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 ipt_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 ipt_standard_target *t = (void *)ipt_get_target_c(s);
235 if (strcmp(t->target.u.kernel.target->name, IPT_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 IPT_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 ipt_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 ipt_standard_target *)t)->verdict;
385 /* Pop from stack? */
386 if (v != IPT_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 == IPT_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 ipt_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 IPT_STANDARD_TARGET) == 0) &&
479 t->verdict < 0 && unconditional(&e->ip)) ||
481 unsigned int oldpos, size;
483 if ((strcmp(t->target.u.user.name,
484 IPT_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 IPT_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 ipt_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 ipt_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 ipt_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 ipt_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 ipt_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 ipt_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 ipt_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 ipt_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 ipt_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 ipt_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 ipt_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 ipt_entry_match *m;
976 const struct ipt_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 ipt_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 ipt_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 ipt_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[IPT_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[IPT_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 info.valid_hooks = t->valid_hooks;
1128 memcpy(info.hook_entry, private->hook_entry,
1129 sizeof(info.hook_entry));
1130 memcpy(info.underflow, private->underflow,
1131 sizeof(info.underflow));
1132 info.num_entries = private->number;
1133 info.size = private->size;
1134 strcpy(info.name, name);
1136 if (copy_to_user(user, &info, *len) != 0)
1144 ret = t ? PTR_ERR(t) : -ENOENT;
1145 #ifdef CONFIG_COMPAT
1147 xt_compat_unlock(AF_INET);
1153 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1157 struct ipt_get_entries get;
1160 if (*len < sizeof(get)) {
1161 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1164 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1166 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1167 duprintf("get_entries: %u != %zu\n",
1168 *len, sizeof(get) + get.size);
1172 t = xt_find_table_lock(net, AF_INET, get.name);
1173 if (t && !IS_ERR(t)) {
1174 const struct xt_table_info *private = t->private;
1175 duprintf("t->private->number = %u\n", private->number);
1176 if (get.size == private->size)
1177 ret = copy_entries_to_user(private->size,
1178 t, uptr->entrytable);
1180 duprintf("get_entries: I've got %u not %u!\n",
1181 private->size, get.size);
1187 ret = t ? PTR_ERR(t) : -ENOENT;
1193 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1194 struct xt_table_info *newinfo, unsigned int num_counters,
1195 void __user *counters_ptr)
1199 struct xt_table_info *oldinfo;
1200 struct xt_counters *counters;
1201 void *loc_cpu_old_entry;
1202 struct ipt_entry *iter;
1205 counters = vmalloc(num_counters * sizeof(struct xt_counters));
1211 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1212 "iptable_%s", name);
1213 if (!t || IS_ERR(t)) {
1214 ret = t ? PTR_ERR(t) : -ENOENT;
1215 goto free_newinfo_counters_untrans;
1219 if (valid_hooks != t->valid_hooks) {
1220 duprintf("Valid hook crap: %08X vs %08X\n",
1221 valid_hooks, t->valid_hooks);
1226 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1230 /* Update module usage count based on number of rules */
1231 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1232 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1233 if ((oldinfo->number > oldinfo->initial_entries) ||
1234 (newinfo->number <= oldinfo->initial_entries))
1236 if ((oldinfo->number > oldinfo->initial_entries) &&
1237 (newinfo->number <= oldinfo->initial_entries))
1240 /* Get the old counters, and synchronize with replace */
