xref: /dragonfly/sys/net/pf/pf_ioctl.c (revision 10cbe914)
1 /*	$OpenBSD: pf_ioctl.c,v 1.182 2007/06/24 11:17:13 mcbride Exp $ */
2 
3 /*
4  * Copyright (c) 2010 The DragonFly Project.  All rights reserved.
5  *
6  * Copyright (c) 2001 Daniel Hartmeier
7  * Copyright (c) 2002,2003 Henning Brauer
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  *
14  *    - Redistributions of source code must retain the above copyright
15  *      notice, this list of conditions and the following disclaimer.
16  *    - Redistributions in binary form must reproduce the above
17  *      copyright notice, this list of conditions and the following
18  *      disclaimer in the documentation and/or other materials provided
19  *      with the distribution.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
29  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
31  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32  * POSSIBILITY OF SUCH DAMAGE.
33  *
34  * Effort sponsored in part by the Defense Advanced Research Projects
35  * Agency (DARPA) and Air Force Research Laboratory, Air Force
36  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
37  *
38  */
39 
40 #include "opt_inet.h"
41 #include "opt_inet6.h"
42 #include "use_pfsync.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/conf.h>
47 #include <sys/device.h>
48 #include <sys/mbuf.h>
49 #include <sys/filio.h>
50 #include <sys/fcntl.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/kernel.h>
54 #include <sys/kthread.h>
55 #include <sys/time.h>
56 #include <sys/proc.h>
57 #include <sys/malloc.h>
58 #include <sys/module.h>
59 #include <vm/vm_zone.h>
60 #include <sys/lock.h>
61 
62 #include <sys/thread2.h>
63 
64 #include <net/if.h>
65 #include <net/if_types.h>
66 #include <net/route.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_icmp.h>
74 
75 #include <net/pf/pfvar.h>
76 #include <sys/md5.h>
77 #include <net/pf/pfvar.h>
78 
79 #if NPFSYNC > 0
80 #include <net/pf/if_pfsync.h>
81 #endif /* NPFSYNC > 0 */
82 
83 #if NPFLOG > 0
84 #include <net/if_pflog.h>
85 #endif /* NPFLOG > 0 */
86 
87 #ifdef INET6
88 #include <netinet/ip6.h>
89 #include <netinet/in_pcb.h>
90 #endif /* INET6 */
91 
92 #ifdef ALTQ
93 #include <net/altq/altq.h>
94 #endif
95 
96 #include <machine/limits.h>
97 #include <net/pfil.h>
98 #include <sys/mutex.h>
99 
100 u_int rt_numfibs = RT_NUMFIBS;
101 
102 void			 init_zone_var(void);
103 void			 cleanup_pf_zone(void);
104 int			 pfattach(void);
105 struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
106 			    u_int8_t, u_int8_t, u_int8_t);
107 
108 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
109 void			 pf_empty_pool(struct pf_palist *);
110 #ifdef ALTQ
111 int			 pf_begin_altq(u_int32_t *);
112 int			 pf_rollback_altq(u_int32_t);
113 int			 pf_commit_altq(u_int32_t);
114 int			 pf_enable_altq(struct pf_altq *);
115 int			 pf_disable_altq(struct pf_altq *);
116 #endif /* ALTQ */
117 int			 pf_begin_rules(u_int32_t *, int, const char *);
118 int			 pf_rollback_rules(u_int32_t, int, char *);
119 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
120 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
121 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
122 int			 pf_commit_rules(u_int32_t, int, char *);
123 void			 pf_state_export(struct pfsync_state *,
124 			    struct pf_state_key *, struct pf_state *);
125 void			 pf_state_import(struct pfsync_state *,
126 			    struct pf_state_key *, struct pf_state *);
127 
128 struct pf_rule		 pf_default_rule;
129 struct lock		 pf_consistency_lock;
130 #ifdef ALTQ
131 static int		 pf_altq_running;
132 #endif
133 
134 #define	TAGID_MAX	 50000
135 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
136 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
137 
138 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
139 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
140 #endif
141 u_int16_t		 tagname2tag(struct pf_tags *, char *);
142 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
143 void			 tag_unref(struct pf_tags *, u_int16_t);
144 int			 pf_rtlabel_add(struct pf_addr_wrap *);
145 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
146 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
147 
148 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
149 
150 static cdev_t	pf_dev;
151 
152 /*
153  * XXX - These are new and need to be checked when moveing to a new version
154  */
155 static void		 pf_clear_states(void);
156 static int		 pf_clear_tables(void);
157 static void		 pf_clear_srcnodes(void);
158 /*
159  * XXX - These are new and need to be checked when moveing to a new version
160  */
161 
162 /*
163  * Wrapper functions for pfil(9) hooks
164  */
165 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
166 		int dir);
167 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
168 		int dir);
169 #ifdef INET6
170 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
171 		int dir);
172 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
173 		int dir);
174 #endif
175 
176 static int 		 hook_pf(void);
177 static int 		 dehook_pf(void);
178 static int 		 shutdown_pf(void);
179 static int 		 pf_load(void);
180 static int 		 pf_unload(void);
181 
182 d_open_t	pfopen;
183 d_close_t	pfclose;
184 d_ioctl_t	pfioctl;
185 
186 static struct dev_ops pf_ops = {	    /* XXX convert to port model */
187 	{ PF_NAME, 73, 0 },
188 	.d_open =	pfopen,
189 	.d_close =	pfclose,
190 	.d_ioctl =	pfioctl
191 };
192 
193 static volatile int pf_pfil_hooked = 0;
194 int pf_end_threads = 0;
195 struct lock pf_mod_lck;
196 
197 int debug_pfugidhack = 0;
198 SYSCTL_INT(_debug, OID_AUTO, pfugidhack, CTLFLAG_RW, &debug_pfugidhack, 0,
199 	"Enable/disable pf user/group rules mpsafe hack");
200 
201 void
202 init_zone_var(void)
203 {
204 	pf_src_tree_pl = pf_rule_pl = NULL;
205 	pf_state_pl = pf_altq_pl = pf_pooladdr_pl = NULL;
206 	pf_frent_pl = pf_frag_pl = pf_cache_pl = pf_cent_pl = NULL;
207 	pf_state_scrub_pl = NULL;
208 	pfr_ktable_pl = pfr_kentry_pl = NULL;
209 }
210 
211 void
212 cleanup_pf_zone(void)
213 {
214 	ZONE_DESTROY(pf_src_tree_pl);
215 	ZONE_DESTROY(pf_rule_pl);
216 	ZONE_DESTROY(pf_state_pl);
217 	ZONE_DESTROY(pf_altq_pl);
218 	ZONE_DESTROY(pf_pooladdr_pl);
219 	ZONE_DESTROY(pf_frent_pl);
220 	ZONE_DESTROY(pf_frag_pl);
221 	ZONE_DESTROY(pf_cache_pl);
222 	ZONE_DESTROY(pf_cent_pl);
223 	ZONE_DESTROY(pfr_ktable_pl);
224 	ZONE_DESTROY(pfr_kentry_pl);
225 	ZONE_DESTROY(pfr_kentry_pl2);
226 	ZONE_DESTROY(pf_state_scrub_pl);
227 	ZONE_DESTROY(pfi_addr_pl);
228 }
229 
230 int
231 pfattach(void)
232 {
233 	u_int32_t *my_timeout = pf_default_rule.timeout;
234 	int error = 1;
235 
236 	do {
237 		ZONE_CREATE(pf_src_tree_pl,struct pf_src_node, "pfsrctrpl");
238 		ZONE_CREATE(pf_rule_pl,    struct pf_rule, "pfrulepl");
239 		ZONE_CREATE(pf_state_pl,   struct pf_state, "pfstatepl");
240 		ZONE_CREATE(pf_state_key_pl, struct pf_state_key, "pfstatekeypl");
241 		ZONE_CREATE(pf_altq_pl,    struct pf_altq, "pfaltqpl");
242 		ZONE_CREATE(pf_pooladdr_pl,struct pf_pooladdr, "pfpooladdrpl");
243 		ZONE_CREATE(pfr_ktable_pl, struct pfr_ktable, "pfrktable");
244 		ZONE_CREATE(pfr_kentry_pl, struct pfr_kentry, "pfrkentry");
245 		ZONE_CREATE(pfr_kentry_pl2, struct pfr_kentry, "pfrkentry2");
246 		ZONE_CREATE(pf_frent_pl,   struct pf_frent, "pffrent");
247 		ZONE_CREATE(pf_frag_pl,    struct pf_fragment, "pffrag");
248 		ZONE_CREATE(pf_cache_pl,   struct pf_fragment, "pffrcache");
249 		ZONE_CREATE(pf_cent_pl,    struct pf_frcache, "pffrcent");
250 		ZONE_CREATE(pf_state_scrub_pl, struct pf_state_scrub,
251 		    "pfstatescrub");
252 		ZONE_CREATE(pfi_addr_pl,   struct pfi_dynaddr, "pfiaddrpl");
253 		error = 0;
254 	} while(0);
255 	if (error) {
256 		cleanup_pf_zone();
257 		return (error);
258 	}
259 	pfr_initialize();
260 	pfi_initialize();
261 	error = pf_osfp_initialize();
262 	if (error) {
263 		cleanup_pf_zone();
264 		pf_osfp_cleanup();
265 		return (error);
266 	}
267 
268 	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
269 	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
270 	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
271 	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
272 	/* XXX uma_zone_set_max(pf_pool_limits[PF_LIMIT_STATES].pp,
273 		pf_pool_limits[PF_LIMIT_STATES].limit);
274 	*/
275 	if (ctob(physmem) <= 100*1024*1024)
276 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
277 		    PFR_KENTRY_HIWAT_SMALL;
278 	RB_INIT(&tree_src_tracking);
279 	RB_INIT(&pf_anchors);
280 	pf_init_ruleset(&pf_main_ruleset);
281 	TAILQ_INIT(&pf_altqs[0]);
282 	TAILQ_INIT(&pf_altqs[1]);
283 	TAILQ_INIT(&pf_pabuf);
284 	pf_altqs_active = &pf_altqs[0];
285 	pf_altqs_inactive = &pf_altqs[1];
286 	TAILQ_INIT(&state_list);
287 
288 	/* default rule should never be garbage collected */
289 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
290 	pf_default_rule.action = PF_PASS;
291 	pf_default_rule.nr = (uint32_t)(-1);
292 	pf_default_rule.rtableid = -1;
293 
294 	/* initialize default timeouts */
295 	my_timeout[PFTM_TCP_FIRST_PACKET] = 120;	/* First TCP packet */
296 	my_timeout[PFTM_TCP_OPENING] = 30; 		/* No response yet */
297 	my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;	/* Established */
298 	my_timeout[PFTM_TCP_CLOSING] = 15 * 60;		/* Half closed */
299 	my_timeout[PFTM_TCP_FIN_WAIT] = 45;		/* Got both FINs */
300 	my_timeout[PFTM_TCP_CLOSED] = 90;		/* Got a RST */
301 	my_timeout[PFTM_UDP_FIRST_PACKET] = 60;		/* First UDP packet */
302 	my_timeout[PFTM_UDP_SINGLE] = 30;		/* Unidirectional */
303 	my_timeout[PFTM_UDP_MULTIPLE] = 60;		/* Bidirectional */
304 	my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;	/* First ICMP packet */
305 	my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;		/* Got error response */
306 	my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;	/* First packet */
307 	my_timeout[PFTM_OTHER_SINGLE] = 30;		/* Unidirectional */
308 	my_timeout[PFTM_OTHER_MULTIPLE] = 60;		/* Bidirectional */
309 	my_timeout[PFTM_FRAG] = 30;			/* Fragment expire */
310 	my_timeout[PFTM_INTERVAL] = 10;			/* Expire interval */
311 	my_timeout[PFTM_SRC_NODE] = 0;		/* Source Tracking */
312 	my_timeout[PFTM_TS_DIFF] = 30;		/* Allowed TS diff */
313 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
314 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
315 
316 	pf_normalize_init();
317 	bzero(&pf_status, sizeof(pf_status));
318 	pf_status.debug = PF_DEBUG_URGENT;
319 
320 	/* XXX do our best to avoid a conflict */
321 	pf_status.hostid = karc4random();
322 
323 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
324 		panic("pfpurge thread");
325 
326 	return (error);
327 }
328 
329 int
330 pfopen(struct dev_open_args *ap)
331 {
332 	lwkt_gettoken(&pf_token);
333 	cdev_t dev = ap->a_head.a_dev;
334 	if (minor(dev) >= 1) {
335 		lwkt_reltoken(&pf_token);
336 		return (ENXIO);
337 	}
338 	lwkt_reltoken(&pf_token);
339 	return (0);
340 }
341 
342 int
343 pfclose(struct dev_close_args *ap)
344 {
345 	lwkt_gettoken(&pf_token);
346 	cdev_t dev = ap->a_head.a_dev;
347 	if (minor(dev) >= 1) {
348 		lwkt_reltoken(&pf_token);
349 		return (ENXIO);
350 	}
351 	lwkt_reltoken(&pf_token);
352 	return (0);
353 }
354 
355 struct pf_pool *
356 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
357     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
358     u_int8_t check_ticket)
359 {
360 	struct pf_ruleset	*ruleset;
361 	struct pf_rule		*rule;
362 	int			 rs_num;
363 
364 	ruleset = pf_find_ruleset(anchor);
365 	if (ruleset == NULL)
366 		return (NULL);
367 	rs_num = pf_get_ruleset_number(rule_action);
368 	if (rs_num >= PF_RULESET_MAX)
369 		return (NULL);
370 	if (active) {
371 		if (check_ticket && ticket !=
372 		    ruleset->rules[rs_num].active.ticket)
373 			return (NULL);
374 		if (r_last)
375 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
376 			    pf_rulequeue);
377 		else
378 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
379 	} else {
380 		if (check_ticket && ticket !=
381 		    ruleset->rules[rs_num].inactive.ticket)
382 			return (NULL);
383 		if (r_last)
384 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
385 			    pf_rulequeue);
386 		else
387 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
388 	}
389 	if (!r_last) {
390 		while ((rule != NULL) && (rule->nr != rule_number))
391 			rule = TAILQ_NEXT(rule, entries);
392 	}
393 	if (rule == NULL)
394 		return (NULL);
395 
396 	return (&rule->rpool);
397 }
398 
399 void
400 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
401 {
402 	struct pf_pooladdr	*mv_pool_pa;
403 
404 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
405 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
406 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
407 	}
408 }
409 
410 void
411 pf_empty_pool(struct pf_palist *poola)
412 {
413 	struct pf_pooladdr	*empty_pool_pa;
414 
415 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
416 		pfi_dynaddr_remove(&empty_pool_pa->addr);
417 		pf_tbladdr_remove(&empty_pool_pa->addr);
418 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
419 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
420 		pool_put(&pf_pooladdr_pl, empty_pool_pa);
421 	}
422 }
423 
424 void
425 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
426 {
427 	if (rulequeue != NULL) {
428 		if (rule->states <= 0) {
429 			/*
430 			 * XXX - we need to remove the table *before* detaching
431 			 * the rule to make sure the table code does not delete
432 			 * the anchor under our feet.
