pf_ioctl.c revision 158486
1/*	$FreeBSD: head/sys/contrib/pf/net/pf_ioctl.c 158486 2006-05-12 16:15:34Z mlaier $	*/
2/*	$OpenBSD: pf_ioctl.c,v 1.139 2005/03/03 07:13:39 dhartmei Exp $ */
3
4/*
5 * Copyright (c) 2001 Daniel Hartmeier
6 * Copyright (c) 2002,2003 Henning Brauer
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 *
13 *    - Redistributions of source code must retain the above copyright
14 *      notice, this list of conditions and the following disclaimer.
15 *    - Redistributions in binary form must reproduce the above
16 *      copyright notice, this list of conditions and the following
17 *      disclaimer in the documentation and/or other materials provided
18 *      with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28 * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30 * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 *
33 * Effort sponsored in part by the Defense Advanced Research Projects
34 * Agency (DARPA) and Air Force Research Laboratory, Air Force
35 * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36 *
37 */
38
39#ifdef __FreeBSD__
40#include "opt_inet.h"
41#include "opt_inet6.h"
42#endif
43
44#ifdef __FreeBSD__
45#include "opt_bpf.h"
46#include "opt_pf.h"
47
48#ifdef DEV_BPF
49#define	NBPFILTER	DEV_BPF
50#else
51#define	NBPFILTER	0
52#endif
53
54#ifdef DEV_PFLOG
55#define	NPFLOG		DEV_PFLOG
56#else
57#define	NPFLOG		0
58#endif
59
60#ifdef DEV_PFSYNC
61#define	NPFSYNC		DEV_PFSYNC
62#else
63#define	NPFSYNC		0
64#endif
65
66#else
67#include "bpfilter.h"
68#include "pflog.h"
69#include "pfsync.h"
70#endif
71
72#include <sys/param.h>
73#include <sys/systm.h>
74#include <sys/mbuf.h>
75#include <sys/filio.h>
76#include <sys/fcntl.h>
77#include <sys/socket.h>
78#include <sys/socketvar.h>
79#include <sys/kernel.h>
80#include <sys/time.h>
81#include <sys/malloc.h>
82#ifdef __FreeBSD__
83#include <sys/module.h>
84#include <sys/conf.h>
85#include <sys/proc.h>
86#else
87#include <sys/timeout.h>
88#include <sys/pool.h>
89#endif
90
91#include <net/if.h>
92#include <net/if_types.h>
93#include <net/route.h>
94
95#include <netinet/in.h>
96#include <netinet/in_var.h>
97#include <netinet/in_systm.h>
98#include <netinet/ip.h>
99#include <netinet/ip_var.h>
100#include <netinet/ip_icmp.h>
101
102#ifndef __FreeBSD__
103#include <dev/rndvar.h>
104#endif
105#include <net/pfvar.h>
106
107#if NPFSYNC > 0
108#include <net/if_pfsync.h>
109#endif /* NPFSYNC > 0 */
110
111#ifdef __FreeBSD__
112#include <net/if_pflog.h>
113#endif
114
115#ifdef INET6
116#include <netinet/ip6.h>
117#include <netinet/in_pcb.h>
118#endif /* INET6 */
119
120#ifdef ALTQ
121#include <altq/altq.h>
122#endif
123
124#ifdef __FreeBSD__
125#include <sys/limits.h>
126#include <sys/lock.h>
127#include <sys/mutex.h>
128#include <net/pfil.h>
129#endif /* __FreeBSD__ */
130
131#ifdef __FreeBSD__
132void			 init_zone_var(void);
133void			 cleanup_pf_zone(void);
134int			 pfattach(void);
135#else
136void			 pfattach(int);
137int			 pfopen(dev_t, int, int, struct proc *);
138int			 pfclose(dev_t, int, int, struct proc *);
139#endif
140struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
141			    u_int8_t, u_int8_t, u_int8_t);
142int			 pf_get_ruleset_number(u_int8_t);
143void			 pf_init_ruleset(struct pf_ruleset *);
144int			 pf_anchor_setup(struct pf_rule *,
145			    const struct pf_ruleset *, const char *);
146int			 pf_anchor_copyout(const struct pf_ruleset *,
147			    const struct pf_rule *, struct pfioc_rule *);
148void			 pf_anchor_remove(struct pf_rule *);
149
150void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
151void			 pf_empty_pool(struct pf_palist *);
152#ifdef __FreeBSD__
153int			 pfioctl(struct cdev *, u_long, caddr_t, int, struct thread *);
154#else
155int			 pfioctl(struct cdev *, u_long, caddr_t, int, struct proc *);
156#endif
157#ifdef ALTQ
158int			 pf_begin_altq(u_int32_t *);
159int			 pf_rollback_altq(u_int32_t);
160int			 pf_commit_altq(u_int32_t);
161int			 pf_enable_altq(struct pf_altq *);
162int			 pf_disable_altq(struct pf_altq *);
163#endif /* ALTQ */
164int			 pf_begin_rules(u_int32_t *, int, const char *);
165int			 pf_rollback_rules(u_int32_t, int, char *);
166int			 pf_commit_rules(u_int32_t, int, char *);
167
168#ifdef __FreeBSD__
169extern struct callout	 pf_expire_to;
170#else
171extern struct timeout	 pf_expire_to;
172#endif
173
174struct pf_rule		 pf_default_rule;
175#ifdef ALTQ
176static int		 pf_altq_running;
177#endif
178
179#define	TAGID_MAX	 50000
180TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
181				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
182
183#if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
184#error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
185#endif
186static u_int16_t	 tagname2tag(struct pf_tags *, char *);
187static void		 tag2tagname(struct pf_tags *, u_int16_t, char *);
188static void		 tag_unref(struct pf_tags *, u_int16_t);
189int			 pf_rtlabel_add(struct pf_addr_wrap *);
190void			 pf_rtlabel_remove(struct pf_addr_wrap *);
191void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
192
193#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
194
195
196#ifdef __FreeBSD__
197static struct cdev	*pf_dev;
198
199/*
200 * XXX - These are new and need to be checked when moveing to a new version
201 */
202static void		 pf_clear_states(void);
203static int		 pf_clear_tables(void);
204static void		 pf_clear_srcnodes(void);
205/*
206 * XXX - These are new and need to be checked when moveing to a new version
207 */
208
209/*
210 * Wrapper functions for pfil(9) hooks
211 */
212static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
213		int dir, struct inpcb *inp);
214static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
215		int dir, struct inpcb *inp);
216#ifdef INET6
217static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
218		int dir, struct inpcb *inp);
219static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
220		int dir, struct inpcb *inp);
221#endif
222
223static int 		 hook_pf(void);
224static int 		 dehook_pf(void);
225static int 		 shutdown_pf(void);
226static int 		 pf_load(void);
227static int 		 pf_unload(void);
228
229static struct cdevsw pf_cdevsw = {
230	.d_ioctl =	pfioctl,
231	.d_name =	PF_NAME,
232	.d_version =	D_VERSION,
233};
234
235static volatile int pf_pfil_hooked = 0;
236struct mtx pf_task_mtx;
237pflog_packet_t *pflog_packet_ptr = NULL;
238
239void
240init_pf_mutex(void)
241{
242	mtx_init(&pf_task_mtx, "pf task mtx", NULL, MTX_DEF);
243}
244
245void
246destroy_pf_mutex(void)
247{
248	mtx_destroy(&pf_task_mtx);
249}
250
251void
252init_zone_var(void)
253{
254	pf_src_tree_pl = pf_rule_pl = NULL;
255	pf_state_pl = pf_altq_pl = pf_pooladdr_pl = NULL;
256	pf_frent_pl = pf_frag_pl = pf_cache_pl = pf_cent_pl = NULL;
257	pf_state_scrub_pl = NULL;
258	pfr_ktable_pl = pfr_kentry_pl = NULL;
259}
260
261void
262cleanup_pf_zone(void)
263{
264	UMA_DESTROY(pf_src_tree_pl);
265	UMA_DESTROY(pf_rule_pl);
266	UMA_DESTROY(pf_state_pl);
267	UMA_DESTROY(pf_altq_pl);
268	UMA_DESTROY(pf_pooladdr_pl);
269	UMA_DESTROY(pf_frent_pl);
270	UMA_DESTROY(pf_frag_pl);
271	UMA_DESTROY(pf_cache_pl);
272	UMA_DESTROY(pf_cent_pl);
273	UMA_DESTROY(pfr_ktable_pl);
274	UMA_DESTROY(pfr_kentry_pl);
275	UMA_DESTROY(pf_state_scrub_pl);
276	UMA_DESTROY(pfi_addr_pl);
277}
278
279int
280pfattach(void)
281{
282	u_int32_t *my_timeout = pf_default_rule.timeout;
283	int error = 1;
284
285	do {
286		UMA_CREATE(pf_src_tree_pl,struct pf_src_node, "pfsrctrpl");
287		UMA_CREATE(pf_rule_pl,	  struct pf_rule, "pfrulepl");
288		UMA_CREATE(pf_state_pl,	  struct pf_state, "pfstatepl");
289		UMA_CREATE(pf_altq_pl,	  struct pf_altq, "pfaltqpl");
290		UMA_CREATE(pf_pooladdr_pl, struct pf_pooladdr, "pfpooladdrpl");
291		UMA_CREATE(pfr_ktable_pl,  struct pfr_ktable, "pfrktable");
292		UMA_CREATE(pfr_kentry_pl,  struct pfr_kentry, "pfrkentry");
293		UMA_CREATE(pfr_kentry_pl2,  struct pfr_kentry, "pfrkentry2");
294		UMA_CREATE(pf_frent_pl,	  struct pf_frent, "pffrent");
295		UMA_CREATE(pf_frag_pl,	  struct pf_fragment, "pffrag");
296		UMA_CREATE(pf_cache_pl,	  struct pf_fragment, "pffrcache");
297		UMA_CREATE(pf_cent_pl,	  struct pf_frcache, "pffrcent");
298		UMA_CREATE(pf_state_scrub_pl, struct pf_state_scrub,
299		    "pfstatescrub");
300		UMA_CREATE(pfi_addr_pl, struct pfi_dynaddr, "pfiaddrpl");
301		error = 0;
302	} while(0);
303	if (error) {
304		cleanup_pf_zone();
305		return (error);
306	}
307	pfr_initialize();
308	pfi_initialize();
309	if ( (error = pf_osfp_initialize()) ) {
310		cleanup_pf_zone();
311		pf_osfp_cleanup();
312		return (error);
313	}
314
315	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
316	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
317	pf_pool_limits[PF_LIMIT_SRC_NODES].pp = pf_src_tree_pl;
318	pf_pool_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
319	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
320	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
321	uma_zone_set_max(pf_pool_limits[PF_LIMIT_STATES].pp,
322		pf_pool_limits[PF_LIMIT_STATES].limit);
323
324	RB_INIT(&tree_src_tracking);
325	RB_INIT(&pf_anchors);
326	pf_init_ruleset(&pf_main_ruleset);
327	TAILQ_INIT(&pf_altqs[0]);
328	TAILQ_INIT(&pf_altqs[1]);
329	TAILQ_INIT(&pf_pabuf);
330	pf_altqs_active = &pf_altqs[0];
331	pf_altqs_inactive = &pf_altqs[1];
332	TAILQ_INIT(&state_updates);
333
334	/* default rule should never be garbage collected */
335	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
336	pf_default_rule.action = PF_PASS;
337	pf_default_rule.nr = -1;
338
339	/* initialize default timeouts */
340	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
341	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
342	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
343	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
344	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
345	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
346	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
347	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
348	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
349	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
350	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
351	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
352	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
353	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
354	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
355	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
356	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
357	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
358
359	callout_init(&pf_expire_to, NET_CALLOUT_MPSAFE);
360	callout_reset(&pf_expire_to, my_timeout[PFTM_INTERVAL] * hz,
361	    pf_purge_timeout, &pf_expire_to);
362
363	pf_normalize_init();
364	bzero(&pf_status, sizeof(pf_status));
365	pf_pfil_hooked = 0;
366
367	/* XXX do our best to avoid a conflict */
368	pf_status.hostid = arc4random();
369
370	return (error);
371}
372#else /* !__FreeBSD__ */
373void
374pfattach(int num)
375{
376	u_int32_t *timeout = pf_default_rule.timeout;
377
378	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
379	    &pool_allocator_nointr);
380	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
381	    "pfsrctrpl", NULL);
382	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
383	    NULL);
384	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
385	    &pool_allocator_nointr);
386	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
387	    "pfpooladdrpl", &pool_allocator_nointr);
388	pfr_initialize();
389	pfi_initialize();
390	pf_osfp_initialize();
391
392	pool_sethardlimit(pf_pool_limits[PF_LIMIT_STATES].pp,
393	    pf_pool_limits[PF_LIMIT_STATES].limit, NULL, 0);
394
395	RB_INIT(&tree_src_tracking);
396	RB_INIT(&pf_anchors);
397	pf_init_ruleset(&pf_main_ruleset);
398	TAILQ_INIT(&pf_altqs[0]);
399	TAILQ_INIT(&pf_altqs[1]);
400	TAILQ_INIT(&pf_pabuf);
401	pf_altqs_active = &pf_altqs[0];
402	pf_altqs_inactive = &pf_altqs[1];
403	TAILQ_INIT(&state_updates);
404
405	/* default rule should never be garbage collected */
406	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
407	pf_default_rule.action = PF_PASS;
408	pf_default_rule.nr = -1;
