net/ipv4/netfilter/ip_tables.c

   1 /*
   2  * Packet matching code.
   3  *
   4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   6  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13 #include <linux/cache.h>
  14 #include <linux/capability.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/kmod.h>
  17 #include <linux/vmalloc.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/module.h>
  20 #include <linux/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/compat.h>
  23 #include <linux/uaccess.h>
  24 #include <linux/mutex.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/err.h>
  27 #include <linux/cpumask.h>
  28
  29 #include <linux/netfilter/x_tables.h>
  30 #include <linux/netfilter_ipv4/ip_tables.h>
  31 #include <net/netfilter/nf_log.h>
  32 #include "../../netfilter/xt_repldata.h"
  33
  34 MODULE_LICENSE("GPL");
  35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36 MODULE_DESCRIPTION("IPv4 packet filter");
  37
  38 void *ipt_alloc_initial_table(const struct xt_table *info)
  39 {
  40         return xt_alloc_initial_table(ipt, IPT);
  41 }
  42 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  43
  44 /* Returns whether matches rule or not. */
  45 /* Performance critical - called for every packet */
  46 static inline bool
  47 ip_packet_match(const struct iphdr *ip,
  48                 const char *indev,
  49                 const char *outdev,
  50                 const struct ipt_ip *ipinfo,
  51                 int isfrag)
  52 {
  53         unsigned long ret;
  54
  55         if (NF_INVF(ipinfo, IPT_INV_SRCIP,
  56                     (ip->saddr & ipinfo->smsk.s_addr) != ipinfo->src.s_addr) ||
  57             NF_INVF(ipinfo, IPT_INV_DSTIP,
  58                     (ip->daddr & ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr))
  59                 return false;
  60
  61         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
  62
  63         if (NF_INVF(ipinfo, IPT_INV_VIA_IN, ret != 0))
  64                 return false;
  65
  66         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
  67
  68         if (NF_INVF(ipinfo, IPT_INV_VIA_OUT, ret != 0))
  69                 return false;
  70
  71         /* Check specific protocol */
  72         if (ipinfo->proto &&
  73             NF_INVF(ipinfo, IPT_INV_PROTO, ip->protocol != ipinfo->proto))
  74                 return false;
  75
  76         /* If we have a fragment rule but the packet is not a fragment
  77          * then we return zero */
  78         if (NF_INVF(ipinfo, IPT_INV_FRAG,
  79                     (ipinfo->flags & IPT_F_FRAG) && !isfrag))
  80                 return false;
  81
  82         return true;
  83 }
  84
  85 static bool
  86 ip_checkentry(const struct ipt_ip *ip)
  87 {
  88         if (ip->flags & ~IPT_F_MASK)
  89                 return false;
  90         if (ip->invflags & ~IPT_INV_MASK)
  91                 return false;
  92         return true;
  93 }
  94
  95 static unsigned int
  96 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
  97 {
  98         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
  99
 100         return NF_DROP;
 101 }
 102
 103 /* Performance critical */
 104 static inline struct ipt_entry *
 105 get_entry(const void *base, unsigned int offset)
 106 {
 107         return (struct ipt_entry *)(base + offset);
 108 }
 109
 110 /* All zeroes == unconditional rule. */
 111 /* Mildly perf critical (only if packet tracing is on) */
 112 static inline bool unconditional(const struct ipt_entry *e)
 113 {
 114         static const struct ipt_ip uncond;
 115
 116         return e->target_offset == sizeof(struct ipt_entry) &&
 117                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
 118 }
 119
 120 /* for const-correctness */
 121 static inline const struct xt_entry_target *
 122 ipt_get_target_c(const struct ipt_entry *e)
 123 {
 124         return ipt_get_target((struct ipt_entry *)e);
 125 }
 126
 127 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 128 static const char *const hooknames[] = {
 129         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 130         [NF_INET_LOCAL_IN]              = "INPUT",
 131         [NF_INET_FORWARD]               = "FORWARD",
 132         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 133         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 134 };
 135
 136 enum nf_ip_trace_comments {
 137         NF_IP_TRACE_COMMENT_RULE,
 138         NF_IP_TRACE_COMMENT_RETURN,
 139         NF_IP_TRACE_COMMENT_POLICY,
 140 };
 141
 142 static const char *const comments[] = {
 143         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 144         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 145         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 146 };
 147
 148 static const struct nf_loginfo trace_loginfo = {
 149         .type = NF_LOG_TYPE_LOG,
 150         .u = {
 151                 .log = {
 152                         .level = 4,
 153                         .logflags = NF_LOG_DEFAULT_MASK,
 154                 },
 155         },
 156 };
 157
 158 /* Mildly perf critical (only if packet tracing is on) */
 159 static inline int
 160 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 161                       const char *hookname, const char **chainname,
 162                       const char **comment, unsigned int *rulenum)
 163 {
 164         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 165
 166         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 167                 /* Head of user chain: ERROR target with chainname */
 168                 *chainname = t->target.data;
 169                 (*rulenum) = 0;
 170         } else if (s == e) {
 171                 (*rulenum)++;
 172
 173                 if (unconditional(s) &&
 174                     strcmp(t->target.u.kernel.target->name,
 175                            XT_STANDARD_TARGET) == 0 &&
 176                    t->verdict < 0) {
 177                         /* Tail of chains: STANDARD target (return/policy) */
 178                         *comment = *chainname == hookname
 179                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 180                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 181                 }
 182                 return 1;
 183         } else
 184                 (*rulenum)++;
 185
 186         return 0;
 187 }
 188
 189 static void trace_packet(struct net *net,
 190                          const struct sk_buff *skb,
 191                          unsigned int hook,
 192                          const struct net_device *in,
 193                          const struct net_device *out,
 194                          const char *tablename,
 195                          const struct xt_table_info *private,
 196                          const struct ipt_entry *e)
 197 {
 198         const struct ipt_entry *root;
 199         const char *hookname, *chainname, *comment;
 200         const struct ipt_entry *iter;
 201         unsigned int rulenum = 0;
 202
 203         root = get_entry(private->entries, private->hook_entry[hook]);
 204
 205         hookname = chainname = hooknames[hook];
 206         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 207
 208         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 209                 if (get_chainname_rulenum(iter, e, hookname,
 210                     &chainname, &comment, &rulenum) != 0)
 211                         break;
 212
 213         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 214                      "TRACE: %s:%s:%s:%u ",
 215                      tablename, chainname, comment, rulenum);
 216 }
 217 #endif
 218
 219 static inline
 220 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 221 {
 222         return (void *)entry + entry->next_offset;
 223 }
 224
 225 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 226 unsigned int
 227 ipt_do_table(struct sk_buff *skb,
 228              const struct nf_hook_state *state,
 229              struct xt_table *table)
 230 {
 231         unsigned int hook = state->hook;
 232         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 233         const struct iphdr *ip;
 234         /* Initializing verdict to NF_DROP keeps gcc happy. */
 235         unsigned int verdict = NF_DROP;
 236         const char *indev, *outdev;
 237         const void *table_base;
 238         struct ipt_entry *e, **jumpstack;
 239         unsigned int stackidx, cpu;
 240         const struct xt_table_info *private;
 241         struct xt_action_param acpar;
 242         unsigned int addend;
 243
 244         /* Initialization */
 245         stackidx = 0;
 246         ip = ip_hdr(skb);
 247         indev = state->in ? state->in->name : nulldevname;
 248         outdev = state->out ? state->out->name : nulldevname;
 249         /* We handle fragments by dealing with the first fragment as
 250          * if it was a normal packet.  All other fragments are treated
 251          * normally, except that they will NEVER match rules that ask
 252          * things we don't know, ie. tcp syn flag or ports).  If the
 253          * rule is also a fragment-specific rule, non-fragments won't
 254          * match it. */
 255         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 256         acpar.thoff   = ip_hdrlen(skb);
 257         acpar.hotdrop = false;
 258         acpar.state   = state;
 259
 260         WARN_ON(!(table->valid_hooks & (1 << hook)));
 261         local_bh_disable();
 262         addend = xt_write_recseq_begin();
 263         private = READ_ONCE(table->private); /* Address dependency. */
 264         cpu        = smp_processor_id();
 265         table_base = private->entries;
 266         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 267
 268         /* Switch to alternate jumpstack if we're being invoked via TEE.
