xref: /freebsd/sys/netpfil/ipfw/ip_fw_sockopt.c (revision a812392203d7c4c3f0db9d8a0f3391374c49c71f)
1 /*-
2  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
3  * Copyright (c) 2014 Yandex LLC
4  * Copyright (c) 2014 Alexander V. Chernikov
5  *
6  * Supported by: Valeria Paoli
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  */
29 
30 #include <sys/cdefs.h>
31 __FBSDID("$FreeBSD$");
32 
33 /*
34  * Control socket and rule management routines for ipfw.
35  * Control is currently implemented via IP_FW3 setsockopt() code.
36  */
37 
38 #include "opt_ipfw.h"
39 #include "opt_inet.h"
40 #ifndef INET
41 #error IPFIREWALL requires INET.
42 #endif /* INET */
43 #include "opt_inet6.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>	/* struct m_tag used by nested headers */
49 #include <sys/kernel.h>
50 #include <sys/lock.h>
51 #include <sys/priv.h>
52 #include <sys/proc.h>
53 #include <sys/rwlock.h>
54 #include <sys/rmlock.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/syslog.h>
59 #include <sys/fnv_hash.h>
60 #include <net/if.h>
61 #include <net/route.h>
62 #include <net/vnet.h>
63 #include <vm/vm.h>
64 #include <vm/vm_extern.h>
65 
66 #include <netinet/in.h>
67 #include <netinet/ip_var.h> /* hooks */
68 #include <netinet/ip_fw.h>
69 
70 #include <netpfil/ipfw/ip_fw_private.h>
71 #include <netpfil/ipfw/ip_fw_table.h>
72 
73 #ifdef MAC
74 #include <security/mac/mac_framework.h>
75 #endif
76 
77 static int ipfw_ctl(struct sockopt *sopt);
78 static int check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len,
79     struct rule_check_info *ci);
80 static int check_ipfw_rule1(struct ip_fw_rule *rule, int size,
81     struct rule_check_info *ci);
82 static int check_ipfw_rule0(struct ip_fw_rule0 *rule, int size,
83     struct rule_check_info *ci);
84 
85 #define	NAMEDOBJ_HASH_SIZE	32
86 
87 struct namedobj_instance {
88 	struct namedobjects_head	*names;
89 	struct namedobjects_head	*values;
90 	uint32_t nn_size;		/* names hash size */
91 	uint32_t nv_size;		/* number hash size */
92 	u_long *idx_mask;		/* used items bitmask */
93 	uint32_t max_blocks;		/* number of "long" blocks in bitmask */
94 	uint32_t count;			/* number of items */
95 	uint16_t free_off[IPFW_MAX_SETS];	/* first possible free offset */
96 	objhash_hash_f	*hash_f;
97 	objhash_cmp_f	*cmp_f;
98 };
99 #define	BLOCK_ITEMS	(8 * sizeof(u_long))	/* Number of items for ffsl() */
100 
101 static uint32_t objhash_hash_name(struct namedobj_instance *ni, void *key,
102     uint32_t kopt);
103 static uint32_t objhash_hash_idx(struct namedobj_instance *ni, uint32_t val);
104 static int objhash_cmp_name(struct named_object *no, void *name, uint32_t set);
105 
106 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
107 
108 static int dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
109     struct sockopt_data *sd);
110 static int add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
111     struct sockopt_data *sd);
112 static int del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
113     struct sockopt_data *sd);
114 static int clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
115     struct sockopt_data *sd);
116 static int move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
117     struct sockopt_data *sd);
118 static int manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
119     struct sockopt_data *sd);
120 static int dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
121     struct sockopt_data *sd);
122 
123 /* ctl3 handler data */
124 struct mtx ctl3_lock;
125 #define	CTL3_LOCK_INIT()	mtx_init(&ctl3_lock, "ctl3_lock", NULL, MTX_DEF)
126 #define	CTL3_LOCK_DESTROY()	mtx_destroy(&ctl3_lock)
127 #define	CTL3_LOCK()		mtx_lock(&ctl3_lock)
128 #define	CTL3_UNLOCK()		mtx_unlock(&ctl3_lock)
129 
130 static struct ipfw_sopt_handler *ctl3_handlers;
131 static size_t ctl3_hsize;
132 static uint64_t ctl3_refct, ctl3_gencnt;
133 #define	CTL3_SMALLBUF	4096			/* small page-size write buffer */
134 #define	CTL3_LARGEBUF	16 * 1024 * 1024	/* handle large rulesets */
135 
136 static int ipfw_flush_sopt_data(struct sockopt_data *sd);
137 
138 static struct ipfw_sopt_handler	scodes[] = {
139 	{ IP_FW_XGET,		0,	HDIR_GET,	dump_config },
140 	{ IP_FW_XADD,		0,	HDIR_BOTH,	add_rules },
141 	{ IP_FW_XDEL,		0,	HDIR_BOTH,	del_rules },
142 	{ IP_FW_XZERO,		0,	HDIR_SET,	clear_rules },
143 	{ IP_FW_XRESETLOG,	0,	HDIR_SET,	clear_rules },
144 	{ IP_FW_XMOVE,		0,	HDIR_SET,	move_rules },
145 	{ IP_FW_SET_SWAP,	0,	HDIR_SET,	manage_sets },
146 	{ IP_FW_SET_MOVE,	0,	HDIR_SET,	manage_sets },
147 	{ IP_FW_SET_ENABLE,	0,	HDIR_SET,	manage_sets },
148 	{ IP_FW_DUMP_SOPTCODES,	0,	HDIR_GET,	dump_soptcodes },
149 };
150 
151 /*
152  * static variables followed by global ones
153  */
154 
155 static VNET_DEFINE(uma_zone_t, ipfw_cntr_zone);
156 #define	V_ipfw_cntr_zone		VNET(ipfw_cntr_zone)
157 
158 void
159 ipfw_init_counters()
160 {
161 
162 	V_ipfw_cntr_zone = uma_zcreate("IPFW counters",
163 	    IPFW_RULE_CNTR_SIZE, NULL, NULL, NULL, NULL,
164 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
165 }
166 
167 void
168 ipfw_destroy_counters()
169 {
170 
171 	uma_zdestroy(V_ipfw_cntr_zone);
172 }
173 
174 struct ip_fw *
175 ipfw_alloc_rule(struct ip_fw_chain *chain, size_t rulesize)
176 {
177 	struct ip_fw *rule;
178 
179 	rule = malloc(rulesize, M_IPFW, M_WAITOK | M_ZERO);
180 	rule->cntr = uma_zalloc(V_ipfw_cntr_zone, M_WAITOK | M_ZERO);
181 
182 	return (rule);
183 }
184 
185 static void
186 free_rule(struct ip_fw *rule)
187 {
188 
189 	uma_zfree(V_ipfw_cntr_zone, rule->cntr);
190 	free(rule, M_IPFW);
191 }
192 
193 
194 /*
195  * Find the smallest rule >= key, id.
196  * We could use bsearch but it is so simple that we code it directly
197  */
198 int
199 ipfw_find_rule(struct ip_fw_chain *chain, uint32_t key, uint32_t id)
200 {
201 	int i, lo, hi;
202 	struct ip_fw *r;
203 
204   	for (lo = 0, hi = chain->n_rules - 1; lo < hi;) {
205 		i = (lo + hi) / 2;
206 		r = chain->map[i];
207 		if (r->rulenum < key)
208 			lo = i + 1;	/* continue from the next one */
209 		else if (r->rulenum > key)
210 			hi = i;		/* this might be good */
211 		else if (r->id < id)
212 			lo = i + 1;	/* continue from the next one */
213 		else /* r->id >= id */
214 			hi = i;		/* this might be good */
215 	};
216 	return hi;
217 }
218 
219 /*
220  * Builds skipto cache on rule set @map.
221  */
222 static void
223 update_skipto_cache(struct ip_fw_chain *chain, struct ip_fw **map)
224 {
225 	int *smap, rulenum;
226 	int i, mi;
227 
228 	IPFW_UH_WLOCK_ASSERT(chain);
229 
230 	mi = 0;
231 	rulenum = map[mi]->rulenum;
232 	smap = chain->idxmap_back;
233 
234 	if (smap == NULL)
235 		return;
236 
237 	for (i = 0; i < 65536; i++) {
238 		smap[i] = mi;
239 		/* Use the same rule index until i < rulenum */
240 		if (i != rulenum || i == 65535)
241 			continue;
242 		/* Find next rule with num > i */
243 		rulenum = map[++mi]->rulenum;
244 		while (rulenum == i)
245 			rulenum = map[++mi]->rulenum;
246 	}
247 }
248 
249 /*
250  * Swaps prepared (backup) index with current one.
251  */
252 static void
253 swap_skipto_cache(struct ip_fw_chain *chain)
254 {
255 	int *map;
256 
257 	IPFW_UH_WLOCK_ASSERT(chain);
258 	IPFW_WLOCK_ASSERT(chain);
259 
260 	map = chain->idxmap;
261 	chain->idxmap = chain->idxmap_back;
262 	chain->idxmap_back = map;
263 }
264 
265 /*
266  * Allocate and initialize skipto cache.
267  */
268 void
269 ipfw_init_skipto_cache(struct ip_fw_chain *chain)
270 {
271 	int *idxmap, *idxmap_back;
272 
273 	idxmap = malloc(65536 * sizeof(uint32_t *), M_IPFW,
274 	    M_WAITOK | M_ZERO);
275 	idxmap_back = malloc(65536 * sizeof(uint32_t *), M_IPFW,
276 	    M_WAITOK | M_ZERO);
277 
278 	/*
279 	 * Note we may be called at any time after initialization,
280 	 * for example, on first skipto rule, so we need to
281 	 * provide valid chain->idxmap on return
282 	 */
283 
284 	IPFW_UH_WLOCK(chain);
285 	if (chain->idxmap != NULL) {
286 		IPFW_UH_WUNLOCK(chain);
287 		free(idxmap, M_IPFW);
288 		free(idxmap_back, M_IPFW);
289 		return;
290 	}
291 
292 	/* Set backup pointer first to permit building cache */
293 	chain->idxmap_back = idxmap_back;
294 	update_skipto_cache(chain, chain->map);
295 	IPFW_WLOCK(chain);
296 	/* It is now safe to set chain->idxmap ptr */
297 	chain->idxmap = idxmap;
298 	swap_skipto_cache(chain);
299 	IPFW_WUNLOCK(chain);
300 	IPFW_UH_WUNLOCK(chain);
301 }
302 
303 /*
304  * Destroys skipto cache.
305  */
306 void
307 ipfw_destroy_skipto_cache(struct ip_fw_chain *chain)
308 {
309 
310 	if (chain->idxmap != NULL)
311 		free(chain->idxmap, M_IPFW);
312 	if (chain->idxmap != NULL)
313 		free(chain->idxmap_back, M_IPFW);
314 }
315 
316 
317 /*
318  * allocate a new map, returns the chain locked. extra is the number
319  * of entries to add or delete.
320  */
321 static struct ip_fw **
322 get_map(struct ip_fw_chain *chain, int extra, int locked)
323 {
324 
325 	for (;;) {
326 		struct ip_fw **map;
327 		int i, mflags;
328 
329 		mflags = M_ZERO | ((locked != 0) ? M_NOWAIT : M_WAITOK);
330 
331 		i = chain->n_rules + extra;
332 		map = malloc(i * sizeof(struct ip_fw *), M_IPFW, mflags);
333 		if (map == NULL) {
334 			printf("%s: cannot allocate map\n", __FUNCTION__);
335 			return NULL;
336 		}
337 		if (!locked)
338 			IPFW_UH_WLOCK(chain);
339 		if (i >= chain->n_rules + extra) /* good */
340 			return map;
341 		/* otherwise we lost the race, free and retry */
342 		if (!locked)
343 			IPFW_UH_WUNLOCK(chain);
344 		free(map, M_IPFW);
345 	}
346 }
347 
348 /*
349  * swap the maps. It is supposed to be called with IPFW_UH_WLOCK
350  */
351 static struct ip_fw **
352 swap_map(struct ip_fw_chain *chain, struct ip_fw **new_map, int new_len)
353 {
354 	struct ip_fw **old_map;
355 
356 	IPFW_WLOCK(chain);
357 	chain->id++;
358 	chain->n_rules = new_len;
359 	old_map = chain->map;
360 	chain->map = new_map;
361 	swap_skipto_cache(chain);
362 	IPFW_WUNLOCK(chain);
363 	return old_map;
364 }
365 
366 
367 static void
368 export_cntr1_base(struct ip_fw *krule, struct ip_fw_bcounter *cntr)
369 {
370 
371 	cntr->size = sizeof(*cntr);
372 
373 	if (krule->cntr != NULL) {
374 		cntr->pcnt = counter_u64_fetch(krule->cntr);
375 		cntr->bcnt = counter_u64_fetch(krule->cntr + 1);
376 		cntr->timestamp = krule->timestamp;
377 	}
378 	if (cntr->timestamp > 0)
379 		cntr->timestamp += boottime.tv_sec;
380 }
381 
382 static void
383 export_cntr0_base(struct ip_fw *krule, struct ip_fw_bcounter0 *cntr)
384 {
385 
386 	if (krule->cntr != NULL) {
387 		cntr->pcnt = counter_u64_fetch(krule->cntr);
388 		cntr->bcnt = counter_u64_fetch(krule->cntr + 1);
389 		cntr->timestamp = krule->timestamp;
390 	}
391 	if (cntr->timestamp > 0)
392 		cntr->timestamp += boottime.tv_sec;
393 }
394 
395 /*
396  * Copies rule @urule from v1 userland format (current).
397  * to kernel @krule.
398  * Assume @krule is zeroed.
399  */
400 static void
401 import_rule1(struct rule_check_info *ci)
402 {
403 	struct ip_fw_rule *urule;
404 	struct ip_fw *krule;
405 
406 	urule = (struct ip_fw_rule *)ci->urule;
407 	krule = (struct ip_fw *)ci->krule;
408 
409 	/* copy header */
410 	krule->act_ofs = urule->act_ofs;
411 	krule->cmd_len = urule->cmd_len;
412 	krule->rulenum = urule->rulenum;
413 	krule->set = urule->set;
414 	krule->flags = urule->flags;
415 
416 	/* Save rulenum offset */
417 	ci->urule_numoff = offsetof(struct ip_fw_rule, rulenum);
418 
419 	/* Copy opcodes */
420 	memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t));
421 }
422 
423 /*
424  * Export rule into v1 format (Current).
425  * Layout:
426  * [ ipfw_obj_tlv(IPFW_TLV_RULE_ENT)
427  *     [ ip_fw_rule ] OR
428  *     [ ip_fw_bcounter ip_fw_rule] (depends on rcntrs).
429  * ]
430  * Assume @data is zeroed.