409
410	/* initialize default timeouts */
411	timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
412	timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
413	timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
414	timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
415	timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
416	timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
417	timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
418	timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
419	timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
420	timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
421	timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
422	timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
423	timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
424	timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
425	timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
426	timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
427	timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
428	timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
429
430	timeout_set(&pf_expire_to, pf_purge_timeout, &pf_expire_to);
431	timeout_add(&pf_expire_to, timeout[PFTM_INTERVAL] * hz);
432
433	pf_normalize_init();
434	bzero(&pf_status, sizeof(pf_status));
435	pf_status.debug = PF_DEBUG_URGENT;
436
437	/* XXX do our best to avoid a conflict */
438	pf_status.hostid = arc4random();
439}
440
441int
442pfopen(struct cdev *dev, int flags, int fmt, struct proc *p)
443{
444	if (minor(dev) >= 1)
445		return (ENXIO);
446	return (0);
447}
448
449int
450pfclose(struct cdev *dev, int flags, int fmt, struct proc *p)
451{
452	if (minor(dev) >= 1)
453		return (ENXIO);
454	return (0);
455}
456#endif /* __FreeBSD__ */
457
458struct pf_pool *
459pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
460    u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
461    u_int8_t check_ticket)
462{
463	struct pf_ruleset	*ruleset;
464	struct pf_rule		*rule;
465	int			 rs_num;
466
467	ruleset = pf_find_ruleset(anchor);
468	if (ruleset == NULL)
469		return (NULL);
470	rs_num = pf_get_ruleset_number(rule_action);
471	if (rs_num >= PF_RULESET_MAX)
472		return (NULL);
473	if (active) {
474		if (check_ticket && ticket !=
475		    ruleset->rules[rs_num].active.ticket)
476			return (NULL);
477		if (r_last)
478			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
479			    pf_rulequeue);
480		else
481			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
482	} else {
483		if (check_ticket && ticket !=
484		    ruleset->rules[rs_num].inactive.ticket)
485			return (NULL);
486		if (r_last)
487			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
488			    pf_rulequeue);
489		else
490			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
491	}
492	if (!r_last) {
493		while ((rule != NULL) && (rule->nr != rule_number))
494			rule = TAILQ_NEXT(rule, entries);
495	}
496	if (rule == NULL)
497		return (NULL);
498
499	return (&rule->rpool);
500}
501
502int
503pf_get_ruleset_number(u_int8_t action)
504{
505	switch (action) {
506	case PF_SCRUB:
507	case PF_NOSCRUB:
508		return (PF_RULESET_SCRUB);
509		break;
510	case PF_PASS:
511	case PF_DROP:
512		return (PF_RULESET_FILTER);
513		break;
514	case PF_NAT:
515	case PF_NONAT:
516		return (PF_RULESET_NAT);
517		break;
518	case PF_BINAT:
519	case PF_NOBINAT:
520		return (PF_RULESET_BINAT);
521		break;
522	case PF_RDR:
523	case PF_NORDR:
524		return (PF_RULESET_RDR);
525		break;
526	default:
527		return (PF_RULESET_MAX);
528		break;
529	}
530}
531
532void
533pf_init_ruleset(struct pf_ruleset *ruleset)
534{
535	int	i;
536
537	memset(ruleset, 0, sizeof(struct pf_ruleset));
538	for (i = 0; i < PF_RULESET_MAX; i++) {
539		TAILQ_INIT(&ruleset->rules[i].queues[0]);
540		TAILQ_INIT(&ruleset->rules[i].queues[1]);
541		ruleset->rules[i].active.ptr = &ruleset->rules[i].queues[0];
542		ruleset->rules[i].inactive.ptr = &ruleset->rules[i].queues[1];
543	}
544}
545
546struct pf_anchor *
547pf_find_anchor(const char *path)
548{
549	static struct pf_anchor	 key;
550
551	memset(&key, 0, sizeof(key));
552	strlcpy(key.path, path, sizeof(key.path));
553	return (RB_FIND(pf_anchor_global, &pf_anchors, &key));
554}
555
556struct pf_ruleset *
557pf_find_ruleset(const char *path)
558{
559	struct pf_anchor	*anchor;
560
561	while (*path == '/')
562		path++;
563	if (!*path)
564		return (&pf_main_ruleset);
565	anchor = pf_find_anchor(path);
566	if (anchor == NULL)
567		return (NULL);
568	else
569		return (&anchor->ruleset);
570}
571
572struct pf_ruleset *
573pf_find_or_create_ruleset(const char *path)
574{
575	static char		 p[MAXPATHLEN];
576	char			*q = NULL, *r;	/* make the compiler happy */
577	struct pf_ruleset	*ruleset;
578	struct pf_anchor	*anchor = NULL, *dup, *parent = NULL;
579
580	while (*path == '/')
581		path++;
582	ruleset = pf_find_ruleset(path);
583	if (ruleset != NULL)
584		return (ruleset);
585	strlcpy(p, path, sizeof(p));
586#ifdef __FreeBSD__
587	while (parent == NULL && (q = rindex(p, '/')) != NULL) {
588#else
589	while (parent == NULL && (q = strrchr(p, '/')) != NULL) {
590#endif
591		*q = 0;
592		if ((ruleset = pf_find_ruleset(p)) != NULL) {
593			parent = ruleset->anchor;
594			break;
595		}
596	}
597	if (q == NULL)
598		q = p;
599	else
600		q++;
601	strlcpy(p, path, sizeof(p));
602	if (!*q)
603		return (NULL);
604#ifdef __FreeBSD__
605	while ((r = index(q, '/')) != NULL || *q) {
606#else
607	while ((r = strchr(q, '/')) != NULL || *q) {
608#endif
609		if (r != NULL)
610			*r = 0;
611		if (!*q || strlen(q) >= PF_ANCHOR_NAME_SIZE ||
612		    (parent != NULL && strlen(parent->path) >=
613		    MAXPATHLEN - PF_ANCHOR_NAME_SIZE - 1))
614			return (NULL);
615		anchor = (struct pf_anchor *)malloc(sizeof(*anchor), M_TEMP,
616		    M_NOWAIT);
617		if (anchor == NULL)
618			return (NULL);
619		memset(anchor, 0, sizeof(*anchor));
620		RB_INIT(&anchor->children);
621		strlcpy(anchor->name, q, sizeof(anchor->name));
622		if (parent != NULL) {
623			strlcpy(anchor->path, parent->path,
624			    sizeof(anchor->path));
625			strlcat(anchor->path, "/", sizeof(anchor->path));
626		}
627		strlcat(anchor->path, anchor->name, sizeof(anchor->path));
628		if ((dup = RB_INSERT(pf_anchor_global, &pf_anchors, anchor)) !=
629		    NULL) {
630			printf("pf_find_or_create_ruleset: RB_INSERT1 "
631			    "'%s' '%s' collides with '%s' '%s'\n",
632			    anchor->path, anchor->name, dup->path, dup->name);
633			free(anchor, M_TEMP);
634			return (NULL);
635		}
636		if (parent != NULL) {
637			anchor->parent = parent;
638			if ((dup = RB_INSERT(pf_anchor_node, &parent->children,
639			    anchor)) != NULL) {
640				printf("pf_find_or_create_ruleset: "
641				    "RB_INSERT2 '%s' '%s' collides with "
642				    "'%s' '%s'\n", anchor->path, anchor->name,
643				    dup->path, dup->name);
644				RB_REMOVE(pf_anchor_global, &pf_anchors,
645				    anchor);
646				free(anchor, M_TEMP);
647				return (NULL);
648			}
649		}
650		pf_init_ruleset(&anchor->ruleset);
651		anchor->ruleset.anchor = anchor;
652		parent = anchor;
653		if (r != NULL)
654			q = r + 1;
655		else
656			*q = 0;
657	}
658	return (&anchor->ruleset);
659}
660
661void
662pf_remove_if_empty_ruleset(struct pf_ruleset *ruleset)
663{
664	struct pf_anchor	*parent;
665	int			 i;
666
667	while (ruleset != NULL) {
668		if (ruleset == &pf_main_ruleset || ruleset->anchor == NULL ||
669		    !RB_EMPTY(&ruleset->anchor->children) ||
670		    ruleset->anchor->refcnt > 0 || ruleset->tables > 0 ||
671		    ruleset->topen)
672			return;
673		for (i = 0; i < PF_RULESET_MAX; ++i)
674			if (!TAILQ_EMPTY(ruleset->rules[i].active.ptr) ||
675			    !TAILQ_EMPTY(ruleset->rules[i].inactive.ptr) ||
676			    ruleset->rules[i].inactive.open)
677				return;
678		RB_REMOVE(pf_anchor_global, &pf_anchors, ruleset->anchor);
679		if ((parent = ruleset->anchor->parent) != NULL)
680			RB_REMOVE(pf_anchor_node, &parent->children,
681			    ruleset->anchor);
682		free(ruleset->anchor, M_TEMP);
683		if (parent == NULL)
684			return;
685		ruleset = &parent->ruleset;
686	}
687}
688
689int
690pf_anchor_setup(struct pf_rule *r, const struct pf_ruleset *s,
691    const char *name)
692{
693	static char		*p, path[MAXPATHLEN];
694	struct pf_ruleset	*ruleset;
695
696	r->anchor = NULL;
697	r->anchor_relative = 0;
698	r->anchor_wildcard = 0;
699	if (!name[0])
700		return (0);
701	if (name[0] == '/')
702		strlcpy(path, name + 1, sizeof(path));
703	else {
704		/* relative path */
705		r->anchor_relative = 1;
706		if (s->anchor == NULL || !s->anchor->path[0])
707			path[0] = 0;
708		else
709			strlcpy(path, s->anchor->path, sizeof(path));
710		while (name[0] == '.' && name[1] == '.' && name[2] == '/') {
711			if (!path[0]) {
712				printf("pf_anchor_setup: .. beyond root\n");
713				return (1);
714			}
715#ifdef __FreeBSD__
716			if ((p = rindex(path, '/')) != NULL)
717#else
718			if ((p = strrchr(path, '/')) != NULL)
719#endif
720				*p = 0;
721			else
722				path[0] = 0;
723			r->anchor_relative++;
724			name += 3;
725		}
726		if (path[0])
727			strlcat(path, "/", sizeof(path));
728		strlcat(path, name, sizeof(path));
729	}
730#ifdef __FreeBSD__
731	if ((p = rindex(path, '/')) != NULL && !strcmp(p, "/*")) {
732#else
733	if ((p = strrchr(path, '/')) != NULL && !strcmp(p, "/*")) {
734#endif
735		r->anchor_wildcard = 1;
736		*p = 0;
737	}
738	ruleset = pf_find_or_create_ruleset(path);
739	if (ruleset == NULL || ruleset->anchor == NULL) {
740		printf("pf_anchor_setup: ruleset\n");
741		return (1);
742	}
743	r->anchor = ruleset->anchor;
744	r->anchor->refcnt++;
745	return (0);
746}
747
748int
749pf_anchor_copyout(const struct pf_ruleset *rs, const struct pf_rule *r,
750    struct pfioc_rule *pr)
751{
752	pr->anchor_call[0] = 0;
753	if (r->anchor == NULL)
754		return (0);
755	if (!r->anchor_relative) {
756		strlcpy(pr->anchor_call, "/", sizeof(pr->anchor_call));
757		strlcat(pr->anchor_call, r->anchor->path,
758		    sizeof(pr->anchor_call));
759	} else {
760		char a[MAXPATHLEN], b[MAXPATHLEN], *p;
761		int i;
762
763		if (rs->anchor == NULL)
764			a[0] = 0;
765		else
766			strlcpy(a, rs->anchor->path, sizeof(a));
767		strlcpy(b, r->anchor->path, sizeof(b));
768		for (i = 1; i < r->anchor_relative; ++i) {
769#ifdef __FreeBSD__
770			if ((p = rindex(a, '/')) == NULL)
771#else
772			if ((p = strrchr(a, '/')) == NULL)
773#endif
774				p = a;
775			*p = 0;
776			strlcat(pr->anchor_call, "../",
777			    sizeof(pr->anchor_call));
778		}
779		if (strncmp(a, b, strlen(a))) {
780			printf("pf_anchor_copyout: '%s' '%s'\n", a, b);
781			return (1);
782		}
783		if (strlen(b) > strlen(a))
784			strlcat(pr->anchor_call, b + (a[0] ? strlen(a) + 1 : 0),
785			    sizeof(pr->anchor_call));
786	}
787	if (r->anchor_wildcard)
788		strlcat(pr->anchor_call, pr->anchor_call[0] ? "/*" : "*",
789		    sizeof(pr->anchor_call));
790	return (0);
791}
792
793void
794pf_anchor_remove(struct pf_rule *r)
795{
796	if (r->anchor == NULL)
797		return;
798	if (r->anchor->refcnt <= 0) {
799		printf("pf_anchor_remove: broken refcount");
800		r->anchor = NULL;
801		return;
802	}
803	if (!--r->anchor->refcnt)
804		pf_remove_if_empty_ruleset(&r->anchor->ruleset);
805	r->anchor = NULL;
806}
807
808void
809pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
810{
811	struct pf_pooladdr	*mv_pool_pa;
812
813	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
814		TAILQ_REMOVE(poola, mv_pool_pa, entries);
815		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
816	}
817}
818
819void
820pf_empty_pool(struct pf_palist *poola)
821{
822	struct pf_pooladdr	*empty_pool_pa;
823
824	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
825		pfi_dynaddr_remove(&empty_pool_pa->addr);
826		pf_tbladdr_remove(&empty_pool_pa->addr);
827		pfi_detach_rule(empty_pool_pa->kif);
828		TAILQ_REMOVE(poola, empty_pool_pa, entries);
829		pool_put(&pf_pooladdr_pl, empty_pool_pa);
830	}
831}
832
833void
834pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
835{
836	if (rulequeue != NULL) {
837		if (rule->states <= 0) {
838			/*
839			 * XXX - we need to remove the table *before* detaching
840			 * the rule to make sure the table code does not delete
841			 * the anchor under our feet.
