800ae854247163f8890ea244060835eccc5f0a99
[linux-2.6.git] / net / ipv6 / netfilter / ip6_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
12 #include <linux/capability.h>
13 #include <linux/in.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/poison.h>
20 #include <linux/icmpv6.h>
21 #include <net/ipv6.h>
22 #include <net/compat.h>
23 #include <asm/uaccess.h>
24 #include <linux/mutex.h>
25 #include <linux/proc_fs.h>
26 #include <linux/err.h>
27 #include <linux/cpumask.h>
28
29 #include <linux/netfilter_ipv6/ip6_tables.h>
30 #include <linux/netfilter/x_tables.h>
31 #include <net/netfilter/nf_log.h>
32
33 MODULE_LICENSE("GPL");
34 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
35 MODULE_DESCRIPTION("IPv6 packet filter");
36
37 /*#define DEBUG_IP_FIREWALL*/
38 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
39 /*#define DEBUG_IP_FIREWALL_USER*/
40
41 #ifdef DEBUG_IP_FIREWALL
42 #define dprintf(format, args...)  printk(format , ## args)
43 #else
44 #define dprintf(format, args...)
45 #endif
46
47 #ifdef DEBUG_IP_FIREWALL_USER
48 #define duprintf(format, args...) printk(format , ## args)
49 #else
50 #define duprintf(format, args...)
51 #endif
52
53 #ifdef CONFIG_NETFILTER_DEBUG
54 #define IP_NF_ASSERT(x)                                         \
55 do {                                                            \
56         if (!(x))                                               \
57                 printk("IP_NF_ASSERT: %s:%s:%u\n",              \
58                        __func__, __FILE__, __LINE__);   \
59 } while(0)
60 #else
61 #define IP_NF_ASSERT(x)
62 #endif
63
64 #if 0
65 /* All the better to debug you with... */
66 #define static
67 #define inline
68 #endif
69
70 /*
71    We keep a set of rules for each CPU, so we can avoid write-locking
72    them in the softirq when updating the counters and therefore
73    only need to read-lock in the softirq; doing a write_lock_bh() in user
74    context stops packets coming through and allows user context to read
75    the counters or update the rules.
76
77    Hence the start of any table is given by get_table() below.  */
78
79 /* Check for an extension */
80 int
81 ip6t_ext_hdr(u8 nexthdr)
82 {
83         return ( (nexthdr == IPPROTO_HOPOPTS)   ||
84                  (nexthdr == IPPROTO_ROUTING)   ||
85                  (nexthdr == IPPROTO_FRAGMENT)  ||
86                  (nexthdr == IPPROTO_ESP)       ||
87                  (nexthdr == IPPROTO_AH)        ||
88                  (nexthdr == IPPROTO_NONE)      ||
89                  (nexthdr == IPPROTO_DSTOPTS) );
90 }
91
92 /* Returns whether matches rule or not. */
93 /* Performance critical - called for every packet */
94 static inline bool
95 ip6_packet_match(const struct sk_buff *skb,
96                  const char *indev,
97                  const char *outdev,
98                  const struct ip6t_ip6 *ip6info,
99                  unsigned int *protoff,
100                  int *fragoff, bool *hotdrop)
101 {
102         unsigned long ret;
103         const struct ipv6hdr *ipv6 = ipv6_hdr(skb);
104
105 #define FWINV(bool, invflg) ((bool) ^ !!(ip6info->invflags & (invflg)))
106
107         if (FWINV(ipv6_masked_addr_cmp(&ipv6->saddr, &ip6info->smsk,
108                                        &ip6info->src), IP6T_INV_SRCIP)
109             || FWINV(ipv6_masked_addr_cmp(&ipv6->daddr, &ip6info->dmsk,
110                                           &ip6info->dst), IP6T_INV_DSTIP)) {
111                 dprintf("Source or dest mismatch.\n");
112 /*
113                 dprintf("SRC: %u. Mask: %u. Target: %u.%s\n", ip->saddr,
114                         ipinfo->smsk.s_addr, ipinfo->src.s_addr,
115                         ipinfo->invflags & IP6T_INV_SRCIP ? " (INV)" : "");
116                 dprintf("DST: %u. Mask: %u. Target: %u.%s\n", ip->daddr,
117                         ipinfo->dmsk.s_addr, ipinfo->dst.s_addr,
118                         ipinfo->invflags & IP6T_INV_DSTIP ? " (INV)" : "");*/
119                 return false;
120         }
121
122         ret = ifname_compare_aligned(indev, ip6info->iniface, ip6info->iniface_mask);
123
124         if (FWINV(ret != 0, IP6T_INV_VIA_IN)) {
125                 dprintf("VIA in mismatch (%s vs %s).%s\n",
126                         indev, ip6info->iniface,
127                         ip6info->invflags&IP6T_INV_VIA_IN ?" (INV)":"");
128                 return false;
129         }
130
131         ret = ifname_compare_aligned(outdev, ip6info->outiface, ip6info->outiface_mask);
132
133         if (FWINV(ret != 0, IP6T_INV_VIA_OUT)) {
134                 dprintf("VIA out mismatch (%s vs %s).%s\n",
135                         outdev, ip6info->outiface,
136                         ip6info->invflags&IP6T_INV_VIA_OUT ?" (INV)":"");
137                 return false;
138         }
139
140 /* ... might want to do something with class and flowlabel here ... */
141
142         /* look for the desired protocol header */
143         if((ip6info->flags & IP6T_F_PROTO)) {
144                 int protohdr;
145                 unsigned short _frag_off;
146
147                 protohdr = ipv6_find_hdr(skb, protoff, -1, &_frag_off);
148                 if (protohdr < 0) {
149                         if (_frag_off == 0)
150                                 *hotdrop = true;
151                         return false;
152                 }
153                 *fragoff = _frag_off;
154
155                 dprintf("Packet protocol %hi ?= %s%hi.\n",
156                                 protohdr,
157                                 ip6info->invflags & IP6T_INV_PROTO ? "!":"",
158                                 ip6info->proto);
159
160                 if (ip6info->proto == protohdr) {
161                         if(ip6info->invflags & IP6T_INV_PROTO) {
162                                 return false;
163                         }
164                         return true;
165                 }
166
167                 /* We need match for the '-p all', too! */
168                 if ((ip6info->proto != 0) &&
169                         !(ip6info->invflags & IP6T_INV_PROTO))
170                         return false;
171         }
172         return true;
173 }
174
175 /* should be ip6 safe */
176 static bool
177 ip6_checkentry(const struct ip6t_ip6 *ipv6)
178 {
179         if (ipv6->flags & ~IP6T_F_MASK) {
180                 duprintf("Unknown flag bits set: %08X\n",
181                          ipv6->flags & ~IP6T_F_MASK);
182                 return false;
183         }
184         if (ipv6->invflags & ~IP6T_INV_MASK) {
185                 duprintf("Unknown invflag bits set: %08X\n",
186                          ipv6->invflags & ~IP6T_INV_MASK);
187                 return false;
188         }
189         return true;
190 }
191
192 static unsigned int
193 ip6t_error(struct sk_buff *skb, const struct xt_target_param *par)
194 {
195         if (net_ratelimit())
196                 printk("ip6_tables: error: `%s'\n",
197                        (const char *)par->targinfo);
198
199         return NF_DROP;
200 }
201
202 /* Performance critical - called for every packet */
203 static inline bool
204 do_match(struct ip6t_entry_match *m, const struct sk_buff *skb,
205          struct xt_match_param *par)
206 {
207         par->match     = m->u.kernel.match;
208         par->matchinfo = m->data;
209
210         /* Stop iteration if it doesn't match */
211         if (!m->u.kernel.match->match(skb, par))
212                 return true;
213         else
214                 return false;
215 }
216
217 static inline struct ip6t_entry *
218 get_entry(void *base, unsigned int offset)
219 {
220         return (struct ip6t_entry *)(base + offset);
221 }
222
223 /* All zeroes == unconditional rule. */
224 /* Mildly perf critical (only if packet tracing is on) */
225 static inline int
226 unconditional(const struct ip6t_ip6 *ipv6)
227 {
228         unsigned int i;
229
230         for (i = 0; i < sizeof(*ipv6); i++)
231                 if (((char *)ipv6)[i])
232                         break;
233
234         return (i == sizeof(*ipv6));
235 }
236
237 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
238     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
239 /* This cries for unification! */
240 static const char *const hooknames[] = {
241         [NF_INET_PRE_ROUTING]           = "PREROUTING",
242         [NF_INET_LOCAL_IN]              = "INPUT",
243         [NF_INET_FORWARD]               = "FORWARD",
244         [NF_INET_LOCAL_OUT]             = "OUTPUT",
245         [NF_INET_POST_ROUTING]          = "POSTROUTING",
246 };
247
248 enum nf_ip_trace_comments {
249         NF_IP6_TRACE_COMMENT_RULE,
250         NF_IP6_TRACE_COMMENT_RETURN,
251         NF_IP6_TRACE_COMMENT_POLICY,
252 };
253
254 static const char *const comments[] = {
255         [NF_IP6_TRACE_COMMENT_RULE]     = "rule",
256         [NF_IP6_TRACE_COMMENT_RETURN]   = "return",
257         [NF_IP6_TRACE_COMMENT_POLICY]   = "policy",
258 };
259
260 static struct nf_loginfo trace_loginfo = {
261         .type = NF_LOG_TYPE_LOG,
262         .u = {
263                 .log = {
264                         .level = 4,
265                         .logflags = NF_LOG_MASK,
266                 },
267         },
268 };
269
270 /* Mildly perf critical (only if packet tracing is on) */
271 static inline int
272 get_chainname_rulenum(struct ip6t_entry *s, struct ip6t_entry *e,
273                       char *hookname, char **chainname,
274                       char **comment, unsigned int *rulenum)
275 {
276         struct ip6t_standard_target *t = (void *)ip6t_get_target(s);
277
278         if (strcmp(t->target.u.kernel.target->name, IP6T_ERROR_TARGET) == 0) {
279                 /* Head of user chain: ERROR target with chainname */
280                 *chainname = t->target.data;
281                 (*rulenum) = 0;
282         } else if (s == e) {
283                 (*rulenum)++;
284
285                 if (s->target_offset == sizeof(struct ip6t_entry)
286                    && strcmp(t->target.u.kernel.target->name,
287                              IP6T_STANDARD_TARGET) == 0
288                    && t->verdict < 0
289                    && unconditional(&s->ipv6)) {
290                         /* Tail of chains: STANDARD target (return/policy) */
291                         *comment = *chainname == hookname
292                                 ? (char *)comments[NF_IP6_TRACE_COMMENT_POLICY]
293                                 : (char *)comments[NF_IP6_TRACE_COMMENT_RETURN];
294                 }
295                 return 1;
296         } else
297                 (*rulenum)++;
298
299         return 0;
300 }
301
302 static void trace_packet(struct sk_buff *skb,
303                          unsigned int hook,
304                          const struct net_device *in,
305                          const struct net_device *out,
306                          const char *tablename,
307                          struct xt_table_info *private,
308                          struct ip6t_entry *e)
309 {
310         void *table_base;
311         const struct ip6t_entry *root;
312         char *hookname, *chainname, *comment;
313         unsigned int rulenum = 0;
314
315         table_base = (void *)private->entries[smp_processor_id()];
316         root = get_entry(table_base, private->hook_entry[hook]);
317
318         hookname = chainname = (char *)hooknames[hook];
319         comment = (char *)comments[NF_IP6_TRACE_COMMENT_RULE];
320
321         IP6T_ENTRY_ITERATE(root,
322                            private->size - private->hook_entry[hook],
323                            get_chainname_rulenum,
324                            e, hookname, &chainname, &comment, &rulenum);
325
326         nf_log_packet(AF_INET6, hook, skb, in, out, &trace_loginfo,
327                       "TRACE: %s:%s:%s:%u ",
328                       tablename, chainname, comment, rulenum);
329 }
330 #endif
331
332 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
333 unsigned int
334 ip6t_do_table(struct sk_buff *skb,
335               unsigned int hook,
336               const struct net_device *in,
337               const struct net_device *out,
