xref: /dflybsd-src/sys/net/pf/pf_ioctl.c (revision e7b4468ce80913950cd099c393f3ce6ece6fcb2c)
1 /*	$FreeBSD: src/sys/contrib/pf/net/pf_ioctl.c,v 1.12 2004/08/12 14:15:42 mlaier Exp $	*/
2 /*	$OpenBSD: pf_ioctl.c,v 1.112.2.2 2004/07/24 18:28:12 brad Exp $ */
3 /*	$DragonFly: src/sys/net/pf/pf_ioctl.c,v 1.13 2008/01/05 14:02:38 swildner Exp $ */
4 
5 /*
6  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
7  *
8  * Copyright (c) 2001 Daniel Hartmeier
9  * Copyright (c) 2002,2003 Henning Brauer
10  * All rights reserved.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  *
16  *    - Redistributions of source code must retain the above copyright
17  *      notice, this list of conditions and the following disclaimer.
18  *    - Redistributions in binary form must reproduce the above
19  *      copyright notice, this list of conditions and the following
20  *      disclaimer in the documentation and/or other materials provided
21  *      with the distribution.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
26  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
27  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
28  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
29  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
30  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
31  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
33  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34  * POSSIBILITY OF SUCH DAMAGE.
35  *
36  * Effort sponsored in part by the Defense Advanced Research Projects
37  * Agency (DARPA) and Air Force Research Laboratory, Air Force
38  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
39  *
40  */
41 
42 #include "opt_inet.h"
43 #include "opt_inet6.h"
44 #include "use_pfsync.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/conf.h>
49 #include <sys/device.h>
50 #include <sys/mbuf.h>
51 #include <sys/filio.h>
52 #include <sys/fcntl.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/kernel.h>
56 #include <sys/thread2.h>
57 #include <sys/time.h>
58 #include <sys/malloc.h>
59 #include <sys/module.h>
60 #include <vm/vm_zone.h>
61 
62 #include <net/if.h>
63 #include <net/if_types.h>
64 #include <net/route.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_icmp.h>
72 
73 #include <net/pf/pfvar.h>
74 
75 #if NPFSYNC > 0
76 #include <net/pf/if_pfsync.h>
77 #endif /* NPFSYNC > 0 */
78 
79 #ifdef INET6
80 #include <netinet/ip6.h>
81 #include <netinet/in_pcb.h>
82 #endif /* INET6 */
83 
84 #ifdef ALTQ
85 #include <net/altq/altq.h>
86 #endif
87 
88 #include <machine/limits.h>
89 #include <net/pfil.h>
90 
91 void			 init_zone_var(void);
92 void			 cleanup_pf_zone(void);
93 int			 pfattach(void);
94 struct pf_pool		*pf_get_pool(char *, char *, u_int32_t,
95 			    u_int8_t, u_int32_t, u_int8_t, u_int8_t, u_int8_t);
96 int			 pf_get_ruleset_number(u_int8_t);
97 void			 pf_init_ruleset(struct pf_ruleset *);
98 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
99 void			 pf_empty_pool(struct pf_palist *);
100 #ifdef ALTQ
101 int			 pf_begin_altq(u_int32_t *);
102 int			 pf_rollback_altq(u_int32_t);
103 int			 pf_commit_altq(u_int32_t);
104 #endif /* ALTQ */
105 int			 pf_begin_rules(u_int32_t *, int, char *, char *);
106 int			 pf_rollback_rules(u_int32_t, int, char *, char *);
107 int			 pf_commit_rules(u_int32_t, int, char *, char *);
108 
109 extern struct callout	 pf_expire_to;
110 
111 struct pf_rule		 pf_default_rule;
112 
113 #define	TAGID_MAX	 50000
114 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
115 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
116 
117 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
118 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
119 #endif
120 static u_int16_t	 tagname2tag(struct pf_tags *, char *);
121 static void		 tag2tagname(struct pf_tags *, u_int16_t, char *);
122 static void		 tag_unref(struct pf_tags *, u_int16_t);
123 
124 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
125 
126 static cdev_t	pf_dev;
127 
128 /*
129  * XXX - These are new and need to be checked when moveing to a new version
130  */
131 static void		 pf_clear_states(void);
132 static int		 pf_clear_tables(void);
133 static void		 pf_clear_srcnodes(void);
134 /*
135  * XXX - These are new and need to be checked when moveing to a new version
136  */
137 
138 /*
139  * Wrapper functions for pfil(9) hooks
140  */
141 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
142 		int dir);
143 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
144 		int dir);
145 #ifdef INET6
146 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
147 		int dir);
148 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
149 		int dir);
150 #endif
151 
152 static int 		 hook_pf(void);
153 static int 		 dehook_pf(void);
154 static int 		 shutdown_pf(void);
155 static int 		 pf_load(void);
156 static int 		 pf_unload(void);
157 
158 d_open_t	pfopen;
159 d_close_t	pfclose;
160 d_ioctl_t	pfioctl;
161 
162 static struct dev_ops pf_ops = {	    /* XXX convert to port model */
163 	{ PF_NAME, 73, 0 },
164 	.d_open =	pfopen,
165 	.d_close =	pfclose,
166 	.d_ioctl =	pfioctl
167 };
168 
169 static volatile int pf_pfil_hooked = 0;
170 
171 void
172 init_zone_var(void)
173 {
174 	pf_src_tree_pl = pf_rule_pl = NULL;
175 	pf_state_pl = pf_altq_pl = pf_pooladdr_pl = NULL;
176 	pf_frent_pl = pf_frag_pl = pf_cache_pl = pf_cent_pl = NULL;
177 	pf_state_scrub_pl = NULL;
178 	pfr_ktable_pl = pfr_kentry_pl = NULL;
179 }
180 
181 void
182 cleanup_pf_zone(void)
183 {
184 	ZONE_DESTROY(pf_src_tree_pl);
185 	ZONE_DESTROY(pf_rule_pl);
186 	ZONE_DESTROY(pf_state_pl);
187 	ZONE_DESTROY(pf_altq_pl);
188 	ZONE_DESTROY(pf_pooladdr_pl);
189 	ZONE_DESTROY(pf_frent_pl);
190 	ZONE_DESTROY(pf_frag_pl);
191 	ZONE_DESTROY(pf_cache_pl);
192 	ZONE_DESTROY(pf_cent_pl);
193 	ZONE_DESTROY(pfr_ktable_pl);
194 	ZONE_DESTROY(pfr_kentry_pl);
195 	ZONE_DESTROY(pf_state_scrub_pl);
196 	ZONE_DESTROY(pfi_addr_pl);
197 }
198 
199 int
200 pfattach(void)
201 {
202 	u_int32_t *my_timeout = pf_default_rule.timeout;
203 	int error = 1;
204 
205 	do {
206 		ZONE_CREATE(pf_src_tree_pl,struct pf_src_node, "pfsrctrpl");
207 		ZONE_CREATE(pf_rule_pl,    struct pf_rule, "pfrulepl");
208 		ZONE_CREATE(pf_state_pl,   struct pf_state, "pfstatepl");
209 		ZONE_CREATE(pf_altq_pl,    struct pf_altq, "pfaltqpl");
210 		ZONE_CREATE(pf_pooladdr_pl,struct pf_pooladdr, "pfpooladdrpl");
211 		ZONE_CREATE(pfr_ktable_pl, struct pfr_ktable, "pfrktable");
212 		ZONE_CREATE(pfr_kentry_pl, struct pfr_kentry, "pfrkentry");
213 		ZONE_CREATE(pf_frent_pl,   struct pf_frent, "pffrent");
214 		ZONE_CREATE(pf_frag_pl,    struct pf_fragment, "pffrag");
215 		ZONE_CREATE(pf_cache_pl,   struct pf_fragment, "pffrcache");
216 		ZONE_CREATE(pf_cent_pl,    struct pf_frcache, "pffrcent");
217 		ZONE_CREATE(pf_state_scrub_pl, struct pf_state_scrub,
218 		    "pfstatescrub");
219 		ZONE_CREATE(pfi_addr_pl,   struct pfi_dynaddr, "pfiaddrpl");
220 		error = 0;
221 	} while(0);
222 	if (error) {
223 		cleanup_pf_zone();
224 		return (error);
225 	}
226 	pfr_initialize();
227 	pfi_initialize();
228 	error = pf_osfp_initialize();
229 	if (error) {
230 		cleanup_pf_zone();
231 		pf_osfp_cleanup();
232 		return (error);
233 	}
234 
235 	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
236 	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
237 	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
238 	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
239 	/* XXX uma_zone_set_max(pf_pool_limits[PF_LIMIT_STATES].pp,
240 		pf_pool_limits[PF_LIMIT_STATES].limit);
241 	*/
242 
243 	RB_INIT(&tree_src_tracking);
244 	TAILQ_INIT(&pf_anchors);
245 	pf_init_ruleset(&pf_main_ruleset);
246 	TAILQ_INIT(&pf_altqs[0]);
247 	TAILQ_INIT(&pf_altqs[1]);
248 	TAILQ_INIT(&pf_pabuf);
249 	pf_altqs_active = &pf_altqs[0];
250 	pf_altqs_inactive = &pf_altqs[1];
251 	TAILQ_INIT(&state_updates);
252 
253 	/* default rule should never be garbage collected */
254 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
255 	pf_default_rule.action = PF_PASS;
256 	pf_default_rule.nr = (uint32_t)(-1);
257 
258 	/* initialize default timeouts */
259 	my_timeout[PFTM_TCP_FIRST_PACKET] = 120;	/* First TCP packet */
260 	my_timeout[PFTM_TCP_OPENING] = 30; 		/* No response yet */
261 	my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;	/* Established */
262 	my_timeout[PFTM_TCP_CLOSING] = 15 * 60;		/* Half closed */
263 	my_timeout[PFTM_TCP_FIN_WAIT] = 45;		/* Got both FINs */
264 	my_timeout[PFTM_TCP_CLOSED] = 90;		/* Got a RST */
265 	my_timeout[PFTM_UDP_FIRST_PACKET] = 60;		/* First UDP packet */
266 	my_timeout[PFTM_UDP_SINGLE] = 30;		/* Unidirectional */
267 	my_timeout[PFTM_UDP_MULTIPLE] = 60;		/* Bidirectional */
268 	my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;	/* First ICMP packet */
269 	my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;		/* Got error response */
270 	my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;	/* First packet */
271 	my_timeout[PFTM_OTHER_SINGLE] = 30;		/* Unidirectional */
272 	my_timeout[PFTM_OTHER_MULTIPLE] = 60;		/* Bidirectional */
273 	my_timeout[PFTM_FRAG] = 30;			/* Fragment expire */