842			 */
843			pf_tbladdr_remove(&rule->src.addr);
844			pf_tbladdr_remove(&rule->dst.addr);
845			if (rule->overload_tbl)
846				pfr_detach_table(rule->overload_tbl);
847		}
848		TAILQ_REMOVE(rulequeue, rule, entries);
849		rule->entries.tqe_prev = NULL;
850		rule->nr = -1;
851	}
852
853	if (rule->states > 0 || rule->src_nodes > 0 ||
854	    rule->entries.tqe_prev != NULL)
855		return;
856	pf_tag_unref(rule->tag);
857	pf_tag_unref(rule->match_tag);
858#ifdef ALTQ
859	if (rule->pqid != rule->qid)
860		pf_qid_unref(rule->pqid);
861	pf_qid_unref(rule->qid);
862#endif
863	pf_rtlabel_remove(&rule->src.addr);
864	pf_rtlabel_remove(&rule->dst.addr);
865	pfi_dynaddr_remove(&rule->src.addr);
866	pfi_dynaddr_remove(&rule->dst.addr);
867	if (rulequeue == NULL) {
868		pf_tbladdr_remove(&rule->src.addr);
869		pf_tbladdr_remove(&rule->dst.addr);
870		if (rule->overload_tbl)
871			pfr_detach_table(rule->overload_tbl);
872	}
873	pfi_detach_rule(rule->kif);
874	pf_anchor_remove(rule);
875	pf_empty_pool(&rule->rpool.list);
876	pool_put(&pf_rule_pl, rule);
877}
878
879static	u_int16_t
880tagname2tag(struct pf_tags *head, char *tagname)
881{
882	struct pf_tagname	*tag, *p = NULL;
883	u_int16_t		 new_tagid = 1;
884
885	TAILQ_FOREACH(tag, head, entries)
886		if (strcmp(tagname, tag->name) == 0) {
887			tag->ref++;
888			return (tag->tag);
889		}
890
891	/*
892	 * to avoid fragmentation, we do a linear search from the beginning
893	 * and take the first free slot we find. if there is none or the list
894	 * is empty, append a new entry at the end.
895	 */
896
897	/* new entry */
898	if (!TAILQ_EMPTY(head))
899		for (p = TAILQ_FIRST(head); p != NULL &&
900		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
901			new_tagid = p->tag + 1;
902
903	if (new_tagid > TAGID_MAX)
904		return (0);
905
906	/* allocate and fill new struct pf_tagname */
907	tag = (struct pf_tagname *)malloc(sizeof(struct pf_tagname),
908	    M_TEMP, M_NOWAIT);
909	if (tag == NULL)
910		return (0);
911	bzero(tag, sizeof(struct pf_tagname));
912	strlcpy(tag->name, tagname, sizeof(tag->name));
913	tag->tag = new_tagid;
914	tag->ref++;
915
916	if (p != NULL)	/* insert new entry before p */
917		TAILQ_INSERT_BEFORE(p, tag, entries);
918	else	/* either list empty or no free slot in between */
919		TAILQ_INSERT_TAIL(head, tag, entries);
920
921	return (tag->tag);
922}
923
924static	void
925tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
926{
927	struct pf_tagname	*tag;
928
929	TAILQ_FOREACH(tag, head, entries)
930		if (tag->tag == tagid) {
931			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
932			return;
933		}
934}
935
936static	void
937tag_unref(struct pf_tags *head, u_int16_t tag)
938{
939	struct pf_tagname	*p, *next;
940
941	if (tag == 0)
942		return;
943
944	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
945		next = TAILQ_NEXT(p, entries);
946		if (tag == p->tag) {
947			if (--p->ref == 0) {
948				TAILQ_REMOVE(head, p, entries);
949				free(p, M_TEMP);
950			}
951			break;
952		}
953	}
954}
955
956u_int16_t
957pf_tagname2tag(char *tagname)
958{
959	return (tagname2tag(&pf_tags, tagname));
960}
961
962void
963pf_tag2tagname(u_int16_t tagid, char *p)
964{
965	return (tag2tagname(&pf_tags, tagid, p));
966}
967
968void
969pf_tag_ref(u_int16_t tag)
970{
971	struct pf_tagname *t;
972
973	TAILQ_FOREACH(t, &pf_tags, entries)
974		if (t->tag == tag)
975			break;
976	if (t != NULL)
977		t->ref++;
978}
979
980void
981pf_tag_unref(u_int16_t tag)
982{
983	return (tag_unref(&pf_tags, tag));
984}
985
986int
987pf_rtlabel_add(struct pf_addr_wrap *a)
988{
989#ifdef __FreeBSD__
990	/* XXX_IMPORT: later */
991	return (0);
992#else
993	if (a->type == PF_ADDR_RTLABEL &&
994	    (a->v.rtlabel = rtlabel_name2id(a->v.rtlabelname)) == 0)
995		return (-1);
996	return (0);
997#endif
998}
999
1000void
1001pf_rtlabel_remove(struct pf_addr_wrap *a)
1002{
1003#ifdef __FreeBSD__
1004	/* XXX_IMPORT: later */
1005#else
1006	if (a->type == PF_ADDR_RTLABEL)
1007		rtlabel_unref(a->v.rtlabel);
1008#endif
1009}
1010
1011void
1012pf_rtlabel_copyout(struct pf_addr_wrap *a)
1013{
1014#ifdef __FreeBSD__
1015	/* XXX_IMPORT: later */
1016	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
1017		strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
1018#else
1019	const char	*name;
1020
1021	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel) {
1022		if ((name = rtlabel_id2name(a->v.rtlabel)) == NULL)
1023			strlcpy(a->v.rtlabelname, "?",
1024			    sizeof(a->v.rtlabelname));
1025		else
1026			strlcpy(a->v.rtlabelname, name,
1027			    sizeof(a->v.rtlabelname));
1028	}
1029#endif
1030}
1031
1032#ifdef ALTQ
1033u_int32_t
1034pf_qname2qid(char *qname)
1035{
1036	return ((u_int32_t)tagname2tag(&pf_qids, qname));
1037}
1038
1039void
1040pf_qid2qname(u_int32_t qid, char *p)
1041{
1042	return (tag2tagname(&pf_qids, (u_int16_t)qid, p));
1043}
1044
1045void
1046pf_qid_unref(u_int32_t qid)
1047{
1048	return (tag_unref(&pf_qids, (u_int16_t)qid));
1049}
1050
1051int
1052pf_begin_altq(u_int32_t *ticket)
1053{
1054	struct pf_altq	*altq;
1055	int		 error = 0;
1056
1057	/* Purge the old altq list */
1058	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
1059		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
1060		if (altq->qname[0] == 0) {
1061			/* detach and destroy the discipline */
1062			error = altq_remove(altq);
1063		} else
1064			pf_qid_unref(altq->qid);
1065		pool_put(&pf_altq_pl, altq);
1066	}
1067	if (error)
1068		return (error);
1069	*ticket = ++ticket_altqs_inactive;
1070	altqs_inactive_open = 1;
1071	return (0);
1072}
1073
1074int
1075pf_rollback_altq(u_int32_t ticket)
1076{
1077	struct pf_altq	*altq;
1078	int		 error = 0;
1079
1080	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
1081		return (0);
1082	/* Purge the old altq list */
1083	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
1084		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
1085		if (altq->qname[0] == 0) {
1086			/* detach and destroy the discipline */
1087			error = altq_remove(altq);
1088		} else
1089			pf_qid_unref(altq->qid);
1090		pool_put(&pf_altq_pl, altq);
1091	}
1092	altqs_inactive_open = 0;
1093	return (error);
1094}
1095
1096int
1097pf_commit_altq(u_int32_t ticket)
1098{
1099	struct pf_altqqueue	*old_altqs;
1100	struct pf_altq		*altq;
1101	int			 s, err, error = 0;
1102
1103	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
1104		return (EBUSY);
1105
1106	/* swap altqs, keep the old. */
1107	s = splsoftnet();
1108	old_altqs = pf_altqs_active;
1109	pf_altqs_active = pf_altqs_inactive;
1110	pf_altqs_inactive = old_altqs;
1111	ticket_altqs_active = ticket_altqs_inactive;
1112
1113	/* Attach new disciplines */
1114	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1115		if (altq->qname[0] == 0) {
1116			/* attach the discipline */
1117			error = altq_pfattach(altq);
1118			if (error == 0 && pf_altq_running)
1119				error = pf_enable_altq(altq);
1120			if (error != 0) {
1121				splx(s);
1122				return (error);
1123			}
1124		}
1125	}
1126
1127	/* Purge the old altq list */
1128	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
1129		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
1130		if (altq->qname[0] == 0) {
1131			/* detach and destroy the discipline */
1132			if (pf_altq_running)
1133				error = pf_disable_altq(altq);
1134			err = altq_pfdetach(altq);
1135			if (err != 0 && error == 0)
1136				error = err;
1137			err = altq_remove(altq);
1138			if (err != 0 && error == 0)
1139				error = err;
1140		} else
1141			pf_qid_unref(altq->qid);
1142		pool_put(&pf_altq_pl, altq);
1143	}
1144	splx(s);
1145
1146	altqs_inactive_open = 0;
1147	return (error);
1148}
1149
1150int
1151pf_enable_altq(struct pf_altq *altq)
1152{
1153	struct ifnet		*ifp;
1154	struct tb_profile	 tb;
1155	int			 s, error = 0;
1156
1157	if ((ifp = ifunit(altq->ifname)) == NULL)
1158		return (EINVAL);
1159
1160	if (ifp->if_snd.altq_type != ALTQT_NONE)
1161		error = altq_enable(&ifp->if_snd);
1162
1163	/* set tokenbucket regulator */
1164	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1165		tb.rate = altq->ifbandwidth;
1166		tb.depth = altq->tbrsize;
1167		s = splimp();
1168#ifdef __FreeBSD__
1169		PF_UNLOCK();
1170#endif
1171		error = tbr_set(&ifp->if_snd, &tb);
1172#ifdef __FreeBSD__
1173		PF_LOCK();
1174#endif
1175		splx(s);
1176	}
1177
1178	return (error);
1179}
1180
1181int
1182pf_disable_altq(struct pf_altq *altq)
1183{
1184	struct ifnet		*ifp;
1185	struct tb_profile	 tb;
1186	int			 s, error;
1187
1188	if ((ifp = ifunit(altq->ifname)) == NULL)
1189		return (EINVAL);
1190
1191	/*
1192	 * when the discipline is no longer referenced, it was overridden
1193	 * by a new one.  if so, just return.
