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.net     = state->net;
 265         acpar.in      = state->in;
 266         acpar.out     = state->out;
 267         acpar.family  = NFPROTO_IPV4;
 268         acpar.hooknum = hook;
 269
 270         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 271         local_bh_disable();
 272         addend = xt_write_recseq_begin();
 273         private = table->private;
 274         cpu        = smp_processor_id();
 275         /*
 276          * Ensure we load private-> members after we've fetched the base
 277          * pointer.
 278          */
 279         smp_read_barrier_depends();
 280         table_base = private->entries;
 281         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 282
 283         /* Switch to alternate jumpstack if we're being invoked via TEE.
 284          * TEE issues XT_CONTINUE verdict on original skb so we must not
 285          * clobber the jumpstack.
 286          *
 287          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 288          * but it is no problem since absolute verdict is issued by these.
 289          */
 290         if (static_key_false(&xt_tee_enabled))
 291                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 292
 293         e = get_entry(table_base, private->hook_entry[hook]);
 294
 295         do {
 296                 const struct xt_entry_target *t;
 297                 const struct xt_entry_match *ematch;
 298                 struct xt_counters *counter;
 299
 300                 IP_NF_ASSERT(e);
 301                 if (!ip_packet_match(ip, indev, outdev,
 302                     &e->ip, acpar.fragoff)) {
 303  no_match:
 304                         e = ipt_next_entry(e);
 305                         continue;
 306                 }
 307
 308                 xt_ematch_foreach(ematch, e) {
 309                         acpar.match     = ematch->u.kernel.match;
 310                         acpar.matchinfo = ematch->data;
 311                         if (!acpar.match->match(skb, &acpar))
 312                                 goto no_match;
 313                 }
 314
 315                 counter = xt_get_this_cpu_counter(&e->counters);
 316                 ADD_COUNTER(*counter, skb->len, 1);
 317
 318                 t = ipt_get_target(e);
 319                 IP_NF_ASSERT(t->u.kernel.target);
 320
 321 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 322                 /* The packet is traced: log it */
 323                 if (unlikely(skb->nf_trace))
 324                         trace_packet(state->net, skb, hook, state->in,
 325                                      state->out, table->name, private, e);
 326 #endif
 327                 /* Standard target? */
 328                 if (!t->u.kernel.target->target) {
 329                         int v;
 330
 331                         v = ((struct xt_standard_target *)t)->verdict;
 332                         if (v < 0) {
 333                                 /* Pop from stack? */
 334                                 if (v != XT_RETURN) {
 335                                         verdict = (unsigned int)(-v) - 1;
 336                                         break;
 337                                 }
 338                                 if (stackidx == 0) {
 339                                         e = get_entry(table_base,
 340                                             private->underflow[hook]);
 341                                 } else {
 342                                         e = jumpstack[--stackidx];
 343                                         e = ipt_next_entry(e);
 344                                 }
 345                                 continue;
 346                         }
 347                         if (table_base + v != ipt_next_entry(e) &&
 348                             !(e->ip.flags & IPT_F_GOTO)) {
 349                                 if (unlikely(stackidx >= private->stacksize)) {
 350                                         verdict = NF_DROP;
 351                                         break;
 352                                 }
 353                                 jumpstack[stackidx++] = e;
 354                         }
 355
 356                         e = get_entry(table_base, v);
 357                         continue;
 358                 }
 359
 360                 acpar.target   = t->u.kernel.target;
 361                 acpar.targinfo = t->data;
 362
 363                 verdict = t->u.kernel.target->target(skb, &acpar);
 364                 /* Target might have changed stuff. */
 365                 ip = ip_hdr(skb);
 366                 if (verdict == XT_CONTINUE)
 367                         e = ipt_next_entry(e);
 368                 else
 369                         /* Verdict */
 370                         break;
 371         } while (!acpar.hotdrop);
 372
 373         xt_write_recseq_end(addend);
 374         local_bh_enable();
 375
 376         if (acpar.hotdrop)
 377                 return NF_DROP;
 378         else return verdict;
 379 }
 380
 381 /* Figures out from what hook each rule can be called: returns 0 if
 382    there are loops.  Puts hook bitmask in comefrom. */
 383 static int
 384 mark_source_chains(const struct xt_table_info *newinfo,
 385                    unsigned int valid_hooks, void *entry0,
 386                    unsigned int *offsets)
 387 {
 388         unsigned int hook;
 389
 390         /* No recursion; use packet counter to save back ptrs (reset
 391            to 0 as we leave), and comefrom to save source hook bitmask */
 392         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 393                 unsigned int pos = newinfo->hook_entry[hook];
 394                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 395
 396                 if (!(valid_hooks & (1 << hook)))
 397                         continue;
 398
 399                 /* Set initial back pointer. */
 400                 e->counters.pcnt = pos;
 401
 402                 for (;;) {
 403                         const struct xt_standard_target *t
 404                                 = (void *)ipt_get_target_c(e);
 405                         int visited = e->comefrom & (1 << hook);
 406
 407                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 408                                 return 0;
 409
 410                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 411
 412                         /* Unconditional return/END. */
 413                         if ((unconditional(e) &&
 414                              (strcmp(t->target.u.user.name,
 415                                      XT_STANDARD_TARGET) == 0) &&
 416                              t->verdict < 0) || visited) {
 417                                 unsigned int oldpos, size;
 418
 419                                 if ((strcmp(t->target.u.user.name,
 420                                             XT_STANDARD_TARGET) == 0) &&
 421                                     t->verdict < -NF_MAX_VERDICT - 1)
 422                                         return 0;
 423
 424                                 /* Return: backtrack through the last
 425                                    big jump. */
 426                                 do {
 427                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 428                                         oldpos = pos;
 429                                         pos = e->counters.pcnt;
 430                                         e->counters.pcnt = 0;
 431
 432                                         /* We're at the start. */
 433                                         if (pos == oldpos)
 434                                                 goto next;
 435
 436                                         e = (struct ipt_entry *)
