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  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13 #include <linux/cache.h>
  14 #include <linux/capability.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/kmod.h>
  17 #include <linux/vmalloc.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/module.h>
  20 #include <linux/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/compat.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/mutex.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/err.h>
  27 #include <linux/cpumask.h>
  28
  29 #include <linux/netfilter/x_tables.h>
  30 #include <linux/netfilter_ipv4/ip_tables.h>
  31 #include <net/netfilter/nf_log.h>
  32 #include "../../netfilter/xt_repldata.h"
  33
  34 MODULE_LICENSE("GPL");
  35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36 MODULE_DESCRIPTION("IPv4 packet filter");
  37
  38 /*#define DEBUG_IP_FIREWALL*/
  39 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  40 /*#define DEBUG_IP_FIREWALL_USER*/
  41
  42 #ifdef DEBUG_IP_FIREWALL
  43 #define dprintf(format, args...) pr_info(format , ## args)
  44 #else
  45 #define dprintf(format, args...)
  46 #endif
  47
  48 #ifdef DEBUG_IP_FIREWALL_USER
  49 #define duprintf(format, args...) pr_info(format , ## args)
  50 #else
  51 #define duprintf(format, args...)
  52 #endif
  53
  54 #ifdef CONFIG_NETFILTER_DEBUG
  55 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  56 #else
  57 #define IP_NF_ASSERT(x)
  58 #endif
  59
  60 #if 0
  61 /* All the better to debug you with... */
  62 #define static
  63 #define inline
  64 #endif
  65
  66 void *ipt_alloc_initial_table(const struct xt_table *info)
  67 {
  68         return xt_alloc_initial_table(ipt, IPT);
  69 }
  70 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  71
  72 /* Returns whether matches rule or not. */
  73 /* Performance critical - called for every packet */
  74 static inline bool
  75 ip_packet_match(const struct iphdr *ip,
  76                 const char *indev,
  77                 const char *outdev,
  78                 const struct ipt_ip *ipinfo,
  79                 int isfrag)
  80 {
  81         unsigned long ret;
  82
  83 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  84
  85         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  86                   IPT_INV_SRCIP) ||
  87             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  88                   IPT_INV_DSTIP)) {
  89                 dprintf("Source or dest mismatch.\n");
  90
  91                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  92                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  93                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  94                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  95                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  96                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  97                 return false;
  98         }
  99
 100         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 101
 102         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 103                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 104                         indev, ipinfo->iniface,
 105                         ipinfo->invflags & IPT_INV_VIA_IN ? " (INV)" : "");
 106                 return false;
 107         }
 108
 109         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 110
 111         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 112                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 113                         outdev, ipinfo->outiface,
 114                         ipinfo->invflags & IPT_INV_VIA_OUT ? " (INV)" : "");
 115                 return false;
 116         }
 117
 118         /* Check specific protocol */
 119         if (ipinfo->proto &&
 120             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 121                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 122                         ip->protocol, ipinfo->proto,
 123                         ipinfo->invflags & IPT_INV_PROTO ? " (INV)" : "");
 124                 return false;
 125         }
 126
 127         /* If we have a fragment rule but the packet is not a fragment
 128          * then we return zero */
 129         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 130                 dprintf("Fragment rule but not fragment.%s\n",
 131                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 132                 return false;
 133         }
 134
 135         return true;
 136 }
 137
 138 static bool
 139 ip_checkentry(const struct ipt_ip *ip)
 140 {
 141         if (ip->flags & ~IPT_F_MASK) {
 142                 duprintf("Unknown flag bits set: %08X\n",
 143                          ip->flags & ~IPT_F_MASK);
 144                 return false;
 145         }
 146         if (ip->invflags & ~IPT_INV_MASK) {
 147                 duprintf("Unknown invflag bits set: %08X\n",
 148                          ip->invflags & ~IPT_INV_MASK);
 149                 return false;
 150         }
 151         return true;
 152 }
 153
 154 static unsigned int
 155 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 156 {
 157         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 158
 159         return NF_DROP;
 160 }
 161
 162 /* Performance critical */
 163 static inline struct ipt_entry *
 164 get_entry(const void *base, unsigned int offset)
 165 {
 166         return (struct ipt_entry *)(base + offset);
 167 }
 168
 169 /* All zeroes == unconditional rule. */
 170 /* Mildly perf critical (only if packet tracing is on) */
 171 static inline bool unconditional(const struct ipt_entry *e)
 172 {
 173         static const struct ipt_ip uncond;
 174
 175         return e->target_offset == sizeof(struct ipt_entry) &&
 176                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
 177 #undef FWINV
 178 }
 179
 180 /* for const-correctness */
 181 static inline const struct xt_entry_target *
 182 ipt_get_target_c(const struct ipt_entry *e)
 183 {
 184         return ipt_get_target((struct ipt_entry *)e);
 185 }
 186
 187 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 188 static const char *const hooknames[] = {
 189         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 190         [NF_INET_LOCAL_IN]              = "INPUT",
 191         [NF_INET_FORWARD]               = "FORWARD",
 192         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 193         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 194 };
 195
 196 enum nf_ip_trace_comments {
 197         NF_IP_TRACE_COMMENT_RULE,
 198         NF_IP_TRACE_COMMENT_RETURN,
 199         NF_IP_TRACE_COMMENT_POLICY,
 200 };
 201
 202 static const char *const comments[] = {
 203         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 204         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 205         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 206 };
 207
 208 static struct nf_loginfo trace_loginfo = {
 209         .type = NF_LOG_TYPE_LOG,
 210         .u = {
 211                 .log = {
 212                         .level = 4,
 213                         .logflags = NF_LOG_MASK,
 214                 },
 215         },
 216 };
 217
 218 /* Mildly perf critical (only if packet tracing is on) */
 219 static inline int
 220 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 221                       const char *hookname, const char **chainname,
 222                       const char **comment, unsigned int *rulenum)
 223 {
 224         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 225
 226         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 227                 /* Head of user chain: ERROR target with chainname */
 228                 *chainname = t->target.data;
 229                 (*rulenum) = 0;
 230         } else if (s == e) {
 231                 (*rulenum)++;
 232
 233                 if (unconditional(s) &&
 234                     strcmp(t->target.u.kernel.target->name,
 235                            XT_STANDARD_TARGET) == 0 &&
 236                    t->verdict < 0) {
 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(struct net *net,
 250                          const struct sk_buff *skb,
 251                          unsigned int hook,
 252                          const struct net_device *in,
 253                          const struct net_device *out,
 254                          const char *tablename,
 255                          const struct xt_table_info *private,
 256                          const struct ipt_entry *e)
 257 {
 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         root = get_entry(private->entries, private->hook_entry[hook]);
 264
 265         hookname = chainname = hooknames[hook];
 266         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 267
 268         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 269                 if (get_chainname_rulenum(iter, e, hookname,
 270                     &chainname, &comment, &rulenum) != 0)
 271                         break;
 272
 273         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 274                      "TRACE: %s:%s:%s:%u ",
 275                      tablename, chainname, comment, rulenum);
 276 }
 277 #endif
 278
 279 static inline
 280 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 281 {
 282         return (void *)entry + entry->next_offset;
 283 }
 284
 285 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 286 unsigned int
 287 ipt_do_table(struct sk_buff *skb,
 288              const struct nf_hook_state *state,
 289              struct xt_table *table)
 290 {
 291         unsigned int hook = state->hook;
 292         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 293         const struct iphdr *ip;
 294         /* Initializing verdict to NF_DROP keeps gcc happy. */
 295         unsigned int verdict = NF_DROP;
 296         const char *indev, *outdev;
 297         const void *table_base;
 298         struct ipt_entry *e, **jumpstack;
 299         unsigned int stackidx, cpu;
 300         const struct xt_table_info *private;
 301         struct xt_action_param acpar;
 302         unsigned int addend;
 303
 304         /* Initialization */
 305         stackidx = 0;
 306         ip = ip_hdr(skb);
 307         indev = state->in ? state->in->name : nulldevname;
 308         outdev = state->out ? state->out->name : nulldevname;
 309         /* We handle fragments by dealing with the first fragment as
 310          * if it was a normal packet.  All other fragments are treated
 311          * normally, except that they will NEVER match rules that ask
 312          * things we don't know, ie. tcp syn flag or ports).  If the
 313          * rule is also a fragment-specific rule, non-fragments won't
 314          * match it. */
 315         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 316         acpar.thoff   = ip_hdrlen(skb);
 317         acpar.hotdrop = false;
 318         acpar.net     = state->net;
 319         acpar.in      = state->in;
 320         acpar.out     = state->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         /*
 330          * Ensure we load private-> members after we've fetched the base
 331          * pointer.
