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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  12 #include <linux/cache.h>
  13 #include <linux/capability.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/icmp.h>
  20 #include <net/ip.h>
  21 #include <net/compat.h>
  22 #include <asm/uaccess.h>
  23 #include <linux/mutex.h>
  24 #include <linux/proc_fs.h>
  25 #include <linux/err.h>
  26 #include <linux/cpumask.h>
  27
  28 #include <linux/netfilter/x_tables.h>
  29 #include <linux/netfilter_ipv4/ip_tables.h>
  30 #include <net/netfilter/nf_log.h>
  31 #include "../../netfilter/xt_repldata.h"
  32
  33 MODULE_LICENSE("GPL");
  34 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  35 MODULE_DESCRIPTION("IPv4 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...) pr_info(format , ## args)
  43 #else
  44 #define dprintf(format, args...)
  45 #endif
  46
  47 #ifdef DEBUG_IP_FIREWALL_USER
  48 #define duprintf(format, args...) pr_info(format , ## args)
  49 #else
  50 #define duprintf(format, args...)
  51 #endif
  52
  53 #ifdef CONFIG_NETFILTER_DEBUG
  54 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  55 #else
  56 #define IP_NF_ASSERT(x)
  57 #endif
  58
  59 #if 0
  60 /* All the better to debug you with... */
  61 #define static
  62 #define inline
  63 #endif
  64
  65 void *ipt_alloc_initial_table(const struct xt_table *info)
  66 {
  67         return xt_alloc_initial_table(ipt, IPT);
  68 }
  69 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  70
  71 /* Returns whether matches rule or not. */
  72 /* Performance critical - called for every packet */
  73 static inline bool
  74 ip_packet_match(const struct iphdr *ip,
  75                 const char *indev,
  76                 const char *outdev,
  77                 const struct ipt_ip *ipinfo,
  78                 int isfrag)
  79 {
  80         unsigned long ret;
  81
  82 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  83
  84         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  85                   IPT_INV_SRCIP) ||
  86             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  87                   IPT_INV_DSTIP)) {
  88                 dprintf("Source or dest mismatch.\n");
  89
  90                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  91                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  92                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  93                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  94                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  95                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  96                 return false;
  97         }
  98
  99         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 100
 101         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 102                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 103                         indev, ipinfo->iniface,
 104                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
 105                 return false;
 106         }
 107
 108         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 109
 110         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 111                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 112                         outdev, ipinfo->outiface,
 113                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
 114                 return false;
 115         }
 116
 117         /* Check specific protocol */
 118         if (ipinfo->proto &&
 119             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 120                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 121                         ip->protocol, ipinfo->proto,
 122                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
 123                 return false;
 124         }
 125
 126         /* If we have a fragment rule but the packet is not a fragment
 127          * then we return zero */
 128         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 129                 dprintf("Fragment rule but not fragment.%s\n",
 130                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 131                 return false;
 132         }
 133
 134         return true;
 135 }
 136
 137 static bool
 138 ip_checkentry(const struct ipt_ip *ip)
 139 {
 140         if (ip->flags & ~IPT_F_MASK) {
 141                 duprintf("Unknown flag bits set: %08X\n",
 142                          ip->flags & ~IPT_F_MASK);
 143                 return false;
 144         }
 145         if (ip->invflags & ~IPT_INV_MASK) {
 146                 duprintf("Unknown invflag bits set: %08X\n",
 147                          ip->invflags & ~IPT_INV_MASK);
 148                 return false;
 149         }
 150         return true;
 151 }
 152
 153 static unsigned int
 154 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 155 {
 156         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 157
 158         return NF_DROP;
 159 }
 160
 161 /* Performance critical */
 162 static inline struct ipt_entry *
 163 get_entry(const void *base, unsigned int offset)
 164 {
 165         return (struct ipt_entry *)(base + offset);
 166 }
 167
 168 /* All zeroes == unconditional rule. */
 169 /* Mildly perf critical (only if packet tracing is on) */
 170 static inline bool unconditional(const struct ipt_ip *ip)
 171 {
 172         static const struct ipt_ip uncond;
 173
 174         return memcmp(ip, &uncond, sizeof(uncond)) == 0;
 175 #undef FWINV
 176 }
 177
 178 /* for const-correctness */
 179 static inline const struct xt_entry_target *
 180 ipt_get_target_c(const struct ipt_entry *e)
 181 {
 182         return ipt_get_target((struct ipt_entry *)e);
 183 }
 184
 185 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
 186     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
 187 static const char *const hooknames[] = {
 188         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 189         [NF_INET_LOCAL_IN]              = "INPUT",
 190         [NF_INET_FORWARD]               = "FORWARD",
 191         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 192         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 193 };
 194
 195 enum nf_ip_trace_comments {
 196         NF_IP_TRACE_COMMENT_RULE,
 197         NF_IP_TRACE_COMMENT_RETURN,
 198         NF_IP_TRACE_COMMENT_POLICY,
 199 };
 200
 201 static const char *const comments[] = {
 202         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 203         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 204         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 205 };
 206
 207 static struct nf_loginfo trace_loginfo = {
 208         .type = NF_LOG_TYPE_LOG,
 209         .u = {
 210                 .log = {
 211                         .level = 4,
 212                         .logflags = NF_LOG_MASK,
 213                 },
 214         },
 215 };
 216
 217 /* Mildly perf critical (only if packet tracing is on) */
 218 static inline int
 219 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 220                       const char *hookname, const char **chainname,
 221                       const char **comment, unsigned int *rulenum)
 222 {
 223         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 224
 225         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 226                 /* Head of user chain: ERROR target with chainname */
 227                 *chainname = t->target.data;
 228                 (*rulenum) = 0;
 229         } else if (s == e) {
 230                 (*rulenum)++;
 231
 232                 if (s->target_offset == sizeof(struct ipt_entry) &&
 233                     strcmp(t->target.u.kernel.target->name,
 234                            XT_STANDARD_TARGET) == 0 &&
 235                    t->verdict < 0 &&
 236                    unconditional(&s->ip)) {
 237                         /* Tail of chains: STANDARD target (return/policy) */
 238                         *comment = *chainname == hookname
 239                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 240                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 241                 }
 242                 return 1;
 243         } else
 244                 (*rulenum)++;
 245
 246         return 0;
 247 }
 248
 249 static void trace_packet(const struct sk_buff *skb,
 250                          unsigned int hook,
 251                          const struct net_device *in,
 252                          const struct net_device *out,
 253                          const char *tablename,
 254                          const struct xt_table_info *private,
 255                          const struct ipt_entry *e)
 256 {
 257         const void *table_base;
 258         const struct ipt_entry *root;
 259         const char *hookname, *chainname, *comment;
 260         const struct ipt_entry *iter;
 261         unsigned int rulenum = 0;
 262
 263         table_base = private->entries[smp_processor_id()];
 264         root = get_entry(table_base, private->hook_entry[hook]);
 265
 266         hookname = chainname = hooknames[hook];
 267         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 268
 269         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 270                 if (get_chainname_rulenum(iter, e, hookname,
 271                     &chainname, &comment, &rulenum) != 0)
 272                         break;
 273
 274         nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
 275                       "TRACE: %s:%s:%s:%u ",
 276                       tablename, chainname, comment, rulenum);
 277 }
 278 #endif
 279
 280 static inline __pure
 281 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 282 {
 283         return (void *)entry + entry->next_offset;
 284 }
 285
 286 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 287 unsigned int
 288 ipt_do_table(struct sk_buff *skb,
 289              unsigned int hook,
 290              const struct net_device *in,
 291              const struct net_device *out,
 292              struct xt_table *table)
 293 {
 294         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 295         const struct iphdr *ip;
 296         /* Initializing verdict to NF_DROP keeps gcc happy. */
 297         unsigned int verdict = NF_DROP;
 298         const char *indev, *outdev;
 299         const void *table_base;
 300         struct ipt_entry *e, **jumpstack;
 301         unsigned int *stackptr, origptr, cpu;
 302         const struct xt_table_info *private;
 303         struct xt_action_param acpar;
 304         unsigned int addend;
 305
 306         /* Initialization */
 307         ip = ip_hdr(skb);
 308         indev = in ? in->name : nulldevname;
 309         outdev = out ? out->name : nulldevname;
 310         /* We handle fragments by dealing with the first fragment as
 311          * if it was a normal packet.  All other fragments are treated
 312          * normally, except that they will NEVER match rules that ask
 313          * things we don't know, ie. tcp syn flag or ports).  If the
 314          * rule is also a fragment-specific rule, non-fragments won't
 315          * match it. */
 316         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 317         acpar.thoff   = ip_hdrlen(skb);
 318         acpar.hotdrop = false;
 319         acpar.in      = in;
 320         acpar.out     = out;
 321         acpar.family  = NFPROTO_IPV4;
 322         acpar.hooknum = hook;
 323
 324         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 325         local_bh_disable();
 326         addend = xt_write_recseq_begin();
 327         private = table->private;
 328         cpu        = smp_processor_id();
 329         table_base = private->entries[cpu];
 330         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 331         stackptr   = per_cpu_ptr(private->stackptr, cpu);
 332         origptr    = *stackptr;
 333
 334         e = get_entry(table_base, private->hook_entry[hook]);
 335
 336         pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
 337                  table->name, hook, origptr,
