xref: /freebsd/sys/netpfil/ipfw/ip_fw_sockopt.c (revision f39bffc62c1395bde25d152c7f68fdf7cbaab414)
1 /*-
2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
3  *
4  * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
5  * Copyright (c) 2014 Yandex LLC
6  * Copyright (c) 2014 Alexander V. Chernikov
7  *
8  * Supported by: Valeria Paoli
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 /*
36  * Control socket and rule management routines for ipfw.
37  * Control is currently implemented via IP_FW3 setsockopt() code.
38  */
39 
40 #include "opt_ipfw.h"
41 #include "opt_inet.h"
42 #ifndef INET
43 #error IPFIREWALL requires INET.
44 #endif /* INET */
45 #include "opt_inet6.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/malloc.h>
50 #include <sys/mbuf.h>	/* struct m_tag used by nested headers */
51 #include <sys/kernel.h>
52 #include <sys/lock.h>
53 #include <sys/priv.h>
54 #include <sys/proc.h>
55 #include <sys/rwlock.h>
56 #include <sys/rmlock.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sysctl.h>
60 #include <sys/syslog.h>
61 #include <sys/fnv_hash.h>
62 #include <net/if.h>
63 #include <net/pfil.h>
64 #include <net/route.h>
65 #include <net/vnet.h>
66 #include <vm/vm.h>
67 #include <vm/vm_extern.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/ip_var.h> /* hooks */
71 #include <netinet/ip_fw.h>
72 
73 #include <netpfil/ipfw/ip_fw_private.h>
74 #include <netpfil/ipfw/ip_fw_table.h>
75 
76 #ifdef MAC
77 #include <security/mac/mac_framework.h>
78 #endif
79 
80 static int ipfw_ctl(struct sockopt *sopt);
81 static int check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len,
82     struct rule_check_info *ci);
83 static int check_ipfw_rule1(struct ip_fw_rule *rule, int size,
84     struct rule_check_info *ci);
85 static int check_ipfw_rule0(struct ip_fw_rule0 *rule, int size,
86     struct rule_check_info *ci);
87 static int rewrite_rule_uidx(struct ip_fw_chain *chain,
88     struct rule_check_info *ci);
89 
90 #define	NAMEDOBJ_HASH_SIZE	32
91 
92 struct namedobj_instance {
93 	struct namedobjects_head	*names;
94 	struct namedobjects_head	*values;
95 	uint32_t nn_size;		/* names hash size */
96 	uint32_t nv_size;		/* number hash size */
97 	u_long *idx_mask;		/* used items bitmask */
98 	uint32_t max_blocks;		/* number of "long" blocks in bitmask */
99 	uint32_t count;			/* number of items */
100 	uint16_t free_off[IPFW_MAX_SETS];	/* first possible free offset */
101 	objhash_hash_f	*hash_f;
102 	objhash_cmp_f	*cmp_f;
103 };
104 #define	BLOCK_ITEMS	(8 * sizeof(u_long))	/* Number of items for ffsl() */
105 
106 static uint32_t objhash_hash_name(struct namedobj_instance *ni,
107     const void *key, uint32_t kopt);
108 static uint32_t objhash_hash_idx(struct namedobj_instance *ni, uint32_t val);
109 static int objhash_cmp_name(struct named_object *no, const void *name,
110     uint32_t set);
111 
112 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
113 
114 static int dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
115     struct sockopt_data *sd);
116 static int add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
117     struct sockopt_data *sd);
118 static int del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
119     struct sockopt_data *sd);
120 static int clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
121     struct sockopt_data *sd);
122 static int move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
123     struct sockopt_data *sd);
124 static int manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
125     struct sockopt_data *sd);
126 static int dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
127     struct sockopt_data *sd);
128 static int dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
129     struct sockopt_data *sd);
130 
131 /* ctl3 handler data */
132 struct mtx ctl3_lock;
133 #define	CTL3_LOCK_INIT()	mtx_init(&ctl3_lock, "ctl3_lock", NULL, MTX_DEF)
134 #define	CTL3_LOCK_DESTROY()	mtx_destroy(&ctl3_lock)
135 #define	CTL3_LOCK()		mtx_lock(&ctl3_lock)
136 #define	CTL3_UNLOCK()		mtx_unlock(&ctl3_lock)
137 
138 static struct ipfw_sopt_handler *ctl3_handlers;
139 static size_t ctl3_hsize;
140 static uint64_t ctl3_refct, ctl3_gencnt;
141 #define	CTL3_SMALLBUF	4096			/* small page-size write buffer */
142 #define	CTL3_LARGEBUF	16 * 1024 * 1024	/* handle large rulesets */
143 
144 static int ipfw_flush_sopt_data(struct sockopt_data *sd);
145 
146 static struct ipfw_sopt_handler	scodes[] = {
147 	{ IP_FW_XGET,		0,	HDIR_GET,	dump_config },
148 	{ IP_FW_XADD,		0,	HDIR_BOTH,	add_rules },
149 	{ IP_FW_XDEL,		0,	HDIR_BOTH,	del_rules },
150 	{ IP_FW_XZERO,		0,	HDIR_SET,	clear_rules },
151 	{ IP_FW_XRESETLOG,	0,	HDIR_SET,	clear_rules },
152 	{ IP_FW_XMOVE,		0,	HDIR_SET,	move_rules },
153 	{ IP_FW_SET_SWAP,	0,	HDIR_SET,	manage_sets },
154 	{ IP_FW_SET_MOVE,	0,	HDIR_SET,	manage_sets },
155 	{ IP_FW_SET_ENABLE,	0,	HDIR_SET,	manage_sets },
156 	{ IP_FW_DUMP_SOPTCODES,	0,	HDIR_GET,	dump_soptcodes },
157 	{ IP_FW_DUMP_SRVOBJECTS,0,	HDIR_GET,	dump_srvobjects },
158 };
159 
160 static int
161 set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule);
162 static struct opcode_obj_rewrite *find_op_rw(ipfw_insn *cmd,
163     uint16_t *puidx, uint8_t *ptype);
164 static int mark_object_kidx(struct ip_fw_chain *ch, struct ip_fw *rule,
165     uint32_t *bmask);
166 static int ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule,
167     struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti);
168 static int ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd,
169     struct tid_info *ti, struct obj_idx *pidx, int *unresolved);
170 static void unref_rule_objects(struct ip_fw_chain *chain, struct ip_fw *rule);
171 static void unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd,
172     struct obj_idx *oib, struct obj_idx *end);
173 static int export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx,
174     struct sockopt_data *sd);
175 
176 /*
177  * Opcode object rewriter variables
178  */
179 struct opcode_obj_rewrite *ctl3_rewriters;
180 static size_t ctl3_rsize;
181 
182 /*
183  * static variables followed by global ones
184  */
185 
186 static VNET_DEFINE(uma_zone_t, ipfw_cntr_zone);
187 #define	V_ipfw_cntr_zone		VNET(ipfw_cntr_zone)
188 
189 void
190 ipfw_init_counters()
191 {
192 
193 	V_ipfw_cntr_zone = uma_zcreate("IPFW counters",
194 	    IPFW_RULE_CNTR_SIZE, NULL, NULL, NULL, NULL,
195 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
196 }
197 
198 void
199 ipfw_destroy_counters()
200 {
201 
202 	uma_zdestroy(V_ipfw_cntr_zone);
203 }
204 
205 struct ip_fw *
206 ipfw_alloc_rule(struct ip_fw_chain *chain, size_t rulesize)
207 {
208 	struct ip_fw *rule;
209 
210 	rule = malloc(rulesize, M_IPFW, M_WAITOK | M_ZERO);
211 	rule->cntr = uma_zalloc_pcpu(V_ipfw_cntr_zone, M_WAITOK | M_ZERO);
212 
213 	return (rule);
214 }
215 
216 static void
217 free_rule(struct ip_fw *rule)
218 {
219 
220 	uma_zfree_pcpu(V_ipfw_cntr_zone, rule->cntr);
221 	free(rule, M_IPFW);
222 }
223 
224 
225 /*
226  * Find the smallest rule >= key, id.
227  * We could use bsearch but it is so simple that we code it directly
228  */
229 int
230 ipfw_find_rule(struct ip_fw_chain *chain, uint32_t key, uint32_t id)
231 {
232 	int i, lo, hi;
233 	struct ip_fw *r;
234 
235   	for (lo = 0, hi = chain->n_rules - 1; lo < hi;) {
236 		i = (lo + hi) / 2;
237 		r = chain->map[i];
238 		if (r->rulenum < key)
239 			lo = i + 1;	/* continue from the next one */
240 		else if (r->rulenum > key)
241 			hi = i;		/* this might be good */
242 		else if (r->id < id)
243 			lo = i + 1;	/* continue from the next one */
244 		else /* r->id >= id */
245 			hi = i;		/* this might be good */
246 	}
247 	return hi;
248 }
249 
250 /*
251  * Builds skipto cache on rule set @map.
252  */
253 static void
254 update_skipto_cache(struct ip_fw_chain *chain, struct ip_fw **map)
255 {
256 	int *smap, rulenum;
257 	int i, mi;
258 
259 	IPFW_UH_WLOCK_ASSERT(chain);
260 
261 	mi = 0;
262 	rulenum = map[mi]->rulenum;
263 	smap = chain->idxmap_back;
264 
265 	if (smap == NULL)
266 		return;
267 
268 	for (i = 0; i < 65536; i++) {
269 		smap[i] = mi;
270 		/* Use the same rule index until i < rulenum */
271 		if (i != rulenum || i == 65535)
272 			continue;
273 		/* Find next rule with num > i */
274 		rulenum = map[++mi]->rulenum;
275 		while (rulenum == i)
276 			rulenum = map[++mi]->rulenum;
277 	}
278 }
279 
280 /*
281  * Swaps prepared (backup) index with current one.
282  */
283 static void
284 swap_skipto_cache(struct ip_fw_chain *chain)
285 {
286 	int *map;
287 
288 	IPFW_UH_WLOCK_ASSERT(chain);
289 	IPFW_WLOCK_ASSERT(chain);
290 
291 	map = chain->idxmap;
292 	chain->idxmap = chain->idxmap_back;
293 	chain->idxmap_back = map;
294 }
295 
296 /*
297  * Allocate and initialize skipto cache.
298  */
299 void
300 ipfw_init_skipto_cache(struct ip_fw_chain *chain)
301 {
302 	int *idxmap, *idxmap_back;
303 
304 	idxmap = malloc(65536 * sizeof(uint32_t *), M_IPFW,
305 	    M_WAITOK | M_ZERO);
306 	idxmap_back = malloc(65536 * sizeof(uint32_t *), M_IPFW,
307 	    M_WAITOK | M_ZERO);
308 
309 	/*
310 	 * Note we may be called at any time after initialization,
311 	 * for example, on first skipto rule, so we need to
312 	 * provide valid chain->idxmap on return
313 	 */
314 
315 	IPFW_UH_WLOCK(chain);
316 	if (chain->idxmap != NULL) {
317 		IPFW_UH_WUNLOCK(chain);
318 		free(idxmap, M_IPFW);
319 		free(idxmap_back, M_IPFW);
320 		return;
321 	}
322 
323 	/* Set backup pointer first to permit building cache */
324 	chain->idxmap_back = idxmap_back;
325 	update_skipto_cache(chain, chain->map);
326 	IPFW_WLOCK(chain);
327 	/* It is now safe to set chain->idxmap ptr */
328 	chain->idxmap = idxmap;
329 	swap_skipto_cache(chain);
330 	IPFW_WUNLOCK(chain);
331 	IPFW_UH_WUNLOCK(chain);
332 }
333 
334 /*
335  * Destroys skipto cache.
336  */
337 void
338 ipfw_destroy_skipto_cache(struct ip_fw_chain *chain)
339 {
340 
341 	if (chain->idxmap != NULL)
342 		free(chain->idxmap, M_IPFW);
343 	if (chain->idxmap != NULL)
344 		free(chain->idxmap_back, M_IPFW);
345 }
346 
347 
348 /*
349  * allocate a new map, returns the chain locked. extra is the number
350  * of entries to add or delete.
351  */
352 static struct ip_fw **
353 get_map(struct ip_fw_chain *chain, int extra, int locked)
354 {
355 
356 	for (;;) {
357 		struct ip_fw **map;
358 		u_int i, mflags;
359 
360 		mflags = M_ZERO | ((locked != 0) ? M_NOWAIT : M_WAITOK);
361 
362 		i = chain->n_rules + extra;
363 		map = malloc(i * sizeof(struct ip_fw *), M_IPFW, mflags);
364 		if (map == NULL) {
365 			printf("%s: cannot allocate map\n", __FUNCTION__);
366 			return NULL;
367 		}
368 		if (!locked)
369 			IPFW_UH_WLOCK(chain);
370 		if (i >= chain->n_rules + extra) /* good */
371 			return map;
372 		/* otherwise we lost the race, free and retry */
373 		if (!locked)
374 			IPFW_UH_WUNLOCK(chain);
375 		free(map, M_IPFW);
376 	}
377 }
378 
379 /*
380  * swap the maps. It is supposed to be called with IPFW_UH_WLOCK
381  */
382 static struct ip_fw **
383 swap_map(struct ip_fw_chain *chain, struct ip_fw **new_map, int new_len)
384 {
385 	struct ip_fw **old_map;
386 
387 	IPFW_WLOCK(chain);
388 	chain->id++;
389 	chain->n_rules = new_len;
390 	old_map = chain->map;
391 	chain->map = new_map;
392 	swap_skipto_cache(chain);
393 	IPFW_WUNLOCK(chain);
394 	return old_map;
395 }
396 
397 
398 static void
399 export_cntr1_base(struct ip_fw *krule, struct ip_fw_bcounter *cntr)
400 {
401 	struct timeval boottime;
402 
403 	cntr->size = sizeof(*cntr);
404 
405 	if (krule->cntr != NULL) {
406 		cntr->pcnt = counter_u64_fetch(krule->cntr);
407 		cntr->bcnt = counter_u64_fetch(krule->cntr + 1);
408 		cntr->timestamp = krule->timestamp;
409 	}
410 	if (cntr->timestamp > 0) {
411 		getboottime(&boottime);
412 		cntr->timestamp += boottime.tv_sec;
413 	}
414 }
415 
416 static void
417 export_cntr0_base(struct ip_fw *krule, struct ip_fw_bcounter0 *cntr)
418 {
419 	struct timeval boottime;
420 
421 	if (krule->cntr != NULL) {
422 		cntr->pcnt = counter_u64_fetch(krule->cntr);
423 		cntr->bcnt = counter_u64_fetch(krule->cntr + 1);
424 		cntr->timestamp = krule->timestamp;
425 	}
426 	if (cntr->timestamp > 0) {
427 		getboottime(&boottime);
428 		cntr->timestamp += boottime.tv_sec;
429 	}
430 }
431 
432 /*
433  * Copies rule @urule from v1 userland format (current).
434  * to kernel @krule.
435  * Assume @krule is zeroed.
436  */
437 static void
438 import_rule1(struct rule_check_info *ci)
439 {
440 	struct ip_fw_rule *urule;
441 	struct ip_fw *krule;
442 
443 	urule = (struct ip_fw_rule *)ci->urule;
444 	krule = (struct ip_fw *)ci->krule;
445 
446 	/* copy header */
447 	krule->act_ofs = urule->act_ofs;
448 	krule->cmd_len = urule->cmd_len;
449 	krule->rulenum = urule->rulenum;
450 	krule->set = urule->set;
451 	krule->flags = urule->flags;
452 
453 	/* Save rulenum offset */
454 	ci->urule_numoff = offsetof(struct ip_fw_rule, rulenum);
455 
456 	/* Copy opcodes */
457 	memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t));
458 }
459 
460 /*
461  * Export rule into v1 format (Current).
462  * Layout:
463  * [ ipfw_obj_tlv(IPFW_TLV_RULE_ENT)
464  *     [ ip_fw_rule ] OR
465  *     [ ip_fw_bcounter ip_fw_rule] (depends on rcntrs).
466  * ]
467  * Assume @data is zeroed.
468  */
469 static void
470 export_rule1(struct ip_fw *krule, caddr_t data, int len, int rcntrs)
471 {
472 	struct ip_fw_bcounter *cntr;
473 	struct ip_fw_rule *urule;
474 	ipfw_obj_tlv *tlv;
475 
476 	/* Fill in TLV header */
477 	tlv = (ipfw_obj_tlv *)data;
478 	tlv->type = IPFW_TLV_RULE_ENT;
479 	tlv->length = len;
480 
481 	if (rcntrs != 0) {
482 		/* Copy counters */
483 		cntr = (struct ip_fw_bcounter *)(tlv + 1);
484 		urule = (struct ip_fw_rule *)(cntr + 1);
485 		export_cntr1_base(krule, cntr);
486 	} else
487 		urule = (struct ip_fw_rule *)(tlv + 1);
488 
489 	/* copy header */
490 	urule->act_ofs = krule->act_ofs;
491 	urule->cmd_len = krule->cmd_len;
492 	urule->rulenum = krule->rulenum;
493 	urule->set = krule->set;
494 	urule->flags = krule->flags;
495 	urule->id = krule->id;
496 
497 	/* Copy opcodes */
498 	memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t));
499 }
500 
501 
502 /*
503  * Copies rule @urule from FreeBSD8 userland format (v0)
504  * to kernel @krule.
505  * Assume @krule is zeroed.
506  */
507 static void
508 import_rule0(struct rule_check_info *ci)
509 {
510 	struct ip_fw_rule0 *urule;
511 	struct ip_fw *krule;
512 	int cmdlen, l;
513 	ipfw_insn *cmd;
514 	ipfw_insn_limit *lcmd;
515 	ipfw_insn_if *cmdif;
516 
517 	urule = (struct ip_fw_rule0 *)ci->urule;
518 	krule = (struct ip_fw *)ci->krule;
519 
520 	/* copy header */
521 	krule->act_ofs = urule->act_ofs;
522 	krule->cmd_len = urule->cmd_len;
523 	krule->rulenum = urule->rulenum;
524 	krule->set = urule->set;
525 	if ((urule->_pad & 1) != 0)
526 		krule->flags |= IPFW_RULE_NOOPT;
527 
528 	/* Save rulenum offset */
529 	ci->urule_numoff = offsetof(struct ip_fw_rule0, rulenum);
530 
531 	/* Copy opcodes */
532 	memcpy(krule->cmd, urule->cmd, krule->cmd_len * sizeof(uint32_t));
533 
534 	/*
535 	 * Alter opcodes:
536 	 * 1) convert tablearg value from 65535 to 0
537 	 * 2) Add high bit to O_SETFIB/O_SETDSCP values (to make room
538 	 *    for targ).
539 	 * 3) convert table number in iface opcodes to u16
540 	 * 4) convert old `nat global` into new 65535
541 	 */
542 	l = krule->cmd_len;
543 	cmd = krule->cmd;
544 	cmdlen = 0;
545 
546 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
547 		cmdlen = F_LEN(cmd);
548 
549 		switch (cmd->opcode) {
550 		/* Opcodes supporting tablearg */
551 		case O_TAG:
552 		case O_TAGGED:
553 		case O_PIPE:
554 		case O_QUEUE:
555 		case O_DIVERT:
556 		case O_TEE:
557 		case O_SKIPTO:
558 		case O_CALLRETURN:
559 		case O_NETGRAPH:
560 		case O_NGTEE:
561 		case O_NAT:
562 			if (cmd->arg1 == IP_FW_TABLEARG)
563 				cmd->arg1 = IP_FW_TARG;
564 			else if (cmd->arg1 == 0)
565 				cmd->arg1 = IP_FW_NAT44_GLOBAL;
566 			break;
567 		case O_SETFIB:
568 		case O_SETDSCP:
569 			if (cmd->arg1 == IP_FW_TABLEARG)
570 				cmd->arg1 = IP_FW_TARG;
571 			else
572 				cmd->arg1 |= 0x8000;
573 			break;
574 		case O_LIMIT:
575 			lcmd = (ipfw_insn_limit *)cmd;
576 			if (lcmd->conn_limit == IP_FW_TABLEARG)
577 				lcmd->conn_limit = IP_FW_TARG;
578 			break;
579 		/* Interface tables */
580 		case O_XMIT:
581 		case O_RECV:
582 		case O_VIA:
583 			/* Interface table, possibly */
584 			cmdif = (ipfw_insn_if *)cmd;
585 			if (cmdif->name[0] != '\1')
586 				break;
587 
588 			cmdif->p.kidx = (uint16_t)cmdif->p.glob;
589 			break;
590 		}
591 	}
592 }
593 
594 /*
595  * Copies rule @krule from kernel to FreeBSD8 userland format (v0)
596  */
597 static void
598 export_rule0(struct ip_fw *krule, struct ip_fw_rule0 *urule, int len)
599 {
600 	int cmdlen, l;
601 	ipfw_insn *cmd;
602 	ipfw_insn_limit *lcmd;
603 	ipfw_insn_if *cmdif;
604 
605 	/* copy header */
606 	memset(urule, 0, len);
607 	urule->act_ofs = krule->act_ofs;
608 	urule->cmd_len = krule->cmd_len;
609 	urule->rulenum = krule->rulenum;
610 	urule->set = krule->set;
611 	if ((krule->flags & IPFW_RULE_NOOPT) != 0)
612 		urule->_pad |= 1;
613 
614 	/* Copy opcodes */
615 	memcpy(urule->cmd, krule->cmd, krule->cmd_len * sizeof(uint32_t));
616 
617 	/* Export counters */
618 	export_cntr0_base(krule, (struct ip_fw_bcounter0 *)&urule->pcnt);
619 
620 	/*
621 	 * Alter opcodes:
622 	 * 1) convert tablearg value from 0 to 65535
623 	 * 2) Remove highest bit from O_SETFIB/O_SETDSCP values.
