2 * Packet matching code.
4 * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
5 * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12 #include <linux/cache.h>
13 #include <linux/capability.h>
14 #include <linux/skbuff.h>
15 #include <linux/kmod.h>
16 #include <linux/vmalloc.h>
17 #include <linux/netdevice.h>
18 #include <linux/module.h>
19 #include <linux/icmp.h>
21 #include <net/compat.h>
22 #include <asm/uaccess.h>
23 #include <linux/mutex.h>
24 #include <linux/proc_fs.h>
25 #include <linux/err.h>
26 #include <linux/cpumask.h>
28 #include <linux/netfilter/x_tables.h>
29 #include <linux/netfilter_ipv4/ip_tables.h>
30 #include <net/netfilter/nf_log.h>
31 #include "../../netfilter/xt_repldata.h"
33 MODULE_LICENSE("GPL");
34 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
35 MODULE_DESCRIPTION("IPv4 packet filter");
37 /*#define DEBUG_IP_FIREWALL*/
38 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
39 /*#define DEBUG_IP_FIREWALL_USER*/
41 #ifdef DEBUG_IP_FIREWALL
42 #define dprintf(format, args...) pr_info(format , ## args)
44 #define dprintf(format, args...)
47 #ifdef DEBUG_IP_FIREWALL_USER
48 #define duprintf(format, args...) pr_info(format , ## args)
50 #define duprintf(format, args...)
53 #ifdef CONFIG_NETFILTER_DEBUG
54 #define IP_NF_ASSERT(x) WARN_ON(!(x))
56 #define IP_NF_ASSERT(x)
60 /* All the better to debug you with... */
65 void *ipt_alloc_initial_table(const struct xt_table *info)
67 return xt_alloc_initial_table(ipt, IPT);
69 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
72 We keep a set of rules for each CPU, so we can avoid write-locking
73 them in the softirq when updating the counters and therefore
74 only need to read-lock in the softirq; doing a write_lock_bh() in user
75 context stops packets coming through and allows user context to read
76 the counters or update the rules.
78 Hence the start of any table is given by get_table() below. */
80 /* Returns whether matches rule or not. */
81 /* Performance critical - called for every packet */
83 ip_packet_match(const struct iphdr *ip,
86 const struct ipt_ip *ipinfo,
91 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
93 if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
95 FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
97 dprintf("Source or dest mismatch.\n");
99 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
100 &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
101 ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
102 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
103 &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
104 ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
108 ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
110 if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
111 dprintf("VIA in mismatch (%s vs %s).%s\n",
112 indev, ipinfo->iniface,
113 ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
117 ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
119 if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
120 dprintf("VIA out mismatch (%s vs %s).%s\n",
121 outdev, ipinfo->outiface,
122 ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
126 /* Check specific protocol */
128 FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
129 dprintf("Packet protocol %hi does not match %hi.%s\n",
130 ip->protocol, ipinfo->proto,
131 ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
135 /* If we have a fragment rule but the packet is not a fragment
136 * then we return zero */
137 if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
138 dprintf("Fragment rule but not fragment.%s\n",
139 ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
147 ip_checkentry(const struct ipt_ip *ip)
149 if (ip->flags & ~IPT_F_MASK) {
150 duprintf("Unknown flag bits set: %08X\n",
151 ip->flags & ~IPT_F_MASK);
154 if (ip->invflags & ~IPT_INV_MASK) {
155 duprintf("Unknown invflag bits set: %08X\n",
156 ip->invflags & ~IPT_INV_MASK);
163 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
166 pr_info("error: `%s'\n", (const char *)par->targinfo);
171 /* Performance critical */
172 static inline struct ipt_entry *
173 get_entry(const void *base, unsigned int offset)
175 return (struct ipt_entry *)(base + offset);
178 /* All zeroes == unconditional rule. */
179 /* Mildly perf critical (only if packet tracing is on) */
180 static inline bool unconditional(const struct ipt_ip *ip)
182 static const struct ipt_ip uncond;
184 return memcmp(ip, &uncond, sizeof(uncond)) == 0;
188 /* for const-correctness */
189 static inline const struct xt_entry_target *
190 ipt_get_target_c(const struct ipt_entry *e)
192 return ipt_get_target((struct ipt_entry *)e);
195 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
196 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
197 static const char *const hooknames[] = {
198 [NF_INET_PRE_ROUTING] = "PREROUTING",
199 [NF_INET_LOCAL_IN] = "INPUT",
200 [NF_INET_FORWARD] = "FORWARD",
201 [NF_INET_LOCAL_OUT] = "OUTPUT",
202 [NF_INET_POST_ROUTING] = "POSTROUTING",
205 enum nf_ip_trace_comments {
206 NF_IP_TRACE_COMMENT_RULE,
207 NF_IP_TRACE_COMMENT_RETURN,
208 NF_IP_TRACE_COMMENT_POLICY,
211 static const char *const comments[] = {
212 [NF_IP_TRACE_COMMENT_RULE] = "rule",
213 [NF_IP_TRACE_COMMENT_RETURN] = "return",
214 [NF_IP_TRACE_COMMENT_POLICY] = "policy",
217 static struct nf_loginfo trace_loginfo = {
218 .type = NF_LOG_TYPE_LOG,
222 .logflags = NF_LOG_MASK,
227 /* Mildly perf critical (only if packet tracing is on) */
229 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
230 const char *hookname, const char **chainname,
231 const char **comment, unsigned int *rulenum)
233 const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
