xref: /netbsd-src/sys/dist/pf/net/pf_ioctl.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: pf_ioctl.c,v 1.49 2014/03/16 05:20:29 dholland Exp $	*/
2 /*	$OpenBSD: pf_ioctl.c,v 1.182 2007/06/24 11:17:13 mcbride Exp $ */
3 
4 /*
5  * Copyright (c) 2001 Daniel Hartmeier
6  * Copyright (c) 2002,2003 Henning Brauer
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  *
13  *    - Redistributions of source code must retain the above copyright
14  *      notice, this list of conditions and the following disclaimer.
15  *    - Redistributions in binary form must reproduce the above
16  *      copyright notice, this list of conditions and the following
17  *      disclaimer in the documentation and/or other materials provided
18  *      with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
28  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
30  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31  * POSSIBILITY OF SUCH DAMAGE.
32  *
33  * Effort sponsored in part by the Defense Advanced Research Projects
34  * Agency (DARPA) and Air Force Research Laboratory, Air Force
35  * Materiel Command, USAF, under agreement number F30602-01-2-0537.
36  *
37  */
38 
39 #include <sys/cdefs.h>
40 __KERNEL_RCSID(0, "$NetBSD: pf_ioctl.c,v 1.49 2014/03/16 05:20:29 dholland Exp $");
41 
42 #ifdef _KERNEL_OPT
43 #include "opt_inet.h"
44 #endif
45 
46 #include "pfsync.h"
47 
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/mbuf.h>
51 #include <sys/filio.h>
52 #include <sys/fcntl.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/kernel.h>
56 #include <sys/time.h>
57 #include <sys/pool.h>
58 #include <sys/proc.h>
59 #include <sys/malloc.h>
60 #include <sys/kthread.h>
61 #include <sys/rwlock.h>
62 #include <uvm/uvm_extern.h>
63 #ifdef __NetBSD__
64 #include <sys/conf.h>
65 #include <sys/lwp.h>
66 #include <sys/kauth.h>
67 #include <sys/module.h>
68 #include <sys/cprng.h>
69 #endif /* __NetBSD__ */
70 
71 #include <net/if.h>
72 #include <net/if_types.h>
73 #include <net/route.h>
74 
75 #include <netinet/in.h>
76 #include <netinet/in_var.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/ip_var.h>
80 #include <netinet/ip_icmp.h>
81 
82 #ifndef __NetBSD__
83 #include <dev/rndvar.h>
84 #include <crypto/md5.h>
85 #else
86 #include <sys/md5.h>
87 #endif /* __NetBSD__ */
88 #include <net/pfvar.h>
89 
90 #if NPFSYNC > 0
91 #include <net/if_pfsync.h>
92 #endif /* NPFSYNC > 0 */
93 
94 #if NPFLOG > 0
95 #include <net/if_pflog.h>
96 #endif /* NPFLOG > 0 */
97 
98 #ifdef INET6
99 #include <netinet/ip6.h>
100 #include <netinet/in_pcb.h>
101 #endif /* INET6 */
102 
103 #ifdef ALTQ
104 #include <altq/altq.h>
105 #endif
106 
107 void			 pfattach(int);
108 #ifdef _MODULE
109 void			 pfdetach(void);
110 #endif /* _MODULE */
111 #ifndef __NetBSD__
112 void			 pf_thread_create(void *);
113 #endif /* !__NetBSD__ */
114 int			 pfopen(dev_t, int, int, struct lwp *);
115 int			 pfclose(dev_t, int, int, struct lwp *);
116 struct pf_pool		*pf_get_pool(char *, u_int32_t, u_int8_t, u_int32_t,
117 			    u_int8_t, u_int8_t, u_int8_t);
118 
119 void			 pf_mv_pool(struct pf_palist *, struct pf_palist *);
120 void			 pf_empty_pool(struct pf_palist *);
121 int			 pfioctl(dev_t, u_long, void *, int, struct lwp *);
122 #ifdef ALTQ
123 int			 pf_begin_altq(u_int32_t *);
124 int			 pf_rollback_altq(u_int32_t);
125 int			 pf_commit_altq(u_int32_t);
126 int			 pf_enable_altq(struct pf_altq *);
127 int			 pf_disable_altq(struct pf_altq *);
128 #endif /* ALTQ */
129 int			 pf_begin_rules(u_int32_t *, int, const char *);
130 int			 pf_rollback_rules(u_int32_t, int, char *);
131 int			 pf_setup_pfsync_matching(struct pf_ruleset *);
132 void			 pf_hash_rule(MD5_CTX *, struct pf_rule *);
133 void			 pf_hash_rule_addr(MD5_CTX *, struct pf_rule_addr *);
134 int			 pf_commit_rules(u_int32_t, int, char *);
135 void			 pf_state_export(struct pfsync_state *,
136 			    struct pf_state_key *, struct pf_state *);
137 void			 pf_state_import(struct pfsync_state *,
138 			    struct pf_state_key *, struct pf_state *);
139 
140 static	int		pf_state_add(struct pfsync_state*);
141 
142 struct pf_rule		 pf_default_rule;
143 #ifdef __NetBSD__
144 krwlock_t		 pf_consistency_lock;
145 #else
146 struct rwlock		 pf_consistency_lock = RWLOCK_INITIALIZER("pfcnslk");
147 #endif /* __NetBSD__ */
148 #ifdef ALTQ
149 static int		 pf_altq_running;
150 #endif
151 
152 int		pf_state_lock = 0;
153 
154 #define	TAGID_MAX	 50000
155 TAILQ_HEAD(pf_tags, pf_tagname)	pf_tags = TAILQ_HEAD_INITIALIZER(pf_tags),
156 				pf_qids = TAILQ_HEAD_INITIALIZER(pf_qids);
157 
158 #if (PF_QNAME_SIZE != PF_TAG_NAME_SIZE)
159 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
160 #endif
161 u_int16_t		 tagname2tag(struct pf_tags *, char *);
162 void			 tag2tagname(struct pf_tags *, u_int16_t, char *);
163 void			 tag_unref(struct pf_tags *, u_int16_t);
164 int			 pf_rtlabel_add(struct pf_addr_wrap *);
165 void			 pf_rtlabel_remove(struct pf_addr_wrap *);
166 void			 pf_rtlabel_copyout(struct pf_addr_wrap *);
167 
168 #define DPFPRINTF(n, x) if (pf_status.debug >= (n)) printf x
169 
170 #ifdef __NetBSD__
171 const struct cdevsw pf_cdevsw = {
172 	.d_open = pfopen,
173 	.d_close = pfclose,
174 	.d_read = noread,
175 	.d_write = nowrite,
176 	.d_ioctl = pfioctl,
177 	.d_stop = nostop,
178 	.d_tty = notty,
179 	.d_poll = nopoll,
180 	.d_mmap = nommap,
181 	.d_kqfilter = nokqfilter,
182 	.d_flag = D_OTHER
183 };
184 
185 static int pfil4_wrapper(void *, struct mbuf **, struct ifnet *, int);
186 #ifdef INET6
187 static int pfil6_wrapper(void *, struct mbuf **, struct ifnet *, int);
188 #endif /* INET6 */
189 
190 static int pf_pfil_attach(void);
191 static int pf_pfil_detach(void);
192 
193 static int pf_pfil_attached;
194 
195 static kauth_listener_t pf_listener;
196 #endif /* __NetBSD__ */
197 
198 #ifdef __NetBSD__
199 static int
200 pf_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
201     void *arg0, void *arg1, void *arg2, void *arg3)
202 {
203 	int result;
204 	enum kauth_network_req req;
205 
206 	result = KAUTH_RESULT_DEFER;
207 	req = (enum kauth_network_req)arg0;
208 
209 	if (action != KAUTH_NETWORK_FIREWALL)
210 		return result;
211 
212 	/* These must have came from device context. */
213 	if ((req == KAUTH_REQ_NETWORK_FIREWALL_FW) ||
214 	    (req == KAUTH_REQ_NETWORK_FIREWALL_NAT))
215 		result = KAUTH_RESULT_ALLOW;
216 
217 	return result;
218 }
219 #endif /* __NetBSD__ */
220 
221 void
222 pfattach(int num)
223 {
224 	u_int32_t *timeout = pf_default_rule.timeout;
225 
226 #ifdef __NetBSD__
227 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
228 	    &pool_allocator_nointr, IPL_NONE);
229 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
230 	    "pfsrctrpl", NULL, IPL_SOFTNET);
231 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
232 	    NULL, IPL_SOFTNET);
233 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0, 0, 0,
234 	    "pfstatekeypl", NULL, IPL_SOFTNET);
235 	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
236 	    &pool_allocator_nointr, IPL_NONE);
237 	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
238 	    "pfpooladdrpl", &pool_allocator_nointr, IPL_NONE);
239 #else
240 	pool_init(&pf_rule_pl, sizeof(struct pf_rule), 0, 0, 0, "pfrulepl",
241 	    &pool_allocator_nointr);
242 	pool_init(&pf_src_tree_pl, sizeof(struct pf_src_node), 0, 0, 0,
243 	    "pfsrctrpl", NULL);
244 	pool_init(&pf_state_pl, sizeof(struct pf_state), 0, 0, 0, "pfstatepl",
245 	    NULL);
246 	pool_init(&pf_state_key_pl, sizeof(struct pf_state_key), 0, 0, 0,
247 	    "pfstatekeypl", NULL);
248 	pool_init(&pf_altq_pl, sizeof(struct pf_altq), 0, 0, 0, "pfaltqpl",
249 	    &pool_allocator_nointr);
250 	pool_init(&pf_pooladdr_pl, sizeof(struct pf_pooladdr), 0, 0, 0,
251 	    "pfpooladdrpl", &pool_allocator_nointr);
252 #endif /* !__NetBSD__ */
253 
254 	pfr_initialize();
255 	pfi_initialize();
256 	pf_osfp_initialize();
257 
258 	pool_sethardlimit(pf_pool_limits[PF_LIMIT_STATES].pp,
259 	    pf_pool_limits[PF_LIMIT_STATES].limit, NULL, 0);
260 
261 	if (ctob(physmem) <= 100*1024*1024)
262 		pf_pool_limits[PF_LIMIT_TABLE_ENTRIES].limit =
263 		    PFR_KENTRY_HIWAT_SMALL;
264 
265 	RB_INIT(&tree_src_tracking);
266 	RB_INIT(&pf_anchors);
267 	pf_init_ruleset(&pf_main_ruleset);
268 	TAILQ_INIT(&pf_altqs[0]);
269 	TAILQ_INIT(&pf_altqs[1]);
270 	TAILQ_INIT(&pf_pabuf);
271 	pf_altqs_active = &pf_altqs[0];
272 	pf_altqs_inactive = &pf_altqs[1];
273 	TAILQ_INIT(&state_list);
274 
275 #ifdef __NetBSD__
276 	rw_init(&pf_consistency_lock);
277 #endif /* __NetBSD__ */
278 
279 	/* default rule should never be garbage collected */
280 	pf_default_rule.entries.tqe_prev = &pf_default_rule.entries.tqe_next;
281 	pf_default_rule.action = PF_PASS;
282 	pf_default_rule.nr = -1;
283 	pf_default_rule.rtableid = -1;
284 
285 	/* initialize default timeouts */
286 	timeout[PFTM_TCP_FIRST_PACKET] = PFTM_TCP_FIRST_PACKET_VAL;
287 	timeout[PFTM_TCP_OPENING] = PFTM_TCP_OPENING_VAL;
288 	timeout[PFTM_TCP_ESTABLISHED] = PFTM_TCP_ESTABLISHED_VAL;
289 	timeout[PFTM_TCP_CLOSING] = PFTM_TCP_CLOSING_VAL;
290 	timeout[PFTM_TCP_FIN_WAIT] = PFTM_TCP_FIN_WAIT_VAL;
291 	timeout[PFTM_TCP_CLOSED] = PFTM_TCP_CLOSED_VAL;
292 	timeout[PFTM_UDP_FIRST_PACKET] = PFTM_UDP_FIRST_PACKET_VAL;
293 	timeout[PFTM_UDP_SINGLE] = PFTM_UDP_SINGLE_VAL;
294 	timeout[PFTM_UDP_MULTIPLE] = PFTM_UDP_MULTIPLE_VAL;
295 	timeout[PFTM_ICMP_FIRST_PACKET] = PFTM_ICMP_FIRST_PACKET_VAL;
296 	timeout[PFTM_ICMP_ERROR_REPLY] = PFTM_ICMP_ERROR_REPLY_VAL;
297 	timeout[PFTM_OTHER_FIRST_PACKET] = PFTM_OTHER_FIRST_PACKET_VAL;
298 	timeout[PFTM_OTHER_SINGLE] = PFTM_OTHER_SINGLE_VAL;
299 	timeout[PFTM_OTHER_MULTIPLE] = PFTM_OTHER_MULTIPLE_VAL;
300 	timeout[PFTM_FRAG] = PFTM_FRAG_VAL;
301 	timeout[PFTM_INTERVAL] = PFTM_INTERVAL_VAL;
302 	timeout[PFTM_SRC_NODE] = PFTM_SRC_NODE_VAL;
303 	timeout[PFTM_TS_DIFF] = PFTM_TS_DIFF_VAL;
304 	timeout[PFTM_ADAPTIVE_START] = PFSTATE_ADAPT_START;
305 	timeout[PFTM_ADAPTIVE_END] = PFSTATE_ADAPT_END;
306 
307 	pf_normalize_init();
308 	bzero(&pf_status, sizeof(pf_status));
309 	pf_status.debug = PF_DEBUG_URGENT;
310 
311 	/* XXX do our best to avoid a conflict */
312 	pf_status.hostid = cprng_fast32();
313 
314 	/* require process context to purge states, so perform in a thread */
315 #ifdef __NetBSD__
316 	if (kthread_create(PRI_NONE, 0, NULL, pf_purge_thread, NULL, NULL,
317 	    "pfpurge"))
318 		panic("pfpurge thread");
319 #else
320 	kthread_create_deferred(pf_thread_create, NULL);
321 #endif /* !__NetBSD__ */
322 
323 #ifdef __NetBSD__
324 	pf_listener = kauth_listen_scope(KAUTH_SCOPE_NETWORK,
325 	    pf_listener_cb, NULL);
326 #endif /* __NetBSD__ */
327 }
328 
329 #ifdef _MODULE
330 void
331 pfdetach(void)
332 {
333 	extern int		 pf_purge_thread_running;
334 	extern int		 pf_purge_thread_stop;
335 	struct pf_anchor	*anchor;
336 	struct pf_state		*state;
337 	struct pf_src_node	*node;
338 	struct pfioc_table	 pt;
339 	u_int32_t		 ticket;
340 	int			 i;
341 	char			 r = '\0';
342 
343 	pf_purge_thread_stop = 1;
344 	wakeup(pf_purge_thread);
345 
346 	/* wait until the kthread exits */
347 	while (pf_purge_thread_running)
348 		tsleep(&pf_purge_thread_running, PWAIT, "pfdown", 0);
349 
350 	(void)pf_pfil_detach();
351 
352 	pf_status.running = 0;
353 
354 	/* clear the rulesets */
355 	for (i = 0; i < PF_RULESET_MAX; i++)
356 		if (pf_begin_rules(&ticket, i, &r) == 0)
357 			pf_commit_rules(ticket, i, &r);
358 #ifdef ALTQ
359 	if (pf_begin_altq(&ticket) == 0)
360 		pf_commit_altq(ticket);
361 #endif /* ALTQ */
362 
363 	/* clear states */
364 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
365 		state->timeout = PFTM_PURGE;
366 #if NPFSYNC > 0
367 		state->sync_flags = PFSTATE_NOSYNC;
368 #endif /* NPFSYNC > 0 */
369 	}
370 	pf_purge_expired_states(pf_status.states);
371 #if NPFSYNC > 0
372 	pfsync_clear_states(pf_status.hostid, NULL);
373 #endif /* NPFSYNC > 0 */
374 
375 	/* clear source nodes */
376 	RB_FOREACH(state, pf_state_tree_id, &tree_id) {
377 		state->src_node = NULL;
378 		state->nat_src_node = NULL;
379 	}
380 	RB_FOREACH(node, pf_src_tree, &tree_src_tracking) {
381 		node->expire = 1;
382 		node->states = 0;
383 	}
384 	pf_purge_expired_src_nodes(0);
385 
386 	/* clear tables */
387 	memset(&pt, '\0', sizeof(pt));
388 	pfr_clr_tables(&pt.pfrio_table, &pt.pfrio_ndel, pt.pfrio_flags);
389 
390 	/* destroy anchors */
391 	while ((anchor = RB_MIN(pf_anchor_global, &pf_anchors)) != NULL) {
392 		for (i = 0; i < PF_RULESET_MAX; i++)
393 			if (pf_begin_rules(&ticket, i, anchor->name) == 0)
394 				pf_commit_rules(ticket, i, anchor->name);
395 	}
396 
397 	/* destroy main ruleset */
398 	pf_remove_if_empty_ruleset(&pf_main_ruleset);
399 
400 	/* destroy the pools */
401 	pool_destroy(&pf_pooladdr_pl);
402 	pool_destroy(&pf_altq_pl);
403 	pool_destroy(&pf_state_key_pl);
404 	pool_destroy(&pf_state_pl);
405 	pool_destroy(&pf_rule_pl);
406 	pool_destroy(&pf_src_tree_pl);
407 
408 	rw_destroy(&pf_consistency_lock);
409 
410 	/* destroy subsystems */
411 	pf_normalize_destroy();
412 	pf_osfp_destroy();
413 	pfr_destroy();
414 	pfi_destroy();
415 
416 	/* cleanup kauth listener */
417 	kauth_unlisten_scope(pf_listener);
418 }
419 #endif /* _MODULE */
420 
421 #ifndef __NetBSD__
422 void
423 pf_thread_create(void *v)
424 {
425 	if (kthread_create(pf_purge_thread, NULL, NULL, "pfpurge"))
426 		panic("pfpurge thread");
427 }
428 #endif /* !__NetBSD__ */
429 
430 int
431 pfopen(dev_t dev, int flags, int fmt, struct lwp *l)
432 {
433 	if (minor(dev) >= 1)
434 		return (ENXIO);
435 	return (0);
436 }
437 
438 int
439 pfclose(dev_t dev, int flags, int fmt, struct lwp *l)
440 {
441 	if (minor(dev) >= 1)
442 		return (ENXIO);
443 	return (0);
444 }
445 
446 struct pf_pool *
447 pf_get_pool(char *anchor, u_int32_t ticket, u_int8_t rule_action,
448     u_int32_t rule_number, u_int8_t r_last, u_int8_t active,
449     u_int8_t check_ticket)
450 {
451 	struct pf_ruleset	*ruleset;
452 	struct pf_rule		*rule;
453 	int			 rs_num;
454 
455 	ruleset = pf_find_ruleset(anchor);
456 	if (ruleset == NULL)
457 		return (NULL);
458 	rs_num = pf_get_ruleset_number(rule_action);
459 	if (rs_num >= PF_RULESET_MAX)
460 		return (NULL);
461 	if (active) {
462 		if (check_ticket && ticket !=
463 		    ruleset->rules[rs_num].active.ticket)
464 			return (NULL);
465 		if (r_last)
466 			rule = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
467 			    pf_rulequeue);
468 		else
469 			rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
470 	} else {
471 		if (check_ticket && ticket !=
472 		    ruleset->rules[rs_num].inactive.ticket)
473 			return (NULL);
474 		if (r_last)
475 			rule = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
476 			    pf_rulequeue);
477 		else
478 			rule = TAILQ_FIRST(ruleset->rules[rs_num].inactive.ptr);
479 	}
480 	if (!r_last) {
481 		while ((rule != NULL) && (rule->nr != rule_number))
482 			rule = TAILQ_NEXT(rule, entries);
483 	}
484 	if (rule == NULL)
485 		return (NULL);
486 
487 	return (&rule->rpool);
488 }
489 
490 void
491 pf_mv_pool(struct pf_palist *poola, struct pf_palist *poolb)
492 {
493 	struct pf_pooladdr	*mv_pool_pa;
494 
495 	while ((mv_pool_pa = TAILQ_FIRST(poola)) != NULL) {
496 		TAILQ_REMOVE(poola, mv_pool_pa, entries);
497 		TAILQ_INSERT_TAIL(poolb, mv_pool_pa, entries);
498 	}
499 }
500 
501 void
502 pf_empty_pool(struct pf_palist *poola)
503 {
504 	struct pf_pooladdr	*empty_pool_pa;
505 
506 	while ((empty_pool_pa = TAILQ_FIRST(poola)) != NULL) {
507 		pfi_dynaddr_remove(&empty_pool_pa->addr);
508 		pf_tbladdr_remove(&empty_pool_pa->addr);
509 		pfi_kif_unref(empty_pool_pa->kif, PFI_KIF_REF_RULE);
510 		TAILQ_REMOVE(poola, empty_pool_pa, entries);
511 		pool_put(&pf_pooladdr_pl, empty_pool_pa);
512 	}
513 }
514 
515 void
516 pf_rm_rule(struct pf_rulequeue *rulequeue, struct pf_rule *rule)
517 {
518 	if (rulequeue != NULL) {
519 		if (rule->states <= 0) {
520 			/*
521 			 * XXX - we need to remove the table *before* detaching
522 			 * the rule to make sure the table code does not delete
523 			 * the anchor under our feet.
