[PATCH] fix iptables __user misannotations
[linux-2.6.22.y-op.git] / net / ipv4 / netfilter / ip_tables.c
blob78a44b01c03516bf18aecd65b7191800c42da947
1 /*
2 * Packet matching code.
4 * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
5 * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
11 * 19 Jan 2002 Harald Welte <laforge@gnumonks.org>
12 * - increase module usage count as soon as we have rules inside
13 * a table
14 * 08 Oct 2005 Harald Welte <lafore@netfilter.org>
15 * - Generalize into "x_tables" layer and "{ip,ip6,arp}_tables"
17 #include <linux/cache.h>
18 #include <linux/capability.h>
19 #include <linux/skbuff.h>
20 #include <linux/kmod.h>
21 #include <linux/vmalloc.h>
22 #include <linux/netdevice.h>
23 #include <linux/module.h>
24 #include <linux/icmp.h>
25 #include <net/ip.h>
26 #include <net/compat.h>
27 #include <asm/uaccess.h>
28 #include <linux/mutex.h>
29 #include <linux/proc_fs.h>
30 #include <linux/err.h>
31 #include <linux/cpumask.h>
33 #include <linux/netfilter/x_tables.h>
34 #include <linux/netfilter_ipv4/ip_tables.h>
36 MODULE_LICENSE("GPL");
37 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
38 MODULE_DESCRIPTION("IPv4 packet filter");
40 /*#define DEBUG_IP_FIREWALL*/
41 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
42 /*#define DEBUG_IP_FIREWALL_USER*/
44 #ifdef DEBUG_IP_FIREWALL
45 #define dprintf(format, args...) printk(format , ## args)
46 #else
47 #define dprintf(format, args...)
48 #endif
50 #ifdef DEBUG_IP_FIREWALL_USER
51 #define duprintf(format, args...) printk(format , ## args)
52 #else
53 #define duprintf(format, args...)
54 #endif
56 #ifdef CONFIG_NETFILTER_DEBUG
57 #define IP_NF_ASSERT(x) \
58 do { \
59 if (!(x)) \
60 printk("IP_NF_ASSERT: %s:%s:%u\n", \
61 __FUNCTION__, __FILE__, __LINE__); \
62 } while(0)
63 #else
64 #define IP_NF_ASSERT(x)
65 #endif
67 #if 0
68 /* All the better to debug you with... */
69 #define static
70 #define inline
71 #endif
74 We keep a set of rules for each CPU, so we can avoid write-locking
75 them in the softirq when updating the counters and therefore
76 only need to read-lock in the softirq; doing a write_lock_bh() in user
77 context stops packets coming through and allows user context to read
78 the counters or update the rules.
80 Hence the start of any table is given by get_table() below. */
82 /* Returns whether matches rule or not. */
83 static inline int
84 ip_packet_match(const struct iphdr *ip,
85 const char *indev,
86 const char *outdev,
87 const struct ipt_ip *ipinfo,
88 int isfrag)
90 size_t i;
91 unsigned long ret;
93 #define FWINV(bool,invflg) ((bool) ^ !!(ipinfo->invflags & invflg))
95 if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
96 IPT_INV_SRCIP)
97 || FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
98 IPT_INV_DSTIP)) {
99 dprintf("Source or dest mismatch.\n");
101 dprintf("SRC: %u.%u.%u.%u. Mask: %u.%u.%u.%u. Target: %u.%u.%u.%u.%s\n",
102 NIPQUAD(ip->saddr),
103 NIPQUAD(ipinfo->smsk.s_addr),
104 NIPQUAD(ipinfo->src.s_addr),
105 ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
106 dprintf("DST: %u.%u.%u.%u Mask: %u.%u.%u.%u Target: %u.%u.%u.%u.%s\n",
107 NIPQUAD(ip->daddr),
108 NIPQUAD(ipinfo->dmsk.s_addr),
109 NIPQUAD(ipinfo->dst.s_addr),
110 ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
111 return 0;
114 /* Look for ifname matches; this should unroll nicely. */
115 for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
116 ret |= (((const unsigned long *)indev)[i]
117 ^ ((const unsigned long *)ipinfo->iniface)[i])
118 & ((const unsigned long *)ipinfo->iniface_mask)[i];
121 if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
122 dprintf("VIA in mismatch (%s vs %s).%s\n",
123 indev, ipinfo->iniface,
124 ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
125 return 0;
128 for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
129 ret |= (((const unsigned long *)outdev)[i]
130 ^ ((const unsigned long *)ipinfo->outiface)[i])
131 & ((const unsigned long *)ipinfo->outiface_mask)[i];
134 if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
135 dprintf("VIA out mismatch (%s vs %s).%s\n",
136 outdev, ipinfo->outiface,
137 ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
138 return 0;
141 /* Check specific protocol */
142 if (ipinfo->proto
143 && FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
144 dprintf("Packet protocol %hi does not match %hi.%s\n",
145 ip->protocol, ipinfo->proto,
146 ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
147 return 0;
150 /* If we have a fragment rule but the packet is not a fragment
151 * then we return zero */
152 if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
153 dprintf("Fragment rule but not fragment.%s\n",
154 ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
155 return 0;
158 return 1;
161 static inline int
162 ip_checkentry(const struct ipt_ip *ip)
164 if (ip->flags & ~IPT_F_MASK) {
165 duprintf("Unknown flag bits set: %08X\n",
166 ip->flags & ~IPT_F_MASK);
167 return 0;
169 if (ip->invflags & ~IPT_INV_MASK) {
170 duprintf("Unknown invflag bits set: %08X\n",
171 ip->invflags & ~IPT_INV_MASK);
172 return 0;
174 return 1;
177 static unsigned int
178 ipt_error(struct sk_buff **pskb,
179 const struct net_device *in,
180 const struct net_device *out,
181 unsigned int hooknum,
182 const struct xt_target *target,
183 const void *targinfo)
185 if (net_ratelimit())
186 printk("ip_tables: error: `%s'\n", (char *)targinfo);
188 return NF_DROP;
191 static inline
192 int do_match(struct ipt_entry_match *m,
193 const struct sk_buff *skb,
194 const struct net_device *in,
195 const struct net_device *out,
196 int offset,
197 int *hotdrop)
199 /* Stop iteration if it doesn't match */
200 if (!m->u.kernel.match->match(skb, in, out, m->u.kernel.match, m->data,
201 offset, skb->nh.iph->ihl*4, hotdrop))
202 return 1;
203 else
204 return 0;
207 static inline struct ipt_entry *
208 get_entry(void *base, unsigned int offset)
210 return (struct ipt_entry *)(base + offset);
213 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
214 unsigned int
215 ipt_do_table(struct sk_buff **pskb,
216 unsigned int hook,
217 const struct net_device *in,
218 const struct net_device *out,
219 struct ipt_table *table)
221 static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
222 u_int16_t offset;
223 struct iphdr *ip;
224 u_int16_t datalen;
225 int hotdrop = 0;
226 /* Initializing verdict to NF_DROP keeps gcc happy. */
227 unsigned int verdict = NF_DROP;
228 const char *indev, *outdev;
229 void *table_base;
230 struct ipt_entry *e, *back;
231 struct xt_table_info *private;
233 /* Initialization */
234 ip = (*pskb)->nh.iph;
235 datalen = (*pskb)->len - ip->ihl * 4;
236 indev = in ? in->name : nulldevname;
237 outdev = out ? out->name : nulldevname;
238 /* We handle fragments by dealing with the first fragment as
239 * if it was a normal packet. All other fragments are treated
240 * normally, except that they will NEVER match rules that ask
241 * things we don't know, ie. tcp syn flag or ports). If the
242 * rule is also a fragment-specific rule, non-fragments won't
243 * match it. */