 269          * TEE issues XT_CONTINUE verdict on original skb so we must not
 270          * clobber the jumpstack.
 271          *
 272          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 273          * but it is no problem since absolute verdict is issued by these.
 274          */
 275         if (static_key_false(&xt_tee_enabled))
 276                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 277
 278         e = get_entry(table_base, private->hook_entry[hook]);
 279
 280         do {
 281                 const struct xt_entry_target *t;
 282                 const struct xt_entry_match *ematch;
 283                 struct xt_counters *counter;
 284
 285                 WARN_ON(!e);
 286                 if (!ip_packet_match(ip, indev, outdev,
 287                     &e->ip, acpar.fragoff)) {
 288  no_match:
 289                         e = ipt_next_entry(e);
 290                         continue;
 291                 }
 292
 293                 xt_ematch_foreach(ematch, e) {
 294                         acpar.match     = ematch->u.kernel.match;
 295                         acpar.matchinfo = ematch->data;
 296                         if (!acpar.match->match(skb, &acpar))
 297                                 goto no_match;
 298                 }
 299
 300                 counter = xt_get_this_cpu_counter(&e->counters);
 301                 ADD_COUNTER(*counter, skb->len, 1);
 302
 303                 t = ipt_get_target(e);
 304                 WARN_ON(!t->u.kernel.target);
 305
 306 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 307                 /* The packet is traced: log it */
 308                 if (unlikely(skb->nf_trace))
 309                         trace_packet(state->net, skb, hook, state->in,
 310                                      state->out, table->name, private, e);
 311 #endif
 312                 /* Standard target? */
 313                 if (!t->u.kernel.target->target) {
 314                         int v;
 315
 316                         v = ((struct xt_standard_target *)t)->verdict;
 317                         if (v < 0) {
 318                                 /* Pop from stack? */
 319                                 if (v != XT_RETURN) {
 320                                         verdict = (unsigned int)(-v) - 1;
 321                                         break;
 322                                 }
 323                                 if (stackidx == 0) {
 324                                         e = get_entry(table_base,
 325                                             private->underflow[hook]);
 326                                 } else {
 327                                         e = jumpstack[--stackidx];
 328                                         e = ipt_next_entry(e);
 329                                 }
 330                                 continue;
 331                         }
 332                         if (table_base + v != ipt_next_entry(e) &&
 333                             !(e->ip.flags & IPT_F_GOTO))
 334                                 jumpstack[stackidx++] = e;
 335
 336                         e = get_entry(table_base, v);
 337                         continue;
 338                 }
 339
 340                 acpar.target   = t->u.kernel.target;
 341                 acpar.targinfo = t->data;
 342
 343                 verdict = t->u.kernel.target->target(skb, &acpar);
 344                 if (verdict == XT_CONTINUE) {
 345                         /* Target might have changed stuff. */
 346                         ip = ip_hdr(skb);
 347                         e = ipt_next_entry(e);
 348                 } else {
 349                         /* Verdict */
 350                         break;
 351                 }
 352         } while (!acpar.hotdrop);
 353
 354         xt_write_recseq_end(addend);
 355         local_bh_enable();
 356
 357         if (acpar.hotdrop)
 358                 return NF_DROP;
 359         else return verdict;
 360 }
 361
 362 /* Figures out from what hook each rule can be called: returns 0 if
 363    there are loops.  Puts hook bitmask in comefrom. */
 364 static int
 365 mark_source_chains(const struct xt_table_info *newinfo,
 366                    unsigned int valid_hooks, void *entry0,
 367                    unsigned int *offsets)
 368 {
 369         unsigned int hook;
 370
 371         /* No recursion; use packet counter to save back ptrs (reset
 372            to 0 as we leave), and comefrom to save source hook bitmask */
 373         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 374                 unsigned int pos = newinfo->hook_entry[hook];
 375                 struct ipt_entry *e = entry0 + pos;
 376
 377                 if (!(valid_hooks & (1 << hook)))
 378                         continue;
 379
 380                 /* Set initial back pointer. */
 381                 e->counters.pcnt = pos;
 382
 383                 for (;;) {
 384                         const struct xt_standard_target *t
 385                                 = (void *)ipt_get_target_c(e);
 386                         int visited = e->comefrom & (1 << hook);
 387
 388                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 389                                 return 0;
 390
 391                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 392
 393                         /* Unconditional return/END. */
 394                         if ((unconditional(e) &&
 395                              (strcmp(t->target.u.user.name,
 396                                      XT_STANDARD_TARGET) == 0) &&
 397                              t->verdict < 0) || visited) {
 398                                 unsigned int oldpos, size;
 399
 400                                 if ((strcmp(t->target.u.user.name,
 401                                             XT_STANDARD_TARGET) == 0) &&
 402                                     t->verdict < -NF_MAX_VERDICT - 1)
 403                                         return 0;
 404
 405                                 /* Return: backtrack through the last
 406                                    big jump. */
 407                                 do {
 408                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 409                                         oldpos = pos;
 410                                         pos = e->counters.pcnt;
 411                                         e->counters.pcnt = 0;
 412
 413                                         /* We're at the start. */
 414                                         if (pos == oldpos)
 415                                                 goto next;
 416
 417                                         e = entry0 + pos;
 418                                 } while (oldpos == pos + e->next_offset);
 419
 420                                 /* Move along one */
 421                                 size = e->next_offset;
 422                                 e = entry0 + pos + size;
 423                                 if (pos + size >= newinfo->size)
 424                                         return 0;
 425                                 e->counters.pcnt = pos;
 426                                 pos += size;
 427                         } else {
 428                                 int newpos = t->verdict;
 429
 430                                 if (strcmp(t->target.u.user.name,
 431                                            XT_STANDARD_TARGET) == 0 &&
 432                                     newpos >= 0) {
 433                                         /* This a jump; chase it. */
 434                                         if (!xt_find_jump_offset(offsets, newpos,
 435                                                                  newinfo->number))
 436                                                 return 0;
 437                                 } else {
 438                                         /* ... this is a fallthru */
 439                                         newpos = pos + e->next_offset;
 440                                         if (newpos >= newinfo->size)
 441                                                 return 0;
 442                                 }
 443                                 e = entry0 + newpos;
 444                                 e->counters.pcnt = pos;
 445                                 pos = newpos;
 446                         }
 447                 }
 448 next:           ;
 449         }
 450         return 1;
 451 }
 452
 453 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 454 {
 455         struct xt_mtdtor_param par;
 456
 457         par.net       = net;
 458         par.match     = m->u.kernel.match;
 459         par.matchinfo = m->data;
 460         par.family    = NFPROTO_IPV4;
 461         if (par.match->destroy != NULL)
 462                 par.match->destroy(&par);
 463         module_put(par.match->me);
 464 }
 465
 466 static int
 467 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 468 {
