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 <linux/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 void *ipt_alloc_initial_table(const struct xt_table *info)
  39 {
  40         return xt_alloc_initial_table(ipt, IPT);
  41 }
  42 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  43
  44 /* Returns whether matches rule or not. */
  45 /* Performance critical - called for every packet */
  46 static inline bool
  47 ip_packet_match(const struct iphdr *ip,
  48                 const char *indev,
  49                 const char *outdev,
  50                 const struct ipt_ip *ipinfo,
  51                 int isfrag)
  52 {
  53         unsigned long ret;
  54
  55         if (NF_INVF(ipinfo, IPT_INV_SRCIP,
  56                     (ip->saddr & ipinfo->smsk.s_addr) != ipinfo->src.s_addr) ||
  57             NF_INVF(ipinfo, IPT_INV_DSTIP,
  58                     (ip->daddr & ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr))
  59                 return false;
  60
  61         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
  62
  63         if (NF_INVF(ipinfo, IPT_INV_VIA_IN, ret != 0))
  64                 return false;
  65
  66         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
  67
  68         if (NF_INVF(ipinfo, IPT_INV_VIA_OUT, ret != 0))
  69                 return false;
  70
  71         /* Check specific protocol */
  72         if (ipinfo->proto &&
  73             NF_INVF(ipinfo, IPT_INV_PROTO, ip->protocol != ipinfo->proto))
  74                 return false;
  75
  76         /* If we have a fragment rule but the packet is not a fragment
  77          * then we return zero */
  78         if (NF_INVF(ipinfo, IPT_INV_FRAG,
  79                     (ipinfo->flags & IPT_F_FRAG) && !isfrag))
  80                 return false;
  81
  82         return true;
  83 }
  84
  85 static bool
  86 ip_checkentry(const struct ipt_ip *ip)
  87 {
  88         if (ip->flags & ~IPT_F_MASK)
  89                 return false;
  90         if (ip->invflags & ~IPT_INV_MASK)
  91                 return false;
  92         return true;
  93 }
  94
  95 static unsigned int
  96 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
  97 {
  98         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
  99
 100         return NF_DROP;
 101 }
 102
 103 /* Performance critical */
 104 static inline struct ipt_entry *
 105 get_entry(const void *base, unsigned int offset)
 106 {
 107         return (struct ipt_entry *)(base + offset);
 108 }
 109
 110 /* All zeroes == unconditional rule. */
 111 /* Mildly perf critical (only if packet tracing is on) */
 112 static inline bool unconditional(const struct ipt_entry *e)
 113 {
 114         static const struct ipt_ip uncond;
 115
 116         return e->target_offset == sizeof(struct ipt_entry) &&
 117                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
 118 }
 119
 120 /* for const-correctness */
 121 static inline const struct xt_entry_target *
 122 ipt_get_target_c(const struct ipt_entry *e)
 123 {
 124         return ipt_get_target((struct ipt_entry *)e);
 125 }
 126
 127 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 128 static const char *const hooknames[] = {
 129         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 130         [NF_INET_LOCAL_IN]              = "INPUT",
 131         [NF_INET_FORWARD]               = "FORWARD",
 132         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 133         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 134 };
 135
 136 enum nf_ip_trace_comments {
 137         NF_IP_TRACE_COMMENT_RULE,
 138         NF_IP_TRACE_COMMENT_RETURN,
 139         NF_IP_TRACE_COMMENT_POLICY,
 140 };
 141
 142 static const char *const comments[] = {
 143         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 144         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 145         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 146 };
 147
 148 static const struct nf_loginfo trace_loginfo = {
 149         .type = NF_LOG_TYPE_LOG,
 150         .u = {
 151                 .log = {
 152                         .level = 4,
 153                         .logflags = NF_LOG_DEFAULT_MASK,
 154                 },
 155         },
 156 };
 157
 158 /* Mildly perf critical (only if packet tracing is on) */
 159 static inline int
 160 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 161                       const char *hookname, const char **chainname,
 162                       const char **comment, unsigned int *rulenum)
 163 {
 164         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 165
 166         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 167                 /* Head of user chain: ERROR target with chainname */
 168                 *chainname = t->target.data;
 169                 (*rulenum) = 0;
 170         } else if (s == e) {
 171                 (*rulenum)++;
 172
 173                 if (unconditional(s) &&
 174                     strcmp(t->target.u.kernel.target->name,
 175                            XT_STANDARD_TARGET) == 0 &&
 176                    t->verdict < 0) {
 177                         /* Tail of chains: STANDARD target (return/policy) */
 178                         *comment = *chainname == hookname
 179                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 180                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 181                 }
 182                 return 1;
 183         } else
 184                 (*rulenum)++;
 185
 186         return 0;
 187 }
 188
 189 static void trace_packet(struct net *net,
 190                          const struct sk_buff *skb,
 191                          unsigned int hook,
 192                          const struct net_device *in,
 193                          const struct net_device *out,
 194                          const char *tablename,
 195                          const struct xt_table_info *private,
 196                          const struct ipt_entry *e)
 197 {
 198         const struct ipt_entry *root;
 199         const char *hookname, *chainname, *comment;
 200         const struct ipt_entry *iter;
 201         unsigned int rulenum = 0;
 202
 203         root = get_entry(private->entries, private->hook_entry[hook]);
 204
 205         hookname = chainname = hooknames[hook];
 206         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 207
 208         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 209                 if (get_chainname_rulenum(iter, e, hookname,
 210                     &chainname, &comment, &rulenum) != 0)
 211                         break;
 212
 213         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 214                      "TRACE: %s:%s:%s:%u ",
 215                      tablename, chainname, comment, rulenum);
 216 }
 217 #endif
 218
 219 static inline
 220 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 221 {
 222         return (void *)entry + entry->next_offset;
 223 }
 224
 225 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 226 unsigned int
 227 ipt_do_table(struct sk_buff *skb,
 228              const struct nf_hook_state *state,
 229              struct xt_table *table)
 230 {
 231         unsigned int hook = state->hook;
 232         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 233         const struct iphdr *ip;
 234         /* Initializing verdict to NF_DROP keeps gcc happy. */
 235         unsigned int verdict = NF_DROP;
 236         const char *indev, *outdev;
 237         const void *table_base;
 238         struct ipt_entry *e, **jumpstack;
 239         unsigned int stackidx, cpu;
 240         const struct xt_table_info *private;
 241         struct xt_action_param acpar;
 242         unsigned int addend;
 243
 244         /* Initialization */
 245         stackidx = 0;
 246         ip = ip_hdr(skb);
 247         indev = state->in ? state->in->name : nulldevname;
 248         outdev = state->out ? state->out->name : nulldevname;
 249         /* We handle fragments by dealing with the first fragment as
 250          * if it was a normal packet.  All other fragments are treated
 251          * normally, except that they will NEVER match rules that ask
 252          * things we don't know, ie. tcp syn flag or ports).  If the
 253          * rule is also a fragment-specific rule, non-fragments won't
 254          * match it. */
 255         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 256         acpar.thoff   = ip_hdrlen(skb);
 257         acpar.hotdrop = false;
 258         acpar.state   = state;
 259
 260         WARN_ON(!(table->valid_hooks & (1 << hook)));
 261         local_bh_disable();
 262         addend = xt_write_recseq_begin();
 263         private = table->private;
 264         cpu        = smp_processor_id();
 265         /*
 266          * Ensure we load private-> members after we've fetched the base
 267          * pointer.
