xref: /dragonfly/sys/net/pf/pf_ioctl.c (revision 31524921)
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 struct lock		 pf_global_statetbl_lock;
121 #ifdef ALTQ
122 static int		 pf_altq_running;
123 #endif
124 
125 #define	TAGID_MAX	 50000
126 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
127 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
128 
129 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
130 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
131 #endif
132 u_int16_t		 tagname2tag(struct pf_tags *, char *);
133 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
134 void			 tag_unref(struct pf_tags *, u_int16_t);
135 int			 pf_rtlabel_add(struct pf_addr_wrap *);
136 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
137 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
138 
139 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) kprintf x
140 
141 static cdev_t	pf_dev;
142 
143 static MALLOC_DEFINE(M_PFRULEPL, "pfrulepl", "pf rule pool list");
144 static MALLOC_DEFINE(M_PFALTQPL, "pfaltqpl", "pf altq pool list");
145 static MALLOC_DEFINE(M_PFPOOLADDRPL, "pfpooladdrpl", "pf pool address pool list");
146 static MALLOC_DEFINE(M_PFFRENTPL, "pffrent", "pf frent pool list");
147 
148 
149 /*
150  * XXX - These are new and need to be checked when moveing to a new version
151  */
152 static void		 pf_clear_states(void);
153 static int		 pf_clear_tables(void);
154 static void		 pf_clear_srcnodes(void);
155 /*
156  * XXX - These are new and need to be checked when moveing to a new version
157  */
158 
159 /*
160  * Wrapper functions for pfil(9) hooks
161  */
162 static int pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp,
163 		int dir);
164 static int pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp,
165 		int dir);
166 #ifdef INET6
167 static int pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp,
168 		int dir);
169 static int pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp,
170 		int dir);
171 #endif
172 
173 static int 		 hook_pf(void);
174 static int 		 dehook_pf(void);
175 static int 		 shutdown_pf(void);
176 static int 		 pf_load(void);
177 static int 		 pf_unload(void);
178 
179 d_open_t	pfopen;
180 d_close_t	pfclose;
181 d_ioctl_t	pfioctl;
182 
183 static struct dev_ops pf_ops = {	    /* XXX convert to port model */
184 	{ PF_NAME, 73, 0 },
185 	.d_open =	pfopen,
186 	.d_close =	pfclose,
187 	.d_ioctl =	pfioctl
188 };
189 
190 static volatile int pf_pfil_hooked = 0;
191 int pf_end_threads = 0;
192 
193 int debug_pfugidhack = 0;
194 SYSCTL_INT(_debug, OID_AUTO, pfugidhack, CTLFLAG_RW, &debug_pfugidhack, 0,
195 	"Enable/disable pf user/group rules mpsafe hack");
196 
197 void
198 pfattach(void)
199 {
200 	u_int32_t *my_timeout = pf_default_rule.timeout;
201 	int nn;
202 
203 	if (!rn_inithead((void **)&pf_maskhead, NULL, 0)) {
204 		kprintf("pf mask radix tree create failed\n");
205 		return;
206 	}
207 	kmalloc_create(&pf_state_pl, "pf state pool list");
208 	kmalloc_raise_limit(pf_state_pl, 0);
209 	kmalloc_create(&pf_frent_pl, "pf fragment pool list");
210 	kmalloc_raise_limit(pf_frent_pl, 0);
211 	kmalloc_create(&pf_cent_pl, "pf cent pool list");
212 	kmalloc_raise_limit(pf_cent_pl, 0);
213 
214 	pfr_initialize();
215 	pfi_initialize();
216 	pf_osfp_initialize();
217 
218 	pf_pool_limits[PF_LIMIT_STATES].pp = pf_state_pl;
219 	pf_pool_limits[PF_LIMIT_STATES].limit = PFSTATE_HIWAT;
220 	pf_pool_limits[PF_LIMIT_FRAGS].pp = pf_frent_pl;
221 	pf_pool_limits[PF_LIMIT_FRAGS].limit = PFFRAG_FRENT_HIWAT;
222 	if (ctob(physmem) <= 100*1024*1024)
223 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
224 		    PFR_KENTRY_HIWAT_SMALL;
225 
226 	for (nn = 0; nn < ncpus; ++nn) {
227 		RB_INIT(&tree_src_tracking[nn]);
228 		RB_INIT(&tree_id[nn]);
229 	}
230 	RB_INIT(&pf_anchors);
231 	pf_init_ruleset(&pf_main_ruleset);
232 	TAILQ_INIT(&pf_altqs[0]);
233 	TAILQ_INIT(&pf_altqs[1]);
234 	TAILQ_INIT(&pf_pabuf);
235 	pf_altqs_active = &pf_altqs[0];
236 	pf_altqs_inactive = &pf_altqs[1];
237 	for (nn = 0; nn < ncpus; ++nn)
238 		TAILQ_INIT(&state_list[nn]);
239 
240 	/* default rule should never be garbage collected */
241 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
242 	pf_default_rule.action = PF_PASS;
243 	pf_default_rule.nr = (uint32_t)(-1);
244 	pf_default_rule.rtableid = -1;
245 
246 	/* initialize default timeouts */
247 	my_timeout[PFTM_TCP_FIRST_PACKET] = 120;	/* First TCP packet */
248 	my_timeout[PFTM_TCP_OPENING] = 30; 		/* No response yet */
249 	my_timeout[PFTM_TCP_ESTABLISHED] = 24*60*60;	/* Established */
250 	my_timeout[PFTM_TCP_CLOSING] = 15 * 60;		/* Half closed */
251 	my_timeout[PFTM_TCP_FIN_WAIT] = 45;		/* Got both FINs */
252 	my_timeout[PFTM_TCP_CLOSED] = 90;		/* Got a RST */
253 	my_timeout[PFTM_UDP_FIRST_PACKET] = 60;		/* First UDP packet */
254 	my_timeout[PFTM_UDP_SINGLE] = 30;		/* Unidirectional */
255 	my_timeout[PFTM_UDP_MULTIPLE] = 60;		/* Bidirectional */
256 	my_timeout[PFTM_ICMP_FIRST_PACKET] = 20;	/* First ICMP packet */
257 	my_timeout[PFTM_ICMP_ERROR_REPLY] = 10;		/* Got error response */
258 	my_timeout[PFTM_OTHER_FIRST_PACKET] = 60;	/* First packet */
259 	my_timeout[PFTM_OTHER_SINGLE] = 30;		/* Unidirectional */
260 	my_timeout[PFTM_OTHER_MULTIPLE] = 60;		/* Bidirectional */
261 	my_timeout[PFTM_FRAG] = 30;			/* Fragment expire */
262 	my_timeout[PFTM_INTERVAL] = 10;			/* Expire interval */
263 	my_timeout[PFTM_SRC_NODE] = 0;		/* Source Tracking */
264 	my_timeout[PFTM_TS_DIFF] = 30;		/* Allowed TS diff */
265 	my_timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
266 	my_timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
267 
268 	pf_normalize_init();
269 	bzero(&pf_status, sizeof(pf_status));
270 	pf_status.debug = PF_DEBUG_URGENT;
271 	/* XXX do our best to avoid a conflict */
272 	pf_status.hostid = karc4random();
273 
274 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
275 		panic("pfpurge thread");
276 }
277 
278 int
279 pfopen(struct dev_open_args *ap)
280 {
281 	lwkt_gettoken(&pf_token);
282 	cdev_t dev = ap->a_head.a_dev;
283 	if (minor(dev) >= 1) {
284 		lwkt_reltoken(&pf_token);
285 		return (ENXIO);
286 	}
287 	lwkt_reltoken(&pf_token);
288 	return (0);
289 }
290 
291 int
292 pfclose(struct dev_close_args *ap)
293 {
294 	lwkt_gettoken(&pf_token);
295 	cdev_t dev = ap->a_head.a_dev;
296 	if (minor(dev) >= 1) {
297 		lwkt_reltoken(&pf_token);
298 		return (ENXIO);
299 	}
300 	lwkt_reltoken(&pf_token);
301 	return (0);
302 }
303 
304 struct pf_pool *
305 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
306     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
307     u_int8_t check_ticket)
308 {
309 	struct pf_ruleset	*ruleset;
310 	struct pf_rule		*rule;
311 	int			 rs_num;
312 
313 	ruleset = pf_find_ruleset(anchor);
314 	if (ruleset == NULL)
315 		return (NULL);
316 	rs_num = pf_get_ruleset_number(rule_action);
317 	if (rs_num >= PF_RULESET_MAX)
318 		return (NULL);
319 	if (active) {
320 		if (check_ticket && ticket !=
321 		    ruleset->rules[rs_num].active.ticket)
322 			return (NULL);
323 		if (r_last)
324 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
325 			    pf_rulequeue);
326 		else
327 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
328 	} else {
329 		if (check_ticket && ticket !=
330 		    ruleset->rules[rs_num].inactive.ticket)
331 			return (NULL);
332 		if (r_last)
333 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
334 			    pf_rulequeue);
335 		else
336 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
337 	}
338 	if (!r_last) {
339 		while ((rule != NULL) && (rule->nr != rule_number))
340 			rule = TAILQ_NEXT(rule, entries);
341 	}
342 	if (rule == NULL)
343 		return (NULL);
344 
345 	return (&rule->rpool);
346 }
347 
348 void
349 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
350 {
351 	struct pf_pooladdr	*mv_pool_pa;
352 
353 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
354 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
355 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
356 	}
357 }
358 
359 void
360 pf_empty_pool(struct pf_palist *poola)
361 {
362 	struct pf_pooladdr	*empty_pool_pa;
363 
364 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
365 		pfi_dynaddr_remove(&empty_pool_pa->addr);
366 		pf_tbladdr_remove(&empty_pool_pa->addr);
367 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
368 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
369 		kfree(empty_pool_pa, M_PFPOOLADDRPL);
370 	}
371 }
372 
373 void
374 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
375 {
376 	if (rulequeue != NULL) {
377 		if (rule->states_cur <= 0) {
378 			/*
379 			 * XXX - we need to remove the table *before* detaching
380 			 * the rule to make sure the table code does not delete
381 			 * the anchor under our feet.
382 			 */
383 			pf_tbladdr_remove(&rule->src.addr);
384 			pf_tbladdr_remove(&rule->dst.addr);
385 			if (rule->overload_tbl)
386 				pfr_detach_table(rule->overload_tbl);
387 		}
388 		TAILQ_REMOVE(rulequeue, rule, entries);
389 		rule->entries.tqe_prev = NULL;
390 		rule->nr = -1;
391 	}
392 
393 	if (rule->states_cur > 0 || rule->src_nodes > 0 ||
394 	    rule->entries.tqe_prev != NULL)
395 		return;
396 	pf_tag_unref(rule->tag);
397 	pf_tag_unref(rule->match_tag);
398 #ifdef ALTQ
399 	if (rule->pqid != rule->qid)
400 		pf_qid_unref(rule->pqid);
401 	pf_qid_unref(rule->qid);
402 #endif
403 	pf_rtlabel_remove(&rule->src.addr);
404 	pf_rtlabel_remove(&rule->dst.addr);
405 	pfi_dynaddr_remove(&rule->src.addr);
406 	pfi_dynaddr_remove(&rule->dst.addr);
407 	if (rulequeue == NULL) {
408 		pf_tbladdr_remove(&rule->src.addr);
409 		pf_tbladdr_remove(&rule->dst.addr);
410 		if (rule->overload_tbl)
411 			pfr_detach_table(rule->overload_tbl);
412 	}
413 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
414 	pf_anchor_remove(rule);
415 	pf_empty_pool(&rule->rpool.list);
416 	kfree(rule, M_PFRULEPL);
417 }
418 
419 u_int16_t
420 tagname2tag(struct pf_tags *head, char *tagname)
421 {
422 	struct pf_tagname	*tag, *p = NULL;
423 	u_int16_t		 new_tagid = 1;
424 
425 	TAILQ_FOREACH(tag, head, entries)
426 		if (strcmp(tagname, tag->name) == 0) {
427 			tag->ref++;
428 			return (tag->tag);
429 		}
430 
431 	/*
432 	 * to avoid fragmentation, we do a linear search from the beginning
433 	 * and take the first free slot we find. if there is none or the list
434 	 * is empty, append a new entry at the end.
