xref: /openbsd-src/sys/net/pf_ioctl.c (revision c1a45aed656e7d5627c30c92421893a76f370ccb)
1 /*	$OpenBSD: pf_ioctl.c,v 1.379 2022/04/09 13:15:44 mbuhl Exp $ */
2 
3 /*
4  * Copyright (c) 2001 Daniel Hartmeier
5  * Copyright (c) 2002 - 2018 Henning Brauer <henning@openbsd.org>
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  *
12  *    - Redistributions of source code must retain the above copyright
13  *      notice, this list of conditions and the following disclaimer.
14  *    - Redistributions in binary form must reproduce the above
15  *      copyright notice, this list of conditions and the following
16  *      disclaimer in the documentation and/or other materials provided
17  *      with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
23  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
25  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
27  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
29  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  *
32  * Effort sponsored in part by the Defense Advanced Research Projects
33  * Agency (DARPA) and Air Force Research Laboratory, Air Force
34  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
35  *
36  */
37 
38 #include "pfsync.h"
39 #include "pflog.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/sysctl.h>
44 #include <sys/mbuf.h>
45 #include <sys/filio.h>
46 #include <sys/fcntl.h>
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/kernel.h>
50 #include <sys/time.h>
51 #include <sys/timeout.h>
52 #include <sys/pool.h>
53 #include <sys/malloc.h>
54 #include <sys/proc.h>
55 #include <sys/rwlock.h>
56 #include <sys/syslog.h>
57 #include <uvm/uvm_extern.h>
58 
59 #include <crypto/md5.h>
60 
61 #include <net/if.h>
62 #include <net/if_var.h>
63 #include <net/route.h>
64 #include <net/hfsc.h>
65 #include <net/fq_codel.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/ip.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_icmp.h>
72 #include <netinet/tcp.h>
73 #include <netinet/udp.h>
74 
75 #ifdef INET6
76 #include <netinet/ip6.h>
77 #include <netinet/icmp6.h>
78 #endif /* INET6 */
79 
80 #include <net/pfvar.h>
81 #include <net/pfvar_priv.h>
82 
83 #if NPFSYNC > 0
84 #include <netinet/ip_ipsp.h>
85 #include <net/if_pfsync.h>
86 #endif /* NPFSYNC > 0 */
87 
88 struct pool		 pf_tag_pl;
89 
90 void			 pfattach(int);
91 void			 pf_thread_create(void *);
92 int			 pfopen(dev_t, int, int, struct proc *);
93 int			 pfclose(dev_t, int, int, struct proc *);
94 int			 pfioctl(dev_t, u_long, caddr_t, int, struct proc *);
95 int			 pf_begin_rules(u_int32_t *, const char *);
96 void			 pf_rollback_rules(u_int32_t, char *);
97 void			 pf_remove_queues(void);
98 int			 pf_commit_queues(void);
99 void			 pf_free_queues(struct pf_queuehead *);
100 void			 pf_calc_chksum(struct pf_ruleset *);
101 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
102 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
103 int			 pf_commit_rules(u_int32_t, char *);
104 int			 pf_addr_setup(struct pf_ruleset *,
105 			    struct pf_addr_wrap *, sa_family_t);
106 struct pfi_kif		*pf_kif_setup(struct pfi_kif *);
107 void			 pf_addr_copyout(struct pf_addr_wrap *);
108 void			 pf_trans_set_commit(void);
109 void			 pf_pool_copyin(struct pf_pool *, struct pf_pool *);
110 int			 pf_validate_range(u_int8_t, u_int16_t[2], int);
111 int			 pf_rule_copyin(struct pf_rule *, struct pf_rule *);
112 int			 pf_rule_checkaf(struct pf_rule *);
113 u_int16_t		 pf_qname2qid(char *, int);
114 void			 pf_qid2qname(u_int16_t, char *);
115 void			 pf_qid_unref(u_int16_t);
116 int			 pf_states_clr(struct pfioc_state_kill *);
117 int			 pf_states_get(struct pfioc_states *);
118 
119 struct pf_rule		 pf_default_rule, pf_default_rule_new;
120 
121 struct {
122 	char		statusif[IFNAMSIZ];
123 	u_int32_t	debug;
124 	u_int32_t	hostid;
125 	u_int32_t	reass;
126 	u_int32_t	mask;
127 } pf_trans_set;
128 
129 #define	PF_ORDER_HOST	0
130 #define	PF_ORDER_NET	1
131 
132 #define	PF_TSET_STATUSIF	0x01
133 #define	PF_TSET_DEBUG		0x02
134 #define	PF_TSET_HOSTID		0x04
135 #define	PF_TSET_REASS		0x08
136 
137 #define	TAGID_MAX	 50000
138 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
139 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
140 
141 /*
142  * pf_lock protects consistency of PF data structures, which don't have
143  * their dedicated lock yet. The pf_lock currently protects:
144  *	- rules,
145  *	- radix tables,
146  *	- source nodes
147  * All callers must grab pf_lock exclusively.
148  *
149  * pf_state_lock protects consistency of state table. Packets, which do state
150  * look up grab the lock as readers. If packet must create state, then it must
151  * grab the lock as writer. Whenever packet creates state it grabs pf_lock
152  * first then it locks pf_state_lock as the writer.
153  */
154 struct rwlock		 pf_lock = RWLOCK_INITIALIZER("pf_lock");
155 struct rwlock		 pf_state_lock = RWLOCK_INITIALIZER("pf_state_lock");
156 struct rwlock		 pfioctl_rw = RWLOCK_INITIALIZER("pfioctl_rw");
157 
158 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
159 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
160 #endif
161 u_int16_t		 tagname2tag(struct pf_tags *, char *, int);
162 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
163 void			 tag_unref(struct pf_tags *, u_int16_t);
164 int			 pf_rtlabel_add(struct pf_addr_wrap *);
165 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
166 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
167 
168 
169 void
170 pfattach(int num)
171 {
172 	u_int32_t *timeout = pf_default_rule.timeout;
173 
174 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0,
175 	    IPL_SOFTNET, 0, "pfrule", NULL);
176 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0,
177 	    IPL_SOFTNET, 0, "pfsrctr", NULL);
178 	pool_init(&pf_sn_item_pl, sizeof(struct pf_sn_item), 0,
179 	    IPL_SOFTNET, 0, "pfsnitem", NULL);
180 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0,
181 	    IPL_SOFTNET, 0, "pfstate", NULL);
182 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0,
183 	    IPL_SOFTNET, 0, "pfstkey", NULL);
184 	pool_init(&pf_state_item_pl, sizeof(struct pf_state_item), 0,
185 	    IPL_SOFTNET, 0, "pfstitem", NULL);
186 	pool_init(&pf_rule_item_pl, sizeof(struct pf_rule_item), 0,
187 	    IPL_SOFTNET, 0, "pfruleitem", NULL);
188 	pool_init(&pf_queue_pl, sizeof(struct pf_queuespec), 0,
189 	    IPL_SOFTNET, 0, "pfqueue", NULL);
190 	pool_init(&pf_tag_pl, sizeof(struct pf_tagname), 0,
191 	    IPL_SOFTNET, 0, "pftag", NULL);
192 	pool_init(&pf_pktdelay_pl, sizeof(struct pf_pktdelay), 0,
193 	    IPL_SOFTNET, 0, "pfpktdelay", NULL);
194 
195 	hfsc_initialize();
196 	pfr_initialize();
197 	pfi_initialize();
198 	pf_osfp_initialize();
199 	pf_syncookies_init();
200 
201 	pool_sethardlimit(pf_pool_limits[PF_LIMIT_STATES].pp,
202 	    pf_pool_limits[PF_LIMIT_STATES].limit, NULL, 0);
203 
204 	if (physmem <= atop(100*1024*1024))
205 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
206 		    PFR_KENTRY_HIWAT_SMALL;
207 
208 	RB_INIT(&tree_src_tracking);
209 	RB_INIT(&pf_anchors);
210 	pf_init_ruleset(&pf_main_ruleset);
211 	TAILQ_INIT(&pf_queues[0]);
212 	TAILQ_INIT(&pf_queues[1]);
213 	pf_queues_active = &pf_queues[0];
214 	pf_queues_inactive = &pf_queues[1];
215 
216 	/* default rule should never be garbage collected */
217 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
218 	pf_default_rule.action = PF_PASS;
219 	pf_default_rule.nr = (u_int32_t)-1;
220 	pf_default_rule.rtableid = -1;
221 
222 	/* initialize default timeouts */
223 	timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
224 	timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
225 	timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
226 	timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
227 	timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
228 	timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
229 	timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
230 	timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
231 	timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
232 	timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
233 	timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
234 	timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
235 	timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
236 	timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
237 	timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
238 	timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
239 	timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
240 	timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
241 	timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
242 	timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
243 
244 	pf_default_rule.src.addr.type =  PF_ADDR_ADDRMASK;
245 	pf_default_rule.dst.addr.type =  PF_ADDR_ADDRMASK;
246 	pf_default_rule.rdr.addr.type =  PF_ADDR_NONE;
247 	pf_default_rule.nat.addr.type =  PF_ADDR_NONE;
248 	pf_default_rule.route.addr.type =  PF_ADDR_NONE;
249 
250 	pf_normalize_init();
251 	memset(&pf_status, 0, sizeof(pf_status));
252 	pf_status.debug = LOG_ERR;
253 	pf_status.reass = PF_REASS_ENABLED;
254 
255 	/* XXX do our best to avoid a conflict */
256 	pf_status.hostid = arc4random();
257 
258 	pf_default_rule_new = pf_default_rule;
259 }
260 
261 int
262 pfopen(dev_t dev, int flags, int fmt, struct proc *p)
263 {
264 	if (minor(dev) >= 1)
265 		return (ENXIO);
266 	return (0);
267 }
268 
269 int
270 pfclose(dev_t dev, int flags, int fmt, struct proc *p)
271 {
272 	if (minor(dev) >= 1)
273 		return (ENXIO);
274 	return (0);
275 }
276 
277 void
278 pf_rule_free(struct pf_rule *rule)
279 {
280 	if (rule == NULL)
281 		return;
282 
283 	pfi_kif_free(rule->kif);
284 	pfi_kif_free(rule->rcv_kif);
285 	pfi_kif_free(rule->rdr.kif);
286 	pfi_kif_free(rule->nat.kif);
287 	pfi_kif_free(rule->route.kif);
288 
289 	pool_put(&pf_rule_pl, rule);
290 }
291 
292 void
293 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
294 {
295 	if (rulequeue != NULL) {
296 		if (rule->states_cur == 0 && rule->src_nodes == 0) {
297 			/*
298 			 * XXX - we need to remove the table *before* detaching
299 			 * the rule to make sure the table code does not delete
300 			 * the anchor under our feet.
301 			 */
302 			pf_tbladdr_remove(&rule->src.addr);
303 			pf_tbladdr_remove(&rule->dst.addr);
304 			pf_tbladdr_remove(&rule->rdr.addr);
305 			pf_tbladdr_remove(&rule->nat.addr);
306 			pf_tbladdr_remove(&rule->route.addr);
307 			if (rule->overload_tbl)
308 				pfr_detach_table(rule->overload_tbl);
309 		}
310 		TAILQ_REMOVE(rulequeue, rule, entries);
311 		rule->entries.tqe_prev = NULL;
312 		rule->nr = (u_int32_t)-1;
313 	}
314 
315 	if (rule->states_cur > 0 || rule->src_nodes > 0 ||
316 	    rule->entries.tqe_prev != NULL)
317 		return;
318 	pf_tag_unref(rule->tag);
319 	pf_tag_unref(rule->match_tag);
320 	pf_rtlabel_remove(&rule->src.addr);
321 	pf_rtlabel_remove(&rule->dst.addr);
322 	pfi_dynaddr_remove(&rule->src.addr);
323 	pfi_dynaddr_remove(&rule->dst.addr);
324 	pfi_dynaddr_remove(&rule->rdr.addr);
325 	pfi_dynaddr_remove(&rule->nat.addr);
326 	pfi_dynaddr_remove(&rule->route.addr);
327 	if (rulequeue == NULL) {
328 		pf_tbladdr_remove(&rule->src.addr);
329 		pf_tbladdr_remove(&rule->dst.addr);
330 		pf_tbladdr_remove(&rule->rdr.addr);
331 		pf_tbladdr_remove(&rule->nat.addr);
332 		pf_tbladdr_remove(&rule->route.addr);
333 		if (rule->overload_tbl)
334 			pfr_detach_table(rule->overload_tbl);
335 	}
336 	pfi_kif_unref(rule->rcv_kif, PFI_KIF_REF_RULE);
337 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
338 	pfi_kif_unref(rule->rdr.kif, PFI_KIF_REF_RULE);
339 	pfi_kif_unref(rule->nat.kif, PFI_KIF_REF_RULE);
340 	pfi_kif_unref(rule->route.kif, PFI_KIF_REF_RULE);
341 	pf_remove_anchor(rule);
342 	pool_put(&pf_rule_pl, rule);
343 }
344 
345 void
346 pf_purge_rule(struct pf_rule *rule)
347 {
348 	u_int32_t		 nr = 0;
349 	struct pf_ruleset	*ruleset;
350 
351 	KASSERT((rule != NULL) && (rule->ruleset != NULL));
352 	ruleset = rule->ruleset;
353 
354 	pf_rm_rule(ruleset->rules.active.ptr, rule);
355 	ruleset->rules.active.rcount--;
356 	TAILQ_FOREACH(rule, ruleset->rules.active.ptr, entries)
357 		rule->nr = nr++;
358 	ruleset->rules.active.ticket++;
359 	pf_calc_skip_steps(ruleset->rules.active.ptr);
360 	pf_remove_if_empty_ruleset(ruleset);
361 
362 	if (ruleset == &pf_main_ruleset)
363 		pf_calc_chksum(ruleset);
364 }
365 
366 u_int16_t
367 tagname2tag(struct pf_tags *head, char *tagname, int create)
368 {
369 	struct pf_tagname	*tag, *p = NULL;
370 	u_int16_t		 new_tagid = 1;
371 
372 	TAILQ_FOREACH(tag, head, entries)
373 		if (strcmp(tagname, tag->name) == 0) {
374 			tag->ref++;
375 			return (tag->tag);
376 		}
377 
378 	if (!create)
379 		return (0);
380 
381 	/*
382 	 * to avoid fragmentation, we do a linear search from the beginning
383 	 * and take the first free slot we find. if there is none or the list
384 	 * is empty, append a new entry at the end.