524 			 */
525 			pf_tbladdr_remove(&rule->src.addr);
526 			pf_tbladdr_remove(&rule->dst.addr);
527 			if (rule->overload_tbl)
528 				pfr_detach_table(rule->overload_tbl);
529 		}
530 		TAILQ_REMOVE(rulequeue, rule, entries);
531 		rule->entries.tqe_prev = NULL;
532 		rule->nr = -1;
533 	}
534 
535 	if (rule->states > 0 || rule->src_nodes > 0 ||
536 	    rule->entries.tqe_prev != NULL)
537 		return;
538 	pf_tag_unref(rule->tag);
539 	pf_tag_unref(rule->match_tag);
540 #ifdef ALTQ
541 	if (rule->pqid != rule->qid)
542 		pf_qid_unref(rule->pqid);
543 	pf_qid_unref(rule->qid);
544 #endif
545 	pf_rtlabel_remove(&rule->src.addr);
546 	pf_rtlabel_remove(&rule->dst.addr);
547 	pfi_dynaddr_remove(&rule->src.addr);
548 	pfi_dynaddr_remove(&rule->dst.addr);
549 	if (rulequeue == NULL) {
550 		pf_tbladdr_remove(&rule->src.addr);
551 		pf_tbladdr_remove(&rule->dst.addr);
552 		if (rule->overload_tbl)
553 			pfr_detach_table(rule->overload_tbl);
554 	}
555 	pfi_kif_unref(rule->kif, PFI_KIF_REF_RULE);
556 	pf_anchor_remove(rule);
557 	pf_empty_pool(&rule->rpool.list);
558 	pool_put(&pf_rule_pl, rule);
559 }
560 
561 u_int16_t
562 tagname2tag(struct pf_tags *head, char *tagname)
563 {
564 	struct pf_tagname	*tag, *p = NULL;
565 	u_int16_t		 new_tagid = 1;
566 
567 	TAILQ_FOREACH(tag, head, entries)
568 		if (strcmp(tagname, tag->name) == 0) {
569 			tag->ref++;
570 			return (tag->tag);
571 		}
572 
573 	/*
574 	 * to avoid fragmentation, we do a linear search from the beginning
575 	 * and take the first free slot we find. if there is none or the list
576 	 * is empty, append a new entry at the end.
577 	 */
578 
579 	/* new entry */
580 	if (!TAILQ_EMPTY(head))
581 		for (p = TAILQ_FIRST(head); p != NULL &&
582 		    p->tag == new_tagid; p = TAILQ_NEXT(p, entries))
583 			new_tagid = p->tag + 1;
584 
585 	if (new_tagid > TAGID_MAX)
586 		return (0);
587 
588 	/* allocate and fill new struct pf_tagname */
589 	tag = (struct pf_tagname *)malloc(sizeof(struct pf_tagname),
590 	    M_TEMP, M_NOWAIT);
591 	if (tag == NULL)
592 		return (0);
593 	bzero(tag, sizeof(struct pf_tagname));
594 	strlcpy(tag->name, tagname, sizeof(tag->name));
595 	tag->tag = new_tagid;
596 	tag->ref++;
597 
598 	if (p != NULL)	/* insert new entry before p */
599 		TAILQ_INSERT_BEFORE(p, tag, entries);
600 	else	/* either list empty or no free slot in between */
601 		TAILQ_INSERT_TAIL(head, tag, entries);
602 
603 	return (tag->tag);
604 }
605 
606 void
607 tag2tagname(struct pf_tags *head, u_int16_t tagid, char *p)
608 {
609 	struct pf_tagname	*tag;
610 
611 	TAILQ_FOREACH(tag, head, entries)
612 		if (tag->tag == tagid) {
613 			strlcpy(p, tag->name, PF_TAG_NAME_SIZE);
614 			return;
615 		}
616 }
617 
618 void
619 tag_unref(struct pf_tags *head, u_int16_t tag)
620 {
621 	struct pf_tagname	*p, *next;
622 
623 	if (tag == 0)
624 		return;
625 
626 	for (p = TAILQ_FIRST(head); p != NULL; p = next) {
627 		next = TAILQ_NEXT(p, entries);
628 		if (tag == p->tag) {
629 			if (--p->ref == 0) {
630 				TAILQ_REMOVE(head, p, entries);
631 				free(p, M_TEMP);
632 			}
633 			break;
634 		}
635 	}
636 }
637 
638 u_int16_t
639 pf_tagname2tag(char *tagname)
640 {
641 	return (tagname2tag(&pf_tags, tagname));
642 }
643 
644 void
645 pf_tag2tagname(u_int16_t tagid, char *p)
646 {
647 	tag2tagname(&pf_tags, tagid, p);
648 }
649 
650 void
651 pf_tag_ref(u_int16_t tag)
652 {
653 	struct pf_tagname *t;
654 
655 	TAILQ_FOREACH(t, &pf_tags, entries)
656 		if (t->tag == tag)
657 			break;
658 	if (t != NULL)
659 		t->ref++;
660 }
661 
662 void
663 pf_tag_unref(u_int16_t tag)
664 {
665 	tag_unref(&pf_tags, tag);
666 }
667 
668 int
669 pf_rtlabel_add(struct pf_addr_wrap *a)
670 {
671 #ifndef __NetBSD__
672 	if (a->type == PF_ADDR_RTLABEL &&
673 	    (a->v.rtlabel = rtlabel_name2id(a->v.rtlabelname)) == 0)
674 		return (-1);
675 #endif /* !__NetBSD__ */
676 	return (0);
677 }
678 
679 void
680 pf_rtlabel_remove(struct pf_addr_wrap *a)
681 {
682 #ifndef __NetBSD__
683 	if (a->type == PF_ADDR_RTLABEL)
684 		rtlabel_unref(a->v.rtlabel);
685 #endif /* !__NetBSD__ */
686 }
687 
688 void
689 pf_rtlabel_copyout(struct pf_addr_wrap *a)
690 {
691 #ifndef __NetBSD__
692 	const char	*name;
693 
694 	if (a->type == PF_ADDR_RTLABEL && a->v.rtlabel) {
695 		if ((name = rtlabel_id2name(a->v.rtlabel)) == NULL)
696 			strlcpy(a->v.rtlabelname, "?",
697 			    sizeof(a->v.rtlabelname));
698 		else
699 			strlcpy(a->v.rtlabelname, name,
700 			    sizeof(a->v.rtlabelname));
701 	}
702 #endif /* !__NetBSD__ */
703 }
704 
705 #ifdef ALTQ
706 u_int32_t
707 pf_qname2qid(char *qname)
708 {
709 	return ((u_int32_t)tagname2tag(&pf_qids, qname));
710 }
711 
712 void
713 pf_qid2qname(u_int32_t qid, char *p)
714 {
715 	tag2tagname(&pf_qids, (u_int16_t)qid, p);
716 }
717 
718 void
719 pf_qid_unref(u_int32_t qid)
720 {
721 	tag_unref(&pf_qids, (u_int16_t)qid);
722 }
723 
724 int
725 pf_begin_altq(u_int32_t *ticket)
726 {
727 	struct pf_altq	*altq;
728 	int		 error = 0;
729 
730 	/* Purge the old altq list */
731 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
732 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
733 		if (altq->qname[0] == 0) {
734 			/* detach and destroy the discipline */
735 			error = altq_remove(altq);
736 		} else
737 			pf_qid_unref(altq->qid);
738 		pool_put(&pf_altq_pl, altq);
739 	}
740 	if (error)
741 		return (error);
742 	*ticket = ++ticket_altqs_inactive;
743 	altqs_inactive_open = 1;
744 	return (0);
745 }
746 
747 int
748 pf_rollback_altq(u_int32_t ticket)
749 {
750 	struct pf_altq	*altq;
751 	int		 error = 0;
752 
753 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
754 		return (0);
755 	/* Purge the old altq list */
756 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
757 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
758 		if (altq->qname[0] == 0) {
759 			/* detach and destroy the discipline */
760 			error = altq_remove(altq);
761 		} else
762 			pf_qid_unref(altq->qid);
763 		pool_put(&pf_altq_pl, altq);
764 	}
765 	altqs_inactive_open = 0;
766 	return (error);
767 }
768 
769 int
770 pf_commit_altq(u_int32_t ticket)
771 {
772 	struct pf_altqqueue	*old_altqs;
773 	struct pf_altq		*altq;
774 	int			 s, err, error = 0;
775 
776 	if (!altqs_inactive_open || ticket != ticket_altqs_inactive)
777 		return (EBUSY);
778 
779 	/* swap altqs, keep the old. */
780 	s = splsoftnet();
781 	old_altqs = pf_altqs_active;
782 	pf_altqs_active = pf_altqs_inactive;
783 	pf_altqs_inactive = old_altqs;
784 	ticket_altqs_active = ticket_altqs_inactive;
785 
786 	/* Attach new disciplines */
787 	TAILQ_FOREACH(altq, pf_altqs_active, entries) {
788 		if (altq->qname[0] == 0) {
789 			/* attach the discipline */
790 			error = altq_pfattach(altq);
791 			if (error == 0 && pf_altq_running)
792 				error = pf_enable_altq(altq);
793 			if (error != 0) {
794 				splx(s);
795 				return (error);
796 			}
797 		}
798 	}
799 
800 	/* Purge the old altq list */
801 	while ((altq = TAILQ_FIRST(pf_altqs_inactive)) != NULL) {
802 		TAILQ_REMOVE(pf_altqs_inactive, altq, entries);
803 		if (altq->qname[0] == 0) {
804 			/* detach and destroy the discipline */
805 			if (pf_altq_running)
806 				error = pf_disable_altq(altq);
807 			err = altq_pfdetach(altq);
808 			if (err != 0 && error == 0)
809 				error = err;
810 			err = altq_remove(altq);
811 			if (err != 0 && error == 0)
812 				error = err;
813 		} else
814 			pf_qid_unref(altq->qid);
815 		pool_put(&pf_altq_pl, altq);
816 	}
817 	splx(s);
818 
819 	altqs_inactive_open = 0;
820 	return (error);
821 }
822 
823 int
824 pf_enable_altq(struct pf_altq *altq)
825 {
826 	struct ifnet		*ifp;
827 	struct tb_profile	 tb;
828 	int			 s, error = 0;
829 
830 	if ((ifp = ifunit(altq->ifname)) == NULL)
831 		return (EINVAL);
832 
833 	if (ifp->if_snd.altq_type != ALTQT_NONE)
834 		error = altq_enable(&ifp->if_snd);
835 
836 	/* set tokenbucket regulator */
837 	if (error == 0 && ifp != NULL && ALTQ_IS_ENABLED(&ifp->if_snd)) {
838 		tb.rate = altq->ifbandwidth;
839 		tb.depth = altq->tbrsize;
840 		s = splnet();
841 		error = tbr_set(&ifp->if_snd, &tb);
842 		splx(s);
843 	}
844 
845 	return (error);
846 }
847 
848 int
849 pf_disable_altq(struct pf_altq *altq)
850 {
851 	struct ifnet		*ifp;
852 	struct tb_profile	 tb;
853 	int			 s, error;
854 
855 	if ((ifp = ifunit(altq->ifname)) == NULL)
856 		return (EINVAL);
857 
858 	/*
859 	 * when the discipline is no longer referenced, it was overridden
860 	 * by a new one.  if so, just return.