 332          */
 333         smp_read_barrier_depends();
 334         table_base = private->entries;
 335         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 336
 337         /* Switch to alternate jumpstack if we're being invoked via TEE.
 338          * TEE issues XT_CONTINUE verdict on original skb so we must not
 339          * clobber the jumpstack.
 340          *
 341          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 342          * but it is no problem since absolute verdict is issued by these.
 343          */
 344         if (static_key_false(&xt_tee_enabled))
 345                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 346
 347         e = get_entry(table_base, private->hook_entry[hook]);
 348
 349         pr_debug("Entering %s(hook %u), UF %p\n",
 350                  table->name, hook,
 351                  get_entry(table_base, private->underflow[hook]));
 352
 353         do {
 354                 const struct xt_entry_target *t;
 355                 const struct xt_entry_match *ematch;
 356                 struct xt_counters *counter;
 357
 358                 IP_NF_ASSERT(e);
 359                 if (!ip_packet_match(ip, indev, outdev,
 360                     &e->ip, acpar.fragoff)) {
 361  no_match:
 362                         e = ipt_next_entry(e);
 363                         continue;
 364                 }
 365
 366                 xt_ematch_foreach(ematch, e) {
 367                         acpar.match     = ematch->u.kernel.match;
 368                         acpar.matchinfo = ematch->data;
 369                         if (!acpar.match->match(skb, &acpar))
 370                                 goto no_match;
 371                 }
 372
 373                 counter = xt_get_this_cpu_counter(&e->counters);
 374                 ADD_COUNTER(*counter, skb->len, 1);
 375
 376                 t = ipt_get_target(e);
 377                 IP_NF_ASSERT(t->u.kernel.target);
 378
 379 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 380                 /* The packet is traced: log it */
 381                 if (unlikely(skb->nf_trace))
 382                         trace_packet(state->net, skb, hook, state->in,
 383                                      state->out, table->name, private, e);
 384 #endif
 385                 /* Standard target? */
 386                 if (!t->u.kernel.target->target) {
 387                         int v;
 388
 389                         v = ((struct xt_standard_target *)t)->verdict;
 390                         if (v < 0) {
 391                                 /* Pop from stack? */
 392                                 if (v != XT_RETURN) {
 393                                         verdict = (unsigned int)(-v) - 1;
 394                                         break;
 395                                 }
 396                                 if (stackidx == 0) {
 397                                         e = get_entry(table_base,
 398                                             private->underflow[hook]);
 399                                         pr_debug("Underflow (this is normal) "
 400                                                  "to %p\n", e);
 401                                 } else {
 402                                         e = jumpstack[--stackidx];
 403                                         pr_debug("Pulled %p out from pos %u\n",
 404                                                  e, stackidx);
 405                                         e = ipt_next_entry(e);
 406                                 }
 407                                 continue;
 408                         }
 409                         if (table_base + v != ipt_next_entry(e) &&
 410                             !(e->ip.flags & IPT_F_GOTO)) {
 411                                 if (unlikely(stackidx >= private->stacksize)) {
 412                                         verdict = NF_DROP;
 413                                         break;
 414                                 }
 415                                 jumpstack[stackidx++] = e;
 416                                 pr_debug("Pushed %p into pos %u\n",
 417                                          e, stackidx - 1);
 418                         }
 419
 420                         e = get_entry(table_base, v);
 421                         continue;
 422                 }
 423
 424                 acpar.target   = t->u.kernel.target;
 425                 acpar.targinfo = t->data;
 426
 427                 verdict = t->u.kernel.target->target(skb, &acpar);
 428                 /* Target might have changed stuff. */
 429                 ip = ip_hdr(skb);
 430                 if (verdict == XT_CONTINUE)
 431                         e = ipt_next_entry(e);
 432                 else
 433                         /* Verdict */
 434                         break;
 435         } while (!acpar.hotdrop);
 436         pr_debug("Exiting %s; sp at %u\n", __func__, stackidx);
 437
 438         xt_write_recseq_end(addend);
 439         local_bh_enable();
 440
 441 #ifdef DEBUG_ALLOW_ALL
 442         return NF_ACCEPT;
 443 #else
 444         if (acpar.hotdrop)
 445                 return NF_DROP;
 446         else return verdict;
 447 #endif
 448 }
 449
 450 /* Figures out from what hook each rule can be called: returns 0 if
 451    there are loops.  Puts hook bitmask in comefrom. */
 452 static int
 453 mark_source_chains(const struct xt_table_info *newinfo,
 454                    unsigned int valid_hooks, void *entry0,
 455                    unsigned int *offsets)
 456 {
 457         unsigned int hook;
 458
 459         /* No recursion; use packet counter to save back ptrs (reset
 460            to 0 as we leave), and comefrom to save source hook bitmask */
 461         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 462                 unsigned int pos = newinfo->hook_entry[hook];
 463                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 464
 465                 if (!(valid_hooks & (1 << hook)))
 466                         continue;
 467
 468                 /* Set initial back pointer. */
 469                 e->counters.pcnt = pos;
 470
 471                 for (;;) {
 472                         const struct xt_standard_target *t
 473                                 = (void *)ipt_get_target_c(e);
 474                         int visited = e->comefrom & (1 << hook);
 475
 476                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 477                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 478                                        hook, pos, e->comefrom);
 479                                 return 0;
 480                         }
 481                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 482
 483                         /* Unconditional return/END. */
 484                         if ((unconditional(e) &&
 485                              (strcmp(t->target.u.user.name,
 486                                      XT_STANDARD_TARGET) == 0) &&
 487                              t->verdict < 0) || visited) {
 488                                 unsigned int oldpos, size;
 489
 490                                 if ((strcmp(t->target.u.user.name,
 491                                             XT_STANDARD_TARGET) == 0) &&
 492                                     t->verdict < -NF_MAX_VERDICT - 1) {
 493                                         duprintf("mark_source_chains: bad "
 494                                                 "negative verdict (%i)\n",
 495                                                                 t->verdict);
 496                                         return 0;
 497                                 }
 498
 499                                 /* Return: backtrack through the last
 500                                    big jump. */
 501                                 do {
 502                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 503 #ifdef DEBUG_IP_FIREWALL_USER
 504                                         if (e->comefrom
 505                                             & (1 << NF_INET_NUMHOOKS)) {
 506                                                 duprintf("Back unset "
 507                                                          "on hook %u "
 508                                                          "rule %u\n",
 509                                                          hook, pos);
 510                                         }
 511 #endif
 512                                         oldpos = pos;
 513                                         pos = e->counters.pcnt;
 514                                         e->counters.pcnt = 0;
 515
 516                                         /* We're at the start. */
 517                                         if (pos == oldpos)
 518                                                 goto next;
 519
 520                                         e = (struct ipt_entry *)
 521                                                 (entry0 + pos);
 522                                 } while (oldpos == pos + e->next_offset);
 523
 524                                 /* Move along one */
 525                                 size = e->next_offset;
 526                                 e = (struct ipt_entry *)
 527                                         (entry0 + pos + size);
 528                                 if (pos + size >= newinfo->size)
 529                                         return 0;
 530                                 e->counters.pcnt = pos;
 531                                 pos += size;
 532                         } else {
 533                                 int newpos = t->verdict;
 534
 535                                 if (strcmp(t->target.u.user.name,
 536                                            XT_STANDARD_TARGET) == 0 &&
 537                                     newpos >= 0) {
 538                                         if (newpos > newinfo->size -
 539                                                 sizeof(struct ipt_entry)) {
 540                                                 duprintf("mark_source_chains: "
 541                                                         "bad verdict (%i)\n",
 542                                                                 newpos);
 543                                                 return 0;
