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 #include <linux/cache.h>
12 #include <linux/capability.h>
13 #include <linux/skbuff.h>
14 #include <linux/kmod.h>
15 #include <linux/vmalloc.h>
16 #include <linux/netdevice.h>
17 #include <linux/module.h>
18 #include <linux/icmp.h>
20 #include <net/compat.h>
21 #include <asm/uaccess.h>
22 #include <linux/mutex.h>
23 #include <linux/proc_fs.h>
24 #include <linux/err.h>
25 #include <linux/cpumask.h>
27 #include <linux/netfilter/x_tables.h>
28 #include <linux/netfilter_ipv4/ip_tables.h>
29 #include <net/netfilter/nf_log.h>
31 MODULE_LICENSE("GPL");
32 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
33 MODULE_DESCRIPTION("IPv4 packet filter");
35 /*#define DEBUG_IP_FIREWALL*/
36 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
37 /*#define DEBUG_IP_FIREWALL_USER*/
39 #ifdef DEBUG_IP_FIREWALL
40 #define dprintf(format, args...) printk(format , ## args)
42 #define dprintf(format, args...)
45 #ifdef DEBUG_IP_FIREWALL_USER
46 #define duprintf(format, args...) printk(format , ## args)
48 #define duprintf(format, args...)
51 #ifdef CONFIG_NETFILTER_DEBUG
52 #define IP_NF_ASSERT(x) \
55 printk("IP_NF_ASSERT: %s:%s:%u\n", \
56 __func__, __FILE__, __LINE__); \
59 #define IP_NF_ASSERT(x)
63 /* All the better to debug you with... */
69 We keep a set of rules for each CPU, so we can avoid write-locking
70 them in the softirq when updating the counters and therefore
71 only need to read-lock in the softirq; doing a write_lock_bh() in user
72 context stops packets coming through and allows user context to read
73 the counters or update the rules.
75 Hence the start of any table is given by get_table() below. */
77 /* Returns whether matches rule or not. */
78 /* Performance critical - called for every packet */
80 ip_packet_match(const struct iphdr *ip,
83 const struct ipt_ip *ipinfo,
89 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
91 if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
93 || FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
95 dprintf("Source or dest mismatch.\n");
97 dprintf("SRC: %u.%u.%u.%u. Mask: %u.%u.%u.%u. Target: %u.%u.%u.%u.%s\n",
99 NIPQUAD(ipinfo->smsk.s_addr),
100 NIPQUAD(ipinfo->src.s_addr),
101 ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
102 dprintf("DST: %u.%u.%u.%u Mask: %u.%u.%u.%u Target: %u.%u.%u.%u.%s\n",
104 NIPQUAD(ipinfo->dmsk.s_addr),
105 NIPQUAD(ipinfo->dst.s_addr),
106 ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
110 /* Look for ifname matches; this should unroll nicely. */
111 for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
112 ret |= (((const unsigned long *)indev)[i]
113 ^ ((const unsigned long *)ipinfo->iniface)[i])
114 & ((const unsigned long *)ipinfo->iniface_mask)[i];
117 if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
118 dprintf("VIA in mismatch (%s vs %s).%s\n",
119 indev, ipinfo->iniface,
120 ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
124 for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
125 ret |= (((const unsigned long *)outdev)[i]
126 ^ ((const unsigned long *)ipinfo->outiface)[i])
127 & ((const unsigned long *)ipinfo->outiface_mask)[i];
130 if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
131 dprintf("VIA out mismatch (%s vs %s).%s\n",
132 outdev, ipinfo->outiface,
133 ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
137 /* Check specific protocol */
139 && FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
140 dprintf("Packet protocol %hi does not match %hi.%s\n",
141 ip->protocol, ipinfo->proto,
142 ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
146 /* If we have a fragment rule but the packet is not a fragment
147 * then we return zero */
148 if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
149 dprintf("Fragment rule but not fragment.%s\n",
150 ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
158 ip_checkentry(const struct ipt_ip *ip)
160 if (ip->flags & ~IPT_F_MASK) {
161 duprintf("Unknown flag bits set: %08X\n",
162 ip->flags & ~IPT_F_MASK);
165 if (ip->invflags & ~IPT_INV_MASK) {
166 duprintf("Unknown invflag bits set: %08X\n",
167 ip->invflags & ~IPT_INV_MASK);
174 ipt_error(struct sk_buff *skb, const struct xt_target_param *par)
177 printk("ip_tables: error: `%s'\n",
178 (const char *)par->targinfo);
183 /* Performance critical - called for every packet */
185 do_match(struct ipt_entry_match *m, const struct sk_buff *skb,
186 struct xt_match_param *par)
188 par->match = m->u.kernel.match;
189 par->matchinfo = m->data;
191 /* Stop iteration if it doesn't match */
192 if (!m->u.kernel.match->match(skb, par))
198 /* Performance critical */
199 static inline struct ipt_entry *
200 get_entry(void *base, unsigned int offset)
202 return (struct ipt_entry *)(base + offset);
205 /* All zeroes == unconditional rule. */
206 /* Mildly perf critical (only if packet tracing is on) */
208 unconditional(const struct ipt_ip *ip)
212 for (i = 0; i < sizeof(*ip)/sizeof(__u32); i++)
213 if (((__u32 *)ip)[i])
220 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
221 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
222 static const char *const hooknames[] = {
223 [NF_INET_PRE_ROUTING] = "PREROUTING",
224 [NF_INET_LOCAL_IN] = "INPUT",
225 [NF_INET_FORWARD] = "FORWARD",
226 [NF_INET_LOCAL_OUT] = "OUTPUT",
227 [NF_INET_POST_ROUTING] = "POSTROUTING",
230 enum nf_ip_trace_comments {
231 NF_IP_TRACE_COMMENT_RULE,
232 NF_IP_TRACE_COMMENT_RETURN,
233 NF_IP_TRACE_COMMENT_POLICY,
236 static const char *const comments[] = {
237 [NF_IP_TRACE_COMMENT_RULE] = "rule",
238 [NF_IP_TRACE_COMMENT_RETURN] = "return",
239 [NF_IP_TRACE_COMMENT_POLICY] = "policy",
242 static struct nf_loginfo trace_loginfo = {
243 .type = NF_LOG_TYPE_LOG,
247 .logflags = NF_LOG_MASK,
252 /* Mildly perf critical (only if packet tracing is on) */
254 get_chainname_rulenum(struct ipt_entry *s, struct ipt_entry *e,
255 char *hookname, char **chainname,
256 char **comment, unsigned int *rulenum)
258 struct ipt_standard_target *t = (void *)ipt_get_target(s);