861 	 */
862 	if (altq->altq_disc != ifp->if_snd.altq_disc)
863 		return (0);
864 
865 	error = altq_disable(&ifp->if_snd);
866 
867 	if (error == 0) {
868 		/* clear tokenbucket regulator */
869 		tb.rate = 0;
870 		s = splnet();
871 		error = tbr_set(&ifp->if_snd, &tb);
872 		splx(s);
873 	}
874 
875 	return (error);
876 }
877 #endif /* ALTQ */
878 
879 int
880 pf_begin_rules(u_int32_t *ticket, int rs_num, const char *anchor)
881 {
882 	struct pf_ruleset	*rs;
883 	struct pf_rule		*rule;
884 
885 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
886 		return (EINVAL);
887 	rs = pf_find_or_create_ruleset(anchor);
888 	if (rs == NULL)
889 		return (EINVAL);
890 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
891 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
892 		rs->rules[rs_num].inactive.rcount--;
893 	}
894 	*ticket = ++rs->rules[rs_num].inactive.ticket;
895 	rs->rules[rs_num].inactive.open = 1;
896 	return (0);
897 }
898 
899 int
900 pf_rollback_rules(u_int32_t ticket, int rs_num, char *anchor)
901 {
902 	struct pf_ruleset	*rs;
903 	struct pf_rule		*rule;
904 
905 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
906 		return (EINVAL);
907 	rs = pf_find_ruleset(anchor);
908 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
909 	    rs->rules[rs_num].inactive.ticket != ticket)
910 		return (0);
911 	while ((rule = TAILQ_FIRST(rs->rules[rs_num].inactive.ptr)) != NULL) {
912 		pf_rm_rule(rs->rules[rs_num].inactive.ptr, rule);
913 		rs->rules[rs_num].inactive.rcount--;
914 	}
915 	rs->rules[rs_num].inactive.open = 0;
916 	return (0);
917 }
918 
919 #define PF_MD5_UPD(st, elm)						\
920 		MD5Update(ctx, (u_int8_t *) &(st)->elm, sizeof((st)->elm))
921 
922 #define PF_MD5_UPD_STR(st, elm)						\
923 		MD5Update(ctx, (u_int8_t *) (st)->elm, strlen((st)->elm))
924 
925 #define PF_MD5_UPD_HTONL(st, elm, stor) do {				\
926 		(stor) = htonl((st)->elm);				\
927 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int32_t));\
928 } while (0)
929 
930 #define PF_MD5_UPD_HTONS(st, elm, stor) do {				\
931 		(stor) = htons((st)->elm);				\
932 		MD5Update(ctx, (u_int8_t *) &(stor), sizeof(u_int16_t));\
933 } while (0)
934 
935 void
936 pf_hash_rule_addr(MD5_CTX *ctx, struct pf_rule_addr *pfr)
937 {
938 	PF_MD5_UPD(pfr, addr.type);
939 	switch (pfr->addr.type) {
940 		case PF_ADDR_DYNIFTL:
941 			PF_MD5_UPD(pfr, addr.v.ifname);
942 			PF_MD5_UPD(pfr, addr.iflags);
943 			break;
944 		case PF_ADDR_TABLE:
945 			PF_MD5_UPD(pfr, addr.v.tblname);
946 			break;
947 		case PF_ADDR_ADDRMASK:
948 			/* XXX ignore af? */
949 			PF_MD5_UPD(pfr, addr.v.a.addr.addr32);
950 			PF_MD5_UPD(pfr, addr.v.a.mask.addr32);
951 			break;
952 		case PF_ADDR_RTLABEL:
953 			PF_MD5_UPD(pfr, addr.v.rtlabelname);
954 			break;
955 	}
956 
957 	PF_MD5_UPD(pfr, port[0]);
958 	PF_MD5_UPD(pfr, port[1]);
959 	PF_MD5_UPD(pfr, neg);
960 	PF_MD5_UPD(pfr, port_op);
961 }
962 
963 void
964 pf_hash_rule(MD5_CTX *ctx, struct pf_rule *rule)
965 {
966 	u_int16_t x;
967 	u_int32_t y;
968 
969 	pf_hash_rule_addr(ctx, &rule->src);
970 	pf_hash_rule_addr(ctx, &rule->dst);
971 	PF_MD5_UPD_STR(rule, label);
972 	PF_MD5_UPD_STR(rule, ifname);
973 	PF_MD5_UPD_STR(rule, match_tagname);
974 	PF_MD5_UPD_HTONS(rule, match_tag, x); /* dup? */
975 	PF_MD5_UPD_HTONL(rule, os_fingerprint, y);
976 	PF_MD5_UPD_HTONL(rule, prob, y);
977 	PF_MD5_UPD_HTONL(rule, uid.uid[0], y);
978 	PF_MD5_UPD_HTONL(rule, uid.uid[1], y);
979 	PF_MD5_UPD(rule, uid.op);
980 	PF_MD5_UPD_HTONL(rule, gid.gid[0], y);
981 	PF_MD5_UPD_HTONL(rule, gid.gid[1], y);
982 	PF_MD5_UPD(rule, gid.op);
983 	PF_MD5_UPD_HTONL(rule, rule_flag, y);
984 	PF_MD5_UPD(rule, action);
985 	PF_MD5_UPD(rule, direction);
986 	PF_MD5_UPD(rule, af);
987 	PF_MD5_UPD(rule, quick);
988 	PF_MD5_UPD(rule, ifnot);
989 	PF_MD5_UPD(rule, match_tag_not);
990 	PF_MD5_UPD(rule, natpass);
991 	PF_MD5_UPD(rule, keep_state);
992 	PF_MD5_UPD(rule, proto);
993 	PF_MD5_UPD(rule, type);
994 	PF_MD5_UPD(rule, code);
995 	PF_MD5_UPD(rule, flags);
996 	PF_MD5_UPD(rule, flagset);
997 	PF_MD5_UPD(rule, allow_opts);
998 	PF_MD5_UPD(rule, rt);
999 	PF_MD5_UPD(rule, tos);
1000 }
1001 
1002 int
1003 pf_commit_rules(u_int32_t ticket, int rs_num, char *anchor)
1004 {
1005 	struct pf_ruleset	*rs;
1006 	struct pf_rule		*rule, **old_array;
1007 	struct pf_rulequeue	*old_rules;
1008 	int			 s, error;
1009 	u_int32_t		 old_rcount;
1010 
1011 	if (rs_num < 0 || rs_num >= PF_RULESET_MAX)
1012 		return (EINVAL);
1013 	rs = pf_find_ruleset(anchor);
1014 	if (rs == NULL || !rs->rules[rs_num].inactive.open ||
1015 	    ticket != rs->rules[rs_num].inactive.ticket)
1016 		return (EBUSY);
1017 
1018 	/* Calculate checksum for the main ruleset */
1019 	if (rs == &pf_main_ruleset) {
1020 		error = pf_setup_pfsync_matching(rs);
1021 		if (error != 0)
1022 			return (error);
1023 	}
1024 
1025 	/* Swap rules, keep the old. */
1026 	s = splsoftnet();
1027 	old_rules = rs->rules[rs_num].active.ptr;
1028 	old_rcount = rs->rules[rs_num].active.rcount;
1029 	old_array = rs->rules[rs_num].active.ptr_array;
1030 
1031 	rs->rules[rs_num].active.ptr =
1032 	    rs->rules[rs_num].inactive.ptr;
1033 	rs->rules[rs_num].active.ptr_array =
1034 	    rs->rules[rs_num].inactive.ptr_array;
1035 	rs->rules[rs_num].active.rcount =
1036 	    rs->rules[rs_num].inactive.rcount;
1037 	rs->rules[rs_num].inactive.ptr = old_rules;
1038 	rs->rules[rs_num].inactive.ptr_array = old_array;
1039 	rs->rules[rs_num].inactive.rcount = old_rcount;
1040 
1041 	rs->rules[rs_num].active.ticket =
1042 	    rs->rules[rs_num].inactive.ticket;
1043 	pf_calc_skip_steps(rs->rules[rs_num].active.ptr);
1044 
1045 
1046 	/* Purge the old rule list. */
1047 	while ((rule = TAILQ_FIRST(old_rules)) != NULL)
1048 		pf_rm_rule(old_rules, rule);
1049 	if (rs->rules[rs_num].inactive.ptr_array)
1050 		free(rs->rules[rs_num].inactive.ptr_array, M_TEMP);
1051 	rs->rules[rs_num].inactive.ptr_array = NULL;
1052 	rs->rules[rs_num].inactive.rcount = 0;
1053 	rs->rules[rs_num].inactive.open = 0;
1054 	pf_remove_if_empty_ruleset(rs);
1055 	splx(s);
1056 	return (0);
1057 }
1058 
1059 void
1060 pf_state_export(struct pfsync_state *sp, struct pf_state_key *sk,
1061    struct pf_state *s)
1062 {
1063 	int secs = time_second;
1064 	bzero(sp, sizeof(struct pfsync_state));
1065 
1066 	/* copy from state key */
1067 	sp->lan.addr = sk->lan.addr;
1068 	sp->lan.port = sk->lan.port;
1069 	sp->gwy.addr = sk->gwy.addr;
1070 	sp->gwy.port = sk->gwy.port;
1071 	sp->ext.addr = sk->ext.addr;
1072 	sp->ext.port = sk->ext.port;
1073 	sp->proto = sk->proto;
1074 	sp->af = sk->af;
1075 	sp->direction = sk->direction;
1076 
1077 	/* copy from state */
1078 	memcpy(&sp->id, &s->id, sizeof(sp->id));
1079 	sp->creatorid = s->creatorid;
1080 	strlcpy(sp->ifname, s->kif->pfik_name, sizeof(sp->ifname));
1081 	pf_state_peer_to_pfsync(&s->src, &sp->src);
1082 	pf_state_peer_to_pfsync(&s->dst, &sp->dst);
1083 
1084 	sp->rule = s->rule.ptr->nr;
1085 	sp->nat_rule = (s->nat_rule.ptr == NULL) ?  -1 : s->nat_rule.ptr->nr;
1086 	sp->anchor = (s->anchor.ptr == NULL) ?  -1 : s->anchor.ptr->nr;
1087 
1088 	pf_state_counter_to_pfsync(s->bytes[0], sp->bytes[0]);
1089 	pf_state_counter_to_pfsync(s->bytes[1], sp->bytes[1]);
1090 	pf_state_counter_to_pfsync(s->packets[0], sp->packets[0]);
1091 	pf_state_counter_to_pfsync(s->packets[1], sp->packets[1]);
1092 	sp->creation = secs - s->creation;
1093 	sp->expire = pf_state_expires(s);
1094 	sp->log = s->log;
1095 	sp->allow_opts = s->allow_opts;
1096 	sp->timeout = s->timeout;
1097 
1098 	if (s->src_node)
1099 		sp->sync_flags |= PFSYNC_FLAG_SRCNODE;
1100 	if (s->nat_src_node)
1101 		sp->sync_flags |= PFSYNC_FLAG_NATSRCNODE;
1102 
1103 	if (sp->expire > secs)
1104 		sp->expire -= secs;
1105 	else
1106 		sp->expire = 0;
1107 
1108 }
1109 
1110 void
1111 pf_state_import(struct pfsync_state *sp, struct pf_state_key *sk,
1112    struct pf_state *s)
1113 {
1114 	/* copy to state key */
1115 	sk->lan.addr = sp->lan.addr;
1116 	sk->lan.port = sp->lan.port;
1117 	sk->gwy.addr = sp->gwy.addr;
1118 	sk->gwy.port = sp->gwy.port;
1119 	sk->ext.addr = sp->ext.addr;
1120 	sk->ext.port = sp->ext.port;
1121 	sk->proto = sp->proto;
1122 	sk->af = sp->af;
1123 	sk->direction = sp->direction;
1124 
1125 	/* copy to state */
1126 	memcpy(&s->id, &sp->id, sizeof(sp->id));
1127 	s->creatorid = sp->creatorid;
1128 	pf_state_peer_from_pfsync(&sp->src, &s->src);
1129 	pf_state_peer_from_pfsync(&sp->dst, &s->dst);
1130 
1131 	s->rule.ptr = &pf_default_rule;
1132 	s->rule.ptr->states++;
1133 	s->nat_rule.ptr = NULL;
1134 	s->anchor.ptr = NULL;
1135 	s->rt_kif = NULL;
1136 	s->creation = time_second;
1137 	s->expire = time_second;
1138 	s->timeout = sp->timeout;
1139 	if (sp->expire > 0)
1140 		s->expire -= pf_default_rule.timeout[sp->timeout] - sp->expire;
1141 	s->pfsync_time = 0;
1142 	s->packets[0] = s->packets[1] = 0;
1143 	s->bytes[0] = s->bytes[1] = 0;
1144 }
1145 
1146 int
1147 pf_state_add(struct pfsync_state* sp)
1148 {
1149 	struct pf_state		*s;
1150 	struct pf_state_key	*sk;
1151 	struct pfi_kif		*kif;
1152 
1153 	if (sp->timeout >= PFTM_MAX &&
1154 			sp->timeout != PFTM_UNTIL_PACKET) {
1155 		return EINVAL;
1156 	}
1157 	s = pool_get(&pf_state_pl, PR_NOWAIT);
1158 	if (s == NULL) {
1159 		return ENOMEM;
1160 	}
1161 	bzero(s, sizeof(struct pf_state));
1162 	if ((sk = pf_alloc_state_key(s)) == NULL) {
1163 		pool_put(&pf_state_pl, s);
1164 		return ENOMEM;
1165 	}
1166 	pf_state_import(sp, sk, s);
1167 	kif = pfi_kif_get(sp->ifname);
1168 	if (kif == NULL) {
1169 		pool_put(&pf_state_pl, s);
1170 		pool_put(&pf_state_key_pl, sk);
1171 		return ENOENT;
1172 	}
1173 	if (pf_insert_state(kif, s)) {
1174 		pfi_kif_unref(kif, PFI_KIF_REF_NONE);
1175 		pool_put(&pf_state_pl, s);
1176 		return ENOMEM;
1177 	}
1178 
1179 	return 0;
1180 }
1181 
1182 
1183 int
1184 pf_setup_pfsync_matching(struct pf_ruleset *rs)
1185 {
1186 	MD5_CTX			 ctx;
1187 	struct pf_rule		*rule;
1188 	int			 rs_cnt;
1189 	u_int8_t		 digest[PF_MD5_DIGEST_LENGTH];
1190 
1191 	MD5Init(&ctx);
1192 	for (rs_cnt = 0; rs_cnt < PF_RULESET_MAX; rs_cnt++) {
1193 		/* XXX PF_RULESET_SCRUB as well? */
1194 		if (rs_cnt == PF_RULESET_SCRUB)
1195 			continue;
1196 
1197 		if (rs->rules[rs_cnt].inactive.ptr_array)
1198 			free(rs->rules[rs_cnt].inactive.ptr_array, M_TEMP);
1199 		rs->rules[rs_cnt].inactive.ptr_array = NULL;
1200 
1201 		if (rs->rules[rs_cnt].inactive.rcount) {
1202 			rs->rules[rs_cnt].inactive.ptr_array =
1203 			    malloc(sizeof(void *) *
1204 			    rs->rules[rs_cnt].inactive.rcount,
1205 			    M_TEMP, M_NOWAIT);
1206 
1207 			if (!rs->rules[rs_cnt].inactive.ptr_array)
1208 				return (ENOMEM);
1209 		}
1210 
1211 		TAILQ_FOREACH(rule, rs->rules[rs_cnt].inactive.ptr,
1212 		    entries) {
1213 			pf_hash_rule(&ctx, rule);
1214 			(rs->rules[rs_cnt].inactive.ptr_array)[rule->nr] = rule;
1215 		}
1216 	}
1217 
1218 	MD5Final(digest, &ctx);
1219 	memcpy(pf_status.pf_chksum, digest, sizeof(pf_status.pf_chksum));
1220 	return (0);
1221 }
1222 
1223 int
1224 pfioctl(dev_t dev, u_long cmd, void *addr, int flags, struct lwp *l)
1225 {
1226 	struct pf_pooladdr	*pa = NULL;
1227 	struct pf_pool		*pool = NULL;
1228 	int			 s;
1229 	int			 error = 0;
1230 
1231 	/* XXX keep in sync with switch() below */
1232 	if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_FIREWALL,
1233 	    KAUTH_REQ_NETWORK_FIREWALL_FW, NULL, NULL, NULL))
1234 		switch (cmd) {
1235 		case DIOCGETRULES:
1236 		case DIOCGETRULE:
1237 		case DIOCGETADDRS:
1238 		case DIOCGETADDR:
1239 		case DIOCGETSTATE:
1240 		case DIOCSETSTATUSIF:
1241 		case DIOCGETSTATUS:
1242 		case DIOCCLRSTATUS:
1243 		case DIOCNATLOOK:
1244 		case DIOCSETDEBUG:
1245 		case DIOCGETSTATES:
1246 		case DIOCGETTIMEOUT:
1247 		case DIOCCLRRULECTRS:
1248 		case DIOCGETLIMIT:
1249 		case DIOCGETALTQS:
1250 		case DIOCGETALTQ:
1251 		case DIOCGETQSTATS:
1252 		case DIOCGETRULESETS:
1253 		case DIOCGETRULESET:
1254 		case DIOCRGETTABLES:
1255 		case DIOCRGETTSTATS:
1256 		case DIOCRCLRTSTATS:
1257 		case DIOCRCLRADDRS:
1258 		case DIOCRADDADDRS:
1259 		case DIOCRDELADDRS:
1260 		case DIOCRSETADDRS:
1261 		case DIOCRGETADDRS:
1262 		case DIOCRGETASTATS:
1263 		case DIOCRCLRASTATS:
1264 		case DIOCRTSTADDRS:
1265 		case DIOCOSFPGET:
1266 		case DIOCGETSRCNODES:
1267 		case DIOCCLRSRCNODES:
1268 		case DIOCIGETIFACES:
1269 		case DIOCSETIFFLAG:
1270 		case DIOCCLRIFFLAG:
1271 		case DIOCSETLCK:
1272 		case DIOCADDSTATES:
1273 			break;
1274 		case DIOCRCLRTABLES:
1275 		case DIOCRADDTABLES:
1276 		case DIOCRDELTABLES:
1277 		case DIOCRSETTFLAGS:
1278 			if (((struct pfioc_table *)addr)->pfrio_flags &
1279 			    PFR_FLAG_DUMMY)
1280 				break; /* dummy operation ok */
1281 			return (EPERM);
1282 		default:
1283 			return (EPERM);
1284 		}
1285 
1286 	if (!(flags & FWRITE))
1287 		switch (cmd) {
1288 		case DIOCGETRULES:
1289 		case DIOCGETADDRS:
1290 		case DIOCGETADDR:
1291 		case DIOCGETSTATE:
1292 		case DIOCGETSTATUS:
1293 		case DIOCGETSTATES:
1294 		case DIOCGETTIMEOUT:
1295 		case DIOCGETLIMIT:
1296 		case DIOCGETALTQS:
1297 		case DIOCGETALTQ:
1298 		case DIOCGETQSTATS:
1299 		case DIOCGETRULESETS:
1300 		case DIOCGETRULESET:
1301 		case DIOCNATLOOK:
1302 		case DIOCRGETTABLES:
1303 		case DIOCRGETTSTATS:
1304 		case DIOCRGETADDRS:
1305 		case DIOCRGETASTATS:
1306 		case DIOCRTSTADDRS:
1307 		case DIOCOSFPGET:
1308 		case DIOCGETSRCNODES:
1309 		case DIOCIGETIFACES:
1310 		case DIOCSETLCK:
1311 			break;
1312 		case DIOCRCLRTABLES:
1313 		case DIOCRADDTABLES:
1314 		case DIOCRDELTABLES:
1315 		case DIOCRCLRTSTATS:
1316 		case DIOCRCLRADDRS:
1317 		case DIOCRADDADDRS:
1318 		case DIOCRDELADDRS:
1319 		case DIOCRSETADDRS:
1320 		case DIOCRSETTFLAGS:
1321 		case DIOCADDSTATES:
1322 			if (((struct pfioc_table *)addr)->pfrio_flags &
1323 			    PFR_FLAG_DUMMY) {
1324 				flags |= FWRITE; /* need write lock for dummy */
1325 				break; /* dummy operation ok */
1326 			}
1327 			return (EACCES);
1328 		case DIOCGETRULE:
1329 			if (((struct pfioc_rule *)addr)->action == PF_GET_CLR_CNTR)
1330 				return (EACCES);
1331 			break;
1332 		default:
1333 			return (EACCES);
1334 		}
1335 
1336 	if (flags & FWRITE)
1337 		rw_enter_write(&pf_consistency_lock);
1338 	else
1339 		rw_enter_read(&pf_consistency_lock);
1340 
1341 	s = splsoftnet();
1342 	switch (cmd) {
1343 
1344 	case DIOCSTART:
1345 		if (pf_status.running)
1346 			error = EEXIST;
1347 		else {
1348 #ifdef __NetBSD__
1349 			error = pf_pfil_attach();
1350 			if (error)
1351 				break;
1352 #endif /* __NetBSD__ */
1353 			pf_status.running = 1;
1354 			pf_status.since = time_second;
1355 			if (pf_status.stateid == 0) {
1356 				pf_status.stateid = time_second;
1357 				pf_status.stateid = pf_status.stateid << 32;
1358 			}
1359 			DPFPRINTF(PF_DEBUG_MISC, ("pf: started\n"));
1360 		}
1361 		break;
1362 
1363 	case DIOCSTOP:
1364 		if (!pf_status.running)
1365 			error = ENOENT;
1366 		else {
1367 #ifdef __NetBSD__
1368 			error = pf_pfil_detach();
1369 			if (error)
1370 				break;
1371 #endif /* __NetBSD__ */
1372 			pf_status.running = 0;
1373 			pf_status.since = time_second;
1374 			DPFPRINTF(PF_DEBUG_MISC, ("pf: stopped\n"));
1375 		}
1376 		break;
1377 
1378 	case DIOCADDRULE: {
1379 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1380 		struct pf_ruleset	*ruleset;
1381 		struct pf_rule		*rule, *tail;
1382 		int			 rs_num;
1383 
1384 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1385 		ruleset = pf_find_ruleset(pr->anchor);
1386 		if (ruleset == NULL) {
1387 			error = EINVAL;
1388 			break;
1389 		}
1390 		rs_num = pf_get_ruleset_number(pr->rule.action);
1391 		if (rs_num >= PF_RULESET_MAX) {
1392 			error = EINVAL;
1393 			break;
1394 		}
1395 		if (pr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1396 			error = EINVAL;
1397 			break;
1398 		}
1399 		if (pr->ticket != ruleset->rules[rs_num].inactive.ticket) {
1400 			error = EBUSY;
1401 			break;
1402 		}
1403 		if (pr->pool_ticket != ticket_pabuf) {
1404 			error = EBUSY;
1405 			break;
1406 		}
1407 		rule = pool_get(&pf_rule_pl, PR_NOWAIT);
1408 		if (rule == NULL) {
1409 			error = ENOMEM;
1410 			break;
1411 		}
1412 		bcopy(&pr->rule, rule, sizeof(struct pf_rule));
1413 #ifdef __NetBSD__
1414 		rule->cuid = kauth_cred_getuid(l->l_cred);
1415 		rule->cpid = l->l_proc->p_pid;
1416 #else
1417 		rule->cuid = p->p_cred->p_ruid;
1418 		rule->cpid = p->p_pid;
1419 #endif /* !__NetBSD__ */
1420 		rule->anchor = NULL;
1421 		rule->kif = NULL;
1422 		TAILQ_INIT(&rule->rpool.list);
1423 		/* initialize refcounting */
1424 		rule->states = 0;
1425 		rule->src_nodes = 0;
1426 		rule->entries.tqe_prev = NULL;
1427 #ifndef INET
1428 		if (rule->af == AF_INET) {
1429 			pool_put(&pf_rule_pl, rule);
1430 			error = EAFNOSUPPORT;
1431 			break;
1432 		}
1433 #endif /* INET */
1434 #ifndef INET6
1435 		if (rule->af == AF_INET6) {
1436 			pool_put(&pf_rule_pl, rule);
1437 			error = EAFNOSUPPORT;
1438 			break;
1439 		}
1440 #endif /* INET6 */
1441 		tail = TAILQ_LAST(ruleset->rules[rs_num].inactive.ptr,
1442 		    pf_rulequeue);
1443 		if (tail)
1444 			rule->nr = tail->nr + 1;
1445 		else
1446 			rule->nr = 0;
1447 		if (rule->ifname[0]) {
1448 			rule->kif = pfi_kif_get(rule->ifname);
1449 			if (rule->kif == NULL) {
1450 				pool_put(&pf_rule_pl, rule);
1451 				error = EINVAL;
1452 				break;
1453 			}
1454 			pfi_kif_ref(rule->kif, PFI_KIF_REF_RULE);
1455 		}
1456 
1457 #ifndef __NetBSD__
1458 		if (rule->rtableid > 0 && !rtable_exists(rule->rtableid))
1459 			error = EBUSY;
1460 #endif /* !__NetBSD__ */
1461 
1462 #ifdef ALTQ
1463 		/* set queue IDs */
1464 		if (rule->qname[0] != 0) {
1465 			if ((rule->qid = pf_qname2qid(rule->qname)) == 0)
1466 				error = EBUSY;
1467 			else if (rule->pqname[0] != 0) {
1468 				if ((rule->pqid =
1469 				    pf_qname2qid(rule->pqname)) == 0)
1470 					error = EBUSY;
1471 			} else
1472 				rule->pqid = rule->qid;
1473 		}
1474 #endif
1475 		if (rule->tagname[0])
1476 			if ((rule->tag = pf_tagname2tag(rule->tagname)) == 0)
1477 				error = EBUSY;
1478 		if (rule->match_tagname[0])
1479 			if ((rule->match_tag =
1480 			    pf_tagname2tag(rule->match_tagname)) == 0)
1481 				error = EBUSY;
1482 		if (rule->rt && !rule->direction)
1483 			error = EINVAL;
1484 #if NPFLOG > 0
1485 		if (!rule->log)
1486 			rule->logif = 0;
1487 		if (rule->logif >= PFLOGIFS_MAX)
1488 			error = EINVAL;
1489 #endif
1490 		if (pf_rtlabel_add(&rule->src.addr) ||
1491 		    pf_rtlabel_add(&rule->dst.addr))
1492 			error = EBUSY;
1493 		if (pfi_dynaddr_setup(&rule->src.addr, rule->af))
1494 			error = EINVAL;
1495 		if (pfi_dynaddr_setup(&rule->dst.addr, rule->af))
1496 			error = EINVAL;
1497 		if (pf_tbladdr_setup(ruleset, &rule->src.addr))
1498 			error = EINVAL;
1499 		if (pf_tbladdr_setup(ruleset, &rule->dst.addr))
1500 			error = EINVAL;
1501 		if (pf_anchor_setup(rule, ruleset, pr->anchor_call))
1502 			error = EINVAL;
1503 		TAILQ_FOREACH(pa, &pf_pabuf, entries)
1504 			if (pf_tbladdr_setup(ruleset, &pa->addr))
1505 				error = EINVAL;
1506 
1507 		rule->overload_tbl = NULL;
1508 		if (rule->overload_tblname[0]) {
1509 			if ((rule->overload_tbl = pfr_attach_table(ruleset,
1510 			    rule->overload_tblname)) == NULL)
1511 				error = EINVAL;
1512 			else
1513 				rule->overload_tbl->pfrkt_flags |=
1514 				    PFR_TFLAG_ACTIVE;
1515 		}
1516 
1517 		pf_mv_pool(&pf_pabuf, &rule->rpool.list);
1518 		if (((((rule->action == PF_NAT) || (rule->action == PF_RDR) ||
1519 		    (rule->action == PF_BINAT)) && rule->anchor == NULL) ||
1520 		    (rule->rt > PF_FASTROUTE)) &&
1521 		    (TAILQ_FIRST(&rule->rpool.list) == NULL))
1522 			error = EINVAL;
1523 
1524 		if (error) {
1525 			pf_rm_rule(NULL, rule);
1526 			break;
1527 		}
1528 		rule->rpool.cur = TAILQ_FIRST(&rule->rpool.list);
1529 		rule->evaluations = rule->packets[0] = rule->packets[1] =
1530 		    rule->bytes[0] = rule->bytes[1] = 0;
1531 		TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr,
1532 		    rule, entries);
1533 		ruleset->rules[rs_num].inactive.rcount++;
1534 		break;
1535 	}
1536 
1537 	case DIOCGETRULES: {
1538 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1539 		struct pf_ruleset	*ruleset;
1540 		struct pf_rule		*tail;
1541 		int			 rs_num;
1542 
1543 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1544 		ruleset = pf_find_ruleset(pr->anchor);
1545 		if (ruleset == NULL) {
1546 			error = EINVAL;
1547 			break;
1548 		}
1549 		rs_num = pf_get_ruleset_number(pr->rule.action);
1550 		if (rs_num >= PF_RULESET_MAX) {
1551 			error = EINVAL;
1552 			break;
1553 		}
1554 		tail = TAILQ_LAST(ruleset->rules[rs_num].active.ptr,
1555 		    pf_rulequeue);
1556 		if (tail)
1557 			pr->nr = tail->nr + 1;
1558 		else
1559 			pr->nr = 0;
1560 		pr->ticket = ruleset->rules[rs_num].active.ticket;
1561 		break;
1562 	}
1563 
1564 	case DIOCGETRULE: {
1565 		struct pfioc_rule	*pr = (struct pfioc_rule *)addr;
1566 		struct pf_ruleset	*ruleset;
1567 		struct pf_rule		*rule;
1568 		int			 rs_num, i;
1569 
1570 		pr->anchor[sizeof(pr->anchor) - 1] = 0;
1571 		ruleset = pf_find_ruleset(pr->anchor);
1572 		if (ruleset == NULL) {
1573 			error = EINVAL;
1574 			break;
1575 		}
1576 		rs_num = pf_get_ruleset_number(pr->rule.action);
1577 		if (rs_num >= PF_RULESET_MAX) {
1578 			error = EINVAL;
1579 			break;
1580 		}
1581 		if (pr->ticket != ruleset->rules[rs_num].active.ticket) {
1582 			error = EBUSY;
1583 			break;
1584 		}
1585 		rule = TAILQ_FIRST(ruleset->rules[rs_num].active.ptr);
1586 		while ((rule != NULL) && (rule->nr != pr->nr))
1587 			rule = TAILQ_NEXT(rule, entries);
1588 		if (rule == NULL) {
1589 			error = EBUSY;
1590 			break;
1591 		}
1592 		bcopy(rule, &pr->rule, sizeof(struct pf_rule));
1593 		if (pf_anchor_copyout(ruleset, rule, pr)) {
1594 			error = EBUSY;
1595 			break;
1596 		}
1597 		pfi_dynaddr_copyout(&pr->rule.src.addr);
1598 		pfi_dynaddr_copyout(&pr->rule.dst.addr);
1599 		pf_tbladdr_copyout(&pr->rule.src.addr);
1600 		pf_tbladdr_copyout(&pr->rule.dst.addr);
1601 		pf_rtlabel_copyout(&pr->rule.src.addr);
1602 		pf_rtlabel_copyout(&pr->rule.dst.addr);
1603 		for (i = 0; i < PF_SKIP_COUNT; ++i)
1604 			if (rule->skip[i].ptr == NULL)
1605 				pr->rule.skip[i].nr = -1;
1606 			else
1607 				pr->rule.skip[i].nr =
1608 				    rule->skip[i].ptr->nr;
1609 
1610 		if (pr->action == PF_GET_CLR_CNTR) {
1611 			rule->evaluations = 0;
1612 			rule->packets[0] = rule->packets[1] = 0;
1613 			rule->bytes[0] = rule->bytes[1] = 0;
1614 		}
1615 		break;
1616 	}
1617 
1618 	case DIOCCHANGERULE: {
1619 		struct pfioc_rule	*pcr = (struct pfioc_rule *)addr;
1620 		struct pf_ruleset	*ruleset;
1621 		struct pf_rule		*oldrule = NULL, *newrule = NULL;
1622 		u_int32_t		 nr = 0;
1623 		int			 rs_num;
1624 
1625 		if (!(pcr->action == PF_CHANGE_REMOVE ||
1626 		    pcr->action == PF_CHANGE_GET_TICKET) &&
1627 		    pcr->pool_ticket != ticket_pabuf) {
1628 			error = EBUSY;
1629 			break;
1630 		}
1631 
1632 		if (pcr->action < PF_CHANGE_ADD_HEAD ||
1633 		    pcr->action > PF_CHANGE_GET_TICKET) {
1634 			error = EINVAL;
1635 			break;
1636 		}
1637 		ruleset = pf_find_ruleset(pcr->anchor);
1638 		if (ruleset == NULL) {
1639 			error = EINVAL;
1640 			break;
1641 		}
1642 		rs_num = pf_get_ruleset_number(pcr->rule.action);
1643 		if (rs_num >= PF_RULESET_MAX) {
1644 			error = EINVAL;
1645 			break;
1646 		}
1647 
1648 		if (pcr->action == PF_CHANGE_GET_TICKET) {
1649 			pcr->ticket = ++ruleset->rules[rs_num].active.ticket;
1650 			break;
1651 		} else {
1652 			if (pcr->ticket !=
1653 			    ruleset->rules[rs_num].active.ticket) {
1654 				error = EINVAL;
1655 				break;
1656 			}
1657 			if (pcr->rule.return_icmp >> 8 > ICMP_MAXTYPE) {
1658 				error = EINVAL;
1659 				break;
1660 			}
1661 		}
1662 
1663 		if (pcr->action != PF_CHANGE_REMOVE) {
1664 			newrule = pool_get(&pf_rule_pl, PR_NOWAIT);
1665 			if (newrule == NULL) {
1666 				error = ENOMEM;
1667 				break;
1668 			}
1669 			bcopy(&pcr->rule, newrule, sizeof(struct pf_rule));
1670 #ifdef __NetBSD__
1671 			newrule->cuid = kauth_cred_getuid(l->l_cred);
1672 			newrule->cpid = l->l_proc->p_pid;
