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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12 #include <linux/cache.h>
13 #include <linux/capability.h>
14 #include <linux/skbuff.h>
15 #include <linux/kmod.h>
16 #include <linux/vmalloc.h>
17 #include <linux/netdevice.h>
18 #include <linux/module.h>
19 #include <linux/icmp.h>
21 #include <net/compat.h>
22 #include <asm/uaccess.h>
23 #include <linux/mutex.h>
24 #include <linux/proc_fs.h>
25 #include <linux/err.h>
26 #include <linux/cpumask.h>
28 #include <linux/netfilter/x_tables.h>
29 #include <linux/netfilter_ipv4/ip_tables.h>
30 #include <net/netfilter/nf_log.h>
31 #include "../../netfilter/xt_repldata.h"
33 MODULE_LICENSE("GPL");
34 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
35 MODULE_DESCRIPTION("IPv4 packet filter");
37 /*#define DEBUG_IP_FIREWALL*/
38 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
39 /*#define DEBUG_IP_FIREWALL_USER*/
41 #ifdef DEBUG_IP_FIREWALL
42 #define dprintf(format, args...) pr_info(format , ## args)
44 #define dprintf(format, args...)
47 #ifdef DEBUG_IP_FIREWALL_USER
48 #define duprintf(format, args...) pr_info(format , ## args)
50 #define duprintf(format, args...)
53 #ifdef CONFIG_NETFILTER_DEBUG
54 #define IP_NF_ASSERT(x) \
57 printk("IP_NF_ASSERT: %s:%s:%u\n", \
58 __func__, __FILE__, __LINE__); \
61 #define IP_NF_ASSERT(x)
65 /* All the better to debug you with... */
70 void *ipt_alloc_initial_table(const struct xt_table *info)
72 return xt_alloc_initial_table(ipt, IPT);
74 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
77 We keep a set of rules for each CPU, so we can avoid write-locking
78 them in the softirq when updating the counters and therefore
79 only need to read-lock in the softirq; doing a write_lock_bh() in user
80 context stops packets coming through and allows user context to read
81 the counters or update the rules.
83 Hence the start of any table is given by get_table() below. */
85 /* Returns whether matches rule or not. */
86 /* Performance critical - called for every packet */
88 ip_packet_match(const struct iphdr *ip,
91 const struct ipt_ip *ipinfo,
96 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
98 if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
100 FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
102 dprintf("Source or dest mismatch.\n");
104 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
105 &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
106 ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
107 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
108 &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
109 ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
113 ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
115 if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
116 dprintf("VIA in mismatch (%s vs %s).%s\n",
117 indev, ipinfo->iniface,
118 ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
122 ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
124 if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
125 dprintf("VIA out mismatch (%s vs %s).%s\n",
126 outdev, ipinfo->outiface,
127 ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
131 /* Check specific protocol */
133 FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
134 dprintf("Packet protocol %hi does not match %hi.%s\n",
135 ip->protocol, ipinfo->proto,
136 ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
140 /* If we have a fragment rule but the packet is not a fragment
141 * then we return zero */
142 if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
143 dprintf("Fragment rule but not fragment.%s\n",
144 ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
152 ip_checkentry(const struct ipt_ip *ip)
154 if (ip->flags & ~IPT_F_MASK) {
155 duprintf("Unknown flag bits set: %08X\n",
156 ip->flags & ~IPT_F_MASK);
159 if (ip->invflags & ~IPT_INV_MASK) {
160 duprintf("Unknown invflag bits set: %08X\n",
161 ip->invflags & ~IPT_INV_MASK);
168 ipt_error(struct sk_buff *skb, const struct xt_target_param *par)
171 pr_info("error: `%s'\n", (const char *)par->targinfo);
176 /* Performance critical - called for every packet */
178 do_match(const struct ipt_entry_match *m, const struct sk_buff *skb,
179 struct xt_match_param *par)
181 par->match = m->u.kernel.match;
182 par->matchinfo = m->data;
184 /* Stop iteration if it doesn't match */
185 if (!m->u.kernel.match->match(skb, par))
191 /* Performance critical */
192 static inline struct ipt_entry *
193 get_entry(const void *base, unsigned int offset)
195 return (struct ipt_entry *)(base + offset);
198 /* All zeroes == unconditional rule. */
199 /* Mildly perf critical (only if packet tracing is on) */
200 static inline bool unconditional(const struct ipt_ip *ip)
202 static const struct ipt_ip uncond;
204 return memcmp(ip, &uncond, sizeof(uncond)) == 0;
208 /* for const-correctness */
209 static inline const struct ipt_entry_target *
210 ipt_get_target_c(const struct ipt_entry *e)
212 return ipt_get_target((struct ipt_entry *)e);
215 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
216 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
217 static const char *const hooknames[] = {
218 [NF_INET_PRE_ROUTING] = "PREROUTING",
219 [NF_INET_LOCAL_IN] = "INPUT",
220 [NF_INET_FORWARD] = "FORWARD",
221 [NF_INET_LOCAL_OUT] = "OUTPUT",
222 [NF_INET_POST_ROUTING] = "POSTROUTING",
225 enum nf_ip_trace_comments {
226 NF_IP_TRACE_COMMENT_RULE,
227 NF_IP_TRACE_COMMENT_RETURN,
228 NF_IP_TRACE_COMMENT_POLICY,
231 static const char *const comments[] = {
232 [NF_IP_TRACE_COMMENT_RULE] = "rule",
233 [NF_IP_TRACE_COMMENT_RETURN] = "return",
234 [NF_IP_TRACE_COMMENT_POLICY] = "policy",
237 static struct nf_loginfo trace_loginfo = {
238 .type = NF_LOG_TYPE_LOG,
242 .logflags = NF_LOG_MASK,
247 /* Mildly perf critical (only if packet tracing is on) */
249 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
250 const char *hookname, const char **chainname,
251 const char **comment, unsigned int *rulenum)
253 const struct ipt_standard_target *t = (void *)ipt_get_target_c(s);
255 if (strcmp(t->target.u.kernel.target->name, IPT_ERROR_TARGET) == 0) {
256 /* Head of user chain: ERROR target with chainname */
257 *chainname = t->target.data;
262 if (s->target_offset == sizeof(struct ipt_entry) &&
263 strcmp(t->target.u.kernel.target->name,
264 IPT_STANDARD_TARGET) == 0 &&
266 unconditional(&s->ip)) {
267 /* Tail of chains: STANDARD target (return/policy) */
268 *comment = *chainname == hookname
269 ? comments[NF_IP_TRACE_COMMENT_POLICY]
270 : comments[NF_IP_TRACE_COMMENT_RETURN];
