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 void *table_base;
258 const struct ipt_entry *root;
259 const char *hookname, *chainname, *comment;
260 const struct ipt_entry *iter;
261 unsigned int rulenum = 0;
262 struct net *net = dev_net(in ? in : out);
264 table_base = private->entries[smp_processor_id()];
265 root = get_entry(table_base, private->hook_entry[hook]);
267 hookname = chainname = hooknames[hook];
268 comment = comments[NF_IP_TRACE_COMMENT_RULE];
270 xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
271 if (get_chainname_rulenum(iter, e, hookname,
272 &chainname, &comment, &rulenum) != 0)
275 nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
276 "TRACE: %s:%s:%s:%u ",
277 tablename, chainname, comment, rulenum);
282 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
284 return (void *)entry + entry->next_offset;
287 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
289 ipt_do_table(struct sk_buff *skb,
291 const struct nf_hook_state *state,
292 struct xt_table *table)
294 static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
295 const struct iphdr *ip;
296 /* Initializing verdict to NF_DROP keeps gcc happy. */
297 unsigned int verdict = NF_DROP;
298 const char *indev, *outdev;
299 const void *table_base;
300 struct ipt_entry *e, **jumpstack;
301 unsigned int *stackptr, origptr, cpu;
302 const struct xt_table_info *private;
303 struct xt_action_param acpar;
308 indev = state->in ? state->in->name : nulldevname;
309 outdev = state->out ? state->out->name : nulldevname;
310 /* We handle fragments by dealing with the first fragment as
311 * if it was a normal packet. All other fragments are treated
312 * normally, except that they will NEVER match rules that ask
313 * things we don't know, ie. tcp syn flag or ports). If the
314 * rule is also a fragment-specific rule, non-fragments won't
316 acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
317 acpar.thoff = ip_hdrlen(skb);
318 acpar.hotdrop = false;
319 acpar.in = state->in;
320 acpar.out = state->out;
321 acpar.family = NFPROTO_IPV4;
322 acpar.hooknum = hook;
324 IP_NF_ASSERT(table->valid_hooks & (1 << hook));
326 addend = xt_write_recseq_begin();
327 private = table->private;
328 cpu = smp_processor_id();
330 * Ensure we load private-> members after we've fetched the base
333 smp_read_barrier_depends();
334 table_base = private->entries[cpu];
335 jumpstack = (struct ipt_entry **)private->jumpstack[cpu];
336 stackptr = per_cpu_ptr(private->stackptr, cpu);
339 e = get_entry(table_base, private->hook_entry[hook]);
341 pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
342 table->name, hook, origptr,
343 get_entry(table_base, private->underflow[hook]));
346 const struct xt_entry_target *t;
347 const struct xt_entry_match *ematch;
350 if (!ip_packet_match(ip, indev, outdev,
351 &e->ip, acpar.fragoff)) {
353 e = ipt_next_entry(e);
357 xt_ematch_foreach(ematch, e) {
358 acpar.match = ematch->u.kernel.match;
359 acpar.matchinfo = ematch->data;
360 if (!acpar.match->match(skb, &acpar))
364 ADD_COUNTER(e->counters, skb->len, 1);
366 t = ipt_get_target(e);
367 IP_NF_ASSERT(t->u.kernel.target);
369 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
370 /* The packet is traced: log it */
371 if (unlikely(skb->nf_trace))
372 trace_packet(skb, hook, state->in, state->out,
373 table->name, private, e);
375 /* Standard target? */
376 if (!t->u.kernel.target->target) {
379 v = ((struct xt_standard_target *)t)->verdict;
381 /* Pop from stack? */
382 if (v != XT_RETURN) {
383 verdict = (unsigned int)(-v) - 1;
386 if (*stackptr <= origptr) {
387 e = get_entry(table_base,
388 private->underflow[hook]);
389 pr_debug("Underflow (this is normal) "
392 e = jumpstack[--*stackptr];
393 pr_debug("Pulled %p out from pos %u\n",
395 e = ipt_next_entry(e);
399 if (table_base + v != ipt_next_entry(e) &&
400 !(e->ip.flags & IPT_F_GOTO)) {
401 if (*stackptr >= private->stacksize) {
405 jumpstack[(*stackptr)++] = e;
406 pr_debug("Pushed %p into pos %u\n",
410 e = get_entry(table_base, v);
414 acpar.target = t->u.kernel.target;
415 acpar.targinfo = t->data;
417 verdict = t->u.kernel.target->target(skb, &acpar);
418 /* Target might have changed stuff. */
420 if (verdict == XT_CONTINUE)
421 e = ipt_next_entry(e);
425 } while (!acpar.hotdrop);
426 pr_debug("Exiting %s; resetting sp from %u to %u\n",
427 __func__, *stackptr, origptr);
429 xt_write_recseq_end(addend);
432 #ifdef DEBUG_ALLOW_ALL
441 /* Figures out from what hook each rule can be called: returns 0 if
442 there are loops. Puts hook bitmask in comefrom. */
444 mark_source_chains(const struct xt_table_info *newinfo,
445 unsigned int valid_hooks, void *entry0)
449 /* No recursion; use packet counter to save back ptrs (reset
450 to 0 as we leave), and comefrom to save source hook bitmask */