624 	 * 3) convert table number in iface opcodes to int
625 	 */
626 	l = urule->cmd_len;
627 	cmd = urule->cmd;
628 	cmdlen = 0;
629 
630 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
631 		cmdlen = F_LEN(cmd);
632 
633 		switch (cmd->opcode) {
634 		/* Opcodes supporting tablearg */
635 		case O_TAG:
636 		case O_TAGGED:
637 		case O_PIPE:
638 		case O_QUEUE:
639 		case O_DIVERT:
640 		case O_TEE:
641 		case O_SKIPTO:
642 		case O_CALLRETURN:
643 		case O_NETGRAPH:
644 		case O_NGTEE:
645 		case O_NAT:
646 			if (cmd->arg1 == IP_FW_TARG)
647 				cmd->arg1 = IP_FW_TABLEARG;
648 			else if (cmd->arg1 == IP_FW_NAT44_GLOBAL)
649 				cmd->arg1 = 0;
650 			break;
651 		case O_SETFIB:
652 		case O_SETDSCP:
653 			if (cmd->arg1 == IP_FW_TARG)
654 				cmd->arg1 = IP_FW_TABLEARG;
655 			else
656 				cmd->arg1 &= ~0x8000;
657 			break;
658 		case O_LIMIT:
659 			lcmd = (ipfw_insn_limit *)cmd;
660 			if (lcmd->conn_limit == IP_FW_TARG)
661 				lcmd->conn_limit = IP_FW_TABLEARG;
662 			break;
663 		/* Interface tables */
664 		case O_XMIT:
665 		case O_RECV:
666 		case O_VIA:
667 			/* Interface table, possibly */
668 			cmdif = (ipfw_insn_if *)cmd;
669 			if (cmdif->name[0] != '\1')
670 				break;
671 
672 			cmdif->p.glob = cmdif->p.kidx;
673 			break;
674 		}
675 	}
676 }
677 
678 /*
679  * Add new rule(s) to the list possibly creating rule number for each.
680  * Update the rule_number in the input struct so the caller knows it as well.
681  * Must be called without IPFW_UH held
682  */
683 static int
684 commit_rules(struct ip_fw_chain *chain, struct rule_check_info *rci, int count)
685 {
686 	int error, i, insert_before, tcount;
687 	uint16_t rulenum, *pnum;
688 	struct rule_check_info *ci;
689 	struct ip_fw *krule;
690 	struct ip_fw **map;	/* the new array of pointers */
691 
692 	/* Check if we need to do table/obj index remap */
693 	tcount = 0;
694 	for (ci = rci, i = 0; i < count; ci++, i++) {
695 		if (ci->object_opcodes == 0)
696 			continue;
697 
698 		/*
699 		 * Rule has some object opcodes.
700 		 * We need to find (and create non-existing)
701 		 * kernel objects, and reference existing ones.
702 		 */
703 		error = rewrite_rule_uidx(chain, ci);
704 		if (error != 0) {
705 
706 			/*
707 			 * rewrite failed, state for current rule
708 			 * has been reverted. Check if we need to
709 			 * revert more.
710 			 */
711 			if (tcount > 0) {
712 
713 				/*
714 				 * We have some more table rules
715 				 * we need to rollback.
716 				 */
717 
718 				IPFW_UH_WLOCK(chain);
719 				while (ci != rci) {
720 					ci--;
721 					if (ci->object_opcodes == 0)
722 						continue;
723 					unref_rule_objects(chain,ci->krule);
724 
725 				}
726 				IPFW_UH_WUNLOCK(chain);
727 
728 			}
729 
730 			return (error);
731 		}
732 
733 		tcount++;
734 	}
735 
736 	/* get_map returns with IPFW_UH_WLOCK if successful */
737 	map = get_map(chain, count, 0 /* not locked */);
738 	if (map == NULL) {
739 		if (tcount > 0) {
740 			/* Unbind tables */
741 			IPFW_UH_WLOCK(chain);
742 			for (ci = rci, i = 0; i < count; ci++, i++) {
743 				if (ci->object_opcodes == 0)
744 					continue;
745 
746 				unref_rule_objects(chain, ci->krule);
747 			}
748 			IPFW_UH_WUNLOCK(chain);
749 		}
750 
751 		return (ENOSPC);
752 	}
753 
754 	if (V_autoinc_step < 1)
755 		V_autoinc_step = 1;
756 	else if (V_autoinc_step > 1000)
757 		V_autoinc_step = 1000;
758 
759 	/* FIXME: Handle count > 1 */
760 	ci = rci;
761 	krule = ci->krule;
762 	rulenum = krule->rulenum;
763 
764 	/* find the insertion point, we will insert before */
765 	insert_before = rulenum ? rulenum + 1 : IPFW_DEFAULT_RULE;
766 	i = ipfw_find_rule(chain, insert_before, 0);
767 	/* duplicate first part */
768 	if (i > 0)
769 		bcopy(chain->map, map, i * sizeof(struct ip_fw *));
770 	map[i] = krule;
771 	/* duplicate remaining part, we always have the default rule */
772 	bcopy(chain->map + i, map + i + 1,
773 		sizeof(struct ip_fw *) *(chain->n_rules - i));
774 	if (rulenum == 0) {
775 		/* Compute rule number and write it back */
776 		rulenum = i > 0 ? map[i-1]->rulenum : 0;
777 		if (rulenum < IPFW_DEFAULT_RULE - V_autoinc_step)
778 			rulenum += V_autoinc_step;
779 		krule->rulenum = rulenum;
780 		/* Save number to userland rule */
781 		pnum = (uint16_t *)((caddr_t)ci->urule + ci->urule_numoff);
782 		*pnum = rulenum;
783 	}
784 
785 	krule->id = chain->id + 1;
786 	update_skipto_cache(chain, map);
787 	map = swap_map(chain, map, chain->n_rules + 1);
788 	chain->static_len += RULEUSIZE0(krule);
789 	IPFW_UH_WUNLOCK(chain);
790 	if (map)
791 		free(map, M_IPFW);
792 	return (0);
793 }
794 
795 int
796 ipfw_add_protected_rule(struct ip_fw_chain *chain, struct ip_fw *rule,
797     int locked)
798 {
799 	struct ip_fw **map;
800 
801 	map = get_map(chain, 1, locked);
802 	if (map == NULL)
803 		return (ENOMEM);
804 	if (chain->n_rules > 0)
805 		bcopy(chain->map, map,
806 		    chain->n_rules * sizeof(struct ip_fw *));
807 	map[chain->n_rules] = rule;
808 	rule->rulenum = IPFW_DEFAULT_RULE;
809 	rule->set = RESVD_SET;
810 	rule->id = chain->id + 1;
811 	/* We add rule in the end of chain, no need to update skipto cache */
812 	map = swap_map(chain, map, chain->n_rules + 1);
813 	chain->static_len += RULEUSIZE0(rule);
814 	IPFW_UH_WUNLOCK(chain);
815 	free(map, M_IPFW);
816 	return (0);
817 }
818 
819 /*
820  * Adds @rule to the list of rules to reap
821  */
822 void
823 ipfw_reap_add(struct ip_fw_chain *chain, struct ip_fw **head,
824     struct ip_fw *rule)
825 {
826 
827 	IPFW_UH_WLOCK_ASSERT(chain);
828 
829 	/* Unlink rule from everywhere */
830 	unref_rule_objects(chain, rule);
831 
832 	*((struct ip_fw **)rule) = *head;
833 	*head = rule;
834 }
835 
836 /*
837  * Reclaim storage associated with a list of rules.  This is
838  * typically the list created using remove_rule.
839  * A NULL pointer on input is handled correctly.
840  */
841 void
842 ipfw_reap_rules(struct ip_fw *head)
843 {
844 	struct ip_fw *rule;
845 
846 	while ((rule = head) != NULL) {
847 		head = *((struct ip_fw **)head);
848 		free_rule(rule);
849 	}
850 }
851 
852 /*
853  * Rules to keep are
854  *	(default || reserved || !match_set || !match_number)
855  * where
856  *   default ::= (rule->rulenum == IPFW_DEFAULT_RULE)
857  *	// the default rule is always protected
858  *
859  *   reserved ::= (cmd == 0 && n == 0 && rule->set == RESVD_SET)
860  *	// RESVD_SET is protected only if cmd == 0 and n == 0 ("ipfw flush")
861  *
862  *   match_set ::= (cmd == 0 || rule->set == set)
863  *	// set number is ignored for cmd == 0
864  *
865  *   match_number ::= (cmd == 1 || n == 0 || n == rule->rulenum)
866  *	// number is ignored for cmd == 1 or n == 0
867  *
868  */
869 int
870 ipfw_match_range(struct ip_fw *rule, ipfw_range_tlv *rt)
871 {
872 
873 	/* Don't match default rule for modification queries */
874 	if (rule->rulenum == IPFW_DEFAULT_RULE &&
875 	    (rt->flags & IPFW_RCFLAG_DEFAULT) == 0)
876 		return (0);
877 
878 	/* Don't match rules in reserved set for flush requests */
879 	if ((rt->flags & IPFW_RCFLAG_ALL) != 0 && rule->set == RESVD_SET)
880 		return (0);
881 
882 	/* If we're filtering by set, don't match other sets */
883 	if ((rt->flags & IPFW_RCFLAG_SET) != 0 && rule->set != rt->set)
884 		return (0);
885 
886 	if ((rt->flags & IPFW_RCFLAG_RANGE) != 0 &&
887 	    (rule->rulenum < rt->start_rule || rule->rulenum > rt->end_rule))
888 		return (0);
889 
890 	return (1);
891 }
892 
893 struct manage_sets_args {
894 	uint16_t	set;
895 	uint8_t		new_set;
896 };
897 
898 static int
899 swap_sets_cb(struct namedobj_instance *ni, struct named_object *no,
900     void *arg)
901 {
902 	struct manage_sets_args *args;
903 
904 	args = (struct manage_sets_args *)arg;
905 	if (no->set == (uint8_t)args->set)
906 		no->set = args->new_set;
907 	else if (no->set == args->new_set)
908 		no->set = (uint8_t)args->set;
909 	return (0);
910 }
911 
912 static int
913 move_sets_cb(struct namedobj_instance *ni, struct named_object *no,
914     void *arg)
915 {
916 	struct manage_sets_args *args;
917 
918 	args = (struct manage_sets_args *)arg;
919 	if (no->set == (uint8_t)args->set)
920 		no->set = args->new_set;
921 	return (0);
922 }
923 
924 static int
925 test_sets_cb(struct namedobj_instance *ni, struct named_object *no,
926     void *arg)
927 {
928 	struct manage_sets_args *args;
929 
930 	args = (struct manage_sets_args *)arg;
931 	if (no->set != (uint8_t)args->set)
932 		return (0);
933 	if (ipfw_objhash_lookup_name_type(ni, args->new_set,
934 	    no->etlv, no->name) != NULL)
935 		return (EEXIST);
936 	return (0);
937 }
938 
939 /*
940  * Generic function to handler moving and swapping sets.
941  */
942 int
943 ipfw_obj_manage_sets(struct namedobj_instance *ni, uint16_t type,
944     uint16_t set, uint8_t new_set, enum ipfw_sets_cmd cmd)
945 {
946 	struct manage_sets_args args;
947 	struct named_object *no;
948 
949 	args.set = set;
950 	args.new_set = new_set;
951 	switch (cmd) {
952 	case SWAP_ALL:
953 		return (ipfw_objhash_foreach_type(ni, swap_sets_cb,
954 		    &args, type));
955 	case TEST_ALL:
956 		return (ipfw_objhash_foreach_type(ni, test_sets_cb,
957 		    &args, type));
958 	case MOVE_ALL:
959 		return (ipfw_objhash_foreach_type(ni, move_sets_cb,
960 		    &args, type));
961 	case COUNT_ONE:
962 		/*
963 		 * @set used to pass kidx.
964 		 * When @new_set is zero - reset object counter,
965 		 * otherwise increment it.
966 		 */
967 		no = ipfw_objhash_lookup_kidx(ni, set);
968 		if (new_set != 0)
969 			no->ocnt++;
970 		else
971 			no->ocnt = 0;
972 		return (0);
973 	case TEST_ONE:
974 		/* @set used to pass kidx */
975 		no = ipfw_objhash_lookup_kidx(ni, set);
976 		/*
977 		 * First check number of references:
978 		 * when it differs, this mean other rules are holding
979 		 * reference to given object, so it is not possible to
980 		 * change its set. Note that refcnt may account references
981 		 * to some going-to-be-added rules. Since we don't know
982 		 * their numbers (and even if they will be added) it is
983 		 * perfectly OK to return error here.
984 		 */
985 		if (no->ocnt != no->refcnt)
986 			return (EBUSY);
987 		if (ipfw_objhash_lookup_name_type(ni, new_set, type,
988 		    no->name) != NULL)
989 			return (EEXIST);
990 		return (0);
991 	case MOVE_ONE:
992 		/* @set used to pass kidx */
993 		no = ipfw_objhash_lookup_kidx(ni, set);
994 		no->set = new_set;
995 		return (0);
996 	}
997 	return (EINVAL);
998 }
999 
1000 /*
1001  * Delete rules matching range @rt.
1002  * Saves number of deleted rules in @ndel.
1003  *
1004  * Returns 0 on success.
1005  */
1006 static int
1007 delete_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int *ndel)
1008 {
1009 	struct ip_fw *reap, *rule, **map;
1010 	int end, start;
1011 	int i, n, ndyn, ofs;
1012 
1013 	reap = NULL;
1014 	IPFW_UH_WLOCK(chain);	/* arbitrate writers */
1015 
1016 	/*
1017 	 * Stage 1: Determine range to inspect.
1018 	 * Range is half-inclusive, e.g [start, end).
1019 	 */
1020 	start = 0;
1021 	end = chain->n_rules - 1;
1022 
1023 	if ((rt->flags & IPFW_RCFLAG_RANGE) != 0) {
1024 		start = ipfw_find_rule(chain, rt->start_rule, 0);
1025 
1026 		if (rt->end_rule >= IPFW_DEFAULT_RULE)
1027 			rt->end_rule = IPFW_DEFAULT_RULE - 1;
1028 		end = ipfw_find_rule(chain, rt->end_rule, UINT32_MAX);
1029 	}
1030 
1031 	/* Allocate new map of the same size */
1032 	map = get_map(chain, 0, 1 /* locked */);
1033 	if (map == NULL) {
1034 		IPFW_UH_WUNLOCK(chain);
1035 		return (ENOMEM);
1036 	}
1037 
1038 	n = 0;
1039 	ndyn = 0;
1040 	ofs = start;
1041 	/* 1. bcopy the initial part of the map */
1042 	if (start > 0)
1043 		bcopy(chain->map, map, start * sizeof(struct ip_fw *));
1044 	/* 2. copy active rules between start and end */
1045 	for (i = start; i < end; i++) {
1046 		rule = chain->map[i];
1047 		if (ipfw_match_range(rule, rt) == 0) {
1048 			map[ofs++] = rule;
1049 			continue;
1050 		}
1051 
1052 		n++;
1053 		if (ipfw_is_dyn_rule(rule) != 0)
1054 			ndyn++;
1055 	}
1056 	/* 3. copy the final part of the map */
1057 	bcopy(chain->map + end, map + ofs,
1058 		(chain->n_rules - end) * sizeof(struct ip_fw *));
1059 	/* 4. recalculate skipto cache */
1060 	update_skipto_cache(chain, map);
1061 	/* 5. swap the maps (under UH_WLOCK + WHLOCK) */
1062 	map = swap_map(chain, map, chain->n_rules - n);
1063 	/* 6. Remove all dynamic states originated by deleted rules */
1064 	if (ndyn > 0)
1065 		ipfw_expire_dyn_states(chain, rt);
1066 	/* 7. now remove the rules deleted from the old map */
1067 	for (i = start; i < end; i++) {
1068 		rule = map[i];
1069 		if (ipfw_match_range(rule, rt) == 0)
1070 			continue;
1071 		chain->static_len -= RULEUSIZE0(rule);
1072 		ipfw_reap_add(chain, &reap, rule);
1073 	}
1074 	IPFW_UH_WUNLOCK(chain);
1075 
1076 	ipfw_reap_rules(reap);
1077 	if (map != NULL)
1078 		free(map, M_IPFW);
1079 	*ndel = n;
1080 	return (0);
1081 }
1082 
1083 static int
1084 move_objects(struct ip_fw_chain *ch, ipfw_range_tlv *rt)
1085 {
1086 	struct opcode_obj_rewrite *rw;
1087 	struct ip_fw *rule;
1088 	ipfw_insn *cmd;
1089 	int cmdlen, i, l, c;
1090 	uint16_t kidx;
1091 
1092 	IPFW_UH_WLOCK_ASSERT(ch);
1093 
1094 	/* Stage 1: count number of references by given rules */
1095 	for (c = 0, i = 0; i < ch->n_rules - 1; i++) {
1096 		rule = ch->map[i];
1097 		if (ipfw_match_range(rule, rt) == 0)
1098 			continue;
1099 		if (rule->set == rt->new_set) /* nothing to do */
1100 			continue;
1101 		/* Search opcodes with named objects */
1102 		for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd;
1103 		    l > 0; l -= cmdlen, cmd += cmdlen) {
1104 			cmdlen = F_LEN(cmd);
1105 			rw = find_op_rw(cmd, &kidx, NULL);
1106 			if (rw == NULL || rw->manage_sets == NULL)
1107 				continue;
1108 			/*
1109 			 * When manage_sets() returns non-zero value to
1110 			 * COUNT_ONE command, consider this as an object
1111 			 * doesn't support sets (e.g. disabled with sysctl).
1112 			 * So, skip checks for this object.
1113 			 */
1114 			if (rw->manage_sets(ch, kidx, 1, COUNT_ONE) != 0)
1115 				continue;
1116 			c++;
1117 		}
1118 	}
1119 	if (c == 0) /* No objects found */
1120 		return (0);
1121 	/* Stage 2: verify "ownership" */
1122 	for (c = 0, i = 0; (i < ch->n_rules - 1) && c == 0; i++) {
1123 		rule = ch->map[i];
1124 		if (ipfw_match_range(rule, rt) == 0)
1125 			continue;
1126 		if (rule->set == rt->new_set) /* nothing to do */
1127 			continue;
1128 		/* Search opcodes with named objects */
1129 		for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd;
1130 		    l > 0 && c == 0; l -= cmdlen, cmd += cmdlen) {
1131 			cmdlen = F_LEN(cmd);
1132 			rw = find_op_rw(cmd, &kidx, NULL);
1133 			if (rw == NULL || rw->manage_sets == NULL)
1134 				continue;
1135 			/* Test for ownership and conflicting names */
1136 			c = rw->manage_sets(ch, kidx,
1137 			    (uint8_t)rt->new_set, TEST_ONE);
1138 		}
1139 	}
1140 	/* Stage 3: change set and cleanup */
1141 	for (i = 0; i < ch->n_rules - 1; i++) {
1142 		rule = ch->map[i];
1143 		if (ipfw_match_range(rule, rt) == 0)
1144 			continue;
1145 		if (rule->set == rt->new_set) /* nothing to do */
1146 			continue;
1147 		/* Search opcodes with named objects */
1148 		for (l = rule->cmd_len, cmdlen = 0, cmd = rule->cmd;
1149 		    l > 0; l -= cmdlen, cmd += cmdlen) {
1150 			cmdlen = F_LEN(cmd);
1151 			rw = find_op_rw(cmd, &kidx, NULL);
1152 			if (rw == NULL || rw->manage_sets == NULL)
1153 				continue;
1154 			/* cleanup object counter */
1155 			rw->manage_sets(ch, kidx,
1156 			    0 /* reset counter */, COUNT_ONE);
1157 			if (c != 0)
1158 				continue;
1159 			/* change set */
1160 			rw->manage_sets(ch, kidx,
1161 			    (uint8_t)rt->new_set, MOVE_ONE);
1162 		}
1163 	}
1164 	return (c);
1165 }/*
1166  * Changes set of given rule rannge @rt
1167  * with each other.
