SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv4/netfilter/ip_tables.c

   1 /*
   2  * Packet matching code.
   3  *
   4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   6  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  */
  11 #include <linux/cache.h>
  12 #include <linux/capability.h>
  13 #include <linux/skbuff.h>
  14 #include <linux/kmod.h>
  15 #include <linux/vmalloc.h>
  16 #include <linux/netdevice.h>
  17 #include <linux/module.h>
  18 #include <linux/icmp.h>
  19 #include <net/ip.h>
  20 #include <net/compat.h>
  21 #include <asm/uaccess.h>
  22 #include <linux/mutex.h>
  23 #include <linux/proc_fs.h>
  24 #include <linux/err.h>
  25 #include <linux/cpumask.h>
  26
  27 #include <linux/netfilter/x_tables.h>
  28 #include <linux/netfilter_ipv4/ip_tables.h>
  29
  30 MODULE_LICENSE("GPL");
  31 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  32 MODULE_DESCRIPTION("IPv4 packet filter");
  33
  34 /*#define DEBUG_IP_FIREWALL*/
  35 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  36 /*#define DEBUG_IP_FIREWALL_USER*/
  37
  38 #ifdef DEBUG_IP_FIREWALL
  39 #define dprintf(format, args...)  printk(format , ## args)
  40 #else
  41 #define dprintf(format, args...)
  42 #endif
  43
  44 #ifdef DEBUG_IP_FIREWALL_USER
  45 #define duprintf(format, args...) printk(format , ## args)
  46 #else
  47 #define duprintf(format, args...)
  48 #endif
  49
  50 #ifdef CONFIG_NETFILTER_DEBUG
  51 #define IP_NF_ASSERT(x)                                         \
  52 do {                                                            \
  53         if (!(x))                                               \
  54                 printk("IP_NF_ASSERT: %s:%s:%u\n",              \
  55                        __FUNCTION__, __FILE__, __LINE__);       \
  56 } while(0)
  57 #else
  58 #define IP_NF_ASSERT(x)
  59 #endif
  60
  61 #if 0
  62 /* All the better to debug you with... */
  63 #define static
  64 #define inline
  65 #endif
  66
  67 /*
  68    We keep a set of rules for each CPU, so we can avoid write-locking
  69    them in the softirq when updating the counters and therefore
  70    only need to read-lock in the softirq; doing a write_lock_bh() in user
  71    context stops packets coming through and allows user context to read
  72    the counters or update the rules.
  73
  74    Hence the start of any table is given by get_table() below.  */
  75
  76 /* Returns whether matches rule or not. */
  77 static inline int
  78 ip_packet_match(const struct iphdr *ip,
  79                 const char *indev,
  80                 const char *outdev,
  81                 const struct ipt_ip *ipinfo,
  82                 int isfrag)
  83 {
  84         size_t i;
  85         unsigned long ret;
  86
  87 #define FWINV(bool,invflg) ((bool) ^ !!(ipinfo->invflags & invflg))
  88
  89         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  90                   IPT_INV_SRCIP)
  91             || FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  92                      IPT_INV_DSTIP)) {
  93                 dprintf("Source or dest mismatch.\n");
  94
  95                 dprintf("SRC: %u.%u.%u.%u. Mask: %u.%u.%u.%u. Target: %u.%u.%u.%u.%s\n",
  96                         NIPQUAD(ip->saddr),
  97                         NIPQUAD(ipinfo->smsk.s_addr),
  98                         NIPQUAD(ipinfo->src.s_addr),
  99                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
 100                 dprintf("DST: %u.%u.%u.%u Mask: %u.%u.%u.%u Target: %u.%u.%u.%u.%s\n",
 101                         NIPQUAD(ip->daddr),
 102                         NIPQUAD(ipinfo->dmsk.s_addr),
 103                         NIPQUAD(ipinfo->dst.s_addr),
 104                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
 105                 return 0;
 106         }
 107
 108         /* Look for ifname matches; this should unroll nicely. */
 109         for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
 110                 ret |= (((const unsigned long *)indev)[i]
 111                         ^ ((const unsigned long *)ipinfo->iniface)[i])
 112                         & ((const unsigned long *)ipinfo->iniface_mask)[i];
 113         }
 114
 115         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 116                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 117                         indev, ipinfo->iniface,
 118                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
 119                 return 0;
 120         }
 121
 122         for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
 123                 ret |= (((const unsigned long *)outdev)[i]
 124                         ^ ((const unsigned long *)ipinfo->outiface)[i])
 125                         & ((const unsigned long *)ipinfo->outiface_mask)[i];
 126         }
 127
 128         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 129                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 130                         outdev, ipinfo->outiface,
 131                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
 132                 return 0;
 133         }
 134
 135         /* Check specific protocol */
 136         if (ipinfo->proto
 137             && FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 138                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 139                         ip->protocol, ipinfo->proto,
 140                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
 141                 return 0;
 142         }
 143
 144         /* If we have a fragment rule but the packet is not a fragment
 145          * then we return zero */
 146         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 147                 dprintf("Fragment rule but not fragment.%s\n",
 148                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 149                 return 0;
 150         }
 151
 152         return 1;
 153 }
 154
 155 static inline bool
 156 ip_checkentry(const struct ipt_ip *ip)
 157 {
 158         if (ip->flags & ~IPT_F_MASK) {
 159                 duprintf("Unknown flag bits set: %08X\n",
 160                          ip->flags & ~IPT_F_MASK);
 161                 return false;
 162         }
 163         if (ip->invflags & ~IPT_INV_MASK) {
 164                 duprintf("Unknown invflag bits set: %08X\n",
 165                          ip->invflags & ~IPT_INV_MASK);
 166                 return false;
 167         }
 168         return true;
 169 }
 170
 171 static unsigned int
 172 ipt_error(struct sk_buff *skb,
 173           const struct net_device *in,
 174           const struct net_device *out,
 175           unsigned int hooknum,
 176           const struct xt_target *target,
 177           const void *targinfo)
 178 {
 179         if (net_ratelimit())
 180                 printk("ip_tables: error: `%s'\n", (char *)targinfo);
 181
 182         return NF_DROP;
 183 }
 184
 185 static inline
 186 bool do_match(struct ipt_entry_match *m,
 187               const struct sk_buff *skb,
 188               const struct net_device *in,
 189               const struct net_device *out,
 190               int offset,
 191               bool *hotdrop)
 192 {
 193         /* Stop iteration if it doesn't match */
 194         if (!m->u.kernel.match->match(skb, in, out, m->u.kernel.match, m->data,
 195                                       offset, ip_hdrlen(skb), hotdrop))
 196                 return true;
 197         else
 198                 return false;
 199 }
 200
 201 static inline struct ipt_entry *
 202 get_entry(void *base, unsigned int offset)
 203 {
 204         return (struct ipt_entry *)(base + offset);
 205 }
 206
 207 /* All zeroes == unconditional rule. */
 208 static inline int
 209 unconditional(const struct ipt_ip *ip)
 210 {
 211         unsigned int i;
 212
 213         for (i = 0; i < sizeof(*ip)/sizeof(__u32); i++)
 214                 if (((__u32 *)ip)[i])
 215                         return 0;
 216
 217         return 1;
 218 }
 219
 220 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
 221     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
 222 static const char *hooknames[] = {
 223         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 224         [NF_INET_LOCAL_IN]              = "INPUT",
 225         [NF_INET_FORWARD]                       = "FORWARD",
 226         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 227         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 228 };
 229
 230 enum nf_ip_trace_comments {
 231         NF_IP_TRACE_COMMENT_RULE,
 232         NF_IP_TRACE_COMMENT_RETURN,
 233         NF_IP_TRACE_COMMENT_POLICY,
 234 };
 235
 236 static const char *comments[] = {
 237         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 238         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 239         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 240 };
 241
 242 static struct nf_loginfo trace_loginfo = {
 243         .type = NF_LOG_TYPE_LOG,
 244         .u = {
 245                 .log = {
 246                         .level = 4,
 247                         .logflags = NF_LOG_MASK,
 248                 },
 249         },
 250 };
 251
 252 static inline int
 253 get_chainname_rulenum(struct ipt_entry *s, struct ipt_entry *e,
 254                       char *hookname, char **chainname,
 255                       char **comment, unsigned int *rulenum)
 256 {
 257         struct ipt_standard_target *t = (void *)ipt_get_target(s);
 258
 259         if (strcmp(t->target.u.kernel.target->name, IPT_ERROR_TARGET) == 0) {
 260                 /* Head of user chain: ERROR target with chainname */
 261                 *chainname = t->target.data;
 262                 (*rulenum) = 0;
 263         } else if (s == e) {
 264                 (*rulenum)++;
 265
 266                 if (s->target_offset == sizeof(struct ipt_entry)
 267                    && strcmp(t->target.u.kernel.target->name,
 268                              IPT_STANDARD_TARGET) == 0
 269                    && t->verdict < 0
 270                    && unconditional(&s->ip)) {
 271                         /* Tail of chains: STANDARD target (return/policy) */
 272                         *comment = *chainname == hookname
 273                                 ? (char *)comments[NF_IP_TRACE_COMMENT_POLICY]
 274                                 : (char *)comments[NF_IP_TRACE_COMMENT_RETURN];
 275                 }
 276                 return 1;
 277         } else
 278                 (*rulenum)++;
 279
 280         return 0;
 281 }
 282
 283 static void trace_packet(struct sk_buff *skb,
 284                          unsigned int hook,
 285                          const struct net_device *in,
 286                          const struct net_device *out,
 287                          char *tablename,
 288                          struct xt_table_info *private,
 289                          struct ipt_entry *e)
 290 {
 291         void *table_base;
 292         struct ipt_entry *root;
 293         char *hookname, *chainname, *comment;
 294         unsigned int rulenum = 0;
 295
 296         table_base = (void *)private->entries[smp_processor_id()];
 297         root = get_entry(table_base, private->hook_entry[hook]);
 298
 299         hookname = chainname = (char *)hooknames[hook];
 300         comment = (char *)comments[NF_IP_TRACE_COMMENT_RULE];
 301
 302         IPT_ENTRY_ITERATE(root,
 303                           private->size - private->hook_entry[hook],
 304                           get_chainname_rulenum,
 305                           e, hookname, &chainname, &comment, &rulenum);
 306
 307         nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
 308                       "TRACE: %s:%s:%s:%u ",
 309                       tablename, chainname, comment, rulenum);
 310 }
 311 #endif
 312
 313 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 314 unsigned int
 315 ipt_do_table(struct sk_buff *skb,
 316              unsigned int hook,
 317              const struct net_device *in,
 318              const struct net_device *out,
 319              struct xt_table *table)
 320 {
 321         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 322         u_int16_t offset;
 323         struct iphdr *ip;
