net/ipv4/netfilter/ip_tables.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * Packet matching code.
   4  *
   5  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   6  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   7  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   8  */
   9 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  10 #include <linux/cache.h>
  11 #include <linux/capability.h>
  12 #include <linux/skbuff.h>
  13 #include <linux/kmod.h>
  14 #include <linux/vmalloc.h>
  15 #include <linux/netdevice.h>
  16 #include <linux/module.h>
  17 #include <net/ip.h>
  18 #include <net/compat.h>
  19 #include <linux/uaccess.h>
  20 #include <linux/mutex.h>
  21 #include <linux/proc_fs.h>
  22 #include <linux/err.h>
  23 #include <linux/cpumask.h>
  24
  25 #include <linux/netfilter/x_tables.h>
  26 #include <linux/netfilter_ipv4/ip_tables.h>
  27 #include <net/netfilter/nf_log.h>
  28 #include "../../netfilter/xt_repldata.h"
  29
  30 MODULE_LICENSE("GPL");
  31 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  32 MODULE_DESCRIPTION("IPv4 packet filter");
  33
  34 void *ipt_alloc_initial_table(const struct xt_table *info)
  35 {
  36         return xt_alloc_initial_table(ipt, IPT);
  37 }
  38 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  39
  40 /* Returns whether matches rule or not. */
  41 /* Performance critical - called for every packet */
  42 static inline bool
  43 ip_packet_match(const struct iphdr *ip,
  44                 const char *indev,
  45                 const char *outdev,
  46                 const struct ipt_ip *ipinfo,
  47                 int isfrag)
  48 {
  49         unsigned long ret;
  50
  51         if (NF_INVF(ipinfo, IPT_INV_SRCIP,
  52                     (ip->saddr & ipinfo->smsk.s_addr) != ipinfo->src.s_addr) ||
  53             NF_INVF(ipinfo, IPT_INV_DSTIP,
  54                     (ip->daddr & ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr))
  55                 return false;
  56
  57         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
  58
  59         if (NF_INVF(ipinfo, IPT_INV_VIA_IN, ret != 0))
  60                 return false;
  61
  62         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
  63
  64         if (NF_INVF(ipinfo, IPT_INV_VIA_OUT, ret != 0))
  65                 return false;
  66
  67         /* Check specific protocol */
  68         if (ipinfo->proto &&
  69             NF_INVF(ipinfo, IPT_INV_PROTO, ip->protocol != ipinfo->proto))
  70                 return false;
  71
  72         /* If we have a fragment rule but the packet is not a fragment
  73          * then we return zero */
  74         if (NF_INVF(ipinfo, IPT_INV_FRAG,
  75                     (ipinfo->flags & IPT_F_FRAG) && !isfrag))
  76                 return false;
  77
  78         return true;
  79 }
  80
  81 static bool
  82 ip_checkentry(const struct ipt_ip *ip)
  83 {
  84         if (ip->flags & ~IPT_F_MASK)
  85                 return false;
  86         if (ip->invflags & ~IPT_INV_MASK)
  87                 return false;
  88         return true;
  89 }
  90
  91 static unsigned int
  92 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
  93 {
  94         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
  95
  96         return NF_DROP;
  97 }
  98
  99 /* Performance critical */
 100 static inline struct ipt_entry *
 101 get_entry(const void *base, unsigned int offset)
 102 {
 103         return (struct ipt_entry *)(base + offset);
 104 }
 105
 106 /* All zeroes == unconditional rule. */
 107 /* Mildly perf critical (only if packet tracing is on) */
 108 static inline bool unconditional(const struct ipt_entry *e)
 109 {
 110         static const struct ipt_ip uncond;
 111
 112         return e->target_offset == sizeof(struct ipt_entry) &&
 113                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
 114 }
 115
 116 /* for const-correctness */
 117 static inline const struct xt_entry_target *
 118 ipt_get_target_c(const struct ipt_entry *e)
 119 {
 120         return ipt_get_target((struct ipt_entry *)e);
 121 }
 122
 123 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 124 static const char *const hooknames[] = {
 125         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 126         [NF_INET_LOCAL_IN]              = "INPUT",
 127         [NF_INET_FORWARD]               = "FORWARD",
 128         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 129         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 130 };
 131
 132 enum nf_ip_trace_comments {
 133         NF_IP_TRACE_COMMENT_RULE,
 134         NF_IP_TRACE_COMMENT_RETURN,
 135         NF_IP_TRACE_COMMENT_POLICY,
 136 };
 137
 138 static const char *const comments[] = {
 139         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 140         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 141         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 142 };
 143
 144 static const struct nf_loginfo trace_loginfo = {
 145         .type = NF_LOG_TYPE_LOG,
 146         .u = {
 147                 .log = {
 148                         .level = 4,
 149                         .logflags = NF_LOG_DEFAULT_MASK,
 150                 },
 151         },
 152 };
 153
 154 /* Mildly perf critical (only if packet tracing is on) */
 155 static inline int
 156 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 157                       const char *hookname, const char **chainname,
 158                       const char **comment, unsigned int *rulenum)
 159 {
 160         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 161
 162         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 163                 /* Head of user chain: ERROR target with chainname */
 164                 *chainname = t->target.data;
 165                 (*rulenum) = 0;
 166         } else if (s == e) {
 167                 (*rulenum)++;
 168
 169                 if (unconditional(s) &&
 170                     strcmp(t->target.u.kernel.target->name,
 171                            XT_STANDARD_TARGET) == 0 &&
 172                    t->verdict < 0) {
 173                         /* Tail of chains: STANDARD target (return/policy) */
 174                         *comment = *chainname == hookname
 175                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 176                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 177                 }
 178                 return 1;
 179         } else
 180                 (*rulenum)++;
 181
 182         return 0;
 183 }
 184
 185 static void trace_packet(struct net *net,
 186                          const struct sk_buff *skb,
 187                          unsigned int hook,
 188                          const struct net_device *in,
 189                          const struct net_device *out,
 190                          const char *tablename,
 191                          const struct xt_table_info *private,
 192                          const struct ipt_entry *e)
 193 {
 194         const struct ipt_entry *root;
 195         const char *hookname, *chainname, *comment;
 196         const struct ipt_entry *iter;
 197         unsigned int rulenum = 0;
 198
 199         root = get_entry(private->entries, private->hook_entry[hook]);
 200
 201         hookname = chainname = hooknames[hook];
 202         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 203
 204         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 205                 if (get_chainname_rulenum(iter, e, hookname,
 206                     &chainname, &comment, &rulenum) != 0)
 207                         break;
 208
 209         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 210                      "TRACE: %s:%s:%s:%u ",
 211                      tablename, chainname, comment, rulenum);
 212 }
 213 #endif
 214
 215 static inline
 216 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 217 {
 218         return (void *)entry + entry->next_offset;
 219 }
 220
 221 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 222 unsigned int
 223 ipt_do_table(void *priv,
 224              struct sk_buff *skb,
 225              const struct nf_hook_state *state)
 226 {
 227         const struct xt_table *table = priv;
 228         unsigned int hook = state->hook;
 229         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 230         const struct iphdr *ip;
 231         /* Initializing verdict to NF_DROP keeps gcc happy. */
 232         unsigned int verdict = NF_DROP;
 233         const char *indev, *outdev;
 234         const void *table_base;
 235         struct ipt_entry *e, **jumpstack;
 236         unsigned int stackidx, cpu;
 237         const struct xt_table_info *private;
 238         struct xt_action_param acpar;
 239         unsigned int addend;
 240
 241         /* Initialization */
 242         stackidx = 0;
 243         ip = ip_hdr(skb);
 244         indev = state->in ? state->in->name : nulldevname;
 245         outdev = state->out ? state->out->name : nulldevname;
 246         /* We handle fragments by dealing with the first fragment as
 247          * if it was a normal packet.  All other fragments are treated
 248          * normally, except that they will NEVER match rules that ask
 249          * things we don't know, ie. tcp syn flag or ports).  If the
 250          * rule is also a fragment-specific rule, non-fragments won't
 251          * match it. */
 252         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 253         acpar.thoff   = ip_hdrlen(skb);
 254         acpar.hotdrop = false;
 255         acpar.state   = state;
 256
 257         WARN_ON(!(table->valid_hooks & (1 << hook)));
 258         local_bh_disable();
 259         addend = xt_write_recseq_begin();
 260         private = READ_ONCE(table->private); /* Address dependency. */
 261         cpu        = smp_processor_id();
 262         table_base = private->entries;
 263         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 264
 265         /* Switch to alternate jumpstack if we're being invoked via TEE.
