xref: /freebsd/sys/netpfil/pf/pf_table.c (revision d06955f9bdb1416d9196043ed781f9b36dae9adc)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2002 Cedric Berger
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  *
11  *    - Redistributions of source code must retain the above copyright
12  *      notice, this list of conditions and the following disclaimer.
13  *    - Redistributions in binary form must reproduce the above
14  *      copyright notice, this list of conditions and the following
15  *      disclaimer in the documentation and/or other materials provided
16  *      with the distribution.
17  *
18  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
20  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
21  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
22  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
23  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
24  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
25  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
26  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
28  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  *
31  *	$OpenBSD: pf_table.c,v 1.79 2008/10/08 06:24:50 mcbride Exp $
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 
40 #include <sys/param.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/mutex.h>
46 #include <sys/refcount.h>
47 #include <sys/rwlock.h>
48 #include <sys/socket.h>
49 #include <vm/uma.h>
50 
51 #include <net/if.h>
52 #include <net/vnet.h>
53 #include <net/pfvar.h>
54 
55 #define	ACCEPT_FLAGS(flags, oklist)		\
56 	do {					\
57 		if ((flags & ~(oklist)) &	\
58 		    PFR_FLAG_ALLMASK)		\
59 			return (EINVAL);	\
60 	} while (0)
61 
62 #define	FILLIN_SIN(sin, addr)			\
63 	do {					\
64 		(sin).sin_len = sizeof(sin);	\
65 		(sin).sin_family = AF_INET;	\
66 		(sin).sin_addr = (addr);	\
67 	} while (0)
68 
69 #define	FILLIN_SIN6(sin6, addr)			\
70 	do {					\
71 		(sin6).sin6_len = sizeof(sin6);	\
72 		(sin6).sin6_family = AF_INET6;	\
73 		(sin6).sin6_addr = (addr);	\
74 	} while (0)
75 
76 #define	SWAP(type, a1, a2)			\
77 	do {					\
78 		type tmp = a1;			\
79 		a1 = a2;			\
80 		a2 = tmp;			\
81 	} while (0)
82 
83 #define	SUNION2PF(su, af) (((af)==AF_INET) ?	\
84     (struct pf_addr *)&(su)->sin.sin_addr :	\
85     (struct pf_addr *)&(su)->sin6.sin6_addr)
86 
87 #define	AF_BITS(af)		(((af)==AF_INET)?32:128)
88 #define	ADDR_NETWORK(ad)	((ad)->pfra_net < AF_BITS((ad)->pfra_af))
89 #define	KENTRY_NETWORK(ke)	((ke)->pfrke_net < AF_BITS((ke)->pfrke_af))
90 #define	KENTRY_RNF_ROOT(ke) \
91 		((((struct radix_node *)(ke))->rn_flags & RNF_ROOT) != 0)
92 
93 #define	NO_ADDRESSES		(-1)
94 #define	ENQUEUE_UNMARKED_ONLY	(1)
95 #define	INVERT_NEG_FLAG		(1)
96 
97 struct pfr_walktree {
98 	enum pfrw_op {
99 		PFRW_MARK,
100 		PFRW_SWEEP,
101 		PFRW_ENQUEUE,
102 		PFRW_GET_ADDRS,
103 		PFRW_GET_ASTATS,
104 		PFRW_POOL_GET,
105 		PFRW_DYNADDR_UPDATE
106 	}	 pfrw_op;
107 	union {
108 		struct pfr_addr		*pfrw1_addr;
109 		struct pfr_astats	*pfrw1_astats;
110 		struct pfr_kentryworkq	*pfrw1_workq;
111 		struct pfr_kentry	*pfrw1_kentry;
112 		struct pfi_dynaddr	*pfrw1_dyn;
113 	}	 pfrw_1;
114 	int	 pfrw_free;
115 };
116 #define	pfrw_addr	pfrw_1.pfrw1_addr
117 #define	pfrw_astats	pfrw_1.pfrw1_astats
118 #define	pfrw_workq	pfrw_1.pfrw1_workq
119 #define	pfrw_kentry	pfrw_1.pfrw1_kentry
120 #define	pfrw_dyn	pfrw_1.pfrw1_dyn
121 #define	pfrw_cnt	pfrw_free
122 
123 #define	senderr(e)	do { rv = (e); goto _bad; } while (0)
124 
125 static MALLOC_DEFINE(M_PFTABLE, "pf_table", "pf(4) tables structures");
126 static VNET_DEFINE(uma_zone_t, pfr_kentry_z);
127 #define	V_pfr_kentry_z		VNET(pfr_kentry_z)
128 static VNET_DEFINE(uma_zone_t, pfr_kcounters_z);
129 #define	V_pfr_kcounters_z	VNET(pfr_kcounters_z)
130 
131 static struct pf_addr	 pfr_ffaddr = {
132 	.addr32 = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }
133 };
134 
135 static void		 pfr_copyout_addr(struct pfr_addr *,
136 			    struct pfr_kentry *ke);
137 static int		 pfr_validate_addr(struct pfr_addr *);
138 static void		 pfr_enqueue_addrs(struct pfr_ktable *,
139 			    struct pfr_kentryworkq *, int *, int);
140 static void		 pfr_mark_addrs(struct pfr_ktable *);
141 static struct pfr_kentry
142 			*pfr_lookup_addr(struct pfr_ktable *,
143 			    struct pfr_addr *, int);
144 static struct pfr_kentry *pfr_create_kentry(struct pfr_addr *);
145 static void		 pfr_destroy_kentries(struct pfr_kentryworkq *);
146 static void		 pfr_destroy_kentry(struct pfr_kentry *);
147 static void		 pfr_insert_kentries(struct pfr_ktable *,
148 			    struct pfr_kentryworkq *, long);
149 static void		 pfr_remove_kentries(struct pfr_ktable *,
150 			    struct pfr_kentryworkq *);
151 static void		 pfr_clstats_kentries(struct pfr_kentryworkq *, long,
152 			    int);
153 static void		 pfr_reset_feedback(struct pfr_addr *, int);
154 static void		 pfr_prepare_network(union sockaddr_union *, int, int);
155 static int		 pfr_route_kentry(struct pfr_ktable *,
156 			    struct pfr_kentry *);
157 static int		 pfr_unroute_kentry(struct pfr_ktable *,
158 			    struct pfr_kentry *);
159 static int		 pfr_walktree(struct radix_node *, void *);
160 static int		 pfr_validate_table(struct pfr_table *, int, int);
161 static int		 pfr_fix_anchor(char *);
162 static void		 pfr_commit_ktable(struct pfr_ktable *, long);
163 static void		 pfr_insert_ktables(struct pfr_ktableworkq *);
164 static void		 pfr_insert_ktable(struct pfr_ktable *);
165 static void		 pfr_setflags_ktables(struct pfr_ktableworkq *);
166 static void		 pfr_setflags_ktable(struct pfr_ktable *, int);
167 static void		 pfr_clstats_ktables(struct pfr_ktableworkq *, long,
168 			    int);
169 static void		 pfr_clstats_ktable(struct pfr_ktable *, long, int);
170 static struct pfr_ktable
171 			*pfr_create_ktable(struct pfr_table *, long, int);
172 static void		 pfr_destroy_ktables(struct pfr_ktableworkq *, int);
173 static void		 pfr_destroy_ktable(struct pfr_ktable *, int);
174 static int		 pfr_ktable_compare(struct pfr_ktable *,
175 			    struct pfr_ktable *);
176 static struct pfr_ktable
177 			*pfr_lookup_table(struct pfr_table *);
178 static void		 pfr_clean_node_mask(struct pfr_ktable *,
179 			    struct pfr_kentryworkq *);
180 static int		 pfr_table_count(struct pfr_table *, int);
181 static int		 pfr_skip_table(struct pfr_table *,
182 			    struct pfr_ktable *, int);
183 static struct pfr_kentry
184 			*pfr_kentry_byidx(struct pfr_ktable *, int, int);
185 
186 static RB_PROTOTYPE(pfr_ktablehead, pfr_ktable, pfrkt_tree, pfr_ktable_compare);
187 static RB_GENERATE(pfr_ktablehead, pfr_ktable, pfrkt_tree, pfr_ktable_compare);
188 
189 static VNET_DEFINE(struct pfr_ktablehead, pfr_ktables);
190 #define	V_pfr_ktables	VNET(pfr_ktables)
191 
192 static VNET_DEFINE(struct pfr_table, pfr_nulltable);
193 #define	V_pfr_nulltable	VNET(pfr_nulltable)
194 
195 static VNET_DEFINE(int, pfr_ktable_cnt);
196 #define V_pfr_ktable_cnt	VNET(pfr_ktable_cnt)
197 
198 void
199 pfr_initialize(void)
200 {
201 
202 	V_pfr_kentry_z = uma_zcreate("pf table entries",
203 	    sizeof(struct pfr_kentry), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR,
204 	    0);
205 	V_pfr_kcounters_z = uma_zcreate("pf table counters",
206 	    sizeof(struct pfr_kcounters), NULL, NULL, NULL, NULL,
207 	    UMA_ALIGN_PTR, 0);
208 	V_pf_limits[PF_LIMIT_TABLE_ENTRIES].zone = V_pfr_kentry_z;
209 	V_pf_limits[PF_LIMIT_TABLE_ENTRIES].limit = PFR_KENTRY_HIWAT;
210 }
211 
212 void
213 pfr_cleanup(void)
214 {
215 
216 	uma_zdestroy(V_pfr_kentry_z);
217 	uma_zdestroy(V_pfr_kcounters_z);
218 }
219 
220 int
221 pfr_clr_addrs(struct pfr_table *tbl, int *ndel, int flags)
222 {
223 	struct pfr_ktable	*kt;
224 	struct pfr_kentryworkq	 workq;
225 
226 	PF_RULES_WASSERT();
227 
228 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY);
229 	if (pfr_validate_table(tbl, 0, flags & PFR_FLAG_USERIOCTL))
230 		return (EINVAL);
231 	kt = pfr_lookup_table(tbl);
232 	if (kt == NULL || !(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
233 		return (ESRCH);
234 	if (kt->pfrkt_flags & PFR_TFLAG_CONST)
235 		return (EPERM);
236 	pfr_enqueue_addrs(kt, &workq, ndel, 0);
237 
238 	if (!(flags & PFR_FLAG_DUMMY)) {
239 		pfr_remove_kentries(kt, &workq);
240 		KASSERT(kt->pfrkt_cnt == 0, ("%s: non-null pfrkt_cnt", __func__));
241 	}
242 	return (0);
243 }
244 
245 int
246 pfr_add_addrs(struct pfr_table *tbl, struct pfr_addr *addr, int size,
247     int *nadd, int flags)
248 {
249 	struct pfr_ktable	*kt, *tmpkt;
250 	struct pfr_kentryworkq	 workq;
251 	struct pfr_kentry	*p, *q;
252 	struct pfr_addr		*ad;
253 	int			 i, rv, xadd = 0;
254 	long			 tzero = time_second;
255 
256 	PF_RULES_WASSERT();
257 
258 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY | PFR_FLAG_FEEDBACK);
259 	if (pfr_validate_table(tbl, 0, flags & PFR_FLAG_USERIOCTL))
260 		return (EINVAL);
261 	kt = pfr_lookup_table(tbl);
262 	if (kt == NULL || !(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
263 		return (ESRCH);
264 	if (kt->pfrkt_flags & PFR_TFLAG_CONST)
265 		return (EPERM);
266 	tmpkt = pfr_create_ktable(&V_pfr_nulltable, 0, 0);
267 	if (tmpkt == NULL)
268 		return (ENOMEM);
269 	SLIST_INIT(&workq);
270 	for (i = 0, ad = addr; i < size; i++, ad++) {
271 		if (pfr_validate_addr(ad))
272 			senderr(EINVAL);
273 		p = pfr_lookup_addr(kt, ad, 1);
274 		q = pfr_lookup_addr(tmpkt, ad, 1);
275 		if (flags & PFR_FLAG_FEEDBACK) {
276 			if (q != NULL)
277 				ad->pfra_fback = PFR_FB_DUPLICATE;
278 			else if (p == NULL)
279 				ad->pfra_fback = PFR_FB_ADDED;
280 			else if (p->pfrke_not != ad->pfra_not)
281 				ad->pfra_fback = PFR_FB_CONFLICT;
282 			else
283 				ad->pfra_fback = PFR_FB_NONE;
284 		}
285 		if (p == NULL && q == NULL) {
286 			p = pfr_create_kentry(ad);
287 			if (p == NULL)
288 				senderr(ENOMEM);
289 			if (pfr_route_kentry(tmpkt, p)) {
290 				pfr_destroy_kentry(p);
291 				ad->pfra_fback = PFR_FB_NONE;
292 			} else {
293 				SLIST_INSERT_HEAD(&workq, p, pfrke_workq);
294 				xadd++;
295 			}
296 		}
297 	}
298 	pfr_clean_node_mask(tmpkt, &workq);
299 	if (!(flags & PFR_FLAG_DUMMY))
300 		pfr_insert_kentries(kt, &workq, tzero);
301 	else
302 		pfr_destroy_kentries(&workq);
303 	if (nadd != NULL)
304 		*nadd = xadd;
305 	pfr_destroy_ktable(tmpkt, 0);
306 	return (0);
307 _bad:
308 	pfr_clean_node_mask(tmpkt, &workq);
309 	pfr_destroy_kentries(&workq);
310 	if (flags & PFR_FLAG_FEEDBACK)
311 		pfr_reset_feedback(addr, size);
312 	pfr_destroy_ktable(tmpkt, 0);
313 	return (rv);
314 }
315 
316 int
317 pfr_del_addrs(struct pfr_table *tbl, struct pfr_addr *addr, int size,
318     int *ndel, int flags)
319 {
320 	struct pfr_ktable	*kt;
321 	struct pfr_kentryworkq	 workq;
322 	struct pfr_kentry	*p;
323 	struct pfr_addr		*ad;
324 	int			 i, rv, xdel = 0, log = 1;
325 
326 	PF_RULES_WASSERT();
327 
328 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY | PFR_FLAG_FEEDBACK);
329 	if (pfr_validate_table(tbl, 0, flags & PFR_FLAG_USERIOCTL))
330 		return (EINVAL);
331 	kt = pfr_lookup_table(tbl);
332 	if (kt == NULL || !(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
333 		return (ESRCH);
334 	if (kt->pfrkt_flags & PFR_TFLAG_CONST)
335 		return (EPERM);
336 	/*
337 	 * there are two algorithms to choose from here.