 324         u_int16_t datalen;
 325         bool hotdrop = false;
 326         /* Initializing verdict to NF_DROP keeps gcc happy. */
 327         unsigned int verdict = NF_DROP;
 328         const char *indev, *outdev;
 329         void *table_base;
 330         struct ipt_entry *e, *back;
 331         struct xt_table_info *private;
 332
 333         /* Initialization */
 334         ip = ip_hdr(skb);
 335         datalen = skb->len - ip->ihl * 4;
 336         indev = in ? in->name : nulldevname;
 337         outdev = out ? out->name : nulldevname;
 338         /* We handle fragments by dealing with the first fragment as
 339          * if it was a normal packet.  All other fragments are treated
 340          * normally, except that they will NEVER match rules that ask
 341          * things we don't know, ie. tcp syn flag or ports).  If the
 342          * rule is also a fragment-specific rule, non-fragments won't
 343          * match it. */
 344         offset = ntohs(ip->frag_off) & IP_OFFSET;
 345
 346         read_lock_bh(&table->lock);
 347         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 348         private = table->private;
 349         table_base = (void *)private->entries[smp_processor_id()];
 350         e = get_entry(table_base, private->hook_entry[hook]);
 351
 352         /* For return from builtin chain */
 353         back = get_entry(table_base, private->underflow[hook]);
 354
 355         do {
 356                 IP_NF_ASSERT(e);
 357                 IP_NF_ASSERT(back);
 358                 if (ip_packet_match(ip, indev, outdev, &e->ip, offset)) {
 359                         struct ipt_entry_target *t;
 360
 361                         if (IPT_MATCH_ITERATE(e, do_match,
 362                                               skb, in, out,
 363                                               offset, &hotdrop) != 0)
 364                                 goto no_match;
 365
 366                         ADD_COUNTER(e->counters, ntohs(ip->tot_len), 1);
 367
 368                         t = ipt_get_target(e);
 369                         IP_NF_ASSERT(t->u.kernel.target);
 370
 371 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
 372     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
 373                         /* The packet is traced: log it */
 374                         if (unlikely(skb->nf_trace))
 375                                 trace_packet(skb, hook, in, out,
 376                                              table->name, private, e);
 377 #endif
 378                         /* Standard target? */
 379                         if (!t->u.kernel.target->target) {
 380                                 int v;
 381
 382                                 v = ((struct ipt_standard_target *)t)->verdict;
 383                                 if (v < 0) {
 384                                         /* Pop from stack? */
 385                                         if (v != IPT_RETURN) {
 386                                                 verdict = (unsigned)(-v) - 1;
 387                                                 break;
 388                                         }
 389                                         e = back;
 390                                         back = get_entry(table_base,
 391                                                          back->comefrom);
 392                                         continue;
 393                                 }
 394                                 if (table_base + v != (void *)e + e->next_offset
 395                                     && !(e->ip.flags & IPT_F_GOTO)) {
 396                                         /* Save old back ptr in next entry */
 397                                         struct ipt_entry *next
 398                                                 = (void *)e + e->next_offset;
 399                                         next->comefrom
 400                                                 = (void *)back - table_base;
 401                                         /* set back pointer to next entry */
 402                                         back = next;
 403                                 }
 404
 405                                 e = get_entry(table_base, v);
 406                         } else {
 407                                 /* Targets which reenter must return
 408                                    abs. verdicts */
 409 #ifdef CONFIG_NETFILTER_DEBUG
 410                                 ((struct ipt_entry *)table_base)->comefrom
 411                                         = 0xeeeeeeec;
 412 #endif
 413                                 verdict = t->u.kernel.target->target(skb,
 414                                                                      in, out,
 415                                                                      hook,
 416                                                                      t->u.kernel.target,
 417                                                                      t->data);
 418
 419 #ifdef CONFIG_NETFILTER_DEBUG
 420                                 if (((struct ipt_entry *)table_base)->comefrom
 421                                     != 0xeeeeeeec
 422                                     && verdict == IPT_CONTINUE) {
 423                                         printk("Target %s reentered!\n",
 424                                                t->u.kernel.target->name);
 425                                         verdict = NF_DROP;
 426                                 }
 427                                 ((struct ipt_entry *)table_base)->comefrom
 428                                         = 0x57acc001;
 429 #endif
 430                                 /* Target might have changed stuff. */
 431                                 ip = ip_hdr(skb);
 432                                 datalen = skb->len - ip->ihl * 4;
 433
 434                                 if (verdict == IPT_CONTINUE)
 435                                         e = (void *)e + e->next_offset;
 436                                 else
 437                                         /* Verdict */
 438                                         break;
 439                         }
 440                 } else {
 441
 442                 no_match:
 443                         e = (void *)e + e->next_offset;
 444                 }
 445         } while (!hotdrop);
 446
 447         read_unlock_bh(&table->lock);
 448
 449 #ifdef DEBUG_ALLOW_ALL
 450         return NF_ACCEPT;
 451 #else
 452         if (hotdrop)
 453                 return NF_DROP;
 454         else return verdict;
 455 #endif
 456 }
 457
 458 /* Figures out from what hook each rule can be called: returns 0 if
 459    there are loops.  Puts hook bitmask in comefrom. */
 460 static int
 461 mark_source_chains(struct xt_table_info *newinfo,
 462                    unsigned int valid_hooks, void *entry0)
 463 {
 464         unsigned int hook;
 465
 466         /* No recursion; use packet counter to save back ptrs (reset
 467            to 0 as we leave), and comefrom to save source hook bitmask */
 468         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 469                 unsigned int pos = newinfo->hook_entry[hook];
 470                 struct ipt_entry *e
 471                         = (struct ipt_entry *)(entry0 + pos);
 472
 473                 if (!(valid_hooks & (1 << hook)))
 474                         continue;
 475
 476                 /* Set initial back pointer. */
 477                 e->counters.pcnt = pos;
 478
 479                 for (;;) {
 480                         struct ipt_standard_target *t
 481                                 = (void *)ipt_get_target(e);
 482                         int visited = e->comefrom & (1 << hook);
 483
 484                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 485                                 printk("iptables: loop hook %u pos %u %08X.\n",
 486                                        hook, pos, e->comefrom);
 487                                 return 0;
 488                         }
 489                         e->comefrom
 490                                 |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 491
 492                         /* Unconditional return/END. */
 493                         if ((e->target_offset == sizeof(struct ipt_entry)
 494                             && (strcmp(t->target.u.user.name,
 495                                        IPT_STANDARD_TARGET) == 0)
 496                             && t->verdict < 0
 497                             && unconditional(&e->ip)) || visited) {
 498                                 unsigned int oldpos, size;
 499
 500                                 if (t->verdict < -NF_MAX_VERDICT - 1) {
 501                                         duprintf("mark_source_chains: bad "
 502                                                 "negative verdict (%i)\n",
 503                                                                 t->verdict);
 504                                         return 0;
 505                                 }
 506
 507                                 /* Return: backtrack through the last
 508                                    big jump. */
 509                                 do {
 510                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 511 #ifdef DEBUG_IP_FIREWALL_USER
 512                                         if (e->comefrom
 513                                             & (1 << NF_INET_NUMHOOKS)) {
 514                                                 duprintf("Back unset "
 515                                                          "on hook %u "
 516                                                          "rule %u\n",
 517                                                          hook, pos);
 518                                         }
 519 #endif
 520                                         oldpos = pos;
 521                                         pos = e->counters.pcnt;
 522                                         e->counters.pcnt = 0;
 523
 524                                         /* We're at the start. */
 525                                         if (pos == oldpos)
 526                                                 goto next;
 527
 528                                         e = (struct ipt_entry *)
 529                                                 (entry0 + pos);
 530                                 } while (oldpos == pos + e->next_offset);
 531
 532                                 /* Move along one */
 533                                 size = e->next_offset;
 534                                 e = (struct ipt_entry *)
 535                                         (entry0 + pos + size);
 536                                 e->counters.pcnt = pos;
 537                                 pos += size;
 538                         } else {
 539                                 int newpos = t->verdict;
 540
 541                                 if (strcmp(t->target.u.user.name,
 542                                            IPT_STANDARD_TARGET) == 0
 543                                     && newpos >= 0) {
 544                                         if (newpos > newinfo->size -
 545                                                 sizeof(struct ipt_entry)) {
 546                                                 duprintf("mark_source_chains: "
 547                                                         "bad verdict (%i)\n",
 548                                                                 newpos);
 549                                                 return 0;
 550                                         }
 551                                         /* This a jump; chase it. */
 552                                         duprintf("Jump rule %u -> %u\n",
 553                                                  pos, newpos);
 554                                 } else {
 555                                         /* ... this is a fallthru */
 556                                         newpos = pos + e->next_offset;
 557                                 }
 558                                 e = (struct ipt_entry *)
 559                                         (entry0 + newpos);
 560                                 e->counters.pcnt = pos;
