xref: /dflybsd-src/sys/net/pf/pf_ioctl.c (revision 7bf41faaed5442d0a7f34840646b4a385034515c)
1 /*	$OpenBSD: pf_ioctl.c,v 1.209 2008/06/29 08:42:15 mcbride Exp $ */
2 /*add $OpenBSD: pf_ioctl.c,v 1.212 2009/02/15 20:42:33 mbalmer Exp $ */
3 
4 /*
5  * Copyright (c) 2010 The DragonFly Project.  All rights reserved.
6  *
7  * Copyright (c) 2001 Daniel Hartmeier
8  * Copyright (c) 2002,2003 Henning Brauer
9  * All rights reserved.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  *
15  *    - Redistributions of source code must retain the above copyright
16  *      notice, this list of conditions and the following disclaimer.
17  *    - Redistributions in binary form must reproduce the above
18  *      copyright notice, this list of conditions and the following
19  *      disclaimer in the documentation and/or other materials provided
20  *      with the distribution.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
28  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
30  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
32  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33  * POSSIBILITY OF SUCH DAMAGE.
34  *
35  * Effort sponsored in part by the Defense Advanced Research Projects
36  * Agency (DARPA) and Air Force Research Laboratory, Air Force
37  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
38  *
39  */
40 
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/conf.h>
47 #include <sys/device.h>
48 #include <sys/mbuf.h>
49 #include <sys/filio.h>
50 #include <sys/fcntl.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/kernel.h>
54 #include <sys/kthread.h>
55 #include <sys/time.h>
56 #include <sys/proc.h>
57 #include <sys/malloc.h>
58 #include <sys/module.h>
59 #include <sys/lock.h>
60 
61 #include <sys/thread2.h>
62 
63 #include <net/if.h>
64 #include <net/if_types.h>
65 #include <net/route.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/ip.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_icmp.h>
73 
74 #include <net/pf/pfvar.h>
75 #include <sys/md5.h>
76 #include <net/pf/pfvar.h>
77 
78 #include <net/pf/if_pflog.h>
79 #include <net/pf/if_pfsync.h>
80 
81 #ifdef INET6
82 #include <netinet/ip6.h>
83 #include <netinet/in_pcb.h>
84 #endif /* INET6 */
85 
86 #ifdef ALTQ
87 #include <net/altq/altq.h>
88 #endif
89 
90 #include <machine/limits.h>
91 #include <net/pfil.h>
92 #include <sys/mutex.h>
93 
94 u_int rt_numfibs = RT_NUMFIBS;
95 
96 void			 pfattach(void);
97 struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
98 			    u_int8_t, u_int8_t, u_int8_t);
99 
100 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
101 void			 pf_empty_pool(struct pf_palist *);
102 #ifdef ALTQ
103 int			 pf_begin_altq(u_int32_t *);
104 int			 pf_rollback_altq(u_int32_t);
105 int			 pf_commit_altq(u_int32_t);
106 int			 pf_enable_altq(struct pf_altq *);
107 int			 pf_disable_altq(struct pf_altq *);
108 #endif /* ALTQ */
109 int			 pf_begin_rules(u_int32_t *, int, const char *);
110 int			 pf_rollback_rules(u_int32_t, int, char *);
111 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
112 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
113 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
114 int			 pf_commit_rules(u_int32_t, int, char *);
115 int			 pf_addr_setup(struct pf_ruleset *,
116 			    struct pf_addr_wrap *, sa_family_t);
117 void			 pf_addr_copyout(struct pf_addr_wrap *);
118 
119 struct pf_rule		 pf_default_rule;
120 struct lock		 pf_consistency_lock;
121 #ifdef ALTQ
122 static int		 pf_altq_running;
123 #endif
124 
125 #define	TAGID_MAX	 50000
126 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
127 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
128 
129 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
130 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
131 #endif
132 u_int16_t		 tagname2tag(struct pf_tags *, char *);
133 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
134 void			 tag_unref(struct pf_tags *, u_int16_t);
135 int			 pf_rtlabel_add(struct pf_addr_wrap *);
136 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
137 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
138 
139 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
140 
141 static cdev_t	pf_dev;
142 
143 static MALLOC_DEFINE(M_PFRULEPL, "pfrulepl", "pf rule pool list");
144 static MALLOC_DEFINE(M_PFALTQPL, "pfaltqpl", "pf altq pool list");
145 static MALLOC_DEFINE(M_PFPOOLADDRPL, "pfpooladdrpl", "pf pool address pool list");
146 static MALLOC_DEFINE(M_PFFRENTPL, "pffrent", "pf frent pool list");
147 
148 
149 /*
150  * XXX - These are new and need to be checked when moveing to a new version
151  */
152 static void		 pf_clear_states(void);
153 static int		 pf_clear_tables(void);
154 static void		 pf_clear_srcnodes(void);
155 /*
156  * XXX - These are new and need to be checked when moveing to a new version
157  */
158 
159 /*
160  * Wrapper functions for pfil(9) hooks
161  */
162 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
163 		int dir);
164 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
165 		int dir);
166 #ifdef INET6
167 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
168 		int dir);
169 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
170 		int dir);
171 #endif
172 
173 static int 		 hook_pf(void);
174 static int 		 dehook_pf(void);
175 static int 		 shutdown_pf(void);
176 static int 		 pf_load(void);
177 static int 		 pf_unload(void);
178 
179 d_open_t	pfopen;
180 d_close_t	pfclose;
181 d_ioctl_t	pfioctl;
182 
183 static struct dev_ops pf_ops = {	    /* XXX convert to port model */
184 	{ PF_NAME, 73, 0 },
185 	.d_open =	pfopen,
186 	.d_close =	pfclose,
187 	.d_ioctl =	pfioctl
188 };
189 
190 static volatile int pf_pfil_hooked = 0;
191 int pf_end_threads = 0;
192 
193 int debug_pfugidhack = 0;
194 SYSCTL_INT(_debug, OID_AUTO, pfugidhack, CTLFLAG_RW, &debug_pfugidhack, 0,
195 	"Enable/disable pf user/group rules mpsafe hack");
196 
197 void
198 pfattach(void)
199 {
200 	u_int32_t *my_timeout = pf_default_rule.timeout;
201 
202 
203 	if (!rn_inithead((void **)&pf_maskhead, NULL, 0)) {
204 		kprintf("pf mask radix tree create failed\n");
205 		return;
206 	}
207 	kmalloc_create(&pf_state_pl, "pf state pool list");
208 	kmalloc_raise_limit(pf_state_pl, 0);
209 	kmalloc_create(&pf_frent_pl, "pf fragment pool list");
210 	kmalloc_raise_limit(pf_frent_pl, 0);
211 	kmalloc_create(&pf_cent_pl, "pf cent pool list");
212 	kmalloc_raise_limit(pf_cent_pl, 0);
213 
214 	pfr_initialize();
215 	pfi_initialize();
216 	pf_osfp_initialize();
217 
218 	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
219 	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
220 	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
221 	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
222 	if (ctob(physmem) <= 100*1024*1024)
223 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
224 		    PFR_KENTRY_HIWAT_SMALL;
225 
226 	RB_INIT(&tree_src_tracking);
227 	RB_INIT(&pf_anchors);
228 	pf_init_ruleset(&pf_main_ruleset);
229 	TAILQ_INIT(&pf_altqs[0]);
230 	TAILQ_INIT(&pf_altqs[1]);
231 	TAILQ_INIT(&pf_pabuf);
232 	pf_altqs_active = &pf_altqs[0];
233 	pf_altqs_inactive = &pf_altqs[1];
234 	TAILQ_INIT(&state_list);
235 
236 	/* default rule should never be garbage collected */
237 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
238 	pf_default_rule.action = PF_PASS;
239 	pf_default_rule.nr = (uint32_t)(-1);
240 	pf_default_rule.rtableid = -1;
241 
242 	/* initialize default timeouts */
243 	my_timeout[PFTM_TCP_FIRST_PACKET] = 120;	/* First TCP packet */
244 	my_timeout[PFTM_TCP_OPENING] = 30; 		/* No response yet */
245 	my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;	/* Established */
246 	my_timeout[PFTM_TCP_CLOSING] = 15 * 60;		/* Half closed */
247 	my_timeout[PFTM_TCP_FIN_WAIT] = 45;		/* Got both FINs */
248 	my_timeout[PFTM_TCP_CLOSED] = 90;		/* Got a RST */
249 	my_timeout[PFTM_UDP_FIRST_PACKET] = 60;		/* First UDP packet */
250 	my_timeout[PFTM_UDP_SINGLE] = 30;		/* Unidirectional */
251 	my_timeout[PFTM_UDP_MULTIPLE] = 60;		/* Bidirectional */
252 	my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;	/* First ICMP packet */
253 	my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;		/* Got error response */
254 	my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;	/* First packet */
255 	my_timeout[PFTM_OTHER_SINGLE] = 30;		/* Unidirectional */
256 	my_timeout[PFTM_OTHER_MULTIPLE] = 60;		/* Bidirectional */
257 	my_timeout[PFTM_FRAG] = 30;			/* Fragment expire */
258 	my_timeout[PFTM_INTERVAL] = 10;			/* Expire interval */
259 	my_timeout[PFTM_SRC_NODE] = 0;		/* Source Tracking */
260 	my_timeout[PFTM_TS_DIFF] = 30;		/* Allowed TS diff */
261 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
262 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
263 
264 	pf_normalize_init();
265 	bzero(&pf_status, sizeof(pf_status));
266 	pf_status.debug = PF_DEBUG_URGENT;
267 	/* XXX do our best to avoid a conflict */
268 	pf_status.hostid = karc4random();
269 
270 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
271 		panic("pfpurge thread");
272 }
273 
274 int
275 pfopen(struct dev_open_args *ap)
276 {
277 	lwkt_gettoken(&pf_token);
278 	cdev_t dev = ap->a_head.a_dev;
279 	if (minor(dev) >= 1) {
280 		lwkt_reltoken(&pf_token);
281 		return (ENXIO);
282 	}
283 	lwkt_reltoken(&pf_token);
284 	return (0);
285 }
286 
287 int
288 pfclose(struct dev_close_args *ap)
289 {
290 	lwkt_gettoken(&pf_token);
291 	cdev_t dev = ap->a_head.a_dev;
292 	if (minor(dev) >= 1) {
293 		lwkt_reltoken(&pf_token);
294 		return (ENXIO);
295 	}
296 	lwkt_reltoken(&pf_token);
297 	return (0);
298 }
299 
300 struct pf_pool *
301 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
302     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
303     u_int8_t check_ticket)
304 {
305 	struct pf_ruleset	*ruleset;
306 	struct pf_rule		*rule;
307 	int			 rs_num;
308 
309 	ruleset = pf_find_ruleset(anchor);
310 	if (ruleset == NULL)
311 		return (NULL);
312 	rs_num = pf_get_ruleset_number(rule_action);
313 	if (rs_num >= PF_RULESET_MAX)
314 		return (NULL);
315 	if (active) {
316 		if (check_ticket && ticket !=
317 		    ruleset->rules[rs_num].active.ticket)
318 			return (NULL);
319 		if (r_last)
320 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
321 			    pf_rulequeue);
322 		else
323 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
324 	} else {
325 		if (check_ticket && ticket !=
326 		    ruleset->rules[rs_num].inactive.ticket)
327 			return (NULL);
328 		if (r_last)
329 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
330 			    pf_rulequeue);
331 		else
332 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
333 	}
334 	if (!r_last) {
335 		while ((rule != NULL) && (rule->nr != rule_number))
336 			rule = TAILQ_NEXT(rule, entries);
337 	}
338 	if (rule == NULL)
339 		return (NULL);
340 
341 	return (&rule->rpool);
342 }
343 
344 void
345 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
346 {
347 	struct pf_pooladdr	*mv_pool_pa;
348 
349 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
350 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
351 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
352 	}
353 }
354 
355 void
356 pf_empty_pool(struct pf_palist *poola)
357 {
358 	struct pf_pooladdr	*empty_pool_pa;
359 
360 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
361 		pfi_dynaddr_remove(&empty_pool_pa->addr);
362 		pf_tbladdr_remove(&empty_pool_pa->addr);
363 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
364 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
365 		kfree(empty_pool_pa, M_PFPOOLADDRPL);
366 	}
367 }
368 
369 void
370 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
371 {
372 	if (rulequeue != NULL) {
373 		if (rule->states_cur <= 0) {
374 			/*
375 			 * XXX - we need to remove the table *before* detaching
376 			 * the rule to make sure the table code does not delete
377 			 * the anchor under our feet.