338 	 * with:
339 	 *   n: number of addresses to delete
340 	 *   N: number of addresses in the table
341 	 *
342 	 * one is O(N) and is better for large 'n'
343 	 * one is O(n*LOG(N)) and is better for small 'n'
344 	 *
345 	 * following code try to decide which one is best.
346 	 */
347 	for (i = kt->pfrkt_cnt; i > 0; i >>= 1)
348 		log++;
349 	if (size > kt->pfrkt_cnt/log) {
350 		/* full table scan */
351 		pfr_mark_addrs(kt);
352 	} else {
353 		/* iterate over addresses to delete */
354 		for (i = 0, ad = addr; i < size; i++, ad++) {
355 			if (pfr_validate_addr(ad))
356 				return (EINVAL);
357 			p = pfr_lookup_addr(kt, ad, 1);
358 			if (p != NULL)
359 				p->pfrke_mark = 0;
360 		}
361 	}
362 	SLIST_INIT(&workq);
363 	for (i = 0, ad = addr; i < size; i++, ad++) {
364 		if (pfr_validate_addr(ad))
365 			senderr(EINVAL);
366 		p = pfr_lookup_addr(kt, ad, 1);
367 		if (flags & PFR_FLAG_FEEDBACK) {
368 			if (p == NULL)
369 				ad->pfra_fback = PFR_FB_NONE;
370 			else if (p->pfrke_not != ad->pfra_not)
371 				ad->pfra_fback = PFR_FB_CONFLICT;
372 			else if (p->pfrke_mark)
373 				ad->pfra_fback = PFR_FB_DUPLICATE;
374 			else
375 				ad->pfra_fback = PFR_FB_DELETED;
376 		}
377 		if (p != NULL && p->pfrke_not == ad->pfra_not &&
378 		    !p->pfrke_mark) {
379 			p->pfrke_mark = 1;
380 			SLIST_INSERT_HEAD(&workq, p, pfrke_workq);
381 			xdel++;
382 		}
383 	}
384 	if (!(flags & PFR_FLAG_DUMMY))
385 		pfr_remove_kentries(kt, &workq);
386 	if (ndel != NULL)
387 		*ndel = xdel;
388 	return (0);
389 _bad:
390 	if (flags & PFR_FLAG_FEEDBACK)
391 		pfr_reset_feedback(addr, size);
392 	return (rv);
393 }
394 
395 int
396 pfr_set_addrs(struct pfr_table *tbl, struct pfr_addr *addr, int size,
397     int *size2, int *nadd, int *ndel, int *nchange, int flags,
398     u_int32_t ignore_pfrt_flags)
399 {
400 	struct pfr_ktable	*kt, *tmpkt;
401 	struct pfr_kentryworkq	 addq, delq, changeq;
402 	struct pfr_kentry	*p, *q;
403 	struct pfr_addr		 ad;
404 	int			 i, rv, xadd = 0, xdel = 0, xchange = 0;
405 	long			 tzero = time_second;
406 
407 	PF_RULES_WASSERT();
408 
409 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY | PFR_FLAG_FEEDBACK);
410 	if (pfr_validate_table(tbl, ignore_pfrt_flags, flags &
411 	    PFR_FLAG_USERIOCTL))
412 		return (EINVAL);
413 	kt = pfr_lookup_table(tbl);
414 	if (kt == NULL || !(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
415 		return (ESRCH);
416 	if (kt->pfrkt_flags & PFR_TFLAG_CONST)
417 		return (EPERM);
418 	tmpkt = pfr_create_ktable(&V_pfr_nulltable, 0, 0);
419 	if (tmpkt == NULL)
420 		return (ENOMEM);
421 	pfr_mark_addrs(kt);
422 	SLIST_INIT(&addq);
423 	SLIST_INIT(&delq);
424 	SLIST_INIT(&changeq);
425 	for (i = 0; i < size; i++) {
426 		/*
427 		 * XXXGL: undertand pf_if usage of this function
428 		 * and make ad a moving pointer
429 		 */
430 		bcopy(addr + i, &ad, sizeof(ad));
431 		if (pfr_validate_addr(&ad))
432 			senderr(EINVAL);
433 		ad.pfra_fback = PFR_FB_NONE;
434 		p = pfr_lookup_addr(kt, &ad, 1);
435 		if (p != NULL) {
436 			if (p->pfrke_mark) {
437 				ad.pfra_fback = PFR_FB_DUPLICATE;
438 				goto _skip;
439 			}
440 			p->pfrke_mark = 1;
441 			if (p->pfrke_not != ad.pfra_not) {
442 				SLIST_INSERT_HEAD(&changeq, p, pfrke_workq);
443 				ad.pfra_fback = PFR_FB_CHANGED;
444 				xchange++;
445 			}
446 		} else {
447 			q = pfr_lookup_addr(tmpkt, &ad, 1);
448 			if (q != NULL) {
449 				ad.pfra_fback = PFR_FB_DUPLICATE;
450 				goto _skip;
451 			}
452 			p = pfr_create_kentry(&ad);
453 			if (p == NULL)
454 				senderr(ENOMEM);
455 			if (pfr_route_kentry(tmpkt, p)) {
456 				pfr_destroy_kentry(p);
457 				ad.pfra_fback = PFR_FB_NONE;
458 			} else {
459 				SLIST_INSERT_HEAD(&addq, p, pfrke_workq);
460 				ad.pfra_fback = PFR_FB_ADDED;
461 				xadd++;
462 			}
463 		}
464 _skip:
465 		if (flags & PFR_FLAG_FEEDBACK)
466 			bcopy(&ad, addr + i, sizeof(ad));
467 	}
468 	pfr_enqueue_addrs(kt, &delq, &xdel, ENQUEUE_UNMARKED_ONLY);
469 	if ((flags & PFR_FLAG_FEEDBACK) && *size2) {
470 		if (*size2 < size+xdel) {
471 			*size2 = size+xdel;
472 			senderr(0);
473 		}
474 		i = 0;
475 		SLIST_FOREACH(p, &delq, pfrke_workq) {
476 			pfr_copyout_addr(&ad, p);
477 			ad.pfra_fback = PFR_FB_DELETED;
478 			bcopy(&ad, addr + size + i, sizeof(ad));
479 			i++;
480 		}
481 	}
482 	pfr_clean_node_mask(tmpkt, &addq);
483 	if (!(flags & PFR_FLAG_DUMMY)) {
484 		pfr_insert_kentries(kt, &addq, tzero);
485 		pfr_remove_kentries(kt, &delq);
486 		pfr_clstats_kentries(&changeq, tzero, INVERT_NEG_FLAG);
487 	} else
488 		pfr_destroy_kentries(&addq);
489 	if (nadd != NULL)
490 		*nadd = xadd;
491 	if (ndel != NULL)
492 		*ndel = xdel;
493 	if (nchange != NULL)
494 		*nchange = xchange;
495 	if ((flags & PFR_FLAG_FEEDBACK) && size2)
496 		*size2 = size+xdel;
497 	pfr_destroy_ktable(tmpkt, 0);
498 	return (0);
499 _bad:
500 	pfr_clean_node_mask(tmpkt, &addq);
501 	pfr_destroy_kentries(&addq);
502 	if (flags & PFR_FLAG_FEEDBACK)
503 		pfr_reset_feedback(addr, size);
504 	pfr_destroy_ktable(tmpkt, 0);
505 	return (rv);
506 }
507 
508 int
509 pfr_tst_addrs(struct pfr_table *tbl, struct pfr_addr *addr, int size,
510 	int *nmatch, int flags)
511 {
512 	struct pfr_ktable	*kt;
513 	struct pfr_kentry	*p;
514 	struct pfr_addr		*ad;
515 	int			 i, xmatch = 0;
516 
517 	PF_RULES_RASSERT();
518 
519 	ACCEPT_FLAGS(flags, PFR_FLAG_REPLACE);
520 	if (pfr_validate_table(tbl, 0, 0))
521 		return (EINVAL);
522 	kt = pfr_lookup_table(tbl);
523 	if (kt == NULL || !(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
524 		return (ESRCH);
525 
526 	for (i = 0, ad = addr; i < size; i++, ad++) {
527 		if (pfr_validate_addr(ad))
528 			return (EINVAL);
529 		if (ADDR_NETWORK(ad))
530 			return (EINVAL);
531 		p = pfr_lookup_addr(kt, ad, 0);
532 		if (flags & PFR_FLAG_REPLACE)
533 			pfr_copyout_addr(ad, p);
534 		ad->pfra_fback = (p == NULL) ? PFR_FB_NONE :
535 		    (p->pfrke_not ? PFR_FB_NOTMATCH : PFR_FB_MATCH);
536 		if (p != NULL && !p->pfrke_not)
537 			xmatch++;
538 	}
539 	if (nmatch != NULL)
540 		*nmatch = xmatch;
541 	return (0);
542 }
543 
544 int
545 pfr_get_addrs(struct pfr_table *tbl, struct pfr_addr *addr, int *size,
546 	int flags)
547 {
548 	struct pfr_ktable	*kt;
549 	struct pfr_walktree	 w;
550 	int			 rv;
551 
552 	PF_RULES_RASSERT();
553 
554 	ACCEPT_FLAGS(flags, 0);
555 	if (pfr_validate_table(tbl, 0, 0))
556 		return (EINVAL);
557 	kt = pfr_lookup_table(tbl);
558 	if (kt == NULL || !(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
559 		return (ESRCH);
560 	if (kt->pfrkt_cnt > *size) {
561 		*size = kt->pfrkt_cnt;
562 		return (0);
563 	}
564 
565 	bzero(&w, sizeof(w));
566 	w.pfrw_op = PFRW_GET_ADDRS;
567 	w.pfrw_addr = addr;
568 	w.pfrw_free = kt->pfrkt_cnt;
569 	rv = kt->pfrkt_ip4->rnh_walktree(&kt->pfrkt_ip4->rh, pfr_walktree, &w);
570 	if (!rv)
571 		rv = kt->pfrkt_ip6->rnh_walktree(&kt->pfrkt_ip6->rh,
572 		    pfr_walktree, &w);
573 	if (rv)
574 		return (rv);
575 
576 	KASSERT(w.pfrw_free == 0, ("%s: corruption detected (%d)", __func__,
577 	    w.pfrw_free));
578 
579 	*size = kt->pfrkt_cnt;
580 	return (0);
581 }
582 
583 int
584 pfr_get_astats(struct pfr_table *tbl, struct pfr_astats *addr, int *size,
585 	int flags)
586 {
587 	struct pfr_ktable	*kt;
588 	struct pfr_walktree	 w;
589 	struct pfr_kentryworkq	 workq;
590 	int			 rv;
591 	long			 tzero = time_second;
592 
593 	PF_RULES_RASSERT();
594 
595 	/* XXX PFR_FLAG_CLSTATS disabled */
596 	ACCEPT_FLAGS(flags, 0);
597 	if (pfr_validate_table(tbl, 0, 0))
598 		return (EINVAL);
599 	kt = pfr_lookup_table(tbl);
600 	if (kt == NULL || !(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
601 		return (ESRCH);
602 	if (kt->pfrkt_cnt > *size) {
603 		*size = kt->pfrkt_cnt;