 266          * TEE issues XT_CONTINUE verdict on original skb so we must not
 267          * clobber the jumpstack.
 268          *
 269          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 270          * but it is no problem since absolute verdict is issued by these.
 271          */
 272         if (static_key_false(&xt_tee_enabled))
 273                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 274
 275         e = get_entry(table_base, private->hook_entry[hook]);
 276
 277         do {
 278                 const struct xt_entry_target *t;
 279                 const struct xt_entry_match *ematch;
 280                 struct xt_counters *counter;
 281
 282                 WARN_ON(!e);
 283                 if (!ip_packet_match(ip, indev, outdev,
 284                     &e->ip, acpar.fragoff)) {
 285  no_match:
 286                         e = ipt_next_entry(e);
 287                         continue;
 288                 }
 289
 290                 xt_ematch_foreach(ematch, e) {
 291                         acpar.match     = ematch->u.kernel.match;
 292                         acpar.matchinfo = ematch->data;
 293                         if (!acpar.match->match(skb, &acpar))
 294                                 goto no_match;
 295                 }
 296
 297                 counter = xt_get_this_cpu_counter(&e->counters);
 298                 ADD_COUNTER(*counter, skb->len, 1);
 299
 300                 t = ipt_get_target_c(e);
 301                 WARN_ON(!t->u.kernel.target);
 302
 303 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 304                 /* The packet is traced: log it */
 305                 if (unlikely(skb->nf_trace))
 306                         trace_packet(state->net, skb, hook, state->in,
 307                                      state->out, table->name, private, e);
 308 #endif
 309                 /* Standard target? */
 310                 if (!t->u.kernel.target->target) {
 311                         int v;
 312
 313                         v = ((struct xt_standard_target *)t)->verdict;
 314                         if (v < 0) {
 315                                 /* Pop from stack? */
 316                                 if (v != XT_RETURN) {
 317                                         verdict = (unsigned int)(-v) - 1;
 318                                         break;
 319                                 }
 320                                 if (stackidx == 0) {
 321                                         e = get_entry(table_base,
 322                                             private->underflow[hook]);
 323                                 } else {
 324                                         e = jumpstack[--stackidx];
 325                                         e = ipt_next_entry(e);
 326                                 }
 327                                 continue;
 328                         }
 329                         if (table_base + v != ipt_next_entry(e) &&
 330                             !(e->ip.flags & IPT_F_GOTO)) {
 331                                 if (unlikely(stackidx >= private->stacksize)) {
 332                                         verdict = NF_DROP;
 333                                         break;
 334                                 }
 335                                 jumpstack[stackidx++] = e;
 336                         }
 337
 338                         e = get_entry(table_base, v);
 339                         continue;
 340                 }
 341
 342                 acpar.target   = t->u.kernel.target;
 343                 acpar.targinfo = t->data;
 344
 345                 verdict = t->u.kernel.target->target(skb, &acpar);
 346                 if (verdict == XT_CONTINUE) {
 347                         /* Target might have changed stuff. */
 348                         ip = ip_hdr(skb);
 349                         e = ipt_next_entry(e);
 350                 } else {
 351                         /* Verdict */
 352                         break;
 353                 }
 354         } while (!acpar.hotdrop);
 355
 356         xt_write_recseq_end(addend);
 357         local_bh_enable();
 358
 359         if (acpar.hotdrop)
 360                 return NF_DROP;
 361         else return verdict;
 362 }
 363
 364 /* Figures out from what hook each rule can be called: returns 0 if
 365    there are loops.  Puts hook bitmask in comefrom. */
 366 static int
 367 mark_source_chains(const struct xt_table_info *newinfo,
 368                    unsigned int valid_hooks, void *entry0,
 369                    unsigned int *offsets)
 370 {
 371         unsigned int hook;
 372
 373         /* No recursion; use packet counter to save back ptrs (reset
 374            to 0 as we leave), and comefrom to save source hook bitmask */
 375         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 376                 unsigned int pos = newinfo->hook_entry[hook];
 377                 struct ipt_entry *e = entry0 + pos;
 378
 379                 if (!(valid_hooks & (1 << hook)))
 380                         continue;
 381
 382                 /* Set initial back pointer. */
 383                 e->counters.pcnt = pos;
 384
 385                 for (;;) {
 386                         const struct xt_standard_target *t
 387                                 = (void *)ipt_get_target_c(e);
 388                         int visited = e->comefrom & (1 << hook);
 389
 390                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 391                                 return 0;
 392
 393                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 394
 395                         /* Unconditional return/END. */
 396                         if ((unconditional(e) &&
 397                              (strcmp(t->target.u.user.name,
 398                                      XT_STANDARD_TARGET) == 0) &&
 399                              t->verdict < 0) || visited) {
 400                                 unsigned int oldpos, size;
 401
 402                                 /* Return: backtrack through the last
 403                                    big jump. */
 404                                 do {
 405                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 406                                         oldpos = pos;
 407                                         pos = e->counters.pcnt;
 408                                         e->counters.pcnt = 0;
 409
 410                                         /* We're at the start. */
 411                                         if (pos == oldpos)
 412                                                 goto next;
 413
 414                                         e = entry0 + pos;
 415                                 } while (oldpos == pos + e->next_offset);
 416
 417                                 /* Move along one */
 418                                 size = e->next_offset;
 419                                 e = entry0 + pos + size;
 420                                 if (pos + size >= newinfo->size)
 421                                         return 0;
 422                                 e->counters.pcnt = pos;
 423                                 pos += size;
 424                         } else {
 425                                 int newpos = t->verdict;
 426
 427                                 if (strcmp(t->target.u.user.name,
 428                                            XT_STANDARD_TARGET) == 0 &&
 429                                     newpos >= 0) {
 430                                         /* This a jump; chase it. */
 431                                         if (!xt_find_jump_offset(offsets, newpos,
 432                                                                  newinfo->number))
 433                                                 return 0;
 434                                 } else {
 435                                         /* ... this is a fallthru */
 436                                         newpos = pos + e->next_offset;
 437                                         if (newpos >= newinfo->size)
 438                                                 return 0;
 439                                 }
 440                                 e = entry0 + newpos;
 441                                 e->counters.pcnt = pos;
 442                                 pos = newpos;
 443                         }
 444                 }
 445 next:           ;
 446         }
 447         return 1;
 448 }
 449
 450 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 451 {
 452         struct xt_mtdtor_param par;
 453
 454         par.net       = net;
 455         par.match     = m->u.kernel.match;
 456         par.matchinfo = m->data;
 457         par.family    = NFPROTO_IPV4;
 458         if (par.match->destroy != NULL)
 459                 par.match->destroy(&par);
 460         module_put(par.match->me);
 461 }
 462
 463 static int
 464 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 465 {
