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);
71 /* Returns whether matches rule or not. */
72 /* Performance critical - called for every packet */
74 ip_packet_match(const struct iphdr *ip,
77 const struct ipt_ip *ipinfo,
82 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
84 if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
86 FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
88 dprintf("Source or dest mismatch.\n");
90 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
91 &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
92 ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
93 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
94 &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
95 ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
99 ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
101 if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
102 dprintf("VIA in mismatch (%s vs %s).%s\n",
103 indev, ipinfo->iniface,
104 ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
108 ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
110 if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
111 dprintf("VIA out mismatch (%s vs %s).%s\n",
112 outdev, ipinfo->outiface,
113 ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
117 /* Check specific protocol */
119 FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
120 dprintf("Packet protocol %hi does not match %hi.%s\n",
121 ip->protocol, ipinfo->proto,
122 ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
126 /* If we have a fragment rule but the packet is not a fragment
127 * then we return zero */
128 if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
129 dprintf("Fragment rule but not fragment.%s\n",
130 ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
138 ip_checkentry(const struct ipt_ip *ip)
140 if (ip->flags & ~IPT_F_MASK) {
141 duprintf("Unknown flag bits set: %08X\n",
142 ip->flags & ~IPT_F_MASK);
145 if (ip->invflags & ~IPT_INV_MASK) {
146 duprintf("Unknown invflag bits set: %08X\n",
147 ip->invflags & ~IPT_INV_MASK);
154 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
156 net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
161 /* Performance critical */
162 static inline struct ipt_entry *
163 get_entry(const void *base, unsigned int offset)
165 return (struct ipt_entry *)(base + offset);
168 /* All zeroes == unconditional rule. */
169 /* Mildly perf critical (only if packet tracing is on) */
170 static inline bool unconditional(const struct ipt_ip *ip)
172 static const struct ipt_ip uncond;
174 return memcmp(ip, &uncond, sizeof(uncond)) == 0;
178 /* for const-correctness */
179 static inline const struct xt_entry_target *
180 ipt_get_target_c(const struct ipt_entry *e)
182 return ipt_get_target((struct ipt_entry *)e);
185 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
186 static const char *const hooknames[] = {
187 [NF_INET_PRE_ROUTING] = "PREROUTING",
188 [NF_INET_LOCAL_IN] = "INPUT",
189 [NF_INET_FORWARD] = "FORWARD",
190 [NF_INET_LOCAL_OUT] = "OUTPUT",
191 [NF_INET_POST_ROUTING] = "POSTROUTING",
194 enum nf_ip_trace_comments {
195 NF_IP_TRACE_COMMENT_RULE,
196 NF_IP_TRACE_COMMENT_RETURN,
197 NF_IP_TRACE_COMMENT_POLICY,
200 static const char *const comments[] = {
201 [NF_IP_TRACE_COMMENT_RULE] = "rule",
202 [NF_IP_TRACE_COMMENT_RETURN] = "return",
203 [NF_IP_TRACE_COMMENT_POLICY] = "policy",
206 static struct nf_loginfo trace_loginfo = {
207 .type = NF_LOG_TYPE_LOG,
211 .logflags = NF_LOG_MASK,
216 /* Mildly perf critical (only if packet tracing is on) */
218 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
219 const char *hookname, const char **chainname,
220 const char **comment, unsigned int *rulenum)
222 const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
224 if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
225 /* Head of user chain: ERROR target with chainname */
226 *chainname = t->target.data;
231 if (s->target_offset == sizeof(struct ipt_entry) &&
232 strcmp(t->target.u.kernel.target->name,
233 XT_STANDARD_TARGET) == 0 &&
235 unconditional(&s->ip)) {
236 /* Tail of chains: STANDARD target (return/policy) */
237 *comment = *chainname == hookname
238 ? comments[NF_IP_TRACE_COMMENT_POLICY]
239 : comments[NF_IP_TRACE_COMMENT_RETURN];
248 static void trace_packet(const struct sk_buff *skb,
250 const struct net_device *in,
251 const struct net_device *out,
252 const char *tablename,
253 const struct xt_table_info *private,
254 const struct ipt_entry *e)
256 const void *table_base;
257 const struct ipt_entry *root;
258 const char *hookname, *chainname, *comment;
259 const struct ipt_entry *iter;
260 unsigned int rulenum = 0;
261 struct net *net = dev_net(in ? in : out);
263 table_base = private->entries[smp_processor_id()];
264 root = get_entry(table_base, private->hook_entry[hook]);
266 hookname = chainname = hooknames[hook];
267 comment = comments[NF_IP_TRACE_COMMENT_RULE];
269 xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
270 if (get_chainname_rulenum(iter, e, hookname,
271 &chainname, &comment, &rulenum) != 0)
274 nf_log_packet(net, AF_INET, hook, skb, in, out, &trace_loginfo,
275 "TRACE: %s:%s:%s:%u ",
276 tablename, chainname, comment, rulenum);
281 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
283 return (void *)entry + entry->next_offset;
286 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
288 ipt_do_table(struct sk_buff *skb,
290 const struct net_device *in,
291 const struct net_device *out,
292 struct xt_table *table)
294 static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
295 const struct iphdr *ip;
296 /* Initializing verdict to NF_DROP keeps gcc happy. */
297 unsigned int verdict = NF_DROP;
298 const char *indev, *outdev;
299 const void *table_base;
300 struct ipt_entry *e, **jumpstack;
301 unsigned int *stackptr, origptr, cpu;
302 const struct xt_table_info *private;
303 struct xt_action_param acpar;
308 indev = in ? in->name : nulldevname;
309 outdev = out ? out->name : nulldevname;
310 /* We handle fragments by dealing with the first fragment as
311 * if it was a normal packet. All other fragments are treated
312 * normally, except that they will NEVER match rules that ask