235 if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
236 /* Head of user chain: ERROR target with chainname */
237 *chainname = t->target.data;
242 if (s->target_offset == sizeof(struct ipt_entry) &&
243 strcmp(t->target.u.kernel.target->name,
244 XT_STANDARD_TARGET) == 0 &&
246 unconditional(&s->ip)) {
247 /* Tail of chains: STANDARD target (return/policy) */
248 *comment = *chainname == hookname
249 ? comments[NF_IP_TRACE_COMMENT_POLICY]
250 : comments[NF_IP_TRACE_COMMENT_RETURN];
259 static void trace_packet(const struct sk_buff *skb,
261 const struct net_device *in,
262 const struct net_device *out,
263 const char *tablename,
264 const struct xt_table_info *private,
265 const struct ipt_entry *e)
267 const void *table_base;
268 const struct ipt_entry *root;
269 const char *hookname, *chainname, *comment;
270 const struct ipt_entry *iter;
271 unsigned int rulenum = 0;
273 table_base = private->entries[smp_processor_id()];
274 root = get_entry(table_base, private->hook_entry[hook]);
276 hookname = chainname = hooknames[hook];
277 comment = comments[NF_IP_TRACE_COMMENT_RULE];
279 xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
280 if (get_chainname_rulenum(iter, e, hookname,
281 &chainname, &comment, &rulenum) != 0)
284 nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
285 "TRACE: %s:%s:%s:%u ",
286 tablename, chainname, comment, rulenum);
291 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
293 return (void *)entry + entry->next_offset;
296 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
298 ipt_do_table(struct sk_buff *skb,
300 const struct net_device *in,
301 const struct net_device *out,
302 struct xt_table *table)
304 static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
305 const struct iphdr *ip;
306 /* Initializing verdict to NF_DROP keeps gcc happy. */
307 unsigned int verdict = NF_DROP;
308 const char *indev, *outdev;
309 const void *table_base;
310 struct ipt_entry *e, **jumpstack;
311 unsigned int *stackptr, origptr, cpu;
312 const struct xt_table_info *private;
313 struct xt_action_param acpar;
317 indev = in ? in->name : nulldevname;
318 outdev = out ? out->name : nulldevname;
319 /* We handle fragments by dealing with the first fragment as
320 * if it was a normal packet. All other fragments are treated
321 * normally, except that they will NEVER match rules that ask
322 * things we don't know, ie. tcp syn flag or ports). If the
323 * rule is also a fragment-specific rule, non-fragments won't
325 acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
326 acpar.thoff = ip_hdrlen(skb);
327 acpar.hotdrop = false;
330 acpar.family = NFPROTO_IPV4;
331 acpar.hooknum = hook;
333 IP_NF_ASSERT(table->valid_hooks & (1 << hook));
335 private = table->private;
336 cpu = smp_processor_id();
337 table_base = private->entries[cpu];
338 jumpstack = (struct ipt_entry **)private->jumpstack[cpu];
339 stackptr = per_cpu_ptr(private->stackptr, cpu);
342 e = get_entry(table_base, private->hook_entry[hook]);
344 pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
345 table->name, hook, origptr,
346 get_entry(table_base, private->underflow[hook]));
349 const struct xt_entry_target *t;
350 const struct xt_entry_match *ematch;
353 if (!ip_packet_match(ip, indev, outdev,
354 &e->ip, acpar.fragoff)) {
356 e = ipt_next_entry(e);
360 xt_ematch_foreach(ematch, e) {
361 acpar.match = ematch->u.kernel.match;
362 acpar.matchinfo = ematch->data;
363 if (!acpar.match->match(skb, &acpar))
367 ADD_COUNTER(e->counters, skb->len, 1);
369 t = ipt_get_target(e);
370 IP_NF_ASSERT(t->u.kernel.target);
372 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
373 defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
374 /* The packet is traced: log it */
375 if (unlikely(skb->nf_trace))
376 trace_packet(skb, hook, in, out,
377 table->name, private, e);
379 /* Standard target? */
380 if (!t->u.kernel.target->target) {
383 v = ((struct xt_standard_target *)t)->verdict;
385 /* Pop from stack? */
386 if (v != XT_RETURN) {
387 verdict = (unsigned)(-v) - 1;
390 if (*stackptr == 0) {
391 e = get_entry(table_base,
392 private->underflow[hook]);
393 pr_debug("Underflow (this is normal) "
396 e = jumpstack[--*stackptr];
397 pr_debug("Pulled %p out from pos %u\n",
399 e = ipt_next_entry(e);
403 if (table_base + v != ipt_next_entry(e) &&
404 !(e->ip.flags & IPT_F_GOTO)) {
405 if (*stackptr >= private->stacksize) {
409 jumpstack[(*stackptr)++] = 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 xt_info_rdunlock_bh();
431 pr_debug("Exiting %s; resetting sp from %u to %u\n",
432 __func__, *stackptr, origptr);
434 #ifdef DEBUG_ALLOW_ALL
443 /* Figures out from what hook each rule can be called: returns 0 if
444 there are loops. Puts hook bitmask in comefrom. */
446 mark_source_chains(const struct xt_table_info *newinfo,
447 unsigned int valid_hooks, void *entry0)
451 /* No recursion; use packet counter to save back ptrs (reset
452 to 0 as we leave), and comefrom to save source hook bitmask */
453 for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
454 unsigned int pos = newinfo->hook_entry[hook];
455 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
457 if (!(valid_hooks & (1 << hook)))
460 /* Set initial back pointer. */
461 e->counters.pcnt = pos;
464 const struct xt_standard_target *t
465 = (void *)ipt_get_target_c(e);
466 int visited = e->comefrom & (1 << hook);
468 if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
469 pr_err("iptables: loop hook %u pos %u %08X.\n",
470 hook, pos, e->comefrom);
473 e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
475 /* Unconditional return/END. */
476 if ((e->target_offset == sizeof(struct ipt_entry) &&
477 (strcmp(t->target.u.user.name,
478 XT_STANDARD_TARGET) == 0) &&
