net/ipv6/netfilter/ip6_tables.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3  * Packet matching code.
   4  *
   5  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   6  * Copyright (C) 2000-2005 Netfilter Core Team <[email protected]>
   7  * Copyright (c) 2006-2010 Patrick McHardy <[email protected]>
   8  */
   9
  10 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  11
  12 #include <linux/kernel.h>
  13 #include <linux/capability.h>
  14 #include <linux/in.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/kmod.h>
  17 #include <linux/vmalloc.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/module.h>
  20 #include <linux/poison.h>
  21 #include <linux/icmpv6.h>
  22 #include <net/ipv6.h>
  23 #include <net/compat.h>
  24 #include <linux/uaccess.h>
  25 #include <linux/mutex.h>
  26 #include <linux/proc_fs.h>
  27 #include <linux/err.h>
  28 #include <linux/cpumask.h>
  29
  30 #include <linux/netfilter_ipv6/ip6_tables.h>
  31 #include <linux/netfilter/x_tables.h>
  32 #include <net/netfilter/nf_log.h>
  33 #include "../../netfilter/xt_repldata.h"
  34
  35 MODULE_LICENSE("GPL");
  36 MODULE_AUTHOR("Netfilter Core Team <[email protected]>");
  37 MODULE_DESCRIPTION("IPv6 packet filter");
  38 MODULE_ALIAS("ip6t_icmp6");
  39
  40 void *ip6t_alloc_initial_table(const struct xt_table *info)
  41 {
  42         return xt_alloc_initial_table(ip6t, IP6T);
  43 }
  44 EXPORT_SYMBOL_GPL(ip6t_alloc_initial_table);
  45
  46 /* Returns whether matches rule or not. */
  47 /* Performance critical - called for every packet */
  48 static inline bool
  49 ip6_packet_match(const struct sk_buff *skb,
  50                  const char *indev,
  51                  const char *outdev,
  52                  const struct ip6t_ip6 *ip6info,
  53                  unsigned int *protoff,
  54                  int *fragoff, bool *hotdrop)
  55 {
  56         unsigned long ret;
  57         const struct ipv6hdr *ipv6 = ipv6_hdr(skb);
  58
  59         if (NF_INVF(ip6info, IP6T_INV_SRCIP,
  60                     ipv6_masked_addr_cmp(&ipv6->saddr, &ip6info->smsk,
  61                                          &ip6info->src)) ||
  62             NF_INVF(ip6info, IP6T_INV_DSTIP,
  63                     ipv6_masked_addr_cmp(&ipv6->daddr, &ip6info->dmsk,
  64                                          &ip6info->dst)))
  65                 return false;
  66
  67         ret = ifname_compare_aligned(indev, ip6info->iniface, ip6info->iniface_mask);
  68
  69         if (NF_INVF(ip6info, IP6T_INV_VIA_IN, ret != 0))
  70                 return false;
  71
  72         ret = ifname_compare_aligned(outdev, ip6info->outiface, ip6info->outiface_mask);
  73
  74         if (NF_INVF(ip6info, IP6T_INV_VIA_OUT, ret != 0))
  75                 return false;
  76
  77 /* ... might want to do something with class and flowlabel here ... */
  78
  79         /* look for the desired protocol header */
  80         if (ip6info->flags & IP6T_F_PROTO) {
  81                 int protohdr;
  82                 unsigned short _frag_off;
  83
  84                 protohdr = ipv6_find_hdr(skb, protoff, -1, &_frag_off, NULL);
  85                 if (protohdr < 0) {
  86                         if (_frag_off == 0)
  87                                 *hotdrop = true;
  88                         return false;
  89                 }
  90                 *fragoff = _frag_off;
  91
  92                 if (ip6info->proto == protohdr) {
  93                         if (ip6info->invflags & IP6T_INV_PROTO)
  94                                 return false;
  95
  96                         return true;
  97                 }
  98
  99                 /* We need match for the '-p all', too! */
 100                 if ((ip6info->proto != 0) &&
 101                         !(ip6info->invflags & IP6T_INV_PROTO))
 102                         return false;
 103         }
 104         return true;
 105 }
 106
 107 /* should be ip6 safe */
 108 static bool
 109 ip6_checkentry(const struct ip6t_ip6 *ipv6)
 110 {
 111         if (ipv6->flags & ~IP6T_F_MASK)
 112                 return false;
 113         if (ipv6->invflags & ~IP6T_INV_MASK)
 114                 return false;
 115
 116         return true;
 117 }
 118
 119 static unsigned int
 120 ip6t_error(struct sk_buff *skb, const struct xt_action_param *par)
 121 {
 122         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 123
 124         return NF_DROP;
 125 }
 126
 127 static inline struct ip6t_entry *
 128 get_entry(const void *base, unsigned int offset)
 129 {
 130         return (struct ip6t_entry *)(base + offset);
 131 }
 132
 133 /* All zeroes == unconditional rule. */
 134 /* Mildly perf critical (only if packet tracing is on) */
 135 static inline bool unconditional(const struct ip6t_entry *e)
 136 {
 137         static const struct ip6t_ip6 uncond;
 138
 139         return e->target_offset == sizeof(struct ip6t_entry) &&
 140                memcmp(&e->ipv6, &uncond, sizeof(uncond)) == 0;
 141 }
 142
 143 static inline const struct xt_entry_target *
 144 ip6t_get_target_c(const struct ip6t_entry *e)
 145 {
 146         return ip6t_get_target((struct ip6t_entry *)e);
 147 }
 148
 149 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 150 /* This cries for unification! */
 151 static const char *const hooknames[] = {
 152         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 153         [NF_INET_LOCAL_IN]              = "INPUT",
 154         [NF_INET_FORWARD]               = "FORWARD",
 155         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 156         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 157 };
 158
 159 enum nf_ip_trace_comments {
 160         NF_IP6_TRACE_COMMENT_RULE,
 161         NF_IP6_TRACE_COMMENT_RETURN,
 162         NF_IP6_TRACE_COMMENT_POLICY,
 163 };
 164
 165 static const char *const comments[] = {
 166         [NF_IP6_TRACE_COMMENT_RULE]     = "rule",
 167         [NF_IP6_TRACE_COMMENT_RETURN]   = "return",
 168         [NF_IP6_TRACE_COMMENT_POLICY]   = "policy",
 169 };
 170
 171 static const struct nf_loginfo trace_loginfo = {
 172         .type = NF_LOG_TYPE_LOG,
 173         .u = {
 174                 .log = {
 175                         .level = LOGLEVEL_WARNING,
 176                         .logflags = NF_LOG_DEFAULT_MASK,
 177                 },
 178         },
 179 };
 180
 181 /* Mildly perf critical (only if packet tracing is on) */
 182 static inline int
 183 get_chainname_rulenum(const struct ip6t_entry *s, const struct ip6t_entry *e,
 184                       const char *hookname, const char **chainname,
 185                       const char **comment, unsigned int *rulenum)
 186 {
 187         const struct xt_standard_target *t = (void *)ip6t_get_target_c(s);
 188
 189         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 190                 /* Head of user chain: ERROR target with chainname */
 191                 *chainname = t->target.data;
 192                 (*rulenum) = 0;
 193         } else if (s == e) {
 194                 (*rulenum)++;
 195
 196                 if (unconditional(s) &&
 197                     strcmp(t->target.u.kernel.target->name,
 198                            XT_STANDARD_TARGET) == 0 &&
 199                     t->verdict < 0) {
 200                         /* Tail of chains: STANDARD target (return/policy) */
 201                         *comment = *chainname == hookname
 202                                 ? comments[NF_IP6_TRACE_COMMENT_POLICY]
 203                                 : comments[NF_IP6_TRACE_COMMENT_RETURN];
 204                 }
 205                 return 1;
 206         } else
 207                 (*rulenum)++;
 208
 209         return 0;
 210 }
 211
 212 static void trace_packet(struct net *net,
 213                          const struct sk_buff *skb,
 214                          unsigned int hook,
 215                          const struct net_device *in,
 216                          const struct net_device *out,
 217                          const char *tablename,
 218                          const struct xt_table_info *private,
 219                          const struct ip6t_entry *e)
 220 {
 221         const struct ip6t_entry *root;
 222         const char *hookname, *chainname, *comment;
 223         const struct ip6t_entry *iter;
 224         unsigned int rulenum = 0;
 225
 226         root = get_entry(private->entries, private->hook_entry[hook]);
 227
 228         hookname = chainname = hooknames[hook];
 229         comment = comments[NF_IP6_TRACE_COMMENT_RULE];
 230
 231         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 232                 if (get_chainname_rulenum(iter, e, hookname,
 233                     &chainname, &comment, &rulenum) != 0)
 234                         break;
 235
 236         nf_log_trace(net, AF_INET6, hook, skb, in, out, &trace_loginfo,
 237                      "TRACE: %s:%s:%s:%u ",
 238                      tablename, chainname, comment, rulenum);
 239 }
 240 #endif
 241
 242 static inline struct ip6t_entry *
 243 ip6t_next_entry(const struct ip6t_entry *entry)