1168  *
1169  * Returns 0 on success.
1170  */
1171 static int
1172 move_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt)
1173 {
1174 	struct ip_fw *rule;
1175 	int i;
1176 
1177 	IPFW_UH_WLOCK(chain);
1178 
1179 	/*
1180 	 * Move rules with matching paramenerts to a new set.
1181 	 * This one is much more complex. We have to ensure
1182 	 * that all referenced tables (if any) are referenced
1183 	 * by given rule subset only. Otherwise, we can't move
1184 	 * them to new set and have to return error.
1185 	 */
1186 	if ((i = move_objects(chain, rt)) != 0) {
1187 		IPFW_UH_WUNLOCK(chain);
1188 		return (i);
1189 	}
1190 
1191 	/* XXX: We have to do swap holding WLOCK */
1192 	for (i = 0; i < chain->n_rules; i++) {
1193 		rule = chain->map[i];
1194 		if (ipfw_match_range(rule, rt) == 0)
1195 			continue;
1196 		rule->set = rt->new_set;
1197 	}
1198 
1199 	IPFW_UH_WUNLOCK(chain);
1200 
1201 	return (0);
1202 }
1203 
1204 /*
1205  * Clear counters for a specific rule.
1206  * Normally run under IPFW_UH_RLOCK, but these are idempotent ops
1207  * so we only care that rules do not disappear.
1208  */
1209 static void
1210 clear_counters(struct ip_fw *rule, int log_only)
1211 {
1212 	ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
1213 
1214 	if (log_only == 0)
1215 		IPFW_ZERO_RULE_COUNTER(rule);
1216 	if (l->o.opcode == O_LOG)
1217 		l->log_left = l->max_log;
1218 }
1219 
1220 /*
1221  * Flushes rules counters and/or log values on matching range.
1222  *
1223  * Returns number of items cleared.
1224  */
1225 static int
1226 clear_range(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int log_only)
1227 {
1228 	struct ip_fw *rule;
1229 	int num;
1230 	int i;
1231 
1232 	num = 0;
1233 	rt->flags |= IPFW_RCFLAG_DEFAULT;
1234 
1235 	IPFW_UH_WLOCK(chain);	/* arbitrate writers */
1236 	for (i = 0; i < chain->n_rules; i++) {
1237 		rule = chain->map[i];
1238 		if (ipfw_match_range(rule, rt) == 0)
1239 			continue;
1240 		clear_counters(rule, log_only);
1241 		num++;
1242 	}
1243 	IPFW_UH_WUNLOCK(chain);
1244 
1245 	return (num);
1246 }
1247 
1248 static int
1249 check_range_tlv(ipfw_range_tlv *rt)
1250 {
1251 
1252 	if (rt->head.length != sizeof(*rt))
1253 		return (1);
1254 	if (rt->start_rule > rt->end_rule)
1255 		return (1);
1256 	if (rt->set >= IPFW_MAX_SETS || rt->new_set >= IPFW_MAX_SETS)
1257 		return (1);
1258 
1259 	if ((rt->flags & IPFW_RCFLAG_USER) != rt->flags)
1260 		return (1);
1261 
1262 	return (0);
1263 }
1264 
1265 /*
1266  * Delete rules matching specified parameters
1267  * Data layout (v0)(current):
1268  * Request: [ ipfw_obj_header ipfw_range_tlv ]
1269  * Reply: [ ipfw_obj_header ipfw_range_tlv ]
1270  *
1271  * Saves number of deleted rules in ipfw_range_tlv->new_set.
1272  *
1273  * Returns 0 on success.
1274  */
1275 static int
1276 del_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1277     struct sockopt_data *sd)
1278 {
1279 	ipfw_range_header *rh;
1280 	int error, ndel;
1281 
1282 	if (sd->valsize != sizeof(*rh))
1283 		return (EINVAL);
1284 
1285 	rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
1286 
1287 	if (check_range_tlv(&rh->range) != 0)
1288 		return (EINVAL);
1289 
1290 	ndel = 0;
1291 	if ((error = delete_range(chain, &rh->range, &ndel)) != 0)
1292 		return (error);
1293 
1294 	/* Save number of rules deleted */
1295 	rh->range.new_set = ndel;
1296 	return (0);
1297 }
1298 
1299 /*
1300  * Move rules/sets matching specified parameters
1301  * Data layout (v0)(current):
1302  * Request: [ ipfw_obj_header ipfw_range_tlv ]
1303  *
1304  * Returns 0 on success.
1305  */
1306 static int
1307 move_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1308     struct sockopt_data *sd)
1309 {
1310 	ipfw_range_header *rh;
1311 
1312 	if (sd->valsize != sizeof(*rh))
1313 		return (EINVAL);
1314 
1315 	rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
1316 
1317 	if (check_range_tlv(&rh->range) != 0)
1318 		return (EINVAL);
1319 
1320 	return (move_range(chain, &rh->range));
1321 }
1322 
1323 /*
1324  * Clear rule accounting data matching specified parameters
1325  * Data layout (v0)(current):
1326  * Request: [ ipfw_obj_header ipfw_range_tlv ]
1327  * Reply: [ ipfw_obj_header ipfw_range_tlv ]
1328  *
1329  * Saves number of cleared rules in ipfw_range_tlv->new_set.
1330  *
1331  * Returns 0 on success.
1332  */
1333 static int
1334 clear_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1335     struct sockopt_data *sd)
1336 {
1337 	ipfw_range_header *rh;
1338 	int log_only, num;
1339 	char *msg;
1340 
1341 	if (sd->valsize != sizeof(*rh))
1342 		return (EINVAL);
1343 
1344 	rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
1345 
1346 	if (check_range_tlv(&rh->range) != 0)
1347 		return (EINVAL);
1348 
1349 	log_only = (op3->opcode == IP_FW_XRESETLOG);
1350 
1351 	num = clear_range(chain, &rh->range, log_only);
1352 
1353 	if (rh->range.flags & IPFW_RCFLAG_ALL)
1354 		msg = log_only ? "All logging counts reset" :
1355 		    "Accounting cleared";
1356 	else
1357 		msg = log_only ? "logging count reset" : "cleared";
1358 
1359 	if (V_fw_verbose) {
1360 		int lev = LOG_SECURITY | LOG_NOTICE;
1361 		log(lev, "ipfw: %s.\n", msg);
1362 	}
1363 
1364 	/* Save number of rules cleared */
1365 	rh->range.new_set = num;
1366 	return (0);
1367 }
1368 
1369 static void
1370 enable_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt)
1371 {
1372 	uint32_t v_set;
1373 
1374 	IPFW_UH_WLOCK_ASSERT(chain);
1375 
1376 	/* Change enabled/disabled sets mask */
1377 	v_set = (V_set_disable | rt->set) & ~rt->new_set;
1378 	v_set &= ~(1 << RESVD_SET); /* set RESVD_SET always enabled */
1379 	IPFW_WLOCK(chain);
1380 	V_set_disable = v_set;
1381 	IPFW_WUNLOCK(chain);
1382 }
1383 
1384 static int
1385 swap_sets(struct ip_fw_chain *chain, ipfw_range_tlv *rt, int mv)
1386 {
1387 	struct opcode_obj_rewrite *rw;
1388 	struct ip_fw *rule;
1389 	int i;
1390 
1391 	IPFW_UH_WLOCK_ASSERT(chain);
1392 
1393 	if (rt->set == rt->new_set) /* nothing to do */
1394 		return (0);
1395 
1396 	if (mv != 0) {
1397 		/*
1398 		 * Berfore moving the rules we need to check that
1399 		 * there aren't any conflicting named objects.
1400 		 */
1401 		for (rw = ctl3_rewriters;
1402 		    rw < ctl3_rewriters + ctl3_rsize; rw++) {
1403 			if (rw->manage_sets == NULL)
1404 				continue;
1405 			i = rw->manage_sets(chain, (uint8_t)rt->set,
1406 			    (uint8_t)rt->new_set, TEST_ALL);
1407 			if (i != 0)
1408 				return (EEXIST);
1409 		}
1410 	}
1411 	/* Swap or move two sets */
1412 	for (i = 0; i < chain->n_rules - 1; i++) {
1413 		rule = chain->map[i];
1414 		if (rule->set == (uint8_t)rt->set)
1415 			rule->set = (uint8_t)rt->new_set;
1416 		else if (rule->set == (uint8_t)rt->new_set && mv == 0)
1417 			rule->set = (uint8_t)rt->set;
1418 	}
1419 	for (rw = ctl3_rewriters; rw < ctl3_rewriters + ctl3_rsize; rw++) {
1420 		if (rw->manage_sets == NULL)
1421 			continue;
1422 		rw->manage_sets(chain, (uint8_t)rt->set,
1423 		    (uint8_t)rt->new_set, mv != 0 ? MOVE_ALL: SWAP_ALL);
1424 	}
1425 	return (0);
1426 }
1427 
1428 /*
1429  * Swaps or moves set
1430  * Data layout (v0)(current):
1431  * Request: [ ipfw_obj_header ipfw_range_tlv ]
1432  *
1433  * Returns 0 on success.
1434  */
1435 static int
1436 manage_sets(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
1437     struct sockopt_data *sd)
1438 {
1439 	ipfw_range_header *rh;
1440 	int ret;
1441 
1442 	if (sd->valsize != sizeof(*rh))
1443 		return (EINVAL);
1444 
1445 	rh = (ipfw_range_header *)ipfw_get_sopt_space(sd, sd->valsize);
1446 
1447 	if (rh->range.head.length != sizeof(ipfw_range_tlv))
1448 		return (1);
1449 	/* enable_sets() expects bitmasks. */
1450 	if (op3->opcode != IP_FW_SET_ENABLE &&
1451 	    (rh->range.set >= IPFW_MAX_SETS ||
1452 	    rh->range.new_set >= IPFW_MAX_SETS))
1453 		return (EINVAL);
1454 
1455 	ret = 0;
1456 	IPFW_UH_WLOCK(chain);
1457 	switch (op3->opcode) {
1458 	case IP_FW_SET_SWAP:
1459 	case IP_FW_SET_MOVE:
1460 		ret = swap_sets(chain, &rh->range,
1461 		    op3->opcode == IP_FW_SET_MOVE);
1462 		break;
1463 	case IP_FW_SET_ENABLE:
1464 		enable_sets(chain, &rh->range);
1465 		break;
1466 	}
1467 	IPFW_UH_WUNLOCK(chain);
1468 
1469 	return (ret);
1470 }
1471 
1472 /**
1473  * Remove all rules with given number, or do set manipulation.
1474  * Assumes chain != NULL && *chain != NULL.
1475  *
1476  * The argument is an uint32_t. The low 16 bit are the rule or set number;
1477  * the next 8 bits are the new set; the top 8 bits indicate the command:
1478  *
1479  *	0	delete rules numbered "rulenum"
1480  *	1	delete rules in set "rulenum"
1481  *	2	move rules "rulenum" to set "new_set"
1482  *	3	move rules from set "rulenum" to set "new_set"
1483  *	4	swap sets "rulenum" and "new_set"
1484  *	5	delete rules "rulenum" and set "new_set"
1485  */
1486 static int
1487 del_entry(struct ip_fw_chain *chain, uint32_t arg)
1488 {
1489 	uint32_t num;	/* rule number or old_set */
1490 	uint8_t cmd, new_set;
1491 	int do_del, ndel;
1492 	int error = 0;
1493 	ipfw_range_tlv rt;
1494 
1495 	num = arg & 0xffff;
1496 	cmd = (arg >> 24) & 0xff;
1497 	new_set = (arg >> 16) & 0xff;
1498 
1499 	if (cmd > 5 || new_set > RESVD_SET)
1500 		return EINVAL;
1501 	if (cmd == 0 || cmd == 2 || cmd == 5) {
1502 		if (num >= IPFW_DEFAULT_RULE)
1503 			return EINVAL;
1504 	} else {
1505 		if (num > RESVD_SET)	/* old_set */
1506 			return EINVAL;
1507 	}
1508 
1509 	/* Convert old requests into new representation */
1510 	memset(&rt, 0, sizeof(rt));
1511 	rt.start_rule = num;
1512 	rt.end_rule = num;
1513 	rt.set = num;
1514 	rt.new_set = new_set;
1515 	do_del = 0;
1516 
1517 	switch (cmd) {
1518 	case 0: /* delete rules numbered "rulenum" */
1519 		if (num == 0)
1520 			rt.flags |= IPFW_RCFLAG_ALL;
1521 		else
1522 			rt.flags |= IPFW_RCFLAG_RANGE;
1523 		do_del = 1;
1524 		break;
1525 	case 1: /* delete rules in set "rulenum" */
1526 		rt.flags |= IPFW_RCFLAG_SET;
1527 		do_del = 1;
1528 		break;
1529 	case 5: /* delete rules "rulenum" and set "new_set" */
1530 		rt.flags |= IPFW_RCFLAG_RANGE | IPFW_RCFLAG_SET;
1531 		rt.set = new_set;
1532 		rt.new_set = 0;
1533 		do_del = 1;
1534 		break;
1535 	case 2: /* move rules "rulenum" to set "new_set" */
1536 		rt.flags |= IPFW_RCFLAG_RANGE;
1537 		break;
1538 	case 3: /* move rules from set "rulenum" to set "new_set" */
1539 		IPFW_UH_WLOCK(chain);
1540 		error = swap_sets(chain, &rt, 1);
1541 		IPFW_UH_WUNLOCK(chain);
1542 		return (error);
1543 	case 4: /* swap sets "rulenum" and "new_set" */
1544 		IPFW_UH_WLOCK(chain);
1545 		error = swap_sets(chain, &rt, 0);
1546 		IPFW_UH_WUNLOCK(chain);
1547 		return (error);
1548 	default:
1549 		return (ENOTSUP);
1550 	}
1551 
1552 	if (do_del != 0) {
1553 		if ((error = delete_range(chain, &rt, &ndel)) != 0)
1554 			return (error);
1555 
1556 		if (ndel == 0 && (cmd != 1 && num != 0))
1557 			return (EINVAL);
1558 
1559 		return (0);
1560 	}
1561 
1562 	return (move_range(chain, &rt));
1563 }
1564 
1565 /**
1566  * Reset some or all counters on firewall rules.
1567  * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
1568  * the next 8 bits are the set number, the top 8 bits are the command:
1569  *	0	work with rules from all set's;
1570  *	1	work with rules only from specified set.
1571  * Specified rule number is zero if we want to clear all entries.
1572  * log_only is 1 if we only want to reset logs, zero otherwise.
1573  */
1574 static int
1575 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
1576 {
1577 	struct ip_fw *rule;
1578 	char *msg;
1579 	int i;
1580 
1581 	uint16_t rulenum = arg & 0xffff;
1582 	uint8_t set = (arg >> 16) & 0xff;
1583 	uint8_t cmd = (arg >> 24) & 0xff;
1584 
1585 	if (cmd > 1)
1586 		return (EINVAL);
1587 	if (cmd == 1 && set > RESVD_SET)
1588 		return (EINVAL);
1589 
1590 	IPFW_UH_RLOCK(chain);
1591 	if (rulenum == 0) {
1592 		V_norule_counter = 0;
1593 		for (i = 0; i < chain->n_rules; i++) {
1594 			rule = chain->map[i];
1595 			/* Skip rules not in our set. */
1596 			if (cmd == 1 && rule->set != set)
1597 				continue;
1598 			clear_counters(rule, log_only);
1599 		}
1600 		msg = log_only ? "All logging counts reset" :
1601 		    "Accounting cleared";
1602 	} else {
1603 		int cleared = 0;
1604 		for (i = 0; i < chain->n_rules; i++) {
1605 			rule = chain->map[i];
1606 			if (rule->rulenum == rulenum) {
1607 				if (cmd == 0 || rule->set == set)
1608 					clear_counters(rule, log_only);
1609 				cleared = 1;
1610 			}
1611 			if (rule->rulenum > rulenum)
1612 				break;
1613 		}
1614 		if (!cleared) {	/* we did not find any matching rules */
1615 			IPFW_UH_RUNLOCK(chain);
1616 			return (EINVAL);
1617 		}
1618 		msg = log_only ? "logging count reset" : "cleared";
1619 	}
1620 	IPFW_UH_RUNLOCK(chain);
1621 
1622 	if (V_fw_verbose) {
1623 		int lev = LOG_SECURITY | LOG_NOTICE;
1624 
1625 		if (rulenum)
1626 			log(lev, "ipfw: Entry %d %s.\n", rulenum, msg);
1627 		else
1628 			log(lev, "ipfw: %s.\n", msg);
1629 	}
1630 	return (0);
1631 }
1632 
1633 
1634 /*
1635  * Check rule head in FreeBSD11 format
1636  *
1637  */
1638 static int
1639 check_ipfw_rule1(struct ip_fw_rule *rule, int size,
1640     struct rule_check_info *ci)
1641 {
1642 	int l;
1643 
1644 	if (size < sizeof(*rule)) {
1645 		printf("ipfw: rule too short\n");
1646 		return (EINVAL);
1647 	}
1648 
1649 	/* Check for valid cmd_len */
1650 	l = roundup2(RULESIZE(rule), sizeof(uint64_t));
1651 	if (l != size) {
1652 		printf("ipfw: size mismatch (have %d want %d)\n", size, l);
1653 		return (EINVAL);
1654 	}
1655 	if (rule->act_ofs >= rule->cmd_len) {
1656 		printf("ipfw: bogus action offset (%u > %u)\n",
1657 		    rule->act_ofs, rule->cmd_len - 1);
1658 		return (EINVAL);
1659 	}
1660 
1661 	if (rule->rulenum > IPFW_DEFAULT_RULE - 1)
1662 		return (EINVAL);
1663 
1664 	return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci));
1665 }
1666 
1667 /*
1668  * Check rule head in FreeBSD8 format
1669  *
1670  */
1671 static int
1672 check_ipfw_rule0(struct ip_fw_rule0 *rule, int size,
1673     struct rule_check_info *ci)
1674 {
1675 	int l;
1676 
1677 	if (size < sizeof(*rule)) {
1678 		printf("ipfw: rule too short\n");
1679 		return (EINVAL);
1680 	}
1681 
1682 	/* Check for valid cmd_len */
1683 	l = sizeof(*rule) + rule->cmd_len * 4 - 4;
1684 	if (l != size) {
1685 		printf("ipfw: size mismatch (have %d want %d)\n", size, l);
1686 		return (EINVAL);
1687 	}
1688 	if (rule->act_ofs >= rule->cmd_len) {
1689 		printf("ipfw: bogus action offset (%u > %u)\n",
1690 		    rule->act_ofs, rule->cmd_len - 1);
1691 		return (EINVAL);
1692 	}
1693 
1694 	if (rule->rulenum > IPFW_DEFAULT_RULE - 1)
1695 		return (EINVAL);
1696 
1697 	return (check_ipfw_rule_body(rule->cmd, rule->cmd_len, ci));
1698 }
1699 
1700 static int
1701 check_ipfw_rule_body(ipfw_insn *cmd, int cmd_len, struct rule_check_info *ci)
1702 {
1703 	int cmdlen, l;
1704 	int have_action;
1705 
1706 	have_action = 0;
1707 
1708 	/*
1709 	 * Now go for the individual checks. Very simple ones, basically only
1710 	 * instruction sizes.
1711 	 */
1712 	for (l = cmd_len; l > 0 ; l -= cmdlen, cmd += cmdlen) {
1713 		cmdlen = F_LEN(cmd);
1714 		if (cmdlen > l) {
1715 			printf("ipfw: opcode %d size truncated\n",
1716 			    cmd->opcode);
1717 			return EINVAL;
1718 		}
1719 		switch (cmd->opcode) {
1720 		case O_PROBE_STATE:
1721 		case O_KEEP_STATE:
1722 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1723 				goto bad_size;
1724 			ci->object_opcodes++;
1725 			break;
1726 		case O_PROTO:
1727 		case O_IP_SRC_ME:
1728 		case O_IP_DST_ME:
1729 		case O_LAYER2:
1730 		case O_IN:
1731 		case O_FRAG:
1732 		case O_DIVERTED:
1733 		case O_IPOPT:
1734 		case O_IPTOS:
1735 		case O_IPPRECEDENCE:
1736 		case O_IPVER:
1737 		case O_SOCKARG:
1738 		case O_TCPFLAGS:
1739 		case O_TCPOPTS:
1740 		case O_ESTAB:
1741 		case O_VERREVPATH:
1742 		case O_VERSRCREACH:
1743 		case O_ANTISPOOF:
1744 		case O_IPSEC:
1745 #ifdef INET6
1746 		case O_IP6_SRC_ME:
1747 		case O_IP6_DST_ME:
1748 		case O_EXT_HDR:
1749 		case O_IP6:
1750 #endif
1751 		case O_IP4:
1752 		case O_TAG:
1753 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1754 				goto bad_size;
1755 			break;
1756 
1757 		case O_EXTERNAL_ACTION:
1758 			if (cmd->arg1 == 0 ||
1759 			    cmdlen != F_INSN_SIZE(ipfw_insn)) {
1760 				printf("ipfw: invalid external "
1761 				    "action opcode\n");
1762 				return (EINVAL);
1763 			}
1764 			ci->object_opcodes++;
1765 			/*
1766 			 * Do we have O_EXTERNAL_INSTANCE or O_EXTERNAL_DATA
1767 			 * opcode?