433 			 */
434 			pf_tbladdr_remove(&rule->src.addr);
435 			pf_tbladdr_remove(&rule->dst.addr);
436 			if (rule->overload_tbl)
437 				pfr_detach_table(rule->overload_tbl);
438 		}
439 		TAILQ_REMOVE(rulequeue, rule, entries);
440 		rule->entries.tqe_prev = NULL;
441 		rule->nr = -1;
442 	}
443 
444 	if (rule->states > 0 || rule->src_nodes > 0 ||
445 	    rule->entries.tqe_prev != NULL)
446 		return;
447 	pf_tag_unref(rule->tag);
448 	pf_tag_unref(rule->match_tag);
449 #ifdef ALTQ
450 	if (rule->pqid != rule->qid)
451 		pf_qid_unref(rule->pqid);
452 	pf_qid_unref(rule->qid);
453 #endif
454 	pf_rtlabel_remove(&rule->src.addr);
455 	pf_rtlabel_remove(&rule->dst.addr);
456 	pfi_dynaddr_remove(&rule->src.addr);
457 	pfi_dynaddr_remove(&rule->dst.addr);
458 	if (rulequeue == NULL) {
459 		pf_tbladdr_remove(&rule->src.addr);
460 		pf_tbladdr_remove(&rule->dst.addr);
461 		if (rule->overload_tbl)
462 			pfr_detach_table(rule->overload_tbl);
463 	}
464 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
465 	pf_anchor_remove(rule);
466 	pf_empty_pool(&rule->rpool.list);
467 	pool_put(&pf_rule_pl, rule);
468 }
469 
470 u_int16_t
471 tagname2tag(struct pf_tags *head, char *tagname)
472 {
473 	struct pf_tagname	*tag, *p = NULL;
474 	u_int16_t		 new_tagid = 1;
475 
476 	TAILQ_FOREACH(tag, head, entries)
477 		if (strcmp(tagname, tag->name) == 0) {
478 			tag->ref++;
479 			return (tag->tag);
480 		}
481 
482 	/*
483 	 * to avoid fragmentation, we do a linear search from the beginning
484 	 * and take the first free slot we find. if there is none or the list
485 	 * is empty, append a new entry at the end.
486 	 */
487 
488 	/* new entry */
489 	if (!TAILQ_EMPTY(head))
490 		for (p = TAILQ_FIRST(head); p != NULL &&
491 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
492 			new_tagid = p->tag + 1;
493 
494 	if (new_tagid > TAGID_MAX)
495 		return (0);
496 
497 	/* allocate and fill new struct pf_tagname */
498 	tag = kmalloc(sizeof(struct pf_tagname), M_TEMP, M_WAITOK);
499 	if (tag == NULL)
500 		return (0);
501 	bzero(tag, sizeof(struct pf_tagname));
502 	strlcpy(tag->name, tagname, sizeof(tag->name));
503 	tag->tag = new_tagid;
504 	tag->ref++;
505 
506 	if (p != NULL)	/* insert new entry before p */
507 		TAILQ_INSERT_BEFORE(p, tag, entries);
508 	else	/* either list empty or no free slot in between */
509 		TAILQ_INSERT_TAIL(head, tag, entries);
510 
511 	return (tag->tag);
512 }
513 
514 void
515 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
516 {
517 	struct pf_tagname	*tag;
518 
519 	TAILQ_FOREACH(tag, head, entries)
520 		if (tag->tag == tagid) {
521 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
522 			return;
523 		}
524 }
525 
526 void
527 tag_unref(struct pf_tags *head, u_int16_t tag)
528 {
529 	struct pf_tagname	*p, *next;
530 
531 	if (tag == 0)
532 		return;
533 
534 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
535 		next = TAILQ_NEXT(p, entries);
536 		if (tag == p->tag) {
537 			if (--p->ref == 0) {
538 				TAILQ_REMOVE(head, p, entries);
539 				kfree(p, M_TEMP);
540 			}
541 			break;
542 		}
543 	}
544 }
545 
546 u_int16_t
547 pf_tagname2tag(char *tagname)
548 {
549 	return (tagname2tag(&pf_tags, tagname));
550 }
551 
552 void
553 pf_tag2tagname(u_int16_t tagid, char *p)
554 {
555 	tag2tagname(&pf_tags, tagid, p);
556 }
557 
558 void
559 pf_tag_ref(u_int16_t tag)
560 {
561 	struct pf_tagname *t;
562 
563 	TAILQ_FOREACH(t, &pf_tags, entries)
564 		if (t->tag == tag)
565 			break;
566 	if (t != NULL)
567 		t->ref++;
568 }
569 
570 void
571 pf_tag_unref(u_int16_t tag)
572 {
573 	tag_unref(&pf_tags, tag);
574 }
575 
576 int
577 pf_rtlabel_add(struct pf_addr_wrap *a)
578 {
579 	return (0);
580 }
581 
582 void
583 pf_rtlabel_remove(struct pf_addr_wrap *a)
584 {
585 }
586 
587 void
588 pf_rtlabel_copyout(struct pf_addr_wrap *a)
589 {
590 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
591 		strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
592 }
593 
594 #ifdef ALTQ
595 u_int32_t
596 pf_qname2qid(char *qname)
597 {
598 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
599 }
600 
601 void
602 pf_qid2qname(u_int32_t qid, char *p)
603 {
604 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
605 }
606 
607 void
608 pf_qid_unref(u_int32_t qid)
609 {
610 	tag_unref(&pf_qids, (u_int16_t)qid);
611 }
612 
613 int
614 pf_begin_altq(u_int32_t *ticket)
615 {
616 	struct pf_altq	*altq;
617 	int		 error = 0;
618 
619 	/* Purge the old altq list */
620 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
621 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
622 		if (altq->qname[0] == 0) {
623 			/* detach and destroy the discipline */
624 			error = altq_remove(altq);
625 		} else
626 			pf_qid_unref(altq->qid);
627 		pool_put(&pf_altq_pl, altq);
628 	}
629 	if (error)
630 		return (error);
631 	*ticket = ++ticket_altqs_inactive;
632 	altqs_inactive_open = 1;
633 	return (0);
634 }
635 
636 int
637 pf_rollback_altq(u_int32_t ticket)
638 {
639 	struct pf_altq	*altq;
640 	int		 error = 0;
641 
642 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
643 		return (0);
644 	/* Purge the old altq list */
645 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
646 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
647 		if (altq->qname[0] == 0) {
648 			/* detach and destroy the discipline */
649 			error = altq_remove(altq);
650 		} else
651 			pf_qid_unref(altq->qid);
652 		pool_put(&pf_altq_pl, altq);
653 	}
654 	altqs_inactive_open = 0;
655 	return (error);
656 }
657 
658 int
659 pf_commit_altq(u_int32_t ticket)
660 {
661 	struct pf_altqqueue	*old_altqs;
662 	struct pf_altq		*altq;
663 	int			 err, error = 0;
664 
665 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
666 		return (EBUSY);
667 
668 	/* swap altqs, keep the old. */
669 	crit_enter();
670 	old_altqs = pf_altqs_active;
671 	pf_altqs_active = pf_altqs_inactive;
672 	pf_altqs_inactive = old_altqs;
673 	ticket_altqs_active = ticket_altqs_inactive;
674 
675 	/* Attach new disciplines */
676 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
677 		if (altq->qname[0] == 0) {
678 			/* attach the discipline */
679 			error = altq_pfattach(altq);
680 			if (error) {
681 				crit_exit();
682 				return (error);
683 			}
684 		}
685 	}
686 
687 	/* Purge the old altq list */
688 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
689 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
690 		if (altq->qname[0] == 0) {
691 			/* detach and destroy the discipline */
692 			if (pf_altq_running)
693 				error = pf_disable_altq(altq);
694 			err = altq_pfdetach(altq);
695 			if (err != 0 && error == 0)
696 				error = err;
697 			err = altq_remove(altq);
698 			if (err != 0 && error == 0)
699 				error = err;
700 		} else
701 			pf_qid_unref(altq->qid);
702 		pool_put(&pf_altq_pl, altq);
703 	}
704 	crit_exit();
705 
706 	altqs_inactive_open = 0;
707 	return (error);
708 }
709 
710 int
711 pf_enable_altq(struct pf_altq *altq)
712 {
713 	struct ifnet		*ifp;
714 	struct tb_profile	 tb;
715 	int			 error = 0;
716 
717 	if ((ifp = ifunit(altq->ifname)) == NULL)
718 		return (EINVAL);
719 
720 	if (ifp->if_snd.altq_type != ALTQT_NONE)
721 		error = altq_enable(&ifp->if_snd);
722 
723 	/* set tokenbucket regulator */
724 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
725 		tb.rate = altq->ifbandwidth;
726 		tb.depth = altq->tbrsize;
727 		crit_enter();
728 		error = tbr_set(&ifp->if_snd, &tb);
729 		crit_exit();
730 	}
731 
732 	return (error);
733 }
734 
735 int
736 pf_disable_altq(struct pf_altq *altq)
737 {
738 	struct ifnet		*ifp;
739 	struct tb_profile	 tb;
740 	int			 error;
741 
742 	if ((ifp = ifunit(altq->ifname)) == NULL)
743 		return (EINVAL);
744 
745 	/*
746 	 * when the discipline is no longer referenced, it was overridden
747 	 * by a new one.  if so, just return.
