pf_ioctl.c revision 147321
1/*	$FreeBSD: head/sys/contrib/pf/net/pf_ioctl.c 147321 2005-06-12 16:46:20Z 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#define	NBPFILTER	DEV_BPF
48#define	NPFLOG		DEV_PFLOG
49#define	NPFSYNC		DEV_PFSYNC
50#else
51#include "bpfilter.h"
52#include "pflog.h"
53#include "pfsync.h"
54#endif
55
56#include <sys/param.h>
57#include <sys/systm.h>
58#include <sys/mbuf.h>
59#include <sys/filio.h>
60#include <sys/fcntl.h>
61#include <sys/socket.h>
62#include <sys/socketvar.h>
63#include <sys/kernel.h>
64#include <sys/time.h>
65#include <sys/malloc.h>
66#ifdef __FreeBSD__
67#include <sys/module.h>
68#include <sys/conf.h>
69#include <sys/proc.h>
70#else
71#include <sys/timeout.h>
72#include <sys/pool.h>
73#endif
74
75#include <net/if.h>
76#include <net/if_types.h>
77#include <net/route.h>
78
79#include <netinet/in.h>
80#include <netinet/in_var.h>
81#include <netinet/in_systm.h>
82#include <netinet/ip.h>
83#include <netinet/ip_var.h>
84#include <netinet/ip_icmp.h>
85
86#ifndef __FreeBSD__
87#include <dev/rndvar.h>
88#endif
89#include <net/pfvar.h>
90
91#if NPFSYNC > 0
92#include <net/if_pfsync.h>
93#endif /* NPFSYNC > 0 */
94
95#ifdef INET6
96#include <netinet/ip6.h>
97#include <netinet/in_pcb.h>
98#endif /* INET6 */
99
100#ifdef ALTQ
101#include <altq/altq.h>
102#endif
103
104#ifdef __FreeBSD__
105#include <sys/limits.h>
106#include <sys/lock.h>
107#include <sys/mutex.h>
108#include <net/pfil.h>
109#endif /* __FreeBSD__ */
110
111#ifdef __FreeBSD__
112void			 init_zone_var(void);
113void			 cleanup_pf_zone(void);
114int			 pfattach(void);
115#else
116void			 pfattach(int);
117int			 pfopen(dev_t, int, int, struct proc *);
118int			 pfclose(dev_t, int, int, struct proc *);
119#endif
120struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
121			    u_int8_t, u_int8_t, u_int8_t);
122int			 pf_get_ruleset_number(u_int8_t);
123void			 pf_init_ruleset(struct pf_ruleset *);
124int			 pf_anchor_setup(struct pf_rule *,
125			    const struct pf_ruleset *, const char *);
126int			 pf_anchor_copyout(const struct pf_ruleset *,
127			    const struct pf_rule *, struct pfioc_rule *);
128void			 pf_anchor_remove(struct pf_rule *);
129
130void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
131void			 pf_empty_pool(struct pf_palist *);
132#ifdef __FreeBSD__
133int			 pfioctl(struct cdev *, u_long, caddr_t, int, struct thread *);
134#else
135int			 pfioctl(struct cdev *, u_long, caddr_t, int, struct proc *);
136#endif
137#ifdef ALTQ
138int			 pf_begin_altq(u_int32_t *);
139int			 pf_rollback_altq(u_int32_t);
140int			 pf_commit_altq(u_int32_t);
141int			 pf_enable_altq(struct pf_altq *);
142int			 pf_disable_altq(struct pf_altq *);
143#endif /* ALTQ */
144int			 pf_begin_rules(u_int32_t *, int, const char *);
145int			 pf_rollback_rules(u_int32_t, int, char *);
146int			 pf_commit_rules(u_int32_t, int, char *);
147
148#ifdef __FreeBSD__
149extern struct callout	 pf_expire_to;
150#else
151extern struct timeout	 pf_expire_to;
152#endif
153
154struct pf_rule		 pf_default_rule;
155#ifdef ALTQ
156static int		 pf_altq_running;
157#endif
158
159#define	TAGID_MAX	 50000
160TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
161				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
162
163#if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
164#error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
165#endif
166static u_int16_t	 tagname2tag(struct pf_tags *, char *);
167static void		 tag2tagname(struct pf_tags *, u_int16_t, char *);
168static void		 tag_unref(struct pf_tags *, u_int16_t);
169int			 pf_rtlabel_add(struct pf_addr_wrap *);
170void			 pf_rtlabel_remove(struct pf_addr_wrap *);
171void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
172
173#define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
174
175
176#ifdef __FreeBSD__
177static struct cdev	*pf_dev;
178
179/*
180 * XXX - These are new and need to be checked when moveing to a new version
181 */
182static void		 pf_clear_states(void);
183static int		 pf_clear_tables(void);
184static void		 pf_clear_srcnodes(void);
185/*
186 * XXX - These are new and need to be checked when moveing to a new version
187 */
188
189/*
190 * Wrapper functions for pfil(9) hooks
191 */
192static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
193		int dir, struct inpcb *inp);
194static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
195		int dir, struct inpcb *inp);
196#ifdef INET6
197static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
198		int dir, struct inpcb *inp);
199static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
200		int dir, struct inpcb *inp);
201#endif
202
203static int 		 hook_pf(void);
204static int 		 dehook_pf(void);
205static int 		 shutdown_pf(void);
206static int 		 pf_load(void);
207static int 		 pf_unload(void);
208
209static struct cdevsw pf_cdevsw = {
210	.d_ioctl =	pfioctl,
211	.d_name =	PF_NAME,
212	.d_version =	D_VERSION,
213};
214
215static volatile int pf_pfil_hooked = 0;
216struct mtx pf_task_mtx;
217
218void
219init_pf_mutex(void)
220{
221	mtx_init(&pf_task_mtx, "pf task mtx", NULL, MTX_DEF);
222}
223
224void
225destroy_pf_mutex(void)
226{
227	mtx_destroy(&pf_task_mtx);
228}
229
230void
231init_zone_var(void)
232{
233	pf_src_tree_pl = pf_rule_pl = NULL;
234	pf_state_pl = pf_altq_pl = pf_pooladdr_pl = NULL;
235	pf_frent_pl = pf_frag_pl = pf_cache_pl = pf_cent_pl = NULL;
236	pf_state_scrub_pl = NULL;
237	pfr_ktable_pl = pfr_kentry_pl = NULL;
238}
239
240void
241cleanup_pf_zone(void)
242{
243	UMA_DESTROY(pf_src_tree_pl);
244	UMA_DESTROY(pf_rule_pl);
245	UMA_DESTROY(pf_state_pl);
246	UMA_DESTROY(pf_altq_pl);
247	UMA_DESTROY(pf_pooladdr_pl);
248	UMA_DESTROY(pf_frent_pl);
249	UMA_DESTROY(pf_frag_pl);
250	UMA_DESTROY(pf_cache_pl);
251	UMA_DESTROY(pf_cent_pl);
252	UMA_DESTROY(pfr_ktable_pl);
253	UMA_DESTROY(pfr_kentry_pl);
254	UMA_DESTROY(pf_state_scrub_pl);
255	UMA_DESTROY(pfi_addr_pl);
256}
257
258int
259pfattach(void)
260{
261	u_int32_t *my_timeout = pf_default_rule.timeout;
262	int error = 1;
263
264	do {
265		UMA_CREATE(pf_src_tree_pl,struct pf_src_node, "pfsrctrpl");
266		UMA_CREATE(pf_rule_pl,	  struct pf_rule, "pfrulepl");
267		UMA_CREATE(pf_state_pl,	  struct pf_state, "pfstatepl");
268		UMA_CREATE(pf_altq_pl,	  struct pf_altq, "pfaltqpl");
269		UMA_CREATE(pf_pooladdr_pl, struct pf_pooladdr, "pfpooladdrpl");
270		UMA_CREATE(pfr_ktable_pl,  struct pfr_ktable, "pfrktable");
271		UMA_CREATE(pfr_kentry_pl,  struct pfr_kentry, "pfrkentry");
272		UMA_CREATE(pfr_kentry_pl2,  struct pfr_kentry, "pfrkentry2");
273		UMA_CREATE(pf_frent_pl,	  struct pf_frent, "pffrent");
274		UMA_CREATE(pf_frag_pl,	  struct pf_fragment, "pffrag");
275		UMA_CREATE(pf_cache_pl,	  struct pf_fragment, "pffrcache");
276		UMA_CREATE(pf_cent_pl,	  struct pf_frcache, "pffrcent");
277		UMA_CREATE(pf_state_scrub_pl, struct pf_state_scrub,
278		    "pfstatescrub");
279		UMA_CREATE(pfi_addr_pl, struct pfi_dynaddr, "pfiaddrpl");
280		error = 0;
281	} while(0);
282	if (error) {
283		cleanup_pf_zone();
284		return (error);
285	}
286	pfr_initialize();
287	pfi_initialize();
288	if ( (error = pf_osfp_initialize()) ) {
289		cleanup_pf_zone();
290		pf_osfp_cleanup();
291		return (error);
292	}
293
294	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
295	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
296	pf_pool_limits[PF_LIMIT_SRC_NODES].pp = pf_src_tree_pl;
297	pf_pool_limits[PF_LIMIT_SRC_NODES].limit = PFSNODE_HIWAT;
298	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
299	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
300	uma_zone_set_max(pf_pool_limits[PF_LIMIT_STATES].pp,
301		pf_pool_limits[PF_LIMIT_STATES].limit);
302
303	RB_INIT(&tree_src_tracking);
304	RB_INIT(&pf_anchors);
305	pf_init_ruleset(&pf_main_ruleset);
306	TAILQ_INIT(&pf_altqs[0]);
307	TAILQ_INIT(&pf_altqs[1]);
308	TAILQ_INIT(&pf_pabuf);
309	pf_altqs_active = &pf_altqs[0];
310	pf_altqs_inactive = &pf_altqs[1];
311	TAILQ_INIT(&state_updates);
312
313	/* default rule should never be garbage collected */
314	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
315	pf_default_rule.action = PF_PASS;
316	pf_default_rule.nr = -1;
317
318	/* initialize default timeouts */
319	my_timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
320	my_timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
321	my_timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
322	my_timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
323	my_timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
324	my_timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
325	my_timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
326	my_timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
327	my_timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
328	my_timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
329	my_timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
330	my_timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
331	my_timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
332	my_timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
333	my_timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
334	my_timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
335	my_timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
336	my_timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
337
338	callout_init(&pf_expire_to, NET_CALLOUT_MPSAFE);
339	callout_reset(&pf_expire_to, my_timeout[PFTM_INTERVAL] * hz,
340	    pf_purge_timeout, &pf_expire_to);
341
342	pf_normalize_init();
343	bzero(&pf_status, sizeof(pf_status));
344	pf_pfil_hooked = 0;
345
346	/* XXX do our best to avoid a conflict */
347	pf_status.hostid = arc4random();
348
349	return (error);
350}
351#else /* !__FreeBSD__ */
352void
353pfattach(int num)
354{
355	u_int32_t *timeout = pf_default_rule.timeout;
356
357	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
358	    &pool_allocator_nointr);
359	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
360	    "pfsrctrpl", NULL);
361	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
362	    NULL);
363	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
364	    &pool_allocator_nointr);
365	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
366	    "pfpooladdrpl", &pool_allocator_nointr);
367	pfr_initialize();
368	pfi_initialize();
369	pf_osfp_initialize();
370
371	pool_sethardlimit(pf_pool_limits[PF_LIMIT_STATES].pp,
372	    pf_pool_limits[PF_LIMIT_STATES].limit, NULL, 0);
373
374	RB_INIT(&tree_src_tracking);
375	RB_INIT(&pf_anchors);
376	pf_init_ruleset(&pf_main_ruleset);
377	TAILQ_INIT(&pf_altqs[0]);
378	TAILQ_INIT(&pf_altqs[1]);
379	TAILQ_INIT(&pf_pabuf);
380	pf_altqs_active = &pf_altqs[0];
381	pf_altqs_inactive = &pf_altqs[1];
382	TAILQ_INIT(&state_updates);
383
384	/* default rule should never be garbage collected */
385	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
386	pf_default_rule.action = PF_PASS;
387	pf_default_rule.nr = -1;
388
389	/* initialize default timeouts */
390	timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
391	timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
392	timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
393	timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
394	timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
395	timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
396	timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
397	timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
398	timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
399	timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
