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
  42 void *ip6t_alloc_initial_table(const struct xt_table *info)
  43 {
  44         return xt_alloc_initial_table(ip6t, IP6T);
  45 }
  46 EXPORT_SYMBOL_GPL(ip6t_alloc_initial_table);
  47
  48 /* Returns whether matches rule or not. */
  49 /* Performance critical - called for every packet */
  50 static inline bool
  51 ip6_packet_match(const struct sk_buff *skb,
  52                  const char *indev,
  53                  const char *outdev,
  54                  const struct ip6t_ip6 *ip6info,
  55                  unsigned int *protoff,
  56                  int *fragoff, bool *hotdrop)
  57 {
  58         unsigned long ret;
  59         const struct ipv6hdr *ipv6 = ipv6_hdr(skb);
  60
  61         if (NF_INVF(ip6info, IP6T_INV_SRCIP,
  62                     ipv6_masked_addr_cmp(&ipv6->saddr, &ip6info->smsk,
  63                                          &ip6info->src)) ||
  64             NF_INVF(ip6info, IP6T_INV_DSTIP,
  65                     ipv6_masked_addr_cmp(&ipv6->daddr, &ip6info->dmsk,
  66                                          &ip6info->dst)))
  67                 return false;
  68
  69         ret = ifname_compare_aligned(indev, ip6info->iniface, ip6info->iniface_mask);
  70
  71         if (NF_INVF(ip6info, IP6T_INV_VIA_IN, ret != 0))
  72                 return false;
  73
  74         ret = ifname_compare_aligned(outdev, ip6info->outiface, ip6info->outiface_mask);
  75
  76         if (NF_INVF(ip6info, IP6T_INV_VIA_OUT, ret != 0))
  77                 return false;
  78
  79 /* ... might want to do something with class and flowlabel here ... */
  80
  81         /* look for the desired protocol header */
  82         if (ip6info->flags & IP6T_F_PROTO) {
  83                 int protohdr;
  84                 unsigned short _frag_off;
  85
  86                 protohdr = ipv6_find_hdr(skb, protoff, -1, &_frag_off, NULL);
  87                 if (protohdr < 0) {
  88                         if (_frag_off == 0)
  89                                 *hotdrop = true;
  90                         return false;
  91                 }
  92                 *fragoff = _frag_off;
  93
  94                 if (ip6info->proto == protohdr) {
  95                         if (ip6info->invflags & IP6T_INV_PROTO)
  96                                 return false;
  97
  98                         return true;
  99                 }
 100
 101                 /* We need match for the '-p all', too! */
 102                 if ((ip6info->proto != 0) &&
 103                         !(ip6info->invflags & IP6T_INV_PROTO))
 104                         return false;
 105         }
 106         return true;
 107 }
 108
 109 /* should be ip6 safe */
 110 static bool
 111 ip6_checkentry(const struct ip6t_ip6 *ipv6)
 112 {
 113         if (ipv6->flags & ~IP6T_F_MASK)
 114                 return false;
 115         if (ipv6->invflags & ~IP6T_INV_MASK)
 116                 return false;
 117
 118         return true;
 119 }
 120
 121 static unsigned int
 122 ip6t_error(struct sk_buff *skb, const struct xt_action_param *par)
 123 {
 124         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 125
 126         return NF_DROP;
 127 }
 128
 129 static inline struct ip6t_entry *
 130 get_entry(const void *base, unsigned int offset)
 131 {
 132         return (struct ip6t_entry *)(base + offset);
 133 }
 134
 135 /* All zeroes == unconditional rule. */
 136 /* Mildly perf critical (only if packet tracing is on) */
 137 static inline bool unconditional(const struct ip6t_entry *e)
 138 {
 139         static const struct ip6t_ip6 uncond;
 140
 141         return e->target_offset == sizeof(struct ip6t_entry) &&
 142                memcmp(&e->ipv6, &uncond, sizeof(uncond)) == 0;
 143 }
 144
 145 static inline const struct xt_entry_target *
 146 ip6t_get_target_c(const struct ip6t_entry *e)
 147 {
 148         return ip6t_get_target((struct ip6t_entry *)e);
 149 }
 150
 151 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 152 /* This cries for unification! */
 153 static const char *const hooknames[] = {
 154         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 155         [NF_INET_LOCAL_IN]              = "INPUT",
 156         [NF_INET_FORWARD]               = "FORWARD",
 157         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 158         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 159 };
 160
 161 enum nf_ip_trace_comments {
 162         NF_IP6_TRACE_COMMENT_RULE,
 163         NF_IP6_TRACE_COMMENT_RETURN,
 164         NF_IP6_TRACE_COMMENT_POLICY,
 165 };
 166
 167 static const char *const comments[] = {
 168         [NF_IP6_TRACE_COMMENT_RULE]     = "rule",
 169         [NF_IP6_TRACE_COMMENT_RETURN]   = "return",
 170         [NF_IP6_TRACE_COMMENT_POLICY]   = "policy",
 171 };
 172
 173 static const struct nf_loginfo trace_loginfo = {
 174         .type = NF_LOG_TYPE_LOG,
 175         .u = {
 176                 .log = {
 177                         .level = LOGLEVEL_WARNING,
 178                         .logflags = NF_LOG_DEFAULT_MASK,