279 static void trace_packet(const struct sk_buff *skb,
281 const struct net_device *in,
282 const struct net_device *out,
283 const char *tablename,
284 const struct xt_table_info *private,
285 const struct ipt_entry *e)
287 const void *table_base;
288 const struct ipt_entry *root;
289 const char *hookname, *chainname, *comment;
290 const struct ipt_entry *iter;
291 unsigned int rulenum = 0;
293 table_base = private->entries[smp_processor_id()];
294 root = get_entry(table_base, private->hook_entry[hook]);
296 hookname = chainname = hooknames[hook];
297 comment = comments[NF_IP_TRACE_COMMENT_RULE];
299 xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
300 if (get_chainname_rulenum(iter, e, hookname,
301 &chainname, &comment, &rulenum) != 0)
304 nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
305 "TRACE: %s:%s:%s:%u ",
306 tablename, chainname, comment, rulenum);
311 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
313 return (void *)entry + entry->next_offset;
316 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
318 ipt_do_table(struct sk_buff *skb,
320 const struct net_device *in,
321 const struct net_device *out,
322 struct xt_table *table)
324 #define tb_comefrom ((struct ipt_entry *)table_base)->comefrom
326 static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
327 const struct iphdr *ip;
328 bool hotdrop = false;
329 /* Initializing verdict to NF_DROP keeps gcc happy. */
330 unsigned int verdict = NF_DROP;
331 const char *indev, *outdev;
332 const void *table_base;
333 struct ipt_entry *e, *back;
334 const struct xt_table_info *private;
335 struct xt_match_param mtpar;
336 struct xt_target_param tgpar;
340 indev = in ? in->name : nulldevname;
341 outdev = out ? out->name : nulldevname;
342 /* We handle fragments by dealing with the first fragment as
343 * if it was a normal packet. All other fragments are treated
344 * normally, except that they will NEVER match rules that ask
345 * things we don't know, ie. tcp syn flag or ports). If the
346 * rule is also a fragment-specific rule, non-fragments won't
348 mtpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
349 mtpar.thoff = ip_hdrlen(skb);
350 mtpar.hotdrop = &hotdrop;
351 mtpar.in = tgpar.in = in;
352 mtpar.out = tgpar.out = out;
353 mtpar.family = tgpar.family = NFPROTO_IPV4;
354 mtpar.hooknum = tgpar.hooknum = hook;
356 IP_NF_ASSERT(table->valid_hooks & (1 << hook));
358 private = table->private;
359 table_base = private->entries[smp_processor_id()];
361 e = get_entry(table_base, private->hook_entry[hook]);
363 /* For return from builtin chain */
364 back = get_entry(table_base, private->underflow[hook]);
367 const struct ipt_entry_target *t;
368 const struct xt_entry_match *ematch;
372 if (!ip_packet_match(ip, indev, outdev,
373 &e->ip, mtpar.fragoff)) {
375 e = ipt_next_entry(e);
379 xt_ematch_foreach(ematch, e)
380 if (do_match(ematch, skb, &mtpar) != 0)
383 ADD_COUNTER(e->counters, ntohs(ip->tot_len), 1);
385 t = ipt_get_target(e);
386 IP_NF_ASSERT(t->u.kernel.target);
388 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
389 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
390 /* The packet is traced: log it */
391 if (unlikely(skb->nf_trace))
392 trace_packet(skb, hook, in, out,
393 table->name, private, e);
395 /* Standard target? */
396 if (!t->u.kernel.target->target) {
399 v = ((struct ipt_standard_target *)t)->verdict;
401 /* Pop from stack? */
402 if (v != IPT_RETURN) {
403 verdict = (unsigned)(-v) - 1;
407 back = get_entry(table_base, back->comefrom);
410 if (table_base + v != ipt_next_entry(e) &&
411 !(e->ip.flags & IPT_F_GOTO)) {
412 /* Save old back ptr in next entry */
413 struct ipt_entry *next = ipt_next_entry(e);
414 next->comefrom = (void *)back - table_base;
415 /* set back pointer to next entry */
419 e = get_entry(table_base, v);
423 /* Targets which reenter must return
425 tgpar.target = t->u.kernel.target;
426 tgpar.targinfo = t->data;
429 #ifdef CONFIG_NETFILTER_DEBUG
430 tb_comefrom = 0xeeeeeeec;
432 verdict = t->u.kernel.target->target(skb, &tgpar);
433 #ifdef CONFIG_NETFILTER_DEBUG
434 if (tb_comefrom != 0xeeeeeeec && verdict == IPT_CONTINUE) {
435 printk("Target %s reentered!\n",
436 t->u.kernel.target->name);
439 tb_comefrom = 0x57acc001;
441 /* Target might have changed stuff. */
443 if (verdict == IPT_CONTINUE)
444 e = ipt_next_entry(e);
449 xt_info_rdunlock_bh();
451 #ifdef DEBUG_ALLOW_ALL
462 /* Figures out from what hook each rule can be called: returns 0 if
463 there are loops. Puts hook bitmask in comefrom. */
465 mark_source_chains(const struct xt_table_info *newinfo,
466 unsigned int valid_hooks, void *entry0)
470 /* No recursion; use packet counter to save back ptrs (reset
471 to 0 as we leave), and comefrom to save source hook bitmask */
472 for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
473 unsigned int pos = newinfo->hook_entry[hook];
474 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
476 if (!(valid_hooks & (1 << hook)))
479 /* Set initial back pointer. */
480 e->counters.pcnt = pos;
483 const struct ipt_standard_target *t
484 = (void *)ipt_get_target_c(e);
485 int visited = e->comefrom & (1 << hook);
487 if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
488 printk("iptables: loop hook %u pos %u %08X.\n",
489 hook, pos, e->comefrom);
492 e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
494 /* Unconditional return/END. */
495 if ((e->target_offset == sizeof(struct ipt_entry) &&
496 (strcmp(t->target.u.user.name,
497 IPT_STANDARD_TARGET) == 0) &&
498 t->verdict < 0 && unconditional(&e->ip)) ||
500 unsigned int oldpos, size;
502 if ((strcmp(t->target.u.user.name,
503 IPT_STANDARD_TARGET) == 0) &&
504 t->verdict < -NF_MAX_VERDICT - 1) {
505 duprintf("mark_source_chains: bad "
506 "negative verdict (%i)\n",
511 /* Return: backtrack through the last
514 e->comefrom ^= (1<<NF_INET_NUMHOOKS);
515 #ifdef DEBUG_IP_FIREWALL_USER
517 & (1 << NF_INET_NUMHOOKS)) {
518 duprintf("Back unset "
525 pos = e->counters.pcnt;
526 e->counters.pcnt = 0;
528 /* We're at the start. */
532 e = (struct ipt_entry *)
534 } while (oldpos == pos + e->next_offset);
537 size = e->next_offset;
538 e = (struct ipt_entry *)
539 (entry0 + pos + size);
540 e->counters.pcnt = pos;
543 int newpos = t->verdict;
545 if (strcmp(t->target.u.user.name,
546 IPT_STANDARD_TARGET) == 0 &&
548 if (newpos > newinfo->size -
549 sizeof(struct ipt_entry)) {
550 duprintf("mark_source_chains: "
551 "bad verdict (%i)\n",
555 /* This a jump; chase it. */
556 duprintf("Jump rule %u -> %u\n",
559 /* ... this is a fallthru */
560 newpos = pos + e->next_offset;
562 e = (struct ipt_entry *)