479 t->verdict < 0 && unconditional(&e->ip)) ||
481 unsigned int oldpos, size;
483 if ((strcmp(t->target.u.user.name,
484 XT_STANDARD_TARGET) == 0) &&
485 t->verdict < -NF_MAX_VERDICT - 1) {
486 duprintf("mark_source_chains: bad "
487 "negative verdict (%i)\n",
492 /* Return: backtrack through the last
495 e->comefrom ^= (1<<NF_INET_NUMHOOKS);
496 #ifdef DEBUG_IP_FIREWALL_USER
498 & (1 << NF_INET_NUMHOOKS)) {
499 duprintf("Back unset "
506 pos = e->counters.pcnt;
507 e->counters.pcnt = 0;
509 /* We're at the start. */
513 e = (struct ipt_entry *)
515 } while (oldpos == pos + e->next_offset);
518 size = e->next_offset;
519 e = (struct ipt_entry *)
520 (entry0 + pos + size);
521 e->counters.pcnt = pos;
524 int newpos = t->verdict;
526 if (strcmp(t->target.u.user.name,
527 XT_STANDARD_TARGET) == 0 &&
529 if (newpos > newinfo->size -
530 sizeof(struct ipt_entry)) {
531 duprintf("mark_source_chains: "
532 "bad verdict (%i)\n",
536 /* This a jump; chase it. */
537 duprintf("Jump rule %u -> %u\n",
540 /* ... this is a fallthru */
541 newpos = pos + e->next_offset;
543 e = (struct ipt_entry *)
545 e->counters.pcnt = pos;
550 duprintf("Finished chain %u\n", hook);
555 static void cleanup_match(struct xt_entry_match *m, struct net *net)
557 struct xt_mtdtor_param par;
560 par.match = m->u.kernel.match;
561 par.matchinfo = m->data;
562 par.family = NFPROTO_IPV4;
563 if (par.match->destroy != NULL)
564 par.match->destroy(&par);
565 module_put(par.match->me);
569 check_entry(const struct ipt_entry *e, const char *name)
571 const struct xt_entry_target *t;
573 if (!ip_checkentry(&e->ip)) {
574 duprintf("ip check failed %p %s.\n", e, par->match->name);
578 if (e->target_offset + sizeof(struct xt_entry_target) >
582 t = ipt_get_target_c(e);
583 if (e->target_offset + t->u.target_size > e->next_offset)
590 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
592 const struct ipt_ip *ip = par->entryinfo;
595 par->match = m->u.kernel.match;
596 par->matchinfo = m->data;
598 ret = xt_check_match(par, m->u.match_size - sizeof(*m),
599 ip->proto, ip->invflags & IPT_INV_PROTO);
601 duprintf("check failed for `%s'.\n", par->match->name);
608 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
610 struct xt_match *match;
613 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
616 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
617 return PTR_ERR(match);
619 m->u.kernel.match = match;
621 ret = check_match(m, par);
627 module_put(m->u.kernel.match->me);
631 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
633 struct xt_entry_target *t = ipt_get_target(e);
634 struct xt_tgchk_param par = {
638 .target = t->u.kernel.target,
640 .hook_mask = e->comefrom,
641 .family = NFPROTO_IPV4,
645 ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
646 e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
648 duprintf("check failed for `%s'.\n",
649 t->u.kernel.target->name);
656 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
659 struct xt_entry_target *t;
660 struct xt_target *target;
663 struct xt_mtchk_param mtpar;
664 struct xt_entry_match *ematch;
666 ret = check_entry(e, name);
673 mtpar.entryinfo = &e->ip;
674 mtpar.hook_mask = e->comefrom;
675 mtpar.family = NFPROTO_IPV4;
676 xt_ematch_foreach(ematch, e) {
677 ret = find_check_match(ematch, &mtpar);
679 goto cleanup_matches;
683 t = ipt_get_target(e);
684 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
686 if (IS_ERR(target)) {
687 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
688 ret = PTR_ERR(target);
689 goto cleanup_matches;
691 t->u.kernel.target = target;
693 ret = check_target(e, net, name);
698 module_put(t->u.kernel.target->me);
700 xt_ematch_foreach(ematch, e) {
703 cleanup_match(ematch, net);
708 static bool check_underflow(const struct ipt_entry *e)
710 const struct xt_entry_target *t;
711 unsigned int verdict;
713 if (!unconditional(&e->ip))
715 t = ipt_get_target_c(e);
716 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
718 verdict = ((struct xt_standard_target *)t)->verdict;
719 verdict = -verdict - 1;
720 return verdict == NF_DROP || verdict == NF_ACCEPT;
724 check_entry_size_and_hooks(struct ipt_entry *e,
725 struct xt_table_info *newinfo,
726 const unsigned char *base,
727 const unsigned char *limit,
728 const unsigned int *hook_entries,
729 const unsigned int *underflows,
730 unsigned int valid_hooks)
734 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
735 (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
736 duprintf("Bad offset %p\n", e);
741 < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
742 duprintf("checking: element %p size %u\n",
747 /* Check hooks & underflows */
748 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
749 if (!(valid_hooks & (1 << h)))
751 if ((unsigned char *)e - base == hook_entries[h])
752 newinfo->hook_entry[h] = hook_entries[h];
753 if ((unsigned char *)e - base == underflows[h]) {
754 if (!check_underflow(e)) {
755 pr_err("Underflows must be unconditional and "
756 "use the STANDARD target with "
760 newinfo->underflow[h] = underflows[h];
764 /* Clear counters and comefrom */
765 e->counters = ((struct xt_counters) { 0, 0 });
771 cleanup_entry(struct ipt_entry *e, struct net *net)
773 struct xt_tgdtor_param par;
774 struct xt_entry_target *t;
775 struct xt_entry_match *ematch;
777 /* Cleanup all matches */
778 xt_ematch_foreach(ematch, e)
779 cleanup_match(ematch, net);
780 t = ipt_get_target(e);
783 par.target = t->u.kernel.target;
784 par.targinfo = t->data;
785 par.family = NFPROTO_IPV4;
786 if (par.target->destroy != NULL)
787 par.target->destroy(&par);
788 module_put(par.target->me);
791 /* Checks and translates the user-supplied table segment (held in