244 offset = ntohs(ip->frag_off) & IP_OFFSET;
246 read_lock_bh(&table->lock);
247 IP_NF_ASSERT(table->valid_hooks & (1 << hook));
248 private = table->private;
249 table_base = (void *)private->entries[smp_processor_id()];
250 e = get_entry(table_base, private->hook_entry[hook]);
252 /* For return from builtin chain */
253 back = get_entry(table_base, private->underflow[hook]);
255 do {
256 IP_NF_ASSERT(e);
257 IP_NF_ASSERT(back);
258 if (ip_packet_match(ip, indev, outdev, &e->ip, offset)) {
259 struct ipt_entry_target *t;
261 if (IPT_MATCH_ITERATE(e, do_match,
262 *pskb, in, out,
263 offset, &hotdrop) != 0)
264 goto no_match;
266 ADD_COUNTER(e->counters, ntohs(ip->tot_len), 1);
268 t = ipt_get_target(e);
269 IP_NF_ASSERT(t->u.kernel.target);
270 /* Standard target? */
271 if (!t->u.kernel.target->target) {
272 int v;
274 v = ((struct ipt_standard_target *)t)->verdict;
275 if (v < 0) {
276 /* Pop from stack? */
277 if (v != IPT_RETURN) {
278 verdict = (unsigned)(-v) - 1;
279 break;
281 e = back;
282 back = get_entry(table_base,
283 back->comefrom);
284 continue;
286 if (table_base + v != (void *)e + e->next_offset
287 && !(e->ip.flags & IPT_F_GOTO)) {
288 /* Save old back ptr in next entry */
289 struct ipt_entry *next
290 = (void *)e + e->next_offset;
291 next->comefrom
292 = (void *)back - table_base;
293 /* set back pointer to next entry */
294 back = next;
297 e = get_entry(table_base, v);
298 } else {
299 /* Targets which reenter must return
300 abs. verdicts */
301 #ifdef CONFIG_NETFILTER_DEBUG
302 ((struct ipt_entry *)table_base)->comefrom
303 = 0xeeeeeeec;
304 #endif
305 verdict = t->u.kernel.target->target(pskb,
306 in, out,
307 hook,
308 t->u.kernel.target,
309 t->data);
311 #ifdef CONFIG_NETFILTER_DEBUG
312 if (((struct ipt_entry *)table_base)->comefrom
313 != 0xeeeeeeec
314 && verdict == IPT_CONTINUE) {
315 printk("Target %s reentered!\n",
316 t->u.kernel.target->name);
317 verdict = NF_DROP;
319 ((struct ipt_entry *)table_base)->comefrom
320 = 0x57acc001;
321 #endif
322 /* Target might have changed stuff. */
323 ip = (*pskb)->nh.iph;
324 datalen = (*pskb)->len - ip->ihl * 4;
326 if (verdict == IPT_CONTINUE)
327 e = (void *)e + e->next_offset;
328 else
329 /* Verdict */
330 break;
332 } else {
334 no_match:
335 e = (void *)e + e->next_offset;
337 } while (!hotdrop);
339 read_unlock_bh(&table->lock);
341 #ifdef DEBUG_ALLOW_ALL
342 return NF_ACCEPT;
343 #else
344 if (hotdrop)
345 return NF_DROP;
346 else return verdict;
347 #endif
350 /* All zeroes == unconditional rule. */
351 static inline int
352 unconditional(const struct ipt_ip *ip)
354 unsigned int i;
356 for (i = 0; i < sizeof(*ip)/sizeof(__u32); i++)
357 if (((__u32 *)ip)[i])
358 return 0;
360 return 1;
363 /* Figures out from what hook each rule can be called: returns 0 if
364 there are loops. Puts hook bitmask in comefrom. */
365 static int
366 mark_source_chains(struct xt_table_info *newinfo,
367 unsigned int valid_hooks, void *entry0)
369 unsigned int hook;
371 /* No recursion; use packet counter to save back ptrs (reset
372 to 0 as we leave), and comefrom to save source hook bitmask */
373 for (hook = 0; hook < NF_IP_NUMHOOKS; hook++) {
374 unsigned int pos = newinfo->hook_entry[hook];
375 struct ipt_entry *e
376 = (struct ipt_entry *)(entry0 + pos);
378 if (!(valid_hooks & (1 << hook)))
379 continue;
381 /* Set initial back pointer. */
382 e->counters.pcnt = pos;
384 for (;;) {
385 struct ipt_standard_target *t
386 = (void *)ipt_get_target(e);
388 if (e->comefrom & (1 << NF_IP_NUMHOOKS)) {
389 printk("iptables: loop hook %u pos %u %08X.\n",
390 hook, pos, e->comefrom);
391 return 0;
393 e->comefrom
394 |= ((1 << hook) | (1 << NF_IP_NUMHOOKS));
396 /* Unconditional return/END. */
397 if (e->target_offset == sizeof(struct ipt_entry)
398 && (strcmp(t->target.u.user.name,
399 IPT_STANDARD_TARGET) == 0)
400 && t->verdict < 0
401 && unconditional(&e->ip)) {
402 unsigned int oldpos, size;
404 /* Return: backtrack through the last
405 big jump. */
406 do {
407 e->comefrom ^= (1<<NF_IP_NUMHOOKS);
408 #ifdef DEBUG_IP_FIREWALL_USER
409 if (e->comefrom
410 & (1 << NF_IP_NUMHOOKS)) {
411 duprintf("Back unset "
412 "on hook %u "
413 "rule %u\n",
414 hook, pos);
416 #endif
417 oldpos = pos;
418 pos = e->counters.pcnt;
419 e->counters.pcnt = 0;
421 /* We're at the start. */
422 if (pos == oldpos)
423 goto next;
425 e = (struct ipt_entry *)
426 (entry0 + pos);
427 } while (oldpos == pos + e->next_offset);
429 /* Move along one */
430 size = e->next_offset;
431 e = (struct ipt_entry *)
432 (entry0 + pos + size);
433 e->counters.pcnt = pos;
434 pos += size;
435 } else {
436 int newpos = t->verdict;
438 if (strcmp(t->target.u.user.name,
439 IPT_STANDARD_TARGET) == 0
440 && newpos >= 0) {
441 /* This a jump; chase it. */
442 duprintf("Jump rule %u -> %u\n",
443 pos, newpos);
444 } else {
445 /* ... this is a fallthru */
446 newpos = pos + e->next_offset;
448 e = (struct ipt_entry *)
449 (entry0 + newpos);
450 e->counters.pcnt = pos;
451 pos = newpos;
454 next:
455 duprintf("Finished chain %u\n", hook);
457 return 1;
460 static inline int
461 cleanup_match(struct ipt_entry_match *m, unsigned int *i)
463 if (i && (*i)-- == 0)
464 return 1;
466 if (m->u.kernel.match->destroy)
467 m->u.kernel.match->destroy(m->u.kernel.match, m->data);
468 module_put(m->u.kernel.match->me);
469 return 0;
472 static inline int
473 standard_check(const struct ipt_entry_target *t,
474 unsigned int max_offset)
476 struct ipt_standard_target *targ = (void *)t;
478 /* Check standard info. */
479 if (targ->verdict >= 0
480 && targ->verdict > max_offset - sizeof(struct ipt_entry)) {
481 duprintf("ipt_standard_check: bad verdict (%i)\n",
482 targ->verdict);
483 return 0;
485 if (targ->verdict < -NF_MAX_VERDICT - 1) {
486 duprintf("ipt_standard_check: bad negative verdict (%i)\n",
487 targ->verdict);
488 return 0;
490 return 1;
493 static inline int
494 check_match(struct ipt_entry_match *m,
495 const char *name,
496 const struct ipt_ip *ip,
497 unsigned int hookmask,
498 unsigned int *i)
500 struct ipt_match *match;
501 int ret;
503 match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
504 m->u.user.revision),
505 "ipt_%s", m->u.user.name);
506 if (IS_ERR(match) || !match) {
507 duprintf("check_match: `%s' not found\n", m->u.user.name);
508 return match ? PTR_ERR(match) : -ENOENT;
510 m->u.kernel.match = match;
512 ret = xt_check_match(match, AF_INET, m->u.match_size - sizeof(*m),
513 name, hookmask, ip->proto,
514 ip->invflags & IPT_INV_PROTO);
515 if (ret)
516 goto err;
518 if (m->u.kernel.match->checkentry
519 && !m->u.kernel.match->checkentry(name, ip, match, m->data,
520 hookmask)) {
521 duprintf("ip_tables: check failed for `%s'.\n",
522 m->u.kernel.match->name);
523 ret = -EINVAL;
524 goto err;
527 (*i)++;
528 return 0;
529 err:
530 module_put(m->u.kernel.match->me);
531 return ret;
534 static struct ipt_target ipt_standard_target;
536 static inline int
537 check_entry(struct ipt_entry *e, const char *name, unsigned int size,
538 unsigned int *i)
540 struct ipt_entry_target *t;
541 struct ipt_target *target;
542 int ret;
543 unsigned int j;