378 			 */
379 			pf_tbladdr_remove(&rule->src.addr);
380 			pf_tbladdr_remove(&rule->dst.addr);
381 			if (rule->overload_tbl)
382 				pfr_detach_table(rule->overload_tbl);
383 		}
384 		TAILQ_REMOVE(rulequeue, rule, entries);
385 		rule->entries.tqe_prev = NULL;
386 		rule->nr = -1;
387 	}
388 
389 	if (rule->states_cur > 0 || rule->src_nodes > 0 ||
390 	    rule->entries.tqe_prev != NULL)
391 		return;
392 	pf_tag_unref(rule->tag);
393 	pf_tag_unref(rule->match_tag);
394 #ifdef ALTQ
395 	if (rule->pqid != rule->qid)
396 		pf_qid_unref(rule->pqid);
397 	pf_qid_unref(rule->qid);
398 #endif
399 	pf_rtlabel_remove(&rule->src.addr);
400 	pf_rtlabel_remove(&rule->dst.addr);
401 	pfi_dynaddr_remove(&rule->src.addr);
402 	pfi_dynaddr_remove(&rule->dst.addr);
403 	if (rulequeue == NULL) {
404 		pf_tbladdr_remove(&rule->src.addr);
405 		pf_tbladdr_remove(&rule->dst.addr);
406 		if (rule->overload_tbl)
407 			pfr_detach_table(rule->overload_tbl);
408 	}
409 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
410 	pf_anchor_remove(rule);
411 	pf_empty_pool(&rule->rpool.list);
412 	kfree(rule, M_PFRULEPL);
413 }
414 
415 u_int16_t
416 tagname2tag(struct pf_tags *head, char *tagname)
417 {
418 	struct pf_tagname	*tag, *p = NULL;
419 	u_int16_t		 new_tagid = 1;
420 
421 	TAILQ_FOREACH(tag, head, entries)
422 		if (strcmp(tagname, tag->name) == 0) {
423 			tag->ref++;
424 			return (tag->tag);
425 		}
426 
427 	/*
428 	 * to avoid fragmentation, we do a linear search from the beginning
429 	 * and take the first free slot we find. if there is none or the list
430 	 * is empty, append a new entry at the end.
431 	 */
432 
433 	/* new entry */
434 	if (!TAILQ_EMPTY(head))
435 		for (p = TAILQ_FIRST(head); p != NULL &&
436 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
437 			new_tagid = p->tag + 1;
438 
439 	if (new_tagid > TAGID_MAX)
440 		return (0);
441 
442 	/* allocate and fill new struct pf_tagname */
443 	tag = kmalloc(sizeof(*tag), M_TEMP, M_WAITOK);
444 	strlcpy(tag->name, tagname, sizeof(tag->name));
445 	tag->tag = new_tagid;
446 	tag->ref++;
447 
448 	if (p != NULL)	/* insert new entry before p */
449 		TAILQ_INSERT_BEFORE(p, tag, entries);
450 	else	/* either list empty or no free slot in between */
451 		TAILQ_INSERT_TAIL(head, tag, entries);
452 
453 	return (tag->tag);
454 }
455 
456 void
457 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
458 {
459 	struct pf_tagname	*tag;
460 
461 	TAILQ_FOREACH(tag, head, entries)
462 		if (tag->tag == tagid) {
463 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
464 			return;
465 		}
466 }
467 
468 void
469 tag_unref(struct pf_tags *head, u_int16_t tag)
470 {
471 	struct pf_tagname	*p, *next;
472 
473 	if (tag == 0)
474 		return;
475 
476 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
477 		next = TAILQ_NEXT(p, entries);
478 		if (tag == p->tag) {
479 			if (--p->ref == 0) {
480 				TAILQ_REMOVE(head, p, entries);
481 				kfree(p, M_TEMP);
482 			}
483 			break;
484 		}
485 	}
486 }
487 
488 u_int16_t
489 pf_tagname2tag(char *tagname)
490 {
491 	return (tagname2tag(&pf_tags, tagname));
492 }
493 
494 void
495 pf_tag2tagname(u_int16_t tagid, char *p)
496 {
497 	tag2tagname(&pf_tags, tagid, p);
498 }
499 
500 void
501 pf_tag_ref(u_int16_t tag)
502 {
503 	struct pf_tagname *t;
504 
505 	TAILQ_FOREACH(t, &pf_tags, entries)
506 		if (t->tag == tag)
507 			break;
508 	if (t != NULL)
509 		t->ref++;
510 }
511 
512 void
513 pf_tag_unref(u_int16_t tag)
514 {
515 	tag_unref(&pf_tags, tag);
516 }
517 
518 int
519 pf_rtlabel_add(struct pf_addr_wrap *a)
520 {
521 	return (0);
522 }
523 
524 void
525 pf_rtlabel_remove(struct pf_addr_wrap *a)
526 {
527 }
528 
529 void
530 pf_rtlabel_copyout(struct pf_addr_wrap *a)
531 {
532 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
533 		strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
534 }
535 
536 #ifdef ALTQ
537 u_int32_t
538 pf_qname2qid(char *qname)
539 {
540 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
541 }
542 
543 void
544 pf_qid2qname(u_int32_t qid, char *p)
545 {
546 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
547 }
548 
549 void
550 pf_qid_unref(u_int32_t qid)
551 {
552 	tag_unref(&pf_qids, (u_int16_t)qid);
553 }
554 
555 int
556 pf_begin_altq(u_int32_t *ticket)
557 {
558 	struct pf_altq	*altq;
559 	int		 error = 0;
560 
561 	/* Purge the old altq list */
562 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
563 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
564 		if (altq->qname[0] == 0) {
565 			/* detach and destroy the discipline */
566 			error = altq_remove(altq);
567 		} else
568 			pf_qid_unref(altq->qid);
569 		kfree(altq, M_PFALTQPL);
570 	}
571 	if (error)
572 		return (error);
573 	*ticket = ++ticket_altqs_inactive;
574 	altqs_inactive_open = 1;
575 	return (0);
576 }
577 
578 int
579 pf_rollback_altq(u_int32_t ticket)
580 {
581 	struct pf_altq	*altq;
582 	int		 error = 0;
583 
584 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
585 		return (0);
586 	/* Purge the old altq list */
587 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
588 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
589 		if (altq->qname[0] == 0) {
590 			/* detach and destroy the discipline */
591 			error = altq_remove(altq);
592 		} else
593 			pf_qid_unref(altq->qid);
594 		kfree(altq, M_PFALTQPL);
595 	}
596 	altqs_inactive_open = 0;
597 	return (error);
598 }
599 
600 int
601 pf_commit_altq(u_int32_t ticket)
602 {
603 	struct pf_altqqueue	*old_altqs;
604 	struct pf_altq		*altq;
605 	int			 err, error = 0;
606 
607 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
608 		return (EBUSY);
609 
610 	/* swap altqs, keep the old. */
611 	crit_enter();
612 	old_altqs = pf_altqs_active;
613 	pf_altqs_active = pf_altqs_inactive;
614 	pf_altqs_inactive = old_altqs;
615 	ticket_altqs_active = ticket_altqs_inactive;
616 
617 	/* Attach new disciplines */
618 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
619 		if (altq->qname[0] == 0) {
620 			/* attach the discipline */
621 			error = altq_pfattach(altq);
622 			if (error) {
623 				crit_exit();
624 				return (error);
625 			}
626 		}
627 	}
628 
629 	/* Purge the old altq list */
630 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
631 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
632 		if (altq->qname[0] == 0) {
633 			/* detach and destroy the discipline */
634 			if (pf_altq_running)
635 				error = pf_disable_altq(altq);
636 			err = altq_pfdetach(altq);
637 			if (err != 0 && error == 0)
638 				error = err;
639 			err = altq_remove(altq);
640 			if (err != 0 && error == 0)
641 				error = err;
642 		} else
643 			pf_qid_unref(altq->qid);
644 		kfree(altq, M_PFALTQPL);
645 	}
646 	crit_exit();
647 
648 	altqs_inactive_open = 0;
649 	return (error);
650 }
651 
652 int
653 pf_enable_altq(struct pf_altq *altq)
654 {
655 	struct ifnet		*ifp;
656 	struct tb_profile	 tb;
657 	int			 error = 0;
658 
659 	if ((ifp = ifunit(altq->ifname)) == NULL)
660 		return (EINVAL);
661 
662 	if (ifp->if_snd.altq_type != ALTQT_NONE)
663 		error = altq_enable(&ifp->if_snd);
664 
665 	/* set tokenbucket regulator */
666 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
667 		tb.rate = altq->ifbandwidth;
668 		tb.depth = altq->tbrsize;
669 		crit_enter();
670 		error = tbr_set(&ifp->if_snd, &tb);
671 		crit_exit();
672 	}
673 
674 	return (error);
675 }
676 
677 int
678 pf_disable_altq(struct pf_altq *altq)
679 {
680 	struct ifnet		*ifp;
681 	struct tb_profile	 tb;
682 	int			 error;
683 
684 	if ((ifp = ifunit(altq->ifname)) == NULL)
685 		return (EINVAL);
686 
687 	/*
688 	 * when the discipline is no longer referenced, it was overridden
689 	 * by a new one.  if so, just return.