260 if (strcmp(t->target.u.kernel.target->name, IPT_ERROR_TARGET) == 0) {
261 /* Head of user chain: ERROR target with chainname */
262 *chainname = t->target.data;
267 if (s->target_offset == sizeof(struct ipt_entry)
268 && strcmp(t->target.u.kernel.target->name,
269 IPT_STANDARD_TARGET) == 0
271 && unconditional(&s->ip)) {
272 /* Tail of chains: STANDARD target (return/policy) */
273 *comment = *chainname == hookname
274 ? (char *)comments[NF_IP_TRACE_COMMENT_POLICY]
275 : (char *)comments[NF_IP_TRACE_COMMENT_RETURN];
284 static void trace_packet(struct sk_buff *skb,
286 const struct net_device *in,
287 const struct net_device *out,
288 const char *tablename,
289 struct xt_table_info *private,
293 const struct ipt_entry *root;
294 char *hookname, *chainname, *comment;
295 unsigned int rulenum = 0;
297 table_base = (void *)private->entries[smp_processor_id()];
298 root = get_entry(table_base, private->hook_entry[hook]);
300 hookname = chainname = (char *)hooknames[hook];
301 comment = (char *)comments[NF_IP_TRACE_COMMENT_RULE];
303 IPT_ENTRY_ITERATE(root,
304 private->size - private->hook_entry[hook],
305 get_chainname_rulenum,
306 e, hookname, &chainname, &comment, &rulenum);
308 nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
309 "TRACE: %s:%s:%s:%u ",
310 tablename, chainname, comment, rulenum);
314 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
316 ipt_do_table(struct sk_buff *skb,
318 const struct net_device *in,
319 const struct net_device *out,
320 struct xt_table *table)
322 static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
323 const struct iphdr *ip;
325 bool hotdrop = false;
326 /* Initializing verdict to NF_DROP keeps gcc happy. */
327 unsigned int verdict = NF_DROP;
328 const char *indev, *outdev;
330 struct ipt_entry *e, *back;
331 struct xt_table_info *private;
332 struct xt_match_param mtpar;
333 struct xt_target_param tgpar;
337 datalen = skb->len - ip->ihl * 4;
338 indev = in ? in->name : nulldevname;
339 outdev = out ? out->name : nulldevname;
340 /* We handle fragments by dealing with the first fragment as
341 * if it was a normal packet. All other fragments are treated
342 * normally, except that they will NEVER match rules that ask
343 * things we don't know, ie. tcp syn flag or ports). If the
344 * rule is also a fragment-specific rule, non-fragments won't
346 mtpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
347 mtpar.thoff = ip_hdrlen(skb);
348 mtpar.hotdrop = &hotdrop;
349 mtpar.in = tgpar.in = in;
350 mtpar.out = tgpar.out = out;
351 tgpar.hooknum = hook;
353 read_lock_bh(&table->lock);
354 IP_NF_ASSERT(table->valid_hooks & (1 << hook));
355 private = table->private;
356 table_base = (void *)private->entries[smp_processor_id()];
357 e = get_entry(table_base, private->hook_entry[hook]);
359 /* For return from builtin chain */
360 back = get_entry(table_base, private->underflow[hook]);
365 if (ip_packet_match(ip, indev, outdev,
366 &e->ip, mtpar.fragoff)) {
367 struct ipt_entry_target *t;
369 if (IPT_MATCH_ITERATE(e, do_match, skb, &mtpar) != 0)
372 ADD_COUNTER(e->counters, ntohs(ip->tot_len), 1);
374 t = ipt_get_target(e);
375 IP_NF_ASSERT(t->u.kernel.target);
377 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
378 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
379 /* The packet is traced: log it */
380 if (unlikely(skb->nf_trace))
381 trace_packet(skb, hook, in, out,
382 table->name, private, e);
384 /* Standard target? */
385 if (!t->u.kernel.target->target) {
388 v = ((struct ipt_standard_target *)t)->verdict;
390 /* Pop from stack? */
391 if (v != IPT_RETURN) {
392 verdict = (unsigned)(-v) - 1;
396 back = get_entry(table_base,
400 if (table_base + v != (void *)e + e->next_offset
401 && !(e->ip.flags & IPT_F_GOTO)) {
402 /* Save old back ptr in next entry */
403 struct ipt_entry *next
404 = (void *)e + e->next_offset;
406 = (void *)back - table_base;
407 /* set back pointer to next entry */
411 e = get_entry(table_base, v);
413 /* Targets which reenter must return
415 tgpar.target = t->u.kernel.target;
416 tgpar.targinfo = t->data;
417 #ifdef CONFIG_NETFILTER_DEBUG
418 ((struct ipt_entry *)table_base)->comefrom
421 verdict = t->u.kernel.target->target(skb,
423 #ifdef CONFIG_NETFILTER_DEBUG
424 if (((struct ipt_entry *)table_base)->comefrom
426 && verdict == IPT_CONTINUE) {
427 printk("Target %s reentered!\n",
428 t->u.kernel.target->name);
431 ((struct ipt_entry *)table_base)->comefrom
434 /* Target might have changed stuff. */
436 datalen = skb->len - ip->ihl * 4;
438 if (verdict == IPT_CONTINUE)
439 e = (void *)e + e->next_offset;
447 e = (void *)e + e->next_offset;
451 read_unlock_bh(&table->lock);
453 #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(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 struct ipt_standard_target *t
484 = (void *)ipt_get_target(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)
499 && unconditional(&e->ip)) || visited) {
500 unsigned int oldpos, size;
502 if (t->verdict < -NF_MAX_VERDICT - 1) {
503 duprintf("mark_source_chains: bad "
504 "negative verdict (%i)\n",
509 /* Return: backtrack through the last
512 e->comefrom ^= (1<<NF_INET_NUMHOOKS);
513 #ifdef DEBUG_IP_FIREWALL_USER
515 & (1 << NF_INET_NUMHOOKS)) {
516 duprintf("Back unset "
523 pos = e->counters.pcnt;
524 e->counters.pcnt = 0;
526 /* We're at the start. */
530 e = (struct ipt_entry *)
532 } while (oldpos == pos + e->next_offset);
535 size = e->next_offset;
536 e = (struct ipt_entry *)
537 (entry0 + pos + size);
538 e->counters.pcnt = pos;
541 int newpos = t->verdict;
543 if (strcmp(t->target.u.user.name,
544 IPT_STANDARD_TARGET) == 0
546 if (newpos > newinfo->size -
547 sizeof(struct ipt_entry)) {
548 duprintf("mark_source_chains: "
549 "bad verdict (%i)\n",
553 /* This a jump; chase it. */
554 duprintf("Jump rule %u -> %u\n",
557 /* ... this is a fallthru */
558 newpos = pos + e->next_offset;
560 e = (struct ipt_entry *)
562 e->counters.pcnt = pos;
567 duprintf("Finished chain %u\n", hook);