748 	 */
749 	if (altq->altq_disc != ifp->if_snd.altq_disc)
750 		return (0);
751 
752 	error = altq_disable(&ifp->if_snd);
753 
754 	if (error == 0) {
755 		/* clear tokenbucket regulator */
756 		tb.rate = 0;
757 		crit_enter();
758 		error = tbr_set(&ifp->if_snd, &tb);
759 		crit_exit();
760 	}
761 
762 	return (error);
763 }
764 #endif /* ALTQ */
765 
766 int
767 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
768 {
769 	struct pf_ruleset	*rs;
770 	struct pf_rule		*rule;
771 
772 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
773 		return (EINVAL);
774 	rs = pf_find_or_create_ruleset(anchor);
775 	if (rs == NULL)
776 		return (EINVAL);
777 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
778 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
779 		rs->rules[rs_num].inactive.rcount--;
780 	}
781 	*ticket = ++rs->rules[rs_num].inactive.ticket;
782 	rs->rules[rs_num].inactive.open = 1;
783 	return (0);
784 }
785 
786 int
787 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
788 {
789 	struct pf_ruleset	*rs;
790 	struct pf_rule		*rule;
791 
792 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
793 		return (EINVAL);
794 	rs = pf_find_ruleset(anchor);
795 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
796 	    rs->rules[rs_num].inactive.ticket != ticket)
797 		return (0);
798 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
799 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
800 		rs->rules[rs_num].inactive.rcount--;
801 	}
802 	rs->rules[rs_num].inactive.open = 0;
803 	return (0);
804 }
805 
806 #define PF_MD5_UPD(st, elm)						\
807 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
808 
809 #define PF_MD5_UPD_STR(st, elm)						\
810 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
811 
812 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
813 		(stor) = htonl((st)->elm);				\
814 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
815 } while (0)
816 
817 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
818 		(stor) = htons((st)->elm);				\
819 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
820 } while (0)
821 
822 void
823 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
824 {
825 	PF_MD5_UPD(pfr, addr.type);
826 	switch (pfr->addr.type) {
827 		case PF_ADDR_DYNIFTL:
828 			PF_MD5_UPD(pfr, addr.v.ifname);
829 			PF_MD5_UPD(pfr, addr.iflags);
830 			break;
831 		case PF_ADDR_TABLE:
832 			PF_MD5_UPD(pfr, addr.v.tblname);
833 			break;
834 		case PF_ADDR_ADDRMASK:
835 			/* XXX ignore af? */
836 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
837 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
838 			break;
839 		case PF_ADDR_RTLABEL:
840 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
841 			break;
842 	}
843 
844 	PF_MD5_UPD(pfr, port[0]);
845 	PF_MD5_UPD(pfr, port[1]);
846 	PF_MD5_UPD(pfr, neg);
847 	PF_MD5_UPD(pfr, port_op);
848 }
849 
850 void
851 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
852 {
853 	u_int16_t x;
854 	u_int32_t y;
855 
856 	pf_hash_rule_addr(ctx, &rule->src);
857 	pf_hash_rule_addr(ctx, &rule->dst);
858 	PF_MD5_UPD_STR(rule, label);
859 	PF_MD5_UPD_STR(rule, ifname);
860 	PF_MD5_UPD_STR(rule, match_tagname);
861 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
862 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
863 	PF_MD5_UPD_HTONL(rule, prob, y);
864 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
865 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
866 	PF_MD5_UPD(rule, uid.op);
867 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
868 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
869 	PF_MD5_UPD(rule, gid.op);
870 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
871 	PF_MD5_UPD(rule, action);
872 	PF_MD5_UPD(rule, direction);
873 	PF_MD5_UPD(rule, af);
874 	PF_MD5_UPD(rule, quick);
875 	PF_MD5_UPD(rule, ifnot);
876 	PF_MD5_UPD(rule, match_tag_not);
877 	PF_MD5_UPD(rule, natpass);
878 	PF_MD5_UPD(rule, keep_state);
879 	PF_MD5_UPD(rule, proto);
880 	PF_MD5_UPD(rule, type);
881 	PF_MD5_UPD(rule, code);
882 	PF_MD5_UPD(rule, flags);
883 	PF_MD5_UPD(rule, flagset);
884 	PF_MD5_UPD(rule, allow_opts);
885 	PF_MD5_UPD(rule, rt);
886 	PF_MD5_UPD(rule, tos);
887 }
888 
889 int
890 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
891 {
892 	struct pf_ruleset	*rs;
893 	struct pf_rule		*rule, **old_array;
894 	struct pf_rulequeue	*old_rules;
895 	int			 error;
896 	u_int32_t		 old_rcount;
897 
898 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
899 		return (EINVAL);
900 	rs = pf_find_ruleset(anchor);
901 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
902 	    ticket != rs->rules[rs_num].inactive.ticket)
903 		return (EBUSY);
904 
905 	/* Calculate checksum for the main ruleset */
906 	if (rs == &pf_main_ruleset) {
907 		error = pf_setup_pfsync_matching(rs);
908 		if (error != 0)
909 			return (error);
910 	}
911 
912 	/* Swap rules, keep the old. */
913 	crit_enter();
914 	old_rules = rs->rules[rs_num].active.ptr;
915 	old_rcount = rs->rules[rs_num].active.rcount;
916 	old_array = rs->rules[rs_num].active.ptr_array;
917 
918 	rs->rules[rs_num].active.ptr =
919 	    rs->rules[rs_num].inactive.ptr;
920 	rs->rules[rs_num].active.ptr_array =
921 	    rs->rules[rs_num].inactive.ptr_array;
922 	rs->rules[rs_num].active.rcount =
923 	    rs->rules[rs_num].inactive.rcount;
924 	rs->rules[rs_num].inactive.ptr = old_rules;
925 	rs->rules[rs_num].inactive.ptr_array = old_array;
926 	rs->rules[rs_num].inactive.rcount = old_rcount;
927 
928 	rs->rules[rs_num].active.ticket =
929 	    rs->rules[rs_num].inactive.ticket;
930 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
931 
932 
933 	/* Purge the old rule list. */
934 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
935 		pf_rm_rule(old_rules, rule);
936 	if (rs->rules[rs_num].inactive.ptr_array)
937 		kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
938 	rs->rules[rs_num].inactive.ptr_array = NULL;
939 	rs->rules[rs_num].inactive.rcount = 0;
940 	rs->rules[rs_num].inactive.open = 0;
941 	pf_remove_if_empty_ruleset(rs);
942 	crit_exit();
943 	return (0);
944 }
945 
946 void
947 pf_state_export(struct pfsync_state *sp, struct pf_state_key *sk,
948    struct pf_state *s)
949 {
950 	int secs = time_second;
951 	bzero(sp, sizeof(struct pfsync_state));
952 
953 	/* copy from state key */
954 	sp->lan.addr = sk->lan.addr;
955 	sp->lan.port = sk->lan.port;
956 	sp->gwy.addr = sk->gwy.addr;
957 	sp->gwy.port = sk->gwy.port;
958 	sp->ext.addr = sk->ext.addr;
959 	sp->ext.port = sk->ext.port;
960 	sp->proto = sk->proto;
961 	sp->af = sk->af;
962 	sp->direction = sk->direction;
963 
964 	/* copy from state */
965 	memcpy(&sp->id, &s->id, sizeof(sp->id));
966 	sp->creatorid = s->creatorid;
967 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
968 	pf_state_peer_to_pfsync(&s->src, &sp->src);
969 	pf_state_peer_to_pfsync(&s->dst, &sp->dst);
970 
971 	sp->rule = s->rule.ptr->nr;
972 	sp->nat_rule = (s->nat_rule.ptr == NULL) ?  -1 : s->nat_rule.ptr->nr;
973 	sp->anchor = (s->anchor.ptr == NULL) ?  -1 : s->anchor.ptr->nr;
974 
975 	pf_state_counter_to_pfsync(s->bytes[0], sp->bytes[0]);
976 	pf_state_counter_to_pfsync(s->bytes[1], sp->bytes[1]);
977 	pf_state_counter_to_pfsync(s->packets[0], sp->packets[0]);
978 	pf_state_counter_to_pfsync(s->packets[1], sp->packets[1]);
979 	sp->creation = secs - s->creation;
980 	sp->expire = pf_state_expires(s);
981 	sp->log = s->log;
982 	sp->allow_opts = s->allow_opts;
983 	sp->timeout = s->timeout;
984 
985 	if (s->src_node)
986 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
987 	if (s->nat_src_node)
988 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
989 
990 	if (sp->expire > secs)
991 		sp->expire -= secs;
992 	else
993 		sp->expire = 0;
994 
995 }
996 
997 void
998 pf_state_import(struct pfsync_state *sp, struct pf_state_key *sk,
999    struct pf_state *s)
1000 {
1001 	/* copy to state key */
1002 	sk->lan.addr = sp->lan.addr;
1003 	sk->lan.port = sp->lan.port;
1004 	sk->gwy.addr = sp->gwy.addr;
1005 	sk->gwy.port = sp->gwy.port;
1006 	sk->ext.addr = sp->ext.addr;
1007 	sk->ext.port = sp->ext.port;
1008 	sk->proto = sp->proto;
1009 	sk->af = sp->af;
1010 	sk->direction = sp->direction;
1011 
1012 	/* copy to state */
1013 	memcpy(&s->id, &sp->id, sizeof(sp->id));
1014 	s->creatorid = sp->creatorid;
1015 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
1016 	pf_state_peer_from_pfsync(&sp->src, &s->src);
1017 	pf_state_peer_from_pfsync(&sp->dst, &s->dst);
1018 
1019 	s->rule.ptr = &pf_default_rule;
1020 	s->nat_rule.ptr = NULL;
1021 	s->anchor.ptr = NULL;
1022 	s->rt_kif = NULL;
1023 	s->creation = time_second;
1024 	s->pfsync_time = 0;
1025 	s->packets[0] = s->packets[1] = 0;
1026 	s->bytes[0] = s->bytes[1] = 0;
1027 }
1028 
1029 int
1030 pf_setup_pfsync_matching(struct pf_ruleset *rs)
1031 {
1032 	MD5_CTX			 ctx;
1033 	struct pf_rule		*rule;
1034 	int			 rs_cnt;
1035 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1036 
1037 	MD5Init(&ctx);
1038 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1039 		/* XXX PF_RULESET_SCRUB as well? */
1040 		if (rs_cnt == PF_RULESET_SCRUB)
1041 			continue;
1042 
1043 		if (rs->rules[rs_cnt].inactive.ptr_array)
1044 			kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1045 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1046 
1047 		if (rs->rules[rs_cnt].inactive.rcount) {
1048 			rs->rules[rs_cnt].inactive.ptr_array =
1049 			    kmalloc(sizeof(caddr_t) *
1050 				    rs->rules[rs_cnt].inactive.rcount,
1051 				    M_TEMP, M_WAITOK);
1052 
1053 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1054 				return (ENOMEM);
1055 		}
1056 
1057 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1058 		    entries) {
1059 			pf_hash_rule(&ctx, rule);
1060 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1061 		}
1062 	}
1063 
1064 	MD5Final(digest, &ctx);
1065 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
1066 	return (0);
1067 }
1068 
1069 int
1070 pfioctl(struct dev_ioctl_args *ap)
1071 {
1072 	u_long cmd = ap->a_cmd;
1073 	caddr_t addr = ap->a_data;
1074 	struct pf_pooladdr	*pa = NULL;
1075 	struct pf_pool		*pool = NULL;
1076 	int			 error = 0;
1077 
1078 	lwkt_gettoken(&pf_token);
1079 
1080 	/* XXX keep in sync with switch() below */
1081 	if (securelevel > 1)
1082 		switch (cmd) {
1083 		case DIOCGETRULES:
1084 		case DIOCGETRULE:
1085 		case DIOCGETADDRS:
1086 		case DIOCGETADDR:
1087 		case DIOCGETSTATE:
1088 		case DIOCSETSTATUSIF:
1089 		case DIOCGETSTATUS:
1090 		case DIOCCLRSTATUS:
1091 		case DIOCNATLOOK:
1092 		case DIOCSETDEBUG:
1093 		case DIOCGETSTATES:
1094 		case DIOCGETTIMEOUT:
1095 		case DIOCCLRRULECTRS:
1096 		case DIOCGETLIMIT:
1097 		case DIOCGETALTQS:
1098 		case DIOCGETALTQ:
1099 		case DIOCGETQSTATS:
1100 		case DIOCGETRULESETS:
1101 		case DIOCGETRULESET:
1102 		case DIOCRGETTABLES:
1103 		case DIOCRGETTSTATS:
1104 		case DIOCRCLRTSTATS:
1105 		case DIOCRCLRADDRS:
1106 		case DIOCRADDADDRS:
1107 		case DIOCRDELADDRS:
1108 		case DIOCRSETADDRS:
1109 		case DIOCRGETADDRS:
1110 		case DIOCRGETASTATS:
1111 		case DIOCRCLRASTATS:
1112 		case DIOCRTSTADDRS:
1113 		case DIOCOSFPGET:
1114 		case DIOCGETSRCNODES:
1115 		case DIOCCLRSRCNODES:
1116 		case DIOCIGETIFACES:
1117 		case DIOCSETIFFLAG:
1118 		case DIOCCLRIFFLAG:
1119 		case DIOCGIFSPEED:
1120 			break;
1121 		case DIOCRCLRTABLES:
1122 		case DIOCRADDTABLES:
1123 		case DIOCRDELTABLES:
1124 		case DIOCRSETTFLAGS:
1125 			if (((struct pfioc_table *)addr)->pfrio_flags &
1126 			    PFR_FLAG_DUMMY)
1127 				break; /* dummy operation ok */
1128 			lwkt_reltoken(&pf_token);
1129 			return (EPERM);