1673 #else
1674 			newrule->cuid = p->p_cred->p_ruid;
1675 			newrule->cpid = p->p_pid;
1676 #endif /* !__NetBSD__ */
1677 			TAILQ_INIT(&newrule->rpool.list);
1678 			/* initialize refcounting */
1679 			newrule->states = 0;
1680 			newrule->entries.tqe_prev = NULL;
1681 #ifndef INET
1682 			if (newrule->af == AF_INET) {
1683 				pool_put(&pf_rule_pl, newrule);
1684 				error = EAFNOSUPPORT;
1685 				break;
1686 			}
1687 #endif /* INET */
1688 #ifndef INET6
1689 			if (newrule->af == AF_INET6) {
1690 				pool_put(&pf_rule_pl, newrule);
1691 				error = EAFNOSUPPORT;
1692 				break;
1693 			}
1694 #endif /* INET6 */
1695 			if (newrule->ifname[0]) {
1696 				newrule->kif = pfi_kif_get(newrule->ifname);
1697 				if (newrule->kif == NULL) {
1698 					pool_put(&pf_rule_pl, newrule);
1699 					error = EINVAL;
1700 					break;
1701 				}
1702 				pfi_kif_ref(newrule->kif, PFI_KIF_REF_RULE);
1703 			} else
1704 				newrule->kif = NULL;
1705 
1706 #ifndef __NetBSD__
1707 			if (newrule->rtableid > 0 &&
1708 			    !rtable_exists(newrule->rtableid))
1709 				error = EBUSY;
1710 #endif /* !__NetBSD__ */
1711 
1712 #ifdef ALTQ
1713 			/* set queue IDs */
1714 			if (newrule->qname[0] != 0) {
1715 				if ((newrule->qid =
1716 				    pf_qname2qid(newrule->qname)) == 0)
1717 					error = EBUSY;
1718 				else if (newrule->pqname[0] != 0) {
1719 					if ((newrule->pqid =
1720 					    pf_qname2qid(newrule->pqname)) == 0)
1721 						error = EBUSY;
1722 				} else
1723 					newrule->pqid = newrule->qid;
1724 			}
1725 #endif /* ALTQ */
1726 			if (newrule->tagname[0])
1727 				if ((newrule->tag =
1728 				    pf_tagname2tag(newrule->tagname)) == 0)
1729 					error = EBUSY;
1730 			if (newrule->match_tagname[0])
1731 				if ((newrule->match_tag = pf_tagname2tag(
1732 				    newrule->match_tagname)) == 0)
1733 					error = EBUSY;
1734 			if (newrule->rt && !newrule->direction)
1735 				error = EINVAL;
1736 #if NPFLOG > 0
1737 			if (!newrule->log)
1738 				newrule->logif = 0;
1739 			if (newrule->logif >= PFLOGIFS_MAX)
1740 				error = EINVAL;
1741 #endif
1742 			if (pf_rtlabel_add(&newrule->src.addr) ||
1743 			    pf_rtlabel_add(&newrule->dst.addr))
1744 				error = EBUSY;
1745 			if (pfi_dynaddr_setup(&newrule->src.addr, newrule->af))
1746 				error = EINVAL;
1747 			if (pfi_dynaddr_setup(&newrule->dst.addr, newrule->af))
1748 				error = EINVAL;
1749 			if (pf_tbladdr_setup(ruleset, &newrule->src.addr))
1750 				error = EINVAL;
1751 			if (pf_tbladdr_setup(ruleset, &newrule->dst.addr))
1752 				error = EINVAL;
1753 			if (pf_anchor_setup(newrule, ruleset, pcr->anchor_call))
1754 				error = EINVAL;
1755 			TAILQ_FOREACH(pa, &pf_pabuf, entries)
1756 				if (pf_tbladdr_setup(ruleset, &pa->addr))
1757 					error = EINVAL;
1758 
1759 			newrule->overload_tbl = NULL;
1760 			if (newrule->overload_tblname[0]) {
1761 				if ((newrule->overload_tbl = pfr_attach_table(
1762 				    ruleset, newrule->overload_tblname)) ==
1763 				    NULL)
1764 					error = EINVAL;
1765 				else
1766 					newrule->overload_tbl->pfrkt_flags |=
1767 					    PFR_TFLAG_ACTIVE;
1768 			}
1769 
1770 			pf_mv_pool(&pf_pabuf, &newrule->rpool.list);
1771 			if (((((newrule->action == PF_NAT) ||
1772 			    (newrule->action == PF_RDR) ||
1773 			    (newrule->action == PF_BINAT) ||
1774 			    (newrule->rt > PF_FASTROUTE)) &&
1775 			    !newrule->anchor)) &&
1776 			    (TAILQ_FIRST(&newrule->rpool.list) == NULL))
1777 				error = EINVAL;
1778 
1779 			if (error) {
1780 				pf_rm_rule(NULL, newrule);
1781 				break;
1782 			}
1783 			newrule->rpool.cur = TAILQ_FIRST(&newrule->rpool.list);
1784 			newrule->evaluations = 0;
1785 			newrule->packets[0] = newrule->packets[1] = 0;
1786 			newrule->bytes[0] = newrule->bytes[1] = 0;
1787 		}
1788 		pf_empty_pool(&pf_pabuf);
1789 
1790 		if (pcr->action == PF_CHANGE_ADD_HEAD)
1791 			oldrule = TAILQ_FIRST(
1792 			    ruleset->rules[rs_num].active.ptr);
1793 		else if (pcr->action == PF_CHANGE_ADD_TAIL)
1794 			oldrule = TAILQ_LAST(
1795 			    ruleset->rules[rs_num].active.ptr, pf_rulequeue);
1796 		else {
1797 			oldrule = TAILQ_FIRST(
1798 			    ruleset->rules[rs_num].active.ptr);
1799 			while ((oldrule != NULL) && (oldrule->nr != pcr->nr))
1800 				oldrule = TAILQ_NEXT(oldrule, entries);
1801 			if (oldrule == NULL) {
1802 				if (newrule != NULL)
1803 					pf_rm_rule(NULL, newrule);
1804 				error = EINVAL;
1805 				break;
1806 			}
1807 		}
1808 
1809 		if (pcr->action == PF_CHANGE_REMOVE) {
1810 			pf_rm_rule(ruleset->rules[rs_num].active.ptr, oldrule);
1811 			ruleset->rules[rs_num].active.rcount--;
1812 		} else {
1813 			if (oldrule == NULL)
1814 				TAILQ_INSERT_TAIL(
1815 				    ruleset->rules[rs_num].active.ptr,
1816 				    newrule, entries);
1817 			else if (pcr->action == PF_CHANGE_ADD_HEAD ||
1818 			    pcr->action == PF_CHANGE_ADD_BEFORE)
1819 				TAILQ_INSERT_BEFORE(oldrule, newrule, entries);
1820 			else
1821 				TAILQ_INSERT_AFTER(
1822 				    ruleset->rules[rs_num].active.ptr,
1823 				    oldrule, newrule, entries);
1824 			ruleset->rules[rs_num].active.rcount++;
1825 		}
1826 
1827 		nr = 0;
1828 		TAILQ_FOREACH(oldrule,
1829 		    ruleset->rules[rs_num].active.ptr, entries)
1830 			oldrule->nr = nr++;
1831 
1832 		ruleset->rules[rs_num].active.ticket++;
1833 
1834 		pf_calc_skip_steps(ruleset->rules[rs_num].active.ptr);
1835 		pf_remove_if_empty_ruleset(ruleset);
1836 
1837 		break;
1838 	}
1839 
1840 	case DIOCCLRSTATES: {
1841 		struct pf_state		*ps, *nexts;
1842 		struct pfioc_state_kill *psk = (struct pfioc_state_kill *)addr;
1843 		int			 killed = 0;
1844 
1845 		for (ps = RB_MIN(pf_state_tree_id, &tree_id); ps; ps = nexts) {
1846 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, ps);
1847 
1848 			if (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1849 			    ps->kif->pfik_name)) {
1850 #if NPFSYNC
1851 				/* don't send out individual delete messages */
1852 				ps->sync_flags = PFSTATE_NOSYNC;
1853 #endif
1854 				pf_unlink_state(ps);
1855 				killed++;
1856 			}
1857 		}
1858 		psk->psk_af = killed;
1859 #if NPFSYNC
1860 		pfsync_clear_states(pf_status.hostid, psk->psk_ifname);
1861 #endif
1862 		break;
1863 	}
1864 
1865 	case DIOCKILLSTATES: {
1866 		struct pf_state		*ps, *nexts;
1867 		struct pf_state_key	*sk;
1868 		struct pf_state_host	*src, *dst;
1869 		struct pfioc_state_kill	*psk = (struct pfioc_state_kill *)addr;
1870 		int			 killed = 0;
1871 
1872 		for (ps = RB_MIN(pf_state_tree_id, &tree_id); ps;
1873 		    ps = nexts) {
1874 			nexts = RB_NEXT(pf_state_tree_id, &tree_id, ps);
1875 			sk = ps->state_key;
1876 
1877 			if (sk->direction == PF_OUT) {
1878 				src = &sk->lan;
1879 				dst = &sk->ext;
1880 			} else {
1881 				src = &sk->ext;
1882 				dst = &sk->lan;
1883 			}
1884 			if ((!psk->psk_af || sk->af == psk->psk_af)
1885 			    && (!psk->psk_proto || psk->psk_proto ==
1886 			    sk->proto) &&
1887 			    PF_MATCHA(psk->psk_src.neg,
1888 			    &psk->psk_src.addr.v.a.addr,
1889 			    &psk->psk_src.addr.v.a.mask,
1890 			    &src->addr, sk->af) &&
1891 			    PF_MATCHA(psk->psk_dst.neg,
1892 			    &psk->psk_dst.addr.v.a.addr,
1893 			    &psk->psk_dst.addr.v.a.mask,
1894 			    &dst->addr, sk->af) &&
1895 			    (psk->psk_src.port_op == 0 ||
1896 			    pf_match_port(psk->psk_src.port_op,
1897 			    psk->psk_src.port[0], psk->psk_src.port[1],
1898 			    src->port)) &&
1899 			    (psk->psk_dst.port_op == 0 ||
1900 			    pf_match_port(psk->psk_dst.port_op,
1901 			    psk->psk_dst.port[0], psk->psk_dst.port[1],
1902 			    dst->port)) &&
1903 			    (!psk->psk_ifname[0] || !strcmp(psk->psk_ifname,
1904 			    ps->kif->pfik_name))) {
1905 #if NPFSYNC > 0
1906 				/* send immediate delete of state */
1907 				pfsync_delete_state(ps);
1908 				ps->sync_flags |= PFSTATE_NOSYNC;
1909 #endif
1910 				pf_unlink_state(ps);
1911 				killed++;
1912 			}
1913 		}
1914 		psk->psk_af = killed;
1915 		break;
1916 	}
1917 
1918 	case DIOCADDSTATE: {
1919 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1920 		struct pfsync_state 	*sp = (struct pfsync_state *)ps->state;
1921 
1922 		error = pf_state_add(sp);
1923 		break;
1924 	}
1925 
1926 	case DIOCADDSTATES: {
1927 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1928 		struct pfsync_state	*p = (struct pfsync_state *) ps->ps_states;
1929 		struct pfsync_state *pk;
1930 		int size = ps->ps_len;
1931 		int i = 0;
1932 		error = 0;
1933 
1934 		pk = malloc(sizeof(*pk), M_TEMP,M_WAITOK);
1935 
1936 		while (error == 0 && i < size)
1937 		{
1938 			if (copyin(p, pk, sizeof(struct pfsync_state)))
1939 			{
1940 				error = EFAULT;
1941 				free(pk, M_TEMP);
1942 			} else {
1943 				error = pf_state_add(pk);
1944 				i += sizeof(*p);
1945 				p++;
1946 			}
1947 		}
1948 
1949 		free(pk, M_TEMP);
1950 		break;
1951 	}
1952 
1953 
1954 	case DIOCGETSTATE: {
1955 		struct pfioc_state	*ps = (struct pfioc_state *)addr;
1956 		struct pf_state		*pfs;
1957 		u_int32_t		 nr;
1958 
1959 		nr = 0;
1960 		RB_FOREACH(pfs, pf_state_tree_id, &tree_id) {
1961 			if (nr >= ps->nr)
1962 				break;
1963 			nr++;
1964 		}
1965 		if (pfs == NULL) {
1966 			error = EBUSY;
1967 			break;
1968 		}
1969 
1970 		pf_state_export((struct pfsync_state *)&ps->state,
1971 		    pfs->state_key, pfs);
1972 		break;
1973 	}
1974 
1975 	case DIOCGETSTATES: {
1976 		struct pfioc_states	*ps = (struct pfioc_states *)addr;
1977 		struct pf_state		*state;
1978 		struct pfsync_state	*p, *pstore;
1979 		u_int32_t		 nr = 0;
1980 
1981 		if (ps->ps_len == 0) {
1982 			nr = pf_status.states;
1983 			ps->ps_len = sizeof(struct pfsync_state) * nr;
1984 			break;
1985 		}
1986 
1987 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
1988 
1989 		p = ps->ps_states;
1990 
1991 		state = TAILQ_FIRST(&state_list);
1992 		while (state) {
1993 			if (state->timeout != PFTM_UNLINKED) {
1994 				if ((nr+1) * sizeof(*p) > (unsigned)ps->ps_len)
1995 					break;
1996 
1997 				pf_state_export(pstore,
1998 				    state->state_key, state);
1999 				error = copyout(pstore, p, sizeof(*p));
2000 				if (error) {
2001 					free(pstore, M_TEMP);
2002 					goto fail;
2003 				}
2004 				p++;
2005 				nr++;
2006 			}
2007 			state = TAILQ_NEXT(state, entry_list);
2008 		}
2009 
2010 		ps->ps_len = sizeof(struct pfsync_state) * nr;
2011 
2012 		free(pstore, M_TEMP);
2013 		break;
2014 	}
2015 
2016 	case DIOCGETSTATUS: {
2017 		struct pf_status *ps = (struct pf_status *)addr;
2018 		bcopy(&pf_status, ps, sizeof(struct pf_status));
2019 		pfi_fill_oldstatus(ps);
2020 		break;
2021 	}
2022 
2023 	case DIOCSETSTATUSIF: {
2024 		struct pfioc_if	*pi = (struct pfioc_if *)addr;
2025 
2026 		if (pi->ifname[0] == 0) {
2027 			bzero(pf_status.ifname, IFNAMSIZ);
2028 			break;
2029 		}
2030 		if (ifunit(pi->ifname) == NULL) {
2031 			error = EINVAL;
2032 			break;
2033 		}
2034 		strlcpy(pf_status.ifname, pi->ifname, IFNAMSIZ);
2035 		break;
2036 	}
2037 
2038 	case DIOCCLRSTATUS: {
2039 		bzero(pf_status.counters, sizeof(pf_status.counters));
2040 		bzero(pf_status.fcounters, sizeof(pf_status.fcounters));
2041 		bzero(pf_status.scounters, sizeof(pf_status.scounters));
2042 		pf_status.since = time_second;
2043 		if (*pf_status.ifname)
2044 			pfi_clr_istats(pf_status.ifname);
2045 		break;
2046 	}
2047 
2048 	case DIOCNATLOOK: {
2049 		struct pfioc_natlook	*pnl = (struct pfioc_natlook *)addr;
2050 		struct pf_state_key	*sk;
2051 		struct pf_state		*state;
2052 		struct pf_state_key_cmp	 key;
2053 		int			 m = 0, direction = pnl->direction;
2054 
2055 		key.af = pnl->af;
2056 		key.proto = pnl->proto;
2057 
2058 		if (!pnl->proto ||
2059 		    PF_AZERO(&pnl->saddr, pnl->af) ||
2060 		    PF_AZERO(&pnl->daddr, pnl->af) ||
2061 		    ((pnl->proto == IPPROTO_TCP ||
2062 		    pnl->proto == IPPROTO_UDP) &&
2063 		    (!pnl->dport || !pnl->sport)))
2064 			error = EINVAL;
2065 		else {
2066 			/*
2067 			 * userland gives us source and dest of connection,
2068 			 * reverse the lookup so we ask for what happens with
2069 			 * the return traffic, enabling us to find it in the
2070 			 * state tree.