573 cleanup_match(struct ipt_entry_match *m, unsigned int *i)
575 struct xt_mtdtor_param par;
577 if (i && (*i)-- == 0)
580 par.match = m->u.kernel.match;
581 par.matchinfo = m->data;
582 if (par.match->destroy != NULL)
583 par.match->destroy(&par);
584 module_put(par.match->me);
589 check_entry(struct ipt_entry *e, const char *name)
591 struct ipt_entry_target *t;
593 if (!ip_checkentry(&e->ip)) {
594 duprintf("ip_tables: ip check failed %p %s.\n", e, name);
598 if (e->target_offset + sizeof(struct ipt_entry_target) >
602 t = ipt_get_target(e);
603 if (e->target_offset + t->u.target_size > e->next_offset)
610 check_match(struct ipt_entry_match *m, struct xt_mtchk_param *par,
613 const struct ipt_ip *ip = par->entryinfo;
616 par->match = m->u.kernel.match;
617 par->matchinfo = m->data;
619 ret = xt_check_match(par, NFPROTO_IPV4, m->u.match_size - sizeof(*m),
620 ip->proto, ip->invflags & IPT_INV_PROTO);
622 duprintf("ip_tables: check failed for `%s'.\n",
631 find_check_match(struct ipt_entry_match *m, struct xt_mtchk_param *par,
634 struct xt_match *match;
637 match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
639 "ipt_%s", m->u.user.name);
640 if (IS_ERR(match) || !match) {
641 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
642 return match ? PTR_ERR(match) : -ENOENT;
644 m->u.kernel.match = match;
646 ret = check_match(m, par, i);
652 module_put(m->u.kernel.match->me);
656 static int check_target(struct ipt_entry *e, const char *name)
658 struct ipt_entry_target *t;
659 struct xt_target *target;
662 t = ipt_get_target(e);
663 target = t->u.kernel.target;
664 ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
665 name, e->comefrom, e->ip.proto,
666 e->ip.invflags & IPT_INV_PROTO, e, t->data);
668 duprintf("ip_tables: check failed for `%s'.\n",
669 t->u.kernel.target->name);
676 find_check_entry(struct ipt_entry *e, const char *name, unsigned int size,
679 struct ipt_entry_target *t;
680 struct xt_target *target;
683 struct xt_mtchk_param mtpar;
685 ret = check_entry(e, name);
691 mtpar.entryinfo = &e->ip;
692 mtpar.hook_mask = e->comefrom;
693 ret = IPT_MATCH_ITERATE(e, find_check_match, &mtpar, &j);
695 goto cleanup_matches;
697 t = ipt_get_target(e);
698 target = try_then_request_module(xt_find_target(AF_INET,
701 "ipt_%s", t->u.user.name);
702 if (IS_ERR(target) || !target) {
703 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
704 ret = target ? PTR_ERR(target) : -ENOENT;
705 goto cleanup_matches;
707 t->u.kernel.target = target;
709 ret = check_target(e, name);
716 module_put(t->u.kernel.target->me);
718 IPT_MATCH_ITERATE(e, cleanup_match, &j);
723 check_entry_size_and_hooks(struct ipt_entry *e,
724 struct xt_table_info *newinfo,
726 unsigned char *limit,
727 const unsigned int *hook_entries,
728 const unsigned int *underflows,
733 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0
734 || (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
735 duprintf("Bad offset %p\n", e);
740 < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) {
741 duprintf("checking: element %p size %u\n",
746 /* Check hooks & underflows */
747 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
748 if ((unsigned char *)e - base == hook_entries[h])
749 newinfo->hook_entry[h] = hook_entries[h];
750 if ((unsigned char *)e - base == underflows[h])
751 newinfo->underflow[h] = underflows[h];
754 /* FIXME: underflows must be unconditional, standard verdicts
755 < 0 (not IPT_RETURN). --RR */
757 /* Clear counters and comefrom */
758 e->counters = ((struct xt_counters) { 0, 0 });
766 cleanup_entry(struct ipt_entry *e, unsigned int *i)
768 struct ipt_entry_target *t;
770 if (i && (*i)-- == 0)
773 /* Cleanup all matches */
774 IPT_MATCH_ITERATE(e, cleanup_match, NULL);
775 t = ipt_get_target(e);
776 if (t->u.kernel.target->destroy)
777 t->u.kernel.target->destroy(t->u.kernel.target, t->data);
778 module_put(t->u.kernel.target->me);
782 /* Checks and translates the user-supplied table segment (held in
785 translate_table(const char *name,
786 unsigned int valid_hooks,
787 struct xt_table_info *newinfo,
791 const unsigned int *hook_entries,
792 const unsigned int *underflows)
797 newinfo->size = size;
798 newinfo->number = number;
800 /* Init all hooks to impossible value. */
801 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
802 newinfo->hook_entry[i] = 0xFFFFFFFF;
803 newinfo->underflow[i] = 0xFFFFFFFF;
806 duprintf("translate_table: size %u\n", newinfo->size);
808 /* Walk through entries, checking offsets. */
809 ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
810 check_entry_size_and_hooks,
814 hook_entries, underflows, &i);
819 duprintf("translate_table: %u not %u entries\n",
824 /* Check hooks all assigned */
825 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
826 /* Only hooks which are valid */
827 if (!(valid_hooks & (1 << i)))
829 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
830 duprintf("Invalid hook entry %u %u\n",
834 if (newinfo->underflow[i] == 0xFFFFFFFF) {
835 duprintf("Invalid underflow %u %u\n",
841 if (!mark_source_chains(newinfo, valid_hooks, entry0))
844 /* Finally, each sanity check must pass */
846 ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
847 find_check_entry, name, size, &i);
850 IPT_ENTRY_ITERATE(entry0, newinfo->size,
855 /* And one copy for every other CPU */
856 for_each_possible_cpu(i) {
857 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
858 memcpy(newinfo->entries[i], entry0, newinfo->size);
866 add_entry_to_counter(const struct ipt_entry *e,
867 struct xt_counters total[],
870 ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
877 set_entry_to_counter(const struct ipt_entry *e,
878 struct ipt_counters total[],
881 SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
888 get_counters(const struct xt_table_info *t,
889 struct xt_counters counters[])
895 /* Instead of clearing (by a previous call to memset())
896 * the counters and using adds, we set the counters
897 * with data used by 'current' CPU
898 * We dont care about preemption here.
