[NETFILTER]: nf_log: move logging stuff to seperate header
[linux-2.6.git] / net / ipv4 / netfilter / ip_tables.c
1 /*
2  * Packet matching code.
3  *
4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
6  *
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.
10  */
11 #include <linux/cache.h>
12 #include <linux/capability.h>
13 #include <linux/skbuff.h>
14 #include <linux/kmod.h>
15 #include <linux/vmalloc.h>
16 #include <linux/netdevice.h>
17 #include <linux/module.h>
18 #include <linux/icmp.h>
19 #include <net/ip.h>
20 #include <net/compat.h>
21 #include <asm/uaccess.h>
22 #include <linux/mutex.h>
23 #include <linux/proc_fs.h>
24 #include <linux/err.h>
25 #include <linux/cpumask.h>
26
27 #include <linux/netfilter/x_tables.h>
28 #include <linux/netfilter_ipv4/ip_tables.h>
29 #include <net/netfilter/nf_log.h>
30
31 MODULE_LICENSE("GPL");
32 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
33 MODULE_DESCRIPTION("IPv4 packet filter");
34
35 /*#define DEBUG_IP_FIREWALL*/
36 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
37 /*#define DEBUG_IP_FIREWALL_USER*/
38
39 #ifdef DEBUG_IP_FIREWALL
40 #define dprintf(format, args...)  printk(format , ## args)
41 #else
42 #define dprintf(format, args...)
43 #endif
44
45 #ifdef DEBUG_IP_FIREWALL_USER
46 #define duprintf(format, args...) printk(format , ## args)
47 #else
48 #define duprintf(format, args...)
49 #endif
50
51 #ifdef CONFIG_NETFILTER_DEBUG
52 #define IP_NF_ASSERT(x)                                         \
53 do {                                                            \
54         if (!(x))                                               \
55                 printk("IP_NF_ASSERT: %s:%s:%u\n",              \
56                        __FUNCTION__, __FILE__, __LINE__);       \
57 } while(0)
58 #else
59 #define IP_NF_ASSERT(x)
60 #endif
61
62 #if 0
63 /* All the better to debug you with... */
64 #define static
65 #define inline
66 #endif
67
68 /*
69    We keep a set of rules for each CPU, so we can avoid write-locking
70    them in the softirq when updating the counters and therefore
71    only need to read-lock in the softirq; doing a write_lock_bh() in user
72    context stops packets coming through and allows user context to read
73    the counters or update the rules.
74
75    Hence the start of any table is given by get_table() below.  */
76
77 /* Returns whether matches rule or not. */
78 static inline bool
79 ip_packet_match(const struct iphdr *ip,
80                 const char *indev,
81                 const char *outdev,
82                 const struct ipt_ip *ipinfo,
83                 int isfrag)
84 {
85         size_t i;
86         unsigned long ret;
87
88 #define FWINV(bool,invflg) ((bool) ^ !!(ipinfo->invflags & invflg))
89
90         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
91                   IPT_INV_SRCIP)
92             || FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
93                      IPT_INV_DSTIP)) {
94                 dprintf("Source or dest mismatch.\n");
95
96                 dprintf("SRC: %u.%u.%u.%u. Mask: %u.%u.%u.%u. Target: %u.%u.%u.%u.%s\n",
97                         NIPQUAD(ip->saddr),
98                         NIPQUAD(ipinfo->smsk.s_addr),
99                         NIPQUAD(ipinfo->src.s_addr),
100                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
101                 dprintf("DST: %u.%u.%u.%u Mask: %u.%u.%u.%u Target: %u.%u.%u.%u.%s\n",
102                         NIPQUAD(ip->daddr),
103                         NIPQUAD(ipinfo->dmsk.s_addr),
104                         NIPQUAD(ipinfo->dst.s_addr),
105                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
106                 return false;
107         }
108
109         /* Look for ifname matches; this should unroll nicely. */
110         for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
111                 ret |= (((const unsigned long *)indev)[i]
112                         ^ ((const unsigned long *)ipinfo->iniface)[i])
113                         & ((const unsigned long *)ipinfo->iniface_mask)[i];
114         }
115
116         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
117                 dprintf("VIA in mismatch (%s vs %s).%s\n",
118                         indev, ipinfo->iniface,
119                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
120                 return false;
121         }
122
123         for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
124                 ret |= (((const unsigned long *)outdev)[i]
125                         ^ ((const unsigned long *)ipinfo->outiface)[i])
126                         & ((const unsigned long *)ipinfo->outiface_mask)[i];
127         }
128
129         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
130                 dprintf("VIA out mismatch (%s vs %s).%s\n",
131                         outdev, ipinfo->outiface,
132                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
133                 return false;
134         }
135
136         /* Check specific protocol */
137         if (ipinfo->proto
138             && FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
139                 dprintf("Packet protocol %hi does not match %hi.%s\n",
140                         ip->protocol, ipinfo->proto,
141                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
142                 return false;
143         }
144
145         /* If we have a fragment rule but the packet is not a fragment
146          * then we return zero */
147         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
148                 dprintf("Fragment rule but not fragment.%s\n",
149                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
150                 return false;
151         }
152
153         return true;
154 }
155
156 static inline bool
157 ip_checkentry(const struct ipt_ip *ip)
158 {
159         if (ip->flags & ~IPT_F_MASK) {
160                 duprintf("Unknown flag bits set: %08X\n",
161                          ip->flags & ~IPT_F_MASK);
162                 return false;
163         }
164         if (ip->invflags & ~IPT_INV_MASK) {
165                 duprintf("Unknown invflag bits set: %08X\n",
166                          ip->invflags & ~IPT_INV_MASK);
167                 return false;
168         }
169         return true;
170 }
171
172 static unsigned int
173 ipt_error(struct sk_buff *skb,
174           const struct net_device *in,
175           const struct net_device *out,
176           unsigned int hooknum,
177           const struct xt_target *target,
178           const void *targinfo)
179 {
180         if (net_ratelimit())
181                 printk("ip_tables: error: `%s'\n", (char *)targinfo);
182
183         return NF_DROP;
184 }
185
186 static inline
187 bool do_match(struct ipt_entry_match *m,
188               const struct sk_buff *skb,
189               const struct net_device *in,
190               const struct net_device *out,
191               int offset,
192               bool *hotdrop)
193 {
194         /* Stop iteration if it doesn't match */
195         if (!m->u.kernel.match->match(skb, in, out, m->u.kernel.match, m->data,
196                                       offset, ip_hdrlen(skb), hotdrop))
197                 return true;
198         else
199                 return false;
200 }
201
202 static inline struct ipt_entry *
203 get_entry(void *base, unsigned int offset)
204 {
205         return (struct ipt_entry *)(base + offset);
206 }
207
208 /* All zeroes == unconditional rule. */
209 static inline int
210 unconditional(const struct ipt_ip *ip)
211 {
212         unsigned int i;
213
214         for (i = 0; i < sizeof(*ip)/sizeof(__u32); i++)
215                 if (((__u32 *)ip)[i])
216                         return 0;
217
218         return 1;
219 }
220
221 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
222     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
223 static const char *hooknames[] = {
224         [NF_INET_PRE_ROUTING]           = "PREROUTING",
225         [NF_INET_LOCAL_IN]              = "INPUT",
226         [NF_INET_FORWARD]               = "FORWARD",
227         [NF_INET_LOCAL_OUT]             = "OUTPUT",
228         [NF_INET_POST_ROUTING]          = "POSTROUTING",
229 };
230
231 enum nf_ip_trace_comments {
232         NF_IP_TRACE_COMMENT_RULE,
233         NF_IP_TRACE_COMMENT_RETURN,
234         NF_IP_TRACE_COMMENT_POLICY,
235 };
236
237 static const char *comments[] = {
238         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
239         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
240         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
241 };
242
243 static struct nf_loginfo trace_loginfo = {
244         .type = NF_LOG_TYPE_LOG,
245         .u = {
246                 .log = {
247                         .level = 4,
248                         .logflags = NF_LOG_MASK,
249                 },
250         },
251 };
252
253 static inline int
254 get_chainname_rulenum(struct ipt_entry *s, struct ipt_entry *e,
255                       char *hookname, char **chainname,
256                       char **comment, unsigned int *rulenum)
257 {
258         struct ipt_standard_target *t = (void *)ipt_get_target(s);
259
260         if (strcmp(t->target.u.kernel.target->name, IPT_ERROR_TARGET) == 0) {
261                 /* Head of user chain: ERROR target with chainname */
262                 *chainname = t->target.data;
263                 (*rulenum) = 0;
264         } else if (s == e) {
265                 (*rulenum)++;
266
267                 if (s->target_offset == sizeof(struct ipt_entry)
268                    && strcmp(t->target.u.kernel.target->name,
269                              IPT_STANDARD_TARGET) == 0
270                    && t->verdict < 0
271                    && unconditional(&s->ip)) {
272                         /* Tail of chains: STANDARD target (return/policy) */
273                         *comment = *chainname == hookname
274                                 ? (char *)comments[NF_IP_TRACE_COMMENT_POLICY]
275                                 : (char *)comments[NF_IP_TRACE_COMMENT_RETURN];
276                 }
277                 return 1;
278         } else
279                 (*rulenum)++;
280
281         return 0;
282 }
283
284 static void trace_packet(struct sk_buff *skb,
285                          unsigned int hook,
286                          const struct net_device *in,
287                          const struct net_device *out,
288                          char *tablename,
289                          struct xt_table_info *private,
290                          struct ipt_entry *e)
291 {
292         void *table_base;
293         struct ipt_entry *root;
294         char *hookname, *chainname, *comment;
295         unsigned int rulenum = 0;
296
297         table_base = (void *)private->entries[smp_processor_id()];
298         root = get_entry(table_base, private->hook_entry[hook]);
299
300         hookname = chainname = (char *)hooknames[hook];
301         comment = (char *)comments[NF_IP_TRACE_COMMENT_RULE];
302
303         IPT_ENTRY_ITERATE(root,
304                           private->size - private->hook_entry[hook],
305                           get_chainname_rulenum,
306                           e, hookname, &chainname, &comment, &rulenum);
307
308         nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
309                       "TRACE: %s:%s:%s:%u ",
310                       tablename, chainname, comment, rulenum);
311 }
312 #endif
313
314 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
315 unsigned int
316 ipt_do_table(struct sk_buff *skb,
317              unsigned int hook,
318              const struct net_device *in,
319              const struct net_device *out,
