1/*-
2 * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
3 *
4 * Supported by: Valeria Paoli
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28#include <sys/cdefs.h>
29__FBSDID("$FreeBSD: head/sys/netinet/ipfw/ip_fw_sockopt.c 225518 2011-09-12 21:09:56Z jhb $");
30
31/*
32 * Sockopt support for ipfw. The routines here implement
33 * the upper half of the ipfw code.
34 */
35
36#include "opt_ipfw.h"
37#if !defined(KLD_MODULE)
38#include "opt_ipdivert.h"
39#include "opt_ipdn.h"
40#include "opt_inet.h"
41#ifndef INET
42#error IPFIREWALL requires INET.
43#endif /* INET */
44#endif
45#include "opt_inet6.h"
46#include "opt_ipsec.h"
47
48#include <sys/param.h>
49#include <sys/systm.h>
50#include <sys/malloc.h>
51#include <sys/mbuf.h> /* struct m_tag used by nested headers */
52#include <sys/kernel.h>
53#include <sys/lock.h>
54#include <sys/priv.h>
55#include <sys/proc.h>
56#include <sys/rwlock.h>
57#include <sys/socket.h>
58#include <sys/socketvar.h>
59#include <sys/sysctl.h>
60#include <sys/syslog.h>
61#include <net/if.h>
62#include <net/route.h>
63#include <net/vnet.h>
64
65#include <netinet/in.h>
66#include <netinet/ip_var.h> /* hooks */
67#include <netinet/ip_fw.h>
68#include <netinet/ipfw/ip_fw_private.h>
69
70#ifdef MAC
71#include <security/mac/mac_framework.h>
72#endif
73
74MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
75
76/*
77 * static variables followed by global ones (none in this file)
78 */
79
80/*
81 * Find the smallest rule >= key, id.
82 * We could use bsearch but it is so simple that we code it directly
83 */
84int
85ipfw_find_rule(struct ip_fw_chain *chain, uint32_t key, uint32_t id)
86{
87 int i, lo, hi;
88 struct ip_fw *r;
89
90 for (lo = 0, hi = chain->n_rules - 1; lo < hi;) {
91 i = (lo + hi) / 2;
92 r = chain->map[i];
93 if (r->rulenum < key)
94 lo = i + 1; /* continue from the next one */
95 else if (r->rulenum > key)
96 hi = i; /* this might be good */
97 else if (r->id < id)
98 lo = i + 1; /* continue from the next one */
99 else /* r->id >= id */
100 hi = i; /* this might be good */
101 };
102 return hi;
103}
104
105/*
106 * allocate a new map, returns the chain locked. extra is the number
107 * of entries to add or delete.
108 */
109static struct ip_fw **
110get_map(struct ip_fw_chain *chain, int extra, int locked)
111{
112
113 for (;;) {
114 struct ip_fw **map;
115 int i;
116
117 i = chain->n_rules + extra;
118 map = malloc(i * sizeof(struct ip_fw *), M_IPFW,
119 locked ? M_NOWAIT : M_WAITOK);
120 if (map == NULL) {
121 printf("%s: cannot allocate map\n", __FUNCTION__);
122 return NULL;
123 }
124 if (!locked)
125 IPFW_UH_WLOCK(chain);
126 if (i >= chain->n_rules + extra) /* good */
127 return map;
128 /* otherwise we lost the race, free and retry */
129 if (!locked)
130 IPFW_UH_WUNLOCK(chain);
131 free(map, M_IPFW);
132 }
133}
134
135/*
136 * swap the maps. It is supposed to be called with IPFW_UH_WLOCK
137 */
138static struct ip_fw **
139swap_map(struct ip_fw_chain *chain, struct ip_fw **new_map, int new_len)
140{
141 struct ip_fw **old_map;
142
143 IPFW_WLOCK(chain);
144 chain->id++;
145 chain->n_rules = new_len;
146 old_map = chain->map;
147 chain->map = new_map;
148 IPFW_WUNLOCK(chain);
149 return old_map;
150}
151
152/*
153 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
154 * possibly create a rule number and add the rule to the list.
155 * Update the rule_number in the input struct so the caller knows it as well.
156 * XXX DO NOT USE FOR THE DEFAULT RULE.
157 * Must be called without IPFW_UH held
158 */
159int
160ipfw_add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule)
161{
162 struct ip_fw *rule;
163 int i, l, insert_before;
164 struct ip_fw **map; /* the new array of pointers */
165
166 if (chain->rules == NULL || input_rule->rulenum > IPFW_DEFAULT_RULE-1)
167 return (EINVAL);
168
169 l = RULESIZE(input_rule);
170 rule = malloc(l, M_IPFW, M_WAITOK | M_ZERO);
171 if (rule == NULL)
172 return (ENOSPC);
173 /* get_map returns with IPFW_UH_WLOCK if successful */
174 map = get_map(chain, 1, 0 /* not locked */);
175 if (map == NULL) {
176 free(rule, M_IPFW);
177 return ENOSPC;
178 }
179
180 bcopy(input_rule, rule, l);
181 /* clear fields not settable from userland */
182 rule->x_next = NULL;
183 rule->next_rule = NULL;
184 rule->pcnt = 0;
185 rule->bcnt = 0;
186 rule->timestamp = 0;
187
188 if (V_autoinc_step < 1)
189 V_autoinc_step = 1;
190 else if (V_autoinc_step > 1000)
191 V_autoinc_step = 1000;
192 /* find the insertion point, we will insert before */
193 insert_before = rule->rulenum ? rule->rulenum + 1 : IPFW_DEFAULT_RULE;
194 i = ipfw_find_rule(chain, insert_before, 0);
195 /* duplicate first part */
196 if (i > 0)
197 bcopy(chain->map, map, i * sizeof(struct ip_fw *));
198 map[i] = rule;
199 /* duplicate remaining part, we always have the default rule */
200 bcopy(chain->map + i, map + i + 1,
201 sizeof(struct ip_fw *) *(chain->n_rules - i));
202 if (rule->rulenum == 0) {
203 /* write back the number */
204 rule->rulenum = i > 0 ? map[i-1]->rulenum : 0;
205 if (rule->rulenum < IPFW_DEFAULT_RULE - V_autoinc_step)
206 rule->rulenum += V_autoinc_step;
207 input_rule->rulenum = rule->rulenum;
208 }
209
210 rule->id = chain->id + 1;
211 map = swap_map(chain, map, chain->n_rules + 1);
212 chain->static_len += l;
213 IPFW_UH_WUNLOCK(chain);
214 if (map)
215 free(map, M_IPFW);
216 return (0);
217}
218
219/*
220 * Reclaim storage associated with a list of rules. This is
221 * typically the list created using remove_rule.
222 * A NULL pointer on input is handled correctly.
223 */
224void
225ipfw_reap_rules(struct ip_fw *head)
226{
227 struct ip_fw *rule;
228
229 while ((rule = head) != NULL) {
230 head = head->x_next;
231 free(rule, M_IPFW);
232 }
233}
234
235/*
236 * Used by del_entry() to check if a rule should be kept.
237 * Returns 1 if the rule must be kept, 0 otherwise.
238 *
239 * Called with cmd = {0,1,5}.
240 * cmd == 0 matches on rule numbers, excludes rules in RESVD_SET if n == 0 ;
241 * cmd == 1 matches on set numbers only, rule numbers are ignored;
242 * cmd == 5 matches on rule and set numbers.
243 *
244 * n == 0 is a wildcard for rule numbers, there is no wildcard for sets.
245 *
246 * Rules to keep are
247 * (default || reserved || !match_set || !match_number)
248 * where
249 * default ::= (rule->rulenum == IPFW_DEFAULT_RULE)
250 * // the default rule is always protected
251 *
252 * reserved ::= (cmd == 0 && n == 0 && rule->set == RESVD_SET)
253 * // RESVD_SET is protected only if cmd == 0 and n == 0 ("ipfw flush")
254 *
255 * match_set ::= (cmd == 0 || rule->set == set)
256 * // set number is ignored for cmd == 0
257 *
258 * match_number ::= (cmd == 1 || n == 0 || n == rule->rulenum)
259 * // number is ignored for cmd == 1 or n == 0
260 *
261 */
262static int
263keep_rule(struct ip_fw *rule, uint8_t cmd, uint8_t set, uint32_t n)
264{
265 return
266 (rule->rulenum == IPFW_DEFAULT_RULE) ||
267 (cmd == 0 && n == 0 && rule->set == RESVD_SET) ||
268 !(cmd == 0 || rule->set == set) ||
269 !(cmd == 1 || n == 0 || n == rule->rulenum);
270}
271
272/**
273 * Remove all rules with given number, or do set manipulation.
274 * Assumes chain != NULL && *chain != NULL.
275 *
276 * The argument is an uint32_t. The low 16 bit are the rule or set number;
277 * the next 8 bits are the new set; the top 8 bits indicate the command:
278 *
279 * 0 delete rules numbered "rulenum"
280 * 1 delete rules in set "rulenum"
281 * 2 move rules "rulenum" to set "new_set"
282 * 3 move rules from set "rulenum" to set "new_set"
283 * 4 swap sets "rulenum" and "new_set"
284 * 5 delete rules "rulenum" and set "new_set"
285 */
286static int
287del_entry(struct ip_fw_chain *chain, uint32_t arg)
288{
289 struct ip_fw *rule;
290 uint32_t num; /* rule number or old_set */
291 uint8_t cmd, new_set;
292 int start, end, i, ofs, n;
293 struct ip_fw **map = NULL;
294 int error = 0;
295
296 num = arg & 0xffff;
297 cmd = (arg >> 24) & 0xff;
298 new_set = (arg >> 16) & 0xff;
299
300 if (cmd > 5 || new_set > RESVD_SET)
301 return EINVAL;
302 if (cmd == 0 || cmd == 2 || cmd == 5) {
303 if (num >= IPFW_DEFAULT_RULE)
304 return EINVAL;
305 } else {
306 if (num > RESVD_SET) /* old_set */
307 return EINVAL;
308 }
309
310 IPFW_UH_WLOCK(chain); /* arbitrate writers */
311 chain->reap = NULL; /* prepare for deletions */
312
313 switch (cmd) {
314 case 0: /* delete rules "num" (num == 0 matches all) */
315 case 1: /* delete all rules in set N */
316 case 5: /* delete rules with number N and set "new_set". */
317
318 /*
319 * Locate first rule to delete (start), the rule after
320 * the last one to delete (end), and count how many
321 * rules to delete (n). Always use keep_rule() to
322 * determine which rules to keep.
323 */
324 n = 0;
325 if (cmd == 1) {
326 /* look for a specific set including RESVD_SET.
327 * Must scan the entire range, ignore num.
