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_DEFAULT_MASK,
 160                 },
 161         },
 162 };
 163
 164 /* Mildly perf critical (only if packet tracing is on) */
 165 static inline int
 166 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 167                       const char *hookname, const char **chainname,
 168                       const char **comment, unsigned int *rulenum)
 169 {
 170         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 171
 172         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 173                 /* Head of user chain: ERROR target with chainname */
 174                 *chainname = t->target.data;
 175                 (*rulenum) = 0;
 176         } else if (s == e) {
 177                 (*rulenum)++;
 178
 179                 if (unconditional(s) &&
 180                     strcmp(t->target.u.kernel.target->name,
 181                            XT_STANDARD_TARGET) == 0 &&
 182                    t->verdict < 0) {
 183                         /* Tail of chains: STANDARD target (return/policy) */
 184                         *comment = *chainname == hookname
 185                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 186                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 187                 }
 188                 return 1;
 189         } else
 190                 (*rulenum)++;
 191
 192         return 0;
 193 }
 194
 195 static void trace_packet(struct net *net,
 196                          const struct sk_buff *skb,
 197                          unsigned int hook,
 198                          const struct net_device *in,
 199                          const struct net_device *out,
 200                          const char *tablename,
 201                          const struct xt_table_info *private,
 202                          const struct ipt_entry *e)
 203 {
 204         const struct ipt_entry *root;
 205         const char *hookname, *chainname, *comment;
 206         const struct ipt_entry *iter;
 207         unsigned int rulenum = 0;
 208
 209         root = get_entry(private->entries, private->hook_entry[hook]);
 210
 211         hookname = chainname = hooknames[hook];
 212         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 213
 214         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 215                 if (get_chainname_rulenum(iter, e, hookname,
 216                     &chainname, &comment, &rulenum) != 0)
 217                         break;
 218
 219         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 220                      "TRACE: %s:%s:%s:%u ",
 221                      tablename, chainname, comment, rulenum);
 222 }
 223 #endif
 224
 225 static inline
 226 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 227 {
 228         return (void *)entry + entry->next_offset;
 229 }
 230
 231 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 232 unsigned int
 233 ipt_do_table(struct sk_buff *skb,
 234              const struct nf_hook_state *state,
 235              struct xt_table *table)
 236 {
 237         unsigned int hook = state->hook;
 238         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 239         const struct iphdr *ip;
 240         /* Initializing verdict to NF_DROP keeps gcc happy. */
 241         unsigned int verdict = NF_DROP;
 242         const char *indev, *outdev;
 243         const void *table_base;
 244         struct ipt_entry *e, **jumpstack;
 245         unsigned int stackidx, cpu;
 246         const struct xt_table_info *private;
 247         struct xt_action_param acpar;
 248         unsigned int addend;
 249
 250         /* Initialization */
 251         stackidx = 0;
 252         ip = ip_hdr(skb);
 253         indev = state->in ? state->in->name : nulldevname;
 254         outdev = state->out ? state->out->name : nulldevname;
 255         /* We handle fragments by dealing with the first fragment as
 256          * if it was a normal packet.  All other fragments are treated
 257          * normally, except that they will NEVER match rules that ask
 258          * things we don't know, ie. tcp syn flag or ports).  If the
 259          * rule is also a fragment-specific rule, non-fragments won't
 260          * match it. */
 261         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 262         acpar.thoff   = ip_hdrlen(skb);
 263         acpar.hotdrop = false;
 264         acpar.state   = state;
 265
 266         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 267         local_bh_disable();
 268         addend = xt_write_recseq_begin();
 269         private = table->private;
 270         cpu        = smp_processor_id();
 271         /*
 272          * Ensure we load private-> members after we've fetched the base
 273          * pointer.
 274          */
 275         smp_read_barrier_depends();
 276         table_base = private->entries;
 277         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 278
 279         /* Switch to alternate jumpstack if we're being invoked via TEE.
 280          * TEE issues XT_CONTINUE verdict on original skb so we must not
 281          * clobber the jumpstack.
 282          *
 283          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 284          * but it is no problem since absolute verdict is issued by these.
 285          */
 286         if (static_key_false(&xt_tee_enabled))
 287                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 288
 289         e = get_entry(table_base, private->hook_entry[hook]);
 290
 291         do {
 292                 const struct xt_entry_target *t;
 293                 const struct xt_entry_match *ematch;
 294                 struct xt_counters *counter;
 295
 296                 IP_NF_ASSERT(e);
 297                 if (!ip_packet_match(ip, indev, outdev,
 298                     &e->ip, acpar.fragoff)) {
 299  no_match:
 300                         e = ipt_next_entry(e);
 301                         continue;
 302                 }
 303
 304                 xt_ematch_foreach(ematch, e) {
 305                         acpar.match     = ematch->u.kernel.match;
 306                         acpar.matchinfo = ematch->data;
 307                         if (!acpar.match->match(skb, &acpar))
 308                                 goto no_match;
 309                 }
 310
 311                 counter = xt_get_this_cpu_counter(&e->counters);
 312                 ADD_COUNTER(*counter, skb->len, 1);
 313
 314                 t = ipt_get_target(e);
 315                 IP_NF_ASSERT(t->u.kernel.target);
 316
 317 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 318                 /* The packet is traced: log it */
 319                 if (unlikely(skb->nf_trace))
 320                         trace_packet(state->net, skb, hook, state->in,
 321                                      state->out, table->name, private, e);
 322 #endif
 323                 /* Standard target? */
 324                 if (!t->u.kernel.target->target) {
 325                         int v;
 326
 327                         v = ((struct xt_standard_target *)t)->verdict;
 328                         if (v < 0) {
 329                                 /* Pop from stack? */
 330                                 if (v != XT_RETURN) {
 331                                         verdict = (unsigned int)(-v) - 1;
 332                                         break;
 333                                 }
 334                                 if (stackidx == 0) {
 335                                         e = get_entry(table_base,
 336                                             private->underflow[hook]);
 337                                 } else {
 338                                         e = jumpstack[--stackidx];
 339                                         e = ipt_next_entry(e);
 340                                 }
 341                                 continue;
 342                         }
 343                         if (table_base + v != ipt_next_entry(e) &&
 344                             !(e->ip.flags & IPT_F_GOTO))
 345                                 jumpstack[stackidx++] = e;
 346
 347                         e = get_entry(table_base, v);
 348                         continue;
 349                 }
 350
 351                 acpar.target   = t->u.kernel.target;
 352                 acpar.targinfo = t->data;
 353
 354                 verdict = t->u.kernel.target->target(skb, &acpar);