400	timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
401	timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
402	timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
403	timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
404	timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
405	timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
406	timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
407	timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
408
409	timeout_set(&pf_expire_to, pf_purge_timeout, &pf_expire_to);
410	timeout_add(&pf_expire_to, timeout[PFTM_INTERVAL] * hz);
411
412	pf_normalize_init();
413	bzero(&pf_status, sizeof(pf_status));
414	pf_status.debug = PF_DEBUG_URGENT;
415
416	/* XXX do our best to avoid a conflict */
417	pf_status.hostid = arc4random();
418}
419
420int
421pfopen(struct cdev *dev, int flags, int fmt, struct proc *p)
422{
423	if (minor(dev) >= 1)
424		return (ENXIO);
425	return (0);
426}
427
428int
429pfclose(struct cdev *dev, int flags, int fmt, struct proc *p)
430{
431	if (minor(dev) >= 1)
432		return (ENXIO);
433	return (0);
434}
435#endif /* __FreeBSD__ */
436
437struct pf_pool *
438pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
439    u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
440    u_int8_t check_ticket)
441{
442	struct pf_ruleset	*ruleset;
443	struct pf_rule		*rule;
444	int			 rs_num;
445
446	ruleset = pf_find_ruleset(anchor);
447	if (ruleset == NULL)
448		return (NULL);
449	rs_num = pf_get_ruleset_number(rule_action);
450	if (rs_num >= PF_RULESET_MAX)
451		return (NULL);
452	if (active) {
453		if (check_ticket && ticket !=
454		    ruleset->rules[rs_num].active.ticket)
455			return (NULL);
456		if (r_last)
457			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
458			    pf_rulequeue);
459		else
460			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
461	} else {
462		if (check_ticket && ticket !=
463		    ruleset->rules[rs_num].inactive.ticket)
464			return (NULL);
465		if (r_last)
466			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
467			    pf_rulequeue);
468		else
469			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
470	}
471	if (!r_last) {
472		while ((rule != NULL) && (rule->nr != rule_number))
473			rule = TAILQ_NEXT(rule, entries);
474	}
475	if (rule == NULL)
476		return (NULL);
477
478	return (&rule->rpool);
479}
480
481int
482pf_get_ruleset_number(u_int8_t action)
483{
484	switch (action) {
485	case PF_SCRUB:
486	case PF_NOSCRUB:
487		return (PF_RULESET_SCRUB);
488		break;
489	case PF_PASS:
490	case PF_DROP:
491		return (PF_RULESET_FILTER);
492		break;
493	case PF_NAT:
494	case PF_NONAT:
495		return (PF_RULESET_NAT);
496		break;
497	case PF_BINAT:
498	case PF_NOBINAT:
499		return (PF_RULESET_BINAT);
500		break;
501	case PF_RDR:
502	case PF_NORDR:
503		return (PF_RULESET_RDR);
504		break;
505	default:
506		return (PF_RULESET_MAX);
507		break;
508	}
509}
510
511void
512pf_init_ruleset(struct pf_ruleset *ruleset)
513{
514	int	i;
515
516	memset(ruleset, 0, sizeof(struct pf_ruleset));
517	for (i = 0; i < PF_RULESET_MAX; i++) {
518		TAILQ_INIT(&ruleset->rules[i].queues[0]);
519		TAILQ_INIT(&ruleset->rules[i].queues[1]);
520		ruleset->rules[i].active.ptr = &ruleset->rules[i].queues[0];
521		ruleset->rules[i].inactive.ptr = &ruleset->rules[i].queues[1];
522	}
523}
524
525struct pf_anchor *
526pf_find_anchor(const char *path)
527{
528	static struct pf_anchor	 key;
529
530	memset(&key, 0, sizeof(key));
531	strlcpy(key.path, path, sizeof(key.path));
532	return (RB_FIND(pf_anchor_global, &pf_anchors, &key));
533}
534
535struct pf_ruleset *
536pf_find_ruleset(const char *path)
537{
538	struct pf_anchor	*anchor;
539
540	while (*path == '/')
541		path++;
542	if (!*path)
543		return (&pf_main_ruleset);
544	anchor = pf_find_anchor(path);
545	if (anchor == NULL)
546		return (NULL);
547	else
548		return (&anchor->ruleset);
549}
550
551struct pf_ruleset *
552pf_find_or_create_ruleset(const char *path)
553{
554	static char		 p[MAXPATHLEN];
555	char			*q = NULL, *r;	/* make the compiler happy */
556	struct pf_ruleset	*ruleset;
557	struct pf_anchor	*anchor = NULL, *dup, *parent = NULL;
558
559	while (*path == '/')
560		path++;
561	ruleset = pf_find_ruleset(path);
562	if (ruleset != NULL)
563		return (ruleset);
564	strlcpy(p, path, sizeof(p));
565#ifdef __FreeBSD__
566	while (parent == NULL && (q = rindex(p, '/')) != NULL) {
567#else
568	while (parent == NULL && (q = strrchr(p, '/')) != NULL) {
569#endif
570		*q = 0;
571		if ((ruleset = pf_find_ruleset(p)) != NULL) {
572			parent = ruleset->anchor;
573			break;
574		}
575	}
576	if (q == NULL)
577		q = p;
578	else
579		q++;
580	strlcpy(p, path, sizeof(p));
581	if (!*q)
582		return (NULL);
583#ifdef __FreeBSD__
584	while ((r = index(q, '/')) != NULL || *q) {
585#else
586	while ((r = strchr(q, '/')) != NULL || *q) {
587#endif
588		if (r != NULL)
589			*r = 0;
590		if (!*q || strlen(q) >= PF_ANCHOR_NAME_SIZE ||
591		    (parent != NULL && strlen(parent->path) >=
592		    MAXPATHLEN - PF_ANCHOR_NAME_SIZE - 1))
593			return (NULL);
594		anchor = (struct pf_anchor *)malloc(sizeof(*anchor), M_TEMP,
595		    M_NOWAIT);
596		if (anchor == NULL)
597			return (NULL);
598		memset(anchor, 0, sizeof(*anchor));
599		RB_INIT(&anchor->children);
600		strlcpy(anchor->name, q, sizeof(anchor->name));
601		if (parent != NULL) {
602			strlcpy(anchor->path, parent->path,
603			    sizeof(anchor->path));
604			strlcat(anchor->path, "/", sizeof(anchor->path));
605		}
606		strlcat(anchor->path, anchor->name, sizeof(anchor->path));
607		if ((dup = RB_INSERT(pf_anchor_global, &pf_anchors, anchor)) !=
608		    NULL) {
609			printf("pf_find_or_create_ruleset: RB_INSERT1 "
610			    "'%s' '%s' collides with '%s' '%s'\n",
611			    anchor->path, anchor->name, dup->path, dup->name);
612			free(anchor, M_TEMP);
613			return (NULL);
614		}
615		if (parent != NULL) {
616			anchor->parent = parent;
617			if ((dup = RB_INSERT(pf_anchor_node, &parent->children,
618			    anchor)) != NULL) {
619				printf("pf_find_or_create_ruleset: "
620				    "RB_INSERT2 '%s' '%s' collides with "
621				    "'%s' '%s'\n", anchor->path, anchor->name,
622				    dup->path, dup->name);
623				RB_REMOVE(pf_anchor_global, &pf_anchors,
624				    anchor);
625				free(anchor, M_TEMP);
626				return (NULL);
627			}
628		}
629		pf_init_ruleset(&anchor->ruleset);
630		anchor->ruleset.anchor = anchor;
631		parent = anchor;
632		if (r != NULL)
633			q = r + 1;
634		else
635			*q = 0;
636	}
637	return (&anchor->ruleset);
638}
639
640void
641pf_remove_if_empty_ruleset(struct pf_ruleset *ruleset)
642{
643	struct pf_anchor	*parent;
644	int			 i;
645
646	while (ruleset != NULL) {
647		if (ruleset == &pf_main_ruleset || ruleset->anchor == NULL ||
648		    !RB_EMPTY(&ruleset->anchor->children) ||
649		    ruleset->anchor->refcnt > 0 || ruleset->tables > 0 ||
650		    ruleset->topen)
651			return;
652		for (i = 0; i < PF_RULESET_MAX; ++i)
653			if (!TAILQ_EMPTY(ruleset->rules[i].active.ptr) ||
654			    !TAILQ_EMPTY(ruleset->rules[i].inactive.ptr) ||
655			    ruleset->rules[i].inactive.open)
656				return;
657		RB_REMOVE(pf_anchor_global, &pf_anchors, ruleset->anchor);
658		if ((parent = ruleset->anchor->parent) != NULL)
659			RB_REMOVE(pf_anchor_node, &parent->children,
660			    ruleset->anchor);
661		free(ruleset->anchor, M_TEMP);
662		if (parent == NULL)
663			return;
664		ruleset = &parent->ruleset;
665	}
666}
667
668int
669pf_anchor_setup(struct pf_rule *r, const struct pf_ruleset *s,
670    const char *name)
671{
672	static char		*p, path[MAXPATHLEN];
673	struct pf_ruleset	*ruleset;
674
675	r->anchor = NULL;
676	r->anchor_relative = 0;
677	r->anchor_wildcard = 0;
678	if (!name[0])
679		return (0);
680	if (name[0] == '/')
681		strlcpy(path, name + 1, sizeof(path));
682	else {
683		/* relative path */
684		r->anchor_relative = 1;
685		if (s->anchor == NULL || !s->anchor->path[0])
686			path[0] = 0;
687		else
688			strlcpy(path, s->anchor->path, sizeof(path));
689		while (name[0] == '.' && name[1] == '.' && name[2] == '/') {
690			if (!path[0]) {
691				printf("pf_anchor_setup: .. beyond root\n");
692				return (1);
693			}
694#ifdef __FreeBSD__
695			if ((p = rindex(path, '/')) != NULL)
696#else
697			if ((p = strrchr(path, '/')) != NULL)
698#endif
699				*p = 0;
700			else
701				path[0] = 0;
702			r->anchor_relative++;
703			name += 3;
704		}
705		if (path[0])
706			strlcat(path, "/", sizeof(path));
707		strlcat(path, name, sizeof(path));
708	}
709#ifdef __FreeBSD__
710	if ((p = rindex(path, '/')) != NULL && !strcmp(p, "/*")) {
711#else
712	if ((p = strrchr(path, '/')) != NULL && !strcmp(p, "/*")) {
713#endif
714		r->anchor_wildcard = 1;
715		*p = 0;
716	}
717	ruleset = pf_find_or_create_ruleset(path);
718	if (ruleset == NULL || ruleset->anchor == NULL) {
719		printf("pf_anchor_setup: ruleset\n");
720		return (1);
721	}
722	r->anchor = ruleset->anchor;
723	r->anchor->refcnt++;
724	return (0);
725}
726
727int
728pf_anchor_copyout(const struct pf_ruleset *rs, const struct pf_rule *r,
729    struct pfioc_rule *pr)
730{
731	pr->anchor_call[0] = 0;
732	if (r->anchor == NULL)
733		return (0);
734	if (!r->anchor_relative) {
735		strlcpy(pr->anchor_call, "/", sizeof(pr->anchor_call));
736		strlcat(pr->anchor_call, r->anchor->path,
737		    sizeof(pr->anchor_call));
738	} else {
739		char a[MAXPATHLEN], b[MAXPATHLEN], *p;
740		int i;
741
742		if (rs->anchor == NULL)
743			a[0] = 0;
744		else
745			strlcpy(a, rs->anchor->path, sizeof(a));
746		strlcpy(b, r->anchor->path, sizeof(b));
747		for (i = 1; i < r->anchor_relative; ++i) {
748#ifdef __FreeBSD__
749			if ((p = rindex(a, '/')) == NULL)
750#else
751			if ((p = strrchr(a, '/')) == NULL)
752#endif
753				p = a;
754			*p = 0;
755			strlcat(pr->anchor_call, "../",
756			    sizeof(pr->anchor_call));
757		}
758		if (strncmp(a, b, strlen(a))) {
759			printf("pf_anchor_copyout: '%s' '%s'\n", a, b);
760			return (1);
761		}
762		if (strlen(b) > strlen(a))
763			strlcat(pr->anchor_call, b + (a[0] ? strlen(a) + 1 : 0),
764			    sizeof(pr->anchor_call));
765	}
766	if (r->anchor_wildcard)
767		strlcat(pr->anchor_call, pr->anchor_call[0] ? "/*" : "*",
768		    sizeof(pr->anchor_call));
769	return (0);
770}
771
772void
773pf_anchor_remove(struct pf_rule *r)
774{
775	if (r->anchor == NULL)
776		return;
777	if (r->anchor->refcnt <= 0) {
778		printf("pf_anchor_remove: broken refcount");
779		r->anchor = NULL;
780		return;
781	}
782	if (!--r->anchor->refcnt)
783		pf_remove_if_empty_ruleset(&r->anchor->ruleset);
784	r->anchor = NULL;
785}
786
787void
788pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
789{
790	struct pf_pooladdr	*mv_pool_pa;
791
792	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
793		TAILQ_REMOVE(poola, mv_pool_pa, entries);
794		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
795	}
796}
797
798void
799pf_empty_pool(struct pf_palist *poola)
800{
801	struct pf_pooladdr	*empty_pool_pa;
802
803	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
804		pfi_dynaddr_remove(&empty_pool_pa->addr);
805		pf_tbladdr_remove(&empty_pool_pa->addr);
806		pfi_detach_rule(empty_pool_pa->kif);
807		TAILQ_REMOVE(poola, empty_pool_pa, entries);
808		pool_put(&pf_pooladdr_pl, empty_pool_pa);
809	}
810}
811
812void
813pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
814{
815	if (rulequeue != NULL) {
816		if (rule->states <= 0) {
817			/*
818			 * XXX - we need to remove the table *before* detaching
819			 * the rule to make sure the table code does not delete
820			 * the anchor under our feet.