431  */
432 static void
433 export_rule1(struct ip_fw *krule, caddr_t data, int len, int rcntrs)
434 {
435 	struct ip_fw_bcounter *cntr;
436 	struct ip_fw_rule *urule;
437 	ipfw_obj_tlv *tlv;
438 
439 	/* Fill in TLV header */
440 	tlv = (ipfw_obj_tlv *)data;
441 	tlv->type = IPFW_TLV_RULE_ENT;
442 	tlv->length = len;
443 
444 	if (rcntrs != 0) {
445 		/* Copy counters */
446 		cntr = (struct ip_fw_bcounter *)(tlv + 1);
447 		urule = (struct ip_fw_rule *)(cntr + 1);
448 		export_cntr1_base(krule, cntr);
449 	} else
450 		urule = (struct ip_fw_rule *)(tlv + 1);
451 
452 	/* copy header */
453 	urule->act_ofs = krule->act_ofs;
454 	urule->cmd_len = krule->cmd_len;
455 	urule->rulenum = krule->rulenum;
456 	urule->set = krule->set;
457 	urule->flags = krule->flags;
458 	urule->id = krule->id;
459 
460 	/* Copy opcodes */
461 	memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t));
462 }
463 
464 
465 /*
466  * Copies rule @urule from FreeBSD8 userland format (v0)
467  * to kernel @krule.
468  * Assume @krule is zeroed.
469  */
470 static void
471 import_rule0(struct rule_check_info *ci)
472 {
473 	struct ip_fw_rule0 *urule;
474 	struct ip_fw *krule;
475 	int cmdlen, l;
476 	ipfw_insn *cmd;
477 	ipfw_insn_limit *lcmd;
478 	ipfw_insn_if *cmdif;
479 
480 	urule = (struct ip_fw_rule0 *)ci->urule;
481 	krule = (struct ip_fw *)ci->krule;
482 
483 	/* copy header */
484 	krule->act_ofs = urule->act_ofs;
485 	krule->cmd_len = urule->cmd_len;
486 	krule->rulenum = urule->rulenum;
487 	krule->set = urule->set;
488 	if ((urule->_pad & 1) != 0)
489 		krule->flags |= IPFW_RULE_NOOPT;
490 
491 	/* Save rulenum offset */
492 	ci->urule_numoff = offsetof(struct ip_fw_rule0, rulenum);
493 
494 	/* Copy opcodes */
495 	memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t));
496 
497 	/*
498 	 * Alter opcodes:
499 	 * 1) convert tablearg value from 65335 to 0
500 	 * 2) Add high bit to O_SETFIB/O_SETDSCP values (to make room for targ).
501 	 * 3) convert table number in iface opcodes to u16
502 	 */
503 	l = krule->cmd_len;
504 	cmd = krule->cmd;
505 	cmdlen = 0;
506 
507 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
508 		cmdlen = F_LEN(cmd);
509 
510 		switch (cmd->opcode) {
511 		/* Opcodes supporting tablearg */
512 		case O_TAG:
513 		case O_TAGGED:
514 		case O_PIPE:
515 		case O_QUEUE:
516 		case O_DIVERT:
517 		case O_TEE:
518 		case O_SKIPTO:
519 		case O_CALLRETURN:
520 		case O_NETGRAPH:
521 		case O_NGTEE:
522 		case O_NAT:
523 			if (cmd->arg1 == 65535)
524 				cmd->arg1 = IP_FW_TARG;
525 			break;
526 		case O_SETFIB:
527 		case O_SETDSCP:
528 			if (cmd->arg1 == 65535)
529 				cmd->arg1 = IP_FW_TARG;
530 			else
531 				cmd->arg1 |= 0x8000;
532 			break;
533 		case O_LIMIT:
534 			lcmd = (ipfw_insn_limit *)cmd;
535 			if (lcmd->conn_limit == 65535)
536 				lcmd->conn_limit = IP_FW_TARG;
537 			break;
538 		/* Interface tables */
539 		case O_XMIT:
540 		case O_RECV:
541 		case O_VIA:
542 			/* Interface table, possibly */
543 			cmdif = (ipfw_insn_if *)cmd;
544 			if (cmdif->name[0] != '\1')
545 				break;
546 
547 			cmdif->p.kidx = (uint16_t)cmdif->p.glob;
548 			break;
549 		}
550 	}
551 }
552 
553 /*
554  * Copies rule @krule from kernel to FreeBSD8 userland format (v0)
555  */
556 static void
557 export_rule0(struct ip_fw *krule, struct ip_fw_rule0 *urule, int len)
558 {
559 	int cmdlen, l;
560 	ipfw_insn *cmd;
561 	ipfw_insn_limit *lcmd;
562 	ipfw_insn_if *cmdif;
563 
564 	/* copy header */
565 	memset(urule, 0, len);
566 	urule->act_ofs = krule->act_ofs;
567 	urule->cmd_len = krule->cmd_len;
568 	urule->rulenum = krule->rulenum;
569 	urule->set = krule->set;
570 	if ((krule->flags & IPFW_RULE_NOOPT) != 0)
571 		urule->_pad |= 1;
572 
573 	/* Copy opcodes */
574 	memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t));
575 
576 	/* Export counters */
577 	export_cntr0_base(krule, (struct ip_fw_bcounter0 *)&urule->pcnt);
578 
579 	/*
580 	 * Alter opcodes:
581 	 * 1) convert tablearg value from 0 to 65335
582 	 * 2) Remove highest bit from O_SETFIB/O_SETDSCP values.
583 	 * 3) convert table number in iface opcodes to int
584 	 */
585 	l = urule->cmd_len;
586 	cmd = urule->cmd;
587 	cmdlen = 0;
588 
589 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
590 		cmdlen = F_LEN(cmd);
591 
592 		switch (cmd->opcode) {
593 		/* Opcodes supporting tablearg */
594 		case O_TAG:
595 		case O_TAGGED:
596 		case O_PIPE:
597 		case O_QUEUE:
598 		case O_DIVERT:
599 		case O_TEE:
600 		case O_SKIPTO:
601 		case O_CALLRETURN:
602 		case O_NETGRAPH:
603 		case O_NGTEE:
604 		case O_NAT:
605 			if (cmd->arg1 == IP_FW_TARG)
606 				cmd->arg1 = 65535;
607 			break;
608 		case O_SETFIB:
609 		case O_SETDSCP:
610 			if (cmd->arg1 == IP_FW_TARG)
611 				cmd->arg1 = 65535;
612 			else
613 				cmd->arg1 &= ~0x8000;
614 			break;
615 		case O_LIMIT:
616 			lcmd = (ipfw_insn_limit *)cmd;
617 			if (lcmd->conn_limit == IP_FW_TARG)
618 				lcmd->conn_limit = 65535;
619 			break;
620 		/* Interface tables */
621 		case O_XMIT:
622 		case O_RECV:
623 		case O_VIA:
624 			/* Interface table, possibly */
625 			cmdif = (ipfw_insn_if *)cmd;
626 			if (cmdif->name[0] != '\1')
627 				break;
628 
629 			cmdif->p.glob = cmdif->p.kidx;
630 			break;
631 		}
632 	}
633 }
634 
635 /*
636  * Add new rule(s) to the list possibly creating rule number for each.
637  * Update the rule_number in the input struct so the caller knows it as well.
638  * Must be called without IPFW_UH held
639  */
640 static int
641 commit_rules(struct ip_fw_chain *chain, struct rule_check_info *rci, int count)
642 {
643 	int error, i, insert_before, tcount;
644 	uint16_t rulenum, *pnum;
645 	struct rule_check_info *ci;
646 	struct ip_fw *krule;
647 	struct ip_fw **map;	/* the new array of pointers */
648 
649 	/* Check if we need to do table remap */
650 	tcount = 0;
651 	for (ci = rci, i = 0; i < count; ci++, i++) {
652 		if (ci->table_opcodes == 0)
653 			continue;
654 
655 		/*
656 		 * Rule has some table opcodes.
657 		 * Reference & allocate needed tables/
658 		 */
659 		error = ipfw_rewrite_table_uidx(chain, ci);
660 		if (error != 0) {
661 
662 			/*
663 			 * rewrite failed, state for current rule
664 			 * has been reverted. Check if we need to
665 			 * revert more.
666 			 */
667 			if (tcount > 0) {
668 
669 				/*
670 				 * We have some more table rules
671 				 * we need to rollback.
672 				 */
673 
674 				IPFW_UH_WLOCK(chain);
675 				while (ci != rci) {
676 					ci--;
677 					if (ci->table_opcodes == 0)
678 						continue;
679 					ipfw_unref_rule_tables(chain,ci->krule);
680 
681 				}
682 				IPFW_UH_WUNLOCK(chain);
683 
684 			}
685 
686 			return (error);
687 		}
688 
689 		tcount++;
690 	}
691 
692 	/* get_map returns with IPFW_UH_WLOCK if successful */
693 	map = get_map(chain, count, 0 /* not locked */);
694 	if (map == NULL) {
695 		if (tcount > 0) {
696 			/* Unbind tables */
697 			IPFW_UH_WLOCK(chain);
698 			for (ci = rci, i = 0; i < count; ci++, i++) {
699 				if (ci->table_opcodes == 0)
700 					continue;
701 
702 				ipfw_unref_rule_tables(chain, ci->krule);
703 			}
704 			IPFW_UH_WUNLOCK(chain);
705 		}
706 
707 		return (ENOSPC);
708 	}
709 
710 	if (V_autoinc_step < 1)
711 		V_autoinc_step = 1;
712 	else if (V_autoinc_step > 1000)
713 		V_autoinc_step = 1000;
714 
715 	/* FIXME: Handle count > 1 */
716 	ci = rci;
717 	krule = ci->krule;
718 	rulenum = krule->rulenum;
719 
720 	/* find the insertion point, we will insert before */
721 	insert_before = rulenum ? rulenum + 1 : IPFW_DEFAULT_RULE;
722 	i = ipfw_find_rule(chain, insert_before, 0);
723 	/* duplicate first part */
724 	if (i > 0)
725 		bcopy(chain->map, map, i * sizeof(struct ip_fw *));
726 	map[i] = krule;
727 	/* duplicate remaining part, we always have the default rule */
728 	bcopy(chain->map + i, map + i + 1,
729 		sizeof(struct ip_fw *) *(chain->n_rules - i));
730 	if (rulenum == 0) {
731 		/* Compute rule number and write it back */
732 		rulenum = i > 0 ? map[i-1]->rulenum : 0;
733 		if (rulenum < IPFW_DEFAULT_RULE - V_autoinc_step)
734 			rulenum += V_autoinc_step;
735 		krule->rulenum = rulenum;
736 		/* Save number to userland rule */
737 		pnum = (uint16_t *)((caddr_t)ci->urule + ci->urule_numoff);
738 		*pnum = rulenum;
739 	}
740 
741 	krule->id = chain->id + 1;
742 	update_skipto_cache(chain, map);
743 	map = swap_map(chain, map, chain->n_rules + 1);
744 	chain->static_len += RULEUSIZE0(krule);
745 	IPFW_UH_WUNLOCK(chain);
746 	if (map)
747 		free(map, M_IPFW);
748 	return (0);
749 }
750 
751 /*
752  * Adds @rule to the list of rules to reap
753  */
754 void
755 ipfw_reap_add(struct ip_fw_chain *chain, struct ip_fw **head,
756     struct ip_fw *rule)
757 {
758 
759 	IPFW_UH_WLOCK_ASSERT(chain);
760 
761 	/* Unlink rule from everywhere */
762 	ipfw_unref_rule_tables(chain, rule);
763 
764 	*((struct ip_fw **)rule) = *head;
765 	*head = rule;
766 }
767 
768 /*
769  * Reclaim storage associated with a list of rules.  This is
770  * typically the list created using remove_rule.
771  * A NULL pointer on input is handled correctly.
772  */
773 void
774 ipfw_reap_rules(struct ip_fw *head)
775 {
776 	struct ip_fw *rule;
777 
778 	while ((rule = head) != NULL) {
779 		head = *((struct ip_fw **)head);
780 		free_rule(rule);
781 	}
782 }
783 
784 /*
785  * Rules to keep are
786  *	(default || reserved || !match_set || !match_number)
787  * where
788  *   default ::= (rule->rulenum == IPFW_DEFAULT_RULE)
789  *	// the default rule is always protected
790  *
791  *   reserved ::= (cmd == 0 && n == 0 && rule->set == RESVD_SET)
792  *	// RESVD_SET is protected only if cmd == 0 and n == 0 ("ipfw flush")
793  *
794  *   match_set ::= (cmd == 0 || rule->set == set)
795  *	// set number is ignored for cmd == 0
796  *
797  *   match_number ::= (cmd == 1 || n == 0 || n == rule->rulenum)
798  *	// number is ignored for cmd == 1 or n == 0
799  *
800  */
801 int
802 ipfw_match_range(struct ip_fw *rule, ipfw_range_tlv *rt)
803 {
804 
805 	/* Don't match default rule for modification queries */
806 	if (rule->rulenum == IPFW_DEFAULT_RULE &&
807 	    (rt->flags & IPFW_RCFLAG_DEFAULT) == 0)
808 		return (0);
809 
810 	/* Don't match rules in reserved set for flush requests */
811 	if ((rt->flags & IPFW_RCFLAG_ALL) != 0 && rule->set == RESVD_SET)
812 		return (0);
813 
814 	/* If we're filtering by set, don't match other sets */
815 	if ((rt->flags & IPFW_RCFLAG_SET) != 0 && rule->set != rt->set)
816 		return (0);
817 
818 	if ((rt->flags & IPFW_RCFLAG_RANGE) != 0 &&
819 	    (rule->rulenum < rt->start_rule || rule->rulenum > rt->end_rule))
820 		return (0);
821 
822 	return (1);
823 }
824 
825 /*
826  * Delete rules matching range @rt.
827  * Saves number of deleted rules in @ndel.
828  *
829  * Returns 0 on success.
830  */
831 static int
832 delete_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int *ndel)
833 {
834 	struct ip_fw *reap, *rule, **map;
835 	int end, start;
836 	int i, n, ndyn, ofs;
837 
838 	reap = NULL;
839 	IPFW_UH_WLOCK(chain);	/* arbitrate writers */
840 
841 	/*
842 	 * Stage 1: Determine range to inspect.
843 	 * Range is half-inclusive, e.g [start, end).
844 	 */
845 	start = 0;
846 	end = chain->n_rules - 1;
847 
848 	if ((rt->flags & IPFW_RCFLAG_RANGE) != 0) {
849 		start = ipfw_find_rule(chain, rt->start_rule, 0);
850 
851 		end = ipfw_find_rule(chain, rt->end_rule, 0);
852 		if (rt->end_rule != IPFW_DEFAULT_RULE)
853 			while (chain->map[end]->rulenum == rt->end_rule)
854 				end++;
855 	}
856 
857 	/* Allocate new map of the same size */
858 	map = get_map(chain, 0, 1 /* locked */);
859 	if (map == NULL) {
860 		IPFW_UH_WUNLOCK(chain);
861 		return (ENOMEM);
862 	}
863 
864 	n = 0;
865 	ndyn = 0;
866 	ofs = start;
867 	/* 1. bcopy the initial part of the map */
868 	if (start > 0)
869 		bcopy(chain->map, map, start * sizeof(struct ip_fw *));
870 	/* 2. copy active rules between start and end */
871 	for (i = start; i < end; i++) {
872 		rule = chain->map[i];
873 		if (ipfw_match_range(rule, rt) == 0) {
874 			map[ofs++] = rule;
875 			continue;
876 		}
877 
878 		n++;
879 		if (ipfw_is_dyn_rule(rule) != 0)
880 			ndyn++;
881 	}
882 	/* 3. copy the final part of the map */
883 	bcopy(chain->map + end, map + ofs,
884 		(chain->n_rules - end) * sizeof(struct ip_fw *));
885 	/* 4. recalculate skipto cache */
886 	update_skipto_cache(chain, map);
887 	/* 5. swap the maps (under UH_WLOCK + WHLOCK) */
888 	map = swap_map(chain, map, chain->n_rules - n);
889 	/* 6. Remove all dynamic states originated by deleted rules */
890 	if (ndyn > 0)
891 		ipfw_expire_dyn_rules(chain, rt);
892 	/* 7. now remove the rules deleted from the old map */
893 	for (i = start; i < end; i++) {
894 		rule = map[i];
895 		if (ipfw_match_range(rule, rt) == 0)
896 			continue;
897 		chain->static_len -= RULEUSIZE0(rule);
898 		ipfw_reap_add(chain, &reap, rule);
899 	}
900 	IPFW_UH_WUNLOCK(chain);
901 
902 	ipfw_reap_rules(reap);
903 	if (map != NULL)
904 		free(map, M_IPFW);
905 	*ndel = n;
906 	return (0);
907 }
908 
909 /*
910  * Changes set of given rule rannge @rt
911  * with each other.