 338                  get_entry(table_base, private->underflow[hook]));
 339
 340         do {
 341                 const struct xt_entry_target *t;
 342                 const struct xt_entry_match *ematch;
 343
 344                 IP_NF_ASSERT(e);
 345                 if (!ip_packet_match(ip, indev, outdev,
 346                     &e->ip, acpar.fragoff)) {
 347  no_match:
 348                         e = ipt_next_entry(e);
 349                         continue;
 350                 }
 351
 352                 xt_ematch_foreach(ematch, e) {
 353                         acpar.match     = ematch->u.kernel.match;
 354                         acpar.matchinfo = ematch->data;
 355                         if (!acpar.match->match(skb, &acpar))
 356                                 goto no_match;
 357                 }
 358
 359                 ADD_COUNTER(e->counters, skb->len, 1);
 360
 361                 t = ipt_get_target(e);
 362                 IP_NF_ASSERT(t->u.kernel.target);
 363
 364 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
 365     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
 366                 /* The packet is traced: log it */
 367                 if (unlikely(skb->nf_trace))
 368                         trace_packet(skb, hook, in, out,
 369                                      table->name, private, e);
 370 #endif
 371                 /* Standard target? */
 372                 if (!t->u.kernel.target->target) {
 373                         int v;
 374
 375                         v = ((struct xt_standard_target *)t)->verdict;
 376                         if (v < 0) {
 377                                 /* Pop from stack? */
 378                                 if (v != XT_RETURN) {
 379                                         verdict = (unsigned int)(-v) - 1;
 380                                         break;
 381                                 }
 382                                 if (*stackptr <= origptr) {
 383                                         e = get_entry(table_base,
 384                                             private->underflow[hook]);
 385                                         pr_debug("Underflow (this is normal) "
 386                                                  "to %p\n", e);
 387                                 } else {
 388                                         e = jumpstack[--*stackptr];
 389                                         pr_debug("Pulled %p out from pos %u\n",
 390                                                  e, *stackptr);
 391                                         e = ipt_next_entry(e);
 392                                 }
 393                                 continue;
 394                         }
 395                         if (table_base + v != ipt_next_entry(e) &&
 396                             !(e->ip.flags & IPT_F_GOTO)) {
 397                                 if (*stackptr >= private->stacksize) {
 398                                         verdict = NF_DROP;
 399                                         break;
 400                                 }
 401                                 jumpstack[(*stackptr)++] = e;
 402                                 pr_debug("Pushed %p into pos %u\n",
 403                                          e, *stackptr - 1);
 404                         }
 405
 406                         e = get_entry(table_base, v);
 407                         continue;
 408                 }
 409
 410                 acpar.target   = t->u.kernel.target;
 411                 acpar.targinfo = t->data;
 412
 413                 verdict = t->u.kernel.target->target(skb, &acpar);
 414                 /* Target might have changed stuff. */
 415                 ip = ip_hdr(skb);
 416                 if (verdict == XT_CONTINUE)
 417                         e = ipt_next_entry(e);
 418                 else
 419                         /* Verdict */
 420                         break;
 421         } while (!acpar.hotdrop);
 422         pr_debug("Exiting %s; resetting sp from %u to %u\n",
 423                  __func__, *stackptr, origptr);
 424         *stackptr = origptr;
 425         xt_write_recseq_end(addend);
 426         local_bh_enable();
 427
 428 #ifdef DEBUG_ALLOW_ALL
 429         return NF_ACCEPT;
 430 #else
 431         if (acpar.hotdrop)
 432                 return NF_DROP;
 433         else return verdict;
 434 #endif
 435 }
 436
 437 /* Figures out from what hook each rule can be called: returns 0 if
 438    there are loops.  Puts hook bitmask in comefrom. */
 439 static int
 440 mark_source_chains(const struct xt_table_info *newinfo,
 441                    unsigned int valid_hooks, void *entry0)
 442 {
 443         unsigned int hook;
 444
 445         /* No recursion; use packet counter to save back ptrs (reset
 446            to 0 as we leave), and comefrom to save source hook bitmask */
 447         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 448                 unsigned int pos = newinfo->hook_entry[hook];
 449                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 450
 451                 if (!(valid_hooks & (1 << hook)))
 452                         continue;
 453
 454                 /* Set initial back pointer. */
 455                 e->counters.pcnt = pos;
 456
 457                 for (;;) {
 458                         const struct xt_standard_target *t
 459                                 = (void *)ipt_get_target_c(e);
 460                         int visited = e->comefrom & (1 << hook);
 461
 462                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 463                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 464                                        hook, pos, e->comefrom);
 465                                 return 0;
 466                         }
 467                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 468
 469                         /* Unconditional return/END. */
 470                         if ((e->target_offset == sizeof(struct ipt_entry) &&
 471                              (strcmp(t->target.u.user.name,
 472                                      XT_STANDARD_TARGET) == 0) &&
 473                              t->verdict < 0 && unconditional(&e->ip)) ||
 474                             visited) {
 475                                 unsigned int oldpos, size;
 476
 477                                 if ((strcmp(t->target.u.user.name,
 478                                             XT_STANDARD_TARGET) == 0) &&
 479                                     t->verdict < -NF_MAX_VERDICT - 1) {
 480                                         duprintf("mark_source_chains: bad "
 481                                                 "negative verdict (%i)\n",
 482                                                                 t->verdict);
 483                                         return 0;
 484                                 }
 485
 486                                 /* Return: backtrack through the last
 487                                    big jump. */
 488                                 do {
 489                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 490 #ifdef DEBUG_IP_FIREWALL_USER
 491                                         if (e->comefrom
 492                                             & (1 << NF_INET_NUMHOOKS)) {
 493                                                 duprintf("Back unset "
 494                                                          "on hook %u "
 495                                                          "rule %u\n",
 496                                                          hook, pos);
 497                                         }
 498 #endif
 499                                         oldpos = pos;
 500                                         pos = e->counters.pcnt;
 501                                         e->counters.pcnt = 0;
 502
 503                                         /* We're at the start. */
 504                                         if (pos == oldpos)
 505                                                 goto next;
 506
 507                                         e = (struct ipt_entry *)
 508                                                 (entry0 + pos);
 509                                 } while (oldpos == pos + e->next_offset);
 510
 511                                 /* Move along one */
 512                                 size = e->next_offset;
 513                                 e = (struct ipt_entry *)
 514                                         (entry0 + pos + size);
 515                                 e->counters.pcnt = pos;
 516                                 pos += size;
 517                         } else {
 518                                 int newpos = t->verdict;
 519
 520                                 if (strcmp(t->target.u.user.name,
 521                                            XT_STANDARD_TARGET) == 0 &&
 522                                     newpos >= 0) {
 523                                         if (newpos > newinfo->size -
 524                                                 sizeof(struct ipt_entry)) {
 525                                                 duprintf("mark_source_chains: "
 526                                                         "bad verdict (%i)\n",
 527                                                                 newpos);
 528                                                 return 0;
 529                                         }
 530                                         /* This a jump; chase it. */
 531                                         duprintf("Jump rule %u -> %u\n",
 532                                                  pos, newpos);
 533                                 } else {
 534                                         /* ... this is a fallthru */
 535                                         newpos = pos + e->next_offset;
 536                                 }
 537                                 e = (struct ipt_entry *)
 538                                         (entry0 + newpos);
 539                                 e->counters.pcnt = pos;
 540                                 pos = newpos;
 541                         }
 542                 }
 543                 next:
 544                 duprintf("Finished chain %u\n", hook);
 545         }
 546         return 1;
 547 }
 548
 549 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 550 {
 551         struct xt_mtdtor_param par;
 552
 553         par.net       = net;
 554         par.match     = m->u.kernel.match;
 555         par.matchinfo = m->data;
 556         par.family    = NFPROTO_IPV4;
 557         if (par.match->destroy != NULL)
 558                 par.match->destroy(&par);
 559         module_put(par.match->me);
 560 }
 561
 562 static int
 563 check_entry(const struct ipt_entry *e, const char *name)
 564 {
 565         const struct xt_entry_target *t;
 566
 567         if (!ip_checkentry(&e->ip)) {
 568                 duprintf("ip check failed %p %s.\n", e, name);
 569                 return -EINVAL;
 570         }
 571
 572         if (e->target_offset + sizeof(struct xt_entry_target) >
 573             e->next_offset)
 574                 return -EINVAL;
 575
 576         t = ipt_get_target_c(e);
 577         if (e->target_offset + t->u.target_size > e->next_offset)
 578                 return -EINVAL;
 579
 580         return 0;
 581 }
 582
 583 static int
 584 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 585 {
 586         const struct ipt_ip *ip = par->entryinfo;
 587         int ret;
 588
 589         par->match     = m->u.kernel.match;
 590         par->matchinfo = m->data;
 591
 592         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 593               ip->proto, ip->invflags & IPT_INV_PROTO);
 594         if (ret < 0) {
 595                 duprintf("check failed for `%s'.\n", par->match->name);
 596                 return ret;
 597         }
 598         return 0;
 599 }
 600
 601 static int
 602 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 603 {
 604         struct xt_match *match;
 605         int ret;
 606
 607         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 608                                       m->u.user.revision);
