net/ipv4/netfilter/ip_tables.c

   1 /*
   2  * Packet matching code.
   3  *
   4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   6  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13 #include <linux/cache.h>
  14 #include <linux/capability.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/kmod.h>
  17 #include <linux/vmalloc.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/module.h>
  20 #include <linux/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/compat.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/mutex.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/err.h>
  27 #include <linux/cpumask.h>
  28
  29 #include <linux/netfilter/x_tables.h>
  30 #include <linux/netfilter_ipv4/ip_tables.h>
  31 #include <net/netfilter/nf_log.h>
  32 #include "../../netfilter/xt_repldata.h"
  33
  34 MODULE_LICENSE("GPL");
  35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36 MODULE_DESCRIPTION("IPv4 packet filter");
  37
  38 #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 struct nf_loginfo trace_loginfo = {
 155         .type = NF_LOG_TYPE_LOG,
 156         .u = {
 157                 .log = {
 158                         .level = 4,
 159                         .logflags = NF_LOG_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.net     = state->net;
 265         acpar.in      = state->in;
 266         acpar.out     = state->out;
 267         acpar.family  = NFPROTO_IPV4;
 268         acpar.hooknum = hook;
 269
 270         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 271         local_bh_disable();
 272         addend = xt_write_recseq_begin();
 273         private = table->private;
 274         cpu        = smp_processor_id();
 275         /*
 276          * Ensure we load private-> members after we've fetched the base
 277          * pointer.
 278          */
 279         smp_read_barrier_depends();
 280         table_base = private->entries;
 281         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 282
 283         /* Switch to alternate jumpstack if we're being invoked via TEE.
 284          * TEE issues XT_CONTINUE verdict on original skb so we must not
 285          * clobber the jumpstack.
 286          *
 287          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 288          * but it is no problem since absolute verdict is issued by these.
 289          */
 290         if (static_key_false(&xt_tee_enabled))
 291                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 292
 293         e = get_entry(table_base, private->hook_entry[hook]);
 294
 295         do {
 296                 const struct xt_entry_target *t;
 297                 const struct xt_entry_match *ematch;
 298                 struct xt_counters *counter;
 299
 300                 IP_NF_ASSERT(e);
 301                 if (!ip_packet_match(ip, indev, outdev,
 302                     &e->ip, acpar.fragoff)) {
 303  no_match:
 304                         e = ipt_next_entry(e);
 305                         continue;
 306                 }
 307
 308                 xt_ematch_foreach(ematch, e) {
 309                         acpar.match     = ematch->u.kernel.match;
 310                         acpar.matchinfo = ematch->data;
 311                         if (!acpar.match->match(skb, &acpar))
 312                                 goto no_match;
 313                 }
 314
 315                 counter = xt_get_this_cpu_counter(&e->counters);
 316                 ADD_COUNTER(*counter, skb->len, 1);
 317
 318                 t = ipt_get_target(e);
 319                 IP_NF_ASSERT(t->u.kernel.target);
 320
 321 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 322                 /* The packet is traced: log it */
 323                 if (unlikely(skb->nf_trace))
 324                         trace_packet(state->net, skb, hook, state->in,
 325                                      state->out, table->name, private, e);
 326 #endif
 327                 /* Standard target? */
 328                 if (!t->u.kernel.target->target) {
 329                         int v;
 330
 331                         v = ((struct xt_standard_target *)t)->verdict;
 332                         if (v < 0) {
 333                                 /* Pop from stack? */
 334                                 if (v != XT_RETURN) {
 335                                         verdict = (unsigned int)(-v) - 1;
 336                                         break;
 337                                 }
 338                                 if (stackidx == 0) {
 339                                         e = get_entry(table_base,
 340                                             private->underflow[hook]);
 341                                 } else {
 342                                         e = jumpstack[--stackidx];
 343                                         e = ipt_next_entry(e);
 344                                 }
 345                                 continue;
 346                         }
 347                         if (table_base + v != ipt_next_entry(e) &&
 348                             !(e->ip.flags & IPT_F_GOTO))
 349                                 jumpstack[stackidx++] = e;
 350
 351                         e = get_entry(table_base, v);
 352                         continue;
 353                 }
 354
 355                 acpar.target   = t->u.kernel.target;
 356                 acpar.targinfo = t->data;
 357
 358                 verdict = t->u.kernel.target->target(skb, &acpar);
 359                 /* Target might have changed stuff. */
 360                 ip = ip_hdr(skb);
 361                 if (verdict == XT_CONTINUE)
 362                         e = ipt_next_entry(e);
 363                 else
 364                         /* Verdict */
 365                         break;
 366         } while (!acpar.hotdrop);
 367
 368         xt_write_recseq_end(addend);
 369         local_bh_enable();
 370
 371         if (acpar.hotdrop)
 372                 return NF_DROP;
 373         else return verdict;
 374 }
 375
 376 /* Figures out from what hook each rule can be called: returns 0 if
 377    there are loops.  Puts hook bitmask in comefrom. */
 378 static int
 379 mark_source_chains(const struct xt_table_info *newinfo,
 380                    unsigned int valid_hooks, void *entry0,
 381                    unsigned int *offsets)
 382 {
 383         unsigned int hook;
 384
 385         /* No recursion; use packet counter to save back ptrs (reset
 386            to 0 as we leave), and comefrom to save source hook bitmask */
 387         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 388                 unsigned int pos = newinfo->hook_entry[hook];
 389                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 390
 391                 if (!(valid_hooks & (1 << hook)))
 392                         continue;
 393
 394                 /* Set initial back pointer. */
 395                 e->counters.pcnt = pos;
 396
 397                 for (;;) {
 398                         const struct xt_standard_target *t
 399                                 = (void *)ipt_get_target_c(e);
 400                         int visited = e->comefrom & (1 << hook);
 401
 402                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 403                                 return 0;
 404
 405                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 406
 407                         /* Unconditional return/END. */
 408                         if ((unconditional(e) &&
 409                              (strcmp(t->target.u.user.name,
 410                                      XT_STANDARD_TARGET) == 0) &&
 411                              t->verdict < 0) || visited) {
 412                                 unsigned int oldpos, size;
 413
 414                                 if ((strcmp(t->target.u.user.name,
 415                                             XT_STANDARD_TARGET) == 0) &&