 466         const struct ipt_ip *ip = par->entryinfo;
 467
 468         par->match     = m->u.kernel.match;
 469         par->matchinfo = m->data;
 470
 471         return xt_check_match(par, m->u.match_size - sizeof(*m),
 472                               ip->proto, ip->invflags & IPT_INV_PROTO);
 473 }
 474
 475 static int
 476 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 477 {
 478         struct xt_match *match;
 479         int ret;
 480
 481         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 482                                       m->u.user.revision);
 483         if (IS_ERR(match))
 484                 return PTR_ERR(match);
 485         m->u.kernel.match = match;
 486
 487         ret = check_match(m, par);
 488         if (ret)
 489                 goto err;
 490
 491         return 0;
 492 err:
 493         module_put(m->u.kernel.match->me);
 494         return ret;
 495 }
 496
 497 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 498 {
 499         struct xt_entry_target *t = ipt_get_target(e);
 500         struct xt_tgchk_param par = {
 501                 .net       = net,
 502                 .table     = name,
 503                 .entryinfo = e,
 504                 .target    = t->u.kernel.target,
 505                 .targinfo  = t->data,
 506                 .hook_mask = e->comefrom,
 507                 .family    = NFPROTO_IPV4,
 508         };
 509
 510         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 511                                e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 512 }
 513
 514 static int
 515 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 516                  unsigned int size,
 517                  struct xt_percpu_counter_alloc_state *alloc_state)
 518 {
 519         struct xt_entry_target *t;
 520         struct xt_target *target;
 521         int ret;
 522         unsigned int j;
 523         struct xt_mtchk_param mtpar;
 524         struct xt_entry_match *ematch;
 525
 526         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 527                 return -ENOMEM;
 528
 529         j = 0;
 530         memset(&mtpar, 0, sizeof(mtpar));
 531         mtpar.net       = net;
 532         mtpar.table     = name;
 533         mtpar.entryinfo = &e->ip;
 534         mtpar.hook_mask = e->comefrom;
 535         mtpar.family    = NFPROTO_IPV4;
 536         xt_ematch_foreach(ematch, e) {
 537                 ret = find_check_match(ematch, &mtpar);
 538                 if (ret != 0)
 539                         goto cleanup_matches;
 540                 ++j;
 541         }
 542
 543         t = ipt_get_target(e);
 544         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 545                                         t->u.user.revision);
 546         if (IS_ERR(target)) {
 547                 ret = PTR_ERR(target);
 548                 goto cleanup_matches;
 549         }
 550         t->u.kernel.target = target;
 551
 552         ret = check_target(e, net, name);
 553         if (ret)
 554                 goto err;
 555
 556         return 0;
 557  err:
 558         module_put(t->u.kernel.target->me);
 559  cleanup_matches:
 560         xt_ematch_foreach(ematch, e) {
 561                 if (j-- == 0)
 562                         break;
 563                 cleanup_match(ematch, net);
 564         }
 565
 566         xt_percpu_counter_free(&e->counters);
 567
 568         return ret;
 569 }
 570
 571 static bool check_underflow(const struct ipt_entry *e)
 572 {
 573         const struct xt_entry_target *t;
 574         unsigned int verdict;
 575
 576         if (!unconditional(e))
 577                 return false;
 578         t = ipt_get_target_c(e);
 579         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 580                 return false;
 581         verdict = ((struct xt_standard_target *)t)->verdict;
 582         verdict = -verdict - 1;
 583         return verdict == NF_DROP || verdict == NF_ACCEPT;
 584 }
 585
 586 static int
 587 check_entry_size_and_hooks(struct ipt_entry *e,
 588                            struct xt_table_info *newinfo,
 589                            const unsigned char *base,
 590                            const unsigned char *limit,
 591                            const unsigned int *hook_entries,
 592                            const unsigned int *underflows,
 593                            unsigned int valid_hooks)
 594 {
 595         unsigned int h;
 596         int err;
 597
 598         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 599             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 600             (unsigned char *)e + e->next_offset > limit)
 601                 return -EINVAL;
 602
 603         if (e->next_offset
 604             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
 605                 return -EINVAL;
 606
 607         if (!ip_checkentry(&e->ip))
 608                 return -EINVAL;
 609
 610         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 611                                      e->next_offset);
 612         if (err)
 613                 return err;
 614
 615         /* Check hooks & underflows */
 616         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 617                 if (!(valid_hooks & (1 << h)))
 618                         continue;
 619                 if ((unsigned char *)e - base == hook_entries[h])
 620                         newinfo->hook_entry[h] = hook_entries[h];
 621                 if ((unsigned char *)e - base == underflows[h]) {
 622                         if (!check_underflow(e))
 623                                 return -EINVAL;
 624
 625                         newinfo->underflow[h] = underflows[h];
 626                 }
 627         }
 628
 629         /* Clear counters and comefrom */
 630         e->counters = ((struct xt_counters) { 0, 0 });
 631         e->comefrom = 0;
 632         return 0;
 633 }
 634
 635 static void
 636 cleanup_entry(struct ipt_entry *e, struct net *net)
 637 {
 638         struct xt_tgdtor_param par;
 639         struct xt_entry_target *t;
 640         struct xt_entry_match *ematch;
 641
 642         /* Cleanup all matches */
 643         xt_ematch_foreach(ematch, e)
 644                 cleanup_match(ematch, net);
 645         t = ipt_get_target(e);
 646
 647         par.net      = net;
 648         par.target   = t->u.kernel.target;
 649         par.targinfo = t->data;
 650         par.family   = NFPROTO_IPV4;
 651         if (par.target->destroy != NULL)
 652                 par.target->destroy(&par);
 653         module_put(par.target->me);
 654         xt_percpu_counter_free(&e->counters);
 655 }
 656
 657 /* Checks and translates the user-supplied table segment (held in
 658    newinfo) */
 659 static int
 660 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 661                 const struct ipt_replace *repl)
 662 {
 663         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 664         struct ipt_entry *iter;
 665         unsigned int *offsets;
 666         unsigned int i;
 667         int ret = 0;
 668
 669         newinfo->size = repl->size;
 670         newinfo->number = repl->num_entries;
 671
 672         /* Init all hooks to impossible value. */
 673         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 674                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 675                 newinfo->underflow[i] = 0xFFFFFFFF;
 676         }
 677
 678         offsets = xt_alloc_entry_offsets(newinfo->number);
 679         if (!offsets)
 680                 return -ENOMEM;
 681         i = 0;
 682         /* Walk through entries, checking offsets. */
 683         xt_entry_foreach(iter, entry0, newinfo->size) {
 684                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 685                                                  entry0 + repl->size,
 686                                                  repl->hook_entry,
 687                                                  repl->underflow,
 688                                                  repl->valid_hooks);
 689                 if (ret != 0)
 690                         goto out_free;
 691                 if (i < repl->num_entries)
 692                         offsets[i] = (void *)iter - entry0;
 693                 ++i;
 694                 if (strcmp(ipt_get_target(iter)->u.user.name,
 695                     XT_ERROR_TARGET) == 0)
 696                         ++newinfo->stacksize;
 697         }
 698
 699         ret = -EINVAL;
 700         if (i != repl->num_entries)
 701                 goto out_free;
 702
 703         ret = xt_check_table_hooks(newinfo, repl->valid_hooks);
 704         if (ret)
 705                 goto out_free;
 706
 707         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 708                 ret = -ELOOP;
 709                 goto out_free;
 710         }
 711         kvfree(offsets);
 712
 713         /* Finally, each sanity check must pass */
 714         i = 0;