 544                                         }
 545                                         /* This a jump; chase it. */
 546                                         duprintf("Jump rule %u -> %u\n",
 547                                                  pos, newpos);
 548                                         if (!xt_find_jump_offset(offsets, newpos,
 549                                                                  newinfo->number))
 550                                                 return 0;
 551                                         e = (struct ipt_entry *)
 552                                                 (entry0 + newpos);
 553                                 } else {
 554                                         /* ... this is a fallthru */
 555                                         newpos = pos + e->next_offset;
 556                                         if (newpos >= newinfo->size)
 557                                                 return 0;
 558                                 }
 559                                 e = (struct ipt_entry *)
 560                                         (entry0 + newpos);
 561                                 e->counters.pcnt = pos;
 562                                 pos = newpos;
 563                         }
 564                 }
 565 next:
 566                 duprintf("Finished chain %u\n", hook);
 567         }
 568         return 1;
 569 }
 570
 571 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 572 {
 573         struct xt_mtdtor_param par;
 574
 575         par.net       = net;
 576         par.match     = m->u.kernel.match;
 577         par.matchinfo = m->data;
 578         par.family    = NFPROTO_IPV4;
 579         if (par.match->destroy != NULL)
 580                 par.match->destroy(&par);
 581         module_put(par.match->me);
 582 }
 583
 584 static int
 585 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 586 {
 587         const struct ipt_ip *ip = par->entryinfo;
 588         int ret;
 589
 590         par->match     = m->u.kernel.match;
 591         par->matchinfo = m->data;
 592
 593         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 594               ip->proto, ip->invflags & IPT_INV_PROTO);
 595         if (ret < 0) {
 596                 duprintf("check failed for `%s'.\n", par->match->name);
 597                 return ret;
 598         }
 599         return 0;
 600 }
 601
 602 static int
 603 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 604 {
 605         struct xt_match *match;
 606         int ret;
 607
 608         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 609                                       m->u.user.revision);
 610         if (IS_ERR(match)) {
 611                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 612                 return PTR_ERR(match);
 613         }
 614         m->u.kernel.match = match;
 615
 616         ret = check_match(m, par);
 617         if (ret)
 618                 goto err;
 619
 620         return 0;
 621 err:
 622         module_put(m->u.kernel.match->me);
 623         return ret;
 624 }
 625
 626 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 627 {
 628         struct xt_entry_target *t = ipt_get_target(e);
 629         struct xt_tgchk_param par = {
 630                 .net       = net,
 631                 .table     = name,
 632                 .entryinfo = e,
 633                 .target    = t->u.kernel.target,
 634                 .targinfo  = t->data,
 635                 .hook_mask = e->comefrom,
 636                 .family    = NFPROTO_IPV4,
 637         };
 638         int ret;
 639
 640         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 641               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 642         if (ret < 0) {
 643                 duprintf("check failed for `%s'.\n",
 644                          t->u.kernel.target->name);
 645                 return ret;
 646         }
 647         return 0;
 648 }
 649
 650 static int
 651 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 652                  unsigned int size,
 653                  struct xt_percpu_counter_alloc_state *alloc_state)
 654 {
 655         struct xt_entry_target *t;
 656         struct xt_target *target;
 657         int ret;
 658         unsigned int j;
 659         struct xt_mtchk_param mtpar;
 660         struct xt_entry_match *ematch;
 661
 662         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 663                 return -ENOMEM;
 664
 665         j = 0;
 666         memset(&mtpar, 0, sizeof(mtpar));
 667         mtpar.net       = net;
 668         mtpar.table     = name;
 669         mtpar.entryinfo = &e->ip;
 670         mtpar.hook_mask = e->comefrom;
 671         mtpar.family    = NFPROTO_IPV4;
 672         xt_ematch_foreach(ematch, e) {
 673                 ret = find_check_match(ematch, &mtpar);
 674                 if (ret != 0)
 675                         goto cleanup_matches;
 676                 ++j;
 677         }
 678
 679         t = ipt_get_target(e);
 680         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 681                                         t->u.user.revision);
 682         if (IS_ERR(target)) {
 683                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 684                 ret = PTR_ERR(target);
 685                 goto cleanup_matches;
 686         }
 687         t->u.kernel.target = target;
 688
 689         ret = check_target(e, net, name);
 690         if (ret)
 691                 goto err;
 692
 693         return 0;
 694  err:
 695         module_put(t->u.kernel.target->me);
 696  cleanup_matches:
 697         xt_ematch_foreach(ematch, e) {
 698                 if (j-- == 0)
 699                         break;
 700                 cleanup_match(ematch, net);
 701         }
 702
 703         xt_percpu_counter_free(&e->counters);
 704
 705         return ret;
 706 }
 707
 708 static bool check_underflow(const struct ipt_entry *e)
 709 {
 710         const struct xt_entry_target *t;
 711         unsigned int verdict;
 712
 713         if (!unconditional(e))
 714                 return false;
 715         t = ipt_get_target_c(e);
 716         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 717                 return false;
 718         verdict = ((struct xt_standard_target *)t)->verdict;
 719         verdict = -verdict - 1;
 720         return verdict == NF_DROP || verdict == NF_ACCEPT;
 721 }
 722
 723 static int
 724 check_entry_size_and_hooks(struct ipt_entry *e,
 725                            struct xt_table_info *newinfo,
 726                            const unsigned char *base,
 727                            const unsigned char *limit,
 728                            const unsigned int *hook_entries,
 729                            const unsigned int *underflows,
 730                            unsigned int valid_hooks)
 731 {
 732         unsigned int h;
 733         int err;
 734
 735         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 736             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 737             (unsigned char *)e + e->next_offset > limit) {
 738                 duprintf("Bad offset %p\n", e);
 739                 return -EINVAL;
 740         }
 741
 742         if (e->next_offset
 743             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 744                 duprintf("checking: element %p size %u\n",
 745                          e, e->next_offset);
 746                 return -EINVAL;
 747         }
 748
 749         if (!ip_checkentry(&e->ip))
 750                 return -EINVAL;
 751
 752         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 753                                      e->next_offset);
 754         if (err)
 755                 return err;
 756
 757         /* Check hooks & underflows */
 758         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 759                 if (!(valid_hooks & (1 << h)))
 760                         continue;
 761                 if ((unsigned char *)e - base == hook_entries[h])
 762                         newinfo->hook_entry[h] = hook_entries[h];
 763                 if ((unsigned char *)e - base == underflows[h]) {
 764                         if (!check_underflow(e)) {
 765                                 pr_debug("Underflows must be unconditional and "
 766                                          "use the STANDARD target with "
 767                                          "ACCEPT/DROP\n");
 768                                 return -EINVAL;
 769                         }
 770                         newinfo->underflow[h] = underflows[h];
 771                 }
 772         }
 773
 774         /* Clear counters and comefrom */
 775         e->counters = ((struct xt_counters) { 0, 0 });
 776         e->comefrom = 0;
 777         return 0;
 778 }
 779
 780 static void
 781 cleanup_entry(struct ipt_entry *e, struct net *net)
 782 {
 783         struct xt_tgdtor_param par;
 784         struct xt_entry_target *t;
 785         struct xt_entry_match *ematch;
 786
 787         /* Cleanup all matches */
 788         xt_ematch_foreach(ematch, e)
 789                 cleanup_match(ematch, net);
 790         t = ipt_get_target(e);
 791
 792         par.net      = net;
 793         par.target   = t->u.kernel.target;
 794         par.targinfo = t->data;
 795         par.family   = NFPROTO_IPV4;
 796         if (par.target->destroy != NULL)
 797                 par.target->destroy(&par);
 798         module_put(par.target->me);
 799         xt_percpu_counter_free(&e->counters);
 800 }
 801
 802 /* Checks and translates the user-supplied table segment (held in
 803    newinfo) */
 804 static int
 805 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 806                 const struct ipt_replace *repl)
 807 {
 808         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 809         struct ipt_entry *iter;
 810         unsigned int *offsets;
 811         unsigned int i;
 812         int ret = 0;
 813
 814         newinfo->size = repl->size;
 815         newinfo->number = repl->num_entries;
 816
 817         /* Init all hooks to impossible value. */
 818         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 819                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 820                 newinfo->underflow[i] = 0xFFFFFFFF;