900 curcpu = raw_smp_processor_id();
903 IPT_ENTRY_ITERATE(t->entries[curcpu],
905 set_entry_to_counter,
909 for_each_possible_cpu(cpu) {
913 IPT_ENTRY_ITERATE(t->entries[cpu],
915 add_entry_to_counter,
921 static struct xt_counters * alloc_counters(struct xt_table *table)
923 unsigned int countersize;
924 struct xt_counters *counters;
925 const struct xt_table_info *private = table->private;
927 /* We need atomic snapshot of counters: rest doesn't change
928 (other than comefrom, which userspace doesn't care
930 countersize = sizeof(struct xt_counters) * private->number;
931 counters = vmalloc_node(countersize, numa_node_id());
933 if (counters == NULL)
934 return ERR_PTR(-ENOMEM);
936 /* First, sum counters... */
937 write_lock_bh(&table->lock);
938 get_counters(private, counters);
939 write_unlock_bh(&table->lock);
945 copy_entries_to_user(unsigned int total_size,
946 struct xt_table *table,
947 void __user *userptr)
949 unsigned int off, num;
951 struct xt_counters *counters;
952 const struct xt_table_info *private = table->private;
954 const void *loc_cpu_entry;
956 counters = alloc_counters(table);
957 if (IS_ERR(counters))
958 return PTR_ERR(counters);
960 /* choose the copy that is on our node/cpu, ...
961 * This choice is lazy (because current thread is
962 * allowed to migrate to another cpu)
964 loc_cpu_entry = private->entries[raw_smp_processor_id()];
965 if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
970 /* FIXME: use iterator macros --RR */
971 /* ... then go back and fix counters and names */
972 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
974 const struct ipt_entry_match *m;
975 const struct ipt_entry_target *t;
977 e = (struct ipt_entry *)(loc_cpu_entry + off);
978 if (copy_to_user(userptr + off
979 + offsetof(struct ipt_entry, counters),
981 sizeof(counters[num])) != 0) {
986 for (i = sizeof(struct ipt_entry);
987 i < e->target_offset;
988 i += m->u.match_size) {
991 if (copy_to_user(userptr + off + i
992 + offsetof(struct ipt_entry_match,
994 m->u.kernel.match->name,
995 strlen(m->u.kernel.match->name)+1)
1002 t = ipt_get_target(e);
1003 if (copy_to_user(userptr + off + e->target_offset
1004 + offsetof(struct ipt_entry_target,
1006 t->u.kernel.target->name,
1007 strlen(t->u.kernel.target->name)+1) != 0) {
1018 #ifdef CONFIG_COMPAT
1019 static void compat_standard_from_user(void *dst, void *src)
1021 int v = *(compat_int_t *)src;
1024 v += xt_compat_calc_jump(AF_INET, v);
1025 memcpy(dst, &v, sizeof(v));
1028 static int compat_standard_to_user(void __user *dst, void *src)
1030 compat_int_t cv = *(int *)src;
1033 cv -= xt_compat_calc_jump(AF_INET, cv);
1034 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1038 compat_calc_match(struct ipt_entry_match *m, int *size)
1040 *size += xt_compat_match_offset(m->u.kernel.match);
1044 static int compat_calc_entry(struct ipt_entry *e,
1045 const struct xt_table_info *info,
1046 void *base, struct xt_table_info *newinfo)
1048 struct ipt_entry_target *t;
1049 unsigned int entry_offset;
1052 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1053 entry_offset = (void *)e - base;
1054 IPT_MATCH_ITERATE(e, compat_calc_match, &off);
1055 t = ipt_get_target(e);
1056 off += xt_compat_target_offset(t->u.kernel.target);
1057 newinfo->size -= off;
1058 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1062 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1063 if (info->hook_entry[i] &&
1064 (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1065 newinfo->hook_entry[i] -= off;
1066 if (info->underflow[i] &&
1067 (e < (struct ipt_entry *)(base + info->underflow[i])))
1068 newinfo->underflow[i] -= off;
1073 static int compat_table_info(const struct xt_table_info *info,
1074 struct xt_table_info *newinfo)
1076 void *loc_cpu_entry;
1078 if (!newinfo || !info)
1081 /* we dont care about newinfo->entries[] */
1082 memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1083 newinfo->initial_entries = 0;
1084 loc_cpu_entry = info->entries[raw_smp_processor_id()];
1085 return IPT_ENTRY_ITERATE(loc_cpu_entry, info->size,
1086 compat_calc_entry, info, loc_cpu_entry,
1091 static int get_info(struct net *net, void __user *user, int *len, int compat)
1093 char name[IPT_TABLE_MAXNAMELEN];
1097 if (*len != sizeof(struct ipt_getinfo)) {
1098 duprintf("length %u != %zu\n", *len,
1099 sizeof(struct ipt_getinfo));
1103 if (copy_from_user(name, user, sizeof(name)) != 0)
1106 name[IPT_TABLE_MAXNAMELEN-1] = '\0';
1107 #ifdef CONFIG_COMPAT
1109 xt_compat_lock(AF_INET);
1111 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1112 "iptable_%s", name);
1113 if (t && !IS_ERR(t)) {
1114 struct ipt_getinfo info;
1115 const struct xt_table_info *private = t->private;
1117 #ifdef CONFIG_COMPAT
1119 struct xt_table_info tmp;
1120 ret = compat_table_info(private, &tmp);
1121 xt_compat_flush_offsets(AF_INET);
1125 info.valid_hooks = t->valid_hooks;
1126 memcpy(info.hook_entry, private->hook_entry,
1127 sizeof(info.hook_entry));
1128 memcpy(info.underflow, private->underflow,
1129 sizeof(info.underflow));
1130 info.num_entries = private->number;
1131 info.size = private->size;
1132 strcpy(info.name, name);
1134 if (copy_to_user(user, &info, *len) != 0)
1142 ret = t ? PTR_ERR(t) : -ENOENT;
1143 #ifdef CONFIG_COMPAT
1145 xt_compat_unlock(AF_INET);
1151 get_entries(struct net *net, struct ipt_get_entries __user *uptr, int *len)
1154 struct ipt_get_entries get;
1157 if (*len < sizeof(get)) {
