xref: /dflybsd-src/sys/net/pf/pf_ioctl.c (revision 0bb7d8c82a64940013681cf515d16f3e62eb7e3c)
1 /*	$FreeBSD: src/sys/contrib/pf/net/pf_ioctl.c,v 1.12 2004/08/12 14:15:42 mlaier Exp $	*/
2 /*	$OpenBSD: pf_ioctl.c,v 1.112.2.2 2004/07/24 18:28:12 brad Exp $ */
3 /*	$OpenBSD: pf_ioctl.c,v 1.175 2007/02/26 22:47:43 deraadt Exp $ */
4 
5 /*
6  * Copyright (c) 2010 The DragonFly Project.  All rights reserved.
7  *
8  * Copyright (c) 2001 Daniel Hartmeier
9  * Copyright (c) 2002,2003 Henning Brauer
10  * All rights reserved.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  *
16  *    - Redistributions of source code must retain the above copyright
17  *      notice, this list of conditions and the following disclaimer.
18  *    - Redistributions in binary form must reproduce the above
19  *      copyright notice, this list of conditions and the following
20  *      disclaimer in the documentation and/or other materials provided
21  *      with the distribution.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
26  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
27  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
28  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
29  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
30  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
31  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
33  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  *
36  * Effort sponsored in part by the Defense Advanced Research Projects
37  * Agency (DARPA) and Air Force Research Laboratory, Air Force
38  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
39  *
40  */
41 
42 #include "opt_inet.h"
43 #include "opt_inet6.h"
44 #include "use_pfsync.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/conf.h>
49 #include <sys/device.h>
50 #include <sys/mbuf.h>
51 #include <sys/filio.h>
52 #include <sys/fcntl.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/kernel.h>
56 #include <sys/kthread.h>
57 #include <sys/time.h>
58 #include <sys/proc.h>
59 #include <sys/malloc.h>
60 #include <sys/module.h>
61 #include <vm/vm_zone.h>
62 #include <sys/lock.h>
63 
64 #include <sys/thread2.h>
65 #include <sys/mplock2.h>
66 
67 #include <net/if.h>
68 #include <net/if_types.h>
69 #include <net/route.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <netinet/in_systm.h>
74 #include <netinet/ip.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_icmp.h>
77 
78 #include <net/pf/pfvar.h>
79 #include <sys/md5.h>
80 #include <net/pf/pfvar.h>
81 
82 #if NPFSYNC > 0
83 #include <net/pf/if_pfsync.h>
84 #endif /* NPFSYNC > 0 */
85 
86 #if NPFLOG > 0
87 #include <net/if_pflog.h>
88 #endif /* NPFLOG > 0 */
89 
90 #ifdef INET6
91 #include <netinet/ip6.h>
92 #include <netinet/in_pcb.h>
93 #endif /* INET6 */
94 
95 #ifdef ALTQ
96 #include <net/altq/altq.h>
97 #endif
98 
99 #include <machine/limits.h>
100 #include <net/pfil.h>
101 #include <sys/mutex.h>
102 
103 u_int rt_numfibs = RT_NUMFIBS;
104 
105 void			 init_zone_var(void);
106 void			 cleanup_pf_zone(void);
107 int			 pfattach(void);
108 struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
109 			    u_int8_t, u_int8_t, u_int8_t);
110 
111 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
112 void			 pf_empty_pool(struct pf_palist *);
113 #ifdef ALTQ
114 int			 pf_begin_altq(u_int32_t *);
115 int			 pf_rollback_altq(u_int32_t);
116 int			 pf_commit_altq(u_int32_t);
117 int			 pf_enable_altq(struct pf_altq *);
118 int			 pf_disable_altq(struct pf_altq *);
119 #endif /* ALTQ */
120 int			 pf_begin_rules(u_int32_t *, int, const char *);
121 int			 pf_rollback_rules(u_int32_t, int, char *);
122 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
123 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
124 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
125 int			 pf_commit_rules(u_int32_t, int, char *);
126 
127 struct pf_rule		 pf_default_rule;
128 struct lock		 pf_consistency_lock;
129 #ifdef ALTQ
130 static int		 pf_altq_running;
131 #endif
132 
133 #define	TAGID_MAX	 50000
134 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
135 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
136 
137 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
138 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
139 #endif
140 u_int16_t		 tagname2tag(struct pf_tags *, char *);
141 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
142 void			 tag_unref(struct pf_tags *, u_int16_t);
143 int			 pf_rtlabel_add(struct pf_addr_wrap *);
144 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
145 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
146 
147 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
148 
149 static cdev_t	pf_dev;
150 
151 /*
152  * XXX - These are new and need to be checked when moveing to a new version
153  */
154 static void		 pf_clear_states(void);
155 static int		 pf_clear_tables(void);
156 static void		 pf_clear_srcnodes(void);
157 /*
158  * XXX - These are new and need to be checked when moveing to a new version
159  */
160 
161 /*
162  * Wrapper functions for pfil(9) hooks
163  */
164 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
165 		int dir);
166 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
167 		int dir);
168 #ifdef INET6
169 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
170 		int dir);
171 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
172 		int dir);
173 #endif
174 
175 static int 		 hook_pf(void);
176 static int 		 dehook_pf(void);
177 static int 		 shutdown_pf(void);
178 static int 		 pf_load(void);
179 static int 		 pf_unload(void);
180 
181 d_open_t	pfopen;
182 d_close_t	pfclose;
183 d_ioctl_t	pfioctl;
184 
185 static struct dev_ops pf_ops = {	    /* XXX convert to port model */
186 	{ PF_NAME, 73, 0 },
187 	.d_open =	pfopen,
188 	.d_close =	pfclose,
189 	.d_ioctl =	pfioctl
190 };
191 
192 static volatile int pf_pfil_hooked = 0;
193 int pf_end_threads = 0;
194 struct lock pf_mod_lck;
195 
196 int debug_pfugidhack = 0;
197 SYSCTL_INT(_debug, OID_AUTO, pfugidhack, CTLFLAG_RW, &debug_pfugidhack, 0,
198 	"Enable/disable pf user/group rules mpsafe hack");
199 
200 void
201 init_zone_var(void)
202 {
203 	pf_src_tree_pl = pf_rule_pl = NULL;
204 	pf_state_pl = pf_altq_pl = pf_pooladdr_pl = NULL;
205 	pf_frent_pl = pf_frag_pl = pf_cache_pl = pf_cent_pl = NULL;
206 	pf_state_scrub_pl = NULL;
207 	pfr_ktable_pl = pfr_kentry_pl = NULL;
208 }
209 
210 void
211 cleanup_pf_zone(void)
212 {
213 	ZONE_DESTROY(pf_src_tree_pl);
214 	ZONE_DESTROY(pf_rule_pl);
215 	ZONE_DESTROY(pf_state_pl);
216 	ZONE_DESTROY(pf_altq_pl);
217 	ZONE_DESTROY(pf_pooladdr_pl);
218 	ZONE_DESTROY(pf_frent_pl);
219 	ZONE_DESTROY(pf_frag_pl);
220 	ZONE_DESTROY(pf_cache_pl);
221 	ZONE_DESTROY(pf_cent_pl);
222 	ZONE_DESTROY(pfr_ktable_pl);
223 	ZONE_DESTROY(pfr_kentry_pl);
224 	ZONE_DESTROY(pf_state_scrub_pl);
225 	ZONE_DESTROY(pfi_addr_pl);
226 }
227 
228 int
229 pfattach(void)
230 {
231 	u_int32_t *my_timeout = pf_default_rule.timeout;
232 	int error = 1;
233 
234 	do {
235 		ZONE_CREATE(pf_src_tree_pl,struct pf_src_node, "pfsrctrpl");
236 		ZONE_CREATE(pf_rule_pl,    struct pf_rule, "pfrulepl");
237 		ZONE_CREATE(pf_state_pl,   struct pf_state, "pfstatepl");
238 		ZONE_CREATE(pf_altq_pl,    struct pf_altq, "pfaltqpl");
239 		ZONE_CREATE(pf_pooladdr_pl,struct pf_pooladdr, "pfpooladdrpl");
240 		ZONE_CREATE(pfr_ktable_pl, struct pfr_ktable, "pfrktable");
241 		ZONE_CREATE(pfr_kentry_pl, struct pfr_kentry, "pfrkentry");
242 		ZONE_CREATE(pf_frent_pl,   struct pf_frent, "pffrent");
243 		ZONE_CREATE(pf_frag_pl,    struct pf_fragment, "pffrag");
244 		ZONE_CREATE(pf_cache_pl,   struct pf_fragment, "pffrcache");
245 		ZONE_CREATE(pf_cent_pl,    struct pf_frcache, "pffrcent");
246 		ZONE_CREATE(pf_state_scrub_pl, struct pf_state_scrub,
247 		    "pfstatescrub");
248 		ZONE_CREATE(pfi_addr_pl,   struct pfi_dynaddr, "pfiaddrpl");
249 		error = 0;
250 	} while(0);
251 	if (error) {
252 		cleanup_pf_zone();
253 		return (error);
254 	}
255 	pfr_initialize();
256 	pfi_initialize();
257 	error = pf_osfp_initialize();
258 	if (error) {
259 		cleanup_pf_zone();
260 		pf_osfp_cleanup();
261 		return (error);
262 	}
263 
264 	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
265 	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
266 	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
267 	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
268 	/* XXX uma_zone_set_max(pf_pool_limits[PF_LIMIT_STATES].pp,
269 		pf_pool_limits[PF_LIMIT_STATES].limit);
270 	*/
271 	if (ctob(physmem) <= 100*1024*1024)
272 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
273 		    PFR_KENTRY_HIWAT_SMALL;
274 	RB_INIT(&tree_src_tracking);
275 	RB_INIT(&pf_anchors);
276 	pf_init_ruleset(&pf_main_ruleset);
277 	TAILQ_INIT(&pf_altqs[0]);
278 	TAILQ_INIT(&pf_altqs[1]);
279 	TAILQ_INIT(&pf_pabuf);
280 	pf_altqs_active = &pf_altqs[0];
281 	pf_altqs_inactive = &pf_altqs[1];
282 	TAILQ_INIT(&state_list);
283 
284 	/* default rule should never be garbage collected */
285 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
286 	pf_default_rule.action = PF_PASS;
287 	pf_default_rule.nr = (uint32_t)(-1);
288 	pf_default_rule.rtableid = -1;
289 
290 	/* initialize default timeouts */
291 	my_timeout[PFTM_TCP_FIRST_PACKET] = 120;	/* First TCP packet */
292 	my_timeout[PFTM_TCP_OPENING] = 30; 		/* No response yet */
293 	my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;	/* Established */
294 	my_timeout[PFTM_TCP_CLOSING] = 15 * 60;		/* Half closed */
295 	my_timeout[PFTM_TCP_FIN_WAIT] = 45;		/* Got both FINs */
296 	my_timeout[PFTM_TCP_CLOSED] = 90;		/* Got a RST */
297 	my_timeout[PFTM_UDP_FIRST_PACKET] = 60;		/* First UDP packet */
298 	my_timeout[PFTM_UDP_SINGLE] = 30;		/* Unidirectional */
299 	my_timeout[PFTM_UDP_MULTIPLE] = 60;		/* Bidirectional */
300 	my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;	/* First ICMP packet */
301 	my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;		/* Got error response */
302 	my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;	/* First packet */
303 	my_timeout[PFTM_OTHER_SINGLE] = 30;		/* Unidirectional */
304 	my_timeout[PFTM_OTHER_MULTIPLE] = 60;		/* Bidirectional */
305 	my_timeout[PFTM_FRAG] = 30;			/* Fragment expire */
306 	my_timeout[PFTM_INTERVAL] = 10;			/* Expire interval */
307 	my_timeout[PFTM_SRC_NODE] = 0;		/* Source Tracking */
308 	my_timeout[PFTM_TS_DIFF] = 30;		/* Allowed TS diff */
309 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
310 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
311 
312 	pf_normalize_init();
313 	bzero(&pf_status, sizeof(pf_status));
314 	pf_status.debug = PF_DEBUG_URGENT;
315 
316 	/* XXX do our best to avoid a conflict */
317 	pf_status.hostid = karc4random();
318 
319 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
320 		panic("pfpurge thread");
321 
322 	return (error);
323 }
324 
325 int
326 pfopen(struct dev_open_args *ap)
327 {
328 	cdev_t dev = ap->a_head.a_dev;
329 	if (minor(dev) >= 1)
330 		return (ENXIO);
331 	return (0);
332 }
333 
334 int
335 pfclose(struct dev_close_args *ap)
336 {
337 	cdev_t dev = ap->a_head.a_dev;
338 	if (minor(dev) >= 1)
339 		return (ENXIO);
340 	return (0);
341 }
342 
343 struct pf_pool *
344 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
345     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
346     u_int8_t check_ticket)
347 {
348 	struct pf_ruleset	*ruleset;
349 	struct pf_rule		*rule;
350 	int			 rs_num;
351 
352 	ruleset = pf_find_ruleset(anchor);
353 	if (ruleset == NULL)
354 		return (NULL);
355 	rs_num = pf_get_ruleset_number(rule_action);
356 	if (rs_num >= PF_RULESET_MAX)
357 		return (NULL);
358 	if (active) {
359 		if (check_ticket && ticket !=
360 		    ruleset->rules[rs_num].active.ticket)
361 			return (NULL);
362 		if (r_last)
363 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
364 			    pf_rulequeue);
365 		else
366 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
367 	} else {
368 		if (check_ticket && ticket !=
369 		    ruleset->rules[rs_num].inactive.ticket)
370 			return (NULL);
371 		if (r_last)
372 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
373 			    pf_rulequeue);
374 		else
375 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
376 	}
377 	if (!r_last) {
378 		while ((rule != NULL) && (rule->nr != rule_number))
379 			rule = TAILQ_NEXT(rule, entries);
380 	}
381 	if (rule == NULL)
382 		return (NULL);
383 
384 	return (&rule->rpool);
385 }
386 
387 void
388 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
389 {
390 	struct pf_pooladdr	*mv_pool_pa;
391 
392 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
393 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
394 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
395 	}
396 }
397 
398 void
399 pf_empty_pool(struct pf_palist *poola)
400 {
401 	struct pf_pooladdr	*empty_pool_pa;
402 
403 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
404 		pfi_dynaddr_remove(&empty_pool_pa->addr);
405 		pf_tbladdr_remove(&empty_pool_pa->addr);
406 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
407 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
408 		pool_put(&pf_pooladdr_pl, empty_pool_pa);
409 	}
410 }
411 
412 void
413 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
414 {
415 	if (rulequeue != NULL) {
416 		if (rule->states <= 0) {
417 			/*
418 			 * XXX - we need to remove the table *before* detaching
419 			 * the rule to make sure the table code does not delete
420 			 * the anchor under our feet.