 821         }
 822
 823         duprintf("translate_table: size %u\n", newinfo->size);
 824         offsets = xt_alloc_entry_offsets(newinfo->number);
 825         if (!offsets)
 826                 return -ENOMEM;
 827         i = 0;
 828         /* Walk through entries, checking offsets. */
 829         xt_entry_foreach(iter, entry0, newinfo->size) {
 830                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 831                                                  entry0 + repl->size,
 832                                                  repl->hook_entry,
 833                                                  repl->underflow,
 834                                                  repl->valid_hooks);
 835                 if (ret != 0)
 836                         goto out_free;
 837                 if (i < repl->num_entries)
 838                         offsets[i] = (void *)iter - entry0;
 839                 ++i;
 840                 if (strcmp(ipt_get_target(iter)->u.user.name,
 841                     XT_ERROR_TARGET) == 0)
 842                         ++newinfo->stacksize;
 843         }
 844
 845         ret = -EINVAL;
 846         if (i != repl->num_entries) {
 847                 duprintf("translate_table: %u not %u entries\n",
 848                          i, repl->num_entries);
 849                 goto out_free;
 850         }
 851
 852         /* Check hooks all assigned */
 853         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 854                 /* Only hooks which are valid */
 855                 if (!(repl->valid_hooks & (1 << i)))
 856                         continue;
 857                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 858                         duprintf("Invalid hook entry %u %u\n",
 859                                  i, repl->hook_entry[i]);
 860                         goto out_free;
 861                 }
 862                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 863                         duprintf("Invalid underflow %u %u\n",
 864                                  i, repl->underflow[i]);
 865                         goto out_free;
 866                 }
 867         }
 868
 869         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 870                 ret = -ELOOP;
 871                 goto out_free;
 872         }
 873         kvfree(offsets);
 874
 875         /* Finally, each sanity check must pass */
 876         i = 0;
 877         xt_entry_foreach(iter, entry0, newinfo->size) {
 878                 ret = find_check_entry(iter, net, repl->name, repl->size,
 879                                        &alloc_state);
 880                 if (ret != 0)
 881                         break;
 882                 ++i;
 883         }
 884
 885         if (ret != 0) {
 886                 xt_entry_foreach(iter, entry0, newinfo->size) {
 887                         if (i-- == 0)
 888                                 break;
 889                         cleanup_entry(iter, net);
 890                 }
 891                 return ret;
 892         }
 893
 894         return ret;
 895  out_free:
 896         kvfree(offsets);
 897         return ret;
 898 }
 899
 900 static void
 901 get_counters(const struct xt_table_info *t,
 902              struct xt_counters counters[])
 903 {
 904         struct ipt_entry *iter;
 905         unsigned int cpu;
 906         unsigned int i;
 907
 908         for_each_possible_cpu(cpu) {
 909                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 910
 911                 i = 0;
 912                 xt_entry_foreach(iter, t->entries, t->size) {
 913                         struct xt_counters *tmp;
 914                         u64 bcnt, pcnt;
 915                         unsigned int start;
 916
 917                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 918                         do {
 919                                 start = read_seqcount_begin(s);
 920                                 bcnt = tmp->bcnt;
 921                                 pcnt = tmp->pcnt;
 922                         } while (read_seqcount_retry(s, start));
 923
 924                         ADD_COUNTER(counters[i], bcnt, pcnt);
 925                         ++i; /* macro does multi eval of i */
 926                 }
 927         }
 928 }
 929
 930 static struct xt_counters *alloc_counters(const struct xt_table *table)
 931 {
 932         unsigned int countersize;
 933         struct xt_counters *counters;
 934         const struct xt_table_info *private = table->private;
 935
 936         /* We need atomic snapshot of counters: rest doesn't change
 937            (other than comefrom, which userspace doesn't care
 938            about). */
 939         countersize = sizeof(struct xt_counters) * private->number;
 940         counters = vzalloc(countersize);
 941
 942         if (counters == NULL)
 943                 return ERR_PTR(-ENOMEM);
 944
 945         get_counters(private, counters);
 946
 947         return counters;
 948 }
 949
 950 static int
 951 copy_entries_to_user(unsigned int total_size,
 952                      const struct xt_table *table,
 953                      void __user *userptr)
 954 {
 955         unsigned int off, num;
 956         const struct ipt_entry *e;
 957         struct xt_counters *counters;
 958         const struct xt_table_info *private = table->private;
 959         int ret = 0;
 960         const void *loc_cpu_entry;
 961
 962         counters = alloc_counters(table);
 963         if (IS_ERR(counters))
 964                 return PTR_ERR(counters);
 965
 966         loc_cpu_entry = private->entries;
 967         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 968                 ret = -EFAULT;
 969                 goto free_counters;
 970         }
 971
 972         /* FIXME: use iterator macros --RR */
 973         /* ... then go back and fix counters and names */
 974         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 975                 unsigned int i;
 976                 const struct xt_entry_match *m;
 977                 const struct xt_entry_target *t;
 978
 979                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 980                 if (copy_to_user(userptr + off
 981                                  + offsetof(struct ipt_entry, counters),
 982                                  &counters[num],
 983                                  sizeof(counters[num])) != 0) {
 984                         ret = -EFAULT;
 985                         goto free_counters;
 986                 }
 987
 988                 for (i = sizeof(struct ipt_entry);
 989                      i < e->target_offset;
 990                      i += m->u.match_size) {
 991                         m = (void *)e + i;
 992
 993                         if (copy_to_user(userptr + off + i
 994                                          + offsetof(struct xt_entry_match,
 995                                                     u.user.name),
 996                                          m->u.kernel.match->name,
 997                                          strlen(m->u.kernel.match->name)+1)
 998                             != 0) {
 999                                 ret = -EFAULT;
1000                                 goto free_counters;
1001                         }
1002                 }
1003
1004                 t = ipt_get_target_c(e);
1005                 if (copy_to_user(userptr + off + e->target_offset
1006                                  + offsetof(struct xt_entry_target,
1007                                             u.user.name),
1008                                  t->u.kernel.target->name,
1009                                  strlen(t->u.kernel.target->name)+1) != 0) {
1010                         ret = -EFAULT;
1011                         goto free_counters;
1012                 }
1013         }
1014
1015  free_counters:
1016         vfree(counters);
1017         return ret;
1018 }
1019
1020 #ifdef CONFIG_COMPAT
1021 static void compat_standard_from_user(void *dst, const void *src)
1022 {
1023         int v = *(compat_int_t *)src;
1024
1025         if (v > 0)
1026                 v += xt_compat_calc_jump(AF_INET, v);
1027         memcpy(dst, &v, sizeof(v));
1028 }
1029
1030 static int compat_standard_to_user(void __user *dst, const void *src)
1031 {
1032         compat_int_t cv = *(int *)src;
1033
1034         if (cv > 0)
1035                 cv -= xt_compat_calc_jump(AF_INET, cv);
1036         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1037 }
1038
1039 static int compat_calc_entry(const struct ipt_entry *e,
1040                              const struct xt_table_info *info,
1041                              const void *base, struct xt_table_info *newinfo)
1042 {
1043         const struct xt_entry_match *ematch;
1044         const struct xt_entry_target *t;
1045         unsigned int entry_offset;
1046         int off, i, ret;
1047
1048         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1049         entry_offset = (void *)e - base;
1050         xt_ematch_foreach(ematch, e)
1051                 off += xt_compat_match_offset(ematch->u.kernel.match);
1052         t = ipt_get_target_c(e);
1053         off += xt_compat_target_offset(t->u.kernel.target);
1054         newinfo->size -= off;
1055         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1056         if (ret)
1057                 return ret;
1058
1059         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1060                 if (info->hook_entry[i] &&
1061                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1062                         newinfo->hook_entry[i] -= off;
1063                 if (info->underflow[i] &&
1064                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1065                         newinfo->underflow[i] -= off;
1066         }
1067         return 0;
1068 }
1069
1070 static int compat_table_info(const struct xt_table_info *info,
1071                              struct xt_table_info *newinfo)
1072 {
1073         struct ipt_entry *iter;
1074         const void *loc_cpu_entry;
1075         int ret;
1076
1077         if (!newinfo || !info)
1078                 return -EINVAL;
1079
1080         /* we dont care about newinfo->entries */