1158 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1161 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1163 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1164 duprintf("get_entries: %u != %zu\n",
1165 *len, sizeof(get) + get.size);
1169 t = xt_find_table_lock(net, AF_INET, get.name);
1170 if (t && !IS_ERR(t)) {
1171 const struct xt_table_info *private = t->private;
1172 duprintf("t->private->number = %u\n", private->number);
1173 if (get.size == private->size)
1174 ret = copy_entries_to_user(private->size,
1175 t, uptr->entrytable);
1177 duprintf("get_entries: I've got %u not %u!\n",
1178 private->size, get.size);
1184 ret = t ? PTR_ERR(t) : -ENOENT;
1190 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1191 struct xt_table_info *newinfo, unsigned int num_counters,
1192 void __user *counters_ptr)
1196 struct xt_table_info *oldinfo;
1197 struct xt_counters *counters;
1198 void *loc_cpu_old_entry;
1201 counters = vmalloc(num_counters * sizeof(struct xt_counters));
1207 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1208 "iptable_%s", name);
1209 if (!t || IS_ERR(t)) {
1210 ret = t ? PTR_ERR(t) : -ENOENT;
1211 goto free_newinfo_counters_untrans;
1215 if (valid_hooks != t->valid_hooks) {
1216 duprintf("Valid hook crap: %08X vs %08X\n",
1217 valid_hooks, t->valid_hooks);
1222 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1226 /* Update module usage count based on number of rules */
1227 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1228 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1229 if ((oldinfo->number > oldinfo->initial_entries) ||
1230 (newinfo->number <= oldinfo->initial_entries))
1232 if ((oldinfo->number > oldinfo->initial_entries) &&
1233 (newinfo->number <= oldinfo->initial_entries))
1236 /* Get the old counters. */
1237 get_counters(oldinfo, counters);
1238 /* Decrease module usage counts and free resource */
1239 loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1240 IPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,
1242 xt_free_table_info(oldinfo);
1243 if (copy_to_user(counters_ptr, counters,
1244 sizeof(struct xt_counters) * num_counters) != 0)
1253 free_newinfo_counters_untrans:
1260 do_replace(struct net *net, void __user *user, unsigned int len)
1263 struct ipt_replace tmp;
1264 struct xt_table_info *newinfo;
1265 void *loc_cpu_entry;
1267 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1270 /* overflow check */
1271 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1274 newinfo = xt_alloc_table_info(tmp.size);
1278 /* choose the copy that is on our node/cpu */
1279 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1280 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1286 ret = translate_table(tmp.name, tmp.valid_hooks,
1287 newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
1288 tmp.hook_entry, tmp.underflow);
1292 duprintf("ip_tables: Translated table\n");
1294 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1295 tmp.num_counters, tmp.counters);
1297 goto free_newinfo_untrans;
1300 free_newinfo_untrans:
1301 IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
1303 xt_free_table_info(newinfo);
1307 /* We're lazy, and add to the first CPU; overflow works its fey magic
1308 * and everything is OK. */
1310 add_counter_to_entry(struct ipt_entry *e,
1311 const struct xt_counters addme[],
1315 duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n",
1317 (long unsigned int)e->counters.pcnt,
1318 (long unsigned int)e->counters.bcnt,
1319 (long unsigned int)addme[*i].pcnt,
1320 (long unsigned int)addme[*i].bcnt);
1323 ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
1330 do_add_counters(struct net *net, void __user *user, unsigned int len, int compat)
1333 struct xt_counters_info tmp;
1334 struct xt_counters *paddc;
1335 unsigned int num_counters;
1340 const struct xt_table_info *private;
1342 void *loc_cpu_entry;
1343 #ifdef CONFIG_COMPAT
1344 struct compat_xt_counters_info compat_tmp;
1348 size = sizeof(struct compat_xt_counters_info);
1353 size = sizeof(struct xt_counters_info);
1356 if (copy_from_user(ptmp, user, size) != 0)
1359 #ifdef CONFIG_COMPAT
1361 num_counters = compat_tmp.num_counters;
1362 name = compat_tmp.name;
1366 num_counters = tmp.num_counters;
1370 if (len != size + num_counters * sizeof(struct xt_counters))
1373 paddc = vmalloc_node(len - size, numa_node_id());
1377 if (copy_from_user(paddc, user + size, len - size) != 0) {
1382 t = xt_find_table_lock(net, AF_INET, name);
1383 if (!t || IS_ERR(t)) {
1384 ret = t ? PTR_ERR(t) : -ENOENT;
1388 write_lock_bh(&t->lock);
1389 private = t->private;
1390 if (private->number != num_counters) {
1392 goto unlock_up_free;
1396 /* Choose the copy that is on our node */
1397 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1398 IPT_ENTRY_ITERATE(loc_cpu_entry,
1400 add_counter_to_entry,
1404 write_unlock_bh(&t->lock);
1413 #ifdef CONFIG_COMPAT
1414 struct compat_ipt_replace {
1415 char name[IPT_TABLE_MAXNAMELEN];
1419 u32 hook_entry[NF_INET_NUMHOOKS];
1420 u32 underflow[NF_INET_NUMHOOKS];
1422 compat_uptr_t counters; /* struct ipt_counters * */
1423 struct compat_ipt_entry entries[0];
1427 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1428 unsigned int *size, struct xt_counters *counters,
1431 struct ipt_entry_target *t;
1432 struct compat_ipt_entry __user *ce;
1433 u_int16_t target_offset, next_offset;
1434 compat_uint_t origsize;
