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>
6 * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 as
10 * published by the Free Software Foundation.
12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
13 #include <linux/cache.h>
14 #include <linux/capability.h>
15 #include <linux/skbuff.h>
16 #include <linux/kmod.h>
17 #include <linux/vmalloc.h>
18 #include <linux/netdevice.h>
19 #include <linux/module.h>
20 #include <linux/icmp.h>
22 #include <net/compat.h>
23 #include <asm/uaccess.h>
24 #include <linux/mutex.h>
25 #include <linux/proc_fs.h>
26 #include <linux/err.h>
27 #include <linux/cpumask.h>
29 #include <linux/netfilter/x_tables.h>
30 #include <linux/netfilter_ipv4/ip_tables.h>
31 #include <net/netfilter/nf_log.h>
32 #include "../../netfilter/xt_repldata.h"
34 MODULE_LICENSE("GPL");
35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
36 MODULE_DESCRIPTION("IPv4 packet filter");
38 /*#define DEBUG_IP_FIREWALL*/
39 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
40 /*#define DEBUG_IP_FIREWALL_USER*/
42 #ifdef DEBUG_IP_FIREWALL
43 #define dprintf(format, args...) pr_info(format , ## args)
45 #define dprintf(format, args...)
48 #ifdef DEBUG_IP_FIREWALL_USER
49 #define duprintf(format, args...) pr_info(format , ## args)
51 #define duprintf(format, args...)
54 #ifdef CONFIG_NETFILTER_DEBUG
55 #define IP_NF_ASSERT(x) WARN_ON(!(x))
57 #define IP_NF_ASSERT(x)
61 /* All the better to debug you with... */
66 void *ipt_alloc_initial_table(const struct xt_table *info)
68 return xt_alloc_initial_table(ipt, IPT);
70 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
72 /* Returns whether matches rule or not. */
73 /* Performance critical - called for every packet */
75 ip_packet_match(const struct iphdr *ip,
78 const struct ipt_ip *ipinfo,
83 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
85 if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
87 FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
89 dprintf("Source or dest mismatch.\n");
91 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
92 &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
93 ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
94 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
95 &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
96 ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
100 ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
102 if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
103 dprintf("VIA in mismatch (%s vs %s).%s\n",
104 indev, ipinfo->iniface,
105 ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
109 ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
111 if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
112 dprintf("VIA out mismatch (%s vs %s).%s\n",
113 outdev, ipinfo->outiface,
114 ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
118 /* Check specific protocol */
120 FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
121 dprintf("Packet protocol %hi does not match %hi.%s\n",
122 ip->protocol, ipinfo->proto,
123 ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
127 /* If we have a fragment rule but the packet is not a fragment
128 * then we return zero */
129 if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
130 dprintf("Fragment rule but not fragment.%s\n",
131 ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
139 ip_checkentry(const struct ipt_ip *ip)
141 if (ip->flags & ~IPT_F_MASK) {
142 duprintf("Unknown flag bits set: %08X\n",
143 ip->flags & ~IPT_F_MASK);
146 if (ip->invflags & ~IPT_INV_MASK) {
147 duprintf("Unknown invflag bits set: %08X\n",
148 ip->invflags & ~IPT_INV_MASK);
155 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
157 net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
162 /* Performance critical */
163 static inline struct ipt_entry *
164 get_entry(const void *base, unsigned int offset)
166 return (struct ipt_entry *)(base + offset);
169 /* All zeroes == unconditional rule. */
170 /* Mildly perf critical (only if packet tracing is on) */
171 static inline bool unconditional(const struct ipt_ip *ip)
173 static const struct ipt_ip uncond;
175 return memcmp(ip, &uncond, sizeof(uncond)) == 0;
179 /* for const-correctness */
180 static inline const struct xt_entry_target *
181 ipt_get_target_c(const struct ipt_entry *e)
183 return ipt_get_target((struct ipt_entry *)e);
186 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
187 static const char *const hooknames[] = {
188 [NF_INET_PRE_ROUTING] = "PREROUTING",
189 [NF_INET_LOCAL_IN] = "INPUT",
190 [NF_INET_FORWARD] = "FORWARD",
191 [NF_INET_LOCAL_OUT] = "OUTPUT",
192 [NF_INET_POST_ROUTING] = "POSTROUTING",
195 enum nf_ip_trace_comments {
196 NF_IP_TRACE_COMMENT_RULE,
197 NF_IP_TRACE_COMMENT_RETURN,
198 NF_IP_TRACE_COMMENT_POLICY,
201 static const char *const comments[] = {
202 [NF_IP_TRACE_COMMENT_RULE] = "rule",
203 [NF_IP_TRACE_COMMENT_RETURN] = "return",
204 [NF_IP_TRACE_COMMENT_POLICY] = "policy",
207 static struct nf_loginfo trace_loginfo = {
208 .type = NF_LOG_TYPE_LOG,
212 .logflags = NF_LOG_MASK,
217 /* Mildly perf critical (only if packet tracing is on) */
219 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
220 const char *hookname, const char **chainname,
221 const char **comment, unsigned int *rulenum)
223 const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
225 if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