 179                 },
 180         },
 181 };
 182
 183 /* Mildly perf critical (only if packet tracing is on) */
 184 static inline int
 185 get_chainname_rulenum(const struct ip6t_entry *s, const struct ip6t_entry *e,
 186                       const char *hookname, const char **chainname,
 187                       const char **comment, unsigned int *rulenum)
 188 {
 189         const struct xt_standard_target *t = (void *)ip6t_get_target_c(s);
 190
 191         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 192                 /* Head of user chain: ERROR target with chainname */
 193                 *chainname = t->target.data;
 194                 (*rulenum) = 0;
 195         } else if (s == e) {
 196                 (*rulenum)++;
 197
 198                 if (unconditional(s) &&
 199                     strcmp(t->target.u.kernel.target->name,
 200                            XT_STANDARD_TARGET) == 0 &&
 201                     t->verdict < 0) {
 202                         /* Tail of chains: STANDARD target (return/policy) */
 203                         *comment = *chainname == hookname
 204                                 ? comments[NF_IP6_TRACE_COMMENT_POLICY]
 205                                 : comments[NF_IP6_TRACE_COMMENT_RETURN];
 206                 }
 207                 return 1;
 208         } else
 209                 (*rulenum)++;
 210
 211         return 0;
 212 }
 213
 214 static void trace_packet(struct net *net,
 215                          const struct sk_buff *skb,
 216                          unsigned int hook,
 217                          const struct net_device *in,
 218                          const struct net_device *out,
 219                          const char *tablename,
 220                          const struct xt_table_info *private,
 221                          const struct ip6t_entry *e)
 222 {
 223         const struct ip6t_entry *root;
 224         const char *hookname, *chainname, *comment;
 225         const struct ip6t_entry *iter;
 226         unsigned int rulenum = 0;
 227
 228         root = get_entry(private->entries, private->hook_entry[hook]);
 229
 230         hookname = chainname = hooknames[hook];
 231         comment = comments[NF_IP6_TRACE_COMMENT_RULE];
 232
 233         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 234                 if (get_chainname_rulenum(iter, e, hookname,
 235                     &chainname, &comment, &rulenum) != 0)
 236                         break;
 237
 238         nf_log_trace(net, AF_INET6, hook, skb, in, out, &trace_loginfo,
 239                      "TRACE: %s:%s:%s:%u ",
 240                      tablename, chainname, comment, rulenum);
 241 }
 242 #endif
 243
 244 static inline struct ip6t_entry *
 245 ip6t_next_entry(const struct ip6t_entry *entry)
 246 {
 247         return (void *)entry + entry->next_offset;
 248 }
 249
 250 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 251 unsigned int
 252 ip6t_do_table(struct sk_buff *skb,
 253               const struct nf_hook_state *state,
 254               struct xt_table *table)
 255 {
 256         unsigned int hook = state->hook;
 257         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 258         /* Initializing verdict to NF_DROP keeps gcc happy. */
 259         unsigned int verdict = NF_DROP;
 260         const char *indev, *outdev;
 261         const void *table_base;
 262         struct ip6t_entry *e, **jumpstack;
 263         unsigned int stackidx, cpu;
 264         const struct xt_table_info *private;
 265         struct xt_action_param acpar;
 266         unsigned int addend;
 267
 268         /* Initialization */
 269         stackidx = 0;
 270         indev = state->in ? state->in->name : nulldevname;
 271         outdev = state->out ? state->out->name : nulldevname;
 272         /* We handle fragments by dealing with the first fragment as
 273          * if it was a normal packet.  All other fragments are treated
 274          * normally, except that they will NEVER match rules that ask
 275          * things we don't know, ie. tcp syn flag or ports).  If the
 276          * rule is also a fragment-specific rule, non-fragments won't
 277          * match it. */
 278         acpar.hotdrop = false;
 279         acpar.state   = state;
 280
 281         WARN_ON(!(table->valid_hooks & (1 << hook)));
 282
 283         local_bh_disable();
 284         addend = xt_write_recseq_begin();
 285         private = READ_ONCE(table->private); /* Address dependency. */
 286         cpu        = smp_processor_id();
 287         table_base = private->entries;
 288         jumpstack  = (struct ip6t_entry **)private->jumpstack[cpu];
 289
 290         /* Switch to alternate jumpstack if we're being invoked via TEE.
 291          * TEE issues XT_CONTINUE verdict on original skb so we must not
 292          * clobber the jumpstack.
 293          *
 294          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 295          * but it is no problem since absolute verdict is issued by these.