1768 			 */
1769 			if (l != cmdlen) {
1770 				l -= cmdlen;
1771 				cmd += cmdlen;
1772 				cmdlen = F_LEN(cmd);
1773 				if (cmd->opcode == O_EXTERNAL_DATA)
1774 					goto check_action;
1775 				if (cmd->opcode != O_EXTERNAL_INSTANCE) {
1776 					printf("ipfw: invalid opcode "
1777 					    "next to external action %u\n",
1778 					    cmd->opcode);
1779 					return (EINVAL);
1780 				}
1781 				if (cmd->arg1 == 0 ||
1782 				    cmdlen != F_INSN_SIZE(ipfw_insn)) {
1783 					printf("ipfw: invalid external "
1784 					    "action instance opcode\n");
1785 					return (EINVAL);
1786 				}
1787 				ci->object_opcodes++;
1788 			}
1789 			goto check_action;
1790 
1791 		case O_FIB:
1792 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1793 				goto bad_size;
1794 			if (cmd->arg1 >= rt_numfibs) {
1795 				printf("ipfw: invalid fib number %d\n",
1796 					cmd->arg1);
1797 				return EINVAL;
1798 			}
1799 			break;
1800 
1801 		case O_SETFIB:
1802 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1803 				goto bad_size;
1804 			if ((cmd->arg1 != IP_FW_TARG) &&
1805 			    ((cmd->arg1 & 0x7FFF) >= rt_numfibs)) {
1806 				printf("ipfw: invalid fib number %d\n",
1807 					cmd->arg1 & 0x7FFF);
1808 				return EINVAL;
1809 			}
1810 			goto check_action;
1811 
1812 		case O_UID:
1813 		case O_GID:
1814 		case O_JAIL:
1815 		case O_IP_SRC:
1816 		case O_IP_DST:
1817 		case O_TCPSEQ:
1818 		case O_TCPACK:
1819 		case O_PROB:
1820 		case O_ICMPTYPE:
1821 			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
1822 				goto bad_size;
1823 			break;
1824 
1825 		case O_LIMIT:
1826 			if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
1827 				goto bad_size;
1828 			ci->object_opcodes++;
1829 			break;
1830 
1831 		case O_LOG:
1832 			if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
1833 				goto bad_size;
1834 
1835 			((ipfw_insn_log *)cmd)->log_left =
1836 			    ((ipfw_insn_log *)cmd)->max_log;
1837 
1838 			break;
1839 
1840 		case O_IP_SRC_MASK:
1841 		case O_IP_DST_MASK:
1842 			/* only odd command lengths */
1843 			if ((cmdlen & 1) == 0)
1844 				goto bad_size;
1845 			break;
1846 
1847 		case O_IP_SRC_SET:
1848 		case O_IP_DST_SET:
1849 			if (cmd->arg1 == 0 || cmd->arg1 > 256) {
1850 				printf("ipfw: invalid set size %d\n",
1851 					cmd->arg1);
1852 				return EINVAL;
1853 			}
1854 			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
1855 			    (cmd->arg1+31)/32 )
1856 				goto bad_size;
1857 			break;
1858 
1859 		case O_IP_SRC_LOOKUP:
1860 			if (cmdlen > F_INSN_SIZE(ipfw_insn_u32))
1861 				goto bad_size;
1862 		case O_IP_DST_LOOKUP:
1863 			if (cmd->arg1 >= V_fw_tables_max) {
1864 				printf("ipfw: invalid table number %d\n",
1865 				    cmd->arg1);
1866 				return (EINVAL);
1867 			}
1868 			if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
1869 			    cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1 &&
1870 			    cmdlen != F_INSN_SIZE(ipfw_insn_u32))
1871 				goto bad_size;
1872 			ci->object_opcodes++;
1873 			break;
1874 		case O_IP_FLOW_LOOKUP:
1875 			if (cmd->arg1 >= V_fw_tables_max) {
1876 				printf("ipfw: invalid table number %d\n",
1877 				    cmd->arg1);
1878 				return (EINVAL);
1879 			}
1880 			if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
1881 			    cmdlen != F_INSN_SIZE(ipfw_insn_u32))
1882 				goto bad_size;
1883 			ci->object_opcodes++;
1884 			break;
1885 		case O_MACADDR2:
1886 			if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
1887 				goto bad_size;
1888 			break;
1889 
1890 		case O_NOP:
1891 		case O_IPID:
1892 		case O_IPTTL:
1893 		case O_IPLEN:
1894 		case O_TCPDATALEN:
1895 		case O_TCPWIN:
1896 		case O_TAGGED:
1897 			if (cmdlen < 1 || cmdlen > 31)
1898 				goto bad_size;
1899 			break;
1900 
1901 		case O_DSCP:
1902 			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1)
1903 				goto bad_size;
1904 			break;
1905 
1906 		case O_MAC_TYPE:
1907 		case O_IP_SRCPORT:
1908 		case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
1909 			if (cmdlen < 2 || cmdlen > 31)
1910 				goto bad_size;
1911 			break;
1912 
1913 		case O_RECV:
1914 		case O_XMIT:
1915 		case O_VIA:
1916 			if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
1917 				goto bad_size;
1918 			ci->object_opcodes++;
1919 			break;
1920 
1921 		case O_ALTQ:
1922 			if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
1923 				goto bad_size;
1924 			break;
1925 
1926 		case O_PIPE:
1927 		case O_QUEUE:
1928 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1929 				goto bad_size;
1930 			goto check_action;
1931 
1932 		case O_FORWARD_IP:
1933 			if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
1934 				goto bad_size;
1935 			goto check_action;
1936 #ifdef INET6
1937 		case O_FORWARD_IP6:
1938 			if (cmdlen != F_INSN_SIZE(ipfw_insn_sa6))
1939 				goto bad_size;
1940 			goto check_action;
1941 #endif /* INET6 */
1942 
1943 		case O_DIVERT:
1944 		case O_TEE:
1945 			if (ip_divert_ptr == NULL)
1946 				return EINVAL;
1947 			else
1948 				goto check_size;
1949 		case O_NETGRAPH:
1950 		case O_NGTEE:
1951 			if (ng_ipfw_input_p == NULL)
1952 				return EINVAL;
1953 			else
1954 				goto check_size;
1955 		case O_NAT:
1956 			if (!IPFW_NAT_LOADED)
1957 				return EINVAL;
1958 			if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
1959  				goto bad_size;
1960  			goto check_action;
1961 		case O_CHECK_STATE:
1962 			ci->object_opcodes++;
1963 			/* FALLTHROUGH */
1964 		case O_FORWARD_MAC: /* XXX not implemented yet */
1965 		case O_COUNT:
1966 		case O_ACCEPT:
1967 		case O_DENY:
1968 		case O_REJECT:
1969 		case O_SETDSCP:
1970 #ifdef INET6
1971 		case O_UNREACH6:
1972 #endif
1973 		case O_SKIPTO:
1974 		case O_REASS:
1975 		case O_CALLRETURN:
1976 check_size:
1977 			if (cmdlen != F_INSN_SIZE(ipfw_insn))
1978 				goto bad_size;
1979 check_action:
1980 			if (have_action) {
1981 				printf("ipfw: opcode %d, multiple actions"
1982 					" not allowed\n",
1983 					cmd->opcode);
1984 				return (EINVAL);
1985 			}
1986 			have_action = 1;
1987 			if (l != cmdlen) {
1988 				printf("ipfw: opcode %d, action must be"
1989 					" last opcode\n",
1990 					cmd->opcode);
1991 				return (EINVAL);
1992 			}
1993 			break;
1994 #ifdef INET6
1995 		case O_IP6_SRC:
1996 		case O_IP6_DST:
1997 			if (cmdlen != F_INSN_SIZE(struct in6_addr) +
1998 			    F_INSN_SIZE(ipfw_insn))
1999 				goto bad_size;
2000 			break;
2001 
2002 		case O_FLOW6ID:
2003 			if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
2004 			    ((ipfw_insn_u32 *)cmd)->o.arg1)
2005 				goto bad_size;
2006 			break;
2007 
2008 		case O_IP6_SRC_MASK:
2009 		case O_IP6_DST_MASK:
2010 			if ( !(cmdlen & 1) || cmdlen > 127)
2011 				goto bad_size;
2012 			break;
2013 		case O_ICMP6TYPE:
2014 			if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
2015 				goto bad_size;
2016 			break;
2017 #endif
2018 
2019 		default:
2020 			switch (cmd->opcode) {
2021 #ifndef INET6
2022 			case O_IP6_SRC_ME:
2023 			case O_IP6_DST_ME:
2024 			case O_EXT_HDR:
2025 			case O_IP6:
2026 			case O_UNREACH6:
2027 			case O_IP6_SRC:
2028 			case O_IP6_DST:
2029 			case O_FLOW6ID:
2030 			case O_IP6_SRC_MASK:
2031 			case O_IP6_DST_MASK:
2032 			case O_ICMP6TYPE:
2033 				printf("ipfw: no IPv6 support in kernel\n");
2034 				return (EPROTONOSUPPORT);
2035 #endif
2036 			default:
2037 				printf("ipfw: opcode %d, unknown opcode\n",
2038 					cmd->opcode);
2039 				return (EINVAL);
2040 			}
2041 		}
2042 	}
2043 	if (have_action == 0) {
2044 		printf("ipfw: missing action\n");
2045 		return (EINVAL);
2046 	}
2047 	return 0;
2048 
2049 bad_size:
2050 	printf("ipfw: opcode %d size %d wrong\n",
2051 		cmd->opcode, cmdlen);
2052 	return (EINVAL);
2053 }
2054 
2055 
2056 /*
2057  * Translation of requests for compatibility with FreeBSD 7.2/8.
2058  * a static variable tells us if we have an old client from userland,
2059  * and if necessary we translate requests and responses between the
2060  * two formats.
2061  */
2062 static int is7 = 0;
2063 
2064 struct ip_fw7 {
2065 	struct ip_fw7	*next;		/* linked list of rules     */
2066 	struct ip_fw7	*next_rule;	/* ptr to next [skipto] rule    */
2067 	/* 'next_rule' is used to pass up 'set_disable' status      */
2068 
2069 	uint16_t	act_ofs;	/* offset of action in 32-bit units */
2070 	uint16_t	cmd_len;	/* # of 32-bit words in cmd */
2071 	uint16_t	rulenum;	/* rule number          */
2072 	uint8_t		set;		/* rule set (0..31)     */
2073 	// #define RESVD_SET   31  /* set for default and persistent rules */
2074 	uint8_t		_pad;		/* padding          */
2075 	// uint32_t        id;             /* rule id, only in v.8 */
2076 	/* These fields are present in all rules.           */
2077 	uint64_t	pcnt;		/* Packet counter       */
2078 	uint64_t	bcnt;		/* Byte counter         */
2079 	uint32_t	timestamp;	/* tv_sec of last match     */
2080 
2081 	ipfw_insn	cmd[1];		/* storage for commands     */
2082 };
2083 
2084 static int convert_rule_to_7(struct ip_fw_rule0 *rule);
2085 static int convert_rule_to_8(struct ip_fw_rule0 *rule);
2086 
2087 #ifndef RULESIZE7
2088 #define RULESIZE7(rule)  (sizeof(struct ip_fw7) + \
2089 	((struct ip_fw7 *)(rule))->cmd_len * 4 - 4)
2090 #endif
2091 
2092 
2093 /*
2094  * Copy the static and dynamic rules to the supplied buffer
2095  * and return the amount of space actually used.
2096  * Must be run under IPFW_UH_RLOCK
2097  */
2098 static size_t
2099 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
2100 {
2101 	char *bp = buf;
2102 	char *ep = bp + space;
2103 	struct ip_fw *rule;
2104 	struct ip_fw_rule0 *dst;
2105 	struct timeval boottime;
2106 	int error, i, l, warnflag;
2107 	time_t	boot_seconds;
2108 
2109 	warnflag = 0;
2110 
2111 	getboottime(&boottime);
2112         boot_seconds = boottime.tv_sec;
2113 	for (i = 0; i < chain->n_rules; i++) {
2114 		rule = chain->map[i];
2115 
2116 		if (is7) {
2117 		    /* Convert rule to FreeBSd 7.2 format */
2118 		    l = RULESIZE7(rule);
2119 		    if (bp + l + sizeof(uint32_t) <= ep) {
2120 			bcopy(rule, bp, l + sizeof(uint32_t));
2121 			error = set_legacy_obj_kidx(chain,
2122 			    (struct ip_fw_rule0 *)bp);
2123 			if (error != 0)
2124 				return (0);
2125 			error = convert_rule_to_7((struct ip_fw_rule0 *) bp);
2126 			if (error)
2127 				return 0; /*XXX correct? */
2128 			/*
2129 			 * XXX HACK. Store the disable mask in the "next"
2130 			 * pointer in a wild attempt to keep the ABI the same.
2131 			 * Why do we do this on EVERY rule?
2132 			 */
2133 			bcopy(&V_set_disable,
2134 				&(((struct ip_fw7 *)bp)->next_rule),
2135 				sizeof(V_set_disable));
2136 			if (((struct ip_fw7 *)bp)->timestamp)
2137 			    ((struct ip_fw7 *)bp)->timestamp += boot_seconds;
2138 			bp += l;
2139 		    }
2140 		    continue; /* go to next rule */
2141 		}
2142 
2143 		l = RULEUSIZE0(rule);
2144 		if (bp + l > ep) { /* should not happen */
2145 			printf("overflow dumping static rules\n");
2146 			break;
2147 		}
2148 		dst = (struct ip_fw_rule0 *)bp;
2149 		export_rule0(rule, dst, l);
2150 		error = set_legacy_obj_kidx(chain, dst);
2151 
2152 		/*
2153 		 * XXX HACK. Store the disable mask in the "next"
2154 		 * pointer in a wild attempt to keep the ABI the same.
2155 		 * Why do we do this on EVERY rule?
2156 		 *
2157 		 * XXX: "ipfw set show" (ab)uses IP_FW_GET to read disabled mask
2158 		 * so we need to fail _after_ saving at least one mask.
2159 		 */
2160 		bcopy(&V_set_disable, &dst->next_rule, sizeof(V_set_disable));
2161 		if (dst->timestamp)
2162 			dst->timestamp += boot_seconds;
2163 		bp += l;
2164 
2165 		if (error != 0) {
2166 			if (error == 2) {
2167 				/* Non-fatal table rewrite error. */
2168 				warnflag = 1;
2169 				continue;
2170 			}
2171 			printf("Stop on rule %d. Fail to convert table\n",
2172 			    rule->rulenum);
2173 			break;
2174 		}
2175 	}
2176 	if (warnflag != 0)
2177 		printf("ipfw: process %s is using legacy interfaces,"
2178 		    " consider rebuilding\n", "");
2179 	ipfw_get_dynamic(chain, &bp, ep); /* protected by the dynamic lock */
2180 	return (bp - (char *)buf);
2181 }
2182 
2183 
2184 struct dump_args {
2185 	uint32_t	b;	/* start rule */
2186 	uint32_t	e;	/* end rule */
2187 	uint32_t	rcount;	/* number of rules */
2188 	uint32_t	rsize;	/* rules size */
2189 	uint32_t	tcount;	/* number of tables */
2190 	int		rcounters;	/* counters */
2191 };
2192 
2193 void
2194 ipfw_export_obj_ntlv(struct named_object *no, ipfw_obj_ntlv *ntlv)
2195 {
2196 
2197 	ntlv->head.type = no->etlv;
2198 	ntlv->head.length = sizeof(*ntlv);
2199 	ntlv->idx = no->kidx;
2200 	strlcpy(ntlv->name, no->name, sizeof(ntlv->name));
2201 }
2202 
2203 /*
2204  * Export named object info in instance @ni, identified by @kidx
2205  * to ipfw_obj_ntlv. TLV is allocated from @sd space.
2206  *
2207  * Returns 0 on success.
2208  */
2209 static int
2210 export_objhash_ntlv(struct namedobj_instance *ni, uint16_t kidx,
2211     struct sockopt_data *sd)
2212 {
2213 	struct named_object *no;
2214 	ipfw_obj_ntlv *ntlv;
2215 
2216 	no = ipfw_objhash_lookup_kidx(ni, kidx);
2217 	KASSERT(no != NULL, ("invalid object kernel index passed"));
2218 
2219 	ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv));
2220 	if (ntlv == NULL)
2221 		return (ENOMEM);
2222 
2223 	ipfw_export_obj_ntlv(no, ntlv);
2224 	return (0);
2225 }
2226 
2227 /*
2228  * Dumps static rules with table TLVs in buffer @sd.
2229  *
2230  * Returns 0 on success.
2231  */
2232 static int
2233 dump_static_rules(struct ip_fw_chain *chain, struct dump_args *da,
2234     uint32_t *bmask, struct sockopt_data *sd)
2235 {
2236 	int error;
2237 	int i, l;
2238 	uint32_t tcount;
2239 	ipfw_obj_ctlv *ctlv;
2240 	struct ip_fw *krule;
2241 	struct namedobj_instance *ni;
2242 	caddr_t dst;
2243 
2244 	/* Dump table names first (if any) */
2245 	if (da->tcount > 0) {
2246 		/* Header first */
2247 		ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv));
2248 		if (ctlv == NULL)
2249 			return (ENOMEM);
2250 		ctlv->head.type = IPFW_TLV_TBLNAME_LIST;
2251 		ctlv->head.length = da->tcount * sizeof(ipfw_obj_ntlv) +
2252 		    sizeof(*ctlv);
2253 		ctlv->count = da->tcount;
2254 		ctlv->objsize = sizeof(ipfw_obj_ntlv);
2255 	}
2256 
2257 	i = 0;
2258 	tcount = da->tcount;
2259 	ni = ipfw_get_table_objhash(chain);
2260 	while (tcount > 0) {
2261 		if ((bmask[i / 32] & (1 << (i % 32))) == 0) {
2262 			i++;
2263 			continue;
2264 		}
2265 
2266 		/* Jump to shared named object bitmask */
2267 		if (i >= IPFW_TABLES_MAX) {
2268 			ni = CHAIN_TO_SRV(chain);
2269 			i -= IPFW_TABLES_MAX;
2270 			bmask += IPFW_TABLES_MAX / 32;
2271 		}
2272 
2273 		if ((error = export_objhash_ntlv(ni, i, sd)) != 0)
2274 			return (error);
2275 
2276 		i++;
2277 		tcount--;
2278 	}
2279 
2280 	/* Dump rules */
2281 	ctlv = (ipfw_obj_ctlv *)ipfw_get_sopt_space(sd, sizeof(*ctlv));
2282 	if (ctlv == NULL)
2283 		return (ENOMEM);
2284 	ctlv->head.type = IPFW_TLV_RULE_LIST;
2285 	ctlv->head.length = da->rsize + sizeof(*ctlv);
2286 	ctlv->count = da->rcount;
2287 
2288 	for (i = da->b; i < da->e; i++) {
2289 		krule = chain->map[i];
2290 
2291 		l = RULEUSIZE1(krule) + sizeof(ipfw_obj_tlv);
2292 		if (da->rcounters != 0)
2293 			l += sizeof(struct ip_fw_bcounter);
2294 		dst = (caddr_t)ipfw_get_sopt_space(sd, l);
2295 		if (dst == NULL)
2296 			return (ENOMEM);
2297 
2298 		export_rule1(krule, dst, l, da->rcounters);
2299 	}
2300 
2301 	return (0);
2302 }
2303 
2304 /*
2305  * Marks every object index used in @rule with bit in @bmask.
2306  * Used to generate bitmask of referenced tables/objects for given ruleset
2307  * or its part.
2308  *
2309  * Returns number of newly-referenced objects.
2310  */
2311 static int
2312 mark_object_kidx(struct ip_fw_chain *ch, struct ip_fw *rule,
2313     uint32_t *bmask)
2314 {
2315 	struct opcode_obj_rewrite *rw;
2316 	ipfw_insn *cmd;
2317 	int bidx, cmdlen, l, count;
2318 	uint16_t kidx;
2319 	uint8_t subtype;
2320 
2321 	l = rule->cmd_len;
2322 	cmd = rule->cmd;
2323 	cmdlen = 0;
2324 	count = 0;
2325 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
2326 		cmdlen = F_LEN(cmd);
2327 
2328 		rw = find_op_rw(cmd, &kidx, &subtype);
2329 		if (rw == NULL)
2330 			continue;
2331 
2332 		bidx = kidx / 32;
2333 		/*
2334 		 * Maintain separate bitmasks for table and
2335 		 * non-table objects.