435 	 */
436 
437 	/* new entry */
438 	if (!TAILQ_EMPTY(head))
439 		for (p = TAILQ_FIRST(head); p != NULL &&
440 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
441 			new_tagid = p->tag + 1;
442 
443 	if (new_tagid > TAGID_MAX)
444 		return (0);
445 
446 	/* allocate and fill new struct pf_tagname */
447 	tag = kmalloc(sizeof(*tag), M_TEMP, M_WAITOK);
448 	strlcpy(tag->name, tagname, sizeof(tag->name));
449 	tag->tag = new_tagid;
450 	tag->ref++;
451 
452 	if (p != NULL)	/* insert new entry before p */
453 		TAILQ_INSERT_BEFORE(p, tag, entries);
454 	else	/* either list empty or no free slot in between */
455 		TAILQ_INSERT_TAIL(head, tag, entries);
456 
457 	return (tag->tag);
458 }
459 
460 void
461 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
462 {
463 	struct pf_tagname	*tag;
464 
465 	TAILQ_FOREACH(tag, head, entries)
466 		if (tag->tag == tagid) {
467 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
468 			return;
469 		}
470 }
471 
472 void
473 tag_unref(struct pf_tags *head, u_int16_t tag)
474 {
475 	struct pf_tagname	*p, *next;
476 
477 	if (tag == 0)
478 		return;
479 
480 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
481 		next = TAILQ_NEXT(p, entries);
482 		if (tag == p->tag) {
483 			if (--p->ref == 0) {
484 				TAILQ_REMOVE(head, p, entries);
485 				kfree(p, M_TEMP);
486 			}
487 			break;
488 		}
489 	}
490 }
491 
492 u_int16_t
493 pf_tagname2tag(char *tagname)
494 {
495 	return (tagname2tag(&pf_tags, tagname));
496 }
497 
498 void
499 pf_tag2tagname(u_int16_t tagid, char *p)
500 {
501 	tag2tagname(&pf_tags, tagid, p);
502 }
503 
504 void
505 pf_tag_ref(u_int16_t tag)
506 {
507 	struct pf_tagname *t;
508 
509 	TAILQ_FOREACH(t, &pf_tags, entries)
510 		if (t->tag == tag)
511 			break;
512 	if (t != NULL)
513 		t->ref++;
514 }
515 
516 void
517 pf_tag_unref(u_int16_t tag)
518 {
519 	tag_unref(&pf_tags, tag);
520 }
521 
522 int
523 pf_rtlabel_add(struct pf_addr_wrap *a)
524 {
525 	return (0);
526 }
527 
528 void
529 pf_rtlabel_remove(struct pf_addr_wrap *a)
530 {
531 }
532 
533 void
534 pf_rtlabel_copyout(struct pf_addr_wrap *a)
535 {
536 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel)
537 		strlcpy(a->v.rtlabelname, "?", sizeof(a->v.rtlabelname));
538 }
539 
540 #ifdef ALTQ
541 u_int32_t
542 pf_qname2qid(char *qname)
543 {
544 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
545 }
546 
547 void
548 pf_qid2qname(u_int32_t qid, char *p)
549 {
550 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
551 }
552 
553 void
554 pf_qid_unref(u_int32_t qid)
555 {
556 	tag_unref(&pf_qids, (u_int16_t)qid);
557 }
558 
559 int
560 pf_begin_altq(u_int32_t *ticket)
561 {
562 	struct pf_altq	*altq;
563 	int		 error = 0;
564 
565 	/* Purge the old altq list */
566 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
567 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
568 		if (altq->qname[0] == 0) {
569 			/* detach and destroy the discipline */
570 			error = altq_remove(altq);
571 		} else
572 			pf_qid_unref(altq->qid);
573 		kfree(altq, M_PFALTQPL);
574 	}
575 	if (error)
576 		return (error);
577 	*ticket = ++ticket_altqs_inactive;
578 	altqs_inactive_open = 1;
579 	return (0);
580 }
581 
582 int
583 pf_rollback_altq(u_int32_t ticket)
584 {
585 	struct pf_altq	*altq;
586 	int		 error = 0;
587 
588 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
589 		return (0);
590 	/* Purge the old altq list */
591 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
592 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
593 		if (altq->qname[0] == 0) {
594 			/* detach and destroy the discipline */
595 			error = altq_remove(altq);
596 		} else
597 			pf_qid_unref(altq->qid);
598 		kfree(altq, M_PFALTQPL);
599 	}
600 	altqs_inactive_open = 0;
601 	return (error);
602 }
603 
604 int
605 pf_commit_altq(u_int32_t ticket)
606 {
607 	struct pf_altqqueue	*old_altqs;
608 	struct pf_altq		*altq;
609 	int			 err, error = 0;
610 
611 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
612 		return (EBUSY);
613 
614 	/* swap altqs, keep the old. */
615 	crit_enter();
616 	old_altqs = pf_altqs_active;
617 	pf_altqs_active = pf_altqs_inactive;
618 	pf_altqs_inactive = old_altqs;
619 	ticket_altqs_active = ticket_altqs_inactive;
620 
621 	/* Attach new disciplines */
622 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
623 		if (altq->qname[0] == 0) {
624 			/* attach the discipline */
625 			error = altq_pfattach(altq);
626 			if (error) {
627 				crit_exit();
628 				return (error);
629 			}
630 		}
631 	}
632 
633 	/* Purge the old altq list */
634 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
635 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
636 		if (altq->qname[0] == 0) {
637 			/* detach and destroy the discipline */
638 			if (pf_altq_running)
639 				error = pf_disable_altq(altq);
640 			err = altq_pfdetach(altq);
641 			if (err != 0 && error == 0)
642 				error = err;
643 			err = altq_remove(altq);
644 			if (err != 0 && error == 0)
645 				error = err;
646 		} else
647 			pf_qid_unref(altq->qid);
648 		kfree(altq, M_PFALTQPL);
649 	}
650 	crit_exit();
651 
652 	altqs_inactive_open = 0;
653 	return (error);
654 }
655 
656 int
657 pf_enable_altq(struct pf_altq *altq)
658 {
659 	struct ifnet		*ifp;
660 	struct tb_profile	 tb;
661 	int			 error = 0;
662 
663 	ifnet_lock();
664 
665 	if ((ifp = ifunit(altq->ifname)) == NULL) {
666 		ifnet_unlock();
667 		return (EINVAL);
668 	}
669 
670 	if (ifp->if_snd.altq_type != ALTQT_NONE)
671 		error = altq_enable(&ifp->if_snd);
672 
673 	/* set tokenbucket regulator */
674 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
675 		tb.rate = altq->ifbandwidth;
676 		tb.depth = altq->tbrsize;
677 		crit_enter();
678 		error = tbr_set(&ifp->if_snd, &tb);
679 		crit_exit();
680 	}
681 
682 	ifnet_unlock();
683 	return (error);
684 }
685 
686 int
687 pf_disable_altq(struct pf_altq *altq)
688 {
689 	struct ifnet		*ifp;
690 	struct tb_profile	 tb;
691 	int			 error;
692 
693 	ifnet_lock();
694 
695 	if ((ifp = ifunit(altq->ifname)) == NULL) {
696 		ifnet_unlock();
697 		return (EINVAL);
698 	}
699 
700 	/*
701 	 * when the discipline is no longer referenced, it was overridden
702 	 * by a new one.  if so, just return.
703 	 */
704 	if (altq->altq_disc != ifp->if_snd.altq_disc) {
705 		ifnet_unlock();
706 		return (0);
707 	}
708 
709 	error = altq_disable(&ifp->if_snd);
710 
711 	if (error == 0) {
712 		/* clear tokenbucket regulator */
713 		tb.rate = 0;
714 		crit_enter();
715 		error = tbr_set(&ifp->if_snd, &tb);
716 		crit_exit();
717 	}
718 
719 	ifnet_unlock();
720 	return (error);
721 }
722 #endif /* ALTQ */
723 
724 int
725 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
726 {
727 	struct pf_ruleset	*rs;
728 	struct pf_rule		*rule;
729 
730 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
731 		return (EINVAL);
732 	rs = pf_find_or_create_ruleset(anchor);
733 	if (rs == NULL)
734 		return (EINVAL);
735 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
736 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
737 		rs->rules[rs_num].inactive.rcount--;
738 	}
739 	*ticket = ++rs->rules[rs_num].inactive.ticket;
740 	rs->rules[rs_num].inactive.open = 1;
741 	return (0);
742 }
743 
744 int
745 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
746 {
747 	struct pf_ruleset	*rs;
748 	struct pf_rule		*rule;
749 
750 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
751 		return (EINVAL);
752 	rs = pf_find_ruleset(anchor);
753 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
754 	    rs->rules[rs_num].inactive.ticket != ticket)
755 		return (0);
756 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
757 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
758 		rs->rules[rs_num].inactive.rcount--;
759 	}
760 	rs->rules[rs_num].inactive.open = 0;
761 	return (0);
762 }
763 
764 #define PF_MD5_UPD(st, elm)						\
765 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
766 
767 #define PF_MD5_UPD_STR(st, elm)						\
768 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
769 
770 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
771 		(stor) = htonl((st)->elm);				\
772 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
773 } while (0)
774 
775 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
776 		(stor) = htons((st)->elm);				\
777 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
778 } while (0)
779 
780 void
781 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
782 {
783 	PF_MD5_UPD(pfr, addr.type);
784 	switch (pfr->addr.type) {
785 		case PF_ADDR_DYNIFTL:
786 			PF_MD5_UPD(pfr, addr.v.ifname);
787 			PF_MD5_UPD(pfr, addr.iflags);
788 			break;
789 		case PF_ADDR_TABLE:
790 			PF_MD5_UPD(pfr, addr.v.tblname);
791 			break;
792 		case PF_ADDR_ADDRMASK:
793 			/* XXX ignore af? */
794 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
795 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
796 			break;
797 		case PF_ADDR_RTLABEL:
798 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
799 			break;
800 	}
801 
802 	PF_MD5_UPD(pfr, port[0]);
803 	PF_MD5_UPD(pfr, port[1]);
804 	PF_MD5_UPD(pfr, neg);
805 	PF_MD5_UPD(pfr, port_op);
806 }
807 
808 void
809 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
810 {
811 	u_int16_t x;
812 	u_int32_t y;
813 
814 	pf_hash_rule_addr(ctx, &rule->src);
815 	pf_hash_rule_addr(ctx, &rule->dst);
816 	PF_MD5_UPD_STR(rule, label);
817 	PF_MD5_UPD_STR(rule, ifname);
818 	PF_MD5_UPD_STR(rule, match_tagname);
819 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
820 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
821 	PF_MD5_UPD_HTONL(rule, prob, y);
822 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
823 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
824 	PF_MD5_UPD(rule, uid.op);
825 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
826 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
827 	PF_MD5_UPD(rule, gid.op);
828 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
829 	PF_MD5_UPD(rule, action);
830 	PF_MD5_UPD(rule, direction);
831 	PF_MD5_UPD(rule, af);
832 	PF_MD5_UPD(rule, quick);
833 	PF_MD5_UPD(rule, ifnot);
834 	PF_MD5_UPD(rule, match_tag_not);
835 	PF_MD5_UPD(rule, natpass);
836 	PF_MD5_UPD(rule, keep_state);
837 	PF_MD5_UPD(rule, proto);
838 	PF_MD5_UPD(rule, type);
839 	PF_MD5_UPD(rule, code);
840 	PF_MD5_UPD(rule, flags);
841 	PF_MD5_UPD(rule, flagset);
842 	PF_MD5_UPD(rule, allow_opts);
843 	PF_MD5_UPD(rule, rt);
844 	PF_MD5_UPD(rule, tos);
845 }
846 
847 int
848 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
849 {
850 	struct pf_ruleset	*rs;
851 	struct pf_rule		*rule, **old_array;
852 	struct pf_rulequeue	*old_rules;
853 	int			 error;
854 	u_int32_t		 old_rcount;
855 
856 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
857 		return (EINVAL);
858 	rs = pf_find_ruleset(anchor);
859 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
860 	    ticket != rs->rules[rs_num].inactive.ticket)
861 		return (EBUSY);
862 
863 	/* Calculate checksum for the main ruleset */
864 	if (rs == &pf_main_ruleset) {
865 		error = pf_setup_pfsync_matching(rs);
866 		if (error != 0)
867 			return (error);
868 	}
869 
870 	/* Swap rules, keep the old. */
871 	crit_enter();
872 	old_rules = rs->rules[rs_num].active.ptr;
873 	old_rcount = rs->rules[rs_num].active.rcount;
874 	old_array = rs->rules[rs_num].active.ptr_array;
875 
876 	rs->rules[rs_num].active.ptr =
877 	    rs->rules[rs_num].inactive.ptr;
878 	rs->rules[rs_num].active.ptr_array =
879 	    rs->rules[rs_num].inactive.ptr_array;
880 	rs->rules[rs_num].active.rcount =
881 	    rs->rules[rs_num].inactive.rcount;
882 	rs->rules[rs_num].inactive.ptr = old_rules;
883 	rs->rules[rs_num].inactive.ptr_array = old_array;
884 	rs->rules[rs_num].inactive.rcount = old_rcount;
885 
886 	rs->rules[rs_num].active.ticket =
887 	    rs->rules[rs_num].inactive.ticket;
888 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
889 
890 
891 	/* Purge the old rule list. */
892 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
893 		pf_rm_rule(old_rules, rule);
894 	if (rs->rules[rs_num].inactive.ptr_array)
895 		kfree(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