564 e->counters.pcnt = pos;
569 duprintf("Finished chain %u\n", hook);
574 static void cleanup_match(struct ipt_entry_match *m, struct net *net)
576 struct xt_mtdtor_param par;
579 par.match = m->u.kernel.match;
580 par.matchinfo = m->data;
581 par.family = NFPROTO_IPV4;
582 if (par.match->destroy != NULL)
583 par.match->destroy(&par);
584 module_put(par.match->me);
588 check_entry(const struct ipt_entry *e, const char *name)
590 const struct ipt_entry_target *t;
592 if (!ip_checkentry(&e->ip)) {
593 duprintf("ip check failed %p %s.\n", e, name);
597 if (e->target_offset + sizeof(struct ipt_entry_target) >
601 t = ipt_get_target_c(e);
602 if (e->target_offset + t->u.target_size > e->next_offset)
609 check_match(struct ipt_entry_match *m, struct xt_mtchk_param *par)
611 const struct ipt_ip *ip = par->entryinfo;
614 par->match = m->u.kernel.match;
615 par->matchinfo = m->data;
617 ret = xt_check_match(par, m->u.match_size - sizeof(*m),
618 ip->proto, ip->invflags & IPT_INV_PROTO);
620 duprintf("check failed for `%s'.\n", par.match->name);
627 find_check_match(struct ipt_entry_match *m, struct xt_mtchk_param *par)
629 struct xt_match *match;
632 match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
634 "ipt_%s", m->u.user.name);
635 if (IS_ERR(match) || !match) {
636 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
637 return match ? PTR_ERR(match) : -ENOENT;
639 m->u.kernel.match = match;
641 ret = check_match(m, par);
647 module_put(m->u.kernel.match->me);
651 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
653 struct ipt_entry_target *t = ipt_get_target(e);
654 struct xt_tgchk_param par = {
658 .target = t->u.kernel.target,
660 .hook_mask = e->comefrom,
661 .family = NFPROTO_IPV4,
665 ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
666 e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
668 duprintf("check failed for `%s'.\n",
669 t->u.kernel.target->name);
676 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
679 struct ipt_entry_target *t;
680 struct xt_target *target;
683 struct xt_mtchk_param mtpar;
684 struct xt_entry_match *ematch;
686 ret = check_entry(e, name);
693 mtpar.entryinfo = &e->ip;
694 mtpar.hook_mask = e->comefrom;
695 mtpar.family = NFPROTO_IPV4;
696 xt_ematch_foreach(ematch, e) {
697 ret = find_check_match(ematch, &mtpar);
699 goto cleanup_matches;
703 t = ipt_get_target(e);
704 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
706 if (IS_ERR(target)) {
707 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
708 ret = PTR_ERR(target);
709 goto cleanup_matches;
711 t->u.kernel.target = target;
713 ret = check_target(e, net, name);
718 module_put(t->u.kernel.target->me);
720 xt_ematch_foreach(ematch, e) {
723 cleanup_match(ematch, net);
728 static bool check_underflow(const struct ipt_entry *e)
730 const struct ipt_entry_target *t;
731 unsigned int verdict;
733 if (!unconditional(&e->ip))
735 t = ipt_get_target_c(e);
736 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
738 verdict = ((struct ipt_standard_target *)t)->verdict;
739 verdict = -verdict - 1;
740 return verdict == NF_DROP || verdict == NF_ACCEPT;
744 check_entry_size_and_hooks(struct ipt_entry *e,
745 struct xt_table_info *newinfo,
746 const unsigned char *base,
747 const unsigned char *limit,
748 const unsigned int *hook_entries,
749 const unsigned int *underflows,
750 unsigned int valid_hooks)
754 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
755 (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
756 duprintf("Bad offset %p\n", e);
761 < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) {
762 duprintf("checking: element %p size %u\n",
767 /* Check hooks & underflows */
768 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
769 if (!(valid_hooks & (1 << h)))
771 if ((unsigned char *)e - base == hook_entries[h])
772 newinfo->hook_entry[h] = hook_entries[h];
773 if ((unsigned char *)e - base == underflows[h]) {
774 if (!check_underflow(e)) {
775 pr_err("Underflows must be unconditional and "
776 "use the STANDARD target with "
780 newinfo->underflow[h] = underflows[h];
784 /* Clear counters and comefrom */
785 e->counters = ((struct xt_counters) { 0, 0 });
791 cleanup_entry(struct ipt_entry *e, struct net *net)
793 struct xt_tgdtor_param par;
794 struct ipt_entry_target *t;
795 struct xt_entry_match *ematch;
797 /* Cleanup all matches */
798 xt_ematch_foreach(ematch, e)
799 cleanup_match(ematch, net);
800 t = ipt_get_target(e);
803 par.target = t->u.kernel.target;
804 par.targinfo = t->data;
805 par.family = NFPROTO_IPV4;
806 if (par.target->destroy != NULL)
807 par.target->destroy(&par);
808 module_put(par.target->me);
811 /* Checks and translates the user-supplied table segment (held in
814 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
815 const struct ipt_replace *repl)
817 struct ipt_entry *iter;
821 newinfo->size = repl->size;
822 newinfo->number = repl->num_entries;
824 /* Init all hooks to impossible value. */
825 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
826 newinfo->hook_entry[i] = 0xFFFFFFFF;
827 newinfo->underflow[i] = 0xFFFFFFFF;
830 duprintf("translate_table: size %u\n", newinfo->size);
832 /* Walk through entries, checking offsets. */
833 xt_entry_foreach(iter, entry0, newinfo->size) {
834 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
844 if (i != repl->num_entries) {
845 duprintf("translate_table: %u not %u entries\n",
846 i, repl->num_entries);
850 /* Check hooks all assigned */
851 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
852 /* Only hooks which are valid */
853 if (!(repl->valid_hooks & (1 << i)))
855 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
856 duprintf("Invalid hook entry %u %u\n",
857 i, repl->hook_entry[i]);
860 if (newinfo->underflow[i] == 0xFFFFFFFF) {
861 duprintf("Invalid underflow %u %u\n",
862 i, repl->underflow[i]);
867 if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
870 /* Finally, each sanity check must pass */
872 xt_entry_foreach(iter, entry0, newinfo->size) {
873 ret = find_check_entry(iter, net, repl->name, repl->size);
880 xt_entry_foreach(iter, entry0, newinfo->size) {
883 cleanup_entry(iter, net);
888 /* And one copy for every other CPU */
889 for_each_possible_cpu(i) {
890 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
891 memcpy(newinfo->entries[i], entry0, newinfo->size);
898 get_counters(const struct xt_table_info *t,
899 struct xt_counters counters[])
901 struct ipt_entry *iter;
906 /* Instead of clearing (by a previous call to memset())
907 * the counters and using adds, we set the counters
908 * with data used by 'current' CPU.