 244 {
 245         return (void *)entry + entry->next_offset;
 246 }
 247
 248 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 249 unsigned int
 250 ip6t_do_table(struct sk_buff *skb,
 251               const struct nf_hook_state *state,
 252               struct xt_table *table)
 253 {
 254         unsigned int hook = state->hook;
 255         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 256         /* Initializing verdict to NF_DROP keeps gcc happy. */
 257         unsigned int verdict = NF_DROP;
 258         const char *indev, *outdev;
 259         const void *table_base;
 260         struct ip6t_entry *e, **jumpstack;
 261         unsigned int stackidx, cpu;
 262         const struct xt_table_info *private;
 263         struct xt_action_param acpar;
 264         unsigned int addend;
 265
 266         /* Initialization */
 267         stackidx = 0;
 268         indev = state->in ? state->in->name : nulldevname;
 269         outdev = state->out ? state->out->name : nulldevname;
 270         /* We handle fragments by dealing with the first fragment as
 271          * if it was a normal packet.  All other fragments are treated
 272          * normally, except that they will NEVER match rules that ask
 273          * things we don't know, ie. tcp syn flag or ports).  If the
 274          * rule is also a fragment-specific rule, non-fragments won't
 275          * match it. */
 276         acpar.hotdrop = false;
 277         acpar.state   = state;
 278
 279         WARN_ON(!(table->valid_hooks & (1 << hook)));
 280
 281         local_bh_disable();
 282         addend = xt_write_recseq_begin();
 283         private = rcu_access_pointer(table->private);
 284         cpu        = smp_processor_id();
 285         table_base = private->entries;
 286         jumpstack  = (struct ip6t_entry **)private->jumpstack[cpu];
 287
 288         /* Switch to alternate jumpstack if we're being invoked via TEE.
 289          * TEE issues XT_CONTINUE verdict on original skb so we must not
 290          * clobber the jumpstack.
 291          *
 292          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 293          * but it is no problem since absolute verdict is issued by these.
 294          */
 295         if (static_key_false(&xt_tee_enabled))
 296                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 297
 298         e = get_entry(table_base, private->hook_entry[hook]);
 299
 300         do {
 301                 const struct xt_entry_target *t;
 302                 const struct xt_entry_match *ematch;
 303                 struct xt_counters *counter;
 304
 305                 WARN_ON(!e);
 306                 acpar.thoff = 0;
 307                 if (!ip6_packet_match(skb, indev, outdev, &e->ipv6,
 308                     &acpar.thoff, &acpar.fragoff, &acpar.hotdrop)) {
 309  no_match:
 310                         e = ip6t_next_entry(e);
 311                         continue;
 312                 }
 313
 314                 xt_ematch_foreach(ematch, e) {
 315                         acpar.match     = ematch->u.kernel.match;
 316                         acpar.matchinfo = ematch->data;
 317                         if (!acpar.match->match(skb, &acpar))
 318                                 goto no_match;
 319                 }
 320
 321                 counter = xt_get_this_cpu_counter(&e->counters);
 322                 ADD_COUNTER(*counter, skb->len, 1);
 323
 324                 t = ip6t_get_target_c(e);
 325                 WARN_ON(!t->u.kernel.target);
 326
 327 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 328                 /* The packet is traced: log it */
 329                 if (unlikely(skb->nf_trace))
 330                         trace_packet(state->net, skb, hook, state->in,
 331                                      state->out, table->name, private, e);
 332 #endif
 333                 /* Standard target? */
 334                 if (!t->u.kernel.target->target) {
 335                         int v;
 336
 337                         v = ((struct xt_standard_target *)t)->verdict;
 338                         if (v < 0) {
 339                                 /* Pop from stack? */
 340                                 if (v != XT_RETURN) {
 341                                         verdict = (unsigned int)(-v) - 1;
 342                                         break;
 343                                 }
 344                                 if (stackidx == 0)
 345                                         e = get_entry(table_base,
 346                                             private->underflow[hook]);
 347                                 else
 348                                         e = ip6t_next_entry(jumpstack[--stackidx]);
 349                                 continue;
 350                         }
 351                         if (table_base + v != ip6t_next_entry(e) &&
 352                             !(e->ipv6.flags & IP6T_F_GOTO)) {
 353                                 if (unlikely(stackidx >= private->stacksize)) {
 354                                         verdict = NF_DROP;
 355                                         break;
 356                                 }
 357                                 jumpstack[stackidx++] = e;
 358                         }
 359
 360                         e = get_entry(table_base, v);
 361                         continue;
 362                 }
 363
 364                 acpar.target   = t->u.kernel.target;
 365                 acpar.targinfo = t->data;
 366
 367                 verdict = t->u.kernel.target->target(skb, &acpar);
 368                 if (verdict == XT_CONTINUE)
 369                         e = ip6t_next_entry(e);
 370                 else
 371                         /* Verdict */
 372                         break;
 373         } while (!acpar.hotdrop);
 374
 375         xt_write_recseq_end(addend);
 376         local_bh_enable();
 377
 378         if (acpar.hotdrop)
 379                 return NF_DROP;
 380         else return verdict;
 381 }
 382
 383 /* Figures out from what hook each rule can be called: returns 0 if
 384    there are loops.  Puts hook bitmask in comefrom. */
 385 static int
 386 mark_source_chains(const struct xt_table_info *newinfo,
 387                    unsigned int valid_hooks, void *entry0,
 388                    unsigned int *offsets)
 389 {
 390         unsigned int hook;
 391
 392         /* No recursion; use packet counter to save back ptrs (reset
 393            to 0 as we leave), and comefrom to save source hook bitmask */
 394         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 395                 unsigned int pos = newinfo->hook_entry[hook];
 396                 struct ip6t_entry *e = entry0 + pos;
 397
 398                 if (!(valid_hooks & (1 << hook)))
 399                         continue;
 400
 401                 /* Set initial back pointer. */
 402                 e->counters.pcnt = pos;
 403
 404                 for (;;) {
 405                         const struct xt_standard_target *t
 406                                 = (void *)ip6t_get_target_c(e);
 407                         int visited = e->comefrom & (1 << hook);
 408
 409                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 410                                 return 0;
 411
 412                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 413
 414                         /* Unconditional return/END. */
 415                         if ((unconditional(e) &&
 416                              (strcmp(t->target.u.user.name,
 417                                      XT_STANDARD_TARGET) == 0) &&
 418                              t->verdict < 0) || visited) {
 419                                 unsigned int oldpos, size;
 420
 421                                 /* Return: backtrack through the last
 422                                    big jump. */
 423                                 do {
 424                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 425                                         oldpos = pos;
 426                                         pos = e->counters.pcnt;
 427                                         e->counters.pcnt = 0;
 428
 429                                         /* We're at the start. */
 430                                         if (pos == oldpos)
 431                                                 goto next;
 432
 433                                         e = entry0 + pos;
 434                                 } while (oldpos == pos + e->next_offset);
 435
 436                                 /* Move along one */
 437                                 size = e->next_offset;
 438                                 e = entry0 + pos + size;
 439                                 if (pos + size >= newinfo->size)
 440                                         return 0;
 441                                 e->counters.pcnt = pos;
 442                                 pos += size;
 443                         } else {
 444                                 int newpos = t->verdict;