 437                                                 (entry0 + pos);
 438                                 } while (oldpos == pos + e->next_offset);
 439
 440                                 /* Move along one */
 441                                 size = e->next_offset;
 442                                 e = (struct ipt_entry *)
 443                                         (entry0 + pos + size);
 444                                 if (pos + size >= newinfo->size)
 445                                         return 0;
 446                                 e->counters.pcnt = pos;
 447                                 pos += size;
 448                         } else {
 449                                 int newpos = t->verdict;
 450
 451                                 if (strcmp(t->target.u.user.name,
 452                                            XT_STANDARD_TARGET) == 0 &&
 453                                     newpos >= 0) {
 454                                         /* This a jump; chase it. */
 455                                         if (!xt_find_jump_offset(offsets, newpos,
 456                                                                  newinfo->number))
 457                                                 return 0;
 458                                         e = (struct ipt_entry *)
 459                                                 (entry0 + newpos);
 460                                 } else {
 461                                         /* ... this is a fallthru */
 462                                         newpos = pos + e->next_offset;
 463                                         if (newpos >= newinfo->size)
 464                                                 return 0;
 465                                 }
 466                                 e = (struct ipt_entry *)
 467                                         (entry0 + newpos);
 468                                 e->counters.pcnt = pos;
 469                                 pos = newpos;
 470                         }
 471                 }
 472 next:           ;
 473         }
 474         return 1;
 475 }
 476
 477 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 478 {
 479         struct xt_mtdtor_param par;
 480
 481         par.net       = net;
 482         par.match     = m->u.kernel.match;
 483         par.matchinfo = m->data;
 484         par.family    = NFPROTO_IPV4;
 485         if (par.match->destroy != NULL)
 486                 par.match->destroy(&par);
 487         module_put(par.match->me);
 488 }
 489
 490 static int
 491 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 492 {
 493         const struct ipt_ip *ip = par->entryinfo;
 494
 495         par->match     = m->u.kernel.match;
 496         par->matchinfo = m->data;
 497
 498         return xt_check_match(par, m->u.match_size - sizeof(*m),
 499                               ip->proto, ip->invflags & IPT_INV_PROTO);
 500 }
 501
 502 static int
 503 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 504 {
 505         struct xt_match *match;
 506         int ret;
 507
 508         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 509                                       m->u.user.revision);
 510         if (IS_ERR(match))
 511                 return PTR_ERR(match);
 512         m->u.kernel.match = match;
 513
 514         ret = check_match(m, par);
 515         if (ret)
 516                 goto err;
 517
 518         return 0;
 519 err:
 520         module_put(m->u.kernel.match->me);
 521         return ret;
 522 }
 523
 524 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 525 {
 526         struct xt_entry_target *t = ipt_get_target(e);
 527         struct xt_tgchk_param par = {
 528                 .net       = net,
 529                 .table     = name,
 530                 .entryinfo = e,
 531                 .target    = t->u.kernel.target,
 532                 .targinfo  = t->data,
 533                 .hook_mask = e->comefrom,
 534                 .family    = NFPROTO_IPV4,
 535         };
 536
 537         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 538                                e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 539 }
 540
 541 static int
 542 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 543                  unsigned int size,
 544                  struct xt_percpu_counter_alloc_state *alloc_state)
 545 {
 546         struct xt_entry_target *t;
 547         struct xt_target *target;
 548         int ret;
 549         unsigned int j;
 550         struct xt_mtchk_param mtpar;
 551         struct xt_entry_match *ematch;
 552
 553         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 554                 return -ENOMEM;
 555
 556         j = 0;
 557         memset(&mtpar, 0, sizeof(mtpar));
 558         mtpar.net       = net;
 559         mtpar.table     = name;
 560         mtpar.entryinfo = &e->ip;
 561         mtpar.hook_mask = e->comefrom;
 562         mtpar.family    = NFPROTO_IPV4;
 563         xt_ematch_foreach(ematch, e) {
 564                 ret = find_check_match(ematch, &mtpar);
 565                 if (ret != 0)
 566                         goto cleanup_matches;
 567                 ++j;
 568         }
 569
 570         t = ipt_get_target(e);
 571         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 572                                         t->u.user.revision);
 573         if (IS_ERR(target)) {
 574                 ret = PTR_ERR(target);
 575                 goto cleanup_matches;
 576         }
 577         t->u.kernel.target = target;
 578
 579         ret = check_target(e, net, name);
 580         if (ret)
 581                 goto err;
 582
 583         return 0;
 584  err:
 585         module_put(t->u.kernel.target->me);
 586  cleanup_matches:
 587         xt_ematch_foreach(ematch, e) {
 588                 if (j-- == 0)
 589                         break;
 590                 cleanup_match(ematch, net);
 591         }
 592
 593         xt_percpu_counter_free(&e->counters);
 594
 595         return ret;
 596 }
 597
 598 static bool check_underflow(const struct ipt_entry *e)
 599 {
 600         const struct xt_entry_target *t;
 601         unsigned int verdict;
 602
 603         if (!unconditional(e))
 604                 return false;
 605         t = ipt_get_target_c(e);
 606         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 607                 return false;
 608         verdict = ((struct xt_standard_target *)t)->verdict;
 609         verdict = -verdict - 1;
 610         return verdict == NF_DROP || verdict == NF_ACCEPT;
 611 }
 612
 613 static int
 614 check_entry_size_and_hooks(struct ipt_entry *e,
 615                            struct xt_table_info *newinfo,
 616                            const unsigned char *base,
 617                            const unsigned char *limit,
 618                            const unsigned int *hook_entries,
 619                            const unsigned int *underflows,
 620                            unsigned int valid_hooks)
 621 {
 622         unsigned int h;
 623         int err;
 624
 625         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 626             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 627             (unsigned char *)e + e->next_offset > limit)
 628                 return -EINVAL;
 629
 630         if (e->next_offset
 631             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target))
 632                 return -EINVAL;
 633
 634         if (!ip_checkentry(&e->ip))
 635                 return -EINVAL;
 636
 637         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 638                                      e->next_offset);
 639         if (err)
 640                 return err;
 641
 642         /* Check hooks & underflows */