 609         if (IS_ERR(match)) {
 610                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 611                 return PTR_ERR(match);
 612         }
 613         m->u.kernel.match = match;
 614
 615         ret = check_match(m, par);
 616         if (ret)
 617                 goto err;
 618
 619         return 0;
 620 err:
 621         module_put(m->u.kernel.match->me);
 622         return ret;
 623 }
 624
 625 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 626 {
 627         struct xt_entry_target *t = ipt_get_target(e);
 628         struct xt_tgchk_param par = {
 629                 .net       = net,
 630                 .table     = name,
 631                 .entryinfo = e,
 632                 .target    = t->u.kernel.target,
 633                 .targinfo  = t->data,
 634                 .hook_mask = e->comefrom,
 635                 .family    = NFPROTO_IPV4,
 636         };
 637         int ret;
 638
 639         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 640               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 641         if (ret < 0) {
 642                 duprintf("check failed for `%s'.\n",
 643                          t->u.kernel.target->name);
 644                 return ret;
 645         }
 646         return 0;
 647 }
 648
 649 static int
 650 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 651                  unsigned int size)
 652 {
 653         struct xt_entry_target *t;
 654         struct xt_target *target;
 655         int ret;
 656         unsigned int j;
 657         struct xt_mtchk_param mtpar;
 658         struct xt_entry_match *ematch;
 659
 660         ret = check_entry(e, name);
 661         if (ret)
 662                 return ret;
 663
 664         j = 0;
 665         mtpar.net       = net;
 666         mtpar.table     = name;
 667         mtpar.entryinfo = &e->ip;
 668         mtpar.hook_mask = e->comefrom;
 669         mtpar.family    = NFPROTO_IPV4;
 670         xt_ematch_foreach(ematch, e) {
 671                 ret = find_check_match(ematch, &mtpar);
 672                 if (ret != 0)
 673                         goto cleanup_matches;
 674                 ++j;
 675         }
 676
 677         t = ipt_get_target(e);
 678         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 679                                         t->u.user.revision);
 680         if (IS_ERR(target)) {
 681                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 682                 ret = PTR_ERR(target);
 683                 goto cleanup_matches;
 684         }
 685         t->u.kernel.target = target;
 686
 687         ret = check_target(e, net, name);
 688         if (ret)
 689                 goto err;
 690         return 0;
 691  err:
 692         module_put(t->u.kernel.target->me);
 693  cleanup_matches:
 694         xt_ematch_foreach(ematch, e) {
 695                 if (j-- == 0)
 696                         break;
 697                 cleanup_match(ematch, net);
 698         }
 699         return ret;
 700 }
 701
 702 static bool check_underflow(const struct ipt_entry *e)
 703 {
 704         const struct xt_entry_target *t;
 705         unsigned int verdict;
 706
 707         if (!unconditional(&e->ip))
 708                 return false;
 709         t = ipt_get_target_c(e);
 710         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 711                 return false;
 712         verdict = ((struct xt_standard_target *)t)->verdict;
 713         verdict = -verdict - 1;
 714         return verdict == NF_DROP || verdict == NF_ACCEPT;
 715 }
 716
 717 static int
 718 check_entry_size_and_hooks(struct ipt_entry *e,
 719                            struct xt_table_info *newinfo,
 720                            const unsigned char *base,
 721                            const unsigned char *limit,
 722                            const unsigned int *hook_entries,
 723                            const unsigned int *underflows,
 724                            unsigned int valid_hooks)
 725 {
 726         unsigned int h;
 727
 728         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 729             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 730                 duprintf("Bad offset %p\n", e);
 731                 return -EINVAL;
 732         }
 733
 734         if (e->next_offset
 735             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 736                 duprintf("checking: element %p size %u\n",
 737                          e, e->next_offset);
 738                 return -EINVAL;
 739         }
 740
 741         /* Check hooks & underflows */
 742         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 743                 if (!(valid_hooks & (1 << h)))
 744                         continue;
 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                         if (!check_underflow(e)) {
 749                                 pr_err("Underflows must be unconditional and "
 750                                        "use the STANDARD target with "
 751                                        "ACCEPT/DROP\n");
 752                                 return -EINVAL;
 753                         }
 754                         newinfo->underflow[h] = underflows[h];
 755                 }
 756         }
 757
 758         /* Clear counters and comefrom */
 759         e->counters = ((struct xt_counters) { 0, 0 });
 760         e->comefrom = 0;
 761         return 0;
 762 }
 763
 764 static void
 765 cleanup_entry(struct ipt_entry *e, struct net *net)
 766 {
 767         struct xt_tgdtor_param par;
 768         struct xt_entry_target *t;
 769         struct xt_entry_match *ematch;
 770
 771         /* Cleanup all matches */
 772         xt_ematch_foreach(ematch, e)
 773                 cleanup_match(ematch, net);
 774         t = ipt_get_target(e);
 775
 776         par.net      = net;
 777         par.target   = t->u.kernel.target;
 778         par.targinfo = t->data;
 779         par.family   = NFPROTO_IPV4;
 780         if (par.target->destroy != NULL)
 781                 par.target->destroy(&par);
 782         module_put(par.target->me);
 783 }
 784
 785 /* Checks and translates the user-supplied table segment (held in
 786    newinfo) */
 787 static int
 788 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 789                 const struct ipt_replace *repl)
 790 {
 791         struct ipt_entry *iter;
 792         unsigned int i;
 793         int ret = 0;
 794
 795         newinfo->size = repl->size;
 796         newinfo->number = repl->num_entries;
 797
 798         /* Init all hooks to impossible value. */
 799         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 800                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 801                 newinfo->underflow[i] = 0xFFFFFFFF;
 802         }
 803
 804         duprintf("translate_table: size %u\n", newinfo->size);
 805         i = 0;
 806         /* Walk through entries, checking offsets. */
 807         xt_entry_foreach(iter, entry0, newinfo->size) {
 808                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 809                                                  entry0 + repl->size,
 810                                                  repl->hook_entry,
 811                                                  repl->underflow,
 812                                                  repl->valid_hooks);
 813                 if (ret != 0)
 814                         return ret;
 815                 ++i;
 816                 if (strcmp(ipt_get_target(iter)->u.user.name,
 817                     XT_ERROR_TARGET) == 0)
 818                         ++newinfo->stacksize;
 819         }
 820
 821         if (i != repl->num_entries) {
 822                 duprintf("translate_table: %u not %u entries\n",
 823                          i, repl->num_entries);
 824                 return -EINVAL;
 825         }
 826
 827         /* Check hooks all assigned */
 828         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 829                 /* Only hooks which are valid */
 830                 if (!(repl->valid_hooks & (1 << i)))
 831                         continue;
 832                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 833                         duprintf("Invalid hook entry %u %u\n",
 834                                  i, repl->hook_entry[i]);
 835                         return -EINVAL;
 836                 }
 837                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 838                         duprintf("Invalid underflow %u %u\n",
 839                                  i, repl->underflow[i]);
 840                         return -EINVAL;
 841                 }
 842         }
 843
 844         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 845                 return -ELOOP;
 846
 847         /* Finally, each sanity check must pass */
 848         i = 0;
 849         xt_entry_foreach(iter, entry0, newinfo->size) {
 850                 ret = find_check_entry(iter, net, repl->name, repl->size);
 851                 if (ret != 0)
 852                         break;
 853                 ++i;
 854         }
 855
 856         if (ret != 0) {
 857                 xt_entry_foreach(iter, entry0, newinfo->size) {
 858                         if (i-- == 0)
 859                                 break;
 860                         cleanup_entry(iter, net);
 861                 }
 862                 return ret;
 863         }
 864
 865         /* And one copy for every other CPU */
 866         for_each_possible_cpu(i) {
 867                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
 868                         memcpy(newinfo->entries[i], entry0, newinfo->size);
 869         }
 870
 871         return ret;
 872 }
 873
 874 static void
 875 get_counters(const struct xt_table_info *t,
 876              struct xt_counters counters[])
 877 {
 878         struct ipt_entry *iter;
 879         unsigned int cpu;
 880         unsigned int i;
 881
 882         for_each_possible_cpu(cpu) {
 883                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 884
 885                 i = 0;
 886                 xt_entry_foreach(iter, t->entries[cpu], t->size) {
 887                         u64 bcnt, pcnt;
 888                         unsigned int start;
 889
 890                         do {
 891                                 start = read_seqcount_begin(s);
 892                                 bcnt = iter->counters.bcnt;
 893                                 pcnt = iter->counters.pcnt;
 894                         } while (read_seqcount_retry(s, start));
 895
 896                         ADD_COUNTER(counters[i], bcnt, pcnt);
 897                         ++i; /* macro does multi eval of i */
 898                 }
 899         }
 900 }
 901
 902 static struct xt_counters *alloc_counters(const struct xt_table *table)
 903 {
 904         unsigned int countersize;
 905         struct xt_counters *counters;
 906         const struct xt_table_info *private = table->private;
 907
 908         /* We need atomic snapshot of counters: rest doesn't change
 909            (other than comefrom, which userspace doesn't care
 910            about). */
 911         countersize = sizeof(struct xt_counters) * private->number;
 912         counters = vzalloc(countersize);
 913
 914         if (counters == NULL)
 915                 return ERR_PTR(-ENOMEM);
 916
 917         get_counters(private, counters);
 918
 919         return counters;
 920 }
 921
 922 static int
 923 copy_entries_to_user(unsigned int total_size,
 924                      const struct xt_table *table,
 925                      void __user *userptr)
 926 {
 927         unsigned int off, num;
 928         const struct ipt_entry *e;
 929         struct xt_counters *counters;
 930         const struct xt_table_info *private = table->private;
 931         int ret = 0;
 932         const void *loc_cpu_entry;
 933
 934         counters = alloc_counters(table);
 935         if (IS_ERR(counters))
 936                 return PTR_ERR(counters);
 937
 938         /* choose the copy that is on our node/cpu, ...