 445
 446                                 if (strcmp(t->target.u.user.name,
 447                                            XT_STANDARD_TARGET) == 0 &&
 448                                     newpos >= 0) {
 449                                         /* This a jump; chase it. */
 450                                         if (!xt_find_jump_offset(offsets, newpos,
 451                                                                  newinfo->number))
 452                                                 return 0;
 453                                 } else {
 454                                         /* ... this is a fallthru */
 455                                         newpos = pos + e->next_offset;
 456                                         if (newpos >= newinfo->size)
 457                                                 return 0;
 458                                 }
 459                                 e = entry0 + newpos;
 460                                 e->counters.pcnt = pos;
 461                                 pos = newpos;
 462                         }
 463                 }
 464 next:           ;
 465         }
 466         return 1;
 467 }
 468
 469 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 470 {
 471         struct xt_mtdtor_param par;
 472
 473         par.net       = net;
 474         par.match     = m->u.kernel.match;
 475         par.matchinfo = m->data;
 476         par.family    = NFPROTO_IPV6;
 477         if (par.match->destroy != NULL)
 478                 par.match->destroy(&par);
 479         module_put(par.match->me);
 480 }
 481
 482 static int check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 483 {
 484         const struct ip6t_ip6 *ipv6 = par->entryinfo;
 485
 486         par->match     = m->u.kernel.match;
 487         par->matchinfo = m->data;
 488
 489         return xt_check_match(par, m->u.match_size - sizeof(*m),
 490                               ipv6->proto, ipv6->invflags & IP6T_INV_PROTO);
 491 }
 492
 493 static int
 494 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 495 {
 496         struct xt_match *match;
 497         int ret;
 498
 499         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
 500                                       m->u.user.revision);
 501         if (IS_ERR(match))
 502                 return PTR_ERR(match);
 503
 504         m->u.kernel.match = match;
 505
 506         ret = check_match(m, par);
 507         if (ret)
 508                 goto err;
 509
 510         return 0;
 511 err:
 512         module_put(m->u.kernel.match->me);
 513         return ret;
 514 }
 515
 516 static int check_target(struct ip6t_entry *e, struct net *net, const char *name)
 517 {
 518         struct xt_entry_target *t = ip6t_get_target(e);
 519         struct xt_tgchk_param par = {
 520                 .net       = net,
 521                 .table     = name,
 522                 .entryinfo = e,
 523                 .target    = t->u.kernel.target,
 524                 .targinfo  = t->data,
 525                 .hook_mask = e->comefrom,
 526                 .family    = NFPROTO_IPV6,
 527         };
 528
 529         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 530                                e->ipv6.proto,
 531                                e->ipv6.invflags & IP6T_INV_PROTO);
 532 }
 533
 534 static int
 535 find_check_entry(struct ip6t_entry *e, struct net *net, const char *name,
 536                  unsigned int size,
 537                  struct xt_percpu_counter_alloc_state *alloc_state)
 538 {
 539         struct xt_entry_target *t;
 540         struct xt_target *target;
 541         int ret;
 542         unsigned int j;
 543         struct xt_mtchk_param mtpar;
 544         struct xt_entry_match *ematch;
 545
 546         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 547                 return -ENOMEM;
 548
 549         j = 0;
 550         memset(&mtpar, 0, sizeof(mtpar));
 551         mtpar.net       = net;
 552         mtpar.table     = name;
 553         mtpar.entryinfo = &e->ipv6;
 554         mtpar.hook_mask = e->comefrom;
 555         mtpar.family    = NFPROTO_IPV6;
 556         xt_ematch_foreach(ematch, e) {
 557                 ret = find_check_match(ematch, &mtpar);
 558                 if (ret != 0)
 559                         goto cleanup_matches;
 560                 ++j;
 561         }
 562
 563         t = ip6t_get_target(e);
 564         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
 565                                         t->u.user.revision);
 566         if (IS_ERR(target)) {
 567                 ret = PTR_ERR(target);
 568                 goto cleanup_matches;
 569         }
 570         t->u.kernel.target = target;
 571
 572         ret = check_target(e, net, name);
 573         if (ret)
 574                 goto err;
 575         return 0;
 576  err:
 577         module_put(t->u.kernel.target->me);
 578  cleanup_matches:
 579         xt_ematch_foreach(ematch, e) {
 580                 if (j-- == 0)
 581                         break;
 582                 cleanup_match(ematch, net);
 583         }
 584
 585         xt_percpu_counter_free(&e->counters);
 586
 587         return ret;
 588 }
 589
 590 static bool check_underflow(const struct ip6t_entry *e)
 591 {
 592         const struct xt_entry_target *t;
 593         unsigned int verdict;
 594
 595         if (!unconditional(e))
 596                 return false;
 597         t = ip6t_get_target_c(e);
 598         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 599                 return false;
 600         verdict = ((struct xt_standard_target *)t)->verdict;
 601         verdict = -verdict - 1;
 602         return verdict == NF_DROP || verdict == NF_ACCEPT;
 603 }
 604
 605 static int
 606 check_entry_size_and_hooks(struct ip6t_entry *e,
 607                            struct xt_table_info *newinfo,
 608                            const unsigned char *base,
 609                            const unsigned char *limit,
 610                            const unsigned int *hook_entries,
 611                            const unsigned int *underflows,
 612                            unsigned int valid_hooks)
 613 {
 614         unsigned int h;
 615         int err;
 616
 617         if ((unsigned long)e % __alignof__(struct ip6t_entry) != 0 ||
 618             (unsigned char *)e + sizeof(struct ip6t_entry) >= limit ||
 619             (unsigned char *)e + e->next_offset > limit)
 620                 return -EINVAL;
 621
 622         if (e->next_offset
 623             < sizeof(struct ip6t_entry) + sizeof(struct xt_entry_target))
 624                 return -EINVAL;
 625
 626         if (!ip6_checkentry(&e->ipv6))
 627                 return -EINVAL;
 628
 629         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 630                                      e->next_offset);
 631         if (err)
 632                 return err;
 633
 634         /* Check hooks & underflows */
 635         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 636                 if (!(valid_hooks & (1 << h)))
 637                         continue;
 638                 if ((unsigned char *)e - base == hook_entries[h])
 639                         newinfo->hook_entry[h] = hook_entries[h];
 640                 if ((unsigned char *)e - base == underflows[h]) {
 641                         if (!check_underflow(e))
 642                                 return -EINVAL;
 643
 644                         newinfo->underflow[h] = underflows[h];
 645                 }
 646         }
 647
 648         /* Clear counters and comefrom */
 649         e->counters = ((struct xt_counters) { 0, 0 });
 650         e->comefrom = 0;
 651         return 0;
 652 }
 653
 654 static void cleanup_entry(struct ip6t_entry *e, struct net *net)
 655 {
 656         struct xt_tgdtor_param par;
 657         struct xt_entry_target *t;
 658         struct xt_entry_match *ematch;
 659
 660         /* Cleanup all matches */
 661         xt_ematch_foreach(ematch, e)
 662                 cleanup_match(ematch, net);
 663         t = ip6t_get_target(e);
 664
 665         par.net      = net;
 666         par.target   = t->u.kernel.target;
 667         par.targinfo = t->data;
 668         par.family   = NFPROTO_IPV6;
 669         if (par.target->destroy != NULL)
 670                 par.target->destroy(&par);
 671         module_put(par.target->me);
 672         xt_percpu_counter_free(&e->counters);
 673 }
 674
 675 /* Checks and translates the user-supplied table segment (held in
 676    newinfo) */
 677 static int
 678 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 679                 const struct ip6t_replace *repl)
 680 {
 681         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 682         struct ip6t_entry *iter;
 683         unsigned int *offsets;
 684         unsigned int i;
 685         int ret = 0;
 686
 687         newinfo->size = repl->size;
 688         newinfo->number = repl->num_entries;
 689
 690         /* Init all hooks to impossible value. */