690 	 */
691 	if (altq->altq_disc != ifp->if_snd.altq_disc)
692 		return (0);
693 
694 	error = altq_disable(&ifp->if_snd);
695 
696 	if (error == 0) {
697 		/* clear tokenbucket regulator */
698 		tb.rate = 0;
699 		crit_enter();
700 		error = tbr_set(&ifp->if_snd, &tb);
701 		crit_exit();
702 	}
703 
704 	return (error);
705 }
706 #endif /* ALTQ */
707 
708 int
709 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
710 {
711 	struct pf_ruleset	*rs;
712 	struct pf_rule		*rule;
713 
714 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
715 		return (EINVAL);
716 	rs = pf_find_or_create_ruleset(anchor);
717 	if (rs == NULL)
718 		return (EINVAL);
719 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
720 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
721 		rs->rules[rs_num].inactive.rcount--;
722 	}
723 	*ticket = ++rs->rules[rs_num].inactive.ticket;
724 	rs->rules[rs_num].inactive.open = 1;
725 	return (0);
726 }
727 
728 int
729 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
730 {
731 	struct pf_ruleset	*rs;
732 	struct pf_rule		*rule;
733 
734 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
735 		return (EINVAL);
736 	rs = pf_find_ruleset(anchor);
737 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
738 	    rs->rules[rs_num].inactive.ticket != ticket)
739 		return (0);
740 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
741 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
742 		rs->rules[rs_num].inactive.rcount--;
743 	}
744 	rs->rules[rs_num].inactive.open = 0;
745 	return (0);
746 }
747 
748 #define PF_MD5_UPD(st, elm)						\
749 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
750 
751 #define PF_MD5_UPD_STR(st, elm)						\
752 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
753 
754 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
755 		(stor) = htonl((st)->elm);				\
756 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
757 } while (0)
758 
759 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
760 		(stor) = htons((st)->elm);				\
761 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
762 } while (0)
763 
764 void
765 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
766 {
767 	PF_MD5_UPD(pfr, addr.type);
768 	switch (pfr->addr.type) {
769 		case PF_ADDR_DYNIFTL:
770 			PF_MD5_UPD(pfr, addr.v.ifname);
771 			PF_MD5_UPD(pfr, addr.iflags);
772 			break;
773 		case PF_ADDR_TABLE:
774 			PF_MD5_UPD(pfr, addr.v.tblname);
775 			break;
776 		case PF_ADDR_ADDRMASK:
777 			/* XXX ignore af? */
778 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
779 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
780 			break;
781 		case PF_ADDR_RTLABEL:
782 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
783 			break;
784 	}
785 
786 	PF_MD5_UPD(pfr, port[0]);
787 	PF_MD5_UPD(pfr, port[1]);
788 	PF_MD5_UPD(pfr, neg);
789 	PF_MD5_UPD(pfr, port_op);
790 }
791 
792 void
793 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
794 {
795 	u_int16_t x;
796 	u_int32_t y;
797 
798 	pf_hash_rule_addr(ctx, &rule->src);
799 	pf_hash_rule_addr(ctx, &rule->dst);
800 	PF_MD5_UPD_STR(rule, label);
801 	PF_MD5_UPD_STR(rule, ifname);
802 	PF_MD5_UPD_STR(rule, match_tagname);
803 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
804 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
805 	PF_MD5_UPD_HTONL(rule, prob, y);
806 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
807 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
808 	PF_MD5_UPD(rule, uid.op);
809 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
810 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
811 	PF_MD5_UPD(rule, gid.op);
812 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
813 	PF_MD5_UPD(rule, action);
814 	PF_MD5_UPD(rule, direction);
815 	PF_MD5_UPD(rule, af);
816 	PF_MD5_UPD(rule, quick);
817 	PF_MD5_UPD(rule, ifnot);
818 	PF_MD5_UPD(rule, match_tag_not);
819 	PF_MD5_UPD(rule, natpass);
820 	PF_MD5_UPD(rule, keep_state);
821 	PF_MD5_UPD(rule, proto);
822 	PF_MD5_UPD(rule, type);
823 	PF_MD5_UPD(rule, code);
824 	PF_MD5_UPD(rule, flags);
825 	PF_MD5_UPD(rule, flagset);
826 	PF_MD5_UPD(rule, allow_opts);
827 	PF_MD5_UPD(rule, rt);
828 	PF_MD5_UPD(rule, tos);
829 }
830 
831 int
832 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
833 {
834 	struct pf_ruleset	*rs;
835 	struct pf_rule		*rule, **old_array;
836 	struct pf_rulequeue	*old_rules;
837 	int			 error;
838 	u_int32_t		 old_rcount;
839 
840 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
841 		return (EINVAL);
842 	rs = pf_find_ruleset(anchor);
843 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
844 	    ticket != rs->rules[rs_num].inactive.ticket)
845 		return (EBUSY);
846 
847 	/* Calculate checksum for the main ruleset */
848 	if (rs == &pf_main_ruleset) {
849 		error = pf_setup_pfsync_matching(rs);
850 		if (error != 0)
851 			return (error);
852 	}
853 
854 	/* Swap rules, keep the old. */
855 	crit_enter();
856 	old_rules = rs->rules[rs_num].active.ptr;
857 	old_rcount = rs->rules[rs_num].active.rcount;
858 	old_array = rs->rules[rs_num].active.ptr_array;
859 
860 	rs->rules[rs_num].active.ptr =
861 	    rs->rules[rs_num].inactive.ptr;
862 	rs->rules[rs_num].active.ptr_array =
863 	    rs->rules[rs_num].inactive.ptr_array;
864 	rs->rules[rs_num].active.rcount =
865 	    rs->rules[rs_num].inactive.rcount;
866 	rs->rules[rs_num].inactive.ptr = old_rules;
867 	rs->rules[rs_num].inactive.ptr_array = old_array;
868 	rs->rules[rs_num].inactive.rcount = old_rcount;
869 
870 	rs->rules[rs_num].active.ticket =
871 	    rs->rules[rs_num].inactive.ticket;
872 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
873 
874 
875 	/* Purge the old rule list. */
876 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
877 		pf_rm_rule(old_rules, rule);
878 	if (rs->rules[rs_num].inactive.ptr_array)
879 		kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
880 	rs->rules[rs_num].inactive.ptr_array = NULL;
881 	rs->rules[rs_num].inactive.rcount = 0;
882 	rs->rules[rs_num].inactive.open = 0;
883 	pf_remove_if_empty_ruleset(rs);
884 	crit_exit();
885 	return (0);
886 }
887 
888 int
889 pf_setup_pfsync_matching(struct pf_ruleset *rs)
890 {
891 	MD5_CTX			 ctx;
892 	struct pf_rule		*rule;
893 	int			 rs_cnt;
894 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
895 
896 	MD5Init(&ctx);
897 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
898 		/* XXX PF_RULESET_SCRUB as well? */
899 		if (rs_cnt == PF_RULESET_SCRUB)
900 			continue;
901 
902 		if (rs->rules[rs_cnt].inactive.ptr_array)
903 			kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
904 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
905 
906 		if (rs->rules[rs_cnt].inactive.rcount) {
907 			rs->rules[rs_cnt].inactive.ptr_array =
908 			    kmalloc(sizeof(caddr_t) *
909 				    rs->rules[rs_cnt].inactive.rcount,
910 				    M_TEMP, M_WAITOK);
911 
912 			if (!rs->rules[rs_cnt].inactive.ptr_array)
913 				return (ENOMEM);
914 		}
915 
916 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
917 		    entries) {
918 			pf_hash_rule(&ctx, rule);
919 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
920 		}
921 	}
922 
923 	MD5Final(digest, &ctx);
924 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
925 	return (0);
926 }
927 
928 int
929 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
930     sa_family_t af)
931 {
932 	if (pfi_dynaddr_setup(addr, af) ||
933 	    pf_tbladdr_setup(ruleset, addr))
934 		return (EINVAL);
935 
936 	return (0);
937 }
938 
939 void
940 pf_addr_copyout(struct pf_addr_wrap *addr)
941 {
942 	pfi_dynaddr_copyout(addr);
943 	pf_tbladdr_copyout(addr);
944 	pf_rtlabel_copyout(addr);
945 }
946 
947 int
948 pfioctl(struct dev_ioctl_args *ap)
949 {
950 	u_long cmd = ap->a_cmd;
951 	caddr_t addr = ap->a_data;
952 	struct pf_pooladdr	*pa = NULL;
953 	struct pf_pool		*pool = NULL;
954 	int			 error = 0;
955 
956 	lwkt_gettoken(&pf_token);
957 
958 	/* XXX keep in sync with switch() below */
959 	if (securelevel > 1)
960 		switch (cmd) {
961 		case DIOCGETRULES:
962 		case DIOCGETRULE:
963 		case DIOCGETADDRS:
964 		case DIOCGETADDR:
965 		case DIOCGETSTATE:
966 		case DIOCSETSTATUSIF:
967 		case DIOCGETSTATUS:
968 		case DIOCCLRSTATUS:
969 		case DIOCNATLOOK:
970 		case DIOCSETDEBUG:
971 		case DIOCGETSTATES:
972 		case DIOCGETTIMEOUT:
973 		case DIOCCLRRULECTRS:
974 		case DIOCGETLIMIT:
975 		case DIOCGETALTQS:
976 		case DIOCGETALTQ:
977 		case DIOCGETQSTATS:
978 		case DIOCGETRULESETS:
979 		case DIOCGETRULESET:
980 		case DIOCRGETTABLES:
981 		case DIOCRGETTSTATS:
982 		case DIOCRCLRTSTATS:
983 		case DIOCRCLRADDRS:
984 		case DIOCRADDADDRS:
985 		case DIOCRDELADDRS:
986 		case DIOCRSETADDRS:
987 		case DIOCRGETADDRS:
988 		case DIOCRGETASTATS:
989 		case DIOCRCLRASTATS:
990 		case DIOCRTSTADDRS:
991 		case DIOCOSFPGET:
992 		case DIOCGETSRCNODES:
993 		case DIOCCLRSRCNODES:
994 		case DIOCIGETIFACES:
995 		case DIOCSETIFFLAG:
996 		case DIOCCLRIFFLAG:
997 		case DIOCGIFSPEED:
998 			break;
999 		case DIOCRCLRTABLES:
1000 		case DIOCRADDTABLES:
1001 		case DIOCRDELTABLES:
1002 		case DIOCRSETTFLAGS:
1003 			if (((struct pfioc_table *)addr)->pfrio_flags &
1004 			    PFR_FLAG_DUMMY)
1005 				break; /* dummy operation ok */
1006 			lwkt_reltoken(&pf_token);
1007 			return (EPERM);
1008 		default:
1009 			lwkt_reltoken(&pf_token);
1010 			return (EPERM);
1011 		}
1012 
1013 	if (!(ap->a_fflag & FWRITE))
1014 		switch (cmd) {
1015 		case DIOCGETRULES:
1016 		case DIOCGETADDRS:
1017 		case DIOCGETADDR:
1018 		case DIOCGETSTATE:
1019 		case DIOCGETSTATUS:
1020 		case DIOCGETSTATES:
1021 		case DIOCGETTIMEOUT:
1022 		case DIOCGETLIMIT:
1023 		case DIOCGETALTQS:
1024 		case DIOCGETALTQ:
1025 		case DIOCGETQSTATS:
1026 		case DIOCGETRULESETS:
1027 		case DIOCGETRULESET:
1028 		case DIOCNATLOOK:
1029 		case DIOCRGETTABLES:
1030 		case DIOCRGETTSTATS:
1031 		case DIOCRGETADDRS:
1032 		case DIOCRGETASTATS:
1033 		case DIOCRTSTADDRS:
1034 		case DIOCOSFPGET:
1035 		case DIOCGETSRCNODES:
1036 		case DIOCIGETIFACES:
1037 		case DIOCGIFSPEED:
1038 			break;
1039 		case DIOCRCLRTABLES:
1040 		case DIOCRADDTABLES:
1041 		case DIOCRDELTABLES:
1042 		case DIOCRCLRTSTATS:
1043 		case DIOCRCLRADDRS:
1044 		case DIOCRADDADDRS:
1045 		case DIOCRDELADDRS:
1046 		case DIOCRSETADDRS:
1047 		case DIOCRSETTFLAGS:
1048 			if (((struct pfioc_table *)addr)->pfrio_flags &
1049 			    PFR_FLAG_DUMMY)
1050 				break; /* dummy operation ok */
1051 			lwkt_reltoken(&pf_token);
1052 			return (EACCES);
1053 		case DIOCGETRULE:
1054 			if (((struct pfioc_rule *)addr)->action ==
1055 			    PF_GET_CLR_CNTR) {
1056 				lwkt_reltoken(&pf_token);
1057 				return (EACCES);
1058 			}
1059 			break;
1060 		default:
1061 			lwkt_reltoken(&pf_token);
1062 			return (EACCES);
1063 		}
1064 
1065 	switch (cmd) {
1066 
1067 	case DIOCSTART:
1068 		if (pf_status.running)
1069 			error = EEXIST;
1070 		else {
1071 			error = hook_pf();
1072 			if (error) {
1073 				DPFPRINTF(PF_DEBUG_MISC,
1074 				    ("pf: pfil registration fail\n"));
1075 				break;
1076 			}
1077 			pf_status.running = 1;
1078 			pf_status.since = time_second;
1079 			if (pf_status.stateid == 0) {
1080 				pf_status.stateid = time_second;
1081 				pf_status.stateid = pf_status.stateid << 32;
1082 			}
1083 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1084 		}
1085 		break;
1086 
1087 	case DIOCSTOP:
1088 		if (!pf_status.running)
1089 			error = ENOENT;
1090 		else {
1091 			pf_status.running = 0;
1092 			error = dehook_pf();
1093 			if (error) {
1094 				pf_status.running = 1;
1095 				DPFPRINTF(PF_DEBUG_MISC,
1096 					("pf: pfil unregistration failed\n"));
1097 			}
1098 			pf_status.since = time_second;
1099 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1100 		}
1101 		break;
1102 
1103 	case DIOCADDRULE: {
1104 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1105 		struct pf_ruleset	*ruleset;
1106 		struct pf_rule		*rule, *tail;
1107 		struct pf_pooladdr	*pa;
1108 		int			 rs_num;
1109 
1110 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1111 		ruleset = pf_find_ruleset(pr->anchor);
1112 		if (ruleset == NULL) {
1113 			error = EINVAL;
1114 			break;
1115 		}
1116 		rs_num = pf_get_ruleset_number(pr->rule.action);
1117 		if (rs_num >= PF_RULESET_MAX) {
1118 			error = EINVAL;
1119 			break;
1120 		}
1121 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1122 			error = EINVAL;
1123 			break;
1124 		}
1125 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1126 			error = EBUSY;
1127 			break;
1128 		}
1129 		if (pr->pool_ticket != ticket_pabuf) {
1130 			error = EBUSY;
1131 			break;
1132 		}
1133 		rule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK);
1134 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1135 		rule->cuid = ap->a_cred->cr_ruid;
1136 		rule->cpid = 0;