912  *
913  * Returns 0 on success.
914  */
915 static int
916 move_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt)
917 {
918 	struct ip_fw *rule;
919 	int i;
920 
921 	IPFW_UH_WLOCK(chain);
922 
923 	/*
924 	 * Move rules with matching paramenerts to a new set.
925 	 * This one is much more complex. We have to ensure
926 	 * that all referenced tables (if any) are referenced
927 	 * by given rule subset only. Otherwise, we can't move
928 	 * them to new set and have to return error.
929 	 */
930 	if (V_fw_tables_sets != 0) {
931 		if (ipfw_move_tables_sets(chain, rt, rt->new_set) != 0) {
932 			IPFW_UH_WUNLOCK(chain);
933 			return (EBUSY);
934 		}
935 	}
936 
937 	/* XXX: We have to do swap holding WLOCK */
938 	for (i = 0; i < chain->n_rules; i++) {
939 		rule = chain->map[i];
940 		if (ipfw_match_range(rule, rt) == 0)
941 			continue;
942 		rule->set = rt->new_set;
943 	}
944 
945 	IPFW_UH_WUNLOCK(chain);
946 
947 	return (0);
948 }
949 
950 /*
951  * Clear counters for a specific rule.
952  * Normally run under IPFW_UH_RLOCK, but these are idempotent ops
953  * so we only care that rules do not disappear.
954  */
955 static void
956 clear_counters(struct ip_fw *rule, int log_only)
957 {
958 	ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
959 
960 	if (log_only == 0)
961 		IPFW_ZERO_RULE_COUNTER(rule);
962 	if (l->o.opcode == O_LOG)
963 		l->log_left = l->max_log;
964 }
965 
966 /*
967  * Flushes rules counters and/or log values on matching range.
968  *
969  * Returns number of items cleared.
970  */
971 static int
972 clear_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int log_only)
973 {
974 	struct ip_fw *rule;
975 	int num;
976 	int i;
977 
978 	num = 0;
979 	rt->flags |= IPFW_RCFLAG_DEFAULT;
980 
981 	IPFW_UH_WLOCK(chain);	/* arbitrate writers */
982 	for (i = 0; i < chain->n_rules; i++) {
983 		rule = chain->map[i];
984 		if (ipfw_match_range(rule, rt) == 0)
985 			continue;
986 		clear_counters(rule, log_only);
987 		num++;
988 	}
989 	IPFW_UH_WUNLOCK(chain);
990 
991 	return (num);
992 }
993 
994 static int
995 check_range_tlv(ipfw_range_tlv *rt)
996 {
997 
998 	if (rt->head.length != sizeof(*rt))
999 		return (1);
1000 	if (rt->start_rule > rt->end_rule)
1001 		return (1);
1002 	if (rt->set >= IPFW_MAX_SETS || rt->new_set >= IPFW_MAX_SETS)
1003 		return (1);
1004 
1005 	if ((rt->flags & IPFW_RCFLAG_USER) != rt->flags)
1006 		return (1);
1007 
1008 	return (0);
1009 }
1010 
1011 /*
1012  * Delete rules matching specified parameters
1013  * Data layout (v0)(current):
1014  * Request: [ ipfw_obj_header ipfw_range_tlv ]
1015  * Reply: [ ipfw_obj_header ipfw_range_tlv ]
1016  *
1017  * Saves number of deleted rules in ipfw_range_tlv->new_set.
1018  *
1019  * Returns 0 on success.
1020  */
1021 static int
1022 del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1023     struct sockopt_data *sd)
1024 {
1025 	ipfw_range_header *rh;
1026 	int error, ndel;
1027 
1028 	if (sd->valsize != sizeof(*rh))
1029 		return (EINVAL);
1030 
1031 	rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
1032 
1033 	if (check_range_tlv(&rh->range) != 0)
1034 		return (EINVAL);
1035 
1036 	ndel = 0;
1037 	if ((error = delete_range(chain, &rh->range, &ndel)) != 0)
1038 		return (error);
1039 
1040 	/* Save number of rules deleted */
1041 	rh->range.new_set = ndel;
1042 	return (0);
1043 }
1044 
1045 /*
1046  * Move rules/sets matching specified parameters
1047  * Data layout (v0)(current):
1048  * Request: [ ipfw_obj_header ipfw_range_tlv ]
1049  *
1050  * Returns 0 on success.
1051  */
1052 static int
1053 move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1054     struct sockopt_data *sd)
1055 {
1056 	ipfw_range_header *rh;
1057 
1058 	if (sd->valsize != sizeof(*rh))
1059 		return (EINVAL);
1060 
1061 	rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
1062 
1063 	if (check_range_tlv(&rh->range) != 0)
1064 		return (EINVAL);
1065 
1066 	return (move_range(chain, &rh->range));
1067 }
1068 
1069 /*
1070  * Clear rule accounting data matching specified parameters
1071  * Data layout (v0)(current):
1072  * Request: [ ipfw_obj_header ipfw_range_tlv ]
1073  * Reply: [ ipfw_obj_header ipfw_range_tlv ]
1074  *
1075  * Saves number of cleared rules in ipfw_range_tlv->new_set.
1076  *
1077  * Returns 0 on success.
1078  */
1079 static int
1080 clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1081     struct sockopt_data *sd)
1082 {
1083 	ipfw_range_header *rh;
1084 	int log_only, num;
1085 	char *msg;
1086 
1087 	if (sd->valsize != sizeof(*rh))
1088 		return (EINVAL);
1089 
1090 	rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
1091 
1092 	if (check_range_tlv(&rh->range) != 0)
1093 		return (EINVAL);
1094 
1095 	log_only = (op3->opcode == IP_FW_XRESETLOG);
1096 
1097 	num = clear_range(chain, &rh->range, log_only);
1098 
1099 	if (rh->range.flags & IPFW_RCFLAG_ALL)
1100 		msg = log_only ? "All logging counts reset" :
1101 		    "Accounting cleared";
1102 	else
1103 		msg = log_only ? "logging count reset" : "cleared";
1104 
1105 	if (V_fw_verbose) {
1106 		int lev = LOG_SECURITY | LOG_NOTICE;
1107 		log(lev, "ipfw: %s.\n", msg);
1108 	}
1109 
1110 	/* Save number of rules cleared */
1111 	rh->range.new_set = num;
1112 	return (0);
1113 }
1114 
1115 static void
1116 enable_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt)
1117 {
1118 	uint32_t v_set;
1119 
1120 	IPFW_UH_WLOCK_ASSERT(chain);
1121 
1122 	/* Change enabled/disabled sets mask */
1123 	v_set = (V_set_disable | rt->set) & ~rt->new_set;
1124 	v_set &= ~(1 << RESVD_SET); /* set RESVD_SET always enabled */
1125 	IPFW_WLOCK(chain);
1126 	V_set_disable = v_set;
1127 	IPFW_WUNLOCK(chain);
1128 }
1129 
1130 static void
1131 swap_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int mv)
1132 {
1133 	struct ip_fw *rule;
1134 	int i;
1135 
1136 	IPFW_UH_WLOCK_ASSERT(chain);
1137 
1138 	/* Swap or move two sets */
1139 	for (i = 0; i < chain->n_rules - 1; i++) {
1140 		rule = chain->map[i];
1141 		if (rule->set == rt->set)
1142 			rule->set = rt->new_set;
1143 		else if (rule->set == rt->new_set && mv == 0)
1144 			rule->set = rt->set;
1145 	}
1146 	if (V_fw_tables_sets != 0)
1147 		ipfw_swap_tables_sets(chain, rt->set, rt->new_set, mv);
1148 }
1149 
1150 /*
1151  * Swaps or moves set
1152  * Data layout (v0)(current):
1153  * Request: [ ipfw_obj_header ipfw_range_tlv ]
1154  *
1155  * Returns 0 on success.
1156  */
1157 static int
1158 manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1159     struct sockopt_data *sd)
1160 {
1161 	ipfw_range_header *rh;
1162 
1163 	if (sd->valsize != sizeof(*rh))
1164 		return (EINVAL);
1165 
1166 	rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
1167 
1168 	if (rh->range.head.length != sizeof(ipfw_range_tlv))
1169 		return (1);
1170 
1171 	IPFW_UH_WLOCK(chain);
1172 	switch (op3->opcode) {
1173 	case IP_FW_SET_SWAP:
1174 	case IP_FW_SET_MOVE:
1175 		swap_sets(chain, &rh->range, op3->opcode == IP_FW_SET_MOVE);
1176 		break;
1177 	case IP_FW_SET_ENABLE:
1178 		enable_sets(chain, &rh->range);
1179 		break;
1180 	}
1181 	IPFW_UH_WUNLOCK(chain);
1182 
1183 	return (0);
1184 }
1185 
1186 /**
1187  * Remove all rules with given number, or do set manipulation.
1188  * Assumes chain != NULL && *chain != NULL.
1189  *
1190  * The argument is an uint32_t. The low 16 bit are the rule or set number;
1191  * the next 8 bits are the new set; the top 8 bits indicate the command:
1192  *
1193  *	0	delete rules numbered "rulenum"
1194  *	1	delete rules in set "rulenum"
1195  *	2	move rules "rulenum" to set "new_set"
1196  *	3	move rules from set "rulenum" to set "new_set"
1197  *	4	swap sets "rulenum" and "new_set"
1198  *	5	delete rules "rulenum" and set "new_set"
1199  */
1200 static int
1201 del_entry(struct ip_fw_chain *chain, uint32_t arg)
1202 {
1203 	uint32_t num;	/* rule number or old_set */
1204 	uint8_t cmd, new_set;
1205 	int do_del, ndel;
1206 	int error = 0;
1207 	ipfw_range_tlv rt;
1208 
1209 	num = arg & 0xffff;
1210 	cmd = (arg >> 24) & 0xff;
1211 	new_set = (arg >> 16) & 0xff;
1212 
1213 	if (cmd > 5 || new_set > RESVD_SET)
1214 		return EINVAL;
1215 	if (cmd == 0 || cmd == 2 || cmd == 5) {
1216 		if (num >= IPFW_DEFAULT_RULE)
1217 			return EINVAL;
1218 	} else {
1219 		if (num > RESVD_SET)	/* old_set */
1220 			return EINVAL;
1221 	}
1222 
1223 	/* Convert old requests into new representation */
1224 	memset(&rt, 0, sizeof(rt));
1225 	rt.start_rule = num;
1226 	rt.end_rule = num;
1227 	rt.set = num;
1228 	rt.new_set = new_set;
1229 	do_del = 0;
1230 
1231 	switch (cmd) {
1232 	case 0: /* delete rules numbered "rulenum" */
1233 		if (num == 0)
1234 			rt.flags |= IPFW_RCFLAG_ALL;
1235 		else
1236 			rt.flags |= IPFW_RCFLAG_RANGE;
1237 		do_del = 1;
1238 		break;
1239 	case 1: /* delete rules in set "rulenum" */
1240 		rt.flags |= IPFW_RCFLAG_SET;
1241 		do_del = 1;
1242 		break;
1243 	case 5: /* delete rules "rulenum" and set "new_set" */
1244 		rt.flags |= IPFW_RCFLAG_RANGE | IPFW_RCFLAG_SET;
1245 		rt.set = new_set;
1246 		rt.new_set = 0;
1247 		do_del = 1;
1248 		break;
1249 	case 2: /* move rules "rulenum" to set "new_set" */
1250 		rt.flags |= IPFW_RCFLAG_RANGE;
1251 		break;
1252 	case 3: /* move rules from set "rulenum" to set "new_set" */
1253 		IPFW_UH_WLOCK(chain);
1254 		swap_sets(chain, &rt, 1);
1255 		IPFW_UH_WUNLOCK(chain);
1256 		return (0);
1257 	case 4: /* swap sets "rulenum" and "new_set" */
1258 		IPFW_UH_WLOCK(chain);
1259 		swap_sets(chain, &rt, 0);
1260 		IPFW_UH_WUNLOCK(chain);
1261 		return (0);
1262 	default:
1263 		return (ENOTSUP);
1264 	}
1265 
1266 	if (do_del != 0) {
1267 		if ((error = delete_range(chain, &rt, &ndel)) != 0)
1268 			return (error);
1269 
1270 		if (ndel == 0 && (cmd != 1 && num != 0))
1271 			return (EINVAL);
1272 
1273 		return (0);
1274 	}
1275 
1276 	return (move_range(chain, &rt));
1277 }
1278 
1279 /**
1280  * Reset some or all counters on firewall rules.
1281  * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
1282  * the next 8 bits are the set number, the top 8 bits are the command:
1283  *	0	work with rules from all set's;
1284  *	1	work with rules only from specified set.
1285  * Specified rule number is zero if we want to clear all entries.
1286  * log_only is 1 if we only want to reset logs, zero otherwise.