 691         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 692                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 693                 newinfo->underflow[i] = 0xFFFFFFFF;
 694         }
 695
 696         offsets = xt_alloc_entry_offsets(newinfo->number);
 697         if (!offsets)
 698                 return -ENOMEM;
 699         i = 0;
 700         /* Walk through entries, checking offsets. */
 701         xt_entry_foreach(iter, entry0, newinfo->size) {
 702                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 703                                                  entry0 + repl->size,
 704                                                  repl->hook_entry,
 705                                                  repl->underflow,
 706                                                  repl->valid_hooks);
 707                 if (ret != 0)
 708                         goto out_free;
 709                 if (i < repl->num_entries)
 710                         offsets[i] = (void *)iter - entry0;
 711                 ++i;
 712                 if (strcmp(ip6t_get_target(iter)->u.user.name,
 713                     XT_ERROR_TARGET) == 0)
 714                         ++newinfo->stacksize;
 715         }
 716
 717         ret = -EINVAL;
 718         if (i != repl->num_entries)
 719                 goto out_free;
 720
 721         ret = xt_check_table_hooks(newinfo, repl->valid_hooks);
 722         if (ret)
 723                 goto out_free;
 724
 725         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 726                 ret = -ELOOP;
 727                 goto out_free;
 728         }
 729         kvfree(offsets);
 730
 731         /* Finally, each sanity check must pass */
 732         i = 0;
 733         xt_entry_foreach(iter, entry0, newinfo->size) {
 734                 ret = find_check_entry(iter, net, repl->name, repl->size,
 735                                        &alloc_state);
 736                 if (ret != 0)
 737                         break;
 738                 ++i;
 739         }
 740
 741         if (ret != 0) {
 742                 xt_entry_foreach(iter, entry0, newinfo->size) {
 743                         if (i-- == 0)
 744                                 break;
 745                         cleanup_entry(iter, net);
 746                 }
 747                 return ret;
 748         }
 749
 750         return ret;
 751  out_free:
 752         kvfree(offsets);
 753         return ret;
 754 }
 755
 756 static void
 757 get_counters(const struct xt_table_info *t,
 758              struct xt_counters counters[])
 759 {
 760         struct ip6t_entry *iter;
 761         unsigned int cpu;
 762         unsigned int i;
 763
 764         for_each_possible_cpu(cpu) {
 765                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 766
 767                 i = 0;
 768                 xt_entry_foreach(iter, t->entries, t->size) {
 769                         struct xt_counters *tmp;
 770                         u64 bcnt, pcnt;
 771                         unsigned int start;
 772
 773                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 774                         do {
 775                                 start = read_seqcount_begin(s);
 776                                 bcnt = tmp->bcnt;
 777                                 pcnt = tmp->pcnt;
 778                         } while (read_seqcount_retry(s, start));
 779
 780                         ADD_COUNTER(counters[i], bcnt, pcnt);
 781                         ++i;
 782                         cond_resched();
 783                 }
 784         }
 785 }
 786
 787 static void get_old_counters(const struct xt_table_info *t,
 788                              struct xt_counters counters[])
 789 {
 790         struct ip6t_entry *iter;
 791         unsigned int cpu, i;
 792
 793         for_each_possible_cpu(cpu) {
 794                 i = 0;
 795                 xt_entry_foreach(iter, t->entries, t->size) {
 796                         const struct xt_counters *tmp;
 797
 798                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 799                         ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
 800                         ++i;
 801                 }
 802                 cond_resched();
 803         }
 804 }
 805
 806 static struct xt_counters *alloc_counters(const struct xt_table *table)
 807 {
 808         unsigned int countersize;
 809         struct xt_counters *counters;
 810         const struct xt_table_info *private = xt_table_get_private_protected(table);
 811
 812         /* We need atomic snapshot of counters: rest doesn't change
 813            (other than comefrom, which userspace doesn't care
 814            about). */
 815         countersize = sizeof(struct xt_counters) * private->number;
 816         counters = vzalloc(countersize);
 817
 818         if (counters == NULL)
 819                 return ERR_PTR(-ENOMEM);
 820
 821         get_counters(private, counters);
 822
 823         return counters;
 824 }
 825
 826 static int
 827 copy_entries_to_user(unsigned int total_size,
 828                      const struct xt_table *table,
 829                      void __user *userptr)
 830 {
 831         unsigned int off, num;
 832         const struct ip6t_entry *e;
 833         struct xt_counters *counters;
 834         const struct xt_table_info *private = xt_table_get_private_protected(table);
 835         int ret = 0;
 836         const void *loc_cpu_entry;
 837
 838         counters = alloc_counters(table);
 839         if (IS_ERR(counters))
 840                 return PTR_ERR(counters);
 841
 842         loc_cpu_entry = private->entries;
 843
 844         /* FIXME: use iterator macros --RR */
 845         /* ... then go back and fix counters and names */
 846         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 847                 unsigned int i;
 848                 const struct xt_entry_match *m;
 849                 const struct xt_entry_target *t;
 850
 851                 e = loc_cpu_entry + off;
 852                 if (copy_to_user(userptr + off, e, sizeof(*e))) {
 853                         ret = -EFAULT;
 854                         goto free_counters;
 855                 }
 856                 if (copy_to_user(userptr + off
 857                                  + offsetof(struct ip6t_entry, counters),
 858                                  &counters[num],
 859                                  sizeof(counters[num])) != 0) {
 860                         ret = -EFAULT;
 861                         goto free_counters;
 862                 }
 863
 864                 for (i = sizeof(struct ip6t_entry);
 865                      i < e->target_offset;
 866                      i += m->u.match_size) {
 867                         m = (void *)e + i;
 868
 869                         if (xt_match_to_user(m, userptr + off + i)) {
 870                                 ret = -EFAULT;
 871                                 goto free_counters;
 872                         }
 873                 }
 874
 875                 t = ip6t_get_target_c(e);
 876                 if (xt_target_to_user(t, userptr + off + e->target_offset)) {
 877                         ret = -EFAULT;
 878                         goto free_counters;
 879                 }
 880         }
 881
 882  free_counters:
 883         vfree(counters);
 884         return ret;
 885 }
 886
 887 #ifdef CONFIG_COMPAT
 888 static void compat_standard_from_user(void *dst, const void *src)
 889 {
 890         int v = *(compat_int_t *)src;
 891
 892         if (v > 0)
 893                 v += xt_compat_calc_jump(AF_INET6, v);
 894         memcpy(dst, &v, sizeof(v));
 895 }
 896
 897 static int compat_standard_to_user(void __user *dst, const void *src)
 898 {
 899         compat_int_t cv = *(int *)src;
 900
 901         if (cv > 0)
 902                 cv -= xt_compat_calc_jump(AF_INET6, cv);
 903         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 904 }
 905
 906 static int compat_calc_entry(const struct ip6t_entry *e,
 907                              const struct xt_table_info *info,
 908                              const void *base, struct xt_table_info *newinfo)
 909 {
 910         const struct xt_entry_match *ematch;
 911         const struct xt_entry_target *t;
 912         unsigned int entry_offset;
 913         int off, i, ret;
 914
 915         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
 916         entry_offset = (void *)e - base;
 917         xt_ematch_foreach(ematch, e)
 918                 off += xt_compat_match_offset(ematch->u.kernel.match);
 919         t = ip6t_get_target_c(e);
 920         off += xt_compat_target_offset(t->u.kernel.target);
 921         newinfo->size -= off;
 922         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
 923         if (ret)
 924                 return ret;
 925
 926         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 927                 if (info->hook_entry[i] &&
 928                     (e < (struct ip6t_entry *)(base + info->hook_entry[i])))
 929                         newinfo->hook_entry[i] -= off;
