xref: /openbsd-src/sys/net/pf_ioctl.c (revision 4b70baf6e17fc8b27fc1f7fa7929335753fa94c3)
1 /*	$OpenBSD: pf_ioctl.c,v 1.343 2019/02/18 13:11:44 bluhm 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/mbuf.h>
44 #include <sys/filio.h>
45 #include <sys/fcntl.h>
46 #include <sys/socket.h>
47 #include <sys/socketvar.h>
48 #include <sys/kernel.h>
49 #include <sys/time.h>
50 #include <sys/timeout.h>
51 #include <sys/pool.h>
52 #include <sys/malloc.h>
53 #include <sys/kthread.h>
54 #include <sys/rwlock.h>
55 #include <sys/syslog.h>
56 #include <uvm/uvm_extern.h>
57 
58 #include <crypto/md5.h>
59 
60 #include <net/if.h>
61 #include <net/if_var.h>
62 #include <net/route.h>
63 #include <net/hfsc.h>
64 #include <net/fq_codel.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/ip.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/ip_icmp.h>
71 #include <netinet/tcp.h>
72 #include <netinet/udp.h>
73 
74 #ifdef INET6
75 #include <netinet/ip6.h>
76 #include <netinet/icmp6.h>
77 #endif /* INET6 */
78 
79 #include <net/pfvar.h>
80 #include <net/pfvar_priv.h>
81 
82 #if NPFSYNC > 0
83 #include <netinet/ip_ipsp.h>
84 #include <net/if_pfsync.h>
85 #endif /* NPFSYNC > 0 */
86 
87 struct pool		 pf_tag_pl;
88 
89 void			 pfattach(int);
90 void			 pf_thread_create(void *);
91 int			 pfopen(dev_t, int, int, struct proc *);
92 int			 pfclose(dev_t, int, int, struct proc *);
93 int			 pfioctl(dev_t, u_long, caddr_t, int, struct proc *);
94 int			 pf_begin_rules(u_int32_t *, const char *);
95 int			 pf_rollback_rules(u_int32_t, char *);
96 void			 pf_remove_queues(void);
97 int			 pf_commit_queues(void);
98 void			 pf_free_queues(struct pf_queuehead *);
99 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
100 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
101 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
102 int			 pf_commit_rules(u_int32_t, char *);
103 int			 pf_addr_setup(struct pf_ruleset *,
104 			    struct pf_addr_wrap *, sa_family_t);
105 int			 pf_kif_setup(char *, struct pfi_kif **);
106 void			 pf_addr_copyout(struct pf_addr_wrap *);
107 void			 pf_trans_set_commit(void);
108 void			 pf_pool_copyin(struct pf_pool *, struct pf_pool *);
109 int			 pf_rule_copyin(struct pf_rule *, struct pf_rule *,
110 			    struct pf_ruleset *);
111 u_int16_t		 pf_qname2qid(char *, int);
112 void			 pf_qid2qname(u_int16_t, char *);
113 void			 pf_qid_unref(u_int16_t);
114 
115 struct pf_rule		 pf_default_rule, pf_default_rule_new;
116 
117 struct {
118 	char		statusif[IFNAMSIZ];
119 	u_int32_t	debug;
120 	u_int32_t	hostid;
121 	u_int32_t	reass;
122 	u_int32_t	mask;
123 } pf_trans_set;
124 
125 #define	PF_TSET_STATUSIF	0x01
126 #define	PF_TSET_DEBUG		0x02
127 #define	PF_TSET_HOSTID		0x04
128 #define	PF_TSET_REASS		0x08
129 
130 #define	TAGID_MAX	 50000
131 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
132 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
133 
134 #ifdef WITH_PF_LOCK
135 /*
136  * pf_lock protects consistency of PF data structures, which don't have
137  * their dedicated lock yet. The pf_lock currently protects:
138  *	- rules,
139  *	- radix tables,
140  *	- source nodes
141  * All callers must grab pf_lock exclusively.
142  *
143  * pf_state_lock protects consistency of state table. Packets, which do state
144  * look up grab the lock as readers. If packet must create state, then it must
145  * grab the lock as writer. Whenever packet creates state it grabs pf_lock
146  * first then it locks pf_state_lock as the writer.
147  */
148 struct rwlock		 pf_lock = RWLOCK_INITIALIZER("pf_lock");
149 struct rwlock		 pf_state_lock = RWLOCK_INITIALIZER("pf_state_lock");
150 #endif /* WITH_PF_LOCK */
151 
152 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
153 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
154 #endif
155 u_int16_t		 tagname2tag(struct pf_tags *, char *, int);
156 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
157 void			 tag_unref(struct pf_tags *, u_int16_t);
158 int			 pf_rtlabel_add(struct pf_addr_wrap *);
159 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
160 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
161 
162 
163 void
164 pfattach(int num)
165 {
166 	u_int32_t *timeout = pf_default_rule.timeout;
167 
168 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0,
169 	    IPL_SOFTNET, 0, "pfrule", NULL);
170 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0,
171 	    IPL_SOFTNET, 0, "pfsrctr", NULL);
172 	pool_init(&pf_sn_item_pl, sizeof(struct pf_sn_item), 0,
173 	    IPL_SOFTNET, 0, "pfsnitem", NULL);
174 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0,
175 	    IPL_SOFTNET, 0, "pfstate", NULL);
176 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0,
177 	    IPL_SOFTNET, 0, "pfstkey", NULL);
178 	pool_init(&pf_state_item_pl, sizeof(struct pf_state_item), 0,
179 	    IPL_SOFTNET, 0, "pfstitem", NULL);
180 	pool_init(&pf_rule_item_pl, sizeof(struct pf_rule_item), 0,
181 	    IPL_SOFTNET, 0, "pfruleitem", NULL);
182 	pool_init(&pf_queue_pl, sizeof(struct pf_queuespec), 0,
183 	    IPL_SOFTNET, 0, "pfqueue", NULL);
184 	pool_init(&pf_tag_pl, sizeof(struct pf_tagname), 0,
185 	    IPL_SOFTNET, 0, "pftag", NULL);
186 	pool_init(&pf_pktdelay_pl, sizeof(struct pf_pktdelay), 0,
187 	    IPL_SOFTNET, 0, "pfpktdelay", NULL);
188 
189 	hfsc_initialize();
190 	pfr_initialize();
191 	pfi_initialize();
192 	pf_osfp_initialize();
193 	pf_syncookies_init();
194 
195 	pool_sethardlimit(pf_pool_limits[PF_LIMIT_STATES].pp,
196 	    pf_pool_limits[PF_LIMIT_STATES].limit, NULL, 0);
197 
198 	if (physmem <= atop(100*1024*1024))
199 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
200 		    PFR_KENTRY_HIWAT_SMALL;
201 
202 	RB_INIT(&tree_src_tracking);
203 	RB_INIT(&pf_anchors);
204 	pf_init_ruleset(&pf_main_ruleset);
205 	TAILQ_INIT(&pf_queues[0]);
206 	TAILQ_INIT(&pf_queues[1]);
207 	pf_queues_active = &pf_queues[0];
208 	pf_queues_inactive = &pf_queues[1];
209 	TAILQ_INIT(&state_list);
210 
211 	/* default rule should never be garbage collected */
212 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
213 	pf_default_rule.action = PF_PASS;
214 	pf_default_rule.nr = (u_int32_t)-1;
215 	pf_default_rule.rtableid = -1;
216 
217 	/* initialize default timeouts */
218 	timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
219 	timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
220 	timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
221 	timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
222 	timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
223 	timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
224 	timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
225 	timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
226 	timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
227 	timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
228 	timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
229 	timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
230 	timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
231 	timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
232 	timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
233 	timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
234 	timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
235 	timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
236 	timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
237 	timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
238 
239 	pf_default_rule.src.addr.type =  PF_ADDR_ADDRMASK;
240 	pf_default_rule.dst.addr.type =  PF_ADDR_ADDRMASK;
241 	pf_default_rule.rdr.addr.type =  PF_ADDR_NONE;
242 	pf_default_rule.nat.addr.type =  PF_ADDR_NONE;
243 	pf_default_rule.route.addr.type =  PF_ADDR_NONE;
244 
245 	pf_normalize_init();
246 	memset(&pf_status, 0, sizeof(pf_status));
247 	pf_status.debug = LOG_ERR;
248 	pf_status.reass = PF_REASS_ENABLED;
249 
250 	/* XXX do our best to avoid a conflict */
251 	pf_status.hostid = arc4random();
252 }
253 
254 int
255 pfopen(dev_t dev, int flags, int fmt, struct proc *p)
256 {
257 	if (minor(dev) >= 1)
258 		return (ENXIO);
259 	return (0);
260 }
261 
262 int
263 pfclose(dev_t dev, int flags, int fmt, struct proc *p)
264 {
265 	if (minor(dev) >= 1)
266 		return (ENXIO);
267 	return (0);
268 }
269 
270 void
271 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
272 {
273 	if (rulequeue != NULL) {
274 		if (rule->states_cur == 0 && rule->src_nodes == 0) {
275 			/*
276 			 * XXX - we need to remove the table *before* detaching
277 			 * the rule to make sure the table code does not delete
278 			 * the anchor under our feet.
279 			 */
280 			pf_tbladdr_remove(&rule->src.addr);
281 			pf_tbladdr_remove(&rule->dst.addr);
282 			pf_tbladdr_remove(&rule->rdr.addr);
283 			pf_tbladdr_remove(&rule->nat.addr);
284 			pf_tbladdr_remove(&rule->route.addr);
285 			if (rule->overload_tbl)
286 				pfr_detach_table(rule->overload_tbl);
287 		}
288 		TAILQ_REMOVE(rulequeue, rule, entries);
289 		rule->entries.tqe_prev = NULL;
290 		rule->nr = (u_int32_t)-1;
291 	}
292 
293 	if (rule->states_cur > 0 || rule->src_nodes > 0 ||
294 	    rule->entries.tqe_prev != NULL)
295 		return;
296 	pf_tag_unref(rule->tag);
297 	pf_tag_unref(rule->match_tag);
298 	pf_rtlabel_remove(&rule->src.addr);
299 	pf_rtlabel_remove(&rule->dst.addr);
300 	pfi_dynaddr_remove(&rule->src.addr);
301 	pfi_dynaddr_remove(&rule->dst.addr);
302 	pfi_dynaddr_remove(&rule->rdr.addr);
303 	pfi_dynaddr_remove(&rule->nat.addr);
304 	pfi_dynaddr_remove(&rule->route.addr);
305 	if (rulequeue == NULL) {
306 		pf_tbladdr_remove(&rule->src.addr);
307 		pf_tbladdr_remove(&rule->dst.addr);
308 		pf_tbladdr_remove(&rule->rdr.addr);
309 		pf_tbladdr_remove(&rule->nat.addr);
310 		pf_tbladdr_remove(&rule->route.addr);
311 		if (rule->overload_tbl)
312 			pfr_detach_table(rule->overload_tbl);
313 	}
314 	pfi_kif_unref(rule->rcv_kif, PFI_KIF_REF_RULE);
315 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
316 	pfi_kif_unref(rule->rdr.kif, PFI_KIF_REF_RULE);
317 	pfi_kif_unref(rule->nat.kif, PFI_KIF_REF_RULE);
318 	pfi_kif_unref(rule->route.kif, PFI_KIF_REF_RULE);
319 	pf_remove_anchor(rule);
320 	pool_put(&pf_rule_pl, rule);
321 }
322 
323 void
324 pf_purge_rule(struct pf_rule *rule)
325 {
326 	u_int32_t		 nr = 0;
327 	struct pf_ruleset	*ruleset;
328 
329 	KASSERT((rule != NULL) && (rule->ruleset != NULL));
330 	ruleset = rule->ruleset;
331 
332 	pf_rm_rule(ruleset->rules.active.ptr, rule);
333 	ruleset->rules.active.rcount--;
334 	TAILQ_FOREACH(rule, ruleset->rules.active.ptr, entries)
335 		rule->nr = nr++;
336 	ruleset->rules.active.ticket++;
337 	pf_calc_skip_steps(ruleset->rules.active.ptr);
338 	pf_remove_if_empty_ruleset(ruleset);
339 }
340 
341 u_int16_t
342 tagname2tag(struct pf_tags *head, char *tagname, int create)
343 {
344 	struct pf_tagname	*tag, *p = NULL;
345 	u_int16_t		 new_tagid = 1;
346 
347 	TAILQ_FOREACH(tag, head, entries)
348 		if (strcmp(tagname, tag->name) == 0) {
349 			tag->ref++;
350 			return (tag->tag);
351 		}
352 
353 	if (!create)
354 		return (0);
355 
356 	/*
357 	 * to avoid fragmentation, we do a linear search from the beginning
358 	 * and take the first free slot we find. if there is none or the list
359 	 * is empty, append a new entry at the end.