1287  */
1288 static int
1289 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
1290 {
1291 	struct ip_fw *rule;
1292 	char *msg;
1293 	int i;
1294 
1295 	uint16_t rulenum = arg & 0xffff;
1296 	uint8_t set = (arg >> 16) & 0xff;
1297 	uint8_t cmd = (arg >> 24) & 0xff;
1298 
1299 	if (cmd > 1)
1300 		return (EINVAL);
1301 	if (cmd == 1 && set > RESVD_SET)
1302 		return (EINVAL);
1303 
1304 	IPFW_UH_RLOCK(chain);
1305 	if (rulenum == 0) {
1306 		V_norule_counter = 0;
1307 		for (i = 0; i < chain->n_rules; i++) {
1308 			rule = chain->map[i];
1309 			/* Skip rules not in our set. */
1310 			if (cmd == 1 && rule->set != set)
1311 				continue;
1312 			clear_counters(rule, log_only);
1313 		}
1314 		msg = log_only ? "All logging counts reset" :
1315 		    "Accounting cleared";
1316 	} else {
1317 		int cleared = 0;
1318 		for (i = 0; i < chain->n_rules; i++) {
1319 			rule = chain->map[i];
1320 			if (rule->rulenum == rulenum) {
1321 				if (cmd == 0 || rule->set == set)
1322 					clear_counters(rule, log_only);
1323 				cleared = 1;
1324 			}
1325 			if (rule->rulenum > rulenum)
1326 				break;
1327 		}
1328 		if (!cleared) {	/* we did not find any matching rules */
1329 			IPFW_UH_RUNLOCK(chain);
1330 			return (EINVAL);
1331 		}
1332 		msg = log_only ? "logging count reset" : "cleared";
1333 	}
1334 	IPFW_UH_RUNLOCK(chain);
1335 
1336 	if (V_fw_verbose) {
1337 		int lev = LOG_SECURITY | LOG_NOTICE;
1338 
1339 		if (rulenum)
1340 			log(lev, "ipfw: Entry %d %s.\n", rulenum, msg);
1341 		else
1342 			log(lev, "ipfw: %s.\n", msg);
1343 	}
1344 	return (0);
1345 }
1346 
1347 
1348 /*
1349  * Check rule head in FreeBSD11 format
1350  *
1351  */
1352 static int
1353 check_ipfw_rule1(struct ip_fw_rule *rule, int size,
1354     struct rule_check_info *ci)
1355 {
1356 	int l;
1357 
1358 	if (size < sizeof(*rule)) {
1359 		printf("ipfw: rule too short\n");
1360 		return (EINVAL);
1361 	}
1362 
1363 	/* Check for valid cmd_len */
1364 	l = roundup2(RULESIZE(rule), sizeof(uint64_t));
1365 	if (l != size) {
1366 		printf("ipfw: size mismatch (have %d want %d)\n", size, l);
1367 		return (EINVAL);
1368 	}
1369 	if (rule->act_ofs >= rule->cmd_len) {
1370 		printf("ipfw: bogus action offset (%u > %u)\n",
1371 		    rule->act_ofs, rule->cmd_len - 1);
1372 		return (EINVAL);
1373 	}
1374 
1375 	if (rule->rulenum > IPFW_DEFAULT_RULE - 1)
1376 		return (EINVAL);
1377 
1378 	return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci));
1379 }
1380 
1381 /*
1382  * Check rule head in FreeBSD8 format
1383  *
1384  */
1385 static int
1386 check_ipfw_rule0(struct ip_fw_rule0 *rule, int size,
1387     struct rule_check_info *ci)
1388 {
1389 	int l;
1390 
1391 	if (size < sizeof(*rule)) {
1392 		printf("ipfw: rule too short\n");
1393 		return (EINVAL);
1394 	}
1395 
1396 	/* Check for valid cmd_len */
1397 	l = sizeof(*rule) + rule->cmd_len * 4 - 4;
1398 	if (l != size) {
1399 		printf("ipfw: size mismatch (have %d want %d)\n", size, l);
1400 		return (EINVAL);
1401 	}
1402 	if (rule->act_ofs >= rule->cmd_len) {
1403 		printf("ipfw: bogus action offset (%u > %u)\n",
1404 		    rule->act_ofs, rule->cmd_len - 1);
1405 		return (EINVAL);
1406 	}
1407 
1408 	if (rule->rulenum > IPFW_DEFAULT_RULE - 1)
1409 		return (EINVAL);
1410 
1411 	return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci));
1412 }
1413 
1414 static int
1415 check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len, struct rule_check_info *ci)
1416 {
1417 	int cmdlen, l;
1418 	int have_action;
1419 
1420 	have_action = 0;
1421 
1422 	/*
1423 	 * Now go for the individual checks. Very simple ones, basically only
1424 	 * instruction sizes.
1425 	 */
1426 	for (l = cmd_len; l > 0 ; l -= cmdlen, cmd += cmdlen) {
1427 		cmdlen = F_LEN(cmd);
1428 		if (cmdlen > l) {
1429 			printf("ipfw: opcode %d size truncated\n",
1430 			    cmd->opcode);
1431 			return EINVAL;
1432 		}
1433 		switch (cmd->opcode) {
1434 		case O_PROBE_STATE:
1435 		case O_KEEP_STATE:
1436 		case O_PROTO:
1437 		case O_IP_SRC_ME:
1438 		case O_IP_DST_ME:
1439 		case O_LAYER2:
1440 		case O_IN:
1441 		case O_FRAG:
1442 		case O_DIVERTED:
1443 		case O_IPOPT:
1444 		case O_IPTOS:
1445 		case O_IPPRECEDENCE:
1446 		case O_IPVER:
1447 		case O_SOCKARG:
1448 		case O_TCPFLAGS:
1449 		case O_TCPOPTS:
1450 		case O_ESTAB:
1451 		case O_VERREVPATH:
1452 		case O_VERSRCREACH:
1453 		case O_ANTISPOOF:
1454 		case O_IPSEC:
1455 #ifdef INET6
1456 		case O_IP6_SRC_ME:
1457 		case O_IP6_DST_ME:
1458 		case O_EXT_HDR:
1459 		case O_IP6:
1460 #endif
1461 		case O_IP4:
1462 		case O_TAG:
1463 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1464 				goto bad_size;
1465 			break;
1466 
1467 		case O_FIB:
1468 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1469 				goto bad_size;
1470 			if (cmd->arg1 >= rt_numfibs) {
1471 				printf("ipfw: invalid fib number %d\n",
1472 					cmd->arg1);
1473 				return EINVAL;
1474 			}
1475 			break;
1476 
1477 		case O_SETFIB:
1478 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1479 				goto bad_size;
1480 			if ((cmd->arg1 != IP_FW_TARG) &&
1481 			    ((cmd->arg1 & 0x7FFFF) >= rt_numfibs)) {
1482 				printf("ipfw: invalid fib number %d\n",
1483 					cmd->arg1 & 0x7FFFF);
1484 				return EINVAL;
1485 			}
1486 			goto check_action;
1487 
1488 		case O_UID:
1489 		case O_GID:
1490 		case O_JAIL:
1491 		case O_IP_SRC:
1492 		case O_IP_DST:
1493 		case O_TCPSEQ:
1494 		case O_TCPACK:
1495 		case O_PROB:
1496 		case O_ICMPTYPE:
1497 			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
1498 				goto bad_size;
1499 			break;
1500 
1501 		case O_LIMIT:
1502 			if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
1503 				goto bad_size;
1504 			break;
1505 
1506 		case O_LOG:
1507 			if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
1508 				goto bad_size;
1509 
1510 			((ipfw_insn_log *)cmd)->log_left =
1511 			    ((ipfw_insn_log *)cmd)->max_log;
1512 
1513 			break;
1514 
1515 		case O_IP_SRC_MASK:
1516 		case O_IP_DST_MASK:
1517 			/* only odd command lengths */
1518 			if ( !(cmdlen & 1) || cmdlen > 31)
1519 				goto bad_size;
1520 			break;
1521 
1522 		case O_IP_SRC_SET:
1523 		case O_IP_DST_SET:
1524 			if (cmd->arg1 == 0 || cmd->arg1 > 256) {
1525 				printf("ipfw: invalid set size %d\n",
1526 					cmd->arg1);
1527 				return EINVAL;
1528 			}
1529 			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
1530 			    (cmd->arg1+31)/32 )
1531 				goto bad_size;
1532 			break;
1533 
1534 		case O_IP_SRC_LOOKUP:
1535 		case O_IP_DST_LOOKUP:
1536 			if (cmd->arg1 >= V_fw_tables_max) {
1537 				printf("ipfw: invalid table number %d\n",
1538 				    cmd->arg1);
1539 				return (EINVAL);
1540 			}
1541 			if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
1542 			    cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1 &&
1543 			    cmdlen != F_INSN_SIZE(ipfw_insn_u32))
1544 				goto bad_size;
1545 			ci->table_opcodes++;
1546 			break;
1547 		case O_IP_FLOW_LOOKUP:
1548 			if (cmd->arg1 >= V_fw_tables_max) {
1549 				printf("ipfw: invalid table number %d\n",
1550 				    cmd->arg1);
1551 				return (EINVAL);
1552 			}
1553 			if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
1554 			    cmdlen != F_INSN_SIZE(ipfw_insn_u32))
1555 				goto bad_size;
1556 			ci->table_opcodes++;
1557 			break;
1558 		case O_MACADDR2:
1559 			if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
1560 				goto bad_size;
1561 			break;
1562 
1563 		case O_NOP:
1564 		case O_IPID:
1565 		case O_IPTTL:
1566 		case O_IPLEN:
1567 		case O_TCPDATALEN:
1568 		case O_TCPWIN:
1569 		case O_TAGGED:
1570 			if (cmdlen < 1 || cmdlen > 31)
1571 				goto bad_size;
1572 			break;
1573 
1574 		case O_DSCP:
1575 			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1)
1576 				goto bad_size;
1577 			break;
1578 
1579 		case O_MAC_TYPE:
1580 		case O_IP_SRCPORT:
1581 		case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
1582 			if (cmdlen < 2 || cmdlen > 31)
1583 				goto bad_size;
1584 			break;
1585 
1586 		case O_RECV:
1587 		case O_XMIT:
1588 		case O_VIA:
1589 			if (((ipfw_insn_if *)cmd)->name[0] == '\1')
1590 				ci->table_opcodes++;
1591 			if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
1592 				goto bad_size;
1593 			break;
1594 
1595 		case O_ALTQ:
1596 			if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
1597 				goto bad_size;
1598 			break;
1599 
1600 		case O_PIPE:
1601 		case O_QUEUE:
1602 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1603 				goto bad_size;
1604 			goto check_action;
1605 
1606 		case O_FORWARD_IP:
1607 			if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
1608 				goto bad_size;
1609 			goto check_action;
1610 #ifdef INET6
1611 		case O_FORWARD_IP6:
1612 			if (cmdlen != F_INSN_SIZE(ipfw_insn_sa6))
1613 				goto bad_size;
1614 			goto check_action;
1615 #endif /* INET6 */
1616 
1617 		case O_DIVERT:
1618 		case O_TEE:
1619 			if (ip_divert_ptr == NULL)
1620 				return EINVAL;
1621 			else
1622 				goto check_size;
1623 		case O_NETGRAPH:
1624 		case O_NGTEE:
1625 			if (ng_ipfw_input_p == NULL)
1626 				return EINVAL;
1627 			else
1628 				goto check_size;
1629 		case O_NAT:
1630 			if (!IPFW_NAT_LOADED)
1631 				return EINVAL;
1632 			if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
1633  				goto bad_size;
1634  			goto check_action;
1635 		case O_FORWARD_MAC: /* XXX not implemented yet */
1636 		case O_CHECK_STATE:
1637 		case O_COUNT:
1638 		case O_ACCEPT:
1639 		case O_DENY:
1640 		case O_REJECT:
1641 		case O_SETDSCP:
1642 #ifdef INET6
1643 		case O_UNREACH6:
1644 #endif
1645 		case O_SKIPTO:
1646 		case O_REASS:
1647 		case O_CALLRETURN:
1648 check_size:
1649 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1650 				goto bad_size;
1651 check_action:
1652 			if (have_action) {
1653 				printf("ipfw: opcode %d, multiple actions"
1654 					" not allowed\n",
1655 					cmd->opcode);
1656 				return (EINVAL);
1657 			}
1658 			have_action = 1;
1659 			if (l != cmdlen) {
1660 				printf("ipfw: opcode %d, action must be"
1661 					" last opcode\n",
1662 					cmd->opcode);
1663 				return (EINVAL);
1664 			}
1665 			break;
1666 #ifdef INET6
1667 		case O_IP6_SRC:
1668 		case O_IP6_DST:
1669 			if (cmdlen != F_INSN_SIZE(struct in6_addr) +
1670 			    F_INSN_SIZE(ipfw_insn))
1671 				goto bad_size;
1672 			break;
1673 
1674 		case O_FLOW6ID:
1675 			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
1676 			    ((ipfw_insn_u32 *)cmd)->o.arg1)
1677 				goto bad_size;
1678 			break;
1679 
1680 		case O_IP6_SRC_MASK:
1681 		case O_IP6_DST_MASK:
1682 			if ( !(cmdlen & 1) || cmdlen > 127)
1683 				goto bad_size;
1684 			break;
1685 		case O_ICMP6TYPE:
1686 			if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
1687 				goto bad_size;
1688 			break;
1689 #endif
1690 
1691 		default:
1692 			switch (cmd->opcode) {
1693 #ifndef INET6
1694 			case O_IP6_SRC_ME:
1695 			case O_IP6_DST_ME:
1696 			case O_EXT_HDR:
1697 			case O_IP6:
1698 			case O_UNREACH6:
1699 			case O_IP6_SRC:
1700 			case O_IP6_DST:
1701 			case O_FLOW6ID:
1702 			case O_IP6_SRC_MASK:
1703 			case O_IP6_DST_MASK:
1704 			case O_ICMP6TYPE:
1705 				printf("ipfw: no IPv6 support in kernel\n");
1706 				return (EPROTONOSUPPORT);
1707 #endif
1708 			default:
1709 				printf("ipfw: opcode %d, unknown opcode\n",
1710 					cmd->opcode);
1711 				return (EINVAL);
1712 			}
1713 		}
1714 	}
1715 	if (have_action == 0) {
1716 		printf("ipfw: missing action\n");
1717 		return (EINVAL);
1718 	}
1719 	return 0;
1720 
1721 bad_size:
1722 	printf("ipfw: opcode %d size %d wrong\n",
1723 		cmd->opcode, cmdlen);
1724 	return (EINVAL);
1725 }
1726 
1727 
1728 /*
1729  * Translation of requests for compatibility with FreeBSD 7.2/8.
1730  * a static variable tells us if we have an old client from userland,
1731  * and if necessary we translate requests and responses between the
1732  * two formats.
1733  */
1734 static int is7 = 0;
1735 
1736 struct ip_fw7 {
1737 	struct ip_fw7	*next;		/* linked list of rules     */
1738 	struct ip_fw7	*next_rule;	/* ptr to next [skipto] rule    */
1739 	/* 'next_rule' is used to pass up 'set_disable' status      */
1740 
1741 	uint16_t	act_ofs;	/* offset of action in 32-bit units */
1742 	uint16_t	cmd_len;	/* # of 32-bit words in cmd */
1743 	uint16_t	rulenum;	/* rule number          */
1744 	uint8_t		set;		/* rule set (0..31)     */
1745 	// #define RESVD_SET   31  /* set for default and persistent rules */
1746 	uint8_t		_pad;		/* padding          */
1747 	// uint32_t        id;             /* rule id, only in v.8 */
1748 	/* These fields are present in all rules.           */
1749 	uint64_t	pcnt;		/* Packet counter       */
1750 	uint64_t	bcnt;		/* Byte counter         */
1751 	uint32_t	timestamp;	/* tv_sec of last match     */
1752 
1753 	ipfw_insn	cmd[1];		/* storage for commands     */
1754 };
1755 
1756 static int convert_rule_to_7(struct ip_fw_rule0 *rule);
1757 static int convert_rule_to_8(struct ip_fw_rule0 *rule);
1758 
1759 #ifndef RULESIZE7
1760 #define RULESIZE7(rule)  (sizeof(struct ip_fw7) + \
1761 	((struct ip_fw7 *)(rule))->cmd_len * 4 - 4)
1762 #endif
1763 
1764 
1765 /*
1766  * Copy the static and dynamic rules to the supplied buffer
1767  * and return the amount of space actually used.