338               struct xt_table *table)
339 {
340         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
341         bool hotdrop = false;
342         /* Initializing verdict to NF_DROP keeps gcc happy. */
343         unsigned int verdict = NF_DROP;
344         const char *indev, *outdev;
345         void *table_base;
346         struct ip6t_entry *e, *back;
347         struct xt_table_info *private;
348         struct xt_match_param mtpar;
349         struct xt_target_param tgpar;
350
351         /* Initialization */
352         indev = in ? in->name : nulldevname;
353         outdev = out ? out->name : nulldevname;
354         /* We handle fragments by dealing with the first fragment as
355          * if it was a normal packet.  All other fragments are treated
356          * normally, except that they will NEVER match rules that ask
357          * things we don't know, ie. tcp syn flag or ports).  If the
358          * rule is also a fragment-specific rule, non-fragments won't
359          * match it. */
360         mtpar.hotdrop = &hotdrop;
361         mtpar.in      = tgpar.in  = in;
362         mtpar.out     = tgpar.out = out;
363         mtpar.family  = tgpar.family = NFPROTO_IPV6;
364         tgpar.hooknum = hook;
365
366         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
367
368         rcu_read_lock_bh();
369         private = rcu_dereference(table->private);
370         table_base = rcu_dereference(private->entries[smp_processor_id()]);
371
372         e = get_entry(table_base, private->hook_entry[hook]);
373
374         /* For return from builtin chain */
375         back = get_entry(table_base, private->underflow[hook]);
376
377         do {
378                 IP_NF_ASSERT(e);
379                 IP_NF_ASSERT(back);
380                 if (ip6_packet_match(skb, indev, outdev, &e->ipv6,
381                         &mtpar.thoff, &mtpar.fragoff, &hotdrop)) {
382                         struct ip6t_entry_target *t;
383
384                         if (IP6T_MATCH_ITERATE(e, do_match, skb, &mtpar) != 0)
385                                 goto no_match;
386
387                         ADD_COUNTER(e->counters,
388                                     ntohs(ipv6_hdr(skb)->payload_len) +
389                                     sizeof(struct ipv6hdr), 1);
390
391                         t = ip6t_get_target(e);
392                         IP_NF_ASSERT(t->u.kernel.target);
393
394 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
395     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
396                         /* The packet is traced: log it */
397                         if (unlikely(skb->nf_trace))
398                                 trace_packet(skb, hook, in, out,
399                                              table->name, private, e);
400 #endif
401                         /* Standard target? */
402                         if (!t->u.kernel.target->target) {
403                                 int v;
404
405                                 v = ((struct ip6t_standard_target *)t)->verdict;
406                                 if (v < 0) {
407                                         /* Pop from stack? */
408                                         if (v != IP6T_RETURN) {
409                                                 verdict = (unsigned)(-v) - 1;
410                                                 break;
411                                         }
412                                         e = back;
413                                         back = get_entry(table_base,
414                                                          back->comefrom);
415                                         continue;
416                                 }
417                                 if (table_base + v != (void *)e + e->next_offset
418                                     && !(e->ipv6.flags & IP6T_F_GOTO)) {
419                                         /* Save old back ptr in next entry */
420                                         struct ip6t_entry *next
421                                                 = (void *)e + e->next_offset;
422                                         next->comefrom
423                                                 = (void *)back - table_base;
424                                         /* set back pointer to next entry */
425                                         back = next;
426                                 }
427
428                                 e = get_entry(table_base, v);
429                         } else {
430                                 /* Targets which reenter must return
431                                    abs. verdicts */
432                                 tgpar.target   = t->u.kernel.target;
433                                 tgpar.targinfo = t->data;
434
435 #ifdef CONFIG_NETFILTER_DEBUG
436                                 ((struct ip6t_entry *)table_base)->comefrom
437                                         = 0xeeeeeeec;
438 #endif
439                                 verdict = t->u.kernel.target->target(skb,
440                                                                      &tgpar);
441
442 #ifdef CONFIG_NETFILTER_DEBUG
443                                 if (((struct ip6t_entry *)table_base)->comefrom
444                                     != 0xeeeeeeec
445                                     && verdict == IP6T_CONTINUE) {
446                                         printk("Target %s reentered!\n",
447                                                t->u.kernel.target->name);
448                                         verdict = NF_DROP;
449                                 }
450                                 ((struct ip6t_entry *)table_base)->comefrom
451                                         = 0x57acc001;
452 #endif
453                                 if (verdict == IP6T_CONTINUE)
454                                         e = (void *)e + e->next_offset;
455                                 else
456                                         /* Verdict */
457                                         break;
458                         }
459                 } else {
460
461                 no_match:
462                         e = (void *)e + e->next_offset;
463                 }
464         } while (!hotdrop);
465
466 #ifdef CONFIG_NETFILTER_DEBUG
467         ((struct ip6t_entry *)table_base)->comefrom = NETFILTER_LINK_POISON;
468 #endif
469         rcu_read_unlock_bh();
470
471 #ifdef DEBUG_ALLOW_ALL
472         return NF_ACCEPT;
473 #else
474         if (hotdrop)
475                 return NF_DROP;
476         else return verdict;
477 #endif
478 }
479
480 /* Figures out from what hook each rule can be called: returns 0 if
481    there are loops.  Puts hook bitmask in comefrom. */
482 static int
483 mark_source_chains(struct xt_table_info *newinfo,
484                    unsigned int valid_hooks, void *entry0)
485 {
486         unsigned int hook;
487
488         /* No recursion; use packet counter to save back ptrs (reset
489            to 0 as we leave), and comefrom to save source hook bitmask */
490         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
491                 unsigned int pos = newinfo->hook_entry[hook];
492                 struct ip6t_entry *e = (struct ip6t_entry *)(entry0 + pos);
493
494                 if (!(valid_hooks & (1 << hook)))
495                         continue;
496
497                 /* Set initial back pointer. */
498                 e->counters.pcnt = pos;
499
500                 for (;;) {
501                         struct ip6t_standard_target *t
502                                 = (void *)ip6t_get_target(e);
503                         int visited = e->comefrom & (1 << hook);
504
505                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
506                                 printk("iptables: loop hook %u pos %u %08X.\n",
507                                        hook, pos, e->comefrom);
508                                 return 0;
509                         }
510                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
511
512                         /* Unconditional return/END. */
513                         if ((e->target_offset == sizeof(struct ip6t_entry)
514                             && (strcmp(t->target.u.user.name,
515                                        IP6T_STANDARD_TARGET) == 0)
516                             && t->verdict < 0
517                             && unconditional(&e->ipv6)) || visited) {
518                                 unsigned int oldpos, size;
519
520                                 if ((strcmp(t->target.u.user.name,
521                                             IP6T_STANDARD_TARGET) == 0) &&
522                                     t->verdict < -NF_MAX_VERDICT - 1) {
523                                         duprintf("mark_source_chains: bad "
524                                                 "negative verdict (%i)\n",
525                                                                 t->verdict);
526                                         return 0;
527                                 }
528
529                                 /* Return: backtrack through the last
530                                    big jump. */
531                                 do {
532                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
533 #ifdef DEBUG_IP_FIREWALL_USER
534                                         if (e->comefrom
535                                             & (1 << NF_INET_NUMHOOKS)) {
536                                                 duprintf("Back unset "
537                                                          "on hook %u "
538                                                          "rule %u\n",
539                                                          hook, pos);
540                                         }
541 #endif
542                                         oldpos = pos;
543                                         pos = e->counters.pcnt;
544                                         e->counters.pcnt = 0;
545
546                                         /* We're at the start. */
547                                         if (pos == oldpos)
548                                                 goto next;
549
550                                         e = (struct ip6t_entry *)
551                                                 (entry0 + pos);
552                                 } while (oldpos == pos + e->next_offset);
553
554                                 /* Move along one */
555                                 size = e->next_offset;
556                                 e = (struct ip6t_entry *)
557                                         (entry0 + pos + size);
558                                 e->counters.pcnt = pos;
559                                 pos += size;
560                         } else {
561                                 int newpos = t->verdict;
562
563                                 if (strcmp(t->target.u.user.name,
564                                            IP6T_STANDARD_TARGET) == 0
565                                     && newpos >= 0) {
566                                         if (newpos > newinfo->size -
567                                                 sizeof(struct ip6t_entry)) {
568                                                 duprintf("mark_source_chains: "
569                                                         "bad verdict (%i)\n",
570                                                                 newpos);
571                                                 return 0;
572                                         }
573                                         /* This a jump; chase it. */
574                                         duprintf("Jump rule %u -> %u\n",
575                                                  pos, newpos);
576                                 } else {
577                                         /* ... this is a fallthru */
578                                         newpos = pos + e->next_offset;
579                                 }
