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