2071 			 */
2072 			if (direction == PF_IN) {
2073 				PF_ACPY(&key.ext.addr, &pnl->daddr, pnl->af);
2074 				key.ext.port = pnl->dport;
2075 				PF_ACPY(&key.gwy.addr, &pnl->saddr, pnl->af);
2076 				key.gwy.port = pnl->sport;
2077 				state = pf_find_state_all(&key, PF_EXT_GWY, &m);
2078 			} else {
2079 				PF_ACPY(&key.lan.addr, &pnl->daddr, pnl->af);
2080 				key.lan.port = pnl->dport;
2081 				PF_ACPY(&key.ext.addr, &pnl->saddr, pnl->af);
2082 				key.ext.port = pnl->sport;
2083 				state = pf_find_state_all(&key, PF_LAN_EXT, &m);
2084 			}
2085 			if (m > 1)
2086 				error = E2BIG;	/* more than one state */
2087 			else if (state != NULL) {
2088 				sk = state->state_key;
2089 				if (direction == PF_IN) {
2090 					PF_ACPY(&pnl->rsaddr, &sk->lan.addr,
2091 					    sk->af);
2092 					pnl->rsport = sk->lan.port;
2093 					PF_ACPY(&pnl->rdaddr, &pnl->daddr,
2094 					    pnl->af);
2095 					pnl->rdport = pnl->dport;
2096 				} else {
2097 					PF_ACPY(&pnl->rdaddr, &sk->gwy.addr,
2098 					    sk->af);
2099 					pnl->rdport = sk->gwy.port;
2100 					PF_ACPY(&pnl->rsaddr, &pnl->saddr,
2101 					    pnl->af);
2102 					pnl->rsport = pnl->sport;
2103 				}
2104 			} else
2105 				error = ENOENT;
2106 		}
2107 		break;
2108 	}
2109 
2110 	case DIOCSETTIMEOUT: {
2111 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2112 		int		 old;
2113 
2114 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX ||
2115 		    pt->seconds < 0) {
2116 			error = EINVAL;
2117 			goto fail;
2118 		}
2119 		old = pf_default_rule.timeout[pt->timeout];
2120 		if (pt->timeout == PFTM_INTERVAL && pt->seconds == 0)
2121 			pt->seconds = 1;
2122 		pf_default_rule.timeout[pt->timeout] = pt->seconds;
2123 		if (pt->timeout == PFTM_INTERVAL && pt->seconds < old)
2124 			wakeup(pf_purge_thread);
2125 		pt->seconds = old;
2126 		break;
2127 	}
2128 
2129 	case DIOCGETTIMEOUT: {
2130 		struct pfioc_tm	*pt = (struct pfioc_tm *)addr;
2131 
2132 		if (pt->timeout < 0 || pt->timeout >= PFTM_MAX) {
2133 			error = EINVAL;
2134 			goto fail;
2135 		}
2136 		pt->seconds = pf_default_rule.timeout[pt->timeout];
2137 		break;
2138 	}
2139 
2140 	case DIOCGETLIMIT: {
2141 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2142 
2143 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX) {
2144 			error = EINVAL;
2145 			goto fail;
2146 		}
2147 		pl->limit = pf_pool_limits[pl->index].limit;
2148 		break;
2149 	}
2150 
2151 	case DIOCSETLIMIT: {
2152 		struct pfioc_limit	*pl = (struct pfioc_limit *)addr;
2153 		int			 old_limit;
2154 
2155 		if (pl->index < 0 || pl->index >= PF_LIMIT_MAX ||
2156 		    pf_pool_limits[pl->index].pp == NULL) {
2157 			error = EINVAL;
2158 			goto fail;
2159 		}
2160 #ifdef __NetBSD__
2161 		pool_sethardlimit(pf_pool_limits[pl->index].pp,
2162 		    pl->limit, NULL, 0);
2163 #else
2164 		if (pool_sethardlimit(pf_pool_limits[pl->index].pp,
2165 		    pl->limit, NULL, 0) != 0) {
2166 			error = EBUSY;
2167 			goto fail;
2168 		}
2169 #endif /* !__NetBSD__ */
2170 		old_limit = pf_pool_limits[pl->index].limit;
2171 		pf_pool_limits[pl->index].limit = pl->limit;
2172 		pl->limit = old_limit;
2173 		break;
2174 	}
2175 
2176 	case DIOCSETDEBUG: {
2177 		u_int32_t	*level = (u_int32_t *)addr;
2178 
2179 		pf_status.debug = *level;
2180 		break;
2181 	}
2182 
2183 	case DIOCCLRRULECTRS: {
2184 		/* obsoleted by DIOCGETRULE with action=PF_GET_CLR_CNTR */
2185 		struct pf_ruleset	*ruleset = &pf_main_ruleset;
2186 		struct pf_rule		*rule;
2187 
2188 		TAILQ_FOREACH(rule,
2189 		    ruleset->rules[PF_RULESET_FILTER].active.ptr, entries) {
2190 			rule->evaluations = 0;
2191 			rule->packets[0] = rule->packets[1] = 0;
2192 			rule->bytes[0] = rule->bytes[1] = 0;
2193 		}
2194 		break;
2195 	}
2196 
2197 #ifdef ALTQ
2198 	case DIOCSTARTALTQ: {
2199 		struct pf_altq		*altq;
2200 
2201 		/* enable all altq interfaces on active list */
2202 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2203 			if (altq->qname[0] == 0) {
2204 				error = pf_enable_altq(altq);
2205 				if (error != 0)
2206 					break;
2207 			}
2208 		}
2209 		if (error == 0)
2210 			pf_altq_running = 1;
2211 		DPFPRINTF(PF_DEBUG_MISC, ("altq: started\n"));
2212 		break;
2213 	}
2214 
2215 	case DIOCSTOPALTQ: {
2216 		struct pf_altq		*altq;
2217 
2218 		/* disable all altq interfaces on active list */
2219 		TAILQ_FOREACH(altq, pf_altqs_active, entries) {
2220 			if (altq->qname[0] == 0) {
2221 				error = pf_disable_altq(altq);
2222 				if (error != 0)
2223 					break;
2224 			}
2225 		}
2226 		if (error == 0)
2227 			pf_altq_running = 0;
2228 		DPFPRINTF(PF_DEBUG_MISC, ("altq: stopped\n"));
2229 		break;
2230 	}
2231 
2232 	case DIOCADDALTQ: {
2233 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2234 		struct pf_altq		*altq, *a;
2235 
2236 		if (paa->ticket != ticket_altqs_inactive) {
2237 			error = EBUSY;
2238 			break;
2239 		}
2240 		altq = pool_get(&pf_altq_pl, PR_NOWAIT);
2241 		if (altq == NULL) {
2242 			error = ENOMEM;
2243 			break;
2244 		}
2245 		bcopy(&paa->altq, altq, sizeof(struct pf_altq));
2246 
2247 		/*
2248 		 * if this is for a queue, find the discipline and
2249 		 * copy the necessary fields
2250 		 */
2251 		if (altq->qname[0] != 0) {
2252 			if ((altq->qid = pf_qname2qid(altq->qname)) == 0) {
2253 				error = EBUSY;
2254 				pool_put(&pf_altq_pl, altq);
2255 				break;
2256 			}
2257 			TAILQ_FOREACH(a, pf_altqs_inactive, entries) {
2258 				if (strncmp(a->ifname, altq->ifname,
2259 				    IFNAMSIZ) == 0 && a->qname[0] == 0) {
2260 					altq->altq_disc = a->altq_disc;
2261 					break;
2262 				}
2263 			}
2264 		}
2265 
2266 		error = altq_add(altq);
2267 		if (error) {
2268 			pool_put(&pf_altq_pl, altq);
2269 			break;
2270 		}
2271 
2272 		TAILQ_INSERT_TAIL(pf_altqs_inactive, altq, entries);
2273 		bcopy(altq, &paa->altq, sizeof(struct pf_altq));
2274 		break;
2275 	}
2276 
2277 	case DIOCGETALTQS: {
2278 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2279 		struct pf_altq		*altq;
2280 
2281 		paa->nr = 0;
2282 		TAILQ_FOREACH(altq, pf_altqs_active, entries)
2283 			paa->nr++;
2284 		paa->ticket = ticket_altqs_active;
2285 		break;
2286 	}
2287 
2288 	case DIOCGETALTQ: {
2289 		struct pfioc_altq	*paa = (struct pfioc_altq *)addr;
2290 		struct pf_altq		*altq;
2291 		u_int32_t		 nr;
2292 
2293 		if (paa->ticket != ticket_altqs_active) {
2294 			error = EBUSY;
2295 			break;
2296 		}
2297 		nr = 0;
2298 		altq = TAILQ_FIRST(pf_altqs_active);
2299 		while ((altq != NULL) && (nr < paa->nr)) {
2300 			altq = TAILQ_NEXT(altq, entries);
2301 			nr++;
2302 		}
2303 		if (altq == NULL) {
2304 			error = EBUSY;
2305 			break;
2306 		}
2307 		bcopy(altq, &paa->altq, sizeof(struct pf_altq));
2308 		break;
2309 	}
2310 
2311 	case DIOCCHANGEALTQ:
2312 		/* CHANGEALTQ not supported yet! */
2313 		error = ENODEV;
2314 		break;
2315 
2316 	case DIOCGETQSTATS: {
2317 		struct pfioc_qstats	*pq = (struct pfioc_qstats *)addr;
2318 		struct pf_altq		*altq;
2319 		u_int32_t		 nr;
2320 		int			 nbytes;
2321 
2322 		if (pq->ticket != ticket_altqs_active) {
2323 			error = EBUSY;
2324 			break;
2325 		}
2326 		nbytes = pq->nbytes;
2327 		nr = 0;
2328 		altq = TAILQ_FIRST(pf_altqs_active);
2329 		while ((altq != NULL) && (nr < pq->nr)) {
2330 			altq = TAILQ_NEXT(altq, entries);
2331 			nr++;
2332 		}
2333 		if (altq == NULL) {
2334 			error = EBUSY;
2335 			break;
2336 		}
2337 		error = altq_getqstats(altq, pq->buf, &nbytes);
2338 		if (error == 0) {
2339 			pq->scheduler = altq->scheduler;
2340 			pq->nbytes = nbytes;
2341 		}
2342 		break;
2343 	}
2344 #endif /* ALTQ */
2345 
2346 	case DIOCBEGINADDRS: {
2347 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2348 
2349 		pf_empty_pool(&pf_pabuf);
2350 		pp->ticket = ++ticket_pabuf;
2351 		break;
2352 	}
2353 
2354 	case DIOCADDADDR: {
2355 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2356 
2357 		if (pp->ticket != ticket_pabuf) {
2358 			error = EBUSY;
2359 			break;
2360 		}
2361 #ifndef INET
2362 		if (pp->af == AF_INET) {
2363 			error = EAFNOSUPPORT;
2364 			break;
2365 		}
2366 #endif /* INET */
2367 #ifndef INET6
2368 		if (pp->af == AF_INET6) {
2369 			error = EAFNOSUPPORT;
2370 			break;
2371 		}
2372 #endif /* INET6 */
2373 		if (pp->addr.addr.type != PF_ADDR_ADDRMASK &&
2374 		    pp->addr.addr.type != PF_ADDR_DYNIFTL &&
2375 		    pp->addr.addr.type != PF_ADDR_TABLE) {
2376 			error = EINVAL;
2377 			break;
2378 		}
2379 		pa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2380 		if (pa == NULL) {
2381 			error = ENOMEM;
2382 			break;
2383 		}
2384 		bcopy(&pp->addr, pa, sizeof(struct pf_pooladdr));
2385 		if (pa->ifname[0]) {
2386 			pa->kif = pfi_kif_get(pa->ifname);
2387 			if (pa->kif == NULL) {
2388 				pool_put(&pf_pooladdr_pl, pa);
2389 				error = EINVAL;
2390 				break;
2391 			}
2392 			pfi_kif_ref(pa->kif, PFI_KIF_REF_RULE);
2393 		}
2394 		if (pfi_dynaddr_setup(&pa->addr, pp->af)) {
2395 			pfi_dynaddr_remove(&pa->addr);
2396 			pfi_kif_unref(pa->kif, PFI_KIF_REF_RULE);
2397 			pool_put(&pf_pooladdr_pl, pa);
2398 			error = EINVAL;
2399 			break;
2400 		}
2401 		TAILQ_INSERT_TAIL(&pf_pabuf, pa, entries);
2402 		break;
2403 	}
2404 
2405 	case DIOCGETADDRS: {
2406 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2407 
2408 		pp->nr = 0;
2409 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2410 		    pp->r_num, 0, 1, 0);
2411 		if (pool == NULL) {
2412 			error = EBUSY;
2413 			break;
2414 		}
2415 		TAILQ_FOREACH(pa, &pool->list, entries)
2416 			pp->nr++;
2417 		break;
2418 	}
2419 
2420 	case DIOCGETADDR: {
2421 		struct pfioc_pooladdr	*pp = (struct pfioc_pooladdr *)addr;
2422 		u_int32_t		 nr = 0;
2423 
2424 		pool = pf_get_pool(pp->anchor, pp->ticket, pp->r_action,
2425 		    pp->r_num, 0, 1, 1);
2426 		if (pool == NULL) {
2427 			error = EBUSY;
2428 			break;
2429 		}
2430 		pa = TAILQ_FIRST(&pool->list);
2431 		while ((pa != NULL) && (nr < pp->nr)) {
2432 			pa = TAILQ_NEXT(pa, entries);
2433 			nr++;
2434 		}
2435 		if (pa == NULL) {
2436 			error = EBUSY;
2437 			break;
2438 		}
2439 		bcopy(pa, &pp->addr, sizeof(struct pf_pooladdr));
2440 		pfi_dynaddr_copyout(&pp->addr.addr);
2441 		pf_tbladdr_copyout(&pp->addr.addr);
2442 		pf_rtlabel_copyout(&pp->addr.addr);
2443 		break;
2444 	}
2445 
2446 	case DIOCCHANGEADDR: {
2447 		struct pfioc_pooladdr	*pca = (struct pfioc_pooladdr *)addr;
2448 		struct pf_pooladdr	*oldpa = NULL, *newpa = NULL;
2449 		struct pf_ruleset	*ruleset;
2450 
2451 		if (pca->action < PF_CHANGE_ADD_HEAD ||
2452 		    pca->action > PF_CHANGE_REMOVE) {
2453 			error = EINVAL;
2454 			break;
2455 		}
2456 		if (pca->addr.addr.type != PF_ADDR_ADDRMASK &&
2457 		    pca->addr.addr.type != PF_ADDR_DYNIFTL &&
2458 		    pca->addr.addr.type != PF_ADDR_TABLE) {
2459 			error = EINVAL;
2460 			break;
2461 		}
2462 
2463 		ruleset = pf_find_ruleset(pca->anchor);
2464 		if (ruleset == NULL) {
2465 			error = EBUSY;
2466 			break;
2467 		}
2468 		pool = pf_get_pool(pca->anchor, pca->ticket, pca->r_action,
2469 		    pca->r_num, pca->r_last, 1, 1);
2470 		if (pool == NULL) {
2471 			error = EBUSY;
2472 			break;
2473 		}
2474 		if (pca->action != PF_CHANGE_REMOVE) {
2475 			newpa = pool_get(&pf_pooladdr_pl, PR_NOWAIT);
2476 			if (newpa == NULL) {
2477 				error = ENOMEM;
2478 				break;
2479 			}
2480 			bcopy(&pca->addr, newpa, sizeof(struct pf_pooladdr));
2481 #ifndef INET
2482 			if (pca->af == AF_INET) {
2483 				pool_put(&pf_pooladdr_pl, newpa);
2484 				error = EAFNOSUPPORT;
2485 				break;
2486 			}
2487 #endif /* INET */
2488 #ifndef INET6
2489 			if (pca->af == AF_INET6) {
2490 				pool_put(&pf_pooladdr_pl, newpa);
2491 				error = EAFNOSUPPORT;
2492 				break;
2493 			}
2494 #endif /* INET6 */
2495 			if (newpa->ifname[0]) {
2496 				newpa->kif = pfi_kif_get(newpa->ifname);
2497 				if (newpa->kif == NULL) {
2498 					pool_put(&pf_pooladdr_pl, newpa);
2499 					error = EINVAL;
2500 					break;
2501 				}
2502 				pfi_kif_ref(newpa->kif, PFI_KIF_REF_RULE);
2503 			} else
2504 				newpa->kif = NULL;
2505 			if (pfi_dynaddr_setup(&newpa->addr, pca->af) ||
2506 			    pf_tbladdr_setup(ruleset, &newpa->addr)) {
2507 				pfi_dynaddr_remove(&newpa->addr);
2508 				pfi_kif_unref(newpa->kif, PFI_KIF_REF_RULE);
2509 				pool_put(&pf_pooladdr_pl, newpa);
2510 				error = EINVAL;
2511 				break;
2512 			}
2513 		}
2514 
2515 		if (pca->action == PF_CHANGE_ADD_HEAD)
2516 			oldpa = TAILQ_FIRST(&pool->list);
2517 		else if (pca->action == PF_CHANGE_ADD_TAIL)
2518 			oldpa = TAILQ_LAST(&pool->list, pf_palist);
2519 		else {
2520 			int	i = 0;
2521 
2522 			oldpa = TAILQ_FIRST(&pool->list);
2523 			while ((oldpa != NULL) && (i < pca->nr)) {
2524 				oldpa = TAILQ_NEXT(oldpa, entries);