545 if (!ip_checkentry(&e->ip)) {
546 duprintf("ip_tables: ip check failed %p %s.\n", e, name);
547 return -EINVAL;
550 j = 0;
551 ret = IPT_MATCH_ITERATE(e, check_match, name, &e->ip, e->comefrom, &j);
552 if (ret != 0)
553 goto cleanup_matches;
555 t = ipt_get_target(e);
556 target = try_then_request_module(xt_find_target(AF_INET,
557 t->u.user.name,
558 t->u.user.revision),
559 "ipt_%s", t->u.user.name);
560 if (IS_ERR(target) || !target) {
561 duprintf("check_entry: `%s' not found\n", t->u.user.name);
562 ret = target ? PTR_ERR(target) : -ENOENT;
563 goto cleanup_matches;
565 t->u.kernel.target = target;
567 ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
568 name, e->comefrom, e->ip.proto,
569 e->ip.invflags & IPT_INV_PROTO);
570 if (ret)
571 goto err;
573 if (t->u.kernel.target == &ipt_standard_target) {
574 if (!standard_check(t, size)) {
575 ret = -EINVAL;
576 goto err;
578 } else if (t->u.kernel.target->checkentry
579 && !t->u.kernel.target->checkentry(name, e, target, t->data,
580 e->comefrom)) {
581 duprintf("ip_tables: check failed for `%s'.\n",
582 t->u.kernel.target->name);
583 ret = -EINVAL;
584 goto err;
587 (*i)++;
588 return 0;
589 err:
590 module_put(t->u.kernel.target->me);
591 cleanup_matches:
592 IPT_MATCH_ITERATE(e, cleanup_match, &j);
593 return ret;
596 static inline int
597 check_entry_size_and_hooks(struct ipt_entry *e,
598 struct xt_table_info *newinfo,
599 unsigned char *base,
600 unsigned char *limit,
601 const unsigned int *hook_entries,
602 const unsigned int *underflows,
603 unsigned int *i)
605 unsigned int h;
607 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0
608 || (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
609 duprintf("Bad offset %p\n", e);
610 return -EINVAL;
613 if (e->next_offset
614 < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) {
615 duprintf("checking: element %p size %u\n",
616 e, e->next_offset);
617 return -EINVAL;
620 /* Check hooks & underflows */
621 for (h = 0; h < NF_IP_NUMHOOKS; h++) {
622 if ((unsigned char *)e - base == hook_entries[h])
623 newinfo->hook_entry[h] = hook_entries[h];
624 if ((unsigned char *)e - base == underflows[h])
625 newinfo->underflow[h] = underflows[h];
628 /* FIXME: underflows must be unconditional, standard verdicts
629 < 0 (not IPT_RETURN). --RR */
631 /* Clear counters and comefrom */
632 e->counters = ((struct xt_counters) { 0, 0 });
633 e->comefrom = 0;
635 (*i)++;
636 return 0;
639 static inline int
640 cleanup_entry(struct ipt_entry *e, unsigned int *i)
642 struct ipt_entry_target *t;
644 if (i && (*i)-- == 0)
645 return 1;
647 /* Cleanup all matches */
648 IPT_MATCH_ITERATE(e, cleanup_match, NULL);
649 t = ipt_get_target(e);
650 if (t->u.kernel.target->destroy)
651 t->u.kernel.target->destroy(t->u.kernel.target, t->data);
652 module_put(t->u.kernel.target->me);
653 return 0;
656 /* Checks and translates the user-supplied table segment (held in
657 newinfo) */
658 static int
659 translate_table(const char *name,
660 unsigned int valid_hooks,
661 struct xt_table_info *newinfo,
662 void *entry0,
663 unsigned int size,
664 unsigned int number,
665 const unsigned int *hook_entries,
666 const unsigned int *underflows)
668 unsigned int i;
669 int ret;
671 newinfo->size = size;
672 newinfo->number = number;
674 /* Init all hooks to impossible value. */
675 for (i = 0; i < NF_IP_NUMHOOKS; i++) {
676 newinfo->hook_entry[i] = 0xFFFFFFFF;
677 newinfo->underflow[i] = 0xFFFFFFFF;
680 duprintf("translate_table: size %u\n", newinfo->size);
681 i = 0;
682 /* Walk through entries, checking offsets. */
683 ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
684 check_entry_size_and_hooks,
685 newinfo,
686 entry0,
687 entry0 + size,
688 hook_entries, underflows, &i);
689 if (ret != 0)
690 return ret;
692 if (i != number) {
693 duprintf("translate_table: %u not %u entries\n",
694 i, number);
695 return -EINVAL;
698 /* Check hooks all assigned */
699 for (i = 0; i < NF_IP_NUMHOOKS; i++) {
700 /* Only hooks which are valid */
701 if (!(valid_hooks & (1 << i)))
702 continue;
703 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
704 duprintf("Invalid hook entry %u %u\n",
705 i, hook_entries[i]);
706 return -EINVAL;
708 if (newinfo->underflow[i] == 0xFFFFFFFF) {
709 duprintf("Invalid underflow %u %u\n",
710 i, underflows[i]);
711 return -EINVAL;
715 if (!mark_source_chains(newinfo, valid_hooks, entry0))
716 return -ELOOP;
718 /* Finally, each sanity check must pass */
719 i = 0;
720 ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
721 check_entry, name, size, &i);
723 if (ret != 0) {
724 IPT_ENTRY_ITERATE(entry0, newinfo->size,
725 cleanup_entry, &i);
726 return ret;
729 /* And one copy for every other CPU */
730 for_each_possible_cpu(i) {
731 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
732 memcpy(newinfo->entries[i], entry0, newinfo->size);
735 return ret;
738 /* Gets counters. */
739 static inline int
740 add_entry_to_counter(const struct ipt_entry *e,
741 struct xt_counters total[],
742 unsigned int *i)
744 ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
746 (*i)++;
747 return 0;
750 static inline int
751 set_entry_to_counter(const struct ipt_entry *e,
752 struct ipt_counters total[],
753 unsigned int *i)
755 SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
757 (*i)++;
758 return 0;
761 static void
762 get_counters(const struct xt_table_info *t,
763 struct xt_counters counters[])
765 unsigned int cpu;
766 unsigned int i;
767 unsigned int curcpu;
769 /* Instead of clearing (by a previous call to memset())
770 * the counters and using adds, we set the counters
771 * with data used by 'current' CPU
772 * We dont care about preemption here.
774 curcpu = raw_smp_processor_id();
776 i = 0;
777 IPT_ENTRY_ITERATE(t->entries[curcpu],
778 t->size,
779 set_entry_to_counter,
780 counters,
781 &i);
783 for_each_possible_cpu(cpu) {
784 if (cpu == curcpu)
785 continue;
786 i = 0;
787 IPT_ENTRY_ITERATE(t->entries[cpu],
788 t->size,
789 add_entry_to_counter,
790 counters,
791 &i);
795 static inline struct xt_counters * alloc_counters(struct ipt_table *table)
797 unsigned int countersize;
798 struct xt_counters *counters;
799 struct xt_table_info *private = table->private;
801 /* We need atomic snapshot of counters: rest doesn't change
802 (other than comefrom, which userspace doesn't care
803 about). */
804 countersize = sizeof(struct xt_counters) * private->number;
805 counters = vmalloc_node(countersize, numa_node_id());
807 if (counters == NULL)
808 return ERR_PTR(-ENOMEM);
810 /* First, sum counters... */
811 write_lock_bh(&table->lock);
812 get_counters(private, counters);
813 write_unlock_bh(&table->lock);
815 return counters;
818 static int
819 copy_entries_to_user(unsigned int total_size,
820 struct ipt_table *table,
821 void __user *userptr)
823 unsigned int off, num;
824 struct ipt_entry *e;
825 struct xt_counters *counters;
826 struct xt_table_info *private = table->private;
827 int ret = 0;
828 void *loc_cpu_entry;
830 counters = alloc_counters(table);
831 if (IS_ERR(counters))
832 return PTR_ERR(counters);
834 /* choose the copy that is on our node/cpu, ...