 355                 /* Target might have changed stuff. */
 356                 ip = ip_hdr(skb);
 357                 if (verdict == XT_CONTINUE)
 358                         e = ipt_next_entry(e);
 359                 else
 360                         /* Verdict */
 361                         break;
 362         } while (!acpar.hotdrop);
 363
 364         xt_write_recseq_end(addend);
 365         local_bh_enable();
 366
 367         if (acpar.hotdrop)
 368                 return NF_DROP;
 369         else return verdict;
 370 }
 371
 372 /* Figures out from what hook each rule can be called: returns 0 if
 373    there are loops.  Puts hook bitmask in comefrom. */
 374 static int
 375 mark_source_chains(const struct xt_table_info *newinfo,
 376                    unsigned int valid_hooks, void *entry0,
 377                    unsigned int *offsets)
 378 {
 379         unsigned int hook;
 380
 381         /* No recursion; use packet counter to save back ptrs (reset
 382            to 0 as we leave), and comefrom to save source hook bitmask */
 383         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 384                 unsigned int pos = newinfo->hook_entry[hook];
 385                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 386
 387                 if (!(valid_hooks & (1 << hook)))
 388                         continue;
 389
 390                 /* Set initial back pointer. */
 391                 e->counters.pcnt = pos;
 392
 393                 for (;;) {
 394                         const struct xt_standard_target *t
 395                                 = (void *)ipt_get_target_c(e);
 396                         int visited = e->comefrom & (1 << hook);
 397
 398                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 399                                 return 0;
 400
 401                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 402
 403                         /* Unconditional return/END. */
 404                         if ((unconditional(e) &&
 405                              (strcmp(t->target.u.user.name,
 406                                      XT_STANDARD_TARGET) == 0) &&
 407                              t->verdict < 0) || visited) {
 408                                 unsigned int oldpos, size;
 409
 410                                 if ((strcmp(t->target.u.user.name,
 411                                             XT_STANDARD_TARGET) == 0) &&
 412                                     t->verdict < -NF_MAX_VERDICT - 1)
 413                                         return 0;
 414
 415                                 /* Return: backtrack through the last
 416                                    big jump. */
 417                                 do {
 418                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 419                                         oldpos = pos;
 420                                         pos = e->counters.pcnt;
 421                                         e->counters.pcnt = 0;
 422
 423                                         /* We're at the start. */
 424                                         if (pos == oldpos)
 425                                                 goto next;
 426
 427                                         e = (struct ipt_entry *)
 428                                                 (entry0 + pos);
 429                                 } while (oldpos == pos + e->next_offset);
 430
 431                                 /* Move along one */
 432                                 size = e->next_offset;
 433                                 e = (struct ipt_entry *)
 434                                         (entry0 + pos + size);
 435                                 if (pos + size >= newinfo->size)
 436                                         return 0;
 437                                 e->counters.pcnt = pos;
 438                                 pos += size;
 439                         } else {
 440                                 int newpos = t->verdict;
 441
 442                                 if (strcmp(t->target.u.user.name,
 443                                            XT_STANDARD_TARGET) == 0 &&
 444                                     newpos >= 0) {
 445                                         /* This a jump; chase it. */
 446                                         if (!xt_find_jump_offset(offsets, newpos,
 447                                                                  newinfo->number))
 448                                                 return 0;
 449                                         e = (struct ipt_entry *)
 450                                                 (entry0 + newpos);
 451                                 } else {
 452                                         /* ... this is a fallthru */
 453                                         newpos = pos + e->next_offset;
 454                                         if (newpos >= newinfo->size)
 455                                                 return 0;
 456                                 }
 457                                 e = (struct ipt_entry *)
 458                                         (entry0 + newpos);
 459                                 e->counters.pcnt = pos;
 460                                 pos = newpos;
 461                         }
 462                 }
 463 next:           ;
 464         }
 465         return 1;
 466 }
 467
 468 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 469 {
 470         struct xt_mtdtor_param par;
 471
 472         par.net       = net;
 473         par.match     = m->u.kernel.match;
 474         par.matchinfo = m->data;
 475         par.family    = NFPROTO_IPV4;
 476         if (par.match->destroy != NULL)
 477                 par.match->destroy(&par);
 478         module_put(par.match->me);
 479 }
 480
 481 static int
 482 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 483 {
 484         const struct ipt_ip *ip = par->entryinfo;
 485
 486         par->match     = m->u.kernel.match;
 487         par->matchinfo = m->data;
 488
 489         return xt_check_match(par, m->u.match_size - sizeof(*m),
 490                               ip->proto, ip->invflags & IPT_INV_PROTO);
 491 }
 492
 493 static int
 494 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 495 {
 496         struct xt_match *match;
 497         int ret;
 498
 499         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 500                                       m->u.user.revision);
 501         if (IS_ERR(match))
 502                 return PTR_ERR(match);
 503         m->u.kernel.match = match;
 504
 505         ret = check_match(m, par);
 506         if (ret)
 507                 goto err;
 508
 509         return 0;
 510 err:
 511         module_put(m->u.kernel.match->me);
 512         return ret;
 513 }
 514
 515 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 516 {
 517         struct xt_entry_target *t = ipt_get_target(e);
 518         struct xt_tgchk_param par = {
 519                 .net       = net,
 520                 .table     = name,
 521                 .entryinfo = e,
 522                 .target    = t->u.kernel.target,
 523                 .targinfo  = t->data,
 524                 .hook_mask = e->comefrom,
 525                 .family    = NFPROTO_IPV4,
 526         };
 527
 528         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 529                                e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 530 }
 531
 532 static int
 533 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 534                  unsigned int size,
 535                  struct xt_percpu_counter_alloc_state *alloc_state)
 536 {
 537         struct xt_entry_target *t;
 538         struct xt_target *target;
 539         int ret;
 540         unsigned int j;
 541         struct xt_mtchk_param mtpar;
 542         struct xt_entry_match *ematch;
 543
 544         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 545                 return -ENOMEM;
 546
 547         j = 0;
 548         mtpar.net       = net;
 549         mtpar.table     = name;
 550         mtpar.entryinfo = &e->ip;
 551         mtpar.hook_mask = e->comefrom;
 552         mtpar.family    = NFPROTO_IPV4;
 553         xt_ematch_foreach(ematch, e) {
 554                 ret = find_check_match(ematch, &mtpar);
 555                 if (ret != 0)
 556                         goto cleanup_matches;