360 	 */
361 
362 	/* new entry */
363 	TAILQ_FOREACH(p, head, entries) {
364 		if (p->tag != new_tagid)
365 			break;
366 		new_tagid = p->tag + 1;
367 	}
368 
369 	if (new_tagid > TAGID_MAX)
370 		return (0);
371 
372 	/* allocate and fill new struct pf_tagname */
373 	tag = pool_get(&pf_tag_pl, PR_NOWAIT | PR_ZERO);
374 	if (tag == NULL)
375 		return (0);
376 	strlcpy(tag->name, tagname, sizeof(tag->name));
377 	tag->tag = new_tagid;
378 	tag->ref++;
379 
380 	if (p != NULL)	/* insert new entry before p */
381 		TAILQ_INSERT_BEFORE(p, tag, entries);
382 	else	/* either list empty or no free slot in between */
383 		TAILQ_INSERT_TAIL(head, tag, entries);
384 
385 	return (tag->tag);
386 }
387 
388 void
389 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
390 {
391 	struct pf_tagname	*tag;
392 
393 	TAILQ_FOREACH(tag, head, entries)
394 		if (tag->tag == tagid) {
395 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
396 			return;
397 		}
398 }
399 
400 void
401 tag_unref(struct pf_tags *head, u_int16_t tag)
402 {
403 	struct pf_tagname	*p, *next;
404 
405 	if (tag == 0)
406 		return;
407 
408 	TAILQ_FOREACH_SAFE(p, head, entries, next) {
409 		if (tag == p->tag) {
410 			if (--p->ref == 0) {
411 				TAILQ_REMOVE(head, p, entries);
412 				pool_put(&pf_tag_pl, p);
413 			}
414 			break;
415 		}
416 	}
417 }
418 
419 u_int16_t
420 pf_tagname2tag(char *tagname, int create)
421 {
422 	return (tagname2tag(&pf_tags, tagname, create));
423 }
424 
425 void
426 pf_tag2tagname(u_int16_t tagid, char *p)
427 {
428 	tag2tagname(&pf_tags, tagid, p);
429 }
430 
431 void
432 pf_tag_ref(u_int16_t tag)
433 {
434 	struct pf_tagname *t;
435 
436 	TAILQ_FOREACH(t, &pf_tags, entries)
437 		if (t->tag == tag)
438 			break;
439 	if (t != NULL)
440 		t->ref++;
441 }
442 
443 void
444 pf_tag_unref(u_int16_t tag)
445 {
446 	tag_unref(&pf_tags, tag);
447 }
448 
449 int
450 pf_rtlabel_add(struct pf_addr_wrap *a)
451 {
452 	if (a->type == PF_ADDR_RTLABEL &&
453 	    (a->v.rtlabel = rtlabel_name2id(a->v.rtlabelname)) == 0)
454 		return (-1);
455 	return (0);
456 }
457 
458 void
459 pf_rtlabel_remove(struct pf_addr_wrap *a)
460 {
461 	if (a->type == PF_ADDR_RTLABEL)
462 		rtlabel_unref(a->v.rtlabel);
463 }
464 
465 void
466 pf_rtlabel_copyout(struct pf_addr_wrap *a)
467 {
468 	const char	*name;
469 
470 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel) {
471 		if ((name = rtlabel_id2name(a->v.rtlabel)) == NULL)
472 			strlcpy(a->v.rtlabelname, "?",
473 			    sizeof(a->v.rtlabelname));
474 		else
475 			strlcpy(a->v.rtlabelname, name,
476 			    sizeof(a->v.rtlabelname));
477 	}
478 }
479 
480 u_int16_t
481 pf_qname2qid(char *qname, int create)
482 {
483 	return (tagname2tag(&pf_qids, qname, create));
484 }
485 
486 void
487 pf_qid2qname(u_int16_t qid, char *p)
488 {
489 	tag2tagname(&pf_qids, qid, p);
490 }
491 
492 void
493 pf_qid_unref(u_int16_t qid)
494 {
495 	tag_unref(&pf_qids, (u_int16_t)qid);
496 }
497 
498 int
499 pf_begin_rules(u_int32_t *ticket, const char *anchor)
500 {
501 	struct pf_ruleset	*rs;
502 	struct pf_rule		*rule;
503 
504 	if ((rs = pf_find_or_create_ruleset(anchor)) == NULL)
505 		return (EINVAL);
506 	while ((rule = TAILQ_FIRST(rs->rules.inactive.ptr)) != NULL) {
507 		pf_rm_rule(rs->rules.inactive.ptr, rule);
508 		rs->rules.inactive.rcount--;
509 	}
510 	*ticket = ++rs->rules.inactive.ticket;
511 	rs->rules.inactive.open = 1;
512 	return (0);
513 }
514 
515 int
516 pf_rollback_rules(u_int32_t ticket, char *anchor)
517 {
518 	struct pf_ruleset	*rs;
519 	struct pf_rule		*rule;
520 
521 	rs = pf_find_ruleset(anchor);
522 	if (rs == NULL || !rs->rules.inactive.open ||
523 	    rs->rules.inactive.ticket != ticket)
524 		return (0);
525 	while ((rule = TAILQ_FIRST(rs->rules.inactive.ptr)) != NULL) {
526 		pf_rm_rule(rs->rules.inactive.ptr, rule);
527 		rs->rules.inactive.rcount--;
528 	}
529 	rs->rules.inactive.open = 0;
530 
531 	/* queue defs only in the main ruleset */
532 	if (anchor[0])
533 		return (0);
534 
535 	pf_free_queues(pf_queues_inactive);
536 
537 	return (0);
538 }
539 
540 void
541 pf_free_queues(struct pf_queuehead *where)
542 {
543 	struct pf_queuespec	*q, *qtmp;
544 
545 	TAILQ_FOREACH_SAFE(q, where, entries, qtmp) {
546 		TAILQ_REMOVE(where, q, entries);
547 		pfi_kif_unref(q->kif, PFI_KIF_REF_RULE);
548 		pool_put(&pf_queue_pl, q);
549 	}
550 }
551 
552 void
553 pf_remove_queues(void)
554 {
555 	struct pf_queuespec	*q;
556 	struct ifnet		*ifp;
557 
558 	/* put back interfaces in normal queueing mode */
559 	TAILQ_FOREACH(q, pf_queues_active, entries) {
560 		if (q->parent_qid != 0)
561 			continue;
562 
563 		ifp = q->kif->pfik_ifp;
564 		if (ifp == NULL)
565 			continue;
566 
567 		ifq_attach(&ifp->if_snd, ifq_priq_ops, NULL);
568 	}
569 }
570 
571 struct pf_queue_if {
572 	struct ifnet		*ifp;
573 	const struct ifq_ops	*ifqops;
574 	const struct pfq_ops	*pfqops;
575 	void			*disc;
576 	struct pf_queue_if	*next;
577 };
578 
579 static inline struct pf_queue_if *
580 pf_ifp2q(struct pf_queue_if *list, struct ifnet *ifp)
581 {
582 	struct pf_queue_if *qif = list;
583 
584 	while (qif != NULL) {
585 		if (qif->ifp == ifp)
586 			return (qif);
587 
588 		qif = qif->next;
589 	}
590 
591 	return (qif);
592 }
593 
594 int
595 pf_create_queues(void)
596 {
597 	struct pf_queuespec	*q;
598 	struct ifnet		*ifp;
599 	struct pf_queue_if		*list = NULL, *qif;
600 	int			 error;
601 
602 	/*
603 	 * Find root queues and allocate traffic conditioner
604 	 * private data for these interfaces
605 	 */
606 	TAILQ_FOREACH(q, pf_queues_active, entries) {
607 		if (q->parent_qid != 0)
608 			continue;
609 
610 		ifp = q->kif->pfik_ifp;
611 		if (ifp == NULL)
612 			continue;
613 
614 		qif = malloc(sizeof(*qif), M_TEMP, M_WAITOK);
615 		qif->ifp = ifp;
616 
617 		if (q->flags & PFQS_ROOTCLASS) {
618 			qif->ifqops = ifq_hfsc_ops;
619 			qif->pfqops = pfq_hfsc_ops;
620 		} else {
621 			qif->ifqops = ifq_fqcodel_ops;
622 			qif->pfqops = pfq_fqcodel_ops;
623 		}
624 
625 		qif->disc = qif->pfqops->pfq_alloc(ifp);
626 
627 		qif->next = list;
628 		list = qif;
629 	}
630 
631 	/* and now everything */
632 	TAILQ_FOREACH(q, pf_queues_active, entries) {
633 		ifp = q->kif->pfik_ifp;
634 		if (ifp == NULL)
635 			continue;
636 
637 		qif = pf_ifp2q(list, ifp);
638 		KASSERT(qif != NULL);
639 
640 		error = qif->pfqops->pfq_addqueue(qif->disc, q);
641 		if (error != 0)
642 			goto error;
643 	}
644 
645 	/* find root queues in old list to disable them if necessary */
646 	TAILQ_FOREACH(q, pf_queues_inactive, entries) {
647 		if (q->parent_qid != 0)
648 			continue;
649 
650 		ifp = q->kif->pfik_ifp;
651 		if (ifp == NULL)
652 			continue;
653 
654 		qif = pf_ifp2q(list, ifp);
655 		if (qif != NULL)
656 			continue;
657 
658 		ifq_attach(&ifp->if_snd, ifq_priq_ops, NULL);
659 	}
660 
661 	/* commit the new queues */
662 	while (list != NULL) {
663 		qif = list;
664 		list = qif->next;
665 
666 		ifp = qif->ifp;
667 
668 		ifq_attach(&ifp->if_snd, qif->ifqops, qif->disc);
669 		free(qif, M_TEMP, sizeof(*qif));
670 	}
671 
672 	return (0);
673 
674 error:
675 	while (list != NULL) {
676 		qif = list;
677 		list = qif->next;
678 
679 		qif->pfqops->pfq_free(qif->disc);
680 		free(qif, M_TEMP, sizeof(*qif));
681 	}
682 
683 	return (error);
684 }
685 
686 int
687 pf_commit_queues(void)
688 {
689 	struct pf_queuehead	*qswap;
690 	int error;
691 
692         /* swap */
693         qswap = pf_queues_active;
694         pf_queues_active = pf_queues_inactive;
695         pf_queues_inactive = qswap;
696 
697 	error = pf_create_queues();
698 	if (error != 0) {
699 		pf_queues_inactive = pf_queues_active;
700 		pf_queues_active = qswap;
701 		return (error);
702 	}
703 
704         pf_free_queues(pf_queues_inactive);
705 
706 	return (0);
707 }
708 
709 const struct pfq_ops *
710 pf_queue_manager(struct pf_queuespec *q)
711 {
712 	if (q->flags & PFQS_FLOWQUEUE)
713 		return pfq_fqcodel_ops;
714 	return (/* pfq_default_ops */ NULL);
715 }
716 
717 #define PF_MD5_UPD(st, elm)						\
718 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
719 
720 #define PF_MD5_UPD_STR(st, elm)						\
721 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
722 
723 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
724 		(stor) = htonl((st)->elm);				\
725 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
726 } while (0)
727 
728 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
729 		(stor) = htons((st)->elm);				\
730 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
731 } while (0)
732 
733 void
734 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
735 {
736 	PF_MD5_UPD(pfr, addr.type);
737 	switch (pfr->addr.type) {
738 		case PF_ADDR_DYNIFTL:
739 			PF_MD5_UPD(pfr, addr.v.ifname);
740 			PF_MD5_UPD(pfr, addr.iflags);
741 			break;
742 		case PF_ADDR_TABLE:
743 			PF_MD5_UPD(pfr, addr.v.tblname);
744 			break;
745 		case PF_ADDR_ADDRMASK:
746 			/* XXX ignore af? */
747 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
748 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
749 			break;
750 		case PF_ADDR_RTLABEL:
751 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
752 			break;
753 	}
754 
755 	PF_MD5_UPD(pfr, port[0]);
756 	PF_MD5_UPD(pfr, port[1]);
757 	PF_MD5_UPD(pfr, neg);
758 	PF_MD5_UPD(pfr, port_op);
759 }
760 
761 void
762 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
763 {
764 	u_int16_t x;
765 	u_int32_t y;
766 
767 	pf_hash_rule_addr(ctx, &rule->src);
768 	pf_hash_rule_addr(ctx, &rule->dst);
769 	PF_MD5_UPD_STR(rule, label);
770 	PF_MD5_UPD_STR(rule, ifname);
771 	PF_MD5_UPD_STR(rule, rcv_ifname);
772 	PF_MD5_UPD_STR(rule, match_tagname);
773 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
774 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
775 	PF_MD5_UPD_HTONL(rule, prob, y);
776 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
777 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
778 	PF_MD5_UPD(rule, uid.op);
779 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
780 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
781 	PF_MD5_UPD(rule, gid.op);
782 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
783 	PF_MD5_UPD(rule, action);
784 	PF_MD5_UPD(rule, direction);
785 	PF_MD5_UPD(rule, af);
