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 = try_then_request_module(xt_find_target(AF_INET,
707 "ipt_%s", t->u.user.name);
708 if (IS_ERR(target) || !target) {
709 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
710 ret = target ? PTR_ERR(target) : -ENOENT;
711 goto cleanup_matches;
713 t->u.kernel.target = target;
715 ret = check_target(e, net, name);
720 module_put(t->u.kernel.target->me);
722 xt_ematch_foreach(ematch, e) {
725 cleanup_match(ematch, net);
730 static bool check_underflow(const struct ipt_entry *e)
732 const struct ipt_entry_target *t;
733 unsigned int verdict;
735 if (!unconditional(&e->ip))
737 t = ipt_get_target_c(e);
738 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
740 verdict = ((struct ipt_standard_target *)t)->verdict;
741 verdict = -verdict - 1;
742 return verdict == NF_DROP || verdict == NF_ACCEPT;
746 check_entry_size_and_hooks(struct ipt_entry *e,
747 struct xt_table_info *newinfo,
748 const unsigned char *base,
749 const unsigned char *limit,
750 const unsigned int *hook_entries,
751 const unsigned int *underflows,
752 unsigned int valid_hooks)
756 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
757 (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
758 duprintf("Bad offset %p\n", e);
763 < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) {
764 duprintf("checking: element %p size %u\n",
769 /* Check hooks & underflows */
770 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
771 if (!(valid_hooks & (1 << h)))
773 if ((unsigned char *)e - base == hook_entries[h])
774 newinfo->hook_entry[h] = hook_entries[h];
775 if ((unsigned char *)e - base == underflows[h]) {
776 if (!check_underflow(e)) {
777 pr_err("Underflows must be unconditional and "
778 "use the STANDARD target with "
782 newinfo->underflow[h] = underflows[h];
786 /* Clear counters and comefrom */
787 e->counters = ((struct xt_counters) { 0, 0 });
793 cleanup_entry(struct ipt_entry *e, struct net *net)
795 struct xt_tgdtor_param par;
796 struct ipt_entry_target *t;
797 struct xt_entry_match *ematch;
799 /* Cleanup all matches */
800 xt_ematch_foreach(ematch, e)
801 cleanup_match(ematch, net);
802 t = ipt_get_target(e);
805 par.target = t->u.kernel.target;
806 par.targinfo = t->data;
807 par.family = NFPROTO_IPV4;
808 if (par.target->destroy != NULL)
809 par.target->destroy(&par);
810 module_put(par.target->me);
813 /* Checks and translates the user-supplied table segment (held in
816 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
817 const struct ipt_replace *repl)
819 struct ipt_entry *iter;
823 newinfo->size = repl->size;
824 newinfo->number = repl->num_entries;
826 /* Init all hooks to impossible value. */
827 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
828 newinfo->hook_entry[i] = 0xFFFFFFFF;
829 newinfo->underflow[i] = 0xFFFFFFFF;
832 duprintf("translate_table: size %u\n", newinfo->size);
834 /* Walk through entries, checking offsets. */
835 xt_entry_foreach(iter, entry0, newinfo->size) {
836 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
846 if (i != repl->num_entries) {
847 duprintf("translate_table: %u not %u entries\n",
848 i, repl->num_entries);
852 /* Check hooks all assigned */
853 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
854 /* Only hooks which are valid */
855 if (!(repl->valid_hooks & (1 << i)))
857 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
858 duprintf("Invalid hook entry %u %u\n",
859 i, repl->hook_entry[i]);
862 if (newinfo->underflow[i] == 0xFFFFFFFF) {
863 duprintf("Invalid underflow %u %u\n",
864 i, repl->underflow[i]);
869 if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
872 /* Finally, each sanity check must pass */
874 xt_entry_foreach(iter, entry0, newinfo->size) {
875 ret = find_check_entry(iter, net, repl->name, repl->size);
882 xt_entry_foreach(iter, entry0, newinfo->size) {
885 cleanup_entry(iter, net);
890 /* And one copy for every other CPU */
891 for_each_possible_cpu(i) {
892 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
893 memcpy(newinfo->entries[i], entry0, newinfo->size);
900 get_counters(const struct xt_table_info *t,
901 struct xt_counters counters[])
903 struct ipt_entry *iter;
908 /* Instead of clearing (by a previous call to memset())
909 * the counters and using adds, we set the counters
910 * with data used by 'current' CPU.