 557                 ++j;
 558         }
 559
 560         t = ipt_get_target(e);
 561         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 562                                         t->u.user.revision);
 563         if (IS_ERR(target)) {
 564                 ret = PTR_ERR(target);
 565                 goto cleanup_matches;
 566         }
 567         t->u.kernel.target = target;
 568
 569         ret = check_target(e, net, name);
 570         if (ret)
 571                 goto err;
 572
 573         return 0;
 574  err:
 575         module_put(t->u.kernel.target->me);
 576  cleanup_matches:
 577         xt_ematch_foreach(ematch, e) {
 578                 if (j-- == 0)
 579                         break;
 580                 cleanup_match(ematch, net);
 581         }
 582
 583         xt_percpu_counter_free(&e->counters);
 584
 585         return ret;
 586 }
 587
 588 static bool check_underflow(const struct ipt_entry *e)
 589 {
 590         const struct xt_entry_target *t;
 591         unsigned int verdict;
 592
 593         if (!unconditional(e))
 594                 return false;
 595         t = ipt_get_target_c(e);
 596         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 597                 return false;
 598         verdict = ((struct xt_standard_target *)t)->verdict;
 599         verdict = -verdict - 1;
 600         return verdict == NF_DROP || verdict == NF_ACCEPT;
 601 }
 602
 603 static int
 604 check_entry_size_and_hooks(struct ipt_entry *e,
 605                            struct xt_table_info *newinfo,
 606                            const unsigned char *base,
 607                            const unsigned char *limit,
 608                            const unsigned int *hook_entries,
 609                            const unsigned int *underflows,
 610                            unsigned int valid_hooks)
 611 {
 612         unsigned int h;
 613         int err;
 614
 615         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 616             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 617             (unsigned char *)e + e->next_offset > limit)
 618                 return -EINVAL;
 619
 620         if (e->next_offset
 621             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
 622                 return -EINVAL;
 623
 624         if (!ip_checkentry(&e->ip))
 625                 return -EINVAL;
 626
 627         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 628                                      e->next_offset);
 629         if (err)
 630                 return err;
 631
 632         /* Check hooks & underflows */
 633         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 634                 if (!(valid_hooks & (1 << h)))
 635                         continue;
 636                 if ((unsigned char *)e - base == hook_entries[h])
 637                         newinfo->hook_entry[h] = hook_entries[h];
 638                 if ((unsigned char *)e - base == underflows[h]) {
 639                         if (!check_underflow(e))
 640                                 return -EINVAL;
 641
 642                         newinfo->underflow[h] = underflows[h];
 643                 }
 644         }
 645
 646         /* Clear counters and comefrom */
 647         e->counters = ((struct xt_counters) { 0, 0 });
 648         e->comefrom = 0;
 649         return 0;
 650 }
 651
 652 static void
 653 cleanup_entry(struct ipt_entry *e, struct net *net)
 654 {
 655         struct xt_tgdtor_param par;
 656         struct xt_entry_target *t;
 657         struct xt_entry_match *ematch;
 658
 659         /* Cleanup all matches */
 660         xt_ematch_foreach(ematch, e)
 661                 cleanup_match(ematch, net);
 662         t = ipt_get_target(e);
 663
 664         par.net      = net;
 665         par.target   = t->u.kernel.target;
 666         par.targinfo = t->data;
 667         par.family   = NFPROTO_IPV4;
 668         if (par.target->destroy != NULL)
 669                 par.target->destroy(&par);
 670         module_put(par.target->me);
 671         xt_percpu_counter_free(&e->counters);
 672 }
 673
 674 /* Checks and translates the user-supplied table segment (held in
 675    newinfo) */
 676 static int
 677 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 678                 const struct ipt_replace *repl)
 679 {
 680         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 681         struct ipt_entry *iter;
 682         unsigned int *offsets;
 683         unsigned int i;
 684         int ret = 0;
 685
 686         newinfo->size = repl->size;
 687         newinfo->number = repl->num_entries;
 688
 689         /* Init all hooks to impossible value. */
 690         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 691                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 692                 newinfo->underflow[i] = 0xFFFFFFFF;
 693         }
 694
 695         offsets = xt_alloc_entry_offsets(newinfo->number);
 696         if (!offsets)
 697                 return -ENOMEM;
 698         i = 0;
 699         /* Walk through entries, checking offsets. */
 700         xt_entry_foreach(iter, entry0, newinfo->size) {
 701                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 702                                                  entry0 + repl->size,
 703                                                  repl->hook_entry,
 704                                                  repl->underflow,
 705                                                  repl->valid_hooks);
 706                 if (ret != 0)
 707                         goto out_free;
 708                 if (i < repl->num_entries)
 709                         offsets[i] = (void *)iter - entry0;
 710                 ++i;
 711                 if (strcmp(ipt_get_target(iter)->u.user.name,
 712                     XT_ERROR_TARGET) == 0)
 713                         ++newinfo->stacksize;
 714         }
 715
 716         ret = -EINVAL;
 717         if (i != repl->num_entries)
 718                 goto out_free;
 719
 720         /* Check hooks all assigned */
 721         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 722                 /* Only hooks which are valid */
 723                 if (!(repl->valid_hooks & (1 << i)))
 724                         continue;
 725                 if (newinfo->hook_entry[i] == 0xFFFFFFFF)
 726                         goto out_free;
 727                 if (newinfo->underflow[i] == 0xFFFFFFFF)
 728                         goto out_free;
 729         }
 730
 731         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 732                 ret = -ELOOP;
 733                 goto out_free;
 734         }
 735         kvfree(offsets);
 736
 737         /* Finally, each sanity check must pass */
 738         i = 0;
 739         xt_entry_foreach(iter, entry0, newinfo->size) {
 740                 ret = find_check_entry(iter, net, repl->name, repl->size,
 741                                        &alloc_state);
 742                 if (ret != 0)
 743                         break;
 744                 ++i;
 745         }
 746
 747         if (ret != 0) {
 748                 xt_entry_foreach(iter, entry0, newinfo->size) {
 749                         if (i-- == 0)
 750                                 break;
 751                         cleanup_entry(iter, net);
 752                 }
 753                 return ret;
 754         }
 755
 756         return ret;
 757  out_free:
 758         kvfree(offsets);
 759         return ret;
 760 }
 761
 762 static void
 763 get_counters(const struct xt_table_info *t,
 764              struct xt_counters counters[])
 765 {
 766         struct ipt_entry *iter;
 767         unsigned int cpu;
 768         unsigned int i;
 769
 770         for_each_possible_cpu(cpu) {
 771                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 772
 773                 i = 0;
 774                 xt_entry_foreach(iter, t->entries, t->size) {
 775                         struct xt_counters *tmp;
 776                         u64 bcnt, pcnt;
 777                         unsigned int start;
 778
 779                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 780                         do {
 781                                 start = read_seqcount_begin(s);
 782                                 bcnt = tmp->bcnt;
 783                                 pcnt = tmp->pcnt;
 784                         } while (read_seqcount_retry(s, start));