1130 		default:
1131 			lwkt_reltoken(&pf_token);
1132 			return (EPERM);
1133 		}
1134 
1135 	if (!(ap->a_fflag & FWRITE))
1136 		switch (cmd) {
1137 		case DIOCGETRULES:
1138 		case DIOCGETADDRS:
1139 		case DIOCGETADDR:
1140 		case DIOCGETSTATE:
1141 		case DIOCGETSTATUS:
1142 		case DIOCGETSTATES:
1143 		case DIOCGETTIMEOUT:
1144 		case DIOCGETLIMIT:
1145 		case DIOCGETALTQS:
1146 		case DIOCGETALTQ:
1147 		case DIOCGETQSTATS:
1148 		case DIOCGETRULESETS:
1149 		case DIOCGETRULESET:
1150 		case DIOCNATLOOK:
1151 		case DIOCRGETTABLES:
1152 		case DIOCRGETTSTATS:
1153 		case DIOCRGETADDRS:
1154 		case DIOCRGETASTATS:
1155 		case DIOCRTSTADDRS:
1156 		case DIOCOSFPGET:
1157 		case DIOCGETSRCNODES:
1158 		case DIOCIGETIFACES:
1159 		case DIOCGIFSPEED:
1160 			break;
1161 		case DIOCRCLRTABLES:
1162 		case DIOCRADDTABLES:
1163 		case DIOCRDELTABLES:
1164 		case DIOCRCLRTSTATS:
1165 		case DIOCRCLRADDRS:
1166 		case DIOCRADDADDRS:
1167 		case DIOCRDELADDRS:
1168 		case DIOCRSETADDRS:
1169 		case DIOCRSETTFLAGS:
1170 			if (((struct pfioc_table *)addr)->pfrio_flags &
1171 			    PFR_FLAG_DUMMY)
1172 				break; /* dummy operation ok */
1173 			lwkt_reltoken(&pf_token);
1174 			return (EACCES);
1175 		case DIOCGETRULE:
1176 			if (((struct pfioc_rule *)addr)->action == PF_GET_CLR_CNTR) {
1177 				lwkt_reltoken(&pf_token);
1178 				return (EACCES);
1179 			}
1180 			break;
1181 		default:
1182 			lwkt_reltoken(&pf_token);
1183 			return (EACCES);
1184 		}
1185 
1186 	switch (cmd) {
1187 
1188 	case DIOCSTART:
1189 		if (pf_status.running)
1190 			error = EEXIST;
1191 		else {
1192 			error = hook_pf();
1193 			if (error) {
1194 				DPFPRINTF(PF_DEBUG_MISC,
1195 				    ("pf: pfil registration fail\n"));
1196 				break;
1197 			}
1198 			pf_status.running = 1;
1199 			pf_status.since = time_second;
1200 			if (pf_status.stateid == 0) {
1201 				pf_status.stateid = time_second;
1202 				pf_status.stateid = pf_status.stateid << 32;
1203 			}
1204 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1205 		}
1206 		break;
1207 
1208 	case DIOCSTOP:
1209 		if (!pf_status.running)
1210 			error = ENOENT;
1211 		else {
1212 			pf_status.running = 0;
1213 			error = dehook_pf();
1214 			if (error) {
1215 				pf_status.running = 1;
1216 				DPFPRINTF(PF_DEBUG_MISC,
1217 					("pf: pfil unregistration failed\n"));
1218 			}
1219 			pf_status.since = time_second;
1220 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1221 		}
1222 		break;
1223 
1224 	case DIOCADDRULE: {
1225 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1226 		struct pf_ruleset	*ruleset;
1227 		struct pf_rule		*rule, *tail;
1228 		struct pf_pooladdr	*pa;
1229 		int			 rs_num;
1230 
1231 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1232 		ruleset = pf_find_ruleset(pr->anchor);
1233 		if (ruleset == NULL) {
1234 			error = EINVAL;
1235 			break;
1236 		}
1237 		rs_num = pf_get_ruleset_number(pr->rule.action);
1238 		if (rs_num >= PF_RULESET_MAX) {
1239 			error = EINVAL;
1240 			break;
1241 		}
1242 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1243 			error = EINVAL;
1244 			break;
1245 		}
1246 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1247 			error = EBUSY;
1248 			break;
1249 		}
1250 		if (pr->pool_ticket != ticket_pabuf) {
1251 			error = EBUSY;
1252 			break;
1253 		}
1254 		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1255 		if (rule == NULL) {
1256 			error = ENOMEM;
1257 			break;
1258 		}
1259 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1260 		rule->cuid = ap->a_cred->cr_ruid;
1261 		rule->cpid = (int)NULL;
1262 		rule->anchor = NULL;
1263 		rule->kif = NULL;
1264 		TAILQ_INIT(&rule->rpool.list);
1265 		/* initialize refcounting */
1266 		rule->states = 0;
1267 		rule->src_nodes = 0;
1268 		rule->entries.tqe_prev = NULL;
1269 #ifndef INET
1270 		if (rule->af == AF_INET) {
1271 			pool_put(&pf_rule_pl, rule);
1272 			error = EAFNOSUPPORT;
1273 			break;
1274 		}
1275 #endif /* INET */
1276 #ifndef INET6
1277 		if (rule->af == AF_INET6) {
1278 			pool_put(&pf_rule_pl, rule);
1279 			error = EAFNOSUPPORT;
1280 			break;
1281 		}
1282 #endif /* INET6 */
1283 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1284 		    pf_rulequeue);
1285 		if (tail)
1286 			rule->nr = tail->nr + 1;
1287 		else
1288 			rule->nr = 0;
1289 		if (rule->ifname[0]) {
1290 			rule->kif = pfi_kif_get(rule->ifname);
1291 			if (rule->kif == NULL) {
1292 				pool_put(&pf_rule_pl, rule);
1293 				error = EINVAL;
1294 				break;
1295 			}
1296 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1297 		}
1298 
1299 		if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1300 			error = EBUSY;
1301 
1302 #ifdef ALTQ
1303 		/* set queue IDs */
1304 		if (rule->qname[0] != 0) {
1305 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1306 				error = EBUSY;
1307 			else if (rule->pqname[0] != 0) {
1308 				if ((rule->pqid =
1309 				    pf_qname2qid(rule->pqname)) == 0)
1310 					error = EBUSY;
1311 			} else
1312 				rule->pqid = rule->qid;
1313 		}
1314 #endif
1315 		if (rule->tagname[0])
1316 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1317 				error = EBUSY;
1318 		if (rule->match_tagname[0])
1319 			if ((rule->match_tag =
1320 			    pf_tagname2tag(rule->match_tagname)) == 0)
1321 				error = EBUSY;
1322 		if (rule->rt && !rule->direction)
1323 			error = EINVAL;
1324 #if NPFLOG > 0
1325 		if (!rule->log)
1326 			rule->logif = 0;
1327 		if (rule->logif >= PFLOGIFS_MAX)
1328 			error = EINVAL;
1329 #endif
1330 		if (pf_rtlabel_add(&rule->src.addr) ||
1331 		    pf_rtlabel_add(&rule->dst.addr))
1332 			error = EBUSY;
1333 		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1334 			error = EINVAL;
1335 		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1336 			error = EINVAL;
1337 		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1338 			error = EINVAL;
1339 		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1340 			error = EINVAL;
1341 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1342 			error = EINVAL;
1343 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1344 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1345 				error = EINVAL;
1346 
1347 		if (rule->overload_tblname[0]) {
1348 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1349 			    rule->overload_tblname)) == NULL)
1350 				error = EINVAL;
1351 			else
1352 				rule->overload_tbl->pfrkt_flags |=
1353 				    PFR_TFLAG_ACTIVE;
1354 		}
1355 
1356 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1357 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1358 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1359 		    (rule->rt > PF_FASTROUTE)) &&
1360 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1361 			error = EINVAL;
1362 
1363 		if (error) {
1364 			pf_rm_rule(NULL, rule);
1365 			break;
1366 		}
1367 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1368 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1369 		    rule->bytes[0] = rule->bytes[1] = 0;
1370 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1371 		    rule, entries);
1372 		ruleset->rules[rs_num].inactive.rcount++;
1373 		break;
1374 	}
1375 
1376 	case DIOCGETRULES: {
1377 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1378 		struct pf_ruleset	*ruleset;
1379 		struct pf_rule		*tail;
1380 		int			 rs_num;
1381 
1382 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1383 		ruleset = pf_find_ruleset(pr->anchor);
1384 		if (ruleset == NULL) {
1385 			error = EINVAL;
1386 			break;
1387 		}
1388 		rs_num = pf_get_ruleset_number(pr->rule.action);
1389 		if (rs_num >= PF_RULESET_MAX) {
1390 			error = EINVAL;
1391 			break;
1392 		}
1393 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1394 		    pf_rulequeue);
1395 		if (tail)
1396 			pr->nr = tail->nr + 1;
1397 		else
1398 			pr->nr = 0;
1399 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1400 		break;
1401 	}
1402 
1403 	case DIOCGETRULE: {
1404 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1405 		struct pf_ruleset	*ruleset;
1406 		struct pf_rule		*rule;
1407 		int			 rs_num, i;
1408 
1409 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1410 		ruleset = pf_find_ruleset(pr->anchor);
1411 		if (ruleset == NULL) {
1412 			error = EINVAL;
1413 			break;
1414 		}
1415 		rs_num = pf_get_ruleset_number(pr->rule.action);
1416 		if (rs_num >= PF_RULESET_MAX) {
1417 			error = EINVAL;
1418 			break;
1419 		}
1420 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1421 			error = EBUSY;
1422 			break;
1423 		}
1424 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1425 		while ((rule != NULL) && (rule->nr != pr->nr))
1426 			rule = TAILQ_NEXT(rule, entries);
1427 		if (rule == NULL) {
1428 			error = EBUSY;
1429 			break;
1430 		}
1431 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1432 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1433 			error = EBUSY;
1434 			break;
1435 		}
1436 		pfi_dynaddr_copyout(&pr->rule.src.addr);
1437 		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1438 		pf_tbladdr_copyout(&pr->rule.src.addr);
1439 		pf_tbladdr_copyout(&pr->rule.dst.addr);
1440 		pf_rtlabel_copyout(&pr->rule.src.addr);
1441 		pf_rtlabel_copyout(&pr->rule.dst.addr);
1442 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1443 			if (rule->skip[i].ptr == NULL)
1444 				pr->rule.skip[i].nr = (uint32_t)(-1);
1445 			else
1446 				pr->rule.skip[i].nr =
1447 				    rule->skip[i].ptr->nr;
1448 
1449 		if (pr->action == PF_GET_CLR_CNTR) {
1450 			rule->evaluations = 0;
1451 			rule->packets[0] = rule->packets[1] = 0;
1452 			rule->bytes[0] = rule->bytes[1] = 0;
1453 		}
1454 		break;
1455 	}
1456 
1457 	case DIOCCHANGERULE: {
1458 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1459 		struct pf_ruleset	*ruleset;
1460 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1461 		u_int32_t		 nr = 0;
1462 		int			 rs_num;
1463 
1464 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1465 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1466 		    pcr->pool_ticket != ticket_pabuf) {
1467 			error = EBUSY;
1468 			break;
1469 		}
1470 
1471 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1472 		    pcr->action > PF_CHANGE_GET_TICKET) {
1473 			error = EINVAL;
1474 			break;
1475 		}
1476 		ruleset = pf_find_ruleset(pcr->anchor);
1477 		if (ruleset == NULL) {
1478 			error = EINVAL;
1479 			break;
1480 		}
1481 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1482 		if (rs_num >= PF_RULESET_MAX) {
1483 			error = EINVAL;
1484 			break;
1485 		}
1486 
1487 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1488 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1489 			break;
1490 		} else {
1491 			if (pcr->ticket !=
1492 			    ruleset->rules[rs_num].active.ticket) {
1493 				error = EINVAL;
1494 				break;
1495 			}
1496 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1497 				error = EINVAL;
1498 				break;
1499 			}
1500 		}
1501 
1502 		if (pcr->action != PF_CHANGE_REMOVE) {
1503 			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1504 			if (newrule == NULL) {
1505 				error = ENOMEM;
1506 				break;
1507 			}
1508 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1509 			newrule->cuid = ap->a_cred->cr_ruid;
1510 			newrule->cpid = (int)NULL;
1511 			TAILQ_INIT(&newrule->rpool.list);
1512 			/* initialize refcounting */
1513 			newrule->states = 0;
1514 			newrule->entries.tqe_prev = NULL;
1515 #ifndef INET
1516 			if (newrule->af == AF_INET) {
1517 				pool_put(&pf_rule_pl, newrule);
1518 				error = EAFNOSUPPORT;
1519 				break;
1520 			}
1521 #endif /* INET */
1522 #ifndef INET6
1523 			if (newrule->af == AF_INET6) {
