net/ipv6/netfilter/ip6_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
  13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  14
  15 #include <linux/kernel.h>
  16 #include <linux/capability.h>
  17 #include <linux/in.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/kmod.h>
  20 #include <linux/vmalloc.h>
  21 #include <linux/netdevice.h>
  22 #include <linux/module.h>
  23 #include <linux/poison.h>
  24 #include <linux/icmpv6.h>
  25 #include <net/ipv6.h>
  26 #include <net/compat.h>
  27 #include <linux/uaccess.h>
  28 #include <linux/mutex.h>
  29 #include <linux/proc_fs.h>
  30 #include <linux/err.h>
  31 #include <linux/cpumask.h>
  32
  33 #include <linux/netfilter_ipv6/ip6_tables.h>
  34 #include <linux/netfilter/x_tables.h>
  35 #include <net/netfilter/nf_log.h>
  36 #include "../../netfilter/xt_repldata.h"
  37
  38 MODULE_LICENSE("GPL");
  39 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  40 MODULE_DESCRIPTION("IPv6 packet filter");
  41 MODULE_ALIAS("ip6t_icmp6");
  42
  43 void *ip6t_alloc_initial_table(const struct xt_table *info)
  44 {
  45         return xt_alloc_initial_table(ip6t, IP6T);
  46 }
  47 EXPORT_SYMBOL_GPL(ip6t_alloc_initial_table);
  48
  49 /* Returns whether matches rule or not. */
  50 /* Performance critical - called for every packet */
  51 static inline bool
  52 ip6_packet_match(const struct sk_buff *skb,
  53                  const char *indev,
  54                  const char *outdev,
  55                  const struct ip6t_ip6 *ip6info,
  56                  unsigned int *protoff,
  57                  int *fragoff, bool *hotdrop)
  58 {
  59         unsigned long ret;
  60         const struct ipv6hdr *ipv6 = ipv6_hdr(skb);
  61
  62         if (NF_INVF(ip6info, IP6T_INV_SRCIP,
  63                     ipv6_masked_addr_cmp(&ipv6->saddr, &ip6info->smsk,
  64                                          &ip6info->src)) ||
  65             NF_INVF(ip6info, IP6T_INV_DSTIP,
  66                     ipv6_masked_addr_cmp(&ipv6->daddr, &ip6info->dmsk,
  67                                          &ip6info->dst)))
  68                 return false;
  69
  70         ret = ifname_compare_aligned(indev, ip6info->iniface, ip6info->iniface_mask);
  71
  72         if (NF_INVF(ip6info, IP6T_INV_VIA_IN, ret != 0))
  73                 return false;
  74
  75         ret = ifname_compare_aligned(outdev, ip6info->outiface, ip6info->outiface_mask);
  76
  77         if (NF_INVF(ip6info, IP6T_INV_VIA_OUT, ret != 0))
  78                 return false;
  79
  80 /* ... might want to do something with class and flowlabel here ... */
  81
  82         /* look for the desired protocol header */
  83         if (ip6info->flags & IP6T_F_PROTO) {
  84                 int protohdr;
  85                 unsigned short _frag_off;
  86
  87                 protohdr = ipv6_find_hdr(skb, protoff, -1, &_frag_off, NULL);
  88                 if (protohdr < 0) {
  89                         if (_frag_off == 0)
  90                                 *hotdrop = true;
  91                         return false;
  92                 }
  93                 *fragoff = _frag_off;
  94
  95                 if (ip6info->proto == protohdr) {
  96                         if (ip6info->invflags & IP6T_INV_PROTO)
  97                                 return false;
  98
  99                         return true;
 100                 }
 101
 102                 /* We need match for the '-p all', too! */
 103                 if ((ip6info->proto != 0) &&
 104                         !(ip6info->invflags & IP6T_INV_PROTO))
 105                         return false;
 106         }
 107         return true;
 108 }
 109
 110 /* should be ip6 safe */
 111 static bool
 112 ip6_checkentry(const struct ip6t_ip6 *ipv6)
 113 {
 114         if (ipv6->flags & ~IP6T_F_MASK)
 115                 return false;
 116         if (ipv6->invflags & ~IP6T_INV_MASK)
 117                 return false;
 118
 119         return true;
 120 }
 121
 122 static unsigned int
 123 ip6t_error(struct sk_buff *skb, const struct xt_action_param *par)
 124 {
 125         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 126
 127         return NF_DROP;
 128 }
 129
 130 static inline struct ip6t_entry *
 131 get_entry(const void *base, unsigned int offset)
 132 {
 133         return (struct ip6t_entry *)(base + offset);
 134 }
 135
 136 /* All zeroes == unconditional rule. */
 137 /* Mildly perf critical (only if packet tracing is on) */
 138 static inline bool unconditional(const struct ip6t_entry *e)
 139 {
 140         static const struct ip6t_ip6 uncond;
 141
 142         return e->target_offset == sizeof(struct ip6t_entry) &&
 143                memcmp(&e->ipv6, &uncond, sizeof(uncond)) == 0;
 144 }
 145
 146 static inline const struct xt_entry_target *
 147 ip6t_get_target_c(const struct ip6t_entry *e)
 148 {
 149         return ip6t_get_target((struct ip6t_entry *)e);
 150 }
 151
 152 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 153 /* This cries for unification! */
 154 static const char *const hooknames[] = {
 155         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 156         [NF_INET_LOCAL_IN]              = "INPUT",
 157         [NF_INET_FORWARD]               = "FORWARD",
 158         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 159         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 160 };
 161
 162 enum nf_ip_trace_comments {
 163         NF_IP6_TRACE_COMMENT_RULE,
 164         NF_IP6_TRACE_COMMENT_RETURN,
 165         NF_IP6_TRACE_COMMENT_POLICY,
 166 };
 167
 168 static const char *const comments[] = {
 169         [NF_IP6_TRACE_COMMENT_RULE]     = "rule",
 170         [NF_IP6_TRACE_COMMENT_RETURN]   = "return",
 171         [NF_IP6_TRACE_COMMENT_POLICY]   = "policy",
 172 };
 173
 174 static const struct nf_loginfo trace_loginfo = {
 175         .type = NF_LOG_TYPE_LOG,
 176         .u = {
 177                 .log = {
 178                         .level = LOGLEVEL_WARNING,
 179                         .logflags = NF_LOG_DEFAULT_MASK,
 180                 },
 181         },
 182 };
 183
 184 /* Mildly perf critical (only if packet tracing is on) */
 185 static inline int
 186 get_chainname_rulenum(const struct ip6t_entry *s, const struct ip6t_entry *e,
 187                       const char *hookname, const char **chainname,
 188                       const char **comment, unsigned int *rulenum)
 189 {
 190         const struct xt_standard_target *t = (void *)ip6t_get_target_c(s);
 191
 192         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 193                 /* Head of user chain: ERROR target with chainname */
 194                 *chainname = t->target.data;
 195                 (*rulenum) = 0;
 196         } else if (s == e) {
 197                 (*rulenum)++;
 198
 199                 if (unconditional(s) &&
 200                     strcmp(t->target.u.kernel.target->name,
 201                            XT_STANDARD_TARGET) == 0 &&
 202                     t->verdict < 0) {
 203                         /* Tail of chains: STANDARD target (return/policy) */
 204                         *comment = *chainname == hookname
 205                                 ? comments[NF_IP6_TRACE_COMMENT_POLICY]
 206                                 : comments[NF_IP6_TRACE_COMMENT_RETURN];
 207                 }
 208                 return 1;
 209         } else
 210                 (*rulenum)++;
 211
 212         return 0;
 213 }
 214
 215 static void trace_packet(struct net *net,
 216                          const struct sk_buff *skb,
 217                          unsigned int hook,
 218                          const struct net_device *in,
 219                          const struct net_device *out,
 220                          const char *tablename,
 221                          const struct xt_table_info *private,
 222                          const struct ip6t_entry *e)
 223 {
 224         const struct ip6t_entry *root;
 225         const char *hookname, *chainname, *comment;
 226         const struct ip6t_entry *iter;
 227         unsigned int rulenum = 0;
 228
 229         root = get_entry(private->entries, private->hook_entry[hook]);
 230
 231         hookname = chainname = hooknames[hook];
 232         comment = comments[NF_IP6_TRACE_COMMENT_RULE];
 233
 234         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 235                 if (get_chainname_rulenum(iter, e, hookname,
 236                     &chainname, &comment, &rulenum) != 0)
 237                         break;
 238
 239         nf_log_trace(net, AF_INET6, hook, skb, in, out, &trace_loginfo,
 240                      "TRACE: %s:%s:%s:%u ",
 241                      tablename, chainname, comment, rulenum);
 242 }
 243 #endif
 244
 245 static inline struct ip6t_entry *
 246 ip6t_next_entry(const struct ip6t_entry *entry)