274 	my_timeout[PFTM_INTERVAL] = 10;			/* Expire interval */
275 
276 	callout_init(&pf_expire_to);
277 	callout_reset(&pf_expire_to, my_timeout[PFTM_INTERVAL] * hz,
278 	    pf_purge_timeout, &pf_expire_to);
279 
280 	pf_normalize_init();
281 	bzero(&pf_status, sizeof(pf_status));
282 	pf_status.debug = PF_DEBUG_URGENT;
283 	pf_pfil_hooked = 0;
284 
285 	/* XXX do our best to avoid a conflict */
286 	pf_status.hostid = karc4random();
287 
288 	return (error);
289 }
290 
291 int
292 pfopen(struct dev_open_args *ap)
293 {
294 	cdev_t dev = ap->a_head.a_dev;
295 	if (minor(dev) >= 1)
296 		return (ENXIO);
297 	return (0);
298 }
299 
300 int
301 pfclose(struct dev_close_args *ap)
302 {
303 	cdev_t dev = ap->a_head.a_dev;
304 	if (minor(dev) >= 1)
305 		return (ENXIO);
306 	return (0);
307 }
308 
309 struct pf_pool *
310 pf_get_pool(char *anchorname, char *rulesetname, u_int32_t ticket,
311     u_int8_t rule_action, u_int32_t rule_number, u_int8_t r_last,
312     u_int8_t active, u_int8_t check_ticket)
313 {
314 	struct pf_ruleset	*ruleset;
315 	struct pf_rule		*rule;
316 	int			 rs_num;
317 
318 	ruleset = pf_find_ruleset(anchorname, rulesetname);
319 	if (ruleset == NULL)
320 		return (NULL);
321 	rs_num = pf_get_ruleset_number(rule_action);
322 	if (rs_num >= PF_RULESET_MAX)
323 		return (NULL);
324 	if (active) {
325 		if (check_ticket && ticket !=
326 		    ruleset->rules[rs_num].active.ticket)
327 			return (NULL);
328 		if (r_last)
329 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
330 			    pf_rulequeue);
331 		else
332 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
333 	} else {
334 		if (check_ticket && ticket !=
335 		    ruleset->rules[rs_num].inactive.ticket)
336 			return (NULL);
337 		if (r_last)
338 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
339 			    pf_rulequeue);
340 		else
341 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
342 	}
343 	if (!r_last) {
344 		while ((rule != NULL) && (rule->nr != rule_number))
345 			rule = TAILQ_NEXT(rule, entries);
346 	}
347 	if (rule == NULL)
348 		return (NULL);
349 
350 	return (&rule->rpool);
351 }
352 
353 int
354 pf_get_ruleset_number(u_int8_t action)
355 {
356 	switch (action) {
357 	case PF_SCRUB:
358 		return (PF_RULESET_SCRUB);
359 		break;
360 	case PF_PASS:
361 	case PF_DROP:
362 		return (PF_RULESET_FILTER);
363 		break;
364 	case PF_NAT:
365 	case PF_NONAT:
366 		return (PF_RULESET_NAT);
367 		break;
368 	case PF_BINAT:
369 	case PF_NOBINAT:
370 		return (PF_RULESET_BINAT);
371 		break;
372 	case PF_RDR:
373 	case PF_NORDR:
374 		return (PF_RULESET_RDR);
375 		break;
376 	default:
377 		return (PF_RULESET_MAX);
378 		break;
379 	}
380 }
381 
382 void
383 pf_init_ruleset(struct pf_ruleset *ruleset)
384 {
385 	int	i;
386 
387 	memset(ruleset, 0, sizeof(struct pf_ruleset));
388 	for (i = 0; i < PF_RULESET_MAX; i++) {
389 		TAILQ_INIT(&ruleset->rules[i].queues[0]);
390 		TAILQ_INIT(&ruleset->rules[i].queues[1]);
391 		ruleset->rules[i].active.ptr = &ruleset->rules[i].queues[0];
392 		ruleset->rules[i].inactive.ptr = &ruleset->rules[i].queues[1];
393 	}
394 }
395 
396 struct pf_anchor *
397 pf_find_anchor(const char *anchorname)
398 {
399 	struct pf_anchor	*anchor;
400 	int			 n = -1;
401 
402 	anchor = TAILQ_FIRST(&pf_anchors);
403 	while (anchor != NULL && (n = strcmp(anchor->name, anchorname)) < 0)
404 		anchor = TAILQ_NEXT(anchor, entries);
405 	if (n == 0)
406 		return (anchor);
407 	else
408 		return (NULL);
409 }
410 
411 struct pf_ruleset *
412 pf_find_ruleset(char *anchorname, char *rulesetname)
413 {
414 	struct pf_anchor	*anchor;
415 	struct pf_ruleset	*ruleset;
416 
417 	if (!anchorname[0] && !rulesetname[0])
418 		return (&pf_main_ruleset);
419 	if (!anchorname[0] || !rulesetname[0])
420 		return (NULL);
421 	anchorname[PF_ANCHOR_NAME_SIZE-1] = 0;
422 	rulesetname[PF_RULESET_NAME_SIZE-1] = 0;
423 	anchor = pf_find_anchor(anchorname);
424 	if (anchor == NULL)
425 		return (NULL);
426 	ruleset = TAILQ_FIRST(&anchor->rulesets);
427 	while (ruleset != NULL && strcmp(ruleset->name, rulesetname) < 0)
428 		ruleset = TAILQ_NEXT(ruleset, entries);
429 	if (ruleset != NULL && !strcmp(ruleset->name, rulesetname))
430 		return (ruleset);
431 	else
432 		return (NULL);
433 }
434 
435 struct pf_ruleset *
436 pf_find_or_create_ruleset(char anchorname[PF_ANCHOR_NAME_SIZE],
437     char rulesetname[PF_RULESET_NAME_SIZE])
438 {
439 	struct pf_anchor	*anchor, *a;
440 	struct pf_ruleset	*ruleset, *r;
441 
442 	if (!anchorname[0] && !rulesetname[0])
443 		return (&pf_main_ruleset);
444 	if (!anchorname[0] || !rulesetname[0])
445 		return (NULL);
446 	anchorname[PF_ANCHOR_NAME_SIZE-1] = 0;
447 	rulesetname[PF_RULESET_NAME_SIZE-1] = 0;
448 	a = TAILQ_FIRST(&pf_anchors);
449 	while (a != NULL && strcmp(a->name, anchorname) < 0)
450 		a = TAILQ_NEXT(a, entries);
451 	if (a != NULL && !strcmp(a->name, anchorname))
452 		anchor = a;
453 	else {
454 		anchor = (struct pf_anchor *)kmalloc(sizeof(struct pf_anchor),
455 		    M_TEMP, M_NOWAIT);
456 		if (anchor == NULL)
457 			return (NULL);
458 		memset(anchor, 0, sizeof(struct pf_anchor));
459 		bcopy(anchorname, anchor->name, sizeof(anchor->name));
460 		TAILQ_INIT(&anchor->rulesets);
461 		if (a != NULL)
462 			TAILQ_INSERT_BEFORE(a, anchor, entries);
463 		else
464 			TAILQ_INSERT_TAIL(&pf_anchors, anchor, entries);
465 	}
466 	r = TAILQ_FIRST(&anchor->rulesets);
467 	while (r != NULL && strcmp(r->name, rulesetname) < 0)
468 		r = TAILQ_NEXT(r, entries);
469 	if (r != NULL && !strcmp(r->name, rulesetname))
470 		return (r);
471 	ruleset = (struct pf_ruleset *)kmalloc(sizeof(struct pf_ruleset),
472 	    M_TEMP, M_NOWAIT);
473 	if (ruleset != NULL) {
474 		pf_init_ruleset(ruleset);
475 		bcopy(rulesetname, ruleset->name, sizeof(ruleset->name));
476 		ruleset->anchor = anchor;
477 		if (r != NULL)
478 			TAILQ_INSERT_BEFORE(r, ruleset, entries);
479 		else
480 			TAILQ_INSERT_TAIL(&anchor->rulesets, ruleset, entries);
481 	}
482 	return (ruleset);
483 }
484 
485 void
486 pf_remove_if_empty_ruleset(struct pf_ruleset *ruleset)
487 {
488 	struct pf_anchor	*anchor;
489 	int			 i;
490 
491 	if (ruleset == NULL || ruleset->anchor == NULL || ruleset->tables > 0 ||
492 	    ruleset->topen)
493 		return;
494 	for (i = 0; i < PF_RULESET_MAX; ++i)
495 		if (!TAILQ_EMPTY(ruleset->rules[i].active.ptr) ||
496 		    !TAILQ_EMPTY(ruleset->rules[i].inactive.ptr) ||
497 		    ruleset->rules[i].inactive.open)
498 			return;
499 
500 	anchor = ruleset->anchor;
501 	TAILQ_REMOVE(&anchor->rulesets, ruleset, entries);
502 	kfree(ruleset, M_TEMP);
503 
504 	if (TAILQ_EMPTY(&anchor->rulesets)) {
505 		TAILQ_REMOVE(&pf_anchors, anchor, entries);
506 		kfree(anchor, M_TEMP);
507 		pf_update_anchor_rules();
508 	}
509 }
510 
511 void
512 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
513 {
514 	struct pf_pooladdr	*mv_pool_pa;
515 
516 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
517 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
518 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
519 	}
520 }
521 
522 void
523 pf_empty_pool(struct pf_palist *poola)
524 {
525 	struct pf_pooladdr	*empty_pool_pa;
526 
527 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
528 		pfi_dynaddr_remove(&empty_pool_pa->addr);
529 		pf_tbladdr_remove(&empty_pool_pa->addr);
530 		pfi_detach_rule(empty_pool_pa->kif);
531 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
532 		pool_put(&pf_pooladdr_pl, empty_pool_pa);
533 	}
534 }
535 
536 void
537 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
538 {
539 	if (rulequeue != NULL) {
540 		if (rule->states <= 0) {
541 			/*
542 			 * XXX - we need to remove the table *before* detaching
543 			 * the rule to make sure the table code does not delete
544 			 * the anchor under our feet.
545 			 */
546 			pf_tbladdr_remove(&rule->src.addr);
547 			pf_tbladdr_remove(&rule->dst.addr);
548 		}
549 		TAILQ_REMOVE(rulequeue, rule, entries);
550 		rule->entries.tqe_prev = NULL;
551 		rule->nr = (uint32_t)(-1);
552 	}
553 
554 	if (rule->states > 0 || rule->src_nodes > 0 ||
555 	    rule->entries.tqe_prev != NULL)
556 		return;
557 	pf_tag_unref(rule->tag);
558 	pf_tag_unref(rule->match_tag);
559 #ifdef ALTQ
560 	if (rule->pqid != rule->qid)
561 		pf_qid_unref(rule->pqid);
562 	pf_qid_unref(rule->qid);
563 #endif
564 	pfi_dynaddr_remove(&rule->src.addr);
565 	pfi_dynaddr_remove(&rule->dst.addr);
566 	if (rulequeue == NULL) {
567 		pf_tbladdr_remove(&rule->src.addr);
568 		pf_tbladdr_remove(&rule->dst.addr);
569 	}
570 	pfi_detach_rule(rule->kif);
571 	pf_empty_pool(&rule->rpool.list);
572 	pool_put(&pf_rule_pl, rule);
573 }
574 
575 static	u_int16_t
576 tagname2tag(struct pf_tags *head, char *tagname)
577 {
578 	struct pf_tagname	*tag, *p = NULL;
579 	u_int16_t		 new_tagid = 1;
580 
581 	TAILQ_FOREACH(tag, head, entries)
582 		if (strcmp(tagname, tag->name) == 0) {
583 			tag->ref++;
584 			return (tag->tag);
585 		}
586 
587 	/*
588 	 * to avoid fragmentation, we do a linear search from the beginning
589 	 * and take the first free slot we find. if there is none or the list
590 	 * is empty, append a new entry at the end.