1194	 */
1195	if (altq->altq_disc != ifp->if_snd.altq_disc)
1196		return (0);
1197
1198	error = altq_disable(&ifp->if_snd);
1199
1200	if (error == 0) {
1201		/* clear tokenbucket regulator */
1202		tb.rate = 0;
1203		s = splimp();
1204#ifdef __FreeBSD__
1205		PF_UNLOCK();
1206#endif
1207		error = tbr_set(&ifp->if_snd, &tb);
1208#ifdef __FreeBSD__
1209		PF_LOCK();
1210#endif
1211		splx(s);
1212	}
1213
1214	return (error);
1215}
1216#endif /* ALTQ */
1217
1218int
1219pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1220{
1221	struct pf_ruleset	*rs;
1222	struct pf_rule		*rule;
1223
1224	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1225		return (EINVAL);
1226	rs = pf_find_or_create_ruleset(anchor);
1227	if (rs == NULL)
1228		return (EINVAL);
1229	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL)
1230		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
1231	*ticket = ++rs->rules[rs_num].inactive.ticket;
1232	rs->rules[rs_num].inactive.open = 1;
1233	return (0);
1234}
1235
1236int
1237pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1238{
1239	struct pf_ruleset	*rs;
1240	struct pf_rule		*rule;
1241
1242	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1243		return (EINVAL);
1244	rs = pf_find_ruleset(anchor);
1245	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1246	    rs->rules[rs_num].inactive.ticket != ticket)
1247		return (0);
1248	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL)
1249		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
1250	rs->rules[rs_num].inactive.open = 0;
1251	return (0);
1252}
1253
1254int
1255pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1256{
1257	struct pf_ruleset	*rs;
1258	struct pf_rule		*rule;
1259	struct pf_rulequeue	*old_rules;
1260	int			 s;
1261
1262	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1263		return (EINVAL);
1264	rs = pf_find_ruleset(anchor);
1265	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1266	    ticket != rs->rules[rs_num].inactive.ticket)
1267		return (EBUSY);
1268
1269	/* Swap rules, keep the old. */
1270	s = splsoftnet();
1271	old_rules = rs->rules[rs_num].active.ptr;
1272	rs->rules[rs_num].active.ptr =
1273	    rs->rules[rs_num].inactive.ptr;
1274	rs->rules[rs_num].inactive.ptr = old_rules;
1275	rs->rules[rs_num].active.ticket =
1276	    rs->rules[rs_num].inactive.ticket;
1277	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1278
1279	/* Purge the old rule list. */
1280	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1281		pf_rm_rule(old_rules, rule);
1282	rs->rules[rs_num].inactive.open = 0;
1283	pf_remove_if_empty_ruleset(rs);
1284	splx(s);
1285	return (0);
1286}
1287
1288#ifdef __FreeBSD__
1289int
1290pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1291#else
1292int
1293pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct proc *p)
1294#endif
1295{
1296	struct pf_pooladdr	*pa = NULL;
1297	struct pf_pool		*pool = NULL;
1298#ifndef __FreeBSD__
1299	int			 s;
1300#endif
1301	int			 error = 0;
1302
1303	/* XXX keep in sync with switch() below */
1304#ifdef __FreeBSD__
1305	if (securelevel_gt(td->td_ucred, 2))
1306#else
1307	if (securelevel > 1)
1308#endif
1309		switch (cmd) {
1310		case DIOCGETRULES:
1311		case DIOCGETRULE:
1312		case DIOCGETADDRS:
1313		case DIOCGETADDR:
1314		case DIOCGETSTATE:
1315		case DIOCSETSTATUSIF:
1316		case DIOCGETSTATUS:
1317		case DIOCCLRSTATUS:
1318		case DIOCNATLOOK:
1319		case DIOCSETDEBUG:
1320		case DIOCGETSTATES:
1321		case DIOCGETTIMEOUT:
1322		case DIOCCLRRULECTRS:
1323		case DIOCGETLIMIT:
1324		case DIOCGETALTQS:
1325		case DIOCGETALTQ:
1326		case DIOCGETQSTATS:
1327		case DIOCGETRULESETS:
1328		case DIOCGETRULESET:
1329		case DIOCRGETTABLES:
1330		case DIOCRGETTSTATS:
1331		case DIOCRCLRTSTATS:
1332		case DIOCRCLRADDRS:
1333		case DIOCRADDADDRS:
1334		case DIOCRDELADDRS:
1335		case DIOCRSETADDRS:
1336		case DIOCRGETADDRS:
1337		case DIOCRGETASTATS:
1338		case DIOCRCLRASTATS:
1339		case DIOCRTSTADDRS:
1340		case DIOCOSFPGET:
1341		case DIOCGETSRCNODES:
1342		case DIOCCLRSRCNODES:
1343		case DIOCIGETIFACES:
1344		case DIOCICLRISTATS:
1345#ifdef __FreeBSD__
1346		case DIOCGIFSPEED:
1347#endif
1348		case DIOCSETIFFLAG:
1349		case DIOCCLRIFFLAG:
1350			break;
1351		case DIOCRCLRTABLES:
1352		case DIOCRADDTABLES:
1353		case DIOCRDELTABLES:
1354		case DIOCRSETTFLAGS:
1355			if (((struct pfioc_table *)addr)->pfrio_flags &
1356			    PFR_FLAG_DUMMY)
1357				break; /* dummy operation ok */
1358			return (EPERM);
1359		default:
1360			return (EPERM);
1361		}
1362
1363	if (!(flags & FWRITE))
1364		switch (cmd) {
1365		case DIOCGETRULES:
1366		case DIOCGETRULE:
1367		case DIOCGETADDRS:
1368		case DIOCGETADDR:
1369		case DIOCGETSTATE:
1370		case DIOCGETSTATUS:
1371		case DIOCGETSTATES:
1372		case DIOCGETTIMEOUT:
1373		case DIOCGETLIMIT:
1374		case DIOCGETALTQS:
1375		case DIOCGETALTQ:
1376		case DIOCGETQSTATS:
1377		case DIOCGETRULESETS:
1378		case DIOCGETRULESET:
1379		case DIOCRGETTABLES:
1380		case DIOCRGETTSTATS:
1381		case DIOCRGETADDRS:
1382		case DIOCRGETASTATS:
1383		case DIOCRTSTADDRS:
1384		case DIOCOSFPGET:
1385		case DIOCGETSRCNODES:
1386		case DIOCIGETIFACES:
1387#ifdef __FreeBSD__
1388		case DIOCGIFSPEED:
1389#endif
1390			break;
1391		case DIOCRCLRTABLES:
1392		case DIOCRADDTABLES:
1393		case DIOCRDELTABLES:
1394		case DIOCRCLRTSTATS:
1395		case DIOCRCLRADDRS:
1396		case DIOCRADDADDRS:
1397		case DIOCRDELADDRS:
1398		case DIOCRSETADDRS:
1399		case DIOCRSETTFLAGS:
1400			if (((struct pfioc_table *)addr)->pfrio_flags &
1401			    PFR_FLAG_DUMMY)
1402				break; /* dummy operation ok */
1403			return (EACCES);
1404		default:
1405			return (EACCES);
1406		}
1407
1408#ifdef __FreeBSD__
1409	PF_LOCK();
1410#else
1411	s = splsoftnet();
1412#endif
1413	switch (cmd) {
1414
1415	case DIOCSTART:
1416		if (pf_status.running)
1417			error = EEXIST;
1418		else {
1419#ifdef __FreeBSD__
1420			PF_UNLOCK();
1421			error = hook_pf();
1422			PF_LOCK();
1423			if (error) {
1424				DPFPRINTF(PF_DEBUG_MISC,
1425				    ("pf: pfil registeration fail\n"));
1426				break;
1427			}
1428#endif
1429			pf_status.running = 1;
1430			pf_status.since = time_second;
1431			if (pf_status.stateid == 0) {
1432				pf_status.stateid = time_second;
1433				pf_status.stateid = pf_status.stateid << 32;
1434			}
1435			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1436		}
1437		break;
1438
1439	case DIOCSTOP:
1440		if (!pf_status.running)
1441			error = ENOENT;
1442		else {
1443			pf_status.running = 0;
1444#ifdef __FreeBSD__
1445			PF_UNLOCK();
1446			error = dehook_pf();
1447			PF_LOCK();
1448			if (error) {
1449				pf_status.running = 1;
1450				DPFPRINTF(PF_DEBUG_MISC,
1451					("pf: pfil unregisteration failed\n"));
1452			}
1453#endif
1454			pf_status.since = time_second;
1455			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1456		}
1457		break;
1458
1459	case DIOCADDRULE: {
1460		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1461		struct pf_ruleset	*ruleset;
1462		struct pf_rule		*rule, *tail;
1463		struct pf_pooladdr	*pa;
1464		int			 rs_num;
1465
1466		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1467		ruleset = pf_find_ruleset(pr->anchor);
1468		if (ruleset == NULL) {
1469			error = EINVAL;
1470			break;
1471		}
1472		rs_num = pf_get_ruleset_number(pr->rule.action);
1473		if (rs_num >= PF_RULESET_MAX) {
1474			error = EINVAL;
1475			break;
1476		}
1477		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1478			error = EINVAL;
1479			break;
1480		}
1481		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1482#ifdef __FreeBSD__
1483			DPFPRINTF(PF_DEBUG_MISC,
1484			    ("ticket: %d != [%d]%d\n", pr->ticket, rs_num,
1485			    ruleset->rules[rs_num].inactive.ticket));
1486#endif
1487			error = EBUSY;
1488			break;
1489		}
1490		if (pr->pool_ticket != ticket_pabuf) {
1491#ifdef __FreeBSD__
1492			DPFPRINTF(PF_DEBUG_MISC,
1493			    ("pool_ticket: %d != %d\n", pr->pool_ticket,
1494			    ticket_pabuf));
1495#endif
1496			error = EBUSY;
1497			break;
1498		}
1499		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1500		if (rule == NULL) {
1501			error = ENOMEM;
1502			break;
1503		}
1504		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1505		rule->anchor = NULL;
1506		rule->kif = NULL;
1507		TAILQ_INIT(&rule->rpool.list);
1508		/* initialize refcounting */
1509		rule->states = 0;
1510		rule->src_nodes = 0;
1511		rule->entries.tqe_prev = NULL;
1512#ifndef INET
1513		if (rule->af == AF_INET) {
1514			pool_put(&pf_rule_pl, rule);
1515			error = EAFNOSUPPORT;
1516			break;
1517		}
1518#endif /* INET */
1519#ifndef INET6
1520		if (rule->af == AF_INET6) {
1521			pool_put(&pf_rule_pl, rule);
1522			error = EAFNOSUPPORT;
1523			break;
1524		}
1525#endif /* INET6 */
1526		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1527		    pf_rulequeue);
1528		if (tail)
1529			rule->nr = tail->nr + 1;
1530		else
1531			rule->nr = 0;
1532		if (rule->ifname[0]) {
1533			rule->kif = pfi_attach_rule(rule->ifname);
1534			if (rule->kif == NULL) {
1535				pool_put(&pf_rule_pl, rule);
1536				error = EINVAL;
1537				break;
1538			}
1539		}
1540
1541#ifdef ALTQ
1542		/* set queue IDs */
1543		if (rule->qname[0] != 0) {
1544			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1545				error = EBUSY;
1546			else if (rule->pqname[0] != 0) {
1547				if ((rule->pqid =
1548				    pf_qname2qid(rule->pqname)) == 0)
1549					error = EBUSY;
1550			} else
1551				rule->pqid = rule->qid;
1552		}
1553#endif
1554		if (rule->tagname[0])
1555			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1556				error = EBUSY;
1557		if (rule->match_tagname[0])
1558			if ((rule->match_tag =
1559			    pf_tagname2tag(rule->match_tagname)) == 0)
1560				error = EBUSY;
1561		if (rule->rt && !rule->direction)
1562			error = EINVAL;
1563		if (pf_rtlabel_add(&rule->src.addr) ||
1564		    pf_rtlabel_add(&rule->dst.addr))
1565			error = EBUSY;
1566		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1567			error = EINVAL;
1568		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1569			error = EINVAL;
1570		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1571			error = EINVAL;
1572		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1573			error = EINVAL;
1574		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1575			error = EINVAL;
1576		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1577			if (pf_tbladdr_setup(ruleset, &pa->addr))
1578				error = EINVAL;
1579
1580		if (rule->overload_tblname[0]) {
1581			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1582			    rule->overload_tblname)) == NULL)
1583				error = EINVAL;
1584			else
1585				rule->overload_tbl->pfrkt_flags |=
1586				    PFR_TFLAG_ACTIVE;
1587		}
1588
1589		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1590		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1591		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1592		    (rule->rt > PF_FASTROUTE)) &&
1593		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1594			error = EINVAL;
1595
1596		if (error) {
1597			pf_rm_rule(NULL, rule);
1598			break;
1599		}
1600		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1601		rule->evaluations = rule->packets = rule->bytes = 0;
1602		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1603		    rule, entries);
1604		break;
1605	}
1606
1607	case DIOCGETRULES: {
1608		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1609		struct pf_ruleset	*ruleset;
1610		struct pf_rule		*tail;
1611		int			 rs_num;
1612
1613		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1614		ruleset = pf_find_ruleset(pr->anchor);
1615		if (ruleset == NULL) {
1616			error = EINVAL;
1617			break;
1618		}
1619		rs_num = pf_get_ruleset_number(pr->rule.action);
1620		if (rs_num >= PF_RULESET_MAX) {
1621			error = EINVAL;
1622			break;
1623		}
1624		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1625		    pf_rulequeue);
1626		if (tail)
1627			pr->nr = tail->nr + 1;
1628		else
1629			pr->nr = 0;
1630		pr->ticket = ruleset->rules[rs_num].active.ticket;
1631		break;
1632	}
1633
1634	case DIOCGETRULE: {
1635		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1636		struct pf_ruleset	*ruleset;
1637		struct pf_rule		*rule;
1638		int			 rs_num, i;
1639
1640		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1641		ruleset = pf_find_ruleset(pr->anchor);
1642		if (ruleset == NULL) {
1643			error = EINVAL;
1644			break;
1645		}
1646		rs_num = pf_get_ruleset_number(pr->rule.action);
1647		if (rs_num >= PF_RULESET_MAX) {
1648			error = EINVAL;
1649			break;
1650		}
1651		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1652			error = EBUSY;
1653			break;
1654		}
1655		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1656		while ((rule != NULL) && (rule->nr != pr->nr))
1657			rule = TAILQ_NEXT(rule, entries);
1658		if (rule == NULL) {
1659			error = EBUSY;
1660			break;
1661		}
1662		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1663		if (pf_anchor_copyout(ruleset, rule, pr)) {
1664			error = EBUSY;
1665			break;
1666		}
1667		pfi_dynaddr_copyout(&pr->rule.src.addr);
1668		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1669		pf_tbladdr_copyout(&pr->rule.src.addr);
1670		pf_tbladdr_copyout(&pr->rule.dst.addr);
1671		pf_rtlabel_copyout(&pr->rule.src.addr);
1672		pf_rtlabel_copyout(&pr->rule.dst.addr);
1673		for (i = 0; i < PF_SKIP_COUNT; ++i)
1674			if (rule->skip[i].ptr == NULL)