 247 {
 248         return (void *)entry + entry->next_offset;
 249 }
 250
 251 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 252 unsigned int
 253 ip6t_do_table(struct sk_buff *skb,
 254               const struct nf_hook_state *state,
 255               struct xt_table *table)
 256 {
 257         unsigned int hook = state->hook;
 258         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 259         /* Initializing verdict to NF_DROP keeps gcc happy. */
 260         unsigned int verdict = NF_DROP;
 261         const char *indev, *outdev;
 262         const void *table_base;
 263         struct ip6t_entry *e, **jumpstack;
 264         unsigned int stackidx, cpu;
 265         const struct xt_table_info *private;
 266         struct xt_action_param acpar;
 267         unsigned int addend;
 268
 269         /* Initialization */
 270         stackidx = 0;
 271         indev = state->in ? state->in->name : nulldevname;
 272         outdev = state->out ? state->out->name : nulldevname;
 273         /* We handle fragments by dealing with the first fragment as
 274          * if it was a normal packet.  All other fragments are treated
 275          * normally, except that they will NEVER match rules that ask
 276          * things we don't know, ie. tcp syn flag or ports).  If the
 277          * rule is also a fragment-specific rule, non-fragments won't
 278          * match it. */
 279         acpar.hotdrop = false;
 280         acpar.state   = state;
 281
 282         WARN_ON(!(table->valid_hooks & (1 << hook)));
 283
 284         local_bh_disable();
 285         addend = xt_write_recseq_begin();
 286         private = READ_ONCE(table->private); /* Address dependency. */
 287         cpu        = smp_processor_id();
 288         table_base = private->entries;
 289         jumpstack  = (struct ip6t_entry **)private->jumpstack[cpu];
 290
 291         /* Switch to alternate jumpstack if we're being invoked via TEE.
 292          * TEE issues XT_CONTINUE verdict on original skb so we must not
 293          * clobber the jumpstack.
 294          *
 295          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 296          * but it is no problem since absolute verdict is issued by these.
 297          */
 298         if (static_key_false(&xt_tee_enabled))
 299                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 300
 301         e = get_entry(table_base, private->hook_entry[hook]);
 302
 303         do {
 304                 const struct xt_entry_target *t;
 305                 const struct xt_entry_match *ematch;
 306                 struct xt_counters *counter;
 307
 308                 WARN_ON(!e);
 309                 acpar.thoff = 0;
 310                 if (!ip6_packet_match(skb, indev, outdev, &e->ipv6,
 311                     &acpar.thoff, &acpar.fragoff, &acpar.hotdrop)) {
 312  no_match:
 313                         e = ip6t_next_entry(e);
 314                         continue;
 315                 }
 316
 317                 xt_ematch_foreach(ematch, e) {
 318                         acpar.match     = ematch->u.kernel.match;
 319                         acpar.matchinfo = ematch->data;
 320                         if (!acpar.match->match(skb, &acpar))
 321                                 goto no_match;
 322                 }
 323
 324                 counter = xt_get_this_cpu_counter(&e->counters);
 325                 ADD_COUNTER(*counter, skb->len, 1);
 326
 327                 t = ip6t_get_target_c(e);
 328                 WARN_ON(!t->u.kernel.target);
 329
 330 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 331                 /* The packet is traced: log it */
 332                 if (unlikely(skb->nf_trace))
 333                         trace_packet(state->net, skb, hook, state->in,
 334                                      state->out, table->name, private, e);
 335 #endif
 336                 /* Standard target? */
 337                 if (!t->u.kernel.target->target) {
 338                         int v;
 339
 340                         v = ((struct xt_standard_target *)t)->verdict;
 341                         if (v < 0) {
 342                                 /* Pop from stack? */
 343                                 if (v != XT_RETURN) {
 344                                         verdict = (unsigned int)(-v) - 1;
 345                                         break;
 346                                 }
 347                                 if (stackidx == 0)
 348                                         e = get_entry(table_base,
 349                                             private->underflow[hook]);
 350                                 else
 351                                         e = ip6t_next_entry(jumpstack[--stackidx]);
 352                                 continue;
 353                         }
 354                         if (table_base + v != ip6t_next_entry(e) &&
 355                             !(e->ipv6.flags & IP6T_F_GOTO)) {
 356                                 if (unlikely(stackidx >= private->stacksize)) {
 357                                         verdict = NF_DROP;
 358                                         break;
 359                                 }
 360                                 jumpstack[stackidx++] = e;
 361                         }
 362
 363                         e = get_entry(table_base, v);
 364                         continue;
 365                 }
 366
 367                 acpar.target   = t->u.kernel.target;
 368                 acpar.targinfo = t->data;
 369
 370                 verdict = t->u.kernel.target->target(skb, &acpar);
 371                 if (verdict == XT_CONTINUE)
 372                         e = ip6t_next_entry(e);
 373                 else
 374                         /* Verdict */
 375                         break;
 376         } while (!acpar.hotdrop);
 377
 378         xt_write_recseq_end(addend);
 379         local_bh_enable();
 380
 381         if (acpar.hotdrop)
 382                 return NF_DROP;
 383         else return verdict;
 384 }
 385
 386 /* Figures out from what hook each rule can be called: returns 0 if
 387    there are loops.  Puts hook bitmask in comefrom. */
 388 static int
 389 mark_source_chains(const struct xt_table_info *newinfo,
 390                    unsigned int valid_hooks, void *entry0,
 391                    unsigned int *offsets)
 392 {
 393         unsigned int hook;
 394
 395         /* No recursion; use packet counter to save back ptrs (reset
 396            to 0 as we leave), and comefrom to save source hook bitmask */
 397         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 398                 unsigned int pos = newinfo->hook_entry[hook];
 399                 struct ip6t_entry *e = entry0 + pos;
 400
 401                 if (!(valid_hooks & (1 << hook)))
 402                         continue;
 403
 404                 /* Set initial back pointer. */
 405                 e->counters.pcnt = pos;
 406
 407                 for (;;) {
 408                         const struct xt_standard_target *t
 409                                 = (void *)ip6t_get_target_c(e);
 410                         int visited = e->comefrom & (1 << hook);
 411
 412                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 413                                 return 0;
 414
 415                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 416
 417                         /* Unconditional return/END. */
 418                         if ((unconditional(e) &&
 419                              (strcmp(t->target.u.user.name,
 420                                      XT_STANDARD_TARGET) == 0) &&
 421                              t->verdict < 0) || visited) {
 422                                 unsigned int oldpos, size;
 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 = entry0 + pos;
 437                                 } while (oldpos == pos + e->next_offset);
 438
 439                                 /* Move along one */
 440                                 size = e->next_offset;
 441                                 e = entry0 + pos + size;
 442                                 if (pos + size >= newinfo->size)
 443                                         return 0;
 444                                 e->counters.pcnt = pos;
 445                                 pos += size;
 446                         } else {
 447                                 int newpos = t->verdict;
 448
 449                                 if (strcmp(t->target.u.user.name,
 450                                            XT_STANDARD_TARGET) == 0 &&
 451                                     newpos >= 0) {
 452                                         /* This a jump; chase it. */
 453                                         if (!xt_find_jump_offset(offsets, newpos,