1524 				pool_put(&pf_rule_pl, newrule);
1525 				error = EAFNOSUPPORT;
1526 				break;
1527 			}
1528 #endif /* INET6 */
1529 			if (newrule->ifname[0]) {
1530 				newrule->kif = pfi_kif_get(newrule->ifname);
1531 				if (newrule->kif == NULL) {
1532 					pool_put(&pf_rule_pl, newrule);
1533 					error = EINVAL;
1534 					break;
1535 				}
1536 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1537 			} else
1538 				newrule->kif = NULL;
1539 
1540 			if (newrule->rtableid > 0 &&
1541 			    newrule->rtableid > rt_numfibs)
1542 				error = EBUSY;
1543 
1544 #ifdef ALTQ
1545 			/* set queue IDs */
1546 			if (newrule->qname[0] != 0) {
1547 				if ((newrule->qid =
1548 				    pf_qname2qid(newrule->qname)) == 0)
1549 					error = EBUSY;
1550 				else if (newrule->pqname[0] != 0) {
1551 					if ((newrule->pqid =
1552 					    pf_qname2qid(newrule->pqname)) == 0)
1553 						error = EBUSY;
1554 				} else
1555 					newrule->pqid = newrule->qid;
1556 			}
1557 #endif /* ALTQ */
1558 			if (newrule->tagname[0])
1559 				if ((newrule->tag =
1560 				    pf_tagname2tag(newrule->tagname)) == 0)
1561 					error = EBUSY;
1562 			if (newrule->match_tagname[0])
1563 				if ((newrule->match_tag = pf_tagname2tag(
1564 				    newrule->match_tagname)) == 0)
1565 					error = EBUSY;
1566 			if (newrule->rt && !newrule->direction)
1567 				error = EINVAL;
1568 #if NPFLOG > 0
1569 			if (!newrule->log)
1570 				newrule->logif = 0;
1571 			if (newrule->logif >= PFLOGIFS_MAX)
1572 				error = EINVAL;
1573 #endif
1574 			if (pf_rtlabel_add(&newrule->src.addr) ||
1575 			    pf_rtlabel_add(&newrule->dst.addr))
1576 				error = EBUSY;
1577 			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1578 				error = EINVAL;
1579 			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1580 				error = EINVAL;
1581 			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1582 				error = EINVAL;
1583 			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1584 				error = EINVAL;
1585 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1586 				error = EINVAL;
1587 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1588 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1589 					error = EINVAL;
1590 
1591 			if (newrule->overload_tblname[0]) {
1592 				if ((newrule->overload_tbl = pfr_attach_table(
1593 				    ruleset, newrule->overload_tblname)) ==
1594 				    NULL)
1595 					error = EINVAL;
1596 				else
1597 					newrule->overload_tbl->pfrkt_flags |=
1598 					    PFR_TFLAG_ACTIVE;
1599 			}
1600 
1601 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1602 			if (((((newrule->action == PF_NAT) ||
1603 			    (newrule->action == PF_RDR) ||
1604 			    (newrule->action == PF_BINAT) ||
1605 			    (newrule->rt > PF_FASTROUTE)) &&
1606 			    !newrule->anchor)) &&
1607 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1608 				error = EINVAL;
1609 
1610 			if (error) {
1611 				pf_rm_rule(NULL, newrule);
1612 				break;
1613 			}
1614 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1615 			newrule->evaluations = 0;
1616 			newrule->packets[0] = newrule->packets[1] = 0;
1617 			newrule->bytes[0] = newrule->bytes[1] = 0;
1618 		}
1619 		pf_empty_pool(&pf_pabuf);
1620 
1621 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1622 			oldrule = TAILQ_FIRST(
1623 			    ruleset->rules[rs_num].active.ptr);
1624 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1625 			oldrule = TAILQ_LAST(
1626 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1627 		else {
1628 			oldrule = TAILQ_FIRST(
1629 			    ruleset->rules[rs_num].active.ptr);
1630 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1631 				oldrule = TAILQ_NEXT(oldrule, entries);
1632 			if (oldrule == NULL) {
1633 				if (newrule != NULL)
1634 					pf_rm_rule(NULL, newrule);
1635 				error = EINVAL;
1636 				break;
1637 			}
1638 		}
1639 
1640 		if (pcr->action == PF_CHANGE_REMOVE) {
1641 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1642 			ruleset->rules[rs_num].active.rcount--;
1643 		} else {
1644 			if (oldrule == NULL)
1645 				TAILQ_INSERT_TAIL(
1646 				    ruleset->rules[rs_num].active.ptr,
1647 				    newrule, entries);
1648 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1649 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1650 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1651 			else
1652 				TAILQ_INSERT_AFTER(
1653 				    ruleset->rules[rs_num].active.ptr,
1654 				    oldrule, newrule, entries);
1655 			ruleset->rules[rs_num].active.rcount++;
1656 		}
1657 
1658 		nr = 0;
1659 		TAILQ_FOREACH(oldrule,
1660 		    ruleset->rules[rs_num].active.ptr, entries)
1661 			oldrule->nr = nr++;
1662 
1663 		ruleset->rules[rs_num].active.ticket++;
1664 
1665 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1666 		pf_remove_if_empty_ruleset(ruleset);
1667 
1668 		break;
1669 	}
1670 
1671 	case DIOCCLRSTATES: {
1672 		struct pf_state		*s, *nexts;
1673 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1674 		int			 killed = 0;
1675 
1676 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
1677 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1678 
1679 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1680 			    s->kif->pfik_name)) {
1681 #if NPFSYNC
1682 				/* don't send out individual delete messages */
1683 				s->sync_flags = PFSTATE_NOSYNC;
1684 #endif
1685 				pf_unlink_state(s);
1686 				killed++;
1687 			}
1688 		}
1689 		psk->psk_af = killed;
1690 #if NPFSYNC
1691 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1692 #endif
1693 		break;
1694 	}
1695 
1696 	case DIOCKILLSTATES: {
1697 		struct pf_state		*s, *nexts;
1698 		struct pf_state_key	*sk;
1699 		struct pf_state_host	*src, *dst;
1700 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1701 		int			 killed = 0;
1702 
1703 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s;
1704 		    s = nexts) {
1705 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1706 			sk = s->state_key;
1707 
1708 			if (sk->direction == PF_OUT) {
1709 				src = &sk->lan;
1710 				dst = &sk->ext;
1711 			} else {
1712 				src = &sk->ext;
1713 				dst = &sk->lan;
1714 			}
1715 			if ((!psk->psk_af || sk->af == psk->psk_af)
1716 			    && (!psk->psk_proto || psk->psk_proto ==
1717 			    sk->proto) &&
1718 			    PF_MATCHA(psk->psk_src.neg,
1719 			    &psk->psk_src.addr.v.a.addr,
1720 			    &psk->psk_src.addr.v.a.mask,
1721 			    &src->addr, sk->af) &&
1722 			    PF_MATCHA(psk->psk_dst.neg,
1723 			    &psk->psk_dst.addr.v.a.addr,
1724 			    &psk->psk_dst.addr.v.a.mask,
1725 			    &dst->addr, sk->af) &&
1726 			    (psk->psk_src.port_op == 0 ||
1727 			    pf_match_port(psk->psk_src.port_op,
1728 			    psk->psk_src.port[0], psk->psk_src.port[1],
1729 			    src->port)) &&
1730 			    (psk->psk_dst.port_op == 0 ||
1731 			    pf_match_port(psk->psk_dst.port_op,
1732 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1733 			    dst->port)) &&
1734 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1735 			    s->kif->pfik_name))) {
1736 #if NPFSYNC > 0
1737 				/* send immediate delete of state */
1738 				pfsync_delete_state(s);
1739 				s->sync_flags |= PFSTATE_NOSYNC;
1740 #endif
1741 				pf_unlink_state(s);
1742 				killed++;
1743 			}
1744 		}
1745 		psk->psk_af = killed;
1746 		break;
1747 	}
1748 
1749 	case DIOCADDSTATE: {
1750 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1751 		struct pfsync_state 	*sp = (struct pfsync_state *)ps->state;
1752 		struct pf_state		*s;
1753 		struct pf_state_key	*sk;
1754 		struct pfi_kif		*kif;
1755 
1756 		if (sp->timeout >= PFTM_MAX &&
1757 		    sp->timeout != PFTM_UNTIL_PACKET) {
1758 			error = EINVAL;
1759 			break;
1760 		}
1761 		s = pool_get(&pf_state_pl, PR_NOWAIT);
1762 		if (s == NULL) {
1763 			error = ENOMEM;
1764 			break;
1765 		}
1766 		bzero(s, sizeof(struct pf_state));
1767 		if ((sk = pf_alloc_state_key(s)) == NULL) {
1768 			error = ENOMEM;
1769 			break;
1770 		}
1771 		pf_state_import(sp, sk, s);
1772 		kif = pfi_kif_get(sp->ifname);
1773 		if (kif == NULL) {
1774 			pool_put(&pf_state_pl, s);
1775 			pool_put(&pf_state_key_pl, sk);
1776 			error = ENOENT;
1777 			break;
1778 		}
1779 		if (pf_insert_state(kif, s)) {
1780 			pfi_kif_unref(kif, PFI_KIF_REF_NONE);
1781 			pool_put(&pf_state_pl, s);
1782 			pool_put(&pf_state_key_pl, sk);
1783 			error = ENOMEM;
1784 		}
1785 		break;
1786 	}
1787 
1788 	case DIOCGETSTATE: {
1789 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1790 		struct pf_state		*s;
1791 		u_int32_t		 nr;
1792 
1793 		nr = 0;
1794 		RB_FOREACH(s, pf_state_tree_id, &tree_id) {
1795 			if (nr >= ps->nr)
1796 				break;
1797 			nr++;
1798 		}
1799 		if (s == NULL) {
1800 			error = EBUSY;
1801 			break;
1802 		}
1803 
1804 		pf_state_export((struct pfsync_state *)&ps->state,
1805 		    s->state_key, s);
1806 		break;
1807 	}
1808 
1809 	case DIOCGETSTATES: {
1810 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1811 		struct pf_state		*state;
1812 		struct pfsync_state	*p, *pstore;
1813 		u_int32_t		 nr = 0;
1814 
1815 		if (ps->ps_len == 0) {
1816 			nr = pf_status.states;
1817 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1818 			break;
1819 		}
1820 
1821 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1822 
1823 		p = ps->ps_states;
1824 
1825 		state = TAILQ_FIRST(&state_list);
1826 		while (state) {
1827 			if (state->timeout != PFTM_UNLINKED) {
1828 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1829 					break;
1830 
1831 				pf_state_export(pstore,
1832 				    state->state_key, state);
1833 				error = copyout(pstore, p, sizeof(*p));
1834 				if (error) {
1835 					kfree(pstore, M_TEMP);
1836 					goto fail;
1837 				}
1838 				p++;
1839 				nr++;
1840 			}
1841 			state = TAILQ_NEXT(state, entry_list);
1842 		}
1843 
1844 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1845 
1846 		kfree(pstore, M_TEMP);
1847 		break;
1848 	}
1849 
1850 	case DIOCGETSTATUS: {
1851 		struct pf_status *s = (struct pf_status *)addr;
1852 		bcopy(&pf_status, s, sizeof(struct pf_status));
1853 		pfi_fill_oldstatus(s);
1854 		break;
1855 	}
1856 
1857 	case DIOCSETSTATUSIF: {
1858 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1859 
1860 		if (pi->ifname[0] == 0) {
1861 			bzero(pf_status.ifname, IFNAMSIZ);
1862 			break;
1863 		}
1864 		if (ifunit(pi->ifname) == NULL) {
1865 			error = EINVAL;
1866 			break;
1867 		}
1868 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1869 		break;
1870 	}
1871 
1872 	case DIOCCLRSTATUS: {
1873 		bzero(pf_status.counters, sizeof(pf_status.counters));
1874 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1875 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
1876 		pf_status.since = time_second;
1877 		if (*pf_status.ifname)
1878 			pfi_clr_istats(pf_status.ifname);
1879 		break;
1880 	}
1881 
1882 	case DIOCNATLOOK: {
1883 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1884 		struct pf_state_key	*sk;
1885 		struct pf_state		*state;
1886 		struct pf_state_key_cmp	 key;
1887 		int			 m = 0, direction = pnl->direction;
1888 
1889 		key.af = pnl->af;
1890 		key.proto = pnl->proto;
1891 
1892 		if (!pnl->proto ||
1893 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1894 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1895 		    ((pnl->proto == IPPROTO_TCP ||
1896 		    pnl->proto == IPPROTO_UDP) &&
1897 		    (!pnl->dport || !pnl->sport)))
1898 			error = EINVAL;
1899 		else {
1900 			/*
1901 			 * userland gives us source and dest of connection,
1902 			 * reverse the lookup so we ask for what happens with
1903 			 * the return traffic, enabling us to find it in the
1904 			 * state tree.