 561                                 pos = newpos;
 562                         }
 563                 }
 564                 next:
 565                 duprintf("Finished chain %u\n", hook);
 566         }
 567         return 1;
 568 }
 569
 570 static inline int
 571 cleanup_match(struct ipt_entry_match *m, unsigned int *i)
 572 {
 573         if (i && (*i)-- == 0)
 574                 return 1;
 575
 576         if (m->u.kernel.match->destroy)
 577                 m->u.kernel.match->destroy(m->u.kernel.match, m->data);
 578         module_put(m->u.kernel.match->me);
 579         return 0;
 580 }
 581
 582 static inline int
 583 check_entry(struct ipt_entry *e, const char *name)
 584 {
 585         struct ipt_entry_target *t;
 586
 587         if (!ip_checkentry(&e->ip)) {
 588                 duprintf("ip_tables: ip check failed %p %s.\n", e, name);
 589                 return -EINVAL;
 590         }
 591
 592         if (e->target_offset + sizeof(struct ipt_entry_target) > e->next_offset)
 593                 return -EINVAL;
 594
 595         t = ipt_get_target(e);
 596         if (e->target_offset + t->u.target_size > e->next_offset)
 597                 return -EINVAL;
 598
 599         return 0;
 600 }
 601
 602 static inline int check_match(struct ipt_entry_match *m, const char *name,
 603                               const struct ipt_ip *ip,
 604                               unsigned int hookmask, unsigned int *i)
 605 {
 606         struct xt_match *match;
 607         int ret;
 608
 609         match = m->u.kernel.match;
 610         ret = xt_check_match(match, AF_INET, m->u.match_size - sizeof(*m),
 611                              name, hookmask, ip->proto,
 612                              ip->invflags & IPT_INV_PROTO);
 613         if (!ret && m->u.kernel.match->checkentry
 614             && !m->u.kernel.match->checkentry(name, ip, match, m->data,
 615                                               hookmask)) {
 616                 duprintf("ip_tables: check failed for `%s'.\n",
 617                          m->u.kernel.match->name);
 618                 ret = -EINVAL;
 619         }
 620         if (!ret)
 621                 (*i)++;
 622         return ret;
 623 }
 624
 625 static inline int
 626 find_check_match(struct ipt_entry_match *m,
 627                  const char *name,
 628                  const struct ipt_ip *ip,
 629                  unsigned int hookmask,
 630                  unsigned int *i)
 631 {
 632         struct xt_match *match;
 633         int ret;
 634
 635         match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
 636                                                    m->u.user.revision),
 637                                         "ipt_%s", m->u.user.name);
 638         if (IS_ERR(match) || !match) {
 639                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 640                 return match ? PTR_ERR(match) : -ENOENT;
 641         }
 642         m->u.kernel.match = match;
 643
 644         ret = check_match(m, name, ip, hookmask, i);
 645         if (ret)
 646                 goto err;
 647
 648         return 0;
 649 err:
 650         module_put(m->u.kernel.match->me);
 651         return ret;
 652 }
 653
 654 static inline int check_target(struct ipt_entry *e, const char *name)
 655 {
 656         struct ipt_entry_target *t;
 657         struct xt_target *target;
 658         int ret;
 659
 660         t = ipt_get_target(e);
 661         target = t->u.kernel.target;
 662         ret = xt_check_target(target, AF_INET, t->u.target_size - sizeof(*t),
 663                               name, e->comefrom, e->ip.proto,
 664                               e->ip.invflags & IPT_INV_PROTO);
 665         if (!ret && t->u.kernel.target->checkentry
 666             && !t->u.kernel.target->checkentry(name, e, target, t->data,
 667                                                e->comefrom)) {
 668                 duprintf("ip_tables: check failed for `%s'.\n",
 669                          t->u.kernel.target->name);
 670                 ret = -EINVAL;
 671         }
 672         return ret;
 673 }
 674
 675 static inline int
 676 find_check_entry(struct ipt_entry *e, const char *name, unsigned int size,
 677                  unsigned int *i)
 678 {
 679         struct ipt_entry_target *t;
 680         struct xt_target *target;
 681         int ret;
 682         unsigned int j;
 683
 684         ret = check_entry(e, name);
 685         if (ret)
 686                 return ret;
 687
 688         j = 0;
 689         ret = IPT_MATCH_ITERATE(e, find_check_match, name, &e->ip,
 690                                 e->comefrom, &j);
 691         if (ret != 0)
 692                 goto cleanup_matches;
 693
 694         t = ipt_get_target(e);
 695         target = try_then_request_module(xt_find_target(AF_INET,
 696                                                         t->u.user.name,
 697                                                         t->u.user.revision),
 698                                          "ipt_%s", t->u.user.name);
 699         if (IS_ERR(target) || !target) {
 700                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 701                 ret = target ? PTR_ERR(target) : -ENOENT;
 702                 goto cleanup_matches;
 703         }
 704         t->u.kernel.target = target;
 705
 706         ret = check_target(e, name);
 707         if (ret)
 708                 goto err;
 709
 710         (*i)++;
 711         return 0;
 712  err:
 713         module_put(t->u.kernel.target->me);
 714  cleanup_matches:
 715         IPT_MATCH_ITERATE(e, cleanup_match, &j);
 716         return ret;
 717 }
 718
 719 static inline int
 720 check_entry_size_and_hooks(struct ipt_entry *e,
 721                            struct xt_table_info *newinfo,
 722                            unsigned char *base,
 723                            unsigned char *limit,
 724                            const unsigned int *hook_entries,
 725                            const unsigned int *underflows,
 726                            unsigned int *i)
 727 {
 728         unsigned int h;
 729
 730         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0
 731             || (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 732                 duprintf("Bad offset %p\n", e);
 733                 return -EINVAL;
 734         }
 735
 736         if (e->next_offset
 737             < sizeof(struct ipt_entry) + sizeof(struct ipt_entry_target)) {
 738                 duprintf("checking: element %p size %u\n",
 739                          e, e->next_offset);
 740                 return -EINVAL;
 741         }
 742
 743         /* Check hooks & underflows */
 744         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 745                 if ((unsigned char *)e - base == hook_entries[h])
 746                         newinfo->hook_entry[h] = hook_entries[h];
 747                 if ((unsigned char *)e - base == underflows[h])
 748                         newinfo->underflow[h] = underflows[h];
 749         }
 750
 751         /* FIXME: underflows must be unconditional, standard verdicts
 752            < 0 (not IPT_RETURN). --RR */
 753
 754         /* Clear counters and comefrom */
 755         e->counters = ((struct xt_counters) { 0, 0 });
 756         e->comefrom = 0;
 757
 758         (*i)++;
 759         return 0;
 760 }
 761
 762 static inline int
 763 cleanup_entry(struct ipt_entry *e, unsigned int *i)
 764 {
 765         struct ipt_entry_target *t;
 766
 767         if (i && (*i)-- == 0)
 768                 return 1;
 769
 770         /* Cleanup all matches */
 771         IPT_MATCH_ITERATE(e, cleanup_match, NULL);
 772         t = ipt_get_target(e);
 773         if (t->u.kernel.target->destroy)
 774                 t->u.kernel.target->destroy(t->u.kernel.target, t->data);
 775         module_put(t->u.kernel.target->me);
 776         return 0;
 777 }
 778
 779 /* Checks and translates the user-supplied table segment (held in
 780    newinfo) */
 781 static int
 782 translate_table(const char *name,
 783                 unsigned int valid_hooks,
 784                 struct xt_table_info *newinfo,
 785                 void *entry0,
 786                 unsigned int size,
 787                 unsigned int number,
 788                 const unsigned int *hook_entries,
 789                 const unsigned int *underflows)
 790 {
 791         unsigned int i;
 792         int ret;
 793
 794         newinfo->size = size;
 795         newinfo->number = number;
 796
 797         /* Init all hooks to impossible value. */
 798         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 799                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 800                 newinfo->underflow[i] = 0xFFFFFFFF;
 801         }
 802
 803         duprintf("translate_table: size %u\n", newinfo->size);
 804         i = 0;
 805         /* Walk through entries, checking offsets. */
 806         ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
 807                                 check_entry_size_and_hooks,
 808                                 newinfo,
 809                                 entry0,
 810                                 entry0 + size,
 811                                 hook_entries, underflows, &i);
 812         if (ret != 0)
 813                 return ret;
 814
 815         if (i != number) {
 816                 duprintf("translate_table: %u not %u entries\n",
 817                          i, number);
 818                 return -EINVAL;
 819         }
 820
 821         /* Check hooks all assigned */
 822         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 823                 /* Only hooks which are valid */
 824                 if (!(valid_hooks & (1 << i)))
 825                         continue;
 826                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 827                         duprintf("Invalid hook entry %u %u\n",
 828                                  i, hook_entries[i]);
 829                         return -EINVAL;
 830                 }
 831                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 832                         duprintf("Invalid underflow %u %u\n",
 833                                  i, underflows[i]);
 834                         return -EINVAL;
 835                 }
 836         }
 837
 838         if (!mark_source_chains(newinfo, valid_hooks, entry0))
 839                 return -ELOOP;
 840
 841         /* Finally, each sanity check must pass */
 842         i = 0;
 843         ret = IPT_ENTRY_ITERATE(entry0, newinfo->size,
 844                                 find_check_entry, name, size, &i);
 845
 846         if (ret != 0) {
 847                 IPT_ENTRY_ITERATE(entry0, newinfo->size,
 848                                 cleanup_entry, &i);
 849                 return ret;
 850         }
 851
 852         /* And one copy for every other CPU */
 853         for_each_possible_cpu(i) {
 854                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
 855                         memcpy(newinfo->entries[i], entry0, newinfo->size);
 856         }
 857
 858         return ret;
 859 }
 860
 861 /* Gets counters. */
 862 static inline int
 863 add_entry_to_counter(const struct ipt_entry *e,
 864                      struct xt_counters total[],
 865                      unsigned int *i)
 866 {
 867         ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
 868
 869         (*i)++;
 870         return 0;
 871 }
 872
 873 static inline int
 874 set_entry_to_counter(const struct ipt_entry *e,
 875                      struct ipt_counters total[],
 876                      unsigned int *i)
 877 {
 878         SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
 879
 880         (*i)++;
 881         return 0;
 882 }
 883
 884 static void
 885 get_counters(const struct xt_table_info *t,
 886              struct xt_counters counters[])
 887 {
 888         unsigned int cpu;
 889         unsigned int i;
 890         unsigned int curcpu;
 891
 892         /* Instead of clearing (by a previous call to memset())
 893          * the counters and using adds, we set the counters
 894          * with data used by 'current' CPU
 895          * We dont care about preemption here.