320              struct xt_table *table)
321 {
322         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
323         u_int16_t offset;
324         struct iphdr *ip;
325         u_int16_t datalen;
326         bool hotdrop = false;
327         /* Initializing verdict to NF_DROP keeps gcc happy. */
328         unsigned int verdict = NF_DROP;
329         const char *indev, *outdev;
330         void *table_base;
331         struct ipt_entry *e, *back;
332         struct xt_table_info *private;
333
334         /* Initialization */
335         ip = ip_hdr(skb);
336         datalen = skb->len - ip->ihl * 4;
337         indev = in ? in->name : nulldevname;
338         outdev = out ? out->name : nulldevname;
339         /* We handle fragments by dealing with the first fragment as
340          * if it was a normal packet.  All other fragments are treated
341          * normally, except that they will NEVER match rules that ask
342          * things we don't know, ie. tcp syn flag or ports).  If the
343          * rule is also a fragment-specific rule, non-fragments won't
344          * match it. */
345         offset = ntohs(ip->frag_off) & IP_OFFSET;
346
347         read_lock_bh(&table->lock);
348         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
349         private = table->private;
350         table_base = (void *)private->entries[smp_processor_id()];
351         e = get_entry(table_base, private->hook_entry[hook]);
352
353         /* For return from builtin chain */
354         back = get_entry(table_base, private->underflow[hook]);
355
356         do {
357                 IP_NF_ASSERT(e);
358                 IP_NF_ASSERT(back);
359                 if (ip_packet_match(ip, indev, outdev, &e->ip, offset)) {
360                         struct ipt_entry_target *t;
361
362                         if (IPT_MATCH_ITERATE(e, do_match,
363                                               skb, in, out,
364                                               offset, &hotdrop) != 0)
365                                 goto no_match;
366
367                         ADD_COUNTER(e->counters, ntohs(ip->tot_len), 1);
368
369                         t = ipt_get_target(e);
370                         IP_NF_ASSERT(t->u.kernel.target);
371
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);
378 #endif
379                         /* Standard target? */
380                         if (!t->u.kernel.target->target) {
381                                 int v;
382
383                                 v = ((struct ipt_standard_target *)t)->verdict;
384                                 if (v < 0) {
385                                         /* Pop from stack? */
386                                         if (v != IPT_RETURN) {
387                                                 verdict = (unsigned)(-v) - 1;
388                                                 break;
389                                         }
390                                         e = back;
391                                         back = get_entry(table_base,
392                                                          back->comefrom);
393                                         continue;
394                                 }
395                                 if (table_base + v != (void *)e + e->next_offset
396                                     && !(e->ip.flags & IPT_F_GOTO)) {
397                                         /* Save old back ptr in next entry */
398                                         struct ipt_entry *next
399                                                 = (void *)e + e->next_offset;
400                                         next->comefrom
401                                                 = (void *)back - table_base;
402                                         /* set back pointer to next entry */
403                                         back = next;
404                                 }
405
406                                 e = get_entry(table_base, v);
407                         } else {
408                                 /* Targets which reenter must return
409                                    abs. verdicts */
410 #ifdef CONFIG_NETFILTER_DEBUG
411                                 ((struct ipt_entry *)table_base)->comefrom
412                                         = 0xeeeeeeec;
413 #endif
414                                 verdict = t->u.kernel.target->target(skb,
415                                                                      in, out,
416                                                                      hook,
417                                                                      t->u.kernel.target,
418                                                                      t->data);
419
420 #ifdef CONFIG_NETFILTER_DEBUG
421                                 if (((struct ipt_entry *)table_base)->comefrom
422                                     != 0xeeeeeeec
423                                     && verdict == IPT_CONTINUE) {
424                                         printk("Target %s reentered!\n",
425                                                t->u.kernel.target->name);
426                                         verdict = NF_DROP;
427                                 }
428                                 ((struct ipt_entry *)table_base)->comefrom
429                                         = 0x57acc001;
430 #endif
431                                 /* Target might have changed stuff. */
432                                 ip = ip_hdr(skb);
433                                 datalen = skb->len - ip->ihl * 4;
434
435                                 if (verdict == IPT_CONTINUE)
436                                         e = (void *)e + e->next_offset;
437                                 else
438                                         /* Verdict */
439                                         break;
440                         }
441                 } else {
442
443                 no_match:
444                         e = (void *)e + e->next_offset;
445                 }
446         } while (!hotdrop);
447
448         read_unlock_bh(&table->lock);
449
450 #ifdef DEBUG_ALLOW_ALL
451         return NF_ACCEPT;
452 #else
453         if (hotdrop)
454                 return NF_DROP;
455         else return verdict;
456 #endif
457 }
458
459 /* Figures out from what hook each rule can be called: returns 0 if
460    there are loops.  Puts hook bitmask in comefrom. */
461 static int
462 mark_source_chains(struct xt_table_info *newinfo,
463                    unsigned int valid_hooks, void *entry0)
464 {
465         unsigned int hook;
466
467         /* No recursion; use packet counter to save back ptrs (reset
468            to 0 as we leave), and comefrom to save source hook bitmask */
469         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
470                 unsigned int pos = newinfo->hook_entry[hook];
471                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
472
473                 if (!(valid_hooks & (1 << hook)))
474                         continue;
475
476                 /* Set initial back pointer. */
477                 e->counters.pcnt = pos;
478
479                 for (;;) {
480                         struct ipt_standard_target *t
481                                 = (void *)ipt_get_target(e);
482                         int visited = e->comefrom & (1 << hook);
483
484                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
485                                 printk("iptables: loop hook %u pos %u %08X.\n",
486                                        hook, pos, e->comefrom);
487                                 return 0;
488                         }
489                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
490
491                         /* Unconditional return/END. */
492                         if ((e->target_offset == sizeof(struct ipt_entry)
493                             && (strcmp(t->target.u.user.name,
494                                        IPT_STANDARD_TARGET) == 0)
495                             && t->verdict < 0
496                             && unconditional(&e->ip)) || visited) {
497                                 unsigned int oldpos, size;
498
499                                 if (t->verdict < -NF_MAX_VERDICT - 1) {
500                                         duprintf("mark_source_chains: bad "
501                                                 "negative verdict (%i)\n",
502                                                                 t->verdict);
503                                         return 0;
504                                 }
505
506                                 /* Return: backtrack through the last
507                                    big jump. */
508                                 do {
509                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
510 #ifdef DEBUG_IP_FIREWALL_USER
511                                         if (e->comefrom
512                                             & (1 << NF_INET_NUMHOOKS)) {
513                                                 duprintf("Back unset "
514                                                          "on hook %u "
515                                                          "rule %u\n",
516                                                          hook, pos);
517                                         }
518 #endif
519                                         oldpos = pos;
520                                         pos = e->counters.pcnt;
521                                         e->counters.pcnt = 0;
522
523                                         /* We're at the start. */
524                                         if (pos == oldpos)
525                                                 goto next;
526
527                                         e = (struct ipt_entry *)
528                                                 (entry0 + pos);
529                                 } while (oldpos == pos + e->next_offset);
530
531                                 /* Move along one */
532                                 size = e->next_offset;
533                                 e = (struct ipt_entry *)
534                                         (entry0 + pos + size);
535                                 e->counters.pcnt = pos;
536                                 pos += size;
537                         } else {
538                                 int newpos = t->verdict;
539
540                                 if (strcmp(t->target.u.user.name,
541                                            IPT_STANDARD_TARGET) == 0
542                                     && newpos >= 0) {
543                                         if (newpos > newinfo->size -
544                                                 sizeof(struct ipt_entry)) {
545                                                 duprintf("mark_source_chains: "
546                                                         "bad verdict (%i)\n",
547                                                                 newpos);
548                                                 return 0;
549                                         }
550                                         /* This a jump; chase it. */
551                                         duprintf("Jump rule %u -> %u\n",
552                                                  pos, newpos);
553                                 } else {
554                                         /* ... this is a fallthru */
555                                         newpos = pos + e->next_offset;
556                                 }
557                                 e = (struct ipt_entry *)
558                                         (entry0 + newpos);
559                                 e->counters.pcnt = pos;