 296          */
 297         if (static_key_false(&xt_tee_enabled))
 298                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 299
 300         e = get_entry(table_base, private->hook_entry[hook]);
 301
 302         do {
 303                 const struct xt_entry_target *t;
 304                 const struct xt_entry_match *ematch;
 305                 struct xt_counters *counter;
 306
 307                 WARN_ON(!e);
 308                 acpar.thoff = 0;
 309                 if (!ip6_packet_match(skb, indev, outdev, &e->ipv6,
 310                     &acpar.thoff, &acpar.fragoff, &acpar.hotdrop)) {
 311  no_match:
 312                         e = ip6t_next_entry(e);
 313                         continue;
 314                 }
 315
 316                 xt_ematch_foreach(ematch, e) {
 317                         acpar.match     = ematch->u.kernel.match;
 318                         acpar.matchinfo = ematch->data;
 319                         if (!acpar.match->match(skb, &acpar))
 320                                 goto no_match;
 321                 }
 322
 323                 counter = xt_get_this_cpu_counter(&e->counters);
 324                 ADD_COUNTER(*counter, skb->len, 1);
 325
 326                 t = ip6t_get_target_c(e);
 327                 WARN_ON(!t->u.kernel.target);
 328
 329 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 330                 /* The packet is traced: log it */
 331                 if (unlikely(skb->nf_trace))
 332                         trace_packet(state->net, skb, hook, state->in,
 333                                      state->out, table->name, private, e);
 334 #endif
 335                 /* Standard target? */
 336                 if (!t->u.kernel.target->target) {
 337                         int v;
 338
 339                         v = ((struct xt_standard_target *)t)->verdict;
 340                         if (v < 0) {
 341                                 /* Pop from stack? */
 342                                 if (v != XT_RETURN) {
 343                                         verdict = (unsigned int)(-v) - 1;
 344                                         break;
 345                                 }
 346                                 if (stackidx == 0)
 347                                         e = get_entry(table_base,
 348                                             private->underflow[hook]);
 349                                 else
 350                                         e = ip6t_next_entry(jumpstack[--stackidx]);
 351                                 continue;
 352                         }
 353                         if (table_base + v != ip6t_next_entry(e) &&
 354                             !(e->ipv6.flags & IP6T_F_GOTO)) {
 355                                 jumpstack[stackidx++] = e;
 356                         }
 357
 358                         e = get_entry(table_base, v);
 359                         continue;
 360                 }
 361
 362                 acpar.target   = t->u.kernel.target;
 363                 acpar.targinfo = t->data;
 364
 365                 verdict = t->u.kernel.target->target(skb, &acpar);
 366                 if (verdict == XT_CONTINUE)
 367                         e = ip6t_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 ip6t_entry *e = 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 *)ip6t_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 = 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         t = ip6t_get_target(e);
 533         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 534                                e->ipv6.proto,
 535                                e->ipv6.invflags & IP6T_INV_PROTO);
 536 }
 537
 538 static int
 539 find_check_entry(struct ip6t_entry *e, struct net *net, const char *name,
 540                  unsigned int size,
 541                  struct xt_percpu_counter_alloc_state *alloc_state)
 542 {
 543         struct xt_entry_target *t;
 544         struct xt_target *target;
 545         int ret;
 546         unsigned int j;
 547         struct xt_mtchk_param mtpar;
 548         struct xt_entry_match *ematch;
 549
 550         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 551                 return -ENOMEM;
 552
 553         j = 0;
 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         /* Check hooks all assigned */
 725         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 726                 /* Only hooks which are valid */
 727                 if (!(repl->valid_hooks & (1 << i)))
 728                         continue;
 729                 if (newinfo->hook_entry[i] == 0xFFFFFFFF)
 730                         goto out_free;
 731                 if (newinfo->underflow[i] == 0xFFFFFFFF)
 732                         goto out_free;
 733         }
 734
 735         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 736                 ret = -ELOOP;
 737                 goto out_free;
 738         }
 739         kvfree(offsets);
 740
 741         /* Finally, each sanity check must pass */
 742         i = 0;
 743         xt_entry_foreach(iter, entry0, newinfo->size) {
 744                 ret = find_check_entry(iter, net, repl->name, repl->size,
 745                                        &alloc_state);
 746                 if (ret != 0)
 747                         break;
 748                 ++i;
 749         }
 750
 751         if (ret != 0) {
 752                 xt_entry_foreach(iter, entry0, newinfo->size) {
 753                         if (i-- == 0)
 754                                 break;
 755                         cleanup_entry(iter, net);
 756                 }
 757                 return ret;
 758         }
 759
 760         return ret;
 761  out_free:
 762         kvfree(offsets);
 763         return ret;
 764 }
 765
 766 static void
 767 get_counters(const struct xt_table_info *t,
 768              struct xt_counters counters[])
 769 {
 770         struct ip6t_entry *iter;
 771         unsigned int cpu;
 772         unsigned int i;
 773
 774         for_each_possible_cpu(cpu) {
 775                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 776
 777                 i = 0;
 778                 xt_entry_foreach(iter, t->entries, t->size) {
 779                         struct xt_counters *tmp;
 780                         u64 bcnt, pcnt;
 781                         unsigned int start;
 782
 783                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 784                         do {
 785                                 start = read_seqcount_begin(s);