1675				pr->rule.skip[i].nr = -1;
1676			else
1677				pr->rule.skip[i].nr =
1678				    rule->skip[i].ptr->nr;
1679		break;
1680	}
1681
1682	case DIOCCHANGERULE: {
1683		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1684		struct pf_ruleset	*ruleset;
1685		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1686		u_int32_t		 nr = 0;
1687		int			 rs_num;
1688
1689		if (!(pcr->action == PF_CHANGE_REMOVE ||
1690		    pcr->action == PF_CHANGE_GET_TICKET) &&
1691		    pcr->pool_ticket != ticket_pabuf) {
1692			error = EBUSY;
1693			break;
1694		}
1695
1696		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1697		    pcr->action > PF_CHANGE_GET_TICKET) {
1698			error = EINVAL;
1699			break;
1700		}
1701		ruleset = pf_find_ruleset(pcr->anchor);
1702		if (ruleset == NULL) {
1703			error = EINVAL;
1704			break;
1705		}
1706		rs_num = pf_get_ruleset_number(pcr->rule.action);
1707		if (rs_num >= PF_RULESET_MAX) {
1708			error = EINVAL;
1709			break;
1710		}
1711
1712		if (pcr->action == PF_CHANGE_GET_TICKET) {
1713			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1714			break;
1715		} else {
1716			if (pcr->ticket !=
1717			    ruleset->rules[rs_num].active.ticket) {
1718				error = EINVAL;
1719				break;
1720			}
1721			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1722				error = EINVAL;
1723				break;
1724			}
1725		}
1726
1727		if (pcr->action != PF_CHANGE_REMOVE) {
1728			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1729			if (newrule == NULL) {
1730				error = ENOMEM;
1731				break;
1732			}
1733			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1734			TAILQ_INIT(&newrule->rpool.list);
1735			/* initialize refcounting */
1736			newrule->states = 0;
1737			newrule->entries.tqe_prev = NULL;
1738#ifndef INET
1739			if (newrule->af == AF_INET) {
1740				pool_put(&pf_rule_pl, newrule);
1741				error = EAFNOSUPPORT;
1742				break;
1743			}
1744#endif /* INET */
1745#ifndef INET6
1746			if (newrule->af == AF_INET6) {
1747				pool_put(&pf_rule_pl, newrule);
1748				error = EAFNOSUPPORT;
1749				break;
1750			}
1751#endif /* INET6 */
1752			if (newrule->ifname[0]) {
1753				newrule->kif = pfi_attach_rule(newrule->ifname);
1754				if (newrule->kif == NULL) {
1755					pool_put(&pf_rule_pl, newrule);
1756					error = EINVAL;
1757					break;
1758				}
1759			} else
1760				newrule->kif = NULL;
1761
1762#ifdef ALTQ
1763			/* set queue IDs */
1764			if (newrule->qname[0] != 0) {
1765				if ((newrule->qid =
1766				    pf_qname2qid(newrule->qname)) == 0)
1767					error = EBUSY;
1768				else if (newrule->pqname[0] != 0) {
1769					if ((newrule->pqid =
1770					    pf_qname2qid(newrule->pqname)) == 0)
1771						error = EBUSY;
1772				} else
1773					newrule->pqid = newrule->qid;
1774			}
1775#endif /* ALTQ */
1776			if (newrule->tagname[0])
1777				if ((newrule->tag =
1778				    pf_tagname2tag(newrule->tagname)) == 0)
1779					error = EBUSY;
1780			if (newrule->match_tagname[0])
1781				if ((newrule->match_tag = pf_tagname2tag(
1782				    newrule->match_tagname)) == 0)
1783					error = EBUSY;
1784			if (newrule->rt && !newrule->direction)
1785				error = EINVAL;
1786			if (pf_rtlabel_add(&newrule->src.addr) ||
1787			    pf_rtlabel_add(&newrule->dst.addr))
1788				error = EBUSY;
1789			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1790				error = EINVAL;
1791			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1792				error = EINVAL;
1793			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1794				error = EINVAL;
1795			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1796				error = EINVAL;
1797			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1798				error = EINVAL;
1799			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1800				if (pf_tbladdr_setup(ruleset, &pa->addr))
1801					error = EINVAL;
1802
1803			if (newrule->overload_tblname[0]) {
1804				if ((newrule->overload_tbl = pfr_attach_table(
1805				    ruleset, newrule->overload_tblname)) ==
1806				    NULL)
1807					error = EINVAL;
1808				else
1809					newrule->overload_tbl->pfrkt_flags |=
1810					    PFR_TFLAG_ACTIVE;
1811			}
1812
1813			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1814			if (((((newrule->action == PF_NAT) ||
1815			    (newrule->action == PF_RDR) ||
1816			    (newrule->action == PF_BINAT) ||
1817			    (newrule->rt > PF_FASTROUTE)) &&
1818			    !pcr->anchor[0])) &&
1819			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1820				error = EINVAL;
1821
1822			if (error) {
1823				pf_rm_rule(NULL, newrule);
1824				break;
1825			}
1826			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1827			newrule->evaluations = newrule->packets = 0;
1828			newrule->bytes = 0;
1829		}
1830		pf_empty_pool(&pf_pabuf);
1831
1832		if (pcr->action == PF_CHANGE_ADD_HEAD)
1833			oldrule = TAILQ_FIRST(
1834			    ruleset->rules[rs_num].active.ptr);
1835		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1836			oldrule = TAILQ_LAST(
1837			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1838		else {
1839			oldrule = TAILQ_FIRST(
1840			    ruleset->rules[rs_num].active.ptr);
1841			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1842				oldrule = TAILQ_NEXT(oldrule, entries);
1843			if (oldrule == NULL) {
1844				if (newrule != NULL)
1845					pf_rm_rule(NULL, newrule);
1846				error = EINVAL;
1847				break;
1848			}
1849		}
1850
1851		if (pcr->action == PF_CHANGE_REMOVE)
1852			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1853		else {
1854			if (oldrule == NULL)
1855				TAILQ_INSERT_TAIL(
1856				    ruleset->rules[rs_num].active.ptr,
1857				    newrule, entries);
1858			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1859			    pcr->action == PF_CHANGE_ADD_BEFORE)
1860				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1861			else
1862				TAILQ_INSERT_AFTER(
1863				    ruleset->rules[rs_num].active.ptr,
1864				    oldrule, newrule, entries);
1865		}
1866
1867		nr = 0;
1868		TAILQ_FOREACH(oldrule,
1869		    ruleset->rules[rs_num].active.ptr, entries)
1870			oldrule->nr = nr++;
1871
1872		ruleset->rules[rs_num].active.ticket++;
1873
1874		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1875		pf_remove_if_empty_ruleset(ruleset);
1876
1877		break;
1878	}
1879
1880	case DIOCCLRSTATES: {
1881		struct pf_state		*state;
1882		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1883		int			 killed = 0;
1884
1885		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1886			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1887			    state->u.s.kif->pfik_name)) {
1888				state->timeout = PFTM_PURGE;
1889#if NPFSYNC
1890				/* don't send out individual delete messages */
1891				state->sync_flags = PFSTATE_NOSYNC;
1892#endif
1893				killed++;
1894			}
1895		}
1896		pf_purge_expired_states();
1897		pf_status.states = 0;
1898		psk->psk_af = killed;
1899#if NPFSYNC
1900		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1901#endif
1902		break;
1903	}
1904
1905	case DIOCKILLSTATES: {
1906		struct pf_state		*state;
1907		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1908		int			 killed = 0;
1909
1910		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1911			if ((!psk->psk_af || state->af == psk->psk_af)
1912			    && (!psk->psk_proto || psk->psk_proto ==
1913			    state->proto) &&
1914			    PF_MATCHA(psk->psk_src.neg,
1915			    &psk->psk_src.addr.v.a.addr,
1916			    &psk->psk_src.addr.v.a.mask,
1917			    &state->lan.addr, state->af) &&
1918			    PF_MATCHA(psk->psk_dst.neg,
1919			    &psk->psk_dst.addr.v.a.addr,
1920			    &psk->psk_dst.addr.v.a.mask,
1921			    &state->ext.addr, state->af) &&
1922			    (psk->psk_src.port_op == 0 ||
1923			    pf_match_port(psk->psk_src.port_op,
1924			    psk->psk_src.port[0], psk->psk_src.port[1],
1925			    state->lan.port)) &&
1926			    (psk->psk_dst.port_op == 0 ||
1927			    pf_match_port(psk->psk_dst.port_op,
1928			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1929			    state->ext.port)) &&
1930			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1931			    state->u.s.kif->pfik_name))) {
1932				state->timeout = PFTM_PURGE;
1933				killed++;
1934			}
1935		}
1936		pf_purge_expired_states();
1937		psk->psk_af = killed;
1938		break;
1939	}
1940
1941	case DIOCADDSTATE: {
1942		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1943		struct pf_state		*state;
1944		struct pfi_kif		*kif;
1945
1946		if (ps->state.timeout >= PFTM_MAX &&
1947		    ps->state.timeout != PFTM_UNTIL_PACKET) {
1948			error = EINVAL;
1949			break;
1950		}
1951		state = pool_get(&pf_state_pl, PR_NOWAIT);
1952		if (state == NULL) {
1953			error = ENOMEM;
1954			break;
1955		}
1956		kif = pfi_lookup_create(ps->state.u.ifname);
1957		if (kif == NULL) {
1958			pool_put(&pf_state_pl, state);
1959			error = ENOENT;
1960			break;
1961		}
1962		bcopy(&ps->state, state, sizeof(struct pf_state));
1963		bzero(&state->u, sizeof(state->u));
1964		state->rule.ptr = &pf_default_rule;
1965		state->nat_rule.ptr = NULL;
1966		state->anchor.ptr = NULL;
1967		state->rt_kif = NULL;
1968		state->creation = time_second;
1969		state->pfsync_time = 0;
1970		state->packets[0] = state->packets[1] = 0;
1971		state->bytes[0] = state->bytes[1] = 0;
1972
1973		if (pf_insert_state(kif, state)) {
1974			pfi_maybe_destroy(kif);
1975			pool_put(&pf_state_pl, state);
1976			error = ENOMEM;
1977		}
1978		break;
1979	}
1980
1981	case DIOCGETSTATE: {
1982		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1983		struct pf_state		*state;
1984		u_int32_t		 nr;
1985
1986		nr = 0;
1987		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1988			if (nr >= ps->nr)
1989				break;
1990			nr++;
1991		}
1992		if (state == NULL) {
1993			error = EBUSY;
1994			break;
1995		}
1996		bcopy(state, &ps->state, sizeof(struct pf_state));
1997		ps->state.rule.nr = state->rule.ptr->nr;
1998		ps->state.nat_rule.nr = (state->nat_rule.ptr == NULL) ?
1999		    -1 : state->nat_rule.ptr->nr;
2000		ps->state.anchor.nr = (state->anchor.ptr == NULL) ?
2001		    -1 : state->anchor.ptr->nr;
2002		ps->state.expire = pf_state_expires(state);
2003		if (ps->state.expire > time_second)
2004			ps->state.expire -= time_second;
2005		else
2006			ps->state.expire = 0;
2007		break;
2008	}
2009
2010	case DIOCGETSTATES: {
2011		struct pfioc_states	*ps = (struct pfioc_states *)addr;
2012		struct pf_state		*state;
2013		struct pf_state		*p, pstore;
2014		struct pfi_kif		*kif;
2015		u_int32_t		 nr = 0;
2016		int			 space = ps->ps_len;
2017
2018		if (space == 0) {
2019			TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states)
2020				nr += kif->pfik_states;
2021			ps->ps_len = sizeof(struct pf_state) * nr;
2022			break;
2023		}
2024
2025		p = ps->ps_states;
2026		TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states)
2027			RB_FOREACH(state, pf_state_tree_ext_gwy,
2028			    &kif->pfik_ext_gwy) {
2029				int	secs = time_second;
2030
2031				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
2032					break;
2033
2034				bcopy(state, &pstore, sizeof(pstore));
2035				strlcpy(pstore.u.ifname, kif->pfik_name,
2036				    sizeof(pstore.u.ifname));
2037				pstore.rule.nr = state->rule.ptr->nr;
2038				pstore.nat_rule.nr = (state->nat_rule.ptr ==
2039				    NULL) ? -1 : state->nat_rule.ptr->nr;
2040				pstore.anchor.nr = (state->anchor.ptr ==
2041				    NULL) ? -1 : state->anchor.ptr->nr;
2042				pstore.creation = secs - pstore.creation;
2043				pstore.expire = pf_state_expires(state);
2044				if (pstore.expire > secs)
2045					pstore.expire -= secs;
2046				else
2047					pstore.expire = 0;
2048#ifdef __FreeBSD__
2049				PF_COPYOUT(&pstore, p, sizeof(*p), error);
2050#else
2051				error = copyout(&pstore, p, sizeof(*p));
2052#endif
2053				if (error)
2054					goto fail;
2055				p++;
2056				nr++;
2057			}
2058		ps->ps_len = sizeof(struct pf_state) * nr;
2059		break;
2060	}
2061
2062	case DIOCGETSTATUS: {
2063		struct pf_status *s = (struct pf_status *)addr;
2064		bcopy(&pf_status, s, sizeof(struct pf_status));
2065		pfi_fill_oldstatus(s);
2066		break;
2067	}
2068
2069	case DIOCSETSTATUSIF: {
2070		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2071
2072		if (pi->ifname[0] == 0) {
2073			bzero(pf_status.ifname, IFNAMSIZ);
2074			break;
2075		}
2076		if (ifunit(pi->ifname) == NULL) {
2077			error = EINVAL;
2078			break;
2079		}
2080		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
2081		break;
2082	}
2083
2084	case DIOCCLRSTATUS: {
2085		bzero(pf_status.counters, sizeof(pf_status.counters));
2086		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
2087		bzero(pf_status.scounters, sizeof(pf_status.scounters));
2088		if (*pf_status.ifname)
2089			pfi_clr_istats(pf_status.ifname, NULL,
2090			    PFI_FLAG_INSTANCE);
2091		break;
2092	}
2093
2094	case DIOCNATLOOK: {
2095		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2096		struct pf_state		*state;
2097		struct pf_state		 key;
2098		int			 m = 0, direction = pnl->direction;
2099
2100		key.af = pnl->af;
2101		key.proto = pnl->proto;
2102
2103		if (!pnl->proto ||
2104		    PF_AZERO(&pnl->saddr, pnl->af) ||
2105		    PF_AZERO(&pnl->daddr, pnl->af) ||
2106		    !pnl->dport || !pnl->sport)
2107			error = EINVAL;
2108		else {
2109			/*
2110			 * userland gives us source and dest of connection,
2111			 * reverse the lookup so we ask for what happens with
2112			 * the return traffic, enabling us to find it in the
2113			 * state tree.
