xref: /dflybsd-src/sys/net/pf/pf_ioctl.c (revision 5f39c7e70ca0960d1868c75a449064df712dbb10)
1 /*	$OpenBSD: pf_ioctl.c,v 1.209 2008/06/29 08:42:15 mcbride Exp $ */
2 /*add $OpenBSD: pf_ioctl.c,v 1.212 2009/02/15 20:42:33 mbalmer Exp $ */
3 
4 /*
5  * Copyright (c) 2010 The DragonFly Project.  All rights reserved.
6  *
7  * Copyright (c) 2001 Daniel Hartmeier
8  * Copyright (c) 2002,2003 Henning Brauer
9  * All rights reserved.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  *
15  *    - Redistributions of source code must retain the above copyright
16  *      notice, this list of conditions and the following disclaimer.
17  *    - Redistributions in binary form must reproduce the above
18  *      copyright notice, this list of conditions and the following
19  *      disclaimer in the documentation and/or other materials provided
20  *      with the distribution.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
25  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
26  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
28  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
29  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
30  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
32  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33  * POSSIBILITY OF SUCH DAMAGE.
34  *
35  * Effort sponsored in part by the Defense Advanced Research Projects
36  * Agency (DARPA) and Air Force Research Laboratory, Air Force
37  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
38  *
39  */
40 
41 #include "opt_inet.h"
42 #include "opt_inet6.h"
43 
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/conf.h>
47 #include <sys/device.h>
48 #include <sys/mbuf.h>
49 #include <sys/filio.h>
50 #include <sys/fcntl.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/kernel.h>
54 #include <sys/kthread.h>
55 #include <sys/time.h>
56 #include <sys/proc.h>
57 #include <sys/malloc.h>
58 #include <sys/module.h>
59 #include <sys/lock.h>
60 
61 #include <sys/thread2.h>
62 
63 #include <net/if.h>
64 #include <net/if_types.h>
65 #include <net/route.h>
66 
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/ip.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_icmp.h>
73 
74 #include <net/pf/pfvar.h>
75 #include <sys/md5.h>
76 
77 #include <net/pf/if_pflog.h>
78 #include <net/pf/if_pfsync.h>
79 
80 #ifdef INET6
81 #include <netinet/ip6.h>
82 #include <netinet/in_pcb.h>
83 #endif /* INET6 */
84 
85 #ifdef ALTQ
86 #include <net/altq/altq.h>
87 #endif
88 
89 #include <machine/limits.h>
90 #include <net/pfil.h>
91 #include <sys/mutex.h>
92 
93 u_int rt_numfibs = RT_NUMFIBS;
94 
95 void			 pfattach(void);
96 struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
97 			    u_int8_t, u_int8_t, u_int8_t);
98 
99 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
100 void			 pf_empty_pool(struct pf_palist *);
101 #ifdef ALTQ
102 int			 pf_begin_altq(u_int32_t *);
103 int			 pf_rollback_altq(u_int32_t);
104 int			 pf_commit_altq(u_int32_t);
105 int			 pf_enable_altq(struct pf_altq *);
106 int			 pf_disable_altq(struct pf_altq *);
107 #endif /* ALTQ */
108 int			 pf_begin_rules(u_int32_t *, int, const char *);
109 int			 pf_rollback_rules(u_int32_t, int, char *);
110 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
111 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
112 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
113 int			 pf_commit_rules(u_int32_t, int, char *);
114 int			 pf_addr_setup(struct pf_ruleset *,
115 			    struct pf_addr_wrap *, sa_family_t);
116 void			 pf_addr_copyout(struct pf_addr_wrap *);
117 
118 struct pf_rule		 pf_default_rule;
119 struct lock		 pf_consistency_lock;
120 #ifdef ALTQ
121 static int		 pf_altq_running;
122 #endif
123 
124 #define	TAGID_MAX	 50000
125 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
126 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
127 
128 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
129 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
130 #endif
131 u_int16_t		 tagname2tag(struct pf_tags *, char *);
132 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
133 void			 tag_unref(struct pf_tags *, u_int16_t);
134 int			 pf_rtlabel_add(struct pf_addr_wrap *);
135 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
136 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
137 
138 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
139 
140 static cdev_t	pf_dev;
141 
142 static MALLOC_DEFINE(M_PFRULEPL, "pfrulepl", "pf rule pool list");
143 static MALLOC_DEFINE(M_PFALTQPL, "pfaltqpl", "pf altq pool list");
144 static MALLOC_DEFINE(M_PFPOOLADDRPL, "pfpooladdrpl", "pf pool address pool list");
145 static MALLOC_DEFINE(M_PFFRENTPL, "pffrent", "pf frent pool list");
146 
147 
148 /*
149  * XXX - These are new and need to be checked when moveing to a new version
150  */
151 static void		 pf_clear_states(void);
152 static int		 pf_clear_tables(void);
153 static void		 pf_clear_srcnodes(void);
154 /*
155  * XXX - These are new and need to be checked when moveing to a new version
156  */
157 
158 /*
159  * Wrapper functions for pfil(9) hooks
160  */
161 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
162 		int dir);
163 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
164 		int dir);
165 #ifdef INET6
166 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
167 		int dir);
168 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
169 		int dir);
170 #endif
171 
172 static int 		 hook_pf(void);
173 static int 		 dehook_pf(void);
174 static int 		 shutdown_pf(void);
175 static int 		 pf_load(void);
176 static int 		 pf_unload(void);
177 
178 d_open_t	pfopen;
179 d_close_t	pfclose;
180 d_ioctl_t	pfioctl;
181 
182 static struct dev_ops pf_ops = {	    /* XXX convert to port model */
183 	{ PF_NAME, 73, 0 },
184 	.d_open =	pfopen,
185 	.d_close =	pfclose,
186 	.d_ioctl =	pfioctl
187 };
188 
189 static volatile int pf_pfil_hooked = 0;
190 int pf_end_threads = 0;
191 
192 int debug_pfugidhack = 0;
193 SYSCTL_INT(_debug, OID_AUTO, pfugidhack, CTLFLAG_RW, &debug_pfugidhack, 0,
194 	"Enable/disable pf user/group rules mpsafe hack");
195 
196 void
197 pfattach(void)
198 {
199 	u_int32_t *my_timeout = pf_default_rule.timeout;
200 	int nn;
201 
202 	if (!rn_inithead((void **)&pf_maskhead, NULL, 0)) {
203 		kprintf("pf mask radix tree create failed\n");
204 		return;
205 	}
206 	kmalloc_create(&pf_state_pl, "pf state pool list");
207 	kmalloc_raise_limit(pf_state_pl, 0);
208 	kmalloc_create(&pf_frent_pl, "pf fragment pool list");
209 	kmalloc_raise_limit(pf_frent_pl, 0);
210 	kmalloc_create(&pf_cent_pl, "pf cent pool list");
211 	kmalloc_raise_limit(pf_cent_pl, 0);
212 
213 	pfr_initialize();
214 	pfi_initialize();
215 	pf_osfp_initialize();
216 
217 	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
218 	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
219 	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
220 	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
221 	if (ctob(physmem) <= 100*1024*1024)
222 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
223 		    PFR_KENTRY_HIWAT_SMALL;
224 
225 	for (nn = 0; nn < ncpus; ++nn) {
226 		RB_INIT(&tree_src_tracking[nn]);
227 		RB_INIT(&tree_id[nn]);
228 	}
229 	RB_INIT(&pf_anchors);
230 	pf_init_ruleset(&pf_main_ruleset);
231 	TAILQ_INIT(&pf_altqs[0]);
232 	TAILQ_INIT(&pf_altqs[1]);
233 	TAILQ_INIT(&pf_pabuf);
234 	pf_altqs_active = &pf_altqs[0];
235 	pf_altqs_inactive = &pf_altqs[1];
236 	for (nn = 0; nn < ncpus; ++nn)
237 		TAILQ_INIT(&state_list[nn]);
238 
239 	/* default rule should never be garbage collected */
240 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
241 	pf_default_rule.action = PF_PASS;
242 	pf_default_rule.nr = (uint32_t)(-1);
243 	pf_default_rule.rtableid = -1;
244 
245 	/* initialize default timeouts */
246 	my_timeout[PFTM_TCP_FIRST_PACKET] = 120;	/* First TCP packet */
247 	my_timeout[PFTM_TCP_OPENING] = 30; 		/* No response yet */
248 	my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;	/* Established */
249 	my_timeout[PFTM_TCP_CLOSING] = 15 * 60;		/* Half closed */
250 	my_timeout[PFTM_TCP_FIN_WAIT] = 45;		/* Got both FINs */
251 	my_timeout[PFTM_TCP_CLOSED] = 90;		/* Got a RST */
252 	my_timeout[PFTM_UDP_FIRST_PACKET] = 60;		/* First UDP packet */
253 	my_timeout[PFTM_UDP_SINGLE] = 30;		/* Unidirectional */
254 	my_timeout[PFTM_UDP_MULTIPLE] = 60;		/* Bidirectional */
255 	my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;	/* First ICMP packet */
256 	my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;		/* Got error response */
257 	my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;	/* First packet */
258 	my_timeout[PFTM_OTHER_SINGLE] = 30;		/* Unidirectional */
259 	my_timeout[PFTM_OTHER_MULTIPLE] = 60;		/* Bidirectional */
260 	my_timeout[PFTM_FRAG] = 30;			/* Fragment expire */
261 	my_timeout[PFTM_INTERVAL] = 10;			/* Expire interval */
262 	my_timeout[PFTM_SRC_NODE] = 0;		/* Source Tracking */
263 	my_timeout[PFTM_TS_DIFF] = 30;		/* Allowed TS diff */
264 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
265 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
266 
267 	pf_normalize_init();
268 	bzero(&pf_status, sizeof(pf_status));
269 	pf_status.debug = PF_DEBUG_URGENT;
270 	/* XXX do our best to avoid a conflict */
271 	pf_status.hostid = karc4random();
272 
273 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
274 		panic("pfpurge thread");
275 }
276 
277 int
278 pfopen(struct dev_open_args *ap)
279 {
280 	lwkt_gettoken(&pf_token);
281 	cdev_t dev = ap->a_head.a_dev;
282 	if (minor(dev) >= 1) {
283 		lwkt_reltoken(&pf_token);
284 		return (ENXIO);
285 	}
286 	lwkt_reltoken(&pf_token);
287 	return (0);
288 }
289 
290 int
291 pfclose(struct dev_close_args *ap)
292 {
293 	lwkt_gettoken(&pf_token);
294 	cdev_t dev = ap->a_head.a_dev;
295 	if (minor(dev) >= 1) {
296 		lwkt_reltoken(&pf_token);
297 		return (ENXIO);
298 	}
299 	lwkt_reltoken(&pf_token);
300 	return (0);
301 }
302 
303 struct pf_pool *
304 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
305     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
306     u_int8_t check_ticket)
307 {
308 	struct pf_ruleset	*ruleset;
309 	struct pf_rule		*rule;
310 	int			 rs_num;
311 
312 	ruleset = pf_find_ruleset(anchor);
313 	if (ruleset == NULL)
314 		return (NULL);
315 	rs_num = pf_get_ruleset_number(rule_action);
316 	if (rs_num >= PF_RULESET_MAX)
317 		return (NULL);
318 	if (active) {
319 		if (check_ticket && ticket !=
320 		    ruleset->rules[rs_num].active.ticket)
321 			return (NULL);
322 		if (r_last)
323 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
324 			    pf_rulequeue);
325 		else
326 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
327 	} else {
328 		if (check_ticket && ticket !=
329 		    ruleset->rules[rs_num].inactive.ticket)
330 			return (NULL);
331 		if (r_last)
332 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
333 			    pf_rulequeue);
334 		else
335 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
336 	}
337 	if (!r_last) {
338 		while ((rule != NULL) && (rule->nr != rule_number))
339 			rule = TAILQ_NEXT(rule, entries);
340 	}
341 	if (rule == NULL)
342 		return (NULL);
343 
344 	return (&rule->rpool);
345 }
346 
347 void
348 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
349 {
350 	struct pf_pooladdr	*mv_pool_pa;
351 
352 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
353 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
354 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
355 	}
356 }
357 
358 void
359 pf_empty_pool(struct pf_palist *poola)
360 {
361 	struct pf_pooladdr	*empty_pool_pa;
362 
363 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
364 		pfi_dynaddr_remove(&empty_pool_pa->addr);
365 		pf_tbladdr_remove(&empty_pool_pa->addr);
366 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
367 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
368 		kfree(empty_pool_pa, M_PFPOOLADDRPL);
369 	}
370 }
371 
372 void
373 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
374 {
375 	if (rulequeue != NULL) {
376 		if (rule->states_cur <= 0) {
377 			/*
378 			 * XXX - we need to remove the table *before* detaching
379 			 * the rule to make sure the table code does not delete
380 			 * the anchor under our feet.