 643         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 644                 if (!(valid_hooks & (1 << h)))
 645                         continue;
 646                 if ((unsigned char *)e - base == hook_entries[h])
 647                         newinfo->hook_entry[h] = hook_entries[h];
 648                 if ((unsigned char *)e - base == underflows[h]) {
 649                         if (!check_underflow(e))
 650                                 return -EINVAL;
 651
 652                         newinfo->underflow[h] = underflows[h];
 653                 }
 654         }
 655
 656         /* Clear counters and comefrom */
 657         e->counters = ((struct xt_counters) { 0, 0 });
 658         e->comefrom = 0;
 659         return 0;
 660 }
 661
 662 static void
 663 cleanup_entry(struct ipt_entry *e, struct net *net)
 664 {
 665         struct xt_tgdtor_param par;
 666         struct xt_entry_target *t;
 667         struct xt_entry_match *ematch;
 668
 669         /* Cleanup all matches */
 670         xt_ematch_foreach(ematch, e)
 671                 cleanup_match(ematch, net);
 672         t = ipt_get_target(e);
 673
 674         par.net      = net;
 675         par.target   = t->u.kernel.target;
 676         par.targinfo = t->data;
 677         par.family   = NFPROTO_IPV4;
 678         if (par.target->destroy != NULL)
 679                 par.target->destroy(&par);
 680         module_put(par.target->me);
 681         xt_percpu_counter_free(&e->counters);
 682 }
 683
 684 /* Checks and translates the user-supplied table segment (held in
 685    newinfo) */
 686 static int
 687 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 688                 const struct ipt_replace *repl)
 689 {
 690         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 691         struct ipt_entry *iter;
 692         unsigned int *offsets;
 693         unsigned int i;
 694         int ret = 0;
 695
 696         newinfo->size = repl->size;
 697         newinfo->number = repl->num_entries;
 698
 699         /* Init all hooks to impossible value. */
 700         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 701                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 702                 newinfo->underflow[i] = 0xFFFFFFFF;
 703         }
 704
 705         offsets = xt_alloc_entry_offsets(newinfo->number);
 706         if (!offsets)
 707                 return -ENOMEM;
 708         i = 0;
 709         /* Walk through entries, checking offsets. */
 710         xt_entry_foreach(iter, entry0, newinfo->size) {
 711                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 712                                                  entry0 + repl->size,
 713                                                  repl->hook_entry,
 714                                                  repl->underflow,
 715                                                  repl->valid_hooks);
 716                 if (ret != 0)
 717                         goto out_free;
 718                 if (i < repl->num_entries)
 719                         offsets[i] = (void *)iter - entry0;
 720                 ++i;
 721                 if (strcmp(ipt_get_target(iter)->u.user.name,
 722                     XT_ERROR_TARGET) == 0)
 723                         ++newinfo->stacksize;
 724         }
 725
 726         ret = -EINVAL;
 727         if (i != repl->num_entries)
 728                 goto out_free;
 729
 730         /* Check hooks all assigned */
 731         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 732                 /* Only hooks which are valid */
 733                 if (!(repl->valid_hooks & (1 << i)))
 734                         continue;
 735                 if (newinfo->hook_entry[i] == 0xFFFFFFFF)
 736                         goto out_free;
 737                 if (newinfo->underflow[i] == 0xFFFFFFFF)
 738                         goto out_free;
 739         }
 740
 741         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 742                 ret = -ELOOP;
 743                 goto out_free;
 744         }
 745         kvfree(offsets);
 746
 747         /* Finally, each sanity check must pass */
 748         i = 0;
 749         xt_entry_foreach(iter, entry0, newinfo->size) {
 750                 ret = find_check_entry(iter, net, repl->name, repl->size,
 751                                        &alloc_state);
 752                 if (ret != 0)
 753                         break;
 754                 ++i;
 755         }
 756
 757         if (ret != 0) {
 758                 xt_entry_foreach(iter, entry0, newinfo->size) {
 759                         if (i-- == 0)
 760                                 break;
 761                         cleanup_entry(iter, net);
 762                 }
 763                 return ret;
 764         }
 765
 766         return ret;
 767  out_free:
 768         kvfree(offsets);
 769         return ret;
 770 }
 771
 772 static void
 773 get_counters(const struct xt_table_info *t,
 774              struct xt_counters counters[])
 775 {
 776         struct ipt_entry *iter;
 777         unsigned int cpu;
 778         unsigned int i;
 779
 780         for_each_possible_cpu(cpu) {
 781                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 782
 783                 i = 0;
 784                 xt_entry_foreach(iter, t->entries, t->size) {
 785                         struct xt_counters *tmp;
 786                         u64 bcnt, pcnt;
 787                         unsigned int start;
 788
 789                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 790                         do {
 791                                 start = read_seqcount_begin(s);
 792                                 bcnt = tmp->bcnt;
 793                                 pcnt = tmp->pcnt;
 794                         } while (read_seqcount_retry(s, start));
 795
 796                         ADD_COUNTER(counters[i], bcnt, pcnt);
 797                         ++i; /* macro does multi eval of i */
 798                 }
 799         }
 800 }
 801
 802 static struct xt_counters *alloc_counters(const struct xt_table *table)
 803 {
 804         unsigned int countersize;
 805         struct xt_counters *counters;
 806         const struct xt_table_info *private = table->private;
 807
 808         /* We need atomic snapshot of counters: rest doesn't change
 809            (other than comefrom, which userspace doesn't care
 810            about). */
 811         countersize = sizeof(struct xt_counters) * private->number;
 812         counters = vzalloc(countersize);
 813
 814         if (counters == NULL)
 815                 return ERR_PTR(-ENOMEM);
 816
 817         get_counters(private, counters);
 818
 819         return counters;
 820 }
 821
 822 static int
 823 copy_entries_to_user(unsigned int total_size,
 824                      const struct xt_table *table,
 825                      void __user *userptr)
 826 {
 827         unsigned int off, num;
 828         const struct ipt_entry *e;
 829         struct xt_counters *counters;
 830         const struct xt_table_info *private = table->private;
 831         int ret = 0;
 832         const void *loc_cpu_entry;
 833
 834         counters = alloc_counters(table);
 835         if (IS_ERR(counters))
 836                 return PTR_ERR(counters);
 837
 838         loc_cpu_entry = private->entries;
 839         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 840                 ret = -EFAULT;
 841                 goto free_counters;
 842         }
 843
 844         /* FIXME: use iterator macros --RR */
 845         /* ... then go back and fix counters and names */
 846         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 847                 unsigned int i;
 848                 const struct xt_entry_match *m;
 849                 const struct xt_entry_target *t;
 850