451 for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
452 unsigned int pos = newinfo->hook_entry[hook];
453 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
455 if (!(valid_hooks & (1 << hook)))
458 /* Set initial back pointer. */
459 e->counters.pcnt = pos;
462 const struct xt_standard_target *t
463 = (void *)ipt_get_target_c(e);
464 int visited = e->comefrom & (1 << hook);
466 if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
467 pr_err("iptables: loop hook %u pos %u %08X.\n",
468 hook, pos, e->comefrom);
471 e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
473 /* Unconditional return/END. */
474 if ((e->target_offset == sizeof(struct ipt_entry) &&
475 (strcmp(t->target.u.user.name,
476 XT_STANDARD_TARGET) == 0) &&
477 t->verdict < 0 && unconditional(&e->ip)) ||
479 unsigned int oldpos, size;
481 if ((strcmp(t->target.u.user.name,
482 XT_STANDARD_TARGET) == 0) &&
483 t->verdict < -NF_MAX_VERDICT - 1) {
484 duprintf("mark_source_chains: bad "
485 "negative verdict (%i)\n",
490 /* Return: backtrack through the last
493 e->comefrom ^= (1<<NF_INET_NUMHOOKS);
494 #ifdef DEBUG_IP_FIREWALL_USER
496 & (1 << NF_INET_NUMHOOKS)) {
497 duprintf("Back unset "
504 pos = e->counters.pcnt;
505 e->counters.pcnt = 0;
507 /* We're at the start. */
511 e = (struct ipt_entry *)
513 } while (oldpos == pos + e->next_offset);
516 size = e->next_offset;
517 e = (struct ipt_entry *)
518 (entry0 + pos + size);
519 e->counters.pcnt = pos;
522 int newpos = t->verdict;
524 if (strcmp(t->target.u.user.name,
525 XT_STANDARD_TARGET) == 0 &&
527 if (newpos > newinfo->size -
528 sizeof(struct ipt_entry)) {
529 duprintf("mark_source_chains: "
530 "bad verdict (%i)\n",
534 /* This a jump; chase it. */
535 duprintf("Jump rule %u -> %u\n",
538 /* ... this is a fallthru */
539 newpos = pos + e->next_offset;
541 e = (struct ipt_entry *)
543 e->counters.pcnt = pos;
548 duprintf("Finished chain %u\n", hook);
553 static void cleanup_match(struct xt_entry_match *m, struct net *net)
555 struct xt_mtdtor_param par;
558 par.match = m->u.kernel.match;
559 par.matchinfo = m->data;
560 par.family = NFPROTO_IPV4;
561 if (par.match->destroy != NULL)
562 par.match->destroy(&par);
563 module_put(par.match->me);
567 check_entry(const struct ipt_entry *e, const char *name)
569 const struct xt_entry_target *t;
571 if (!ip_checkentry(&e->ip)) {
572 duprintf("ip check failed %p %s.\n", e, name);
576 if (e->target_offset + sizeof(struct xt_entry_target) >
580 t = ipt_get_target_c(e);
581 if (e->target_offset + t->u.target_size > e->next_offset)
588 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
590 const struct ipt_ip *ip = par->entryinfo;
593 par->match = m->u.kernel.match;
594 par->matchinfo = m->data;
596 ret = xt_check_match(par, m->u.match_size - sizeof(*m),
597 ip->proto, ip->invflags & IPT_INV_PROTO);
599 duprintf("check failed for `%s'.\n", par->match->name);
606 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
608 struct xt_match *match;
611 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
614 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
615 return PTR_ERR(match);
617 m->u.kernel.match = match;
619 ret = check_match(m, par);
625 module_put(m->u.kernel.match->me);
629 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
631 struct xt_entry_target *t = ipt_get_target(e);
632 struct xt_tgchk_param par = {
636 .target = t->u.kernel.target,
638 .hook_mask = e->comefrom,
639 .family = NFPROTO_IPV4,
643 ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
644 e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
646 duprintf("check failed for `%s'.\n",
647 t->u.kernel.target->name);
654 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
657 struct xt_entry_target *t;
658 struct xt_target *target;
661 struct xt_mtchk_param mtpar;
662 struct xt_entry_match *ematch;
664 ret = check_entry(e, name);
671 mtpar.entryinfo = &e->ip;
672 mtpar.hook_mask = e->comefrom;
673 mtpar.family = NFPROTO_IPV4;
674 xt_ematch_foreach(ematch, e) {
675 ret = find_check_match(ematch, &mtpar);
677 goto cleanup_matches;
681 t = ipt_get_target(e);
682 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
684 if (IS_ERR(target)) {
685 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
686 ret = PTR_ERR(target);
687 goto cleanup_matches;
689 t->u.kernel.target = target;
691 ret = check_target(e, net, name);
696 module_put(t->u.kernel.target->me);
698 xt_ematch_foreach(ematch, e) {
701 cleanup_match(ematch, net);
706 static bool check_underflow(const struct ipt_entry *e)
708 const struct xt_entry_target *t;
709 unsigned int verdict;
711 if (!unconditional(&e->ip))
713 t = ipt_get_target_c(e);
714 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
716 verdict = ((struct xt_standard_target *)t)->verdict;
717 verdict = -verdict - 1;