835 * This choice is lazy (because current thread is
836 * allowed to migrate to another cpu)
838 loc_cpu_entry = private->entries[raw_smp_processor_id()];
839 /* ... then copy entire thing ... */
840 if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
841 ret = -EFAULT;
842 goto free_counters;
845 /* FIXME: use iterator macros --RR */
846 /* ... then go back and fix counters and names */
847 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
848 unsigned int i;
849 struct ipt_entry_match *m;
850 struct ipt_entry_target *t;
852 e = (struct ipt_entry *)(loc_cpu_entry + off);
853 if (copy_to_user(userptr + off
854 + offsetof(struct ipt_entry, counters),
855 &counters[num],
856 sizeof(counters[num])) != 0) {
857 ret = -EFAULT;
858 goto free_counters;
861 for (i = sizeof(struct ipt_entry);
862 i < e->target_offset;
863 i += m->u.match_size) {
864 m = (void *)e + i;
866 if (copy_to_user(userptr + off + i
867 + offsetof(struct ipt_entry_match,
868 u.user.name),
869 m->u.kernel.match->name,
870 strlen(m->u.kernel.match->name)+1)
871 != 0) {
872 ret = -EFAULT;
873 goto free_counters;
877 t = ipt_get_target(e);
878 if (copy_to_user(userptr + off + e->target_offset
879 + offsetof(struct ipt_entry_target,
880 u.user.name),
881 t->u.kernel.target->name,
882 strlen(t->u.kernel.target->name)+1) != 0) {
883 ret = -EFAULT;
884 goto free_counters;
888 free_counters:
889 vfree(counters);
890 return ret;
893 #ifdef CONFIG_COMPAT
894 struct compat_delta {
895 struct compat_delta *next;
896 u_int16_t offset;
897 short delta;
900 static struct compat_delta *compat_offsets = NULL;
902 static int compat_add_offset(u_int16_t offset, short delta)
904 struct compat_delta *tmp;
906 tmp = kmalloc(sizeof(struct compat_delta), GFP_KERNEL);
907 if (!tmp)
908 return -ENOMEM;
909 tmp->offset = offset;
910 tmp->delta = delta;
911 if (compat_offsets) {
912 tmp->next = compat_offsets->next;
913 compat_offsets->next = tmp;
914 } else {
915 compat_offsets = tmp;
916 tmp->next = NULL;
918 return 0;
921 static void compat_flush_offsets(void)
923 struct compat_delta *tmp, *next;
925 if (compat_offsets) {
926 for(tmp = compat_offsets; tmp; tmp = next) {
927 next = tmp->next;
928 kfree(tmp);
930 compat_offsets = NULL;
934 static short compat_calc_jump(u_int16_t offset)
936 struct compat_delta *tmp;
937 short delta;
939 for(tmp = compat_offsets, delta = 0; tmp; tmp = tmp->next)
940 if (tmp->offset < offset)
941 delta += tmp->delta;
942 return delta;
945 static void compat_standard_from_user(void *dst, void *src)
947 int v = *(compat_int_t *)src;
949 if (v > 0)
950 v += compat_calc_jump(v);
951 memcpy(dst, &v, sizeof(v));
954 static int compat_standard_to_user(void __user *dst, void *src)
956 compat_int_t cv = *(int *)src;
958 if (cv > 0)
959 cv -= compat_calc_jump(cv);
960 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
963 static inline int
964 compat_calc_match(struct ipt_entry_match *m, int * size)
966 *size += xt_compat_match_offset(m->u.kernel.match);
967 return 0;
970 static int compat_calc_entry(struct ipt_entry *e, struct xt_table_info *info,
971 void *base, struct xt_table_info *newinfo)
973 struct ipt_entry_target *t;
974 u_int16_t entry_offset;
975 int off, i, ret;
977 off = 0;
978 entry_offset = (void *)e - base;
979 IPT_MATCH_ITERATE(e, compat_calc_match, &off);
980 t = ipt_get_target(e);
981 off += xt_compat_target_offset(t->u.kernel.target);
982 newinfo->size -= off;
983 ret = compat_add_offset(entry_offset, off);
984 if (ret)
985 return ret;
987 for (i = 0; i< NF_IP_NUMHOOKS; i++) {
988 if (info->hook_entry[i] && (e < (struct ipt_entry *)
989 (base + info->hook_entry[i])))
990 newinfo->hook_entry[i] -= off;
991 if (info->underflow[i] && (e < (struct ipt_entry *)
992 (base + info->underflow[i])))
993 newinfo->underflow[i] -= off;
995 return 0;
998 static int compat_table_info(struct xt_table_info *info,
999 struct xt_table_info *newinfo)
1001 void *loc_cpu_entry;
1002 int i;
1004 if (!newinfo || !info)
1005 return -EINVAL;
1007 memset(newinfo, 0, sizeof(struct xt_table_info));
1008 newinfo->size = info->size;
1009 newinfo->number = info->number;
1010 for (i = 0; i < NF_IP_NUMHOOKS; i++) {
1011 newinfo->hook_entry[i] = info->hook_entry[i];
1012 newinfo->underflow[i] = info->underflow[i];
1014 loc_cpu_entry = info->entries[raw_smp_processor_id()];
1015 return IPT_ENTRY_ITERATE(loc_cpu_entry, info->size,
1016 compat_calc_entry, info, loc_cpu_entry, newinfo);
1018 #endif
1020 static int get_info(void __user *user, int *len, int compat)
1022 char name[IPT_TABLE_MAXNAMELEN];
1023 struct ipt_table *t;
1024 int ret;
1026 if (*len != sizeof(struct ipt_getinfo)) {
1027 duprintf("length %u != %u\n", *len,
1028 (unsigned int)sizeof(struct ipt_getinfo));
1029 return -EINVAL;
1032 if (copy_from_user(name, user, sizeof(name)) != 0)
1033 return -EFAULT;
1035 name[IPT_TABLE_MAXNAMELEN-1] = '\0';
1036 #ifdef CONFIG_COMPAT
1037 if (compat)
1038 xt_compat_lock(AF_INET);
1039 #endif
1040 t = try_then_request_module(xt_find_table_lock(AF_INET, name),
1041 "iptable_%s", name);
1042 if (t && !IS_ERR(t)) {
1043 struct ipt_getinfo info;
1044 struct xt_table_info *private = t->private;
1046 #ifdef CONFIG_COMPAT
1047 if (compat) {
1048 struct xt_table_info tmp;
1049 ret = compat_table_info(private, &tmp);
1050 compat_flush_offsets();
1051 private = &tmp;
1053 #endif
1054 info.valid_hooks = t->valid_hooks;
1055 memcpy(info.hook_entry, private->hook_entry,
1056 sizeof(info.hook_entry));
1057 memcpy(info.underflow, private->underflow,
1058 sizeof(info.underflow));
1059 info.num_entries = private->number;
1060 info.size = private->size;
1061 strcpy(info.name, name);
1063 if (copy_to_user(user, &info, *len) != 0)
1064 ret = -EFAULT;
1065 else
1066 ret = 0;
1068 xt_table_unlock(t);
1069 module_put(t->me);
1070 } else
1071 ret = t ? PTR_ERR(t) : -ENOENT;
1072 #ifdef CONFIG_COMPAT
1073 if (compat)
1074 xt_compat_unlock(AF_INET);
1075 #endif
1076 return ret;
1079 static int
1080 get_entries(struct ipt_get_entries __user *uptr, int *len)
1082 int ret;
1083 struct ipt_get_entries get;
1084 struct ipt_table *t;
1086 if (*len < sizeof(get)) {
1087 duprintf("get_entries: %u < %d\n", *len,
1088 (unsigned int)sizeof(get));
1089 return -EINVAL;
1091 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1092 return -EFAULT;
1093 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1094 duprintf("get_entries: %u != %u\n", *len,
1095 (unsigned int)(sizeof(struct ipt_get_entries) +
1096 get.size));
1097 return -EINVAL;
1100 t = xt_find_table_lock(AF_INET, get.name);
1101 if (t && !IS_ERR(t)) {
1102 struct xt_table_info *private = t->private;
1103 duprintf("t->private->number = %u\n",
1104 private->number);
1105 if (get.size == private->size)
1106 ret = copy_entries_to_user(private->size,
1107 t, uptr->entrytable);
1108 else {
1109 duprintf("get_entries: I've got %u not %u!\n",
1110 private->size,
1111 get.size);
1112 ret = -EINVAL;
1114 module_put(t->me);
1115 xt_table_unlock(t);
1116 } else
1117 ret = t ? PTR_ERR(t) : -ENOENT;
1119 return ret;
1122 static int
1123 __do_replace(const char *name, unsigned int valid_hooks,
1124 struct xt_table_info *newinfo, unsigned int num_counters,