1768  * Must be run under IPFW_UH_RLOCK
1769  */
1770 static size_t
1771 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
1772 {
1773 	char *bp = buf;
1774 	char *ep = bp + space;
1775 	struct ip_fw *rule;
1776 	struct ip_fw_rule0 *dst;
1777 	int error, i, l, warnflag;
1778 	time_t	boot_seconds;
1779 
1780 	warnflag = 0;
1781 
1782         boot_seconds = boottime.tv_sec;
1783 	for (i = 0; i < chain->n_rules; i++) {
1784 		rule = chain->map[i];
1785 
1786 		if (is7) {
1787 		    /* Convert rule to FreeBSd 7.2 format */
1788 		    l = RULESIZE7(rule);
1789 		    if (bp + l + sizeof(uint32_t) <= ep) {
1790 			bcopy(rule, bp, l + sizeof(uint32_t));
1791 			error = ipfw_rewrite_table_kidx(chain,
1792 			    (struct ip_fw_rule0 *)bp);
1793 			if (error != 0)
1794 				return (0);
1795 			error = convert_rule_to_7((struct ip_fw_rule0 *) bp);
1796 			if (error)
1797 				return 0; /*XXX correct? */
1798 			/*
1799 			 * XXX HACK. Store the disable mask in the "next"
1800 			 * pointer in a wild attempt to keep the ABI the same.
1801 			 * Why do we do this on EVERY rule?
1802 			 */
1803 			bcopy(&V_set_disable,
1804 				&(((struct ip_fw7 *)bp)->next_rule),
1805 				sizeof(V_set_disable));
1806 			if (((struct ip_fw7 *)bp)->timestamp)
1807 			    ((struct ip_fw7 *)bp)->timestamp += boot_seconds;
1808 			bp += l;
1809 		    }
1810 		    continue; /* go to next rule */
1811 		}
1812 
1813 		l = RULEUSIZE0(rule);
1814 		if (bp + l > ep) { /* should not happen */
1815 			printf("overflow dumping static rules\n");
1816 			break;
1817 		}
1818 		dst = (struct ip_fw_rule0 *)bp;
1819 		export_rule0(rule, dst, l);
1820 		error = ipfw_rewrite_table_kidx(chain, dst);
1821 
1822 		/*
1823 		 * XXX HACK. Store the disable mask in the "next"
1824 		 * pointer in a wild attempt to keep the ABI the same.
1825 		 * Why do we do this on EVERY rule?
1826 		 *
1827 		 * XXX: "ipfw set show" (ab)uses IP_FW_GET to read disabled mask
1828 		 * so we need to fail _after_ saving at least one mask.
1829 		 */
1830 		bcopy(&V_set_disable, &dst->next_rule, sizeof(V_set_disable));
1831 		if (dst->timestamp)
1832 			dst->timestamp += boot_seconds;
1833 		bp += l;
1834 
1835 		if (error != 0) {
1836 			if (error == 2) {
1837 				/* Non-fatal table rewrite error. */
1838 				warnflag = 1;
1839 				continue;
1840 			}
1841 			printf("Stop on rule %d. Fail to convert table\n",
1842 			    rule->rulenum);
1843 			break;
1844 		}
1845 	}
1846 	if (warnflag != 0)
1847 		printf("ipfw: process %s is using legacy interfaces,"
1848 		    " consider rebuilding\n", "");
1849 	ipfw_get_dynamic(chain, &bp, ep); /* protected by the dynamic lock */
1850 	return (bp - (char *)buf);
1851 }
1852 
1853 
1854 struct dump_args {
1855 	uint32_t	b;	/* start rule */
1856 	uint32_t	e;	/* end rule */
1857 	uint32_t	rcount;	/* number of rules */
1858 	uint32_t	rsize;	/* rules size */
1859 	uint32_t	tcount;	/* number of tables */
1860 	int		rcounters;	/* counters */
1861 };
1862 
1863 /*
1864  * Dumps static rules with table TLVs in buffer @sd.
1865  *
1866  * Returns 0 on success.
1867  */
1868 static int
1869 dump_static_rules(struct ip_fw_chain *chain, struct dump_args *da,
1870     uint32_t *bmask, struct sockopt_data *sd)
1871 {
1872 	int error;
1873 	int i, l;
1874 	uint32_t tcount;
1875 	ipfw_obj_ctlv *ctlv;
1876 	struct ip_fw *krule;
1877 	caddr_t dst;
1878 
1879 	/* Dump table names first (if any) */
1880 	if (da->tcount > 0) {
1881 		/* Header first */
1882 		ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv));
1883 		if (ctlv == NULL)
1884 			return (ENOMEM);
1885 		ctlv->head.type = IPFW_TLV_TBLNAME_LIST;
1886 		ctlv->head.length = da->tcount * sizeof(ipfw_obj_ntlv) +
1887 		    sizeof(*ctlv);
1888 		ctlv->count = da->tcount;
1889 		ctlv->objsize = sizeof(ipfw_obj_ntlv);
1890 	}
1891 
1892 	i = 0;
1893 	tcount = da->tcount;
1894 	while (tcount > 0) {
1895 		if ((bmask[i / 32] & (1 << (i % 32))) == 0) {
1896 			i++;
1897 			continue;
1898 		}
1899 
1900 		if ((error = ipfw_export_table_ntlv(chain, i, sd)) != 0)
1901 			return (error);
1902 
1903 		i++;
1904 		tcount--;
1905 	}
1906 
1907 	/* Dump rules */
1908 	ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv));
1909 	if (ctlv == NULL)
1910 		return (ENOMEM);
1911 	ctlv->head.type = IPFW_TLV_RULE_LIST;
1912 	ctlv->head.length = da->rsize + sizeof(*ctlv);
1913 	ctlv->count = da->rcount;
1914 
1915 	for (i = da->b; i < da->e; i++) {
1916 		krule = chain->map[i];
1917 
1918 		l = RULEUSIZE1(krule) + sizeof(ipfw_obj_tlv);
1919 		if (da->rcounters != 0)
1920 			l += sizeof(struct ip_fw_bcounter);
1921 		dst = (caddr_t)ipfw_get_sopt_space(sd, l);
1922 		if (dst == NULL)
1923 			return (ENOMEM);
1924 
1925 		export_rule1(krule, dst, l, da->rcounters);
1926 	}
1927 
1928 	return (0);
1929 }
1930 
1931 /*
1932  * Dumps requested objects data
1933  * Data layout (version 0)(current):
1934  * Request: [ ipfw_cfg_lheader ] + IPFW_CFG_GET_* flags
1935  *   size = ipfw_cfg_lheader.size
1936  * Reply: [ ipfw_cfg_lheader
1937  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional)
1938  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST)
1939  *     ipfw_obj_tlv(IPFW_TLV_RULE_ENT) [ ip_fw_bcounter (optional) ip_fw_rule ]
1940  *   ] (optional)
1941  *   [ ipfw_obj_ctlv(IPFW_TLV_STATE_LIST) ipfw_obj_dyntlv x N ] (optional)
1942  * ]
1943  * * NOTE IPFW_TLV_STATE_LIST has the single valid field: objsize.
1944  * The rest (size, count) are set to zero and needs to be ignored.
1945  *
1946  * Returns 0 on success.
1947  */
1948 static int
1949 dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1950     struct sockopt_data *sd)
1951 {
1952 	ipfw_cfg_lheader *hdr;
1953 	struct ip_fw *rule;
1954 	size_t sz, rnum;
1955 	uint32_t hdr_flags;
1956 	int error, i;
1957 	struct dump_args da;
1958 	uint32_t *bmask;
1959 
1960 	hdr = (ipfw_cfg_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr));
1961 	if (hdr == NULL)
1962 		return (EINVAL);
1963 
1964 	error = 0;
1965 	bmask = NULL;
1966 	/* Allocate needed state */
1967 	if (hdr->flags & IPFW_CFG_GET_STATIC)
1968 		bmask = malloc(IPFW_TABLES_MAX / 8, M_TEMP, M_WAITOK | M_ZERO);
1969 
1970 	IPFW_UH_RLOCK(chain);
1971 
1972 	/*
1973 	 * STAGE 1: Determine size/count for objects in range.
1974 	 * Prepare used tables bitmask.
1975 	 */
1976 	sz = sizeof(ipfw_cfg_lheader);
1977 	memset(&da, 0, sizeof(da));
1978 
1979 	da.b = 0;
1980 	da.e = chain->n_rules;
1981 
1982 	if (hdr->end_rule != 0) {
1983 		/* Handle custom range */
1984 		if ((rnum = hdr->start_rule) > IPFW_DEFAULT_RULE)
1985 			rnum = IPFW_DEFAULT_RULE;
1986 		da.b = ipfw_find_rule(chain, rnum, 0);
1987 		rnum = hdr->end_rule;
1988 		rnum = (rnum < IPFW_DEFAULT_RULE) ? rnum+1 : IPFW_DEFAULT_RULE;
1989 		da.e = ipfw_find_rule(chain, rnum, 0) + 1;
1990 	}
1991 
1992 	if (hdr->flags & IPFW_CFG_GET_STATIC) {
1993 		for (i = da.b; i < da.e; i++) {
1994 			rule = chain->map[i];
1995 			da.rsize += RULEUSIZE1(rule) + sizeof(ipfw_obj_tlv);
1996 			da.rcount++;
1997 			da.tcount += ipfw_mark_table_kidx(chain, rule, bmask);
1998 		}
1999 		/* Add counters if requested */
2000 		if (hdr->flags & IPFW_CFG_GET_COUNTERS) {
2001 			da.rsize += sizeof(struct ip_fw_bcounter) * da.rcount;
2002 			da.rcounters = 1;
2003 		}
2004 
2005 		if (da.tcount > 0)
2006 			sz += da.tcount * sizeof(ipfw_obj_ntlv) +
2007 			    sizeof(ipfw_obj_ctlv);
2008 		sz += da.rsize + sizeof(ipfw_obj_ctlv);
2009 	}
2010 
2011 	if (hdr->flags & IPFW_CFG_GET_STATES)
2012 		sz += ipfw_dyn_get_count() * sizeof(ipfw_obj_dyntlv) +
2013 		     sizeof(ipfw_obj_ctlv);
2014 
2015 
2016 	/*
2017 	 * Fill header anyway.
2018 	 * Note we have to save header fields to stable storage
2019 	 * buffer inside @sd can be flushed after dumping rules
2020 	 */
2021 	hdr->size = sz;
2022 	hdr->set_mask = ~V_set_disable;
2023 	hdr_flags = hdr->flags;
2024 	hdr = NULL;
2025 
2026 	if (sd->valsize < sz) {
2027 		error = ENOMEM;
2028 		goto cleanup;
2029 	}
2030 
2031 	/* STAGE2: Store actual data */
2032 	if (hdr_flags & IPFW_CFG_GET_STATIC) {
2033 		error = dump_static_rules(chain, &da, bmask, sd);
2034 		if (error != 0)
2035 			goto cleanup;
2036 	}
2037 
2038 	if (hdr_flags & IPFW_CFG_GET_STATES)
2039 		error = ipfw_dump_states(chain, sd);
2040 
2041 cleanup:
2042 	IPFW_UH_RUNLOCK(chain);
2043 
2044 	if (bmask != NULL)
2045 		free(bmask, M_TEMP);
2046 
2047 	return (error);
2048 }
2049 
2050 static int
2051 check_object_name(ipfw_obj_ntlv *ntlv)
2052 {
2053 	int error;
2054 
2055 	switch (ntlv->head.type) {
2056 	case IPFW_TLV_TBL_NAME:
2057 		error = ipfw_check_table_name(ntlv->name);
2058 		break;
2059 	default:
2060 		error = ENOTSUP;
2061 	}
2062 
2063 	return (0);
2064 }
2065 
2066 /*
2067  * Adds one or more rules to ipfw @chain.
2068  * Data layout (version 0)(current):
2069  * Request:
2070  * [
2071  *   ip_fw3_opheader
2072  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional *1)
2073  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] (*2) (*3)
2074  * ]
2075  * Reply:
2076  * [
2077  *   ip_fw3_opheader
2078  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional)
2079  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ]
2080  * ]
2081  *
2082  * Rules in reply are modified to store their actual ruleset number.
2083  *
2084  * (*1) TLVs inside IPFW_TLV_TBL_LIST needs to be sorted ascending
2085  * accoring to their idx field and there has to be no duplicates.
2086  * (*2) Numbered rules inside IPFW_TLV_RULE_LIST needs to be sorted ascending.
2087  * (*3) Each ip_fw structure needs to be aligned to u64 boundary.
2088  *
2089  * Returns 0 on success.
2090  */
2091 static int
2092 add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
2093     struct sockopt_data *sd)
2094 {
2095 	ipfw_obj_ctlv *ctlv, *rtlv, *tstate;
2096 	ipfw_obj_ntlv *ntlv;
2097 	int clen, error, idx;
2098 	uint32_t count, read;
2099 	struct ip_fw_rule *r;
2100 	struct rule_check_info rci, *ci, *cbuf;
2101 	int i, rsize;
2102 
2103 	op3 = (ip_fw3_opheader *)ipfw_get_sopt_space(sd, sd->valsize);
2104 	ctlv = (ipfw_obj_ctlv *)(op3 + 1);
2105 
2106 	read = sizeof(ip_fw3_opheader);
2107 	rtlv = NULL;
2108 	tstate = NULL;
2109 	cbuf = NULL;
2110 	memset(&rci, 0, sizeof(struct rule_check_info));
2111 
2112 	if (read + sizeof(*ctlv) > sd->valsize)
2113 		return (EINVAL);
2114 
2115 	if (ctlv->head.type == IPFW_TLV_TBLNAME_LIST) {
2116 		clen = ctlv->head.length;
2117 		/* Check size and alignment */
2118 		if (clen > sd->valsize || clen < sizeof(*ctlv))
2119 			return (EINVAL);
2120 		if ((clen % sizeof(uint64_t)) != 0)
2121 			return (EINVAL);
2122 
2123 		/*
2124 		 * Some table names or other named objects.
2125 		 * Check for validness.
2126 		 */
2127 		count = (ctlv->head.length - sizeof(*ctlv)) / sizeof(*ntlv);
2128 		if (ctlv->count != count || ctlv->objsize != sizeof(*ntlv))
2129 			return (EINVAL);
2130 
2131 		/*
2132 		 * Check each TLV.
2133 		 * Ensure TLVs are sorted ascending and
2134 		 * there are no duplicates.
2135 		 */
2136 		idx = -1;
2137 		ntlv = (ipfw_obj_ntlv *)(ctlv + 1);
2138 		while (count > 0) {
2139 			if (ntlv->head.length != sizeof(ipfw_obj_ntlv))
2140 				return (EINVAL);
2141 
2142 			error = check_object_name(ntlv);
2143 			if (error != 0)
2144 				return (error);
2145 
2146 			if (ntlv->idx <= idx)
2147 				return (EINVAL);
2148 
2149 			idx = ntlv->idx;
2150 			count--;
2151 			ntlv++;
2152 		}
2153 
2154 		tstate = ctlv;
2155 		read += ctlv->head.length;
2156 		ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length);
2157 	}
2158 
2159 	if (read + sizeof(*ctlv) > sd->valsize)
2160 		return (EINVAL);
2161 
2162 	if (ctlv->head.type == IPFW_TLV_RULE_LIST) {
2163 		clen = ctlv->head.length;
2164 		if (clen + read > sd->valsize || clen < sizeof(*ctlv))
2165 			return (EINVAL);
2166 		if ((clen % sizeof(uint64_t)) != 0)
2167 			return (EINVAL);
2168 
2169 		/*
2170 		 * TODO: Permit adding multiple rules at once
2171 		 */
2172 		if (ctlv->count != 1)
2173 			return (ENOTSUP);
2174 
2175 		clen -= sizeof(*ctlv);
2176 
2177 		if (ctlv->count > clen / sizeof(struct ip_fw_rule))
2178 			return (EINVAL);
2179 
2180 		/* Allocate state for each rule or use stack */
2181 		if (ctlv->count == 1) {
2182 			memset(&rci, 0, sizeof(struct rule_check_info));
2183 			cbuf = &rci;
2184 		} else
2185 			cbuf = malloc(ctlv->count * sizeof(*ci), M_TEMP,
2186 			    M_WAITOK | M_ZERO);
2187 		ci = cbuf;
2188 
2189 		/*
2190 		 * Check each rule for validness.