 469         const struct ipt_ip *ip = par->entryinfo;
 470
 471         par->match     = m->u.kernel.match;
 472         par->matchinfo = m->data;
 473
 474         return xt_check_match(par, m->u.match_size - sizeof(*m),
 475                               ip->proto, ip->invflags & IPT_INV_PROTO);
 476 }
 477
 478 static int
 479 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 480 {
 481         struct xt_match *match;
 482         int ret;
 483
 484         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 485                                       m->u.user.revision);
 486         if (IS_ERR(match))
 487                 return PTR_ERR(match);
 488         m->u.kernel.match = match;
 489
 490         ret = check_match(m, par);
 491         if (ret)
 492                 goto err;
 493
 494         return 0;
 495 err:
 496         module_put(m->u.kernel.match->me);
 497         return ret;
 498 }
 499
 500 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 501 {
 502         struct xt_entry_target *t = ipt_get_target(e);
 503         struct xt_tgchk_param par = {
 504                 .net       = net,
 505                 .table     = name,
 506                 .entryinfo = e,
 507                 .target    = t->u.kernel.target,
 508                 .targinfo  = t->data,
 509                 .hook_mask = e->comefrom,
 510                 .family    = NFPROTO_IPV4,
 511         };
 512
 513         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 514                                e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 515 }
 516
 517 static int
 518 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 519                  unsigned int size,
 520                  struct xt_percpu_counter_alloc_state *alloc_state)
 521 {
 522         struct xt_entry_target *t;
 523         struct xt_target *target;
 524         int ret;
 525         unsigned int j;
 526         struct xt_mtchk_param mtpar;
 527         struct xt_entry_match *ematch;
 528
 529         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 530                 return -ENOMEM;
 531
 532         j = 0;
 533         mtpar.net       = net;
 534         mtpar.table     = name;
 535         mtpar.entryinfo = &e->ip;
 536         mtpar.hook_mask = e->comefrom;
 537         mtpar.family    = NFPROTO_IPV4;
 538         xt_ematch_foreach(ematch, e) {
 539                 ret = find_check_match(ematch, &mtpar);
 540                 if (ret != 0)
 541                         goto cleanup_matches;
 542                 ++j;
 543         }
 544
 545         t = ipt_get_target(e);
 546         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 547                                         t->u.user.revision);
 548         if (IS_ERR(target)) {
 549                 ret = PTR_ERR(target);
 550                 goto cleanup_matches;
 551         }
 552         t->u.kernel.target = target;
 553
 554         ret = check_target(e, net, name);
 555         if (ret)
 556                 goto err;
 557
 558         return 0;
 559  err:
 560         module_put(t->u.kernel.target->me);
 561  cleanup_matches:
 562         xt_ematch_foreach(ematch, e) {
 563                 if (j-- == 0)
 564                         break;
 565                 cleanup_match(ematch, net);
 566         }
 567
 568         xt_percpu_counter_free(&e->counters);
 569
 570         return ret;
 571 }
 572
 573 static bool check_underflow(const struct ipt_entry *e)
 574 {
 575         const struct xt_entry_target *t;
 576         unsigned int verdict;
 577
 578         if (!unconditional(e))
 579                 return false;
 580         t = ipt_get_target_c(e);
 581         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 582                 return false;
 583         verdict = ((struct xt_standard_target *)t)->verdict;
 584         verdict = -verdict - 1;
 585         return verdict == NF_DROP || verdict == NF_ACCEPT;
 586 }
 587
 588 static int
 589 check_entry_size_and_hooks(struct ipt_entry *e,
 590                            struct xt_table_info *newinfo,
 591                            const unsigned char *base,
 592                            const unsigned char *limit,
 593                            const unsigned int *hook_entries,
 594                            const unsigned int *underflows,
 595                            unsigned int valid_hooks)
 596 {
 597         unsigned int h;
 598         int err;
 599
 600         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 601             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 602             (unsigned char *)e + e->next_offset > limit)
 603                 return -EINVAL;
 604
 605         if (e->next_offset
 606             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
 607                 return -EINVAL;
 608
 609         if (!ip_checkentry(&e->ip))
 610                 return -EINVAL;
 611
 612         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 613                                      e->next_offset);
 614         if (err)
 615                 return err;
 616
 617         /* Check hooks & underflows */
 618         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 619                 if (!(valid_hooks & (1 << h)))
 620                         continue;
 621                 if ((unsigned char *)e - base == hook_entries[h])
 622                         newinfo->hook_entry[h] = hook_entries[h];
 623                 if ((unsigned char *)e - base == underflows[h]) {
 624                         if (!check_underflow(e))
 625                                 return -EINVAL;
 626
 627                         newinfo->underflow[h] = underflows[h];
 628                 }
 629         }
 630
 631         /* Clear counters and comefrom */
 632         e->counters = ((struct xt_counters) { 0, 0 });
 633         e->comefrom = 0;
 634         return 0;
 635 }
 636
 637 static void
 638 cleanup_entry(struct ipt_entry *e, struct net *net)
 639 {
 640         struct xt_tgdtor_param par;
 641         struct xt_entry_target *t;
 642         struct xt_entry_match *ematch;
 643
 644         /* Cleanup all matches */
 645         xt_ematch_foreach(ematch, e)
 646                 cleanup_match(ematch, net);
 647         t = ipt_get_target(e);
 648
 649         par.net      = net;
 650         par.target   = t->u.kernel.target;
 651         par.targinfo = t->data;
 652         par.family   = NFPROTO_IPV4;
 653         if (par.target->destroy != NULL)
 654                 par.target->destroy(&par);
 655         module_put(par.target->me);
 656         xt_percpu_counter_free(&e->counters);
 657 }
 658
 659 /* Checks and translates the user-supplied table segment (held in
 660    newinfo) */
 661 static int
 662 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 663                 const struct ipt_replace *repl)
 664 {
 665         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 666         struct ipt_entry *iter;
 667         unsigned int *offsets;
 668         unsigned int i;
 669         int ret = 0;
 670
 671         newinfo->size = repl->size;
 672         newinfo->number = repl->num_entries;
 673
 674         /* Init all hooks to impossible value. */
 675         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 676                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 677                 newinfo->underflow[i] = 0xFFFFFFFF;
 678         }
 679
 680         offsets = xt_alloc_entry_offsets(newinfo->number);
 681         if (!offsets)
 682                 return -ENOMEM;
 683         i = 0;
 684         /* Walk through entries, checking offsets. */
 685         xt_entry_foreach(iter, entry0, newinfo->size) {
 686                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 687                                                  entry0 + repl->size,
 688                                                  repl->hook_entry,
 689                                                  repl->underflow,
 690                                                  repl->valid_hooks);
 691                 if (ret != 0)
 692                         goto out_free;
 693                 if (i < repl->num_entries)
 694                         offsets[i] = (void *)iter - entry0;
 695                 ++i;
 696                 if (strcmp(ipt_get_target(iter)->u.user.name,
 697                     XT_ERROR_TARGET) == 0)
 698                         ++newinfo->stacksize;
 699         }
 700
 701         ret = -EINVAL;
 702         if (i != repl->num_entries)
 703                 goto out_free;
 704
 705         /* Check hooks all assigned */
 706         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 707                 /* Only hooks which are valid */
 708                 if (!(repl->valid_hooks & (1 << i)))
 709                         continue;
 710                 if (newinfo->hook_entry[i] == 0xFFFFFFFF)
 711                         goto out_free;
 712                 if (newinfo->underflow[i] == 0xFFFFFFFF)