1081         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1082         newinfo->initial_entries = 0;
1083         loc_cpu_entry = info->entries;
1084         xt_compat_init_offsets(AF_INET, info->number);
1085         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1086                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1087                 if (ret != 0)
1088                         return ret;
1089         }
1090         return 0;
1091 }
1092 #endif
1093
1094 static int get_info(struct net *net, void __user *user,
1095                     const int *len, int compat)
1096 {
1097         char name[XT_TABLE_MAXNAMELEN];
1098         struct xt_table *t;
1099         int ret;
1100
1101         if (*len != sizeof(struct ipt_getinfo)) {
1102                 duprintf("length %u != %zu\n", *len,
1103                          sizeof(struct ipt_getinfo));
1104                 return -EINVAL;
1105         }
1106
1107         if (copy_from_user(name, user, sizeof(name)) != 0)
1108                 return -EFAULT;
1109
1110         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1111 #ifdef CONFIG_COMPAT
1112         if (compat)
1113                 xt_compat_lock(AF_INET);
1114 #endif
1115         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1116                                     "iptable_%s", name);
1117         if (!IS_ERR_OR_NULL(t)) {
1118                 struct ipt_getinfo info;
1119                 const struct xt_table_info *private = t->private;
1120 #ifdef CONFIG_COMPAT
1121                 struct xt_table_info tmp;
1122
1123                 if (compat) {
1124                         ret = compat_table_info(private, &tmp);
1125                         xt_compat_flush_offsets(AF_INET);
1126                         private = &tmp;
1127                 }
1128 #endif
1129                 memset(&info, 0, sizeof(info));
1130                 info.valid_hooks = t->valid_hooks;
1131                 memcpy(info.hook_entry, private->hook_entry,
1132                        sizeof(info.hook_entry));
1133                 memcpy(info.underflow, private->underflow,
1134                        sizeof(info.underflow));
1135                 info.num_entries = private->number;
1136                 info.size = private->size;
1137                 strcpy(info.name, name);
1138
1139                 if (copy_to_user(user, &info, *len) != 0)
1140                         ret = -EFAULT;
1141                 else
1142                         ret = 0;
1143
1144                 xt_table_unlock(t);
1145                 module_put(t->me);
1146         } else
1147                 ret = t ? PTR_ERR(t) : -ENOENT;
1148 #ifdef CONFIG_COMPAT
1149         if (compat)
1150                 xt_compat_unlock(AF_INET);
1151 #endif
1152         return ret;
1153 }
1154
1155 static int
1156 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1157             const int *len)
1158 {
1159         int ret;
1160         struct ipt_get_entries get;
1161         struct xt_table *t;
1162
1163         if (*len < sizeof(get)) {
1164                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1165                 return -EINVAL;
1166         }
1167         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1168                 return -EFAULT;
1169         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1170                 duprintf("get_entries: %u != %zu\n",
1171                          *len, sizeof(get) + get.size);
1172                 return -EINVAL;
1173         }
1174         get.name[sizeof(get.name) - 1] = '\0';
1175
1176         t = xt_find_table_lock(net, AF_INET, get.name);
1177         if (!IS_ERR_OR_NULL(t)) {
1178                 const struct xt_table_info *private = t->private;
1179                 duprintf("t->private->number = %u\n", private->number);
1180                 if (get.size == private->size)
1181                         ret = copy_entries_to_user(private->size,
1182                                                    t, uptr->entrytable);
1183                 else {
1184                         duprintf("get_entries: I've got %u not %u!\n",
1185                                  private->size, get.size);
1186                         ret = -EAGAIN;
1187                 }
1188                 module_put(t->me);
1189                 xt_table_unlock(t);
1190         } else
1191                 ret = t ? PTR_ERR(t) : -ENOENT;
1192
1193         return ret;
1194 }
1195
1196 static int
1197 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1198              struct xt_table_info *newinfo, unsigned int num_counters,
1199              void __user *counters_ptr)
1200 {
1201         int ret;
1202         struct xt_table *t;
1203         struct xt_table_info *oldinfo;
1204         struct xt_counters *counters;
1205         struct ipt_entry *iter;
1206
1207         ret = 0;
1208         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1209         if (!counters) {
1210                 ret = -ENOMEM;
1211                 goto out;
1212         }
1213
1214         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1215                                     "iptable_%s", name);
1216         if (IS_ERR_OR_NULL(t)) {
1217                 ret = t ? PTR_ERR(t) : -ENOENT;
1218                 goto free_newinfo_counters_untrans;
1219         }
1220
1221         /* You lied! */
1222         if (valid_hooks != t->valid_hooks) {
1223                 duprintf("Valid hook crap: %08X vs %08X\n",
1224                          valid_hooks, t->valid_hooks);
1225                 ret = -EINVAL;
1226                 goto put_module;
1227         }
1228
1229         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1230         if (!oldinfo)
1231                 goto put_module;
1232
1233         /* Update module usage count based on number of rules */
1234         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1235                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1236         if ((oldinfo->number > oldinfo->initial_entries) ||
1237             (newinfo->number <= oldinfo->initial_entries))
1238                 module_put(t->me);
1239         if ((oldinfo->number > oldinfo->initial_entries) &&
1240             (newinfo->number <= oldinfo->initial_entries))
1241                 module_put(t->me);
1242
1243         /* Get the old counters, and synchronize with replace */
1244         get_counters(oldinfo, counters);
1245
1246         /* Decrease module usage counts and free resource */
1247         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1248                 cleanup_entry(iter, net);
1249
1250         xt_free_table_info(oldinfo);
1251         if (copy_to_user(counters_ptr, counters,
1252                          sizeof(struct xt_counters) * num_counters) != 0) {
1253                 /* Silent error, can't fail, new table is already in place */
1254                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1255         }
1256         vfree(counters);
1257         xt_table_unlock(t);
1258         return ret;
1259
1260  put_module:
1261         module_put(t->me);
1262         xt_table_unlock(t);
1263  free_newinfo_counters_untrans:
1264         vfree(counters);
1265  out:
1266         return ret;
1267 }
1268
1269 static int
1270 do_replace(struct net *net, const void __user *user, unsigned int len)
1271 {
1272         int ret;
1273         struct ipt_replace tmp;
1274         struct xt_table_info *newinfo;
1275         void *loc_cpu_entry;
1276         struct ipt_entry *iter;
1277
1278         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1279                 return -EFAULT;
1280
1281         /* overflow check */
1282         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1283                 return -ENOMEM;
1284         if (tmp.num_counters == 0)
1285                 return -EINVAL;
1286
1287         tmp.name[sizeof(tmp.name)-1] = 0;
1288
1289         newinfo = xt_alloc_table_info(tmp.size);
1290         if (!newinfo)
1291                 return -ENOMEM;
1292
1293         loc_cpu_entry = newinfo->entries;
1294         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1295                            tmp.size) != 0) {
1296                 ret = -EFAULT;
1297                 goto free_newinfo;
1298         }
1299
1300         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1301         if (ret != 0)
1302                 goto free_newinfo;
1303
1304         duprintf("Translated table\n");
1305
1306         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1307                            tmp.num_counters, tmp.counters);
1308         if (ret)
1309                 goto free_newinfo_untrans;
1310         return 0;
1311
1312  free_newinfo_untrans:
1313         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1314                 cleanup_entry(iter, net);
1315  free_newinfo:
1316         xt_free_table_info(newinfo);
1317         return ret;
1318 }
1319
1320 static int
1321 do_add_counters(struct net *net, const void __user *user,
1322                 unsigned int len, int compat)
1323 {
1324         unsigned int i;
1325         struct xt_counters_info tmp;
1326         struct xt_counters *paddc;
1327         struct xt_table *t;
1328         const struct xt_table_info *private;
1329         int ret = 0;
1330         struct ipt_entry *iter;
1331         unsigned int addend;
1332
1333         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1334         if (IS_ERR(paddc))
1335                 return PTR_ERR(paddc);
1336
1337         t = xt_find_table_lock(net, AF_INET, tmp.name);
1338         if (IS_ERR_OR_NULL(t)) {
1339                 ret = t ? PTR_ERR(t) : -ENOENT;
1340                 goto free;
1341         }
1342
1343         local_bh_disable();
1344         private = t->private;
1345         if (private->number != tmp.num_counters) {
1346                 ret = -EINVAL;
1347                 goto unlock_up_free;
1348         }
1349
1350         i = 0;
1351         addend = xt_write_recseq_begin();
1352         xt_entry_foreach(iter, private->entries, private->size) {
1353                 struct xt_counters *tmp;
1354
1355                 tmp = xt_get_this_cpu_counter(&iter->counters);
1356                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1357                 ++i;
1358         }
1359         xt_write_recseq_end(addend);
1360  unlock_up_free:
1361         local_bh_enable();