1241 get_counters(oldinfo, counters);
1243 /* Decrease module usage counts and free resource */
1244 loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1245 xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1246 cleanup_entry(iter, net);
1248 xt_free_table_info(oldinfo);
1249 if (copy_to_user(counters_ptr, counters,
1250 sizeof(struct xt_counters) * num_counters) != 0)
1259 free_newinfo_counters_untrans:
1266 do_replace(struct net *net, const void __user *user, unsigned int len)
1269 struct ipt_replace tmp;
1270 struct xt_table_info *newinfo;
1271 void *loc_cpu_entry;
1272 struct ipt_entry *iter;
1274 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1277 /* overflow check */
1278 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1281 newinfo = xt_alloc_table_info(tmp.size);
1285 /* choose the copy that is on our node/cpu */
1286 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1287 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1293 ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1297 duprintf("Translated table\n");
1299 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1300 tmp.num_counters, tmp.counters);
1302 goto free_newinfo_untrans;
1305 free_newinfo_untrans:
1306 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1307 cleanup_entry(iter, net);
1309 xt_free_table_info(newinfo);
1314 do_add_counters(struct net *net, const void __user *user,
1315 unsigned int len, int compat)
1317 unsigned int i, curcpu;
1318 struct xt_counters_info tmp;
1319 struct xt_counters *paddc;
1320 unsigned int num_counters;
1325 const struct xt_table_info *private;
1327 void *loc_cpu_entry;
1328 struct ipt_entry *iter;
1329 #ifdef CONFIG_COMPAT
1330 struct compat_xt_counters_info compat_tmp;
1334 size = sizeof(struct compat_xt_counters_info);
1339 size = sizeof(struct xt_counters_info);
1342 if (copy_from_user(ptmp, user, size) != 0)
1345 #ifdef CONFIG_COMPAT
1347 num_counters = compat_tmp.num_counters;
1348 name = compat_tmp.name;
1352 num_counters = tmp.num_counters;
1356 if (len != size + num_counters * sizeof(struct xt_counters))
1359 paddc = vmalloc(len - size);
1363 if (copy_from_user(paddc, user + size, len - size) != 0) {
1368 t = xt_find_table_lock(net, AF_INET, name);
1369 if (!t || IS_ERR(t)) {
1370 ret = t ? PTR_ERR(t) : -ENOENT;
1375 private = t->private;
1376 if (private->number != num_counters) {
1378 goto unlock_up_free;
1382 /* Choose the copy that is on our node */
1383 curcpu = smp_processor_id();
1384 loc_cpu_entry = private->entries[curcpu];
1385 xt_info_wrlock(curcpu);
1386 xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1387 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1390 xt_info_wrunlock(curcpu);
1401 #ifdef CONFIG_COMPAT
1402 struct compat_ipt_replace {
1403 char name[IPT_TABLE_MAXNAMELEN];
1407 u32 hook_entry[NF_INET_NUMHOOKS];
1408 u32 underflow[NF_INET_NUMHOOKS];
1410 compat_uptr_t counters; /* struct ipt_counters * */
1411 struct compat_ipt_entry entries[0];
1415 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1416 unsigned int *size, struct xt_counters *counters,
1419 struct ipt_entry_target *t;
1420 struct compat_ipt_entry __user *ce;
1421 u_int16_t target_offset, next_offset;
1422 compat_uint_t origsize;
1423 const struct xt_entry_match *ematch;
1427 ce = (struct compat_ipt_entry __user *)*dstptr;
1428 if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1429 copy_to_user(&ce->counters, &counters[i],
1430 sizeof(counters[i])) != 0)
1433 *dstptr += sizeof(struct compat_ipt_entry);
1434 *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1436 xt_ematch_foreach(ematch, e) {
1437 ret = xt_compat_match_to_user(ematch, dstptr, size);
1441 target_offset = e->target_offset - (origsize - *size);
1442 t = ipt_get_target(e);
1443 ret = xt_compat_target_to_user(t, dstptr, size);
1446 next_offset = e->next_offset - (origsize - *size);
1447 if (put_user(target_offset, &ce->target_offset) != 0 ||
1448 put_user(next_offset, &ce->next_offset) != 0)
1454 compat_find_calc_match(struct ipt_entry_match *m,
1456 const struct ipt_ip *ip,
1457 unsigned int hookmask,
1460 struct xt_match *match;
1462 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1463 m->u.user.revision);
1464 if (IS_ERR(match)) {
1465 duprintf("compat_check_calc_match: `%s' not found\n",
1467 return PTR_ERR(match);
1469 m->u.kernel.match = match;
1470 *size += xt_compat_match_offset(match);