 713                         goto out_free;
 714         }
 715
 716         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 717                 ret = -ELOOP;
 718                 goto out_free;
 719         }
 720         kvfree(offsets);
 721
 722         /* Finally, each sanity check must pass */
 723         i = 0;
 724         xt_entry_foreach(iter, entry0, newinfo->size) {
 725                 ret = find_check_entry(iter, net, repl->name, repl->size,
 726                                        &alloc_state);
 727                 if (ret != 0)
 728                         break;
 729                 ++i;
 730         }
 731
 732         if (ret != 0) {
 733                 xt_entry_foreach(iter, entry0, newinfo->size) {
 734                         if (i-- == 0)
 735                                 break;
 736                         cleanup_entry(iter, net);
 737                 }
 738                 return ret;
 739         }
 740
 741         return ret;
 742  out_free:
 743         kvfree(offsets);
 744         return ret;
 745 }
 746
 747 static void
 748 get_counters(const struct xt_table_info *t,
 749              struct xt_counters counters[])
 750 {
 751         struct ipt_entry *iter;
 752         unsigned int cpu;
 753         unsigned int i;
 754
 755         for_each_possible_cpu(cpu) {
 756                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 757
 758                 i = 0;
 759                 xt_entry_foreach(iter, t->entries, t->size) {
 760                         struct xt_counters *tmp;
 761                         u64 bcnt, pcnt;
 762                         unsigned int start;
 763
 764                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 765                         do {
 766                                 start = read_seqcount_begin(s);
 767                                 bcnt = tmp->bcnt;
 768                                 pcnt = tmp->pcnt;
 769                         } while (read_seqcount_retry(s, start));
 770
 771                         ADD_COUNTER(counters[i], bcnt, pcnt);
 772                         ++i; /* macro does multi eval of i */
 773                         cond_resched();
 774                 }
 775         }
 776 }
 777
 778 static void get_old_counters(const struct xt_table_info *t,
 779                              struct xt_counters counters[])
 780 {
 781         struct ipt_entry *iter;
 782         unsigned int cpu, i;
 783
 784         for_each_possible_cpu(cpu) {
 785                 i = 0;
 786                 xt_entry_foreach(iter, t->entries, t->size) {
 787                         const struct xt_counters *tmp;
 788
 789                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 790                         ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
 791                         ++i; /* macro does multi eval of i */
 792                 }
 793
 794                 cond_resched();
 795         }
 796 }
 797
 798 static struct xt_counters *alloc_counters(const struct xt_table *table)
 799 {
 800         unsigned int countersize;
 801         struct xt_counters *counters;
 802         const struct xt_table_info *private = table->private;
 803
 804         /* We need atomic snapshot of counters: rest doesn't change
 805            (other than comefrom, which userspace doesn't care
 806            about). */
 807         countersize = sizeof(struct xt_counters) * private->number;
 808         counters = vzalloc(countersize);
 809
 810         if (counters == NULL)
 811                 return ERR_PTR(-ENOMEM);
 812
 813         get_counters(private, counters);
 814
 815         return counters;
 816 }
 817
 818 static int
 819 copy_entries_to_user(unsigned int total_size,
 820                      const struct xt_table *table,
 821                      void __user *userptr)
 822 {
 823         unsigned int off, num;
 824         const struct ipt_entry *e;
 825         struct xt_counters *counters;
 826         const struct xt_table_info *private = table->private;
 827         int ret = 0;
 828         const void *loc_cpu_entry;
 829
 830         counters = alloc_counters(table);
 831         if (IS_ERR(counters))
 832                 return PTR_ERR(counters);
 833
 834         loc_cpu_entry = private->entries;
 835
 836         /* FIXME: use iterator macros --RR */
 837         /* ... then go back and fix counters and names */
 838         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 839                 unsigned int i;
 840                 const struct xt_entry_match *m;
 841                 const struct xt_entry_target *t;
 842
 843                 e = loc_cpu_entry + off;
 844                 if (copy_to_user(userptr + off, e, sizeof(*e))) {
 845                         ret = -EFAULT;
 846                         goto free_counters;
 847                 }
 848                 if (copy_to_user(userptr + off
 849                                  + offsetof(struct ipt_entry, counters),
 850                                  &counters[num],
 851                                  sizeof(counters[num])) != 0) {
 852                         ret = -EFAULT;
 853                         goto free_counters;
 854                 }
 855
 856                 for (i = sizeof(struct ipt_entry);
 857                      i < e->target_offset;
 858                      i += m->u.match_size) {
 859                         m = (void *)e + i;
 860
 861                         if (xt_match_to_user(m, userptr + off + i)) {
 862                                 ret = -EFAULT;
 863                                 goto free_counters;
 864                         }
 865                 }
 866
 867                 t = ipt_get_target_c(e);
 868                 if (xt_target_to_user(t, userptr + off + e->target_offset)) {
 869                         ret = -EFAULT;
 870                         goto free_counters;
 871                 }
 872         }
 873
 874  free_counters:
 875         vfree(counters);
 876         return ret;
 877 }
 878
 879 #ifdef CONFIG_COMPAT
 880 static void compat_standard_from_user(void *dst, const void *src)
 881 {
 882         int v = *(compat_int_t *)src;
 883
 884         if (v > 0)
 885                 v += xt_compat_calc_jump(AF_INET, v);
 886         memcpy(dst, &v, sizeof(v));
 887 }
 888
 889 static int compat_standard_to_user(void __user *dst, const void *src)
 890 {
 891         compat_int_t cv = *(int *)src;
 892
 893         if (cv > 0)
 894                 cv -= xt_compat_calc_jump(AF_INET, cv);
 895         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 896 }
 897
 898 static int compat_calc_entry(const struct ipt_entry *e,
 899                              const struct xt_table_info *info,
 900                              const void *base, struct xt_table_info *newinfo)
 901 {
 902         const struct xt_entry_match *ematch;
 903         const struct xt_entry_target *t;
 904         unsigned int entry_offset;
 905         int off, i, ret;
 906
 907         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
 908         entry_offset = (void *)e - base;
 909         xt_ematch_foreach(ematch, e)
 910                 off += xt_compat_match_offset(ematch->u.kernel.match);
 911         t = ipt_get_target_c(e);
 912         off += xt_compat_target_offset(t->u.kernel.target);
 913         newinfo->size -= off;
 914         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
 915         if (ret)
 916                 return ret;
 917
 918         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 919                 if (info->hook_entry[i] &&
 920                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
 921                         newinfo->hook_entry[i] -= off;
 922                 if (info->underflow[i] &&
 923                     (e < (struct ipt_entry *)(base + info->underflow[i])))
 924                         newinfo->underflow[i] -= off;
 925         }
 926         return 0;
 927 }
 928
 929 static int compat_table_info(const struct xt_table_info *info,
 930                              struct xt_table_info *newinfo)
 931 {
 932         struct ipt_entry *iter;
 933         const void *loc_cpu_entry;
 934         int ret;
 935
 936         if (!newinfo || !info)
 937                 return -EINVAL;
 938
 939         /* we dont care about newinfo->entries */
 940         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 941         newinfo->initial_entries = 0;
 942         loc_cpu_entry = info->entries;
 943         xt_compat_init_offsets(AF_INET, info->number);
 944         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 945                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 946                 if (ret != 0)
 947                         return ret;
 948         }
 949         return 0;
 950 }
 951 #endif
 952
 953 static int get_info(struct net *net, void __user *user,
 954                     const int *len, int compat)
 955 {
 956         char name[XT_TABLE_MAXNAMELEN];