226 /* Head of user chain: ERROR target with chainname */
227 *chainname = t->target.data;
232 if (s->target_offset == sizeof(struct ipt_entry) &&
233 strcmp(t->target.u.kernel.target->name,
234 XT_STANDARD_TARGET) == 0 &&
236 unconditional(&s->ip)) {
237 /* Tail of chains: STANDARD target (return/policy) */
238 *comment = *chainname == hookname
239 ? comments[NF_IP_TRACE_COMMENT_POLICY]
240 : comments[NF_IP_TRACE_COMMENT_RETURN];
249 static void trace_packet(const struct sk_buff *skb,
251 const struct net_device *in,
252 const struct net_device *out,
253 const char *tablename,
254 const struct xt_table_info *private,
255 const struct ipt_entry *e)
257 const struct ipt_entry *root;
258 const char *hookname, *chainname, *comment;
259 const struct ipt_entry *iter;
260 unsigned int rulenum = 0;
261 struct net *net = dev_net(in ? in : out);
263 root = get_entry(private->entries, private->hook_entry[hook]);
265 hookname = chainname = hooknames[hook];
266 comment = comments[NF_IP_TRACE_COMMENT_RULE];
268 xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
269 if (get_chainname_rulenum(iter, e, hookname,
270 &chainname, &comment, &rulenum) != 0)
273 nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
274 "TRACE: %s:%s:%s:%u ",
275 tablename, chainname, comment, rulenum);
280 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
282 return (void *)entry + entry->next_offset;
285 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
287 ipt_do_table(struct sk_buff *skb,
289 const struct nf_hook_state *state,
290 struct xt_table *table)
292 static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
293 const struct iphdr *ip;
294 /* Initializing verdict to NF_DROP keeps gcc happy. */
295 unsigned int verdict = NF_DROP;
296 const char *indev, *outdev;
297 const void *table_base;
298 struct ipt_entry *e, **jumpstack;
299 unsigned int stackidx, cpu;
300 const struct xt_table_info *private;
301 struct xt_action_param acpar;
307 indev = state->in ? state->in->name : nulldevname;
308 outdev = state->out ? state->out->name : nulldevname;
309 /* We handle fragments by dealing with the first fragment as
310 * if it was a normal packet. All other fragments are treated
311 * normally, except that they will NEVER match rules that ask
312 * things we don't know, ie. tcp syn flag or ports). If the
313 * rule is also a fragment-specific rule, non-fragments won't
315 acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
316 acpar.thoff = ip_hdrlen(skb);
317 acpar.hotdrop = false;
318 acpar.in = state->in;
319 acpar.out = state->out;
320 acpar.family = NFPROTO_IPV4;
321 acpar.hooknum = hook;
323 IP_NF_ASSERT(table->valid_hooks & (1 << hook));
325 addend = xt_write_recseq_begin();
326 private = table->private;
327 cpu = smp_processor_id();
329 * Ensure we load private-> members after we've fetched the base
332 smp_read_barrier_depends();
333 table_base = private->entries;
334 jumpstack = (struct ipt_entry **)private->jumpstack[cpu];
336 /* Switch to alternate jumpstack if we're being invoked via TEE.
337 * TEE issues XT_CONTINUE verdict on original skb so we must not
338 * clobber the jumpstack.
340 * For recursion via REJECT or SYNPROXY the stack will be clobbered
341 * but it is no problem since absolute verdict is issued by these.
343 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
345 e = get_entry(table_base, private->hook_entry[hook]);
347 pr_debug("Entering %s(hook %u), UF %p\n",
349 get_entry(table_base, private->underflow[hook]));
352 const struct xt_entry_target *t;
353 const struct xt_entry_match *ematch;
354 struct xt_counters *counter;
357 if (!ip_packet_match(ip, indev, outdev,
358 &e->ip, acpar.fragoff)) {
360 e = ipt_next_entry(e);
364 xt_ematch_foreach(ematch, e) {
365 acpar.match = ematch->u.kernel.match;
366 acpar.matchinfo = ematch->data;
367 if (!acpar.match->match(skb, &acpar))
371 counter = xt_get_this_cpu_counter(&e->counters);
372 ADD_COUNTER(*counter, skb->len, 1);
374 t = ipt_get_target(e);
375 IP_NF_ASSERT(t->u.kernel.target);
377 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
378 /* The packet is traced: log it */
379 if (unlikely(skb->nf_trace))
380 trace_packet(skb, hook, state->in, state->out,
381 table->name, private, e);
383 /* Standard target? */
384 if (!t->u.kernel.target->target) {
387 v = ((struct xt_standard_target *)t)->verdict;
389 /* Pop from stack? */
390 if (v != XT_RETURN) {
391 verdict = (unsigned int)(-v) - 1;
395 e = get_entry(table_base,
396 private->underflow[hook]);
397 pr_debug("Underflow (this is normal) "
400 e = jumpstack[--stackidx];
401 pr_debug("Pulled %p out from pos %u\n",
403 e = ipt_next_entry(e);
407 if (table_base + v != ipt_next_entry(e) &&
408 !(e->ip.flags & IPT_F_GOTO)) {
409 jumpstack[stackidx++] = e;
410 pr_debug("Pushed %p into pos %u\n",
414 e = get_entry(table_base, v);
418 acpar.target = t->u.kernel.target;
419 acpar.targinfo = t->data;
421 verdict = t->u.kernel.target->target(skb, &acpar);
422 /* Target might have changed stuff. */
424 if (verdict == XT_CONTINUE)
425 e = ipt_next_entry(e);
429 } while (!acpar.hotdrop);
430 pr_debug("Exiting %s; sp at %u\n", __func__, stackidx);
432 xt_write_recseq_end(addend);
435 #ifdef DEBUG_ALLOW_ALL
444 /* Figures out from what hook each rule can be called: returns 0 if
445 * there are loops. Puts hook bitmask in comefrom.
447 * Keeps track of largest call depth seen and stores it in newinfo->stacksize.