896 	rs->rules[rs_num].inactive.ptr_array = NULL;
897 	rs->rules[rs_num].inactive.rcount = 0;
898 	rs->rules[rs_num].inactive.open = 0;
899 	pf_remove_if_empty_ruleset(rs);
900 	crit_exit();
901 	return (0);
902 }
903 
904 int
905 pf_setup_pfsync_matching(struct pf_ruleset *rs)
906 {
907 	MD5_CTX			 ctx;
908 	struct pf_rule		*rule;
909 	int			 rs_cnt;
910 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
911 
912 	MD5Init(&ctx);
913 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
914 		/* XXX PF_RULESET_SCRUB as well? */
915 		if (rs_cnt == PF_RULESET_SCRUB)
916 			continue;
917 
918 		if (rs->rules[rs_cnt].inactive.ptr_array)
919 			kfree(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
920 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
921 
922 		if (rs->rules[rs_cnt].inactive.rcount) {
923 			rs->rules[rs_cnt].inactive.ptr_array =
924 			    kmalloc(sizeof(caddr_t) *
925 				    rs->rules[rs_cnt].inactive.rcount,
926 				    M_TEMP, M_WAITOK);
927 		}
928 
929 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
930 		    entries) {
931 			pf_hash_rule(&ctx, rule);
932 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
933 		}
934 	}
935 
936 	MD5Final(digest, &ctx);
937 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
938 	return (0);
939 }
940 
941 int
942 pf_addr_setup(struct pf_ruleset *ruleset, struct pf_addr_wrap *addr,
943     sa_family_t af)
944 {
945 	if (pfi_dynaddr_setup(addr, af) ||
946 	    pf_tbladdr_setup(ruleset, addr))
947 		return (EINVAL);
948 
949 	return (0);
950 }
951 
952 void
953 pf_addr_copyout(struct pf_addr_wrap *addr)
954 {
955 	pfi_dynaddr_copyout(addr);
956 	pf_tbladdr_copyout(addr);
957 	pf_rtlabel_copyout(addr);
958 }
959 
960 int
961 pfioctl(struct dev_ioctl_args *ap)
962 {
963 	u_long cmd = ap->a_cmd;
964 	caddr_t addr = ap->a_data;
965 	struct pf_pooladdr	*pa = NULL;
966 	struct pf_pool		*pool = NULL;
967 	int			 error = 0;
968 
969 	lwkt_gettoken(&pf_token);
970 
971 	/* XXX keep in sync with switch() below */
972 	if (securelevel > 1) {
973 		switch (cmd) {
974 		case DIOCGETRULES:
975 		case DIOCGETRULE:
976 		case DIOCGETADDRS:
977 		case DIOCGETADDR:
978 		case DIOCGETSTATE:
979 		case DIOCSETSTATUSIF:
980 		case DIOCGETSTATUS:
981 		case DIOCCLRSTATUS:
982 		case DIOCNATLOOK:
983 		case DIOCSETDEBUG:
984 		case DIOCGETSTATES:
985 		case DIOCGETTIMEOUT:
986 		case DIOCCLRRULECTRS:
987 		case DIOCGETLIMIT:
988 		case DIOCGETALTQS:
989 		case DIOCGETALTQ:
990 		case DIOCGETQSTATS:
991 		case DIOCGETRULESETS:
992 		case DIOCGETRULESET:
993 		case DIOCRGETTABLES:
994 		case DIOCRGETTSTATS:
995 		case DIOCRCLRTSTATS:
996 		case DIOCRCLRADDRS:
997 		case DIOCRADDADDRS:
998 		case DIOCRDELADDRS:
999 		case DIOCRSETADDRS:
1000 		case DIOCRGETADDRS:
1001 		case DIOCRGETASTATS:
1002 		case DIOCRCLRASTATS:
1003 		case DIOCRTSTADDRS:
1004 		case DIOCOSFPGET:
1005 		case DIOCGETSRCNODES:
1006 		case DIOCCLRSRCNODES:
1007 		case DIOCIGETIFACES:
1008 		case DIOCSETIFFLAG:
1009 		case DIOCCLRIFFLAG:
1010 		case DIOCGIFSPEED:
1011 			break;
1012 		case DIOCRCLRTABLES:
1013 		case DIOCRADDTABLES:
1014 		case DIOCRDELTABLES:
1015 		case DIOCRSETTFLAGS:
1016 			if (((struct pfioc_table *)addr)->pfrio_flags &
1017 			    PFR_FLAG_DUMMY)
1018 				break; /* dummy operation ok */
1019 			lwkt_reltoken(&pf_token);
1020 			return (EPERM);
1021 		default:
1022 			lwkt_reltoken(&pf_token);
1023 			return (EPERM);
1024 		}
1025 	}
1026 
1027 	if (!(ap->a_fflag & FWRITE)) {
1028 		switch (cmd) {
1029 		case DIOCGETRULES:
1030 		case DIOCGETADDRS:
1031 		case DIOCGETADDR:
1032 		case DIOCGETSTATE:
1033 		case DIOCGETSTATUS:
1034 		case DIOCGETSTATES:
1035 		case DIOCGETTIMEOUT:
1036 		case DIOCGETLIMIT:
1037 		case DIOCGETALTQS:
1038 		case DIOCGETALTQ:
1039 		case DIOCGETQSTATS:
1040 		case DIOCGETRULESETS:
1041 		case DIOCGETRULESET:
1042 		case DIOCNATLOOK:
1043 		case DIOCRGETTABLES:
1044 		case DIOCRGETTSTATS:
1045 		case DIOCRGETADDRS:
1046 		case DIOCRGETASTATS:
1047 		case DIOCRTSTADDRS:
1048 		case DIOCOSFPGET:
1049 		case DIOCGETSRCNODES:
1050 		case DIOCIGETIFACES:
1051 		case DIOCGIFSPEED:
1052 			break;
1053 		case DIOCRCLRTABLES:
1054 		case DIOCRADDTABLES:
1055 		case DIOCRDELTABLES:
1056 		case DIOCRCLRTSTATS:
1057 		case DIOCRCLRADDRS:
1058 		case DIOCRADDADDRS:
1059 		case DIOCRDELADDRS:
1060 		case DIOCRSETADDRS:
1061 		case DIOCRSETTFLAGS:
1062 			if (((struct pfioc_table *)addr)->pfrio_flags &
1063 			    PFR_FLAG_DUMMY)
1064 				break; /* dummy operation ok */
1065 			lwkt_reltoken(&pf_token);
1066 			return (EACCES);
1067 		case DIOCGETRULE:
1068 			if (((struct pfioc_rule *)addr)->action ==
1069 			    PF_GET_CLR_CNTR) {
1070 				lwkt_reltoken(&pf_token);
1071 				return (EACCES);
1072 			}
1073 			break;
1074 		default:
1075 			lwkt_reltoken(&pf_token);
1076 			return (EACCES);
1077 		}
1078 	}
1079 
1080 	switch (cmd) {
1081 	case DIOCSTART:
1082 		if (pf_status.running)
1083 			error = EEXIST;
1084 		else {
1085 			error = hook_pf();
1086 			if (error) {
1087 				DPFPRINTF(PF_DEBUG_MISC,
1088 				    ("pf: pfil registration fail\n"));
1089 				break;
1090 			}
1091 			pf_status.running = 1;
1092 			pf_status.since = time_second;
1093 			if (pf_status.stateid == 0) {
1094 				pf_status.stateid = time_second;
1095 				pf_status.stateid = pf_status.stateid << 32;
1096 			}
1097 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1098 		}
1099 		break;
1100 
1101 	case DIOCSTOP:
1102 		if (!pf_status.running)
1103 			error = ENOENT;
1104 		else {
1105 			pf_status.running = 0;
1106 			error = dehook_pf();
1107 			if (error) {
1108 				pf_status.running = 1;
1109 				DPFPRINTF(PF_DEBUG_MISC,
1110 					("pf: pfil unregistration failed\n"));
1111 			}
1112 			pf_status.since = time_second;
1113 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1114 		}
1115 		break;
1116 
1117 	case DIOCADDRULE: {
1118 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1119 		struct pf_ruleset	*ruleset;
1120 		struct pf_rule		*rule, *tail;
1121 		struct pf_pooladdr	*pa;
1122 		int			 rs_num;
1123 
1124 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1125 		ruleset = pf_find_ruleset(pr->anchor);
1126 		if (ruleset == NULL) {
1127 			error = EINVAL;
1128 			break;
1129 		}
1130 		rs_num = pf_get_ruleset_number(pr->rule.action);
1131 		if (rs_num >= PF_RULESET_MAX) {
1132 			error = EINVAL;
1133 			break;
1134 		}
1135 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1136 			error = EINVAL;
1137 			break;
1138 		}
1139 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1140 			error = EBUSY;
1141 			break;
1142 		}
1143 		if (pr->pool_ticket != ticket_pabuf) {
1144 			error = EBUSY;
1145 			break;
1146 		}
1147 		rule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK);
1148 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1149 		rule->cuid = ap->a_cred->cr_ruid;
1150 		rule->cpid = 0;
1151 		rule->anchor = NULL;
1152 		rule->kif = NULL;
1153 		TAILQ_INIT(&rule->rpool.list);
1154 		/* initialize refcounting */
1155 		rule->states_cur = 0;
1156 		rule->src_nodes = 0;
1157 		rule->entries.tqe_prev = NULL;
1158 #ifndef INET
1159 		if (rule->af == AF_INET) {
1160 			kfree(rule, M_PFRULEPL);
1161 			error = EAFNOSUPPORT;
1162 			break;
1163 		}
1164 #endif /* INET */
1165 #ifndef INET6
1166 		if (rule->af == AF_INET6) {
1167 			kfree(rule, M_PFRULEPL);
1168 			error = EAFNOSUPPORT;
1169 			break;
1170 		}
1171 #endif /* INET6 */
1172 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1173 		    pf_rulequeue);
1174 		if (tail)
1175 			rule->nr = tail->nr + 1;
1176 		else
1177 			rule->nr = 0;
1178 		if (rule->ifname[0]) {
1179 			rule->kif = pfi_kif_get(rule->ifname);
1180 			if (rule->kif == NULL) {
1181 				kfree(rule, M_PFRULEPL);
1182 				error = EINVAL;
1183 				break;
1184 			}
1185 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1186 		}
1187 
1188 		if (rule->rtableid > 0 && rule->rtableid > rt_numfibs)
1189 			error = EBUSY;
1190 
1191 #ifdef ALTQ
1192 		/* set queue IDs */
1193 		if (rule->qname[0] != 0) {
1194 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1195 				error = EBUSY;
1196 			else if (rule->pqname[0] != 0) {
1197 				if ((rule->pqid =
1198 				    pf_qname2qid(rule->pqname)) == 0)
1199 					error = EBUSY;
1200 			} else
1201 				rule->pqid = rule->qid;
1202 		}
1203 #endif
1204 		if (rule->tagname[0])
1205 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1206 				error = EBUSY;
1207 		if (rule->match_tagname[0])
1208 			if ((rule->match_tag =
1209 			    pf_tagname2tag(rule->match_tagname)) == 0)
1210 				error = EBUSY;
1211 		if (rule->rt && !rule->direction)
1212 			error = EINVAL;
1213 #if NPFLOG > 0
1214 		if (!rule->log)
1215 			rule->logif = 0;
1216 		if (rule->logif >= PFLOGIFS_MAX)
1217 			error = EINVAL;
1218 #endif
1219 		if (pf_rtlabel_add(&rule->src.addr) ||
1220 		    pf_rtlabel_add(&rule->dst.addr))
1221 			error = EBUSY;
1222 		if (pf_addr_setup(ruleset, &rule->src.addr, rule->af))
1223 			error = EINVAL;
1224 		if (pf_addr_setup(ruleset, &rule->dst.addr, rule->af))
1225 			error = EINVAL;
1226 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1227 			error = EINVAL;
1228 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1229 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1230 				error = EINVAL;
1231 
1232 		if (rule->overload_tblname[0]) {
1233 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1234 			    rule->overload_tblname)) == NULL)
1235 				error = EINVAL;
1236 			else
1237 				rule->overload_tbl->pfrkt_flags |=
1238 				    PFR_TFLAG_ACTIVE;
1239 		}
1240 
1241 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1242 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1243 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1244 		    (rule->rt > PF_FASTROUTE)) &&
1245 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1246 			error = EINVAL;
1247 
1248 		if (error) {
1249 			pf_rm_rule(NULL, rule);
1250 			break;
1251 		}
1252 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1253 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1254 		    rule->bytes[0] = rule->bytes[1] = 0;
1255 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1256 		    rule, entries);
1257 		ruleset->rules[rs_num].inactive.rcount++;
1258 		break;
1259 	}
1260 
1261 	case DIOCGETRULES: {
1262 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1263 		struct pf_ruleset	*ruleset;
1264 		struct pf_rule		*tail;
1265 		int			 rs_num;
1266 
1267 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1268 		ruleset = pf_find_ruleset(pr->anchor);
1269 		if (ruleset == NULL) {
1270 			error = EINVAL;
1271 			break;
1272 		}
1273 		rs_num = pf_get_ruleset_number(pr->rule.action);
1274 		if (rs_num >= PF_RULESET_MAX) {
1275 			error = EINVAL;
1276 			break;
1277 		}
1278 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1279 		    pf_rulequeue);
1280 		if (tail)
1281 			pr->nr = tail->nr + 1;
1282 		else
1283 			pr->nr = 0;
1284 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1285 		break;
1286 	}
1287 
1288 	case DIOCGETRULE: {
1289 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1290 		struct pf_ruleset	*ruleset;
1291 		struct pf_rule		*rule;
1292 		int			 rs_num, i;
1293 
1294 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1295 		ruleset = pf_find_ruleset(pr->anchor);
1296 		if (ruleset == NULL) {
1297 			error = EINVAL;
1298 			break;
1299 		}
1300 		rs_num = pf_get_ruleset_number(pr->rule.action);
1301 		if (rs_num >= PF_RULESET_MAX) {
1302 			error = EINVAL;
1303 			break;
1304 		}
1305 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1306 			error = EBUSY;
1307 			break;
1308 		}
1309 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1310 		while ((rule != NULL) && (rule->nr != pr->nr))
1311 			rule = TAILQ_NEXT(rule, entries);
1312 		if (rule == NULL) {
1313 			error = EBUSY;
1314 			break;
1315 		}
1316 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1317 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1318 			error = EBUSY;
1319 			break;
1320 		}
1321 		pf_addr_copyout(&pr->rule.src.addr);
1322 		pf_addr_copyout(&pr->rule.dst.addr);
1323 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1324 			if (rule->skip[i].ptr == NULL)
