xref: /freebsd/sys/netpfil/pf/pf_ioctl.c (revision 644b7b5a2e31fc1d21e7e90764a7b043e5b66d1a)
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_rollup_protected(&rule->evaluations,
1412 			    pf_counter_u64_fetch(&old_rule->evaluations));
1413 			pf_counter_u64_rollup_protected(&rule->packets[0],
1414 			    pf_counter_u64_fetch(&old_rule->packets[0]));
1415 			pf_counter_u64_rollup_protected(&rule->packets[1],
1416 			    pf_counter_u64_fetch(&old_rule->packets[1]));
1417 			pf_counter_u64_rollup_protected(&rule->bytes[0],
1418 			    pf_counter_u64_fetch(&old_rule->bytes[0]));
1419 			pf_counter_u64_rollup_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 int
2447 pf_ioctl_set_timeout(int timeout, int seconds, int *prev_seconds)
2448 {
2449 	uint32_t old;
2450 
2451 	if (timeout < 0 || timeout >= PFTM_MAX ||
2452 	    seconds < 0)
2453 		return (EINVAL);
2454 
2455 	PF_RULES_WLOCK();
2456 	old = V_pf_default_rule.timeout[timeout];
2457 	if (timeout == PFTM_INTERVAL && seconds == 0)
2458 		seconds = 1;
2459 	V_pf_default_rule.timeout[timeout] = seconds;
2460 	if (timeout == PFTM_INTERVAL && seconds < old)
2461 		wakeup(pf_purge_thread);
2462 
2463 	if (prev_seconds != NULL)
2464 		*prev_seconds = old;
2465 
2466 	PF_RULES_WUNLOCK();
2467 
2468 	return (0);
2469 }
2470 
2471 int
2472 pf_ioctl_get_timeout(int timeout, int *seconds)
2473 {
2474 	PF_RULES_RLOCK_TRACKER;
2475 
2476 	if (timeout < 0 || timeout >= PFTM_MAX)
2477 		return (EINVAL);
2478 
2479 	PF_RULES_RLOCK();
2480 	*seconds = V_pf_default_rule.timeout[timeout];
2481 	PF_RULES_RUNLOCK();
2482 
2483 	return (0);
2484 }
2485 
2486 int
2487 pf_ioctl_set_limit(int index, unsigned int limit, unsigned int *old_limit)
2488 {
2489 
2490 	PF_RULES_WLOCK();
2491 	if (index < 0 || index >= PF_LIMIT_MAX ||
2492 	    V_pf_limits[index].zone == NULL) {
2493 		PF_RULES_WUNLOCK();
2494 		return (EINVAL);
2495 	}
2496 	uma_zone_set_max(V_pf_limits[index].zone, limit);
2497 	if (old_limit != NULL)
2498 		*old_limit = V_pf_limits[index].limit;
2499 	V_pf_limits[index].limit = limit;
2500 	PF_RULES_WUNLOCK();
2501 
2502 	return (0);
2503 }
2504 
2505 int
2506 pf_ioctl_get_limit(int index, unsigned int *limit)
2507 {
2508 	PF_RULES_RLOCK_TRACKER;
2509 
2510 	if (index < 0 || index >= PF_LIMIT_MAX)
2511 		return (EINVAL);
2512 
2513 	PF_RULES_RLOCK();
2514 	*limit = V_pf_limits[index].limit;
2515 	PF_RULES_RUNLOCK();
2516 
2517 	return (0);
2518 }
2519 
2520 int
2521 pf_ioctl_begin_addrs(uint32_t *ticket)
2522 {
2523 	PF_RULES_WLOCK();
2524 	pf_empty_kpool(&V_pf_pabuf);
2525 	*ticket = ++V_ticket_pabuf;
2526 	PF_RULES_WUNLOCK();
2527 
2528 	return (0);
2529 }
2530 
2531 int
2532 pf_ioctl_add_addr(struct pfioc_pooladdr *pp)
2533 {
2534 	struct pf_kpooladdr	*pa = NULL;
2535 	struct pfi_kkif		*kif = NULL;
2536 	int error;
2537 
2538 #ifndef INET
2539 	if (pp->af == AF_INET)
2540 		return (EAFNOSUPPORT);
2541 #endif /* INET */
2542 #ifndef INET6
2543 	if (pp->af == AF_INET6)
2544 		return (EAFNOSUPPORT);
2545 #endif /* INET6 */
2546 
2547 	if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2548 	    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2549 	    pp->addr.addr.type != PF_ADDR_TABLE)
2550 		return (EINVAL);
2551 
2552 	if (pp->addr.addr.p.dyn != NULL)
2553 		return (EINVAL);
2554 
2555 	pa = malloc(sizeof(*pa), M_PFRULE, M_WAITOK);
2556 	error = pf_pooladdr_to_kpooladdr(&pp->addr, pa);
2557 	if (error != 0)
2558 		goto out;
2559 	if (pa->ifname[0])
2560 		kif = pf_kkif_create(M_WAITOK);
2561 	PF_RULES_WLOCK();
2562 	if (pp->ticket != V_ticket_pabuf) {
2563 		PF_RULES_WUNLOCK();
2564 		if (pa->ifname[0])
2565 			pf_kkif_free(kif);
2566 		error = EBUSY;
2567 		goto out;
2568 	}
2569 	if (pa->ifname[0]) {
2570 		pa->kif = pfi_kkif_attach(kif, pa->ifname);
2571 		kif = NULL;
2572 		pfi_kkif_ref(pa->kif);
2573 	} else
2574 		pa->kif = NULL;
2575 	if (pa->addr.type == PF_ADDR_DYNIFTL && ((error =
2576 	    pfi_dynaddr_setup(&pa->addr, pp->af)) != 0)) {
2577 		if (pa->ifname[0])
2578 			pfi_kkif_unref(pa->kif);
2579 		PF_RULES_WUNLOCK();
2580 		goto out;
2581 	}
2582 	TAILQ_INSERT_TAIL(&V_pf_pabuf, pa, entries);
2583 	PF_RULES_WUNLOCK();
2584 
2585 	return (0);
2586 
2587 out:
2588 	free(pa, M_PFRULE);
2589 	return (error);
2590 }
2591 
2592 int
2593 pf_ioctl_get_addrs(struct pfioc_pooladdr *pp)
2594 {
2595 	struct pf_kpool		*pool;
2596 	struct pf_kpooladdr	*pa;
2597 
2598 	PF_RULES_RLOCK_TRACKER;
2599 
2600 	pp->anchor[sizeof(pp->anchor) - 1] = 0;
2601 	pp->nr = 0;
2602 
2603 	PF_RULES_RLOCK();
2604 	pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
2605 	    pp->r_num, 0, 1, 0);
2606 	if (pool == NULL) {
2607 		PF_RULES_RUNLOCK();
2608 		return (EBUSY);
2609 	}
2610 	TAILQ_FOREACH(pa, &pool->list, entries)
2611 		pp->nr++;
2612 	PF_RULES_RUNLOCK();
2613 
2614 	return (0);
2615 }
2616 
2617 static int
2618 pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
2619 {
2620 	int			 error = 0;
2621 	PF_RULES_RLOCK_TRACKER;
2622 
2623 #define	ERROUT_IOCTL(target, x)					\
2624     do {								\
2625 	    error = (x);						\
2626 	    SDT_PROBE3(pf, ioctl, ioctl, error, cmd, error, __LINE__);	\
2627 	    goto target;						\
2628     } while (0)
2629 
2630 
2631 	/* XXX keep in sync with switch() below */
2632 	if (securelevel_gt(td->td_ucred, 2))
2633 		switch (cmd) {
2634 		case DIOCGETRULES:
2635 		case DIOCGETRULENV:
2636 		case DIOCGETADDRS:
2637 		case DIOCGETADDR:
2638 		case DIOCGETSTATE:
2639 		case DIOCGETSTATENV:
2640 		case DIOCSETSTATUSIF:
2641 		case DIOCGETSTATUSNV:
2642 		case DIOCCLRSTATUS:
2643 		case DIOCNATLOOK:
2644 		case DIOCSETDEBUG:
2645 #ifdef COMPAT_FREEBSD14
2646 		case DIOCGETSTATES:
2647 		case DIOCGETSTATESV2:
2648 #endif
2649 		case DIOCGETTIMEOUT:
2650 		case DIOCCLRRULECTRS:
2651 		case DIOCGETLIMIT:
2652 		case DIOCGETALTQSV0:
2653 		case DIOCGETALTQSV1:
2654 		case DIOCGETALTQV0:
2655 		case DIOCGETALTQV1:
2656 		case DIOCGETQSTATSV0:
2657 		case DIOCGETQSTATSV1:
2658 		case DIOCGETRULESETS:
2659 		case DIOCGETRULESET:
2660 		case DIOCRGETTABLES:
2661 		case DIOCRGETTSTATS:
2662 		case DIOCRCLRTSTATS:
2663 		case DIOCRCLRADDRS:
2664 		case DIOCRADDADDRS:
2665 		case DIOCRDELADDRS:
2666 		case DIOCRSETADDRS:
2667 		case DIOCRGETADDRS:
2668 		case DIOCRGETASTATS:
2669 		case DIOCRCLRASTATS:
2670 		case DIOCRTSTADDRS:
2671 		case DIOCOSFPGET:
2672 		case DIOCGETSRCNODES:
2673 		case DIOCCLRSRCNODES:
2674 		case DIOCGETSYNCOOKIES:
2675 		case DIOCIGETIFACES:
2676 		case DIOCGIFSPEEDV0:
2677 		case DIOCGIFSPEEDV1:
2678 		case DIOCSETIFFLAG:
2679 		case DIOCCLRIFFLAG:
2680 		case DIOCGETETHRULES:
2681 		case DIOCGETETHRULE:
2682 		case DIOCGETETHRULESETS:
2683 		case DIOCGETETHRULESET:
2684 			break;
2685 		case DIOCRCLRTABLES:
2686 		case DIOCRADDTABLES:
2687 		case DIOCRDELTABLES:
2688 		case DIOCRSETTFLAGS:
2689 			if (((struct pfioc_table *)addr)->pfrio_flags &
2690 			    PFR_FLAG_DUMMY)
2691 				break; /* dummy operation ok */
2692 			return (EPERM);
2693 		default:
2694 			return (EPERM);
2695 		}
2696 
2697 	if (!(flags & FWRITE))
2698 		switch (cmd) {
2699 		case DIOCGETRULES:
2700 		case DIOCGETADDRS:
2701 		case DIOCGETADDR:
2702 		case DIOCGETSTATE:
2703 		case DIOCGETSTATENV:
2704 		case DIOCGETSTATUSNV:
2705 #ifdef COMPAT_FREEBSD14
2706 		case DIOCGETSTATES:
2707 		case DIOCGETSTATESV2:
2708 #endif
2709 		case DIOCGETTIMEOUT:
2710 		case DIOCGETLIMIT:
2711 		case DIOCGETALTQSV0:
2712 		case DIOCGETALTQSV1:
2713 		case DIOCGETALTQV0:
2714 		case DIOCGETALTQV1:
2715 		case DIOCGETQSTATSV0:
2716 		case DIOCGETQSTATSV1:
2717 		case DIOCGETRULESETS:
2718 		case DIOCGETRULESET:
2719 		case DIOCNATLOOK:
2720 		case DIOCRGETTABLES:
2721 		case DIOCRGETTSTATS:
2722 		case DIOCRGETADDRS:
2723 		case DIOCRGETASTATS:
2724 		case DIOCRTSTADDRS:
2725 		case DIOCOSFPGET:
2726 		case DIOCGETSRCNODES:
2727 		case DIOCGETSYNCOOKIES:
2728 		case DIOCIGETIFACES:
2729 		case DIOCGIFSPEEDV1:
2730 		case DIOCGIFSPEEDV0:
2731 		case DIOCGETRULENV:
2732 		case DIOCGETETHRULES:
2733 		case DIOCGETETHRULE:
2734 		case DIOCGETETHRULESETS:
2735 		case DIOCGETETHRULESET:
2736 			break;
2737 		case DIOCRCLRTABLES:
2738 		case DIOCRADDTABLES:
2739 		case DIOCRDELTABLES:
2740 		case DIOCRCLRTSTATS:
2741 		case DIOCRCLRADDRS:
2742 		case DIOCRADDADDRS:
2743 		case DIOCRDELADDRS:
2744 		case DIOCRSETADDRS:
2745 		case DIOCRSETTFLAGS:
2746 			if (((struct pfioc_table *)addr)->pfrio_flags &
2747 			    PFR_FLAG_DUMMY) {
2748 				flags |= FWRITE; /* need write lock for dummy */
2749 				break; /* dummy operation ok */
2750 			}
2751 			return (EACCES);
2752 		default:
2753 			return (EACCES);
2754 		}
2755 
2756 	CURVNET_SET(TD_TO_VNET(td));
2757 
2758 	switch (cmd) {
2759 #ifdef COMPAT_FREEBSD14
2760 	case DIOCSTART:
2761 		error = pf_start();
2762 		break;
2763 
2764 	case DIOCSTOP:
2765 		error = pf_stop();
2766 		break;
2767 #endif
2768 
2769 	case DIOCGETETHRULES: {
2770 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2771 		nvlist_t		*nvl;
2772 		void			*packed;
2773 		struct pf_keth_rule	*tail;
2774 		struct pf_keth_ruleset	*rs;
2775 		u_int32_t		 ticket, nr;
2776 		const char		*anchor = "";
2777 
2778 		nvl = NULL;
2779 		packed = NULL;
2780 
2781 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULES_error, x)
2782 
2783 		if (nv->len > pf_ioctl_maxcount)
2784 			ERROUT(ENOMEM);
2785 
2786 		/* Copy the request in */
2787 		packed = malloc(nv->len, M_NVLIST, M_WAITOK);
2788 		if (packed == NULL)
2789 			ERROUT(ENOMEM);
2790 
2791 		error = copyin(nv->data, packed, nv->len);
2792 		if (error)
2793 			ERROUT(error);
2794 
2795 		nvl = nvlist_unpack(packed, nv->len, 0);
2796 		if (nvl == NULL)
2797 			ERROUT(EBADMSG);
2798 
2799 		if (! nvlist_exists_string(nvl, "anchor"))
2800 			ERROUT(EBADMSG);
2801 
2802 		anchor = nvlist_get_string(nvl, "anchor");
2803 
2804 		rs = pf_find_keth_ruleset(anchor);
2805 
2806 		nvlist_destroy(nvl);
2807 		nvl = NULL;
2808 		free(packed, M_NVLIST);
2809 		packed = NULL;
2810 
2811 		if (rs == NULL)
2812 			ERROUT(ENOENT);
2813 
2814 		/* Reply */
2815 		nvl = nvlist_create(0);
2816 		if (nvl == NULL)
2817 			ERROUT(ENOMEM);
2818 
2819 		PF_RULES_RLOCK();
2820 
2821 		ticket = rs->active.ticket;
2822 		tail = TAILQ_LAST(rs->active.rules, pf_keth_ruleq);
2823 		if (tail)
2824 			nr = tail->nr + 1;
2825 		else
2826 			nr = 0;
2827 
2828 		PF_RULES_RUNLOCK();
2829 
2830 		nvlist_add_number(nvl, "ticket", ticket);
2831 		nvlist_add_number(nvl, "nr", nr);
2832 
2833 		packed = nvlist_pack(nvl, &nv->len);
2834 		if (packed == NULL)
2835 			ERROUT(ENOMEM);
2836 
2837 		if (nv->size == 0)
2838 			ERROUT(0);
2839 		else if (nv->size < nv->len)
2840 			ERROUT(ENOSPC);
2841 
2842 		error = copyout(packed, nv->data, nv->len);
2843 
2844 #undef ERROUT
2845 DIOCGETETHRULES_error:
2846 		free(packed, M_NVLIST);
2847 		nvlist_destroy(nvl);
2848 		break;
2849 	}
2850 
2851 	case DIOCGETETHRULE: {
2852 		struct epoch_tracker	 et;
2853 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2854 		nvlist_t		*nvl = NULL;
2855 		void			*nvlpacked = NULL;
2856 		struct pf_keth_rule	*rule = NULL;
2857 		struct pf_keth_ruleset	*rs;
2858 		u_int32_t		 ticket, nr;
2859 		bool			 clear = false;
2860 		const char		*anchor;
2861 
2862 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULE_error, x)
2863 
2864 		if (nv->len > pf_ioctl_maxcount)
2865 			ERROUT(ENOMEM);
2866 
2867 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2868 		if (nvlpacked == NULL)
2869 			ERROUT(ENOMEM);
2870 
2871 		error = copyin(nv->data, nvlpacked, nv->len);
2872 		if (error)
2873 			ERROUT(error);
2874 
2875 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2876 		if (nvl == NULL)
2877 			ERROUT(EBADMSG);
2878 		if (! nvlist_exists_number(nvl, "ticket"))
2879 			ERROUT(EBADMSG);
2880 		ticket = nvlist_get_number(nvl, "ticket");
2881 		if (! nvlist_exists_string(nvl, "anchor"))
2882 			ERROUT(EBADMSG);
2883 		anchor = nvlist_get_string(nvl, "anchor");
2884 
2885 		if (nvlist_exists_bool(nvl, "clear"))
2886 			clear = nvlist_get_bool(nvl, "clear");
2887 
2888 		if (clear && !(flags & FWRITE))
2889 			ERROUT(EACCES);
2890 
2891 		if (! nvlist_exists_number(nvl, "nr"))
2892 			ERROUT(EBADMSG);
2893 		nr = nvlist_get_number(nvl, "nr");
2894 
2895 		PF_RULES_RLOCK();
2896 		rs = pf_find_keth_ruleset(anchor);
2897 		if (rs == NULL) {
2898 			PF_RULES_RUNLOCK();
2899 			ERROUT(ENOENT);
2900 		}
2901 		if (ticket != rs->active.ticket) {
2902 			PF_RULES_RUNLOCK();
2903 			ERROUT(EBUSY);
2904 		}
2905 
2906 		nvlist_destroy(nvl);
2907 		nvl = NULL;
2908 		free(nvlpacked, M_NVLIST);
2909 		nvlpacked = NULL;
2910 
2911 		rule = TAILQ_FIRST(rs->active.rules);
2912 		while ((rule != NULL) && (rule->nr != nr))
2913 			rule = TAILQ_NEXT(rule, entries);
2914 		if (rule == NULL) {
2915 			PF_RULES_RUNLOCK();
2916 			ERROUT(ENOENT);
2917 		}
2918 		/* Make sure rule can't go away. */
2919 		NET_EPOCH_ENTER(et);
2920 		PF_RULES_RUNLOCK();
2921 		nvl = pf_keth_rule_to_nveth_rule(rule);
2922 		if (pf_keth_anchor_nvcopyout(rs, rule, nvl))
2923 			ERROUT(EBUSY);
2924 		NET_EPOCH_EXIT(et);
2925 		if (nvl == NULL)
2926 			ERROUT(ENOMEM);
2927 
2928 		nvlpacked = nvlist_pack(nvl, &nv->len);
2929 		if (nvlpacked == NULL)
2930 			ERROUT(ENOMEM);
2931 
2932 		if (nv->size == 0)
2933 			ERROUT(0);
2934 		else if (nv->size < nv->len)
2935 			ERROUT(ENOSPC);
2936 
2937 		error = copyout(nvlpacked, nv->data, nv->len);
2938 		if (error == 0 && clear) {
2939 			counter_u64_zero(rule->evaluations);
2940 			for (int i = 0; i < 2; i++) {
2941 				counter_u64_zero(rule->packets[i]);
2942 				counter_u64_zero(rule->bytes[i]);
2943 			}
2944 		}
2945 
2946 #undef ERROUT
2947 DIOCGETETHRULE_error:
2948 		free(nvlpacked, M_NVLIST);
2949 		nvlist_destroy(nvl);
2950 		break;
2951 	}
2952 
2953 	case DIOCADDETHRULE: {
2954 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
2955 		nvlist_t		*nvl = NULL;
2956 		void			*nvlpacked = NULL;
2957 		struct pf_keth_rule	*rule = NULL, *tail = NULL;
2958 		struct pf_keth_ruleset	*ruleset = NULL;
2959 		struct pfi_kkif		*kif = NULL, *bridge_to_kif = NULL;
2960 		const char		*anchor = "", *anchor_call = "";
2961 
2962 #define ERROUT(x)	ERROUT_IOCTL(DIOCADDETHRULE_error, x)
2963 
2964 		if (nv->len > pf_ioctl_maxcount)
2965 			ERROUT(ENOMEM);
2966 
2967 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
2968 		if (nvlpacked == NULL)
2969 			ERROUT(ENOMEM);
2970 
2971 		error = copyin(nv->data, nvlpacked, nv->len);
2972 		if (error)
2973 			ERROUT(error);
2974 
2975 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
2976 		if (nvl == NULL)
2977 			ERROUT(EBADMSG);
2978 
2979 		if (! nvlist_exists_number(nvl, "ticket"))
2980 			ERROUT(EBADMSG);
2981 
2982 		if (nvlist_exists_string(nvl, "anchor"))
2983 			anchor = nvlist_get_string(nvl, "anchor");
2984 		if (nvlist_exists_string(nvl, "anchor_call"))
2985 			anchor_call = nvlist_get_string(nvl, "anchor_call");
2986 
2987 		ruleset = pf_find_keth_ruleset(anchor);
2988 		if (ruleset == NULL)
2989 			ERROUT(EINVAL);
2990 
2991 		if (nvlist_get_number(nvl, "ticket") !=
2992 		    ruleset->inactive.ticket) {
2993 			DPFPRINTF(PF_DEBUG_MISC,
2994 			    ("ticket: %d != %d\n",
2995 			    (u_int32_t)nvlist_get_number(nvl, "ticket"),
2996 			    ruleset->inactive.ticket));
2997 			ERROUT(EBUSY);
2998 		}
2999 
3000 		rule = malloc(sizeof(*rule), M_PFRULE, M_WAITOK);
3001 		if (rule == NULL)
3002 			ERROUT(ENOMEM);
3003 		rule->timestamp = NULL;
3004 
3005 		error = pf_nveth_rule_to_keth_rule(nvl, rule);
3006 		if (error != 0)
3007 			ERROUT(error);
3008 