421 			 */
422 			pf_tbladdr_remove(&rule->src.addr);
423 			pf_tbladdr_remove(&rule->dst.addr);
424 			if (rule->overload_tbl)
425 				pfr_detach_table(rule->overload_tbl);
426 		}
427 		TAILQ_REMOVE(rulequeue, rule, entries);
428 		rule->entries.tqe_prev = NULL;
429 		rule->nr = -1;
430 	}
431 
432 	if (rule->states > 0 || rule->src_nodes > 0 ||
433 	    rule->entries.tqe_prev != NULL)
434 		return;
435 	pf_tag_unref(rule->tag);
436 	pf_tag_unref(rule->match_tag);
437 #ifdef ALTQ
438 	if (rule->pqid != rule->qid)
439 		pf_qid_unref(rule->pqid);
440 	pf_qid_unref(rule->qid);
441 #endif
442 	pf_rtlabel_remove(&rule->src.addr);
443 	pf_rtlabel_remove(&rule->dst.addr);
444 	pfi_dynaddr_remove(&rule->src.addr);
445 	pfi_dynaddr_remove(&rule->dst.addr);
446 	if (rulequeue == NULL) {
447 		pf_tbladdr_remove(&rule->src.addr);
448 		pf_tbladdr_remove(&rule->dst.addr);
449 		if (rule->overload_tbl)
450 			pfr_detach_table(rule->overload_tbl);
451 	}
452 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
453 	pf_anchor_remove(rule);
454 	pf_empty_pool(&rule->rpool.list);
455 	pool_put(&pf_rule_pl, rule);
456 }
457 
458 u_int16_t
459 tagname2tag(struct pf_tags *head, char *tagname)
460 {
461 	struct pf_tagname	*tag, *p = NULL;
462 	u_int16_t		 new_tagid = 1;
463 
464 	TAILQ_FOREACH(tag, head, entries)
465 		if (strcmp(tagname, tag->name) == 0) {
466 			tag->ref++;
467 			return (tag->tag);
468 		}
469 
470 	/*
471 	 * to avoid fragmentation, we do a linear search from the beginning
472 	 * and take the first free slot we find. if there is none or the list
473 	 * is empty, append a new entry at the end.
474 	 */
475 
476 	/* new entry */
477 	if (!TAILQ_EMPTY(head))
478 		for (p = TAILQ_FIRST(head); p != NULL &&
479 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
480 			new_tagid = p->tag + 1;
481 
482 	if (new_tagid > TAGID_MAX)
483 		return (0);
484 
485 	/* allocate and fill new struct pf_tagname */
486 	tag = kmalloc(sizeof(struct pf_tagname), M_TEMP, M_WAITOK);
487 	if (tag == NULL)
488 		return (0);
489 	bzero(tag, sizeof(struct pf_tagname));
490 	strlcpy(tag->name, tagname, sizeof(tag->name));
491 	tag->tag = new_tagid;
492 	tag->ref++;
493 
494 	if (p != NULL)	/* insert new entry before p */
495 		TAILQ_INSERT_BEFORE(p, tag, entries);
496 	else	/* either list empty or no free slot in between */
497 		TAILQ_INSERT_TAIL(head, tag, entries);
498 
499 	return (tag->tag);
500 }
501 
502 void
503 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
504 {
505 	struct pf_tagname	*tag;
506 
507 	TAILQ_FOREACH(tag, head, entries)
508 		if (tag->tag == tagid) {
509 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
510 			return;
511 		}
512 }
513 
514 void
515 tag_unref(struct pf_tags *head, u_int16_t tag)
516 {
517 	struct pf_tagname	*p, *next;
518 
519 	if (tag == 0)
520 		return;
521 
522 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
523 		next = TAILQ_NEXT(p, entries);
524 		if (tag == p->tag) {
525 			if (--p->ref == 0) {
526 				TAILQ_REMOVE(head, p, entries);
527 				kfree(p, M_TEMP);
528 			}
529 			break;
530 		}
531 	}
532 }
533 
534 u_int16_t
535 pf_tagname2tag(char *tagname)
536 {
537 	return (tagname2tag(&pf_tags, tagname));
538 }
539 
540 void
541 pf_tag2tagname(u_int16_t tagid, char *p)
542 {
543 	tag2tagname(&pf_tags, tagid, p);
544 }
545 
546 void
547 pf_tag_ref(u_int16_t tag)
548 {
549 	struct pf_tagname *t;
550 
551 	TAILQ_FOREACH(t, &pf_tags, entries)
552 		if (t->tag == tag)
553 			break;
554 	if (t != NULL)
555 		t->ref++;
556 }
557 
558 void
559 pf_tag_unref(u_int16_t tag)
560 {
561 	tag_unref(&pf_tags, tag);
562 }
563 
564 int
565 pf_rtlabel_add(struct pf_addr_wrap *a)
566 {
567 	return (0);
568 }
569 
570 void
571 pf_rtlabel_remove(struct pf_addr_wrap *a)
572 {
573 }
574 
575 void
576 pf_rtlabel_copyout(struct pf_addr_wrap *a)
577 {
578 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
579 		strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
580 }
581 
582 #ifdef ALTQ
583 u_int32_t
584 pf_qname2qid(char *qname)
585 {
586 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
587 }
588 
589 void
590 pf_qid2qname(u_int32_t qid, char *p)
591 {
592 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
593 }
594 
595 void
596 pf_qid_unref(u_int32_t qid)
597 {
598 	tag_unref(&pf_qids, (u_int16_t)qid);
599 }
600 
601 int
602 pf_begin_altq(u_int32_t *ticket)
603 {
604 	struct pf_altq	*altq;
605 	int		 error = 0;
606 
607 	/* Purge the old altq list */
608 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
609 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
610 		if (altq->qname[0] == 0) {
611 			/* detach and destroy the discipline */
612 			error = altq_remove(altq);
613 		} else
614 			pf_qid_unref(altq->qid);
615 		pool_put(&pf_altq_pl, altq);
616 	}
617 	if (error)
618 		return (error);
619 	*ticket = ++ticket_altqs_inactive;
620 	altqs_inactive_open = 1;
621 	return (0);
622 }
623 
624 int
625 pf_rollback_altq(u_int32_t ticket)
626 {
627 	struct pf_altq	*altq;
628 	int		 error = 0;
629 
630 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
631 		return (0);
632 	/* Purge the old altq list */
633 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
634 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
635 		if (altq->qname[0] == 0) {
636 			/* detach and destroy the discipline */
637 			error = altq_remove(altq);
638 		} else
639 			pf_qid_unref(altq->qid);
640 		pool_put(&pf_altq_pl, altq);
641 	}
642 	altqs_inactive_open = 0;
643 	return (error);
644 }
645 
646 int
647 pf_commit_altq(u_int32_t ticket)
648 {
649 	struct pf_altqqueue	*old_altqs;
650 	struct pf_altq		*altq;
651 	int			 err, error = 0;
652 
653 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
654 		return (EBUSY);
655 
656 	/* swap altqs, keep the old. */
657 	crit_enter();
658 	old_altqs = pf_altqs_active;
659 	pf_altqs_active = pf_altqs_inactive;
660 	pf_altqs_inactive = old_altqs;
661 	ticket_altqs_active = ticket_altqs_inactive;
662 
663 	/* Attach new disciplines */
664 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
665 		if (altq->qname[0] == 0) {
666 			/* attach the discipline */
667 			error = altq_pfattach(altq);
668 			if (error) {
669 				crit_exit();
670 				return (error);
671 			}
672 		}
673 	}
674 
675 	/* Purge the old altq list */
676 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
677 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
678 		if (altq->qname[0] == 0) {
679 			/* detach and destroy the discipline */
680 			if (pf_altq_running)
681 				error = pf_disable_altq(altq);
682 			err = altq_pfdetach(altq);
683 			if (err != 0 && error == 0)
684 				error = err;
685 			err = altq_remove(altq);
686 			if (err != 0 && error == 0)
687 				error = err;
688 		} else
689 			pf_qid_unref(altq->qid);
690 		pool_put(&pf_altq_pl, altq);
691 	}
692 	crit_exit();
693 
694 	altqs_inactive_open = 0;
695 	return (error);
696 }
697 
698 int
699 pf_enable_altq(struct pf_altq *altq)
700 {
701 	struct ifnet		*ifp;
702 	struct tb_profile	 tb;
703 	int			 error = 0;
704 
705 	if ((ifp = ifunit(altq->ifname)) == NULL)
706 		return (EINVAL);
707 
708 	if (ifp->if_snd.altq_type != ALTQT_NONE)
709 		error = altq_enable(&ifp->if_snd);
710 
711 	/* set tokenbucket regulator */
712 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
713 		tb.rate = altq->ifbandwidth;
714 		tb.depth = altq->tbrsize;
715 		crit_enter();
716 		error = tbr_set(&ifp->if_snd, &tb);
717 		crit_exit();
718 	}
719 
720 	return (error);
721 }
722 
723 int
724 pf_disable_altq(struct pf_altq *altq)
725 {
726 	struct ifnet		*ifp;
727 	struct tb_profile	 tb;
728 	int			 error;
729 
730 	if ((ifp = ifunit(altq->ifname)) == NULL)
731 		return (EINVAL);
732 
733 	/*
734 	 * when the discipline is no longer referenced, it was overridden
735 	 * by a new one.  if so, just return.