 851                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 852                 if (copy_to_user(userptr + off
 853                                  + offsetof(struct ipt_entry, counters),
 854                                  &counters[num],
 855                                  sizeof(counters[num])) != 0) {
 856                         ret = -EFAULT;
 857                         goto free_counters;
 858                 }
 859
 860                 for (i = sizeof(struct ipt_entry);
 861                      i < e->target_offset;
 862                      i += m->u.match_size) {
 863                         m = (void *)e + i;
 864
 865                         if (copy_to_user(userptr + off + i
 866                                          + offsetof(struct xt_entry_match,
 867                                                     u.user.name),
 868                                          m->u.kernel.match->name,
 869                                          strlen(m->u.kernel.match->name)+1)
 870                             != 0) {
 871                                 ret = -EFAULT;
 872                                 goto free_counters;
 873                         }
 874                 }
 875
 876                 t = ipt_get_target_c(e);
 877                 if (copy_to_user(userptr + off + e->target_offset
 878                                  + offsetof(struct xt_entry_target,
 879                                             u.user.name),
 880                                  t->u.kernel.target->name,
 881                                  strlen(t->u.kernel.target->name)+1) != 0) {
 882                         ret = -EFAULT;
 883                         goto free_counters;
 884                 }
 885         }
 886
 887  free_counters:
 888         vfree(counters);
 889         return ret;
 890 }
 891
 892 #ifdef CONFIG_COMPAT
 893 static void compat_standard_from_user(void *dst, const void *src)
 894 {
 895         int v = *(compat_int_t *)src;
 896
 897         if (v > 0)
 898                 v += xt_compat_calc_jump(AF_INET, v);
 899         memcpy(dst, &v, sizeof(v));
 900 }
 901
 902 static int compat_standard_to_user(void __user *dst, const void *src)
 903 {
 904         compat_int_t cv = *(int *)src;
 905
 906         if (cv > 0)
 907                 cv -= xt_compat_calc_jump(AF_INET, cv);
 908         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 909 }
 910
 911 static int compat_calc_entry(const struct ipt_entry *e,
 912                              const struct xt_table_info *info,
 913                              const void *base, struct xt_table_info *newinfo)
 914 {
 915         const struct xt_entry_match *ematch;
 916         const struct xt_entry_target *t;
 917         unsigned int entry_offset;
 918         int off, i, ret;
 919
 920         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
 921         entry_offset = (void *)e - base;
 922         xt_ematch_foreach(ematch, e)
 923                 off += xt_compat_match_offset(ematch->u.kernel.match);
 924         t = ipt_get_target_c(e);
 925         off += xt_compat_target_offset(t->u.kernel.target);
 926         newinfo->size -= off;
 927         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
 928         if (ret)
 929                 return ret;
 930
 931         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 932                 if (info->hook_entry[i] &&
 933                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
 934                         newinfo->hook_entry[i] -= off;
 935                 if (info->underflow[i] &&
 936                     (e < (struct ipt_entry *)(base + info->underflow[i])))
 937                         newinfo->underflow[i] -= off;
 938         }
 939         return 0;
 940 }
 941
 942 static int compat_table_info(const struct xt_table_info *info,
 943                              struct xt_table_info *newinfo)
 944 {
 945         struct ipt_entry *iter;
 946         const void *loc_cpu_entry;
 947         int ret;
 948
 949         if (!newinfo || !info)
 950                 return -EINVAL;
 951
 952         /* we dont care about newinfo->entries */
 953         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 954         newinfo->initial_entries = 0;
 955         loc_cpu_entry = info->entries;
 956         xt_compat_init_offsets(AF_INET, info->number);
 957         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 958                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 959                 if (ret != 0)
 960                         return ret;
 961         }
 962         return 0;
 963 }
 964 #endif
 965
 966 static int get_info(struct net *net, void __user *user,
 967                     const int *len, int compat)
 968 {
 969         char name[XT_TABLE_MAXNAMELEN];
 970         struct xt_table *t;
 971         int ret;
 972
 973         if (*len != sizeof(struct ipt_getinfo))
 974                 return -EINVAL;
 975
 976         if (copy_from_user(name, user, sizeof(name)) != 0)
 977                 return -EFAULT;
 978
 979         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 980 #ifdef CONFIG_COMPAT
 981         if (compat)
 982                 xt_compat_lock(AF_INET);
 983 #endif
 984         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
 985                                     "iptable_%s", name);
 986         if (!IS_ERR_OR_NULL(t)) {
 987                 struct ipt_getinfo info;
 988                 const struct xt_table_info *private = t->private;
 989 #ifdef CONFIG_COMPAT
 990                 struct xt_table_info tmp;
 991
 992                 if (compat) {
 993                         ret = compat_table_info(private, &tmp);
 994                         xt_compat_flush_offsets(AF_INET);
 995                         private = &tmp;
 996                 }
 997 #endif
 998                 memset(&info, 0, sizeof(info));
 999                 info.valid_hooks = t->valid_hooks;
1000                 memcpy(info.hook_entry, private->hook_entry,
1001                        sizeof(info.hook_entry));
1002                 memcpy(info.underflow, private->underflow,
1003                        sizeof(info.underflow));
1004                 info.num_entries = private->number;
1005                 info.size = private->size;
1006                 strcpy(info.name, name);
1007
1008                 if (copy_to_user(user, &info, *len) != 0)
1009                         ret = -EFAULT;
1010                 else
1011                         ret = 0;
1012
1013                 xt_table_unlock(t);
1014                 module_put(t->me);
1015         } else
1016                 ret = t ? PTR_ERR(t) : -ENOENT;
1017 #ifdef CONFIG_COMPAT
1018         if (compat)
1019                 xt_compat_unlock(AF_INET);
1020 #endif
1021         return ret;
1022 }
1023
1024 static int
1025 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1026             const int *len)
1027 {
1028         int ret;
1029         struct ipt_get_entries get;
1030         struct xt_table *t;
1031
1032         if (*len < sizeof(get))
1033                 return -EINVAL;
1034         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1035                 return -EFAULT;
1036         if (*len != sizeof(struct ipt_get_entries) + get.size)
1037                 return -EINVAL;
1038         get.name[sizeof(get.name) - 1] = '\0';
1039
1040         t = xt_find_table_lock(net, AF_INET, get.name);
1041         if (!IS_ERR_OR_NULL(t)) {
1042                 const struct xt_table_info *private = t->private;
1043                 if (get.size == private->size)
1044                         ret = copy_entries_to_user(private->size,
1045                                                    t, uptr->entrytable);
1046                 else
1047                         ret = -EAGAIN;
1048
1049                 module_put(t->me);