3009 		if (rule->ifname[0])
3010 			kif = pf_kkif_create(M_WAITOK);
3011 		if (rule->bridge_to_name[0])
3012 			bridge_to_kif = pf_kkif_create(M_WAITOK);
3013 		rule->evaluations = counter_u64_alloc(M_WAITOK);
3014 		for (int i = 0; i < 2; i++) {
3015 			rule->packets[i] = counter_u64_alloc(M_WAITOK);
3016 			rule->bytes[i] = counter_u64_alloc(M_WAITOK);
3017 		}
3018 		rule->timestamp = uma_zalloc_pcpu(pf_timestamp_pcpu_zone,
3019 		    M_WAITOK | M_ZERO);
3020 
3021 		PF_RULES_WLOCK();
3022 
3023 		if (rule->ifname[0]) {
3024 			rule->kif = pfi_kkif_attach(kif, rule->ifname);
3025 			pfi_kkif_ref(rule->kif);
3026 		} else
3027 			rule->kif = NULL;
3028 		if (rule->bridge_to_name[0]) {
3029 			rule->bridge_to = pfi_kkif_attach(bridge_to_kif,
3030 			    rule->bridge_to_name);
3031 			pfi_kkif_ref(rule->bridge_to);
3032 		} else
3033 			rule->bridge_to = NULL;
3034 
3035 #ifdef ALTQ
3036 		/* set queue IDs */
3037 		if (rule->qname[0] != 0) {
3038 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
3039 				error = EBUSY;
3040 			else
3041 				rule->qid = rule->qid;
3042 		}
3043 #endif
3044 		if (rule->tagname[0])
3045 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
3046 				error = EBUSY;
3047 		if (rule->match_tagname[0])
3048 			if ((rule->match_tag = pf_tagname2tag(
3049 			    rule->match_tagname)) == 0)
3050 				error = EBUSY;
3051 
3052 		if (error == 0 && rule->ipdst.addr.type == PF_ADDR_TABLE)
3053 			error = pf_eth_addr_setup(ruleset, &rule->ipdst.addr);
3054 		if (error == 0 && rule->ipsrc.addr.type == PF_ADDR_TABLE)
3055 			error = pf_eth_addr_setup(ruleset, &rule->ipsrc.addr);
3056 
3057 		if (error) {
3058 			pf_free_eth_rule(rule);
3059 			PF_RULES_WUNLOCK();
3060 			ERROUT(error);
3061 		}
3062 
3063 		if (pf_keth_anchor_setup(rule, ruleset, anchor_call)) {
3064 			pf_free_eth_rule(rule);
3065 			PF_RULES_WUNLOCK();
3066 			ERROUT(EINVAL);
3067 		}
3068 
3069 		tail = TAILQ_LAST(ruleset->inactive.rules, pf_keth_ruleq);
3070 		if (tail)
3071 			rule->nr = tail->nr + 1;
3072 		else
3073 			rule->nr = 0;
3074 
3075 		TAILQ_INSERT_TAIL(ruleset->inactive.rules, rule, entries);
3076 
3077 		PF_RULES_WUNLOCK();
3078 
3079 #undef ERROUT
3080 DIOCADDETHRULE_error:
3081 		nvlist_destroy(nvl);
3082 		free(nvlpacked, M_NVLIST);
3083 		break;
3084 	}
3085 
3086 	case DIOCGETETHRULESETS: {
3087 		struct epoch_tracker	 et;
3088 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3089 		nvlist_t		*nvl = NULL;
3090 		void			*nvlpacked = NULL;
3091 		struct pf_keth_ruleset	*ruleset;
3092 		struct pf_keth_anchor	*anchor;
3093 		int			 nr = 0;
3094 
3095 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESETS_error, x)
3096 
3097 		if (nv->len > pf_ioctl_maxcount)
3098 			ERROUT(ENOMEM);
3099 
3100 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3101 		if (nvlpacked == NULL)
3102 			ERROUT(ENOMEM);
3103 
3104 		error = copyin(nv->data, nvlpacked, nv->len);
3105 		if (error)
3106 			ERROUT(error);
3107 
3108 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3109 		if (nvl == NULL)
3110 			ERROUT(EBADMSG);
3111 		if (! nvlist_exists_string(nvl, "path"))
3112 			ERROUT(EBADMSG);
3113 
3114 		NET_EPOCH_ENTER(et);
3115 
3116 		if ((ruleset = pf_find_keth_ruleset(
3117 		    nvlist_get_string(nvl, "path"))) == NULL) {
3118 			NET_EPOCH_EXIT(et);
3119 			ERROUT(ENOENT);
3120 		}
3121 
3122 		if (ruleset->anchor == NULL) {
3123 			RB_FOREACH(anchor, pf_keth_anchor_global, &V_pf_keth_anchors)
3124 				if (anchor->parent == NULL)
3125 					nr++;
3126 		} else {
3127 			RB_FOREACH(anchor, pf_keth_anchor_node,
3128 			    &ruleset->anchor->children)
3129 				nr++;
3130 		}
3131 
3132 		NET_EPOCH_EXIT(et);
3133 
3134 		nvlist_destroy(nvl);
3135 		nvl = NULL;
3136 		free(nvlpacked, M_NVLIST);
3137 		nvlpacked = NULL;
3138 
3139 		nvl = nvlist_create(0);
3140 		if (nvl == NULL)
3141 			ERROUT(ENOMEM);
3142 
3143 		nvlist_add_number(nvl, "nr", nr);
3144 
3145 		nvlpacked = nvlist_pack(nvl, &nv->len);
3146 		if (nvlpacked == NULL)
3147 			ERROUT(ENOMEM);
3148 
3149 		if (nv->size == 0)
3150 			ERROUT(0);
3151 		else if (nv->size < nv->len)
3152 			ERROUT(ENOSPC);
3153 
3154 		error = copyout(nvlpacked, nv->data, nv->len);
3155 
3156 #undef ERROUT
3157 DIOCGETETHRULESETS_error:
3158 		free(nvlpacked, M_NVLIST);
3159 		nvlist_destroy(nvl);
3160 		break;
3161 	}
3162 
3163 	case DIOCGETETHRULESET: {
3164 		struct epoch_tracker	 et;
3165 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3166 		nvlist_t		*nvl = NULL;
3167 		void			*nvlpacked = NULL;
3168 		struct pf_keth_ruleset	*ruleset;
3169 		struct pf_keth_anchor	*anchor;
3170 		int			 nr = 0, req_nr = 0;
3171 		bool			 found = false;
3172 
3173 #define ERROUT(x)	ERROUT_IOCTL(DIOCGETETHRULESET_error, x)
3174 
3175 		if (nv->len > pf_ioctl_maxcount)
3176 			ERROUT(ENOMEM);
3177 
3178 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3179 		if (nvlpacked == NULL)
3180 			ERROUT(ENOMEM);
3181 
3182 		error = copyin(nv->data, nvlpacked, nv->len);
3183 		if (error)
3184 			ERROUT(error);
3185 
3186 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3187 		if (nvl == NULL)
3188 			ERROUT(EBADMSG);
3189 		if (! nvlist_exists_string(nvl, "path"))
3190 			ERROUT(EBADMSG);
3191 		if (! nvlist_exists_number(nvl, "nr"))
3192 			ERROUT(EBADMSG);
3193 
3194 		req_nr = nvlist_get_number(nvl, "nr");
3195 
3196 		NET_EPOCH_ENTER(et);
3197 
3198 		if ((ruleset = pf_find_keth_ruleset(
3199 		    nvlist_get_string(nvl, "path"))) == NULL) {
3200 			NET_EPOCH_EXIT(et);
3201 			ERROUT(ENOENT);
3202 		}
3203 
3204 		nvlist_destroy(nvl);
3205 		nvl = NULL;
3206 		free(nvlpacked, M_NVLIST);
3207 		nvlpacked = NULL;
3208 
3209 		nvl = nvlist_create(0);
3210 		if (nvl == NULL) {
3211 			NET_EPOCH_EXIT(et);
3212 			ERROUT(ENOMEM);
3213 		}
3214 
3215 		if (ruleset->anchor == NULL) {
3216 			RB_FOREACH(anchor, pf_keth_anchor_global,
3217 			    &V_pf_keth_anchors) {
3218 				if (anchor->parent == NULL && nr++ == req_nr) {
3219 					found = true;
3220 					break;
3221 				}
3222 			}
3223 		} else {
3224 			RB_FOREACH(anchor, pf_keth_anchor_node,
3225 			     &ruleset->anchor->children) {
3226 				if (nr++ == req_nr) {
3227 					found = true;
3228 					break;
3229 				}
3230 			}
3231 		}
3232 
3233 		NET_EPOCH_EXIT(et);
3234 		if (found) {
3235 			nvlist_add_number(nvl, "nr", nr);
3236 			nvlist_add_string(nvl, "name", anchor->name);
3237 			if (ruleset->anchor)
3238 				nvlist_add_string(nvl, "path",
3239 				    ruleset->anchor->path);
3240 			else
3241 				nvlist_add_string(nvl, "path", "");
3242 		} else {
3243 			ERROUT(EBUSY);
3244 		}
3245 
3246 		nvlpacked = nvlist_pack(nvl, &nv->len);
3247 		if (nvlpacked == NULL)
3248 			ERROUT(ENOMEM);
3249 
3250 		if (nv->size == 0)
3251 			ERROUT(0);
3252 		else if (nv->size < nv->len)
3253 			ERROUT(ENOSPC);
3254 
3255 		error = copyout(nvlpacked, nv->data, nv->len);
3256 
3257 #undef ERROUT
3258 DIOCGETETHRULESET_error:
3259 		free(nvlpacked, M_NVLIST);
3260 		nvlist_destroy(nvl);
3261 		break;
3262 	}
3263 
3264 	case DIOCADDRULENV: {
3265 		struct pfioc_nv	*nv = (struct pfioc_nv *)addr;
3266 		nvlist_t	*nvl = NULL;
3267 		void		*nvlpacked = NULL;
3268 		struct pf_krule	*rule = NULL;
3269 		const char	*anchor = "", *anchor_call = "";
3270 		uint32_t	 ticket = 0, pool_ticket = 0;
3271 
3272 #define	ERROUT(x)	ERROUT_IOCTL(DIOCADDRULENV_error, x)
3273 
3274 		if (nv->len > pf_ioctl_maxcount)
3275 			ERROUT(ENOMEM);
3276 
3277 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3278 		error = copyin(nv->data, nvlpacked, nv->len);
3279 		if (error)
3280 			ERROUT(error);
3281 
3282 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3283 		if (nvl == NULL)
3284 			ERROUT(EBADMSG);
3285 
3286 		if (! nvlist_exists_number(nvl, "ticket"))
3287 			ERROUT(EINVAL);
3288 		ticket = nvlist_get_number(nvl, "ticket");
3289 
3290 		if (! nvlist_exists_number(nvl, "pool_ticket"))
3291 			ERROUT(EINVAL);
3292 		pool_ticket = nvlist_get_number(nvl, "pool_ticket");
3293 
3294 		if (! nvlist_exists_nvlist(nvl, "rule"))
3295 			ERROUT(EINVAL);
3296 
3297 		rule = pf_krule_alloc();
3298 		error = pf_nvrule_to_krule(nvlist_get_nvlist(nvl, "rule"),
3299 		    rule);
3300 		if (error)
3301 			ERROUT(error);
3302 
3303 		if (nvlist_exists_string(nvl, "anchor"))
3304 			anchor = nvlist_get_string(nvl, "anchor");
3305 		if (nvlist_exists_string(nvl, "anchor_call"))
3306 			anchor_call = nvlist_get_string(nvl, "anchor_call");
3307 
3308 		if ((error = nvlist_error(nvl)))
3309 			ERROUT(error);
3310 
3311 		/* Frees rule on error */
3312 		error = pf_ioctl_addrule(rule, ticket, pool_ticket, anchor,
3313 		    anchor_call, td->td_ucred->cr_ruid,
3314 		    td->td_proc ? td->td_proc->p_pid : 0);
3315 
3316 		nvlist_destroy(nvl);
3317 		free(nvlpacked, M_NVLIST);
3318 		break;
3319 #undef ERROUT
3320 DIOCADDRULENV_error:
3321 		pf_krule_free(rule);
3322 		nvlist_destroy(nvl);
3323 		free(nvlpacked, M_NVLIST);
3324 
3325 		break;
3326 	}
3327 	case DIOCADDRULE: {
3328 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3329 		struct pf_krule		*rule;
3330 
3331 		rule = pf_krule_alloc();
3332 		error = pf_rule_to_krule(&pr->rule, rule);
3333 		if (error != 0) {
3334 			pf_krule_free(rule);
3335 			break;
3336 		}
3337 
3338 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3339 
3340 		/* Frees rule on error */
3341 		error = pf_ioctl_addrule(rule, pr->ticket, pr->pool_ticket,
3342 		    pr->anchor, pr->anchor_call, td->td_ucred->cr_ruid,
3343 		    td->td_proc ? td->td_proc->p_pid : 0);
3344 		break;
3345 	}
3346 
3347 	case DIOCGETRULES: {
3348 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
3349 
3350 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
3351 
3352 		error = pf_ioctl_getrules(pr);
3353 
3354 		break;
3355 	}
3356 
3357 	case DIOCGETRULENV: {
3358 		struct pfioc_nv		*nv = (struct pfioc_nv *)addr;
3359 		nvlist_t		*nvrule = NULL;
3360 		nvlist_t		*nvl = NULL;
3361 		struct pf_kruleset	*ruleset;
3362 		struct pf_krule		*rule;
3363 		void			*nvlpacked = NULL;
3364 		int			 rs_num, nr;
3365 		bool			 clear_counter = false;
3366 
3367 #define	ERROUT(x)	ERROUT_IOCTL(DIOCGETRULENV_error, x)
3368 
3369 		if (nv->len > pf_ioctl_maxcount)
3370 			ERROUT(ENOMEM);
3371 
3372 		/* Copy the request in */
3373 		nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
3374 		if (nvlpacked == NULL)
3375 			ERROUT(ENOMEM);
3376 
3377 		error = copyin(nv->data, nvlpacked, nv->len);
3378 		if (error)
3379 			ERROUT(error);
3380 
3381 		nvl = nvlist_unpack(nvlpacked, nv->len, 0);
3382 		if (nvl == NULL)
3383 			ERROUT(EBADMSG);
3384 
3385 		if (! nvlist_exists_string(nvl, "anchor"))
3386 			ERROUT(EBADMSG);
3387 		if (! nvlist_exists_number(nvl, "ruleset"))
3388 			ERROUT(EBADMSG);
3389 		if (! nvlist_exists_number(nvl, "ticket"))
3390 			ERROUT(EBADMSG);
3391 		if (! nvlist_exists_number(nvl, "nr"))
3392 			ERROUT(EBADMSG);
3393 
3394 		if (nvlist_exists_bool(nvl, "clear_counter"))
3395 			clear_counter = nvlist_get_bool(nvl, "clear_counter");
3396 
3397 		if (clear_counter && !(flags & FWRITE))
3398 			ERROUT(EACCES);
3399 
3400 		nr = nvlist_get_number(nvl, "nr");
3401 
3402 		PF_RULES_WLOCK();
3403 		ruleset = pf_find_kruleset(nvlist_get_string(nvl, "anchor"));
3404 		if (ruleset == NULL) {
3405 			PF_RULES_WUNLOCK();
3406 			ERROUT(ENOENT);
3407 		}
3408 
3409 		rs_num = pf_get_ruleset_number(nvlist_get_number(nvl, "ruleset"));
3410 		if (rs_num >= PF_RULESET_MAX) {
3411 			PF_RULES_WUNLOCK();
3412 			ERROUT(EINVAL);
3413 		}
3414 
3415 		if (nvlist_get_number(nvl, "ticket") !=
3416 		    ruleset->rules[rs_num].active.ticket) {
3417 			PF_RULES_WUNLOCK();
3418 			ERROUT(EBUSY);
3419 		}
3420 
3421 		if ((error = nvlist_error(nvl))) {
3422 			PF_RULES_WUNLOCK();
3423 			ERROUT(error);
3424 		}
3425 
3426 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
3427 		while ((rule != NULL) && (rule->nr != nr))
3428 			rule = TAILQ_NEXT(rule, entries);
3429 		if (rule == NULL) {
3430 			PF_RULES_WUNLOCK();
3431 			ERROUT(EBUSY);
3432 		}
3433 
3434 		nvrule = pf_krule_to_nvrule(rule);
3435 
3436 		nvlist_destroy(nvl);
3437 		nvl = nvlist_create(0);
3438 		if (nvl == NULL) {
3439 			PF_RULES_WUNLOCK();
3440 			ERROUT(ENOMEM);
3441 		}
3442 		nvlist_add_number(nvl, "nr", nr);
3443 		nvlist_add_nvlist(nvl, "rule", nvrule);
3444 		nvlist_destroy(nvrule);
3445 		nvrule = NULL;
3446 		if (pf_kanchor_nvcopyout(ruleset, rule, nvl)) {
3447 			PF_RULES_WUNLOCK();
3448 			ERROUT(EBUSY);
3449 		}
3450 
3451 		free(nvlpacked, M_NVLIST);
3452 		nvlpacked = nvlist_pack(nvl, &nv->len);
3453 		if (nvlpacked == NULL) {
3454 			PF_RULES_WUNLOCK();
3455 			ERROUT(ENOMEM);
3456 		}
3457 
3458 		if (nv->size == 0) {
3459 			PF_RULES_WUNLOCK();
3460 			ERROUT(0);
3461 		}
3462 		else if (nv->size < nv->len) {
3463 			PF_RULES_WUNLOCK();
3464 			ERROUT(ENOSPC);
3465 		}
3466 
3467 		if (clear_counter)
3468 			pf_krule_clear_counters(rule);
3469 
3470 		PF_RULES_WUNLOCK();
3471 
3472 		error = copyout(nvlpacked, nv->data, nv->len);
3473 
3474 #undef ERROUT
3475 DIOCGETRULENV_error:
3476 		free(nvlpacked, M_NVLIST);
3477 		nvlist_destroy(nvrule);
3478 		nvlist_destroy(nvl);
3479 
3480 		break;
3481 	}
3482 
3483 	case DIOCCHANGERULE: {
3484 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
3485 		struct pf_kruleset	*ruleset;
3486 		struct pf_krule		*oldrule = NULL, *newrule = NULL;
3487 		struct pfi_kkif		*kif = NULL;
3488 		struct pf_kpooladdr	*pa;
3489 		u_int32_t		 nr = 0;
3490 		int			 rs_num;
3491 
3492 		pcr->anchor[sizeof(pcr->anchor) - 1] = 0;
3493 
3494 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
3495 		    pcr->action > PF_CHANGE_GET_TICKET) {
3496 			error = EINVAL;
3497 			break;
3498 		}
3499 		if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
3500 			error = EINVAL;
3501 			break;
3502 		}
3503 
3504 		if (pcr->action != PF_CHANGE_REMOVE) {
3505 			newrule = pf_krule_alloc();
3506 			error = pf_rule_to_krule(&pcr->rule, newrule);
3507 			if (error != 0) {
3508 				pf_krule_free(newrule);
3509 				break;
3510 			}
3511 
3512 			if (newrule->ifname[0])
3513 				kif = pf_kkif_create(M_WAITOK);
3514 			pf_counter_u64_init(&newrule->evaluations, M_WAITOK);
3515 			for (int i = 0; i < 2; i++) {
3516 				pf_counter_u64_init(&newrule->packets[i], M_WAITOK);
3517 				pf_counter_u64_init(&newrule->bytes[i], M_WAITOK);
3518 			}
3519 			newrule->states_cur = counter_u64_alloc(M_WAITOK);
3520 			newrule->states_tot = counter_u64_alloc(M_WAITOK);
3521 			newrule->src_nodes = counter_u64_alloc(M_WAITOK);
3522 			newrule->cuid = td->td_ucred->cr_ruid;
3523 			newrule->cpid = td->td_proc ? td->td_proc->p_pid : 0;
3524 			TAILQ_INIT(&newrule->rpool.list);
3525 		}
3526 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGERULE_error, x)
3527 
3528 		PF_CONFIG_LOCK();
3529 		PF_RULES_WLOCK();
3530 #ifdef PF_WANT_32_TO_64_COUNTER
3531 		if (newrule != NULL) {
3532 			LIST_INSERT_HEAD(&V_pf_allrulelist, newrule, allrulelist);
3533 			newrule->allrulelinked = true;
3534 			V_pf_allrulecount++;
3535 		}
3536 #endif
3537 
3538 		if (!(pcr->action == PF_CHANGE_REMOVE ||
3539 		    pcr->action == PF_CHANGE_GET_TICKET) &&
3540 		    pcr->pool_ticket != V_ticket_pabuf)
3541 			ERROUT(EBUSY);
3542 
3543 		ruleset = pf_find_kruleset(pcr->anchor);
3544 		if (ruleset == NULL)
3545 			ERROUT(EINVAL);
3546 
3547 		rs_num = pf_get_ruleset_number(pcr->rule.action);
3548 		if (rs_num >= PF_RULESET_MAX)
3549 			ERROUT(EINVAL);
3550 
3551 		/*
3552 		 * XXXMJG: there is no guarantee that the ruleset was
3553 		 * created by the usual route of calling DIOCXBEGIN.