736 	 */
737 	if (altq->altq_disc != ifp->if_snd.altq_disc)
738 		return (0);
739 
740 	error = altq_disable(&ifp->if_snd);
741 
742 	if (error == 0) {
743 		/* clear tokenbucket regulator */
744 		tb.rate = 0;
745 		crit_enter();
746 		error = tbr_set(&ifp->if_snd, &tb);
747 		crit_exit();
748 	}
749 
750 	return (error);
751 }
752 #endif /* ALTQ */
753 
754 int
755 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
756 {
757 	struct pf_ruleset	*rs;
758 	struct pf_rule		*rule;
759 
760 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
761 		return (EINVAL);
762 	rs = pf_find_or_create_ruleset(anchor);
763 	if (rs == NULL)
764 		return (EINVAL);
765 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
766 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
767 		rs->rules[rs_num].inactive.rcount--;
768 	}
769 	*ticket = ++rs->rules[rs_num].inactive.ticket;
770 	rs->rules[rs_num].inactive.open = 1;
771 	return (0);
772 }
773 
774 int
775 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
776 {
777 	struct pf_ruleset	*rs;
778 	struct pf_rule		*rule;
779 
780 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
781 		return (EINVAL);
782 	rs = pf_find_ruleset(anchor);
783 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
784 	    rs->rules[rs_num].inactive.ticket != ticket)
785 		return (0);
786 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
787 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
788 		rs->rules[rs_num].inactive.rcount--;
789 	}
790 	rs->rules[rs_num].inactive.open = 0;
791 	return (0);
792 }
793 
794 #define PF_MD5_UPD(st, elm)						\
795 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
796 
797 #define PF_MD5_UPD_STR(st, elm)						\
798 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
799 
800 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
801 		(stor) = htonl((st)->elm);				\
802 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
803 } while (0)
804 
805 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
806 		(stor) = htons((st)->elm);				\
807 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
808 } while (0)
809 
810 void
811 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
812 {
813 	PF_MD5_UPD(pfr, addr.type);
814 	switch (pfr->addr.type) {
815 		case PF_ADDR_DYNIFTL:
816 			PF_MD5_UPD(pfr, addr.v.ifname);
817 			PF_MD5_UPD(pfr, addr.iflags);
818 			break;
819 		case PF_ADDR_TABLE:
820 			PF_MD5_UPD(pfr, addr.v.tblname);
821 			break;
822 		case PF_ADDR_ADDRMASK:
823 			/* XXX ignore af? */
824 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
825 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
826 			break;
827 		case PF_ADDR_RTLABEL:
828 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
829 			break;
830 	}
831 
832 	PF_MD5_UPD(pfr, port[0]);
833 	PF_MD5_UPD(pfr, port[1]);
834 	PF_MD5_UPD(pfr, neg);
835 	PF_MD5_UPD(pfr, port_op);
836 }
837 
838 void
839 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
840 {
841 	u_int16_t x;
842 	u_int32_t y;
843 
844 	pf_hash_rule_addr(ctx, &rule->src);
845 	pf_hash_rule_addr(ctx, &rule->dst);
846 	PF_MD5_UPD_STR(rule, label);
847 	PF_MD5_UPD_STR(rule, ifname);
848 	PF_MD5_UPD_STR(rule, match_tagname);
849 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
850 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
851 	PF_MD5_UPD_HTONL(rule, prob, y);
852 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
853 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
854 	PF_MD5_UPD(rule, uid.op);
855 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
856 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
857 	PF_MD5_UPD(rule, gid.op);
858 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
859 	PF_MD5_UPD(rule, action);
860 	PF_MD5_UPD(rule, direction);
861 	PF_MD5_UPD(rule, af);
862 	PF_MD5_UPD(rule, quick);
863 	PF_MD5_UPD(rule, ifnot);
864 	PF_MD5_UPD(rule, match_tag_not);
865 	PF_MD5_UPD(rule, natpass);
866 	PF_MD5_UPD(rule, keep_state);
867 	PF_MD5_UPD(rule, proto);
868 	PF_MD5_UPD(rule, type);
869 	PF_MD5_UPD(rule, code);
870 	PF_MD5_UPD(rule, flags);
871 	PF_MD5_UPD(rule, flagset);
872 	PF_MD5_UPD(rule, allow_opts);
873 	PF_MD5_UPD(rule, rt);
874 	PF_MD5_UPD(rule, tos);
875 }
876 
877 int
878 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
879 {
880 	struct pf_ruleset	*rs;
881 	struct pf_rule		*rule, **old_array;
882 	struct pf_rulequeue	*old_rules;
883 	int			 error;
884 	u_int32_t		 old_rcount;
885 
886 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
887 		return (EINVAL);
888 	rs = pf_find_ruleset(anchor);
889 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
890 	    ticket != rs->rules[rs_num].inactive.ticket)
891 		return (EBUSY);
892 
893 	/* Calculate checksum for the main ruleset */
894 	if (rs == &pf_main_ruleset) {
895 		error = pf_setup_pfsync_matching(rs);
896 		if (error != 0)
897 			return (error);
898 	}
899 
900 	/* Swap rules, keep the old. */
901 	crit_enter();
902 	old_rules = rs->rules[rs_num].active.ptr;
903 	old_rcount = rs->rules[rs_num].active.rcount;
904 	old_array = rs->rules[rs_num].active.ptr_array;
905 
906 	rs->rules[rs_num].active.ptr =
907 	    rs->rules[rs_num].inactive.ptr;
908 	rs->rules[rs_num].active.ptr_array =
909 	    rs->rules[rs_num].inactive.ptr_array;
910 	rs->rules[rs_num].active.rcount =
911 	    rs->rules[rs_num].inactive.rcount;
912 	rs->rules[rs_num].inactive.ptr = old_rules;
913 	rs->rules[rs_num].inactive.ptr_array = old_array;
914 	rs->rules[rs_num].inactive.rcount = old_rcount;
915 
916 	rs->rules[rs_num].active.ticket =
917 	    rs->rules[rs_num].inactive.ticket;
918 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
919 
920 
921 	/* Purge the old rule list. */
922 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
923 		pf_rm_rule(old_rules, rule);
924 	if (rs->rules[rs_num].inactive.ptr_array)
925 		kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
926 	rs->rules[rs_num].inactive.ptr_array = NULL;
927 	rs->rules[rs_num].inactive.rcount = 0;
928 	rs->rules[rs_num].inactive.open = 0;
929 	pf_remove_if_empty_ruleset(rs);
930 	crit_exit();
931 	return (0);
932 }
933 
934 int
935 pf_setup_pfsync_matching(struct pf_ruleset *rs)
936 {
937 	MD5_CTX			 ctx;
938 	struct pf_rule		*rule;
939 	int			 rs_cnt;
940 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
941 
942 	MD5Init(&ctx);
943 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
944 		/* XXX PF_RULESET_SCRUB as well? */
945 		if (rs_cnt == PF_RULESET_SCRUB)
946 			continue;
947 
948 		if (rs->rules[rs_cnt].inactive.ptr_array)
949 			kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
950 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
951 
952 		if (rs->rules[rs_cnt].inactive.rcount) {
953 			rs->rules[rs_cnt].inactive.ptr_array =
954 			    kmalloc(sizeof(caddr_t) *
955 				    rs->rules[rs_cnt].inactive.rcount,
956 				    M_TEMP, M_WAITOK);
957 
958 			if (!rs->rules[rs_cnt].inactive.ptr_array)
959 				return (ENOMEM);
960 		}
961 
962 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
963 		    entries) {
964 			pf_hash_rule(&ctx, rule);
965 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
966 		}
967 	}
968 
969 	MD5Final(digest, &ctx);
970 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
971 	return (0);
972 }
973 
974 int
975 pfioctl(struct dev_ioctl_args *ap)
976 {
977 	u_long cmd = ap->a_cmd;
978 	caddr_t addr = ap->a_data;
979 	struct pf_pooladdr	*pa = NULL;
980 	struct pf_pool		*pool = NULL;
981 	int			 error = 0;
982 
983 	/* XXX keep in sync with switch() below */
984 	if (securelevel > 1)
985 		switch (cmd) {
986 		case DIOCGETRULES:
987 		case DIOCGETRULE:
988 		case DIOCGETADDRS:
989 		case DIOCGETADDR:
990 		case DIOCGETSTATE:
991 		case DIOCSETSTATUSIF:
992 		case DIOCGETSTATUS:
993 		case DIOCCLRSTATUS:
994 		case DIOCNATLOOK:
995 		case DIOCSETDEBUG:
996 		case DIOCGETSTATES:
997 		case DIOCGETTIMEOUT:
998 		case DIOCCLRRULECTRS:
999 		case DIOCGETLIMIT:
1000 		case DIOCGETALTQS:
1001 		case DIOCGETALTQ:
1002 		case DIOCGETQSTATS:
1003 		case DIOCGETRULESETS:
1004 		case DIOCGETRULESET:
1005 		case DIOCRGETTABLES:
1006 		case DIOCRGETTSTATS:
1007 		case DIOCRCLRTSTATS:
1008 		case DIOCRCLRADDRS:
1009 		case DIOCRADDADDRS:
1010 		case DIOCRDELADDRS:
1011 		case DIOCRSETADDRS:
1012 		case DIOCRGETADDRS:
1013 		case DIOCRGETASTATS:
1014 		case DIOCRCLRASTATS:
1015 		case DIOCRTSTADDRS:
1016 		case DIOCOSFPGET:
1017 		case DIOCGETSRCNODES:
1018 		case DIOCCLRSRCNODES:
1019 		case DIOCIGETIFACES:
1020 		case DIOCSETIFFLAG:
1021 		case DIOCCLRIFFLAG:
1022 		case DIOCGIFSPEED:
1023 			break;
1024 		case DIOCRCLRTABLES:
1025 		case DIOCRADDTABLES:
1026 		case DIOCRDELTABLES:
1027 		case DIOCRSETTFLAGS:
1028 			if (((struct pfioc_table *)addr)->pfrio_flags &
1029 			    PFR_FLAG_DUMMY)
1030 				break; /* dummy operation ok */
1031 			return (EPERM);
1032 		default:
1033 			return (EPERM);
1034 		}
1035 
1036 	if (!(ap->a_fflag & FWRITE))
1037 		switch (cmd) {
1038 		case DIOCGETRULES:
1039 		case DIOCGETADDRS:
1040 		case DIOCGETADDR:
1041 		case DIOCGETSTATE:
1042 		case DIOCGETSTATUS:
1043 		case DIOCGETSTATES:
1044 		case DIOCGETTIMEOUT:
1045 		case DIOCGETLIMIT:
1046 		case DIOCGETALTQS:
1047 		case DIOCGETALTQ:
1048 		case DIOCGETQSTATS:
1049 		case DIOCGETRULESETS:
1050 		case DIOCGETRULESET:
1051 		case DIOCNATLOOK:
1052 		case DIOCRGETTABLES:
1053 		case DIOCRGETTSTATS:
1054 		case DIOCRGETADDRS:
1055 		case DIOCRGETASTATS:
1056 		case DIOCRTSTADDRS:
1057 		case DIOCOSFPGET:
1058 		case DIOCGETSRCNODES:
1059 		case DIOCIGETIFACES:
1060 		case DIOCGIFSPEED:
1061 			break;
1062 		case DIOCRCLRTABLES:
1063 		case DIOCRADDTABLES:
1064 		case DIOCRDELTABLES:
1065 		case DIOCRCLRTSTATS:
1066 		case DIOCRCLRADDRS:
1067 		case DIOCRADDADDRS:
1068 		case DIOCRDELADDRS:
1069 		case DIOCRSETADDRS:
1070 		case DIOCRSETTFLAGS:
1071 			if (((struct pfioc_table *)addr)->pfrio_flags &
1072 			    PFR_FLAG_DUMMY)
1073 				break; /* dummy operation ok */
1074 			return (EACCES);
1075 		case DIOCGETRULE:
1076 			if (((struct pfioc_rule *)addr)->action == PF_GET_CLR_CNTR)
1077 				return (EACCES);
1078 			break;
1079 		default:
1080 			return (EACCES);
1081 		}
1082 
1083 	switch (cmd) {
1084 
1085 	case DIOCSTART:
1086 		if (pf_status.running)
1087 			error = EEXIST;
1088 		else {
1089 			error = hook_pf();
1090 			if (error) {
1091 				DPFPRINTF(PF_DEBUG_MISC,
1092 				    ("pf: pfil registration fail\n"));
1093 				break;
1094 			}
1095 			pf_status.running = 1;
1096 			pf_status.since = time_second;
1097 			if (pf_status.stateid == 0) {
1098 				pf_status.stateid = time_second;
1099 				pf_status.stateid = pf_status.stateid << 32;
1100 			}
1101 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1102 		}
1103 		break;
1104 
1105 	case DIOCSTOP:
1106 		if (!pf_status.running)
1107 			error = ENOENT;
1108 		else {
1109 			pf_status.running = 0;
1110 			error = dehook_pf();
1111 			if (error) {
1112 				pf_status.running = 1;
1113 				DPFPRINTF(PF_DEBUG_MISC,
1114 					("pf: pfil unregistration failed\n"));
1115 			}