 454                                                                  newinfo->number))
 455                                                 return 0;
 456                                 } else {
 457                                         /* ... this is a fallthru */
 458                                         newpos = pos + e->next_offset;
 459                                         if (newpos >= newinfo->size)
 460                                                 return 0;
 461                                 }
 462                                 e = entry0 + newpos;
 463                                 e->counters.pcnt = pos;
 464                                 pos = newpos;
 465                         }
 466                 }
 467 next:           ;
 468         }
 469         return 1;
 470 }
 471
 472 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 473 {
 474         struct xt_mtdtor_param par;
 475
 476         par.net       = net;
 477         par.match     = m->u.kernel.match;
 478         par.matchinfo = m->data;
 479         par.family    = NFPROTO_IPV6;
 480         if (par.match->destroy != NULL)
 481                 par.match->destroy(&par);
 482         module_put(par.match->me);
 483 }
 484
 485 static int check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 486 {
 487         const struct ip6t_ip6 *ipv6 = par->entryinfo;
 488
 489         par->match     = m->u.kernel.match;
 490         par->matchinfo = m->data;
 491
 492         return xt_check_match(par, m->u.match_size - sizeof(*m),
 493                               ipv6->proto, ipv6->invflags & IP6T_INV_PROTO);
 494 }
 495
 496 static int
 497 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 498 {
 499         struct xt_match *match;
 500         int ret;
 501
 502         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
 503                                       m->u.user.revision);
 504         if (IS_ERR(match))
 505                 return PTR_ERR(match);
 506
 507         m->u.kernel.match = match;
 508
 509         ret = check_match(m, par);
 510         if (ret)
 511                 goto err;
 512
 513         return 0;
 514 err:
 515         module_put(m->u.kernel.match->me);
 516         return ret;
 517 }
 518
 519 static int check_target(struct ip6t_entry *e, struct net *net, const char *name)
 520 {
 521         struct xt_entry_target *t = ip6t_get_target(e);
 522         struct xt_tgchk_param par = {
 523                 .net       = net,
 524                 .table     = name,
 525                 .entryinfo = e,
 526                 .target    = t->u.kernel.target,
 527                 .targinfo  = t->data,
 528                 .hook_mask = e->comefrom,
 529                 .family    = NFPROTO_IPV6,
 530         };
 531
 532         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 533                                e->ipv6.proto,
 534                                e->ipv6.invflags & IP6T_INV_PROTO);
 535 }
 536
 537 static int
 538 find_check_entry(struct ip6t_entry *e, struct net *net, const char *name,
 539                  unsigned int size,
 540                  struct xt_percpu_counter_alloc_state *alloc_state)
 541 {
 542         struct xt_entry_target *t;
 543         struct xt_target *target;
 544         int ret;
 545         unsigned int j;
 546         struct xt_mtchk_param mtpar;
 547         struct xt_entry_match *ematch;
 548
 549         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 550                 return -ENOMEM;
 551
 552         j = 0;
 553         memset(&mtpar, 0, sizeof(mtpar));
 554         mtpar.net       = net;
 555         mtpar.table     = name;
 556         mtpar.entryinfo = &e->ipv6;
 557         mtpar.hook_mask = e->comefrom;
 558         mtpar.family    = NFPROTO_IPV6;
 559         xt_ematch_foreach(ematch, e) {
 560                 ret = find_check_match(ematch, &mtpar);
 561                 if (ret != 0)
 562                         goto cleanup_matches;
 563                 ++j;
 564         }
 565
 566         t = ip6t_get_target(e);
 567         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
 568                                         t->u.user.revision);
 569         if (IS_ERR(target)) {
 570                 ret = PTR_ERR(target);
 571                 goto cleanup_matches;
 572         }
 573         t->u.kernel.target = target;
 574
 575         ret = check_target(e, net, name);
 576         if (ret)
 577                 goto err;
 578         return 0;
 579  err:
 580         module_put(t->u.kernel.target->me);
 581  cleanup_matches:
 582         xt_ematch_foreach(ematch, e) {
 583                 if (j-- == 0)
 584                         break;
 585                 cleanup_match(ematch, net);
 586         }
 587
 588         xt_percpu_counter_free(&e->counters);
 589
 590         return ret;
 591 }
 592
 593 static bool check_underflow(const struct ip6t_entry *e)
 594 {
 595         const struct xt_entry_target *t;
 596         unsigned int verdict;
 597
 598         if (!unconditional(e))
 599                 return false;
 600         t = ip6t_get_target_c(e);
 601         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 602                 return false;
 603         verdict = ((struct xt_standard_target *)t)->verdict;
 604         verdict = -verdict - 1;
 605         return verdict == NF_DROP || verdict == NF_ACCEPT;
 606 }
 607
 608 static int
 609 check_entry_size_and_hooks(struct ip6t_entry *e,
 610                            struct xt_table_info *newinfo,
 611                            const unsigned char *base,
 612                            const unsigned char *limit,
 613                            const unsigned int *hook_entries,
 614                            const unsigned int *underflows,
 615                            unsigned int valid_hooks)
 616 {
 617         unsigned int h;
 618         int err;
 619
 620         if ((unsigned long)e % __alignof__(struct ip6t_entry) != 0 ||
 621             (unsigned char *)e + sizeof(struct ip6t_entry) >= limit ||
 622             (unsigned char *)e + e->next_offset > limit)
 623                 return -EINVAL;
 624
 625         if (e->next_offset
 626             < sizeof(struct ip6t_entry) + sizeof(struct xt_entry_target))
 627                 return -EINVAL;
 628
 629         if (!ip6_checkentry(&e->ipv6))
 630                 return -EINVAL;
 631
 632         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 633                                      e->next_offset);
 634         if (err)
 635                 return err;
 636
 637         /* Check hooks & underflows */
 638         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 639                 if (!(valid_hooks & (1 << h)))
 640                         continue;
 641                 if ((unsigned char *)e - base == hook_entries[h])
 642                         newinfo->hook_entry[h] = hook_entries[h];
 643                 if ((unsigned char *)e - base == underflows[h]) {
 644                         if (!check_underflow(e))
 645                                 return -EINVAL;
 646
 647                         newinfo->underflow[h] = underflows[h];
 648                 }
 649         }
 650
 651         /* Clear counters and comefrom */
 652         e->counters = ((struct xt_counters) { 0, 0 });
 653         e->comefrom = 0;
 654         return 0;
 655 }
 656
 657 static void cleanup_entry(struct ip6t_entry *e, struct net *net)
 658 {
 659         struct xt_tgdtor_param par;
 660         struct xt_entry_target *t;
 661         struct xt_entry_match *ematch;
 662
 663         /* Cleanup all matches */
 664         xt_ematch_foreach(ematch, e)
 665                 cleanup_match(ematch, net);
 666         t = ip6t_get_target(e);
 667
 668         par.net      = net;
 669         par.target   = t->u.kernel.target;
 670         par.targinfo = t->data;
 671         par.family   = NFPROTO_IPV6;
 672         if (par.target->destroy != NULL)
 673                 par.target->destroy(&par);
 674         module_put(par.target->me);
 675         xt_percpu_counter_free(&e->counters);
 676 }
 677
 678 /* Checks and translates the user-supplied table segment (held in
 679    newinfo) */
 680 static int
 681 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 682                 const struct ip6t_replace *repl)
 683 {
 684         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 685         struct ip6t_entry *iter;
 686         unsigned int *offsets;
 687         unsigned int i;
 688         int ret = 0;
 689
 690         newinfo->size = repl->size;
 691         newinfo->number = repl->num_entries;
 692
 693         /* Init all hooks to impossible value. */
 694         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 695                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 696                 newinfo->underflow[i] = 0xFFFFFFFF;