3554 		 * As a result it is possible the rule tree will not
3555 		 * be allocated yet. Hack around it by doing it here.
3556 		 * Note it is fine to let the tree persist in case of
3557 		 * error as it will be freed down the road on future
3558 		 * updates (if need be).
3559 		 */
3560 		if (ruleset->rules[rs_num].active.tree == NULL) {
3561 			ruleset->rules[rs_num].active.tree = pf_rule_tree_alloc(M_NOWAIT);
3562 			if (ruleset->rules[rs_num].active.tree == NULL) {
3563 				ERROUT(ENOMEM);
3564 			}
3565 		}
3566 
3567 		if (pcr->action == PF_CHANGE_GET_TICKET) {
3568 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
3569 			ERROUT(0);
3570 		} else if (pcr->ticket !=
3571 			    ruleset->rules[rs_num].active.ticket)
3572 				ERROUT(EINVAL);
3573 
3574 		if (pcr->action != PF_CHANGE_REMOVE) {
3575 			if (newrule->ifname[0]) {
3576 				newrule->kif = pfi_kkif_attach(kif,
3577 				    newrule->ifname);
3578 				kif = NULL;
3579 				pfi_kkif_ref(newrule->kif);
3580 			} else
3581 				newrule->kif = NULL;
3582 
3583 			if (newrule->rtableid > 0 &&
3584 			    newrule->rtableid >= rt_numfibs)
3585 				error = EBUSY;
3586 
3587 #ifdef ALTQ
3588 			/* set queue IDs */
3589 			if (newrule->qname[0] != 0) {
3590 				if ((newrule->qid =
3591 				    pf_qname2qid(newrule->qname)) == 0)
3592 					error = EBUSY;
3593 				else if (newrule->pqname[0] != 0) {
3594 					if ((newrule->pqid =
3595 					    pf_qname2qid(newrule->pqname)) == 0)
3596 						error = EBUSY;
3597 				} else
3598 					newrule->pqid = newrule->qid;
3599 			}
3600 #endif /* ALTQ */
3601 			if (newrule->tagname[0])
3602 				if ((newrule->tag =
3603 				    pf_tagname2tag(newrule->tagname)) == 0)
3604 					error = EBUSY;
3605 			if (newrule->match_tagname[0])
3606 				if ((newrule->match_tag = pf_tagname2tag(
3607 				    newrule->match_tagname)) == 0)
3608 					error = EBUSY;
3609 			if (newrule->rt && !newrule->direction)
3610 				error = EINVAL;
3611 			if (!newrule->log)
3612 				newrule->logif = 0;
3613 			if (newrule->logif >= PFLOGIFS_MAX)
3614 				error = EINVAL;
3615 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
3616 				error = ENOMEM;
3617 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
3618 				error = ENOMEM;
3619 			if (pf_kanchor_setup(newrule, ruleset, pcr->anchor_call))
3620 				error = EINVAL;
3621 			TAILQ_FOREACH(pa, &V_pf_pabuf, entries)
3622 				if (pa->addr.type == PF_ADDR_TABLE) {
3623 					pa->addr.p.tbl =
3624 					    pfr_attach_table(ruleset,
3625 					    pa->addr.v.tblname);
3626 					if (pa->addr.p.tbl == NULL)
3627 						error = ENOMEM;
3628 				}
3629 
3630 			newrule->overload_tbl = NULL;
3631 			if (newrule->overload_tblname[0]) {
3632 				if ((newrule->overload_tbl = pfr_attach_table(
3633 				    ruleset, newrule->overload_tblname)) ==
3634 				    NULL)
3635 					error = EINVAL;
3636 				else
3637 					newrule->overload_tbl->pfrkt_flags |=
3638 					    PFR_TFLAG_ACTIVE;
3639 			}
3640 
3641 			pf_mv_kpool(&V_pf_pabuf, &newrule->rpool.list);
3642 			if (((((newrule->action == PF_NAT) ||
3643 			    (newrule->action == PF_RDR) ||
3644 			    (newrule->action == PF_BINAT) ||
3645 			    (newrule->rt > PF_NOPFROUTE)) &&
3646 			    !newrule->anchor)) &&
3647 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
3648 				error = EINVAL;
3649 
3650 			if (error) {
3651 				pf_free_rule(newrule);
3652 				PF_RULES_WUNLOCK();
3653 				PF_CONFIG_UNLOCK();
3654 				break;
3655 			}
3656 
3657 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
3658 		}
3659 		pf_empty_kpool(&V_pf_pabuf);
3660 
3661 		if (pcr->action == PF_CHANGE_ADD_HEAD)
3662 			oldrule = TAILQ_FIRST(
3663 			    ruleset->rules[rs_num].active.ptr);
3664 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
3665 			oldrule = TAILQ_LAST(
3666 			    ruleset->rules[rs_num].active.ptr, pf_krulequeue);
3667 		else {
3668 			oldrule = TAILQ_FIRST(
3669 			    ruleset->rules[rs_num].active.ptr);
3670 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
3671 				oldrule = TAILQ_NEXT(oldrule, entries);
3672 			if (oldrule == NULL) {
3673 				if (newrule != NULL)
3674 					pf_free_rule(newrule);
3675 				PF_RULES_WUNLOCK();
3676 				PF_CONFIG_UNLOCK();
3677 				error = EINVAL;
3678 				break;
3679 			}
3680 		}
3681 
3682 		if (pcr->action == PF_CHANGE_REMOVE) {
3683 			pf_unlink_rule(ruleset->rules[rs_num].active.ptr,
3684 			    oldrule);
3685 			RB_REMOVE(pf_krule_global,
3686 			    ruleset->rules[rs_num].active.tree, oldrule);
3687 			ruleset->rules[rs_num].active.rcount--;
3688 		} else {
3689 			pf_hash_rule(newrule);
3690 			if (RB_INSERT(pf_krule_global,
3691 			    ruleset->rules[rs_num].active.tree, newrule) != NULL) {
3692 				pf_free_rule(newrule);
3693 				PF_RULES_WUNLOCK();
3694 				PF_CONFIG_UNLOCK();
3695 				error = EEXIST;
3696 				break;
3697 			}
3698 
3699 			if (oldrule == NULL)
3700 				TAILQ_INSERT_TAIL(
3701 				    ruleset->rules[rs_num].active.ptr,
3702 				    newrule, entries);
3703 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
3704 			    pcr->action == PF_CHANGE_ADD_BEFORE)
3705 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
3706 			else
3707 				TAILQ_INSERT_AFTER(
3708 				    ruleset->rules[rs_num].active.ptr,
3709 				    oldrule, newrule, entries);
3710 			ruleset->rules[rs_num].active.rcount++;
3711 		}
3712 
3713 		nr = 0;
3714 		TAILQ_FOREACH(oldrule,
3715 		    ruleset->rules[rs_num].active.ptr, entries)
3716 			oldrule->nr = nr++;
3717 
3718 		ruleset->rules[rs_num].active.ticket++;
3719 
3720 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
3721 		pf_remove_if_empty_kruleset(ruleset);
3722 
3723 		PF_RULES_WUNLOCK();
3724 		PF_CONFIG_UNLOCK();
3725 		break;
3726 
3727 #undef ERROUT
3728 DIOCCHANGERULE_error:
3729 		PF_RULES_WUNLOCK();
3730 		PF_CONFIG_UNLOCK();
3731 		pf_krule_free(newrule);
3732 		pf_kkif_free(kif);
3733 		break;
3734 	}
3735 
3736 	case DIOCCLRSTATESNV: {
3737 		error = pf_clearstates_nv((struct pfioc_nv *)addr);
3738 		break;
3739 	}
3740 
3741 	case DIOCKILLSTATESNV: {
3742 		error = pf_killstates_nv((struct pfioc_nv *)addr);
3743 		break;
3744 	}
3745 
3746 	case DIOCADDSTATE: {
3747 		struct pfioc_state		*ps = (struct pfioc_state *)addr;
3748 		struct pfsync_state_1301	*sp = &ps->state;
3749 
3750 		if (sp->timeout >= PFTM_MAX) {
3751 			error = EINVAL;
3752 			break;
3753 		}
3754 		if (V_pfsync_state_import_ptr != NULL) {
3755 			PF_RULES_RLOCK();
3756 			error = V_pfsync_state_import_ptr(
3757 			    (union pfsync_state_union *)sp, PFSYNC_SI_IOCTL,
3758 			    PFSYNC_MSG_VERSION_1301);
3759 			PF_RULES_RUNLOCK();
3760 		} else
3761 			error = EOPNOTSUPP;
3762 		break;
3763 	}
3764 
3765 	case DIOCGETSTATE: {
3766 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
3767 		struct pf_kstate	*s;
3768 
3769 		s = pf_find_state_byid(ps->state.id, ps->state.creatorid);
3770 		if (s == NULL) {
3771 			error = ENOENT;
3772 			break;
3773 		}
3774 
3775 		pfsync_state_export((union pfsync_state_union*)&ps->state,
3776 		    s, PFSYNC_MSG_VERSION_1301);
3777 		PF_STATE_UNLOCK(s);
3778 		break;
3779 	}
3780 
3781 	case DIOCGETSTATENV: {
3782 		error = pf_getstate((struct pfioc_nv *)addr);
3783 		break;
3784 	}
3785 
3786 #ifdef COMPAT_FREEBSD14
3787 	case DIOCGETSTATES: {
3788 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
3789 		struct pf_kstate	*s;
3790 		struct pfsync_state_1301	*pstore, *p;
3791 		int			 i, nr;
3792 		size_t			 slice_count = 16, count;
3793 		void			*out;
3794 
3795 		if (ps->ps_len <= 0) {
3796 			nr = uma_zone_get_cur(V_pf_state_z);
3797 			ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3798 			break;
3799 		}
3800 
3801 		out = ps->ps_states;
3802 		pstore = mallocarray(slice_count,
3803 		    sizeof(struct pfsync_state_1301), M_TEMP, M_WAITOK | M_ZERO);
3804 		nr = 0;
3805 
3806 		for (i = 0; i <= V_pf_hashmask; i++) {
3807 			struct pf_idhash *ih = &V_pf_idhash[i];
3808 
3809 DIOCGETSTATES_retry:
3810 			p = pstore;
3811 
3812 			if (LIST_EMPTY(&ih->states))
3813 				continue;
3814 
3815 			PF_HASHROW_LOCK(ih);
3816 			count = 0;
3817 			LIST_FOREACH(s, &ih->states, entry) {
3818 				if (s->timeout == PFTM_UNLINKED)
3819 					continue;
3820 				count++;
3821 			}
3822 
3823 			if (count > slice_count) {
3824 				PF_HASHROW_UNLOCK(ih);
3825 				free(pstore, M_TEMP);
3826 				slice_count = count * 2;
3827 				pstore = mallocarray(slice_count,
3828 				    sizeof(struct pfsync_state_1301), M_TEMP,
3829 				    M_WAITOK | M_ZERO);
3830 				goto DIOCGETSTATES_retry;
3831 			}
3832 
3833 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3834 				PF_HASHROW_UNLOCK(ih);
3835 				goto DIOCGETSTATES_full;
3836 			}
3837 
3838 			LIST_FOREACH(s, &ih->states, entry) {
3839 				if (s->timeout == PFTM_UNLINKED)
3840 					continue;
3841 
3842 				pfsync_state_export((union pfsync_state_union*)p,
3843 				    s, PFSYNC_MSG_VERSION_1301);
3844 				p++;
3845 				nr++;
3846 			}
3847 			PF_HASHROW_UNLOCK(ih);
3848 			error = copyout(pstore, out,
3849 			    sizeof(struct pfsync_state_1301) * count);
3850 			if (error)
3851 				break;
3852 			out = ps->ps_states + nr;
3853 		}
3854 DIOCGETSTATES_full:
3855 		ps->ps_len = sizeof(struct pfsync_state_1301) * nr;
3856 		free(pstore, M_TEMP);
3857 
3858 		break;
3859 	}
3860 
3861 	case DIOCGETSTATESV2: {
3862 		struct pfioc_states_v2	*ps = (struct pfioc_states_v2 *)addr;
3863 		struct pf_kstate	*s;
3864 		struct pf_state_export	*pstore, *p;
3865 		int i, nr;
3866 		size_t slice_count = 16, count;
3867 		void *out;
3868 
3869 		if (ps->ps_req_version > PF_STATE_VERSION) {
3870 			error = ENOTSUP;
3871 			break;
3872 		}
3873 
3874 		if (ps->ps_len <= 0) {
3875 			nr = uma_zone_get_cur(V_pf_state_z);
3876 			ps->ps_len = sizeof(struct pf_state_export) * nr;
3877 			break;
3878 		}
3879 
3880 		out = ps->ps_states;
3881 		pstore = mallocarray(slice_count,
3882 		    sizeof(struct pf_state_export), M_TEMP, M_WAITOK | M_ZERO);
3883 		nr = 0;
3884 
3885 		for (i = 0; i <= V_pf_hashmask; i++) {
3886 			struct pf_idhash *ih = &V_pf_idhash[i];
3887 
3888 DIOCGETSTATESV2_retry:
3889 			p = pstore;
3890 
3891 			if (LIST_EMPTY(&ih->states))
3892 				continue;
3893 
3894 			PF_HASHROW_LOCK(ih);
3895 			count = 0;
3896 			LIST_FOREACH(s, &ih->states, entry) {
3897 				if (s->timeout == PFTM_UNLINKED)
3898 					continue;
3899 				count++;
3900 			}
3901 
3902 			if (count > slice_count) {
3903 				PF_HASHROW_UNLOCK(ih);
3904 				free(pstore, M_TEMP);
3905 				slice_count = count * 2;
3906 				pstore = mallocarray(slice_count,
3907 				    sizeof(struct pf_state_export), M_TEMP,
3908 				    M_WAITOK | M_ZERO);
3909 				goto DIOCGETSTATESV2_retry;
3910 			}
3911 
3912 			if ((nr+count) * sizeof(*p) > ps->ps_len) {
3913 				PF_HASHROW_UNLOCK(ih);
3914 				goto DIOCGETSTATESV2_full;
3915 			}
3916 
3917 			LIST_FOREACH(s, &ih->states, entry) {
3918 				if (s->timeout == PFTM_UNLINKED)
3919 					continue;
3920 
3921 				pf_state_export(p, s);
3922 				p++;
3923 				nr++;
3924 			}
3925 			PF_HASHROW_UNLOCK(ih);
3926 			error = copyout(pstore, out,
3927 			    sizeof(struct pf_state_export) * count);
3928 			if (error)
3929 				break;
3930 			out = ps->ps_states + nr;
3931 		}
3932 DIOCGETSTATESV2_full:
3933 		ps->ps_len = nr * sizeof(struct pf_state_export);
3934 		free(pstore, M_TEMP);
3935 
3936 		break;
3937 	}
3938 #endif
3939 	case DIOCGETSTATUSNV: {
3940 		error = pf_getstatus((struct pfioc_nv *)addr);
3941 		break;
3942 	}
3943 
3944 	case DIOCSETSTATUSIF: {
3945 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
3946 
3947 		if (pi->ifname[0] == 0) {
3948 			bzero(V_pf_status.ifname, IFNAMSIZ);
3949 			break;
3950 		}
3951 		PF_RULES_WLOCK();
3952 		error = pf_user_strcpy(V_pf_status.ifname, pi->ifname, IFNAMSIZ);
3953 		PF_RULES_WUNLOCK();
3954 		break;
3955 	}
3956 
3957 	case DIOCCLRSTATUS: {
3958 		pf_ioctl_clear_status();
3959 		break;
3960 	}
3961 
3962 	case DIOCNATLOOK: {
3963 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
3964 		struct pf_state_key	*sk;
3965 		struct pf_kstate	*state;
3966 		struct pf_state_key_cmp	 key;
3967 		int			 m = 0, direction = pnl->direction;
3968 		int			 sidx, didx;
3969 
3970 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
3971 		sidx = (direction == PF_IN) ? 1 : 0;
3972 		didx = (direction == PF_IN) ? 0 : 1;
3973 
3974 		if (!pnl->proto ||
3975 		    PF_AZERO(&pnl->saddr, pnl->af) ||
3976 		    PF_AZERO(&pnl->daddr, pnl->af) ||
3977 		    ((pnl->proto == IPPROTO_TCP ||
3978 		    pnl->proto == IPPROTO_UDP) &&
3979 		    (!pnl->dport || !pnl->sport)))
3980 			error = EINVAL;
3981 		else {
3982 			bzero(&key, sizeof(key));
3983 			key.af = pnl->af;
3984 			key.proto = pnl->proto;
3985 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
3986 			key.port[sidx] = pnl->sport;
3987 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
3988 			key.port[didx] = pnl->dport;
3989 
3990 			state = pf_find_state_all(&key, direction, &m);
3991 			if (state == NULL) {
3992 				error = ENOENT;
3993 			} else {
3994 				if (m > 1) {
3995 					PF_STATE_UNLOCK(state);
3996 					error = E2BIG;	/* more than one state */
3997 				} else {
3998 					sk = state->key[sidx];
3999 					PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
4000 					pnl->rsport = sk->port[sidx];
4001 					PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
4002 					pnl->rdport = sk->port[didx];
4003 					PF_STATE_UNLOCK(state);
4004 				}
4005 			}
4006 		}
4007 		break;
4008 	}
4009 
4010 	case DIOCSETTIMEOUT: {
4011 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4012 
4013 		error = pf_ioctl_set_timeout(pt->timeout, pt->seconds,
4014 		    &pt->seconds);
4015 		break;
4016 	}
4017 
4018 	case DIOCGETTIMEOUT: {
4019 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
4020 
4021 		error = pf_ioctl_get_timeout(pt->timeout, &pt->seconds);
4022 		break;
4023 	}
4024 
4025 	case DIOCGETLIMIT: {
4026 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4027 
4028 		error = pf_ioctl_get_limit(pl->index, &pl->limit);
4029 		break;
4030 	}
4031 
4032 	case DIOCSETLIMIT: {
4033 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
4034 		unsigned int old_limit;
4035 
4036 		error = pf_ioctl_set_limit(pl->index, pl->limit, &old_limit);
4037 		pl->limit = old_limit;
4038 		break;
4039 	}
4040 
4041 	case DIOCSETDEBUG: {
4042 		u_int32_t	*level = (u_int32_t *)addr;
4043 
4044 		PF_RULES_WLOCK();
4045 		V_pf_status.debug = *level;
4046 		PF_RULES_WUNLOCK();
4047 		break;
4048 	}
4049 
4050 	case DIOCCLRRULECTRS: {
4051 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
4052 		struct pf_kruleset	*ruleset = &pf_main_ruleset;
4053 		struct pf_krule		*rule;
4054 
4055 		PF_RULES_WLOCK();
4056 		TAILQ_FOREACH(rule,
4057 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
4058 			pf_counter_u64_zero(&rule->evaluations);
4059 			for (int i = 0; i < 2; i++) {
4060 				pf_counter_u64_zero(&rule->packets[i]);
4061 				pf_counter_u64_zero(&rule->bytes[i]);
4062 			}
4063 		}
4064 		PF_RULES_WUNLOCK();
4065 		break;
4066 	}
4067 
4068 	case DIOCGIFSPEEDV0:
4069 	case DIOCGIFSPEEDV1: {
4070 		struct pf_ifspeed_v1	*psp = (struct pf_ifspeed_v1 *)addr;
4071 		struct pf_ifspeed_v1	ps;
4072 		struct ifnet		*ifp;
4073 
4074 		if (psp->ifname[0] == '\0') {
4075 			error = EINVAL;
4076 			break;
4077 		}
4078 
4079 		error = pf_user_strcpy(ps.ifname, psp->ifname, IFNAMSIZ);
4080 		if (error != 0)
4081 			break;
4082 		ifp = ifunit(ps.ifname);
4083 		if (ifp != NULL) {
4084 			psp->baudrate32 =
4085 			    (u_int32_t)uqmin(ifp->if_baudrate, UINT_MAX);
4086 			if (cmd == DIOCGIFSPEEDV1)
4087 				psp->baudrate = ifp->if_baudrate;
4088 		} else {
4089 			error = EINVAL;
4090 		}
4091 		break;
4092 	}
4093 
4094 #ifdef ALTQ
4095 	case DIOCSTARTALTQ: {
4096 		struct pf_altq		*altq;
4097 
4098 		PF_RULES_WLOCK();
4099 		/* enable all altq interfaces on active list */
4100 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4101 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4102 				error = pf_enable_altq(altq);
4103 				if (error != 0)
4104 					break;
4105 			}
4106 		}
4107 		if (error == 0)
4108 			V_pf_altq_running = 1;
4109 		PF_RULES_WUNLOCK();
4110 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
4111 		break;
4112 	}
4113 
4114 	case DIOCSTOPALTQ: {
4115 		struct pf_altq		*altq;
4116 
4117 		PF_RULES_WLOCK();
4118 		/* disable all altq interfaces on active list */
4119 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries) {