718 return verdict == NF_DROP || verdict == NF_ACCEPT;
722 check_entry_size_and_hooks(struct ipt_entry *e,
723 struct xt_table_info *newinfo,
724 const unsigned char *base,
725 const unsigned char *limit,
726 const unsigned int *hook_entries,
727 const unsigned int *underflows,
728 unsigned int valid_hooks)
732 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
733 (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
734 duprintf("Bad offset %p\n", e);
739 < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
740 duprintf("checking: element %p size %u\n",
745 /* Check hooks & underflows */
746 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
747 if (!(valid_hooks & (1 << h)))
749 if ((unsigned char *)e - base == hook_entries[h])
750 newinfo->hook_entry[h] = hook_entries[h];
751 if ((unsigned char *)e - base == underflows[h]) {
752 if (!check_underflow(e)) {
753 pr_err("Underflows must be unconditional and "
754 "use the STANDARD target with "
758 newinfo->underflow[h] = underflows[h];
762 /* Clear counters and comefrom */
763 e->counters = ((struct xt_counters) { 0, 0 });
769 cleanup_entry(struct ipt_entry *e, struct net *net)
771 struct xt_tgdtor_param par;
772 struct xt_entry_target *t;
773 struct xt_entry_match *ematch;
775 /* Cleanup all matches */
776 xt_ematch_foreach(ematch, e)
777 cleanup_match(ematch, net);
778 t = ipt_get_target(e);
781 par.target = t->u.kernel.target;
782 par.targinfo = t->data;
783 par.family = NFPROTO_IPV4;
784 if (par.target->destroy != NULL)
785 par.target->destroy(&par);
786 module_put(par.target->me);
789 /* Checks and translates the user-supplied table segment (held in
792 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
793 const struct ipt_replace *repl)
795 struct ipt_entry *iter;
799 newinfo->size = repl->size;
800 newinfo->number = repl->num_entries;
802 /* Init all hooks to impossible value. */
803 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
804 newinfo->hook_entry[i] = 0xFFFFFFFF;
805 newinfo->underflow[i] = 0xFFFFFFFF;
808 duprintf("translate_table: size %u\n", newinfo->size);
810 /* Walk through entries, checking offsets. */
811 xt_entry_foreach(iter, entry0, newinfo->size) {
812 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
820 if (strcmp(ipt_get_target(iter)->u.user.name,
821 XT_ERROR_TARGET) == 0)
822 ++newinfo->stacksize;
825 if (i != repl->num_entries) {
826 duprintf("translate_table: %u not %u entries\n",
827 i, repl->num_entries);
831 /* Check hooks all assigned */
832 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
833 /* Only hooks which are valid */
834 if (!(repl->valid_hooks & (1 << i)))
836 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
837 duprintf("Invalid hook entry %u %u\n",
838 i, repl->hook_entry[i]);
841 if (newinfo->underflow[i] == 0xFFFFFFFF) {
842 duprintf("Invalid underflow %u %u\n",
843 i, repl->underflow[i]);
848 if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
851 /* Finally, each sanity check must pass */
853 xt_entry_foreach(iter, entry0, newinfo->size) {
854 ret = find_check_entry(iter, net, repl->name, repl->size);
861 xt_entry_foreach(iter, entry0, newinfo->size) {
864 cleanup_entry(iter, net);
869 /* And one copy for every other CPU */
870 for_each_possible_cpu(i) {
871 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
872 memcpy(newinfo->entries[i], entry0, newinfo->size);
879 get_counters(const struct xt_table_info *t,
880 struct xt_counters counters[])
882 struct ipt_entry *iter;
886 for_each_possible_cpu(cpu) {
887 seqcount_t *s = &per_cpu(xt_recseq, cpu);
890 xt_entry_foreach(iter, t->entries[cpu], t->size) {
895 start = read_seqcount_begin(s);
896 bcnt = iter->counters.bcnt;
897 pcnt = iter->counters.pcnt;
898 } while (read_seqcount_retry(s, start));
900 ADD_COUNTER(counters[i], bcnt, pcnt);
901 ++i; /* macro does multi eval of i */
906 static struct xt_counters *alloc_counters(const struct xt_table *table)
908 unsigned int countersize;
909 struct xt_counters *counters;
910 const struct xt_table_info *private = table->private;
912 /* We need atomic snapshot of counters: rest doesn't change
913 (other than comefrom, which userspace doesn't care
915 countersize = sizeof(struct xt_counters) * private->number;
916 counters = vzalloc(countersize);
918 if (counters == NULL)
919 return ERR_PTR(-ENOMEM);
921 get_counters(private, counters);
927 copy_entries_to_user(unsigned int total_size,
928 const struct xt_table *table,
929 void __user *userptr)
931 unsigned int off, num;
932 const struct ipt_entry *e;
933 struct xt_counters *counters;
934 const struct xt_table_info *private = table->private;
936 const void *loc_cpu_entry;
938 counters = alloc_counters(table);
939 if (IS_ERR(counters))
940 return PTR_ERR(counters);
942 /* choose the copy that is on our node/cpu, ...