450 mark_source_chains(struct xt_table_info *newinfo,
451 unsigned int valid_hooks, void *entry0)
453 unsigned int calldepth, max_calldepth = 0;
456 /* No recursion; use packet counter to save back ptrs (reset
457 to 0 as we leave), and comefrom to save source hook bitmask */
458 for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
459 unsigned int pos = newinfo->hook_entry[hook];
460 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
462 if (!(valid_hooks & (1 << hook)))
465 /* Set initial back pointer. */
466 e->counters.pcnt = pos;
470 const struct xt_standard_target *t
471 = (void *)ipt_get_target_c(e);
472 int visited = e->comefrom & (1 << hook);
474 if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
475 pr_err("iptables: loop hook %u pos %u %08X.\n",
476 hook, pos, e->comefrom);
479 e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
481 /* Unconditional return/END. */
482 if ((e->target_offset == sizeof(struct ipt_entry) &&
483 (strcmp(t->target.u.user.name,
484 XT_STANDARD_TARGET) == 0) &&
485 t->verdict < 0 && unconditional(&e->ip)) ||
487 unsigned int oldpos, size;
489 if ((strcmp(t->target.u.user.name,
490 XT_STANDARD_TARGET) == 0) &&
491 t->verdict < -NF_MAX_VERDICT - 1) {
492 duprintf("mark_source_chains: bad "
493 "negative verdict (%i)\n",
498 /* Return: backtrack through the last
501 e->comefrom ^= (1<<NF_INET_NUMHOOKS);
502 #ifdef DEBUG_IP_FIREWALL_USER
504 & (1 << NF_INET_NUMHOOKS)) {
505 duprintf("Back unset "
512 pos = e->counters.pcnt;
513 e->counters.pcnt = 0;
515 /* We're at the start. */
519 e = (struct ipt_entry *)
521 } while (oldpos == pos + e->next_offset);
524 size = e->next_offset;
525 e = (struct ipt_entry *)
526 (entry0 + pos + size);
527 e->counters.pcnt = pos;
529 WARN_ON_ONCE(calldepth == 0);
533 int newpos = t->verdict;
535 if (strcmp(t->target.u.user.name,
536 XT_STANDARD_TARGET) == 0 &&
538 if (newpos > newinfo->size -
539 sizeof(struct ipt_entry)) {
540 duprintf("mark_source_chains: "
541 "bad verdict (%i)\n",
545 if (entry0 + newpos != ipt_next_entry(e) &&
546 !(e->ip.flags & IPT_F_GOTO) &&
547 ++calldepth > max_calldepth)
548 max_calldepth = calldepth;
550 /* This a jump; chase it. */
551 duprintf("Jump rule %u -> %u, calldepth %d\n",
552 pos, newpos, calldepth);
554 /* ... this is a fallthru */
555 newpos = pos + e->next_offset;
557 e = (struct ipt_entry *)
559 e->counters.pcnt = pos;
564 duprintf("Finished chain %u\n", hook);
566 newinfo->stacksize = max_calldepth;
570 static void cleanup_match(struct xt_entry_match *m, struct net *net)
572 struct xt_mtdtor_param par;
575 par.match = m->u.kernel.match;
576 par.matchinfo = m->data;
577 par.family = NFPROTO_IPV4;
578 if (par.match->destroy != NULL)
579 par.match->destroy(&par);
580 module_put(par.match->me);
584 check_entry(const struct ipt_entry *e, const char *name)
586 const struct xt_entry_target *t;
588 if (!ip_checkentry(&e->ip)) {
589 duprintf("ip check failed %p %s.\n", e, name);
593 if (e->target_offset + sizeof(struct xt_entry_target) >
597 t = ipt_get_target_c(e);
598 if (e->target_offset + t->u.target_size > e->next_offset)
605 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
607 const struct ipt_ip *ip = par->entryinfo;
610 par->match = m->u.kernel.match;
611 par->matchinfo = m->data;
613 ret = xt_check_match(par, m->u.match_size - sizeof(*m),
614 ip->proto, ip->invflags & IPT_INV_PROTO);
616 duprintf("check failed for `%s'.\n", par->match->name);
623 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
625 struct xt_match *match;
628 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
631 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
632 return PTR_ERR(match);
634 m->u.kernel.match = match;
636 ret = check_match(m, par);
642 module_put(m->u.kernel.match->me);
646 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
648 struct xt_entry_target *t = ipt_get_target(e);
649 struct xt_tgchk_param par = {
653 .target = t->u.kernel.target,
655 .hook_mask = e->comefrom,
656 .family = NFPROTO_IPV4,
660 ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
661 e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
663 duprintf("check failed for `%s'.\n",
664 t->u.kernel.target->name);
671 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
674 struct xt_entry_target *t;
675 struct xt_target *target;
678 struct xt_mtchk_param mtpar;
679 struct xt_entry_match *ematch;
681 ret = check_entry(e, name);
685 e->counters.pcnt = xt_percpu_counter_alloc();
686 if (IS_ERR_VALUE(e->counters.pcnt))
692 mtpar.entryinfo = &e->ip;
693 mtpar.hook_mask = e->comefrom;
694 mtpar.family = NFPROTO_IPV4;
695 xt_ematch_foreach(ematch, e) {
696 ret = find_check_match(ematch, &mtpar);
698 goto cleanup_matches;
702 t = ipt_get_target(e);
703 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
705 if (IS_ERR(target)) {
706 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
707 ret = PTR_ERR(target);
708 goto cleanup_matches;
710 t->u.kernel.target = target;
712 ret = check_target(e, net, name);
718 module_put(t->u.kernel.target->me);
720 xt_ematch_foreach(ematch, e) {
723 cleanup_match(ematch, net);
726 xt_percpu_counter_free(e->counters.pcnt);
731 static bool check_underflow(const struct ipt_entry *e)
733 const struct xt_entry_target *t;
734 unsigned int verdict;
736 if (!unconditional(&e->ip))
738 t = ipt_get_target_c(e);
739 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
741 verdict = ((struct xt_standard_target *)t)->verdict;
742 verdict = -verdict - 1;
743 return verdict == NF_DROP || verdict == NF_ACCEPT;
747 check_entry_size_and_hooks(struct ipt_entry *e,
748 struct xt_table_info *newinfo,
749 const unsigned char *base,
750 const unsigned char *limit,
751 const unsigned int *hook_entries,
752 const unsigned int *underflows,
753 unsigned int valid_hooks)
757 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
758 (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
759 duprintf("Bad offset %p\n", e);