560                                 pos = newpos;
561                         }
562                 }
563                 next:
564                 duprintf("Finished chain %u\n", hook);
565         }
566         return 1;
567 }
568
569 static inline int
570 cleanup_match(struct ipt_entry_match *m, unsigned int *i)
571 {
572         if (i && (*i)-- == 0)
573                 return 1;
574
575         if (m->u.kernel.match->destroy)
576                 m->u.kernel.match->destroy(m->u.kernel.match, m->data);
577         module_put(m->u.kernel.match->me);
578         return 0;
579 }
580
581 static inline int
582 check_entry(struct ipt_entry *e, const char *name)
583 {
584         struct ipt_entry_target *t;
585
586         if (!ip_checkentry(&e->ip)) {
587                 duprintf("ip_tables: ip check failed %p %s.\n", e, name);
588                 return -EINVAL;
589         }
590
591         if (e->target_offset + sizeof(struct ipt_entry_target) >
592             e->next_offset)
593                 return -EINVAL;
594
595         t = ipt_get_target(e);
596         if (e->target_offset + t->u.target_size > e->next_offset)
597                 return -EINVAL;
598
599         return 0;
600 }
601
602 static inline int check_match(struct ipt_entry_match *m, const char *name,
603                               const struct ipt_ip *ip,
604                               unsigned int hookmask, unsigned int *i)
605 {
606         struct xt_match *match;
607         int ret;
608
609         match = m->u.kernel.match;
610         ret = xt_check_match(match, AF_INET, m->u.match_size - sizeof(*m),
611                              name, hookmask, ip->proto,
612                              ip->invflags & IPT_INV_PROTO);
613         if (!ret && m->u.kernel.match->checkentry
614             && !m->u.kernel.match->checkentry(name, ip, match, m->data,
615                                               hookmask)) {
616                 duprintf("ip_tables: check failed for `%s'.\n",
617                          m->u.kernel.match->name);
618                 ret = -EINVAL;
619         }
620         if (!ret)
621                 (*i)++;
622         return ret;
623 }
624
625 static inline int
626 find_check_match(struct ipt_entry_match *m,
627                  const char *name,
628                  const struct ipt_ip *ip,
629                  unsigned int hookmask,
630                  unsigned int *i)
631 {
632         struct xt_match *match;
633         int ret;
634
635         match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
636                                                       m->u.user.revision),
637                                         "ipt_%s", m->u.user.name);
638         if (IS_ERR(match) || !match) {
639                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
640                 return match ? PTR_ERR(match) : -ENOENT;
641         }
642         m->u.kernel.match = match;
643
644         ret = check_match(m, name, ip, hookmask, i);
645         if (ret)
646                 goto err;
647
648         return 0;
649 err:
650         module_put(m->u.kernel.match->me);
651         return ret;
652 }
653
654 static inline int check_target(struct ipt_entry *e, const char *name)
655 {
656         struct ipt_entry_target *t;
657         struct xt_target *target;
658         int ret;
659
660         t = ipt_get_target(e);
661         target = t->u.kernel.target;
662         ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
663                               name, e->comefrom, e->ip.proto,
664                               e->ip.invflags & IPT_INV_PROTO);
665         if (!ret && t->u.kernel.target->checkentry
666             && !t->u.kernel.target->checkentry(name, e, target, t->data,
667                                                e->comefrom)) {
668                 duprintf("ip_tables: check failed for `%s'.\n",
669                          t->u.kernel.target->name);
670                 ret = -EINVAL;
671         }
672         return ret;
673 }
674
675 static inline int
676 find_check_entry(struct ipt_entry *e, const char *name, unsigned int size,
677                  unsigned int *i)
678 {
679         struct ipt_entry_target *t;
680         struct xt_target *target;
681         int ret;
682         unsigned int j;
683
684         ret = check_entry(e, name);
685         if (ret)
686                 return ret;
687
688         j = 0;
689         ret = IPT_MATCH_ITERATE(e, find_check_match, name, &e->ip,
690                                 e->comefrom, &j);
691         if (ret != 0)
692                 goto cleanup_matches;
693
694         t = ipt_get_target(e);
695         target = try_then_request_module(xt_find_target(AF_INET,
696                                                         t->u.user.name,
697                                                         t->u.user.revision),
698                                          "ipt_%s", t->u.user.name);
699         if (IS_ERR(target) || !target) {
700                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
701                 ret = target ? PTR_ERR(target) : -ENOENT;
702                 goto cleanup_matches;
703         }
704         t->u.kernel.target = target;
705
706         ret = check_target(e, name);
707         if (ret)
708                 goto err;
709
710         (*i)++;
711         return 0;
712  err:
713         module_put(t->u.kernel.target->me);
714  cleanup_matches:
715         IPT_MATCH_ITERATE(e, cleanup_match, &j);
716         return ret;
717 }
718
719 static inline int
720 check_entry_size_and_hooks(struct ipt_entry *e,
721                            struct xt_table_info *newinfo,
722                            unsigned char *base,
723                            unsigned char *limit,
724                            const unsigned int *hook_entries,
725                            const unsigned int *underflows,
726                            unsigned int *i)
727 {
728         unsigned int h;
729
730         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0
731             || (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
732                 duprintf("Bad offset %p\n", e);
733                 return -EINVAL;
734         }
735
736         if (e->next_offset
737             < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) {
738                 duprintf("checking: element %p size %u\n",
739                          e, e->next_offset);
740                 return -EINVAL;
741         }
742
743         /* Check hooks & underflows */
744         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
745                 if ((unsigned char *)e - base == hook_entries[h])
746                         newinfo->hook_entry[h] = hook_entries[h];
747                 if ((unsigned char *)e - base == underflows[h])
748                         newinfo->underflow[h] = underflows[h];
749         }
750
751         /* FIXME: underflows must be unconditional, standard verdicts
752            < 0 (not IPT_RETURN). --RR */
753
754         /* Clear counters and comefrom */
755         e->counters = ((struct xt_counters) { 0, 0 });
756         e->comefrom = 0;
757
758         (*i)++;
759         return 0;
760 }
761
762 static inline int
763 cleanup_entry(struct ipt_entry *e, unsigned int *i)
764 {
765         struct ipt_entry_target *t;
766
767         if (i && (*i)-- == 0)
768                 return 1;
769
770         /* Cleanup all matches */
771         IPT_MATCH_ITERATE(e, cleanup_match, NULL);
772         t = ipt_get_target(e);
773         if (t->u.kernel.target->destroy)
774                 t->u.kernel.target->destroy(t->u.kernel.target, t->data);
775         module_put(t->u.kernel.target->me);
776         return 0;
777 }
778
779 /* Checks and translates the user-supplied table segment (held in
780    newinfo) */
781 static int
782 translate_table(const char *name,
783                 unsigned int valid_hooks,
784                 struct xt_table_info *newinfo,
785                 void *entry0,
786                 unsigned int size,
787                 unsigned int number,
788                 const unsigned int *hook_entries,
789                 const unsigned int *underflows)
790 {
791         unsigned int i;
792         int ret;
793
794         newinfo->size = size;
795         newinfo->number = number;
796
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;
801         }
802
803         duprintf("translate_table: size %u\n", newinfo->size);
804         i = 0;
805         /* Walk through entries, checking offsets. */
806         ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
807                                 check_entry_size_and_hooks,
808                                 newinfo,
809                                 entry0,
810                                 entry0 + size,
811                                 hook_entries, underflows, &i);
812         if (ret != 0)
813                 return ret;
814
815         if (i != number) {
816                 duprintf("translate_table: %u not %u entries\n",
817                          i, number);
818                 return -EINVAL;
819         }
820
821         /* Check hooks all assigned */
822         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
823                 /* Only hooks which are valid */
824                 if (!(valid_hooks & (1 << i)))
825                         continue;
826                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
827                         duprintf("Invalid hook entry %u %u\n",
828                                  i, hook_entries[i]);
829                         return -EINVAL;
830                 }
831                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
832                         duprintf("Invalid underflow %u %u\n",
833                                  i, underflows[i]);
834                         return -EINVAL;
835                 }
836         }
837
838         if (!mark_source_chains(newinfo, valid_hooks, entry0))
839                 return -ELOOP;
840
841         /* Finally, each sanity check must pass */
842         i = 0;
843         ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
844                                 find_check_entry, name, size, &i);
845
846         if (ret != 0) {
847                 IPT_ENTRY_ITERATE(entry0, newinfo->size,
848                                 cleanup_entry, &i);
849                 return ret;
850         }
851
852         /* And one copy for every other CPU */
853         for_each_possible_cpu(i) {
854                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
855                         memcpy(newinfo->entries[i], entry0, newinfo->size);
856         }
857
858         return ret;
859 }
860
861 /* Gets counters. */
862 static inline int
863 add_entry_to_counter(const struct ipt_entry *e,
864                      struct xt_counters total[],
865                      unsigned int *i)
866 {
867         ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
868
869         (*i)++;
870         return 0;
871 }
872
873 static inline int
874 set_entry_to_counter(const struct ipt_entry *e,
875                      struct ipt_counters total[],
876                      unsigned int *i)
877 {
878         SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
879
880         (*i)++;
881         return 0;
882 }
883
884 static void
885 get_counters(const struct xt_table_info *t,
886              struct xt_counters counters[])
887 {
888         unsigned int cpu;
889         unsigned int i;
890         unsigned int curcpu;
891
892         /* Instead of clearing (by a previous call to memset())
893          * the counters and using adds, we set the counters
894          * with data used by 'current' CPU
895          * We dont care about preemption here.