2336 		 */
2337 		if (rw->etlv != IPFW_TLV_TBL_NAME)
2338 			bidx += IPFW_TABLES_MAX / 32;
2339 
2340 		if ((bmask[bidx] & (1 << (kidx % 32))) == 0)
2341 			count++;
2342 
2343 		bmask[bidx] |= 1 << (kidx % 32);
2344 	}
2345 
2346 	return (count);
2347 }
2348 
2349 /*
2350  * Dumps requested objects data
2351  * Data layout (version 0)(current):
2352  * Request: [ ipfw_cfg_lheader ] + IPFW_CFG_GET_* flags
2353  *   size = ipfw_cfg_lheader.size
2354  * Reply: [ ipfw_cfg_lheader
2355  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional)
2356  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST)
2357  *     ipfw_obj_tlv(IPFW_TLV_RULE_ENT) [ ip_fw_bcounter (optional) ip_fw_rule ]
2358  *   ] (optional)
2359  *   [ ipfw_obj_ctlv(IPFW_TLV_STATE_LIST) ipfw_obj_dyntlv x N ] (optional)
2360  * ]
2361  * * NOTE IPFW_TLV_STATE_LIST has the single valid field: objsize.
2362  * The rest (size, count) are set to zero and needs to be ignored.
2363  *
2364  * Returns 0 on success.
2365  */
2366 static int
2367 dump_config(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
2368     struct sockopt_data *sd)
2369 {
2370 	ipfw_cfg_lheader *hdr;
2371 	struct ip_fw *rule;
2372 	size_t sz, rnum;
2373 	uint32_t hdr_flags;
2374 	int error, i;
2375 	struct dump_args da;
2376 	uint32_t *bmask;
2377 
2378 	hdr = (ipfw_cfg_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr));
2379 	if (hdr == NULL)
2380 		return (EINVAL);
2381 
2382 	error = 0;
2383 	bmask = NULL;
2384 	/* Allocate needed state. Note we allocate 2xspace mask, for table&srv  */
2385 	if (hdr->flags & IPFW_CFG_GET_STATIC)
2386 		bmask = malloc(IPFW_TABLES_MAX / 4, M_TEMP, M_WAITOK | M_ZERO);
2387 
2388 	IPFW_UH_RLOCK(chain);
2389 
2390 	/*
2391 	 * STAGE 1: Determine size/count for objects in range.
2392 	 * Prepare used tables bitmask.
2393 	 */
2394 	sz = sizeof(ipfw_cfg_lheader);
2395 	memset(&da, 0, sizeof(da));
2396 
2397 	da.b = 0;
2398 	da.e = chain->n_rules;
2399 
2400 	if (hdr->end_rule != 0) {
2401 		/* Handle custom range */
2402 		if ((rnum = hdr->start_rule) > IPFW_DEFAULT_RULE)
2403 			rnum = IPFW_DEFAULT_RULE;
2404 		da.b = ipfw_find_rule(chain, rnum, 0);
2405 		rnum = (hdr->end_rule < IPFW_DEFAULT_RULE) ?
2406 		    hdr->end_rule + 1: IPFW_DEFAULT_RULE;
2407 		da.e = ipfw_find_rule(chain, rnum, UINT32_MAX) + 1;
2408 	}
2409 
2410 	if (hdr->flags & IPFW_CFG_GET_STATIC) {
2411 		for (i = da.b; i < da.e; i++) {
2412 			rule = chain->map[i];
2413 			da.rsize += RULEUSIZE1(rule) + sizeof(ipfw_obj_tlv);
2414 			da.rcount++;
2415 			/* Update bitmask of used objects for given range */
2416 			da.tcount += mark_object_kidx(chain, rule, bmask);
2417 		}
2418 		/* Add counters if requested */
2419 		if (hdr->flags & IPFW_CFG_GET_COUNTERS) {
2420 			da.rsize += sizeof(struct ip_fw_bcounter) * da.rcount;
2421 			da.rcounters = 1;
2422 		}
2423 
2424 		if (da.tcount > 0)
2425 			sz += da.tcount * sizeof(ipfw_obj_ntlv) +
2426 			    sizeof(ipfw_obj_ctlv);
2427 		sz += da.rsize + sizeof(ipfw_obj_ctlv);
2428 	}
2429 
2430 	if (hdr->flags & IPFW_CFG_GET_STATES)
2431 		sz += ipfw_dyn_get_count() * sizeof(ipfw_obj_dyntlv) +
2432 		     sizeof(ipfw_obj_ctlv);
2433 
2434 
2435 	/*
2436 	 * Fill header anyway.
2437 	 * Note we have to save header fields to stable storage
2438 	 * buffer inside @sd can be flushed after dumping rules
2439 	 */
2440 	hdr->size = sz;
2441 	hdr->set_mask = ~V_set_disable;
2442 	hdr_flags = hdr->flags;
2443 	hdr = NULL;
2444 
2445 	if (sd->valsize < sz) {
2446 		error = ENOMEM;
2447 		goto cleanup;
2448 	}
2449 
2450 	/* STAGE2: Store actual data */
2451 	if (hdr_flags & IPFW_CFG_GET_STATIC) {
2452 		error = dump_static_rules(chain, &da, bmask, sd);
2453 		if (error != 0)
2454 			goto cleanup;
2455 	}
2456 
2457 	if (hdr_flags & IPFW_CFG_GET_STATES)
2458 		error = ipfw_dump_states(chain, sd);
2459 
2460 cleanup:
2461 	IPFW_UH_RUNLOCK(chain);
2462 
2463 	if (bmask != NULL)
2464 		free(bmask, M_TEMP);
2465 
2466 	return (error);
2467 }
2468 
2469 int
2470 ipfw_check_object_name_generic(const char *name)
2471 {
2472 	int nsize;
2473 
2474 	nsize = sizeof(((ipfw_obj_ntlv *)0)->name);
2475 	if (strnlen(name, nsize) == nsize)
2476 		return (EINVAL);
2477 	if (name[0] == '\0')
2478 		return (EINVAL);
2479 	return (0);
2480 }
2481 
2482 /*
2483  * Creates non-existent objects referenced by rule.
2484  *
2485  * Return 0 on success.
2486  */
2487 int
2488 create_objects_compat(struct ip_fw_chain *ch, ipfw_insn *cmd,
2489     struct obj_idx *oib, struct obj_idx *pidx, struct tid_info *ti)
2490 {
2491 	struct opcode_obj_rewrite *rw;
2492 	struct obj_idx *p;
2493 	uint16_t kidx;
2494 	int error;
2495 
2496 	/*
2497 	 * Compatibility stuff: do actual creation for non-existing,
2498 	 * but referenced objects.
2499 	 */
2500 	for (p = oib; p < pidx; p++) {
2501 		if (p->kidx != 0)
2502 			continue;
2503 
2504 		ti->uidx = p->uidx;
2505 		ti->type = p->type;
2506 		ti->atype = 0;
2507 
2508 		rw = find_op_rw(cmd + p->off, NULL, NULL);
2509 		KASSERT(rw != NULL, ("Unable to find handler for op %d",
2510 		    (cmd + p->off)->opcode));
2511 
2512 		if (rw->create_object == NULL)
2513 			error = EOPNOTSUPP;
2514 		else
2515 			error = rw->create_object(ch, ti, &kidx);
2516 		if (error == 0) {
2517 			p->kidx = kidx;
2518 			continue;
2519 		}
2520 
2521 		/*
2522 		 * Error happened. We have to rollback everything.
2523 		 * Drop all already acquired references.
2524 		 */
2525 		IPFW_UH_WLOCK(ch);
2526 		unref_oib_objects(ch, cmd, oib, pidx);
2527 		IPFW_UH_WUNLOCK(ch);
2528 
2529 		return (error);
2530 	}
2531 
2532 	return (0);
2533 }
2534 
2535 /*
2536  * Compatibility function for old ipfw(8) binaries.
2537  * Rewrites table/nat kernel indices with userland ones.
2538  * Convert tables matching '/^\d+$/' to their atoi() value.
2539  * Use number 65535 for other tables.
2540  *
2541  * Returns 0 on success.
2542  */
2543 static int
2544 set_legacy_obj_kidx(struct ip_fw_chain *ch, struct ip_fw_rule0 *rule)
2545 {
2546 	struct opcode_obj_rewrite *rw;
2547 	struct named_object *no;
2548 	ipfw_insn *cmd;
2549 	char *end;
2550 	long val;
2551 	int cmdlen, error, l;
2552 	uint16_t kidx, uidx;
2553 	uint8_t subtype;
2554 
2555 	error = 0;
2556 
2557 	l = rule->cmd_len;
2558 	cmd = rule->cmd;
2559 	cmdlen = 0;
2560 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
2561 		cmdlen = F_LEN(cmd);
2562 
2563 		/* Check if is index in given opcode */
2564 		rw = find_op_rw(cmd, &kidx, &subtype);
2565 		if (rw == NULL)
2566 			continue;
2567 
2568 		/* Try to find referenced kernel object */
2569 		no = rw->find_bykidx(ch, kidx);
2570 		if (no == NULL)
2571 			continue;
2572 
2573 		val = strtol(no->name, &end, 10);
2574 		if (*end == '\0' && val < 65535) {
2575 			uidx = val;
2576 		} else {
2577 
2578 			/*
2579 			 * We are called via legacy opcode.
2580 			 * Save error and show table as fake number
2581 			 * not to make ipfw(8) hang.
2582 			 */
2583 			uidx = 65535;
2584 			error = 2;
2585 		}
2586 
2587 		rw->update(cmd, uidx);
2588 	}
2589 
2590 	return (error);
2591 }
2592 
2593 
2594 /*
2595  * Unreferences all already-referenced objects in given @cmd rule,
2596  * using information in @oib.
2597  *
2598  * Used to rollback partially converted rule on error.
2599  */
2600 static void
2601 unref_oib_objects(struct ip_fw_chain *ch, ipfw_insn *cmd, struct obj_idx *oib,
2602     struct obj_idx *end)
2603 {
2604 	struct opcode_obj_rewrite *rw;
2605 	struct named_object *no;
2606 	struct obj_idx *p;
2607 
2608 	IPFW_UH_WLOCK_ASSERT(ch);
2609 
2610 	for (p = oib; p < end; p++) {
2611 		if (p->kidx == 0)
2612 			continue;
2613 
2614 		rw = find_op_rw(cmd + p->off, NULL, NULL);
2615 		KASSERT(rw != NULL, ("Unable to find handler for op %d",
2616 		    (cmd + p->off)->opcode));
2617 
2618 		/* Find & unref by existing idx */
2619 		no = rw->find_bykidx(ch, p->kidx);
2620 		KASSERT(no != NULL, ("Ref'd object %d disappeared", p->kidx));
2621 		no->refcnt--;
2622 	}
2623 }
2624 
2625 /*
2626  * Remove references from every object used in @rule.
2627  * Used at rule removal code.
2628  */
2629 static void
2630 unref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule)
2631 {
2632 	struct opcode_obj_rewrite *rw;
2633 	struct named_object *no;
2634 	ipfw_insn *cmd;
2635 	int cmdlen, l;
2636 	uint16_t kidx;
2637 	uint8_t subtype;
2638 
2639 	IPFW_UH_WLOCK_ASSERT(ch);
2640 
2641 	l = rule->cmd_len;
2642 	cmd = rule->cmd;
2643 	cmdlen = 0;
2644 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
2645 		cmdlen = F_LEN(cmd);
2646 
2647 		rw = find_op_rw(cmd, &kidx, &subtype);
2648 		if (rw == NULL)
2649 			continue;
2650 		no = rw->find_bykidx(ch, kidx);
2651 
2652 		KASSERT(no != NULL, ("object id %d not found", kidx));
2653 		KASSERT(no->subtype == subtype,
2654 		    ("wrong type %d (%d) for object id %d",
2655 		    no->subtype, subtype, kidx));
2656 		KASSERT(no->refcnt > 0, ("refcount for object %d is %d",
2657 		    kidx, no->refcnt));
2658 
2659 		if (no->refcnt == 1 && rw->destroy_object != NULL)
2660 			rw->destroy_object(ch, no);
2661 		else
2662 			no->refcnt--;
2663 	}
2664 }
2665 
2666 
2667 /*
2668  * Find and reference object (if any) stored in instruction @cmd.
2669  *
2670  * Saves object info in @pidx, sets
2671  *  - @unresolved to 1 if object should exists but not found
2672  *
2673  * Returns non-zero value in case of error.
2674  */
2675 static int
2676 ref_opcode_object(struct ip_fw_chain *ch, ipfw_insn *cmd, struct tid_info *ti,
2677     struct obj_idx *pidx, int *unresolved)
2678 {
2679 	struct named_object *no;
2680 	struct opcode_obj_rewrite *rw;
2681 	int error;
2682 
2683 	/* Check if this opcode is candidate for rewrite */
2684 	rw = find_op_rw(cmd, &ti->uidx, &ti->type);
2685 	if (rw == NULL)
2686 		return (0);
2687 
2688 	/* Need to rewrite. Save necessary fields */
2689 	pidx->uidx = ti->uidx;
2690 	pidx->type = ti->type;
2691 
2692 	/* Try to find referenced kernel object */
2693 	error = rw->find_byname(ch, ti, &no);
2694 	if (error != 0)
2695 		return (error);
2696 	if (no == NULL) {
2697 		/*
2698 		 * Report about unresolved object for automaic
2699 		 * creation.
2700 		 */
2701 		*unresolved = 1;
2702 		return (0);
2703 	}
2704 
2705 	/*
2706 	 * Object is already exist.
2707 	 * Its subtype should match with expected value.
2708 	 */
2709 	if (ti->type != no->subtype)
2710 		return (EINVAL);
2711 
2712 	/* Bump refcount and update kidx. */
2713 	no->refcnt++;
2714 	rw->update(cmd, no->kidx);
2715 	return (0);
2716 }
2717 
2718 /*
2719  * Finds and bumps refcount for objects referenced by given @rule.
2720  * Auto-creates non-existing tables.
2721  * Fills in @oib array with userland/kernel indexes.
2722  *
2723  * Returns 0 on success.
2724  */
2725 static int
2726 ref_rule_objects(struct ip_fw_chain *ch, struct ip_fw *rule,
2727     struct rule_check_info *ci, struct obj_idx *oib, struct tid_info *ti)
2728 {
2729 	struct obj_idx *pidx;
2730 	ipfw_insn *cmd;
2731 	int cmdlen, error, l, unresolved;
2732 
2733 	pidx = oib;
2734 	l = rule->cmd_len;
2735 	cmd = rule->cmd;
2736 	cmdlen = 0;
2737 	error = 0;
2738 
2739 	IPFW_UH_WLOCK(ch);
2740 
2741 	/* Increase refcount on each existing referenced table. */
2742 	for ( ;	l > 0 ; l -= cmdlen, cmd += cmdlen) {
2743 		cmdlen = F_LEN(cmd);
2744 		unresolved = 0;
2745 
2746 		error = ref_opcode_object(ch, cmd, ti, pidx, &unresolved);
2747 		if (error != 0)
2748 			break;
2749 		/*
2750 		 * Compatibility stuff for old clients:
2751 		 * prepare to automaitcally create non-existing objects.
2752 		 */
2753 		if (unresolved != 0) {
2754 			pidx->off = rule->cmd_len - l;
2755 			pidx++;
2756 		}
2757 	}
2758 
2759 	if (error != 0) {
2760 		/* Unref everything we have already done */
2761 		unref_oib_objects(ch, rule->cmd, oib, pidx);
2762 		IPFW_UH_WUNLOCK(ch);
2763 		return (error);
2764 	}
2765 	IPFW_UH_WUNLOCK(ch);
2766 
2767 	/* Perform auto-creation for non-existing objects */
2768 	if (pidx != oib)
2769 		error = create_objects_compat(ch, rule->cmd, oib, pidx, ti);
2770 
2771 	/* Calculate real number of dynamic objects */
2772 	ci->object_opcodes = (uint16_t)(pidx - oib);
2773 
2774 	return (error);
2775 }
2776 
2777 /*
2778  * Checks is opcode is referencing table of appropriate type.
2779  * Adds reference count for found table if true.
2780  * Rewrites user-supplied opcode values with kernel ones.
2781  *
2782  * Returns 0 on success and appropriate error code otherwise.
2783  */
2784 static int
2785 rewrite_rule_uidx(struct ip_fw_chain *chain, struct rule_check_info *ci)
2786 {
2787 	int error;
2788 	ipfw_insn *cmd;
2789 	uint8_t type;
2790 	struct obj_idx *p, *pidx_first, *pidx_last;
2791 	struct tid_info ti;
2792 
2793 	/*
2794 	 * Prepare an array for storing opcode indices.
2795 	 * Use stack allocation by default.
2796 	 */
2797 	if (ci->object_opcodes <= (sizeof(ci->obuf)/sizeof(ci->obuf[0]))) {
2798 		/* Stack */
2799 		pidx_first = ci->obuf;
2800 	} else
2801 		pidx_first = malloc(
2802 		    ci->object_opcodes * sizeof(struct obj_idx),
2803 		    M_IPFW, M_WAITOK | M_ZERO);
2804 
2805 	error = 0;
2806 	type = 0;
2807 	memset(&ti, 0, sizeof(ti));
2808 
2809 	/* Use set rule is assigned to. */
2810 	ti.set = ci->krule->set;
2811 	if (ci->ctlv != NULL) {
2812 		ti.tlvs = (void *)(ci->ctlv + 1);
2813 		ti.tlen = ci->ctlv->head.length - sizeof(ipfw_obj_ctlv);
2814 	}
2815 
2816 	/* Reference all used tables and other objects */
2817 	error = ref_rule_objects(chain, ci->krule, ci, pidx_first, &ti);
2818 	if (error != 0)
2819 		goto free;
2820 	/*
2821 	 * Note that ref_rule_objects() might have updated ci->object_opcodes
2822 	 * to reflect actual number of object opcodes.
2823 	 */
2824 
2825 	/* Perform rewrite of remaining opcodes */
2826 	p = pidx_first;
2827 	pidx_last = pidx_first + ci->object_opcodes;
2828 	for (p = pidx_first; p < pidx_last; p++) {
2829 		cmd = ci->krule->cmd + p->off;
2830 		update_opcode_kidx(cmd, p->kidx);
2831 	}
2832 
2833 free:
2834 	if (pidx_first != ci->obuf)
2835 		free(pidx_first, M_IPFW);
2836 
2837 	return (error);
2838 }
2839 
2840 /*
2841  * Adds one or more rules to ipfw @chain.
2842  * Data layout (version 0)(current):
2843  * Request:
2844  * [
2845  *   ip_fw3_opheader
2846  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional *1)
2847  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ] (*2) (*3)
2848  * ]
2849  * Reply:
2850  * [
2851  *   ip_fw3_opheader
2852  *   [ ipfw_obj_ctlv(IPFW_TLV_TBL_LIST) ipfw_obj_ntlv x N ] (optional)
2853  *   [ ipfw_obj_ctlv(IPFW_TLV_RULE_LIST) ip_fw x N ]
2854  * ]
2855  *
2856  * Rules in reply are modified to store their actual ruleset number.
2857  *
2858  * (*1) TLVs inside IPFW_TLV_TBL_LIST needs to be sorted ascending
2859  * according to their idx field and there has to be no duplicates.
2860  * (*2) Numbered rules inside IPFW_TLV_RULE_LIST needs to be sorted ascending.
2861  * (*3) Each ip_fw structure needs to be aligned to u64 boundary.
2862  *
2863  * Returns 0 on success.
2864  */
2865 static int
2866 add_rules(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
2867     struct sockopt_data *sd)
2868 {
2869 	ipfw_obj_ctlv *ctlv, *rtlv, *tstate;
2870 	ipfw_obj_ntlv *ntlv;
2871 	int clen, error, idx;
2872 	uint32_t count, read;
2873 	struct ip_fw_rule *r;
2874 	struct rule_check_info rci, *ci, *cbuf;
2875 	int i, rsize;
2876 
2877 	op3 = (ip_fw3_opheader *)ipfw_get_sopt_space(sd, sd->valsize);
2878 	ctlv = (ipfw_obj_ctlv *)(op3 + 1);
2879 
2880 	read = sizeof(ip_fw3_opheader);
2881 	rtlv = NULL;
2882 	tstate = NULL;
2883 	cbuf = NULL;
2884 	memset(&rci, 0, sizeof(struct rule_check_info));
2885 
2886 	if (read + sizeof(*ctlv) > sd->valsize)
2887 		return (EINVAL);
2888 
2889 	if (ctlv->head.type == IPFW_TLV_TBLNAME_LIST) {
2890 		clen = ctlv->head.length;
2891 		/* Check size and alignment */
2892 		if (clen > sd->valsize || clen < sizeof(*ctlv))
2893 			return (EINVAL);
2894 		if ((clen % sizeof(uint64_t)) != 0)
2895 			return (EINVAL);
2896 
2897 		/*
2898 		 * Some table names or other named objects.