381 			 */
382 			pf_tbladdr_remove(&rule->src.addr);
383 			pf_tbladdr_remove(&rule->dst.addr);
384 			if (rule->overload_tbl)
385 				pfr_detach_table(rule->overload_tbl);
386 		}
387 		TAILQ_REMOVE(rulequeue, rule, entries);
388 		rule->entries.tqe_prev = NULL;
389 		rule->nr = -1;
390 	}
391 
392 	if (rule->states_cur > 0 || rule->src_nodes > 0 ||
393 	    rule->entries.tqe_prev != NULL)
394 		return;
395 	pf_tag_unref(rule->tag);
396 	pf_tag_unref(rule->match_tag);
397 #ifdef ALTQ
398 	if (rule->pqid != rule->qid)
399 		pf_qid_unref(rule->pqid);
400 	pf_qid_unref(rule->qid);
401 #endif
402 	pf_rtlabel_remove(&rule->src.addr);
403 	pf_rtlabel_remove(&rule->dst.addr);
404 	pfi_dynaddr_remove(&rule->src.addr);
405 	pfi_dynaddr_remove(&rule->dst.addr);
406 	if (rulequeue == NULL) {
407 		pf_tbladdr_remove(&rule->src.addr);
408 		pf_tbladdr_remove(&rule->dst.addr);
409 		if (rule->overload_tbl)
410 			pfr_detach_table(rule->overload_tbl);
411 	}
412 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
413 	pf_anchor_remove(rule);
414 	pf_empty_pool(&rule->rpool.list);
415 	kfree(rule, M_PFRULEPL);
416 }
417 
418 u_int16_t
419 tagname2tag(struct pf_tags *head, char *tagname)
420 {
421 	struct pf_tagname	*tag, *p = NULL;
422 	u_int16_t		 new_tagid = 1;
423 
424 	TAILQ_FOREACH(tag, head, entries)
425 		if (strcmp(tagname, tag->name) == 0) {
426 			tag->ref++;
427 			return (tag->tag);
428 		}
429 
430 	/*
431 	 * to avoid fragmentation, we do a linear search from the beginning
432 	 * and take the first free slot we find. if there is none or the list
433 	 * is empty, append a new entry at the end.
434 	 */
435 
436 	/* new entry */
437 	if (!TAILQ_EMPTY(head))
438 		for (p = TAILQ_FIRST(head); p != NULL &&
439 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
440 			new_tagid = p->tag + 1;
441 
442 	if (new_tagid > TAGID_MAX)
443 		return (0);
444 
445 	/* allocate and fill new struct pf_tagname */
446 	tag = kmalloc(sizeof(*tag), M_TEMP, M_WAITOK);
447 	strlcpy(tag->name, tagname, sizeof(tag->name));
448 	tag->tag = new_tagid;
449 	tag->ref++;
450 
451 	if (p != NULL)	/* insert new entry before p */
452 		TAILQ_INSERT_BEFORE(p, tag, entries);
453 	else	/* either list empty or no free slot in between */
454 		TAILQ_INSERT_TAIL(head, tag, entries);
455 
456 	return (tag->tag);
457 }
458 
459 void
460 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
461 {
462 	struct pf_tagname	*tag;
463 
464 	TAILQ_FOREACH(tag, head, entries)
465 		if (tag->tag == tagid) {
466 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
467 			return;
468 		}
469 }
470 
471 void
472 tag_unref(struct pf_tags *head, u_int16_t tag)
473 {
474 	struct pf_tagname	*p, *next;
475 
476 	if (tag == 0)
477 		return;
478 
479 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
480 		next = TAILQ_NEXT(p, entries);
481 		if (tag == p->tag) {
482 			if (--p->ref == 0) {
483 				TAILQ_REMOVE(head, p, entries);
484 				kfree(p, M_TEMP);
485 			}
486 			break;
487 		}
488 	}
489 }
490 
491 u_int16_t
492 pf_tagname2tag(char *tagname)
493 {
494 	return (tagname2tag(&pf_tags, tagname));
495 }
496 
497 void
498 pf_tag2tagname(u_int16_t tagid, char *p)
499 {
500 	tag2tagname(&pf_tags, tagid, p);
501 }
502 
503 void
504 pf_tag_ref(u_int16_t tag)
505 {
506 	struct pf_tagname *t;
507 
508 	TAILQ_FOREACH(t, &pf_tags, entries)
509 		if (t->tag == tag)
510 			break;
511 	if (t != NULL)
512 		t->ref++;
513 }
514 
515 void
516 pf_tag_unref(u_int16_t tag)
517 {
518 	tag_unref(&pf_tags, tag);
519 }
520 
521 int
522 pf_rtlabel_add(struct pf_addr_wrap *a)
523 {
524 	return (0);
525 }
526 
527 void
528 pf_rtlabel_remove(struct pf_addr_wrap *a)
529 {
530 }
531 
532 void
533 pf_rtlabel_copyout(struct pf_addr_wrap *a)
534 {
535 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
536 		strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
537 }
538 
539 #ifdef ALTQ
540 u_int32_t
541 pf_qname2qid(char *qname)
542 {
543 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
544 }
545 
546 void
547 pf_qid2qname(u_int32_t qid, char *p)
548 {
549 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
550 }
551 
552 void
553 pf_qid_unref(u_int32_t qid)
554 {
555 	tag_unref(&pf_qids, (u_int16_t)qid);
556 }
557 
558 int
559 pf_begin_altq(u_int32_t *ticket)
560 {
561 	struct pf_altq	*altq;
562 	int		 error = 0;
563 
564 	/* Purge the old altq list */
565 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
566 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
567 		if (altq->qname[0] == 0) {
568 			/* detach and destroy the discipline */
569 			error = altq_remove(altq);
570 		} else
571 			pf_qid_unref(altq->qid);
572 		kfree(altq, M_PFALTQPL);
573 	}
574 	if (error)
575 		return (error);
576 	*ticket = ++ticket_altqs_inactive;
577 	altqs_inactive_open = 1;
578 	return (0);
579 }
580 
581 int
582 pf_rollback_altq(u_int32_t ticket)
583 {
584 	struct pf_altq	*altq;
585 	int		 error = 0;
586 
587 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
588 		return (0);
589 	/* Purge the old altq list */
590 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
591 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
592 		if (altq->qname[0] == 0) {
593 			/* detach and destroy the discipline */
594 			error = altq_remove(altq);
595 		} else
596 			pf_qid_unref(altq->qid);
597 		kfree(altq, M_PFALTQPL);
598 	}
599 	altqs_inactive_open = 0;
600 	return (error);
601 }
602 
603 int
604 pf_commit_altq(u_int32_t ticket)
605 {
606 	struct pf_altqqueue	*old_altqs;
607 	struct pf_altq		*altq;
608 	int			 err, error = 0;
609 
610 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
611 		return (EBUSY);
612 
613 	/* swap altqs, keep the old. */
614 	crit_enter();
615 	old_altqs = pf_altqs_active;
616 	pf_altqs_active = pf_altqs_inactive;
617 	pf_altqs_inactive = old_altqs;
618 	ticket_altqs_active = ticket_altqs_inactive;
619 
620 	/* Attach new disciplines */
621 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
622 		if (altq->qname[0] == 0) {
623 			/* attach the discipline */
624 			error = altq_pfattach(altq);
625 			if (error) {
626 				crit_exit();
627 				return (error);
628 			}
629 		}
630 	}
631 
632 	/* Purge the old altq list */
633 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
634 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
635 		if (altq->qname[0] == 0) {
636 			/* detach and destroy the discipline */
637 			if (pf_altq_running)
638 				error = pf_disable_altq(altq);
639 			err = altq_pfdetach(altq);
640 			if (err != 0 && error == 0)
641 				error = err;
642 			err = altq_remove(altq);
643 			if (err != 0 && error == 0)
644 				error = err;
645 		} else
646 			pf_qid_unref(altq->qid);
647 		kfree(altq, M_PFALTQPL);
648 	}
649 	crit_exit();
650 
651 	altqs_inactive_open = 0;
652 	return (error);
653 }
654 
655 int
656 pf_enable_altq(struct pf_altq *altq)
657 {
658 	struct ifnet		*ifp;
659 	struct tb_profile	 tb;
660 	int			 error = 0;
661 
662 	if ((ifp = ifunit(altq->ifname)) == NULL)
663 		return (EINVAL);
664 
665 	if (ifp->if_snd.altq_type != ALTQT_NONE)
666 		error = altq_enable(&ifp->if_snd);
667 
668 	/* set tokenbucket regulator */
669 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
670 		tb.rate = altq->ifbandwidth;
671 		tb.depth = altq->tbrsize;
672 		crit_enter();
673 		error = tbr_set(&ifp->if_snd, &tb);
674 		crit_exit();
675 	}
676 
677 	return (error);
678 }
679 
680 int
681 pf_disable_altq(struct pf_altq *altq)
682 {
683 	struct ifnet		*ifp;
684 	struct tb_profile	 tb;
685 	int			 error;
686 
687 	if ((ifp = ifunit(altq->ifname)) == NULL)
688 		return (EINVAL);
689 
690 	/*
691 	 * when the discipline is no longer referenced, it was overridden
692 	 * by a new one.  if so, just return.
693 	 */
694 	if (altq->altq_disc != ifp->if_snd.altq_disc)
695 		return (0);
696 
697 	error = altq_disable(&ifp->if_snd);
698 
699 	if (error == 0) {
700 		/* clear tokenbucket regulator */
701 		tb.rate = 0;
702 		crit_enter();
703 		error = tbr_set(&ifp->if_snd, &tb);
704 		crit_exit();
705 	}
706 
707 	return (error);
708 }
709 #endif /* ALTQ */
710 
711 int
712 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
713 {
714 	struct pf_ruleset	*rs;
715 	struct pf_rule		*rule;
716 
717 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
718 		return (EINVAL);
719 	rs = pf_find_or_create_ruleset(anchor);
720 	if (rs == NULL)
721 		return (EINVAL);
722 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
723 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
724 		rs->rules[rs_num].inactive.rcount--;
725 	}
726 	*ticket = ++rs->rules[rs_num].inactive.ticket;
727 	rs->rules[rs_num].inactive.open = 1;
728 	return (0);
729 }
730 
731 int
732 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
733 {
734 	struct pf_ruleset	*rs;
735 	struct pf_rule		*rule;
736 
737 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
738 		return (EINVAL);
739 	rs = pf_find_ruleset(anchor);
740 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
741 	    rs->rules[rs_num].inactive.ticket != ticket)
742 		return (0);
743 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
744 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
745 		rs->rules[rs_num].inactive.rcount--;
746 	}
747 	rs->rules[rs_num].inactive.open = 0;
748 	return (0);
749 }
750 
751 #define PF_MD5_UPD(st, elm)						\
752 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
753 
754 #define PF_MD5_UPD_STR(st, elm)						\
755 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
756 
757 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
758 		(stor) = htonl((st)->elm);				\
759 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
760 } while (0)
761 
762 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
763 		(stor) = htons((st)->elm);				\
764 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
765 } while (0)
766 
767 void
768 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
769 {
770 	PF_MD5_UPD(pfr, addr.type);
771 	switch (pfr->addr.type) {
772 		case PF_ADDR_DYNIFTL:
773 			PF_MD5_UPD(pfr, addr.v.ifname);
774 			PF_MD5_UPD(pfr, addr.iflags);
775 			break;
776 		case PF_ADDR_TABLE:
777 			PF_MD5_UPD(pfr, addr.v.tblname);
778 			break;
779 		case PF_ADDR_ADDRMASK:
780 			/* XXX ignore af? */
781 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
782 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
783 			break;
784 		case PF_ADDR_RTLABEL:
785 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
786 			break;
787 	}
788 
789 	PF_MD5_UPD(pfr, port[0]);
790 	PF_MD5_UPD(pfr, port[1]);
791 	PF_MD5_UPD(pfr, neg);
792 	PF_MD5_UPD(pfr, port_op);
793 }
794 
795 void
796 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
797 {
798 	u_int16_t x;
799 	u_int32_t y;
800 
801 	pf_hash_rule_addr(ctx, &rule->src);
802 	pf_hash_rule_addr(ctx, &rule->dst);
803 	PF_MD5_UPD_STR(rule, label);
804 	PF_MD5_UPD_STR(rule, ifname);
805 	PF_MD5_UPD_STR(rule, match_tagname);
806 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
807 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
808 	PF_MD5_UPD_HTONL(rule, prob, y);
809 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
810 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
811 	PF_MD5_UPD(rule, uid.op);
812 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
813 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
814 	PF_MD5_UPD(rule, gid.op);
815 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
816 	PF_MD5_UPD(rule, action);
817 	PF_MD5_UPD(rule, direction);
818 	PF_MD5_UPD(rule, af);
819 	PF_MD5_UPD(rule, quick);
820 	PF_MD5_UPD(rule, ifnot);
821 	PF_MD5_UPD(rule, match_tag_not);
822 	PF_MD5_UPD(rule, natpass);
823 	PF_MD5_UPD(rule, keep_state);
824 	PF_MD5_UPD(rule, proto);
825 	PF_MD5_UPD(rule, type);
826 	PF_MD5_UPD(rule, code);
827 	PF_MD5_UPD(rule, flags);
828 	PF_MD5_UPD(rule, flagset);
829 	PF_MD5_UPD(rule, allow_opts);
830 	PF_MD5_UPD(rule, rt);
831 	PF_MD5_UPD(rule, tos);
832 }
833 
834 int
835 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
836 {
837 	struct pf_ruleset	*rs;