580                                 e = (struct ip6t_entry *)
581                                         (entry0 + newpos);
582                                 e->counters.pcnt = pos;
583                                 pos = newpos;
584                         }
585                 }
586                 next:
587                 duprintf("Finished chain %u\n", hook);
588         }
589         return 1;
590 }
591
592 static int
593 cleanup_match(struct ip6t_entry_match *m, unsigned int *i)
594 {
595         struct xt_mtdtor_param par;
596
597         if (i && (*i)-- == 0)
598                 return 1;
599
600         par.match     = m->u.kernel.match;
601         par.matchinfo = m->data;
602         par.family    = NFPROTO_IPV6;
603         if (par.match->destroy != NULL)
604                 par.match->destroy(&par);
605         module_put(par.match->me);
606         return 0;
607 }
608
609 static int
610 check_entry(struct ip6t_entry *e, const char *name)
611 {
612         struct ip6t_entry_target *t;
613
614         if (!ip6_checkentry(&e->ipv6)) {
615                 duprintf("ip_tables: ip check failed %p %s.\n", e, name);
616                 return -EINVAL;
617         }
618
619         if (e->target_offset + sizeof(struct ip6t_entry_target) >
620             e->next_offset)
621                 return -EINVAL;
622
623         t = ip6t_get_target(e);
624         if (e->target_offset + t->u.target_size > e->next_offset)
625                 return -EINVAL;
626
627         return 0;
628 }
629
630 static int check_match(struct ip6t_entry_match *m, struct xt_mtchk_param *par,
631                        unsigned int *i)
632 {
633         const struct ip6t_ip6 *ipv6 = par->entryinfo;
634         int ret;
635
636         par->match     = m->u.kernel.match;
637         par->matchinfo = m->data;
638
639         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
640                              ipv6->proto, ipv6->invflags & IP6T_INV_PROTO);
641         if (ret < 0) {
642                 duprintf("ip_tables: check failed for `%s'.\n",
643                          par.match->name);
644                 return ret;
645         }
646         ++*i;
647         return 0;
648 }
649
650 static int
651 find_check_match(struct ip6t_entry_match *m, struct xt_mtchk_param *par,
652                  unsigned int *i)
653 {
654         struct xt_match *match;
655         int ret;
656
657         match = try_then_request_module(xt_find_match(AF_INET6, m->u.user.name,
658                                                       m->u.user.revision),
659                                         "ip6t_%s", m->u.user.name);
660         if (IS_ERR(match) || !match) {
661                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
662                 return match ? PTR_ERR(match) : -ENOENT;
663         }
664         m->u.kernel.match = match;
665
666         ret = check_match(m, par, i);
667         if (ret)
668                 goto err;
669
670         return 0;
671 err:
672         module_put(m->u.kernel.match->me);
673         return ret;
674 }
675
676 static int check_target(struct ip6t_entry *e, const char *name)
677 {
678         struct ip6t_entry_target *t = ip6t_get_target(e);
679         struct xt_tgchk_param par = {
680                 .table     = name,
681                 .entryinfo = e,
682                 .target    = t->u.kernel.target,
683                 .targinfo  = t->data,
684                 .hook_mask = e->comefrom,
685                 .family    = NFPROTO_IPV6,
686         };
687         int ret;
688
689         t = ip6t_get_target(e);
690         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
691               e->ipv6.proto, e->ipv6.invflags & IP6T_INV_PROTO);
692         if (ret < 0) {
693                 duprintf("ip_tables: check failed for `%s'.\n",
694                          t->u.kernel.target->name);
695                 return ret;
696         }
697         return 0;
698 }
699
700 static int
701 find_check_entry(struct ip6t_entry *e, const char *name, unsigned int size,
702                  unsigned int *i)
703 {
704         struct ip6t_entry_target *t;
705         struct xt_target *target;
706         int ret;
707         unsigned int j;
708         struct xt_mtchk_param mtpar;
709
710         ret = check_entry(e, name);
711         if (ret)
712                 return ret;
713
714         j = 0;
715         mtpar.table     = name;
716         mtpar.entryinfo = &e->ipv6;
717         mtpar.hook_mask = e->comefrom;
718         mtpar.family    = NFPROTO_IPV6;
719         ret = IP6T_MATCH_ITERATE(e, find_check_match, &mtpar, &j);
720         if (ret != 0)
721                 goto cleanup_matches;
722
723         t = ip6t_get_target(e);
724         target = try_then_request_module(xt_find_target(AF_INET6,
725                                                         t->u.user.name,
726                                                         t->u.user.revision),
727                                          "ip6t_%s", t->u.user.name);
728         if (IS_ERR(target) || !target) {
729                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
730                 ret = target ? PTR_ERR(target) : -ENOENT;
731                 goto cleanup_matches;
732         }
733         t->u.kernel.target = target;
734
735         ret = check_target(e, name);
736         if (ret)
737                 goto err;
738
739         (*i)++;
740         return 0;
741  err:
742         module_put(t->u.kernel.target->me);
743  cleanup_matches:
744         IP6T_MATCH_ITERATE(e, cleanup_match, &j);
745         return ret;
746 }
747
748 static int
749 check_entry_size_and_hooks(struct ip6t_entry *e,
750                            struct xt_table_info *newinfo,
751                            unsigned char *base,
752                            unsigned char *limit,
753                            const unsigned int *hook_entries,
754                            const unsigned int *underflows,
755                            unsigned int *i)
756 {
757         unsigned int h;
758
759         if ((unsigned long)e % __alignof__(struct ip6t_entry) != 0
760             || (unsigned char *)e + sizeof(struct ip6t_entry) >= limit) {
761                 duprintf("Bad offset %p\n", e);
762                 return -EINVAL;
763         }
764
765         if (e->next_offset
766             < sizeof(struct ip6t_entry) + sizeof(struct ip6t_entry_target)) {
767                 duprintf("checking: element %p size %u\n",
768                          e, e->next_offset);
769                 return -EINVAL;
770         }
771
772         /* Check hooks & underflows */
773         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
774                 if ((unsigned char *)e - base == hook_entries[h])
775                         newinfo->hook_entry[h] = hook_entries[h];
776                 if ((unsigned char *)e - base == underflows[h])
777                         newinfo->underflow[h] = underflows[h];
778         }
779
780         /* FIXME: underflows must be unconditional, standard verdicts
781            < 0 (not IP6T_RETURN). --RR */
782
783         /* Clear counters and comefrom */
784         e->counters = ((struct xt_counters) { 0, 0 });
785         e->comefrom = 0;
786
787         (*i)++;
788         return 0;
789 }
790
791 static int
792 cleanup_entry(struct ip6t_entry *e, unsigned int *i)
793 {
794         struct xt_tgdtor_param par;
795         struct ip6t_entry_target *t;
796
797         if (i && (*i)-- == 0)
798                 return 1;
799
800         /* Cleanup all matches */
801         IP6T_MATCH_ITERATE(e, cleanup_match, NULL);
802         t = ip6t_get_target(e);
803
804         par.target   = t->u.kernel.target;
805         par.targinfo = t->data;
806         par.family   = NFPROTO_IPV6;
807         if (par.target->destroy != NULL)
808                 par.target->destroy(&par);
809         module_put(par.target->me);
810         return 0;
811 }
812
813 /* Checks and translates the user-supplied table segment (held in
814    newinfo) */
815 static int
816 translate_table(const char *name,
817                 unsigned int valid_hooks,
818                 struct xt_table_info *newinfo,
819                 void *entry0,
820                 unsigned int size,
821                 unsigned int number,
822                 const unsigned int *hook_entries,
823                 const unsigned int *underflows)
824 {
825         unsigned int i;
826         int ret;
827
828         newinfo->size = size;
829         newinfo->number = number;
830
831         /* Init all hooks to impossible value. */
832         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
833                 newinfo->hook_entry[i] = 0xFFFFFFFF;
834                 newinfo->underflow[i] = 0xFFFFFFFF;
835         }
836
837         duprintf("translate_table: size %u\n", newinfo->size);
838         i = 0;
839         /* Walk through entries, checking offsets. */
840         ret = IP6T_ENTRY_ITERATE(entry0, newinfo->size,
841                                 check_entry_size_and_hooks,
842                                 newinfo,
843                                 entry0,
844                                 entry0 + size,
845                                 hook_entries, underflows, &i);
846         if (ret != 0)
847                 return ret;
848
849         if (i != number) {
850                 duprintf("translate_table: %u not %u entries\n",
851                          i, number);
852                 return -EINVAL;
853         }
854
855         /* Check hooks all assigned */
856         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
857                 /* Only hooks which are valid */
858                 if (!(valid_hooks & (1 << i)))
859                         continue;
860                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
861                         duprintf("Invalid hook entry %u %u\n",
862                                  i, hook_entries[i]);
863                         return -EINVAL;
864                 }
865                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
866                         duprintf("Invalid underflow %u %u\n",
867                                  i, underflows[i]);
868                         return -EINVAL;
869                 }
870         }
871
872         if (!mark_source_chains(newinfo, valid_hooks, entry0))
873                 return -ELOOP;
874
875         /* Finally, each sanity check must pass */
876         i = 0;
877         ret = IP6T_ENTRY_ITERATE(entry0, newinfo->size,
878                                 find_check_entry, name, size, &i);
879
880         if (ret != 0) {
881                 IP6T_ENTRY_ITERATE(entry0, newinfo->size,
882                                    cleanup_entry, &i);
883                 return ret;
884         }
885
886         /* And one copy for every other CPU */
887         for_each_possible_cpu(i) {
888                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
889                         memcpy(newinfo->entries[i], entry0, newinfo->size);
890         }
891
892         return ret;
893 }
894
895 /* Gets counters. */
896 static inline int
897 add_entry_to_counter(const struct ip6t_entry *e,
898                      struct xt_counters total[],
899                      unsigned int *i)
900 {
901         ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
902
903         (*i)++;
904         return 0;
905 }
906
907 static inline int
908 set_entry_to_counter(const struct ip6t_entry *e,
909                      struct ip6t_counters total[],
910                      unsigned int *i)
911 {
912         SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
913
914         (*i)++;
915         return 0;
916 }
917
918 static void
919 get_counters(const struct xt_table_info *t,
920              struct xt_counters counters[])
921 {
922         unsigned int cpu;
923         unsigned int i;
924         unsigned int curcpu;
925
926         /* Instead of clearing (by a previous call to memset())
927          * the counters and using adds, we set the counters
928          * with data used by 'current' CPU
929          * We dont care about preemption here.