1050                 xt_table_unlock(t);
1051         } else
1052                 ret = t ? PTR_ERR(t) : -ENOENT;
1053
1054         return ret;
1055 }
1056
1057 static int
1058 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1059              struct xt_table_info *newinfo, unsigned int num_counters,
1060              void __user *counters_ptr)
1061 {
1062         int ret;
1063         struct xt_table *t;
1064         struct xt_table_info *oldinfo;
1065         struct xt_counters *counters;
1066         struct ipt_entry *iter;
1067
1068         ret = 0;
1069         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1070         if (!counters) {
1071                 ret = -ENOMEM;
1072                 goto out;
1073         }
1074
1075         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1076                                     "iptable_%s", name);
1077         if (IS_ERR_OR_NULL(t)) {
1078                 ret = t ? PTR_ERR(t) : -ENOENT;
1079                 goto free_newinfo_counters_untrans;
1080         }
1081
1082         /* You lied! */
1083         if (valid_hooks != t->valid_hooks) {
1084                 ret = -EINVAL;
1085                 goto put_module;
1086         }
1087
1088         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1089         if (!oldinfo)
1090                 goto put_module;
1091
1092         /* Update module usage count based on number of rules */
1093         if ((oldinfo->number > oldinfo->initial_entries) ||
1094             (newinfo->number <= oldinfo->initial_entries))
1095                 module_put(t->me);
1096         if ((oldinfo->number > oldinfo->initial_entries) &&
1097             (newinfo->number <= oldinfo->initial_entries))
1098                 module_put(t->me);
1099
1100         /* Get the old counters, and synchronize with replace */
1101         get_counters(oldinfo, counters);
1102
1103         /* Decrease module usage counts and free resource */
1104         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1105                 cleanup_entry(iter, net);
1106
1107         xt_free_table_info(oldinfo);
1108         if (copy_to_user(counters_ptr, counters,
1109                          sizeof(struct xt_counters) * num_counters) != 0) {
1110                 /* Silent error, can't fail, new table is already in place */
1111                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1112         }
1113         vfree(counters);
1114         xt_table_unlock(t);
1115         return ret;
1116
1117  put_module:
1118         module_put(t->me);
1119         xt_table_unlock(t);
1120  free_newinfo_counters_untrans:
1121         vfree(counters);
1122  out:
1123         return ret;
1124 }
1125
1126 static int
1127 do_replace(struct net *net, const void __user *user, unsigned int len)
1128 {
1129         int ret;
1130         struct ipt_replace tmp;
1131         struct xt_table_info *newinfo;
1132         void *loc_cpu_entry;
1133         struct ipt_entry *iter;
1134
1135         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1136                 return -EFAULT;
1137
1138         /* overflow check */
1139         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1140                 return -ENOMEM;
1141         if (tmp.num_counters == 0)
1142                 return -EINVAL;
1143
1144         tmp.name[sizeof(tmp.name)-1] = 0;
1145
1146         newinfo = xt_alloc_table_info(tmp.size);
1147         if (!newinfo)
1148                 return -ENOMEM;
1149
1150         loc_cpu_entry = newinfo->entries;
1151         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1152                            tmp.size) != 0) {
1153                 ret = -EFAULT;
1154                 goto free_newinfo;
1155         }
1156
1157         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1158         if (ret != 0)
1159                 goto free_newinfo;
1160
1161         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1162                            tmp.num_counters, tmp.counters);
1163         if (ret)
1164                 goto free_newinfo_untrans;
1165         return 0;
1166
1167  free_newinfo_untrans:
1168         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1169                 cleanup_entry(iter, net);
1170  free_newinfo:
1171         xt_free_table_info(newinfo);
1172         return ret;
1173 }
1174
1175 static int
1176 do_add_counters(struct net *net, const void __user *user,
1177                 unsigned int len, int compat)
1178 {
1179         unsigned int i;
1180         struct xt_counters_info tmp;
1181         struct xt_counters *paddc;
1182         struct xt_table *t;
1183         const struct xt_table_info *private;
1184         int ret = 0;
1185         struct ipt_entry *iter;
1186         unsigned int addend;
1187
1188         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1189         if (IS_ERR(paddc))
1190                 return PTR_ERR(paddc);
1191
1192         t = xt_find_table_lock(net, AF_INET, tmp.name);
1193         if (IS_ERR_OR_NULL(t)) {
1194                 ret = t ? PTR_ERR(t) : -ENOENT;
1195                 goto free;
1196         }
1197
1198         local_bh_disable();
1199         private = t->private;
1200         if (private->number != tmp.num_counters) {
1201                 ret = -EINVAL;
1202                 goto unlock_up_free;
1203         }
1204
1205         i = 0;
1206         addend = xt_write_recseq_begin();
1207         xt_entry_foreach(iter, private->entries, private->size) {
1208                 struct xt_counters *tmp;
1209
1210                 tmp = xt_get_this_cpu_counter(&iter->counters);
1211                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1212                 ++i;
1213         }
1214         xt_write_recseq_end(addend);
1215  unlock_up_free:
1216         local_bh_enable();
1217         xt_table_unlock(t);
1218         module_put(t->me);
1219  free:
1220         vfree(paddc);
1221
1222         return ret;
1223 }
1224
1225 #ifdef CONFIG_COMPAT
1226 struct compat_ipt_replace {
1227         char                    name[XT_TABLE_MAXNAMELEN];
1228         u32                     valid_hooks;
1229         u32                     num_entries;
1230         u32                     size;
1231         u32                     hook_entry[NF_INET_NUMHOOKS];
1232         u32                     underflow[NF_INET_NUMHOOKS];
1233         u32                     num_counters;
1234         compat_uptr_t           counters;       /* struct xt_counters * */
1235         struct compat_ipt_entry entries[0];
1236 };
1237
1238 static int
1239 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1240                           unsigned int *size, struct xt_counters *counters,
1241                           unsigned int i)
1242 {
1243         struct xt_entry_target *t;
1244         struct compat_ipt_entry __user *ce;
1245         u_int16_t target_offset, next_offset;
1246         compat_uint_t origsize;
1247         const struct xt_entry_match *ematch;
1248         int ret = 0;
1249
1250         origsize = *size;
1251         ce = (struct compat_ipt_entry __user *)*dstptr;
1252         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1253             copy_to_user(&ce->counters, &counters[i],
1254             sizeof(counters[i])) != 0)
1255                 return -EFAULT;
1256
1257         *dstptr += sizeof(struct compat_ipt_entry);
1258         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1259
1260         xt_ematch_foreach(ematch, e) {
1261                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1262                 if (ret != 0)
1263                         return ret;
1264         }
1265         target_offset = e->target_offset - (origsize - *size);
1266         t = ipt_get_target(e);
1267         ret = xt_compat_target_to_user(t, dstptr, size);
1268         if (ret)