2114			 */
2115			if (direction == PF_IN) {
2116				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
2117				key.ext.port = pnl->dport;
2118				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
2119				key.gwy.port = pnl->sport;
2120				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
2121			} else {
2122				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
2123				key.lan.port = pnl->dport;
2124				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
2125				key.ext.port = pnl->sport;
2126				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
2127			}
2128			if (m > 1)
2129				error = E2BIG;	/* more than one state */
2130			else if (state != NULL) {
2131				if (direction == PF_IN) {
2132					PF_ACPY(&pnl->rsaddr, &state->lan.addr,
2133					    state->af);
2134					pnl->rsport = state->lan.port;
2135					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
2136					    pnl->af);
2137					pnl->rdport = pnl->dport;
2138				} else {
2139					PF_ACPY(&pnl->rdaddr, &state->gwy.addr,
2140					    state->af);
2141					pnl->rdport = state->gwy.port;
2142					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
2143					    pnl->af);
2144					pnl->rsport = pnl->sport;
2145				}
2146			} else
2147				error = ENOENT;
2148		}
2149		break;
2150	}
2151
2152	case DIOCSETTIMEOUT: {
2153		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2154		int		 old;
2155
2156		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2157		    pt->seconds < 0) {
2158			error = EINVAL;
2159			goto fail;
2160		}
2161		old = pf_default_rule.timeout[pt->timeout];
2162		pf_default_rule.timeout[pt->timeout] = pt->seconds;
2163		pt->seconds = old;
2164		break;
2165	}
2166
2167	case DIOCGETTIMEOUT: {
2168		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2169
2170		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2171			error = EINVAL;
2172			goto fail;
2173		}
2174		pt->seconds = pf_default_rule.timeout[pt->timeout];
2175		break;
2176	}
2177
2178	case DIOCGETLIMIT: {
2179		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2180
2181		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2182			error = EINVAL;
2183			goto fail;
2184		}
2185		pl->limit = pf_pool_limits[pl->index].limit;
2186		break;
2187	}
2188
2189	case DIOCSETLIMIT: {
2190		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2191		int			 old_limit;
2192
2193		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2194		    pf_pool_limits[pl->index].pp == NULL) {
2195			error = EINVAL;
2196			goto fail;
2197		}
2198#ifdef __FreeBSD__
2199		uma_zone_set_max(pf_pool_limits[pl->index].pp, pl->limit);
2200#else
2201		if (pool_sethardlimit(pf_pool_limits[pl->index].pp,
2202		    pl->limit, NULL, 0) != 0) {
2203			error = EBUSY;
2204			goto fail;
2205		}
2206#endif
2207		old_limit = pf_pool_limits[pl->index].limit;
2208		pf_pool_limits[pl->index].limit = pl->limit;
2209		pl->limit = old_limit;
2210		break;
2211	}
2212
2213	case DIOCSETDEBUG: {
2214		u_int32_t	*level = (u_int32_t *)addr;
2215
2216		pf_status.debug = *level;
2217		break;
2218	}
2219
2220	case DIOCCLRRULECTRS: {
2221		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2222		struct pf_rule		*rule;
2223
2224		TAILQ_FOREACH(rule,
2225		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries)
2226			rule->evaluations = rule->packets =
2227			    rule->bytes = 0;
2228		break;
2229	}
2230
2231#ifdef __FreeBSD__
2232	case DIOCGIFSPEED: {
2233		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
2234		struct pf_ifspeed	ps;
2235		struct ifnet		*ifp;
2236
2237		if (psp->ifname[0] != 0) {
2238			/* Can we completely trust user-land? */
2239			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2240			ifp = ifunit(ps.ifname);
2241			if (ifp != NULL)
2242				psp->baudrate = ifp->if_baudrate;
2243			else
2244				error = EINVAL;
2245		} else
2246			error = EINVAL;
2247		break;
2248	}
2249#endif /* __FreeBSD__ */
2250
2251#ifdef ALTQ
2252	case DIOCSTARTALTQ: {
2253		struct pf_altq		*altq;
2254
2255		/* enable all altq interfaces on active list */
2256		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2257			if (altq->qname[0] == 0) {
2258				error = pf_enable_altq(altq);
2259				if (error != 0)
2260					break;
2261			}
2262		}
2263		if (error == 0)
2264			pf_altq_running = 1;
2265		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2266		break;
2267	}
2268
2269	case DIOCSTOPALTQ: {
2270		struct pf_altq		*altq;
2271
2272		/* disable all altq interfaces on active list */
2273		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2274			if (altq->qname[0] == 0) {
2275				error = pf_disable_altq(altq);
2276				if (error != 0)
2277					break;
2278			}
2279		}
2280		if (error == 0)
2281			pf_altq_running = 0;
2282		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2283		break;
2284	}
2285
2286	case DIOCADDALTQ: {
2287		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2288		struct pf_altq		*altq, *a;
2289
2290		if (pa->ticket != ticket_altqs_inactive) {
2291			error = EBUSY;
2292			break;
2293		}
2294		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2295		if (altq == NULL) {
2296			error = ENOMEM;
2297			break;
2298		}
2299		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2300
2301		/*
2302		 * if this is for a queue, find the discipline and
2303		 * copy the necessary fields
2304		 */
2305		if (altq->qname[0] != 0) {
2306			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2307				error = EBUSY;
2308				pool_put(&pf_altq_pl, altq);
2309				break;
2310			}
2311			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2312				if (strncmp(a->ifname, altq->ifname,
2313				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2314					altq->altq_disc = a->altq_disc;
2315					break;
2316				}
2317			}
2318		}
2319
2320#ifdef __FreeBSD__
2321		PF_UNLOCK();
2322#endif
2323		error = altq_add(altq);
2324#ifdef __FreeBSD__
2325		PF_LOCK();
2326#endif
2327		if (error) {
2328			pool_put(&pf_altq_pl, altq);
2329			break;
2330		}
2331
2332		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2333		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2334		break;
2335	}
2336
2337	case DIOCGETALTQS: {
2338		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2339		struct pf_altq		*altq;
2340
2341		pa->nr = 0;
2342		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2343			pa->nr++;
2344		pa->ticket = ticket_altqs_active;
2345		break;
2346	}
2347
2348	case DIOCGETALTQ: {
2349		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2350		struct pf_altq		*altq;
2351		u_int32_t		 nr;
2352
2353		if (pa->ticket != ticket_altqs_active) {
2354			error = EBUSY;
2355			break;
2356		}
2357		nr = 0;
2358		altq = TAILQ_FIRST(pf_altqs_active);
2359		while ((altq != NULL) && (nr < pa->nr)) {
2360			altq = TAILQ_NEXT(altq, entries);
2361			nr++;
2362		}
2363		if (altq == NULL) {
2364			error = EBUSY;
2365			break;
2366		}
2367		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2368		break;
2369	}
2370
2371	case DIOCCHANGEALTQ:
2372		/* CHANGEALTQ not supported yet! */
2373		error = ENODEV;
2374		break;
2375
2376	case DIOCGETQSTATS: {
2377		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2378		struct pf_altq		*altq;
2379		u_int32_t		 nr;
2380		int			 nbytes;
2381
2382		if (pq->ticket != ticket_altqs_active) {
2383			error = EBUSY;
2384			break;
2385		}
2386		nbytes = pq->nbytes;
2387		nr = 0;
2388		altq = TAILQ_FIRST(pf_altqs_active);
2389		while ((altq != NULL) && (nr < pq->nr)) {
2390			altq = TAILQ_NEXT(altq, entries);
2391			nr++;
2392		}
2393		if (altq == NULL) {
2394			error = EBUSY;
2395			break;
2396		}
2397#ifdef __FreeBSD__
2398		PF_UNLOCK();
2399#endif
2400		error = altq_getqstats(altq, pq->buf, &nbytes);
2401#ifdef __FreeBSD__
2402		PF_LOCK();
2403#endif
2404		if (error == 0) {
2405			pq->scheduler = altq->scheduler;
2406			pq->nbytes = nbytes;
2407		}
2408		break;
2409	}
2410#endif /* ALTQ */
2411
2412	case DIOCBEGINADDRS: {
2413		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2414
2415		pf_empty_pool(&pf_pabuf);
2416		pp->ticket = ++ticket_pabuf;
2417		break;
2418	}
2419
2420	case DIOCADDADDR: {
2421		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2422
2423		if (pp->ticket != ticket_pabuf) {
2424			error = EBUSY;
2425			break;
2426		}
2427#ifndef INET
2428		if (pp->af == AF_INET) {
2429			error = EAFNOSUPPORT;
2430			break;
2431		}
2432#endif /* INET */
2433#ifndef INET6
2434		if (pp->af == AF_INET6) {
2435			error = EAFNOSUPPORT;
2436			break;
2437		}
2438#endif /* INET6 */
2439		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2440		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2441		    pp->addr.addr.type != PF_ADDR_TABLE) {
2442			error = EINVAL;
2443			break;
2444		}
2445		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2446		if (pa == NULL) {
2447			error = ENOMEM;
2448			break;
2449		}
2450		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2451		if (pa->ifname[0]) {
2452			pa->kif = pfi_attach_rule(pa->ifname);
2453			if (pa->kif == NULL) {
2454				pool_put(&pf_pooladdr_pl, pa);
2455				error = EINVAL;
2456				break;
2457			}
2458		}
2459		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2460			pfi_dynaddr_remove(&pa->addr);
2461			pfi_detach_rule(pa->kif);
2462			pool_put(&pf_pooladdr_pl, pa);
2463			error = EINVAL;
2464			break;
2465		}
2466		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2467		break;
2468	}
2469
2470	case DIOCGETADDRS: {
2471		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2472
2473		pp->nr = 0;
2474		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2475		    pp->r_num, 0, 1, 0);
2476		if (pool == NULL) {
2477			error = EBUSY;
2478			break;
2479		}
2480		TAILQ_FOREACH(pa, &pool->list, entries)
2481			pp->nr++;
2482		break;
2483	}
2484
2485	case DIOCGETADDR: {
2486		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2487		u_int32_t		 nr = 0;
2488
2489		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2490		    pp->r_num, 0, 1, 1);
2491		if (pool == NULL) {
2492			error = EBUSY;
2493			break;
2494		}
2495		pa = TAILQ_FIRST(&pool->list);
2496		while ((pa != NULL) && (nr < pp->nr)) {
2497			pa = TAILQ_NEXT(pa, entries);
2498			nr++;
2499		}
2500		if (pa == NULL) {
2501			error = EBUSY;
2502			break;
2503		}
2504		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2505		pfi_dynaddr_copyout(&pp->addr.addr);
2506		pf_tbladdr_copyout(&pp->addr.addr);
2507		pf_rtlabel_copyout(&pp->addr.addr);
2508		break;
2509	}
2510
2511	case DIOCCHANGEADDR: {
2512		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2513		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2514		struct pf_ruleset	*ruleset;
2515
2516		if (pca->action < PF_CHANGE_ADD_HEAD ||
2517		    pca->action > PF_CHANGE_REMOVE) {
2518			error = EINVAL;
2519			break;
2520		}
2521		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2522		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2523		    pca->addr.addr.type != PF_ADDR_TABLE) {
2524			error = EINVAL;
2525			break;
2526		}
2527
2528		ruleset = pf_find_ruleset(pca->anchor);
2529		if (ruleset == NULL) {
2530			error = EBUSY;
2531			break;
2532		}
2533		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2534		    pca->r_num, pca->r_last, 1, 1);
2535		if (pool == NULL) {
2536			error = EBUSY;
2537			break;
2538		}
2539		if (pca->action != PF_CHANGE_REMOVE) {
2540			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2541			if (newpa == NULL) {
2542				error = ENOMEM;
2543				break;
2544			}
2545			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2546#ifndef INET
2547			if (pca->af == AF_INET) {
2548				pool_put(&pf_pooladdr_pl, newpa);
2549				error = EAFNOSUPPORT;
2550				break;
2551			}
2552#endif /* INET */
2553#ifndef INET6
2554			if (pca->af == AF_INET6) {
2555				pool_put(&pf_pooladdr_pl, newpa);
2556				error = EAFNOSUPPORT;
2557				break;
2558			}
2559#endif /* INET6 */
2560			if (newpa->ifname[0]) {
2561				newpa->kif = pfi_attach_rule(newpa->ifname);
2562				if (newpa->kif == NULL) {
2563					pool_put(&pf_pooladdr_pl, newpa);
2564					error = EINVAL;
2565					break;
2566				}
2567			} else
2568				newpa->kif = NULL;
2569			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2570			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2571				pfi_dynaddr_remove(&newpa->addr);
2572				pfi_detach_rule(newpa->kif);
2573				pool_put(&pf_pooladdr_pl, newpa);
2574				error = EINVAL;
2575				break;
2576			}
2577		}
2578
2579		if (pca->action == PF_CHANGE_ADD_HEAD)
2580			oldpa = TAILQ_FIRST(&pool->list);
2581		else if (pca->action == PF_CHANGE_ADD_TAIL)
2582			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2583		else {
2584			int	i = 0;
2585
2586			oldpa = TAILQ_FIRST(&pool->list);
2587			while ((oldpa != NULL) && (i < pca->nr)) {
2588				oldpa = TAILQ_NEXT(oldpa, entries);
2589				i++;
2590			}
2591			if (oldpa == NULL) {
2592				error = EINVAL;
2593				break;
2594			}
2595		}
2596
2597		if (pca->action == PF_CHANGE_REMOVE) {
2598			TAILQ_REMOVE(&pool->list, oldpa, entries);
2599			pfi_dynaddr_remove(&oldpa->addr);
2600			pf_tbladdr_remove(&oldpa->addr);
2601			pfi_detach_rule(oldpa->kif);
2602			pool_put(&pf_pooladdr_pl, oldpa);
2603		} else {
2604			if (oldpa == NULL)
2605				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2606			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2607			    pca->action == PF_CHANGE_ADD_BEFORE)
2608				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2609			else
2610				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2611				    newpa, entries);
2612		}
2613
2614		pool->cur = TAILQ_FIRST(&pool->list);
2615		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2616		    pca->af);
2617		break;
2618	}
2619
2620	case DIOCGETRULESETS: {
2621		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2622		struct pf_ruleset	*ruleset;
2623		struct pf_anchor	*anchor;
2624
2625		pr->path[sizeof(pr->path) - 1] = 0;
2626		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2627			error = EINVAL;
2628			break;
2629		}
2630		pr->nr = 0;
2631		if (ruleset->anchor == NULL) {
2632			/* XXX kludge for pf_main_ruleset */
2633			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2634				if (anchor->parent == NULL)
2635					pr->nr++;
2636		} else {
2637			RB_FOREACH(anchor, pf_anchor_node,
2638			    &ruleset->anchor->children)
2639				pr->nr++;
2640		}
2641		break;
2642	}
2643
2644	case DIOCGETRULESET: {
2645		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2646		struct pf_ruleset	*ruleset;
2647		struct pf_anchor	*anchor;
2648		u_int32_t		 nr = 0;
2649
2650		pr->path[sizeof(pr->path) - 1] = 0;
2651		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2652			error = EINVAL;
2653			break;
2654		}
2655		pr->name[0] = 0;
2656		if (ruleset->anchor == NULL) {
2657			/* XXX kludge for pf_main_ruleset */
2658			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2659				if (anchor->parent == NULL && nr++ == pr->nr) {
2660					strlcpy(pr->name, anchor->name,
2661					    sizeof(pr->name));
2662					break;
2663				}
2664		} else {
2665			RB_FOREACH(anchor, pf_anchor_node,