912 * Bottom half has to be disabled to prevent deadlock
913 * if new softirq were to run and call ipt_do_table
916 curcpu = smp_processor_id();
919 xt_entry_foreach(iter, t->entries[curcpu], t->size) {
920 SET_COUNTER(counters[i], iter->counters.bcnt,
921 iter->counters.pcnt);
925 for_each_possible_cpu(cpu) {
930 xt_entry_foreach(iter, t->entries[cpu], t->size) {
931 ADD_COUNTER(counters[i], iter->counters.bcnt,
932 iter->counters.pcnt);
933 ++i; /* macro does multi eval of i */
935 xt_info_wrunlock(cpu);
940 static struct xt_counters *alloc_counters(const struct xt_table *table)
942 unsigned int countersize;
943 struct xt_counters *counters;
944 const struct xt_table_info *private = table->private;
946 /* We need atomic snapshot of counters: rest doesn't change
947 (other than comefrom, which userspace doesn't care
949 countersize = sizeof(struct xt_counters) * private->number;
950 counters = vmalloc_node(countersize, numa_node_id());
952 if (counters == NULL)
953 return ERR_PTR(-ENOMEM);
955 get_counters(private, counters);
961 copy_entries_to_user(unsigned int total_size,
962 const struct xt_table *table,
963 void __user *userptr)
965 unsigned int off, num;
966 const struct ipt_entry *e;
967 struct xt_counters *counters;
968 const struct xt_table_info *private = table->private;
970 const void *loc_cpu_entry;
972 counters = alloc_counters(table);
973 if (IS_ERR(counters))
974 return PTR_ERR(counters);
976 /* choose the copy that is on our node/cpu, ...
977 * This choice is lazy (because current thread is
978 * allowed to migrate to another cpu)
980 loc_cpu_entry = private->entries[raw_smp_processor_id()];
981 if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
986 /* FIXME: use iterator macros --RR */
987 /* ... then go back and fix counters and names */
988 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
990 const struct ipt_entry_match *m;
991 const struct ipt_entry_target *t;
993 e = (struct ipt_entry *)(loc_cpu_entry + off);
994 if (copy_to_user(userptr + off
995 + offsetof(struct ipt_entry, counters),
997 sizeof(counters[num])) != 0) {
1002 for (i = sizeof(struct ipt_entry);
1003 i < e->target_offset;
1004 i += m->u.match_size) {
1007 if (copy_to_user(userptr + off + i
1008 + offsetof(struct ipt_entry_match,
1010 m->u.kernel.match->name,
1011 strlen(m->u.kernel.match->name)+1)
1018 t = ipt_get_target_c(e);
1019 if (copy_to_user(userptr + off + e->target_offset
1020 + offsetof(struct ipt_entry_target,
1022 t->u.kernel.target->name,
1023 strlen(t->u.kernel.target->name)+1) != 0) {
1034 #ifdef CONFIG_COMPAT
1035 static void compat_standard_from_user(void *dst, const void *src)
1037 int v = *(compat_int_t *)src;
1040 v += xt_compat_calc_jump(AF_INET, v);
1041 memcpy(dst, &v, sizeof(v));
1044 static int compat_standard_to_user(void __user *dst, const void *src)
1046 compat_int_t cv = *(int *)src;
1049 cv -= xt_compat_calc_jump(AF_INET, cv);
1050 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1053 static int compat_calc_entry(const struct ipt_entry *e,
1054 const struct xt_table_info *info,
1055 const void *base, struct xt_table_info *newinfo)
1057 const struct xt_entry_match *ematch;
1058 const struct ipt_entry_target *t;
1059 unsigned int entry_offset;
1062 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1063 entry_offset = (void *)e - base;
1064 xt_ematch_foreach(ematch, e)
1065 off += xt_compat_match_offset(ematch->u.kernel.match);
1066 t = ipt_get_target_c(e);
1067 off += xt_compat_target_offset(t->u.kernel.target);
1068 newinfo->size -= off;
1069 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1073 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1074 if (info->hook_entry[i] &&
1075 (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1076 newinfo->hook_entry[i] -= off;
1077 if (info->underflow[i] &&
1078 (e < (struct ipt_entry *)(base + info->underflow[i])))
1079 newinfo->underflow[i] -= off;
1084 static int compat_table_info(const struct xt_table_info *info,
1085 struct xt_table_info *newinfo)
1087 struct ipt_entry *iter;
1088 void *loc_cpu_entry;