896          */
897         curcpu = raw_smp_processor_id();
898
899         i = 0;
900         IPT_ENTRY_ITERATE(t->entries[curcpu],
901                           t->size,
902                           set_entry_to_counter,
903                           counters,
904                           &i);
905
906         for_each_possible_cpu(cpu) {
907                 if (cpu == curcpu)
908                         continue;
909                 i = 0;
910                 IPT_ENTRY_ITERATE(t->entries[cpu],
911                                   t->size,
912                                   add_entry_to_counter,
913                                   counters,
914                                   &i);
915         }
916 }
917
918 static inline struct xt_counters * alloc_counters(struct xt_table *table)
919 {
920         unsigned int countersize;
921         struct xt_counters *counters;
922         struct xt_table_info *private = table->private;
923
924         /* We need atomic snapshot of counters: rest doesn't change
925            (other than comefrom, which userspace doesn't care
926            about). */
927         countersize = sizeof(struct xt_counters) * private->number;
928         counters = vmalloc_node(countersize, numa_node_id());
929
930         if (counters == NULL)
931                 return ERR_PTR(-ENOMEM);
932
933         /* First, sum counters... */
934         write_lock_bh(&table->lock);
935         get_counters(private, counters);
936         write_unlock_bh(&table->lock);
937
938         return counters;
939 }
940
941 static int
942 copy_entries_to_user(unsigned int total_size,
943                      struct xt_table *table,
944                      void __user *userptr)
945 {
946         unsigned int off, num;
947         struct ipt_entry *e;
948         struct xt_counters *counters;
949         struct xt_table_info *private = table->private;
950         int ret = 0;
951         void *loc_cpu_entry;
952
953         counters = alloc_counters(table);
954         if (IS_ERR(counters))
955                 return PTR_ERR(counters);
956
957         /* choose the copy that is on our node/cpu, ...
958          * This choice is lazy (because current thread is
959          * allowed to migrate to another cpu)
960          */
961         loc_cpu_entry = private->entries[raw_smp_processor_id()];
962         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
963                 ret = -EFAULT;
964                 goto free_counters;
965         }
966
967         /* FIXME: use iterator macros --RR */
968         /* ... then go back and fix counters and names */
969         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
970                 unsigned int i;
971                 struct ipt_entry_match *m;
972                 struct ipt_entry_target *t;
973
974                 e = (struct ipt_entry *)(loc_cpu_entry + off);
975                 if (copy_to_user(userptr + off
976                                  + offsetof(struct ipt_entry, counters),
977                                  &counters[num],
978                                  sizeof(counters[num])) != 0) {
979                         ret = -EFAULT;
980                         goto free_counters;
981                 }
982
983                 for (i = sizeof(struct ipt_entry);
984                      i < e->target_offset;
985                      i += m->u.match_size) {
986                         m = (void *)e + i;
987
988                         if (copy_to_user(userptr + off + i
989                                          + offsetof(struct ipt_entry_match,
990                                                     u.user.name),
991                                          m->u.kernel.match->name,
992                                          strlen(m->u.kernel.match->name)+1)
993                             != 0) {
994                                 ret = -EFAULT;
995                                 goto free_counters;
996                         }
997                 }
998
999                 t = ipt_get_target(e);
1000                 if (copy_to_user(userptr + off + e->target_offset
1001                                  + offsetof(struct ipt_entry_target,
1002                                             u.user.name),
1003                                  t->u.kernel.target->name,
1004                                  strlen(t->u.kernel.target->name)+1) != 0) {
1005                         ret = -EFAULT;
1006                         goto free_counters;
1007                 }
1008         }
1009
1010  free_counters:
1011         vfree(counters);
1012         return ret;
1013 }
1014
1015 #ifdef CONFIG_COMPAT
1016 static void compat_standard_from_user(void *dst, void *src)
1017 {
1018         int v = *(compat_int_t *)src;
1019
1020         if (v > 0)
1021                 v += xt_compat_calc_jump(AF_INET, v);
1022         memcpy(dst, &v, sizeof(v));
1023 }
1024
1025 static int compat_standard_to_user(void __user *dst, void *src)
1026 {
1027         compat_int_t cv = *(int *)src;
1028
1029         if (cv > 0)
1030                 cv -= xt_compat_calc_jump(AF_INET, cv);
1031         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1032 }
1033
1034 static inline int
1035 compat_calc_match(struct ipt_entry_match *m, int *size)
1036 {
1037         *size += xt_compat_match_offset(m->u.kernel.match);
1038         return 0;
1039 }
1040
1041 static int compat_calc_entry(struct ipt_entry *e,
1042                              const struct xt_table_info *info,
1043                              void *base, struct xt_table_info *newinfo)
1044 {
1045         struct ipt_entry_target *t;
1046         unsigned int entry_offset;
1047         int off, i, ret;
1048
1049         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1050         entry_offset = (void *)e - base;
1051         IPT_MATCH_ITERATE(e, compat_calc_match, &off);
1052         t = ipt_get_target(e);
1053         off += xt_compat_target_offset(t->u.kernel.target);
1054         newinfo->size -= off;
1055         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1056         if (ret)
1057                 return ret;
1058
1059         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1060                 if (info->hook_entry[i] &&
1061                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1062                         newinfo->hook_entry[i] -= off;
1063                 if (info->underflow[i] &&
1064                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1065                         newinfo->underflow[i] -= off;
1066         }
1067         return 0;
1068 }
1069
1070 static int compat_table_info(const struct xt_table_info *info,
1071                              struct xt_table_info *newinfo)
1072 {
1073         void *loc_cpu_entry;
1074
1075         if (!newinfo || !info)
1076                 return -EINVAL;
1077
1078         /* we dont care about newinfo->entries[] */
1079         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1080         newinfo->initial_entries = 0;
1081         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1082         return IPT_ENTRY_ITERATE(loc_cpu_entry, info->size,
1083                                  compat_calc_entry, info, loc_cpu_entry,
1084                                  newinfo);
1085 }
1086 #endif
1087
1088 static int get_info(void __user *user, int *len, int compat)
1089 {
1090         char name[IPT_TABLE_MAXNAMELEN];
1091         struct xt_table *t;
1092         int ret;
1093
1094         if (*len != sizeof(struct ipt_getinfo)) {
1095                 duprintf("length %u != %zu\n", *len,
1096                          sizeof(struct ipt_getinfo));
1097                 return -EINVAL;
1098         }
1099
1100         if (copy_from_user(name, user, sizeof(name)) != 0)
1101                 return -EFAULT;
1102
1103         name[IPT_TABLE_MAXNAMELEN-1] = '\0';
1104 #ifdef CONFIG_COMPAT
1105         if (compat)
1106                 xt_compat_lock(AF_INET);
1107 #endif
1108         t = try_then_request_module(xt_find_table_lock(AF_INET, name),
1109                                     "iptable_%s", name);
1110         if (t && !IS_ERR(t)) {
1111                 struct ipt_getinfo info;
1112                 struct xt_table_info *private = t->private;
1113
1114 #ifdef CONFIG_COMPAT
1115                 if (compat) {
1116                         struct xt_table_info tmp;
1117                         ret = compat_table_info(private, &tmp);
1118                         xt_compat_flush_offsets(AF_INET);
1119                         private = &tmp;
1120                 }
1121 #endif
1122                 info.valid_hooks = t->valid_hooks;
1123                 memcpy(info.hook_entry, private->hook_entry,
1124                        sizeof(info.hook_entry));
1125                 memcpy(info.underflow, private->underflow,
1126                        sizeof(info.underflow));
1127                 info.num_entries = private->number;
1128                 info.size = private->size;
1129                 strcpy(info.name, name);
1130
1131                 if (copy_to_user(user, &info, *len) != 0)
1132                         ret = -EFAULT;
1133                 else
1134                         ret = 0;
1135
1136                 xt_table_unlock(t);
1137                 module_put(t->me);
1138         } else
1139                 ret = t ? PTR_ERR(t) : -ENOENT;
1140 #ifdef CONFIG_COMPAT
1141         if (compat)
1142                 xt_compat_unlock(AF_INET);
1143 #endif
1144         return ret;
1145 }
1146
1147 static int
1148 get_entries(struct ipt_get_entries __user *uptr, int *len)
1149 {
1150         int ret;
1151         struct ipt_get_entries get;
1152         struct xt_table *t;
1153
1154         if (*len < sizeof(get)) {
1155                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1156                 return -EINVAL;
1157         }
1158         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1159                 return -EFAULT;