1269                 return ret;
1270         next_offset = e->next_offset - (origsize - *size);
1271         if (put_user(target_offset, &ce->target_offset) != 0 ||
1272             put_user(next_offset, &ce->next_offset) != 0)
1273                 return -EFAULT;
1274         return 0;
1275 }
1276
1277 static int
1278 compat_find_calc_match(struct xt_entry_match *m,
1279                        const struct ipt_ip *ip,
1280                        int *size)
1281 {
1282         struct xt_match *match;
1283
1284         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1285                                       m->u.user.revision);
1286         if (IS_ERR(match))
1287                 return PTR_ERR(match);
1288
1289         m->u.kernel.match = match;
1290         *size += xt_compat_match_offset(match);
1291         return 0;
1292 }
1293
1294 static void compat_release_entry(struct compat_ipt_entry *e)
1295 {
1296         struct xt_entry_target *t;
1297         struct xt_entry_match *ematch;
1298
1299         /* Cleanup all matches */
1300         xt_ematch_foreach(ematch, e)
1301                 module_put(ematch->u.kernel.match->me);
1302         t = compat_ipt_get_target(e);
1303         module_put(t->u.kernel.target->me);
1304 }
1305
1306 static int
1307 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1308                                   struct xt_table_info *newinfo,
1309                                   unsigned int *size,
1310                                   const unsigned char *base,
1311                                   const unsigned char *limit)
1312 {
1313         struct xt_entry_match *ematch;
1314         struct xt_entry_target *t;
1315         struct xt_target *target;
1316         unsigned int entry_offset;
1317         unsigned int j;
1318         int ret, off;
1319
1320         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1321             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1322             (unsigned char *)e + e->next_offset > limit)
1323                 return -EINVAL;
1324
1325         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1326                              sizeof(struct compat_xt_entry_target))
1327                 return -EINVAL;
1328
1329         if (!ip_checkentry(&e->ip))
1330                 return -EINVAL;
1331
1332         ret = xt_compat_check_entry_offsets(e, e->elems,
1333                                             e->target_offset, e->next_offset);
1334         if (ret)
1335                 return ret;
1336
1337         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1338         entry_offset = (void *)e - (void *)base;
1339         j = 0;
1340         xt_ematch_foreach(ematch, e) {
1341                 ret = compat_find_calc_match(ematch, &e->ip, &off);
1342                 if (ret != 0)
1343                         goto release_matches;
1344                 ++j;
1345         }
1346
1347         t = compat_ipt_get_target(e);
1348         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1349                                         t->u.user.revision);
1350         if (IS_ERR(target)) {
1351                 ret = PTR_ERR(target);
1352                 goto release_matches;
1353         }
1354         t->u.kernel.target = target;
1355
1356         off += xt_compat_target_offset(target);
1357         *size += off;
1358         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1359         if (ret)
1360                 goto out;
1361
1362         return 0;
1363
1364 out:
1365         module_put(t->u.kernel.target->me);
1366 release_matches:
1367         xt_ematch_foreach(ematch, e) {
1368                 if (j-- == 0)
1369                         break;
1370                 module_put(ematch->u.kernel.match->me);
1371         }
1372         return ret;
1373 }
1374
1375 static void
1376 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1377                             unsigned int *size,
1378                             struct xt_table_info *newinfo, unsigned char *base)
1379 {
1380         struct xt_entry_target *t;
1381         struct xt_target *target;
1382         struct ipt_entry *de;
1383         unsigned int origsize;
1384         int h;
1385         struct xt_entry_match *ematch;
1386
1387         origsize = *size;
1388         de = (struct ipt_entry *)*dstptr;
1389         memcpy(de, e, sizeof(struct ipt_entry));
1390         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1391
1392         *dstptr += sizeof(struct ipt_entry);
1393         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1394
1395         xt_ematch_foreach(ematch, e)
1396                 xt_compat_match_from_user(ematch, dstptr, size);
1397
1398         de->target_offset = e->target_offset - (origsize - *size);
1399         t = compat_ipt_get_target(e);
1400         target = t->u.kernel.target;
1401         xt_compat_target_from_user(t, dstptr, size);
1402
1403         de->next_offset = e->next_offset - (origsize - *size);
1404
1405         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1406                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1407                         newinfo->hook_entry[h] -= origsize - *size;
1408                 if ((unsigned char *)de - base < newinfo->underflow[h])
1409                         newinfo->underflow[h] -= origsize - *size;
1410         }
1411 }
1412
1413 static int
1414 translate_compat_table(struct net *net,
1415                        struct xt_table_info **pinfo,
1416                        void **pentry0,
1417                        const struct compat_ipt_replace *compatr)
1418 {
1419         unsigned int i, j;
1420         struct xt_table_info *newinfo, *info;
1421         void *pos, *entry0, *entry1;
1422         struct compat_ipt_entry *iter0;
1423         struct ipt_replace repl;
1424         unsigned int size;
1425         int ret;
1426
1427         info = *pinfo;
1428         entry0 = *pentry0;
1429         size = compatr->size;
1430         info->number = compatr->num_entries;
1431
1432         j = 0;
1433         xt_compat_lock(AF_INET);
1434         xt_compat_init_offsets(AF_INET, compatr->num_entries);
1435         /* Walk through entries, checking offsets. */
1436         xt_entry_foreach(iter0, entry0, compatr->size) {
1437                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1438                                                         entry0,
1439                                                         entry0 + compatr->size);
1440                 if (ret != 0)
1441                         goto out_unlock;
1442                 ++j;
1443         }
1444
1445         ret = -EINVAL;
1446         if (j != compatr->num_entries)
1447                 goto out_unlock;
1448
1449         ret = -ENOMEM;
1450         newinfo = xt_alloc_table_info(size);
1451         if (!newinfo)
1452                 goto out_unlock;
1453
1454         memset(newinfo->entries, 0, size);
1455
1456         newinfo->number = compatr->num_entries;
1457         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1458                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1459                 newinfo->underflow[i] = compatr->underflow[i];
1460         }
1461         entry1 = newinfo->entries;
1462         pos = entry1;
1463         size = compatr->size;
1464         xt_entry_foreach(iter0, entry0, compatr->size)
1465                 compat_copy_entry_from_user(iter0, &pos, &size,
1466                                             newinfo, entry1);
1467
1468         /* all module references in entry0 are now gone.
1469          * entry1/newinfo contains a 64bit ruleset that looks exactly as
1470          * generated by 64bit userspace.
1471          *
1472          * Call standard translate_table() to validate all hook_entrys,
1473          * underflows, check for loops, etc.
