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