2899 		 * Check for validness.
2900 		 */
2901 		count = (ctlv->head.length - sizeof(*ctlv)) / sizeof(*ntlv);
2902 		if (ctlv->count != count || ctlv->objsize != sizeof(*ntlv))
2903 			return (EINVAL);
2904 
2905 		/*
2906 		 * Check each TLV.
2907 		 * Ensure TLVs are sorted ascending and
2908 		 * there are no duplicates.
2909 		 */
2910 		idx = -1;
2911 		ntlv = (ipfw_obj_ntlv *)(ctlv + 1);
2912 		while (count > 0) {
2913 			if (ntlv->head.length != sizeof(ipfw_obj_ntlv))
2914 				return (EINVAL);
2915 
2916 			error = ipfw_check_object_name_generic(ntlv->name);
2917 			if (error != 0)
2918 				return (error);
2919 
2920 			if (ntlv->idx <= idx)
2921 				return (EINVAL);
2922 
2923 			idx = ntlv->idx;
2924 			count--;
2925 			ntlv++;
2926 		}
2927 
2928 		tstate = ctlv;
2929 		read += ctlv->head.length;
2930 		ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length);
2931 	}
2932 
2933 	if (read + sizeof(*ctlv) > sd->valsize)
2934 		return (EINVAL);
2935 
2936 	if (ctlv->head.type == IPFW_TLV_RULE_LIST) {
2937 		clen = ctlv->head.length;
2938 		if (clen + read > sd->valsize || clen < sizeof(*ctlv))
2939 			return (EINVAL);
2940 		if ((clen % sizeof(uint64_t)) != 0)
2941 			return (EINVAL);
2942 
2943 		/*
2944 		 * TODO: Permit adding multiple rules at once
2945 		 */
2946 		if (ctlv->count != 1)
2947 			return (ENOTSUP);
2948 
2949 		clen -= sizeof(*ctlv);
2950 
2951 		if (ctlv->count > clen / sizeof(struct ip_fw_rule))
2952 			return (EINVAL);
2953 
2954 		/* Allocate state for each rule or use stack */
2955 		if (ctlv->count == 1) {
2956 			memset(&rci, 0, sizeof(struct rule_check_info));
2957 			cbuf = &rci;
2958 		} else
2959 			cbuf = malloc(ctlv->count * sizeof(*ci), M_TEMP,
2960 			    M_WAITOK | M_ZERO);
2961 		ci = cbuf;
2962 
2963 		/*
2964 		 * Check each rule for validness.
2965 		 * Ensure numbered rules are sorted ascending
2966 		 * and properly aligned
2967 		 */
2968 		idx = 0;
2969 		r = (struct ip_fw_rule *)(ctlv + 1);
2970 		count = 0;
2971 		error = 0;
2972 		while (clen > 0) {
2973 			rsize = roundup2(RULESIZE(r), sizeof(uint64_t));
2974 			if (rsize > clen || ctlv->count <= count) {
2975 				error = EINVAL;
2976 				break;
2977 			}
2978 
2979 			ci->ctlv = tstate;
2980 			error = check_ipfw_rule1(r, rsize, ci);
2981 			if (error != 0)
2982 				break;
2983 
2984 			/* Check sorting */
2985 			if (r->rulenum != 0 && r->rulenum < idx) {
2986 				printf("rulenum %d idx %d\n", r->rulenum, idx);
2987 				error = EINVAL;
2988 				break;
2989 			}
2990 			idx = r->rulenum;
2991 
2992 			ci->urule = (caddr_t)r;
2993 
2994 			rsize = roundup2(rsize, sizeof(uint64_t));
2995 			clen -= rsize;
2996 			r = (struct ip_fw_rule *)((caddr_t)r + rsize);
2997 			count++;
2998 			ci++;
2999 		}
3000 
3001 		if (ctlv->count != count || error != 0) {
3002 			if (cbuf != &rci)
3003 				free(cbuf, M_TEMP);
3004 			return (EINVAL);
3005 		}
3006 
3007 		rtlv = ctlv;
3008 		read += ctlv->head.length;
3009 		ctlv = (ipfw_obj_ctlv *)((caddr_t)ctlv + ctlv->head.length);
3010 	}
3011 
3012 	if (read != sd->valsize || rtlv == NULL || rtlv->count == 0) {
3013 		if (cbuf != NULL && cbuf != &rci)
3014 			free(cbuf, M_TEMP);
3015 		return (EINVAL);
3016 	}
3017 
3018 	/*
3019 	 * Passed rules seems to be valid.
3020 	 * Allocate storage and try to add them to chain.
3021 	 */
3022 	for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++) {
3023 		clen = RULEKSIZE1((struct ip_fw_rule *)ci->urule);
3024 		ci->krule = ipfw_alloc_rule(chain, clen);
3025 		import_rule1(ci);
3026 	}
3027 
3028 	if ((error = commit_rules(chain, cbuf, rtlv->count)) != 0) {
3029 		/* Free allocate krules */
3030 		for (i = 0, ci = cbuf; i < rtlv->count; i++, ci++)
3031 			free_rule(ci->krule);
3032 	}
3033 
3034 	if (cbuf != NULL && cbuf != &rci)
3035 		free(cbuf, M_TEMP);
3036 
3037 	return (error);
3038 }
3039 
3040 /*
3041  * Lists all sopts currently registered.
3042  * Data layout (v0)(current):
3043  * Request: [ ipfw_obj_lheader ], size = ipfw_obj_lheader.size
3044  * Reply: [ ipfw_obj_lheader ipfw_sopt_info x N ]
3045  *
3046  * Returns 0 on success
3047  */
3048 static int
3049 dump_soptcodes(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
3050     struct sockopt_data *sd)
3051 {
3052 	struct _ipfw_obj_lheader *olh;
3053 	ipfw_sopt_info *i;
3054 	struct ipfw_sopt_handler *sh;
3055 	uint32_t count, n, size;
3056 
3057 	olh = (struct _ipfw_obj_lheader *)ipfw_get_sopt_header(sd,sizeof(*olh));
3058 	if (olh == NULL)
3059 		return (EINVAL);
3060 	if (sd->valsize < olh->size)
3061 		return (EINVAL);
3062 
3063 	CTL3_LOCK();
3064 	count = ctl3_hsize;
3065 	size = count * sizeof(ipfw_sopt_info) + sizeof(ipfw_obj_lheader);
3066 
3067 	/* Fill in header regadless of buffer size */
3068 	olh->count = count;
3069 	olh->objsize = sizeof(ipfw_sopt_info);
3070 
3071 	if (size > olh->size) {
3072 		olh->size = size;
3073 		CTL3_UNLOCK();
3074 		return (ENOMEM);
3075 	}
3076 	olh->size = size;
3077 
3078 	for (n = 1; n <= count; n++) {
3079 		i = (ipfw_sopt_info *)ipfw_get_sopt_space(sd, sizeof(*i));
3080 		KASSERT(i != NULL, ("previously checked buffer is not enough"));
3081 		sh = &ctl3_handlers[n];
3082 		i->opcode = sh->opcode;
3083 		i->version = sh->version;
3084 		i->refcnt = sh->refcnt;
3085 	}
3086 	CTL3_UNLOCK();
3087 
3088 	return (0);
3089 }
3090 
3091 /*
3092  * Compares two opcodes.
3093  * Used both in qsort() and bsearch().
3094  *
3095  * Returns 0 if match is found.
3096  */
3097 static int
3098 compare_opcodes(const void *_a, const void *_b)
3099 {
3100 	const struct opcode_obj_rewrite *a, *b;
3101 
3102 	a = (const struct opcode_obj_rewrite *)_a;
3103 	b = (const struct opcode_obj_rewrite *)_b;
3104 
3105 	if (a->opcode < b->opcode)
3106 		return (-1);
3107 	else if (a->opcode > b->opcode)
3108 		return (1);
3109 
3110 	return (0);
3111 }
3112 
3113 /*
3114  * XXX: Rewrite bsearch()
3115  */
3116 static int
3117 find_op_rw_range(uint16_t op, struct opcode_obj_rewrite **plo,
3118     struct opcode_obj_rewrite **phi)
3119 {
3120 	struct opcode_obj_rewrite *ctl3_max, *lo, *hi, h, *rw;
3121 
3122 	memset(&h, 0, sizeof(h));
3123 	h.opcode = op;
3124 
3125 	rw = (struct opcode_obj_rewrite *)bsearch(&h, ctl3_rewriters,
3126 	    ctl3_rsize, sizeof(h), compare_opcodes);
3127 	if (rw == NULL)
3128 		return (1);
3129 
3130 	/* Find the first element matching the same opcode */
3131 	lo = rw;
3132 	for ( ; lo > ctl3_rewriters && (lo - 1)->opcode == op; lo--)
3133 		;
3134 
3135 	/* Find the last element matching the same opcode */
3136 	hi = rw;
3137 	ctl3_max = ctl3_rewriters + ctl3_rsize;
3138 	for ( ; (hi + 1) < ctl3_max && (hi + 1)->opcode == op; hi++)
3139 		;
3140 
3141 	*plo = lo;
3142 	*phi = hi;
3143 
3144 	return (0);
3145 }
3146 
3147 /*
3148  * Finds opcode object rewriter based on @code.
3149  *
3150  * Returns pointer to handler or NULL.
3151  */
3152 static struct opcode_obj_rewrite *
3153 find_op_rw(ipfw_insn *cmd, uint16_t *puidx, uint8_t *ptype)
3154 {
3155 	struct opcode_obj_rewrite *rw, *lo, *hi;
3156 	uint16_t uidx;
3157 	uint8_t subtype;
3158 
3159 	if (find_op_rw_range(cmd->opcode, &lo, &hi) != 0)
3160 		return (NULL);
3161 
3162 	for (rw = lo; rw <= hi; rw++) {
3163 		if (rw->classifier(cmd, &uidx, &subtype) == 0) {
3164 			if (puidx != NULL)
3165 				*puidx = uidx;
3166 			if (ptype != NULL)
3167 				*ptype = subtype;
3168 			return (rw);
3169 		}
3170 	}
3171 
3172 	return (NULL);
3173 }
3174 int
3175 classify_opcode_kidx(ipfw_insn *cmd, uint16_t *puidx)
3176 {
3177 
3178 	if (find_op_rw(cmd, puidx, NULL) == NULL)
3179 		return (1);
3180 	return (0);
3181 }
3182 
3183 void
3184 update_opcode_kidx(ipfw_insn *cmd, uint16_t idx)
3185 {
3186 	struct opcode_obj_rewrite *rw;
3187 
3188 	rw = find_op_rw(cmd, NULL, NULL);
3189 	KASSERT(rw != NULL, ("No handler to update opcode %d", cmd->opcode));
3190 	rw->update(cmd, idx);
3191 }
3192 
3193 void
3194 ipfw_init_obj_rewriter()
3195 {
3196 
3197 	ctl3_rewriters = NULL;
3198 	ctl3_rsize = 0;
3199 }
3200 
3201 void
3202 ipfw_destroy_obj_rewriter()
3203 {
3204 
3205 	if (ctl3_rewriters != NULL)
3206 		free(ctl3_rewriters, M_IPFW);
3207 	ctl3_rewriters = NULL;
3208 	ctl3_rsize = 0;
3209 }
3210 
3211 /*
3212  * Adds one or more opcode object rewrite handlers to the global array.
3213  * Function may sleep.
3214  */
3215 void
3216 ipfw_add_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count)
3217 {
3218 	size_t sz;
3219 	struct opcode_obj_rewrite *tmp;
3220 
3221 	CTL3_LOCK();
3222 
3223 	for (;;) {
3224 		sz = ctl3_rsize + count;
3225 		CTL3_UNLOCK();
3226 		tmp = malloc(sizeof(*rw) * sz, M_IPFW, M_WAITOK | M_ZERO);
3227 		CTL3_LOCK();
3228 		if (ctl3_rsize + count <= sz)
3229 			break;
3230 
3231 		/* Retry */
3232 		free(tmp, M_IPFW);
3233 	}
3234 
3235 	/* Merge old & new arrays */
3236 	sz = ctl3_rsize + count;
3237 	memcpy(tmp, ctl3_rewriters, ctl3_rsize * sizeof(*rw));
3238 	memcpy(&tmp[ctl3_rsize], rw, count * sizeof(*rw));
3239 	qsort(tmp, sz, sizeof(*rw), compare_opcodes);
3240 	/* Switch new and free old */
3241 	if (ctl3_rewriters != NULL)
3242 		free(ctl3_rewriters, M_IPFW);
3243 	ctl3_rewriters = tmp;
3244 	ctl3_rsize = sz;
3245 
3246 	CTL3_UNLOCK();
3247 }
3248 
3249 /*
3250  * Removes one or more object rewrite handlers from the global array.
3251  */
3252 int
3253 ipfw_del_obj_rewriter(struct opcode_obj_rewrite *rw, size_t count)
3254 {
3255 	size_t sz;
3256 	struct opcode_obj_rewrite *ctl3_max, *ktmp, *lo, *hi;
3257 	int i;
3258 
3259 	CTL3_LOCK();
3260 
3261 	for (i = 0; i < count; i++) {
3262 		if (find_op_rw_range(rw[i].opcode, &lo, &hi) != 0)
3263 			continue;
3264 
3265 		for (ktmp = lo; ktmp <= hi; ktmp++) {
3266 			if (ktmp->classifier != rw[i].classifier)
3267 				continue;
3268 
3269 			ctl3_max = ctl3_rewriters + ctl3_rsize;
3270 			sz = (ctl3_max - (ktmp + 1)) * sizeof(*ktmp);
3271 			memmove(ktmp, ktmp + 1, sz);
3272 			ctl3_rsize--;
3273 			break;
3274 		}
3275 
3276 	}
3277 
3278 	if (ctl3_rsize == 0) {
3279 		if (ctl3_rewriters != NULL)
3280 			free(ctl3_rewriters, M_IPFW);
3281 		ctl3_rewriters = NULL;
3282 	}
3283 
3284 	CTL3_UNLOCK();
3285 
3286 	return (0);
3287 }
3288 
3289 static int
3290 export_objhash_ntlv_internal(struct namedobj_instance *ni,
3291     struct named_object *no, void *arg)
3292 {
3293 	struct sockopt_data *sd;
3294 	ipfw_obj_ntlv *ntlv;
3295 
3296 	sd = (struct sockopt_data *)arg;
3297 	ntlv = (ipfw_obj_ntlv *)ipfw_get_sopt_space(sd, sizeof(*ntlv));
3298 	if (ntlv == NULL)
3299 		return (ENOMEM);
3300 	ipfw_export_obj_ntlv(no, ntlv);
3301 	return (0);
3302 }
3303 
3304 /*
3305  * Lists all service objects.
3306  * Data layout (v0)(current):
3307  * Request: [ ipfw_obj_lheader ] size = ipfw_obj_lheader.size
3308  * Reply: [ ipfw_obj_lheader [ ipfw_obj_ntlv x N ] (optional) ]
3309  * Returns 0 on success
3310  */
3311 static int
3312 dump_srvobjects(struct ip_fw_chain *chain, ip_fw3_opheader *op3,
3313     struct sockopt_data *sd)
3314 {
3315 	ipfw_obj_lheader *hdr;
3316 	int count;
3317 
3318 	hdr = (ipfw_obj_lheader *)ipfw_get_sopt_header(sd, sizeof(*hdr));
3319 	if (hdr == NULL)
3320 		return (EINVAL);
3321 
3322 	IPFW_UH_RLOCK(chain);
3323 	count = ipfw_objhash_count(CHAIN_TO_SRV(chain));
3324 	hdr->size = sizeof(ipfw_obj_lheader) + count * sizeof(ipfw_obj_ntlv);
3325 	if (sd->valsize < hdr->size) {
3326 		IPFW_UH_RUNLOCK(chain);
3327 		return (ENOMEM);
3328 	}
3329 	hdr->count = count;
3330 	hdr->objsize = sizeof(ipfw_obj_ntlv);
3331 	if (count > 0)
3332 		ipfw_objhash_foreach(CHAIN_TO_SRV(chain),
3333 		    export_objhash_ntlv_internal, sd);
3334 	IPFW_UH_RUNLOCK(chain);
3335 	return (0);
3336 }
3337 
3338 /*
3339  * Compares two sopt handlers (code, version and handler ptr).
3340  * Used both as qsort() and bsearch().
3341  * Does not compare handler for latter case.
3342  *
3343  * Returns 0 if match is found.
3344  */
3345 static int
3346 compare_sh(const void *_a, const void *_b)
3347 {
3348 	const struct ipfw_sopt_handler *a, *b;
3349 
3350 	a = (const struct ipfw_sopt_handler *)_a;
3351 	b = (const struct ipfw_sopt_handler *)_b;
3352 
3353 	if (a->opcode < b->opcode)
3354 		return (-1);
3355 	else if (a->opcode > b->opcode)
3356 		return (1);
3357 
3358 	if (a->version < b->version)
3359 		return (-1);
3360 	else if (a->version > b->version)
3361 		return (1);
3362 
3363 	/* bsearch helper */
3364 	if (a->handler == NULL)
3365 		return (0);
3366 
3367 	if ((uintptr_t)a->handler < (uintptr_t)b->handler)
3368 		return (-1);
3369 	else if ((uintptr_t)a->handler > (uintptr_t)b->handler)
3370 		return (1);
3371 
3372 	return (0);
3373 }
3374 
3375 /*
3376  * Finds sopt handler based on @code and @version.
3377  *
3378  * Returns pointer to handler or NULL.
3379  */
3380 static struct ipfw_sopt_handler *
3381 find_sh(uint16_t code, uint8_t version, sopt_handler_f *handler)
3382 {
3383 	struct ipfw_sopt_handler *sh, h;
3384 
3385 	memset(&h, 0, sizeof(h));
3386 	h.opcode = code;
3387 	h.version = version;
3388 	h.handler = handler;
3389 
3390 	sh = (struct ipfw_sopt_handler *)bsearch(&h, ctl3_handlers,
3391 	    ctl3_hsize, sizeof(h), compare_sh);
3392 
3393 	return (sh);
3394 }
3395 
3396 static int
3397 find_ref_sh(uint16_t opcode, uint8_t version, struct ipfw_sopt_handler *psh)
3398 {
3399 	struct ipfw_sopt_handler *sh;
3400 
3401 	CTL3_LOCK();
3402 	if ((sh = find_sh(opcode, version, NULL)) == NULL) {
3403 		CTL3_UNLOCK();
3404 		printf("ipfw: ipfw_ctl3 invalid option %d""v""%d\n",
3405 		    opcode, version);
3406 		return (EINVAL);
3407 	}
3408 	sh->refcnt++;
3409 	ctl3_refct++;
3410 	/* Copy handler data to requested buffer */
3411 	*psh = *sh;
3412 	CTL3_UNLOCK();
3413 
3414 	return (0);
3415 }
3416 
3417 static void
3418 find_unref_sh(struct ipfw_sopt_handler *psh)
3419 {
3420 	struct ipfw_sopt_handler *sh;
3421 
3422 	CTL3_LOCK();
3423 	sh = find_sh(psh->opcode, psh->version, NULL);
3424 	KASSERT(sh != NULL, ("ctl3 handler disappeared"));
3425 	sh->refcnt--;
3426 	ctl3_refct--;
3427 	CTL3_UNLOCK();
3428 }
3429 
3430 void
3431 ipfw_init_sopt_handler()
3432 {
3433 
3434 	CTL3_LOCK_INIT();
3435 	IPFW_ADD_SOPT_HANDLER(1, scodes);
3436 }
3437 
3438 void
3439 ipfw_destroy_sopt_handler()
3440 {
3441 
3442 	IPFW_DEL_SOPT_HANDLER(1, scodes);
3443 	CTL3_LOCK_DESTROY();
3444 }
3445 
3446 /*
3447  * Adds one or more sockopt handlers to the global array.
3448  * Function may sleep.
3449  */
3450 void
3451 ipfw_add_sopt_handler(struct ipfw_sopt_handler *sh, size_t count)
3452 {
3453 	size_t sz;
3454 	struct ipfw_sopt_handler *tmp;
3455 
3456 	CTL3_LOCK();
3457 
3458 	for (;;) {
3459 		sz = ctl3_hsize + count;
3460 		CTL3_UNLOCK();
3461 		tmp = malloc(sizeof(*sh) * sz, M_IPFW, M_WAITOK | M_ZERO);
3462 		CTL3_LOCK();
3463 		if (ctl3_hsize + count <= sz)
3464 			break;
3465 
3466 		/* Retry */
3467 		free(tmp, M_IPFW);
3468 	}
3469 
3470 	/* Merge old & new arrays */
3471 	sz = ctl3_hsize + count;
3472 	memcpy(tmp, ctl3_handlers, ctl3_hsize * sizeof(*sh));
3473 	memcpy(&tmp[ctl3_hsize], sh, count * sizeof(*sh));
3474 	qsort(tmp, sz, sizeof(*sh), compare_sh);
3475 	/* Switch new and free old */
3476 	if (ctl3_handlers != NULL)
3477 		free(ctl3_handlers, M_IPFW);
3478 	ctl3_handlers = tmp;
3479 	ctl3_hsize = sz;
3480 	ctl3_gencnt++;
3481 
3482 	CTL3_UNLOCK();
3483 }
3484 
3485 /*
3486  * Removes one or more sockopt handlers from the global array.