1474          */
1475         xt_compat_flush_offsets(AF_INET);
1476         xt_compat_unlock(AF_INET);
1477
1478         memcpy(&repl, compatr, sizeof(*compatr));
1479
1480         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1481                 repl.hook_entry[i] = newinfo->hook_entry[i];
1482                 repl.underflow[i] = newinfo->underflow[i];
1483         }
1484
1485         repl.num_counters = 0;
1486         repl.counters = NULL;
1487         repl.size = newinfo->size;
1488         ret = translate_table(net, newinfo, entry1, &repl);
1489         if (ret)
1490                 goto free_newinfo;
1491
1492         *pinfo = newinfo;
1493         *pentry0 = entry1;
1494         xt_free_table_info(info);
1495         return 0;
1496
1497 free_newinfo:
1498         xt_free_table_info(newinfo);
1499         return ret;
1500 out_unlock:
1501         xt_compat_flush_offsets(AF_INET);
1502         xt_compat_unlock(AF_INET);
1503         xt_entry_foreach(iter0, entry0, compatr->size) {
1504                 if (j-- == 0)
1505                         break;
1506                 compat_release_entry(iter0);
1507         }
1508         return ret;
1509 }
1510
1511 static int
1512 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1513 {
1514         int ret;
1515         struct compat_ipt_replace tmp;
1516         struct xt_table_info *newinfo;
1517         void *loc_cpu_entry;
1518         struct ipt_entry *iter;
1519
1520         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1521                 return -EFAULT;
1522
1523         /* overflow check */
1524         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1525                 return -ENOMEM;
1526         if (tmp.num_counters == 0)
1527                 return -EINVAL;
1528
1529         tmp.name[sizeof(tmp.name)-1] = 0;
1530
1531         newinfo = xt_alloc_table_info(tmp.size);
1532         if (!newinfo)
1533                 return -ENOMEM;
1534
1535         loc_cpu_entry = newinfo->entries;
1536         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1537                            tmp.size) != 0) {
1538                 ret = -EFAULT;
1539                 goto free_newinfo;
1540         }
1541
1542         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1543         if (ret != 0)
1544                 goto free_newinfo;
1545
1546         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1547                            tmp.num_counters, compat_ptr(tmp.counters));
1548         if (ret)
1549                 goto free_newinfo_untrans;
1550         return 0;
1551
1552  free_newinfo_untrans:
1553         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1554                 cleanup_entry(iter, net);
1555  free_newinfo:
1556         xt_free_table_info(newinfo);
1557         return ret;
1558 }
1559
1560 static int
1561 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1562                       unsigned int len)
1563 {
1564         int ret;
1565
1566         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1567                 return -EPERM;
1568
1569         switch (cmd) {
1570         case IPT_SO_SET_REPLACE:
1571                 ret = compat_do_replace(sock_net(sk), user, len);
1572                 break;
1573
1574         case IPT_SO_SET_ADD_COUNTERS:
1575                 ret = do_add_counters(sock_net(sk), user, len, 1);
1576                 break;
1577
1578         default:
1579                 ret = -EINVAL;
1580         }
1581
1582         return ret;
1583 }
1584
1585 struct compat_ipt_get_entries {
1586         char name[XT_TABLE_MAXNAMELEN];
1587         compat_uint_t size;
1588         struct compat_ipt_entry entrytable[0];
1589 };
1590
1591 static int
1592 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1593                             void __user *userptr)
1594 {
1595         struct xt_counters *counters;
1596         const struct xt_table_info *private = table->private;
1597         void __user *pos;
1598         unsigned int size;
1599         int ret = 0;
1600         unsigned int i = 0;
1601         struct ipt_entry *iter;
1602
1603         counters = alloc_counters(table);
1604         if (IS_ERR(counters))
1605                 return PTR_ERR(counters);
1606
1607         pos = userptr;
1608         size = total_size;
1609         xt_entry_foreach(iter, private->entries, total_size) {
1610                 ret = compat_copy_entry_to_user(iter, &pos,
1611                                                 &size, counters, i++);
1612                 if (ret != 0)
1613                         break;
1614         }
1615
1616         vfree(counters);
1617         return ret;
1618 }
1619
1620 static int
1621 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1622                    int *len)
1623 {
1624         int ret;
1625         struct compat_ipt_get_entries get;
1626         struct xt_table *t;
1627
1628         if (*len < sizeof(get))
1629                 return -EINVAL;
1630
1631         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1632                 return -EFAULT;
1633
1634         if (*len != sizeof(struct compat_ipt_get_entries) + get.size)
1635                 return -EINVAL;
1636
1637         get.name[sizeof(get.name) - 1] = '\0';
1638
1639         xt_compat_lock(AF_INET);
1640         t = xt_find_table_lock(net, AF_INET, get.name);
1641         if (!IS_ERR_OR_NULL(t)) {
1642                 const struct xt_table_info *private = t->private;
1643                 struct xt_table_info info;
1644                 ret = compat_table_info(private, &info);
1645                 if (!ret && get.size == info.size)
1646                         ret = compat_copy_entries_to_user(private->size,
1647                                                           t, uptr->entrytable);
1648                 else if (!ret)
1649                         ret = -EAGAIN;
1650
1651                 xt_compat_flush_offsets(AF_INET);
1652                 module_put(t->me);
1653                 xt_table_unlock(t);
1654         } else
1655                 ret = t ? PTR_ERR(t) : -ENOENT;
1656
1657         xt_compat_unlock(AF_INET);
1658         return ret;
1659 }
1660
1661 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1662
1663 static int
1664 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1665 {
1666         int ret;
1667
1668         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1669                 return -EPERM;
1670
1671         switch (cmd) {
1672         case IPT_SO_GET_INFO:
1673                 ret = get_info(sock_net(sk), user, len, 1);
1674                 break;
1675         case IPT_SO_GET_ENTRIES:
1676                 ret = compat_get_entries(sock_net(sk), user, len);
1677                 break;
1678         default:
1679                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1680         }
1681         return ret;
1682 }
1683 #endif
1684
1685 static int
1686 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1687 {
1688         int ret;
1689
1690         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1691                 return -EPERM;
1692
1693         switch (cmd) {
1694         case IPT_SO_SET_REPLACE:
1695                 ret = do_replace(sock_net(sk), user, len);
1696                 break;
1697
1698         case IPT_SO_SET_ADD_COUNTERS:
1699                 ret = do_add_counters(sock_net(sk), user, len, 0);
1700                 break;
1701
1702         default:
1703                 ret = -EINVAL;
1704         }
1705
1706         return ret;
1707 }
1708
1709 static int
1710 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1711 {
1712         int ret;
1713
1714         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1715                 return -EPERM;
1716
1717         switch (cmd) {
1718         case IPT_SO_GET_INFO:
1719                 ret = get_info(sock_net(sk), user, len, 0);
1720                 break;
1721
1722         case IPT_SO_GET_ENTRIES:
1723                 ret = get_entries(sock_net(sk), user, len);
1724                 break;
1725
1726         case IPT_SO_GET_REVISION_MATCH:
1727         case IPT_SO_GET_REVISION_TARGET: {
1728                 struct xt_get_revision rev;
1729                 int target;
1730
1731                 if (*len != sizeof(rev)) {
1732                         ret = -EINVAL;
1733                         break;
1734                 }
1735                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1736                         ret = -EFAULT;
1737                         break;
1738                 }
1739                 rev.name[sizeof(rev.name)-1] = 0;