1362         xt_table_unlock(t);
1363         module_put(t->me);
1364  free:
1365         vfree(paddc);
1366
1367         return ret;
1368 }
1369
1370 #ifdef CONFIG_COMPAT
1371 struct compat_ipt_replace {
1372         char                    name[XT_TABLE_MAXNAMELEN];
1373         u32                     valid_hooks;
1374         u32                     num_entries;
1375         u32                     size;
1376         u32                     hook_entry[NF_INET_NUMHOOKS];
1377         u32                     underflow[NF_INET_NUMHOOKS];
1378         u32                     num_counters;
1379         compat_uptr_t           counters;       /* struct xt_counters * */
1380         struct compat_ipt_entry entries[0];
1381 };
1382
1383 static int
1384 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1385                           unsigned int *size, struct xt_counters *counters,
1386                           unsigned int i)
1387 {
1388         struct xt_entry_target *t;
1389         struct compat_ipt_entry __user *ce;
1390         u_int16_t target_offset, next_offset;
1391         compat_uint_t origsize;
1392         const struct xt_entry_match *ematch;
1393         int ret = 0;
1394
1395         origsize = *size;
1396         ce = (struct compat_ipt_entry __user *)*dstptr;
1397         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1398             copy_to_user(&ce->counters, &counters[i],
1399             sizeof(counters[i])) != 0)
1400                 return -EFAULT;
1401
1402         *dstptr += sizeof(struct compat_ipt_entry);
1403         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1404
1405         xt_ematch_foreach(ematch, e) {
1406                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1407                 if (ret != 0)
1408                         return ret;
1409         }
1410         target_offset = e->target_offset - (origsize - *size);
1411         t = ipt_get_target(e);
1412         ret = xt_compat_target_to_user(t, dstptr, size);
1413         if (ret)
1414                 return ret;
1415         next_offset = e->next_offset - (origsize - *size);
1416         if (put_user(target_offset, &ce->target_offset) != 0 ||
1417             put_user(next_offset, &ce->next_offset) != 0)
1418                 return -EFAULT;
1419         return 0;
1420 }
1421
1422 static int
1423 compat_find_calc_match(struct xt_entry_match *m,
1424                        const struct ipt_ip *ip,
1425                        int *size)
1426 {
1427         struct xt_match *match;
1428
1429         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1430                                       m->u.user.revision);
1431         if (IS_ERR(match)) {
1432                 duprintf("compat_check_calc_match: `%s' not found\n",
1433                          m->u.user.name);
1434                 return PTR_ERR(match);
1435         }
1436         m->u.kernel.match = match;
1437         *size += xt_compat_match_offset(match);
1438         return 0;
1439 }
1440
1441 static void compat_release_entry(struct compat_ipt_entry *e)
1442 {
1443         struct xt_entry_target *t;
1444         struct xt_entry_match *ematch;
1445
1446         /* Cleanup all matches */
1447         xt_ematch_foreach(ematch, e)
1448                 module_put(ematch->u.kernel.match->me);
1449         t = compat_ipt_get_target(e);
1450         module_put(t->u.kernel.target->me);
1451 }
1452
1453 static int
1454 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1455                                   struct xt_table_info *newinfo,
1456                                   unsigned int *size,
1457                                   const unsigned char *base,
1458                                   const unsigned char *limit)
1459 {
1460         struct xt_entry_match *ematch;
1461         struct xt_entry_target *t;
1462         struct xt_target *target;
1463         unsigned int entry_offset;
1464         unsigned int j;
1465         int ret, off;
1466
1467         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1468         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1469             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1470             (unsigned char *)e + e->next_offset > limit) {
1471                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1472                 return -EINVAL;
1473         }
1474
1475         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1476                              sizeof(struct compat_xt_entry_target)) {
1477                 duprintf("checking: element %p size %u\n",
1478                          e, e->next_offset);
1479                 return -EINVAL;
1480         }
1481
1482         if (!ip_checkentry(&e->ip))
1483                 return -EINVAL;
1484
1485         ret = xt_compat_check_entry_offsets(e, e->elems,
1486                                             e->target_offset, e->next_offset);
1487         if (ret)
1488                 return ret;
1489
1490         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1491         entry_offset = (void *)e - (void *)base;
1492         j = 0;
1493         xt_ematch_foreach(ematch, e) {
1494                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1495                 if (ret != 0)
1496                         goto release_matches;
1497                 ++j;
1498         }
1499
1500         t = compat_ipt_get_target(e);
1501         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1502                                         t->u.user.revision);
1503         if (IS_ERR(target)) {
1504                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1505                          t->u.user.name);
1506                 ret = PTR_ERR(target);
1507                 goto release_matches;
1508         }
1509         t->u.kernel.target = target;
1510
1511         off += xt_compat_target_offset(target);
1512         *size += off;
1513         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1514         if (ret)
1515                 goto out;
1516
1517         return 0;
1518
1519 out:
1520         module_put(t->u.kernel.target->me);
1521 release_matches:
1522         xt_ematch_foreach(ematch, e) {
1523                 if (j-- == 0)
1524                         break;
1525                 module_put(ematch->u.kernel.match->me);
1526         }
1527         return ret;
1528 }
1529
1530 static void
1531 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1532                             unsigned int *size,
1533                             struct xt_table_info *newinfo, unsigned char *base)
1534 {
1535         struct xt_entry_target *t;
1536         struct xt_target *target;
1537         struct ipt_entry *de;
1538         unsigned int origsize;
1539         int h;
1540         struct xt_entry_match *ematch;
1541
1542         origsize = *size;
1543         de = (struct ipt_entry *)*dstptr;
1544         memcpy(de, e, sizeof(struct ipt_entry));
1545         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1546
1547         *dstptr += sizeof(struct ipt_entry);
1548         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1549
1550         xt_ematch_foreach(ematch, e)
1551                 xt_compat_match_from_user(ematch, dstptr, size);
1552
1553         de->target_offset = e->target_offset - (origsize - *size);
1554         t = compat_ipt_get_target(e);
1555         target = t->u.kernel.target;
1556         xt_compat_target_from_user(t, dstptr, size);
1557
1558         de->next_offset = e->next_offset - (origsize - *size);
1559
1560         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1561                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1562                         newinfo->hook_entry[h] -= origsize - *size;
1563                 if ((unsigned char *)de - base < newinfo->underflow[h])
1564                         newinfo->underflow[h] -= origsize - *size;
1565         }
1566 }
1567
1568 static int
1569 translate_compat_table(struct net *net,
1570                        struct xt_table_info **pinfo,
1571                        void **pentry0,
1572                        const struct compat_ipt_replace *compatr)
1573 {
1574         unsigned int i, j;
1575         struct xt_table_info *newinfo, *info;
1576         void *pos, *entry0, *entry1;
1577         struct compat_ipt_entry *iter0;
1578         struct ipt_replace repl;
1579         unsigned int size;
1580         int ret;
1581
1582         info = *pinfo;
1583         entry0 = *pentry0;
1584         size = compatr->size;
1585         info->number = compatr->num_entries;
1586
1587         duprintf("translate_compat_table: size %u\n", info->size);
1588         j = 0;
1589         xt_compat_lock(AF_INET);
1590         xt_compat_init_offsets(AF_INET, compatr->num_entries);
1591         /* Walk through entries, checking offsets. */
1592         xt_entry_foreach(iter0, entry0, compatr->size) {
1593                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1594                                                         entry0,
1595                                                         entry0 + compatr->size);
1596                 if (ret != 0)
1597                         goto out_unlock;
1598                 ++j;
1599         }
1600
1601         ret = -EINVAL;
1602         if (j != compatr->num_entries) {
1603                 duprintf("translate_compat_table: %u not %u entries\n",
1604                          j, compatr->num_entries);
1605                 goto out_unlock;
1606         }
1607
1608         ret = -ENOMEM;
1609         newinfo = xt_alloc_table_info(size);
1610         if (!newinfo)
1611                 goto out_unlock;
1612
1613         memset(newinfo->entries, 0, size);
1614
1615         newinfo->number = compatr->num_entries;
1616         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1617                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1618                 newinfo->underflow[i] = compatr->underflow[i];
1619         }
1620         entry1 = newinfo->entries;
1621         pos = entry1;
1622         size = compatr->size;
1623         xt_entry_foreach(iter0, entry0, compatr->size)
1624                 compat_copy_entry_from_user(iter0, &pos, &size,
1625                                             newinfo, entry1);
1626
1627         /* all module references in entry0 are now gone.