1439 ce = (struct compat_ipt_entry __user *)*dstptr;
1440 if (copy_to_user(ce, e, sizeof(struct ipt_entry)))
1443 if (copy_to_user(&ce->counters, &counters[*i], sizeof(counters[*i])))
1446 *dstptr += sizeof(struct compat_ipt_entry);
1447 *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1449 ret = IPT_MATCH_ITERATE(e, xt_compat_match_to_user, dstptr, size);
1450 target_offset = e->target_offset - (origsize - *size);
1453 t = ipt_get_target(e);
1454 ret = xt_compat_target_to_user(t, dstptr, size);
1458 next_offset = e->next_offset - (origsize - *size);
1459 if (put_user(target_offset, &ce->target_offset))
1461 if (put_user(next_offset, &ce->next_offset))
1471 compat_find_calc_match(struct ipt_entry_match *m,
1473 const struct ipt_ip *ip,
1474 unsigned int hookmask,
1475 int *size, unsigned int *i)
1477 struct xt_match *match;
1479 match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
1480 m->u.user.revision),
1481 "ipt_%s", m->u.user.name);
1482 if (IS_ERR(match) || !match) {
1483 duprintf("compat_check_calc_match: `%s' not found\n",
1485 return match ? PTR_ERR(match) : -ENOENT;
1487 m->u.kernel.match = match;
1488 *size += xt_compat_match_offset(match);
1495 compat_release_match(struct ipt_entry_match *m, unsigned int *i)
1497 if (i && (*i)-- == 0)
1500 module_put(m->u.kernel.match->me);
1505 compat_release_entry(struct compat_ipt_entry *e, unsigned int *i)
1507 struct ipt_entry_target *t;
1509 if (i && (*i)-- == 0)
1512 /* Cleanup all matches */
1513 COMPAT_IPT_MATCH_ITERATE(e, compat_release_match, NULL);
1514 t = compat_ipt_get_target(e);
1515 module_put(t->u.kernel.target->me);
1520 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1521 struct xt_table_info *newinfo,
1523 unsigned char *base,
1524 unsigned char *limit,
1525 unsigned int *hook_entries,
1526 unsigned int *underflows,
1530 struct ipt_entry_target *t;
1531 struct xt_target *target;
1532 unsigned int entry_offset;
1536 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1537 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0
1538 || (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1539 duprintf("Bad offset %p, limit = %p\n", e, limit);
1543 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1544 sizeof(struct compat_xt_entry_target)) {
1545 duprintf("checking: element %p size %u\n",
1550 /* For purposes of check_entry casting the compat entry is fine */
1551 ret = check_entry((struct ipt_entry *)e, name);
1555 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1556 entry_offset = (void *)e - (void *)base;
1558 ret = COMPAT_IPT_MATCH_ITERATE(e, compat_find_calc_match, name,
1559 &e->ip, e->comefrom, &off, &j);
1561 goto release_matches;
1563 t = compat_ipt_get_target(e);
1564 target = try_then_request_module(xt_find_target(AF_INET,
1566 t->u.user.revision),
1567 "ipt_%s", t->u.user.name);
1568 if (IS_ERR(target) || !target) {
1569 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1571 ret = target ? PTR_ERR(target) : -ENOENT;
1572 goto release_matches;
1574 t->u.kernel.target = target;
1576 off += xt_compat_target_offset(target);
1578 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1582 /* Check hooks & underflows */
1583 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1584 if ((unsigned char *)e - base == hook_entries[h])
1585 newinfo->hook_entry[h] = hook_entries[h];
1586 if ((unsigned char *)e - base == underflows[h])
1587 newinfo->underflow[h] = underflows[h];
1590 /* Clear counters and comefrom */
1591 memset(&e->counters, 0, sizeof(e->counters));
1598 module_put(t->u.kernel.target->me);
1600 IPT_MATCH_ITERATE(e, compat_release_match, &j);
1605 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1606 unsigned int *size, const char *name,
1607 struct xt_table_info *newinfo, unsigned char *base)
1609 struct ipt_entry_target *t;
1610 struct xt_target *target;
1611 struct ipt_entry *de;
1612 unsigned int origsize;
1617 de = (struct ipt_entry *)*dstptr;
1618 memcpy(de, e, sizeof(struct ipt_entry));
1619 memcpy(&de->counters, &e->counters, sizeof(e->counters));
1621 *dstptr += sizeof(struct ipt_entry);
1622 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1624 ret = COMPAT_IPT_MATCH_ITERATE(e, xt_compat_match_from_user,
1628 de->target_offset = e->target_offset - (origsize - *size);
1629 t = compat_ipt_get_target(e);
1630 target = t->u.kernel.target;
1631 xt_compat_target_from_user(t, dstptr, size);
1633 de->next_offset = e->next_offset - (origsize - *size);
1634 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1635 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1636 newinfo->hook_entry[h] -= origsize - *size;
1637 if ((unsigned char *)de - base < newinfo->underflow[h])
1638 newinfo->underflow[h] -= origsize - *size;
1644 compat_check_entry(struct ipt_entry *e, const char *name,
1647 struct xt_mtchk_param mtpar;
1653 mtpar.entryinfo = &e->ip;
1654 mtpar.hook_mask = e->comefrom;
1655 ret = IPT_MATCH_ITERATE(e, check_match, &mtpar, &j);
1657 goto cleanup_matches;
1659 ret = check_target(e, name);
1661 goto cleanup_matches;
1667 IPT_MATCH_ITERATE(e, cleanup_match, &j);
1672 translate_compat_table(const char *name,
1673 unsigned int valid_hooks,
1674 struct xt_table_info **pinfo,
1676 unsigned int total_size,
1677 unsigned int number,
1678 unsigned int *hook_entries,
1679 unsigned int *underflows)
1682 struct xt_table_info *newinfo, *info;
1683 void *pos, *entry0, *entry1;
1690 info->number = number;
1692 /* Init all hooks to impossible value. */