 416                                     t->verdict < -NF_MAX_VERDICT - 1)
 417                                         return 0;
 418
 419                                 /* Return: backtrack through the last
 420                                    big jump. */
 421                                 do {
 422                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 423                                         oldpos = pos;
 424                                         pos = e->counters.pcnt;
 425                                         e->counters.pcnt = 0;
 426
 427                                         /* We're at the start. */
 428                                         if (pos == oldpos)
 429                                                 goto next;
 430
 431                                         e = (struct ipt_entry *)
 432                                                 (entry0 + pos);
 433                                 } while (oldpos == pos + e->next_offset);
 434
 435                                 /* Move along one */
 436                                 size = e->next_offset;
 437                                 e = (struct ipt_entry *)
 438                                         (entry0 + pos + size);
 439                                 if (pos + size >= newinfo->size)
 440                                         return 0;
 441                                 e->counters.pcnt = pos;
 442                                 pos += size;
 443                         } else {
 444                                 int newpos = t->verdict;
 445
 446                                 if (strcmp(t->target.u.user.name,
 447                                            XT_STANDARD_TARGET) == 0 &&
 448                                     newpos >= 0) {
 449                                         /* This a jump; chase it. */
 450                                         if (!xt_find_jump_offset(offsets, newpos,
 451                                                                  newinfo->number))
 452                                                 return 0;
 453                                         e = (struct ipt_entry *)
 454                                                 (entry0 + newpos);
 455                                 } else {
 456                                         /* ... this is a fallthru */
 457                                         newpos = pos + e->next_offset;
 458                                         if (newpos >= newinfo->size)
 459                                                 return 0;
 460                                 }
 461                                 e = (struct ipt_entry *)
 462                                         (entry0 + newpos);
 463                                 e->counters.pcnt = pos;
 464                                 pos = newpos;
 465                         }
 466                 }
 467 next:           ;
 468         }
 469         return 1;
 470 }
 471
 472 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 473 {
 474         struct xt_mtdtor_param par;
 475
 476         par.net       = net;
 477         par.match     = m->u.kernel.match;
 478         par.matchinfo = m->data;
 479         par.family    = NFPROTO_IPV4;
 480         if (par.match->destroy != NULL)
 481                 par.match->destroy(&par);
 482         module_put(par.match->me);
 483 }
 484
 485 static int
 486 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 487 {
 488         const struct ipt_ip *ip = par->entryinfo;
 489
 490         par->match     = m->u.kernel.match;
 491         par->matchinfo = m->data;
 492
 493         return xt_check_match(par, m->u.match_size - sizeof(*m),
 494                               ip->proto, ip->invflags & IPT_INV_PROTO);
 495 }
 496
 497 static int
 498 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 499 {
 500         struct xt_match *match;
 501         int ret;
 502
 503         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 504                                       m->u.user.revision);
 505         if (IS_ERR(match))
 506                 return PTR_ERR(match);
 507         m->u.kernel.match = match;
 508
 509         ret = check_match(m, par);
 510         if (ret)
 511                 goto err;
 512
 513         return 0;
 514 err:
 515         module_put(m->u.kernel.match->me);
 516         return ret;
 517 }
 518
 519 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 520 {
 521         struct xt_entry_target *t = ipt_get_target(e);
 522         struct xt_tgchk_param par = {
 523                 .net       = net,
 524                 .table     = name,
 525                 .entryinfo = e,
 526                 .target    = t->u.kernel.target,
 527                 .targinfo  = t->data,
 528                 .hook_mask = e->comefrom,
 529                 .family    = NFPROTO_IPV4,
 530         };
 531
 532         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 533                                e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 534 }
 535
 536 static int
 537 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 538                  unsigned int size)
 539 {
 540         struct xt_entry_target *t;
 541         struct xt_target *target;
 542         int ret;
 543         unsigned int j;
 544         struct xt_mtchk_param mtpar;
 545         struct xt_entry_match *ematch;
 546         unsigned long pcnt;
 547
 548         pcnt = xt_percpu_counter_alloc();
 549         if (IS_ERR_VALUE(pcnt))
 550                 return -ENOMEM;
 551         e->counters.pcnt = pcnt;
 552
 553         j = 0;
 554         mtpar.net       = net;
 555         mtpar.table     = name;
 556         mtpar.entryinfo = &e->ip;
 557         mtpar.hook_mask = e->comefrom;
 558         mtpar.family    = NFPROTO_IPV4;
 559         xt_ematch_foreach(ematch, e) {
 560                 ret = find_check_match(ematch, &mtpar);
 561                 if (ret != 0)
 562                         goto cleanup_matches;
 563                 ++j;
 564         }
 565
 566         t = ipt_get_target(e);
 567         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 568                                         t->u.user.revision);
 569         if (IS_ERR(target)) {
 570                 ret = PTR_ERR(target);
 571                 goto cleanup_matches;
 572         }
 573         t->u.kernel.target = target;
 574
 575         ret = check_target(e, net, name);
 576         if (ret)
 577                 goto err;
 578
 579         return 0;
 580  err:
 581         module_put(t->u.kernel.target->me);
 582  cleanup_matches:
 583         xt_ematch_foreach(ematch, e) {
 584                 if (j-- == 0)
 585                         break;
 586                 cleanup_match(ematch, net);
 587         }
 588
 589         xt_percpu_counter_free(e->counters.pcnt);
 590
 591         return ret;
 592 }
 593
 594 static bool check_underflow(const struct ipt_entry *e)
 595 {
 596         const struct xt_entry_target *t;
 597         unsigned int verdict;
 598
 599         if (!unconditional(e))
 600                 return false;
 601         t = ipt_get_target_c(e);
 602         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 603                 return false;
 604         verdict = ((struct xt_standard_target *)t)->verdict;
 605         verdict = -verdict - 1;
 606         return verdict == NF_DROP || verdict == NF_ACCEPT;
 607 }
 608
 609 static int
 610 check_entry_size_and_hooks(struct ipt_entry *e,
 611                            struct xt_table_info *newinfo,
 612                            const unsigned char *base,
 613                            const unsigned char *limit,
 614                            const unsigned int *hook_entries,
 615                            const unsigned int *underflows,
 616                            unsigned int valid_hooks)
 617 {
 618         unsigned int h;
 619         int err;
 620