 930                 if (info->underflow[i] &&
 931                     (e < (struct ip6t_entry *)(base + info->underflow[i])))
 932                         newinfo->underflow[i] -= off;
 933         }
 934         return 0;
 935 }
 936
 937 static int compat_table_info(const struct xt_table_info *info,
 938                              struct xt_table_info *newinfo)
 939 {
 940         struct ip6t_entry *iter;
 941         const void *loc_cpu_entry;
 942         int ret;
 943
 944         if (!newinfo || !info)
 945                 return -EINVAL;
 946
 947         /* we dont care about newinfo->entries */
 948         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 949         newinfo->initial_entries = 0;
 950         loc_cpu_entry = info->entries;
 951         ret = xt_compat_init_offsets(AF_INET6, info->number);
 952         if (ret)
 953                 return ret;
 954         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 955                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 956                 if (ret != 0)
 957                         return ret;
 958         }
 959         return 0;
 960 }
 961 #endif
 962
 963 static int get_info(struct net *net, void __user *user, const int *len)
 964 {
 965         char name[XT_TABLE_MAXNAMELEN];
 966         struct xt_table *t;
 967         int ret;
 968
 969         if (*len != sizeof(struct ip6t_getinfo))
 970                 return -EINVAL;
 971
 972         if (copy_from_user(name, user, sizeof(name)) != 0)
 973                 return -EFAULT;
 974
 975         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 976 #ifdef CONFIG_COMPAT
 977         if (in_compat_syscall())
 978                 xt_compat_lock(AF_INET6);
 979 #endif
 980         t = xt_request_find_table_lock(net, AF_INET6, name);
 981         if (!IS_ERR(t)) {
 982                 struct ip6t_getinfo info;
 983                 const struct xt_table_info *private = xt_table_get_private_protected(t);
 984 #ifdef CONFIG_COMPAT
 985                 struct xt_table_info tmp;
 986
 987                 if (in_compat_syscall()) {
 988                         ret = compat_table_info(private, &tmp);
 989                         xt_compat_flush_offsets(AF_INET6);
 990                         private = &tmp;
 991                 }
 992 #endif
 993                 memset(&info, 0, sizeof(info));
 994                 info.valid_hooks = t->valid_hooks;
 995                 memcpy(info.hook_entry, private->hook_entry,
 996                        sizeof(info.hook_entry));
 997                 memcpy(info.underflow, private->underflow,
 998                        sizeof(info.underflow));
 999                 info.num_entries = private->number;
1000                 info.size = private->size;
1001                 strcpy(info.name, name);
1002
1003                 if (copy_to_user(user, &info, *len) != 0)
1004                         ret = -EFAULT;
1005                 else
1006                         ret = 0;
1007
1008                 xt_table_unlock(t);
1009                 module_put(t->me);
1010         } else
1011                 ret = PTR_ERR(t);
1012 #ifdef CONFIG_COMPAT
1013         if (in_compat_syscall())
1014                 xt_compat_unlock(AF_INET6);
1015 #endif
1016         return ret;
1017 }
1018
1019 static int
1020 get_entries(struct net *net, struct ip6t_get_entries __user *uptr,
1021             const int *len)
1022 {
1023         int ret;
1024         struct ip6t_get_entries get;
1025         struct xt_table *t;
1026
1027         if (*len < sizeof(get))
1028                 return -EINVAL;
1029         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1030                 return -EFAULT;
1031         if (*len != sizeof(struct ip6t_get_entries) + get.size)
1032                 return -EINVAL;
1033
1034         get.name[sizeof(get.name) - 1] = '\0';
1035
1036         t = xt_find_table_lock(net, AF_INET6, get.name);
1037         if (!IS_ERR(t)) {
1038                 struct xt_table_info *private = xt_table_get_private_protected(t);
1039                 if (get.size == private->size)
1040                         ret = copy_entries_to_user(private->size,
1041                                                    t, uptr->entrytable);
1042                 else
1043                         ret = -EAGAIN;
1044
1045                 module_put(t->me);
1046                 xt_table_unlock(t);
1047         } else
1048                 ret = PTR_ERR(t);
1049
1050         return ret;
1051 }
1052
1053 static int
1054 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1055              struct xt_table_info *newinfo, unsigned int num_counters,
1056              void __user *counters_ptr)
1057 {
1058         int ret;
1059         struct xt_table *t;
1060         struct xt_table_info *oldinfo;
1061         struct xt_counters *counters;
1062         struct ip6t_entry *iter;
1063
1064         ret = 0;
1065         counters = xt_counters_alloc(num_counters);
1066         if (!counters) {
1067                 ret = -ENOMEM;
1068                 goto out;
1069         }
1070
1071         t = xt_request_find_table_lock(net, AF_INET6, name);
1072         if (IS_ERR(t)) {
1073                 ret = PTR_ERR(t);
1074                 goto free_newinfo_counters_untrans;
1075         }
1076
1077         /* You lied! */
1078         if (valid_hooks != t->valid_hooks) {
1079                 ret = -EINVAL;
1080                 goto put_module;
1081         }
1082
1083         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1084         if (!oldinfo)
1085                 goto put_module;
1086
1087         /* Update module usage count based on number of rules */
1088         if ((oldinfo->number > oldinfo->initial_entries) ||
1089             (newinfo->number <= oldinfo->initial_entries))
1090                 module_put(t->me);
1091         if ((oldinfo->number > oldinfo->initial_entries) &&
1092             (newinfo->number <= oldinfo->initial_entries))
1093                 module_put(t->me);
1094
1095         xt_table_unlock(t);
1096
1097         get_old_counters(oldinfo, counters);
1098
1099         /* Decrease module usage counts and free resource */
1100         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1101                 cleanup_entry(iter, net);
1102
1103         xt_free_table_info(oldinfo);
1104         if (copy_to_user(counters_ptr, counters,
1105                          sizeof(struct xt_counters) * num_counters) != 0) {
1106                 /* Silent error, can't fail, new table is already in place */
1107                 net_warn_ratelimited("ip6tables: counters copy to user failed while replacing table\n");
1108         }
1109         vfree(counters);
1110         return ret;
1111
1112  put_module:
1113         module_put(t->me);
1114         xt_table_unlock(t);
1115  free_newinfo_counters_untrans:
1116         vfree(counters);
1117  out:
1118         return ret;
1119 }
1120
1121 static int
1122 do_replace(struct net *net, sockptr_t arg, unsigned int len)
1123 {
1124         int ret;
1125         struct ip6t_replace tmp;
1126         struct xt_table_info *newinfo;
1127         void *loc_cpu_entry;
1128         struct ip6t_entry *iter;
1129
1130         if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0)
1131                 return -EFAULT;
1132
1133         /* overflow check */
1134         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1135                 return -ENOMEM;
1136         if (tmp.num_counters == 0)
1137                 return -EINVAL;
1138
1139         tmp.name[sizeof(tmp.name)-1] = 0;
1140
1141         newinfo = xt_alloc_table_info(tmp.size);
1142         if (!newinfo)
1143                 return -ENOMEM;
1144
1145         loc_cpu_entry = newinfo->entries;
1146         if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp),
1147                         tmp.size) != 0) {
1148                 ret = -EFAULT;
1149                 goto free_newinfo;
1150         }
1151
1152         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1153         if (ret != 0)
1154                 goto free_newinfo;
1155
1156         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1157                            tmp.num_counters, tmp.counters);
1158         if (ret)
1159                 goto free_newinfo_untrans;
1160         return 0;
1161
1162  free_newinfo_untrans:
1163         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1164                 cleanup_entry(iter, net);
1165  free_newinfo:
1166         xt_free_table_info(newinfo);
1167         return ret;
1168 }
1169
1170 static int
1171 do_add_counters(struct net *net, sockptr_t arg, unsigned int len)
1172 {
1173         unsigned int i;
1174         struct xt_counters_info tmp;
1175         struct xt_counters *paddc;
1176         struct xt_table *t;
1177         const struct xt_table_info *private;
1178         int ret = 0;
1179         struct ip6t_entry *iter;
1180         unsigned int addend;
1181
1182         paddc = xt_copy_counters(arg, len, &tmp);
1183         if (IS_ERR(paddc))
1184                 return PTR_ERR(paddc);