385 	 */
386 
387 	/* new entry */
388 	TAILQ_FOREACH(p, head, entries) {
389 		if (p->tag != new_tagid)
390 			break;
391 		new_tagid = p->tag + 1;
392 	}
393 
394 	if (new_tagid > TAGID_MAX)
395 		return (0);
396 
397 	/* allocate and fill new struct pf_tagname */
398 	tag = pool_get(&pf_tag_pl, PR_NOWAIT | PR_ZERO);
399 	if (tag == NULL)
400 		return (0);
401 	strlcpy(tag->name, tagname, sizeof(tag->name));
402 	tag->tag = new_tagid;
403 	tag->ref++;
404 
405 	if (p != NULL)	/* insert new entry before p */
406 		TAILQ_INSERT_BEFORE(p, tag, entries);
407 	else	/* either list empty or no free slot in between */
408 		TAILQ_INSERT_TAIL(head, tag, entries);
409 
410 	return (tag->tag);
411 }
412 
413 void
414 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
415 {
416 	struct pf_tagname	*tag;
417 
418 	TAILQ_FOREACH(tag, head, entries)
419 		if (tag->tag == tagid) {
420 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
421 			return;
422 		}
423 }
424 
425 void
426 tag_unref(struct pf_tags *head, u_int16_t tag)
427 {
428 	struct pf_tagname	*p, *next;
429 
430 	if (tag == 0)
431 		return;
432 
433 	TAILQ_FOREACH_SAFE(p, head, entries, next) {
434 		if (tag == p->tag) {
435 			if (--p->ref == 0) {
436 				TAILQ_REMOVE(head, p, entries);
437 				pool_put(&pf_tag_pl, p);
438 			}
439 			break;
440 		}
441 	}
442 }
443 
444 u_int16_t
445 pf_tagname2tag(char *tagname, int create)
446 {
447 	return (tagname2tag(&pf_tags, tagname, create));
448 }
449 
450 void
451 pf_tag2tagname(u_int16_t tagid, char *p)
452 {
453 	tag2tagname(&pf_tags, tagid, p);
454 }
455 
456 void
457 pf_tag_ref(u_int16_t tag)
458 {
459 	struct pf_tagname *t;
460 
461 	TAILQ_FOREACH(t, &pf_tags, entries)
462 		if (t->tag == tag)
463 			break;
464 	if (t != NULL)
465 		t->ref++;
466 }
467 
468 void
469 pf_tag_unref(u_int16_t tag)
470 {
471 	tag_unref(&pf_tags, tag);
472 }
473 
474 int
475 pf_rtlabel_add(struct pf_addr_wrap *a)
476 {
477 	if (a->type == PF_ADDR_RTLABEL &&
478 	    (a->v.rtlabel = rtlabel_name2id(a->v.rtlabelname)) == 0)
479 		return (-1);
480 	return (0);
481 }
482 
483 void
484 pf_rtlabel_remove(struct pf_addr_wrap *a)
485 {
486 	if (a->type == PF_ADDR_RTLABEL)
487 		rtlabel_unref(a->v.rtlabel);
488 }
489 
490 void
491 pf_rtlabel_copyout(struct pf_addr_wrap *a)
492 {
493 	const char	*name;
494 
495 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel) {
496 		if ((name = rtlabel_id2name(a->v.rtlabel)) == NULL)
497 			strlcpy(a->v.rtlabelname, "?",
498 			    sizeof(a->v.rtlabelname));
499 		else
500 			strlcpy(a->v.rtlabelname, name,
501 			    sizeof(a->v.rtlabelname));
502 	}
503 }
504 
505 u_int16_t
506 pf_qname2qid(char *qname, int create)
507 {
508 	return (tagname2tag(&pf_qids, qname, create));
509 }
510 
511 void
512 pf_qid2qname(u_int16_t qid, char *p)
513 {
514 	tag2tagname(&pf_qids, qid, p);
515 }
516 
517 void
518 pf_qid_unref(u_int16_t qid)
519 {
520 	tag_unref(&pf_qids, (u_int16_t)qid);
521 }
522 
523 int
524 pf_begin_rules(u_int32_t *ticket, const char *anchor)
525 {
526 	struct pf_ruleset	*rs;
527 	struct pf_rule		*rule;
528 
529 	if ((rs = pf_find_or_create_ruleset(anchor)) == NULL)
530 		return (EINVAL);
531 	while ((rule = TAILQ_FIRST(rs->rules.inactive.ptr)) != NULL) {
532 		pf_rm_rule(rs->rules.inactive.ptr, rule);
533 		rs->rules.inactive.rcount--;
534 	}
535 	*ticket = ++rs->rules.inactive.ticket;
536 	rs->rules.inactive.open = 1;
537 	return (0);
538 }
539 
540 void
541 pf_rollback_rules(u_int32_t ticket, char *anchor)
542 {
543 	struct pf_ruleset	*rs;
544 	struct pf_rule		*rule;
545 
546 	rs = pf_find_ruleset(anchor);
547 	if (rs == NULL || !rs->rules.inactive.open ||
548 	    rs->rules.inactive.ticket != ticket)
549 		return;
550 	while ((rule = TAILQ_FIRST(rs->rules.inactive.ptr)) != NULL) {
551 		pf_rm_rule(rs->rules.inactive.ptr, rule);
552 		rs->rules.inactive.rcount--;
553 	}
554 	rs->rules.inactive.open = 0;
555 
556 	/* queue defs only in the main ruleset */
557 	if (anchor[0])
558 		return;
559 
560 	pf_free_queues(pf_queues_inactive);
561 }
562 
563 void
564 pf_free_queues(struct pf_queuehead *where)
565 {
566 	struct pf_queuespec	*q, *qtmp;
567 
568 	TAILQ_FOREACH_SAFE(q, where, entries, qtmp) {
569 		TAILQ_REMOVE(where, q, entries);
570 		pfi_kif_unref(q->kif, PFI_KIF_REF_RULE);
571 		pool_put(&pf_queue_pl, q);
572 	}
573 }
574 
575 void
576 pf_remove_queues(void)
577 {
578 	struct pf_queuespec	*q;
579 	struct ifnet		*ifp;
580 
581 	/* put back interfaces in normal queueing mode */
582 	TAILQ_FOREACH(q, pf_queues_active, entries) {
583 		if (q->parent_qid != 0)
584 			continue;
585 
586 		ifp = q->kif->pfik_ifp;
587 		if (ifp == NULL)
588 			continue;
589 
590 		ifq_attach(&ifp->if_snd, ifq_priq_ops, NULL);
591 	}
592 }
593 
594 struct pf_queue_if {
595 	struct ifnet		*ifp;
596 	const struct ifq_ops	*ifqops;
597 	const struct pfq_ops	*pfqops;
598 	void			*disc;
599 	struct pf_queue_if	*next;
600 };
601 
602 static inline struct pf_queue_if *
603 pf_ifp2q(struct pf_queue_if *list, struct ifnet *ifp)
604 {
605 	struct pf_queue_if *qif = list;
606 
607 	while (qif != NULL) {
608 		if (qif->ifp == ifp)
609 			return (qif);
610 
611 		qif = qif->next;
612 	}
613 
614 	return (qif);
615 }
616 
617 int
618 pf_create_queues(void)
619 {
620 	struct pf_queuespec	*q;
621 	struct ifnet		*ifp;
622 	struct pf_queue_if		*list = NULL, *qif;
623 	int			 error;
624 
625 	/*
626 	 * Find root queues and allocate traffic conditioner
627 	 * private data for these interfaces
628 	 */
629 	TAILQ_FOREACH(q, pf_queues_active, entries) {
630 		if (q->parent_qid != 0)
631 			continue;
632 
633 		ifp = q->kif->pfik_ifp;
634 		if (ifp == NULL)
635 			continue;
636 
637 		qif = malloc(sizeof(*qif), M_TEMP, M_WAITOK);
638 		qif->ifp = ifp;
639 
640 		if (q->flags & PFQS_ROOTCLASS) {
641 			qif->ifqops = ifq_hfsc_ops;
642 			qif->pfqops = pfq_hfsc_ops;
643 		} else {
644 			qif->ifqops = ifq_fqcodel_ops;
645 			qif->pfqops = pfq_fqcodel_ops;
646 		}
647 
648 		qif->disc = qif->pfqops->pfq_alloc(ifp);
649 
650 		qif->next = list;
651 		list = qif;
652 	}
653 
654 	/* and now everything */
655 	TAILQ_FOREACH(q, pf_queues_active, entries) {
656 		ifp = q->kif->pfik_ifp;
657 		if (ifp == NULL)
658 			continue;
659 
660 		qif = pf_ifp2q(list, ifp);
661 		KASSERT(qif != NULL);
662 
663 		error = qif->pfqops->pfq_addqueue(qif->disc, q);
664 		if (error != 0)
665 			goto error;
666 	}
667 
668 	/* find root queues in old list to disable them if necessary */
669 	TAILQ_FOREACH(q, pf_queues_inactive, entries) {
670 		if (q->parent_qid != 0)
671 			continue;
672 
673 		ifp = q->kif->pfik_ifp;
674 		if (ifp == NULL)
675 			continue;
676 
677 		qif = pf_ifp2q(list, ifp);
678 		if (qif != NULL)
679 			continue;
680 
681 		ifq_attach(&ifp->if_snd, ifq_priq_ops, NULL);
682 	}
683 
684 	/* commit the new queues */
685 	while (list != NULL) {
686 		qif = list;
687 		list = qif->next;
688 
689 		ifp = qif->ifp;
690 
691 		ifq_attach(&ifp->if_snd, qif->ifqops, qif->disc);
692 		free(qif, M_TEMP, sizeof(*qif));
693 	}
694 
695 	return (0);
696 
697 error:
698 	while (list != NULL) {
699 		qif = list;
700 		list = qif->next;
701 
702 		qif->pfqops->pfq_free(qif->disc);
703 		free(qif, M_TEMP, sizeof(*qif));
704 	}
705 
706 	return (error);
707 }
708 
709 int
710 pf_commit_queues(void)
711 {
712 	struct pf_queuehead	*qswap;
713 	int error;
714 
715 	/* swap */
716 	qswap = pf_queues_active;
717 	pf_queues_active = pf_queues_inactive;
718 	pf_queues_inactive = qswap;
719 
720 	error = pf_create_queues();
721 	if (error != 0) {
722 		pf_queues_inactive = pf_queues_active;
723 		pf_queues_active = qswap;
724 		return (error);
725 	}
726 
727 	pf_free_queues(pf_queues_inactive);
728 
729 	return (0);
730 }
731 
732 const struct pfq_ops *
733 pf_queue_manager(struct pf_queuespec *q)
734 {
735 	if (q->flags & PFQS_FLOWQUEUE)
736 		return pfq_fqcodel_ops;
737 	return (/* pfq_default_ops */ NULL);
738 }
739 
740 #define PF_MD5_UPD(st, elm)						\
741 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
742 
743 #define PF_MD5_UPD_STR(st, elm)						\
744 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
745 
746 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
747 		(stor) = htonl((st)->elm);				\
748 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
749 } while (0)
750 
751 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
752 		(stor) = htons((st)->elm);				\
753 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
754 } while (0)
755 
756 void
757 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
758 {
759 	PF_MD5_UPD(pfr, addr.type);
760 	switch (pfr->addr.type) {
761 		case PF_ADDR_DYNIFTL:
762 			PF_MD5_UPD(pfr, addr.v.ifname);
763 			PF_MD5_UPD(pfr, addr.iflags);
764 			break;
765 		case PF_ADDR_TABLE:
766 			if (strncmp(pfr->addr.v.tblname, PF_OPTIMIZER_TABLE_PFX,
767 			    strlen(PF_OPTIMIZER_TABLE_PFX)))
768 				PF_MD5_UPD(pfr, addr.v.tblname);
769 			break;
770 		case PF_ADDR_ADDRMASK:
771 			/* XXX ignore af? */
772 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
773 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
774 			break;
775 		case PF_ADDR_RTLABEL:
776 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
777 			break;
778 	}
779 
780 	PF_MD5_UPD(pfr, port[0]);
781 	PF_MD5_UPD(pfr, port[1]);
782 	PF_MD5_UPD(pfr, neg);
783 	PF_MD5_UPD(pfr, port_op);
784 }
785 
786 void
787 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
788 {
789 	u_int16_t x;
790 	u_int32_t y;
791 
792 	pf_hash_rule_addr(ctx, &rule->src);
793 	pf_hash_rule_addr(ctx, &rule->dst);
794 	PF_MD5_UPD_STR(rule, label);
795 	PF_MD5_UPD_STR(rule, ifname);
796 	PF_MD5_UPD_STR(rule, rcv_ifname);
797 	PF_MD5_UPD_STR(rule, match_tagname);
798 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
799 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
800 	PF_MD5_UPD_HTONL(rule, prob, y);
801 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
802 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
803 	PF_MD5_UPD(rule, uid.op);
804 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
805 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
806 	PF_MD5_UPD(rule, gid.op);
807 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
808 	PF_MD5_UPD(rule, action);
809 	PF_MD5_UPD(rule, direction);
810 	PF_MD5_UPD(rule, af);
811 	PF_MD5_UPD(rule, quick);
812 	PF_MD5_UPD(rule, ifnot);
813 	PF_MD5_UPD(rule, rcvifnot);
814 	PF_MD5_UPD(rule, match_tag_not);
815 	PF_MD5_UPD(rule, keep_state);
816 	PF_MD5_UPD(rule, proto);
817 	PF_MD5_UPD(rule, type);
818 	PF_MD5_UPD(rule, code);
819 	PF_MD5_UPD(rule, flags);
820 	PF_MD5_UPD(rule, flagset);
821 	PF_MD5_UPD(rule, allow_opts);
822 	PF_MD5_UPD(rule, rt);
823 	PF_MD5_UPD(rule, tos);
824 }
825 
826 int
827 pf_commit_rules(u_int32_t ticket, char *anchor)
828 {
829 	struct pf_ruleset	*rs;
830 	struct pf_rule		*rule;
831 	struct pf_rulequeue	*old_rules;
832 	u_int32_t		 old_rcount;
833 
834 	/* Make sure any expired rules get removed from active rules first. */
835 	pf_purge_expired_rules();
836 
837 	rs = pf_find_ruleset(anchor);
838 	if (rs == NULL || !rs->rules.inactive.open ||
839 	    ticket != rs->rules.inactive.ticket)
840 		return (EBUSY);
841 
842 	if (rs == &pf_main_ruleset)
843 		pf_calc_chksum(rs);
844 
845 	/* Swap rules, keep the old. */
846 	old_rules = rs->rules.active.ptr;