604 		return (0);
605 	}
606 
607 	bzero(&w, sizeof(w));
608 	w.pfrw_op = PFRW_GET_ASTATS;
609 	w.pfrw_astats = addr;
610 	w.pfrw_free = kt->pfrkt_cnt;
611 	rv = kt->pfrkt_ip4->rnh_walktree(&kt->pfrkt_ip4->rh, pfr_walktree, &w);
612 	if (!rv)
613 		rv = kt->pfrkt_ip6->rnh_walktree(&kt->pfrkt_ip6->rh,
614 		    pfr_walktree, &w);
615 	if (!rv && (flags & PFR_FLAG_CLSTATS)) {
616 		pfr_enqueue_addrs(kt, &workq, NULL, 0);
617 		pfr_clstats_kentries(&workq, tzero, 0);
618 	}
619 	if (rv)
620 		return (rv);
621 
622 	if (w.pfrw_free) {
623 		printf("pfr_get_astats: corruption detected (%d).\n",
624 		    w.pfrw_free);
625 		return (ENOTTY);
626 	}
627 	*size = kt->pfrkt_cnt;
628 	return (0);
629 }
630 
631 int
632 pfr_clr_astats(struct pfr_table *tbl, struct pfr_addr *addr, int size,
633     int *nzero, int flags)
634 {
635 	struct pfr_ktable	*kt;
636 	struct pfr_kentryworkq	 workq;
637 	struct pfr_kentry	*p;
638 	struct pfr_addr		*ad;
639 	int			 i, rv, xzero = 0;
640 
641 	PF_RULES_WASSERT();
642 
643 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY | PFR_FLAG_FEEDBACK);
644 	if (pfr_validate_table(tbl, 0, 0))
645 		return (EINVAL);
646 	kt = pfr_lookup_table(tbl);
647 	if (kt == NULL || !(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
648 		return (ESRCH);
649 	SLIST_INIT(&workq);
650 	for (i = 0, ad = addr; i < size; i++, ad++) {
651 		if (pfr_validate_addr(ad))
652 			senderr(EINVAL);
653 		p = pfr_lookup_addr(kt, ad, 1);
654 		if (flags & PFR_FLAG_FEEDBACK) {
655 			ad->pfra_fback = (p != NULL) ?
656 			    PFR_FB_CLEARED : PFR_FB_NONE;
657 		}
658 		if (p != NULL) {
659 			SLIST_INSERT_HEAD(&workq, p, pfrke_workq);
660 			xzero++;
661 		}
662 	}
663 
664 	if (!(flags & PFR_FLAG_DUMMY))
665 		pfr_clstats_kentries(&workq, 0, 0);
666 	if (nzero != NULL)
667 		*nzero = xzero;
668 	return (0);
669 _bad:
670 	if (flags & PFR_FLAG_FEEDBACK)
671 		pfr_reset_feedback(addr, size);
672 	return (rv);
673 }
674 
675 static int
676 pfr_validate_addr(struct pfr_addr *ad)
677 {
678 	int i;
679 
680 	switch (ad->pfra_af) {
681 #ifdef INET
682 	case AF_INET:
683 		if (ad->pfra_net > 32)
684 			return (-1);
685 		break;
686 #endif /* INET */
687 #ifdef INET6
688 	case AF_INET6:
689 		if (ad->pfra_net > 128)
690 			return (-1);
691 		break;
692 #endif /* INET6 */
693 	default:
694 		return (-1);
695 	}
696 	if (ad->pfra_net < 128 &&
697 		(((caddr_t)ad)[ad->pfra_net/8] & (0xFF >> (ad->pfra_net%8))))
698 			return (-1);
699 	for (i = (ad->pfra_net+7)/8; i < sizeof(ad->pfra_u); i++)
700 		if (((caddr_t)ad)[i])
701 			return (-1);
702 	if (ad->pfra_not && ad->pfra_not != 1)
703 		return (-1);
704 	if (ad->pfra_fback)
705 		return (-1);
706 	return (0);
707 }
708 
709 static void
710 pfr_enqueue_addrs(struct pfr_ktable *kt, struct pfr_kentryworkq *workq,
711 	int *naddr, int sweep)
712 {
713 	struct pfr_walktree	w;
714 
715 	SLIST_INIT(workq);
716 	bzero(&w, sizeof(w));
717 	w.pfrw_op = sweep ? PFRW_SWEEP : PFRW_ENQUEUE;
718 	w.pfrw_workq = workq;
719 	if (kt->pfrkt_ip4 != NULL)
720 		if (kt->pfrkt_ip4->rnh_walktree(&kt->pfrkt_ip4->rh,
721 		    pfr_walktree, &w))
722 			printf("pfr_enqueue_addrs: IPv4 walktree failed.\n");
723 	if (kt->pfrkt_ip6 != NULL)
724 		if (kt->pfrkt_ip6->rnh_walktree(&kt->pfrkt_ip6->rh,
725 		    pfr_walktree, &w))
726 			printf("pfr_enqueue_addrs: IPv6 walktree failed.\n");
727 	if (naddr != NULL)
728 		*naddr = w.pfrw_cnt;
729 }
730 
731 static void
732 pfr_mark_addrs(struct pfr_ktable *kt)
733 {
734 	struct pfr_walktree	w;
735 
736 	bzero(&w, sizeof(w));
737 	w.pfrw_op = PFRW_MARK;
738 	if (kt->pfrkt_ip4->rnh_walktree(&kt->pfrkt_ip4->rh, pfr_walktree, &w))
739 		printf("pfr_mark_addrs: IPv4 walktree failed.\n");
740 	if (kt->pfrkt_ip6->rnh_walktree(&kt->pfrkt_ip6->rh, pfr_walktree, &w))
741 		printf("pfr_mark_addrs: IPv6 walktree failed.\n");
742 }
743 
744 
745 static struct pfr_kentry *
746 pfr_lookup_addr(struct pfr_ktable *kt, struct pfr_addr *ad, int exact)
747 {
748 	union sockaddr_union	 sa, mask;
749 	struct radix_head	*head = NULL;
750 	struct pfr_kentry	*ke;
751 
752 	PF_RULES_ASSERT();
753 
754 	bzero(&sa, sizeof(sa));
755 	if (ad->pfra_af == AF_INET) {
756 		FILLIN_SIN(sa.sin, ad->pfra_ip4addr);
757 		head = &kt->pfrkt_ip4->rh;
758 	} else if ( ad->pfra_af == AF_INET6 ) {
759 		FILLIN_SIN6(sa.sin6, ad->pfra_ip6addr);
760 		head = &kt->pfrkt_ip6->rh;
761 	}
762 	if (ADDR_NETWORK(ad)) {
763 		pfr_prepare_network(&mask, ad->pfra_af, ad->pfra_net);
764 		ke = (struct pfr_kentry *)rn_lookup(&sa, &mask, head);
765 		if (ke && KENTRY_RNF_ROOT(ke))
766 			ke = NULL;
767 	} else {
768 		ke = (struct pfr_kentry *)rn_match(&sa, head);
769 		if (ke && KENTRY_RNF_ROOT(ke))
770 			ke = NULL;
771 		if (exact && ke && KENTRY_NETWORK(ke))
772 			ke = NULL;
773 	}
774 	return (ke);
775 }
776 
777 static struct pfr_kentry *
778 pfr_create_kentry(struct pfr_addr *ad)
779 {
780 	struct pfr_kentry	*ke;
781 
782 	ke =  uma_zalloc(V_pfr_kentry_z, M_NOWAIT | M_ZERO);
783 	if (ke == NULL)
784 		return (NULL);
785 
786 	if (ad->pfra_af == AF_INET)
787 		FILLIN_SIN(ke->pfrke_sa.sin, ad->pfra_ip4addr);
788 	else if (ad->pfra_af == AF_INET6)
789 		FILLIN_SIN6(ke->pfrke_sa.sin6, ad->pfra_ip6addr);
790 	ke->pfrke_af = ad->pfra_af;
791 	ke->pfrke_net = ad->pfra_net;
792 	ke->pfrke_not = ad->pfra_not;
793 	return (ke);
794 }
795 
796 static void
797 pfr_destroy_kentries(struct pfr_kentryworkq *workq)
798 {
799 	struct pfr_kentry	*p, *q;
800 
801 	for (p = SLIST_FIRST(workq); p != NULL; p = q) {
802 		q = SLIST_NEXT(p, pfrke_workq);
803 		pfr_destroy_kentry(p);
804 	}
805 }
806 
807 static void
808 pfr_destroy_kentry(struct pfr_kentry *ke)
809 {
810 	if (ke->pfrke_counters)
811 		uma_zfree(V_pfr_kcounters_z, ke->pfrke_counters);
812 	uma_zfree(V_pfr_kentry_z, ke);
813 }
814 
815 static void
816 pfr_insert_kentries(struct pfr_ktable *kt,
817     struct pfr_kentryworkq *workq, long tzero)
818 {
819 	struct pfr_kentry	*p;
820 	int			 rv, n = 0;
821 
822 	SLIST_FOREACH(p, workq, pfrke_workq) {
823 		rv = pfr_route_kentry(kt, p);
824 		if (rv) {
825 			printf("pfr_insert_kentries: cannot route entry "
826 			    "(code=%d).\n", rv);
827 			break;
828 		}
829 		p->pfrke_tzero = tzero;
830 		n++;
831 	}
832 	kt->pfrkt_cnt += n;
833 }
834 
835 int
836 pfr_insert_kentry(struct pfr_ktable *kt, struct pfr_addr *ad, long tzero)
837 {
838 	struct pfr_kentry	*p;
839 	int			 rv;
840 
841 	p = pfr_lookup_addr(kt, ad, 1);
842 	if (p != NULL)
843 		return (0);
844 	p = pfr_create_kentry(ad);
845 	if (p == NULL)
846 		return (ENOMEM);
847 
848 	rv = pfr_route_kentry(kt, p);
849 	if (rv)
850 		return (rv);
851 
852 	p->pfrke_tzero = tzero;
853 	kt->pfrkt_cnt++;
854 
855 	return (0);
856 }
857 
858 static void
859 pfr_remove_kentries(struct pfr_ktable *kt,
860     struct pfr_kentryworkq *workq)
861 {
862 	struct pfr_kentry	*p;
863 	int			 n = 0;
864 
865 	SLIST_FOREACH(p, workq, pfrke_workq) {
866 		pfr_unroute_kentry(kt, p);
867 		n++;
868 	}
869 	kt->pfrkt_cnt -= n;
870 	pfr_destroy_kentries(workq);
871 }
872 
873 static void
874 pfr_clean_node_mask(struct pfr_ktable *kt,
875     struct pfr_kentryworkq *workq)
876 {
877 	struct pfr_kentry	*p;
878 
879 	SLIST_FOREACH(p, workq, pfrke_workq)
880 		pfr_unroute_kentry(kt, p);
881 }
882 
883 static void
884 pfr_clstats_kentries(struct pfr_kentryworkq *workq, long tzero, int negchange)
885 {
886 	struct pfr_kentry	*p;
887 
888 	SLIST_FOREACH(p, workq, pfrke_workq) {
889 		if (negchange)
890 			p->pfrke_not = !p->pfrke_not;
891 		if (p->pfrke_counters) {
892 			uma_zfree(V_pfr_kcounters_z, p->pfrke_counters);
893 			p->pfrke_counters = NULL;
894 		}
895 		p->pfrke_tzero = tzero;
896 	}
897 }
898 
899 static void