 786                                 bcnt = tmp->bcnt;
 787                                 pcnt = tmp->pcnt;
 788                         } while (read_seqcount_retry(s, start));
 789
 790                         ADD_COUNTER(counters[i], bcnt, pcnt);
 791                         ++i;
 792                         cond_resched();
 793                 }
 794         }
 795 }
 796
 797 static void get_old_counters(const struct xt_table_info *t,
 798                              struct xt_counters counters[])
 799 {
 800         struct ip6t_entry *iter;
 801         unsigned int cpu, i;
 802
 803         for_each_possible_cpu(cpu) {
 804                 i = 0;
 805                 xt_entry_foreach(iter, t->entries, t->size) {
 806                         const struct xt_counters *tmp;
 807
 808                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 809                         ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
 810                         ++i;
 811                 }
 812                 cond_resched();
 813         }
 814 }
 815
 816 static struct xt_counters *alloc_counters(const struct xt_table *table)
 817 {
 818         unsigned int countersize;
 819         struct xt_counters *counters;
 820         const struct xt_table_info *private = table->private;
 821
 822         /* We need atomic snapshot of counters: rest doesn't change
 823            (other than comefrom, which userspace doesn't care
 824            about). */
 825         countersize = sizeof(struct xt_counters) * private->number;
 826         counters = vzalloc(countersize);
 827
 828         if (counters == NULL)
 829                 return ERR_PTR(-ENOMEM);
 830
 831         get_counters(private, counters);
 832
 833         return counters;
 834 }
 835
 836 static int
 837 copy_entries_to_user(unsigned int total_size,
 838                      const struct xt_table *table,
 839                      void __user *userptr)
 840 {
 841         unsigned int off, num;
 842         const struct ip6t_entry *e;
 843         struct xt_counters *counters;
 844         const struct xt_table_info *private = table->private;
 845         int ret = 0;
 846         const void *loc_cpu_entry;
 847
 848         counters = alloc_counters(table);
 849         if (IS_ERR(counters))
 850                 return PTR_ERR(counters);
 851
 852         loc_cpu_entry = private->entries;
 853
 854         /* FIXME: use iterator macros --RR */
 855         /* ... then go back and fix counters and names */
 856         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 857                 unsigned int i;
 858                 const struct xt_entry_match *m;
 859                 const struct xt_entry_target *t;
 860
 861                 e = loc_cpu_entry + off;
 862                 if (copy_to_user(userptr + off, e, sizeof(*e))) {
 863                         ret = -EFAULT;
 864                         goto free_counters;
 865                 }
 866                 if (copy_to_user(userptr + off
 867                                  + offsetof(struct ip6t_entry, counters),
 868                                  &counters[num],
 869                                  sizeof(counters[num])) != 0) {
 870                         ret = -EFAULT;
 871                         goto free_counters;
 872                 }
 873
 874                 for (i = sizeof(struct ip6t_entry);
 875                      i < e->target_offset;
 876                      i += m->u.match_size) {
 877                         m = (void *)e + i;
 878
 879                         if (xt_match_to_user(m, userptr + off + i)) {
 880                                 ret = -EFAULT;
 881                                 goto free_counters;
 882                         }
 883                 }
 884
 885                 t = ip6t_get_target_c(e);
 886                 if (xt_target_to_user(t, userptr + off + e->target_offset)) {
 887                         ret = -EFAULT;
 888                         goto free_counters;
 889                 }
 890         }
 891
 892  free_counters:
 893         vfree(counters);
 894         return ret;
 895 }
 896
 897 #ifdef CONFIG_COMPAT
 898 static void compat_standard_from_user(void *dst, const void *src)
 899 {
 900         int v = *(compat_int_t *)src;
 901
 902         if (v > 0)
 903                 v += xt_compat_calc_jump(AF_INET6, v);
 904         memcpy(dst, &v, sizeof(v));
 905 }
 906
 907 static int compat_standard_to_user(void __user *dst, const void *src)
 908 {
 909         compat_int_t cv = *(int *)src;
 910
 911         if (cv > 0)
 912                 cv -= xt_compat_calc_jump(AF_INET6, cv);
 913         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 914 }
 915
 916 static int compat_calc_entry(const struct ip6t_entry *e,
 917                              const struct xt_table_info *info,
 918                              const void *base, struct xt_table_info *newinfo)
 919 {
 920         const struct xt_entry_match *ematch;
 921         const struct xt_entry_target *t;
 922         unsigned int entry_offset;
 923         int off, i, ret;
 924
 925         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
 926         entry_offset = (void *)e - base;
 927         xt_ematch_foreach(ematch, e)
 928                 off += xt_compat_match_offset(ematch->u.kernel.match);
 929         t = ip6t_get_target_c(e);
 930         off += xt_compat_target_offset(t->u.kernel.target);
 931         newinfo->size -= off;
 932         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
 933         if (ret)
 934                 return ret;
 935
 936         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 937                 if (info->hook_entry[i] &&
 938                     (e < (struct ip6t_entry *)(base + info->hook_entry[i])))
 939                         newinfo->hook_entry[i] -= off;
 940                 if (info->underflow[i] &&
 941                     (e < (struct ip6t_entry *)(base + info->underflow[i])))
 942                         newinfo->underflow[i] -= off;
 943         }
 944         return 0;
 945 }
 946
 947 static int compat_table_info(const struct xt_table_info *info,
 948                              struct xt_table_info *newinfo)
 949 {
 950         struct ip6t_entry *iter;
 951         const void *loc_cpu_entry;
 952         int ret;
 953
 954         if (!newinfo || !info)
 955                 return -EINVAL;
 956
 957         /* we dont care about newinfo->entries */
 958         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 959         newinfo->initial_entries = 0;