 621         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 622             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 623             (unsigned char *)e + e->next_offset > limit)
 624                 return -EINVAL;
 625
 626         if (e->next_offset
 627             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
 628                 return -EINVAL;
 629
 630         if (!ip_checkentry(&e->ip))
 631                 return -EINVAL;
 632
 633         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 634                                      e->next_offset);
 635         if (err)
 636                 return err;
 637
 638         /* Check hooks & underflows */
 639         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 640                 if (!(valid_hooks & (1 << h)))
 641                         continue;
 642                 if ((unsigned char *)e - base == hook_entries[h])
 643                         newinfo->hook_entry[h] = hook_entries[h];
 644                 if ((unsigned char *)e - base == underflows[h]) {
 645                         if (!check_underflow(e))
 646                                 return -EINVAL;
 647
 648                         newinfo->underflow[h] = underflows[h];
 649                 }
 650         }
 651
 652         /* Clear counters and comefrom */
 653         e->counters = ((struct xt_counters) { 0, 0 });
 654         e->comefrom = 0;
 655         return 0;
 656 }
 657
 658 static void
 659 cleanup_entry(struct ipt_entry *e, struct net *net)
 660 {
 661         struct xt_tgdtor_param par;
 662         struct xt_entry_target *t;
 663         struct xt_entry_match *ematch;
 664
 665         /* Cleanup all matches */
 666         xt_ematch_foreach(ematch, e)
 667                 cleanup_match(ematch, net);
 668         t = ipt_get_target(e);
 669
 670         par.net      = net;
 671         par.target   = t->u.kernel.target;
 672         par.targinfo = t->data;
 673         par.family   = NFPROTO_IPV4;
 674         if (par.target->destroy != NULL)
 675                 par.target->destroy(&par);
 676         module_put(par.target->me);
 677         xt_percpu_counter_free(e->counters.pcnt);
 678 }
 679
 680 /* Checks and translates the user-supplied table segment (held in
 681    newinfo) */
 682 static int
 683 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 684                 const struct ipt_replace *repl)
 685 {
 686         struct ipt_entry *iter;
 687         unsigned int *offsets;
 688         unsigned int i;
 689         int ret = 0;
 690
 691         newinfo->size = repl->size;
 692         newinfo->number = repl->num_entries;
 693
 694         /* Init all hooks to impossible value. */
 695         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 696                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 697                 newinfo->underflow[i] = 0xFFFFFFFF;
 698         }
 699
 700         offsets = xt_alloc_entry_offsets(newinfo->number);
 701         if (!offsets)
 702                 return -ENOMEM;
 703         i = 0;
 704         /* Walk through entries, checking offsets. */
 705         xt_entry_foreach(iter, entry0, newinfo->size) {
 706                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 707                                                  entry0 + repl->size,
 708                                                  repl->hook_entry,
 709                                                  repl->underflow,
 710                                                  repl->valid_hooks);
 711                 if (ret != 0)
 712                         goto out_free;
 713                 if (i < repl->num_entries)
 714                         offsets[i] = (void *)iter - entry0;
 715                 ++i;
 716                 if (strcmp(ipt_get_target(iter)->u.user.name,
 717                     XT_ERROR_TARGET) == 0)
 718                         ++newinfo->stacksize;
 719         }
 720
 721         ret = -EINVAL;
 722         if (i != repl->num_entries)
 723                 goto out_free;
 724
 725         /* Check hooks all assigned */
 726         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 727                 /* Only hooks which are valid */
 728                 if (!(repl->valid_hooks & (1 << i)))
 729                         continue;
 730                 if (newinfo->hook_entry[i] == 0xFFFFFFFF)
 731                         goto out_free;
 732                 if (newinfo->underflow[i] == 0xFFFFFFFF)
 733                         goto out_free;
 734         }
 735
 736         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 737                 ret = -ELOOP;
 738                 goto out_free;
 739         }
 740         kvfree(offsets);
 741
 742         /* Finally, each sanity check must pass */
 743         i = 0;
 744         xt_entry_foreach(iter, entry0, newinfo->size) {
 745                 ret = find_check_entry(iter, net, repl->name, repl->size);
 746                 if (ret != 0)
 747                         break;
 748                 ++i;
 749         }
 750
 751         if (ret != 0) {
 752                 xt_entry_foreach(iter, entry0, newinfo->size) {
 753                         if (i-- == 0)
 754                                 break;
 755                         cleanup_entry(iter, net);
 756                 }
 757                 return ret;
 758         }
 759
 760         return ret;
 761  out_free:
 762         kvfree(offsets);
 763         return ret;
 764 }
 765
 766 static void
 767 get_counters(const struct xt_table_info *t,
 768              struct xt_counters counters[])
 769 {
 770         struct ipt_entry *iter;
 771         unsigned int cpu;
 772         unsigned int i;
 773
 774         for_each_possible_cpu(cpu) {
 775                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 776
 777                 i = 0;
 778                 xt_entry_foreach(iter, t->entries, t->size) {
 779                         struct xt_counters *tmp;
 780                         u64 bcnt, pcnt;
 781                         unsigned int start;
 782
 783                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 784                         do {
 785                                 start = read_seqcount_begin(s);
 786                                 bcnt = tmp->bcnt;
 787                                 pcnt = tmp->pcnt;
 788                         } while (read_seqcount_retry(s, start));
 789
 790                         ADD_COUNTER(counters[i], bcnt, pcnt);
 791                         ++i; /* macro does multi eval of i */
 792                 }
 793         }
 794 }
 795
 796 static struct xt_counters *alloc_counters(const struct xt_table *table)
 797 {
 798         unsigned int countersize;
 799         struct xt_counters *counters;
 800         const struct xt_table_info *private = table->private;
 801
 802         /* We need atomic snapshot of counters: rest doesn't change
 803            (other than comefrom, which userspace doesn't care
 804            about). */
 805         countersize = sizeof(struct xt_counters) * private->number;
 806         counters = vzalloc(countersize);
 807
 808         if (counters == NULL)
 809                 return ERR_PTR(-ENOMEM);
 810
 811         get_counters(private, counters);
 812
 813         return counters;
 814 }
 815
 816 static int
 817 copy_entries_to_user(unsigned int total_size,
 818                      const struct xt_table *table,
 819                      void __user *userptr)
 820 {
 821         unsigned int off, num;
 822         const struct ipt_entry *e;
 823         struct xt_counters *counters;
 824         const struct xt_table_info *private = table->private;
 825         int ret = 0;
 826         const void *loc_cpu_entry;
 827
 828         counters = alloc_counters(table);
 829         if (IS_ERR(counters))
 830                 return PTR_ERR(counters);
 831
 832         loc_cpu_entry = private->entries;
 833         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 834                 ret = -EFAULT;