 268          */
 269         smp_read_barrier_depends();
 270         table_base = private->entries;
 271         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 272
 273         /* Switch to alternate jumpstack if we're being invoked via TEE.
 274          * TEE issues XT_CONTINUE verdict on original skb so we must not
 275          * clobber the jumpstack.
 276          *
 277          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 278          * but it is no problem since absolute verdict is issued by these.
 279          */
 280         if (static_key_false(&xt_tee_enabled))
 281                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 282
 283         e = get_entry(table_base, private->hook_entry[hook]);
 284
 285         do {
 286                 const struct xt_entry_target *t;
 287                 const struct xt_entry_match *ematch;
 288                 struct xt_counters *counter;
 289
 290                 WARN_ON(!e);
 291                 if (!ip_packet_match(ip, indev, outdev,
 292                     &e->ip, acpar.fragoff)) {
 293  no_match:
 294                         e = ipt_next_entry(e);
 295                         continue;
 296                 }
 297
 298                 xt_ematch_foreach(ematch, e) {
 299                         acpar.match     = ematch->u.kernel.match;
 300                         acpar.matchinfo = ematch->data;
 301                         if (!acpar.match->match(skb, &acpar))
 302                                 goto no_match;
 303                 }
 304
 305                 counter = xt_get_this_cpu_counter(&e->counters);
 306                 ADD_COUNTER(*counter, skb->len, 1);
 307
 308                 t = ipt_get_target(e);
 309                 WARN_ON(!t->u.kernel.target);
 310
 311 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 312                 /* The packet is traced: log it */
 313                 if (unlikely(skb->nf_trace))
 314                         trace_packet(state->net, skb, hook, state->in,
 315                                      state->out, table->name, private, e);
 316 #endif
 317                 /* Standard target? */
 318                 if (!t->u.kernel.target->target) {
 319                         int v;
 320
 321                         v = ((struct xt_standard_target *)t)->verdict;
 322                         if (v < 0) {
 323                                 /* Pop from stack? */
 324                                 if (v != XT_RETURN) {
 325                                         verdict = (unsigned int)(-v) - 1;
 326                                         break;
 327                                 }
 328                                 if (stackidx == 0) {
 329                                         e = get_entry(table_base,
 330                                             private->underflow[hook]);
 331                                 } else {
 332                                         e = jumpstack[--stackidx];
 333                                         e = ipt_next_entry(e);
 334                                 }
 335                                 continue;
 336                         }
 337                         if (table_base + v != ipt_next_entry(e) &&
 338                             !(e->ip.flags & IPT_F_GOTO)) {
 339                                 if (unlikely(stackidx >= private->stacksize)) {
 340                                         verdict = NF_DROP;
 341                                         break;
 342                                 }
 343                                 jumpstack[stackidx++] = e;
 344                         }
 345
 346                         e = get_entry(table_base, v);
 347                         continue;
 348                 }
 349
 350                 acpar.target   = t->u.kernel.target;
 351                 acpar.targinfo = t->data;
 352
 353                 verdict = t->u.kernel.target->target(skb, &acpar);
 354                 if (verdict == XT_CONTINUE) {
 355                         /* Target might have changed stuff. */
 356                         ip = ip_hdr(skb);
 357                         e = ipt_next_entry(e);
 358                 } else {
 359                         /* Verdict */
 360                         break;
 361                 }
 362         } while (!acpar.hotdrop);
 363
 364         xt_write_recseq_end(addend);
 365         local_bh_enable();
 366
 367         if (acpar.hotdrop)
 368                 return NF_DROP;
 369         else return verdict;
 370 }
 371
 372 /* Figures out from what hook each rule can be called: returns 0 if
 373    there are loops.  Puts hook bitmask in comefrom. */
 374 static int
 375 mark_source_chains(const struct xt_table_info *newinfo,
 376                    unsigned int valid_hooks, void *entry0,
 377                    unsigned int *offsets)
 378 {
 379         unsigned int hook;
 380
 381         /* No recursion; use packet counter to save back ptrs (reset
 382            to 0 as we leave), and comefrom to save source hook bitmask */
 383         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 384                 unsigned int pos = newinfo->hook_entry[hook];
 385                 struct ipt_entry *e = entry0 + pos;
 386
 387                 if (!(valid_hooks & (1 << hook)))
 388                         continue;
 389
 390                 /* Set initial back pointer. */
 391                 e->counters.pcnt = pos;
 392
 393                 for (;;) {
 394                         const struct xt_standard_target *t
 395                                 = (void *)ipt_get_target_c(e);
 396                         int visited = e->comefrom & (1 << hook);
 397
 398                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 399                                 return 0;
 400
 401                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 402
 403                         /* Unconditional return/END. */
 404                         if ((unconditional(e) &&
 405                              (strcmp(t->target.u.user.name,
 406                                      XT_STANDARD_TARGET) == 0) &&
 407                              t->verdict < 0) || visited) {
 408                                 unsigned int oldpos, size;
 409
 410                                 if ((strcmp(t->target.u.user.name,
 411                                             XT_STANDARD_TARGET) == 0) &&
 412                                     t->verdict < -NF_MAX_VERDICT - 1)
 413                                         return 0;
 414
 415                                 /* Return: backtrack through the last
 416                                    big jump. */
 417                                 do {
 418                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 419                                         oldpos = pos;
 420                                         pos = e->counters.pcnt;
 421                                         e->counters.pcnt = 0;
 422
 423                                         /* We're at the start. */
 424                                         if (pos == oldpos)
 425                                                 goto next;
 426
 427                                         e = entry0 + pos;
 428                                 } while (oldpos == pos + e->next_offset);
 429
 430                                 /* Move along one */
 431                                 size = e->next_offset;
 432                                 e = entry0 + pos + size;
 433                                 if (pos + size >= newinfo->size)
 434                                         return 0;
 435                                 e->counters.pcnt = pos;
 436                                 pos += size;
 437                         } else {
 438                                 int newpos = t->verdict;
 439
 440                                 if (strcmp(t->target.u.user.name,
 441                                            XT_STANDARD_TARGET) == 0 &&
 442                                     newpos >= 0) {
 443                                         /* This a jump; chase it. */
 444                                         if (!xt_find_jump_offset(offsets, newpos,
 445                                                                  newinfo->number))
 446                                                 return 0;
 447                                         e = entry0 + newpos;
 448                                 } else {
 449                                         /* ... this is a fallthru */
 450                                         newpos = pos + e->next_offset;
 451                                         if (newpos >= newinfo->size)
 452                                                 return 0;
 453                                 }
 454                                 e = entry0 + newpos;
 455                                 e->counters.pcnt = pos;
 456                                 pos = newpos;
 457                         }
 458                 }
 459 next:           ;
 460         }