328 */
329 new_set = num;
330 for (start = -1, end = i = 0; i < chain->n_rules; i++) {
331 if (keep_rule(chain->map[i], cmd, new_set, 0))
332 continue;
333 if (start < 0)
334 start = i;
335 end = i;
336 n++;
337 }
338 end++; /* first non-matching */
339 } else {
340 /* Optimized search on rule numbers */
341 start = ipfw_find_rule(chain, num, 0);
342 for (end = start; end < chain->n_rules; end++) {
343 rule = chain->map[end];
344 if (num > 0 && rule->rulenum != num)
345 break;
346 if (!keep_rule(rule, cmd, new_set, num))
347 n++;
348 }
349 }
350
351 if (n == 0) {
352 /* A flush request (arg == 0 or cmd == 1) on empty
353 * ruleset returns with no error. On the contrary,
354 * if there is no match on a specific request,
355 * we return EINVAL.
356 */
357 if (arg != 0 && cmd != 1)
358 error = EINVAL;
359 break;
360 }
361
362 /* We have something to delete. Allocate the new map */
363 map = get_map(chain, -n, 1 /* locked */);
364 if (map == NULL) {
365 error = EINVAL;
366 break;
367 }
368
369 /* 1. bcopy the initial part of the map */
370 if (start > 0)
371 bcopy(chain->map, map, start * sizeof(struct ip_fw *));
372 /* 2. copy active rules between start and end */
373 for (i = ofs = start; i < end; i++) {
374 rule = chain->map[i];
375 if (keep_rule(rule, cmd, new_set, num))
376 map[ofs++] = rule;
377 }
378 /* 3. copy the final part of the map */
379 bcopy(chain->map + end, map + ofs,
380 (chain->n_rules - end) * sizeof(struct ip_fw *));
381 /* 4. swap the maps (under BH_LOCK) */
382 map = swap_map(chain, map, chain->n_rules - n);
383 /* 5. now remove the rules deleted from the old map */
384 for (i = start; i < end; i++) {
385 int l;
386 rule = map[i];
387 if (keep_rule(rule, cmd, new_set, num))
388 continue;
389 l = RULESIZE(rule);
390 chain->static_len -= l;
391 ipfw_remove_dyn_children(rule);
392 rule->x_next = chain->reap;
393 chain->reap = rule;
394 }
395 break;
396
397 /*
398 * In the next 3 cases the loop stops at (n_rules - 1)
399 * because the default rule is never eligible..
400 */
401
402 case 2: /* move rules with given RULE number to new set */
403 for (i = 0; i < chain->n_rules - 1; i++) {
404 rule = chain->map[i];
405 if (rule->rulenum == num)
406 rule->set = new_set;
407 }
408 break;
409
410 case 3: /* move rules with given SET number to new set */
411 for (i = 0; i < chain->n_rules - 1; i++) {
412 rule = chain->map[i];
413 if (rule->set == num)
414 rule->set = new_set;
415 }
416 break;
417
418 case 4: /* swap two sets */
419 for (i = 0; i < chain->n_rules - 1; i++) {
420 rule = chain->map[i];
421 if (rule->set == num)
422 rule->set = new_set;
423 else if (rule->set == new_set)
424 rule->set = num;
425 }
426 break;
427 }
428
429 rule = chain->reap;
430 chain->reap = NULL;
431 IPFW_UH_WUNLOCK(chain);
432 ipfw_reap_rules(rule);
433 if (map)
434 free(map, M_IPFW);
435 return error;
436}
437
438/*
439 * Clear counters for a specific rule.
440 * Normally run under IPFW_UH_RLOCK, but these are idempotent ops
441 * so we only care that rules do not disappear.
442 */
443static void
444clear_counters(struct ip_fw *rule, int log_only)
445{
446 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
447
448 if (log_only == 0) {
449 rule->bcnt = rule->pcnt = 0;
450 rule->timestamp = 0;
451 }
452 if (l->o.opcode == O_LOG)
453 l->log_left = l->max_log;
454}
455
456/**
457 * Reset some or all counters on firewall rules.
458 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
459 * the next 8 bits are the set number, the top 8 bits are the command:
460 * 0 work with rules from all set's;
461 * 1 work with rules only from specified set.
462 * Specified rule number is zero if we want to clear all entries.
463 * log_only is 1 if we only want to reset logs, zero otherwise.
464 */
465static int
466zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
467{
468 struct ip_fw *rule;
469 char *msg;
470 int i;
471
472 uint16_t rulenum = arg & 0xffff;
473 uint8_t set = (arg >> 16) & 0xff;
474 uint8_t cmd = (arg >> 24) & 0xff;
475
476 if (cmd > 1)
477 return (EINVAL);
478 if (cmd == 1 && set > RESVD_SET)
479 return (EINVAL);
480
481 IPFW_UH_RLOCK(chain);
482 if (rulenum == 0) {
483 V_norule_counter = 0;
484 for (i = 0; i < chain->n_rules; i++) {
485 rule = chain->map[i];
486 /* Skip rules not in our set. */
487 if (cmd == 1 && rule->set != set)
488 continue;
489 clear_counters(rule, log_only);
490 }
491 msg = log_only ? "All logging counts reset" :
492 "Accounting cleared";
493 } else {
494 int cleared = 0;
495 for (i = 0; i < chain->n_rules; i++) {
496 rule = chain->map[i];
497 if (rule->rulenum == rulenum) {
498 if (cmd == 0 || rule->set == set)
499 clear_counters(rule, log_only);
500 cleared = 1;
501 }
502 if (rule->rulenum > rulenum)
503 break;
504 }
505 if (!cleared) { /* we did not find any matching rules */
506 IPFW_UH_RUNLOCK(chain);
507 return (EINVAL);
508 }
509 msg = log_only ? "logging count reset" : "cleared";
510 }
511 IPFW_UH_RUNLOCK(chain);
512
513 if (V_fw_verbose) {
514 int lev = LOG_SECURITY | LOG_NOTICE;
515
516 if (rulenum)
517 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg);
518 else
519 log(lev, "ipfw: %s.\n", msg);
520 }
521 return (0);
522}
523
524/*
525 * Check validity of the structure before insert.
526 * Rules are simple, so this mostly need to check rule sizes.
527 */
528static int
529check_ipfw_struct(struct ip_fw *rule, int size)
530{
531 int l, cmdlen = 0;
532 int have_action=0;
533 ipfw_insn *cmd;
534
535 if (size < sizeof(*rule)) {
536 printf("ipfw: rule too short\n");
537 return (EINVAL);
538 }
539 /* first, check for valid size */
540 l = RULESIZE(rule);
541 if (l != size) {
542 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
543 return (EINVAL);
544 }
545 if (rule->act_ofs >= rule->cmd_len) {
546 printf("ipfw: bogus action offset (%u > %u)\n",
547 rule->act_ofs, rule->cmd_len - 1);
548 return (EINVAL);
549 }
550 /*
551 * Now go for the individual checks. Very simple ones, basically only
552 * instruction sizes.
553 */
554 for (l = rule->cmd_len, cmd = rule->cmd ;
555 l > 0 ; l -= cmdlen, cmd += cmdlen) {
556 cmdlen = F_LEN(cmd);
557 if (cmdlen > l) {
558 printf("ipfw: opcode %d size truncated\n",
559 cmd->opcode);
560 return EINVAL;
561 }
562 switch (cmd->opcode) {
563 case O_PROBE_STATE:
564 case O_KEEP_STATE:
565 case O_PROTO:
566 case O_IP_SRC_ME:
567 case O_IP_DST_ME:
568 case O_LAYER2:
569 case O_IN:
570 case O_FRAG:
571 case O_DIVERTED:
572 case O_IPOPT:
573 case O_IPTOS:
574 case O_IPPRECEDENCE:
575 case O_IPVER:
576 case O_SOCKARG:
577 case O_TCPWIN:
578 case O_TCPFLAGS:
579 case O_TCPOPTS:
580 case O_ESTAB:
581 case O_VERREVPATH:
582 case O_VERSRCREACH:
583 case O_ANTISPOOF:
584 case O_IPSEC:
585#ifdef INET6
586 case O_IP6_SRC_ME:
587 case O_IP6_DST_ME:
588 case O_EXT_HDR:
589 case O_IP6:
590#endif
591 case O_IP4:
592 case O_TAG:
593 if (cmdlen != F_INSN_SIZE(ipfw_insn))
594 goto bad_size;
595 break;
596
597 case O_FIB:
598 if (cmdlen != F_INSN_SIZE(ipfw_insn))
599 goto bad_size;
600 if (cmd->arg1 >= rt_numfibs) {
601 printf("ipfw: invalid fib number %d\n",
602 cmd->arg1);
603 return EINVAL;
604 }
605 break;
606
607 case O_SETFIB:
608 if (cmdlen != F_INSN_SIZE(ipfw_insn))
609 goto bad_size;
610 if ((cmd->arg1 != IP_FW_TABLEARG) &&
611 (cmd->arg1 >= rt_numfibs)) {
612 printf("ipfw: invalid fib number %d\n",
613 cmd->arg1);
614 return EINVAL;
615 }
616 goto check_action;
617
618 case O_UID:
619 case O_GID:
620 case O_JAIL:
621 case O_IP_SRC:
622 case O_IP_DST:
623 case O_TCPSEQ:
624 case O_TCPACK:
625 case O_PROB:
626 case O_ICMPTYPE:
627 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
628 goto bad_size;
629 break;
630
631 case O_LIMIT:
632 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
633 goto bad_size;
634 break;
635
636 case O_LOG:
637 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
638 goto bad_size;
639
640 ((ipfw_insn_log *)cmd)->log_left =
641 ((ipfw_insn_log *)cmd)->max_log;
642
643 break;
644
645 case O_IP_SRC_MASK:
646 case O_IP_DST_MASK:
647 /* only odd command lengths */
648 if ( !(cmdlen & 1) || cmdlen > 31)
649 goto bad_size;
650 break;
651
652 case O_IP_SRC_SET:
653 case O_IP_DST_SET:
654 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
655 printf("ipfw: invalid set size %d\n",
656 cmd->arg1);
657 return EINVAL;
658 }
659 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
660 (cmd->arg1+31)/32 )
661 goto bad_size;
662 break;
663
664 case O_IP_SRC_LOOKUP:
665 case O_IP_DST_LOOKUP:
666 if (cmd->arg1 >= IPFW_TABLES_MAX) {
667 printf("ipfw: invalid table number %d\n",
668 cmd->arg1);
669 return (EINVAL);
670 }
671 if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
672 cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1 &&
673 cmdlen != F_INSN_SIZE(ipfw_insn_u32))
674 goto bad_size;
675 break;
676
677 case O_MACADDR2:
678 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
679 goto bad_size;
680 break;
681
682 case O_NOP:
683 case O_IPID:
684 case O_IPTTL:
685 case O_IPLEN:
686 case O_TCPDATALEN:
687 case O_TAGGED:
688 if (cmdlen < 1 || cmdlen > 31)
689 goto bad_size;