1137 		rule->anchor = NULL;
1138 		rule->kif = NULL;
1139 		TAILQ_INIT(&rule->rpool.list);
1140 		/* initialize refcounting */
1141 		rule->states_cur = 0;
1142 		rule->src_nodes = 0;
1143 		rule->entries.tqe_prev = NULL;
1144 #ifndef INET
1145 		if (rule->af == AF_INET) {
1146 			kfree(rule, M_PFRULEPL);
1147 			error = EAFNOSUPPORT;
1148 			break;
1149 		}
1150 #endif /* INET */
1151 #ifndef INET6
1152 		if (rule->af == AF_INET6) {
1153 			kfree(rule, M_PFRULEPL);
1154 			error = EAFNOSUPPORT;
1155 			break;
1156 		}
1157 #endif /* INET6 */
1158 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1159 		    pf_rulequeue);
1160 		if (tail)
1161 			rule->nr = tail->nr + 1;
1162 		else
1163 			rule->nr = 0;
1164 		if (rule->ifname[0]) {
1165 			rule->kif = pfi_kif_get(rule->ifname);
1166 			if (rule->kif == NULL) {
1167 				kfree(rule, M_PFRULEPL);
1168 				error = EINVAL;
1169 				break;
1170 			}
1171 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1172 		}
1173 
1174 		if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1175 			error = EBUSY;
1176 
1177 #ifdef ALTQ
1178 		/* set queue IDs */
1179 		if (rule->qname[0] != 0) {
1180 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1181 				error = EBUSY;
1182 			else if (rule->pqname[0] != 0) {
1183 				if ((rule->pqid =
1184 				    pf_qname2qid(rule->pqname)) == 0)
1185 					error = EBUSY;
1186 			} else
1187 				rule->pqid = rule->qid;
1188 		}
1189 #endif
1190 		if (rule->tagname[0])
1191 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1192 				error = EBUSY;
1193 		if (rule->match_tagname[0])
1194 			if ((rule->match_tag =
1195 			    pf_tagname2tag(rule->match_tagname)) == 0)
1196 				error = EBUSY;
1197 		if (rule->rt && !rule->direction)
1198 			error = EINVAL;
1199 #if NPFLOG > 0
1200 		if (!rule->log)
1201 			rule->logif = 0;
1202 		if (rule->logif >= PFLOGIFS_MAX)
1203 			error = EINVAL;
1204 #endif
1205 		if (pf_rtlabel_add(&rule->src.addr) ||
1206 		    pf_rtlabel_add(&rule->dst.addr))
1207 			error = EBUSY;
1208 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1209 			error = EINVAL;
1210 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1211 			error = EINVAL;
1212 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1213 			error = EINVAL;
1214 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1215 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1216 				error = EINVAL;
1217 
1218 		if (rule->overload_tblname[0]) {
1219 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1220 			    rule->overload_tblname)) == NULL)
1221 				error = EINVAL;
1222 			else
1223 				rule->overload_tbl->pfrkt_flags |=
1224 				    PFR_TFLAG_ACTIVE;
1225 		}
1226 
1227 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1228 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1229 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1230 		    (rule->rt > PF_FASTROUTE)) &&
1231 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1232 			error = EINVAL;
1233 
1234 		if (error) {
1235 			pf_rm_rule(NULL, rule);
1236 			break;
1237 		}
1238 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1239 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1240 		    rule->bytes[0] = rule->bytes[1] = 0;
1241 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1242 		    rule, entries);
1243 		ruleset->rules[rs_num].inactive.rcount++;
1244 		break;
1245 	}
1246 
1247 	case DIOCGETRULES: {
1248 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1249 		struct pf_ruleset	*ruleset;
1250 		struct pf_rule		*tail;
1251 		int			 rs_num;
1252 
1253 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1254 		ruleset = pf_find_ruleset(pr->anchor);
1255 		if (ruleset == NULL) {
1256 			error = EINVAL;
1257 			break;
1258 		}
1259 		rs_num = pf_get_ruleset_number(pr->rule.action);
1260 		if (rs_num >= PF_RULESET_MAX) {
1261 			error = EINVAL;
1262 			break;
1263 		}
1264 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1265 		    pf_rulequeue);
1266 		if (tail)
1267 			pr->nr = tail->nr + 1;
1268 		else
1269 			pr->nr = 0;
1270 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1271 		break;
1272 	}
1273 
1274 	case DIOCGETRULE: {
1275 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1276 		struct pf_ruleset	*ruleset;
1277 		struct pf_rule		*rule;
1278 		int			 rs_num, i;
1279 
1280 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1281 		ruleset = pf_find_ruleset(pr->anchor);
1282 		if (ruleset == NULL) {
1283 			error = EINVAL;
1284 			break;
1285 		}
1286 		rs_num = pf_get_ruleset_number(pr->rule.action);
1287 		if (rs_num >= PF_RULESET_MAX) {
1288 			error = EINVAL;
1289 			break;
1290 		}
1291 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1292 			error = EBUSY;
1293 			break;
1294 		}
1295 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1296 		while ((rule != NULL) && (rule->nr != pr->nr))
1297 			rule = TAILQ_NEXT(rule, entries);
1298 		if (rule == NULL) {
1299 			error = EBUSY;
1300 			break;
1301 		}
1302 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1303 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1304 			error = EBUSY;
1305 			break;
1306 		}
1307 		pf_addr_copyout(&pr->rule.src.addr);
1308 		pf_addr_copyout(&pr->rule.dst.addr);
1309 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1310 			if (rule->skip[i].ptr == NULL)
1311 				pr->rule.skip[i].nr = (uint32_t)(-1);
1312 			else
1313 				pr->rule.skip[i].nr =
1314 				    rule->skip[i].ptr->nr;
1315 
1316 		if (pr->action == PF_GET_CLR_CNTR) {
1317 			rule->evaluations = 0;
1318 			rule->packets[0] = rule->packets[1] = 0;
1319 			rule->bytes[0] = rule->bytes[1] = 0;
1320 			rule->states_tot = 0;
1321 		}
1322 		break;
1323 	}
1324 
1325 	case DIOCCHANGERULE: {
1326 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1327 		struct pf_ruleset	*ruleset;
1328 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1329 		u_int32_t		 nr = 0;
1330 		int			 rs_num;
1331 
1332 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1333 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1334 		    pcr->pool_ticket != ticket_pabuf) {
1335 			error = EBUSY;
1336 			break;
1337 		}
1338 
1339 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1340 		    pcr->action > PF_CHANGE_GET_TICKET) {
1341 			error = EINVAL;
1342 			break;
1343 		}
1344 		ruleset = pf_find_ruleset(pcr->anchor);
1345 		if (ruleset == NULL) {
1346 			error = EINVAL;
1347 			break;
1348 		}
1349 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1350 		if (rs_num >= PF_RULESET_MAX) {
1351 			error = EINVAL;
1352 			break;
1353 		}
1354 
1355 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1356 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1357 			break;
1358 		} else {
1359 			if (pcr->ticket !=
1360 			    ruleset->rules[rs_num].active.ticket) {
1361 				error = EINVAL;
1362 				break;
1363 			}
1364 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1365 				error = EINVAL;
1366 				break;
1367 			}
1368 		}
1369 
1370 		if (pcr->action != PF_CHANGE_REMOVE) {
1371 			newrule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK|M_NULLOK);
1372 			if (newrule == NULL) {
1373 				error = ENOMEM;
1374 				break;
1375 			}
1376 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1377 			newrule->cuid = ap->a_cred->cr_ruid;
1378 			newrule->cpid = 0;
1379 			TAILQ_INIT(&newrule->rpool.list);
1380 			/* initialize refcounting */
1381 			newrule->states_cur = 0;
1382 			newrule->entries.tqe_prev = NULL;
1383 #ifndef INET
1384 			if (newrule->af == AF_INET) {
1385 				kfree(newrule, M_PFRULEPL);
1386 				error = EAFNOSUPPORT;
1387 				break;
1388 			}
1389 #endif /* INET */
1390 #ifndef INET6
1391 			if (newrule->af == AF_INET6) {
1392 				kfree(newrule, M_PFRULEPL);
1393 				error = EAFNOSUPPORT;
1394 				break;
1395 			}
1396 #endif /* INET6 */
1397 			if (newrule->ifname[0]) {
1398 				newrule->kif = pfi_kif_get(newrule->ifname);
1399 				if (newrule->kif == NULL) {
1400 					kfree(newrule, M_PFRULEPL);
1401 					error = EINVAL;
1402 					break;
1403 				}
1404 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1405 			} else
1406 				newrule->kif = NULL;
1407 
1408 			if (newrule->rtableid > 0 &&
1409 			    newrule->rtableid > rt_numfibs)
1410 				error = EBUSY;
1411 
1412 #ifdef ALTQ
1413 			/* set queue IDs */
1414 			if (newrule->qname[0] != 0) {
1415 				if ((newrule->qid =
1416 				    pf_qname2qid(newrule->qname)) == 0)
1417 					error = EBUSY;
1418 				else if (newrule->pqname[0] != 0) {
1419 					if ((newrule->pqid =
1420 					    pf_qname2qid(newrule->pqname)) == 0)
1421 						error = EBUSY;
1422 				} else
1423 					newrule->pqid = newrule->qid;
1424 			}
1425 #endif /* ALTQ */
1426 			if (newrule->tagname[0])
1427 				if ((newrule->tag =
1428 				    pf_tagname2tag(newrule->tagname)) == 0)
1429 					error = EBUSY;
1430 			if (newrule->match_tagname[0])
1431 				if ((newrule->match_tag = pf_tagname2tag(
1432 				    newrule->match_tagname)) == 0)
1433 					error = EBUSY;
1434 			if (newrule->rt && !newrule->direction)
1435 				error = EINVAL;
1436 #if NPFLOG > 0
1437 			if (!newrule->log)
1438 				newrule->logif = 0;
1439 			if (newrule->logif >= PFLOGIFS_MAX)
1440 				error = EINVAL;
1441 #endif
1442 			if (pf_rtlabel_add(&newrule->src.addr) ||
1443 			    pf_rtlabel_add(&newrule->dst.addr))
1444 				error = EBUSY;
1445 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1446 				error = EINVAL;
1447 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1448 				error = EINVAL;
1449 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1450 				error = EINVAL;
1451 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1452 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1453 					error = EINVAL;
1454 
1455 			if (newrule->overload_tblname[0]) {
1456 				if ((newrule->overload_tbl = pfr_attach_table(
1457 				    ruleset, newrule->overload_tblname)) ==
1458 				    NULL)
1459 					error = EINVAL;
1460 				else
1461 					newrule->overload_tbl->pfrkt_flags |=
1462 					    PFR_TFLAG_ACTIVE;
1463 			}
1464 
1465 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1466 			if (((((newrule->action == PF_NAT) ||
1467 			    (newrule->action == PF_RDR) ||
1468 			    (newrule->action == PF_BINAT) ||
1469 			    (newrule->rt > PF_FASTROUTE)) &&
1470 			    !newrule->anchor)) &&
1471 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1472 				error = EINVAL;
1473 
1474 			if (error) {
1475 				pf_rm_rule(NULL, newrule);
1476 				break;
1477 			}
1478 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1479 			newrule->evaluations = 0;
1480 			newrule->packets[0] = newrule->packets[1] = 0;
1481 			newrule->bytes[0] = newrule->bytes[1] = 0;
1482 		}
1483 		pf_empty_pool(&pf_pabuf);
1484 
1485 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1486 			oldrule = TAILQ_FIRST(
1487 			    ruleset->rules[rs_num].active.ptr);
1488 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1489 			oldrule = TAILQ_LAST(
1490 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1491 		else {
1492 			oldrule = TAILQ_FIRST(
1493 			    ruleset->rules[rs_num].active.ptr);
1494 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1495 				oldrule = TAILQ_NEXT(oldrule, entries);
1496 			if (oldrule == NULL) {
1497 				if (newrule != NULL)
1498 					pf_rm_rule(NULL, newrule);
1499 				error = EINVAL;
1500 				break;
1501 			}
1502 		}
1503 
1504 		if (pcr->action == PF_CHANGE_REMOVE) {
1505 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1506 			ruleset->rules[rs_num].active.rcount--;
1507 		} else {
1508 			if (oldrule == NULL)
1509 				TAILQ_INSERT_TAIL(
1510 				    ruleset->rules[rs_num].active.ptr,
1511 				    newrule, entries);
1512 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1513 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1514 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1515 			else
1516 				TAILQ_INSERT_AFTER(
1517 				    ruleset->rules[rs_num].active.ptr,
1518 				    oldrule, newrule, entries);
1519 			ruleset->rules[rs_num].active.rcount++;
1520 		}
1521 
1522 		nr = 0;
1523 		TAILQ_FOREACH(oldrule,
1524 		    ruleset->rules[rs_num].active.ptr, entries)
1525 			oldrule->nr = nr++;
1526 
1527 		ruleset->rules[rs_num].active.ticket++;
1528 
1529 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1530 		pf_remove_if_empty_ruleset(ruleset);
1531 
1532 		break;
1533 	}
1534 
1535 	case DIOCCLRSTATES: {
1536 		struct pf_state		*s, *nexts;
1537 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1538 		u_int			 killed = 0;
1539 
1540 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
1541 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1542 
1543 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1544 			    s->kif->pfik_name)) {
1545 				/* don't send out individual delete messages */
1546 				s->sync_flags = PFSTATE_NOSYNC;