 896          */
 897         curcpu = raw_smp_processor_id();
 898
 899         i = 0;
 900         IPT_ENTRY_ITERATE(t->entries[curcpu],
 901                           t->size,
 902                           set_entry_to_counter,
 903                           counters,
 904                           &i);
 905
 906         for_each_possible_cpu(cpu) {
 907                 if (cpu == curcpu)
 908                         continue;
 909                 i = 0;
 910                 IPT_ENTRY_ITERATE(t->entries[cpu],
 911                                   t->size,
 912                                   add_entry_to_counter,
 913                                   counters,
 914                                   &i);
 915         }
 916 }
 917
 918 static inline struct xt_counters * alloc_counters(struct xt_table *table)
 919 {
 920         unsigned int countersize;
 921         struct xt_counters *counters;
 922         struct xt_table_info *private = table->private;
 923
 924         /* We need atomic snapshot of counters: rest doesn't change
 925            (other than comefrom, which userspace doesn't care
 926            about). */
 927         countersize = sizeof(struct xt_counters) * private->number;
 928         counters = vmalloc_node(countersize, numa_node_id());
 929
 930         if (counters == NULL)
 931                 return ERR_PTR(-ENOMEM);
 932
 933         /* First, sum counters... */
 934         write_lock_bh(&table->lock);
 935         get_counters(private, counters);
 936         write_unlock_bh(&table->lock);
 937
 938         return counters;
 939 }
 940
 941 static int
 942 copy_entries_to_user(unsigned int total_size,
 943                      struct xt_table *table,
 944                      void __user *userptr)
 945 {
 946         unsigned int off, num;
 947         struct ipt_entry *e;
 948         struct xt_counters *counters;
 949         struct xt_table_info *private = table->private;
 950         int ret = 0;
 951         void *loc_cpu_entry;
 952
 953         counters = alloc_counters(table);
 954         if (IS_ERR(counters))
 955                 return PTR_ERR(counters);
 956
 957         /* choose the copy that is on our node/cpu, ...
 958          * This choice is lazy (because current thread is
 959          * allowed to migrate to another cpu)
 960          */
 961         loc_cpu_entry = private->entries[raw_smp_processor_id()];
 962         /* ... then copy entire thing ... */
 963         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 964                 ret = -EFAULT;
 965                 goto free_counters;
 966         }
 967
 968         /* FIXME: use iterator macros --RR */
 969         /* ... then go back and fix counters and names */
 970         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 971                 unsigned int i;
 972                 struct ipt_entry_match *m;
 973                 struct ipt_entry_target *t;
 974
 975                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 976                 if (copy_to_user(userptr + off
 977                                  + offsetof(struct ipt_entry, counters),
 978                                  &counters[num],
 979                                  sizeof(counters[num])) != 0) {
 980                         ret = -EFAULT;
 981                         goto free_counters;
 982                 }
 983
 984                 for (i = sizeof(struct ipt_entry);
 985                      i < e->target_offset;
 986                      i += m->u.match_size) {
 987                         m = (void *)e + i;
 988
 989                         if (copy_to_user(userptr + off + i
 990                                          + offsetof(struct ipt_entry_match,
 991                                                     u.user.name),
 992                                          m->u.kernel.match->name,
 993                                          strlen(m->u.kernel.match->name)+1)
 994                             != 0) {
 995                                 ret = -EFAULT;
 996                                 goto free_counters;
 997                         }
 998                 }
 999
1000                 t = ipt_get_target(e);
1001                 if (copy_to_user(userptr + off + e->target_offset
1002                                  + offsetof(struct ipt_entry_target,
1003                                             u.user.name),
1004                                  t->u.kernel.target->name,
1005                                  strlen(t->u.kernel.target->name)+1) != 0) {
1006                         ret = -EFAULT;
1007                         goto free_counters;
1008                 }
1009         }
1010
1011  free_counters:
1012         vfree(counters);
1013         return ret;
1014 }
1015
1016 #ifdef CONFIG_COMPAT
1017 struct compat_delta {
1018         struct compat_delta *next;
1019         unsigned int offset;
1020         short delta;
1021 };
1022
1023 static struct compat_delta *compat_offsets;
1024
1025 static int compat_add_offset(unsigned int offset, short delta)
1026 {
1027         struct compat_delta *tmp;
1028
1029         tmp = kmalloc(sizeof(struct compat_delta), GFP_KERNEL);
1030         if (!tmp)
1031                 return -ENOMEM;
1032         tmp->offset = offset;
1033         tmp->delta = delta;
1034         if (compat_offsets) {
1035                 tmp->next = compat_offsets->next;
1036                 compat_offsets->next = tmp;
1037         } else {
1038                 compat_offsets = tmp;
1039                 tmp->next = NULL;
1040         }
1041         return 0;
1042 }
1043
1044 static void compat_flush_offsets(void)
1045 {
1046         struct compat_delta *tmp, *next;
1047
1048         if (compat_offsets) {
1049                 for (tmp = compat_offsets; tmp; tmp = next) {
1050                         next = tmp->next;
1051                         kfree(tmp);
1052                 }
1053                 compat_offsets = NULL;
1054         }
1055 }
1056
1057 static short compat_calc_jump(unsigned int offset)
1058 {
1059         struct compat_delta *tmp;
1060         short delta;
1061
1062         for (tmp = compat_offsets, delta = 0; tmp; tmp = tmp->next)
1063                 if (tmp->offset < offset)
1064                         delta += tmp->delta;
1065         return delta;
1066 }
1067
1068 static void compat_standard_from_user(void *dst, void *src)
1069 {
1070         int v = *(compat_int_t *)src;
1071
1072         if (v > 0)
1073                 v += compat_calc_jump(v);
1074         memcpy(dst, &v, sizeof(v));
1075 }
1076
1077 static int compat_standard_to_user(void __user *dst, void *src)
1078 {
1079         compat_int_t cv = *(int *)src;
1080
1081         if (cv > 0)
1082                 cv -= compat_calc_jump(cv);
1083         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1084 }
1085
1086 static inline int
1087 compat_calc_match(struct ipt_entry_match *m, int *size)
1088 {
1089         *size += xt_compat_match_offset(m->u.kernel.match);
1090         return 0;
1091 }
1092
1093 static int compat_calc_entry(struct ipt_entry *e,
1094                              const struct xt_table_info *info,
1095                              void *base, struct xt_table_info *newinfo)
1096 {
1097         struct ipt_entry_target *t;
1098         unsigned int entry_offset;
1099         int off, i, ret;
1100
1101         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1102         entry_offset = (void *)e - base;
1103         IPT_MATCH_ITERATE(e, compat_calc_match, &off);
1104         t = ipt_get_target(e);
1105         off += xt_compat_target_offset(t->u.kernel.target);
1106         newinfo->size -= off;
1107         ret = compat_add_offset(entry_offset, off);
1108         if (ret)
1109                 return ret;
1110
1111         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1112                 if (info->hook_entry[i] &&
1113                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1114                         newinfo->hook_entry[i] -= off;
1115                 if (info->underflow[i] &&
1116                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1117                         newinfo->underflow[i] -= off;
1118         }
1119         return 0;
1120 }
1121
1122 static int compat_table_info(const struct xt_table_info *info,
1123                              struct xt_table_info *newinfo)
1124 {
1125         void *loc_cpu_entry;
1126
1127         if (!newinfo || !info)
1128                 return -EINVAL;
1129
1130         /* we dont care about newinfo->entries[] */
1131         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1132         newinfo->initial_entries = 0;
1133         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1134         return IPT_ENTRY_ITERATE(loc_cpu_entry, info->size,
1135                                  compat_calc_entry, info, loc_cpu_entry,
1136                                  newinfo);
1137 }
1138 #endif
1139
1140 static int get_info(void __user *user, int *len, int compat)
1141 {
1142         char name[IPT_TABLE_MAXNAMELEN];
1143         struct xt_table *t;
1144         int ret;
1145
1146         if (*len != sizeof(struct ipt_getinfo)) {
1147                 duprintf("length %u != %u\n", *len,
1148                         (unsigned int)sizeof(struct ipt_getinfo));
1149                 return -EINVAL;
1150         }
1151
1152         if (copy_from_user(name, user, sizeof(name)) != 0)
1153                 return -EFAULT;
1154
1155         name[IPT_TABLE_MAXNAMELEN-1] = '\0';
1156 #ifdef CONFIG_COMPAT
1157         if (compat)
1158                 xt_compat_lock(AF_INET);
1159 #endif
1160         t = try_then_request_module(xt_find_table_lock(AF_INET, name),
1161                                     "iptable_%s", name);