900 pfr_reset_feedback(struct pfr_addr *addr, int size)
901 {
902 	struct pfr_addr	*ad;
903 	int		i;
904 
905 	for (i = 0, ad = addr; i < size; i++, ad++)
906 		ad->pfra_fback = PFR_FB_NONE;
907 }
908 
909 static void
910 pfr_prepare_network(union sockaddr_union *sa, int af, int net)
911 {
912 	int	i;
913 
914 	bzero(sa, sizeof(*sa));
915 	if (af == AF_INET) {
916 		sa->sin.sin_len = sizeof(sa->sin);
917 		sa->sin.sin_family = AF_INET;
918 		sa->sin.sin_addr.s_addr = net ? htonl(-1 << (32-net)) : 0;
919 	} else if (af == AF_INET6) {
920 		sa->sin6.sin6_len = sizeof(sa->sin6);
921 		sa->sin6.sin6_family = AF_INET6;
922 		for (i = 0; i < 4; i++) {
923 			if (net <= 32) {
924 				sa->sin6.sin6_addr.s6_addr32[i] =
925 				    net ? htonl(-1 << (32-net)) : 0;
926 				break;
927 			}
928 			sa->sin6.sin6_addr.s6_addr32[i] = 0xFFFFFFFF;
929 			net -= 32;
930 		}
931 	}
932 }
933 
934 static int
935 pfr_route_kentry(struct pfr_ktable *kt, struct pfr_kentry *ke)
936 {
937 	union sockaddr_union	 mask;
938 	struct radix_node	*rn;
939 	struct radix_head	*head = NULL;
940 
941 	PF_RULES_WASSERT();
942 
943 	bzero(ke->pfrke_node, sizeof(ke->pfrke_node));
944 	if (ke->pfrke_af == AF_INET)
945 		head = &kt->pfrkt_ip4->rh;
946 	else if (ke->pfrke_af == AF_INET6)
947 		head = &kt->pfrkt_ip6->rh;
948 
949 	if (KENTRY_NETWORK(ke)) {
950 		pfr_prepare_network(&mask, ke->pfrke_af, ke->pfrke_net);
951 		rn = rn_addroute(&ke->pfrke_sa, &mask, head, ke->pfrke_node);
952 	} else
953 		rn = rn_addroute(&ke->pfrke_sa, NULL, head, ke->pfrke_node);
954 
955 	return (rn == NULL ? -1 : 0);
956 }
957 
958 static int
959 pfr_unroute_kentry(struct pfr_ktable *kt, struct pfr_kentry *ke)
960 {
961 	union sockaddr_union	 mask;
962 	struct radix_node	*rn;
963 	struct radix_head	*head = NULL;
964 
965 	if (ke->pfrke_af == AF_INET)
966 		head = &kt->pfrkt_ip4->rh;
967 	else if (ke->pfrke_af == AF_INET6)
968 		head = &kt->pfrkt_ip6->rh;
969 
970 	if (KENTRY_NETWORK(ke)) {
971 		pfr_prepare_network(&mask, ke->pfrke_af, ke->pfrke_net);
972 		rn = rn_delete(&ke->pfrke_sa, &mask, head);
973 	} else
974 		rn = rn_delete(&ke->pfrke_sa, NULL, head);
975 
976 	if (rn == NULL) {
977 		printf("pfr_unroute_kentry: delete failed.\n");
978 		return (-1);
979 	}
980 	return (0);
981 }
982 
983 static void
984 pfr_copyout_addr(struct pfr_addr *ad, struct pfr_kentry *ke)
985 {
986 	bzero(ad, sizeof(*ad));
987 	if (ke == NULL)
988 		return;
989 	ad->pfra_af = ke->pfrke_af;
990 	ad->pfra_net = ke->pfrke_net;
991 	ad->pfra_not = ke->pfrke_not;
992 	if (ad->pfra_af == AF_INET)
993 		ad->pfra_ip4addr = ke->pfrke_sa.sin.sin_addr;
994 	else if (ad->pfra_af == AF_INET6)
995 		ad->pfra_ip6addr = ke->pfrke_sa.sin6.sin6_addr;
996 }
997 
998 static int
999 pfr_walktree(struct radix_node *rn, void *arg)
1000 {
1001 	struct pfr_kentry	*ke = (struct pfr_kentry *)rn;
1002 	struct pfr_walktree	*w = arg;
1003 
1004 	switch (w->pfrw_op) {
1005 	case PFRW_MARK:
1006 		ke->pfrke_mark = 0;
1007 		break;
1008 	case PFRW_SWEEP:
1009 		if (ke->pfrke_mark)
1010 			break;
1011 		/* FALLTHROUGH */
1012 	case PFRW_ENQUEUE:
1013 		SLIST_INSERT_HEAD(w->pfrw_workq, ke, pfrke_workq);
1014 		w->pfrw_cnt++;
1015 		break;
1016 	case PFRW_GET_ADDRS:
1017 		if (w->pfrw_free-- > 0) {
1018 			pfr_copyout_addr(w->pfrw_addr, ke);
1019 			w->pfrw_addr++;
1020 		}
1021 		break;
1022 	case PFRW_GET_ASTATS:
1023 		if (w->pfrw_free-- > 0) {
1024 			struct pfr_astats as;
1025 
1026 			pfr_copyout_addr(&as.pfras_a, ke);
1027 
1028 			if (ke->pfrke_counters) {
1029 				bcopy(ke->pfrke_counters->pfrkc_packets,
1030 				    as.pfras_packets, sizeof(as.pfras_packets));
1031 				bcopy(ke->pfrke_counters->pfrkc_bytes,
1032 				    as.pfras_bytes, sizeof(as.pfras_bytes));
1033 			} else {
1034 				bzero(as.pfras_packets, sizeof(as.pfras_packets));
1035 				bzero(as.pfras_bytes, sizeof(as.pfras_bytes));
1036 				as.pfras_a.pfra_fback = PFR_FB_NOCOUNT;
1037 			}
1038 			as.pfras_tzero = ke->pfrke_tzero;
1039 
1040 			bcopy(&as, w->pfrw_astats, sizeof(as));
1041 			w->pfrw_astats++;
1042 		}
1043 		break;
1044 	case PFRW_POOL_GET:
1045 		if (ke->pfrke_not)
1046 			break; /* negative entries are ignored */
1047 		if (!w->pfrw_cnt--) {
1048 			w->pfrw_kentry = ke;
1049 			return (1); /* finish search */
1050 		}
1051 		break;
1052 	case PFRW_DYNADDR_UPDATE:
1053 	    {
1054 		union sockaddr_union	pfr_mask;
1055 
1056 		if (ke->pfrke_af == AF_INET) {
1057 			if (w->pfrw_dyn->pfid_acnt4++ > 0)
1058 				break;
1059 			pfr_prepare_network(&pfr_mask, AF_INET, ke->pfrke_net);
1060 			w->pfrw_dyn->pfid_addr4 = *SUNION2PF(&ke->pfrke_sa,
1061 			    AF_INET);
1062 			w->pfrw_dyn->pfid_mask4 = *SUNION2PF(&pfr_mask,
1063 			    AF_INET);
1064 		} else if (ke->pfrke_af == AF_INET6){
1065 			if (w->pfrw_dyn->pfid_acnt6++ > 0)
1066 				break;
1067 			pfr_prepare_network(&pfr_mask, AF_INET6, ke->pfrke_net);
1068 			w->pfrw_dyn->pfid_addr6 = *SUNION2PF(&ke->pfrke_sa,
1069 			    AF_INET6);
1070 			w->pfrw_dyn->pfid_mask6 = *SUNION2PF(&pfr_mask,
1071 			    AF_INET6);
1072 		}
1073 		break;
1074 	    }
1075 	}
1076 	return (0);
1077 }
1078 
1079 int
1080 pfr_clr_tables(struct pfr_table *filter, int *ndel, int flags)
1081 {
1082 	struct pfr_ktableworkq	 workq;
1083 	struct pfr_ktable	*p;
1084 	int			 xdel = 0;
1085 
1086 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY | PFR_FLAG_ALLRSETS);
1087 	if (pfr_fix_anchor(filter->pfrt_anchor))
1088 		return (EINVAL);
1089 	if (pfr_table_count(filter, flags) < 0)
1090 		return (ENOENT);
1091 
1092 	SLIST_INIT(&workq);
1093 	RB_FOREACH(p, pfr_ktablehead, &V_pfr_ktables) {
1094 		if (pfr_skip_table(filter, p, flags))
1095 			continue;
1096 		if (!strcmp(p->pfrkt_anchor, PF_RESERVED_ANCHOR))
1097 			continue;
1098 		if (!(p->pfrkt_flags & PFR_TFLAG_ACTIVE))
1099 			continue;
1100 		p->pfrkt_nflags = p->pfrkt_flags & ~PFR_TFLAG_ACTIVE;
1101 		SLIST_INSERT_HEAD(&workq, p, pfrkt_workq);
1102 		xdel++;
1103 	}
1104 	if (!(flags & PFR_FLAG_DUMMY))
1105 		pfr_setflags_ktables(&workq);
1106 	if (ndel != NULL)
1107 		*ndel = xdel;
1108 	return (0);
1109 }
1110 
1111 int
1112 pfr_add_tables(struct pfr_table *tbl, int size, int *nadd, int flags)
1113 {
1114 	struct pfr_ktableworkq	 addq, changeq;
1115 	struct pfr_ktable	*p, *q, *r, key;
1116 	int			 i, rv, xadd = 0;
1117 	long			 tzero = time_second;
1118 
1119 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY);
1120 	SLIST_INIT(&addq);
1121 	SLIST_INIT(&changeq);
1122 	for (i = 0; i < size; i++) {
1123 		bcopy(tbl+i, &key.pfrkt_t, sizeof(key.pfrkt_t));
1124 		if (pfr_validate_table(&key.pfrkt_t, PFR_TFLAG_USRMASK,
1125 		    flags & PFR_FLAG_USERIOCTL))
1126 			senderr(EINVAL);
1127 		key.pfrkt_flags |= PFR_TFLAG_ACTIVE;
1128 		p = RB_FIND(pfr_ktablehead, &V_pfr_ktables, &key);
1129 		if (p == NULL) {
1130 			p = pfr_create_ktable(&key.pfrkt_t, tzero, 1);
1131 			if (p == NULL)
1132 				senderr(ENOMEM);
1133 			SLIST_FOREACH(q, &addq, pfrkt_workq) {
1134 				if (!pfr_ktable_compare(p, q))
1135 					goto _skip;
1136 			}
1137 			SLIST_INSERT_HEAD(&addq, p, pfrkt_workq);
1138 			xadd++;
1139 			if (!key.pfrkt_anchor[0])
1140 				goto _skip;
1141 
1142 			/* find or create root table */
1143 			bzero(key.pfrkt_anchor, sizeof(key.pfrkt_anchor));
1144 			r = RB_FIND(pfr_ktablehead, &V_pfr_ktables, &key);
1145 			if (r != NULL) {
1146 				p->pfrkt_root = r;
1147 				goto _skip;
1148 			}
1149 			SLIST_FOREACH(q, &addq, pfrkt_workq) {
1150 				if (!pfr_ktable_compare(&key, q)) {
1151 					p->pfrkt_root = q;
1152 					goto _skip;
1153 				}
1154 			}
1155 			key.pfrkt_flags = 0;
1156 			r = pfr_create_ktable(&key.pfrkt_t, 0, 1);
1157 			if (r == NULL)
1158 				senderr(ENOMEM);
1159 			SLIST_INSERT_HEAD(&addq, r, pfrkt_workq);