930          */
931         curcpu = raw_smp_processor_id();
932
933         i = 0;
934         IP6T_ENTRY_ITERATE(t->entries[curcpu],
935                            t->size,
936                            set_entry_to_counter,
937                            counters,
938                            &i);
939
940         for_each_possible_cpu(cpu) {
941                 if (cpu == curcpu)
942                         continue;
943                 i = 0;
944                 IP6T_ENTRY_ITERATE(t->entries[cpu],
945                                   t->size,
946                                   add_entry_to_counter,
947                                   counters,
948                                   &i);
949         }
950 }
951
952 /* We're lazy, and add to the first CPU; overflow works its fey magic
953  * and everything is OK. */
954 static int
955 add_counter_to_entry(struct ip6t_entry *e,
956                      const struct xt_counters addme[],
957                      unsigned int *i)
958 {
959         ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
960
961         (*i)++;
962         return 0;
963 }
964
965 /* Take values from counters and add them back onto the current cpu */
966 static void put_counters(struct xt_table_info *t,
967                          const struct xt_counters counters[])
968 {
969         unsigned int i, cpu;
970
971         local_bh_disable();
972         cpu = smp_processor_id();
973         i = 0;
974         IP6T_ENTRY_ITERATE(t->entries[cpu],
975                            t->size,
976                            add_counter_to_entry,
977                            counters,
978                            &i);
979         local_bh_enable();
980 }
981
982 static inline int
983 zero_entry_counter(struct ip6t_entry *e, void *arg)
984 {
985         e->counters.bcnt = 0;
986         e->counters.pcnt = 0;
987         return 0;
988 }
989
990 static void
991 clone_counters(struct xt_table_info *newinfo, const struct xt_table_info *info)
992 {
993         unsigned int cpu;
994         const void *loc_cpu_entry = info->entries[raw_smp_processor_id()];
995
996         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
997         for_each_possible_cpu(cpu) {
998                 memcpy(newinfo->entries[cpu], loc_cpu_entry, info->size);
999                 IP6T_ENTRY_ITERATE(newinfo->entries[cpu], newinfo->size,
1000                                    zero_entry_counter, NULL);
1001         }
1002 }
1003
1004 static struct xt_counters *alloc_counters(struct xt_table *table)
1005 {
1006         unsigned int countersize;
1007         struct xt_counters *counters;
1008         struct xt_table_info *private = table->private;
1009         struct xt_table_info *info;
1010
1011         /* We need atomic snapshot of counters: rest doesn't change
1012            (other than comefrom, which userspace doesn't care
1013            about). */
1014         countersize = sizeof(struct xt_counters) * private->number;
1015         counters = vmalloc_node(countersize, numa_node_id());
1016
1017         if (counters == NULL)
1018                 goto nomem;
1019
1020         info = xt_alloc_table_info(private->size);
1021         if (!info)
1022                 goto free_counters;
1023
1024         clone_counters(info, private);
1025
1026         mutex_lock(&table->lock);
1027         xt_table_entry_swap_rcu(private, info);
1028         synchronize_net();      /* Wait until smoke has cleared */
1029
1030         get_counters(info, counters);
1031         put_counters(private, counters);
1032         mutex_unlock(&table->lock);
1033
1034         xt_free_table_info(info);
1035
1036         return counters;
1037
1038  free_counters:
1039         vfree(counters);
1040  nomem:
1041         return ERR_PTR(-ENOMEM);
1042 }
1043
1044 static int
1045 copy_entries_to_user(unsigned int total_size,
1046                      struct xt_table *table,
1047                      void __user *userptr)
1048 {
1049         unsigned int off, num;
1050         struct ip6t_entry *e;
1051         struct xt_counters *counters;
1052         const struct xt_table_info *private = table->private;
1053         int ret = 0;
1054         const void *loc_cpu_entry;
1055
1056         counters = alloc_counters(table);
1057         if (IS_ERR(counters))
1058                 return PTR_ERR(counters);
1059
1060         /* choose the copy that is on our node/cpu, ...
1061          * This choice is lazy (because current thread is
1062          * allowed to migrate to another cpu)
1063          */
1064         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1065         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
1066                 ret = -EFAULT;
1067                 goto free_counters;
1068         }
1069
1070         /* FIXME: use iterator macros --RR */
1071         /* ... then go back and fix counters and names */
1072         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
1073                 unsigned int i;
1074                 const struct ip6t_entry_match *m;
1075                 const struct ip6t_entry_target *t;
1076
1077                 e = (struct ip6t_entry *)(loc_cpu_entry + off);
1078                 if (copy_to_user(userptr + off
1079                                  + offsetof(struct ip6t_entry, counters),
1080                                  &counters[num],
1081                                  sizeof(counters[num])) != 0) {
1082                         ret = -EFAULT;
1083                         goto free_counters;
1084                 }
1085
1086                 for (i = sizeof(struct ip6t_entry);
1087                      i < e->target_offset;
1088                      i += m->u.match_size) {
1089                         m = (void *)e + i;
1090
1091                         if (copy_to_user(userptr + off + i
1092                                          + offsetof(struct ip6t_entry_match,
1093                                                     u.user.name),
1094                                          m->u.kernel.match->name,
1095                                          strlen(m->u.kernel.match->name)+1)
1096                             != 0) {
1097                                 ret = -EFAULT;
1098                                 goto free_counters;
1099                         }
1100                 }
1101
1102                 t = ip6t_get_target(e);
1103                 if (copy_to_user(userptr + off + e->target_offset
1104                                  + offsetof(struct ip6t_entry_target,
1105                                             u.user.name),
1106                                  t->u.kernel.target->name,
1107                                  strlen(t->u.kernel.target->name)+1) != 0) {
1108                         ret = -EFAULT;
1109                         goto free_counters;
1110                 }
1111         }
1112
1113  free_counters:
1114         vfree(counters);
1115         return ret;
1116 }
1117
1118 #ifdef CONFIG_COMPAT
1119 static void compat_standard_from_user(void *dst, void *src)
1120 {
1121         int v = *(compat_int_t *)src;
1122
1123         if (v > 0)
1124                 v += xt_compat_calc_jump(AF_INET6, v);
1125         memcpy(dst, &v, sizeof(v));
1126 }
1127
1128 static int compat_standard_to_user(void __user *dst, void *src)
1129 {
1130         compat_int_t cv = *(int *)src;
1131
1132         if (cv > 0)
1133                 cv -= xt_compat_calc_jump(AF_INET6, cv);
1134         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1135 }
1136
1137 static inline int
1138 compat_calc_match(struct ip6t_entry_match *m, int *size)
1139 {
1140         *size += xt_compat_match_offset(m->u.kernel.match);
1141         return 0;
1142 }
1143
1144 static int compat_calc_entry(struct ip6t_entry *e,
1145                              const struct xt_table_info *info,
1146                              void *base, struct xt_table_info *newinfo)
1147 {
1148         struct ip6t_entry_target *t;
1149         unsigned int entry_offset;
1150         int off, i, ret;
1151
1152         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1153         entry_offset = (void *)e - base;
1154         IP6T_MATCH_ITERATE(e, compat_calc_match, &off);
1155         t = ip6t_get_target(e);
1156         off += xt_compat_target_offset(t->u.kernel.target);
1157         newinfo->size -= off;
1158         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1159         if (ret)
1160                 return ret;
1161
1162         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1163                 if (info->hook_entry[i] &&
1164                     (e < (struct ip6t_entry *)(base + info->hook_entry[i])))
1165                         newinfo->hook_entry[i] -= off;
1166                 if (info->underflow[i] &&
1167                     (e < (struct ip6t_entry *)(base + info->underflow[i])))
1168                         newinfo->underflow[i] -= off;
1169         }
1170         return 0;
1171 }
1172
1173 static int compat_table_info(const struct xt_table_info *info,
1174                              struct xt_table_info *newinfo)
1175 {
1176         void *loc_cpu_entry;
1177
1178         if (!newinfo || !info)
1179                 return -EINVAL;
1180
1181         /* we dont care about newinfo->entries[] */
1182         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1183         newinfo->initial_entries = 0;
1184         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1185         return IP6T_ENTRY_ITERATE(loc_cpu_entry, info->size,
1186                                   compat_calc_entry, info, loc_cpu_entry,
1187                                   newinfo);
1188 }
1189 #endif
1190
1191 static int get_info(struct net *net, void __user *user, int *len, int compat)
1192 {
1193         char name[IP6T_TABLE_MAXNAMELEN];
1194         struct xt_table *t;
1195         int ret;
1196
1197         if (*len != sizeof(struct ip6t_getinfo)) {
1198                 duprintf("length %u != %zu\n", *len,
1199                          sizeof(struct ip6t_getinfo));
1200                 return -EINVAL;
1201         }
1202
1203         if (copy_from_user(name, user, sizeof(name)) != 0)
1204                 return -EFAULT;
1205
1206         name[IP6T_TABLE_MAXNAMELEN-1] = '\0';
1207 #ifdef CONFIG_COMPAT
1208         if (compat)
1209                 xt_compat_lock(AF_INET6);
1210 #endif
1211         t = try_then_request_module(xt_find_table_lock(net, AF_INET6, name),
1212                                     "ip6table_%s", name);
1213         if (t && !IS_ERR(t)) {
1214                 struct ip6t_getinfo info;
1215                 const struct xt_table_info *private = t->private;
1216
1217 #ifdef CONFIG_COMPAT
1218                 if (compat) {
1219                         struct xt_table_info tmp;
1220                         ret = compat_table_info(private, &tmp);
1221                         xt_compat_flush_offsets(AF_INET6);
1222                         private = &tmp;
1223                 }
1224 #endif
1225                 info.valid_hooks = t->valid_hooks;
1226                 memcpy(info.hook_entry, private->hook_entry,
1227                        sizeof(info.hook_entry));
1228                 memcpy(info.underflow, private->underflow,
1229                        sizeof(info.underflow));
1230                 info.num_entries = private->number;
1231                 info.size = private->size;
1232                 strcpy(info.name, name);
1233
1234                 if (copy_to_user(user, &info, *len) != 0)
1235                         ret = -EFAULT;
1236                 else
1237                         ret = 0;
1238
1239                 xt_table_unlock(t);
1240                 module_put(t->me);
1241         } else
1242                 ret = t ? PTR_ERR(t) : -ENOENT;
1243 #ifdef CONFIG_COMPAT
1244         if (compat)
1245                 xt_compat_unlock(AF_INET6);
1246 #endif
1247         return ret;
1248 }
1249
1250 static int
1251 get_entries(struct net *net, struct ip6t_get_entries __user *uptr, int *len)