690 break;
691
692 case O_MAC_TYPE:
693 case O_IP_SRCPORT:
694 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
695 if (cmdlen < 2 || cmdlen > 31)
696 goto bad_size;
697 break;
698
699 case O_RECV:
700 case O_XMIT:
701 case O_VIA:
702 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
703 goto bad_size;
704 break;
705
706 case O_ALTQ:
707 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
708 goto bad_size;
709 break;
710
711 case O_PIPE:
712 case O_QUEUE:
713 if (cmdlen != F_INSN_SIZE(ipfw_insn))
714 goto bad_size;
715 goto check_action;
716
717 case O_FORWARD_IP:
718#ifdef IPFIREWALL_FORWARD
719 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
720 goto bad_size;
721 goto check_action;
722#else
723 return EINVAL;
724#endif
725
726#ifdef INET6
727 case O_FORWARD_IP6:
728#ifdef IPFIREWALL_FORWARD
729 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa6))
730 goto bad_size;
731 goto check_action;
732#else
733 return (EINVAL);
734#endif
735#endif /* INET6 */
736
737 case O_DIVERT:
738 case O_TEE:
739 if (ip_divert_ptr == NULL)
740 return EINVAL;
741 else
742 goto check_size;
743 case O_NETGRAPH:
744 case O_NGTEE:
745 if (ng_ipfw_input_p == NULL)
746 return EINVAL;
747 else
748 goto check_size;
749 case O_NAT:
750 if (!IPFW_NAT_LOADED)
751 return EINVAL;
752 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
753 goto bad_size;
754 goto check_action;
755 case O_FORWARD_MAC: /* XXX not implemented yet */
756 case O_CHECK_STATE:
757 case O_COUNT:
758 case O_ACCEPT:
759 case O_DENY:
760 case O_REJECT:
761#ifdef INET6
762 case O_UNREACH6:
763#endif
764 case O_SKIPTO:
765 case O_REASS:
766 case O_CALLRETURN:
767check_size:
768 if (cmdlen != F_INSN_SIZE(ipfw_insn))
769 goto bad_size;
770check_action:
771 if (have_action) {
772 printf("ipfw: opcode %d, multiple actions"
773 " not allowed\n",
774 cmd->opcode);
775 return EINVAL;
776 }
777 have_action = 1;
778 if (l != cmdlen) {
779 printf("ipfw: opcode %d, action must be"
780 " last opcode\n",
781 cmd->opcode);
782 return EINVAL;
783 }
784 break;
785#ifdef INET6
786 case O_IP6_SRC:
787 case O_IP6_DST:
788 if (cmdlen != F_INSN_SIZE(struct in6_addr) +
789 F_INSN_SIZE(ipfw_insn))
790 goto bad_size;
791 break;
792
793 case O_FLOW6ID:
794 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
795 ((ipfw_insn_u32 *)cmd)->o.arg1)
796 goto bad_size;
797 break;
798
799 case O_IP6_SRC_MASK:
800 case O_IP6_DST_MASK:
801 if ( !(cmdlen & 1) || cmdlen > 127)
802 goto bad_size;
803 break;
804 case O_ICMP6TYPE:
805 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
806 goto bad_size;
807 break;
808#endif
809
810 default:
811 switch (cmd->opcode) {
812#ifndef INET6
813 case O_IP6_SRC_ME:
814 case O_IP6_DST_ME:
815 case O_EXT_HDR:
816 case O_IP6:
817 case O_UNREACH6:
818 case O_IP6_SRC:
819 case O_IP6_DST:
820 case O_FLOW6ID:
821 case O_IP6_SRC_MASK:
822 case O_IP6_DST_MASK:
823 case O_ICMP6TYPE:
824 printf("ipfw: no IPv6 support in kernel\n");
825 return EPROTONOSUPPORT;
826#endif
827 default:
828 printf("ipfw: opcode %d, unknown opcode\n",
829 cmd->opcode);
830 return EINVAL;
831 }
832 }
833 }
834 if (have_action == 0) {
835 printf("ipfw: missing action\n");
836 return EINVAL;
837 }
838 return 0;
839
840bad_size:
841 printf("ipfw: opcode %d size %d wrong\n",
842 cmd->opcode, cmdlen);
843 return EINVAL;
844}
845
846
847/*
848 * Translation of requests for compatibility with FreeBSD 7.2/8.
849 * a static variable tells us if we have an old client from userland,
850 * and if necessary we translate requests and responses between the
851 * two formats.
852 */
853static int is7 = 0;
854
855struct ip_fw7 {
856 struct ip_fw7 *next; /* linked list of rules */
857 struct ip_fw7 *next_rule; /* ptr to next [skipto] rule */
858 /* 'next_rule' is used to pass up 'set_disable' status */
859
860 uint16_t act_ofs; /* offset of action in 32-bit units */
861 uint16_t cmd_len; /* # of 32-bit words in cmd */
862 uint16_t rulenum; /* rule number */
863 uint8_t set; /* rule set (0..31) */
864 // #define RESVD_SET 31 /* set for default and persistent rules */
865 uint8_t _pad; /* padding */
866 // uint32_t id; /* rule id, only in v.8 */
867 /* These fields are present in all rules. */
868 uint64_t pcnt; /* Packet counter */
869 uint64_t bcnt; /* Byte counter */
870 uint32_t timestamp; /* tv_sec of last match */
871
872 ipfw_insn cmd[1]; /* storage for commands */
873};
874
875 int convert_rule_to_7(struct ip_fw *rule);
876int convert_rule_to_8(struct ip_fw *rule);
877
878#ifndef RULESIZE7
879#define RULESIZE7(rule) (sizeof(struct ip_fw7) + \
880 ((struct ip_fw7 *)(rule))->cmd_len * 4 - 4)
881#endif
882
883
884/*
885 * Copy the static and dynamic rules to the supplied buffer
886 * and return the amount of space actually used.
887 * Must be run under IPFW_UH_RLOCK
888 */
889static size_t
890ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
891{
892 char *bp = buf;
893 char *ep = bp + space;
894 struct ip_fw *rule, *dst;
895 int l, i;
896 time_t boot_seconds;
897
898 boot_seconds = boottime.tv_sec;
899 for (i = 0; i < chain->n_rules; i++) {
900 rule = chain->map[i];
901
902 if (is7) {
903 /* Convert rule to FreeBSd 7.2 format */
904 l = RULESIZE7(rule);
905 if (bp + l + sizeof(uint32_t) <= ep) {
906 int error;
907 bcopy(rule, bp, l + sizeof(uint32_t));
908 error = convert_rule_to_7((struct ip_fw *) bp);
909 if (error)
910 return 0; /*XXX correct? */
911 /*
912 * XXX HACK. Store the disable mask in the "next"
913 * pointer in a wild attempt to keep the ABI the same.
914 * Why do we do this on EVERY rule?
915 */
916 bcopy(&V_set_disable,
917 &(((struct ip_fw7 *)bp)->next_rule),
918 sizeof(V_set_disable));
919 if (((struct ip_fw7 *)bp)->timestamp)
920 ((struct ip_fw7 *)bp)->timestamp += boot_seconds;
921 bp += l;
922 }
923 continue; /* go to next rule */
924 }
925
926 /* normal mode, don't touch rules */
927 l = RULESIZE(rule);
928 if (bp + l > ep) { /* should not happen */
929 printf("overflow dumping static rules\n");
930 break;
931 }
932 dst = (struct ip_fw *)bp;
933 bcopy(rule, dst, l);
934 /*
935 * XXX HACK. Store the disable mask in the "next"
936 * pointer in a wild attempt to keep the ABI the same.
937 * Why do we do this on EVERY rule?
938 */
939 bcopy(&V_set_disable, &dst->next_rule, sizeof(V_set_disable));
940 if (dst->timestamp)
941 dst->timestamp += boot_seconds;
942 bp += l;
943 }
944 ipfw_get_dynamic(&bp, ep); /* protected by the dynamic lock */
945 return (bp - (char *)buf);
946}
947
948
949/**
950 * {set|get}sockopt parser.
951 */
952int
953ipfw_ctl(struct sockopt *sopt)
954{
955#define RULE_MAXSIZE (256*sizeof(u_int32_t))
956 int error;
957 size_t size;
958 struct ip_fw *buf, *rule;
959 struct ip_fw_chain *chain;
960 u_int32_t rulenum[2];
961
962 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
963 if (error)
964 return (error);
965
966 /*
967 * Disallow modifications in really-really secure mode, but still allow
968 * the logging counters to be reset.
969 */
970 if (sopt->sopt_name == IP_FW_ADD ||
971 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
972 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
973 if (error)
974 return (error);
975 }
976
977 chain = &V_layer3_chain;
978 error = 0;
979
980 switch (sopt->sopt_name) {
981 case IP_FW_GET:
982 /*
983 * pass up a copy of the current rules. Static rules
984 * come first (the last of which has number IPFW_DEFAULT_RULE),
985 * followed by a possibly empty list of dynamic rule.
986 * The last dynamic rule has NULL in the "next" field.
987 *
988 * Note that the calculated size is used to bound the
989 * amount of data returned to the user. The rule set may
990 * change between calculating the size and returning the
991 * data in which case we'll just return what fits.
992 */
993 for (;;) {
994 int len = 0, want;
995
996 size = chain->static_len;
997 size += ipfw_dyn_len();
998 if (size >= sopt->sopt_valsize)
999 break;
1000 buf = malloc(size, M_TEMP, M_WAITOK);
1001 if (buf == NULL)
1002 break;
1003 IPFW_UH_RLOCK(chain);
1004 /* check again how much space we need */
1005 want = chain->static_len + ipfw_dyn_len();
1006 if (size >= want)
1007 len = ipfw_getrules(chain, buf, size);
1008 IPFW_UH_RUNLOCK(chain);
1009 if (size >= want)
1010 error = sooptcopyout(sopt, buf, len);
1011 free(buf, M_TEMP);
1012 if (size >= want)
1013 break;
1014 }
1015 break;
1016
1017 case IP_FW_FLUSH:
1018 /* locking is done within del_entry() */
1019 error = del_entry(chain, 0); /* special case, rule=0, cmd=0 means all */
1020 break;
1021
1022 case IP_FW_ADD:
1023 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
1024 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
1025 sizeof(struct ip_fw7) );
1026
1027 /*
1028 * If the size of commands equals RULESIZE7 then we assume
1029 * a FreeBSD7.2 binary is talking to us (set is7=1).
1030 * is7 is persistent so the next 'ipfw list' command
1031 * will use this format.
1032 * NOTE: If wrong version is guessed (this can happen if
1033 * the first ipfw command is 'ipfw [pipe] list')
1034 * the ipfw binary may crash or loop infinitly...