 715         xt_entry_foreach(iter, entry0, newinfo->size) {
 716                 ret = find_check_entry(iter, net, repl->name, repl->size,
 717                                        &alloc_state);
 718                 if (ret != 0)
 719                         break;
 720                 ++i;
 721         }
 722
 723         if (ret != 0) {
 724                 xt_entry_foreach(iter, entry0, newinfo->size) {
 725                         if (i-- == 0)
 726                                 break;
 727                         cleanup_entry(iter, net);
 728                 }
 729                 return ret;
 730         }
 731
 732         return ret;
 733  out_free:
 734         kvfree(offsets);
 735         return ret;
 736 }
 737
 738 static void
 739 get_counters(const struct xt_table_info *t,
 740              struct xt_counters counters[])
 741 {
 742         struct ipt_entry *iter;
 743         unsigned int cpu;
 744         unsigned int i;
 745
 746         for_each_possible_cpu(cpu) {
 747                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 748
 749                 i = 0;
 750                 xt_entry_foreach(iter, t->entries, t->size) {
 751                         struct xt_counters *tmp;
 752                         u64 bcnt, pcnt;
 753                         unsigned int start;
 754
 755                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 756                         do {
 757                                 start = read_seqcount_begin(s);
 758                                 bcnt = tmp->bcnt;
 759                                 pcnt = tmp->pcnt;
 760                         } while (read_seqcount_retry(s, start));
 761
 762                         ADD_COUNTER(counters[i], bcnt, pcnt);
 763                         ++i; /* macro does multi eval of i */
 764                         cond_resched();
 765                 }
 766         }
 767 }
 768
 769 static void get_old_counters(const struct xt_table_info *t,
 770                              struct xt_counters counters[])
 771 {
 772         struct ipt_entry *iter;
 773         unsigned int cpu, i;
 774
 775         for_each_possible_cpu(cpu) {
 776                 i = 0;
 777                 xt_entry_foreach(iter, t->entries, t->size) {
 778                         const struct xt_counters *tmp;
 779
 780                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 781                         ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
 782                         ++i; /* macro does multi eval of i */
 783                 }
 784
 785                 cond_resched();
 786         }
 787 }
 788
 789 static struct xt_counters *alloc_counters(const struct xt_table *table)
 790 {
 791         unsigned int countersize;
 792         struct xt_counters *counters;
 793         const struct xt_table_info *private = table->private;
 794
 795         /* We need atomic snapshot of counters: rest doesn't change
 796            (other than comefrom, which userspace doesn't care
 797            about). */
 798         countersize = sizeof(struct xt_counters) * private->number;
 799         counters = vzalloc(countersize);
 800
 801         if (counters == NULL)
 802                 return ERR_PTR(-ENOMEM);
 803
 804         get_counters(private, counters);
 805
 806         return counters;
 807 }
 808
 809 static int
 810 copy_entries_to_user(unsigned int total_size,
 811                      const struct xt_table *table,
 812                      void __user *userptr)
 813 {
 814         unsigned int off, num;
 815         const struct ipt_entry *e;
 816         struct xt_counters *counters;
 817         const struct xt_table_info *private = table->private;
 818         int ret = 0;
 819         const void *loc_cpu_entry;
 820
 821         counters = alloc_counters(table);
 822         if (IS_ERR(counters))
 823                 return PTR_ERR(counters);
 824
 825         loc_cpu_entry = private->entries;
 826
 827         /* FIXME: use iterator macros --RR */
 828         /* ... then go back and fix counters and names */
 829         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 830                 unsigned int i;
 831                 const struct xt_entry_match *m;
 832                 const struct xt_entry_target *t;
 833
 834                 e = loc_cpu_entry + off;
 835                 if (copy_to_user(userptr + off, e, sizeof(*e))) {
 836                         ret = -EFAULT;
 837                         goto free_counters;
 838                 }
 839                 if (copy_to_user(userptr + off
 840                                  + offsetof(struct ipt_entry, counters),
 841                                  &counters[num],
 842                                  sizeof(counters[num])) != 0) {
 843                         ret = -EFAULT;
 844                         goto free_counters;
 845                 }
 846
 847                 for (i = sizeof(struct ipt_entry);
 848                      i < e->target_offset;
 849                      i += m->u.match_size) {
 850                         m = (void *)e + i;
 851
 852                         if (xt_match_to_user(m, userptr + off + i)) {
 853                                 ret = -EFAULT;
 854                                 goto free_counters;
 855                         }
 856                 }
 857
 858                 t = ipt_get_target_c(e);
 859                 if (xt_target_to_user(t, userptr + off + e->target_offset)) {
 860                         ret = -EFAULT;
 861                         goto free_counters;
 862                 }
 863         }
 864
 865  free_counters:
 866         vfree(counters);
 867         return ret;
 868 }
 869
 870 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
 871 static void compat_standard_from_user(void *dst, const void *src)
 872 {
 873         int v = *(compat_int_t *)src;
 874
 875         if (v > 0)
 876                 v += xt_compat_calc_jump(AF_INET, v);
 877         memcpy(dst, &v, sizeof(v));
 878 }
 879
 880 static int compat_standard_to_user(void __user *dst, const void *src)
 881 {
 882         compat_int_t cv = *(int *)src;
 883
 884         if (cv > 0)
 885                 cv -= xt_compat_calc_jump(AF_INET, cv);
 886         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 887 }
 888
 889 static int compat_calc_entry(const struct ipt_entry *e,
 890                              const struct xt_table_info *info,
 891                              const void *base, struct xt_table_info *newinfo)
 892 {
 893         const struct xt_entry_match *ematch;
 894         const struct xt_entry_target *t;
 895         unsigned int entry_offset;
 896         int off, i, ret;
 897
 898         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
 899         entry_offset = (void *)e - base;
 900         xt_ematch_foreach(ematch, e)
 901                 off += xt_compat_match_offset(ematch->u.kernel.match);
 902         t = ipt_get_target_c(e);
 903         off += xt_compat_target_offset(t->u.kernel.target);
 904         newinfo->size -= off;
 905         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
 906         if (ret)
 907                 return ret;
 908
 909         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 910                 if (info->hook_entry[i] &&
 911                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
 912                         newinfo->hook_entry[i] -= off;
 913                 if (info->underflow[i] &&
 914                     (e < (struct ipt_entry *)(base + info->underflow[i])))
 915                         newinfo->underflow[i] -= off;
 916         }
 917         return 0;
 918 }
 919
 920 static int compat_table_info(const struct xt_table_info *info,
 921                              struct xt_table_info *newinfo)
 922 {
 923         struct ipt_entry *iter;
 924         const void *loc_cpu_entry;
 925         int ret;
 926
 927         if (!newinfo || !info)
 928                 return -EINVAL;
 929
 930         /* we dont care about newinfo->entries */
 931         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 932         newinfo->initial_entries = 0;
 933         loc_cpu_entry = info->entries;
 934         ret = xt_compat_init_offsets(AF_INET, info->number);
 935         if (ret)
 936                 return ret;
 937         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 938                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 939                 if (ret != 0)
 940                         return ret;
 941         }
 942         return 0;
 943 }
 944 #endif
 945
 946 static int get_info(struct net *net, void __user *user, const int *len)
 947 {
 948         char name[XT_TABLE_MAXNAMELEN];
 949         struct xt_table *t;
 950         int ret;
 951
 952         if (*len != sizeof(struct ipt_getinfo))
 953                 return -EINVAL;
 954
 955         if (copy_from_user(name, user, sizeof(name)) != 0)
 956                 return -EFAULT;
 957