1628          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1629          * generated by 64bit userspace.
1630          *
1631          * Call standard translate_table() to validate all hook_entrys,
1632          * underflows, check for loops, etc.
1633          */
1634         xt_compat_flush_offsets(AF_INET);
1635         xt_compat_unlock(AF_INET);
1636
1637         memcpy(&repl, compatr, sizeof(*compatr));
1638
1639         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1640                 repl.hook_entry[i] = newinfo->hook_entry[i];
1641                 repl.underflow[i] = newinfo->underflow[i];
1642         }
1643
1644         repl.num_counters = 0;
1645         repl.counters = NULL;
1646         repl.size = newinfo->size;
1647         ret = translate_table(net, newinfo, entry1, &repl);
1648         if (ret)
1649                 goto free_newinfo;
1650
1651         *pinfo = newinfo;
1652         *pentry0 = entry1;
1653         xt_free_table_info(info);
1654         return 0;
1655
1656 free_newinfo:
1657         xt_free_table_info(newinfo);
1658         return ret;
1659 out_unlock:
1660         xt_compat_flush_offsets(AF_INET);
1661         xt_compat_unlock(AF_INET);
1662         xt_entry_foreach(iter0, entry0, compatr->size) {
1663                 if (j-- == 0)
1664                         break;
1665                 compat_release_entry(iter0);
1666         }
1667         return ret;
1668 }
1669
1670 static int
1671 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1672 {
1673         int ret;
1674         struct compat_ipt_replace tmp;
1675         struct xt_table_info *newinfo;
1676         void *loc_cpu_entry;
1677         struct ipt_entry *iter;
1678
1679         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1680                 return -EFAULT;
1681
1682         /* overflow check */
1683         if (tmp.size >= INT_MAX / num_possible_cpus())
1684                 return -ENOMEM;
1685         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1686                 return -ENOMEM;
1687         if (tmp.num_counters == 0)
1688                 return -EINVAL;
1689
1690         tmp.name[sizeof(tmp.name)-1] = 0;
1691
1692         newinfo = xt_alloc_table_info(tmp.size);
1693         if (!newinfo)
1694                 return -ENOMEM;
1695
1696         loc_cpu_entry = newinfo->entries;
1697         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1698                            tmp.size) != 0) {
1699                 ret = -EFAULT;
1700                 goto free_newinfo;
1701         }
1702
1703         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1704         if (ret != 0)
1705                 goto free_newinfo;
1706
1707         duprintf("compat_do_replace: Translated table\n");
1708
1709         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1710                            tmp.num_counters, compat_ptr(tmp.counters));
1711         if (ret)
1712                 goto free_newinfo_untrans;
1713         return 0;
1714
1715  free_newinfo_untrans:
1716         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1717                 cleanup_entry(iter, net);
1718  free_newinfo:
1719         xt_free_table_info(newinfo);
1720         return ret;
1721 }
1722
1723 static int
1724 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1725                       unsigned int len)
1726 {
1727         int ret;
1728
1729         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1730                 return -EPERM;
1731
1732         switch (cmd) {
1733         case IPT_SO_SET_REPLACE:
1734                 ret = compat_do_replace(sock_net(sk), user, len);
1735                 break;
1736
1737         case IPT_SO_SET_ADD_COUNTERS:
1738                 ret = do_add_counters(sock_net(sk), user, len, 1);
1739                 break;
1740
1741         default:
1742                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1743                 ret = -EINVAL;
1744         }
1745
1746         return ret;
1747 }
1748
1749 struct compat_ipt_get_entries {
1750         char name[XT_TABLE_MAXNAMELEN];
1751         compat_uint_t size;
1752         struct compat_ipt_entry entrytable[0];
1753 };
1754
1755 static int
1756 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1757                             void __user *userptr)
1758 {
1759         struct xt_counters *counters;
1760         const struct xt_table_info *private = table->private;
1761         void __user *pos;
1762         unsigned int size;
1763         int ret = 0;
1764         unsigned int i = 0;
1765         struct ipt_entry *iter;
1766
1767         counters = alloc_counters(table);
1768         if (IS_ERR(counters))
1769                 return PTR_ERR(counters);
1770
1771         pos = userptr;
1772         size = total_size;
1773         xt_entry_foreach(iter, private->entries, total_size) {
1774                 ret = compat_copy_entry_to_user(iter, &pos,
1775                                                 &size, counters, i++);
1776                 if (ret != 0)
1777                         break;
1778         }
1779
1780         vfree(counters);
1781         return ret;
1782 }
1783
1784 static int
1785 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1786                    int *len)
1787 {
1788         int ret;
1789         struct compat_ipt_get_entries get;
1790         struct xt_table *t;
1791
1792         if (*len < sizeof(get)) {
1793                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1794                 return -EINVAL;
1795         }
1796
1797         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1798                 return -EFAULT;
1799
1800         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1801                 duprintf("compat_get_entries: %u != %zu\n",
1802                          *len, sizeof(get) + get.size);
1803                 return -EINVAL;
1804         }
1805         get.name[sizeof(get.name) - 1] = '\0';
1806
1807         xt_compat_lock(AF_INET);
1808         t = xt_find_table_lock(net, AF_INET, get.name);
1809         if (!IS_ERR_OR_NULL(t)) {
1810                 const struct xt_table_info *private = t->private;
1811                 struct xt_table_info info;
1812                 duprintf("t->private->number = %u\n", private->number);
1813                 ret = compat_table_info(private, &info);
1814                 if (!ret && get.size == info.size) {
1815                         ret = compat_copy_entries_to_user(private->size,
1816                                                           t, uptr->entrytable);
1817                 } else if (!ret) {
1818                         duprintf("compat_get_entries: I've got %u not %u!\n",
1819                                  private->size, get.size);
1820                         ret = -EAGAIN;
1821                 }
1822                 xt_compat_flush_offsets(AF_INET);
1823                 module_put(t->me);
1824                 xt_table_unlock(t);
1825         } else
1826                 ret = t ? PTR_ERR(t) : -ENOENT;
1827
1828         xt_compat_unlock(AF_INET);
1829         return ret;
1830 }
1831
1832 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1833
1834 static int
1835 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1836 {
1837         int ret;
1838
1839         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1840                 return -EPERM;
1841
1842         switch (cmd) {
1843         case IPT_SO_GET_INFO:
1844                 ret = get_info(sock_net(sk), user, len, 1);
1845                 break;
1846         case IPT_SO_GET_ENTRIES:
1847                 ret = compat_get_entries(sock_net(sk), user, len);
1848                 break;
1849         default:
1850                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1851         }
1852         return ret;
1853 }
1854 #endif
1855
1856 static int
1857 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1858 {
1859         int ret;
1860
1861         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1862                 return -EPERM;
1863
1864         switch (cmd) {
1865         case IPT_SO_SET_REPLACE:
1866                 ret = do_replace(sock_net(sk), user, len);
1867                 break;
1868
1869         case IPT_SO_SET_ADD_COUNTERS:
1870                 ret = do_add_counters(sock_net(sk), user, len, 0);
1871                 break;
1872
1873         default:
1874                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1875                 ret = -EINVAL;
1876         }
1877
1878         return ret;
1879 }
1880
1881 static int
1882 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1883 {
1884         int ret;
1885
1886         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1887                 return -EPERM;
1888
1889         switch (cmd) {
1890         case IPT_SO_GET_INFO:
1891                 ret = get_info(sock_net(sk), user, len, 0);