313 * things we don't know, ie. tcp syn flag or ports). If the
314 * rule is also a fragment-specific rule, non-fragments won't
316 acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
317 acpar.thoff = ip_hdrlen(skb);
318 acpar.hotdrop = false;
321 acpar.family = NFPROTO_IPV4;
322 acpar.hooknum = hook;
324 IP_NF_ASSERT(table->valid_hooks & (1 << hook));
326 addend = xt_write_recseq_begin();
327 private = table->private;
328 cpu = smp_processor_id();
329 table_base = private->entries[cpu];
330 jumpstack = (struct ipt_entry **)private->jumpstack[cpu];
331 stackptr = per_cpu_ptr(private->stackptr, cpu);
334 e = get_entry(table_base, private->hook_entry[hook]);
336 pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
337 table->name, hook, origptr,
338 get_entry(table_base, private->underflow[hook]));
341 const struct xt_entry_target *t;
342 const struct xt_entry_match *ematch;
345 if (!ip_packet_match(ip, indev, outdev,
346 &e->ip, acpar.fragoff)) {
348 e = ipt_next_entry(e);
352 xt_ematch_foreach(ematch, e) {
353 acpar.match = ematch->u.kernel.match;
354 acpar.matchinfo = ematch->data;
355 if (!acpar.match->match(skb, &acpar))
359 ADD_COUNTER(e->counters, skb->len, 1);
361 t = ipt_get_target(e);
362 IP_NF_ASSERT(t->u.kernel.target);
364 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
365 /* The packet is traced: log it */
366 if (unlikely(skb->nf_trace))
367 trace_packet(skb, hook, in, out,
368 table->name, private, e);
370 /* Standard target? */
371 if (!t->u.kernel.target->target) {
374 v = ((struct xt_standard_target *)t)->verdict;
376 /* Pop from stack? */
377 if (v != XT_RETURN) {
378 verdict = (unsigned int)(-v) - 1;
381 if (*stackptr <= origptr) {
382 e = get_entry(table_base,
383 private->underflow[hook]);
384 pr_debug("Underflow (this is normal) "
387 e = jumpstack[--*stackptr];
388 pr_debug("Pulled %p out from pos %u\n",
390 e = ipt_next_entry(e);
394 if (table_base + v != ipt_next_entry(e) &&
395 !(e->ip.flags & IPT_F_GOTO)) {
396 if (*stackptr >= private->stacksize) {
400 jumpstack[(*stackptr)++] = e;
401 pr_debug("Pushed %p into pos %u\n",
405 e = get_entry(table_base, v);
409 acpar.target = t->u.kernel.target;
410 acpar.targinfo = t->data;
412 verdict = t->u.kernel.target->target(skb, &acpar);
413 /* Target might have changed stuff. */
415 if (verdict == XT_CONTINUE)
416 e = ipt_next_entry(e);
420 } while (!acpar.hotdrop);
421 pr_debug("Exiting %s; resetting sp from %u to %u\n",
422 __func__, *stackptr, origptr);
424 xt_write_recseq_end(addend);
427 #ifdef DEBUG_ALLOW_ALL
436 /* Figures out from what hook each rule can be called: returns 0 if
437 there are loops. Puts hook bitmask in comefrom. */
439 mark_source_chains(const struct xt_table_info *newinfo,
440 unsigned int valid_hooks, void *entry0)
444 /* No recursion; use packet counter to save back ptrs (reset
445 to 0 as we leave), and comefrom to save source hook bitmask */
446 for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
447 unsigned int pos = newinfo->hook_entry[hook];
448 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
450 if (!(valid_hooks & (1 << hook)))
453 /* Set initial back pointer. */
454 e->counters.pcnt = pos;
457 const struct xt_standard_target *t
458 = (void *)ipt_get_target_c(e);
459 int visited = e->comefrom & (1 << hook);
461 if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
462 pr_err("iptables: loop hook %u pos %u %08X.\n",
463 hook, pos, e->comefrom);
466 e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
468 /* Unconditional return/END. */
469 if ((e->target_offset == sizeof(struct ipt_entry) &&
470 (strcmp(t->target.u.user.name,
471 XT_STANDARD_TARGET) == 0) &&
472 t->verdict < 0 && unconditional(&e->ip)) ||
474 unsigned int oldpos, size;
476 if ((strcmp(t->target.u.user.name,
477 XT_STANDARD_TARGET) == 0) &&
478 t->verdict < -NF_MAX_VERDICT - 1) {
479 duprintf("mark_source_chains: bad "
480 "negative verdict (%i)\n",
485 /* Return: backtrack through the last
488 e->comefrom ^= (1<<NF_INET_NUMHOOKS);
489 #ifdef DEBUG_IP_FIREWALL_USER
491 & (1 << NF_INET_NUMHOOKS)) {
492 duprintf("Back unset "
499 pos = e->counters.pcnt;
500 e->counters.pcnt = 0;
502 /* We're at the start. */
506 e = (struct ipt_entry *)
508 } while (oldpos == pos + e->next_offset);
511 size = e->next_offset;
512 e = (struct ipt_entry *)
513 (entry0 + pos + size);
514 e->counters.pcnt = pos;
517 int newpos = t->verdict;
519 if (strcmp(t->target.u.user.name,
520 XT_STANDARD_TARGET) == 0 &&
522 if (newpos > newinfo->size -
523 sizeof(struct ipt_entry)) {
524 duprintf("mark_source_chains: "
525 "bad verdict (%i)\n",
529 /* This a jump; chase it. */
530 duprintf("Jump rule %u -> %u\n",
533 /* ... this is a fallthru */
534 newpos = pos + e->next_offset;
536 e = (struct ipt_entry *)
538 e->counters.pcnt = pos;
543 duprintf("Finished chain %u\n", hook);
548 static void cleanup_match(struct xt_entry_match *m, struct net *net)
550 struct xt_mtdtor_param par;
553 par.match = m->u.kernel.match;
554 par.matchinfo = m->data;
555 par.family = NFPROTO_IPV4;
556 if (par.match->destroy != NULL)
557 par.match->destroy(&par);
558 module_put(par.match->me);
562 check_entry(const struct ipt_entry *e, const char *name)
564 const struct xt_entry_target *t;
566 if (!ip_checkentry(&e->ip)) {
567 duprintf("ip check failed %p %s.\n", e, name);
571 if (e->target_offset + sizeof(struct xt_entry_target) >
575 t = ipt_get_target_c(e);
576 if (e->target_offset + t->u.target_size > e->next_offset)
583 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
585 const struct ipt_ip *ip = par->entryinfo;
588 par->match = m->u.kernel.match;
589 par->matchinfo = m->data;
591 ret = xt_check_match(par, m->u.match_size - sizeof(*m),
592 ip->proto, ip->invflags & IPT_INV_PROTO);
594 duprintf("check failed for `%s'.\n", par->match->name);
601 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
603 struct xt_match *match;
606 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
609 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
610 return PTR_ERR(match);
612 m->u.kernel.match = match;
614 ret = check_match(m, par);
620 module_put(m->u.kernel.match->me);
624 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
626 struct xt_entry_target *t = ipt_get_target(e);
627 struct xt_tgchk_param par = {