1325 				pr->rule.skip[i].nr = (uint32_t)(-1);
1326 			else
1327 				pr->rule.skip[i].nr =
1328 				    rule->skip[i].ptr->nr;
1329 
1330 		if (pr->action == PF_GET_CLR_CNTR) {
1331 			rule->evaluations = 0;
1332 			rule->packets[0] = rule->packets[1] = 0;
1333 			rule->bytes[0] = rule->bytes[1] = 0;
1334 			rule->states_tot = 0;
1335 		}
1336 		break;
1337 	}
1338 
1339 	case DIOCCHANGERULE: {
1340 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1341 		struct pf_ruleset	*ruleset;
1342 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1343 		u_int32_t		 nr = 0;
1344 		int			 rs_num;
1345 
1346 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1347 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1348 		    pcr->pool_ticket != ticket_pabuf) {
1349 			error = EBUSY;
1350 			break;
1351 		}
1352 
1353 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1354 		    pcr->action > PF_CHANGE_GET_TICKET) {
1355 			error = EINVAL;
1356 			break;
1357 		}
1358 		ruleset = pf_find_ruleset(pcr->anchor);
1359 		if (ruleset == NULL) {
1360 			error = EINVAL;
1361 			break;
1362 		}
1363 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1364 		if (rs_num >= PF_RULESET_MAX) {
1365 			error = EINVAL;
1366 			break;
1367 		}
1368 
1369 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1370 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1371 			break;
1372 		} else {
1373 			if (pcr->ticket !=
1374 			    ruleset->rules[rs_num].active.ticket) {
1375 				error = EINVAL;
1376 				break;
1377 			}
1378 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1379 				error = EINVAL;
1380 				break;
1381 			}
1382 		}
1383 
1384 		if (pcr->action != PF_CHANGE_REMOVE) {
1385 			newrule = kmalloc(sizeof(struct pf_rule), M_PFRULEPL, M_WAITOK|M_NULLOK);
1386 			if (newrule == NULL) {
1387 				error = ENOMEM;
1388 				break;
1389 			}
1390 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1391 			newrule->cuid = ap->a_cred->cr_ruid;
1392 			newrule->cpid = 0;
1393 			TAILQ_INIT(&newrule->rpool.list);
1394 			/* initialize refcounting */
1395 			newrule->states_cur = 0;
1396 			newrule->entries.tqe_prev = NULL;
1397 #ifndef INET
1398 			if (newrule->af == AF_INET) {
1399 				kfree(newrule, M_PFRULEPL);
1400 				error = EAFNOSUPPORT;
1401 				break;
1402 			}
1403 #endif /* INET */
1404 #ifndef INET6
1405 			if (newrule->af == AF_INET6) {
1406 				kfree(newrule, M_PFRULEPL);
1407 				error = EAFNOSUPPORT;
1408 				break;
1409 			}
1410 #endif /* INET6 */
1411 			if (newrule->ifname[0]) {
1412 				newrule->kif = pfi_kif_get(newrule->ifname);
1413 				if (newrule->kif == NULL) {
1414 					kfree(newrule, M_PFRULEPL);
1415 					error = EINVAL;
1416 					break;
1417 				}
1418 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1419 			} else
1420 				newrule->kif = NULL;
1421 
1422 			if (newrule->rtableid > 0 &&
1423 			    newrule->rtableid > rt_numfibs)
1424 				error = EBUSY;
1425 
1426 #ifdef ALTQ
1427 			/* set queue IDs */
1428 			if (newrule->qname[0] != 0) {
1429 				if ((newrule->qid =
1430 				    pf_qname2qid(newrule->qname)) == 0)
1431 					error = EBUSY;
1432 				else if (newrule->pqname[0] != 0) {
1433 					if ((newrule->pqid =
1434 					    pf_qname2qid(newrule->pqname)) == 0)
1435 						error = EBUSY;
1436 				} else
1437 					newrule->pqid = newrule->qid;
1438 			}
1439 #endif /* ALTQ */
1440 			if (newrule->tagname[0])
1441 				if ((newrule->tag =
1442 				    pf_tagname2tag(newrule->tagname)) == 0)
1443 					error = EBUSY;
1444 			if (newrule->match_tagname[0])
1445 				if ((newrule->match_tag = pf_tagname2tag(
1446 				    newrule->match_tagname)) == 0)
1447 					error = EBUSY;
1448 			if (newrule->rt && !newrule->direction)
1449 				error = EINVAL;
1450 #if NPFLOG > 0
1451 			if (!newrule->log)
1452 				newrule->logif = 0;
1453 			if (newrule->logif >= PFLOGIFS_MAX)
1454 				error = EINVAL;
1455 #endif
1456 			if (pf_rtlabel_add(&newrule->src.addr) ||
1457 			    pf_rtlabel_add(&newrule->dst.addr))
1458 				error = EBUSY;
1459 			if (pf_addr_setup(ruleset, &newrule->src.addr, newrule->af))
1460 				error = EINVAL;
1461 			if (pf_addr_setup(ruleset, &newrule->dst.addr, newrule->af))
1462 				error = EINVAL;
1463 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1464 				error = EINVAL;
1465 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1466 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1467 					error = EINVAL;
1468 
1469 			if (newrule->overload_tblname[0]) {
1470 				if ((newrule->overload_tbl = pfr_attach_table(
1471 				    ruleset, newrule->overload_tblname)) ==
1472 				    NULL)
1473 					error = EINVAL;
1474 				else
1475 					newrule->overload_tbl->pfrkt_flags |=
1476 					    PFR_TFLAG_ACTIVE;
1477 			}
1478 
1479 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1480 			if (((((newrule->action == PF_NAT) ||
1481 			    (newrule->action == PF_RDR) ||
1482 			    (newrule->action == PF_BINAT) ||
1483 			    (newrule->rt > PF_FASTROUTE)) &&
1484 			    !newrule->anchor)) &&
1485 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1486 				error = EINVAL;
1487 
1488 			if (error) {
1489 				pf_rm_rule(NULL, newrule);
1490 				break;
1491 			}
1492 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1493 			newrule->evaluations = 0;
1494 			newrule->packets[0] = newrule->packets[1] = 0;
1495 			newrule->bytes[0] = newrule->bytes[1] = 0;
1496 		}
1497 		pf_empty_pool(&pf_pabuf);
1498 
1499 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1500 			oldrule = TAILQ_FIRST(
1501 			    ruleset->rules[rs_num].active.ptr);
1502 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1503 			oldrule = TAILQ_LAST(
1504 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1505 		else {
1506 			oldrule = TAILQ_FIRST(
1507 			    ruleset->rules[rs_num].active.ptr);
1508 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1509 				oldrule = TAILQ_NEXT(oldrule, entries);
1510 			if (oldrule == NULL) {
1511 				if (newrule != NULL)
1512 					pf_rm_rule(NULL, newrule);
1513 				error = EINVAL;
1514 				break;
1515 			}
1516 		}
1517 
1518 		if (pcr->action == PF_CHANGE_REMOVE) {
1519 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1520 			ruleset->rules[rs_num].active.rcount--;
1521 		} else {
1522 			if (oldrule == NULL)
1523 				TAILQ_INSERT_TAIL(
1524 				    ruleset->rules[rs_num].active.ptr,
1525 				    newrule, entries);
1526 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1527 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1528 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1529 			else
1530 				TAILQ_INSERT_AFTER(
1531 				    ruleset->rules[rs_num].active.ptr,
1532 				    oldrule, newrule, entries);
1533 			ruleset->rules[rs_num].active.rcount++;
1534 		}
1535 
1536 		nr = 0;
1537 		TAILQ_FOREACH(oldrule,
1538 		    ruleset->rules[rs_num].active.ptr, entries)
1539 			oldrule->nr = nr++;
1540 
1541 		ruleset->rules[rs_num].active.ticket++;
1542 
1543 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1544 		pf_remove_if_empty_ruleset(ruleset);
1545 
1546 		break;
1547 	}
1548 
1549 	case DIOCCLRSTATES: {
1550 		struct pf_state		*s, *nexts;
1551 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1552 		u_int			 killed = 0;
1553 		globaldata_t save_gd = mycpu;
1554 		int nn;
1555 
1556 		for (nn = 0; nn < ncpus; ++nn) {
1557 			lwkt_setcpu_self(globaldata_find(nn));
1558 			for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]);
1559 			     s; s = nexts) {
1560 				nexts = RB_NEXT(pf_state_tree_id,
1561 						&tree_id[nn], s);
1562 
1563 				if (!psk->psk_ifname[0] ||
1564 				    !strcmp(psk->psk_ifname,
1565 					    s->kif->pfik_name)) {
1566 					/*
1567 					 * don't send out individual
1568 					 * delete messages
1569 					 */
1570 					s->sync_flags = PFSTATE_NOSYNC;
1571 					pf_unlink_state(s);
1572 					killed++;
1573 				}
1574 			}
1575 		}
1576 		lwkt_setcpu_self(save_gd);
1577 		psk->psk_killed = killed;
1578 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1579 		break;
1580 	}
1581 
1582 	case DIOCKILLSTATES: {
1583 		struct pf_state		*s, *nexts;
1584 		struct pf_state_key	*sk;
1585 		struct pf_addr		*srcaddr, *dstaddr;
1586 		u_int16_t		 srcport, dstport;
1587 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1588 		u_int			 killed = 0;
1589 		globaldata_t save_gd = mycpu;
1590 		int nn;
1591 
1592 		if (psk->psk_pfcmp.id) {
1593 			if (psk->psk_pfcmp.creatorid == 0)
1594 				psk->psk_pfcmp.creatorid = pf_status.hostid;
1595 			for (nn = 0; nn < ncpus; ++nn) {
1596 				lwkt_setcpu_self(globaldata_find(nn));
1597 				if ((s = pf_find_state_byid(&psk->psk_pfcmp))) {
1598 					/* send immediate delete of state */
1599 					pfsync_delete_state(s);
1600 					s->sync_flags |= PFSTATE_NOSYNC;
1601 					pf_unlink_state(s);
1602 					++psk->psk_killed;
1603 				}
1604 			}
1605 			lwkt_setcpu_self(save_gd);
1606 			break;
1607 		}
1608 
1609 		for (nn = 0; nn < ncpus; ++nn) {
1610 		    lwkt_setcpu_self(globaldata_find(nn));
1611 		    for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]);
1612 			 s; s = nexts) {
1613 			    nexts = RB_NEXT(pf_state_tree_id, &tree_id[nn], s);
1614 			    sk = s->key[PF_SK_WIRE];
1615 
1616 			    if (s->direction == PF_OUT) {
1617 				    srcaddr = &sk->addr[1];
1618 				    dstaddr = &sk->addr[0];
1619 				    srcport = sk->port[0];
1620 				    dstport = sk->port[0];
1621 			    } else {
1622 				    srcaddr = &sk->addr[0];
1623 				    dstaddr = &sk->addr[1];
1624 				    srcport = sk->port[0];
1625 				    dstport = sk->port[0];
1626 			    }
1627 			    if ((!psk->psk_af || sk->af == psk->psk_af)
1628 				&& (!psk->psk_proto || psk->psk_proto ==
1629 						       sk->proto) &&
1630 				PF_MATCHA(psk->psk_src.neg,
1631 					  &psk->psk_src.addr.v.a.addr,
1632 					  &psk->psk_src.addr.v.a.mask,
1633 					  srcaddr, sk->af) &&
1634 				PF_MATCHA(psk->psk_dst.neg,
1635 					  &psk->psk_dst.addr.v.a.addr,
1636 					  &psk->psk_dst.addr.v.a.mask,
1637 					  dstaddr, sk->af) &&
1638 				(psk->psk_src.port_op == 0 ||
1639 				 pf_match_port(psk->psk_src.port_op,
1640 					       psk->psk_src.port[0],
1641 					       psk->psk_src.port[1],
1642 					       srcport)) &&
1643 				(psk->psk_dst.port_op == 0 ||
1644 				 pf_match_port(psk->psk_dst.port_op,
1645 					       psk->psk_dst.port[0],
1646 					       psk->psk_dst.port[1],
1647 					       dstport)) &&
1648 				(!psk->psk_label[0] ||
1649 				 (s->rule.ptr->label[0] &&
1650 				  !strcmp(psk->psk_label, s->rule.ptr->label))) &&
1651 				(!psk->psk_ifname[0] ||
1652 				 !strcmp(psk->psk_ifname, s->kif->pfik_name))) {
1653 				    /* send immediate delete of state */
1654 				    pfsync_delete_state(s);
1655 				    s->sync_flags |= PFSTATE_NOSYNC;
1656 				    pf_unlink_state(s);
1657 				    killed++;
1658 			    }
1659 		    }
1660 		}
1661 		lwkt_setcpu_self(save_gd);
1662 		psk->psk_killed = killed;
1663 		break;
1664 	}
1665 
1666 	case DIOCADDSTATE: {
1667 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1668 		struct pfsync_state	*sp = &ps->state;
1669 
1670 		if (sp->timeout >= PFTM_MAX &&
1671 		    sp->timeout != PFTM_UNTIL_PACKET) {
1672 			error = EINVAL;
1673 			break;
1674 		}
1675 		error = pfsync_state_import(sp, PFSYNC_SI_IOCTL);
1676 		break;
1677 	}
1678 
1679 	case DIOCGETSTATE: {
1680 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1681 		struct pf_state		*s;
1682 		struct pf_state_cmp	 id_key;
1683 		globaldata_t save_gd = mycpu;
1684 		int nn;
1685 
1686 		bcopy(ps->state.id, &id_key.id, sizeof(id_key.id));
1687 		id_key.creatorid = ps->state.creatorid;
1688 		s = NULL;
1689 		for (nn = 0; nn < ncpus; ++nn) {
1690 			lwkt_setcpu_self(globaldata_find(nn));
1691 			s = pf_find_state_byid(&id_key);
1692 			if (s)
1693 				break;
1694 		}
1695 		if (s) {
1696 			pfsync_state_export(&ps->state, s);
1697 		} else {
1698 			error = ENOENT;
1699 		}
1700 		lwkt_setcpu_self(save_gd);
1701 		break;
1702 	}
1703 
1704 	case DIOCGETSTATES: {
1705 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1706 		struct pf_state		*state;
1707 		struct pfsync_state	*p, *pstore;
1708 		u_int32_t		 nr = 0;
1709 		globaldata_t save_gd = mycpu;
1710 		int nn;
1711 
1712 		if (ps->ps_len == 0) {