 957         struct xt_table *t;
 958         int ret;
 959
 960         if (*len != sizeof(struct ipt_getinfo))
 961                 return -EINVAL;
 962
 963         if (copy_from_user(name, user, sizeof(name)) != 0)
 964                 return -EFAULT;
 965
 966         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 967 #ifdef CONFIG_COMPAT
 968         if (compat)
 969                 xt_compat_lock(AF_INET);
 970 #endif
 971         t = xt_request_find_table_lock(net, AF_INET, name);
 972         if (!IS_ERR(t)) {
 973                 struct ipt_getinfo info;
 974                 const struct xt_table_info *private = t->private;
 975 #ifdef CONFIG_COMPAT
 976                 struct xt_table_info tmp;
 977
 978                 if (compat) {
 979                         ret = compat_table_info(private, &tmp);
 980                         xt_compat_flush_offsets(AF_INET);
 981                         private = &tmp;
 982                 }
 983 #endif
 984                 memset(&info, 0, sizeof(info));
 985                 info.valid_hooks = t->valid_hooks;
 986                 memcpy(info.hook_entry, private->hook_entry,
 987                        sizeof(info.hook_entry));
 988                 memcpy(info.underflow, private->underflow,
 989                        sizeof(info.underflow));
 990                 info.num_entries = private->number;
 991                 info.size = private->size;
 992                 strcpy(info.name, name);
 993
 994                 if (copy_to_user(user, &info, *len) != 0)
 995                         ret = -EFAULT;
 996                 else
 997                         ret = 0;
 998
 999                 xt_table_unlock(t);
1000                 module_put(t->me);
1001         } else
1002                 ret = PTR_ERR(t);
1003 #ifdef CONFIG_COMPAT
1004         if (compat)
1005                 xt_compat_unlock(AF_INET);
1006 #endif
1007         return ret;
1008 }
1009
1010 static int
1011 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1012             const int *len)
1013 {
1014         int ret;
1015         struct ipt_get_entries get;
1016         struct xt_table *t;
1017
1018         if (*len < sizeof(get))
1019                 return -EINVAL;
1020         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1021                 return -EFAULT;
1022         if (*len != sizeof(struct ipt_get_entries) + get.size)
1023                 return -EINVAL;
1024         get.name[sizeof(get.name) - 1] = '\0';
1025
1026         t = xt_find_table_lock(net, AF_INET, get.name);
1027         if (!IS_ERR(t)) {
1028                 const struct xt_table_info *private = t->private;
1029                 if (get.size == private->size)
1030                         ret = copy_entries_to_user(private->size,
1031                                                    t, uptr->entrytable);
1032                 else
1033                         ret = -EAGAIN;
1034
1035                 module_put(t->me);
1036                 xt_table_unlock(t);
1037         } else
1038                 ret = PTR_ERR(t);
1039
1040         return ret;
1041 }
1042
1043 static int
1044 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1045              struct xt_table_info *newinfo, unsigned int num_counters,
1046              void __user *counters_ptr)
1047 {
1048         int ret;
1049         struct xt_table *t;
1050         struct xt_table_info *oldinfo;
1051         struct xt_counters *counters;
1052         struct ipt_entry *iter;
1053
1054         ret = 0;
1055         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1056         if (!counters) {
1057                 ret = -ENOMEM;
1058                 goto out;
1059         }
1060
1061         t = xt_request_find_table_lock(net, AF_INET, name);
1062         if (IS_ERR(t)) {
1063                 ret = PTR_ERR(t);
1064                 goto free_newinfo_counters_untrans;
1065         }
1066
1067         /* You lied! */
1068         if (valid_hooks != t->valid_hooks) {
1069                 ret = -EINVAL;
1070                 goto put_module;
1071         }
1072
1073         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1074         if (!oldinfo)
1075                 goto put_module;
1076
1077         /* Update module usage count based on number of rules */
1078         if ((oldinfo->number > oldinfo->initial_entries) ||
1079             (newinfo->number <= oldinfo->initial_entries))
1080                 module_put(t->me);
1081         if ((oldinfo->number > oldinfo->initial_entries) &&
1082             (newinfo->number <= oldinfo->initial_entries))
1083                 module_put(t->me);
1084
1085         get_old_counters(oldinfo, counters);
1086
1087         /* Decrease module usage counts and free resource */
1088         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1089                 cleanup_entry(iter, net);
1090
1091         xt_free_table_info(oldinfo);
1092         if (copy_to_user(counters_ptr, counters,
1093                          sizeof(struct xt_counters) * num_counters) != 0) {
1094                 /* Silent error, can't fail, new table is already in place */
1095                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1096         }
1097         vfree(counters);
1098         xt_table_unlock(t);
1099         return ret;
1100
1101  put_module:
1102         module_put(t->me);
1103         xt_table_unlock(t);
1104  free_newinfo_counters_untrans:
1105         vfree(counters);
1106  out:
1107         return ret;
1108 }
1109
1110 static int
1111 do_replace(struct net *net, const void __user *user, unsigned int len)
1112 {
1113         int ret;
1114         struct ipt_replace tmp;
1115         struct xt_table_info *newinfo;
1116         void *loc_cpu_entry;
1117         struct ipt_entry *iter;
1118
1119         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1120                 return -EFAULT;
1121
1122         /* overflow check */
1123         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1124                 return -ENOMEM;
1125         if (tmp.num_counters == 0)
1126                 return -EINVAL;
1127
1128         tmp.name[sizeof(tmp.name)-1] = 0;
1129
1130         newinfo = xt_alloc_table_info(tmp.size);
1131         if (!newinfo)
1132                 return -ENOMEM;
1133
1134         loc_cpu_entry = newinfo->entries;
1135         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1136                            tmp.size) != 0) {
1137                 ret = -EFAULT;
1138                 goto free_newinfo;
1139         }
1140
1141         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1142         if (ret != 0)
1143                 goto free_newinfo;
1144
1145         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1146                            tmp.num_counters, tmp.counters);
1147         if (ret)
1148                 goto free_newinfo_untrans;
1149         return 0;
1150
1151  free_newinfo_untrans:
1152         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1153                 cleanup_entry(iter, net);
1154  free_newinfo:
1155         xt_free_table_info(newinfo);
1156         return ret;
1157 }
1158
1159 static int
1160 do_add_counters(struct net *net, const void __user *user,
1161                 unsigned int len, int compat)
1162 {
1163         unsigned int i;
1164         struct xt_counters_info tmp;
1165         struct xt_counters *paddc;
1166         struct xt_table *t;
1167         const struct xt_table_info *private;
1168         int ret = 0;
1169         struct ipt_entry *iter;
1170         unsigned int addend;
1171
1172         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1173         if (IS_ERR(paddc))
1174                 return PTR_ERR(paddc);
1175
1176         t = xt_find_table_lock(net, AF_INET, tmp.name);
1177         if (IS_ERR(t)) {
1178                 ret = PTR_ERR(t);
1179                 goto free;
1180         }
1181
1182         local_bh_disable();
1183         private = t->private;
1184         if (private->number != tmp.num_counters) {
1185                 ret = -EINVAL;
1186                 goto unlock_up_free;
1187         }
1188
1189         i = 0;
1190         addend = xt_write_recseq_begin();
1191         xt_entry_foreach(iter, private->entries, private->size) {
1192                 struct xt_counters *tmp;
1193
1194                 tmp = xt_get_this_cpu_counter(&iter->counters);
1195                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1196                 ++i;
1197         }
1198         xt_write_recseq_end(addend);
1199  unlock_up_free:
1200         local_bh_enable();
1201         xt_table_unlock(t);
1202         module_put(t->me);
1203  free:
1204         vfree(paddc);
1205
1206         return ret;
1207 }
1208
1209 #ifdef CONFIG_COMPAT
1210 struct compat_ipt_replace {
1211         char                    name[XT_TABLE_MAXNAMELEN];
1212         u32                     valid_hooks;
1213         u32                     num_entries;
1214         u32                     size;