1116 			pf_status.since = time_second;
1117 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1118 		}
1119 		break;
1120 
1121 	case DIOCADDRULE: {
1122 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1123 		struct pf_ruleset	*ruleset;
1124 		struct pf_rule		*rule, *tail;
1125 		struct pf_pooladdr	*pa;
1126 		int			 rs_num;
1127 
1128 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1129 		ruleset = pf_find_ruleset(pr->anchor);
1130 		if (ruleset == NULL) {
1131 			error = EINVAL;
1132 			break;
1133 		}
1134 		rs_num = pf_get_ruleset_number(pr->rule.action);
1135 		if (rs_num >= PF_RULESET_MAX) {
1136 			error = EINVAL;
1137 			break;
1138 		}
1139 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1140 			error = EINVAL;
1141 			break;
1142 		}
1143 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1144 			error = EBUSY;
1145 			break;
1146 		}
1147 		if (pr->pool_ticket != ticket_pabuf) {
1148 			error = EBUSY;
1149 			break;
1150 		}
1151 		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1152 		if (rule == NULL) {
1153 			error = ENOMEM;
1154 			break;
1155 		}
1156 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1157 		rule->cuid = ap->a_cred->cr_ruid;
1158 		rule->cpid = (int)NULL;
1159 		rule->anchor = NULL;
1160 		rule->kif = NULL;
1161 		TAILQ_INIT(&rule->rpool.list);
1162 		/* initialize refcounting */
1163 		rule->states = 0;
1164 		rule->src_nodes = 0;
1165 		rule->entries.tqe_prev = NULL;
1166 #ifndef INET
1167 		if (rule->af == AF_INET) {
1168 			pool_put(&pf_rule_pl, rule);
1169 			error = EAFNOSUPPORT;
1170 			break;
1171 		}
1172 #endif /* INET */
1173 #ifndef INET6
1174 		if (rule->af == AF_INET6) {
1175 			pool_put(&pf_rule_pl, rule);
1176 			error = EAFNOSUPPORT;
1177 			break;
1178 		}
1179 #endif /* INET6 */
1180 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1181 		    pf_rulequeue);
1182 		if (tail)
1183 			rule->nr = tail->nr + 1;
1184 		else
1185 			rule->nr = 0;
1186 		if (rule->ifname[0]) {
1187 			rule->kif = pfi_kif_get(rule->ifname);
1188 			if (rule->kif == NULL) {
1189 				pool_put(&pf_rule_pl, rule);
1190 				error = EINVAL;
1191 				break;
1192 			}
1193 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1194 		}
1195 
1196 		if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1197 			error = EBUSY;
1198 
1199 #ifdef ALTQ
1200 		/* set queue IDs */
1201 		if (rule->qname[0] != 0) {
1202 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1203 				error = EBUSY;
1204 			else if (rule->pqname[0] != 0) {
1205 				if ((rule->pqid =
1206 				    pf_qname2qid(rule->pqname)) == 0)
1207 					error = EBUSY;
1208 			} else
1209 				rule->pqid = rule->qid;
1210 		}
1211 #endif
1212 		if (rule->tagname[0])
1213 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1214 				error = EBUSY;
1215 		if (rule->match_tagname[0])
1216 			if ((rule->match_tag =
1217 			    pf_tagname2tag(rule->match_tagname)) == 0)
1218 				error = EBUSY;
1219 		if (rule->rt && !rule->direction)
1220 			error = EINVAL;
1221 #if NPFLOG > 0
1222 		if (rule->logif >= PFLOGIFS_MAX)
1223 			error = EINVAL;
1224 #endif
1225 		if (pf_rtlabel_add(&rule->src.addr) ||
1226 		    pf_rtlabel_add(&rule->dst.addr))
1227 			error = EBUSY;
1228 		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1229 			error = EINVAL;
1230 		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1231 			error = EINVAL;
1232 		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1233 			error = EINVAL;
1234 		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1235 			error = EINVAL;
1236 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1237 			error = EINVAL;
1238 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1239 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1240 				error = EINVAL;
1241 
1242 		if (rule->overload_tblname[0]) {
1243 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1244 			    rule->overload_tblname)) == NULL)
1245 				error = EINVAL;
1246 			else
1247 				rule->overload_tbl->pfrkt_flags |=
1248 				    PFR_TFLAG_ACTIVE;
1249 		}
1250 
1251 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1252 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1253 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1254 		    (rule->rt > PF_FASTROUTE)) &&
1255 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1256 			error = EINVAL;
1257 
1258 		if (error) {
1259 			pf_rm_rule(NULL, rule);
1260 			break;
1261 		}
1262 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1263 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1264 		    rule->bytes[0] = rule->bytes[1] = 0;
1265 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1266 		    rule, entries);
1267 		ruleset->rules[rs_num].inactive.rcount++;
1268 		break;
1269 	}
1270 
1271 	case DIOCGETRULES: {
1272 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1273 		struct pf_ruleset	*ruleset;
1274 		struct pf_rule		*tail;
1275 		int			 rs_num;
1276 
1277 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1278 		ruleset = pf_find_ruleset(pr->anchor);
1279 		if (ruleset == NULL) {
1280 			error = EINVAL;
1281 			break;
1282 		}
1283 		rs_num = pf_get_ruleset_number(pr->rule.action);
1284 		if (rs_num >= PF_RULESET_MAX) {
1285 			error = EINVAL;
1286 			break;
1287 		}
1288 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1289 		    pf_rulequeue);
1290 		if (tail)
1291 			pr->nr = tail->nr + 1;
1292 		else
1293 			pr->nr = 0;
1294 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1295 		break;
1296 	}
1297 
1298 	case DIOCGETRULE: {
1299 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1300 		struct pf_ruleset	*ruleset;
1301 		struct pf_rule		*rule;
1302 		int			 rs_num, i;
1303 
1304 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1305 		ruleset = pf_find_ruleset(pr->anchor);
1306 		if (ruleset == NULL) {
1307 			error = EINVAL;
1308 			break;
1309 		}
1310 		rs_num = pf_get_ruleset_number(pr->rule.action);
1311 		if (rs_num >= PF_RULESET_MAX) {
1312 			error = EINVAL;
1313 			break;
1314 		}
1315 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1316 			error = EBUSY;
1317 			break;
1318 		}
1319 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1320 		while ((rule != NULL) && (rule->nr != pr->nr))
1321 			rule = TAILQ_NEXT(rule, entries);
1322 		if (rule == NULL) {
1323 			error = EBUSY;
1324 			break;
1325 		}
1326 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1327 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1328 			error = EBUSY;
1329 			break;
1330 		}
1331 		pfi_dynaddr_copyout(&pr->rule.src.addr);
1332 		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1333 		pf_tbladdr_copyout(&pr->rule.src.addr);
1334 		pf_tbladdr_copyout(&pr->rule.dst.addr);
1335 		pf_rtlabel_copyout(&pr->rule.src.addr);
1336 		pf_rtlabel_copyout(&pr->rule.dst.addr);
1337 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1338 			if (rule->skip[i].ptr == NULL)
1339 				pr->rule.skip[i].nr = (uint32_t)(-1);
1340 			else
1341 				pr->rule.skip[i].nr =
1342 				    rule->skip[i].ptr->nr;
1343 
1344 		if (pr->action == PF_GET_CLR_CNTR) {
1345 			rule->evaluations = 0;
1346 			rule->packets[0] = rule->packets[1] = 0;
1347 			rule->bytes[0] = rule->bytes[1] = 0;
1348 		}
1349 		break;
1350 	}
1351 
1352 	case DIOCCHANGERULE: {
1353 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1354 		struct pf_ruleset	*ruleset;
1355 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1356 		u_int32_t		 nr = 0;
1357 		int			 rs_num;
1358 
1359 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1360 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1361 		    pcr->pool_ticket != ticket_pabuf) {
1362 			error = EBUSY;
1363 			break;
1364 		}
1365 
1366 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1367 		    pcr->action > PF_CHANGE_GET_TICKET) {
1368 			error = EINVAL;
1369 			break;
1370 		}
1371 		ruleset = pf_find_ruleset(pcr->anchor);
1372 		if (ruleset == NULL) {
1373 			error = EINVAL;
1374 			break;
1375 		}
1376 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1377 		if (rs_num >= PF_RULESET_MAX) {
1378 			error = EINVAL;
1379 			break;
1380 		}
1381 
1382 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1383 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1384 			break;
1385 		} else {
1386 			if (pcr->ticket !=
1387 			    ruleset->rules[rs_num].active.ticket) {
1388 				error = EINVAL;
1389 				break;
1390 			}
1391 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1392 				error = EINVAL;
1393 				break;
1394 			}
1395 		}
1396 
1397 		if (pcr->action != PF_CHANGE_REMOVE) {
1398 			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1399 			if (newrule == NULL) {
1400 				error = ENOMEM;
1401 				break;
1402 			}
1403 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1404 			newrule->cuid = ap->a_cred->cr_ruid;
1405 			newrule->cpid = (int)NULL;
1406 			TAILQ_INIT(&newrule->rpool.list);
1407 			/* initialize refcounting */
1408 			newrule->states = 0;
1409 			newrule->entries.tqe_prev = NULL;
1410 #ifndef INET
1411 			if (newrule->af == AF_INET) {
1412 				pool_put(&pf_rule_pl, newrule);
1413 				error = EAFNOSUPPORT;
1414 				break;
1415 			}
1416 #endif /* INET */
1417 #ifndef INET6
1418 			if (newrule->af == AF_INET6) {
1419 				pool_put(&pf_rule_pl, newrule);
1420 				error = EAFNOSUPPORT;
1421 				break;
1422 			}
1423 #endif /* INET6 */
1424 			if (newrule->ifname[0]) {
1425 				newrule->kif = pfi_kif_get(newrule->ifname);
1426 				if (newrule->kif == NULL) {
1427 					pool_put(&pf_rule_pl, newrule);
1428 					error = EINVAL;
1429 					break;
1430 				}
1431 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1432 			} else
1433 				newrule->kif = NULL;
1434 
1435 			if (newrule->rtableid > 0 &&
1436 			    newrule->rtableid > rt_numfibs)
1437 				error = EBUSY;
1438 
1439 #ifdef ALTQ
1440 			/* set queue IDs */
1441 			if (newrule->qname[0] != 0) {
1442 				if ((newrule->qid =
1443 				    pf_qname2qid(newrule->qname)) == 0)
1444 					error = EBUSY;
1445 				else if (newrule->pqname[0] != 0) {
1446 					if ((newrule->pqid =
1447 					    pf_qname2qid(newrule->pqname)) == 0)
1448 						error = EBUSY;
1449 				} else
1450 					newrule->pqid = newrule->qid;
1451 			}
1452 #endif /* ALTQ */
1453 			if (newrule->tagname[0])
1454 				if ((newrule->tag =
1455 				    pf_tagname2tag(newrule->tagname)) == 0)
1456 					error = EBUSY;
1457 			if (newrule->match_tagname[0])
1458 				if ((newrule->match_tag = pf_tagname2tag(
1459 				    newrule->match_tagname)) == 0)
1460 					error = EBUSY;
1461 			if (newrule->rt && !newrule->direction)
1462 				error = EINVAL;
1463 			if (pf_rtlabel_add(&newrule->src.addr) ||
1464 			    pf_rtlabel_add(&newrule->dst.addr))
1465 				error = EBUSY;
1466 			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1467 				error = EINVAL;
1468 			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1469 				error = EINVAL;
1470 			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1471 				error = EINVAL;
1472 			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1473 				error = EINVAL;