 835                 goto free_counters;
 836         }
 837
 838         /* FIXME: use iterator macros --RR */
 839         /* ... then go back and fix counters and names */
 840         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 841                 unsigned int i;
 842                 const struct xt_entry_match *m;
 843                 const struct xt_entry_target *t;
 844
 845                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 846                 if (copy_to_user(userptr + off
 847                                  + offsetof(struct ipt_entry, counters),
 848                                  &counters[num],
 849                                  sizeof(counters[num])) != 0) {
 850                         ret = -EFAULT;
 851                         goto free_counters;
 852                 }
 853
 854                 for (i = sizeof(struct ipt_entry);
 855                      i < e->target_offset;
 856                      i += m->u.match_size) {
 857                         m = (void *)e + i;
 858
 859                         if (copy_to_user(userptr + off + i
 860                                          + offsetof(struct xt_entry_match,
 861                                                     u.user.name),
 862                                          m->u.kernel.match->name,
 863                                          strlen(m->u.kernel.match->name)+1)
 864                             != 0) {
 865                                 ret = -EFAULT;
 866                                 goto free_counters;
 867                         }
 868                 }
 869
 870                 t = ipt_get_target_c(e);
 871                 if (copy_to_user(userptr + off + e->target_offset
 872                                  + offsetof(struct xt_entry_target,
 873                                             u.user.name),
 874                                  t->u.kernel.target->name,
 875                                  strlen(t->u.kernel.target->name)+1) != 0) {
 876                         ret = -EFAULT;
 877                         goto free_counters;
 878                 }
 879         }
 880
 881  free_counters:
 882         vfree(counters);
 883         return ret;
 884 }
 885
 886 #ifdef CONFIG_COMPAT
 887 static void compat_standard_from_user(void *dst, const void *src)
 888 {
 889         int v = *(compat_int_t *)src;
 890
 891         if (v > 0)
 892                 v += xt_compat_calc_jump(AF_INET, v);
 893         memcpy(dst, &v, sizeof(v));
 894 }
 895
 896 static int compat_standard_to_user(void __user *dst, const void *src)
 897 {
 898         compat_int_t cv = *(int *)src;
 899
 900         if (cv > 0)
 901                 cv -= xt_compat_calc_jump(AF_INET, cv);
 902         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 903 }
 904
 905 static int compat_calc_entry(const struct ipt_entry *e,
 906                              const struct xt_table_info *info,
 907                              const void *base, struct xt_table_info *newinfo)
 908 {
 909         const struct xt_entry_match *ematch;
 910         const struct xt_entry_target *t;
 911         unsigned int entry_offset;
 912         int off, i, ret;
 913
 914         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
 915         entry_offset = (void *)e - base;
 916         xt_ematch_foreach(ematch, e)
 917                 off += xt_compat_match_offset(ematch->u.kernel.match);
 918         t = ipt_get_target_c(e);
 919         off += xt_compat_target_offset(t->u.kernel.target);
 920         newinfo->size -= off;
 921         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
 922         if (ret)
 923                 return ret;
 924
 925         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 926                 if (info->hook_entry[i] &&
 927                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
 928                         newinfo->hook_entry[i] -= off;
 929                 if (info->underflow[i] &&
 930                     (e < (struct ipt_entry *)(base + info->underflow[i])))
 931                         newinfo->underflow[i] -= off;
 932         }
 933         return 0;
 934 }
 935
 936 static int compat_table_info(const struct xt_table_info *info,
 937                              struct xt_table_info *newinfo)
 938 {
 939         struct ipt_entry *iter;
 940         const void *loc_cpu_entry;
 941         int ret;
 942
 943         if (!newinfo || !info)
 944                 return -EINVAL;
 945
 946         /* we dont care about newinfo->entries */
 947         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 948         newinfo->initial_entries = 0;
 949         loc_cpu_entry = info->entries;
 950         xt_compat_init_offsets(AF_INET, info->number);
 951         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 952                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 953                 if (ret != 0)
 954                         return ret;
 955         }
 956         return 0;
 957 }
 958 #endif
 959
 960 static int get_info(struct net *net, void __user *user,
 961                     const int *len, int compat)
 962 {
 963         char name[XT_TABLE_MAXNAMELEN];
 964         struct xt_table *t;
 965         int ret;
 966
 967         if (*len != sizeof(struct ipt_getinfo))
 968                 return -EINVAL;
 969
 970         if (copy_from_user(name, user, sizeof(name)) != 0)
 971                 return -EFAULT;
 972
 973         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 974 #ifdef CONFIG_COMPAT
 975         if (compat)
 976                 xt_compat_lock(AF_INET);
 977 #endif
 978         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
 979                                     "iptable_%s", name);
 980         if (!IS_ERR_OR_NULL(t)) {
 981                 struct ipt_getinfo info;
 982                 const struct xt_table_info *private = t->private;
 983 #ifdef CONFIG_COMPAT
 984                 struct xt_table_info tmp;
 985
 986                 if (compat) {
 987                         ret = compat_table_info(private, &tmp);
 988                         xt_compat_flush_offsets(AF_INET);
 989                         private = &tmp;
 990                 }
 991 #endif
 992                 memset(&info, 0, sizeof(info));
 993                 info.valid_hooks = t->valid_hooks;
 994                 memcpy(info.hook_entry, private->hook_entry,
 995                        sizeof(info.hook_entry));
 996                 memcpy(info.underflow, private->underflow,
 997                        sizeof(info.underflow));
 998                 info.num_entries = private->number;
 999                 info.size = private->size;
1000                 strcpy(info.name, name);
1001
1002                 if (copy_to_user(user, &info, *len) != 0)
1003                         ret = -EFAULT;
1004                 else
1005                         ret = 0;
1006
1007                 xt_table_unlock(t);
1008                 module_put(t->me);
1009         } else
1010                 ret = t ? PTR_ERR(t) : -ENOENT;
1011 #ifdef CONFIG_COMPAT
1012         if (compat)
1013                 xt_compat_unlock(AF_INET);
1014 #endif
1015         return ret;
1016 }
1017
1018 static int
1019 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1020             const int *len)
1021 {
1022         int ret;
1023         struct ipt_get_entries get;
1024         struct xt_table *t;
1025
1026         if (*len < sizeof(get))
1027                 return -EINVAL;
1028         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1029                 return -EFAULT;
1030         if (*len != sizeof(struct ipt_get_entries) + get.size)
1031                 return -EINVAL;
1032         get.name[sizeof(get.name) - 1] = '\0';
1033
1034         t = xt_find_table_lock(net, AF_INET, get.name);
1035         if (!IS_ERR_OR_NULL(t)) {
1036                 const struct xt_table_info *private = t->private;