1892                 break;
1893
1894         case IPT_SO_GET_ENTRIES:
1895                 ret = get_entries(sock_net(sk), user, len);
1896                 break;
1897
1898         case IPT_SO_GET_REVISION_MATCH:
1899         case IPT_SO_GET_REVISION_TARGET: {
1900                 struct xt_get_revision rev;
1901                 int target;
1902
1903                 if (*len != sizeof(rev)) {
1904                         ret = -EINVAL;
1905                         break;
1906                 }
1907                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1908                         ret = -EFAULT;
1909                         break;
1910                 }
1911                 rev.name[sizeof(rev.name)-1] = 0;
1912
1913                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1914                         target = 1;
1915                 else
1916                         target = 0;
1917
1918                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1919                                                          rev.revision,
1920                                                          target, &ret),
1921                                         "ipt_%s", rev.name);
1922                 break;
1923         }
1924
1925         default:
1926                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
1927                 ret = -EINVAL;
1928         }
1929
1930         return ret;
1931 }
1932
1933 struct xt_table *ipt_register_table(struct net *net,
1934                                     const struct xt_table *table,
1935                                     const struct ipt_replace *repl)
1936 {
1937         int ret;
1938         struct xt_table_info *newinfo;
1939         struct xt_table_info bootstrap = {0};
1940         void *loc_cpu_entry;
1941         struct xt_table *new_table;
1942
1943         newinfo = xt_alloc_table_info(repl->size);
1944         if (!newinfo) {
1945                 ret = -ENOMEM;
1946                 goto out;
1947         }
1948
1949         loc_cpu_entry = newinfo->entries;
1950         memcpy(loc_cpu_entry, repl->entries, repl->size);
1951
1952         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1953         if (ret != 0)
1954                 goto out_free;
1955
1956         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1957         if (IS_ERR(new_table)) {
1958                 ret = PTR_ERR(new_table);
1959                 goto out_free;
1960         }
1961
1962         return new_table;
1963
1964 out_free:
1965         xt_free_table_info(newinfo);
1966 out:
1967         return ERR_PTR(ret);
1968 }
1969
1970 void ipt_unregister_table(struct net *net, struct xt_table *table)
1971 {
1972         struct xt_table_info *private;
1973         void *loc_cpu_entry;
1974         struct module *table_owner = table->me;
1975         struct ipt_entry *iter;
1976
1977         private = xt_unregister_table(table);
1978
1979         /* Decrease module usage counts and free resources */
1980         loc_cpu_entry = private->entries;
1981         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1982                 cleanup_entry(iter, net);
1983         if (private->number > private->initial_entries)
1984                 module_put(table_owner);
1985         xt_free_table_info(private);
1986 }
1987
1988 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1989 static inline bool
1990 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1991                      u_int8_t type, u_int8_t code,
1992                      bool invert)
1993 {
1994         return ((test_type == 0xFF) ||
1995                 (type == test_type && code >= min_code && code <= max_code))
1996                 ^ invert;
1997 }
1998
1999 static bool
2000 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2001 {
2002         const struct icmphdr *ic;
2003         struct icmphdr _icmph;
2004         const struct ipt_icmp *icmpinfo = par->matchinfo;
2005
2006         /* Must not be a fragment. */
2007         if (par->fragoff != 0)
2008                 return false;
2009
2010         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2011         if (ic == NULL) {
2012                 /* We've been asked to examine this packet, and we
2013                  * can't.  Hence, no choice but to drop.
2014                  */
2015                 duprintf("Dropping evil ICMP tinygram.\n");
2016                 par->hotdrop = true;
2017                 return false;
2018         }
2019
2020         return icmp_type_code_match(icmpinfo->type,
2021                                     icmpinfo->code[0],
2022                                     icmpinfo->code[1],
2023                                     ic->type, ic->code,
2024                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2025 }
2026
2027 static int icmp_checkentry(const struct xt_mtchk_param *par)
2028 {
2029         const struct ipt_icmp *icmpinfo = par->matchinfo;
2030
2031         /* Must specify no unknown invflags */
2032         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2033 }
2034
2035 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2036         {
2037                 .name             = XT_STANDARD_TARGET,
2038                 .targetsize       = sizeof(int),
2039                 .family           = NFPROTO_IPV4,
2040 #ifdef CONFIG_COMPAT
2041                 .compatsize       = sizeof(compat_int_t),
2042                 .compat_from_user = compat_standard_from_user,
2043                 .compat_to_user   = compat_standard_to_user,
2044 #endif
2045         },
2046         {
2047                 .name             = XT_ERROR_TARGET,
2048                 .target           = ipt_error,
2049                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2050                 .family           = NFPROTO_IPV4,
2051         },
2052 };
2053
2054 static struct nf_sockopt_ops ipt_sockopts = {
2055         .pf             = PF_INET,
2056         .set_optmin     = IPT_BASE_CTL,
2057         .set_optmax     = IPT_SO_SET_MAX+1,
2058         .set            = do_ipt_set_ctl,
2059 #ifdef CONFIG_COMPAT
2060         .compat_set     = compat_do_ipt_set_ctl,
2061 #endif
2062         .get_optmin     = IPT_BASE_CTL,
2063         .get_optmax     = IPT_SO_GET_MAX+1,
2064         .get            = do_ipt_get_ctl,
2065 #ifdef CONFIG_COMPAT
2066         .compat_get     = compat_do_ipt_get_ctl,
2067 #endif
2068         .owner          = THIS_MODULE,
2069 };
2070
2071 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2072         {
2073                 .name       = "icmp",
2074                 .match      = icmp_match,
2075                 .matchsize  = sizeof(struct ipt_icmp),
2076                 .checkentry = icmp_checkentry,
2077                 .proto      = IPPROTO_ICMP,
2078                 .family     = NFPROTO_IPV4,
2079                 .me         = THIS_MODULE,
2080         },
2081 };
2082
2083 static int __net_init ip_tables_net_init(struct net *net)
2084 {
2085         return xt_proto_init(net, NFPROTO_IPV4);
2086 }
2087
2088 static void __net_exit ip_tables_net_exit(struct net *net)
2089 {
2090         xt_proto_fini(net, NFPROTO_IPV4);
2091 }
2092
2093 static struct pernet_operations ip_tables_net_ops = {
2094         .init = ip_tables_net_init,
2095         .exit = ip_tables_net_exit,
2096 };
2097
2098 static int __init ip_tables_init(void)
2099 {
2100         int ret;
2101
2102         ret = register_pernet_subsys(&ip_tables_net_ops);
2103         if (ret < 0)
2104                 goto err1;
2105
2106         /* No one else will be downing sem now, so we won't sleep */
2107         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2108         if (ret < 0)
2109                 goto err2;
2110         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2111         if (ret < 0)
2112                 goto err4;
2113
2114         /* Register setsockopt */
2115         ret = nf_register_sockopt(&ipt_sockopts);
2116         if (ret < 0)
2117                 goto err5;
2118
2119         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2120         return 0;
2121
2122 err5:
2123         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2124 err4:
2125         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2126 err2:
2127         unregister_pernet_subsys(&ip_tables_net_ops);
2128 err1:
2129         return ret;
2130 }
2131
2132 static void __exit ip_tables_fini(void)
2133 {
2134         nf_unregister_sockopt(&ipt_sockopts);
2135
2136         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2137         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2138         unregister_pernet_subsys(&ip_tables_net_ops);
2139 }
2140
2141 EXPORT_SYMBOL(ipt_register_table);
2142 EXPORT_SYMBOL(ipt_unregister_table);
2143 EXPORT_SYMBOL(ipt_do_table);
2144 module_init(ip_tables_init);
2145 module_exit(ip_tables_fini);