1474 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1475 				error = EINVAL;
1476 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1477 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1478 					error = EINVAL;
1479 
1480 			if (newrule->overload_tblname[0]) {
1481 				if ((newrule->overload_tbl = pfr_attach_table(
1482 				    ruleset, newrule->overload_tblname)) ==
1483 				    NULL)
1484 					error = EINVAL;
1485 				else
1486 					newrule->overload_tbl->pfrkt_flags |=
1487 					    PFR_TFLAG_ACTIVE;
1488 			}
1489 
1490 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1491 			if (((((newrule->action == PF_NAT) ||
1492 			    (newrule->action == PF_RDR) ||
1493 			    (newrule->action == PF_BINAT) ||
1494 			    (newrule->rt > PF_FASTROUTE)) &&
1495 			    !newrule->anchor)) &&
1496 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1497 				error = EINVAL;
1498 
1499 			if (error) {
1500 				pf_rm_rule(NULL, newrule);
1501 				break;
1502 			}
1503 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1504 			newrule->evaluations = 0;
1505 			newrule->packets[0] = newrule->packets[1] = 0;
1506 			newrule->bytes[0] = newrule->bytes[1] = 0;
1507 		}
1508 		pf_empty_pool(&pf_pabuf);
1509 
1510 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1511 			oldrule = TAILQ_FIRST(
1512 			    ruleset->rules[rs_num].active.ptr);
1513 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1514 			oldrule = TAILQ_LAST(
1515 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1516 		else {
1517 			oldrule = TAILQ_FIRST(
1518 			    ruleset->rules[rs_num].active.ptr);
1519 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1520 				oldrule = TAILQ_NEXT(oldrule, entries);
1521 			if (oldrule == NULL) {
1522 				if (newrule != NULL)
1523 					pf_rm_rule(NULL, newrule);
1524 				error = EINVAL;
1525 				break;
1526 			}
1527 		}
1528 
1529 		if (pcr->action == PF_CHANGE_REMOVE) {
1530 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1531 			ruleset->rules[rs_num].active.rcount--;
1532 		} else {
1533 			if (oldrule == NULL)
1534 				TAILQ_INSERT_TAIL(
1535 				    ruleset->rules[rs_num].active.ptr,
1536 				    newrule, entries);
1537 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1538 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1539 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1540 			else
1541 				TAILQ_INSERT_AFTER(
1542 				    ruleset->rules[rs_num].active.ptr,
1543 				    oldrule, newrule, entries);
1544 			ruleset->rules[rs_num].active.rcount++;
1545 		}
1546 
1547 		nr = 0;
1548 		TAILQ_FOREACH(oldrule,
1549 		    ruleset->rules[rs_num].active.ptr, entries)
1550 			oldrule->nr = nr++;
1551 
1552 		ruleset->rules[rs_num].active.ticket++;
1553 
1554 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1555 		pf_remove_if_empty_ruleset(ruleset);
1556 
1557 		break;
1558 	}
1559 
1560 	case DIOCCLRSTATES: {
1561 		struct pf_state		*state, *nexts;
1562 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1563 		int			 killed = 0;
1564 
1565 		for (state = RB_MIN(pf_state_tree_id, &tree_id); state;
1566 		    state = nexts) {
1567 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, state);
1568 
1569 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1570 			    state->u.s.kif->pfik_name)) {
1571 #if NPFSYNC
1572 				/* don't send out individual delete messages */
1573 				state->sync_flags = PFSTATE_NOSYNC;
1574 #endif
1575 				pf_unlink_state(state);
1576 				killed++;
1577 			}
1578 		}
1579 		psk->psk_af = killed;
1580 #if NPFSYNC
1581 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1582 #endif
1583 		break;
1584 	}
1585 
1586 	case DIOCKILLSTATES: {
1587 		struct pf_state		*state, *nexts;
1588 		struct pf_state_host	*src, *dst;
1589 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1590 		int			 killed = 0;
1591 
1592 		for (state = RB_MIN(pf_state_tree_id, &tree_id); state;
1593 		    state = nexts) {
1594 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, state);
1595 
1596 			if (state->direction == PF_OUT) {
1597 				src = &state->lan;
1598 				dst = &state->ext;
1599 			} else {
1600 				src = &state->ext;
1601 				dst = &state->lan;
1602 			}
1603 			if ((!psk->psk_af || state->af == psk->psk_af)
1604 			    && (!psk->psk_proto || psk->psk_proto ==
1605 			    state->proto) &&
1606 			    PF_MATCHA(psk->psk_src.neg,
1607 			    &psk->psk_src.addr.v.a.addr,
1608 			    &psk->psk_src.addr.v.a.mask,
1609 			    &src->addr, state->af) &&
1610 			    PF_MATCHA(psk->psk_dst.neg,
1611 			    &psk->psk_dst.addr.v.a.addr,
1612 			    &psk->psk_dst.addr.v.a.mask,
1613 			    &dst->addr, state->af) &&
1614 			    (psk->psk_src.port_op == 0 ||
1615 			    pf_match_port(psk->psk_src.port_op,
1616 			    psk->psk_src.port[0], psk->psk_src.port[1],
1617 			    src->port)) &&
1618 			    (psk->psk_dst.port_op == 0 ||
1619 			    pf_match_port(psk->psk_dst.port_op,
1620 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1621 			    dst->port)) &&
1622 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1623 			    state->u.s.kif->pfik_name))) {
1624 #if NPFSYNC > 0
1625 				/* send immediate delete of state */
1626 				pfsync_delete_state(state);
1627 				state->sync_flags |= PFSTATE_NOSYNC;
1628 #endif
1629 				pf_unlink_state(state);
1630 				killed++;
1631 			}
1632 		}
1633 		psk->psk_af = killed;
1634 		break;
1635 	}
1636 
1637 	case DIOCADDSTATE: {
1638 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1639 		struct pf_state		*state;
1640 		struct pfi_kif		*kif;
1641 
1642 		if (ps->state.timeout >= PFTM_MAX &&
1643 		    ps->state.timeout != PFTM_UNTIL_PACKET) {
1644 			error = EINVAL;
1645 			break;
1646 		}
1647 		state = pool_get(&pf_state_pl, PR_NOWAIT);
1648 		if (state == NULL) {
1649 			error = ENOMEM;
1650 			break;
1651 		}
1652 		kif = pfi_kif_get(ps->state.u.ifname);
1653 		if (kif == NULL) {
1654 			pool_put(&pf_state_pl, state);
1655 			error = ENOENT;
1656 			break;
1657 		}
1658 		bcopy(&ps->state, state, sizeof(struct pf_state));
1659 		bzero(&state->u, sizeof(state->u));
1660 		state->rule.ptr = &pf_default_rule;
1661 		state->nat_rule.ptr = NULL;
1662 		state->anchor.ptr = NULL;
1663 		state->rt_kif = NULL;
1664 		state->creation = time_second;
1665 		state->pfsync_time = 0;
1666 		state->packets[0] = state->packets[1] = 0;
1667 		state->bytes[0] = state->bytes[1] = 0;
1668 		state->hash = pf_state_hash(state);
1669 
1670 		if (pf_insert_state(kif, state)) {
1671 			pfi_kif_unref(kif, PFI_KIF_REF_NONE);
1672 			pool_put(&pf_state_pl, state);
1673 			error = ENOMEM;
1674 		}
1675 		break;
1676 	}
1677 
1678 	case DIOCGETSTATE: {
1679 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1680 		struct pf_state		*state;
1681 		u_int32_t		 nr;
1682 		int			 secs;
1683 
1684 		nr = 0;
1685 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1686 			if (nr >= ps->nr)
1687 				break;
1688 			nr++;
1689 		}
1690 		if (state == NULL) {
1691 			error = EBUSY;
1692 			break;
1693 		}
1694 		secs = time_second;
1695 		bcopy(state, &ps->state, sizeof(ps->state));
1696 		strlcpy(ps->state.u.ifname, state->u.s.kif->pfik_name,
1697 		    sizeof(ps->state.u.ifname));
1698 		ps->state.rule.nr = state->rule.ptr->nr;
1699 		ps->state.nat_rule.nr = (state->nat_rule.ptr == NULL) ?
1700 		    (uint32_t)(-1) : state->nat_rule.ptr->nr;
1701 		ps->state.anchor.nr = (state->anchor.ptr == NULL) ?
1702 		    -1 : state->anchor.ptr->nr;
1703 		ps->state.creation = secs - ps->state.creation;
1704 		ps->state.expire = pf_state_expires(state);
1705 		if (ps->state.expire > secs)
1706 			ps->state.expire -= secs;
1707 		else
1708 			ps->state.expire = 0;
1709 		break;
1710 	}
1711 
1712 	case DIOCGETSTATES: {
1713 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1714 		struct pf_state		*state;
1715 		struct pf_state		*p, *pstore;
1716 		u_int32_t		 nr = 0;
1717 		int			 space = ps->ps_len;
1718 
1719 		if (space == 0) {
1720 			nr = pf_status.states;
1721 			ps->ps_len = sizeof(struct pf_state) * nr;
1722 			break;
1723 		}
1724 
1725 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1726 
1727 		p = ps->ps_states;
1728 
1729 		state = TAILQ_FIRST(&state_list);
1730 		while (state) {
1731 			if (state->timeout != PFTM_UNLINKED) {
1732 				int	secs = time_second;
1733 
1734 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1735 					break;
1736 
1737 				bcopy(state, pstore, sizeof(*pstore));
1738 				strlcpy(pstore->u.ifname,
1739 				    state->u.s.kif->pfik_name,
1740 				    sizeof(pstore->u.ifname));
1741 				pstore->rule.nr = state->rule.ptr->nr;
1742 				pstore->nat_rule.nr = (state->nat_rule.ptr ==
1743 				    NULL) ? -1 : state->nat_rule.ptr->nr;
1744 				pstore->anchor.nr = (state->anchor.ptr ==
1745 				    NULL) ? -1 : state->anchor.ptr->nr;
1746 				pstore->creation = secs - pstore->creation;
1747 				pstore->expire = pf_state_expires(state);
1748 				if (pstore->expire > secs)
1749 					pstore->expire -= secs;
1750 				else
1751 					pstore->expire = 0;
1752 				error = copyout(pstore, p, sizeof(*p));
1753 				if (error) {
1754 					kfree(pstore, M_TEMP);
1755 					goto fail;
1756 				}
1757 				p++;
1758 				nr++;
1759 			}
1760 			state = TAILQ_NEXT(state, u.s.entry_list);
1761 		}
1762 
1763 		ps->ps_len = sizeof(struct pf_state) * nr;
1764 
1765 		kfree(pstore, M_TEMP);
1766 		break;
1767 	}
1768 
1769 	case DIOCGETSTATUS: {
1770 		struct pf_status *s = (struct pf_status *)addr;
1771 		bcopy(&pf_status, s, sizeof(struct pf_status));
1772 		pfi_fill_oldstatus(s);
1773 		break;
1774 	}
1775 
1776 	case DIOCSETSTATUSIF: {
1777 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1778 
1779 		if (pi->ifname[0] == 0) {
1780 			bzero(pf_status.ifname, IFNAMSIZ);
1781 			break;
1782 		}
1783 		if (ifunit(pi->ifname) == NULL) {
1784 			error = EINVAL;
1785 			break;
1786 		}
1787 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1788 		break;
1789 	}
1790 
1791 	case DIOCCLRSTATUS: {
1792 		bzero(pf_status.counters, sizeof(pf_status.counters));
1793 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1794 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
1795 		pf_status.since = time_second;
1796 		if (*pf_status.ifname)
1797 			pfi_clr_istats(pf_status.ifname);
1798 		break;
1799 	}
1800 
1801 	case DIOCNATLOOK: {
1802 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1803 		struct pf_state		*state;
1804 		struct pf_state_cmp	 key;
1805 		int			 m = 0, direction = pnl->direction;
1806 
1807 		key.af = pnl->af;
1808 		key.proto = pnl->proto;
1809 
1810 		if (!pnl->proto ||
1811 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1812 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1813 		    ((pnl->proto == IPPROTO_TCP ||
1814 		    pnl->proto == IPPROTO_UDP) &&
1815 		    (!pnl->dport || !pnl->sport)))
1816 			error = EINVAL;
1817 		else {
1818 			/*
1819 			 * userland gives us source and dest of connection,
1820 			 * reverse the lookup so we ask for what happens with
1821 			 * the return traffic, enabling us to find it in the
1822 			 * state tree.