1162         if (t && !IS_ERR(t)) {
1163                 struct ipt_getinfo info;
1164                 struct xt_table_info *private = t->private;
1165
1166 #ifdef CONFIG_COMPAT
1167                 if (compat) {
1168                         struct xt_table_info tmp;
1169                         ret = compat_table_info(private, &tmp);
1170                         compat_flush_offsets();
1171                         private = &tmp;
1172                 }
1173 #endif
1174                 info.valid_hooks = t->valid_hooks;
1175                 memcpy(info.hook_entry, private->hook_entry,
1176                        sizeof(info.hook_entry));
1177                 memcpy(info.underflow, private->underflow,
1178                        sizeof(info.underflow));
1179                 info.num_entries = private->number;
1180                 info.size = private->size;
1181                 strcpy(info.name, name);
1182
1183                 if (copy_to_user(user, &info, *len) != 0)
1184                         ret = -EFAULT;
1185                 else
1186                         ret = 0;
1187
1188                 xt_table_unlock(t);
1189                 module_put(t->me);
1190         } else
1191                 ret = t ? PTR_ERR(t) : -ENOENT;
1192 #ifdef CONFIG_COMPAT
1193         if (compat)
1194                 xt_compat_unlock(AF_INET);
1195 #endif
1196         return ret;
1197 }
1198
1199 static int
1200 get_entries(struct ipt_get_entries __user *uptr, int *len)
1201 {
1202         int ret;
1203         struct ipt_get_entries get;
1204         struct xt_table *t;
1205
1206         if (*len < sizeof(get)) {
1207                 duprintf("get_entries: %u < %d\n", *len,
1208                                 (unsigned int)sizeof(get));
1209                 return -EINVAL;
1210         }
1211         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1212                 return -EFAULT;
1213         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1214                 duprintf("get_entries: %u != %u\n", *len,
1215                                 (unsigned int)(sizeof(struct ipt_get_entries) +
1216                                 get.size));
1217                 return -EINVAL;
1218         }
1219
1220         t = xt_find_table_lock(AF_INET, get.name);
1221         if (t && !IS_ERR(t)) {
1222                 struct xt_table_info *private = t->private;
1223                 duprintf("t->private->number = %u\n",
1224                          private->number);
1225                 if (get.size == private->size)
1226                         ret = copy_entries_to_user(private->size,
1227                                                    t, uptr->entrytable);
1228                 else {
1229                         duprintf("get_entries: I've got %u not %u!\n",
1230                                  private->size,
1231                                  get.size);
1232                         ret = -EINVAL;
1233                 }
1234                 module_put(t->me);
1235                 xt_table_unlock(t);
1236         } else
1237                 ret = t ? PTR_ERR(t) : -ENOENT;
1238
1239         return ret;
1240 }
1241
1242 static int
1243 __do_replace(const char *name, unsigned int valid_hooks,
1244              struct xt_table_info *newinfo, unsigned int num_counters,
1245              void __user *counters_ptr)
1246 {
1247         int ret;
1248         struct xt_table *t;
1249         struct xt_table_info *oldinfo;
1250         struct xt_counters *counters;
1251         void *loc_cpu_old_entry;
1252
1253         ret = 0;
1254         counters = vmalloc(num_counters * sizeof(struct xt_counters));
1255         if (!counters) {
1256                 ret = -ENOMEM;
1257                 goto out;
1258         }
1259
1260         t = try_then_request_module(xt_find_table_lock(AF_INET, name),
1261                                     "iptable_%s", name);
1262         if (!t || IS_ERR(t)) {
1263                 ret = t ? PTR_ERR(t) : -ENOENT;
1264                 goto free_newinfo_counters_untrans;
1265         }
1266
1267         /* You lied! */
1268         if (valid_hooks != t->valid_hooks) {
1269                 duprintf("Valid hook crap: %08X vs %08X\n",
1270                          valid_hooks, t->valid_hooks);
1271                 ret = -EINVAL;
1272                 goto put_module;
1273         }
1274
1275         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1276         if (!oldinfo)
1277                 goto put_module;
1278
1279         /* Update module usage count based on number of rules */
1280         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1281                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1282         if ((oldinfo->number > oldinfo->initial_entries) ||
1283             (newinfo->number <= oldinfo->initial_entries))
1284                 module_put(t->me);
1285         if ((oldinfo->number > oldinfo->initial_entries) &&
1286             (newinfo->number <= oldinfo->initial_entries))
1287                 module_put(t->me);
1288
1289         /* Get the old counters. */
1290         get_counters(oldinfo, counters);
1291         /* Decrease module usage counts and free resource */
1292         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1293         IPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,
1294                           NULL);
1295         xt_free_table_info(oldinfo);
1296         if (copy_to_user(counters_ptr, counters,
1297                          sizeof(struct xt_counters) * num_counters) != 0)
1298                 ret = -EFAULT;
1299         vfree(counters);
1300         xt_table_unlock(t);
1301         return ret;
1302
1303  put_module:
1304         module_put(t->me);
1305         xt_table_unlock(t);
1306  free_newinfo_counters_untrans:
1307         vfree(counters);
1308  out:
1309         return ret;
1310 }
1311
1312 static int
1313 do_replace(void __user *user, unsigned int len)
1314 {
1315         int ret;
1316         struct ipt_replace tmp;
1317         struct xt_table_info *newinfo;
1318         void *loc_cpu_entry;
1319
1320         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1321                 return -EFAULT;
1322
1323         /* Hack: Causes ipchains to give correct error msg --RR */
1324         if (len != sizeof(tmp) + tmp.size)
1325                 return -ENOPROTOOPT;
1326
1327         /* overflow check */
1328         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1329                 return -ENOMEM;
1330
1331         newinfo = xt_alloc_table_info(tmp.size);
1332         if (!newinfo)
1333                 return -ENOMEM;
1334
1335         /* choose the copy that is our node/cpu */
1336         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1337         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1338                            tmp.size) != 0) {
1339                 ret = -EFAULT;
1340                 goto free_newinfo;
1341         }
1342
1343         ret = translate_table(tmp.name, tmp.valid_hooks,
1344                               newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
1345                               tmp.hook_entry, tmp.underflow);
1346         if (ret != 0)
1347                 goto free_newinfo;
1348
1349         duprintf("ip_tables: Translated table\n");
1350
1351         ret = __do_replace(tmp.name, tmp.valid_hooks, newinfo,
1352                            tmp.num_counters, tmp.counters);
1353         if (ret)
1354                 goto free_newinfo_untrans;
1355         return 0;
1356
1357  free_newinfo_untrans:
1358         IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry,NULL);
1359  free_newinfo:
1360         xt_free_table_info(newinfo);
1361         return ret;
1362 }
1363
1364 /* We're lazy, and add to the first CPU; overflow works its fey magic
1365  * and everything is OK. */
1366 static inline int
1367 add_counter_to_entry(struct ipt_entry *e,
1368                      const struct xt_counters addme[],
1369                      unsigned int *i)
1370 {
1371 #if 0
1372         duprintf("add_counter: Entry %u %lu/%lu + %lu/%lu\n",
1373                  *i,
1374                  (long unsigned int)e->counters.pcnt,
1375                  (long unsigned int)e->counters.bcnt,
1376                  (long unsigned int)addme[*i].pcnt,
1377                  (long unsigned int)addme[*i].bcnt);
1378 #endif
1379
1380         ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
1381
1382         (*i)++;
1383         return 0;
1384 }
1385
1386 static int
1387 do_add_counters(void __user *user, unsigned int len, int compat)
1388 {
1389         unsigned int i;
1390         struct xt_counters_info tmp;
1391         struct xt_counters *paddc;
1392         unsigned int num_counters;
1393         char *name;
1394         int size;
1395         void *ptmp;
1396         struct xt_table *t;
1397         struct xt_table_info *private;
1398         int ret = 0;
1399         void *loc_cpu_entry;
1400 #ifdef CONFIG_COMPAT
1401         struct compat_xt_counters_info compat_tmp;
1402
1403         if (compat) {
1404                 ptmp = &compat_tmp;
1405                 size = sizeof(struct compat_xt_counters_info);
1406         } else
1407 #endif
1408         {
1409                 ptmp = &tmp;
1410                 size = sizeof(struct xt_counters_info);
1411         }
1412
1413         if (copy_from_user(ptmp, user, size) != 0)
1414                 return -EFAULT;
1415
1416 #ifdef CONFIG_COMPAT
1417         if (compat) {
1418                 num_counters = compat_tmp.num_counters;
1419                 name = compat_tmp.name;
1420         } else
1421 #endif
1422         {
1423                 num_counters = tmp.num_counters;
1424                 name = tmp.name;
1425         }
1426