910 * Bottom half has to be disabled to prevent deadlock
911 * if new softirq were to run and call ipt_do_table
914 curcpu = smp_processor_id();
917 xt_entry_foreach(iter, t->entries[curcpu], t->size) {
918 SET_COUNTER(counters[i], iter->counters.bcnt,
919 iter->counters.pcnt);
923 for_each_possible_cpu(cpu) {
928 xt_entry_foreach(iter, t->entries[cpu], t->size) {
929 ADD_COUNTER(counters[i], iter->counters.bcnt,
930 iter->counters.pcnt);
931 ++i; /* macro does multi eval of i */
933 xt_info_wrunlock(cpu);
938 static struct xt_counters *alloc_counters(const struct xt_table *table)
940 unsigned int countersize;
941 struct xt_counters *counters;
942 const struct xt_table_info *private = table->private;
944 /* We need atomic snapshot of counters: rest doesn't change
945 (other than comefrom, which userspace doesn't care
947 countersize = sizeof(struct xt_counters) * private->number;
948 counters = vmalloc_node(countersize, numa_node_id());
950 if (counters == NULL)
951 return ERR_PTR(-ENOMEM);
953 get_counters(private, counters);
959 copy_entries_to_user(unsigned int total_size,
960 const struct xt_table *table,
961 void __user *userptr)
963 unsigned int off, num;
964 const struct ipt_entry *e;
965 struct xt_counters *counters;
966 const struct xt_table_info *private = table->private;
968 const void *loc_cpu_entry;
970 counters = alloc_counters(table);
971 if (IS_ERR(counters))
972 return PTR_ERR(counters);
974 /* choose the copy that is on our node/cpu, ...
975 * This choice is lazy (because current thread is
976 * allowed to migrate to another cpu)
978 loc_cpu_entry = private->entries[raw_smp_processor_id()];
979 if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
984 /* FIXME: use iterator macros --RR */
985 /* ... then go back and fix counters and names */
986 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
988 const struct ipt_entry_match *m;
989 const struct ipt_entry_target *t;
991 e = (struct ipt_entry *)(loc_cpu_entry + off);
992 if (copy_to_user(userptr + off
993 + offsetof(struct ipt_entry, counters),
995 sizeof(counters[num])) != 0) {
1000 for (i = sizeof(struct ipt_entry);
1001 i < e->target_offset;
1002 i += m->u.match_size) {
1005 if (copy_to_user(userptr + off + i
1006 + offsetof(struct ipt_entry_match,
1008 m->u.kernel.match->name,
1009 strlen(m->u.kernel.match->name)+1)
1016 t = ipt_get_target_c(e);
1017 if (copy_to_user(userptr + off + e->target_offset
1018 + offsetof(struct ipt_entry_target,
1020 t->u.kernel.target->name,
1021 strlen(t->u.kernel.target->name)+1) != 0) {
1032 #ifdef CONFIG_COMPAT
1033 static void compat_standard_from_user(void *dst, const void *src)
1035 int v = *(compat_int_t *)src;
1038 v += xt_compat_calc_jump(AF_INET, v);
1039 memcpy(dst, &v, sizeof(v));
1042 static int compat_standard_to_user(void __user *dst, const void *src)
1044 compat_int_t cv = *(int *)src;
1047 cv -= xt_compat_calc_jump(AF_INET, cv);
1048 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1051 static int compat_calc_entry(const struct ipt_entry *e,
1052 const struct xt_table_info *info,
1053 const void *base, struct xt_table_info *newinfo)
1055 const struct xt_entry_match *ematch;
1056 const struct ipt_entry_target *t;
1057 unsigned int entry_offset;
1060 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1061 entry_offset = (void *)e - base;
1062 xt_ematch_foreach(ematch, e)
1063 off += xt_compat_match_offset(ematch->u.kernel.match);
1064 t = ipt_get_target_c(e);
1065 off += xt_compat_target_offset(t->u.kernel.target);
1066 newinfo->size -= off;
1067 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1071 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1072 if (info->hook_entry[i] &&
1073 (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1074 newinfo->hook_entry[i] -= off;
1075 if (info->underflow[i] &&
1076 (e < (struct ipt_entry *)(base + info->underflow[i])))
1077 newinfo->underflow[i] -= off;
1082 static int compat_table_info(const struct xt_table_info *info,
1083 struct xt_table_info *newinfo)
1085 struct ipt_entry *iter;
1086 void *loc_cpu_entry;
1089 if (!newinfo || !info)
1092 /* we dont care about newinfo->entries[] */
1093 memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1094 newinfo->initial_entries = 0;
1095 loc_cpu_entry = info->entries[raw_smp_processor_id()];
1096 xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1097 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1105 static int get_info(struct net *net, void __user *user,
1106 const int *len, int compat)
1108 char name[IPT_TABLE_MAXNAMELEN];
1112 if (*len != sizeof(struct ipt_getinfo)) {
1113 duprintf("length %u != %zu\n", *len,
1114 sizeof(struct ipt_getinfo));
1118 if (copy_from_user(name, user, sizeof(name)) != 0)
1121 name[IPT_TABLE_MAXNAMELEN-1] = '\0';
1122 #ifdef CONFIG_COMPAT
1124 xt_compat_lock(AF_INET);
1126 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1127 "iptable_%s", name);
1128 if (t && !IS_ERR(t)) {
1129 struct ipt_getinfo info;
1130 const struct xt_table_info *private = t->private;
1131 #ifdef CONFIG_COMPAT
1132 struct xt_table_info tmp;
1135 ret = compat_table_info(private, &tmp);
1136 xt_compat_flush_offsets(AF_INET);
1140 info.valid_hooks = t->valid_hooks;
1141 memcpy(info.hook_entry, private->hook_entry,
1142 sizeof(info.hook_entry));
1143 memcpy(info.underflow, private->underflow,
1144 sizeof(info.underflow));
1145 info.num_entries = private->number;
1146 info.size = private->size;
1147 strcpy(info.name, name);
1149 if (copy_to_user(user, &info, *len) != 0)
1157 ret = t ? PTR_ERR(t) : -ENOENT;
1158 #ifdef CONFIG_COMPAT
1160 xt_compat_unlock(AF_INET);
1166 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1170 struct ipt_get_entries get;
1173 if (*len < sizeof(get)) {
1174 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1177 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1179 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1180 duprintf("get_entries: %u != %zu\n",
1181 *len, sizeof(get) + get.size);
1185 t = xt_find_table_lock(net, AF_INET, get.name);
1186 if (t && !IS_ERR(t)) {