821			 */
822			pf_tbladdr_remove(&rule->src.addr);
823			pf_tbladdr_remove(&rule->dst.addr);
824			if (rule->overload_tbl)
825				pfr_detach_table(rule->overload_tbl);
826		}
827		TAILQ_REMOVE(rulequeue, rule, entries);
828		rule->entries.tqe_prev = NULL;
829		rule->nr = -1;
830	}
831
832	if (rule->states > 0 || rule->src_nodes > 0 ||
833	    rule->entries.tqe_prev != NULL)
834		return;
835	pf_tag_unref(rule->tag);
836	pf_tag_unref(rule->match_tag);
837#ifdef ALTQ
838	if (rule->pqid != rule->qid)
839		pf_qid_unref(rule->pqid);
840	pf_qid_unref(rule->qid);
841#endif
842	pf_rtlabel_remove(&rule->src.addr);
843	pf_rtlabel_remove(&rule->dst.addr);
844	pfi_dynaddr_remove(&rule->src.addr);
845	pfi_dynaddr_remove(&rule->dst.addr);
846	if (rulequeue == NULL) {
847		pf_tbladdr_remove(&rule->src.addr);
848		pf_tbladdr_remove(&rule->dst.addr);
849		if (rule->overload_tbl)
850			pfr_detach_table(rule->overload_tbl);
851	}
852	pfi_detach_rule(rule->kif);
853	pf_anchor_remove(rule);
854	pf_empty_pool(&rule->rpool.list);
855	pool_put(&pf_rule_pl, rule);
856}
857
858static	u_int16_t
859tagname2tag(struct pf_tags *head, char *tagname)
860{
861	struct pf_tagname	*tag, *p = NULL;
862	u_int16_t		 new_tagid = 1;
863
864	TAILQ_FOREACH(tag, head, entries)
865		if (strcmp(tagname, tag->name) == 0) {
866			tag->ref++;
867			return (tag->tag);
868		}
869
870	/*
871	 * to avoid fragmentation, we do a linear search from the beginning
872	 * and take the first free slot we find. if there is none or the list
873	 * is empty, append a new entry at the end.
874	 */
875
876	/* new entry */
877	if (!TAILQ_EMPTY(head))
878		for (p = TAILQ_FIRST(head); p != NULL &&
879		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
880			new_tagid = p->tag + 1;
881
882	if (new_tagid > TAGID_MAX)
883		return (0);
884
885	/* allocate and fill new struct pf_tagname */
886	tag = (struct pf_tagname *)malloc(sizeof(struct pf_tagname),
887	    M_TEMP, M_NOWAIT);
888	if (tag == NULL)
889		return (0);
890	bzero(tag, sizeof(struct pf_tagname));
891	strlcpy(tag->name, tagname, sizeof(tag->name));
892	tag->tag = new_tagid;
893	tag->ref++;
894
895	if (p != NULL)	/* insert new entry before p */
896		TAILQ_INSERT_BEFORE(p, tag, entries);
897	else	/* either list empty or no free slot in between */
898		TAILQ_INSERT_TAIL(head, tag, entries);
899
900	return (tag->tag);
901}
902
903static	void
904tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
905{
906	struct pf_tagname	*tag;
907
908	TAILQ_FOREACH(tag, head, entries)
909		if (tag->tag == tagid) {
910			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
911			return;
912		}
913}
914
915static	void
916tag_unref(struct pf_tags *head, u_int16_t tag)
917{
918	struct pf_tagname	*p, *next;
919
920	if (tag == 0)
921		return;
922
923	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
924		next = TAILQ_NEXT(p, entries);
925		if (tag == p->tag) {
926			if (--p->ref == 0) {
927				TAILQ_REMOVE(head, p, entries);
928				free(p, M_TEMP);
929			}
930			break;
931		}
932	}
933}
934
935u_int16_t
936pf_tagname2tag(char *tagname)
937{
938	return (tagname2tag(&pf_tags, tagname));
939}
940
941void
942pf_tag2tagname(u_int16_t tagid, char *p)
943{
944	return (tag2tagname(&pf_tags, tagid, p));
945}
946
947void
948pf_tag_ref(u_int16_t tag)
949{
950	struct pf_tagname *t;
951
952	TAILQ_FOREACH(t, &pf_tags, entries)
953		if (t->tag == tag)
954			break;
955	if (t != NULL)
956		t->ref++;
957}
958
959void
960pf_tag_unref(u_int16_t tag)
961{
962	return (tag_unref(&pf_tags, tag));
963}
964
965int
966pf_rtlabel_add(struct pf_addr_wrap *a)
967{
968#ifdef __FreeBSD__
969	/* XXX_IMPORT: later */
970	return (0);
971#else
972	if (a->type == PF_ADDR_RTLABEL &&
973	    (a->v.rtlabel = rtlabel_name2id(a->v.rtlabelname)) == 0)
974		return (-1);
975	return (0);
976#endif
977}
978
979void
980pf_rtlabel_remove(struct pf_addr_wrap *a)
981{
982#ifdef __FreeBSD__
983	/* XXX_IMPORT: later */
984#else
985	if (a->type == PF_ADDR_RTLABEL)
986		rtlabel_unref(a->v.rtlabel);
987#endif
988}
989
990void
991pf_rtlabel_copyout(struct pf_addr_wrap *a)
992{
993#ifdef __FreeBSD__
994	/* XXX_IMPORT: later */
995	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
996		strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
997#else
998	const char	*name;
999
1000	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel) {
1001		if ((name = rtlabel_id2name(a->v.rtlabel)) == NULL)
1002			strlcpy(a->v.rtlabelname, "?",
1003			    sizeof(a->v.rtlabelname));
1004		else
1005			strlcpy(a->v.rtlabelname, name,
1006			    sizeof(a->v.rtlabelname));
1007	}
1008#endif
1009}
1010
1011#ifdef ALTQ
1012u_int32_t
1013pf_qname2qid(char *qname)
1014{
1015	return ((u_int32_t)tagname2tag(&pf_qids, qname));
1016}
1017
1018void
1019pf_qid2qname(u_int32_t qid, char *p)
1020{
1021	return (tag2tagname(&pf_qids, (u_int16_t)qid, p));
1022}
1023
1024void
1025pf_qid_unref(u_int32_t qid)
1026{
1027	return (tag_unref(&pf_qids, (u_int16_t)qid));
1028}
1029
1030int
1031pf_begin_altq(u_int32_t *ticket)
1032{
1033	struct pf_altq	*altq;
1034	int		 error = 0;
1035
1036	/* Purge the old altq list */
1037	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
1038		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
1039		if (altq->qname[0] == 0) {
1040			/* detach and destroy the discipline */
1041			error = altq_remove(altq);
1042		} else
1043			pf_qid_unref(altq->qid);
1044		pool_put(&pf_altq_pl, altq);
1045	}
1046	if (error)
1047		return (error);
1048	*ticket = ++ticket_altqs_inactive;
1049	altqs_inactive_open = 1;
1050	return (0);
1051}
1052
1053int
1054pf_rollback_altq(u_int32_t ticket)
1055{
1056	struct pf_altq	*altq;
1057	int		 error = 0;
1058
1059	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
1060		return (0);
1061	/* Purge the old altq list */
1062	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
1063		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
1064		if (altq->qname[0] == 0) {
1065			/* detach and destroy the discipline */
1066			error = altq_remove(altq);
1067		} else
1068			pf_qid_unref(altq->qid);
1069		pool_put(&pf_altq_pl, altq);
1070	}
1071	altqs_inactive_open = 0;
1072	return (error);
1073}
1074
1075int
1076pf_commit_altq(u_int32_t ticket)
1077{
1078	struct pf_altqqueue	*old_altqs;
1079	struct pf_altq		*altq;
1080	int			 s, err, error = 0;
1081
1082	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
1083		return (EBUSY);
1084
1085	/* swap altqs, keep the old. */
1086	s = splsoftnet();
1087	old_altqs = pf_altqs_active;
1088	pf_altqs_active = pf_altqs_inactive;
1089	pf_altqs_inactive = old_altqs;
1090	ticket_altqs_active = ticket_altqs_inactive;
1091
1092	/* Attach new disciplines */
1093	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1094		if (altq->qname[0] == 0) {
1095			/* attach the discipline */
1096			error = altq_pfattach(altq);
1097			if (error == 0 && pf_altq_running)
1098				error = pf_enable_altq(altq);
1099			if (error != 0) {
1100				splx(s);
1101				return (error);
1102			}
1103		}
1104	}
1105
1106	/* Purge the old altq list */
1107	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
1108		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
1109		if (altq->qname[0] == 0) {
1110			/* detach and destroy the discipline */
1111			if (pf_altq_running)
1112				error = pf_disable_altq(altq);
1113			err = altq_pfdetach(altq);
1114			if (err != 0 && error == 0)
1115				error = err;
1116			err = altq_remove(altq);
1117			if (err != 0 && error == 0)
1118				error = err;
1119		} else
1120			pf_qid_unref(altq->qid);
1121		pool_put(&pf_altq_pl, altq);
1122	}
1123	splx(s);
1124
1125	altqs_inactive_open = 0;
1126	return (error);
1127}
1128
1129int
1130pf_enable_altq(struct pf_altq *altq)
1131{
1132	struct ifnet		*ifp;
1133	struct tb_profile	 tb;
1134	int			 s, error = 0;
1135
1136	if ((ifp = ifunit(altq->ifname)) == NULL)
1137		return (EINVAL);
1138
1139	if (ifp->if_snd.altq_type != ALTQT_NONE)
1140		error = altq_enable(&ifp->if_snd);
1141
1142	/* set tokenbucket regulator */
1143	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
1144		tb.rate = altq->ifbandwidth;
1145		tb.depth = altq->tbrsize;
1146		s = splimp();
1147#ifdef __FreeBSD__
1148		PF_UNLOCK();
1149#endif
1150		error = tbr_set(&ifp->if_snd, &tb);
1151#ifdef __FreeBSD__
1152		PF_LOCK();
1153#endif
1154		splx(s);
1155	}
1156
1157	return (error);
1158}
1159
1160int
1161pf_disable_altq(struct pf_altq *altq)
1162{
1163	struct ifnet		*ifp;
1164	struct tb_profile	 tb;
1165	int			 s, error;
1166
1167	if ((ifp = ifunit(altq->ifname)) == NULL)
1168		return (EINVAL);
1169
1170	/*
1171	 * when the discipline is no longer referenced, it was overridden
1172	 * by a new one.  if so, just return.