 697         }
 698
 699         offsets = xt_alloc_entry_offsets(newinfo->number);
 700         if (!offsets)
 701                 return -ENOMEM;
 702         i = 0;
 703         /* Walk through entries, checking offsets. */
 704         xt_entry_foreach(iter, entry0, newinfo->size) {
 705                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 706                                                  entry0 + repl->size,
 707                                                  repl->hook_entry,
 708                                                  repl->underflow,
 709                                                  repl->valid_hooks);
 710                 if (ret != 0)
 711                         goto out_free;
 712                 if (i < repl->num_entries)
 713                         offsets[i] = (void *)iter - entry0;
 714                 ++i;
 715                 if (strcmp(ip6t_get_target(iter)->u.user.name,
 716                     XT_ERROR_TARGET) == 0)
 717                         ++newinfo->stacksize;
 718         }
 719
 720         ret = -EINVAL;
 721         if (i != repl->num_entries)
 722                 goto out_free;
 723
 724         ret = xt_check_table_hooks(newinfo, repl->valid_hooks);
 725         if (ret)
 726                 goto out_free;
 727
 728         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 729                 ret = -ELOOP;
 730                 goto out_free;
 731         }
 732         kvfree(offsets);
 733
 734         /* Finally, each sanity check must pass */
 735         i = 0;
 736         xt_entry_foreach(iter, entry0, newinfo->size) {
 737                 ret = find_check_entry(iter, net, repl->name, repl->size,
 738                                        &alloc_state);
 739                 if (ret != 0)
 740                         break;
 741                 ++i;
 742         }
 743
 744         if (ret != 0) {
 745                 xt_entry_foreach(iter, entry0, newinfo->size) {
 746                         if (i-- == 0)
 747                                 break;
 748                         cleanup_entry(iter, net);
 749                 }
 750                 return ret;
 751         }
 752
 753         return ret;
 754  out_free:
 755         kvfree(offsets);
 756         return ret;
 757 }
 758
 759 static void
 760 get_counters(const struct xt_table_info *t,
 761              struct xt_counters counters[])
 762 {
 763         struct ip6t_entry *iter;
 764         unsigned int cpu;
 765         unsigned int i;
 766
 767         for_each_possible_cpu(cpu) {
 768                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 769
 770                 i = 0;
 771                 xt_entry_foreach(iter, t->entries, t->size) {
 772                         struct xt_counters *tmp;
 773                         u64 bcnt, pcnt;
 774                         unsigned int start;
 775
 776                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 777                         do {
 778                                 start = read_seqcount_begin(s);
 779                                 bcnt = tmp->bcnt;
 780                                 pcnt = tmp->pcnt;
 781                         } while (read_seqcount_retry(s, start));
 782
 783                         ADD_COUNTER(counters[i], bcnt, pcnt);
 784                         ++i;
 785                         cond_resched();
 786                 }
 787         }
 788 }
 789
 790 static void get_old_counters(const struct xt_table_info *t,
 791                              struct xt_counters counters[])
 792 {
 793         struct ip6t_entry *iter;
 794         unsigned int cpu, i;
 795
 796         for_each_possible_cpu(cpu) {
 797                 i = 0;
 798                 xt_entry_foreach(iter, t->entries, t->size) {
 799                         const struct xt_counters *tmp;
 800
 801                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 802                         ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
 803                         ++i;
 804                 }
 805                 cond_resched();
 806         }
 807 }
 808
 809 static struct xt_counters *alloc_counters(const struct xt_table *table)
 810 {
 811         unsigned int countersize;
 812         struct xt_counters *counters;
 813         const struct xt_table_info *private = table->private;
 814
 815         /* We need atomic snapshot of counters: rest doesn't change
 816            (other than comefrom, which userspace doesn't care
 817            about). */
 818         countersize = sizeof(struct xt_counters) * private->number;
 819         counters = vzalloc(countersize);
 820
 821         if (counters == NULL)
 822                 return ERR_PTR(-ENOMEM);
 823
 824         get_counters(private, counters);
 825
 826         return counters;
 827 }
 828
 829 static int
 830 copy_entries_to_user(unsigned int total_size,
 831                      const struct xt_table *table,
 832                      void __user *userptr)
 833 {
 834         unsigned int off, num;
 835         const struct ip6t_entry *e;
 836         struct xt_counters *counters;
 837         const struct xt_table_info *private = table->private;
 838         int ret = 0;
 839         const void *loc_cpu_entry;
 840
 841         counters = alloc_counters(table);
 842         if (IS_ERR(counters))
 843                 return PTR_ERR(counters);
 844
 845         loc_cpu_entry = private->entries;
 846
 847         /* FIXME: use iterator macros --RR */
 848         /* ... then go back and fix counters and names */
 849         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 850                 unsigned int i;
 851                 const struct xt_entry_match *m;
 852                 const struct xt_entry_target *t;
 853
 854                 e = loc_cpu_entry + off;
 855                 if (copy_to_user(userptr + off, e, sizeof(*e))) {
 856                         ret = -EFAULT;
 857                         goto free_counters;
 858                 }
 859                 if (copy_to_user(userptr + off
 860                                  + offsetof(struct ip6t_entry, counters),
 861                                  &counters[num],
 862                                  sizeof(counters[num])) != 0) {
 863                         ret = -EFAULT;
 864                         goto free_counters;
 865                 }
 866
 867                 for (i = sizeof(struct ip6t_entry);
 868                      i < e->target_offset;
 869                      i += m->u.match_size) {
 870                         m = (void *)e + i;
 871
 872                         if (xt_match_to_user(m, userptr + off + i)) {
 873                                 ret = -EFAULT;
 874                                 goto free_counters;
 875                         }
 876                 }
 877
 878                 t = ip6t_get_target_c(e);
 879                 if (xt_target_to_user(t, userptr + off + e->target_offset)) {
 880                         ret = -EFAULT;
 881                         goto free_counters;
 882                 }
 883         }
 884
 885  free_counters:
 886         vfree(counters);
 887         return ret;
 888 }
 889
 890 #ifdef CONFIG_COMPAT
 891 static void compat_standard_from_user(void *dst, const void *src)
 892 {
 893         int v = *(compat_int_t *)src;
 894
 895         if (v > 0)
 896                 v += xt_compat_calc_jump(AF_INET6, v);
 897         memcpy(dst, &v, sizeof(v));
 898 }
 899
 900 static int compat_standard_to_user(void __user *dst, const void *src)
 901 {
 902         compat_int_t cv = *(int *)src;
 903
 904         if (cv > 0)
 905                 cv -= xt_compat_calc_jump(AF_INET6, cv);
 906         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 907 }
 908
 909 static int compat_calc_entry(const struct ip6t_entry *e,
 910                              const struct xt_table_info *info,
 911                              const void *base, struct xt_table_info *newinfo)
 912 {
 913         const struct xt_entry_match *ematch;
 914         const struct xt_entry_target *t;
 915         unsigned int entry_offset;
 916         int off, i, ret;
 917
 918         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
 919         entry_offset = (void *)e - base;
 920         xt_ematch_foreach(ematch, e)
 921                 off += xt_compat_match_offset(ematch->u.kernel.match);
 922         t = ip6t_get_target_c(e);
 923         off += xt_compat_target_offset(t->u.kernel.target);
 924         newinfo->size -= off;
 925         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
 926         if (ret)
 927                 return ret;
 928
 929         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 930                 if (info->hook_entry[i] &&