1035 */
1036 if (sopt->sopt_valsize == RULESIZE7(rule)) {
1037 is7 = 1;
1038 error = convert_rule_to_8(rule);
1039 if (error)
1040 return error;
1041 if (error == 0)
1042 error = check_ipfw_struct(rule, RULESIZE(rule));
1043 } else {
1044 is7 = 0;
1045 if (error == 0)
1046 error = check_ipfw_struct(rule, sopt->sopt_valsize);
1047 }
1048 if (error == 0) {
1049 /* locking is done within ipfw_add_rule() */
1050 error = ipfw_add_rule(chain, rule);
1051 size = RULESIZE(rule);
1052 if (!error && sopt->sopt_dir == SOPT_GET) {
1053 if (is7) {
1054 error = convert_rule_to_7(rule);
1055 size = RULESIZE7(rule);
1056 if (error)
1057 return error;
1058 }
1059 error = sooptcopyout(sopt, rule, size);
1060 }
1061 }
1062 free(rule, M_TEMP);
1063 break;
1064
1065 case IP_FW_DEL:
1066 /*
1067 * IP_FW_DEL is used for deleting single rules or sets,
1068 * and (ab)used to atomically manipulate sets. Argument size
1069 * is used to distinguish between the two:
1070 * sizeof(u_int32_t)
1071 * delete single rule or set of rules,
1072 * or reassign rules (or sets) to a different set.
1073 * 2*sizeof(u_int32_t)
1074 * atomic disable/enable sets.
1075 * first u_int32_t contains sets to be disabled,
1076 * second u_int32_t contains sets to be enabled.
1077 */
1078 error = sooptcopyin(sopt, rulenum,
1079 2*sizeof(u_int32_t), sizeof(u_int32_t));
1080 if (error)
1081 break;
1082 size = sopt->sopt_valsize;
1083 if (size == sizeof(u_int32_t) && rulenum[0] != 0) {
1084 /* delete or reassign, locking done in del_entry() */
1085 error = del_entry(chain, rulenum[0]);
1086 } else if (size == 2*sizeof(u_int32_t)) { /* set enable/disable */
1087 IPFW_UH_WLOCK(chain);
1088 V_set_disable =
1089 (V_set_disable | rulenum[0]) & ~rulenum[1] &
1090 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
1091 IPFW_UH_WUNLOCK(chain);
1092 } else
1093 error = EINVAL;
1094 break;
1095
1096 case IP_FW_ZERO:
1097 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
1098 rulenum[0] = 0;
1099 if (sopt->sopt_val != 0) {
1100 error = sooptcopyin(sopt, rulenum,
1101 sizeof(u_int32_t), sizeof(u_int32_t));
1102 if (error)
1103 break;
1104 }
1105 error = zero_entry(chain, rulenum[0],
1106 sopt->sopt_name == IP_FW_RESETLOG);
1107 break;
1108
1109 /*--- TABLE manipulations are protected by the IPFW_LOCK ---*/
1110 case IP_FW_TABLE_ADD:
1111 {
1112 ipfw_table_entry ent;
1113
1114 error = sooptcopyin(sopt, &ent,
1115 sizeof(ent), sizeof(ent));
1116 if (error)
1117 break;
1118 error = ipfw_add_table_entry(chain, ent.tbl,
1119 ent.addr, ent.masklen, ent.value);
1120 }
1121 break;
1122
1123 case IP_FW_TABLE_DEL:
1124 {
1125 ipfw_table_entry ent;
1126
1127 error = sooptcopyin(sopt, &ent,
1128 sizeof(ent), sizeof(ent));
1129 if (error)
1130 break;
1131 error = ipfw_del_table_entry(chain, ent.tbl,
1132 ent.addr, ent.masklen);
1133 }
1134 break;
1135
1136 case IP_FW_TABLE_FLUSH:
1137 {
1138 u_int16_t tbl;
1139
1140 error = sooptcopyin(sopt, &tbl,
1141 sizeof(tbl), sizeof(tbl));
1142 if (error)
1143 break;
1144 IPFW_WLOCK(chain);
1145 error = ipfw_flush_table(chain, tbl);
1146 IPFW_WUNLOCK(chain);
1147 }
1148 break;
1149
1150 case IP_FW_TABLE_GETSIZE:
1151 {
1152 u_int32_t tbl, cnt;
1153
1154 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
1155 sizeof(tbl))))
1156 break;
1157 IPFW_RLOCK(chain);
1158 error = ipfw_count_table(chain, tbl, &cnt);
1159 IPFW_RUNLOCK(chain);
1160 if (error)
1161 break;
1162 error = sooptcopyout(sopt, &cnt, sizeof(cnt));
1163 }
1164 break;
1165
1166 case IP_FW_TABLE_LIST:
1167 {
1168 ipfw_table *tbl;
1169
1170 if (sopt->sopt_valsize < sizeof(*tbl)) {
1171 error = EINVAL;
1172 break;
1173 }
1174 size = sopt->sopt_valsize;
1175 tbl = malloc(size, M_TEMP, M_WAITOK);
1176 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
1177 if (error) {
1178 free(tbl, M_TEMP);
1179 break;
1180 }
1181 tbl->size = (size - sizeof(*tbl)) /
1182 sizeof(ipfw_table_entry);
1183 IPFW_RLOCK(chain);
1184 error = ipfw_dump_table(chain, tbl);
1185 IPFW_RUNLOCK(chain);
1186 if (error) {
1187 free(tbl, M_TEMP);
1188 break;
1189 }
1190 error = sooptcopyout(sopt, tbl, size);
1191 free(tbl, M_TEMP);
1192 }
1193 break;
1194
1195 /*--- NAT operations are protected by the IPFW_LOCK ---*/
1196 case IP_FW_NAT_CFG:
1197 if (IPFW_NAT_LOADED)
1198 error = ipfw_nat_cfg_ptr(sopt);
1199 else {
1200 printf("IP_FW_NAT_CFG: %s\n",
1201 "ipfw_nat not present, please load it");
1202 error = EINVAL;
1203 }
1204 break;
1205
1206 case IP_FW_NAT_DEL:
1207 if (IPFW_NAT_LOADED)
1208 error = ipfw_nat_del_ptr(sopt);
1209 else {
1210 printf("IP_FW_NAT_DEL: %s\n",
1211 "ipfw_nat not present, please load it");
1212 error = EINVAL;
1213 }
1214 break;
1215
1216 case IP_FW_NAT_GET_CONFIG:
1217 if (IPFW_NAT_LOADED)
1218 error = ipfw_nat_get_cfg_ptr(sopt);
1219 else {
1220 printf("IP_FW_NAT_GET_CFG: %s\n",
1221 "ipfw_nat not present, please load it");
1222 error = EINVAL;
1223 }
1224 break;
1225
1226 case IP_FW_NAT_GET_LOG:
1227 if (IPFW_NAT_LOADED)
1228 error = ipfw_nat_get_log_ptr(sopt);
1229 else {
1230 printf("IP_FW_NAT_GET_LOG: %s\n",
1231 "ipfw_nat not present, please load it");
1232 error = EINVAL;
1233 }
1234 break;
1235
1236 default:
1237 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
1238 error = EINVAL;
1239 }
1240
1241 return (error);
1242#undef RULE_MAXSIZE
1243}
1244
1245
1246#define RULE_MAXSIZE (256*sizeof(u_int32_t))
1247
1248/* Functions to convert rules 7.2 <==> 8.0 */
1249int
1250convert_rule_to_7(struct ip_fw *rule)
1251{
1252 /* Used to modify original rule */
1253 struct ip_fw7 *rule7 = (struct ip_fw7 *)rule;
1254 /* copy of original rule, version 8 */
1255 struct ip_fw *tmp;
1256
1257 /* Used to copy commands */
1258 ipfw_insn *ccmd, *dst;
1259 int ll = 0, ccmdlen = 0;
1260
1261 tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
1262 if (tmp == NULL) {
1263 return 1; //XXX error
1264 }
1265 bcopy(rule, tmp, RULE_MAXSIZE);
1266
1267 /* Copy fields */
1268 rule7->_pad = tmp->_pad;
1269 rule7->set = tmp->set;
1270 rule7->rulenum = tmp->rulenum;
1271 rule7->cmd_len = tmp->cmd_len;
1272 rule7->act_ofs = tmp->act_ofs;
1273 rule7->next_rule = (struct ip_fw7 *)tmp->next_rule;
1274 rule7->next = (struct ip_fw7 *)tmp->x_next;
1275 rule7->cmd_len = tmp->cmd_len;
1276 rule7->pcnt = tmp->pcnt;
1277 rule7->bcnt = tmp->bcnt;
1278 rule7->timestamp = tmp->timestamp;
1279
1280 /* Copy commands */
1281 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule7->cmd ;
1282 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
1283 ccmdlen = F_LEN(ccmd);
1284
1285 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
1286
1287 if (dst->opcode > O_NAT)
1288 /* O_REASS doesn't exists in 7.2 version, so
1289 * decrement opcode if it is after O_REASS
1290 */
1291 dst->opcode--;
1292
1293 if (ccmdlen > ll) {
1294 printf("ipfw: opcode %d size truncated\n",
1295 ccmd->opcode);
1296 return EINVAL;
1297 }
1298 }
1299 free(tmp, M_TEMP);
1300
1301 return 0;
1302}
1303
1304int
1305convert_rule_to_8(struct ip_fw *rule)
1306{
1307 /* Used to modify original rule */
1308 struct ip_fw7 *rule7 = (struct ip_fw7 *) rule;
1309
1310 /* Used to copy commands */
1311 ipfw_insn *ccmd, *dst;
1312 int ll = 0, ccmdlen = 0;
1313
1314 /* Copy of original rule */
1315 struct ip_fw7 *tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
1316 if (tmp == NULL) {
1317 return 1; //XXX error
1318 }
1319
1320 bcopy(rule7, tmp, RULE_MAXSIZE);
1321
1322 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule->cmd ;
1323 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
1324 ccmdlen = F_LEN(ccmd);
1325
1326 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
1327
1328 if (dst->opcode > O_NAT)
1329 /* O_REASS doesn't exists in 7.2 version, so
1330 * increment opcode if it is after O_REASS
1331 */
1332 dst->opcode++;
1333
1334 if (ccmdlen > ll) {
1335 printf("ipfw: opcode %d size truncated\n",
1336 ccmd->opcode);
1337 return EINVAL;
1338 }
1339 }
1340
1341 rule->_pad = tmp->_pad;
1342 rule->set = tmp->set;
1343 rule->rulenum = tmp->rulenum;
1344 rule->cmd_len = tmp->cmd_len;
1345 rule->act_ofs = tmp->act_ofs;
1346 rule->next_rule = (struct ip_fw *)tmp->next_rule;
1347 rule->x_next = (struct ip_fw *)tmp->next;
1348 rule->cmd_len = tmp->cmd_len;
1349 rule->id = 0; /* XXX see if is ok = 0 */
1350 rule->pcnt = tmp->pcnt;
1351 rule->bcnt = tmp->bcnt;
1352 rule->timestamp = tmp->timestamp;
1353
1354 free (tmp, M_TEMP);
1355 return 0;
1356}
1357
1358/* end of file */
2 * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
3 *
4 * Supported by: Valeria Paoli
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 *
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
25 * SUCH DAMAGE.