631 .target = t->u.kernel.target,
633 .hook_mask = e->comefrom,
634 .family = NFPROTO_IPV4,
638 ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
639 e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
641 duprintf("check failed for `%s'.\n",
642 t->u.kernel.target->name);
649 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
652 struct xt_entry_target *t;
653 struct xt_target *target;
656 struct xt_mtchk_param mtpar;
657 struct xt_entry_match *ematch;
659 ret = check_entry(e, name);
666 mtpar.entryinfo = &e->ip;
667 mtpar.hook_mask = e->comefrom;
668 mtpar.family = NFPROTO_IPV4;
669 xt_ematch_foreach(ematch, e) {
670 ret = find_check_match(ematch, &mtpar);
672 goto cleanup_matches;
676 t = ipt_get_target(e);
677 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
679 if (IS_ERR(target)) {
680 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
681 ret = PTR_ERR(target);
682 goto cleanup_matches;
684 t->u.kernel.target = target;
686 ret = check_target(e, net, name);
691 module_put(t->u.kernel.target->me);
693 xt_ematch_foreach(ematch, e) {
696 cleanup_match(ematch, net);
701 static bool check_underflow(const struct ipt_entry *e)
703 const struct xt_entry_target *t;
704 unsigned int verdict;
706 if (!unconditional(&e->ip))
708 t = ipt_get_target_c(e);
709 if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
711 verdict = ((struct xt_standard_target *)t)->verdict;
712 verdict = -verdict - 1;
713 return verdict == NF_DROP || verdict == NF_ACCEPT;
717 check_entry_size_and_hooks(struct ipt_entry *e,
718 struct xt_table_info *newinfo,
719 const unsigned char *base,
720 const unsigned char *limit,
721 const unsigned int *hook_entries,
722 const unsigned int *underflows,
723 unsigned int valid_hooks)
727 if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
728 (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
729 duprintf("Bad offset %p\n", e);
734 < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
735 duprintf("checking: element %p size %u\n",
740 /* Check hooks & underflows */
741 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
742 if (!(valid_hooks & (1 << h)))
744 if ((unsigned char *)e - base == hook_entries[h])
745 newinfo->hook_entry[h] = hook_entries[h];
746 if ((unsigned char *)e - base == underflows[h]) {
747 if (!check_underflow(e)) {
748 pr_err("Underflows must be unconditional and "
749 "use the STANDARD target with "
753 newinfo->underflow[h] = underflows[h];
757 /* Clear counters and comefrom */
758 e->counters = ((struct xt_counters) { 0, 0 });
764 cleanup_entry(struct ipt_entry *e, struct net *net)
766 struct xt_tgdtor_param par;
767 struct xt_entry_target *t;
768 struct xt_entry_match *ematch;
770 /* Cleanup all matches */
771 xt_ematch_foreach(ematch, e)
772 cleanup_match(ematch, net);
773 t = ipt_get_target(e);
776 par.target = t->u.kernel.target;
777 par.targinfo = t->data;
778 par.family = NFPROTO_IPV4;
779 if (par.target->destroy != NULL)
780 par.target->destroy(&par);
781 module_put(par.target->me);
784 /* Checks and translates the user-supplied table segment (held in
787 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
788 const struct ipt_replace *repl)
790 struct ipt_entry *iter;
794 newinfo->size = repl->size;
795 newinfo->number = repl->num_entries;
797 /* Init all hooks to impossible value. */
798 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
799 newinfo->hook_entry[i] = 0xFFFFFFFF;
800 newinfo->underflow[i] = 0xFFFFFFFF;
803 duprintf("translate_table: size %u\n", newinfo->size);
805 /* Walk through entries, checking offsets. */
806 xt_entry_foreach(iter, entry0, newinfo->size) {
807 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
815 if (strcmp(ipt_get_target(iter)->u.user.name,
816 XT_ERROR_TARGET) == 0)
817 ++newinfo->stacksize;
820 if (i != repl->num_entries) {
821 duprintf("translate_table: %u not %u entries\n",
822 i, repl->num_entries);
826 /* Check hooks all assigned */
827 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
828 /* Only hooks which are valid */
829 if (!(repl->valid_hooks & (1 << i)))
831 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
832 duprintf("Invalid hook entry %u %u\n",
833 i, repl->hook_entry[i]);
836 if (newinfo->underflow[i] == 0xFFFFFFFF) {
837 duprintf("Invalid underflow %u %u\n",
838 i, repl->underflow[i]);
843 if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
846 /* Finally, each sanity check must pass */
848 xt_entry_foreach(iter, entry0, newinfo->size) {
849 ret = find_check_entry(iter, net, repl->name, repl->size);
856 xt_entry_foreach(iter, entry0, newinfo->size) {
859 cleanup_entry(iter, net);
864 /* And one copy for every other CPU */
865 for_each_possible_cpu(i) {
866 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
867 memcpy(newinfo->entries[i], entry0, newinfo->size);
874 get_counters(const struct xt_table_info *t,
875 struct xt_counters counters[])
877 struct ipt_entry *iter;
881 for_each_possible_cpu(cpu) {
882 seqcount_t *s = &per_cpu(xt_recseq, cpu);
885 xt_entry_foreach(iter, t->entries[cpu], t->size) {
890 start = read_seqcount_begin(s);
891 bcnt = iter->counters.bcnt;
892 pcnt = iter->counters.pcnt;
893 } while (read_seqcount_retry(s, start));
895 ADD_COUNTER(counters[i], bcnt, pcnt);
896 ++i; /* macro does multi eval of i */
901 static struct xt_counters *alloc_counters(const struct xt_table *table)
903 unsigned int countersize;
904 struct xt_counters *counters;
905 const struct xt_table_info *private = table->private;
907 /* We need atomic snapshot of counters: rest doesn't change
908 (other than comefrom, which userspace doesn't care
910 countersize = sizeof(struct xt_counters) * private->number;
911 counters = vzalloc(countersize);
913 if (counters == NULL)
914 return ERR_PTR(-ENOMEM);
916 get_counters(private, counters);
922 copy_entries_to_user(unsigned int total_size,
923 const struct xt_table *table,
924 void __user *userptr)
926 unsigned int off, num;
927 const struct ipt_entry *e;
928 struct xt_counters *counters;
929 const struct xt_table_info *private = table->private;
931 const void *loc_cpu_entry;
933 counters = alloc_counters(table);
934 if (IS_ERR(counters))
935 return PTR_ERR(counters);
937 /* choose the copy that is on our node/cpu, ...