 461         return 1;
 462 }
 463
 464 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 465 {
 466         struct xt_mtdtor_param par;
 467
 468         par.net       = net;
 469         par.match     = m->u.kernel.match;
 470         par.matchinfo = m->data;
 471         par.family    = NFPROTO_IPV4;
 472         if (par.match->destroy != NULL)
 473                 par.match->destroy(&par);
 474         module_put(par.match->me);
 475 }
 476
 477 static int
 478 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 479 {
 480         const struct ipt_ip *ip = par->entryinfo;
 481
 482         par->match     = m->u.kernel.match;
 483         par->matchinfo = m->data;
 484
 485         return xt_check_match(par, m->u.match_size - sizeof(*m),
 486                               ip->proto, ip->invflags & IPT_INV_PROTO);
 487 }
 488
 489 static int
 490 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 491 {
 492         struct xt_match *match;
 493         int ret;
 494
 495         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 496                                       m->u.user.revision);
 497         if (IS_ERR(match))
 498                 return PTR_ERR(match);
 499         m->u.kernel.match = match;
 500
 501         ret = check_match(m, par);
 502         if (ret)
 503                 goto err;
 504
 505         return 0;
 506 err:
 507         module_put(m->u.kernel.match->me);
 508         return ret;
 509 }
 510
 511 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 512 {
 513         struct xt_entry_target *t = ipt_get_target(e);
 514         struct xt_tgchk_param par = {
 515                 .net       = net,
 516                 .table     = name,
 517                 .entryinfo = e,
 518                 .target    = t->u.kernel.target,
 519                 .targinfo  = t->data,
 520                 .hook_mask = e->comefrom,
 521                 .family    = NFPROTO_IPV4,
 522         };
 523
 524         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 525                                e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 526 }
 527
 528 static int
 529 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 530                  unsigned int size,
 531                  struct xt_percpu_counter_alloc_state *alloc_state)
 532 {
 533         struct xt_entry_target *t;
 534         struct xt_target *target;
 535         int ret;
 536         unsigned int j;
 537         struct xt_mtchk_param mtpar;
 538         struct xt_entry_match *ematch;
 539
 540         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 541                 return -ENOMEM;
 542
 543         j = 0;
 544         memset(&mtpar, 0, sizeof(mtpar));
 545         mtpar.net       = net;
 546         mtpar.table     = name;
 547         mtpar.entryinfo = &e->ip;
 548         mtpar.hook_mask = e->comefrom;
 549         mtpar.family    = NFPROTO_IPV4;
 550         xt_ematch_foreach(ematch, e) {
 551                 ret = find_check_match(ematch, &mtpar);
 552                 if (ret != 0)
 553                         goto cleanup_matches;
 554                 ++j;
 555         }
 556
 557         t = ipt_get_target(e);
 558         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 559                                         t->u.user.revision);
 560         if (IS_ERR(target)) {
 561                 ret = PTR_ERR(target);
 562                 goto cleanup_matches;
 563         }
 564         t->u.kernel.target = target;
 565
 566         ret = check_target(e, net, name);
 567         if (ret)
 568                 goto err;
 569
 570         return 0;
 571  err:
 572         module_put(t->u.kernel.target->me);
 573  cleanup_matches:
 574         xt_ematch_foreach(ematch, e) {
 575                 if (j-- == 0)
 576                         break;
 577                 cleanup_match(ematch, net);
 578         }
 579
 580         xt_percpu_counter_free(&e->counters);
 581
 582         return ret;
 583 }
 584
 585 static bool check_underflow(const struct ipt_entry *e)
 586 {
 587         const struct xt_entry_target *t;
 588         unsigned int verdict;
 589
 590         if (!unconditional(e))
 591                 return false;
 592         t = ipt_get_target_c(e);
 593         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 594                 return false;
 595         verdict = ((struct xt_standard_target *)t)->verdict;
 596         verdict = -verdict - 1;
 597         return verdict == NF_DROP || verdict == NF_ACCEPT;
 598 }
 599
 600 static int
 601 check_entry_size_and_hooks(struct ipt_entry *e,
 602                            struct xt_table_info *newinfo,
 603                            const unsigned char *base,
 604                            const unsigned char *limit,
 605                            const unsigned int *hook_entries,
 606                            const unsigned int *underflows,
 607                            unsigned int valid_hooks)
 608 {
 609         unsigned int h;
 610         int err;
 611
 612         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 613             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 614             (unsigned char *)e + e->next_offset > limit)
 615                 return -EINVAL;
 616
 617         if (e->next_offset
 618             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
 619                 return -EINVAL;
 620
 621         if (!ip_checkentry(&e->ip))
 622                 return -EINVAL;
 623
 624         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 625                                      e->next_offset);
 626         if (err)
 627                 return err;
 628
 629         /* Check hooks & underflows */
 630         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 631                 if (!(valid_hooks & (1 << h)))
 632                         continue;
 633                 if ((unsigned char *)e - base == hook_entries[h])
 634                         newinfo->hook_entry[h] = hook_entries[h];
 635                 if ((unsigned char *)e - base == underflows[h]) {
 636                         if (!check_underflow(e))
 637                                 return -EINVAL;
 638
 639                         newinfo->underflow[h] = underflows[h];
 640                 }
 641         }
 642
 643         /* Clear counters and comefrom */
 644         e->counters = ((struct xt_counters) { 0, 0 });
 645         e->comefrom = 0;
 646         return 0;
 647 }
 648
 649 static void
 650 cleanup_entry(struct ipt_entry *e, struct net *net)
 651 {
 652         struct xt_tgdtor_param par;
 653         struct xt_entry_target *t;
 654         struct xt_entry_match *ematch;
 655
 656         /* Cleanup all matches */
 657         xt_ematch_foreach(ematch, e)
 658                 cleanup_match(ematch, net);
 659         t = ipt_get_target(e);
 660
 661         par.net      = net;
 662         par.target   = t->u.kernel.target;
 663         par.targinfo = t->data;
 664         par.family   = NFPROTO_IPV4;
 665         if (par.target->destroy != NULL)
 666                 par.target->destroy(&par);
 667         module_put(par.target->me);
 668         xt_percpu_counter_free(&e->counters);
 669 }
 670
 671 /* Checks and translates the user-supplied table segment (held in
 672    newinfo) */
 673 static int
 674 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 675                 const struct ipt_replace *repl)
 676 {
 677         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 678         struct ipt_entry *iter;
 679         unsigned int *offsets;
 680         unsigned int i;
 681         int ret = 0;
 682
 683         newinfo->size = repl->size;
 684         newinfo->number = repl->num_entries;
 685
 686         /* Init all hooks to impossible value. */
 687         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 688                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 689                 newinfo->underflow[i] = 0xFFFFFFFF;
 690         }
 691
 692         offsets = xt_alloc_entry_offsets(newinfo->number);
 693         if (!offsets)
 694                 return -ENOMEM;
 695         i = 0;
 696         /* Walk through entries, checking offsets. */
 697         xt_entry_foreach(iter, entry0, newinfo->size) {
 698                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 699                                                  entry0 + repl->size,
 700                                                  repl->hook_entry,
 701                                                  repl->underflow,
 702                                                  repl->valid_hooks);
 703                 if (ret != 0)
 704                         goto out_free;
 705                 if (i < repl->num_entries)
 706                         offsets[i] = (void *)iter - entry0;
 707                 ++i;
 708                 if (strcmp(ipt_get_target(iter)->u.user.name,
 709                     XT_ERROR_TARGET) == 0)
 710                         ++newinfo->stacksize;
 711         }
 712
 713         ret = -EINVAL;