26 */
27
28#include <sys/cdefs.h>
29__FBSDID("$FreeBSD: head/sys/netinet/ipfw/ip_fw_sockopt.c 225518 2011-09-12 21:09:56Z jhb $");
30
31/*
32 * Sockopt support for ipfw. The routines here implement
33 * the upper half of the ipfw code.
34 */
35
36#include "opt_ipfw.h"
37#if !defined(KLD_MODULE)
38#include "opt_ipdivert.h"
39#include "opt_ipdn.h"
40#include "opt_inet.h"
41#ifndef INET
42#error IPFIREWALL requires INET.
43#endif /* INET */
44#endif
45#include "opt_inet6.h"
46#include "opt_ipsec.h"
47
48#include <sys/param.h>
49#include <sys/systm.h>
50#include <sys/malloc.h>
51#include <sys/mbuf.h> /* struct m_tag used by nested headers */
52#include <sys/kernel.h>
53#include <sys/lock.h>
54#include <sys/priv.h>
55#include <sys/proc.h>
56#include <sys/rwlock.h>
57#include <sys/socket.h>
58#include <sys/socketvar.h>
59#include <sys/sysctl.h>
60#include <sys/syslog.h>
61#include <net/if.h>
62#include <net/route.h>
63#include <net/vnet.h>
64
65#include <netinet/in.h>
66#include <netinet/ip_var.h> /* hooks */
67#include <netinet/ip_fw.h>
68#include <netinet/ipfw/ip_fw_private.h>
69
70#ifdef MAC
71#include <security/mac/mac_framework.h>
72#endif
73
74MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
75
76/*
77 * static variables followed by global ones (none in this file)
78 */
79
80/*
81 * Find the smallest rule >= key, id.
82 * We could use bsearch but it is so simple that we code it directly
83 */
84int
85ipfw_find_rule(struct ip_fw_chain *chain, uint32_t key, uint32_t id)
86{
87 int i, lo, hi;
88 struct ip_fw *r;
89
90 for (lo = 0, hi = chain->n_rules - 1; lo < hi;) {
91 i = (lo + hi) / 2;
92 r = chain->map[i];
93 if (r->rulenum < key)
94 lo = i + 1; /* continue from the next one */
95 else if (r->rulenum > key)
96 hi = i; /* this might be good */
97 else if (r->id < id)
98 lo = i + 1; /* continue from the next one */
99 else /* r->id >= id */
100 hi = i; /* this might be good */
101 };
102 return hi;
103}
104
105/*
106 * allocate a new map, returns the chain locked. extra is the number
107 * of entries to add or delete.
108 */
109static struct ip_fw **
110get_map(struct ip_fw_chain *chain, int extra, int locked)
111{
112
113 for (;;) {
114 struct ip_fw **map;
115 int i;
116
117 i = chain->n_rules + extra;
118 map = malloc(i * sizeof(struct ip_fw *), M_IPFW,
119 locked ? M_NOWAIT : M_WAITOK);
120 if (map == NULL) {
121 printf("%s: cannot allocate map\n", __FUNCTION__);
122 return NULL;
123 }
124 if (!locked)
125 IPFW_UH_WLOCK(chain);
126 if (i >= chain->n_rules + extra) /* good */
127 return map;
128 /* otherwise we lost the race, free and retry */
129 if (!locked)
130 IPFW_UH_WUNLOCK(chain);
131 free(map, M_IPFW);
132 }
133}
134
135/*
136 * swap the maps. It is supposed to be called with IPFW_UH_WLOCK
137 */
138static struct ip_fw **
139swap_map(struct ip_fw_chain *chain, struct ip_fw **new_map, int new_len)
140{
141 struct ip_fw **old_map;
142
143 IPFW_WLOCK(chain);
144 chain->id++;
145 chain->n_rules = new_len;
146 old_map = chain->map;
147 chain->map = new_map;
148 IPFW_WUNLOCK(chain);
149 return old_map;
150}
151
152/*
153 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
154 * possibly create a rule number and add the rule to the list.
155 * Update the rule_number in the input struct so the caller knows it as well.
156 * XXX DO NOT USE FOR THE DEFAULT RULE.
157 * Must be called without IPFW_UH held
158 */
159int
160ipfw_add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule)
161{
162 struct ip_fw *rule;
163 int i, l, insert_before;
164 struct ip_fw **map; /* the new array of pointers */
165
166 if (chain->rules == NULL || input_rule->rulenum > IPFW_DEFAULT_RULE-1)
167 return (EINVAL);
168
169 l = RULESIZE(input_rule);
170 rule = malloc(l, M_IPFW, M_WAITOK | M_ZERO);
171 if (rule == NULL)
172 return (ENOSPC);
173 /* get_map returns with IPFW_UH_WLOCK if successful */
174 map = get_map(chain, 1, 0 /* not locked */);
175 if (map == NULL) {
176 free(rule, M_IPFW);
177 return ENOSPC;
178 }
179
180 bcopy(input_rule, rule, l);
181 /* clear fields not settable from userland */
182 rule->x_next = NULL;
183 rule->next_rule = NULL;
184 rule->pcnt = 0;
185 rule->bcnt = 0;
186 rule->timestamp = 0;
187
188 if (V_autoinc_step < 1)
189 V_autoinc_step = 1;
190 else if (V_autoinc_step > 1000)
191 V_autoinc_step = 1000;
192 /* find the insertion point, we will insert before */
193 insert_before = rule->rulenum ? rule->rulenum + 1 : IPFW_DEFAULT_RULE;
194 i = ipfw_find_rule(chain, insert_before, 0);
195 /* duplicate first part */
196 if (i > 0)
197 bcopy(chain->map, map, i * sizeof(struct ip_fw *));
198 map[i] = rule;
199 /* duplicate remaining part, we always have the default rule */
200 bcopy(chain->map + i, map + i + 1,
201 sizeof(struct ip_fw *) *(chain->n_rules - i));
202 if (rule->rulenum == 0) {
203 /* write back the number */
204 rule->rulenum = i > 0 ? map[i-1]->rulenum : 0;
205 if (rule->rulenum < IPFW_DEFAULT_RULE - V_autoinc_step)
206 rule->rulenum += V_autoinc_step;
207 input_rule->rulenum = rule->rulenum;
208 }
209
210 rule->id = chain->id + 1;
211 map = swap_map(chain, map, chain->n_rules + 1);
212 chain->static_len += l;
213 IPFW_UH_WUNLOCK(chain);
214 if (map)
215 free(map, M_IPFW);
216 return (0);
217}
218
219/*
220 * Reclaim storage associated with a list of rules. This is
221 * typically the list created using remove_rule.
222 * A NULL pointer on input is handled correctly.
223 */
224void
225ipfw_reap_rules(struct ip_fw *head)
226{
227 struct ip_fw *rule;
228
229 while ((rule = head) != NULL) {
230 head = head->x_next;
231 free(rule, M_IPFW);
232 }
233}
234
235/*
236 * Used by del_entry() to check if a rule should be kept.
237 * Returns 1 if the rule must be kept, 0 otherwise.
238 *
239 * Called with cmd = {0,1,5}.
240 * cmd == 0 matches on rule numbers, excludes rules in RESVD_SET if n == 0 ;
241 * cmd == 1 matches on set numbers only, rule numbers are ignored;
242 * cmd == 5 matches on rule and set numbers.
243 *
244 * n == 0 is a wildcard for rule numbers, there is no wildcard for sets.
245 *
246 * Rules to keep are
247 * (default || reserved || !match_set || !match_number)
248 * where
249 * default ::= (rule->rulenum == IPFW_DEFAULT_RULE)
250 * // the default rule is always protected
251 *
252 * reserved ::= (cmd == 0 && n == 0 && rule->set == RESVD_SET)
253 * // RESVD_SET is protected only if cmd == 0 and n == 0 ("ipfw flush")
254 *
255 * match_set ::= (cmd == 0 || rule->set == set)
256 * // set number is ignored for cmd == 0
257 *
258 * match_number ::= (cmd == 1 || n == 0 || n == rule->rulenum)
259 * // number is ignored for cmd == 1 or n == 0
260 *
261 */
262static int
263keep_rule(struct ip_fw *rule, uint8_t cmd, uint8_t set, uint32_t n)
264{
265 return
266 (rule->rulenum == IPFW_DEFAULT_RULE) ||
267 (cmd == 0 && n == 0 && rule->set == RESVD_SET) ||
268 !(cmd == 0 || rule->set == set) ||
269 !(cmd == 1 || n == 0 || n == rule->rulenum);
270}
271
272/**
273 * Remove all rules with given number, or do set manipulation.
274 * Assumes chain != NULL && *chain != NULL.
275 *
276 * The argument is an uint32_t. The low 16 bit are the rule or set number;
277 * the next 8 bits are the new set; the top 8 bits indicate the command:
278 *
279 * 0 delete rules numbered "rulenum"
280 * 1 delete rules in set "rulenum"
281 * 2 move rules "rulenum" to set "new_set"
282 * 3 move rules from set "rulenum" to set "new_set"
283 * 4 swap sets "rulenum" and "new_set"
284 * 5 delete rules "rulenum" and set "new_set"
285 */
286static int
287del_entry(struct ip_fw_chain *chain, uint32_t arg)
288{
289 struct ip_fw *rule;
290 uint32_t num; /* rule number or old_set */
291 uint8_t cmd, new_set;
292 int start, end, i, ofs, n;
293 struct ip_fw **map = NULL;
294 int error = 0;
295
296 num = arg & 0xffff;
297 cmd = (arg >> 24) & 0xff;
298 new_set = (arg >> 16) & 0xff;
299
300 if (cmd > 5 || new_set > RESVD_SET)
301 return EINVAL;
302 if (cmd == 0 || cmd == 2 || cmd == 5) {
303 if (num >= IPFW_DEFAULT_RULE)
304 return EINVAL;
305 } else {
306 if (num > RESVD_SET) /* old_set */
307 return EINVAL;
308 }
309
310 IPFW_UH_WLOCK(chain); /* arbitrate writers */
311 chain->reap = NULL; /* prepare for deletions */
312
313 switch (cmd) {
314 case 0: /* delete rules "num" (num == 0 matches all) */
315 case 1: /* delete all rules in set N */
316 case 5: /* delete rules with number N and set "new_set". */
317
318 /*
319 * Locate first rule to delete (start), the rule after
320 * the last one to delete (end), and count how many
321 * rules to delete (n). Always use keep_rule() to
322 * determine which rules to keep.
323 */
324 n = 0;
325 if (cmd == 1) {
326 /* look for a specific set including RESVD_SET.
327 * Must scan the entire range, ignore num.