1185         t = xt_find_table_lock(net, AF_INET6, tmp.name);
1186         if (IS_ERR(t)) {
1187                 ret = PTR_ERR(t);
1188                 goto free;
1189         }
1190
1191         local_bh_disable();
1192         private = xt_table_get_private_protected(t);
1193         if (private->number != tmp.num_counters) {
1194                 ret = -EINVAL;
1195                 goto unlock_up_free;
1196         }
1197
1198         i = 0;
1199         addend = xt_write_recseq_begin();
1200         xt_entry_foreach(iter, private->entries, private->size) {
1201                 struct xt_counters *tmp;
1202
1203                 tmp = xt_get_this_cpu_counter(&iter->counters);
1204                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1205                 ++i;
1206         }
1207         xt_write_recseq_end(addend);
1208  unlock_up_free:
1209         local_bh_enable();
1210         xt_table_unlock(t);
1211         module_put(t->me);
1212  free:
1213         vfree(paddc);
1214
1215         return ret;
1216 }
1217
1218 #ifdef CONFIG_COMPAT
1219 struct compat_ip6t_replace {
1220         char                    name[XT_TABLE_MAXNAMELEN];
1221         u32                     valid_hooks;
1222         u32                     num_entries;
1223         u32                     size;
1224         u32                     hook_entry[NF_INET_NUMHOOKS];
1225         u32                     underflow[NF_INET_NUMHOOKS];
1226         u32                     num_counters;
1227         compat_uptr_t           counters;       /* struct xt_counters * */
1228         struct compat_ip6t_entry entries[];
1229 };
1230
1231 static int
1232 compat_copy_entry_to_user(struct ip6t_entry *e, void __user **dstptr,
1233                           unsigned int *size, struct xt_counters *counters,
1234                           unsigned int i)
1235 {
1236         struct xt_entry_target *t;
1237         struct compat_ip6t_entry __user *ce;
1238         u_int16_t target_offset, next_offset;
1239         compat_uint_t origsize;
1240         const struct xt_entry_match *ematch;
1241         int ret = 0;
1242
1243         origsize = *size;
1244         ce = *dstptr;
1245         if (copy_to_user(ce, e, sizeof(struct ip6t_entry)) != 0 ||
1246             copy_to_user(&ce->counters, &counters[i],
1247             sizeof(counters[i])) != 0)
1248                 return -EFAULT;
1249
1250         *dstptr += sizeof(struct compat_ip6t_entry);
1251         *size -= sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1252
1253         xt_ematch_foreach(ematch, e) {
1254                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1255                 if (ret != 0)
1256                         return ret;
1257         }
1258         target_offset = e->target_offset - (origsize - *size);
1259         t = ip6t_get_target(e);
1260         ret = xt_compat_target_to_user(t, dstptr, size);
1261         if (ret)
1262                 return ret;
1263         next_offset = e->next_offset - (origsize - *size);
1264         if (put_user(target_offset, &ce->target_offset) != 0 ||
1265             put_user(next_offset, &ce->next_offset) != 0)
1266                 return -EFAULT;
1267         return 0;
1268 }
1269
1270 static int
1271 compat_find_calc_match(struct xt_entry_match *m,
1272                        const struct ip6t_ip6 *ipv6,
1273                        int *size)
1274 {
1275         struct xt_match *match;
1276
1277         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
1278                                       m->u.user.revision);
1279         if (IS_ERR(match))
1280                 return PTR_ERR(match);
1281
1282         m->u.kernel.match = match;
1283         *size += xt_compat_match_offset(match);
1284         return 0;
1285 }
1286
1287 static void compat_release_entry(struct compat_ip6t_entry *e)
1288 {
1289         struct xt_entry_target *t;
1290         struct xt_entry_match *ematch;
1291
1292         /* Cleanup all matches */
1293         xt_ematch_foreach(ematch, e)
1294                 module_put(ematch->u.kernel.match->me);
1295         t = compat_ip6t_get_target(e);
1296         module_put(t->u.kernel.target->me);
1297 }
1298
1299 static int
1300 check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
1301                                   struct xt_table_info *newinfo,
1302                                   unsigned int *size,
1303                                   const unsigned char *base,
1304                                   const unsigned char *limit)
1305 {
1306         struct xt_entry_match *ematch;
1307         struct xt_entry_target *t;
1308         struct xt_target *target;
1309         unsigned int entry_offset;
1310         unsigned int j;
1311         int ret, off;
1312
1313         if ((unsigned long)e % __alignof__(struct compat_ip6t_entry) != 0 ||
1314             (unsigned char *)e + sizeof(struct compat_ip6t_entry) >= limit ||
1315             (unsigned char *)e + e->next_offset > limit)
1316                 return -EINVAL;
1317
1318         if (e->next_offset < sizeof(struct compat_ip6t_entry) +
1319                              sizeof(struct compat_xt_entry_target))
1320                 return -EINVAL;
1321
1322         if (!ip6_checkentry(&e->ipv6))
1323                 return -EINVAL;
1324
1325         ret = xt_compat_check_entry_offsets(e, e->elems,
1326                                             e->target_offset, e->next_offset);
1327         if (ret)
1328                 return ret;
1329
1330         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1331         entry_offset = (void *)e - (void *)base;
1332         j = 0;
1333         xt_ematch_foreach(ematch, e) {
1334                 ret = compat_find_calc_match(ematch, &e->ipv6, &off);
1335                 if (ret != 0)
1336                         goto release_matches;
1337                 ++j;
1338         }
1339
1340         t = compat_ip6t_get_target(e);
1341         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
1342                                         t->u.user.revision);
1343         if (IS_ERR(target)) {
1344                 ret = PTR_ERR(target);
1345                 goto release_matches;
1346         }
1347         t->u.kernel.target = target;
1348
1349         off += xt_compat_target_offset(target);
1350         *size += off;
1351         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1352         if (ret)
1353                 goto out;
1354
1355         return 0;
1356
1357 out:
1358         module_put(t->u.kernel.target->me);
1359 release_matches:
1360         xt_ematch_foreach(ematch, e) {
1361                 if (j-- == 0)
1362                         break;
1363                 module_put(ematch->u.kernel.match->me);
1364         }
1365         return ret;
1366 }
1367
1368 static void
1369 compat_copy_entry_from_user(struct compat_ip6t_entry *e, void **dstptr,
1370                             unsigned int *size,
1371                             struct xt_table_info *newinfo, unsigned char *base)
1372 {
1373         struct xt_entry_target *t;
1374         struct ip6t_entry *de;
1375         unsigned int origsize;
1376         int h;
1377         struct xt_entry_match *ematch;
1378
1379         origsize = *size;
1380         de = *dstptr;
1381         memcpy(de, e, sizeof(struct ip6t_entry));
1382         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1383
1384         *dstptr += sizeof(struct ip6t_entry);
1385         *size += sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1386
1387         xt_ematch_foreach(ematch, e)
1388                 xt_compat_match_from_user(ematch, dstptr, size);
1389
1390         de->target_offset = e->target_offset - (origsize - *size);
1391         t = compat_ip6t_get_target(e);
1392         xt_compat_target_from_user(t, dstptr, size);
1393
1394         de->next_offset = e->next_offset - (origsize - *size);
1395         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1396                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1397                         newinfo->hook_entry[h] -= origsize - *size;
1398                 if ((unsigned char *)de - base < newinfo->underflow[h])
1399                         newinfo->underflow[h] -= origsize - *size;
1400         }
1401 }
1402
1403 static int
1404 translate_compat_table(struct net *net,
1405                        struct xt_table_info **pinfo,
1406                        void **pentry0,
1407                        const struct compat_ip6t_replace *compatr)
1408 {
1409         unsigned int i, j;
1410         struct xt_table_info *newinfo, *info;
1411         void *pos, *entry0, *entry1;
1412         struct compat_ip6t_entry *iter0;
1413         struct ip6t_replace repl;
1414         unsigned int size;
1415         int ret;
1416
1417         info = *pinfo;
1418         entry0 = *pentry0;
1419         size = compatr->size;
1420         info->number = compatr->num_entries;
1421
1422         j = 0;
1423         xt_compat_lock(AF_INET6);
1424         ret = xt_compat_init_offsets(AF_INET6, compatr->num_entries);
1425         if (ret)