1037                 if (get.size == private->size)
1038                         ret = copy_entries_to_user(private->size,
1039                                                    t, uptr->entrytable);
1040                 else
1041                         ret = -EAGAIN;
1042
1043                 module_put(t->me);
1044                 xt_table_unlock(t);
1045         } else
1046                 ret = t ? PTR_ERR(t) : -ENOENT;
1047
1048         return ret;
1049 }
1050
1051 static int
1052 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1053              struct xt_table_info *newinfo, unsigned int num_counters,
1054              void __user *counters_ptr)
1055 {
1056         int ret;
1057         struct xt_table *t;
1058         struct xt_table_info *oldinfo;
1059         struct xt_counters *counters;
1060         struct ipt_entry *iter;
1061
1062         ret = 0;
1063         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1064         if (!counters) {
1065                 ret = -ENOMEM;
1066                 goto out;
1067         }
1068
1069         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1070                                     "iptable_%s", name);
1071         if (IS_ERR_OR_NULL(t)) {
1072                 ret = t ? PTR_ERR(t) : -ENOENT;
1073                 goto free_newinfo_counters_untrans;
1074         }
1075
1076         /* You lied! */
1077         if (valid_hooks != t->valid_hooks) {
1078                 ret = -EINVAL;
1079                 goto put_module;
1080         }
1081
1082         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1083         if (!oldinfo)
1084                 goto put_module;
1085
1086         /* Update module usage count based on number of rules */
1087         if ((oldinfo->number > oldinfo->initial_entries) ||
1088             (newinfo->number <= oldinfo->initial_entries))
1089                 module_put(t->me);
1090         if ((oldinfo->number > oldinfo->initial_entries) &&
1091             (newinfo->number <= oldinfo->initial_entries))
1092                 module_put(t->me);
1093
1094         /* Get the old counters, and synchronize with replace */
1095         get_counters(oldinfo, counters);
1096
1097         /* Decrease module usage counts and free resource */
1098         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1099                 cleanup_entry(iter, net);
1100
1101         xt_free_table_info(oldinfo);
1102         if (copy_to_user(counters_ptr, counters,
1103                          sizeof(struct xt_counters) * num_counters) != 0) {
1104                 /* Silent error, can't fail, new table is already in place */
1105                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1106         }
1107         vfree(counters);
1108         xt_table_unlock(t);
1109         return ret;
1110
1111  put_module:
1112         module_put(t->me);
1113         xt_table_unlock(t);
1114  free_newinfo_counters_untrans:
1115         vfree(counters);
1116  out:
1117         return ret;
1118 }
1119
1120 static int
1121 do_replace(struct net *net, const void __user *user, unsigned int len)
1122 {
1123         int ret;
1124         struct ipt_replace tmp;
1125         struct xt_table_info *newinfo;
1126         void *loc_cpu_entry;
1127         struct ipt_entry *iter;
1128
1129         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1130                 return -EFAULT;
1131
1132         /* overflow check */
1133         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1134                 return -ENOMEM;
1135         if (tmp.num_counters == 0)
1136                 return -EINVAL;
1137
1138         tmp.name[sizeof(tmp.name)-1] = 0;
1139
1140         newinfo = xt_alloc_table_info(tmp.size);
1141         if (!newinfo)
1142                 return -ENOMEM;
1143
1144         loc_cpu_entry = newinfo->entries;
1145         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1146                            tmp.size) != 0) {
1147                 ret = -EFAULT;
1148                 goto free_newinfo;
1149         }
1150
1151         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1152         if (ret != 0)
1153                 goto free_newinfo;
1154
1155         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1156                            tmp.num_counters, tmp.counters);
1157         if (ret)
1158                 goto free_newinfo_untrans;
1159         return 0;
1160
1161  free_newinfo_untrans:
1162         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1163                 cleanup_entry(iter, net);
1164  free_newinfo:
1165         xt_free_table_info(newinfo);
1166         return ret;
1167 }
1168
1169 static int
1170 do_add_counters(struct net *net, const void __user *user,
1171                 unsigned int len, int compat)
1172 {
1173         unsigned int i;
1174         struct xt_counters_info tmp;
1175         struct xt_counters *paddc;
1176         struct xt_table *t;
1177         const struct xt_table_info *private;
1178         int ret = 0;
1179         struct ipt_entry *iter;
1180         unsigned int addend;
1181
1182         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1183         if (IS_ERR(paddc))
1184                 return PTR_ERR(paddc);
1185
1186         t = xt_find_table_lock(net, AF_INET, tmp.name);
1187         if (IS_ERR_OR_NULL(t)) {
1188                 ret = t ? PTR_ERR(t) : -ENOENT;
1189                 goto free;
1190         }
1191
1192         local_bh_disable();
1193         private = t->private;
1194         if (private->number != tmp.num_counters) {
1195                 ret = -EINVAL;
1196                 goto unlock_up_free;
1197         }
1198
1199         i = 0;
1200         addend = xt_write_recseq_begin();
1201         xt_entry_foreach(iter, private->entries, private->size) {
1202                 struct xt_counters *tmp;
1203
1204                 tmp = xt_get_this_cpu_counter(&iter->counters);
1205                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1206                 ++i;
1207         }
1208         xt_write_recseq_end(addend);
1209  unlock_up_free:
1210         local_bh_enable();
1211         xt_table_unlock(t);
1212         module_put(t->me);
1213  free:
1214         vfree(paddc);
1215
1216         return ret;
1217 }
1218
1219 #ifdef CONFIG_COMPAT
1220 struct compat_ipt_replace {
1221         char                    name[XT_TABLE_MAXNAMELEN];
1222         u32                     valid_hooks;
1223         u32                     num_entries;
1224         u32                     size;
1225         u32                     hook_entry[NF_INET_NUMHOOKS];
1226         u32                     underflow[NF_INET_NUMHOOKS];
1227         u32                     num_counters;
1228         compat_uptr_t           counters;       /* struct xt_counters * */
1229         struct compat_ipt_entry entries[0];
1230 };
1231
1232 static int
1233 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1234                           unsigned int *size, struct xt_counters *counters,
1235                           unsigned int i)
1236 {
1237         struct xt_entry_target *t;
1238         struct compat_ipt_entry __user *ce;
1239         u_int16_t target_offset, next_offset;
1240         compat_uint_t origsize;
1241         const struct xt_entry_match *ematch;
1242         int ret = 0;
1243
1244         origsize = *size;
1245         ce = (struct compat_ipt_entry __user *)*dstptr;
1246         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1247             copy_to_user(&ce->counters, &counters[i],
1248             sizeof(counters[i])) != 0)
1249                 return -EFAULT;
1250
1251         *dstptr += sizeof(struct compat_ipt_entry);
1252         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1253
1254         xt_ematch_foreach(ematch, e) {
1255                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1256                 if (ret != 0)
1257                         return ret;
1258         }