943 * This choice is lazy (because current thread is
944 * allowed to migrate to another cpu)
946 loc_cpu_entry = private->entries[raw_smp_processor_id()];
947 if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
952 /* FIXME: use iterator macros --RR */
953 /* ... then go back and fix counters and names */
954 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
956 const struct xt_entry_match *m;
957 const struct xt_entry_target *t;
959 e = (struct ipt_entry *)(loc_cpu_entry + off);
960 if (copy_to_user(userptr + off
961 + offsetof(struct ipt_entry, counters),
963 sizeof(counters[num])) != 0) {
968 for (i = sizeof(struct ipt_entry);
969 i < e->target_offset;
970 i += m->u.match_size) {
973 if (copy_to_user(userptr + off + i
974 + offsetof(struct xt_entry_match,
976 m->u.kernel.match->name,
977 strlen(m->u.kernel.match->name)+1)
984 t = ipt_get_target_c(e);
985 if (copy_to_user(userptr + off + e->target_offset
986 + offsetof(struct xt_entry_target,
988 t->u.kernel.target->name,
989 strlen(t->u.kernel.target->name)+1) != 0) {
1000 #ifdef CONFIG_COMPAT
1001 static void compat_standard_from_user(void *dst, const void *src)
1003 int v = *(compat_int_t *)src;
1006 v += xt_compat_calc_jump(AF_INET, v);
1007 memcpy(dst, &v, sizeof(v));
1010 static int compat_standard_to_user(void __user *dst, const void *src)
1012 compat_int_t cv = *(int *)src;
1015 cv -= xt_compat_calc_jump(AF_INET, cv);
1016 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1019 static int compat_calc_entry(const struct ipt_entry *e,
1020 const struct xt_table_info *info,
1021 const void *base, struct xt_table_info *newinfo)
1023 const struct xt_entry_match *ematch;
1024 const struct xt_entry_target *t;
1025 unsigned int entry_offset;
1028 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1029 entry_offset = (void *)e - base;
1030 xt_ematch_foreach(ematch, e)
1031 off += xt_compat_match_offset(ematch->u.kernel.match);
1032 t = ipt_get_target_c(e);
1033 off += xt_compat_target_offset(t->u.kernel.target);
1034 newinfo->size -= off;
1035 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1039 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1040 if (info->hook_entry[i] &&
1041 (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1042 newinfo->hook_entry[i] -= off;
1043 if (info->underflow[i] &&
1044 (e < (struct ipt_entry *)(base + info->underflow[i])))
1045 newinfo->underflow[i] -= off;
1050 static int compat_table_info(const struct xt_table_info *info,
1051 struct xt_table_info *newinfo)
1053 struct ipt_entry *iter;
1054 void *loc_cpu_entry;
1057 if (!newinfo || !info)
1060 /* we dont care about newinfo->entries[] */
1061 memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1062 newinfo->initial_entries = 0;
1063 loc_cpu_entry = info->entries[raw_smp_processor_id()];
1064 xt_compat_init_offsets(AF_INET, info->number);
1065 xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1066 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1074 static int get_info(struct net *net, void __user *user,
1075 const int *len, int compat)
1077 char name[XT_TABLE_MAXNAMELEN];
1081 if (*len != sizeof(struct ipt_getinfo)) {
1082 duprintf("length %u != %zu\n", *len,
1083 sizeof(struct ipt_getinfo));
1087 if (copy_from_user(name, user, sizeof(name)) != 0)
1090 name[XT_TABLE_MAXNAMELEN-1] = '\0';
1091 #ifdef CONFIG_COMPAT
1093 xt_compat_lock(AF_INET);
1095 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1096 "iptable_%s", name);
1097 if (!IS_ERR_OR_NULL(t)) {
1098 struct ipt_getinfo info;
1099 const struct xt_table_info *private = t->private;
1100 #ifdef CONFIG_COMPAT
1101 struct xt_table_info tmp;
1104 ret = compat_table_info(private, &tmp);
1105 xt_compat_flush_offsets(AF_INET);
1109 memset(&info, 0, sizeof(info));
1110 info.valid_hooks = t->valid_hooks;
1111 memcpy(info.hook_entry, private->hook_entry,
1112 sizeof(info.hook_entry));
1113 memcpy(info.underflow, private->underflow,
1114 sizeof(info.underflow));
1115 info.num_entries = private->number;
1116 info.size = private->size;
1117 strcpy(info.name, name);
1119 if (copy_to_user(user, &info, *len) != 0)
1127 ret = t ? PTR_ERR(t) : -ENOENT;
1128 #ifdef CONFIG_COMPAT
1130 xt_compat_unlock(AF_INET);
1136 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1140 struct ipt_get_entries get;
1143 if (*len < sizeof(get)) {
1144 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1147 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1149 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1150 duprintf("get_entries: %u != %zu\n",
1151 *len, sizeof(get) + get.size);
1155 t = xt_find_table_lock(net, AF_INET, get.name);
1156 if (!IS_ERR_OR_NULL(t)) {
1157 const struct xt_table_info *private = t->private;
1158 duprintf("t->private->number = %u\n", private->number);
1159 if (get.size == private->size)
1160 ret = copy_entries_to_user(private->size,
1161 t, uptr->entrytable);
1163 duprintf("get_entries: I've got %u not %u!\n",
1164 private->size, get.size);
1170 ret = t ? PTR_ERR(t) : -ENOENT;
1176 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1177 struct xt_table_info *newinfo, unsigned int num_counters,
1178 void __user *counters_ptr)
1182 struct xt_table_info *oldinfo;
1183 struct xt_counters *counters;
1184 void *loc_cpu_old_entry;
1185 struct ipt_entry *iter;
1188 counters = vzalloc(num_counters * sizeof(struct xt_counters));
1194 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1195 "iptable_%s", name);
1196 if (IS_ERR_OR_NULL(t)) {
1197 ret = t ? PTR_ERR(t) : -ENOENT;
1198 goto free_newinfo_counters_untrans;
1202 if (valid_hooks != t->valid_hooks) {
1203 duprintf("Valid hook crap: %08X vs %08X\n",
1204 valid_hooks, t->valid_hooks);