838 	struct pf_rule		*rule, **old_array;
839 	struct pf_rulequeue	*old_rules;
840 	int			 error;
841 	u_int32_t		 old_rcount;
842 
843 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
844 		return (EINVAL);
845 	rs = pf_find_ruleset(anchor);
846 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
847 	    ticket != rs->rules[rs_num].inactive.ticket)
848 		return (EBUSY);
849 
850 	/* Calculate checksum for the main ruleset */
851 	if (rs == &pf_main_ruleset) {
852 		error = pf_setup_pfsync_matching(rs);
853 		if (error != 0)
854 			return (error);
855 	}
856 
857 	/* Swap rules, keep the old. */
858 	crit_enter();
859 	old_rules = rs->rules[rs_num].active.ptr;
860 	old_rcount = rs->rules[rs_num].active.rcount;
861 	old_array = rs->rules[rs_num].active.ptr_array;
862 
863 	rs->rules[rs_num].active.ptr =
864 	    rs->rules[rs_num].inactive.ptr;
865 	rs->rules[rs_num].active.ptr_array =
866 	    rs->rules[rs_num].inactive.ptr_array;
867 	rs->rules[rs_num].active.rcount =
868 	    rs->rules[rs_num].inactive.rcount;
869 	rs->rules[rs_num].inactive.ptr = old_rules;
870 	rs->rules[rs_num].inactive.ptr_array = old_array;
871 	rs->rules[rs_num].inactive.rcount = old_rcount;
872 
873 	rs->rules[rs_num].active.ticket =
874 	    rs->rules[rs_num].inactive.ticket;
875 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
876 
877 
878 	/* Purge the old rule list. */
879 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
880 		pf_rm_rule(old_rules, rule);
881 	if (rs->rules[rs_num].inactive.ptr_array)
882 		kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
883 	rs->rules[rs_num].inactive.ptr_array = NULL;
884 	rs->rules[rs_num].inactive.rcount = 0;
885 	rs->rules[rs_num].inactive.open = 0;
886 	pf_remove_if_empty_ruleset(rs);
887 	crit_exit();
888 	return (0);
889 }
890 
891 int
892 pf_setup_pfsync_matching(struct pf_ruleset *rs)
893 {
894 	MD5_CTX			 ctx;
895 	struct pf_rule		*rule;
896 	int			 rs_cnt;
897 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
898 
899 	MD5Init(&ctx);
900 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
901 		/* XXX PF_RULESET_SCRUB as well? */
902 		if (rs_cnt == PF_RULESET_SCRUB)
903 			continue;
904 
905 		if (rs->rules[rs_cnt].inactive.ptr_array)
906 			kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
907 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
908 
909 		if (rs->rules[rs_cnt].inactive.rcount) {
910 			rs->rules[rs_cnt].inactive.ptr_array =
911 			    kmalloc(sizeof(caddr_t) *
912 				    rs->rules[rs_cnt].inactive.rcount,
913 				    M_TEMP, M_WAITOK);
914 		}
915 
916 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
917 		    entries) {
918 			pf_hash_rule(&ctx, rule);
919 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
920 		}
921 	}
922 
923 	MD5Final(digest, &ctx);
924 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
925 	return (0);
926 }
927 
928 int
929 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
930     sa_family_t af)
931 {
932 	if (pfi_dynaddr_setup(addr, af) ||
933 	    pf_tbladdr_setup(ruleset, addr))
934 		return (EINVAL);
935 
936 	return (0);
937 }
938 
939 void
940 pf_addr_copyout(struct pf_addr_wrap *addr)
941 {
942 	pfi_dynaddr_copyout(addr);
943 	pf_tbladdr_copyout(addr);
944 	pf_rtlabel_copyout(addr);
945 }
946 
947 int
948 pfioctl(struct dev_ioctl_args *ap)
949 {
950 	u_long cmd = ap->a_cmd;
951 	caddr_t addr = ap->a_data;
952 	struct pf_pooladdr	*pa = NULL;
953 	struct pf_pool		*pool = NULL;
954 	int			 error = 0;
955 
956 	lwkt_gettoken(&pf_token);
957 
958 	/* XXX keep in sync with switch() below */
959 	if (securelevel > 1) {
960 		switch (cmd) {
961 		case DIOCGETRULES:
962 		case DIOCGETRULE:
963 		case DIOCGETADDRS:
964 		case DIOCGETADDR:
965 		case DIOCGETSTATE:
966 		case DIOCSETSTATUSIF:
967 		case DIOCGETSTATUS:
968 		case DIOCCLRSTATUS:
969 		case DIOCNATLOOK:
970 		case DIOCSETDEBUG:
971 		case DIOCGETSTATES:
972 		case DIOCGETTIMEOUT:
973 		case DIOCCLRRULECTRS:
974 		case DIOCGETLIMIT:
975 		case DIOCGETALTQS:
976 		case DIOCGETALTQ:
977 		case DIOCGETQSTATS:
978 		case DIOCGETRULESETS:
979 		case DIOCGETRULESET:
980 		case DIOCRGETTABLES:
981 		case DIOCRGETTSTATS:
982 		case DIOCRCLRTSTATS:
983 		case DIOCRCLRADDRS:
984 		case DIOCRADDADDRS:
985 		case DIOCRDELADDRS:
986 		case DIOCRSETADDRS:
987 		case DIOCRGETADDRS:
988 		case DIOCRGETASTATS:
989 		case DIOCRCLRASTATS:
990 		case DIOCRTSTADDRS:
991 		case DIOCOSFPGET:
992 		case DIOCGETSRCNODES:
993 		case DIOCCLRSRCNODES:
994 		case DIOCIGETIFACES:
995 		case DIOCSETIFFLAG:
996 		case DIOCCLRIFFLAG:
997 		case DIOCGIFSPEED:
998 			break;
999 		case DIOCRCLRTABLES:
1000 		case DIOCRADDTABLES:
1001 		case DIOCRDELTABLES:
1002 		case DIOCRSETTFLAGS:
1003 			if (((struct pfioc_table *)addr)->pfrio_flags &
1004 			    PFR_FLAG_DUMMY)
1005 				break; /* dummy operation ok */
1006 			lwkt_reltoken(&pf_token);
1007 			return (EPERM);
1008 		default:
1009 			lwkt_reltoken(&pf_token);
1010 			return (EPERM);
1011 		}
1012 	}
1013 
1014 	if (!(ap->a_fflag & FWRITE)) {
1015 		switch (cmd) {
1016 		case DIOCGETRULES:
1017 		case DIOCGETADDRS:
1018 		case DIOCGETADDR:
1019 		case DIOCGETSTATE:
1020 		case DIOCGETSTATUS:
1021 		case DIOCGETSTATES:
1022 		case DIOCGETTIMEOUT:
1023 		case DIOCGETLIMIT:
1024 		case DIOCGETALTQS:
1025 		case DIOCGETALTQ:
1026 		case DIOCGETQSTATS:
1027 		case DIOCGETRULESETS:
1028 		case DIOCGETRULESET:
1029 		case DIOCNATLOOK:
1030 		case DIOCRGETTABLES:
1031 		case DIOCRGETTSTATS:
1032 		case DIOCRGETADDRS:
1033 		case DIOCRGETASTATS:
1034 		case DIOCRTSTADDRS:
1035 		case DIOCOSFPGET:
1036 		case DIOCGETSRCNODES:
1037 		case DIOCIGETIFACES:
1038 		case DIOCGIFSPEED:
1039 			break;
1040 		case DIOCRCLRTABLES:
1041 		case DIOCRADDTABLES:
1042 		case DIOCRDELTABLES:
1043 		case DIOCRCLRTSTATS:
1044 		case DIOCRCLRADDRS:
1045 		case DIOCRADDADDRS:
1046 		case DIOCRDELADDRS:
1047 		case DIOCRSETADDRS:
1048 		case DIOCRSETTFLAGS:
1049 			if (((struct pfioc_table *)addr)->pfrio_flags &
1050 			    PFR_FLAG_DUMMY)
1051 				break; /* dummy operation ok */
1052 			lwkt_reltoken(&pf_token);
1053 			return (EACCES);
1054 		case DIOCGETRULE:
1055 			if (((struct pfioc_rule *)addr)->action ==
1056 			    PF_GET_CLR_CNTR) {
1057 				lwkt_reltoken(&pf_token);
1058 				return (EACCES);
1059 			}
1060 			break;
1061 		default:
1062 			lwkt_reltoken(&pf_token);
1063 			return (EACCES);
1064 		}
1065 	}
1066 
1067 	switch (cmd) {
1068 	case DIOCSTART:
1069 		if (pf_status.running)
1070 			error = EEXIST;
1071 		else {
1072 			error = hook_pf();
1073 			if (error) {
1074 				DPFPRINTF(PF_DEBUG_MISC,
1075 				    ("pf: pfil registration fail\n"));
1076 				break;
1077 			}
1078 			pf_status.running = 1;
1079 			pf_status.since = time_second;
1080 			if (pf_status.stateid == 0) {
1081 				pf_status.stateid = time_second;
1082 				pf_status.stateid = pf_status.stateid << 32;
1083 			}
1084 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1085 		}
1086 		break;
1087 
1088 	case DIOCSTOP:
1089 		if (!pf_status.running)
1090 			error = ENOENT;
1091 		else {
1092 			pf_status.running = 0;
1093 			error = dehook_pf();
1094 			if (error) {
1095 				pf_status.running = 1;
1096 				DPFPRINTF(PF_DEBUG_MISC,
1097 					("pf: pfil unregistration failed\n"));
1098 			}
1099 			pf_status.since = time_second;
1100 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1101 		}
1102 		break;
1103 
1104 	case DIOCADDRULE: {
1105 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1106 		struct pf_ruleset	*ruleset;
1107 		struct pf_rule		*rule, *tail;
1108 		struct pf_pooladdr	*pa;
1109 		int			 rs_num;
1110 
1111 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1112 		ruleset = pf_find_ruleset(pr->anchor);
1113 		if (ruleset == NULL) {
1114 			error = EINVAL;
1115 			break;
1116 		}
1117 		rs_num = pf_get_ruleset_number(pr->rule.action);
1118 		if (rs_num >= PF_RULESET_MAX) {
1119 			error = EINVAL;
1120 			break;
1121 		}
1122 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1123 			error = EINVAL;
1124 			break;
1125 		}
1126 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1127 			error = EBUSY;
1128 			break;
1129 		}
1130 		if (pr->pool_ticket != ticket_pabuf) {
1131 			error = EBUSY;
1132 			break;
1133 		}
1134 		rule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK);
1135 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1136 		rule->cuid = ap->a_cred->cr_ruid;
1137 		rule->cpid = 0;
1138 		rule->anchor = NULL;
1139 		rule->kif = NULL;
1140 		TAILQ_INIT(&rule->rpool.list);
1141 		/* initialize refcounting */
1142 		rule->states_cur = 0;
1143 		rule->src_nodes = 0;
1144 		rule->entries.tqe_prev = NULL;
1145 #ifndef INET
1146 		if (rule->af == AF_INET) {
1147 			kfree(rule, M_PFRULEPL);
1148 			error = EAFNOSUPPORT;
1149 			break;
1150 		}
1151 #endif /* INET */
1152 #ifndef INET6
1153 		if (rule->af == AF_INET6) {
1154 			kfree(rule, M_PFRULEPL);
1155 			error = EAFNOSUPPORT;
1156 			break;
1157 		}
1158 #endif /* INET6 */
1159 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1160 		    pf_rulequeue);
1161 		if (tail)
1162 			rule->nr = tail->nr + 1;
1163 		else
1164 			rule->nr = 0;
1165 		if (rule->ifname[0]) {
1166 			rule->kif = pfi_kif_get(rule->ifname);
1167 			if (rule->kif == NULL) {
1168 				kfree(rule, M_PFRULEPL);
1169 				error = EINVAL;
1170 				break;
1171 			}
1172 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1173 		}
1174 
1175 		if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1176 			error = EBUSY;
1177 
1178 #ifdef ALTQ
1179 		/* set queue IDs */
1180 		if (rule->qname[0] != 0) {
1181 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1182 				error = EBUSY;
1183 			else if (rule->pqname[0] != 0) {
1184 				if ((rule->pqid =
1185 				    pf_qname2qid(rule->pqname)) == 0)
1186 					error = EBUSY;
1187 			} else
1188 				rule->pqid = rule->qid;
1189 		}
1190 #endif
1191 		if (rule->tagname[0])
1192 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1193 				error = EBUSY;
1194 		if (rule->match_tagname[0])
1195 			if ((rule->match_tag =
1196 			    pf_tagname2tag(rule->match_tagname)) == 0)
1197 				error = EBUSY;
1198 		if (rule->rt && !rule->direction)
1199 			error = EINVAL;
1200 #if NPFLOG > 0
1201 		if (!rule->log)
1202 			rule->logif = 0;
1203 		if (rule->logif >= PFLOGIFS_MAX)
1204 			error = EINVAL;
1205 #endif
1206 		if (pf_rtlabel_add(&rule->src.addr) ||
1207 		    pf_rtlabel_add(&rule->dst.addr))
1208 			error = EBUSY;
1209 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1210 			error = EINVAL;
1211 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1212 			error = EINVAL;
1213 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1214 			error = EINVAL;
1215 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1216 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1217 				error = EINVAL;
1218 
1219 		if (rule->overload_tblname[0]) {
1220 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1221 			    rule->overload_tblname)) == NULL)
1222 				error = EINVAL;
1223 			else
1224 				rule->overload_tbl->pfrkt_flags |=
1225 				    PFR_TFLAG_ACTIVE;
1226 		}
1227 
1228 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1229 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1230 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1231 		    (rule->rt > PF_FASTROUTE)) &&
1232 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1233 			error = EINVAL;
1234 
1235 		if (error) {
1236 			pf_rm_rule(NULL, rule);
1237 			break;
1238 		}
1239 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1240 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1241 		    rule->bytes[0] = rule->bytes[1] = 0;
1242 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1243 		    rule, entries);
1244 		ruleset->rules[rs_num].inactive.rcount++;
1245 		break;
1246 	}
1247 
1248 	case DIOCGETRULES: {
1249 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1250 		struct pf_ruleset	*ruleset;
1251 		struct pf_rule		*tail;
1252 		int			 rs_num;
1253 