 785
 786                         ADD_COUNTER(counters[i], bcnt, pcnt);
 787                         ++i; /* macro does multi eval of i */
 788                 }
 789         }
 790 }
 791
 792 static struct xt_counters *alloc_counters(const struct xt_table *table)
 793 {
 794         unsigned int countersize;
 795         struct xt_counters *counters;
 796         const struct xt_table_info *private = table->private;
 797
 798         /* We need atomic snapshot of counters: rest doesn't change
 799            (other than comefrom, which userspace doesn't care
 800            about). */
 801         countersize = sizeof(struct xt_counters) * private->number;
 802         counters = vzalloc(countersize);
 803
 804         if (counters == NULL)
 805                 return ERR_PTR(-ENOMEM);
 806
 807         get_counters(private, counters);
 808
 809         return counters;
 810 }
 811
 812 static int
 813 copy_entries_to_user(unsigned int total_size,
 814                      const struct xt_table *table,
 815                      void __user *userptr)
 816 {
 817         unsigned int off, num;
 818         const struct ipt_entry *e;
 819         struct xt_counters *counters;
 820         const struct xt_table_info *private = table->private;
 821         int ret = 0;
 822         const void *loc_cpu_entry;
 823
 824         counters = alloc_counters(table);
 825         if (IS_ERR(counters))
 826                 return PTR_ERR(counters);
 827
 828         loc_cpu_entry = private->entries;
 829         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 830                 ret = -EFAULT;
 831                 goto free_counters;
 832         }
 833
 834         /* FIXME: use iterator macros --RR */
 835         /* ... then go back and fix counters and names */
 836         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 837                 unsigned int i;
 838                 const struct xt_entry_match *m;
 839                 const struct xt_entry_target *t;
 840
 841                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 842                 if (copy_to_user(userptr + off
 843                                  + offsetof(struct ipt_entry, counters),
 844                                  &counters[num],
 845                                  sizeof(counters[num])) != 0) {
 846                         ret = -EFAULT;
 847                         goto free_counters;
 848                 }
 849
 850                 for (i = sizeof(struct ipt_entry);
 851                      i < e->target_offset;
 852                      i += m->u.match_size) {
 853                         m = (void *)e + i;
 854
 855                         if (copy_to_user(userptr + off + i
 856                                          + offsetof(struct xt_entry_match,
 857                                                     u.user.name),
 858                                          m->u.kernel.match->name,
 859                                          strlen(m->u.kernel.match->name)+1)
 860                             != 0) {
 861                                 ret = -EFAULT;
 862                                 goto free_counters;
 863                         }
 864                 }
 865
 866                 t = ipt_get_target_c(e);
 867                 if (copy_to_user(userptr + off + e->target_offset
 868                                  + offsetof(struct xt_entry_target,
 869                                             u.user.name),
 870                                  t->u.kernel.target->name,
 871                                  strlen(t->u.kernel.target->name)+1) != 0) {
 872                         ret = -EFAULT;
 873                         goto free_counters;
 874                 }
 875         }
 876
 877  free_counters:
 878         vfree(counters);
 879         return ret;
 880 }
 881
 882 #ifdef CONFIG_COMPAT
 883 static void compat_standard_from_user(void *dst, const void *src)
 884 {
 885         int v = *(compat_int_t *)src;
 886
 887         if (v > 0)
 888                 v += xt_compat_calc_jump(AF_INET, v);
 889         memcpy(dst, &v, sizeof(v));
 890 }
 891
 892 static int compat_standard_to_user(void __user *dst, const void *src)
 893 {
 894         compat_int_t cv = *(int *)src;
 895
 896         if (cv > 0)
 897                 cv -= xt_compat_calc_jump(AF_INET, cv);
 898         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 899 }
 900
 901 static int compat_calc_entry(const struct ipt_entry *e,
 902                              const struct xt_table_info *info,
 903                              const void *base, struct xt_table_info *newinfo)
 904 {
 905         const struct xt_entry_match *ematch;
 906         const struct xt_entry_target *t;
 907         unsigned int entry_offset;
 908         int off, i, ret;
 909
 910         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
 911         entry_offset = (void *)e - base;
 912         xt_ematch_foreach(ematch, e)
 913                 off += xt_compat_match_offset(ematch->u.kernel.match);
 914         t = ipt_get_target_c(e);
 915         off += xt_compat_target_offset(t->u.kernel.target);
 916         newinfo->size -= off;
 917         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
 918         if (ret)
 919                 return ret;
 920
 921         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 922                 if (info->hook_entry[i] &&
 923                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
 924                         newinfo->hook_entry[i] -= off;
 925                 if (info->underflow[i] &&
 926                     (e < (struct ipt_entry *)(base + info->underflow[i])))
 927                         newinfo->underflow[i] -= off;
 928         }
 929         return 0;
 930 }
 931
 932 static int compat_table_info(const struct xt_table_info *info,
 933                              struct xt_table_info *newinfo)
 934 {
 935         struct ipt_entry *iter;
 936         const void *loc_cpu_entry;
 937         int ret;
 938
 939         if (!newinfo || !info)
 940                 return -EINVAL;
 941
 942         /* we dont care about newinfo->entries */
 943         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 944         newinfo->initial_entries = 0;
 945         loc_cpu_entry = info->entries;
 946         xt_compat_init_offsets(AF_INET, info->number);
 947         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 948                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 949                 if (ret != 0)
 950                         return ret;
 951         }
 952         return 0;
 953 }
 954 #endif
 955
 956 static int get_info(struct net *net, void __user *user,
 957                     const int *len, int compat)
 958 {
 959         char name[XT_TABLE_MAXNAMELEN];
 960         struct xt_table *t;
 961         int ret;
 962
 963         if (*len != sizeof(struct ipt_getinfo))
 964                 return -EINVAL;
 965
 966         if (copy_from_user(name, user, sizeof(name)) != 0)
 967                 return -EFAULT;
 968
 969         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 970 #ifdef CONFIG_COMPAT
 971         if (compat)
 972                 xt_compat_lock(AF_INET);
 973 #endif
 974         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
 975                                     "iptable_%s", name);
 976         if (t) {
 977                 struct ipt_getinfo info;
 978                 const struct xt_table_info *private = t->private;
 979 #ifdef CONFIG_COMPAT
 980                 struct xt_table_info tmp;
 981
 982                 if (compat) {
 983                         ret = compat_table_info(private, &tmp);
 984                         xt_compat_flush_offsets(AF_INET);
 985                         private = &tmp;
 986                 }
 987 #endif
 988                 memset(&info, 0, sizeof(info));
 989                 info.valid_hooks = t->valid_hooks;
 990                 memcpy(info.hook_entry, private->hook_entry,
 991                        sizeof(info.hook_entry));
 992                 memcpy(info.underflow, private->underflow,
 993                        sizeof(info.underflow));
 994                 info.num_entries = private->number;
 995                 info.size = private->size;
 996                 strcpy(info.name, name);
 997
 998                 if (copy_to_user(user, &info, *len) != 0)
 999                         ret = -EFAULT;