764 < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
765 duprintf("checking: element %p size %u\n",
770 /* Check hooks & underflows */
771 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
772 if (!(valid_hooks & (1 << h)))
774 if ((unsigned char *)e - base == hook_entries[h])
775 newinfo->hook_entry[h] = hook_entries[h];
776 if ((unsigned char *)e - base == underflows[h]) {
777 if (!check_underflow(e)) {
778 pr_err("Underflows must be unconditional and "
779 "use the STANDARD target with "
783 newinfo->underflow[h] = underflows[h];
787 /* Clear counters and comefrom */
788 e->counters = ((struct xt_counters) { 0, 0 });
794 cleanup_entry(struct ipt_entry *e, struct net *net)
796 struct xt_tgdtor_param par;
797 struct xt_entry_target *t;
798 struct xt_entry_match *ematch;
800 /* Cleanup all matches */
801 xt_ematch_foreach(ematch, e)
802 cleanup_match(ematch, net);
803 t = ipt_get_target(e);
806 par.target = t->u.kernel.target;
807 par.targinfo = t->data;
808 par.family = NFPROTO_IPV4;
809 if (par.target->destroy != NULL)
810 par.target->destroy(&par);
811 module_put(par.target->me);
812 xt_percpu_counter_free(e->counters.pcnt);
815 /* Checks and translates the user-supplied table segment (held in
818 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
819 const struct ipt_replace *repl)
821 struct ipt_entry *iter;
825 newinfo->size = repl->size;
826 newinfo->number = repl->num_entries;
828 /* Init all hooks to impossible value. */
829 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
830 newinfo->hook_entry[i] = 0xFFFFFFFF;
831 newinfo->underflow[i] = 0xFFFFFFFF;
834 duprintf("translate_table: size %u\n", newinfo->size);
836 /* Walk through entries, checking offsets. */
837 xt_entry_foreach(iter, entry0, newinfo->size) {
838 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
848 if (i != repl->num_entries) {
849 duprintf("translate_table: %u not %u entries\n",
850 i, repl->num_entries);
854 /* Check hooks all assigned */
855 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
856 /* Only hooks which are valid */
857 if (!(repl->valid_hooks & (1 << i)))
859 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
860 duprintf("Invalid hook entry %u %u\n",
861 i, repl->hook_entry[i]);
864 if (newinfo->underflow[i] == 0xFFFFFFFF) {
865 duprintf("Invalid underflow %u %u\n",
866 i, repl->underflow[i]);
871 if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
874 /* Finally, each sanity check must pass */
876 xt_entry_foreach(iter, entry0, newinfo->size) {
877 ret = find_check_entry(iter, net, repl->name, repl->size);
884 xt_entry_foreach(iter, entry0, newinfo->size) {
887 cleanup_entry(iter, net);
896 get_counters(const struct xt_table_info *t,
897 struct xt_counters counters[])
899 struct ipt_entry *iter;
903 for_each_possible_cpu(cpu) {
904 seqcount_t *s = &per_cpu(xt_recseq, cpu);
907 xt_entry_foreach(iter, t->entries, t->size) {
908 struct xt_counters *tmp;
912 tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
914 start = read_seqcount_begin(s);
917 } while (read_seqcount_retry(s, start));
919 ADD_COUNTER(counters[i], bcnt, pcnt);
920 ++i; /* macro does multi eval of i */
925 static struct xt_counters *alloc_counters(const struct xt_table *table)
927 unsigned int countersize;
928 struct xt_counters *counters;
929 const struct xt_table_info *private = table->private;
931 /* We need atomic snapshot of counters: rest doesn't change
932 (other than comefrom, which userspace doesn't care
934 countersize = sizeof(struct xt_counters) * private->number;
935 counters = vzalloc(countersize);
937 if (counters == NULL)
938 return ERR_PTR(-ENOMEM);
940 get_counters(private, counters);
946 copy_entries_to_user(unsigned int total_size,
947 const struct xt_table *table,
948 void __user *userptr)
950 unsigned int off, num;
951 const struct ipt_entry *e;
952 struct xt_counters *counters;
953 const struct xt_table_info *private = table->private;
955 const void *loc_cpu_entry;
957 counters = alloc_counters(table);
958 if (IS_ERR(counters))
959 return PTR_ERR(counters);
961 loc_cpu_entry = private->entries;
962 if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
967 /* FIXME: use iterator macros --RR */
968 /* ... then go back and fix counters and names */
969 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
971 const struct xt_entry_match *m;
972 const struct xt_entry_target *t;
974 e = (struct ipt_entry *)(loc_cpu_entry + off);
975 if (copy_to_user(userptr + off
976 + offsetof(struct ipt_entry, counters),
978 sizeof(counters[num])) != 0) {
983 for (i = sizeof(struct ipt_entry);
984 i < e->target_offset;
985 i += m->u.match_size) {
988 if (copy_to_user(userptr + off + i
989 + offsetof(struct xt_entry_match,
991 m->u.kernel.match->name,
992 strlen(m->u.kernel.match->name)+1)
999 t = ipt_get_target_c(e);
1000 if (copy_to_user(userptr + off + e->target_offset
1001 + offsetof(struct xt_entry_target,
1003 t->u.kernel.target->name,
1004 strlen(t->u.kernel.target->name)+1) != 0) {
1015 #ifdef CONFIG_COMPAT
1016 static void compat_standard_from_user(void *dst, const void *src)
1018 int v = *(compat_int_t *)src;
1021 v += xt_compat_calc_jump(AF_INET, v);
1022 memcpy(dst, &v, sizeof(v));
1025 static int compat_standard_to_user(void __user *dst, const void *src)
1027 compat_int_t cv = *(int *)src;
1030 cv -= xt_compat_calc_jump(AF_INET, cv);
1031 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1034 static int compat_calc_entry(const struct ipt_entry *e,
1035 const struct xt_table_info *info,
1036 const void *base, struct xt_table_info *newinfo)
1038 const struct xt_entry_match *ematch;
1039 const struct xt_entry_target *t;
1040 unsigned int entry_offset;
1043 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1044 entry_offset = (void *)e - base;
1045 xt_ematch_foreach(ematch, e)
1046 off += xt_compat_match_offset(ematch->u.kernel.match);
1047 t = ipt_get_target_c(e);
1048 off += xt_compat_target_offset(t->u.kernel.target);
1049 newinfo->size -= off;