328 */
329 new_set = num;
330 for (start = -1, end = i = 0; i < chain->n_rules; i++) {
331 if (keep_rule(chain->map[i], cmd, new_set, 0))
332 continue;
333 if (start < 0)
334 start = i;
335 end = i;
336 n++;
337 }
338 end++; /* first non-matching */
339 } else {
340 /* Optimized search on rule numbers */
341 start = ipfw_find_rule(chain, num, 0);
342 for (end = start; end < chain->n_rules; end++) {
343 rule = chain->map[end];
344 if (num > 0 && rule->rulenum != num)
345 break;
346 if (!keep_rule(rule, cmd, new_set, num))
347 n++;
348 }
349 }
350
351 if (n == 0) {
352 /* A flush request (arg == 0 or cmd == 1) on empty
353 * ruleset returns with no error. On the contrary,
354 * if there is no match on a specific request,
355 * we return EINVAL.
356 */
357 if (arg != 0 && cmd != 1)
358 error = EINVAL;
359 break;
360 }
361
362 /* We have something to delete. Allocate the new map */
363 map = get_map(chain, -n, 1 /* locked */);
364 if (map == NULL) {
365 error = EINVAL;
366 break;
367 }
368
369 /* 1. bcopy the initial part of the map */
370 if (start > 0)
371 bcopy(chain->map, map, start * sizeof(struct ip_fw *));
372 /* 2. copy active rules between start and end */
373 for (i = ofs = start; i < end; i++) {
374 rule = chain->map[i];
375 if (keep_rule(rule, cmd, new_set, num))
376 map[ofs++] = rule;
377 }
378 /* 3. copy the final part of the map */
379 bcopy(chain->map + end, map + ofs,
380 (chain->n_rules - end) * sizeof(struct ip_fw *));
381 /* 4. swap the maps (under BH_LOCK) */
382 map = swap_map(chain, map, chain->n_rules - n);
383 /* 5. now remove the rules deleted from the old map */
384 for (i = start; i < end; i++) {
385 int l;
386 rule = map[i];
387 if (keep_rule(rule, cmd, new_set, num))
388 continue;
389 l = RULESIZE(rule);
390 chain->static_len -= l;
391 ipfw_remove_dyn_children(rule);
392 rule->x_next = chain->reap;
393 chain->reap = rule;
394 }
395 break;
396
397 /*
398 * In the next 3 cases the loop stops at (n_rules - 1)
399 * because the default rule is never eligible..
400 */
401
402 case 2: /* move rules with given RULE number to new set */
403 for (i = 0; i < chain->n_rules - 1; i++) {
404 rule = chain->map[i];
405 if (rule->rulenum == num)
406 rule->set = new_set;
407 }
408 break;
409
410 case 3: /* move rules with given SET number to new set */
411 for (i = 0; i < chain->n_rules - 1; i++) {
412 rule = chain->map[i];
413 if (rule->set == num)
414 rule->set = new_set;
415 }
416 break;
417
418 case 4: /* swap two sets */
419 for (i = 0; i < chain->n_rules - 1; i++) {
420 rule = chain->map[i];
421 if (rule->set == num)
422 rule->set = new_set;
423 else if (rule->set == new_set)
424 rule->set = num;
425 }
426 break;
427 }
428
429 rule = chain->reap;
430 chain->reap = NULL;
431 IPFW_UH_WUNLOCK(chain);
432 ipfw_reap_rules(rule);
433 if (map)
434 free(map, M_IPFW);
435 return error;
436}
437
438/*
439 * Clear counters for a specific rule.
440 * Normally run under IPFW_UH_RLOCK, but these are idempotent ops
441 * so we only care that rules do not disappear.
442 */
443static void
444clear_counters(struct ip_fw *rule, int log_only)
445{
446 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
447
448 if (log_only == 0) {
449 rule->bcnt = rule->pcnt = 0;
450 rule->timestamp = 0;
451 }
452 if (l->o.opcode == O_LOG)
453 l->log_left = l->max_log;
454}
455
456/**
457 * Reset some or all counters on firewall rules.
458 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
459 * the next 8 bits are the set number, the top 8 bits are the command:
460 * 0 work with rules from all set's;
461 * 1 work with rules only from specified set.
462 * Specified rule number is zero if we want to clear all entries.
463 * log_only is 1 if we only want to reset logs, zero otherwise.
464 */
465static int
466zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
467{
468 struct ip_fw *rule;
469 char *msg;
470 int i;
471
472 uint16_t rulenum = arg & 0xffff;
473 uint8_t set = (arg >> 16) & 0xff;
474 uint8_t cmd = (arg >> 24) & 0xff;
475
476 if (cmd > 1)
477 return (EINVAL);
478 if (cmd == 1 && set > RESVD_SET)
479 return (EINVAL);
480
481 IPFW_UH_RLOCK(chain);
482 if (rulenum == 0) {
483 V_norule_counter = 0;
484 for (i = 0; i < chain->n_rules; i++) {
485 rule = chain->map[i];
486 /* Skip rules not in our set. */
487 if (cmd == 1 && rule->set != set)
488 continue;
489 clear_counters(rule, log_only);
490 }
491 msg = log_only ? "All logging counts reset" :
492 "Accounting cleared";
493 } else {
494 int cleared = 0;
495 for (i = 0; i < chain->n_rules; i++) {
496 rule = chain->map[i];
497 if (rule->rulenum == rulenum) {
498 if (cmd == 0 || rule->set == set)
499 clear_counters(rule, log_only);
500 cleared = 1;
501 }
502 if (rule->rulenum > rulenum)
503 break;
504 }
505 if (!cleared) { /* we did not find any matching rules */
506 IPFW_UH_RUNLOCK(chain);
507 return (EINVAL);
508 }
509 msg = log_only ? "logging count reset" : "cleared";
510 }
511 IPFW_UH_RUNLOCK(chain);
512
513 if (V_fw_verbose) {
514 int lev = LOG_SECURITY | LOG_NOTICE;
515
516 if (rulenum)
517 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg);
518 else
519 log(lev, "ipfw: %s.\n", msg);
520 }
521 return (0);
522}
523
524/*
525 * Check validity of the structure before insert.
526 * Rules are simple, so this mostly need to check rule sizes.
527 */
528static int
529check_ipfw_struct(struct ip_fw *rule, int size)
530{
531 int l, cmdlen = 0;
532 int have_action=0;
533 ipfw_insn *cmd;
534
535 if (size < sizeof(*rule)) {
536 printf("ipfw: rule too short\n");
537 return (EINVAL);
538 }
539 /* first, check for valid size */
540 l = RULESIZE(rule);
541 if (l != size) {
542 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
543 return (EINVAL);
544 }
545 if (rule->act_ofs >= rule->cmd_len) {
546 printf("ipfw: bogus action offset (%u > %u)\n",
547 rule->act_ofs, rule->cmd_len - 1);
548 return (EINVAL);
549 }
550 /*
551 * Now go for the individual checks. Very simple ones, basically only
552 * instruction sizes.
553 */
554 for (l = rule->cmd_len, cmd = rule->cmd ;
555 l > 0 ; l -= cmdlen, cmd += cmdlen) {
556 cmdlen = F_LEN(cmd);
557 if (cmdlen > l) {
558 printf("ipfw: opcode %d size truncated\n",
559 cmd->opcode);
560 return EINVAL;
561 }
562 switch (cmd->opcode) {
563 case O_PROBE_STATE:
564 case O_KEEP_STATE:
565 case O_PROTO:
566 case O_IP_SRC_ME:
567 case O_IP_DST_ME:
568 case O_LAYER2:
569 case O_IN:
570 case O_FRAG:
571 case O_DIVERTED:
572 case O_IPOPT:
573 case O_IPTOS:
574 case O_IPPRECEDENCE:
575 case O_IPVER:
576 case O_SOCKARG:
577 case O_TCPWIN:
578 case O_TCPFLAGS:
579 case O_TCPOPTS:
580 case O_ESTAB:
581 case O_VERREVPATH:
582 case O_VERSRCREACH:
583 case O_ANTISPOOF:
584 case O_IPSEC:
585#ifdef INET6
586 case O_IP6_SRC_ME:
587 case O_IP6_DST_ME:
588 case O_EXT_HDR:
589 case O_IP6:
590#endif
591 case O_IP4:
592 case O_TAG:
593 if (cmdlen != F_INSN_SIZE(ipfw_insn))
594 goto bad_size;
595 break;
596
597 case O_FIB:
598 if (cmdlen != F_INSN_SIZE(ipfw_insn))
599 goto bad_size;
600 if (cmd->arg1 >= rt_numfibs) {
601 printf("ipfw: invalid fib number %d\n",
602 cmd->arg1);
603 return EINVAL;
604 }
605 break;
606
607 case O_SETFIB:
608 if (cmdlen != F_INSN_SIZE(ipfw_insn))
609 goto bad_size;
610 if ((cmd->arg1 != IP_FW_TABLEARG) &&
611 (cmd->arg1 >= rt_numfibs)) {
612 printf("ipfw: invalid fib number %d\n",
613 cmd->arg1);
614 return EINVAL;
615 }
616 goto check_action;
617
618 case O_UID:
619 case O_GID:
620 case O_JAIL:
621 case O_IP_SRC:
622 case O_IP_DST:
623 case O_TCPSEQ:
624 case O_TCPACK:
625 case O_PROB:
626 case O_ICMPTYPE:
627 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
628 goto bad_size;
629 break;
630
631 case O_LIMIT:
632 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
633 goto bad_size;
634 break;
635
636 case O_LOG:
637 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
638 goto bad_size;
639
640 ((ipfw_insn_log *)cmd)->log_left =
641 ((ipfw_insn_log *)cmd)->max_log;
642
643 break;
644
645 case O_IP_SRC_MASK:
646 case O_IP_DST_MASK:
647 /* only odd command lengths */
648 if ( !(cmdlen & 1) || cmdlen > 31)
649 goto bad_size;
650 break;
651
652 case O_IP_SRC_SET:
653 case O_IP_DST_SET:
654 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
655 printf("ipfw: invalid set size %d\n",
656 cmd->arg1);
657 return EINVAL;
658 }
659 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
660 (cmd->arg1+31)/32 )
661 goto bad_size;
662 break;
663
664 case O_IP_SRC_LOOKUP:
665 case O_IP_DST_LOOKUP:
666 if (cmd->arg1 >= IPFW_TABLES_MAX) {
667 printf("ipfw: invalid table number %d\n",
668 cmd->arg1);
669 return (EINVAL);
670 }
671 if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
672 cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 1 &&
673 cmdlen != F_INSN_SIZE(ipfw_insn_u32))
674 goto bad_size;
675 break;
676
677 case O_MACADDR2:
678 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
679 goto bad_size;
680 break;
681
682 case O_NOP:
683 case O_IPID:
684 case O_IPTTL:
685 case O_IPLEN:
686 case O_TCPDATALEN:
687 case O_TAGGED:
688 if (cmdlen < 1 || cmdlen > 31)
689 goto bad_size;