938 * This choice is lazy (because current thread is
939 * allowed to migrate to another cpu)
941 loc_cpu_entry = private->entries[raw_smp_processor_id()];
942 if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
947 /* FIXME: use iterator macros --RR */
948 /* ... then go back and fix counters and names */
949 for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
951 const struct xt_entry_match *m;
952 const struct xt_entry_target *t;
954 e = (struct ipt_entry *)(loc_cpu_entry + off);
955 if (copy_to_user(userptr + off
956 + offsetof(struct ipt_entry, counters),
958 sizeof(counters[num])) != 0) {
963 for (i = sizeof(struct ipt_entry);
964 i < e->target_offset;
965 i += m->u.match_size) {
968 if (copy_to_user(userptr + off + i
969 + offsetof(struct xt_entry_match,
971 m->u.kernel.match->name,
972 strlen(m->u.kernel.match->name)+1)
979 t = ipt_get_target_c(e);
980 if (copy_to_user(userptr + off + e->target_offset
981 + offsetof(struct xt_entry_target,
983 t->u.kernel.target->name,
984 strlen(t->u.kernel.target->name)+1) != 0) {
996 static void compat_standard_from_user(void *dst, const void *src)
998 int v = *(compat_int_t *)src;
1001 v += xt_compat_calc_jump(AF_INET, v);
1002 memcpy(dst, &v, sizeof(v));
1005 static int compat_standard_to_user(void __user *dst, const void *src)
1007 compat_int_t cv = *(int *)src;
1010 cv -= xt_compat_calc_jump(AF_INET, cv);
1011 return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1014 static int compat_calc_entry(const struct ipt_entry *e,
1015 const struct xt_table_info *info,
1016 const void *base, struct xt_table_info *newinfo)
1018 const struct xt_entry_match *ematch;
1019 const struct xt_entry_target *t;
1020 unsigned int entry_offset;
1023 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1024 entry_offset = (void *)e - base;
1025 xt_ematch_foreach(ematch, e)
1026 off += xt_compat_match_offset(ematch->u.kernel.match);
1027 t = ipt_get_target_c(e);
1028 off += xt_compat_target_offset(t->u.kernel.target);
1029 newinfo->size -= off;
1030 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1034 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1035 if (info->hook_entry[i] &&
1036 (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1037 newinfo->hook_entry[i] -= off;
1038 if (info->underflow[i] &&
1039 (e < (struct ipt_entry *)(base + info->underflow[i])))
1040 newinfo->underflow[i] -= off;
1045 static int compat_table_info(const struct xt_table_info *info,
1046 struct xt_table_info *newinfo)
1048 struct ipt_entry *iter;
1049 void *loc_cpu_entry;
1052 if (!newinfo || !info)
1055 /* we dont care about newinfo->entries[] */
1056 memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1057 newinfo->initial_entries = 0;
1058 loc_cpu_entry = info->entries[raw_smp_processor_id()];
1059 xt_compat_init_offsets(AF_INET, info->number);
1060 xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1061 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1069 static int get_info(struct net *net, void __user *user,
1070 const int *len, int compat)
1072 char name[XT_TABLE_MAXNAMELEN];
1076 if (*len != sizeof(struct ipt_getinfo)) {
1077 duprintf("length %u != %zu\n", *len,
1078 sizeof(struct ipt_getinfo));
1082 if (copy_from_user(name, user, sizeof(name)) != 0)
1085 name[XT_TABLE_MAXNAMELEN-1] = '\0';
1086 #ifdef CONFIG_COMPAT
1088 xt_compat_lock(AF_INET);
1090 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1091 "iptable_%s", name);
1092 if (!IS_ERR_OR_NULL(t)) {
1093 struct ipt_getinfo info;
1094 const struct xt_table_info *private = t->private;
1095 #ifdef CONFIG_COMPAT
1096 struct xt_table_info tmp;
1099 ret = compat_table_info(private, &tmp);
1100 xt_compat_flush_offsets(AF_INET);
1104 memset(&info, 0, sizeof(info));
1105 info.valid_hooks = t->valid_hooks;
1106 memcpy(info.hook_entry, private->hook_entry,
1107 sizeof(info.hook_entry));
1108 memcpy(info.underflow, private->underflow,
1109 sizeof(info.underflow));
1110 info.num_entries = private->number;
1111 info.size = private->size;
1112 strcpy(info.name, name);
1114 if (copy_to_user(user, &info, *len) != 0)
1122 ret = t ? PTR_ERR(t) : -ENOENT;
1123 #ifdef CONFIG_COMPAT
1125 xt_compat_unlock(AF_INET);
1131 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1135 struct ipt_get_entries get;
1138 if (*len < sizeof(get)) {
1139 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1142 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1144 if (*len != sizeof(struct ipt_get_entries) + get.size) {
1145 duprintf("get_entries: %u != %zu\n",
1146 *len, sizeof(get) + get.size);
1150 t = xt_find_table_lock(net, AF_INET, get.name);
1151 if (!IS_ERR_OR_NULL(t)) {
1152 const struct xt_table_info *private = t->private;
1153 duprintf("t->private->number = %u\n", private->number);
1154 if (get.size == private->size)
1155 ret = copy_entries_to_user(private->size,
1156 t, uptr->entrytable);
1158 duprintf("get_entries: I've got %u not %u!\n",
1159 private->size, get.size);
1165 ret = t ? PTR_ERR(t) : -ENOENT;
1171 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1172 struct xt_table_info *newinfo, unsigned int num_counters,
1173 void __user *counters_ptr)
1177 struct xt_table_info *oldinfo;
1178 struct xt_counters *counters;
1179 void *loc_cpu_old_entry;
1180 struct ipt_entry *iter;
1183 counters = vzalloc(num_counters * sizeof(struct xt_counters));
1189 t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1190 "iptable_%s", name);
1191 if (IS_ERR_OR_NULL(t)) {
1192 ret = t ? PTR_ERR(t) : -ENOENT;
1193 goto free_newinfo_counters_untrans;
1197 if (valid_hooks != t->valid_hooks) {
1198 duprintf("Valid hook crap: %08X vs %08X\n",
1199 valid_hooks, t->valid_hooks);