1215         u32                     hook_entry[NF_INET_NUMHOOKS];
1216         u32                     underflow[NF_INET_NUMHOOKS];
1217         u32                     num_counters;
1218         compat_uptr_t           counters;       /* struct xt_counters * */
1219         struct compat_ipt_entry entries[0];
1220 };
1221
1222 static int
1223 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1224                           unsigned int *size, struct xt_counters *counters,
1225                           unsigned int i)
1226 {
1227         struct xt_entry_target *t;
1228         struct compat_ipt_entry __user *ce;
1229         u_int16_t target_offset, next_offset;
1230         compat_uint_t origsize;
1231         const struct xt_entry_match *ematch;
1232         int ret = 0;
1233
1234         origsize = *size;
1235         ce = *dstptr;
1236         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1237             copy_to_user(&ce->counters, &counters[i],
1238             sizeof(counters[i])) != 0)
1239                 return -EFAULT;
1240
1241         *dstptr += sizeof(struct compat_ipt_entry);
1242         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1243
1244         xt_ematch_foreach(ematch, e) {
1245                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1246                 if (ret != 0)
1247                         return ret;
1248         }
1249         target_offset = e->target_offset - (origsize - *size);
1250         t = ipt_get_target(e);
1251         ret = xt_compat_target_to_user(t, dstptr, size);
1252         if (ret)
1253                 return ret;
1254         next_offset = e->next_offset - (origsize - *size);
1255         if (put_user(target_offset, &ce->target_offset) != 0 ||
1256             put_user(next_offset, &ce->next_offset) != 0)
1257                 return -EFAULT;
1258         return 0;
1259 }
1260
1261 static int
1262 compat_find_calc_match(struct xt_entry_match *m,
1263                        const struct ipt_ip *ip,
1264                        int *size)
1265 {
1266         struct xt_match *match;
1267
1268         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1269                                       m->u.user.revision);
1270         if (IS_ERR(match))
1271                 return PTR_ERR(match);
1272
1273         m->u.kernel.match = match;
1274         *size += xt_compat_match_offset(match);
1275         return 0;
1276 }
1277
1278 static void compat_release_entry(struct compat_ipt_entry *e)
1279 {
1280         struct xt_entry_target *t;
1281         struct xt_entry_match *ematch;
1282
1283         /* Cleanup all matches */
1284         xt_ematch_foreach(ematch, e)
1285                 module_put(ematch->u.kernel.match->me);
1286         t = compat_ipt_get_target(e);
1287         module_put(t->u.kernel.target->me);
1288 }
1289
1290 static int
1291 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1292                                   struct xt_table_info *newinfo,
1293                                   unsigned int *size,
1294                                   const unsigned char *base,
1295                                   const unsigned char *limit)
1296 {
1297         struct xt_entry_match *ematch;
1298         struct xt_entry_target *t;
1299         struct xt_target *target;
1300         unsigned int entry_offset;
1301         unsigned int j;
1302         int ret, off;
1303
1304         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1305             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1306             (unsigned char *)e + e->next_offset > limit)
1307                 return -EINVAL;
1308
1309         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1310                              sizeof(struct compat_xt_entry_target))
1311                 return -EINVAL;
1312
1313         if (!ip_checkentry(&e->ip))
1314                 return -EINVAL;
1315
1316         ret = xt_compat_check_entry_offsets(e, e->elems,
1317                                             e->target_offset, e->next_offset);
1318         if (ret)
1319                 return ret;
1320
1321         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1322         entry_offset = (void *)e - (void *)base;
1323         j = 0;
1324         xt_ematch_foreach(ematch, e) {
1325                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1326                 if (ret != 0)
1327                         goto release_matches;
1328                 ++j;
1329         }
1330
1331         t = compat_ipt_get_target(e);
1332         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1333                                         t->u.user.revision);
1334         if (IS_ERR(target)) {
1335                 ret = PTR_ERR(target);
1336                 goto release_matches;
1337         }
1338         t->u.kernel.target = target;
1339
1340         off += xt_compat_target_offset(target);
1341         *size += off;
1342         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1343         if (ret)
1344                 goto out;
1345
1346         return 0;
1347
1348 out:
1349         module_put(t->u.kernel.target->me);
1350 release_matches:
1351         xt_ematch_foreach(ematch, e) {
1352                 if (j-- == 0)
1353                         break;
1354                 module_put(ematch->u.kernel.match->me);
1355         }
1356         return ret;
1357 }
1358
1359 static void
1360 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1361                             unsigned int *size,
1362                             struct xt_table_info *newinfo, unsigned char *base)
1363 {
1364         struct xt_entry_target *t;
1365         struct ipt_entry *de;
1366         unsigned int origsize;
1367         int h;
1368         struct xt_entry_match *ematch;
1369
1370         origsize = *size;
1371         de = *dstptr;
1372         memcpy(de, e, sizeof(struct ipt_entry));
1373         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1374
1375         *dstptr += sizeof(struct ipt_entry);
1376         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1377
1378         xt_ematch_foreach(ematch, e)
1379                 xt_compat_match_from_user(ematch, dstptr, size);
1380
1381         de->target_offset = e->target_offset - (origsize - *size);
1382         t = compat_ipt_get_target(e);
1383         xt_compat_target_from_user(t, dstptr, size);
1384
1385         de->next_offset = e->next_offset - (origsize - *size);
1386
1387         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1388                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1389                         newinfo->hook_entry[h] -= origsize - *size;
1390                 if ((unsigned char *)de - base < newinfo->underflow[h])
1391                         newinfo->underflow[h] -= origsize - *size;
1392         }
1393 }
1394
1395 static int
1396 translate_compat_table(struct net *net,
1397                        struct xt_table_info **pinfo,
1398                        void **pentry0,
1399                        const struct compat_ipt_replace *compatr)
1400 {
1401         unsigned int i, j;
1402         struct xt_table_info *newinfo, *info;
1403         void *pos, *entry0, *entry1;
1404         struct compat_ipt_entry *iter0;
1405         struct ipt_replace repl;
1406         unsigned int size;
1407         int ret;
1408
1409         info = *pinfo;
1410         entry0 = *pentry0;
1411         size = compatr->size;
1412         info->number = compatr->num_entries;
1413
1414         j = 0;
1415         xt_compat_lock(AF_INET);
1416         xt_compat_init_offsets(AF_INET, compatr->num_entries);
1417         /* Walk through entries, checking offsets. */
1418         xt_entry_foreach(iter0, entry0, compatr->size) {
1419                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1420                                                         entry0,
1421                                                         entry0 + compatr->size);
1422                 if (ret != 0)
1423                         goto out_unlock;
1424                 ++j;
1425         }
1426
1427         ret = -EINVAL;
1428         if (j != compatr->num_entries)
1429                 goto out_unlock;
1430
1431         ret = -ENOMEM;
1432         newinfo = xt_alloc_table_info(size);
1433         if (!newinfo)
1434                 goto out_unlock;
1435
1436         newinfo->number = compatr->num_entries;
1437         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1438                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1439                 newinfo->underflow[i] = compatr->underflow[i];
1440         }
1441         entry1 = newinfo->entries;
1442         pos = entry1;
1443         size = compatr->size;
1444         xt_entry_foreach(iter0, entry0, compatr->size)
1445                 compat_copy_entry_from_user(iter0, &pos, &size,
1446                                             newinfo, entry1);
1447
1448         /* all module references in entry0 are now gone.