1187 const struct xt_table_info *private = t->private;
1188 duprintf("t->private->number = %u\n", private->number);
1189 if (get.size == private->size)
1190 ret = copy_entries_to_user(private->size,
1191 t, uptr->entrytable);
1193 duprintf("get_entries: I've got %u not %u!\n",
1194 private->size, get.size);
1200 ret = t ? PTR_ERR(t) : -ENOENT;
1206 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1207 struct xt_table_info *newinfo, unsigned int num_counters,
1208 void __user *counters_ptr)
1212 struct xt_table_info *oldinfo;
1213 struct xt_counters *counters;
1214 void *loc_cpu_old_entry;
1215 struct ipt_entry *iter;
1218 counters = vmalloc(num_counters * sizeof(struct xt_counters));
1224 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1225 "iptable_%s", name);
1226 if (!t || IS_ERR(t)) {
1227 ret = t ? PTR_ERR(t) : -ENOENT;
1228 goto free_newinfo_counters_untrans;
1232 if (valid_hooks != t->valid_hooks) {
1233 duprintf("Valid hook crap: %08X vs %08X\n",
1234 valid_hooks, t->valid_hooks);
1239 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1243 /* Update module usage count based on number of rules */
1244 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1245 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1246 if ((oldinfo->number > oldinfo->initial_entries) ||
1247 (newinfo->number <= oldinfo->initial_entries))
1249 if ((oldinfo->number > oldinfo->initial_entries) &&
1250 (newinfo->number <= oldinfo->initial_entries))
1253 /* Get the old counters, and synchronize with replace */
1254 get_counters(oldinfo, counters);
1256 /* Decrease module usage counts and free resource */
1257 loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1258 xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1259 cleanup_entry(iter, net);
1261 xt_free_table_info(oldinfo);
1262 if (copy_to_user(counters_ptr, counters,
1263 sizeof(struct xt_counters) * num_counters) != 0)
1272 free_newinfo_counters_untrans:
1279 do_replace(struct net *net, const void __user *user, unsigned int len)
1282 struct ipt_replace tmp;
1283 struct xt_table_info *newinfo;
1284 void *loc_cpu_entry;
1285 struct ipt_entry *iter;
1287 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1290 /* overflow check */
1291 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1294 newinfo = xt_alloc_table_info(tmp.size);
1298 /* choose the copy that is on our node/cpu */
1299 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1300 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1306 ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1310 duprintf("Translated table\n");
1312 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1313 tmp.num_counters, tmp.counters);
1315 goto free_newinfo_untrans;
1318 free_newinfo_untrans:
1319 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1320 cleanup_entry(iter, net);
1322 xt_free_table_info(newinfo);
1327 do_add_counters(struct net *net, const void __user *user,
1328 unsigned int len, int compat)
1330 unsigned int i, curcpu;
1331 struct xt_counters_info tmp;
1332 struct xt_counters *paddc;
1333 unsigned int num_counters;
1338 const struct xt_table_info *private;
1340 void *loc_cpu_entry;
1341 struct ipt_entry *iter;
1342 #ifdef CONFIG_COMPAT
1343 struct compat_xt_counters_info compat_tmp;
1347 size = sizeof(struct compat_xt_counters_info);
1352 size = sizeof(struct xt_counters_info);
1355 if (copy_from_user(ptmp, user, size) != 0)
1358 #ifdef CONFIG_COMPAT
1360 num_counters = compat_tmp.num_counters;
1361 name = compat_tmp.name;
1365 num_counters = tmp.num_counters;
1369 if (len != size + num_counters * sizeof(struct xt_counters))
1372 paddc = vmalloc_node(len - size, numa_node_id());
1376 if (copy_from_user(paddc, user + size, len - size) != 0) {
1381 t = xt_find_table_lock(net, AF_INET, name);
1382 if (!t || IS_ERR(t)) {
1383 ret = t ? PTR_ERR(t) : -ENOENT;
1388 private = t->private;
1389 if (private->number != num_counters) {
1391 goto unlock_up_free;
1395 /* Choose the copy that is on our node */
1396 curcpu = smp_processor_id();
1397 loc_cpu_entry = private->entries[curcpu];
1398 xt_info_wrlock(curcpu);
1399 xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1400 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1403 xt_info_wrunlock(curcpu);
1414 #ifdef CONFIG_COMPAT
1415 struct compat_ipt_replace {
1416 char name[IPT_TABLE_MAXNAMELEN];
1420 u32 hook_entry[NF_INET_NUMHOOKS];
1421 u32 underflow[NF_INET_NUMHOOKS];
1423 compat_uptr_t counters; /* struct ipt_counters * */
1424 struct compat_ipt_entry entries[0];
1428 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1429 unsigned int *size, struct xt_counters *counters,
1432 struct ipt_entry_target *t;
1433 struct compat_ipt_entry __user *ce;
1434 u_int16_t target_offset, next_offset;
1435 compat_uint_t origsize;
1436 const struct xt_entry_match *ematch;
1440 ce = (struct compat_ipt_entry __user *)*dstptr;
1441 if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1442 copy_to_user(&ce->counters, &counters[i],
1443 sizeof(counters[i])) != 0)
1446 *dstptr += sizeof(struct compat_ipt_entry);
1447 *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1449 xt_ematch_foreach(ematch, e) {
1450 ret = xt_compat_match_to_user(ematch, dstptr, size);
1454 target_offset = e->target_offset - (origsize - *size);
1455 t = ipt_get_target(e);
1456 ret = xt_compat_target_to_user(t, dstptr, size);
1459 next_offset = e->next_offset - (origsize - *size);
1460 if (put_user(target_offset, &ce->target_offset) != 0 ||
1461 put_user(next_offset, &ce->next_offset) != 0)
1467 compat_find_calc_match(struct ipt_entry_match *m,
1469 const struct ipt_ip *ip,
1470 unsigned int hookmask,
1473 struct xt_match *match;
1475 match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
1476 m->u.user.revision),
1477 "ipt_%s", m->u.user.name);
1478 if (IS_ERR(match) || !match) {
1479 duprintf("compat_check_calc_match: `%s' not found\n",
1481 return match ? PTR_ERR(match) : -ENOENT;
1483 m->u.kernel.match = match;
1484 *size += xt_compat_match_offset(match);
1488 static void compat_release_entry(struct compat_ipt_entry *e)