2666			    &ruleset->anchor->children)
2667				if (nr++ == pr->nr) {
2668					strlcpy(pr->name, anchor->name,
2669					    sizeof(pr->name));
2670					break;
2671				}
2672		}
2673		if (!pr->name[0])
2674			error = EBUSY;
2675		break;
2676	}
2677
2678	case DIOCRCLRTABLES: {
2679		struct pfioc_table *io = (struct pfioc_table *)addr;
2680
2681		if (io->pfrio_esize != 0) {
2682			error = ENODEV;
2683			break;
2684		}
2685		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2686		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2687		break;
2688	}
2689
2690	case DIOCRADDTABLES: {
2691		struct pfioc_table *io = (struct pfioc_table *)addr;
2692
2693		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2694			error = ENODEV;
2695			break;
2696		}
2697		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2698		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2699		break;
2700	}
2701
2702	case DIOCRDELTABLES: {
2703		struct pfioc_table *io = (struct pfioc_table *)addr;
2704
2705		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2706			error = ENODEV;
2707			break;
2708		}
2709		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2710		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2711		break;
2712	}
2713
2714	case DIOCRGETTABLES: {
2715		struct pfioc_table *io = (struct pfioc_table *)addr;
2716
2717		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2718			error = ENODEV;
2719			break;
2720		}
2721		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2722		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2723		break;
2724	}
2725
2726	case DIOCRGETTSTATS: {
2727		struct pfioc_table *io = (struct pfioc_table *)addr;
2728
2729		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2730			error = ENODEV;
2731			break;
2732		}
2733		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2734		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2735		break;
2736	}
2737
2738	case DIOCRCLRTSTATS: {
2739		struct pfioc_table *io = (struct pfioc_table *)addr;
2740
2741		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2742			error = ENODEV;
2743			break;
2744		}
2745		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2746		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2747		break;
2748	}
2749
2750	case DIOCRSETTFLAGS: {
2751		struct pfioc_table *io = (struct pfioc_table *)addr;
2752
2753		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2754			error = ENODEV;
2755			break;
2756		}
2757		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2758		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2759		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2760		break;
2761	}
2762
2763	case DIOCRCLRADDRS: {
2764		struct pfioc_table *io = (struct pfioc_table *)addr;
2765
2766		if (io->pfrio_esize != 0) {
2767			error = ENODEV;
2768			break;
2769		}
2770		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2771		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2772		break;
2773	}
2774
2775	case DIOCRADDADDRS: {
2776		struct pfioc_table *io = (struct pfioc_table *)addr;
2777
2778		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2779			error = ENODEV;
2780			break;
2781		}
2782		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2783		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2784		    PFR_FLAG_USERIOCTL);
2785		break;
2786	}
2787
2788	case DIOCRDELADDRS: {
2789		struct pfioc_table *io = (struct pfioc_table *)addr;
2790
2791		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2792			error = ENODEV;
2793			break;
2794		}
2795		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2796		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2797		    PFR_FLAG_USERIOCTL);
2798		break;
2799	}
2800
2801	case DIOCRSETADDRS: {
2802		struct pfioc_table *io = (struct pfioc_table *)addr;
2803
2804		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2805			error = ENODEV;
2806			break;
2807		}
2808		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2809		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2810		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2811		    PFR_FLAG_USERIOCTL);
2812		break;
2813	}
2814
2815	case DIOCRGETADDRS: {
2816		struct pfioc_table *io = (struct pfioc_table *)addr;
2817
2818		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2819			error = ENODEV;
2820			break;
2821		}
2822		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2823		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2824		break;
2825	}
2826
2827	case DIOCRGETASTATS: {
2828		struct pfioc_table *io = (struct pfioc_table *)addr;
2829
2830		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2831			error = ENODEV;
2832			break;
2833		}
2834		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2835		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2836		break;
2837	}
2838
2839	case DIOCRCLRASTATS: {
2840		struct pfioc_table *io = (struct pfioc_table *)addr;
2841
2842		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2843			error = ENODEV;
2844			break;
2845		}
2846		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2847		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2848		    PFR_FLAG_USERIOCTL);
2849		break;
2850	}
2851
2852	case DIOCRTSTADDRS: {
2853		struct pfioc_table *io = (struct pfioc_table *)addr;
2854
2855		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2856			error = ENODEV;
2857			break;
2858		}
2859		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2860		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2861		    PFR_FLAG_USERIOCTL);
2862		break;
2863	}
2864
2865	case DIOCRINADEFINE: {
2866		struct pfioc_table *io = (struct pfioc_table *)addr;
2867
2868		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2869			error = ENODEV;
2870			break;
2871		}
2872		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2873		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2874		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2875		break;
2876	}
2877
2878	case DIOCOSFPADD: {
2879		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2880		error = pf_osfp_add(io);
2881		break;
2882	}
2883
2884	case DIOCOSFPGET: {
2885		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2886		error = pf_osfp_get(io);
2887		break;
2888	}
2889
2890	case DIOCXBEGIN: {
2891		struct pfioc_trans		*io = (struct pfioc_trans *)
2892						    addr;
2893		static struct pfioc_trans_e	 ioe;
2894		static struct pfr_table		 table;
2895		int				 i;
2896
2897		if (io->esize != sizeof(ioe)) {
2898			error = ENODEV;
2899			goto fail;
2900		}
2901		for (i = 0; i < io->size; i++) {
2902#ifdef __FreeBSD__
2903			PF_COPYIN(io->array+i, &ioe, sizeof(ioe), error);
2904			if (error) {
2905#else
2906			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2907#endif
2908				error = EFAULT;
2909				goto fail;
2910			}
2911			switch (ioe.rs_num) {
2912#ifdef ALTQ
2913			case PF_RULESET_ALTQ:
2914				if (ioe.anchor[0]) {
2915					error = EINVAL;
2916					goto fail;
2917				}
2918				if ((error = pf_begin_altq(&ioe.ticket)))
2919					goto fail;
2920				break;
2921#endif /* ALTQ */
2922			case PF_RULESET_TABLE:
2923				bzero(&table, sizeof(table));
2924				strlcpy(table.pfrt_anchor, ioe.anchor,
2925				    sizeof(table.pfrt_anchor));
2926				if ((error = pfr_ina_begin(&table,
2927				    &ioe.ticket, NULL, 0)))
2928					goto fail;
2929				break;
2930			default:
2931				if ((error = pf_begin_rules(&ioe.ticket,
2932				    ioe.rs_num, ioe.anchor)))
2933					goto fail;
2934				break;
2935			}
2936#ifdef __FreeBSD__
2937			PF_COPYOUT(&ioe, io->array+i, sizeof(io->array[i]),
2938			    error);
2939			if (error) {
2940#else
2941			if (copyout(&ioe, io->array+i, sizeof(io->array[i]))) {
2942#endif
2943				error = EFAULT;
2944				goto fail;
2945			}
2946		}
2947		break;
2948	}
2949
2950	case DIOCXROLLBACK: {
2951		struct pfioc_trans		*io = (struct pfioc_trans *)
2952						    addr;
2953		static struct pfioc_trans_e	 ioe;
2954		static struct pfr_table		 table;
2955		int				 i;
2956
2957		if (io->esize != sizeof(ioe)) {
2958			error = ENODEV;
2959			goto fail;
2960		}
2961		for (i = 0; i < io->size; i++) {
2962#ifdef __FreeBSD__
2963			PF_COPYIN(io->array+i, &ioe, sizeof(ioe), error);
2964			if (error) {
2965#else
2966			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2967#endif
2968				error = EFAULT;
2969				goto fail;
2970			}
2971			switch (ioe.rs_num) {
2972#ifdef ALTQ
2973			case PF_RULESET_ALTQ:
2974				if (ioe.anchor[0]) {
2975					error = EINVAL;
2976					goto fail;
2977				}
2978				if ((error = pf_rollback_altq(ioe.ticket)))
2979					goto fail; /* really bad */
2980				break;
2981#endif /* ALTQ */
2982			case PF_RULESET_TABLE:
2983				bzero(&table, sizeof(table));
2984				strlcpy(table.pfrt_anchor, ioe.anchor,
2985				    sizeof(table.pfrt_anchor));
2986				if ((error = pfr_ina_rollback(&table,
2987				    ioe.ticket, NULL, 0)))
2988					goto fail; /* really bad */
2989				break;
2990			default:
2991				if ((error = pf_rollback_rules(ioe.ticket,
2992				    ioe.rs_num, ioe.anchor)))
2993					goto fail; /* really bad */
2994				break;
2995			}
2996		}
2997		break;
2998	}
2999
3000	case DIOCXCOMMIT: {
3001		struct pfioc_trans		*io = (struct pfioc_trans *)
3002						    addr;
3003		static struct pfioc_trans_e	 ioe;
3004		static struct pfr_table		 table;
3005		struct pf_ruleset		*rs;
3006		int				 i;
3007
3008		if (io->esize != sizeof(ioe)) {
3009			error = ENODEV;
3010			goto fail;
3011		}
3012		/* first makes sure everything will succeed */
3013		for (i = 0; i < io->size; i++) {
3014#ifdef __FreeBSD__
3015			PF_COPYIN(io->array+i, &ioe, sizeof(ioe), error);
3016			if (error) {
3017#else
3018			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
3019#endif
3020				error = EFAULT;
3021				goto fail;
3022			}
3023			switch (ioe.rs_num) {
3024#ifdef ALTQ
3025			case PF_RULESET_ALTQ:
3026				if (ioe.anchor[0]) {
3027					error = EINVAL;
3028					goto fail;
3029				}
3030				if (!altqs_inactive_open || ioe.ticket !=
3031				    ticket_altqs_inactive) {
3032					error = EBUSY;
3033					goto fail;
3034				}
3035				break;
3036#endif /* ALTQ */
3037			case PF_RULESET_TABLE:
3038				rs = pf_find_ruleset(ioe.anchor);
3039				if (rs == NULL || !rs->topen || ioe.ticket !=
3040				     rs->tticket) {
3041					error = EBUSY;
3042					goto fail;
3043				}
3044				break;
3045			default:
3046				if (ioe.rs_num < 0 || ioe.rs_num >=
3047				    PF_RULESET_MAX) {
3048					error = EINVAL;
3049					goto fail;
3050				}
3051				rs = pf_find_ruleset(ioe.anchor);
3052				if (rs == NULL ||
3053				    !rs->rules[ioe.rs_num].inactive.open ||
3054				    rs->rules[ioe.rs_num].inactive.ticket !=
3055				    ioe.ticket) {
3056					error = EBUSY;
3057					goto fail;
3058				}
3059				break;
3060			}
3061		}
3062		/* now do the commit - no errors should happen here */
3063		for (i = 0; i < io->size; i++) {
3064#ifdef __FreeBSD__
3065			PF_COPYIN(io->array+i, &ioe, sizeof(ioe), error);
3066			if (error) {
3067#else
3068			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
3069#endif
3070				error = EFAULT;
3071				goto fail;
3072			}
3073			switch (ioe.rs_num) {
3074#ifdef ALTQ
3075			case PF_RULESET_ALTQ:
3076				if ((error = pf_commit_altq(ioe.ticket)))
3077					goto fail; /* really bad */
3078				break;
3079#endif /* ALTQ */
3080			case PF_RULESET_TABLE:
3081				bzero(&table, sizeof(table));
3082				strlcpy(table.pfrt_anchor, ioe.anchor,
3083				    sizeof(table.pfrt_anchor));
3084				if ((error = pfr_ina_commit(&table, ioe.ticket,
3085				    NULL, NULL, 0)))
3086					goto fail; /* really bad */
3087				break;
3088			default:
3089				if ((error = pf_commit_rules(ioe.ticket,
3090				    ioe.rs_num, ioe.anchor)))
3091					goto fail; /* really bad */
3092				break;
3093			}
3094		}
3095		break;
3096	}
3097
3098	case DIOCGETSRCNODES: {
3099		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3100		struct pf_src_node	*n;
3101		struct pf_src_node *p, pstore;
3102		u_int32_t		 nr = 0;
3103		int			 space = psn->psn_len;
3104
3105		if (space == 0) {
3106			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
3107				nr++;
3108			psn->psn_len = sizeof(struct pf_src_node) * nr;
3109			break;
3110		}
3111
3112		p = psn->psn_src_nodes;
3113		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3114			int	secs = time_second, diff;
3115
3116			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3117				break;
3118
3119			bcopy(n, &pstore, sizeof(pstore));
3120			if (n->rule.ptr != NULL)
3121				pstore.rule.nr = n->rule.ptr->nr;
3122			pstore.creation = secs - pstore.creation;
3123			if (pstore.expire > secs)
3124				pstore.expire -= secs;
3125			else
3126				pstore.expire = 0;
3127
3128			/* adjust the connection rate estimate */
3129			diff = secs - n->conn_rate.last;
3130			if (diff >= n->conn_rate.seconds)
3131				pstore.conn_rate.count = 0;
3132			else
3133				pstore.conn_rate.count -=
3134				    n->conn_rate.count * diff /
3135				    n->conn_rate.seconds;
3136
3137#ifdef __FreeBSD__
3138			PF_COPYOUT(&pstore, p, sizeof(*p), error);
3139#else
3140			error = copyout(&pstore, p, sizeof(*p));
3141#endif
3142			if (error)
3143				goto fail;
3144			p++;
3145			nr++;
3146		}
3147		psn->psn_len = sizeof(struct pf_src_node) * nr;
3148		break;
3149	}
3150
3151	case DIOCCLRSRCNODES: {
3152		struct pf_src_node	*n;
3153		struct pf_state		*state;
3154
3155		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3156			state->src_node = NULL;
3157			state->nat_src_node = NULL;
3158		}
3159		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3160			n->expire = 1;
3161			n->states = 0;
3162		}
3163		pf_purge_expired_src_nodes();
3164		pf_status.src_nodes = 0;
3165		break;
3166	}
3167
3168	case DIOCSETHOSTID: {
3169		u_int32_t	*hostid = (u_int32_t *)addr;
3170
3171		if (*hostid == 0)
3172			pf_status.hostid = arc4random();
3173		else
3174			pf_status.hostid = *hostid;
3175		break;
3176	}
3177
3178	case DIOCOSFPFLUSH:
3179		pf_osfp_flush();
3180		break;
3181
3182	case DIOCIGETIFACES: {
3183		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3184
3185		if (io->pfiio_esize != sizeof(struct pfi_if)) {
3186			error = ENODEV;
3187			break;
3188		}
3189		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3190		    &io->pfiio_size, io->pfiio_flags);
3191		break;
3192	}
3193
3194	case DIOCICLRISTATS: {
3195		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3196
3197		error = pfi_clr_istats(io->pfiio_name, &io->pfiio_nzero,
3198		    io->pfiio_flags);
3199		break;
3200	}
3201
3202	case DIOCSETIFFLAG: {
3203		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3204
3205		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3206		break;
3207	}
3208
3209	case DIOCCLRIFFLAG: {
3210		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3211
3212		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3213		break;
3214	}
3215
3216	default:
3217		error = ENODEV;
3218		break;
3219	}
3220fail:
3221#ifdef __FreeBSD__
3222	PF_UNLOCK();
3223#else
3224	splx(s);
3225#endif
3226	return (error);
3227}
3228
3229#ifdef __FreeBSD__
3230/*
3231 * XXX - Check for version missmatch!!!