1050 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1054 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1055 if (info->hook_entry[i] &&
1056 (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1057 newinfo->hook_entry[i] -= off;
1058 if (info->underflow[i] &&
1059 (e < (struct ipt_entry *)(base + info->underflow[i])))
1060 newinfo->underflow[i] -= off;
1065 static int compat_table_info(const struct xt_table_info *info,
1066 struct xt_table_info *newinfo)
1068 struct ipt_entry *iter;
1069 const void *loc_cpu_entry;
1072 if (!newinfo || !info)
1075 /* we dont care about newinfo->entries */
1076 memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1077 newinfo->initial_entries = 0;
1078 loc_cpu_entry = info->entries;
1079 xt_compat_init_offsets(AF_INET, info->number);
1080 xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1081 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1089 static int get_info(struct net *net, void __user *user,
1090 const int *len, int compat)
1092 char name[XT_TABLE_MAXNAMELEN];
1096 if (*len != sizeof(struct ipt_getinfo)) {
1097 duprintf("length %u != %zu\n", *len,
1098 sizeof(struct ipt_getinfo));
1102 if (copy_from_user(name, user, sizeof(name)) != 0)
1105 name[XT_TABLE_MAXNAMELEN-1] = '\0';
1106 #ifdef CONFIG_COMPAT
1108 xt_compat_lock(AF_INET);
1110 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1111 "iptable_%s", name);
1112 if (!IS_ERR_OR_NULL(t)) {
1113 struct ipt_getinfo info;
1114 const struct xt_table_info *private = t->private;
1115 #ifdef CONFIG_COMPAT
1116 struct xt_table_info tmp;
1119 ret = compat_table_info(private, &tmp);
1120 xt_compat_flush_offsets(AF_INET);
1124 memset(&info, 0, sizeof(info));
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,
1155 struct ipt_get_entries get;
1158 if (*len < sizeof(get)) {
1159 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1162 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1164 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1165 duprintf("get_entries: %u != %zu\n",
1166 *len, sizeof(get) + get.size);
1170 t = xt_find_table_lock(net, AF_INET, get.name);
1171 if (!IS_ERR_OR_NULL(t)) {
1172 const struct xt_table_info *private = t->private;
1173 duprintf("t->private->number = %u\n", private->number);
1174 if (get.size == private->size)
1175 ret = copy_entries_to_user(private->size,
1176 t, uptr->entrytable);
1178 duprintf("get_entries: I've got %u not %u!\n",
1179 private->size, get.size);
1185 ret = t ? PTR_ERR(t) : -ENOENT;
1191 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1192 struct xt_table_info *newinfo, unsigned int num_counters,
1193 void __user *counters_ptr)
1197 struct xt_table_info *oldinfo;
1198 struct xt_counters *counters;
1199 struct ipt_entry *iter;
1202 counters = vzalloc(num_counters * sizeof(struct xt_counters));
1208 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1209 "iptable_%s", name);
1210 if (IS_ERR_OR_NULL(t)) {
1211 ret = t ? PTR_ERR(t) : -ENOENT;
1212 goto free_newinfo_counters_untrans;
1216 if (valid_hooks != t->valid_hooks) {
1217 duprintf("Valid hook crap: %08X vs %08X\n",
1218 valid_hooks, t->valid_hooks);
1223 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1227 /* Update module usage count based on number of rules */
1228 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1229 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1230 if ((oldinfo->number > oldinfo->initial_entries) ||
1231 (newinfo->number <= oldinfo->initial_entries))
1233 if ((oldinfo->number > oldinfo->initial_entries) &&
1234 (newinfo->number <= oldinfo->initial_entries))
1237 /* Get the old counters, and synchronize with replace */
1238 get_counters(oldinfo, counters);
1240 /* Decrease module usage counts and free resource */
1241 xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1242 cleanup_entry(iter, net);
1244 xt_free_table_info(oldinfo);
1245 if (copy_to_user(counters_ptr, counters,
1246 sizeof(struct xt_counters) * num_counters) != 0) {
1247 /* Silent error, can't fail, new table is already in place */
1248 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1257 free_newinfo_counters_untrans:
1264 do_replace(struct net *net, const void __user *user, unsigned int len)
1267 struct ipt_replace tmp;
1268 struct xt_table_info *newinfo;
1269 void *loc_cpu_entry;
1270 struct ipt_entry *iter;
1272 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1275 /* overflow check */
1276 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1278 if (tmp.num_counters == 0)
1281 tmp.name[sizeof(tmp.name)-1] = 0;
1283 newinfo = xt_alloc_table_info(tmp.size);
1287 loc_cpu_entry = newinfo->entries;
1288 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1294 ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1298 duprintf("Translated table\n");
1300 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1301 tmp.num_counters, tmp.counters);
1303 goto free_newinfo_untrans;
1306 free_newinfo_untrans:
1307 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1308 cleanup_entry(iter, net);
1310 xt_free_table_info(newinfo);
1315 do_add_counters(struct net *net, const void __user *user,
1316 unsigned int len, int compat)
1319 struct xt_counters_info tmp;
1320 struct xt_counters *paddc;
1321 unsigned int num_counters;
1326 const struct xt_table_info *private;
1328 struct ipt_entry *iter;
1329 unsigned int addend;
1330 #ifdef CONFIG_COMPAT
1331 struct compat_xt_counters_info compat_tmp;
1335 size = sizeof(struct compat_xt_counters_info);
1340 size = sizeof(struct xt_counters_info);
1343 if (copy_from_user(ptmp, user, size) != 0)
1346 #ifdef CONFIG_COMPAT
1348 num_counters = compat_tmp.num_counters;
1349 name = compat_tmp.name;
1353 num_counters = tmp.num_counters;
1357 if (len != size + num_counters * sizeof(struct xt_counters))
1360 paddc = vmalloc(len - size);
1364 if (copy_from_user(paddc, user + size, len - size) != 0) {
1369 t = xt_find_table_lock(net, AF_INET, name);
1370 if (IS_ERR_OR_NULL(t)) {