591 	 */
592 
593 	/* new entry */
594 	if (!TAILQ_EMPTY(head))
595 		for (p = TAILQ_FIRST(head); p != NULL &&
596 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
597 			new_tagid = p->tag + 1;
598 
599 	if (new_tagid > TAGID_MAX)
600 		return (0);
601 
602 	/* allocate and fill new struct pf_tagname */
603 	tag = (struct pf_tagname *)kmalloc(sizeof(struct pf_tagname),
604 	    M_TEMP, M_NOWAIT | M_ZERO);
605 	if (tag == NULL)
606 		return (0);
607 	strlcpy(tag->name, tagname, sizeof(tag->name));
608 	tag->tag = new_tagid;
609 	tag->ref++;
610 
611 	if (p != NULL)	/* insert new entry before p */
612 		TAILQ_INSERT_BEFORE(p, tag, entries);
613 	else	/* either list empty or no free slot in between */
614 		TAILQ_INSERT_TAIL(head, tag, entries);
615 
616 	return (tag->tag);
617 }
618 
619 static	void
620 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
621 {
622 	struct pf_tagname	*tag;
623 
624 	TAILQ_FOREACH(tag, head, entries)
625 		if (tag->tag == tagid) {
626 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
627 			return;
628 		}
629 }
630 
631 static	void
632 tag_unref(struct pf_tags *head, u_int16_t tag)
633 {
634 	struct pf_tagname	*p, *next;
635 
636 	if (tag == 0)
637 		return;
638 
639 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
640 		next = TAILQ_NEXT(p, entries);
641 		if (tag == p->tag) {
642 			if (--p->ref == 0) {
643 				TAILQ_REMOVE(head, p, entries);
644 				kfree(p, M_TEMP);
645 			}
646 			break;
647 		}
648 	}
649 }
650 
651 u_int16_t
652 pf_tagname2tag(char *tagname)
653 {
654 	return (tagname2tag(&pf_tags, tagname));
655 }
656 
657 void
658 pf_tag2tagname(u_int16_t tagid, char *p)
659 {
660 	return (tag2tagname(&pf_tags, tagid, p));
661 }
662 
663 void
664 pf_tag_unref(u_int16_t tag)
665 {
666 	return (tag_unref(&pf_tags, tag));
667 }
668 
669 #ifdef ALTQ
670 u_int32_t
671 pf_qname2qid(char *qname)
672 {
673 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
674 }
675 
676 void
677 pf_qid2qname(u_int32_t qid, char *p)
678 {
679 	return (tag2tagname(&pf_qids, (u_int16_t)qid, p));
680 }
681 
682 void
683 pf_qid_unref(u_int32_t qid)
684 {
685 	return (tag_unref(&pf_qids, (u_int16_t)qid));
686 }
687 
688 int
689 pf_begin_altq(u_int32_t *ticket)
690 {
691 	struct pf_altq	*altq;
692 	int		 error = 0;
693 
694 	/* Purge the old altq list */
695 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
696 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
697 		if (altq->qname[0] == 0) {
698 			/* detach and destroy the discipline */
699 			error = altq_remove(altq);
700 		} else
701 			pf_qid_unref(altq->qid);
702 		pool_put(&pf_altq_pl, altq);
703 	}
704 	if (error)
705 		return (error);
706 	*ticket = ++ticket_altqs_inactive;
707 	altqs_inactive_open = 1;
708 	return (0);
709 }
710 
711 int
712 pf_rollback_altq(u_int32_t ticket)
713 {
714 	struct pf_altq	*altq;
715 	int		 error = 0;
716 
717 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
718 		return (0);
719 	/* Purge the old altq list */
720 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
721 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
722 		if (altq->qname[0] == 0) {
723 			/* detach and destroy the discipline */
724 			error = altq_remove(altq);
725 		} else
726 			pf_qid_unref(altq->qid);
727 		pool_put(&pf_altq_pl, altq);
728 	}
729 	altqs_inactive_open = 0;
730 	return (error);
731 }
732 
733 int
734 pf_commit_altq(u_int32_t ticket)
735 {
736 	struct pf_altqqueue	*old_altqs;
737 	struct pf_altq		*altq;
738 	int			 err, error = 0;
739 
740 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
741 		return (EBUSY);
742 
743 	/* swap altqs, keep the old. */
744 	crit_enter();
745 	old_altqs = pf_altqs_active;
746 	pf_altqs_active = pf_altqs_inactive;
747 	pf_altqs_inactive = old_altqs;
748 	ticket_altqs_active = ticket_altqs_inactive;
749 
750 	/* Attach new disciplines */
751 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
752 		if (altq->qname[0] == 0) {
753 			/* attach the discipline */
754 			error = altq_pfattach(altq);
755 			if (error) {
756 				crit_exit();
757 				return (error);
758 			}
759 		}
760 	}
761 
762 	/* Purge the old altq list */
763 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
764 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
765 		if (altq->qname[0] == 0) {
766 			/* detach and destroy the discipline */
767 			err = altq_pfdetach(altq);
768 			if (err != 0 && error == 0)
769 				error = err;
770 			err = altq_remove(altq);
771 			if (err != 0 && error == 0)
772 				error = err;
773 		} else
774 			pf_qid_unref(altq->qid);
775 		pool_put(&pf_altq_pl, altq);
776 	}
777 	crit_exit();
778 
779 	altqs_inactive_open = 0;
780 	return (error);
781 }
782 #endif /* ALTQ */
783 
784 int
785 pf_begin_rules(u_int32_t *ticket, int rs_num, char *anchor, char *ruleset)
786 {
787 	struct pf_ruleset	*rs;
788 	struct pf_rule		*rule;
789 
790 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
791 		return (EINVAL);
792 	rs = pf_find_or_create_ruleset(anchor, ruleset);
793 	if (rs == NULL)
794 		return (EINVAL);
795 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL)
796 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
797 	*ticket = ++rs->rules[rs_num].inactive.ticket;
798 	rs->rules[rs_num].inactive.open = 1;
799 	return (0);
800 }
801 
802 int
803 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor, char *ruleset)
804 {
805 	struct pf_ruleset	*rs;
806 	struct pf_rule		*rule;
807 
808 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
809 		return (EINVAL);
810 	rs = pf_find_ruleset(anchor, ruleset);
811 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
812 	    rs->rules[rs_num].inactive.ticket != ticket)
813 		return (0);
814 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL)
815 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
816 	rs->rules[rs_num].inactive.open = 0;
817 	return (0);
818 }
819 
820 int
821 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor, char *ruleset)
822 {
823 	struct pf_ruleset	*rs;
824 	struct pf_rule		*rule;
825 	struct pf_rulequeue	*old_rules;
826 
827 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
828 		return (EINVAL);
829 	rs = pf_find_ruleset(anchor, ruleset);
830 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
831 	    ticket != rs->rules[rs_num].inactive.ticket)
832 		return (EBUSY);
833 
834 	/* Swap rules, keep the old. */
835 	crit_enter();
836 	old_rules = rs->rules[rs_num].active.ptr;
837 	rs->rules[rs_num].active.ptr =
838 	    rs->rules[rs_num].inactive.ptr;
839 	rs->rules[rs_num].inactive.ptr = old_rules;
840 	rs->rules[rs_num].active.ticket =
841 	    rs->rules[rs_num].inactive.ticket;
842 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
843 
844 	/* Purge the old rule list. */
845 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
846 		pf_rm_rule(old_rules, rule);
847 	rs->rules[rs_num].inactive.open = 0;
848 	pf_remove_if_empty_ruleset(rs);
849 	pf_update_anchor_rules();
850 	crit_exit();
851 	return (0);
852 }
853 
854 int
855 pfioctl(struct dev_ioctl_args *ap)
856 {
857 	u_long cmd = ap->a_cmd;
858 	caddr_t addr = ap->a_data;
859 	struct pf_pooladdr	*pa = NULL;
860 	struct pf_pool		*pool = NULL;
861 	int			 error = 0;
862 
863 	/* XXX keep in sync with switch() below */
864 	if (securelevel > 1)
865 		switch (cmd) {
866 		case DIOCGETRULES:
867 		case DIOCGETRULE:
868 		case DIOCGETADDRS:
869 		case DIOCGETADDR:
870 		case DIOCGETSTATE:
871 		case DIOCSETSTATUSIF:
872 		case DIOCGETSTATUS:
873 		case DIOCCLRSTATUS:
874 		case DIOCNATLOOK:
875 		case DIOCSETDEBUG:
876 		case DIOCGETSTATES:
877 		case DIOCGETTIMEOUT:
878 		case DIOCCLRRULECTRS:
879 		case DIOCGETLIMIT:
880 		case DIOCGETALTQS:
881 		case DIOCGETALTQ:
882 		case DIOCGETQSTATS:
883 		case DIOCGETANCHORS:
884 		case DIOCGETANCHOR:
885 		case DIOCGETRULESETS:
886 		case DIOCGETRULESET:
887 		case DIOCRGETTABLES:
888 		case DIOCRGETTSTATS:
889 		case DIOCRCLRTSTATS:
890 		case DIOCRCLRADDRS:
891 		case DIOCRADDADDRS:
892 		case DIOCRDELADDRS:
893 		case DIOCRSETADDRS:
894 		case DIOCRGETADDRS:
895 		case DIOCRGETASTATS:
896 		case DIOCRCLRASTATS:
897 		case DIOCRTSTADDRS:
898 		case DIOCOSFPGET:
899 		case DIOCGETSRCNODES:
900 		case DIOCCLRSRCNODES:
901 		case DIOCIGETIFACES:
902 		case DIOCICLRISTATS:
903 		case DIOCGIFSPEED:
904 			break;
905 		case DIOCRCLRTABLES:
906 		case DIOCRADDTABLES:
907 		case DIOCRDELTABLES:
908 		case DIOCRSETTFLAGS:
909 			if (((struct pfioc_table *)addr)->pfrio_flags &
910 			    PFR_FLAG_DUMMY)
911 				break; /* dummy operation ok */
912 			return (EPERM);
913 		default:
914 			return (EPERM);
915 		}
916 
917 	if (!(ap->a_fflag & FWRITE))
918 		switch (cmd) {
919 		case DIOCGETRULES:
920 		case DIOCGETRULE:
921 		case DIOCGETADDRS:
922 		case DIOCGETADDR:
923 		case DIOCGETSTATE:
924 		case DIOCGETSTATUS:
925 		case DIOCGETSTATES:
926 		case DIOCGETTIMEOUT:
927 		case DIOCGETLIMIT:
928 		case DIOCGETALTQS:
929 		case DIOCGETALTQ:
930 		case DIOCGETQSTATS:
931 		case DIOCGETANCHORS:
932 		case DIOCGETANCHOR:
933 		case DIOCGETRULESETS:
934 		case DIOCGETRULESET:
935 		case DIOCRGETTABLES:
936 		case DIOCRGETTSTATS:
937 		case DIOCRGETADDRS:
938 		case DIOCRGETASTATS:
939 		case DIOCRTSTADDRS:
940 		case DIOCOSFPGET:
941 		case DIOCGETSRCNODES:
942 		case DIOCIGETIFACES:
943 		case DIOCGIFSPEED:
944 			break;
945 		case DIOCRCLRTABLES:
946 		case DIOCRADDTABLES:
947 		case DIOCRDELTABLES:
948 		case DIOCRCLRTSTATS:
949 		case DIOCRCLRADDRS:
950 		case DIOCRADDADDRS:
951 		case DIOCRDELADDRS:
952 		case DIOCRSETADDRS:
953 		case DIOCRSETTFLAGS:
954 			if (((struct pfioc_table *)addr)->pfrio_flags &
955 			    PFR_FLAG_DUMMY)
956 				break; /* dummy operation ok */
957 			return (EACCES);
958 		default:
959 			return (EACCES);
960 		}
961 
962 	switch (cmd) {
963 
964 	case DIOCSTART:
965 		if (pf_status.running)
966 			error = EEXIST;
967 		else {
968 			error = hook_pf();
969 			if (error) {
970 				DPFPRINTF(PF_DEBUG_MISC,
971 				    ("pf: pfil registration fail\n"));
972 				break;
973 			}
974 			pf_status.running = 1;
975 			pf_status.since = time_second;
976 			if (pf_status.stateid == 0) {
977 				pf_status.stateid = time_second;
978 				pf_status.stateid = pf_status.stateid << 32;
979 			}
980 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
981 		}
982 		break;
983 
984 	case DIOCSTOP:
985 		if (!pf_status.running)
986 			error = ENOENT;
987 		else {
988 			pf_status.running = 0;
989 			error = dehook_pf();
990 			if (error) {
991 				pf_status.running = 1;
992 				DPFPRINTF(PF_DEBUG_MISC,
993 					("pf: pfil unregistration failed\n"));
994 			}
995 			pf_status.since = time_second;
996 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
997 		}
998 		break;
999 
1000 	case DIOCBEGINRULES: {
1001 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1002 
1003 		error = pf_begin_rules(&pr->ticket, pf_get_ruleset_number(
1004 		    pr->rule.action), pr->anchor, pr->ruleset);
1005 		break;
1006 	}
1007 
1008 	case DIOCADDRULE: {
1009 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1010 		struct pf_ruleset	*ruleset;
1011 		struct pf_rule		*rule, *tail;
1012 		struct pf_pooladdr	*pa;
1013 		int			 rs_num;
1014 
1015 		ruleset = pf_find_ruleset(pr->anchor, pr->ruleset);
1016 		if (ruleset == NULL) {
1017 			error = EINVAL;
1018 			break;
1019 		}
1020 		rs_num = pf_get_ruleset_number(pr->rule.action);
1021 		if (rs_num >= PF_RULESET_MAX) {
1022 			error = EINVAL;
1023 			break;
1024 		}
1025 		if (pr->rule.anchorname[0] && ruleset != &pf_main_ruleset) {
1026 			error = EINVAL;
1027 			break;
1028 		}
1029 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1030 			error = EINVAL;
1031 			break;
1032 		}
1033 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1034 			error = EBUSY;
1035 			break;
1036 		}
1037 		if (pr->pool_ticket != ticket_pabuf) {
1038 			error = EBUSY;
1039 			break;
1040 		}
1041 		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1042 		if (rule == NULL) {
1043 			error = ENOMEM;
1044 			break;
1045 		}
1046 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1047 		rule->anchor = NULL;
1048 		rule->kif = NULL;
1049 		TAILQ_INIT(&rule->rpool.list);
1050 		/* initialize refcounting */
1051 		rule->states = 0;
1052 		rule->src_nodes = 0;
1053 		rule->entries.tqe_prev = NULL;
1054 #ifndef INET
1055 		if (rule->af == AF_INET) {
1056 			pool_put(&pf_rule_pl, rule);
1057 			error = EAFNOSUPPORT;
1058 			break;
1059 		}
1060 #endif /* INET */
1061 #ifndef INET6
1062 		if (rule->af == AF_INET6) {
1063 			pool_put(&pf_rule_pl, rule);
1064 			error = EAFNOSUPPORT;
1065 			break;
1066 		}
1067 #endif /* INET6 */
1068 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1069 		    pf_rulequeue);
1070 		if (tail)
1071 			rule->nr = tail->nr + 1;
1072 		else
1073 			rule->nr = 0;
1074 		if (rule->ifname[0]) {
1075 			rule->kif = pfi_attach_rule(rule->ifname);
1076 			if (rule->kif == NULL) {
1077 				pool_put(&pf_rule_pl, rule);
1078 				error = EINVAL;
1079 				break;
1080 			}
1081 		}
1082 
1083 #ifdef ALTQ
1084 		/* set queue IDs */
1085 		if (rule->qname[0] != 0) {
1086 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1087 				error = EBUSY;
1088 			else if (rule->pqname[0] != 0) {
1089 				if ((rule->pqid =
1090 				    pf_qname2qid(rule->pqname)) == 0)
1091 					error = EBUSY;
1092 			} else
1093 				rule->pqid = rule->qid;
1094 		}
1095 #endif