794 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
795 const struct ipt_replace *repl)
797 struct ipt_entry *iter;
801 newinfo->size = repl->size;
802 newinfo->number = repl->num_entries;
804 /* Init all hooks to impossible value. */
805 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
806 newinfo->hook_entry[i] = 0xFFFFFFFF;
807 newinfo->underflow[i] = 0xFFFFFFFF;
810 duprintf("translate_table: size %u\n", newinfo->size);
812 /* Walk through entries, checking offsets. */
813 xt_entry_foreach(iter, entry0, newinfo->size) {
814 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
822 if (strcmp(ipt_get_target(iter)->u.user.name,
823 XT_ERROR_TARGET) == 0)
824 ++newinfo->stacksize;
827 if (i != repl->num_entries) {
828 duprintf("translate_table: %u not %u entries\n",
829 i, repl->num_entries);
833 /* Check hooks all assigned */
834 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
835 /* Only hooks which are valid */
836 if (!(repl->valid_hooks & (1 << i)))
838 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
839 duprintf("Invalid hook entry %u %u\n",
840 i, repl->hook_entry[i]);
843 if (newinfo->underflow[i] == 0xFFFFFFFF) {
844 duprintf("Invalid underflow %u %u\n",
845 i, repl->underflow[i]);
850 if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
853 /* Finally, each sanity check must pass */
855 xt_entry_foreach(iter, entry0, newinfo->size) {
856 ret = find_check_entry(iter, net, repl->name, repl->size);
863 xt_entry_foreach(iter, entry0, newinfo->size) {
866 cleanup_entry(iter, net);
871 /* And one copy for every other CPU */
872 for_each_possible_cpu(i) {
873 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
874 memcpy(newinfo->entries[i], entry0, newinfo->size);
881 get_counters(const struct xt_table_info *t,
882 struct xt_counters counters[])
884 struct ipt_entry *iter;
887 unsigned int curcpu = get_cpu();
889 /* Instead of clearing (by a previous call to memset())
890 * the counters and using adds, we set the counters
891 * with data used by 'current' CPU.
893 * Bottom half has to be disabled to prevent deadlock
894 * if new softirq were to run and call ipt_do_table
898 xt_entry_foreach(iter, t->entries[curcpu], t->size) {
899 SET_COUNTER(counters[i], iter->counters.bcnt,
900 iter->counters.pcnt);
904 /* Processing counters from other cpus, we can let bottom half enabled,
905 * (preemption is disabled)
908 for_each_possible_cpu(cpu) {
914 xt_entry_foreach(iter, t->entries[cpu], t->size) {
915 ADD_COUNTER(counters[i], iter->counters.bcnt,
916 iter->counters.pcnt);
917 ++i; /* macro does multi eval of i */
919 xt_info_wrunlock(cpu);
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 = vmalloc(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 /* choose the copy that is on our node/cpu, ...
962 * This choice is lazy (because current thread is
963 * allowed to migrate to another cpu)
965 loc_cpu_entry = private->entries[raw_smp_processor_id()];
966 if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
971 /* FIXME: use iterator macros --RR */
972 /* ... then go back and fix counters and names */
973 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
975 const struct xt_entry_match *m;
976 const struct xt_entry_target *t;
978 e = (struct ipt_entry *)(loc_cpu_entry + off);
979 if (copy_to_user(userptr + off
980 + offsetof(struct ipt_entry, counters),
982 sizeof(counters[num])) != 0) {
987 for (i = sizeof(struct ipt_entry);
988 i < e->target_offset;
989 i += m->u.match_size) {
992 if (copy_to_user(userptr + off + i
993 + offsetof(struct xt_entry_match,
995 m->u.kernel.match->name,
996 strlen(m->u.kernel.match->name)+1)
1003 t = ipt_get_target_c(e);
1004 if (copy_to_user(userptr + off + e->target_offset
1005 + offsetof(struct xt_entry_target,
1007 t->u.kernel.target->name,
1008 strlen(t->u.kernel.target->name)+1) != 0) {
1019 #ifdef CONFIG_COMPAT
1020 static void compat_standard_from_user(void *dst, const void *src)
1022 int v = *(compat_int_t *)src;
1025 v += xt_compat_calc_jump(AF_INET, v);
1026 memcpy(dst, &v, sizeof(v));
1029 static int compat_standard_to_user(void __user *dst, const void *src)
1031 compat_int_t cv = *(int *)src;
1034 cv -= xt_compat_calc_jump(AF_INET, cv);
1035 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1038 static int compat_calc_entry(const struct ipt_entry *e,
1039 const struct xt_table_info *info,
1040 const void *base, struct xt_table_info *newinfo)
1042 const struct xt_entry_match *ematch;
1043 const struct xt_entry_target *t;
1044 unsigned int entry_offset;
1047 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1048 entry_offset = (void *)e - base;
1049 xt_ematch_foreach(ematch, e)
1050 off += xt_compat_match_offset(ematch->u.kernel.match);
1051 t = ipt_get_target_c(e);
1052 off += xt_compat_target_offset(t->u.kernel.target);
1053 newinfo->size -= off;
1054 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1058 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1059 if (info->hook_entry[i] &&
1060 (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1061 newinfo->hook_entry[i] -= off;
1062 if (info->underflow[i] &&
1063 (e < (struct ipt_entry *)(base + info->underflow[i])))
1064 newinfo->underflow[i] -= off;
1069 static int compat_table_info(const struct xt_table_info *info,
1070 struct xt_table_info *newinfo)
1072 struct ipt_entry *iter;
1073 void *loc_cpu_entry;
1076 if (!newinfo || !info)
1079 /* we dont care about newinfo->entries[] */
1080 memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1081 newinfo->initial_entries = 0;
1082 loc_cpu_entry = info->entries[raw_smp_processor_id()];
1083 xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1084 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1092 static int get_info(struct net *net, void __user *user,