1209 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1213 /* Update module usage count based on number of rules */
1214 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1215 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1216 if ((oldinfo->number > oldinfo->initial_entries) ||
1217 (newinfo->number <= oldinfo->initial_entries))
1219 if ((oldinfo->number > oldinfo->initial_entries) &&
1220 (newinfo->number <= oldinfo->initial_entries))
1223 /* Get the old counters, and synchronize with replace */
1224 get_counters(oldinfo, counters);
1226 /* Decrease module usage counts and free resource */
1227 loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1228 xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1229 cleanup_entry(iter, net);
1231 xt_free_table_info(oldinfo);
1232 if (copy_to_user(counters_ptr, counters,
1233 sizeof(struct xt_counters) * num_counters) != 0) {
1234 /* Silent error, can't fail, new table is already in place */
1235 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1244 free_newinfo_counters_untrans:
1251 do_replace(struct net *net, const void __user *user, unsigned int len)
1254 struct ipt_replace tmp;
1255 struct xt_table_info *newinfo;
1256 void *loc_cpu_entry;
1257 struct ipt_entry *iter;
1259 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1262 /* overflow check */
1263 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1265 if (tmp.num_counters == 0)
1268 tmp.name[sizeof(tmp.name)-1] = 0;
1270 newinfo = xt_alloc_table_info(tmp.size);
1274 /* choose the copy that is on our node/cpu */
1275 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1276 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1282 ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1286 duprintf("Translated table\n");
1288 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1289 tmp.num_counters, tmp.counters);
1291 goto free_newinfo_untrans;
1294 free_newinfo_untrans:
1295 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1296 cleanup_entry(iter, net);
1298 xt_free_table_info(newinfo);
1303 do_add_counters(struct net *net, const void __user *user,
1304 unsigned int len, int compat)
1306 unsigned int i, curcpu;
1307 struct xt_counters_info tmp;
1308 struct xt_counters *paddc;
1309 unsigned int num_counters;
1314 const struct xt_table_info *private;
1316 void *loc_cpu_entry;
1317 struct ipt_entry *iter;
1318 unsigned int addend;
1319 #ifdef CONFIG_COMPAT
1320 struct compat_xt_counters_info compat_tmp;
1324 size = sizeof(struct compat_xt_counters_info);
1329 size = sizeof(struct xt_counters_info);
1332 if (copy_from_user(ptmp, user, size) != 0)
1335 #ifdef CONFIG_COMPAT
1337 num_counters = compat_tmp.num_counters;
1338 name = compat_tmp.name;
1342 num_counters = tmp.num_counters;
1346 if (len != size + num_counters * sizeof(struct xt_counters))
1349 paddc = vmalloc(len - size);
1353 if (copy_from_user(paddc, user + size, len - size) != 0) {
1358 t = xt_find_table_lock(net, AF_INET, name);
1359 if (IS_ERR_OR_NULL(t)) {
1360 ret = t ? PTR_ERR(t) : -ENOENT;
1365 private = t->private;
1366 if (private->number != num_counters) {
1368 goto unlock_up_free;
1372 /* Choose the copy that is on our node */
1373 curcpu = smp_processor_id();
1374 loc_cpu_entry = private->entries[curcpu];
1375 addend = xt_write_recseq_begin();
1376 xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1377 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1380 xt_write_recseq_end(addend);
1391 #ifdef CONFIG_COMPAT
1392 struct compat_ipt_replace {
1393 char name[XT_TABLE_MAXNAMELEN];
1397 u32 hook_entry[NF_INET_NUMHOOKS];
1398 u32 underflow[NF_INET_NUMHOOKS];
1400 compat_uptr_t counters; /* struct xt_counters * */
1401 struct compat_ipt_entry entries[0];
1405 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1406 unsigned int *size, struct xt_counters *counters,
1409 struct xt_entry_target *t;
1410 struct compat_ipt_entry __user *ce;
1411 u_int16_t target_offset, next_offset;
1412 compat_uint_t origsize;
1413 const struct xt_entry_match *ematch;
1417 ce = (struct compat_ipt_entry __user *)*dstptr;
1418 if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1419 copy_to_user(&ce->counters, &counters[i],
1420 sizeof(counters[i])) != 0)
1423 *dstptr += sizeof(struct compat_ipt_entry);
1424 *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1426 xt_ematch_foreach(ematch, e) {
1427 ret = xt_compat_match_to_user(ematch, dstptr, size);
1431 target_offset = e->target_offset - (origsize - *size);
1432 t = ipt_get_target(e);
1433 ret = xt_compat_target_to_user(t, dstptr, size);
1436 next_offset = e->next_offset - (origsize - *size);
1437 if (put_user(target_offset, &ce->target_offset) != 0 ||
1438 put_user(next_offset, &ce->next_offset) != 0)
1444 compat_find_calc_match(struct xt_entry_match *m,
1446 const struct ipt_ip *ip,
1449 struct xt_match *match;
1451 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1452 m->u.user.revision);
1453 if (IS_ERR(match)) {
1454 duprintf("compat_check_calc_match: `%s' not found\n",
1456 return PTR_ERR(match);
1458 m->u.kernel.match = match;
1459 *size += xt_compat_match_offset(match);
1463 static void compat_release_entry(struct compat_ipt_entry *e)
1465 struct xt_entry_target *t;
1466 struct xt_entry_match *ematch;
1468 /* Cleanup all matches */
1469 xt_ematch_foreach(ematch, e)
1470 module_put(ematch->u.kernel.match->me);
1471 t = compat_ipt_get_target(e);
1472 module_put(t->u.kernel.target->me);
1476 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1477 struct xt_table_info *newinfo,
1479 const unsigned char *base,
1480 const unsigned char *limit,
1481 const unsigned int *hook_entries,
1482 const unsigned int *underflows,