2525 				i++;
2526 			}
2527 			if (oldpa == NULL) {
2528 				error = EINVAL;
2529 				break;
2530 			}
2531 		}
2532 
2533 		if (pca->action == PF_CHANGE_REMOVE) {
2534 			TAILQ_REMOVE(&pool->list, oldpa, entries);
2535 			pfi_dynaddr_remove(&oldpa->addr);
2536 			pf_tbladdr_remove(&oldpa->addr);
2537 			pfi_kif_unref(oldpa->kif, PFI_KIF_REF_RULE);
2538 			pool_put(&pf_pooladdr_pl, oldpa);
2539 		} else {
2540 			if (oldpa == NULL)
2541 				TAILQ_INSERT_TAIL(&pool->list, newpa, entries);
2542 			else if (pca->action == PF_CHANGE_ADD_HEAD ||
2543 			    pca->action == PF_CHANGE_ADD_BEFORE)
2544 				TAILQ_INSERT_BEFORE(oldpa, newpa, entries);
2545 			else
2546 				TAILQ_INSERT_AFTER(&pool->list, oldpa,
2547 				    newpa, entries);
2548 		}
2549 
2550 		pool->cur = TAILQ_FIRST(&pool->list);
2551 		PF_ACPY(&pool->counter, &pool->cur->addr.v.a.addr,
2552 		    pca->af);
2553 		break;
2554 	}
2555 
2556 	case DIOCGETRULESETS: {
2557 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2558 		struct pf_ruleset	*ruleset;
2559 		struct pf_anchor	*anchor;
2560 
2561 		pr->path[sizeof(pr->path) - 1] = 0;
2562 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2563 			error = EINVAL;
2564 			break;
2565 		}
2566 		pr->nr = 0;
2567 		if (ruleset->anchor == NULL) {
2568 			/* XXX kludge for pf_main_ruleset */
2569 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2570 				if (anchor->parent == NULL)
2571 					pr->nr++;
2572 		} else {
2573 			RB_FOREACH(anchor, pf_anchor_node,
2574 			    &ruleset->anchor->children)
2575 				pr->nr++;
2576 		}
2577 		break;
2578 	}
2579 
2580 	case DIOCGETRULESET: {
2581 		struct pfioc_ruleset	*pr = (struct pfioc_ruleset *)addr;
2582 		struct pf_ruleset	*ruleset;
2583 		struct pf_anchor	*anchor;
2584 		u_int32_t		 nr = 0;
2585 
2586 		pr->path[sizeof(pr->path) - 1] = 0;
2587 		if ((ruleset = pf_find_ruleset(pr->path)) == NULL) {
2588 			error = EINVAL;
2589 			break;
2590 		}
2591 		pr->name[0] = 0;
2592 		if (ruleset->anchor == NULL) {
2593 			/* XXX kludge for pf_main_ruleset */
2594 			RB_FOREACH(anchor, pf_anchor_global, &pf_anchors)
2595 				if (anchor->parent == NULL && nr++ == pr->nr) {
2596 					strlcpy(pr->name, anchor->name,
2597 					    sizeof(pr->name));
2598 					break;
2599 				}
2600 		} else {
2601 			RB_FOREACH(anchor, pf_anchor_node,
2602 			    &ruleset->anchor->children)
2603 				if (nr++ == pr->nr) {
2604 					strlcpy(pr->name, anchor->name,
2605 					    sizeof(pr->name));
2606 					break;
2607 				}
2608 		}
2609 		if (!pr->name[0])
2610 			error = EBUSY;
2611 		break;
2612 	}
2613 
2614 	case DIOCRCLRTABLES: {
2615 		struct pfioc_table *io = (struct pfioc_table *)addr;
2616 
2617 		if (io->pfrio_esize != 0) {
2618 			error = ENODEV;
2619 			break;
2620 		}
2621 		error = pfr_clr_tables(&io->pfrio_table, &io->pfrio_ndel,
2622 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2623 		break;
2624 	}
2625 
2626 	case DIOCRADDTABLES: {
2627 		struct pfioc_table *io = (struct pfioc_table *)addr;
2628 
2629 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2630 			error = ENODEV;
2631 			break;
2632 		}
2633 		error = pfr_add_tables(io->pfrio_buffer, io->pfrio_size,
2634 		    &io->pfrio_nadd, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2635 		break;
2636 	}
2637 
2638 	case DIOCRDELTABLES: {
2639 		struct pfioc_table *io = (struct pfioc_table *)addr;
2640 
2641 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2642 			error = ENODEV;
2643 			break;
2644 		}
2645 		error = pfr_del_tables(io->pfrio_buffer, io->pfrio_size,
2646 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2647 		break;
2648 	}
2649 
2650 	case DIOCRGETTABLES: {
2651 		struct pfioc_table *io = (struct pfioc_table *)addr;
2652 
2653 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2654 			error = ENODEV;
2655 			break;
2656 		}
2657 		error = pfr_get_tables(&io->pfrio_table, io->pfrio_buffer,
2658 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2659 		break;
2660 	}
2661 
2662 	case DIOCRGETTSTATS: {
2663 		struct pfioc_table *io = (struct pfioc_table *)addr;
2664 
2665 		if (io->pfrio_esize != sizeof(struct pfr_tstats)) {
2666 			error = ENODEV;
2667 			break;
2668 		}
2669 		error = pfr_get_tstats(&io->pfrio_table, io->pfrio_buffer,
2670 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2671 		break;
2672 	}
2673 
2674 	case DIOCRCLRTSTATS: {
2675 		struct pfioc_table *io = (struct pfioc_table *)addr;
2676 
2677 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2678 			error = ENODEV;
2679 			break;
2680 		}
2681 		error = pfr_clr_tstats(io->pfrio_buffer, io->pfrio_size,
2682 		    &io->pfrio_nzero, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2683 		break;
2684 	}
2685 
2686 	case DIOCRSETTFLAGS: {
2687 		struct pfioc_table *io = (struct pfioc_table *)addr;
2688 
2689 		if (io->pfrio_esize != sizeof(struct pfr_table)) {
2690 			error = ENODEV;
2691 			break;
2692 		}
2693 		error = pfr_set_tflags(io->pfrio_buffer, io->pfrio_size,
2694 		    io->pfrio_setflag, io->pfrio_clrflag, &io->pfrio_nchange,
2695 		    &io->pfrio_ndel, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2696 		break;
2697 	}
2698 
2699 	case DIOCRCLRADDRS: {
2700 		struct pfioc_table *io = (struct pfioc_table *)addr;
2701 
2702 		if (io->pfrio_esize != 0) {
2703 			error = ENODEV;
2704 			break;
2705 		}
2706 		error = pfr_clr_addrs(&io->pfrio_table, &io->pfrio_ndel,
2707 		    io->pfrio_flags | PFR_FLAG_USERIOCTL);
2708 		break;
2709 	}
2710 
2711 	case DIOCRADDADDRS: {
2712 		struct pfioc_table *io = (struct pfioc_table *)addr;
2713 
2714 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2715 			error = ENODEV;
2716 			break;
2717 		}
2718 		error = pfr_add_addrs(&io->pfrio_table, io->pfrio_buffer,
2719 		    io->pfrio_size, &io->pfrio_nadd, io->pfrio_flags |
2720 		    PFR_FLAG_USERIOCTL);
2721 		break;
2722 	}
2723 
2724 	case DIOCRDELADDRS: {
2725 		struct pfioc_table *io = (struct pfioc_table *)addr;
2726 
2727 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2728 			error = ENODEV;
2729 			break;
2730 		}
2731 		error = pfr_del_addrs(&io->pfrio_table, io->pfrio_buffer,
2732 		    io->pfrio_size, &io->pfrio_ndel, io->pfrio_flags |
2733 		    PFR_FLAG_USERIOCTL);
2734 		break;
2735 	}
2736 
2737 	case DIOCRSETADDRS: {
2738 		struct pfioc_table *io = (struct pfioc_table *)addr;
2739 
2740 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2741 			error = ENODEV;
2742 			break;
2743 		}
2744 		error = pfr_set_addrs(&io->pfrio_table, io->pfrio_buffer,
2745 		    io->pfrio_size, &io->pfrio_size2, &io->pfrio_nadd,
2746 		    &io->pfrio_ndel, &io->pfrio_nchange, io->pfrio_flags |
2747 		    PFR_FLAG_USERIOCTL, 0);
2748 		break;
2749 	}
2750 
2751 	case DIOCRGETADDRS: {
2752 		struct pfioc_table *io = (struct pfioc_table *)addr;
2753 
2754 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2755 			error = ENODEV;
2756 			break;
2757 		}
2758 		error = pfr_get_addrs(&io->pfrio_table, io->pfrio_buffer,
2759 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2760 		break;
2761 	}
2762 
2763 	case DIOCRGETASTATS: {
2764 		struct pfioc_table *io = (struct pfioc_table *)addr;
2765 
2766 		if (io->pfrio_esize != sizeof(struct pfr_astats)) {
2767 			error = ENODEV;
2768 			break;
2769 		}
2770 		error = pfr_get_astats(&io->pfrio_table, io->pfrio_buffer,
2771 		    &io->pfrio_size, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2772 		break;
2773 	}
2774 
2775 	case DIOCRCLRASTATS: {
2776 		struct pfioc_table *io = (struct pfioc_table *)addr;
2777 
2778 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2779 			error = ENODEV;
2780 			break;
2781 		}
2782 		error = pfr_clr_astats(&io->pfrio_table, io->pfrio_buffer,
2783 		    io->pfrio_size, &io->pfrio_nzero, io->pfrio_flags |
2784 		    PFR_FLAG_USERIOCTL);
2785 		break;
2786 	}
2787 
2788 	case DIOCRTSTADDRS: {
2789 		struct pfioc_table *io = (struct pfioc_table *)addr;
2790 
2791 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2792 			error = ENODEV;
2793 			break;
2794 		}
2795 		error = pfr_tst_addrs(&io->pfrio_table, io->pfrio_buffer,
2796 		    io->pfrio_size, &io->pfrio_nmatch, io->pfrio_flags |
2797 		    PFR_FLAG_USERIOCTL);
2798 		break;
2799 	}
2800 
2801 	case DIOCRINADEFINE: {
2802 		struct pfioc_table *io = (struct pfioc_table *)addr;
2803 
2804 		if (io->pfrio_esize != sizeof(struct pfr_addr)) {
2805 			error = ENODEV;
2806 			break;
2807 		}
2808 		error = pfr_ina_define(&io->pfrio_table, io->pfrio_buffer,
2809 		    io->pfrio_size, &io->pfrio_nadd, &io->pfrio_naddr,
2810 		    io->pfrio_ticket, io->pfrio_flags | PFR_FLAG_USERIOCTL);
2811 		break;
2812 	}
2813 
2814 	case DIOCOSFPADD: {
2815 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2816 		error = pf_osfp_add(io);
2817 		break;
2818 	}
2819 
2820 	case DIOCOSFPGET: {
2821 		struct pf_osfp_ioctl *io = (struct pf_osfp_ioctl *)addr;
2822 		error = pf_osfp_get(io);
2823 		break;
2824 	}
2825 
2826 	case DIOCXBEGIN: {
2827 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2828 		struct pfioc_trans_e	*ioe;
2829 		struct pfr_table	*table;
2830 		int			 i;
2831 
2832 		if (io->esize != sizeof(*ioe)) {
2833 			error = ENODEV;
2834 			goto fail;
2835 		}
2836 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2837 		    M_TEMP, M_WAITOK);
2838 		table = (struct pfr_table *)malloc(sizeof(*table),
2839 		    M_TEMP, M_WAITOK);
2840 		for (i = 0; i < io->size; i++) {
2841 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2842 				free(table, M_TEMP);
2843 				free(ioe, M_TEMP);
2844 				error = EFAULT;
2845 				goto fail;
2846 			}
2847 			switch (ioe->rs_num) {
2848 #ifdef ALTQ
2849 			case PF_RULESET_ALTQ:
2850 				if (ioe->anchor[0]) {
2851 					free(table, M_TEMP);
2852 					free(ioe, M_TEMP);
2853 					error = EINVAL;
2854 					goto fail;
2855 				}
2856 				if ((error = pf_begin_altq(&ioe->ticket))) {
2857 					free(table, M_TEMP);
2858 					free(ioe, M_TEMP);
2859 					goto fail;
2860 				}
2861 				break;
2862 #endif /* ALTQ */
2863 			case PF_RULESET_TABLE:
2864 				bzero(table, sizeof(*table));
2865 				strlcpy(table->pfrt_anchor, ioe->anchor,
2866 				    sizeof(table->pfrt_anchor));
2867 				if ((error = pfr_ina_begin(table,
2868 				    &ioe->ticket, NULL, 0))) {
2869 					free(table, M_TEMP);
2870 					free(ioe, M_TEMP);
2871 					goto fail;
2872 				}
2873 				break;
2874 			default:
2875 				if ((error = pf_begin_rules(&ioe->ticket,
2876 				    ioe->rs_num, ioe->anchor))) {
2877 					free(table, M_TEMP);
2878 					free(ioe, M_TEMP);
2879 					goto fail;
2880 				}
2881 				break;
2882 			}
2883 			if (copyout(ioe, io->array+i, sizeof(io->array[i]))) {
2884 				free(table, M_TEMP);
2885 				free(ioe, M_TEMP);
2886 				error = EFAULT;
2887 				goto fail;
2888 			}
2889 		}
2890 		free(table, M_TEMP);
2891 		free(ioe, M_TEMP);
2892 		break;
2893 	}
2894 
2895 	case DIOCXROLLBACK: {
2896 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2897 		struct pfioc_trans_e	*ioe;
2898 		struct pfr_table	*table;
2899 		int			 i;
2900 
2901 		if (io->esize != sizeof(*ioe)) {
2902 			error = ENODEV;
2903 			goto fail;
2904 		}
2905 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2906 		    M_TEMP, M_WAITOK);
2907 		table = (struct pfr_table *)malloc(sizeof(*table),
2908 		    M_TEMP, M_WAITOK);
2909 		for (i = 0; i < io->size; i++) {
2910 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2911 				free(table, M_TEMP);
2912 				free(ioe, M_TEMP);
2913 				error = EFAULT;
2914 				goto fail;
2915 			}
2916 			switch (ioe->rs_num) {
2917 #ifdef ALTQ
2918 			case PF_RULESET_ALTQ:
2919 				if (ioe->anchor[0]) {
2920 					free(table, M_TEMP);
2921 					free(ioe, M_TEMP);
2922 					error = EINVAL;
2923 					goto fail;
2924 				}
2925 				if ((error = pf_rollback_altq(ioe->ticket))) {
2926 					free(table, M_TEMP);
2927 					free(ioe, M_TEMP);
2928 					goto fail; /* really bad */
2929 				}
2930 				break;
2931 #endif /* ALTQ */
2932 			case PF_RULESET_TABLE:
2933 				bzero(table, sizeof(*table));
2934 				strlcpy(table->pfrt_anchor, ioe->anchor,
2935 				    sizeof(table->pfrt_anchor));
2936 				if ((error = pfr_ina_rollback(table,
2937 				    ioe->ticket, NULL, 0))) {
2938 					free(table, M_TEMP);
2939 					free(ioe, M_TEMP);
2940 					goto fail; /* really bad */
2941 				}
2942 				break;
2943 			default:
2944 				if ((error = pf_rollback_rules(ioe->ticket,
2945 				    ioe->rs_num, ioe->anchor))) {
2946 					free(table, M_TEMP);
2947 					free(ioe, M_TEMP);
2948 					goto fail; /* really bad */
2949 				}
2950 				break;
2951 			}
2952 		}
2953 		free(table, M_TEMP);
2954 		free(ioe, M_TEMP);
2955 		break;
2956 	}
2957 
2958 	case DIOCXCOMMIT: {
2959 		struct pfioc_trans	*io = (struct pfioc_trans *)addr;