 939          * This choice is lazy (because current thread is
 940          * allowed to migrate to another cpu)
 941          */
 942         loc_cpu_entry = private->entries[raw_smp_processor_id()];
 943         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 944                 ret = -EFAULT;
 945                 goto free_counters;
 946         }
 947
 948         /* FIXME: use iterator macros --RR */
 949         /* ... then go back and fix counters and names */
 950         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 951                 unsigned int i;
 952                 const struct xt_entry_match *m;
 953                 const struct xt_entry_target *t;
 954
 955                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 956                 if (copy_to_user(userptr + off
 957                                  + offsetof(struct ipt_entry, counters),
 958                                  &counters[num],
 959                                  sizeof(counters[num])) != 0) {
 960                         ret = -EFAULT;
 961                         goto free_counters;
 962                 }
 963
 964                 for (i = sizeof(struct ipt_entry);
 965                      i < e->target_offset;
 966                      i += m->u.match_size) {
 967                         m = (void *)e + i;
 968
 969                         if (copy_to_user(userptr + off + i
 970                                          + offsetof(struct xt_entry_match,
 971                                                     u.user.name),
 972                                          m->u.kernel.match->name,
 973                                          strlen(m->u.kernel.match->name)+1)
 974                             != 0) {
 975                                 ret = -EFAULT;
 976                                 goto free_counters;
 977                         }
 978                 }
 979
 980                 t = ipt_get_target_c(e);
 981                 if (copy_to_user(userptr + off + e->target_offset
 982                                  + offsetof(struct xt_entry_target,
 983                                             u.user.name),
 984                                  t->u.kernel.target->name,
 985                                  strlen(t->u.kernel.target->name)+1) != 0) {
 986                         ret = -EFAULT;
 987                         goto free_counters;
 988                 }
 989         }
 990
 991  free_counters:
 992         vfree(counters);
 993         return ret;
 994 }
 995
 996 #ifdef CONFIG_COMPAT
 997 static void compat_standard_from_user(void *dst, const void *src)
 998 {
 999         int v = *(compat_int_t *)src;
1000
1001         if (v > 0)
1002                 v += xt_compat_calc_jump(AF_INET, v);
1003         memcpy(dst, &v, sizeof(v));
1004 }
1005
1006 static int compat_standard_to_user(void __user *dst, const void *src)
1007 {
1008         compat_int_t cv = *(int *)src;
1009
1010         if (cv > 0)
1011                 cv -= xt_compat_calc_jump(AF_INET, cv);
1012         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1013 }
1014
1015 static int compat_calc_entry(const struct ipt_entry *e,
1016                              const struct xt_table_info *info,
1017                              const void *base, struct xt_table_info *newinfo)
1018 {
1019         const struct xt_entry_match *ematch;
1020         const struct xt_entry_target *t;
1021         unsigned int entry_offset;
1022         int off, i, ret;
1023
1024         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1025         entry_offset = (void *)e - base;
1026         xt_ematch_foreach(ematch, e)
1027                 off += xt_compat_match_offset(ematch->u.kernel.match);
1028         t = ipt_get_target_c(e);
1029         off += xt_compat_target_offset(t->u.kernel.target);
1030         newinfo->size -= off;
1031         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1032         if (ret)
1033                 return ret;
1034
1035         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1036                 if (info->hook_entry[i] &&
1037                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1038                         newinfo->hook_entry[i] -= off;
1039                 if (info->underflow[i] &&
1040                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1041                         newinfo->underflow[i] -= off;
1042         }
1043         return 0;
1044 }
1045
1046 static int compat_table_info(const struct xt_table_info *info,
1047                              struct xt_table_info *newinfo)
1048 {
1049         struct ipt_entry *iter;
1050         void *loc_cpu_entry;
1051         int ret;
1052
1053         if (!newinfo || !info)
1054                 return -EINVAL;
1055
1056         /* we dont care about newinfo->entries[] */
1057         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1058         newinfo->initial_entries = 0;
1059         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1060         xt_compat_init_offsets(AF_INET, info->number);
1061         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1062                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1063                 if (ret != 0)
1064                         return ret;
1065         }
1066         return 0;
1067 }
1068 #endif
1069
1070 static int get_info(struct net *net, void __user *user,
1071                     const int *len, int compat)
1072 {
1073         char name[XT_TABLE_MAXNAMELEN];
1074         struct xt_table *t;
1075         int ret;
1076
1077         if (*len != sizeof(struct ipt_getinfo)) {
1078                 duprintf("length %u != %zu\n", *len,
1079                          sizeof(struct ipt_getinfo));
1080                 return -EINVAL;
1081         }
1082
1083         if (copy_from_user(name, user, sizeof(name)) != 0)
1084                 return -EFAULT;
1085
1086         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1087 #ifdef CONFIG_COMPAT
1088         if (compat)
1089                 xt_compat_lock(AF_INET);
1090 #endif
1091         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1092                                     "iptable_%s", name);
1093         if (t && !IS_ERR(t)) {
1094                 struct ipt_getinfo info;
1095                 const struct xt_table_info *private = t->private;
1096 #ifdef CONFIG_COMPAT
1097                 struct xt_table_info tmp;
1098
1099                 if (compat) {
1100                         ret = compat_table_info(private, &tmp);
1101                         xt_compat_flush_offsets(AF_INET);
1102                         private = &tmp;
1103                 }
1104 #endif
1105                 memset(&info, 0, sizeof(info));
1106                 info.valid_hooks = t->valid_hooks;
1107                 memcpy(info.hook_entry, private->hook_entry,
1108                        sizeof(info.hook_entry));
1109                 memcpy(info.underflow, private->underflow,
1110                        sizeof(info.underflow));
1111                 info.num_entries = private->number;
1112                 info.size = private->size;
1113                 strcpy(info.name, name);
1114
1115                 if (copy_to_user(user, &info, *len) != 0)
1116                         ret = -EFAULT;
1117                 else
1118                         ret = 0;
1119
1120                 xt_table_unlock(t);
1121                 module_put(t->me);
1122         } else
1123                 ret = t ? PTR_ERR(t) : -ENOENT;
1124 #ifdef CONFIG_COMPAT
1125         if (compat)
1126                 xt_compat_unlock(AF_INET);
1127 #endif
1128         return ret;
1129 }
1130
1131 static int
1132 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1133             const int *len)
1134 {
1135         int ret;
1136         struct ipt_get_entries get;
1137         struct xt_table *t;
1138
1139         if (*len < sizeof(get)) {
1140                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1141                 return -EINVAL;
1142         }
1143         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1144                 return -EFAULT;
1145         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1146                 duprintf("get_entries: %u != %zu\n",
1147                          *len, sizeof(get) + get.size);
1148                 return -EINVAL;
1149         }
1150
1151         t = xt_find_table_lock(net, AF_INET, get.name);
1152         if (t && !IS_ERR(t)) {
1153                 const struct xt_table_info *private = t->private;
1154                 duprintf("t->private->number = %u\n", private->number);
1155                 if (get.size == private->size)
1156                         ret = copy_entries_to_user(private->size,
1157                                                    t, uptr->entrytable);
1158                 else {
1159                         duprintf("get_entries: I've got %u not %u!\n",
1160                                  private->size, get.size);
1161                         ret = -EAGAIN;
1162                 }
1163                 module_put(t->me);
1164                 xt_table_unlock(t);
1165         } else
1166                 ret = t ? PTR_ERR(t) : -ENOENT;
1167
1168         return ret;
1169 }
1170
1171 static int
1172 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1173              struct xt_table_info *newinfo, unsigned int num_counters,
1174              void __user *counters_ptr)
1175 {
1176         int ret;
1177         struct xt_table *t;
1178         struct xt_table_info *oldinfo;
1179         struct xt_counters *counters;
1180         void *loc_cpu_old_entry;
1181         struct ipt_entry *iter;
1182
1183         ret = 0;
1184         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1185         if (!counters) {
1186                 ret = -ENOMEM;
1187                 goto out;
1188         }
1189
1190         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1191                                     "iptable_%s", name);
1192         if (!t || IS_ERR(t)) {
1193                 ret = t ? PTR_ERR(t) : -ENOENT;
1194                 goto free_newinfo_counters_untrans;
1195         }
1196
1197         /* You lied! */
1198         if (valid_hooks != t->valid_hooks) {
1199                 duprintf("Valid hook crap: %08X vs %08X\n",
1200                          valid_hooks, t->valid_hooks);