1474 static void compat_release_entry(struct compat_ipt_entry *e)
1476 struct ipt_entry_target *t;
1477 struct xt_entry_match *ematch;
1479 /* Cleanup all matches */
1480 xt_ematch_foreach(ematch, e)
1481 module_put(ematch->u.kernel.match->me);
1482 t = compat_ipt_get_target(e);
1483 module_put(t->u.kernel.target->me);
1487 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1488 struct xt_table_info *newinfo,
1490 const unsigned char *base,
1491 const unsigned char *limit,
1492 const unsigned int *hook_entries,
1493 const unsigned int *underflows,
1496 struct xt_entry_match *ematch;
1497 struct ipt_entry_target *t;
1498 struct xt_target *target;
1499 unsigned int entry_offset;
1503 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1504 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1505 (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1506 duprintf("Bad offset %p, limit = %p\n", e, limit);
1510 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1511 sizeof(struct compat_xt_entry_target)) {
1512 duprintf("checking: element %p size %u\n",
1517 /* For purposes of check_entry casting the compat entry is fine */
1518 ret = check_entry((struct ipt_entry *)e, name);
1522 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1523 entry_offset = (void *)e - (void *)base;
1525 xt_ematch_foreach(ematch, e) {
1526 ret = compat_find_calc_match(ematch, name,
1527 &e->ip, e->comefrom, &off);
1529 goto release_matches;
1533 t = compat_ipt_get_target(e);
1534 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1535 t->u.user.revision);
1536 if (IS_ERR(target)) {
1537 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1539 ret = PTR_ERR(target);
1540 goto release_matches;
1542 t->u.kernel.target = target;
1544 off += xt_compat_target_offset(target);
1546 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1550 /* Check hooks & underflows */
1551 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1552 if ((unsigned char *)e - base == hook_entries[h])
1553 newinfo->hook_entry[h] = hook_entries[h];
1554 if ((unsigned char *)e - base == underflows[h])
1555 newinfo->underflow[h] = underflows[h];
1558 /* Clear counters and comefrom */
1559 memset(&e->counters, 0, sizeof(e->counters));
1564 module_put(t->u.kernel.target->me);
1566 xt_ematch_foreach(ematch, e) {
1569 module_put(ematch->u.kernel.match->me);
1575 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1576 unsigned int *size, const char *name,
1577 struct xt_table_info *newinfo, unsigned char *base)
1579 struct ipt_entry_target *t;
1580 struct xt_target *target;
1581 struct ipt_entry *de;
1582 unsigned int origsize;
1584 struct xt_entry_match *ematch;
1588 de = (struct ipt_entry *)*dstptr;
1589 memcpy(de, e, sizeof(struct ipt_entry));
1590 memcpy(&de->counters, &e->counters, sizeof(e->counters));
1592 *dstptr += sizeof(struct ipt_entry);
1593 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1595 xt_ematch_foreach(ematch, e) {
1596 ret = xt_compat_match_from_user(ematch, dstptr, size);
1600 de->target_offset = e->target_offset - (origsize - *size);
1601 t = compat_ipt_get_target(e);
1602 target = t->u.kernel.target;
1603 xt_compat_target_from_user(t, dstptr, size);
1605 de->next_offset = e->next_offset - (origsize - *size);
1606 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1607 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1608 newinfo->hook_entry[h] -= origsize - *size;
1609 if ((unsigned char *)de - base < newinfo->underflow[h])
1610 newinfo->underflow[h] -= origsize - *size;
1616 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1618 struct xt_entry_match *ematch;
1619 struct xt_mtchk_param mtpar;
1626 mtpar.entryinfo = &e->ip;
1627 mtpar.hook_mask = e->comefrom;
1628 mtpar.family = NFPROTO_IPV4;
1629 xt_ematch_foreach(ematch, e) {
1630 ret = check_match(ematch, &mtpar);
1632 goto cleanup_matches;
1636 ret = check_target(e, net, name);
1638 goto cleanup_matches;
1642 xt_ematch_foreach(ematch, e) {
1645 cleanup_match(ematch, net);
1651 translate_compat_table(struct net *net,
1653 unsigned int valid_hooks,
1654 struct xt_table_info **pinfo,
1656 unsigned int total_size,
1657 unsigned int number,
1658 unsigned int *hook_entries,
1659 unsigned int *underflows)
1662 struct xt_table_info *newinfo, *info;
1663 void *pos, *entry0, *entry1;
1664 struct compat_ipt_entry *iter0;
1665 struct ipt_entry *iter1;