1547 				pf_unlink_state(s);
1548 				killed++;
1549 			}
1550 		}
1551 		psk->psk_killed = killed;
1552 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1553 		break;
1554 	}
1555 
1556 	case DIOCKILLSTATES: {
1557 		struct pf_state		*s, *nexts;
1558 		struct pf_state_key	*sk;
1559 		struct pf_addr		*srcaddr, *dstaddr;
1560 		u_int16_t		 srcport, dstport;
1561 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1562 		u_int			 killed = 0;
1563 
1564 		if (psk->psk_pfcmp.id) {
1565 			if (psk->psk_pfcmp.creatorid == 0)
1566 				psk->psk_pfcmp.creatorid = pf_status.hostid;
1567 			if ((s = pf_find_state_byid(&psk->psk_pfcmp))) {
1568 				/* send immediate delete of state */
1569 				pfsync_delete_state(s);
1570 				s->sync_flags |= PFSTATE_NOSYNC;
1571 				pf_unlink_state(s);
1572 				psk->psk_killed = 1;
1573 			}
1574 			break;
1575 		}
1576 
1577 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s;
1578 		    s = nexts) {
1579 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1580 			sk = s->key[PF_SK_WIRE];
1581 
1582 			if (s->direction == PF_OUT) {
1583 				srcaddr = &sk->addr[1];
1584 				dstaddr = &sk->addr[0];
1585 				srcport = sk->port[0];
1586 				dstport = sk->port[0];
1587 			} else {
1588 				srcaddr = &sk->addr[0];
1589 				dstaddr = &sk->addr[1];
1590 				srcport = sk->port[0];
1591 				dstport = sk->port[0];
1592 			}
1593 			if ((!psk->psk_af || sk->af == psk->psk_af)
1594 			    && (!psk->psk_proto || psk->psk_proto ==
1595 			    sk->proto) &&
1596 			    PF_MATCHA(psk->psk_src.neg,
1597 			    &psk->psk_src.addr.v.a.addr,
1598 			    &psk->psk_src.addr.v.a.mask,
1599 			    srcaddr, sk->af) &&
1600 			    PF_MATCHA(psk->psk_dst.neg,
1601 			    &psk->psk_dst.addr.v.a.addr,
1602 			    &psk->psk_dst.addr.v.a.mask,
1603 			    dstaddr, sk->af) &&
1604 			    (psk->psk_src.port_op == 0 ||
1605 			    pf_match_port(psk->psk_src.port_op,
1606 			    psk->psk_src.port[0], psk->psk_src.port[1],
1607 			    srcport)) &&
1608 			    (psk->psk_dst.port_op == 0 ||
1609 			    pf_match_port(psk->psk_dst.port_op,
1610 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1611 			    dstport)) &&
1612 			    (!psk->psk_label[0] || (s->rule.ptr->label[0] &&
1613 			    !strcmp(psk->psk_label, s->rule.ptr->label))) &&
1614 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1615 			    s->kif->pfik_name))) {
1616 				/* send immediate delete of state */
1617 				pfsync_delete_state(s);
1618 				s->sync_flags |= PFSTATE_NOSYNC;
1619 				pf_unlink_state(s);
1620 				killed++;
1621 			}
1622 		}
1623 		psk->psk_killed = killed;
1624 		break;
1625 	}
1626 
1627 	case DIOCADDSTATE: {
1628 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1629 		struct pfsync_state	*sp = &ps->state;
1630 
1631 		if (sp->timeout >= PFTM_MAX &&
1632 		    sp->timeout != PFTM_UNTIL_PACKET) {
1633 			error = EINVAL;
1634 			break;
1635 		}
1636 		error = pfsync_state_import(sp, PFSYNC_SI_IOCTL);
1637 		break;
1638 	}
1639 
1640 	case DIOCGETSTATE: {
1641 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1642 		struct pf_state		*s;
1643 		struct pf_state_cmp	 id_key;
1644 
1645 		bcopy(ps->state.id, &id_key.id, sizeof(id_key.id));
1646 		id_key.creatorid = ps->state.creatorid;
1647 
1648 		s = pf_find_state_byid(&id_key);
1649 		if (s == NULL) {
1650 			error = ENOENT;
1651 			break;
1652 		}
1653 		pfsync_state_export(&ps->state, s);
1654 		break;
1655 	}
1656 
1657 	case DIOCGETSTATES: {
1658 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1659 		struct pf_state		*state;
1660 		struct pfsync_state	*p, *pstore;
1661 		u_int32_t		 nr = 0;
1662 
1663 		if (ps->ps_len == 0) {
1664 			nr = pf_status.states;
1665 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1666 			break;
1667 		}
1668 
1669 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1670 
1671 		p = ps->ps_states;
1672 
1673 		state = TAILQ_FIRST(&state_list);
1674 		while (state) {
1675 			if (state->timeout != PFTM_UNLINKED) {
1676 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1677 					break;
1678 				pfsync_state_export(pstore, state);
1679 				error = copyout(pstore, p, sizeof(*p));
1680 				if (error) {
1681 					kfree(pstore, M_TEMP);
1682 					goto fail;
1683 				}
1684 				p++;
1685 				nr++;
1686 			}
1687 			state = TAILQ_NEXT(state, entry_list);
1688 		}
1689 
1690 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1691 
1692 		kfree(pstore, M_TEMP);
1693 		break;
1694 	}
1695 
1696 	case DIOCGETSTATUS: {
1697 		struct pf_status *s = (struct pf_status *)addr;
1698 		bcopy(&pf_status, s, sizeof(struct pf_status));
1699 		pfi_update_status(s->ifname, s);
1700 		break;
1701 	}
1702 
1703 	case DIOCSETSTATUSIF: {
1704 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1705 
1706 		if (pi->ifname[0] == 0) {
1707 			bzero(pf_status.ifname, IFNAMSIZ);
1708 			break;
1709 		}
1710 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1711 		break;
1712 	}
1713 
1714 	case DIOCCLRSTATUS: {
1715 		bzero(pf_status.counters, sizeof(pf_status.counters));
1716 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1717 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
1718 		pf_status.since = time_second;
1719 		if (*pf_status.ifname)
1720 			pfi_update_status(pf_status.ifname, NULL);
1721 		break;
1722 	}
1723 
1724 	case DIOCNATLOOK: {
1725 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1726 		struct pf_state_key	*sk;
1727 		struct pf_state		*state;
1728 		struct pf_state_key_cmp	 key;
1729 		int			 m = 0, direction = pnl->direction;
1730 		int			 sidx, didx;
1731 
1732 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1733 		sidx = (direction == PF_IN) ? 1 : 0;
1734 		didx = (direction == PF_IN) ? 0 : 1;
1735 
1736 		if (!pnl->proto ||
1737 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1738 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1739 		    ((pnl->proto == IPPROTO_TCP ||
1740 		    pnl->proto == IPPROTO_UDP) &&
1741 		    (!pnl->dport || !pnl->sport)))
1742 			error = EINVAL;
1743 		else {
1744 			key.af = pnl->af;
1745 			key.proto = pnl->proto;
1746 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1747 			key.port[sidx] = pnl->sport;
1748 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1749 			key.port[didx] = pnl->dport;
1750 
1751 			state = pf_find_state_all(&key, direction, &m);
1752 
1753 			if (m > 1)
1754 				error = E2BIG;	/* more than one state */
1755 			else if (state != NULL) {
1756 				sk = state->key[sidx];
1757 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1758 				pnl->rsport = sk->port[sidx];
1759 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1760 				pnl->rdport = sk->port[didx];
1761 			} else
1762 				error = ENOENT;
1763 		}
1764 		break;
1765 	}
1766 
1767 	case DIOCSETTIMEOUT: {
1768 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1769 		int		 old;
1770 
1771 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1772 		    pt->seconds < 0) {
1773 			error = EINVAL;
1774 			goto fail;
1775 		}
1776 		old = pf_default_rule.timeout[pt->timeout];
1777 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1778 			pt->seconds = 1;
1779 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
1780 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1781 			wakeup(pf_purge_thread);
1782 		pt->seconds = old;
1783 		break;
1784 	}
1785 
1786 	case DIOCGETTIMEOUT: {
1787 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1788 
1789 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1790 			error = EINVAL;
1791 			goto fail;
1792 		}
1793 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1794 		break;
1795 	}
1796 
1797 	case DIOCGETLIMIT: {
1798 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1799 
1800 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1801 			error = EINVAL;
1802 			goto fail;
1803 		}
1804 		pl->limit = pf_pool_limits[pl->index].limit;
1805 		break;
1806 	}
1807 
1808 	case DIOCSETLIMIT: {
1809 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1810 		int			 old_limit;
1811 
1812 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1813 		    pf_pool_limits[pl->index].pp == NULL) {
1814 			error = EINVAL;
1815 			goto fail;
1816 		}
1817 
1818 		/* XXX Get an API to set limits on the zone/pool */
1819 		old_limit = pf_pool_limits[pl->index].limit;
1820 		pf_pool_limits[pl->index].limit = pl->limit;
1821 		pl->limit = old_limit;
1822 		break;
1823 	}
1824 
1825 	case DIOCSETDEBUG: {
1826 		u_int32_t	*level = (u_int32_t *)addr;
1827 
1828 		pf_status.debug = *level;
1829 		break;
1830 	}
1831 
1832 	case DIOCCLRRULECTRS: {
1833 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1834 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1835 		struct pf_rule		*rule;
1836 
1837 		TAILQ_FOREACH(rule,
1838 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1839 			rule->evaluations = 0;
1840 			rule->packets[0] = rule->packets[1] = 0;
1841 			rule->bytes[0] = rule->bytes[1] = 0;
1842 		}
1843 		break;
1844 	}
1845 
1846 	case DIOCGIFSPEED: {
1847 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1848 		struct pf_ifspeed	ps;
1849 		struct ifnet		*ifp;
1850 
1851 		if (psp->ifname[0] != 0) {
1852 			/* Can we completely trust user-land? */
1853 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1854 			ifp = ifunit(ps.ifname);
1855 			if (ifp )
1856 				psp->baudrate = ifp->if_baudrate;
1857 			else
1858 				error = EINVAL;
1859 		} else
1860 			error = EINVAL;
1861 		break;
1862 	}
1863 #ifdef ALTQ
1864 	case DIOCSTARTALTQ: {
1865 		struct pf_altq		*altq;
1866 
1867 		/* enable all altq interfaces on active list */
1868 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1869 			if (altq->qname[0] == 0) {
1870 				error = pf_enable_altq(altq);
1871 				if (error != 0)
1872 					break;
1873 			}
1874 		}
1875 		if (error == 0)
1876 			pf_altq_running = 1;
1877 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1878 		break;
1879 	}
1880 
1881 	case DIOCSTOPALTQ: {
1882 		struct pf_altq		*altq;
1883 
1884 		/* disable all altq interfaces on active list */
1885 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1886 			if (altq->qname[0] == 0) {
1887 				error = pf_disable_altq(altq);
1888 				if (error != 0)
1889 					break;
1890 			}
1891 		}
1892 		if (error == 0)
1893 			pf_altq_running = 0;
1894 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1895 		break;
1896 	}
1897 
1898 	case DIOCADDALTQ: {
1899 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1900 		struct pf_altq		*altq, *a;
1901 
1902 		if (pa->ticket != ticket_altqs_inactive) {
1903 			error = EBUSY;
1904 			break;
1905 		}
1906 		altq = kmalloc(sizeof(struct pf_altq), M_PFALTQPL, M_WAITOK|M_NULLOK);
1907 		if (altq == NULL) {
1908 			error = ENOMEM;
1909 			break;
1910 		}
1911 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
1912 
1913 		/*
1914 		 * if this is for a queue, find the discipline and
1915 		 * copy the necessary fields
1916 		 */
1917 		if (altq->qname[0] != 0) {
1918 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
1919 				error = EBUSY;
1920 				kfree(altq, M_PFALTQPL);
1921 				break;
1922 			}
1923 			altq->altq_disc = NULL;
1924 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
1925 				if (strncmp(a->ifname, altq->ifname,
1926 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
1927 					altq->altq_disc = a->altq_disc;
1928 					break;
1929 				}
1930 			}
1931 		}
1932 
1933 		error = altq_add(altq);
1934 		if (error) {
1935 			kfree(altq, M_PFALTQPL);
1936 			break;
1937 		}
1938 
1939 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
1940 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1941 		break;
1942 	}
1943 
1944 	case DIOCGETALTQS: {
1945 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1946 		struct pf_altq		*altq;
1947 
1948 		pa->nr = 0;
1949 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
1950 			pa->nr++;
1951 		pa->ticket = ticket_altqs_active;
1952 		break;
1953 	}
1954 
1955 	case DIOCGETALTQ: {
1956 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1957 		struct pf_altq		*altq;
1958 		u_int32_t		 nr;
1959 
1960 		if (pa->ticket != ticket_altqs_active) {
1961 			error = EBUSY;
1962 			break;
1963 		}
1964 		nr = 0;
1965 		altq = TAILQ_FIRST(pf_altqs_active);
1966 		while ((altq != NULL) && (nr < pa->nr)) {
1967 			altq = TAILQ_NEXT(altq, entries);
1968 			nr++;
1969 		}
1970 		if (altq == NULL) {
1971 			error = EBUSY;
1972 			break;
1973 		}
1974 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1975 		break;
1976 	}
1977 
1978 	case DIOCCHANGEALTQ:
1979 		/* CHANGEALTQ not supported yet! */
1980 		error = ENODEV;
1981 		break;
1982 
1983 	case DIOCGETQSTATS: {
1984 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
1985 		struct pf_altq		*altq;
1986 		u_int32_t		 nr;
1987 		int			 nbytes;
1988 