1254 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1255 		ruleset = pf_find_ruleset(pr->anchor);
1256 		if (ruleset == NULL) {
1257 			error = EINVAL;
1258 			break;
1259 		}
1260 		rs_num = pf_get_ruleset_number(pr->rule.action);
1261 		if (rs_num >= PF_RULESET_MAX) {
1262 			error = EINVAL;
1263 			break;
1264 		}
1265 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1266 		    pf_rulequeue);
1267 		if (tail)
1268 			pr->nr = tail->nr + 1;
1269 		else
1270 			pr->nr = 0;
1271 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1272 		break;
1273 	}
1274 
1275 	case DIOCGETRULE: {
1276 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1277 		struct pf_ruleset	*ruleset;
1278 		struct pf_rule		*rule;
1279 		int			 rs_num, i;
1280 
1281 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1282 		ruleset = pf_find_ruleset(pr->anchor);
1283 		if (ruleset == NULL) {
1284 			error = EINVAL;
1285 			break;
1286 		}
1287 		rs_num = pf_get_ruleset_number(pr->rule.action);
1288 		if (rs_num >= PF_RULESET_MAX) {
1289 			error = EINVAL;
1290 			break;
1291 		}
1292 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1293 			error = EBUSY;
1294 			break;
1295 		}
1296 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1297 		while ((rule != NULL) && (rule->nr != pr->nr))
1298 			rule = TAILQ_NEXT(rule, entries);
1299 		if (rule == NULL) {
1300 			error = EBUSY;
1301 			break;
1302 		}
1303 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1304 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1305 			error = EBUSY;
1306 			break;
1307 		}
1308 		pf_addr_copyout(&pr->rule.src.addr);
1309 		pf_addr_copyout(&pr->rule.dst.addr);
1310 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1311 			if (rule->skip[i].ptr == NULL)
1312 				pr->rule.skip[i].nr = (uint32_t)(-1);
1313 			else
1314 				pr->rule.skip[i].nr =
1315 				    rule->skip[i].ptr->nr;
1316 
1317 		if (pr->action == PF_GET_CLR_CNTR) {
1318 			rule->evaluations = 0;
1319 			rule->packets[0] = rule->packets[1] = 0;
1320 			rule->bytes[0] = rule->bytes[1] = 0;
1321 			rule->states_tot = 0;
1322 		}
1323 		break;
1324 	}
1325 
1326 	case DIOCCHANGERULE: {
1327 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1328 		struct pf_ruleset	*ruleset;
1329 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1330 		u_int32_t		 nr = 0;
1331 		int			 rs_num;
1332 
1333 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1334 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1335 		    pcr->pool_ticket != ticket_pabuf) {
1336 			error = EBUSY;
1337 			break;
1338 		}
1339 
1340 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1341 		    pcr->action > PF_CHANGE_GET_TICKET) {
1342 			error = EINVAL;
1343 			break;
1344 		}
1345 		ruleset = pf_find_ruleset(pcr->anchor);
1346 		if (ruleset == NULL) {
1347 			error = EINVAL;
1348 			break;
1349 		}
1350 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1351 		if (rs_num >= PF_RULESET_MAX) {
1352 			error = EINVAL;
1353 			break;
1354 		}
1355 
1356 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1357 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1358 			break;
1359 		} else {
1360 			if (pcr->ticket !=
1361 			    ruleset->rules[rs_num].active.ticket) {
1362 				error = EINVAL;
1363 				break;
1364 			}
1365 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1366 				error = EINVAL;
1367 				break;
1368 			}
1369 		}
1370 
1371 		if (pcr->action != PF_CHANGE_REMOVE) {
1372 			newrule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK|M_NULLOK);
1373 			if (newrule == NULL) {
1374 				error = ENOMEM;
1375 				break;
1376 			}
1377 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1378 			newrule->cuid = ap->a_cred->cr_ruid;
1379 			newrule->cpid = 0;
1380 			TAILQ_INIT(&newrule->rpool.list);
1381 			/* initialize refcounting */
1382 			newrule->states_cur = 0;
1383 			newrule->entries.tqe_prev = NULL;
1384 #ifndef INET
1385 			if (newrule->af == AF_INET) {
1386 				kfree(newrule, M_PFRULEPL);
1387 				error = EAFNOSUPPORT;
1388 				break;
1389 			}
1390 #endif /* INET */
1391 #ifndef INET6
1392 			if (newrule->af == AF_INET6) {
1393 				kfree(newrule, M_PFRULEPL);
1394 				error = EAFNOSUPPORT;
1395 				break;
1396 			}
1397 #endif /* INET6 */
1398 			if (newrule->ifname[0]) {
1399 				newrule->kif = pfi_kif_get(newrule->ifname);
1400 				if (newrule->kif == NULL) {
1401 					kfree(newrule, M_PFRULEPL);
1402 					error = EINVAL;
1403 					break;
1404 				}
1405 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1406 			} else
1407 				newrule->kif = NULL;
1408 
1409 			if (newrule->rtableid > 0 &&
1410 			    newrule->rtableid > rt_numfibs)
1411 				error = EBUSY;
1412 
1413 #ifdef ALTQ
1414 			/* set queue IDs */
1415 			if (newrule->qname[0] != 0) {
1416 				if ((newrule->qid =
1417 				    pf_qname2qid(newrule->qname)) == 0)
1418 					error = EBUSY;
1419 				else if (newrule->pqname[0] != 0) {
1420 					if ((newrule->pqid =
1421 					    pf_qname2qid(newrule->pqname)) == 0)
1422 						error = EBUSY;
1423 				} else
1424 					newrule->pqid = newrule->qid;
1425 			}
1426 #endif /* ALTQ */
1427 			if (newrule->tagname[0])
1428 				if ((newrule->tag =
1429 				    pf_tagname2tag(newrule->tagname)) == 0)
1430 					error = EBUSY;
1431 			if (newrule->match_tagname[0])
1432 				if ((newrule->match_tag = pf_tagname2tag(
1433 				    newrule->match_tagname)) == 0)
1434 					error = EBUSY;
1435 			if (newrule->rt && !newrule->direction)
1436 				error = EINVAL;
1437 #if NPFLOG > 0
1438 			if (!newrule->log)
1439 				newrule->logif = 0;
1440 			if (newrule->logif >= PFLOGIFS_MAX)
1441 				error = EINVAL;
1442 #endif
1443 			if (pf_rtlabel_add(&newrule->src.addr) ||
1444 			    pf_rtlabel_add(&newrule->dst.addr))
1445 				error = EBUSY;
1446 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1447 				error = EINVAL;
1448 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1449 				error = EINVAL;
1450 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1451 				error = EINVAL;
1452 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1453 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1454 					error = EINVAL;
1455 
1456 			if (newrule->overload_tblname[0]) {
1457 				if ((newrule->overload_tbl = pfr_attach_table(
1458 				    ruleset, newrule->overload_tblname)) ==
1459 				    NULL)
1460 					error = EINVAL;
1461 				else
1462 					newrule->overload_tbl->pfrkt_flags |=
1463 					    PFR_TFLAG_ACTIVE;
1464 			}
1465 
1466 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1467 			if (((((newrule->action == PF_NAT) ||
1468 			    (newrule->action == PF_RDR) ||
1469 			    (newrule->action == PF_BINAT) ||
1470 			    (newrule->rt > PF_FASTROUTE)) &&
1471 			    !newrule->anchor)) &&
1472 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1473 				error = EINVAL;
1474 
1475 			if (error) {
1476 				pf_rm_rule(NULL, newrule);
1477 				break;
1478 			}
1479 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1480 			newrule->evaluations = 0;
1481 			newrule->packets[0] = newrule->packets[1] = 0;
1482 			newrule->bytes[0] = newrule->bytes[1] = 0;
1483 		}
1484 		pf_empty_pool(&pf_pabuf);
1485 
1486 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1487 			oldrule = TAILQ_FIRST(
1488 			    ruleset->rules[rs_num].active.ptr);
1489 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1490 			oldrule = TAILQ_LAST(
1491 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1492 		else {
1493 			oldrule = TAILQ_FIRST(
1494 			    ruleset->rules[rs_num].active.ptr);
1495 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1496 				oldrule = TAILQ_NEXT(oldrule, entries);
1497 			if (oldrule == NULL) {
1498 				if (newrule != NULL)
1499 					pf_rm_rule(NULL, newrule);
1500 				error = EINVAL;
1501 				break;
1502 			}
1503 		}
1504 
1505 		if (pcr->action == PF_CHANGE_REMOVE) {
1506 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1507 			ruleset->rules[rs_num].active.rcount--;
1508 		} else {
1509 			if (oldrule == NULL)
1510 				TAILQ_INSERT_TAIL(
1511 				    ruleset->rules[rs_num].active.ptr,
1512 				    newrule, entries);
1513 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1514 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1515 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1516 			else
1517 				TAILQ_INSERT_AFTER(
1518 				    ruleset->rules[rs_num].active.ptr,
1519 				    oldrule, newrule, entries);
1520 			ruleset->rules[rs_num].active.rcount++;
1521 		}
1522 
1523 		nr = 0;
1524 		TAILQ_FOREACH(oldrule,
1525 		    ruleset->rules[rs_num].active.ptr, entries)
1526 			oldrule->nr = nr++;
1527 
1528 		ruleset->rules[rs_num].active.ticket++;
1529 
1530 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1531 		pf_remove_if_empty_ruleset(ruleset);
1532 
1533 		break;
1534 	}
1535 
1536 	case DIOCCLRSTATES: {
1537 		struct pf_state		*s, *nexts;
1538 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1539 		u_int			 killed = 0;
1540 		globaldata_t save_gd = mycpu;
1541 		int nn;
1542 
1543 		for (nn = 0; nn < ncpus; ++nn) {
1544 			lwkt_setcpu_self(globaldata_find(nn));
1545 			for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]);
1546 			     s; s = nexts) {
1547 				nexts = RB_NEXT(pf_state_tree_id,
1548 						&tree_id[nn], s);
1549 
1550 				if (!psk->psk_ifname[0] ||
1551 				    !strcmp(psk->psk_ifname,
1552 					    s->kif->pfik_name)) {
1553 					/*
1554 					 * don't send out individual
1555 					 * delete messages
1556 					 */
1557 					s->sync_flags = PFSTATE_NOSYNC;
1558 					pf_unlink_state(s);
1559 					killed++;
1560 				}
1561 			}
1562 		}
1563 		lwkt_setcpu_self(save_gd);
1564 		psk->psk_killed = killed;
1565 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1566 		break;
1567 	}
1568 
1569 	case DIOCKILLSTATES: {
1570 		struct pf_state		*s, *nexts;
1571 		struct pf_state_key	*sk;
1572 		struct pf_addr		*srcaddr, *dstaddr;
1573 		u_int16_t		 srcport, dstport;
1574 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1575 		u_int			 killed = 0;
1576 		globaldata_t save_gd = mycpu;
1577 		int nn;
1578 
1579 		if (psk->psk_pfcmp.id) {
1580 			if (psk->psk_pfcmp.creatorid == 0)
1581 				psk->psk_pfcmp.creatorid = pf_status.hostid;
1582 			for (nn = 0; nn < ncpus; ++nn) {
1583 				lwkt_setcpu_self(globaldata_find(nn));
1584 				if ((s = pf_find_state_byid(&psk->psk_pfcmp))) {
1585 					/* send immediate delete of state */
1586 					pfsync_delete_state(s);
1587 					s->sync_flags |= PFSTATE_NOSYNC;
1588 					pf_unlink_state(s);
1589 					++psk->psk_killed;
1590 				}
1591 			}
1592 			lwkt_setcpu_self(save_gd);
1593 			break;
1594 		}
1595 
1596 		for (nn = 0; nn < ncpus; ++nn) {
1597 		    lwkt_setcpu_self(globaldata_find(nn));
1598 		    for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]);
1599 			 s; s = nexts) {
1600 			    nexts = RB_NEXT(pf_state_tree_id, &tree_id[nn], s);
1601 			    sk = s->key[PF_SK_WIRE];
1602 
1603 			    if (s->direction == PF_OUT) {
1604 				    srcaddr = &sk->addr[1];
1605 				    dstaddr = &sk->addr[0];
1606 				    srcport = sk->port[0];
1607 				    dstport = sk->port[0];
1608 			    } else {
1609 				    srcaddr = &sk->addr[0];
1610 				    dstaddr = &sk->addr[1];
1611 				    srcport = sk->port[0];
1612 				    dstport = sk->port[0];
1613 			    }
1614 			    if ((!psk->psk_af || sk->af == psk->psk_af)
1615 				&& (!psk->psk_proto || psk->psk_proto ==
1616 						       sk->proto) &&
1617 				PF_MATCHA(psk->psk_src.neg,
1618 					  &psk->psk_src.addr.v.a.addr,
1619 					  &psk->psk_src.addr.v.a.mask,
1620 					  srcaddr, sk->af) &&
1621 				PF_MATCHA(psk->psk_dst.neg,
1622 					  &psk->psk_dst.addr.v.a.addr,
1623 					  &psk->psk_dst.addr.v.a.mask,
1624 					  dstaddr, sk->af) &&
1625 				(psk->psk_src.port_op == 0 ||
1626 				 pf_match_port(psk->psk_src.port_op,
1627 					       psk->psk_src.port[0],
1628 					       psk->psk_src.port[1],
1629 					       srcport)) &&
1630 				(psk->psk_dst.port_op == 0 ||
1631 				 pf_match_port(psk->psk_dst.port_op,
1632 					       psk->psk_dst.port[0],
1633 					       psk->psk_dst.port[1],