1091 if (!newinfo || !info)
1094 /* we dont care about newinfo->entries[] */
1095 memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1096 newinfo->initial_entries = 0;
1097 loc_cpu_entry = info->entries[raw_smp_processor_id()];
1098 xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1099 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1107 static int get_info(struct net *net, void __user *user,
1108 const int *len, int compat)
1110 char name[IPT_TABLE_MAXNAMELEN];
1114 if (*len != sizeof(struct ipt_getinfo)) {
1115 duprintf("length %u != %zu\n", *len,
1116 sizeof(struct ipt_getinfo));
1120 if (copy_from_user(name, user, sizeof(name)) != 0)
1123 name[IPT_TABLE_MAXNAMELEN-1] = '\0';
1124 #ifdef CONFIG_COMPAT
1126 xt_compat_lock(AF_INET);
1128 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1129 "iptable_%s", name);
1130 if (t && !IS_ERR(t)) {
1131 struct ipt_getinfo info;
1132 const struct xt_table_info *private = t->private;
1133 #ifdef CONFIG_COMPAT
1134 struct xt_table_info tmp;
1137 ret = compat_table_info(private, &tmp);
1138 xt_compat_flush_offsets(AF_INET);
1142 info.valid_hooks = t->valid_hooks;
1143 memcpy(info.hook_entry, private->hook_entry,
1144 sizeof(info.hook_entry));
1145 memcpy(info.underflow, private->underflow,
1146 sizeof(info.underflow));
1147 info.num_entries = private->number;
1148 info.size = private->size;
1149 strcpy(info.name, name);
1151 if (copy_to_user(user, &info, *len) != 0)
1159 ret = t ? PTR_ERR(t) : -ENOENT;
1160 #ifdef CONFIG_COMPAT
1162 xt_compat_unlock(AF_INET);
1168 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1172 struct ipt_get_entries get;
1175 if (*len < sizeof(get)) {
1176 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1179 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1181 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1182 duprintf("get_entries: %u != %zu\n",
1183 *len, sizeof(get) + get.size);
1187 t = xt_find_table_lock(net, AF_INET, get.name);
1188 if (t && !IS_ERR(t)) {
1189 const struct xt_table_info *private = t->private;
1190 duprintf("t->private->number = %u\n", private->number);
1191 if (get.size == private->size)
1192 ret = copy_entries_to_user(private->size,
1193 t, uptr->entrytable);
1195 duprintf("get_entries: I've got %u not %u!\n",
1196 private->size, get.size);
1202 ret = t ? PTR_ERR(t) : -ENOENT;
1208 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1209 struct xt_table_info *newinfo, unsigned int num_counters,
1210 void __user *counters_ptr)
1214 struct xt_table_info *oldinfo;
1215 struct xt_counters *counters;
1216 void *loc_cpu_old_entry;
1217 struct ipt_entry *iter;
1220 counters = vmalloc(num_counters * sizeof(struct xt_counters));
1226 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1227 "iptable_%s", name);
1228 if (!t || IS_ERR(t)) {
1229 ret = t ? PTR_ERR(t) : -ENOENT;
1230 goto free_newinfo_counters_untrans;
1234 if (valid_hooks != t->valid_hooks) {
1235 duprintf("Valid hook crap: %08X vs %08X\n",
1236 valid_hooks, t->valid_hooks);
1241 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1245 /* Update module usage count based on number of rules */
1246 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1247 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1248 if ((oldinfo->number > oldinfo->initial_entries) ||
1249 (newinfo->number <= oldinfo->initial_entries))
1251 if ((oldinfo->number > oldinfo->initial_entries) &&
1252 (newinfo->number <= oldinfo->initial_entries))
1255 /* Get the old counters, and synchronize with replace */
1256 get_counters(oldinfo, counters);
1258 /* Decrease module usage counts and free resource */
1259 loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1260 xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1261 cleanup_entry(iter, net);
1263 xt_free_table_info(oldinfo);
1264 if (copy_to_user(counters_ptr, counters,
1265 sizeof(struct xt_counters) * num_counters) != 0)
1274 free_newinfo_counters_untrans:
1281 do_replace(struct net *net, const void __user *user, unsigned int len)
1284 struct ipt_replace tmp;
1285 struct xt_table_info *newinfo;