 714         if (i != repl->num_entries)
 715                 goto out_free;
 716
 717         /* Check hooks all assigned */
 718         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 719                 /* Only hooks which are valid */
 720                 if (!(repl->valid_hooks & (1 << i)))
 721                         continue;
 722                 if (newinfo->hook_entry[i] == 0xFFFFFFFF)
 723                         goto out_free;
 724                 if (newinfo->underflow[i] == 0xFFFFFFFF)
 725                         goto out_free;
 726         }
 727
 728         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 729                 ret = -ELOOP;
 730                 goto out_free;
 731         }
 732         kvfree(offsets);
 733
 734         /* Finally, each sanity check must pass */
 735         i = 0;
 736         xt_entry_foreach(iter, entry0, newinfo->size) {
 737                 ret = find_check_entry(iter, net, repl->name, repl->size,
 738                                        &alloc_state);
 739                 if (ret != 0)
 740                         break;
 741                 ++i;
 742         }
 743
 744         if (ret != 0) {
 745                 xt_entry_foreach(iter, entry0, newinfo->size) {
 746                         if (i-- == 0)
 747                                 break;
 748                         cleanup_entry(iter, net);
 749                 }
 750                 return ret;
 751         }
 752
 753         return ret;
 754  out_free:
 755         kvfree(offsets);
 756         return ret;
 757 }
 758
 759 static void
 760 get_counters(const struct xt_table_info *t,
 761              struct xt_counters counters[])
 762 {
 763         struct ipt_entry *iter;
 764         unsigned int cpu;
 765         unsigned int i;
 766
 767         for_each_possible_cpu(cpu) {
 768                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 769
 770                 i = 0;
 771                 xt_entry_foreach(iter, t->entries, t->size) {
 772                         struct xt_counters *tmp;
 773                         u64 bcnt, pcnt;
 774                         unsigned int start;
 775
 776                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 777                         do {
 778                                 start = read_seqcount_begin(s);
 779                                 bcnt = tmp->bcnt;
 780                                 pcnt = tmp->pcnt;
 781                         } while (read_seqcount_retry(s, start));
 782
 783                         ADD_COUNTER(counters[i], bcnt, pcnt);
 784                         ++i; /* macro does multi eval of i */
 785                         cond_resched();
 786                 }
 787         }
 788 }
 789
 790 static struct xt_counters *alloc_counters(const struct xt_table *table)
 791 {
 792         unsigned int countersize;
 793         struct xt_counters *counters;
 794         const struct xt_table_info *private = table->private;
 795
 796         /* We need atomic snapshot of counters: rest doesn't change
 797            (other than comefrom, which userspace doesn't care
 798            about). */
 799         countersize = sizeof(struct xt_counters) * private->number;
 800         counters = vzalloc(countersize);
 801
 802         if (counters == NULL)
 803                 return ERR_PTR(-ENOMEM);
 804
 805         get_counters(private, counters);
 806
 807         return counters;
 808 }
 809
 810 static int
 811 copy_entries_to_user(unsigned int total_size,
 812                      const struct xt_table *table,
 813                      void __user *userptr)
 814 {
 815         unsigned int off, num;
 816         const struct ipt_entry *e;
 817         struct xt_counters *counters;
 818         const struct xt_table_info *private = table->private;
 819         int ret = 0;
 820         const void *loc_cpu_entry;
 821
 822         counters = alloc_counters(table);
 823         if (IS_ERR(counters))
 824                 return PTR_ERR(counters);
 825
 826         loc_cpu_entry = private->entries;
 827
 828         /* FIXME: use iterator macros --RR */
 829         /* ... then go back and fix counters and names */
 830         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 831                 unsigned int i;
 832                 const struct xt_entry_match *m;
 833                 const struct xt_entry_target *t;
 834
 835                 e = loc_cpu_entry + off;
 836                 if (copy_to_user(userptr + off, e, sizeof(*e))) {
 837                         ret = -EFAULT;
 838                         goto free_counters;
 839                 }
 840                 if (copy_to_user(userptr + off
 841                                  + offsetof(struct ipt_entry, counters),
 842                                  &counters[num],
 843                                  sizeof(counters[num])) != 0) {
 844                         ret = -EFAULT;
 845                         goto free_counters;
 846                 }
 847
 848                 for (i = sizeof(struct ipt_entry);
 849                      i < e->target_offset;
 850                      i += m->u.match_size) {
 851                         m = (void *)e + i;
 852
 853                         if (xt_match_to_user(m, userptr + off + i)) {
 854                                 ret = -EFAULT;
 855                                 goto free_counters;
 856                         }
 857                 }
 858
 859                 t = ipt_get_target_c(e);
 860                 if (xt_target_to_user(t, userptr + off + e->target_offset)) {
 861                         ret = -EFAULT;
 862                         goto free_counters;
 863                 }
 864         }
 865
 866  free_counters:
 867         vfree(counters);
 868         return ret;
 869 }
 870
 871 #ifdef CONFIG_COMPAT
 872 static void compat_standard_from_user(void *dst, const void *src)
 873 {
 874         int v = *(compat_int_t *)src;
 875
 876         if (v > 0)
 877                 v += xt_compat_calc_jump(AF_INET, v);
 878         memcpy(dst, &v, sizeof(v));
 879 }
 880
 881 static int compat_standard_to_user(void __user *dst, const void *src)
 882 {
 883         compat_int_t cv = *(int *)src;
 884
 885         if (cv > 0)
 886                 cv -= xt_compat_calc_jump(AF_INET, cv);
 887         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 888 }
 889
 890 static int compat_calc_entry(const struct ipt_entry *e,
 891                              const struct xt_table_info *info,
 892                              const void *base, struct xt_table_info *newinfo)
 893 {
 894         const struct xt_entry_match *ematch;
 895         const struct xt_entry_target *t;
 896         unsigned int entry_offset;
 897         int off, i, ret;
 898
 899         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
 900         entry_offset = (void *)e - base;
 901         xt_ematch_foreach(ematch, e)
 902                 off += xt_compat_match_offset(ematch->u.kernel.match);
 903         t = ipt_get_target_c(e);
 904         off += xt_compat_target_offset(t->u.kernel.target);
 905         newinfo->size -= off;
 906         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
 907         if (ret)
 908                 return ret;
 909
 910         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 911                 if (info->hook_entry[i] &&
 912                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
 913                         newinfo->hook_entry[i] -= off;
 914                 if (info->underflow[i] &&
 915                     (e < (struct ipt_entry *)(base + info->underflow[i])))
 916                         newinfo->underflow[i] -= off;
 917         }
 918         return 0;
 919 }
 920
 921 static int compat_table_info(const struct xt_table_info *info,
 922                              struct xt_table_info *newinfo)
 923 {
 924         struct ipt_entry *iter;
 925         const void *loc_cpu_entry;
 926         int ret;
 927
 928         if (!newinfo || !info)
 929                 return -EINVAL;
 930
 931         /* we dont care about newinfo->entries */
 932         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 933         newinfo->initial_entries = 0;
 934         loc_cpu_entry = info->entries;
 935         ret = xt_compat_init_offsets(AF_INET, info->number);
 936         if (ret)
 937                 return ret;
 938         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 939                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 940                 if (ret != 0)
 941                         return ret;
 942         }
 943         return 0;
 944 }
 945 #endif
 946
 947 static int get_info(struct net *net, void __user *user,
 948                     const int *len, int compat)
 949 {
 950         char name[XT_TABLE_MAXNAMELEN];
 951         struct xt_table *t;
 952         int ret;
 953
 954         if (*len != sizeof(struct ipt_getinfo))
 955                 return -EINVAL;
 956
 957         if (copy_from_user(name, user, sizeof(name)) != 0)