1160 			p->pfrkt_root = r;
1161 		} else if (!(p->pfrkt_flags & PFR_TFLAG_ACTIVE)) {
1162 			SLIST_FOREACH(q, &changeq, pfrkt_workq)
1163 				if (!pfr_ktable_compare(&key, q))
1164 					goto _skip;
1165 			p->pfrkt_nflags = (p->pfrkt_flags &
1166 			    ~PFR_TFLAG_USRMASK) | key.pfrkt_flags;
1167 			SLIST_INSERT_HEAD(&changeq, p, pfrkt_workq);
1168 			xadd++;
1169 		}
1170 _skip:
1171 	;
1172 	}
1173 	if (!(flags & PFR_FLAG_DUMMY)) {
1174 		pfr_insert_ktables(&addq);
1175 		pfr_setflags_ktables(&changeq);
1176 	} else
1177 		 pfr_destroy_ktables(&addq, 0);
1178 	if (nadd != NULL)
1179 		*nadd = xadd;
1180 	return (0);
1181 _bad:
1182 	pfr_destroy_ktables(&addq, 0);
1183 	return (rv);
1184 }
1185 
1186 int
1187 pfr_del_tables(struct pfr_table *tbl, int size, int *ndel, int flags)
1188 {
1189 	struct pfr_ktableworkq	 workq;
1190 	struct pfr_ktable	*p, *q, key;
1191 	int			 i, xdel = 0;
1192 
1193 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY);
1194 	SLIST_INIT(&workq);
1195 	for (i = 0; i < size; i++) {
1196 		bcopy(tbl+i, &key.pfrkt_t, sizeof(key.pfrkt_t));
1197 		if (pfr_validate_table(&key.pfrkt_t, 0,
1198 		    flags & PFR_FLAG_USERIOCTL))
1199 			return (EINVAL);
1200 		p = RB_FIND(pfr_ktablehead, &V_pfr_ktables, &key);
1201 		if (p != NULL && (p->pfrkt_flags & PFR_TFLAG_ACTIVE)) {
1202 			SLIST_FOREACH(q, &workq, pfrkt_workq)
1203 				if (!pfr_ktable_compare(p, q))
1204 					goto _skip;
1205 			p->pfrkt_nflags = p->pfrkt_flags & ~PFR_TFLAG_ACTIVE;
1206 			SLIST_INSERT_HEAD(&workq, p, pfrkt_workq);
1207 			xdel++;
1208 		}
1209 _skip:
1210 	;
1211 	}
1212 
1213 	if (!(flags & PFR_FLAG_DUMMY))
1214 		pfr_setflags_ktables(&workq);
1215 	if (ndel != NULL)
1216 		*ndel = xdel;
1217 	return (0);
1218 }
1219 
1220 int
1221 pfr_get_tables(struct pfr_table *filter, struct pfr_table *tbl, int *size,
1222 	int flags)
1223 {
1224 	struct pfr_ktable	*p;
1225 	int			 n, nn;
1226 
1227 	PF_RULES_RASSERT();
1228 
1229 	ACCEPT_FLAGS(flags, PFR_FLAG_ALLRSETS);
1230 	if (pfr_fix_anchor(filter->pfrt_anchor))
1231 		return (EINVAL);
1232 	n = nn = pfr_table_count(filter, flags);
1233 	if (n < 0)
1234 		return (ENOENT);
1235 	if (n > *size) {
1236 		*size = n;
1237 		return (0);
1238 	}
1239 	RB_FOREACH(p, pfr_ktablehead, &V_pfr_ktables) {
1240 		if (pfr_skip_table(filter, p, flags))
1241 			continue;
1242 		if (n-- <= 0)
1243 			continue;
1244 		bcopy(&p->pfrkt_t, tbl++, sizeof(*tbl));
1245 	}
1246 
1247 	KASSERT(n == 0, ("%s: corruption detected (%d)", __func__, n));
1248 
1249 	*size = nn;
1250 	return (0);
1251 }
1252 
1253 int
1254 pfr_get_tstats(struct pfr_table *filter, struct pfr_tstats *tbl, int *size,
1255 	int flags)
1256 {
1257 	struct pfr_ktable	*p;
1258 	struct pfr_ktableworkq	 workq;
1259 	int			 n, nn;
1260 	long			 tzero = time_second;
1261 
1262 	/* XXX PFR_FLAG_CLSTATS disabled */
1263 	ACCEPT_FLAGS(flags, PFR_FLAG_ALLRSETS);
1264 	if (pfr_fix_anchor(filter->pfrt_anchor))
1265 		return (EINVAL);
1266 	n = nn = pfr_table_count(filter, flags);
1267 	if (n < 0)
1268 		return (ENOENT);
1269 	if (n > *size) {
1270 		*size = n;
1271 		return (0);
1272 	}
1273 	SLIST_INIT(&workq);
1274 	RB_FOREACH(p, pfr_ktablehead, &V_pfr_ktables) {
1275 		if (pfr_skip_table(filter, p, flags))
1276 			continue;
1277 		if (n-- <= 0)
1278 			continue;
1279 		bcopy(&p->pfrkt_ts, tbl++, sizeof(*tbl));
1280 		SLIST_INSERT_HEAD(&workq, p, pfrkt_workq);
1281 	}
1282 	if (flags & PFR_FLAG_CLSTATS)
1283 		pfr_clstats_ktables(&workq, tzero,
1284 		    flags & PFR_FLAG_ADDRSTOO);
1285 
1286 	KASSERT(n == 0, ("%s: corruption detected (%d)", __func__, n));
1287 
1288 	*size = nn;
1289 	return (0);
1290 }
1291 
1292 int
1293 pfr_clr_tstats(struct pfr_table *tbl, int size, int *nzero, int flags)
1294 {
1295 	struct pfr_ktableworkq	 workq;
1296 	struct pfr_ktable	*p, key;
1297 	int			 i, xzero = 0;
1298 	long			 tzero = time_second;
1299 
1300 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY | PFR_FLAG_ADDRSTOO);
1301 	SLIST_INIT(&workq);
1302 	for (i = 0; i < size; i++) {
1303 		bcopy(tbl + i, &key.pfrkt_t, sizeof(key.pfrkt_t));
1304 		if (pfr_validate_table(&key.pfrkt_t, 0, 0))
1305 			return (EINVAL);
1306 		p = RB_FIND(pfr_ktablehead, &V_pfr_ktables, &key);
1307 		if (p != NULL) {
1308 			SLIST_INSERT_HEAD(&workq, p, pfrkt_workq);
1309 			xzero++;
1310 		}
1311 	}
1312 	if (!(flags & PFR_FLAG_DUMMY))
1313 		pfr_clstats_ktables(&workq, tzero, flags & PFR_FLAG_ADDRSTOO);
1314 	if (nzero != NULL)
1315 		*nzero = xzero;
1316 	return (0);
1317 }
1318 
1319 int
1320 pfr_set_tflags(struct pfr_table *tbl, int size, int setflag, int clrflag,
1321 	int *nchange, int *ndel, int flags)
1322 {
1323 	struct pfr_ktableworkq	 workq;
1324 	struct pfr_ktable	*p, *q, key;
1325 	int			 i, xchange = 0, xdel = 0;
1326 
1327 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY);
1328 	if ((setflag & ~PFR_TFLAG_USRMASK) ||
1329 	    (clrflag & ~PFR_TFLAG_USRMASK) ||
1330 	    (setflag & clrflag))
1331 		return (EINVAL);
1332 	SLIST_INIT(&workq);
1333 	for (i = 0; i < size; i++) {
1334 		bcopy(tbl + i, &key.pfrkt_t, sizeof(key.pfrkt_t));
1335 		if (pfr_validate_table(&key.pfrkt_t, 0,
1336 		    flags & PFR_FLAG_USERIOCTL))
1337 			return (EINVAL);
1338 		p = RB_FIND(pfr_ktablehead, &V_pfr_ktables, &key);
1339 		if (p != NULL && (p->pfrkt_flags & PFR_TFLAG_ACTIVE)) {
1340 			p->pfrkt_nflags = (p->pfrkt_flags | setflag) &
1341 			    ~clrflag;
1342 			if (p->pfrkt_nflags == p->pfrkt_flags)
1343 				goto _skip;
1344 			SLIST_FOREACH(q, &workq, pfrkt_workq)
1345 				if (!pfr_ktable_compare(p, q))
1346 					goto _skip;
1347 			SLIST_INSERT_HEAD(&workq, p, pfrkt_workq);
1348 			if ((p->pfrkt_flags & PFR_TFLAG_PERSIST) &&
1349 			    (clrflag & PFR_TFLAG_PERSIST) &&
1350 			    !(p->pfrkt_flags & PFR_TFLAG_REFERENCED))
1351 				xdel++;
1352 			else
1353 				xchange++;
1354 		}
1355 _skip:
1356 	;
1357 	}
1358 	if (!(flags & PFR_FLAG_DUMMY))
1359 		pfr_setflags_ktables(&workq);
1360 	if (nchange != NULL)
1361 		*nchange = xchange;
1362 	if (ndel != NULL)
1363 		*ndel = xdel;
1364 	return (0);
1365 }
1366 
1367 int
1368 pfr_ina_begin(struct pfr_table *trs, u_int32_t *ticket, int *ndel, int flags)
1369 {
1370 	struct pfr_ktableworkq	 workq;
1371 	struct pfr_ktable	*p;
1372 	struct pf_ruleset	*rs;
1373 	int			 xdel = 0;
1374 
1375 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY);
1376 	rs = pf_find_or_create_ruleset(trs->pfrt_anchor);
1377 	if (rs == NULL)
1378 		return (ENOMEM);
1379 	SLIST_INIT(&workq);
1380 	RB_FOREACH(p, pfr_ktablehead, &V_pfr_ktables) {
1381 		if (!(p->pfrkt_flags & PFR_TFLAG_INACTIVE) ||
1382 		    pfr_skip_table(trs, p, 0))
1383 			continue;
1384 		p->pfrkt_nflags = p->pfrkt_flags & ~PFR_TFLAG_INACTIVE;
1385 		SLIST_INSERT_HEAD(&workq, p, pfrkt_workq);
1386 		xdel++;
1387 	}
1388 	if (!(flags & PFR_FLAG_DUMMY)) {
1389 		pfr_setflags_ktables(&workq);
1390 		if (ticket != NULL)
1391 			*ticket = ++rs->tticket;
1392 		rs->topen = 1;
1393 	} else
1394 		pf_remove_if_empty_ruleset(rs);
1395 	if (ndel != NULL)
1396 		*ndel = xdel;
1397 	return (0);
1398 }
1399 
1400 int
1401 pfr_ina_define(struct pfr_table *tbl, struct pfr_addr *addr, int size,
1402     int *nadd, int *naddr, u_int32_t ticket, int flags)
1403 {
1404 	struct pfr_ktableworkq	 tableq;
1405 	struct pfr_kentryworkq	 addrq;
1406 	struct pfr_ktable	*kt, *rt, *shadow, key;
1407 	struct pfr_kentry	*p;
1408 	struct pfr_addr		*ad;
1409 	struct pf_ruleset	*rs;
1410 	int			 i, rv, xadd = 0, xaddr = 0;
1411 
1412 	PF_RULES_WASSERT();
1413 
1414 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY | PFR_FLAG_ADDRSTOO);
1415 	if (size && !(flags & PFR_FLAG_ADDRSTOO))
1416 		return (EINVAL);
1417 	if (pfr_validate_table(tbl, PFR_TFLAG_USRMASK,