3487  */
3488 int
3489 ipfw_del_sopt_handler(struct ipfw_sopt_handler *sh, size_t count)
3490 {
3491 	size_t sz;
3492 	struct ipfw_sopt_handler *tmp, *h;
3493 	int i;
3494 
3495 	CTL3_LOCK();
3496 
3497 	for (i = 0; i < count; i++) {
3498 		tmp = &sh[i];
3499 		h = find_sh(tmp->opcode, tmp->version, tmp->handler);
3500 		if (h == NULL)
3501 			continue;
3502 
3503 		sz = (ctl3_handlers + ctl3_hsize - (h + 1)) * sizeof(*h);
3504 		memmove(h, h + 1, sz);
3505 		ctl3_hsize--;
3506 	}
3507 
3508 	if (ctl3_hsize == 0) {
3509 		if (ctl3_handlers != NULL)
3510 			free(ctl3_handlers, M_IPFW);
3511 		ctl3_handlers = NULL;
3512 	}
3513 
3514 	ctl3_gencnt++;
3515 
3516 	CTL3_UNLOCK();
3517 
3518 	return (0);
3519 }
3520 
3521 /*
3522  * Writes data accumulated in @sd to sockopt buffer.
3523  * Zeroes internal @sd buffer.
3524  */
3525 static int
3526 ipfw_flush_sopt_data(struct sockopt_data *sd)
3527 {
3528 	struct sockopt *sopt;
3529 	int error;
3530 	size_t sz;
3531 
3532 	sz = sd->koff;
3533 	if (sz == 0)
3534 		return (0);
3535 
3536 	sopt = sd->sopt;
3537 
3538 	if (sopt->sopt_dir == SOPT_GET) {
3539 		error = copyout(sd->kbuf, sopt->sopt_val, sz);
3540 		if (error != 0)
3541 			return (error);
3542 	}
3543 
3544 	memset(sd->kbuf, 0, sd->ksize);
3545 	sd->ktotal += sz;
3546 	sd->koff = 0;
3547 	if (sd->ktotal + sd->ksize < sd->valsize)
3548 		sd->kavail = sd->ksize;
3549 	else
3550 		sd->kavail = sd->valsize - sd->ktotal;
3551 
3552 	/* Update sopt buffer data */
3553 	sopt->sopt_valsize = sd->ktotal;
3554 	sopt->sopt_val = sd->sopt_val + sd->ktotal;
3555 
3556 	return (0);
3557 }
3558 
3559 /*
3560  * Ensures that @sd buffer has contiguous @neeeded number of
3561  * bytes.
3562  *
3563  * Returns pointer to requested space or NULL.
3564  */
3565 caddr_t
3566 ipfw_get_sopt_space(struct sockopt_data *sd, size_t needed)
3567 {
3568 	int error;
3569 	caddr_t addr;
3570 
3571 	if (sd->kavail < needed) {
3572 		/*
3573 		 * Flush data and try another time.
3574 		 */
3575 		error = ipfw_flush_sopt_data(sd);
3576 
3577 		if (sd->kavail < needed || error != 0)
3578 			return (NULL);
3579 	}
3580 
3581 	addr = sd->kbuf + sd->koff;
3582 	sd->koff += needed;
3583 	sd->kavail -= needed;
3584 	return (addr);
3585 }
3586 
3587 /*
3588  * Requests @needed contiguous bytes from @sd buffer.
3589  * Function is used to notify subsystem that we are
3590  * interesed in first @needed bytes (request header)
3591  * and the rest buffer can be safely zeroed.
3592  *
3593  * Returns pointer to requested space or NULL.
3594  */
3595 caddr_t
3596 ipfw_get_sopt_header(struct sockopt_data *sd, size_t needed)
3597 {
3598 	caddr_t addr;
3599 
3600 	if ((addr = ipfw_get_sopt_space(sd, needed)) == NULL)
3601 		return (NULL);
3602 
3603 	if (sd->kavail > 0)
3604 		memset(sd->kbuf + sd->koff, 0, sd->kavail);
3605 
3606 	return (addr);
3607 }
3608 
3609 /*
3610  * New sockopt handler.
3611  */
3612 int
3613 ipfw_ctl3(struct sockopt *sopt)
3614 {
3615 	int error, locked;
3616 	size_t size, valsize;
3617 	struct ip_fw_chain *chain;
3618 	char xbuf[256];
3619 	struct sockopt_data sdata;
3620 	struct ipfw_sopt_handler h;
3621 	ip_fw3_opheader *op3 = NULL;
3622 
3623 	error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
3624 	if (error != 0)
3625 		return (error);
3626 
3627 	if (sopt->sopt_name != IP_FW3)
3628 		return (ipfw_ctl(sopt));
3629 
3630 	chain = &V_layer3_chain;
3631 	error = 0;
3632 
3633 	/* Save original valsize before it is altered via sooptcopyin() */
3634 	valsize = sopt->sopt_valsize;
3635 	memset(&sdata, 0, sizeof(sdata));
3636 	/* Read op3 header first to determine actual operation */
3637 	op3 = (ip_fw3_opheader *)xbuf;
3638 	error = sooptcopyin(sopt, op3, sizeof(*op3), sizeof(*op3));
3639 	if (error != 0)
3640 		return (error);
3641 	sopt->sopt_valsize = valsize;
3642 
3643 	/*
3644 	 * Find and reference command.
3645 	 */
3646 	error = find_ref_sh(op3->opcode, op3->version, &h);
3647 	if (error != 0)
3648 		return (error);
3649 
3650 	/*
3651 	 * Disallow modifications in really-really secure mode, but still allow
3652 	 * the logging counters to be reset.
3653 	 */
3654 	if ((h.dir & HDIR_SET) != 0 && h.opcode != IP_FW_XRESETLOG) {
3655 		error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
3656 		if (error != 0) {
3657 			find_unref_sh(&h);
3658 			return (error);
3659 		}
3660 	}
3661 
3662 	/*
3663 	 * Fill in sockopt_data structure that may be useful for
3664 	 * IP_FW3 get requests.
3665 	 */
3666 	locked = 0;
3667 	if (valsize <= sizeof(xbuf)) {
3668 		/* use on-stack buffer */
3669 		sdata.kbuf = xbuf;
3670 		sdata.ksize = sizeof(xbuf);
3671 		sdata.kavail = valsize;
3672 	} else {
3673 
3674 		/*
3675 		 * Determine opcode type/buffer size:
3676 		 * allocate sliding-window buf for data export or
3677 		 * contiguous buffer for special ops.
3678 		 */
3679 		if ((h.dir & HDIR_SET) != 0) {
3680 			/* Set request. Allocate contigous buffer. */
3681 			if (valsize > CTL3_LARGEBUF) {
3682 				find_unref_sh(&h);
3683 				return (EFBIG);
3684 			}
3685 
3686 			size = valsize;
3687 		} else {
3688 			/* Get request. Allocate sliding window buffer */
3689 			size = (valsize<CTL3_SMALLBUF) ? valsize:CTL3_SMALLBUF;
3690 
3691 			if (size < valsize) {
3692 				/* We have to wire user buffer */
3693 				error = vslock(sopt->sopt_val, valsize);
3694 				if (error != 0)
3695 					return (error);
3696 				locked = 1;
3697 			}
3698 		}
3699 
3700 		sdata.kbuf = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
3701 		sdata.ksize = size;
3702 		sdata.kavail = size;
3703 	}
3704 
3705 	sdata.sopt = sopt;
3706 	sdata.sopt_val = sopt->sopt_val;
3707 	sdata.valsize = valsize;
3708 
3709 	/*
3710 	 * Copy either all request (if valsize < bsize_max)
3711 	 * or first bsize_max bytes to guarantee most consumers
3712 	 * that all necessary data has been copied).
3713 	 * Anyway, copy not less than sizeof(ip_fw3_opheader).
3714 	 */
3715 	if ((error = sooptcopyin(sopt, sdata.kbuf, sdata.ksize,
3716 	    sizeof(ip_fw3_opheader))) != 0)
3717 		return (error);
3718 	op3 = (ip_fw3_opheader *)sdata.kbuf;
3719 
3720 	/* Finally, run handler */
3721 	error = h.handler(chain, op3, &sdata);
3722 	find_unref_sh(&h);
3723 
3724 	/* Flush state and free buffers */
3725 	if (error == 0)
3726 		error = ipfw_flush_sopt_data(&sdata);
3727 	else
3728 		ipfw_flush_sopt_data(&sdata);
3729 
3730 	if (locked != 0)
3731 		vsunlock(sdata.sopt_val, valsize);
3732 
3733 	/* Restore original pointer and set number of bytes written */
3734 	sopt->sopt_val = sdata.sopt_val;
3735 	sopt->sopt_valsize = sdata.ktotal;
3736 	if (sdata.kbuf != xbuf)
3737 		free(sdata.kbuf, M_TEMP);
3738 
3739 	return (error);
3740 }
3741 
3742 /**
3743  * {set|get}sockopt parser.
3744  */
3745 int
3746 ipfw_ctl(struct sockopt *sopt)
3747 {
3748 #define	RULE_MAXSIZE	(512*sizeof(u_int32_t))
3749 	int error;
3750 	size_t size, valsize;
3751 	struct ip_fw *buf;
3752 	struct ip_fw_rule0 *rule;
3753 	struct ip_fw_chain *chain;
3754 	u_int32_t rulenum[2];
3755 	uint32_t opt;
3756 	struct rule_check_info ci;
3757 	IPFW_RLOCK_TRACKER;
3758 
3759 	chain = &V_layer3_chain;
3760 	error = 0;
3761 
3762 	/* Save original valsize before it is altered via sooptcopyin() */
3763 	valsize = sopt->sopt_valsize;
3764 	opt = sopt->sopt_name;
3765 
3766 	/*
3767 	 * Disallow modifications in really-really secure mode, but still allow
3768 	 * the logging counters to be reset.
3769 	 */
3770 	if (opt == IP_FW_ADD ||
3771 	    (sopt->sopt_dir == SOPT_SET && opt != IP_FW_RESETLOG)) {
3772 		error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
3773 		if (error != 0)
3774 			return (error);
3775 	}
3776 
3777 	switch (opt) {
3778 	case IP_FW_GET:
3779 		/*
3780 		 * pass up a copy of the current rules. Static rules
3781 		 * come first (the last of which has number IPFW_DEFAULT_RULE),
3782 		 * followed by a possibly empty list of dynamic rule.
3783 		 * The last dynamic rule has NULL in the "next" field.
3784 		 *
3785 		 * Note that the calculated size is used to bound the
3786 		 * amount of data returned to the user.  The rule set may
3787 		 * change between calculating the size and returning the
3788 		 * data in which case we'll just return what fits.
3789 		 */
3790 		for (;;) {
3791 			int len = 0, want;
3792 
3793 			size = chain->static_len;
3794 			size += ipfw_dyn_len();
3795 			if (size >= sopt->sopt_valsize)
3796 				break;
3797 			buf = malloc(size, M_TEMP, M_WAITOK | M_ZERO);
3798 			IPFW_UH_RLOCK(chain);
3799 			/* check again how much space we need */
3800 			want = chain->static_len + ipfw_dyn_len();
3801 			if (size >= want)
3802 				len = ipfw_getrules(chain, buf, size);
3803 			IPFW_UH_RUNLOCK(chain);
3804 			if (size >= want)
3805 				error = sooptcopyout(sopt, buf, len);
3806 			free(buf, M_TEMP);
3807 			if (size >= want)
3808 				break;
3809 		}
3810 		break;
3811 
3812 	case IP_FW_FLUSH:
3813 		/* locking is done within del_entry() */
3814 		error = del_entry(chain, 0); /* special case, rule=0, cmd=0 means all */
3815 		break;
3816 
3817 	case IP_FW_ADD:
3818 		rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
3819 		error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
3820 			sizeof(struct ip_fw7) );
3821 
3822 		memset(&ci, 0, sizeof(struct rule_check_info));
3823 
3824 		/*
3825 		 * If the size of commands equals RULESIZE7 then we assume
3826 		 * a FreeBSD7.2 binary is talking to us (set is7=1).
3827 		 * is7 is persistent so the next 'ipfw list' command
3828 		 * will use this format.
3829 		 * NOTE: If wrong version is guessed (this can happen if
3830 		 *       the first ipfw command is 'ipfw [pipe] list')
3831 		 *       the ipfw binary may crash or loop infinitly...
3832 		 */
3833 		size = sopt->sopt_valsize;
3834 		if (size == RULESIZE7(rule)) {
3835 		    is7 = 1;
3836 		    error = convert_rule_to_8(rule);
3837 		    if (error) {
3838 			free(rule, M_TEMP);
3839 			return error;
3840 		    }
3841 		    size = RULESIZE(rule);
3842 		} else
3843 		    is7 = 0;
3844 		if (error == 0)
3845 			error = check_ipfw_rule0(rule, size, &ci);
3846 		if (error == 0) {
3847 			/* locking is done within add_rule() */
3848 			struct ip_fw *krule;
3849 			krule = ipfw_alloc_rule(chain, RULEKSIZE0(rule));
3850 			ci.urule = (caddr_t)rule;
3851 			ci.krule = krule;
3852 			import_rule0(&ci);
3853 			error = commit_rules(chain, &ci, 1);
3854 			if (error != 0)
3855 				free_rule(ci.krule);
3856 			else if (sopt->sopt_dir == SOPT_GET) {
3857 				if (is7) {
3858 					error = convert_rule_to_7(rule);
3859 					size = RULESIZE7(rule);
3860 					if (error) {
3861 						free(rule, M_TEMP);
3862 						return error;
3863 					}
3864 				}
3865 				error = sooptcopyout(sopt, rule, size);
3866 			}
3867 		}
3868 		free(rule, M_TEMP);
3869 		break;
3870 
3871 	case IP_FW_DEL:
3872 		/*
3873 		 * IP_FW_DEL is used for deleting single rules or sets,
3874 		 * and (ab)used to atomically manipulate sets. Argument size
3875 		 * is used to distinguish between the two:
3876 		 *    sizeof(u_int32_t)
3877 		 *	delete single rule or set of rules,
3878 		 *	or reassign rules (or sets) to a different set.
3879 		 *    2*sizeof(u_int32_t)
3880 		 *	atomic disable/enable sets.
3881 		 *	first u_int32_t contains sets to be disabled,
3882 		 *	second u_int32_t contains sets to be enabled.
3883 		 */
3884 		error = sooptcopyin(sopt, rulenum,
3885 			2*sizeof(u_int32_t), sizeof(u_int32_t));
3886 		if (error)
3887 			break;
3888 		size = sopt->sopt_valsize;
3889 		if (size == sizeof(u_int32_t) && rulenum[0] != 0) {
3890 			/* delete or reassign, locking done in del_entry() */
3891 			error = del_entry(chain, rulenum[0]);
3892 		} else if (size == 2*sizeof(u_int32_t)) { /* set enable/disable */
3893 			IPFW_UH_WLOCK(chain);
3894 			V_set_disable =
3895 			    (V_set_disable | rulenum[0]) & ~rulenum[1] &
3896 			    ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
3897 			IPFW_UH_WUNLOCK(chain);
3898 		} else
3899 			error = EINVAL;
3900 		break;
3901 
3902 	case IP_FW_ZERO:
3903 	case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
3904 		rulenum[0] = 0;
3905 		if (sopt->sopt_val != 0) {
3906 		    error = sooptcopyin(sopt, rulenum,
3907 			    sizeof(u_int32_t), sizeof(u_int32_t));
3908 		    if (error)
3909 			break;
3910 		}
3911 		error = zero_entry(chain, rulenum[0],
3912 			sopt->sopt_name == IP_FW_RESETLOG);
3913 		break;
3914 
3915 	/*--- TABLE opcodes ---*/
3916 	case IP_FW_TABLE_ADD:
3917 	case IP_FW_TABLE_DEL:
3918 		{
3919 			ipfw_table_entry ent;
3920 			struct tentry_info tei;
3921 			struct tid_info ti;
3922 			struct table_value v;
3923 
3924 			error = sooptcopyin(sopt, &ent,
3925 			    sizeof(ent), sizeof(ent));
3926 			if (error)
3927 				break;
3928 
3929 			memset(&tei, 0, sizeof(tei));
3930 			tei.paddr = &ent.addr;
3931 			tei.subtype = AF_INET;
3932 			tei.masklen = ent.masklen;
3933 			ipfw_import_table_value_legacy(ent.value, &v);
3934 			tei.pvalue = &v;
3935 			memset(&ti, 0, sizeof(ti));
3936 			ti.uidx = ent.tbl;
3937 			ti.type = IPFW_TABLE_CIDR;
3938 
3939 			error = (opt == IP_FW_TABLE_ADD) ?