1093 const int *len, int compat)
1095 char name[XT_TABLE_MAXNAMELEN];
1099 if (*len != sizeof(struct ipt_getinfo)) {
1100 duprintf("length %u != %zu\n", *len,
1101 sizeof(struct ipt_getinfo));
1105 if (copy_from_user(name, user, sizeof(name)) != 0)
1108 name[XT_TABLE_MAXNAMELEN-1] = '\0';
1109 #ifdef CONFIG_COMPAT
1111 xt_compat_lock(AF_INET);
1113 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1114 "iptable_%s", name);
1115 if (t && !IS_ERR(t)) {
1116 struct ipt_getinfo info;
1117 const struct xt_table_info *private = t->private;
1118 #ifdef CONFIG_COMPAT
1119 struct xt_table_info tmp;
1122 ret = compat_table_info(private, &tmp);
1123 xt_compat_flush_offsets(AF_INET);
1127 memset(&info, 0, sizeof(info));
1128 info.valid_hooks = t->valid_hooks;
1129 memcpy(info.hook_entry, private->hook_entry,
1130 sizeof(info.hook_entry));
1131 memcpy(info.underflow, private->underflow,
1132 sizeof(info.underflow));
1133 info.num_entries = private->number;
1134 info.size = private->size;
1135 strcpy(info.name, name);
1137 if (copy_to_user(user, &info, *len) != 0)
1145 ret = t ? PTR_ERR(t) : -ENOENT;
1146 #ifdef CONFIG_COMPAT
1148 xt_compat_unlock(AF_INET);
1154 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1158 struct ipt_get_entries get;
1161 if (*len < sizeof(get)) {
1162 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1165 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1167 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1168 duprintf("get_entries: %u != %zu\n",
1169 *len, sizeof(get) + get.size);
1173 t = xt_find_table_lock(net, AF_INET, get.name);
1174 if (t && !IS_ERR(t)) {
1175 const struct xt_table_info *private = t->private;
1176 duprintf("t->private->number = %u\n", private->number);
1177 if (get.size == private->size)
1178 ret = copy_entries_to_user(private->size,
1179 t, uptr->entrytable);
1181 duprintf("get_entries: I've got %u not %u!\n",
1182 private->size, get.size);
1188 ret = t ? PTR_ERR(t) : -ENOENT;
1194 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1195 struct xt_table_info *newinfo, unsigned int num_counters,
1196 void __user *counters_ptr)
1200 struct xt_table_info *oldinfo;
1201 struct xt_counters *counters;
1202 void *loc_cpu_old_entry;
1203 struct ipt_entry *iter;
1206 counters = vmalloc(num_counters * sizeof(struct xt_counters));
1212 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1213 "iptable_%s", name);
1214 if (!t || IS_ERR(t)) {
1215 ret = t ? PTR_ERR(t) : -ENOENT;
1216 goto free_newinfo_counters_untrans;
1220 if (valid_hooks != t->valid_hooks) {
1221 duprintf("Valid hook crap: %08X vs %08X\n",
1222 valid_hooks, t->valid_hooks);
1227 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1231 /* Update module usage count based on number of rules */
1232 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1233 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1234 if ((oldinfo->number > oldinfo->initial_entries) ||
1235 (newinfo->number <= oldinfo->initial_entries))
1237 if ((oldinfo->number > oldinfo->initial_entries) &&
1238 (newinfo->number <= oldinfo->initial_entries))
1241 /* Get the old counters, and synchronize with replace */
1242 get_counters(oldinfo, counters);
1244 /* Decrease module usage counts and free resource */
1245 loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1246 xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1247 cleanup_entry(iter, net);
1249 xt_free_table_info(oldinfo);
1250 if (copy_to_user(counters_ptr, counters,
1251 sizeof(struct xt_counters) * num_counters) != 0)
1260 free_newinfo_counters_untrans:
1267 do_replace(struct net *net, const void __user *user, unsigned int len)
1270 struct ipt_replace tmp;
1271 struct xt_table_info *newinfo;
1272 void *loc_cpu_entry;
1273 struct ipt_entry *iter;
1275 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1278 /* overflow check */
1279 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1282 newinfo = xt_alloc_table_info(tmp.size);
1286 /* choose the copy that is on our node/cpu */
1287 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
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)
1318 unsigned int i, curcpu;
1319 struct xt_counters_info tmp;
1320 struct xt_counters *paddc;
1321 unsigned int num_counters;
1326 const struct xt_table_info *private;
1328 void *loc_cpu_entry;
1329 struct ipt_entry *iter;
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 (!t || IS_ERR(t)) {
1371 ret = t ? PTR_ERR(t) : -ENOENT;
1376 private = t->private;
1377 if (private->number != num_counters) {
1379 goto unlock_up_free;
1383 /* Choose the copy that is on our node */
1384 curcpu = smp_processor_id();
1385 loc_cpu_entry = private->entries[curcpu];
1386 xt_info_wrlock(curcpu);
1387 xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1388 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1391 xt_info_wrunlock(curcpu);
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,
1458 unsigned int hookmask,
1461 struct xt_match *match;
1463 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1464 m->u.user.revision);
1465 if (IS_ERR(match)) {
1466 duprintf("compat_check_calc_match: `%s' not found\n",
1468 return PTR_ERR(match);
1470 m->u.kernel.match = match;
1471 *size += xt_compat_match_offset(match);
1475 static void compat_release_entry(struct compat_ipt_entry *e)
1477 struct xt_entry_target *t;
1478 struct xt_entry_match *ematch;
1480 /* Cleanup all matches */
1481 xt_ematch_foreach(ematch, e)
1482 module_put(ematch->u.kernel.match->me);
1483 t = compat_ipt_get_target(e);
1484 module_put(t->u.kernel.target->me);
1488 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1489 struct xt_table_info *newinfo,
1491 const unsigned char *base,
1492 const unsigned char *limit,