2191 		 * Ensure numbered rules are sorted ascending
2192 		 * and properly aligned
2193 		 */
2194 		idx = 0;
2195 		r = (struct ip_fw_rule *)(ctlv + 1);
2196 		count = 0;
2197 		error = 0;
2198 		while (clen > 0) {
2199 			rsize = roundup2(RULESIZE(r), sizeof(uint64_t));
2200 			if (rsize > clen || ctlv->count <= count) {
2201 				error = EINVAL;
2202 				break;
2203 			}
2204 
2205 			ci->ctlv = tstate;
2206 			error = check_ipfw_rule1(r, rsize, ci);
2207 			if (error != 0)
2208 				break;
2209 
2210 			/* Check sorting */
2211 			if (r->rulenum != 0 && r->rulenum < idx) {
2212 				printf("rulenum %d idx %d\n", r->rulenum, idx);
2213 				error = EINVAL;
2214 				break;
2215 			}
2216 			idx = r->rulenum;
2217 
2218 			ci->urule = (caddr_t)r;
2219 
2220 			rsize = roundup2(rsize, sizeof(uint64_t));
2221 			clen -= rsize;
2222 			r = (struct ip_fw_rule *)((caddr_t)r + rsize);
2223 			count++;
2224 			ci++;
2225 		}
2226 
2227 		if (ctlv->count != count || error != 0) {
2228 			if (cbuf != &rci)
2229 				free(cbuf, M_TEMP);
2230 			return (EINVAL);
2231 		}
2232 
2233 		rtlv = ctlv;
2234 		read += ctlv->head.length;
2235 		ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length);
2236 	}
2237 
2238 	if (read != sd->valsize || rtlv == NULL || rtlv->count == 0) {
2239 		if (cbuf != NULL && cbuf != &rci)
2240 			free(cbuf, M_TEMP);
2241 		return (EINVAL);
2242 	}
2243 
2244 	/*
2245 	 * Passed rules seems to be valid.
2246 	 * Allocate storage and try to add them to chain.
2247 	 */
2248 	for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) {
2249 		clen = RULEKSIZE1((struct ip_fw_rule *)ci->urule);
2250 		ci->krule = ipfw_alloc_rule(chain, clen);
2251 		import_rule1(ci);
2252 	}
2253 
2254 	if ((error = commit_rules(chain, cbuf, rtlv->count)) != 0) {
2255 		/* Free allocate krules */
2256 		for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++)
2257 			free(ci->krule, M_IPFW);
2258 	}
2259 
2260 	if (cbuf != NULL && cbuf != &rci)
2261 		free(cbuf, M_TEMP);
2262 
2263 	return (error);
2264 }
2265 
2266 /*
2267  * Lists all sopts currently registered.
2268  * Data layout (v0)(current):
2269  * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size
2270  * Reply: [ ipfw_obj_lheader ipfw_sopt_info x N ]
2271  *
2272  * Returns 0 on success
2273  */
2274 static int
2275 dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
2276     struct sockopt_data *sd)
2277 {
2278 	struct _ipfw_obj_lheader *olh;
2279 	ipfw_sopt_info *i;
2280 	struct ipfw_sopt_handler *sh;
2281 	uint32_t count, n, size;
2282 
2283 	olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh));
2284 	if (olh == NULL)
2285 		return (EINVAL);
2286 	if (sd->valsize < olh->size)
2287 		return (EINVAL);
2288 
2289 	CTL3_LOCK();
2290 	count = ctl3_hsize;
2291 	size = count * sizeof(ipfw_sopt_info) + sizeof(ipfw_obj_lheader);
2292 
2293 	/* Fill in header regadless of buffer size */
2294 	olh->count = count;
2295 	olh->objsize = sizeof(ipfw_sopt_info);
2296 
2297 	if (size > olh->size) {
2298 		olh->size = size;
2299 		CTL3_UNLOCK();
2300 		return (ENOMEM);
2301 	}
2302 	olh->size = size;
2303 
2304 	for (n = 1; n <= count; n++) {
2305 		i = (ipfw_sopt_info *)ipfw_get_sopt_space(sd, sizeof(*i));
2306 		KASSERT(i != 0, ("previously checked buffer is not enough"));
2307 		sh = &ctl3_handlers[n];
2308 		i->opcode = sh->opcode;
2309 		i->version = sh->version;
2310 		i->refcnt = sh->refcnt;
2311 	}
2312 	CTL3_UNLOCK();
2313 
2314 	return (0);
2315 }
2316 
2317 /*
2318  * Compares two sopt handlers (code, version and handler ptr).
2319  * Used both as qsort() and bsearch().
2320  * Does not compare handler for latter case.
2321  *
2322  * Returns 0 if match is found.
2323  */
2324 static int
2325 compare_sh(const void *_a, const void *_b)
2326 {
2327 	const struct ipfw_sopt_handler *a, *b;
2328 
2329 	a = (const struct ipfw_sopt_handler *)_a;
2330 	b = (const struct ipfw_sopt_handler *)_b;
2331 
2332 	if (a->opcode < b->opcode)
2333 		return (-1);
2334 	else if (a->opcode > b->opcode)
2335 		return (1);
2336 
2337 	if (a->version < b->version)
2338 		return (-1);
2339 	else if (a->version > b->version)
2340 		return (1);
2341 
2342 	/* bsearch helper */
2343 	if (a->handler == NULL)
2344 		return (0);
2345 
2346 	if ((uintptr_t)a->handler < (uintptr_t)b->handler)
2347 		return (-1);
2348 	else if ((uintptr_t)b->handler > (uintptr_t)b->handler)
2349 		return (1);
2350 
2351 	return (0);
2352 }
2353 
2354 /*
2355  * Finds sopt handler based on @code and @version.
2356  *
2357  * Returns pointer to handler or NULL.
2358  */
2359 static struct ipfw_sopt_handler *
2360 find_sh(uint16_t code, uint8_t version, void *handler)
2361 {
2362 	struct ipfw_sopt_handler *sh, h;
2363 
2364 	memset(&h, 0, sizeof(h));
2365 	h.opcode = code;
2366 	h.version = version;
2367 	h.handler = handler;
2368 
2369 	sh = (struct ipfw_sopt_handler *)bsearch(&h, ctl3_handlers,
2370 	    ctl3_hsize, sizeof(h), compare_sh);
2371 
2372 	return (sh);
2373 }
2374 
2375 static int
2376 find_ref_sh(uint16_t opcode, uint8_t version, struct ipfw_sopt_handler *psh)
2377 {
2378 	struct ipfw_sopt_handler *sh;
2379 
2380 	CTL3_LOCK();
2381 	if ((sh = find_sh(opcode, version, NULL)) == NULL) {
2382 		CTL3_UNLOCK();
2383 		printf("ipfw: ipfw_ctl3 invalid option %d""v""%d\n",
2384 		    opcode, version);
2385 		return (EINVAL);
2386 	}
2387 	sh->refcnt++;
2388 	ctl3_refct++;
2389 	/* Copy handler data to requested buffer */
2390 	*psh = *sh;
2391 	CTL3_UNLOCK();
2392 
2393 	return (0);
2394 }
2395 
2396 static void
2397 find_unref_sh(struct ipfw_sopt_handler *psh)
2398 {
2399 	struct ipfw_sopt_handler *sh;
2400 
2401 	CTL3_LOCK();
2402 	sh = find_sh(psh->opcode, psh->version, NULL);
2403 	KASSERT(sh != NULL, ("ctl3 handler disappeared"));
2404 	sh->refcnt--;
2405 	ctl3_refct--;
2406 	CTL3_UNLOCK();
2407 }
2408 
2409 void
2410 ipfw_init_sopt_handler()
2411 {
2412 
2413 	CTL3_LOCK_INIT();
2414 	IPFW_ADD_SOPT_HANDLER(1, scodes);
2415 }
2416 
2417 void
2418 ipfw_destroy_sopt_handler()
2419 {
2420 
2421 	IPFW_DEL_SOPT_HANDLER(1, scodes);
2422 	CTL3_LOCK_DESTROY();
2423 }
2424 
2425 /*
2426  * Adds one or more sockopt handlers to the global array.
2427  * Function may sleep.
2428  */
2429 void
2430 ipfw_add_sopt_handler(struct ipfw_sopt_handler *sh, size_t count)
2431 {
2432 	size_t sz;
2433 	struct ipfw_sopt_handler *tmp;
2434 
2435 	CTL3_LOCK();
2436 
2437 	for (;;) {
2438 		sz = ctl3_hsize + count;
2439 		CTL3_UNLOCK();
2440 		tmp = malloc(sizeof(*sh) * sz, M_IPFW, M_WAITOK | M_ZERO);
2441 		CTL3_LOCK();
2442 		if (ctl3_hsize + count <= sz)
2443 			break;
2444 
2445 		/* Retry */
2446 		free(tmp, M_IPFW);
2447 	}
2448 
2449 	/* Merge old & new arrays */
2450 	sz = ctl3_hsize + count;
2451 	memcpy(tmp, ctl3_handlers, ctl3_hsize * sizeof(*sh));
2452 	memcpy(&tmp[ctl3_hsize], sh, count * sizeof(*sh));
2453 	qsort(tmp, sz, sizeof(*sh), compare_sh);
2454 	/* Switch new and free old */
2455 	if (ctl3_handlers != NULL)
2456 		free(ctl3_handlers, M_IPFW);
2457 	ctl3_handlers = tmp;
2458 	ctl3_hsize = sz;
2459 	ctl3_gencnt++;
2460 
2461 	CTL3_UNLOCK();
2462 }
2463 
2464 /*
2465  * Removes one or more sockopt handlers from the global array.
2466  */
2467 int
2468 ipfw_del_sopt_handler(struct ipfw_sopt_handler *sh, size_t count)
2469 {
2470 	size_t sz;
2471 	struct ipfw_sopt_handler *tmp, *h;
2472 	int i;
2473 
2474 	CTL3_LOCK();
2475 
2476 	for (i = 0; i < count; i++) {
2477 		tmp = &sh[i];
2478 		h = find_sh(tmp->opcode, tmp->version, tmp->handler);
2479 		if (h == NULL)
2480 			continue;
2481 
2482 		sz = (ctl3_handlers + ctl3_hsize - (h + 1)) * sizeof(*h);
2483 		memmove(h, h + 1, sz);
2484 		ctl3_hsize--;
2485 	}
2486 
2487 	if (ctl3_hsize == 0) {
2488 		if (ctl3_handlers != NULL)
2489 			free(ctl3_handlers, M_IPFW);
2490 		ctl3_handlers = NULL;
2491 	}
2492 
2493 	ctl3_gencnt++;
2494 
2495 	CTL3_UNLOCK();
2496 
2497 	return (0);
2498 }
2499 
2500 /*
2501  * Writes data accumulated in @sd to sockopt buffer.
2502  * Zeroes internal @sd buffer.
2503  */
2504 static int
2505 ipfw_flush_sopt_data(struct sockopt_data *sd)
2506 {
2507 	struct sockopt *sopt;
2508 	int error;
2509 	size_t sz;
2510 
2511 	sz = sd->koff;
2512 	if (sz == 0)
2513 		return (0);
2514 
2515 	sopt = sd->sopt;
2516 
2517 	if (sopt->sopt_dir == SOPT_GET) {
2518 		error = copyout(sd->kbuf, sopt->sopt_val, sz);
2519 		if (error != 0)
2520 			return (error);
2521 	}
2522 
2523 	memset(sd->kbuf, 0, sd->ksize);
2524 	sd->ktotal += sz;
2525 	sd->koff = 0;
2526 	if (sd->ktotal + sd->ksize < sd->valsize)
2527 		sd->kavail = sd->ksize;
2528 	else
2529 		sd->kavail = sd->valsize - sd->ktotal;
2530 
2531 	/* Update sopt buffer data */
2532 	sopt->sopt_valsize = sd->ktotal;
2533 	sopt->sopt_val = sd->sopt_val + sd->ktotal;
2534 
2535 	return (0);
2536 }
2537 
2538 /*
2539  * Ensures that @sd buffer has contigious @neeeded number of
2540  * bytes.
2541  *
2542  * Returns pointer to requested space or NULL.
2543  */
2544 caddr_t
2545 ipfw_get_sopt_space(struct sockopt_data *sd, size_t needed)
2546 {
2547 	int error;
2548 	caddr_t addr;
2549 
2550 	if (sd->kavail < needed) {
2551 		/*
2552 		 * Flush data and try another time.
2553 		 */
2554 		error = ipfw_flush_sopt_data(sd);
2555 
2556 		if (sd->kavail < needed || error != 0)
2557 			return (NULL);
2558 	}
2559 
2560 	addr = sd->kbuf + sd->koff;
2561 	sd->koff += needed;
2562 	sd->kavail -= needed;
2563 	return (addr);
2564 }
2565 
2566 /*
2567  * Requests @needed contigious bytes from @sd buffer.
2568  * Function is used to notify subsystem that we are
2569  * interesed in first @needed bytes (request header)
2570  * and the rest buffer can be safely zeroed.
2571  *
2572  * Returns pointer to requested space or NULL.
2573  */
2574 caddr_t
2575 ipfw_get_sopt_header(struct sockopt_data *sd, size_t needed)
2576 {
2577 	caddr_t addr;
2578 
2579 	if ((addr = ipfw_get_sopt_space(sd, needed)) == NULL)
2580 		return (NULL);
2581 
2582 	if (sd->kavail > 0)
2583 		memset(sd->kbuf + sd->koff, 0, sd->kavail);
2584 
2585 	return (addr);
2586 }
2587 
2588 /*
2589  * New sockopt handler.
2590  */
2591 int
2592 ipfw_ctl3(struct sockopt *sopt)
2593 {
2594 	int error, locked;
2595 	size_t size, valsize;
2596 	struct ip_fw_chain *chain;
2597 	char xbuf[256];
2598 	struct sockopt_data sdata;
2599 	struct ipfw_sopt_handler h;
2600 	ip_fw3_opheader *op3 = NULL;
2601 
2602 	error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
2603 	if (error != 0)
2604 		return (error);
2605 
2606 	if (sopt->sopt_name != IP_FW3)
2607 		return (ipfw_ctl(sopt));
2608 
2609 	chain = &V_layer3_chain;
2610 	error = 0;
2611 
2612 	/* Save original valsize before it is altered via sooptcopyin() */
2613 	valsize = sopt->sopt_valsize;
2614 	memset(&sdata, 0, sizeof(sdata));
2615 	/* Read op3 header first to determine actual operation */
2616 	op3 = (ip_fw3_opheader *)xbuf;
2617 	error = sooptcopyin(sopt, op3, sizeof(*op3), sizeof(*op3));
2618 	if (error != 0)
2619 		return (error);
2620 	sopt->sopt_valsize = valsize;
2621 
2622 	/*
2623 	 * Find and reference command.