1426                 goto out_unlock;
1427         /* Walk through entries, checking offsets. */
1428         xt_entry_foreach(iter0, entry0, compatr->size) {
1429                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1430                                                         entry0,
1431                                                         entry0 + compatr->size);
1432                 if (ret != 0)
1433                         goto out_unlock;
1434                 ++j;
1435         }
1436
1437         ret = -EINVAL;
1438         if (j != compatr->num_entries)
1439                 goto out_unlock;
1440
1441         ret = -ENOMEM;
1442         newinfo = xt_alloc_table_info(size);
1443         if (!newinfo)
1444                 goto out_unlock;
1445
1446         newinfo->number = compatr->num_entries;
1447         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1448                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1449                 newinfo->underflow[i] = compatr->underflow[i];
1450         }
1451         entry1 = newinfo->entries;
1452         pos = entry1;
1453         size = compatr->size;
1454         xt_entry_foreach(iter0, entry0, compatr->size)
1455                 compat_copy_entry_from_user(iter0, &pos, &size,
1456                                             newinfo, entry1);
1457
1458         /* all module references in entry0 are now gone. */
1459         xt_compat_flush_offsets(AF_INET6);
1460         xt_compat_unlock(AF_INET6);
1461
1462         memcpy(&repl, compatr, sizeof(*compatr));
1463
1464         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1465                 repl.hook_entry[i] = newinfo->hook_entry[i];
1466                 repl.underflow[i] = newinfo->underflow[i];
1467         }
1468
1469         repl.num_counters = 0;
1470         repl.counters = NULL;
1471         repl.size = newinfo->size;
1472         ret = translate_table(net, newinfo, entry1, &repl);
1473         if (ret)
1474                 goto free_newinfo;
1475
1476         *pinfo = newinfo;
1477         *pentry0 = entry1;
1478         xt_free_table_info(info);
1479         return 0;
1480
1481 free_newinfo:
1482         xt_free_table_info(newinfo);
1483         return ret;
1484 out_unlock:
1485         xt_compat_flush_offsets(AF_INET6);
1486         xt_compat_unlock(AF_INET6);
1487         xt_entry_foreach(iter0, entry0, compatr->size) {
1488                 if (j-- == 0)
1489                         break;
1490                 compat_release_entry(iter0);
1491         }
1492         return ret;
1493 }
1494
1495 static int
1496 compat_do_replace(struct net *net, sockptr_t arg, unsigned int len)
1497 {
1498         int ret;
1499         struct compat_ip6t_replace tmp;
1500         struct xt_table_info *newinfo;
1501         void *loc_cpu_entry;
1502         struct ip6t_entry *iter;
1503
1504         if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0)
1505                 return -EFAULT;
1506
1507         /* overflow check */
1508         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1509                 return -ENOMEM;
1510         if (tmp.num_counters == 0)
1511                 return -EINVAL;
1512
1513         tmp.name[sizeof(tmp.name)-1] = 0;
1514
1515         newinfo = xt_alloc_table_info(tmp.size);
1516         if (!newinfo)
1517                 return -ENOMEM;
1518
1519         loc_cpu_entry = newinfo->entries;
1520         if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp),
1521                         tmp.size) != 0) {
1522                 ret = -EFAULT;
1523                 goto free_newinfo;
1524         }
1525
1526         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1527         if (ret != 0)
1528                 goto free_newinfo;
1529
1530         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1531                            tmp.num_counters, compat_ptr(tmp.counters));
1532         if (ret)
1533                 goto free_newinfo_untrans;
1534         return 0;
1535
1536  free_newinfo_untrans:
1537         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1538                 cleanup_entry(iter, net);
1539  free_newinfo:
1540         xt_free_table_info(newinfo);
1541         return ret;
1542 }
1543
1544 struct compat_ip6t_get_entries {
1545         char name[XT_TABLE_MAXNAMELEN];
1546         compat_uint_t size;
1547         struct compat_ip6t_entry entrytable[];
1548 };
1549
1550 static int
1551 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1552                             void __user *userptr)
1553 {
1554         struct xt_counters *counters;
1555         const struct xt_table_info *private = xt_table_get_private_protected(table);
1556         void __user *pos;
1557         unsigned int size;
1558         int ret = 0;
1559         unsigned int i = 0;
1560         struct ip6t_entry *iter;
1561
1562         counters = alloc_counters(table);
1563         if (IS_ERR(counters))
1564                 return PTR_ERR(counters);
1565
1566         pos = userptr;
1567         size = total_size;
1568         xt_entry_foreach(iter, private->entries, total_size) {
1569                 ret = compat_copy_entry_to_user(iter, &pos,
1570                                                 &size, counters, i++);
1571                 if (ret != 0)
1572                         break;
1573         }
1574
1575         vfree(counters);
1576         return ret;
1577 }
1578
1579 static int
1580 compat_get_entries(struct net *net, struct compat_ip6t_get_entries __user *uptr,
1581                    int *len)
1582 {
1583         int ret;
1584         struct compat_ip6t_get_entries get;
1585         struct xt_table *t;
1586
1587         if (*len < sizeof(get))
1588                 return -EINVAL;
1589
1590         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1591                 return -EFAULT;
1592
1593         if (*len != sizeof(struct compat_ip6t_get_entries) + get.size)
1594                 return -EINVAL;
1595
1596         get.name[sizeof(get.name) - 1] = '\0';
1597
1598         xt_compat_lock(AF_INET6);
1599         t = xt_find_table_lock(net, AF_INET6, get.name);
1600         if (!IS_ERR(t)) {
1601                 const struct xt_table_info *private = xt_table_get_private_protected(t);
1602                 struct xt_table_info info;
1603                 ret = compat_table_info(private, &info);
1604                 if (!ret && get.size == info.size)
1605                         ret = compat_copy_entries_to_user(private->size,
1606                                                           t, uptr->entrytable);
1607                 else if (!ret)
1608                         ret = -EAGAIN;
1609
1610                 xt_compat_flush_offsets(AF_INET6);
1611                 module_put(t->me);
1612                 xt_table_unlock(t);
1613         } else
1614                 ret = PTR_ERR(t);
1615
1616         xt_compat_unlock(AF_INET6);
1617         return ret;
1618 }
1619 #endif
1620
1621 static int
1622 do_ip6t_set_ctl(struct sock *sk, int cmd, sockptr_t arg, unsigned int len)
1623 {
1624         int ret;
1625
1626         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1627                 return -EPERM;
1628
1629         switch (cmd) {
1630         case IP6T_SO_SET_REPLACE:
1631 #ifdef CONFIG_COMPAT
1632                 if (in_compat_syscall())
1633                         ret = compat_do_replace(sock_net(sk), arg, len);
1634                 else
1635 #endif
1636                         ret = do_replace(sock_net(sk), arg, len);
1637                 break;
1638
1639         case IP6T_SO_SET_ADD_COUNTERS:
1640                 ret = do_add_counters(sock_net(sk), arg, len);
1641                 break;
1642
1643         default:
1644                 ret = -EINVAL;
1645         }
1646
1647         return ret;
1648 }
1649
1650 static int
1651 do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1652 {
1653         int ret;
1654
1655         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1656                 return -EPERM;
1657
1658         switch (cmd) {
1659         case IP6T_SO_GET_INFO:
1660                 ret = get_info(sock_net(sk), user, len);
1661                 break;
1662
1663         case IP6T_SO_GET_ENTRIES:
1664 #ifdef CONFIG_COMPAT
1665                 if (in_compat_syscall())
1666                         ret = compat_get_entries(sock_net(sk), user, len);
1667                 else
1668 #endif
1669                         ret = get_entries(sock_net(sk), user, len);
1670                 break;
1671
1672         case IP6T_SO_GET_REVISION_MATCH:
1673         case IP6T_SO_GET_REVISION_TARGET: {
1674                 struct xt_get_revision rev;
1675                 int target;
1676
1677                 if (*len != sizeof(rev)) {
1678                         ret = -EINVAL;
1679                         break;
1680                 }
1681                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1682                         ret = -EFAULT;
1683                         break;
1684                 }
1685                 rev.name[sizeof(rev.name)-1] = 0;
1686