1125 void __user *counters_ptr)
1127 int ret;
1128 struct ipt_table *t;
1129 struct xt_table_info *oldinfo;
1130 struct xt_counters *counters;
1131 void *loc_cpu_old_entry;
1133 ret = 0;
1134 counters = vmalloc(num_counters * sizeof(struct xt_counters));
1135 if (!counters) {
1136 ret = -ENOMEM;
1137 goto out;
1140 t = try_then_request_module(xt_find_table_lock(AF_INET, name),
1141 "iptable_%s", name);
1142 if (!t || IS_ERR(t)) {
1143 ret = t ? PTR_ERR(t) : -ENOENT;
1144 goto free_newinfo_counters_untrans;
1147 /* You lied! */
1148 if (valid_hooks != t->valid_hooks) {
1149 duprintf("Valid hook crap: %08X vs %08X\n",
1150 valid_hooks, t->valid_hooks);
1151 ret = -EINVAL;
1152 goto put_module;
1155 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1156 if (!oldinfo)
1157 goto put_module;
1159 /* Update module usage count based on number of rules */
1160 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1161 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1162 if ((oldinfo->number > oldinfo->initial_entries) ||
1163 (newinfo->number <= oldinfo->initial_entries))
1164 module_put(t->me);
1165 if ((oldinfo->number > oldinfo->initial_entries) &&
1166 (newinfo->number <= oldinfo->initial_entries))
1167 module_put(t->me);
1169 /* Get the old counters. */
1170 get_counters(oldinfo, counters);
1171 /* Decrease module usage counts and free resource */
1172 loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1173 IPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,NULL);
1174 xt_free_table_info(oldinfo);
1175 if (copy_to_user(counters_ptr, counters,
1176 sizeof(struct xt_counters) * num_counters) != 0)
1177 ret = -EFAULT;
1178 vfree(counters);
1179 xt_table_unlock(t);
1180 return ret;
1182 put_module:
1183 module_put(t->me);
1184 xt_table_unlock(t);
1185 free_newinfo_counters_untrans:
1186 vfree(counters);
1187 out:
1188 return ret;
1191 static int
1192 do_replace(void __user *user, unsigned int len)
1194 int ret;
1195 struct ipt_replace tmp;
1196 struct xt_table_info *newinfo;
1197 void *loc_cpu_entry;
1199 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1200 return -EFAULT;
1202 /* Hack: Causes ipchains to give correct error msg --RR */
1203 if (len != sizeof(tmp) + tmp.size)
1204 return -ENOPROTOOPT;
1206 /* overflow check */
1207 if (tmp.size >= (INT_MAX - sizeof(struct xt_table_info)) / NR_CPUS -
1208 SMP_CACHE_BYTES)
1209 return -ENOMEM;
1210 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1211 return -ENOMEM;
1213 newinfo = xt_alloc_table_info(tmp.size);
1214 if (!newinfo)
1215 return -ENOMEM;
1217 /* choose the copy that is our node/cpu */
1218 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1219 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1220 tmp.size) != 0) {
1221 ret = -EFAULT;
1222 goto free_newinfo;
1225 ret = translate_table(tmp.name, tmp.valid_hooks,
1226 newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
1227 tmp.hook_entry, tmp.underflow);
1228 if (ret != 0)
1229 goto free_newinfo;
1231 duprintf("ip_tables: Translated table\n");
1233 ret = __do_replace(tmp.name, tmp.valid_hooks,
1234 newinfo, tmp.num_counters,
1235 tmp.counters);
1236 if (ret)
1237 goto free_newinfo_untrans;
1238 return 0;
1240 free_newinfo_untrans:
1241 IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry,NULL);
1242 free_newinfo:
1243 xt_free_table_info(newinfo);
1244 return ret;
1247 /* We're lazy, and add to the first CPU; overflow works its fey magic
1248 * and everything is OK. */
1249 static inline int
1250 add_counter_to_entry(struct ipt_entry *e,
1251 const struct xt_counters addme[],
1252 unsigned int *i)
1254 #if 0
1255 duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n",
1257 (long unsigned int)e->counters.pcnt,
1258 (long unsigned int)e->counters.bcnt,
1259 (long unsigned int)addme[*i].pcnt,
1260 (long unsigned int)addme[*i].bcnt);
1261 #endif
1263 ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
1265 (*i)++;
1266 return 0;
1269 static int
1270 do_add_counters(void __user *user, unsigned int len, int compat)
1272 unsigned int i;
1273 struct xt_counters_info tmp;
1274 struct xt_counters *paddc;
1275 unsigned int num_counters;
1276 char *name;
1277 int size;
1278 void *ptmp;
1279 struct ipt_table *t;
1280 struct xt_table_info *private;
1281 int ret = 0;
1282 void *loc_cpu_entry;
1283 #ifdef CONFIG_COMPAT
1284 struct compat_xt_counters_info compat_tmp;
1286 if (compat) {
1287 ptmp = &compat_tmp;
1288 size = sizeof(struct compat_xt_counters_info);
1289 } else
1290 #endif
1292 ptmp = &tmp;
1293 size = sizeof(struct xt_counters_info);
1296 if (copy_from_user(ptmp, user, size) != 0)
1297 return -EFAULT;
1299 #ifdef CONFIG_COMPAT
1300 if (compat) {
1301 num_counters = compat_tmp.num_counters;
1302 name = compat_tmp.name;
1303 } else
1304 #endif
1306 num_counters = tmp.num_counters;
1307 name = tmp.name;
1310 if (len != size + num_counters * sizeof(struct xt_counters))
1311 return -EINVAL;
1313 paddc = vmalloc_node(len - size, numa_node_id());
1314 if (!paddc)
1315 return -ENOMEM;
1317 if (copy_from_user(paddc, user + size, len - size) != 0) {
1318 ret = -EFAULT;
1319 goto free;
1322 t = xt_find_table_lock(AF_INET, name);
1323 if (!t || IS_ERR(t)) {
1324 ret = t ? PTR_ERR(t) : -ENOENT;
1325 goto free;
1328 write_lock_bh(&t->lock);
1329 private = t->private;
1330 if (private->number != num_counters) {
1331 ret = -EINVAL;
1332 goto unlock_up_free;
1335 i = 0;
1336 /* Choose the copy that is on our node */
1337 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1338 IPT_ENTRY_ITERATE(loc_cpu_entry,
1339 private->size,
1340 add_counter_to_entry,
1341 paddc,
1342 &i);
1343 unlock_up_free:
1344 write_unlock_bh(&t->lock);
1345 xt_table_unlock(t);
1346 module_put(t->me);
1347 free:
1348 vfree(paddc);
1350 return ret;
1353 #ifdef CONFIG_COMPAT
1354 struct compat_ipt_replace {
1355 char name[IPT_TABLE_MAXNAMELEN];
1356 u32 valid_hooks;
1357 u32 num_entries;
1358 u32 size;
1359 u32 hook_entry[NF_IP_NUMHOOKS];
1360 u32 underflow[NF_IP_NUMHOOKS];
1361 u32 num_counters;
1362 compat_uptr_t counters; /* struct ipt_counters * */
1363 struct compat_ipt_entry entries[0];
1366 static inline int compat_copy_match_to_user(struct ipt_entry_match *m,
1367 void __user **dstptr, compat_uint_t *size)
1369 return xt_compat_match_to_user(m, dstptr, size);
1372 static int compat_copy_entry_to_user(struct ipt_entry *e,
1373 void __user **dstptr, compat_uint_t *size)
1375 struct ipt_entry_target *t;
1376 struct compat_ipt_entry __user *ce;
1377 u_int16_t target_offset, next_offset;
1378 compat_uint_t origsize;
1379 int ret;
1381 ret = -EFAULT;
1382 origsize = *size;
1383 ce = (struct compat_ipt_entry __user *)*dstptr;
1384 if (copy_to_user(ce, e, sizeof(struct ipt_entry)))
1385 goto out;
1387 *dstptr += sizeof(struct compat_ipt_entry);
1388 ret = IPT_MATCH_ITERATE(e, compat_copy_match_to_user, dstptr, size);
1389 target_offset = e->target_offset - (origsize - *size);
1390 if (ret)
1391 goto out;
1392 t = ipt_get_target(e);
1393 ret = xt_compat_target_to_user(t, dstptr, size);
1394 if (ret)
1395 goto out;
1396 ret = -EFAULT;
1397 next_offset = e->next_offset - (origsize - *size);