 958         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 959 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
 960         if (in_compat_syscall())
 961                 xt_compat_lock(AF_INET);
 962 #endif
 963         t = xt_request_find_table_lock(net, AF_INET, name);
 964         if (!IS_ERR(t)) {
 965                 struct ipt_getinfo info;
 966                 const struct xt_table_info *private = t->private;
 967 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
 968                 struct xt_table_info tmp;
 969
 970                 if (in_compat_syscall()) {
 971                         ret = compat_table_info(private, &tmp);
 972                         xt_compat_flush_offsets(AF_INET);
 973                         private = &tmp;
 974                 }
 975 #endif
 976                 memset(&info, 0, sizeof(info));
 977                 info.valid_hooks = t->valid_hooks;
 978                 memcpy(info.hook_entry, private->hook_entry,
 979                        sizeof(info.hook_entry));
 980                 memcpy(info.underflow, private->underflow,
 981                        sizeof(info.underflow));
 982                 info.num_entries = private->number;
 983                 info.size = private->size;
 984                 strcpy(info.name, name);
 985
 986                 if (copy_to_user(user, &info, *len) != 0)
 987                         ret = -EFAULT;
 988                 else
 989                         ret = 0;
 990
 991                 xt_table_unlock(t);
 992                 module_put(t->me);
 993         } else
 994                 ret = PTR_ERR(t);
 995 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
 996         if (in_compat_syscall())
 997                 xt_compat_unlock(AF_INET);
 998 #endif
 999         return ret;
1000 }
1001
1002 static int
1003 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1004             const int *len)
1005 {
1006         int ret;
1007         struct ipt_get_entries get;
1008         struct xt_table *t;
1009
1010         if (*len < sizeof(get))
1011                 return -EINVAL;
1012         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1013                 return -EFAULT;
1014         if (*len != sizeof(struct ipt_get_entries) + get.size)
1015                 return -EINVAL;
1016         get.name[sizeof(get.name) - 1] = '\0';
1017
1018         t = xt_find_table_lock(net, AF_INET, get.name);
1019         if (!IS_ERR(t)) {
1020                 const struct xt_table_info *private = t->private;
1021                 if (get.size == private->size)
1022                         ret = copy_entries_to_user(private->size,
1023                                                    t, uptr->entrytable);
1024                 else
1025                         ret = -EAGAIN;
1026
1027                 module_put(t->me);
1028                 xt_table_unlock(t);
1029         } else
1030                 ret = PTR_ERR(t);
1031
1032         return ret;
1033 }
1034
1035 static int
1036 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1037              struct xt_table_info *newinfo, unsigned int num_counters,
1038              void __user *counters_ptr)
1039 {
1040         int ret;
1041         struct xt_table *t;
1042         struct xt_table_info *oldinfo;
1043         struct xt_counters *counters;
1044         struct ipt_entry *iter;
1045
1046         counters = xt_counters_alloc(num_counters);
1047         if (!counters) {
1048                 ret = -ENOMEM;
1049                 goto out;
1050         }
1051
1052         t = xt_request_find_table_lock(net, AF_INET, name);
1053         if (IS_ERR(t)) {
1054                 ret = PTR_ERR(t);
1055                 goto free_newinfo_counters_untrans;
1056         }
1057
1058         /* You lied! */
1059         if (valid_hooks != t->valid_hooks) {
1060                 ret = -EINVAL;
1061                 goto put_module;
1062         }
1063
1064         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1065         if (!oldinfo)
1066                 goto put_module;
1067
1068         /* Update module usage count based on number of rules */
1069         if ((oldinfo->number > oldinfo->initial_entries) ||
1070             (newinfo->number <= oldinfo->initial_entries))
1071                 module_put(t->me);
1072         if ((oldinfo->number > oldinfo->initial_entries) &&
1073             (newinfo->number <= oldinfo->initial_entries))
1074                 module_put(t->me);
1075
1076         xt_table_unlock(t);
1077
1078         get_old_counters(oldinfo, counters);
1079
1080         /* Decrease module usage counts and free resource */
1081         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1082                 cleanup_entry(iter, net);
1083
1084         xt_free_table_info(oldinfo);
1085         if (copy_to_user(counters_ptr, counters,
1086                          sizeof(struct xt_counters) * num_counters) != 0) {
1087                 /* Silent error, can't fail, new table is already in place */
1088                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1089         }
1090         vfree(counters);
1091         return 0;
1092
1093  put_module:
1094         module_put(t->me);
1095         xt_table_unlock(t);
1096  free_newinfo_counters_untrans:
1097         vfree(counters);
1098  out:
1099         return ret;
1100 }
1101
1102 static int
1103 do_replace(struct net *net, sockptr_t arg, unsigned int len)
1104 {
1105         int ret;
1106         struct ipt_replace tmp;
1107         struct xt_table_info *newinfo;
1108         void *loc_cpu_entry;
1109         struct ipt_entry *iter;
1110
1111         if (len < sizeof(tmp))
1112                 return -EINVAL;
1113         if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0)
1114                 return -EFAULT;
1115
1116         /* overflow check */
1117         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1118                 return -ENOMEM;
1119         if (tmp.num_counters == 0)
1120                 return -EINVAL;
1121
1122         tmp.name[sizeof(tmp.name)-1] = 0;
1123
1124         newinfo = xt_alloc_table_info(tmp.size);
1125         if (!newinfo)
1126                 return -ENOMEM;
1127
1128         loc_cpu_entry = newinfo->entries;
1129         if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp),
1130                         tmp.size) != 0) {
1131                 ret = -EFAULT;
1132                 goto free_newinfo;
1133         }
1134
1135         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1136         if (ret != 0)
1137                 goto free_newinfo;
1138
1139         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1140                            tmp.num_counters, tmp.counters);
1141         if (ret)
1142                 goto free_newinfo_untrans;
1143         return 0;
1144
1145  free_newinfo_untrans:
1146         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1147                 cleanup_entry(iter, net);
1148  free_newinfo:
1149         xt_free_table_info(newinfo);
1150         return ret;
1151 }
1152
1153 static int
1154 do_add_counters(struct net *net, sockptr_t arg, unsigned int len)
1155 {
1156         unsigned int i;
1157         struct xt_counters_info tmp;
1158         struct xt_counters *paddc;
1159         struct xt_table *t;
1160         const struct xt_table_info *private;
1161         int ret = 0;
1162         struct ipt_entry *iter;
1163         unsigned int addend;
1164
1165         paddc = xt_copy_counters(arg, len, &tmp);
1166         if (IS_ERR(paddc))
1167                 return PTR_ERR(paddc);
1168
1169         t = xt_find_table_lock(net, AF_INET, tmp.name);
1170         if (IS_ERR(t)) {
1171                 ret = PTR_ERR(t);
1172                 goto free;
1173         }
1174
1175         local_bh_disable();
1176         private = t->private;
1177         if (private->number != tmp.num_counters) {
1178                 ret = -EINVAL;
1179                 goto unlock_up_free;
1180         }
1181
1182         i = 0;
1183         addend = xt_write_recseq_begin();
1184         xt_entry_foreach(iter, private->entries, private->size) {
1185                 struct xt_counters *tmp;
1186
1187                 tmp = xt_get_this_cpu_counter(&iter->counters);
1188                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1189                 ++i;
1190         }
1191         xt_write_recseq_end(addend);
1192  unlock_up_free:
1193         local_bh_enable();
1194         xt_table_unlock(t);
1195         module_put(t->me);
1196  free:
1197         vfree(paddc);
1198
1199         return ret;
1200 }
1201
1202 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1203 struct compat_ipt_replace {
1204         char                    name[XT_TABLE_MAXNAMELEN];
1205         u32                     valid_hooks;
1206         u32                     num_entries;
1207         u32                     size;
1208         u32                     hook_entry[NF_INET_NUMHOOKS];
1209         u32                     underflow[NF_INET_NUMHOOKS];