 960         loc_cpu_entry = info->entries;
 961         xt_compat_init_offsets(AF_INET6, info->number);
 962         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 963                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 964                 if (ret != 0)
 965                         return ret;
 966         }
 967         return 0;
 968 }
 969 #endif
 970
 971 static int get_info(struct net *net, void __user *user,
 972                     const int *len, int compat)
 973 {
 974         char name[XT_TABLE_MAXNAMELEN];
 975         struct xt_table *t;
 976         int ret;
 977
 978         if (*len != sizeof(struct ip6t_getinfo))
 979                 return -EINVAL;
 980
 981         if (copy_from_user(name, user, sizeof(name)) != 0)
 982                 return -EFAULT;
 983
 984         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 985 #ifdef CONFIG_COMPAT
 986         if (compat)
 987                 xt_compat_lock(AF_INET6);
 988 #endif
 989         t = xt_request_find_table_lock(net, AF_INET6, name);
 990         if (!IS_ERR(t)) {
 991                 struct ip6t_getinfo info;
 992                 const struct xt_table_info *private = t->private;
 993 #ifdef CONFIG_COMPAT
 994                 struct xt_table_info tmp;
 995
 996                 if (compat) {
 997                         ret = compat_table_info(private, &tmp);
 998                         xt_compat_flush_offsets(AF_INET6);
 999                         private = &tmp;
1000                 }
1001 #endif
1002                 memset(&info, 0, sizeof(info));
1003                 info.valid_hooks = t->valid_hooks;
1004                 memcpy(info.hook_entry, private->hook_entry,
1005                        sizeof(info.hook_entry));
1006                 memcpy(info.underflow, private->underflow,
1007                        sizeof(info.underflow));
1008                 info.num_entries = private->number;
1009                 info.size = private->size;
1010                 strcpy(info.name, name);
1011
1012                 if (copy_to_user(user, &info, *len) != 0)
1013                         ret = -EFAULT;
1014                 else
1015                         ret = 0;
1016
1017                 xt_table_unlock(t);
1018                 module_put(t->me);
1019         } else
1020                 ret = PTR_ERR(t);
1021 #ifdef CONFIG_COMPAT
1022         if (compat)
1023                 xt_compat_unlock(AF_INET6);
1024 #endif
1025         return ret;
1026 }
1027
1028 static int
1029 get_entries(struct net *net, struct ip6t_get_entries __user *uptr,
1030             const int *len)
1031 {
1032         int ret;
1033         struct ip6t_get_entries get;
1034         struct xt_table *t;
1035
1036         if (*len < sizeof(get))
1037                 return -EINVAL;
1038         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1039                 return -EFAULT;
1040         if (*len != sizeof(struct ip6t_get_entries) + get.size)
1041                 return -EINVAL;
1042
1043         get.name[sizeof(get.name) - 1] = '\0';
1044
1045         t = xt_find_table_lock(net, AF_INET6, get.name);
1046         if (!IS_ERR(t)) {
1047                 struct xt_table_info *private = t->private;
1048                 if (get.size == private->size)
1049                         ret = copy_entries_to_user(private->size,
1050                                                    t, uptr->entrytable);
1051                 else
1052                         ret = -EAGAIN;
1053
1054                 module_put(t->me);
1055                 xt_table_unlock(t);
1056         } else
1057                 ret = PTR_ERR(t);
1058
1059         return ret;
1060 }
1061
1062 static int
1063 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1064              struct xt_table_info *newinfo, unsigned int num_counters,
1065              void __user *counters_ptr)
1066 {
1067         int ret;
1068         struct xt_table *t;
1069         struct xt_table_info *oldinfo;
1070         struct xt_counters *counters;
1071         struct ip6t_entry *iter;
1072
1073         ret = 0;
1074         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1075         if (!counters) {
1076                 ret = -ENOMEM;
1077                 goto out;
1078         }
1079
1080         t = xt_request_find_table_lock(net, AF_INET6, name);
1081         if (IS_ERR(t)) {
1082                 ret = PTR_ERR(t);
1083                 goto free_newinfo_counters_untrans;
1084         }
1085
1086         /* You lied! */
1087         if (valid_hooks != t->valid_hooks) {
1088                 ret = -EINVAL;
1089                 goto put_module;
1090         }
1091
1092         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1093         if (!oldinfo)
1094                 goto put_module;
1095
1096         /* Update module usage count based on number of rules */
1097         if ((oldinfo->number > oldinfo->initial_entries) ||
1098             (newinfo->number <= oldinfo->initial_entries))
1099                 module_put(t->me);
1100         if ((oldinfo->number > oldinfo->initial_entries) &&
1101             (newinfo->number <= oldinfo->initial_entries))
1102                 module_put(t->me);
1103
1104         get_old_counters(oldinfo, counters);
1105
1106         /* Decrease module usage counts and free resource */
1107         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1108                 cleanup_entry(iter, net);
1109
1110         xt_free_table_info(oldinfo);
1111         if (copy_to_user(counters_ptr, counters,
1112                          sizeof(struct xt_counters) * num_counters) != 0) {
1113                 /* Silent error, can't fail, new table is already in place */
1114                 net_warn_ratelimited("ip6tables: counters copy to user failed while replacing table\n");
1115         }
1116         vfree(counters);
1117         xt_table_unlock(t);
1118         return ret;
1119
1120  put_module:
1121         module_put(t->me);
1122         xt_table_unlock(t);
1123  free_newinfo_counters_untrans:
1124         vfree(counters);
1125  out:
1126         return ret;
1127 }
1128
1129 static int
1130 do_replace(struct net *net, const void __user *user, unsigned int len)
1131 {
1132         int ret;
1133         struct ip6t_replace tmp;
1134         struct xt_table_info *newinfo;
1135         void *loc_cpu_entry;
1136         struct ip6t_entry *iter;
1137
1138         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1139                 return -EFAULT;
1140
1141         /* overflow check */