1398 if (put_user(target_offset, &ce->target_offset))
1399 goto out;
1400 if (put_user(next_offset, &ce->next_offset))
1401 goto out;
1402 return 0;
1403 out:
1404 return ret;
1407 static inline int
1408 compat_check_calc_match(struct ipt_entry_match *m,
1409 const char *name,
1410 const struct ipt_ip *ip,
1411 unsigned int hookmask,
1412 int *size, int *i)
1414 struct ipt_match *match;
1416 match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
1417 m->u.user.revision),
1418 "ipt_%s", m->u.user.name);
1419 if (IS_ERR(match) || !match) {
1420 duprintf("compat_check_calc_match: `%s' not found\n",
1421 m->u.user.name);
1422 return match ? PTR_ERR(match) : -ENOENT;
1424 m->u.kernel.match = match;
1425 *size += xt_compat_match_offset(match);
1427 (*i)++;
1428 return 0;
1431 static inline int
1432 check_compat_entry_size_and_hooks(struct ipt_entry *e,
1433 struct xt_table_info *newinfo,
1434 unsigned int *size,
1435 unsigned char *base,
1436 unsigned char *limit,
1437 unsigned int *hook_entries,
1438 unsigned int *underflows,
1439 unsigned int *i,
1440 const char *name)
1442 struct ipt_entry_target *t;
1443 struct ipt_target *target;
1444 u_int16_t entry_offset;
1445 int ret, off, h, j;
1447 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1448 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0
1449 || (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1450 duprintf("Bad offset %p, limit = %p\n", e, limit);
1451 return -EINVAL;
1454 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1455 sizeof(struct compat_xt_entry_target)) {
1456 duprintf("checking: element %p size %u\n",
1457 e, e->next_offset);
1458 return -EINVAL;
1461 if (!ip_checkentry(&e->ip)) {
1462 duprintf("ip_tables: ip check failed %p %s.\n", e, name);
1463 return -EINVAL;
1466 off = 0;
1467 entry_offset = (void *)e - (void *)base;
1468 j = 0;
1469 ret = IPT_MATCH_ITERATE(e, compat_check_calc_match, name, &e->ip,
1470 e->comefrom, &off, &j);
1471 if (ret != 0)
1472 goto cleanup_matches;
1474 t = ipt_get_target(e);
1475 target = try_then_request_module(xt_find_target(AF_INET,
1476 t->u.user.name,
1477 t->u.user.revision),
1478 "ipt_%s", t->u.user.name);
1479 if (IS_ERR(target) || !target) {
1480 duprintf("check_entry: `%s' not found\n", t->u.user.name);
1481 ret = target ? PTR_ERR(target) : -ENOENT;
1482 goto cleanup_matches;
1484 t->u.kernel.target = target;
1486 off += xt_compat_target_offset(target);
1487 *size += off;
1488 ret = compat_add_offset(entry_offset, off);
1489 if (ret)
1490 goto out;
1492 /* Check hooks & underflows */
1493 for (h = 0; h < NF_IP_NUMHOOKS; h++) {
1494 if ((unsigned char *)e - base == hook_entries[h])
1495 newinfo->hook_entry[h] = hook_entries[h];
1496 if ((unsigned char *)e - base == underflows[h])
1497 newinfo->underflow[h] = underflows[h];
1500 /* Clear counters and comefrom */
1501 e->counters = ((struct ipt_counters) { 0, 0 });
1502 e->comefrom = 0;
1504 (*i)++;
1505 return 0;
1507 out:
1508 module_put(t->u.kernel.target->me);
1509 cleanup_matches:
1510 IPT_MATCH_ITERATE(e, cleanup_match, &j);
1511 return ret;
1514 static inline int compat_copy_match_from_user(struct ipt_entry_match *m,
1515 void **dstptr, compat_uint_t *size, const char *name,
1516 const struct ipt_ip *ip, unsigned int hookmask, int *i)
1518 struct ipt_entry_match *dm;
1519 struct ipt_match *match;
1520 int ret;
1522 dm = (struct ipt_entry_match *)*dstptr;
1523 match = m->u.kernel.match;
1524 xt_compat_match_from_user(m, dstptr, size);
1526 ret = xt_check_match(match, AF_INET, dm->u.match_size - sizeof(*dm),
1527 name, hookmask, ip->proto,
1528 ip->invflags & IPT_INV_PROTO);
1529 if (ret)
1530 goto err;
1532 if (m->u.kernel.match->checkentry
1533 && !m->u.kernel.match->checkentry(name, ip, match, dm->data,
1534 hookmask)) {
1535 duprintf("ip_tables: check failed for `%s'.\n",
1536 m->u.kernel.match->name);
1537 ret = -EINVAL;
1538 goto err;
1540 (*i)++;
1541 return 0;
1543 err:
1544 module_put(m->u.kernel.match->me);
1545 return ret;
1548 static int compat_copy_entry_from_user(struct ipt_entry *e, void **dstptr,
1549 unsigned int *size, const char *name,
1550 struct xt_table_info *newinfo, unsigned char *base)
1552 struct ipt_entry_target *t;
1553 struct ipt_target *target;
1554 struct ipt_entry *de;
1555 unsigned int origsize;
1556 int ret, h, j;
1558 ret = 0;
1559 origsize = *size;
1560 de = (struct ipt_entry *)*dstptr;
1561 memcpy(de, e, sizeof(struct ipt_entry));
1563 j = 0;
1564 *dstptr += sizeof(struct compat_ipt_entry);
1565 ret = IPT_MATCH_ITERATE(e, compat_copy_match_from_user, dstptr, size,
1566 name, &de->ip, de->comefrom, &j);
1567 if (ret)
1568 goto cleanup_matches;
1569 de->target_offset = e->target_offset - (origsize - *size);
1570 t = ipt_get_target(e);
1571 target = t->u.kernel.target;
1572 xt_compat_target_from_user(t, dstptr, size);
1574 de->next_offset = e->next_offset - (origsize - *size);
1575 for (h = 0; h < NF_IP_NUMHOOKS; h++) {
1576 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1577 newinfo->hook_entry[h] -= origsize - *size;
1578 if ((unsigned char *)de - base < newinfo->underflow[h])
1579 newinfo->underflow[h] -= origsize - *size;
1582 t = ipt_get_target(de);
1583 target = t->u.kernel.target;
1584 ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
1585 name, e->comefrom, e->ip.proto,
1586 e->ip.invflags & IPT_INV_PROTO);
1587 if (ret)
1588 goto err;
1590 ret = -EINVAL;
1591 if (t->u.kernel.target == &ipt_standard_target) {
1592 if (!standard_check(t, *size))
1593 goto err;
1594 } else if (t->u.kernel.target->checkentry
1595 && !t->u.kernel.target->checkentry(name, de, target,
1596 t->data, de->comefrom)) {
1597 duprintf("ip_tables: compat: check failed for `%s'.\n",
1598 t->u.kernel.target->name);
1599 goto err;
1601 ret = 0;
1602 return ret;
1604 err:
1605 module_put(t->u.kernel.target->me);
1606 cleanup_matches:
1607 IPT_MATCH_ITERATE(e, cleanup_match, &j);
1608 return ret;
1611 static int
1612 translate_compat_table(const char *name,
1613 unsigned int valid_hooks,
1614 struct xt_table_info **pinfo,
1615 void **pentry0,
1616 unsigned int total_size,
1617 unsigned int number,
1618 unsigned int *hook_entries,
1619 unsigned int *underflows)
1621 unsigned int i;
1622 struct xt_table_info *newinfo, *info;
1623 void *pos, *entry0, *entry1;
1624 unsigned int size;
1625 int ret;
1627 info = *pinfo;
1628 entry0 = *pentry0;
1629 size = total_size;
1630 info->number = number;
1632 /* Init all hooks to impossible value. */
1633 for (i = 0; i < NF_IP_NUMHOOKS; i++) {
1634 info->hook_entry[i] = 0xFFFFFFFF;
1635 info->underflow[i] = 0xFFFFFFFF;
1638 duprintf("translate_compat_table: size %u\n", info->size);
1639 i = 0;
1640 xt_compat_lock(AF_INET);
1641 /* Walk through entries, checking offsets. */
1642 ret = IPT_ENTRY_ITERATE(entry0, total_size,
1643 check_compat_entry_size_and_hooks,
1644 info, &size, entry0,
1645 entry0 + total_size,
1646 hook_entries, underflows, &i, name);
1647 if (ret != 0)
1648 goto out_unlock;
1650 ret = -EINVAL;
1651 if (i != number) {
1652 duprintf("translate_compat_table: %u not %u entries\n",
1653 i, number);
1654 goto out_unlock;
1657 /* Check hooks all assigned */