1173	 */
1174	if (altq->altq_disc != ifp->if_snd.altq_disc)
1175		return (0);
1176
1177	error = altq_disable(&ifp->if_snd);
1178
1179	if (error == 0) {
1180		/* clear tokenbucket regulator */
1181		tb.rate = 0;
1182		s = splimp();
1183#ifdef __FreeBSD__
1184		PF_UNLOCK();
1185#endif
1186		error = tbr_set(&ifp->if_snd, &tb);
1187#ifdef __FreeBSD__
1188		PF_LOCK();
1189#endif
1190		splx(s);
1191	}
1192
1193	return (error);
1194}
1195#endif /* ALTQ */
1196
1197int
1198pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
1199{
1200	struct pf_ruleset	*rs;
1201	struct pf_rule		*rule;
1202
1203	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1204		return (EINVAL);
1205	rs = pf_find_or_create_ruleset(anchor);
1206	if (rs == NULL)
1207		return (EINVAL);
1208	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL)
1209		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
1210	*ticket = ++rs->rules[rs_num].inactive.ticket;
1211	rs->rules[rs_num].inactive.open = 1;
1212	return (0);
1213}
1214
1215int
1216pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
1217{
1218	struct pf_ruleset	*rs;
1219	struct pf_rule		*rule;
1220
1221	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1222		return (EINVAL);
1223	rs = pf_find_ruleset(anchor);
1224	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1225	    rs->rules[rs_num].inactive.ticket != ticket)
1226		return (0);
1227	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL)
1228		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
1229	rs->rules[rs_num].inactive.open = 0;
1230	return (0);
1231}
1232
1233int
1234pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1235{
1236	struct pf_ruleset	*rs;
1237	struct pf_rule		*rule;
1238	struct pf_rulequeue	*old_rules;
1239	int			 s;
1240
1241	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1242		return (EINVAL);
1243	rs = pf_find_ruleset(anchor);
1244	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1245	    ticket != rs->rules[rs_num].inactive.ticket)
1246		return (EBUSY);
1247
1248	/* Swap rules, keep the old. */
1249	s = splsoftnet();
1250	old_rules = rs->rules[rs_num].active.ptr;
1251	rs->rules[rs_num].active.ptr =
1252	    rs->rules[rs_num].inactive.ptr;
1253	rs->rules[rs_num].inactive.ptr = old_rules;
1254	rs->rules[rs_num].active.ticket =
1255	    rs->rules[rs_num].inactive.ticket;
1256	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1257
1258	/* Purge the old rule list. */
1259	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1260		pf_rm_rule(old_rules, rule);
1261	rs->rules[rs_num].inactive.open = 0;
1262	pf_remove_if_empty_ruleset(rs);
1263	splx(s);
1264	return (0);
1265}
1266
1267#ifdef __FreeBSD__
1268int
1269pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct thread *td)
1270#else
1271int
1272pfioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flags, struct proc *p)
1273#endif
1274{
1275	struct pf_pooladdr	*pa = NULL;
1276	struct pf_pool		*pool = NULL;
1277#ifndef __FreeBSD__
1278	int			 s;
1279#endif
1280	int			 error = 0;
1281
1282	/* XXX keep in sync with switch() below */
1283#ifdef __FreeBSD__
1284	if (securelevel_gt(td->td_ucred, 2))
1285#else
1286	if (securelevel > 1)
1287#endif
1288		switch (cmd) {
1289		case DIOCGETRULES:
1290		case DIOCGETRULE:
1291		case DIOCGETADDRS:
1292		case DIOCGETADDR:
1293		case DIOCGETSTATE:
1294		case DIOCSETSTATUSIF:
1295		case DIOCGETSTATUS:
1296		case DIOCCLRSTATUS:
1297		case DIOCNATLOOK:
1298		case DIOCSETDEBUG:
1299		case DIOCGETSTATES:
1300		case DIOCGETTIMEOUT:
1301		case DIOCCLRRULECTRS:
1302		case DIOCGETLIMIT:
1303		case DIOCGETALTQS:
1304		case DIOCGETALTQ:
1305		case DIOCGETQSTATS:
1306		case DIOCGETRULESETS:
1307		case DIOCGETRULESET:
1308		case DIOCRGETTABLES:
1309		case DIOCRGETTSTATS:
1310		case DIOCRCLRTSTATS:
1311		case DIOCRCLRADDRS:
1312		case DIOCRADDADDRS:
1313		case DIOCRDELADDRS:
1314		case DIOCRSETADDRS:
1315		case DIOCRGETADDRS:
1316		case DIOCRGETASTATS:
1317		case DIOCRCLRASTATS:
1318		case DIOCRTSTADDRS:
1319		case DIOCOSFPGET:
1320		case DIOCGETSRCNODES:
1321		case DIOCCLRSRCNODES:
1322		case DIOCIGETIFACES:
1323		case DIOCICLRISTATS:
1324#ifdef __FreeBSD__
1325		case DIOCGIFSPEED:
1326#endif
1327		case DIOCSETIFFLAG:
1328		case DIOCCLRIFFLAG:
1329			break;
1330		case DIOCRCLRTABLES:
1331		case DIOCRADDTABLES:
1332		case DIOCRDELTABLES:
1333		case DIOCRSETTFLAGS:
1334			if (((struct pfioc_table *)addr)->pfrio_flags &
1335			    PFR_FLAG_DUMMY)
1336				break; /* dummy operation ok */
1337			return (EPERM);
1338		default:
1339			return (EPERM);
1340		}
1341
1342	if (!(flags & FWRITE))
1343		switch (cmd) {
1344		case DIOCGETRULES:
1345		case DIOCGETRULE:
1346		case DIOCGETADDRS:
1347		case DIOCGETADDR:
1348		case DIOCGETSTATE:
1349		case DIOCGETSTATUS:
1350		case DIOCGETSTATES:
1351		case DIOCGETTIMEOUT:
1352		case DIOCGETLIMIT:
1353		case DIOCGETALTQS:
1354		case DIOCGETALTQ:
1355		case DIOCGETQSTATS:
1356		case DIOCGETRULESETS:
1357		case DIOCGETRULESET:
1358		case DIOCRGETTABLES:
1359		case DIOCRGETTSTATS:
1360		case DIOCRGETADDRS:
1361		case DIOCRGETASTATS:
1362		case DIOCRTSTADDRS:
1363		case DIOCOSFPGET:
1364		case DIOCGETSRCNODES:
1365		case DIOCIGETIFACES:
1366#ifdef __FreeBSD__
1367		case DIOCGIFSPEED:
1368#endif
1369			break;
1370		case DIOCRCLRTABLES:
1371		case DIOCRADDTABLES:
1372		case DIOCRDELTABLES:
1373		case DIOCRCLRTSTATS:
1374		case DIOCRCLRADDRS:
1375		case DIOCRADDADDRS:
1376		case DIOCRDELADDRS:
1377		case DIOCRSETADDRS:
1378		case DIOCRSETTFLAGS:
1379			if (((struct pfioc_table *)addr)->pfrio_flags &
1380			    PFR_FLAG_DUMMY)
1381				break; /* dummy operation ok */
1382			return (EACCES);
1383		default:
1384			return (EACCES);
1385		}
1386
1387#ifdef __FreeBSD__
1388	PF_LOCK();
1389#else
1390	s = splsoftnet();
1391#endif
1392	switch (cmd) {
1393
1394	case DIOCSTART:
1395		if (pf_status.running)
1396			error = EEXIST;
1397		else {
1398#ifdef __FreeBSD__
1399			PF_UNLOCK();
1400			error = hook_pf();
1401			PF_LOCK();
1402			if (error) {
1403				DPFPRINTF(PF_DEBUG_MISC,
1404				    ("pf: pfil registeration fail\n"));
1405				break;
1406			}
1407#endif
1408			pf_status.running = 1;
1409			pf_status.since = time_second;
1410			if (pf_status.stateid == 0) {
1411				pf_status.stateid = time_second;
1412				pf_status.stateid = pf_status.stateid << 32;
1413			}
1414			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1415		}
1416		break;
1417
1418	case DIOCSTOP:
1419		if (!pf_status.running)
1420			error = ENOENT;
1421		else {
1422			pf_status.running = 0;
1423#ifdef __FreeBSD__
1424			PF_UNLOCK();
1425			error = dehook_pf();
1426			PF_LOCK();
1427			if (error) {
1428				pf_status.running = 1;
1429				DPFPRINTF(PF_DEBUG_MISC,
1430					("pf: pfil unregisteration failed\n"));
1431			}
1432#endif
1433			pf_status.since = time_second;
1434			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1435		}
1436		break;
1437
1438	case DIOCADDRULE: {
1439		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1440		struct pf_ruleset	*ruleset;
1441		struct pf_rule		*rule, *tail;
1442		struct pf_pooladdr	*pa;
1443		int			 rs_num;
1444
1445		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1446		ruleset = pf_find_ruleset(pr->anchor);
1447		if (ruleset == NULL) {
1448			error = EINVAL;
1449			break;
1450		}
1451		rs_num = pf_get_ruleset_number(pr->rule.action);
1452		if (rs_num >= PF_RULESET_MAX) {
1453			error = EINVAL;
1454			break;
1455		}
1456		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1457			error = EINVAL;
1458			break;
1459		}
1460		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1461			printf("ticket: %d != [%d]%d\n", pr->ticket,
1462			    rs_num, ruleset->rules[rs_num].inactive.ticket);
1463			error = EBUSY;
1464			break;
1465		}
1466		if (pr->pool_ticket != ticket_pabuf) {
1467			printf("pool_ticket: %d != %d\n", pr->pool_ticket,
1468			    ticket_pabuf);
1469			error = EBUSY;
1470			break;
1471		}
1472		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1473		if (rule == NULL) {
1474			error = ENOMEM;
1475			break;
1476		}
1477		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1478		rule->anchor = NULL;
1479		rule->kif = NULL;
1480		TAILQ_INIT(&rule->rpool.list);
1481		/* initialize refcounting */
1482		rule->states = 0;
1483		rule->src_nodes = 0;
1484		rule->entries.tqe_prev = NULL;
1485#ifndef INET
1486		if (rule->af == AF_INET) {
1487			pool_put(&pf_rule_pl, rule);
1488			error = EAFNOSUPPORT;
1489			break;
1490		}
1491#endif /* INET */
1492#ifndef INET6
1493		if (rule->af == AF_INET6) {
1494			pool_put(&pf_rule_pl, rule);
1495			error = EAFNOSUPPORT;
1496			break;
1497		}
1498#endif /* INET6 */
1499		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1500		    pf_rulequeue);
1501		if (tail)
1502			rule->nr = tail->nr + 1;
1503		else
1504			rule->nr = 0;
1505		if (rule->ifname[0]) {
1506			rule->kif = pfi_attach_rule(rule->ifname);
1507			if (rule->kif == NULL) {
1508				pool_put(&pf_rule_pl, rule);
1509				error = EINVAL;
1510				break;
1511			}
1512		}
1513
1514#ifdef ALTQ
1515		/* set queue IDs */
1516		if (rule->qname[0] != 0) {
1517			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1518				error = EBUSY;
1519			else if (rule->pqname[0] != 0) {
1520				if ((rule->pqid =
1521				    pf_qname2qid(rule->pqname)) == 0)
1522					error = EBUSY;
1523			} else
1524				rule->pqid = rule->qid;
1525		}
1526#endif
1527		if (rule->tagname[0])
1528			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1529				error = EBUSY;
1530		if (rule->match_tagname[0])
1531			if ((rule->match_tag =
1532			    pf_tagname2tag(rule->match_tagname)) == 0)
1533				error = EBUSY;
1534		if (rule->rt && !rule->direction)
1535			error = EINVAL;
1536		if (pf_rtlabel_add(&rule->src.addr) ||
1537		    pf_rtlabel_add(&rule->dst.addr))
1538			error = EBUSY;
1539		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1540			error = EINVAL;
1541		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1542			error = EINVAL;
1543		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1544			error = EINVAL;
1545		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1546			error = EINVAL;
1547		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1548			error = EINVAL;
1549		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1550			if (pf_tbladdr_setup(ruleset, &pa->addr))
1551				error = EINVAL;
1552
1553		if (rule->overload_tblname[0]) {
1554			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1555			    rule->overload_tblname)) == NULL)
1556				error = EINVAL;
1557			else
1558				rule->overload_tbl->pfrkt_flags |=
1559				    PFR_TFLAG_ACTIVE;
1560		}
1561
1562		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1563		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1564		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1565		    (rule->rt > PF_FASTROUTE)) &&
1566		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1567			error = EINVAL;
1568
1569		if (error) {
1570			pf_rm_rule(NULL, rule);
1571			break;
1572		}
1573		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1574		rule->evaluations = rule->packets = rule->bytes = 0;
1575		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1576		    rule, entries);
1577		break;
1578	}
1579
1580	case DIOCGETRULES: {
1581		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1582		struct pf_ruleset	*ruleset;
1583		struct pf_rule		*tail;
1584		int			 rs_num;
1585
1586		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1587		ruleset = pf_find_ruleset(pr->anchor);
1588		if (ruleset == NULL) {
1589			error = EINVAL;
1590			break;
1591		}
1592		rs_num = pf_get_ruleset_number(pr->rule.action);
1593		if (rs_num >= PF_RULESET_MAX) {
1594			error = EINVAL;
1595			break;
1596		}
1597		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1598		    pf_rulequeue);
1599		if (tail)
1600			pr->nr = tail->nr + 1;
1601		else
1602			pr->nr = 0;
1603		pr->ticket = ruleset->rules[rs_num].active.ticket;
1604		break;
1605	}
1606
1607	case DIOCGETRULE: {
1608		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1609		struct pf_ruleset	*ruleset;
1610		struct pf_rule		*rule;
1611		int			 rs_num, i;
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		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1625			error = EBUSY;
1626			break;
1627		}
1628		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1629		while ((rule != NULL) && (rule->nr != pr->nr))
1630			rule = TAILQ_NEXT(rule, entries);
1631		if (rule == NULL) {
1632			error = EBUSY;
1633			break;
1634		}
1635		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1636		if (pf_anchor_copyout(ruleset, rule, pr)) {
1637			error = EBUSY;
1638			break;
1639		}
1640		pfi_dynaddr_copyout(&pr->rule.src.addr);
1641		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1642		pf_tbladdr_copyout(&pr->rule.src.addr);
1643		pf_tbladdr_copyout(&pr->rule.dst.addr);
1644		pf_rtlabel_copyout(&pr->rule.src.addr);
1645		pf_rtlabel_copyout(&pr->rule.dst.addr);
1646		for (i = 0; i < PF_SKIP_COUNT; ++i)
1647			if (rule->skip[i].ptr == NULL)
1648				pr->rule.skip[i].nr = -1;
1649			else
1650				pr->rule.skip[i].nr =
1651				    rule->skip[i].ptr->nr;
1652		break;