1672 info->number = number;
1674 /* Init all hooks to impossible value. */
1675 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1676 info->hook_entry[i] = 0xFFFFFFFF;
1677 info->underflow[i] = 0xFFFFFFFF;
1680 duprintf("translate_compat_table: size %u\n", info->size);
1682 xt_compat_lock(AF_INET);
1683 /* Walk through entries, checking offsets. */
1684 xt_entry_foreach(iter0, entry0, total_size) {
1685 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1687 entry0 + total_size,
1698 duprintf("translate_compat_table: %u not %u entries\n",
1703 /* Check hooks all assigned */
1704 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1705 /* Only hooks which are valid */
1706 if (!(valid_hooks & (1 << i)))
1708 if (info->hook_entry[i] == 0xFFFFFFFF) {
1709 duprintf("Invalid hook entry %u %u\n",
1710 i, hook_entries[i]);
1713 if (info->underflow[i] == 0xFFFFFFFF) {
1714 duprintf("Invalid underflow %u %u\n",
1721 newinfo = xt_alloc_table_info(size);
1725 newinfo->number = number;
1726 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1727 newinfo->hook_entry[i] = info->hook_entry[i];
1728 newinfo->underflow[i] = info->underflow[i];
1730 entry1 = newinfo->entries[raw_smp_processor_id()];
1733 xt_entry_foreach(iter0, entry0, total_size) {
1734 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1735 name, newinfo, entry1);
1739 xt_compat_flush_offsets(AF_INET);
1740 xt_compat_unlock(AF_INET);
1745 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1749 xt_entry_foreach(iter1, entry1, newinfo->size) {
1750 ret = compat_check_entry(iter1, net, name);
1754 if (strcmp(ipt_get_target(iter1)->u.user.name,
1755 XT_ERROR_TARGET) == 0)
1756 ++newinfo->stacksize;
1760 * The first i matches need cleanup_entry (calls ->destroy)
1761 * because they had called ->check already. The other j-i
1762 * entries need only release.
1766 xt_entry_foreach(iter0, entry0, newinfo->size) {
1771 compat_release_entry(iter0);
1773 xt_entry_foreach(iter1, entry1, newinfo->size) {
1776 cleanup_entry(iter1, net);
1778 xt_free_table_info(newinfo);
1782 /* And one copy for every other CPU */
1783 for_each_possible_cpu(i)
1784 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1785 memcpy(newinfo->entries[i], entry1, newinfo->size);
1789 xt_free_table_info(info);
1793 xt_free_table_info(newinfo);
1795 xt_entry_foreach(iter0, entry0, total_size) {
1798 compat_release_entry(iter0);
1802 xt_compat_flush_offsets(AF_INET);
1803 xt_compat_unlock(AF_INET);
1808 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1811 struct compat_ipt_replace tmp;
1812 struct xt_table_info *newinfo;
1813 void *loc_cpu_entry;
1814 struct ipt_entry *iter;
1816 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1819 /* overflow check */
1820 if (tmp.size >= INT_MAX / num_possible_cpus())
1822 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1825 newinfo = xt_alloc_table_info(tmp.size);
1829 /* choose the copy that is on our node/cpu */
1830 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1831 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1837 ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1838 &newinfo, &loc_cpu_entry, tmp.size,
1839 tmp.num_entries, tmp.hook_entry,
1844 duprintf("compat_do_replace: Translated table\n");
1846 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1847 tmp.num_counters, compat_ptr(tmp.counters));
1849 goto free_newinfo_untrans;
1852 free_newinfo_untrans:
1853 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1854 cleanup_entry(iter, net);
1856 xt_free_table_info(newinfo);
1861 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1866 if (!capable(CAP_NET_ADMIN))
1870 case IPT_SO_SET_REPLACE:
1871 ret = compat_do_replace(sock_net(sk), user, len);
1874 case IPT_SO_SET_ADD_COUNTERS:
1875 ret = do_add_counters(sock_net(sk), user, len, 1);
1879 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1886 struct compat_ipt_get_entries {
1887 char name[IPT_TABLE_MAXNAMELEN];
1889 struct compat_ipt_entry entrytable[0];
1893 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1894 void __user *userptr)
1896 struct xt_counters *counters;
1897 const struct xt_table_info *private = table->private;
1901 const void *loc_cpu_entry;
1903 struct ipt_entry *iter;
1905 counters = alloc_counters(table);
1906 if (IS_ERR(counters))
1907 return PTR_ERR(counters);
1909 /* choose the copy that is on our node/cpu, ...