1905 			 */
1906 			if (direction == PF_IN) {
1907 				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
1908 				key.ext.port = pnl->dport;
1909 				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
1910 				key.gwy.port = pnl->sport;
1911 				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
1912 			} else {
1913 				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
1914 				key.lan.port = pnl->dport;
1915 				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
1916 				key.ext.port = pnl->sport;
1917 				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
1918 			}
1919 			if (m > 1)
1920 				error = E2BIG;	/* more than one state */
1921 			else if (state != NULL) {
1922 				sk = state->state_key;
1923 				if (direction == PF_IN) {
1924 					PF_ACPY(&pnl->rsaddr, &sk->lan.addr,
1925 					    sk->af);
1926 					pnl->rsport = sk->lan.port;
1927 					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
1928 					    pnl->af);
1929 					pnl->rdport = pnl->dport;
1930 				} else {
1931 					PF_ACPY(&pnl->rdaddr, &sk->gwy.addr,
1932 					    sk->af);
1933 					pnl->rdport = sk->gwy.port;
1934 					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
1935 					    pnl->af);
1936 					pnl->rsport = pnl->sport;
1937 				}
1938 			} else
1939 				error = ENOENT;
1940 		}
1941 		break;
1942 	}
1943 
1944 	case DIOCSETTIMEOUT: {
1945 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1946 		int		 old;
1947 
1948 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1949 		    pt->seconds < 0) {
1950 			error = EINVAL;
1951 			goto fail;
1952 		}
1953 		old = pf_default_rule.timeout[pt->timeout];
1954 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1955 			pt->seconds = 1;
1956 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
1957 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1958 			wakeup(pf_purge_thread);
1959 		pt->seconds = old;
1960 		break;
1961 	}
1962 
1963 	case DIOCGETTIMEOUT: {
1964 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1965 
1966 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1967 			error = EINVAL;
1968 			goto fail;
1969 		}
1970 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1971 		break;
1972 	}
1973 
1974 	case DIOCGETLIMIT: {
1975 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1976 
1977 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1978 			error = EINVAL;
1979 			goto fail;
1980 		}
1981 		pl->limit = pf_pool_limits[pl->index].limit;
1982 		break;
1983 	}
1984 
1985 	case DIOCSETLIMIT: {
1986 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1987 		int			 old_limit;
1988 
1989 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1990 		    pf_pool_limits[pl->index].pp == NULL) {
1991 			error = EINVAL;
1992 			goto fail;
1993 		}
1994 
1995 		/* XXX Get an API to set limits on the zone/pool */
1996 		old_limit = pf_pool_limits[pl->index].limit;
1997 		pf_pool_limits[pl->index].limit = pl->limit;
1998 		pl->limit = old_limit;
1999 		break;
2000 	}
2001 
2002 	case DIOCSETDEBUG: {
2003 		u_int32_t	*level = (u_int32_t *)addr;
2004 
2005 		pf_status.debug = *level;
2006 		break;
2007 	}
2008 
2009 	case DIOCCLRRULECTRS: {
2010 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2011 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2012 		struct pf_rule		*rule;
2013 
2014 		TAILQ_FOREACH(rule,
2015 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2016 			rule->evaluations = 0;
2017 			rule->packets[0] = rule->packets[1] = 0;
2018 			rule->bytes[0] = rule->bytes[1] = 0;
2019 		}
2020 		break;
2021 	}
2022 
2023 	case DIOCGIFSPEED: {
2024 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
2025 		struct pf_ifspeed	ps;
2026 		struct ifnet		*ifp;
2027 
2028 		if (psp->ifname[0] != 0) {
2029 			/* Can we completely trust user-land? */
2030 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2031 			ifp = ifunit(ps.ifname);
2032 			if (ifp )
2033 				psp->baudrate = ifp->if_baudrate;
2034 			else
2035 				error = EINVAL;
2036 		} else
2037 			error = EINVAL;
2038 		break;
2039 	}
2040 #ifdef ALTQ
2041 	case DIOCSTARTALTQ: {
2042 		struct pf_altq		*altq;
2043 
2044 		/* enable all altq interfaces on active list */
2045 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2046 			if (altq->qname[0] == 0) {
2047 				error = pf_enable_altq(altq);
2048 				if (error != 0)
2049 					break;
2050 			}
2051 		}
2052 		if (error == 0)
2053 			pf_altq_running = 1;
2054 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2055 		break;
2056 	}
2057 
2058 	case DIOCSTOPALTQ: {
2059 		struct pf_altq		*altq;
2060 
2061 		/* disable all altq interfaces on active list */
2062 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2063 			if (altq->qname[0] == 0) {
2064 				error = pf_disable_altq(altq);
2065 				if (error != 0)
2066 					break;
2067 			}
2068 		}
2069 		if (error == 0)
2070 			pf_altq_running = 0;
2071 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2072 		break;
2073 	}
2074 
2075 	case DIOCADDALTQ: {
2076 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2077 		struct pf_altq		*altq, *a;
2078 
2079 		if (pa->ticket != ticket_altqs_inactive) {
2080 			error = EBUSY;
2081 			break;
2082 		}
2083 		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2084 		if (altq == NULL) {
2085 			error = ENOMEM;
2086 			break;
2087 		}
2088 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2089 
2090 		/*
2091 		 * if this is for a queue, find the discipline and
2092 		 * copy the necessary fields
2093 		 */
2094 		if (altq->qname[0] != 0) {
2095 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2096 				error = EBUSY;
2097 				pool_put(&pf_altq_pl, altq);
2098 				break;
2099 			}
2100 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2101 				if (strncmp(a->ifname, altq->ifname,
2102 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2103 					altq->altq_disc = a->altq_disc;
2104 					break;
2105 				}
2106 			}
2107 		}
2108 
2109 		error = altq_add(altq);
2110 		if (error) {
2111 			pool_put(&pf_altq_pl, altq);
2112 			break;
2113 		}
2114 
2115 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2116 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2117 		break;
2118 	}
2119 
2120 	case DIOCGETALTQS: {
2121 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2122 		struct pf_altq		*altq;
2123 
2124 		pa->nr = 0;
2125 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2126 			pa->nr++;
2127 		pa->ticket = ticket_altqs_active;
2128 		break;
2129 	}
2130 
2131 	case DIOCGETALTQ: {
2132 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2133 		struct pf_altq		*altq;
2134 		u_int32_t		 nr;
2135 
2136 		if (pa->ticket != ticket_altqs_active) {
2137 			error = EBUSY;
2138 			break;
2139 		}
2140 		nr = 0;
2141 		altq = TAILQ_FIRST(pf_altqs_active);
2142 		while ((altq != NULL) && (nr < pa->nr)) {
2143 			altq = TAILQ_NEXT(altq, entries);
2144 			nr++;
2145 		}
2146 		if (altq == NULL) {
2147 			error = EBUSY;
2148 			break;
2149 		}
2150 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2151 		break;
2152 	}
2153 
2154 	case DIOCCHANGEALTQ:
2155 		/* CHANGEALTQ not supported yet! */
2156 		error = ENODEV;
2157 		break;
2158 
2159 	case DIOCGETQSTATS: {
2160 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2161 		struct pf_altq		*altq;
2162 		u_int32_t		 nr;
2163 		int			 nbytes;
2164 
2165 		if (pq->ticket != ticket_altqs_active) {
2166 			error = EBUSY;
2167 			break;
2168 		}
2169 		nbytes = pq->nbytes;
2170 		nr = 0;
2171 		altq = TAILQ_FIRST(pf_altqs_active);
2172 		while ((altq != NULL) && (nr < pq->nr)) {
2173 			altq = TAILQ_NEXT(altq, entries);
2174 			nr++;
2175 		}
2176 		if (altq == NULL) {
2177 			error = EBUSY;
2178 			break;
2179 		}
2180 		error = altq_getqstats(altq, pq->buf, &nbytes);
2181 		if (error == 0) {
2182 			pq->scheduler = altq->scheduler;
2183 			pq->nbytes = nbytes;
2184 		}
2185 		break;
2186 	}
2187 #endif /* ALTQ */
2188 
2189 	case DIOCBEGINADDRS: {
2190 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2191 
2192 		pf_empty_pool(&pf_pabuf);
2193 		pp->ticket = ++ticket_pabuf;
2194 		break;
2195 	}
2196 
2197 	case DIOCADDADDR: {
2198 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2199 
2200 		if (pp->ticket != ticket_pabuf) {
2201 			error = EBUSY;
2202 			break;
2203 		}
2204 #ifndef INET
2205 		if (pp->af == AF_INET) {
2206 			error = EAFNOSUPPORT;
2207 			break;
2208 		}
2209 #endif /* INET */
2210 #ifndef INET6
2211 		if (pp->af == AF_INET6) {
2212 			error = EAFNOSUPPORT;
2213 			break;
2214 		}
2215 #endif /* INET6 */
2216 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2217 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2218 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2219 			error = EINVAL;
2220 			break;
2221 		}
2222 		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2223 		if (pa == NULL) {
2224 			error = ENOMEM;
2225 			break;
2226 		}
2227 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2228 		if (pa->ifname[0]) {
2229 			pa->kif = pfi_kif_get(pa->ifname);
2230 			if (pa->kif == NULL) {
2231 				pool_put(&pf_pooladdr_pl, pa);
2232 				error = EINVAL;
2233 				break;
2234 			}
2235 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2236 		}
2237 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2238 			pfi_dynaddr_remove(&pa->addr);
2239 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2240 			pool_put(&pf_pooladdr_pl, pa);
2241 			error = EINVAL;
2242 			break;
2243 		}
2244 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2245 		break;
2246 	}
2247 
2248 	case DIOCGETADDRS: {
2249 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2250 
2251 		pp->nr = 0;
2252 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2253 		    pp->r_num, 0, 1, 0);
2254 		if (pool == NULL) {
2255 			error = EBUSY;
2256 			break;
2257 		}
2258 		TAILQ_FOREACH(pa, &pool->list, entries)
2259 			pp->nr++;
2260 		break;
2261 	}
2262 
2263 	case DIOCGETADDR: {
2264 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2265 		u_int32_t		 nr = 0;
2266 
2267 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2268 		    pp->r_num, 0, 1, 1);
2269 		if (pool == NULL) {
2270 			error = EBUSY;
2271 			break;
2272 		}
2273 		pa = TAILQ_FIRST(&pool->list);
2274 		while ((pa != NULL) && (nr < pp->nr)) {
2275 			pa = TAILQ_NEXT(pa, entries);
2276 			nr++;
2277 		}
2278 		if (pa == NULL) {
2279 			error = EBUSY;
2280 			break;
2281 		}
2282 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2283 		pfi_dynaddr_copyout(&pp->addr.addr);
2284 		pf_tbladdr_copyout(&pp->addr.addr);
2285 		pf_rtlabel_copyout(&pp->addr.addr);
2286 		break;
2287 	}
2288 
2289 	case DIOCCHANGEADDR: {
2290 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2291 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2292 		struct pf_ruleset	*ruleset;
2293 
2294 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2295 		    pca->action > PF_CHANGE_REMOVE) {
2296 			error = EINVAL;
2297 			break;
2298 		}
2299 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2300 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2301 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2302 			error = EINVAL;
2303 			break;
2304 		}
2305 
2306 		ruleset = pf_find_ruleset(pca->anchor);
2307 		if (ruleset == NULL) {
2308 			error = EBUSY;
2309 			break;
2310 		}
2311 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2312 		    pca->r_num, pca->r_last, 1, 1);
2313 		if (pool == NULL) {
2314 			error = EBUSY;
2315 			break;
2316 		}
2317 		if (pca->action != PF_CHANGE_REMOVE) {
2318 			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2319 			if (newpa == NULL) {
2320 				error = ENOMEM;
2321 				break;
2322 			}
2323 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2324 #ifndef INET
2325 			if (pca->af == AF_INET) {
2326 				pool_put(&pf_pooladdr_pl, newpa);
2327 				error = EAFNOSUPPORT;
2328 				break;
2329 			}
2330 #endif /* INET */
2331 #ifndef INET6
2332 			if (pca->af == AF_INET6) {
2333 				pool_put(&pf_pooladdr_pl, newpa);
2334 				error = EAFNOSUPPORT;
2335 				break;
2336 			}
2337 #endif /* INET6 */
2338 			if (newpa->ifname[0]) {
2339 				newpa->kif = pfi_kif_get(newpa->ifname);
2340 				if (newpa->kif == NULL) {
2341 					pool_put(&pf_pooladdr_pl, newpa);
2342 					error = EINVAL;
2343 					break;
2344 				}
2345 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2346 			} else
2347 				newpa->kif = NULL;
2348 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2349 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2350 				pfi_dynaddr_remove(&newpa->addr);
2351 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2352 				pool_put(&pf_pooladdr_pl, newpa);
2353 				error = EINVAL;
2354 				break;
2355 			}
2356 		}
2357 
2358 		if (pca->action == PF_CHANGE_ADD_HEAD)
2359 			oldpa = TAILQ_FIRST(&pool->list);
2360 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2361 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2362 		else {
2363 			int	i = 0;
2364 
2365 			oldpa = TAILQ_FIRST(&pool->list);
2366 			while ((oldpa != NULL) && (i < pca->nr)) {
2367 				oldpa = TAILQ_NEXT(oldpa, entries);
2368 				i++;
2369 			}
2370 			if (oldpa == NULL) {
2371 				error = EINVAL;
2372 				break;
2373 			}
2374 		}
2375 
2376 		if (pca->action == PF_CHANGE_REMOVE) {
2377 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2378 			pfi_dynaddr_remove(&oldpa->addr);
2379 			pf_tbladdr_remove(&oldpa->addr);
2380 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2381 			pool_put(&pf_pooladdr_pl, oldpa);
2382 		} else {
2383 			if (oldpa == NULL)
2384 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2385 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2386 			    pca->action == PF_CHANGE_ADD_BEFORE)
2387 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2388 			else
2389 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2390 				    newpa, entries);
2391 		}
2392 
2393 		pool->cur = TAILQ_FIRST(&pool->list);
2394 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2395 		    pca->af);
2396 		break;
2397 	}
2398 
2399 	case DIOCGETRULESETS: {
2400 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2401 		struct pf_ruleset	*ruleset;
2402 		struct pf_anchor	*anchor;
2403 
2404 		pr->path[sizeof(pr->path) - 1] = 0;
2405 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2406 			error = EINVAL;
2407 			break;
2408 		}
2409 		pr->nr = 0;
2410 		if (ruleset->anchor == NULL) {
2411 			/* XXX kludge for pf_main_ruleset */
2412 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2413 				if (anchor->parent == NULL)
2414 					pr->nr++;
2415 		} else {
2416 			RB_FOREACH(anchor, pf_anchor_node,
2417 			    &ruleset->anchor->children)
2418 				pr->nr++;
2419 		}
2420 		break;
2421 	}
2422 
2423 	case DIOCGETRULESET: {
2424 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2425 		struct pf_ruleset	*ruleset;
2426 		struct pf_anchor	*anchor;
2427 		u_int32_t		 nr = 0;
2428 
2429 		pr->path[sizeof(pr->path) - 1] = 0;
2430 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2431 			error = EINVAL;
2432 			break;
2433 		}
2434 		pr->name[0] = 0;
2435 		if (ruleset->anchor == NULL) {
2436 			/* XXX kludge for pf_main_ruleset */
2437 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2438 				if (anchor->parent == NULL && nr++ == pr->nr) {
2439 					strlcpy(pr->name, anchor->name,
2440 					    sizeof(pr->name));
2441 					break;
2442 				}
2443 		} else {
2444 			RB_FOREACH(anchor, pf_anchor_node,
2445 			    &ruleset->anchor->children)
2446 				if (nr++ == pr->nr) {
2447 					strlcpy(pr->name, anchor->name,
2448 					    sizeof(pr->name));
2449 					break;
2450 				}
2451 		}
2452 		if (!pr->name[0])
2453 			error = EBUSY;
2454 		break;
2455 	}
2456 
2457 	case DIOCRCLRTABLES: {
2458 		struct pfioc_table *io = (struct pfioc_table *)addr;
2459 
2460 		if (io->pfrio_esize != 0) {
2461 			error = ENODEV;
2462 			break;
2463 		}
2464 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2465 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2466 		break;
2467 	}
2468 
2469 	case DIOCRADDTABLES: {
2470 		struct pfioc_table *io = (struct pfioc_table *)addr;
2471 
2472 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2473 			error = ENODEV;
2474 			break;
2475 		}
2476 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2477 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2478 		break;
2479 	}
2480 
2481 	case DIOCRDELTABLES: {
2482 		struct pfioc_table *io = (struct pfioc_table *)addr;
2483 
2484 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2485 			error = ENODEV;
2486 			break;
2487 		}