1160         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1161                 duprintf("get_entries: %u != %zu\n",
1162                          *len, sizeof(get) + get.size);
1163                 return -EINVAL;
1164         }
1165
1166         t = xt_find_table_lock(AF_INET, get.name);
1167         if (t && !IS_ERR(t)) {
1168                 struct xt_table_info *private = t->private;
1169                 duprintf("t->private->number = %u\n", private->number);
1170                 if (get.size == private->size)
1171                         ret = copy_entries_to_user(private->size,
1172                                                    t, uptr->entrytable);
1173                 else {
1174                         duprintf("get_entries: I've got %u not %u!\n",
1175                                  private->size, get.size);
1176                         ret = -EINVAL;
1177                 }
1178                 module_put(t->me);
1179                 xt_table_unlock(t);
1180         } else
1181                 ret = t ? PTR_ERR(t) : -ENOENT;
1182
1183         return ret;
1184 }
1185
1186 static int
1187 __do_replace(const char *name, unsigned int valid_hooks,
1188              struct xt_table_info *newinfo, unsigned int num_counters,
1189              void __user *counters_ptr)
1190 {
1191         int ret;
1192         struct xt_table *t;
1193         struct xt_table_info *oldinfo;
1194         struct xt_counters *counters;
1195         void *loc_cpu_old_entry;
1196
1197         ret = 0;
1198         counters = vmalloc(num_counters * sizeof(struct xt_counters));
1199         if (!counters) {
1200                 ret = -ENOMEM;
1201                 goto out;
1202         }
1203
1204         t = try_then_request_module(xt_find_table_lock(AF_INET, name),
1205                                     "iptable_%s", name);
1206         if (!t || IS_ERR(t)) {
1207                 ret = t ? PTR_ERR(t) : -ENOENT;
1208                 goto free_newinfo_counters_untrans;
1209         }
1210
1211         /* You lied! */
1212         if (valid_hooks != t->valid_hooks) {
1213                 duprintf("Valid hook crap: %08X vs %08X\n",
1214                          valid_hooks, t->valid_hooks);
1215                 ret = -EINVAL;
1216                 goto put_module;
1217         }
1218
1219         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1220         if (!oldinfo)
1221                 goto put_module;
1222
1223         /* Update module usage count based on number of rules */
1224         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1225                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1226         if ((oldinfo->number > oldinfo->initial_entries) ||
1227             (newinfo->number <= oldinfo->initial_entries))
1228                 module_put(t->me);
1229         if ((oldinfo->number > oldinfo->initial_entries) &&
1230             (newinfo->number <= oldinfo->initial_entries))
1231                 module_put(t->me);
1232
1233         /* Get the old counters. */
1234         get_counters(oldinfo, counters);
1235         /* Decrease module usage counts and free resource */
1236         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1237         IPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,
1238                           NULL);
1239         xt_free_table_info(oldinfo);
1240         if (copy_to_user(counters_ptr, counters,
1241                          sizeof(struct xt_counters) * num_counters) != 0)
1242                 ret = -EFAULT;
1243         vfree(counters);
1244         xt_table_unlock(t);
1245         return ret;
1246
1247  put_module:
1248         module_put(t->me);
1249         xt_table_unlock(t);
1250  free_newinfo_counters_untrans:
1251         vfree(counters);
1252  out:
1253         return ret;
1254 }
1255
1256 static int
1257 do_replace(void __user *user, unsigned int len)
1258 {
1259         int ret;
1260         struct ipt_replace tmp;
1261         struct xt_table_info *newinfo;
1262         void *loc_cpu_entry;
1263
1264         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1265                 return -EFAULT;
1266
1267         /* overflow check */
1268         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1269                 return -ENOMEM;
1270
1271         newinfo = xt_alloc_table_info(tmp.size);
1272         if (!newinfo)
1273                 return -ENOMEM;
1274
1275         /* choose the copy that is on our node/cpu */
1276         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1277         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1278                            tmp.size) != 0) {
1279                 ret = -EFAULT;
1280                 goto free_newinfo;
1281         }
1282
1283         ret = translate_table(tmp.name, tmp.valid_hooks,
1284                               newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
1285                               tmp.hook_entry, tmp.underflow);
1286         if (ret != 0)
1287                 goto free_newinfo;
1288
1289         duprintf("ip_tables: Translated table\n");
1290
1291         ret = __do_replace(tmp.name, tmp.valid_hooks, newinfo,
1292                            tmp.num_counters, tmp.counters);
1293         if (ret)
1294                 goto free_newinfo_untrans;
1295         return 0;
1296
1297  free_newinfo_untrans:
1298         IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
1299  free_newinfo:
1300         xt_free_table_info(newinfo);
1301         return ret;
1302 }
1303
1304 /* We're lazy, and add to the first CPU; overflow works its fey magic
1305  * and everything is OK. */
1306 static inline int
1307 add_counter_to_entry(struct ipt_entry *e,
1308                      const struct xt_counters addme[],
1309                      unsigned int *i)
1310 {
1311 #if 0
1312         duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n",
1313                  *i,
1314                  (long unsigned int)e->counters.pcnt,
1315                  (long unsigned int)e->counters.bcnt,
1316                  (long unsigned int)addme[*i].pcnt,
1317                  (long unsigned int)addme[*i].bcnt);
1318 #endif
1319
1320         ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
1321
1322         (*i)++;
1323         return 0;
1324 }
1325
1326 static int
1327 do_add_counters(void __user *user, unsigned int len, int compat)
1328 {
1329         unsigned int i;
1330         struct xt_counters_info tmp;
1331         struct xt_counters *paddc;
1332         unsigned int num_counters;
1333         char *name;
1334         int size;
1335         void *ptmp;
1336         struct xt_table *t;
1337         struct xt_table_info *private;
1338         int ret = 0;
1339         void *loc_cpu_entry;
1340 #ifdef CONFIG_COMPAT
1341         struct compat_xt_counters_info compat_tmp;
1342
1343         if (compat) {
1344                 ptmp = &compat_tmp;
1345                 size = sizeof(struct compat_xt_counters_info);
1346         } else
1347 #endif
1348         {
1349                 ptmp = &tmp;
1350                 size = sizeof(struct xt_counters_info);
1351         }
1352
1353         if (copy_from_user(ptmp, user, size) != 0)
1354                 return -EFAULT;
1355
1356 #ifdef CONFIG_COMPAT
1357         if (compat) {
1358                 num_counters = compat_tmp.num_counters;
1359                 name = compat_tmp.name;
1360         } else
1361 #endif
1362         {
1363                 num_counters = tmp.num_counters;
1364                 name = tmp.name;
1365         }
1366
1367         if (len != size + num_counters * sizeof(struct xt_counters))
1368                 return -EINVAL;
1369
1370         paddc = vmalloc_node(len - size, numa_node_id());
1371         if (!paddc)
1372                 return -ENOMEM;
1373
1374         if (copy_from_user(paddc, user + size, len - size) != 0) {
1375                 ret = -EFAULT;
1376                 goto free;
1377         }
1378
1379         t = xt_find_table_lock(AF_INET, name);
1380         if (!t || IS_ERR(t)) {
1381                 ret = t ? PTR_ERR(t) : -ENOENT;
1382                 goto free;
1383         }
1384
1385         write_lock_bh(&t->lock);
1386         private = t->private;
1387         if (private->number != num_counters) {
1388                 ret = -EINVAL;
1389                 goto unlock_up_free;
1390         }
1391
1392         i = 0;
1393         /* Choose the copy that is on our node */
1394         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1395         IPT_ENTRY_ITERATE(loc_cpu_entry,
1396                           private->size,
1397                           add_counter_to_entry,
1398                           paddc,
1399                           &i);
1400  unlock_up_free:
1401         write_unlock_bh(&t->lock);
1402         xt_table_unlock(t);
1403         module_put(t->me);
1404  free:
1405         vfree(paddc);
1406
1407         return ret;
1408 }
1409
1410 #ifdef CONFIG_COMPAT
1411 struct compat_ipt_replace {
1412         char                    name[IPT_TABLE_MAXNAMELEN];
1413         u32                     valid_hooks;
1414         u32                     num_entries;
1415         u32                     size;
1416         u32                     hook_entry[NF_INET_NUMHOOKS];
1417         u32                     underflow[NF_INET_NUMHOOKS];
1418         u32                     num_counters;
1419         compat_uptr_t           counters;       /* struct ipt_counters * */
1420         struct compat_ipt_entry entries[0];
1421 };
1422
1423 static int
1424 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1425                           compat_uint_t *size, struct xt_counters *counters,
1426                           unsigned int *i)
1427 {
1428         struct ipt_entry_target *t;
1429         struct compat_ipt_entry __user *ce;
1430         u_int16_t target_offset, next_offset;
1431         compat_uint_t origsize;