1653	}
1654
1655	case DIOCCHANGERULE: {
1656		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1657		struct pf_ruleset	*ruleset;
1658		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1659		u_int32_t		 nr = 0;
1660		int			 rs_num;
1661
1662		if (!(pcr->action == PF_CHANGE_REMOVE ||
1663		    pcr->action == PF_CHANGE_GET_TICKET) &&
1664		    pcr->pool_ticket != ticket_pabuf) {
1665			error = EBUSY;
1666			break;
1667		}
1668
1669		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1670		    pcr->action > PF_CHANGE_GET_TICKET) {
1671			error = EINVAL;
1672			break;
1673		}
1674		ruleset = pf_find_ruleset(pcr->anchor);
1675		if (ruleset == NULL) {
1676			error = EINVAL;
1677			break;
1678		}
1679		rs_num = pf_get_ruleset_number(pcr->rule.action);
1680		if (rs_num >= PF_RULESET_MAX) {
1681			error = EINVAL;
1682			break;
1683		}
1684
1685		if (pcr->action == PF_CHANGE_GET_TICKET) {
1686			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1687			break;
1688		} else {
1689			if (pcr->ticket !=
1690			    ruleset->rules[rs_num].active.ticket) {
1691				error = EINVAL;
1692				break;
1693			}
1694			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1695				error = EINVAL;
1696				break;
1697			}
1698		}
1699
1700		if (pcr->action != PF_CHANGE_REMOVE) {
1701			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1702			if (newrule == NULL) {
1703				error = ENOMEM;
1704				break;
1705			}
1706			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1707			TAILQ_INIT(&newrule->rpool.list);
1708			/* initialize refcounting */
1709			newrule->states = 0;
1710			newrule->entries.tqe_prev = NULL;
1711#ifndef INET
1712			if (newrule->af == AF_INET) {
1713				pool_put(&pf_rule_pl, newrule);
1714				error = EAFNOSUPPORT;
1715				break;
1716			}
1717#endif /* INET */
1718#ifndef INET6
1719			if (newrule->af == AF_INET6) {
1720				pool_put(&pf_rule_pl, newrule);
1721				error = EAFNOSUPPORT;
1722				break;
1723			}
1724#endif /* INET6 */
1725			if (newrule->ifname[0]) {
1726				newrule->kif = pfi_attach_rule(newrule->ifname);
1727				if (newrule->kif == NULL) {
1728					pool_put(&pf_rule_pl, newrule);
1729					error = EINVAL;
1730					break;
1731				}
1732			} else
1733				newrule->kif = NULL;
1734
1735#ifdef ALTQ
1736			/* set queue IDs */
1737			if (newrule->qname[0] != 0) {
1738				if ((newrule->qid =
1739				    pf_qname2qid(newrule->qname)) == 0)
1740					error = EBUSY;
1741				else if (newrule->pqname[0] != 0) {
1742					if ((newrule->pqid =
1743					    pf_qname2qid(newrule->pqname)) == 0)
1744						error = EBUSY;
1745				} else
1746					newrule->pqid = newrule->qid;
1747			}
1748#endif /* ALTQ */
1749			if (newrule->tagname[0])
1750				if ((newrule->tag =
1751				    pf_tagname2tag(newrule->tagname)) == 0)
1752					error = EBUSY;
1753			if (newrule->match_tagname[0])
1754				if ((newrule->match_tag = pf_tagname2tag(
1755				    newrule->match_tagname)) == 0)
1756					error = EBUSY;
1757			if (newrule->rt && !newrule->direction)
1758				error = EINVAL;
1759			if (pf_rtlabel_add(&newrule->src.addr) ||
1760			    pf_rtlabel_add(&newrule->dst.addr))
1761				error = EBUSY;
1762			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1763				error = EINVAL;
1764			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1765				error = EINVAL;
1766			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1767				error = EINVAL;
1768			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1769				error = EINVAL;
1770			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1771				error = EINVAL;
1772
1773			if (newrule->overload_tblname[0]) {
1774				if ((newrule->overload_tbl = pfr_attach_table(
1775				    ruleset, newrule->overload_tblname)) ==
1776				    NULL)
1777					error = EINVAL;
1778				else
1779					newrule->overload_tbl->pfrkt_flags |=
1780					    PFR_TFLAG_ACTIVE;
1781			}
1782
1783			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1784			if (((((newrule->action == PF_NAT) ||
1785			    (newrule->action == PF_RDR) ||
1786			    (newrule->action == PF_BINAT) ||
1787			    (newrule->rt > PF_FASTROUTE)) &&
1788			    !pcr->anchor[0])) &&
1789			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1790				error = EINVAL;
1791
1792			if (error) {
1793				pf_rm_rule(NULL, newrule);
1794				break;
1795			}
1796			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1797			newrule->evaluations = newrule->packets = 0;
1798			newrule->bytes = 0;
1799		}
1800		pf_empty_pool(&pf_pabuf);
1801
1802		if (pcr->action == PF_CHANGE_ADD_HEAD)
1803			oldrule = TAILQ_FIRST(
1804			    ruleset->rules[rs_num].active.ptr);
1805		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1806			oldrule = TAILQ_LAST(
1807			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1808		else {
1809			oldrule = TAILQ_FIRST(
1810			    ruleset->rules[rs_num].active.ptr);
1811			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1812				oldrule = TAILQ_NEXT(oldrule, entries);
1813			if (oldrule == NULL) {
1814				if (newrule != NULL)
1815					pf_rm_rule(NULL, newrule);
1816				error = EINVAL;
1817				break;
1818			}
1819		}
1820
1821		if (pcr->action == PF_CHANGE_REMOVE)
1822			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1823		else {
1824			if (oldrule == NULL)
1825				TAILQ_INSERT_TAIL(
1826				    ruleset->rules[rs_num].active.ptr,
1827				    newrule, entries);
1828			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1829			    pcr->action == PF_CHANGE_ADD_BEFORE)
1830				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1831			else
1832				TAILQ_INSERT_AFTER(
1833				    ruleset->rules[rs_num].active.ptr,
1834				    oldrule, newrule, entries);
1835		}
1836
1837		nr = 0;
1838		TAILQ_FOREACH(oldrule,
1839		    ruleset->rules[rs_num].active.ptr, entries)
1840			oldrule->nr = nr++;
1841
1842		ruleset->rules[rs_num].active.ticket++;
1843
1844		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1845		pf_remove_if_empty_ruleset(ruleset);
1846
1847		break;
1848	}
1849
1850	case DIOCCLRSTATES: {
1851		struct pf_state		*state;
1852		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1853		int			 killed = 0;
1854
1855		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1856			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1857			    state->u.s.kif->pfik_name)) {
1858				state->timeout = PFTM_PURGE;
1859#if NPFSYNC
1860				/* don't send out individual delete messages */
1861				state->sync_flags = PFSTATE_NOSYNC;
1862#endif
1863				killed++;
1864			}
1865		}
1866		pf_purge_expired_states();
1867		pf_status.states = 0;
1868		psk->psk_af = killed;
1869#if NPFSYNC
1870		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1871#endif
1872		break;
1873	}
1874
1875	case DIOCKILLSTATES: {
1876		struct pf_state		*state;
1877		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1878		int			 killed = 0;
1879
1880		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1881			if ((!psk->psk_af || state->af == psk->psk_af)
1882			    && (!psk->psk_proto || psk->psk_proto ==
1883			    state->proto) &&
1884			    PF_MATCHA(psk->psk_src.neg,
1885			    &psk->psk_src.addr.v.a.addr,
1886			    &psk->psk_src.addr.v.a.mask,
1887			    &state->lan.addr, state->af) &&
1888			    PF_MATCHA(psk->psk_dst.neg,
1889			    &psk->psk_dst.addr.v.a.addr,
1890			    &psk->psk_dst.addr.v.a.mask,
1891			    &state->ext.addr, state->af) &&
1892			    (psk->psk_src.port_op == 0 ||
1893			    pf_match_port(psk->psk_src.port_op,
1894			    psk->psk_src.port[0], psk->psk_src.port[1],
1895			    state->lan.port)) &&
1896			    (psk->psk_dst.port_op == 0 ||
1897			    pf_match_port(psk->psk_dst.port_op,
1898			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1899			    state->ext.port)) &&
1900			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1901			    state->u.s.kif->pfik_name))) {
1902				state->timeout = PFTM_PURGE;
1903				killed++;
1904			}
1905		}
1906		pf_purge_expired_states();
1907		psk->psk_af = killed;
1908		break;
1909	}
1910
1911	case DIOCADDSTATE: {
1912		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1913		struct pf_state		*state;
1914		struct pfi_kif		*kif;
1915
1916		if (ps->state.timeout >= PFTM_MAX &&
1917		    ps->state.timeout != PFTM_UNTIL_PACKET) {
1918			error = EINVAL;
1919			break;
1920		}
1921		state = pool_get(&pf_state_pl, PR_NOWAIT);
1922		if (state == NULL) {
1923			error = ENOMEM;
1924			break;
1925		}
1926		kif = pfi_lookup_create(ps->state.u.ifname);
1927		if (kif == NULL) {
1928			pool_put(&pf_state_pl, state);
1929			error = ENOENT;
1930			break;
1931		}
1932		bcopy(&ps->state, state, sizeof(struct pf_state));
1933		bzero(&state->u, sizeof(state->u));
1934		state->rule.ptr = &pf_default_rule;
1935		state->nat_rule.ptr = NULL;
1936		state->anchor.ptr = NULL;
1937		state->rt_kif = NULL;
1938		state->creation = time_second;
1939		state->pfsync_time = 0;
1940		state->packets[0] = state->packets[1] = 0;
1941		state->bytes[0] = state->bytes[1] = 0;
1942
1943		if (pf_insert_state(kif, state)) {
1944			pfi_maybe_destroy(kif);
1945			pool_put(&pf_state_pl, state);
1946			error = ENOMEM;
1947		}
1948		break;
1949	}
1950
1951	case DIOCGETSTATE: {
1952		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1953		struct pf_state		*state;
1954		u_int32_t		 nr;
1955
1956		nr = 0;
1957		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
1958			if (nr >= ps->nr)
1959				break;
1960			nr++;
1961		}
1962		if (state == NULL) {
1963			error = EBUSY;
1964			break;
1965		}
1966		bcopy(state, &ps->state, sizeof(struct pf_state));
1967		ps->state.rule.nr = state->rule.ptr->nr;
1968		ps->state.nat_rule.nr = (state->nat_rule.ptr == NULL) ?
1969		    -1 : state->nat_rule.ptr->nr;
1970		ps->state.anchor.nr = (state->anchor.ptr == NULL) ?
1971		    -1 : state->anchor.ptr->nr;
1972		ps->state.expire = pf_state_expires(state);
1973		if (ps->state.expire > time_second)
1974			ps->state.expire -= time_second;
1975		else
1976			ps->state.expire = 0;
1977		break;
1978	}
1979
1980	case DIOCGETSTATES: {
1981		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1982		struct pf_state		*state;
1983		struct pf_state		*p, pstore;
1984		struct pfi_kif		*kif;
1985		u_int32_t		 nr = 0;
1986		int			 space = ps->ps_len;
1987
1988		if (space == 0) {
1989			TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states)
1990				nr += kif->pfik_states;
1991			ps->ps_len = sizeof(struct pf_state) * nr;
1992			break;
1993		}
1994
1995		p = ps->ps_states;
1996		TAILQ_FOREACH(kif, &pfi_statehead, pfik_w_states)
1997			RB_FOREACH(state, pf_state_tree_ext_gwy,
1998			    &kif->pfik_ext_gwy) {
1999				int	secs = time_second;
2000
2001				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
2002					break;
2003
2004				bcopy(state, &pstore, sizeof(pstore));
2005				strlcpy(pstore.u.ifname, kif->pfik_name,
2006				    sizeof(pstore.u.ifname));
2007				pstore.rule.nr = state->rule.ptr->nr;
2008				pstore.nat_rule.nr = (state->nat_rule.ptr ==
2009				    NULL) ? -1 : state->nat_rule.ptr->nr;
2010				pstore.anchor.nr = (state->anchor.ptr ==
2011				    NULL) ? -1 : state->anchor.ptr->nr;
2012				pstore.creation = secs - pstore.creation;
2013				pstore.expire = pf_state_expires(state);
2014				if (pstore.expire > secs)
2015					pstore.expire -= secs;
2016				else
2017					pstore.expire = 0;
2018#ifdef __FreeBSD__
2019				PF_COPYOUT(&pstore, p, sizeof(*p), error);
2020#else
2021				error = copyout(&pstore, p, sizeof(*p));
2022#endif
2023				if (error)
2024					goto fail;
2025				p++;
2026				nr++;
2027			}
2028		ps->ps_len = sizeof(struct pf_state) * nr;
2029		break;
2030	}
2031
2032	case DIOCGETSTATUS: {
2033		struct pf_status *s = (struct pf_status *)addr;
2034		bcopy(&pf_status, s, sizeof(struct pf_status));
2035		pfi_fill_oldstatus(s);
2036		break;
2037	}
2038
2039	case DIOCSETSTATUSIF: {
2040		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2041
2042		if (pi->ifname[0] == 0) {
2043			bzero(pf_status.ifname, IFNAMSIZ);
2044			break;
2045		}
2046		if (ifunit(pi->ifname) == NULL) {
2047			error = EINVAL;
2048			break;
2049		}
2050		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
2051		break;
2052	}
2053
2054	case DIOCCLRSTATUS: {
2055		bzero(pf_status.counters, sizeof(pf_status.counters));
2056		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
2057		bzero(pf_status.scounters, sizeof(pf_status.scounters));
2058		if (*pf_status.ifname)
2059			pfi_clr_istats(pf_status.ifname, NULL,
2060			    PFI_FLAG_INSTANCE);
2061		break;
2062	}
2063
2064	case DIOCNATLOOK: {
2065		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2066		struct pf_state		*state;
2067		struct pf_state		 key;
2068		int			 m = 0, direction = pnl->direction;
2069
2070		key.af = pnl->af;
2071		key.proto = pnl->proto;
2072
2073		if (!pnl->proto ||
2074		    PF_AZERO(&pnl->saddr, pnl->af) ||
2075		    PF_AZERO(&pnl->daddr, pnl->af) ||
2076		    !pnl->dport || !pnl->sport)
2077			error = EINVAL;
2078		else {
2079			/*
2080			 * userland gives us source and dest of connection,
2081			 * reverse the lookup so we ask for what happens with
2082			 * the return traffic, enabling us to find it in the
2083			 * state tree.