847 	old_rcount = rs->rules.active.rcount;
848 
849 	rs->rules.active.ptr = rs->rules.inactive.ptr;
850 	rs->rules.active.rcount = rs->rules.inactive.rcount;
851 	rs->rules.inactive.ptr = old_rules;
852 	rs->rules.inactive.rcount = old_rcount;
853 
854 	rs->rules.active.ticket = rs->rules.inactive.ticket;
855 	pf_calc_skip_steps(rs->rules.active.ptr);
856 
857 
858 	/* Purge the old rule list. */
859 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
860 		pf_rm_rule(old_rules, rule);
861 	rs->rules.inactive.rcount = 0;
862 	rs->rules.inactive.open = 0;
863 	pf_remove_if_empty_ruleset(rs);
864 
865 	/* queue defs only in the main ruleset */
866 	if (anchor[0])
867 		return (0);
868 	return (pf_commit_queues());
869 }
870 
871 void
872 pf_calc_chksum(struct pf_ruleset *rs)
873 {
874 	MD5_CTX			 ctx;
875 	struct pf_rule		*rule;
876 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
877 
878 	MD5Init(&ctx);
879 
880 	if (rs->rules.inactive.rcount) {
881 		TAILQ_FOREACH(rule, rs->rules.inactive.ptr, entries) {
882 			pf_hash_rule(&ctx, rule);
883 		}
884 	}
885 
886 	MD5Final(digest, &ctx);
887 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
888 }
889 
890 int
891 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
892     sa_family_t af)
893 {
894 	if (pfi_dynaddr_setup(addr, af) ||
895 	    pf_tbladdr_setup(ruleset, addr) ||
896 	    pf_rtlabel_add(addr))
897 		return (EINVAL);
898 
899 	return (0);
900 }
901 
902 struct pfi_kif *
903 pf_kif_setup(struct pfi_kif *kif_buf)
904 {
905 	struct pfi_kif *kif;
906 
907 	if (kif_buf == NULL)
908 		return (NULL);
909 
910 	KASSERT(kif_buf->pfik_name[0] != '\0');
911 
912 	kif = pfi_kif_get(kif_buf->pfik_name, &kif_buf);
913 	if (kif_buf != NULL)
914 		pfi_kif_free(kif_buf);
915 	pfi_kif_ref(kif, PFI_KIF_REF_RULE);
916 
917 	return (kif);
918 }
919 
920 void
921 pf_addr_copyout(struct pf_addr_wrap *addr)
922 {
923 	pfi_dynaddr_copyout(addr);
924 	pf_tbladdr_copyout(addr);
925 	pf_rtlabel_copyout(addr);
926 }
927 
928 int
929 pf_states_clr(struct pfioc_state_kill *psk)
930 {
931 	struct pf_state		*s, *nexts;
932 	struct pf_state		*head, *tail;
933 	u_int			 killed = 0;
934 	int			 error;
935 
936 	NET_LOCK();
937 
938 	/* lock against the gc removing an item from the list */
939 	error = rw_enter(&pf_state_list.pfs_rwl, RW_READ|RW_INTR);
940 	if (error != 0)
941 		goto unlock;
942 
943 	/* get a snapshot view of the ends of the list to traverse between */
944 	mtx_enter(&pf_state_list.pfs_mtx);
945 	head = TAILQ_FIRST(&pf_state_list.pfs_list);
946 	tail = TAILQ_LAST(&pf_state_list.pfs_list, pf_state_queue);
947 	mtx_leave(&pf_state_list.pfs_mtx);
948 
949 	s = NULL;
950 	nexts = head;
951 
952 	PF_LOCK();
953 	PF_STATE_ENTER_WRITE();
954 
955 	while (s != tail) {
956 		s = nexts;
957 		nexts = TAILQ_NEXT(s, entry_list);
958 
959 		if (s->timeout == PFTM_UNLINKED)
960 			continue;
961 
962 		if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
963 		    s->kif->pfik_name)) {
964 #if NPFSYNC > 0
965 			/* don't send out individual delete messages */
966 			SET(s->state_flags, PFSTATE_NOSYNC);
967 #endif	/* NPFSYNC > 0 */
968 			pf_remove_state(s);
969 			killed++;
970 		}
971 	}
972 
973 	PF_STATE_EXIT_WRITE();
974 #if NPFSYNC > 0
975 	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
976 #endif	/* NPFSYNC > 0 */
977 	PF_UNLOCK();
978 	rw_exit(&pf_state_list.pfs_rwl);
979 
980 	psk->psk_killed = killed;
981 unlock:
982 	NET_UNLOCK();
983 
984 	return (error);
985 }
986 
987 int
988 pf_states_get(struct pfioc_states *ps)
989 {
990 	struct pf_state		*head, *tail;
991 	struct pf_state		*next, *state;
992 	struct pfsync_state	*p, pstore;
993 	u_int32_t		 nr = 0;
994 	int			 error;
995 
996 	if (ps->ps_len == 0) {
997 		nr = pf_status.states;
998 		ps->ps_len = sizeof(struct pfsync_state) * nr;
999 		return (0);
1000 	}
1001 
1002 	p = ps->ps_states;
1003 
1004 	/* lock against the gc removing an item from the list */
1005 	error = rw_enter(&pf_state_list.pfs_rwl, RW_READ|RW_INTR);
1006 	if (error != 0)
1007 		return (error);
1008 
1009 	/* get a snapshot view of the ends of the list to traverse between */
1010 	mtx_enter(&pf_state_list.pfs_mtx);
1011 	head = TAILQ_FIRST(&pf_state_list.pfs_list);
1012 	tail = TAILQ_LAST(&pf_state_list.pfs_list, pf_state_queue);
1013 	mtx_leave(&pf_state_list.pfs_mtx);
1014 
1015 	state = NULL;
1016 	next = head;
1017 
1018 	while (state != tail) {
1019 		state = next;
1020 		next = TAILQ_NEXT(state, entry_list);
1021 
1022 		if (state->timeout == PFTM_UNLINKED)
1023 			continue;
1024 
1025 		if ((nr+1) * sizeof(*p) > ps->ps_len)
1026 			break;
1027 
1028 		pf_state_export(&pstore, state);
1029 		error = copyout(&pstore, p, sizeof(*p));
1030 		if (error)
1031 			goto fail;
1032 
1033 		p++;
1034 		nr++;
1035 	}
1036 	ps->ps_len = sizeof(struct pfsync_state) * nr;
1037 
1038 fail:
1039 	rw_exit(&pf_state_list.pfs_rwl);
1040 
1041 	return (error);
1042 }
1043 
1044 int
1045 pfioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p)
1046 {
1047 	int			 error = 0;
1048 
1049 	/* XXX keep in sync with switch() below */
1050 	if (securelevel > 1)
1051 		switch (cmd) {
1052 		case DIOCGETRULES:
1053 		case DIOCGETRULE:
1054 		case DIOCGETSTATE:
1055 		case DIOCSETSTATUSIF:
1056 		case DIOCGETSTATUS:
1057 		case DIOCCLRSTATUS:
1058 		case DIOCNATLOOK:
1059 		case DIOCSETDEBUG:
1060 		case DIOCGETSTATES:
1061 		case DIOCGETTIMEOUT:
1062 		case DIOCGETLIMIT:
1063 		case DIOCGETRULESETS:
1064 		case DIOCGETRULESET:
1065 		case DIOCGETQUEUES:
1066 		case DIOCGETQUEUE:
1067 		case DIOCGETQSTATS:
1068 		case DIOCRGETTABLES:
1069 		case DIOCRGETTSTATS:
1070 		case DIOCRCLRTSTATS:
1071 		case DIOCRCLRADDRS:
1072 		case DIOCRADDADDRS:
1073 		case DIOCRDELADDRS:
1074 		case DIOCRSETADDRS:
1075 		case DIOCRGETADDRS:
1076 		case DIOCRGETASTATS:
1077 		case DIOCRCLRASTATS:
1078 		case DIOCRTSTADDRS:
1079 		case DIOCOSFPGET:
1080 		case DIOCGETSRCNODES:
1081 		case DIOCCLRSRCNODES:
1082 		case DIOCIGETIFACES:
1083 		case DIOCSETIFFLAG:
1084 		case DIOCCLRIFFLAG:
1085 		case DIOCGETSYNFLWATS:
1086 			break;
1087 		case DIOCRCLRTABLES:
1088 		case DIOCRADDTABLES:
1089 		case DIOCRDELTABLES:
1090 		case DIOCRSETTFLAGS:
1091 			if (((struct pfioc_table *)addr)->pfrio_flags &
1092 			    PFR_FLAG_DUMMY)
1093 				break; /* dummy operation ok */
1094 			return (EPERM);
1095 		default:
1096 			return (EPERM);
1097 		}
1098 
1099 	if (!(flags & FWRITE))
1100 		switch (cmd) {
1101 		case DIOCGETRULES:
1102 		case DIOCGETSTATE:
1103 		case DIOCGETSTATUS:
1104 		case DIOCGETSTATES:
1105 		case DIOCGETTIMEOUT:
1106 		case DIOCGETLIMIT:
1107 		case DIOCGETRULESETS:
1108 		case DIOCGETRULESET:
1109 		case DIOCGETQUEUES:
1110 		case DIOCGETQUEUE:
1111 		case DIOCGETQSTATS:
1112 		case DIOCNATLOOK:
1113 		case DIOCRGETTABLES:
1114 		case DIOCRGETTSTATS:
1115 		case DIOCRGETADDRS:
1116 		case DIOCRGETASTATS:
1117 		case DIOCRTSTADDRS:
1118 		case DIOCOSFPGET:
1119 		case DIOCGETSRCNODES:
1120 		case DIOCIGETIFACES:
1121 		case DIOCGETSYNFLWATS:
1122 			break;
1123 		case DIOCRCLRTABLES:
1124 		case DIOCRADDTABLES:
1125 		case DIOCRDELTABLES:
1126 		case DIOCRCLRTSTATS:
1127 		case DIOCRCLRADDRS:
1128 		case DIOCRADDADDRS:
1129 		case DIOCRDELADDRS:
1130 		case DIOCRSETADDRS:
1131 		case DIOCRSETTFLAGS:
1132 			if (((struct pfioc_table *)addr)->pfrio_flags &
1133 			    PFR_FLAG_DUMMY) {
1134 				flags |= FWRITE; /* need write lock for dummy */
1135 				break; /* dummy operation ok */
1136 			}
1137 			return (EACCES);
1138 		case DIOCGETRULE:
1139 			if (((struct pfioc_rule *)addr)->action ==
1140 			    PF_GET_CLR_CNTR)
1141 				return (EACCES);
1142 			break;
1143 		default:
1144 			return (EACCES);
1145 		}
1146 
1147 	if (flags & FWRITE)
1148 		rw_enter_write(&pfioctl_rw);
1149 	else
1150 		rw_enter_read(&pfioctl_rw);
1151 
1152 	switch (cmd) {
1153 
1154 	case DIOCSTART:
1155 		NET_LOCK();
1156 		PF_LOCK();
1157 		if (pf_status.running)
1158 			error = EEXIST;
1159 		else {
1160 			pf_status.running = 1;
1161 			pf_status.since = getuptime();
1162 			if (pf_status.stateid == 0) {
1163 				pf_status.stateid = gettime();
1164 				pf_status.stateid = pf_status.stateid << 32;
1165 			}
1166 			timeout_add_sec(&pf_purge_to, 1);
1167 			pf_create_queues();
1168 			DPFPRINTF(LOG_NOTICE, "pf: started");
1169 		}
1170 		PF_UNLOCK();
1171 		NET_UNLOCK();
1172 		break;
1173 
1174 	case DIOCSTOP:
1175 		NET_LOCK();
1176 		PF_LOCK();
1177 		if (!pf_status.running)
1178 			error = ENOENT;
1179 		else {
1180 			pf_status.running = 0;
1181 			pf_status.since = getuptime();
1182 			pf_remove_queues();
1183 			DPFPRINTF(LOG_NOTICE, "pf: stopped");
1184 		}
1185 		PF_UNLOCK();
1186 		NET_UNLOCK();
1187 		break;
1188 
1189 	case DIOCGETQUEUES: {
1190 		struct pfioc_queue	*pq = (struct pfioc_queue *)addr;
1191 		struct pf_queuespec	*qs;
1192 		u_int32_t		 nr = 0;
1193 
1194 		NET_LOCK();
1195 		PF_LOCK();
1196 		pq->ticket = pf_main_ruleset.rules.active.ticket;
1197 
1198 		/* save state to not run over them all each time? */
1199 		qs = TAILQ_FIRST(pf_queues_active);
1200 		while (qs != NULL) {
1201 			qs = TAILQ_NEXT(qs, entries);
1202 			nr++;
1203 		}
1204 		pq->nr = nr;
1205 		PF_UNLOCK();
1206 		NET_UNLOCK();
1207 		break;
1208 	}
1209 
1210 	case DIOCGETQUEUE: {
1211 		struct pfioc_queue	*pq = (struct pfioc_queue *)addr;
1212 		struct pf_queuespec	*qs;
1213 		u_int32_t		 nr = 0;
1214 
1215 		NET_LOCK();
1216 		PF_LOCK();
1217 		if (pq->ticket != pf_main_ruleset.rules.active.ticket) {
1218 			error = EBUSY;
1219 			PF_UNLOCK();
1220 			NET_UNLOCK();
1221 			goto fail;
1222 		}
1223 
1224 		/* save state to not run over them all each time? */
1225 		qs = TAILQ_FIRST(pf_queues_active);
1226 		while ((qs != NULL) && (nr++ < pq->nr))
1227 			qs = TAILQ_NEXT(qs, entries);
1228 		if (qs == NULL) {
1229 			error = EBUSY;
1230 			PF_UNLOCK();
1231 			NET_UNLOCK();
1232 			goto fail;
1233 		}
1234 		memcpy(&pq->queue, qs, sizeof(pq->queue));
1235 		PF_UNLOCK();
1236 		NET_UNLOCK();
1237 		break;
1238 	}
1239 
1240 	case DIOCGETQSTATS: {
1241 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
1242 		struct pf_queuespec	*qs;
1243 		u_int32_t		 nr;
1244 		int			 nbytes;
1245 
1246 		NET_LOCK();
1247 		PF_LOCK();
1248 		if (pq->ticket != pf_main_ruleset.rules.active.ticket) {
1249 			error = EBUSY;
1250 			PF_UNLOCK();
1251 			NET_UNLOCK();
1252 			goto fail;
1253 		}
1254 		nbytes = pq->nbytes;
1255 		nr = 0;
1256 
1257 		/* save state to not run over them all each time? */
1258 		qs = TAILQ_FIRST(pf_queues_active);
1259 		while ((qs != NULL) && (nr++ < pq->nr))
1260 			qs = TAILQ_NEXT(qs, entries);
1261 		if (qs == NULL) {
1262 			error = EBUSY;
1263 			PF_UNLOCK();
1264 			NET_UNLOCK();
1265 			goto fail;
1266 		}
1267 		memcpy(&pq->queue, qs, sizeof(pq->queue));
1268 		/* It's a root flow queue but is not an HFSC root class */
1269 		if ((qs->flags & PFQS_FLOWQUEUE) && qs->parent_qid == 0 &&
1270 		    !(qs->flags & PFQS_ROOTCLASS))
1271 			error = pfq_fqcodel_ops->pfq_qstats(qs, pq->buf,
1272 			    &nbytes);
1273 		else
1274 			error = pfq_hfsc_ops->pfq_qstats(qs, pq->buf,
1275 			    &nbytes);
1276 		if (error == 0)
1277 			pq->nbytes = nbytes;
1278 		PF_UNLOCK();
1279 		NET_UNLOCK();
1280 		break;
1281 	}
1282 
1283 	case DIOCADDQUEUE: {