1449          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1450          * generated by 64bit userspace.
1451          *
1452          * Call standard translate_table() to validate all hook_entrys,
1453          * underflows, check for loops, etc.
1454          */
1455         xt_compat_flush_offsets(AF_INET);
1456         xt_compat_unlock(AF_INET);
1457
1458         memcpy(&repl, compatr, sizeof(*compatr));
1459
1460         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1461                 repl.hook_entry[i] = newinfo->hook_entry[i];
1462                 repl.underflow[i] = newinfo->underflow[i];
1463         }
1464
1465         repl.num_counters = 0;
1466         repl.counters = NULL;
1467         repl.size = newinfo->size;
1468         ret = translate_table(net, newinfo, entry1, &repl);
1469         if (ret)
1470                 goto free_newinfo;
1471
1472         *pinfo = newinfo;
1473         *pentry0 = entry1;
1474         xt_free_table_info(info);
1475         return 0;
1476
1477 free_newinfo:
1478         xt_free_table_info(newinfo);
1479         return ret;
1480 out_unlock:
1481         xt_compat_flush_offsets(AF_INET);
1482         xt_compat_unlock(AF_INET);
1483         xt_entry_foreach(iter0, entry0, compatr->size) {
1484                 if (j-- == 0)
1485                         break;
1486                 compat_release_entry(iter0);
1487         }
1488         return ret;
1489 }
1490
1491 static int
1492 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1493 {
1494         int ret;
1495         struct compat_ipt_replace tmp;
1496         struct xt_table_info *newinfo;
1497         void *loc_cpu_entry;
1498         struct ipt_entry *iter;
1499
1500         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1501                 return -EFAULT;
1502
1503         /* overflow check */
1504         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1505                 return -ENOMEM;
1506         if (tmp.num_counters == 0)
1507                 return -EINVAL;
1508
1509         tmp.name[sizeof(tmp.name)-1] = 0;
1510
1511         newinfo = xt_alloc_table_info(tmp.size);
1512         if (!newinfo)
1513                 return -ENOMEM;
1514
1515         loc_cpu_entry = newinfo->entries;
1516         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1517                            tmp.size) != 0) {
1518                 ret = -EFAULT;
1519                 goto free_newinfo;
1520         }
1521
1522         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1523         if (ret != 0)
1524                 goto free_newinfo;
1525
1526         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1527                            tmp.num_counters, compat_ptr(tmp.counters));
1528         if (ret)
1529                 goto free_newinfo_untrans;
1530         return 0;
1531
1532  free_newinfo_untrans:
1533         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1534                 cleanup_entry(iter, net);
1535  free_newinfo:
1536         xt_free_table_info(newinfo);
1537         return ret;
1538 }
1539
1540 static int
1541 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1542                       unsigned int len)
1543 {
1544         int ret;
1545
1546         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1547                 return -EPERM;
1548
1549         switch (cmd) {
1550         case IPT_SO_SET_REPLACE:
1551                 ret = compat_do_replace(sock_net(sk), user, len);
1552                 break;
1553
1554         case IPT_SO_SET_ADD_COUNTERS:
1555                 ret = do_add_counters(sock_net(sk), user, len, 1);
1556                 break;
1557
1558         default:
1559                 ret = -EINVAL;
1560         }
1561
1562         return ret;
1563 }
1564
1565 struct compat_ipt_get_entries {
1566         char name[XT_TABLE_MAXNAMELEN];
1567         compat_uint_t size;
1568         struct compat_ipt_entry entrytable[0];
1569 };
1570
1571 static int
1572 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1573                             void __user *userptr)
1574 {
1575         struct xt_counters *counters;
1576         const struct xt_table_info *private = table->private;
1577         void __user *pos;
1578         unsigned int size;
1579         int ret = 0;
1580         unsigned int i = 0;
1581         struct ipt_entry *iter;
1582
1583         counters = alloc_counters(table);
1584         if (IS_ERR(counters))
1585                 return PTR_ERR(counters);
1586
1587         pos = userptr;
1588         size = total_size;
1589         xt_entry_foreach(iter, private->entries, total_size) {
1590                 ret = compat_copy_entry_to_user(iter, &pos,
1591                                                 &size, counters, i++);
1592                 if (ret != 0)
1593                         break;
1594         }
1595
1596         vfree(counters);
1597         return ret;
1598 }
1599
1600 static int
1601 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1602                    int *len)
1603 {
1604         int ret;
1605         struct compat_ipt_get_entries get;
1606         struct xt_table *t;
1607
1608         if (*len < sizeof(get))
1609                 return -EINVAL;
1610
1611         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1612                 return -EFAULT;
1613
1614         if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1615                 return -EINVAL;
1616
1617         get.name[sizeof(get.name) - 1] = '\0';
1618
1619         xt_compat_lock(AF_INET);
1620         t = xt_find_table_lock(net, AF_INET, get.name);
1621         if (!IS_ERR(t)) {
1622                 const struct xt_table_info *private = t->private;
1623                 struct xt_table_info info;
1624                 ret = compat_table_info(private, &info);
1625                 if (!ret && get.size == info.size)
1626                         ret = compat_copy_entries_to_user(private->size,
1627                                                           t, uptr->entrytable);
1628                 else if (!ret)
1629                         ret = -EAGAIN;
1630
1631                 xt_compat_flush_offsets(AF_INET);
1632                 module_put(t->me);
1633                 xt_table_unlock(t);
1634         } else
1635                 ret = PTR_ERR(t);
1636
1637         xt_compat_unlock(AF_INET);
1638         return ret;
1639 }
1640
1641 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1642
1643 static int
1644 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1645 {
1646         int ret;
1647
1648         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1649                 return -EPERM;
1650
1651         switch (cmd) {
1652         case IPT_SO_GET_INFO:
1653                 ret = get_info(sock_net(sk), user, len, 1);
1654                 break;
1655         case IPT_SO_GET_ENTRIES:
1656                 ret = compat_get_entries(sock_net(sk), user, len);
1657                 break;
1658         default:
1659                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1660         }
1661         return ret;
1662 }
1663 #endif
1664
1665 static int
1666 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1667 {
1668         int ret;
1669
1670         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1671                 return -EPERM;
1672
1673         switch (cmd) {
1674         case IPT_SO_SET_REPLACE:
1675                 ret = do_replace(sock_net(sk), user, len);
1676                 break;
1677
1678         case IPT_SO_SET_ADD_COUNTERS:
1679                 ret = do_add_counters(sock_net(sk), user, len, 0);
1680                 break;
1681
1682         default:
1683                 ret = -EINVAL;
1684         }
1685
1686         return ret;
1687 }
1688
1689 static int
1690 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1691 {
1692         int ret;
1693
1694         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1695                 return -EPERM;
1696
1697         switch (cmd) {
1698         case IPT_SO_GET_INFO:
1699                 ret = get_info(sock_net(sk), user, len, 0);
1700                 break;
1701
1702         case IPT_SO_GET_ENTRIES:
1703                 ret = get_entries(sock_net(sk), user, len);
1704                 break;
1705
1706         case IPT_SO_GET_REVISION_MATCH:
1707         case IPT_SO_GET_REVISION_TARGET: {
1708                 struct xt_get_revision rev;
1709                 int target;
1710
1711                 if (*len != sizeof(rev)) {
1712                         ret = -EINVAL;
1713                         break;
1714                 }