2084			 */
2085			if (direction == PF_IN) {
2086				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
2087				key.ext.port = pnl->dport;
2088				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
2089				key.gwy.port = pnl->sport;
2090				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
2091			} else {
2092				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
2093				key.lan.port = pnl->dport;
2094				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
2095				key.ext.port = pnl->sport;
2096				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
2097			}
2098			if (m > 1)
2099				error = E2BIG;	/* more than one state */
2100			else if (state != NULL) {
2101				if (direction == PF_IN) {
2102					PF_ACPY(&pnl->rsaddr, &state->lan.addr,
2103					    state->af);
2104					pnl->rsport = state->lan.port;
2105					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
2106					    pnl->af);
2107					pnl->rdport = pnl->dport;
2108				} else {
2109					PF_ACPY(&pnl->rdaddr, &state->gwy.addr,
2110					    state->af);
2111					pnl->rdport = state->gwy.port;
2112					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
2113					    pnl->af);
2114					pnl->rsport = pnl->sport;
2115				}
2116			} else
2117				error = ENOENT;
2118		}
2119		break;
2120	}
2121
2122	case DIOCSETTIMEOUT: {
2123		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2124		int		 old;
2125
2126		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2127		    pt->seconds < 0) {
2128			error = EINVAL;
2129			goto fail;
2130		}
2131		old = pf_default_rule.timeout[pt->timeout];
2132		pf_default_rule.timeout[pt->timeout] = pt->seconds;
2133		pt->seconds = old;
2134		break;
2135	}
2136
2137	case DIOCGETTIMEOUT: {
2138		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2139
2140		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2141			error = EINVAL;
2142			goto fail;
2143		}
2144		pt->seconds = pf_default_rule.timeout[pt->timeout];
2145		break;
2146	}
2147
2148	case DIOCGETLIMIT: {
2149		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2150
2151		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2152			error = EINVAL;
2153			goto fail;
2154		}
2155		pl->limit = pf_pool_limits[pl->index].limit;
2156		break;
2157	}
2158
2159	case DIOCSETLIMIT: {
2160		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2161		int			 old_limit;
2162
2163		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2164		    pf_pool_limits[pl->index].pp == NULL) {
2165			error = EINVAL;
2166			goto fail;
2167		}
2168#ifdef __FreeBSD__
2169		uma_zone_set_max(pf_pool_limits[pl->index].pp, pl->limit);
2170#else
2171		if (pool_sethardlimit(pf_pool_limits[pl->index].pp,
2172		    pl->limit, NULL, 0) != 0) {
2173			error = EBUSY;
2174			goto fail;
2175		}
2176#endif
2177		old_limit = pf_pool_limits[pl->index].limit;
2178		pf_pool_limits[pl->index].limit = pl->limit;
2179		pl->limit = old_limit;
2180		break;
2181	}
2182
2183	case DIOCSETDEBUG: {
2184		u_int32_t	*level = (u_int32_t *)addr;
2185
2186		pf_status.debug = *level;
2187		break;
2188	}
2189
2190	case DIOCCLRRULECTRS: {
2191		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2192		struct pf_rule		*rule;
2193
2194		TAILQ_FOREACH(rule,
2195		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries)
2196			rule->evaluations = rule->packets =
2197			    rule->bytes = 0;
2198		break;
2199	}
2200
2201#ifdef __FreeBSD__
2202	case DIOCGIFSPEED: {
2203		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
2204		struct pf_ifspeed	ps;
2205		struct ifnet		*ifp;
2206
2207		if (psp->ifname[0] != 0) {
2208			/* Can we completely trust user-land? */
2209			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
2210			ifp = ifunit(ps.ifname);
2211			if (ifp != NULL)
2212				psp->baudrate = ifp->if_baudrate;
2213			else
2214				error = EINVAL;
2215		} else
2216			error = EINVAL;
2217		break;
2218	}
2219#endif /* __FreeBSD__ */
2220
2221#ifdef ALTQ
2222	case DIOCSTARTALTQ: {
2223		struct pf_altq		*altq;
2224
2225		/* enable all altq interfaces on active list */
2226		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2227			if (altq->qname[0] == 0) {
2228				error = pf_enable_altq(altq);
2229				if (error != 0)
2230					break;
2231			}
2232		}
2233		if (error == 0)
2234			pf_altq_running = 1;
2235		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2236		break;
2237	}
2238
2239	case DIOCSTOPALTQ: {
2240		struct pf_altq		*altq;
2241
2242		/* disable all altq interfaces on active list */
2243		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2244			if (altq->qname[0] == 0) {
2245				error = pf_disable_altq(altq);
2246				if (error != 0)
2247					break;
2248			}
2249		}
2250		if (error == 0)
2251			pf_altq_running = 0;
2252		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2253		break;
2254	}
2255
2256	case DIOCADDALTQ: {
2257		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2258		struct pf_altq		*altq, *a;
2259
2260		if (pa->ticket != ticket_altqs_inactive) {
2261			error = EBUSY;
2262			break;
2263		}
2264		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2265		if (altq == NULL) {
2266			error = ENOMEM;
2267			break;
2268		}
2269		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
2270
2271		/*
2272		 * if this is for a queue, find the discipline and
2273		 * copy the necessary fields
2274		 */
2275		if (altq->qname[0] != 0) {
2276			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2277				error = EBUSY;
2278				pool_put(&pf_altq_pl, altq);
2279				break;
2280			}
2281			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2282				if (strncmp(a->ifname, altq->ifname,
2283				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2284					altq->altq_disc = a->altq_disc;
2285					break;
2286				}
2287			}
2288		}
2289
2290#ifdef __FreeBSD__
2291		PF_UNLOCK();
2292#endif
2293		error = altq_add(altq);
2294#ifdef __FreeBSD__
2295		PF_LOCK();
2296#endif
2297		if (error) {
2298			pool_put(&pf_altq_pl, altq);
2299			break;
2300		}
2301
2302		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2303		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2304		break;
2305	}
2306
2307	case DIOCGETALTQS: {
2308		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2309		struct pf_altq		*altq;
2310
2311		pa->nr = 0;
2312		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2313			pa->nr++;
2314		pa->ticket = ticket_altqs_active;
2315		break;
2316	}
2317
2318	case DIOCGETALTQ: {
2319		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2320		struct pf_altq		*altq;
2321		u_int32_t		 nr;
2322
2323		if (pa->ticket != ticket_altqs_active) {
2324			error = EBUSY;
2325			break;
2326		}
2327		nr = 0;
2328		altq = TAILQ_FIRST(pf_altqs_active);
2329		while ((altq != NULL) && (nr < pa->nr)) {
2330			altq = TAILQ_NEXT(altq, entries);
2331			nr++;
2332		}
2333		if (altq == NULL) {
2334			error = EBUSY;
2335			break;
2336		}
2337		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2338		break;
2339	}
2340
2341	case DIOCCHANGEALTQ:
2342		/* CHANGEALTQ not supported yet! */
2343		error = ENODEV;
2344		break;
2345
2346	case DIOCGETQSTATS: {
2347		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2348		struct pf_altq		*altq;
2349		u_int32_t		 nr;
2350		int			 nbytes;
2351
2352		if (pq->ticket != ticket_altqs_active) {
2353			error = EBUSY;
2354			break;
2355		}
2356		nbytes = pq->nbytes;
2357		nr = 0;
2358		altq = TAILQ_FIRST(pf_altqs_active);
2359		while ((altq != NULL) && (nr < pq->nr)) {
2360			altq = TAILQ_NEXT(altq, entries);
2361			nr++;
2362		}
2363		if (altq == NULL) {
2364			error = EBUSY;
2365			break;
2366		}
2367#ifdef __FreeBSD__
2368		PF_UNLOCK();
2369#endif
2370		error = altq_getqstats(altq, pq->buf, &nbytes);
2371#ifdef __FreeBSD__
2372		PF_LOCK();
2373#endif
2374		if (error == 0) {
2375			pq->scheduler = altq->scheduler;
2376			pq->nbytes = nbytes;
2377		}
2378		break;
2379	}
2380#endif /* ALTQ */
2381
2382	case DIOCBEGINADDRS: {
2383		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2384
2385		pf_empty_pool(&pf_pabuf);
2386		pp->ticket = ++ticket_pabuf;
2387		break;
2388	}
2389
2390	case DIOCADDADDR: {
2391		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2392
2393#ifndef INET
2394		if (pp->af == AF_INET) {
2395			error = EAFNOSUPPORT;
2396			break;
2397		}
2398#endif /* INET */
2399#ifndef INET6
2400		if (pp->af == AF_INET6) {
2401			error = EAFNOSUPPORT;
2402			break;
2403		}
2404#endif /* INET6 */
2405		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2406		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2407		    pp->addr.addr.type != PF_ADDR_TABLE) {
2408			error = EINVAL;
2409			break;
2410		}
2411		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2412		if (pa == NULL) {
2413			error = ENOMEM;
2414			break;
2415		}
2416		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2417		if (pa->ifname[0]) {
2418			pa->kif = pfi_attach_rule(pa->ifname);
2419			if (pa->kif == NULL) {
2420				pool_put(&pf_pooladdr_pl, pa);
2421				error = EINVAL;
2422				break;
2423			}
2424		}
2425		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2426			pfi_dynaddr_remove(&pa->addr);
2427			pfi_detach_rule(pa->kif);
2428			pool_put(&pf_pooladdr_pl, pa);
2429			error = EINVAL;
2430			break;
2431		}
2432		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2433		break;
2434	}
2435
2436	case DIOCGETADDRS: {
2437		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2438
2439		pp->nr = 0;
2440		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2441		    pp->r_num, 0, 1, 0);
2442		if (pool == NULL) {
2443			error = EBUSY;
2444			break;
2445		}
2446		TAILQ_FOREACH(pa, &pool->list, entries)
2447			pp->nr++;
2448		break;
2449	}
2450
2451	case DIOCGETADDR: {
2452		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2453		u_int32_t		 nr = 0;
2454
2455		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2456		    pp->r_num, 0, 1, 1);
2457		if (pool == NULL) {
2458			error = EBUSY;
2459			break;
2460		}
2461		pa = TAILQ_FIRST(&pool->list);
2462		while ((pa != NULL) && (nr < pp->nr)) {
2463			pa = TAILQ_NEXT(pa, entries);
2464			nr++;
2465		}
2466		if (pa == NULL) {
2467			error = EBUSY;
2468			break;
2469		}
2470		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2471		pfi_dynaddr_copyout(&pp->addr.addr);
2472		pf_tbladdr_copyout(&pp->addr.addr);
2473		pf_rtlabel_copyout(&pp->addr.addr);
2474		break;
2475	}
2476
2477	case DIOCCHANGEADDR: {
2478		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2479		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2480		struct pf_ruleset	*ruleset;
2481
2482		if (pca->action < PF_CHANGE_ADD_HEAD ||
2483		    pca->action > PF_CHANGE_REMOVE) {
2484			error = EINVAL;
2485			break;
2486		}
2487		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2488		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2489		    pca->addr.addr.type != PF_ADDR_TABLE) {
2490			error = EINVAL;
2491			break;
2492		}
2493
2494		ruleset = pf_find_ruleset(pca->anchor);
2495		if (ruleset == NULL) {
2496			error = EBUSY;
2497			break;
2498		}
2499		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2500		    pca->r_num, pca->r_last, 1, 1);
2501		if (pool == NULL) {
2502			error = EBUSY;
2503			break;
2504		}
2505		if (pca->action != PF_CHANGE_REMOVE) {
2506			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2507			if (newpa == NULL) {
2508				error = ENOMEM;
2509				break;
2510			}
2511			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2512#ifndef INET
2513			if (pca->af == AF_INET) {
2514				pool_put(&pf_pooladdr_pl, newpa);
2515				error = EAFNOSUPPORT;
2516				break;
2517			}
2518#endif /* INET */
2519#ifndef INET6
2520			if (pca->af == AF_INET6) {
2521				pool_put(&pf_pooladdr_pl, newpa);
2522				error = EAFNOSUPPORT;
2523				break;
2524			}
2525#endif /* INET6 */
2526			if (newpa->ifname[0]) {
2527				newpa->kif = pfi_attach_rule(newpa->ifname);
2528				if (newpa->kif == NULL) {
2529					pool_put(&pf_pooladdr_pl, newpa);
2530					error = EINVAL;
2531					break;
2532				}
2533			} else
2534				newpa->kif = NULL;
2535			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2536			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2537				pfi_dynaddr_remove(&newpa->addr);
2538				pfi_detach_rule(newpa->kif);
2539				pool_put(&pf_pooladdr_pl, newpa);
2540				error = EINVAL;
2541				break;
2542			}
2543		}
2544
2545		if (pca->action == PF_CHANGE_ADD_HEAD)
2546			oldpa = TAILQ_FIRST(&pool->list);
2547		else if (pca->action == PF_CHANGE_ADD_TAIL)
2548			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2549		else {
2550			int	i = 0;
2551
2552			oldpa = TAILQ_FIRST(&pool->list);
2553			while ((oldpa != NULL) && (i < pca->nr)) {
2554				oldpa = TAILQ_NEXT(oldpa, entries);
2555				i++;
2556			}
2557			if (oldpa == NULL) {
2558				error = EINVAL;
2559				break;
2560			}
2561		}
2562
2563		if (pca->action == PF_CHANGE_REMOVE) {
2564			TAILQ_REMOVE(&pool->list, oldpa, entries);
2565			pfi_dynaddr_remove(&oldpa->addr);
2566			pf_tbladdr_remove(&oldpa->addr);
2567			pfi_detach_rule(oldpa->kif);
2568			pool_put(&pf_pooladdr_pl, oldpa);
2569		} else {
2570			if (oldpa == NULL)
2571				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2572			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2573			    pca->action == PF_CHANGE_ADD_BEFORE)
2574				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2575			else
2576				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2577				    newpa, entries);
2578		}
2579
2580		pool->cur = TAILQ_FIRST(&pool->list);
2581		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2582		    pca->af);
2583		break;
2584	}
2585
2586	case DIOCGETRULESETS: {
2587		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2588		struct pf_ruleset	*ruleset;
2589		struct pf_anchor	*anchor;
2590
2591		pr->path[sizeof(pr->path) - 1] = 0;
2592		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2593			error = EINVAL;
2594			break;
2595		}
2596		pr->nr = 0;
2597		if (ruleset->anchor == NULL) {
2598			/* XXX kludge for pf_main_ruleset */
2599			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2600				if (anchor->parent == NULL)
2601					pr->nr++;
2602		} else {
2603			RB_FOREACH(anchor, pf_anchor_node,
2604			    &ruleset->anchor->children)
2605				pr->nr++;
2606		}
2607		break;
2608	}
2609
2610	case DIOCGETRULESET: {
2611		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2612		struct pf_ruleset	*ruleset;
2613		struct pf_anchor	*anchor;
2614		u_int32_t		 nr = 0;
2615
2616		pr->path[sizeof(pr->path) - 1] = 0;
2617		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2618			error = EINVAL;
2619			break;
2620		}
2621		pr->name[0] = 0;
2622		if (ruleset->anchor == NULL) {
2623			/* XXX kludge for pf_main_ruleset */
2624			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2625				if (anchor->parent == NULL && nr++ == pr->nr) {
2626					strlcpy(pr->name, anchor->name,
2627					    sizeof(pr->name));
2628					break;
2629				}
2630		} else {
2631			RB_FOREACH(anchor, pf_anchor_node,
2632			    &ruleset->anchor->children)
2633				if (nr++ == pr->nr) {