1713 			nr = pf_status.states;
1714 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1715 			break;
1716 		}
1717 
1718 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1719 
1720 		p = ps->ps_states;
1721 
1722 		for (nn = 0; nn < ncpus; ++nn) {
1723 			lwkt_setcpu_self(globaldata_find(nn));
1724 			state = TAILQ_FIRST(&state_list[nn]);
1725 			while (state) {
1726 				if (state->timeout != PFTM_UNLINKED) {
1727 					if ((nr + 1) * sizeof(*p) >
1728 					    (unsigned)ps->ps_len) {
1729 						break;
1730 					}
1731 					pfsync_state_export(pstore, state);
1732 					error = copyout(pstore, p, sizeof(*p));
1733 					if (error) {
1734 						kfree(pstore, M_TEMP);
1735 						lwkt_setcpu_self(save_gd);
1736 						goto fail;
1737 					}
1738 					p++;
1739 					nr++;
1740 				}
1741 				state = TAILQ_NEXT(state, entry_list);
1742 			}
1743 		}
1744 		lwkt_setcpu_self(save_gd);
1745 		ps->ps_len = sizeof(struct pfsync_state) * nr;
1746 		kfree(pstore, M_TEMP);
1747 		break;
1748 	}
1749 
1750 	case DIOCGETSTATUS: {
1751 		struct pf_status *s = (struct pf_status *)addr;
1752 		bcopy(&pf_status, s, sizeof(struct pf_status));
1753 		pfi_update_status(s->ifname, s);
1754 		break;
1755 	}
1756 
1757 	case DIOCSETSTATUSIF: {
1758 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
1759 
1760 		if (pi->ifname[0] == 0) {
1761 			bzero(pf_status.ifname, IFNAMSIZ);
1762 			break;
1763 		}
1764 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
1765 		break;
1766 	}
1767 
1768 	case DIOCCLRSTATUS: {
1769 		bzero(pf_status.counters, sizeof(pf_status.counters));
1770 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
1771 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
1772 		pf_status.since = time_second;
1773 		if (*pf_status.ifname)
1774 			pfi_update_status(pf_status.ifname, NULL);
1775 		break;
1776 	}
1777 
1778 	case DIOCNATLOOK: {
1779 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
1780 		struct pf_state_key	*sk;
1781 		struct pf_state		*state;
1782 		struct pf_state_key_cmp	 key;
1783 		int			 m = 0, direction = pnl->direction;
1784 		int			 sidx, didx;
1785 		globaldata_t save_gd = mycpu;
1786 		int nn;
1787 
1788 		/* NATLOOK src and dst are reversed, so reverse sidx/didx */
1789 		sidx = (direction == PF_IN) ? 1 : 0;
1790 		didx = (direction == PF_IN) ? 0 : 1;
1791 
1792 		if (!pnl->proto ||
1793 		    PF_AZERO(&pnl->saddr, pnl->af) ||
1794 		    PF_AZERO(&pnl->daddr, pnl->af) ||
1795 		    ((pnl->proto == IPPROTO_TCP ||
1796 		    pnl->proto == IPPROTO_UDP) &&
1797 		    (!pnl->dport || !pnl->sport)))
1798 			error = EINVAL;
1799 		else {
1800 			key.af = pnl->af;
1801 			key.proto = pnl->proto;
1802 			PF_ACPY(&key.addr[sidx], &pnl->saddr, pnl->af);
1803 			key.port[sidx] = pnl->sport;
1804 			PF_ACPY(&key.addr[didx], &pnl->daddr, pnl->af);
1805 			key.port[didx] = pnl->dport;
1806 
1807 			state = NULL;
1808 			for (nn = 0; nn < ncpus; ++nn) {
1809 				lwkt_setcpu_self(globaldata_find(nn));
1810 				state = pf_find_state_all(&key, direction, &m);
1811 				if (state || m > 1)
1812 					break;
1813 				m = 0;
1814 			}
1815 
1816 			if (m > 1) {
1817 				error = E2BIG;	/* more than one state */
1818 			} else if (state != NULL) {
1819 				sk = state->key[sidx];
1820 				PF_ACPY(&pnl->rsaddr, &sk->addr[sidx], sk->af);
1821 				pnl->rsport = sk->port[sidx];
1822 				PF_ACPY(&pnl->rdaddr, &sk->addr[didx], sk->af);
1823 				pnl->rdport = sk->port[didx];
1824 			} else {
1825 				error = ENOENT;
1826 			}
1827 			lwkt_setcpu_self(save_gd);
1828 		}
1829 		break;
1830 	}
1831 
1832 	case DIOCSETTIMEOUT: {
1833 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1834 		int		 old;
1835 
1836 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
1837 		    pt->seconds < 0) {
1838 			error = EINVAL;
1839 			goto fail;
1840 		}
1841 		old = pf_default_rule.timeout[pt->timeout];
1842 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
1843 			pt->seconds = 1;
1844 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
1845 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
1846 			wakeup(pf_purge_thread);
1847 		pt->seconds = old;
1848 		break;
1849 	}
1850 
1851 	case DIOCGETTIMEOUT: {
1852 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
1853 
1854 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
1855 			error = EINVAL;
1856 			goto fail;
1857 		}
1858 		pt->seconds = pf_default_rule.timeout[pt->timeout];
1859 		break;
1860 	}
1861 
1862 	case DIOCGETLIMIT: {
1863 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1864 
1865 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
1866 			error = EINVAL;
1867 			goto fail;
1868 		}
1869 		pl->limit = pf_pool_limits[pl->index].limit;
1870 		break;
1871 	}
1872 
1873 	case DIOCSETLIMIT: {
1874 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
1875 		int			 old_limit;
1876 
1877 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
1878 		    pf_pool_limits[pl->index].pp == NULL) {
1879 			error = EINVAL;
1880 			goto fail;
1881 		}
1882 
1883 		/* XXX Get an API to set limits on the zone/pool */
1884 		old_limit = pf_pool_limits[pl->index].limit;
1885 		pf_pool_limits[pl->index].limit = pl->limit;
1886 		pl->limit = old_limit;
1887 		break;
1888 	}
1889 
1890 	case DIOCSETDEBUG: {
1891 		u_int32_t	*level = (u_int32_t *)addr;
1892 
1893 		pf_status.debug = *level;
1894 		break;
1895 	}
1896 
1897 	case DIOCCLRRULECTRS: {
1898 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
1899 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
1900 		struct pf_rule		*rule;
1901 
1902 		TAILQ_FOREACH(rule,
1903 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
1904 			rule->evaluations = 0;
1905 			rule->packets[0] = rule->packets[1] = 0;
1906 			rule->bytes[0] = rule->bytes[1] = 0;
1907 		}
1908 		break;
1909 	}
1910 
1911 	case DIOCGIFSPEED: {
1912 		struct pf_ifspeed	*psp = (struct pf_ifspeed *)addr;
1913 		struct pf_ifspeed	ps;
1914 		struct ifnet		*ifp;
1915 
1916 		if (psp->ifname[0] != 0) {
1917 			/* Can we completely trust user-land? */
1918 			strlcpy(ps.ifname, psp->ifname, IFNAMSIZ);
1919 			ifnet_lock();
1920 			ifp = ifunit(ps.ifname);
1921 			if (ifp )
1922 				psp->baudrate = ifp->if_baudrate;
1923 			else
1924 				error = EINVAL;
1925 			ifnet_unlock();
1926 		} else
1927 			error = EINVAL;
1928 		break;
1929 	}
1930 #ifdef ALTQ
1931 	case DIOCSTARTALTQ: {
1932 		struct pf_altq		*altq;
1933 
1934 		/* enable all altq interfaces on active list */
1935 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1936 			if (altq->qname[0] == 0) {
1937 				error = pf_enable_altq(altq);
1938 				if (error != 0)
1939 					break;
1940 			}
1941 		}
1942 		if (error == 0)
1943 			pf_altq_running = 1;
1944 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
1945 		break;
1946 	}
1947 
1948 	case DIOCSTOPALTQ: {
1949 		struct pf_altq		*altq;
1950 
1951 		/* disable all altq interfaces on active list */
1952 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
1953 			if (altq->qname[0] == 0) {
1954 				error = pf_disable_altq(altq);
1955 				if (error != 0)
1956 					break;
1957 			}
1958 		}
1959 		if (error == 0)
1960 			pf_altq_running = 0;
1961 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
1962 		break;
1963 	}
1964 
1965 	case DIOCADDALTQ: {
1966 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
1967 		struct pf_altq		*altq, *a;
1968 
1969 		if (pa->ticket != ticket_altqs_inactive) {
1970 			error = EBUSY;
1971 			break;
1972 		}
1973 		altq = kmalloc(sizeof(struct pf_altq), M_PFALTQPL, M_WAITOK|M_NULLOK);
1974 		if (altq == NULL) {
1975 			error = ENOMEM;
1976 			break;
1977 		}
1978 		bcopy(&pa->altq, altq, sizeof(struct pf_altq));
1979 
1980 		/*
1981 		 * if this is for a queue, find the discipline and
1982 		 * copy the necessary fields
1983 		 */
1984 		if (altq->qname[0] != 0) {
1985 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
1986 				error = EBUSY;
1987 				kfree(altq, M_PFALTQPL);
1988 				break;
1989 			}
1990 			altq->altq_disc = NULL;
1991 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
1992 				if (strncmp(a->ifname, altq->ifname,
1993 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
1994 					altq->altq_disc = a->altq_disc;
1995 					break;
1996 				}
1997 			}
1998 		}
1999 
2000 		error = altq_add(altq);
2001 		if (error) {
2002 			kfree(altq, M_PFALTQPL);
2003 			break;
2004 		}
2005 
2006 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2007 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2008 		break;
2009 	}
2010 
2011 	case DIOCGETALTQS: {
2012 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2013 		struct pf_altq		*altq;
2014 
2015 		pa->nr = 0;
2016 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2017 			pa->nr++;
2018 		pa->ticket = ticket_altqs_active;
2019 		break;
2020 	}
2021 
2022 	case DIOCGETALTQ: {
2023 		struct pfioc_altq	*pa = (struct pfioc_altq *)addr;
2024 		struct pf_altq		*altq;
2025 		u_int32_t		 nr;
2026 
2027 		if (pa->ticket != ticket_altqs_active) {
2028 			error = EBUSY;
2029 			break;
2030 		}
2031 		nr = 0;
2032 		altq = TAILQ_FIRST(pf_altqs_active);
2033 		while ((altq != NULL) && (nr < pa->nr)) {
2034 			altq = TAILQ_NEXT(altq, entries);
2035 			nr++;
2036 		}
2037 		if (altq == NULL) {
2038 			error = EBUSY;
2039 			break;
2040 		}
2041 		bcopy(altq, &pa->altq, sizeof(struct pf_altq));
2042 		break;
2043 	}
2044 
2045 	case DIOCCHANGEALTQ:
2046 		/* CHANGEALTQ not supported yet! */
2047 		error = ENODEV;
2048 		break;
2049 
2050 	case DIOCGETQSTATS: {
2051 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2052 		struct pf_altq		*altq;
2053 		u_int32_t		 nr;
2054 		int			 nbytes;
2055 
2056 		if (pq->ticket != ticket_altqs_active) {
2057 			error = EBUSY;
2058 			break;
2059 		}
2060 		nbytes = pq->nbytes;
2061 		nr = 0;
2062 		altq = TAILQ_FIRST(pf_altqs_active);
2063 		while ((altq != NULL) && (nr < pq->nr)) {
2064 			altq = TAILQ_NEXT(altq, entries);
2065 			nr++;
2066 		}
2067 		if (altq == NULL) {
2068 			error = EBUSY;
2069 			break;
2070 		}
2071 		error = altq_getqstats(altq, pq->buf, &nbytes);
2072 		if (error == 0) {
2073 			pq->scheduler = altq->scheduler;
2074 			pq->nbytes = nbytes;
2075 		}
2076 		break;
2077 	}
2078 #endif /* ALTQ */
2079 
2080 	case DIOCBEGINADDRS: {
2081 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2082 
2083 		pf_empty_pool(&pf_pabuf);
2084 		pp->ticket = ++ticket_pabuf;
2085 		break;
2086 	}
2087 
2088 	case DIOCADDADDR: {
2089 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2090 
2091 		if (pp->ticket != ticket_pabuf) {
2092 			error = EBUSY;
2093 			break;
2094 		}
2095 #ifndef INET
2096 		if (pp->af == AF_INET) {
2097 			error = EAFNOSUPPORT;
2098 			break;
2099 		}
2100 #endif /* INET */
2101 #ifndef INET6
2102 		if (pp->af == AF_INET6) {
2103 			error = EAFNOSUPPORT;
2104 			break;
2105 		}
2106 #endif /* INET6 */
2107 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2108 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2109 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2110 			error = EINVAL;
2111 			break;
2112 		}
2113 		pa = kmalloc(sizeof(struct pf_altq), M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2114 		if (pa == NULL) {
2115 			error = ENOMEM;
2116 			break;
2117 		}
2118 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2119 		if (pa->ifname[0]) {
2120 			pa->kif = pfi_kif_get(pa->ifname);
2121 			if (pa->kif == NULL) {
2122 				kfree(ap, M_PFPOOLADDRPL);
2123 				error = EINVAL;
2124 				break;
2125 			}
2126 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2127 		}
2128 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2129 			pfi_dynaddr_remove(&pa->addr);
2130 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2131 			kfree(pa, M_PFPOOLADDRPL);
2132 			error = EINVAL;
2133 			break;
2134 		}
2135 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2136 		break;
2137 	}
2138 
2139 	case DIOCGETADDRS: {
2140 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2141 
2142 		pp->nr = 0;
2143 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2144 		    pp->r_num, 0, 1, 0);
2145 		if (pool == NULL) {
2146 			error = EBUSY;
2147 			break;
2148 		}