2960 		struct pfioc_trans_e	*ioe;
2961 		struct pfr_table	*table;
2962 		struct pf_ruleset	*rs;
2963 		int			 i;
2964 
2965 		if (io->esize != sizeof(*ioe)) {
2966 			error = ENODEV;
2967 			goto fail;
2968 		}
2969 		ioe = (struct pfioc_trans_e *)malloc(sizeof(*ioe),
2970 		    M_TEMP, M_WAITOK);
2971 		table = (struct pfr_table *)malloc(sizeof(*table),
2972 		    M_TEMP, M_WAITOK);
2973 		/* first makes sure everything will succeed */
2974 		for (i = 0; i < io->size; i++) {
2975 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
2976 				free(table, M_TEMP);
2977 				free(ioe, M_TEMP);
2978 				error = EFAULT;
2979 				goto fail;
2980 			}
2981 			switch (ioe->rs_num) {
2982 #ifdef ALTQ
2983 			case PF_RULESET_ALTQ:
2984 				if (ioe->anchor[0]) {
2985 					free(table, M_TEMP);
2986 					free(ioe, M_TEMP);
2987 					error = EINVAL;
2988 					goto fail;
2989 				}
2990 				if (!altqs_inactive_open || ioe->ticket !=
2991 				    ticket_altqs_inactive) {
2992 					free(table, M_TEMP);
2993 					free(ioe, M_TEMP);
2994 					error = EBUSY;
2995 					goto fail;
2996 				}
2997 				break;
2998 #endif /* ALTQ */
2999 			case PF_RULESET_TABLE:
3000 				rs = pf_find_ruleset(ioe->anchor);
3001 				if (rs == NULL || !rs->topen || ioe->ticket !=
3002 				     rs->tticket) {
3003 					free(table, M_TEMP);
3004 					free(ioe, M_TEMP);
3005 					error = EBUSY;
3006 					goto fail;
3007 				}
3008 				break;
3009 			default:
3010 				if (ioe->rs_num < 0 || ioe->rs_num >=
3011 				    PF_RULESET_MAX) {
3012 					free(table, M_TEMP);
3013 					free(ioe, M_TEMP);
3014 					error = EINVAL;
3015 					goto fail;
3016 				}
3017 				rs = pf_find_ruleset(ioe->anchor);
3018 				if (rs == NULL ||
3019 				    !rs->rules[ioe->rs_num].inactive.open ||
3020 				    rs->rules[ioe->rs_num].inactive.ticket !=
3021 				    ioe->ticket) {
3022 					free(table, M_TEMP);
3023 					free(ioe, M_TEMP);
3024 					error = EBUSY;
3025 					goto fail;
3026 				}
3027 				break;
3028 			}
3029 		}
3030 		/* now do the commit - no errors should happen here */
3031 		for (i = 0; i < io->size; i++) {
3032 			if (copyin(io->array+i, ioe, sizeof(*ioe))) {
3033 				free(table, M_TEMP);
3034 				free(ioe, M_TEMP);
3035 				error = EFAULT;
3036 				goto fail;
3037 			}
3038 			switch (ioe->rs_num) {
3039 #ifdef ALTQ
3040 			case PF_RULESET_ALTQ:
3041 				if ((error = pf_commit_altq(ioe->ticket))) {
3042 					free(table, M_TEMP);
3043 					free(ioe, M_TEMP);
3044 					goto fail; /* really bad */
3045 				}
3046 				break;
3047 #endif /* ALTQ */
3048 			case PF_RULESET_TABLE:
3049 				bzero(table, sizeof(*table));
3050 				strlcpy(table->pfrt_anchor, ioe->anchor,
3051 				    sizeof(table->pfrt_anchor));
3052 				if ((error = pfr_ina_commit(table, ioe->ticket,
3053 				    NULL, NULL, 0))) {
3054 					free(table, M_TEMP);
3055 					free(ioe, M_TEMP);
3056 					goto fail; /* really bad */
3057 				}
3058 				break;
3059 			default:
3060 				if ((error = pf_commit_rules(ioe->ticket,
3061 				    ioe->rs_num, ioe->anchor))) {
3062 					free(table, M_TEMP);
3063 					free(ioe, M_TEMP);
3064 					goto fail; /* really bad */
3065 				}
3066 				break;
3067 			}
3068 		}
3069 		free(table, M_TEMP);
3070 		free(ioe, M_TEMP);
3071 		break;
3072 	}
3073 
3074 	case DIOCGETSRCNODES: {
3075 		struct pfioc_src_nodes	*psn = (struct pfioc_src_nodes *)addr;
3076 		struct pf_src_node	*n, *p, *pstore;
3077 		u_int32_t		 nr = 0;
3078 		int			 space = psn->psn_len;
3079 
3080 		if (space == 0) {
3081 			RB_FOREACH(n, pf_src_tree, &tree_src_tracking)
3082 				nr++;
3083 			psn->psn_len = sizeof(struct pf_src_node) * nr;
3084 			break;
3085 		}
3086 
3087 		pstore = malloc(sizeof(*pstore), M_TEMP, M_WAITOK);
3088 
3089 		p = psn->psn_src_nodes;
3090 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3091 			int	secs = time_second, diff;
3092 
3093 			if ((nr + 1) * sizeof(*p) > (unsigned)psn->psn_len)
3094 				break;
3095 
3096 			bcopy(n, pstore, sizeof(*pstore));
3097 			if (n->rule.ptr != NULL)
3098 				pstore->rule.nr = n->rule.ptr->nr;
3099 			pstore->creation = secs - pstore->creation;
3100 			if (pstore->expire > secs)
3101 				pstore->expire -= secs;
3102 			else
3103 				pstore->expire = 0;
3104 
3105 			/* adjust the connection rate estimate */
3106 			diff = secs - n->conn_rate.last;
3107 			if (diff >= n->conn_rate.seconds)
3108 				pstore->conn_rate.count = 0;
3109 			else
3110 				pstore->conn_rate.count -=
3111 				    n->conn_rate.count * diff /
3112 				    n->conn_rate.seconds;
3113 
3114 			error = copyout(pstore, p, sizeof(*p));
3115 			if (error) {
3116 				free(pstore, M_TEMP);
3117 				goto fail;
3118 			}
3119 			p++;
3120 			nr++;
3121 		}
3122 		psn->psn_len = sizeof(struct pf_src_node) * nr;
3123 
3124 		free(pstore, M_TEMP);
3125 		break;
3126 	}
3127 
3128 	case DIOCCLRSRCNODES: {
3129 		struct pf_src_node	*n;
3130 		struct pf_state		*state;
3131 
3132 		RB_FOREACH(state, pf_state_tree_id, &tree_id) {
3133 			state->src_node = NULL;
3134 			state->nat_src_node = NULL;
3135 		}
3136 		RB_FOREACH(n, pf_src_tree, &tree_src_tracking) {
3137 			n->expire = 1;
3138 			n->states = 0;
3139 		}
3140 		pf_purge_expired_src_nodes(1);
3141 		pf_status.src_nodes = 0;
3142 		break;
3143 	}
3144 
3145 	case DIOCKILLSRCNODES: {
3146 		struct pf_src_node	*sn;
3147 		struct pf_state		*ps;
3148 		struct pfioc_src_node_kill *psnk = \
3149 			(struct pfioc_src_node_kill *) addr;
3150 		int			killed = 0;
3151 
3152 		RB_FOREACH(sn, pf_src_tree, &tree_src_tracking) {
3153         		if (PF_MATCHA(psnk->psnk_src.neg, \
3154 				      &psnk->psnk_src.addr.v.a.addr, \
3155 				      &psnk->psnk_src.addr.v.a.mask, \
3156 				      &sn->addr, sn->af) &&
3157 			    PF_MATCHA(psnk->psnk_dst.neg, \
3158 				      &psnk->psnk_dst.addr.v.a.addr, \
3159 				      &psnk->psnk_dst.addr.v.a.mask, \
3160 				      &sn->raddr, sn->af)) {
3161 				/* Handle state to src_node linkage */
3162 				if (sn->states != 0) {
3163 					RB_FOREACH(ps, pf_state_tree_id,
3164 					    &tree_id) {
3165 						if (ps->src_node == sn)
3166 							ps->src_node = NULL;
3167 						if (ps->nat_src_node == sn)
3168 							ps->nat_src_node = NULL;
3169 					}
3170 					sn->states = 0;
3171 				}
3172 				sn->expire = 1;
3173 				killed++;
3174 			}
3175 		}
3176 
3177 		if (killed > 0)
3178 			pf_purge_expired_src_nodes(1);
3179 
3180 		psnk->psnk_af = killed;
3181 		break;
3182 	}
3183 
3184 	case DIOCSETHOSTID: {
3185 		u_int32_t	*hid = (u_int32_t *)addr;
3186 
3187 		if (*hid == 0)
3188 			pf_status.hostid = cprng_fast32();
3189 		else
3190 			pf_status.hostid = *hid;
3191 		break;
3192 	}
3193 
3194 	case DIOCOSFPFLUSH:
3195 		pf_osfp_flush();
3196 		break;
3197 
3198 	case DIOCIGETIFACES: {
3199 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3200 
3201 		if (io->pfiio_esize != sizeof(struct pfi_kif)) {
3202 			error = ENODEV;
3203 			break;
3204 		}
3205 		error = pfi_get_ifaces(io->pfiio_name, io->pfiio_buffer,
3206 		    &io->pfiio_size);
3207 		break;
3208 	}
3209 
3210 	case DIOCSETIFFLAG: {
3211 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3212 
3213 		error = pfi_set_flags(io->pfiio_name, io->pfiio_flags);
3214 		break;
3215 	}
3216 
3217 	case DIOCCLRIFFLAG: {
3218 		struct pfioc_iface *io = (struct pfioc_iface *)addr;
3219 
3220 		error = pfi_clear_flags(io->pfiio_name, io->pfiio_flags);
3221 		break;
3222 	}
3223 
3224 	case DIOCSETLCK: {
3225 		pf_state_lock = *(uint32_t*)addr;
3226 		break;
3227 	}
3228 
3229 	default:
3230 		error = ENODEV;
3231 		break;
3232 	}
3233 fail:
3234 	splx(s);
3235 	if (flags & FWRITE)
3236 		rw_exit_write(&pf_consistency_lock);
3237 	else
3238 		rw_exit_read(&pf_consistency_lock);
3239 	return (error);
3240 }
3241 
3242 #ifdef __NetBSD__
3243 #ifdef INET
3244 static int
3245 pfil4_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
3246 {
3247 	int error;
3248 
3249 	/*
3250 	 * ensure that mbufs are writable beforehand
3251 	 * as it's assumed by pf code.
3252 	 * ip hdr (60 bytes) + tcp hdr (60 bytes) should be enough.
3253 	 * XXX inefficient
3254 	 */
3255 	error = m_makewritable(mp, 0, 60 + 60, M_DONTWAIT);
3256 	if (error) {
3257 		m_freem(*mp);
3258 		*mp = NULL;
3259 		return error;
3260 	}
3261 
3262 	/*
3263 	 * If the packet is out-bound, we can't delay checksums
3264 	 * here.  For in-bound, the checksum has already been
3265 	 * validated.
3266 	 */
3267 	if (dir == PFIL_OUT) {
3268 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
3269 			in_delayed_cksum(*mp);
3270 			(*mp)->m_pkthdr.csum_flags &=
3271 			    ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
3272 		}
3273 	}
3274 
3275 	if (pf_test(dir == PFIL_OUT ? PF_OUT : PF_IN, ifp, mp, NULL)
3276 	    != PF_PASS) {
3277 		m_freem(*mp);
3278 		*mp = NULL;
3279 		return EHOSTUNREACH;
3280 	}
3281 
3282 	/*
3283 	 * we're not compatible with fast-forward.
3284 	 */
3285 
3286 	if (dir == PFIL_IN && *mp) {
3287 		(*mp)->m_flags &= ~M_CANFASTFWD;
3288 	}
3289 
3290 	return (0);
3291 }
3292 #endif /* INET */
3293 
3294 #ifdef INET6
3295 static int
3296 pfil6_wrapper(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
3297 {
3298 	int error;
3299 
3300 	/*
3301 	 * ensure that mbufs are writable beforehand
3302 	 * as it's assumed by pf code.
3303 	 * XXX inefficient
3304 	 */
3305 	error = m_makewritable(mp, 0, M_COPYALL, M_DONTWAIT);
3306 	if (error) {
3307 		m_freem(*mp);
3308 		*mp = NULL;
3309 		return error;
3310 	}
3311 
3312 	/*
3313 	 * If the packet is out-bound, we can't delay checksums
3314 	 * here.  For in-bound, the checksum has already been
3315 	 * validated.
3316 	 */
3317 	if (dir == PFIL_OUT) {
3318 		if ((*mp)->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
3319 			in6_delayed_cksum(*mp);
3320 			(*mp)->m_pkthdr.csum_flags &=
3321 			    ~(M_CSUM_TCPv6|M_CSUM_UDPv6);
3322 		}
3323 	}
3324 
3325 	if (pf_test6(dir == PFIL_OUT ? PF_OUT : PF_IN, ifp, mp, NULL)
3326 	    != PF_PASS) {
3327 		m_freem(*mp);
3328 		*mp = NULL;
3329 		return EHOSTUNREACH;
3330 	} else
3331 		return (0);
3332 }
3333 #endif /* INET6 */
3334 
3335 static int
3336 pf_pfil_attach(void)
3337 {
3338 	pfil_head_t *ph_inet;
3339 #ifdef INET6
3340 	pfil_head_t *ph_inet6;
3341 #endif /* INET6 */
3342 	int error;
3343 
3344 	if (pf_pfil_attached)
3345 		return (EBUSY);
3346 
3347 	ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
3348 	if (ph_inet)
3349 		error = pfil_add_hook((void *)pfil4_wrapper, NULL,
3350 		    PFIL_IN|PFIL_OUT, ph_inet);
3351 	else
3352 		error = ENOENT;
3353 	if (error)
3354 		return (error);
3355 
3356 #ifdef INET6
3357 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
3358 	if (ph_inet6)
3359 		error = pfil_add_hook((void *)pfil6_wrapper, NULL,
3360 		    PFIL_IN|PFIL_OUT, ph_inet6);
3361 	else
3362 		error = ENOENT;
3363 	if (error)
3364 		goto bad;
3365 #endif /* INET6 */
3366 
3367 	pf_pfil_attached = 1;
3368 
3369 	return (0);
3370 
3371 #ifdef INET6
3372 bad:
3373 	pfil_remove_hook(pfil4_wrapper, NULL, PFIL_IN|PFIL_OUT, ph_inet);
3374 #endif /* INET6 */
3375 
3376 	return (error);
3377 }
3378 
3379 static int
3380 pf_pfil_detach(void)
3381 {
3382 	pfil_head_t *ph_inet;
3383 #ifdef INET6
3384 	pfil_head_t *ph_inet6;
3385 #endif /* INET6 */
3386 
3387 	if (pf_pfil_attached == 0)
3388 		return (EBUSY);
3389 
3390 	ph_inet = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET);
3391 	if (ph_inet)
3392 		pfil_remove_hook((void *)pfil4_wrapper, NULL,
3393 		    PFIL_IN|PFIL_OUT, ph_inet);
3394 #ifdef INET6
3395 	ph_inet6 = pfil_head_get(PFIL_TYPE_AF, (void *)AF_INET6);
3396 	if (ph_inet6)
3397 		pfil_remove_hook((void *)pfil6_wrapper, NULL,
3398 		    PFIL_IN|PFIL_OUT, ph_inet6);
3399 #endif /* INET6 */
3400 	pf_pfil_attached = 0;
3401 
3402 	return (0);
3403 }
3404 #endif /* __NetBSD__ */
3405 
3406 #if defined(__NetBSD__)
3407 MODULE(MODULE_CLASS_DRIVER, pf, "bpf");
3408 
3409 static int
3410 pf_modcmd(modcmd_t cmd, void *opaque)
3411 {
3412 #ifdef _MODULE
3413 	extern void pflogattach(int);
3414 	extern void pflogdetach(void);
3415 
3416 	devmajor_t cmajor = NODEVMAJOR, bmajor = NODEVMAJOR;
3417 	int err;
3418 
3419 	switch (cmd) {
3420 	case MODULE_CMD_INIT:
3421 		err = devsw_attach("pf", NULL, &bmajor, &pf_cdevsw, &cmajor);
3422 		if (err)
3423 			return err;
3424 		pfattach(1);
3425 		pflogattach(1);
3426 		return 0;
3427 	case MODULE_CMD_FINI:
3428 		if (pf_status.running) {
3429 			return EBUSY;
3430 		} else {
3431 			pfdetach();
3432 			pflogdetach();
3433 			return devsw_detach(NULL, &pf_cdevsw);
3434 		}
3435 	default:
3436 		return ENOTTY;
3437 	}
3438 #else
3439 	if (cmd == MODULE_CMD_INIT)
3440 		return 0;
3441 	return ENOTTY;
3442 #endif
3443 }
3444 #endif
3445