 931                     (e < (struct ip6t_entry *)(base + info->hook_entry[i])))
 932                         newinfo->hook_entry[i] -= off;
 933                 if (info->underflow[i] &&
 934                     (e < (struct ip6t_entry *)(base + info->underflow[i])))
 935                         newinfo->underflow[i] -= off;
 936         }
 937         return 0;
 938 }
 939
 940 static int compat_table_info(const struct xt_table_info *info,
 941                              struct xt_table_info *newinfo)
 942 {
 943         struct ip6t_entry *iter;
 944         const void *loc_cpu_entry;
 945         int ret;
 946
 947         if (!newinfo || !info)
 948                 return -EINVAL;
 949
 950         /* we dont care about newinfo->entries */
 951         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 952         newinfo->initial_entries = 0;
 953         loc_cpu_entry = info->entries;
 954         ret = xt_compat_init_offsets(AF_INET6, info->number);
 955         if (ret)
 956                 return ret;
 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 ip6t_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_INET6);
 983 #endif
 984         t = xt_request_find_table_lock(net, AF_INET6, name);
 985         if (!IS_ERR(t)) {
 986                 struct ip6t_getinfo info;
 987                 const struct xt_table_info *private = t->private;
 988 #ifdef CONFIG_COMPAT
 989                 struct xt_table_info tmp;
 990
 991                 if (compat) {
 992                         ret = compat_table_info(private, &tmp);
 993                         xt_compat_flush_offsets(AF_INET6);
 994                         private = &tmp;
 995                 }
 996 #endif
 997                 memset(&info, 0, sizeof(info));
 998                 info.valid_hooks = t->valid_hooks;
 999                 memcpy(info.hook_entry, private->hook_entry,
1000                        sizeof(info.hook_entry));
1001                 memcpy(info.underflow, private->underflow,
1002                        sizeof(info.underflow));
1003                 info.num_entries = private->number;
1004                 info.size = private->size;
1005                 strcpy(info.name, name);
1006
1007                 if (copy_to_user(user, &info, *len) != 0)
1008                         ret = -EFAULT;
1009                 else
1010                         ret = 0;
1011
1012                 xt_table_unlock(t);
1013                 module_put(t->me);
1014         } else
1015                 ret = PTR_ERR(t);
1016 #ifdef CONFIG_COMPAT
1017         if (compat)
1018                 xt_compat_unlock(AF_INET6);
1019 #endif
1020         return ret;
1021 }
1022
1023 static int
1024 get_entries(struct net *net, struct ip6t_get_entries __user *uptr,
1025             const int *len)
1026 {
1027         int ret;
1028         struct ip6t_get_entries get;
1029         struct xt_table *t;
1030
1031         if (*len < sizeof(get))
1032                 return -EINVAL;
1033         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1034                 return -EFAULT;
1035         if (*len != sizeof(struct ip6t_get_entries) + get.size)
1036                 return -EINVAL;
1037
1038         get.name[sizeof(get.name) - 1] = '\0';
1039
1040         t = xt_find_table_lock(net, AF_INET6, get.name);
1041         if (!IS_ERR(t)) {
1042                 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 = PTR_ERR(t);
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 ip6t_entry *iter;
1067
1068         ret = 0;
1069         counters = xt_counters_alloc(num_counters);
1070         if (!counters) {
1071                 ret = -ENOMEM;
1072                 goto out;
1073         }
1074
1075         t = xt_request_find_table_lock(net, AF_INET6, name);
1076         if (IS_ERR(t)) {
1077                 ret = PTR_ERR(t);
1078                 goto free_newinfo_counters_untrans;
1079         }
1080
1081         /* You lied! */
1082         if (valid_hooks != t->valid_hooks) {
1083                 ret = -EINVAL;
1084                 goto put_module;
1085         }
1086
1087         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1088         if (!oldinfo)
1089                 goto put_module;
1090
1091         /* Update module usage count based on number of rules */
1092         if ((oldinfo->number > oldinfo->initial_entries) ||
1093             (newinfo->number <= oldinfo->initial_entries))
1094                 module_put(t->me);
1095         if ((oldinfo->number > oldinfo->initial_entries) &&
1096             (newinfo->number <= oldinfo->initial_entries))
1097                 module_put(t->me);
1098
1099         xt_table_unlock(t);
1100
1101         get_old_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("ip6tables: counters copy to user failed while replacing table\n");
1112         }
1113         vfree(counters);
1114         return ret;
1115
1116  put_module:
1117         module_put(t->me);
1118         xt_table_unlock(t);
1119  free_newinfo_counters_untrans:
1120         vfree(counters);
1121  out:
1122         return ret;
1123 }
1124
1125 static int
1126 do_replace(struct net *net, const void __user *user, unsigned int len)
1127 {
1128         int ret;
1129         struct ip6t_replace tmp;
1130         struct xt_table_info *newinfo;
1131         void *loc_cpu_entry;
1132         struct ip6t_entry *iter;
1133
1134         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1135                 return -EFAULT;
1136
1137         /* overflow check */
1138         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1139                 return -ENOMEM;
1140         if (tmp.num_counters == 0)
1141                 return -EINVAL;
1142
1143         tmp.name[sizeof(tmp.name)-1] = 0;
1144
1145         newinfo = xt_alloc_table_info(tmp.size);
1146         if (!newinfo)
1147                 return -ENOMEM;
1148
1149         loc_cpu_entry = newinfo->entries;
1150         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1151                            tmp.size) != 0) {
1152                 ret = -EFAULT;
1153                 goto free_newinfo;
1154         }
1155
1156         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1157         if (ret != 0)
1158                 goto free_newinfo;
1159
1160         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1161                            tmp.num_counters, tmp.counters);
1162         if (ret)
1163                 goto free_newinfo_untrans;
1164         return 0;
1165
1166  free_newinfo_untrans:
1167         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1168                 cleanup_entry(iter, net);
1169  free_newinfo:
1170         xt_free_table_info(newinfo);
1171         return ret;
1172 }
1173
1174 static int
1175 do_add_counters(struct net *net, const void __user *user, unsigned int len,
1176                 int compat)
1177 {
1178         unsigned int i;
1179         struct xt_counters_info tmp;
1180         struct xt_counters *paddc;
1181         struct xt_table *t;
1182         const struct xt_table_info *private;
1183         int ret = 0;
1184         struct ip6t_entry *iter;
1185         unsigned int addend;
1186
1187         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1188         if (IS_ERR(paddc))
1189                 return PTR_ERR(paddc);
1190         t = xt_find_table_lock(net, AF_INET6, tmp.name);
1191         if (IS_ERR(t)) {
1192                 ret = PTR_ERR(t);
1193                 goto free;
1194         }
1195
1196         local_bh_disable();
1197         private = t->private;
1198         if (private->number != tmp.num_counters) {
1199                 ret = -EINVAL;
1200                 goto unlock_up_free;
1201         }
1202
1203         i = 0;
1204         addend = xt_write_recseq_begin();
1205         xt_entry_foreach(iter, private->entries, private->size) {
1206                 struct xt_counters *tmp;
1207
1208                 tmp = xt_get_this_cpu_counter(&iter->counters);
1209                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1210                 ++i;
1211         }
1212         xt_write_recseq_end(addend);
1213  unlock_up_free:
1214         local_bh_enable();
1215         xt_table_unlock(t);
1216         module_put(t->me);
1217  free:
1218         vfree(paddc);
1219
1220         return ret;
1221 }
1222
1223 #ifdef CONFIG_COMPAT
1224 struct compat_ip6t_replace {
1225         char                    name[XT_TABLE_MAXNAMELEN];
1226         u32                     valid_hooks;
1227         u32                     num_entries;
1228         u32                     size;
1229         u32                     hook_entry[NF_INET_NUMHOOKS];
1230         u32                     underflow[NF_INET_NUMHOOKS];
1231         u32                     num_counters;
1232         compat_uptr_t           counters;       /* struct xt_counters * */