1371 ret = t ? PTR_ERR(t) : -ENOENT;
1376 private = t->private;
1377 if (private->number != num_counters) {
1379 goto unlock_up_free;
1383 addend = xt_write_recseq_begin();
1384 xt_entry_foreach(iter, private->entries, private->size) {
1385 struct xt_counters *tmp;
1387 tmp = xt_get_this_cpu_counter(&iter->counters);
1388 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1391 xt_write_recseq_end(addend);
1402 #ifdef CONFIG_COMPAT
1403 struct compat_ipt_replace {
1404 char name[XT_TABLE_MAXNAMELEN];
1408 u32 hook_entry[NF_INET_NUMHOOKS];
1409 u32 underflow[NF_INET_NUMHOOKS];
1411 compat_uptr_t counters; /* struct xt_counters * */
1412 struct compat_ipt_entry entries[0];
1416 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1417 unsigned int *size, struct xt_counters *counters,
1420 struct xt_entry_target *t;
1421 struct compat_ipt_entry __user *ce;
1422 u_int16_t target_offset, next_offset;
1423 compat_uint_t origsize;
1424 const struct xt_entry_match *ematch;
1428 ce = (struct compat_ipt_entry __user *)*dstptr;
1429 if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1430 copy_to_user(&ce->counters, &counters[i],
1431 sizeof(counters[i])) != 0)
1434 *dstptr += sizeof(struct compat_ipt_entry);
1435 *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1437 xt_ematch_foreach(ematch, e) {
1438 ret = xt_compat_match_to_user(ematch, dstptr, size);
1442 target_offset = e->target_offset - (origsize - *size);
1443 t = ipt_get_target(e);
1444 ret = xt_compat_target_to_user(t, dstptr, size);
1447 next_offset = e->next_offset - (origsize - *size);
1448 if (put_user(target_offset, &ce->target_offset) != 0 ||
1449 put_user(next_offset, &ce->next_offset) != 0)
1455 compat_find_calc_match(struct xt_entry_match *m,
1457 const struct ipt_ip *ip,
1460 struct xt_match *match;
1462 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1463 m->u.user.revision);
1464 if (IS_ERR(match)) {
1465 duprintf("compat_check_calc_match: `%s' not found\n",
1467 return PTR_ERR(match);
1469 m->u.kernel.match = match;
1470 *size += xt_compat_match_offset(match);
1474 static void compat_release_entry(struct compat_ipt_entry *e)
1476 struct xt_entry_target *t;
1477 struct xt_entry_match *ematch;
1479 /* Cleanup all matches */
1480 xt_ematch_foreach(ematch, e)
1481 module_put(ematch->u.kernel.match->me);
1482 t = compat_ipt_get_target(e);
1483 module_put(t->u.kernel.target->me);
1487 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1488 struct xt_table_info *newinfo,
1490 const unsigned char *base,
1491 const unsigned char *limit,
1492 const unsigned int *hook_entries,
1493 const unsigned int *underflows,
1496 struct xt_entry_match *ematch;
1497 struct xt_entry_target *t;
1498 struct xt_target *target;
1499 unsigned int entry_offset;
1503 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1504 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1505 (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1506 duprintf("Bad offset %p, limit = %p\n", e, limit);
1510 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1511 sizeof(struct compat_xt_entry_target)) {
1512 duprintf("checking: element %p size %u\n",
1517 /* For purposes of check_entry casting the compat entry is fine */
1518 ret = check_entry((struct ipt_entry *)e, name);
1522 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1523 entry_offset = (void *)e - (void *)base;
1525 xt_ematch_foreach(ematch, e) {
1526 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1528 goto release_matches;
1532 t = compat_ipt_get_target(e);
1533 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1534 t->u.user.revision);
1535 if (IS_ERR(target)) {
1536 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1538 ret = PTR_ERR(target);
1539 goto release_matches;
1541 t->u.kernel.target = target;
1543 off += xt_compat_target_offset(target);
1545 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1549 /* Check hooks & underflows */
1550 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1551 if ((unsigned char *)e - base == hook_entries[h])
1552 newinfo->hook_entry[h] = hook_entries[h];
1553 if ((unsigned char *)e - base == underflows[h])
1554 newinfo->underflow[h] = underflows[h];
1557 /* Clear counters and comefrom */
1558 memset(&e->counters, 0, sizeof(e->counters));
1563 module_put(t->u.kernel.target->me);
1565 xt_ematch_foreach(ematch, e) {
1568 module_put(ematch->u.kernel.match->me);
1574 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1575 unsigned int *size, const char *name,
1576 struct xt_table_info *newinfo, unsigned char *base)
1578 struct xt_entry_target *t;
1579 struct xt_target *target;
1580 struct ipt_entry *de;
1581 unsigned int origsize;
1583 struct xt_entry_match *ematch;
1587 de = (struct ipt_entry *)*dstptr;
1588 memcpy(de, e, sizeof(struct ipt_entry));
1589 memcpy(&de->counters, &e->counters, sizeof(e->counters));
1591 *dstptr += sizeof(struct ipt_entry);
1592 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1594 xt_ematch_foreach(ematch, e) {
1595 ret = xt_compat_match_from_user(ematch, dstptr, size);
1599 de->target_offset = e->target_offset - (origsize - *size);
1600 t = compat_ipt_get_target(e);
1601 target = t->u.kernel.target;
1602 xt_compat_target_from_user(t, dstptr, size);
1604 de->next_offset = e->next_offset - (origsize - *size);
1605 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1606 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1607 newinfo->hook_entry[h] -= origsize - *size;
1608 if ((unsigned char *)de - base < newinfo->underflow[h])
1609 newinfo->underflow[h] -= origsize - *size;
1615 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1617 struct xt_entry_match *ematch;
1618 struct xt_mtchk_param mtpar;
1622 e->counters.pcnt = xt_percpu_counter_alloc();
1623 if (IS_ERR_VALUE(e->counters.pcnt))
1629 mtpar.entryinfo = &e->ip;
1630 mtpar.hook_mask = e->comefrom;
1631 mtpar.family = NFPROTO_IPV4;
1632 xt_ematch_foreach(ematch, e) {
1633 ret = check_match(ematch, &mtpar);