1485 struct xt_entry_match *ematch;
1486 struct xt_entry_target *t;
1487 struct xt_target *target;
1488 unsigned int entry_offset;
1492 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1493 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1494 (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1495 duprintf("Bad offset %p, limit = %p\n", e, limit);
1499 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1500 sizeof(struct compat_xt_entry_target)) {
1501 duprintf("checking: element %p size %u\n",
1506 /* For purposes of check_entry casting the compat entry is fine */
1507 ret = check_entry((struct ipt_entry *)e, name);
1511 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1512 entry_offset = (void *)e - (void *)base;
1514 xt_ematch_foreach(ematch, e) {
1515 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1517 goto release_matches;
1521 t = compat_ipt_get_target(e);
1522 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1523 t->u.user.revision);
1524 if (IS_ERR(target)) {
1525 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1527 ret = PTR_ERR(target);
1528 goto release_matches;
1530 t->u.kernel.target = target;
1532 off += xt_compat_target_offset(target);
1534 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1538 /* Check hooks & underflows */
1539 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1540 if ((unsigned char *)e - base == hook_entries[h])
1541 newinfo->hook_entry[h] = hook_entries[h];
1542 if ((unsigned char *)e - base == underflows[h])
1543 newinfo->underflow[h] = underflows[h];
1546 /* Clear counters and comefrom */
1547 memset(&e->counters, 0, sizeof(e->counters));
1552 module_put(t->u.kernel.target->me);
1554 xt_ematch_foreach(ematch, e) {
1557 module_put(ematch->u.kernel.match->me);
1563 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1564 unsigned int *size, const char *name,
1565 struct xt_table_info *newinfo, unsigned char *base)
1567 struct xt_entry_target *t;
1568 struct xt_target *target;
1569 struct ipt_entry *de;
1570 unsigned int origsize;
1572 struct xt_entry_match *ematch;
1576 de = (struct ipt_entry *)*dstptr;
1577 memcpy(de, e, sizeof(struct ipt_entry));
1578 memcpy(&de->counters, &e->counters, sizeof(e->counters));
1580 *dstptr += sizeof(struct ipt_entry);
1581 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1583 xt_ematch_foreach(ematch, e) {
1584 ret = xt_compat_match_from_user(ematch, dstptr, size);
1588 de->target_offset = e->target_offset - (origsize - *size);
1589 t = compat_ipt_get_target(e);
1590 target = t->u.kernel.target;
1591 xt_compat_target_from_user(t, dstptr, size);
1593 de->next_offset = e->next_offset - (origsize - *size);
1594 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1595 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1596 newinfo->hook_entry[h] -= origsize - *size;
1597 if ((unsigned char *)de - base < newinfo->underflow[h])
1598 newinfo->underflow[h] -= origsize - *size;
1604 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1606 struct xt_entry_match *ematch;
1607 struct xt_mtchk_param mtpar;
1614 mtpar.entryinfo = &e->ip;
1615 mtpar.hook_mask = e->comefrom;
1616 mtpar.family = NFPROTO_IPV4;
1617 xt_ematch_foreach(ematch, e) {
1618 ret = check_match(ematch, &mtpar);
1620 goto cleanup_matches;
1624 ret = check_target(e, net, name);
1626 goto cleanup_matches;
1630 xt_ematch_foreach(ematch, e) {
1633 cleanup_match(ematch, net);
1639 translate_compat_table(struct net *net,
1641 unsigned int valid_hooks,
1642 struct xt_table_info **pinfo,
1644 unsigned int total_size,
1645 unsigned int number,
1646 unsigned int *hook_entries,
1647 unsigned int *underflows)
1650 struct xt_table_info *newinfo, *info;
1651 void *pos, *entry0, *entry1;
1652 struct compat_ipt_entry *iter0;
1653 struct ipt_entry *iter1;
1660 info->number = number;
1662 /* Init all hooks to impossible value. */
1663 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1664 info->hook_entry[i] = 0xFFFFFFFF;
1665 info->underflow[i] = 0xFFFFFFFF;
1668 duprintf("translate_compat_table: size %u\n", info->size);
1670 xt_compat_lock(AF_INET);
1671 xt_compat_init_offsets(AF_INET, number);
1672 /* Walk through entries, checking offsets. */
1673 xt_entry_foreach(iter0, entry0, total_size) {
1674 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1676 entry0 + total_size,
1687 duprintf("translate_compat_table: %u not %u entries\n",
1692 /* Check hooks all assigned */
1693 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1694 /* Only hooks which are valid */
1695 if (!(valid_hooks & (1 << i)))
1697 if (info->hook_entry[i] == 0xFFFFFFFF) {
1698 duprintf("Invalid hook entry %u %u\n",
1699 i, hook_entries[i]);
1702 if (info->underflow[i] == 0xFFFFFFFF) {
1703 duprintf("Invalid underflow %u %u\n",
1710 newinfo = xt_alloc_table_info(size);
1714 newinfo->number = number;
1715 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1716 newinfo->hook_entry[i] = info->hook_entry[i];
1717 newinfo->underflow[i] = info->underflow[i];
1719 entry1 = newinfo->entries[raw_smp_processor_id()];
1722 xt_entry_foreach(iter0, entry0, total_size) {
1723 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1724 name, newinfo, entry1);
1728 xt_compat_flush_offsets(AF_INET);
1729 xt_compat_unlock(AF_INET);
1734 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1738 xt_entry_foreach(iter1, entry1, newinfo->size) {
1739 ret = compat_check_entry(iter1, net, name);
1743 if (strcmp(ipt_get_target(iter1)->u.user.name,
1744 XT_ERROR_TARGET) == 0)
1745 ++newinfo->stacksize;
1749 * The first i matches need cleanup_entry (calls ->destroy)
1750 * because they had called ->check already. The other j-i
1751 * entries need only release.