1490 struct ipt_entry_target *t;
1491 struct xt_entry_match *ematch;
1493 /* Cleanup all matches */
1494 xt_ematch_foreach(ematch, e)
1495 module_put(ematch->u.kernel.match->me);
1496 t = compat_ipt_get_target(e);
1497 module_put(t->u.kernel.target->me);
1501 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1502 struct xt_table_info *newinfo,
1504 const unsigned char *base,
1505 const unsigned char *limit,
1506 const unsigned int *hook_entries,
1507 const unsigned int *underflows,
1510 struct xt_entry_match *ematch;
1511 struct ipt_entry_target *t;
1512 struct xt_target *target;
1513 unsigned int entry_offset;
1517 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1518 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1519 (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1520 duprintf("Bad offset %p, limit = %p\n", e, limit);
1524 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1525 sizeof(struct compat_xt_entry_target)) {
1526 duprintf("checking: element %p size %u\n",
1531 /* For purposes of check_entry casting the compat entry is fine */
1532 ret = check_entry((struct ipt_entry *)e, name);
1536 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1537 entry_offset = (void *)e - (void *)base;
1539 xt_ematch_foreach(ematch, e) {
1540 ret = compat_find_calc_match(ematch, name,
1541 &e->ip, e->comefrom, &off);
1543 goto release_matches;
1547 t = compat_ipt_get_target(e);
1548 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1549 t->u.user.revision);
1550 if (IS_ERR(target)) {
1551 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1553 ret = PTR_ERR(target);
1554 goto release_matches;
1556 t->u.kernel.target = target;
1558 off += xt_compat_target_offset(target);
1560 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1564 /* Check hooks & underflows */
1565 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1566 if ((unsigned char *)e - base == hook_entries[h])
1567 newinfo->hook_entry[h] = hook_entries[h];
1568 if ((unsigned char *)e - base == underflows[h])
1569 newinfo->underflow[h] = underflows[h];
1572 /* Clear counters and comefrom */
1573 memset(&e->counters, 0, sizeof(e->counters));
1578 module_put(t->u.kernel.target->me);
1580 xt_ematch_foreach(ematch, e) {
1583 module_put(ematch->u.kernel.match->me);
1589 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1590 unsigned int *size, const char *name,
1591 struct xt_table_info *newinfo, unsigned char *base)
1593 struct ipt_entry_target *t;
1594 struct xt_target *target;
1595 struct ipt_entry *de;
1596 unsigned int origsize;
1598 struct xt_entry_match *ematch;
1602 de = (struct ipt_entry *)*dstptr;
1603 memcpy(de, e, sizeof(struct ipt_entry));
1604 memcpy(&de->counters, &e->counters, sizeof(e->counters));
1606 *dstptr += sizeof(struct ipt_entry);
1607 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1609 xt_ematch_foreach(ematch, e) {
1610 ret = xt_compat_match_from_user(ematch, dstptr, size);
1614 de->target_offset = e->target_offset - (origsize - *size);
1615 t = compat_ipt_get_target(e);
1616 target = t->u.kernel.target;
1617 xt_compat_target_from_user(t, dstptr, size);
1619 de->next_offset = e->next_offset - (origsize - *size);
1620 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1621 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1622 newinfo->hook_entry[h] -= origsize - *size;
1623 if ((unsigned char *)de - base < newinfo->underflow[h])
1624 newinfo->underflow[h] -= origsize - *size;
1630 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1632 struct xt_entry_match *ematch;
1633 struct xt_mtchk_param mtpar;
1640 mtpar.entryinfo = &e->ip;
1641 mtpar.hook_mask = e->comefrom;
1642 mtpar.family = NFPROTO_IPV4;
1643 xt_ematch_foreach(ematch, e) {
1644 ret = check_match(ematch, &mtpar);
1646 goto cleanup_matches;
1650 ret = check_target(e, net, name);
1652 goto cleanup_matches;
1656 xt_ematch_foreach(ematch, e) {
1659 cleanup_match(ematch, net);
1665 translate_compat_table(struct net *net,
1667 unsigned int valid_hooks,
1668 struct xt_table_info **pinfo,
1670 unsigned int total_size,
1671 unsigned int number,
1672 unsigned int *hook_entries,
1673 unsigned int *underflows)
1676 struct xt_table_info *newinfo, *info;
1677 void *pos, *entry0, *entry1;
1678 struct compat_ipt_entry *iter0;
1679 struct ipt_entry *iter1;
1686 info->number = number;
1688 /* Init all hooks to impossible value. */
1689 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1690 info->hook_entry[i] = 0xFFFFFFFF;
1691 info->underflow[i] = 0xFFFFFFFF;
1694 duprintf("translate_compat_table: size %u\n", info->size);
1696 xt_compat_lock(AF_INET);
1697 /* Walk through entries, checking offsets. */
1698 xt_entry_foreach(iter0, entry0, total_size) {
1699 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1701 entry0 + total_size,
1712 duprintf("translate_compat_table: %u not %u entries\n",
1717 /* Check hooks all assigned */
1718 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1719 /* Only hooks which are valid */
1720 if (!(valid_hooks & (1 << i)))
1722 if (info->hook_entry[i] == 0xFFFFFFFF) {
1723 duprintf("Invalid hook entry %u %u\n",
1724 i, hook_entries[i]);
1727 if (info->underflow[i] == 0xFFFFFFFF) {
1728 duprintf("Invalid underflow %u %u\n",
1735 newinfo = xt_alloc_table_info(size);
1739 newinfo->number = number;
1740 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1741 newinfo->hook_entry[i] = info->hook_entry[i];
1742 newinfo->underflow[i] = info->underflow[i];
1744 entry1 = newinfo->entries[raw_smp_processor_id()];
1747 xt_entry_foreach(iter0, entry0, total_size) {
1748 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1749 name, newinfo, entry1);
1753 xt_compat_flush_offsets(AF_INET);
1754 xt_compat_unlock(AF_INET);
1759 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1763 xt_entry_foreach(iter1, entry1, newinfo->size) {
1764 ret = compat_check_entry(iter1, net, name);
1771 * The first i matches need cleanup_entry (calls ->destroy)
1772 * because they had called ->check already. The other j-i
1773 * entries need only release.