4120 			if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) == 0) {
4121 				error = pf_disable_altq(altq);
4122 				if (error != 0)
4123 					break;
4124 			}
4125 		}
4126 		if (error == 0)
4127 			V_pf_altq_running = 0;
4128 		PF_RULES_WUNLOCK();
4129 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
4130 		break;
4131 	}
4132 
4133 	case DIOCADDALTQV0:
4134 	case DIOCADDALTQV1: {
4135 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4136 		struct pf_altq		*altq, *a;
4137 		struct ifnet		*ifp;
4138 
4139 		altq = malloc(sizeof(*altq), M_PFALTQ, M_WAITOK | M_ZERO);
4140 		error = pf_import_kaltq(pa, altq, IOCPARM_LEN(cmd));
4141 		if (error)
4142 			break;
4143 		altq->local_flags = 0;
4144 
4145 		PF_RULES_WLOCK();
4146 		if (pa->ticket != V_ticket_altqs_inactive) {
4147 			PF_RULES_WUNLOCK();
4148 			free(altq, M_PFALTQ);
4149 			error = EBUSY;
4150 			break;
4151 		}
4152 
4153 		/*
4154 		 * if this is for a queue, find the discipline and
4155 		 * copy the necessary fields
4156 		 */
4157 		if (altq->qname[0] != 0) {
4158 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
4159 				PF_RULES_WUNLOCK();
4160 				error = EBUSY;
4161 				free(altq, M_PFALTQ);
4162 				break;
4163 			}
4164 			altq->altq_disc = NULL;
4165 			TAILQ_FOREACH(a, V_pf_altq_ifs_inactive, entries) {
4166 				if (strncmp(a->ifname, altq->ifname,
4167 				    IFNAMSIZ) == 0) {
4168 					altq->altq_disc = a->altq_disc;
4169 					break;
4170 				}
4171 			}
4172 		}
4173 
4174 		if ((ifp = ifunit(altq->ifname)) == NULL)
4175 			altq->local_flags |= PFALTQ_FLAG_IF_REMOVED;
4176 		else
4177 			error = altq_add(ifp, altq);
4178 
4179 		if (error) {
4180 			PF_RULES_WUNLOCK();
4181 			free(altq, M_PFALTQ);
4182 			break;
4183 		}
4184 
4185 		if (altq->qname[0] != 0)
4186 			TAILQ_INSERT_TAIL(V_pf_altqs_inactive, altq, entries);
4187 		else
4188 			TAILQ_INSERT_TAIL(V_pf_altq_ifs_inactive, altq, entries);
4189 		/* version error check done on import above */
4190 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4191 		PF_RULES_WUNLOCK();
4192 		break;
4193 	}
4194 
4195 	case DIOCGETALTQSV0:
4196 	case DIOCGETALTQSV1: {
4197 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4198 		struct pf_altq		*altq;
4199 
4200 		PF_RULES_RLOCK();
4201 		pa->nr = 0;
4202 		TAILQ_FOREACH(altq, V_pf_altq_ifs_active, entries)
4203 			pa->nr++;
4204 		TAILQ_FOREACH(altq, V_pf_altqs_active, entries)
4205 			pa->nr++;
4206 		pa->ticket = V_ticket_altqs_active;
4207 		PF_RULES_RUNLOCK();
4208 		break;
4209 	}
4210 
4211 	case DIOCGETALTQV0:
4212 	case DIOCGETALTQV1: {
4213 		struct pfioc_altq_v1	*pa = (struct pfioc_altq_v1 *)addr;
4214 		struct pf_altq		*altq;
4215 
4216 		PF_RULES_RLOCK();
4217 		if (pa->ticket != V_ticket_altqs_active) {
4218 			PF_RULES_RUNLOCK();
4219 			error = EBUSY;
4220 			break;
4221 		}
4222 		altq = pf_altq_get_nth_active(pa->nr);
4223 		if (altq == NULL) {
4224 			PF_RULES_RUNLOCK();
4225 			error = EBUSY;
4226 			break;
4227 		}
4228 		pf_export_kaltq(altq, pa, IOCPARM_LEN(cmd));
4229 		PF_RULES_RUNLOCK();
4230 		break;
4231 	}
4232 
4233 	case DIOCCHANGEALTQV0:
4234 	case DIOCCHANGEALTQV1:
4235 		/* CHANGEALTQ not supported yet! */
4236 		error = ENODEV;
4237 		break;
4238 
4239 	case DIOCGETQSTATSV0:
4240 	case DIOCGETQSTATSV1: {
4241 		struct pfioc_qstats_v1	*pq = (struct pfioc_qstats_v1 *)addr;
4242 		struct pf_altq		*altq;
4243 		int			 nbytes;
4244 		u_int32_t		 version;
4245 
4246 		PF_RULES_RLOCK();
4247 		if (pq->ticket != V_ticket_altqs_active) {
4248 			PF_RULES_RUNLOCK();
4249 			error = EBUSY;
4250 			break;
4251 		}
4252 		nbytes = pq->nbytes;
4253 		altq = pf_altq_get_nth_active(pq->nr);
4254 		if (altq == NULL) {
4255 			PF_RULES_RUNLOCK();
4256 			error = EBUSY;
4257 			break;
4258 		}
4259 
4260 		if ((altq->local_flags & PFALTQ_FLAG_IF_REMOVED) != 0) {
4261 			PF_RULES_RUNLOCK();
4262 			error = ENXIO;
4263 			break;
4264 		}
4265 		PF_RULES_RUNLOCK();
4266 		if (cmd == DIOCGETQSTATSV0)
4267 			version = 0;  /* DIOCGETQSTATSV0 means stats struct v0 */
4268 		else
4269 			version = pq->version;
4270 		error = altq_getqstats(altq, pq->buf, &nbytes, version);
4271 		if (error == 0) {
4272 			pq->scheduler = altq->scheduler;
4273 			pq->nbytes = nbytes;
4274 		}
4275 		break;
4276 	}
4277 #endif /* ALTQ */
4278 
4279 	case DIOCBEGINADDRS: {
4280 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4281 
4282 		error = pf_ioctl_begin_addrs(&pp->ticket);
4283 		break;
4284 	}
4285 
4286 	case DIOCADDADDR: {
4287 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4288 
4289 		error = pf_ioctl_add_addr(pp);
4290 		break;
4291 	}
4292 
4293 	case DIOCGETADDRS: {
4294 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4295 
4296 		error = pf_ioctl_get_addrs(pp);
4297 		break;
4298 	}
4299 
4300 	case DIOCGETADDR: {
4301 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
4302 		struct pf_kpool		*pool;
4303 		struct pf_kpooladdr	*pa;
4304 		u_int32_t		 nr = 0;
4305 
4306 		pp->anchor[sizeof(pp->anchor) - 1] = 0;
4307 
4308 		PF_RULES_RLOCK();
4309 		pool = pf_get_kpool(pp->anchor, pp->ticket, pp->r_action,
4310 		    pp->r_num, 0, 1, 1);
4311 		if (pool == NULL) {
4312 			PF_RULES_RUNLOCK();
4313 			error = EBUSY;
4314 			break;
4315 		}
4316 		pa = TAILQ_FIRST(&pool->list);
4317 		while ((pa != NULL) && (nr < pp->nr)) {
4318 			pa = TAILQ_NEXT(pa, entries);
4319 			nr++;
4320 		}
4321 		if (pa == NULL) {
4322 			PF_RULES_RUNLOCK();
4323 			error = EBUSY;
4324 			break;
4325 		}
4326 		pf_kpooladdr_to_pooladdr(pa, &pp->addr);
4327 		pf_addr_copyout(&pp->addr.addr);
4328 		PF_RULES_RUNLOCK();
4329 		break;
4330 	}
4331 
4332 	case DIOCCHANGEADDR: {
4333 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
4334 		struct pf_kpool		*pool;
4335 		struct pf_kpooladdr	*oldpa = NULL, *newpa = NULL;
4336 		struct pf_kruleset	*ruleset;
4337 		struct pfi_kkif		*kif = NULL;
4338 
4339 		pca->anchor[sizeof(pca->anchor) - 1] = 0;
4340 
4341 		if (pca->action < PF_CHANGE_ADD_HEAD ||
4342 		    pca->action > PF_CHANGE_REMOVE) {
4343 			error = EINVAL;
4344 			break;
4345 		}
4346 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
4347 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
4348 		    pca->addr.addr.type != PF_ADDR_TABLE) {
4349 			error = EINVAL;
4350 			break;
4351 		}
4352 		if (pca->addr.addr.p.dyn != NULL) {
4353 			error = EINVAL;
4354 			break;
4355 		}
4356 
4357 		if (pca->action != PF_CHANGE_REMOVE) {
4358 #ifndef INET
4359 			if (pca->af == AF_INET) {
4360 				error = EAFNOSUPPORT;
4361 				break;
4362 			}
4363 #endif /* INET */
4364 #ifndef INET6
4365 			if (pca->af == AF_INET6) {
4366 				error = EAFNOSUPPORT;
4367 				break;
4368 			}
4369 #endif /* INET6 */
4370 			newpa = malloc(sizeof(*newpa), M_PFRULE, M_WAITOK);
4371 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
4372 			if (newpa->ifname[0])
4373 				kif = pf_kkif_create(M_WAITOK);
4374 			newpa->kif = NULL;
4375 		}
4376 #define	ERROUT(x)	ERROUT_IOCTL(DIOCCHANGEADDR_error, x)
4377 		PF_RULES_WLOCK();
4378 		ruleset = pf_find_kruleset(pca->anchor);
4379 		if (ruleset == NULL)
4380 			ERROUT(EBUSY);
4381 
4382 		pool = pf_get_kpool(pca->anchor, pca->ticket, pca->r_action,
4383 		    pca->r_num, pca->r_last, 1, 1);
4384 		if (pool == NULL)
4385 			ERROUT(EBUSY);
4386 
4387 		if (pca->action != PF_CHANGE_REMOVE) {
4388 			if (newpa->ifname[0]) {
4389 				newpa->kif = pfi_kkif_attach(kif, newpa->ifname);
4390 				pfi_kkif_ref(newpa->kif);
4391 				kif = NULL;
4392 			}
4393 
4394 			switch (newpa->addr.type) {
4395 			case PF_ADDR_DYNIFTL:
4396 				error = pfi_dynaddr_setup(&newpa->addr,
4397 				    pca->af);
4398 				break;
4399 			case PF_ADDR_TABLE:
4400 				newpa->addr.p.tbl = pfr_attach_table(ruleset,
4401 				    newpa->addr.v.tblname);
4402 				if (newpa->addr.p.tbl == NULL)
4403 					error = ENOMEM;
4404 				break;
4405 			}
4406 			if (error)
4407 				goto DIOCCHANGEADDR_error;
4408 		}
4409 
4410 		switch (pca->action) {
4411 		case PF_CHANGE_ADD_HEAD:
4412 			oldpa = TAILQ_FIRST(&pool->list);
4413 			break;
4414 		case PF_CHANGE_ADD_TAIL:
4415 			oldpa = TAILQ_LAST(&pool->list, pf_kpalist);
4416 			break;
4417 		default:
4418 			oldpa = TAILQ_FIRST(&pool->list);
4419 			for (int i = 0; oldpa && i < pca->nr; i++)
4420 				oldpa = TAILQ_NEXT(oldpa, entries);
4421 
4422 			if (oldpa == NULL)
4423 				ERROUT(EINVAL);
4424 		}
4425 
4426 		if (pca->action == PF_CHANGE_REMOVE) {
4427 			TAILQ_REMOVE(&pool->list, oldpa, entries);
4428 			switch (oldpa->addr.type) {
4429 			case PF_ADDR_DYNIFTL:
4430 				pfi_dynaddr_remove(oldpa->addr.p.dyn);
4431 				break;
4432 			case PF_ADDR_TABLE:
4433 				pfr_detach_table(oldpa->addr.p.tbl);
4434 				break;
4435 			}
4436 			if (oldpa->kif)
4437 				pfi_kkif_unref(oldpa->kif);
4438 			free(oldpa, M_PFRULE);
4439 		} else {
4440 			if (oldpa == NULL)
4441 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
4442 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
4443 			    pca->action == PF_CHANGE_ADD_BEFORE)
4444 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
4445 			else
4446 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
4447 				    newpa, entries);
4448 		}
4449 
4450 		pool->cur = TAILQ_FIRST(&pool->list);
4451 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr, pca->af);
4452 		PF_RULES_WUNLOCK();
4453 		break;
4454 
4455 #undef ERROUT
4456 DIOCCHANGEADDR_error:
4457 		if (newpa != NULL) {
4458 			if (newpa->kif)
4459 				pfi_kkif_unref(newpa->kif);
4460 			free(newpa, M_PFRULE);
4461 		}
4462 		PF_RULES_WUNLOCK();
4463 		pf_kkif_free(kif);
4464 		break;
4465 	}
4466 
4467 	case DIOCGETRULESETS: {
4468 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4469 		struct pf_kruleset	*ruleset;
4470 		struct pf_kanchor	*anchor;
4471 
4472 		pr->path[sizeof(pr->path) - 1] = 0;
4473 
4474 		PF_RULES_RLOCK();
4475 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4476 			PF_RULES_RUNLOCK();
4477 			error = ENOENT;
4478 			break;
4479 		}
4480 		pr->nr = 0;
4481 		if (ruleset->anchor == NULL) {
4482 			/* XXX kludge for pf_main_ruleset */
4483 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4484 				if (anchor->parent == NULL)
4485 					pr->nr++;
4486 		} else {
4487 			RB_FOREACH(anchor, pf_kanchor_node,
4488 			    &ruleset->anchor->children)
4489 				pr->nr++;
4490 		}
4491 		PF_RULES_RUNLOCK();
4492 		break;
4493 	}
4494 
4495 	case DIOCGETRULESET: {
4496 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
4497 		struct pf_kruleset	*ruleset;
4498 		struct pf_kanchor	*anchor;
4499 		u_int32_t		 nr = 0;
4500 
4501 		pr->path[sizeof(pr->path) - 1] = 0;
4502 
4503 		PF_RULES_RLOCK();
4504 		if ((ruleset = pf_find_kruleset(pr->path)) == NULL) {
4505 			PF_RULES_RUNLOCK();
4506 			error = ENOENT;
4507 			break;
4508 		}
4509 		pr->name[0] = 0;
4510 		if (ruleset->anchor == NULL) {
4511 			/* XXX kludge for pf_main_ruleset */
4512 			RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors)
4513 				if (anchor->parent == NULL && nr++ == pr->nr) {
4514 					strlcpy(pr->name, anchor->name,
4515 					    sizeof(pr->name));
4516 					break;
4517 				}
4518 		} else {
4519 			RB_FOREACH(anchor, pf_kanchor_node,
4520 			    &ruleset->anchor->children)
4521 				if (nr++ == pr->nr) {
4522 					strlcpy(pr->name, anchor->name,
4523 					    sizeof(pr->name));
4524 					break;
4525 				}
4526 		}
4527 		if (!pr->name[0])
4528 			error = EBUSY;
4529 		PF_RULES_RUNLOCK();
4530 		break;
4531 	}
4532 
4533 	case DIOCRCLRTABLES: {
4534 		struct pfioc_table *io = (struct pfioc_table *)addr;
4535 
4536 		if (io->pfrio_esize != 0) {
4537 			error = ENODEV;
4538 			break;
4539 		}
4540 		PF_RULES_WLOCK();
4541 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
4542 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4543 		PF_RULES_WUNLOCK();
4544 		break;
4545 	}
4546 
4547 	case DIOCRADDTABLES: {
4548 		struct pfioc_table *io = (struct pfioc_table *)addr;
4549 		struct pfr_table *pfrts;
4550 		size_t totlen;
4551 
4552 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4553 			error = ENODEV;
4554 			break;
4555 		}
4556 
4557 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4558 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4559 			error = ENOMEM;
4560 			break;
4561 		}
4562 
4563 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4564 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4565 		    M_TEMP, M_WAITOK);
4566 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4567 		if (error) {
4568 			free(pfrts, M_TEMP);
4569 			break;
4570 		}
4571 		PF_RULES_WLOCK();
4572 		error = pfr_add_tables(pfrts, io->pfrio_size,
4573 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4574 		PF_RULES_WUNLOCK();
4575 		free(pfrts, M_TEMP);
4576 		break;
4577 	}
4578 
4579 	case DIOCRDELTABLES: {
4580 		struct pfioc_table *io = (struct pfioc_table *)addr;
4581 		struct pfr_table *pfrts;
4582 		size_t totlen;
4583 
4584 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4585 			error = ENODEV;
4586 			break;
4587 		}
4588 
4589 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4590 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4591 			error = ENOMEM;
4592 			break;
4593 		}
4594 
4595 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4596 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4597 		    M_TEMP, M_WAITOK);
4598 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4599 		if (error) {
4600 			free(pfrts, M_TEMP);
4601 			break;
4602 		}
4603 		PF_RULES_WLOCK();
4604 		error = pfr_del_tables(pfrts, io->pfrio_size,
4605 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4606 		PF_RULES_WUNLOCK();
4607 		free(pfrts, M_TEMP);
4608 		break;
4609 	}
4610 
4611 	case DIOCRGETTABLES: {
4612 		struct pfioc_table *io = (struct pfioc_table *)addr;
4613 		struct pfr_table *pfrts;
4614 		size_t totlen;
4615 		int n;
4616 
4617 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4618 			error = ENODEV;
4619 			break;
4620 		}
4621 		PF_RULES_RLOCK();
4622 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4623 		if (n < 0) {
4624 			PF_RULES_RUNLOCK();
4625 			error = EINVAL;
4626 			break;
4627 		}
4628 		io->pfrio_size = min(io->pfrio_size, n);
4629 
4630 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4631 
4632 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4633 		    M_TEMP, M_NOWAIT | M_ZERO);
4634 		if (pfrts == NULL) {
4635 			error = ENOMEM;
4636 			PF_RULES_RUNLOCK();
4637 			break;
4638 		}
4639 		error = pfr_get_tables(&io->pfrio_table, pfrts,
4640 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4641 		PF_RULES_RUNLOCK();
4642 		if (error == 0)
4643 			error = copyout(pfrts, io->pfrio_buffer, totlen);
4644 		free(pfrts, M_TEMP);
4645 		break;
4646 	}
4647 
4648 	case DIOCRGETTSTATS: {
4649 		struct pfioc_table *io = (struct pfioc_table *)addr;
4650 		struct pfr_tstats *pfrtstats;
4651 		size_t totlen;
4652 		int n;
4653 
4654 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
4655 			error = ENODEV;
4656 			break;
4657 		}
4658 		PF_TABLE_STATS_LOCK();
4659 		PF_RULES_RLOCK();
4660 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4661 		if (n < 0) {
4662 			PF_RULES_RUNLOCK();
4663 			PF_TABLE_STATS_UNLOCK();
4664 			error = EINVAL;
4665 			break;
4666 		}
4667 		io->pfrio_size = min(io->pfrio_size, n);
4668 
4669 		totlen = io->pfrio_size * sizeof(struct pfr_tstats);
4670 		pfrtstats = mallocarray(io->pfrio_size,
4671 		    sizeof(struct pfr_tstats), M_TEMP, M_NOWAIT | M_ZERO);
4672 		if (pfrtstats == NULL) {
4673 			error = ENOMEM;
4674 			PF_RULES_RUNLOCK();
4675 			PF_TABLE_STATS_UNLOCK();
4676 			break;
4677 		}
4678 		error = pfr_get_tstats(&io->pfrio_table, pfrtstats,
4679 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4680 		PF_RULES_RUNLOCK();
4681 		PF_TABLE_STATS_UNLOCK();
4682 		if (error == 0)
4683 			error = copyout(pfrtstats, io->pfrio_buffer, totlen);
4684 		free(pfrtstats, M_TEMP);
4685 		break;
4686 	}
4687 
4688 	case DIOCRCLRTSTATS: {
4689 		struct pfioc_table *io = (struct pfioc_table *)addr;
4690 		struct pfr_table *pfrts;
4691 		size_t totlen;
4692 
4693 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4694 			error = ENODEV;
4695 			break;
4696 		}
4697 
4698 		if (io->pfrio_size < 0 || io->pfrio_size > pf_ioctl_maxcount ||
4699 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_table))) {
4700 			/* We used to count tables and use the minimum required
4701 			 * size, so we didn't fail on overly large requests.