1432         int ret;
1433
1434         ret = -EFAULT;
1435         origsize = *size;
1436         ce = (struct compat_ipt_entry __user *)*dstptr;
1437         if (copy_to_user(ce, e, sizeof(struct ipt_entry)))
1438                 goto out;
1439
1440         if (copy_to_user(&ce->counters, &counters[*i], sizeof(counters[*i])))
1441                 goto out;
1442
1443         *dstptr += sizeof(struct compat_ipt_entry);
1444         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1445
1446         ret = IPT_MATCH_ITERATE(e, xt_compat_match_to_user, dstptr, size);
1447         target_offset = e->target_offset - (origsize - *size);
1448         if (ret)
1449                 goto out;
1450         t = ipt_get_target(e);
1451         ret = xt_compat_target_to_user(t, dstptr, size);
1452         if (ret)
1453                 goto out;
1454         ret = -EFAULT;
1455         next_offset = e->next_offset - (origsize - *size);
1456         if (put_user(target_offset, &ce->target_offset))
1457                 goto out;
1458         if (put_user(next_offset, &ce->next_offset))
1459                 goto out;
1460
1461         (*i)++;
1462         return 0;
1463 out:
1464         return ret;
1465 }
1466
1467 static inline int
1468 compat_find_calc_match(struct ipt_entry_match *m,
1469                        const char *name,
1470                        const struct ipt_ip *ip,
1471                        unsigned int hookmask,
1472                        int *size, int *i)
1473 {
1474         struct xt_match *match;
1475
1476         match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
1477                                                       m->u.user.revision),
1478                                         "ipt_%s", m->u.user.name);
1479         if (IS_ERR(match) || !match) {
1480                 duprintf("compat_check_calc_match: `%s' not found\n",
1481                          m->u.user.name);
1482                 return match ? PTR_ERR(match) : -ENOENT;
1483         }
1484         m->u.kernel.match = match;
1485         *size += xt_compat_match_offset(match);
1486
1487         (*i)++;
1488         return 0;
1489 }
1490
1491 static inline int
1492 compat_release_match(struct ipt_entry_match *m, unsigned int *i)
1493 {
1494         if (i && (*i)-- == 0)
1495                 return 1;
1496
1497         module_put(m->u.kernel.match->me);
1498         return 0;
1499 }
1500
1501 static inline int
1502 compat_release_entry(struct compat_ipt_entry *e, unsigned int *i)
1503 {
1504         struct ipt_entry_target *t;
1505
1506         if (i && (*i)-- == 0)
1507                 return 1;
1508
1509         /* Cleanup all matches */
1510         COMPAT_IPT_MATCH_ITERATE(e, compat_release_match, NULL);
1511         t = compat_ipt_get_target(e);
1512         module_put(t->u.kernel.target->me);
1513         return 0;
1514 }
1515
1516 static inline int
1517 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1518                                   struct xt_table_info *newinfo,
1519                                   unsigned int *size,
1520                                   unsigned char *base,
1521                                   unsigned char *limit,
1522                                   unsigned int *hook_entries,
1523                                   unsigned int *underflows,
1524                                   unsigned int *i,
1525                                   const char *name)
1526 {
1527         struct ipt_entry_target *t;
1528         struct xt_target *target;
1529         unsigned int entry_offset;
1530         int ret, off, h, j;
1531
1532         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1533         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0
1534             || (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1535                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1536                 return -EINVAL;
1537         }
1538
1539         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1540                              sizeof(struct compat_xt_entry_target)) {
1541                 duprintf("checking: element %p size %u\n",
1542                          e, e->next_offset);
1543                 return -EINVAL;
1544         }
1545
1546         /* For purposes of check_entry casting the compat entry is fine */
1547         ret = check_entry((struct ipt_entry *)e, name);
1548         if (ret)
1549                 return ret;
1550
1551         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1552         entry_offset = (void *)e - (void *)base;
1553         j = 0;
1554         ret = COMPAT_IPT_MATCH_ITERATE(e, compat_find_calc_match, name,
1555                                        &e->ip, e->comefrom, &off, &j);
1556         if (ret != 0)
1557                 goto release_matches;
1558
1559         t = compat_ipt_get_target(e);
1560         target = try_then_request_module(xt_find_target(AF_INET,
1561                                                         t->u.user.name,
1562                                                         t->u.user.revision),
1563                                          "ipt_%s", t->u.user.name);
1564         if (IS_ERR(target) || !target) {
1565                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1566                          t->u.user.name);
1567                 ret = target ? PTR_ERR(target) : -ENOENT;
1568                 goto release_matches;
1569         }
1570         t->u.kernel.target = target;
1571
1572         off += xt_compat_target_offset(target);
1573         *size += off;
1574         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1575         if (ret)
1576                 goto out;
1577
1578         /* Check hooks & underflows */
1579         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1580                 if ((unsigned char *)e - base == hook_entries[h])
1581                         newinfo->hook_entry[h] = hook_entries[h];
1582                 if ((unsigned char *)e - base == underflows[h])
1583                         newinfo->underflow[h] = underflows[h];
1584         }
1585
1586         /* Clear counters and comefrom */
1587         memset(&e->counters, 0, sizeof(e->counters));
1588         e->comefrom = 0;
1589
1590         (*i)++;
1591         return 0;
1592
1593 out:
1594         module_put(t->u.kernel.target->me);
1595 release_matches:
1596         IPT_MATCH_ITERATE(e, compat_release_match, &j);
1597         return ret;
1598 }
1599
1600 static int
1601 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1602                             unsigned int *size, const char *name,
1603                             struct xt_table_info *newinfo, unsigned char *base)
1604 {
1605         struct ipt_entry_target *t;
1606         struct xt_target *target;
1607         struct ipt_entry *de;
1608         unsigned int origsize;
1609         int ret, h;
1610
1611         ret = 0;
1612         origsize = *size;
1613         de = (struct ipt_entry *)*dstptr;
1614         memcpy(de, e, sizeof(struct ipt_entry));
1615         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1616
1617         *dstptr += sizeof(struct ipt_entry);
1618         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1619
1620         ret = COMPAT_IPT_MATCH_ITERATE(e, xt_compat_match_from_user,
1621                                        dstptr, size);
1622         if (ret)
1623                 return ret;
1624         de->target_offset = e->target_offset - (origsize - *size);
1625         t = compat_ipt_get_target(e);
1626         target = t->u.kernel.target;
1627         xt_compat_target_from_user(t, dstptr, size);
1628
1629         de->next_offset = e->next_offset - (origsize - *size);
1630         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1631                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1632                         newinfo->hook_entry[h] -= origsize - *size;
1633                 if ((unsigned char *)de - base < newinfo->underflow[h])
1634                         newinfo->underflow[h] -= origsize - *size;
1635         }
1636         return ret;
1637 }
1638
1639 static inline int compat_check_entry(struct ipt_entry *e, const char *name,
1640                                      unsigned int *i)
1641 {
1642         int j, ret;
1643
1644         j = 0;
1645         ret = IPT_MATCH_ITERATE(e, check_match, name, &e->ip,
1646                                 e->comefrom, &j);
1647         if (ret)
1648                 goto cleanup_matches;
1649
1650         ret = check_target(e, name);
1651         if (ret)
1652                 goto cleanup_matches;
1653
1654         (*i)++;
1655         return 0;
1656
1657  cleanup_matches:
1658         IPT_MATCH_ITERATE(e, cleanup_match, &j);
1659         return ret;
1660 }
1661
1662 static int
1663 translate_compat_table(const char *name,
1664                        unsigned int valid_hooks,
1665                        struct xt_table_info **pinfo,
1666                        void **pentry0,
1667                        unsigned int total_size,
1668                        unsigned int number,
1669                        unsigned int *hook_entries,
1670                        unsigned int *underflows)
1671 {
1672         unsigned int i, j;
1673         struct xt_table_info *newinfo, *info;
1674         void *pos, *entry0, *entry1;
1675         unsigned int size;
1676         int ret;
1677
1678         info = *pinfo;
1679         entry0 = *pentry0;
1680         size = total_size;
1681         info->number = number;
1682
1683         /* Init all hooks to impossible value. */
1684         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1685                 info->hook_entry[i] = 0xFFFFFFFF;
1686                 info->underflow[i] = 0xFFFFFFFF;
1687         }
1688
1689         duprintf("translate_compat_table: size %u\n", info->size);
1690         j = 0;
1691         xt_compat_lock(AF_INET);