1252 {
1253         int ret;
1254         struct ip6t_get_entries get;
1255         struct xt_table *t;
1256
1257         if (*len < sizeof(get)) {
1258                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1259                 return -EINVAL;
1260         }
1261         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1262                 return -EFAULT;
1263         if (*len != sizeof(struct ip6t_get_entries) + get.size) {
1264                 duprintf("get_entries: %u != %zu\n",
1265                          *len, sizeof(get) + get.size);
1266                 return -EINVAL;
1267         }
1268
1269         t = xt_find_table_lock(net, AF_INET6, get.name);
1270         if (t && !IS_ERR(t)) {
1271                 struct xt_table_info *private = t->private;
1272                 duprintf("t->private->number = %u\n", private->number);
1273                 if (get.size == private->size)
1274                         ret = copy_entries_to_user(private->size,
1275                                                    t, uptr->entrytable);
1276                 else {
1277                         duprintf("get_entries: I've got %u not %u!\n",
1278                                  private->size, get.size);
1279                         ret = -EAGAIN;
1280                 }
1281                 module_put(t->me);
1282                 xt_table_unlock(t);
1283         } else
1284                 ret = t ? PTR_ERR(t) : -ENOENT;
1285
1286         return ret;
1287 }
1288
1289 static int
1290 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1291              struct xt_table_info *newinfo, unsigned int num_counters,
1292              void __user *counters_ptr)
1293 {
1294         int ret;
1295         struct xt_table *t;
1296         struct xt_table_info *oldinfo;
1297         struct xt_counters *counters;
1298         const void *loc_cpu_old_entry;
1299
1300         ret = 0;
1301         counters = vmalloc_node(num_counters * sizeof(struct xt_counters),
1302                                 numa_node_id());
1303         if (!counters) {
1304                 ret = -ENOMEM;
1305                 goto out;
1306         }
1307
1308         t = try_then_request_module(xt_find_table_lock(net, AF_INET6, name),
1309                                     "ip6table_%s", name);
1310         if (!t || IS_ERR(t)) {
1311                 ret = t ? PTR_ERR(t) : -ENOENT;
1312                 goto free_newinfo_counters_untrans;
1313         }
1314
1315         /* You lied! */
1316         if (valid_hooks != t->valid_hooks) {
1317                 duprintf("Valid hook crap: %08X vs %08X\n",
1318                          valid_hooks, t->valid_hooks);
1319                 ret = -EINVAL;
1320                 goto put_module;
1321         }
1322
1323         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1324         if (!oldinfo)
1325                 goto put_module;
1326
1327         /* Update module usage count based on number of rules */
1328         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1329                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1330         if ((oldinfo->number > oldinfo->initial_entries) ||
1331             (newinfo->number <= oldinfo->initial_entries))
1332                 module_put(t->me);
1333         if ((oldinfo->number > oldinfo->initial_entries) &&
1334             (newinfo->number <= oldinfo->initial_entries))
1335                 module_put(t->me);
1336
1337         /* Get the old counters. */
1338         get_counters(oldinfo, counters);
1339         /* Decrease module usage counts and free resource */
1340         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1341         IP6T_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,
1342                            NULL);
1343         xt_free_table_info(oldinfo);
1344         if (copy_to_user(counters_ptr, counters,
1345                          sizeof(struct xt_counters) * num_counters) != 0)
1346                 ret = -EFAULT;
1347         vfree(counters);
1348         xt_table_unlock(t);
1349         return ret;
1350
1351  put_module:
1352         module_put(t->me);
1353         xt_table_unlock(t);
1354  free_newinfo_counters_untrans:
1355         vfree(counters);
1356  out:
1357         return ret;
1358 }
1359
1360 static int
1361 do_replace(struct net *net, void __user *user, unsigned int len)
1362 {
1363         int ret;
1364         struct ip6t_replace tmp;
1365         struct xt_table_info *newinfo;
1366         void *loc_cpu_entry;
1367
1368         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1369                 return -EFAULT;
1370
1371         /* overflow check */
1372         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1373                 return -ENOMEM;
1374
1375         newinfo = xt_alloc_table_info(tmp.size);
1376         if (!newinfo)
1377                 return -ENOMEM;
1378
1379         /* choose the copy that is on our node/cpu */
1380         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1381         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1382                            tmp.size) != 0) {
1383                 ret = -EFAULT;
1384                 goto free_newinfo;
1385         }
1386
1387         ret = translate_table(tmp.name, tmp.valid_hooks,
1388                               newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
1389                               tmp.hook_entry, tmp.underflow);
1390         if (ret != 0)
1391                 goto free_newinfo;
1392
1393         duprintf("ip_tables: Translated table\n");
1394
1395         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1396                            tmp.num_counters, tmp.counters);
1397         if (ret)
1398                 goto free_newinfo_untrans;
1399         return 0;
1400
1401  free_newinfo_untrans:
1402         IP6T_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
1403  free_newinfo:
1404         xt_free_table_info(newinfo);
1405         return ret;
1406 }
1407
1408 static int
1409 do_add_counters(struct net *net, void __user *user, unsigned int len,
1410                 int compat)
1411 {
1412         unsigned int i;
1413         struct xt_counters_info tmp;
1414         struct xt_counters *paddc;
1415         unsigned int num_counters;
1416         char *name;
1417         int size;
1418         void *ptmp;
1419         struct xt_table *t;
1420         const struct xt_table_info *private;
1421         int ret = 0;
1422         const void *loc_cpu_entry;
1423 #ifdef CONFIG_COMPAT
1424         struct compat_xt_counters_info compat_tmp;
1425
1426         if (compat) {
1427                 ptmp = &compat_tmp;
1428                 size = sizeof(struct compat_xt_counters_info);
1429         } else
1430 #endif
1431         {
1432                 ptmp = &tmp;
1433                 size = sizeof(struct xt_counters_info);
1434         }
1435
1436         if (copy_from_user(ptmp, user, size) != 0)
1437                 return -EFAULT;
1438
1439 #ifdef CONFIG_COMPAT
1440         if (compat) {
1441                 num_counters = compat_tmp.num_counters;
1442                 name = compat_tmp.name;
1443         } else
1444 #endif
1445         {
1446                 num_counters = tmp.num_counters;
1447                 name = tmp.name;
1448         }
1449
1450         if (len != size + num_counters * sizeof(struct xt_counters))
1451                 return -EINVAL;
1452
1453         paddc = vmalloc_node(len - size, numa_node_id());
1454         if (!paddc)
1455                 return -ENOMEM;
1456
1457         if (copy_from_user(paddc, user + size, len - size) != 0) {
1458                 ret = -EFAULT;
1459                 goto free;
1460         }
1461
1462         t = xt_find_table_lock(net, AF_INET6, name);
1463         if (!t || IS_ERR(t)) {
1464                 ret = t ? PTR_ERR(t) : -ENOENT;
1465                 goto free;
1466         }
1467
1468         mutex_lock(&t->lock);
1469         private = t->private;
1470         if (private->number != num_counters) {
1471                 ret = -EINVAL;
1472                 goto unlock_up_free;
1473         }
1474
1475         preempt_disable();
1476         i = 0;
1477         /* Choose the copy that is on our node */
1478         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1479         IP6T_ENTRY_ITERATE(loc_cpu_entry,
1480                           private->size,
1481                           add_counter_to_entry,
1482                           paddc,
1483                           &i);
1484         preempt_enable();
1485  unlock_up_free:
1486         mutex_unlock(&t->lock);
1487         xt_table_unlock(t);
1488         module_put(t->me);
1489  free:
1490         vfree(paddc);
1491
1492         return ret;
1493 }
1494
1495 #ifdef CONFIG_COMPAT
1496 struct compat_ip6t_replace {
1497         char                    name[IP6T_TABLE_MAXNAMELEN];
1498         u32                     valid_hooks;
1499         u32                     num_entries;
1500         u32                     size;
1501         u32                     hook_entry[NF_INET_NUMHOOKS];
1502         u32                     underflow[NF_INET_NUMHOOKS];
1503         u32                     num_counters;
1504         compat_uptr_t           counters;       /* struct ip6t_counters * */
1505         struct compat_ip6t_entry entries[0];
1506 };
1507
1508 static int
1509 compat_copy_entry_to_user(struct ip6t_entry *e, void __user **dstptr,
1510                           unsigned int *size, struct xt_counters *counters,
1511                           unsigned int *i)
1512 {
1513         struct ip6t_entry_target *t;
1514         struct compat_ip6t_entry __user *ce;
1515         u_int16_t target_offset, next_offset;
1516         compat_uint_t origsize;
1517         int ret;
1518
1519         ret = -EFAULT;
1520         origsize = *size;
1521         ce = (struct compat_ip6t_entry __user *)*dstptr;
1522         if (copy_to_user(ce, e, sizeof(struct ip6t_entry)))
1523                 goto out;
1524
1525         if (copy_to_user(&ce->counters, &counters[*i], sizeof(counters[*i])))
1526                 goto out;
1527
1528         *dstptr += sizeof(struct compat_ip6t_entry);
1529         *size -= sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1530
1531         ret = IP6T_MATCH_ITERATE(e, xt_compat_match_to_user, dstptr, size);
1532         target_offset = e->target_offset - (origsize - *size);
1533         if (ret)
1534                 goto out;
1535         t = ip6t_get_target(e);
1536         ret = xt_compat_target_to_user(t, dstptr, size);
1537         if (ret)
1538                 goto out;
1539         ret = -EFAULT;
1540         next_offset = e->next_offset - (origsize - *size);
1541         if (put_user(target_offset, &ce->target_offset))
1542                 goto out;
1543         if (put_user(next_offset, &ce->next_offset))
1544                 goto out;
1545
1546         (*i)++;
1547         return 0;
1548 out:
1549         return ret;
1550 }
1551
1552 static int
1553 compat_find_calc_match(struct ip6t_entry_match *m,
1554                        const char *name,
1555                        const struct ip6t_ip6 *ipv6,
1556                        unsigned int hookmask,
1557                        int *size, unsigned int *i)
1558 {
1559         struct xt_match *match;
1560
1561         match = try_then_request_module(xt_find_match(AF_INET6, m->u.user.name,
1562                                                       m->u.user.revision),
1563                                         "ip6t_%s", m->u.user.name);
1564         if (IS_ERR(match) || !match) {
1565                 duprintf("compat_check_calc_match: `%s' not found\n",
1566                          m->u.user.name);
1567                 return match ? PTR_ERR(match) : -ENOENT;
1568         }
1569         m->u.kernel.match = match;
1570         *size += xt_compat_match_offset(match);
1571
1572         (*i)++;
1573         return 0;
1574 }
1575
1576 static int