1201                 ret = -EINVAL;
1202                 goto put_module;
1203         }
1204
1205         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1206         if (!oldinfo)
1207                 goto put_module;
1208
1209         /* Update module usage count based on number of rules */
1210         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1211                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1212         if ((oldinfo->number > oldinfo->initial_entries) ||
1213             (newinfo->number <= oldinfo->initial_entries))
1214                 module_put(t->me);
1215         if ((oldinfo->number > oldinfo->initial_entries) &&
1216             (newinfo->number <= oldinfo->initial_entries))
1217                 module_put(t->me);
1218
1219         /* Get the old counters, and synchronize with replace */
1220         get_counters(oldinfo, counters);
1221
1222         /* Decrease module usage counts and free resource */
1223         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1224         xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1225                 cleanup_entry(iter, net);
1226
1227         xt_free_table_info(oldinfo);
1228         if (copy_to_user(counters_ptr, counters,
1229                          sizeof(struct xt_counters) * num_counters) != 0)
1230                 ret = -EFAULT;
1231         vfree(counters);
1232         xt_table_unlock(t);
1233         return ret;
1234
1235  put_module:
1236         module_put(t->me);
1237         xt_table_unlock(t);
1238  free_newinfo_counters_untrans:
1239         vfree(counters);
1240  out:
1241         return ret;
1242 }
1243
1244 static int
1245 do_replace(struct net *net, const void __user *user, unsigned int len)
1246 {
1247         int ret;
1248         struct ipt_replace tmp;
1249         struct xt_table_info *newinfo;
1250         void *loc_cpu_entry;
1251         struct ipt_entry *iter;
1252
1253         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1254                 return -EFAULT;
1255
1256         /* overflow check */
1257         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1258                 return -ENOMEM;
1259         tmp.name[sizeof(tmp.name)-1] = 0;
1260
1261         newinfo = xt_alloc_table_info(tmp.size);
1262         if (!newinfo)
1263                 return -ENOMEM;
1264
1265         /* choose the copy that is on our node/cpu */
1266         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1267         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1268                            tmp.size) != 0) {
1269                 ret = -EFAULT;
1270                 goto free_newinfo;
1271         }
1272
1273         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1274         if (ret != 0)
1275                 goto free_newinfo;
1276
1277         duprintf("Translated table\n");
1278
1279         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1280                            tmp.num_counters, tmp.counters);
1281         if (ret)
1282                 goto free_newinfo_untrans;
1283         return 0;
1284
1285  free_newinfo_untrans:
1286         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1287                 cleanup_entry(iter, net);
1288  free_newinfo:
1289         xt_free_table_info(newinfo);
1290         return ret;
1291 }
1292
1293 static int
1294 do_add_counters(struct net *net, const void __user *user,
1295                 unsigned int len, int compat)
1296 {
1297         unsigned int i, curcpu;
1298         struct xt_counters_info tmp;
1299         struct xt_counters *paddc;
1300         unsigned int num_counters;
1301         const char *name;
1302         int size;
1303         void *ptmp;
1304         struct xt_table *t;
1305         const struct xt_table_info *private;
1306         int ret = 0;
1307         void *loc_cpu_entry;
1308         struct ipt_entry *iter;
1309         unsigned int addend;
1310 #ifdef CONFIG_COMPAT
1311         struct compat_xt_counters_info compat_tmp;
1312
1313         if (compat) {
1314                 ptmp = &compat_tmp;
1315                 size = sizeof(struct compat_xt_counters_info);
1316         } else
1317 #endif
1318         {
1319                 ptmp = &tmp;
1320                 size = sizeof(struct xt_counters_info);
1321         }
1322
1323         if (copy_from_user(ptmp, user, size) != 0)
1324                 return -EFAULT;
1325
1326 #ifdef CONFIG_COMPAT
1327         if (compat) {
1328                 num_counters = compat_tmp.num_counters;
1329                 name = compat_tmp.name;
1330         } else
1331 #endif
1332         {
1333                 num_counters = tmp.num_counters;
1334                 name = tmp.name;
1335         }
1336
1337         if (len != size + num_counters * sizeof(struct xt_counters))
1338                 return -EINVAL;
1339
1340         paddc = vmalloc(len - size);
1341         if (!paddc)
1342                 return -ENOMEM;
1343
1344         if (copy_from_user(paddc, user + size, len - size) != 0) {
1345                 ret = -EFAULT;
1346                 goto free;
1347         }
1348
1349         t = xt_find_table_lock(net, AF_INET, name);
1350         if (!t || IS_ERR(t)) {
1351                 ret = t ? PTR_ERR(t) : -ENOENT;
1352                 goto free;
1353         }
1354
1355         local_bh_disable();
1356         private = t->private;
1357         if (private->number != num_counters) {
1358                 ret = -EINVAL;
1359                 goto unlock_up_free;
1360         }
1361
1362         i = 0;
1363         /* Choose the copy that is on our node */
1364         curcpu = smp_processor_id();
1365         loc_cpu_entry = private->entries[curcpu];
1366         addend = xt_write_recseq_begin();
1367         xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1368                 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1369                 ++i;
1370         }
1371         xt_write_recseq_end(addend);
1372  unlock_up_free:
1373         local_bh_enable();
1374         xt_table_unlock(t);
1375         module_put(t->me);
1376  free:
1377         vfree(paddc);
1378
1379         return ret;
1380 }
1381
1382 #ifdef CONFIG_COMPAT
1383 struct compat_ipt_replace {
1384         char                    name[XT_TABLE_MAXNAMELEN];
1385         u32                     valid_hooks;
1386         u32                     num_entries;
1387         u32                     size;
1388         u32                     hook_entry[NF_INET_NUMHOOKS];
1389         u32                     underflow[NF_INET_NUMHOOKS];
1390         u32                     num_counters;
1391         compat_uptr_t           counters;       /* struct xt_counters * */
1392         struct compat_ipt_entry entries[0];
1393 };
1394
1395 static int
1396 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1397                           unsigned int *size, struct xt_counters *counters,
1398                           unsigned int i)
1399 {
1400         struct xt_entry_target *t;
1401         struct compat_ipt_entry __user *ce;
1402         u_int16_t target_offset, next_offset;
1403         compat_uint_t origsize;
1404         const struct xt_entry_match *ematch;
1405         int ret = 0;
1406
1407         origsize = *size;
1408         ce = (struct compat_ipt_entry __user *)*dstptr;
1409         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1410             copy_to_user(&ce->counters, &counters[i],
1411             sizeof(counters[i])) != 0)
1412                 return -EFAULT;
1413
1414         *dstptr += sizeof(struct compat_ipt_entry);
1415         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1416
1417         xt_ematch_foreach(ematch, e) {
1418                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1419                 if (ret != 0)
1420                         return ret;
1421         }
1422         target_offset = e->target_offset - (origsize - *size);
1423         t = ipt_get_target(e);
1424         ret = xt_compat_target_to_user(t, dstptr, size);
1425         if (ret)
1426                 return ret;
1427         next_offset = e->next_offset - (origsize - *size);
1428         if (put_user(target_offset, &ce->target_offset) != 0 ||
1429             put_user(next_offset, &ce->next_offset) != 0)
1430                 return -EFAULT;
1431         return 0;
1432 }
1433
1434 static int
1435 compat_find_calc_match(struct xt_entry_match *m,
1436                        const char *name,
1437                        const struct ipt_ip *ip,
1438                        unsigned int hookmask,
1439                        int *size)
1440 {
1441         struct xt_match *match;
1442
1443         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1444                                       m->u.user.revision);
1445         if (IS_ERR(match)) {
1446                 duprintf("compat_check_calc_match: `%s' not found\n",
1447                          m->u.user.name);
1448                 return PTR_ERR(match);
1449         }
1450         m->u.kernel.match = match;
1451         *size += xt_compat_match_offset(match);
1452         return 0;
1453 }
1454
1455 static void compat_release_entry(struct compat_ipt_entry *e)
1456 {
1457         struct xt_entry_target *t;
1458         struct xt_entry_match *ematch;
1459
1460         /* Cleanup all matches */
1461         xt_ematch_foreach(ematch, e)
1462                 module_put(ematch->u.kernel.match->me);
1463         t = compat_ipt_get_target(e);
1464         module_put(t->u.kernel.target->me);
1465 }
1466
1467 static int
1468 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1469                                   struct xt_table_info *newinfo,
1470                                   unsigned int *size,
1471                                   const unsigned char *base,
1472                                   const unsigned char *limit,
1473                                   const unsigned int *hook_entries,
1474                                   const unsigned int *underflows,
1475                                   const char *name)
1476 {
1477         struct xt_entry_match *ematch;
1478         struct xt_entry_target *t;
1479         struct xt_target *target;
1480         unsigned int entry_offset;
1481         unsigned int j;
1482         int ret, off, h;
1483