 958                 return -EFAULT;
 959
 960         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 961 #ifdef CONFIG_COMPAT
 962         if (compat)
 963                 xt_compat_lock(AF_INET);
 964 #endif
 965         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
 966                                     "iptable_%s", name);
 967         if (t) {
 968                 struct ipt_getinfo info;
 969                 const struct xt_table_info *private = t->private;
 970 #ifdef CONFIG_COMPAT
 971                 struct xt_table_info tmp;
 972
 973                 if (compat) {
 974                         ret = compat_table_info(private, &tmp);
 975                         xt_compat_flush_offsets(AF_INET);
 976                         private = &tmp;
 977                 }
 978 #endif
 979                 memset(&info, 0, sizeof(info));
 980                 info.valid_hooks = t->valid_hooks;
 981                 memcpy(info.hook_entry, private->hook_entry,
 982                        sizeof(info.hook_entry));
 983                 memcpy(info.underflow, private->underflow,
 984                        sizeof(info.underflow));
 985                 info.num_entries = private->number;
 986                 info.size = private->size;
 987                 strcpy(info.name, name);
 988
 989                 if (copy_to_user(user, &info, *len) != 0)
 990                         ret = -EFAULT;
 991                 else
 992                         ret = 0;
 993
 994                 xt_table_unlock(t);
 995                 module_put(t->me);
 996         } else
 997                 ret = -ENOENT;
 998 #ifdef CONFIG_COMPAT
 999         if (compat)
1000                 xt_compat_unlock(AF_INET);
1001 #endif
1002         return ret;
1003 }
1004
1005 static int
1006 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1007             const int *len)
1008 {
1009         int ret;
1010         struct ipt_get_entries get;
1011         struct xt_table *t;
1012
1013         if (*len < sizeof(get))
1014                 return -EINVAL;
1015         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1016                 return -EFAULT;
1017         if (*len != sizeof(struct ipt_get_entries) + get.size)
1018                 return -EINVAL;
1019         get.name[sizeof(get.name) - 1] = '\0';
1020
1021         t = xt_find_table_lock(net, AF_INET, get.name);
1022         if (t) {
1023                 const struct xt_table_info *private = t->private;
1024                 if (get.size == private->size)
1025                         ret = copy_entries_to_user(private->size,
1026                                                    t, uptr->entrytable);
1027                 else
1028                         ret = -EAGAIN;
1029
1030                 module_put(t->me);
1031                 xt_table_unlock(t);
1032         } else
1033                 ret = -ENOENT;
1034
1035         return ret;
1036 }
1037
1038 static int
1039 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1040              struct xt_table_info *newinfo, unsigned int num_counters,
1041              void __user *counters_ptr)
1042 {
1043         int ret;
1044         struct xt_table *t;
1045         struct xt_table_info *oldinfo;
1046         struct xt_counters *counters;
1047         struct ipt_entry *iter;
1048
1049         ret = 0;
1050         counters = xt_counters_alloc(num_counters);
1051         if (!counters) {
1052                 ret = -ENOMEM;
1053                 goto out;
1054         }
1055
1056         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1057                                     "iptable_%s", name);
1058         if (!t) {
1059                 ret = -ENOENT;
1060                 goto free_newinfo_counters_untrans;
1061         }
1062
1063         /* You lied! */
1064         if (valid_hooks != t->valid_hooks) {
1065                 ret = -EINVAL;
1066                 goto put_module;
1067         }
1068
1069         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1070         if (!oldinfo)
1071                 goto put_module;
1072
1073         /* Update module usage count based on number of rules */
1074         if ((oldinfo->number > oldinfo->initial_entries) ||
1075             (newinfo->number <= oldinfo->initial_entries))
1076                 module_put(t->me);
1077         if ((oldinfo->number > oldinfo->initial_entries) &&
1078             (newinfo->number <= oldinfo->initial_entries))
1079                 module_put(t->me);
1080
1081         /* Get the old counters, and synchronize with replace */
1082         get_counters(oldinfo, counters);
1083
1084         /* Decrease module usage counts and free resource */
1085         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1086                 cleanup_entry(iter, net);
1087
1088         xt_free_table_info(oldinfo);
1089         if (copy_to_user(counters_ptr, counters,
1090                          sizeof(struct xt_counters) * num_counters) != 0) {
1091                 /* Silent error, can't fail, new table is already in place */
1092                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1093         }
1094         vfree(counters);
1095         xt_table_unlock(t);
1096         return ret;
1097
1098  put_module:
1099         module_put(t->me);
1100         xt_table_unlock(t);
1101  free_newinfo_counters_untrans:
1102         vfree(counters);
1103  out:
1104         return ret;
1105 }
1106
1107 static int
1108 do_replace(struct net *net, const void __user *user, unsigned int len)
1109 {
1110         int ret;
1111         struct ipt_replace tmp;
1112         struct xt_table_info *newinfo;
1113         void *loc_cpu_entry;
1114         struct ipt_entry *iter;
1115
1116         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1117                 return -EFAULT;
1118
1119         /* overflow check */
1120         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1121                 return -ENOMEM;
1122         if (tmp.num_counters == 0)
1123                 return -EINVAL;
1124
1125         tmp.name[sizeof(tmp.name)-1] = 0;
1126
1127         newinfo = xt_alloc_table_info(tmp.size);
1128         if (!newinfo)
1129                 return -ENOMEM;
1130
1131         loc_cpu_entry = newinfo->entries;
1132         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1133                            tmp.size) != 0) {
1134                 ret = -EFAULT;
1135                 goto free_newinfo;
1136         }
1137
1138         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1139         if (ret != 0)
1140                 goto free_newinfo;
1141
1142         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1143                            tmp.num_counters, tmp.counters);
1144         if (ret)
1145                 goto free_newinfo_untrans;
1146         return 0;
1147
1148  free_newinfo_untrans:
1149         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1150                 cleanup_entry(iter, net);
1151  free_newinfo:
1152         xt_free_table_info(newinfo);
1153         return ret;
1154 }
1155
1156 static int
1157 do_add_counters(struct net *net, const void __user *user,
1158                 unsigned int len, int compat)
1159 {
1160         unsigned int i;
1161         struct xt_counters_info tmp;
1162         struct xt_counters *paddc;
1163         struct xt_table *t;
1164         const struct xt_table_info *private;
1165         int ret = 0;
1166         struct ipt_entry *iter;
1167         unsigned int addend;
1168
1169         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1170         if (IS_ERR(paddc))
1171                 return PTR_ERR(paddc);
1172
1173         t = xt_find_table_lock(net, AF_INET, tmp.name);
1174         if (!t) {
1175                 ret = -ENOENT;
1176                 goto free;
1177         }
1178
1179         local_bh_disable();
1180         private = t->private;
1181         if (private->number != tmp.num_counters) {
1182                 ret = -EINVAL;
1183                 goto unlock_up_free;
1184         }
1185
1186         i = 0;
1187         addend = xt_write_recseq_begin();
1188         xt_entry_foreach(iter, private->entries, private->size) {
1189                 struct xt_counters *tmp;
1190
1191                 tmp = xt_get_this_cpu_counter(&iter->counters);
1192                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1193                 ++i;
1194         }
1195         xt_write_recseq_end(addend);
1196  unlock_up_free:
1197         local_bh_enable();
1198         xt_table_unlock(t);
1199         module_put(t->me);
1200  free:
1201         vfree(paddc);
1202
1203         return ret;
1204 }
1205
1206 #ifdef CONFIG_COMPAT
1207 struct compat_ipt_replace {
1208         char                    name[XT_TABLE_MAXNAMELEN];
1209         u32                     valid_hooks;
1210         u32                     num_entries;
1211         u32                     size;
1212         u32                     hook_entry[NF_INET_NUMHOOKS];
1213         u32                     underflow[NF_INET_NUMHOOKS];
1214         u32                     num_counters;