1286 void *loc_cpu_entry;
1287 struct ipt_entry *iter;
1289 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1292 /* overflow check */
1293 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1296 newinfo = xt_alloc_table_info(tmp.size);
1300 /* choose the copy that is on our node/cpu */
1301 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1302 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1308 ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1312 duprintf("Translated table\n");
1314 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1315 tmp.num_counters, tmp.counters);
1317 goto free_newinfo_untrans;
1320 free_newinfo_untrans:
1321 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1322 cleanup_entry(iter, net);
1324 xt_free_table_info(newinfo);
1329 do_add_counters(struct net *net, const void __user *user,
1330 unsigned int len, int compat)
1332 unsigned int i, curcpu;
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 struct ipt_entry *iter;
1344 #ifdef CONFIG_COMPAT
1345 struct compat_xt_counters_info compat_tmp;
1349 size = sizeof(struct compat_xt_counters_info);
1354 size = sizeof(struct xt_counters_info);
1357 if (copy_from_user(ptmp, user, size) != 0)
1360 #ifdef CONFIG_COMPAT
1362 num_counters = compat_tmp.num_counters;
1363 name = compat_tmp.name;
1367 num_counters = tmp.num_counters;
1371 if (len != size + num_counters * sizeof(struct xt_counters))
1374 paddc = vmalloc_node(len - size, numa_node_id());
1378 if (copy_from_user(paddc, user + size, len - size) != 0) {
1383 t = xt_find_table_lock(net, AF_INET, name);
1384 if (!t || IS_ERR(t)) {
1385 ret = t ? PTR_ERR(t) : -ENOENT;
1390 private = t->private;
1391 if (private->number != num_counters) {
1393 goto unlock_up_free;
1397 /* Choose the copy that is on our node */
1398 curcpu = smp_processor_id();
1399 loc_cpu_entry = private->entries[curcpu];
1400 xt_info_wrlock(curcpu);
1401 xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1402 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1405 xt_info_wrunlock(curcpu);
1416 #ifdef CONFIG_COMPAT
1417 struct compat_ipt_replace {
1418 char name[IPT_TABLE_MAXNAMELEN];
1422 u32 hook_entry[NF_INET_NUMHOOKS];
1423 u32 underflow[NF_INET_NUMHOOKS];
1425 compat_uptr_t counters; /* struct ipt_counters * */
1426 struct compat_ipt_entry entries[0];
1430 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1431 unsigned int *size, struct xt_counters *counters,
1434 struct ipt_entry_target *t;
1435 struct compat_ipt_entry __user *ce;
1436 u_int16_t target_offset, next_offset;
1437 compat_uint_t origsize;
1438 const struct xt_entry_match *ematch;
1442 ce = (struct compat_ipt_entry __user *)*dstptr;
1443 if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1444 copy_to_user(&ce->counters, &counters[i],
1445 sizeof(counters[i])) != 0)
1448 *dstptr += sizeof(struct compat_ipt_entry);
1449 *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1451 xt_ematch_foreach(ematch, e) {
1452 ret = xt_compat_match_to_user(ematch, dstptr, size);
1456 target_offset = e->target_offset - (origsize - *size);
1457 t = ipt_get_target(e);
1458 ret = xt_compat_target_to_user(t, dstptr, size);
1461 next_offset = e->next_offset - (origsize - *size);
1462 if (put_user(target_offset, &ce->target_offset) != 0 ||
1463 put_user(next_offset, &ce->next_offset) != 0)
1469 compat_find_calc_match(struct ipt_entry_match *m,
1471 const struct ipt_ip *ip,
1472 unsigned int hookmask,
1475 struct xt_match *match;
1477 match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
1478 m->u.user.revision),
1479 "ipt_%s", m->u.user.name);
1480 if (IS_ERR(match) || !match) {
1481 duprintf("compat_check_calc_match: `%s' not found\n",
1483 return match ? PTR_ERR(match) : -ENOENT;
1485 m->u.kernel.match = match;
1486 *size += xt_compat_match_offset(match);
1490 static void compat_release_entry(struct compat_ipt_entry *e)
1492 struct ipt_entry_target *t;
1493 struct xt_entry_match *ematch;
1495 /* Cleanup all matches */
1496 xt_ematch_foreach(ematch, e)
1497 module_put(ematch->u.kernel.match->me);