1233         struct compat_ip6t_entry entries[0];
1234 };
1235
1236 static int
1237 compat_copy_entry_to_user(struct ip6t_entry *e, void __user **dstptr,
1238                           unsigned int *size, struct xt_counters *counters,
1239                           unsigned int i)
1240 {
1241         struct xt_entry_target *t;
1242         struct compat_ip6t_entry __user *ce;
1243         u_int16_t target_offset, next_offset;
1244         compat_uint_t origsize;
1245         const struct xt_entry_match *ematch;
1246         int ret = 0;
1247
1248         origsize = *size;
1249         ce = *dstptr;
1250         if (copy_to_user(ce, e, sizeof(struct ip6t_entry)) != 0 ||
1251             copy_to_user(&ce->counters, &counters[i],
1252             sizeof(counters[i])) != 0)
1253                 return -EFAULT;
1254
1255         *dstptr += sizeof(struct compat_ip6t_entry);
1256         *size -= sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1257
1258         xt_ematch_foreach(ematch, e) {
1259                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1260                 if (ret != 0)
1261                         return ret;
1262         }
1263         target_offset = e->target_offset - (origsize - *size);
1264         t = ip6t_get_target(e);
1265         ret = xt_compat_target_to_user(t, dstptr, size);
1266         if (ret)
1267                 return ret;
1268         next_offset = e->next_offset - (origsize - *size);
1269         if (put_user(target_offset, &ce->target_offset) != 0 ||
1270             put_user(next_offset, &ce->next_offset) != 0)
1271                 return -EFAULT;
1272         return 0;
1273 }
1274
1275 static int
1276 compat_find_calc_match(struct xt_entry_match *m,
1277                        const struct ip6t_ip6 *ipv6,
1278                        int *size)
1279 {
1280         struct xt_match *match;
1281
1282         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
1283                                       m->u.user.revision);
1284         if (IS_ERR(match))
1285                 return PTR_ERR(match);
1286
1287         m->u.kernel.match = match;
1288         *size += xt_compat_match_offset(match);
1289         return 0;
1290 }
1291
1292 static void compat_release_entry(struct compat_ip6t_entry *e)
1293 {
1294         struct xt_entry_target *t;
1295         struct xt_entry_match *ematch;
1296
1297         /* Cleanup all matches */
1298         xt_ematch_foreach(ematch, e)
1299                 module_put(ematch->u.kernel.match->me);
1300         t = compat_ip6t_get_target(e);
1301         module_put(t->u.kernel.target->me);
1302 }
1303
1304 static int
1305 check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
1306                                   struct xt_table_info *newinfo,
1307                                   unsigned int *size,
1308                                   const unsigned char *base,
1309                                   const unsigned char *limit)
1310 {
1311         struct xt_entry_match *ematch;
1312         struct xt_entry_target *t;
1313         struct xt_target *target;
1314         unsigned int entry_offset;
1315         unsigned int j;
1316         int ret, off;
1317
1318         if ((unsigned long)e % __alignof__(struct compat_ip6t_entry) != 0 ||
1319             (unsigned char *)e + sizeof(struct compat_ip6t_entry) >= limit ||
1320             (unsigned char *)e + e->next_offset > limit)
1321                 return -EINVAL;
1322
1323         if (e->next_offset < sizeof(struct compat_ip6t_entry) +
1324                              sizeof(struct compat_xt_entry_target))
1325                 return -EINVAL;
1326
1327         if (!ip6_checkentry(&e->ipv6))
1328                 return -EINVAL;
1329
1330         ret = xt_compat_check_entry_offsets(e, e->elems,
1331                                             e->target_offset, e->next_offset);
1332         if (ret)
1333                 return ret;
1334
1335         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1336         entry_offset = (void *)e - (void *)base;
1337         j = 0;
1338         xt_ematch_foreach(ematch, e) {
1339                 ret = compat_find_calc_match(ematch, &e->ipv6, &off);
1340                 if (ret != 0)
1341                         goto release_matches;
1342                 ++j;
1343         }
1344
1345         t = compat_ip6t_get_target(e);
1346         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
1347                                         t->u.user.revision);
1348         if (IS_ERR(target)) {
1349                 ret = PTR_ERR(target);
1350                 goto release_matches;
1351         }
1352         t->u.kernel.target = target;
1353
1354         off += xt_compat_target_offset(target);
1355         *size += off;
1356         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1357         if (ret)
1358                 goto out;
1359
1360         return 0;
1361
1362 out:
1363         module_put(t->u.kernel.target->me);
1364 release_matches:
1365         xt_ematch_foreach(ematch, e) {
1366                 if (j-- == 0)
1367                         break;
1368                 module_put(ematch->u.kernel.match->me);
1369         }
1370         return ret;
1371 }
1372
1373 static void
1374 compat_copy_entry_from_user(struct compat_ip6t_entry *e, void **dstptr,
1375                             unsigned int *size,
1376                             struct xt_table_info *newinfo, unsigned char *base)
1377 {
1378         struct xt_entry_target *t;
1379         struct ip6t_entry *de;
1380         unsigned int origsize;
1381         int h;
1382         struct xt_entry_match *ematch;
1383
1384         origsize = *size;
1385         de = *dstptr;
1386         memcpy(de, e, sizeof(struct ip6t_entry));
1387         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1388
1389         *dstptr += sizeof(struct ip6t_entry);
1390         *size += sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1391
1392         xt_ematch_foreach(ematch, e)
1393                 xt_compat_match_from_user(ematch, dstptr, size);
1394
1395         de->target_offset = e->target_offset - (origsize - *size);
1396         t = compat_ip6t_get_target(e);
1397         xt_compat_target_from_user(t, dstptr, size);
1398
1399         de->next_offset = e->next_offset - (origsize - *size);
1400         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1401                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1402                         newinfo->hook_entry[h] -= origsize - *size;
1403                 if ((unsigned char *)de - base < newinfo->underflow[h])
1404                         newinfo->underflow[h] -= origsize - *size;
1405         }
1406 }
1407
1408 static int
1409 translate_compat_table(struct net *net,
1410                        struct xt_table_info **pinfo,
1411                        void **pentry0,
1412                        const struct compat_ip6t_replace *compatr)
1413 {
1414         unsigned int i, j;
1415         struct xt_table_info *newinfo, *info;
1416         void *pos, *entry0, *entry1;
1417         struct compat_ip6t_entry *iter0;
1418         struct ip6t_replace repl;
1419         unsigned int size;
1420         int ret;
1421
1422         info = *pinfo;
1423         entry0 = *pentry0;
1424         size = compatr->size;
1425         info->number = compatr->num_entries;
1426
1427         j = 0;
1428         xt_compat_lock(AF_INET6);
1429         ret = xt_compat_init_offsets(AF_INET6, compatr->num_entries);
1430         if (ret)
1431                 goto out_unlock;
1432         /* Walk through entries, checking offsets. */
1433         xt_entry_foreach(iter0, entry0, compatr->size) {
1434                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1435                                                         entry0,
1436                                                         entry0 + compatr->size);
1437                 if (ret != 0)
1438                         goto out_unlock;
1439                 ++j;
1440         }
1441
1442         ret = -EINVAL;
1443         if (j != compatr->num_entries)
1444                 goto out_unlock;
1445
1446         ret = -ENOMEM;
1447         newinfo = xt_alloc_table_info(size);
1448         if (!newinfo)
1449                 goto out_unlock;
1450
1451         memset(newinfo->entries, 0, size);
1452
1453         newinfo->number = compatr->num_entries;
1454         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1455                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1456                 newinfo->underflow[i] = compatr->underflow[i];
1457         }
1458         entry1 = newinfo->entries;
1459         pos = entry1;
1460         size = compatr->size;
1461         xt_entry_foreach(iter0, entry0, compatr->size)