786 	PF_MD5_UPD(rule, quick);
787 	PF_MD5_UPD(rule, ifnot);
788 	PF_MD5_UPD(rule, rcvifnot);
789 	PF_MD5_UPD(rule, match_tag_not);
790 	PF_MD5_UPD(rule, keep_state);
791 	PF_MD5_UPD(rule, proto);
792 	PF_MD5_UPD(rule, type);
793 	PF_MD5_UPD(rule, code);
794 	PF_MD5_UPD(rule, flags);
795 	PF_MD5_UPD(rule, flagset);
796 	PF_MD5_UPD(rule, allow_opts);
797 	PF_MD5_UPD(rule, rt);
798 	PF_MD5_UPD(rule, tos);
799 }
800 
801 int
802 pf_commit_rules(u_int32_t ticket, char *anchor)
803 {
804 	struct pf_ruleset	*rs;
805 	struct pf_rule		*rule, **old_array;
806 	struct pf_rulequeue	*old_rules;
807 	int			 error;
808 	u_int32_t		 old_rcount;
809 
810 	/* Make sure any expired rules get removed from active rules first. */
811 	pf_purge_expired_rules();
812 
813 	rs = pf_find_ruleset(anchor);
814 	if (rs == NULL || !rs->rules.inactive.open ||
815 	    ticket != rs->rules.inactive.ticket)
816 		return (EBUSY);
817 
818 	/* Calculate checksum for the main ruleset */
819 	if (rs == &pf_main_ruleset) {
820 		error = pf_setup_pfsync_matching(rs);
821 		if (error != 0)
822 			return (error);
823 	}
824 
825 	/* Swap rules, keep the old. */
826 	old_rules = rs->rules.active.ptr;
827 	old_rcount = rs->rules.active.rcount;
828 	old_array = rs->rules.active.ptr_array;
829 
830 	rs->rules.active.ptr = rs->rules.inactive.ptr;
831 	rs->rules.active.ptr_array = rs->rules.inactive.ptr_array;
832 	rs->rules.active.rcount = rs->rules.inactive.rcount;
833 	rs->rules.inactive.ptr = old_rules;
834 	rs->rules.inactive.ptr_array = old_array;
835 	rs->rules.inactive.rcount = old_rcount;
836 
837 	rs->rules.active.ticket = rs->rules.inactive.ticket;
838 	pf_calc_skip_steps(rs->rules.active.ptr);
839 
840 
841 	/* Purge the old rule list. */
842 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
843 		pf_rm_rule(old_rules, rule);
844 	if (rs->rules.inactive.ptr_array)
845 		free(rs->rules.inactive.ptr_array, M_TEMP, 0);
846 	rs->rules.inactive.ptr_array = NULL;
847 	rs->rules.inactive.rcount = 0;
848 	rs->rules.inactive.open = 0;
849 	pf_remove_if_empty_ruleset(rs);
850 
851 	/* queue defs only in the main ruleset */
852 	if (anchor[0])
853 		return (0);
854 	return (pf_commit_queues());
855 }
856 
857 int
858 pf_setup_pfsync_matching(struct pf_ruleset *rs)
859 {
860 	MD5_CTX			 ctx;
861 	struct pf_rule		*rule;
862 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
863 
864 	MD5Init(&ctx);
865 	if (rs->rules.inactive.ptr_array)
866 		free(rs->rules.inactive.ptr_array, M_TEMP, 0);
867 	rs->rules.inactive.ptr_array = NULL;
868 
869 	if (rs->rules.inactive.rcount) {
870 		rs->rules.inactive.ptr_array =
871 		    mallocarray(rs->rules.inactive.rcount, sizeof(caddr_t),
872 		    M_TEMP, M_NOWAIT);
873 
874 		if (!rs->rules.inactive.ptr_array)
875 			return (ENOMEM);
876 
877 		TAILQ_FOREACH(rule, rs->rules.inactive.ptr, entries) {
878 			pf_hash_rule(&ctx, rule);
879 			(rs->rules.inactive.ptr_array)[rule->nr] = rule;
880 		}
881 	}
882 
883 	MD5Final(digest, &ctx);
884 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
885 	return (0);
886 }
887 
888 int
889 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
890     sa_family_t af)
891 {
892 	if (pfi_dynaddr_setup(addr, af) ||
893 	    pf_tbladdr_setup(ruleset, addr) ||
894 	    pf_rtlabel_add(addr))
895 		return (EINVAL);
896 
897 	return (0);
898 }
899 
900 int
901 pf_kif_setup(char *ifname, struct pfi_kif **kif)
902 {
903 	if (ifname[0]) {
904 		*kif = pfi_kif_get(ifname);
905 		if (*kif == NULL)
906 			return (EINVAL);
907 
908 		pfi_kif_ref(*kif, PFI_KIF_REF_RULE);
909 	} else
910 		*kif = NULL;
911 
912 	return (0);
913 }
914 
915 void
916 pf_addr_copyout(struct pf_addr_wrap *addr)
917 {
918 	pfi_dynaddr_copyout(addr);
919 	pf_tbladdr_copyout(addr);
920 	pf_rtlabel_copyout(addr);
921 }
922 
923 int
924 pfioctl(dev_t dev, u_long cmd, caddr_t addr, int flags, struct proc *p)
925 {
926 	int			 error = 0;
927 
928 	/* XXX keep in sync with switch() below */
929 	if (securelevel > 1)
930 		switch (cmd) {
931 		case DIOCGETRULES:
932 		case DIOCGETRULE:
933 		case DIOCGETSTATE:
934 		case DIOCSETSTATUSIF:
935 		case DIOCGETSTATUS:
936 		case DIOCCLRSTATUS:
937 		case DIOCNATLOOK:
938 		case DIOCSETDEBUG:
939 		case DIOCGETSTATES:
940 		case DIOCGETTIMEOUT:
941 		case DIOCGETLIMIT:
942 		case DIOCGETRULESETS:
943 		case DIOCGETRULESET:
944 		case DIOCGETQUEUES:
945 		case DIOCGETQUEUE:
946 		case DIOCGETQSTATS:
947 		case DIOCRGETTABLES:
948 		case DIOCRGETTSTATS:
949 		case DIOCRCLRTSTATS:
950 		case DIOCRCLRADDRS:
951 		case DIOCRADDADDRS:
952 		case DIOCRDELADDRS:
953 		case DIOCRSETADDRS:
954 		case DIOCRGETADDRS:
955 		case DIOCRGETASTATS:
956 		case DIOCRCLRASTATS:
957 		case DIOCRTSTADDRS:
958 		case DIOCOSFPGET:
959 		case DIOCGETSRCNODES:
960 		case DIOCCLRSRCNODES:
961 		case DIOCIGETIFACES:
962 		case DIOCSETIFFLAG:
963 		case DIOCCLRIFFLAG:
964 		case DIOCGETSYNFLWATS:
965 			break;
966 		case DIOCRCLRTABLES:
967 		case DIOCRADDTABLES:
968 		case DIOCRDELTABLES:
969 		case DIOCRSETTFLAGS:
970 			if (((struct pfioc_table *)addr)->pfrio_flags &
971 			    PFR_FLAG_DUMMY)
972 				break; /* dummy operation ok */
973 			return (EPERM);
974 		default:
975 			return (EPERM);
976 		}
977 
978 	if (!(flags & FWRITE))
979 		switch (cmd) {
980 		case DIOCGETRULES:
981 		case DIOCGETSTATE:
982 		case DIOCGETSTATUS:
983 		case DIOCGETSTATES:
984 		case DIOCGETTIMEOUT:
985 		case DIOCGETLIMIT:
986 		case DIOCGETRULESETS:
987 		case DIOCGETRULESET:
988 		case DIOCGETQUEUES:
989 		case DIOCGETQUEUE:
990 		case DIOCGETQSTATS:
991 		case DIOCNATLOOK:
992 		case DIOCRGETTABLES:
993 		case DIOCRGETTSTATS:
994 		case DIOCRGETADDRS:
995 		case DIOCRGETASTATS:
996 		case DIOCRTSTADDRS:
997 		case DIOCOSFPGET:
998 		case DIOCGETSRCNODES:
999 		case DIOCIGETIFACES:
1000 		case DIOCGETSYNFLWATS:
1001 			break;
1002 		case DIOCRCLRTABLES:
1003 		case DIOCRADDTABLES:
1004 		case DIOCRDELTABLES:
1005 		case DIOCRCLRTSTATS:
1006 		case DIOCRCLRADDRS:
1007 		case DIOCRADDADDRS:
1008 		case DIOCRDELADDRS:
1009 		case DIOCRSETADDRS:
1010 		case DIOCRSETTFLAGS:
1011 			if (((struct pfioc_table *)addr)->pfrio_flags &
1012 			    PFR_FLAG_DUMMY) {
1013 				flags |= FWRITE; /* need write lock for dummy */
1014 				break; /* dummy operation ok */
1015 			}
1016 			return (EACCES);
1017 		case DIOCGETRULE:
1018 			if (((struct pfioc_rule *)addr)->action ==
1019 			    PF_GET_CLR_CNTR)
1020 				return (EACCES);
1021 			break;
1022 		default:
1023 			return (EACCES);
1024 		}
1025 
1026 	NET_LOCK();
1027 	switch (cmd) {
1028 
1029 	case DIOCSTART:
1030 		PF_LOCK();
1031 		if (pf_status.running)
1032 			error = EEXIST;
1033 		else {
1034 			pf_status.running = 1;
1035 			pf_status.since = time_uptime;
1036 			if (pf_status.stateid == 0) {
1037 				pf_status.stateid = time_second;
1038 				pf_status.stateid = pf_status.stateid << 32;
1039 			}
1040 			timeout_add_sec(&pf_purge_to, 1);
1041 			pf_create_queues();
1042 			DPFPRINTF(LOG_NOTICE, "pf: started");
1043 		}
1044 		PF_UNLOCK();
1045 		break;
1046 
1047 	case DIOCSTOP:
1048 		PF_LOCK();
1049 		if (!pf_status.running)
1050 			error = ENOENT;
1051 		else {
1052 			pf_status.running = 0;
1053 			pf_status.since = time_uptime;
1054 			pf_remove_queues();
1055 			DPFPRINTF(LOG_NOTICE, "pf: stopped");
1056 		}
1057 		PF_UNLOCK();
1058 		break;
1059 
1060 	case DIOCGETQUEUES: {
1061 		struct pfioc_queue	*pq = (struct pfioc_queue *)addr;
1062 		struct pf_queuespec	*qs;
1063 		u_int32_t		 nr = 0;
1064 
1065 		PF_LOCK();
1066 		pq->ticket = pf_main_ruleset.rules.active.ticket;
1067 
1068 		/* save state to not run over them all each time? */
1069 		qs = TAILQ_FIRST(pf_queues_active);
1070 		while (qs != NULL) {
1071 			qs = TAILQ_NEXT(qs, entries);
1072 			nr++;
1073 		}
1074 		pq->nr = nr;
1075 		PF_UNLOCK();
1076 		break;
1077 	}
1078 
1079 	case DIOCGETQUEUE: {
1080 		struct pfioc_queue	*pq = (struct pfioc_queue *)addr;
1081 		struct pf_queuespec	*qs;
1082 		u_int32_t		 nr = 0;
1083 
1084 		PF_LOCK();
1085 		if (pq->ticket != pf_main_ruleset.rules.active.ticket) {
1086 			error = EBUSY;
1087 			PF_UNLOCK();
1088 			break;
1089 		}
1090 
1091 		/* save state to not run over them all each time? */
1092 		qs = TAILQ_FIRST(pf_queues_active);
1093 		while ((qs != NULL) && (nr++ < pq->nr))
1094 			qs = TAILQ_NEXT(qs, entries);
1095 		if (qs == NULL) {
1096 			error = EBUSY;
1097 			PF_UNLOCK();
1098 			break;
1099 		}
1100 		memcpy(&pq->queue, qs, sizeof(pq->queue));
1101 		PF_UNLOCK();
1102 		break;
1103 	}
1104 
1105 	case DIOCGETQSTATS: {
1106 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
1107 		struct pf_queuespec	*qs;
1108 		u_int32_t		 nr;
1109 		int			 nbytes;
1110 
1111 		PF_LOCK();
1112 		if (pq->ticket != pf_main_ruleset.rules.active.ticket) {
1113 			error = EBUSY;
1114 			PF_UNLOCK();
1115 			break;
1116 		}
1117 		nbytes = pq->nbytes;
1118 		nr = 0;
1119 
1120 		/* save state to not run over them all each time? */
1121 		qs = TAILQ_FIRST(pf_queues_active);
1122 		while ((qs != NULL) && (nr++ < pq->nr))
1123 			qs = TAILQ_NEXT(qs, entries);
1124 		if (qs == NULL) {
1125 			error = EBUSY;
1126 			PF_UNLOCK();
1127 			break;
1128 		}
1129 		memcpy(&pq->queue, qs, sizeof(pq->queue));
1130 		/* It's a root flow queue but is not an HFSC root class */
1131 		if ((qs->flags & PFQS_FLOWQUEUE) && qs->parent_qid == 0 &&
1132 		    !(qs->flags & PFQS_ROOTCLASS))
1133 			error = pfq_fqcodel_ops->pfq_qstats(qs, pq->buf,
1134 			    &nbytes);
1135 		else
1136 			error = pfq_hfsc_ops->pfq_qstats(qs, pq->buf,
1137 			    &nbytes);
1138 		if (error == 0)
1139 			pq->nbytes = nbytes;
1140 		PF_UNLOCK();
1141 		break;
1142 	}
1143 
1144 	case DIOCADDQUEUE: {
1145 		struct pfioc_queue	*q = (struct pfioc_queue *)addr;
1146 		struct pf_queuespec	*qs;
1147 
1148 		PF_LOCK();
1149 		if (q->ticket != pf_main_ruleset.rules.inactive.ticket) {
1150 			error = EBUSY;
1151 			PF_UNLOCK();
1152 			break;
1153 		}
1154 		qs = pool_get(&pf_queue_pl, PR_WAITOK|PR_LIMITFAIL|PR_ZERO);
1155 		if (qs == NULL) {
1156 			error = ENOMEM;
1157 			PF_UNLOCK();
1158 			break;
1159 		}
1160 		memcpy(qs, &q->queue, sizeof(*qs));
1161 		qs->qid = pf_qname2qid(qs->qname, 1);
1162 		if (qs->qid == 0) {
1163 			pool_put(&pf_queue_pl, qs);