1755 xt_entry_foreach(iter0, entry0, newinfo->size) {
1760 compat_release_entry(iter0);
1762 xt_entry_foreach(iter1, entry1, newinfo->size) {
1765 cleanup_entry(iter1, net);
1767 xt_free_table_info(newinfo);
1771 /* And one copy for every other CPU */
1772 for_each_possible_cpu(i)
1773 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1774 memcpy(newinfo->entries[i], entry1, newinfo->size);
1778 xt_free_table_info(info);
1782 xt_free_table_info(newinfo);
1784 xt_entry_foreach(iter0, entry0, total_size) {
1787 compat_release_entry(iter0);
1791 xt_compat_flush_offsets(AF_INET);
1792 xt_compat_unlock(AF_INET);
1797 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1800 struct compat_ipt_replace tmp;
1801 struct xt_table_info *newinfo;
1802 void *loc_cpu_entry;
1803 struct ipt_entry *iter;
1805 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1808 /* overflow check */
1809 if (tmp.size >= INT_MAX / num_possible_cpus())
1811 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1813 if (tmp.num_counters == 0)
1816 tmp.name[sizeof(tmp.name)-1] = 0;
1818 newinfo = xt_alloc_table_info(tmp.size);
1822 /* choose the copy that is on our node/cpu */
1823 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1824 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1830 ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1831 &newinfo, &loc_cpu_entry, tmp.size,
1832 tmp.num_entries, tmp.hook_entry,
1837 duprintf("compat_do_replace: Translated table\n");
1839 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1840 tmp.num_counters, compat_ptr(tmp.counters));
1842 goto free_newinfo_untrans;
1845 free_newinfo_untrans:
1846 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1847 cleanup_entry(iter, net);
1849 xt_free_table_info(newinfo);
1854 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1859 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1863 case IPT_SO_SET_REPLACE:
1864 ret = compat_do_replace(sock_net(sk), user, len);
1867 case IPT_SO_SET_ADD_COUNTERS:
1868 ret = do_add_counters(sock_net(sk), user, len, 1);
1872 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1879 struct compat_ipt_get_entries {
1880 char name[XT_TABLE_MAXNAMELEN];
1882 struct compat_ipt_entry entrytable[0];
1886 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1887 void __user *userptr)
1889 struct xt_counters *counters;
1890 const struct xt_table_info *private = table->private;
1894 const void *loc_cpu_entry;
1896 struct ipt_entry *iter;
1898 counters = alloc_counters(table);
1899 if (IS_ERR(counters))
1900 return PTR_ERR(counters);
1902 /* choose the copy that is on our node/cpu, ...
1903 * This choice is lazy (because current thread is
1904 * allowed to migrate to another cpu)
1906 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1909 xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1910 ret = compat_copy_entry_to_user(iter, &pos,
1911 &size, counters, i++);
1921 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1925 struct compat_ipt_get_entries get;
1928 if (*len < sizeof(get)) {
1929 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1933 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1936 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1937 duprintf("compat_get_entries: %u != %zu\n",
1938 *len, sizeof(get) + get.size);
1942 xt_compat_lock(AF_INET);
1943 t = xt_find_table_lock(net, AF_INET, get.name);
1944 if (!IS_ERR_OR_NULL(t)) {
1945 const struct xt_table_info *private = t->private;
1946 struct xt_table_info info;
1947 duprintf("t->private->number = %u\n", private->number);
1948 ret = compat_table_info(private, &info);
1949 if (!ret && get.size == info.size) {
1950 ret = compat_copy_entries_to_user(private->size,
1951 t, uptr->entrytable);
1953 duprintf("compat_get_entries: I've got %u not %u!\n",
1954 private->size, get.size);
1957 xt_compat_flush_offsets(AF_INET);
1961 ret = t ? PTR_ERR(t) : -ENOENT;
1963 xt_compat_unlock(AF_INET);
1967 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1970 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1974 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1978 case IPT_SO_GET_INFO:
1979 ret = get_info(sock_net(sk), user, len, 1);
1981 case IPT_SO_GET_ENTRIES:
1982 ret = compat_get_entries(sock_net(sk), user, len);
1985 ret = do_ipt_get_ctl(sk, cmd, user, len);
1992 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1996 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2000 case IPT_SO_SET_REPLACE:
2001 ret = do_replace(sock_net(sk), user, len);
2004 case IPT_SO_SET_ADD_COUNTERS:
2005 ret = do_add_counters(sock_net(sk), user, len, 0);
2009 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
2017 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2021 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2025 case IPT_SO_GET_INFO:
2026 ret = get_info(sock_net(sk), user, len, 0);
2029 case IPT_SO_GET_ENTRIES:
2030 ret = get_entries(sock_net(sk), user, len);
2033 case IPT_SO_GET_REVISION_MATCH:
2034 case IPT_SO_GET_REVISION_TARGET: {
2035 struct xt_get_revision rev;
2038 if (*len != sizeof(rev)) {