1484         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1485         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1486             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1487                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1488                 return -EINVAL;
1489         }
1490
1491         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1492                              sizeof(struct compat_xt_entry_target)) {
1493                 duprintf("checking: element %p size %u\n",
1494                          e, e->next_offset);
1495                 return -EINVAL;
1496         }
1497
1498         /* For purposes of check_entry casting the compat entry is fine */
1499         ret = check_entry((struct ipt_entry *)e, name);
1500         if (ret)
1501                 return ret;
1502
1503         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1504         entry_offset = (void *)e - (void *)base;
1505         j = 0;
1506         xt_ematch_foreach(ematch, e) {
1507                 ret = compat_find_calc_match(ematch, name,
1508                                              &e->ip, e->comefrom, &off);
1509                 if (ret != 0)
1510                         goto release_matches;
1511                 ++j;
1512         }
1513
1514         t = compat_ipt_get_target(e);
1515         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1516                                         t->u.user.revision);
1517         if (IS_ERR(target)) {
1518                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1519                          t->u.user.name);
1520                 ret = PTR_ERR(target);
1521                 goto release_matches;
1522         }
1523         t->u.kernel.target = target;
1524
1525         off += xt_compat_target_offset(target);
1526         *size += off;
1527         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1528         if (ret)
1529                 goto out;
1530
1531         /* Check hooks & underflows */
1532         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1533                 if ((unsigned char *)e - base == hook_entries[h])
1534                         newinfo->hook_entry[h] = hook_entries[h];
1535                 if ((unsigned char *)e - base == underflows[h])
1536                         newinfo->underflow[h] = underflows[h];
1537         }
1538
1539         /* Clear counters and comefrom */
1540         memset(&e->counters, 0, sizeof(e->counters));
1541         e->comefrom = 0;
1542         return 0;
1543
1544 out:
1545         module_put(t->u.kernel.target->me);
1546 release_matches:
1547         xt_ematch_foreach(ematch, e) {
1548                 if (j-- == 0)
1549                         break;
1550                 module_put(ematch->u.kernel.match->me);
1551         }
1552         return ret;
1553 }
1554
1555 static int
1556 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1557                             unsigned int *size, const char *name,
1558                             struct xt_table_info *newinfo, unsigned char *base)
1559 {
1560         struct xt_entry_target *t;
1561         struct xt_target *target;
1562         struct ipt_entry *de;
1563         unsigned int origsize;
1564         int ret, h;
1565         struct xt_entry_match *ematch;
1566
1567         ret = 0;
1568         origsize = *size;
1569         de = (struct ipt_entry *)*dstptr;
1570         memcpy(de, e, sizeof(struct ipt_entry));
1571         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1572
1573         *dstptr += sizeof(struct ipt_entry);
1574         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1575
1576         xt_ematch_foreach(ematch, e) {
1577                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1578                 if (ret != 0)
1579                         return ret;
1580         }
1581         de->target_offset = e->target_offset - (origsize - *size);
1582         t = compat_ipt_get_target(e);
1583         target = t->u.kernel.target;
1584         xt_compat_target_from_user(t, dstptr, size);
1585
1586         de->next_offset = e->next_offset - (origsize - *size);
1587         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1588                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1589                         newinfo->hook_entry[h] -= origsize - *size;
1590                 if ((unsigned char *)de - base < newinfo->underflow[h])
1591                         newinfo->underflow[h] -= origsize - *size;
1592         }
1593         return ret;
1594 }
1595
1596 static int
1597 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1598 {
1599         struct xt_entry_match *ematch;
1600         struct xt_mtchk_param mtpar;
1601         unsigned int j;
1602         int ret = 0;
1603
1604         j = 0;
1605         mtpar.net       = net;
1606         mtpar.table     = name;
1607         mtpar.entryinfo = &e->ip;
1608         mtpar.hook_mask = e->comefrom;
1609         mtpar.family    = NFPROTO_IPV4;
1610         xt_ematch_foreach(ematch, e) {
1611                 ret = check_match(ematch, &mtpar);
1612                 if (ret != 0)
1613                         goto cleanup_matches;
1614                 ++j;
1615         }
1616
1617         ret = check_target(e, net, name);
1618         if (ret)
1619                 goto cleanup_matches;
1620         return 0;
1621
1622  cleanup_matches:
1623         xt_ematch_foreach(ematch, e) {
1624                 if (j-- == 0)
1625                         break;
1626                 cleanup_match(ematch, net);
1627         }
1628         return ret;
1629 }
1630
1631 static int
1632 translate_compat_table(struct net *net,
1633                        const char *name,
1634                        unsigned int valid_hooks,
1635                        struct xt_table_info **pinfo,
1636                        void **pentry0,
1637                        unsigned int total_size,
1638                        unsigned int number,
1639                        unsigned int *hook_entries,
1640                        unsigned int *underflows)
1641 {
1642         unsigned int i, j;
1643         struct xt_table_info *newinfo, *info;
1644         void *pos, *entry0, *entry1;
1645         struct compat_ipt_entry *iter0;
1646         struct ipt_entry *iter1;
1647         unsigned int size;
1648         int ret;
1649
1650         info = *pinfo;
1651         entry0 = *pentry0;
1652         size = total_size;
1653         info->number = number;
1654
1655         /* Init all hooks to impossible value. */
1656         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1657                 info->hook_entry[i] = 0xFFFFFFFF;
1658                 info->underflow[i] = 0xFFFFFFFF;
1659         }
1660
1661         duprintf("translate_compat_table: size %u\n", info->size);
1662         j = 0;
1663         xt_compat_lock(AF_INET);
1664         xt_compat_init_offsets(AF_INET, number);
1665         /* Walk through entries, checking offsets. */
1666         xt_entry_foreach(iter0, entry0, total_size) {
1667                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1668                                                         entry0,
1669                                                         entry0 + total_size,
1670                                                         hook_entries,
1671                                                         underflows,
1672                                                         name);
1673                 if (ret != 0)
1674                         goto out_unlock;
1675                 ++j;
1676         }
1677
1678         ret = -EINVAL;
1679         if (j != number) {
1680                 duprintf("translate_compat_table: %u not %u entries\n",
1681                          j, number);
1682                 goto out_unlock;
1683         }
1684
1685         /* Check hooks all assigned */
1686         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1687                 /* Only hooks which are valid */
1688                 if (!(valid_hooks & (1 << i)))
1689                         continue;
1690                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1691                         duprintf("Invalid hook entry %u %u\n",
1692                                  i, hook_entries[i]);
1693                         goto out_unlock;
1694                 }
1695                 if (info->underflow[i] == 0xFFFFFFFF) {
1696                         duprintf("Invalid underflow %u %u\n",
1697                                  i, underflows[i]);
1698                         goto out_unlock;
1699                 }
1700         }
1701
1702         ret = -ENOMEM;
1703         newinfo = xt_alloc_table_info(size);
1704         if (!newinfo)
1705                 goto out_unlock;
1706
1707         newinfo->number = number;
1708         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1709                 newinfo->hook_entry[i] = info->hook_entry[i];
1710                 newinfo->underflow[i] = info->underflow[i];
1711         }
1712         entry1 = newinfo->entries[raw_smp_processor_id()];
1713         pos = entry1;
1714         size = total_size;
1715         xt_entry_foreach(iter0, entry0, total_size) {
1716                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1717                                                   name, newinfo, entry1);
1718                 if (ret != 0)
1719                         break;
1720         }
1721         xt_compat_flush_offsets(AF_INET);
1722         xt_compat_unlock(AF_INET);
1723         if (ret)
1724                 goto free_newinfo;
1725
1726         ret = -ELOOP;
1727         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1728                 goto free_newinfo;
1729
1730         i = 0;
1731         xt_entry_foreach(iter1, entry1, newinfo->size) {
1732                 ret = compat_check_entry(iter1, net, name);
1733                 if (ret != 0)
1734                         break;
1735                 ++i;
1736                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1737                     XT_ERROR_TARGET) == 0)
1738                         ++newinfo->stacksize;
1739         }
1740         if (ret) {
1741                 /*
1742                  * The first i matches need cleanup_entry (calls ->destroy)
1743                  * because they had called ->check already. The other j-i
1744                  * entries need only release.