1634 					       dstport)) &&
1635 				(!psk->psk_label[0] ||
1636 				 (s->rule.ptr->label[0] &&
1637 				  !strcmp(psk->psk_label, s->rule.ptr->label))) &&
1638 				(!psk->psk_ifname[0] ||
1639 				 !strcmp(psk->psk_ifname, s->kif->pfik_name))) {
1640 				    /* send immediate delete of state */
1641 				    pfsync_delete_state(s);
1642 				    s->sync_flags |= PFSTATE_NOSYNC;
1643 				    pf_unlink_state(s);
1644 				    killed++;
1645 			    }
1646 		    }
1647 		}
1648 		lwkt_setcpu_self(save_gd);
1649 		psk->psk_killed = killed;
1650 		break;
1651 	}
1652 
1653 	case DIOCADDSTATE: {
1654 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1655 		struct pfsync_state	*sp = &ps->state;
1656 
1657 		if (sp->timeout >= PFTM_MAX &&
1658 		    sp->timeout != PFTM_UNTIL_PACKET) {
1659 			error = EINVAL;
1660 			break;
1661 		}
1662 		error = pfsync_state_import(sp, PFSYNC_SI_IOCTL);
1663 		break;
1664 	}
1665 
1666 	case DIOCGETSTATE: {
1667 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1668 		struct pf_state		*s;
1669 		struct pf_state_cmp	 id_key;
1670 		globaldata_t save_gd = mycpu;
1671 		int nn;
1672 
1673 		bcopy(ps->state.id, &id_key.id, sizeof(id_key.id));
1674 		id_key.creatorid = ps->state.creatorid;
1675 		s = NULL;
1676 		for (nn = 0; nn < ncpus; ++nn) {
1677 			lwkt_setcpu_self(globaldata_find(nn));
1678 			s = pf_find_state_byid(&id_key);
1679 			if (s)
1680 				break;
1681 		}
1682 		if (s) {
1683 			pfsync_state_export(&ps->state, s);
1684 		} else {
1685 			error = ENOENT;
1686 		}
1687 		lwkt_setcpu_self(save_gd);
1688 		break;
1689 	}
1690 
1691 	case DIOCGETSTATES: {
1692 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1693 		struct pf_state		*state;
1694 		struct pfsync_state	*p, *pstore;
1695 		u_int32_t		 nr = 0;
1696 		globaldata_t save_gd = mycpu;
1697 		int nn;
1698 
1699 		if (ps->ps_len == 0) {
1700 			nr = pf_status.states;
1701 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1702 			break;
1703 		}
1704 
1705 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1706 
1707 		p = ps->ps_states;
1708 
1709 		for (nn = 0; nn < ncpus; ++nn) {
1710 			lwkt_setcpu_self(globaldata_find(nn));
1711 			state = TAILQ_FIRST(&state_list[nn]);
1712 			while (state) {
1713 				if (state->timeout != PFTM_UNLINKED) {
1714 					if ((nr + 1) * sizeof(*p) >
1715 					    (unsigned)ps->ps_len) {
1716 						break;
1717 					}
1718 					pfsync_state_export(pstore, state);
1719 					error = copyout(pstore, p, sizeof(*p));
1720 					if (error) {
1721 						kfree(pstore, M_TEMP);
1722 						lwkt_setcpu_self(save_gd);
1723 						goto fail;
1724 					}
1725 					p++;
1726 					nr++;
1727 				}
1728 				state = TAILQ_NEXT(state, entry_list);
1729 			}
1730 		}
1731 		lwkt_setcpu_self(save_gd);
1732 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1733 		kfree(pstore, M_TEMP);
1734 		break;
1735 	}
1736 
1737 	case DIOCGETSTATUS: {
1738 		struct pf_status *s = (struct pf_status *)addr;
1739 		bcopy(&pf_status, s, sizeof(struct pf_status));
1740 		pfi_update_status(s->ifname, s);
1741 		break;
1742 	}
1743 
1744 	case DIOCSETSTATUSIF: {
1745 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1746 
1747 		if (pi->ifname[0] == 0) {
1748 			bzero(pf_status.ifname, IFNAMSIZ);
1749 			break;
1750 		}
1751 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1752 		break;
1753 	}
1754 
1755 	case DIOCCLRSTATUS: {
1756 		bzero(pf_status.counters, sizeof(pf_status.counters));
1757 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1758 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
1759 		pf_status.since = time_second;
1760 		if (*pf_status.ifname)
1761 			pfi_update_status(pf_status.ifname, NULL);
1762 		break;
1763 	}
1764 
1765 	case DIOCNATLOOK: {
1766 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1767 		struct pf_state_key	*sk;
1768 		struct pf_state		*state;
1769 		struct pf_state_key_cmp	 key;
1770 		int			 m = 0, direction = pnl->direction;
1771 		int			 sidx, didx;
1772 		globaldata_t save_gd = mycpu;
1773 		int nn;
1774 
1775 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1776 		sidx = (direction == PF_IN) ? 1 : 0;
1777 		didx = (direction == PF_IN) ? 0 : 1;
1778 
1779 		if (!pnl->proto ||
1780 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1781 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1782 		    ((pnl->proto == IPPROTO_TCP ||
1783 		    pnl->proto == IPPROTO_UDP) &&
1784 		    (!pnl->dport || !pnl->sport)))
1785 			error = EINVAL;
1786 		else {
1787 			key.af = pnl->af;
1788 			key.proto = pnl->proto;
1789 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1790 			key.port[sidx] = pnl->sport;
1791 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1792 			key.port[didx] = pnl->dport;
1793 
1794 			state = NULL;
1795 			for (nn = 0; nn < ncpus; ++nn) {
1796 				lwkt_setcpu_self(globaldata_find(nn));
1797 				state = pf_find_state_all(&key, direction, &m);
1798 				if (state || m > 1)
1799 					break;
1800 				m = 0;
1801 			}
1802 
1803 			if (m > 1) {
1804 				error = E2BIG;	/* more than one state */
1805 			} else if (state != NULL) {
1806 				sk = state->key[sidx];
1807 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1808 				pnl->rsport = sk->port[sidx];
1809 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1810 				pnl->rdport = sk->port[didx];
1811 			} else {
1812 				error = ENOENT;
1813 			}
1814 			lwkt_setcpu_self(save_gd);
1815 		}
1816 		break;
1817 	}
1818 
1819 	case DIOCSETTIMEOUT: {
1820 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1821 		int		 old;
1822 
1823 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1824 		    pt->seconds < 0) {
1825 			error = EINVAL;
1826 			goto fail;
1827 		}
1828 		old = pf_default_rule.timeout[pt->timeout];
1829 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1830 			pt->seconds = 1;
1831 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
1832 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1833 			wakeup(pf_purge_thread);
1834 		pt->seconds = old;
1835 		break;
1836 	}
1837 
1838 	case DIOCGETTIMEOUT: {
1839 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1840 
1841 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1842 			error = EINVAL;
1843 			goto fail;
1844 		}
1845 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1846 		break;
1847 	}
1848 
1849 	case DIOCGETLIMIT: {
1850 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1851 
1852 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1853 			error = EINVAL;
1854 			goto fail;
1855 		}
1856 		pl->limit = pf_pool_limits[pl->index].limit;
1857 		break;
1858 	}
1859 
1860 	case DIOCSETLIMIT: {
1861 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1862 		int			 old_limit;
1863 
1864 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1865 		    pf_pool_limits[pl->index].pp == NULL) {
1866 			error = EINVAL;
1867 			goto fail;
1868 		}
1869 
1870 		/* XXX Get an API to set limits on the zone/pool */
1871 		old_limit = pf_pool_limits[pl->index].limit;
1872 		pf_pool_limits[pl->index].limit = pl->limit;
1873 		pl->limit = old_limit;
1874 		break;
1875 	}
1876 
1877 	case DIOCSETDEBUG: {
1878 		u_int32_t	*level = (u_int32_t *)addr;
1879 
1880 		pf_status.debug = *level;
1881 		break;
1882 	}
1883 
1884 	case DIOCCLRRULECTRS: {
1885 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1886 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1887 		struct pf_rule		*rule;
1888 
1889 		TAILQ_FOREACH(rule,
1890 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1891 			rule->evaluations = 0;
1892 			rule->packets[0] = rule->packets[1] = 0;
1893 			rule->bytes[0] = rule->bytes[1] = 0;
1894 		}
1895 		break;
1896 	}
1897 
1898 	case DIOCGIFSPEED: {
1899 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1900 		struct pf_ifspeed	ps;
1901 		struct ifnet		*ifp;
1902 
1903 		if (psp->ifname[0] != 0) {
1904 			/* Can we completely trust user-land? */
1905 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1906 			ifp = ifunit(ps.ifname);
1907 			if (ifp )
1908 				psp->baudrate = ifp->if_baudrate;
1909 			else
1910 				error = EINVAL;
1911 		} else
1912 			error = EINVAL;
1913 		break;
1914 	}
1915 #ifdef ALTQ
1916 	case DIOCSTARTALTQ: {
1917 		struct pf_altq		*altq;
1918 
1919 		/* enable all altq interfaces on active list */
1920 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1921 			if (altq->qname[0] == 0) {
1922 				error = pf_enable_altq(altq);
1923 				if (error != 0)
1924 					break;
1925 			}
1926 		}
1927 		if (error == 0)
1928 			pf_altq_running = 1;
1929 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1930 		break;
1931 	}
1932 
1933 	case DIOCSTOPALTQ: {
1934 		struct pf_altq		*altq;
1935 
1936 		/* disable all altq interfaces on active list */
1937 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1938 			if (altq->qname[0] == 0) {
1939 				error = pf_disable_altq(altq);
1940 				if (error != 0)
1941 					break;
1942 			}
1943 		}
1944 		if (error == 0)
1945 			pf_altq_running = 0;
1946 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1947 		break;
1948 	}
1949 
1950 	case DIOCADDALTQ: {
1951 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1952 		struct pf_altq		*altq, *a;
1953 
1954 		if (pa->ticket != ticket_altqs_inactive) {
1955 			error = EBUSY;
1956 			break;
1957 		}
1958 		altq = kmalloc(sizeof(struct pf_altq), M_PFALTQPL, M_WAITOK|M_NULLOK);
1959 		if (altq == NULL) {
1960 			error = ENOMEM;
1961 			break;
1962 		}
1963 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
1964 
1965 		/*
1966 		 * if this is for a queue, find the discipline and
1967 		 * copy the necessary fields
1968 		 */
1969 		if (altq->qname[0] != 0) {
1970 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
1971 				error = EBUSY;
1972 				kfree(altq, M_PFALTQPL);
1973 				break;
1974 			}
1975 			altq->altq_disc = NULL;
1976 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
1977 				if (strncmp(a->ifname, altq->ifname,
1978 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
1979 					altq->altq_disc = a->altq_disc;
1980 					break;
1981 				}
1982 			}
1983 		}
1984 
1985 		error = altq_add(altq);
1986 		if (error) {
1987 			kfree(altq, M_PFALTQPL);
1988 			break;
1989 		}
1990 
1991 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
1992 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
1993 		break;
1994 	}
1995 
1996 	case DIOCGETALTQS: {
1997 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1998 		struct pf_altq		*altq;
1999 
2000 		pa->nr = 0;
2001 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2002 			pa->nr++;
2003 		pa->ticket = ticket_altqs_active;
2004 		break;
2005 	}
2006 
2007 	case DIOCGETALTQ: {
2008 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2009 		struct pf_altq		*altq;
2010 		u_int32_t		 nr;
2011 
2012 		if (pa->ticket != ticket_altqs_active) {
2013 			error = EBUSY;
2014 			break;
2015 		}
2016 		nr = 0;
2017 		altq = TAILQ_FIRST(pf_altqs_active);
2018 		while ((altq != NULL) && (nr < pa->nr)) {
2019 			altq = TAILQ_NEXT(altq, entries);
2020 			nr++;
2021 		}
2022 		if (altq == NULL) {
2023 			error = EBUSY;
2024 			break;
2025 		}
2026 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2027 		break;
2028 	}
2029 
2030 	case DIOCCHANGEALTQ:
2031 		/* CHANGEALTQ not supported yet! */
2032 		error = ENODEV;
2033 		break;
2034 
2035 	case DIOCGETQSTATS: {
2036 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2037 		struct pf_altq		*altq;
2038 		u_int32_t		 nr;
2039 		int			 nbytes;
2040 
2041 		if (pq->ticket != ticket_altqs_active) {
2042 			error = EBUSY;
2043 			break;
2044 		}
2045 		nbytes = pq->nbytes;
2046 		nr = 0;
2047 		altq = TAILQ_FIRST(pf_altqs_active);
2048 		while ((altq != NULL) && (nr < pq->nr)) {
2049 			altq = TAILQ_NEXT(altq, entries);
2050 			nr++;
2051 		}
2052 		if (altq == NULL) {
2053 			error = EBUSY;
2054 			break;
2055 		}
2056 		error = altq_getqstats(altq, pq->buf, &nbytes);
2057 		if (error == 0) {
2058 			pq->scheduler = altq->scheduler;
2059 			pq->nbytes = nbytes;
2060 		}
2061 		break;
2062 	}