1493 const unsigned int *hook_entries,
1494 const unsigned int *underflows,
1497 struct xt_entry_match *ematch;
1498 struct xt_entry_target *t;
1499 struct xt_target *target;
1500 unsigned int entry_offset;
1504 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1505 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1506 (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1507 duprintf("Bad offset %p, limit = %p\n", e, limit);
1511 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1512 sizeof(struct compat_xt_entry_target)) {
1513 duprintf("checking: element %p size %u\n",
1518 /* For purposes of check_entry casting the compat entry is fine */
1519 ret = check_entry((struct ipt_entry *)e, name);
1523 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1524 entry_offset = (void *)e - (void *)base;
1526 xt_ematch_foreach(ematch, e) {
1527 ret = compat_find_calc_match(ematch, name,
1528 &e->ip, e->comefrom, &off);
1530 goto release_matches;
1534 t = compat_ipt_get_target(e);
1535 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1536 t->u.user.revision);
1537 if (IS_ERR(target)) {
1538 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1540 ret = PTR_ERR(target);
1541 goto release_matches;
1543 t->u.kernel.target = target;
1545 off += xt_compat_target_offset(target);
1547 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1551 /* Check hooks & underflows */
1552 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1553 if ((unsigned char *)e - base == hook_entries[h])
1554 newinfo->hook_entry[h] = hook_entries[h];
1555 if ((unsigned char *)e - base == underflows[h])
1556 newinfo->underflow[h] = underflows[h];
1559 /* Clear counters and comefrom */
1560 memset(&e->counters, 0, sizeof(e->counters));
1565 module_put(t->u.kernel.target->me);
1567 xt_ematch_foreach(ematch, e) {
1570 module_put(ematch->u.kernel.match->me);
1576 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1577 unsigned int *size, const char *name,
1578 struct xt_table_info *newinfo, unsigned char *base)
1580 struct xt_entry_target *t;
1581 struct xt_target *target;
1582 struct ipt_entry *de;
1583 unsigned int origsize;
1585 struct xt_entry_match *ematch;
1589 de = (struct ipt_entry *)*dstptr;
1590 memcpy(de, e, sizeof(struct ipt_entry));
1591 memcpy(&de->counters, &e->counters, sizeof(e->counters));
1593 *dstptr += sizeof(struct ipt_entry);
1594 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1596 xt_ematch_foreach(ematch, e) {
1597 ret = xt_compat_match_from_user(ematch, dstptr, size);
1601 de->target_offset = e->target_offset - (origsize - *size);
1602 t = compat_ipt_get_target(e);
1603 target = t->u.kernel.target;
1604 xt_compat_target_from_user(t, dstptr, size);
1606 de->next_offset = e->next_offset - (origsize - *size);
1607 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1608 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1609 newinfo->hook_entry[h] -= origsize - *size;
1610 if ((unsigned char *)de - base < newinfo->underflow[h])
1611 newinfo->underflow[h] -= origsize - *size;
1617 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1619 struct xt_entry_match *ematch;
1620 struct xt_mtchk_param mtpar;
1627 mtpar.entryinfo = &e->ip;
1628 mtpar.hook_mask = e->comefrom;
1629 mtpar.family = NFPROTO_IPV4;
1630 xt_ematch_foreach(ematch, e) {
1631 ret = check_match(ematch, &mtpar);
1633 goto cleanup_matches;
1637 ret = check_target(e, net, name);
1639 goto cleanup_matches;
1643 xt_ematch_foreach(ematch, e) {
1646 cleanup_match(ematch, net);
1652 translate_compat_table(struct net *net,
1654 unsigned int valid_hooks,
1655 struct xt_table_info **pinfo,
1657 unsigned int total_size,
1658 unsigned int number,
1659 unsigned int *hook_entries,
1660 unsigned int *underflows)
1663 struct xt_table_info *newinfo, *info;
1664 void *pos, *entry0, *entry1;
1665 struct compat_ipt_entry *iter0;
1666 struct ipt_entry *iter1;
1673 info->number = number;
1675 /* Init all hooks to impossible value. */
1676 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1677 info->hook_entry[i] = 0xFFFFFFFF;
1678 info->underflow[i] = 0xFFFFFFFF;
1681 duprintf("translate_compat_table: size %u\n", info->size);
1683 xt_compat_lock(AF_INET);
1684 /* Walk through entries, checking offsets. */
1685 xt_entry_foreach(iter0, entry0, total_size) {
1686 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1688 entry0 + total_size,
1699 duprintf("translate_compat_table: %u not %u entries\n",
1704 /* Check hooks all assigned */
1705 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1706 /* Only hooks which are valid */
1707 if (!(valid_hooks & (1 << i)))
1709 if (info->hook_entry[i] == 0xFFFFFFFF) {
1710 duprintf("Invalid hook entry %u %u\n",
1711 i, hook_entries[i]);
1714 if (info->underflow[i] == 0xFFFFFFFF) {
1715 duprintf("Invalid underflow %u %u\n",
1722 newinfo = xt_alloc_table_info(size);
1726 newinfo->number = number;
1727 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1728 newinfo->hook_entry[i] = info->hook_entry[i];
1729 newinfo->underflow[i] = info->underflow[i];
1731 entry1 = newinfo->entries[raw_smp_processor_id()];
1734 xt_entry_foreach(iter0, entry0, total_size) {
1735 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1736 name, newinfo, entry1);
1740 xt_compat_flush_offsets(AF_INET);
1741 xt_compat_unlock(AF_INET);
1746 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1750 xt_entry_foreach(iter1, entry1, newinfo->size) {
1751 ret = compat_check_entry(iter1, net, name);
1755 if (strcmp(ipt_get_target(iter1)->u.user.name,
1756 XT_ERROR_TARGET) == 0)
1757 ++newinfo->stacksize;
1761 * The first i matches need cleanup_entry (calls ->destroy)
1762 * because they had called ->check already. The other j-i
1763 * entries need only release.