1210         u32                     num_counters;
1211         compat_uptr_t           counters;       /* struct xt_counters * */
1212         struct compat_ipt_entry entries[];
1213 };
1214
1215 static int
1216 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1217                           unsigned int *size, struct xt_counters *counters,
1218                           unsigned int i)
1219 {
1220         struct xt_entry_target *t;
1221         struct compat_ipt_entry __user *ce;
1222         u_int16_t target_offset, next_offset;
1223         compat_uint_t origsize;
1224         const struct xt_entry_match *ematch;
1225         int ret = 0;
1226
1227         origsize = *size;
1228         ce = *dstptr;
1229         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1230             copy_to_user(&ce->counters, &counters[i],
1231             sizeof(counters[i])) != 0)
1232                 return -EFAULT;
1233
1234         *dstptr += sizeof(struct compat_ipt_entry);
1235         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1236
1237         xt_ematch_foreach(ematch, e) {
1238                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1239                 if (ret != 0)
1240                         return ret;
1241         }
1242         target_offset = e->target_offset - (origsize - *size);
1243         t = ipt_get_target(e);
1244         ret = xt_compat_target_to_user(t, dstptr, size);
1245         if (ret)
1246                 return ret;
1247         next_offset = e->next_offset - (origsize - *size);
1248         if (put_user(target_offset, &ce->target_offset) != 0 ||
1249             put_user(next_offset, &ce->next_offset) != 0)
1250                 return -EFAULT;
1251         return 0;
1252 }
1253
1254 static int
1255 compat_find_calc_match(struct xt_entry_match *m,
1256                        const struct ipt_ip *ip,
1257                        int *size)
1258 {
1259         struct xt_match *match;
1260
1261         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1262                                       m->u.user.revision);
1263         if (IS_ERR(match))
1264                 return PTR_ERR(match);
1265
1266         m->u.kernel.match = match;
1267         *size += xt_compat_match_offset(match);
1268         return 0;
1269 }
1270
1271 static void compat_release_entry(struct compat_ipt_entry *e)
1272 {
1273         struct xt_entry_target *t;
1274         struct xt_entry_match *ematch;
1275
1276         /* Cleanup all matches */
1277         xt_ematch_foreach(ematch, e)
1278                 module_put(ematch->u.kernel.match->me);
1279         t = compat_ipt_get_target(e);
1280         module_put(t->u.kernel.target->me);
1281 }
1282
1283 static int
1284 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1285                                   struct xt_table_info *newinfo,
1286                                   unsigned int *size,
1287                                   const unsigned char *base,
1288                                   const unsigned char *limit)
1289 {
1290         struct xt_entry_match *ematch;
1291         struct xt_entry_target *t;
1292         struct xt_target *target;
1293         unsigned int entry_offset;
1294         unsigned int j;
1295         int ret, off;
1296
1297         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1298             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1299             (unsigned char *)e + e->next_offset > limit)
1300                 return -EINVAL;
1301
1302         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1303                              sizeof(struct compat_xt_entry_target))
1304                 return -EINVAL;
1305
1306         if (!ip_checkentry(&e->ip))
1307                 return -EINVAL;
1308
1309         ret = xt_compat_check_entry_offsets(e, e->elems,
1310                                             e->target_offset, e->next_offset);
1311         if (ret)
1312                 return ret;
1313
1314         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1315         entry_offset = (void *)e - (void *)base;
1316         j = 0;
1317         xt_ematch_foreach(ematch, e) {
1318                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1319                 if (ret != 0)
1320                         goto release_matches;
1321                 ++j;
1322         }
1323
1324         t = compat_ipt_get_target(e);
1325         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1326                                         t->u.user.revision);
1327         if (IS_ERR(target)) {
1328                 ret = PTR_ERR(target);
1329                 goto release_matches;
1330         }
1331         t->u.kernel.target = target;
1332
1333         off += xt_compat_target_offset(target);
1334         *size += off;
1335         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1336         if (ret)
1337                 goto out;
1338
1339         return 0;
1340
1341 out:
1342         module_put(t->u.kernel.target->me);
1343 release_matches:
1344         xt_ematch_foreach(ematch, e) {
1345                 if (j-- == 0)
1346                         break;
1347                 module_put(ematch->u.kernel.match->me);
1348         }
1349         return ret;
1350 }
1351
1352 static void
1353 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1354                             unsigned int *size,
1355                             struct xt_table_info *newinfo, unsigned char *base)
1356 {
1357         struct xt_entry_target *t;
1358         struct ipt_entry *de;
1359         unsigned int origsize;
1360         int h;
1361         struct xt_entry_match *ematch;
1362
1363         origsize = *size;
1364         de = *dstptr;
1365         memcpy(de, e, sizeof(struct ipt_entry));
1366         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1367
1368         *dstptr += sizeof(struct ipt_entry);
1369         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1370
1371         xt_ematch_foreach(ematch, e)
1372                 xt_compat_match_from_user(ematch, dstptr, size);
1373
1374         de->target_offset = e->target_offset - (origsize - *size);
1375         t = compat_ipt_get_target(e);
1376         xt_compat_target_from_user(t, dstptr, size);
1377
1378         de->next_offset = e->next_offset - (origsize - *size);
1379
1380         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1381                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1382                         newinfo->hook_entry[h] -= origsize - *size;
1383                 if ((unsigned char *)de - base < newinfo->underflow[h])
1384                         newinfo->underflow[h] -= origsize - *size;
1385         }
1386 }
1387
1388 static int
1389 translate_compat_table(struct net *net,
1390                        struct xt_table_info **pinfo,
1391                        void **pentry0,
1392                        const struct compat_ipt_replace *compatr)
1393 {
1394         unsigned int i, j;
1395         struct xt_table_info *newinfo, *info;
1396         void *pos, *entry0, *entry1;
1397         struct compat_ipt_entry *iter0;
1398         struct ipt_replace repl;
1399         unsigned int size;
1400         int ret;
1401
1402         info = *pinfo;
1403         entry0 = *pentry0;
1404         size = compatr->size;
1405         info->number = compatr->num_entries;
1406
1407         j = 0;
1408         xt_compat_lock(AF_INET);
1409         ret = xt_compat_init_offsets(AF_INET, compatr->num_entries);
1410         if (ret)
1411                 goto out_unlock;
1412         /* Walk through entries, checking offsets. */
1413         xt_entry_foreach(iter0, entry0, compatr->size) {
1414                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1415                                                         entry0,
1416                                                         entry0 + compatr->size);
1417                 if (ret != 0)
1418                         goto out_unlock;
1419                 ++j;
1420         }
1421
1422         ret = -EINVAL;
1423         if (j != compatr->num_entries)
1424                 goto out_unlock;
1425
1426         ret = -ENOMEM;
1427         newinfo = xt_alloc_table_info(size);
1428         if (!newinfo)
1429                 goto out_unlock;
1430
1431         memset(newinfo->entries, 0, size);
1432
1433         newinfo->number = compatr->num_entries;
1434         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1435                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1436                 newinfo->underflow[i] = compatr->underflow[i];
1437         }
1438         entry1 = newinfo->entries;
1439         pos = entry1;
1440         size = compatr->size;
1441         xt_entry_foreach(iter0, entry0, compatr->size)
1442                 compat_copy_entry_from_user(iter0, &pos, &size,
1443                                             newinfo, entry1);
1444
1445         /* all module references in entry0 are now gone.