2063 #endif /* ALTQ */
2064 
2065 	case DIOCBEGINADDRS: {
2066 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2067 
2068 		pf_empty_pool(&pf_pabuf);
2069 		pp->ticket = ++ticket_pabuf;
2070 		break;
2071 	}
2072 
2073 	case DIOCADDADDR: {
2074 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2075 
2076 		if (pp->ticket != ticket_pabuf) {
2077 			error = EBUSY;
2078 			break;
2079 		}
2080 #ifndef INET
2081 		if (pp->af == AF_INET) {
2082 			error = EAFNOSUPPORT;
2083 			break;
2084 		}
2085 #endif /* INET */
2086 #ifndef INET6
2087 		if (pp->af == AF_INET6) {
2088 			error = EAFNOSUPPORT;
2089 			break;
2090 		}
2091 #endif /* INET6 */
2092 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2093 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2094 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2095 			error = EINVAL;
2096 			break;
2097 		}
2098 		pa = kmalloc(sizeof(struct pf_altq), M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2099 		if (pa == NULL) {
2100 			error = ENOMEM;
2101 			break;
2102 		}
2103 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2104 		if (pa->ifname[0]) {
2105 			pa->kif = pfi_kif_get(pa->ifname);
2106 			if (pa->kif == NULL) {
2107 				kfree(ap, M_PFPOOLADDRPL);
2108 				error = EINVAL;
2109 				break;
2110 			}
2111 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2112 		}
2113 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2114 			pfi_dynaddr_remove(&pa->addr);
2115 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2116 			kfree(pa, M_PFPOOLADDRPL);
2117 			error = EINVAL;
2118 			break;
2119 		}
2120 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2121 		break;
2122 	}
2123 
2124 	case DIOCGETADDRS: {
2125 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2126 
2127 		pp->nr = 0;
2128 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2129 		    pp->r_num, 0, 1, 0);
2130 		if (pool == NULL) {
2131 			error = EBUSY;
2132 			break;
2133 		}
2134 		TAILQ_FOREACH(pa, &pool->list, entries)
2135 			pp->nr++;
2136 		break;
2137 	}
2138 
2139 	case DIOCGETADDR: {
2140 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2141 		u_int32_t		 nr = 0;
2142 
2143 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2144 		    pp->r_num, 0, 1, 1);
2145 		if (pool == NULL) {
2146 			error = EBUSY;
2147 			break;
2148 		}
2149 		pa = TAILQ_FIRST(&pool->list);
2150 		while ((pa != NULL) && (nr < pp->nr)) {
2151 			pa = TAILQ_NEXT(pa, entries);
2152 			nr++;
2153 		}
2154 		if (pa == NULL) {
2155 			error = EBUSY;
2156 			break;
2157 		}
2158 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2159 		pf_addr_copyout(&pp->addr.addr);
2160 		break;
2161 	}
2162 
2163 	case DIOCCHANGEADDR: {
2164 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2165 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2166 		struct pf_ruleset	*ruleset;
2167 
2168 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2169 		    pca->action > PF_CHANGE_REMOVE) {
2170 			error = EINVAL;
2171 			break;
2172 		}
2173 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2174 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2175 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2176 			error = EINVAL;
2177 			break;
2178 		}
2179 
2180 		ruleset = pf_find_ruleset(pca->anchor);
2181 		if (ruleset == NULL) {
2182 			error = EBUSY;
2183 			break;
2184 		}
2185 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2186 		    pca->r_num, pca->r_last, 1, 1);
2187 		if (pool == NULL) {
2188 			error = EBUSY;
2189 			break;
2190 		}
2191 		if (pca->action != PF_CHANGE_REMOVE) {
2192 			newpa = kmalloc(sizeof(struct pf_pooladdr),
2193 				M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2194 			if (newpa == NULL) {
2195 				error = ENOMEM;
2196 				break;
2197 			}
2198 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2199 #ifndef INET
2200 			if (pca->af == AF_INET) {
2201 				kfree(newpa, M_PFPOOLADDRPL);
2202 				error = EAFNOSUPPORT;
2203 				break;
2204 			}
2205 #endif /* INET */
2206 #ifndef INET6
2207 			if (pca->af == AF_INET6) {
2208 				kfree(newpa, M_PFPOOLADDRPL);
2209 				error = EAFNOSUPPORT;
2210 				break;
2211 			}
2212 #endif /* INET6 */
2213 			if (newpa->ifname[0]) {
2214 				newpa->kif = pfi_kif_get(newpa->ifname);
2215 				if (newpa->kif == NULL) {
2216 					kfree(newpa, M_PFPOOLADDRPL);
2217 					error = EINVAL;
2218 					break;
2219 				}
2220 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2221 			} else
2222 				newpa->kif = NULL;
2223 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2224 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2225 				pfi_dynaddr_remove(&newpa->addr);
2226 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2227 				kfree(newpa, M_PFPOOLADDRPL);
2228 				error = EINVAL;
2229 				break;
2230 			}
2231 		}
2232 
2233 		if (pca->action == PF_CHANGE_ADD_HEAD)
2234 			oldpa = TAILQ_FIRST(&pool->list);
2235 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2236 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2237 		else {
2238 			int	i = 0;
2239 
2240 			oldpa = TAILQ_FIRST(&pool->list);
2241 			while ((oldpa != NULL) && (i < pca->nr)) {
2242 				oldpa = TAILQ_NEXT(oldpa, entries);
2243 				i++;
2244 			}
2245 			if (oldpa == NULL) {
2246 				error = EINVAL;
2247 				break;
2248 			}
2249 		}
2250 
2251 		if (pca->action == PF_CHANGE_REMOVE) {
2252 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2253 			pfi_dynaddr_remove(&oldpa->addr);
2254 			pf_tbladdr_remove(&oldpa->addr);
2255 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2256 			kfree(oldpa, M_PFPOOLADDRPL);
2257 		} else {
2258 			if (oldpa == NULL)
2259 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2260 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2261 			    pca->action == PF_CHANGE_ADD_BEFORE)
2262 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2263 			else
2264 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2265 				    newpa, entries);
2266 		}
2267 
2268 		pool->cur = TAILQ_FIRST(&pool->list);
2269 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2270 		    pca->af);
2271 		break;
2272 	}
2273 
2274 	case DIOCGETRULESETS: {
2275 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2276 		struct pf_ruleset	*ruleset;
2277 		struct pf_anchor	*anchor;
2278 
2279 		pr->path[sizeof(pr->path) - 1] = 0;
2280 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2281 			error = EINVAL;
2282 			break;
2283 		}
2284 		pr->nr = 0;
2285 		if (ruleset->anchor == NULL) {
2286 			/* XXX kludge for pf_main_ruleset */
2287 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2288 				if (anchor->parent == NULL)
2289 					pr->nr++;
2290 		} else {
2291 			RB_FOREACH(anchor, pf_anchor_node,
2292 			    &ruleset->anchor->children)
2293 				pr->nr++;
2294 		}
2295 		break;
2296 	}
2297 
2298 	case DIOCGETRULESET: {
2299 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2300 		struct pf_ruleset	*ruleset;
2301 		struct pf_anchor	*anchor;
2302 		u_int32_t		 nr = 0;
2303 
2304 		pr->path[sizeof(pr->path) - 1] = 0;
2305 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2306 			error = EINVAL;
2307 			break;
2308 		}
2309 		pr->name[0] = 0;
2310 		if (ruleset->anchor == NULL) {
2311 			/* XXX kludge for pf_main_ruleset */
2312 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2313 				if (anchor->parent == NULL && nr++ == pr->nr) {
2314 					strlcpy(pr->name, anchor->name,
2315 					    sizeof(pr->name));
2316 					break;
2317 				}
2318 		} else {
2319 			RB_FOREACH(anchor, pf_anchor_node,
2320 			    &ruleset->anchor->children)
2321 				if (nr++ == pr->nr) {
2322 					strlcpy(pr->name, anchor->name,
2323 					    sizeof(pr->name));
2324 					break;
2325 				}
2326 		}
2327 		if (!pr->name[0])
2328 			error = EBUSY;
2329 		break;
2330 	}
2331 
2332 	case DIOCRCLRTABLES: {
2333 		struct pfioc_table *io = (struct pfioc_table *)addr;
2334 
2335 		if (io->pfrio_esize != 0) {
2336 			error = ENODEV;
2337 			break;
2338 		}
2339 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2340 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2341 		break;
2342 	}
2343 
2344 	case DIOCRADDTABLES: {
2345 		struct pfioc_table *io = (struct pfioc_table *)addr;
2346 
2347 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2348 			error = ENODEV;
2349 			break;
2350 		}
2351 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2352 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2353 		break;
2354 	}
2355 
2356 	case DIOCRDELTABLES: {
2357 		struct pfioc_table *io = (struct pfioc_table *)addr;
2358 
2359 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2360 			error = ENODEV;
2361 			break;
2362 		}
2363 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2364 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2365 		break;
2366 	}
2367 
2368 	case DIOCRGETTABLES: {
2369 		struct pfioc_table *io = (struct pfioc_table *)addr;
2370 
2371 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2372 			error = ENODEV;
2373 			break;
2374 		}
2375 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2376 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2377 		break;
2378 	}
2379 
2380 	case DIOCRGETTSTATS: {
2381 		struct pfioc_table *io = (struct pfioc_table *)addr;
2382 
2383 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2384 			error = ENODEV;
2385 			break;
2386 		}
2387 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2388 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2389 		break;
2390 	}
2391 
2392 	case DIOCRCLRTSTATS: {
2393 		struct pfioc_table *io = (struct pfioc_table *)addr;
2394 
2395 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2396 			error = ENODEV;
2397 			break;
2398 		}
2399 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2400 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2401 		break;
2402 	}
2403 
2404 	case DIOCRSETTFLAGS: {
2405 		struct pfioc_table *io = (struct pfioc_table *)addr;
2406 
2407 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2408 			error = ENODEV;
2409 			break;
2410 		}
2411 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2412 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2413 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2414 		break;
2415 	}
2416 
2417 	case DIOCRCLRADDRS: {
2418 		struct pfioc_table *io = (struct pfioc_table *)addr;
2419 
2420 		if (io->pfrio_esize != 0) {
2421 			error = ENODEV;
2422 			break;
2423 		}
2424 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2425 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2426 		break;
2427 	}
2428 
2429 	case DIOCRADDADDRS: {
2430 		struct pfioc_table *io = (struct pfioc_table *)addr;
2431 
2432 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2433 			error = ENODEV;
2434 			break;
2435 		}
2436 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2437 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2438 		    PFR_FLAG_USERIOCTL);
2439 		break;
2440 	}
2441 
2442 	case DIOCRDELADDRS: {
2443 		struct pfioc_table *io = (struct pfioc_table *)addr;
2444 
2445 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2446 			error = ENODEV;
2447 			break;
2448 		}
2449 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2450 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2451 		    PFR_FLAG_USERIOCTL);
2452 		break;
2453 	}
2454 
2455 	case DIOCRSETADDRS: {
2456 		struct pfioc_table *io = (struct pfioc_table *)addr;
2457 
2458 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2459 			error = ENODEV;
2460 			break;
2461 		}
2462 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2463 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2464 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2465 		    PFR_FLAG_USERIOCTL, 0);
2466 		break;
2467 	}
2468 
2469 	case DIOCRGETADDRS: {
2470 		struct pfioc_table *io = (struct pfioc_table *)addr;
2471 
2472 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2473 			error = ENODEV;
2474 			break;
2475 		}
2476 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2477 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2478 		break;