1284 		struct pfioc_queue	*q = (struct pfioc_queue *)addr;
1285 		struct pf_queuespec	*qs;
1286 
1287 		qs = pool_get(&pf_queue_pl, PR_WAITOK|PR_LIMITFAIL|PR_ZERO);
1288 		if (qs == NULL) {
1289 			error = ENOMEM;
1290 			goto fail;
1291 		}
1292 
1293 		NET_LOCK();
1294 		PF_LOCK();
1295 		if (q->ticket != pf_main_ruleset.rules.inactive.ticket) {
1296 			error = EBUSY;
1297 			PF_UNLOCK();
1298 			NET_UNLOCK();
1299 			pool_put(&pf_queue_pl, qs);
1300 			goto fail;
1301 		}
1302 		memcpy(qs, &q->queue, sizeof(*qs));
1303 		qs->qid = pf_qname2qid(qs->qname, 1);
1304 		if (qs->qid == 0) {
1305 			error = EBUSY;
1306 			PF_UNLOCK();
1307 			NET_UNLOCK();
1308 			pool_put(&pf_queue_pl, qs);
1309 			goto fail;
1310 		}
1311 		if (qs->parent[0] && (qs->parent_qid =
1312 		    pf_qname2qid(qs->parent, 0)) == 0) {
1313 			error = ESRCH;
1314 			PF_UNLOCK();
1315 			NET_UNLOCK();
1316 			pool_put(&pf_queue_pl, qs);
1317 			goto fail;
1318 		}
1319 		qs->kif = pfi_kif_get(qs->ifname, NULL);
1320 		if (qs->kif == NULL) {
1321 			error = ESRCH;
1322 			PF_UNLOCK();
1323 			NET_UNLOCK();
1324 			pool_put(&pf_queue_pl, qs);
1325 			goto fail;
1326 		}
1327 		/* XXX resolve bw percentage specs */
1328 		pfi_kif_ref(qs->kif, PFI_KIF_REF_RULE);
1329 
1330 		TAILQ_INSERT_TAIL(pf_queues_inactive, qs, entries);
1331 		PF_UNLOCK();
1332 		NET_UNLOCK();
1333 
1334 		break;
1335 	}
1336 
1337 	case DIOCADDRULE: {
1338 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1339 		struct pf_ruleset	*ruleset;
1340 		struct pf_rule		*rule, *tail;
1341 
1342 		rule = pool_get(&pf_rule_pl, PR_WAITOK|PR_LIMITFAIL|PR_ZERO);
1343 		if (rule == NULL) {
1344 			error = ENOMEM;
1345 			goto fail;
1346 		}
1347 
1348 		if ((error = pf_rule_copyin(&pr->rule, rule))) {
1349 			pf_rule_free(rule);
1350 			rule = NULL;
1351 			goto fail;
1352 		}
1353 
1354 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1355 			error = EINVAL;
1356 			pf_rule_free(rule);
1357 			rule = NULL;
1358 			goto fail;
1359 		}
1360 		if ((error = pf_rule_checkaf(rule))) {
1361 			pf_rule_free(rule);
1362 			rule = NULL;
1363 			goto fail;
1364 		}
1365 		if (rule->src.addr.type == PF_ADDR_NONE ||
1366 		    rule->dst.addr.type == PF_ADDR_NONE) {
1367 			error = EINVAL;
1368 			pf_rule_free(rule);
1369 			rule = NULL;
1370 			goto fail;
1371 		}
1372 
1373 		if (rule->rt && !rule->direction) {
1374 			error = EINVAL;
1375 			pf_rule_free(rule);
1376 			rule = NULL;
1377 			goto fail;
1378 		}
1379 
1380 		NET_LOCK();
1381 		PF_LOCK();
1382 		pr->anchor[sizeof(pr->anchor) - 1] = '\0';
1383 		ruleset = pf_find_ruleset(pr->anchor);
1384 		if (ruleset == NULL) {
1385 			error = EINVAL;
1386 			PF_UNLOCK();
1387 			NET_UNLOCK();
1388 			pf_rule_free(rule);
1389 			goto fail;
1390 		}
1391 		if (pr->ticket != ruleset->rules.inactive.ticket) {
1392 			error = EBUSY;
1393 			PF_UNLOCK();
1394 			NET_UNLOCK();
1395 			pf_rule_free(rule);
1396 			goto fail;
1397 		}
1398 		rule->cuid = p->p_ucred->cr_ruid;
1399 		rule->cpid = p->p_p->ps_pid;
1400 
1401 		tail = TAILQ_LAST(ruleset->rules.inactive.ptr,
1402 		    pf_rulequeue);
1403 		if (tail)
1404 			rule->nr = tail->nr + 1;
1405 		else
1406 			rule->nr = 0;
1407 
1408 		rule->kif = pf_kif_setup(rule->kif);
1409 		rule->rcv_kif = pf_kif_setup(rule->rcv_kif);
1410 		rule->rdr.kif = pf_kif_setup(rule->rdr.kif);
1411 		rule->nat.kif = pf_kif_setup(rule->nat.kif);
1412 		rule->route.kif = pf_kif_setup(rule->route.kif);
1413 
1414 		if (rule->overload_tblname[0]) {
1415 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1416 			    rule->overload_tblname, 0)) == NULL)
1417 				error = EINVAL;
1418 			else
1419 				rule->overload_tbl->pfrkt_flags |= PFR_TFLAG_ACTIVE;
1420 		}
1421 
1422 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1423 			error = EINVAL;
1424 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1425 			error = EINVAL;
1426 		if (pf_addr_setup(ruleset, &rule->rdr.addr, rule->af))
1427 			error = EINVAL;
1428 		if (pf_addr_setup(ruleset, &rule->nat.addr, rule->af))
1429 			error = EINVAL;
1430 		if (pf_addr_setup(ruleset, &rule->route.addr, rule->af))
1431 			error = EINVAL;
1432 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1433 			error = EINVAL;
1434 
1435 		if (error) {
1436 			pf_rm_rule(NULL, rule);
1437 			PF_UNLOCK();
1438 			NET_UNLOCK();
1439 			goto fail;
1440 		}
1441 		TAILQ_INSERT_TAIL(ruleset->rules.inactive.ptr,
1442 		    rule, entries);
1443 		rule->ruleset = ruleset;
1444 		ruleset->rules.inactive.rcount++;
1445 		PF_UNLOCK();
1446 		NET_UNLOCK();
1447 		break;
1448 	}
1449 
1450 	case DIOCGETRULES: {
1451 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1452 		struct pf_ruleset	*ruleset;
1453 		struct pf_rule		*tail;
1454 
1455 		NET_LOCK();
1456 		PF_LOCK();
1457 		pr->anchor[sizeof(pr->anchor) - 1] = '\0';
1458 		ruleset = pf_find_ruleset(pr->anchor);
1459 		if (ruleset == NULL) {
1460 			error = EINVAL;
1461 			PF_UNLOCK();
1462 			NET_UNLOCK();
1463 			goto fail;
1464 		}
1465 		tail = TAILQ_LAST(ruleset->rules.active.ptr, pf_rulequeue);
1466 		if (tail)
1467 			pr->nr = tail->nr + 1;
1468 		else
1469 			pr->nr = 0;
1470 		pr->ticket = ruleset->rules.active.ticket;
1471 		PF_UNLOCK();
1472 		NET_UNLOCK();
1473 		break;
1474 	}
1475 
1476 	case DIOCGETRULE: {
1477 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1478 		struct pf_ruleset	*ruleset;
1479 		struct pf_rule		*rule;
1480 		int			 i;
1481 
1482 		NET_LOCK();
1483 		PF_LOCK();
1484 		pr->anchor[sizeof(pr->anchor) - 1] = '\0';
1485 		ruleset = pf_find_ruleset(pr->anchor);
1486 		if (ruleset == NULL) {
1487 			error = EINVAL;
1488 			PF_UNLOCK();
1489 			NET_UNLOCK();
1490 			goto fail;
1491 		}
1492 		if (pr->ticket != ruleset->rules.active.ticket) {
1493 			error = EBUSY;
1494 			PF_UNLOCK();
1495 			NET_UNLOCK();
1496 			goto fail;
1497 		}
1498 		rule = TAILQ_FIRST(ruleset->rules.active.ptr);
1499 		while ((rule != NULL) && (rule->nr != pr->nr))
1500 			rule = TAILQ_NEXT(rule, entries);
1501 		if (rule == NULL) {
1502 			error = EBUSY;
1503 			PF_UNLOCK();
1504 			NET_UNLOCK();
1505 			goto fail;
1506 		}
1507 		memcpy(&pr->rule, rule, sizeof(struct pf_rule));
1508 		memset(&pr->rule.entries, 0, sizeof(pr->rule.entries));
1509 		pr->rule.kif = NULL;
1510 		pr->rule.nat.kif = NULL;
1511 		pr->rule.rdr.kif = NULL;
1512 		pr->rule.route.kif = NULL;
1513 		pr->rule.rcv_kif = NULL;
1514 		pr->rule.anchor = NULL;
1515 		pr->rule.overload_tbl = NULL;
1516 		pr->rule.pktrate.limit /= PF_THRESHOLD_MULT;
1517 		memset(&pr->rule.gcle, 0, sizeof(pr->rule.gcle));
1518 		pr->rule.ruleset = NULL;
1519 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1520 			error = EBUSY;
1521 			PF_UNLOCK();
1522 			NET_UNLOCK();
1523 			goto fail;
1524 		}
1525 		pf_addr_copyout(&pr->rule.src.addr);
1526 		pf_addr_copyout(&pr->rule.dst.addr);
1527 		pf_addr_copyout(&pr->rule.rdr.addr);
1528 		pf_addr_copyout(&pr->rule.nat.addr);
1529 		pf_addr_copyout(&pr->rule.route.addr);
1530 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1531 			if (rule->skip[i].ptr == NULL)
1532 				pr->rule.skip[i].nr = (u_int32_t)-1;
1533 			else
1534 				pr->rule.skip[i].nr =
1535 				    rule->skip[i].ptr->nr;
1536 
1537 		if (pr->action == PF_GET_CLR_CNTR) {
1538 			rule->evaluations = 0;
1539 			rule->packets[0] = rule->packets[1] = 0;
1540 			rule->bytes[0] = rule->bytes[1] = 0;
1541 			rule->states_tot = 0;
1542 		}
1543 		PF_UNLOCK();
1544 		NET_UNLOCK();
1545 		break;
1546 	}
1547 
1548 	case DIOCCHANGERULE: {
1549 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1550 		struct pf_ruleset	*ruleset;
1551 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1552 		u_int32_t		 nr = 0;
1553 
1554 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1555 		    pcr->action > PF_CHANGE_GET_TICKET) {
1556 			error = EINVAL;
1557 			goto fail;
1558 		}
1559 
1560 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1561 			NET_LOCK();
1562 			PF_LOCK();
1563 
1564 			ruleset = pf_find_ruleset(pcr->anchor);
1565 			if (ruleset == NULL)
1566 				error = EINVAL;
1567 			else
1568 				pcr->ticket = ++ruleset->rules.active.ticket;
1569 
1570 			PF_UNLOCK();
1571 			NET_UNLOCK();
1572 			goto fail;
1573 		}
1574 
1575 		if (pcr->action != PF_CHANGE_REMOVE) {
1576 			newrule = pool_get(&pf_rule_pl,
1577 			    PR_WAITOK|PR_LIMITFAIL|PR_ZERO);
1578 			if (newrule == NULL) {
1579 				error = ENOMEM;
1580 				goto fail;
1581 			}
1582 
1583 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1584 				error = EINVAL;
1585 				pool_put(&pf_rule_pl, newrule);
1586 				goto fail;
1587 			}
1588 			error = pf_rule_copyin(&pcr->rule, newrule);
1589 			if (error != 0) {
1590 				pf_rule_free(newrule);
1591 				newrule = NULL;
1592 				goto fail;
1593 			}
1594 			if ((error = pf_rule_checkaf(newrule))) {
1595 				pf_rule_free(newrule);
1596 				newrule = NULL;
1597 				goto fail;
1598 			}
1599 			if (newrule->rt && !newrule->direction) {
1600 				pf_rule_free(newrule);
1601 				error = EINVAL;
1602 				newrule = NULL;
1603 				goto fail;
1604 			}
1605 		}
1606 
1607 		NET_LOCK();
1608 		PF_LOCK();
1609 		ruleset = pf_find_ruleset(pcr->anchor);
1610 		if (ruleset == NULL) {
1611 			error = EINVAL;
1612 			PF_UNLOCK();
1613 			NET_UNLOCK();
1614 			pf_rule_free(newrule);
1615 			goto fail;
1616 		}
1617 
1618 		if (pcr->ticket != ruleset->rules.active.ticket) {
1619 			error = EINVAL;
1620 			PF_UNLOCK();
1621 			NET_UNLOCK();
1622 			pf_rule_free(newrule);
1623 			goto fail;
1624 		}
1625 
1626 		if (pcr->action != PF_CHANGE_REMOVE) {
1627 			KASSERT(newrule != NULL);
1628 			newrule->cuid = p->p_ucred->cr_ruid;
1629 			newrule->cpid = p->p_p->ps_pid;
1630 
1631 			newrule->kif = pf_kif_setup(newrule->kif);
1632 			newrule->rcv_kif = pf_kif_setup(newrule->rcv_kif);
1633 			newrule->rdr.kif = pf_kif_setup(newrule->rdr.kif);
1634 			newrule->nat.kif = pf_kif_setup(newrule->nat.kif);
1635 			newrule->route.kif = pf_kif_setup(newrule->route.kif);
1636 
1637 			if (newrule->overload_tblname[0]) {
1638 				newrule->overload_tbl = pfr_attach_table(
1639 				    ruleset, newrule->overload_tblname, 0);
1640 				if (newrule->overload_tbl == NULL)
1641 					error = EINVAL;
1642 				else
1643 					newrule->overload_tbl->pfrkt_flags |=
1644 					    PFR_TFLAG_ACTIVE;
1645 			}
1646 
1647 			if (pf_addr_setup(ruleset, &newrule->src.addr,
1648 			    newrule->af))
1649 				error = EINVAL;
1650 			if (pf_addr_setup(ruleset, &newrule->dst.addr,
1651 			    newrule->af))
1652 				error = EINVAL;
1653 			if (pf_addr_setup(ruleset, &newrule->rdr.addr,
1654 			    newrule->af))
1655 				error = EINVAL;
1656 			if (pf_addr_setup(ruleset, &newrule->nat.addr,
1657 			    newrule->af))
1658 				error = EINVAL;
1659 			if (pf_addr_setup(ruleset, &newrule->route.addr,
1660 			    newrule->af))
1661 				error = EINVAL;
1662 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1663 				error = EINVAL;
1664 
1665 			if (error) {
1666 				pf_rm_rule(NULL, newrule);
1667 				PF_UNLOCK();
1668 				NET_UNLOCK();
1669 				goto fail;
1670 			}
1671 		}
1672 
1673 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1674 			oldrule = TAILQ_FIRST(ruleset->rules.active.ptr);
1675 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1676 			oldrule = TAILQ_LAST(ruleset->rules.active.ptr,
1677 			    pf_rulequeue);