2488 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2489 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2490 		break;
2491 	}
2492 
2493 	case DIOCRGETTABLES: {
2494 		struct pfioc_table *io = (struct pfioc_table *)addr;
2495 
2496 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2497 			error = ENODEV;
2498 			break;
2499 		}
2500 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2501 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2502 		break;
2503 	}
2504 
2505 	case DIOCRGETTSTATS: {
2506 		struct pfioc_table *io = (struct pfioc_table *)addr;
2507 
2508 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2509 			error = ENODEV;
2510 			break;
2511 		}
2512 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2513 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2514 		break;
2515 	}
2516 
2517 	case DIOCRCLRTSTATS: {
2518 		struct pfioc_table *io = (struct pfioc_table *)addr;
2519 
2520 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2521 			error = ENODEV;
2522 			break;
2523 		}
2524 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2525 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2526 		break;
2527 	}
2528 
2529 	case DIOCRSETTFLAGS: {
2530 		struct pfioc_table *io = (struct pfioc_table *)addr;
2531 
2532 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2533 			error = ENODEV;
2534 			break;
2535 		}
2536 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2537 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2538 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2539 		break;
2540 	}
2541 
2542 	case DIOCRCLRADDRS: {
2543 		struct pfioc_table *io = (struct pfioc_table *)addr;
2544 
2545 		if (io->pfrio_esize != 0) {
2546 			error = ENODEV;
2547 			break;
2548 		}
2549 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2550 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2551 		break;
2552 	}
2553 
2554 	case DIOCRADDADDRS: {
2555 		struct pfioc_table *io = (struct pfioc_table *)addr;
2556 
2557 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2558 			error = ENODEV;
2559 			break;
2560 		}
2561 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2562 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2563 		    PFR_FLAG_USERIOCTL);
2564 		break;
2565 	}
2566 
2567 	case DIOCRDELADDRS: {
2568 		struct pfioc_table *io = (struct pfioc_table *)addr;
2569 
2570 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2571 			error = ENODEV;
2572 			break;
2573 		}
2574 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2575 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2576 		    PFR_FLAG_USERIOCTL);
2577 		break;
2578 	}
2579 
2580 	case DIOCRSETADDRS: {
2581 		struct pfioc_table *io = (struct pfioc_table *)addr;
2582 
2583 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2584 			error = ENODEV;
2585 			break;
2586 		}
2587 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2588 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2589 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2590 		    PFR_FLAG_USERIOCTL, 0);
2591 		break;
2592 	}
2593 
2594 	case DIOCRGETADDRS: {
2595 		struct pfioc_table *io = (struct pfioc_table *)addr;
2596 
2597 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2598 			error = ENODEV;
2599 			break;
2600 		}
2601 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2602 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2603 		break;
2604 	}
2605 
2606 	case DIOCRGETASTATS: {
2607 		struct pfioc_table *io = (struct pfioc_table *)addr;
2608 
2609 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2610 			error = ENODEV;
2611 			break;
2612 		}
2613 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2614 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2615 		break;
2616 	}
2617 
2618 	case DIOCRCLRASTATS: {
2619 		struct pfioc_table *io = (struct pfioc_table *)addr;
2620 
2621 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2622 			error = ENODEV;
2623 			break;
2624 		}
2625 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2626 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2627 		    PFR_FLAG_USERIOCTL);
2628 		break;
2629 	}
2630 
2631 	case DIOCRTSTADDRS: {
2632 		struct pfioc_table *io = (struct pfioc_table *)addr;
2633 
2634 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2635 			error = ENODEV;
2636 			break;
2637 		}
2638 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2639 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2640 		    PFR_FLAG_USERIOCTL);
2641 		break;
2642 	}
2643 
2644 	case DIOCRINADEFINE: {
2645 		struct pfioc_table *io = (struct pfioc_table *)addr;
2646 
2647 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2648 			error = ENODEV;
2649 			break;
2650 		}
2651 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2652 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2653 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2654 		break;
2655 	}
2656 
2657 	case DIOCOSFPADD: {
2658 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2659 		error = pf_osfp_add(io);
2660 		break;
2661 	}
2662 
2663 	case DIOCOSFPGET: {
2664 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2665 		error = pf_osfp_get(io);
2666 		break;
2667 	}
2668 
2669 	case DIOCXBEGIN: {
2670 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2671 		struct pfioc_trans_e	*ioe;
2672 		struct pfr_table	*table;
2673 		int			 i;
2674 
2675 		if (io->esize != sizeof(*ioe)) {
2676 			error = ENODEV;
2677 			goto fail;
2678 		}
2679 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2680 		    M_TEMP, M_WAITOK);
2681 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2682 		    M_TEMP, M_WAITOK);
2683 		for (i = 0; i < io->size; i++) {
2684 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2685 				kfree(table, M_TEMP);
2686 				kfree(ioe, M_TEMP);
2687 				error = EFAULT;
2688 				goto fail;
2689 			}
2690 			switch (ioe->rs_num) {
2691 #ifdef ALTQ
2692 			case PF_RULESET_ALTQ:
2693 				if (ioe->anchor[0]) {
2694 					kfree(table, M_TEMP);
2695 					kfree(ioe, M_TEMP);
2696 					error = EINVAL;
2697 					goto fail;
2698 				}
2699 				if ((error = pf_begin_altq(&ioe->ticket))) {
2700 					kfree(table, M_TEMP);
2701 					kfree(ioe, M_TEMP);
2702 					goto fail;
2703 				}
2704 				break;
2705 #endif /* ALTQ */
2706 			case PF_RULESET_TABLE:
2707 				bzero(table, sizeof(*table));
2708 				strlcpy(table->pfrt_anchor, ioe->anchor,
2709 				    sizeof(table->pfrt_anchor));
2710 				if ((error = pfr_ina_begin(table,
2711 				    &ioe->ticket, NULL, 0))) {
2712 					kfree(table, M_TEMP);
2713 					kfree(ioe, M_TEMP);
2714 					goto fail;
2715 				}
2716 				break;
2717 			default:
2718 				if ((error = pf_begin_rules(&ioe->ticket,
2719 				    ioe->rs_num, ioe->anchor))) {
2720 					kfree(table, M_TEMP);
2721 					kfree(ioe, M_TEMP);
2722 					goto fail;
2723 				}
2724 				break;
2725 			}
2726 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2727 				kfree(table, M_TEMP);
2728 				kfree(ioe, M_TEMP);
2729 				error = EFAULT;
2730 				goto fail;
2731 			}
2732 		}
2733 		kfree(table, M_TEMP);
2734 		kfree(ioe, M_TEMP);
2735 		break;
2736 	}
2737 
2738 	case DIOCXROLLBACK: {
2739 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2740 		struct pfioc_trans_e	*ioe;
2741 		struct pfr_table	*table;
2742 		int			 i;
2743 
2744 		if (io->esize != sizeof(*ioe)) {
2745 			error = ENODEV;
2746 			goto fail;
2747 		}
2748 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2749 		    M_TEMP, M_WAITOK);
2750 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2751 		    M_TEMP, M_WAITOK);
2752 		for (i = 0; i < io->size; i++) {
2753 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2754 				kfree(table, M_TEMP);
2755 				kfree(ioe, M_TEMP);
2756 				error = EFAULT;
2757 				goto fail;
2758 			}
2759 			switch (ioe->rs_num) {
2760 #ifdef ALTQ
2761 			case PF_RULESET_ALTQ:
2762 				if (ioe->anchor[0]) {
2763 					kfree(table, M_TEMP);
2764 					kfree(ioe, M_TEMP);
2765 					error = EINVAL;
2766 					goto fail;
2767 				}
2768 				if ((error = pf_rollback_altq(ioe->ticket))) {
2769 					kfree(table, M_TEMP);
2770 					kfree(ioe, M_TEMP);
2771 					goto fail; /* really bad */
2772 				}
2773 				break;
2774 #endif /* ALTQ */
2775 			case PF_RULESET_TABLE:
2776 				bzero(table, sizeof(*table));
2777 				strlcpy(table->pfrt_anchor, ioe->anchor,
2778 				    sizeof(table->pfrt_anchor));
2779 				if ((error = pfr_ina_rollback(table,
2780 				    ioe->ticket, NULL, 0))) {
2781 					kfree(table, M_TEMP);
2782 					kfree(ioe, M_TEMP);
2783 					goto fail; /* really bad */
2784 				}
2785 				break;
2786 			default:
2787 				if ((error = pf_rollback_rules(ioe->ticket,
2788 				    ioe->rs_num, ioe->anchor))) {
2789 					kfree(table, M_TEMP);
2790 					kfree(ioe, M_TEMP);
2791 					goto fail; /* really bad */
2792 				}
2793 				break;
2794 			}
2795 		}
2796 		kfree(table, M_TEMP);
2797 		kfree(ioe, M_TEMP);
2798 		break;
2799 	}
2800 
2801 	case DIOCXCOMMIT: {
2802 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2803 		struct pfioc_trans_e	*ioe;
2804 		struct pfr_table	*table;
2805 		struct pf_ruleset	*rs;
2806 		int			 i;
2807 
2808 		if (io->esize != sizeof(*ioe)) {
2809 			error = ENODEV;
2810 			goto fail;
2811 		}
2812 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2813 		    M_TEMP, M_WAITOK);
2814 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2815 		    M_TEMP, M_WAITOK);
2816 		/* first makes sure everything will succeed */
2817 		for (i = 0; i < io->size; i++) {
2818 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2819 				kfree(table, M_TEMP);
2820 				kfree(ioe, M_TEMP);
2821 				error = EFAULT;
2822 				goto fail;
2823 			}
2824 			switch (ioe->rs_num) {
2825 #ifdef ALTQ
2826 			case PF_RULESET_ALTQ:
2827 				if (ioe->anchor[0]) {
2828 					kfree(table, M_TEMP);
2829 					kfree(ioe, M_TEMP);
2830 					error = EINVAL;
2831 					goto fail;
2832 				}
2833 				if (!altqs_inactive_open || ioe->ticket !=
2834 				    ticket_altqs_inactive) {
2835 					kfree(table, M_TEMP);
2836 					kfree(ioe, M_TEMP);
2837 					error = EBUSY;
2838 					goto fail;
2839 				}
2840 				break;
2841 #endif /* ALTQ */
2842 			case PF_RULESET_TABLE:
2843 				rs = pf_find_ruleset(ioe->anchor);
2844 				if (rs == NULL || !rs->topen || ioe->ticket !=
2845 				     rs->tticket) {
2846 					kfree(table, M_TEMP);
2847 					kfree(ioe, M_TEMP);
2848 					error = EBUSY;
2849 					goto fail;
2850 				}
2851 				break;
2852 			default:
2853 				if (ioe->rs_num < 0 || ioe->rs_num >=
2854 				    PF_RULESET_MAX) {
2855 					kfree(table, M_TEMP);
2856 					kfree(ioe, M_TEMP);
2857 					error = EINVAL;
2858 					goto fail;
2859 				}
2860 				rs = pf_find_ruleset(ioe->anchor);
2861 				if (rs == NULL ||
2862 				    !rs->rules[ioe->rs_num].inactive.open ||
2863 				    rs->rules[ioe->rs_num].inactive.ticket !=
2864 				    ioe->ticket) {
2865 					kfree(table, M_TEMP);
2866 					kfree(ioe, M_TEMP);
2867 					error = EBUSY;
2868 					goto fail;
2869 				}
2870 				break;
2871 			}
2872 		}
2873 		/* now do the commit - no errors should happen here */
2874 		for (i = 0; i < io->size; i++) {
2875 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2876 				kfree(table, M_TEMP);
2877 				kfree(ioe, M_TEMP);
2878 				error = EFAULT;
2879 				goto fail;
2880 			}
2881 			switch (ioe->rs_num) {
2882 #ifdef ALTQ
2883 			case PF_RULESET_ALTQ:
2884 				if ((error = pf_commit_altq(ioe->ticket))) {
2885 					kfree(table, M_TEMP);
2886 					kfree(ioe, M_TEMP);
2887 					goto fail; /* really bad */
2888 				}
2889 				break;
2890 #endif /* ALTQ */
2891 			case PF_RULESET_TABLE:
2892 				bzero(table, sizeof(*table));
2893 				strlcpy(table->pfrt_anchor, ioe->anchor,
2894 				    sizeof(table->pfrt_anchor));
2895 				if ((error = pfr_ina_commit(table, ioe->ticket,
2896 				    NULL, NULL, 0))) {
2897 					kfree(table, M_TEMP);
2898 					kfree(ioe, M_TEMP);
2899 					goto fail; /* really bad */
2900 				}
2901 				break;
2902 			default:
2903 				if ((error = pf_commit_rules(ioe->ticket,
2904 				    ioe->rs_num, ioe->anchor))) {
2905 					kfree(table, M_TEMP);
2906 					kfree(ioe, M_TEMP);
2907 					goto fail; /* really bad */
2908 				}
2909 				break;
2910 			}
2911 		}
2912 		kfree(table, M_TEMP);
2913 		kfree(ioe, M_TEMP);
2914 		break;
2915 	}
2916 
2917 	case DIOCGETSRCNODES: {
2918 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2919 		struct pf_src_node	*n, *p, *pstore;
2920 		u_int32_t		 nr = 0;
2921 		int			 space = psn->psn_len;
2922 
2923 		if (space == 0) {
2924 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2925 				nr++;
2926 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2927 			break;
2928 		}
2929 
2930 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2931 
2932 		p = psn->psn_src_nodes;
2933 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2934 			int	secs = time_second, diff;
2935 
2936 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
2937 				break;
2938 
2939 			bcopy(n, pstore, sizeof(*pstore));
2940 			if (n->rule.ptr != NULL)
2941 				pstore->rule.nr = n->rule.ptr->nr;
2942 			pstore->creation = secs - pstore->creation;
2943 			if (pstore->expire > secs)
2944 				pstore->expire -= secs;
2945 			else
2946 				pstore->expire = 0;
2947 
2948 			/* adjust the connection rate estimate */
2949 			diff = secs - n->conn_rate.last;
2950 			if (diff >= n->conn_rate.seconds)
2951 				pstore->conn_rate.count = 0;
2952 			else
2953 				pstore->conn_rate.count -=
2954 				    n->conn_rate.count * diff /
2955 				    n->conn_rate.seconds;
2956 
2957 			error = copyout(pstore, p, sizeof(*p));
2958 			if (error) {
2959 				kfree(pstore, M_TEMP);
2960 				goto fail;
2961 			}
2962 			p++;
2963 			nr++;
2964 		}
2965 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2966 
2967 		kfree(pstore, M_TEMP);
2968 		break;
2969 	}
2970 
2971 	case DIOCCLRSRCNODES: {
2972 		struct pf_src_node	*n;
2973 		struct pf_state		*state;
2974 
2975 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2976 			state->src_node = NULL;
2977 			state->nat_src_node = NULL;
2978 		}
2979 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2980 			n->expire = 1;
2981 			n->states = 0;
2982 		}
2983 		pf_purge_expired_src_nodes(1);
2984 		pf_status.src_nodes = 0;
2985 		break;
2986 	}
2987 
2988 	case DIOCKILLSRCNODES: {
2989 		struct pf_src_node	*sn;
2990 		struct pf_state		*s;
2991 		struct pfioc_src_node_kill *psnk = \
2992 			(struct pfioc_src_node_kill *) addr;
2993 		int			killed = 0;
2994 
2995 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
2996         		if (PF_MATCHA(psnk->psnk_src.neg, \
2997 				      &psnk->psnk_src.addr.v.a.addr, \
2998 				      &psnk->psnk_src.addr.v.a.mask, \
2999 				      &sn->addr, sn->af) &&
3000 			    PF_MATCHA(psnk->psnk_dst.neg, \
3001 				      &psnk->psnk_dst.addr.v.a.addr, \
3002 				      &psnk->psnk_dst.addr.v.a.mask, \
3003 				      &sn->raddr, sn->af)) {
3004 				/* Handle state to src_node linkage */
3005 				if (sn->states != 0) {
3006 					RB_FOREACH(s, pf_state_tree_id,
3007 					    &tree_id) {
3008 						if (s->src_node == sn)
3009 							s->src_node = NULL;
3010 						if (s->nat_src_node == sn)
3011 							s->nat_src_node = NULL;
3012 					}
3013 					sn->states = 0;
3014 				}
3015 				sn->expire = 1;
3016 				killed++;
3017 			}
3018 		}
3019 
3020 		if (killed > 0)
3021 			pf_purge_expired_src_nodes(1);
3022 
3023 		psnk->psnk_af = killed;
3024 		break;
3025 	}
3026 
3027 	case DIOCSETHOSTID: {
3028 		u_int32_t	*hostid = (u_int32_t *)addr;
3029 
3030 		if (*hostid == 0)
3031 			pf_status.hostid = karc4random();
3032 		else
3033 			pf_status.hostid = *hostid;
3034 		break;
3035 	}
3036 
3037 	case DIOCOSFPFLUSH:
3038 		crit_enter();
3039 		pf_osfp_flush();
3040 		crit_exit();
3041 		break;
3042 
3043 	case DIOCIGETIFACES: {
3044 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3045 
3046 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3047 			error = ENODEV;
3048 			break;
3049 		}
3050 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3051 		    &io->pfiio_size);
3052 		break;
3053 	}
3054 
3055 	case DIOCSETIFFLAG: {
3056 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3057 
3058 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3059 		break;
3060 	}
3061 
3062 	case DIOCCLRIFFLAG: {
3063 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3064 
3065 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3066 		break;
3067 	}
3068 
3069 	default:
3070 		error = ENODEV;
3071 		break;
3072 	}
3073 fail:
3074 	lwkt_reltoken(&pf_token);
3075 	return (error);
3076 }
3077 
3078 /*
3079  * XXX - Check for version missmatch!!!