1823 			 */
1824 			if (direction == PF_IN) {
1825 				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
1826 				key.ext.port = pnl->dport;
1827 				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
1828 				key.gwy.port = pnl->sport;
1829 				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
1830 			} else {
1831 				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
1832 				key.lan.port = pnl->dport;
1833 				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
1834 				key.ext.port = pnl->sport;
1835 				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
1836 			}
1837 			if (m > 1)
1838 				error = E2BIG;	/* more than one state */
1839 			else if (state != NULL) {
1840 				if (direction == PF_IN) {
1841 					PF_ACPY(&pnl->rsaddr, &state->lan.addr,
1842 					    state->af);
1843 					pnl->rsport = state->lan.port;
1844 					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
1845 					    pnl->af);
1846 					pnl->rdport = pnl->dport;
1847 				} else {
1848 					PF_ACPY(&pnl->rdaddr, &state->gwy.addr,
1849 					    state->af);
1850 					pnl->rdport = state->gwy.port;
1851 					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
1852 					    pnl->af);
1853 					pnl->rsport = pnl->sport;
1854 				}
1855 			} else
1856 				error = ENOENT;
1857 		}
1858 		break;
1859 	}
1860 
1861 	case DIOCSETTIMEOUT: {
1862 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1863 		int		 old;
1864 
1865 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1866 		    pt->seconds < 0) {
1867 			error = EINVAL;
1868 			goto fail;
1869 		}
1870 		old = pf_default_rule.timeout[pt->timeout];
1871 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1872 			pt->seconds = 1;
1873 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
1874 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1875 			wakeup(pf_purge_thread);
1876 		pt->seconds = old;
1877 		break;
1878 	}
1879 
1880 	case DIOCGETTIMEOUT: {
1881 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1882 
1883 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1884 			error = EINVAL;
1885 			goto fail;
1886 		}
1887 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1888 		break;
1889 	}
1890 
1891 	case DIOCGETLIMIT: {
1892 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1893 
1894 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1895 			error = EINVAL;
1896 			goto fail;
1897 		}
1898 		pl->limit = pf_pool_limits[pl->index].limit;
1899 		break;
1900 	}
1901 
1902 	case DIOCSETLIMIT: {
1903 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1904 		int			 old_limit;
1905 
1906 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1907 		    pf_pool_limits[pl->index].pp == NULL) {
1908 			error = EINVAL;
1909 			goto fail;
1910 		}
1911 
1912 		/* XXX Get an API to set limits on the zone/pool */
1913 		old_limit = pf_pool_limits[pl->index].limit;
1914 		pf_pool_limits[pl->index].limit = pl->limit;
1915 		pl->limit = old_limit;
1916 		break;
1917 	}
1918 
1919 	case DIOCSETDEBUG: {
1920 		u_int32_t	*level = (u_int32_t *)addr;
1921 
1922 		pf_status.debug = *level;
1923 		break;
1924 	}
1925 
1926 	case DIOCCLRRULECTRS: {
1927 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1928 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1929 		struct pf_rule		*rule;
1930 
1931 		TAILQ_FOREACH(rule,
1932 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1933 			rule->evaluations = 0;
1934 			rule->packets[0] = rule->packets[1] = 0;
1935 			rule->bytes[0] = rule->bytes[1] = 0;
1936 		}
1937 		break;
1938 	}
1939 
1940 	case DIOCGIFSPEED: {
1941 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1942 		struct pf_ifspeed	ps;
1943 		struct ifnet		*ifp;
1944 
1945 		if (psp->ifname[0] != 0) {
1946 			/* Can we completely trust user-land? */
1947 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1948 			ifp = ifunit(ps.ifname);
1949 			if (ifp )
1950 				psp->baudrate = ifp->if_baudrate;
1951 			else
1952 				error = EINVAL;
1953 		} else
1954 			error = EINVAL;
1955 		break;
1956 	}
1957 #ifdef ALTQ
1958 	case DIOCSTARTALTQ: {
1959 		struct pf_altq		*altq;
1960 
1961 		/* enable all altq interfaces on active list */
1962 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1963 			if (altq->qname[0] == 0) {
1964 				error = pf_enable_altq(altq);
1965 				if (error != 0)
1966 					break;
1967 			}
1968 		}
1969 		if (error == 0)
1970 			pf_altq_running = 1;
1971 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1972 		break;
1973 	}
1974 
1975 	case DIOCSTOPALTQ: {
1976 		struct pf_altq		*altq;
1977 
1978 		/* disable all altq interfaces on active list */
1979 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1980 			if (altq->qname[0] == 0) {
1981 				error = pf_disable_altq(altq);
1982 				if (error != 0)
1983 					break;
1984 			}
1985 		}
1986 		if (error == 0)
1987 			pf_altq_running = 0;
1988 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1989 		break;
1990 	}
1991 
1992 	case DIOCADDALTQ: {
1993 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1994 		struct pf_altq		*altq, *a;
1995 
1996 		if (pa->ticket != ticket_altqs_inactive) {
1997 			error = EBUSY;
1998 			break;
1999 		}
2000 		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2001 		if (altq == NULL) {
2002 			error = ENOMEM;
2003 			break;
2004 		}
2005 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2006 
2007 		/*
2008 		 * if this is for a queue, find the discipline and
2009 		 * copy the necessary fields
2010 		 */
2011 		if (altq->qname[0] != 0) {
2012 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2013 				error = EBUSY;
2014 				pool_put(&pf_altq_pl, altq);
2015 				break;
2016 			}
2017 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2018 				if (strncmp(a->ifname, altq->ifname,
2019 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2020 					altq->altq_disc = a->altq_disc;
2021 					break;
2022 				}
2023 			}
2024 		}
2025 
2026 		error = altq_add(altq);
2027 		if (error) {
2028 			pool_put(&pf_altq_pl, altq);
2029 			break;
2030 		}
2031 
2032 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2033 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2034 		break;
2035 	}
2036 
2037 	case DIOCGETALTQS: {
2038 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2039 		struct pf_altq		*altq;
2040 
2041 		pa->nr = 0;
2042 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2043 			pa->nr++;
2044 		pa->ticket = ticket_altqs_active;
2045 		break;
2046 	}
2047 
2048 	case DIOCGETALTQ: {
2049 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2050 		struct pf_altq		*altq;
2051 		u_int32_t		 nr;
2052 
2053 		if (pa->ticket != ticket_altqs_active) {
2054 			error = EBUSY;
2055 			break;
2056 		}
2057 		nr = 0;
2058 		altq = TAILQ_FIRST(pf_altqs_active);
2059 		while ((altq != NULL) && (nr < pa->nr)) {
2060 			altq = TAILQ_NEXT(altq, entries);
2061 			nr++;
2062 		}
2063 		if (altq == NULL) {
2064 			error = EBUSY;
2065 			break;
2066 		}
2067 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2068 		break;
2069 	}
2070 
2071 	case DIOCCHANGEALTQ:
2072 		/* CHANGEALTQ not supported yet! */
2073 		error = ENODEV;
2074 		break;
2075 
2076 	case DIOCGETQSTATS: {
2077 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2078 		struct pf_altq		*altq;
2079 		u_int32_t		 nr;
2080 		int			 nbytes;
2081 
2082 		if (pq->ticket != ticket_altqs_active) {
2083 			error = EBUSY;
2084 			break;
2085 		}
2086 		nbytes = pq->nbytes;
2087 		nr = 0;
2088 		altq = TAILQ_FIRST(pf_altqs_active);
2089 		while ((altq != NULL) && (nr < pq->nr)) {
2090 			altq = TAILQ_NEXT(altq, entries);
2091 			nr++;
2092 		}
2093 		if (altq == NULL) {
2094 			error = EBUSY;
2095 			break;
2096 		}
2097 		error = altq_getqstats(altq, pq->buf, &nbytes);
2098 		if (error == 0) {
2099 			pq->scheduler = altq->scheduler;
2100 			pq->nbytes = nbytes;
2101 		}
2102 		break;
2103 	}
2104 #endif /* ALTQ */
2105 
2106 	case DIOCBEGINADDRS: {
2107 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2108 
2109 		pf_empty_pool(&pf_pabuf);
2110 		pp->ticket = ++ticket_pabuf;
2111 		break;
2112 	}
2113 
2114 	case DIOCADDADDR: {
2115 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2116 
2117 		if (pp->ticket != ticket_pabuf) {
2118 			error = EBUSY;
2119 			break;
2120 		}
2121 #ifndef INET
2122 		if (pp->af == AF_INET) {
2123 			error = EAFNOSUPPORT;
2124 			break;
2125 		}
2126 #endif /* INET */
2127 #ifndef INET6
2128 		if (pp->af == AF_INET6) {
2129 			error = EAFNOSUPPORT;
2130 			break;
2131 		}
2132 #endif /* INET6 */
2133 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2134 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2135 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2136 			error = EINVAL;
2137 			break;
2138 		}
2139 		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2140 		if (pa == NULL) {
2141 			error = ENOMEM;
2142 			break;
2143 		}
2144 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2145 		if (pa->ifname[0]) {
2146 			pa->kif = pfi_kif_get(pa->ifname);
2147 			if (pa->kif == NULL) {
2148 				pool_put(&pf_pooladdr_pl, pa);
2149 				error = EINVAL;
2150 				break;
2151 			}
2152 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2153 		}
2154 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2155 			pfi_dynaddr_remove(&pa->addr);
2156 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2157 			pool_put(&pf_pooladdr_pl, pa);
2158 			error = EINVAL;
2159 			break;
2160 		}
2161 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2162 		break;
2163 	}
2164 
2165 	case DIOCGETADDRS: {
2166 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2167 
2168 		pp->nr = 0;
2169 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2170 		    pp->r_num, 0, 1, 0);
2171 		if (pool == NULL) {
2172 			error = EBUSY;
2173 			break;
2174 		}
2175 		TAILQ_FOREACH(pa, &pool->list, entries)
2176 			pp->nr++;
2177 		break;
2178 	}
2179 
2180 	case DIOCGETADDR: {
2181 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2182 		u_int32_t		 nr = 0;
2183 
2184 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2185 		    pp->r_num, 0, 1, 1);
2186 		if (pool == NULL) {
2187 			error = EBUSY;
2188 			break;
2189 		}
2190 		pa = TAILQ_FIRST(&pool->list);
2191 		while ((pa != NULL) && (nr < pp->nr)) {
2192 			pa = TAILQ_NEXT(pa, entries);
2193 			nr++;
2194 		}
2195 		if (pa == NULL) {
2196 			error = EBUSY;
2197 			break;
2198 		}
2199 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2200 		pfi_dynaddr_copyout(&pp->addr.addr);
2201 		pf_tbladdr_copyout(&pp->addr.addr);
2202 		pf_rtlabel_copyout(&pp->addr.addr);
2203 		break;
2204 	}
2205 
2206 	case DIOCCHANGEADDR: {
2207 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2208 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2209 		struct pf_ruleset	*ruleset;
2210 
2211 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2212 		    pca->action > PF_CHANGE_REMOVE) {
2213 			error = EINVAL;
2214 			break;
2215 		}
2216 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2217 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2218 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2219 			error = EINVAL;
2220 			break;
2221 		}
2222 
2223 		ruleset = pf_find_ruleset(pca->anchor);
2224 		if (ruleset == NULL) {
2225 			error = EBUSY;
2226 			break;
2227 		}
2228 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2229 		    pca->r_num, pca->r_last, 1, 1);
2230 		if (pool == NULL) {
2231 			error = EBUSY;
2232 			break;
2233 		}
2234 		if (pca->action != PF_CHANGE_REMOVE) {
2235 			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2236 			if (newpa == NULL) {
2237 				error = ENOMEM;
2238 				break;
2239 			}
2240 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2241 #ifndef INET
2242 			if (pca->af == AF_INET) {
2243 				pool_put(&pf_pooladdr_pl, newpa);
2244 				error = EAFNOSUPPORT;
2245 				break;
2246 			}
2247 #endif /* INET */
2248 #ifndef INET6
2249 			if (pca->af == AF_INET6) {
2250 				pool_put(&pf_pooladdr_pl, newpa);
2251 				error = EAFNOSUPPORT;
2252 				break;
2253 			}
2254 #endif /* INET6 */
2255 			if (newpa->ifname[0]) {
2256 				newpa->kif = pfi_kif_get(newpa->ifname);
2257 				if (newpa->kif == NULL) {
2258 					pool_put(&pf_pooladdr_pl, newpa);
2259 					error = EINVAL;
2260 					break;
2261 				}
2262 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2263 			} else
2264 				newpa->kif = NULL;
2265 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2266 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2267 				pfi_dynaddr_remove(&newpa->addr);
2268 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2269 				pool_put(&pf_pooladdr_pl, newpa);
2270 				error = EINVAL;
2271 				break;
2272 			}
2273 		}
2274 
2275 		if (pca->action == PF_CHANGE_ADD_HEAD)
2276 			oldpa = TAILQ_FIRST(&pool->list);
2277 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2278 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2279 		else {
2280 			int	i = 0;
2281 
2282 			oldpa = TAILQ_FIRST(&pool->list);
2283 			while ((oldpa != NULL) && (i < pca->nr)) {
2284 				oldpa = TAILQ_NEXT(oldpa, entries);
2285 				i++;
2286 			}
2287 			if (oldpa == NULL) {
2288 				error = EINVAL;
2289 				break;
2290 			}
2291 		}
2292 
2293 		if (pca->action == PF_CHANGE_REMOVE) {
2294 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2295 			pfi_dynaddr_remove(&oldpa->addr);
2296 			pf_tbladdr_remove(&oldpa->addr);
2297 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2298 			pool_put(&pf_pooladdr_pl, oldpa);
2299 		} else {
2300 			if (oldpa == NULL)
2301 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2302 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2303 			    pca->action == PF_CHANGE_ADD_BEFORE)
2304 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2305 			else
2306 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2307 				    newpa, entries);
2308 		}
2309 
2310 		pool->cur = TAILQ_FIRST(&pool->list);
2311 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2312 		    pca->af);
2313 		break;
2314 	}
2315 