1635 goto cleanup_matches;
1639 ret = check_target(e, net, name);
1641 goto cleanup_matches;
1645 xt_ematch_foreach(ematch, e) {
1648 cleanup_match(ematch, net);
1651 xt_percpu_counter_free(e->counters.pcnt);
1657 translate_compat_table(struct net *net,
1659 unsigned int valid_hooks,
1660 struct xt_table_info **pinfo,
1662 unsigned int total_size,
1663 unsigned int number,
1664 unsigned int *hook_entries,
1665 unsigned int *underflows)
1668 struct xt_table_info *newinfo, *info;
1669 void *pos, *entry0, *entry1;
1670 struct compat_ipt_entry *iter0;
1671 struct ipt_entry *iter1;
1678 info->number = number;
1680 /* Init all hooks to impossible value. */
1681 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1682 info->hook_entry[i] = 0xFFFFFFFF;
1683 info->underflow[i] = 0xFFFFFFFF;
1686 duprintf("translate_compat_table: size %u\n", info->size);
1688 xt_compat_lock(AF_INET);
1689 xt_compat_init_offsets(AF_INET, number);
1690 /* Walk through entries, checking offsets. */
1691 xt_entry_foreach(iter0, entry0, total_size) {
1692 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1694 entry0 + total_size,
1705 duprintf("translate_compat_table: %u not %u entries\n",
1710 /* Check hooks all assigned */
1711 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1712 /* Only hooks which are valid */
1713 if (!(valid_hooks & (1 << i)))
1715 if (info->hook_entry[i] == 0xFFFFFFFF) {
1716 duprintf("Invalid hook entry %u %u\n",
1717 i, hook_entries[i]);
1720 if (info->underflow[i] == 0xFFFFFFFF) {
1721 duprintf("Invalid underflow %u %u\n",
1728 newinfo = xt_alloc_table_info(size);
1732 newinfo->number = number;
1733 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1734 newinfo->hook_entry[i] = info->hook_entry[i];
1735 newinfo->underflow[i] = info->underflow[i];
1737 entry1 = newinfo->entries;
1740 xt_entry_foreach(iter0, entry0, total_size) {
1741 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1742 name, newinfo, entry1);
1746 xt_compat_flush_offsets(AF_INET);
1747 xt_compat_unlock(AF_INET);
1752 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1756 xt_entry_foreach(iter1, entry1, newinfo->size) {
1757 ret = compat_check_entry(iter1, net, name);
1764 * The first i matches need cleanup_entry (calls ->destroy)
1765 * because they had called ->check already. The other j-i
1766 * entries need only release.
1770 xt_entry_foreach(iter0, entry0, newinfo->size) {
1775 compat_release_entry(iter0);
1777 xt_entry_foreach(iter1, entry1, newinfo->size) {
1780 cleanup_entry(iter1, net);
1782 xt_free_table_info(newinfo);
1788 xt_free_table_info(info);
1792 xt_free_table_info(newinfo);
1794 xt_entry_foreach(iter0, entry0, total_size) {
1797 compat_release_entry(iter0);
1801 xt_compat_flush_offsets(AF_INET);
1802 xt_compat_unlock(AF_INET);
1807 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1810 struct compat_ipt_replace tmp;
1811 struct xt_table_info *newinfo;
1812 void *loc_cpu_entry;
1813 struct ipt_entry *iter;
1815 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1818 /* overflow check */
1819 if (tmp.size >= INT_MAX / num_possible_cpus())
1821 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1823 if (tmp.num_counters == 0)
1826 tmp.name[sizeof(tmp.name)-1] = 0;
1828 newinfo = xt_alloc_table_info(tmp.size);
1832 loc_cpu_entry = newinfo->entries;
1833 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1839 ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1840 &newinfo, &loc_cpu_entry, tmp.size,
1841 tmp.num_entries, tmp.hook_entry,
1846 duprintf("compat_do_replace: Translated table\n");
1848 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1849 tmp.num_counters, compat_ptr(tmp.counters));
1851 goto free_newinfo_untrans;
1854 free_newinfo_untrans:
1855 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1856 cleanup_entry(iter, net);
1858 xt_free_table_info(newinfo);
1863 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1868 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1872 case IPT_SO_SET_REPLACE:
1873 ret = compat_do_replace(sock_net(sk), user, len);
1876 case IPT_SO_SET_ADD_COUNTERS:
1877 ret = do_add_counters(sock_net(sk), user, len, 1);
1881 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1888 struct compat_ipt_get_entries {
1889 char name[XT_TABLE_MAXNAMELEN];
1891 struct compat_ipt_entry entrytable[0];
1895 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1896 void __user *userptr)
1898 struct xt_counters *counters;
1899 const struct xt_table_info *private = table->private;
1904 struct ipt_entry *iter;
1906 counters = alloc_counters(table);
1907 if (IS_ERR(counters))
1908 return PTR_ERR(counters);
1912 xt_entry_foreach(iter, private->entries, total_size) {
1913 ret = compat_copy_entry_to_user(iter, &pos,
1914 &size, counters, i++);
1924 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1928 struct compat_ipt_get_entries get;
1931 if (*len < sizeof(get)) {
1932 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1936 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1939 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1940 duprintf("compat_get_entries: %u != %zu\n",
1941 *len, sizeof(get) + get.size);
1945 xt_compat_lock(AF_INET);
1946 t = xt_find_table_lock(net, AF_INET, get.name);
1947 if (!IS_ERR_OR_NULL(t)) {
1948 const struct xt_table_info *private = t->private;
1949 struct xt_table_info info;
1950 duprintf("t->private->number = %u\n", private->number);
1951 ret = compat_table_info(private, &info);
1952 if (!ret && get.size == info.size) {
1953 ret = compat_copy_entries_to_user(private->size,
1954 t, uptr->entrytable);
1956 duprintf("compat_get_entries: I've got %u not %u!\n",
1957 private->size, get.size);
1960 xt_compat_flush_offsets(AF_INET);
1964 ret = t ? PTR_ERR(t) : -ENOENT;
1966 xt_compat_unlock(AF_INET);
1970 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1973 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1977 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1981 case IPT_SO_GET_INFO:
1982 ret = get_info(sock_net(sk), user, len, 1);
1984 case IPT_SO_GET_ENTRIES:
1985 ret = compat_get_entries(sock_net(sk), user, len);
1988 ret = do_ipt_get_ctl(sk, cmd, user, len);
1995 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1999 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2003 case IPT_SO_SET_REPLACE:
2004 ret = do_replace(sock_net(sk), user, len);
2007 case IPT_SO_SET_ADD_COUNTERS:
2008 ret = do_add_counters(sock_net(sk), user, len, 0);
2012 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
2020 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2024 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2028 case IPT_SO_GET_INFO:
2029 ret = get_info(sock_net(sk), user, len, 0);
2032 case IPT_SO_GET_ENTRIES:
2033 ret = get_entries(sock_net(sk), user, len);
2036 case IPT_SO_GET_REVISION_MATCH:
2037 case IPT_SO_GET_REVISION_TARGET: {
2038 struct xt_get_revision rev;
2041 if (*len != sizeof(rev)) {
2045 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2049 rev.name[sizeof(rev.name)-1] = 0;
2051 if (cmd == IPT_SO_GET_REVISION_TARGET)
2056 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2059 "ipt_%s", rev.name);
2064 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2071 struct xt_table *ipt_register_table(struct net *net,
2072 const struct xt_table *table,
2073 const struct ipt_replace *repl)
2076 struct xt_table_info *newinfo;
2077 struct xt_table_info bootstrap = {0};
2078 void *loc_cpu_entry;
2079 struct xt_table *new_table;
2081 newinfo = xt_alloc_table_info(repl->size);
2087 loc_cpu_entry = newinfo->entries;
2088 memcpy(loc_cpu_entry, repl->entries, repl->size);
2090 ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2094 new_table = xt_register_table(net, table, &bootstrap, newinfo);
2095 if (IS_ERR(new_table)) {
2096 ret = PTR_ERR(new_table);
2103 xt_free_table_info(newinfo);
2105 return ERR_PTR(ret);
2108 void ipt_unregister_table(struct net *net, struct xt_table *table)
2110 struct xt_table_info *private;
2111 void *loc_cpu_entry;
2112 struct module *table_owner = table->me;
2113 struct ipt_entry *iter;
2115 private = xt_unregister_table(table);
2117 /* Decrease module usage counts and free resources */
2118 loc_cpu_entry = private->entries;
2119 xt_entry_foreach(iter, loc_cpu_entry, private->size)
2120 cleanup_entry(iter, net);
2121 if (private->number > private->initial_entries)
2122 module_put(table_owner);
2123 xt_free_table_info(private);
2126 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2128 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2129 u_int8_t type, u_int8_t code,
2132 return ((test_type == 0xFF) ||
2133 (type == test_type && code >= min_code && code <= max_code))
2138 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2140 const struct icmphdr *ic;
2141 struct icmphdr _icmph;
2142 const struct ipt_icmp *icmpinfo = par->matchinfo;
2144 /* Must not be a fragment. */
2145 if (par->fragoff != 0)
2148 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2150 /* We've been asked to examine this packet, and we
2151 * can't. Hence, no choice but to drop.
2153 duprintf("Dropping evil ICMP tinygram.\n");
2154 par->hotdrop = true;
2158 return icmp_type_code_match(icmpinfo->type,
2162 !!(icmpinfo->invflags&IPT_ICMP_INV));
2165 static int icmp_checkentry(const struct xt_mtchk_param *par)
2167 const struct ipt_icmp *icmpinfo = par->matchinfo;
2169 /* Must specify no unknown invflags */
2170 return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2173 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2175 .name = XT_STANDARD_TARGET,
2176 .targetsize = sizeof(int),
2177 .family = NFPROTO_IPV4,
2178 #ifdef CONFIG_COMPAT
2179 .compatsize = sizeof(compat_int_t),
2180 .compat_from_user = compat_standard_from_user,
2181 .compat_to_user = compat_standard_to_user,
2185 .name = XT_ERROR_TARGET,
2186 .target = ipt_error,
2187 .targetsize = XT_FUNCTION_MAXNAMELEN,
2188 .family = NFPROTO_IPV4,
2192 static struct nf_sockopt_ops ipt_sockopts = {
2194 .set_optmin = IPT_BASE_CTL,
2195 .set_optmax = IPT_SO_SET_MAX+1,
2196 .set = do_ipt_set_ctl,
2197 #ifdef CONFIG_COMPAT
2198 .compat_set = compat_do_ipt_set_ctl,
2200 .get_optmin = IPT_BASE_CTL,
2201 .get_optmax = IPT_SO_GET_MAX+1,
2202 .get = do_ipt_get_ctl,
2203 #ifdef CONFIG_COMPAT
2204 .compat_get = compat_do_ipt_get_ctl,
2206 .owner = THIS_MODULE,
2209 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2212 .match = icmp_match,
2213 .matchsize = sizeof(struct ipt_icmp),
2214 .checkentry = icmp_checkentry,
2215 .proto = IPPROTO_ICMP,
2216 .family = NFPROTO_IPV4,
2220 static int __net_init ip_tables_net_init(struct net *net)
2222 return xt_proto_init(net, NFPROTO_IPV4);
2225 static void __net_exit ip_tables_net_exit(struct net *net)
2227 xt_proto_fini(net, NFPROTO_IPV4);
2230 static struct pernet_operations ip_tables_net_ops = {
2231 .init = ip_tables_net_init,
2232 .exit = ip_tables_net_exit,
2235 static int __init ip_tables_init(void)
2239 ret = register_pernet_subsys(&ip_tables_net_ops);
2243 /* No one else will be downing sem now, so we won't sleep */
2244 ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2247 ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2251 /* Register setsockopt */
2252 ret = nf_register_sockopt(&ipt_sockopts);
2256 pr_info("(C) 2000-2006 Netfilter Core Team\n");
2260 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2262 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2264 unregister_pernet_subsys(&ip_tables_net_ops);
2269 static void __exit ip_tables_fini(void)
2271 nf_unregister_sockopt(&ipt_sockopts);
2273 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2274 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2275 unregister_pernet_subsys(&ip_tables_net_ops);
2278 EXPORT_SYMBOL(ipt_register_table);
2279 EXPORT_SYMBOL(ipt_unregister_table);
2280 EXPORT_SYMBOL(ipt_do_table);
2281 module_init(ip_tables_init);
2282 module_exit(ip_tables_fini);