2624 	 */
2625 	error = find_ref_sh(op3->opcode, op3->version, &h);
2626 	if (error != 0)
2627 		return (error);
2628 
2629 	/*
2630 	 * Disallow modifications in really-really secure mode, but still allow
2631 	 * the logging counters to be reset.
2632 	 */
2633 	if ((h.dir & HDIR_SET) != 0 && h.opcode != IP_FW_XRESETLOG) {
2634 		error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
2635 		if (error != 0) {
2636 			find_unref_sh(&h);
2637 			return (error);
2638 		}
2639 	}
2640 
2641 	/*
2642 	 * Fill in sockopt_data structure that may be useful for
2643 	 * IP_FW3 get requests.
2644 	 */
2645 	locked = 0;
2646 	if (valsize <= sizeof(xbuf)) {
2647 		/* use on-stack buffer */
2648 		sdata.kbuf = xbuf;
2649 		sdata.ksize = sizeof(xbuf);
2650 		sdata.kavail = valsize;
2651 	} else {
2652 
2653 		/*
2654 		 * Determine opcode type/buffer size:
2655 		 * allocate sliding-window buf for data export or
2656 		 * contigious buffer for special ops.
2657 		 */
2658 		if ((h.dir & HDIR_SET) != 0) {
2659 			/* Set request. Allocate contigous buffer. */
2660 			if (valsize > CTL3_LARGEBUF) {
2661 				find_unref_sh(&h);
2662 				return (EFBIG);
2663 			}
2664 
2665 			size = valsize;
2666 		} else {
2667 			/* Get request. Allocate sliding window buffer */
2668 			size = (valsize<CTL3_SMALLBUF) ? valsize:CTL3_SMALLBUF;
2669 
2670 			if (size < valsize) {
2671 				/* We have to wire user buffer */
2672 				error = vslock(sopt->sopt_val, valsize);
2673 				if (error != 0)
2674 					return (error);
2675 				locked = 1;
2676 			}
2677 		}
2678 
2679 		sdata.kbuf = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
2680 		sdata.ksize = size;
2681 		sdata.kavail = size;
2682 	}
2683 
2684 	sdata.sopt = sopt;
2685 	sdata.sopt_val = sopt->sopt_val;
2686 	sdata.valsize = valsize;
2687 
2688 	/*
2689 	 * Copy either all request (if valsize < bsize_max)
2690 	 * or first bsize_max bytes to guarantee most consumers
2691 	 * that all necessary data has been copied).
2692 	 * Anyway, copy not less than sizeof(ip_fw3_opheader).
2693 	 */
2694 	if ((error = sooptcopyin(sopt, sdata.kbuf, sdata.ksize,
2695 	    sizeof(ip_fw3_opheader))) != 0)
2696 		return (error);
2697 	op3 = (ip_fw3_opheader *)sdata.kbuf;
2698 
2699 	/* Finally, run handler */
2700 	error = h.handler(chain, op3, &sdata);
2701 	find_unref_sh(&h);
2702 
2703 	/* Flush state and free buffers */
2704 	if (error == 0)
2705 		error = ipfw_flush_sopt_data(&sdata);
2706 	else
2707 		ipfw_flush_sopt_data(&sdata);
2708 
2709 	if (locked != 0)
2710 		vsunlock(sdata.sopt_val, valsize);
2711 
2712 	/* Restore original pointer and set number of bytes written */
2713 	sopt->sopt_val = sdata.sopt_val;
2714 	sopt->sopt_valsize = sdata.ktotal;
2715 	if (sdata.kbuf != xbuf)
2716 		free(sdata.kbuf, M_TEMP);
2717 
2718 	return (error);
2719 }
2720 
2721 /**
2722  * {set|get}sockopt parser.
2723  */
2724 int
2725 ipfw_ctl(struct sockopt *sopt)
2726 {
2727 #define	RULE_MAXSIZE	(512*sizeof(u_int32_t))
2728 	int error;
2729 	size_t size, valsize;
2730 	struct ip_fw *buf;
2731 	struct ip_fw_rule0 *rule;
2732 	struct ip_fw_chain *chain;
2733 	u_int32_t rulenum[2];
2734 	uint32_t opt;
2735 	struct rule_check_info ci;
2736 	IPFW_RLOCK_TRACKER;
2737 
2738 	chain = &V_layer3_chain;
2739 	error = 0;
2740 
2741 	/* Save original valsize before it is altered via sooptcopyin() */
2742 	valsize = sopt->sopt_valsize;
2743 	opt = sopt->sopt_name;
2744 
2745 	/*
2746 	 * Disallow modifications in really-really secure mode, but still allow
2747 	 * the logging counters to be reset.
2748 	 */
2749 	if (opt == IP_FW_ADD ||
2750 	    (sopt->sopt_dir == SOPT_SET && opt != IP_FW_RESETLOG)) {
2751 		error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
2752 		if (error != 0)
2753 			return (error);
2754 	}
2755 
2756 	switch (opt) {
2757 	case IP_FW_GET:
2758 		/*
2759 		 * pass up a copy of the current rules. Static rules
2760 		 * come first (the last of which has number IPFW_DEFAULT_RULE),
2761 		 * followed by a possibly empty list of dynamic rule.
2762 		 * The last dynamic rule has NULL in the "next" field.
2763 		 *
2764 		 * Note that the calculated size is used to bound the
2765 		 * amount of data returned to the user.  The rule set may
2766 		 * change between calculating the size and returning the
2767 		 * data in which case we'll just return what fits.
2768 		 */
2769 		for (;;) {
2770 			int len = 0, want;
2771 
2772 			size = chain->static_len;
2773 			size += ipfw_dyn_len();
2774 			if (size >= sopt->sopt_valsize)
2775 				break;
2776 			buf = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
2777 			IPFW_UH_RLOCK(chain);
2778 			/* check again how much space we need */
2779 			want = chain->static_len + ipfw_dyn_len();
2780 			if (size >= want)
2781 				len = ipfw_getrules(chain, buf, size);
2782 			IPFW_UH_RUNLOCK(chain);
2783 			if (size >= want)
2784 				error = sooptcopyout(sopt, buf, len);
2785 			free(buf, M_TEMP);
2786 			if (size >= want)
2787 				break;
2788 		}
2789 		break;
2790 
2791 	case IP_FW_FLUSH:
2792 		/* locking is done within del_entry() */
2793 		error = del_entry(chain, 0); /* special case, rule=0, cmd=0 means all */
2794 		break;
2795 
2796 	case IP_FW_ADD:
2797 		rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
2798 		error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
2799 			sizeof(struct ip_fw7) );
2800 
2801 		memset(&ci, 0, sizeof(struct rule_check_info));
2802 
2803 		/*
2804 		 * If the size of commands equals RULESIZE7 then we assume
2805 		 * a FreeBSD7.2 binary is talking to us (set is7=1).
2806 		 * is7 is persistent so the next 'ipfw list' command
2807 		 * will use this format.
2808 		 * NOTE: If wrong version is guessed (this can happen if
2809 		 *       the first ipfw command is 'ipfw [pipe] list')
2810 		 *       the ipfw binary may crash or loop infinitly...
2811 		 */
2812 		size = sopt->sopt_valsize;
2813 		if (size == RULESIZE7(rule)) {
2814 		    is7 = 1;
2815 		    error = convert_rule_to_8(rule);
2816 		    if (error) {
2817 			free(rule, M_TEMP);
2818 			return error;
2819 		    }
2820 		    size = RULESIZE(rule);
2821 		} else
2822 		    is7 = 0;
2823 		if (error == 0)
2824 			error = check_ipfw_rule0(rule, size, &ci);
2825 		if (error == 0) {
2826 			/* locking is done within add_rule() */
2827 			struct ip_fw *krule;
2828 			krule = ipfw_alloc_rule(chain, RULEKSIZE0(rule));
2829 			ci.urule = (caddr_t)rule;
2830 			ci.krule = krule;
2831 			import_rule0(&ci);
2832 			error = commit_rules(chain, &ci, 1);
2833 			if (!error && sopt->sopt_dir == SOPT_GET) {
2834 				if (is7) {
2835 					error = convert_rule_to_7(rule);
2836 					size = RULESIZE7(rule);
2837 					if (error) {
2838 						free(rule, M_TEMP);
2839 						return error;
2840 					}
2841 				}
2842 				error = sooptcopyout(sopt, rule, size);
2843 			}
2844 		}
2845 		free(rule, M_TEMP);
2846 		break;
2847 
2848 	case IP_FW_DEL:
2849 		/*
2850 		 * IP_FW_DEL is used for deleting single rules or sets,
2851 		 * and (ab)used to atomically manipulate sets. Argument size
2852 		 * is used to distinguish between the two:
2853 		 *    sizeof(u_int32_t)
2854 		 *	delete single rule or set of rules,
2855 		 *	or reassign rules (or sets) to a different set.
2856 		 *    2*sizeof(u_int32_t)
2857 		 *	atomic disable/enable sets.
2858 		 *	first u_int32_t contains sets to be disabled,
2859 		 *	second u_int32_t contains sets to be enabled.
2860 		 */
2861 		error = sooptcopyin(sopt, rulenum,
2862 			2*sizeof(u_int32_t), sizeof(u_int32_t));
2863 		if (error)
2864 			break;
2865 		size = sopt->sopt_valsize;
2866 		if (size == sizeof(u_int32_t) && rulenum[0] != 0) {
2867 			/* delete or reassign, locking done in del_entry() */
2868 			error = del_entry(chain, rulenum[0]);
2869 		} else if (size == 2*sizeof(u_int32_t)) { /* set enable/disable */
2870 			IPFW_UH_WLOCK(chain);
2871 			V_set_disable =
2872 			    (V_set_disable | rulenum[0]) & ~rulenum[1] &
2873 			    ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
2874 			IPFW_UH_WUNLOCK(chain);
2875 		} else
2876 			error = EINVAL;
2877 		break;
2878 
2879 	case IP_FW_ZERO:
2880 	case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
2881 		rulenum[0] = 0;
2882 		if (sopt->sopt_val != 0) {
2883 		    error = sooptcopyin(sopt, rulenum,
2884 			    sizeof(u_int32_t), sizeof(u_int32_t));
2885 		    if (error)
2886 			break;
2887 		}
2888 		error = zero_entry(chain, rulenum[0],
2889 			sopt->sopt_name == IP_FW_RESETLOG);
2890 		break;
2891 
2892 	/*--- TABLE opcodes ---*/
2893 	case IP_FW_TABLE_ADD:
2894 	case IP_FW_TABLE_DEL:
2895 		{
2896 			ipfw_table_entry ent;
2897 			struct tentry_info tei;
2898 			struct tid_info ti;
2899 			struct table_value v;
2900 
2901 			error = sooptcopyin(sopt, &ent,
2902 			    sizeof(ent), sizeof(ent));
2903 			if (error)
2904 				break;
2905 
2906 			memset(&tei, 0, sizeof(tei));
2907 			tei.paddr = &ent.addr;
2908 			tei.subtype = AF_INET;
2909 			tei.masklen = ent.masklen;
2910 			ipfw_import_table_value_legacy(ent.value, &v);
2911 			tei.pvalue = &v;
2912 			memset(&ti, 0, sizeof(ti));
2913 			ti.uidx = ent.tbl;
2914 			ti.type = IPFW_TABLE_CIDR;
2915 
2916 			error = (opt == IP_FW_TABLE_ADD) ?
