xref: /dflybsd-src/sys/net/pf/pf_ioctl.c (revision 67bf99c4e3c62e257027c8f0d3b312f44cfe622f)
1 /*	$OpenBSD: pf_ioctl.c,v 1.182 2007/06/24 11:17:13 mcbride Exp $ */
2 
3 /*
4  * Copyright (c) 2010 The DragonFly Project.  All rights reserved.
5  *
6  * Copyright (c) 2001 Daniel Hartmeier
7  * Copyright (c) 2002,2003 Henning Brauer
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  *
14  *    - Redistributions of source code must retain the above copyright
15  *      notice, this list of conditions and the following disclaimer.
16  *    - Redistributions in binary form must reproduce the above
17  *      copyright notice, this list of conditions and the following
18  *      disclaimer in the documentation and/or other materials provided
19  *      with the distribution.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
24  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
25  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
27  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
28  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
29  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
31  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
32  * POSSIBILITY OF SUCH DAMAGE.
33  *
34  * Effort sponsored in part by the Defense Advanced Research Projects
35  * Agency (DARPA) and Air Force Research Laboratory, Air Force
36  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
37  *
38  */
39 
40 #include "opt_inet.h"
41 #include "opt_inet6.h"
42 #include "use_pfsync.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/conf.h>
47 #include <sys/device.h>
48 #include <sys/mbuf.h>
49 #include <sys/filio.h>
50 #include <sys/fcntl.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/kernel.h>
54 #include <sys/kthread.h>
55 #include <sys/time.h>
56 #include <sys/proc.h>
57 #include <sys/malloc.h>
58 #include <sys/module.h>
59 #include <vm/vm_zone.h>
60 #include <sys/lock.h>
61 
62 #include <sys/thread2.h>
63 #include <sys/mplock2.h>
64 
65 #include <net/if.h>
66 #include <net/if_types.h>
67 #include <net/route.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_var.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/ip_icmp.h>
75 
76 #include <net/pf/pfvar.h>
77 #include <sys/md5.h>
78 #include <net/pf/pfvar.h>
79 
80 #if NPFSYNC > 0
81 #include <net/pf/if_pfsync.h>
82 #endif /* NPFSYNC > 0 */
83 
84 #if NPFLOG > 0
85 #include <net/if_pflog.h>
86 #endif /* NPFLOG > 0 */
87 
88 #ifdef INET6
89 #include <netinet/ip6.h>
90 #include <netinet/in_pcb.h>
91 #endif /* INET6 */
92 
93 #ifdef ALTQ
94 #include <net/altq/altq.h>
95 #endif
96 
97 #include <machine/limits.h>
98 #include <net/pfil.h>
99 #include <sys/mutex.h>
100 
101 u_int rt_numfibs = RT_NUMFIBS;
102 
103 void			 init_zone_var(void);
104 void			 cleanup_pf_zone(void);
105 int			 pfattach(void);
106 struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
107 			    u_int8_t, u_int8_t, u_int8_t);
108 
109 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
110 void			 pf_empty_pool(struct pf_palist *);
111 #ifdef ALTQ
112 int			 pf_begin_altq(u_int32_t *);
113 int			 pf_rollback_altq(u_int32_t);
114 int			 pf_commit_altq(u_int32_t);
115 int			 pf_enable_altq(struct pf_altq *);
116 int			 pf_disable_altq(struct pf_altq *);
117 #endif /* ALTQ */
118 int			 pf_begin_rules(u_int32_t *, int, const char *);
119 int			 pf_rollback_rules(u_int32_t, int, char *);
120 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
121 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
122 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
123 int			 pf_commit_rules(u_int32_t, int, char *);
124 void			 pf_state_export(struct pfsync_state *,
125 			    struct pf_state_key *, struct pf_state *);
126 void			 pf_state_import(struct pfsync_state *,
127 			    struct pf_state_key *, struct pf_state *);
128 
129 struct pf_rule		 pf_default_rule;
130 struct lock		 pf_consistency_lock;
131 #ifdef ALTQ
132 static int		 pf_altq_running;
133 #endif
134 
135 #define	TAGID_MAX	 50000
136 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
137 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
138 
139 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
140 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
141 #endif
142 u_int16_t		 tagname2tag(struct pf_tags *, char *);
143 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
144 void			 tag_unref(struct pf_tags *, u_int16_t);
145 int			 pf_rtlabel_add(struct pf_addr_wrap *);
146 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
147 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
148 
149 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
150 
151 static cdev_t	pf_dev;
152 
153 /*
154  * XXX - These are new and need to be checked when moveing to a new version
155  */
156 static void		 pf_clear_states(void);
157 static int		 pf_clear_tables(void);
158 static void		 pf_clear_srcnodes(void);
159 /*
160  * XXX - These are new and need to be checked when moveing to a new version
161  */
162 
163 /*
164  * Wrapper functions for pfil(9) hooks
165  */
166 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
167 		int dir);
168 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
169 		int dir);
170 #ifdef INET6
171 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
172 		int dir);
173 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
174 		int dir);
175 #endif
176 
177 static int 		 hook_pf(void);
178 static int 		 dehook_pf(void);
179 static int 		 shutdown_pf(void);
180 static int 		 pf_load(void);
181 static int 		 pf_unload(void);
182 
183 d_open_t	pfopen;
184 d_close_t	pfclose;
185 d_ioctl_t	pfioctl;
186 
187 static struct dev_ops pf_ops = {	    /* XXX convert to port model */
188 	{ PF_NAME, 73, 0 },
189 	.d_open =	pfopen,
190 	.d_close =	pfclose,
191 	.d_ioctl =	pfioctl
192 };
193 
194 static volatile int pf_pfil_hooked = 0;
195 int pf_end_threads = 0;
196 struct lock pf_mod_lck;
197 
198 int debug_pfugidhack = 0;
199 SYSCTL_INT(_debug, OID_AUTO, pfugidhack, CTLFLAG_RW, &debug_pfugidhack, 0,
200 	"Enable/disable pf user/group rules mpsafe hack");
201 
202 void
203 init_zone_var(void)
204 {
205 	pf_src_tree_pl = pf_rule_pl = NULL;
206 	pf_state_pl = pf_altq_pl = pf_pooladdr_pl = NULL;
207 	pf_frent_pl = pf_frag_pl = pf_cache_pl = pf_cent_pl = NULL;
208 	pf_state_scrub_pl = NULL;
209 	pfr_ktable_pl = pfr_kentry_pl = NULL;
210 }
211 
212 void
213 cleanup_pf_zone(void)
214 {
215 	ZONE_DESTROY(pf_src_tree_pl);
216 	ZONE_DESTROY(pf_rule_pl);
217 	ZONE_DESTROY(pf_state_pl);
218 	ZONE_DESTROY(pf_altq_pl);
219 	ZONE_DESTROY(pf_pooladdr_pl);
220 	ZONE_DESTROY(pf_frent_pl);
221 	ZONE_DESTROY(pf_frag_pl);
222 	ZONE_DESTROY(pf_cache_pl);
223 	ZONE_DESTROY(pf_cent_pl);
224 	ZONE_DESTROY(pfr_ktable_pl);
225 	ZONE_DESTROY(pfr_kentry_pl);
226 	ZONE_DESTROY(pfr_kentry_pl2);
227 	ZONE_DESTROY(pf_state_scrub_pl);
228 	ZONE_DESTROY(pfi_addr_pl);
229 }
230 
231 int
232 pfattach(void)
233 {
234 	u_int32_t *my_timeout = pf_default_rule.timeout;
235 	int error = 1;
236 
237 	do {
238 		ZONE_CREATE(pf_src_tree_pl,struct pf_src_node, "pfsrctrpl");
239 		ZONE_CREATE(pf_rule_pl,    struct pf_rule, "pfrulepl");
240 		ZONE_CREATE(pf_state_pl,   struct pf_state, "pfstatepl");
241 		ZONE_CREATE(pf_state_key_pl, struct pf_state_key, "pfstatekeypl");
242 		ZONE_CREATE(pf_altq_pl,    struct pf_altq, "pfaltqpl");
243 		ZONE_CREATE(pf_pooladdr_pl,struct pf_pooladdr, "pfpooladdrpl");
244 		ZONE_CREATE(pfr_ktable_pl, struct pfr_ktable, "pfrktable");
245 		ZONE_CREATE(pfr_kentry_pl, struct pfr_kentry, "pfrkentry");
246 		ZONE_CREATE(pfr_kentry_pl2, struct pfr_kentry, "pfrkentry2");
247 		ZONE_CREATE(pf_frent_pl,   struct pf_frent, "pffrent");
248 		ZONE_CREATE(pf_frag_pl,    struct pf_fragment, "pffrag");
249 		ZONE_CREATE(pf_cache_pl,   struct pf_fragment, "pffrcache");
250 		ZONE_CREATE(pf_cent_pl,    struct pf_frcache, "pffrcent");
251 		ZONE_CREATE(pf_state_scrub_pl, struct pf_state_scrub,
252 		    "pfstatescrub");
253 		ZONE_CREATE(pfi_addr_pl,   struct pfi_dynaddr, "pfiaddrpl");
254 		error = 0;
255 	} while(0);
256 	if (error) {
257 		cleanup_pf_zone();
258 		return (error);
259 	}
260 	pfr_initialize();
261 	pfi_initialize();
262 	error = pf_osfp_initialize();
263 	if (error) {
264 		cleanup_pf_zone();
265 		pf_osfp_cleanup();
266 		return (error);
267 	}
268 
269 	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
270 	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
271 	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
272 	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
273 	/* XXX uma_zone_set_max(pf_pool_limits[PF_LIMIT_STATES].pp,
274 		pf_pool_limits[PF_LIMIT_STATES].limit);
275 	*/
276 	if (ctob(physmem) <= 100*1024*1024)
277 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
278 		    PFR_KENTRY_HIWAT_SMALL;
279 	RB_INIT(&tree_src_tracking);
280 	RB_INIT(&pf_anchors);
281 	pf_init_ruleset(&pf_main_ruleset);
282 	TAILQ_INIT(&pf_altqs[0]);
283 	TAILQ_INIT(&pf_altqs[1]);
284 	TAILQ_INIT(&pf_pabuf);
285 	pf_altqs_active = &pf_altqs[0];
286 	pf_altqs_inactive = &pf_altqs[1];
287 	TAILQ_INIT(&state_list);
288 
289 	/* default rule should never be garbage collected */
290 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
291 	pf_default_rule.action = PF_PASS;
292 	pf_default_rule.nr = (uint32_t)(-1);
293 	pf_default_rule.rtableid = -1;
294 
295 	/* initialize default timeouts */
296 	my_timeout[PFTM_TCP_FIRST_PACKET] = 120;	/* First TCP packet */
297 	my_timeout[PFTM_TCP_OPENING] = 30; 		/* No response yet */
298 	my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;	/* Established */
299 	my_timeout[PFTM_TCP_CLOSING] = 15 * 60;		/* Half closed */
300 	my_timeout[PFTM_TCP_FIN_WAIT] = 45;		/* Got both FINs */
301 	my_timeout[PFTM_TCP_CLOSED] = 90;		/* Got a RST */
302 	my_timeout[PFTM_UDP_FIRST_PACKET] = 60;		/* First UDP packet */
303 	my_timeout[PFTM_UDP_SINGLE] = 30;		/* Unidirectional */
304 	my_timeout[PFTM_UDP_MULTIPLE] = 60;		/* Bidirectional */
305 	my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;	/* First ICMP packet */
306 	my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;		/* Got error response */
307 	my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;	/* First packet */
308 	my_timeout[PFTM_OTHER_SINGLE] = 30;		/* Unidirectional */
309 	my_timeout[PFTM_OTHER_MULTIPLE] = 60;		/* Bidirectional */
310 	my_timeout[PFTM_FRAG] = 30;			/* Fragment expire */
311 	my_timeout[PFTM_INTERVAL] = 10;			/* Expire interval */
312 	my_timeout[PFTM_SRC_NODE] = 0;		/* Source Tracking */
313 	my_timeout[PFTM_TS_DIFF] = 30;		/* Allowed TS diff */
314 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
315 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
316 
317 	pf_normalize_init();
318 	bzero(&pf_status, sizeof(pf_status));
319 	pf_status.debug = PF_DEBUG_URGENT;
320 
321 	/* XXX do our best to avoid a conflict */
322 	pf_status.hostid = karc4random();
323 
324 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
325 		panic("pfpurge thread");
326 
327 	return (error);
328 }
329 
330 int
331 pfopen(struct dev_open_args *ap)
332 {
333 	lwkt_gettoken(&pf_token);
334 	cdev_t dev = ap->a_head.a_dev;
335 	if (minor(dev) >= 1) {
336 		lwkt_reltoken(&pf_token);
337 		return (ENXIO);
338 	}
339 	lwkt_reltoken(&pf_token);
340 	return (0);
341 }
342 
343 int
344 pfclose(struct dev_close_args *ap)
345 {
346 	lwkt_gettoken(&pf_token);
347 	cdev_t dev = ap->a_head.a_dev;
348 	if (minor(dev) >= 1) {
349 		lwkt_reltoken(&pf_token);
350 		return (ENXIO);
351 	}
352 	lwkt_reltoken(&pf_token);
353 	return (0);
354 }
355 
356 struct pf_pool *
357 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
358     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
359     u_int8_t check_ticket)
360 {
361 	struct pf_ruleset	*ruleset;
362 	struct pf_rule		*rule;
363 	int			 rs_num;
364 
365 	ruleset = pf_find_ruleset(anchor);
366 	if (ruleset == NULL)
367 		return (NULL);
368 	rs_num = pf_get_ruleset_number(rule_action);
369 	if (rs_num >= PF_RULESET_MAX)
370 		return (NULL);
371 	if (active) {
372 		if (check_ticket && ticket !=
373 		    ruleset->rules[rs_num].active.ticket)
374 			return (NULL);
375 		if (r_last)
376 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
377 			    pf_rulequeue);
378 		else
379 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
380 	} else {
381 		if (check_ticket && ticket !=
382 		    ruleset->rules[rs_num].inactive.ticket)
383 			return (NULL);
384 		if (r_last)
385 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
386 			    pf_rulequeue);
387 		else
388 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
389 	}
390 	if (!r_last) {
391 		while ((rule != NULL) && (rule->nr != rule_number))
392 			rule = TAILQ_NEXT(rule, entries);
393 	}
394 	if (rule == NULL)
395 		return (NULL);
396 
397 	return (&rule->rpool);
398 }
399 
400 void
401 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
402 {
403 	struct pf_pooladdr	*mv_pool_pa;
404 
405 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
406 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
407 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
408 	}
409 }
410 
411 void
412 pf_empty_pool(struct pf_palist *poola)
413 {
414 	struct pf_pooladdr	*empty_pool_pa;
415 
416 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
417 		pfi_dynaddr_remove(&empty_pool_pa->addr);
418 		pf_tbladdr_remove(&empty_pool_pa->addr);
419 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
420 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
421 		pool_put(&pf_pooladdr_pl, empty_pool_pa);
422 	}
423 }
424 
425 void
426 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
427 {
428 	if (rulequeue != NULL) {
429 		if (rule->states <= 0) {
430 			/*
431 			 * XXX - we need to remove the table *before* detaching
432 			 * the rule to make sure the table code does not delete
433 			 * the anchor under our feet.