1687                 if (cmd == IP6T_SO_GET_REVISION_TARGET)
1688                         target = 1;
1689                 else
1690                         target = 0;
1691
1692                 try_then_request_module(xt_find_revision(AF_INET6, rev.name,
1693                                                          rev.revision,
1694                                                          target, &ret),
1695                                         "ip6t_%s", rev.name);
1696                 break;
1697         }
1698
1699         default:
1700                 ret = -EINVAL;
1701         }
1702
1703         return ret;
1704 }
1705
1706 static void __ip6t_unregister_table(struct net *net, struct xt_table *table)
1707 {
1708         struct xt_table_info *private;
1709         void *loc_cpu_entry;
1710         struct module *table_owner = table->me;
1711         struct ip6t_entry *iter;
1712
1713         private = xt_unregister_table(table);
1714
1715         /* Decrease module usage counts and free resources */
1716         loc_cpu_entry = private->entries;
1717         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1718                 cleanup_entry(iter, net);
1719         if (private->number > private->initial_entries)
1720                 module_put(table_owner);
1721         xt_free_table_info(private);
1722 }
1723
1724 int ip6t_register_table(struct net *net, const struct xt_table *table,
1725                         const struct ip6t_replace *repl,
1726                         const struct nf_hook_ops *ops,
1727                         struct xt_table **res)
1728 {
1729         int ret;
1730         struct xt_table_info *newinfo;
1731         struct xt_table_info bootstrap = {0};
1732         void *loc_cpu_entry;
1733         struct xt_table *new_table;
1734
1735         newinfo = xt_alloc_table_info(repl->size);
1736         if (!newinfo)
1737                 return -ENOMEM;
1738
1739         loc_cpu_entry = newinfo->entries;
1740         memcpy(loc_cpu_entry, repl->entries, repl->size);
1741
1742         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1743         if (ret != 0)
1744                 goto out_free;
1745
1746         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1747         if (IS_ERR(new_table)) {
1748                 ret = PTR_ERR(new_table);
1749                 goto out_free;
1750         }
1751
1752         /* set res now, will see skbs right after nf_register_net_hooks */
1753         WRITE_ONCE(*res, new_table);
1754         if (!ops)
1755                 return 0;
1756
1757         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
1758         if (ret != 0) {
1759                 __ip6t_unregister_table(net, new_table);
1760                 *res = NULL;
1761         }
1762
1763         return ret;
1764
1765 out_free:
1766         xt_free_table_info(newinfo);
1767         return ret;
1768 }
1769
1770 void ip6t_unregister_table_pre_exit(struct net *net, struct xt_table *table,
1771                                     const struct nf_hook_ops *ops)
1772 {
1773         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
1774 }
1775
1776 void ip6t_unregister_table_exit(struct net *net, struct xt_table *table)
1777 {
1778         __ip6t_unregister_table(net, table);
1779 }
1780
1781 void ip6t_unregister_table(struct net *net, struct xt_table *table,
1782                            const struct nf_hook_ops *ops)
1783 {
1784         if (ops)
1785                 ip6t_unregister_table_pre_exit(net, table, ops);
1786         __ip6t_unregister_table(net, table);
1787 }
1788
1789 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1790 static inline bool
1791 icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1792                      u_int8_t type, u_int8_t code,
1793                      bool invert)
1794 {
1795         return (type == test_type && code >= min_code && code <= max_code)
1796                 ^ invert;
1797 }
1798
1799 static bool
1800 icmp6_match(const struct sk_buff *skb, struct xt_action_param *par)
1801 {
1802         const struct icmp6hdr *ic;
1803         struct icmp6hdr _icmph;
1804         const struct ip6t_icmp *icmpinfo = par->matchinfo;
1805
1806         /* Must not be a fragment. */
1807         if (par->fragoff != 0)
1808                 return false;
1809
1810         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1811         if (ic == NULL) {
1812                 /* We've been asked to examine this packet, and we
1813                  * can't.  Hence, no choice but to drop.
1814                  */
1815                 par->hotdrop = true;
1816                 return false;
1817         }
1818
1819         return icmp6_type_code_match(icmpinfo->type,
1820                                      icmpinfo->code[0],
1821                                      icmpinfo->code[1],
1822                                      ic->icmp6_type, ic->icmp6_code,
1823                                      !!(icmpinfo->invflags&IP6T_ICMP_INV));
1824 }
1825
1826 /* Called when user tries to insert an entry of this type. */
1827 static int icmp6_checkentry(const struct xt_mtchk_param *par)
1828 {
1829         const struct ip6t_icmp *icmpinfo = par->matchinfo;
1830
1831         /* Must specify no unknown invflags */
1832         return (icmpinfo->invflags & ~IP6T_ICMP_INV) ? -EINVAL : 0;
1833 }
1834
1835 /* The built-in targets: standard (NULL) and error. */
1836 static struct xt_target ip6t_builtin_tg[] __read_mostly = {
1837         {
1838                 .name             = XT_STANDARD_TARGET,
1839                 .targetsize       = sizeof(int),
1840                 .family           = NFPROTO_IPV6,
1841 #ifdef CONFIG_COMPAT
1842                 .compatsize       = sizeof(compat_int_t),
1843                 .compat_from_user = compat_standard_from_user,
1844                 .compat_to_user   = compat_standard_to_user,
1845 #endif
1846         },
1847         {
1848                 .name             = XT_ERROR_TARGET,
1849                 .target           = ip6t_error,
1850                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1851                 .family           = NFPROTO_IPV6,
1852         },
1853 };
1854
1855 static struct nf_sockopt_ops ip6t_sockopts = {
1856         .pf             = PF_INET6,
1857         .set_optmin     = IP6T_BASE_CTL,
1858         .set_optmax     = IP6T_SO_SET_MAX+1,
1859         .set            = do_ip6t_set_ctl,
1860         .get_optmin     = IP6T_BASE_CTL,
1861         .get_optmax     = IP6T_SO_GET_MAX+1,
1862         .get            = do_ip6t_get_ctl,
1863         .owner          = THIS_MODULE,
1864 };
1865
1866 static struct xt_match ip6t_builtin_mt[] __read_mostly = {
1867         {
1868                 .name       = "icmp6",
1869                 .match      = icmp6_match,
1870                 .matchsize  = sizeof(struct ip6t_icmp),
1871                 .checkentry = icmp6_checkentry,
1872                 .proto      = IPPROTO_ICMPV6,
1873                 .family     = NFPROTO_IPV6,
1874                 .me         = THIS_MODULE,
1875         },
1876 };
1877
1878 static int __net_init ip6_tables_net_init(struct net *net)
1879 {
1880         return xt_proto_init(net, NFPROTO_IPV6);
1881 }
1882
1883 static void __net_exit ip6_tables_net_exit(struct net *net)
1884 {
1885         xt_proto_fini(net, NFPROTO_IPV6);
1886 }
1887
1888 static struct pernet_operations ip6_tables_net_ops = {
1889         .init = ip6_tables_net_init,
1890         .exit = ip6_tables_net_exit,
1891 };
1892
1893 static int __init ip6_tables_init(void)
1894 {
1895         int ret;
1896
1897         ret = register_pernet_subsys(&ip6_tables_net_ops);
1898         if (ret < 0)
1899                 goto err1;
1900
1901         /* No one else will be downing sem now, so we won't sleep */
1902         ret = xt_register_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1903         if (ret < 0)
1904                 goto err2;
1905         ret = xt_register_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1906         if (ret < 0)
1907                 goto err4;
1908
1909         /* Register setsockopt */
1910         ret = nf_register_sockopt(&ip6t_sockopts);
1911         if (ret < 0)
1912                 goto err5;
1913
1914         return 0;
1915
1916 err5:
1917         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1918 err4:
1919         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1920 err2:
1921         unregister_pernet_subsys(&ip6_tables_net_ops);
1922 err1:
1923         return ret;
1924 }
1925
1926 static void __exit ip6_tables_fini(void)
1927 {
1928         nf_unregister_sockopt(&ip6t_sockopts);
1929
1930         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1931         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1932         unregister_pernet_subsys(&ip6_tables_net_ops);
1933 }
1934
1935 EXPORT_SYMBOL(ip6t_register_table);
1936 EXPORT_SYMBOL(ip6t_unregister_table);
1937 EXPORT_SYMBOL(ip6t_unregister_table_pre_exit);
1938 EXPORT_SYMBOL(ip6t_unregister_table_exit);
1939 EXPORT_SYMBOL(ip6t_do_table);
1940
1941 module_init(ip6_tables_init);
1942 module_exit(ip6_tables_fini);