2149 		TAILQ_FOREACH(pa, &pool->list, entries)
2150 			pp->nr++;
2151 		break;
2152 	}
2153 
2154 	case DIOCGETADDR: {
2155 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2156 		u_int32_t		 nr = 0;
2157 
2158 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2159 		    pp->r_num, 0, 1, 1);
2160 		if (pool == NULL) {
2161 			error = EBUSY;
2162 			break;
2163 		}
2164 		pa = TAILQ_FIRST(&pool->list);
2165 		while ((pa != NULL) && (nr < pp->nr)) {
2166 			pa = TAILQ_NEXT(pa, entries);
2167 			nr++;
2168 		}
2169 		if (pa == NULL) {
2170 			error = EBUSY;
2171 			break;
2172 		}
2173 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2174 		pf_addr_copyout(&pp->addr.addr);
2175 		break;
2176 	}
2177 
2178 	case DIOCCHANGEADDR: {
2179 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2180 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2181 		struct pf_ruleset	*ruleset;
2182 
2183 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2184 		    pca->action > PF_CHANGE_REMOVE) {
2185 			error = EINVAL;
2186 			break;
2187 		}
2188 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2189 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2190 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2191 			error = EINVAL;
2192 			break;
2193 		}
2194 
2195 		ruleset = pf_find_ruleset(pca->anchor);
2196 		if (ruleset == NULL) {
2197 			error = EBUSY;
2198 			break;
2199 		}
2200 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2201 		    pca->r_num, pca->r_last, 1, 1);
2202 		if (pool == NULL) {
2203 			error = EBUSY;
2204 			break;
2205 		}
2206 		if (pca->action != PF_CHANGE_REMOVE) {
2207 			newpa = kmalloc(sizeof(struct pf_pooladdr),
2208 				M_PFPOOLADDRPL, M_WAITOK|M_NULLOK);
2209 			if (newpa == NULL) {
2210 				error = ENOMEM;
2211 				break;
2212 			}
2213 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2214 #ifndef INET
2215 			if (pca->af == AF_INET) {
2216 				kfree(newpa, M_PFPOOLADDRPL);
2217 				error = EAFNOSUPPORT;
2218 				break;
2219 			}
2220 #endif /* INET */
2221 #ifndef INET6
2222 			if (pca->af == AF_INET6) {
2223 				kfree(newpa, M_PFPOOLADDRPL);
2224 				error = EAFNOSUPPORT;
2225 				break;
2226 			}
2227 #endif /* INET6 */
2228 			if (newpa->ifname[0]) {
2229 				newpa->kif = pfi_kif_get(newpa->ifname);
2230 				if (newpa->kif == NULL) {
2231 					kfree(newpa, M_PFPOOLADDRPL);
2232 					error = EINVAL;
2233 					break;
2234 				}
2235 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2236 			} else
2237 				newpa->kif = NULL;
2238 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2239 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2240 				pfi_dynaddr_remove(&newpa->addr);
2241 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2242 				kfree(newpa, M_PFPOOLADDRPL);
2243 				error = EINVAL;
2244 				break;
2245 			}
2246 		}
2247 
2248 		if (pca->action == PF_CHANGE_ADD_HEAD)
2249 			oldpa = TAILQ_FIRST(&pool->list);
2250 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2251 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2252 		else {
2253 			int	i = 0;
2254 
2255 			oldpa = TAILQ_FIRST(&pool->list);
2256 			while ((oldpa != NULL) && (i < pca->nr)) {
2257 				oldpa = TAILQ_NEXT(oldpa, entries);
2258 				i++;
2259 			}
2260 			if (oldpa == NULL) {
2261 				error = EINVAL;
2262 				break;
2263 			}
2264 		}
2265 
2266 		if (pca->action == PF_CHANGE_REMOVE) {
2267 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2268 			pfi_dynaddr_remove(&oldpa->addr);
2269 			pf_tbladdr_remove(&oldpa->addr);
2270 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2271 			kfree(oldpa, M_PFPOOLADDRPL);
2272 		} else {
2273 			if (oldpa == NULL)
2274 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2275 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2276 			    pca->action == PF_CHANGE_ADD_BEFORE)
2277 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2278 			else
2279 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2280 				    newpa, entries);
2281 		}
2282 
2283 		pool->cur = TAILQ_FIRST(&pool->list);
2284 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2285 		    pca->af);
2286 		break;
2287 	}
2288 
2289 	case DIOCGETRULESETS: {
2290 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2291 		struct pf_ruleset	*ruleset;
2292 		struct pf_anchor	*anchor;
2293 
2294 		pr->path[sizeof(pr->path) - 1] = 0;
2295 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2296 			error = EINVAL;
2297 			break;
2298 		}
2299 		pr->nr = 0;
2300 		if (ruleset->anchor == NULL) {
2301 			/* XXX kludge for pf_main_ruleset */
2302 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2303 				if (anchor->parent == NULL)
2304 					pr->nr++;
2305 		} else {
2306 			RB_FOREACH(anchor, pf_anchor_node,
2307 			    &ruleset->anchor->children)
2308 				pr->nr++;
2309 		}
2310 		break;
2311 	}
2312 
2313 	case DIOCGETRULESET: {
2314 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2315 		struct pf_ruleset	*ruleset;
2316 		struct pf_anchor	*anchor;
2317 		u_int32_t		 nr = 0;
2318 
2319 		pr->path[sizeof(pr->path) - 1] = 0;
2320 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2321 			error = EINVAL;
2322 			break;
2323 		}
2324 		pr->name[0] = 0;
2325 		if (ruleset->anchor == NULL) {
2326 			/* XXX kludge for pf_main_ruleset */
2327 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2328 				if (anchor->parent == NULL && nr++ == pr->nr) {
2329 					strlcpy(pr->name, anchor->name,
2330 					    sizeof(pr->name));
2331 					break;
2332 				}
2333 		} else {
2334 			RB_FOREACH(anchor, pf_anchor_node,
2335 			    &ruleset->anchor->children)
2336 				if (nr++ == pr->nr) {
2337 					strlcpy(pr->name, anchor->name,
2338 					    sizeof(pr->name));
2339 					break;
2340 				}
2341 		}
2342 		if (!pr->name[0])
2343 			error = EBUSY;
2344 		break;
2345 	}
2346 
2347 	case DIOCRCLRTABLES: {
2348 		struct pfioc_table *io = (struct pfioc_table *)addr;
2349 
2350 		if (io->pfrio_esize != 0) {
2351 			error = ENODEV;
2352 			break;
2353 		}
2354 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2355 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2356 		break;
2357 	}
2358 
2359 	case DIOCRADDTABLES: {
2360 		struct pfioc_table *io = (struct pfioc_table *)addr;
2361 
2362 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2363 			error = ENODEV;
2364 			break;
2365 		}
2366 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2367 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2368 		break;
2369 	}
2370 
2371 	case DIOCRDELTABLES: {
2372 		struct pfioc_table *io = (struct pfioc_table *)addr;
2373 
2374 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2375 			error = ENODEV;
2376 			break;
2377 		}
2378 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2379 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2380 		break;
2381 	}
2382 
2383 	case DIOCRGETTABLES: {
2384 		struct pfioc_table *io = (struct pfioc_table *)addr;
2385 
2386 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2387 			error = ENODEV;
2388 			break;
2389 		}
2390 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2391 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2392 		break;
2393 	}
2394 
2395 	case DIOCRGETTSTATS: {
2396 		struct pfioc_table *io = (struct pfioc_table *)addr;
2397 
2398 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2399 			error = ENODEV;
2400 			break;
2401 		}
2402 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2403 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2404 		break;
2405 	}
2406 
2407 	case DIOCRCLRTSTATS: {
2408 		struct pfioc_table *io = (struct pfioc_table *)addr;
2409 
2410 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2411 			error = ENODEV;
2412 			break;
2413 		}
2414 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2415 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2416 		break;
2417 	}
2418 
2419 	case DIOCRSETTFLAGS: {
2420 		struct pfioc_table *io = (struct pfioc_table *)addr;
2421 
2422 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2423 			error = ENODEV;
2424 			break;
2425 		}
2426 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2427 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2428 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2429 		break;
2430 	}
2431 
2432 	case DIOCRCLRADDRS: {
2433 		struct pfioc_table *io = (struct pfioc_table *)addr;
2434 
2435 		if (io->pfrio_esize != 0) {
2436 			error = ENODEV;
2437 			break;
2438 		}
2439 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2440 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2441 		break;
2442 	}
2443 
2444 	case DIOCRADDADDRS: {
2445 		struct pfioc_table *io = (struct pfioc_table *)addr;
2446 
2447 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2448 			error = ENODEV;
2449 			break;
2450 		}
2451 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2452 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2453 		    PFR_FLAG_USERIOCTL);
2454 		break;
2455 	}
2456 
2457 	case DIOCRDELADDRS: {
2458 		struct pfioc_table *io = (struct pfioc_table *)addr;
2459 
2460 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2461 			error = ENODEV;
2462 			break;
2463 		}
2464 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2465 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2466 		    PFR_FLAG_USERIOCTL);
2467 		break;
2468 	}
2469 
2470 	case DIOCRSETADDRS: {
2471 		struct pfioc_table *io = (struct pfioc_table *)addr;
2472 
2473 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2474 			error = ENODEV;
2475 			break;
2476 		}
2477 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2478 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2479 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2480 		    PFR_FLAG_USERIOCTL, 0);
2481 		break;
2482 	}
2483 
2484 	case DIOCRGETADDRS: {
2485 		struct pfioc_table *io = (struct pfioc_table *)addr;
2486 
2487 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2488 			error = ENODEV;
2489 			break;
2490 		}
2491 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2492 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2493 		break;
2494 	}
2495 
2496 	case DIOCRGETASTATS: {
2497 		struct pfioc_table *io = (struct pfioc_table *)addr;
2498 
2499 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2500 			error = ENODEV;
2501 			break;
2502 		}
2503 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2504 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2505 		break;
2506 	}
2507 
2508 	case DIOCRCLRASTATS: {
2509 		struct pfioc_table *io = (struct pfioc_table *)addr;
2510 
2511 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2512 			error = ENODEV;
2513 			break;
2514 		}
2515 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2516 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2517 		    PFR_FLAG_USERIOCTL);
2518 		break;
2519 	}
2520 
2521 	case DIOCRTSTADDRS: {
2522 		struct pfioc_table *io = (struct pfioc_table *)addr;
2523 
2524 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2525 			error = ENODEV;
2526 			break;
2527 		}
2528 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2529 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2530 		    PFR_FLAG_USERIOCTL);
2531 		break;
2532 	}
2533 
2534 	case DIOCRINADEFINE: {
2535 		struct pfioc_table *io = (struct pfioc_table *)addr;
2536 
2537 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2538 			error = ENODEV;
2539 			break;
2540 		}
2541 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2542 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2543 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2544 		break;
2545 	}
2546 
2547 	case DIOCOSFPADD: {
2548 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2549 		error = pf_osfp_add(io);
2550 		break;
2551 	}
2552 
2553 	case DIOCOSFPGET: {
2554 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2555 		error = pf_osfp_get(io);
2556 		break;
2557 	}
2558 
2559 	case DIOCXBEGIN: {
2560 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2561 		struct pfioc_trans_e	*ioe;
2562 		struct pfr_table	*table;
2563 		int			 i;
2564 
2565 		if (io->esize != sizeof(*ioe)) {
2566 			error = ENODEV;
2567 			goto fail;
2568 		}