1989 		if (pq->ticket != ticket_altqs_active) {
1990 			error = EBUSY;
1991 			break;
1992 		}
1993 		nbytes = pq->nbytes;
1994 		nr = 0;
1995 		altq = TAILQ_FIRST(pf_altqs_active);
1996 		while ((altq != NULL) && (nr < pq->nr)) {
1997 			altq = TAILQ_NEXT(altq, entries);
1998 			nr++;
1999 		}
2000 		if (altq == NULL) {
2001 			error = EBUSY;
2002 			break;
2003 		}
2004 		error = altq_getqstats(altq, pq->buf, &nbytes);
2005 		if (error == 0) {
2006 			pq->scheduler = altq->scheduler;
2007 			pq->nbytes = nbytes;
2008 		}
2009 		break;
2010 	}
2011 #endif /* ALTQ */
2012 
2013 	case DIOCBEGINADDRS: {
2014 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2015 
2016 		pf_empty_pool(&pf_pabuf);
2017 		pp->ticket = ++ticket_pabuf;
2018 		break;
2019 	}
2020 
2021 	case DIOCADDADDR: {
2022 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2023 
2024 		if (pp->ticket != ticket_pabuf) {
2025 			error = EBUSY;
2026 			break;
2027 		}
2028 #ifndef INET
2029 		if (pp->af == AF_INET) {
2030 			error = EAFNOSUPPORT;
2031 			break;
2032 		}
2033 #endif /* INET */
2034 #ifndef INET6
2035 		if (pp->af == AF_INET6) {
2036 			error = EAFNOSUPPORT;
2037 			break;
2038 		}
2039 #endif /* INET6 */
2040 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2041 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2042 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2043 			error = EINVAL;
2044 			break;
2045 		}
2046 		pa = kmalloc(sizeof(struct pf_altq), M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2047 		if (pa == NULL) {
2048 			error = ENOMEM;
2049 			break;
2050 		}
2051 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2052 		if (pa->ifname[0]) {
2053 			pa->kif = pfi_kif_get(pa->ifname);
2054 			if (pa->kif == NULL) {
2055 				kfree(ap, M_PFPOOLADDRPL);
2056 				error = EINVAL;
2057 				break;
2058 			}
2059 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2060 		}
2061 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2062 			pfi_dynaddr_remove(&pa->addr);
2063 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2064 			kfree(pa, M_PFPOOLADDRPL);
2065 			error = EINVAL;
2066 			break;
2067 		}
2068 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2069 		break;
2070 	}
2071 
2072 	case DIOCGETADDRS: {
2073 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2074 
2075 		pp->nr = 0;
2076 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2077 		    pp->r_num, 0, 1, 0);
2078 		if (pool == NULL) {
2079 			error = EBUSY;
2080 			break;
2081 		}
2082 		TAILQ_FOREACH(pa, &pool->list, entries)
2083 			pp->nr++;
2084 		break;
2085 	}
2086 
2087 	case DIOCGETADDR: {
2088 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2089 		u_int32_t		 nr = 0;
2090 
2091 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2092 		    pp->r_num, 0, 1, 1);
2093 		if (pool == NULL) {
2094 			error = EBUSY;
2095 			break;
2096 		}
2097 		pa = TAILQ_FIRST(&pool->list);
2098 		while ((pa != NULL) && (nr < pp->nr)) {
2099 			pa = TAILQ_NEXT(pa, entries);
2100 			nr++;
2101 		}
2102 		if (pa == NULL) {
2103 			error = EBUSY;
2104 			break;
2105 		}
2106 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2107 		pf_addr_copyout(&pp->addr.addr);
2108 		break;
2109 	}
2110 
2111 	case DIOCCHANGEADDR: {
2112 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2113 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2114 		struct pf_ruleset	*ruleset;
2115 
2116 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2117 		    pca->action > PF_CHANGE_REMOVE) {
2118 			error = EINVAL;
2119 			break;
2120 		}
2121 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2122 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2123 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2124 			error = EINVAL;
2125 			break;
2126 		}
2127 
2128 		ruleset = pf_find_ruleset(pca->anchor);
2129 		if (ruleset == NULL) {
2130 			error = EBUSY;
2131 			break;
2132 		}
2133 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2134 		    pca->r_num, pca->r_last, 1, 1);
2135 		if (pool == NULL) {
2136 			error = EBUSY;
2137 			break;
2138 		}
2139 		if (pca->action != PF_CHANGE_REMOVE) {
2140 			newpa = kmalloc(sizeof(struct pf_pooladdr),
2141 				M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2142 			if (newpa == NULL) {
2143 				error = ENOMEM;
2144 				break;
2145 			}
2146 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2147 #ifndef INET
2148 			if (pca->af == AF_INET) {
2149 				kfree(newpa, M_PFPOOLADDRPL);
2150 				error = EAFNOSUPPORT;
2151 				break;
2152 			}
2153 #endif /* INET */
2154 #ifndef INET6
2155 			if (pca->af == AF_INET6) {
2156 				kfree(newpa, M_PFPOOLADDRPL);
2157 				error = EAFNOSUPPORT;
2158 				break;
2159 			}
2160 #endif /* INET6 */
2161 			if (newpa->ifname[0]) {
2162 				newpa->kif = pfi_kif_get(newpa->ifname);
2163 				if (newpa->kif == NULL) {
2164 					kfree(newpa, M_PFPOOLADDRPL);
2165 					error = EINVAL;
2166 					break;
2167 				}
2168 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2169 			} else
2170 				newpa->kif = NULL;
2171 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2172 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2173 				pfi_dynaddr_remove(&newpa->addr);
2174 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2175 				kfree(newpa, M_PFPOOLADDRPL);
2176 				error = EINVAL;
2177 				break;
2178 			}
2179 		}
2180 
2181 		if (pca->action == PF_CHANGE_ADD_HEAD)
2182 			oldpa = TAILQ_FIRST(&pool->list);
2183 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2184 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2185 		else {
2186 			int	i = 0;
2187 
2188 			oldpa = TAILQ_FIRST(&pool->list);
2189 			while ((oldpa != NULL) && (i < pca->nr)) {
2190 				oldpa = TAILQ_NEXT(oldpa, entries);
2191 				i++;
2192 			}
2193 			if (oldpa == NULL) {
2194 				error = EINVAL;
2195 				break;
2196 			}
2197 		}
2198 
2199 		if (pca->action == PF_CHANGE_REMOVE) {
2200 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2201 			pfi_dynaddr_remove(&oldpa->addr);
2202 			pf_tbladdr_remove(&oldpa->addr);
2203 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2204 			kfree(oldpa, M_PFPOOLADDRPL);
2205 		} else {
2206 			if (oldpa == NULL)
2207 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2208 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2209 			    pca->action == PF_CHANGE_ADD_BEFORE)
2210 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2211 			else
2212 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2213 				    newpa, entries);
2214 		}
2215 
2216 		pool->cur = TAILQ_FIRST(&pool->list);
2217 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2218 		    pca->af);
2219 		break;
2220 	}
2221 
2222 	case DIOCGETRULESETS: {
2223 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2224 		struct pf_ruleset	*ruleset;
2225 		struct pf_anchor	*anchor;
2226 
2227 		pr->path[sizeof(pr->path) - 1] = 0;
2228 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2229 			error = EINVAL;
2230 			break;
2231 		}
2232 		pr->nr = 0;
2233 		if (ruleset->anchor == NULL) {
2234 			/* XXX kludge for pf_main_ruleset */
2235 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2236 				if (anchor->parent == NULL)
2237 					pr->nr++;
2238 		} else {
2239 			RB_FOREACH(anchor, pf_anchor_node,
2240 			    &ruleset->anchor->children)
2241 				pr->nr++;
2242 		}
2243 		break;
2244 	}
2245 
2246 	case DIOCGETRULESET: {
2247 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2248 		struct pf_ruleset	*ruleset;
2249 		struct pf_anchor	*anchor;
2250 		u_int32_t		 nr = 0;
2251 
2252 		pr->path[sizeof(pr->path) - 1] = 0;
2253 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2254 			error = EINVAL;
2255 			break;
2256 		}
2257 		pr->name[0] = 0;
2258 		if (ruleset->anchor == NULL) {
2259 			/* XXX kludge for pf_main_ruleset */
2260 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2261 				if (anchor->parent == NULL && nr++ == pr->nr) {
2262 					strlcpy(pr->name, anchor->name,
2263 					    sizeof(pr->name));
2264 					break;
2265 				}
2266 		} else {
2267 			RB_FOREACH(anchor, pf_anchor_node,
2268 			    &ruleset->anchor->children)
2269 				if (nr++ == pr->nr) {
2270 					strlcpy(pr->name, anchor->name,
2271 					    sizeof(pr->name));
2272 					break;
2273 				}
2274 		}
2275 		if (!pr->name[0])
2276 			error = EBUSY;
2277 		break;
2278 	}
2279 
2280 	case DIOCRCLRTABLES: {
2281 		struct pfioc_table *io = (struct pfioc_table *)addr;
2282 
2283 		if (io->pfrio_esize != 0) {
2284 			error = ENODEV;
2285 			break;
2286 		}
2287 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2288 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2289 		break;
2290 	}
2291 
2292 	case DIOCRADDTABLES: {
2293 		struct pfioc_table *io = (struct pfioc_table *)addr;
2294 
2295 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2296 			error = ENODEV;
2297 			break;
2298 		}
2299 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2300 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2301 		break;
2302 	}
2303 
2304 	case DIOCRDELTABLES: {
2305 		struct pfioc_table *io = (struct pfioc_table *)addr;
2306 
2307 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2308 			error = ENODEV;
2309 			break;
2310 		}
2311 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2312 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2313 		break;
2314 	}
2315 
2316 	case DIOCRGETTABLES: {
2317 		struct pfioc_table *io = (struct pfioc_table *)addr;
2318 
2319 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2320 			error = ENODEV;
2321 			break;
2322 		}
2323 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2324 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2325 		break;
2326 	}
2327 
2328 	case DIOCRGETTSTATS: {
2329 		struct pfioc_table *io = (struct pfioc_table *)addr;
2330 
2331 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2332 			error = ENODEV;
2333 			break;
2334 		}
2335 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2336 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2337 		break;
2338 	}
2339 
2340 	case DIOCRCLRTSTATS: {
2341 		struct pfioc_table *io = (struct pfioc_table *)addr;
2342 
2343 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2344 			error = ENODEV;
2345 			break;
2346 		}
2347 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2348 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2349 		break;
2350 	}
2351 
2352 	case DIOCRSETTFLAGS: {
2353 		struct pfioc_table *io = (struct pfioc_table *)addr;
2354 
2355 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2356 			error = ENODEV;
2357 			break;
2358 		}
2359 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2360 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2361 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2362 		break;
2363 	}
2364 
2365 	case DIOCRCLRADDRS: {
2366 		struct pfioc_table *io = (struct pfioc_table *)addr;
2367 
2368 		if (io->pfrio_esize != 0) {
2369 			error = ENODEV;
2370 			break;
2371 		}
2372 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2373 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2374 		break;
2375 	}
2376 
2377 	case DIOCRADDADDRS: {
2378 		struct pfioc_table *io = (struct pfioc_table *)addr;
2379 
2380 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2381 			error = ENODEV;
2382 			break;
2383 		}
2384 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2385 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2386 		    PFR_FLAG_USERIOCTL);
2387 		break;
2388 	}
2389 
2390 	case DIOCRDELADDRS: {
2391 		struct pfioc_table *io = (struct pfioc_table *)addr;
2392 
2393 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2394 			error = ENODEV;
2395 			break;
2396 		}
2397 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2398 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2399 		    PFR_FLAG_USERIOCTL);
2400 		break;
2401 	}
2402 
2403 	case DIOCRSETADDRS: {
2404 		struct pfioc_table *io = (struct pfioc_table *)addr;
2405 
2406 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2407 			error = ENODEV;
2408 			break;
2409 		}
2410 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2411 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2412 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2413 		    PFR_FLAG_USERIOCTL, 0);
2414 		break;
2415 	}
2416 
2417 	case DIOCRGETADDRS: {
2418 		struct pfioc_table *io = (struct pfioc_table *)addr;
2419 
2420 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2421 			error = ENODEV;
2422 			break;
2423 		}
2424 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2425 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2426 		break;
2427 	}
2428 
2429 	case DIOCRGETASTATS: {
2430 		struct pfioc_table *io = (struct pfioc_table *)addr;
2431 
2432 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2433 			error = ENODEV;