1427         if (len != size + num_counters * sizeof(struct xt_counters))
1428                 return -EINVAL;
1429
1430         paddc = vmalloc_node(len - size, numa_node_id());
1431         if (!paddc)
1432                 return -ENOMEM;
1433
1434         if (copy_from_user(paddc, user + size, len - size) != 0) {
1435                 ret = -EFAULT;
1436                 goto free;
1437         }
1438
1439         t = xt_find_table_lock(AF_INET, name);
1440         if (!t || IS_ERR(t)) {
1441                 ret = t ? PTR_ERR(t) : -ENOENT;
1442                 goto free;
1443         }
1444
1445         write_lock_bh(&t->lock);
1446         private = t->private;
1447         if (private->number != num_counters) {
1448                 ret = -EINVAL;
1449                 goto unlock_up_free;
1450         }
1451
1452         i = 0;
1453         /* Choose the copy that is on our node */
1454         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1455         IPT_ENTRY_ITERATE(loc_cpu_entry,
1456                           private->size,
1457                           add_counter_to_entry,
1458                           paddc,
1459                           &i);
1460  unlock_up_free:
1461         write_unlock_bh(&t->lock);
1462         xt_table_unlock(t);
1463         module_put(t->me);
1464  free:
1465         vfree(paddc);
1466
1467         return ret;
1468 }
1469
1470 #ifdef CONFIG_COMPAT
1471 struct compat_ipt_replace {
1472         char                    name[IPT_TABLE_MAXNAMELEN];
1473         u32                     valid_hooks;
1474         u32                     num_entries;
1475         u32                     size;
1476         u32                     hook_entry[NF_INET_NUMHOOKS];
1477         u32                     underflow[NF_INET_NUMHOOKS];
1478         u32                     num_counters;
1479         compat_uptr_t           counters;       /* struct ipt_counters * */
1480         struct compat_ipt_entry entries[0];
1481 };
1482
1483 static int
1484 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1485                           compat_uint_t *size, struct xt_counters *counters,
1486                           unsigned int *i)
1487 {
1488         struct ipt_entry_target *t;
1489         struct compat_ipt_entry __user *ce;
1490         u_int16_t target_offset, next_offset;
1491         compat_uint_t origsize;
1492         int ret;
1493
1494         ret = -EFAULT;
1495         origsize = *size;
1496         ce = (struct compat_ipt_entry __user *)*dstptr;
1497         if (copy_to_user(ce, e, sizeof(struct ipt_entry)))
1498                 goto out;
1499
1500         if (copy_to_user(&ce->counters, &counters[*i], sizeof(counters[*i])))
1501                 goto out;
1502
1503         *dstptr += sizeof(struct compat_ipt_entry);
1504         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1505
1506         ret = IPT_MATCH_ITERATE(e, xt_compat_match_to_user, dstptr, size);
1507         target_offset = e->target_offset - (origsize - *size);
1508         if (ret)
1509                 goto out;
1510         t = ipt_get_target(e);
1511         ret = xt_compat_target_to_user(t, dstptr, size);
1512         if (ret)
1513                 goto out;
1514         ret = -EFAULT;
1515         next_offset = e->next_offset - (origsize - *size);
1516         if (put_user(target_offset, &ce->target_offset))
1517                 goto out;
1518         if (put_user(next_offset, &ce->next_offset))
1519                 goto out;
1520
1521         (*i)++;
1522         return 0;
1523 out:
1524         return ret;
1525 }
1526
1527 static inline int
1528 compat_find_calc_match(struct ipt_entry_match *m,
1529                        const char *name,
1530                        const struct ipt_ip *ip,
1531                        unsigned int hookmask,
1532                        int *size, int *i)
1533 {
1534         struct xt_match *match;
1535
1536         match = try_then_request_module(xt_find_match(AF_INET, m->u.user.name,
1537                                                       m->u.user.revision),
1538                                         "ipt_%s", m->u.user.name);
1539         if (IS_ERR(match) || !match) {
1540                 duprintf("compat_check_calc_match: `%s' not found\n",
1541                          m->u.user.name);
1542                 return match ? PTR_ERR(match) : -ENOENT;
1543         }
1544         m->u.kernel.match = match;
1545         *size += xt_compat_match_offset(match);
1546
1547         (*i)++;
1548         return 0;
1549 }
1550
1551 static inline int
1552 compat_release_match(struct ipt_entry_match *m, unsigned int *i)
1553 {
1554         if (i && (*i)-- == 0)
1555                 return 1;
1556
1557         module_put(m->u.kernel.match->me);
1558         return 0;
1559 }
1560
1561 static inline int
1562 compat_release_entry(struct compat_ipt_entry *e, unsigned int *i)
1563 {
1564         struct ipt_entry_target *t;
1565
1566         if (i && (*i)-- == 0)
1567                 return 1;
1568
1569         /* Cleanup all matches */
1570         COMPAT_IPT_MATCH_ITERATE(e, compat_release_match, NULL);
1571         t = compat_ipt_get_target(e);
1572         module_put(t->u.kernel.target->me);
1573         return 0;
1574 }
1575
1576 static inline int
1577 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1578                                   struct xt_table_info *newinfo,
1579                                   unsigned int *size,
1580                                   unsigned char *base,
1581                                   unsigned char *limit,
1582                                   unsigned int *hook_entries,
1583                                   unsigned int *underflows,
1584                                   unsigned int *i,
1585                                   const char *name)
1586 {
1587         struct ipt_entry_target *t;
1588         struct xt_target *target;
1589         unsigned int entry_offset;
1590         int ret, off, h, j;
1591
1592         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1593         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0
1594             || (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1595                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1596                 return -EINVAL;
1597         }
1598
1599         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1600                              sizeof(struct compat_xt_entry_target)) {
1601                 duprintf("checking: element %p size %u\n",
1602                          e, e->next_offset);
1603                 return -EINVAL;
1604         }
1605
1606         /* For purposes of check_entry casting the compat entry is fine */
1607         ret = check_entry((struct ipt_entry *)e, name);
1608         if (ret)
1609                 return ret;
1610
1611         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1612         entry_offset = (void *)e - (void *)base;
1613         j = 0;
1614         ret = COMPAT_IPT_MATCH_ITERATE(e, compat_find_calc_match, name,
1615                                        &e->ip, e->comefrom, &off, &j);
1616         if (ret != 0)
1617                 goto release_matches;
1618
1619         t = compat_ipt_get_target(e);
1620         target = try_then_request_module(xt_find_target(AF_INET,
1621                                                         t->u.user.name,
1622                                                         t->u.user.revision),
1623                                          "ipt_%s", t->u.user.name);
1624         if (IS_ERR(target) || !target) {
1625                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1626                          t->u.user.name);
1627                 ret = target ? PTR_ERR(target) : -ENOENT;
1628                 goto release_matches;
1629         }
1630         t->u.kernel.target = target;
1631
1632         off += xt_compat_target_offset(target);
1633         *size += off;
1634         ret = compat_add_offset(entry_offset, off);
1635         if (ret)
1636                 goto out;
1637
1638         /* Check hooks & underflows */
1639         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1640                 if ((unsigned char *)e - base == hook_entries[h])
1641                         newinfo->hook_entry[h] = hook_entries[h];
1642                 if ((unsigned char *)e - base == underflows[h])
1643                         newinfo->underflow[h] = underflows[h];
1644         }
1645
1646         /* Clear counters and comefrom */
1647         memset(&e->counters, 0, sizeof(e->counters));
1648         e->comefrom = 0;
1649
1650         (*i)++;
1651         return 0;
1652
1653 out:
1654         module_put(t->u.kernel.target->me);
1655 release_matches:
1656         IPT_MATCH_ITERATE(e, compat_release_match, &j);
1657         return ret;
1658 }
1659
1660 static int
1661 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1662                             unsigned int *size, const char *name,
1663                             struct xt_table_info *newinfo, unsigned char *base)
1664 {
1665         struct ipt_entry_target *t;
1666         struct xt_target *target;
1667         struct ipt_entry *de;
1668         unsigned int origsize;
1669         int ret, h;
1670
1671         ret = 0;
1672         origsize = *size;
1673         de = (struct ipt_entry *)*dstptr;
1674         memcpy(de, e, sizeof(struct ipt_entry));
1675         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1676
1677         *dstptr += sizeof(struct ipt_entry);
1678         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1679
1680         ret = COMPAT_IPT_MATCH_ITERATE(e, xt_compat_match_from_user,
1681                                        dstptr, size);
1682         if (ret)
1683                 return ret;