1096 		if (rule->tagname[0])
1097 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1098 				error = EBUSY;
1099 		if (rule->match_tagname[0])
1100 			if ((rule->match_tag =
1101 			    pf_tagname2tag(rule->match_tagname)) == 0)
1102 				error = EBUSY;
1103 		if (rule->rt && !rule->direction)
1104 			error = EINVAL;
1105 		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1106 			error = EINVAL;
1107 		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1108 			error = EINVAL;
1109 		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1110 			error = EINVAL;
1111 		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1112 			error = EINVAL;
1113 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1114 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1115 				error = EINVAL;
1116 
1117 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1118 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1119 		    (rule->action == PF_BINAT)) && !rule->anchorname[0]) ||
1120 		    (rule->rt > PF_FASTROUTE)) &&
1121 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1122 			error = EINVAL;
1123 
1124 		if (error) {
1125 			pf_rm_rule(NULL, rule);
1126 			break;
1127 		}
1128 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1129 		rule->evaluations = rule->packets = rule->bytes = 0;
1130 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1131 		    rule, entries);
1132 		break;
1133 	}
1134 
1135 	case DIOCCOMMITRULES: {
1136 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1137 
1138 		error = pf_commit_rules(pr->ticket, pf_get_ruleset_number(
1139 		    pr->rule.action), pr->anchor, pr->ruleset);
1140 		break;
1141 	}
1142 
1143 	case DIOCGETRULES: {
1144 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1145 		struct pf_ruleset	*ruleset;
1146 		struct pf_rule		*tail;
1147 		int			 rs_num;
1148 
1149 		ruleset = pf_find_ruleset(pr->anchor, pr->ruleset);
1150 		if (ruleset == NULL) {
1151 			error = EINVAL;
1152 			break;
1153 		}
1154 		rs_num = pf_get_ruleset_number(pr->rule.action);
1155 		if (rs_num >= PF_RULESET_MAX) {
1156 			error = EINVAL;
1157 			break;
1158 		}
1159 		crit_enter();
1160 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1161 		    pf_rulequeue);
1162 		if (tail)
1163 			pr->nr = tail->nr + 1;
1164 		else
1165 			pr->nr = 0;
1166 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1167 		crit_exit();
1168 		break;
1169 	}
1170 
1171 	case DIOCGETRULE: {
1172 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1173 		struct pf_ruleset	*ruleset;
1174 		struct pf_rule		*rule;
1175 		int			 rs_num, i;
1176 
1177 		ruleset = pf_find_ruleset(pr->anchor, pr->ruleset);
1178 		if (ruleset == NULL) {
1179 			error = EINVAL;
1180 			break;
1181 		}
1182 		rs_num = pf_get_ruleset_number(pr->rule.action);
1183 		if (rs_num >= PF_RULESET_MAX) {
1184 			error = EINVAL;
1185 			break;
1186 		}
1187 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1188 			error = EBUSY;
1189 			break;
1190 		}
1191 		crit_enter();
1192 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1193 		while ((rule != NULL) && (rule->nr != pr->nr))
1194 			rule = TAILQ_NEXT(rule, entries);
1195 		if (rule == NULL) {
1196 			error = EBUSY;
1197 			crit_exit();
1198 			break;
1199 		}
1200 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1201 		pfi_dynaddr_copyout(&pr->rule.src.addr);
1202 		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1203 		pf_tbladdr_copyout(&pr->rule.src.addr);
1204 		pf_tbladdr_copyout(&pr->rule.dst.addr);
1205 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1206 			if (rule->skip[i].ptr == NULL)
1207 				pr->rule.skip[i].nr = (uint32_t)(-1);
1208 			else
1209 				pr->rule.skip[i].nr =
1210 				    rule->skip[i].ptr->nr;
1211 		crit_exit();
1212 		break;
1213 	}
1214 
1215 	case DIOCCHANGERULE: {
1216 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1217 		struct pf_ruleset	*ruleset;
1218 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1219 		u_int32_t		 nr = 0;
1220 		int			 rs_num;
1221 
1222 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1223 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1224 		    pcr->pool_ticket != ticket_pabuf) {
1225 			error = EBUSY;
1226 			break;
1227 		}
1228 
1229 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1230 		    pcr->action > PF_CHANGE_GET_TICKET) {
1231 			error = EINVAL;
1232 			break;
1233 		}
1234 		ruleset = pf_find_ruleset(pcr->anchor, pcr->ruleset);
1235 		if (ruleset == NULL) {
1236 			error = EINVAL;
1237 			break;
1238 		}
1239 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1240 		if (rs_num >= PF_RULESET_MAX) {
1241 			error = EINVAL;
1242 			break;
1243 		}
1244 
1245 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1246 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1247 			break;
1248 		} else {
1249 			if (pcr->ticket !=
1250 			    ruleset->rules[rs_num].active.ticket) {
1251 				error = EINVAL;
1252 				break;
1253 			}
1254 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1255 				error = EINVAL;
1256 				break;
1257 			}
1258 		}
1259 
1260 		if (pcr->action != PF_CHANGE_REMOVE) {
1261 			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1262 			if (newrule == NULL) {
1263 				error = ENOMEM;
1264 				break;
1265 			}
1266 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1267 			TAILQ_INIT(&newrule->rpool.list);
1268 			/* initialize refcounting */
1269 			newrule->states = 0;
1270 			newrule->entries.tqe_prev = NULL;
1271 #ifndef INET
1272 			if (newrule->af == AF_INET) {
1273 				pool_put(&pf_rule_pl, newrule);
1274 				error = EAFNOSUPPORT;
1275 				break;
1276 			}
1277 #endif /* INET */
1278 #ifndef INET6
1279 			if (newrule->af == AF_INET6) {
1280 				pool_put(&pf_rule_pl, newrule);
1281 				error = EAFNOSUPPORT;
1282 				break;
1283 			}
1284 #endif /* INET6 */
1285 			if (newrule->ifname[0]) {
1286 				newrule->kif = pfi_attach_rule(newrule->ifname);
1287 				if (newrule->kif == NULL) {
1288 					pool_put(&pf_rule_pl, newrule);
1289 					error = EINVAL;
1290 					break;
1291 				}
1292 			} else
1293 				newrule->kif = NULL;
1294 
1295 #ifdef ALTQ
1296 			/* set queue IDs */
1297 			if (newrule->qname[0] != 0) {
1298 				if ((newrule->qid =
1299 				    pf_qname2qid(newrule->qname)) == 0)
1300 					error = EBUSY;
1301 				else if (newrule->pqname[0] != 0) {
1302 					if ((newrule->pqid =
1303 					    pf_qname2qid(newrule->pqname)) == 0)
1304 						error = EBUSY;
1305 				} else
1306 					newrule->pqid = newrule->qid;
1307 			}
1308 #endif
1309 			if (newrule->tagname[0])
1310 				if ((newrule->tag =
1311 				    pf_tagname2tag(newrule->tagname)) == 0)
1312 					error = EBUSY;
1313 			if (newrule->match_tagname[0])
1314 				if ((newrule->match_tag = pf_tagname2tag(
1315 				    newrule->match_tagname)) == 0)
1316 					error = EBUSY;
1317 
1318 			if (newrule->rt && !newrule->direction)
1319 				error = EINVAL;
1320 			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1321 				error = EINVAL;
1322 			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1323 				error = EINVAL;
1324 			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1325 				error = EINVAL;
1326 			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1327 				error = EINVAL;
1328 
1329 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1330 			if (((((newrule->action == PF_NAT) ||
1331 			    (newrule->action == PF_RDR) ||
1332 			    (newrule->action == PF_BINAT) ||
1333 			    (newrule->rt > PF_FASTROUTE)) &&
1334 			    !newrule->anchorname[0])) &&
1335 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1336 				error = EINVAL;
1337 
1338 			if (error) {
1339 				pf_rm_rule(NULL, newrule);
1340 				break;
1341 			}
1342 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1343 			newrule->evaluations = newrule->packets = 0;
1344 			newrule->bytes = 0;
1345 		}
1346 		pf_empty_pool(&pf_pabuf);
1347 
1348 		crit_enter();
1349 
1350 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1351 			oldrule = TAILQ_FIRST(
1352 			    ruleset->rules[rs_num].active.ptr);
1353 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1354 			oldrule = TAILQ_LAST(
1355 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1356 		else {
1357 			oldrule = TAILQ_FIRST(
1358 			    ruleset->rules[rs_num].active.ptr);
1359 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1360 				oldrule = TAILQ_NEXT(oldrule, entries);
1361 			if (oldrule == NULL) {
1362 				pf_rm_rule(NULL, newrule);
1363 				error = EINVAL;
1364 				crit_exit();
1365 				break;
1366 			}
1367 		}
1368 
1369 		if (pcr->action == PF_CHANGE_REMOVE)
1370 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1371 		else {
1372 			if (oldrule == NULL)
1373 				TAILQ_INSERT_TAIL(
1374 				    ruleset->rules[rs_num].active.ptr,
1375 				    newrule, entries);
1376 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1377 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1378 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1379 			else
1380 				TAILQ_INSERT_AFTER(
1381 				    ruleset->rules[rs_num].active.ptr,
1382 				    oldrule, newrule, entries);
1383 		}
1384 
1385 		nr = 0;
1386 		TAILQ_FOREACH(oldrule,
1387 		    ruleset->rules[rs_num].active.ptr, entries)
1388 			oldrule->nr = nr++;
1389 
1390 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1391 		pf_remove_if_empty_ruleset(ruleset);
1392 		pf_update_anchor_rules();
1393 
1394 		ruleset->rules[rs_num].active.ticket++;
1395 		crit_exit();
1396 		break;
1397 	}
1398 
1399 	case DIOCCLRSTATES: {
1400 		struct pf_state		*state;
1401 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1402 		int			 killed = 0;
1403 
1404 		crit_enter();
1405 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1406 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1407 			    state->u.s.kif->pfik_name)) {
1408 				state->timeout = PFTM_PURGE;
1409 #if NPFSYNC
1410 				/* don't send out individual delete messages */
1411 				state->sync_flags = PFSTATE_NOSYNC;
1412 #endif
1413 				killed++;
1414 			}
1415 		}
1416 		pf_purge_expired_states();
1417 		pf_status.states = 0;
1418 		psk->psk_af = killed;
1419 #if NPFSYNC
1420 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1421 #endif
1422 		crit_exit();
1423 		break;
1424 	}
1425 
1426 	case DIOCKILLSTATES: {
1427 		struct pf_state		*state;
1428 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1429 		int			 killed = 0;
1430 
1431 		crit_enter();
1432 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1433 			if ((!psk->psk_af || state->af == psk->psk_af)
1434 			    && (!psk->psk_proto || psk->psk_proto ==
1435 			    state->proto) &&
1436 			    PF_MATCHA(psk->psk_src.not,
1437 			    &psk->psk_src.addr.v.a.addr,
1438 			    &psk->psk_src.addr.v.a.mask,
1439 			    &state->lan.addr, state->af) &&
1440 			    PF_MATCHA(psk->psk_dst.not,
1441 			    &psk->psk_dst.addr.v.a.addr,
1442 			    &psk->psk_dst.addr.v.a.mask,
1443 			    &state->ext.addr, state->af) &&
1444 			    (psk->psk_src.port_op == 0 ||
1445 			    pf_match_port(psk->psk_src.port_op,
1446 			    psk->psk_src.port[0], psk->psk_src.port[1],
1447 			    state->lan.port)) &&
1448 			    (psk->psk_dst.port_op == 0 ||
1449 			    pf_match_port(psk->psk_dst.port_op,
1450 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1451 			    state->ext.port)) &&
1452 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1453 			    state->u.s.kif->pfik_name))) {
1454 				state->timeout = PFTM_PURGE;
1455 				killed++;
1456 			}
1457 		}
1458 		pf_purge_expired_states();
1459 		crit_exit();
1460 		psk->psk_af = killed;
1461 		break;
1462 	}
1463 
1464 	case DIOCADDSTATE: {
1465 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1466 		struct pf_state		*state;
1467 		struct pfi_kif		*kif;
1468 
1469 		if (ps->state.timeout >= PFTM_MAX &&
1470 		    ps->state.timeout != PFTM_UNTIL_PACKET) {
1471 			error = EINVAL;
1472 			break;
1473 		}
1474 		state = pool_get(&pf_state_pl, PR_NOWAIT);
1475 		if (state == NULL) {
1476 			error = ENOMEM;
1477 			break;
1478 		}
1479 		crit_enter();
1480 		kif = pfi_lookup_create(ps->state.u.ifname);
1481 		if (kif == NULL) {
1482 			pool_put(&pf_state_pl, state);
1483 			error = ENOENT;
1484 			crit_exit();
1485 			break;
1486 		}
1487 		bcopy(&ps->state, state, sizeof(struct pf_state));
1488 		bzero(&state->u, sizeof(state->u));
1489 		state->rule.ptr = &pf_default_rule;
1490 		state->nat_rule.ptr = NULL;
1491 		state->anchor.ptr = NULL;
1492 		state->rt_kif = NULL;
1493 		state->creation = time_second;
1494 		state->pfsync_time = 0;
1495 		state->packets[0] = state->packets[1] = 0;
1496 		state->bytes[0] = state->bytes[1] = 0;
1497 
1498 		if (pf_insert_state(kif, state)) {
1499 			pfi_maybe_destroy(kif);
1500 			pool_put(&pf_state_pl, state);
1501 			error = ENOMEM;
1502 		}
1503 		crit_exit();
1504 		break;
1505 	}
1506 
1507 	case DIOCGETSTATE: {
1508 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1509 		struct pf_state		*state;
1510 		u_int32_t		 nr;
1511 
1512 		nr = 0;
1513 		crit_enter();
1514 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1515 			if (nr >= ps->nr)
1516 				break;
1517 			nr++;
1518 		}
1519 		if (state == NULL) {
1520 			error = EBUSY;
1521 			crit_exit();
1522 			break;
1523 		}
1524 		bcopy(state, &ps->state, sizeof(struct pf_state));
1525 		ps->state.rule.nr = state->rule.ptr->nr;
1526 		ps->state.nat_rule.nr = (state->nat_rule.ptr == NULL) ?