1418 	    flags & PFR_FLAG_USERIOCTL))
1419 		return (EINVAL);
1420 	rs = pf_find_ruleset(tbl->pfrt_anchor);
1421 	if (rs == NULL || !rs->topen || ticket != rs->tticket)
1422 		return (EBUSY);
1423 	tbl->pfrt_flags |= PFR_TFLAG_INACTIVE;
1424 	SLIST_INIT(&tableq);
1425 	kt = RB_FIND(pfr_ktablehead, &V_pfr_ktables, (struct pfr_ktable *)tbl);
1426 	if (kt == NULL) {
1427 		kt = pfr_create_ktable(tbl, 0, 1);
1428 		if (kt == NULL)
1429 			return (ENOMEM);
1430 		SLIST_INSERT_HEAD(&tableq, kt, pfrkt_workq);
1431 		xadd++;
1432 		if (!tbl->pfrt_anchor[0])
1433 			goto _skip;
1434 
1435 		/* find or create root table */
1436 		bzero(&key, sizeof(key));
1437 		strlcpy(key.pfrkt_name, tbl->pfrt_name, sizeof(key.pfrkt_name));
1438 		rt = RB_FIND(pfr_ktablehead, &V_pfr_ktables, &key);
1439 		if (rt != NULL) {
1440 			kt->pfrkt_root = rt;
1441 			goto _skip;
1442 		}
1443 		rt = pfr_create_ktable(&key.pfrkt_t, 0, 1);
1444 		if (rt == NULL) {
1445 			pfr_destroy_ktables(&tableq, 0);
1446 			return (ENOMEM);
1447 		}
1448 		SLIST_INSERT_HEAD(&tableq, rt, pfrkt_workq);
1449 		kt->pfrkt_root = rt;
1450 	} else if (!(kt->pfrkt_flags & PFR_TFLAG_INACTIVE))
1451 		xadd++;
1452 _skip:
1453 	shadow = pfr_create_ktable(tbl, 0, 0);
1454 	if (shadow == NULL) {
1455 		pfr_destroy_ktables(&tableq, 0);
1456 		return (ENOMEM);
1457 	}
1458 	SLIST_INIT(&addrq);
1459 	for (i = 0, ad = addr; i < size; i++, ad++) {
1460 		if (pfr_validate_addr(ad))
1461 			senderr(EINVAL);
1462 		if (pfr_lookup_addr(shadow, ad, 1) != NULL)
1463 			continue;
1464 		p = pfr_create_kentry(ad);
1465 		if (p == NULL)
1466 			senderr(ENOMEM);
1467 		if (pfr_route_kentry(shadow, p)) {
1468 			pfr_destroy_kentry(p);
1469 			continue;
1470 		}
1471 		SLIST_INSERT_HEAD(&addrq, p, pfrke_workq);
1472 		xaddr++;
1473 	}
1474 	if (!(flags & PFR_FLAG_DUMMY)) {
1475 		if (kt->pfrkt_shadow != NULL)
1476 			pfr_destroy_ktable(kt->pfrkt_shadow, 1);
1477 		kt->pfrkt_flags |= PFR_TFLAG_INACTIVE;
1478 		pfr_insert_ktables(&tableq);
1479 		shadow->pfrkt_cnt = (flags & PFR_FLAG_ADDRSTOO) ?
1480 		    xaddr : NO_ADDRESSES;
1481 		kt->pfrkt_shadow = shadow;
1482 	} else {
1483 		pfr_clean_node_mask(shadow, &addrq);
1484 		pfr_destroy_ktable(shadow, 0);
1485 		pfr_destroy_ktables(&tableq, 0);
1486 		pfr_destroy_kentries(&addrq);
1487 	}
1488 	if (nadd != NULL)
1489 		*nadd = xadd;
1490 	if (naddr != NULL)
1491 		*naddr = xaddr;
1492 	return (0);
1493 _bad:
1494 	pfr_destroy_ktable(shadow, 0);
1495 	pfr_destroy_ktables(&tableq, 0);
1496 	pfr_destroy_kentries(&addrq);
1497 	return (rv);
1498 }
1499 
1500 int
1501 pfr_ina_rollback(struct pfr_table *trs, u_int32_t ticket, int *ndel, int flags)
1502 {
1503 	struct pfr_ktableworkq	 workq;
1504 	struct pfr_ktable	*p;
1505 	struct pf_ruleset	*rs;
1506 	int			 xdel = 0;
1507 
1508 	PF_RULES_WASSERT();
1509 
1510 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY);
1511 	rs = pf_find_ruleset(trs->pfrt_anchor);
1512 	if (rs == NULL || !rs->topen || ticket != rs->tticket)
1513 		return (0);
1514 	SLIST_INIT(&workq);
1515 	RB_FOREACH(p, pfr_ktablehead, &V_pfr_ktables) {
1516 		if (!(p->pfrkt_flags & PFR_TFLAG_INACTIVE) ||
1517 		    pfr_skip_table(trs, p, 0))
1518 			continue;
1519 		p->pfrkt_nflags = p->pfrkt_flags & ~PFR_TFLAG_INACTIVE;
1520 		SLIST_INSERT_HEAD(&workq, p, pfrkt_workq);
1521 		xdel++;
1522 	}
1523 	if (!(flags & PFR_FLAG_DUMMY)) {
1524 		pfr_setflags_ktables(&workq);
1525 		rs->topen = 0;
1526 		pf_remove_if_empty_ruleset(rs);
1527 	}
1528 	if (ndel != NULL)
1529 		*ndel = xdel;
1530 	return (0);
1531 }
1532 
1533 int
1534 pfr_ina_commit(struct pfr_table *trs, u_int32_t ticket, int *nadd,
1535     int *nchange, int flags)
1536 {
1537 	struct pfr_ktable	*p, *q;
1538 	struct pfr_ktableworkq	 workq;
1539 	struct pf_ruleset	*rs;
1540 	int			 xadd = 0, xchange = 0;
1541 	long			 tzero = time_second;
1542 
1543 	PF_RULES_WASSERT();
1544 
1545 	ACCEPT_FLAGS(flags, PFR_FLAG_DUMMY);
1546 	rs = pf_find_ruleset(trs->pfrt_anchor);
1547 	if (rs == NULL || !rs->topen || ticket != rs->tticket)
1548 		return (EBUSY);
1549 
1550 	SLIST_INIT(&workq);
1551 	RB_FOREACH(p, pfr_ktablehead, &V_pfr_ktables) {
1552 		if (!(p->pfrkt_flags & PFR_TFLAG_INACTIVE) ||
1553 		    pfr_skip_table(trs, p, 0))
1554 			continue;
1555 		SLIST_INSERT_HEAD(&workq, p, pfrkt_workq);
1556 		if (p->pfrkt_flags & PFR_TFLAG_ACTIVE)
1557 			xchange++;
1558 		else
1559 			xadd++;
1560 	}
1561 
1562 	if (!(flags & PFR_FLAG_DUMMY)) {
1563 		for (p = SLIST_FIRST(&workq); p != NULL; p = q) {
1564 			q = SLIST_NEXT(p, pfrkt_workq);
1565 			pfr_commit_ktable(p, tzero);
1566 		}
1567 		rs->topen = 0;
1568 		pf_remove_if_empty_ruleset(rs);
1569 	}
1570 	if (nadd != NULL)
1571 		*nadd = xadd;
1572 	if (nchange != NULL)
1573 		*nchange = xchange;
1574 
1575 	return (0);
1576 }
1577 
1578 static void
1579 pfr_commit_ktable(struct pfr_ktable *kt, long tzero)
1580 {
1581 	struct pfr_ktable	*shadow = kt->pfrkt_shadow;
1582 	int			 nflags;
1583 
1584 	PF_RULES_WASSERT();
1585 
1586 	if (shadow->pfrkt_cnt == NO_ADDRESSES) {
1587 		if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
1588 			pfr_clstats_ktable(kt, tzero, 1);
1589 	} else if (kt->pfrkt_flags & PFR_TFLAG_ACTIVE) {
1590 		/* kt might contain addresses */
1591 		struct pfr_kentryworkq	 addrq, addq, changeq, delq, garbageq;
1592 		struct pfr_kentry	*p, *q, *next;
1593 		struct pfr_addr		 ad;
1594 
1595 		pfr_enqueue_addrs(shadow, &addrq, NULL, 0);
1596 		pfr_mark_addrs(kt);
1597 		SLIST_INIT(&addq);
1598 		SLIST_INIT(&changeq);
1599 		SLIST_INIT(&delq);
1600 		SLIST_INIT(&garbageq);
1601 		pfr_clean_node_mask(shadow, &addrq);
1602 		for (p = SLIST_FIRST(&addrq); p != NULL; p = next) {
1603 			next = SLIST_NEXT(p, pfrke_workq);	/* XXX */
1604 			pfr_copyout_addr(&ad, p);
1605 			q = pfr_lookup_addr(kt, &ad, 1);
1606 			if (q != NULL) {
1607 				if (q->pfrke_not != p->pfrke_not)
1608 					SLIST_INSERT_HEAD(&changeq, q,
1609 					    pfrke_workq);
1610 				q->pfrke_mark = 1;
1611 				SLIST_INSERT_HEAD(&garbageq, p, pfrke_workq);
1612 			} else {
1613 				p->pfrke_tzero = tzero;
1614 				SLIST_INSERT_HEAD(&addq, p, pfrke_workq);
1615 			}
1616 		}
1617 		pfr_enqueue_addrs(kt, &delq, NULL, ENQUEUE_UNMARKED_ONLY);
1618 		pfr_insert_kentries(kt, &addq, tzero);
1619 		pfr_remove_kentries(kt, &delq);
1620 		pfr_clstats_kentries(&changeq, tzero, INVERT_NEG_FLAG);
1621 		pfr_destroy_kentries(&garbageq);
1622 	} else {
1623 		/* kt cannot contain addresses */
1624 		SWAP(struct radix_node_head *, kt->pfrkt_ip4,
1625 		    shadow->pfrkt_ip4);
1626 		SWAP(struct radix_node_head *, kt->pfrkt_ip6,
1627 		    shadow->pfrkt_ip6);
1628 		SWAP(int, kt->pfrkt_cnt, shadow->pfrkt_cnt);
1629 		pfr_clstats_ktable(kt, tzero, 1);
1630 	}
1631 	nflags = ((shadow->pfrkt_flags & PFR_TFLAG_USRMASK) |
1632 	    (kt->pfrkt_flags & PFR_TFLAG_SETMASK) | PFR_TFLAG_ACTIVE)
1633 		& ~PFR_TFLAG_INACTIVE;
1634 	pfr_destroy_ktable(shadow, 0);
1635 	kt->pfrkt_shadow = NULL;
1636 	pfr_setflags_ktable(kt, nflags);
1637 }
1638 
1639 static int
1640 pfr_validate_table(struct pfr_table *tbl, int allowedflags, int no_reserved)
1641 {
1642 	int i;
1643 
1644 	if (!tbl->pfrt_name[0])
1645 		return (-1);
1646 	if (no_reserved && !strcmp(tbl->pfrt_anchor, PF_RESERVED_ANCHOR))
1647 		 return (-1);
1648 	if (tbl->pfrt_name[PF_TABLE_NAME_SIZE-1])
1649 		return (-1);
1650 	for (i = strlen(tbl->pfrt_name); i < PF_TABLE_NAME_SIZE; i++)
1651 		if (tbl->pfrt_name[i])
1652 			return (-1);
1653 	if (pfr_fix_anchor(tbl->pfrt_anchor))
1654 		return (-1);
1655 	if (tbl->pfrt_flags & ~allowedflags)
1656 		return (-1);
1657 	return (0);
1658 }
1659 
1660 /*
1661  * Rewrite anchors referenced by tables to remove slashes
1662  * and check for validity.