1658 for (i = 0; i < NF_IP_NUMHOOKS; i++) {
1659 /* Only hooks which are valid */
1660 if (!(valid_hooks & (1 << i)))
1661 continue;
1662 if (info->hook_entry[i] == 0xFFFFFFFF) {
1663 duprintf("Invalid hook entry %u %u\n",
1664 i, hook_entries[i]);
1665 goto out_unlock;
1667 if (info->underflow[i] == 0xFFFFFFFF) {
1668 duprintf("Invalid underflow %u %u\n",
1669 i, underflows[i]);
1670 goto out_unlock;
1674 ret = -ENOMEM;
1675 newinfo = xt_alloc_table_info(size);
1676 if (!newinfo)
1677 goto out_unlock;
1679 newinfo->number = number;
1680 for (i = 0; i < NF_IP_NUMHOOKS; i++) {
1681 newinfo->hook_entry[i] = info->hook_entry[i];
1682 newinfo->underflow[i] = info->underflow[i];
1684 entry1 = newinfo->entries[raw_smp_processor_id()];
1685 pos = entry1;
1686 size = total_size;
1687 ret = IPT_ENTRY_ITERATE(entry0, total_size,
1688 compat_copy_entry_from_user, &pos, &size,
1689 name, newinfo, entry1);
1690 compat_flush_offsets();
1691 xt_compat_unlock(AF_INET);
1692 if (ret)
1693 goto free_newinfo;
1695 ret = -ELOOP;
1696 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1697 goto free_newinfo;
1699 /* And one copy for every other CPU */
1700 for_each_possible_cpu(i)
1701 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1702 memcpy(newinfo->entries[i], entry1, newinfo->size);
1704 *pinfo = newinfo;
1705 *pentry0 = entry1;
1706 xt_free_table_info(info);
1707 return 0;
1709 free_newinfo:
1710 xt_free_table_info(newinfo);
1711 out:
1712 return ret;
1713 out_unlock:
1714 xt_compat_unlock(AF_INET);
1715 goto out;
1718 static int
1719 compat_do_replace(void __user *user, unsigned int len)
1721 int ret;
1722 struct compat_ipt_replace tmp;
1723 struct xt_table_info *newinfo;
1724 void *loc_cpu_entry;
1726 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1727 return -EFAULT;
1729 /* Hack: Causes ipchains to give correct error msg --RR */
1730 if (len != sizeof(tmp) + tmp.size)
1731 return -ENOPROTOOPT;
1733 /* overflow check */
1734 if (tmp.size >= (INT_MAX - sizeof(struct xt_table_info)) / NR_CPUS -
1735 SMP_CACHE_BYTES)
1736 return -ENOMEM;
1737 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1738 return -ENOMEM;
1740 newinfo = xt_alloc_table_info(tmp.size);
1741 if (!newinfo)
1742 return -ENOMEM;
1744 /* choose the copy that is our node/cpu */
1745 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1746 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1747 tmp.size) != 0) {
1748 ret = -EFAULT;
1749 goto free_newinfo;
1752 ret = translate_compat_table(tmp.name, tmp.valid_hooks,
1753 &newinfo, &loc_cpu_entry, tmp.size,
1754 tmp.num_entries, tmp.hook_entry, tmp.underflow);
1755 if (ret != 0)
1756 goto free_newinfo;
1758 duprintf("compat_do_replace: Translated table\n");
1760 ret = __do_replace(tmp.name, tmp.valid_hooks,
1761 newinfo, tmp.num_counters,
1762 compat_ptr(tmp.counters));
1763 if (ret)
1764 goto free_newinfo_untrans;
1765 return 0;
1767 free_newinfo_untrans:
1768 IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry,NULL);
1769 free_newinfo:
1770 xt_free_table_info(newinfo);
1771 return ret;
1774 static int
1775 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1776 unsigned int len)
1778 int ret;
1780 if (!capable(CAP_NET_ADMIN))
1781 return -EPERM;
1783 switch (cmd) {
1784 case IPT_SO_SET_REPLACE:
1785 ret = compat_do_replace(user, len);
1786 break;
1788 case IPT_SO_SET_ADD_COUNTERS:
1789 ret = do_add_counters(user, len, 1);
1790 break;
1792 default:
1793 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1794 ret = -EINVAL;
1797 return ret;
1800 struct compat_ipt_get_entries
1802 char name[IPT_TABLE_MAXNAMELEN];
1803 compat_uint_t size;
1804 struct compat_ipt_entry entrytable[0];
1807 static int compat_copy_entries_to_user(unsigned int total_size,
1808 struct ipt_table *table, void __user *userptr)
1810 unsigned int off, num;
1811 struct compat_ipt_entry e;
1812 struct xt_counters *counters;
1813 struct xt_table_info *private = table->private;
1814 void __user *pos;
1815 unsigned int size;
1816 int ret = 0;
1817 void *loc_cpu_entry;
1819 counters = alloc_counters(table);
1820 if (IS_ERR(counters))
1821 return PTR_ERR(counters);
1823 /* choose the copy that is on our node/cpu, ...
1824 * This choice is lazy (because current thread is
1825 * allowed to migrate to another cpu)
1827 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1828 pos = userptr;
1829 size = total_size;
1830 ret = IPT_ENTRY_ITERATE(loc_cpu_entry, total_size,
1831 compat_copy_entry_to_user, &pos, &size);
1832 if (ret)
1833 goto free_counters;
1835 /* ... then go back and fix counters and names */
1836 for (off = 0, num = 0; off < size; off += e.next_offset, num++) {
1837 unsigned int i;
1838 struct ipt_entry_match m;
1839 struct ipt_entry_target t;
1841 ret = -EFAULT;
1842 if (copy_from_user(&e, userptr + off,
1843 sizeof(struct compat_ipt_entry)))
1844 goto free_counters;
1845 if (copy_to_user(userptr + off +
1846 offsetof(struct compat_ipt_entry, counters),
1847 &counters[num], sizeof(counters[num])))
1848 goto free_counters;
1850 for (i = sizeof(struct compat_ipt_entry);
1851 i < e.target_offset; i += m.u.match_size) {
1852 if (copy_from_user(&m, userptr + off + i,
1853 sizeof(struct ipt_entry_match)))
1854 goto free_counters;
1855 if (copy_to_user(userptr + off + i +
1856 offsetof(struct ipt_entry_match, u.user.name),
1857 m.u.kernel.match->name,
1858 strlen(m.u.kernel.match->name) + 1))
1859 goto free_counters;
1862 if (copy_from_user(&t, userptr + off + e.target_offset,
1863 sizeof(struct ipt_entry_target)))
1864 goto free_counters;
1865 if (copy_to_user(userptr + off + e.target_offset +
1866 offsetof(struct ipt_entry_target, u.user.name),
1867 t.u.kernel.target->name,
1868 strlen(t.u.kernel.target->name) + 1))
1869 goto free_counters;
1871 ret = 0;
1872 free_counters:
1873 vfree(counters);
1874 return ret;
1877 static int
1878 compat_get_entries(struct compat_ipt_get_entries __user *uptr, int *len)
1880 int ret;
1881 struct compat_ipt_get_entries get;
1882 struct ipt_table *t;
1885 if (*len < sizeof(get)) {
1886 duprintf("compat_get_entries: %u < %u\n",
1887 *len, (unsigned int)sizeof(get));
1888 return -EINVAL;
1891 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1892 return -EFAULT;
1894 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1895 duprintf("compat_get_entries: %u != %u\n", *len,
1896 (unsigned int)(sizeof(struct compat_ipt_get_entries) +
1897 get.size));
1898 return -EINVAL;
1901 xt_compat_lock(AF_INET);
1902 t = xt_find_table_lock(AF_INET, get.name);
1903 if (t && !IS_ERR(t)) {
1904 struct xt_table_info *private = t->private;
1905 struct xt_table_info info;
1906 duprintf("t->private->number = %u\n",
1907 private->number);
1908 ret = compat_table_info(private, &info);
1909 if (!ret && get.size == info.size) {
1910 ret = compat_copy_entries_to_user(private->size,
1911 t, uptr->entrytable);
1912 } else if (!ret) {
1913 duprintf("compat_get_entries: I've got %u not %u!\n",
1914 private->size,
1915 get.size);
1916 ret = -EINVAL;
1918 compat_flush_offsets();
1919 module_put(t->me);
1920 xt_table_unlock(t);
1921 } else
1922 ret = t ? PTR_ERR(t) : -ENOENT;
1924 xt_compat_unlock(AF_INET);
1925 return ret;