4702 			 * Keep doing so. */
4703 			io->pfrio_size = pf_ioctl_maxcount;
4704 			break;
4705 		}
4706 
4707 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4708 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4709 		    M_TEMP, M_WAITOK);
4710 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4711 		if (error) {
4712 			free(pfrts, M_TEMP);
4713 			break;
4714 		}
4715 
4716 		PF_TABLE_STATS_LOCK();
4717 		PF_RULES_RLOCK();
4718 		error = pfr_clr_tstats(pfrts, io->pfrio_size,
4719 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4720 		PF_RULES_RUNLOCK();
4721 		PF_TABLE_STATS_UNLOCK();
4722 		free(pfrts, M_TEMP);
4723 		break;
4724 	}
4725 
4726 	case DIOCRSETTFLAGS: {
4727 		struct pfioc_table *io = (struct pfioc_table *)addr;
4728 		struct pfr_table *pfrts;
4729 		size_t totlen;
4730 		int n;
4731 
4732 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
4733 			error = ENODEV;
4734 			break;
4735 		}
4736 
4737 		PF_RULES_RLOCK();
4738 		n = pfr_table_count(&io->pfrio_table, io->pfrio_flags);
4739 		if (n < 0) {
4740 			PF_RULES_RUNLOCK();
4741 			error = EINVAL;
4742 			break;
4743 		}
4744 
4745 		io->pfrio_size = min(io->pfrio_size, n);
4746 		PF_RULES_RUNLOCK();
4747 
4748 		totlen = io->pfrio_size * sizeof(struct pfr_table);
4749 		pfrts = mallocarray(io->pfrio_size, sizeof(struct pfr_table),
4750 		    M_TEMP, M_WAITOK);
4751 		error = copyin(io->pfrio_buffer, pfrts, totlen);
4752 		if (error) {
4753 			free(pfrts, M_TEMP);
4754 			break;
4755 		}
4756 		PF_RULES_WLOCK();
4757 		error = pfr_set_tflags(pfrts, io->pfrio_size,
4758 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
4759 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4760 		PF_RULES_WUNLOCK();
4761 		free(pfrts, M_TEMP);
4762 		break;
4763 	}
4764 
4765 	case DIOCRCLRADDRS: {
4766 		struct pfioc_table *io = (struct pfioc_table *)addr;
4767 
4768 		if (io->pfrio_esize != 0) {
4769 			error = ENODEV;
4770 			break;
4771 		}
4772 		PF_RULES_WLOCK();
4773 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
4774 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
4775 		PF_RULES_WUNLOCK();
4776 		break;
4777 	}
4778 
4779 	case DIOCRADDADDRS: {
4780 		struct pfioc_table *io = (struct pfioc_table *)addr;
4781 		struct pfr_addr *pfras;
4782 		size_t totlen;
4783 
4784 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4785 			error = ENODEV;
4786 			break;
4787 		}
4788 		if (io->pfrio_size < 0 ||
4789 		    io->pfrio_size > pf_ioctl_maxcount ||
4790 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4791 			error = EINVAL;
4792 			break;
4793 		}
4794 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4795 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4796 		    M_TEMP, M_WAITOK);
4797 		error = copyin(io->pfrio_buffer, pfras, totlen);
4798 		if (error) {
4799 			free(pfras, M_TEMP);
4800 			break;
4801 		}
4802 		PF_RULES_WLOCK();
4803 		error = pfr_add_addrs(&io->pfrio_table, pfras,
4804 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
4805 		    PFR_FLAG_USERIOCTL);
4806 		PF_RULES_WUNLOCK();
4807 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4808 			error = copyout(pfras, io->pfrio_buffer, totlen);
4809 		free(pfras, M_TEMP);
4810 		break;
4811 	}
4812 
4813 	case DIOCRDELADDRS: {
4814 		struct pfioc_table *io = (struct pfioc_table *)addr;
4815 		struct pfr_addr *pfras;
4816 		size_t totlen;
4817 
4818 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4819 			error = ENODEV;
4820 			break;
4821 		}
4822 		if (io->pfrio_size < 0 ||
4823 		    io->pfrio_size > pf_ioctl_maxcount ||
4824 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4825 			error = EINVAL;
4826 			break;
4827 		}
4828 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4829 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4830 		    M_TEMP, M_WAITOK);
4831 		error = copyin(io->pfrio_buffer, pfras, totlen);
4832 		if (error) {
4833 			free(pfras, M_TEMP);
4834 			break;
4835 		}
4836 		PF_RULES_WLOCK();
4837 		error = pfr_del_addrs(&io->pfrio_table, pfras,
4838 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
4839 		    PFR_FLAG_USERIOCTL);
4840 		PF_RULES_WUNLOCK();
4841 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4842 			error = copyout(pfras, io->pfrio_buffer, totlen);
4843 		free(pfras, M_TEMP);
4844 		break;
4845 	}
4846 
4847 	case DIOCRSETADDRS: {
4848 		struct pfioc_table *io = (struct pfioc_table *)addr;
4849 		struct pfr_addr *pfras;
4850 		size_t totlen, count;
4851 
4852 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4853 			error = ENODEV;
4854 			break;
4855 		}
4856 		if (io->pfrio_size < 0 || io->pfrio_size2 < 0) {
4857 			error = EINVAL;
4858 			break;
4859 		}
4860 		count = max(io->pfrio_size, io->pfrio_size2);
4861 		if (count > pf_ioctl_maxcount ||
4862 		    WOULD_OVERFLOW(count, sizeof(struct pfr_addr))) {
4863 			error = EINVAL;
4864 			break;
4865 		}
4866 		totlen = count * sizeof(struct pfr_addr);
4867 		pfras = mallocarray(count, sizeof(struct pfr_addr), M_TEMP,
4868 		    M_WAITOK);
4869 		error = copyin(io->pfrio_buffer, pfras, totlen);
4870 		if (error) {
4871 			free(pfras, M_TEMP);
4872 			break;
4873 		}
4874 		PF_RULES_WLOCK();
4875 		error = pfr_set_addrs(&io->pfrio_table, pfras,
4876 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
4877 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
4878 		    PFR_FLAG_USERIOCTL, 0);
4879 		PF_RULES_WUNLOCK();
4880 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4881 			error = copyout(pfras, io->pfrio_buffer, totlen);
4882 		free(pfras, M_TEMP);
4883 		break;
4884 	}
4885 
4886 	case DIOCRGETADDRS: {
4887 		struct pfioc_table *io = (struct pfioc_table *)addr;
4888 		struct pfr_addr *pfras;
4889 		size_t totlen;
4890 
4891 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4892 			error = ENODEV;
4893 			break;
4894 		}
4895 		if (io->pfrio_size < 0 ||
4896 		    io->pfrio_size > pf_ioctl_maxcount ||
4897 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4898 			error = EINVAL;
4899 			break;
4900 		}
4901 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4902 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4903 		    M_TEMP, M_WAITOK | M_ZERO);
4904 		PF_RULES_RLOCK();
4905 		error = pfr_get_addrs(&io->pfrio_table, pfras,
4906 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4907 		PF_RULES_RUNLOCK();
4908 		if (error == 0)
4909 			error = copyout(pfras, io->pfrio_buffer, totlen);
4910 		free(pfras, M_TEMP);
4911 		break;
4912 	}
4913 
4914 	case DIOCRGETASTATS: {
4915 		struct pfioc_table *io = (struct pfioc_table *)addr;
4916 		struct pfr_astats *pfrastats;
4917 		size_t totlen;
4918 
4919 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
4920 			error = ENODEV;
4921 			break;
4922 		}
4923 		if (io->pfrio_size < 0 ||
4924 		    io->pfrio_size > pf_ioctl_maxcount ||
4925 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_astats))) {
4926 			error = EINVAL;
4927 			break;
4928 		}
4929 		totlen = io->pfrio_size * sizeof(struct pfr_astats);
4930 		pfrastats = mallocarray(io->pfrio_size,
4931 		    sizeof(struct pfr_astats), M_TEMP, M_WAITOK | M_ZERO);
4932 		PF_RULES_RLOCK();
4933 		error = pfr_get_astats(&io->pfrio_table, pfrastats,
4934 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
4935 		PF_RULES_RUNLOCK();
4936 		if (error == 0)
4937 			error = copyout(pfrastats, io->pfrio_buffer, totlen);
4938 		free(pfrastats, M_TEMP);
4939 		break;
4940 	}
4941 
4942 	case DIOCRCLRASTATS: {
4943 		struct pfioc_table *io = (struct pfioc_table *)addr;
4944 		struct pfr_addr *pfras;
4945 		size_t totlen;
4946 
4947 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4948 			error = ENODEV;
4949 			break;
4950 		}
4951 		if (io->pfrio_size < 0 ||
4952 		    io->pfrio_size > pf_ioctl_maxcount ||
4953 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4954 			error = EINVAL;
4955 			break;
4956 		}
4957 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4958 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4959 		    M_TEMP, M_WAITOK);
4960 		error = copyin(io->pfrio_buffer, pfras, totlen);
4961 		if (error) {
4962 			free(pfras, M_TEMP);
4963 			break;
4964 		}
4965 		PF_RULES_WLOCK();
4966 		error = pfr_clr_astats(&io->pfrio_table, pfras,
4967 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
4968 		    PFR_FLAG_USERIOCTL);
4969 		PF_RULES_WUNLOCK();
4970 		if (error == 0 && io->pfrio_flags & PFR_FLAG_FEEDBACK)
4971 			error = copyout(pfras, io->pfrio_buffer, totlen);
4972 		free(pfras, M_TEMP);
4973 		break;
4974 	}
4975 
4976 	case DIOCRTSTADDRS: {
4977 		struct pfioc_table *io = (struct pfioc_table *)addr;
4978 		struct pfr_addr *pfras;
4979 		size_t totlen;
4980 
4981 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
4982 			error = ENODEV;
4983 			break;
4984 		}
4985 		if (io->pfrio_size < 0 ||
4986 		    io->pfrio_size > pf_ioctl_maxcount ||
4987 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
4988 			error = EINVAL;
4989 			break;
4990 		}
4991 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
4992 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
4993 		    M_TEMP, M_WAITOK);
4994 		error = copyin(io->pfrio_buffer, pfras, totlen);
4995 		if (error) {
4996 			free(pfras, M_TEMP);
4997 			break;
4998 		}
4999 		PF_RULES_RLOCK();
5000 		error = pfr_tst_addrs(&io->pfrio_table, pfras,
5001 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
5002 		    PFR_FLAG_USERIOCTL);
5003 		PF_RULES_RUNLOCK();
5004 		if (error == 0)
5005 			error = copyout(pfras, io->pfrio_buffer, totlen);
5006 		free(pfras, M_TEMP);
5007 		break;
5008 	}
5009 
5010 	case DIOCRINADEFINE: {
5011 		struct pfioc_table *io = (struct pfioc_table *)addr;
5012 		struct pfr_addr *pfras;
5013 		size_t totlen;
5014 
5015 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
5016 			error = ENODEV;
5017 			break;
5018 		}
5019 		if (io->pfrio_size < 0 ||
5020 		    io->pfrio_size > pf_ioctl_maxcount ||
5021 		    WOULD_OVERFLOW(io->pfrio_size, sizeof(struct pfr_addr))) {
5022 			error = EINVAL;
5023 			break;
5024 		}
5025 		totlen = io->pfrio_size * sizeof(struct pfr_addr);
5026 		pfras = mallocarray(io->pfrio_size, sizeof(struct pfr_addr),
5027 		    M_TEMP, M_WAITOK);
5028 		error = copyin(io->pfrio_buffer, pfras, totlen);
5029 		if (error) {
5030 			free(pfras, M_TEMP);
5031 			break;
5032 		}
5033 		PF_RULES_WLOCK();
5034 		error = pfr_ina_define(&io->pfrio_table, pfras,
5035 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
5036 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
5037 		PF_RULES_WUNLOCK();
5038 		free(pfras, M_TEMP);
5039 		break;
5040 	}
5041 
5042 	case DIOCOSFPADD: {
5043 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5044 		PF_RULES_WLOCK();
5045 		error = pf_osfp_add(io);
5046 		PF_RULES_WUNLOCK();
5047 		break;
5048 	}
5049 
5050 	case DIOCOSFPGET: {
5051 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
5052 		PF_RULES_RLOCK();
5053 		error = pf_osfp_get(io);
5054 		PF_RULES_RUNLOCK();
5055 		break;
5056 	}
5057 
5058 	case DIOCXBEGIN: {
5059 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5060 		struct pfioc_trans_e	*ioes, *ioe;
5061 		size_t			 totlen;
5062 		int			 i;
5063 
5064 		if (io->esize != sizeof(*ioe)) {
5065 			error = ENODEV;
5066 			break;
5067 		}
5068 		if (io->size < 0 ||
5069 		    io->size > pf_ioctl_maxcount ||
5070 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5071 			error = EINVAL;
5072 			break;
5073 		}
5074 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5075 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5076 		    M_TEMP, M_WAITOK);
5077 		error = copyin(io->array, ioes, totlen);
5078 		if (error) {
5079 			free(ioes, M_TEMP);
5080 			break;
5081 		}
5082 		/* Ensure there's no more ethernet rules to clean up. */
5083 		NET_EPOCH_DRAIN_CALLBACKS();
5084 		PF_RULES_WLOCK();
5085 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5086 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5087 			switch (ioe->rs_num) {
5088 			case PF_RULESET_ETH:
5089 				if ((error = pf_begin_eth(&ioe->ticket, ioe->anchor))) {
5090 					PF_RULES_WUNLOCK();
5091 					free(ioes, M_TEMP);
5092 					goto fail;
5093 				}
5094 				break;
5095 #ifdef ALTQ
5096 			case PF_RULESET_ALTQ:
5097 				if (ioe->anchor[0]) {
5098 					PF_RULES_WUNLOCK();
5099 					free(ioes, M_TEMP);
5100 					error = EINVAL;
5101 					goto fail;
5102 				}
5103 				if ((error = pf_begin_altq(&ioe->ticket))) {
5104 					PF_RULES_WUNLOCK();
5105 					free(ioes, M_TEMP);
5106 					goto fail;
5107 				}
5108 				break;
5109 #endif /* ALTQ */
5110 			case PF_RULESET_TABLE:
5111 			    {
5112 				struct pfr_table table;
5113 
5114 				bzero(&table, sizeof(table));
5115 				strlcpy(table.pfrt_anchor, ioe->anchor,
5116 				    sizeof(table.pfrt_anchor));
5117 				if ((error = pfr_ina_begin(&table,
5118 				    &ioe->ticket, NULL, 0))) {
5119 					PF_RULES_WUNLOCK();
5120 					free(ioes, M_TEMP);
5121 					goto fail;
5122 				}
5123 				break;
5124 			    }
5125 			default:
5126 				if ((error = pf_begin_rules(&ioe->ticket,
5127 				    ioe->rs_num, ioe->anchor))) {
5128 					PF_RULES_WUNLOCK();
5129 					free(ioes, M_TEMP);
5130 					goto fail;
5131 				}
5132 				break;
5133 			}
5134 		}
5135 		PF_RULES_WUNLOCK();
5136 		error = copyout(ioes, io->array, totlen);
5137 		free(ioes, M_TEMP);
5138 		break;
5139 	}
5140 
5141 	case DIOCXROLLBACK: {
5142 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5143 		struct pfioc_trans_e	*ioe, *ioes;
5144 		size_t			 totlen;
5145 		int			 i;
5146 
5147 		if (io->esize != sizeof(*ioe)) {
5148 			error = ENODEV;
5149 			break;
5150 		}
5151 		if (io->size < 0 ||
5152 		    io->size > pf_ioctl_maxcount ||
5153 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5154 			error = EINVAL;
5155 			break;
5156 		}
5157 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5158 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5159 		    M_TEMP, M_WAITOK);
5160 		error = copyin(io->array, ioes, totlen);
5161 		if (error) {
5162 			free(ioes, M_TEMP);
5163 			break;
5164 		}
5165 		PF_RULES_WLOCK();
5166 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5167 			ioe->anchor[sizeof(ioe->anchor) - 1] = '\0';
5168 			switch (ioe->rs_num) {
5169 			case PF_RULESET_ETH:
5170 				if ((error = pf_rollback_eth(ioe->ticket,
5171 				    ioe->anchor))) {
5172 					PF_RULES_WUNLOCK();
5173 					free(ioes, M_TEMP);
5174 					goto fail; /* really bad */
5175 				}
5176 				break;
5177 #ifdef ALTQ
5178 			case PF_RULESET_ALTQ:
5179 				if (ioe->anchor[0]) {
5180 					PF_RULES_WUNLOCK();
5181 					free(ioes, M_TEMP);
5182 					error = EINVAL;
5183 					goto fail;
5184 				}
5185 				if ((error = pf_rollback_altq(ioe->ticket))) {
5186 					PF_RULES_WUNLOCK();
5187 					free(ioes, M_TEMP);
5188 					goto fail; /* really bad */
5189 				}
5190 				break;
5191 #endif /* ALTQ */
5192 			case PF_RULESET_TABLE:
5193 			    {
5194 				struct pfr_table table;
5195 
5196 				bzero(&table, sizeof(table));
5197 				strlcpy(table.pfrt_anchor, ioe->anchor,
5198 				    sizeof(table.pfrt_anchor));
5199 				if ((error = pfr_ina_rollback(&table,
5200 				    ioe->ticket, NULL, 0))) {
5201 					PF_RULES_WUNLOCK();
5202 					free(ioes, M_TEMP);
5203 					goto fail; /* really bad */
5204 				}
5205 				break;
5206 			    }
5207 			default:
5208 				if ((error = pf_rollback_rules(ioe->ticket,
5209 				    ioe->rs_num, ioe->anchor))) {
5210 					PF_RULES_WUNLOCK();
5211 					free(ioes, M_TEMP);
5212 					goto fail; /* really bad */
5213 				}
5214 				break;
5215 			}
5216 		}
5217 		PF_RULES_WUNLOCK();
5218 		free(ioes, M_TEMP);
5219 		break;
5220 	}
5221 
5222 	case DIOCXCOMMIT: {
5223 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
5224 		struct pfioc_trans_e	*ioe, *ioes;