1215         compat_uptr_t           counters;       /* struct xt_counters * */
1216         struct compat_ipt_entry entries[0];
1217 };
1218
1219 static int
1220 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1221                           unsigned int *size, struct xt_counters *counters,
1222                           unsigned int i)
1223 {
1224         struct xt_entry_target *t;
1225         struct compat_ipt_entry __user *ce;
1226         u_int16_t target_offset, next_offset;
1227         compat_uint_t origsize;
1228         const struct xt_entry_match *ematch;
1229         int ret = 0;
1230
1231         origsize = *size;
1232         ce = *dstptr;
1233         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1234             copy_to_user(&ce->counters, &counters[i],
1235             sizeof(counters[i])) != 0)
1236                 return -EFAULT;
1237
1238         *dstptr += sizeof(struct compat_ipt_entry);
1239         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1240
1241         xt_ematch_foreach(ematch, e) {
1242                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1243                 if (ret != 0)
1244                         return ret;
1245         }
1246         target_offset = e->target_offset - (origsize - *size);
1247         t = ipt_get_target(e);
1248         ret = xt_compat_target_to_user(t, dstptr, size);
1249         if (ret)
1250                 return ret;
1251         next_offset = e->next_offset - (origsize - *size);
1252         if (put_user(target_offset, &ce->target_offset) != 0 ||
1253             put_user(next_offset, &ce->next_offset) != 0)
1254                 return -EFAULT;
1255         return 0;
1256 }
1257
1258 static int
1259 compat_find_calc_match(struct xt_entry_match *m,
1260                        const struct ipt_ip *ip,
1261                        int *size)
1262 {
1263         struct xt_match *match;
1264
1265         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1266                                       m->u.user.revision);
1267         if (IS_ERR(match))
1268                 return PTR_ERR(match);
1269
1270         m->u.kernel.match = match;
1271         *size += xt_compat_match_offset(match);
1272         return 0;
1273 }
1274
1275 static void compat_release_entry(struct compat_ipt_entry *e)
1276 {
1277         struct xt_entry_target *t;
1278         struct xt_entry_match *ematch;
1279
1280         /* Cleanup all matches */
1281         xt_ematch_foreach(ematch, e)
1282                 module_put(ematch->u.kernel.match->me);
1283         t = compat_ipt_get_target(e);
1284         module_put(t->u.kernel.target->me);
1285 }
1286
1287 static int
1288 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1289                                   struct xt_table_info *newinfo,
1290                                   unsigned int *size,
1291                                   const unsigned char *base,
1292                                   const unsigned char *limit)
1293 {
1294         struct xt_entry_match *ematch;
1295         struct xt_entry_target *t;
1296         struct xt_target *target;
1297         unsigned int entry_offset;
1298         unsigned int j;
1299         int ret, off;
1300
1301         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1302             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1303             (unsigned char *)e + e->next_offset > limit)
1304                 return -EINVAL;
1305
1306         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1307                              sizeof(struct compat_xt_entry_target))
1308                 return -EINVAL;
1309
1310         if (!ip_checkentry(&e->ip))
1311                 return -EINVAL;
1312
1313         ret = xt_compat_check_entry_offsets(e, e->elems,
1314                                             e->target_offset, e->next_offset);
1315         if (ret)
1316                 return ret;
1317
1318         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1319         entry_offset = (void *)e - (void *)base;
1320         j = 0;
1321         xt_ematch_foreach(ematch, e) {
1322                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1323                 if (ret != 0)
1324                         goto release_matches;
1325                 ++j;
1326         }
1327
1328         t = compat_ipt_get_target(e);
1329         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1330                                         t->u.user.revision);
1331         if (IS_ERR(target)) {
1332                 ret = PTR_ERR(target);
1333                 goto release_matches;
1334         }
1335         t->u.kernel.target = target;
1336
1337         off += xt_compat_target_offset(target);
1338         *size += off;
1339         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1340         if (ret)
1341                 goto out;
1342
1343         return 0;
1344
1345 out:
1346         module_put(t->u.kernel.target->me);
1347 release_matches:
1348         xt_ematch_foreach(ematch, e) {
1349                 if (j-- == 0)
1350                         break;
1351                 module_put(ematch->u.kernel.match->me);
1352         }
1353         return ret;
1354 }
1355
1356 static void
1357 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1358                             unsigned int *size,
1359                             struct xt_table_info *newinfo, unsigned char *base)
1360 {
1361         struct xt_entry_target *t;
1362         struct ipt_entry *de;
1363         unsigned int origsize;
1364         int h;
1365         struct xt_entry_match *ematch;
1366
1367         origsize = *size;
1368         de = *dstptr;
1369         memcpy(de, e, sizeof(struct ipt_entry));
1370         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1371
1372         *dstptr += sizeof(struct ipt_entry);
1373         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1374
1375         xt_ematch_foreach(ematch, e)
1376                 xt_compat_match_from_user(ematch, dstptr, size);
1377
1378         de->target_offset = e->target_offset - (origsize - *size);
1379         t = compat_ipt_get_target(e);
1380         xt_compat_target_from_user(t, dstptr, size);
1381
1382         de->next_offset = e->next_offset - (origsize - *size);
1383
1384         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1385                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1386                         newinfo->hook_entry[h] -= origsize - *size;
1387                 if ((unsigned char *)de - base < newinfo->underflow[h])
1388                         newinfo->underflow[h] -= origsize - *size;
1389         }
1390 }
1391
1392 static int
1393 translate_compat_table(struct net *net,
1394                        struct xt_table_info **pinfo,
1395                        void **pentry0,
1396                        const struct compat_ipt_replace *compatr)
1397 {
1398         unsigned int i, j;
1399         struct xt_table_info *newinfo, *info;
1400         void *pos, *entry0, *entry1;
1401         struct compat_ipt_entry *iter0;
1402         struct ipt_replace repl;
1403         unsigned int size;
1404         int ret;
1405
1406         info = *pinfo;
1407         entry0 = *pentry0;
1408         size = compatr->size;
1409         info->number = compatr->num_entries;
1410
1411         j = 0;
1412         xt_compat_lock(AF_INET);
1413         ret = xt_compat_init_offsets(AF_INET, compatr->num_entries);
1414         if (ret)
1415                 goto out_unlock;
1416         /* Walk through entries, checking offsets. */
1417         xt_entry_foreach(iter0, entry0, compatr->size) {
1418                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1419                                                         entry0,
1420                                                         entry0 + compatr->size);
1421                 if (ret != 0)
1422                         goto out_unlock;
1423                 ++j;
1424         }
1425
1426         ret = -EINVAL;
1427         if (j != compatr->num_entries)
1428                 goto out_unlock;
1429
1430         ret = -ENOMEM;
1431         newinfo = xt_alloc_table_info(size);
1432         if (!newinfo)
1433                 goto out_unlock;
1434
1435         memset(newinfo->entries, 0, size);
1436
1437         newinfo->number = compatr->num_entries;
1438         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1439                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1440                 newinfo->underflow[i] = compatr->underflow[i];
1441         }
1442         entry1 = newinfo->entries;
1443         pos = entry1;
1444         size = compatr->size;
1445         xt_entry_foreach(iter0, entry0, compatr->size)
1446                 compat_copy_entry_from_user(iter0, &pos, &size,
1447                                             newinfo, entry1);
1448
1449         /* all module references in entry0 are now gone.
1450          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1451          * generated by 64bit userspace.