2316 	case DIOCGETRULESETS: {
2317 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2318 		struct pf_ruleset	*ruleset;
2319 		struct pf_anchor	*anchor;
2320 
2321 		pr->path[sizeof(pr->path) - 1] = 0;
2322 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2323 			error = EINVAL;
2324 			break;
2325 		}
2326 		pr->nr = 0;
2327 		if (ruleset->anchor == NULL) {
2328 			/* XXX kludge for pf_main_ruleset */
2329 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2330 				if (anchor->parent == NULL)
2331 					pr->nr++;
2332 		} else {
2333 			RB_FOREACH(anchor, pf_anchor_node,
2334 			    &ruleset->anchor->children)
2335 				pr->nr++;
2336 		}
2337 		break;
2338 	}
2339 
2340 	case DIOCGETRULESET: {
2341 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2342 		struct pf_ruleset	*ruleset;
2343 		struct pf_anchor	*anchor;
2344 		u_int32_t		 nr = 0;
2345 
2346 		pr->path[sizeof(pr->path) - 1] = 0;
2347 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2348 			error = EINVAL;
2349 			break;
2350 		}
2351 		pr->name[0] = 0;
2352 		if (ruleset->anchor == NULL) {
2353 			/* XXX kludge for pf_main_ruleset */
2354 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2355 				if (anchor->parent == NULL && nr++ == pr->nr) {
2356 					strlcpy(pr->name, anchor->name,
2357 					    sizeof(pr->name));
2358 					break;
2359 				}
2360 		} else {
2361 			RB_FOREACH(anchor, pf_anchor_node,
2362 			    &ruleset->anchor->children)
2363 				if (nr++ == pr->nr) {
2364 					strlcpy(pr->name, anchor->name,
2365 					    sizeof(pr->name));
2366 					break;
2367 				}
2368 		}
2369 		if (!pr->name[0])
2370 			error = EBUSY;
2371 		break;
2372 	}
2373 
2374 	case DIOCRCLRTABLES: {
2375 		struct pfioc_table *io = (struct pfioc_table *)addr;
2376 
2377 		if (io->pfrio_esize != 0) {
2378 			error = ENODEV;
2379 			break;
2380 		}
2381 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2382 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2383 		break;
2384 	}
2385 
2386 	case DIOCRADDTABLES: {
2387 		struct pfioc_table *io = (struct pfioc_table *)addr;
2388 
2389 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2390 			error = ENODEV;
2391 			break;
2392 		}
2393 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2394 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2395 		break;
2396 	}
2397 
2398 	case DIOCRDELTABLES: {
2399 		struct pfioc_table *io = (struct pfioc_table *)addr;
2400 
2401 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2402 			error = ENODEV;
2403 			break;
2404 		}
2405 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2406 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2407 		break;
2408 	}
2409 
2410 	case DIOCRGETTABLES: {
2411 		struct pfioc_table *io = (struct pfioc_table *)addr;
2412 
2413 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2414 			error = ENODEV;
2415 			break;
2416 		}
2417 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2418 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2419 		break;
2420 	}
2421 
2422 	case DIOCRGETTSTATS: {
2423 		struct pfioc_table *io = (struct pfioc_table *)addr;
2424 
2425 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2426 			error = ENODEV;
2427 			break;
2428 		}
2429 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2430 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2431 		break;
2432 	}
2433 
2434 	case DIOCRCLRTSTATS: {
2435 		struct pfioc_table *io = (struct pfioc_table *)addr;
2436 
2437 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2438 			error = ENODEV;
2439 			break;
2440 		}
2441 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2442 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2443 		break;
2444 	}
2445 
2446 	case DIOCRSETTFLAGS: {
2447 		struct pfioc_table *io = (struct pfioc_table *)addr;
2448 
2449 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2450 			error = ENODEV;
2451 			break;
2452 		}
2453 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2454 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2455 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2456 		break;
2457 	}
2458 
2459 	case DIOCRCLRADDRS: {
2460 		struct pfioc_table *io = (struct pfioc_table *)addr;
2461 
2462 		if (io->pfrio_esize != 0) {
2463 			error = ENODEV;
2464 			break;
2465 		}
2466 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2467 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2468 		break;
2469 	}
2470 
2471 	case DIOCRADDADDRS: {
2472 		struct pfioc_table *io = (struct pfioc_table *)addr;
2473 
2474 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2475 			error = ENODEV;
2476 			break;
2477 		}
2478 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2479 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2480 		    PFR_FLAG_USERIOCTL);
2481 		break;
2482 	}
2483 
2484 	case DIOCRDELADDRS: {
2485 		struct pfioc_table *io = (struct pfioc_table *)addr;
2486 
2487 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2488 			error = ENODEV;
2489 			break;
2490 		}
2491 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2492 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2493 		    PFR_FLAG_USERIOCTL);
2494 		break;
2495 	}
2496 
2497 	case DIOCRSETADDRS: {
2498 		struct pfioc_table *io = (struct pfioc_table *)addr;
2499 
2500 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2501 			error = ENODEV;
2502 			break;
2503 		}
2504 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2505 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2506 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2507 		    PFR_FLAG_USERIOCTL, 0);
2508 		break;
2509 	}
2510 
2511 	case DIOCRGETADDRS: {
2512 		struct pfioc_table *io = (struct pfioc_table *)addr;
2513 
2514 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2515 			error = ENODEV;
2516 			break;
2517 		}
2518 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2519 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2520 		break;
2521 	}
2522 
2523 	case DIOCRGETASTATS: {
2524 		struct pfioc_table *io = (struct pfioc_table *)addr;
2525 
2526 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2527 			error = ENODEV;
2528 			break;
2529 		}
2530 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2531 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2532 		break;
2533 	}
2534 
2535 	case DIOCRCLRASTATS: {
2536 		struct pfioc_table *io = (struct pfioc_table *)addr;
2537 
2538 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2539 			error = ENODEV;
2540 			break;
2541 		}
2542 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2543 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2544 		    PFR_FLAG_USERIOCTL);
2545 		break;
2546 	}
2547 
2548 	case DIOCRTSTADDRS: {
2549 		struct pfioc_table *io = (struct pfioc_table *)addr;
2550 
2551 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2552 			error = ENODEV;
2553 			break;
2554 		}
2555 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2556 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2557 		    PFR_FLAG_USERIOCTL);
2558 		break;
2559 	}
2560 
2561 	case DIOCRINADEFINE: {
2562 		struct pfioc_table *io = (struct pfioc_table *)addr;
2563 
2564 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2565 			error = ENODEV;
2566 			break;
2567 		}
2568 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2569 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2570 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2571 		break;
2572 	}
2573 
2574 	case DIOCOSFPADD: {
2575 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2576 		error = pf_osfp_add(io);
2577 		break;
2578 	}
2579 
2580 	case DIOCOSFPGET: {
2581 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2582 		error = pf_osfp_get(io);
2583 		break;
2584 	}
2585 
2586 	case DIOCXBEGIN: {
2587 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2588 		struct pfioc_trans_e	*ioe;
2589 		struct pfr_table	*table;
2590 		int			 i;
2591 
2592 		if (io->esize != sizeof(*ioe)) {
2593 			error = ENODEV;
2594 			goto fail;
2595 		}
2596 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2597 		    M_TEMP, M_WAITOK);
2598 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2599 		    M_TEMP, M_WAITOK);
2600 		for (i = 0; i < io->size; i++) {
2601 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2602 				kfree(table, M_TEMP);
2603 				kfree(ioe, M_TEMP);
2604 				error = EFAULT;
2605 				goto fail;
2606 			}
2607 			switch (ioe->rs_num) {
2608 #ifdef ALTQ
2609 			case PF_RULESET_ALTQ:
2610 				if (ioe->anchor[0]) {
2611 					kfree(table, M_TEMP);
2612 					kfree(ioe, M_TEMP);
2613 					error = EINVAL;
2614 					goto fail;
2615 				}
2616 				if ((error = pf_begin_altq(&ioe->ticket))) {
2617 					kfree(table, M_TEMP);
2618 					kfree(ioe, M_TEMP);
2619 					goto fail;
2620 				}
2621 				break;
2622 #endif /* ALTQ */
2623 			case PF_RULESET_TABLE:
2624 				bzero(table, sizeof(*table));
2625 				strlcpy(table->pfrt_anchor, ioe->anchor,
2626 				    sizeof(table->pfrt_anchor));
2627 				if ((error = pfr_ina_begin(table,
2628 				    &ioe->ticket, NULL, 0))) {
2629 					kfree(table, M_TEMP);
2630 					kfree(ioe, M_TEMP);
2631 					goto fail;
2632 				}
2633 				break;
2634 			default:
2635 				if ((error = pf_begin_rules(&ioe->ticket,
2636 				    ioe->rs_num, ioe->anchor))) {
2637 					kfree(table, M_TEMP);
2638 					kfree(ioe, M_TEMP);
2639 					goto fail;
2640 				}
2641 				break;
2642 			}
2643 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2644 				kfree(table, M_TEMP);
2645 				kfree(ioe, M_TEMP);
2646 				error = EFAULT;
2647 				goto fail;
2648 			}
2649 		}
2650 		kfree(table, M_TEMP);
2651 		kfree(ioe, M_TEMP);
2652 		break;
2653 	}
2654 
2655 	case DIOCXROLLBACK: {
2656 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2657 		struct pfioc_trans_e	*ioe;
2658 		struct pfr_table	*table;
2659 		int			 i;
2660 
2661 		if (io->esize != sizeof(*ioe)) {
2662 			error = ENODEV;
2663 			goto fail;
2664 		}
2665 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2666 		    M_TEMP, M_WAITOK);
2667 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2668 		    M_TEMP, M_WAITOK);
2669 		for (i = 0; i < io->size; i++) {
2670 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2671 				kfree(table, M_TEMP);
2672 				kfree(ioe, M_TEMP);
2673 				error = EFAULT;
2674 				goto fail;
2675 			}
2676 			switch (ioe->rs_num) {
2677 #ifdef ALTQ
2678 			case PF_RULESET_ALTQ:
2679 				if (ioe->anchor[0]) {
2680 					kfree(table, M_TEMP);
2681 					kfree(ioe, M_TEMP);
2682 					error = EINVAL;
2683 					goto fail;
2684 				}
2685 				if ((error = pf_rollback_altq(ioe->ticket))) {
2686 					kfree(table, M_TEMP);
2687 					kfree(ioe, M_TEMP);
2688 					goto fail; /* really bad */
2689 				}
2690 				break;
2691 #endif /* ALTQ */
2692 			case PF_RULESET_TABLE:
2693 				bzero(table, sizeof(*table));
2694 				strlcpy(table->pfrt_anchor, ioe->anchor,
2695 				    sizeof(table->pfrt_anchor));
2696 				if ((error = pfr_ina_rollback(table,
2697 				    ioe->ticket, NULL, 0))) {
2698 					kfree(table, M_TEMP);
2699 					kfree(ioe, M_TEMP);
2700 					goto fail; /* really bad */
2701 				}
2702 				break;
2703 			default:
2704 				if ((error = pf_rollback_rules(ioe->ticket,
2705 				    ioe->rs_num, ioe->anchor))) {
2706 					kfree(table, M_TEMP);
2707 					kfree(ioe, M_TEMP);
2708 					goto fail; /* really bad */
2709 				}
2710 				break;
2711 			}
2712 		}
2713 		kfree(table, M_TEMP);
2714 		kfree(ioe, M_TEMP);
2715 		break;
2716 	}
2717 
2718 	case DIOCXCOMMIT: {
2719 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2720 		struct pfioc_trans_e	*ioe;
2721 		struct pfr_table	*table;
2722 		struct pf_ruleset	*rs;
2723 		int			 i;
2724 
2725 		if (io->esize != sizeof(*ioe)) {
2726 			error = ENODEV;
2727 			goto fail;
2728 		}
2729 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2730 		    M_TEMP, M_WAITOK);
2731 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2732 		    M_TEMP, M_WAITOK);
2733 		/* first makes sure everything will succeed */
2734 		for (i = 0; i < io->size; i++) {
2735 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2736 				kfree(table, M_TEMP);
2737 				kfree(ioe, M_TEMP);
2738 				error = EFAULT;
2739 				goto fail;
2740 			}
2741 			switch (ioe->rs_num) {
2742 #ifdef ALTQ
2743 			case PF_RULESET_ALTQ:
2744 				if (ioe->anchor[0]) {
2745 					kfree(table, M_TEMP);
2746 					kfree(ioe, M_TEMP);
2747 					error = EINVAL;
2748 					goto fail;
2749 				}
2750 				if (!altqs_inactive_open || ioe->ticket !=
2751 				    ticket_altqs_inactive) {
2752 					kfree(table, M_TEMP);
2753 					kfree(ioe, M_TEMP);
2754 					error = EBUSY;
2755 					goto fail;
2756 				}
2757 				break;
2758 #endif /* ALTQ */
2759 			case PF_RULESET_TABLE:
2760 				rs = pf_find_ruleset(ioe->anchor);
2761 				if (rs == NULL || !rs->topen || ioe->ticket !=
2762 				     rs->tticket) {
2763 					kfree(table, M_TEMP);
2764 					kfree(ioe, M_TEMP);
2765 					error = EBUSY;
2766 					goto fail;
2767 				}
2768 				break;
2769 			default:
2770 				if (ioe->rs_num < 0 || ioe->rs_num >=
2771 				    PF_RULESET_MAX) {
2772 					kfree(table, M_TEMP);
2773 					kfree(ioe, M_TEMP);
2774 					error = EINVAL;
2775 					goto fail;
2776 				}
2777 				rs = pf_find_ruleset(ioe->anchor);
2778 				if (rs == NULL ||
2779 				    !rs->rules[ioe->rs_num].inactive.open ||
2780 				    rs->rules[ioe->rs_num].inactive.ticket !=
2781 				    ioe->ticket) {
2782 					kfree(table, M_TEMP);
2783 					kfree(ioe, M_TEMP);
2784 					error = EBUSY;
2785 					goto fail;
2786 				}
2787 				break;
2788 			}
2789 		}
2790 		/* now do the commit - no errors should happen here */
2791 		for (i = 0; i < io->size; i++) {
2792 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2793 				kfree(table, M_TEMP);
2794 				kfree(ioe, M_TEMP);
2795 				error = EFAULT;
2796 				goto fail;
2797 			}
2798 			switch (ioe->rs_num) {
2799 #ifdef ALTQ
2800 			case PF_RULESET_ALTQ:
2801 				if ((error = pf_commit_altq(ioe->ticket))) {
2802 					kfree(table, M_TEMP);
2803 					kfree(ioe, M_TEMP);