1692         /* Walk through entries, checking offsets. */
1693         ret = COMPAT_IPT_ENTRY_ITERATE(entry0, total_size,
1694                                        check_compat_entry_size_and_hooks,
1695                                        info, &size, entry0,
1696                                        entry0 + total_size,
1697                                        hook_entries, underflows, &j, name);
1698         if (ret != 0)
1699                 goto out_unlock;
1700
1701         ret = -EINVAL;
1702         if (j != number) {
1703                 duprintf("translate_compat_table: %u not %u entries\n",
1704                          j, number);
1705                 goto out_unlock;
1706         }
1707
1708         /* Check hooks all assigned */
1709         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1710                 /* Only hooks which are valid */
1711                 if (!(valid_hooks & (1 << i)))
1712                         continue;
1713                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1714                         duprintf("Invalid hook entry %u %u\n",
1715                                  i, hook_entries[i]);
1716                         goto out_unlock;
1717                 }
1718                 if (info->underflow[i] == 0xFFFFFFFF) {
1719                         duprintf("Invalid underflow %u %u\n",
1720                                  i, underflows[i]);
1721                         goto out_unlock;
1722                 }
1723         }
1724
1725         ret = -ENOMEM;
1726         newinfo = xt_alloc_table_info(size);
1727         if (!newinfo)
1728                 goto out_unlock;
1729
1730         newinfo->number = number;
1731         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1732                 newinfo->hook_entry[i] = info->hook_entry[i];
1733                 newinfo->underflow[i] = info->underflow[i];
1734         }
1735         entry1 = newinfo->entries[raw_smp_processor_id()];
1736         pos = entry1;
1737         size = total_size;
1738         ret = COMPAT_IPT_ENTRY_ITERATE(entry0, total_size,
1739                                        compat_copy_entry_from_user,
1740                                        &pos, &size, name, newinfo, entry1);
1741         xt_compat_flush_offsets(AF_INET);
1742         xt_compat_unlock(AF_INET);
1743         if (ret)
1744                 goto free_newinfo;
1745
1746         ret = -ELOOP;
1747         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1748                 goto free_newinfo;
1749
1750         i = 0;
1751         ret = IPT_ENTRY_ITERATE(entry1, newinfo->size, compat_check_entry,
1752                                 name, &i);
1753         if (ret) {
1754                 j -= i;
1755                 COMPAT_IPT_ENTRY_ITERATE_CONTINUE(entry0, newinfo->size, i,
1756                                                   compat_release_entry, &j);
1757                 IPT_ENTRY_ITERATE(entry1, newinfo->size, cleanup_entry, &i);
1758                 xt_free_table_info(newinfo);
1759                 return ret;
1760         }
1761
1762         /* And one copy for every other CPU */
1763         for_each_possible_cpu(i)
1764                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1765                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1766
1767         *pinfo = newinfo;
1768         *pentry0 = entry1;
1769         xt_free_table_info(info);
1770         return 0;
1771
1772 free_newinfo:
1773         xt_free_table_info(newinfo);
1774 out:
1775         COMPAT_IPT_ENTRY_ITERATE(entry0, total_size, compat_release_entry, &j);
1776         return ret;
1777 out_unlock:
1778         xt_compat_flush_offsets(AF_INET);
1779         xt_compat_unlock(AF_INET);
1780         goto out;
1781 }
1782
1783 static int
1784 compat_do_replace(void __user *user, unsigned int len)
1785 {
1786         int ret;
1787         struct compat_ipt_replace tmp;
1788         struct xt_table_info *newinfo;
1789         void *loc_cpu_entry;
1790
1791         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1792                 return -EFAULT;
1793
1794         /* overflow check */
1795         if (tmp.size >= INT_MAX / num_possible_cpus())
1796                 return -ENOMEM;
1797         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1798                 return -ENOMEM;
1799
1800         newinfo = xt_alloc_table_info(tmp.size);
1801         if (!newinfo)
1802                 return -ENOMEM;
1803
1804         /* choose the copy that is on our node/cpu */
1805         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1806         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1807                            tmp.size) != 0) {
1808                 ret = -EFAULT;
1809                 goto free_newinfo;
1810         }
1811
1812         ret = translate_compat_table(tmp.name, tmp.valid_hooks,
1813                                      &newinfo, &loc_cpu_entry, tmp.size,
1814                                      tmp.num_entries, tmp.hook_entry,
1815                                      tmp.underflow);
1816         if (ret != 0)
1817                 goto free_newinfo;
1818
1819         duprintf("compat_do_replace: Translated table\n");
1820
1821         ret = __do_replace(tmp.name, tmp.valid_hooks, newinfo,
1822                            tmp.num_counters, compat_ptr(tmp.counters));
1823         if (ret)
1824                 goto free_newinfo_untrans;
1825         return 0;
1826
1827  free_newinfo_untrans:
1828         IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
1829  free_newinfo:
1830         xt_free_table_info(newinfo);
1831         return ret;
1832 }
1833
1834 static int
1835 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1836                       unsigned int len)
1837 {
1838         int ret;
1839
1840         if (!capable(CAP_NET_ADMIN))
1841                 return -EPERM;
1842
1843         switch (cmd) {
1844         case IPT_SO_SET_REPLACE:
1845                 ret = compat_do_replace(user, len);
1846                 break;
1847
1848         case IPT_SO_SET_ADD_COUNTERS:
1849                 ret = do_add_counters(user, len, 1);
1850                 break;
1851
1852         default:
1853                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1854                 ret = -EINVAL;
1855         }
1856
1857         return ret;
1858 }
1859
1860 struct compat_ipt_get_entries {
1861         char name[IPT_TABLE_MAXNAMELEN];
1862         compat_uint_t size;
1863         struct compat_ipt_entry entrytable[0];
1864 };
1865
1866 static int
1867 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1868                             void __user *userptr)
1869 {
1870         struct xt_counters *counters;
1871         struct xt_table_info *private = table->private;
1872         void __user *pos;
1873         unsigned int size;
1874         int ret = 0;
1875         void *loc_cpu_entry;
1876         unsigned int i = 0;
1877
1878         counters = alloc_counters(table);
1879         if (IS_ERR(counters))
1880                 return PTR_ERR(counters);
1881
1882         /* choose the copy that is on our node/cpu, ...
1883          * This choice is lazy (because current thread is
1884          * allowed to migrate to another cpu)
1885          */
1886         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1887         pos = userptr;
1888         size = total_size;
1889         ret = IPT_ENTRY_ITERATE(loc_cpu_entry, total_size,
1890                                 compat_copy_entry_to_user,
1891                                 &pos, &size, counters, &i);
1892
1893         vfree(counters);
1894         return ret;
1895 }
1896
1897 static int
1898 compat_get_entries(struct compat_ipt_get_entries __user *uptr, int *len)
1899 {
1900         int ret;
1901         struct compat_ipt_get_entries get;
1902         struct xt_table *t;
1903
1904         if (*len < sizeof(get)) {
1905                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1906                 return -EINVAL;
1907         }
1908
1909         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1910                 return -EFAULT;
1911
1912         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1913                 duprintf("compat_get_entries: %u != %zu\n",
1914                          *len, sizeof(get) + get.size);
1915                 return -EINVAL;
1916         }
1917
1918         xt_compat_lock(AF_INET);
1919         t = xt_find_table_lock(AF_INET, get.name);
1920         if (t && !IS_ERR(t)) {
1921                 struct xt_table_info *private = t->private;
1922                 struct xt_table_info info;
1923                 duprintf("t->private->number = %u\n", private->number);
1924                 ret = compat_table_info(private, &info);
1925                 if (!ret && get.size == info.size) {
1926                         ret = compat_copy_entries_to_user(private->size,
1927                                                           t, uptr->entrytable);
1928                 } else if (!ret) {
1929                         duprintf("compat_get_entries: I've got %u not %u!\n",
1930                                  private->size, get.size);
1931                         ret = -EINVAL;
1932                 }
1933                 xt_compat_flush_offsets(AF_INET);
1934                 module_put(t->me);
1935                 xt_table_unlock(t);
1936         } else
1937                 ret = t ? PTR_ERR(t) : -ENOENT;
1938
1939         xt_compat_unlock(AF_INET);
1940         return ret;
1941 }
1942
1943 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1944
1945 static int
1946 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1947 {
1948         int ret;
1949
1950         if (!capable(CAP_NET_ADMIN))
1951                 return -EPERM;
1952
1953         switch (cmd) {
1954         case IPT_SO_GET_INFO:
1955                 ret = get_info(user, len, 1);
1956                 break;