1527 		    (uint32_t)(-1) : state->nat_rule.ptr->nr;
1528 		ps->state.anchor.nr = (state->anchor.ptr == NULL) ?
1529 		    (uint32_t)(-1) : state->anchor.ptr->nr;
1530 		crit_exit();
1531 		ps->state.expire = pf_state_expires(state);
1532 		if (ps->state.expire > time_second)
1533 			ps->state.expire -= time_second;
1534 		else
1535 			ps->state.expire = 0;
1536 		break;
1537 	}
1538 
1539 	case DIOCGETSTATES: {
1540 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1541 		struct pf_state		*state;
1542 		struct pf_state		*p, pstore;
1543 		struct pfi_kif		*kif;
1544 		u_int32_t		 nr = 0;
1545 		int			 space = ps->ps_len;
1546 
1547 		if (space == 0) {
1548 			crit_enter();
1549 			TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states)
1550 				nr += kif->pfik_states;
1551 			crit_exit();
1552 			ps->ps_len = sizeof(struct pf_state) * nr;
1553 			return (0);
1554 		}
1555 
1556 		crit_enter();
1557 		p = ps->ps_states;
1558 		TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states)
1559 			RB_FOREACH(state, pf_state_tree_ext_gwy,
1560 			    &kif->pfik_ext_gwy) {
1561 				int	secs = time_second;
1562 
1563 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1564 					break;
1565 
1566 				bcopy(state, &pstore, sizeof(pstore));
1567 				strlcpy(pstore.u.ifname, kif->pfik_name,
1568 				    sizeof(pstore.u.ifname));
1569 				pstore.rule.nr = state->rule.ptr->nr;
1570 				pstore.nat_rule.nr = (state->nat_rule.ptr ==
1571 				    NULL) ? (uint32_t)(-1)
1572 					  : state->nat_rule.ptr->nr;
1573 				pstore.anchor.nr = (state->anchor.ptr ==
1574 				    NULL) ? (uint32_t)(-1)
1575 					  : state->anchor.ptr->nr;
1576 				pstore.creation = secs - pstore.creation;
1577 				pstore.expire = pf_state_expires(state);
1578 				if (pstore.expire > secs)
1579 					pstore.expire -= secs;
1580 				else
1581 					pstore.expire = 0;
1582 				error = copyout(&pstore, p, sizeof(*p));
1583 				if (error) {
1584 					crit_exit();
1585 					goto fail;
1586 				}
1587 				p++;
1588 				nr++;
1589 			}
1590 		ps->ps_len = sizeof(struct pf_state) * nr;
1591 		crit_exit();
1592 		break;
1593 	}
1594 
1595 	case DIOCGETSTATUS: {
1596 		struct pf_status *s = (struct pf_status *)addr;
1597 		bcopy(&pf_status, s, sizeof(struct pf_status));
1598 		pfi_fill_oldstatus(s);
1599 		break;
1600 	}
1601 
1602 	case DIOCSETSTATUSIF: {
1603 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1604 
1605 		if (pi->ifname[0] == 0) {
1606 			bzero(pf_status.ifname, IFNAMSIZ);
1607 			break;
1608 		}
1609 		if (ifunit(pi->ifname) == NULL) {
1610 			error = EINVAL;
1611 			break;
1612 		}
1613 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1614 		break;
1615 	}
1616 
1617 	case DIOCCLRSTATUS: {
1618 		bzero(pf_status.counters, sizeof(pf_status.counters));
1619 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1620 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
1621 		if (*pf_status.ifname)
1622 			pfi_clr_istats(pf_status.ifname, NULL,
1623 			    PFI_FLAG_INSTANCE);
1624 		break;
1625 	}
1626 
1627 	case DIOCNATLOOK: {
1628 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1629 		struct pf_state		*state;
1630 		struct pf_state		 key;
1631 		int			 m = 0, direction = pnl->direction;
1632 
1633 		key.af = pnl->af;
1634 		key.proto = pnl->proto;
1635 
1636 		if (!pnl->proto ||
1637 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1638 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1639 		    !pnl->dport || !pnl->sport)
1640 			error = EINVAL;
1641 		else {
1642 			crit_enter();
1643 
1644 			/*
1645 			 * userland gives us source and dest of connection,
1646 			 * reverse the lookup so we ask for what happens with
1647 			 * the return traffic, enabling us to find it in the
1648 			 * state tree.
1649 			 */
1650 			if (direction == PF_IN) {
1651 				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
1652 				key.ext.port = pnl->dport;
1653 				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
1654 				key.gwy.port = pnl->sport;
1655 				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
1656 			} else {
1657 				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
1658 				key.lan.port = pnl->dport;
1659 				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
1660 				key.ext.port = pnl->sport;
1661 				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
1662 			}
1663 			if (m > 1)
1664 				error = E2BIG;	/* more than one state */
1665 			else if (state != NULL) {
1666 				if (direction == PF_IN) {
1667 					PF_ACPY(&pnl->rsaddr, &state->lan.addr,
1668 					    state->af);
1669 					pnl->rsport = state->lan.port;
1670 					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
1671 					    pnl->af);
1672 					pnl->rdport = pnl->dport;
1673 				} else {
1674 					PF_ACPY(&pnl->rdaddr, &state->gwy.addr,
1675 					    state->af);
1676 					pnl->rdport = state->gwy.port;
1677 					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
1678 					    pnl->af);
1679 					pnl->rsport = pnl->sport;
1680 				}
1681 			} else
1682 				error = ENOENT;
1683 			crit_exit();
1684 		}
1685 		break;
1686 	}
1687 
1688 	case DIOCSETTIMEOUT: {
1689 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1690 		int		 old;
1691 
1692 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1693 		    pt->seconds < 0) {
1694 			error = EINVAL;
1695 			goto fail;
1696 		}
1697 		old = pf_default_rule.timeout[pt->timeout];
1698 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
1699 		pt->seconds = old;
1700 		break;
1701 	}
1702 
1703 	case DIOCGETTIMEOUT: {
1704 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1705 
1706 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1707 			error = EINVAL;
1708 			goto fail;
1709 		}
1710 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1711 		break;
1712 	}
1713 
1714 	case DIOCGETLIMIT: {
1715 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1716 
1717 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1718 			error = EINVAL;
1719 			goto fail;
1720 		}
1721 		pl->limit = pf_pool_limits[pl->index].limit;
1722 		break;
1723 	}
1724 
1725 	case DIOCSETLIMIT: {
1726 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1727 		int			 old_limit;
1728 
1729 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1730 		    pf_pool_limits[pl->index].pp == NULL) {
1731 			error = EINVAL;
1732 			goto fail;
1733 		}
1734 
1735 		/* XXX Get an API to set limits on the zone/pool */
1736 		old_limit = pf_pool_limits[pl->index].limit;
1737 		pf_pool_limits[pl->index].limit = pl->limit;
1738 		pl->limit = old_limit;
1739 		break;
1740 	}
1741 
1742 	case DIOCSETDEBUG: {
1743 		u_int32_t	*level = (u_int32_t *)addr;
1744 
1745 		pf_status.debug = *level;
1746 		break;
1747 	}
1748 
1749 	case DIOCCLRRULECTRS: {
1750 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1751 		struct pf_rule		*rule;
1752 
1753 		crit_enter();
1754 		TAILQ_FOREACH(rule,
1755 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries)
1756 			rule->evaluations = rule->packets =
1757 			    rule->bytes = 0;
1758 		crit_exit();
1759 		break;
1760 	}
1761 
1762 	case DIOCGIFSPEED: {
1763 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1764 		struct pf_ifspeed	ps;
1765 		struct ifnet		*ifp;
1766 
1767 		if (psp->ifname[0] != 0) {
1768 			/* Can we completely trust user-land? */
1769 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1770 			ifp = ifunit(ps.ifname);
1771 			if (ifp )
1772 				psp->baudrate = ifp->if_baudrate;
1773 			else
1774 				error = EINVAL;
1775 		} else
1776 			error = EINVAL;
1777 		break;
1778 	}
1779 #ifdef ALTQ
1780 	case DIOCSTARTALTQ: {
1781 		struct pf_altq		*altq;
1782 		struct ifnet		*ifp;
1783 		struct tb_profile	 tb;
1784 
1785 		/* enable all altq interfaces on active list */
1786 		crit_enter();
1787 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1788 			if (altq->qname[0] == 0) {
1789 				if ((ifp = ifunit(altq->ifname)) == NULL) {
1790 					error = EINVAL;
1791 					break;
1792 				}
1793 				if (ifp->if_snd.altq_type != ALTQT_NONE)
1794 					error = altq_enable(&ifp->if_snd);
1795 				if (error != 0)
1796 					break;
1797 				/* set tokenbucket regulator */
1798 				tb.rate = altq->ifbandwidth;
1799 				tb.depth = altq->tbrsize;
1800 				error = tbr_set(&ifp->if_snd, &tb);
1801 				if (error != 0)
1802 					break;
1803 			}
1804 		}
1805 		crit_exit();
1806 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1807 		break;
1808 	}
1809 
1810 	case DIOCSTOPALTQ: {
1811 		struct pf_altq		*altq;
1812 		struct ifnet		*ifp;
1813 		struct tb_profile	 tb;
1814 		int			 err;
1815 
1816 		/* disable all altq interfaces on active list */
1817 		crit_enter();
1818 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1819 			if (altq->qname[0] == 0) {
1820 				if ((ifp = ifunit(altq->ifname)) == NULL) {
1821 					error = EINVAL;
1822 					break;
1823 				}
1824 				if (ifp->if_snd.altq_type != ALTQT_NONE) {
1825 					err = altq_disable(&ifp->if_snd);
1826 					if (err != 0 && error == 0)
1827 						error = err;
1828 				}
1829 				/* clear tokenbucket regulator */
1830 				tb.rate = 0;
1831 				err = tbr_set(&ifp->if_snd, &tb);
1832 				if (err != 0 && error == 0)
1833 					error = err;
1834 			}
1835 		}
1836 		crit_exit();
1837 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1838 		break;
1839 	}
1840 
1841 	case DIOCBEGINALTQS: {
1842 		u_int32_t	*ticket = (u_int32_t *)addr;
1843 
1844 		error = pf_begin_altq(ticket);
1845 		break;
1846 	}
1847 
1848 	case DIOCADDALTQ: {
1849 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1850 		struct pf_altq		*altq, *a;
1851 
1852 		if (pa->ticket != ticket_altqs_inactive) {
1853 			error = EBUSY;
1854 			break;
1855 		}
1856 		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
1857 		if (altq == NULL) {
1858 			error = ENOMEM;
1859 			break;
1860 		}
1861 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
1862 
1863 		/*
1864 		 * if this is for a queue, find the discipline and
1865 		 * copy the necessary fields
1866 		 */
1867 		if (altq->qname[0] != 0) {
1868 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
1869 				error = EBUSY;
1870 				pool_put(&pf_altq_pl, altq);
1871 				break;
1872 			}
1873 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
1874 				if (strncmp(a->ifname, altq->ifname,
1875 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
1876 					altq->altq_disc = a->altq_disc;
1877 					break;
1878 				}
1879 			}
1880 		}
1881 
1882 		error = altq_add(altq);
1883 		if (error) {
1884 			pool_put(&pf_altq_pl, altq);
1885 			break;
1886 		}
1887 
1888 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
1889 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1890 		break;
1891 	}
1892 
1893 	case DIOCCOMMITALTQS: {
1894 		u_int32_t		ticket = *(u_int32_t *)addr;
1895 
1896 		error = pf_commit_altq(ticket);
1897 		break;
1898 	}
1899 
1900 	case DIOCGETALTQS: {
1901 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1902 		struct pf_altq		*altq;
1903 