434 			 */
435 			pf_tbladdr_remove(&rule->src.addr);
436 			pf_tbladdr_remove(&rule->dst.addr);
437 			if (rule->overload_tbl)
438 				pfr_detach_table(rule->overload_tbl);
439 		}
440 		TAILQ_REMOVE(rulequeue, rule, entries);
441 		rule->entries.tqe_prev = NULL;
442 		rule->nr = -1;
443 	}
444 
445 	if (rule->states > 0 || rule->src_nodes > 0 ||
446 	    rule->entries.tqe_prev != NULL)
447 		return;
448 	pf_tag_unref(rule->tag);
449 	pf_tag_unref(rule->match_tag);
450 #ifdef ALTQ
451 	if (rule->pqid != rule->qid)
452 		pf_qid_unref(rule->pqid);
453 	pf_qid_unref(rule->qid);
454 #endif
455 	pf_rtlabel_remove(&rule->src.addr);
456 	pf_rtlabel_remove(&rule->dst.addr);
457 	pfi_dynaddr_remove(&rule->src.addr);
458 	pfi_dynaddr_remove(&rule->dst.addr);
459 	if (rulequeue == NULL) {
460 		pf_tbladdr_remove(&rule->src.addr);
461 		pf_tbladdr_remove(&rule->dst.addr);
462 		if (rule->overload_tbl)
463 			pfr_detach_table(rule->overload_tbl);
464 	}
465 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
466 	pf_anchor_remove(rule);
467 	pf_empty_pool(&rule->rpool.list);
468 	pool_put(&pf_rule_pl, rule);
469 }
470 
471 u_int16_t
472 tagname2tag(struct pf_tags *head, char *tagname)
473 {
474 	struct pf_tagname	*tag, *p = NULL;
475 	u_int16_t		 new_tagid = 1;
476 
477 	TAILQ_FOREACH(tag, head, entries)
478 		if (strcmp(tagname, tag->name) == 0) {
479 			tag->ref++;
480 			return (tag->tag);
481 		}
482 
483 	/*
484 	 * to avoid fragmentation, we do a linear search from the beginning
485 	 * and take the first free slot we find. if there is none or the list
486 	 * is empty, append a new entry at the end.
487 	 */
488 
489 	/* new entry */
490 	if (!TAILQ_EMPTY(head))
491 		for (p = TAILQ_FIRST(head); p != NULL &&
492 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
493 			new_tagid = p->tag + 1;
494 
495 	if (new_tagid > TAGID_MAX)
496 		return (0);
497 
498 	/* allocate and fill new struct pf_tagname */
499 	tag = kmalloc(sizeof(struct pf_tagname), M_TEMP, M_WAITOK);
500 	if (tag == NULL)
501 		return (0);
502 	bzero(tag, sizeof(struct pf_tagname));
503 	strlcpy(tag->name, tagname, sizeof(tag->name));
504 	tag->tag = new_tagid;
505 	tag->ref++;
506 
507 	if (p != NULL)	/* insert new entry before p */
508 		TAILQ_INSERT_BEFORE(p, tag, entries);
509 	else	/* either list empty or no free slot in between */
510 		TAILQ_INSERT_TAIL(head, tag, entries);
511 
512 	return (tag->tag);
513 }
514 
515 void
516 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
517 {
518 	struct pf_tagname	*tag;
519 
520 	TAILQ_FOREACH(tag, head, entries)
521 		if (tag->tag == tagid) {
522 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
523 			return;
524 		}
525 }
526 
527 void
528 tag_unref(struct pf_tags *head, u_int16_t tag)
529 {
530 	struct pf_tagname	*p, *next;
531 
532 	if (tag == 0)
533 		return;
534 
535 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
536 		next = TAILQ_NEXT(p, entries);
537 		if (tag == p->tag) {
538 			if (--p->ref == 0) {
539 				TAILQ_REMOVE(head, p, entries);
540 				kfree(p, M_TEMP);
541 			}
542 			break;
543 		}
544 	}
545 }
546 
547 u_int16_t
548 pf_tagname2tag(char *tagname)
549 {
550 	return (tagname2tag(&pf_tags, tagname));
551 }
552 
553 void
554 pf_tag2tagname(u_int16_t tagid, char *p)
555 {
556 	tag2tagname(&pf_tags, tagid, p);
557 }
558 
559 void
560 pf_tag_ref(u_int16_t tag)
561 {
562 	struct pf_tagname *t;
563 
564 	TAILQ_FOREACH(t, &pf_tags, entries)
565 		if (t->tag == tag)
566 			break;
567 	if (t != NULL)
568 		t->ref++;
569 }
570 
571 void
572 pf_tag_unref(u_int16_t tag)
573 {
574 	tag_unref(&pf_tags, tag);
575 }
576 
577 int
578 pf_rtlabel_add(struct pf_addr_wrap *a)
579 {
580 	return (0);
581 }
582 
583 void
584 pf_rtlabel_remove(struct pf_addr_wrap *a)
585 {
586 }
587 
588 void
589 pf_rtlabel_copyout(struct pf_addr_wrap *a)
590 {
591 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
592 		strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
593 }
594 
595 #ifdef ALTQ
596 u_int32_t
597 pf_qname2qid(char *qname)
598 {
599 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
600 }
601 
602 void
603 pf_qid2qname(u_int32_t qid, char *p)
604 {
605 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
606 }
607 
608 void
609 pf_qid_unref(u_int32_t qid)
610 {
611 	tag_unref(&pf_qids, (u_int16_t)qid);
612 }
613 
614 int
615 pf_begin_altq(u_int32_t *ticket)
616 {
617 	struct pf_altq	*altq;
618 	int		 error = 0;
619 
620 	/* Purge the old altq list */
621 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
622 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
623 		if (altq->qname[0] == 0) {
624 			/* detach and destroy the discipline */
625 			error = altq_remove(altq);
626 		} else
627 			pf_qid_unref(altq->qid);
628 		pool_put(&pf_altq_pl, altq);
629 	}
630 	if (error)
631 		return (error);
632 	*ticket = ++ticket_altqs_inactive;
633 	altqs_inactive_open = 1;
634 	return (0);
635 }
636 
637 int
638 pf_rollback_altq(u_int32_t ticket)
639 {
640 	struct pf_altq	*altq;
641 	int		 error = 0;
642 
643 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
644 		return (0);
645 	/* Purge the old altq list */
646 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
647 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
648 		if (altq->qname[0] == 0) {
649 			/* detach and destroy the discipline */
650 			error = altq_remove(altq);
651 		} else
652 			pf_qid_unref(altq->qid);
653 		pool_put(&pf_altq_pl, altq);
654 	}
655 	altqs_inactive_open = 0;
656 	return (error);
657 }
658 
659 int
660 pf_commit_altq(u_int32_t ticket)
661 {
662 	struct pf_altqqueue	*old_altqs;
663 	struct pf_altq		*altq;
664 	int			 err, error = 0;
665 
666 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
667 		return (EBUSY);
668 
669 	/* swap altqs, keep the old. */
670 	crit_enter();
671 	old_altqs = pf_altqs_active;
672 	pf_altqs_active = pf_altqs_inactive;
673 	pf_altqs_inactive = old_altqs;
674 	ticket_altqs_active = ticket_altqs_inactive;
675 
676 	/* Attach new disciplines */
677 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
678 		if (altq->qname[0] == 0) {
679 			/* attach the discipline */
680 			error = altq_pfattach(altq);
681 			if (error) {
682 				crit_exit();
683 				return (error);
684 			}
685 		}
686 	}
687 
688 	/* Purge the old altq list */
689 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
690 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
691 		if (altq->qname[0] == 0) {
692 			/* detach and destroy the discipline */
693 			if (pf_altq_running)
694 				error = pf_disable_altq(altq);
695 			err = altq_pfdetach(altq);
696 			if (err != 0 && error == 0)
697 				error = err;
698 			err = altq_remove(altq);
699 			if (err != 0 && error == 0)
700 				error = err;
701 		} else
702 			pf_qid_unref(altq->qid);
703 		pool_put(&pf_altq_pl, altq);
704 	}
705 	crit_exit();
706 
707 	altqs_inactive_open = 0;
708 	return (error);
709 }
710 
711 int
712 pf_enable_altq(struct pf_altq *altq)
713 {
714 	struct ifnet		*ifp;
715 	struct tb_profile	 tb;
716 	int			 error = 0;
717 
718 	if ((ifp = ifunit(altq->ifname)) == NULL)
719 		return (EINVAL);
720 
721 	if (ifp->if_snd.altq_type != ALTQT_NONE)
722 		error = altq_enable(&ifp->if_snd);
723 
724 	/* set tokenbucket regulator */
725 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
726 		tb.rate = altq->ifbandwidth;
727 		tb.depth = altq->tbrsize;
728 		crit_enter();
729 		error = tbr_set(&ifp->if_snd, &tb);
730 		crit_exit();
731 	}
732 
733 	return (error);
734 }
735 
736 int
737 pf_disable_altq(struct pf_altq *altq)
738 {
739 	struct ifnet		*ifp;
740 	struct tb_profile	 tb;
741 	int			 error;
742 
743 	if ((ifp = ifunit(altq->ifname)) == NULL)
744 		return (EINVAL);
745 
746 	/*
747 	 * when the discipline is no longer referenced, it was overridden
748 	 * by a new one.  if so, just return.