1498 t = compat_ipt_get_target(e);
1499 module_put(t->u.kernel.target->me);
1503 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1504 struct xt_table_info *newinfo,
1506 const unsigned char *base,
1507 const unsigned char *limit,
1508 const unsigned int *hook_entries,
1509 const unsigned int *underflows,
1512 struct xt_entry_match *ematch;
1513 struct ipt_entry_target *t;
1514 struct xt_target *target;
1515 unsigned int entry_offset;
1519 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1520 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1521 (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1522 duprintf("Bad offset %p, limit = %p\n", e, limit);
1526 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1527 sizeof(struct compat_xt_entry_target)) {
1528 duprintf("checking: element %p size %u\n",
1533 /* For purposes of check_entry casting the compat entry is fine */
1534 ret = check_entry((struct ipt_entry *)e, name);
1538 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1539 entry_offset = (void *)e - (void *)base;
1541 xt_ematch_foreach(ematch, e) {
1542 ret = compat_find_calc_match(ematch, name,
1543 &e->ip, e->comefrom, &off);
1545 goto release_matches;
1549 t = compat_ipt_get_target(e);
1550 target = try_then_request_module(xt_find_target(AF_INET,
1552 t->u.user.revision),
1553 "ipt_%s", t->u.user.name);
1554 if (IS_ERR(target) || !target) {
1555 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1557 ret = target ? PTR_ERR(target) : -ENOENT;
1558 goto release_matches;
1560 t->u.kernel.target = target;
1562 off += xt_compat_target_offset(target);
1564 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1568 /* Check hooks & underflows */
1569 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1570 if ((unsigned char *)e - base == hook_entries[h])
1571 newinfo->hook_entry[h] = hook_entries[h];
1572 if ((unsigned char *)e - base == underflows[h])
1573 newinfo->underflow[h] = underflows[h];
1576 /* Clear counters and comefrom */
1577 memset(&e->counters, 0, sizeof(e->counters));
1582 module_put(t->u.kernel.target->me);
1584 xt_ematch_foreach(ematch, e) {
1587 module_put(ematch->u.kernel.match->me);
1593 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1594 unsigned int *size, const char *name,
1595 struct xt_table_info *newinfo, unsigned char *base)
1597 struct ipt_entry_target *t;
1598 struct xt_target *target;
1599 struct ipt_entry *de;
1600 unsigned int origsize;
1602 struct xt_entry_match *ematch;
1606 de = (struct ipt_entry *)*dstptr;
1607 memcpy(de, e, sizeof(struct ipt_entry));
1608 memcpy(&de->counters, &e->counters, sizeof(e->counters));
1610 *dstptr += sizeof(struct ipt_entry);
1611 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1613 xt_ematch_foreach(ematch, e) {
1614 ret = xt_compat_match_from_user(ematch, dstptr, size);
1618 de->target_offset = e->target_offset - (origsize - *size);
1619 t = compat_ipt_get_target(e);
1620 target = t->u.kernel.target;
1621 xt_compat_target_from_user(t, dstptr, size);
1623 de->next_offset = e->next_offset - (origsize - *size);
1624 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1625 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1626 newinfo->hook_entry[h] -= origsize - *size;
1627 if ((unsigned char *)de - base < newinfo->underflow[h])
1628 newinfo->underflow[h] -= origsize - *size;
1634 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1636 struct xt_entry_match *ematch;
1637 struct xt_mtchk_param mtpar;
1644 mtpar.entryinfo = &e->ip;
1645 mtpar.hook_mask = e->comefrom;
1646 mtpar.family = NFPROTO_IPV4;
1647 xt_ematch_foreach(ematch, e) {
1648 ret = check_match(ematch, &mtpar);
1650 goto cleanup_matches;
1654 ret = check_target(e, net, name);
1656 goto cleanup_matches;
1660 xt_ematch_foreach(ematch, e) {
1663 cleanup_match(ematch, net);
1669 translate_compat_table(struct net *net,
1671 unsigned int valid_hooks,
1672 struct xt_table_info **pinfo,
1674 unsigned int total_size,
1675 unsigned int number,
1676 unsigned int *hook_entries,
1677 unsigned int *underflows)
1680 struct xt_table_info *newinfo, *info;