1678 		else {
1679 			oldrule = TAILQ_FIRST(ruleset->rules.active.ptr);
1680 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1681 				oldrule = TAILQ_NEXT(oldrule, entries);
1682 			if (oldrule == NULL) {
1683 				if (newrule != NULL)
1684 					pf_rm_rule(NULL, newrule);
1685 				error = EINVAL;
1686 				PF_UNLOCK();
1687 				NET_UNLOCK();
1688 				goto fail;
1689 			}
1690 		}
1691 
1692 		if (pcr->action == PF_CHANGE_REMOVE) {
1693 			pf_rm_rule(ruleset->rules.active.ptr, oldrule);
1694 			ruleset->rules.active.rcount--;
1695 		} else {
1696 			if (oldrule == NULL)
1697 				TAILQ_INSERT_TAIL(
1698 				    ruleset->rules.active.ptr,
1699 				    newrule, entries);
1700 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1701 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1702 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1703 			else
1704 				TAILQ_INSERT_AFTER(
1705 				    ruleset->rules.active.ptr,
1706 				    oldrule, newrule, entries);
1707 			ruleset->rules.active.rcount++;
1708 			newrule->ruleset = ruleset;
1709 		}
1710 
1711 		nr = 0;
1712 		TAILQ_FOREACH(oldrule, ruleset->rules.active.ptr, entries)
1713 			oldrule->nr = nr++;
1714 
1715 		ruleset->rules.active.ticket++;
1716 
1717 		pf_calc_skip_steps(ruleset->rules.active.ptr);
1718 		pf_remove_if_empty_ruleset(ruleset);
1719 
1720 		PF_UNLOCK();
1721 		NET_UNLOCK();
1722 		break;
1723 	}
1724 
1725 	case DIOCCLRSTATES:
1726 		error = pf_states_clr((struct pfioc_state_kill *)addr);
1727 		break;
1728 
1729 	case DIOCKILLSTATES: {
1730 		struct pf_state		*s, *nexts;
1731 		struct pf_state_item	*si, *sit;
1732 		struct pf_state_key	*sk, key;
1733 		struct pf_addr		*srcaddr, *dstaddr;
1734 		u_int16_t		 srcport, dstport;
1735 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1736 		u_int			 i, killed = 0;
1737 		const int		 dirs[] = { PF_IN, PF_OUT };
1738 		int			 sidx, didx;
1739 
1740 		if (psk->psk_pfcmp.id) {
1741 			if (psk->psk_pfcmp.creatorid == 0)
1742 				psk->psk_pfcmp.creatorid = pf_status.hostid;
1743 			NET_LOCK();
1744 			PF_LOCK();
1745 			PF_STATE_ENTER_WRITE();
1746 			if ((s = pf_find_state_byid(&psk->psk_pfcmp))) {
1747 				pf_remove_state(s);
1748 				psk->psk_killed = 1;
1749 			}
1750 			PF_STATE_EXIT_WRITE();
1751 			PF_UNLOCK();
1752 			NET_UNLOCK();
1753 			goto fail;
1754 		}
1755 
1756 		if (psk->psk_af && psk->psk_proto &&
1757 		    psk->psk_src.port_op == PF_OP_EQ &&
1758 		    psk->psk_dst.port_op == PF_OP_EQ) {
1759 
1760 			key.af = psk->psk_af;
1761 			key.proto = psk->psk_proto;
1762 			key.rdomain = psk->psk_rdomain;
1763 
1764 			NET_LOCK();
1765 			PF_LOCK();
1766 			PF_STATE_ENTER_WRITE();
1767 			for (i = 0; i < nitems(dirs); i++) {
1768 				if (dirs[i] == PF_IN) {
1769 					sidx = 0;
1770 					didx = 1;
1771 				} else {
1772 					sidx = 1;
1773 					didx = 0;
1774 				}
1775 				pf_addrcpy(&key.addr[sidx],
1776 				    &psk->psk_src.addr.v.a.addr, key.af);
1777 				pf_addrcpy(&key.addr[didx],
1778 				    &psk->psk_dst.addr.v.a.addr, key.af);
1779 				key.port[sidx] = psk->psk_src.port[0];
1780 				key.port[didx] = psk->psk_dst.port[0];
1781 
1782 				sk = RB_FIND(pf_state_tree, &pf_statetbl, &key);
1783 				if (sk == NULL)
1784 					continue;
1785 
1786 				TAILQ_FOREACH_SAFE(si, &sk->states, entry, sit)
1787 					if (((si->s->key[PF_SK_WIRE]->af ==
1788 					    si->s->key[PF_SK_STACK]->af &&
1789 					    sk == (dirs[i] == PF_IN ?
1790 					    si->s->key[PF_SK_WIRE] :
1791 					    si->s->key[PF_SK_STACK])) ||
1792 					    (si->s->key[PF_SK_WIRE]->af !=
1793 					    si->s->key[PF_SK_STACK]->af &&
1794 					    dirs[i] == PF_IN &&
1795 					    (sk == si->s->key[PF_SK_STACK] ||
1796 					    sk == si->s->key[PF_SK_WIRE]))) &&
1797 					    (!psk->psk_ifname[0] ||
1798 					    (si->s->kif != pfi_all &&
1799 					    !strcmp(psk->psk_ifname,
1800 					    si->s->kif->pfik_name)))) {
1801 						pf_remove_state(si->s);
1802 						killed++;
1803 					}
1804 			}
1805 			if (killed)
1806 				psk->psk_killed = killed;
1807 			PF_STATE_EXIT_WRITE();
1808 			PF_UNLOCK();
1809 			NET_UNLOCK();
1810 			goto fail;
1811 		}
1812 
1813 		NET_LOCK();
1814 		PF_LOCK();
1815 		PF_STATE_ENTER_WRITE();
1816 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s;
1817 		    s = nexts) {
1818 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1819 
1820 			if (s->direction == PF_OUT) {
1821 				sk = s->key[PF_SK_STACK];
1822 				srcaddr = &sk->addr[1];
1823 				dstaddr = &sk->addr[0];
1824 				srcport = sk->port[1];
1825 				dstport = sk->port[0];
1826 			} else {
1827 				sk = s->key[PF_SK_WIRE];
1828 				srcaddr = &sk->addr[0];
1829 				dstaddr = &sk->addr[1];
1830 				srcport = sk->port[0];
1831 				dstport = sk->port[1];
1832 			}
1833 			if ((!psk->psk_af || sk->af == psk->psk_af)
1834 			    && (!psk->psk_proto || psk->psk_proto ==
1835 			    sk->proto) && psk->psk_rdomain == sk->rdomain &&
1836 			    pf_match_addr(psk->psk_src.neg,
1837 			    &psk->psk_src.addr.v.a.addr,
1838 			    &psk->psk_src.addr.v.a.mask,
1839 			    srcaddr, sk->af) &&
1840 			    pf_match_addr(psk->psk_dst.neg,
1841 			    &psk->psk_dst.addr.v.a.addr,
1842 			    &psk->psk_dst.addr.v.a.mask,
1843 			    dstaddr, sk->af) &&
1844 			    (psk->psk_src.port_op == 0 ||
1845 			    pf_match_port(psk->psk_src.port_op,
1846 			    psk->psk_src.port[0], psk->psk_src.port[1],
1847 			    srcport)) &&
1848 			    (psk->psk_dst.port_op == 0 ||
1849 			    pf_match_port(psk->psk_dst.port_op,
1850 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1851 			    dstport)) &&
1852 			    (!psk->psk_label[0] || (s->rule.ptr->label[0] &&
1853 			    !strcmp(psk->psk_label, s->rule.ptr->label))) &&
1854 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1855 			    s->kif->pfik_name))) {
1856 				pf_remove_state(s);
1857 				killed++;
1858 			}
1859 		}
1860 		psk->psk_killed = killed;
1861 		PF_STATE_EXIT_WRITE();
1862 		PF_UNLOCK();
1863 		NET_UNLOCK();
1864 		break;
1865 	}
1866 
1867 #if NPFSYNC > 0
1868 	case DIOCADDSTATE: {
1869 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1870 		struct pfsync_state	*sp = &ps->state;
1871 
1872 		if (sp->timeout >= PFTM_MAX) {
1873 			error = EINVAL;
1874 			goto fail;
1875 		}
1876 		NET_LOCK();
1877 		PF_LOCK();
1878 		error = pfsync_state_import(sp, PFSYNC_SI_IOCTL);
1879 		PF_UNLOCK();
1880 		NET_UNLOCK();
1881 		break;
1882 	}
1883 #endif	/* NPFSYNC > 0 */
1884 
1885 	case DIOCGETSTATE: {
1886 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1887 		struct pf_state		*s;
1888 		struct pf_state_cmp	 id_key;
1889 
1890 		memset(&id_key, 0, sizeof(id_key));
1891 		id_key.id = ps->state.id;
1892 		id_key.creatorid = ps->state.creatorid;
1893 
1894 		NET_LOCK();
1895 		PF_STATE_ENTER_READ();
1896 		s = pf_find_state_byid(&id_key);
1897 		s = pf_state_ref(s);
1898 		PF_STATE_EXIT_READ();
1899 		NET_UNLOCK();
1900 		if (s == NULL) {
1901 			error = ENOENT;
1902 			goto fail;
1903 		}
1904 
1905 		pf_state_export(&ps->state, s);
1906 		pf_state_unref(s);
1907 		break;
1908 	}
1909 
1910 	case DIOCGETSTATES:
1911 		error = pf_states_get((struct pfioc_states *)addr);
1912 		break;
1913 
1914 	case DIOCGETSTATUS: {
1915 		struct pf_status *s = (struct pf_status *)addr;
1916 		NET_LOCK();
1917 		PF_LOCK();
1918 		memcpy(s, &pf_status, sizeof(struct pf_status));
1919 		pfi_update_status(s->ifname, s);
1920 		PF_UNLOCK();
1921 		NET_UNLOCK();
1922 		break;
1923 	}
1924 
1925 	case DIOCSETSTATUSIF: {
1926 		struct pfioc_iface	*pi = (struct pfioc_iface *)addr;
1927 
1928 		NET_LOCK();
1929 		PF_LOCK();
1930 		if (pi->pfiio_name[0] == 0) {
1931 			memset(pf_status.ifname, 0, IFNAMSIZ);
1932 			PF_UNLOCK();
1933 			NET_UNLOCK();
1934 			goto fail;
1935 		}
1936 		strlcpy(pf_trans_set.statusif, pi->pfiio_name, IFNAMSIZ);
1937 		pf_trans_set.mask |= PF_TSET_STATUSIF;
1938 		PF_UNLOCK();
1939 		NET_UNLOCK();
1940 		break;
1941 	}
1942 
1943 	case DIOCCLRSTATUS: {
1944 		struct pfioc_iface	*pi = (struct pfioc_iface *)addr;
1945 
1946 		NET_LOCK();
1947 		PF_LOCK();
1948 		/* if ifname is specified, clear counters there only */
1949 		if (pi->pfiio_name[0]) {
1950 			pfi_update_status(pi->pfiio_name, NULL);
1951 			PF_UNLOCK();
1952 			NET_UNLOCK();
1953 			goto fail;
1954 		}
1955 
1956 		memset(pf_status.counters, 0, sizeof(pf_status.counters));
1957 		memset(pf_status.fcounters, 0, sizeof(pf_status.fcounters));
1958 		memset(pf_status.scounters, 0, sizeof(pf_status.scounters));
1959 		pf_status.since = getuptime();
1960 
1961 		PF_UNLOCK();
1962 		NET_UNLOCK();
1963 		break;
1964 	}
1965 
1966 	case DIOCNATLOOK: {
1967 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1968 		struct pf_state_key	*sk;
1969 		struct pf_state		*state;
1970 		struct pf_state_key_cmp	 key;
1971 		int			 m = 0, direction = pnl->direction;
1972 		int			 sidx, didx;
1973 
1974 		switch (pnl->af) {
1975 		case AF_INET:
1976 			break;
1977 #ifdef INET6
1978 		case AF_INET6:
1979 			break;
1980 #endif /* INET6 */
1981 		default:
1982 			error = EAFNOSUPPORT;
1983 			goto fail;
1984 		}
1985 
1986 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1987 		sidx = (direction == PF_IN) ? 1 : 0;
1988 		didx = (direction == PF_IN) ? 0 : 1;
1989 
1990 		if (!pnl->proto ||
1991 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1992 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1993 		    ((pnl->proto == IPPROTO_TCP ||
1994 		    pnl->proto == IPPROTO_UDP) &&
1995 		    (!pnl->dport || !pnl->sport)) ||
1996 		    pnl->rdomain > RT_TABLEID_MAX)
1997 			error = EINVAL;
1998 		else {
1999 			key.af = pnl->af;
2000 			key.proto = pnl->proto;
2001 			key.rdomain = pnl->rdomain;
2002 			pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af);
2003 			key.port[sidx] = pnl->sport;
2004 			pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af);
2005 			key.port[didx] = pnl->dport;
2006 
2007 			NET_LOCK();
2008 			PF_STATE_ENTER_READ();
2009 			state = pf_find_state_all(&key, direction, &m);
2010 			state = pf_state_ref(state);
2011 			PF_STATE_EXIT_READ();
2012 			NET_UNLOCK();
2013 
2014 			if (m > 1)
2015 				error = E2BIG;	/* more than one state */
2016 			else if (state != NULL) {
2017 				sk = state->key[sidx];
2018 				pf_addrcpy(&pnl->rsaddr, &sk->addr[sidx],
2019 				    sk->af);
2020 				pnl->rsport = sk->port[sidx];
2021 				pf_addrcpy(&pnl->rdaddr, &sk->addr[didx],
2022 				    sk->af);
2023 				pnl->rdport = sk->port[didx];
2024 				pnl->rrdomain = sk->rdomain;
2025 			} else
2026 				error = ENOENT;
2027 			pf_state_unref(state);
2028 		}
2029 		break;
2030 	}
2031 
2032 	case DIOCSETTIMEOUT: {
2033 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2034 
2035 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2036 		    pt->seconds < 0) {
2037 			error = EINVAL;
2038 			goto fail;
2039 		}
2040 		NET_LOCK();
2041 		PF_LOCK();
2042 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2043 			pt->seconds = 1;
2044 		pf_default_rule_new.timeout[pt->timeout] = pt->seconds;
2045 		pt->seconds = pf_default_rule.timeout[pt->timeout];
2046 		PF_UNLOCK();
2047 		NET_UNLOCK();
2048 		break;
2049 	}
2050 
2051 	case DIOCGETTIMEOUT: {
2052 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2053 
2054 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2055 			error = EINVAL;
2056 			goto fail;
2057 		}