2569 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2570 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2571 		for (i = 0; i < io->size; i++) {
2572 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2573 				kfree(table, M_TEMP);
2574 				kfree(ioe, M_TEMP);
2575 				error = EFAULT;
2576 				goto fail;
2577 			}
2578 			switch (ioe->rs_num) {
2579 #ifdef ALTQ
2580 			case PF_RULESET_ALTQ:
2581 				if (ioe->anchor[0]) {
2582 					kfree(table, M_TEMP);
2583 					kfree(ioe, M_TEMP);
2584 					error = EINVAL;
2585 					goto fail;
2586 				}
2587 				if ((error = pf_begin_altq(&ioe->ticket))) {
2588 					kfree(table, M_TEMP);
2589 					kfree(ioe, M_TEMP);
2590 					goto fail;
2591 				}
2592 				break;
2593 #endif /* ALTQ */
2594 			case PF_RULESET_TABLE:
2595 				bzero(table, sizeof(*table));
2596 				strlcpy(table->pfrt_anchor, ioe->anchor,
2597 				    sizeof(table->pfrt_anchor));
2598 				if ((error = pfr_ina_begin(table,
2599 				    &ioe->ticket, NULL, 0))) {
2600 					kfree(table, M_TEMP);
2601 					kfree(ioe, M_TEMP);
2602 					goto fail;
2603 				}
2604 				break;
2605 			default:
2606 				if ((error = pf_begin_rules(&ioe->ticket,
2607 				    ioe->rs_num, ioe->anchor))) {
2608 					kfree(table, M_TEMP);
2609 					kfree(ioe, M_TEMP);
2610 					goto fail;
2611 				}
2612 				break;
2613 			}
2614 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2615 				kfree(table, M_TEMP);
2616 				kfree(ioe, M_TEMP);
2617 				error = EFAULT;
2618 				goto fail;
2619 			}
2620 		}
2621 		kfree(table, M_TEMP);
2622 		kfree(ioe, M_TEMP);
2623 		break;
2624 	}
2625 
2626 	case DIOCXROLLBACK: {
2627 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2628 		struct pfioc_trans_e	*ioe;
2629 		struct pfr_table	*table;
2630 		int			 i;
2631 
2632 		if (io->esize != sizeof(*ioe)) {
2633 			error = ENODEV;
2634 			goto fail;
2635 		}
2636 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2637 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2638 		for (i = 0; i < io->size; i++) {
2639 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2640 				kfree(table, M_TEMP);
2641 				kfree(ioe, M_TEMP);
2642 				error = EFAULT;
2643 				goto fail;
2644 			}
2645 			switch (ioe->rs_num) {
2646 #ifdef ALTQ
2647 			case PF_RULESET_ALTQ:
2648 				if (ioe->anchor[0]) {
2649 					kfree(table, M_TEMP);
2650 					kfree(ioe, M_TEMP);
2651 					error = EINVAL;
2652 					goto fail;
2653 				}
2654 				if ((error = pf_rollback_altq(ioe->ticket))) {
2655 					kfree(table, M_TEMP);
2656 					kfree(ioe, M_TEMP);
2657 					goto fail; /* really bad */
2658 				}
2659 				break;
2660 #endif /* ALTQ */
2661 			case PF_RULESET_TABLE:
2662 				bzero(table, sizeof(*table));
2663 				strlcpy(table->pfrt_anchor, ioe->anchor,
2664 				    sizeof(table->pfrt_anchor));
2665 				if ((error = pfr_ina_rollback(table,
2666 				    ioe->ticket, NULL, 0))) {
2667 					kfree(table, M_TEMP);
2668 					kfree(ioe, M_TEMP);
2669 					goto fail; /* really bad */
2670 				}
2671 				break;
2672 			default:
2673 				if ((error = pf_rollback_rules(ioe->ticket,
2674 				    ioe->rs_num, ioe->anchor))) {
2675 					kfree(table, M_TEMP);
2676 					kfree(ioe, M_TEMP);
2677 					goto fail; /* really bad */
2678 				}
2679 				break;
2680 			}
2681 		}
2682 		kfree(table, M_TEMP);
2683 		kfree(ioe, M_TEMP);
2684 		break;
2685 	}
2686 
2687 	case DIOCXCOMMIT: {
2688 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2689 		struct pfioc_trans_e	*ioe;
2690 		struct pfr_table	*table;
2691 		struct pf_ruleset	*rs;
2692 		int			 i;
2693 
2694 		if (io->esize != sizeof(*ioe)) {
2695 			error = ENODEV;
2696 			goto fail;
2697 		}
2698 		ioe = kmalloc(sizeof(*ioe), M_TEMP, M_WAITOK);
2699 		table = kmalloc(sizeof(*table), M_TEMP, M_WAITOK);
2700 		/* first makes sure everything will succeed */
2701 		for (i = 0; i < io->size; i++) {
2702 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2703 				kfree(table, M_TEMP);
2704 				kfree(ioe, M_TEMP);
2705 				error = EFAULT;
2706 				goto fail;
2707 			}
2708 			switch (ioe->rs_num) {
2709 #ifdef ALTQ
2710 			case PF_RULESET_ALTQ:
2711 				if (ioe->anchor[0]) {
2712 					kfree(table, M_TEMP);
2713 					kfree(ioe, M_TEMP);
2714 					error = EINVAL;
2715 					goto fail;
2716 				}
2717 				if (!altqs_inactive_open || ioe->ticket !=
2718 				    ticket_altqs_inactive) {
2719 					kfree(table, M_TEMP);
2720 					kfree(ioe, M_TEMP);
2721 					error = EBUSY;
2722 					goto fail;
2723 				}
2724 				break;
2725 #endif /* ALTQ */
2726 			case PF_RULESET_TABLE:
2727 				rs = pf_find_ruleset(ioe->anchor);
2728 				if (rs == NULL || !rs->topen || ioe->ticket !=
2729 				     rs->tticket) {
2730 					kfree(table, M_TEMP);
2731 					kfree(ioe, M_TEMP);
2732 					error = EBUSY;
2733 					goto fail;
2734 				}
2735 				break;
2736 			default:
2737 				if (ioe->rs_num < 0 || ioe->rs_num >=
2738 				    PF_RULESET_MAX) {
2739 					kfree(table, M_TEMP);
2740 					kfree(ioe, M_TEMP);
2741 					error = EINVAL;
2742 					goto fail;
2743 				}
2744 				rs = pf_find_ruleset(ioe->anchor);
2745 				if (rs == NULL ||
2746 				    !rs->rules[ioe->rs_num].inactive.open ||
2747 				    rs->rules[ioe->rs_num].inactive.ticket !=
2748 				    ioe->ticket) {
2749 					kfree(table, M_TEMP);
2750 					kfree(ioe, M_TEMP);
2751 					error = EBUSY;
2752 					goto fail;
2753 				}
2754 				break;
2755 			}
2756 		}
2757 		/* now do the commit - no errors should happen here */
2758 		for (i = 0; i < io->size; i++) {
2759 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2760 				kfree(table, M_TEMP);
2761 				kfree(ioe, M_TEMP);
2762 				error = EFAULT;
2763 				goto fail;
2764 			}
2765 			switch (ioe->rs_num) {
2766 #ifdef ALTQ
2767 			case PF_RULESET_ALTQ:
2768 				if ((error = pf_commit_altq(ioe->ticket))) {
2769 					kfree(table, M_TEMP);
2770 					kfree(ioe, M_TEMP);
2771 					goto fail; /* really bad */
2772 				}
2773 				break;
2774 #endif /* ALTQ */
2775 			case PF_RULESET_TABLE:
2776 				bzero(table, sizeof(*table));
2777 				strlcpy(table->pfrt_anchor, ioe->anchor,
2778 				    sizeof(table->pfrt_anchor));
2779 				if ((error = pfr_ina_commit(table, ioe->ticket,
2780 				    NULL, NULL, 0))) {
2781 					kfree(table, M_TEMP);
2782 					kfree(ioe, M_TEMP);
2783 					goto fail; /* really bad */
2784 				}
2785 				break;
2786 			default:
2787 				if ((error = pf_commit_rules(ioe->ticket,
2788 				    ioe->rs_num, ioe->anchor))) {
2789 					kfree(table, M_TEMP);
2790 					kfree(ioe, M_TEMP);
2791 					goto fail; /* really bad */
2792 				}
2793 				break;
2794 			}
2795 		}
2796 		kfree(table, M_TEMP);
2797 		kfree(ioe, M_TEMP);
2798 		break;
2799 	}
2800 
2801 	case DIOCGETSRCNODES: {
2802 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
2803 		struct pf_src_node	*n, *p, *pstore;
2804 		u_int32_t nr = 0;
2805 		int	space = psn->psn_len;
2806 		int	nn;
2807 
2808 		if (space == 0) {
2809 			for (nn = 0; nn < ncpus; ++nn) {
2810 				RB_FOREACH(n, pf_src_tree,
2811 					   &tree_src_tracking[nn]) {
2812 					nr++;
2813 				}
2814 			}
2815 			psn->psn_len = sizeof(struct pf_src_node) * nr;
2816 			break;
2817 		}
2818 
2819 		pstore = kmalloc(sizeof(*pstore), M_TEMP, M_WAITOK);
2820 
2821 		p = psn->psn_src_nodes;
2822 
2823 		/*
2824 		 * WARNING: We are not switching cpus so we cannot call
2825 		 *	    nominal pf.c support routines for cpu-specific
2826 		 *	    data.
2827 		 */
2828 		for (nn = 0; nn < ncpus; ++nn) {
2829 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
2830 				int	secs = time_second, diff;
2831 
2832 				if ((nr + 1) * sizeof(*p) >
2833 				    (unsigned)psn->psn_len) {
2834 					break;
2835 				}
2836 
2837 				bcopy(n, pstore, sizeof(*pstore));
2838 				if (n->rule.ptr != NULL)
2839 					pstore->rule.nr = n->rule.ptr->nr;
2840 				pstore->creation = secs - pstore->creation;
2841 				if (pstore->expire > secs)
2842 					pstore->expire -= secs;
2843 				else
2844 					pstore->expire = 0;
2845 
2846 				/* adjust the connection rate estimate */
2847 				diff = secs - n->conn_rate.last;
2848 				if (diff >= n->conn_rate.seconds)
2849 					pstore->conn_rate.count = 0;
2850 				else
2851 					pstore->conn_rate.count -=
2852 					    n->conn_rate.count * diff /
2853 					    n->conn_rate.seconds;
2854 
2855 				error = copyout(pstore, p, sizeof(*p));
2856 				if (error) {
2857 					kfree(pstore, M_TEMP);
2858 					goto fail;
2859 				}
2860 				p++;
2861 				nr++;
2862 			}
2863 		}
2864 		psn->psn_len = sizeof(struct pf_src_node) * nr;
2865 		kfree(pstore, M_TEMP);
2866 		break;
2867 	}
2868 
2869 	case DIOCCLRSRCNODES: {
2870 		struct pf_src_node	*n;
2871 		struct pf_state		*state;
2872 		globaldata_t save_gd = mycpu;
2873 		int nn;
2874 
2875 		/*
2876 		 * WARNING: We are not switching cpus so we cannot call
2877 		 *	    nominal pf.c support routines for cpu-specific
2878 		 *	    data.
2879 		 */
2880 		for (nn = 0; nn < ncpus; ++nn) {
2881 			RB_FOREACH(state, pf_state_tree_id, &tree_id[nn]) {
2882 				state->src_node = NULL;
2883 				state->nat_src_node = NULL;
2884 			}
2885 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
2886 				n->expire = 1;
2887 				n->states = 0;
2888 			}
2889 		}
2890 
2891 		/*
2892 		 * WARNING: Must move to the target cpu for nominal calls
2893 		 *	    into pf.c
2894 		 */
2895 		for (nn = 0; nn < ncpus; ++nn) {
2896 			lwkt_setcpu_self(globaldata_find(nn));
2897 			pf_purge_expired_src_nodes(1);
2898 		}
2899 		lwkt_setcpu_self(save_gd);
2900 		pf_status.src_nodes = 0;
2901 		break;
2902 	}
2903 
2904 	case DIOCKILLSRCNODES: {
2905 		struct pf_src_node	*sn;
2906 		struct pf_state		*s;
2907 		struct pfioc_src_node_kill *psnk =
2908 		    (struct pfioc_src_node_kill *)addr;
2909 		u_int			killed = 0;
2910 		globaldata_t save_gd = mycpu;
2911 		int nn;
2912 
2913 		/*
2914 		 * WARNING: We are not switching cpus so we cannot call
2915 		 *	    nominal pf.c support routines for cpu-specific
2916 		 *	    data.
2917 		 */
2918 		for (nn = 0; nn < ncpus; ++nn) {
2919 		    RB_FOREACH(sn, pf_src_tree, &tree_src_tracking[nn]) {
2920 			if (PF_MATCHA(psnk->psnk_src.neg,
2921 				&psnk->psnk_src.addr.v.a.addr,
2922 				&psnk->psnk_src.addr.v.a.mask,
2923 				&sn->addr, sn->af) &&
2924 			    PF_MATCHA(psnk->psnk_dst.neg,
2925 				&psnk->psnk_dst.addr.v.a.addr,
2926 				&psnk->psnk_dst.addr.v.a.mask,
2927 				&sn->raddr, sn->af)) {
2928 				/* Handle state to src_node linkage */
2929 				if (sn->states != 0) {
2930 					RB_FOREACH(s, pf_state_tree_id,
2931 					    &tree_id[nn]) {
2932 						if (s->src_node == sn)
2933 							s->src_node = NULL;
2934 						if (s->nat_src_node == sn)
2935 							s->nat_src_node = NULL;
2936 					}
2937 					sn->states = 0;
2938 				}
2939 				sn->expire = 1;
2940 				killed++;
2941 			}
2942 		    }
2943 		}
2944 		if (killed > 0) {
2945 			for (nn = 0; nn < ncpus; ++nn) {
2946 				lwkt_setcpu_self(globaldata_find(nn));
2947 				pf_purge_expired_src_nodes(1);
2948 			}
2949 			lwkt_setcpu_self(save_gd);
2950 		}
2951 
2952 		psnk->psnk_killed = killed;
2953 		break;
2954 	}
2955 
2956 	case DIOCSETHOSTID: {
2957 		u_int32_t	*hostid = (u_int32_t *)addr;
2958 
2959 		if (*hostid == 0)
2960 			pf_status.hostid = karc4random();
2961 		else
2962 			pf_status.hostid = *hostid;
2963 		break;
2964 	}
2965 
2966 	case DIOCOSFPFLUSH:
2967 		crit_enter();
2968 		pf_osfp_flush();
2969 		crit_exit();
2970 		break;
2971 
2972 	case DIOCIGETIFACES: {
2973 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2974 
2975 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
2976 			error = ENODEV;
2977 			break;
2978 		}
2979 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
2980 		    &io->pfiio_size);
2981 		break;
2982 	}
2983 
2984 	case DIOCSETIFFLAG: {
2985 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2986 
2987 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
2988 		break;
2989 	}
2990 
2991 	case DIOCCLRIFFLAG: {
2992 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
2993 
2994 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
2995 		break;
2996 	}
2997 
2998 	default:
2999 		error = ENODEV;
3000 		break;
3001 	}
3002 fail:
3003 	lwkt_reltoken(&pf_token);
3004 	return (error);
3005 }
3006 
3007 /*
3008  * XXX - Check for version missmatch!!!