1681 void *pos, *entry0, *entry1;
1682 struct compat_ipt_entry *iter0;
1683 struct ipt_entry *iter1;
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 xt_entry_foreach(iter0, entry0, total_size) {
1703 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1705 entry0 + total_size,
1716 duprintf("translate_compat_table: %u not %u entries\n",
1721 /* Check hooks all assigned */
1722 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1723 /* Only hooks which are valid */
1724 if (!(valid_hooks & (1 << i)))
1726 if (info->hook_entry[i] == 0xFFFFFFFF) {
1727 duprintf("Invalid hook entry %u %u\n",
1728 i, hook_entries[i]);
1731 if (info->underflow[i] == 0xFFFFFFFF) {
1732 duprintf("Invalid underflow %u %u\n",
1739 newinfo = xt_alloc_table_info(size);
1743 newinfo->number = number;
1744 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1745 newinfo->hook_entry[i] = info->hook_entry[i];
1746 newinfo->underflow[i] = info->underflow[i];
1748 entry1 = newinfo->entries[raw_smp_processor_id()];
1751 xt_entry_foreach(iter0, entry0, total_size) {
1752 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1753 name, newinfo, entry1);
1757 xt_compat_flush_offsets(AF_INET);
1758 xt_compat_unlock(AF_INET);
1763 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1767 xt_entry_foreach(iter1, entry1, newinfo->size) {
1768 ret = compat_check_entry(iter1, net, name);
1775 * The first i matches need cleanup_entry (calls ->destroy)
1776 * because they had called ->check already. The other j-i
1777 * entries need only release.
1781 xt_entry_foreach(iter0, entry0, newinfo->size) {
1786 compat_release_entry(iter0);
1788 xt_entry_foreach(iter1, entry1, newinfo->size) {
1791 cleanup_entry(iter1, net);
1793 xt_free_table_info(newinfo);
1797 /* And one copy for every other CPU */
1798 for_each_possible_cpu(i)
1799 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1800 memcpy(newinfo->entries[i], entry1, newinfo->size);
1804 xt_free_table_info(info);
1808 xt_free_table_info(newinfo);
1810 xt_entry_foreach(iter0, entry0, total_size) {
1813 compat_release_entry(iter0);
1817 xt_compat_flush_offsets(AF_INET);
1818 xt_compat_unlock(AF_INET);
1823 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1826 struct compat_ipt_replace tmp;
1827 struct xt_table_info *newinfo;
1828 void *loc_cpu_entry;
1829 struct ipt_entry *iter;
1831 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1834 /* overflow check */
1835 if (tmp.size >= INT_MAX / num_possible_cpus())
1837 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1840 newinfo = xt_alloc_table_info(tmp.size);
1844 /* choose the copy that is on our node/cpu */
1845 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1846 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1852 ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1853 &newinfo, &loc_cpu_entry, tmp.size,
1854 tmp.num_entries, tmp.hook_entry,
1859 duprintf("compat_do_replace: Translated table\n");
1861 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1862 tmp.num_counters, compat_ptr(tmp.counters));
1864 goto free_newinfo_untrans;
1867 free_newinfo_untrans:
1868 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1869 cleanup_entry(iter, net);
1871 xt_free_table_info(newinfo);
1876 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1881 if (!capable(CAP_NET_ADMIN))
1885 case IPT_SO_SET_REPLACE:
1886 ret = compat_do_replace(sock_net(sk), user, len);
1889 case IPT_SO_SET_ADD_COUNTERS:
1890 ret = do_add_counters(sock_net(sk), user, len, 1);
1894 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1901 struct compat_ipt_get_entries {
1902 char name[IPT_TABLE_MAXNAMELEN];
1904 struct compat_ipt_entry entrytable[0];
1908 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1909 void __user *userptr)
1911 struct xt_counters *counters;
1912 const struct xt_table_info *private = table->private;
1916 const void *loc_cpu_entry;
1918 struct ipt_entry *iter;
1920 counters = alloc_counters(table);
1921 if (IS_ERR(counters))
1922 return PTR_ERR(counters);
1924 /* choose the copy that is on our node/cpu, ...