1164 			error = EBUSY;
1165 			PF_UNLOCK();
1166 			break;
1167 		}
1168 		if (qs->parent[0] && (qs->parent_qid =
1169 		    pf_qname2qid(qs->parent, 0)) == 0) {
1170 			pool_put(&pf_queue_pl, qs);
1171 			error = ESRCH;
1172 			PF_UNLOCK();
1173 			break;
1174 		}
1175 		qs->kif = pfi_kif_get(qs->ifname);
1176 		if (qs->kif == NULL) {
1177 			pool_put(&pf_queue_pl, qs);
1178 			error = ESRCH;
1179 			PF_UNLOCK();
1180 			break;
1181 		}
1182 		/* XXX resolve bw percentage specs */
1183 		pfi_kif_ref(qs->kif, PFI_KIF_REF_RULE);
1184 
1185 		TAILQ_INSERT_TAIL(pf_queues_inactive, qs, entries);
1186 		PF_UNLOCK();
1187 
1188 		break;
1189 	}
1190 
1191 	case DIOCADDRULE: {
1192 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1193 		struct pf_ruleset	*ruleset;
1194 		struct pf_rule		*rule, *tail;
1195 
1196 		PF_LOCK();
1197 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1198 		ruleset = pf_find_ruleset(pr->anchor);
1199 		if (ruleset == NULL) {
1200 			error = EINVAL;
1201 			PF_UNLOCK();
1202 			break;
1203 		}
1204 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1205 			error = EINVAL;
1206 			PF_UNLOCK();
1207 			break;
1208 		}
1209 		if (pr->ticket != ruleset->rules.inactive.ticket) {
1210 			error = EBUSY;
1211 			PF_UNLOCK();
1212 			break;
1213 		}
1214 		rule = pool_get(&pf_rule_pl, PR_WAITOK|PR_LIMITFAIL|PR_ZERO);
1215 		if (rule == NULL) {
1216 			error = ENOMEM;
1217 			PF_UNLOCK();
1218 			break;
1219 		}
1220 		if ((error = pf_rule_copyin(&pr->rule, rule, ruleset))) {
1221 			pf_rm_rule(NULL, rule);
1222 			rule = NULL;
1223 			PF_UNLOCK();
1224 			break;
1225 		}
1226 		rule->cuid = p->p_ucred->cr_ruid;
1227 		rule->cpid = p->p_p->ps_pid;
1228 
1229 		switch (rule->af) {
1230 		case 0:
1231 			break;
1232 		case AF_INET:
1233 			break;
1234 #ifdef INET6
1235 		case AF_INET6:
1236 			break;
1237 #endif /* INET6 */
1238 		default:
1239 			pf_rm_rule(NULL, rule);
1240 			rule = NULL;
1241 			error = EAFNOSUPPORT;
1242 			PF_UNLOCK();
1243 			goto fail;
1244 		}
1245 		tail = TAILQ_LAST(ruleset->rules.inactive.ptr,
1246 		    pf_rulequeue);
1247 		if (tail)
1248 			rule->nr = tail->nr + 1;
1249 		else
1250 			rule->nr = 0;
1251 
1252 		if (rule->src.addr.type == PF_ADDR_NONE ||
1253 		    rule->dst.addr.type == PF_ADDR_NONE)
1254 			error = EINVAL;
1255 
1256 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1257 			error = EINVAL;
1258 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1259 			error = EINVAL;
1260 		if (pf_addr_setup(ruleset, &rule->rdr.addr, rule->af))
1261 			error = EINVAL;
1262 		if (pf_addr_setup(ruleset, &rule->nat.addr, rule->af))
1263 			error = EINVAL;
1264 		if (pf_addr_setup(ruleset, &rule->route.addr, rule->af))
1265 			error = EINVAL;
1266 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1267 			error = EINVAL;
1268 		if (rule->rt && !rule->direction)
1269 			error = EINVAL;
1270 		if (rule->scrub_flags & PFSTATE_SETPRIO &&
1271 		    (rule->set_prio[0] > IFQ_MAXPRIO ||
1272 		    rule->set_prio[1] > IFQ_MAXPRIO))
1273 			error = EINVAL;
1274 
1275 		if (error) {
1276 			pf_rm_rule(NULL, rule);
1277 			PF_UNLOCK();
1278 			break;
1279 		}
1280 		TAILQ_INSERT_TAIL(ruleset->rules.inactive.ptr,
1281 		    rule, entries);
1282 		rule->ruleset = ruleset;
1283 		ruleset->rules.inactive.rcount++;
1284 		PF_UNLOCK();
1285 		break;
1286 	}
1287 
1288 	case DIOCGETRULES: {
1289 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1290 		struct pf_ruleset	*ruleset;
1291 		struct pf_rule		*tail;
1292 
1293 		PF_LOCK();
1294 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1295 		ruleset = pf_find_ruleset(pr->anchor);
1296 		if (ruleset == NULL) {
1297 			error = EINVAL;
1298 			PF_UNLOCK();
1299 			break;
1300 		}
1301 		tail = TAILQ_LAST(ruleset->rules.active.ptr, pf_rulequeue);
1302 		if (tail)
1303 			pr->nr = tail->nr + 1;
1304 		else
1305 			pr->nr = 0;
1306 		pr->ticket = ruleset->rules.active.ticket;
1307 		PF_UNLOCK();
1308 		break;
1309 	}
1310 
1311 	case DIOCGETRULE: {
1312 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1313 		struct pf_ruleset	*ruleset;
1314 		struct pf_rule		*rule;
1315 		int			 i;
1316 
1317 		PF_LOCK();
1318 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1319 		ruleset = pf_find_ruleset(pr->anchor);
1320 		if (ruleset == NULL) {
1321 			error = EINVAL;
1322 			PF_UNLOCK();
1323 			break;
1324 		}
1325 		if (pr->ticket != ruleset->rules.active.ticket) {
1326 			error = EBUSY;
1327 			PF_UNLOCK();
1328 			break;
1329 		}
1330 		rule = TAILQ_FIRST(ruleset->rules.active.ptr);
1331 		while ((rule != NULL) && (rule->nr != pr->nr))
1332 			rule = TAILQ_NEXT(rule, entries);
1333 		if (rule == NULL) {
1334 			error = EBUSY;
1335 			PF_UNLOCK();
1336 			break;
1337 		}
1338 		memcpy(&pr->rule, rule, sizeof(struct pf_rule));
1339 		memset(&pr->rule.entries, 0, sizeof(pr->rule.entries));
1340 		pr->rule.kif = NULL;
1341 		pr->rule.nat.kif = NULL;
1342 		pr->rule.rdr.kif = NULL;
1343 		pr->rule.route.kif = NULL;
1344 		pr->rule.rcv_kif = NULL;
1345 		pr->rule.anchor = NULL;
1346 		pr->rule.overload_tbl = NULL;
1347 		pr->rule.pktrate.limit /= PF_THRESHOLD_MULT;
1348 		memset(&pr->rule.gcle, 0, sizeof(pr->rule.gcle));
1349 		pr->rule.ruleset = NULL;
1350 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1351 			error = EBUSY;
1352 			PF_UNLOCK();
1353 			break;
1354 		}
1355 		pf_addr_copyout(&pr->rule.src.addr);
1356 		pf_addr_copyout(&pr->rule.dst.addr);
1357 		pf_addr_copyout(&pr->rule.rdr.addr);
1358 		pf_addr_copyout(&pr->rule.nat.addr);
1359 		pf_addr_copyout(&pr->rule.route.addr);
1360 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1361 			if (rule->skip[i].ptr == NULL)
1362 				pr->rule.skip[i].nr = (u_int32_t)-1;
1363 			else
1364 				pr->rule.skip[i].nr =
1365 				    rule->skip[i].ptr->nr;
1366 
1367 		if (pr->action == PF_GET_CLR_CNTR) {
1368 			rule->evaluations = 0;
1369 			rule->packets[0] = rule->packets[1] = 0;
1370 			rule->bytes[0] = rule->bytes[1] = 0;
1371 			rule->states_tot = 0;
1372 		}
1373 		PF_UNLOCK();
1374 		break;
1375 	}
1376 
1377 	case DIOCCHANGERULE: {
1378 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1379 		struct pf_ruleset	*ruleset;
1380 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1381 		u_int32_t		 nr = 0;
1382 
1383 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1384 		    pcr->action > PF_CHANGE_GET_TICKET) {
1385 			error = EINVAL;
1386 			break;
1387 		}
1388 		PF_LOCK();
1389 		ruleset = pf_find_ruleset(pcr->anchor);
1390 		if (ruleset == NULL) {
1391 			error = EINVAL;
1392 			PF_UNLOCK();
1393 			break;
1394 		}
1395 
1396 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1397 			pcr->ticket = ++ruleset->rules.active.ticket;
1398 			PF_UNLOCK();
1399 			break;
1400 		} else {
1401 			if (pcr->ticket !=
1402 			    ruleset->rules.active.ticket) {
1403 				error = EINVAL;
1404 				PF_UNLOCK();
1405 				break;
1406 			}
1407 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1408 				error = EINVAL;
1409 				PF_UNLOCK();
1410 				break;
1411 			}
1412 		}
1413 
1414 		if (pcr->action != PF_CHANGE_REMOVE) {
1415 			newrule = pool_get(&pf_rule_pl,
1416 			    PR_WAITOK|PR_LIMITFAIL|PR_ZERO);
1417 			if (newrule == NULL) {
1418 				error = ENOMEM;
1419 				PF_UNLOCK();
1420 				break;
1421 			}
1422 			pf_rule_copyin(&pcr->rule, newrule, ruleset);
1423 			newrule->cuid = p->p_ucred->cr_ruid;
1424 			newrule->cpid = p->p_p->ps_pid;
1425 
1426 			switch (newrule->af) {
1427 			case 0:
1428 				break;
1429 			case AF_INET:
1430 				break;
1431 #ifdef INET6
1432 			case AF_INET6:
1433 				break;
1434 #endif /* INET6 */
1435 			default:
1436 				pf_rm_rule(NULL, newrule);
1437 				error = EAFNOSUPPORT;
1438 				PF_UNLOCK();
1439 				goto fail;
1440 			}
1441 
1442 			if (newrule->rt && !newrule->direction)
1443 				error = EINVAL;
1444 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1445 				error = EINVAL;
1446 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1447 				error = EINVAL;
1448 			if (pf_addr_setup(ruleset, &newrule->rdr.addr, newrule->af))
1449 				error = EINVAL;
1450 			if (pf_addr_setup(ruleset, &newrule->nat.addr, newrule->af))
1451 				error = EINVAL;
1452 			if (pf_addr_setup(ruleset, &newrule->route.addr, newrule->af))
1453 				error = EINVAL;
1454 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1455 				error = EINVAL;
1456 
1457 			if (error) {
1458 				pf_rm_rule(NULL, newrule);
1459 				PF_UNLOCK();
1460 				break;
1461 			}
1462 		}
1463 
1464 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1465 			oldrule = TAILQ_FIRST(ruleset->rules.active.ptr);
1466 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1467 			oldrule = TAILQ_LAST(ruleset->rules.active.ptr,
1468 			    pf_rulequeue);
1469 		else {
1470 			oldrule = TAILQ_FIRST(ruleset->rules.active.ptr);
1471 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1472 				oldrule = TAILQ_NEXT(oldrule, entries);
1473 			if (oldrule == NULL) {
1474 				if (newrule != NULL)
1475 					pf_rm_rule(NULL, newrule);
1476 				error = EINVAL;
1477 				PF_UNLOCK();
1478 				break;
1479 			}
1480 		}
1481 
1482 		if (pcr->action == PF_CHANGE_REMOVE) {
1483 			pf_rm_rule(ruleset->rules.active.ptr, oldrule);
1484 			ruleset->rules.active.rcount--;
1485 		} else {
1486 			if (oldrule == NULL)
1487 				TAILQ_INSERT_TAIL(
1488 				    ruleset->rules.active.ptr,
1489 				    newrule, entries);
1490 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1491 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1492 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1493 			else
1494 				TAILQ_INSERT_AFTER(
1495 				    ruleset->rules.active.ptr,
1496 				    oldrule, newrule, entries);
1497 			ruleset->rules.active.rcount++;
1498 		}
1499 
1500 		nr = 0;
1501 		TAILQ_FOREACH(oldrule, ruleset->rules.active.ptr, entries)
1502 			oldrule->nr = nr++;
1503 
1504 		ruleset->rules.active.ticket++;
1505 
1506 		pf_calc_skip_steps(ruleset->rules.active.ptr);
1507 		pf_remove_if_empty_ruleset(ruleset);
1508 
1509 		PF_UNLOCK();
1510 		break;
1511 	}
1512 
1513 	case DIOCCLRSTATES: {
1514 		struct pf_state		*s, *nexts;
1515 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1516 		u_int			 killed = 0;
1517 
1518 		PF_LOCK();