3009  */
3010 static void
3011 pf_clear_states(void)
3012 {
3013 	struct pf_state		*s, *nexts;
3014 	u_int			killed = 0;
3015 	globaldata_t save_gd = mycpu;
3016 	int nn;
3017 
3018 	for (nn = 0; nn < ncpus; ++nn) {
3019 		lwkt_setcpu_self(globaldata_find(nn));
3020 		for (s = RB_MIN(pf_state_tree_id, &tree_id[nn]); s; s = nexts) {
3021 			nexts = RB_NEXT(pf_state_tree_id, &tree_id[nn], s);
3022 
3023 			/* don't send out individual delete messages */
3024 			s->sync_flags = PFSTATE_NOSYNC;
3025 			pf_unlink_state(s);
3026 			killed++;
3027 		}
3028 
3029 	}
3030 	lwkt_setcpu_self(save_gd);
3031 
3032 #if 0 /* PFSYNC */
3033 /*
3034  * XXX This is called on module unload, we do not want to sync that over? */
3035  */
3036 	pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
3037 #endif
3038 }
3039 
3040 static int
3041 pf_clear_tables(void)
3042 {
3043 	struct pfioc_table io;
3044 	int error;
3045 
3046 	bzero(&io, sizeof(io));
3047 
3048 	error = pfr_clr_tables(&io.pfrio_table, &io.pfrio_ndel,
3049 	    io.pfrio_flags);
3050 
3051 	return (error);
3052 }
3053 
3054 static void
3055 pf_clear_srcnodes(void)
3056 {
3057 	struct pf_src_node	*n;
3058 	struct pf_state		*state;
3059 	globaldata_t save_gd = mycpu;
3060 	int nn;
3061 
3062 	for (nn = 0; nn < ncpus; ++nn) {
3063 		lwkt_setcpu_self(globaldata_find(nn));
3064 		RB_FOREACH(state, pf_state_tree_id, &tree_id[nn]) {
3065 			state->src_node = NULL;
3066 			state->nat_src_node = NULL;
3067 		}
3068 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking[nn]) {
3069 			n->expire = 1;
3070 			n->states = 0;
3071 		}
3072 		pf_purge_expired_src_nodes(0);
3073 	}
3074 	lwkt_setcpu_self(save_gd);
3075 
3076 	pf_status.src_nodes = 0;
3077 }
3078 
3079 /*
3080  * XXX - Check for version missmatch!!!
3081  */
3082 
3083 /*
3084  * Duplicate pfctl -Fa operation to get rid of as much as we can.
3085  */
3086 static int
3087 shutdown_pf(void)
3088 {
3089 	int error = 0;
3090 	u_int32_t t[5];
3091 	char nn = '\0';
3092 
3093 
3094 	pf_status.running = 0;
3095 	error = dehook_pf();
3096 	if (error) {
3097 		pf_status.running = 1;
3098 		DPFPRINTF(PF_DEBUG_MISC,
3099 		    ("pf: pfil unregistration failed\n"));
3100 		return(error);
3101 	}
3102 	do {
3103 		if ((error = pf_begin_rules(&t[0], PF_RULESET_SCRUB, &nn)) != 0) {
3104 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: SCRUB\n"));
3105 			break;
3106 		}
3107 		if ((error = pf_begin_rules(&t[1], PF_RULESET_FILTER, &nn)) != 0) {
3108 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: FILTER\n"));
3109 			break;		/* XXX: rollback? */
3110 		}
3111 		if ((error = pf_begin_rules(&t[2], PF_RULESET_NAT, &nn))    != 0) {
3112 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: NAT\n"));
3113 			break;		/* XXX: rollback? */
3114 		}
3115 		if ((error = pf_begin_rules(&t[3], PF_RULESET_BINAT, &nn))
3116 		    != 0) {
3117 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: BINAT\n"));
3118 			break;		/* XXX: rollback? */
3119 		}
3120 		if ((error = pf_begin_rules(&t[4], PF_RULESET_RDR, &nn))
3121 		    != 0) {
3122 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: RDR\n"));
3123 			break;		/* XXX: rollback? */
3124 		}
3125 
3126 		/* XXX: these should always succeed here */
3127 		pf_commit_rules(t[0], PF_RULESET_SCRUB, &nn);
3128 		pf_commit_rules(t[1], PF_RULESET_FILTER, &nn);
3129 		pf_commit_rules(t[2], PF_RULESET_NAT, &nn);
3130 		pf_commit_rules(t[3], PF_RULESET_BINAT, &nn);
3131 		pf_commit_rules(t[4], PF_RULESET_RDR, &nn);
3132 
3133 		if ((error = pf_clear_tables()) != 0)
3134 			break;
3135 #ifdef ALTQ
3136 		if ((error = pf_begin_altq(&t[0])) != 0) {
3137 			DPFPRINTF(PF_DEBUG_MISC, ("shutdown_pf: ALTQ\n"));
3138 			break;
3139 		}
3140 		pf_commit_altq(t[0]);
3141 #endif
3142 		pf_clear_states();
3143 		pf_clear_srcnodes();
3144 
3145 		/* status does not use malloced mem so no need to cleanup */
3146 		/* fingerprints and interfaces have their own cleanup code */
3147 	} while(0);
3148         return (error);
3149 }
3150 
3151 static int
3152 pf_check_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3153 {
3154 	/*
3155 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3156 	 * for ip_len/ip_off. This is why we don't have to change byte order
3157 	 * like the FreeBSD-5 version does.
3158 	 */
3159 	int chk;
3160 
3161 	lwkt_gettoken_shared(&pf_token);
3162 
3163 	chk = pf_test(PF_IN, ifp, m, NULL, NULL);
3164 	if (chk && *m) {
3165 		m_freem(*m);
3166 		*m = NULL;
3167 	}
3168 	lwkt_reltoken(&pf_token);
3169 	return chk;
3170 }
3171 
3172 static int
3173 pf_check_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3174 {
3175 	/*
3176 	 * DragonFly's version of pf uses FreeBSD's native host byte ordering
3177 	 * for ip_len/ip_off. This is why we don't have to change byte order
3178 	 * like the FreeBSD-5 version does.
3179 	 */
3180 	int chk;
3181 
3182 	lwkt_gettoken_shared(&pf_token);
3183 
3184 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3185 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3186 		in_delayed_cksum(*m);
3187 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3188 	}
3189 	chk = pf_test(PF_OUT, ifp, m, NULL, NULL);
3190 	if (chk && *m) {
3191 		m_freem(*m);
3192 		*m = NULL;
3193 	}
3194 	lwkt_reltoken(&pf_token);
3195 	return chk;
3196 }
3197 
3198 #ifdef INET6
3199 static int
3200 pf_check6_in(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3201 {
3202 	/*
3203 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3204 	 */
3205 	int chk;
3206 
3207 	lwkt_gettoken_shared(&pf_token);
3208 
3209 	chk = pf_test6(PF_IN, ifp, m, NULL, NULL);
3210 	if (chk && *m) {
3211 		m_freem(*m);
3212 		*m = NULL;
3213 	}
3214 	lwkt_reltoken(&pf_token);
3215 	return chk;
3216 }
3217 
3218 static int
3219 pf_check6_out(void *arg, struct mbuf **m, struct ifnet *ifp, int dir)
3220 {
3221 	/*
3222 	 * IPv6 is not affected by ip_len/ip_off byte order changes.
3223 	 */
3224 	int chk;
3225 
3226 	lwkt_gettoken_shared(&pf_token);
3227 
3228 	/* We need a proper CSUM befor we start (s. OpenBSD ip_output) */
3229 	if ((*m)->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
3230 		in_delayed_cksum(*m);
3231 		(*m)->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
3232 	}
3233 	chk = pf_test6(PF_OUT, ifp, m, NULL, NULL);
3234 	if (chk && *m) {
3235 		m_freem(*m);
3236 		*m = NULL;
3237 	}
3238 	lwkt_reltoken(&pf_token);
3239 	return chk;
3240 }
3241 #endif /* INET6 */
3242 
3243 static int
3244 hook_pf(void)
3245 {
3246 	struct pfil_head *pfh_inet;
3247 #ifdef INET6
3248 	struct pfil_head *pfh_inet6;
3249 #endif
3250 
3251 	lwkt_gettoken(&pf_token);
3252 
3253 	if (pf_pfil_hooked) {
3254 		lwkt_reltoken(&pf_token);
3255 		return (0);
3256 	}
3257 
3258 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3259 	if (pfh_inet == NULL) {
3260 		lwkt_reltoken(&pf_token);
3261 		return (ENODEV);
3262 	}
3263 	pfil_add_hook(pf_check_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet);
3264 	pfil_add_hook(pf_check_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet);
3265 #ifdef INET6
3266 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3267 	if (pfh_inet6 == NULL) {
3268 		pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3269 		pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3270 		lwkt_reltoken(&pf_token);
3271 		return (ENODEV);
3272 	}
3273 	pfil_add_hook(pf_check6_in, NULL, PFIL_IN | PFIL_MPSAFE, pfh_inet6);
3274 	pfil_add_hook(pf_check6_out, NULL, PFIL_OUT | PFIL_MPSAFE, pfh_inet6);
3275 #endif
3276 
3277 	pf_pfil_hooked = 1;
3278 	lwkt_reltoken(&pf_token);
3279 	return (0);
3280 }
3281 
3282 static int
3283 dehook_pf(void)
3284 {
3285 	struct pfil_head *pfh_inet;
3286 #ifdef INET6
3287 	struct pfil_head *pfh_inet6;
3288 #endif
3289 
3290 	lwkt_gettoken(&pf_token);
3291 
3292 	if (pf_pfil_hooked == 0) {
3293 		lwkt_reltoken(&pf_token);
3294 		return (0);
3295 	}
3296 
3297 	pfh_inet = pfil_head_get(PFIL_TYPE_AF, AF_INET);
3298 	if (pfh_inet == NULL) {
3299 		lwkt_reltoken(&pf_token);
3300 		return (ENODEV);
3301 	}
3302 	pfil_remove_hook(pf_check_in, NULL, PFIL_IN, pfh_inet);
3303 	pfil_remove_hook(pf_check_out, NULL, PFIL_OUT, pfh_inet);
3304 #ifdef INET6
3305 	pfh_inet6 = pfil_head_get(PFIL_TYPE_AF, AF_INET6);
3306 	if (pfh_inet6 == NULL) {
3307 		lwkt_reltoken(&pf_token);
3308 		return (ENODEV);
3309 	}
3310 	pfil_remove_hook(pf_check6_in, NULL, PFIL_IN, pfh_inet6);
3311 	pfil_remove_hook(pf_check6_out, NULL, PFIL_OUT, pfh_inet6);
3312 #endif
3313 
3314 	pf_pfil_hooked = 0;
3315 	lwkt_reltoken(&pf_token);
3316 	return (0);
3317 }
3318 
3319 static int
3320 pf_load(void)
3321 {
3322 	lwkt_gettoken(&pf_token);
3323 
3324 	pf_dev = make_dev(&pf_ops, 0, 0, 0, 0600, PF_NAME);
3325 	pfattach();
3326 	lockinit(&pf_consistency_lock, "pfconslck", 0, LK_CANRECURSE);
3327 	lockinit(&pf_global_statetbl_lock, "pfglstlk", 0, 0);
3328 	lwkt_reltoken(&pf_token);
3329 	return (0);
3330 }
3331 
3332 static int
3333 pf_mask_del(struct radix_node *rn, void *arg)
3334 {
3335 	struct radix_node_head *rnh = arg;
3336 
3337 	rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
3338 	Free(rn);
3339 	return 0;
3340 }
3341 
3342 static int
3343 pf_unload(void)
3344 {
3345 	int error;
3346 	pf_status.running = 0;
3347 
3348 	lwkt_gettoken(&pf_token);
3349 
3350 	error = dehook_pf();
3351 	if (error) {
3352 		/*
3353 		 * Should not happen!
3354 		 * XXX Due to error code ESRCH, kldunload will show
3355 		 * a message like 'No such process'.
3356 		 */
3357 		kprintf("pfil unregistration fail\n");
3358 		lwkt_reltoken(&pf_token);
3359 		return error;
3360 	}
3361 	shutdown_pf();
3362 	pf_end_threads = 1;
3363 	while (pf_end_threads < 2) {
3364 		wakeup_one(pf_purge_thread);
3365 		tsleep(pf_purge_thread, 0, "pftmo", hz);
3366 	}
3367 	pfi_cleanup();
3368 	pf_osfp_flush();
3369 	dev_ops_remove_all(&pf_ops);
3370 	lockuninit(&pf_consistency_lock);
3371 	lwkt_reltoken(&pf_token);
3372 
3373 	if (pf_maskhead != NULL) {
3374 		pf_maskhead->rnh_walktree(pf_maskhead,
3375 			pf_mask_del, pf_maskhead);
3376 		Free(pf_maskhead);
3377 		pf_maskhead = NULL;
3378 	}
3379 	kmalloc_destroy(&pf_state_pl);
3380 	kmalloc_destroy(&pf_frent_pl);
3381 	kmalloc_destroy(&pf_cent_pl);
3382 	return 0;
3383 }
3384 
3385 static int
3386 pf_modevent(module_t mod, int type, void *data)
3387 {
3388 	int error = 0;
3389 
3390 	lwkt_gettoken(&pf_token);
3391 
3392 	switch(type) {
3393 	case MOD_LOAD:
3394 		error = pf_load();
3395 		break;
3396 
3397 	case MOD_UNLOAD:
3398 		error = pf_unload();
3399 		break;
3400 	default:
3401 		error = EINVAL;
3402 		break;
3403 	}
3404 	lwkt_reltoken(&pf_token);
3405 	return error;
3406 }
3407 
3408 static moduledata_t pf_mod = {
3409 	"pf",
3410 	pf_modevent,
3411 	0
3412 };
3413 DECLARE_MODULE(pf, pf_mod, SI_SUB_PSEUDO, SI_ORDER_FIRST);
3414 MODULE_VERSION(pf, PF_MODVER);
3415