1925 * This choice is lazy (because current thread is
1926 * allowed to migrate to another cpu)
1928 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1931 xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1932 ret = compat_copy_entry_to_user(iter, &pos,
1933 &size, counters, i++);
1943 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1947 struct compat_ipt_get_entries get;
1950 if (*len < sizeof(get)) {
1951 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1955 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1958 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1959 duprintf("compat_get_entries: %u != %zu\n",
1960 *len, sizeof(get) + get.size);
1964 xt_compat_lock(AF_INET);
1965 t = xt_find_table_lock(net, AF_INET, get.name);
1966 if (t && !IS_ERR(t)) {
1967 const struct xt_table_info *private = t->private;
1968 struct xt_table_info info;
1969 duprintf("t->private->number = %u\n", private->number);
1970 ret = compat_table_info(private, &info);
1971 if (!ret && get.size == info.size) {
1972 ret = compat_copy_entries_to_user(private->size,
1973 t, uptr->entrytable);
1975 duprintf("compat_get_entries: I've got %u not %u!\n",
1976 private->size, get.size);
1979 xt_compat_flush_offsets(AF_INET);
1983 ret = t ? PTR_ERR(t) : -ENOENT;
1985 xt_compat_unlock(AF_INET);
1989 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1992 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1996 if (!capable(CAP_NET_ADMIN))
2000 case IPT_SO_GET_INFO:
2001 ret = get_info(sock_net(sk), user, len, 1);
2003 case IPT_SO_GET_ENTRIES:
2004 ret = compat_get_entries(sock_net(sk), user, len);
2007 ret = do_ipt_get_ctl(sk, cmd, user, len);
2014 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2018 if (!capable(CAP_NET_ADMIN))
2022 case IPT_SO_SET_REPLACE:
2023 ret = do_replace(sock_net(sk), user, len);
2026 case IPT_SO_SET_ADD_COUNTERS:
2027 ret = do_add_counters(sock_net(sk), user, len, 0);
2031 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
2039 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2043 if (!capable(CAP_NET_ADMIN))
2047 case IPT_SO_GET_INFO:
2048 ret = get_info(sock_net(sk), user, len, 0);
2051 case IPT_SO_GET_ENTRIES:
2052 ret = get_entries(sock_net(sk), user, len);
2055 case IPT_SO_GET_REVISION_MATCH:
2056 case IPT_SO_GET_REVISION_TARGET: {
2057 struct ipt_get_revision rev;
2060 if (*len != sizeof(rev)) {
2064 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2069 if (cmd == IPT_SO_GET_REVISION_TARGET)
2074 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2077 "ipt_%s", rev.name);
2082 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2089 struct xt_table *ipt_register_table(struct net *net,
2090 const struct xt_table *table,
2091 const struct ipt_replace *repl)
2094 struct xt_table_info *newinfo;
2095 struct xt_table_info bootstrap
2096 = { 0, 0, 0, { 0 }, { 0 }, { } };
2097 void *loc_cpu_entry;
2098 struct xt_table *new_table;
2100 newinfo = xt_alloc_table_info(repl->size);
2106 /* choose the copy on our node/cpu, but dont care about preemption */
2107 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2108 memcpy(loc_cpu_entry, repl->entries, repl->size);
2110 ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2114 new_table = xt_register_table(net, table, &bootstrap, newinfo);
2115 if (IS_ERR(new_table)) {
2116 ret = PTR_ERR(new_table);
2123 xt_free_table_info(newinfo);
2125 return ERR_PTR(ret);
2128 void ipt_unregister_table(struct net *net, struct xt_table *table)
2130 struct xt_table_info *private;
2131 void *loc_cpu_entry;
2132 struct module *table_owner = table->me;
2133 struct ipt_entry *iter;
2135 private = xt_unregister_table(table);
2137 /* Decrease module usage counts and free resources */
2138 loc_cpu_entry = private->entries[raw_smp_processor_id()];
2139 xt_entry_foreach(iter, loc_cpu_entry, private->size)
2140 cleanup_entry(iter, net);
2141 if (private->number > private->initial_entries)
2142 module_put(table_owner);
2143 xt_free_table_info(private);
2146 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2148 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2149 u_int8_t type, u_int8_t code,
2152 return ((test_type == 0xFF) ||
2153 (type == test_type && code >= min_code && code <= max_code))
2158 icmp_match(const struct sk_buff *skb, const struct xt_match_param *par)
2160 const struct icmphdr *ic;
2161 struct icmphdr _icmph;
2162 const struct ipt_icmp *icmpinfo = par->matchinfo;
2164 /* Must not be a fragment. */
2165 if (par->fragoff != 0)
2168 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2170 /* We've been asked to examine this packet, and we
2171 * can't. Hence, no choice but to drop.