1446          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1447          * generated by 64bit userspace.
1448          *
1449          * Call standard translate_table() to validate all hook_entrys,
1450          * underflows, check for loops, etc.
1451          */
1452         xt_compat_flush_offsets(AF_INET);
1453         xt_compat_unlock(AF_INET);
1454
1455         memcpy(&repl, compatr, sizeof(*compatr));
1456
1457         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1458                 repl.hook_entry[i] = newinfo->hook_entry[i];
1459                 repl.underflow[i] = newinfo->underflow[i];
1460         }
1461
1462         repl.num_counters = 0;
1463         repl.counters = NULL;
1464         repl.size = newinfo->size;
1465         ret = translate_table(net, newinfo, entry1, &repl);
1466         if (ret)
1467                 goto free_newinfo;
1468
1469         *pinfo = newinfo;
1470         *pentry0 = entry1;
1471         xt_free_table_info(info);
1472         return 0;
1473
1474 free_newinfo:
1475         xt_free_table_info(newinfo);
1476         return ret;
1477 out_unlock:
1478         xt_compat_flush_offsets(AF_INET);
1479         xt_compat_unlock(AF_INET);
1480         xt_entry_foreach(iter0, entry0, compatr->size) {
1481                 if (j-- == 0)
1482                         break;
1483                 compat_release_entry(iter0);
1484         }
1485         return ret;
1486 }
1487
1488 static int
1489 compat_do_replace(struct net *net, sockptr_t arg, unsigned int len)
1490 {
1491         int ret;
1492         struct compat_ipt_replace tmp;
1493         struct xt_table_info *newinfo;
1494         void *loc_cpu_entry;
1495         struct ipt_entry *iter;
1496
1497         if (len < sizeof(tmp))
1498                 return -EINVAL;
1499         if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0)
1500                 return -EFAULT;
1501
1502         /* overflow check */
1503         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1504                 return -ENOMEM;
1505         if (tmp.num_counters == 0)
1506                 return -EINVAL;
1507
1508         tmp.name[sizeof(tmp.name)-1] = 0;
1509
1510         newinfo = xt_alloc_table_info(tmp.size);
1511         if (!newinfo)
1512                 return -ENOMEM;
1513
1514         loc_cpu_entry = newinfo->entries;
1515         if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp),
1516                         tmp.size) != 0) {
1517                 ret = -EFAULT;
1518                 goto free_newinfo;
1519         }
1520
1521         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1522         if (ret != 0)
1523                 goto free_newinfo;
1524
1525         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1526                            tmp.num_counters, compat_ptr(tmp.counters));
1527         if (ret)
1528                 goto free_newinfo_untrans;
1529         return 0;
1530
1531  free_newinfo_untrans:
1532         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1533                 cleanup_entry(iter, net);
1534  free_newinfo:
1535         xt_free_table_info(newinfo);
1536         return ret;
1537 }
1538
1539 struct compat_ipt_get_entries {
1540         char name[XT_TABLE_MAXNAMELEN];
1541         compat_uint_t size;
1542         struct compat_ipt_entry entrytable[];
1543 };
1544
1545 static int
1546 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1547                             void __user *userptr)
1548 {
1549         struct xt_counters *counters;
1550         const struct xt_table_info *private = table->private;
1551         void __user *pos;
1552         unsigned int size;
1553         int ret = 0;
1554         unsigned int i = 0;
1555         struct ipt_entry *iter;
1556
1557         counters = alloc_counters(table);
1558         if (IS_ERR(counters))
1559                 return PTR_ERR(counters);
1560
1561         pos = userptr;
1562         size = total_size;
1563         xt_entry_foreach(iter, private->entries, total_size) {
1564                 ret = compat_copy_entry_to_user(iter, &pos,
1565                                                 &size, counters, i++);
1566                 if (ret != 0)
1567                         break;
1568         }
1569
1570         vfree(counters);
1571         return ret;
1572 }
1573
1574 static int
1575 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1576                    int *len)
1577 {
1578         int ret;
1579         struct compat_ipt_get_entries get;
1580         struct xt_table *t;
1581
1582         if (*len < sizeof(get))
1583                 return -EINVAL;
1584
1585         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1586                 return -EFAULT;
1587
1588         if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1589                 return -EINVAL;
1590
1591         get.name[sizeof(get.name) - 1] = '\0';
1592
1593         xt_compat_lock(AF_INET);
1594         t = xt_find_table_lock(net, AF_INET, get.name);
1595         if (!IS_ERR(t)) {
1596                 const struct xt_table_info *private = t->private;
1597                 struct xt_table_info info;
1598                 ret = compat_table_info(private, &info);
1599                 if (!ret && get.size == info.size)
1600                         ret = compat_copy_entries_to_user(private->size,
1601                                                           t, uptr->entrytable);
1602                 else if (!ret)
1603                         ret = -EAGAIN;
1604
1605                 xt_compat_flush_offsets(AF_INET);
1606                 module_put(t->me);
1607                 xt_table_unlock(t);
1608         } else
1609                 ret = PTR_ERR(t);
1610
1611         xt_compat_unlock(AF_INET);
1612         return ret;
1613 }
1614 #endif
1615
1616 static int
1617 do_ipt_set_ctl(struct sock *sk, int cmd, sockptr_t arg, unsigned int len)
1618 {
1619         int ret;
1620
1621         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1622                 return -EPERM;
1623
1624         switch (cmd) {
1625         case IPT_SO_SET_REPLACE:
1626 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1627                 if (in_compat_syscall())
1628                         ret = compat_do_replace(sock_net(sk), arg, len);
1629                 else
1630 #endif
1631                         ret = do_replace(sock_net(sk), arg, len);
1632                 break;
1633
1634         case IPT_SO_SET_ADD_COUNTERS:
1635                 ret = do_add_counters(sock_net(sk), arg, len);
1636                 break;
1637
1638         default:
1639                 ret = -EINVAL;
1640         }
1641
1642         return ret;
1643 }
1644
1645 static int
1646 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1647 {
1648         int ret;
1649
1650         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1651                 return -EPERM;
1652
1653         switch (cmd) {
1654         case IPT_SO_GET_INFO:
1655                 ret = get_info(sock_net(sk), user, len);
1656                 break;
1657
1658         case IPT_SO_GET_ENTRIES:
1659 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1660                 if (in_compat_syscall())
1661                         ret = compat_get_entries(sock_net(sk), user, len);
1662                 else
1663 #endif
1664                         ret = get_entries(sock_net(sk), user, len);
1665                 break;
1666
1667         case IPT_SO_GET_REVISION_MATCH:
1668         case IPT_SO_GET_REVISION_TARGET: {
1669                 struct xt_get_revision rev;