2042 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2046 rev.name[sizeof(rev.name)-1] = 0;
2048 if (cmd == IPT_SO_GET_REVISION_TARGET)
2053 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2056 "ipt_%s", rev.name);
2061 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2068 struct xt_table *ipt_register_table(struct net *net,
2069 const struct xt_table *table,
2070 const struct ipt_replace *repl)
2073 struct xt_table_info *newinfo;
2074 struct xt_table_info bootstrap = {0};
2075 void *loc_cpu_entry;
2076 struct xt_table *new_table;
2078 newinfo = xt_alloc_table_info(repl->size);
2084 /* choose the copy on our node/cpu, but dont care about preemption */
2085 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2086 memcpy(loc_cpu_entry, repl->entries, repl->size);
2088 ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2092 new_table = xt_register_table(net, table, &bootstrap, newinfo);
2093 if (IS_ERR(new_table)) {
2094 ret = PTR_ERR(new_table);
2101 xt_free_table_info(newinfo);
2103 return ERR_PTR(ret);
2106 void ipt_unregister_table(struct net *net, struct xt_table *table)
2108 struct xt_table_info *private;
2109 void *loc_cpu_entry;
2110 struct module *table_owner = table->me;
2111 struct ipt_entry *iter;
2113 private = xt_unregister_table(table);
2115 /* Decrease module usage counts and free resources */
2116 loc_cpu_entry = private->entries[raw_smp_processor_id()];
2117 xt_entry_foreach(iter, loc_cpu_entry, private->size)
2118 cleanup_entry(iter, net);
2119 if (private->number > private->initial_entries)
2120 module_put(table_owner);
2121 xt_free_table_info(private);
2124 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2126 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2127 u_int8_t type, u_int8_t code,
2130 return ((test_type == 0xFF) ||
2131 (type == test_type && code >= min_code && code <= max_code))
2136 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2138 const struct icmphdr *ic;
2139 struct icmphdr _icmph;
2140 const struct ipt_icmp *icmpinfo = par->matchinfo;
2142 /* Must not be a fragment. */
2143 if (par->fragoff != 0)
2146 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2148 /* We've been asked to examine this packet, and we
2149 * can't. Hence, no choice but to drop.
2151 duprintf("Dropping evil ICMP tinygram.\n");
2152 par->hotdrop = true;
2156 return icmp_type_code_match(icmpinfo->type,
2160 !!(icmpinfo->invflags&IPT_ICMP_INV));
2163 static int icmp_checkentry(const struct xt_mtchk_param *par)
2165 const struct ipt_icmp *icmpinfo = par->matchinfo;
2167 /* Must specify no unknown invflags */
2168 return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2171 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2173 .name = XT_STANDARD_TARGET,
2174 .targetsize = sizeof(int),
2175 .family = NFPROTO_IPV4,
2176 #ifdef CONFIG_COMPAT
2177 .compatsize = sizeof(compat_int_t),
2178 .compat_from_user = compat_standard_from_user,
2179 .compat_to_user = compat_standard_to_user,
2183 .name = XT_ERROR_TARGET,
2184 .target = ipt_error,
2185 .targetsize = XT_FUNCTION_MAXNAMELEN,
2186 .family = NFPROTO_IPV4,
2190 static struct nf_sockopt_ops ipt_sockopts = {
2192 .set_optmin = IPT_BASE_CTL,
2193 .set_optmax = IPT_SO_SET_MAX+1,
2194 .set = do_ipt_set_ctl,
2195 #ifdef CONFIG_COMPAT
2196 .compat_set = compat_do_ipt_set_ctl,
2198 .get_optmin = IPT_BASE_CTL,
2199 .get_optmax = IPT_SO_GET_MAX+1,
2200 .get = do_ipt_get_ctl,
2201 #ifdef CONFIG_COMPAT
2202 .compat_get = compat_do_ipt_get_ctl,
2204 .owner = THIS_MODULE,
2207 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2210 .match = icmp_match,
2211 .matchsize = sizeof(struct ipt_icmp),
2212 .checkentry = icmp_checkentry,
2213 .proto = IPPROTO_ICMP,
2214 .family = NFPROTO_IPV4,
2218 static int __net_init ip_tables_net_init(struct net *net)
2220 return xt_proto_init(net, NFPROTO_IPV4);
2223 static void __net_exit ip_tables_net_exit(struct net *net)
2225 xt_proto_fini(net, NFPROTO_IPV4);
2228 static struct pernet_operations ip_tables_net_ops = {
2229 .init = ip_tables_net_init,
2230 .exit = ip_tables_net_exit,
2233 static int __init ip_tables_init(void)
2237 ret = register_pernet_subsys(&ip_tables_net_ops);
2241 /* No one else will be downing sem now, so we won't sleep */
2242 ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2245 ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2249 /* Register setsockopt */
2250 ret = nf_register_sockopt(&ipt_sockopts);
2254 pr_info("(C) 2000-2006 Netfilter Core Team\n");
2258 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2260 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2262 unregister_pernet_subsys(&ip_tables_net_ops);
2267 static void __exit ip_tables_fini(void)
2269 nf_unregister_sockopt(&ipt_sockopts);
2271 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2272 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2273 unregister_pernet_subsys(&ip_tables_net_ops);
2276 EXPORT_SYMBOL(ipt_register_table);
2277 EXPORT_SYMBOL(ipt_unregister_table);
2278 EXPORT_SYMBOL(ipt_do_table);
2279 module_init(ip_tables_init);
2280 module_exit(ip_tables_fini);