2479 	}
2480 
2481 	case DIOCRGETASTATS: {
2482 		struct pfioc_table *io = (struct pfioc_table *)addr;
2483 
2484 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2485 			error = ENODEV;
2486 			break;
2487 		}
2488 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2489 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2490 		break;
2491 	}
2492 
2493 	case DIOCRCLRASTATS: {
2494 		struct pfioc_table *io = (struct pfioc_table *)addr;
2495 
2496 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2497 			error = ENODEV;
2498 			break;
2499 		}
2500 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2501 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2502 		    PFR_FLAG_USERIOCTL);
2503 		break;
2504 	}
2505 
2506 	case DIOCRTSTADDRS: {
2507 		struct pfioc_table *io = (struct pfioc_table *)addr;
2508 
2509 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2510 			error = ENODEV;
2511 			break;
2512 		}
2513 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2514 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2515 		    PFR_FLAG_USERIOCTL);
2516 		break;
2517 	}
2518 
2519 	case DIOCRINADEFINE: {
2520 		struct pfioc_table *io = (struct pfioc_table *)addr;
2521 
2522 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2523 			error = ENODEV;
2524 			break;
2525 		}
2526 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2527 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2528 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2529 		break;
2530 	}
2531 
2532 	case DIOCOSFPADD: {
2533 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2534 		error = pf_osfp_add(io);
2535 		break;
2536 	}
2537 
2538 	case DIOCOSFPGET: {
2539 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2540 		error = pf_osfp_get(io);
2541 		break;
2542 	}
2543 
2544 	case DIOCXBEGIN: {
2545 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2546 		struct pfioc_trans_e	*ioe;
2547 		struct pfr_table	*table;
2548 		int			 i;
2549 
2550 		if (io->esize != sizeof(*ioe)) {
2551 			error = ENODEV;
2552 			goto fail;
2553 		}
2554 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2555 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2556 		for (i = 0; i < io->size; i++) {
2557 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2558 				kfree(table, M_TEMP);
2559 				kfree(ioe, M_TEMP);
2560 				error = EFAULT;
2561 				goto fail;
2562 			}
2563 			switch (ioe->rs_num) {
2564 #ifdef ALTQ
2565 			case PF_RULESET_ALTQ:
2566 				if (ioe->anchor[0]) {
2567 					kfree(table, M_TEMP);
2568 					kfree(ioe, M_TEMP);
2569 					error = EINVAL;
2570 					goto fail;
2571 				}
2572 				if ((error = pf_begin_altq(&ioe->ticket))) {
2573 					kfree(table, M_TEMP);
2574 					kfree(ioe, M_TEMP);
2575 					goto fail;
2576 				}
2577 				break;
2578 #endif /* ALTQ */
2579 			case PF_RULESET_TABLE:
2580 				bzero(table, sizeof(*table));
2581 				strlcpy(table->pfrt_anchor, ioe->anchor,
2582 				    sizeof(table->pfrt_anchor));
2583 				if ((error = pfr_ina_begin(table,
2584 				    &ioe->ticket, NULL, 0))) {
2585 					kfree(table, M_TEMP);
2586 					kfree(ioe, M_TEMP);
2587 					goto fail;
2588 				}
2589 				break;
2590 			default:
2591 				if ((error = pf_begin_rules(&ioe->ticket,
2592 				    ioe->rs_num, ioe->anchor))) {
2593 					kfree(table, M_TEMP);
2594 					kfree(ioe, M_TEMP);
2595 					goto fail;
2596 				}
2597 				break;
2598 			}
2599 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2600 				kfree(table, M_TEMP);
2601 				kfree(ioe, M_TEMP);
2602 				error = EFAULT;
2603 				goto fail;
2604 			}
2605 		}
2606 		kfree(table, M_TEMP);
2607 		kfree(ioe, M_TEMP);
2608 		break;
2609 	}
2610 
2611 	case DIOCXROLLBACK: {
2612 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2613 		struct pfioc_trans_e	*ioe;
2614 		struct pfr_table	*table;
2615 		int			 i;
2616 
2617 		if (io->esize != sizeof(*ioe)) {
2618 			error = ENODEV;
2619 			goto fail;
2620 		}
2621 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2622 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2623 		for (i = 0; i < io->size; i++) {
2624 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2625 				kfree(table, M_TEMP);
2626 				kfree(ioe, M_TEMP);
2627 				error = EFAULT;
2628 				goto fail;
2629 			}
2630 			switch (ioe->rs_num) {
2631 #ifdef ALTQ
2632 			case PF_RULESET_ALTQ:
2633 				if (ioe->anchor[0]) {
2634 					kfree(table, M_TEMP);
2635 					kfree(ioe, M_TEMP);
2636 					error = EINVAL;
2637 					goto fail;
2638 				}
2639 				if ((error = pf_rollback_altq(ioe->ticket))) {
2640 					kfree(table, M_TEMP);
2641 					kfree(ioe, M_TEMP);
2642 					goto fail; /* really bad */
2643 				}
2644 				break;
2645 #endif /* ALTQ */
2646 			case PF_RULESET_TABLE:
2647 				bzero(table, sizeof(*table));
2648 				strlcpy(table->pfrt_anchor, ioe->anchor,
2649 				    sizeof(table->pfrt_anchor));
2650 				if ((error = pfr_ina_rollback(table,
2651 				    ioe->ticket, NULL, 0))) {
2652 					kfree(table, M_TEMP);
2653 					kfree(ioe, M_TEMP);
2654 					goto fail; /* really bad */
2655 				}
2656 				break;
2657 			default:
2658 				if ((error = pf_rollback_rules(ioe->ticket,
2659 				    ioe->rs_num, ioe->anchor))) {
2660 					kfree(table, M_TEMP);
2661 					kfree(ioe, M_TEMP);
2662 					goto fail; /* really bad */
2663 				}
2664 				break;
2665 			}
2666 		}
2667 		kfree(table, M_TEMP);
2668 		kfree(ioe, M_TEMP);
2669 		break;
2670 	}
2671 
2672 	case DIOCXCOMMIT: {
2673 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2674 		struct pfioc_trans_e	*ioe;
2675 		struct pfr_table	*table;
2676 		struct pf_ruleset	*rs;
2677 		int			 i;
2678 
2679 		if (io->esize != sizeof(*ioe)) {
2680 			error = ENODEV;
2681 			goto fail;
2682 		}
2683 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2684 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2685 		/* first makes sure everything will succeed */
2686 		for (i = 0; i < io->size; i++) {
2687 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2688 				kfree(table, M_TEMP);
2689 				kfree(ioe, M_TEMP);
2690 				error = EFAULT;
2691 				goto fail;
2692 			}
2693 			switch (ioe->rs_num) {
2694 #ifdef ALTQ
2695 			case PF_RULESET_ALTQ:
2696 				if (ioe->anchor[0]) {
2697 					kfree(table, M_TEMP);
2698 					kfree(ioe, M_TEMP);
2699 					error = EINVAL;
2700 					goto fail;
2701 				}
2702 				if (!altqs_inactive_open || ioe->ticket !=
2703 				    ticket_altqs_inactive) {
2704 					kfree(table, M_TEMP);
2705 					kfree(ioe, M_TEMP);
2706 					error = EBUSY;
2707 					goto fail;
2708 				}
2709 				break;
2710 #endif /* ALTQ */
2711 			case PF_RULESET_TABLE:
2712 				rs = pf_find_ruleset(ioe->anchor);
2713 				if (rs == NULL || !rs->topen || ioe->ticket !=
2714 				     rs->tticket) {
2715 					kfree(table, M_TEMP);
2716 					kfree(ioe, M_TEMP);
2717 					error = EBUSY;
2718 					goto fail;
2719 				}
2720 				break;
2721 			default:
2722 				if (ioe->rs_num < 0 || ioe->rs_num >=
2723 				    PF_RULESET_MAX) {
2724 					kfree(table, M_TEMP);
2725 					kfree(ioe, M_TEMP);
2726 					error = EINVAL;
2727 					goto fail;
2728 				}
2729 				rs = pf_find_ruleset(ioe->anchor);
2730 				if (rs == NULL ||
2731 				    !rs->rules[ioe->rs_num].inactive.open ||
2732 				    rs->rules[ioe->rs_num].inactive.ticket !=
2733 				    ioe->ticket) {
2734 					kfree(table, M_TEMP);
2735 					kfree(ioe, M_TEMP);
2736 					error = EBUSY;
2737 					goto fail;
2738 				}
2739 				break;
2740 			}
2741 		}
2742 		/* now do the commit - no errors should happen here */
2743 		for (i = 0; i < io->size; i++) {
2744 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2745 				kfree(table, M_TEMP);
2746 				kfree(ioe, M_TEMP);
2747 				error = EFAULT;
2748 				goto fail;
2749 			}
2750 			switch (ioe->rs_num) {
2751 #ifdef ALTQ
2752 			case PF_RULESET_ALTQ:
2753 				if ((error = pf_commit_altq(ioe->ticket))) {
2754 					kfree(table, M_TEMP);
2755 					kfree(ioe, M_TEMP);
2756 					goto fail; /* really bad */
2757 				}
2758 				break;
2759 #endif /* ALTQ */
2760 			case PF_RULESET_TABLE:
2761 				bzero(table, sizeof(*table));
2762 				strlcpy(table->pfrt_anchor, ioe->anchor,
2763 				    sizeof(table->pfrt_anchor));
2764 				if ((error = pfr_ina_commit(table, ioe->ticket,
2765 				    NULL, NULL, 0))) {
2766 					kfree(table, M_TEMP);
2767 					kfree(ioe, M_TEMP);
2768 					goto fail; /* really bad */
2769 				}
2770 				break;
2771 			default:
2772 				if ((error = pf_commit_rules(ioe->ticket,
2773 				    ioe->rs_num, ioe->anchor))) {
2774 					kfree(table, M_TEMP);
2775 					kfree(ioe, M_TEMP);
2776 					goto fail; /* really bad */
2777 				}
2778 				break;
2779 			}
2780 		}
2781 		kfree(table, M_TEMP);
2782 		kfree(ioe, M_TEMP);
2783 		break;
2784 	}
2785 
2786 	case DIOCGETSRCNODES: {
2787 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2788 		struct pf_src_node	*n, *p, *pstore;
2789 		u_int32_t nr = 0;
2790 		int	space = psn->psn_len;
2791 		int	nn;
2792 
2793 		if (space == 0) {
2794 			for (nn = 0; nn < ncpus; ++nn) {
2795 				RB_FOREACH(n, pf_src_tree,
2796 					   &tree_src_tracking[nn]) {
2797 					nr++;
2798 				}
2799 			}
2800 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2801 			break;
2802 		}
2803 
2804 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2805 
2806 		p = psn->psn_src_nodes;
2807 
2808 		/*
2809 		 * WARNING: We are not switching cpus so we cannot call
2810 		 *	    nominal pf.c support routines for cpu-specific
2811 		 *	    data.
2812 		 */
2813 		for (nn = 0; nn < ncpus; ++nn) {
2814 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
2815 				int	secs = time_second, diff;
2816 
2817 				if ((nr + 1) * sizeof(*p) >
2818 				    (unsigned)psn->psn_len) {
2819 					break;
2820 				}
2821 
2822 				bcopy(n, pstore, sizeof(*pstore));
2823 				if (n->rule.ptr != NULL)
2824 					pstore->rule.nr = n->rule.ptr->nr;
2825 				pstore->creation = secs - pstore->creation;
2826 				if (pstore->expire > secs)
2827 					pstore->expire -= secs;
2828 				else
2829 					pstore->expire = 0;
2830 
2831 				/* adjust the connection rate estimate */
2832 				diff = secs - n->conn_rate.last;
2833 				if (diff >= n->conn_rate.seconds)
2834 					pstore->conn_rate.count = 0;
2835 				else
2836 					pstore->conn_rate.count -=
2837 					    n->conn_rate.count * diff /
2838 					    n->conn_rate.seconds;
2839 
2840 				error = copyout(pstore, p, sizeof(*p));
2841 				if (error) {
2842 					kfree(pstore, M_TEMP);
2843 					goto fail;
2844 				}
2845 				p++;
2846 				nr++;
2847 			}
2848 		}
2849 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2850 		kfree(pstore, M_TEMP);
2851 		break;
2852 	}
2853 
2854 	case DIOCCLRSRCNODES: {
2855 		struct pf_src_node	*n;
2856 		struct pf_state		*state;
2857 		globaldata_t save_gd = mycpu;
2858 		int nn;
2859 
2860 		/*
2861 		 * WARNING: We are not switching cpus so we cannot call
2862 		 *	    nominal pf.c support routines for cpu-specific
2863 		 *	    data.
2864 		 */
2865 		for (nn = 0; nn < ncpus; ++nn) {
2866 			RB_FOREACH(state, pf_state_tree_id, &tree_id[nn]) {
2867 				state->src_node = NULL;
2868 				state->nat_src_node = NULL;
2869 			}
2870 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
2871 				n->expire = 1;
2872 				n->states = 0;
2873 			}
2874 		}
2875 
2876 		/*
2877 		 * WARNING: Must move to the target cpu for nominal calls
2878 		 *	    into pf.c
2879 		 */
2880 		for (nn = 0; nn < ncpus; ++nn) {
2881 			lwkt_setcpu_self(globaldata_find(nn));
2882 			pf_purge_expired_src_nodes(1);
2883 		}
2884 		lwkt_setcpu_self(save_gd);
2885 		pf_status.src_nodes = 0;
2886 		break;
2887 	}
2888 
2889 	case DIOCKILLSRCNODES: {
2890 		struct pf_src_node	*sn;
2891 		struct pf_state		*s;
2892 		struct pfioc_src_node_kill *psnk =
2893 		    (struct pfioc_src_node_kill *)addr;
2894 		u_int			killed = 0;
2895 		globaldata_t save_gd = mycpu;
2896 		int nn;
2897 
2898 		/*
2899 		 * WARNING: We are not switching cpus so we cannot call
2900 		 *	    nominal pf.c support routines for cpu-specific
2901 		 *	    data.