5225 		struct pf_kruleset	*rs;
5226 		struct pf_keth_ruleset	*ers;
5227 		size_t			 totlen;
5228 		int			 i;
5229 
5230 		if (io->esize != sizeof(*ioe)) {
5231 			error = ENODEV;
5232 			break;
5233 		}
5234 
5235 		if (io->size < 0 ||
5236 		    io->size > pf_ioctl_maxcount ||
5237 		    WOULD_OVERFLOW(io->size, sizeof(struct pfioc_trans_e))) {
5238 			error = EINVAL;
5239 			break;
5240 		}
5241 
5242 		totlen = sizeof(struct pfioc_trans_e) * io->size;
5243 		ioes = mallocarray(io->size, sizeof(struct pfioc_trans_e),
5244 		    M_TEMP, M_WAITOK);
5245 		error = copyin(io->array, ioes, totlen);
5246 		if (error) {
5247 			free(ioes, M_TEMP);
5248 			break;
5249 		}
5250 		PF_RULES_WLOCK();
5251 		/* First makes sure everything will succeed. */
5252 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5253 			ioe->anchor[sizeof(ioe->anchor) - 1] = 0;
5254 			switch (ioe->rs_num) {
5255 			case PF_RULESET_ETH:
5256 				ers = pf_find_keth_ruleset(ioe->anchor);
5257 				if (ers == NULL || ioe->ticket == 0 ||
5258 				    ioe->ticket != ers->inactive.ticket) {
5259 					PF_RULES_WUNLOCK();
5260 					free(ioes, M_TEMP);
5261 					error = EINVAL;
5262 					goto fail;
5263 				}
5264 				break;
5265 #ifdef ALTQ
5266 			case PF_RULESET_ALTQ:
5267 				if (ioe->anchor[0]) {
5268 					PF_RULES_WUNLOCK();
5269 					free(ioes, M_TEMP);
5270 					error = EINVAL;
5271 					goto fail;
5272 				}
5273 				if (!V_altqs_inactive_open || ioe->ticket !=
5274 				    V_ticket_altqs_inactive) {
5275 					PF_RULES_WUNLOCK();
5276 					free(ioes, M_TEMP);
5277 					error = EBUSY;
5278 					goto fail;
5279 				}
5280 				break;
5281 #endif /* ALTQ */
5282 			case PF_RULESET_TABLE:
5283 				rs = pf_find_kruleset(ioe->anchor);
5284 				if (rs == NULL || !rs->topen || ioe->ticket !=
5285 				    rs->tticket) {
5286 					PF_RULES_WUNLOCK();
5287 					free(ioes, M_TEMP);
5288 					error = EBUSY;
5289 					goto fail;
5290 				}
5291 				break;
5292 			default:
5293 				if (ioe->rs_num < 0 || ioe->rs_num >=
5294 				    PF_RULESET_MAX) {
5295 					PF_RULES_WUNLOCK();
5296 					free(ioes, M_TEMP);
5297 					error = EINVAL;
5298 					goto fail;
5299 				}
5300 				rs = pf_find_kruleset(ioe->anchor);
5301 				if (rs == NULL ||
5302 				    !rs->rules[ioe->rs_num].inactive.open ||
5303 				    rs->rules[ioe->rs_num].inactive.ticket !=
5304 				    ioe->ticket) {
5305 					PF_RULES_WUNLOCK();
5306 					free(ioes, M_TEMP);
5307 					error = EBUSY;
5308 					goto fail;
5309 				}
5310 				break;
5311 			}
5312 		}
5313 		/* Now do the commit - no errors should happen here. */
5314 		for (i = 0, ioe = ioes; i < io->size; i++, ioe++) {
5315 			switch (ioe->rs_num) {
5316 			case PF_RULESET_ETH:
5317 				if ((error = pf_commit_eth(ioe->ticket, ioe->anchor))) {
5318 					PF_RULES_WUNLOCK();
5319 					free(ioes, M_TEMP);
5320 					goto fail; /* really bad */
5321 				}
5322 				break;
5323 #ifdef ALTQ
5324 			case PF_RULESET_ALTQ:
5325 				if ((error = pf_commit_altq(ioe->ticket))) {
5326 					PF_RULES_WUNLOCK();
5327 					free(ioes, M_TEMP);
5328 					goto fail; /* really bad */
5329 				}
5330 				break;
5331 #endif /* ALTQ */
5332 			case PF_RULESET_TABLE:
5333 			    {
5334 				struct pfr_table table;
5335 
5336 				bzero(&table, sizeof(table));
5337 				(void)strlcpy(table.pfrt_anchor, ioe->anchor,
5338 				    sizeof(table.pfrt_anchor));
5339 				if ((error = pfr_ina_commit(&table,
5340 				    ioe->ticket, NULL, NULL, 0))) {
5341 					PF_RULES_WUNLOCK();
5342 					free(ioes, M_TEMP);
5343 					goto fail; /* really bad */
5344 				}
5345 				break;
5346 			    }
5347 			default:
5348 				if ((error = pf_commit_rules(ioe->ticket,
5349 				    ioe->rs_num, ioe->anchor))) {
5350 					PF_RULES_WUNLOCK();
5351 					free(ioes, M_TEMP);
5352 					goto fail; /* really bad */
5353 				}
5354 				break;
5355 			}
5356 		}
5357 		PF_RULES_WUNLOCK();
5358 
5359 		/* Only hook into EtherNet taffic if we've got rules for it. */
5360 		if (! TAILQ_EMPTY(V_pf_keth->active.rules))
5361 			hook_pf_eth();
5362 		else
5363 			dehook_pf_eth();
5364 
5365 		free(ioes, M_TEMP);
5366 		break;
5367 	}
5368 
5369 	case DIOCGETSRCNODES: {
5370 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
5371 		struct pf_srchash	*sh;
5372 		struct pf_ksrc_node	*n;
5373 		struct pf_src_node	*p, *pstore;
5374 		uint32_t		 i, nr = 0;
5375 
5376 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5377 				i++, sh++) {
5378 			PF_HASHROW_LOCK(sh);
5379 			LIST_FOREACH(n, &sh->nodes, entry)
5380 				nr++;
5381 			PF_HASHROW_UNLOCK(sh);
5382 		}
5383 
5384 		psn->psn_len = min(psn->psn_len,
5385 		    sizeof(struct pf_src_node) * nr);
5386 
5387 		if (psn->psn_len == 0) {
5388 			psn->psn_len = sizeof(struct pf_src_node) * nr;
5389 			break;
5390 		}
5391 
5392 		nr = 0;
5393 
5394 		p = pstore = malloc(psn->psn_len, M_TEMP, M_WAITOK | M_ZERO);
5395 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5396 		    i++, sh++) {
5397 		    PF_HASHROW_LOCK(sh);
5398 		    LIST_FOREACH(n, &sh->nodes, entry) {
5399 
5400 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
5401 				break;
5402 
5403 			pf_src_node_copy(n, p);
5404 
5405 			p++;
5406 			nr++;
5407 		    }
5408 		    PF_HASHROW_UNLOCK(sh);
5409 		}
5410 		error = copyout(pstore, psn->psn_src_nodes,
5411 		    sizeof(struct pf_src_node) * nr);
5412 		if (error) {
5413 			free(pstore, M_TEMP);
5414 			break;
5415 		}
5416 		psn->psn_len = sizeof(struct pf_src_node) * nr;
5417 		free(pstore, M_TEMP);
5418 		break;
5419 	}
5420 
5421 	case DIOCCLRSRCNODES: {
5422 		pf_clear_srcnodes(NULL);
5423 		pf_purge_expired_src_nodes();
5424 		break;
5425 	}
5426 
5427 	case DIOCKILLSRCNODES:
5428 		pf_kill_srcnodes((struct pfioc_src_node_kill *)addr);
5429 		break;
5430 
5431 #ifdef COMPAT_FREEBSD13
5432 	case DIOCKEEPCOUNTERS_FREEBSD13:
5433 #endif
5434 	case DIOCKEEPCOUNTERS:
5435 		error = pf_keepcounters((struct pfioc_nv *)addr);
5436 		break;
5437 
5438 	case DIOCGETSYNCOOKIES:
5439 		error = pf_get_syncookies((struct pfioc_nv *)addr);
5440 		break;
5441 
5442 	case DIOCSETSYNCOOKIES:
5443 		error = pf_set_syncookies((struct pfioc_nv *)addr);
5444 		break;
5445 
5446 	case DIOCSETHOSTID: {
5447 		u_int32_t	*hostid = (u_int32_t *)addr;
5448 
5449 		PF_RULES_WLOCK();
5450 		if (*hostid == 0)
5451 			V_pf_status.hostid = arc4random();
5452 		else
5453 			V_pf_status.hostid = *hostid;
5454 		PF_RULES_WUNLOCK();
5455 		break;
5456 	}
5457 
5458 	case DIOCOSFPFLUSH:
5459 		PF_RULES_WLOCK();
5460 		pf_osfp_flush();
5461 		PF_RULES_WUNLOCK();
5462 		break;
5463 
5464 	case DIOCIGETIFACES: {
5465 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5466 		struct pfi_kif *ifstore;
5467 		size_t bufsiz;
5468 
5469 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
5470 			error = ENODEV;
5471 			break;
5472 		}
5473 
5474 		if (io->pfiio_size < 0 ||
5475 		    io->pfiio_size > pf_ioctl_maxcount ||
5476 		    WOULD_OVERFLOW(io->pfiio_size, sizeof(struct pfi_kif))) {
5477 			error = EINVAL;
5478 			break;
5479 		}
5480 
5481 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5482 
5483 		bufsiz = io->pfiio_size * sizeof(struct pfi_kif);
5484 		ifstore = mallocarray(io->pfiio_size, sizeof(struct pfi_kif),
5485 		    M_TEMP, M_WAITOK | M_ZERO);
5486 
5487 		PF_RULES_RLOCK();
5488 		pfi_get_ifaces(io->pfiio_name, ifstore, &io->pfiio_size);
5489 		PF_RULES_RUNLOCK();
5490 		error = copyout(ifstore, io->pfiio_buffer, bufsiz);
5491 		free(ifstore, M_TEMP);
5492 		break;
5493 	}
5494 
5495 	case DIOCSETIFFLAG: {
5496 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5497 
5498 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5499 
5500 		PF_RULES_WLOCK();
5501 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
5502 		PF_RULES_WUNLOCK();
5503 		break;
5504 	}
5505 
5506 	case DIOCCLRIFFLAG: {
5507 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
5508 
5509 		io->pfiio_name[sizeof(io->pfiio_name) - 1] = '\0';
5510 
5511 		PF_RULES_WLOCK();
5512 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
5513 		PF_RULES_WUNLOCK();
5514 		break;
5515 	}
5516 
5517 	case DIOCSETREASS: {
5518 		u_int32_t	*reass = (u_int32_t *)addr;
5519 
5520 		V_pf_status.reass = *reass & (PF_REASS_ENABLED|PF_REASS_NODF);
5521 		/* Removal of DF flag without reassembly enabled is not a
5522 		 * valid combination. Disable reassembly in such case. */
5523 		if (!(V_pf_status.reass & PF_REASS_ENABLED))
5524 			V_pf_status.reass = 0;
5525 		break;
5526 	}
5527 
5528 	default:
5529 		error = ENODEV;
5530 		break;
5531 	}
5532 fail:
5533 	CURVNET_RESTORE();
5534 
5535 #undef ERROUT_IOCTL
5536 
5537 	return (error);
5538 }
5539 
5540 void
5541 pfsync_state_export(union pfsync_state_union *sp, struct pf_kstate *st, int msg_version)
5542 {
5543 	bzero(sp, sizeof(union pfsync_state_union));
5544 
5545 	/* copy from state key */
5546 	sp->pfs_1301.key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5547 	sp->pfs_1301.key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5548 	sp->pfs_1301.key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5549 	sp->pfs_1301.key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5550 	sp->pfs_1301.key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5551 	sp->pfs_1301.key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5552 	sp->pfs_1301.key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5553 	sp->pfs_1301.key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5554 	sp->pfs_1301.proto = st->key[PF_SK_WIRE]->proto;
5555 	sp->pfs_1301.af = st->key[PF_SK_WIRE]->af;
5556 
5557 	/* copy from state */
5558 	strlcpy(sp->pfs_1301.ifname, st->kif->pfik_name, sizeof(sp->pfs_1301.ifname));
5559 	bcopy(&st->rt_addr, &sp->pfs_1301.rt_addr, sizeof(sp->pfs_1301.rt_addr));
5560 	sp->pfs_1301.creation = htonl(time_uptime - (st->creation / 1000));
5561 	sp->pfs_1301.expire = pf_state_expires(st);
5562 	if (sp->pfs_1301.expire <= time_uptime)
5563 		sp->pfs_1301.expire = htonl(0);
5564 	else
5565 		sp->pfs_1301.expire = htonl(sp->pfs_1301.expire - time_uptime);
5566 
5567 	sp->pfs_1301.direction = st->direction;
5568 	sp->pfs_1301.log = st->act.log;
5569 	sp->pfs_1301.timeout = st->timeout;
5570 
5571 	switch (msg_version) {
5572 		case PFSYNC_MSG_VERSION_1301:
5573 			sp->pfs_1301.state_flags = st->state_flags;
5574 			break;
5575 		case PFSYNC_MSG_VERSION_1400:
5576 			sp->pfs_1400.state_flags = htons(st->state_flags);
5577 			sp->pfs_1400.qid = htons(st->act.qid);
5578 			sp->pfs_1400.pqid = htons(st->act.pqid);
5579 			sp->pfs_1400.dnpipe = htons(st->act.dnpipe);
5580 			sp->pfs_1400.dnrpipe = htons(st->act.dnrpipe);
5581 			sp->pfs_1400.rtableid = htonl(st->act.rtableid);
5582 			sp->pfs_1400.min_ttl = st->act.min_ttl;
5583 			sp->pfs_1400.set_tos = st->act.set_tos;
5584 			sp->pfs_1400.max_mss = htons(st->act.max_mss);
5585 			sp->pfs_1400.set_prio[0] = st->act.set_prio[0];
5586 			sp->pfs_1400.set_prio[1] = st->act.set_prio[1];
5587 			sp->pfs_1400.rt = st->rt;
5588 			if (st->rt_kif)
5589 				strlcpy(sp->pfs_1400.rt_ifname,
5590 				    st->rt_kif->pfik_name,
5591 				    sizeof(sp->pfs_1400.rt_ifname));
5592 			break;
5593 		default:
5594 			panic("%s: Unsupported pfsync_msg_version %d",
5595 			    __func__, msg_version);
5596 	}
5597 
5598 	if (st->src_node)
5599 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_SRCNODE;
5600 	if (st->nat_src_node)
5601 		sp->pfs_1301.sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5602 
5603 	sp->pfs_1301.id = st->id;
5604 	sp->pfs_1301.creatorid = st->creatorid;
5605 	pf_state_peer_hton(&st->src, &sp->pfs_1301.src);
5606 	pf_state_peer_hton(&st->dst, &sp->pfs_1301.dst);
5607 
5608 	if (st->rule.ptr == NULL)
5609 		sp->pfs_1301.rule = htonl(-1);
5610 	else
5611 		sp->pfs_1301.rule = htonl(st->rule.ptr->nr);
5612 	if (st->anchor.ptr == NULL)
5613 		sp->pfs_1301.anchor = htonl(-1);
5614 	else
5615 		sp->pfs_1301.anchor = htonl(st->anchor.ptr->nr);
5616 	if (st->nat_rule.ptr == NULL)
5617 		sp->pfs_1301.nat_rule = htonl(-1);
5618 	else
5619 		sp->pfs_1301.nat_rule = htonl(st->nat_rule.ptr->nr);
5620 
5621 	pf_state_counter_hton(st->packets[0], sp->pfs_1301.packets[0]);
5622 	pf_state_counter_hton(st->packets[1], sp->pfs_1301.packets[1]);
5623 	pf_state_counter_hton(st->bytes[0], sp->pfs_1301.bytes[0]);
5624 	pf_state_counter_hton(st->bytes[1], sp->pfs_1301.bytes[1]);
5625 }
5626 
5627 void
5628 pf_state_export(struct pf_state_export *sp, struct pf_kstate *st)
5629 {
5630 	bzero(sp, sizeof(*sp));
5631 
5632 	sp->version = PF_STATE_VERSION;
5633 
5634 	/* copy from state key */
5635 	sp->key[PF_SK_WIRE].addr[0] = st->key[PF_SK_WIRE]->addr[0];
5636 	sp->key[PF_SK_WIRE].addr[1] = st->key[PF_SK_WIRE]->addr[1];
5637 	sp->key[PF_SK_WIRE].port[0] = st->key[PF_SK_WIRE]->port[0];
5638 	sp->key[PF_SK_WIRE].port[1] = st->key[PF_SK_WIRE]->port[1];
5639 	sp->key[PF_SK_STACK].addr[0] = st->key[PF_SK_STACK]->addr[0];
5640 	sp->key[PF_SK_STACK].addr[1] = st->key[PF_SK_STACK]->addr[1];
5641 	sp->key[PF_SK_STACK].port[0] = st->key[PF_SK_STACK]->port[0];
5642 	sp->key[PF_SK_STACK].port[1] = st->key[PF_SK_STACK]->port[1];
5643 	sp->proto = st->key[PF_SK_WIRE]->proto;
5644 	sp->af = st->key[PF_SK_WIRE]->af;
5645 
5646 	/* copy from state */
5647 	strlcpy(sp->ifname, st->kif->pfik_name, sizeof(sp->ifname));
5648 	strlcpy(sp->orig_ifname, st->orig_kif->pfik_name,
5649 	    sizeof(sp->orig_ifname));
5650 	bcopy(&st->rt_addr, &sp->rt_addr, sizeof(sp->rt_addr));
5651 	sp->creation = htonl(time_uptime - (st->creation / 1000));
5652 	sp->expire = pf_state_expires(st);
5653 	if (sp->expire <= time_uptime)
5654 		sp->expire = htonl(0);
5655 	else
5656 		sp->expire = htonl(sp->expire - time_uptime);
5657 
5658 	sp->direction = st->direction;
5659 	sp->log = st->act.log;
5660 	sp->timeout = st->timeout;
5661 	/* 8 bits for the old libpfctl, 16 bits for the new libpfctl */
5662 	sp->state_flags_compat = st->state_flags;
5663 	sp->state_flags = htons(st->state_flags);
5664 	if (st->src_node)
5665 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
5666 	if (st->nat_src_node)
5667 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
5668 
5669 	sp->id = st->id;
5670 	sp->creatorid = st->creatorid;
5671 	pf_state_peer_hton(&st->src, &sp->src);
5672 	pf_state_peer_hton(&st->dst, &sp->dst);
5673 
5674 	if (st->rule.ptr == NULL)
5675 		sp->rule = htonl(-1);
5676 	else
5677 		sp->rule = htonl(st->rule.ptr->nr);
5678 	if (st->anchor.ptr == NULL)
5679 		sp->anchor = htonl(-1);
5680 	else
5681 		sp->anchor = htonl(st->anchor.ptr->nr);
5682 	if (st->nat_rule.ptr == NULL)
5683 		sp->nat_rule = htonl(-1);
5684 	else
5685 		sp->nat_rule = htonl(st->nat_rule.ptr->nr);
5686 
5687 	sp->packets[0] = st->packets[0];
5688 	sp->packets[1] = st->packets[1];
5689 	sp->bytes[0] = st->bytes[0];
5690 	sp->bytes[1] = st->bytes[1];
5691 
5692 	sp->qid = htons(st->act.qid);
5693 	sp->pqid = htons(st->act.pqid);
5694 	sp->dnpipe = htons(st->act.dnpipe);
5695 	sp->dnrpipe = htons(st->act.dnrpipe);
5696 	sp->rtableid = htonl(st->act.rtableid);
5697 	sp->min_ttl = st->act.min_ttl;
5698 	sp->set_tos = st->act.set_tos;
5699 	sp->max_mss = htons(st->act.max_mss);
5700 	sp->rt = st->rt;
5701 	if (st->rt_kif)
5702 		strlcpy(sp->rt_ifname, st->rt_kif->pfik_name,
5703 		    sizeof(sp->rt_ifname));
5704 	sp->set_prio[0] = st->act.set_prio[0];
5705 	sp->set_prio[1] = st->act.set_prio[1];
5706 
5707 }
5708 
5709 static void
5710 pf_tbladdr_copyout(struct pf_addr_wrap *aw)
5711 {
5712 	struct pfr_ktable *kt;
5713 
5714 	KASSERT(aw->type == PF_ADDR_TABLE, ("%s: type %u", __func__, aw->type));
5715 
5716 	kt = aw->p.tbl;
5717 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
5718 		kt = kt->pfrkt_root;
5719 	aw->p.tbl = NULL;
5720 	aw->p.tblcnt = (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) ?