1740
1741                 if (cmd == IPT_SO_GET_REVISION_TARGET)
1742                         target = 1;
1743                 else
1744                         target = 0;
1745
1746                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
1747                                                          rev.revision,
1748                                                          target, &ret),
1749                                         "ipt_%s", rev.name);
1750                 break;
1751         }
1752
1753         default:
1754                 ret = -EINVAL;
1755         }
1756
1757         return ret;
1758 }
1759
1760 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
1761 {
1762         struct xt_table_info *private;
1763         void *loc_cpu_entry;
1764         struct module *table_owner = table->me;
1765         struct ipt_entry *iter;
1766
1767         private = xt_unregister_table(table);
1768
1769         /* Decrease module usage counts and free resources */
1770         loc_cpu_entry = private->entries;
1771         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1772                 cleanup_entry(iter, net);
1773         if (private->number > private->initial_entries)
1774                 module_put(table_owner);
1775         xt_free_table_info(private);
1776 }
1777
1778 int ipt_register_table(struct net *net, const struct xt_table *table,
1779                        const struct ipt_replace *repl,
1780                        const struct nf_hook_ops *ops, struct xt_table **res)
1781 {
1782         int ret;
1783         struct xt_table_info *newinfo;
1784         struct xt_table_info bootstrap = {0};
1785         void *loc_cpu_entry;
1786         struct xt_table *new_table;
1787
1788         newinfo = xt_alloc_table_info(repl->size);
1789         if (!newinfo)
1790                 return -ENOMEM;
1791
1792         loc_cpu_entry = newinfo->entries;
1793         memcpy(loc_cpu_entry, repl->entries, repl->size);
1794
1795         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1796         if (ret != 0)
1797                 goto out_free;
1798
1799         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1800         if (IS_ERR(new_table)) {
1801                 ret = PTR_ERR(new_table);
1802                 goto out_free;
1803         }
1804
1805         /* set res now, will see skbs right after nf_register_net_hooks */
1806         WRITE_ONCE(*res, new_table);
1807
1808         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1809         if (ret != 0) {
1810                 __ipt_unregister_table(net, new_table);
1811                 *res = NULL;
1812         }
1813
1814         return ret;
1815
1816 out_free:
1817         xt_free_table_info(newinfo);
1818         return ret;
1819 }
1820
1821 void ipt_unregister_table(struct net *net, struct xt_table *table,
1822                           const struct nf_hook_ops *ops)
1823 {
1824         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1825         __ipt_unregister_table(net, table);
1826 }
1827
1828 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1829 static inline bool
1830 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1831                      u_int8_t type, u_int8_t code,
1832                      bool invert)
1833 {
1834         return ((test_type == 0xFF) ||
1835                 (type == test_type && code >= min_code && code <= max_code))
1836                 ^ invert;
1837 }
1838
1839 static bool
1840 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
1841 {
1842         const struct icmphdr *ic;
1843         struct icmphdr _icmph;
1844         const struct ipt_icmp *icmpinfo = par->matchinfo;
1845
1846         /* Must not be a fragment. */
1847         if (par->fragoff != 0)
1848                 return false;
1849
1850         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1851         if (ic == NULL) {
1852                 /* We've been asked to examine this packet, and we
1853                  * can't.  Hence, no choice but to drop.
1854                  */
1855                 par->hotdrop = true;
1856                 return false;
1857         }
1858
1859         return icmp_type_code_match(icmpinfo->type,
1860                                     icmpinfo->code[0],
1861                                     icmpinfo->code[1],
1862                                     ic->type, ic->code,
1863                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
1864 }
1865
1866 static int icmp_checkentry(const struct xt_mtchk_param *par)
1867 {
1868         const struct ipt_icmp *icmpinfo = par->matchinfo;
1869
1870         /* Must specify no unknown invflags */
1871         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
1872 }
1873
1874 static struct xt_target ipt_builtin_tg[] __read_mostly = {
1875         {
1876                 .name             = XT_STANDARD_TARGET,
1877                 .targetsize       = sizeof(int),
1878                 .family           = NFPROTO_IPV4,
1879 #ifdef CONFIG_COMPAT
1880                 .compatsize       = sizeof(compat_int_t),
1881                 .compat_from_user = compat_standard_from_user,
1882                 .compat_to_user   = compat_standard_to_user,
1883 #endif
1884         },
1885         {
1886                 .name             = XT_ERROR_TARGET,
1887                 .target           = ipt_error,
1888                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1889                 .family           = NFPROTO_IPV4,
1890         },
1891 };
1892
1893 static struct nf_sockopt_ops ipt_sockopts = {
1894         .pf             = PF_INET,
1895         .set_optmin     = IPT_BASE_CTL,
1896         .set_optmax     = IPT_SO_SET_MAX+1,
1897         .set            = do_ipt_set_ctl,
1898 #ifdef CONFIG_COMPAT
1899         .compat_set     = compat_do_ipt_set_ctl,
1900 #endif
1901         .get_optmin     = IPT_BASE_CTL,
1902         .get_optmax     = IPT_SO_GET_MAX+1,
1903         .get            = do_ipt_get_ctl,
1904 #ifdef CONFIG_COMPAT
1905         .compat_get     = compat_do_ipt_get_ctl,
1906 #endif
1907         .owner          = THIS_MODULE,
1908 };
1909
1910 static struct xt_match ipt_builtin_mt[] __read_mostly = {
1911         {
1912                 .name       = "icmp",
1913                 .match      = icmp_match,
1914                 .matchsize  = sizeof(struct ipt_icmp),
1915                 .checkentry = icmp_checkentry,
1916                 .proto      = IPPROTO_ICMP,
1917                 .family     = NFPROTO_IPV4,
1918                 .me         = THIS_MODULE,
1919         },
1920 };
1921
1922 static int __net_init ip_tables_net_init(struct net *net)
1923 {
1924         return xt_proto_init(net, NFPROTO_IPV4);
1925 }
1926
1927 static void __net_exit ip_tables_net_exit(struct net *net)
1928 {
1929         xt_proto_fini(net, NFPROTO_IPV4);
1930 }
1931
1932 static struct pernet_operations ip_tables_net_ops = {
1933         .init = ip_tables_net_init,
1934         .exit = ip_tables_net_exit,
1935 };
1936
1937 static int __init ip_tables_init(void)
1938 {
1939         int ret;
1940
1941         ret = register_pernet_subsys(&ip_tables_net_ops);
1942         if (ret < 0)
1943                 goto err1;
1944
1945         /* No one else will be downing sem now, so we won't sleep */
1946         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1947         if (ret < 0)
1948                 goto err2;
1949         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1950         if (ret < 0)
1951                 goto err4;
1952
1953         /* Register setsockopt */
1954         ret = nf_register_sockopt(&ipt_sockopts);
1955         if (ret < 0)
1956                 goto err5;
1957
1958         pr_info("(C) 2000-2006 Netfilter Core Team\n");
1959         return 0;
1960
1961 err5:
1962         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1963 err4:
1964         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1965 err2:
1966         unregister_pernet_subsys(&ip_tables_net_ops);
1967 err1:
1968         return ret;
1969 }
1970
1971 static void __exit ip_tables_fini(void)
1972 {
1973         nf_unregister_sockopt(&ipt_sockopts);
1974
1975         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
1976         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
1977         unregister_pernet_subsys(&ip_tables_net_ops);
1978 }
1979
1980 EXPORT_SYMBOL(ipt_register_table);
1981 EXPORT_SYMBOL(ipt_unregister_table);
1982 EXPORT_SYMBOL(ipt_do_table);
1983 module_init(ip_tables_init);
1984 module_exit(ip_tables_fini);