1462                 compat_copy_entry_from_user(iter0, &pos, &size,
1463                                             newinfo, entry1);
1464
1465         /* all module references in entry0 are now gone. */
1466         xt_compat_flush_offsets(AF_INET6);
1467         xt_compat_unlock(AF_INET6);
1468
1469         memcpy(&repl, compatr, sizeof(*compatr));
1470
1471         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1472                 repl.hook_entry[i] = newinfo->hook_entry[i];
1473                 repl.underflow[i] = newinfo->underflow[i];
1474         }
1475
1476         repl.num_counters = 0;
1477         repl.counters = NULL;
1478         repl.size = newinfo->size;
1479         ret = translate_table(net, newinfo, entry1, &repl);
1480         if (ret)
1481                 goto free_newinfo;
1482
1483         *pinfo = newinfo;
1484         *pentry0 = entry1;
1485         xt_free_table_info(info);
1486         return 0;
1487
1488 free_newinfo:
1489         xt_free_table_info(newinfo);
1490         return ret;
1491 out_unlock:
1492         xt_compat_flush_offsets(AF_INET6);
1493         xt_compat_unlock(AF_INET6);
1494         xt_entry_foreach(iter0, entry0, compatr->size) {
1495                 if (j-- == 0)
1496                         break;
1497                 compat_release_entry(iter0);
1498         }
1499         return ret;
1500 }
1501
1502 static int
1503 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1504 {
1505         int ret;
1506         struct compat_ip6t_replace tmp;
1507         struct xt_table_info *newinfo;
1508         void *loc_cpu_entry;
1509         struct ip6t_entry *iter;
1510
1511         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1512                 return -EFAULT;
1513
1514         /* overflow check */
1515         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1516                 return -ENOMEM;
1517         if (tmp.num_counters == 0)
1518                 return -EINVAL;
1519
1520         tmp.name[sizeof(tmp.name)-1] = 0;
1521
1522         newinfo = xt_alloc_table_info(tmp.size);
1523         if (!newinfo)
1524                 return -ENOMEM;
1525
1526         loc_cpu_entry = newinfo->entries;
1527         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1528                            tmp.size) != 0) {
1529                 ret = -EFAULT;
1530                 goto free_newinfo;
1531         }
1532
1533         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1534         if (ret != 0)
1535                 goto free_newinfo;
1536
1537         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1538                            tmp.num_counters, compat_ptr(tmp.counters));
1539         if (ret)
1540                 goto free_newinfo_untrans;
1541         return 0;
1542
1543  free_newinfo_untrans:
1544         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1545                 cleanup_entry(iter, net);
1546  free_newinfo:
1547         xt_free_table_info(newinfo);
1548         return ret;
1549 }
1550
1551 static int
1552 compat_do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user,
1553                        unsigned int len)
1554 {
1555         int ret;
1556
1557         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1558                 return -EPERM;
1559
1560         switch (cmd) {
1561         case IP6T_SO_SET_REPLACE:
1562                 ret = compat_do_replace(sock_net(sk), user, len);
1563                 break;
1564
1565         case IP6T_SO_SET_ADD_COUNTERS:
1566                 ret = do_add_counters(sock_net(sk), user, len, 1);
1567                 break;
1568
1569         default:
1570                 ret = -EINVAL;
1571         }
1572
1573         return ret;
1574 }
1575
1576 struct compat_ip6t_get_entries {
1577         char name[XT_TABLE_MAXNAMELEN];
1578         compat_uint_t size;
1579         struct compat_ip6t_entry entrytable[0];
1580 };
1581
1582 static int
1583 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1584                             void __user *userptr)
1585 {
1586         struct xt_counters *counters;
1587         const struct xt_table_info *private = table->private;
1588         void __user *pos;
1589         unsigned int size;
1590         int ret = 0;
1591         unsigned int i = 0;
1592         struct ip6t_entry *iter;
1593
1594         counters = alloc_counters(table);
1595         if (IS_ERR(counters))
1596                 return PTR_ERR(counters);
1597
1598         pos = userptr;
1599         size = total_size;
1600         xt_entry_foreach(iter, private->entries, total_size) {
1601                 ret = compat_copy_entry_to_user(iter, &pos,
1602                                                 &size, counters, i++);
1603                 if (ret != 0)
1604                         break;
1605         }
1606
1607         vfree(counters);
1608         return ret;
1609 }
1610
1611 static int
1612 compat_get_entries(struct net *net, struct compat_ip6t_get_entries __user *uptr,
1613                    int *len)
1614 {
1615         int ret;
1616         struct compat_ip6t_get_entries get;
1617         struct xt_table *t;
1618
1619         if (*len < sizeof(get))
1620                 return -EINVAL;
1621
1622         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1623                 return -EFAULT;
1624
1625         if (*len != sizeof(struct compat_ip6t_get_entries) + get.size)
1626                 return -EINVAL;
1627
1628         get.name[sizeof(get.name) - 1] = '\0';
1629
1630         xt_compat_lock(AF_INET6);
1631         t = xt_find_table_lock(net, AF_INET6, get.name);
1632         if (!IS_ERR(t)) {
1633                 const struct xt_table_info *private = t->private;
1634                 struct xt_table_info info;
1635                 ret = compat_table_info(private, &info);
1636                 if (!ret && get.size == info.size)
1637                         ret = compat_copy_entries_to_user(private->size,
1638                                                           t, uptr->entrytable);
1639                 else if (!ret)
1640                         ret = -EAGAIN;
1641
1642                 xt_compat_flush_offsets(AF_INET6);
1643                 module_put(t->me);
1644                 xt_table_unlock(t);
1645         } else
1646                 ret = PTR_ERR(t);
1647
1648         xt_compat_unlock(AF_INET6);
1649         return ret;
1650 }
1651
1652 static int do_ip6t_get_ctl(struct sock *, int, void __user *, int *);
1653
1654 static int
1655 compat_do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1656 {
1657         int ret;
1658
1659         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1660                 return -EPERM;
1661
1662         switch (cmd) {
1663         case IP6T_SO_GET_INFO:
1664                 ret = get_info(sock_net(sk), user, len, 1);
1665                 break;
1666         case IP6T_SO_GET_ENTRIES:
1667                 ret = compat_get_entries(sock_net(sk), user, len);
1668                 break;
1669         default:
1670                 ret = do_ip6t_get_ctl(sk, cmd, user, len);
1671         }
1672         return ret;
1673 }
1674 #endif
1675
1676 static int
1677 do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1678 {
1679         int ret;
1680
1681         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1682                 return -EPERM;
1683
1684         switch (cmd) {
1685         case IP6T_SO_SET_REPLACE:
1686                 ret = do_replace(sock_net(sk), user, len);
1687                 break;
1688
1689         case IP6T_SO_SET_ADD_COUNTERS:
1690                 ret = do_add_counters(sock_net(sk), user, len, 0);
1691                 break;
1692
1693         default:
1694                 ret = -EINVAL;
1695         }
1696
1697         return ret;
1698 }
1699
1700 static int
1701 do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1702 {
1703         int ret;
1704
1705         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1706                 return -EPERM;
1707
1708         switch (cmd) {
1709         case IP6T_SO_GET_INFO:
1710                 ret = get_info(sock_net(sk), user, len, 0);
1711                 break;
1712
1713         case IP6T_SO_GET_ENTRIES:
1714                 ret = get_entries(sock_net(sk), user, len);
1715                 break;
1716
1717         case IP6T_SO_GET_REVISION_MATCH:
1718         case IP6T_SO_GET_REVISION_TARGET: {
1719                 struct xt_get_revision rev;
1720                 int target;
1721
1722                 if (*len != sizeof(rev)) {
1723                         ret = -EINVAL;
1724                         break;
1725                 }
1726                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1727                         ret = -EFAULT;
1728                         break;
1729                 }
1730                 rev.name[sizeof(rev.name)-1] = 0;
1731