1715                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1716                         ret = -EFAULT;
1717                         break;
1718                 }
1719                 rev.name[sizeof(rev.name)-1] = 0;
1720
1721                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1722                         target = 1;
1723                 else
1724                         target = 0;
1725
1726                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1727                                                          rev.revision,
1728                                                          target, &ret),
1729                                         "ipt_%s", rev.name);
1730                 break;
1731         }
1732
1733         default:
1734                 ret = -EINVAL;
1735         }
1736
1737         return ret;
1738 }
1739
1740 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1741 {
1742         struct xt_table_info *private;
1743         void *loc_cpu_entry;
1744         struct module *table_owner = table->me;
1745         struct ipt_entry *iter;
1746
1747         private = xt_unregister_table(table);
1748
1749         /* Decrease module usage counts and free resources */
1750         loc_cpu_entry = private->entries;
1751         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1752                 cleanup_entry(iter, net);
1753         if (private->number > private->initial_entries)
1754                 module_put(table_owner);
1755         xt_free_table_info(private);
1756 }
1757
1758 int ipt_register_table(struct net *net, const struct xt_table *table,
1759                        const struct ipt_replace *repl,
1760                        const struct nf_hook_ops *ops, struct xt_table **res)
1761 {
1762         int ret;
1763         struct xt_table_info *newinfo;
1764         struct xt_table_info bootstrap = {0};
1765         void *loc_cpu_entry;
1766         struct xt_table *new_table;
1767
1768         newinfo = xt_alloc_table_info(repl->size);
1769         if (!newinfo)
1770                 return -ENOMEM;
1771
1772         loc_cpu_entry = newinfo->entries;
1773         memcpy(loc_cpu_entry, repl->entries, repl->size);
1774
1775         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1776         if (ret != 0)
1777                 goto out_free;
1778
1779         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1780         if (IS_ERR(new_table)) {
1781                 ret = PTR_ERR(new_table);
1782                 goto out_free;
1783         }
1784
1785         /* set res now, will see skbs right after nf_register_net_hooks */
1786         WRITE_ONCE(*res, new_table);
1787
1788         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1789         if (ret != 0) {
1790                 __ipt_unregister_table(net, new_table);
1791                 *res = NULL;
1792         }
1793
1794         return ret;
1795
1796 out_free:
1797         xt_free_table_info(newinfo);
1798         return ret;
1799 }
1800
1801 void ipt_unregister_table(struct net *net, struct xt_table *table,
1802                           const struct nf_hook_ops *ops)
1803 {
1804         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1805         __ipt_unregister_table(net, table);
1806 }
1807
1808 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1809 static inline bool
1810 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1811                      u_int8_t type, u_int8_t code,
1812                      bool invert)
1813 {
1814         return ((test_type == 0xFF) ||
1815                 (type == test_type && code >= min_code && code <= max_code))
1816                 ^ invert;
1817 }
1818
1819 static bool
1820 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
1821 {
1822         const struct icmphdr *ic;
1823         struct icmphdr _icmph;
1824         const struct ipt_icmp *icmpinfo = par->matchinfo;
1825
1826         /* Must not be a fragment. */
1827         if (par->fragoff != 0)
1828                 return false;
1829
1830         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1831         if (ic == NULL) {
1832                 /* We've been asked to examine this packet, and we
1833                  * can't.  Hence, no choice but to drop.
1834                  */
1835                 par->hotdrop = true;
1836                 return false;
1837         }
1838
1839         return icmp_type_code_match(icmpinfo->type,
1840                                     icmpinfo->code[0],
1841                                     icmpinfo->code[1],
1842                                     ic->type, ic->code,
1843                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
1844 }
1845
1846 static int icmp_checkentry(const struct xt_mtchk_param *par)
1847 {
1848         const struct ipt_icmp *icmpinfo = par->matchinfo;
1849
1850         /* Must specify no unknown invflags */
1851         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
1852 }
1853
1854 static struct xt_target ipt_builtin_tg[] __read_mostly = {
1855         {
1856                 .name             = XT_STANDARD_TARGET,
1857                 .targetsize       = sizeof(int),
1858                 .family           = NFPROTO_IPV4,
1859 #ifdef CONFIG_COMPAT
1860                 .compatsize       = sizeof(compat_int_t),
1861                 .compat_from_user = compat_standard_from_user,
1862                 .compat_to_user   = compat_standard_to_user,
1863 #endif
1864         },
1865         {
1866                 .name             = XT_ERROR_TARGET,
1867                 .target           = ipt_error,
1868                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1869                 .family           = NFPROTO_IPV4,
1870         },
1871 };
1872
1873 static struct nf_sockopt_ops ipt_sockopts = {
1874         .pf             = PF_INET,
1875         .set_optmin     = IPT_BASE_CTL,
1876         .set_optmax     = IPT_SO_SET_MAX+1,
1877         .set            = do_ipt_set_ctl,
1878 #ifdef CONFIG_COMPAT
1879         .compat_set     = compat_do_ipt_set_ctl,
1880 #endif
1881         .get_optmin     = IPT_BASE_CTL,
1882         .get_optmax     = IPT_SO_GET_MAX+1,
1883         .get            = do_ipt_get_ctl,
1884 #ifdef CONFIG_COMPAT
1885         .compat_get     = compat_do_ipt_get_ctl,
1886 #endif
1887         .owner          = THIS_MODULE,
1888 };
1889
1890 static struct xt_match ipt_builtin_mt[] __read_mostly = {
1891         {
1892                 .name       = "icmp",
1893                 .match      = icmp_match,
1894                 .matchsize  = sizeof(struct ipt_icmp),
1895                 .checkentry = icmp_checkentry,
1896                 .proto      = IPPROTO_ICMP,
1897                 .family     = NFPROTO_IPV4,
1898         },
1899 };
1900
1901 static int __net_init ip_tables_net_init(struct net *net)
1902 {
1903         return xt_proto_init(net, NFPROTO_IPV4);
1904 }
1905
1906 static void __net_exit ip_tables_net_exit(struct net *net)
1907 {
1908         xt_proto_fini(net, NFPROTO_IPV4);
1909 }
1910
1911 static struct pernet_operations ip_tables_net_ops = {
1912         .init = ip_tables_net_init,
1913         .exit = ip_tables_net_exit,
1914 };
1915
1916 static int __init ip_tables_init(void)
1917 {
1918         int ret;
1919
1920         ret = register_pernet_subsys(&ip_tables_net_ops);
1921         if (ret < 0)
1922                 goto err1;
1923
1924         /* No one else will be downing sem now, so we won't sleep */
1925         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1926         if (ret < 0)
1927                 goto err2;
1928         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1929         if (ret < 0)
1930                 goto err4;
1931
1932         /* Register setsockopt */
1933         ret = nf_register_sockopt(&ipt_sockopts);
1934         if (ret < 0)
1935                 goto err5;
1936
1937         return 0;
1938
1939 err5:
1940         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1941 err4:
1942         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1943 err2:
1944         unregister_pernet_subsys(&ip_tables_net_ops);
1945 err1:
1946         return ret;
1947 }
1948
1949 static void __exit ip_tables_fini(void)
1950 {
1951         nf_unregister_sockopt(&ipt_sockopts);
1952
1953         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1954         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1955         unregister_pernet_subsys(&ip_tables_net_ops);
1956 }
1957
1958 EXPORT_SYMBOL(ipt_register_table);
1959 EXPORT_SYMBOL(ipt_unregister_table);
1960 EXPORT_SYMBOL(ipt_do_table);
1961 module_init(ip_tables_init);
1962 module_exit(ip_tables_fini);