749 	 */
750 	if (altq->altq_disc != ifp->if_snd.altq_disc)
751 		return (0);
752 
753 	error = altq_disable(&ifp->if_snd);
754 
755 	if (error == 0) {
756 		/* clear tokenbucket regulator */
757 		tb.rate = 0;
758 		crit_enter();
759 		error = tbr_set(&ifp->if_snd, &tb);
760 		crit_exit();
761 	}
762 
763 	return (error);
764 }
765 #endif /* ALTQ */
766 
767 int
768 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
769 {
770 	struct pf_ruleset	*rs;
771 	struct pf_rule		*rule;
772 
773 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
774 		return (EINVAL);
775 	rs = pf_find_or_create_ruleset(anchor);
776 	if (rs == NULL)
777 		return (EINVAL);
778 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
779 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
780 		rs->rules[rs_num].inactive.rcount--;
781 	}
782 	*ticket = ++rs->rules[rs_num].inactive.ticket;
783 	rs->rules[rs_num].inactive.open = 1;
784 	return (0);
785 }
786 
787 int
788 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
789 {
790 	struct pf_ruleset	*rs;
791 	struct pf_rule		*rule;
792 
793 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
794 		return (EINVAL);
795 	rs = pf_find_ruleset(anchor);
796 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
797 	    rs->rules[rs_num].inactive.ticket != ticket)
798 		return (0);
799 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
800 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
801 		rs->rules[rs_num].inactive.rcount--;
802 	}
803 	rs->rules[rs_num].inactive.open = 0;
804 	return (0);
805 }
806 
807 #define PF_MD5_UPD(st, elm)						\
808 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
809 
810 #define PF_MD5_UPD_STR(st, elm)						\
811 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
812 
813 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
814 		(stor) = htonl((st)->elm);				\
815 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
816 } while (0)
817 
818 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
819 		(stor) = htons((st)->elm);				\
820 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
821 } while (0)
822 
823 void
824 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
825 {
826 	PF_MD5_UPD(pfr, addr.type);
827 	switch (pfr->addr.type) {
828 		case PF_ADDR_DYNIFTL:
829 			PF_MD5_UPD(pfr, addr.v.ifname);
830 			PF_MD5_UPD(pfr, addr.iflags);
831 			break;
832 		case PF_ADDR_TABLE:
833 			PF_MD5_UPD(pfr, addr.v.tblname);
834 			break;
835 		case PF_ADDR_ADDRMASK:
836 			/* XXX ignore af? */
837 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
838 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
839 			break;
840 		case PF_ADDR_RTLABEL:
841 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
842 			break;
843 	}
844 
845 	PF_MD5_UPD(pfr, port[0]);
846 	PF_MD5_UPD(pfr, port[1]);
847 	PF_MD5_UPD(pfr, neg);
848 	PF_MD5_UPD(pfr, port_op);
849 }
850 
851 void
852 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
853 {
854 	u_int16_t x;
855 	u_int32_t y;
856 
857 	pf_hash_rule_addr(ctx, &rule->src);
858 	pf_hash_rule_addr(ctx, &rule->dst);
859 	PF_MD5_UPD_STR(rule, label);
860 	PF_MD5_UPD_STR(rule, ifname);
861 	PF_MD5_UPD_STR(rule, match_tagname);
862 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
863 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
864 	PF_MD5_UPD_HTONL(rule, prob, y);
865 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
866 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
867 	PF_MD5_UPD(rule, uid.op);
868 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
869 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
870 	PF_MD5_UPD(rule, gid.op);
871 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
872 	PF_MD5_UPD(rule, action);
873 	PF_MD5_UPD(rule, direction);
874 	PF_MD5_UPD(rule, af);
875 	PF_MD5_UPD(rule, quick);
876 	PF_MD5_UPD(rule, ifnot);
877 	PF_MD5_UPD(rule, match_tag_not);
878 	PF_MD5_UPD(rule, natpass);
879 	PF_MD5_UPD(rule, keep_state);
880 	PF_MD5_UPD(rule, proto);
881 	PF_MD5_UPD(rule, type);
882 	PF_MD5_UPD(rule, code);
883 	PF_MD5_UPD(rule, flags);
884 	PF_MD5_UPD(rule, flagset);
885 	PF_MD5_UPD(rule, allow_opts);
886 	PF_MD5_UPD(rule, rt);
887 	PF_MD5_UPD(rule, tos);
888 }
889 
890 int
891 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
892 {
893 	struct pf_ruleset	*rs;
894 	struct pf_rule		*rule, **old_array;
895 	struct pf_rulequeue	*old_rules;
896 	int			 error;
897 	u_int32_t		 old_rcount;
898 
899 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
900 		return (EINVAL);
901 	rs = pf_find_ruleset(anchor);
902 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
903 	    ticket != rs->rules[rs_num].inactive.ticket)
904 		return (EBUSY);
905 
906 	/* Calculate checksum for the main ruleset */
907 	if (rs == &pf_main_ruleset) {
908 		error = pf_setup_pfsync_matching(rs);
909 		if (error != 0)
910 			return (error);
911 	}
912 
913 	/* Swap rules, keep the old. */
914 	crit_enter();
915 	old_rules = rs->rules[rs_num].active.ptr;
916 	old_rcount = rs->rules[rs_num].active.rcount;
917 	old_array = rs->rules[rs_num].active.ptr_array;
918 
919 	rs->rules[rs_num].active.ptr =
920 	    rs->rules[rs_num].inactive.ptr;
921 	rs->rules[rs_num].active.ptr_array =
922 	    rs->rules[rs_num].inactive.ptr_array;
923 	rs->rules[rs_num].active.rcount =
924 	    rs->rules[rs_num].inactive.rcount;
925 	rs->rules[rs_num].inactive.ptr = old_rules;
926 	rs->rules[rs_num].inactive.ptr_array = old_array;
927 	rs->rules[rs_num].inactive.rcount = old_rcount;
928 
929 	rs->rules[rs_num].active.ticket =
930 	    rs->rules[rs_num].inactive.ticket;
931 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
932 
933 
934 	/* Purge the old rule list. */
935 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
936 		pf_rm_rule(old_rules, rule);
937 	if (rs->rules[rs_num].inactive.ptr_array)
938 		kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
939 	rs->rules[rs_num].inactive.ptr_array = NULL;
940 	rs->rules[rs_num].inactive.rcount = 0;
941 	rs->rules[rs_num].inactive.open = 0;
942 	pf_remove_if_empty_ruleset(rs);
943 	crit_exit();
944 	return (0);
945 }
946 
947 void
948 pf_state_export(struct pfsync_state *sp, struct pf_state_key *sk,
949    struct pf_state *s)
950 {
951 	int secs = time_second;
952 	bzero(sp, sizeof(struct pfsync_state));
953 
954 	/* copy from state key */
955 	sp->lan.addr = sk->lan.addr;
956 	sp->lan.port = sk->lan.port;
957 	sp->gwy.addr = sk->gwy.addr;
958 	sp->gwy.port = sk->gwy.port;
959 	sp->ext.addr = sk->ext.addr;
960 	sp->ext.port = sk->ext.port;
961 	sp->proto = sk->proto;
962 	sp->af = sk->af;
963 	sp->direction = sk->direction;
964 
965 	/* copy from state */
966 	memcpy(&sp->id, &s->id, sizeof(sp->id));
967 	sp->creatorid = s->creatorid;
968 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
969 	pf_state_peer_to_pfsync(&s->src, &sp->src);
970 	pf_state_peer_to_pfsync(&s->dst, &sp->dst);
971 
972 	sp->rule = s->rule.ptr->nr;
973 	sp->nat_rule = (s->nat_rule.ptr == NULL) ?  -1 : s->nat_rule.ptr->nr;
974 	sp->anchor = (s->anchor.ptr == NULL) ?  -1 : s->anchor.ptr->nr;
975 
976 	pf_state_counter_to_pfsync(s->bytes[0], sp->bytes[0]);
977 	pf_state_counter_to_pfsync(s->bytes[1], sp->bytes[1]);
978 	pf_state_counter_to_pfsync(s->packets[0], sp->packets[0]);
979 	pf_state_counter_to_pfsync(s->packets[1], sp->packets[1]);
980 	sp->creation = secs - s->creation;
981 	sp->expire = pf_state_expires(s);
982 	sp->log = s->log;
983 	sp->allow_opts = s->allow_opts;
984 	sp->timeout = s->timeout;
985 
986 	if (s->src_node)
987 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
988 	if (s->nat_src_node)
989 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
990 
991 	if (sp->expire > secs)
992 		sp->expire -= secs;
993 	else
994 		sp->expire = 0;
995 
996 }
997 
998 void
999 pf_state_import(struct pfsync_state *sp, struct pf_state_key *sk,
1000    struct pf_state *s)
1001 {
1002 	/* copy to state key */
1003 	sk->lan.addr = sp->lan.addr;
1004 	sk->lan.port = sp->lan.port;
1005 	sk->gwy.addr = sp->gwy.addr;
1006 	sk->gwy.port = sp->gwy.port;
1007 	sk->ext.addr = sp->ext.addr;
1008 	sk->ext.port = sp->ext.port;
1009 	sk->proto = sp->proto;
1010 	sk->af = sp->af;
1011 	sk->direction = sp->direction;
1012 
1013 	/* copy to state */
1014 	memcpy(&s->id, &sp->id, sizeof(sp->id));
1015 	s->creatorid = sp->creatorid;
1016 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
1017 	pf_state_peer_from_pfsync(&sp->src, &s->src);
1018 	pf_state_peer_from_pfsync(&sp->dst, &s->dst);
1019 
1020 	s->rule.ptr = &pf_default_rule;
1021 	s->nat_rule.ptr = NULL;
1022 	s->anchor.ptr = NULL;
1023 	s->rt_kif = NULL;
1024 	s->creation = time_second;
1025 	s->pfsync_time = 0;
1026 	s->packets[0] = s->packets[1] = 0;
1027 	s->bytes[0] = s->bytes[1] = 0;
1028 }
1029 
1030 int
1031 pf_setup_pfsync_matching(struct pf_ruleset *rs)
1032 {
1033 	MD5_CTX			 ctx;
1034 	struct pf_rule		*rule;
1035 	int			 rs_cnt;
1036 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1037 
1038 	MD5Init(&ctx);
1039 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1040 		/* XXX PF_RULESET_SCRUB as well? */
1041 		if (rs_cnt == PF_RULESET_SCRUB)
1042 			continue;
1043 
1044 		if (rs->rules[rs_cnt].inactive.ptr_array)
1045 			kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1046 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1047 
1048 		if (rs->rules[rs_cnt].inactive.rcount) {
1049 			rs->rules[rs_cnt].inactive.ptr_array =
1050 			    kmalloc(sizeof(caddr_t) *
1051 				    rs->rules[rs_cnt].inactive.rcount,
1052 				    M_TEMP, M_WAITOK);
1053 
1054 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1055 				return (ENOMEM);
1056 		}
1057 
1058 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1059 		    entries) {
1060 			pf_hash_rule(&ctx, rule);
1061 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1062 		}
1063 	}
1064 
1065 	MD5Final(digest, &ctx);
1066 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
1067 	return (0);
1068 }
1069 
1070 int
1071 pfioctl(struct dev_ioctl_args *ap)
1072 {
1073 	u_long cmd = ap->a_cmd;
1074 	caddr_t addr = ap->a_data;
1075 	struct pf_pooladdr	*pa = NULL;
1076 	struct pf_pool		*pool = NULL;
1077 	int			 error = 0;
1078 
1079 	lwkt_gettoken(&pf_token);
1080 
1081 	/* XXX keep in sync with switch() below */
1082 	if (securelevel > 1)
1083 		switch (cmd) {
1084 		case DIOCGETRULES:
1085 		case DIOCGETRULE:
1086 		case DIOCGETADDRS:
1087 		case DIOCGETADDR:
1088 		case DIOCGETSTATE:
1089 		case DIOCSETSTATUSIF:
1090 		case DIOCGETSTATUS:
1091 		case DIOCCLRSTATUS:
1092 		case DIOCNATLOOK:
1093 		case DIOCSETDEBUG:
1094 		case DIOCGETSTATES:
1095 		case DIOCGETTIMEOUT:
1096 		case DIOCCLRRULECTRS:
1097 		case DIOCGETLIMIT:
1098 		case DIOCGETALTQS:
1099 		case DIOCGETALTQ:
1100 		case DIOCGETQSTATS:
1101 		case DIOCGETRULESETS:
1102 		case DIOCGETRULESET:
1103 		case DIOCRGETTABLES:
1104 		case DIOCRGETTSTATS:
1105 		case DIOCRCLRTSTATS:
1106 		case DIOCRCLRADDRS:
1107 		case DIOCRADDADDRS:
1108 		case DIOCRDELADDRS:
1109 		case DIOCRSETADDRS:
1110 		case DIOCRGETADDRS:
1111 		case DIOCRGETASTATS:
1112 		case DIOCRCLRASTATS:
1113 		case DIOCRTSTADDRS:
1114 		case DIOCOSFPGET:
1115 		case DIOCGETSRCNODES:
1116 		case DIOCCLRSRCNODES:
1117 		case DIOCIGETIFACES:
1118 		case DIOCSETIFFLAG:
1119 		case DIOCCLRIFFLAG:
1120 		case DIOCGIFSPEED:
1121 			break;
1122 		case DIOCRCLRTABLES:
1123 		case DIOCRADDTABLES:
1124 		case DIOCRDELTABLES:
1125 		case DIOCRSETTFLAGS:
1126 			if (((struct pfioc_table *)addr)->pfrio_flags &
1127 			    PFR_FLAG_DUMMY)
1128 				break; /* dummy operation ok */
1129 			lwkt_reltoken(&pf_token);
1130 			return (EPERM);