1142         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1143                 return -ENOMEM;
1144         if (tmp.num_counters == 0)
1145                 return -EINVAL;
1146
1147         tmp.name[sizeof(tmp.name)-1] = 0;
1148
1149         newinfo = xt_alloc_table_info(tmp.size);
1150         if (!newinfo)
1151                 return -ENOMEM;
1152
1153         loc_cpu_entry = newinfo->entries;
1154         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1155                            tmp.size) != 0) {
1156                 ret = -EFAULT;
1157                 goto free_newinfo;
1158         }
1159
1160         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1161         if (ret != 0)
1162                 goto free_newinfo;
1163
1164         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1165                            tmp.num_counters, tmp.counters);
1166         if (ret)
1167                 goto free_newinfo_untrans;
1168         return 0;
1169
1170  free_newinfo_untrans:
1171         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1172                 cleanup_entry(iter, net);
1173  free_newinfo:
1174         xt_free_table_info(newinfo);
1175         return ret;
1176 }
1177
1178 static int
1179 do_add_counters(struct net *net, const void __user *user, unsigned int len,
1180                 int compat)
1181 {
1182         unsigned int i;
1183         struct xt_counters_info tmp;
1184         struct xt_counters *paddc;
1185         struct xt_table *t;
1186         const struct xt_table_info *private;
1187         int ret = 0;
1188         struct ip6t_entry *iter;
1189         unsigned int addend;
1190
1191         paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1192         if (IS_ERR(paddc))
1193                 return PTR_ERR(paddc);
1194         t = xt_find_table_lock(net, AF_INET6, tmp.name);
1195         if (IS_ERR(t)) {
1196                 ret = PTR_ERR(t);
1197                 goto free;
1198         }
1199
1200         local_bh_disable();
1201         private = t->private;
1202         if (private->number != tmp.num_counters) {
1203                 ret = -EINVAL;
1204                 goto unlock_up_free;
1205         }
1206
1207         i = 0;
1208         addend = xt_write_recseq_begin();
1209         xt_entry_foreach(iter, private->entries, private->size) {
1210                 struct xt_counters *tmp;
1211
1212                 tmp = xt_get_this_cpu_counter(&iter->counters);
1213                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1214                 ++i;
1215         }
1216         xt_write_recseq_end(addend);
1217  unlock_up_free:
1218         local_bh_enable();
1219         xt_table_unlock(t);
1220         module_put(t->me);
1221  free:
1222         vfree(paddc);
1223
1224         return ret;
1225 }
1226
1227 #ifdef CONFIG_COMPAT
1228 struct compat_ip6t_replace {
1229         char                    name[XT_TABLE_MAXNAMELEN];
1230         u32                     valid_hooks;
1231         u32                     num_entries;
1232         u32                     size;
1233         u32                     hook_entry[NF_INET_NUMHOOKS];
1234         u32                     underflow[NF_INET_NUMHOOKS];
1235         u32                     num_counters;
1236         compat_uptr_t           counters;       /* struct xt_counters * */
1237         struct compat_ip6t_entry entries[0];
1238 };
1239
1240 static int
1241 compat_copy_entry_to_user(struct ip6t_entry *e, void __user **dstptr,
1242                           unsigned int *size, struct xt_counters *counters,
1243                           unsigned int i)
1244 {
1245         struct xt_entry_target *t;
1246         struct compat_ip6t_entry __user *ce;
1247         u_int16_t target_offset, next_offset;
1248         compat_uint_t origsize;
1249         const struct xt_entry_match *ematch;
1250         int ret = 0;
1251
1252         origsize = *size;
1253         ce = *dstptr;
1254         if (copy_to_user(ce, e, sizeof(struct ip6t_entry)) != 0 ||
1255             copy_to_user(&ce->counters, &counters[i],
1256             sizeof(counters[i])) != 0)
1257                 return -EFAULT;
1258
1259         *dstptr += sizeof(struct compat_ip6t_entry);
1260         *size -= sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1261
1262         xt_ematch_foreach(ematch, e) {
1263                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1264                 if (ret != 0)
1265                         return ret;
1266         }
1267         target_offset = e->target_offset - (origsize - *size);
1268         t = ip6t_get_target(e);
1269         ret = xt_compat_target_to_user(t, dstptr, size);
1270         if (ret)
1271                 return ret;
1272         next_offset = e->next_offset - (origsize - *size);
1273         if (put_user(target_offset, &ce->target_offset) != 0 ||
1274             put_user(next_offset, &ce->next_offset) != 0)
1275                 return -EFAULT;
1276         return 0;
1277 }
1278
1279 static int
1280 compat_find_calc_match(struct xt_entry_match *m,
1281                        const struct ip6t_ip6 *ipv6,
1282                        int *size)
1283 {
1284         struct xt_match *match;
1285
1286         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
1287                                       m->u.user.revision);
1288         if (IS_ERR(match))
1289                 return PTR_ERR(match);
1290
1291         m->u.kernel.match = match;
1292         *size += xt_compat_match_offset(match);
1293         return 0;
1294 }
1295
1296 static void compat_release_entry(struct compat_ip6t_entry *e)
1297 {
1298         struct xt_entry_target *t;
1299         struct xt_entry_match *ematch;
1300
1301         /* Cleanup all matches */
1302         xt_ematch_foreach(ematch, e)
1303                 module_put(ematch->u.kernel.match->me);
1304         t = compat_ip6t_get_target(e);
1305         module_put(t->u.kernel.target->me);
1306 }
1307
1308 static int
1309 check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
1310                                   struct xt_table_info *newinfo,
1311                                   unsigned int *size,
1312                                   const unsigned char *base,
1313                                   const unsigned char *limit)
1314 {
1315         struct xt_entry_match *ematch;
1316         struct xt_entry_target *t;
1317         struct xt_target *target;
1318         unsigned int entry_offset;
1319         unsigned int j;
1320         int ret, off;
1321
1322         if ((unsigned long)e % __alignof__(struct compat_ip6t_entry) != 0 ||