1452          *
1453          * Call standard translate_table() to validate all hook_entrys,
1454          * underflows, check for loops, etc.
1455          */
1456         xt_compat_flush_offsets(AF_INET);
1457         xt_compat_unlock(AF_INET);
1458
1459         memcpy(&repl, compatr, sizeof(*compatr));
1460
1461         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1462                 repl.hook_entry[i] = newinfo->hook_entry[i];
1463                 repl.underflow[i] = newinfo->underflow[i];
1464         }
1465
1466         repl.num_counters = 0;
1467         repl.counters = NULL;
1468         repl.size = newinfo->size;
1469         ret = translate_table(net, newinfo, entry1, &repl);
1470         if (ret)
1471                 goto free_newinfo;
1472
1473         *pinfo = newinfo;
1474         *pentry0 = entry1;
1475         xt_free_table_info(info);
1476         return 0;
1477
1478 free_newinfo:
1479         xt_free_table_info(newinfo);
1480         return ret;
1481 out_unlock:
1482         xt_compat_flush_offsets(AF_INET);
1483         xt_compat_unlock(AF_INET);
1484         xt_entry_foreach(iter0, entry0, compatr->size) {
1485                 if (j-- == 0)
1486                         break;
1487                 compat_release_entry(iter0);
1488         }
1489         return ret;
1490 }
1491
1492 static int
1493 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1494 {
1495         int ret;
1496         struct compat_ipt_replace tmp;
1497         struct xt_table_info *newinfo;
1498         void *loc_cpu_entry;
1499         struct ipt_entry *iter;
1500
1501         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1502                 return -EFAULT;
1503
1504         /* overflow check */
1505         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1506                 return -ENOMEM;
1507         if (tmp.num_counters == 0)
1508                 return -EINVAL;
1509
1510         tmp.name[sizeof(tmp.name)-1] = 0;
1511
1512         newinfo = xt_alloc_table_info(tmp.size);
1513         if (!newinfo)
1514                 return -ENOMEM;
1515
1516         loc_cpu_entry = newinfo->entries;
1517         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1518                            tmp.size) != 0) {
1519                 ret = -EFAULT;
1520                 goto free_newinfo;
1521         }
1522
1523         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1524         if (ret != 0)
1525                 goto free_newinfo;
1526
1527         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1528                            tmp.num_counters, compat_ptr(tmp.counters));
1529         if (ret)
1530                 goto free_newinfo_untrans;
1531         return 0;
1532
1533  free_newinfo_untrans:
1534         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1535                 cleanup_entry(iter, net);
1536  free_newinfo:
1537         xt_free_table_info(newinfo);
1538         return ret;
1539 }
1540
1541 static int
1542 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1543                       unsigned int len)
1544 {
1545         int ret;
1546
1547         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1548                 return -EPERM;
1549
1550         switch (cmd) {
1551         case IPT_SO_SET_REPLACE:
1552                 ret = compat_do_replace(sock_net(sk), user, len);
1553                 break;
1554
1555         case IPT_SO_SET_ADD_COUNTERS:
1556                 ret = do_add_counters(sock_net(sk), user, len, 1);
1557                 break;
1558
1559         default:
1560                 ret = -EINVAL;
1561         }
1562
1563         return ret;
1564 }
1565
1566 struct compat_ipt_get_entries {
1567         char name[XT_TABLE_MAXNAMELEN];
1568         compat_uint_t size;
1569         struct compat_ipt_entry entrytable[0];
1570 };
1571
1572 static int
1573 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1574                             void __user *userptr)
1575 {
1576         struct xt_counters *counters;
1577         const struct xt_table_info *private = table->private;
1578         void __user *pos;
1579         unsigned int size;
1580         int ret = 0;
1581         unsigned int i = 0;
1582         struct ipt_entry *iter;
1583
1584         counters = alloc_counters(table);
1585         if (IS_ERR(counters))
1586                 return PTR_ERR(counters);
1587
1588         pos = userptr;
1589         size = total_size;
1590         xt_entry_foreach(iter, private->entries, total_size) {
1591                 ret = compat_copy_entry_to_user(iter, &pos,
1592                                                 &size, counters, i++);
1593                 if (ret != 0)
1594                         break;
1595         }
1596
1597         vfree(counters);
1598         return ret;
1599 }
1600
1601 static int
1602 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1603                    int *len)
1604 {
1605         int ret;
1606         struct compat_ipt_get_entries get;
1607         struct xt_table *t;
1608
1609         if (*len < sizeof(get))
1610                 return -EINVAL;
1611
1612         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1613                 return -EFAULT;
1614
1615         if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1616                 return -EINVAL;
1617
1618         get.name[sizeof(get.name) - 1] = '\0';
1619
1620         xt_compat_lock(AF_INET);
1621         t = xt_find_table_lock(net, AF_INET, get.name);
1622         if (t) {
1623                 const struct xt_table_info *private = t->private;
1624                 struct xt_table_info info;
1625                 ret = compat_table_info(private, &info);
1626                 if (!ret && get.size == info.size)
1627                         ret = compat_copy_entries_to_user(private->size,
1628                                                           t, uptr->entrytable);
1629                 else if (!ret)
1630                         ret = -EAGAIN;
1631
1632                 xt_compat_flush_offsets(AF_INET);
1633                 module_put(t->me);
1634                 xt_table_unlock(t);
1635         } else
1636                 ret = -ENOENT;
1637
1638         xt_compat_unlock(AF_INET);
1639         return ret;
1640 }
1641
1642 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1643
1644 static int
1645 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1646 {
1647         int ret;
1648
1649         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1650                 return -EPERM;
1651
1652         switch (cmd) {
1653         case IPT_SO_GET_INFO:
1654                 ret = get_info(sock_net(sk), user, len, 1);
1655                 break;
1656         case IPT_SO_GET_ENTRIES:
1657                 ret = compat_get_entries(sock_net(sk), user, len);
1658                 break;
1659         default:
1660                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1661         }
1662         return ret;
1663 }
1664 #endif
1665
1666 static int
1667 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1668 {
1669         int ret;
1670
1671         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1672                 return -EPERM;
1673
1674         switch (cmd) {
1675         case IPT_SO_SET_REPLACE:
1676                 ret = do_replace(sock_net(sk), user, len);
1677                 break;
1678
1679         case IPT_SO_SET_ADD_COUNTERS:
1680                 ret = do_add_counters(sock_net(sk), user, len, 0);
1681                 break;
1682
1683         default:
1684                 ret = -EINVAL;
1685         }
1686
1687         return ret;
1688 }
1689
1690 static int
1691 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1692 {
1693         int ret;
1694
1695         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1696                 return -EPERM;
1697
1698         switch (cmd) {
1699         case IPT_SO_GET_INFO:
1700                 ret = get_info(sock_net(sk), user, len, 0);
1701                 break;
1702
1703         case IPT_SO_GET_ENTRIES:
1704                 ret = get_entries(sock_net(sk), user, len);
1705                 break;
1706
1707         case IPT_SO_GET_REVISION_MATCH:
1708         case IPT_SO_GET_REVISION_TARGET: {
1709                 struct xt_get_revision rev;
1710                 int target;
1711
1712                 if (*len != sizeof(rev)) {
1713                         ret = -EINVAL;
1714                         break;
1715                 }
1716                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1717                         ret = -EFAULT;
1718                         break;