1131 		default:
1132 			lwkt_reltoken(&pf_token);
1133 			return (EPERM);
1134 		}
1135 
1136 	if (!(ap->a_fflag & FWRITE))
1137 		switch (cmd) {
1138 		case DIOCGETRULES:
1139 		case DIOCGETADDRS:
1140 		case DIOCGETADDR:
1141 		case DIOCGETSTATE:
1142 		case DIOCGETSTATUS:
1143 		case DIOCGETSTATES:
1144 		case DIOCGETTIMEOUT:
1145 		case DIOCGETLIMIT:
1146 		case DIOCGETALTQS:
1147 		case DIOCGETALTQ:
1148 		case DIOCGETQSTATS:
1149 		case DIOCGETRULESETS:
1150 		case DIOCGETRULESET:
1151 		case DIOCNATLOOK:
1152 		case DIOCRGETTABLES:
1153 		case DIOCRGETTSTATS:
1154 		case DIOCRGETADDRS:
1155 		case DIOCRGETASTATS:
1156 		case DIOCRTSTADDRS:
1157 		case DIOCOSFPGET:
1158 		case DIOCGETSRCNODES:
1159 		case DIOCIGETIFACES:
1160 		case DIOCGIFSPEED:
1161 			break;
1162 		case DIOCRCLRTABLES:
1163 		case DIOCRADDTABLES:
1164 		case DIOCRDELTABLES:
1165 		case DIOCRCLRTSTATS:
1166 		case DIOCRCLRADDRS:
1167 		case DIOCRADDADDRS:
1168 		case DIOCRDELADDRS:
1169 		case DIOCRSETADDRS:
1170 		case DIOCRSETTFLAGS:
1171 			if (((struct pfioc_table *)addr)->pfrio_flags &
1172 			    PFR_FLAG_DUMMY)
1173 				break; /* dummy operation ok */
1174 			lwkt_reltoken(&pf_token);
1175 			return (EACCES);
1176 		case DIOCGETRULE:
1177 			if (((struct pfioc_rule *)addr)->action == PF_GET_CLR_CNTR) {
1178 				lwkt_reltoken(&pf_token);
1179 				return (EACCES);
1180 			}
1181 			break;
1182 		default:
1183 			lwkt_reltoken(&pf_token);
1184 			return (EACCES);
1185 		}
1186 
1187 	switch (cmd) {
1188 
1189 	case DIOCSTART:
1190 		if (pf_status.running)
1191 			error = EEXIST;
1192 		else {
1193 			error = hook_pf();
1194 			if (error) {
1195 				DPFPRINTF(PF_DEBUG_MISC,
1196 				    ("pf: pfil registration fail\n"));
1197 				break;
1198 			}
1199 			pf_status.running = 1;
1200 			pf_status.since = time_second;
1201 			if (pf_status.stateid == 0) {
1202 				pf_status.stateid = time_second;
1203 				pf_status.stateid = pf_status.stateid << 32;
1204 			}
1205 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1206 		}
1207 		break;
1208 
1209 	case DIOCSTOP:
1210 		if (!pf_status.running)
1211 			error = ENOENT;
1212 		else {
1213 			pf_status.running = 0;
1214 			error = dehook_pf();
1215 			if (error) {
1216 				pf_status.running = 1;
1217 				DPFPRINTF(PF_DEBUG_MISC,
1218 					("pf: pfil unregistration failed\n"));
1219 			}
1220 			pf_status.since = time_second;
1221 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1222 		}
1223 		break;
1224 
1225 	case DIOCADDRULE: {
1226 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1227 		struct pf_ruleset	*ruleset;
1228 		struct pf_rule		*rule, *tail;
1229 		struct pf_pooladdr	*pa;
1230 		int			 rs_num;
1231 
1232 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1233 		ruleset = pf_find_ruleset(pr->anchor);
1234 		if (ruleset == NULL) {
1235 			error = EINVAL;
1236 			break;
1237 		}
1238 		rs_num = pf_get_ruleset_number(pr->rule.action);
1239 		if (rs_num >= PF_RULESET_MAX) {
1240 			error = EINVAL;
1241 			break;
1242 		}
1243 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1244 			error = EINVAL;
1245 			break;
1246 		}
1247 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1248 			error = EBUSY;
1249 			break;
1250 		}
1251 		if (pr->pool_ticket != ticket_pabuf) {
1252 			error = EBUSY;
1253 			break;
1254 		}
1255 		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1256 		if (rule == NULL) {
1257 			error = ENOMEM;
1258 			break;
1259 		}
1260 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1261 		rule->cuid = ap->a_cred->cr_ruid;
1262 		rule->cpid = (int)NULL;
1263 		rule->anchor = NULL;
1264 		rule->kif = NULL;
1265 		TAILQ_INIT(&rule->rpool.list);
1266 		/* initialize refcounting */
1267 		rule->states = 0;
1268 		rule->src_nodes = 0;
1269 		rule->entries.tqe_prev = NULL;
1270 #ifndef INET
1271 		if (rule->af == AF_INET) {
1272 			pool_put(&pf_rule_pl, rule);
1273 			error = EAFNOSUPPORT;
1274 			break;
1275 		}
1276 #endif /* INET */
1277 #ifndef INET6
1278 		if (rule->af == AF_INET6) {
1279 			pool_put(&pf_rule_pl, rule);
1280 			error = EAFNOSUPPORT;
1281 			break;
1282 		}
1283 #endif /* INET6 */
1284 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1285 		    pf_rulequeue);
1286 		if (tail)
1287 			rule->nr = tail->nr + 1;
1288 		else
1289 			rule->nr = 0;
1290 		if (rule->ifname[0]) {
1291 			rule->kif = pfi_kif_get(rule->ifname);
1292 			if (rule->kif == NULL) {
1293 				pool_put(&pf_rule_pl, rule);
1294 				error = EINVAL;
1295 				break;
1296 			}
1297 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1298 		}
1299 
1300 		if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1301 			error = EBUSY;
1302 
1303 #ifdef ALTQ
1304 		/* set queue IDs */
1305 		if (rule->qname[0] != 0) {
1306 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1307 				error = EBUSY;
1308 			else if (rule->pqname[0] != 0) {
1309 				if ((rule->pqid =
1310 				    pf_qname2qid(rule->pqname)) == 0)
1311 					error = EBUSY;
1312 			} else
1313 				rule->pqid = rule->qid;
1314 		}
1315 #endif
1316 		if (rule->tagname[0])
1317 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1318 				error = EBUSY;
1319 		if (rule->match_tagname[0])
1320 			if ((rule->match_tag =
1321 			    pf_tagname2tag(rule->match_tagname)) == 0)
1322 				error = EBUSY;
1323 		if (rule->rt && !rule->direction)
1324 			error = EINVAL;
1325 #if NPFLOG > 0
1326 		if (!rule->log)
1327 			rule->logif = 0;
1328 		if (rule->logif >= PFLOGIFS_MAX)
1329 			error = EINVAL;
1330 #endif
1331 		if (pf_rtlabel_add(&rule->src.addr) ||
1332 		    pf_rtlabel_add(&rule->dst.addr))
1333 			error = EBUSY;
1334 		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1335 			error = EINVAL;
1336 		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1337 			error = EINVAL;
1338 		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1339 			error = EINVAL;
1340 		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1341 			error = EINVAL;
1342 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1343 			error = EINVAL;
1344 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1345 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1346 				error = EINVAL;
1347 
1348 		if (rule->overload_tblname[0]) {
1349 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1350 			    rule->overload_tblname)) == NULL)
1351 				error = EINVAL;
1352 			else
1353 				rule->overload_tbl->pfrkt_flags |=
1354 				    PFR_TFLAG_ACTIVE;
1355 		}
1356 
1357 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1358 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1359 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1360 		    (rule->rt > PF_FASTROUTE)) &&
1361 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1362 			error = EINVAL;
1363 
1364 		if (error) {
1365 			pf_rm_rule(NULL, rule);
1366 			break;
1367 		}
1368 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1369 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1370 		    rule->bytes[0] = rule->bytes[1] = 0;
1371 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1372 		    rule, entries);
1373 		ruleset->rules[rs_num].inactive.rcount++;
1374 		break;
1375 	}
1376 
1377 	case DIOCGETRULES: {
1378 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1379 		struct pf_ruleset	*ruleset;
1380 		struct pf_rule		*tail;
1381 		int			 rs_num;
1382 
1383 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1384 		ruleset = pf_find_ruleset(pr->anchor);
1385 		if (ruleset == NULL) {
1386 			error = EINVAL;
1387 			break;
1388 		}
1389 		rs_num = pf_get_ruleset_number(pr->rule.action);
1390 		if (rs_num >= PF_RULESET_MAX) {
1391 			error = EINVAL;
1392 			break;
1393 		}
1394 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1395 		    pf_rulequeue);
1396 		if (tail)
1397 			pr->nr = tail->nr + 1;
1398 		else
1399 			pr->nr = 0;
1400 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1401 		break;
1402 	}
1403 
1404 	case DIOCGETRULE: {
1405 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1406 		struct pf_ruleset	*ruleset;
1407 		struct pf_rule		*rule;
1408 		int			 rs_num, i;
1409 
1410 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1411 		ruleset = pf_find_ruleset(pr->anchor);
1412 		if (ruleset == NULL) {
1413 			error = EINVAL;
1414 			break;
1415 		}
1416 		rs_num = pf_get_ruleset_number(pr->rule.action);
1417 		if (rs_num >= PF_RULESET_MAX) {
1418 			error = EINVAL;
1419 			break;
1420 		}
1421 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1422 			error = EBUSY;
1423 			break;
1424 		}
1425 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1426 		while ((rule != NULL) && (rule->nr != pr->nr))
1427 			rule = TAILQ_NEXT(rule, entries);
1428 		if (rule == NULL) {
1429 			error = EBUSY;
1430 			break;
1431 		}
1432 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1433 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1434 			error = EBUSY;
1435 			break;
1436 		}
1437 		pfi_dynaddr_copyout(&pr->rule.src.addr);
1438 		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1439 		pf_tbladdr_copyout(&pr->rule.src.addr);
1440 		pf_tbladdr_copyout(&pr->rule.dst.addr);
1441 		pf_rtlabel_copyout(&pr->rule.src.addr);
1442 		pf_rtlabel_copyout(&pr->rule.dst.addr);
1443 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1444 			if (rule->skip[i].ptr == NULL)
1445 				pr->rule.skip[i].nr = (uint32_t)(-1);
1446 			else
1447 				pr->rule.skip[i].nr =
1448 				    rule->skip[i].ptr->nr;
1449 
1450 		if (pr->action == PF_GET_CLR_CNTR) {
1451 			rule->evaluations = 0;
1452 			rule->packets[0] = rule->packets[1] = 0;
1453 			rule->bytes[0] = rule->bytes[1] = 0;
1454 		}
1455 		break;
1456 	}
1457 
1458 	case DIOCCHANGERULE: {
1459 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1460 		struct pf_ruleset	*ruleset;
1461 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1462 		u_int32_t		 nr = 0;
1463 		int			 rs_num;
1464 
1465 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1466 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1467 		    pcr->pool_ticket != ticket_pabuf) {
1468 			error = EBUSY;
1469 			break;
1470 		}
1471 
1472 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1473 		    pcr->action > PF_CHANGE_GET_TICKET) {
1474 			error = EINVAL;
1475 			break;
1476 		}
1477 		ruleset = pf_find_ruleset(pcr->anchor);
1478 		if (ruleset == NULL) {
1479 			error = EINVAL;
1480 			break;
1481 		}
1482 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1483 		if (rs_num >= PF_RULESET_MAX) {
1484 			error = EINVAL;
1485 			break;
1486 		}
1487 
1488 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1489 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1490 			break;
1491 		} else {
1492 			if (pcr->ticket !=
1493 			    ruleset->rules[rs_num].active.ticket) {
1494 				error = EINVAL;
1495 				break;
1496 			}
1497 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1498 				error = EINVAL;
1499 				break;
1500 			}
1501 		}
1502 
1503 		if (pcr->action != PF_CHANGE_REMOVE) {
1504 			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1505 			if (newrule == NULL) {
1506 				error = ENOMEM;
1507 				break;
1508 			}
1509 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1510 			newrule->cuid = ap->a_cred->cr_ruid;
1511 			newrule->cpid = (int)NULL;
1512 			TAILQ_INIT(&newrule->rpool.list);
1513 			/* initialize refcounting */
1514 			newrule->states = 0;
1515 			newrule->entries.tqe_prev = NULL;
1516 #ifndef INET
1517 			if (newrule->af == AF_INET) {
1518 				pool_put(&pf_rule_pl, newrule);
1519 				error = EAFNOSUPPORT;
1520 				break;
1521 			}
1522 #endif /* INET */
1523 #ifndef INET6
1524 			if (newrule->af == AF_INET6) {