1577 compat_release_match(struct ip6t_entry_match *m, unsigned int *i)
1578 {
1579         if (i && (*i)-- == 0)
1580                 return 1;
1581
1582         module_put(m->u.kernel.match->me);
1583         return 0;
1584 }
1585
1586 static int
1587 compat_release_entry(struct compat_ip6t_entry *e, unsigned int *i)
1588 {
1589         struct ip6t_entry_target *t;
1590
1591         if (i && (*i)-- == 0)
1592                 return 1;
1593
1594         /* Cleanup all matches */
1595         COMPAT_IP6T_MATCH_ITERATE(e, compat_release_match, NULL);
1596         t = compat_ip6t_get_target(e);
1597         module_put(t->u.kernel.target->me);
1598         return 0;
1599 }
1600
1601 static int
1602 check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
1603                                   struct xt_table_info *newinfo,
1604                                   unsigned int *size,
1605                                   unsigned char *base,
1606                                   unsigned char *limit,
1607                                   unsigned int *hook_entries,
1608                                   unsigned int *underflows,
1609                                   unsigned int *i,
1610                                   const char *name)
1611 {
1612         struct ip6t_entry_target *t;
1613         struct xt_target *target;
1614         unsigned int entry_offset;
1615         unsigned int j;
1616         int ret, off, h;
1617
1618         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1619         if ((unsigned long)e % __alignof__(struct compat_ip6t_entry) != 0
1620             || (unsigned char *)e + sizeof(struct compat_ip6t_entry) >= limit) {
1621                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1622                 return -EINVAL;
1623         }
1624
1625         if (e->next_offset < sizeof(struct compat_ip6t_entry) +
1626                              sizeof(struct compat_xt_entry_target)) {
1627                 duprintf("checking: element %p size %u\n",
1628                          e, e->next_offset);
1629                 return -EINVAL;
1630         }
1631
1632         /* For purposes of check_entry casting the compat entry is fine */
1633         ret = check_entry((struct ip6t_entry *)e, name);
1634         if (ret)
1635                 return ret;
1636
1637         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1638         entry_offset = (void *)e - (void *)base;
1639         j = 0;
1640         ret = COMPAT_IP6T_MATCH_ITERATE(e, compat_find_calc_match, name,
1641                                         &e->ipv6, e->comefrom, &off, &j);
1642         if (ret != 0)
1643                 goto release_matches;
1644
1645         t = compat_ip6t_get_target(e);
1646         target = try_then_request_module(xt_find_target(AF_INET6,
1647                                                         t->u.user.name,
1648                                                         t->u.user.revision),
1649                                          "ip6t_%s", t->u.user.name);
1650         if (IS_ERR(target) || !target) {
1651                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1652                          t->u.user.name);
1653                 ret = target ? PTR_ERR(target) : -ENOENT;
1654                 goto release_matches;
1655         }
1656         t->u.kernel.target = target;
1657
1658         off += xt_compat_target_offset(target);
1659         *size += off;
1660         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1661         if (ret)
1662                 goto out;
1663
1664         /* Check hooks & underflows */
1665         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1666                 if ((unsigned char *)e - base == hook_entries[h])
1667                         newinfo->hook_entry[h] = hook_entries[h];
1668                 if ((unsigned char *)e - base == underflows[h])
1669                         newinfo->underflow[h] = underflows[h];
1670         }
1671
1672         /* Clear counters and comefrom */
1673         memset(&e->counters, 0, sizeof(e->counters));
1674         e->comefrom = 0;
1675
1676         (*i)++;
1677         return 0;
1678
1679 out:
1680         module_put(t->u.kernel.target->me);
1681 release_matches:
1682         IP6T_MATCH_ITERATE(e, compat_release_match, &j);
1683         return ret;
1684 }
1685
1686 static int
1687 compat_copy_entry_from_user(struct compat_ip6t_entry *e, void **dstptr,
1688                             unsigned int *size, const char *name,
1689                             struct xt_table_info *newinfo, unsigned char *base)
1690 {
1691         struct ip6t_entry_target *t;
1692         struct xt_target *target;
1693         struct ip6t_entry *de;
1694         unsigned int origsize;
1695         int ret, h;
1696
1697         ret = 0;
1698         origsize = *size;
1699         de = (struct ip6t_entry *)*dstptr;
1700         memcpy(de, e, sizeof(struct ip6t_entry));
1701         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1702
1703         *dstptr += sizeof(struct ip6t_entry);
1704         *size += sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1705
1706         ret = COMPAT_IP6T_MATCH_ITERATE(e, xt_compat_match_from_user,
1707                                         dstptr, size);
1708         if (ret)
1709                 return ret;
1710         de->target_offset = e->target_offset - (origsize - *size);
1711         t = compat_ip6t_get_target(e);
1712         target = t->u.kernel.target;
1713         xt_compat_target_from_user(t, dstptr, size);
1714
1715         de->next_offset = e->next_offset - (origsize - *size);
1716         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1717                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1718                         newinfo->hook_entry[h] -= origsize - *size;
1719                 if ((unsigned char *)de - base < newinfo->underflow[h])
1720                         newinfo->underflow[h] -= origsize - *size;
1721         }
1722         return ret;
1723 }
1724
1725 static int compat_check_entry(struct ip6t_entry *e, const char *name,
1726                                      unsigned int *i)
1727 {
1728         unsigned int j;
1729         int ret;
1730         struct xt_mtchk_param mtpar;
1731
1732         j = 0;
1733         mtpar.table     = name;
1734         mtpar.entryinfo = &e->ipv6;
1735         mtpar.hook_mask = e->comefrom;
1736         mtpar.family    = NFPROTO_IPV6;
1737         ret = IP6T_MATCH_ITERATE(e, check_match, &mtpar, &j);
1738         if (ret)
1739                 goto cleanup_matches;
1740
1741         ret = check_target(e, name);
1742         if (ret)
1743                 goto cleanup_matches;
1744
1745         (*i)++;
1746         return 0;
1747
1748  cleanup_matches:
1749         IP6T_MATCH_ITERATE(e, cleanup_match, &j);
1750         return ret;
1751 }
1752
1753 static int
1754 translate_compat_table(const char *name,
1755                        unsigned int valid_hooks,
1756                        struct xt_table_info **pinfo,
1757                        void **pentry0,
1758                        unsigned int total_size,
1759                        unsigned int number,
1760                        unsigned int *hook_entries,
1761                        unsigned int *underflows)
1762 {
1763         unsigned int i, j;
1764         struct xt_table_info *newinfo, *info;
1765         void *pos, *entry0, *entry1;
1766         unsigned int size;
1767         int ret;
1768
1769         info = *pinfo;
1770         entry0 = *pentry0;
1771         size = total_size;
1772         info->number = number;
1773
1774         /* Init all hooks to impossible value. */
1775         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1776                 info->hook_entry[i] = 0xFFFFFFFF;
1777                 info->underflow[i] = 0xFFFFFFFF;
1778         }
1779
1780         duprintf("translate_compat_table: size %u\n", info->size);
1781         j = 0;
1782         xt_compat_lock(AF_INET6);
1783         /* Walk through entries, checking offsets. */
1784         ret = COMPAT_IP6T_ENTRY_ITERATE(entry0, total_size,
1785                                         check_compat_entry_size_and_hooks,
1786                                         info, &size, entry0,
1787                                         entry0 + total_size,
1788                                         hook_entries, underflows, &j, name);
1789         if (ret != 0)
1790                 goto out_unlock;
1791
1792         ret = -EINVAL;
1793         if (j != number) {
1794                 duprintf("translate_compat_table: %u not %u entries\n",
1795                          j, number);
1796                 goto out_unlock;
1797         }
1798
1799         /* Check hooks all assigned */
1800         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1801                 /* Only hooks which are valid */
1802                 if (!(valid_hooks & (1 << i)))
1803                         continue;
1804                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1805                         duprintf("Invalid hook entry %u %u\n",
1806                                  i, hook_entries[i]);
1807                         goto out_unlock;
1808                 }
1809                 if (info->underflow[i] == 0xFFFFFFFF) {
1810                         duprintf("Invalid underflow %u %u\n",
1811                                  i, underflows[i]);
1812                         goto out_unlock;
1813                 }
1814         }
1815
1816         ret = -ENOMEM;
1817         newinfo = xt_alloc_table_info(size);
1818         if (!newinfo)
1819                 goto out_unlock;
1820
1821         newinfo->number = number;
1822         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1823                 newinfo->hook_entry[i] = info->hook_entry[i];
1824                 newinfo->underflow[i] = info->underflow[i];
1825         }
1826         entry1 = newinfo->entries[raw_smp_processor_id()];
1827         pos = entry1;
1828         size = total_size;
1829         ret = COMPAT_IP6T_ENTRY_ITERATE(entry0, total_size,
1830                                         compat_copy_entry_from_user,
1831                                         &pos, &size, name, newinfo, entry1);
1832         xt_compat_flush_offsets(AF_INET6);
1833         xt_compat_unlock(AF_INET6);
1834         if (ret)
1835                 goto free_newinfo;
1836
1837         ret = -ELOOP;
1838         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1839                 goto free_newinfo;
1840
1841         i = 0;
1842         ret = IP6T_ENTRY_ITERATE(entry1, newinfo->size, compat_check_entry,
1843                                  name, &i);
1844         if (ret) {
1845                 j -= i;
1846                 COMPAT_IP6T_ENTRY_ITERATE_CONTINUE(entry0, newinfo->size, i,
1847                                                    compat_release_entry, &j);
1848                 IP6T_ENTRY_ITERATE(entry1, newinfo->size, cleanup_entry, &i);
1849                 xt_free_table_info(newinfo);
1850                 return ret;
1851         }
1852
1853         /* And one copy for every other CPU */
1854         for_each_possible_cpu(i)
1855                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1856                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1857
1858         *pinfo = newinfo;
1859         *pentry0 = entry1;
1860         xt_free_table_info(info);
1861         return 0;
1862
1863 free_newinfo:
1864         xt_free_table_info(newinfo);
1865 out:
1866         COMPAT_IP6T_ENTRY_ITERATE(entry0, total_size, compat_release_entry, &j);
1867         return ret;
1868 out_unlock:
1869         xt_compat_flush_offsets(AF_INET6);
1870         xt_compat_unlock(AF_INET6);
1871         goto out;
1872 }
1873
1874 static int
1875 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1876 {
1877         int ret;
1878         struct compat_ip6t_replace tmp;
1879         struct xt_table_info *newinfo;