1000                 else
1001                         ret = 0;
1002
1003                 xt_table_unlock(t);
1004                 module_put(t->me);
1005         } else
1006                 ret = -ENOENT;
1007 #ifdef CONFIG_COMPAT
1008         if (compat)
1009                 xt_compat_unlock(AF_INET);
1010 #endif
1011         return ret;
1012 }
1013
1014 static int
1015 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1016             const int *len)
1017 {
1018         int ret;
1019         struct ipt_get_entries get;
1020         struct xt_table *t;
1021
1022         if (*len < sizeof(get))
1023                 return -EINVAL;
1024         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1025                 return -EFAULT;
1026         if (*len != sizeof(struct ipt_get_entries) + get.size)
1027                 return -EINVAL;
1028         get.name[sizeof(get.name) - 1] = '\0';
1029
1030         t = xt_find_table_lock(net, AF_INET, get.name);
1031         if (t) {
1032                 const struct xt_table_info *private = t->private;
1033                 if (get.size == private->size)
1034                         ret = copy_entries_to_user(private->size,
1035                                                    t, uptr->entrytable);
1036                 else
1037                         ret = -EAGAIN;
1038
1039                 module_put(t->me);
1040                 xt_table_unlock(t);
1041         } else
1042                 ret = -ENOENT;
1043
1044         return ret;
1045 }
1046
1047 static int
1048 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1049              struct xt_table_info *newinfo, unsigned int num_counters,
1050              void __user *counters_ptr)
1051 {
1052         int ret;
1053         struct xt_table *t;
1054         struct xt_table_info *oldinfo;
1055         struct xt_counters *counters;
1056         struct ipt_entry *iter;
1057
1058         ret = 0;
1059         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1060         if (!counters) {
1061                 ret = -ENOMEM;
1062                 goto out;
1063         }
1064
1065         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1066                                     "iptable_%s", name);
1067         if (!t) {
1068                 ret = -ENOENT;
1069                 goto free_newinfo_counters_untrans;
1070         }
1071
1072         /* You lied! */
1073         if (valid_hooks != t->valid_hooks) {
1074                 ret = -EINVAL;
1075                 goto put_module;
1076         }
1077
1078         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1079         if (!oldinfo)
1080                 goto put_module;
1081
1082         /* Update module usage count based on number of rules */
1083         if ((oldinfo->number > oldinfo->initial_entries) ||
1084             (newinfo->number <= oldinfo->initial_entries))
1085                 module_put(t->me);
1086         if ((oldinfo->number > oldinfo->initial_entries) &&
1087             (newinfo->number <= oldinfo->initial_entries))
1088                 module_put(t->me);
1089
1090         /* Get the old counters, and synchronize with replace */
1091         get_counters(oldinfo, counters);
1092
1093         /* Decrease module usage counts and free resource */
1094         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1095                 cleanup_entry(iter, net);
1096
1097         xt_free_table_info(oldinfo);
1098         if (copy_to_user(counters_ptr, counters,
1099                          sizeof(struct xt_counters) * num_counters) != 0) {
1100                 /* Silent error, can't fail, new table is already in place */
1101                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1102         }
1103         vfree(counters);
1104         xt_table_unlock(t);
1105         return ret;
1106
1107  put_module:
1108         module_put(t->me);
1109         xt_table_unlock(t);
1110  free_newinfo_counters_untrans:
1111         vfree(counters);
1112  out:
1113         return ret;
1114 }
1115
1116 static int
1117 do_replace(struct net *net, const void __user *user, unsigned int len)
1118 {
1119         int ret;
1120         struct ipt_replace tmp;
1121         struct xt_table_info *newinfo;
1122         void *loc_cpu_entry;
1123         struct ipt_entry *iter;
1124
1125         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1126                 return -EFAULT;
1127
1128         /* overflow check */
1129         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1130                 return -ENOMEM;
1131         if (tmp.num_counters == 0)
1132                 return -EINVAL;
1133
1134         tmp.name[sizeof(tmp.name)-1] = 0;
1135
1136         newinfo = xt_alloc_table_info(tmp.size);
1137         if (!newinfo)
1138                 return -ENOMEM;
1139
1140         loc_cpu_entry = newinfo->entries;
1141         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1142                            tmp.size) != 0) {
1143                 ret = -EFAULT;
1144                 goto free_newinfo;
1145         }
1146
1147         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1148         if (ret != 0)
1149                 goto free_newinfo;
1150
1151         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1152                            tmp.num_counters, tmp.counters);
1153         if (ret)
1154                 goto free_newinfo_untrans;
1155         return 0;
1156
1157  free_newinfo_untrans:
1158         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1159                 cleanup_entry(iter, net);
1160  free_newinfo:
1161         xt_free_table_info(newinfo);
1162         return ret;
1163 }
1164
1165 static int
1166 do_add_counters(struct net *net, const void __user *user,
1167                 unsigned int len, int compat)
1168 {
1169         unsigned int i;
1170         struct xt_counters_info tmp;
1171         struct xt_counters *paddc;
1172         struct xt_table *t;
1173         const struct xt_table_info *private;
1174         int ret = 0;
1175         struct ipt_entry *iter;
1176         unsigned int addend;
1177
1178         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1179         if (IS_ERR(paddc))
1180                 return PTR_ERR(paddc);
1181
1182         t = xt_find_table_lock(net, AF_INET, tmp.name);
1183         if (!t) {
1184                 ret = -ENOENT;
1185                 goto free;
1186         }
1187
1188         local_bh_disable();
1189         private = t->private;
1190         if (private->number != tmp.num_counters) {
1191                 ret = -EINVAL;
1192                 goto unlock_up_free;
1193         }
1194
1195         i = 0;
1196         addend = xt_write_recseq_begin();
1197         xt_entry_foreach(iter, private->entries, private->size) {
1198                 struct xt_counters *tmp;
1199
1200                 tmp = xt_get_this_cpu_counter(&iter->counters);
1201                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1202                 ++i;
1203         }
1204         xt_write_recseq_end(addend);
1205  unlock_up_free:
1206         local_bh_enable();
1207         xt_table_unlock(t);
1208         module_put(t->me);
1209  free:
1210         vfree(paddc);
1211
1212         return ret;
1213 }
1214
1215 #ifdef CONFIG_COMPAT
1216 struct compat_ipt_replace {
1217         char                    name[XT_TABLE_MAXNAMELEN];
1218         u32                     valid_hooks;
1219         u32                     num_entries;
1220         u32                     size;
1221         u32                     hook_entry[NF_INET_NUMHOOKS];
1222         u32                     underflow[NF_INET_NUMHOOKS];
1223         u32                     num_counters;
1224         compat_uptr_t           counters;       /* struct xt_counters * */
1225         struct compat_ipt_entry entries[0];
1226 };
1227
1228 static int
1229 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1230                           unsigned int *size, struct xt_counters *counters,
1231                           unsigned int i)
1232 {
1233         struct xt_entry_target *t;
1234         struct compat_ipt_entry __user *ce;
1235         u_int16_t target_offset, next_offset;
1236         compat_uint_t origsize;
1237         const struct xt_entry_match *ematch;
1238         int ret = 0;
1239
1240         origsize = *size;
1241         ce = (struct compat_ipt_entry __user *)*dstptr;