1519 		PF_STATE_ENTER_WRITE();
1520 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
1521 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1522 
1523 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1524 			    s->kif->pfik_name)) {
1525 #if NPFSYNC > 0
1526 				/* don't send out individual delete messages */
1527 				SET(s->state_flags, PFSTATE_NOSYNC);
1528 #endif	/* NPFSYNC > 0 */
1529 				pf_remove_state(s);
1530 				killed++;
1531 			}
1532 		}
1533 		PF_STATE_EXIT_WRITE();
1534 		psk->psk_killed = killed;
1535 #if NPFSYNC > 0
1536 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1537 #endif	/* NPFSYNC > 0 */
1538 		PF_UNLOCK();
1539 		break;
1540 	}
1541 
1542 	case DIOCKILLSTATES: {
1543 		struct pf_state		*s, *nexts;
1544 		struct pf_state_item	*si, *sit;
1545 		struct pf_state_key	*sk, key;
1546 		struct pf_addr		*srcaddr, *dstaddr;
1547 		u_int16_t		 srcport, dstport;
1548 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1549 		u_int			 i, killed = 0;
1550 		const int 		 dirs[] = { PF_IN, PF_OUT };
1551 		int			 sidx, didx;
1552 
1553 		if (psk->psk_pfcmp.id) {
1554 			if (psk->psk_pfcmp.creatorid == 0)
1555 				psk->psk_pfcmp.creatorid = pf_status.hostid;
1556 			PF_LOCK();
1557 			PF_STATE_ENTER_WRITE();
1558 			if ((s = pf_find_state_byid(&psk->psk_pfcmp))) {
1559 				pf_remove_state(s);
1560 				psk->psk_killed = 1;
1561 			}
1562 			PF_STATE_EXIT_WRITE();
1563 			PF_UNLOCK();
1564 			break;
1565 		}
1566 
1567 		if (psk->psk_af && psk->psk_proto &&
1568 		    psk->psk_src.port_op == PF_OP_EQ &&
1569 		    psk->psk_dst.port_op == PF_OP_EQ) {
1570 
1571 			key.af = psk->psk_af;
1572 			key.proto = psk->psk_proto;
1573 			key.rdomain = psk->psk_rdomain;
1574 
1575 			PF_LOCK();
1576 			PF_STATE_ENTER_WRITE();
1577 			for (i = 0; i < nitems(dirs); i++) {
1578 				if (dirs[i] == PF_IN) {
1579 					sidx = 0;
1580 					didx = 1;
1581 				} else {
1582 					sidx = 1;
1583 					didx = 0;
1584 				}
1585 				pf_addrcpy(&key.addr[sidx],
1586 				    &psk->psk_src.addr.v.a.addr, key.af);
1587 				pf_addrcpy(&key.addr[didx],
1588 				    &psk->psk_dst.addr.v.a.addr, key.af);
1589 				key.port[sidx] = psk->psk_src.port[0];
1590 				key.port[didx] = psk->psk_dst.port[0];
1591 
1592 				sk = RB_FIND(pf_state_tree, &pf_statetbl, &key);
1593 				if (sk == NULL)
1594 					continue;
1595 
1596 				TAILQ_FOREACH_SAFE(si, &sk->states, entry, sit)
1597 					if (((si->s->key[PF_SK_WIRE]->af ==
1598 					    si->s->key[PF_SK_STACK]->af &&
1599 					    sk == (dirs[i] == PF_IN ?
1600 					    si->s->key[PF_SK_WIRE] :
1601 					    si->s->key[PF_SK_STACK])) ||
1602 					    (si->s->key[PF_SK_WIRE]->af !=
1603 					    si->s->key[PF_SK_STACK]->af &&
1604 					    dirs[i] == PF_IN &&
1605 					    (sk == si->s->key[PF_SK_STACK] ||
1606 					    sk == si->s->key[PF_SK_WIRE]))) &&
1607 					    (!psk->psk_ifname[0] ||
1608 					    (si->s->kif != pfi_all &&
1609 					    !strcmp(psk->psk_ifname,
1610 					    si->s->kif->pfik_name)))) {
1611 						pf_remove_state(si->s);
1612 						killed++;
1613 					}
1614 			}
1615 			if (killed)
1616 				psk->psk_killed = killed;
1617 			PF_STATE_EXIT_WRITE();
1618 			PF_UNLOCK();
1619 			break;
1620 		}
1621 
1622 		PF_LOCK();
1623 		PF_STATE_ENTER_WRITE();
1624 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s;
1625 		    s = nexts) {
1626 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1627 
1628 			if (s->direction == PF_OUT) {
1629 				sk = s->key[PF_SK_STACK];
1630 				srcaddr = &sk->addr[1];
1631 				dstaddr = &sk->addr[0];
1632 				srcport = sk->port[1];
1633 				dstport = sk->port[0];
1634 			} else {
1635 				sk = s->key[PF_SK_WIRE];
1636 				srcaddr = &sk->addr[0];
1637 				dstaddr = &sk->addr[1];
1638 				srcport = sk->port[0];
1639 				dstport = sk->port[1];
1640 			}
1641 			if ((!psk->psk_af || sk->af == psk->psk_af)
1642 			    && (!psk->psk_proto || psk->psk_proto ==
1643 			    sk->proto) && psk->psk_rdomain == sk->rdomain &&
1644 			    pf_match_addr(psk->psk_src.neg,
1645 			    &psk->psk_src.addr.v.a.addr,
1646 			    &psk->psk_src.addr.v.a.mask,
1647 			    srcaddr, sk->af) &&
1648 			    pf_match_addr(psk->psk_dst.neg,
1649 			    &psk->psk_dst.addr.v.a.addr,
1650 			    &psk->psk_dst.addr.v.a.mask,
1651 			    dstaddr, sk->af) &&
1652 			    (psk->psk_src.port_op == 0 ||
1653 			    pf_match_port(psk->psk_src.port_op,
1654 			    psk->psk_src.port[0], psk->psk_src.port[1],
1655 			    srcport)) &&
1656 			    (psk->psk_dst.port_op == 0 ||
1657 			    pf_match_port(psk->psk_dst.port_op,
1658 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1659 			    dstport)) &&
1660 			    (!psk->psk_label[0] || (s->rule.ptr->label[0] &&
1661 			    !strcmp(psk->psk_label, s->rule.ptr->label))) &&
1662 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1663 			    s->kif->pfik_name))) {
1664 				pf_remove_state(s);
1665 				killed++;
1666 			}
1667 		}
1668 		psk->psk_killed = killed;
1669 		PF_STATE_EXIT_WRITE();
1670 		PF_UNLOCK();
1671 		break;
1672 	}
1673 
1674 #if NPFSYNC > 0
1675 	case DIOCADDSTATE: {
1676 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1677 		struct pfsync_state	*sp = &ps->state;
1678 
1679 		if (sp->timeout >= PFTM_MAX) {
1680 			error = EINVAL;
1681 			break;
1682 		}
1683 		PF_LOCK();
1684 		PF_STATE_ENTER_WRITE();
1685 		error = pfsync_state_import(sp, PFSYNC_SI_IOCTL);
1686 		PF_STATE_EXIT_WRITE();
1687 		PF_UNLOCK();
1688 		break;
1689 	}
1690 #endif	/* NPFSYNC > 0 */
1691 
1692 	case DIOCGETSTATE: {
1693 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1694 		struct pf_state		*s;
1695 		struct pf_state_cmp	 id_key;
1696 
1697 		memset(&id_key, 0, sizeof(id_key));
1698 		id_key.id = ps->state.id;
1699 		id_key.creatorid = ps->state.creatorid;
1700 
1701 		PF_STATE_ENTER_READ();
1702 		s = pf_find_state_byid(&id_key);
1703 		s = pf_state_ref(s);
1704 		PF_STATE_EXIT_READ();
1705 		if (s == NULL) {
1706 			error = ENOENT;
1707 			break;
1708 		}
1709 
1710 		pf_state_export(&ps->state, s);
1711 		pf_state_unref(s);
1712 		break;
1713 	}
1714 
1715 	case DIOCGETSTATES: {
1716 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1717 		struct pf_state		*state;
1718 		struct pfsync_state	*p, *pstore;
1719 		u_int32_t		 nr = 0;
1720 
1721 		if (ps->ps_len == 0) {
1722 			nr = pf_status.states;
1723 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1724 			break;
1725 		}
1726 
1727 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1728 
1729 		p = ps->ps_states;
1730 
1731 		PF_STATE_ENTER_READ();
1732 		state = TAILQ_FIRST(&state_list);
1733 		while (state) {
1734 			if (state->timeout != PFTM_UNLINKED) {
1735 				if ((nr+1) * sizeof(*p) > ps->ps_len)
1736 					break;
1737 				pf_state_export(pstore, state);
1738 				error = copyout(pstore, p, sizeof(*p));
1739 				if (error) {
1740 					free(pstore, M_TEMP, sizeof(*pstore));
1741 					PF_STATE_EXIT_READ();
1742 					goto fail;
1743 				}
1744 				p++;
1745 				nr++;
1746 			}
1747 			state = TAILQ_NEXT(state, entry_list);
1748 		}
1749 		PF_STATE_EXIT_READ();
1750 
1751 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1752 
1753 		free(pstore, M_TEMP, sizeof(*pstore));
1754 		break;
1755 	}
1756 
1757 	case DIOCGETSTATUS: {
1758 		struct pf_status *s = (struct pf_status *)addr;
1759 		PF_LOCK();
1760 		memcpy(s, &pf_status, sizeof(struct pf_status));
1761 		pfi_update_status(s->ifname, s);
1762 		PF_UNLOCK();
1763 		break;
1764 	}
1765 
1766 	case DIOCSETSTATUSIF: {
1767 		struct pfioc_iface	*pi = (struct pfioc_iface *)addr;
1768 
1769 		PF_LOCK();
1770 		if (pi->pfiio_name[0] == 0) {
1771 			memset(pf_status.ifname, 0, IFNAMSIZ);
1772 			PF_UNLOCK();
1773 			break;
1774 		}
1775 		strlcpy(pf_trans_set.statusif, pi->pfiio_name, IFNAMSIZ);
1776 		pf_trans_set.mask |= PF_TSET_STATUSIF;
1777 		PF_UNLOCK();
1778 		break;
1779 	}
1780 
1781 	case DIOCCLRSTATUS: {
1782 		struct pfioc_iface	*pi = (struct pfioc_iface *)addr;
1783 
1784 		PF_LOCK();
1785 		/* if ifname is specified, clear counters there only */
1786 		if (pi->pfiio_name[0]) {
1787 			pfi_update_status(pi->pfiio_name, NULL);
1788 			PF_UNLOCK();
1789 			break;
1790 		}
1791 
1792 		memset(pf_status.counters, 0, sizeof(pf_status.counters));
1793 		memset(pf_status.fcounters, 0, sizeof(pf_status.fcounters));
1794 		memset(pf_status.scounters, 0, sizeof(pf_status.scounters));
1795 		pf_status.since = time_uptime;
1796 
1797 		PF_UNLOCK();
1798 		break;
1799 	}
1800 
1801 	case DIOCNATLOOK: {
1802 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1803 		struct pf_state_key	*sk;
1804 		struct pf_state		*state;
1805 		struct pf_state_key_cmp	 key;
1806 		int			 m = 0, direction = pnl->direction;
1807 		int			 sidx, didx;
1808 
1809 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1810 		sidx = (direction == PF_IN) ? 1 : 0;
1811 		didx = (direction == PF_IN) ? 0 : 1;
1812 
1813 		if (!pnl->proto ||
1814 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1815 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1816 		    ((pnl->proto == IPPROTO_TCP ||
1817 		    pnl->proto == IPPROTO_UDP) &&
1818 		    (!pnl->dport || !pnl->sport)) ||
1819 		    pnl->rdomain > RT_TABLEID_MAX)
1820 			error = EINVAL;
1821 		else {
1822 			key.af = pnl->af;
1823 			key.proto = pnl->proto;
1824 			key.rdomain = pnl->rdomain;
1825 			pf_addrcpy(&key.addr[sidx], &pnl->saddr, pnl->af);
1826 			key.port[sidx] = pnl->sport;
1827 			pf_addrcpy(&key.addr[didx], &pnl->daddr, pnl->af);
1828 			key.port[didx] = pnl->dport;
1829 
1830 			PF_STATE_ENTER_READ();
1831 			state = pf_find_state_all(&key, direction, &m);
1832 			state = pf_state_ref(state);
1833 			PF_STATE_EXIT_READ();
1834 
1835 			if (m > 1)
1836 				error = E2BIG;	/* more than one state */
1837 			else if (state != NULL) {
1838 				sk = state->key[sidx];
1839 				pf_addrcpy(&pnl->rsaddr, &sk->addr[sidx],
1840 				    sk->af);
1841 				pnl->rsport = sk->port[sidx];
1842 				pf_addrcpy(&pnl->rdaddr, &sk->addr[didx],
1843 				    sk->af);
1844 				pnl->rdport = sk->port[didx];
1845 				pnl->rrdomain = sk->rdomain;
1846 			} else
1847 				error = ENOENT;
1848 			pf_state_unref(state);
1849 		}
1850 		break;
1851 	}
1852 
1853 	case DIOCSETTIMEOUT: {
1854 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1855 