2434 			break;
2435 		}
2436 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2437 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2438 		break;
2439 	}
2440 
2441 	case DIOCRCLRASTATS: {
2442 		struct pfioc_table *io = (struct pfioc_table *)addr;
2443 
2444 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2445 			error = ENODEV;
2446 			break;
2447 		}
2448 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2449 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2450 		    PFR_FLAG_USERIOCTL);
2451 		break;
2452 	}
2453 
2454 	case DIOCRTSTADDRS: {
2455 		struct pfioc_table *io = (struct pfioc_table *)addr;
2456 
2457 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2458 			error = ENODEV;
2459 			break;
2460 		}
2461 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2462 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2463 		    PFR_FLAG_USERIOCTL);
2464 		break;
2465 	}
2466 
2467 	case DIOCRINADEFINE: {
2468 		struct pfioc_table *io = (struct pfioc_table *)addr;
2469 
2470 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2471 			error = ENODEV;
2472 			break;
2473 		}
2474 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2475 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2476 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2477 		break;
2478 	}
2479 
2480 	case DIOCOSFPADD: {
2481 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2482 		error = pf_osfp_add(io);
2483 		break;
2484 	}
2485 
2486 	case DIOCOSFPGET: {
2487 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2488 		error = pf_osfp_get(io);
2489 		break;
2490 	}
2491 
2492 	case DIOCXBEGIN: {
2493 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2494 		struct pfioc_trans_e	*ioe;
2495 		struct pfr_table	*table;
2496 		int			 i;
2497 
2498 		if (io->esize != sizeof(*ioe)) {
2499 			error = ENODEV;
2500 			goto fail;
2501 		}
2502 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2503 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2504 		for (i = 0; i < io->size; i++) {
2505 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2506 				kfree(table, M_TEMP);
2507 				kfree(ioe, M_TEMP);
2508 				error = EFAULT;
2509 				goto fail;
2510 			}
2511 			switch (ioe->rs_num) {
2512 #ifdef ALTQ
2513 			case PF_RULESET_ALTQ:
2514 				if (ioe->anchor[0]) {
2515 					kfree(table, M_TEMP);
2516 					kfree(ioe, M_TEMP);
2517 					error = EINVAL;
2518 					goto fail;
2519 				}
2520 				if ((error = pf_begin_altq(&ioe->ticket))) {
2521 					kfree(table, M_TEMP);
2522 					kfree(ioe, M_TEMP);
2523 					goto fail;
2524 				}
2525 				break;
2526 #endif /* ALTQ */
2527 			case PF_RULESET_TABLE:
2528 				bzero(table, sizeof(*table));
2529 				strlcpy(table->pfrt_anchor, ioe->anchor,
2530 				    sizeof(table->pfrt_anchor));
2531 				if ((error = pfr_ina_begin(table,
2532 				    &ioe->ticket, NULL, 0))) {
2533 					kfree(table, M_TEMP);
2534 					kfree(ioe, M_TEMP);
2535 					goto fail;
2536 				}
2537 				break;
2538 			default:
2539 				if ((error = pf_begin_rules(&ioe->ticket,
2540 				    ioe->rs_num, ioe->anchor))) {
2541 					kfree(table, M_TEMP);
2542 					kfree(ioe, M_TEMP);
2543 					goto fail;
2544 				}
2545 				break;
2546 			}
2547 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2548 				kfree(table, M_TEMP);
2549 				kfree(ioe, M_TEMP);
2550 				error = EFAULT;
2551 				goto fail;
2552 			}
2553 		}
2554 		kfree(table, M_TEMP);
2555 		kfree(ioe, M_TEMP);
2556 		break;
2557 	}
2558 
2559 	case DIOCXROLLBACK: {
2560 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2561 		struct pfioc_trans_e	*ioe;
2562 		struct pfr_table	*table;
2563 		int			 i;
2564 
2565 		if (io->esize != sizeof(*ioe)) {
2566 			error = ENODEV;
2567 			goto fail;
2568 		}
2569 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2570 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2571 		for (i = 0; i < io->size; i++) {
2572 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2573 				kfree(table, M_TEMP);
2574 				kfree(ioe, M_TEMP);
2575 				error = EFAULT;
2576 				goto fail;
2577 			}
2578 			switch (ioe->rs_num) {
2579 #ifdef ALTQ
2580 			case PF_RULESET_ALTQ:
2581 				if (ioe->anchor[0]) {
2582 					kfree(table, M_TEMP);
2583 					kfree(ioe, M_TEMP);
2584 					error = EINVAL;
2585 					goto fail;
2586 				}
2587 				if ((error = pf_rollback_altq(ioe->ticket))) {
2588 					kfree(table, M_TEMP);
2589 					kfree(ioe, M_TEMP);
2590 					goto fail; /* really bad */
2591 				}
2592 				break;
2593 #endif /* ALTQ */
2594 			case PF_RULESET_TABLE:
2595 				bzero(table, sizeof(*table));
2596 				strlcpy(table->pfrt_anchor, ioe->anchor,
2597 				    sizeof(table->pfrt_anchor));
2598 				if ((error = pfr_ina_rollback(table,
2599 				    ioe->ticket, NULL, 0))) {
2600 					kfree(table, M_TEMP);
2601 					kfree(ioe, M_TEMP);
2602 					goto fail; /* really bad */
2603 				}
2604 				break;
2605 			default:
2606 				if ((error = pf_rollback_rules(ioe->ticket,
2607 				    ioe->rs_num, ioe->anchor))) {
2608 					kfree(table, M_TEMP);
2609 					kfree(ioe, M_TEMP);
2610 					goto fail; /* really bad */
2611 				}
2612 				break;
2613 			}
2614 		}
2615 		kfree(table, M_TEMP);
2616 		kfree(ioe, M_TEMP);
2617 		break;
2618 	}
2619 
2620 	case DIOCXCOMMIT: {
2621 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2622 		struct pfioc_trans_e	*ioe;
2623 		struct pfr_table	*table;
2624 		struct pf_ruleset	*rs;
2625 		int			 i;
2626 
2627 		if (io->esize != sizeof(*ioe)) {
2628 			error = ENODEV;
2629 			goto fail;
2630 		}
2631 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2632 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2633 		/* first makes sure everything will succeed */
2634 		for (i = 0; i < io->size; i++) {
2635 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2636 				kfree(table, M_TEMP);
2637 				kfree(ioe, M_TEMP);
2638 				error = EFAULT;
2639 				goto fail;
2640 			}
2641 			switch (ioe->rs_num) {
2642 #ifdef ALTQ
2643 			case PF_RULESET_ALTQ:
2644 				if (ioe->anchor[0]) {
2645 					kfree(table, M_TEMP);
2646 					kfree(ioe, M_TEMP);
2647 					error = EINVAL;
2648 					goto fail;
2649 				}
2650 				if (!altqs_inactive_open || ioe->ticket !=
2651 				    ticket_altqs_inactive) {
2652 					kfree(table, M_TEMP);
2653 					kfree(ioe, M_TEMP);
2654 					error = EBUSY;
2655 					goto fail;
2656 				}
2657 				break;
2658 #endif /* ALTQ */
2659 			case PF_RULESET_TABLE:
2660 				rs = pf_find_ruleset(ioe->anchor);
2661 				if (rs == NULL || !rs->topen || ioe->ticket !=
2662 				     rs->tticket) {
2663 					kfree(table, M_TEMP);
2664 					kfree(ioe, M_TEMP);
2665 					error = EBUSY;
2666 					goto fail;
2667 				}
2668 				break;
2669 			default:
2670 				if (ioe->rs_num < 0 || ioe->rs_num >=
2671 				    PF_RULESET_MAX) {
2672 					kfree(table, M_TEMP);
2673 					kfree(ioe, M_TEMP);
2674 					error = EINVAL;
2675 					goto fail;
2676 				}
2677 				rs = pf_find_ruleset(ioe->anchor);
2678 				if (rs == NULL ||
2679 				    !rs->rules[ioe->rs_num].inactive.open ||
2680 				    rs->rules[ioe->rs_num].inactive.ticket !=
2681 				    ioe->ticket) {
2682 					kfree(table, M_TEMP);
2683 					kfree(ioe, M_TEMP);
2684 					error = EBUSY;
2685 					goto fail;
2686 				}
2687 				break;
2688 			}
2689 		}
2690 		/* now do the commit - no errors should happen here */
2691 		for (i = 0; i < io->size; i++) {
2692 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2693 				kfree(table, M_TEMP);
2694 				kfree(ioe, M_TEMP);
2695 				error = EFAULT;
2696 				goto fail;
2697 			}
2698 			switch (ioe->rs_num) {
2699 #ifdef ALTQ
2700 			case PF_RULESET_ALTQ:
2701 				if ((error = pf_commit_altq(ioe->ticket))) {
2702 					kfree(table, M_TEMP);
2703 					kfree(ioe, M_TEMP);
2704 					goto fail; /* really bad */
2705 				}
2706 				break;
2707 #endif /* ALTQ */
2708 			case PF_RULESET_TABLE:
2709 				bzero(table, sizeof(*table));
2710 				strlcpy(table->pfrt_anchor, ioe->anchor,
2711 				    sizeof(table->pfrt_anchor));
2712 				if ((error = pfr_ina_commit(table, ioe->ticket,
2713 				    NULL, NULL, 0))) {
2714 					kfree(table, M_TEMP);
2715 					kfree(ioe, M_TEMP);
2716 					goto fail; /* really bad */
2717 				}
2718 				break;
2719 			default:
2720 				if ((error = pf_commit_rules(ioe->ticket,
2721 				    ioe->rs_num, ioe->anchor))) {
2722 					kfree(table, M_TEMP);
2723 					kfree(ioe, M_TEMP);
2724 					goto fail; /* really bad */
2725 				}
2726 				break;
2727 			}
2728 		}
2729 		kfree(table, M_TEMP);
2730 		kfree(ioe, M_TEMP);
2731 		break;
2732 	}
2733 
2734 	case DIOCGETSRCNODES: {
2735 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2736 		struct pf_src_node	*n, *p, *pstore;
2737 		u_int32_t		 nr = 0;
2738 		int			 space = psn->psn_len;
2739 
2740 		if (space == 0) {
2741 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2742 				nr++;
2743 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2744 			break;
2745 		}
2746 
2747 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2748 
2749 		p = psn->psn_src_nodes;
2750 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2751 			int	secs = time_second, diff;
2752 
2753 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
2754 				break;
2755 
2756 			bcopy(n, pstore, sizeof(*pstore));
2757 			if (n->rule.ptr != NULL)
2758 				pstore->rule.nr = n->rule.ptr->nr;
2759 			pstore->creation = secs - pstore->creation;
2760 			if (pstore->expire > secs)
2761 				pstore->expire -= secs;
2762 			else
2763 				pstore->expire = 0;
2764 
2765 			/* adjust the connection rate estimate */
2766 			diff = secs - n->conn_rate.last;
2767 			if (diff >= n->conn_rate.seconds)
2768 				pstore->conn_rate.count = 0;
2769 			else
2770 				pstore->conn_rate.count -=
2771 				    n->conn_rate.count * diff /
2772 				    n->conn_rate.seconds;
2773 
2774 			error = copyout(pstore, p, sizeof(*p));
2775 			if (error) {
2776 				kfree(pstore, M_TEMP);
2777 				goto fail;
2778 			}
2779 			p++;
2780 			nr++;
2781 		}
2782 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2783 
2784 		kfree(pstore, M_TEMP);
2785 		break;
2786 	}
2787 
2788 	case DIOCCLRSRCNODES: {
2789 		struct pf_src_node	*n;
2790 		struct pf_state		*state;
2791 
2792 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2793 			state->src_node = NULL;
2794 			state->nat_src_node = NULL;
2795 		}
2796 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2797 			n->expire = 1;
2798 			n->states = 0;
2799 		}
2800 		pf_purge_expired_src_nodes(1);
2801 		pf_status.src_nodes = 0;
2802 		break;
2803 	}
2804 
2805 	case DIOCKILLSRCNODES: {
2806 		struct pf_src_node	*sn;
2807 		struct pf_state		*s;
2808 		struct pfioc_src_node_kill *psnk =
2809 		    (struct pfioc_src_node_kill *)addr;
2810 		u_int			killed = 0;
2811 
2812 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
2813 			if (PF_MATCHA(psnk->psnk_src.neg,
2814 				&psnk->psnk_src.addr.v.a.addr,
2815 				&psnk->psnk_src.addr.v.a.mask,
2816 				&sn->addr, sn->af) &&
2817 			    PF_MATCHA(psnk->psnk_dst.neg,
2818 				&psnk->psnk_dst.addr.v.a.addr,
2819 				&psnk->psnk_dst.addr.v.a.mask,
2820 				&sn->raddr, sn->af)) {
2821 				/* Handle state to src_node linkage */
2822 				if (sn->states != 0) {
2823 					RB_FOREACH(s, pf_state_tree_id,
2824 					    &tree_id) {
2825 						if (s->src_node == sn)
2826 							s->src_node = NULL;
2827 						if (s->nat_src_node == sn)
2828 							s->nat_src_node = NULL;
2829 					}
2830 					sn->states = 0;
2831 				}
2832 				sn->expire = 1;
2833 				killed++;
2834 			}
2835 		}
2836 
2837 		if (killed > 0)
2838 			pf_purge_expired_src_nodes(1);
2839 
2840 		psnk->psnk_killed = killed;
2841 		break;
2842 	}
2843 
2844 	case DIOCSETHOSTID: {
2845 		u_int32_t	*hostid = (u_int32_t *)addr;
2846 
2847 		if (*hostid == 0)
2848 			pf_status.hostid = karc4random();
2849 		else
2850 			pf_status.hostid = *hostid;
2851 		break;
2852 	}
2853 
2854 	case DIOCOSFPFLUSH:
2855 		crit_enter();
2856 		pf_osfp_flush();
2857 		crit_exit();
2858 		break;
2859 
2860 	case DIOCIGETIFACES: {
2861 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2862 
2863 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
2864 			error = ENODEV;
2865 			break;
2866 		}
2867 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2868 		    &io->pfiio_size);
2869 		break;
2870 	}
2871 
2872 	case DIOCSETIFFLAG: {
2873 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2874 
2875 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
2876 		break;
2877 	}
2878 
2879 	case DIOCCLRIFFLAG: {
2880 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2881 
2882 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
2883 		break;
2884 	}
2885 
2886 	default:
2887 		error = ENODEV;
2888 		break;
2889 	}
2890 fail:
2891 	lwkt_reltoken(&pf_token);
2892 	return (error);
2893 }
2894 
2895 /*
2896  * XXX - Check for version missmatch!!!