1663  */
1664 static int
1665 pfr_fix_anchor(char *anchor)
1666 {
1667 	size_t siz = MAXPATHLEN;
1668 	int i;
1669 
1670 	if (anchor[0] == '/') {
1671 		char *path;
1672 		int off;
1673 
1674 		path = anchor;
1675 		off = 1;
1676 		while (*++path == '/')
1677 			off++;
1678 		bcopy(path, anchor, siz - off);
1679 		memset(anchor + siz - off, 0, off);
1680 	}
1681 	if (anchor[siz - 1])
1682 		return (-1);
1683 	for (i = strlen(anchor); i < siz; i++)
1684 		if (anchor[i])
1685 			return (-1);
1686 	return (0);
1687 }
1688 
1689 static int
1690 pfr_table_count(struct pfr_table *filter, int flags)
1691 {
1692 	struct pf_ruleset *rs;
1693 
1694 	PF_RULES_ASSERT();
1695 
1696 	if (flags & PFR_FLAG_ALLRSETS)
1697 		return (V_pfr_ktable_cnt);
1698 	if (filter->pfrt_anchor[0]) {
1699 		rs = pf_find_ruleset(filter->pfrt_anchor);
1700 		return ((rs != NULL) ? rs->tables : -1);
1701 	}
1702 	return (pf_main_ruleset.tables);
1703 }
1704 
1705 static int
1706 pfr_skip_table(struct pfr_table *filter, struct pfr_ktable *kt, int flags)
1707 {
1708 	if (flags & PFR_FLAG_ALLRSETS)
1709 		return (0);
1710 	if (strcmp(filter->pfrt_anchor, kt->pfrkt_anchor))
1711 		return (1);
1712 	return (0);
1713 }
1714 
1715 static void
1716 pfr_insert_ktables(struct pfr_ktableworkq *workq)
1717 {
1718 	struct pfr_ktable	*p;
1719 
1720 	SLIST_FOREACH(p, workq, pfrkt_workq)
1721 		pfr_insert_ktable(p);
1722 }
1723 
1724 static void
1725 pfr_insert_ktable(struct pfr_ktable *kt)
1726 {
1727 
1728 	PF_RULES_WASSERT();
1729 
1730 	RB_INSERT(pfr_ktablehead, &V_pfr_ktables, kt);
1731 	V_pfr_ktable_cnt++;
1732 	if (kt->pfrkt_root != NULL)
1733 		if (!kt->pfrkt_root->pfrkt_refcnt[PFR_REFCNT_ANCHOR]++)
1734 			pfr_setflags_ktable(kt->pfrkt_root,
1735 			    kt->pfrkt_root->pfrkt_flags|PFR_TFLAG_REFDANCHOR);
1736 }
1737 
1738 static void
1739 pfr_setflags_ktables(struct pfr_ktableworkq *workq)
1740 {
1741 	struct pfr_ktable	*p, *q;
1742 
1743 	for (p = SLIST_FIRST(workq); p; p = q) {
1744 		q = SLIST_NEXT(p, pfrkt_workq);
1745 		pfr_setflags_ktable(p, p->pfrkt_nflags);
1746 	}
1747 }
1748 
1749 static void
1750 pfr_setflags_ktable(struct pfr_ktable *kt, int newf)
1751 {
1752 	struct pfr_kentryworkq	addrq;
1753 
1754 	PF_RULES_WASSERT();
1755 
1756 	if (!(newf & PFR_TFLAG_REFERENCED) &&
1757 	    !(newf & PFR_TFLAG_PERSIST))
1758 		newf &= ~PFR_TFLAG_ACTIVE;
1759 	if (!(newf & PFR_TFLAG_ACTIVE))
1760 		newf &= ~PFR_TFLAG_USRMASK;
1761 	if (!(newf & PFR_TFLAG_SETMASK)) {
1762 		RB_REMOVE(pfr_ktablehead, &V_pfr_ktables, kt);
1763 		if (kt->pfrkt_root != NULL)
1764 			if (!--kt->pfrkt_root->pfrkt_refcnt[PFR_REFCNT_ANCHOR])
1765 				pfr_setflags_ktable(kt->pfrkt_root,
1766 				    kt->pfrkt_root->pfrkt_flags &
1767 					~PFR_TFLAG_REFDANCHOR);
1768 		pfr_destroy_ktable(kt, 1);
1769 		V_pfr_ktable_cnt--;
1770 		return;
1771 	}
1772 	if (!(newf & PFR_TFLAG_ACTIVE) && kt->pfrkt_cnt) {
1773 		pfr_enqueue_addrs(kt, &addrq, NULL, 0);
1774 		pfr_remove_kentries(kt, &addrq);
1775 	}
1776 	if (!(newf & PFR_TFLAG_INACTIVE) && kt->pfrkt_shadow != NULL) {
1777 		pfr_destroy_ktable(kt->pfrkt_shadow, 1);
1778 		kt->pfrkt_shadow = NULL;
1779 	}
1780 	kt->pfrkt_flags = newf;
1781 }
1782 
1783 static void
1784 pfr_clstats_ktables(struct pfr_ktableworkq *workq, long tzero, int recurse)
1785 {
1786 	struct pfr_ktable	*p;
1787 
1788 	SLIST_FOREACH(p, workq, pfrkt_workq)
1789 		pfr_clstats_ktable(p, tzero, recurse);
1790 }
1791 
1792 static void
1793 pfr_clstats_ktable(struct pfr_ktable *kt, long tzero, int recurse)
1794 {
1795 	struct pfr_kentryworkq	 addrq;
1796 
1797 	if (recurse) {
1798 		pfr_enqueue_addrs(kt, &addrq, NULL, 0);
1799 		pfr_clstats_kentries(&addrq, tzero, 0);
1800 	}
1801 	bzero(kt->pfrkt_packets, sizeof(kt->pfrkt_packets));
1802 	bzero(kt->pfrkt_bytes, sizeof(kt->pfrkt_bytes));
1803 	kt->pfrkt_match = kt->pfrkt_nomatch = 0;
1804 	kt->pfrkt_tzero = tzero;
1805 }
1806 
1807 static struct pfr_ktable *
1808 pfr_create_ktable(struct pfr_table *tbl, long tzero, int attachruleset)
1809 {
1810 	struct pfr_ktable	*kt;
1811 	struct pf_ruleset	*rs;
1812 
1813 	PF_RULES_WASSERT();
1814 
1815 	kt = malloc(sizeof(*kt), M_PFTABLE, M_NOWAIT|M_ZERO);
1816 	if (kt == NULL)
1817 		return (NULL);
1818 	kt->pfrkt_t = *tbl;
1819 
1820 	if (attachruleset) {
1821 		rs = pf_find_or_create_ruleset(tbl->pfrt_anchor);
1822 		if (!rs) {
1823 			pfr_destroy_ktable(kt, 0);
1824 			return (NULL);
1825 		}
1826 		kt->pfrkt_rs = rs;
1827 		rs->tables++;
1828 	}
1829 
1830 	if (!rn_inithead((void **)&kt->pfrkt_ip4,
1831 	    offsetof(struct sockaddr_in, sin_addr) * 8) ||
1832 	    !rn_inithead((void **)&kt->pfrkt_ip6,
1833 	    offsetof(struct sockaddr_in6, sin6_addr) * 8)) {
1834 		pfr_destroy_ktable(kt, 0);
1835 		return (NULL);
1836 	}
1837 	kt->pfrkt_tzero = tzero;
1838 
1839 	return (kt);
1840 }
1841 
1842 static void
1843 pfr_destroy_ktables(struct pfr_ktableworkq *workq, int flushaddr)
1844 {
1845 	struct pfr_ktable	*p, *q;
1846 
1847 	for (p = SLIST_FIRST(workq); p; p = q) {
1848 		q = SLIST_NEXT(p, pfrkt_workq);
1849 		pfr_destroy_ktable(p, flushaddr);
1850 	}
1851 }
1852 
1853 static void
1854 pfr_destroy_ktable(struct pfr_ktable *kt, int flushaddr)
1855 {
1856 	struct pfr_kentryworkq	 addrq;
1857 
1858 	if (flushaddr) {
1859 		pfr_enqueue_addrs(kt, &addrq, NULL, 0);
1860 		pfr_clean_node_mask(kt, &addrq);
1861 		pfr_destroy_kentries(&addrq);
1862 	}
1863 	if (kt->pfrkt_ip4 != NULL)
1864 		rn_detachhead((void **)&kt->pfrkt_ip4);
1865 	if (kt->pfrkt_ip6 != NULL)
1866 		rn_detachhead((void **)&kt->pfrkt_ip6);
1867 	if (kt->pfrkt_shadow != NULL)
1868 		pfr_destroy_ktable(kt->pfrkt_shadow, flushaddr);
1869 	if (kt->pfrkt_rs != NULL) {
1870 		kt->pfrkt_rs->tables--;
1871 		pf_remove_if_empty_ruleset(kt->pfrkt_rs);
1872 	}
1873 	free(kt, M_PFTABLE);
1874 }
1875 
1876 static int
1877 pfr_ktable_compare(struct pfr_ktable *p, struct pfr_ktable *q)
1878 {
1879 	int d;
1880 
1881 	if ((d = strncmp(p->pfrkt_name, q->pfrkt_name, PF_TABLE_NAME_SIZE)))
1882 		return (d);
1883 	return (strcmp(p->pfrkt_anchor, q->pfrkt_anchor));
1884 }
1885 
1886 static struct pfr_ktable *
1887 pfr_lookup_table(struct pfr_table *tbl)
1888 {
1889 	/* struct pfr_ktable start like a struct pfr_table */
1890 	return (RB_FIND(pfr_ktablehead, &V_pfr_ktables,
1891 	    (struct pfr_ktable *)tbl));
1892 }
1893 
1894 int
1895 pfr_match_addr(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af)
1896 {
1897 	struct pfr_kentry	*ke = NULL;
1898 	int			 match;
1899 
1900 	PF_RULES_RASSERT();
1901 
1902 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1903 		kt = kt->pfrkt_root;
1904 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
1905 		return (0);
1906 
1907 	switch (af) {
1908 #ifdef INET
1909 	case AF_INET:
1910 	    {
1911 		struct sockaddr_in sin;
1912 
1913 		bzero(&sin, sizeof(sin));
1914 		sin.sin_len = sizeof(sin);
1915 		sin.sin_family = AF_INET;
1916 		sin.sin_addr.s_addr = a->addr32[0];
1917 		ke = (struct pfr_kentry *)rn_match(&sin, &kt->pfrkt_ip4->rh);
1918 		if (ke && KENTRY_RNF_ROOT(ke))
1919 			ke = NULL;
1920 		break;
1921 	    }
1922 #endif /* INET */
1923 #ifdef INET6
1924 	case AF_INET6:
1925 	    {
1926 		struct sockaddr_in6 sin6;
1927 
1928 		bzero(&sin6, sizeof(sin6));
1929 		sin6.sin6_len = sizeof(sin6);
1930 		sin6.sin6_family = AF_INET6;
1931 		bcopy(a, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1932 		ke = (struct pfr_kentry *)rn_match(&sin6, &kt->pfrkt_ip6->rh);
1933 		if (ke && KENTRY_RNF_ROOT(ke))
1934 			ke = NULL;
1935 		break;
1936 	    }
1937 #endif /* INET6 */
1938 	}
1939 	match = (ke && !ke->pfrke_not);
1940 	if (match)
1941 		kt->pfrkt_match++;
1942 	else