1525 				pool_put(&pf_rule_pl, newrule);
1526 				error = EAFNOSUPPORT;
1527 				break;
1528 			}
1529 #endif /* INET6 */
1530 			if (newrule->ifname[0]) {
1531 				newrule->kif = pfi_kif_get(newrule->ifname);
1532 				if (newrule->kif == NULL) {
1533 					pool_put(&pf_rule_pl, newrule);
1534 					error = EINVAL;
1535 					break;
1536 				}
1537 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1538 			} else
1539 				newrule->kif = NULL;
1540 
1541 			if (newrule->rtableid > 0 &&
1542 			    newrule->rtableid > rt_numfibs)
1543 				error = EBUSY;
1544 
1545 #ifdef ALTQ
1546 			/* set queue IDs */
1547 			if (newrule->qname[0] != 0) {
1548 				if ((newrule->qid =
1549 				    pf_qname2qid(newrule->qname)) == 0)
1550 					error = EBUSY;
1551 				else if (newrule->pqname[0] != 0) {
1552 					if ((newrule->pqid =
1553 					    pf_qname2qid(newrule->pqname)) == 0)
1554 						error = EBUSY;
1555 				} else
1556 					newrule->pqid = newrule->qid;
1557 			}
1558 #endif /* ALTQ */
1559 			if (newrule->tagname[0])
1560 				if ((newrule->tag =
1561 				    pf_tagname2tag(newrule->tagname)) == 0)
1562 					error = EBUSY;
1563 			if (newrule->match_tagname[0])
1564 				if ((newrule->match_tag = pf_tagname2tag(
1565 				    newrule->match_tagname)) == 0)
1566 					error = EBUSY;
1567 			if (newrule->rt && !newrule->direction)
1568 				error = EINVAL;
1569 #if NPFLOG > 0
1570 			if (!newrule->log)
1571 				newrule->logif = 0;
1572 			if (newrule->logif >= PFLOGIFS_MAX)
1573 				error = EINVAL;
1574 #endif
1575 			if (pf_rtlabel_add(&newrule->src.addr) ||
1576 			    pf_rtlabel_add(&newrule->dst.addr))
1577 				error = EBUSY;
1578 			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1579 				error = EINVAL;
1580 			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1581 				error = EINVAL;
1582 			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1583 				error = EINVAL;
1584 			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1585 				error = EINVAL;
1586 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1587 				error = EINVAL;
1588 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1589 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1590 					error = EINVAL;
1591 
1592 			if (newrule->overload_tblname[0]) {
1593 				if ((newrule->overload_tbl = pfr_attach_table(
1594 				    ruleset, newrule->overload_tblname)) ==
1595 				    NULL)
1596 					error = EINVAL;
1597 				else
1598 					newrule->overload_tbl->pfrkt_flags |=
1599 					    PFR_TFLAG_ACTIVE;
1600 			}
1601 
1602 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1603 			if (((((newrule->action == PF_NAT) ||
1604 			    (newrule->action == PF_RDR) ||
1605 			    (newrule->action == PF_BINAT) ||
1606 			    (newrule->rt > PF_FASTROUTE)) &&
1607 			    !newrule->anchor)) &&
1608 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1609 				error = EINVAL;
1610 
1611 			if (error) {
1612 				pf_rm_rule(NULL, newrule);
1613 				break;
1614 			}
1615 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1616 			newrule->evaluations = 0;
1617 			newrule->packets[0] = newrule->packets[1] = 0;
1618 			newrule->bytes[0] = newrule->bytes[1] = 0;
1619 		}
1620 		pf_empty_pool(&pf_pabuf);
1621 
1622 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1623 			oldrule = TAILQ_FIRST(
1624 			    ruleset->rules[rs_num].active.ptr);
1625 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1626 			oldrule = TAILQ_LAST(
1627 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1628 		else {
1629 			oldrule = TAILQ_FIRST(
1630 			    ruleset->rules[rs_num].active.ptr);
1631 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1632 				oldrule = TAILQ_NEXT(oldrule, entries);
1633 			if (oldrule == NULL) {
1634 				if (newrule != NULL)
1635 					pf_rm_rule(NULL, newrule);
1636 				error = EINVAL;
1637 				break;
1638 			}
1639 		}
1640 
1641 		if (pcr->action == PF_CHANGE_REMOVE) {
1642 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1643 			ruleset->rules[rs_num].active.rcount--;
1644 		} else {
1645 			if (oldrule == NULL)
1646 				TAILQ_INSERT_TAIL(
1647 				    ruleset->rules[rs_num].active.ptr,
1648 				    newrule, entries);
1649 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1650 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1651 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1652 			else
1653 				TAILQ_INSERT_AFTER(
1654 				    ruleset->rules[rs_num].active.ptr,
1655 				    oldrule, newrule, entries);
1656 			ruleset->rules[rs_num].active.rcount++;
1657 		}
1658 
1659 		nr = 0;
1660 		TAILQ_FOREACH(oldrule,
1661 		    ruleset->rules[rs_num].active.ptr, entries)
1662 			oldrule->nr = nr++;
1663 
1664 		ruleset->rules[rs_num].active.ticket++;
1665 
1666 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1667 		pf_remove_if_empty_ruleset(ruleset);
1668 
1669 		break;
1670 	}
1671 
1672 	case DIOCCLRSTATES: {
1673 		struct pf_state		*s, *nexts;
1674 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1675 		int			 killed = 0;
1676 
1677 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s; s = nexts) {
1678 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1679 
1680 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1681 			    s->kif->pfik_name)) {
1682 #if NPFSYNC
1683 				/* don't send out individual delete messages */
1684 				s->sync_flags = PFSTATE_NOSYNC;
1685 #endif
1686 				pf_unlink_state(s);
1687 				killed++;
1688 			}
1689 		}
1690 		psk->psk_af = killed;
1691 #if NPFSYNC
1692 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1693 #endif
1694 		break;
1695 	}
1696 
1697 	case DIOCKILLSTATES: {
1698 		struct pf_state		*s, *nexts;
1699 		struct pf_state_key	*sk;
1700 		struct pf_state_host	*src, *dst;
1701 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1702 		int			 killed = 0;
1703 
1704 		for (s = RB_MIN(pf_state_tree_id, &tree_id); s;
1705 		    s = nexts) {
1706 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, s);
1707 			sk = s->state_key;
1708 
1709 			if (sk->direction == PF_OUT) {
1710 				src = &sk->lan;
1711 				dst = &sk->ext;
1712 			} else {
1713 				src = &sk->ext;
1714 				dst = &sk->lan;
1715 			}
1716 			if ((!psk->psk_af || sk->af == psk->psk_af)
1717 			    && (!psk->psk_proto || psk->psk_proto ==
1718 			    sk->proto) &&
1719 			    PF_MATCHA(psk->psk_src.neg,
1720 			    &psk->psk_src.addr.v.a.addr,
1721 			    &psk->psk_src.addr.v.a.mask,
1722 			    &src->addr, sk->af) &&
1723 			    PF_MATCHA(psk->psk_dst.neg,
1724 			    &psk->psk_dst.addr.v.a.addr,
1725 			    &psk->psk_dst.addr.v.a.mask,
1726 			    &dst->addr, sk->af) &&
1727 			    (psk->psk_src.port_op == 0 ||
1728 			    pf_match_port(psk->psk_src.port_op,
1729 			    psk->psk_src.port[0], psk->psk_src.port[1],
1730 			    src->port)) &&
1731 			    (psk->psk_dst.port_op == 0 ||
1732 			    pf_match_port(psk->psk_dst.port_op,
1733 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1734 			    dst->port)) &&
1735 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1736 			    s->kif->pfik_name))) {
1737 #if NPFSYNC > 0
1738 				/* send immediate delete of state */
1739 				pfsync_delete_state(s);
1740 				s->sync_flags |= PFSTATE_NOSYNC;
1741 #endif
1742 				pf_unlink_state(s);
1743 				killed++;
1744 			}
1745 		}
1746 		psk->psk_af = killed;
1747 		break;
1748 	}
1749 
1750 	case DIOCADDSTATE: {
1751 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1752 		struct pfsync_state 	*sp = (struct pfsync_state *)ps->state;
1753 		struct pf_state		*s;
1754 		struct pf_state_key	*sk;
1755 		struct pfi_kif		*kif;
1756 
1757 		if (sp->timeout >= PFTM_MAX &&
1758 		    sp->timeout != PFTM_UNTIL_PACKET) {
1759 			error = EINVAL;
1760 			break;
1761 		}
1762 		s = pool_get(&pf_state_pl, PR_NOWAIT);
1763 		if (s == NULL) {
1764 			error = ENOMEM;
1765 			break;
1766 		}
1767 		bzero(s, sizeof(struct pf_state));
1768 		if ((sk = pf_alloc_state_key(s)) == NULL) {
1769 			error = ENOMEM;
1770 			break;
1771 		}
1772 		pf_state_import(sp, sk, s);
1773 		kif = pfi_kif_get(sp->ifname);
1774 		if (kif == NULL) {
1775 			pool_put(&pf_state_pl, s);
1776 			pool_put(&pf_state_key_pl, sk);
1777 			error = ENOENT;
1778 			break;
1779 		}
1780 		if (pf_insert_state(kif, s)) {
1781 			pfi_kif_unref(kif, PFI_KIF_REF_NONE);
1782 			pool_put(&pf_state_pl, s);
1783 			pool_put(&pf_state_key_pl, sk);
1784 			error = ENOMEM;
1785 		}
1786 		break;
1787 	}
1788 
1789 	case DIOCGETSTATE: {
1790 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1791 		struct pf_state		*s;
1792 		u_int32_t		 nr;
1793 
1794 		nr = 0;
1795 		RB_FOREACH(s, pf_state_tree_id, &tree_id) {
1796 			if (nr >= ps->nr)
1797 				break;
1798 			nr++;
1799 		}
1800 		if (s == NULL) {
1801 			error = EBUSY;
1802 			break;
1803 		}
1804 
1805 		pf_state_export((struct pfsync_state *)&ps->state,
1806 		    s->state_key, s);
1807 		break;
1808 	}
1809 
1810 	case DIOCGETSTATES: {
1811 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1812 		struct pf_state		*state;
1813 		struct pfsync_state	*p, *pstore;
1814 		u_int32_t		 nr = 0;
1815 
1816 		if (ps->ps_len == 0) {
1817 			nr = pf_status.states;
1818 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1819 			break;
1820 		}
1821 
1822 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1823 
1824 		p = ps->ps_states;
1825 
1826 		state = TAILQ_FIRST(&state_list);
1827 		while (state) {
1828 			if (state->timeout != PFTM_UNLINKED) {
1829 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1830 					break;
1831 
1832 				pf_state_export(pstore,
1833 				    state->state_key, state);
1834 				error = copyout(pstore, p, sizeof(*p));
1835 				if (error) {
1836 					kfree(pstore, M_TEMP);
1837 					goto fail;
1838 				}
1839 				p++;
1840 				nr++;
1841 			}
1842 			state = TAILQ_NEXT(state, entry_list);
1843 		}
1844 
1845 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1846 
1847 		kfree(pstore, M_TEMP);
1848 		break;
1849 	}
1850 
1851 	case DIOCGETSTATUS: {
1852 		struct pf_status *s = (struct pf_status *)addr;
1853 		bcopy(&pf_status, s, sizeof(struct pf_status));
1854 		pfi_fill_oldstatus(s);
1855 		break;
1856 	}
1857 
1858 	case DIOCSETSTATUSIF: {
1859 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1860 
1861 		if (pi->ifname[0] == 0) {
1862 			bzero(pf_status.ifname, IFNAMSIZ);
1863 			break;
1864 		}
1865 		if (ifunit(pi->ifname) == NULL) {
1866 			error = EINVAL;
1867 			break;
1868 		}
1869 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1870 		break;
1871 	}
1872 
1873 	case DIOCCLRSTATUS: {
1874 		bzero(pf_status.counters, sizeof(pf_status.counters));
1875 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1876 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
1877 		pf_status.since = time_second;
1878 		if (*pf_status.ifname)
1879 			pfi_clr_istats(pf_status.ifname);
1880 		break;
1881 	}
1882 
1883 	case DIOCNATLOOK: {
1884 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1885 		struct pf_state_key	*sk;
1886 		struct pf_state		*state;
1887 		struct pf_state_key_cmp	 key;
1888 		int			 m = 0, direction = pnl->direction;
1889 
1890 		key.af = pnl->af;
1891 		key.proto = pnl->proto;
1892 
1893 		if (!pnl->proto ||
1894 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1895 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1896 		    ((pnl->proto == IPPROTO_TCP ||
1897 		    pnl->proto == IPPROTO_UDP) &&
1898 		    (!pnl->dport || !pnl->sport)))
1899 			error = EINVAL;
1900 		else {
1901 			/*
1902 			 * userland gives us source and dest of connection,
1903 			 * reverse the lookup so we ask for what happens with
1904 			 * the return traffic, enabling us to find it in the
1905 			 * state tree.