1910 * This choice is lazy (because current thread is
1911 * allowed to migrate to another cpu)
1913 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1916 xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1917 ret = compat_copy_entry_to_user(iter, &pos,
1918 &size, counters, i++);
1928 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1932 struct compat_ipt_get_entries get;
1935 if (*len < sizeof(get)) {
1936 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1940 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1943 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1944 duprintf("compat_get_entries: %u != %zu\n",
1945 *len, sizeof(get) + get.size);
1949 xt_compat_lock(AF_INET);
1950 t = xt_find_table_lock(net, AF_INET, get.name);
1951 if (t && !IS_ERR(t)) {
1952 const struct xt_table_info *private = t->private;
1953 struct xt_table_info info;
1954 duprintf("t->private->number = %u\n", private->number);
1955 ret = compat_table_info(private, &info);
1956 if (!ret && get.size == info.size) {
1957 ret = compat_copy_entries_to_user(private->size,
1958 t, uptr->entrytable);
1960 duprintf("compat_get_entries: I've got %u not %u!\n",
1961 private->size, get.size);
1964 xt_compat_flush_offsets(AF_INET);
1968 ret = t ? PTR_ERR(t) : -ENOENT;
1970 xt_compat_unlock(AF_INET);
1974 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1977 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1981 if (!capable(CAP_NET_ADMIN))
1985 case IPT_SO_GET_INFO:
1986 ret = get_info(sock_net(sk), user, len, 1);
1988 case IPT_SO_GET_ENTRIES:
1989 ret = compat_get_entries(sock_net(sk), user, len);
1992 ret = do_ipt_get_ctl(sk, cmd, user, len);
1999 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2003 if (!capable(CAP_NET_ADMIN))
2007 case IPT_SO_SET_REPLACE:
2008 ret = do_replace(sock_net(sk), user, len);
2011 case IPT_SO_SET_ADD_COUNTERS:
2012 ret = do_add_counters(sock_net(sk), user, len, 0);
2016 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
2024 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2028 if (!capable(CAP_NET_ADMIN))
2032 case IPT_SO_GET_INFO:
2033 ret = get_info(sock_net(sk), user, len, 0);
2036 case IPT_SO_GET_ENTRIES:
2037 ret = get_entries(sock_net(sk), user, len);
2040 case IPT_SO_GET_REVISION_MATCH:
2041 case IPT_SO_GET_REVISION_TARGET: {
2042 struct ipt_get_revision rev;
2045 if (*len != sizeof(rev)) {
2049 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2054 if (cmd == IPT_SO_GET_REVISION_TARGET)
2059 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2062 "ipt_%s", rev.name);
2067 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2074 struct xt_table *ipt_register_table(struct net *net,
2075 const struct xt_table *table,
2076 const struct ipt_replace *repl)
2079 struct xt_table_info *newinfo;
2080 struct xt_table_info bootstrap = {0};
2081 void *loc_cpu_entry;
2082 struct xt_table *new_table;
2084 newinfo = xt_alloc_table_info(repl->size);
2090 /* choose the copy on our node/cpu, but dont care about preemption */
2091 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2092 memcpy(loc_cpu_entry, repl->entries, repl->size);
2094 ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2098 new_table = xt_register_table(net, table, &bootstrap, newinfo);
2099 if (IS_ERR(new_table)) {
2100 ret = PTR_ERR(new_table);
2107 xt_free_table_info(newinfo);
2109 return ERR_PTR(ret);
2112 void ipt_unregister_table(struct net *net, struct xt_table *table)
2114 struct xt_table_info *private;
2115 void *loc_cpu_entry;
2116 struct module *table_owner = table->me;
2117 struct ipt_entry *iter;
2119 private = xt_unregister_table(table);
2121 /* Decrease module usage counts and free resources */
2122 loc_cpu_entry = private->entries[raw_smp_processor_id()];
2123 xt_entry_foreach(iter, loc_cpu_entry, private->size)
2124 cleanup_entry(iter, net);
2125 if (private->number > private->initial_entries)
2126 module_put(table_owner);
2127 xt_free_table_info(private);
2130 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2132 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2133 u_int8_t type, u_int8_t code,
2136 return ((test_type == 0xFF) ||
2137 (type == test_type && code >= min_code && code <= max_code))
2142 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2144 const struct icmphdr *ic;
2145 struct icmphdr _icmph;
2146 const struct ipt_icmp *icmpinfo = par->matchinfo;
2148 /* Must not be a fragment. */
2149 if (par->fragoff != 0)
2152 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2154 /* We've been asked to examine this packet, and we
2155 * can't. Hence, no choice but to drop.