1767 xt_entry_foreach(iter0, entry0, newinfo->size) {
1772 compat_release_entry(iter0);
1774 xt_entry_foreach(iter1, entry1, newinfo->size) {
1777 cleanup_entry(iter1, net);
1779 xt_free_table_info(newinfo);
1783 /* And one copy for every other CPU */
1784 for_each_possible_cpu(i)
1785 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1786 memcpy(newinfo->entries[i], entry1, newinfo->size);
1790 xt_free_table_info(info);
1794 xt_free_table_info(newinfo);
1796 xt_entry_foreach(iter0, entry0, total_size) {
1799 compat_release_entry(iter0);
1803 xt_compat_flush_offsets(AF_INET);
1804 xt_compat_unlock(AF_INET);
1809 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1812 struct compat_ipt_replace tmp;
1813 struct xt_table_info *newinfo;
1814 void *loc_cpu_entry;
1815 struct ipt_entry *iter;
1817 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1820 /* overflow check */
1821 if (tmp.size >= INT_MAX / num_possible_cpus())
1823 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1826 newinfo = xt_alloc_table_info(tmp.size);
1830 /* choose the copy that is on our node/cpu */
1831 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1832 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1838 ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1839 &newinfo, &loc_cpu_entry, tmp.size,
1840 tmp.num_entries, tmp.hook_entry,
1845 duprintf("compat_do_replace: Translated table\n");
1847 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1848 tmp.num_counters, compat_ptr(tmp.counters));
1850 goto free_newinfo_untrans;
1853 free_newinfo_untrans:
1854 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1855 cleanup_entry(iter, net);
1857 xt_free_table_info(newinfo);
1862 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1867 if (!capable(CAP_NET_ADMIN))
1871 case IPT_SO_SET_REPLACE:
1872 ret = compat_do_replace(sock_net(sk), user, len);
1875 case IPT_SO_SET_ADD_COUNTERS:
1876 ret = do_add_counters(sock_net(sk), user, len, 1);
1880 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1887 struct compat_ipt_get_entries {
1888 char name[XT_TABLE_MAXNAMELEN];
1890 struct compat_ipt_entry entrytable[0];
1894 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1895 void __user *userptr)
1897 struct xt_counters *counters;
1898 const struct xt_table_info *private = table->private;
1902 const void *loc_cpu_entry;
1904 struct ipt_entry *iter;
1906 counters = alloc_counters(table);
1907 if (IS_ERR(counters))
1908 return PTR_ERR(counters);
1910 /* choose the copy that is on our node/cpu, ...
1911 * This choice is lazy (because current thread is
1912 * allowed to migrate to another cpu)
1914 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1917 xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1918 ret = compat_copy_entry_to_user(iter, &pos,
1919 &size, counters, i++);
1929 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1933 struct compat_ipt_get_entries get;
1936 if (*len < sizeof(get)) {
1937 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1941 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1944 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1945 duprintf("compat_get_entries: %u != %zu\n",
1946 *len, sizeof(get) + get.size);
1950 xt_compat_lock(AF_INET);
1951 t = xt_find_table_lock(net, AF_INET, get.name);
1952 if (t && !IS_ERR(t)) {
1953 const struct xt_table_info *private = t->private;
1954 struct xt_table_info info;
1955 duprintf("t->private->number = %u\n", private->number);
1956 ret = compat_table_info(private, &info);
1957 if (!ret && get.size == info.size) {
1958 ret = compat_copy_entries_to_user(private->size,
1959 t, uptr->entrytable);
1961 duprintf("compat_get_entries: I've got %u not %u!\n",
1962 private->size, get.size);
1965 xt_compat_flush_offsets(AF_INET);
1969 ret = t ? PTR_ERR(t) : -ENOENT;
1971 xt_compat_unlock(AF_INET);
1975 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1978 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1982 if (!capable(CAP_NET_ADMIN))
1986 case IPT_SO_GET_INFO:
1987 ret = get_info(sock_net(sk), user, len, 1);
1989 case IPT_SO_GET_ENTRIES:
1990 ret = compat_get_entries(sock_net(sk), user, len);
1993 ret = do_ipt_get_ctl(sk, cmd, user, len);
2000 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2004 if (!capable(CAP_NET_ADMIN))
2008 case IPT_SO_SET_REPLACE:
2009 ret = do_replace(sock_net(sk), user, len);
2012 case IPT_SO_SET_ADD_COUNTERS:
2013 ret = do_add_counters(sock_net(sk), user, len, 0);
2017 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
2025 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2029 if (!capable(CAP_NET_ADMIN))
2033 case IPT_SO_GET_INFO:
2034 ret = get_info(sock_net(sk), user, len, 0);
2037 case IPT_SO_GET_ENTRIES:
2038 ret = get_entries(sock_net(sk), user, len);
2041 case IPT_SO_GET_REVISION_MATCH:
2042 case IPT_SO_GET_REVISION_TARGET: {
2043 struct xt_get_revision rev;
2046 if (*len != sizeof(rev)) {
2050 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2055 if (cmd == IPT_SO_GET_REVISION_TARGET)