3232 */
3233static void
3234pf_clear_states(void)
3235{
3236	struct pf_state		*state;
3237
3238	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3239		state->timeout = PFTM_PURGE;
3240#if NPFSYNC
3241		/* don't send out individual delete messages */
3242		state->sync_flags = PFSTATE_NOSYNC;
3243#endif
3244	}
3245	pf_purge_expired_states();
3246	pf_status.states = 0;
3247#if 0 /* NPFSYNC */
3248/*
3249 * XXX This is called on module unload, we do not want to sync that over? */
3250 */
3251	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
3252#endif
3253}
3254
3255static int
3256pf_clear_tables(void)
3257{
3258	struct pfioc_table io;
3259	int error;
3260
3261	bzero(&io, sizeof(io));
3262
3263	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3264	    io.pfrio_flags);
3265
3266	return (error);
3267}
3268
3269static void
3270pf_clear_srcnodes(void)
3271{
3272	struct pf_src_node	*n;
3273	struct pf_state		*state;
3274
3275	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3276		state->src_node = NULL;
3277		state->nat_src_node = NULL;
3278	}
3279	RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3280		n->expire = 1;
3281		n->states = 0;
3282	}
3283	pf_purge_expired_src_nodes();
3284	pf_status.src_nodes = 0;
3285}
3286/*
3287 * XXX - Check for version missmatch!!!
3288 */
3289
3290/*
3291 * Duplicate pfctl -Fa operation to get rid of as much as we can.
3292 */
3293static int
3294shutdown_pf(void)
3295{
3296	int error = 0;
3297	u_int32_t t[5];
3298	char nn = '\0';
3299
3300	callout_stop(&pf_expire_to);
3301
3302	pf_status.running = 0;
3303	do {
3304		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3305		    != 0) {
3306			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3307			break;
3308		}
3309		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3310		    != 0) {
3311			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3312			break;		/* XXX: rollback? */
3313		}
3314		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3315		    != 0) {
3316			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3317			break;		/* XXX: rollback? */
3318		}
3319		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3320		    != 0) {
3321			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3322			break;		/* XXX: rollback? */
3323		}
3324		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3325		    != 0) {
3326			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3327			break;		/* XXX: rollback? */
3328		}
3329
3330		/* XXX: these should always succeed here */
3331		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3332		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3333		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3334		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3335		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3336
3337		if ((error = pf_clear_tables()) != 0)
3338			break;
3339
3340#ifdef ALTQ
3341		if ((error = pf_begin_altq(&t[0])) != 0) {
3342			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3343			break;
3344		}
3345		pf_commit_altq(t[0]);
3346#endif
3347
3348		pf_clear_states();
3349
3350		pf_clear_srcnodes();
3351
3352		/* status does not use malloced mem so no need to cleanup */
3353		/* fingerprints and interfaces have thier own cleanup code */
3354	} while(0);
3355
3356        return (error);
3357}
3358
3359static int
3360pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3361    struct inpcb *inp)
3362{
3363	/*
3364	 * XXX Wed Jul 9 22:03:16 2003 UTC
3365	 * OpenBSD has changed its byte ordering convention on ip_len/ip_off
3366	 * in network stack. OpenBSD's network stack have converted
3367	 * ip_len/ip_off to host byte order frist as FreeBSD.
3368	 * Now this is not true anymore , so we should convert back to network
3369	 * byte order.
3370	 */
3371	struct ip *h = NULL;
3372	int chk;
3373
3374	if ((*m)->m_pkthdr.len >= (int)sizeof(struct ip)) {
3375		/* if m_pkthdr.len is less than ip header, pf will handle. */
3376		h = mtod(*m, struct ip *);
3377	        HTONS(h->ip_len);
3378	        HTONS(h->ip_off);
3379	}
3380	chk = pf_test(PF_IN, ifp, m, NULL, inp);
3381	if (chk && *m) {
3382		m_freem(*m);
3383		*m = NULL;
3384	}
3385	if (*m != NULL) {
3386		/* pf_test can change ip header location */
3387		h = mtod(*m, struct ip *);
3388		NTOHS(h->ip_len);
3389		NTOHS(h->ip_off);
3390	}
3391	return chk;
3392}
3393
3394static int
3395pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3396    struct inpcb *inp)
3397{
3398	/*
3399	 * XXX Wed Jul 9 22:03:16 2003 UTC
3400	 * OpenBSD has changed its byte ordering convention on ip_len/ip_off
3401	 * in network stack. OpenBSD's network stack have converted
3402	 * ip_len/ip_off to host byte order frist as FreeBSD.
3403	 * Now this is not true anymore , so we should convert back to network
3404	 * byte order.
3405	 */
3406	struct ip *h = NULL;
3407	int chk;
3408
3409	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3410	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3411		in_delayed_cksum(*m);
3412		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3413	}
3414	if ((*m)->m_pkthdr.len >= (int)sizeof(*h)) {
3415		/* if m_pkthdr.len is less than ip header, pf will handle. */
3416		h = mtod(*m, struct ip *);
3417	        HTONS(h->ip_len);
3418	        HTONS(h->ip_off);
3419	}
3420	chk = pf_test(PF_OUT, ifp, m, NULL, inp);
3421	if (chk && *m) {
3422		m_freem(*m);
3423		*m = NULL;
3424	}
3425	if (*m != NULL) {
3426		/* pf_test can change ip header location */
3427		h = mtod(*m, struct ip *);
3428		NTOHS(h->ip_len);
3429		NTOHS(h->ip_off);
3430	}
3431	return chk;
3432}
3433
3434#ifdef INET6
3435static int
3436pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3437    struct inpcb *inp)
3438{
3439	/*
3440	 * IPv6 does not affected ip_len/ip_off byte order changes.
3441	 */
3442	int chk;
3443
3444	chk = pf_test6(PF_IN, ifp, m, NULL, inp);
3445	if (chk && *m) {
3446		m_freem(*m);
3447		*m = NULL;
3448	}
3449	return chk;
3450}
3451
3452static int
3453pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3454    struct inpcb *inp)
3455{
3456	/*
3457	 * IPv6 does not affected ip_len/ip_off byte order changes.
3458	 */
3459	int chk;
3460
3461	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3462	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3463		in_delayed_cksum(*m);
3464		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3465	}
3466	chk = pf_test6(PF_OUT, ifp, m, NULL, inp);
3467	if (chk && *m) {
3468		m_freem(*m);
3469		*m = NULL;
3470	}
3471	return chk;
3472}
3473#endif /* INET6 */
3474
3475static int
3476hook_pf(void)
3477{
3478	struct pfil_head *pfh_inet;
3479#ifdef INET6
3480	struct pfil_head *pfh_inet6;
3481#endif
3482
3483	PF_ASSERT(MA_NOTOWNED);
3484
3485	if (pf_pfil_hooked)
3486		return (0);
3487
3488	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3489	if (pfh_inet == NULL)
3490		return (ESRCH); /* XXX */
3491	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3492	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3493#ifdef INET6
3494	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3495	if (pfh_inet6 == NULL) {
3496		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3497		    pfh_inet);
3498		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3499		    pfh_inet);
3500		return (ESRCH); /* XXX */
3501	}
3502	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3503	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3504#endif
3505
3506	pf_pfil_hooked = 1;
3507	return (0);
3508}
3509
3510static int
3511dehook_pf(void)
3512{
3513	struct pfil_head *pfh_inet;
3514#ifdef INET6
3515	struct pfil_head *pfh_inet6;
3516#endif
3517
3518	PF_ASSERT(MA_NOTOWNED);
3519
3520	if (pf_pfil_hooked == 0)
3521		return (0);
3522
3523	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3524	if (pfh_inet == NULL)
3525		return (ESRCH); /* XXX */
3526	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3527	    pfh_inet);
3528	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3529	    pfh_inet);
3530#ifdef INET6
3531	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3532	if (pfh_inet6 == NULL)
3533		return (ESRCH); /* XXX */
3534	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3535	    pfh_inet6);
3536	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3537	    pfh_inet6);
3538#endif
3539
3540	pf_pfil_hooked = 0;
3541	return (0);
3542}
3543
3544static int
3545pf_load(void)
3546{
3547	init_zone_var();
3548	init_pf_mutex();
3549	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3550	if (pfattach() < 0) {
3551		destroy_dev(pf_dev);
3552		destroy_pf_mutex();
3553		return (ENOMEM);
3554	}
3555	return (0);
3556}
3557
3558static int
3559pf_unload(void)
3560{
3561	int error = 0;
3562
3563	PF_LOCK();
3564	pf_status.running = 0;
3565	PF_UNLOCK();
3566	error = dehook_pf();
3567	if (error) {
3568		/*
3569		 * Should not happen!
3570		 * XXX Due to error code ESRCH, kldunload will show
3571		 * a message like 'No such process'.
3572		 */
3573		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3574		return error;
3575	}
3576	PF_LOCK();
3577	shutdown_pf();
3578	pfi_cleanup();
3579	pf_osfp_flush();
3580	pf_osfp_cleanup();
3581	cleanup_pf_zone();
3582	PF_UNLOCK();
3583	destroy_dev(pf_dev);
3584	destroy_pf_mutex();
3585	return error;
3586}
3587
3588static int
3589pf_modevent(module_t mod, int type, void *data)
3590{
3591	int error = 0;
3592
3593	switch(type) {
3594	case MOD_LOAD:
3595		error = pf_load();
3596		break;
3597
3598	case MOD_UNLOAD:
3599		error = pf_unload();
3600		break;
3601	default:
3602		error = EINVAL;
3603		break;
3604	}
3605	return error;
3606}
3607
3608static moduledata_t pf_mod = {
3609	"pf",
3610	pf_modevent,
3611	0
3612};
3613
3614DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST);
3615MODULE_VERSION(pf, PF_MODVER);
3616#endif	/* __FreeBSD__ */
3617