1684         de->target_offset = e->target_offset - (origsize - *size);
1685         t = compat_ipt_get_target(e);
1686         target = t->u.kernel.target;
1687         xt_compat_target_from_user(t, dstptr, size);
1688
1689         de->next_offset = e->next_offset - (origsize - *size);
1690         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1691                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1692                         newinfo->hook_entry[h] -= origsize - *size;
1693                 if ((unsigned char *)de - base < newinfo->underflow[h])
1694                         newinfo->underflow[h] -= origsize - *size;
1695         }
1696         return ret;
1697 }
1698
1699 static inline int compat_check_entry(struct ipt_entry *e, const char *name,
1700                                      unsigned int *i)
1701 {
1702         int j, ret;
1703
1704         j = 0;
1705         ret = IPT_MATCH_ITERATE(e, check_match, name, &e->ip, e->comefrom, &j);
1706         if (ret)
1707                 goto cleanup_matches;
1708
1709         ret = check_target(e, name);
1710         if (ret)
1711                 goto cleanup_matches;
1712
1713         (*i)++;
1714         return 0;
1715
1716  cleanup_matches:
1717         IPT_MATCH_ITERATE(e, cleanup_match, &j);
1718         return ret;
1719 }
1720
1721 static int
1722 translate_compat_table(const char *name,
1723                        unsigned int valid_hooks,
1724                        struct xt_table_info **pinfo,
1725                        void **pentry0,
1726                        unsigned int total_size,
1727                        unsigned int number,
1728                        unsigned int *hook_entries,
1729                        unsigned int *underflows)
1730 {
1731         unsigned int i, j;
1732         struct xt_table_info *newinfo, *info;
1733         void *pos, *entry0, *entry1;
1734         unsigned int size;
1735         int ret;
1736
1737         info = *pinfo;
1738         entry0 = *pentry0;
1739         size = total_size;
1740         info->number = number;
1741
1742         /* Init all hooks to impossible value. */
1743         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1744                 info->hook_entry[i] = 0xFFFFFFFF;
1745                 info->underflow[i] = 0xFFFFFFFF;
1746         }
1747
1748         duprintf("translate_compat_table: size %u\n", info->size);
1749         j = 0;
1750         xt_compat_lock(AF_INET);
1751         /* Walk through entries, checking offsets. */
1752         ret = COMPAT_IPT_ENTRY_ITERATE(entry0, total_size,
1753                                        check_compat_entry_size_and_hooks,
1754                                        info, &size, entry0,
1755                                        entry0 + total_size,
1756                                        hook_entries, underflows, &j, name);
1757         if (ret != 0)
1758                 goto out_unlock;
1759
1760         ret = -EINVAL;
1761         if (j != number) {
1762                 duprintf("translate_compat_table: %u not %u entries\n",
1763                          j, number);
1764                 goto out_unlock;
1765         }
1766
1767         /* Check hooks all assigned */
1768         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1769                 /* Only hooks which are valid */
1770                 if (!(valid_hooks & (1 << i)))
1771                         continue;
1772                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1773                         duprintf("Invalid hook entry %u %u\n",
1774                                  i, hook_entries[i]);
1775                         goto out_unlock;
1776                 }
1777                 if (info->underflow[i] == 0xFFFFFFFF) {
1778                         duprintf("Invalid underflow %u %u\n",
1779                                  i, underflows[i]);
1780                         goto out_unlock;
1781                 }
1782         }
1783
1784         ret = -ENOMEM;
1785         newinfo = xt_alloc_table_info(size);
1786         if (!newinfo)
1787                 goto out_unlock;
1788
1789         newinfo->number = number;
1790         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1791                 newinfo->hook_entry[i] = info->hook_entry[i];
1792                 newinfo->underflow[i] = info->underflow[i];
1793         }
1794         entry1 = newinfo->entries[raw_smp_processor_id()];
1795         pos = entry1;
1796         size = total_size;
1797         ret = COMPAT_IPT_ENTRY_ITERATE(entry0, total_size,
1798                                        compat_copy_entry_from_user, &pos, &size,
1799                                        name, newinfo, entry1);
1800         compat_flush_offsets();
1801         xt_compat_unlock(AF_INET);
1802         if (ret)
1803                 goto free_newinfo;
1804
1805         ret = -ELOOP;
1806         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1807                 goto free_newinfo;
1808
1809         i = 0;
1810         ret = IPT_ENTRY_ITERATE(entry1, newinfo->size, compat_check_entry,
1811                                 name, &i);
1812         if (ret) {
1813                 j -= i;
1814                 COMPAT_IPT_ENTRY_ITERATE_CONTINUE(entry0, newinfo->size, i,
1815                                                   compat_release_entry, &j);
1816                 IPT_ENTRY_ITERATE(entry1, newinfo->size, cleanup_entry, &i);
1817                 xt_free_table_info(newinfo);
1818                 return ret;
1819         }
1820
1821         /* And one copy for every other CPU */
1822         for_each_possible_cpu(i)
1823                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1824                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1825
1826         *pinfo = newinfo;
1827         *pentry0 = entry1;
1828         xt_free_table_info(info);
1829         return 0;
1830
1831 free_newinfo:
1832         xt_free_table_info(newinfo);
1833 out:
1834         COMPAT_IPT_ENTRY_ITERATE(entry0, total_size, compat_release_entry, &j);
1835         return ret;
1836 out_unlock:
1837         compat_flush_offsets();
1838         xt_compat_unlock(AF_INET);
1839         goto out;
1840 }
1841
1842 static int
1843 compat_do_replace(void __user *user, unsigned int len)
1844 {
1845         int ret;
1846         struct compat_ipt_replace tmp;
1847         struct xt_table_info *newinfo;
1848         void *loc_cpu_entry;
1849
1850         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1851                 return -EFAULT;
1852
1853         /* Hack: Causes ipchains to give correct error msg --RR */
1854         if (len != sizeof(tmp) + tmp.size)
1855                 return -ENOPROTOOPT;
1856
1857         /* overflow check */
1858         if (tmp.size >= INT_MAX / num_possible_cpus())
1859                 return -ENOMEM;
1860         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1861                 return -ENOMEM;
1862
1863         newinfo = xt_alloc_table_info(tmp.size);
1864         if (!newinfo)
1865                 return -ENOMEM;
1866
1867         /* choose the copy that is our node/cpu */
1868         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1869         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1870                            tmp.size) != 0) {
1871                 ret = -EFAULT;
1872                 goto free_newinfo;
1873         }
1874
1875         ret = translate_compat_table(tmp.name, tmp.valid_hooks,
1876                                      &newinfo, &loc_cpu_entry, tmp.size,
1877                                      tmp.num_entries, tmp.hook_entry,
1878                                      tmp.underflow);
1879         if (ret != 0)
1880                 goto free_newinfo;
1881
1882         duprintf("compat_do_replace: Translated table\n");
1883
1884         ret = __do_replace(tmp.name, tmp.valid_hooks, newinfo,
1885                            tmp.num_counters, compat_ptr(tmp.counters));
1886         if (ret)
1887                 goto free_newinfo_untrans;
1888         return 0;
1889
1890  free_newinfo_untrans:
1891         IPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
1892  free_newinfo:
1893         xt_free_table_info(newinfo);
1894         return ret;
1895 }
1896
1897 static int
1898 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1899                       unsigned int len)
1900 {
1901         int ret;
1902
1903         if (!capable(CAP_NET_ADMIN))
1904                 return -EPERM;
1905
1906         switch (cmd) {
1907         case IPT_SO_SET_REPLACE:
1908                 ret = compat_do_replace(user, len);
1909                 break;
1910
1911         case IPT_SO_SET_ADD_COUNTERS:
1912                 ret = do_add_counters(user, len, 1);
1913                 break;
1914
1915         default:
1916                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1917                 ret = -EINVAL;
1918         }
1919
1920         return ret;
1921 }
1922
1923 struct compat_ipt_get_entries {
1924         char name[IPT_TABLE_MAXNAMELEN];
1925         compat_uint_t size;
1926         struct compat_ipt_entry entrytable[0];
1927 };
1928
1929 static int
1930 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1931                             void __user *userptr)
1932 {
1933         struct xt_counters *counters;
1934         struct xt_table_info *private = table->private;
1935         void __user *pos;
1936         unsigned int size;
1937         int ret = 0;
1938         void *loc_cpu_entry;
1939         unsigned int i = 0;
1940
1941         counters = alloc_counters(table);
1942         if (IS_ERR(counters))
1943                 return PTR_ERR(counters);
1944
1945         /* choose the copy that is on our node/cpu, ...