1745                  */
1746                 int skip = i;
1747                 j -= i;
1748                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1749                         if (skip-- > 0)
1750                                 continue;
1751                         if (j-- == 0)
1752                                 break;
1753                         compat_release_entry(iter0);
1754                 }
1755                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1756                         if (i-- == 0)
1757                                 break;
1758                         cleanup_entry(iter1, net);
1759                 }
1760                 xt_free_table_info(newinfo);
1761                 return ret;
1762         }
1763
1764         /* And one copy for every other CPU */
1765         for_each_possible_cpu(i)
1766                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1767                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1768
1769         *pinfo = newinfo;
1770         *pentry0 = entry1;
1771         xt_free_table_info(info);
1772         return 0;
1773
1774 free_newinfo:
1775         xt_free_table_info(newinfo);
1776 out:
1777         xt_entry_foreach(iter0, entry0, total_size) {
1778                 if (j-- == 0)
1779                         break;
1780                 compat_release_entry(iter0);
1781         }
1782         return ret;
1783 out_unlock:
1784         xt_compat_flush_offsets(AF_INET);
1785         xt_compat_unlock(AF_INET);
1786         goto out;
1787 }
1788
1789 static int
1790 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1791 {
1792         int ret;
1793         struct compat_ipt_replace tmp;
1794         struct xt_table_info *newinfo;
1795         void *loc_cpu_entry;
1796         struct ipt_entry *iter;
1797
1798         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1799                 return -EFAULT;
1800
1801         /* overflow check */
1802         if (tmp.size >= INT_MAX / num_possible_cpus())
1803                 return -ENOMEM;
1804         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1805                 return -ENOMEM;
1806         tmp.name[sizeof(tmp.name)-1] = 0;
1807
1808         newinfo = xt_alloc_table_info(tmp.size);
1809         if (!newinfo)
1810                 return -ENOMEM;
1811
1812         /* choose the copy that is on our node/cpu */
1813         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1814         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1815                            tmp.size) != 0) {
1816                 ret = -EFAULT;
1817                 goto free_newinfo;
1818         }
1819
1820         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1821                                      &newinfo, &loc_cpu_entry, tmp.size,
1822                                      tmp.num_entries, tmp.hook_entry,
1823                                      tmp.underflow);
1824         if (ret != 0)
1825                 goto free_newinfo;
1826
1827         duprintf("compat_do_replace: Translated table\n");
1828
1829         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1830                            tmp.num_counters, compat_ptr(tmp.counters));
1831         if (ret)
1832                 goto free_newinfo_untrans;
1833         return 0;
1834
1835  free_newinfo_untrans:
1836         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1837                 cleanup_entry(iter, net);
1838  free_newinfo:
1839         xt_free_table_info(newinfo);
1840         return ret;
1841 }
1842
1843 static int
1844 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1845                       unsigned int len)
1846 {
1847         int ret;
1848
1849         if (!capable(CAP_NET_ADMIN))
1850                 return -EPERM;
1851
1852         switch (cmd) {
1853         case IPT_SO_SET_REPLACE:
1854                 ret = compat_do_replace(sock_net(sk), user, len);
1855                 break;
1856
1857         case IPT_SO_SET_ADD_COUNTERS:
1858                 ret = do_add_counters(sock_net(sk), user, len, 1);
1859                 break;
1860
1861         default:
1862                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1863                 ret = -EINVAL;
1864         }
1865
1866         return ret;
1867 }
1868
1869 struct compat_ipt_get_entries {
1870         char name[XT_TABLE_MAXNAMELEN];
1871         compat_uint_t size;
1872         struct compat_ipt_entry entrytable[0];
1873 };
1874
1875 static int
1876 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1877                             void __user *userptr)
1878 {
1879         struct xt_counters *counters;
1880         const struct xt_table_info *private = table->private;
1881         void __user *pos;
1882         unsigned int size;
1883         int ret = 0;
1884         const void *loc_cpu_entry;
1885         unsigned int i = 0;
1886         struct ipt_entry *iter;
1887
1888         counters = alloc_counters(table);
1889         if (IS_ERR(counters))
1890                 return PTR_ERR(counters);
1891
1892         /* choose the copy that is on our node/cpu, ...