1719                 }
1720                 rev.name[sizeof(rev.name)-1] = 0;
1721
1722                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1723                         target = 1;
1724                 else
1725                         target = 0;
1726
1727                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1728                                                          rev.revision,
1729                                                          target, &ret),
1730                                         "ipt_%s", rev.name);
1731                 break;
1732         }
1733
1734         default:
1735                 ret = -EINVAL;
1736         }
1737
1738         return ret;
1739 }
1740
1741 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1742 {
1743         struct xt_table_info *private;
1744         void *loc_cpu_entry;
1745         struct module *table_owner = table->me;
1746         struct ipt_entry *iter;
1747
1748         private = xt_unregister_table(table);
1749
1750         /* Decrease module usage counts and free resources */
1751         loc_cpu_entry = private->entries;
1752         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1753                 cleanup_entry(iter, net);
1754         if (private->number > private->initial_entries)
1755                 module_put(table_owner);
1756         xt_free_table_info(private);
1757 }
1758
1759 int ipt_register_table(struct net *net, const struct xt_table *table,
1760                        const struct ipt_replace *repl,
1761                        const struct nf_hook_ops *ops, struct xt_table **res)
1762 {
1763         int ret;
1764         struct xt_table_info *newinfo;
1765         struct xt_table_info bootstrap = {0};
1766         void *loc_cpu_entry;
1767         struct xt_table *new_table;
1768
1769         newinfo = xt_alloc_table_info(repl->size);
1770         if (!newinfo)
1771                 return -ENOMEM;
1772
1773         loc_cpu_entry = newinfo->entries;
1774         memcpy(loc_cpu_entry, repl->entries, repl->size);
1775
1776         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1777         if (ret != 0)
1778                 goto out_free;
1779
1780         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1781         if (IS_ERR(new_table)) {
1782                 ret = PTR_ERR(new_table);
1783                 goto out_free;
1784         }
1785
1786         /* set res now, will see skbs right after nf_register_net_hooks */
1787         WRITE_ONCE(*res, new_table);
1788
1789         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1790         if (ret != 0) {
1791                 __ipt_unregister_table(net, new_table);
1792                 *res = NULL;
1793         }
1794
1795         return ret;
1796
1797 out_free:
1798         xt_free_table_info(newinfo);
1799         return ret;
1800 }
1801
1802 void ipt_unregister_table(struct net *net, struct xt_table *table,
1803                           const struct nf_hook_ops *ops)
1804 {
1805         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1806         __ipt_unregister_table(net, table);
1807 }
1808
1809 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1810 static inline bool
1811 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1812                      u_int8_t type, u_int8_t code,
1813                      bool invert)
1814 {
1815         return ((test_type == 0xFF) ||
1816                 (type == test_type && code >= min_code && code <= max_code))
1817                 ^ invert;
1818 }
1819
1820 static bool
1821 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
1822 {
1823         const struct icmphdr *ic;
1824         struct icmphdr _icmph;
1825         const struct ipt_icmp *icmpinfo = par->matchinfo;
1826
1827         /* Must not be a fragment. */
1828         if (par->fragoff != 0)
1829                 return false;
1830
1831         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1832         if (ic == NULL) {
1833                 /* We've been asked to examine this packet, and we
1834                  * can't.  Hence, no choice but to drop.
1835                  */
1836                 par->hotdrop = true;
1837                 return false;
1838         }
1839
1840         return icmp_type_code_match(icmpinfo->type,
1841                                     icmpinfo->code[0],
1842                                     icmpinfo->code[1],
1843                                     ic->type, ic->code,
1844                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
1845 }
1846
1847 static int icmp_checkentry(const struct xt_mtchk_param *par)
1848 {
1849         const struct ipt_icmp *icmpinfo = par->matchinfo;
1850
1851         /* Must specify no unknown invflags */
1852         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
1853 }
1854
1855 static struct xt_target ipt_builtin_tg[] __read_mostly = {
1856         {
1857                 .name             = XT_STANDARD_TARGET,
1858                 .targetsize       = sizeof(int),
1859                 .family           = NFPROTO_IPV4,
1860 #ifdef CONFIG_COMPAT
1861                 .compatsize       = sizeof(compat_int_t),
1862                 .compat_from_user = compat_standard_from_user,
1863                 .compat_to_user   = compat_standard_to_user,
1864 #endif
1865         },
1866         {
1867                 .name             = XT_ERROR_TARGET,
1868                 .target           = ipt_error,
1869                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1870                 .family           = NFPROTO_IPV4,
1871         },
1872 };
1873
1874 static struct nf_sockopt_ops ipt_sockopts = {
1875         .pf             = PF_INET,
1876         .set_optmin     = IPT_BASE_CTL,
1877         .set_optmax     = IPT_SO_SET_MAX+1,
1878         .set            = do_ipt_set_ctl,
1879 #ifdef CONFIG_COMPAT
1880         .compat_set     = compat_do_ipt_set_ctl,
1881 #endif
1882         .get_optmin     = IPT_BASE_CTL,
1883         .get_optmax     = IPT_SO_GET_MAX+1,
1884         .get            = do_ipt_get_ctl,
1885 #ifdef CONFIG_COMPAT
1886         .compat_get     = compat_do_ipt_get_ctl,
1887 #endif
1888         .owner          = THIS_MODULE,
1889 };
1890
1891 static struct xt_match ipt_builtin_mt[] __read_mostly = {
1892         {
1893                 .name       = "icmp",
1894                 .match      = icmp_match,
1895                 .matchsize  = sizeof(struct ipt_icmp),
1896                 .checkentry = icmp_checkentry,
1897                 .proto      = IPPROTO_ICMP,
1898                 .family     = NFPROTO_IPV4,
1899                 .me         = THIS_MODULE,
1900         },
1901 };
1902
1903 static int __net_init ip_tables_net_init(struct net *net)
1904 {
1905         return xt_proto_init(net, NFPROTO_IPV4);
1906 }
1907
1908 static void __net_exit ip_tables_net_exit(struct net *net)
1909 {
1910         xt_proto_fini(net, NFPROTO_IPV4);
1911 }
1912
1913 static struct pernet_operations ip_tables_net_ops = {
1914         .init = ip_tables_net_init,
1915         .exit = ip_tables_net_exit,
1916 };
1917
1918 static int __init ip_tables_init(void)
1919 {
1920         int ret;
1921
1922         ret = register_pernet_subsys(&ip_tables_net_ops);
1923         if (ret < 0)
1924                 goto err1;
1925
1926         /* No one else will be downing sem now, so we won't sleep */
1927         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1928         if (ret < 0)
1929                 goto err2;
1930         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1931         if (ret < 0)
1932                 goto err4;
1933
1934         /* Register setsockopt */
1935         ret = nf_register_sockopt(&ipt_sockopts);
1936         if (ret < 0)
1937                 goto err5;
1938
1939         pr_info("(C) 2000-2006 Netfilter Core Team\n");
1940         return 0;
1941
1942 err5:
1943         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1944 err4:
1945         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1946 err2:
1947         unregister_pernet_subsys(&ip_tables_net_ops);
1948 err1:
1949         return ret;
1950 }
1951
1952 static void __exit ip_tables_fini(void)
1953 {
1954         nf_unregister_sockopt(&ipt_sockopts);
1955
1956         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1957         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1958         unregister_pernet_subsys(&ip_tables_net_ops);
1959 }
1960
1961 EXPORT_SYMBOL(ipt_register_table);
1962 EXPORT_SYMBOL(ipt_unregister_table);
1963 EXPORT_SYMBOL(ipt_do_table);
1964 module_init(ip_tables_init);
1965 module_exit(ip_tables_fini);