1906 			 */
1907 			if (direction == PF_IN) {
1908 				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
1909 				key.ext.port = pnl->dport;
1910 				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
1911 				key.gwy.port = pnl->sport;
1912 				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
1913 			} else {
1914 				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
1915 				key.lan.port = pnl->dport;
1916 				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
1917 				key.ext.port = pnl->sport;
1918 				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
1919 			}
1920 			if (m > 1)
1921 				error = E2BIG;	/* more than one state */
1922 			else if (state != NULL) {
1923 				sk = state->state_key;
1924 				if (direction == PF_IN) {
1925 					PF_ACPY(&pnl->rsaddr, &sk->lan.addr,
1926 					    sk->af);
1927 					pnl->rsport = sk->lan.port;
1928 					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
1929 					    pnl->af);
1930 					pnl->rdport = pnl->dport;
1931 				} else {
1932 					PF_ACPY(&pnl->rdaddr, &sk->gwy.addr,
1933 					    sk->af);
1934 					pnl->rdport = sk->gwy.port;
1935 					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
1936 					    pnl->af);
1937 					pnl->rsport = pnl->sport;
1938 				}
1939 			} else
1940 				error = ENOENT;
1941 		}
1942 		break;
1943 	}
1944 
1945 	case DIOCSETTIMEOUT: {
1946 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1947 		int		 old;
1948 
1949 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1950 		    pt->seconds < 0) {
1951 			error = EINVAL;
1952 			goto fail;
1953 		}
1954 		old = pf_default_rule.timeout[pt->timeout];
1955 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1956 			pt->seconds = 1;
1957 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
1958 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1959 			wakeup(pf_purge_thread);
1960 		pt->seconds = old;
1961 		break;
1962 	}
1963 
1964 	case DIOCGETTIMEOUT: {
1965 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1966 
1967 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1968 			error = EINVAL;
1969 			goto fail;
1970 		}
1971 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1972 		break;
1973 	}
1974 
1975 	case DIOCGETLIMIT: {
1976 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1977 
1978 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1979 			error = EINVAL;
1980 			goto fail;
1981 		}
1982 		pl->limit = pf_pool_limits[pl->index].limit;
1983 		break;
1984 	}
1985 
1986 	case DIOCSETLIMIT: {
1987 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1988 		int			 old_limit;
1989 
1990 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1991 		    pf_pool_limits[pl->index].pp == NULL) {
1992 			error = EINVAL;
1993 			goto fail;
1994 		}
1995 
1996 		/* XXX Get an API to set limits on the zone/pool */
1997 		old_limit = pf_pool_limits[pl->index].limit;
1998 		pf_pool_limits[pl->index].limit = pl->limit;
1999 		pl->limit = old_limit;
2000 		break;
2001 	}
2002 
2003 	case DIOCSETDEBUG: {
2004 		u_int32_t	*level = (u_int32_t *)addr;
2005 
2006 		pf_status.debug = *level;
2007 		break;
2008 	}
2009 
2010 	case DIOCCLRRULECTRS: {
2011 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2012 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2013 		struct pf_rule		*rule;
2014 
2015 		TAILQ_FOREACH(rule,
2016 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2017 			rule->evaluations = 0;
2018 			rule->packets[0] = rule->packets[1] = 0;
2019 			rule->bytes[0] = rule->bytes[1] = 0;
2020 		}
2021 		break;
2022 	}
2023 
2024 	case DIOCGIFSPEED: {
2025 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
2026 		struct pf_ifspeed	ps;
2027 		struct ifnet		*ifp;
2028 
2029 		if (psp->ifname[0] != 0) {
2030 			/* Can we completely trust user-land? */
2031 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2032 			ifp = ifunit(ps.ifname);
2033 			if (ifp )
2034 				psp->baudrate = ifp->if_baudrate;
2035 			else
2036 				error = EINVAL;
2037 		} else
2038 			error = EINVAL;
2039 		break;
2040 	}
2041 #ifdef ALTQ
2042 	case DIOCSTARTALTQ: {
2043 		struct pf_altq		*altq;
2044 
2045 		/* enable all altq interfaces on active list */
2046 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2047 			if (altq->qname[0] == 0) {
2048 				error = pf_enable_altq(altq);
2049 				if (error != 0)
2050 					break;
2051 			}
2052 		}
2053 		if (error == 0)
2054 			pf_altq_running = 1;
2055 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2056 		break;
2057 	}
2058 
2059 	case DIOCSTOPALTQ: {
2060 		struct pf_altq		*altq;
2061 
2062 		/* disable all altq interfaces on active list */
2063 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2064 			if (altq->qname[0] == 0) {
2065 				error = pf_disable_altq(altq);
2066 				if (error != 0)
2067 					break;
2068 			}
2069 		}
2070 		if (error == 0)
2071 			pf_altq_running = 0;
2072 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2073 		break;
2074 	}
2075 
2076 	case DIOCADDALTQ: {
2077 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2078 		struct pf_altq		*altq, *a;
2079 
2080 		if (pa->ticket != ticket_altqs_inactive) {
2081 			error = EBUSY;
2082 			break;
2083 		}
2084 		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2085 		if (altq == NULL) {
2086 			error = ENOMEM;
2087 			break;
2088 		}
2089 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2090 
2091 		/*
2092 		 * if this is for a queue, find the discipline and
2093 		 * copy the necessary fields
2094 		 */
2095 		if (altq->qname[0] != 0) {
2096 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2097 				error = EBUSY;
2098 				pool_put(&pf_altq_pl, altq);
2099 				break;
2100 			}
2101 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2102 				if (strncmp(a->ifname, altq->ifname,
2103 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2104 					altq->altq_disc = a->altq_disc;
2105 					break;
2106 				}
2107 			}
2108 		}
2109 
2110 		error = altq_add(altq);
2111 		if (error) {
2112 			pool_put(&pf_altq_pl, altq);
2113 			break;
2114 		}
2115 
2116 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2117 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2118 		break;
2119 	}
2120 
2121 	case DIOCGETALTQS: {
2122 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2123 		struct pf_altq		*altq;
2124 
2125 		pa->nr = 0;
2126 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2127 			pa->nr++;
2128 		pa->ticket = ticket_altqs_active;
2129 		break;
2130 	}
2131 
2132 	case DIOCGETALTQ: {
2133 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2134 		struct pf_altq		*altq;
2135 		u_int32_t		 nr;
2136 
2137 		if (pa->ticket != ticket_altqs_active) {
2138 			error = EBUSY;
2139 			break;
2140 		}
2141 		nr = 0;
2142 		altq = TAILQ_FIRST(pf_altqs_active);
2143 		while ((altq != NULL) && (nr < pa->nr)) {
2144 			altq = TAILQ_NEXT(altq, entries);
2145 			nr++;
2146 		}
2147 		if (altq == NULL) {
2148 			error = EBUSY;
2149 			break;
2150 		}
2151 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2152 		break;
2153 	}
2154 
2155 	case DIOCCHANGEALTQ:
2156 		/* CHANGEALTQ not supported yet! */
2157 		error = ENODEV;
2158 		break;
2159 
2160 	case DIOCGETQSTATS: {
2161 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2162 		struct pf_altq		*altq;
2163 		u_int32_t		 nr;
2164 		int			 nbytes;
2165 
2166 		if (pq->ticket != ticket_altqs_active) {
2167 			error = EBUSY;
2168 			break;
2169 		}
2170 		nbytes = pq->nbytes;
2171 		nr = 0;
2172 		altq = TAILQ_FIRST(pf_altqs_active);
2173 		while ((altq != NULL) && (nr < pq->nr)) {
2174 			altq = TAILQ_NEXT(altq, entries);
2175 			nr++;
2176 		}
2177 		if (altq == NULL) {
2178 			error = EBUSY;
2179 			break;
2180 		}
2181 		error = altq_getqstats(altq, pq->buf, &nbytes);
2182 		if (error == 0) {
2183 			pq->scheduler = altq->scheduler;
2184 			pq->nbytes = nbytes;
2185 		}
2186 		break;
2187 	}
2188 #endif /* ALTQ */
2189 
2190 	case DIOCBEGINADDRS: {
2191 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2192 
2193 		pf_empty_pool(&pf_pabuf);
2194 		pp->ticket = ++ticket_pabuf;
2195 		break;
2196 	}
2197 
2198 	case DIOCADDADDR: {
2199 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2200 
2201 		if (pp->ticket != ticket_pabuf) {
2202 			error = EBUSY;
2203 			break;
2204 		}
2205 #ifndef INET
2206 		if (pp->af == AF_INET) {
2207 			error = EAFNOSUPPORT;
2208 			break;
2209 		}
2210 #endif /* INET */
2211 #ifndef INET6
2212 		if (pp->af == AF_INET6) {
2213 			error = EAFNOSUPPORT;
2214 			break;
2215 		}
2216 #endif /* INET6 */
2217 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2218 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2219 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2220 			error = EINVAL;
2221 			break;
2222 		}
2223 		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2224 		if (pa == NULL) {
2225 			error = ENOMEM;
2226 			break;
2227 		}
2228 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2229 		if (pa->ifname[0]) {
2230 			pa->kif = pfi_kif_get(pa->ifname);
2231 			if (pa->kif == NULL) {
2232 				pool_put(&pf_pooladdr_pl, pa);
2233 				error = EINVAL;
2234 				break;
2235 			}
2236 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2237 		}
2238 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2239 			pfi_dynaddr_remove(&pa->addr);
2240 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2241 			pool_put(&pf_pooladdr_pl, pa);
2242 			error = EINVAL;
2243 			break;
2244 		}
2245 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2246 		break;
2247 	}
2248 
2249 	case DIOCGETADDRS: {
2250 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2251 
2252 		pp->nr = 0;
2253 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2254 		    pp->r_num, 0, 1, 0);
2255 		if (pool == NULL) {
2256 			error = EBUSY;
2257 			break;
2258 		}
2259 		TAILQ_FOREACH(pa, &pool->list, entries)
2260 			pp->nr++;
2261 		break;
2262 	}
2263 
2264 	case DIOCGETADDR: {
2265 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2266 		u_int32_t		 nr = 0;
2267 
2268 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2269 		    pp->r_num, 0, 1, 1);
2270 		if (pool == NULL) {
2271 			error = EBUSY;
2272 			break;
2273 		}
2274 		pa = TAILQ_FIRST(&pool->list);
2275 		while ((pa != NULL) && (nr < pp->nr)) {
2276 			pa = TAILQ_NEXT(pa, entries);
2277 			nr++;
2278 		}
2279 		if (pa == NULL) {
2280 			error = EBUSY;
2281 			break;
2282 		}
2283 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2284 		pfi_dynaddr_copyout(&pp->addr.addr);
2285 		pf_tbladdr_copyout(&pp->addr.addr);
2286 		pf_rtlabel_copyout(&pp->addr.addr);
2287 		break;
2288 	}
2289 
2290 	case DIOCCHANGEADDR: {
2291 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2292 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2293 		struct pf_ruleset	*ruleset;
2294 
2295 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2296 		    pca->action > PF_CHANGE_REMOVE) {
2297 			error = EINVAL;
2298 			break;
2299 		}
2300 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2301 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2302 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2303 			error = EINVAL;
2304 			break;
2305 		}
2306 
2307 		ruleset = pf_find_ruleset(pca->anchor);
2308 		if (ruleset == NULL) {
2309 			error = EBUSY;
2310 			break;
2311 		}
2312 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2313 		    pca->r_num, pca->r_last, 1, 1);
2314 		if (pool == NULL) {
2315 			error = EBUSY;
2316 			break;
2317 		}
2318 		if (pca->action != PF_CHANGE_REMOVE) {
2319 			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2320 			if (newpa == NULL) {
2321 				error = ENOMEM;
2322 				break;
2323 			}
2324 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2325 #ifndef INET
2326 			if (pca->af == AF_INET) {
2327 				pool_put(&pf_pooladdr_pl, newpa);
2328 				error = EAFNOSUPPORT;
2329 				break;
2330 			}
2331 #endif /* INET */
2332 #ifndef INET6
2333 			if (pca->af == AF_INET6) {
2334 				pool_put(&pf_pooladdr_pl, newpa);
2335 				error = EAFNOSUPPORT;
2336 				break;
2337 			}
2338 #endif /* INET6 */
2339 			if (newpa->ifname[0]) {
2340 				newpa->kif = pfi_kif_get(newpa->ifname);
2341 				if (newpa->kif == NULL) {
2342 					pool_put(&pf_pooladdr_pl, newpa);
2343 					error = EINVAL;
2344 					break;
2345 				}
2346 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2347 			} else
2348 				newpa->kif = NULL;
2349 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2350 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2351 				pfi_dynaddr_remove(&newpa->addr);
2352 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2353 				pool_put(&pf_pooladdr_pl, newpa);
2354 				error = EINVAL;
2355 				break;
2356 			}
2357 		}
2358 
2359 		if (pca->action == PF_CHANGE_ADD_HEAD)
2360 			oldpa = TAILQ_FIRST(&pool->list);
2361 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2362 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2363 		else {
2364 			int	i = 0;
2365 
2366 			oldpa = TAILQ_FIRST(&pool->list);
2367 			while ((oldpa != NULL) && (i < pca->nr)) {
2368 				oldpa = TAILQ_NEXT(oldpa, entries);
2369 				i++;
2370 			}
2371 			if (oldpa == NULL) {
2372 				error = EINVAL;
2373 				break;
2374 			}
2375 		}
2376 
2377 		if (pca->action == PF_CHANGE_REMOVE) {
2378 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2379 			pfi_dynaddr_remove(&oldpa->addr);
2380 			pf_tbladdr_remove(&oldpa->addr);
2381 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2382 			pool_put(&pf_pooladdr_pl, oldpa);
2383 		} else {
2384 			if (oldpa == NULL)
2385 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2386 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2387 			    pca->action == PF_CHANGE_ADD_BEFORE)
2388 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2389 			else
2390 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2391 				    newpa, entries);
2392 		}
2393 
2394 		pool->cur = TAILQ_FIRST(&pool->list);
2395 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2396 		    pca->af);
2397 		break;
2398 	}
2399 
2400 	case DIOCGETRULESETS: {
2401 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2402 		struct pf_ruleset	*ruleset;
2403 		struct pf_anchor	*anchor;
2404 
2405 		pr->path[sizeof(pr->path) - 1] = 0;
2406 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2407 			error = EINVAL;
2408 			break;
2409 		}
2410 		pr->nr = 0;
2411 		if (ruleset->anchor == NULL) {
2412 			/* XXX kludge for pf_main_ruleset */
2413 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2414 				if (anchor->parent == NULL)
2415 					pr->nr++;
2416 		} else {
2417 			RB_FOREACH(anchor, pf_anchor_node,
2418 			    &ruleset->anchor->children)
2419 				pr->nr++;
2420 		}
2421 		break;
2422 	}
2423 
2424 	case DIOCGETRULESET: {
2425 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2426 		struct pf_ruleset	*ruleset;
2427 		struct pf_anchor	*anchor;
2428 		u_int32_t		 nr = 0;
2429 
2430 		pr->path[sizeof(pr->path) - 1] = 0;
2431 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2432 			error = EINVAL;
2433 			break;
2434 		}
2435 		pr->name[0] = 0;
2436 		if (ruleset->anchor == NULL) {
2437 			/* XXX kludge for pf_main_ruleset */
2438 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2439 				if (anchor->parent == NULL && nr++ == pr->nr) {
2440 					strlcpy(pr->name, anchor->name,
2441 					    sizeof(pr->name));
2442 					break;
2443 				}
2444 		} else {
2445 			RB_FOREACH(anchor, pf_anchor_node,
2446 			    &ruleset->anchor->children)
2447 				if (nr++ == pr->nr) {
2448 					strlcpy(pr->name, anchor->name,
2449 					    sizeof(pr->name));
2450 					break;
2451 				}
2452 		}
2453 		if (!pr->name[0])
2454 			error = EBUSY;
2455 		break;
2456 	}
2457 
2458 	case DIOCRCLRTABLES: {
2459 		struct pfioc_table *io = (struct pfioc_table *)addr;
2460 
2461 		if (io->pfrio_esize != 0) {
2462 			error = ENODEV;
2463 			break;
2464 		}
2465 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2466 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2467 		break;
2468 	}
2469 
2470 	case DIOCRADDTABLES: {
2471 		struct pfioc_table *io = (struct pfioc_table *)addr;
2472 
2473 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2474 			error = ENODEV;
2475 			break;
2476 		}
2477 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2478 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2479 		break;
2480 	}
2481 
2482 	case DIOCRDELTABLES: {
2483 		struct pfioc_table *io = (struct pfioc_table *)addr;
2484 
2485 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2486 			error = ENODEV;
2487 			break;
2488 		}