1242         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1243             copy_to_user(&ce->counters, &counters[i],
1244             sizeof(counters[i])) != 0)
1245                 return -EFAULT;
1246
1247         *dstptr += sizeof(struct compat_ipt_entry);
1248         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1249
1250         xt_ematch_foreach(ematch, e) {
1251                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1252                 if (ret != 0)
1253                         return ret;
1254         }
1255         target_offset = e->target_offset - (origsize - *size);
1256         t = ipt_get_target(e);
1257         ret = xt_compat_target_to_user(t, dstptr, size);
1258         if (ret)
1259                 return ret;
1260         next_offset = e->next_offset - (origsize - *size);
1261         if (put_user(target_offset, &ce->target_offset) != 0 ||
1262             put_user(next_offset, &ce->next_offset) != 0)
1263                 return -EFAULT;
1264         return 0;
1265 }
1266
1267 static int
1268 compat_find_calc_match(struct xt_entry_match *m,
1269                        const struct ipt_ip *ip,
1270                        int *size)
1271 {
1272         struct xt_match *match;
1273
1274         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1275                                       m->u.user.revision);
1276         if (IS_ERR(match))
1277                 return PTR_ERR(match);
1278
1279         m->u.kernel.match = match;
1280         *size += xt_compat_match_offset(match);
1281         return 0;
1282 }
1283
1284 static void compat_release_entry(struct compat_ipt_entry *e)
1285 {
1286         struct xt_entry_target *t;
1287         struct xt_entry_match *ematch;
1288
1289         /* Cleanup all matches */
1290         xt_ematch_foreach(ematch, e)
1291                 module_put(ematch->u.kernel.match->me);
1292         t = compat_ipt_get_target(e);
1293         module_put(t->u.kernel.target->me);
1294 }
1295
1296 static int
1297 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1298                                   struct xt_table_info *newinfo,
1299                                   unsigned int *size,
1300                                   const unsigned char *base,
1301                                   const unsigned char *limit)
1302 {
1303         struct xt_entry_match *ematch;
1304         struct xt_entry_target *t;
1305         struct xt_target *target;
1306         unsigned int entry_offset;
1307         unsigned int j;
1308         int ret, off;
1309
1310         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1311             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1312             (unsigned char *)e + e->next_offset > limit)
1313                 return -EINVAL;
1314
1315         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1316                              sizeof(struct compat_xt_entry_target))
1317                 return -EINVAL;
1318
1319         if (!ip_checkentry(&e->ip))
1320                 return -EINVAL;
1321
1322         ret = xt_compat_check_entry_offsets(e, e->elems,
1323                                             e->target_offset, e->next_offset);
1324         if (ret)
1325                 return ret;
1326
1327         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1328         entry_offset = (void *)e - (void *)base;
1329         j = 0;
1330         xt_ematch_foreach(ematch, e) {
1331                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1332                 if (ret != 0)
1333                         goto release_matches;
1334                 ++j;
1335         }
1336
1337         t = compat_ipt_get_target(e);
1338         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1339                                         t->u.user.revision);
1340         if (IS_ERR(target)) {
1341                 ret = PTR_ERR(target);
1342                 goto release_matches;
1343         }
1344         t->u.kernel.target = target;
1345
1346         off += xt_compat_target_offset(target);
1347         *size += off;
1348         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1349         if (ret)
1350                 goto out;
1351
1352         return 0;
1353
1354 out:
1355         module_put(t->u.kernel.target->me);
1356 release_matches:
1357         xt_ematch_foreach(ematch, e) {
1358                 if (j-- == 0)
1359                         break;
1360                 module_put(ematch->u.kernel.match->me);
1361         }
1362         return ret;
1363 }
1364
1365 static void
1366 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1367                             unsigned int *size,
1368                             struct xt_table_info *newinfo, unsigned char *base)
1369 {
1370         struct xt_entry_target *t;
1371         struct xt_target *target;
1372         struct ipt_entry *de;
1373         unsigned int origsize;
1374         int h;
1375         struct xt_entry_match *ematch;
1376
1377         origsize = *size;
1378         de = (struct ipt_entry *)*dstptr;
1379         memcpy(de, e, sizeof(struct ipt_entry));
1380         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1381
1382         *dstptr += sizeof(struct ipt_entry);
1383         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1384
1385         xt_ematch_foreach(ematch, e)
1386                 xt_compat_match_from_user(ematch, dstptr, size);
1387
1388         de->target_offset = e->target_offset - (origsize - *size);
1389         t = compat_ipt_get_target(e);
1390         target = t->u.kernel.target;
1391         xt_compat_target_from_user(t, dstptr, size);
1392
1393         de->next_offset = e->next_offset - (origsize - *size);
1394
1395         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1396                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1397                         newinfo->hook_entry[h] -= origsize - *size;
1398                 if ((unsigned char *)de - base < newinfo->underflow[h])
1399                         newinfo->underflow[h] -= origsize - *size;
1400         }
1401 }
1402
1403 static int
1404 translate_compat_table(struct net *net,
1405                        struct xt_table_info **pinfo,
1406                        void **pentry0,
1407                        const struct compat_ipt_replace *compatr)
1408 {
1409         unsigned int i, j;
1410         struct xt_table_info *newinfo, *info;
1411         void *pos, *entry0, *entry1;
1412         struct compat_ipt_entry *iter0;
1413         struct ipt_replace repl;
1414         unsigned int size;
1415         int ret;
1416
1417         info = *pinfo;
1418         entry0 = *pentry0;
1419         size = compatr->size;
1420         info->number = compatr->num_entries;
1421
1422         j = 0;
1423         xt_compat_lock(AF_INET);
1424         xt_compat_init_offsets(AF_INET, compatr->num_entries);
1425         /* Walk through entries, checking offsets. */
1426         xt_entry_foreach(iter0, entry0, compatr->size) {
1427                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1428                                                         entry0,
1429                                                         entry0 + compatr->size);
1430                 if (ret != 0)
1431                         goto out_unlock;
1432                 ++j;
1433         }
1434
1435         ret = -EINVAL;
1436         if (j != compatr->num_entries)
1437                 goto out_unlock;
1438
1439         ret = -ENOMEM;
1440         newinfo = xt_alloc_table_info(size);
1441         if (!newinfo)
1442                 goto out_unlock;
1443
1444         newinfo->number = compatr->num_entries;
1445         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1446                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1447                 newinfo->underflow[i] = compatr->underflow[i];
1448         }
1449         entry1 = newinfo->entries;
1450         pos = entry1;
1451         size = compatr->size;
1452         xt_entry_foreach(iter0, entry0, compatr->size)
1453                 compat_copy_entry_from_user(iter0, &pos, &size,
1454                                             newinfo, entry1);
1455
1456         /* all module references in entry0 are now gone.
1457          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1458          * generated by 64bit userspace.
1459          *
1460          * Call standard translate_table() to validate all hook_entrys,
1461          * underflows, check for loops, etc.