2897  */
2898 static void
2899 pf_clear_states(void)
2900 {
2901 	struct pf_state		*s, *nexts;
2902 	u_int			killed = 0;
2903 
2904 	for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
2905 		nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
2906 
2907 		/* don't send out individual delete messages */
2908 		s->sync_flags = PFSTATE_NOSYNC;
2909 		pf_unlink_state(s);
2910 		killed++;
2911 
2912 	}
2913 
2914 #if 0 /* PFSYNC */
2915 /*
2916  * XXX This is called on module unload, we do not want to sync that over? */
2917  */
2918 	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
2919 #endif
2920 }
2921 
2922 static int
2923 pf_clear_tables(void)
2924 {
2925 	struct pfioc_table io;
2926 	int error;
2927 
2928 	bzero(&io, sizeof(io));
2929 
2930 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
2931 	    io.pfrio_flags);
2932 
2933 	return (error);
2934 }
2935 
2936 static void
2937 pf_clear_srcnodes(void)
2938 {
2939 	struct pf_src_node	*n;
2940 	struct pf_state		*state;
2941 
2942 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2943 		state->src_node = NULL;
2944 		state->nat_src_node = NULL;
2945 	}
2946 	RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2947 		n->expire = 1;
2948 		n->states = 0;
2949 	}
2950 	pf_purge_expired_src_nodes(0);
2951 	pf_status.src_nodes = 0;
2952 }
2953 
2954 /*
2955  * XXX - Check for version missmatch!!!
2956  */
2957 
2958 /*
2959  * Duplicate pfctl -Fa operation to get rid of as much as we can.
2960  */
2961 static int
2962 shutdown_pf(void)
2963 {
2964 	int error = 0;
2965 	u_int32_t t[5];
2966 	char nn = '\0';
2967 
2968 
2969 	pf_status.running = 0;
2970 	error = dehook_pf();
2971 	if (error) {
2972 		pf_status.running = 1;
2973 		DPFPRINTF(PF_DEBUG_MISC,
2974 		    ("pf: pfil unregistration failed\n"));
2975 		return(error);
2976 	}
2977 	do {
2978 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
2979 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
2980 			break;
2981 		}
2982 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
2983 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
2984 			break;		/* XXX: rollback? */
2985 		}
2986 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
2987 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
2988 			break;		/* XXX: rollback? */
2989 		}
2990 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
2991 		    != 0) {
2992 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
2993 			break;		/* XXX: rollback? */
2994 		}
2995 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
2996 		    != 0) {
2997 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
2998 			break;		/* XXX: rollback? */
2999 		}
3000 
3001 		/* XXX: these should always succeed here */
3002 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3003 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3004 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3005 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3006 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3007 
3008 		if ((error = pf_clear_tables()) != 0)
3009 			break;
3010 #ifdef ALTQ
3011 		if ((error = pf_begin_altq(&t[0])) != 0) {
3012 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3013 			break;
3014 		}
3015 		pf_commit_altq(t[0]);
3016 #endif
3017 		pf_clear_states();
3018 		pf_clear_srcnodes();
3019 
3020 		/* status does not use malloced mem so no need to cleanup */
3021 		/* fingerprints and interfaces have their own cleanup code */
3022 	} while(0);
3023         return (error);
3024 }
3025 
3026 static int
3027 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3028 {
3029 	/*
3030 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3031 	 * for ip_len/ip_off. This is why we don't have to change byte order
3032 	 * like the FreeBSD-5 version does.
3033 	 */
3034 	int chk;
3035 
3036 	lwkt_gettoken(&pf_token);
3037 
3038 	chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3039 	if (chk && *m) {
3040 		m_freem(*m);
3041 		*m = NULL;
3042 	}
3043 	lwkt_reltoken(&pf_token);
3044 	return chk;
3045 }
3046 
3047 static int
3048 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3049 {
3050 	/*
3051 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3052 	 * for ip_len/ip_off. This is why we don't have to change byte order
3053 	 * like the FreeBSD-5 version does.
3054 	 */
3055 	int chk;
3056 
3057 	lwkt_gettoken(&pf_token);
3058 
3059 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3060 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3061 		in_delayed_cksum(*m);
3062 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3063 	}
3064 	chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3065 	if (chk && *m) {
3066 		m_freem(*m);
3067 		*m = NULL;
3068 	}
3069 	lwkt_reltoken(&pf_token);
3070 	return chk;
3071 }
3072 
3073 #ifdef INET6
3074 static int
3075 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3076 {
3077 	/*
3078 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3079 	 */
3080 	int chk;
3081 
3082 	lwkt_gettoken(&pf_token);
3083 
3084 	chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3085 	if (chk && *m) {
3086 		m_freem(*m);
3087 		*m = NULL;
3088 	}
3089 	lwkt_reltoken(&pf_token);
3090 	return chk;
3091 }
3092 
3093 static int
3094 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3095 {
3096 	/*
3097 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3098 	 */
3099 	int chk;
3100 
3101 	lwkt_gettoken(&pf_token);
3102 
3103 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3104 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3105 		in_delayed_cksum(*m);
3106 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3107 	}
3108 	chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3109 	if (chk && *m) {
3110 		m_freem(*m);
3111 		*m = NULL;
3112 	}
3113 	lwkt_reltoken(&pf_token);
3114 	return chk;
3115 }
3116 #endif /* INET6 */
3117 
3118 static int
3119 hook_pf(void)
3120 {
3121 	struct pfil_head *pfh_inet;
3122 #ifdef INET6
3123 	struct pfil_head *pfh_inet6;
3124 #endif
3125 
3126 	lwkt_gettoken(&pf_token);
3127 
3128 	if (pf_pfil_hooked) {
3129 		lwkt_reltoken(&pf_token);
3130 		return (0);
3131 	}
3132 
3133 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3134 	if (pfh_inet == NULL) {
3135 		lwkt_reltoken(&pf_token);
3136 		return (ENODEV);
3137 	}
3138 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet);
3139 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet);
3140 #ifdef INET6
3141 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3142 	if (pfh_inet6 == NULL) {
3143 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3144 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3145 		lwkt_reltoken(&pf_token);
3146 		return (ENODEV);
3147 	}
3148 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet6);
3149 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet6);
3150 #endif
3151 
3152 	pf_pfil_hooked = 1;
3153 	lwkt_reltoken(&pf_token);
3154 	return (0);
3155 }
3156 
3157 static int
3158 dehook_pf(void)
3159 {
3160 	struct pfil_head *pfh_inet;
3161 #ifdef INET6
3162 	struct pfil_head *pfh_inet6;
3163 #endif
3164 
3165 	lwkt_gettoken(&pf_token);
3166 
3167 	if (pf_pfil_hooked == 0) {
3168 		lwkt_reltoken(&pf_token);
3169 		return (0);
3170 	}
3171 
3172 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3173 	if (pfh_inet == NULL) {
3174 		lwkt_reltoken(&pf_token);
3175 		return (ENODEV);
3176 	}
3177 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3178 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3179 #ifdef INET6
3180 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3181 	if (pfh_inet6 == NULL) {
3182 		lwkt_reltoken(&pf_token);
3183 		return (ENODEV);
3184 	}
3185 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3186 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3187 #endif
3188 
3189 	pf_pfil_hooked = 0;
3190 	lwkt_reltoken(&pf_token);
3191 	return (0);
3192 }
3193 
3194 static int
3195 pf_load(void)
3196 {
3197 	lwkt_gettoken(&pf_token);
3198 
3199 	pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3200 	pfattach();
3201 	lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3202 	lwkt_reltoken(&pf_token);
3203 	return (0);
3204 }
3205 
3206 static int
3207 pf_mask_del(struct radix_node *rn, void *arg)
3208 {
3209 	struct radix_node_head *rnh = arg;
3210 
3211 	rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
3212 	Free(rn);
3213 	return 0;
3214 }
3215 
3216 static int
3217 pf_unload(void)
3218 {
3219 	int error;
3220 	pf_status.running = 0;
3221 
3222 	lwkt_gettoken(&pf_token);
3223 
3224 	error = dehook_pf();
3225 	if (error) {
3226 		/*
3227 		 * Should not happen!
3228 		 * XXX Due to error code ESRCH, kldunload will show
3229 		 * a message like 'No such process'.
3230 		 */
3231 		kprintf("pfil unregistration fail\n");
3232 		lwkt_reltoken(&pf_token);
3233 		return error;
3234 	}
3235 	shutdown_pf();
3236 	pf_end_threads = 1;
3237 	while (pf_end_threads < 2) {
3238 		wakeup_one(pf_purge_thread);
3239 		tsleep(pf_purge_thread, 0, "pftmo", hz);
3240 	}
3241 	pfi_cleanup();
3242 	pf_osfp_flush();
3243 	dev_ops_remove_all(&pf_ops);
3244 	lockuninit(&pf_consistency_lock);
3245 	lwkt_reltoken(&pf_token);
3246 
3247 	if (pf_maskhead != NULL) {
3248 		pf_maskhead->rnh_walktree(pf_maskhead,
3249 			pf_mask_del, pf_maskhead);
3250 		Free(pf_maskhead);
3251 		pf_maskhead = NULL;
3252 	}
3253 	kmalloc_destroy(&pf_state_pl);
3254 	kmalloc_destroy(&pf_frent_pl);
3255 	kmalloc_destroy(&pf_cent_pl);
3256 	return 0;
3257 }
3258 
3259 static int
3260 pf_modevent(module_t mod, int type, void *data)
3261 {
3262 	int error = 0;
3263 
3264 	lwkt_gettoken(&pf_token);
3265 
3266 	switch(type) {
3267 	case MOD_LOAD:
3268 		error = pf_load();
3269 		break;
3270 
3271 	case MOD_UNLOAD:
3272 		error = pf_unload();
3273 		break;
3274 	default:
3275 		error = EINVAL;
3276 		break;
3277 	}
3278 	lwkt_reltoken(&pf_token);
3279 	return error;
3280 }
3281 
3282 static moduledata_t pf_mod = {
3283 	"pf",
3284 	pf_modevent,
3285 	0
3286 };
3287 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3288 MODULE_VERSION(pf, PF_MODVER);
3289