2058 		NET_LOCK();
2059 		PF_LOCK();
2060 		pt->seconds = pf_default_rule.timeout[pt->timeout];
2061 		PF_UNLOCK();
2062 		NET_UNLOCK();
2063 		break;
2064 	}
2065 
2066 	case DIOCGETLIMIT: {
2067 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2068 
2069 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2070 			error = EINVAL;
2071 			goto fail;
2072 		}
2073 		NET_LOCK();
2074 		PF_LOCK();
2075 		pl->limit = pf_pool_limits[pl->index].limit;
2076 		PF_UNLOCK();
2077 		NET_UNLOCK();
2078 		break;
2079 	}
2080 
2081 	case DIOCSETLIMIT: {
2082 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2083 
2084 		NET_LOCK();
2085 		PF_LOCK();
2086 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2087 		    pf_pool_limits[pl->index].pp == NULL) {
2088 			error = EINVAL;
2089 			PF_UNLOCK();
2090 			NET_UNLOCK();
2091 			goto fail;
2092 		}
2093 		if (((struct pool *)pf_pool_limits[pl->index].pp)->pr_nout >
2094 		    pl->limit) {
2095 			error = EBUSY;
2096 			PF_UNLOCK();
2097 			NET_UNLOCK();
2098 			goto fail;
2099 		}
2100 		/* Fragments reference mbuf clusters. */
2101 		if (pl->index == PF_LIMIT_FRAGS && pl->limit > nmbclust) {
2102 			error = EINVAL;
2103 			PF_UNLOCK();
2104 			NET_UNLOCK();
2105 			goto fail;
2106 		}
2107 
2108 		pf_pool_limits[pl->index].limit_new = pl->limit;
2109 		pl->limit = pf_pool_limits[pl->index].limit;
2110 		PF_UNLOCK();
2111 		NET_UNLOCK();
2112 		break;
2113 	}
2114 
2115 	case DIOCSETDEBUG: {
2116 		u_int32_t	*level = (u_int32_t *)addr;
2117 
2118 		NET_LOCK();
2119 		PF_LOCK();
2120 		pf_trans_set.debug = *level;
2121 		pf_trans_set.mask |= PF_TSET_DEBUG;
2122 		PF_UNLOCK();
2123 		NET_UNLOCK();
2124 		break;
2125 	}
2126 
2127 	case DIOCGETRULESETS: {
2128 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2129 		struct pf_ruleset	*ruleset;
2130 		struct pf_anchor	*anchor;
2131 
2132 		NET_LOCK();
2133 		PF_LOCK();
2134 		pr->path[sizeof(pr->path) - 1] = '\0';
2135 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2136 			error = EINVAL;
2137 			PF_UNLOCK();
2138 			NET_UNLOCK();
2139 			goto fail;
2140 		}
2141 		pr->nr = 0;
2142 		if (ruleset == &pf_main_ruleset) {
2143 			/* XXX kludge for pf_main_ruleset */
2144 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2145 				if (anchor->parent == NULL)
2146 					pr->nr++;
2147 		} else {
2148 			RB_FOREACH(anchor, pf_anchor_node,
2149 			    &ruleset->anchor->children)
2150 				pr->nr++;
2151 		}
2152 		PF_UNLOCK();
2153 		NET_UNLOCK();
2154 		break;
2155 	}
2156 
2157 	case DIOCGETRULESET: {
2158 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2159 		struct pf_ruleset	*ruleset;
2160 		struct pf_anchor	*anchor;
2161 		u_int32_t		 nr = 0;
2162 
2163 		NET_LOCK();
2164 		PF_LOCK();
2165 		pr->path[sizeof(pr->path) - 1] = '\0';
2166 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2167 			error = EINVAL;
2168 			PF_UNLOCK();
2169 			NET_UNLOCK();
2170 			goto fail;
2171 		}
2172 		pr->name[0] = '\0';
2173 		if (ruleset == &pf_main_ruleset) {
2174 			/* XXX kludge for pf_main_ruleset */
2175 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2176 				if (anchor->parent == NULL && nr++ == pr->nr) {
2177 					strlcpy(pr->name, anchor->name,
2178 					    sizeof(pr->name));
2179 					break;
2180 				}
2181 		} else {
2182 			RB_FOREACH(anchor, pf_anchor_node,
2183 			    &ruleset->anchor->children)
2184 				if (nr++ == pr->nr) {
2185 					strlcpy(pr->name, anchor->name,
2186 					    sizeof(pr->name));
2187 					break;
2188 				}
2189 		}
2190 		PF_UNLOCK();
2191 		NET_UNLOCK();
2192 		if (!pr->name[0])
2193 			error = EBUSY;
2194 		break;
2195 	}
2196 
2197 	case DIOCRCLRTABLES: {
2198 		struct pfioc_table *io = (struct pfioc_table *)addr;
2199 
2200 		if (io->pfrio_esize != 0) {
2201 			error = ENODEV;
2202 			goto fail;
2203 		}
2204 		NET_LOCK();
2205 		PF_LOCK();
2206 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2207 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2208 		PF_UNLOCK();
2209 		NET_UNLOCK();
2210 		break;
2211 	}
2212 
2213 	case DIOCRADDTABLES: {
2214 		struct pfioc_table *io = (struct pfioc_table *)addr;
2215 
2216 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2217 			error = ENODEV;
2218 			goto fail;
2219 		}
2220 		NET_LOCK();
2221 		PF_LOCK();
2222 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2223 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2224 		PF_UNLOCK();
2225 		NET_UNLOCK();
2226 		break;
2227 	}
2228 
2229 	case DIOCRDELTABLES: {
2230 		struct pfioc_table *io = (struct pfioc_table *)addr;
2231 
2232 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2233 			error = ENODEV;
2234 			goto fail;
2235 		}
2236 		NET_LOCK();
2237 		PF_LOCK();
2238 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2239 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2240 		PF_UNLOCK();
2241 		NET_UNLOCK();
2242 		break;
2243 	}
2244 
2245 	case DIOCRGETTABLES: {
2246 		struct pfioc_table *io = (struct pfioc_table *)addr;
2247 
2248 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2249 			error = ENODEV;
2250 			goto fail;
2251 		}
2252 		NET_LOCK();
2253 		PF_LOCK();
2254 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2255 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2256 		PF_UNLOCK();
2257 		NET_UNLOCK();
2258 		break;
2259 	}
2260 
2261 	case DIOCRGETTSTATS: {
2262 		struct pfioc_table *io = (struct pfioc_table *)addr;
2263 
2264 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2265 			error = ENODEV;
2266 			goto fail;
2267 		}
2268 		NET_LOCK();
2269 		PF_LOCK();
2270 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2271 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2272 		PF_UNLOCK();
2273 		NET_UNLOCK();
2274 		break;
2275 	}
2276 
2277 	case DIOCRCLRTSTATS: {
2278 		struct pfioc_table *io = (struct pfioc_table *)addr;
2279 
2280 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2281 			error = ENODEV;
2282 			goto fail;
2283 		}
2284 		NET_LOCK();
2285 		PF_LOCK();
2286 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2287 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2288 		PF_UNLOCK();
2289 		NET_UNLOCK();
2290 		break;
2291 	}
2292 
2293 	case DIOCRSETTFLAGS: {
2294 		struct pfioc_table *io = (struct pfioc_table *)addr;
2295 
2296 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2297 			error = ENODEV;
2298 			goto fail;
2299 		}
2300 		NET_LOCK();
2301 		PF_LOCK();
2302 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2303 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2304 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2305 		PF_UNLOCK();
2306 		NET_UNLOCK();
2307 		break;
2308 	}
2309 
2310 	case DIOCRCLRADDRS: {
2311 		struct pfioc_table *io = (struct pfioc_table *)addr;
2312 
2313 		if (io->pfrio_esize != 0) {
2314 			error = ENODEV;
2315 			goto fail;
2316 		}
2317 		NET_LOCK();
2318 		PF_LOCK();
2319 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2320 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2321 		PF_UNLOCK();
2322 		NET_UNLOCK();
2323 		break;
2324 	}
2325 
2326 	case DIOCRADDADDRS: {
2327 		struct pfioc_table *io = (struct pfioc_table *)addr;
2328 
2329 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2330 			error = ENODEV;
2331 			goto fail;
2332 		}
2333 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2334 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2335 		    PFR_FLAG_USERIOCTL);
2336 		break;
2337 	}
2338 
2339 	case DIOCRDELADDRS: {
2340 		struct pfioc_table *io = (struct pfioc_table *)addr;
2341 
2342 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2343 			error = ENODEV;
2344 			goto fail;
2345 		}
2346 		NET_LOCK();
2347 		PF_LOCK();
2348 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2349 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2350 		    PFR_FLAG_USERIOCTL);
2351 		PF_UNLOCK();
2352 		NET_UNLOCK();
2353 		break;
2354 	}
2355 
2356 	case DIOCRSETADDRS: {
2357 		struct pfioc_table *io = (struct pfioc_table *)addr;
2358 
2359 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2360 			error = ENODEV;
2361 			goto fail;
2362 		}
2363 		NET_LOCK();
2364 		PF_LOCK();
2365 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2366 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2367 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2368 		    PFR_FLAG_USERIOCTL, 0);
2369 		PF_UNLOCK();
2370 		NET_UNLOCK();
2371 		break;
2372 	}
2373 
2374 	case DIOCRGETADDRS: {
2375 		struct pfioc_table *io = (struct pfioc_table *)addr;
2376 
2377 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2378 			error = ENODEV;
2379 			goto fail;
2380 		}
2381 		NET_LOCK();
2382 		PF_LOCK();
2383 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2384 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2385 		PF_UNLOCK();
2386 		NET_UNLOCK();
2387 		break;
2388 	}
2389 
2390 	case DIOCRGETASTATS: {
2391 		struct pfioc_table *io = (struct pfioc_table *)addr;
2392 
2393 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2394 			error = ENODEV;
2395 			goto fail;
2396 		}
2397 		NET_LOCK();
2398 		PF_LOCK();
2399 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2400 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2401 		PF_UNLOCK();
2402 		NET_UNLOCK();
2403 		break;
2404 	}
2405 
2406 	case DIOCRCLRASTATS: {
2407 		struct pfioc_table *io = (struct pfioc_table *)addr;
2408 
2409 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2410 			error = ENODEV;
2411 			goto fail;
2412 		}
2413 		NET_LOCK();
2414 		PF_LOCK();
2415 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2416 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2417 		    PFR_FLAG_USERIOCTL);
2418 		PF_UNLOCK();
2419 		NET_UNLOCK();
2420 		break;
2421 	}
2422 
2423 	case DIOCRTSTADDRS: {
2424 		struct pfioc_table *io = (struct pfioc_table *)addr;
2425 
2426 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2427 			error = ENODEV;
2428 			goto fail;
2429 		}
2430 		NET_LOCK();
2431 		PF_LOCK();
2432 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2433 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2434 		    PFR_FLAG_USERIOCTL);
2435 		PF_UNLOCK();
2436 		NET_UNLOCK();
2437 		break;
2438 	}
2439 
2440 	case DIOCRINADEFINE: {
2441 		struct pfioc_table *io = (struct pfioc_table *)addr;
2442 
2443 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2444 			error = ENODEV;
2445 			goto fail;
2446 		}
2447 		NET_LOCK();
2448 		PF_LOCK();
2449 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2450 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2451 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2452 		PF_UNLOCK();
2453 		NET_UNLOCK();
2454 		break;
2455 	}
2456 
2457 	case DIOCOSFPADD: {
2458 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2459 		error = pf_osfp_add(io);
2460 		break;
2461 	}
2462 
2463 	case DIOCOSFPGET: {
2464 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2465 		error = pf_osfp_get(io);
2466 		break;
2467 	}
2468 
2469 	case DIOCXBEGIN: {
2470 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2471 		struct pfioc_trans_e	*ioe;
2472 		struct pfr_table	*table;
2473 		int			 i;
2474 
2475 		if (io->esize != sizeof(*ioe)) {
2476 			error = ENODEV;
2477 			goto fail;
2478 		}
2479 		ioe = malloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2480 		table = malloc(sizeof(*table), M_TEMP, M_WAITOK);