1880         void *loc_cpu_entry;
1881
1882         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1883                 return -EFAULT;
1884
1885         /* overflow check */
1886         if (tmp.size >= INT_MAX / num_possible_cpus())
1887                 return -ENOMEM;
1888         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1889                 return -ENOMEM;
1890
1891         newinfo = xt_alloc_table_info(tmp.size);
1892         if (!newinfo)
1893                 return -ENOMEM;
1894
1895         /* choose the copy that is on our node/cpu */
1896         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1897         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1898                            tmp.size) != 0) {
1899                 ret = -EFAULT;
1900                 goto free_newinfo;
1901         }
1902
1903         ret = translate_compat_table(tmp.name, tmp.valid_hooks,
1904                                      &newinfo, &loc_cpu_entry, tmp.size,
1905                                      tmp.num_entries, tmp.hook_entry,
1906                                      tmp.underflow);
1907         if (ret != 0)
1908                 goto free_newinfo;
1909
1910         duprintf("compat_do_replace: Translated table\n");
1911
1912         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1913                            tmp.num_counters, compat_ptr(tmp.counters));
1914         if (ret)
1915                 goto free_newinfo_untrans;
1916         return 0;
1917
1918  free_newinfo_untrans:
1919         IP6T_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
1920  free_newinfo:
1921         xt_free_table_info(newinfo);
1922         return ret;
1923 }
1924
1925 static int
1926 compat_do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user,
1927                        unsigned int len)
1928 {
1929         int ret;
1930
1931         if (!capable(CAP_NET_ADMIN))
1932                 return -EPERM;
1933
1934         switch (cmd) {
1935         case IP6T_SO_SET_REPLACE:
1936                 ret = compat_do_replace(sock_net(sk), user, len);
1937                 break;
1938
1939         case IP6T_SO_SET_ADD_COUNTERS:
1940                 ret = do_add_counters(sock_net(sk), user, len, 1);
1941                 break;
1942
1943         default:
1944                 duprintf("do_ip6t_set_ctl:  unknown request %i\n", cmd);
1945                 ret = -EINVAL;
1946         }
1947
1948         return ret;
1949 }
1950
1951 struct compat_ip6t_get_entries {
1952         char name[IP6T_TABLE_MAXNAMELEN];
1953         compat_uint_t size;
1954         struct compat_ip6t_entry entrytable[0];
1955 };
1956
1957 static int
1958 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1959                             void __user *userptr)
1960 {
1961         struct xt_counters *counters;
1962         const struct xt_table_info *private = table->private;
1963         void __user *pos;
1964         unsigned int size;
1965         int ret = 0;
1966         const void *loc_cpu_entry;
1967         unsigned int i = 0;
1968
1969         counters = alloc_counters(table);
1970         if (IS_ERR(counters))
1971                 return PTR_ERR(counters);
1972
1973         /* choose the copy that is on our node/cpu, ...
1974          * This choice is lazy (because current thread is
1975          * allowed to migrate to another cpu)
1976          */
1977         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1978         pos = userptr;
1979         size = total_size;
1980         ret = IP6T_ENTRY_ITERATE(loc_cpu_entry, total_size,
1981                                  compat_copy_entry_to_user,
1982                                  &pos, &size, counters, &i);
1983
1984         vfree(counters);
1985         return ret;
1986 }
1987
1988 static int
1989 compat_get_entries(struct net *net, struct compat_ip6t_get_entries __user *uptr,
1990                    int *len)
1991 {
1992         int ret;
1993         struct compat_ip6t_get_entries get;
1994         struct xt_table *t;
1995
1996         if (*len < sizeof(get)) {
1997                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1998                 return -EINVAL;
1999         }
2000
2001         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
2002                 return -EFAULT;
2003
2004         if (*len != sizeof(struct compat_ip6t_get_entries) + get.size) {
2005                 duprintf("compat_get_entries: %u != %zu\n",
2006                          *len, sizeof(get) + get.size);
2007                 return -EINVAL;
2008         }
2009
2010         xt_compat_lock(AF_INET6);
2011         t = xt_find_table_lock(net, AF_INET6, get.name);
2012         if (t && !IS_ERR(t)) {
2013                 const struct xt_table_info *private = t->private;
2014                 struct xt_table_info info;
2015                 duprintf("t->private->number = %u\n", private->number);
2016                 ret = compat_table_info(private, &info);
2017                 if (!ret && get.size == info.size) {
2018                         ret = compat_copy_entries_to_user(private->size,
2019                                                           t, uptr->entrytable);
2020                 } else if (!ret) {
2021                         duprintf("compat_get_entries: I've got %u not %u!\n",
2022                                  private->size, get.size);
2023                         ret = -EAGAIN;
2024                 }
2025                 xt_compat_flush_offsets(AF_INET6);
2026                 module_put(t->me);
2027                 xt_table_unlock(t);
2028         } else
2029                 ret = t ? PTR_ERR(t) : -ENOENT;
2030
2031         xt_compat_unlock(AF_INET6);
2032         return ret;
2033 }
2034
2035 static int do_ip6t_get_ctl(struct sock *, int, void __user *, int *);
2036
2037 static int
2038 compat_do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2039 {
2040         int ret;
2041
2042         if (!capable(CAP_NET_ADMIN))
2043                 return -EPERM;
2044
2045         switch (cmd) {
2046         case IP6T_SO_GET_INFO:
2047                 ret = get_info(sock_net(sk), user, len, 1);
2048                 break;
2049         case IP6T_SO_GET_ENTRIES:
2050                 ret = compat_get_entries(sock_net(sk), user, len);
2051                 break;
2052         default:
2053                 ret = do_ip6t_get_ctl(sk, cmd, user, len);
2054         }
2055         return ret;
2056 }
2057 #endif
2058
2059 static int
2060 do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2061 {
2062         int ret;
2063
2064         if (!capable(CAP_NET_ADMIN))
2065                 return -EPERM;
2066
2067         switch (cmd) {
2068         case IP6T_SO_SET_REPLACE:
2069                 ret = do_replace(sock_net(sk), user, len);
2070                 break;
2071
2072         case IP6T_SO_SET_ADD_COUNTERS:
2073                 ret = do_add_counters(sock_net(sk), user, len, 0);
2074                 break;
2075
2076         default:
2077                 duprintf("do_ip6t_set_ctl:  unknown request %i\n", cmd);
2078                 ret = -EINVAL;
2079         }
2080
2081         return ret;
2082 }
2083
2084 static int
2085 do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2086 {
2087         int ret;
2088
2089         if (!capable(CAP_NET_ADMIN))
2090                 return -EPERM;
2091
2092         switch (cmd) {
2093         case IP6T_SO_GET_INFO:
2094                 ret = get_info(sock_net(sk), user, len, 0);
2095                 break;
2096
2097         case IP6T_SO_GET_ENTRIES:
2098                 ret = get_entries(sock_net(sk), user, len);
2099                 break;
2100
2101         case IP6T_SO_GET_REVISION_MATCH:
2102         case IP6T_SO_GET_REVISION_TARGET: {
2103                 struct ip6t_get_revision rev;
2104                 int target;
2105
2106                 if (*len != sizeof(rev)) {
2107                         ret = -EINVAL;
2108                         break;
2109                 }
2110                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2111                         ret = -EFAULT;
2112                         break;
2113                 }
2114
2115                 if (cmd == IP6T_SO_GET_REVISION_TARGET)
2116                         target = 1;
2117                 else
2118                         target = 0;
2119
2120                 try_then_request_module(xt_find_revision(AF_INET6, rev.name,
2121                                                          rev.revision,
2122                                                          target, &ret),
2123                                         "ip6t_%s", rev.name);
2124                 break;
2125         }
2126
2127         default:
2128                 duprintf("do_ip6t_get_ctl: unknown request %i\n", cmd);
2129                 ret = -EINVAL;
2130         }
2131
2132         return ret;
2133 }
2134
2135 struct xt_table *ip6t_register_table(struct net *net, struct xt_table *table,
2136                                      const struct ip6t_replace *repl)
2137 {
2138         int ret;
2139         struct xt_table_info *newinfo;
2140         struct xt_table_info bootstrap
2141                 = { 0, 0, 0, { 0 }, { 0 }, { } };
2142         void *loc_cpu_entry;
2143         struct xt_table *new_table;
2144
2145         newinfo = xt_alloc_table_info(repl->size);
2146         if (!newinfo) {
2147                 ret = -ENOMEM;
2148                 goto out;
2149         }
2150
2151         /* choose the copy on our node/cpu, but dont care about preemption */
2152         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2153         memcpy(loc_cpu_entry, repl->entries, repl->size);
2154
2155         ret = translate_table(table->name, table->valid_hooks,
2156                               newinfo, loc_cpu_entry, repl->size,
2157                               repl->num_entries,
2158                               repl->hook_entry,
2159                               repl->underflow);
2160         if (ret != 0)
2161                 goto out_free;
2162
2163         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2164         if (IS_ERR(new_table)) {
2165                 ret = PTR_ERR(new_table);
2166                 goto out_free;
2167         }
2168         return new_table;
2169
2170 out_free:
2171         xt_free_table_info(newinfo);
2172 out:
2173         return ERR_PTR(ret);
2174 }
2175
2176 void ip6t_unregister_table(struct xt_table *table)
2177 {
2178         struct xt_table_info *private;
2179         void *loc_cpu_entry;
2180         struct module *table_owner = table->me;
2181
2182         private = xt_unregister_table(table);
2183
2184         /* Decrease module usage counts and free resources */
2185         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2186         IP6T_ENTRY_ITERATE(loc_cpu_entry, private->size, cleanup_entry, NULL);
2187         if (private->number > private->initial_entries)
2188                 module_put(table_owner);
2189         xt_free_table_info(private);
2190 }
2191
2192 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2193 static inline bool
2194 icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2195                      u_int8_t type, u_int8_t code,
2196                      bool invert)
2197 {
2198         return (type == test_type && code >= min_code && code <= max_code)
2199                 ^ invert;
2200 }
2201
2202 static bool
2203 icmp6_match(const struct sk_buff *skb, const struct xt_match_param *par)
2204 {
2205         const struct icmp6hdr *ic;
2206         struct icmp6hdr _icmph;
2207         const struct ip6t_icmp *icmpinfo = par->matchinfo;
2208
2209         /* Must not be a fragment. */
2210         if (par->fragoff != 0)
2211                 return false;
2212
2213         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2214         if (ic == NULL) {
2215                 /* We've been asked to examine this packet, and we
2216                  * can't.  Hence, no choice but to drop.