2489 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2490 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2491 		break;
2492 	}
2493 
2494 	case DIOCRGETTABLES: {
2495 		struct pfioc_table *io = (struct pfioc_table *)addr;
2496 
2497 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2498 			error = ENODEV;
2499 			break;
2500 		}
2501 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2502 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2503 		break;
2504 	}
2505 
2506 	case DIOCRGETTSTATS: {
2507 		struct pfioc_table *io = (struct pfioc_table *)addr;
2508 
2509 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2510 			error = ENODEV;
2511 			break;
2512 		}
2513 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2514 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2515 		break;
2516 	}
2517 
2518 	case DIOCRCLRTSTATS: {
2519 		struct pfioc_table *io = (struct pfioc_table *)addr;
2520 
2521 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2522 			error = ENODEV;
2523 			break;
2524 		}
2525 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2526 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2527 		break;
2528 	}
2529 
2530 	case DIOCRSETTFLAGS: {
2531 		struct pfioc_table *io = (struct pfioc_table *)addr;
2532 
2533 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2534 			error = ENODEV;
2535 			break;
2536 		}
2537 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2538 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2539 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2540 		break;
2541 	}
2542 
2543 	case DIOCRCLRADDRS: {
2544 		struct pfioc_table *io = (struct pfioc_table *)addr;
2545 
2546 		if (io->pfrio_esize != 0) {
2547 			error = ENODEV;
2548 			break;
2549 		}
2550 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2551 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2552 		break;
2553 	}
2554 
2555 	case DIOCRADDADDRS: {
2556 		struct pfioc_table *io = (struct pfioc_table *)addr;
2557 
2558 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2559 			error = ENODEV;
2560 			break;
2561 		}
2562 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2563 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2564 		    PFR_FLAG_USERIOCTL);
2565 		break;
2566 	}
2567 
2568 	case DIOCRDELADDRS: {
2569 		struct pfioc_table *io = (struct pfioc_table *)addr;
2570 
2571 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2572 			error = ENODEV;
2573 			break;
2574 		}
2575 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2576 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2577 		    PFR_FLAG_USERIOCTL);
2578 		break;
2579 	}
2580 
2581 	case DIOCRSETADDRS: {
2582 		struct pfioc_table *io = (struct pfioc_table *)addr;
2583 
2584 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2585 			error = ENODEV;
2586 			break;
2587 		}
2588 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2589 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2590 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2591 		    PFR_FLAG_USERIOCTL, 0);
2592 		break;
2593 	}
2594 
2595 	case DIOCRGETADDRS: {
2596 		struct pfioc_table *io = (struct pfioc_table *)addr;
2597 
2598 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2599 			error = ENODEV;
2600 			break;
2601 		}
2602 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2603 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2604 		break;
2605 	}
2606 
2607 	case DIOCRGETASTATS: {
2608 		struct pfioc_table *io = (struct pfioc_table *)addr;
2609 
2610 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2611 			error = ENODEV;
2612 			break;
2613 		}
2614 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2615 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2616 		break;
2617 	}
2618 
2619 	case DIOCRCLRASTATS: {
2620 		struct pfioc_table *io = (struct pfioc_table *)addr;
2621 
2622 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2623 			error = ENODEV;
2624 			break;
2625 		}
2626 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2627 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2628 		    PFR_FLAG_USERIOCTL);
2629 		break;
2630 	}
2631 
2632 	case DIOCRTSTADDRS: {
2633 		struct pfioc_table *io = (struct pfioc_table *)addr;
2634 
2635 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2636 			error = ENODEV;
2637 			break;
2638 		}
2639 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2640 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2641 		    PFR_FLAG_USERIOCTL);
2642 		break;
2643 	}
2644 
2645 	case DIOCRINADEFINE: {
2646 		struct pfioc_table *io = (struct pfioc_table *)addr;
2647 
2648 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2649 			error = ENODEV;
2650 			break;
2651 		}
2652 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2653 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2654 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2655 		break;
2656 	}
2657 
2658 	case DIOCOSFPADD: {
2659 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2660 		error = pf_osfp_add(io);
2661 		break;
2662 	}
2663 
2664 	case DIOCOSFPGET: {
2665 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2666 		error = pf_osfp_get(io);
2667 		break;
2668 	}
2669 
2670 	case DIOCXBEGIN: {
2671 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2672 		struct pfioc_trans_e	*ioe;
2673 		struct pfr_table	*table;
2674 		int			 i;
2675 
2676 		if (io->esize != sizeof(*ioe)) {
2677 			error = ENODEV;
2678 			goto fail;
2679 		}
2680 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2681 		    M_TEMP, M_WAITOK);
2682 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2683 		    M_TEMP, M_WAITOK);
2684 		for (i = 0; i < io->size; i++) {
2685 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2686 				kfree(table, M_TEMP);
2687 				kfree(ioe, M_TEMP);
2688 				error = EFAULT;
2689 				goto fail;
2690 			}
2691 			switch (ioe->rs_num) {
2692 #ifdef ALTQ
2693 			case PF_RULESET_ALTQ:
2694 				if (ioe->anchor[0]) {
2695 					kfree(table, M_TEMP);
2696 					kfree(ioe, M_TEMP);
2697 					error = EINVAL;
2698 					goto fail;
2699 				}
2700 				if ((error = pf_begin_altq(&ioe->ticket))) {
2701 					kfree(table, M_TEMP);
2702 					kfree(ioe, M_TEMP);
2703 					goto fail;
2704 				}
2705 				break;
2706 #endif /* ALTQ */
2707 			case PF_RULESET_TABLE:
2708 				bzero(table, sizeof(*table));
2709 				strlcpy(table->pfrt_anchor, ioe->anchor,
2710 				    sizeof(table->pfrt_anchor));
2711 				if ((error = pfr_ina_begin(table,
2712 				    &ioe->ticket, NULL, 0))) {
2713 					kfree(table, M_TEMP);
2714 					kfree(ioe, M_TEMP);
2715 					goto fail;
2716 				}
2717 				break;
2718 			default:
2719 				if ((error = pf_begin_rules(&ioe->ticket,
2720 				    ioe->rs_num, ioe->anchor))) {
2721 					kfree(table, M_TEMP);
2722 					kfree(ioe, M_TEMP);
2723 					goto fail;
2724 				}
2725 				break;
2726 			}
2727 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2728 				kfree(table, M_TEMP);
2729 				kfree(ioe, M_TEMP);
2730 				error = EFAULT;
2731 				goto fail;
2732 			}
2733 		}
2734 		kfree(table, M_TEMP);
2735 		kfree(ioe, M_TEMP);
2736 		break;
2737 	}
2738 
2739 	case DIOCXROLLBACK: {
2740 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2741 		struct pfioc_trans_e	*ioe;
2742 		struct pfr_table	*table;
2743 		int			 i;
2744 
2745 		if (io->esize != sizeof(*ioe)) {
2746 			error = ENODEV;
2747 			goto fail;
2748 		}
2749 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2750 		    M_TEMP, M_WAITOK);
2751 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2752 		    M_TEMP, M_WAITOK);
2753 		for (i = 0; i < io->size; i++) {
2754 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2755 				kfree(table, M_TEMP);
2756 				kfree(ioe, M_TEMP);
2757 				error = EFAULT;
2758 				goto fail;
2759 			}
2760 			switch (ioe->rs_num) {
2761 #ifdef ALTQ
2762 			case PF_RULESET_ALTQ:
2763 				if (ioe->anchor[0]) {
2764 					kfree(table, M_TEMP);
2765 					kfree(ioe, M_TEMP);
2766 					error = EINVAL;
2767 					goto fail;
2768 				}
2769 				if ((error = pf_rollback_altq(ioe->ticket))) {
2770 					kfree(table, M_TEMP);
2771 					kfree(ioe, M_TEMP);
2772 					goto fail; /* really bad */
2773 				}
2774 				break;
2775 #endif /* ALTQ */
2776 			case PF_RULESET_TABLE:
2777 				bzero(table, sizeof(*table));
2778 				strlcpy(table->pfrt_anchor, ioe->anchor,
2779 				    sizeof(table->pfrt_anchor));
2780 				if ((error = pfr_ina_rollback(table,
2781 				    ioe->ticket, NULL, 0))) {
2782 					kfree(table, M_TEMP);
2783 					kfree(ioe, M_TEMP);
2784 					goto fail; /* really bad */
2785 				}
2786 				break;
2787 			default:
2788 				if ((error = pf_rollback_rules(ioe->ticket,
2789 				    ioe->rs_num, ioe->anchor))) {
2790 					kfree(table, M_TEMP);
2791 					kfree(ioe, M_TEMP);
2792 					goto fail; /* really bad */
2793 				}
2794 				break;
2795 			}
2796 		}
2797 		kfree(table, M_TEMP);
2798 		kfree(ioe, M_TEMP);
2799 		break;
2800 	}
2801 
2802 	case DIOCXCOMMIT: {
2803 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2804 		struct pfioc_trans_e	*ioe;
2805 		struct pfr_table	*table;
2806 		struct pf_ruleset	*rs;
2807 		int			 i;
2808 
2809 		if (io->esize != sizeof(*ioe)) {
2810 			error = ENODEV;
2811 			goto fail;
2812 		}
2813 		ioe = (struct pfioc_trans_e *)kmalloc(sizeof(*ioe),
2814 		    M_TEMP, M_WAITOK);
2815 		table = (struct pfr_table *)kmalloc(sizeof(*table),
2816 		    M_TEMP, M_WAITOK);
2817 		/* first makes sure everything will succeed */
2818 		for (i = 0; i < io->size; i++) {
2819 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2820 				kfree(table, M_TEMP);
2821 				kfree(ioe, M_TEMP);
2822 				error = EFAULT;
2823 				goto fail;
2824 			}
2825 			switch (ioe->rs_num) {
2826 #ifdef ALTQ
2827 			case PF_RULESET_ALTQ:
2828 				if (ioe->anchor[0]) {
2829 					kfree(table, M_TEMP);
2830 					kfree(ioe, M_TEMP);
2831 					error = EINVAL;
2832 					goto fail;
2833 				}
2834 				if (!altqs_inactive_open || ioe->ticket !=
2835 				    ticket_altqs_inactive) {
2836 					kfree(table, M_TEMP);
2837 					kfree(ioe, M_TEMP);
2838 					error = EBUSY;
2839 					goto fail;
2840 				}
2841 				break;
2842 #endif /* ALTQ */
2843 			case PF_RULESET_TABLE:
2844 				rs = pf_find_ruleset(ioe->anchor);
2845 				if (rs == NULL || !rs->topen || ioe->ticket !=
2846 				     rs->tticket) {
2847 					kfree(table, M_TEMP);
2848 					kfree(ioe, M_TEMP);
2849 					error = EBUSY;
2850 					goto fail;
2851 				}
2852 				break;
2853 			default:
2854 				if (ioe->rs_num < 0 || ioe->rs_num >=
2855 				    PF_RULESET_MAX) {
2856 					kfree(table, M_TEMP);
2857 					kfree(ioe, M_TEMP);
2858 					error = EINVAL;
2859 					goto fail;
2860 				}
2861 				rs = pf_find_ruleset(ioe->anchor);
2862 				if (rs == NULL ||
2863 				    !rs->rules[ioe->rs_num].inactive.open ||
2864 				    rs->rules[ioe->rs_num].inactive.ticket !=
2865 				    ioe->ticket) {
2866 					kfree(table, M_TEMP);
2867 					kfree(ioe, M_TEMP);
2868 					error = EBUSY;
2869 					goto fail;
2870 				}
2871 				break;
2872 			}
2873 		}
2874 		/* now do the commit - no errors should happen here */
2875 		for (i = 0; i < io->size; i++) {
2876 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2877 				kfree(table, M_TEMP);
2878 				kfree(ioe, M_TEMP);
2879 				error = EFAULT;
2880 				goto fail;
2881 			}
2882 			switch (ioe->rs_num) {
2883 #ifdef ALTQ
2884 			case PF_RULESET_ALTQ:
2885 				if ((error = pf_commit_altq(ioe->ticket))) {
2886 					kfree(table, M_TEMP);
2887 					kfree(ioe, M_TEMP);
2888 					goto fail; /* really bad */
2889 				}
2890 				break;
2891 #endif /* ALTQ */
2892 			case PF_RULESET_TABLE:
2893 				bzero(table, sizeof(*table));
2894 				strlcpy(table->pfrt_anchor, ioe->anchor,
2895 				    sizeof(table->pfrt_anchor));
2896 				if ((error = pfr_ina_commit(table, ioe->ticket,
2897 				    NULL, NULL, 0))) {
2898 					kfree(table, M_TEMP);
2899 					kfree(ioe, M_TEMP);
2900 					goto fail; /* really bad */
2901 				}
2902 				break;
2903 			default:
2904 				if ((error = pf_commit_rules(ioe->ticket,
2905 				    ioe->rs_num, ioe->anchor))) {
2906 					kfree(table, M_TEMP);
2907 					kfree(ioe, M_TEMP);
2908 					goto fail; /* really bad */
2909 				}
2910 				break;
2911 			}
2912 		}
2913 		kfree(table, M_TEMP);
2914 		kfree(ioe, M_TEMP);
2915 		break;
2916 	}
2917 
2918 	case DIOCGETSRCNODES: {
2919 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2920 		struct pf_src_node	*n, *p, *pstore;
2921 		u_int32_t		 nr = 0;
2922 		int			 space = psn->psn_len;
2923 
2924 		if (space == 0) {
2925 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
2926 				nr++;
2927 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2928 			break;
2929 		}
2930 
2931 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2932 
2933 		p = psn->psn_src_nodes;
2934 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2935 			int	secs = time_second, diff;
2936 
2937 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
2938 				break;
2939 
2940 			bcopy(n, pstore, sizeof(*pstore));
2941 			if (n->rule.ptr != NULL)
2942 				pstore->rule.nr = n->rule.ptr->nr;
2943 			pstore->creation = secs - pstore->creation;
2944 			if (pstore->expire > secs)
2945 				pstore->expire -= secs;
2946 			else
2947 				pstore->expire = 0;
2948 
2949 			/* adjust the connection rate estimate */
2950 			diff = secs - n->conn_rate.last;
2951 			if (diff >= n->conn_rate.seconds)
2952 				pstore->conn_rate.count = 0;
2953 			else
2954 				pstore->conn_rate.count -=
2955 				    n->conn_rate.count * diff /
2956 				    n->conn_rate.seconds;
2957 
2958 			error = copyout(pstore, p, sizeof(*p));
2959 			if (error) {
2960 				kfree(pstore, M_TEMP);
2961 				goto fail;
2962 			}
2963 			p++;
2964 			nr++;
2965 		}
2966 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2967 
2968 		kfree(pstore, M_TEMP);
2969 		break;
2970 	}
2971 
2972 	case DIOCCLRSRCNODES: {
2973 		struct pf_src_node	*n;
2974 		struct pf_state		*state;
2975 
2976 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
2977 			state->src_node = NULL;
2978 			state->nat_src_node = NULL;
2979 		}
2980 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
2981 			n->expire = 1;
2982 			n->states = 0;
2983 		}
2984 		pf_purge_expired_src_nodes(1);
2985 		pf_status.src_nodes = 0;
2986 		break;
2987 	}
2988 
2989 	case DIOCKILLSRCNODES: {
2990 		struct pf_src_node	*sn;
2991 		struct pf_state		*s;
2992 		struct pfioc_src_node_kill *psnk = \
2993 			(struct pfioc_src_node_kill *) addr;
2994 		int			killed = 0;
2995 
2996 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
2997         		if (PF_MATCHA(psnk->psnk_src.neg, \
2998 				      &psnk->psnk_src.addr.v.a.addr, \
2999 				      &psnk->psnk_src.addr.v.a.mask, \
3000 				      &sn->addr, sn->af) &&
3001 			    PF_MATCHA(psnk->psnk_dst.neg, \
3002 				      &psnk->psnk_dst.addr.v.a.addr, \
3003 				      &psnk->psnk_dst.addr.v.a.mask, \
3004 				      &sn->raddr, sn->af)) {
3005 				/* Handle state to src_node linkage */
3006 				if (sn->states != 0) {
3007 					RB_FOREACH(s, pf_state_tree_id,
3008 					    &tree_id) {
3009 						if (s->src_node == sn)
3010 							s->src_node = NULL;
3011 						if (s->nat_src_node == sn)
3012 							s->nat_src_node = NULL;
3013 					}
3014 					sn->states = 0;
3015 				}
3016 				sn->expire = 1;
3017 				killed++;
3018 			}
3019 		}
3020 
3021 		if (killed > 0)
3022 			pf_purge_expired_src_nodes(1);
3023 
3024 		psnk->psnk_af = killed;
3025 		break;
3026 	}
3027 
3028 	case DIOCSETHOSTID: {
3029 		u_int32_t	*hostid = (u_int32_t *)addr;
3030 
3031 		if (*hostid == 0)
3032 			pf_status.hostid = karc4random();
3033 		else
3034 			pf_status.hostid = *hostid;
3035 		break;
3036 	}
3037 
3038 	case DIOCOSFPFLUSH:
3039 		crit_enter();
3040 		pf_osfp_flush();
3041 		crit_exit();
3042 		break;
3043 
3044 	case DIOCIGETIFACES: {
3045 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3046 
3047 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3048 			error = ENODEV;
3049 			break;
3050 		}
3051 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3052 		    &io->pfiio_size);
3053 		break;
3054 	}
3055 
3056 	case DIOCSETIFFLAG: {
3057 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3058 
3059 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3060 		break;
3061 	}
3062 
3063 	case DIOCCLRIFFLAG: {
3064 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3065 
3066 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3067 		break;
3068 	}
3069 
3070 	default:
3071 		error = ENODEV;
3072 		break;
3073 	}
3074 fail:
3075 	lwkt_reltoken(&pf_token);
3076 	return (error);
3077 }
3078 
3079 /*
3080  * XXX - Check for version missmatch!!!