2634					strlcpy(pr->name, anchor->name,
2635					    sizeof(pr->name));
2636					break;
2637				}
2638		}
2639		if (!pr->name[0])
2640			error = EBUSY;
2641		break;
2642	}
2643
2644	case DIOCRCLRTABLES: {
2645		struct pfioc_table *io = (struct pfioc_table *)addr;
2646
2647		if (io->pfrio_esize != 0) {
2648			error = ENODEV;
2649			break;
2650		}
2651		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2652		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2653		break;
2654	}
2655
2656	case DIOCRADDTABLES: {
2657		struct pfioc_table *io = (struct pfioc_table *)addr;
2658
2659		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2660			error = ENODEV;
2661			break;
2662		}
2663		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2664		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2665		break;
2666	}
2667
2668	case DIOCRDELTABLES: {
2669		struct pfioc_table *io = (struct pfioc_table *)addr;
2670
2671		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2672			error = ENODEV;
2673			break;
2674		}
2675		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2676		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2677		break;
2678	}
2679
2680	case DIOCRGETTABLES: {
2681		struct pfioc_table *io = (struct pfioc_table *)addr;
2682
2683		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2684			error = ENODEV;
2685			break;
2686		}
2687		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2688		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2689		break;
2690	}
2691
2692	case DIOCRGETTSTATS: {
2693		struct pfioc_table *io = (struct pfioc_table *)addr;
2694
2695		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2696			error = ENODEV;
2697			break;
2698		}
2699		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2700		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2701		break;
2702	}
2703
2704	case DIOCRCLRTSTATS: {
2705		struct pfioc_table *io = (struct pfioc_table *)addr;
2706
2707		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2708			error = ENODEV;
2709			break;
2710		}
2711		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2712		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2713		break;
2714	}
2715
2716	case DIOCRSETTFLAGS: {
2717		struct pfioc_table *io = (struct pfioc_table *)addr;
2718
2719		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2720			error = ENODEV;
2721			break;
2722		}
2723		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2724		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2725		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2726		break;
2727	}
2728
2729	case DIOCRCLRADDRS: {
2730		struct pfioc_table *io = (struct pfioc_table *)addr;
2731
2732		if (io->pfrio_esize != 0) {
2733			error = ENODEV;
2734			break;
2735		}
2736		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2737		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2738		break;
2739	}
2740
2741	case DIOCRADDADDRS: {
2742		struct pfioc_table *io = (struct pfioc_table *)addr;
2743
2744		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2745			error = ENODEV;
2746			break;
2747		}
2748		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2749		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2750		    PFR_FLAG_USERIOCTL);
2751		break;
2752	}
2753
2754	case DIOCRDELADDRS: {
2755		struct pfioc_table *io = (struct pfioc_table *)addr;
2756
2757		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2758			error = ENODEV;
2759			break;
2760		}
2761		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2762		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2763		    PFR_FLAG_USERIOCTL);
2764		break;
2765	}
2766
2767	case DIOCRSETADDRS: {
2768		struct pfioc_table *io = (struct pfioc_table *)addr;
2769
2770		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2771			error = ENODEV;
2772			break;
2773		}
2774		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2775		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2776		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2777		    PFR_FLAG_USERIOCTL);
2778		break;
2779	}
2780
2781	case DIOCRGETADDRS: {
2782		struct pfioc_table *io = (struct pfioc_table *)addr;
2783
2784		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2785			error = ENODEV;
2786			break;
2787		}
2788		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2789		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2790		break;
2791	}
2792
2793	case DIOCRGETASTATS: {
2794		struct pfioc_table *io = (struct pfioc_table *)addr;
2795
2796		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2797			error = ENODEV;
2798			break;
2799		}
2800		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2801		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2802		break;
2803	}
2804
2805	case DIOCRCLRASTATS: {
2806		struct pfioc_table *io = (struct pfioc_table *)addr;
2807
2808		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2809			error = ENODEV;
2810			break;
2811		}
2812		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2813		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2814		    PFR_FLAG_USERIOCTL);
2815		break;
2816	}
2817
2818	case DIOCRTSTADDRS: {
2819		struct pfioc_table *io = (struct pfioc_table *)addr;
2820
2821		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2822			error = ENODEV;
2823			break;
2824		}
2825		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2826		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2827		    PFR_FLAG_USERIOCTL);
2828		break;
2829	}
2830
2831	case DIOCRINADEFINE: {
2832		struct pfioc_table *io = (struct pfioc_table *)addr;
2833
2834		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2835			error = ENODEV;
2836			break;
2837		}
2838		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2839		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2840		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2841		break;
2842	}
2843
2844	case DIOCOSFPADD: {
2845		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2846		error = pf_osfp_add(io);
2847		break;
2848	}
2849
2850	case DIOCOSFPGET: {
2851		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2852		error = pf_osfp_get(io);
2853		break;
2854	}
2855
2856	case DIOCXBEGIN: {
2857		struct pfioc_trans		*io = (struct pfioc_trans *)
2858						    addr;
2859		static struct pfioc_trans_e	 ioe;
2860		static struct pfr_table		 table;
2861		int				 i;
2862
2863		if (io->esize != sizeof(ioe)) {
2864			error = ENODEV;
2865			goto fail;
2866		}
2867		for (i = 0; i < io->size; i++) {
2868#ifdef __FreeBSD__
2869			PF_COPYIN(io->array+i, &ioe, sizeof(ioe), error);
2870			if (error) {
2871#else
2872			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2873#endif
2874				error = EFAULT;
2875				goto fail;
2876			}
2877			switch (ioe.rs_num) {
2878#ifdef ALTQ
2879			case PF_RULESET_ALTQ:
2880				if (ioe.anchor[0]) {
2881					error = EINVAL;
2882					goto fail;
2883				}
2884				if ((error = pf_begin_altq(&ioe.ticket)))
2885					goto fail;
2886				break;
2887#endif /* ALTQ */
2888			case PF_RULESET_TABLE:
2889				bzero(&table, sizeof(table));
2890				strlcpy(table.pfrt_anchor, ioe.anchor,
2891				    sizeof(table.pfrt_anchor));
2892				if ((error = pfr_ina_begin(&table,
2893				    &ioe.ticket, NULL, 0)))
2894					goto fail;
2895				break;
2896			default:
2897				if ((error = pf_begin_rules(&ioe.ticket,
2898				    ioe.rs_num, ioe.anchor)))
2899					goto fail;
2900				break;
2901			}
2902#ifdef __FreeBSD__
2903			PF_COPYOUT(&ioe, io->array+i, sizeof(io->array[i]),
2904			    error);
2905			if (error) {
2906#else
2907			if (copyout(&ioe, io->array+i, sizeof(io->array[i]))) {
2908#endif
2909				error = EFAULT;
2910				goto fail;
2911			}
2912		}
2913		break;
2914	}
2915
2916	case DIOCXROLLBACK: {
2917		struct pfioc_trans		*io = (struct pfioc_trans *)
2918						    addr;
2919		static struct pfioc_trans_e	 ioe;
2920		static struct pfr_table		 table;
2921		int				 i;
2922
2923		if (io->esize != sizeof(ioe)) {
2924			error = ENODEV;
2925			goto fail;
2926		}
2927		for (i = 0; i < io->size; i++) {
2928#ifdef __FreeBSD__
2929			PF_COPYIN(io->array+i, &ioe, sizeof(ioe), error);
2930			if (error) {
2931#else
2932			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2933#endif
2934				error = EFAULT;
2935				goto fail;
2936			}
2937			switch (ioe.rs_num) {
2938#ifdef ALTQ
2939			case PF_RULESET_ALTQ:
2940				if (ioe.anchor[0]) {
2941					error = EINVAL;
2942					goto fail;
2943				}
2944				if ((error = pf_rollback_altq(ioe.ticket)))
2945					goto fail; /* really bad */
2946				break;
2947#endif /* ALTQ */
2948			case PF_RULESET_TABLE:
2949				bzero(&table, sizeof(table));
2950				strlcpy(table.pfrt_anchor, ioe.anchor,
2951				    sizeof(table.pfrt_anchor));
2952				if ((error = pfr_ina_rollback(&table,
2953				    ioe.ticket, NULL, 0)))
2954					goto fail; /* really bad */
2955				break;
2956			default:
2957				if ((error = pf_rollback_rules(ioe.ticket,
2958				    ioe.rs_num, ioe.anchor)))
2959					goto fail; /* really bad */
2960				break;
2961			}
2962		}
2963		break;
2964	}
2965
2966	case DIOCXCOMMIT: {
2967		struct pfioc_trans		*io = (struct pfioc_trans *)
2968						    addr;
2969		static struct pfioc_trans_e	 ioe;
2970		static struct pfr_table		 table;
2971		struct pf_ruleset		*rs;
2972		int				 i;
2973
2974		if (io->esize != sizeof(ioe)) {
2975			error = ENODEV;
2976			goto fail;
2977		}
2978		/* first makes sure everything will succeed */
2979		for (i = 0; i < io->size; i++) {
2980#ifdef __FreeBSD__
2981			PF_COPYIN(io->array+i, &ioe, sizeof(ioe), error);
2982			if (error) {
2983#else
2984			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
2985#endif
2986				error = EFAULT;
2987				goto fail;
2988			}
2989			switch (ioe.rs_num) {
2990#ifdef ALTQ
2991			case PF_RULESET_ALTQ:
2992				if (ioe.anchor[0]) {
2993					error = EINVAL;
2994					goto fail;
2995				}
2996				if (!altqs_inactive_open || ioe.ticket !=
2997				    ticket_altqs_inactive) {
2998					error = EBUSY;
2999					goto fail;
3000				}
3001				break;
3002#endif /* ALTQ */
3003			case PF_RULESET_TABLE:
3004				rs = pf_find_ruleset(ioe.anchor);
3005				if (rs == NULL || !rs->topen || ioe.ticket !=
3006				     rs->tticket) {
3007					error = EBUSY;
3008					goto fail;
3009				}
3010				break;
3011			default:
3012				if (ioe.rs_num < 0 || ioe.rs_num >=
3013				    PF_RULESET_MAX) {
3014					error = EINVAL;
3015					goto fail;
3016				}
3017				rs = pf_find_ruleset(ioe.anchor);
3018				if (rs == NULL ||
3019				    !rs->rules[ioe.rs_num].inactive.open ||
3020				    rs->rules[ioe.rs_num].inactive.ticket !=
3021				    ioe.ticket) {
3022					error = EBUSY;
3023					goto fail;
3024				}
3025				break;
3026			}
3027		}
3028		/* now do the commit - no errors should happen here */
3029		for (i = 0; i < io->size; i++) {
3030#ifdef __FreeBSD__
3031			PF_COPYIN(io->array+i, &ioe, sizeof(ioe), error);
3032			if (error) {
3033#else
3034			if (copyin(io->array+i, &ioe, sizeof(ioe))) {
3035#endif
3036				error = EFAULT;
3037				goto fail;
3038			}
3039			switch (ioe.rs_num) {
3040#ifdef ALTQ
3041			case PF_RULESET_ALTQ:
3042				if ((error = pf_commit_altq(ioe.ticket)))
3043					goto fail; /* really bad */
3044				break;
3045#endif /* ALTQ */
3046			case PF_RULESET_TABLE:
3047				bzero(&table, sizeof(table));
3048				strlcpy(table.pfrt_anchor, ioe.anchor,
3049				    sizeof(table.pfrt_anchor));
3050				if ((error = pfr_ina_commit(&table, ioe.ticket,
3051				    NULL, NULL, 0)))
3052					goto fail; /* really bad */
3053				break;
3054			default:
3055				if ((error = pf_commit_rules(ioe.ticket,
3056				    ioe.rs_num, ioe.anchor)))
3057					goto fail; /* really bad */
3058				break;
3059			}
3060		}
3061		break;
3062	}
3063
3064	case DIOCGETSRCNODES: {
3065		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3066		struct pf_src_node	*n;
3067		struct pf_src_node *p, pstore;
3068		u_int32_t		 nr = 0;
3069		int			 space = psn->psn_len;
3070
3071		if (space == 0) {
3072			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
3073				nr++;
3074			psn->psn_len = sizeof(struct pf_src_node) * nr;
3075			break;
3076		}
3077
3078		p = psn->psn_src_nodes;
3079		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3080			int	secs = time_second, diff;
3081
3082			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3083				break;
3084
3085			bcopy(n, &pstore, sizeof(pstore));
3086			if (n->rule.ptr != NULL)
3087				pstore.rule.nr = n->rule.ptr->nr;
3088			pstore.creation = secs - pstore.creation;
3089			if (pstore.expire > secs)
3090				pstore.expire -= secs;
3091			else
3092				pstore.expire = 0;
3093
3094			/* adjust the connection rate estimate */
3095			diff = secs - n->conn_rate.last;
3096			if (diff >= n->conn_rate.seconds)
3097				pstore.conn_rate.count = 0;
3098			else
3099				pstore.conn_rate.count -=
3100				    n->conn_rate.count * diff /
3101				    n->conn_rate.seconds;
3102
3103#ifdef __FreeBSD__
3104			PF_COPYOUT(&pstore, p, sizeof(*p), error);
3105#else
3106			error = copyout(&pstore, p, sizeof(*p));
3107#endif
3108			if (error)
3109				goto fail;
3110			p++;
3111			nr++;
3112		}
3113		psn->psn_len = sizeof(struct pf_src_node) * nr;
3114		break;
3115	}
3116
3117	case DIOCCLRSRCNODES: {
3118		struct pf_src_node	*n;
3119		struct pf_state		*state;
3120
3121		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3122			state->src_node = NULL;
3123			state->nat_src_node = NULL;
3124		}
3125		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3126			n->expire = 1;
3127			n->states = 0;
3128		}
3129		pf_purge_expired_src_nodes();
3130		pf_status.src_nodes = 0;
3131		break;
3132	}
3133
3134	case DIOCSETHOSTID: {
3135		u_int32_t	*hostid = (u_int32_t *)addr;
3136
3137		if (*hostid == 0)
3138			pf_status.hostid = arc4random();
3139		else
3140			pf_status.hostid = *hostid;
3141		break;
3142	}
3143
3144	case DIOCOSFPFLUSH:
3145		pf_osfp_flush();
3146		break;
3147
3148	case DIOCIGETIFACES: {
3149		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3150
3151		if (io->pfiio_esize != sizeof(struct pfi_if)) {
3152			error = ENODEV;
3153			break;
3154		}
3155		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3156		    &io->pfiio_size, io->pfiio_flags);
3157		break;
3158	}
3159
3160	case DIOCICLRISTATS: {
3161		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3162
3163		error = pfi_clr_istats(io->pfiio_name, &io->pfiio_nzero,
3164		    io->pfiio_flags);
3165		break;
3166	}
3167
3168	case DIOCSETIFFLAG: {
3169		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3170
3171		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3172		break;
3173	}
3174
3175	case DIOCCLRIFFLAG: {
3176		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3177
3178		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3179		break;
3180	}
3181
3182	default:
3183		error = ENODEV;
3184		break;
3185	}
3186fail:
3187#ifdef __FreeBSD__
3188	PF_UNLOCK();
3189#else
3190	splx(s);
3191#endif
3192	return (error);
3193}
3194
3195#ifdef __FreeBSD__
3196/*
3197 * XXX - Check for version missmatch!!!