1204 oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1208 /* Update module usage count based on number of rules */
1209 duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1210 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1211 if ((oldinfo->number > oldinfo->initial_entries) ||
1212 (newinfo->number <= oldinfo->initial_entries))
1214 if ((oldinfo->number > oldinfo->initial_entries) &&
1215 (newinfo->number <= oldinfo->initial_entries))
1218 /* Get the old counters, and synchronize with replace */
1219 get_counters(oldinfo, counters);
1221 /* Decrease module usage counts and free resource */
1222 loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1223 xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1224 cleanup_entry(iter, net);
1226 xt_free_table_info(oldinfo);
1227 if (copy_to_user(counters_ptr, counters,
1228 sizeof(struct xt_counters) * num_counters) != 0)
1237 free_newinfo_counters_untrans:
1244 do_replace(struct net *net, const void __user *user, unsigned int len)
1247 struct ipt_replace tmp;
1248 struct xt_table_info *newinfo;
1249 void *loc_cpu_entry;
1250 struct ipt_entry *iter;
1252 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1255 /* overflow check */
1256 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1258 tmp.name[sizeof(tmp.name)-1] = 0;
1260 newinfo = xt_alloc_table_info(tmp.size);
1264 /* choose the copy that is on our node/cpu */
1265 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1266 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1272 ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1276 duprintf("Translated table\n");
1278 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1279 tmp.num_counters, tmp.counters);
1281 goto free_newinfo_untrans;
1284 free_newinfo_untrans:
1285 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1286 cleanup_entry(iter, net);
1288 xt_free_table_info(newinfo);
1293 do_add_counters(struct net *net, const void __user *user,
1294 unsigned int len, int compat)
1296 unsigned int i, curcpu;
1297 struct xt_counters_info tmp;
1298 struct xt_counters *paddc;
1299 unsigned int num_counters;
1304 const struct xt_table_info *private;
1306 void *loc_cpu_entry;
1307 struct ipt_entry *iter;
1308 unsigned int addend;
1309 #ifdef CONFIG_COMPAT
1310 struct compat_xt_counters_info compat_tmp;
1314 size = sizeof(struct compat_xt_counters_info);
1319 size = sizeof(struct xt_counters_info);
1322 if (copy_from_user(ptmp, user, size) != 0)
1325 #ifdef CONFIG_COMPAT
1327 num_counters = compat_tmp.num_counters;
1328 name = compat_tmp.name;
1332 num_counters = tmp.num_counters;
1336 if (len != size + num_counters * sizeof(struct xt_counters))
1339 paddc = vmalloc(len - size);
1343 if (copy_from_user(paddc, user + size, len - size) != 0) {
1348 t = xt_find_table_lock(net, AF_INET, name);
1349 if (IS_ERR_OR_NULL(t)) {
1350 ret = t ? PTR_ERR(t) : -ENOENT;
1355 private = t->private;
1356 if (private->number != num_counters) {
1358 goto unlock_up_free;
1362 /* Choose the copy that is on our node */
1363 curcpu = smp_processor_id();
1364 loc_cpu_entry = private->entries[curcpu];
1365 addend = xt_write_recseq_begin();
1366 xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1367 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1370 xt_write_recseq_end(addend);
1381 #ifdef CONFIG_COMPAT
1382 struct compat_ipt_replace {
1383 char name[XT_TABLE_MAXNAMELEN];
1387 u32 hook_entry[NF_INET_NUMHOOKS];
1388 u32 underflow[NF_INET_NUMHOOKS];
1390 compat_uptr_t counters; /* struct xt_counters * */
1391 struct compat_ipt_entry entries[0];
1395 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1396 unsigned int *size, struct xt_counters *counters,
1399 struct xt_entry_target *t;
1400 struct compat_ipt_entry __user *ce;
1401 u_int16_t target_offset, next_offset;
1402 compat_uint_t origsize;
1403 const struct xt_entry_match *ematch;
1407 ce = (struct compat_ipt_entry __user *)*dstptr;
1408 if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1409 copy_to_user(&ce->counters, &counters[i],
1410 sizeof(counters[i])) != 0)
1413 *dstptr += sizeof(struct compat_ipt_entry);
1414 *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1416 xt_ematch_foreach(ematch, e) {
1417 ret = xt_compat_match_to_user(ematch, dstptr, size);
1421 target_offset = e->target_offset - (origsize - *size);
1422 t = ipt_get_target(e);
1423 ret = xt_compat_target_to_user(t, dstptr, size);
1426 next_offset = e->next_offset - (origsize - *size);
1427 if (put_user(target_offset, &ce->target_offset) != 0 ||
1428 put_user(next_offset, &ce->next_offset) != 0)
1434 compat_find_calc_match(struct xt_entry_match *m,
1436 const struct ipt_ip *ip,
1437 unsigned int hookmask,
1440 struct xt_match *match;
1442 match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1443 m->u.user.revision);
1444 if (IS_ERR(match)) {
1445 duprintf("compat_check_calc_match: `%s' not found\n",
1447 return PTR_ERR(match);
1449 m->u.kernel.match = match;
1450 *size += xt_compat_match_offset(match);
1454 static void compat_release_entry(struct compat_ipt_entry *e)
1456 struct xt_entry_target *t;
1457 struct xt_entry_match *ematch;
1459 /* Cleanup all matches */
1460 xt_ematch_foreach(ematch, e)
1461 module_put(ematch->u.kernel.match->me);
1462 t = compat_ipt_get_target(e);
1463 module_put(t->u.kernel.target->me);
1467 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1468 struct xt_table_info *newinfo,
1470 const unsigned char *base,
1471 const unsigned char *limit,
1472 const unsigned int *hook_entries,
1473 const unsigned int *underflows,
1476 struct xt_entry_match *ematch;
1477 struct xt_entry_target *t;