1928 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1930 static int
1931 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1933 int ret;
1935 switch (cmd) {
1936 case IPT_SO_GET_INFO:
1937 ret = get_info(user, len, 1);
1938 break;
1939 case IPT_SO_GET_ENTRIES:
1940 ret = compat_get_entries(user, len);
1941 break;
1942 default:
1943 ret = do_ipt_get_ctl(sk, cmd, user, len);
1945 return ret;
1947 #endif
1949 static int
1950 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1952 int ret;
1954 if (!capable(CAP_NET_ADMIN))
1955 return -EPERM;
1957 switch (cmd) {
1958 case IPT_SO_SET_REPLACE:
1959 ret = do_replace(user, len);
1960 break;
1962 case IPT_SO_SET_ADD_COUNTERS:
1963 ret = do_add_counters(user, len, 0);
1964 break;
1966 default:
1967 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1968 ret = -EINVAL;
1971 return ret;
1974 static int
1975 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1977 int ret;
1979 if (!capable(CAP_NET_ADMIN))
1980 return -EPERM;
1982 switch (cmd) {
1983 case IPT_SO_GET_INFO:
1984 ret = get_info(user, len, 0);
1985 break;
1987 case IPT_SO_GET_ENTRIES:
1988 ret = get_entries(user, len);
1989 break;
1991 case IPT_SO_GET_REVISION_MATCH:
1992 case IPT_SO_GET_REVISION_TARGET: {
1993 struct ipt_get_revision rev;
1994 int target;
1996 if (*len != sizeof(rev)) {
1997 ret = -EINVAL;
1998 break;
2000 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2001 ret = -EFAULT;
2002 break;
2005 if (cmd == IPT_SO_GET_REVISION_TARGET)
2006 target = 1;
2007 else
2008 target = 0;
2010 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2011 rev.revision,
2012 target, &ret),
2013 "ipt_%s", rev.name);
2014 break;
2017 default:
2018 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2019 ret = -EINVAL;
2022 return ret;
2025 int ipt_register_table(struct xt_table *table, const struct ipt_replace *repl)
2027 int ret;
2028 struct xt_table_info *newinfo;
2029 static struct xt_table_info bootstrap
2030 = { 0, 0, 0, { 0 }, { 0 }, { } };
2031 void *loc_cpu_entry;
2033 newinfo = xt_alloc_table_info(repl->size);
2034 if (!newinfo)
2035 return -ENOMEM;
2037 /* choose the copy on our node/cpu
2038 * but dont care of preemption
2040 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2041 memcpy(loc_cpu_entry, repl->entries, repl->size);
2043 ret = translate_table(table->name, table->valid_hooks,
2044 newinfo, loc_cpu_entry, repl->size,
2045 repl->num_entries,
2046 repl->hook_entry,
2047 repl->underflow);
2048 if (ret != 0) {
2049 xt_free_table_info(newinfo);
2050 return ret;
2053 ret = xt_register_table(table, &bootstrap, newinfo);
2054 if (ret != 0) {
2055 xt_free_table_info(newinfo);
2056 return ret;
2059 return 0;
2062 void ipt_unregister_table(struct ipt_table *table)
2064 struct xt_table_info *private;
2065 void *loc_cpu_entry;
2067 private = xt_unregister_table(table);
2069 /* Decrease module usage counts and free resources */
2070 loc_cpu_entry = private->entries[raw_smp_processor_id()];
2071 IPT_ENTRY_ITERATE(loc_cpu_entry, private->size, cleanup_entry, NULL);
2072 xt_free_table_info(private);
2075 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2076 static inline int
2077 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2078 u_int8_t type, u_int8_t code,
2079 int invert)
2081 return ((test_type == 0xFF) || (type == test_type && code >= min_code && code <= max_code))
2082 ^ invert;
2085 static int
2086 icmp_match(const struct sk_buff *skb,
2087 const struct net_device *in,
2088 const struct net_device *out,
2089 const struct xt_match *match,
2090 const void *matchinfo,
2091 int offset,
2092 unsigned int protoff,
2093 int *hotdrop)
2095 struct icmphdr _icmph, *ic;
2096 const struct ipt_icmp *icmpinfo = matchinfo;
2098 /* Must not be a fragment. */
2099 if (offset)
2100 return 0;
2102 ic = skb_header_pointer(skb, protoff, sizeof(_icmph), &_icmph);
2103 if (ic == NULL) {
2104 /* We've been asked to examine this packet, and we
2105 * can't. Hence, no choice but to drop.
2107 duprintf("Dropping evil ICMP tinygram.\n");
2108 *hotdrop = 1;
2109 return 0;
2112 return icmp_type_code_match(icmpinfo->type,
2113 icmpinfo->code[0],
2114 icmpinfo->code[1],
2115 ic->type, ic->code,
2116 !!(icmpinfo->invflags&IPT_ICMP_INV));
2119 /* Called when user tries to insert an entry of this type. */
2120 static int
2121 icmp_checkentry(const char *tablename,
2122 const void *info,
2123 const struct xt_match *match,
2124 void *matchinfo,
2125 unsigned int hook_mask)
2127 const struct ipt_icmp *icmpinfo = matchinfo;
2129 /* Must specify no unknown invflags */
2130 return !(icmpinfo->invflags & ~IPT_ICMP_INV);
2133 /* The built-in targets: standard (NULL) and error. */
2134 static struct ipt_target ipt_standard_target = {
2135 .name = IPT_STANDARD_TARGET,
2136 .targetsize = sizeof(int),
2137 .family = AF_INET,
2138 #ifdef CONFIG_COMPAT
2139 .compatsize = sizeof(compat_int_t),
2140 .compat_from_user = compat_standard_from_user,
2141 .compat_to_user = compat_standard_to_user,
2142 #endif
2145 static struct ipt_target ipt_error_target = {
2146 .name = IPT_ERROR_TARGET,
2147 .target = ipt_error,
2148 .targetsize = IPT_FUNCTION_MAXNAMELEN,
2149 .family = AF_INET,
2152 static struct nf_sockopt_ops ipt_sockopts = {
2153 .pf = PF_INET,
2154 .set_optmin = IPT_BASE_CTL,
2155 .set_optmax = IPT_SO_SET_MAX+1,
2156 .set = do_ipt_set_ctl,
2157 #ifdef CONFIG_COMPAT
2158 .compat_set = compat_do_ipt_set_ctl,
2159 #endif
2160 .get_optmin = IPT_BASE_CTL,
2161 .get_optmax = IPT_SO_GET_MAX+1,
2162 .get = do_ipt_get_ctl,
2163 #ifdef CONFIG_COMPAT
2164 .compat_get = compat_do_ipt_get_ctl,
2165 #endif
2168 static struct ipt_match icmp_matchstruct = {
2169 .name = "icmp",
2170 .match = icmp_match,
2171 .matchsize = sizeof(struct ipt_icmp),
2172 .proto = IPPROTO_ICMP,
2173 .family = AF_INET,
2174 .checkentry = icmp_checkentry,
2177 static int __init ip_tables_init(void)
2179 int ret;
2181 ret = xt_proto_init(AF_INET);
2182 if (ret < 0)
2183 goto err1;
2185 /* Noone else will be downing sem now, so we won't sleep */
2186 ret = xt_register_target(&ipt_standard_target);
2187 if (ret < 0)
2188 goto err2;
2189 ret = xt_register_target(&ipt_error_target);
2190 if (ret < 0)
2191 goto err3;
2192 ret = xt_register_match(&icmp_matchstruct);
2193 if (ret < 0)
2194 goto err4;
2196 /* Register setsockopt */
2197 ret = nf_register_sockopt(&ipt_sockopts);
2198 if (ret < 0)
2199 goto err5;
2201 printk("ip_tables: (C) 2000-2006 Netfilter Core Team\n");
2202 return 0;
2204 err5:
2205 xt_unregister_match(&icmp_matchstruct);
2206 err4:
2207 xt_unregister_target(&ipt_error_target);
2208 err3:
2209 xt_unregister_target(&ipt_standard_target);
2210 err2:
2211 xt_proto_fini(AF_INET);
2212 err1:
2213 return ret;
2216 static void __exit ip_tables_fini(void)
2218 nf_unregister_sockopt(&ipt_sockopts);
2220 xt_unregister_match(&icmp_matchstruct);
2221 xt_unregister_target(&ipt_error_target);
2222 xt_unregister_target(&ipt_standard_target);
2224 xt_proto_fini(AF_INET);
2227 EXPORT_SYMBOL(ipt_register_table);
2228 EXPORT_SYMBOL(ipt_unregister_table);
2229 EXPORT_SYMBOL(ipt_do_table);
2230 module_init(ip_tables_init);
2231 module_exit(ip_tables_fini);