3081  */
3082 static void
3083 pf_clear_states(void)
3084 {
3085 	struct pf_state		*state;
3086 
3087 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3088 		state->timeout = PFTM_PURGE;
3089 #if NPFSYNC
3090 		/* don't send out individual delete messages */
3091 		state->sync_flags = PFSTATE_NOSYNC;
3092 #endif
3093 		pf_unlink_state(state);
3094 	}
3095 	pf_status.states = 0;
3096 #if 0 /* NPFSYNC */
3097 /*
3098  * XXX This is called on module unload, we do not want to sync that over? */
3099  */
3100 	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
3101 #endif
3102 }
3103 
3104 static int
3105 pf_clear_tables(void)
3106 {
3107 	struct pfioc_table io;
3108 	int error;
3109 
3110 	bzero(&io, sizeof(io));
3111 
3112 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3113 	    io.pfrio_flags);
3114 
3115 	return (error);
3116 }
3117 
3118 static void
3119 pf_clear_srcnodes(void)
3120 {
3121 	struct pf_src_node	*n;
3122 	struct pf_state		*state;
3123 
3124 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3125 		state->src_node = NULL;
3126 		state->nat_src_node = NULL;
3127 	}
3128 	RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3129 		n->expire = 1;
3130 		n->states = 0;
3131 	}
3132 	pf_purge_expired_src_nodes(0);
3133 	pf_status.src_nodes = 0;
3134 }
3135 /*
3136  * XXX - Check for version missmatch!!!
3137  */
3138 
3139 /*
3140  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3141  */
3142 static int
3143 shutdown_pf(void)
3144 {
3145 	int error = 0;
3146 	u_int32_t t[5];
3147 	char nn = '\0';
3148 
3149 	pf_status.running = 0;
3150 	do {
3151 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
3152 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3153 			break;
3154 		}
3155 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
3156 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3157 			break;		/* XXX: rollback? */
3158 		}
3159 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
3160 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3161 			break;		/* XXX: rollback? */
3162 		}
3163 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3164 		    != 0) {
3165 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3166 			break;		/* XXX: rollback? */
3167 		}
3168 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3169 		    != 0) {
3170 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3171 			break;		/* XXX: rollback? */
3172 		}
3173 
3174 		/* XXX: these should always succeed here */
3175 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3176 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3177 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3178 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3179 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3180 
3181 		if ((error = pf_clear_tables()) != 0)
3182 			break;
3183 
3184 #ifdef ALTQ
3185 		if ((error = pf_begin_altq(&t[0])) != 0) {
3186 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3187 			break;
3188 		}
3189 		pf_commit_altq(t[0]);
3190 #endif
3191 
3192 		pf_clear_states();
3193 
3194 		pf_clear_srcnodes();
3195 
3196 		/* status does not use malloced mem so no need to cleanup */
3197 		/* fingerprints and interfaces have their own cleanup code */
3198 	} while(0);
3199 
3200         return (error);
3201 }
3202 
3203 static int
3204 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3205 {
3206 	/*
3207 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3208 	 * for ip_len/ip_off. This is why we don't have to change byte order
3209 	 * like the FreeBSD-5 version does.
3210 	 */
3211 	int chk;
3212 
3213 	lwkt_gettoken(&pf_token);
3214 
3215 	chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3216 	if (chk && *m) {
3217 		m_freem(*m);
3218 		*m = NULL;
3219 	}
3220 	lwkt_reltoken(&pf_token);
3221 	return chk;
3222 }
3223 
3224 static int
3225 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3226 {
3227 	/*
3228 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3229 	 * for ip_len/ip_off. This is why we don't have to change byte order
3230 	 * like the FreeBSD-5 version does.
3231 	 */
3232 	int chk;
3233 
3234 	lwkt_gettoken(&pf_token);
3235 
3236 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3237 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3238 		in_delayed_cksum(*m);
3239 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3240 	}
3241 	chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3242 	if (chk && *m) {
3243 		m_freem(*m);
3244 		*m = NULL;
3245 	}
3246 	lwkt_reltoken(&pf_token);
3247 	return chk;
3248 }
3249 
3250 #ifdef INET6
3251 static int
3252 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3253 {
3254 	/*
3255 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3256 	 */
3257 	int chk;
3258 
3259 	lwkt_gettoken(&pf_token);
3260 
3261 	chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3262 	if (chk && *m) {
3263 		m_freem(*m);
3264 		*m = NULL;
3265 	}
3266 	lwkt_reltoken(&pf_token);
3267 	return chk;
3268 }
3269 
3270 static int
3271 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3272 {
3273 	/*
3274 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3275 	 */
3276 	int chk;
3277 
3278 	lwkt_gettoken(&pf_token);
3279 
3280 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3281 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3282 		in_delayed_cksum(*m);
3283 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3284 	}
3285 	chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3286 	if (chk && *m) {
3287 		m_freem(*m);
3288 		*m = NULL;
3289 	}
3290 	lwkt_reltoken(&pf_token);
3291 	return chk;
3292 }
3293 #endif /* INET6 */
3294 
3295 static int
3296 hook_pf(void)
3297 {
3298 	struct pfil_head *pfh_inet;
3299 #ifdef INET6
3300 	struct pfil_head *pfh_inet6;
3301 #endif
3302 
3303 	lwkt_gettoken(&pf_token);
3304 
3305 	if (pf_pfil_hooked) {
3306 		lwkt_reltoken(&pf_token);
3307 		return (0);
3308 	}
3309 
3310 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3311 	if (pfh_inet == NULL) {
3312 		lwkt_reltoken(&pf_token);
3313 		return (ENODEV);
3314 	}
3315 	pfil_add_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3316 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3317 #ifdef INET6
3318 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3319 	if (pfh_inet6 == NULL) {
3320 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3321 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3322 		lwkt_reltoken(&pf_token);
3323 		return (ENODEV);
3324 	}
3325 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3326 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3327 #endif
3328 
3329 	pf_pfil_hooked = 1;
3330 	lwkt_reltoken(&pf_token);
3331 	return (0);
3332 }
3333 
3334 static int
3335 dehook_pf(void)
3336 {
3337 	struct pfil_head *pfh_inet;
3338 #ifdef INET6
3339 	struct pfil_head *pfh_inet6;
3340 #endif
3341 
3342 	lwkt_gettoken(&pf_token);
3343 
3344 	if (pf_pfil_hooked == 0) {
3345 		lwkt_reltoken(&pf_token);
3346 		return (0);
3347 	}
3348 
3349 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3350 	if (pfh_inet == NULL) {
3351 		lwkt_reltoken(&pf_token);
3352 		return (ENODEV);
3353 	}
3354 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3355 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3356 #ifdef INET6
3357 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3358 	if (pfh_inet6 == NULL) {
3359 		lwkt_reltoken(&pf_token);
3360 		return (ENODEV);
3361 	}
3362 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3363 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3364 #endif
3365 
3366 	pf_pfil_hooked = 0;
3367 	lwkt_reltoken(&pf_token);
3368 	return (0);
3369 }
3370 
3371 static int
3372 pf_load(void)
3373 {
3374 	int error;
3375 
3376 	lwkt_gettoken(&pf_token);
3377 
3378 	init_zone_var();
3379 	lockinit(&pf_mod_lck, "pf task lck", 0, LK_CANRECURSE);
3380 	pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3381 	error = pfattach();
3382 	if (error) {
3383 		dev_ops_remove_all(&pf_ops);
3384 		lockuninit(&pf_mod_lck);
3385 		lwkt_reltoken(&pf_token);
3386 		return (error);
3387 	}
3388 	lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3389 	lwkt_reltoken(&pf_token);
3390 	return (0);
3391 }
3392 
3393 static int
3394 pf_unload(void)
3395 {
3396 	int error;
3397 	pf_status.running = 0;
3398 
3399 	lwkt_gettoken(&pf_token);
3400 
3401 	error = dehook_pf();
3402 	if (error) {
3403 		/*
3404 		 * Should not happen!
3405 		 * XXX Due to error code ESRCH, kldunload will show
3406 		 * a message like 'No such process'.
3407 		 */
3408 		kprintf("pfil unregistration fail\n");
3409 		lwkt_reltoken(&pf_token);
3410 		return error;
3411 	}
3412 	shutdown_pf();
3413 	pf_end_threads = 1;
3414 	while (pf_end_threads < 2) {
3415 		wakeup_one(pf_purge_thread);
3416 		lksleep(pf_purge_thread, &pf_mod_lck, 0, "pftmo", hz);
3417 
3418 	}
3419 	pfi_cleanup();
3420 	pf_osfp_flush();
3421 	pf_osfp_cleanup();
3422 	cleanup_pf_zone();
3423 	dev_ops_remove_all(&pf_ops);
3424 	lockuninit(&pf_consistency_lock);
3425 	lockuninit(&pf_mod_lck);
3426 	lwkt_reltoken(&pf_token);
3427 	return 0;
3428 }
3429 
3430 static int
3431 pf_modevent(module_t mod, int type, void *data)
3432 {
3433 	int error = 0;
3434 
3435 	lwkt_gettoken(&pf_token);
3436 
3437 	switch(type) {
3438 	case MOD_LOAD:
3439 		error = pf_load();
3440 		break;
3441 
3442 	case MOD_UNLOAD:
3443 		error = pf_unload();
3444 		break;
3445 	default:
3446 		error = EINVAL;
3447 		break;
3448 	}
3449 	lwkt_reltoken(&pf_token);
3450 	return error;
3451 }
3452 
3453 static moduledata_t pf_mod = {
3454 	"pf",
3455 	pf_modevent,
3456 	0
3457 };
3458 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3459 MODULE_VERSION(pf, PF_MODVER);
3460