1323             (unsigned char *)e + sizeof(struct compat_ip6t_entry) >= limit ||
1324             (unsigned char *)e + e->next_offset > limit)
1325                 return -EINVAL;
1326
1327         if (e->next_offset < sizeof(struct compat_ip6t_entry) +
1328                              sizeof(struct compat_xt_entry_target))
1329                 return -EINVAL;
1330
1331         if (!ip6_checkentry(&e->ipv6))
1332                 return -EINVAL;
1333
1334         ret = xt_compat_check_entry_offsets(e, e->elems,
1335                                             e->target_offset, e->next_offset);
1336         if (ret)
1337                 return ret;
1338
1339         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1340         entry_offset = (void *)e - (void *)base;
1341         j = 0;
1342         xt_ematch_foreach(ematch, e) {
1343                 ret = compat_find_calc_match(ematch, &e->ipv6, &off);
1344                 if (ret != 0)
1345                         goto release_matches;
1346                 ++j;
1347         }
1348
1349         t = compat_ip6t_get_target(e);
1350         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
1351                                         t->u.user.revision);
1352         if (IS_ERR(target)) {
1353                 ret = PTR_ERR(target);
1354                 goto release_matches;
1355         }
1356         t->u.kernel.target = target;
1357
1358         off += xt_compat_target_offset(target);
1359         *size += off;
1360         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1361         if (ret)
1362                 goto out;
1363
1364         return 0;
1365
1366 out:
1367         module_put(t->u.kernel.target->me);
1368 release_matches:
1369         xt_ematch_foreach(ematch, e) {
1370                 if (j-- == 0)
1371                         break;
1372                 module_put(ematch->u.kernel.match->me);
1373         }
1374         return ret;
1375 }
1376
1377 static void
1378 compat_copy_entry_from_user(struct compat_ip6t_entry *e, void **dstptr,
1379                             unsigned int *size,
1380                             struct xt_table_info *newinfo, unsigned char *base)
1381 {
1382         struct xt_entry_target *t;
1383         struct ip6t_entry *de;
1384         unsigned int origsize;
1385         int h;
1386         struct xt_entry_match *ematch;
1387
1388         origsize = *size;
1389         de = *dstptr;
1390         memcpy(de, e, sizeof(struct ip6t_entry));
1391         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1392
1393         *dstptr += sizeof(struct ip6t_entry);
1394         *size += sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1395
1396         xt_ematch_foreach(ematch, e)
1397                 xt_compat_match_from_user(ematch, dstptr, size);
1398
1399         de->target_offset = e->target_offset - (origsize - *size);
1400         t = compat_ip6t_get_target(e);
1401         xt_compat_target_from_user(t, dstptr, size);
1402
1403         de->next_offset = e->next_offset - (origsize - *size);
1404         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1405                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1406                         newinfo->hook_entry[h] -= origsize - *size;
1407                 if ((unsigned char *)de - base < newinfo->underflow[h])
1408                         newinfo->underflow[h] -= origsize - *size;
1409         }
1410 }
1411
1412 static int
1413 translate_compat_table(struct net *net,
1414                        struct xt_table_info **pinfo,
1415                        void **pentry0,
1416                        const struct compat_ip6t_replace *compatr)
1417 {
1418         unsigned int i, j;
1419         struct xt_table_info *newinfo, *info;
1420         void *pos, *entry0, *entry1;
1421         struct compat_ip6t_entry *iter0;
1422         struct ip6t_replace repl;
1423         unsigned int size;
1424         int ret = 0;
1425
1426         info = *pinfo;
1427         entry0 = *pentry0;
1428         size = compatr->size;
1429         info->number = compatr->num_entries;
1430
1431         j = 0;
1432         xt_compat_lock(AF_INET6);
1433         xt_compat_init_offsets(AF_INET6, compatr->num_entries);
1434         /* Walk through entries, checking offsets. */
1435         xt_entry_foreach(iter0, entry0, compatr->size) {
1436                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1437                                                         entry0,
1438                                                         entry0 + compatr->size);
1439                 if (ret != 0)
1440                         goto out_unlock;
1441                 ++j;
1442         }
1443
1444         ret = -EINVAL;
1445         if (j != compatr->num_entries)
1446                 goto out_unlock;
1447
1448         ret = -ENOMEM;
1449         newinfo = xt_alloc_table_info(size);
1450         if (!newinfo)
1451                 goto out_unlock;
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
1800         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1801         if (ret != 0) {
1802                 __ip6t_unregister_table(net, new_table);
1803                 *res = NULL;
1804         }
1805
1806         return ret;
1807
1808 out_free:
1809         xt_free_table_info(newinfo);
1810         return ret;
1811 }
1812
1813 void ip6t_unregister_table(struct net *net, struct xt_table *table,
1814                            const struct nf_hook_ops *ops)
1815 {
1816         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1817         __ip6t_unregister_table(net, table);
1818 }
1819
1820 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1821 static inline bool
1822 icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1823                      u_int8_t type, u_int8_t code,
1824                      bool invert)
1825 {
1826         return (type == test_type && code >= min_code && code <= max_code)
1827                 ^ invert;
1828 }
1829
1830 static bool
1831 icmp6_match(const struct sk_buff *skb, struct xt_action_param *par)
1832 {
1833         const struct icmp6hdr *ic;
1834         struct icmp6hdr _icmph;
1835         const struct ip6t_icmp *icmpinfo = par->matchinfo;
1836
1837         /* Must not be a fragment. */
1838         if (par->fragoff != 0)
1839                 return false;
1840
1841         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1842         if (ic == NULL) {
1843                 /* We've been asked to examine this packet, and we
1844                  * can't.  Hence, no choice but to drop.