1478 struct xt_target *target;
1479 unsigned int entry_offset;
1483 duprintf("check_compat_entry_size_and_hooks %p\n", e);
1484 if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1485 (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1486 duprintf("Bad offset %p, limit = %p\n", e, limit);
1490 if (e->next_offset < sizeof(struct compat_ipt_entry) +
1491 sizeof(struct compat_xt_entry_target)) {
1492 duprintf("checking: element %p size %u\n",
1497 /* For purposes of check_entry casting the compat entry is fine */
1498 ret = check_entry((struct ipt_entry *)e, name);
1502 off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1503 entry_offset = (void *)e - (void *)base;
1505 xt_ematch_foreach(ematch, e) {
1506 ret = compat_find_calc_match(ematch, name,
1507 &e->ip, e->comefrom, &off);
1509 goto release_matches;
1513 t = compat_ipt_get_target(e);
1514 target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1515 t->u.user.revision);
1516 if (IS_ERR(target)) {
1517 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1519 ret = PTR_ERR(target);
1520 goto release_matches;
1522 t->u.kernel.target = target;
1524 off += xt_compat_target_offset(target);
1526 ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1530 /* Check hooks & underflows */
1531 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1532 if ((unsigned char *)e - base == hook_entries[h])
1533 newinfo->hook_entry[h] = hook_entries[h];
1534 if ((unsigned char *)e - base == underflows[h])
1535 newinfo->underflow[h] = underflows[h];
1538 /* Clear counters and comefrom */
1539 memset(&e->counters, 0, sizeof(e->counters));
1544 module_put(t->u.kernel.target->me);
1546 xt_ematch_foreach(ematch, e) {
1549 module_put(ematch->u.kernel.match->me);
1555 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1556 unsigned int *size, const char *name,
1557 struct xt_table_info *newinfo, unsigned char *base)
1559 struct xt_entry_target *t;
1560 struct xt_target *target;
1561 struct ipt_entry *de;
1562 unsigned int origsize;
1564 struct xt_entry_match *ematch;
1568 de = (struct ipt_entry *)*dstptr;
1569 memcpy(de, e, sizeof(struct ipt_entry));
1570 memcpy(&de->counters, &e->counters, sizeof(e->counters));
1572 *dstptr += sizeof(struct ipt_entry);
1573 *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1575 xt_ematch_foreach(ematch, e) {
1576 ret = xt_compat_match_from_user(ematch, dstptr, size);
1580 de->target_offset = e->target_offset - (origsize - *size);
1581 t = compat_ipt_get_target(e);
1582 target = t->u.kernel.target;
1583 xt_compat_target_from_user(t, dstptr, size);
1585 de->next_offset = e->next_offset - (origsize - *size);
1586 for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1587 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1588 newinfo->hook_entry[h] -= origsize - *size;
1589 if ((unsigned char *)de - base < newinfo->underflow[h])
1590 newinfo->underflow[h] -= origsize - *size;
1596 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1598 struct xt_entry_match *ematch;
1599 struct xt_mtchk_param mtpar;
1606 mtpar.entryinfo = &e->ip;
1607 mtpar.hook_mask = e->comefrom;
1608 mtpar.family = NFPROTO_IPV4;
1609 xt_ematch_foreach(ematch, e) {
1610 ret = check_match(ematch, &mtpar);
1612 goto cleanup_matches;
1616 ret = check_target(e, net, name);
1618 goto cleanup_matches;
1622 xt_ematch_foreach(ematch, e) {
1625 cleanup_match(ematch, net);
1631 translate_compat_table(struct net *net,
1633 unsigned int valid_hooks,
1634 struct xt_table_info **pinfo,
1636 unsigned int total_size,
1637 unsigned int number,
1638 unsigned int *hook_entries,
1639 unsigned int *underflows)
1642 struct xt_table_info *newinfo, *info;
1643 void *pos, *entry0, *entry1;
1644 struct compat_ipt_entry *iter0;
1645 struct ipt_entry *iter1;
1652 info->number = number;
1654 /* Init all hooks to impossible value. */
1655 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1656 info->hook_entry[i] = 0xFFFFFFFF;
1657 info->underflow[i] = 0xFFFFFFFF;
1660 duprintf("translate_compat_table: size %u\n", info->size);
1662 xt_compat_lock(AF_INET);
1663 xt_compat_init_offsets(AF_INET, number);
1664 /* Walk through entries, checking offsets. */
1665 xt_entry_foreach(iter0, entry0, total_size) {
1666 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1668 entry0 + total_size,
1679 duprintf("translate_compat_table: %u not %u entries\n",
1684 /* Check hooks all assigned */
1685 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1686 /* Only hooks which are valid */
1687 if (!(valid_hooks & (1 << i)))
1689 if (info->hook_entry[i] == 0xFFFFFFFF) {
1690 duprintf("Invalid hook entry %u %u\n",
1691 i, hook_entries[i]);
1694 if (info->underflow[i] == 0xFFFFFFFF) {
1695 duprintf("Invalid underflow %u %u\n",
1702 newinfo = xt_alloc_table_info(size);
1706 newinfo->number = number;
1707 for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1708 newinfo->hook_entry[i] = info->hook_entry[i];
1709 newinfo->underflow[i] = info->underflow[i];
1711 entry1 = newinfo->entries[raw_smp_processor_id()];
1714 xt_entry_foreach(iter0, entry0, total_size) {
1715 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1716 name, newinfo, entry1);
1720 xt_compat_flush_offsets(AF_INET);
1721 xt_compat_unlock(AF_INET);
1726 if (!mark_source_chains(newinfo, valid_hooks, entry1))
1730 xt_entry_foreach(iter1, entry1, newinfo->size) {
1731 ret = compat_check_entry(iter1, net, name);
1735 if (strcmp(ipt_get_target(iter1)->u.user.name,
1736 XT_ERROR_TARGET) == 0)
1737 ++newinfo->stacksize;
1741 * The first i matches need cleanup_entry (calls ->destroy)
1742 * because they had called ->check already. The other j-i
1743 * entries need only release.