2060 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2063 "ipt_%s", rev.name);
2068 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2075 struct xt_table *ipt_register_table(struct net *net,
2076 const struct xt_table *table,
2077 const struct ipt_replace *repl)
2080 struct xt_table_info *newinfo;
2081 struct xt_table_info bootstrap = {0};
2082 void *loc_cpu_entry;
2083 struct xt_table *new_table;
2085 newinfo = xt_alloc_table_info(repl->size);
2091 /* choose the copy on our node/cpu, but dont care about preemption */
2092 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2093 memcpy(loc_cpu_entry, repl->entries, repl->size);
2095 ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2099 new_table = xt_register_table(net, table, &bootstrap, newinfo);
2100 if (IS_ERR(new_table)) {
2101 ret = PTR_ERR(new_table);
2108 xt_free_table_info(newinfo);
2110 return ERR_PTR(ret);
2113 void ipt_unregister_table(struct net *net, struct xt_table *table)
2115 struct xt_table_info *private;
2116 void *loc_cpu_entry;
2117 struct module *table_owner = table->me;
2118 struct ipt_entry *iter;
2120 private = xt_unregister_table(table);
2122 /* Decrease module usage counts and free resources */
2123 loc_cpu_entry = private->entries[raw_smp_processor_id()];
2124 xt_entry_foreach(iter, loc_cpu_entry, private->size)
2125 cleanup_entry(iter, net);
2126 if (private->number > private->initial_entries)
2127 module_put(table_owner);
2128 xt_free_table_info(private);
2131 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2133 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2134 u_int8_t type, u_int8_t code,
2137 return ((test_type == 0xFF) ||
2138 (type == test_type && code >= min_code && code <= max_code))
2143 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2145 const struct icmphdr *ic;
2146 struct icmphdr _icmph;
2147 const struct ipt_icmp *icmpinfo = par->matchinfo;
2149 /* Must not be a fragment. */
2150 if (par->fragoff != 0)
2153 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2155 /* We've been asked to examine this packet, and we
2156 * can't. Hence, no choice but to drop.
2158 duprintf("Dropping evil ICMP tinygram.\n");
2159 par->hotdrop = true;
2163 return icmp_type_code_match(icmpinfo->type,
2167 !!(icmpinfo->invflags&IPT_ICMP_INV));
2170 static int icmp_checkentry(const struct xt_mtchk_param *par)
2172 const struct ipt_icmp *icmpinfo = par->matchinfo;
2174 /* Must specify no unknown invflags */
2175 return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2178 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2180 .name = XT_STANDARD_TARGET,
2181 .targetsize = sizeof(int),
2182 .family = NFPROTO_IPV4,
2183 #ifdef CONFIG_COMPAT
2184 .compatsize = sizeof(compat_int_t),
2185 .compat_from_user = compat_standard_from_user,
2186 .compat_to_user = compat_standard_to_user,
2190 .name = XT_ERROR_TARGET,
2191 .target = ipt_error,
2192 .targetsize = XT_FUNCTION_MAXNAMELEN,
2193 .family = NFPROTO_IPV4,
2197 static struct nf_sockopt_ops ipt_sockopts = {
2199 .set_optmin = IPT_BASE_CTL,
2200 .set_optmax = IPT_SO_SET_MAX+1,
2201 .set = do_ipt_set_ctl,
2202 #ifdef CONFIG_COMPAT
2203 .compat_set = compat_do_ipt_set_ctl,
2205 .get_optmin = IPT_BASE_CTL,
2206 .get_optmax = IPT_SO_GET_MAX+1,
2207 .get = do_ipt_get_ctl,
2208 #ifdef CONFIG_COMPAT
2209 .compat_get = compat_do_ipt_get_ctl,
2211 .owner = THIS_MODULE,
2214 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2217 .match = icmp_match,
2218 .matchsize = sizeof(struct ipt_icmp),
2219 .checkentry = icmp_checkentry,
2220 .proto = IPPROTO_ICMP,
2221 .family = NFPROTO_IPV4,
2225 static int __net_init ip_tables_net_init(struct net *net)
2227 return xt_proto_init(net, NFPROTO_IPV4);
2230 static void __net_exit ip_tables_net_exit(struct net *net)
2232 xt_proto_fini(net, NFPROTO_IPV4);
2235 static struct pernet_operations ip_tables_net_ops = {
2236 .init = ip_tables_net_init,
2237 .exit = ip_tables_net_exit,
2240 static int __init ip_tables_init(void)
2244 ret = register_pernet_subsys(&ip_tables_net_ops);
2248 /* Noone else will be downing sem now, so we won't sleep */
2249 ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2252 ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2256 /* Register setsockopt */
2257 ret = nf_register_sockopt(&ipt_sockopts);
2261 pr_info("(C) 2000-2006 Netfilter Core Team\n");
2265 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2267 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2269 unregister_pernet_subsys(&ip_tables_net_ops);
2274 static void __exit ip_tables_fini(void)
2276 nf_unregister_sockopt(&ipt_sockopts);
2278 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2279 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2280 unregister_pernet_subsys(&ip_tables_net_ops);
2283 EXPORT_SYMBOL(ipt_register_table);
2284 EXPORT_SYMBOL(ipt_unregister_table);
2285 EXPORT_SYMBOL(ipt_do_table);
2286 module_init(ip_tables_init);
2287 module_exit(ip_tables_fini);