1856 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1857 		    pt->seconds < 0) {
1858 			error = EINVAL;
1859 			goto fail;
1860 		}
1861 		PF_LOCK();
1862 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1863 			pt->seconds = 1;
1864 		pf_default_rule_new.timeout[pt->timeout] = pt->seconds;
1865 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1866 		PF_UNLOCK();
1867 		break;
1868 	}
1869 
1870 	case DIOCGETTIMEOUT: {
1871 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1872 
1873 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1874 			error = EINVAL;
1875 			goto fail;
1876 		}
1877 		PF_LOCK();
1878 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1879 		PF_UNLOCK();
1880 		break;
1881 	}
1882 
1883 	case DIOCGETLIMIT: {
1884 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1885 
1886 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1887 			error = EINVAL;
1888 			goto fail;
1889 		}
1890 		PF_LOCK();
1891 		pl->limit = pf_pool_limits[pl->index].limit;
1892 		PF_UNLOCK();
1893 		break;
1894 	}
1895 
1896 	case DIOCSETLIMIT: {
1897 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1898 
1899 		PF_LOCK();
1900 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1901 		    pf_pool_limits[pl->index].pp == NULL) {
1902 			error = EINVAL;
1903 			PF_UNLOCK();
1904 			goto fail;
1905 		}
1906 		if (((struct pool *)pf_pool_limits[pl->index].pp)->pr_nout >
1907 		    pl->limit) {
1908 			error = EBUSY;
1909 			PF_UNLOCK();
1910 			goto fail;
1911 		}
1912 		/* Fragments reference mbuf clusters. */
1913 		if (pl->index == PF_LIMIT_FRAGS && pl->limit > nmbclust) {
1914 			error = EINVAL;
1915 			PF_UNLOCK();
1916 			goto fail;
1917 		}
1918 
1919 		pf_pool_limits[pl->index].limit_new = pl->limit;
1920 		pl->limit = pf_pool_limits[pl->index].limit;
1921 		PF_UNLOCK();
1922 		break;
1923 	}
1924 
1925 	case DIOCSETDEBUG: {
1926 		u_int32_t	*level = (u_int32_t *)addr;
1927 
1928 		PF_LOCK();
1929 		pf_trans_set.debug = *level;
1930 		pf_trans_set.mask |= PF_TSET_DEBUG;
1931 		PF_UNLOCK();
1932 		break;
1933 	}
1934 
1935 	case DIOCGETRULESETS: {
1936 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
1937 		struct pf_ruleset	*ruleset;
1938 		struct pf_anchor	*anchor;
1939 
1940 		PF_LOCK();
1941 		pr->path[sizeof(pr->path) - 1] = 0;
1942 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
1943 			error = EINVAL;
1944 			PF_UNLOCK();
1945 			break;
1946 		}
1947 		pr->nr = 0;
1948 		if (ruleset == &pf_main_ruleset) {
1949 			/* XXX kludge for pf_main_ruleset */
1950 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
1951 				if (anchor->parent == NULL)
1952 					pr->nr++;
1953 		} else {
1954 			RB_FOREACH(anchor, pf_anchor_node,
1955 			    &ruleset->anchor->children)
1956 				pr->nr++;
1957 		}
1958 		PF_UNLOCK();
1959 		break;
1960 	}
1961 
1962 	case DIOCGETRULESET: {
1963 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
1964 		struct pf_ruleset	*ruleset;
1965 		struct pf_anchor	*anchor;
1966 		u_int32_t		 nr = 0;
1967 
1968 		PF_LOCK();
1969 		pr->path[sizeof(pr->path) - 1] = 0;
1970 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
1971 			error = EINVAL;
1972 			PF_UNLOCK();
1973 			break;
1974 		}
1975 		pr->name[0] = 0;
1976 		if (ruleset == &pf_main_ruleset) {
1977 			/* XXX kludge for pf_main_ruleset */
1978 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
1979 				if (anchor->parent == NULL && nr++ == pr->nr) {
1980 					strlcpy(pr->name, anchor->name,
1981 					    sizeof(pr->name));
1982 					PF_UNLOCK();
1983 					break;
1984 				}
1985 		} else {
1986 			RB_FOREACH(anchor, pf_anchor_node,
1987 			    &ruleset->anchor->children)
1988 				if (nr++ == pr->nr) {
1989 					strlcpy(pr->name, anchor->name,
1990 					    sizeof(pr->name));
1991 					PF_UNLOCK();
1992 					break;
1993 				}
1994 		}
1995 		if (!pr->name[0])
1996 			error = EBUSY;
1997 		PF_UNLOCK();
1998 		break;
1999 	}
2000 
2001 	case DIOCRCLRTABLES: {
2002 		struct pfioc_table *io = (struct pfioc_table *)addr;
2003 
2004 		if (io->pfrio_esize != 0) {
2005 			error = ENODEV;
2006 			break;
2007 		}
2008 		PF_LOCK();
2009 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2010 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2011 		PF_UNLOCK();
2012 		break;
2013 	}
2014 
2015 	case DIOCRADDTABLES: {
2016 		struct pfioc_table *io = (struct pfioc_table *)addr;
2017 
2018 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2019 			error = ENODEV;
2020 			break;
2021 		}
2022 		PF_LOCK();
2023 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2024 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2025 		PF_UNLOCK();
2026 		break;
2027 	}
2028 
2029 	case DIOCRDELTABLES: {
2030 		struct pfioc_table *io = (struct pfioc_table *)addr;
2031 
2032 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2033 			error = ENODEV;
2034 			break;
2035 		}
2036 		PF_LOCK();
2037 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2038 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2039 		PF_UNLOCK();
2040 		break;
2041 	}
2042 
2043 	case DIOCRGETTABLES: {
2044 		struct pfioc_table *io = (struct pfioc_table *)addr;
2045 
2046 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2047 			error = ENODEV;
2048 			break;
2049 		}
2050 		PF_LOCK();
2051 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2052 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2053 		PF_UNLOCK();
2054 		break;
2055 	}
2056 
2057 	case DIOCRGETTSTATS: {
2058 		struct pfioc_table *io = (struct pfioc_table *)addr;
2059 
2060 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2061 			error = ENODEV;
2062 			break;
2063 		}
2064 		PF_LOCK();
2065 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2066 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2067 		PF_UNLOCK();
2068 		break;
2069 	}
2070 
2071 	case DIOCRCLRTSTATS: {
2072 		struct pfioc_table *io = (struct pfioc_table *)addr;
2073 
2074 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2075 			error = ENODEV;
2076 			break;
2077 		}
2078 		PF_LOCK();
2079 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2080 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2081 		PF_UNLOCK();
2082 		break;
2083 	}
2084 
2085 	case DIOCRSETTFLAGS: {
2086 		struct pfioc_table *io = (struct pfioc_table *)addr;
2087 
2088 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2089 			error = ENODEV;
2090 			break;
2091 		}
2092 		PF_LOCK();
2093 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2094 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2095 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2096 		PF_UNLOCK();
2097 		break;
2098 	}
2099 
2100 	case DIOCRCLRADDRS: {
2101 		struct pfioc_table *io = (struct pfioc_table *)addr;
2102 
2103 		if (io->pfrio_esize != 0) {
2104 			error = ENODEV;
2105 			break;
2106 		}
2107 		PF_LOCK();
2108 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2109 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2110 		PF_UNLOCK();
2111 		break;
2112 	}
2113 
2114 	case DIOCRADDADDRS: {
2115 		struct pfioc_table *io = (struct pfioc_table *)addr;
2116 
2117 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2118 			error = ENODEV;
2119 			break;
2120 		}
2121 		PF_LOCK();
2122 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2123 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2124 		    PFR_FLAG_USERIOCTL);
2125 		PF_UNLOCK();
2126 		break;
2127 	}
2128 
2129 	case DIOCRDELADDRS: {
2130 		struct pfioc_table *io = (struct pfioc_table *)addr;
2131 
2132 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2133 			error = ENODEV;
2134 			break;
2135 		}
2136 		PF_LOCK();
2137 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2138 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2139 		    PFR_FLAG_USERIOCTL);
2140 		PF_UNLOCK();
2141 		break;
2142 	}
2143 
2144 	case DIOCRSETADDRS: {
2145 		struct pfioc_table *io = (struct pfioc_table *)addr;
2146 
2147 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2148 			error = ENODEV;
2149 			break;
2150 		}
2151 		PF_LOCK();
2152 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2153 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2154 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2155 		    PFR_FLAG_USERIOCTL, 0);
2156 		PF_UNLOCK();
2157 		break;
2158 	}
2159 
2160 	case DIOCRGETADDRS: {
2161 		struct pfioc_table *io = (struct pfioc_table *)addr;
2162 
2163 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2164 			error = ENODEV;
2165 			break;
2166 		}
2167 		PF_LOCK();
2168 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2169 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2170 		PF_UNLOCK();
2171 		break;
2172 	}
2173 
2174 	case DIOCRGETASTATS: {
2175 		struct pfioc_table *io = (struct pfioc_table *)addr;
2176 
2177 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2178 			error = ENODEV;
2179 			break;
2180 		}
2181 		PF_LOCK();
2182 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2183 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2184 		PF_UNLOCK();
2185 		break;
2186 	}
2187 
2188 	case DIOCRCLRASTATS: {
2189 		struct pfioc_table *io = (struct pfioc_table *)addr;
2190 
2191 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2192 			error = ENODEV;
2193 			break;
2194 		}
2195 		PF_LOCK();
2196 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2197 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2198 		    PFR_FLAG_USERIOCTL);
2199 		PF_UNLOCK();
2200 		break;
2201 	}
2202 
2203 	case DIOCRTSTADDRS: {
2204 		struct pfioc_table *io = (struct pfioc_table *)addr;
2205 
2206 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2207 			error = ENODEV;
2208 			break;
2209 		}
2210 		PF_LOCK();
2211 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2212 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2213 		    PFR_FLAG_USERIOCTL);
2214 		PF_UNLOCK();
2215 		break;
2216 	}
2217 
2218 	case DIOCRINADEFINE: {
2219 		struct pfioc_table *io = (struct pfioc_table *)addr;
2220 
2221 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2222 			error = ENODEV;
2223 			break;
2224 		}
2225 		PF_LOCK();
2226 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2227 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2228 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2229 		PF_UNLOCK();