1957         case IPT_SO_GET_ENTRIES:
1958                 ret = compat_get_entries(user, len);
1959                 break;
1960         default:
1961                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1962         }
1963         return ret;
1964 }
1965 #endif
1966
1967 static int
1968 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1969 {
1970         int ret;
1971
1972         if (!capable(CAP_NET_ADMIN))
1973                 return -EPERM;
1974
1975         switch (cmd) {
1976         case IPT_SO_SET_REPLACE:
1977                 ret = do_replace(user, len);
1978                 break;
1979
1980         case IPT_SO_SET_ADD_COUNTERS:
1981                 ret = do_add_counters(user, len, 0);
1982                 break;
1983
1984         default:
1985                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1986                 ret = -EINVAL;
1987         }
1988
1989         return ret;
1990 }
1991
1992 static int
1993 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1994 {
1995         int ret;
1996
1997         if (!capable(CAP_NET_ADMIN))
1998                 return -EPERM;
1999
2000         switch (cmd) {
2001         case IPT_SO_GET_INFO:
2002                 ret = get_info(user, len, 0);
2003                 break;
2004
2005         case IPT_SO_GET_ENTRIES:
2006                 ret = get_entries(user, len);
2007                 break;
2008
2009         case IPT_SO_GET_REVISION_MATCH:
2010         case IPT_SO_GET_REVISION_TARGET: {
2011                 struct ipt_get_revision rev;
2012                 int target;
2013
2014                 if (*len != sizeof(rev)) {
2015                         ret = -EINVAL;
2016                         break;
2017                 }
2018                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2019                         ret = -EFAULT;
2020                         break;
2021                 }
2022
2023                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2024                         target = 1;
2025                 else
2026                         target = 0;
2027
2028                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2029                                                          rev.revision,
2030                                                          target, &ret),
2031                                         "ipt_%s", rev.name);
2032                 break;
2033         }
2034
2035         default:
2036                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2037                 ret = -EINVAL;
2038         }
2039
2040         return ret;
2041 }
2042
2043 int ipt_register_table(struct xt_table *table, const struct ipt_replace *repl)
2044 {
2045         int ret;
2046         struct xt_table_info *newinfo;
2047         struct xt_table_info bootstrap
2048                 = { 0, 0, 0, { 0 }, { 0 }, { } };
2049         void *loc_cpu_entry;
2050
2051         newinfo = xt_alloc_table_info(repl->size);
2052         if (!newinfo)
2053                 return -ENOMEM;
2054
2055         /* choose the copy on our node/cpu, but dont care about preemption */
2056         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2057         memcpy(loc_cpu_entry, repl->entries, repl->size);
2058
2059         ret = translate_table(table->name, table->valid_hooks,
2060                               newinfo, loc_cpu_entry, repl->size,
2061                               repl->num_entries,
2062                               repl->hook_entry,
2063                               repl->underflow);
2064         if (ret != 0) {
2065                 xt_free_table_info(newinfo);
2066                 return ret;
2067         }
2068
2069         ret = xt_register_table(table, &bootstrap, newinfo);
2070         if (ret != 0) {
2071                 xt_free_table_info(newinfo);
2072                 return ret;
2073         }
2074
2075         return 0;
2076 }
2077
2078 void ipt_unregister_table(struct xt_table *table)
2079 {
2080         struct xt_table_info *private;
2081         void *loc_cpu_entry;
2082
2083         private = xt_unregister_table(table);
2084
2085         /* Decrease module usage counts and free resources */
2086         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2087         IPT_ENTRY_ITERATE(loc_cpu_entry, private->size, cleanup_entry, NULL);
2088         xt_free_table_info(private);
2089 }
2090
2091 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2092 static inline bool
2093 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2094                      u_int8_t type, u_int8_t code,
2095                      bool invert)
2096 {
2097         return ((test_type == 0xFF) ||
2098                 (type == test_type && code >= min_code && code <= max_code))
2099                 ^ invert;
2100 }
2101
2102 static bool
2103 icmp_match(const struct sk_buff *skb,
2104            const struct net_device *in,
2105            const struct net_device *out,
2106            const struct xt_match *match,
2107            const void *matchinfo,
2108            int offset,
2109            unsigned int protoff,
2110            bool *hotdrop)
2111 {
2112         struct icmphdr _icmph, *ic;
2113         const struct ipt_icmp *icmpinfo = matchinfo;
2114
2115         /* Must not be a fragment. */
2116         if (offset)
2117                 return false;
2118
2119         ic = skb_header_pointer(skb, protoff, sizeof(_icmph), &_icmph);
2120         if (ic == NULL) {
2121                 /* We've been asked to examine this packet, and we
2122                  * can't.  Hence, no choice but to drop.
2123                  */
2124                 duprintf("Dropping evil ICMP tinygram.\n");
2125                 *hotdrop = true;
2126                 return false;
2127         }
2128
2129         return icmp_type_code_match(icmpinfo->type,
2130                                     icmpinfo->code[0],
2131                                     icmpinfo->code[1],
2132                                     ic->type, ic->code,
2133                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2134 }
2135
2136 /* Called when user tries to insert an entry of this type. */
2137 static bool
2138 icmp_checkentry(const char *tablename,
2139            const void *entry,
2140            const struct xt_match *match,
2141            void *matchinfo,
2142            unsigned int hook_mask)
2143 {
2144         const struct ipt_icmp *icmpinfo = matchinfo;
2145
2146         /* Must specify no unknown invflags */
2147         return !(icmpinfo->invflags & ~IPT_ICMP_INV);
2148 }
2149
2150 /* The built-in targets: standard (NULL) and error. */
2151 static struct xt_target ipt_standard_target __read_mostly = {
2152         .name           = IPT_STANDARD_TARGET,
2153         .targetsize     = sizeof(int),
2154         .family         = AF_INET,
2155 #ifdef CONFIG_COMPAT
2156         .compatsize     = sizeof(compat_int_t),
2157         .compat_from_user = compat_standard_from_user,
2158         .compat_to_user = compat_standard_to_user,
2159 #endif
2160 };
2161
2162 static struct xt_target ipt_error_target __read_mostly = {
2163         .name           = IPT_ERROR_TARGET,
2164         .target         = ipt_error,
2165         .targetsize     = IPT_FUNCTION_MAXNAMELEN,
2166         .family         = AF_INET,
2167 };
2168
2169 static struct nf_sockopt_ops ipt_sockopts = {
2170         .pf             = PF_INET,
2171         .set_optmin     = IPT_BASE_CTL,
2172         .set_optmax     = IPT_SO_SET_MAX+1,
2173         .set            = do_ipt_set_ctl,
2174 #ifdef CONFIG_COMPAT
2175         .compat_set     = compat_do_ipt_set_ctl,
2176 #endif
2177         .get_optmin     = IPT_BASE_CTL,
2178         .get_optmax     = IPT_SO_GET_MAX+1,
2179         .get            = do_ipt_get_ctl,
2180 #ifdef CONFIG_COMPAT
2181         .compat_get     = compat_do_ipt_get_ctl,
2182 #endif
2183         .owner          = THIS_MODULE,
2184 };
2185
2186 static struct xt_match icmp_matchstruct __read_mostly = {
2187         .name           = "icmp",
2188         .match          = icmp_match,
2189         .matchsize      = sizeof(struct ipt_icmp),
2190         .checkentry     = icmp_checkentry,
2191         .proto          = IPPROTO_ICMP,
2192         .family         = AF_INET,
2193 };
2194
2195 static int __init ip_tables_init(void)
2196 {
2197         int ret;
2198
2199         ret = xt_proto_init(AF_INET);
2200         if (ret < 0)
2201                 goto err1;
2202
2203         /* Noone else will be downing sem now, so we won't sleep */
2204         ret = xt_register_target(&ipt_standard_target);
2205         if (ret < 0)
2206                 goto err2;
2207         ret = xt_register_target(&ipt_error_target);
2208         if (ret < 0)
2209                 goto err3;
2210         ret = xt_register_match(&icmp_matchstruct);
2211         if (ret < 0)
2212                 goto err4;
2213
2214         /* Register setsockopt */
2215         ret = nf_register_sockopt(&ipt_sockopts);
2216         if (ret < 0)
2217                 goto err5;
2218
2219         printk(KERN_INFO "ip_tables: (C) 2000-2006 Netfilter Core Team\n");
2220         return 0;
2221
2222 err5:
2223         xt_unregister_match(&icmp_matchstruct);
2224 err4:
2225         xt_unregister_target(&ipt_error_target);
2226 err3:
2227         xt_unregister_target(&ipt_standard_target);
2228 err2:
2229         xt_proto_fini(AF_INET);
2230 err1:
2231         return ret;
2232 }
2233
2234 static void __exit ip_tables_fini(void)
2235 {
2236         nf_unregister_sockopt(&ipt_sockopts);
2237
2238         xt_unregister_match(&icmp_matchstruct);
2239         xt_unregister_target(&ipt_error_target);
2240         xt_unregister_target(&ipt_standard_target);
2241
2242         xt_proto_fini(AF_INET);
2243 }
2244
2245 EXPORT_SYMBOL(ipt_register_table);
2246 EXPORT_SYMBOL(ipt_unregister_table);
2247 EXPORT_SYMBOL(ipt_do_table);
2248 module_init(ip_tables_init);
2249 module_exit(ip_tables_fini);