1943 		kt->pfrkt_nomatch++;
1944 	return (match);
1945 }
1946 
1947 void
1948 pfr_update_stats(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af,
1949     u_int64_t len, int dir_out, int op_pass, int notrule)
1950 {
1951 	struct pfr_kentry	*ke = NULL;
1952 
1953 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
1954 		kt = kt->pfrkt_root;
1955 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
1956 		return;
1957 
1958 	switch (af) {
1959 #ifdef INET
1960 	case AF_INET:
1961 	    {
1962 		struct sockaddr_in sin;
1963 
1964 		bzero(&sin, sizeof(sin));
1965 		sin.sin_len = sizeof(sin);
1966 		sin.sin_family = AF_INET;
1967 		sin.sin_addr.s_addr = a->addr32[0];
1968 		ke = (struct pfr_kentry *)rn_match(&sin, &kt->pfrkt_ip4->rh);
1969 		if (ke && KENTRY_RNF_ROOT(ke))
1970 			ke = NULL;
1971 		break;
1972 	    }
1973 #endif /* INET */
1974 #ifdef INET6
1975 	case AF_INET6:
1976 	    {
1977 		struct sockaddr_in6 sin6;
1978 
1979 		bzero(&sin6, sizeof(sin6));
1980 		sin6.sin6_len = sizeof(sin6);
1981 		sin6.sin6_family = AF_INET6;
1982 		bcopy(a, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1983 		ke = (struct pfr_kentry *)rn_match(&sin6, &kt->pfrkt_ip6->rh);
1984 		if (ke && KENTRY_RNF_ROOT(ke))
1985 			ke = NULL;
1986 		break;
1987 	    }
1988 #endif /* INET6 */
1989 	default:
1990 		panic("%s: unknown address family %u", __func__, af);
1991 	}
1992 	if ((ke == NULL || ke->pfrke_not) != notrule) {
1993 		if (op_pass != PFR_OP_PASS)
1994 			printf("pfr_update_stats: assertion failed.\n");
1995 		op_pass = PFR_OP_XPASS;
1996 	}
1997 	kt->pfrkt_packets[dir_out][op_pass]++;
1998 	kt->pfrkt_bytes[dir_out][op_pass] += len;
1999 	if (ke != NULL && op_pass != PFR_OP_XPASS &&
2000 	    (kt->pfrkt_flags & PFR_TFLAG_COUNTERS)) {
2001 		if (ke->pfrke_counters == NULL)
2002 			ke->pfrke_counters = uma_zalloc(V_pfr_kcounters_z,
2003 			    M_NOWAIT | M_ZERO);
2004 		if (ke->pfrke_counters != NULL) {
2005 			ke->pfrke_counters->pfrkc_packets[dir_out][op_pass]++;
2006 			ke->pfrke_counters->pfrkc_bytes[dir_out][op_pass] += len;
2007 		}
2008 	}
2009 }
2010 
2011 struct pfr_ktable *
2012 pfr_attach_table(struct pf_ruleset *rs, char *name)
2013 {
2014 	struct pfr_ktable	*kt, *rt;
2015 	struct pfr_table	 tbl;
2016 	struct pf_anchor	*ac = rs->anchor;
2017 
2018 	PF_RULES_WASSERT();
2019 
2020 	bzero(&tbl, sizeof(tbl));
2021 	strlcpy(tbl.pfrt_name, name, sizeof(tbl.pfrt_name));
2022 	if (ac != NULL)
2023 		strlcpy(tbl.pfrt_anchor, ac->path, sizeof(tbl.pfrt_anchor));
2024 	kt = pfr_lookup_table(&tbl);
2025 	if (kt == NULL) {
2026 		kt = pfr_create_ktable(&tbl, time_second, 1);
2027 		if (kt == NULL)
2028 			return (NULL);
2029 		if (ac != NULL) {
2030 			bzero(tbl.pfrt_anchor, sizeof(tbl.pfrt_anchor));
2031 			rt = pfr_lookup_table(&tbl);
2032 			if (rt == NULL) {
2033 				rt = pfr_create_ktable(&tbl, 0, 1);
2034 				if (rt == NULL) {
2035 					pfr_destroy_ktable(kt, 0);
2036 					return (NULL);
2037 				}
2038 				pfr_insert_ktable(rt);
2039 			}
2040 			kt->pfrkt_root = rt;
2041 		}
2042 		pfr_insert_ktable(kt);
2043 	}
2044 	if (!kt->pfrkt_refcnt[PFR_REFCNT_RULE]++)
2045 		pfr_setflags_ktable(kt, kt->pfrkt_flags|PFR_TFLAG_REFERENCED);
2046 	return (kt);
2047 }
2048 
2049 void
2050 pfr_detach_table(struct pfr_ktable *kt)
2051 {
2052 
2053 	PF_RULES_WASSERT();
2054 	KASSERT(kt->pfrkt_refcnt[PFR_REFCNT_RULE] > 0, ("%s: refcount %d\n",
2055 	    __func__, kt->pfrkt_refcnt[PFR_REFCNT_RULE]));
2056 
2057 	if (!--kt->pfrkt_refcnt[PFR_REFCNT_RULE])
2058 		pfr_setflags_ktable(kt, kt->pfrkt_flags&~PFR_TFLAG_REFERENCED);
2059 }
2060 
2061 int
2062 pfr_pool_get(struct pfr_ktable *kt, int *pidx, struct pf_addr *counter,
2063     sa_family_t af)
2064 {
2065 	struct pf_addr		 *addr, *cur, *mask;
2066 	union sockaddr_union	 uaddr, umask;
2067 	struct pfr_kentry	*ke, *ke2 = NULL;
2068 	int			 idx = -1, use_counter = 0;
2069 
2070 	switch (af) {
2071 	case AF_INET:
2072 		uaddr.sin.sin_len = sizeof(struct sockaddr_in);
2073 		uaddr.sin.sin_family = AF_INET;
2074 		break;
2075 	case AF_INET6:
2076 		uaddr.sin6.sin6_len = sizeof(struct sockaddr_in6);
2077 		uaddr.sin6.sin6_family = AF_INET6;
2078 		break;
2079 	}
2080 	addr = SUNION2PF(&uaddr, af);
2081 
2082 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE) && kt->pfrkt_root != NULL)
2083 		kt = kt->pfrkt_root;
2084 	if (!(kt->pfrkt_flags & PFR_TFLAG_ACTIVE))
2085 		return (-1);
2086 
2087 	if (pidx != NULL)
2088 		idx = *pidx;
2089 	if (counter != NULL && idx >= 0)
2090 		use_counter = 1;
2091 	if (idx < 0)
2092 		idx = 0;
2093 
2094 _next_block:
2095 	ke = pfr_kentry_byidx(kt, idx, af);
2096 	if (ke == NULL) {
2097 		kt->pfrkt_nomatch++;
2098 		return (1);
2099 	}
2100 	pfr_prepare_network(&umask, af, ke->pfrke_net);
2101 	cur = SUNION2PF(&ke->pfrke_sa, af);
2102 	mask = SUNION2PF(&umask, af);
2103 
2104 	if (use_counter) {
2105 		/* is supplied address within block? */
2106 		if (!PF_MATCHA(0, cur, mask, counter, af)) {
2107 			/* no, go to next block in table */
2108 			idx++;
2109 			use_counter = 0;
2110 			goto _next_block;
2111 		}
2112 		PF_ACPY(addr, counter, af);
2113 	} else {
2114 		/* use first address of block */
2115 		PF_ACPY(addr, cur, af);
2116 	}
2117 
2118 	if (!KENTRY_NETWORK(ke)) {
2119 		/* this is a single IP address - no possible nested block */
2120 		PF_ACPY(counter, addr, af);
2121 		*pidx = idx;
2122 		kt->pfrkt_match++;
2123 		return (0);
2124 	}
2125 	for (;;) {
2126 		/* we don't want to use a nested block */
2127 		switch (af) {
2128 		case AF_INET:
2129 			ke2 = (struct pfr_kentry *)rn_match(&uaddr,
2130 			    &kt->pfrkt_ip4->rh);
2131 			break;
2132 		case AF_INET6:
2133 			ke2 = (struct pfr_kentry *)rn_match(&uaddr,
2134 			    &kt->pfrkt_ip6->rh);
2135 			break;
2136 		}
2137 		/* no need to check KENTRY_RNF_ROOT() here */
2138 		if (ke2 == ke) {
2139 			/* lookup return the same block - perfect */
2140 			PF_ACPY(counter, addr, af);
2141 			*pidx = idx;
2142 			kt->pfrkt_match++;
2143 			return (0);
2144 		}
2145 
2146 		/* we need to increase the counter past the nested block */
2147 		pfr_prepare_network(&umask, AF_INET, ke2->pfrke_net);
2148 		PF_POOLMASK(addr, addr, SUNION2PF(&umask, af), &pfr_ffaddr, af);
2149 		PF_AINC(addr, af);
2150 		if (!PF_MATCHA(0, cur, mask, addr, af)) {
2151 			/* ok, we reached the end of our main block */
2152 			/* go to next block in table */
2153 			idx++;
2154 			use_counter = 0;
2155 			goto _next_block;
2156 		}
2157 	}
2158 }
2159 
2160 static struct pfr_kentry *
2161 pfr_kentry_byidx(struct pfr_ktable *kt, int idx, int af)
2162 {
2163 	struct pfr_walktree	w;
2164 
2165 	bzero(&w, sizeof(w));
2166 	w.pfrw_op = PFRW_POOL_GET;
2167 	w.pfrw_cnt = idx;
2168 
2169 	switch (af) {
2170 #ifdef INET
2171 	case AF_INET:
2172 		kt->pfrkt_ip4->rnh_walktree(&kt->pfrkt_ip4->rh, pfr_walktree, &w);
2173 		return (w.pfrw_kentry);
2174 #endif /* INET */
2175 #ifdef INET6
2176 	case AF_INET6:
2177 		kt->pfrkt_ip6->rnh_walktree(&kt->pfrkt_ip6->rh, pfr_walktree, &w);
2178 		return (w.pfrw_kentry);
2179 #endif /* INET6 */
2180 	default:
2181 		return (NULL);
2182 	}
2183 }
2184 
2185 void
2186 pfr_dynaddr_update(struct pfr_ktable *kt, struct pfi_dynaddr *dyn)
2187 {
2188 	struct pfr_walktree	w;
2189 
2190 	bzero(&w, sizeof(w));
2191 	w.pfrw_op = PFRW_DYNADDR_UPDATE;
2192 	w.pfrw_dyn = dyn;
2193 
2194 	dyn->pfid_acnt4 = 0;
2195 	dyn->pfid_acnt6 = 0;
2196 	if (!dyn->pfid_af || dyn->pfid_af == AF_INET)
2197 		kt->pfrkt_ip4->rnh_walktree(&kt->pfrkt_ip4->rh, pfr_walktree, &w);
2198 	if (!dyn->pfid_af || dyn->pfid_af == AF_INET6)
2199 		kt->pfrkt_ip6->rnh_walktree(&kt->pfrkt_ip6->rh, pfr_walktree, &w);
2200 }
2201