1747 xt_entry_foreach(iter0, entry0, newinfo->size) {
1752 compat_release_entry(iter0);
1754 xt_entry_foreach(iter1, entry1, newinfo->size) {
1757 cleanup_entry(iter1, net);
1759 xt_free_table_info(newinfo);
1763 /* And one copy for every other CPU */
1764 for_each_possible_cpu(i)
1765 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1766 memcpy(newinfo->entries[i], entry1, newinfo->size);
1770 xt_free_table_info(info);
1774 xt_free_table_info(newinfo);
1776 xt_entry_foreach(iter0, entry0, total_size) {
1779 compat_release_entry(iter0);
1783 xt_compat_flush_offsets(AF_INET);
1784 xt_compat_unlock(AF_INET);
1789 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1792 struct compat_ipt_replace tmp;
1793 struct xt_table_info *newinfo;
1794 void *loc_cpu_entry;
1795 struct ipt_entry *iter;
1797 if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1800 /* overflow check */
1801 if (tmp.size >= INT_MAX / num_possible_cpus())
1803 if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1805 tmp.name[sizeof(tmp.name)-1] = 0;
1807 newinfo = xt_alloc_table_info(tmp.size);
1811 /* choose the copy that is on our node/cpu */
1812 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1813 if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1819 ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1820 &newinfo, &loc_cpu_entry, tmp.size,
1821 tmp.num_entries, tmp.hook_entry,
1826 duprintf("compat_do_replace: Translated table\n");
1828 ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1829 tmp.num_counters, compat_ptr(tmp.counters));
1831 goto free_newinfo_untrans;
1834 free_newinfo_untrans:
1835 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1836 cleanup_entry(iter, net);
1838 xt_free_table_info(newinfo);
1843 compat_do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user,
1848 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1852 case IPT_SO_SET_REPLACE:
1853 ret = compat_do_replace(sock_net(sk), user, len);
1856 case IPT_SO_SET_ADD_COUNTERS:
1857 ret = do_add_counters(sock_net(sk), user, len, 1);
1861 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
1868 struct compat_ipt_get_entries {
1869 char name[XT_TABLE_MAXNAMELEN];
1871 struct compat_ipt_entry entrytable[0];
1875 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1876 void __user *userptr)
1878 struct xt_counters *counters;
1879 const struct xt_table_info *private = table->private;
1883 const void *loc_cpu_entry;
1885 struct ipt_entry *iter;
1887 counters = alloc_counters(table);
1888 if (IS_ERR(counters))
1889 return PTR_ERR(counters);
1891 /* choose the copy that is on our node/cpu, ...
1892 * This choice is lazy (because current thread is
1893 * allowed to migrate to another cpu)
1895 loc_cpu_entry = private->entries[raw_smp_processor_id()];
1898 xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1899 ret = compat_copy_entry_to_user(iter, &pos,
1900 &size, counters, i++);
1910 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1914 struct compat_ipt_get_entries get;
1917 if (*len < sizeof(get)) {
1918 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1922 if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1925 if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1926 duprintf("compat_get_entries: %u != %zu\n",
1927 *len, sizeof(get) + get.size);
1931 xt_compat_lock(AF_INET);
1932 t = xt_find_table_lock(net, AF_INET, get.name);
1933 if (!IS_ERR_OR_NULL(t)) {
1934 const struct xt_table_info *private = t->private;
1935 struct xt_table_info info;
1936 duprintf("t->private->number = %u\n", private->number);
1937 ret = compat_table_info(private, &info);
1938 if (!ret && get.size == info.size) {
1939 ret = compat_copy_entries_to_user(private->size,
1940 t, uptr->entrytable);
1942 duprintf("compat_get_entries: I've got %u not %u!\n",
1943 private->size, get.size);
1946 xt_compat_flush_offsets(AF_INET);
1950 ret = t ? PTR_ERR(t) : -ENOENT;
1952 xt_compat_unlock(AF_INET);
1956 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1959 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1963 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1967 case IPT_SO_GET_INFO:
1968 ret = get_info(sock_net(sk), user, len, 1);
1970 case IPT_SO_GET_ENTRIES:
1971 ret = compat_get_entries(sock_net(sk), user, len);
1974 ret = do_ipt_get_ctl(sk, cmd, user, len);
1981 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1985 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1989 case IPT_SO_SET_REPLACE:
1990 ret = do_replace(sock_net(sk), user, len);
1993 case IPT_SO_SET_ADD_COUNTERS:
1994 ret = do_add_counters(sock_net(sk), user, len, 0);
1998 duprintf("do_ipt_set_ctl: unknown request %i\n", cmd);
2006 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2010 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2014 case IPT_SO_GET_INFO:
2015 ret = get_info(sock_net(sk), user, len, 0);
2018 case IPT_SO_GET_ENTRIES:
2019 ret = get_entries(sock_net(sk), user, len);
2022 case IPT_SO_GET_REVISION_MATCH:
2023 case IPT_SO_GET_REVISION_TARGET: {
2024 struct xt_get_revision rev;
2027 if (*len != sizeof(rev)) {