1732                 if (cmd == IP6T_SO_GET_REVISION_TARGET)
1733                         target = 1;
1734                 else
1735                         target = 0;
1736
1737                 try_then_request_module(xt_find_revision(AF_INET6, rev.name,
1738                                                          rev.revision,
1739                                                          target, &ret),
1740                                         "ip6t_%s", rev.name);
1741                 break;
1742         }
1743
1744         default:
1745                 ret = -EINVAL;
1746         }
1747
1748         return ret;
1749 }
1750
1751 static void __ip6t_unregister_table(struct net *net, struct xt_table *table)
1752 {
1753         struct xt_table_info *private;
1754         void *loc_cpu_entry;
1755         struct module *table_owner = table->me;
1756         struct ip6t_entry *iter;
1757
1758         private = xt_unregister_table(table);
1759
1760         /* Decrease module usage counts and free resources */
1761         loc_cpu_entry = private->entries;
1762         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1763                 cleanup_entry(iter, net);
1764         if (private->number > private->initial_entries)
1765                 module_put(table_owner);
1766         xt_free_table_info(private);
1767 }
1768
1769 int ip6t_register_table(struct net *net, const struct xt_table *table,
1770                         const struct ip6t_replace *repl,
1771                         const struct nf_hook_ops *ops,
1772                         struct xt_table **res)
1773 {
1774         int ret;
1775         struct xt_table_info *newinfo;
1776         struct xt_table_info bootstrap = {0};
1777         void *loc_cpu_entry;
1778         struct xt_table *new_table;
1779
1780         newinfo = xt_alloc_table_info(repl->size);
1781         if (!newinfo)
1782                 return -ENOMEM;
1783
1784         loc_cpu_entry = newinfo->entries;
1785         memcpy(loc_cpu_entry, repl->entries, repl->size);
1786
1787         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1788         if (ret != 0)
1789                 goto out_free;
1790
1791         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1792         if (IS_ERR(new_table)) {
1793                 ret = PTR_ERR(new_table);
1794                 goto out_free;
1795         }
1796
1797         /* set res now, will see skbs right after nf_register_net_hooks */
1798         WRITE_ONCE(*res, new_table);
1799         if (!ops)
1800                 return 0;
1801
1802         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1803         if (ret != 0) {
1804                 __ip6t_unregister_table(net, new_table);
1805                 *res = NULL;
1806         }
1807
1808         return ret;
1809
1810 out_free:
1811         xt_free_table_info(newinfo);
1812         return ret;
1813 }
1814
1815 void ip6t_unregister_table(struct net *net, struct xt_table *table,
1816                            const struct nf_hook_ops *ops)
1817 {
1818         if (ops)
1819                 nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1820         __ip6t_unregister_table(net, table);
1821 }
1822
1823 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1824 static inline bool
1825 icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1826                      u_int8_t type, u_int8_t code,
1827                      bool invert)
1828 {
1829         return (type == test_type && code >= min_code && code <= max_code)
1830                 ^ invert;
1831 }
1832
1833 static bool
1834 icmp6_match(const struct sk_buff *skb, struct xt_action_param *par)
1835 {
1836         const struct icmp6hdr *ic;
1837         struct icmp6hdr _icmph;
1838         const struct ip6t_icmp *icmpinfo = par->matchinfo;
1839
1840         /* Must not be a fragment. */
1841         if (par->fragoff != 0)
1842                 return false;
1843
1844         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1845         if (ic == NULL) {
1846                 /* We've been asked to examine this packet, and we
1847                  * can't.  Hence, no choice but to drop.
1848                  */
1849                 par->hotdrop = true;
1850                 return false;
1851         }
1852
1853         return icmp6_type_code_match(icmpinfo->type,
1854                                      icmpinfo->code[0],
1855                                      icmpinfo->code[1],
1856                                      ic->icmp6_type, ic->icmp6_code,
1857                                      !!(icmpinfo->invflags&IP6T_ICMP_INV));
1858 }
1859
1860 /* Called when user tries to insert an entry of this type. */
1861 static int icmp6_checkentry(const struct xt_mtchk_param *par)
1862 {
1863         const struct ip6t_icmp *icmpinfo = par->matchinfo;
1864
1865         /* Must specify no unknown invflags */
1866         return (icmpinfo->invflags & ~IP6T_ICMP_INV) ? -EINVAL : 0;
1867 }
1868
1869 /* The built-in targets: standard (NULL) and error. */
1870 static struct xt_target ip6t_builtin_tg[] __read_mostly = {
1871         {
1872                 .name             = XT_STANDARD_TARGET,
1873                 .targetsize       = sizeof(int),
1874                 .family           = NFPROTO_IPV6,
1875 #ifdef CONFIG_COMPAT
1876                 .compatsize       = sizeof(compat_int_t),
1877                 .compat_from_user = compat_standard_from_user,
1878                 .compat_to_user   = compat_standard_to_user,
1879 #endif
1880         },
1881         {
1882                 .name             = XT_ERROR_TARGET,
1883                 .target           = ip6t_error,
1884                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1885                 .family           = NFPROTO_IPV6,
1886         },
1887 };
1888
1889 static struct nf_sockopt_ops ip6t_sockopts = {
1890         .pf             = PF_INET6,
1891         .set_optmin     = IP6T_BASE_CTL,
1892         .set_optmax     = IP6T_SO_SET_MAX+1,
1893         .set            = do_ip6t_set_ctl,
1894 #ifdef CONFIG_COMPAT
1895         .compat_set     = compat_do_ip6t_set_ctl,
1896 #endif
1897         .get_optmin     = IP6T_BASE_CTL,
1898         .get_optmax     = IP6T_SO_GET_MAX+1,
1899         .get            = do_ip6t_get_ctl,
1900 #ifdef CONFIG_COMPAT
1901         .compat_get     = compat_do_ip6t_get_ctl,
1902 #endif
1903         .owner          = THIS_MODULE,
1904 };
1905
1906 static struct xt_match ip6t_builtin_mt[] __read_mostly = {
1907         {
1908                 .name       = "icmp6",
1909                 .match      = icmp6_match,
1910                 .matchsize  = sizeof(struct ip6t_icmp),
1911                 .checkentry = icmp6_checkentry,
1912                 .proto      = IPPROTO_ICMPV6,
1913                 .family     = NFPROTO_IPV6,
1914                 .me         = THIS_MODULE,
1915         },
1916 };
1917
1918 static int __net_init ip6_tables_net_init(struct net *net)
1919 {
1920         return xt_proto_init(net, NFPROTO_IPV6);
1921 }
1922
1923 static void __net_exit ip6_tables_net_exit(struct net *net)
1924 {
1925         xt_proto_fini(net, NFPROTO_IPV6);
1926 }
1927
1928 static struct pernet_operations ip6_tables_net_ops = {
1929         .init = ip6_tables_net_init,
1930         .exit = ip6_tables_net_exit,
1931 };
1932
1933 static int __init ip6_tables_init(void)
1934 {
1935         int ret;
1936
1937         ret = register_pernet_subsys(&ip6_tables_net_ops);
1938         if (ret < 0)
1939                 goto err1;
1940
1941         /* No one else will be downing sem now, so we won't sleep */
1942         ret = xt_register_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1943         if (ret < 0)
1944                 goto err2;
1945         ret = xt_register_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1946         if (ret < 0)
1947                 goto err4;
1948
1949         /* Register setsockopt */
1950         ret = nf_register_sockopt(&ip6t_sockopts);
1951         if (ret < 0)
1952                 goto err5;
1953
1954         return 0;
1955
1956 err5:
1957         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1958 err4:
1959         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1960 err2:
1961         unregister_pernet_subsys(&ip6_tables_net_ops);
1962 err1:
1963         return ret;
1964 }
1965
1966 static void __exit ip6_tables_fini(void)
1967 {
1968         nf_unregister_sockopt(&ip6t_sockopts);
1969
1970         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1971         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1972         unregister_pernet_subsys(&ip6_tables_net_ops);
1973 }
1974
1975 EXPORT_SYMBOL(ip6t_register_table);
1976 EXPORT_SYMBOL(ip6t_unregister_table);
1977 EXPORT_SYMBOL(ip6t_do_table);
1978
1979 module_init(ip6_tables_init);
1980 module_exit(ip6_tables_fini);