1259         target_offset = e->target_offset - (origsize - *size);
1260         t = ipt_get_target(e);
1261         ret = xt_compat_target_to_user(t, dstptr, size);
1262         if (ret)
1263                 return ret;
1264         next_offset = e->next_offset - (origsize - *size);
1265         if (put_user(target_offset, &ce->target_offset) != 0 ||
1266             put_user(next_offset, &ce->next_offset) != 0)
1267                 return -EFAULT;
1268         return 0;
1269 }
1270
1271 static int
1272 compat_find_calc_match(struct xt_entry_match *m,
1273                        const struct ipt_ip *ip,
1274                        int *size)
1275 {
1276         struct xt_match *match;
1277
1278         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1279                                       m->u.user.revision);
1280         if (IS_ERR(match))
1281                 return PTR_ERR(match);
1282
1283         m->u.kernel.match = match;
1284         *size += xt_compat_match_offset(match);
1285         return 0;
1286 }
1287
1288 static void compat_release_entry(struct compat_ipt_entry *e)
1289 {
1290         struct xt_entry_target *t;
1291         struct xt_entry_match *ematch;
1292
1293         /* Cleanup all matches */
1294         xt_ematch_foreach(ematch, e)
1295                 module_put(ematch->u.kernel.match->me);
1296         t = compat_ipt_get_target(e);
1297         module_put(t->u.kernel.target->me);
1298 }
1299
1300 static int
1301 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1302                                   struct xt_table_info *newinfo,
1303                                   unsigned int *size,
1304                                   const unsigned char *base,
1305                                   const unsigned char *limit)
1306 {
1307         struct xt_entry_match *ematch;
1308         struct xt_entry_target *t;
1309         struct xt_target *target;
1310         unsigned int entry_offset;
1311         unsigned int j;
1312         int ret, off;
1313
1314         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1315             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1316             (unsigned char *)e + e->next_offset > limit)
1317                 return -EINVAL;
1318
1319         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1320                              sizeof(struct compat_xt_entry_target))
1321                 return -EINVAL;
1322
1323         if (!ip_checkentry(&e->ip))
1324                 return -EINVAL;
1325
1326         ret = xt_compat_check_entry_offsets(e, e->elems,
1327                                             e->target_offset, e->next_offset);
1328         if (ret)
1329                 return ret;
1330
1331         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1332         entry_offset = (void *)e - (void *)base;
1333         j = 0;
1334         xt_ematch_foreach(ematch, e) {
1335                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1336                 if (ret != 0)
1337                         goto release_matches;
1338                 ++j;
1339         }
1340
1341         t = compat_ipt_get_target(e);
1342         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1343                                         t->u.user.revision);
1344         if (IS_ERR(target)) {
1345                 ret = PTR_ERR(target);
1346                 goto release_matches;
1347         }
1348         t->u.kernel.target = target;
1349
1350         off += xt_compat_target_offset(target);
1351         *size += off;
1352         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1353         if (ret)
1354                 goto out;
1355
1356         return 0;
1357
1358 out:
1359         module_put(t->u.kernel.target->me);
1360 release_matches:
1361         xt_ematch_foreach(ematch, e) {
1362                 if (j-- == 0)
1363                         break;
1364                 module_put(ematch->u.kernel.match->me);
1365         }
1366         return ret;
1367 }
1368
1369 static void
1370 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1371                             unsigned int *size,
1372                             struct xt_table_info *newinfo, unsigned char *base)
1373 {
1374         struct xt_entry_target *t;
1375         struct xt_target *target;
1376         struct ipt_entry *de;
1377         unsigned int origsize;
1378         int h;
1379         struct xt_entry_match *ematch;
1380
1381         origsize = *size;
1382         de = (struct ipt_entry *)*dstptr;
1383         memcpy(de, e, sizeof(struct ipt_entry));
1384         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1385
1386         *dstptr += sizeof(struct ipt_entry);
1387         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1388
1389         xt_ematch_foreach(ematch, e)
1390                 xt_compat_match_from_user(ematch, dstptr, size);
1391
1392         de->target_offset = e->target_offset - (origsize - *size);
1393         t = compat_ipt_get_target(e);
1394         target = t->u.kernel.target;
1395         xt_compat_target_from_user(t, dstptr, size);
1396
1397         de->next_offset = e->next_offset - (origsize - *size);
1398
1399         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1400                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1401                         newinfo->hook_entry[h] -= origsize - *size;
1402                 if ((unsigned char *)de - base < newinfo->underflow[h])
1403                         newinfo->underflow[h] -= origsize - *size;
1404         }
1405 }
1406
1407 static int
1408 translate_compat_table(struct net *net,
1409                        struct xt_table_info **pinfo,
1410                        void **pentry0,
1411                        const struct compat_ipt_replace *compatr)
1412 {
1413         unsigned int i, j;
1414         struct xt_table_info *newinfo, *info;
1415         void *pos, *entry0, *entry1;
1416         struct compat_ipt_entry *iter0;
1417         struct ipt_replace repl;
1418         unsigned int size;
1419         int ret;
1420
1421         info = *pinfo;
1422         entry0 = *pentry0;
1423         size = compatr->size;
1424         info->number = compatr->num_entries;
1425
1426         j = 0;
1427         xt_compat_lock(AF_INET);
1428         xt_compat_init_offsets(AF_INET, compatr->num_entries);
1429         /* Walk through entries, checking offsets. */
1430         xt_entry_foreach(iter0, entry0, compatr->size) {
1431                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1432                                                         entry0,
1433                                                         entry0 + compatr->size);
1434                 if (ret != 0)
1435                         goto out_unlock;
1436                 ++j;
1437         }
1438
1439         ret = -EINVAL;
1440         if (j != compatr->num_entries)
1441                 goto out_unlock;
1442
1443         ret = -ENOMEM;
1444         newinfo = xt_alloc_table_info(size);
1445         if (!newinfo)
1446                 goto out_unlock;
1447
1448         newinfo->number = compatr->num_entries;
1449         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1450                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1451                 newinfo->underflow[i] = compatr->underflow[i];
1452         }
1453         entry1 = newinfo->entries;
1454         pos = entry1;
1455         size = compatr->size;
1456         xt_entry_foreach(iter0, entry0, compatr->size)
1457                 compat_copy_entry_from_user(iter0, &pos, &size,
1458                                             newinfo, entry1);
1459
1460         /* all module references in entry0 are now gone.
1461          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1462          * generated by 64bit userspace.
1463          *
1464          * Call standard translate_table() to validate all hook_entrys,
1465          * underflows, check for loops, etc.
