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