1904 		pa->nr = 0;
1905 		crit_enter();
1906 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
1907 			pa->nr++;
1908 		pa->ticket = ticket_altqs_active;
1909 		crit_exit();
1910 		break;
1911 	}
1912 
1913 	case DIOCGETALTQ: {
1914 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1915 		struct pf_altq		*altq;
1916 		u_int32_t		 nr;
1917 
1918 		if (pa->ticket != ticket_altqs_active) {
1919 			error = EBUSY;
1920 			break;
1921 		}
1922 		nr = 0;
1923 		crit_enter();
1924 		altq = TAILQ_FIRST(pf_altqs_active);
1925 		while ((altq != NULL) && (nr < pa->nr)) {
1926 			altq = TAILQ_NEXT(altq, entries);
1927 			nr++;
1928 		}
1929 		if (altq == NULL) {
1930 			error = EBUSY;
1931 			crit_exit();
1932 			break;
1933 		}
1934 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1935 		crit_exit();
1936 		break;
1937 	}
1938 
1939 	case DIOCCHANGEALTQ:
1940 		/* CHANGEALTQ not supported yet! */
1941 		error = ENODEV;
1942 		break;
1943 
1944 	case DIOCGETQSTATS: {
1945 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
1946 		struct pf_altq		*altq;
1947 		u_int32_t		 nr;
1948 		int			 nbytes;
1949 
1950 		if (pq->ticket != ticket_altqs_active) {
1951 			error = EBUSY;
1952 			break;
1953 		}
1954 		nbytes = pq->nbytes;
1955 		nr = 0;
1956 		crit_enter();
1957 		altq = TAILQ_FIRST(pf_altqs_active);
1958 		while ((altq != NULL) && (nr < pq->nr)) {
1959 			altq = TAILQ_NEXT(altq, entries);
1960 			nr++;
1961 		}
1962 		if (altq == NULL) {
1963 			error = EBUSY;
1964 			crit_exit();
1965 			break;
1966 		}
1967 		error = altq_getqstats(altq, pq->buf, &nbytes);
1968 		crit_exit();
1969 		if (error == 0) {
1970 			pq->scheduler = altq->scheduler;
1971 			pq->nbytes = nbytes;
1972 		}
1973 		break;
1974 	}
1975 #endif /* ALTQ */
1976 
1977 	case DIOCBEGINADDRS: {
1978 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
1979 
1980 		pf_empty_pool(&pf_pabuf);
1981 		pp->ticket = ++ticket_pabuf;
1982 		break;
1983 	}
1984 
1985 	case DIOCADDADDR: {
1986 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
1987 
1988 #ifndef INET
1989 		if (pp->af == AF_INET) {
1990 			error = EAFNOSUPPORT;
1991 			break;
1992 		}
1993 #endif /* INET */
1994 #ifndef INET6
1995 		if (pp->af == AF_INET6) {
1996 			error = EAFNOSUPPORT;
1997 			break;
1998 		}
1999 #endif /* INET6 */
2000 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2001 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2002 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2003 			error = EINVAL;
2004 			break;
2005 		}
2006 		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2007 		if (pa == NULL) {
2008 			error = ENOMEM;
2009 			break;
2010 		}
2011 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2012 		if (pa->ifname[0]) {
2013 			pa->kif = pfi_attach_rule(pa->ifname);
2014 			if (pa->kif == NULL) {
2015 				pool_put(&pf_pooladdr_pl, pa);
2016 				error = EINVAL;
2017 				break;
2018 			}
2019 		}
2020 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2021 			pfi_dynaddr_remove(&pa->addr);
2022 			pfi_detach_rule(pa->kif);
2023 			pool_put(&pf_pooladdr_pl, pa);
2024 			error = EINVAL;
2025 			break;
2026 		}
2027 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2028 		break;
2029 	}
2030 
2031 	case DIOCGETADDRS: {
2032 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2033 
2034 		pp->nr = 0;
2035 		crit_enter();
2036 		pool = pf_get_pool(pp->anchor, pp->ruleset, pp->ticket,
2037 		    pp->r_action, pp->r_num, 0, 1, 0);
2038 		if (pool == NULL) {
2039 			error = EBUSY;
2040 			crit_exit();
2041 			break;
2042 		}
2043 		TAILQ_FOREACH(pa, &pool->list, entries)
2044 			pp->nr++;
2045 		crit_exit();
2046 		break;
2047 	}
2048 
2049 	case DIOCGETADDR: {
2050 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2051 		u_int32_t		 nr = 0;
2052 
2053 		crit_enter();
2054 		pool = pf_get_pool(pp->anchor, pp->ruleset, pp->ticket,
2055 		    pp->r_action, pp->r_num, 0, 1, 1);
2056 		if (pool == NULL) {
2057 			error = EBUSY;
2058 			crit_exit();
2059 			break;
2060 		}
2061 		pa = TAILQ_FIRST(&pool->list);
2062 		while ((pa != NULL) && (nr < pp->nr)) {
2063 			pa = TAILQ_NEXT(pa, entries);
2064 			nr++;
2065 		}
2066 		if (pa == NULL) {
2067 			error = EBUSY;
2068 			crit_exit();
2069 			break;
2070 		}
2071 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2072 		pfi_dynaddr_copyout(&pp->addr.addr);
2073 		pf_tbladdr_copyout(&pp->addr.addr);
2074 		crit_exit();
2075 		break;
2076 	}
2077 
2078 	case DIOCCHANGEADDR: {
2079 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2080 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2081 		struct pf_ruleset	*ruleset;
2082 
2083 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2084 		    pca->action > PF_CHANGE_REMOVE) {
2085 			error = EINVAL;
2086 			break;
2087 		}
2088 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2089 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2090 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2091 			error = EINVAL;
2092 			break;
2093 		}
2094 
2095 		ruleset = pf_find_ruleset(pca->anchor, pca->ruleset);
2096 		if (ruleset == NULL) {
2097 			error = EBUSY;
2098 			break;
2099 		}
2100 		pool = pf_get_pool(pca->anchor, pca->ruleset, pca->ticket,
2101 		    pca->r_action, pca->r_num, pca->r_last, 1, 1);
2102 		if (pool == NULL) {
2103 			error = EBUSY;
2104 			break;
2105 		}
2106 		if (pca->action != PF_CHANGE_REMOVE) {
2107 			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2108 			if (newpa == NULL) {
2109 				error = ENOMEM;
2110 				break;
2111 			}
2112 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2113 #ifndef INET
2114 			if (pca->af == AF_INET) {
2115 				pool_put(&pf_pooladdr_pl, newpa);
2116 				error = EAFNOSUPPORT;
2117 				break;
2118 			}
2119 #endif /* INET */
2120 #ifndef INET6
2121 			if (pca->af == AF_INET6) {
2122 				pool_put(&pf_pooladdr_pl, newpa);
2123 				error = EAFNOSUPPORT;
2124 				break;
2125 			}
2126 #endif /* INET6 */
2127 			if (newpa->ifname[0]) {
2128 				newpa->kif = pfi_attach_rule(newpa->ifname);
2129 				if (newpa->kif == NULL) {
2130 					pool_put(&pf_pooladdr_pl, newpa);
2131 					error = EINVAL;
2132 					break;
2133 				}
2134 			} else
2135 				newpa->kif = NULL;
2136 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2137 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2138 				pfi_dynaddr_remove(&newpa->addr);
2139 				pfi_detach_rule(newpa->kif);
2140 				pool_put(&pf_pooladdr_pl, newpa);
2141 				error = EINVAL;
2142 				break;
2143 			}
2144 		}
2145 
2146 		crit_enter();
2147 
2148 		if (pca->action == PF_CHANGE_ADD_HEAD)
2149 			oldpa = TAILQ_FIRST(&pool->list);
2150 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2151 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2152 		else {
2153 			int	i = 0;
2154 
2155 			oldpa = TAILQ_FIRST(&pool->list);
2156 			while ((oldpa != NULL) && (i < pca->nr)) {
2157 				oldpa = TAILQ_NEXT(oldpa, entries);
2158 				i++;
2159 			}
2160 			if (oldpa == NULL) {
2161 				error = EINVAL;
2162 				crit_exit();
2163 				break;
2164 			}
2165 		}
2166 
2167 		if (pca->action == PF_CHANGE_REMOVE) {
2168 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2169 			pfi_dynaddr_remove(&oldpa->addr);
2170 			pf_tbladdr_remove(&oldpa->addr);
2171 			pfi_detach_rule(oldpa->kif);
2172 			pool_put(&pf_pooladdr_pl, oldpa);
2173 		} else {
2174 			if (oldpa == NULL)
2175 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2176 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2177 			    pca->action == PF_CHANGE_ADD_BEFORE)
2178 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2179 			else
2180 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2181 				    newpa, entries);
2182 		}
2183 
2184 		pool->cur = TAILQ_FIRST(&pool->list);
2185 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2186 		    pca->af);
2187 		crit_exit();
2188 		break;
2189 	}
2190 
2191 	case DIOCGETANCHORS: {
2192 		struct pfioc_anchor	*pa = (struct pfioc_anchor *)addr;
2193 		struct pf_anchor	*anchor;
2194 
2195 		pa->nr = 0;
2196 		TAILQ_FOREACH(anchor, &pf_anchors, entries)
2197 			pa->nr++;
2198 		break;
2199 	}
2200 
2201 	case DIOCGETANCHOR: {
2202 		struct pfioc_anchor	*pa = (struct pfioc_anchor *)addr;
2203 		struct pf_anchor	*anchor;
2204 		u_int32_t		 nr = 0;
2205 
2206 		anchor = TAILQ_FIRST(&pf_anchors);
2207 		while (anchor != NULL && nr < pa->nr) {
2208 			anchor = TAILQ_NEXT(anchor, entries);
2209 			nr++;
2210 		}
2211 		if (anchor == NULL)
2212 			error = EBUSY;
2213 		else
2214 			bcopy(anchor->name, pa->name, sizeof(pa->name));
2215 		break;
2216 	}
2217 
2218 	case DIOCGETRULESETS: {
2219 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2220 		struct pf_anchor	*anchor;
2221 		struct pf_ruleset	*ruleset;
2222 
2223 		pr->anchor[PF_ANCHOR_NAME_SIZE-1] = 0;
2224 		if ((anchor = pf_find_anchor(pr->anchor)) == NULL) {
2225 			error = EINVAL;
2226 			break;
2227 		}
2228 		pr->nr = 0;
2229 		TAILQ_FOREACH(ruleset, &anchor->rulesets, entries)
2230 			pr->nr++;
2231 		break;
2232 	}
2233 
2234 	case DIOCGETRULESET: {
2235 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2236 		struct pf_anchor	*anchor;
2237 		struct pf_ruleset	*ruleset;
2238 		u_int32_t		 nr = 0;
2239 
2240 		if ((anchor = pf_find_anchor(pr->anchor)) == NULL) {
2241 			error = EINVAL;
2242 			break;
2243 		}
2244 		ruleset = TAILQ_FIRST(&anchor->rulesets);
2245 		while (ruleset != NULL && nr < pr->nr) {
2246 			ruleset = TAILQ_NEXT(ruleset, entries);
2247 			nr++;
2248 		}
2249 		if (ruleset == NULL)
2250 			error = EBUSY;
2251 		else
2252 			bcopy(ruleset->name, pr->name, sizeof(pr->name));
2253 		break;
2254 	}
2255 
2256 	case DIOCRCLRTABLES: {
2257 		struct pfioc_table *io = (struct pfioc_table *)addr;
2258 
2259 		if (io->pfrio_esize != 0) {
2260 			error = ENODEV;
2261 			break;
2262 		}
2263 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2264 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2265 		break;
2266 	}
2267 
2268 	case DIOCRADDTABLES: {
2269 		struct pfioc_table *io = (struct pfioc_table *)addr;
2270 
2271 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2272 			error = ENODEV;
2273 			break;
2274 		}
2275 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2276 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2277 		break;
2278 	}
2279 
2280 	case DIOCRDELTABLES: {
2281 		struct pfioc_table *io = (struct pfioc_table *)addr;
2282 
2283 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2284 			error = ENODEV;
2285 			break;
2286 		}
2287 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2288 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2289 		break;
2290 	}
2291 
2292 	case DIOCRGETTABLES: {
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_get_tables(&io->pfrio_table, io->pfrio_buffer,
2300 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2301 		break;
2302 	}
2303 
2304 	case DIOCRGETTSTATS: {
2305 		struct pfioc_table *io = (struct pfioc_table *)addr;
2306 