1466          */
1467         xt_compat_flush_offsets(AF_INET);
1468         xt_compat_unlock(AF_INET);
1469
1470         memcpy(&repl, compatr, sizeof(*compatr));
1471
1472         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1473                 repl.hook_entry[i] = newinfo->hook_entry[i];
1474                 repl.underflow[i] = newinfo->underflow[i];
1475         }
1476
1477         repl.num_counters = 0;
1478         repl.counters = NULL;
1479         repl.size = newinfo->size;
1480         ret = translate_table(net, newinfo, entry1, &repl);
1481         if (ret)
1482                 goto free_newinfo;
1483
1484         *pinfo = newinfo;
1485         *pentry0 = entry1;
1486         xt_free_table_info(info);
1487         return 0;
1488
1489 free_newinfo:
1490         xt_free_table_info(newinfo);
1491         return ret;
1492 out_unlock:
1493         xt_compat_flush_offsets(AF_INET);
1494         xt_compat_unlock(AF_INET);
1495         xt_entry_foreach(iter0, entry0, compatr->size) {
1496                 if (j-- == 0)
1497                         break;
1498                 compat_release_entry(iter0);
1499         }
1500         return ret;
1501 }
1502
1503 static int
1504 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1505 {
1506         int ret;
1507         struct compat_ipt_replace tmp;
1508         struct xt_table_info *newinfo;
1509         void *loc_cpu_entry;
1510         struct ipt_entry *iter;
1511
1512         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1513                 return -EFAULT;
1514
1515         /* overflow check */
1516         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1517                 return -ENOMEM;
1518         if (tmp.num_counters == 0)
1519                 return -EINVAL;
1520
1521         tmp.name[sizeof(tmp.name)-1] = 0;
1522
1523         newinfo = xt_alloc_table_info(tmp.size);
1524         if (!newinfo)
1525                 return -ENOMEM;
1526
1527         loc_cpu_entry = newinfo->entries;
1528         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1529                            tmp.size) != 0) {
1530                 ret = -EFAULT;
1531                 goto free_newinfo;
1532         }
1533
1534         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1535         if (ret != 0)
1536                 goto free_newinfo;
1537
1538         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1539                            tmp.num_counters, compat_ptr(tmp.counters));
1540         if (ret)
1541                 goto free_newinfo_untrans;
1542         return 0;
1543
1544  free_newinfo_untrans:
1545         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1546                 cleanup_entry(iter, net);
1547  free_newinfo:
1548         xt_free_table_info(newinfo);
1549         return ret;
1550 }
1551
1552 static int
1553 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1554                       unsigned int len)
1555 {
1556         int ret;
1557
1558         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1559                 return -EPERM;
1560
1561         switch (cmd) {
1562         case IPT_SO_SET_REPLACE:
1563                 ret = compat_do_replace(sock_net(sk), user, len);
1564                 break;
1565
1566         case IPT_SO_SET_ADD_COUNTERS:
1567                 ret = do_add_counters(sock_net(sk), user, len, 1);
1568                 break;
1569
1570         default:
1571                 ret = -EINVAL;
1572         }
1573
1574         return ret;
1575 }
1576
1577 struct compat_ipt_get_entries {
1578         char name[XT_TABLE_MAXNAMELEN];
1579         compat_uint_t size;
1580         struct compat_ipt_entry entrytable[0];
1581 };
1582
1583 static int
1584 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1585                             void __user *userptr)
1586 {
1587         struct xt_counters *counters;
1588         const struct xt_table_info *private = table->private;
1589         void __user *pos;
1590         unsigned int size;
1591         int ret = 0;
1592         unsigned int i = 0;
1593         struct ipt_entry *iter;
1594
1595         counters = alloc_counters(table);
1596         if (IS_ERR(counters))
1597                 return PTR_ERR(counters);
1598
1599         pos = userptr;
1600         size = total_size;
1601         xt_entry_foreach(iter, private->entries, total_size) {
1602                 ret = compat_copy_entry_to_user(iter, &pos,
1603                                                 &size, counters, i++);
1604                 if (ret != 0)
1605                         break;
1606         }
1607
1608         vfree(counters);
1609         return ret;
1610 }
1611
1612 static int
1613 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1614                    int *len)
1615 {
1616         int ret;
1617         struct compat_ipt_get_entries get;
1618         struct xt_table *t;
1619
1620         if (*len < sizeof(get))
1621                 return -EINVAL;
1622
1623         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1624                 return -EFAULT;
1625
1626         if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1627                 return -EINVAL;
1628
1629         get.name[sizeof(get.name) - 1] = '\0';
1630
1631         xt_compat_lock(AF_INET);
1632         t = xt_find_table_lock(net, AF_INET, get.name);
1633         if (!IS_ERR_OR_NULL(t)) {
1634                 const struct xt_table_info *private = t->private;
1635                 struct xt_table_info info;
1636                 ret = compat_table_info(private, &info);
1637                 if (!ret && get.size == info.size)
1638                         ret = compat_copy_entries_to_user(private->size,
1639                                                           t, uptr->entrytable);
1640                 else if (!ret)
1641                         ret = -EAGAIN;
1642
1643                 xt_compat_flush_offsets(AF_INET);
1644                 module_put(t->me);
1645                 xt_table_unlock(t);
1646         } else
1647                 ret = t ? PTR_ERR(t) : -ENOENT;
1648
1649         xt_compat_unlock(AF_INET);
1650         return ret;
1651 }
1652
1653 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1654
1655 static int
1656 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1657 {
1658         int ret;
1659
1660         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1661                 return -EPERM;
1662
1663         switch (cmd) {
1664         case IPT_SO_GET_INFO:
1665                 ret = get_info(sock_net(sk), user, len, 1);
1666                 break;
1667         case IPT_SO_GET_ENTRIES:
1668                 ret = compat_get_entries(sock_net(sk), user, len);
1669                 break;
1670         default:
1671                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1672         }
1673         return ret;
1674 }
1675 #endif
1676
1677 static int
1678 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1679 {
1680         int ret;
1681
1682         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1683                 return -EPERM;
1684
1685         switch (cmd) {
1686         case IPT_SO_SET_REPLACE:
1687                 ret = do_replace(sock_net(sk), user, len);
1688                 break;
1689
1690         case IPT_SO_SET_ADD_COUNTERS:
1691                 ret = do_add_counters(sock_net(sk), user, len, 0);
1692                 break;
1693
1694         default:
1695                 ret = -EINVAL;
1696         }
1697
1698         return ret;
1699 }
1700
1701 static int
1702 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1703 {
1704         int ret;
1705
1706         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1707                 return -EPERM;
1708
1709         switch (cmd) {
1710         case IPT_SO_GET_INFO:
1711                 ret = get_info(sock_net(sk), user, len, 0);
1712                 break;
1713
1714         case IPT_SO_GET_ENTRIES:
1715                 ret = get_entries(sock_net(sk), user, len);
1716                 break;
1717
1718         case IPT_SO_GET_REVISION_MATCH:
1719         case IPT_SO_GET_REVISION_TARGET: {
1720                 struct xt_get_revision rev;
1721                 int target;
1722
1723                 if (*len != sizeof(rev)) {
1724                         ret = -EINVAL;
1725                         break;
1726                 }
1727                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1728                         ret = -EFAULT;
1729                         break;
1730                 }