1693 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1694 info->hook_entry[i] = 0xFFFFFFFF;
1695 info->underflow[i] = 0xFFFFFFFF;
1698 duprintf("translate_compat_table: size %u\n", info->size);
1700 xt_compat_lock(AF_INET);
1701 /* Walk through entries, checking offsets. */
1702 ret = COMPAT_IPT_ENTRY_ITERATE(entry0, total_size,
1703 check_compat_entry_size_and_hooks,
1704 info, &size, entry0,
1705 entry0 + total_size,
1706 hook_entries, underflows, &j, name);
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 ret = COMPAT_IPT_ENTRY_ITERATE(entry0, total_size,
1748 compat_copy_entry_from_user,
1749 &pos, &size, name, newinfo, entry1);
1750 xt_compat_flush_offsets(AF_INET);
1751 xt_compat_unlock(AF_INET);
1756 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1760 ret = IPT_ENTRY_ITERATE(entry1, newinfo->size, compat_check_entry,
1764 COMPAT_IPT_ENTRY_ITERATE_CONTINUE(entry0, newinfo->size, i,
1765 compat_release_entry, &j);
1766 IPT_ENTRY_ITERATE(entry1, newinfo->size, cleanup_entry, &i);
1767 xt_free_table_info(newinfo);
1771 /* And one copy for every other CPU */
1772 for_each_possible_cpu(i)
1773 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1774 memcpy(newinfo->entries[i], entry1, newinfo->size);
1778 xt_free_table_info(info);
1782 xt_free_table_info(newinfo);
1784 COMPAT_IPT_ENTRY_ITERATE(entry0, total_size, compat_release_entry, &j);
1787 xt_compat_flush_offsets(AF_INET);
1788 xt_compat_unlock(AF_INET);
1793 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1796 struct compat_ipt_replace tmp;
1797 struct xt_table_info *newinfo;
1798 void *loc_cpu_entry;
1800 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1803 /* overflow check */
1804 if (tmp.size >= INT_MAX / num_possible_cpus())
1806 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1809 newinfo = xt_alloc_table_info(tmp.size);
1813 /* choose the copy that is on our node/cpu */
1814 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1815 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1821 ret = translate_compat_table(tmp.name, tmp.valid_hooks,
1822 &newinfo, &loc_cpu_entry, tmp.size,
1823 tmp.num_entries, tmp.hook_entry,
1828 duprintf("compat_do_replace: Translated table\n");
1830 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1831 tmp.num_counters, compat_ptr(tmp.counters));
1833 goto free_newinfo_untrans;
1836 free_newinfo_untrans:
1837 IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
1839 xt_free_table_info(newinfo);
1844 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1849 if (!capable(CAP_NET_ADMIN))
1853 case IPT_SO_SET_REPLACE:
1854 ret = compat_do_replace(sock_net(sk), user, len);
1857 case IPT_SO_SET_ADD_COUNTERS:
1858 ret = do_add_counters(sock_net(sk), user, len, 1);
1862 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1869 struct compat_ipt_get_entries {
1870 char name[IPT_TABLE_MAXNAMELEN];
1872 struct compat_ipt_entry entrytable[0];
1876 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1877 void __user *userptr)
1879 struct xt_counters *counters;
1880 const struct xt_table_info *private = table->private;
1884 const void *loc_cpu_entry;
1887 counters = alloc_counters(table);
1888 if (IS_ERR(counters))
1889 return PTR_ERR(counters);
1891 /* choose the copy that is on our node/cpu, ...
1892 * This choice is lazy (because current thread is
1893 * allowed to migrate to another cpu)
1895 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1898 ret = IPT_ENTRY_ITERATE(loc_cpu_entry, total_size,
1899 compat_copy_entry_to_user,
1900 &pos, &size, counters, &i);
1907 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1911 struct compat_ipt_get_entries get;
1914 if (*len < sizeof(get)) {
1915 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1919 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1922 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1923 duprintf("compat_get_entries: %u != %zu\n",
1924 *len, sizeof(get) + get.size);
1928 xt_compat_lock(AF_INET);
1929 t = xt_find_table_lock(net, AF_INET, get.name);
1930 if (t && !IS_ERR(t)) {
1931 const struct xt_table_info *private = t->private;
1932 struct xt_table_info info;
1933 duprintf("t->private->number = %u\n", private->number);
1934 ret = compat_table_info(private, &info);
1935 if (!ret && get.size == info.size) {
1936 ret = compat_copy_entries_to_user(private->size,
1937 t, uptr->entrytable);
1939 duprintf("compat_get_entries: I've got %u not %u!\n",
1940 private->size, get.size);
1943 xt_compat_flush_offsets(AF_INET);
1947 ret = t ? PTR_ERR(t) : -ENOENT;
1949 xt_compat_unlock(AF_INET);
1953 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1956 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1960 if (!capable(CAP_NET_ADMIN))
1964 case IPT_SO_GET_INFO:
1965 ret = get_info(sock_net(sk), user, len, 1);
1967 case IPT_SO_GET_ENTRIES:
1968 ret = compat_get_entries(sock_net(sk), user, len);
1971 ret = do_ipt_get_ctl(sk, cmd, user, len);
1978 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1982 if (!capable(CAP_NET_ADMIN))
1986 case IPT_SO_SET_REPLACE:
1987 ret = do_replace(sock_net(sk), user, len);
1990 case IPT_SO_SET_ADD_COUNTERS:
1991 ret = do_add_counters(sock_net(sk), user, len, 0);
1995 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
2003 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2007 if (!capable(CAP_NET_ADMIN))
2011 case IPT_SO_GET_INFO:
2012 ret = get_info(sock_net(sk), user, len, 0);