2307 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2308 			error = ENODEV;
2309 			break;
2310 		}
2311 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2312 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2313 		break;
2314 	}
2315 
2316 	case DIOCRCLRTSTATS: {
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_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2324 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2325 		break;
2326 	}
2327 
2328 	case DIOCRSETTFLAGS: {
2329 		struct pfioc_table *io = (struct pfioc_table *)addr;
2330 
2331 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2332 			error = ENODEV;
2333 			break;
2334 		}
2335 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2336 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2337 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2338 		break;
2339 	}
2340 
2341 	case DIOCRCLRADDRS: {
2342 		struct pfioc_table *io = (struct pfioc_table *)addr;
2343 
2344 		if (io->pfrio_esize != 0) {
2345 			error = ENODEV;
2346 			break;
2347 		}
2348 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2349 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2350 		break;
2351 	}
2352 
2353 	case DIOCRADDADDRS: {
2354 		struct pfioc_table *io = (struct pfioc_table *)addr;
2355 
2356 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2357 			error = ENODEV;
2358 			break;
2359 		}
2360 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2361 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2362 		    PFR_FLAG_USERIOCTL);
2363 		break;
2364 	}
2365 
2366 	case DIOCRDELADDRS: {
2367 		struct pfioc_table *io = (struct pfioc_table *)addr;
2368 
2369 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2370 			error = ENODEV;
2371 			break;
2372 		}
2373 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2374 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2375 		    PFR_FLAG_USERIOCTL);
2376 		break;
2377 	}
2378 
2379 	case DIOCRSETADDRS: {
2380 		struct pfioc_table *io = (struct pfioc_table *)addr;
2381 
2382 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2383 			error = ENODEV;
2384 			break;
2385 		}
2386 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2387 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2388 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2389 		    PFR_FLAG_USERIOCTL);
2390 		break;
2391 	}
2392 
2393 	case DIOCRGETADDRS: {
2394 		struct pfioc_table *io = (struct pfioc_table *)addr;
2395 
2396 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2397 			error = ENODEV;
2398 			break;
2399 		}
2400 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2401 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2402 		break;
2403 	}
2404 
2405 	case DIOCRGETASTATS: {
2406 		struct pfioc_table *io = (struct pfioc_table *)addr;
2407 
2408 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2409 			error = ENODEV;
2410 			break;
2411 		}
2412 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2413 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2414 		break;
2415 	}
2416 
2417 	case DIOCRCLRASTATS: {
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_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2425 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2426 		    PFR_FLAG_USERIOCTL);
2427 		break;
2428 	}
2429 
2430 	case DIOCRTSTADDRS: {
2431 		struct pfioc_table *io = (struct pfioc_table *)addr;
2432 
2433 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2434 			error = ENODEV;
2435 			break;
2436 		}
2437 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2438 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2439 		    PFR_FLAG_USERIOCTL);
2440 		break;
2441 	}
2442 
2443 	case DIOCRINABEGIN: {
2444 		struct pfioc_table *io = (struct pfioc_table *)addr;
2445 
2446 		if (io->pfrio_esize != 0) {
2447 			error = ENODEV;
2448 			break;
2449 		}
2450 		error = pfr_ina_begin(&io->pfrio_table, &io->pfrio_ticket,
2451 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2452 		break;
2453 	}
2454 
2455 	case DIOCRINACOMMIT: {
2456 		struct pfioc_table *io = (struct pfioc_table *)addr;
2457 
2458 		if (io->pfrio_esize != 0) {
2459 			error = ENODEV;
2460 			break;
2461 		}
2462 		error = pfr_ina_commit(&io->pfrio_table, io->pfrio_ticket,
2463 		    &io->pfrio_nadd, &io->pfrio_nchange, io->pfrio_flags |
2464 		    PFR_FLAG_USERIOCTL);
2465 		break;
2466 	}
2467 
2468 	case DIOCRINADEFINE: {
2469 		struct pfioc_table *io = (struct pfioc_table *)addr;
2470 
2471 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2472 			error = ENODEV;
2473 			break;
2474 		}
2475 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2476 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2477 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2478 		break;
2479 	}
2480 
2481 	case DIOCOSFPADD: {
2482 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2483 		crit_enter();
2484 		error = pf_osfp_add(io);
2485 		crit_exit();
2486 		break;
2487 	}
2488 
2489 	case DIOCOSFPGET: {
2490 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2491 		crit_enter();
2492 		error = pf_osfp_get(io);
2493 		crit_exit();
2494 		break;
2495 	}
2496 
2497 	case DIOCXBEGIN: {
2498 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2499 		struct pfioc_trans_e	 ioe;
2500 		struct pfr_table	 table;
2501 		int			 i;
2502 
2503 		if (io->esize != sizeof(ioe)) {
2504 			error = ENODEV;
2505 			goto fail;
2506 		}
2507 		for (i = 0; i < io->size; i++) {
2508 			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2509 				error = EFAULT;
2510 				goto fail;
2511 			}
2512 			switch (ioe.rs_num) {
2513 #ifdef ALTQ
2514 			case PF_RULESET_ALTQ:
2515 				if (ioe.anchor[0] || ioe.ruleset[0]) {
2516 					error = EINVAL;
2517 					goto fail;
2518 				}
2519 				if ((error = pf_begin_altq(&ioe.ticket)))
2520 					goto fail;
2521 				break;
2522 #endif /* ALTQ */
2523 			case PF_RULESET_TABLE:
2524 				bzero(&table, sizeof(table));
2525 				strlcpy(table.pfrt_anchor, ioe.anchor,
2526 				    sizeof(table.pfrt_anchor));
2527 				strlcpy(table.pfrt_ruleset, ioe.ruleset,
2528 				    sizeof(table.pfrt_ruleset));
2529 				if ((error = pfr_ina_begin(&table,
2530 				    &ioe.ticket, NULL, 0)))
2531 					goto fail;
2532 				break;
2533 			default:
2534 				if ((error = pf_begin_rules(&ioe.ticket,
2535 				    ioe.rs_num, ioe.anchor, ioe.ruleset)))
2536 					goto fail;
2537 				break;
2538 			}
2539 			if (copyout(&ioe, io->array+i, sizeof(io->array[i]))) {
2540 				error = EFAULT;
2541 				goto fail;
2542 			}
2543 		}
2544 		break;
2545 	}
2546 
2547 	case DIOCXROLLBACK: {
2548 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2549 		struct pfioc_trans_e	 ioe;
2550 		struct pfr_table	 table;
2551 		int			 i;
2552 
2553 		if (io->esize != sizeof(ioe)) {
2554 			error = ENODEV;
2555 			goto fail;
2556 		}
2557 		for (i = 0; i < io->size; i++) {
2558 			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2559 				error = EFAULT;
2560 				goto fail;
2561 			}
2562 			switch (ioe.rs_num) {
2563 #ifdef ALTQ
2564 			case PF_RULESET_ALTQ:
2565 				if (ioe.anchor[0] || ioe.ruleset[0]) {
2566 					error = EINVAL;
2567 					goto fail;
2568 				}
2569 				if ((error = pf_rollback_altq(ioe.ticket)))
2570 					goto fail; /* really bad */
2571 				break;
2572 #endif /* ALTQ */
2573 			case PF_RULESET_TABLE:
2574 				bzero(&table, sizeof(table));
2575 				strlcpy(table.pfrt_anchor, ioe.anchor,
2576 				    sizeof(table.pfrt_anchor));
2577 				strlcpy(table.pfrt_ruleset, ioe.ruleset,
2578 				    sizeof(table.pfrt_ruleset));
2579 				if ((error = pfr_ina_rollback(&table,
2580 				    ioe.ticket, NULL, 0)))
2581 					goto fail; /* really bad */
2582 				break;
2583 			default:
2584 				if ((error = pf_rollback_rules(ioe.ticket,
2585 				    ioe.rs_num, ioe.anchor, ioe.ruleset)))
2586 					goto fail; /* really bad */
2587 				break;
2588 			}
2589 		}
2590 		break;
2591 	}
2592 
2593 	case DIOCXCOMMIT: {
2594 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2595 		struct pfioc_trans_e	 ioe;
2596 		struct pfr_table	 table;
2597 		struct pf_ruleset	*rs;
2598 		int			 i;
2599 
2600 		if (io->esize != sizeof(ioe)) {
2601 			error = ENODEV;
2602 			goto fail;
2603 		}
2604 		/* first makes sure everything will succeed */
2605 		for (i = 0; i < io->size; i++) {
2606 			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2607 				error = EFAULT;
2608 				goto fail;
2609 			}
2610 			switch (ioe.rs_num) {
2611 #ifdef ALTQ
2612 			case PF_RULESET_ALTQ:
2613 				if (ioe.anchor[0] || ioe.ruleset[0]) {
2614 					error = EINVAL;
2615 					goto fail;
2616 				}
2617 				if (!altqs_inactive_open || ioe.ticket !=
2618 				    ticket_altqs_inactive) {
2619 					error = EBUSY;
2620 					goto fail;
2621 				}
2622 				break;
2623 #endif /* ALTQ */
2624 			case PF_RULESET_TABLE:
2625 				rs = pf_find_ruleset(ioe.anchor, ioe.ruleset);
2626 				if (rs == NULL || !rs->topen || ioe.ticket !=
2627 				     rs->tticket) {
2628 					error = EBUSY;
2629 					goto fail;
2630 				}
2631 				break;
2632 			default:
2633 				if (ioe.rs_num < 0 || ioe.rs_num >=
2634 				    PF_RULESET_MAX) {
2635 					error = EINVAL;
2636 					goto fail;
2637 				}
2638 				rs = pf_find_ruleset(ioe.anchor, ioe.ruleset);
2639 				if (rs == NULL ||
2640 				    !rs->rules[ioe.rs_num].inactive.open ||
2641 				    rs->rules[ioe.rs_num].inactive.ticket !=
2642 				    ioe.ticket) {
2643 					error = EBUSY;
2644 					goto fail;
2645 				}
2646 				break;
2647 			}
2648 		}
2649 		/* now do the commit - no errors should happen here */
2650 		for (i = 0; i < io->size; i++) {
2651 			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2652 				error = EFAULT;
2653 				goto fail;
2654 			}
2655 			switch (ioe.rs_num) {
2656 #ifdef ALTQ
2657 			case PF_RULESET_ALTQ:
2658 				if ((error = pf_commit_altq(ioe.ticket)))
2659 					goto fail; /* really bad */
2660 				break;
2661 #endif /* ALTQ */
2662 			case PF_RULESET_TABLE:
2663 				bzero(&table, sizeof(table));
2664 				strlcpy(table.pfrt_anchor, ioe.anchor,
2665 				    sizeof(table.pfrt_anchor));
2666 				strlcpy(table.pfrt_ruleset, ioe.ruleset,
2667 				    sizeof(table.pfrt_ruleset));
2668 				if ((error = pfr_ina_commit(&table, ioe.ticket,
2669 				    NULL, NULL, 0)))
2670 					goto fail; /* really bad */
2671 				break;
2672 			default:
2673 				if ((error = pf_commit_rules(ioe.ticket,
2674 				    ioe.rs_num, ioe.anchor, ioe.ruleset)))
2675 					goto fail; /* really bad */
2676 				break;
2677 			}
2678 		}
2679 		break;
2680 	}
2681 
2682 	case DIOCGETSRCNODES: {
2683 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2684 		struct pf_src_node	*n;
2685 		struct pf_src_node *p, pstore;
2686 		u_int32_t		 nr = 0;
2687 		int			 space = psn->psn_len;
2688 
2689 		if (space == 0) {
2690 			crit_enter();
2691 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2692 				nr++;
2693 			crit_exit();
2694 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2695 			return (0);
2696 		}
2697 