2481 		NET_LOCK();
2482 		PF_LOCK();
2483 		pf_default_rule_new = pf_default_rule;
2484 		PF_UNLOCK();
2485 		NET_UNLOCK();
2486 		memset(&pf_trans_set, 0, sizeof(pf_trans_set));
2487 		for (i = 0; i < io->size; i++) {
2488 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2489 				free(table, M_TEMP, sizeof(*table));
2490 				free(ioe, M_TEMP, sizeof(*ioe));
2491 				error = EFAULT;
2492 				goto fail;
2493 			}
2494 			if (strnlen(ioe->anchor, sizeof(ioe->anchor)) ==
2495 			    sizeof(ioe->anchor)) {
2496 				free(table, M_TEMP, sizeof(*table));
2497 				free(ioe, M_TEMP, sizeof(*ioe));
2498 				error = ENAMETOOLONG;
2499 				goto fail;
2500 			}
2501 			NET_LOCK();
2502 			PF_LOCK();
2503 			switch (ioe->type) {
2504 			case PF_TRANS_TABLE:
2505 				memset(table, 0, sizeof(*table));
2506 				strlcpy(table->pfrt_anchor, ioe->anchor,
2507 				    sizeof(table->pfrt_anchor));
2508 				if ((error = pfr_ina_begin(table,
2509 				    &ioe->ticket, NULL, 0))) {
2510 					PF_UNLOCK();
2511 					NET_UNLOCK();
2512 					free(table, M_TEMP, sizeof(*table));
2513 					free(ioe, M_TEMP, sizeof(*ioe));
2514 					goto fail;
2515 				}
2516 				break;
2517 			case PF_TRANS_RULESET:
2518 				if ((error = pf_begin_rules(&ioe->ticket,
2519 				    ioe->anchor))) {
2520 					PF_UNLOCK();
2521 					NET_UNLOCK();
2522 					free(table, M_TEMP, sizeof(*table));
2523 					free(ioe, M_TEMP, sizeof(*ioe));
2524 					goto fail;
2525 				}
2526 				break;
2527 			default:
2528 				PF_UNLOCK();
2529 				NET_UNLOCK();
2530 				free(table, M_TEMP, sizeof(*table));
2531 				free(ioe, M_TEMP, sizeof(*ioe));
2532 				error = EINVAL;
2533 				goto fail;
2534 			}
2535 			PF_UNLOCK();
2536 			NET_UNLOCK();
2537 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2538 				free(table, M_TEMP, sizeof(*table));
2539 				free(ioe, M_TEMP, sizeof(*ioe));
2540 				error = EFAULT;
2541 				goto fail;
2542 			}
2543 		}
2544 		free(table, M_TEMP, sizeof(*table));
2545 		free(ioe, M_TEMP, sizeof(*ioe));
2546 		break;
2547 	}
2548 
2549 	case DIOCXROLLBACK: {
2550 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2551 		struct pfioc_trans_e	*ioe;
2552 		struct pfr_table	*table;
2553 		int			 i;
2554 
2555 		if (io->esize != sizeof(*ioe)) {
2556 			error = ENODEV;
2557 			goto fail;
2558 		}
2559 		ioe = malloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2560 		table = malloc(sizeof(*table), M_TEMP, M_WAITOK);
2561 		for (i = 0; i < io->size; i++) {
2562 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2563 				free(table, M_TEMP, sizeof(*table));
2564 				free(ioe, M_TEMP, sizeof(*ioe));
2565 				error = EFAULT;
2566 				goto fail;
2567 			}
2568 			if (strnlen(ioe->anchor, sizeof(ioe->anchor)) ==
2569 			    sizeof(ioe->anchor)) {
2570 				free(table, M_TEMP, sizeof(*table));
2571 				free(ioe, M_TEMP, sizeof(*ioe));
2572 				error = ENAMETOOLONG;
2573 				goto fail;
2574 			}
2575 			NET_LOCK();
2576 			PF_LOCK();
2577 			switch (ioe->type) {
2578 			case PF_TRANS_TABLE:
2579 				memset(table, 0, sizeof(*table));
2580 				strlcpy(table->pfrt_anchor, ioe->anchor,
2581 				    sizeof(table->pfrt_anchor));
2582 				if ((error = pfr_ina_rollback(table,
2583 				    ioe->ticket, NULL, 0))) {
2584 					PF_UNLOCK();
2585 					NET_UNLOCK();
2586 					free(table, M_TEMP, sizeof(*table));
2587 					free(ioe, M_TEMP, sizeof(*ioe));
2588 					goto fail; /* really bad */
2589 				}
2590 				break;
2591 			case PF_TRANS_RULESET:
2592 				pf_rollback_rules(ioe->ticket, ioe->anchor);
2593 				break;
2594 			default:
2595 				PF_UNLOCK();
2596 				NET_UNLOCK();
2597 				free(table, M_TEMP, sizeof(*table));
2598 				free(ioe, M_TEMP, sizeof(*ioe));
2599 				error = EINVAL;
2600 				goto fail; /* really bad */
2601 			}
2602 			PF_UNLOCK();
2603 			NET_UNLOCK();
2604 		}
2605 		free(table, M_TEMP, sizeof(*table));
2606 		free(ioe, M_TEMP, sizeof(*ioe));
2607 		break;
2608 	}
2609 
2610 	case DIOCXCOMMIT: {
2611 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2612 		struct pfioc_trans_e	*ioe;
2613 		struct pfr_table	*table;
2614 		struct pf_ruleset	*rs;
2615 		int			 i;
2616 
2617 		if (io->esize != sizeof(*ioe)) {
2618 			error = ENODEV;
2619 			goto fail;
2620 		}
2621 		ioe = malloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2622 		table = malloc(sizeof(*table), M_TEMP, M_WAITOK);
2623 		NET_LOCK();
2624 		PF_LOCK();
2625 		/* first makes sure everything will succeed */
2626 		for (i = 0; i < io->size; i++) {
2627 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2628 				PF_UNLOCK();
2629 				NET_UNLOCK();
2630 				free(table, M_TEMP, sizeof(*table));
2631 				free(ioe, M_TEMP, sizeof(*ioe));
2632 				error = EFAULT;
2633 				goto fail;
2634 			}
2635 			if (strnlen(ioe->anchor, sizeof(ioe->anchor)) ==
2636 			    sizeof(ioe->anchor)) {
2637 				PF_UNLOCK();
2638 				NET_UNLOCK();
2639 				free(table, M_TEMP, sizeof(*table));
2640 				free(ioe, M_TEMP, sizeof(*ioe));
2641 				error = ENAMETOOLONG;
2642 				goto fail;
2643 			}
2644 			switch (ioe->type) {
2645 			case PF_TRANS_TABLE:
2646 				rs = pf_find_ruleset(ioe->anchor);
2647 				if (rs == NULL || !rs->topen || ioe->ticket !=
2648 				     rs->tticket) {
2649 					PF_UNLOCK();
2650 					NET_UNLOCK();
2651 					free(table, M_TEMP, sizeof(*table));
2652 					free(ioe, M_TEMP, sizeof(*ioe));
2653 					error = EBUSY;
2654 					goto fail;
2655 				}
2656 				break;
2657 			case PF_TRANS_RULESET:
2658 				rs = pf_find_ruleset(ioe->anchor);
2659 				if (rs == NULL ||
2660 				    !rs->rules.inactive.open ||
2661 				    rs->rules.inactive.ticket !=
2662 				    ioe->ticket) {
2663 					PF_UNLOCK();
2664 					NET_UNLOCK();
2665 					free(table, M_TEMP, sizeof(*table));
2666 					free(ioe, M_TEMP, sizeof(*ioe));
2667 					error = EBUSY;
2668 					goto fail;
2669 				}
2670 				break;
2671 			default:
2672 				PF_UNLOCK();
2673 				NET_UNLOCK();
2674 				free(table, M_TEMP, sizeof(*table));
2675 				free(ioe, M_TEMP, sizeof(*ioe));
2676 				error = EINVAL;
2677 				goto fail;
2678 			}
2679 		}
2680 
2681 		/*
2682 		 * Checked already in DIOCSETLIMIT, but check again as the
2683 		 * situation might have changed.
2684 		 */
2685 		for (i = 0; i < PF_LIMIT_MAX; i++) {
2686 			if (((struct pool *)pf_pool_limits[i].pp)->pr_nout >
2687 			    pf_pool_limits[i].limit_new) {
2688 				PF_UNLOCK();
2689 				NET_UNLOCK();
2690 				free(table, M_TEMP, sizeof(*table));
2691 				free(ioe, M_TEMP, sizeof(*ioe));
2692 				error = EBUSY;
2693 				goto fail;
2694 			}
2695 		}
2696 		/* now do the commit - no errors should happen here */
2697 		for (i = 0; i < io->size; i++) {
2698 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2699 				PF_UNLOCK();
2700 				NET_UNLOCK();
2701 				free(table, M_TEMP, sizeof(*table));
2702 				free(ioe, M_TEMP, sizeof(*ioe));
2703 				error = EFAULT;
2704 				goto fail;
2705 			}
2706 			if (strnlen(ioe->anchor, sizeof(ioe->anchor)) ==
2707 			    sizeof(ioe->anchor)) {
2708 				PF_UNLOCK();
2709 				NET_UNLOCK();
2710 				free(table, M_TEMP, sizeof(*table));
2711 				free(ioe, M_TEMP, sizeof(*ioe));
2712 				error = ENAMETOOLONG;
2713 				goto fail;
2714 			}
2715 			switch (ioe->type) {
2716 			case PF_TRANS_TABLE:
2717 				memset(table, 0, sizeof(*table));
2718 				strlcpy(table->pfrt_anchor, ioe->anchor,
2719 				    sizeof(table->pfrt_anchor));
2720 				if ((error = pfr_ina_commit(table, ioe->ticket,
2721 				    NULL, NULL, 0))) {
2722 					PF_UNLOCK();
2723 					NET_UNLOCK();
2724 					free(table, M_TEMP, sizeof(*table));
2725 					free(ioe, M_TEMP, sizeof(*ioe));
2726 					goto fail; /* really bad */
2727 				}
2728 				break;
2729 			case PF_TRANS_RULESET:
2730 				if ((error = pf_commit_rules(ioe->ticket,
2731 				    ioe->anchor))) {
2732 					PF_UNLOCK();
2733 					NET_UNLOCK();
2734 					free(table, M_TEMP, sizeof(*table));
2735 					free(ioe, M_TEMP, sizeof(*ioe));
2736 					goto fail; /* really bad */
2737 				}
2738 				break;
2739 			default:
2740 				PF_UNLOCK();
2741 				NET_UNLOCK();
2742 				free(table, M_TEMP, sizeof(*table));
2743 				free(ioe, M_TEMP, sizeof(*ioe));
2744 				error = EINVAL;
2745 				goto fail; /* really bad */
2746 			}
2747 		}
2748 		for (i = 0; i < PF_LIMIT_MAX; i++) {
2749 			if (pf_pool_limits[i].limit_new !=
2750 			    pf_pool_limits[i].limit &&
2751 			    pool_sethardlimit(pf_pool_limits[i].pp,
2752 			    pf_pool_limits[i].limit_new, NULL, 0) != 0) {
2753 				PF_UNLOCK();
2754 				NET_UNLOCK();
2755 				free(table, M_TEMP, sizeof(*table));
2756 				free(ioe, M_TEMP, sizeof(*ioe));
2757 				error = EBUSY;
2758 				goto fail; /* really bad */
2759 			}
2760 			pf_pool_limits[i].limit = pf_pool_limits[i].limit_new;
2761 		}
2762 		for (i = 0; i < PFTM_MAX; i++) {
2763 			int old = pf_default_rule.timeout[i];
2764 
2765 			pf_default_rule.timeout[i] =
2766 			    pf_default_rule_new.timeout[i];
2767 			if (pf_default_rule.timeout[i] == PFTM_INTERVAL &&
2768 			    pf_default_rule.timeout[i] < old)
2769 				task_add(net_tq(0), &pf_purge_task);
2770 		}
2771 		pfi_xcommit();
2772 		pf_trans_set_commit();
2773 		PF_UNLOCK();
2774 		NET_UNLOCK();
2775 		free(table, M_TEMP, sizeof(*table));
2776 		free(ioe, M_TEMP, sizeof(*ioe));
2777 		break;
2778 	}
2779 
2780 	case DIOCGETSRCNODES: {
2781 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2782 		struct pf_src_node	*n, *p, *pstore;
2783 		u_int32_t		 nr = 0;
2784 		size_t			 space = psn->psn_len;
2785 
2786 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2787 
2788 		NET_LOCK();
2789 		PF_LOCK();
2790 		if (space == 0) {
2791 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2792 				nr++;
2793 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2794 			PF_UNLOCK();
2795 			NET_UNLOCK();
2796 			free(pstore, M_TEMP, sizeof(*pstore));
2797 			goto fail;
2798 		}
2799 
2800 		p = psn->psn_src_nodes;
2801 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2802 			int	secs = getuptime(), diff;
2803 
2804 			if ((nr + 1) * sizeof(*p) > psn->psn_len)
2805 				break;
2806 
2807 			memcpy(pstore, n, sizeof(*pstore));
2808 			memset(&pstore->entry, 0, sizeof(pstore->entry));
2809 			pstore->rule.ptr = NULL;
2810 			pstore->kif = NULL;
2811 			pstore->rule.nr = n->rule.ptr->nr;
2812 			pstore->creation = secs - pstore->creation;
2813 			if (pstore->expire > secs)
2814 				pstore->expire -= secs;
2815 			else
2816 				pstore->expire = 0;
2817 
2818 			/* adjust the connection rate estimate */
2819 			diff = secs - n->conn_rate.last;
2820 			if (diff >= n->conn_rate.seconds)
2821 				pstore->conn_rate.count = 0;
2822 			else
2823 				pstore->conn_rate.count -=
2824 				    n->conn_rate.count * diff /
2825 				    n->conn_rate.seconds;
2826 
2827 			error = copyout(pstore, p, sizeof(*p));
2828 			if (error) {
2829 				PF_UNLOCK();
2830 				NET_UNLOCK();
2831 				free(pstore, M_TEMP, sizeof(*pstore));
2832 				goto fail;
2833 			}
2834 			p++;
2835 			nr++;
2836 		}
2837 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2838 
2839 		PF_UNLOCK();
2840 		NET_UNLOCK();
2841 		free(pstore, M_TEMP, sizeof(*pstore));
2842 		break;
2843 	}
2844 
2845 	case DIOCCLRSRCNODES: {
2846 		struct pf_src_node	*n;
2847 		struct pf_state		*state;
2848 
2849 		NET_LOCK();
2850 		PF_LOCK();
2851 		PF_STATE_ENTER_WRITE();
2852 		RB_FOREACH(state, pf_state_tree_id, &tree_id)