2917 			    add_table_entry(chain, &ti, &tei, 0, 1) :
2918 			    del_table_entry(chain, &ti, &tei, 0, 1);
2919 		}
2920 		break;
2921 
2922 
2923 	case IP_FW_TABLE_FLUSH:
2924 		{
2925 			u_int16_t tbl;
2926 			struct tid_info ti;
2927 
2928 			error = sooptcopyin(sopt, &tbl,
2929 			    sizeof(tbl), sizeof(tbl));
2930 			if (error)
2931 				break;
2932 			memset(&ti, 0, sizeof(ti));
2933 			ti.uidx = tbl;
2934 			error = flush_table(chain, &ti);
2935 		}
2936 		break;
2937 
2938 	case IP_FW_TABLE_GETSIZE:
2939 		{
2940 			u_int32_t tbl, cnt;
2941 			struct tid_info ti;
2942 
2943 			if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
2944 			    sizeof(tbl))))
2945 				break;
2946 			memset(&ti, 0, sizeof(ti));
2947 			ti.uidx = tbl;
2948 			IPFW_RLOCK(chain);
2949 			error = ipfw_count_table(chain, &ti, &cnt);
2950 			IPFW_RUNLOCK(chain);
2951 			if (error)
2952 				break;
2953 			error = sooptcopyout(sopt, &cnt, sizeof(cnt));
2954 		}
2955 		break;
2956 
2957 	case IP_FW_TABLE_LIST:
2958 		{
2959 			ipfw_table *tbl;
2960 			struct tid_info ti;
2961 
2962 			if (sopt->sopt_valsize < sizeof(*tbl)) {
2963 				error = EINVAL;
2964 				break;
2965 			}
2966 			size = sopt->sopt_valsize;
2967 			tbl = malloc(size, M_TEMP, M_WAITOK);
2968 			error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
2969 			if (error) {
2970 				free(tbl, M_TEMP);
2971 				break;
2972 			}
2973 			tbl->size = (size - sizeof(*tbl)) /
2974 			    sizeof(ipfw_table_entry);
2975 			memset(&ti, 0, sizeof(ti));
2976 			ti.uidx = tbl->tbl;
2977 			IPFW_RLOCK(chain);
2978 			error = ipfw_dump_table_legacy(chain, &ti, tbl);
2979 			IPFW_RUNLOCK(chain);
2980 			if (error) {
2981 				free(tbl, M_TEMP);
2982 				break;
2983 			}
2984 			error = sooptcopyout(sopt, tbl, size);
2985 			free(tbl, M_TEMP);
2986 		}
2987 		break;
2988 
2989 	/*--- NAT operations are protected by the IPFW_LOCK ---*/
2990 	case IP_FW_NAT_CFG:
2991 		if (IPFW_NAT_LOADED)
2992 			error = ipfw_nat_cfg_ptr(sopt);
2993 		else {
2994 			printf("IP_FW_NAT_CFG: %s\n",
2995 			    "ipfw_nat not present, please load it");
2996 			error = EINVAL;
2997 		}
2998 		break;
2999 
3000 	case IP_FW_NAT_DEL:
3001 		if (IPFW_NAT_LOADED)
3002 			error = ipfw_nat_del_ptr(sopt);
3003 		else {
3004 			printf("IP_FW_NAT_DEL: %s\n",
3005 			    "ipfw_nat not present, please load it");
3006 			error = EINVAL;
3007 		}
3008 		break;
3009 
3010 	case IP_FW_NAT_GET_CONFIG:
3011 		if (IPFW_NAT_LOADED)
3012 			error = ipfw_nat_get_cfg_ptr(sopt);
3013 		else {
3014 			printf("IP_FW_NAT_GET_CFG: %s\n",
3015 			    "ipfw_nat not present, please load it");
3016 			error = EINVAL;
3017 		}
3018 		break;
3019 
3020 	case IP_FW_NAT_GET_LOG:
3021 		if (IPFW_NAT_LOADED)
3022 			error = ipfw_nat_get_log_ptr(sopt);
3023 		else {
3024 			printf("IP_FW_NAT_GET_LOG: %s\n",
3025 			    "ipfw_nat not present, please load it");
3026 			error = EINVAL;
3027 		}
3028 		break;
3029 
3030 	default:
3031 		printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
3032 		error = EINVAL;
3033 	}
3034 
3035 	return (error);
3036 #undef RULE_MAXSIZE
3037 }
3038 #define	RULE_MAXSIZE	(256*sizeof(u_int32_t))
3039 
3040 /* Functions to convert rules 7.2 <==> 8.0 */
3041 static int
3042 convert_rule_to_7(struct ip_fw_rule0 *rule)
3043 {
3044 	/* Used to modify original rule */
3045 	struct ip_fw7 *rule7 = (struct ip_fw7 *)rule;
3046 	/* copy of original rule, version 8 */
3047 	struct ip_fw_rule0 *tmp;
3048 
3049 	/* Used to copy commands */
3050 	ipfw_insn *ccmd, *dst;
3051 	int ll = 0, ccmdlen = 0;
3052 
3053 	tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
3054 	if (tmp == NULL) {
3055 		return 1; //XXX error
3056 	}
3057 	bcopy(rule, tmp, RULE_MAXSIZE);
3058 
3059 	/* Copy fields */
3060 	//rule7->_pad = tmp->_pad;
3061 	rule7->set = tmp->set;
3062 	rule7->rulenum = tmp->rulenum;
3063 	rule7->cmd_len = tmp->cmd_len;
3064 	rule7->act_ofs = tmp->act_ofs;
3065 	rule7->next_rule = (struct ip_fw7 *)tmp->next_rule;
3066 	rule7->cmd_len = tmp->cmd_len;
3067 	rule7->pcnt = tmp->pcnt;
3068 	rule7->bcnt = tmp->bcnt;
3069 	rule7->timestamp = tmp->timestamp;
3070 
3071 	/* Copy commands */
3072 	for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule7->cmd ;
3073 			ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
3074 		ccmdlen = F_LEN(ccmd);
3075 
3076 		bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
3077 
3078 		if (dst->opcode > O_NAT)
3079 			/* O_REASS doesn't exists in 7.2 version, so
3080 			 * decrement opcode if it is after O_REASS
3081 			 */
3082 			dst->opcode--;
3083 
3084 		if (ccmdlen > ll) {
3085 			printf("ipfw: opcode %d size truncated\n",
3086 				ccmd->opcode);
3087 			return EINVAL;
3088 		}
3089 	}
3090 	free(tmp, M_TEMP);
3091 
3092 	return 0;
3093 }
3094 
3095 static int
3096 convert_rule_to_8(struct ip_fw_rule0 *rule)
3097 {
3098 	/* Used to modify original rule */
3099 	struct ip_fw7 *rule7 = (struct ip_fw7 *) rule;
3100 
3101 	/* Used to copy commands */
3102 	ipfw_insn *ccmd, *dst;
3103 	int ll = 0, ccmdlen = 0;
3104 
3105 	/* Copy of original rule */
3106 	struct ip_fw7 *tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
3107 	if (tmp == NULL) {
3108 		return 1; //XXX error
3109 	}
3110 
3111 	bcopy(rule7, tmp, RULE_MAXSIZE);
3112 
3113 	for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule->cmd ;
3114 			ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
3115 		ccmdlen = F_LEN(ccmd);
3116 
3117 		bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
3118 
3119 		if (dst->opcode > O_NAT)
3120 			/* O_REASS doesn't exists in 7.2 version, so
3121 			 * increment opcode if it is after O_REASS
3122 			 */
3123 			dst->opcode++;
3124 
3125 		if (ccmdlen > ll) {
3126 			printf("ipfw: opcode %d size truncated\n",
3127 			    ccmd->opcode);
3128 			return EINVAL;
3129 		}
3130 	}
3131 
3132 	rule->_pad = tmp->_pad;
3133 	rule->set = tmp->set;
3134 	rule->rulenum = tmp->rulenum;
3135 	rule->cmd_len = tmp->cmd_len;
3136 	rule->act_ofs = tmp->act_ofs;
3137 	rule->next_rule = (struct ip_fw *)tmp->next_rule;
3138 	rule->cmd_len = tmp->cmd_len;
3139 	rule->id = 0; /* XXX see if is ok = 0 */
3140 	rule->pcnt = tmp->pcnt;
3141 	rule->bcnt = tmp->bcnt;
3142 	rule->timestamp = tmp->timestamp;
3143 
3144 	free (tmp, M_TEMP);
3145 	return 0;
3146 }
3147 
3148 /*
3149  * Named object api
3150  *
3151  */
3152 
3153 /*
3154  * Allocate new bitmask which can be used to enlarge/shrink
3155  * named instance index.
3156  */
3157 void
3158 ipfw_objhash_bitmap_alloc(uint32_t items, void **idx, int *pblocks)
3159 {
3160 	size_t size;
3161 	int max_blocks;
3162 	u_long *idx_mask;
3163 
3164 	KASSERT((items % BLOCK_ITEMS) == 0,
3165 	   ("bitmask size needs to power of 2 and greater or equal to %zu",
3166 	    BLOCK_ITEMS));
3167 
3168 	max_blocks = items / BLOCK_ITEMS;
3169 	size = items / 8;
3170 	idx_mask = malloc(size * IPFW_MAX_SETS, M_IPFW, M_WAITOK);
3171 	/* Mark all as free */
3172 	memset(idx_mask, 0xFF, size * IPFW_MAX_SETS);
3173 	*idx_mask &= ~(u_long)1; /* Skip index 0 */
3174 
3175 	*idx = idx_mask;
3176 	*pblocks = max_blocks;
3177 }
3178 
3179 /*
3180  * Copy current bitmask index to new one.
3181  */
3182 void
3183 ipfw_objhash_bitmap_merge(struct namedobj_instance *ni, void **idx, int *blocks)
3184 {
3185 	int old_blocks, new_blocks;
3186 	u_long *old_idx, *new_idx;
3187 	int i;
3188 
3189 	old_idx = ni->idx_mask;
3190 	old_blocks = ni->max_blocks;
3191 	new_idx = *idx;
3192 	new_blocks = *blocks;
3193 
3194 	for (i = 0; i < IPFW_MAX_SETS; i++) {
3195 		memcpy(&new_idx[new_blocks * i], &old_idx[old_blocks * i],
3196 		    old_blocks * sizeof(u_long));
3197 	}
3198 }
3199 
3200 /*
3201  * Swaps current @ni index with new one.
3202  */
3203 void
3204 ipfw_objhash_bitmap_swap(struct namedobj_instance *ni, void **idx, int *blocks)
3205 {
3206 	int old_blocks;
3207 	u_long *old_idx;
3208 
3209 	old_idx = ni->idx_mask;
3210 	old_blocks = ni->max_blocks;
3211 
3212 	ni->idx_mask = *idx;
3213 	ni->max_blocks = *blocks;
3214 
3215 	/* Save old values */
3216 	*idx = old_idx;
3217 	*blocks = old_blocks;
3218 }
3219 
3220 void
3221 ipfw_objhash_bitmap_free(void *idx, int blocks)
3222 {
3223 
3224 	free(idx, M_IPFW);
3225 }
3226 
3227 /*
3228  * Creates named hash instance.
3229  * Must be called without holding any locks.
3230  * Return pointer to new instance.
3231  */
3232 struct namedobj_instance *
3233 ipfw_objhash_create(uint32_t items)
3234 {
3235 	struct namedobj_instance *ni;
3236 	int i;
3237 	size_t size;
3238 
3239 	size = sizeof(struct namedobj_instance) +
3240 	    sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE +
3241 	    sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE;
3242 
3243 	ni = malloc(size, M_IPFW, M_WAITOK | M_ZERO);
3244 	ni->nn_size = NAMEDOBJ_HASH_SIZE;
3245 	ni->nv_size = NAMEDOBJ_HASH_SIZE;
3246 
3247 	ni->names = (struct namedobjects_head *)(ni +1);
3248 	ni->values = &ni->names[ni->nn_size];
3249 
3250 	for (i = 0; i < ni->nn_size; i++)
3251 		TAILQ_INIT(&ni->names[i]);
3252 
3253 	for (i = 0; i < ni->nv_size; i++)
3254 		TAILQ_INIT(&ni->values[i]);
3255 
3256 	/* Set default hashing/comparison functions */
3257 	ni->hash_f = objhash_hash_name;
3258 	ni->cmp_f = objhash_cmp_name;
3259 
3260 	/* Allocate bitmask separately due to possible resize */
3261 	ipfw_objhash_bitmap_alloc(items, (void*)&ni->idx_mask, &ni->max_blocks);
3262 
3263 	return (ni);
3264 }
3265 
3266 void
3267 ipfw_objhash_destroy(struct namedobj_instance *ni)
3268 {
3269 
3270 	free(ni->idx_mask, M_IPFW);
3271 	free(ni, M_IPFW);
3272 }
3273 
3274 void
3275 ipfw_objhash_set_funcs(struct namedobj_instance *ni, objhash_hash_f *hash_f,
3276     objhash_cmp_f *cmp_f)
3277 {
3278 
3279 	ni->hash_f = hash_f;
3280 	ni->cmp_f = cmp_f;
3281 }
3282 
3283 static uint32_t
3284 objhash_hash_name(struct namedobj_instance *ni, void *name, uint32_t set)
3285 {
3286 
3287 	return (fnv_32_str((char *)name, FNV1_32_INIT));
3288 }
3289 
3290 static int
3291 objhash_cmp_name(struct named_object *no, void *name, uint32_t set)
3292 {
3293 
3294 	if ((strcmp(no->name, (char *)name) == 0) && (no->set == set))
3295 		return (0);
3296 
3297 	return (1);
3298 }
3299 
3300 static uint32_t
3301 objhash_hash_idx(struct namedobj_instance *ni, uint32_t val)
3302 {
3303 	uint32_t v;
3304 
3305 	v = val % (ni->nv_size - 1);
3306 
3307 	return (v);
3308 }
3309 
3310 struct named_object *
3311 ipfw_objhash_lookup_name(struct namedobj_instance *ni, uint32_t set, char *name)
3312 {
3313 	struct named_object *no;
3314 	uint32_t hash;
3315 
3316 	hash = ni->hash_f(ni, name, set) % ni->nn_size;
3317 
3318 	TAILQ_FOREACH(no, &ni->names[hash], nn_next) {
3319 		if (ni->cmp_f(no, name, set) == 0)
3320 			return (no);
3321 	}
3322 
3323 	return (NULL);
3324 }
3325 
3326 struct named_object *
3327 ipfw_objhash_lookup_kidx(struct namedobj_instance *ni, uint16_t kidx)
3328 {
3329 	struct named_object *no;
3330 	uint32_t hash;
3331 
3332 	hash = objhash_hash_idx(ni, kidx);
3333 
3334 	TAILQ_FOREACH(no, &ni->values[hash], nv_next) {
3335 		if (no->kidx == kidx)
3336 			return (no);
3337 	}
3338 
3339 	return (NULL);
3340 }
3341 
3342 int
3343 ipfw_objhash_same_name(struct namedobj_instance *ni, struct named_object *a,
3344     struct named_object *b)
3345 {
3346 
3347 	if ((strcmp(a->name, b->name) == 0) && a->set == b->set)
3348 		return (1);
3349 
3350 	return (0);
3351 }
3352 
3353 void
3354 ipfw_objhash_add(struct namedobj_instance *ni, struct named_object *no)
3355 {
3356 	uint32_t hash;
3357 
3358 	hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size;
3359 	TAILQ_INSERT_HEAD(&ni->names[hash], no, nn_next);
3360 
3361 	hash = objhash_hash_idx(ni, no->kidx);
3362 	TAILQ_INSERT_HEAD(&ni->values[hash], no, nv_next);
3363 
3364 	ni->count++;
3365 }
3366 
3367 void
3368 ipfw_objhash_del(struct namedobj_instance *ni, struct named_object *no)
3369 {
3370 	uint32_t hash;
3371 
3372 	hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size;
3373 	TAILQ_REMOVE(&ni->names[hash], no, nn_next);
3374 
3375 	hash = objhash_hash_idx(ni, no->kidx);
3376 	TAILQ_REMOVE(&ni->values[hash], no, nv_next);
3377 
3378 	ni->count--;
3379 }
3380 
3381 uint32_t
3382 ipfw_objhash_count(struct namedobj_instance *ni)
3383 {
3384 
3385 	return (ni->count);
3386 }
3387 
3388 /*
3389  * Runs @func for each found named object.
3390  * It is safe to delete objects from callback
3391  */
3392 void
3393 ipfw_objhash_foreach(struct namedobj_instance *ni, objhash_cb_t *f, void *arg)
3394 {
3395 	struct named_object *no, *no_tmp;
3396 	int i;
3397 
3398 	for (i = 0; i < ni->nn_size; i++) {
3399 		TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp)
3400 			f(ni, no, arg);
3401 	}
3402 }
3403 
3404 /*
3405  * Removes index from given set.
3406  * Returns 0 on success.
3407  */
3408 int
3409 ipfw_objhash_free_idx(struct namedobj_instance *ni, uint16_t idx)
3410 {
3411 	u_long *mask;
3412 	int i, v;
3413 
3414 	i = idx / BLOCK_ITEMS;
3415 	v = idx % BLOCK_ITEMS;
3416 
3417 	if (i >= ni->max_blocks)
3418 		return (1);
3419 
3420 	mask = &ni->idx_mask[i];
3421 
3422 	if ((*mask & ((u_long)1 << v)) != 0)
3423 		return (1);
3424 
3425 	/* Mark as free */
3426 	*mask |= (u_long)1 << v;
3427 
3428 	/* Update free offset */
3429 	if (ni->free_off[0] > i)
3430 		ni->free_off[0] = i;
3431 
3432 	return (0);
3433 }
3434 
3435 /*
3436  * Allocate new index in given instance and stores in in @pidx.
3437  * Returns 0 on success.
3438  */
3439 int
3440 ipfw_objhash_alloc_idx(void *n, uint16_t *pidx)
3441 {
3442 	struct namedobj_instance *ni;
3443 	u_long *mask;
3444 	int i, off, v;
3445 
3446 	ni = (struct namedobj_instance *)n;
3447 
3448 	off = ni->free_off[0];
3449 	mask = &ni->idx_mask[off];
3450 
3451 	for (i = off; i < ni->max_blocks; i++, mask++) {
3452 		if ((v = ffsl(*mask)) == 0)
3453 			continue;
3454 
3455 		/* Mark as busy */
3456 		*mask &= ~ ((u_long)1 << (v - 1));
3457 
3458 		ni->free_off[0] = i;
3459 
3460 		v = BLOCK_ITEMS * i + v - 1;
3461 
3462 		*pidx = v;
3463 		return (0);
3464 	}
3465 
3466 	return (1);
3467 }
3468 
3469 /* end of file */
3470