3940 			    add_table_entry(chain, &ti, &tei, 0, 1) :
3941 			    del_table_entry(chain, &ti, &tei, 0, 1);
3942 		}
3943 		break;
3944 
3945 
3946 	case IP_FW_TABLE_FLUSH:
3947 		{
3948 			u_int16_t tbl;
3949 			struct tid_info ti;
3950 
3951 			error = sooptcopyin(sopt, &tbl,
3952 			    sizeof(tbl), sizeof(tbl));
3953 			if (error)
3954 				break;
3955 			memset(&ti, 0, sizeof(ti));
3956 			ti.uidx = tbl;
3957 			error = flush_table(chain, &ti);
3958 		}
3959 		break;
3960 
3961 	case IP_FW_TABLE_GETSIZE:
3962 		{
3963 			u_int32_t tbl, cnt;
3964 			struct tid_info ti;
3965 
3966 			if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
3967 			    sizeof(tbl))))
3968 				break;
3969 			memset(&ti, 0, sizeof(ti));
3970 			ti.uidx = tbl;
3971 			IPFW_RLOCK(chain);
3972 			error = ipfw_count_table(chain, &ti, &cnt);
3973 			IPFW_RUNLOCK(chain);
3974 			if (error)
3975 				break;
3976 			error = sooptcopyout(sopt, &cnt, sizeof(cnt));
3977 		}
3978 		break;
3979 
3980 	case IP_FW_TABLE_LIST:
3981 		{
3982 			ipfw_table *tbl;
3983 			struct tid_info ti;
3984 
3985 			if (sopt->sopt_valsize < sizeof(*tbl)) {
3986 				error = EINVAL;
3987 				break;
3988 			}
3989 			size = sopt->sopt_valsize;
3990 			tbl = malloc(size, M_TEMP, M_WAITOK);
3991 			error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
3992 			if (error) {
3993 				free(tbl, M_TEMP);
3994 				break;
3995 			}
3996 			tbl->size = (size - sizeof(*tbl)) /
3997 			    sizeof(ipfw_table_entry);
3998 			memset(&ti, 0, sizeof(ti));
3999 			ti.uidx = tbl->tbl;
4000 			IPFW_RLOCK(chain);
4001 			error = ipfw_dump_table_legacy(chain, &ti, tbl);
4002 			IPFW_RUNLOCK(chain);
4003 			if (error) {
4004 				free(tbl, M_TEMP);
4005 				break;
4006 			}
4007 			error = sooptcopyout(sopt, tbl, size);
4008 			free(tbl, M_TEMP);
4009 		}
4010 		break;
4011 
4012 	/*--- NAT operations are protected by the IPFW_LOCK ---*/
4013 	case IP_FW_NAT_CFG:
4014 		if (IPFW_NAT_LOADED)
4015 			error = ipfw_nat_cfg_ptr(sopt);
4016 		else {
4017 			printf("IP_FW_NAT_CFG: %s\n",
4018 			    "ipfw_nat not present, please load it");
4019 			error = EINVAL;
4020 		}
4021 		break;
4022 
4023 	case IP_FW_NAT_DEL:
4024 		if (IPFW_NAT_LOADED)
4025 			error = ipfw_nat_del_ptr(sopt);
4026 		else {
4027 			printf("IP_FW_NAT_DEL: %s\n",
4028 			    "ipfw_nat not present, please load it");
4029 			error = EINVAL;
4030 		}
4031 		break;
4032 
4033 	case IP_FW_NAT_GET_CONFIG:
4034 		if (IPFW_NAT_LOADED)
4035 			error = ipfw_nat_get_cfg_ptr(sopt);
4036 		else {
4037 			printf("IP_FW_NAT_GET_CFG: %s\n",
4038 			    "ipfw_nat not present, please load it");
4039 			error = EINVAL;
4040 		}
4041 		break;
4042 
4043 	case IP_FW_NAT_GET_LOG:
4044 		if (IPFW_NAT_LOADED)
4045 			error = ipfw_nat_get_log_ptr(sopt);
4046 		else {
4047 			printf("IP_FW_NAT_GET_LOG: %s\n",
4048 			    "ipfw_nat not present, please load it");
4049 			error = EINVAL;
4050 		}
4051 		break;
4052 
4053 	default:
4054 		printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
4055 		error = EINVAL;
4056 	}
4057 
4058 	return (error);
4059 #undef RULE_MAXSIZE
4060 }
4061 #define	RULE_MAXSIZE	(256*sizeof(u_int32_t))
4062 
4063 /* Functions to convert rules 7.2 <==> 8.0 */
4064 static int
4065 convert_rule_to_7(struct ip_fw_rule0 *rule)
4066 {
4067 	/* Used to modify original rule */
4068 	struct ip_fw7 *rule7 = (struct ip_fw7 *)rule;
4069 	/* copy of original rule, version 8 */
4070 	struct ip_fw_rule0 *tmp;
4071 
4072 	/* Used to copy commands */
4073 	ipfw_insn *ccmd, *dst;
4074 	int ll = 0, ccmdlen = 0;
4075 
4076 	tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
4077 	if (tmp == NULL) {
4078 		return 1; //XXX error
4079 	}
4080 	bcopy(rule, tmp, RULE_MAXSIZE);
4081 
4082 	/* Copy fields */
4083 	//rule7->_pad = tmp->_pad;
4084 	rule7->set = tmp->set;
4085 	rule7->rulenum = tmp->rulenum;
4086 	rule7->cmd_len = tmp->cmd_len;
4087 	rule7->act_ofs = tmp->act_ofs;
4088 	rule7->next_rule = (struct ip_fw7 *)tmp->next_rule;
4089 	rule7->cmd_len = tmp->cmd_len;
4090 	rule7->pcnt = tmp->pcnt;
4091 	rule7->bcnt = tmp->bcnt;
4092 	rule7->timestamp = tmp->timestamp;
4093 
4094 	/* Copy commands */
4095 	for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule7->cmd ;
4096 			ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
4097 		ccmdlen = F_LEN(ccmd);
4098 
4099 		bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
4100 
4101 		if (dst->opcode > O_NAT)
4102 			/* O_REASS doesn't exists in 7.2 version, so
4103 			 * decrement opcode if it is after O_REASS
4104 			 */
4105 			dst->opcode--;
4106 
4107 		if (ccmdlen > ll) {
4108 			printf("ipfw: opcode %d size truncated\n",
4109 				ccmd->opcode);
4110 			return EINVAL;
4111 		}
4112 	}
4113 	free(tmp, M_TEMP);
4114 
4115 	return 0;
4116 }
4117 
4118 static int
4119 convert_rule_to_8(struct ip_fw_rule0 *rule)
4120 {
4121 	/* Used to modify original rule */
4122 	struct ip_fw7 *rule7 = (struct ip_fw7 *) rule;
4123 
4124 	/* Used to copy commands */
4125 	ipfw_insn *ccmd, *dst;
4126 	int ll = 0, ccmdlen = 0;
4127 
4128 	/* Copy of original rule */
4129 	struct ip_fw7 *tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
4130 	if (tmp == NULL) {
4131 		return 1; //XXX error
4132 	}
4133 
4134 	bcopy(rule7, tmp, RULE_MAXSIZE);
4135 
4136 	for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule->cmd ;
4137 			ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
4138 		ccmdlen = F_LEN(ccmd);
4139 
4140 		bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
4141 
4142 		if (dst->opcode > O_NAT)
4143 			/* O_REASS doesn't exists in 7.2 version, so
4144 			 * increment opcode if it is after O_REASS
4145 			 */
4146 			dst->opcode++;
4147 
4148 		if (ccmdlen > ll) {
4149 			printf("ipfw: opcode %d size truncated\n",
4150 			    ccmd->opcode);
4151 			return EINVAL;
4152 		}
4153 	}
4154 
4155 	rule->_pad = tmp->_pad;
4156 	rule->set = tmp->set;
4157 	rule->rulenum = tmp->rulenum;
4158 	rule->cmd_len = tmp->cmd_len;
4159 	rule->act_ofs = tmp->act_ofs;
4160 	rule->next_rule = (struct ip_fw *)tmp->next_rule;
4161 	rule->cmd_len = tmp->cmd_len;
4162 	rule->id = 0; /* XXX see if is ok = 0 */
4163 	rule->pcnt = tmp->pcnt;
4164 	rule->bcnt = tmp->bcnt;
4165 	rule->timestamp = tmp->timestamp;
4166 
4167 	free (tmp, M_TEMP);
4168 	return 0;
4169 }
4170 
4171 /*
4172  * Named object api
4173  *
4174  */
4175 
4176 void
4177 ipfw_init_srv(struct ip_fw_chain *ch)
4178 {
4179 
4180 	ch->srvmap = ipfw_objhash_create(IPFW_OBJECTS_DEFAULT);
4181 	ch->srvstate = malloc(sizeof(void *) * IPFW_OBJECTS_DEFAULT,
4182 	    M_IPFW, M_WAITOK | M_ZERO);
4183 }
4184 
4185 void
4186 ipfw_destroy_srv(struct ip_fw_chain *ch)
4187 {
4188 
4189 	free(ch->srvstate, M_IPFW);
4190 	ipfw_objhash_destroy(ch->srvmap);
4191 }
4192 
4193 /*
4194  * Allocate new bitmask which can be used to enlarge/shrink
4195  * named instance index.
4196  */
4197 void
4198 ipfw_objhash_bitmap_alloc(uint32_t items, void **idx, int *pblocks)
4199 {
4200 	size_t size;
4201 	int max_blocks;
4202 	u_long *idx_mask;
4203 
4204 	KASSERT((items % BLOCK_ITEMS) == 0,
4205 	   ("bitmask size needs to power of 2 and greater or equal to %zu",
4206 	    BLOCK_ITEMS));
4207 
4208 	max_blocks = items / BLOCK_ITEMS;
4209 	size = items / 8;
4210 	idx_mask = malloc(size * IPFW_MAX_SETS, M_IPFW, M_WAITOK);
4211 	/* Mark all as free */
4212 	memset(idx_mask, 0xFF, size * IPFW_MAX_SETS);
4213 	*idx_mask &= ~(u_long)1; /* Skip index 0 */
4214 
4215 	*idx = idx_mask;
4216 	*pblocks = max_blocks;
4217 }
4218 
4219 /*
4220  * Copy current bitmask index to new one.
4221  */
4222 void
4223 ipfw_objhash_bitmap_merge(struct namedobj_instance *ni, void **idx, int *blocks)
4224 {
4225 	int old_blocks, new_blocks;
4226 	u_long *old_idx, *new_idx;
4227 	int i;
4228 
4229 	old_idx = ni->idx_mask;
4230 	old_blocks = ni->max_blocks;
4231 	new_idx = *idx;
4232 	new_blocks = *blocks;
4233 
4234 	for (i = 0; i < IPFW_MAX_SETS; i++) {
4235 		memcpy(&new_idx[new_blocks * i], &old_idx[old_blocks * i],
4236 		    old_blocks * sizeof(u_long));
4237 	}
4238 }
4239 
4240 /*
4241  * Swaps current @ni index with new one.
4242  */
4243 void
4244 ipfw_objhash_bitmap_swap(struct namedobj_instance *ni, void **idx, int *blocks)
4245 {
4246 	int old_blocks;
4247 	u_long *old_idx;
4248 
4249 	old_idx = ni->idx_mask;
4250 	old_blocks = ni->max_blocks;
4251 
4252 	ni->idx_mask = *idx;
4253 	ni->max_blocks = *blocks;
4254 
4255 	/* Save old values */
4256 	*idx = old_idx;
4257 	*blocks = old_blocks;
4258 }
4259 
4260 void
4261 ipfw_objhash_bitmap_free(void *idx, int blocks)
4262 {
4263 
4264 	free(idx, M_IPFW);
4265 }
4266 
4267 /*
4268  * Creates named hash instance.
4269  * Must be called without holding any locks.
4270  * Return pointer to new instance.
4271  */
4272 struct namedobj_instance *
4273 ipfw_objhash_create(uint32_t items)
4274 {
4275 	struct namedobj_instance *ni;
4276 	int i;
4277 	size_t size;
4278 
4279 	size = sizeof(struct namedobj_instance) +
4280 	    sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE +
4281 	    sizeof(struct namedobjects_head) * NAMEDOBJ_HASH_SIZE;
4282 
4283 	ni = malloc(size, M_IPFW, M_WAITOK | M_ZERO);
4284 	ni->nn_size = NAMEDOBJ_HASH_SIZE;
4285 	ni->nv_size = NAMEDOBJ_HASH_SIZE;
4286 
4287 	ni->names = (struct namedobjects_head *)(ni +1);
4288 	ni->values = &ni->names[ni->nn_size];
4289 
4290 	for (i = 0; i < ni->nn_size; i++)
4291 		TAILQ_INIT(&ni->names[i]);
4292 
4293 	for (i = 0; i < ni->nv_size; i++)
4294 		TAILQ_INIT(&ni->values[i]);
4295 
4296 	/* Set default hashing/comparison functions */
4297 	ni->hash_f = objhash_hash_name;
4298 	ni->cmp_f = objhash_cmp_name;
4299 
4300 	/* Allocate bitmask separately due to possible resize */
4301 	ipfw_objhash_bitmap_alloc(items, (void*)&ni->idx_mask, &ni->max_blocks);
4302 
4303 	return (ni);
4304 }
4305 
4306 void
4307 ipfw_objhash_destroy(struct namedobj_instance *ni)
4308 {
4309 
4310 	free(ni->idx_mask, M_IPFW);
4311 	free(ni, M_IPFW);
4312 }
4313 
4314 void
4315 ipfw_objhash_set_funcs(struct namedobj_instance *ni, objhash_hash_f *hash_f,
4316     objhash_cmp_f *cmp_f)
4317 {
4318 
4319 	ni->hash_f = hash_f;
4320 	ni->cmp_f = cmp_f;
4321 }
4322 
4323 static uint32_t
4324 objhash_hash_name(struct namedobj_instance *ni, const void *name, uint32_t set)
4325 {
4326 
4327 	return (fnv_32_str((const char *)name, FNV1_32_INIT));
4328 }
4329 
4330 static int
4331 objhash_cmp_name(struct named_object *no, const void *name, uint32_t set)
4332 {
4333 
4334 	if ((strcmp(no->name, (const char *)name) == 0) && (no->set == set))
4335 		return (0);
4336 
4337 	return (1);
4338 }
4339 
4340 static uint32_t
4341 objhash_hash_idx(struct namedobj_instance *ni, uint32_t val)
4342 {
4343 	uint32_t v;
4344 
4345 	v = val % (ni->nv_size - 1);
4346 
4347 	return (v);
4348 }
4349 
4350 struct named_object *
4351 ipfw_objhash_lookup_name(struct namedobj_instance *ni, uint32_t set, char *name)
4352 {
4353 	struct named_object *no;
4354 	uint32_t hash;
4355 
4356 	hash = ni->hash_f(ni, name, set) % ni->nn_size;
4357 
4358 	TAILQ_FOREACH(no, &ni->names[hash], nn_next) {
4359 		if (ni->cmp_f(no, name, set) == 0)
4360 			return (no);
4361 	}
4362 
4363 	return (NULL);
4364 }
4365 
4366 /*
4367  * Find named object by @uid.
4368  * Check @tlvs for valid data inside.
4369  *
4370  * Returns pointer to found TLV or NULL.
4371  */
4372 ipfw_obj_ntlv *
4373 ipfw_find_name_tlv_type(void *tlvs, int len, uint16_t uidx, uint32_t etlv)
4374 {
4375 	ipfw_obj_ntlv *ntlv;
4376 	uintptr_t pa, pe;
4377 	int l;
4378 
4379 	pa = (uintptr_t)tlvs;
4380 	pe = pa + len;
4381 	l = 0;
4382 	for (; pa < pe; pa += l) {
4383 		ntlv = (ipfw_obj_ntlv *)pa;
4384 		l = ntlv->head.length;
4385 
4386 		if (l != sizeof(*ntlv))
4387 			return (NULL);
4388 
4389 		if (ntlv->idx != uidx)
4390 			continue;
4391 		/*
4392 		 * When userland has specified zero TLV type, do
4393 		 * not compare it with eltv. In some cases userland
4394 		 * doesn't know what type should it have. Use only
4395 		 * uidx and name for search named_object.
4396 		 */
4397 		if (ntlv->head.type != 0 &&
4398 		    ntlv->head.type != (uint16_t)etlv)
4399 			continue;
4400 
4401 		if (ipfw_check_object_name_generic(ntlv->name) != 0)
4402 			return (NULL);
4403 
4404 		return (ntlv);
4405 	}
4406 
4407 	return (NULL);
4408 }
4409 
4410 /*
4411  * Finds object config based on either legacy index
4412  * or name in ntlv.
4413  * Note @ti structure contains unchecked data from userland.
4414  *
4415  * Returns 0 in success and fills in @pno with found config
4416  */
4417 int
4418 ipfw_objhash_find_type(struct namedobj_instance *ni, struct tid_info *ti,
4419     uint32_t etlv, struct named_object **pno)
4420 {
4421 	char *name;
4422 	ipfw_obj_ntlv *ntlv;
4423 	uint32_t set;
4424 
4425 	if (ti->tlvs == NULL)
4426 		return (EINVAL);
4427 
4428 	ntlv = ipfw_find_name_tlv_type(ti->tlvs, ti->tlen, ti->uidx, etlv);
4429 	if (ntlv == NULL)
4430 		return (EINVAL);
4431 	name = ntlv->name;
4432 
4433 	/*
4434 	 * Use set provided by @ti instead of @ntlv one.
4435 	 * This is needed due to different sets behavior
4436 	 * controlled by V_fw_tables_sets.
4437 	 */
4438 	set = ti->set;
4439 	*pno = ipfw_objhash_lookup_name(ni, set, name);
4440 	if (*pno == NULL)
4441 		return (ESRCH);
4442 	return (0);
4443 }
4444 
4445 /*
4446  * Find named object by name, considering also its TLV type.
4447  */
4448 struct named_object *
4449 ipfw_objhash_lookup_name_type(struct namedobj_instance *ni, uint32_t set,
4450     uint32_t type, const char *name)
4451 {
4452 	struct named_object *no;
4453 	uint32_t hash;
4454 
4455 	hash = ni->hash_f(ni, name, set) % ni->nn_size;
4456 
4457 	TAILQ_FOREACH(no, &ni->names[hash], nn_next) {
4458 		if (ni->cmp_f(no, name, set) == 0 &&
4459 		    no->etlv == (uint16_t)type)
4460 			return (no);
4461 	}
4462 
4463 	return (NULL);
4464 }
4465 
4466 struct named_object *
4467 ipfw_objhash_lookup_kidx(struct namedobj_instance *ni, uint16_t kidx)
4468 {
4469 	struct named_object *no;
4470 	uint32_t hash;
4471 
4472 	hash = objhash_hash_idx(ni, kidx);
4473 
4474 	TAILQ_FOREACH(no, &ni->values[hash], nv_next) {
4475 		if (no->kidx == kidx)
4476 			return (no);
4477 	}
4478 
4479 	return (NULL);
4480 }
4481 
4482 int
4483 ipfw_objhash_same_name(struct namedobj_instance *ni, struct named_object *a,
4484     struct named_object *b)
4485 {
4486 
4487 	if ((strcmp(a->name, b->name) == 0) && a->set == b->set)
4488 		return (1);
4489 
4490 	return (0);
4491 }
4492 
4493 void
4494 ipfw_objhash_add(struct namedobj_instance *ni, struct named_object *no)
4495 {
4496 	uint32_t hash;
4497 
4498 	hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size;
4499 	TAILQ_INSERT_HEAD(&ni->names[hash], no, nn_next);
4500 
4501 	hash = objhash_hash_idx(ni, no->kidx);
4502 	TAILQ_INSERT_HEAD(&ni->values[hash], no, nv_next);
4503 
4504 	ni->count++;
4505 }
4506 
4507 void
4508 ipfw_objhash_del(struct namedobj_instance *ni, struct named_object *no)
4509 {
4510 	uint32_t hash;
4511 
4512 	hash = ni->hash_f(ni, no->name, no->set) % ni->nn_size;
4513 	TAILQ_REMOVE(&ni->names[hash], no, nn_next);
4514 
4515 	hash = objhash_hash_idx(ni, no->kidx);
4516 	TAILQ_REMOVE(&ni->values[hash], no, nv_next);
4517 
4518 	ni->count--;
4519 }
4520 
4521 uint32_t
4522 ipfw_objhash_count(struct namedobj_instance *ni)
4523 {
4524 
4525 	return (ni->count);
4526 }
4527 
4528 uint32_t
4529 ipfw_objhash_count_type(struct namedobj_instance *ni, uint16_t type)
4530 {
4531 	struct named_object *no;
4532 	uint32_t count;
4533 	int i;
4534 
4535 	count = 0;
4536 	for (i = 0; i < ni->nn_size; i++) {
4537 		TAILQ_FOREACH(no, &ni->names[i], nn_next) {
4538 			if (no->etlv == type)
4539 				count++;
4540 		}
4541 	}
4542 	return (count);
4543 }
4544 
4545 /*
4546  * Runs @func for each found named object.
4547  * It is safe to delete objects from callback
4548  */
4549 int
4550 ipfw_objhash_foreach(struct namedobj_instance *ni, objhash_cb_t *f, void *arg)
4551 {
4552 	struct named_object *no, *no_tmp;
4553 	int i, ret;
4554 
4555 	for (i = 0; i < ni->nn_size; i++) {
4556 		TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) {
4557 			ret = f(ni, no, arg);
4558 			if (ret != 0)
4559 				return (ret);
4560 		}
4561 	}
4562 	return (0);
4563 }
4564 
4565 /*
4566  * Runs @f for each found named object with type @type.
4567  * It is safe to delete objects from callback
4568  */
4569 int
4570 ipfw_objhash_foreach_type(struct namedobj_instance *ni, objhash_cb_t *f,
4571     void *arg, uint16_t type)
4572 {
4573 	struct named_object *no, *no_tmp;
4574 	int i, ret;
4575 
4576 	for (i = 0; i < ni->nn_size; i++) {
4577 		TAILQ_FOREACH_SAFE(no, &ni->names[i], nn_next, no_tmp) {
4578 			if (no->etlv != type)
4579 				continue;
4580 			ret = f(ni, no, arg);
4581 			if (ret != 0)
4582 				return (ret);
4583 		}
4584 	}
4585 	return (0);
4586 }
4587 
4588 /*
4589  * Removes index from given set.
4590  * Returns 0 on success.
4591  */
4592 int
4593 ipfw_objhash_free_idx(struct namedobj_instance *ni, uint16_t idx)
4594 {
4595 	u_long *mask;
4596 	int i, v;
4597 
4598 	i = idx / BLOCK_ITEMS;
4599 	v = idx % BLOCK_ITEMS;
4600 
4601 	if (i >= ni->max_blocks)
4602 		return (1);
4603 
4604 	mask = &ni->idx_mask[i];
4605 
4606 	if ((*mask & ((u_long)1 << v)) != 0)
4607 		return (1);
4608 
4609 	/* Mark as free */
4610 	*mask |= (u_long)1 << v;
4611 
4612 	/* Update free offset */
4613 	if (ni->free_off[0] > i)
4614 		ni->free_off[0] = i;
4615 
4616 	return (0);
4617 }
4618 
4619 /*
4620  * Allocate new index in given instance and stores in in @pidx.
4621  * Returns 0 on success.
4622  */
4623 int
4624 ipfw_objhash_alloc_idx(void *n, uint16_t *pidx)
4625 {
4626 	struct namedobj_instance *ni;
4627 	u_long *mask;
4628 	int i, off, v;
4629 
4630 	ni = (struct namedobj_instance *)n;
4631 
4632 	off = ni->free_off[0];
4633 	mask = &ni->idx_mask[off];
4634 
4635 	for (i = off; i < ni->max_blocks; i++, mask++) {
4636 		if ((v = ffsl(*mask)) == 0)
4637 			continue;
4638 
4639 		/* Mark as busy */
4640 		*mask &= ~ ((u_long)1 << (v - 1));
4641 
4642 		ni->free_off[0] = i;
4643 
4644 		v = BLOCK_ITEMS * i + v - 1;
4645 
4646 		*pidx = v;
4647 		return (0);
4648 	}
4649 
4650 	return (1);
4651 }
4652 
4653 /* end of file */
4654