690 break;
691
692 case O_MAC_TYPE:
693 case O_IP_SRCPORT:
694 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
695 if (cmdlen < 2 || cmdlen > 31)
696 goto bad_size;
697 break;
698
699 case O_RECV:
700 case O_XMIT:
701 case O_VIA:
702 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
703 goto bad_size;
704 break;
705
706 case O_ALTQ:
707 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
708 goto bad_size;
709 break;
710
711 case O_PIPE:
712 case O_QUEUE:
713 if (cmdlen != F_INSN_SIZE(ipfw_insn))
714 goto bad_size;
715 goto check_action;
716
717 case O_FORWARD_IP:
718#ifdef IPFIREWALL_FORWARD
719 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
720 goto bad_size;
721 goto check_action;
722#else
723 return EINVAL;
724#endif
725
726#ifdef INET6
727 case O_FORWARD_IP6:
728#ifdef IPFIREWALL_FORWARD
729 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa6))
730 goto bad_size;
731 goto check_action;
732#else
733 return (EINVAL);
734#endif
735#endif /* INET6 */
736
737 case O_DIVERT:
738 case O_TEE:
739 if (ip_divert_ptr == NULL)
740 return EINVAL;
741 else
742 goto check_size;
743 case O_NETGRAPH:
744 case O_NGTEE:
745 if (ng_ipfw_input_p == NULL)
746 return EINVAL;
747 else
748 goto check_size;
749 case O_NAT:
750 if (!IPFW_NAT_LOADED)
751 return EINVAL;
752 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
753 goto bad_size;
754 goto check_action;
755 case O_FORWARD_MAC: /* XXX not implemented yet */
756 case O_CHECK_STATE:
757 case O_COUNT:
758 case O_ACCEPT:
759 case O_DENY:
760 case O_REJECT:
761#ifdef INET6
762 case O_UNREACH6:
763#endif
764 case O_SKIPTO:
765 case O_REASS:
766 case O_CALLRETURN:
767check_size:
768 if (cmdlen != F_INSN_SIZE(ipfw_insn))
769 goto bad_size;
770check_action:
771 if (have_action) {
772 printf("ipfw: opcode %d, multiple actions"
773 " not allowed\n",
774 cmd->opcode);
775 return EINVAL;
776 }
777 have_action = 1;
778 if (l != cmdlen) {
779 printf("ipfw: opcode %d, action must be"
780 " last opcode\n",
781 cmd->opcode);
782 return EINVAL;
783 }
784 break;
785#ifdef INET6
786 case O_IP6_SRC:
787 case O_IP6_DST:
788 if (cmdlen != F_INSN_SIZE(struct in6_addr) +
789 F_INSN_SIZE(ipfw_insn))
790 goto bad_size;
791 break;
792
793 case O_FLOW6ID:
794 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
795 ((ipfw_insn_u32 *)cmd)->o.arg1)
796 goto bad_size;
797 break;
798
799 case O_IP6_SRC_MASK:
800 case O_IP6_DST_MASK:
801 if ( !(cmdlen & 1) || cmdlen > 127)
802 goto bad_size;
803 break;
804 case O_ICMP6TYPE:
805 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
806 goto bad_size;
807 break;
808#endif
809
810 default:
811 switch (cmd->opcode) {
812#ifndef INET6
813 case O_IP6_SRC_ME:
814 case O_IP6_DST_ME:
815 case O_EXT_HDR:
816 case O_IP6:
817 case O_UNREACH6:
818 case O_IP6_SRC:
819 case O_IP6_DST:
820 case O_FLOW6ID:
821 case O_IP6_SRC_MASK:
822 case O_IP6_DST_MASK:
823 case O_ICMP6TYPE:
824 printf("ipfw: no IPv6 support in kernel\n");
825 return EPROTONOSUPPORT;
826#endif
827 default:
828 printf("ipfw: opcode %d, unknown opcode\n",
829 cmd->opcode);
830 return EINVAL;
831 }
832 }
833 }
834 if (have_action == 0) {
835 printf("ipfw: missing action\n");
836 return EINVAL;
837 }
838 return 0;
839
840bad_size:
841 printf("ipfw: opcode %d size %d wrong\n",
842 cmd->opcode, cmdlen);
843 return EINVAL;
844}
845
846
847/*
848 * Translation of requests for compatibility with FreeBSD 7.2/8.
849 * a static variable tells us if we have an old client from userland,
850 * and if necessary we translate requests and responses between the
851 * two formats.
852 */
853static int is7 = 0;
854
855struct ip_fw7 {
856 struct ip_fw7 *next; /* linked list of rules */
857 struct ip_fw7 *next_rule; /* ptr to next [skipto] rule */
858 /* 'next_rule' is used to pass up 'set_disable' status */
859
860 uint16_t act_ofs; /* offset of action in 32-bit units */
861 uint16_t cmd_len; /* # of 32-bit words in cmd */
862 uint16_t rulenum; /* rule number */
863 uint8_t set; /* rule set (0..31) */
864 // #define RESVD_SET 31 /* set for default and persistent rules */
865 uint8_t _pad; /* padding */
866 // uint32_t id; /* rule id, only in v.8 */
867 /* These fields are present in all rules. */
868 uint64_t pcnt; /* Packet counter */
869 uint64_t bcnt; /* Byte counter */
870 uint32_t timestamp; /* tv_sec of last match */
871
872 ipfw_insn cmd[1]; /* storage for commands */
873};
874
875 int convert_rule_to_7(struct ip_fw *rule);
876int convert_rule_to_8(struct ip_fw *rule);
877
878#ifndef RULESIZE7
879#define RULESIZE7(rule) (sizeof(struct ip_fw7) + \
880 ((struct ip_fw7 *)(rule))->cmd_len * 4 - 4)
881#endif
882
883
884/*
885 * Copy the static and dynamic rules to the supplied buffer
886 * and return the amount of space actually used.
887 * Must be run under IPFW_UH_RLOCK
888 */
889static size_t
890ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
891{
892 char *bp = buf;
893 char *ep = bp + space;
894 struct ip_fw *rule, *dst;
895 int l, i;
896 time_t boot_seconds;
897
898 boot_seconds = boottime.tv_sec;
899 for (i = 0; i < chain->n_rules; i++) {
900 rule = chain->map[i];
901
902 if (is7) {
903 /* Convert rule to FreeBSd 7.2 format */
904 l = RULESIZE7(rule);
905 if (bp + l + sizeof(uint32_t) <= ep) {
906 int error;
907 bcopy(rule, bp, l + sizeof(uint32_t));
908 error = convert_rule_to_7((struct ip_fw *) bp);
909 if (error)
910 return 0; /*XXX correct? */
911 /*
912 * XXX HACK. Store the disable mask in the "next"
913 * pointer in a wild attempt to keep the ABI the same.
914 * Why do we do this on EVERY rule?
915 */
916 bcopy(&V_set_disable,
917 &(((struct ip_fw7 *)bp)->next_rule),
918 sizeof(V_set_disable));
919 if (((struct ip_fw7 *)bp)->timestamp)
920 ((struct ip_fw7 *)bp)->timestamp += boot_seconds;
921 bp += l;
922 }
923 continue; /* go to next rule */
924 }
925
926 /* normal mode, don't touch rules */
927 l = RULESIZE(rule);
928 if (bp + l > ep) { /* should not happen */
929 printf("overflow dumping static rules\n");
930 break;
931 }
932 dst = (struct ip_fw *)bp;
933 bcopy(rule, dst, l);
934 /*
935 * XXX HACK. Store the disable mask in the "next"
936 * pointer in a wild attempt to keep the ABI the same.
937 * Why do we do this on EVERY rule?
938 */
939 bcopy(&V_set_disable, &dst->next_rule, sizeof(V_set_disable));
940 if (dst->timestamp)
941 dst->timestamp += boot_seconds;
942 bp += l;
943 }
944 ipfw_get_dynamic(&bp, ep); /* protected by the dynamic lock */
945 return (bp - (char *)buf);
946}
947
948
949/**
950 * {set|get}sockopt parser.
951 */
952int
953ipfw_ctl(struct sockopt *sopt)
954{
955#define RULE_MAXSIZE (256*sizeof(u_int32_t))
956 int error;
957 size_t size;
958 struct ip_fw *buf, *rule;
959 struct ip_fw_chain *chain;
960 u_int32_t rulenum[2];
961
962 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
963 if (error)
964 return (error);
965
966 /*
967 * Disallow modifications in really-really secure mode, but still allow
968 * the logging counters to be reset.
969 */
970 if (sopt->sopt_name == IP_FW_ADD ||
971 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
972 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
973 if (error)
974 return (error);
975 }
976
977 chain = &V_layer3_chain;
978 error = 0;
979
980 switch (sopt->sopt_name) {
981 case IP_FW_GET:
982 /*
983 * pass up a copy of the current rules. Static rules
984 * come first (the last of which has number IPFW_DEFAULT_RULE),
985 * followed by a possibly empty list of dynamic rule.
986 * The last dynamic rule has NULL in the "next" field.
987 *
988 * Note that the calculated size is used to bound the
989 * amount of data returned to the user. The rule set may
990 * change between calculating the size and returning the
991 * data in which case we'll just return what fits.
992 */
993 for (;;) {
994 int len = 0, want;
995
996 size = chain->static_len;
997 size += ipfw_dyn_len();
998 if (size >= sopt->sopt_valsize)
999 break;
1000 buf = malloc(size, M_TEMP, M_WAITOK);
1001 if (buf == NULL)
1002 break;
1003 IPFW_UH_RLOCK(chain);
1004 /* check again how much space we need */
1005 want = chain->static_len + ipfw_dyn_len();
1006 if (size >= want)
1007 len = ipfw_getrules(chain, buf, size);
1008 IPFW_UH_RUNLOCK(chain);
1009 if (size >= want)
1010 error = sooptcopyout(sopt, buf, len);
1011 free(buf, M_TEMP);
1012 if (size >= want)
1013 break;
1014 }
1015 break;
1016
1017 case IP_FW_FLUSH:
1018 /* locking is done within del_entry() */
1019 error = del_entry(chain, 0); /* special case, rule=0, cmd=0 means all */
1020 break;
1021
1022 case IP_FW_ADD:
1023 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
1024 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
1025 sizeof(struct ip_fw7) );
1026
1027 /*
1028 * If the size of commands equals RULESIZE7 then we assume
1029 * a FreeBSD7.2 binary is talking to us (set is7=1).
1030 * is7 is persistent so the next 'ipfw list' command
1031 * will use this format.
1032 * NOTE: If wrong version is guessed (this can happen if
1033 * the first ipfw command is 'ipfw [pipe] list')
1034 * the ipfw binary may crash or loop infinitly...