2217                  */
2218                 duprintf("Dropping evil ICMP tinygram.\n");
2219                 *par->hotdrop = true;
2220                 return false;
2221         }
2222
2223         return icmp6_type_code_match(icmpinfo->type,
2224                                      icmpinfo->code[0],
2225                                      icmpinfo->code[1],
2226                                      ic->icmp6_type, ic->icmp6_code,
2227                                      !!(icmpinfo->invflags&IP6T_ICMP_INV));
2228 }
2229
2230 /* Called when user tries to insert an entry of this type. */
2231 static bool icmp6_checkentry(const struct xt_mtchk_param *par)
2232 {
2233         const struct ip6t_icmp *icmpinfo = par->matchinfo;
2234
2235         /* Must specify no unknown invflags */
2236         return !(icmpinfo->invflags & ~IP6T_ICMP_INV);
2237 }
2238
2239 /* The built-in targets: standard (NULL) and error. */
2240 static struct xt_target ip6t_standard_target __read_mostly = {
2241         .name           = IP6T_STANDARD_TARGET,
2242         .targetsize     = sizeof(int),
2243         .family         = AF_INET6,
2244 #ifdef CONFIG_COMPAT
2245         .compatsize     = sizeof(compat_int_t),
2246         .compat_from_user = compat_standard_from_user,
2247         .compat_to_user = compat_standard_to_user,
2248 #endif
2249 };
2250
2251 static struct xt_target ip6t_error_target __read_mostly = {
2252         .name           = IP6T_ERROR_TARGET,
2253         .target         = ip6t_error,
2254         .targetsize     = IP6T_FUNCTION_MAXNAMELEN,
2255         .family         = AF_INET6,
2256 };
2257
2258 static struct nf_sockopt_ops ip6t_sockopts = {
2259         .pf             = PF_INET6,
2260         .set_optmin     = IP6T_BASE_CTL,
2261         .set_optmax     = IP6T_SO_SET_MAX+1,
2262         .set            = do_ip6t_set_ctl,
2263 #ifdef CONFIG_COMPAT
2264         .compat_set     = compat_do_ip6t_set_ctl,
2265 #endif
2266         .get_optmin     = IP6T_BASE_CTL,
2267         .get_optmax     = IP6T_SO_GET_MAX+1,
2268         .get            = do_ip6t_get_ctl,
2269 #ifdef CONFIG_COMPAT
2270         .compat_get     = compat_do_ip6t_get_ctl,
2271 #endif
2272         .owner          = THIS_MODULE,
2273 };
2274
2275 static struct xt_match icmp6_matchstruct __read_mostly = {
2276         .name           = "icmp6",
2277         .match          = icmp6_match,
2278         .matchsize      = sizeof(struct ip6t_icmp),
2279         .checkentry     = icmp6_checkentry,
2280         .proto          = IPPROTO_ICMPV6,
2281         .family         = AF_INET6,
2282 };
2283
2284 static int __net_init ip6_tables_net_init(struct net *net)
2285 {
2286         return xt_proto_init(net, AF_INET6);
2287 }
2288
2289 static void __net_exit ip6_tables_net_exit(struct net *net)
2290 {
2291         xt_proto_fini(net, AF_INET6);
2292 }
2293
2294 static struct pernet_operations ip6_tables_net_ops = {
2295         .init = ip6_tables_net_init,
2296         .exit = ip6_tables_net_exit,
2297 };
2298
2299 static int __init ip6_tables_init(void)
2300 {
2301         int ret;
2302
2303         ret = register_pernet_subsys(&ip6_tables_net_ops);
2304         if (ret < 0)
2305                 goto err1;
2306
2307         /* Noone else will be downing sem now, so we won't sleep */
2308         ret = xt_register_target(&ip6t_standard_target);
2309         if (ret < 0)
2310                 goto err2;
2311         ret = xt_register_target(&ip6t_error_target);
2312         if (ret < 0)
2313                 goto err3;
2314         ret = xt_register_match(&icmp6_matchstruct);
2315         if (ret < 0)
2316                 goto err4;
2317
2318         /* Register setsockopt */
2319         ret = nf_register_sockopt(&ip6t_sockopts);
2320         if (ret < 0)
2321                 goto err5;
2322
2323         printk(KERN_INFO "ip6_tables: (C) 2000-2006 Netfilter Core Team\n");
2324         return 0;
2325
2326 err5:
2327         xt_unregister_match(&icmp6_matchstruct);
2328 err4:
2329         xt_unregister_target(&ip6t_error_target);
2330 err3:
2331         xt_unregister_target(&ip6t_standard_target);
2332 err2:
2333         unregister_pernet_subsys(&ip6_tables_net_ops);
2334 err1:
2335         return ret;
2336 }
2337
2338 static void __exit ip6_tables_fini(void)
2339 {
2340         nf_unregister_sockopt(&ip6t_sockopts);
2341
2342         xt_unregister_match(&icmp6_matchstruct);
2343         xt_unregister_target(&ip6t_error_target);
2344         xt_unregister_target(&ip6t_standard_target);
2345
2346         unregister_pernet_subsys(&ip6_tables_net_ops);
2347 }
2348
2349 /*
2350  * find the offset to specified header or the protocol number of last header
2351  * if target < 0. "last header" is transport protocol header, ESP, or
2352  * "No next header".
2353  *
2354  * If target header is found, its offset is set in *offset and return protocol
2355  * number. Otherwise, return -1.
2356  *
2357  * If the first fragment doesn't contain the final protocol header or
2358  * NEXTHDR_NONE it is considered invalid.
2359  *
2360  * Note that non-1st fragment is special case that "the protocol number
2361  * of last header" is "next header" field in Fragment header. In this case,
2362  * *offset is meaningless and fragment offset is stored in *fragoff if fragoff
2363  * isn't NULL.
2364  *
2365  */
2366 int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset,
2367                   int target, unsigned short *fragoff)
2368 {
2369         unsigned int start = skb_network_offset(skb) + sizeof(struct ipv6hdr);
2370         u8 nexthdr = ipv6_hdr(skb)->nexthdr;
2371         unsigned int len = skb->len - start;
2372
2373         if (fragoff)
2374                 *fragoff = 0;
2375
2376         while (nexthdr != target) {
2377                 struct ipv6_opt_hdr _hdr, *hp;
2378                 unsigned int hdrlen;
2379
2380                 if ((!ipv6_ext_hdr(nexthdr)) || nexthdr == NEXTHDR_NONE) {
2381                         if (target < 0)
2382                                 break;
2383                         return -ENOENT;
2384                 }
2385
2386                 hp = skb_header_pointer(skb, start, sizeof(_hdr), &_hdr);
2387                 if (hp == NULL)
2388                         return -EBADMSG;
2389                 if (nexthdr == NEXTHDR_FRAGMENT) {
2390                         unsigned short _frag_off;
2391                         __be16 *fp;
2392                         fp = skb_header_pointer(skb,
2393                                                 start+offsetof(struct frag_hdr,
2394                                                                frag_off),
2395                                                 sizeof(_frag_off),
2396                                                 &_frag_off);
2397                         if (fp == NULL)
2398                                 return -EBADMSG;
2399
2400                         _frag_off = ntohs(*fp) & ~0x7;
2401                         if (_frag_off) {
2402                                 if (target < 0 &&
2403                                     ((!ipv6_ext_hdr(hp->nexthdr)) ||
2404                                      hp->nexthdr == NEXTHDR_NONE)) {
2405                                         if (fragoff)
2406                                                 *fragoff = _frag_off;
2407                                         return hp->nexthdr;
2408                                 }
2409                                 return -ENOENT;
2410                         }
2411                         hdrlen = 8;
2412                 } else if (nexthdr == NEXTHDR_AUTH)
2413                         hdrlen = (hp->hdrlen + 2) << 2;
2414                 else
2415                         hdrlen = ipv6_optlen(hp);
2416
2417                 nexthdr = hp->nexthdr;
2418                 len -= hdrlen;
2419                 start += hdrlen;
2420         }
2421
2422         *offset = start;
2423         return nexthdr;
2424 }
2425
2426 EXPORT_SYMBOL(ip6t_register_table);
2427 EXPORT_SYMBOL(ip6t_unregister_table);
2428 EXPORT_SYMBOL(ip6t_do_table);
2429 EXPORT_SYMBOL(ip6t_ext_hdr);
2430 EXPORT_SYMBOL(ipv6_find_hdr);
2431
2432 module_init(ip6_tables_init);
2433 module_exit(ip6_tables_fini);