2157 duprintf("Dropping evil ICMP tinygram.\n");
2158 par->hotdrop = true;
2162 return icmp_type_code_match(icmpinfo->type,
2166 !!(icmpinfo->invflags&IPT_ICMP_INV));
2169 static int icmp_checkentry(const struct xt_mtchk_param *par)
2171 const struct ipt_icmp *icmpinfo = par->matchinfo;
2173 /* Must specify no unknown invflags */
2174 return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2177 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2179 .name = IPT_STANDARD_TARGET,
2180 .targetsize = sizeof(int),
2181 .family = NFPROTO_IPV4,
2182 #ifdef CONFIG_COMPAT
2183 .compatsize = sizeof(compat_int_t),
2184 .compat_from_user = compat_standard_from_user,
2185 .compat_to_user = compat_standard_to_user,
2189 .name = IPT_ERROR_TARGET,
2190 .target = ipt_error,
2191 .targetsize = IPT_FUNCTION_MAXNAMELEN,
2192 .family = NFPROTO_IPV4,
2196 static struct nf_sockopt_ops ipt_sockopts = {
2198 .set_optmin = IPT_BASE_CTL,
2199 .set_optmax = IPT_SO_SET_MAX+1,
2200 .set = do_ipt_set_ctl,
2201 #ifdef CONFIG_COMPAT
2202 .compat_set = compat_do_ipt_set_ctl,
2204 .get_optmin = IPT_BASE_CTL,
2205 .get_optmax = IPT_SO_GET_MAX+1,
2206 .get = do_ipt_get_ctl,
2207 #ifdef CONFIG_COMPAT
2208 .compat_get = compat_do_ipt_get_ctl,
2210 .owner = THIS_MODULE,
2213 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2216 .match = icmp_match,
2217 .matchsize = sizeof(struct ipt_icmp),
2218 .checkentry = icmp_checkentry,
2219 .proto = IPPROTO_ICMP,
2220 .family = NFPROTO_IPV4,
2224 static int __net_init ip_tables_net_init(struct net *net)
2226 return xt_proto_init(net, NFPROTO_IPV4);
2229 static void __net_exit ip_tables_net_exit(struct net *net)
2231 xt_proto_fini(net, NFPROTO_IPV4);
2234 static struct pernet_operations ip_tables_net_ops = {
2235 .init = ip_tables_net_init,
2236 .exit = ip_tables_net_exit,
2239 static int __init ip_tables_init(void)
2243 ret = register_pernet_subsys(&ip_tables_net_ops);
2247 /* Noone else will be downing sem now, so we won't sleep */
2248 ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2251 ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2255 /* Register setsockopt */
2256 ret = nf_register_sockopt(&ipt_sockopts);
2260 pr_info("(C) 2000-2006 Netfilter Core Team\n");
2264 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2266 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2268 unregister_pernet_subsys(&ip_tables_net_ops);
2273 static void __exit ip_tables_fini(void)
2275 nf_unregister_sockopt(&ipt_sockopts);
2277 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2278 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2279 unregister_pernet_subsys(&ip_tables_net_ops);
2282 EXPORT_SYMBOL(ipt_register_table);
2283 EXPORT_SYMBOL(ipt_unregister_table);
2284 EXPORT_SYMBOL(ipt_do_table);
2285 module_init(ip_tables_init);
2286 module_exit(ip_tables_fini);