1035 */
1036 if (sopt->sopt_valsize == RULESIZE7(rule)) {
1037 is7 = 1;
1038 error = convert_rule_to_8(rule);
1039 if (error)
1040 return error;
1041 if (error == 0)
1042 error = check_ipfw_struct(rule, RULESIZE(rule));
1043 } else {
1044 is7 = 0;
1045 if (error == 0)
1046 error = check_ipfw_struct(rule, sopt->sopt_valsize);
1047 }
1048 if (error == 0) {
1049 /* locking is done within ipfw_add_rule() */
1050 error = ipfw_add_rule(chain, rule);
1051 size = RULESIZE(rule);
1052 if (!error && sopt->sopt_dir == SOPT_GET) {
1053 if (is7) {
1054 error = convert_rule_to_7(rule);
1055 size = RULESIZE7(rule);
1056 if (error)
1057 return error;
1058 }
1059 error = sooptcopyout(sopt, rule, size);
1060 }
1061 }
1062 free(rule, M_TEMP);
1063 break;
1064
1065 case IP_FW_DEL:
1066 /*
1067 * IP_FW_DEL is used for deleting single rules or sets,
1068 * and (ab)used to atomically manipulate sets. Argument size
1069 * is used to distinguish between the two:
1070 * sizeof(u_int32_t)
1071 * delete single rule or set of rules,
1072 * or reassign rules (or sets) to a different set.
1073 * 2*sizeof(u_int32_t)
1074 * atomic disable/enable sets.
1075 * first u_int32_t contains sets to be disabled,
1076 * second u_int32_t contains sets to be enabled.
1077 */
1078 error = sooptcopyin(sopt, rulenum,
1079 2*sizeof(u_int32_t), sizeof(u_int32_t));
1080 if (error)
1081 break;
1082 size = sopt->sopt_valsize;
1083 if (size == sizeof(u_int32_t) && rulenum[0] != 0) {
1084 /* delete or reassign, locking done in del_entry() */
1085 error = del_entry(chain, rulenum[0]);
1086 } else if (size == 2*sizeof(u_int32_t)) { /* set enable/disable */
1087 IPFW_UH_WLOCK(chain);
1088 V_set_disable =
1089 (V_set_disable | rulenum[0]) & ~rulenum[1] &
1090 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
1091 IPFW_UH_WUNLOCK(chain);
1092 } else
1093 error = EINVAL;
1094 break;
1095
1096 case IP_FW_ZERO:
1097 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
1098 rulenum[0] = 0;
1099 if (sopt->sopt_val != 0) {
1100 error = sooptcopyin(sopt, rulenum,
1101 sizeof(u_int32_t), sizeof(u_int32_t));
1102 if (error)
1103 break;
1104 }
1105 error = zero_entry(chain, rulenum[0],
1106 sopt->sopt_name == IP_FW_RESETLOG);
1107 break;
1108
1109 /*--- TABLE manipulations are protected by the IPFW_LOCK ---*/
1110 case IP_FW_TABLE_ADD:
1111 {
1112 ipfw_table_entry ent;
1113
1114 error = sooptcopyin(sopt, &ent,
1115 sizeof(ent), sizeof(ent));
1116 if (error)
1117 break;
1118 error = ipfw_add_table_entry(chain, ent.tbl,
1119 ent.addr, ent.masklen, ent.value);
1120 }
1121 break;
1122
1123 case IP_FW_TABLE_DEL:
1124 {
1125 ipfw_table_entry ent;
1126
1127 error = sooptcopyin(sopt, &ent,
1128 sizeof(ent), sizeof(ent));
1129 if (error)
1130 break;
1131 error = ipfw_del_table_entry(chain, ent.tbl,
1132 ent.addr, ent.masklen);
1133 }
1134 break;
1135
1136 case IP_FW_TABLE_FLUSH:
1137 {
1138 u_int16_t tbl;
1139
1140 error = sooptcopyin(sopt, &tbl,
1141 sizeof(tbl), sizeof(tbl));
1142 if (error)
1143 break;
1144 IPFW_WLOCK(chain);
1145 error = ipfw_flush_table(chain, tbl);
1146 IPFW_WUNLOCK(chain);
1147 }
1148 break;
1149
1150 case IP_FW_TABLE_GETSIZE:
1151 {
1152 u_int32_t tbl, cnt;
1153
1154 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
1155 sizeof(tbl))))
1156 break;
1157 IPFW_RLOCK(chain);
1158 error = ipfw_count_table(chain, tbl, &cnt);
1159 IPFW_RUNLOCK(chain);
1160 if (error)
1161 break;
1162 error = sooptcopyout(sopt, &cnt, sizeof(cnt));
1163 }
1164 break;
1165
1166 case IP_FW_TABLE_LIST:
1167 {
1168 ipfw_table *tbl;
1169
1170 if (sopt->sopt_valsize < sizeof(*tbl)) {
1171 error = EINVAL;
1172 break;
1173 }
1174 size = sopt->sopt_valsize;
1175 tbl = malloc(size, M_TEMP, M_WAITOK);
1176 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
1177 if (error) {
1178 free(tbl, M_TEMP);
1179 break;
1180 }
1181 tbl->size = (size - sizeof(*tbl)) /
1182 sizeof(ipfw_table_entry);
1183 IPFW_RLOCK(chain);
1184 error = ipfw_dump_table(chain, tbl);
1185 IPFW_RUNLOCK(chain);
1186 if (error) {
1187 free(tbl, M_TEMP);
1188 break;
1189 }
1190 error = sooptcopyout(sopt, tbl, size);
1191 free(tbl, M_TEMP);
1192 }
1193 break;
1194
1195 /*--- NAT operations are protected by the IPFW_LOCK ---*/
1196 case IP_FW_NAT_CFG:
1197 if (IPFW_NAT_LOADED)
1198 error = ipfw_nat_cfg_ptr(sopt);
1199 else {
1200 printf("IP_FW_NAT_CFG: %s\n",
1201 "ipfw_nat not present, please load it");
1202 error = EINVAL;
1203 }
1204 break;
1205
1206 case IP_FW_NAT_DEL:
1207 if (IPFW_NAT_LOADED)
1208 error = ipfw_nat_del_ptr(sopt);
1209 else {
1210 printf("IP_FW_NAT_DEL: %s\n",
1211 "ipfw_nat not present, please load it");
1212 error = EINVAL;
1213 }
1214 break;
1215
1216 case IP_FW_NAT_GET_CONFIG:
1217 if (IPFW_NAT_LOADED)
1218 error = ipfw_nat_get_cfg_ptr(sopt);
1219 else {
1220 printf("IP_FW_NAT_GET_CFG: %s\n",
1221 "ipfw_nat not present, please load it");
1222 error = EINVAL;
1223 }
1224 break;
1225
1226 case IP_FW_NAT_GET_LOG:
1227 if (IPFW_NAT_LOADED)
1228 error = ipfw_nat_get_log_ptr(sopt);
1229 else {
1230 printf("IP_FW_NAT_GET_LOG: %s\n",
1231 "ipfw_nat not present, please load it");
1232 error = EINVAL;
1233 }
1234 break;
1235
1236 default:
1237 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
1238 error = EINVAL;
1239 }
1240
1241 return (error);
1242#undef RULE_MAXSIZE
1243}
1244
1245
1246#define RULE_MAXSIZE (256*sizeof(u_int32_t))
1247
1248/* Functions to convert rules 7.2 <==> 8.0 */
1249int
1250convert_rule_to_7(struct ip_fw *rule)
1251{
1252 /* Used to modify original rule */
1253 struct ip_fw7 *rule7 = (struct ip_fw7 *)rule;
1254 /* copy of original rule, version 8 */
1255 struct ip_fw *tmp;
1256
1257 /* Used to copy commands */
1258 ipfw_insn *ccmd, *dst;
1259 int ll = 0, ccmdlen = 0;
1260
1261 tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
1262 if (tmp == NULL) {
1263 return 1; //XXX error
1264 }
1265 bcopy(rule, tmp, RULE_MAXSIZE);
1266
1267 /* Copy fields */
1268 rule7->_pad = tmp->_pad;
1269 rule7->set = tmp->set;
1270 rule7->rulenum = tmp->rulenum;
1271 rule7->cmd_len = tmp->cmd_len;
1272 rule7->act_ofs = tmp->act_ofs;
1273 rule7->next_rule = (struct ip_fw7 *)tmp->next_rule;
1274 rule7->next = (struct ip_fw7 *)tmp->x_next;
1275 rule7->cmd_len = tmp->cmd_len;
1276 rule7->pcnt = tmp->pcnt;
1277 rule7->bcnt = tmp->bcnt;
1278 rule7->timestamp = tmp->timestamp;
1279
1280 /* Copy commands */
1281 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule7->cmd ;
1282 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
1283 ccmdlen = F_LEN(ccmd);
1284
1285 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
1286
1287 if (dst->opcode > O_NAT)
1288 /* O_REASS doesn't exists in 7.2 version, so
1289 * decrement opcode if it is after O_REASS
1290 */
1291 dst->opcode--;
1292
1293 if (ccmdlen > ll) {
1294 printf("ipfw: opcode %d size truncated\n",
1295 ccmd->opcode);
1296 return EINVAL;
1297 }
1298 }
1299 free(tmp, M_TEMP);
1300
1301 return 0;
1302}
1303
1304int
1305convert_rule_to_8(struct ip_fw *rule)
1306{
1307 /* Used to modify original rule */
1308 struct ip_fw7 *rule7 = (struct ip_fw7 *) rule;
1309
1310 /* Used to copy commands */
1311 ipfw_insn *ccmd, *dst;
1312 int ll = 0, ccmdlen = 0;
1313
1314 /* Copy of original rule */
1315 struct ip_fw7 *tmp = malloc(RULE_MAXSIZE, M_TEMP, M_NOWAIT | M_ZERO);
1316 if (tmp == NULL) {
1317 return 1; //XXX error
1318 }
1319
1320 bcopy(rule7, tmp, RULE_MAXSIZE);
1321
1322 for (ll = tmp->cmd_len, ccmd = tmp->cmd, dst = rule->cmd ;
1323 ll > 0 ; ll -= ccmdlen, ccmd += ccmdlen, dst += ccmdlen) {
1324 ccmdlen = F_LEN(ccmd);
1325
1326 bcopy(ccmd, dst, F_LEN(ccmd)*sizeof(uint32_t));
1327
1328 if (dst->opcode > O_NAT)
1329 /* O_REASS doesn't exists in 7.2 version, so
1330 * increment opcode if it is after O_REASS
1331 */
1332 dst->opcode++;
1333
1334 if (ccmdlen > ll) {
1335 printf("ipfw: opcode %d size truncated\n",
1336 ccmd->opcode);
1337 return EINVAL;
1338 }
1339 }
1340
1341 rule->_pad = tmp->_pad;
1342 rule->set = tmp->set;
1343 rule->rulenum = tmp->rulenum;
1344 rule->cmd_len = tmp->cmd_len;
1345 rule->act_ofs = tmp->act_ofs;
1346 rule->next_rule = (struct ip_fw *)tmp->next_rule;
1347 rule->x_next = (struct ip_fw *)tmp->next;
1348 rule->cmd_len = tmp->cmd_len;
1349 rule->id = 0; /* XXX see if is ok = 0 */
1350 rule->pcnt = tmp->pcnt;
1351 rule->bcnt = tmp->bcnt;
1352 rule->timestamp = tmp->timestamp;
1353
1354 free (tmp, M_TEMP);
1355 return 0;
1356}
1357
1358/* end of file */