1670                 int target;
1671
1672                 if (*len != sizeof(rev)) {
1673                         ret = -EINVAL;
1674                         break;
1675                 }
1676                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1677                         ret = -EFAULT;
1678                         break;
1679                 }
1680                 rev.name[sizeof(rev.name)-1] = 0;
1681
1682                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1683                         target = 1;
1684                 else
1685                         target = 0;
1686
1687                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1688                                                          rev.revision,
1689                                                          target, &ret),
1690                                         "ipt_%s", rev.name);
1691                 break;
1692         }
1693
1694         default:
1695                 ret = -EINVAL;
1696         }
1697
1698         return ret;
1699 }
1700
1701 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1702 {
1703         struct xt_table_info *private;
1704         void *loc_cpu_entry;
1705         struct module *table_owner = table->me;
1706         struct ipt_entry *iter;
1707
1708         private = xt_unregister_table(table);
1709
1710         /* Decrease module usage counts and free resources */
1711         loc_cpu_entry = private->entries;
1712         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1713                 cleanup_entry(iter, net);
1714         if (private->number > private->initial_entries)
1715                 module_put(table_owner);
1716         xt_free_table_info(private);
1717 }
1718
1719 int ipt_register_table(struct net *net, const struct xt_table *table,
1720                        const struct ipt_replace *repl,
1721                        const struct nf_hook_ops *template_ops)
1722 {
1723         struct nf_hook_ops *ops;
1724         unsigned int num_ops;
1725         int ret, i;
1726         struct xt_table_info *newinfo;
1727         struct xt_table_info bootstrap = {0};
1728         void *loc_cpu_entry;
1729         struct xt_table *new_table;
1730
1731         newinfo = xt_alloc_table_info(repl->size);
1732         if (!newinfo)
1733                 return -ENOMEM;
1734
1735         loc_cpu_entry = newinfo->entries;
1736         memcpy(loc_cpu_entry, repl->entries, repl->size);
1737
1738         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1739         if (ret != 0) {
1740                 xt_free_table_info(newinfo);
1741                 return ret;
1742         }
1743
1744         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1745         if (IS_ERR(new_table)) {
1746                 struct ipt_entry *iter;
1747
1748                 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1749                         cleanup_entry(iter, net);
1750                 xt_free_table_info(newinfo);
1751                 return PTR_ERR(new_table);
1752         }
1753
1754         /* No template? No need to do anything. This is used by 'nat' table, it registers
1755          * with the nat core instead of the netfilter core.
1756          */
1757         if (!template_ops)
1758                 return 0;
1759
1760         num_ops = hweight32(table->valid_hooks);
1761         if (num_ops == 0) {
1762                 ret = -EINVAL;
1763                 goto out_free;
1764         }
1765
1766         ops = kmemdup(template_ops, sizeof(*ops) * num_ops, GFP_KERNEL);
1767         if (!ops) {
1768                 ret = -ENOMEM;
1769                 goto out_free;
1770         }
1771
1772         for (i = 0; i < num_ops; i++)
1773                 ops[i].priv = new_table;
1774
1775         new_table->ops = ops;
1776
1777         ret = nf_register_net_hooks(net, ops, num_ops);
1778         if (ret != 0)
1779                 goto out_free;
1780
1781         return ret;
1782
1783 out_free:
1784         __ipt_unregister_table(net, new_table);
1785         return ret;
1786 }
1787
1788 void ipt_unregister_table_pre_exit(struct net *net, const char *name)
1789 {
1790         struct xt_table *table = xt_find_table(net, NFPROTO_IPV4, name);
1791
1792         if (table)
1793                 nf_unregister_net_hooks(net, table->ops, hweight32(table->valid_hooks));
1794 }
1795
1796 void ipt_unregister_table_exit(struct net *net, const char *name)
1797 {
1798         struct xt_table *table = xt_find_table(net, NFPROTO_IPV4, name);
1799
1800         if (table)
1801                 __ipt_unregister_table(net, table);
1802 }
1803
1804 static struct xt_target ipt_builtin_tg[] __read_mostly = {
1805         {
1806                 .name             = XT_STANDARD_TARGET,
1807                 .targetsize       = sizeof(int),
1808                 .family           = NFPROTO_IPV4,
1809 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1810                 .compatsize       = sizeof(compat_int_t),
1811                 .compat_from_user = compat_standard_from_user,
1812                 .compat_to_user   = compat_standard_to_user,
1813 #endif
1814         },
1815         {
1816                 .name             = XT_ERROR_TARGET,
1817                 .target           = ipt_error,
1818                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1819                 .family           = NFPROTO_IPV4,
1820         },
1821 };
1822
1823 static struct nf_sockopt_ops ipt_sockopts = {
1824         .pf             = PF_INET,
1825         .set_optmin     = IPT_BASE_CTL,
1826         .set_optmax     = IPT_SO_SET_MAX+1,
1827         .set            = do_ipt_set_ctl,
1828         .get_optmin     = IPT_BASE_CTL,
1829         .get_optmax     = IPT_SO_GET_MAX+1,
1830         .get            = do_ipt_get_ctl,
1831         .owner          = THIS_MODULE,
1832 };
1833
1834 static int __net_init ip_tables_net_init(struct net *net)
1835 {
1836         return xt_proto_init(net, NFPROTO_IPV4);
1837 }
1838
1839 static void __net_exit ip_tables_net_exit(struct net *net)
1840 {
1841         xt_proto_fini(net, NFPROTO_IPV4);
1842 }
1843
1844 static struct pernet_operations ip_tables_net_ops = {
1845         .init = ip_tables_net_init,
1846         .exit = ip_tables_net_exit,
1847 };
1848
1849 static int __init ip_tables_init(void)
1850 {
1851         int ret;
1852
1853         ret = register_pernet_subsys(&ip_tables_net_ops);
1854         if (ret < 0)
1855                 goto err1;
1856
1857         /* No one else will be downing sem now, so we won't sleep */
1858         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1859         if (ret < 0)
1860                 goto err2;
1861
1862         /* Register setsockopt */
1863         ret = nf_register_sockopt(&ipt_sockopts);
1864         if (ret < 0)
1865                 goto err4;
1866
1867         return 0;
1868
1869 err4:
1870         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1871 err2:
1872         unregister_pernet_subsys(&ip_tables_net_ops);
1873 err1:
1874         return ret;
1875 }
1876
1877 static void __exit ip_tables_fini(void)
1878 {
1879         nf_unregister_sockopt(&ipt_sockopts);
1880
1881         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1882         unregister_pernet_subsys(&ip_tables_net_ops);
1883 }
1884
1885 EXPORT_SYMBOL(ipt_register_table);
1886 EXPORT_SYMBOL(ipt_unregister_table_pre_exit);
1887 EXPORT_SYMBOL(ipt_unregister_table_exit);
1888 EXPORT_SYMBOL(ipt_do_table);
1889 module_init(ip_tables_init);
1890 module_exit(ip_tables_fini);