2031 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2035 rev.name[sizeof(rev.name)-1] = 0;
2037 if (cmd == IPT_SO_GET_REVISION_TARGET)
2042 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2045 "ipt_%s", rev.name);
2050 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2057 struct xt_table *ipt_register_table(struct net *net,
2058 const struct xt_table *table,
2059 const struct ipt_replace *repl)
2062 struct xt_table_info *newinfo;
2063 struct xt_table_info bootstrap = {0};
2064 void *loc_cpu_entry;
2065 struct xt_table *new_table;
2067 newinfo = xt_alloc_table_info(repl->size);
2073 /* choose the copy on our node/cpu, but dont care about preemption */
2074 loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2075 memcpy(loc_cpu_entry, repl->entries, repl->size);
2077 ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2081 new_table = xt_register_table(net, table, &bootstrap, newinfo);
2082 if (IS_ERR(new_table)) {
2083 ret = PTR_ERR(new_table);
2090 xt_free_table_info(newinfo);
2092 return ERR_PTR(ret);
2095 void ipt_unregister_table(struct net *net, struct xt_table *table)
2097 struct xt_table_info *private;
2098 void *loc_cpu_entry;
2099 struct module *table_owner = table->me;
2100 struct ipt_entry *iter;
2102 private = xt_unregister_table(table);
2104 /* Decrease module usage counts and free resources */
2105 loc_cpu_entry = private->entries[raw_smp_processor_id()];
2106 xt_entry_foreach(iter, loc_cpu_entry, private->size)
2107 cleanup_entry(iter, net);
2108 if (private->number > private->initial_entries)
2109 module_put(table_owner);
2110 xt_free_table_info(private);
2113 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2115 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2116 u_int8_t type, u_int8_t code,
2119 return ((test_type == 0xFF) ||
2120 (type == test_type && code >= min_code && code <= max_code))
2125 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2127 const struct icmphdr *ic;
2128 struct icmphdr _icmph;
2129 const struct ipt_icmp *icmpinfo = par->matchinfo;
2131 /* Must not be a fragment. */
2132 if (par->fragoff != 0)
2135 ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2137 /* We've been asked to examine this packet, and we
2138 * can't. Hence, no choice but to drop.
2140 duprintf("Dropping evil ICMP tinygram.\n");
2141 par->hotdrop = true;
2145 return icmp_type_code_match(icmpinfo->type,
2149 !!(icmpinfo->invflags&IPT_ICMP_INV));
2152 static int icmp_checkentry(const struct xt_mtchk_param *par)
2154 const struct ipt_icmp *icmpinfo = par->matchinfo;
2156 /* Must specify no unknown invflags */
2157 return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2160 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2162 .name = XT_STANDARD_TARGET,
2163 .targetsize = sizeof(int),
2164 .family = NFPROTO_IPV4,
2165 #ifdef CONFIG_COMPAT
2166 .compatsize = sizeof(compat_int_t),
2167 .compat_from_user = compat_standard_from_user,
2168 .compat_to_user = compat_standard_to_user,
2172 .name = XT_ERROR_TARGET,
2173 .target = ipt_error,
2174 .targetsize = XT_FUNCTION_MAXNAMELEN,
2175 .family = NFPROTO_IPV4,
2179 static struct nf_sockopt_ops ipt_sockopts = {
2181 .set_optmin = IPT_BASE_CTL,
2182 .set_optmax = IPT_SO_SET_MAX+1,
2183 .set = do_ipt_set_ctl,
2184 #ifdef CONFIG_COMPAT
2185 .compat_set = compat_do_ipt_set_ctl,
2187 .get_optmin = IPT_BASE_CTL,
2188 .get_optmax = IPT_SO_GET_MAX+1,
2189 .get = do_ipt_get_ctl,
2190 #ifdef CONFIG_COMPAT
2191 .compat_get = compat_do_ipt_get_ctl,
2193 .owner = THIS_MODULE,
2196 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2199 .match = icmp_match,
2200 .matchsize = sizeof(struct ipt_icmp),
2201 .checkentry = icmp_checkentry,
2202 .proto = IPPROTO_ICMP,
2203 .family = NFPROTO_IPV4,
2207 static int __net_init ip_tables_net_init(struct net *net)
2209 return xt_proto_init(net, NFPROTO_IPV4);
2212 static void __net_exit ip_tables_net_exit(struct net *net)
2214 xt_proto_fini(net, NFPROTO_IPV4);
2217 static struct pernet_operations ip_tables_net_ops = {
2218 .init = ip_tables_net_init,
2219 .exit = ip_tables_net_exit,
2222 static int __init ip_tables_init(void)
2226 ret = register_pernet_subsys(&ip_tables_net_ops);
2230 /* No one else will be downing sem now, so we won't sleep */
2231 ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2234 ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2238 /* Register setsockopt */
2239 ret = nf_register_sockopt(&ipt_sockopts);
2243 pr_info("(C) 2000-2006 Netfilter Core Team\n");
2247 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2249 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2251 unregister_pernet_subsys(&ip_tables_net_ops);
2256 static void __exit ip_tables_fini(void)
2258 nf_unregister_sockopt(&ipt_sockopts);
2260 xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2261 xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2262 unregister_pernet_subsys(&ip_tables_net_ops);
2265 EXPORT_SYMBOL(ipt_register_table);
2266 EXPORT_SYMBOL(ipt_unregister_table);
2267 EXPORT_SYMBOL(ipt_do_table);
2268 module_init(ip_tables_init);
2269 module_exit(ip_tables_fini);