2853 			pf_src_tree_remove_state(state);
2854 		PF_STATE_EXIT_WRITE();
2855 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2856 			n->expire = 1;
2857 		pf_purge_expired_src_nodes();
2858 		PF_UNLOCK();
2859 		NET_UNLOCK();
2860 		break;
2861 	}
2862 
2863 	case DIOCKILLSRCNODES: {
2864 		struct pf_src_node	*sn;
2865 		struct pf_state		*s;
2866 		struct pfioc_src_node_kill *psnk =
2867 		    (struct pfioc_src_node_kill *)addr;
2868 		u_int			killed = 0;
2869 
2870 		NET_LOCK();
2871 		PF_LOCK();
2872 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
2873 			if (pf_match_addr(psnk->psnk_src.neg,
2874 				&psnk->psnk_src.addr.v.a.addr,
2875 				&psnk->psnk_src.addr.v.a.mask,
2876 				&sn->addr, sn->af) &&
2877 			    pf_match_addr(psnk->psnk_dst.neg,
2878 				&psnk->psnk_dst.addr.v.a.addr,
2879 				&psnk->psnk_dst.addr.v.a.mask,
2880 				&sn->raddr, sn->af)) {
2881 				/* Handle state to src_node linkage */
2882 				if (sn->states != 0) {
2883 					PF_ASSERT_LOCKED();
2884 					PF_STATE_ENTER_WRITE();
2885 					RB_FOREACH(s, pf_state_tree_id,
2886 					   &tree_id)
2887 						pf_state_rm_src_node(s, sn);
2888 					PF_STATE_EXIT_WRITE();
2889 				}
2890 				sn->expire = 1;
2891 				killed++;
2892 			}
2893 		}
2894 
2895 		if (killed > 0)
2896 			pf_purge_expired_src_nodes();
2897 
2898 		psnk->psnk_killed = killed;
2899 		PF_UNLOCK();
2900 		NET_UNLOCK();
2901 		break;
2902 	}
2903 
2904 	case DIOCSETHOSTID: {
2905 		u_int32_t	*hostid = (u_int32_t *)addr;
2906 
2907 		NET_LOCK();
2908 		PF_LOCK();
2909 		if (*hostid == 0)
2910 			pf_trans_set.hostid = arc4random();
2911 		else
2912 			pf_trans_set.hostid = *hostid;
2913 		pf_trans_set.mask |= PF_TSET_HOSTID;
2914 		PF_UNLOCK();
2915 		NET_UNLOCK();
2916 		break;
2917 	}
2918 
2919 	case DIOCOSFPFLUSH:
2920 		pf_osfp_flush();
2921 		break;
2922 
2923 	case DIOCIGETIFACES: {
2924 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2925 
2926 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
2927 			error = ENODEV;
2928 			goto fail;
2929 		}
2930 		NET_LOCK();
2931 		PF_LOCK();
2932 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2933 		    &io->pfiio_size);
2934 		PF_UNLOCK();
2935 		NET_UNLOCK();
2936 		break;
2937 	}
2938 
2939 	case DIOCSETIFFLAG: {
2940 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2941 
2942 		if (io == NULL) {
2943 			error = EINVAL;
2944 			goto fail;
2945 		}
2946 
2947 		NET_LOCK();
2948 		PF_LOCK();
2949 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
2950 		PF_UNLOCK();
2951 		NET_UNLOCK();
2952 		break;
2953 	}
2954 
2955 	case DIOCCLRIFFLAG: {
2956 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2957 
2958 		if (io == NULL) {
2959 			error = EINVAL;
2960 			goto fail;
2961 		}
2962 
2963 		NET_LOCK();
2964 		PF_LOCK();
2965 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
2966 		PF_UNLOCK();
2967 		NET_UNLOCK();
2968 		break;
2969 	}
2970 
2971 	case DIOCSETREASS: {
2972 		u_int32_t	*reass = (u_int32_t *)addr;
2973 
2974 		NET_LOCK();
2975 		PF_LOCK();
2976 		pf_trans_set.reass = *reass;
2977 		pf_trans_set.mask |= PF_TSET_REASS;
2978 		PF_UNLOCK();
2979 		NET_UNLOCK();
2980 		break;
2981 	}
2982 
2983 	case DIOCSETSYNFLWATS: {
2984 		struct pfioc_synflwats *io = (struct pfioc_synflwats *)addr;
2985 
2986 		NET_LOCK();
2987 		PF_LOCK();
2988 		error = pf_syncookies_setwats(io->hiwat, io->lowat);
2989 		PF_UNLOCK();
2990 		NET_UNLOCK();
2991 		break;
2992 	}
2993 
2994 	case DIOCGETSYNFLWATS: {
2995 		struct pfioc_synflwats *io = (struct pfioc_synflwats *)addr;
2996 
2997 		NET_LOCK();
2998 		PF_LOCK();
2999 		error = pf_syncookies_getwats(io);
3000 		PF_UNLOCK();
3001 		NET_UNLOCK();
3002 		break;
3003 	}
3004 
3005 	case DIOCSETSYNCOOKIES: {
3006 		u_int8_t	*mode = (u_int8_t *)addr;
3007 
3008 		NET_LOCK();
3009 		PF_LOCK();
3010 		error = pf_syncookies_setmode(*mode);
3011 		PF_UNLOCK();
3012 		NET_UNLOCK();
3013 		break;
3014 	}
3015 
3016 	default:
3017 		error = ENODEV;
3018 		break;
3019 	}
3020 fail:
3021 	if (flags & FWRITE)
3022 		rw_exit_write(&pfioctl_rw);
3023 	else
3024 		rw_exit_read(&pfioctl_rw);
3025 
3026 	return (error);
3027 }
3028 
3029 void
3030 pf_trans_set_commit(void)
3031 {
3032 	if (pf_trans_set.mask & PF_TSET_STATUSIF)
3033 		strlcpy(pf_status.ifname, pf_trans_set.statusif, IFNAMSIZ);
3034 	if (pf_trans_set.mask & PF_TSET_DEBUG)
3035 		pf_status.debug = pf_trans_set.debug;
3036 	if (pf_trans_set.mask & PF_TSET_HOSTID)
3037 		pf_status.hostid = pf_trans_set.hostid;
3038 	if (pf_trans_set.mask & PF_TSET_REASS)
3039 		pf_status.reass = pf_trans_set.reass;
3040 }
3041 
3042 void
3043 pf_pool_copyin(struct pf_pool *from, struct pf_pool *to)
3044 {
3045 	memmove(to, from, sizeof(*to));
3046 	to->kif = NULL;
3047 	to->addr.p.tbl = NULL;
3048 }
3049 
3050 int
3051 pf_validate_range(u_int8_t op, u_int16_t port[2], int order)
3052 {
3053 	u_int16_t a = (order == PF_ORDER_NET) ? ntohs(port[0]) : port[0];
3054 	u_int16_t b = (order == PF_ORDER_NET) ? ntohs(port[1]) : port[1];
3055 
3056 	if ((op == PF_OP_RRG && a > b) ||  /* 34:12,  i.e. none */
3057 	    (op == PF_OP_IRG && a >= b) || /* 34><12, i.e. none */
3058 	    (op == PF_OP_XRG && a > b))    /* 34<>22, i.e. all */
3059 		return 1;
3060 	return 0;
3061 }
3062 
3063 int
3064 pf_rule_copyin(struct pf_rule *from, struct pf_rule *to)
3065 {
3066 	int i;
3067 
3068 	if (from->scrub_flags & PFSTATE_SETPRIO &&
3069 	    (from->set_prio[0] > IFQ_MAXPRIO ||
3070 	    from->set_prio[1] > IFQ_MAXPRIO))
3071 		return (EINVAL);
3072 
3073 	to->src = from->src;
3074 	to->src.addr.p.tbl = NULL;
3075 	to->dst = from->dst;
3076 	to->dst.addr.p.tbl = NULL;
3077 
3078 	if (pf_validate_range(to->src.port_op, to->src.port, PF_ORDER_NET))
3079 		return (EINVAL);
3080 	if (pf_validate_range(to->dst.port_op, to->dst.port, PF_ORDER_NET))
3081 		return (EINVAL);
3082 
3083 	/* XXX union skip[] */
3084 
3085 	strlcpy(to->label, from->label, sizeof(to->label));
3086 	strlcpy(to->ifname, from->ifname, sizeof(to->ifname));
3087 	strlcpy(to->rcv_ifname, from->rcv_ifname, sizeof(to->rcv_ifname));
3088 	strlcpy(to->qname, from->qname, sizeof(to->qname));
3089 	strlcpy(to->pqname, from->pqname, sizeof(to->pqname));
3090 	strlcpy(to->tagname, from->tagname, sizeof(to->tagname));
3091 	strlcpy(to->match_tagname, from->match_tagname,
3092 	    sizeof(to->match_tagname));
3093 	strlcpy(to->overload_tblname, from->overload_tblname,
3094 	    sizeof(to->overload_tblname));
3095 
3096 	pf_pool_copyin(&from->nat, &to->nat);
3097 	pf_pool_copyin(&from->rdr, &to->rdr);
3098 	pf_pool_copyin(&from->route, &to->route);
3099 
3100 	if (pf_validate_range(to->rdr.port_op, to->rdr.proxy_port,
3101 	    PF_ORDER_HOST))
3102 		return (EINVAL);
3103 
3104 	to->kif = (to->ifname[0]) ?
3105 	    pfi_kif_alloc(to->ifname, M_WAITOK) : NULL;
3106 	to->rcv_kif = (to->rcv_ifname[0]) ?
3107 	    pfi_kif_alloc(to->rcv_ifname, M_WAITOK) : NULL;
3108 	to->rdr.kif = (to->rdr.ifname[0]) ?
3109 	    pfi_kif_alloc(to->rdr.ifname, M_WAITOK) : NULL;
3110 	to->nat.kif = (to->nat.ifname[0]) ?
3111 	    pfi_kif_alloc(to->nat.ifname, M_WAITOK) : NULL;
3112 	to->route.kif = (to->route.ifname[0]) ?
3113 	    pfi_kif_alloc(to->route.ifname, M_WAITOK) : NULL;
3114 
3115 	to->os_fingerprint = from->os_fingerprint;
3116 
3117 	to->rtableid = from->rtableid;
3118 	if (to->rtableid >= 0 && !rtable_exists(to->rtableid))
3119 		return (EBUSY);
3120 	to->onrdomain = from->onrdomain;
3121 	if (to->onrdomain != -1 && (to->onrdomain < 0 ||
3122 	    to->onrdomain > RT_TABLEID_MAX))
3123 		return (EINVAL);
3124 
3125 	for (i = 0; i < PFTM_MAX; i++)
3126 		to->timeout[i] = from->timeout[i];
3127 	to->states_tot = from->states_tot;
3128 	to->max_states = from->max_states;
3129 	to->max_src_nodes = from->max_src_nodes;
3130 	to->max_src_states = from->max_src_states;
3131 	to->max_src_conn = from->max_src_conn;
3132 	to->max_src_conn_rate.limit = from->max_src_conn_rate.limit;
3133 	to->max_src_conn_rate.seconds = from->max_src_conn_rate.seconds;
3134 	pf_init_threshold(&to->pktrate, from->pktrate.limit,
3135 	    from->pktrate.seconds);
3136 
3137 	if (to->qname[0] != 0) {
3138 		if ((to->qid = pf_qname2qid(to->qname, 0)) == 0)
3139 			return (EBUSY);
3140 		if (to->pqname[0] != 0) {
3141 			if ((to->pqid = pf_qname2qid(to->pqname, 0)) == 0)
3142 				return (EBUSY);
3143 		} else
3144 			to->pqid = to->qid;
3145 	}
3146 	to->rt_listid = from->rt_listid;
3147 	to->prob = from->prob;
3148 	to->return_icmp = from->return_icmp;
3149 	to->return_icmp6 = from->return_icmp6;
3150 	to->max_mss = from->max_mss;
3151 	if (to->tagname[0])
3152 		if ((to->tag = pf_tagname2tag(to->tagname, 1)) == 0)
3153 			return (EBUSY);
3154 	if (to->match_tagname[0])
3155 		if ((to->match_tag = pf_tagname2tag(to->match_tagname, 1)) == 0)
3156 			return (EBUSY);
3157 	to->scrub_flags = from->scrub_flags;
3158 	to->delay = from->delay;
3159 	to->uid = from->uid;
3160 	to->gid = from->gid;
3161 	to->rule_flag = from->rule_flag;
3162 	to->action = from->action;
3163 	to->direction = from->direction;
3164 	to->log = from->log;
3165 	to->logif = from->logif;
3166 #if NPFLOG > 0
3167 	if (!to->log)
3168 		to->logif = 0;
3169 #endif	/* NPFLOG > 0 */
3170 	to->quick = from->quick;
3171 	to->ifnot = from->ifnot;
3172 	to->rcvifnot = from->rcvifnot;
3173 	to->match_tag_not = from->match_tag_not;
3174 	to->keep_state = from->keep_state;
3175 	to->af = from->af;
3176 	to->naf = from->naf;
3177 	to->proto = from->proto;
3178 	to->type = from->type;
3179 	to->code = from->code;
3180 	to->flags = from->flags;
3181 	to->flagset = from->flagset;
3182 	to->min_ttl = from->min_ttl;
3183 	to->allow_opts = from->allow_opts;
3184 	to->rt = from->rt;
3185 	to->return_ttl = from->return_ttl;
3186 	to->tos = from->tos;
3187 	to->set_tos = from->set_tos;
3188 	to->anchor_relative = from->anchor_relative; /* XXX */
3189 	to->anchor_wildcard = from->anchor_wildcard; /* XXX */
3190 	to->flush = from->flush;
3191 	to->divert.addr = from->divert.addr;
3192 	to->divert.port = from->divert.port;
3193 	to->divert.type = from->divert.type;
3194 	to->prio = from->prio;
3195 	to->set_prio[0] = from->set_prio[0];
3196 	to->set_prio[1] = from->set_prio[1];
3197 
3198 	return (0);
3199 }
3200 
3201 int
3202 pf_rule_checkaf(struct pf_rule *r)
3203 {
3204 	switch (r->af) {
3205 	case 0:
3206 		if (r->rule_flag & PFRULE_AFTO)
3207 			return (EPFNOSUPPORT);
3208 		break;
3209 	case AF_INET:
3210 		if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET6)
3211 			return (EPFNOSUPPORT);
3212 		break;
3213 #ifdef INET6
3214 	case AF_INET6:
3215 		if ((r->rule_flag & PFRULE_AFTO) && r->naf != AF_INET)
3216 			return (EPFNOSUPPORT);
3217 		break;
3218 #endif /* INET6 */
3219 	default:
3220 		return (EPFNOSUPPORT);
3221 	}
3222 
3223 	if ((r->rule_flag & PFRULE_AFTO) == 0 && r->naf != 0)
3224 		return (EPFNOSUPPORT);
3225 
3226 	return (0);
3227 }
3228 
3229 int
3230 pf_sysctl(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
3231 {
3232 	struct pf_status	pfs;
3233 
3234 	NET_RLOCK_IN_IOCTL();
3235 	PF_LOCK();
3236 	memcpy(&pfs, &pf_status, sizeof(struct pf_status));
3237 	pfi_update_status(pfs.ifname, &pfs);
3238 	PF_UNLOCK();
3239 	NET_RUNLOCK_IN_IOCTL();
3240 
3241 	return sysctl_rdstruct(oldp, oldlenp, newp, &pfs, sizeof(pfs));
3242 }
3243