3080  */
3081 static void
3082 pf_clear_states(void)
3083 {
3084 	struct pf_state		*state;
3085 
3086 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3087 		state->timeout = PFTM_PURGE;
3088 #if NPFSYNC
3089 		/* don't send out individual delete messages */
3090 		state->sync_flags = PFSTATE_NOSYNC;
3091 #endif
3092 		pf_unlink_state(state);
3093 	}
3094 	pf_status.states = 0;
3095 #if 0 /* NPFSYNC */
3096 /*
3097  * XXX This is called on module unload, we do not want to sync that over? */
3098  */
3099 	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
3100 #endif
3101 }
3102 
3103 static int
3104 pf_clear_tables(void)
3105 {
3106 	struct pfioc_table io;
3107 	int error;
3108 
3109 	bzero(&io, sizeof(io));
3110 
3111 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3112 	    io.pfrio_flags);
3113 
3114 	return (error);
3115 }
3116 
3117 static void
3118 pf_clear_srcnodes(void)
3119 {
3120 	struct pf_src_node	*n;
3121 	struct pf_state		*state;
3122 
3123 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3124 		state->src_node = NULL;
3125 		state->nat_src_node = NULL;
3126 	}
3127 	RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3128 		n->expire = 1;
3129 		n->states = 0;
3130 	}
3131 	pf_purge_expired_src_nodes(0);
3132 	pf_status.src_nodes = 0;
3133 }
3134 /*
3135  * XXX - Check for version missmatch!!!
3136  */
3137 
3138 /*
3139  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3140  */
3141 static int
3142 shutdown_pf(void)
3143 {
3144 	int error = 0;
3145 	u_int32_t t[5];
3146 	char nn = '\0';
3147 
3148 	pf_status.running = 0;
3149 	do {
3150 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
3151 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3152 			break;
3153 		}
3154 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
3155 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3156 			break;		/* XXX: rollback? */
3157 		}
3158 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
3159 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3160 			break;		/* XXX: rollback? */
3161 		}
3162 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3163 		    != 0) {
3164 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3165 			break;		/* XXX: rollback? */
3166 		}
3167 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3168 		    != 0) {
3169 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3170 			break;		/* XXX: rollback? */
3171 		}
3172 
3173 		/* XXX: these should always succeed here */
3174 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3175 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3176 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3177 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3178 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3179 
3180 		if ((error = pf_clear_tables()) != 0)
3181 			break;
3182 
3183 #ifdef ALTQ
3184 		if ((error = pf_begin_altq(&t[0])) != 0) {
3185 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3186 			break;
3187 		}
3188 		pf_commit_altq(t[0]);
3189 #endif
3190 
3191 		pf_clear_states();
3192 
3193 		pf_clear_srcnodes();
3194 
3195 		/* status does not use malloced mem so no need to cleanup */
3196 		/* fingerprints and interfaces have their own cleanup code */
3197 	} while(0);
3198 
3199         return (error);
3200 }
3201 
3202 static int
3203 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3204 {
3205 	/*
3206 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3207 	 * for ip_len/ip_off. This is why we don't have to change byte order
3208 	 * like the FreeBSD-5 version does.
3209 	 */
3210 	int chk;
3211 
3212 	lwkt_gettoken(&pf_token);
3213 
3214 	chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3215 	if (chk && *m) {
3216 		m_freem(*m);
3217 		*m = NULL;
3218 	}
3219 	lwkt_reltoken(&pf_token);
3220 	return chk;
3221 }
3222 
3223 static int
3224 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3225 {
3226 	/*
3227 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3228 	 * for ip_len/ip_off. This is why we don't have to change byte order
3229 	 * like the FreeBSD-5 version does.
3230 	 */
3231 	int chk;
3232 
3233 	lwkt_gettoken(&pf_token);
3234 
3235 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3236 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3237 		in_delayed_cksum(*m);
3238 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3239 	}
3240 	chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3241 	if (chk && *m) {
3242 		m_freem(*m);
3243 		*m = NULL;
3244 	}
3245 	lwkt_reltoken(&pf_token);
3246 	return chk;
3247 }
3248 
3249 #ifdef INET6
3250 static int
3251 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3252 {
3253 	/*
3254 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3255 	 */
3256 	int chk;
3257 
3258 	lwkt_gettoken(&pf_token);
3259 
3260 	chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3261 	if (chk && *m) {
3262 		m_freem(*m);
3263 		*m = NULL;
3264 	}
3265 	lwkt_reltoken(&pf_token);
3266 	return chk;
3267 }
3268 
3269 static int
3270 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3271 {
3272 	/*
3273 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3274 	 */
3275 	int chk;
3276 
3277 	lwkt_gettoken(&pf_token);
3278 
3279 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3280 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3281 		in_delayed_cksum(*m);
3282 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3283 	}
3284 	chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3285 	if (chk && *m) {
3286 		m_freem(*m);
3287 		*m = NULL;
3288 	}
3289 	lwkt_reltoken(&pf_token);
3290 	return chk;
3291 }
3292 #endif /* INET6 */
3293 
3294 static int
3295 hook_pf(void)
3296 {
3297 	struct pfil_head *pfh_inet;
3298 #ifdef INET6
3299 	struct pfil_head *pfh_inet6;
3300 #endif
3301 
3302 	lwkt_gettoken(&pf_token);
3303 
3304 	if (pf_pfil_hooked) {
3305 		lwkt_reltoken(&pf_token);
3306 		return (0);
3307 	}
3308 
3309 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3310 	if (pfh_inet == NULL) {
3311 		lwkt_reltoken(&pf_token);
3312 		return (ENODEV);
3313 	}
3314 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet);
3315 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet);
3316 #ifdef INET6
3317 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3318 	if (pfh_inet6 == NULL) {
3319 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3320 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3321 		lwkt_reltoken(&pf_token);
3322 		return (ENODEV);
3323 	}
3324 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet6);
3325 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet6);
3326 #endif
3327 
3328 	pf_pfil_hooked = 1;
3329 	lwkt_reltoken(&pf_token);
3330 	return (0);
3331 }
3332 
3333 static int
3334 dehook_pf(void)
3335 {
3336 	struct pfil_head *pfh_inet;
3337 #ifdef INET6
3338 	struct pfil_head *pfh_inet6;
3339 #endif
3340 
3341 	lwkt_gettoken(&pf_token);
3342 
3343 	if (pf_pfil_hooked == 0) {
3344 		lwkt_reltoken(&pf_token);
3345 		return (0);
3346 	}
3347 
3348 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3349 	if (pfh_inet == NULL) {
3350 		lwkt_reltoken(&pf_token);
3351 		return (ENODEV);
3352 	}
3353 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3354 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3355 #ifdef INET6
3356 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3357 	if (pfh_inet6 == NULL) {
3358 		lwkt_reltoken(&pf_token);
3359 		return (ENODEV);
3360 	}
3361 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3362 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3363 #endif
3364 
3365 	pf_pfil_hooked = 0;
3366 	lwkt_reltoken(&pf_token);
3367 	return (0);
3368 }
3369 
3370 static int
3371 pf_load(void)
3372 {
3373 	int error;
3374 
3375 	lwkt_gettoken(&pf_token);
3376 
3377 	init_zone_var();
3378 	lockinit(&pf_mod_lck, "pf task lck", 0, LK_CANRECURSE);
3379 	pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3380 	error = pfattach();
3381 	if (error) {
3382 		dev_ops_remove_all(&pf_ops);
3383 		lockuninit(&pf_mod_lck);
3384 		lwkt_reltoken(&pf_token);
3385 		return (error);
3386 	}
3387 	lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3388 	lwkt_reltoken(&pf_token);
3389 	return (0);
3390 }
3391 
3392 static int
3393 pf_unload(void)
3394 {
3395 	int error;
3396 	pf_status.running = 0;
3397 
3398 	lwkt_gettoken(&pf_token);
3399 
3400 	error = dehook_pf();
3401 	if (error) {
3402 		/*
3403 		 * Should not happen!
3404 		 * XXX Due to error code ESRCH, kldunload will show
3405 		 * a message like 'No such process'.
3406 		 */
3407 		kprintf("pfil unregistration fail\n");
3408 		lwkt_reltoken(&pf_token);
3409 		return error;
3410 	}
3411 	shutdown_pf();
3412 	pf_end_threads = 1;
3413 	while (pf_end_threads < 2) {
3414 		wakeup_one(pf_purge_thread);
3415 		lksleep(pf_purge_thread, &pf_mod_lck, 0, "pftmo", hz);
3416 
3417 	}
3418 	pfi_cleanup();
3419 	pf_osfp_flush();
3420 	pf_osfp_cleanup();
3421 	cleanup_pf_zone();
3422 	dev_ops_remove_all(&pf_ops);
3423 	lockuninit(&pf_consistency_lock);
3424 	lockuninit(&pf_mod_lck);
3425 	lwkt_reltoken(&pf_token);
3426 	return 0;
3427 }
3428 
3429 static int
3430 pf_modevent(module_t mod, int type, void *data)
3431 {
3432 	int error = 0;
3433 
3434 	lwkt_gettoken(&pf_token);
3435 
3436 	switch(type) {
3437 	case MOD_LOAD:
3438 		error = pf_load();
3439 		break;
3440 
3441 	case MOD_UNLOAD:
3442 		error = pf_unload();
3443 		break;
3444 	default:
3445 		error = EINVAL;
3446 		break;
3447 	}
3448 	lwkt_reltoken(&pf_token);
3449 	return error;
3450 }
3451 
3452 static moduledata_t pf_mod = {
3453 	"pf",
3454 	pf_modevent,
3455 	0
3456 };
3457 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3458 MODULE_VERSION(pf, PF_MODVER);
3459