2230 		break;
2231 	}
2232 
2233 	case DIOCOSFPADD: {
2234 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2235 		PF_LOCK();
2236 		error = pf_osfp_add(io);
2237 		PF_UNLOCK();
2238 		break;
2239 	}
2240 
2241 	case DIOCOSFPGET: {
2242 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2243 		PF_LOCK();
2244 		error = pf_osfp_get(io);
2245 		PF_UNLOCK();
2246 		break;
2247 	}
2248 
2249 	case DIOCXBEGIN: {
2250 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2251 		struct pfioc_trans_e	*ioe;
2252 		struct pfr_table	*table;
2253 		int			 i;
2254 
2255 		if (io->esize != sizeof(*ioe)) {
2256 			error = ENODEV;
2257 			goto fail;
2258 		}
2259 		PF_LOCK();
2260 		ioe = malloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2261 		table = malloc(sizeof(*table), M_TEMP, M_WAITOK);
2262 		pf_default_rule_new = pf_default_rule;
2263 		memset(&pf_trans_set, 0, sizeof(pf_trans_set));
2264 		for (i = 0; i < io->size; i++) {
2265 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2266 				free(table, M_TEMP, sizeof(*table));
2267 				free(ioe, M_TEMP, sizeof(*ioe));
2268 				error = EFAULT;
2269 				PF_UNLOCK();
2270 				goto fail;
2271 			}
2272 			if (strnlen(ioe->anchor, sizeof(ioe->anchor)) ==
2273 			    sizeof(ioe->anchor)) {
2274 				free(table, M_TEMP, sizeof(*table));
2275 				free(ioe, M_TEMP, sizeof(*ioe));
2276 				error = ENAMETOOLONG;
2277 				PF_UNLOCK();
2278 				goto fail;
2279 			}
2280 			switch (ioe->type) {
2281 			case PF_TRANS_TABLE:
2282 				memset(table, 0, sizeof(*table));
2283 				strlcpy(table->pfrt_anchor, ioe->anchor,
2284 				    sizeof(table->pfrt_anchor));
2285 				if ((error = pfr_ina_begin(table,
2286 				    &ioe->ticket, NULL, 0))) {
2287 					free(table, M_TEMP, sizeof(*table));
2288 					free(ioe, M_TEMP, sizeof(*ioe));
2289 					PF_UNLOCK();
2290 					goto fail;
2291 				}
2292 				break;
2293 			case PF_TRANS_RULESET:
2294 				if ((error = pf_begin_rules(&ioe->ticket,
2295 				    ioe->anchor))) {
2296 					free(table, M_TEMP, sizeof(*table));
2297 					free(ioe, M_TEMP, sizeof(*ioe));
2298 					PF_UNLOCK();
2299 					goto fail;
2300 				}
2301 				break;
2302 			default:
2303 				free(table, M_TEMP, sizeof(*table));
2304 				free(ioe, M_TEMP, sizeof(*ioe));
2305 				error = EINVAL;
2306 				PF_UNLOCK();
2307 				goto fail;
2308 			}
2309 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2310 				free(table, M_TEMP, sizeof(*table));
2311 				free(ioe, M_TEMP, sizeof(*ioe));
2312 				error = EFAULT;
2313 				PF_UNLOCK();
2314 				goto fail;
2315 			}
2316 		}
2317 		free(table, M_TEMP, sizeof(*table));
2318 		free(ioe, M_TEMP, sizeof(*ioe));
2319 		PF_UNLOCK();
2320 		break;
2321 	}
2322 
2323 	case DIOCXROLLBACK: {
2324 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2325 		struct pfioc_trans_e	*ioe;
2326 		struct pfr_table	*table;
2327 		int			 i;
2328 
2329 		if (io->esize != sizeof(*ioe)) {
2330 			error = ENODEV;
2331 			goto fail;
2332 		}
2333 		PF_LOCK();
2334 		ioe = malloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2335 		table = malloc(sizeof(*table), M_TEMP, M_WAITOK);
2336 		for (i = 0; i < io->size; i++) {
2337 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2338 				free(table, M_TEMP, sizeof(*table));
2339 				free(ioe, M_TEMP, sizeof(*ioe));
2340 				error = EFAULT;
2341 				PF_UNLOCK();
2342 				goto fail;
2343 			}
2344 			if (strnlen(ioe->anchor, sizeof(ioe->anchor)) ==
2345 			    sizeof(ioe->anchor)) {
2346 				free(table, M_TEMP, sizeof(*table));
2347 				free(ioe, M_TEMP, sizeof(*ioe));
2348 				error = ENAMETOOLONG;
2349 				PF_UNLOCK();
2350 				goto fail;
2351 			}
2352 			switch (ioe->type) {
2353 			case PF_TRANS_TABLE:
2354 				memset(table, 0, sizeof(*table));
2355 				strlcpy(table->pfrt_anchor, ioe->anchor,
2356 				    sizeof(table->pfrt_anchor));
2357 				if ((error = pfr_ina_rollback(table,
2358 				    ioe->ticket, NULL, 0))) {
2359 					free(table, M_TEMP, sizeof(*table));
2360 					free(ioe, M_TEMP, sizeof(*ioe));
2361 					PF_UNLOCK();
2362 					goto fail; /* really bad */
2363 				}
2364 				break;
2365 			case PF_TRANS_RULESET:
2366 				if ((error = pf_rollback_rules(ioe->ticket,
2367 				    ioe->anchor))) {
2368 					free(table, M_TEMP, sizeof(*table));
2369 					free(ioe, M_TEMP, sizeof(*ioe));
2370 					PF_UNLOCK();
2371 					goto fail; /* really bad */
2372 				}
2373 				break;
2374 			default:
2375 				free(table, M_TEMP, sizeof(*table));
2376 				free(ioe, M_TEMP, sizeof(*ioe));
2377 				error = EINVAL;
2378 				PF_UNLOCK();
2379 				goto fail; /* really bad */
2380 			}
2381 		}
2382 		free(table, M_TEMP, sizeof(*table));
2383 		free(ioe, M_TEMP, sizeof(*ioe));
2384 		PF_UNLOCK();
2385 		break;
2386 	}
2387 
2388 	case DIOCXCOMMIT: {
2389 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2390 		struct pfioc_trans_e	*ioe;
2391 		struct pfr_table	*table;
2392 		struct pf_ruleset	*rs;
2393 		int			 i;
2394 
2395 		if (io->esize != sizeof(*ioe)) {
2396 			error = ENODEV;
2397 			goto fail;
2398 		}
2399 		PF_LOCK();
2400 		ioe = malloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2401 		table = malloc(sizeof(*table), M_TEMP, M_WAITOK);
2402 		/* first makes sure everything will succeed */
2403 		for (i = 0; i < io->size; i++) {
2404 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2405 				free(table, M_TEMP, sizeof(*table));
2406 				free(ioe, M_TEMP, sizeof(*ioe));
2407 				error = EFAULT;
2408 				PF_UNLOCK();
2409 				goto fail;
2410 			}
2411 			if (strnlen(ioe->anchor, sizeof(ioe->anchor)) ==
2412 			    sizeof(ioe->anchor)) {
2413 				free(table, M_TEMP, sizeof(*table));
2414 				free(ioe, M_TEMP, sizeof(*ioe));
2415 				error = ENAMETOOLONG;
2416 				PF_UNLOCK();
2417 				goto fail;
2418 			}
2419 			switch (ioe->type) {
2420 			case PF_TRANS_TABLE:
2421 				rs = pf_find_ruleset(ioe->anchor);
2422 				if (rs == NULL || !rs->topen || ioe->ticket !=
2423 				     rs->tticket) {
2424 					free(table, M_TEMP, sizeof(*table));
2425 					free(ioe, M_TEMP, sizeof(*ioe));
2426 					error = EBUSY;
2427 					PF_UNLOCK();
2428 					goto fail;
2429 				}
2430 				break;
2431 			case PF_TRANS_RULESET:
2432 				rs = pf_find_ruleset(ioe->anchor);
2433 				if (rs == NULL ||
2434 				    !rs->rules.inactive.open ||
2435 				    rs->rules.inactive.ticket !=
2436 				    ioe->ticket) {
2437 					free(table, M_TEMP, sizeof(*table));
2438 					free(ioe, M_TEMP, sizeof(*ioe));
2439 					error = EBUSY;
2440 					PF_UNLOCK();
2441 					goto fail;
2442 				}
2443 				break;
2444 			default:
2445 				free(table, M_TEMP, sizeof(*table));
2446 				free(ioe, M_TEMP, sizeof(*ioe));
2447 				error = EINVAL;
2448 				PF_UNLOCK();
2449 				goto fail;
2450 			}
2451 		}
2452 
2453 		/*
2454 		 * Checked already in DIOCSETLIMIT, but check again as the
2455 		 * situation might have changed.
2456 		 */
2457 		for (i = 0; i < PF_LIMIT_MAX; i++) {
2458 			if (((struct pool *)pf_pool_limits[i].pp)->pr_nout >
2459 			    pf_pool_limits[i].limit_new) {
2460 				free(table, M_TEMP, sizeof(*table));
2461 				free(ioe, M_TEMP, sizeof(*ioe));
2462 				error = EBUSY;
2463 				PF_UNLOCK();
2464 				goto fail;
2465 			}
2466 		}
2467 		/* now do the commit - no errors should happen here */
2468 		for (i = 0; i < io->size; i++) {
2469 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2470 				free(table, M_TEMP, sizeof(*table));
2471 				free(ioe, M_TEMP, sizeof(*ioe));
2472 				error = EFAULT;
2473 				PF_UNLOCK();
2474 				goto fail;
2475 			}
2476 			if (strnlen(ioe->anchor, sizeof(ioe->anchor)) ==
2477 			    sizeof(ioe->anchor)) {
2478 				free(table, M_TEMP, sizeof(*table));
2479 				free(ioe, M_TEMP, sizeof(*ioe));
2480 				error = ENAMETOOLONG;
2481 				PF_UNLOCK();
2482 				goto fail;
2483 			}
2484 			switch (ioe->type) {
2485 			case PF_TRANS_TABLE:
2486 				memset(table, 0, sizeof(*table));
2487 				strlcpy(table->pfrt_anchor, ioe->anchor,
2488 				    sizeof(table->pfrt_anchor));
2489 				if ((error = pfr_ina_commit(table, ioe->ticket,
2490 				    NULL, NULL, 0))) {
2491 					free(table, M_TEMP, sizeof(*table));
2492 					free(ioe, M_TEMP, sizeof(*ioe));
2493 					PF_UNLOCK();
2494 					goto fail; /* really bad */
2495 				}
2496 				break;
2497 			case PF_TRANS_RULESET:
2498 				if ((error = pf_commit_rules(ioe->ticket,
2499 				    ioe->anchor))) {
2500 					free(table, M_TEMP, sizeof(*table));
2501 					free(ioe, M_TEMP, sizeof(*ioe));
2502 					PF_UNLOCK();
2503 					goto fail; /* really bad */
2504 				}
2505 				break;
2506 			default:
2507 				free(table, M_TEMP, sizeof(*table));
2508 				free(ioe, M_TEMP, sizeof(*ioe));
2509 				error = EINVAL;
2510 				PF_UNLOCK();
2511 				goto fail; /* really bad */
2512 			}
2513 		}
2514 		for (i = 0; i < PF_LIMIT_MAX; i++) {
2515 			if (pf_pool_limits[i].limit_new !=
2516 			    pf_pool_limits[i].limit &&
2517 			    pool_sethardlimit(pf_pool_limits[i].pp,
2518 			    pf_pool_limits[i].limit_new, NULL, 0) != 0) {
2519 				free(table, M_TEMP, sizeof(*table));
2520 				free(ioe, M_TEMP, sizeof(*ioe));
2521 				error = EBUSY;
2522 				PF_UNLOCK();
2523 				goto fail; /* really bad */
2524 			}
2525 			pf_pool_limits[i].limit = pf_pool_limits[i].limit_new;
2526 		}
2527 		for (i = 0; i < PFTM_MAX; i++) {
2528 			int old = pf_default_rule.timeout[i];
2529 
2530 			pf_default_rule.timeout[i] =
2531 			    pf_default_rule_new.timeout[i];
2532 			if (pf_default_rule.timeout[i] == PFTM_INTERVAL &&
2533 			    pf_default_rule.timeout[i] < old)
2534 				task_add(net_tq(0), &pf_purge_task);
2535 		}
2536 		pfi_xcommit();
2537 		pf_trans_set_commit();
2538 		free(table, M_TEMP, sizeof(*table));
2539 		free(ioe, M_TEMP, sizeof(*ioe));
2540 		PF_UNLOCK();
2541 		break;
2542 	}
2543 
2544 	case DIOCGETSRCNODES: {
2545 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2546 		struct pf_src_node	*n, *p, *pstore;
2547 		u_int32_t		 nr = 0;
2548 		size_t			 space = psn->psn_len;
2549 
2550 		PF_LOCK();
2551 		if (space == 0) {
2552 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2553 				nr++;
2554 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2555 			PF_UNLOCK();
2556 			break;
2557 		}
2558 
2559 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2560 