2902 		 */
2903 		for (nn = 0; nn < ncpus; ++nn) {
2904 		    RB_FOREACH(sn, pf_src_tree, &tree_src_tracking[nn]) {
2905 			if (PF_MATCHA(psnk->psnk_src.neg,
2906 				&psnk->psnk_src.addr.v.a.addr,
2907 				&psnk->psnk_src.addr.v.a.mask,
2908 				&sn->addr, sn->af) &&
2909 			    PF_MATCHA(psnk->psnk_dst.neg,
2910 				&psnk->psnk_dst.addr.v.a.addr,
2911 				&psnk->psnk_dst.addr.v.a.mask,
2912 				&sn->raddr, sn->af)) {
2913 				/* Handle state to src_node linkage */
2914 				if (sn->states != 0) {
2915 					RB_FOREACH(s, pf_state_tree_id,
2916 					    &tree_id[nn]) {
2917 						if (s->src_node == sn)
2918 							s->src_node = NULL;
2919 						if (s->nat_src_node == sn)
2920 							s->nat_src_node = NULL;
2921 					}
2922 					sn->states = 0;
2923 				}
2924 				sn->expire = 1;
2925 				killed++;
2926 			}
2927 		    }
2928 		}
2929 		if (killed > 0) {
2930 			for (nn = 0; nn < ncpus; ++nn) {
2931 				lwkt_setcpu_self(globaldata_find(nn));
2932 				pf_purge_expired_src_nodes(1);
2933 			}
2934 			lwkt_setcpu_self(save_gd);
2935 		}
2936 
2937 		psnk->psnk_killed = killed;
2938 		break;
2939 	}
2940 
2941 	case DIOCSETHOSTID: {
2942 		u_int32_t	*hostid = (u_int32_t *)addr;
2943 
2944 		if (*hostid == 0)
2945 			pf_status.hostid = karc4random();
2946 		else
2947 			pf_status.hostid = *hostid;
2948 		break;
2949 	}
2950 
2951 	case DIOCOSFPFLUSH:
2952 		crit_enter();
2953 		pf_osfp_flush();
2954 		crit_exit();
2955 		break;
2956 
2957 	case DIOCIGETIFACES: {
2958 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2959 
2960 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
2961 			error = ENODEV;
2962 			break;
2963 		}
2964 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2965 		    &io->pfiio_size);
2966 		break;
2967 	}
2968 
2969 	case DIOCSETIFFLAG: {
2970 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2971 
2972 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
2973 		break;
2974 	}
2975 
2976 	case DIOCCLRIFFLAG: {
2977 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2978 
2979 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
2980 		break;
2981 	}
2982 
2983 	default:
2984 		error = ENODEV;
2985 		break;
2986 	}
2987 fail:
2988 	lwkt_reltoken(&pf_token);
2989 	return (error);
2990 }
2991 
2992 /*
2993  * XXX - Check for version missmatch!!!
2994  */
2995 static void
2996 pf_clear_states(void)
2997 {
2998 	struct pf_state		*s, *nexts;
2999 	u_int			killed = 0;
3000 	globaldata_t save_gd = mycpu;
3001 	int nn;
3002 
3003 	for (nn = 0; nn < ncpus; ++nn) {
3004 		lwkt_setcpu_self(globaldata_find(nn));
3005 		for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]); s; s = nexts) {
3006 			nexts = RB_NEXT(pf_state_tree_id, &tree_id[nn], s);
3007 
3008 			/* don't send out individual delete messages */
3009 			s->sync_flags = PFSTATE_NOSYNC;
3010 			pf_unlink_state(s);
3011 			killed++;
3012 		}
3013 
3014 	}
3015 	lwkt_setcpu_self(save_gd);
3016 
3017 #if 0 /* PFSYNC */
3018 /*
3019  * XXX This is called on module unload, we do not want to sync that over? */
3020  */
3021 	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
3022 #endif
3023 }
3024 
3025 static int
3026 pf_clear_tables(void)
3027 {
3028 	struct pfioc_table io;
3029 	int error;
3030 
3031 	bzero(&io, sizeof(io));
3032 
3033 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3034 	    io.pfrio_flags);
3035 
3036 	return (error);
3037 }
3038 
3039 static void
3040 pf_clear_srcnodes(void)
3041 {
3042 	struct pf_src_node	*n;
3043 	struct pf_state		*state;
3044 	globaldata_t save_gd = mycpu;
3045 	int nn;
3046 
3047 	for (nn = 0; nn < ncpus; ++nn) {
3048 		lwkt_setcpu_self(globaldata_find(nn));
3049 		RB_FOREACH(state, pf_state_tree_id, &tree_id[nn]) {
3050 			state->src_node = NULL;
3051 			state->nat_src_node = NULL;
3052 		}
3053 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
3054 			n->expire = 1;
3055 			n->states = 0;
3056 		}
3057 		pf_purge_expired_src_nodes(0);
3058 	}
3059 	lwkt_setcpu_self(save_gd);
3060 
3061 	pf_status.src_nodes = 0;
3062 }
3063 
3064 /*
3065  * XXX - Check for version missmatch!!!
3066  */
3067 
3068 /*
3069  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3070  */
3071 static int
3072 shutdown_pf(void)
3073 {
3074 	int error = 0;
3075 	u_int32_t t[5];
3076 	char nn = '\0';
3077 
3078 
3079 	pf_status.running = 0;
3080 	error = dehook_pf();
3081 	if (error) {
3082 		pf_status.running = 1;
3083 		DPFPRINTF(PF_DEBUG_MISC,
3084 		    ("pf: pfil unregistration failed\n"));
3085 		return(error);
3086 	}
3087 	do {
3088 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
3089 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3090 			break;
3091 		}
3092 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
3093 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3094 			break;		/* XXX: rollback? */
3095 		}
3096 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
3097 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3098 			break;		/* XXX: rollback? */
3099 		}
3100 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3101 		    != 0) {
3102 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3103 			break;		/* XXX: rollback? */
3104 		}
3105 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3106 		    != 0) {
3107 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3108 			break;		/* XXX: rollback? */
3109 		}
3110 
3111 		/* XXX: these should always succeed here */
3112 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3113 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3114 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3115 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3116 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3117 
3118 		if ((error = pf_clear_tables()) != 0)
3119 			break;
3120 #ifdef ALTQ
3121 		if ((error = pf_begin_altq(&t[0])) != 0) {
3122 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3123 			break;
3124 		}
3125 		pf_commit_altq(t[0]);
3126 #endif
3127 		pf_clear_states();
3128 		pf_clear_srcnodes();
3129 
3130 		/* status does not use malloced mem so no need to cleanup */
3131 		/* fingerprints and interfaces have their own cleanup code */
3132 	} while(0);
3133         return (error);
3134 }
3135 
3136 static int
3137 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3138 {
3139 	/*
3140 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3141 	 * for ip_len/ip_off. This is why we don't have to change byte order
3142 	 * like the FreeBSD-5 version does.
3143 	 */
3144 	int chk;
3145 
3146 	lwkt_gettoken_shared(&pf_token);
3147 
3148 	chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3149 	if (chk && *m) {
3150 		m_freem(*m);
3151 		*m = NULL;
3152 	}
3153 	lwkt_reltoken(&pf_token);
3154 	return chk;
3155 }
3156 
3157 static int
3158 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3159 {
3160 	/*
3161 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3162 	 * for ip_len/ip_off. This is why we don't have to change byte order
3163 	 * like the FreeBSD-5 version does.
3164 	 */
3165 	int chk;
3166 
3167 	lwkt_gettoken_shared(&pf_token);
3168 
3169 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3170 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3171 		in_delayed_cksum(*m);
3172 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3173 	}
3174 	chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3175 	if (chk && *m) {
3176 		m_freem(*m);
3177 		*m = NULL;
3178 	}
3179 	lwkt_reltoken(&pf_token);
3180 	return chk;
3181 }
3182 
3183 #ifdef INET6
3184 static int
3185 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3186 {
3187 	/*
3188 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3189 	 */
3190 	int chk;
3191 
3192 	lwkt_gettoken_shared(&pf_token);
3193 
3194 	chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3195 	if (chk && *m) {
3196 		m_freem(*m);
3197 		*m = NULL;
3198 	}
3199 	lwkt_reltoken(&pf_token);
3200 	return chk;
3201 }
3202 
3203 static int
3204 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3205 {
3206 	/*
3207 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3208 	 */
3209 	int chk;
3210 
3211 	lwkt_gettoken_shared(&pf_token);
3212 
3213 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3214 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3215 		in_delayed_cksum(*m);
3216 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3217 	}
3218 	chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3219 	if (chk && *m) {
3220 		m_freem(*m);
3221 		*m = NULL;
3222 	}
3223 	lwkt_reltoken(&pf_token);
3224 	return chk;
3225 }
3226 #endif /* INET6 */
3227 
3228 static int
3229 hook_pf(void)
3230 {
3231 	struct pfil_head *pfh_inet;
3232 #ifdef INET6
3233 	struct pfil_head *pfh_inet6;
3234 #endif
3235 
3236 	lwkt_gettoken(&pf_token);
3237 
3238 	if (pf_pfil_hooked) {
3239 		lwkt_reltoken(&pf_token);
3240 		return (0);
3241 	}
3242 
3243 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3244 	if (pfh_inet == NULL) {
3245 		lwkt_reltoken(&pf_token);
3246 		return (ENODEV);
3247 	}
3248 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet);
3249 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet);
3250 #ifdef INET6
3251 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3252 	if (pfh_inet6 == NULL) {
3253 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3254 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3255 		lwkt_reltoken(&pf_token);
3256 		return (ENODEV);
3257 	}
3258 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet6);
3259 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet6);
3260 #endif
3261 
3262 	pf_pfil_hooked = 1;
3263 	lwkt_reltoken(&pf_token);
3264 	return (0);
3265 }
3266 
3267 static int
3268 dehook_pf(void)
3269 {
3270 	struct pfil_head *pfh_inet;
3271 #ifdef INET6
3272 	struct pfil_head *pfh_inet6;
3273 #endif
3274 
3275 	lwkt_gettoken(&pf_token);
3276 
3277 	if (pf_pfil_hooked == 0) {
3278 		lwkt_reltoken(&pf_token);
3279 		return (0);
3280 	}
3281 
3282 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3283 	if (pfh_inet == NULL) {
3284 		lwkt_reltoken(&pf_token);
3285 		return (ENODEV);
3286 	}
3287 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3288 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3289 #ifdef INET6
3290 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3291 	if (pfh_inet6 == NULL) {
3292 		lwkt_reltoken(&pf_token);
3293 		return (ENODEV);
3294 	}
3295 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3296 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3297 #endif
3298 
3299 	pf_pfil_hooked = 0;
3300 	lwkt_reltoken(&pf_token);
3301 	return (0);
3302 }
3303 
3304 static int
3305 pf_load(void)
3306 {
3307 	lwkt_gettoken(&pf_token);
3308 
3309 	pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3310 	pfattach();
3311 	lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3312 	lwkt_reltoken(&pf_token);
3313 	return (0);
3314 }
3315 
3316 static int
3317 pf_mask_del(struct radix_node *rn, void *arg)
3318 {
3319 	struct radix_node_head *rnh = arg;
3320 
3321 	rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
3322 	Free(rn);
3323 	return 0;
3324 }
3325 
3326 static int
3327 pf_unload(void)
3328 {
3329 	int error;
3330 	pf_status.running = 0;
3331 
3332 	lwkt_gettoken(&pf_token);
3333 
3334 	error = dehook_pf();
3335 	if (error) {
3336 		/*
3337 		 * Should not happen!
3338 		 * XXX Due to error code ESRCH, kldunload will show
3339 		 * a message like 'No such process'.
3340 		 */
3341 		kprintf("pfil unregistration fail\n");
3342 		lwkt_reltoken(&pf_token);
3343 		return error;
3344 	}
3345 	shutdown_pf();
3346 	pf_end_threads = 1;
3347 	while (pf_end_threads < 2) {
3348 		wakeup_one(pf_purge_thread);
3349 		tsleep(pf_purge_thread, 0, "pftmo", hz);
3350 	}
3351 	pfi_cleanup();
3352 	pf_osfp_flush();
3353 	dev_ops_remove_all(&pf_ops);
3354 	lockuninit(&pf_consistency_lock);
3355 	lwkt_reltoken(&pf_token);
3356 
3357 	if (pf_maskhead != NULL) {
3358 		pf_maskhead->rnh_walktree(pf_maskhead,
3359 			pf_mask_del, pf_maskhead);
3360 		Free(pf_maskhead);
3361 		pf_maskhead = NULL;
3362 	}
3363 	kmalloc_destroy(&pf_state_pl);
3364 	kmalloc_destroy(&pf_frent_pl);
3365 	kmalloc_destroy(&pf_cent_pl);
3366 	return 0;
3367 }
3368 
3369 static int
3370 pf_modevent(module_t mod, int type, void *data)
3371 {
3372 	int error = 0;
3373 
3374 	lwkt_gettoken(&pf_token);
3375 
3376 	switch(type) {
3377 	case MOD_LOAD:
3378 		error = pf_load();
3379 		break;
3380 
3381 	case MOD_UNLOAD:
3382 		error = pf_unload();
3383 		break;
3384 	default:
3385 		error = EINVAL;
3386 		break;
3387 	}
3388 	lwkt_reltoken(&pf_token);
3389 	return error;
3390 }
3391 
3392 static moduledata_t pf_mod = {
3393 	"pf",
3394 	pf_modevent,
3395 	0
3396 };
3397 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3398 MODULE_VERSION(pf, PF_MODVER);
3399