1893          * This choice is lazy (because current thread is
1894          * allowed to migrate to another cpu)
1895          */
1896         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1897         pos = userptr;
1898         size = total_size;
1899         xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1900                 ret = compat_copy_entry_to_user(iter, &pos,
1901                                                 &size, counters, i++);
1902                 if (ret != 0)
1903                         break;
1904         }
1905
1906         vfree(counters);
1907         return ret;
1908 }
1909
1910 static int
1911 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1912                    int *len)
1913 {
1914         int ret;
1915         struct compat_ipt_get_entries get;
1916         struct xt_table *t;
1917
1918         if (*len < sizeof(get)) {
1919                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1920                 return -EINVAL;
1921         }
1922
1923         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1924                 return -EFAULT;
1925
1926         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1927                 duprintf("compat_get_entries: %u != %zu\n",
1928                          *len, sizeof(get) + get.size);
1929                 return -EINVAL;
1930         }
1931
1932         xt_compat_lock(AF_INET);
1933         t = xt_find_table_lock(net, AF_INET, get.name);
1934         if (t && !IS_ERR(t)) {
1935                 const struct xt_table_info *private = t->private;
1936                 struct xt_table_info info;
1937                 duprintf("t->private->number = %u\n", private->number);
1938                 ret = compat_table_info(private, &info);
1939                 if (!ret && get.size == info.size) {
1940                         ret = compat_copy_entries_to_user(private->size,
1941                                                           t, uptr->entrytable);
1942                 } else if (!ret) {
1943                         duprintf("compat_get_entries: I've got %u not %u!\n",
1944                                  private->size, get.size);
1945                         ret = -EAGAIN;
1946                 }
1947                 xt_compat_flush_offsets(AF_INET);
1948                 module_put(t->me);
1949                 xt_table_unlock(t);
1950         } else
1951                 ret = t ? PTR_ERR(t) : -ENOENT;
1952
1953         xt_compat_unlock(AF_INET);
1954         return ret;
1955 }
1956
1957 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1958
1959 static int
1960 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1961 {
1962         int ret;
1963
1964         if (!capable(CAP_NET_ADMIN))
1965                 return -EPERM;
1966
1967         switch (cmd) {
1968         case IPT_SO_GET_INFO:
1969                 ret = get_info(sock_net(sk), user, len, 1);
1970                 break;
1971         case IPT_SO_GET_ENTRIES:
1972                 ret = compat_get_entries(sock_net(sk), user, len);
1973                 break;
1974         default:
1975                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1976         }
1977         return ret;
1978 }
1979 #endif
1980
1981 static int
1982 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1983 {
1984         int ret;
1985
1986         if (!capable(CAP_NET_ADMIN))
1987                 return -EPERM;
1988
1989         switch (cmd) {
1990         case IPT_SO_SET_REPLACE:
1991                 ret = do_replace(sock_net(sk), user, len);
1992                 break;
1993
1994         case IPT_SO_SET_ADD_COUNTERS:
1995                 ret = do_add_counters(sock_net(sk), user, len, 0);
1996                 break;
1997
1998         default:
1999                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2000                 ret = -EINVAL;
2001         }
2002
2003         return ret;
2004 }
2005
2006 static int
2007 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2008 {
2009         int ret;
2010
2011         if (!capable(CAP_NET_ADMIN))
2012                 return -EPERM;
2013
2014         switch (cmd) {
2015         case IPT_SO_GET_INFO:
2016                 ret = get_info(sock_net(sk), user, len, 0);
2017                 break;
2018
2019         case IPT_SO_GET_ENTRIES:
2020                 ret = get_entries(sock_net(sk), user, len);
2021                 break;
2022
2023         case IPT_SO_GET_REVISION_MATCH:
2024         case IPT_SO_GET_REVISION_TARGET: {
2025                 struct xt_get_revision rev;
2026                 int target;
2027
2028                 if (*len != sizeof(rev)) {
2029                         ret = -EINVAL;
2030                         break;
2031                 }
2032                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2033                         ret = -EFAULT;
2034                         break;
2035                 }
2036                 rev.name[sizeof(rev.name)-1] = 0;
2037
2038                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2039                         target = 1;
2040                 else
2041                         target = 0;
2042
2043                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2044                                                          rev.revision,
2045                                                          target, &ret),
2046                                         "ipt_%s", rev.name);
2047                 break;
2048         }
2049
2050         default:
2051                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2052                 ret = -EINVAL;
2053         }
2054
2055         return ret;
2056 }
2057
2058 struct xt_table *ipt_register_table(struct net *net,
2059                                     const struct xt_table *table,
2060                                     const struct ipt_replace *repl)
2061 {
2062         int ret;
2063         struct xt_table_info *newinfo;
2064         struct xt_table_info bootstrap = {0};
2065         void *loc_cpu_entry;
2066         struct xt_table *new_table;
2067
2068         newinfo = xt_alloc_table_info(repl->size);
2069         if (!newinfo) {
2070                 ret = -ENOMEM;
2071                 goto out;
2072         }
2073
2074         /* choose the copy on our node/cpu, but dont care about preemption */
2075         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2076         memcpy(loc_cpu_entry, repl->entries, repl->size);
2077
2078         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2079         if (ret != 0)
2080                 goto out_free;
2081
2082         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2083         if (IS_ERR(new_table)) {
2084                 ret = PTR_ERR(new_table);
2085                 goto out_free;
2086         }
2087
2088         return new_table;
2089
2090 out_free:
2091         xt_free_table_info(newinfo);
2092 out:
2093         return ERR_PTR(ret);
2094 }
2095
2096 void ipt_unregister_table(struct net *net, struct xt_table *table)
2097 {
2098         struct xt_table_info *private;
2099         void *loc_cpu_entry;
2100         struct module *table_owner = table->me;
2101         struct ipt_entry *iter;
2102
2103         private = xt_unregister_table(table);
2104
2105         /* Decrease module usage counts and free resources */
2106         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2107         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2108                 cleanup_entry(iter, net);
2109         if (private->number > private->initial_entries)
2110                 module_put(table_owner);
2111         xt_free_table_info(private);
2112 }
2113
2114 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2115 static inline bool
2116 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2117                      u_int8_t type, u_int8_t code,
2118                      bool invert)
2119 {
2120         return ((test_type == 0xFF) ||
2121                 (type == test_type && code >= min_code && code <= max_code))
2122                 ^ invert;
2123 }
2124
2125 static bool
2126 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2127 {
2128         const struct icmphdr *ic;
2129         struct icmphdr _icmph;
2130         const struct ipt_icmp *icmpinfo = par->matchinfo;
2131
2132         /* Must not be a fragment. */
2133         if (par->fragoff != 0)
2134                 return false;
2135
2136         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2137         if (ic == NULL) {
2138                 /* We've been asked to examine this packet, and we
2139                  * can't.  Hence, no choice but to drop.
2140                  */
2141                 duprintf("Dropping evil ICMP tinygram.\n");
2142                 par->hotdrop = true;
2143                 return false;
2144         }
2145
2146         return icmp_type_code_match(icmpinfo->type,
2147                                     icmpinfo->code[0],
2148                                     icmpinfo->code[1],
2149                                     ic->type, ic->code,
2150                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2151 }
2152
2153 static int icmp_checkentry(const struct xt_mtchk_param *par)
2154 {
2155         const struct ipt_icmp *icmpinfo = par->matchinfo;
2156
2157         /* Must specify no unknown invflags */
2158         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2159 }
2160
2161 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2162         {
2163                 .name             = XT_STANDARD_TARGET,
2164                 .targetsize       = sizeof(int),
2165                 .family           = NFPROTO_IPV4,
2166 #ifdef CONFIG_COMPAT
2167                 .compatsize       = sizeof(compat_int_t),
2168                 .compat_from_user = compat_standard_from_user,
2169                 .compat_to_user   = compat_standard_to_user,
2170 #endif
2171         },
2172         {
2173                 .name             = XT_ERROR_TARGET,
2174                 .target           = ipt_error,
2175                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2176                 .family           = NFPROTO_IPV4,
2177         },
2178 };
2179
2180 static struct nf_sockopt_ops ipt_sockopts = {
2181         .pf             = PF_INET,
2182         .set_optmin     = IPT_BASE_CTL,
2183         .set_optmax     = IPT_SO_SET_MAX+1,
2184         .set            = do_ipt_set_ctl,
2185 #ifdef CONFIG_COMPAT
2186         .compat_set     = compat_do_ipt_set_ctl,
2187 #endif
2188         .get_optmin     = IPT_BASE_CTL,
2189         .get_optmax     = IPT_SO_GET_MAX+1,
2190         .get            = do_ipt_get_ctl,
2191 #ifdef CONFIG_COMPAT
2192         .compat_get     = compat_do_ipt_get_ctl,
2193 #endif
2194         .owner          = THIS_MODULE,
2195 };
2196
2197 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2198         {
2199                 .name       = "icmp",
2200                 .match      = icmp_match,
2201                 .matchsize  = sizeof(struct ipt_icmp),
2202                 .checkentry = icmp_checkentry,
2203                 .proto      = IPPROTO_ICMP,
2204                 .family     = NFPROTO_IPV4,
2205         },
2206 };
2207
2208 static int __net_init ip_tables_net_init(struct net *net)
2209 {
2210         return xt_proto_init(net, NFPROTO_IPV4);
2211 }
2212
2213 static void __net_exit ip_tables_net_exit(struct net *net)
2214 {
2215         xt_proto_fini(net, NFPROTO_IPV4);
2216 }
2217
2218 static struct pernet_operations ip_tables_net_ops = {
2219         .init = ip_tables_net_init,
2220         .exit = ip_tables_net_exit,
2221 };
2222
2223 static int __init ip_tables_init(void)
2224 {
2225         int ret;
2226
2227         ret = register_pernet_subsys(&ip_tables_net_ops);
2228         if (ret < 0)
2229                 goto err1;
2230
2231         /* No one else will be downing sem now, so we won't sleep */
2232         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2233         if (ret < 0)
2234                 goto err2;
2235         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2236         if (ret < 0)
2237                 goto err4;
2238
2239         /* Register setsockopt */
2240         ret = nf_register_sockopt(&ipt_sockopts);
2241         if (ret < 0)
2242                 goto err5;
2243
2244         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2245         return 0;
2246
2247 err5:
2248         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2249 err4:
2250         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2251 err2:
2252         unregister_pernet_subsys(&ip_tables_net_ops);
2253 err1:
2254         return ret;
2255 }
2256
2257 static void __exit ip_tables_fini(void)
2258 {
2259         nf_unregister_sockopt(&ipt_sockopts);
2260
2261         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2262         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2263         unregister_pernet_subsys(&ip_tables_net_ops);
2264 }
2265
2266 EXPORT_SYMBOL(ipt_register_table);
2267 EXPORT_SYMBOL(ipt_unregister_table);
2268 EXPORT_SYMBOL(ipt_do_table);
2269 module_init(ip_tables_init);
2270 module_exit(ip_tables_fini);