5721 		kt->pfrkt_cnt : -1;
5722 }
5723 
5724 static int
5725 pf_add_status_counters(nvlist_t *nvl, const char *name, counter_u64_t *counters,
5726     size_t number, char **names)
5727 {
5728 	nvlist_t        *nvc;
5729 
5730 	nvc = nvlist_create(0);
5731 	if (nvc == NULL)
5732 		return (ENOMEM);
5733 
5734 	for (int i = 0; i < number; i++) {
5735 		nvlist_append_number_array(nvc, "counters",
5736 		    counter_u64_fetch(counters[i]));
5737 		nvlist_append_string_array(nvc, "names",
5738 		    names[i]);
5739 		nvlist_append_number_array(nvc, "ids",
5740 		    i);
5741 	}
5742 	nvlist_add_nvlist(nvl, name, nvc);
5743 	nvlist_destroy(nvc);
5744 
5745 	return (0);
5746 }
5747 
5748 static int
5749 pf_getstatus(struct pfioc_nv *nv)
5750 {
5751 	nvlist_t        *nvl = NULL, *nvc = NULL;
5752 	void            *nvlpacked = NULL;
5753 	int              error;
5754 	struct pf_status s;
5755 	char *pf_reasons[PFRES_MAX+1] = PFRES_NAMES;
5756 	char *pf_lcounter[KLCNT_MAX+1] = KLCNT_NAMES;
5757 	char *pf_fcounter[FCNT_MAX+1] = FCNT_NAMES;
5758 	PF_RULES_RLOCK_TRACKER;
5759 
5760 #define ERROUT(x)      ERROUT_FUNCTION(errout, x)
5761 
5762 	PF_RULES_RLOCK();
5763 
5764 	nvl = nvlist_create(0);
5765 	if (nvl == NULL)
5766 		ERROUT(ENOMEM);
5767 
5768 	nvlist_add_bool(nvl, "running", V_pf_status.running);
5769 	nvlist_add_number(nvl, "since", V_pf_status.since);
5770 	nvlist_add_number(nvl, "debug", V_pf_status.debug);
5771 	nvlist_add_number(nvl, "hostid", V_pf_status.hostid);
5772 	nvlist_add_number(nvl, "states", V_pf_status.states);
5773 	nvlist_add_number(nvl, "src_nodes", V_pf_status.src_nodes);
5774 	nvlist_add_number(nvl, "reass", V_pf_status.reass);
5775 	nvlist_add_bool(nvl, "syncookies_active",
5776 	    V_pf_status.syncookies_active);
5777 	nvlist_add_number(nvl, "halfopen_states", V_pf_status.states_halfopen);
5778 
5779 	/* counters */
5780 	error = pf_add_status_counters(nvl, "counters", V_pf_status.counters,
5781 	    PFRES_MAX, pf_reasons);
5782 	if (error != 0)
5783 		ERROUT(error);
5784 
5785 	/* lcounters */
5786 	error = pf_add_status_counters(nvl, "lcounters", V_pf_status.lcounters,
5787 	    KLCNT_MAX, pf_lcounter);
5788 	if (error != 0)
5789 		ERROUT(error);
5790 
5791 	/* fcounters */
5792 	nvc = nvlist_create(0);
5793 	if (nvc == NULL)
5794 		ERROUT(ENOMEM);
5795 
5796 	for (int i = 0; i < FCNT_MAX; i++) {
5797 		nvlist_append_number_array(nvc, "counters",
5798 		    pf_counter_u64_fetch(&V_pf_status.fcounters[i]));
5799 		nvlist_append_string_array(nvc, "names",
5800 		    pf_fcounter[i]);
5801 		nvlist_append_number_array(nvc, "ids",
5802 		    i);
5803 	}
5804 	nvlist_add_nvlist(nvl, "fcounters", nvc);
5805 	nvlist_destroy(nvc);
5806 	nvc = NULL;
5807 
5808 	/* scounters */
5809 	error = pf_add_status_counters(nvl, "scounters", V_pf_status.scounters,
5810 	    SCNT_MAX, pf_fcounter);
5811 	if (error != 0)
5812 		ERROUT(error);
5813 
5814 	nvlist_add_string(nvl, "ifname", V_pf_status.ifname);
5815 	nvlist_add_binary(nvl, "chksum", V_pf_status.pf_chksum,
5816 	    PF_MD5_DIGEST_LENGTH);
5817 
5818 	pfi_update_status(V_pf_status.ifname, &s);
5819 
5820 	/* pcounters / bcounters */
5821 	for (int i = 0; i < 2; i++) {
5822 		for (int j = 0; j < 2; j++) {
5823 			for (int k = 0; k < 2; k++) {
5824 				nvlist_append_number_array(nvl, "pcounters",
5825 				    s.pcounters[i][j][k]);
5826 			}
5827 			nvlist_append_number_array(nvl, "bcounters",
5828 			    s.bcounters[i][j]);
5829 		}
5830 	}
5831 
5832 	nvlpacked = nvlist_pack(nvl, &nv->len);
5833 	if (nvlpacked == NULL)
5834 		ERROUT(ENOMEM);
5835 
5836 	if (nv->size == 0)
5837 		ERROUT(0);
5838 	else if (nv->size < nv->len)
5839 		ERROUT(ENOSPC);
5840 
5841 	PF_RULES_RUNLOCK();
5842 	error = copyout(nvlpacked, nv->data, nv->len);
5843 	goto done;
5844 
5845 #undef ERROUT
5846 errout:
5847 	PF_RULES_RUNLOCK();
5848 done:
5849 	free(nvlpacked, M_NVLIST);
5850 	nvlist_destroy(nvc);
5851 	nvlist_destroy(nvl);
5852 
5853 	return (error);
5854 }
5855 
5856 /*
5857  * XXX - Check for version mismatch!!!
5858  */
5859 static void
5860 pf_clear_all_states(void)
5861 {
5862 	struct epoch_tracker	 et;
5863 	struct pf_kstate	*s;
5864 	u_int i;
5865 
5866 	NET_EPOCH_ENTER(et);
5867 	for (i = 0; i <= V_pf_hashmask; i++) {
5868 		struct pf_idhash *ih = &V_pf_idhash[i];
5869 relock:
5870 		PF_HASHROW_LOCK(ih);
5871 		LIST_FOREACH(s, &ih->states, entry) {
5872 			s->timeout = PFTM_PURGE;
5873 			/* Don't send out individual delete messages. */
5874 			s->state_flags |= PFSTATE_NOSYNC;
5875 			pf_unlink_state(s);
5876 			goto relock;
5877 		}
5878 		PF_HASHROW_UNLOCK(ih);
5879 	}
5880 	NET_EPOCH_EXIT(et);
5881 }
5882 
5883 static int
5884 pf_clear_tables(void)
5885 {
5886 	struct pfioc_table io;
5887 	int error;
5888 
5889 	bzero(&io, sizeof(io));
5890 	io.pfrio_flags |= PFR_FLAG_ALLRSETS;
5891 
5892 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
5893 	    io.pfrio_flags);
5894 
5895 	return (error);
5896 }
5897 
5898 static void
5899 pf_clear_srcnodes(struct pf_ksrc_node *n)
5900 {
5901 	struct pf_kstate *s;
5902 	int i;
5903 
5904 	for (i = 0; i <= V_pf_hashmask; i++) {
5905 		struct pf_idhash *ih = &V_pf_idhash[i];
5906 
5907 		PF_HASHROW_LOCK(ih);
5908 		LIST_FOREACH(s, &ih->states, entry) {
5909 			if (n == NULL || n == s->src_node)
5910 				s->src_node = NULL;
5911 			if (n == NULL || n == s->nat_src_node)
5912 				s->nat_src_node = NULL;
5913 		}
5914 		PF_HASHROW_UNLOCK(ih);
5915 	}
5916 
5917 	if (n == NULL) {
5918 		struct pf_srchash *sh;
5919 
5920 		for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask;
5921 		    i++, sh++) {
5922 			PF_HASHROW_LOCK(sh);
5923 			LIST_FOREACH(n, &sh->nodes, entry) {
5924 				n->expire = 1;
5925 				n->states = 0;
5926 			}
5927 			PF_HASHROW_UNLOCK(sh);
5928 		}
5929 	} else {
5930 		/* XXX: hash slot should already be locked here. */
5931 		n->expire = 1;
5932 		n->states = 0;
5933 	}
5934 }
5935 
5936 static void
5937 pf_kill_srcnodes(struct pfioc_src_node_kill *psnk)
5938 {
5939 	struct pf_ksrc_node_list	 kill;
5940 
5941 	LIST_INIT(&kill);
5942 	for (int i = 0; i <= V_pf_srchashmask; i++) {
5943 		struct pf_srchash *sh = &V_pf_srchash[i];
5944 		struct pf_ksrc_node *sn, *tmp;
5945 
5946 		PF_HASHROW_LOCK(sh);
5947 		LIST_FOREACH_SAFE(sn, &sh->nodes, entry, tmp)
5948 			if (PF_MATCHA(psnk->psnk_src.neg,
5949 			      &psnk->psnk_src.addr.v.a.addr,
5950 			      &psnk->psnk_src.addr.v.a.mask,
5951 			      &sn->addr, sn->af) &&
5952 			    PF_MATCHA(psnk->psnk_dst.neg,
5953 			      &psnk->psnk_dst.addr.v.a.addr,
5954 			      &psnk->psnk_dst.addr.v.a.mask,
5955 			      &sn->raddr, sn->af)) {
5956 				pf_unlink_src_node(sn);
5957 				LIST_INSERT_HEAD(&kill, sn, entry);
5958 				sn->expire = 1;
5959 			}
5960 		PF_HASHROW_UNLOCK(sh);
5961 	}
5962 
5963 	for (int i = 0; i <= V_pf_hashmask; i++) {
5964 		struct pf_idhash *ih = &V_pf_idhash[i];
5965 		struct pf_kstate *s;
5966 
5967 		PF_HASHROW_LOCK(ih);
5968 		LIST_FOREACH(s, &ih->states, entry) {
5969 			if (s->src_node && s->src_node->expire == 1)
5970 				s->src_node = NULL;
5971 			if (s->nat_src_node && s->nat_src_node->expire == 1)
5972 				s->nat_src_node = NULL;
5973 		}
5974 		PF_HASHROW_UNLOCK(ih);
5975 	}
5976 
5977 	psnk->psnk_killed = pf_free_src_nodes(&kill);
5978 }
5979 
5980 static int
5981 pf_keepcounters(struct pfioc_nv *nv)
5982 {
5983 	nvlist_t	*nvl = NULL;
5984 	void		*nvlpacked = NULL;
5985 	int		 error = 0;
5986 
5987 #define	ERROUT(x)	ERROUT_FUNCTION(on_error, x)
5988 
5989 	if (nv->len > pf_ioctl_maxcount)
5990 		ERROUT(ENOMEM);
5991 
5992 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
5993 	if (nvlpacked == NULL)
5994 		ERROUT(ENOMEM);
5995 
5996 	error = copyin(nv->data, nvlpacked, nv->len);
5997 	if (error)
5998 		ERROUT(error);
5999 
6000 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6001 	if (nvl == NULL)
6002 		ERROUT(EBADMSG);
6003 
6004 	if (! nvlist_exists_bool(nvl, "keep_counters"))
6005 		ERROUT(EBADMSG);
6006 
6007 	V_pf_status.keep_counters = nvlist_get_bool(nvl, "keep_counters");
6008 
6009 on_error:
6010 	nvlist_destroy(nvl);
6011 	free(nvlpacked, M_NVLIST);
6012 	return (error);
6013 }
6014 
6015 unsigned int
6016 pf_clear_states(const struct pf_kstate_kill *kill)
6017 {
6018 	struct pf_state_key_cmp	 match_key;
6019 	struct pf_kstate	*s;
6020 	struct pfi_kkif	*kif;
6021 	int		 idx;
6022 	unsigned int	 killed = 0, dir;
6023 
6024 	NET_EPOCH_ASSERT();
6025 
6026 	for (unsigned int i = 0; i <= V_pf_hashmask; i++) {
6027 		struct pf_idhash *ih = &V_pf_idhash[i];
6028 
6029 relock_DIOCCLRSTATES:
6030 		PF_HASHROW_LOCK(ih);
6031 		LIST_FOREACH(s, &ih->states, entry) {
6032 			/* For floating states look at the original kif. */
6033 			kif = s->kif == V_pfi_all ? s->orig_kif : s->kif;
6034 
6035 			if (kill->psk_ifname[0] &&
6036 			    strcmp(kill->psk_ifname,
6037 			    kif->pfik_name))
6038 				continue;
6039 
6040 			if (kill->psk_kill_match) {
6041 				bzero(&match_key, sizeof(match_key));
6042 
6043 				if (s->direction == PF_OUT) {
6044 					dir = PF_IN;
6045 					idx = PF_SK_STACK;
6046 				} else {
6047 					dir = PF_OUT;
6048 					idx = PF_SK_WIRE;
6049 				}
6050 
6051 				match_key.af = s->key[idx]->af;
6052 				match_key.proto = s->key[idx]->proto;
6053 				PF_ACPY(&match_key.addr[0],
6054 				    &s->key[idx]->addr[1], match_key.af);
6055 				match_key.port[0] = s->key[idx]->port[1];
6056 				PF_ACPY(&match_key.addr[1],
6057 				    &s->key[idx]->addr[0], match_key.af);
6058 				match_key.port[1] = s->key[idx]->port[0];
6059 			}
6060 
6061 			/*
6062 			 * Don't send out individual
6063 			 * delete messages.
6064 			 */
6065 			s->state_flags |= PFSTATE_NOSYNC;
6066 			pf_unlink_state(s);
6067 			killed++;
6068 
6069 			if (kill->psk_kill_match)
6070 				killed += pf_kill_matching_state(&match_key,
6071 				    dir);
6072 
6073 			goto relock_DIOCCLRSTATES;
6074 		}
6075 		PF_HASHROW_UNLOCK(ih);
6076 	}
6077 
6078 	if (V_pfsync_clear_states_ptr != NULL)
6079 		V_pfsync_clear_states_ptr(V_pf_status.hostid, kill->psk_ifname);
6080 
6081 	return (killed);
6082 }
6083 
6084 void
6085 pf_killstates(struct pf_kstate_kill *kill, unsigned int *killed)
6086 {
6087 	struct pf_kstate	*s;
6088 
6089 	NET_EPOCH_ASSERT();
6090 	if (kill->psk_pfcmp.id) {
6091 		if (kill->psk_pfcmp.creatorid == 0)
6092 			kill->psk_pfcmp.creatorid = V_pf_status.hostid;
6093 		if ((s = pf_find_state_byid(kill->psk_pfcmp.id,
6094 		    kill->psk_pfcmp.creatorid))) {
6095 			pf_unlink_state(s);
6096 			*killed = 1;
6097 		}
6098 		return;
6099 	}
6100 
6101 	for (unsigned int i = 0; i <= V_pf_hashmask; i++)
6102 		*killed += pf_killstates_row(kill, &V_pf_idhash[i]);
6103 }
6104 
6105 static int
6106 pf_killstates_nv(struct pfioc_nv *nv)
6107 {
6108 	struct pf_kstate_kill	 kill;
6109 	struct epoch_tracker	 et;
6110 	nvlist_t		*nvl = NULL;
6111 	void			*nvlpacked = NULL;
6112 	int			 error = 0;
6113 	unsigned int		 killed = 0;
6114 
6115 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6116 
6117 	if (nv->len > pf_ioctl_maxcount)
6118 		ERROUT(ENOMEM);
6119 
6120 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6121 	if (nvlpacked == NULL)
6122 		ERROUT(ENOMEM);
6123 
6124 	error = copyin(nv->data, nvlpacked, nv->len);
6125 	if (error)
6126 		ERROUT(error);
6127 
6128 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6129 	if (nvl == NULL)
6130 		ERROUT(EBADMSG);
6131 
6132 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6133 	if (error)
6134 		ERROUT(error);
6135 
6136 	NET_EPOCH_ENTER(et);
6137 	pf_killstates(&kill, &killed);
6138 	NET_EPOCH_EXIT(et);
6139 
6140 	free(nvlpacked, M_NVLIST);
6141 	nvlpacked = NULL;
6142 	nvlist_destroy(nvl);
6143 	nvl = nvlist_create(0);
6144 	if (nvl == NULL)
6145 		ERROUT(ENOMEM);
6146 
6147 	nvlist_add_number(nvl, "killed", killed);
6148 
6149 	nvlpacked = nvlist_pack(nvl, &nv->len);
6150 	if (nvlpacked == NULL)
6151 		ERROUT(ENOMEM);
6152 
6153 	if (nv->size == 0)
6154 		ERROUT(0);
6155 	else if (nv->size < nv->len)
6156 		ERROUT(ENOSPC);
6157 
6158 	error = copyout(nvlpacked, nv->data, nv->len);
6159 
6160 on_error:
6161 	nvlist_destroy(nvl);
6162 	free(nvlpacked, M_NVLIST);
6163 	return (error);
6164 }
6165 
6166 static int
6167 pf_clearstates_nv(struct pfioc_nv *nv)
6168 {
6169 	struct pf_kstate_kill	 kill;
6170 	struct epoch_tracker	 et;
6171 	nvlist_t		*nvl = NULL;
6172 	void			*nvlpacked = NULL;
6173 	int			 error = 0;
6174 	unsigned int		 killed;
6175 
6176 #define ERROUT(x)	ERROUT_FUNCTION(on_error, x)
6177 
6178 	if (nv->len > pf_ioctl_maxcount)
6179 		ERROUT(ENOMEM);
6180 
6181 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6182 	if (nvlpacked == NULL)
6183 		ERROUT(ENOMEM);
6184 
6185 	error = copyin(nv->data, nvlpacked, nv->len);
6186 	if (error)
6187 		ERROUT(error);
6188 
6189 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6190 	if (nvl == NULL)
6191 		ERROUT(EBADMSG);
6192 
6193 	error = pf_nvstate_kill_to_kstate_kill(nvl, &kill);
6194 	if (error)
6195 		ERROUT(error);
6196 
6197 	NET_EPOCH_ENTER(et);
6198 	killed = pf_clear_states(&kill);
6199 	NET_EPOCH_EXIT(et);
6200 
6201 	free(nvlpacked, M_NVLIST);
6202 	nvlpacked = NULL;
6203 	nvlist_destroy(nvl);
6204 	nvl = nvlist_create(0);
6205 	if (nvl == NULL)
6206 		ERROUT(ENOMEM);
6207 
6208 	nvlist_add_number(nvl, "killed", killed);
6209 
6210 	nvlpacked = nvlist_pack(nvl, &nv->len);
6211 	if (nvlpacked == NULL)
6212 		ERROUT(ENOMEM);
6213 
6214 	if (nv->size == 0)
6215 		ERROUT(0);
6216 	else if (nv->size < nv->len)
6217 		ERROUT(ENOSPC);
6218 
6219 	error = copyout(nvlpacked, nv->data, nv->len);
6220 
6221 #undef ERROUT
6222 on_error:
6223 	nvlist_destroy(nvl);
6224 	free(nvlpacked, M_NVLIST);
6225 	return (error);
6226 }
6227 
6228 static int
6229 pf_getstate(struct pfioc_nv *nv)
6230 {
6231 	nvlist_t		*nvl = NULL, *nvls;
6232 	void			*nvlpacked = NULL;
6233 	struct pf_kstate	*s = NULL;
6234 	int			 error = 0;
6235 	uint64_t		 id, creatorid;
6236 
6237 #define ERROUT(x)	ERROUT_FUNCTION(errout, x)
6238 
6239 	if (nv->len > pf_ioctl_maxcount)
6240 		ERROUT(ENOMEM);
6241 
6242 	nvlpacked = malloc(nv->len, M_NVLIST, M_WAITOK);
6243 	if (nvlpacked == NULL)
6244 		ERROUT(ENOMEM);
6245 
6246 	error = copyin(nv->data, nvlpacked, nv->len);
6247 	if (error)
6248 		ERROUT(error);
6249 
6250 	nvl = nvlist_unpack(nvlpacked, nv->len, 0);
6251 	if (nvl == NULL)
6252 		ERROUT(EBADMSG);
6253 
6254 	PFNV_CHK(pf_nvuint64(nvl, "id", &id));
6255 	PFNV_CHK(pf_nvuint64(nvl, "creatorid", &creatorid));
6256 
6257 	s = pf_find_state_byid(id, creatorid);
6258 	if (s == NULL)
6259 		ERROUT(ENOENT);
6260 
6261 	free(nvlpacked, M_NVLIST);
6262 	nvlpacked = NULL;
6263 	nvlist_destroy(nvl);
6264 	nvl = nvlist_create(0);
6265 	if (nvl == NULL)
6266 		ERROUT(ENOMEM);
6267 
6268 	nvls = pf_state_to_nvstate(s);
6269 	if (nvls == NULL)
6270 		ERROUT(ENOMEM);
6271 
6272 	nvlist_add_nvlist(nvl, "state", nvls);
6273 	nvlist_destroy(nvls);
6274 
6275 	nvlpacked = nvlist_pack(nvl, &nv->len);
6276 	if (nvlpacked == NULL)
6277 		ERROUT(ENOMEM);
6278 
6279 	if (nv->size == 0)
6280 		ERROUT(0);
6281 	else if (nv->size < nv->len)
6282 		ERROUT(ENOSPC);
6283 
6284 	error = copyout(nvlpacked, nv->data, nv->len);
6285 
6286 #undef ERROUT
6287 errout:
6288 	if (s != NULL)
6289 		PF_STATE_UNLOCK(s);
6290 	free(nvlpacked, M_NVLIST);
6291 	nvlist_destroy(nvl);
6292 	return (error);
6293 }
6294 
6295 /*
6296  * XXX - Check for version mismatch!!!
6297  */
6298 
6299 /*
6300  * Duplicate pfctl -Fa operation to get rid of as much as we can.