1845                  */
1846                 par->hotdrop = true;
1847                 return false;
1848         }
1849
1850         return icmp6_type_code_match(icmpinfo->type,
1851                                      icmpinfo->code[0],
1852                                      icmpinfo->code[1],
1853                                      ic->icmp6_type, ic->icmp6_code,
1854                                      !!(icmpinfo->invflags&IP6T_ICMP_INV));
1855 }
1856
1857 /* Called when user tries to insert an entry of this type. */
1858 static int icmp6_checkentry(const struct xt_mtchk_param *par)
1859 {
1860         const struct ip6t_icmp *icmpinfo = par->matchinfo;
1861
1862         /* Must specify no unknown invflags */
1863         return (icmpinfo->invflags & ~IP6T_ICMP_INV) ? -EINVAL : 0;
1864 }
1865
1866 /* The built-in targets: standard (NULL) and error. */
1867 static struct xt_target ip6t_builtin_tg[] __read_mostly = {
1868         {
1869                 .name             = XT_STANDARD_TARGET,
1870                 .targetsize       = sizeof(int),
1871                 .family           = NFPROTO_IPV6,
1872 #ifdef CONFIG_COMPAT
1873                 .compatsize       = sizeof(compat_int_t),
1874                 .compat_from_user = compat_standard_from_user,
1875                 .compat_to_user   = compat_standard_to_user,
1876 #endif
1877         },
1878         {
1879                 .name             = XT_ERROR_TARGET,
1880                 .target           = ip6t_error,
1881                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1882                 .family           = NFPROTO_IPV6,
1883         },
1884 };
1885
1886 static struct nf_sockopt_ops ip6t_sockopts = {
1887         .pf             = PF_INET6,
1888         .set_optmin     = IP6T_BASE_CTL,
1889         .set_optmax     = IP6T_SO_SET_MAX+1,
1890         .set            = do_ip6t_set_ctl,
1891 #ifdef CONFIG_COMPAT
1892         .compat_set     = compat_do_ip6t_set_ctl,
1893 #endif
1894         .get_optmin     = IP6T_BASE_CTL,
1895         .get_optmax     = IP6T_SO_GET_MAX+1,
1896         .get            = do_ip6t_get_ctl,
1897 #ifdef CONFIG_COMPAT
1898         .compat_get     = compat_do_ip6t_get_ctl,
1899 #endif
1900         .owner          = THIS_MODULE,
1901 };
1902
1903 static struct xt_match ip6t_builtin_mt[] __read_mostly = {
1904         {
1905                 .name       = "icmp6",
1906                 .match      = icmp6_match,
1907                 .matchsize  = sizeof(struct ip6t_icmp),
1908                 .checkentry = icmp6_checkentry,
1909                 .proto      = IPPROTO_ICMPV6,
1910                 .family     = NFPROTO_IPV6,
1911         },
1912 };
1913
1914 static int __net_init ip6_tables_net_init(struct net *net)
1915 {
1916         return xt_proto_init(net, NFPROTO_IPV6);
1917 }
1918
1919 static void __net_exit ip6_tables_net_exit(struct net *net)
1920 {
1921         xt_proto_fini(net, NFPROTO_IPV6);
1922 }
1923
1924 static struct pernet_operations ip6_tables_net_ops = {
1925         .init = ip6_tables_net_init,
1926         .exit = ip6_tables_net_exit,
1927 };
1928
1929 static int __init ip6_tables_init(void)
1930 {
1931         int ret;
1932
1933         ret = register_pernet_subsys(&ip6_tables_net_ops);
1934         if (ret < 0)
1935                 goto err1;
1936
1937         /* No one else will be downing sem now, so we won't sleep */
1938         ret = xt_register_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1939         if (ret < 0)
1940                 goto err2;
1941         ret = xt_register_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1942         if (ret < 0)
1943                 goto err4;
1944
1945         /* Register setsockopt */
1946         ret = nf_register_sockopt(&ip6t_sockopts);
1947         if (ret < 0)
1948                 goto err5;
1949
1950         return 0;
1951
1952 err5:
1953         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1954 err4:
1955         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1956 err2:
1957         unregister_pernet_subsys(&ip6_tables_net_ops);
1958 err1:
1959         return ret;
1960 }
1961
1962 static void __exit ip6_tables_fini(void)
1963 {
1964         nf_unregister_sockopt(&ip6t_sockopts);
1965
1966         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1967         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1968         unregister_pernet_subsys(&ip6_tables_net_ops);
1969 }
1970
1971 EXPORT_SYMBOL(ip6t_register_table);
1972 EXPORT_SYMBOL(ip6t_unregister_table);
1973 EXPORT_SYMBOL(ip6t_do_table);
1974
1975 module_init(ip6_tables_init);
1976 module_exit(ip6_tables_fini);