1462          */
1463         xt_compat_flush_offsets(AF_INET);
1464         xt_compat_unlock(AF_INET);
1465
1466         memcpy(&repl, compatr, sizeof(*compatr));
1467
1468         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1469                 repl.hook_entry[i] = newinfo->hook_entry[i];
1470                 repl.underflow[i] = newinfo->underflow[i];
1471         }
1472
1473         repl.num_counters = 0;
1474         repl.counters = NULL;
1475         repl.size = newinfo->size;
1476         ret = translate_table(net, newinfo, entry1, &repl);
1477         if (ret)
1478                 goto free_newinfo;
1479
1480         *pinfo = newinfo;
1481         *pentry0 = entry1;
1482         xt_free_table_info(info);
1483         return 0;
1484
1485 free_newinfo:
1486         xt_free_table_info(newinfo);
1487         return ret;
1488 out_unlock:
1489         xt_compat_flush_offsets(AF_INET);
1490         xt_compat_unlock(AF_INET);
1491         xt_entry_foreach(iter0, entry0, compatr->size) {
1492                 if (j-- == 0)
1493                         break;
1494                 compat_release_entry(iter0);
1495         }
1496         return ret;
1497 }
1498
1499 static int
1500 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1501 {
1502         int ret;
1503         struct compat_ipt_replace tmp;
1504         struct xt_table_info *newinfo;
1505         void *loc_cpu_entry;
1506         struct ipt_entry *iter;
1507
1508         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1509                 return -EFAULT;
1510
1511         /* overflow check */
1512         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1513                 return -ENOMEM;
1514         if (tmp.num_counters == 0)
1515                 return -EINVAL;
1516
1517         tmp.name[sizeof(tmp.name)-1] = 0;
1518
1519         newinfo = xt_alloc_table_info(tmp.size);
1520         if (!newinfo)
1521                 return -ENOMEM;
1522
1523         loc_cpu_entry = newinfo->entries;
1524         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1525                            tmp.size) != 0) {
1526                 ret = -EFAULT;
1527                 goto free_newinfo;
1528         }
1529
1530         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1531         if (ret != 0)
1532                 goto free_newinfo;
1533
1534         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1535                            tmp.num_counters, compat_ptr(tmp.counters));
1536         if (ret)
1537                 goto free_newinfo_untrans;
1538         return 0;
1539
1540  free_newinfo_untrans:
1541         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1542                 cleanup_entry(iter, net);
1543  free_newinfo:
1544         xt_free_table_info(newinfo);
1545         return ret;
1546 }
1547
1548 static int
1549 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1550                       unsigned int len)
1551 {
1552         int ret;
1553
1554         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1555                 return -EPERM;
1556
1557         switch (cmd) {
1558         case IPT_SO_SET_REPLACE:
1559                 ret = compat_do_replace(sock_net(sk), user, len);
1560                 break;
1561
1562         case IPT_SO_SET_ADD_COUNTERS:
1563                 ret = do_add_counters(sock_net(sk), user, len, 1);
1564                 break;
1565
1566         default:
1567                 ret = -EINVAL;
1568         }
1569
1570         return ret;
1571 }
1572
1573 struct compat_ipt_get_entries {
1574         char name[XT_TABLE_MAXNAMELEN];
1575         compat_uint_t size;
1576         struct compat_ipt_entry entrytable[0];
1577 };
1578
1579 static int
1580 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1581                             void __user *userptr)
1582 {
1583         struct xt_counters *counters;
1584         const struct xt_table_info *private = table->private;
1585         void __user *pos;
1586         unsigned int size;
1587         int ret = 0;
1588         unsigned int i = 0;
1589         struct ipt_entry *iter;
1590
1591         counters = alloc_counters(table);
1592         if (IS_ERR(counters))
1593                 return PTR_ERR(counters);
1594
1595         pos = userptr;
1596         size = total_size;
1597         xt_entry_foreach(iter, private->entries, total_size) {
1598                 ret = compat_copy_entry_to_user(iter, &pos,
1599                                                 &size, counters, i++);
1600                 if (ret != 0)
1601                         break;
1602         }
1603
1604         vfree(counters);
1605         return ret;
1606 }
1607
1608 static int
1609 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1610                    int *len)
1611 {
1612         int ret;
1613         struct compat_ipt_get_entries get;
1614         struct xt_table *t;
1615
1616         if (*len < sizeof(get))
1617                 return -EINVAL;
1618
1619         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1620                 return -EFAULT;
1621
1622         if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1623                 return -EINVAL;
1624
1625         get.name[sizeof(get.name) - 1] = '\0';
1626
1627         xt_compat_lock(AF_INET);
1628         t = xt_find_table_lock(net, AF_INET, get.name);
1629         if (t) {
1630                 const struct xt_table_info *private = t->private;
1631                 struct xt_table_info info;
1632                 ret = compat_table_info(private, &info);
1633                 if (!ret && get.size == info.size)
1634                         ret = compat_copy_entries_to_user(private->size,
1635                                                           t, uptr->entrytable);
1636                 else if (!ret)
1637                         ret = -EAGAIN;
1638
1639                 xt_compat_flush_offsets(AF_INET);
1640                 module_put(t->me);
1641                 xt_table_unlock(t);
1642         } else
1643                 ret = -ENOENT;
1644
1645         xt_compat_unlock(AF_INET);
1646         return ret;
1647 }
1648
1649 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1650
1651 static int
1652 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1653 {
1654         int ret;
1655
1656         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1657                 return -EPERM;
1658
1659         switch (cmd) {
1660         case IPT_SO_GET_INFO:
1661                 ret = get_info(sock_net(sk), user, len, 1);
1662                 break;
1663         case IPT_SO_GET_ENTRIES:
1664                 ret = compat_get_entries(sock_net(sk), user, len);
1665                 break;
1666         default:
1667                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1668         }
1669         return ret;
1670 }
1671 #endif
1672
1673 static int
1674 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1675 {
1676         int ret;
1677
1678         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1679                 return -EPERM;
1680
1681         switch (cmd) {
1682         case IPT_SO_SET_REPLACE:
1683                 ret = do_replace(sock_net(sk), user, len);
1684                 break;
1685
1686         case IPT_SO_SET_ADD_COUNTERS:
1687                 ret = do_add_counters(sock_net(sk), user, len, 0);
1688                 break;
1689
1690         default:
1691                 ret = -EINVAL;
1692         }
1693
1694         return ret;
1695 }
1696
1697 static int
1698 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1699 {
1700         int ret;
1701
1702         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1703                 return -EPERM;
1704
1705         switch (cmd) {
1706         case IPT_SO_GET_INFO:
1707                 ret = get_info(sock_net(sk), user, len, 0);
1708                 break;
1709
1710         case IPT_SO_GET_ENTRIES:
1711                 ret = get_entries(sock_net(sk), user, len);
1712                 break;
1713
1714         case IPT_SO_GET_REVISION_MATCH:
1715         case IPT_SO_GET_REVISION_TARGET: {
1716                 struct xt_get_revision rev;
1717                 int target;
1718
1719                 if (*len != sizeof(rev)) {
1720                         ret = -EINVAL;
1721                         break;
1722                 }
1723                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1724                         ret = -EFAULT;
1725                         break;
1726                 }
1727                 rev.name[sizeof(rev.name)-1] = 0;