2173 duprintf("Dropping evil ICMP tinygram.\n");
2174 *par->hotdrop = true;
2178 return icmp_type_code_match(icmpinfo->type,
2182 !!(icmpinfo->invflags&IPT_ICMP_INV));
2185 static bool icmp_checkentry(const struct xt_mtchk_param *par)
2187 const struct ipt_icmp *icmpinfo = par->matchinfo;
2189 /* Must specify no unknown invflags */
2190 return !(icmpinfo->invflags & ~IPT_ICMP_INV);
2193 /* The built-in targets: standard (NULL) and error. */
2194 static struct xt_target ipt_standard_target __read_mostly = {
2195 .name = IPT_STANDARD_TARGET,
2196 .targetsize = sizeof(int),
2197 .family = NFPROTO_IPV4,
2198 #ifdef CONFIG_COMPAT
2199 .compatsize = sizeof(compat_int_t),
2200 .compat_from_user = compat_standard_from_user,
2201 .compat_to_user = compat_standard_to_user,
2205 static struct xt_target ipt_error_target __read_mostly = {
2206 .name = IPT_ERROR_TARGET,
2207 .target = ipt_error,
2208 .targetsize = IPT_FUNCTION_MAXNAMELEN,
2209 .family = NFPROTO_IPV4,
2212 static struct nf_sockopt_ops ipt_sockopts = {
2214 .set_optmin = IPT_BASE_CTL,
2215 .set_optmax = IPT_SO_SET_MAX+1,
2216 .set = do_ipt_set_ctl,
2217 #ifdef CONFIG_COMPAT
2218 .compat_set = compat_do_ipt_set_ctl,
2220 .get_optmin = IPT_BASE_CTL,
2221 .get_optmax = IPT_SO_GET_MAX+1,
2222 .get = do_ipt_get_ctl,
2223 #ifdef CONFIG_COMPAT
2224 .compat_get = compat_do_ipt_get_ctl,
2226 .owner = THIS_MODULE,
2229 static struct xt_match icmp_matchstruct __read_mostly = {
2231 .match = icmp_match,
2232 .matchsize = sizeof(struct ipt_icmp),
2233 .checkentry = icmp_checkentry,
2234 .proto = IPPROTO_ICMP,
2235 .family = NFPROTO_IPV4,
2238 static int __net_init ip_tables_net_init(struct net *net)
2240 return xt_proto_init(net, NFPROTO_IPV4);
2243 static void __net_exit ip_tables_net_exit(struct net *net)
2245 xt_proto_fini(net, NFPROTO_IPV4);
2248 static struct pernet_operations ip_tables_net_ops = {
2249 .init = ip_tables_net_init,
2250 .exit = ip_tables_net_exit,
2253 static int __init ip_tables_init(void)
2257 ret = register_pernet_subsys(&ip_tables_net_ops);
2261 /* Noone else will be downing sem now, so we won't sleep */
2262 ret = xt_register_target(&ipt_standard_target);
2265 ret = xt_register_target(&ipt_error_target);
2268 ret = xt_register_match(&icmp_matchstruct);
2272 /* Register setsockopt */
2273 ret = nf_register_sockopt(&ipt_sockopts);
2277 pr_info("(C) 2000-2006 Netfilter Core Team\n");
2281 xt_unregister_match(&icmp_matchstruct);
2283 xt_unregister_target(&ipt_error_target);
2285 xt_unregister_target(&ipt_standard_target);
2287 unregister_pernet_subsys(&ip_tables_net_ops);
2292 static void __exit ip_tables_fini(void)
2294 nf_unregister_sockopt(&ipt_sockopts);
2296 xt_unregister_match(&icmp_matchstruct);
2297 xt_unregister_target(&ipt_error_target);
2298 xt_unregister_target(&ipt_standard_target);
2300 unregister_pernet_subsys(&ip_tables_net_ops);
2303 EXPORT_SYMBOL(ipt_register_table);
2304 EXPORT_SYMBOL(ipt_unregister_table);
2305 EXPORT_SYMBOL(ipt_do_table);
2306 module_init(ip_tables_init);
2307 module_exit(ip_tables_fini);