2561 		p = psn->psn_src_nodes;
2562 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2563 			int	secs = time_uptime, diff;
2564 
2565 			if ((nr + 1) * sizeof(*p) > psn->psn_len)
2566 				break;
2567 
2568 			memcpy(pstore, n, sizeof(*pstore));
2569 			memset(&pstore->entry, 0, sizeof(pstore->entry));
2570 			pstore->rule.ptr = NULL;
2571 			pstore->kif = NULL;
2572 			pstore->rule.nr = n->rule.ptr->nr;
2573 			pstore->creation = secs - pstore->creation;
2574 			if (pstore->expire > secs)
2575 				pstore->expire -= secs;
2576 			else
2577 				pstore->expire = 0;
2578 
2579 			/* adjust the connection rate estimate */
2580 			diff = secs - n->conn_rate.last;
2581 			if (diff >= n->conn_rate.seconds)
2582 				pstore->conn_rate.count = 0;
2583 			else
2584 				pstore->conn_rate.count -=
2585 				    n->conn_rate.count * diff /
2586 				    n->conn_rate.seconds;
2587 
2588 			error = copyout(pstore, p, sizeof(*p));
2589 			if (error) {
2590 				free(pstore, M_TEMP, sizeof(*pstore));
2591 				PF_UNLOCK();
2592 				goto fail;
2593 			}
2594 			p++;
2595 			nr++;
2596 		}
2597 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2598 
2599 		free(pstore, M_TEMP, sizeof(*pstore));
2600 		PF_UNLOCK();
2601 		break;
2602 	}
2603 
2604 	case DIOCCLRSRCNODES: {
2605 		struct pf_src_node	*n;
2606 		struct pf_state		*state;
2607 
2608 		PF_LOCK();
2609 		PF_STATE_ENTER_WRITE();
2610 		RB_FOREACH(state, pf_state_tree_id, &tree_id)
2611 			pf_src_tree_remove_state(state);
2612 		PF_STATE_EXIT_WRITE();
2613 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2614 			n->expire = 1;
2615 		pf_purge_expired_src_nodes();
2616 		PF_UNLOCK();
2617 		break;
2618 	}
2619 
2620 	case DIOCKILLSRCNODES: {
2621 		struct pf_src_node	*sn;
2622 		struct pf_state		*s;
2623 		struct pfioc_src_node_kill *psnk =
2624 		    (struct pfioc_src_node_kill *)addr;
2625 		u_int			killed = 0;
2626 
2627 		PF_LOCK();
2628 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
2629 			if (pf_match_addr(psnk->psnk_src.neg,
2630 				&psnk->psnk_src.addr.v.a.addr,
2631 				&psnk->psnk_src.addr.v.a.mask,
2632 				&sn->addr, sn->af) &&
2633 			    pf_match_addr(psnk->psnk_dst.neg,
2634 				&psnk->psnk_dst.addr.v.a.addr,
2635 				&psnk->psnk_dst.addr.v.a.mask,
2636 				&sn->raddr, sn->af)) {
2637 				/* Handle state to src_node linkage */
2638 				if (sn->states != 0) {
2639 					PF_ASSERT_LOCKED();
2640 					PF_STATE_ENTER_WRITE();
2641 					RB_FOREACH(s, pf_state_tree_id,
2642 					   &tree_id)
2643 						pf_state_rm_src_node(s, sn);
2644 					PF_STATE_EXIT_WRITE();
2645 				}
2646 				sn->expire = 1;
2647 				killed++;
2648 			}
2649 		}
2650 
2651 		if (killed > 0)
2652 			pf_purge_expired_src_nodes();
2653 
2654 		psnk->psnk_killed = killed;
2655 		PF_UNLOCK();
2656 		break;
2657 	}
2658 
2659 	case DIOCSETHOSTID: {
2660 		u_int32_t	*hostid = (u_int32_t *)addr;
2661 
2662 		PF_LOCK();
2663 		if (*hostid == 0)
2664 			pf_trans_set.hostid = arc4random();
2665 		else
2666 			pf_trans_set.hostid = *hostid;
2667 		pf_trans_set.mask |= PF_TSET_HOSTID;
2668 		PF_UNLOCK();
2669 		break;
2670 	}
2671 
2672 	case DIOCOSFPFLUSH:
2673 		PF_LOCK();
2674 		pf_osfp_flush();
2675 		PF_UNLOCK();
2676 		break;
2677 
2678 	case DIOCIGETIFACES: {
2679 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2680 
2681 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
2682 			error = ENODEV;
2683 			break;
2684 		}
2685 		PF_LOCK();
2686 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2687 		    &io->pfiio_size);
2688 		PF_UNLOCK();
2689 		break;
2690 	}
2691 
2692 	case DIOCSETIFFLAG: {
2693 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2694 
2695 		PF_LOCK();
2696 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
2697 		PF_UNLOCK();
2698 		break;
2699 	}
2700 
2701 	case DIOCCLRIFFLAG: {
2702 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2703 
2704 		PF_LOCK();
2705 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
2706 		PF_UNLOCK();
2707 		break;
2708 	}
2709 
2710 	case DIOCSETREASS: {
2711 		u_int32_t	*reass = (u_int32_t *)addr;
2712 
2713 		PF_LOCK();
2714 		pf_trans_set.reass = *reass;
2715 		pf_trans_set.mask |= PF_TSET_REASS;
2716 		PF_UNLOCK();
2717 		break;
2718 	}
2719 
2720 	case DIOCSETSYNFLWATS: {
2721 		struct pfioc_synflwats *io = (struct pfioc_synflwats *)addr;
2722 
2723 		PF_LOCK();
2724 		error = pf_syncookies_setwats(io->hiwat, io->lowat);
2725 		PF_UNLOCK();
2726 		break;
2727 	}
2728 
2729 	case DIOCGETSYNFLWATS: {
2730 		struct pfioc_synflwats *io = (struct pfioc_synflwats *)addr;
2731 
2732 		PF_LOCK();
2733 		error = pf_syncookies_getwats(io);
2734 		PF_UNLOCK();
2735 		break;
2736 	}
2737 
2738 	case DIOCSETSYNCOOKIES: {
2739 		u_int8_t	*mode = (u_int8_t *)addr;
2740 
2741 		PF_LOCK();
2742 		error = pf_syncookies_setmode(*mode);
2743 		PF_UNLOCK();
2744 		break;
2745 	}
2746 
2747 	default:
2748 		error = ENODEV;
2749 		break;
2750 	}
2751 fail:
2752 	NET_UNLOCK();
2753 	return (error);
2754 }
2755 
2756 void
2757 pf_trans_set_commit(void)
2758 {
2759 	if (pf_trans_set.mask & PF_TSET_STATUSIF)
2760 		strlcpy(pf_status.ifname, pf_trans_set.statusif, IFNAMSIZ);
2761 	if (pf_trans_set.mask & PF_TSET_DEBUG)
2762 		pf_status.debug = pf_trans_set.debug;
2763 	if (pf_trans_set.mask & PF_TSET_HOSTID)
2764 		pf_status.hostid = pf_trans_set.hostid;
2765 	if (pf_trans_set.mask & PF_TSET_REASS)
2766 		pf_status.reass = pf_trans_set.reass;
2767 }
2768 
2769 void
2770 pf_pool_copyin(struct pf_pool *from, struct pf_pool *to)
2771 {
2772 	memmove(to, from, sizeof(*to));
2773 	to->kif = NULL;
2774 }
2775 
2776 int
2777 pf_rule_copyin(struct pf_rule *from, struct pf_rule *to,
2778     struct pf_ruleset *ruleset)
2779 {
2780 	int i;
2781 
2782 	to->src = from->src;
2783 	to->dst = from->dst;
2784 
2785 	/* XXX union skip[] */
2786 
2787 	strlcpy(to->label, from->label, sizeof(to->label));
2788 	strlcpy(to->ifname, from->ifname, sizeof(to->ifname));
2789 	strlcpy(to->rcv_ifname, from->rcv_ifname, sizeof(to->rcv_ifname));
2790 	strlcpy(to->qname, from->qname, sizeof(to->qname));
2791 	strlcpy(to->pqname, from->pqname, sizeof(to->pqname));
2792 	strlcpy(to->tagname, from->tagname, sizeof(to->tagname));
2793 	strlcpy(to->match_tagname, from->match_tagname,
2794 	    sizeof(to->match_tagname));
2795 	strlcpy(to->overload_tblname, from->overload_tblname,
2796 	    sizeof(to->overload_tblname));
2797 
2798 	pf_pool_copyin(&from->nat, &to->nat);
2799 	pf_pool_copyin(&from->rdr, &to->rdr);
2800 	pf_pool_copyin(&from->route, &to->route);
2801 
2802 	if (pf_kif_setup(to->ifname, &to->kif))
2803 		return (EINVAL);
2804 	if (pf_kif_setup(to->rcv_ifname, &to->rcv_kif))
2805 		return (EINVAL);
2806 	if (to->overload_tblname[0]) {
2807 		if ((to->overload_tbl = pfr_attach_table(ruleset,
2808 		    to->overload_tblname, 0)) == NULL)
2809 			return (EINVAL);
2810 		else
2811 			to->overload_tbl->pfrkt_flags |= PFR_TFLAG_ACTIVE;
2812 	}
2813 
2814 	if (pf_kif_setup(to->rdr.ifname, &to->rdr.kif))
2815 		return (EINVAL);
2816 	if (pf_kif_setup(to->nat.ifname, &to->nat.kif))
2817 		return (EINVAL);
2818 	if (pf_kif_setup(to->route.ifname, &to->route.kif))
2819 		return (EINVAL);
2820 
2821 	to->os_fingerprint = from->os_fingerprint;
2822 
2823 	to->rtableid = from->rtableid;
2824 	if (to->rtableid >= 0 && !rtable_exists(to->rtableid))
2825 		return (EBUSY);
2826 	to->onrdomain = from->onrdomain;
2827 	if (to->onrdomain >= 0 && !rtable_exists(to->onrdomain))
2828 		return (EBUSY);
2829 	if (to->onrdomain >= 0)		/* make sure it is a real rdomain */
2830 		to->onrdomain = rtable_l2(to->onrdomain);
2831 
2832 	for (i = 0; i < PFTM_MAX; i++)
2833 		to->timeout[i] = from->timeout[i];
2834 	to->states_tot = from->states_tot;
2835 	to->max_states = from->max_states;
2836 	to->max_src_nodes = from->max_src_nodes;
2837 	to->max_src_states = from->max_src_states;
2838 	to->max_src_conn = from->max_src_conn;
2839 	to->max_src_conn_rate.limit = from->max_src_conn_rate.limit;
2840 	to->max_src_conn_rate.seconds = from->max_src_conn_rate.seconds;
2841 	pf_init_threshold(&to->pktrate, from->pktrate.limit,
2842 	    from->pktrate.seconds);
2843 
2844 	if (to->qname[0] != 0) {
2845 		if ((to->qid = pf_qname2qid(to->qname, 0)) == 0)
2846 			return (EBUSY);
2847 		if (to->pqname[0] != 0) {
2848 			if ((to->pqid = pf_qname2qid(to->pqname, 0)) == 0)
2849 				return (EBUSY);
2850 		} else
2851 			to->pqid = to->qid;
2852 	}
2853 	to->rt_listid = from->rt_listid;
2854 	to->prob = from->prob;
2855 	to->return_icmp = from->return_icmp;
2856 	to->return_icmp6 = from->return_icmp6;
2857 	to->max_mss = from->max_mss;
2858 	if (to->tagname[0])
2859 		if ((to->tag = pf_tagname2tag(to->tagname, 1)) == 0)
2860 			return (EBUSY);
2861 	if (to->match_tagname[0])
2862 		if ((to->match_tag = pf_tagname2tag(to->match_tagname, 1)) == 0)
2863 			return (EBUSY);
2864 	to->scrub_flags = from->scrub_flags;
2865 	to->uid = from->uid;
2866 	to->gid = from->gid;
2867 	to->rule_flag = from->rule_flag;
2868 	to->action = from->action;
2869 	to->direction = from->direction;
2870 	to->log = from->log;
2871 	to->logif = from->logif;
2872 #if NPFLOG > 0
2873 	if (!to->log)
2874 		to->logif = 0;
2875 #endif	/* NPFLOG > 0 */
2876 	to->quick = from->quick;
2877 	to->ifnot = from->ifnot;
2878 	to->rcvifnot = from->rcvifnot;
2879 	to->match_tag_not = from->match_tag_not;
2880 	to->keep_state = from->keep_state;
2881 	to->af = from->af;
2882 	to->naf = from->naf;
2883 	to->proto = from->proto;
2884 	to->type = from->type;
2885 	to->code = from->code;
2886 	to->flags = from->flags;
2887 	to->flagset = from->flagset;
2888 	to->min_ttl = from->min_ttl;
2889 	to->allow_opts = from->allow_opts;
2890 	to->rt = from->rt;
2891 	to->return_ttl = from->return_ttl;
2892 	to->tos = from->tos;
2893 	to->set_tos = from->set_tos;
2894 	to->anchor_relative = from->anchor_relative; /* XXX */
2895 	to->anchor_wildcard = from->anchor_wildcard; /* XXX */
2896 	to->flush = from->flush;
2897 	to->divert.addr = from->divert.addr;
2898 	to->divert.port = from->divert.port;
2899 	to->divert.type = from->divert.type;
2900 	to->prio = from->prio;
2901 	to->set_prio[0] = from->set_prio[0];
2902 	to->set_prio[1] = from->set_prio[1];
2903 
2904 	return (0);
2905 }
2906