1777 xt_entry_foreach(iter0, entry0, newinfo->size) {
1782 compat_release_entry(iter0);
1784 xt_entry_foreach(iter1, entry1, newinfo->size) {
1787 cleanup_entry(iter1, net);
1789 xt_free_table_info(newinfo);
1793 /* And one copy for every other CPU */
1794 for_each_possible_cpu(i)
1795 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1796 memcpy(newinfo->entries[i], entry1, newinfo->size);
1800 xt_free_table_info(info);
1804 xt_free_table_info(newinfo);
1806 xt_entry_foreach(iter0, entry0, total_size) {
1809 compat_release_entry(iter0);
1813 xt_compat_flush_offsets(AF_INET);
1814 xt_compat_unlock(AF_INET);
1819 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1822 struct compat_ipt_replace tmp;
1823 struct xt_table_info *newinfo;
1824 void *loc_cpu_entry;
1825 struct ipt_entry *iter;
1827 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1830 /* overflow check */
1831 if (tmp.size >= INT_MAX / num_possible_cpus())
1833 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1836 newinfo = xt_alloc_table_info(tmp.size);
1840 /* choose the copy that is on our node/cpu */
1841 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1842 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1848 ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1849 &newinfo, &loc_cpu_entry, tmp.size,
1850 tmp.num_entries, tmp.hook_entry,
1855 duprintf("compat_do_replace: Translated table\n");
1857 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1858 tmp.num_counters, compat_ptr(tmp.counters));
1860 goto free_newinfo_untrans;
1863 free_newinfo_untrans:
1864 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1865 cleanup_entry(iter, net);
1867 xt_free_table_info(newinfo);
1872 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1877 if (!capable(CAP_NET_ADMIN))
1881 case IPT_SO_SET_REPLACE:
1882 ret = compat_do_replace(sock_net(sk), user, len);
1885 case IPT_SO_SET_ADD_COUNTERS:
1886 ret = do_add_counters(sock_net(sk), user, len, 1);
1890 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1897 struct compat_ipt_get_entries {
1898 char name[IPT_TABLE_MAXNAMELEN];
1900 struct compat_ipt_entry entrytable[0];
1904 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1905 void __user *userptr)
1907 struct xt_counters *counters;
1908 const struct xt_table_info *private = table->private;
1912 const void *loc_cpu_entry;
1914 struct ipt_entry *iter;
1916 counters = alloc_counters(table);
1917 if (IS_ERR(counters))
1918 return PTR_ERR(counters);
1920 /* choose the copy that is on our node/cpu, ...
1921 * This choice is lazy (because current thread is
1922 * allowed to migrate to another cpu)
1924 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1927 xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1928 ret = compat_copy_entry_to_user(iter, &pos,
1929 &size, counters, i++);
1939 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1943 struct compat_ipt_get_entries get;
1946 if (*len < sizeof(get)) {
1947 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1951 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1954 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1955 duprintf("compat_get_entries: %u != %zu\n",
1956 *len, sizeof(get) + get.size);
1960 xt_compat_lock(AF_INET);
1961 t = xt_find_table_lock(net, AF_INET, get.name);
1962 if (t && !IS_ERR(t)) {
1963 const struct xt_table_info *private = t->private;
1964 struct xt_table_info info;
1965 duprintf("t->private->number = %u\n", private->number);
1966 ret = compat_table_info(private, &info);
1967 if (!ret && get.size == info.size) {
1968 ret = compat_copy_entries_to_user(private->size,
1969 t, uptr->entrytable);
1971 duprintf("compat_get_entries: I've got %u not %u!\n",
1972 private->size, get.size);
1975 xt_compat_flush_offsets(AF_INET);
1979 ret = t ? PTR_ERR(t) : -ENOENT;
1981 xt_compat_unlock(AF_INET);
1985 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1988 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1992 if (!capable(CAP_NET_ADMIN))
1996 case IPT_SO_GET_INFO:
1997 ret = get_info(sock_net(sk), user, len, 1);
1999 case IPT_SO_GET_ENTRIES:
2000 ret = compat_get_entries(sock_net(sk), user, len);
2003 ret = do_ipt_get_ctl(sk, cmd, user, len);
2010 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2014 if (!capable(CAP_NET_ADMIN))
2018 case IPT_SO_SET_REPLACE:
2019 ret = do_replace(sock_net(sk), user, len);
2022 case IPT_SO_SET_ADD_COUNTERS:
2023 ret = do_add_counters(sock_net(sk), user, len, 0);
2027 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
2035 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2039 if (!capable(CAP_NET_ADMIN))
2043 case IPT_SO_GET_INFO:
2044 ret = get_info(sock_net(sk), user, len, 0);
2047 case IPT_SO_GET_ENTRIES:
2048 ret = get_entries(sock_net(sk), user, len);
2051 case IPT_SO_GET_REVISION_MATCH:
2052 case IPT_SO_GET_REVISION_TARGET: {
2053 struct ipt_get_revision rev;
2056 if (*len != sizeof(rev)) {
2060 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2065 if (cmd == IPT_SO_GET_REVISION_TARGET)
2070 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2073 "ipt_%s", rev.name);