1946          * This choice is lazy (because current thread is
1947          * allowed to migrate to another cpu)
1948          */
1949         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1950         pos = userptr;
1951         size = total_size;
1952         ret = IPT_ENTRY_ITERATE(loc_cpu_entry, total_size,
1953                                 compat_copy_entry_to_user,
1954                                 &pos, &size, counters, &i);
1955
1956         vfree(counters);
1957         return ret;
1958 }
1959
1960 static int
1961 compat_get_entries(struct compat_ipt_get_entries __user *uptr, int *len)
1962 {
1963         int ret;
1964         struct compat_ipt_get_entries get;
1965         struct xt_table *t;
1966
1967         if (*len < sizeof(get)) {
1968                 duprintf("compat_get_entries: %u < %u\n",
1969                          *len, (unsigned int)sizeof(get));
1970                 return -EINVAL;
1971         }
1972
1973         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1974                 return -EFAULT;
1975
1976         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1977                 duprintf("compat_get_entries: %u != %u\n", *len,
1978                          (unsigned int)(sizeof(struct compat_ipt_get_entries) +
1979                          get.size));
1980                 return -EINVAL;
1981         }
1982
1983         xt_compat_lock(AF_INET);
1984         t = xt_find_table_lock(AF_INET, get.name);
1985         if (t && !IS_ERR(t)) {
1986                 struct xt_table_info *private = t->private;
1987                 struct xt_table_info info;
1988                 duprintf("t->private->number = %u\n",
1989                          private->number);
1990                 ret = compat_table_info(private, &info);
1991                 if (!ret && get.size == info.size) {
1992                         ret = compat_copy_entries_to_user(private->size,
1993                                                           t, uptr->entrytable);
1994                 } else if (!ret) {
1995                         duprintf("compat_get_entries: I've got %u not %u!\n",
1996                                  private->size,
1997                                  get.size);
1998                         ret = -EINVAL;
1999                 }
2000                 compat_flush_offsets();
2001                 module_put(t->me);
2002                 xt_table_unlock(t);
2003         } else
2004                 ret = t ? PTR_ERR(t) : -ENOENT;
2005
2006         xt_compat_unlock(AF_INET);
2007         return ret;
2008 }
2009
2010 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
2011
2012 static int
2013 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2014 {
2015         int ret;
2016
2017         if (!capable(CAP_NET_ADMIN))
2018                 return -EPERM;
2019
2020         switch (cmd) {
2021         case IPT_SO_GET_INFO:
2022                 ret = get_info(user, len, 1);
2023                 break;
2024         case IPT_SO_GET_ENTRIES:
2025                 ret = compat_get_entries(user, len);
2026                 break;
2027         default:
2028                 ret = do_ipt_get_ctl(sk, cmd, user, len);
2029         }
2030         return ret;
2031 }
2032 #endif
2033
2034 static int
2035 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2036 {
2037         int ret;
2038
2039         if (!capable(CAP_NET_ADMIN))
2040                 return -EPERM;
2041
2042         switch (cmd) {
2043         case IPT_SO_SET_REPLACE:
2044                 ret = do_replace(user, len);
2045                 break;
2046
2047         case IPT_SO_SET_ADD_COUNTERS:
2048                 ret = do_add_counters(user, len, 0);
2049                 break;
2050
2051         default:
2052                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2053                 ret = -EINVAL;
2054         }
2055
2056         return ret;
2057 }
2058
2059 static int
2060 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2061 {
2062         int ret;
2063
2064         if (!capable(CAP_NET_ADMIN))
2065                 return -EPERM;
2066
2067         switch (cmd) {
2068         case IPT_SO_GET_INFO:
2069                 ret = get_info(user, len, 0);
2070                 break;
2071
2072         case IPT_SO_GET_ENTRIES:
2073                 ret = get_entries(user, len);
2074                 break;
2075
2076         case IPT_SO_GET_REVISION_MATCH:
2077         case IPT_SO_GET_REVISION_TARGET: {
2078                 struct ipt_get_revision rev;
2079                 int target;
2080
2081                 if (*len != sizeof(rev)) {
2082                         ret = -EINVAL;
2083                         break;
2084                 }
2085                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2086                         ret = -EFAULT;
2087                         break;
2088                 }
2089
2090                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2091                         target = 1;
2092                 else
2093                         target = 0;
2094
2095                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2096                                                          rev.revision,
2097                                                          target, &ret),
2098                                         "ipt_%s", rev.name);
2099                 break;
2100         }
2101
2102         default:
2103                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2104                 ret = -EINVAL;
2105         }
2106
2107         return ret;
2108 }
2109
2110 int ipt_register_table(struct xt_table *table, const struct ipt_replace *repl)
2111 {
2112         int ret;
2113         struct xt_table_info *newinfo;
2114         struct xt_table_info bootstrap
2115                 = { 0, 0, 0, { 0 }, { 0 }, { } };
2116         void *loc_cpu_entry;
2117
2118         newinfo = xt_alloc_table_info(repl->size);
2119         if (!newinfo)
2120                 return -ENOMEM;
2121
2122         /* choose the copy on our node/cpu
2123          * but dont care of preemption
2124          */
2125         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2126         memcpy(loc_cpu_entry, repl->entries, repl->size);
2127
2128         ret = translate_table(table->name, table->valid_hooks,
2129                               newinfo, loc_cpu_entry, repl->size,
2130                               repl->num_entries,
2131                               repl->hook_entry,
2132                               repl->underflow);
2133         if (ret != 0) {
2134                 xt_free_table_info(newinfo);
2135                 return ret;
2136         }
2137
2138         ret = xt_register_table(table, &bootstrap, newinfo);
2139         if (ret != 0) {
2140                 xt_free_table_info(newinfo);
2141                 return ret;
2142         }
2143
2144         return 0;
2145 }
2146
2147 void ipt_unregister_table(struct xt_table *table)
2148 {
2149         struct xt_table_info *private;
2150         void *loc_cpu_entry;
2151
2152         private = xt_unregister_table(table);
2153
2154         /* Decrease module usage counts and free resources */
2155         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2156         IPT_ENTRY_ITERATE(loc_cpu_entry, private->size, cleanup_entry, NULL);
2157         xt_free_table_info(private);
2158 }
2159
2160 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2161 static inline bool
2162 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2163                      u_int8_t type, u_int8_t code,
2164                      bool invert)
2165 {
2166         return ((test_type == 0xFF) || (type == test_type && code >= min_code && code <= max_code))
2167                 ^ invert;
2168 }
2169
2170 static bool
2171 icmp_match(const struct sk_buff *skb,
2172            const struct net_device *in,
2173            const struct net_device *out,
2174            const struct xt_match *match,
2175            const void *matchinfo,
2176            int offset,
2177            unsigned int protoff,
2178            bool *hotdrop)
2179 {
2180         struct icmphdr _icmph, *ic;
2181         const struct ipt_icmp *icmpinfo = matchinfo;
2182
2183         /* Must not be a fragment. */
2184         if (offset)
2185                 return false;
2186
2187         ic = skb_header_pointer(skb, protoff, sizeof(_icmph), &_icmph);
2188         if (ic == NULL) {
2189                 /* We've been asked to examine this packet, and we
2190                  * can't.  Hence, no choice but to drop.
2191                  */
2192                 duprintf("Dropping evil ICMP tinygram.\n");
2193                 *hotdrop = true;
2194                 return false;
2195         }
2196
2197         return icmp_type_code_match(icmpinfo->type,
2198                                     icmpinfo->code[0],
2199                                     icmpinfo->code[1],
2200                                     ic->type, ic->code,
2201                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2202 }
2203
2204 /* Called when user tries to insert an entry of this type. */
2205 static bool
2206 icmp_checkentry(const char *tablename,
2207            const void *info,
2208            const struct xt_match *match,
2209            void *matchinfo,
2210            unsigned int hook_mask)
2211 {
2212         const struct ipt_icmp *icmpinfo = matchinfo;
2213
2214         /* Must specify no unknown invflags */
2215         return !(icmpinfo->invflags & ~IPT_ICMP_INV);
2216 }
2217
2218 /* The built-in targets: standard (NULL) and error. */
2219 static struct xt_target ipt_standard_target __read_mostly = {
2220         .name           = IPT_STANDARD_TARGET,
2221         .targetsize     = sizeof(int),
2222         .family         = AF_INET,
2223 #ifdef CONFIG_COMPAT
2224         .compatsize     = sizeof(compat_int_t),
2225         .compat_from_user = compat_standard_from_user,
2226         .compat_to_user = compat_standard_to_user,
2227 #endif
2228 };
2229
2230 static struct xt_target ipt_error_target __read_mostly = {
2231         .name           = IPT_ERROR_TARGET,
2232         .target         = ipt_error,
2233         .targetsize     = IPT_FUNCTION_MAXNAMELEN,
2234         .family         = AF_INET,
2235 };
2236
2237 static struct nf_sockopt_ops ipt_sockopts = {
2238         .pf             = PF_INET,
2239         .set_optmin     = IPT_BASE_CTL,
2240         .set_optmax     = IPT_SO_SET_MAX+1,
2241         .set            = do_ipt_set_ctl,
2242 #ifdef CONFIG_COMPAT
2243         .compat_set     = compat_do_ipt_set_ctl,
2244 #endif
2245         .get_optmin     = IPT_BASE_CTL,
2246         .get_optmax     = IPT_SO_GET_MAX+1,
2247         .get            = do_ipt_get_ctl,
2248 #ifdef CONFIG_COMPAT
2249         .compat_get     = compat_do_ipt_get_ctl,
2250 #endif
2251         .owner          = THIS_MODULE,
2252 };
2253
2254 static struct xt_match icmp_matchstruct __read_mostly = {
2255         .name           = "icmp",
2256         .match          = icmp_match,
2257         .matchsize      = sizeof(struct ipt_icmp),
2258         .proto          = IPPROTO_ICMP,
2259         .family         = AF_INET,
2260         .checkentry     = icmp_checkentry,
2261 };
2262
2263 static int __init ip_tables_init(void)
2264 {
2265         int ret;
2266
2267         ret = xt_proto_init(AF_INET);
2268         if (ret < 0)
2269                 goto err1;
2270
2271         /* Noone else will be downing sem now, so we won't sleep */
2272         ret = xt_register_target(&ipt_standard_target);
2273         if (ret < 0)
2274                 goto err2;
2275         ret = xt_register_target(&ipt_error_target);
2276         if (ret < 0)
2277                 goto err3;
2278         ret = xt_register_match(&icmp_matchstruct);
2279         if (ret < 0)
2280                 goto err4;
2281
2282         /* Register setsockopt */
2283         ret = nf_register_sockopt(&ipt_sockopts);
2284         if (ret < 0)
2285                 goto err5;
2286
2287         printk(KERN_INFO "ip_tables: (C) 2000-2006 Netfilter Core Team\n");
2288         return 0;
2289
2290 err5:
2291         xt_unregister_match(&icmp_matchstruct);
2292 err4:
2293         xt_unregister_target(&ipt_error_target);
2294 err3:
2295         xt_unregister_target(&ipt_standard_target);
2296 err2:
2297         xt_proto_fini(AF_INET);
2298 err1:
2299         return ret;
2300 }
2301
2302 static void __exit ip_tables_fini(void)
2303 {
2304         nf_unregister_sockopt(&ipt_sockopts);
2305
2306         xt_unregister_match(&icmp_matchstruct);
2307         xt_unregister_target(&ipt_error_target);
2308         xt_unregister_target(&ipt_standard_target);
2309
2310         xt_proto_fini(AF_INET);
2311 }
2312
2313 EXPORT_SYMBOL(ipt_register_table);
2314 EXPORT_SYMBOL(ipt_unregister_table);
2315 EXPORT_SYMBOL(ipt_do_table);
2316 module_init(ip_tables_init);
2317 module_exit(ip_tables_fini);