2804 					goto fail; /* really bad */
2805 				}
2806 				break;
2807 #endif /* ALTQ */
2808 			case PF_RULESET_TABLE:
2809 				bzero(table, sizeof(*table));
2810 				strlcpy(table->pfrt_anchor, ioe->anchor,
2811 				    sizeof(table->pfrt_anchor));
2812 				if ((error = pfr_ina_commit(table, ioe->ticket,
2813 				    NULL, NULL, 0))) {
2814 					kfree(table, M_TEMP);
2815 					kfree(ioe, M_TEMP);
2816 					goto fail; /* really bad */
2817 				}
2818 				break;
2819 			default:
2820 				if ((error = pf_commit_rules(ioe->ticket,
2821 				    ioe->rs_num, ioe->anchor))) {
2822 					kfree(table, M_TEMP);
2823 					kfree(ioe, M_TEMP);
2824 					goto fail; /* really bad */
2825 				}
2826 				break;
2827 			}
2828 		}
2829 		kfree(table, M_TEMP);
2830 		kfree(ioe, M_TEMP);
2831 		break;
2832 	}
2833 
2834 	case DIOCGETSRCNODES: {
2835 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2836 		struct pf_src_node	*n, *p, *pstore;
2837 		u_int32_t		 nr = 0;
2838 		int			 space = psn->psn_len;
2839 
2840 		if (space == 0) {
2841 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2842 				nr++;
2843 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2844 			break;
2845 		}
2846 
2847 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2848 
2849 		p = psn->psn_src_nodes;
2850 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2851 			int	secs = time_second, diff;
2852 
2853 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
2854 				break;
2855 
2856 			bcopy(n, pstore, sizeof(*pstore));
2857 			if (n->rule.ptr != NULL)
2858 				pstore->rule.nr = n->rule.ptr->nr;
2859 			pstore->creation = secs - pstore->creation;
2860 			if (pstore->expire > secs)
2861 				pstore->expire -= secs;
2862 			else
2863 				pstore->expire = 0;
2864 
2865 			/* adjust the connection rate estimate */
2866 			diff = secs - n->conn_rate.last;
2867 			if (diff >= n->conn_rate.seconds)
2868 				pstore->conn_rate.count = 0;
2869 			else
2870 				pstore->conn_rate.count -=
2871 				    n->conn_rate.count * diff /
2872 				    n->conn_rate.seconds;
2873 
2874 			error = copyout(pstore, p, sizeof(*p));
2875 			if (error) {
2876 				kfree(pstore, M_TEMP);
2877 				goto fail;
2878 			}
2879 			p++;
2880 			nr++;
2881 		}
2882 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2883 
2884 		kfree(pstore, M_TEMP);
2885 		break;
2886 	}
2887 
2888 	case DIOCCLRSRCNODES: {
2889 		struct pf_src_node	*n;
2890 		struct pf_state		*state;
2891 
2892 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2893 			state->src_node = NULL;
2894 			state->nat_src_node = NULL;
2895 		}
2896 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2897 			n->expire = 1;
2898 			n->states = 0;
2899 		}
2900 		pf_purge_expired_src_nodes(1);
2901 		pf_status.src_nodes = 0;
2902 		break;
2903 	}
2904 
2905 	case DIOCKILLSRCNODES: {
2906 		struct pf_src_node	*sn;
2907 		struct pf_state		*s;
2908 		struct pfioc_src_node_kill *psnk = \
2909 			(struct pfioc_src_node_kill *) addr;
2910 		int			killed = 0;
2911 
2912 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
2913         		if (PF_MATCHA(psnk->psnk_src.neg, \
2914 				      &psnk->psnk_src.addr.v.a.addr, \
2915 				      &psnk->psnk_src.addr.v.a.mask, \
2916 				      &sn->addr, sn->af) &&
2917 			    PF_MATCHA(psnk->psnk_dst.neg, \
2918 				      &psnk->psnk_dst.addr.v.a.addr, \
2919 				      &psnk->psnk_dst.addr.v.a.mask, \
2920 				      &sn->raddr, sn->af)) {
2921 				/* Handle state to src_node linkage */
2922 				if (sn->states != 0) {
2923 					RB_FOREACH(s, pf_state_tree_id,
2924 					    &tree_id) {
2925 						if (s->src_node == sn)
2926 							s->src_node = NULL;
2927 						if (s->nat_src_node == sn)
2928 							s->nat_src_node = NULL;
2929 					}
2930 					sn->states = 0;
2931 				}
2932 				sn->expire = 1;
2933 				killed++;
2934 			}
2935 		}
2936 
2937 		if (killed > 0)
2938 			pf_purge_expired_src_nodes(1);
2939 
2940 		psnk->psnk_af = killed;
2941 		break;
2942 	}
2943 
2944 	case DIOCSETHOSTID: {
2945 		u_int32_t	*hostid = (u_int32_t *)addr;
2946 
2947 		if (*hostid == 0)
2948 			pf_status.hostid = karc4random();
2949 		else
2950 			pf_status.hostid = *hostid;
2951 		break;
2952 	}
2953 
2954 	case DIOCOSFPFLUSH:
2955 		crit_enter();
2956 		pf_osfp_flush();
2957 		crit_exit();
2958 		break;
2959 
2960 	case DIOCIGETIFACES: {
2961 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2962 
2963 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
2964 			error = ENODEV;
2965 			break;
2966 		}
2967 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2968 		    &io->pfiio_size);
2969 		break;
2970 	}
2971 
2972 	case DIOCSETIFFLAG: {
2973 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2974 
2975 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
2976 		break;
2977 	}
2978 
2979 	case DIOCCLRIFFLAG: {
2980 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2981 
2982 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
2983 		break;
2984 	}
2985 
2986 	default:
2987 		error = ENODEV;
2988 		break;
2989 	}
2990 fail:
2991 	return (error);
2992 }
2993 
2994 /*
2995  * XXX - Check for version missmatch!!!
2996  */
2997 static void
2998 pf_clear_states(void)
2999 {
3000 	struct pf_state		*state;
3001 
3002 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3003 		state->timeout = PFTM_PURGE;
3004 #if NPFSYNC
3005 		/* don't send out individual delete messages */
3006 		state->sync_flags = PFSTATE_NOSYNC;
3007 #endif
3008 		pf_unlink_state(state);
3009 	}
3010 	pf_status.states = 0;
3011 #if 0 /* NPFSYNC */
3012 /*
3013  * XXX This is called on module unload, we do not want to sync that over? */
3014  */
3015 	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
3016 #endif
3017 }
3018 
3019 static int
3020 pf_clear_tables(void)
3021 {
3022 	struct pfioc_table io;
3023 	int error;
3024 
3025 	bzero(&io, sizeof(io));
3026 
3027 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3028 	    io.pfrio_flags);
3029 
3030 	return (error);
3031 }
3032 
3033 static void
3034 pf_clear_srcnodes(void)
3035 {
3036 	struct pf_src_node	*n;
3037 	struct pf_state		*state;
3038 
3039 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3040 		state->src_node = NULL;
3041 		state->nat_src_node = NULL;
3042 	}
3043 	RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3044 		n->expire = 1;
3045 		n->states = 0;
3046 	}
3047 	pf_purge_expired_src_nodes(0);
3048 	pf_status.src_nodes = 0;
3049 }
3050 /*
3051  * XXX - Check for version missmatch!!!
3052  */
3053 
3054 /*
3055  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3056  */
3057 static int
3058 shutdown_pf(void)
3059 {
3060 	int error = 0;
3061 	u_int32_t t[5];
3062 	char nn = '\0';
3063 
3064 	pf_status.running = 0;
3065 	do {
3066 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
3067 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3068 			break;
3069 		}
3070 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
3071 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3072 			break;		/* XXX: rollback? */
3073 		}
3074 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
3075 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3076 			break;		/* XXX: rollback? */
3077 		}
3078 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3079 		    != 0) {
3080 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3081 			break;		/* XXX: rollback? */
3082 		}
3083 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3084 		    != 0) {
3085 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3086 			break;		/* XXX: rollback? */
3087 		}
3088 
3089 		/* XXX: these should always succeed here */
3090 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3091 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3092 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3093 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3094 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3095 
3096 		if ((error = pf_clear_tables()) != 0)
3097 			break;
3098 
3099 #ifdef ALTQ
3100 		if ((error = pf_begin_altq(&t[0])) != 0) {
3101 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3102 			break;
3103 		}
3104 		pf_commit_altq(t[0]);
3105 #endif
3106 
3107 		pf_clear_states();
3108 
3109 		pf_clear_srcnodes();
3110 
3111 		/* status does not use malloced mem so no need to cleanup */
3112 		/* fingerprints and interfaces have their own cleanup code */
3113 	} while(0);
3114 
3115         return (error);
3116 }
3117 
3118 static int
3119 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3120 {
3121 	/*
3122 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3123 	 * for ip_len/ip_off. This is why we don't have to change byte order
3124 	 * like the FreeBSD-5 version does.
3125 	 */
3126 	int chk;
3127 
3128 	chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3129 	if (chk && *m) {
3130 		m_freem(*m);
3131 		*m = NULL;
3132 	}
3133 	return chk;
3134 }
3135 
3136 static int
3137 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3138 {
3139 	/*
3140 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3141 	 * for ip_len/ip_off. This is why we don't have to change byte order
3142 	 * like the FreeBSD-5 version does.
3143 	 */
3144 	int chk;
3145 
3146 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3147 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3148 		in_delayed_cksum(*m);
3149 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3150 	}
3151 	chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3152 	if (chk && *m) {
3153 		m_freem(*m);
3154 		*m = NULL;
3155 	}
3156 	return chk;
3157 }
3158 
3159 #ifdef INET6
3160 static int
3161 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3162 {
3163 	/*
3164 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3165 	 */
3166 	int chk;
3167 
3168 	chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3169 	if (chk && *m) {
3170 		m_freem(*m);
3171 		*m = NULL;
3172 	}
3173 	return chk;
3174 }
3175 
3176 static int
3177 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3178 {
3179 	/*
3180 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3181 	 */
3182 	int chk;
3183 
3184 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3185 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3186 		in_delayed_cksum(*m);
3187 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3188 	}
3189 	chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3190 	if (chk && *m) {
3191 		m_freem(*m);
3192 		*m = NULL;
3193 	}
3194 	return chk;
3195 }
3196 #endif /* INET6 */
3197 
3198 static int
3199 hook_pf(void)
3200 {
3201 	struct pfil_head *pfh_inet;
3202 #ifdef INET6
3203 	struct pfil_head *pfh_inet6;
3204 #endif
3205 
3206 	if (pf_pfil_hooked)
3207 		return (0);
3208 
3209 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3210 	if (pfh_inet == NULL)
3211 		return (ENODEV);
3212 	pfil_add_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3213 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3214 #ifdef INET6
3215 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3216 	if (pfh_inet6 == NULL) {
3217 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3218 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3219 		return (ENODEV);
3220 	}
3221 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3222 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3223 #endif
3224 
3225 	pf_pfil_hooked = 1;
3226 	return (0);
3227 }
3228 
3229 static int
3230 dehook_pf(void)
3231 {
3232 	struct pfil_head *pfh_inet;
3233 #ifdef INET6
3234 	struct pfil_head *pfh_inet6;
3235 #endif
3236 
3237 	if (pf_pfil_hooked == 0)
3238 		return (0);
3239 
3240 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3241 	if (pfh_inet == NULL)
3242 		return (ENODEV);
3243 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3244 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3245 #ifdef INET6
3246 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3247 	if (pfh_inet6 == NULL)
3248 		return (ENODEV);
3249 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3250 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3251 #endif
3252 
3253 	pf_pfil_hooked = 0;
3254 	return (0);
3255 }
3256 
3257 static int
3258 pf_load(void)
3259 {
3260 	int error;
3261 
3262 	init_zone_var();
3263 	lockinit(&pf_mod_lck, "pf task lck", 0, LK_CANRECURSE);
3264 	kprintf("pf_load() p1\n"); /*XXX*/
3265 	pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3266 	kprintf("pf_load() p2\n"); /*XXX*/
3267 	error = pfattach();
3268 	kprintf("pf_load() p3\n"); /*XXX*/
3269 	if (error) {
3270 		dev_ops_remove_all(&pf_ops);
3271 		lockuninit(&pf_mod_lck);
3272 		return (error);
3273 	}
3274 	lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3275 	kprintf("pf_load() p4\n"); /*XXX*/
3276 	return (0);
3277 }
3278 
3279 static int
3280 pf_unload(void)
3281 {
3282 	int error;
3283 
3284 	pf_status.running = 0;
3285 	error = dehook_pf();
3286 	if (error) {
3287 		/*
3288 		 * Should not happen!
3289 		 * XXX Due to error code ESRCH, kldunload will show
3290 		 * a message like 'No such process'.
3291 		 */
3292 		kprintf("pfil unregistration fail\n");
3293 		return error;
3294 	}
3295 	shutdown_pf();
3296 	pf_end_threads = 1;
3297 	while (pf_end_threads < 2) {
3298 		wakeup_one(pf_purge_thread);
3299 		lksleep(pf_purge_thread, &pf_mod_lck, 0, "pftmo", hz);
3300 
3301 	}
3302 	pfi_cleanup();
3303 	pf_osfp_flush();
3304 	pf_osfp_cleanup();
3305 	cleanup_pf_zone();
3306 	dev_ops_remove_all(&pf_ops);
3307 	lockuninit(&pf_consistency_lock);
3308 	lockuninit(&pf_mod_lck);
3309 	return 0;
3310 }
3311 
3312 static int
3313 pf_modevent(module_t mod, int type, void *data)
3314 {
3315 	int error = 0;
3316 
3317 	switch(type) {
3318 	case MOD_LOAD:
3319 		error = pf_load();
3320 		break;
3321 
3322 	case MOD_UNLOAD:
3323 		error = pf_unload();
3324 		break;
3325 	default:
3326 		error = EINVAL;
3327 		break;
3328 	}
3329 	return error;
3330 }
3331 
3332 static moduledata_t pf_mod = {
3333 	"pf",
3334 	pf_modevent,
3335 	0
3336 };
3337 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3338 MODULE_VERSION(pf, PF_MODVER);
3339