2698 		crit_enter();
2699 		p = psn->psn_src_nodes;
2700 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2701 			int	secs = time_second;
2702 
2703 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
2704 				break;
2705 
2706 			bcopy(n, &pstore, sizeof(pstore));
2707 			if (n->rule.ptr != NULL)
2708 				pstore.rule.nr = n->rule.ptr->nr;
2709 			pstore.creation = secs - pstore.creation;
2710 			if (pstore.expire > secs)
2711 				pstore.expire -= secs;
2712 			else
2713 				pstore.expire = 0;
2714 			error = copyout(&pstore, p, sizeof(*p));
2715 			if (error) {
2716 				crit_exit();
2717 				goto fail;
2718 			}
2719 			p++;
2720 			nr++;
2721 		}
2722 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2723 		crit_exit();
2724 		break;
2725 	}
2726 
2727 	case DIOCCLRSRCNODES: {
2728 		struct pf_src_node	*n;
2729 		struct pf_state		*state;
2730 
2731 		crit_enter();
2732 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2733 			state->src_node = NULL;
2734 			state->nat_src_node = NULL;
2735 		}
2736 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2737 			n->expire = 1;
2738 			n->states = 0;
2739 		}
2740 		pf_purge_expired_src_nodes();
2741 		pf_status.src_nodes = 0;
2742 		crit_exit();
2743 		break;
2744 	}
2745 
2746 	case DIOCSETHOSTID: {
2747 		u_int32_t	*hostid = (u_int32_t *)addr;
2748 
2749 		if (*hostid == 0) {
2750 			error = EINVAL;
2751 			goto fail;
2752 		}
2753 		pf_status.hostid = *hostid;
2754 		break;
2755 	}
2756 
2757 	case DIOCOSFPFLUSH:
2758 		crit_enter();
2759 		pf_osfp_flush();
2760 		crit_exit();
2761 		break;
2762 
2763 	case DIOCIGETIFACES: {
2764 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2765 
2766 		if (io->pfiio_esize != sizeof(struct pfi_if)) {
2767 			error = ENODEV;
2768 			break;
2769 		}
2770 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2771 		    &io->pfiio_size, io->pfiio_flags);
2772 		break;
2773 	}
2774 
2775 	case DIOCICLRISTATS: {
2776 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2777 
2778 		error = pfi_clr_istats(io->pfiio_name, &io->pfiio_nzero,
2779 		    io->pfiio_flags);
2780 		break;
2781 	}
2782 
2783 	default:
2784 		error = ENODEV;
2785 		break;
2786 	}
2787 fail:
2788 	return (error);
2789 }
2790 
2791 /*
2792  * XXX - Check for version missmatch!!!
2793  */
2794 static void
2795 pf_clear_states(void)
2796 {
2797 	struct pf_state		*state;
2798 
2799 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2800 		state->timeout = PFTM_PURGE;
2801 #if NPFSYNC
2802 		/* don't send out individual delete messages */
2803 		state->sync_flags = PFSTATE_NOSYNC;
2804 #endif
2805 	}
2806 	pf_purge_expired_states();
2807 	pf_status.states = 0;
2808 #if 0 /* NPFSYNC */
2809 /*
2810  * XXX This is called on module unload, we do not want to sync that over? */
2811  */
2812 	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
2813 #endif
2814 }
2815 
2816 static int
2817 pf_clear_tables(void)
2818 {
2819 	struct pfioc_table io;
2820 	int error;
2821 
2822 	bzero(&io, sizeof(io));
2823 
2824 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
2825 	    io.pfrio_flags);
2826 
2827 	return (error);
2828 }
2829 
2830 static void
2831 pf_clear_srcnodes(void)
2832 {
2833 	struct pf_src_node	*n;
2834 	struct pf_state		*state;
2835 
2836 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2837 		state->src_node = NULL;
2838 		state->nat_src_node = NULL;
2839 	}
2840 	RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2841 		n->expire = 1;
2842 		n->states = 0;
2843 	}
2844 	pf_purge_expired_src_nodes();
2845 	pf_status.src_nodes = 0;
2846 }
2847 /*
2848  * XXX - Check for version missmatch!!!
2849  */
2850 
2851 /*
2852  * Duplicate pfctl -Fa operation to get rid of as much as we can.
2853  */
2854 static int
2855 shutdown_pf(void)
2856 {
2857 	int error = 0;
2858 	u_int32_t t[5];
2859 	char nn = '\0';
2860 
2861 	callout_stop(&pf_expire_to);
2862 
2863 	pf_status.running = 0;
2864 	do {
2865 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn,
2866 		    &nn)) != 0) {
2867 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
2868 			break;
2869 		}
2870 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn,
2871 		    &nn)) != 0) {
2872 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
2873 			break;		/* XXX: rollback? */
2874 		}
2875 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn, &nn))
2876 		    != 0) {
2877 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
2878 			break;		/* XXX: rollback? */
2879 		}
2880 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn, &nn))
2881 		    != 0) {
2882 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
2883 			break;		/* XXX: rollback? */
2884 		}
2885 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn, &nn))
2886 		    != 0) {
2887 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
2888 			break;		/* XXX: rollback? */
2889 		}
2890 
2891 		/* XXX: these should always succeed here */
2892 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn, &nn);
2893 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn, &nn);
2894 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn, &nn);
2895 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn, &nn);
2896 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn, &nn);
2897 
2898 		if ((error = pf_clear_tables()) != 0)
2899 			break;
2900 
2901 #ifdef ALTQ
2902 		if ((error = pf_begin_altq(&t[0])) != 0) {
2903 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
2904 			break;
2905 		}
2906 		pf_commit_altq(t[0]);
2907 #endif
2908 
2909 		pf_clear_states();
2910 
2911 		pf_clear_srcnodes();
2912 
2913 		/* status does not use malloced mem so no need to cleanup */
2914 		/* fingerprints and interfaces have their own cleanup code */
2915 	} while(0);
2916 
2917         return (error);
2918 }
2919 
2920 static int
2921 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
2922 {
2923 	/*
2924 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
2925 	 * for ip_len/ip_off. This is why we don't have to change byte order
2926 	 * like the FreeBSD-5 version does.
2927 	 */
2928 	int chk;
2929 
2930 	chk = pf_test(PF_IN, ifp, m);
2931 	if (chk && *m) {
2932 		m_freem(*m);
2933 		*m = NULL;
2934 	}
2935 	return chk;
2936 }
2937 
2938 static int
2939 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
2940 {
2941 	/*
2942 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
2943 	 * for ip_len/ip_off. This is why we don't have to change byte order
2944 	 * like the FreeBSD-5 version does.
2945 	 */
2946 	int chk;
2947 
2948 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
2949 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2950 		in_delayed_cksum(*m);
2951 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2952 	}
2953 	chk = pf_test(PF_OUT, ifp, m);
2954 	if (chk && *m) {
2955 		m_freem(*m);
2956 		*m = NULL;
2957 	}
2958 	return chk;
2959 }
2960 
2961 #ifdef INET6
2962 static int
2963 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
2964 {
2965 	/*
2966 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
2967 	 */
2968 	int chk;
2969 
2970 	chk = pf_test6(PF_IN, ifp, m);
2971 	if (chk && *m) {
2972 		m_freem(*m);
2973 		*m = NULL;
2974 	}
2975 	return chk;
2976 }
2977 
2978 static int
2979 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
2980 {
2981 	/*
2982 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
2983 	 */
2984 	int chk;
2985 
2986 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
2987 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2988 		in_delayed_cksum(*m);
2989 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2990 	}
2991 	chk = pf_test6(PF_OUT, ifp, m);
2992 	if (chk && *m) {
2993 		m_freem(*m);
2994 		*m = NULL;
2995 	}
2996 	return chk;
2997 }
2998 #endif /* INET6 */
2999 
3000 static int
3001 hook_pf(void)
3002 {
3003 	struct pfil_head *pfh_inet;
3004 #ifdef INET6
3005 	struct pfil_head *pfh_inet6;
3006 #endif
3007 
3008 	if (pf_pfil_hooked)
3009 		return (0);
3010 
3011 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3012 	if (pfh_inet == NULL)
3013 		return (ENODEV);
3014 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3015 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3016 #ifdef INET6
3017 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3018 	if (pfh_inet6 == NULL) {
3019 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3020 		    pfh_inet);
3021 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3022 		    pfh_inet);
3023 		return (ENODEV);
3024 	}
3025 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3026 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3027 #endif
3028 
3029 	pf_pfil_hooked = 1;
3030 	return (0);
3031 }
3032 
3033 static int
3034 dehook_pf(void)
3035 {
3036 	struct pfil_head *pfh_inet;
3037 #ifdef INET6
3038 	struct pfil_head *pfh_inet6;
3039 #endif
3040 
3041 	if (pf_pfil_hooked == 0)
3042 		return (0);
3043 
3044 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3045 	if (pfh_inet == NULL)
3046 		return (ENODEV);
3047 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3048 	    pfh_inet);
3049 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3050 	    pfh_inet);
3051 #ifdef INET6
3052 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3053 	if (pfh_inet6 == NULL)
3054 		return (ENODEV);
3055 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3056 	    pfh_inet6);
3057 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3058 	    pfh_inet6);
3059 #endif
3060 
3061 	pf_pfil_hooked = 0;
3062 	return (0);
3063 }
3064 
3065 static int
3066 pf_load(void)
3067 {
3068 	int error;
3069 
3070 	init_zone_var();
3071 	error = dev_ops_add(&pf_ops, 0, 0);
3072 	if (error)
3073 		return (error);
3074 	pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3075 	error = pfattach();
3076 	if (error) {
3077 		dev_ops_remove(&pf_ops, 0, 0);
3078 		return (error);
3079 	}
3080 	return (0);
3081 }
3082 
3083 static int
3084 pf_unload(void)
3085 {
3086 	int error;
3087 
3088 	pf_status.running = 0;
3089 	error = dehook_pf();
3090 	if (error) {
3091 		/*
3092 		 * Should not happen!
3093 		 * XXX Due to error code ESRCH, kldunload will show
3094 		 * a message like 'No such process'.
3095 		 */
3096 		kprintf("pfil unregistration fail\n");
3097 		return error;
3098 	}
3099 	shutdown_pf();
3100 	pfi_cleanup();
3101 	pf_osfp_flush();
3102 	pf_osfp_cleanup();
3103 	cleanup_pf_zone();
3104 	dev_ops_remove(&pf_ops, 0, 0);
3105 	return 0;
3106 }
3107 
3108 static int
3109 pf_modevent(module_t mod, int type, void *data)
3110 {
3111 	int error = 0;
3112 
3113 	switch(type) {
3114 	case MOD_LOAD:
3115 		error = pf_load();
3116 		break;
3117 
3118 	case MOD_UNLOAD:
3119 		error = pf_unload();
3120 		break;
3121 	default:
3122 		error = EINVAL;
3123 		break;
3124 	}
3125 	return error;
3126 }
3127 
3128 static moduledata_t pf_mod = {
3129 	"pf",
3130 	pf_modevent,
3131 	0
3132 };
3133 
3134 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3135 MODULE_VERSION(pf, PF_MODVER);
3136