1728
1729                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1730                         target = 1;
1731                 else
1732                         target = 0;
1733
1734                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1735                                                          rev.revision,
1736                                                          target, &ret),
1737                                         "ipt_%s", rev.name);
1738                 break;
1739         }
1740
1741         default:
1742                 ret = -EINVAL;
1743         }
1744
1745         return ret;
1746 }
1747
1748 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1749 {
1750         struct xt_table_info *private;
1751         void *loc_cpu_entry;
1752         struct module *table_owner = table->me;
1753         struct ipt_entry *iter;
1754
1755         private = xt_unregister_table(table);
1756
1757         /* Decrease module usage counts and free resources */
1758         loc_cpu_entry = private->entries;
1759         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1760                 cleanup_entry(iter, net);
1761         if (private->number > private->initial_entries)
1762                 module_put(table_owner);
1763         xt_free_table_info(private);
1764 }
1765
1766 int ipt_register_table(struct net *net, const struct xt_table *table,
1767                        const struct ipt_replace *repl,
1768                        const struct nf_hook_ops *ops, struct xt_table **res)
1769 {
1770         int ret;
1771         struct xt_table_info *newinfo;
1772         struct xt_table_info bootstrap = {0};
1773         void *loc_cpu_entry;
1774         struct xt_table *new_table;
1775
1776         newinfo = xt_alloc_table_info(repl->size);
1777         if (!newinfo)
1778                 return -ENOMEM;
1779
1780         loc_cpu_entry = newinfo->entries;
1781         memcpy(loc_cpu_entry, repl->entries, repl->size);
1782
1783         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1784         if (ret != 0)
1785                 goto out_free;
1786
1787         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1788         if (IS_ERR(new_table)) {
1789                 ret = PTR_ERR(new_table);
1790                 goto out_free;
1791         }
1792
1793         /* set res now, will see skbs right after nf_register_net_hooks */
1794         WRITE_ONCE(*res, new_table);
1795
1796         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1797         if (ret != 0) {
1798                 __ipt_unregister_table(net, new_table);
1799                 *res = NULL;
1800         }
1801
1802         return ret;
1803
1804 out_free:
1805         xt_free_table_info(newinfo);
1806         return ret;
1807 }
1808
1809 void ipt_unregister_table(struct net *net, struct xt_table *table,
1810                           const struct nf_hook_ops *ops)
1811 {
1812         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1813         __ipt_unregister_table(net, table);
1814 }
1815
1816 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1817 static inline bool
1818 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1819                      u_int8_t type, u_int8_t code,
1820                      bool invert)
1821 {
1822         return ((test_type == 0xFF) ||
1823                 (type == test_type && code >= min_code && code <= max_code))
1824                 ^ invert;
1825 }
1826
1827 static bool
1828 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
1829 {
1830         const struct icmphdr *ic;
1831         struct icmphdr _icmph;
1832         const struct ipt_icmp *icmpinfo = par->matchinfo;
1833
1834         /* Must not be a fragment. */
1835         if (par->fragoff != 0)
1836                 return false;
1837
1838         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1839         if (ic == NULL) {
1840                 /* We've been asked to examine this packet, and we
1841                  * can't.  Hence, no choice but to drop.
1842                  */
1843                 par->hotdrop = true;
1844                 return false;
1845         }
1846
1847         return icmp_type_code_match(icmpinfo->type,
1848                                     icmpinfo->code[0],
1849                                     icmpinfo->code[1],
1850                                     ic->type, ic->code,
1851                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
1852 }
1853
1854 static int icmp_checkentry(const struct xt_mtchk_param *par)
1855 {
1856         const struct ipt_icmp *icmpinfo = par->matchinfo;
1857
1858         /* Must specify no unknown invflags */
1859         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
1860 }
1861
1862 static struct xt_target ipt_builtin_tg[] __read_mostly = {
1863         {
1864                 .name             = XT_STANDARD_TARGET,
1865                 .targetsize       = sizeof(int),
1866                 .family           = NFPROTO_IPV4,
1867 #ifdef CONFIG_COMPAT
1868                 .compatsize       = sizeof(compat_int_t),
1869                 .compat_from_user = compat_standard_from_user,
1870                 .compat_to_user   = compat_standard_to_user,
1871 #endif
1872         },
1873         {
1874                 .name             = XT_ERROR_TARGET,
1875                 .target           = ipt_error,
1876                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1877                 .family           = NFPROTO_IPV4,
1878         },
1879 };
1880
1881 static struct nf_sockopt_ops ipt_sockopts = {
1882         .pf             = PF_INET,
1883         .set_optmin     = IPT_BASE_CTL,
1884         .set_optmax     = IPT_SO_SET_MAX+1,
1885         .set            = do_ipt_set_ctl,
1886 #ifdef CONFIG_COMPAT
1887         .compat_set     = compat_do_ipt_set_ctl,
1888 #endif
1889         .get_optmin     = IPT_BASE_CTL,
1890         .get_optmax     = IPT_SO_GET_MAX+1,
1891         .get            = do_ipt_get_ctl,
1892 #ifdef CONFIG_COMPAT
1893         .compat_get     = compat_do_ipt_get_ctl,
1894 #endif
1895         .owner          = THIS_MODULE,
1896 };
1897
1898 static struct xt_match ipt_builtin_mt[] __read_mostly = {
1899         {
1900                 .name       = "icmp",
1901                 .match      = icmp_match,
1902                 .matchsize  = sizeof(struct ipt_icmp),
1903                 .checkentry = icmp_checkentry,
1904                 .proto      = IPPROTO_ICMP,
1905                 .family     = NFPROTO_IPV4,
1906         },
1907 };
1908
1909 static int __net_init ip_tables_net_init(struct net *net)
1910 {
1911         return xt_proto_init(net, NFPROTO_IPV4);
1912 }
1913
1914 static void __net_exit ip_tables_net_exit(struct net *net)
1915 {
1916         xt_proto_fini(net, NFPROTO_IPV4);
1917 }
1918
1919 static struct pernet_operations ip_tables_net_ops = {
1920         .init = ip_tables_net_init,
1921         .exit = ip_tables_net_exit,
1922 };
1923
1924 static int __init ip_tables_init(void)
1925 {
1926         int ret;
1927
1928         ret = register_pernet_subsys(&ip_tables_net_ops);
1929         if (ret < 0)
1930                 goto err1;
1931
1932         /* No one else will be downing sem now, so we won't sleep */
1933         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1934         if (ret < 0)
1935                 goto err2;
1936         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1937         if (ret < 0)
1938                 goto err4;
1939
1940         /* Register setsockopt */
1941         ret = nf_register_sockopt(&ipt_sockopts);
1942         if (ret < 0)
1943                 goto err5;
1944
1945         pr_info("(C) 2000-2006 Netfilter Core Team\n");
1946         return 0;
1947
1948 err5:
1949         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1950 err4:
1951         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1952 err2:
1953         unregister_pernet_subsys(&ip_tables_net_ops);
1954 err1:
1955         return ret;
1956 }
1957
1958 static void __exit ip_tables_fini(void)
1959 {
1960         nf_unregister_sockopt(&ipt_sockopts);
1961
1962         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1963         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1964         unregister_pernet_subsys(&ip_tables_net_ops);
1965 }
1966
1967 EXPORT_SYMBOL(ipt_register_table);
1968 EXPORT_SYMBOL(ipt_unregister_table);
1969 EXPORT_SYMBOL(ipt_do_table);
1970 module_init(ip_tables_init);
1971 module_exit(ip_tables_fini);