2078 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2085 struct xt_table *ipt_register_table(struct net *net,
2086 const struct xt_table *table,
2087 const struct ipt_replace *repl)
2090 struct xt_table_info *newinfo;
2091 struct xt_table_info bootstrap
2092 = { 0, 0, 0, { 0 }, { 0 }, { } };
2093 void *loc_cpu_entry;
2094 struct xt_table *new_table;
2096 newinfo = xt_alloc_table_info(repl->size);
2102 /* choose the copy on our node/cpu, but dont care about preemption */
2103 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2104 memcpy(loc_cpu_entry, repl->entries, repl->size);
2106 ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2110 new_table = xt_register_table(net, table, &bootstrap, newinfo);
2111 if (IS_ERR(new_table)) {
2112 ret = PTR_ERR(new_table);
2119 xt_free_table_info(newinfo);
2121 return ERR_PTR(ret);
2124 void ipt_unregister_table(struct net *net, struct xt_table *table)
2126 struct xt_table_info *private;
2127 void *loc_cpu_entry;
2128 struct module *table_owner = table->me;
2129 struct ipt_entry *iter;
2131 private = xt_unregister_table(table);
2133 /* Decrease module usage counts and free resources */
2134 loc_cpu_entry = private->entries[raw_smp_processor_id()];
2135 xt_entry_foreach(iter, loc_cpu_entry, private->size)
2136 cleanup_entry(iter, net);
2137 if (private->number > private->initial_entries)
2138 module_put(table_owner);
2139 xt_free_table_info(private);
2142 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2144 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2145 u_int8_t type, u_int8_t code,
2148 return ((test_type == 0xFF) ||
2149 (type == test_type && code >= min_code && code <= max_code))
2154 icmp_match(const struct sk_buff *skb, const struct xt_match_param *par)
2156 const struct icmphdr *ic;
2157 struct icmphdr _icmph;
2158 const struct ipt_icmp *icmpinfo = par->matchinfo;
2160 /* Must not be a fragment. */
2161 if (par->fragoff != 0)
2164 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2166 /* We've been asked to examine this packet, and we
2167 * can't. Hence, no choice but to drop.
2169 duprintf("Dropping evil ICMP tinygram.\n");
2170 *par->hotdrop = true;
2174 return icmp_type_code_match(icmpinfo->type,
2178 !!(icmpinfo->invflags&IPT_ICMP_INV));
2181 static bool icmp_checkentry(const struct xt_mtchk_param *par)
2183 const struct ipt_icmp *icmpinfo = par->matchinfo;
2185 /* Must specify no unknown invflags */
2186 return !(icmpinfo->invflags & ~IPT_ICMP_INV);
2189 /* The built-in targets: standard (NULL) and error. */
2190 static struct xt_target ipt_standard_target __read_mostly = {
2191 .name = IPT_STANDARD_TARGET,
2192 .targetsize = sizeof(int),
2193 .family = NFPROTO_IPV4,
2194 #ifdef CONFIG_COMPAT
2195 .compatsize = sizeof(compat_int_t),
2196 .compat_from_user = compat_standard_from_user,
2197 .compat_to_user = compat_standard_to_user,
2201 static struct xt_target ipt_error_target __read_mostly = {
2202 .name = IPT_ERROR_TARGET,
2203 .target = ipt_error,
2204 .targetsize = IPT_FUNCTION_MAXNAMELEN,
2205 .family = NFPROTO_IPV4,
2208 static struct nf_sockopt_ops ipt_sockopts = {
2210 .set_optmin = IPT_BASE_CTL,
2211 .set_optmax = IPT_SO_SET_MAX+1,
2212 .set = do_ipt_set_ctl,
2213 #ifdef CONFIG_COMPAT
2214 .compat_set = compat_do_ipt_set_ctl,
2216 .get_optmin = IPT_BASE_CTL,
2217 .get_optmax = IPT_SO_GET_MAX+1,
2218 .get = do_ipt_get_ctl,
2219 #ifdef CONFIG_COMPAT
2220 .compat_get = compat_do_ipt_get_ctl,
2222 .owner = THIS_MODULE,
2225 static struct xt_match icmp_matchstruct __read_mostly = {
2227 .match = icmp_match,
2228 .matchsize = sizeof(struct ipt_icmp),
2229 .checkentry = icmp_checkentry,
2230 .proto = IPPROTO_ICMP,
2231 .family = NFPROTO_IPV4,
2234 static int __net_init ip_tables_net_init(struct net *net)
2236 return xt_proto_init(net, NFPROTO_IPV4);
2239 static void __net_exit ip_tables_net_exit(struct net *net)
2241 xt_proto_fini(net, NFPROTO_IPV4);
2244 static struct pernet_operations ip_tables_net_ops = {
2245 .init = ip_tables_net_init,
2246 .exit = ip_tables_net_exit,
2249 static int __init ip_tables_init(void)
2253 ret = register_pernet_subsys(&ip_tables_net_ops);
2257 /* Noone else will be downing sem now, so we won't sleep */
2258 ret = xt_register_target(&ipt_standard_target);
2261 ret = xt_register_target(&ipt_error_target);
2264 ret = xt_register_match(&icmp_matchstruct);
2268 /* Register setsockopt */
2269 ret = nf_register_sockopt(&ipt_sockopts);
2273 pr_info("(C) 2000-2006 Netfilter Core Team\n");
2277 xt_unregister_match(&icmp_matchstruct);
2279 xt_unregister_target(&ipt_error_target);
2281 xt_unregister_target(&ipt_standard_target);
2283 unregister_pernet_subsys(&ip_tables_net_ops);
2288 static void __exit ip_tables_fini(void)
2290 nf_unregister_sockopt(&ipt_sockopts);
2292 xt_unregister_match(&icmp_matchstruct);
2293 xt_unregister_target(&ipt_error_target);
2294 xt_unregister_target(&ipt_standard_target);
2296 unregister_pernet_subsys(&ip_tables_net_ops);
2299 EXPORT_SYMBOL(ipt_register_table);
2300 EXPORT_SYMBOL(ipt_unregister_table);
2301 EXPORT_SYMBOL(ipt_do_table);
2302 module_init(ip_tables_init);
2303 module_exit(ip_tables_fini);