6301  */
6302 static int
6303 shutdown_pf(void)
6304 {
6305 	int error = 0;
6306 	u_int32_t t[5];
6307 	char nn = '\0';
6308 	struct pf_kanchor *anchor;
6309 	struct pf_keth_anchor *eth_anchor;
6310 	int rs_num;
6311 
6312 	do {
6313 		/* Unlink rules of all user defined anchors */
6314 		RB_FOREACH(anchor, pf_kanchor_global, &V_pf_anchors) {
6315 			/* Wildcard based anchors may not have a respective
6316 			 * explicit anchor rule or they may be left empty
6317 			 * without rules. It leads to anchor.refcnt=0, and the
6318 			 * rest of the logic does not expect it. */
6319 			if (anchor->refcnt == 0)
6320 				anchor->refcnt = 1;
6321 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6322 				if ((error = pf_begin_rules(&t[rs_num], rs_num,
6323 				    anchor->path)) != 0) {
6324 					DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: "
6325 					    "anchor.path=%s rs_num=%d\n",
6326 					    anchor->path, rs_num));
6327 					goto error;	/* XXX: rollback? */
6328 				}
6329 			}
6330 			for (rs_num = 0; rs_num < PF_RULESET_MAX; ++rs_num) {
6331 				error = pf_commit_rules(t[rs_num], rs_num,
6332 				    anchor->path);
6333 				MPASS(error == 0);
6334 			}
6335 		}
6336 
6337 		/* Unlink rules of all user defined ether anchors */
6338 		RB_FOREACH(eth_anchor, pf_keth_anchor_global,
6339 		    &V_pf_keth_anchors) {
6340 			/* Wildcard based anchors may not have a respective
6341 			 * explicit anchor rule or they may be left empty
6342 			 * without rules. It leads to anchor.refcnt=0, and the
6343 			 * rest of the logic does not expect it. */
6344 			if (eth_anchor->refcnt == 0)
6345 				eth_anchor->refcnt = 1;
6346 			if ((error = pf_begin_eth(&t[0], eth_anchor->path))
6347 			    != 0) {
6348 				DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth "
6349 				    "anchor.path=%s\n", eth_anchor->path));
6350 				goto error;
6351 			}
6352 			error = pf_commit_eth(t[0], eth_anchor->path);
6353 			MPASS(error == 0);
6354 		}
6355 
6356 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
6357 		    != 0) {
6358 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
6359 			break;
6360 		}
6361 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
6362 		    != 0) {
6363 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
6364 			break;		/* XXX: rollback? */
6365 		}
6366 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
6367 		    != 0) {
6368 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
6369 			break;		/* XXX: rollback? */
6370 		}
6371 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
6372 		    != 0) {
6373 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
6374 			break;		/* XXX: rollback? */
6375 		}
6376 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
6377 		    != 0) {
6378 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
6379 			break;		/* XXX: rollback? */
6380 		}
6381 
6382 		error = pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
6383 		MPASS(error == 0);
6384 		error = pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
6385 		MPASS(error == 0);
6386 		error = pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
6387 		MPASS(error == 0);
6388 		error = pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
6389 		MPASS(error == 0);
6390 		error = pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
6391 		MPASS(error == 0);
6392 
6393 		if ((error = pf_clear_tables()) != 0)
6394 			break;
6395 
6396 		if ((error = pf_begin_eth(&t[0], &nn)) != 0) {
6397 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: eth\n"));
6398 			break;
6399 		}
6400 		error = pf_commit_eth(t[0], &nn);
6401 		MPASS(error == 0);
6402 
6403 #ifdef ALTQ
6404 		if ((error = pf_begin_altq(&t[0])) != 0) {
6405 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
6406 			break;
6407 		}
6408 		pf_commit_altq(t[0]);
6409 #endif
6410 
6411 		pf_clear_all_states();
6412 
6413 		pf_clear_srcnodes(NULL);
6414 
6415 		/* status does not use malloced mem so no need to cleanup */
6416 		/* fingerprints and interfaces have their own cleanup code */
6417 	} while(0);
6418 
6419 error:
6420 	return (error);
6421 }
6422 
6423 static pfil_return_t
6424 pf_check_return(int chk, struct mbuf **m)
6425 {
6426 
6427 	switch (chk) {
6428 	case PF_PASS:
6429 		if (*m == NULL)
6430 			return (PFIL_CONSUMED);
6431 		else
6432 			return (PFIL_PASS);
6433 		break;
6434 	default:
6435 		if (*m != NULL) {
6436 			m_freem(*m);
6437 			*m = NULL;
6438 		}
6439 		return (PFIL_DROPPED);
6440 	}
6441 }
6442 
6443 static pfil_return_t
6444 pf_eth_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6445     void *ruleset __unused, struct inpcb *inp)
6446 {
6447 	int chk;
6448 
6449 	chk = pf_test_eth(PF_IN, flags, ifp, m, inp);
6450 
6451 	return (pf_check_return(chk, m));
6452 }
6453 
6454 static pfil_return_t
6455 pf_eth_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6456     void *ruleset __unused, struct inpcb *inp)
6457 {
6458 	int chk;
6459 
6460 	chk = pf_test_eth(PF_OUT, flags, ifp, m, inp);
6461 
6462 	return (pf_check_return(chk, m));
6463 }
6464 
6465 #ifdef INET
6466 static pfil_return_t
6467 pf_check_in(struct mbuf **m, struct ifnet *ifp, int flags,
6468     void *ruleset __unused, struct inpcb *inp)
6469 {
6470 	int chk;
6471 
6472 	chk = pf_test(PF_IN, flags, ifp, m, inp, NULL);
6473 
6474 	return (pf_check_return(chk, m));
6475 }
6476 
6477 static pfil_return_t
6478 pf_check_out(struct mbuf **m, struct ifnet *ifp, int flags,
6479     void *ruleset __unused,  struct inpcb *inp)
6480 {
6481 	int chk;
6482 
6483 	chk = pf_test(PF_OUT, flags, ifp, m, inp, NULL);
6484 
6485 	return (pf_check_return(chk, m));
6486 }
6487 #endif
6488 
6489 #ifdef INET6
6490 static pfil_return_t
6491 pf_check6_in(struct mbuf **m, struct ifnet *ifp, int flags,
6492     void *ruleset __unused,  struct inpcb *inp)
6493 {
6494 	int chk;
6495 
6496 	/*
6497 	 * In case of loopback traffic IPv6 uses the real interface in
6498 	 * order to support scoped addresses. In order to support stateful
6499 	 * filtering we have change this to lo0 as it is the case in IPv4.
6500 	 */
6501 	CURVNET_SET(ifp->if_vnet);
6502 	chk = pf_test6(PF_IN, flags, (*m)->m_flags & M_LOOP ? V_loif : ifp,
6503 	    m, inp, NULL);
6504 	CURVNET_RESTORE();
6505 
6506 	return (pf_check_return(chk, m));
6507 }
6508 
6509 static pfil_return_t
6510 pf_check6_out(struct mbuf **m, struct ifnet *ifp, int flags,
6511     void *ruleset __unused,  struct inpcb *inp)
6512 {
6513 	int chk;
6514 
6515 	CURVNET_SET(ifp->if_vnet);
6516 	chk = pf_test6(PF_OUT, flags, ifp, m, inp, NULL);
6517 	CURVNET_RESTORE();
6518 
6519 	return (pf_check_return(chk, m));
6520 }
6521 #endif /* INET6 */
6522 
6523 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_in_hook);
6524 VNET_DEFINE_STATIC(pfil_hook_t, pf_eth_out_hook);
6525 #define	V_pf_eth_in_hook	VNET(pf_eth_in_hook)
6526 #define	V_pf_eth_out_hook	VNET(pf_eth_out_hook)
6527 
6528 #ifdef INET
6529 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_in_hook);
6530 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip4_out_hook);
6531 #define	V_pf_ip4_in_hook	VNET(pf_ip4_in_hook)
6532 #define	V_pf_ip4_out_hook	VNET(pf_ip4_out_hook)
6533 #endif
6534 #ifdef INET6
6535 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_in_hook);
6536 VNET_DEFINE_STATIC(pfil_hook_t, pf_ip6_out_hook);
6537 #define	V_pf_ip6_in_hook	VNET(pf_ip6_in_hook)
6538 #define	V_pf_ip6_out_hook	VNET(pf_ip6_out_hook)
6539 #endif
6540 
6541 static void
6542 hook_pf_eth(void)
6543 {
6544 	struct pfil_hook_args pha = {
6545 		.pa_version = PFIL_VERSION,
6546 		.pa_modname = "pf",
6547 		.pa_type = PFIL_TYPE_ETHERNET,
6548 	};
6549 	struct pfil_link_args pla = {
6550 		.pa_version = PFIL_VERSION,
6551 	};
6552 	int ret __diagused;
6553 
6554 	if (atomic_load_bool(&V_pf_pfil_eth_hooked))
6555 		return;
6556 
6557 	pha.pa_mbuf_chk = pf_eth_check_in;
6558 	pha.pa_flags = PFIL_IN;
6559 	pha.pa_rulname = "eth-in";
6560 	V_pf_eth_in_hook = pfil_add_hook(&pha);
6561 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6562 	pla.pa_head = V_link_pfil_head;
6563 	pla.pa_hook = V_pf_eth_in_hook;
6564 	ret = pfil_link(&pla);
6565 	MPASS(ret == 0);
6566 	pha.pa_mbuf_chk = pf_eth_check_out;
6567 	pha.pa_flags = PFIL_OUT;
6568 	pha.pa_rulname = "eth-out";
6569 	V_pf_eth_out_hook = pfil_add_hook(&pha);
6570 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6571 	pla.pa_head = V_link_pfil_head;
6572 	pla.pa_hook = V_pf_eth_out_hook;
6573 	ret = pfil_link(&pla);
6574 	MPASS(ret == 0);
6575 
6576 	atomic_store_bool(&V_pf_pfil_eth_hooked, true);
6577 }
6578 
6579 static void
6580 hook_pf(void)
6581 {
6582 	struct pfil_hook_args pha = {
6583 		.pa_version = PFIL_VERSION,
6584 		.pa_modname = "pf",
6585 	};
6586 	struct pfil_link_args pla = {
6587 		.pa_version = PFIL_VERSION,
6588 	};
6589 	int ret __diagused;
6590 
6591 	if (atomic_load_bool(&V_pf_pfil_hooked))
6592 		return;
6593 
6594 #ifdef INET
6595 	pha.pa_type = PFIL_TYPE_IP4;
6596 	pha.pa_mbuf_chk = pf_check_in;
6597 	pha.pa_flags = PFIL_IN;
6598 	pha.pa_rulname = "default-in";
6599 	V_pf_ip4_in_hook = pfil_add_hook(&pha);
6600 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6601 	pla.pa_head = V_inet_pfil_head;
6602 	pla.pa_hook = V_pf_ip4_in_hook;
6603 	ret = pfil_link(&pla);
6604 	MPASS(ret == 0);
6605 	pha.pa_mbuf_chk = pf_check_out;
6606 	pha.pa_flags = PFIL_OUT;
6607 	pha.pa_rulname = "default-out";
6608 	V_pf_ip4_out_hook = pfil_add_hook(&pha);
6609 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6610 	pla.pa_head = V_inet_pfil_head;
6611 	pla.pa_hook = V_pf_ip4_out_hook;
6612 	ret = pfil_link(&pla);
6613 	MPASS(ret == 0);
6614 	if (V_pf_filter_local) {
6615 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6616 		pla.pa_head = V_inet_local_pfil_head;
6617 		pla.pa_hook = V_pf_ip4_out_hook;
6618 		ret = pfil_link(&pla);
6619 		MPASS(ret == 0);
6620 	}
6621 #endif
6622 #ifdef INET6
6623 	pha.pa_type = PFIL_TYPE_IP6;
6624 	pha.pa_mbuf_chk = pf_check6_in;
6625 	pha.pa_flags = PFIL_IN;
6626 	pha.pa_rulname = "default-in6";
6627 	V_pf_ip6_in_hook = pfil_add_hook(&pha);
6628 	pla.pa_flags = PFIL_IN | PFIL_HEADPTR | PFIL_HOOKPTR;
6629 	pla.pa_head = V_inet6_pfil_head;
6630 	pla.pa_hook = V_pf_ip6_in_hook;
6631 	ret = pfil_link(&pla);
6632 	MPASS(ret == 0);
6633 	pha.pa_mbuf_chk = pf_check6_out;
6634 	pha.pa_rulname = "default-out6";
6635 	pha.pa_flags = PFIL_OUT;
6636 	V_pf_ip6_out_hook = pfil_add_hook(&pha);
6637 	pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6638 	pla.pa_head = V_inet6_pfil_head;
6639 	pla.pa_hook = V_pf_ip6_out_hook;
6640 	ret = pfil_link(&pla);
6641 	MPASS(ret == 0);
6642 	if (V_pf_filter_local) {
6643 		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
6644 		pla.pa_head = V_inet6_local_pfil_head;
6645 		pla.pa_hook = V_pf_ip6_out_hook;
6646 		ret = pfil_link(&pla);
6647 		MPASS(ret == 0);
6648 	}
6649 #endif
6650 
6651 	atomic_store_bool(&V_pf_pfil_hooked, true);
6652 }
6653 
6654 static void
6655 dehook_pf_eth(void)
6656 {
6657 
6658 	if (!atomic_load_bool(&V_pf_pfil_eth_hooked))
6659 		return;
6660 
6661 	pfil_remove_hook(V_pf_eth_in_hook);
6662 	pfil_remove_hook(V_pf_eth_out_hook);
6663 
6664 	atomic_store_bool(&V_pf_pfil_eth_hooked, false);
6665 }
6666 
6667 static void
6668 dehook_pf(void)
6669 {
6670 
6671 	if (!atomic_load_bool(&V_pf_pfil_hooked))
6672 		return;
6673 
6674 #ifdef INET
6675 	pfil_remove_hook(V_pf_ip4_in_hook);
6676 	pfil_remove_hook(V_pf_ip4_out_hook);
6677 #endif
6678 #ifdef INET6
6679 	pfil_remove_hook(V_pf_ip6_in_hook);
6680 	pfil_remove_hook(V_pf_ip6_out_hook);
6681 #endif
6682 
6683 	atomic_store_bool(&V_pf_pfil_hooked, false);
6684 }
6685 
6686 static void
6687 pf_load_vnet(void)
6688 {
6689 	V_pf_tag_z = uma_zcreate("pf tags", sizeof(struct pf_tagname),
6690 	    NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
6691 
6692 	rm_init_flags(&V_pf_rules_lock, "pf rulesets", RM_RECURSE);
6693 	sx_init(&V_pf_ioctl_lock, "pf ioctl");
6694 
6695 	pf_init_tagset(&V_pf_tags, &pf_rule_tag_hashsize,
6696 	    PF_RULE_TAG_HASH_SIZE_DEFAULT);
6697 #ifdef ALTQ
6698 	pf_init_tagset(&V_pf_qids, &pf_queue_tag_hashsize,
6699 	    PF_QUEUE_TAG_HASH_SIZE_DEFAULT);
6700 #endif
6701 
6702 	V_pf_keth = &V_pf_main_keth_anchor.ruleset;
6703 
6704 	pfattach_vnet();
6705 	V_pf_vnet_active = 1;
6706 }
6707 
6708 static int
6709 pf_load(void)
6710 {
6711 	int error;
6712 
6713 	sx_init(&pf_end_lock, "pf end thread");
6714 
6715 	pf_mtag_initialize();
6716 
6717 	pf_dev = make_dev(&pf_cdevsw, 0, UID_ROOT, GID_WHEEL, 0600, PF_NAME);
6718 	if (pf_dev == NULL)
6719 		return (ENOMEM);
6720 
6721 	pf_end_threads = 0;
6722 	error = kproc_create(pf_purge_thread, NULL, &pf_purge_proc, 0, 0, "pf purge");
6723 	if (error != 0)
6724 		return (error);
6725 
6726 	pfi_initialize();
6727 
6728 	return (0);
6729 }
6730 
6731 static void
6732 pf_unload_vnet(void)
6733 {
6734 	int ret __diagused;
6735 
6736 	V_pf_vnet_active = 0;
6737 	V_pf_status.running = 0;
6738 	dehook_pf();
6739 	dehook_pf_eth();
6740 
6741 	PF_RULES_WLOCK();
6742 	pf_syncookies_cleanup();
6743 	shutdown_pf();
6744 	PF_RULES_WUNLOCK();
6745 
6746 	/* Make sure we've cleaned up ethernet rules before we continue. */
6747 	NET_EPOCH_DRAIN_CALLBACKS();
6748 
6749 	ret = swi_remove(V_pf_swi_cookie);
6750 	MPASS(ret == 0);
6751 	ret = intr_event_destroy(V_pf_swi_ie);
6752 	MPASS(ret == 0);
6753 
6754 	pf_unload_vnet_purge();
6755 
6756 	pf_normalize_cleanup();
6757 	PF_RULES_WLOCK();
6758 	pfi_cleanup_vnet();
6759 	PF_RULES_WUNLOCK();
6760 	pfr_cleanup();
6761 	pf_osfp_flush();
6762 	pf_cleanup();
6763 	if (IS_DEFAULT_VNET(curvnet))
6764 		pf_mtag_cleanup();
6765 
6766 	pf_cleanup_tagset(&V_pf_tags);
6767 #ifdef ALTQ
6768 	pf_cleanup_tagset(&V_pf_qids);
6769 #endif
6770 	uma_zdestroy(V_pf_tag_z);
6771 
6772 #ifdef PF_WANT_32_TO_64_COUNTER
6773 	PF_RULES_WLOCK();
6774 	LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist);
6775 
6776 	MPASS(LIST_EMPTY(&V_pf_allkiflist));
6777 	MPASS(V_pf_allkifcount == 0);
6778 
6779 	LIST_REMOVE(&V_pf_default_rule, allrulelist);
6780 	V_pf_allrulecount--;
6781 	LIST_REMOVE(V_pf_rulemarker, allrulelist);
6782 
6783 	MPASS(LIST_EMPTY(&V_pf_allrulelist));
6784 	MPASS(V_pf_allrulecount == 0);
6785 
6786 	PF_RULES_WUNLOCK();
6787 
6788 	free(V_pf_kifmarker, PFI_MTYPE);
6789 	free(V_pf_rulemarker, M_PFRULE);
6790 #endif
6791 
6792 	/* Free counters last as we updated them during shutdown. */
6793 	pf_counter_u64_deinit(&V_pf_default_rule.evaluations);
6794 	for (int i = 0; i < 2; i++) {
6795 		pf_counter_u64_deinit(&V_pf_default_rule.packets[i]);
6796 		pf_counter_u64_deinit(&V_pf_default_rule.bytes[i]);
6797 	}
6798 	counter_u64_free(V_pf_default_rule.states_cur);
6799 	counter_u64_free(V_pf_default_rule.states_tot);
6800 	counter_u64_free(V_pf_default_rule.src_nodes);
6801 	uma_zfree_pcpu(pf_timestamp_pcpu_zone, V_pf_default_rule.timestamp);
6802 
6803 	for (int i = 0; i < PFRES_MAX; i++)
6804 		counter_u64_free(V_pf_status.counters[i]);
6805 	for (int i = 0; i < KLCNT_MAX; i++)
6806 		counter_u64_free(V_pf_status.lcounters[i]);
6807 	for (int i = 0; i < FCNT_MAX; i++)
6808 		pf_counter_u64_deinit(&V_pf_status.fcounters[i]);
6809 	for (int i = 0; i < SCNT_MAX; i++)
6810 		counter_u64_free(V_pf_status.scounters[i]);
6811 
6812 	rm_destroy(&V_pf_rules_lock);
6813 	sx_destroy(&V_pf_ioctl_lock);
6814 }
6815 
6816 static void
6817 pf_unload(void)
6818 {
6819 
6820 	sx_xlock(&pf_end_lock);
6821 	pf_end_threads = 1;
6822 	while (pf_end_threads < 2) {
6823 		wakeup_one(pf_purge_thread);
6824 		sx_sleep(pf_purge_proc, &pf_end_lock, 0, "pftmo", 0);
6825 	}
6826 	sx_xunlock(&pf_end_lock);
6827 
6828 	pf_nl_unregister();
6829 
6830 	if (pf_dev != NULL)
6831 		destroy_dev(pf_dev);
6832 
6833 	pfi_cleanup();
6834 
6835 	sx_destroy(&pf_end_lock);
6836 }
6837 
6838 static void
6839 vnet_pf_init(void *unused __unused)
6840 {
6841 
6842 	pf_load_vnet();
6843 }
6844 VNET_SYSINIT(vnet_pf_init, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6845     vnet_pf_init, NULL);
6846 
6847 static void
6848 vnet_pf_uninit(const void *unused __unused)
6849 {
6850 
6851 	pf_unload_vnet();
6852 }
6853 SYSUNINIT(pf_unload, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND, pf_unload, NULL);
6854 VNET_SYSUNINIT(vnet_pf_uninit, SI_SUB_PROTO_FIREWALL, SI_ORDER_THIRD,
6855     vnet_pf_uninit, NULL);
6856 
6857 static int
6858 pf_modevent(module_t mod, int type, void *data)
6859 {
6860 	int error = 0;
6861 
6862 	switch(type) {
6863 	case MOD_LOAD:
6864 		error = pf_load();
6865 		pf_nl_register();
6866 		break;
6867 	case MOD_UNLOAD:
6868 		/* Handled in SYSUNINIT(pf_unload) to ensure it's done after
6869 		 * the vnet_pf_uninit()s */
6870 		break;
6871 	default:
6872 		error = EINVAL;
6873 		break;
6874 	}
6875 
6876 	return (error);
6877 }
6878 
6879 static moduledata_t pf_mod = {
6880 	"pf",
6881 	pf_modevent,
6882 	0
6883 };
6884 
6885 DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_FIREWALL, SI_ORDER_SECOND);
6886 MODULE_DEPEND(pf, netlink, 1, 1, 1);
6887 MODULE_VERSION(pf, PF_MODVER);
6888