1731                 rev.name[sizeof(rev.name)-1] = 0;
1732
1733                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1734                         target = 1;
1735                 else
1736                         target = 0;
1737
1738                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1739                                                          rev.revision,
1740                                                          target, &ret),
1741                                         "ipt_%s", rev.name);
1742                 break;
1743         }
1744
1745         default:
1746                 ret = -EINVAL;
1747         }
1748
1749         return ret;
1750 }
1751
1752 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1753 {
1754         struct xt_table_info *private;
1755         void *loc_cpu_entry;
1756         struct module *table_owner = table->me;
1757         struct ipt_entry *iter;
1758
1759         private = xt_unregister_table(table);
1760
1761         /* Decrease module usage counts and free resources */
1762         loc_cpu_entry = private->entries;
1763         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1764                 cleanup_entry(iter, net);
1765         if (private->number > private->initial_entries)
1766                 module_put(table_owner);
1767         xt_free_table_info(private);
1768 }
1769
1770 int ipt_register_table(struct net *net, const struct xt_table *table,
1771                        const struct ipt_replace *repl,
1772                        const struct nf_hook_ops *ops, struct xt_table **res)
1773 {
1774         int ret;
1775         struct xt_table_info *newinfo;
1776         struct xt_table_info bootstrap = {0};
1777         void *loc_cpu_entry;
1778         struct xt_table *new_table;
1779
1780         newinfo = xt_alloc_table_info(repl->size);
1781         if (!newinfo)
1782                 return -ENOMEM;
1783
1784         loc_cpu_entry = newinfo->entries;
1785         memcpy(loc_cpu_entry, repl->entries, repl->size);
1786
1787         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1788         if (ret != 0)
1789                 goto out_free;
1790
1791         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1792         if (IS_ERR(new_table)) {
1793                 ret = PTR_ERR(new_table);
1794                 goto out_free;
1795         }
1796
1797         /* set res now, will see skbs right after nf_register_net_hooks */
1798         WRITE_ONCE(*res, new_table);
1799
1800         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1801         if (ret != 0) {
1802                 __ipt_unregister_table(net, new_table);
1803                 *res = NULL;
1804         }
1805
1806         return ret;
1807
1808 out_free:
1809         xt_free_table_info(newinfo);
1810         return ret;
1811 }
1812
1813 void ipt_unregister_table(struct net *net, struct xt_table *table,
1814                           const struct nf_hook_ops *ops)
1815 {
1816         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1817         __ipt_unregister_table(net, table);
1818 }
1819
1820 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1821 static inline bool
1822 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1823                      u_int8_t type, u_int8_t code,
1824                      bool invert)
1825 {
1826         return ((test_type == 0xFF) ||
1827                 (type == test_type && code >= min_code && code <= max_code))
1828                 ^ invert;
1829 }
1830
1831 static bool
1832 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
1833 {
1834         const struct icmphdr *ic;
1835         struct icmphdr _icmph;
1836         const struct ipt_icmp *icmpinfo = par->matchinfo;
1837
1838         /* Must not be a fragment. */
1839         if (par->fragoff != 0)
1840                 return false;
1841
1842         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1843         if (ic == NULL) {
1844                 /* We've been asked to examine this packet, and we
1845                  * can't.  Hence, no choice but to drop.
1846                  */
1847                 par->hotdrop = true;
1848                 return false;
1849         }
1850
1851         return icmp_type_code_match(icmpinfo->type,
1852                                     icmpinfo->code[0],
1853                                     icmpinfo->code[1],
1854                                     ic->type, ic->code,
1855                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
1856 }
1857
1858 static int icmp_checkentry(const struct xt_mtchk_param *par)
1859 {
1860         const struct ipt_icmp *icmpinfo = par->matchinfo;
1861
1862         /* Must specify no unknown invflags */
1863         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
1864 }
1865
1866 static struct xt_target ipt_builtin_tg[] __read_mostly = {
1867         {
1868                 .name             = XT_STANDARD_TARGET,
1869                 .targetsize       = sizeof(int),
1870                 .family           = NFPROTO_IPV4,
1871 #ifdef CONFIG_COMPAT
1872                 .compatsize       = sizeof(compat_int_t),
1873                 .compat_from_user = compat_standard_from_user,
1874                 .compat_to_user   = compat_standard_to_user,
1875 #endif
1876         },
1877         {
1878                 .name             = XT_ERROR_TARGET,
1879                 .target           = ipt_error,
1880                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1881                 .family           = NFPROTO_IPV4,
1882         },
1883 };
1884
1885 static struct nf_sockopt_ops ipt_sockopts = {
1886         .pf             = PF_INET,
1887         .set_optmin     = IPT_BASE_CTL,
1888         .set_optmax     = IPT_SO_SET_MAX+1,
1889         .set            = do_ipt_set_ctl,
1890 #ifdef CONFIG_COMPAT
1891         .compat_set     = compat_do_ipt_set_ctl,
1892 #endif
1893         .get_optmin     = IPT_BASE_CTL,
1894         .get_optmax     = IPT_SO_GET_MAX+1,
1895         .get            = do_ipt_get_ctl,
1896 #ifdef CONFIG_COMPAT
1897         .compat_get     = compat_do_ipt_get_ctl,
1898 #endif
1899         .owner          = THIS_MODULE,
1900 };
1901
1902 static struct xt_match ipt_builtin_mt[] __read_mostly = {
1903         {
1904                 .name       = "icmp",
1905                 .match      = icmp_match,
1906                 .matchsize  = sizeof(struct ipt_icmp),
1907                 .checkentry = icmp_checkentry,
1908                 .proto      = IPPROTO_ICMP,
1909                 .family     = NFPROTO_IPV4,
1910         },
1911 };
1912
1913 static int __net_init ip_tables_net_init(struct net *net)
1914 {
1915         return xt_proto_init(net, NFPROTO_IPV4);
1916 }
1917
1918 static void __net_exit ip_tables_net_exit(struct net *net)
1919 {
1920         xt_proto_fini(net, NFPROTO_IPV4);
1921 }
1922
1923 static struct pernet_operations ip_tables_net_ops = {
1924         .init = ip_tables_net_init,
1925         .exit = ip_tables_net_exit,
1926 };
1927
1928 static int __init ip_tables_init(void)
1929 {
1930         int ret;
1931
1932         ret = register_pernet_subsys(&ip_tables_net_ops);
1933         if (ret < 0)
1934                 goto err1;
1935
1936         /* No one else will be downing sem now, so we won't sleep */
1937         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1938         if (ret < 0)
1939                 goto err2;
1940         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1941         if (ret < 0)
1942                 goto err4;
1943
1944         /* Register setsockopt */
1945         ret = nf_register_sockopt(&ipt_sockopts);
1946         if (ret < 0)
1947                 goto err5;
1948
1949         pr_info("(C) 2000-2006 Netfilter Core Team\n");
1950         return 0;
1951
1952 err5:
1953         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1954 err4:
1955         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1956 err2:
1957         unregister_pernet_subsys(&ip_tables_net_ops);
1958 err1:
1959         return ret;
1960 }
1961
1962 static void __exit ip_tables_fini(void)
1963 {
1964         nf_unregister_sockopt(&ipt_sockopts);
1965
1966         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1967         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1968         unregister_pernet_subsys(&ip_tables_net_ops);
1969 }
1970
1971 EXPORT_SYMBOL(ipt_register_table);
1972 EXPORT_SYMBOL(ipt_unregister_table);
1973 EXPORT_SYMBOL(ipt_do_table);
1974 module_init(ip_tables_init);
1975 module_exit(ip_tables_fini);