3198 */
3199static void
3200pf_clear_states(void)
3201{
3202	struct pf_state		*state;
3203
3204	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3205		state->timeout = PFTM_PURGE;
3206#if NPFSYNC
3207		/* don't send out individual delete messages */
3208		state->sync_flags = PFSTATE_NOSYNC;
3209#endif
3210	}
3211	pf_purge_expired_states();
3212	pf_status.states = 0;
3213#if 0 /* NPFSYNC */
3214/*
3215 * XXX This is called on module unload, we do not want to sync that over? */
3216 */
3217	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
3218#endif
3219}
3220
3221static int
3222pf_clear_tables(void)
3223{
3224	struct pfioc_table io;
3225	int error;
3226
3227	bzero(&io, sizeof(io));
3228
3229	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3230	    io.pfrio_flags);
3231
3232	return (error);
3233}
3234
3235static void
3236pf_clear_srcnodes(void)
3237{
3238	struct pf_src_node	*n;
3239	struct pf_state		*state;
3240
3241	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3242		state->src_node = NULL;
3243		state->nat_src_node = NULL;
3244	}
3245	RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3246		n->expire = 1;
3247		n->states = 0;
3248	}
3249	pf_purge_expired_src_nodes();
3250	pf_status.src_nodes = 0;
3251}
3252/*
3253 * XXX - Check for version missmatch!!!
3254 */
3255
3256/*
3257 * Duplicate pfctl -Fa operation to get rid of as much as we can.
3258 */
3259static int
3260shutdown_pf(void)
3261{
3262	int error = 0;
3263	u_int32_t t[5];
3264	char nn = '\0';
3265
3266	callout_stop(&pf_expire_to);
3267
3268	pf_status.running = 0;
3269	do {
3270		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn))
3271		    != 0) {
3272			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3273			break;
3274		}
3275		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn))
3276		    != 0) {
3277			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3278			break;		/* XXX: rollback? */
3279		}
3280		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))
3281		    != 0) {
3282			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3283			break;		/* XXX: rollback? */
3284		}
3285		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3286		    != 0) {
3287			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3288			break;		/* XXX: rollback? */
3289		}
3290		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3291		    != 0) {
3292			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3293			break;		/* XXX: rollback? */
3294		}
3295
3296		/* XXX: these should always succeed here */
3297		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3298		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3299		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3300		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3301		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3302
3303		if ((error = pf_clear_tables()) != 0)
3304			break;
3305
3306#ifdef ALTQ
3307		if ((error = pf_begin_altq(&t[0])) != 0) {
3308			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3309			break;
3310		}
3311		pf_commit_altq(t[0]);
3312#endif
3313
3314		pf_clear_states();
3315
3316		pf_clear_srcnodes();
3317
3318		/* status does not use malloced mem so no need to cleanup */
3319		/* fingerprints and interfaces have thier own cleanup code */
3320	} while(0);
3321
3322        return (error);
3323}
3324
3325static int
3326pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3327    struct inpcb *inp)
3328{
3329	/*
3330	 * XXX Wed Jul 9 22:03:16 2003 UTC
3331	 * OpenBSD has changed its byte ordering convention on ip_len/ip_off
3332	 * in network stack. OpenBSD's network stack have converted
3333	 * ip_len/ip_off to host byte order frist as FreeBSD.
3334	 * Now this is not true anymore , so we should convert back to network
3335	 * byte order.
3336	 */
3337	struct ip *h = NULL;
3338	int chk;
3339
3340	if ((*m)->m_pkthdr.len >= (int)sizeof(struct ip)) {
3341		/* if m_pkthdr.len is less than ip header, pf will handle. */
3342		h = mtod(*m, struct ip *);
3343	        HTONS(h->ip_len);
3344	        HTONS(h->ip_off);
3345	}
3346	chk = pf_test(PF_IN, ifp, m, NULL, inp);
3347	if (chk && *m) {
3348		m_freem(*m);
3349		*m = NULL;
3350	}
3351	if (*m != NULL) {
3352		/* pf_test can change ip header location */
3353		h = mtod(*m, struct ip *);
3354		NTOHS(h->ip_len);
3355		NTOHS(h->ip_off);
3356	}
3357	return chk;
3358}
3359
3360static int
3361pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3362    struct inpcb *inp)
3363{
3364	/*
3365	 * XXX Wed Jul 9 22:03:16 2003 UTC
3366	 * OpenBSD has changed its byte ordering convention on ip_len/ip_off
3367	 * in network stack. OpenBSD's network stack have converted
3368	 * ip_len/ip_off to host byte order frist as FreeBSD.
3369	 * Now this is not true anymore , so we should convert back to network
3370	 * byte order.
3371	 */
3372	struct ip *h = NULL;
3373	int chk;
3374
3375	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3376	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3377		in_delayed_cksum(*m);
3378		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3379	}
3380	if ((*m)->m_pkthdr.len >= (int)sizeof(*h)) {
3381		/* if m_pkthdr.len is less than ip header, pf will handle. */
3382		h = mtod(*m, struct ip *);
3383	        HTONS(h->ip_len);
3384	        HTONS(h->ip_off);
3385	}
3386	chk = pf_test(PF_OUT, ifp, m, NULL, inp);
3387	if (chk && *m) {
3388		m_freem(*m);
3389		*m = NULL;
3390	}
3391	if (*m != NULL) {
3392		/* pf_test can change ip header location */
3393		h = mtod(*m, struct ip *);
3394		NTOHS(h->ip_len);
3395		NTOHS(h->ip_off);
3396	}
3397	return chk;
3398}
3399
3400#ifdef INET6
3401static int
3402pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3403    struct inpcb *inp)
3404{
3405	/*
3406	 * IPv6 does not affected ip_len/ip_off byte order changes.
3407	 */
3408	int chk;
3409
3410	chk = pf_test6(PF_IN, ifp, m, NULL, inp);
3411	if (chk && *m) {
3412		m_freem(*m);
3413		*m = NULL;
3414	}
3415	return chk;
3416}
3417
3418static int
3419pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir,
3420    struct inpcb *inp)
3421{
3422	/*
3423	 * IPv6 does not affected ip_len/ip_off byte order changes.
3424	 */
3425	int chk;
3426
3427	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3428	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3429		in_delayed_cksum(*m);
3430		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3431	}
3432	chk = pf_test6(PF_OUT, ifp, m, NULL, inp);
3433	if (chk && *m) {
3434		m_freem(*m);
3435		*m = NULL;
3436	}
3437	return chk;
3438}
3439#endif /* INET6 */
3440
3441static int
3442hook_pf(void)
3443{
3444	struct pfil_head *pfh_inet;
3445#ifdef INET6
3446	struct pfil_head *pfh_inet6;
3447#endif
3448
3449	PF_ASSERT(MA_NOTOWNED);
3450
3451	if (pf_pfil_hooked)
3452		return (0);
3453
3454	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3455	if (pfh_inet == NULL)
3456		return (ESRCH); /* XXX */
3457	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet);
3458	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet);
3459#ifdef INET6
3460	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3461	if (pfh_inet6 == NULL) {
3462		pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3463		    pfh_inet);
3464		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3465		    pfh_inet);
3466		return (ESRCH); /* XXX */
3467	}
3468	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK, pfh_inet6);
3469	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK, pfh_inet6);
3470#endif
3471
3472	pf_pfil_hooked = 1;
3473	return (0);
3474}
3475
3476static int
3477dehook_pf(void)
3478{
3479	struct pfil_head *pfh_inet;
3480#ifdef INET6
3481	struct pfil_head *pfh_inet6;
3482#endif
3483
3484	PF_ASSERT(MA_NOTOWNED);
3485
3486	if (pf_pfil_hooked == 0)
3487		return (0);
3488
3489	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3490	if (pfh_inet == NULL)
3491		return (ESRCH); /* XXX */
3492	pfil_remove_hook(pf_check_in, NULL, PFIL_IN | PFIL_WAITOK,
3493	    pfh_inet);
3494	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT | PFIL_WAITOK,
3495	    pfh_inet);
3496#ifdef INET6
3497	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3498	if (pfh_inet6 == NULL)
3499		return (ESRCH); /* XXX */
3500	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN | PFIL_WAITOK,
3501	    pfh_inet6);
3502	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_WAITOK,
3503	    pfh_inet6);
3504#endif
3505
3506	pf_pfil_hooked = 0;
3507	return (0);
3508}
3509
3510static int
3511pf_load(void)
3512{
3513	init_zone_var();
3514	init_pf_mutex();
3515	pf_dev = make_dev(&pf_cdevsw, 0, 0, 0, 0600, PF_NAME);
3516	if (pfattach() < 0) {
3517		destroy_dev(pf_dev);
3518		destroy_pf_mutex();
3519		return (ENOMEM);
3520	}
3521	return (0);
3522}
3523
3524static int
3525pf_unload(void)
3526{
3527	int error = 0;
3528
3529	PF_LOCK();
3530	pf_status.running = 0;
3531	PF_UNLOCK();
3532	error = dehook_pf();
3533	if (error) {
3534		/*
3535		 * Should not happen!
3536		 * XXX Due to error code ESRCH, kldunload will show
3537		 * a message like 'No such process'.
3538		 */
3539		printf("%s : pfil unregisteration fail\n", __FUNCTION__);
3540		return error;
3541	}
3542	PF_LOCK();
3543	shutdown_pf();
3544	pfi_cleanup();
3545	pf_osfp_flush();
3546	pf_osfp_cleanup();
3547	cleanup_pf_zone();
3548	PF_UNLOCK();
3549	destroy_dev(pf_dev);
3550	destroy_pf_mutex();
3551	return error;
3552}
3553
3554static int
3555pf_modevent(module_t mod, int type, void *data)
3556{
3557	int error = 0;
3558
3559	switch(type) {
3560	case MOD_LOAD:
3561		error = pf_load();
3562		break;
3563
3564	case MOD_UNLOAD:
3565		error = pf_unload();
3566		break;
3567	default:
3568		error = EINVAL;
3569		break;
3570	}
3571	return error;
3572}
3573
3574static moduledata_t pf_mod = {
3575	"pf",
3576	pf_modevent,
3577	0
3578};
3579
3580DECLARE_MODULE(pf, pf_mod, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_FIRST);
3581MODULE_VERSION(pf, PF_MODVER);
3582#endif	/* __FreeBSD__ */
3583