2015 case IPT_SO_GET_ENTRIES:
2016 ret = get_entries(sock_net(sk), user, len);
2019 case IPT_SO_GET_REVISION_MATCH:
2020 case IPT_SO_GET_REVISION_TARGET: {
2021 struct ipt_get_revision rev;
2024 if (*len != sizeof(rev)) {
2028 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2033 if (cmd == IPT_SO_GET_REVISION_TARGET)
2038 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2041 "ipt_%s", rev.name);
2046 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2053 struct xt_table *ipt_register_table(struct net *net, struct xt_table *table,
2054 const struct ipt_replace *repl)
2057 struct xt_table_info *newinfo;
2058 struct xt_table_info bootstrap
2059 = { 0, 0, 0, { 0 }, { 0 }, { } };
2060 void *loc_cpu_entry;
2061 struct xt_table *new_table;
2063 newinfo = xt_alloc_table_info(repl->size);
2069 /* choose the copy on our node/cpu, but dont care about preemption */
2070 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2071 memcpy(loc_cpu_entry, repl->entries, repl->size);
2073 ret = translate_table(table->name, table->valid_hooks,
2074 newinfo, loc_cpu_entry, repl->size,
2081 new_table = xt_register_table(net, table, &bootstrap, newinfo);
2082 if (IS_ERR(new_table)) {
2083 ret = PTR_ERR(new_table);
2090 xt_free_table_info(newinfo);
2092 return ERR_PTR(ret);
2095 void ipt_unregister_table(struct xt_table *table)
2097 struct xt_table_info *private;
2098 void *loc_cpu_entry;
2099 struct module *table_owner = table->me;
2101 private = xt_unregister_table(table);
2103 /* Decrease module usage counts and free resources */
2104 loc_cpu_entry = private->entries[raw_smp_processor_id()];
2105 IPT_ENTRY_ITERATE(loc_cpu_entry, private->size, cleanup_entry, NULL);
2106 if (private->number > private->initial_entries)
2107 module_put(table_owner);
2108 xt_free_table_info(private);
2111 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2113 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2114 u_int8_t type, u_int8_t code,
2117 return ((test_type == 0xFF) ||
2118 (type == test_type && code >= min_code && code <= max_code))
2123 icmp_match(const struct sk_buff *skb, const struct xt_match_param *par)
2125 const struct icmphdr *ic;
2126 struct icmphdr _icmph;
2127 const struct ipt_icmp *icmpinfo = par->matchinfo;
2129 /* Must not be a fragment. */
2130 if (par->fragoff != 0)
2133 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2135 /* We've been asked to examine this packet, and we
2136 * can't. Hence, no choice but to drop.
2138 duprintf("Dropping evil ICMP tinygram.\n");
2139 *par->hotdrop = true;
2143 return icmp_type_code_match(icmpinfo->type,
2147 !!(icmpinfo->invflags&IPT_ICMP_INV));
2150 static bool icmp_checkentry(const struct xt_mtchk_param *par)
2152 const struct ipt_icmp *icmpinfo = par->matchinfo;
2154 /* Must specify no unknown invflags */
2155 return !(icmpinfo->invflags & ~IPT_ICMP_INV);
2158 /* The built-in targets: standard (NULL) and error. */
2159 static struct xt_target ipt_standard_target __read_mostly = {
2160 .name = IPT_STANDARD_TARGET,
2161 .targetsize = sizeof(int),
2163 #ifdef CONFIG_COMPAT
2164 .compatsize = sizeof(compat_int_t),
2165 .compat_from_user = compat_standard_from_user,
2166 .compat_to_user = compat_standard_to_user,
2170 static struct xt_target ipt_error_target __read_mostly = {
2171 .name = IPT_ERROR_TARGET,
2172 .target = ipt_error,
2173 .targetsize = IPT_FUNCTION_MAXNAMELEN,
2177 static struct nf_sockopt_ops ipt_sockopts = {
2179 .set_optmin = IPT_BASE_CTL,
2180 .set_optmax = IPT_SO_SET_MAX+1,
2181 .set = do_ipt_set_ctl,
2182 #ifdef CONFIG_COMPAT
2183 .compat_set = compat_do_ipt_set_ctl,
2185 .get_optmin = IPT_BASE_CTL,
2186 .get_optmax = IPT_SO_GET_MAX+1,
2187 .get = do_ipt_get_ctl,
2188 #ifdef CONFIG_COMPAT
2189 .compat_get = compat_do_ipt_get_ctl,
2191 .owner = THIS_MODULE,
2194 static struct xt_match icmp_matchstruct __read_mostly = {
2196 .match = icmp_match,
2197 .matchsize = sizeof(struct ipt_icmp),
2198 .checkentry = icmp_checkentry,
2199 .proto = IPPROTO_ICMP,
2203 static int __net_init ip_tables_net_init(struct net *net)
2205 return xt_proto_init(net, AF_INET);
2208 static void __net_exit ip_tables_net_exit(struct net *net)
2210 xt_proto_fini(net, AF_INET);
2213 static struct pernet_operations ip_tables_net_ops = {
2214 .init = ip_tables_net_init,
2215 .exit = ip_tables_net_exit,
2218 static int __init ip_tables_init(void)
2222 ret = register_pernet_subsys(&ip_tables_net_ops);
2226 /* Noone else will be downing sem now, so we won't sleep */
2227 ret = xt_register_target(&ipt_standard_target);
2230 ret = xt_register_target(&ipt_error_target);
2233 ret = xt_register_match(&icmp_matchstruct);
2237 /* Register setsockopt */
2238 ret = nf_register_sockopt(&ipt_sockopts);
2242 printk(KERN_INFO "ip_tables: (C) 2000-2006 Netfilter Core Team\n");
2246 xt_unregister_match(&icmp_matchstruct);
2248 xt_unregister_target(&ipt_error_target);
2250 xt_unregister_target(&ipt_standard_target);
2252 unregister_pernet_subsys(&ip_tables_net_ops);
2257 static void __exit ip_tables_fini(void)
2259 nf_unregister_sockopt(&ipt_sockopts);
2261 xt_unregister_match(&icmp_matchstruct);
2262 xt_unregister_target(&ipt_error_target);
2263 xt_unregister_target(&ipt_standard_target);
2265 unregister_pernet_subsys(&ip_tables_net_ops);
2268 EXPORT_SYMBOL(ipt_register_table);
2269 EXPORT_SYMBOL(ipt_unregister_table);
2270 EXPORT_SYMBOL(ipt_do_table);
2271 module_init(ip_tables_init);
2272 module_exit(ip_tables_fini);