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 <coreteam@netfilter.org>
   7  * Copyright (c) 2006-2010 Patrick McHardy <kaber@trash.net>
   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 <coreteam@netfilter.org>");
  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                  u16 *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(void *priv, struct sk_buff *skb,
 251               const struct nf_hook_state *state)
 252 {
 253         const struct xt_table *table = priv;
 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.fragoff = 0;
 277         acpar.hotdrop = false;
 278         acpar.state   = state;
 279
 280         WARN_ON(!(table->valid_hooks & (1 << hook)));
 281
 282         local_bh_disable();
 283         addend = xt_write_recseq_begin();
 284         private = READ_ONCE(table->private); /* Address dependency. */
 285         cpu        = smp_processor_id();
 286         table_base = private->entries;
 287         jumpstack  = (struct ip6t_entry **)private->jumpstack[cpu];
 288
 289         /* Switch to alternate jumpstack if we're being invoked via TEE.
 290          * TEE issues XT_CONTINUE verdict on original skb so we must not
 291          * clobber the jumpstack.
 292          *
 293          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 294          * but it is no problem since absolute verdict is issued by these.
 295          */
 296         if (static_key_false(&xt_tee_enabled))
 297                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 298
 299         e = get_entry(table_base, private->hook_entry[hook]);
 300
 301         do {
 302                 const struct xt_entry_target *t;
 303                 const struct xt_entry_match *ematch;
 304                 struct xt_counters *counter;
 305
 306                 WARN_ON(!e);
 307                 acpar.thoff = 0;
 308                 if (!ip6_packet_match(skb, indev, outdev, &e->ipv6,
 309                     &acpar.thoff, &acpar.fragoff, &acpar.hotdrop)) {
 310  no_match:
 311                         e = ip6t_next_entry(e);
 312                         continue;
 313                 }
 314
 315                 xt_ematch_foreach(ematch, e) {
 316                         acpar.match     = ematch->u.kernel.match;
 317                         acpar.matchinfo = ematch->data;
 318                         if (!acpar.match->match(skb, &acpar))
 319                                 goto no_match;
 320                 }
 321
 322                 counter = xt_get_this_cpu_counter(&e->counters);
 323                 ADD_COUNTER(*counter, skb->len, 1);
 324
 325                 t = ip6t_get_target_c(e);
 326                 WARN_ON(!t->u.kernel.target);
 327
 328 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 329                 /* The packet is traced: log it */
 330                 if (unlikely(skb->nf_trace))
 331                         trace_packet(state->net, skb, hook, state->in,
 332                                      state->out, table->name, private, e);
 333 #endif
 334                 /* Standard target? */
 335                 if (!t->u.kernel.target->target) {
 336                         int v;
 337
 338                         v = ((struct xt_standard_target *)t)->verdict;
 339                         if (v < 0) {
 340                                 /* Pop from stack? */
 341                                 if (v != XT_RETURN) {
 342                                         verdict = (unsigned int)(-v) - 1;
 343                                         break;
 344                                 }
 345                                 if (stackidx == 0)
 346                                         e = get_entry(table_base,
 347                                             private->underflow[hook]);
 348                                 else
 349                                         e = ip6t_next_entry(jumpstack[--stackidx]);
 350                                 continue;
 351                         }
 352                         if (table_base + v != ip6t_next_entry(e) &&
 353                             !(e->ipv6.flags & IP6T_F_GOTO)) {
 354                                 if (unlikely(stackidx >= private->stacksize)) {
 355                                         verdict = NF_DROP;
 356                                         break;
 357                                 }
 358                                 jumpstack[stackidx++] = e;
 359                         }
 360
 361                         e = get_entry(table_base, v);
 362                         continue;
 363                 }
 364
 365                 acpar.target   = t->u.kernel.target;
 366                 acpar.targinfo = t->data;
 367
 368                 verdict = t->u.kernel.target->target(skb, &acpar);
 369                 if (verdict == XT_CONTINUE)
 370                         e = ip6t_next_entry(e);
 371                 else
 372                         /* Verdict */
 373                         break;
 374         } while (!acpar.hotdrop);
 375
 376         xt_write_recseq_end(addend);
 377         local_bh_enable();
 378
 379         if (acpar.hotdrop)
 380                 return NF_DROP;
 381         else return verdict;
 382 }
 383
 384 /* Figures out from what hook each rule can be called: returns 0 if
 385    there are loops.  Puts hook bitmask in comefrom. */
 386 static int
 387 mark_source_chains(const struct xt_table_info *newinfo,
 388                    unsigned int valid_hooks, void *entry0,
 389                    unsigned int *offsets)
 390 {
 391         unsigned int hook;
 392
 393         /* No recursion; use packet counter to save back ptrs (reset
 394            to 0 as we leave), and comefrom to save source hook bitmask */
 395         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 396                 unsigned int pos = newinfo->hook_entry[hook];
 397                 struct ip6t_entry *e = entry0 + pos;
 398
 399                 if (!(valid_hooks & (1 << hook)))
 400                         continue;
 401
 402                 /* Set initial back pointer. */
 403                 e->counters.pcnt = pos;
 404
 405                 for (;;) {
 406                         const struct xt_standard_target *t
 407                                 = (void *)ip6t_get_target_c(e);
 408                         int visited = e->comefrom & (1 << hook);
 409
 410                         if (e->comefrom & (1 << NF_INET_NUMHOOKS))
 411                                 return 0;
 412
 413                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 414
 415                         /* Unconditional return/END. */
 416                         if ((unconditional(e) &&
 417                              (strcmp(t->target.u.user.name,
 418                                      XT_STANDARD_TARGET) == 0) &&
 419                              t->verdict < 0) || visited) {
 420                                 unsigned int oldpos, size;
 421
 422                                 /* Return: backtrack through the last
 423                                    big jump. */
 424                                 do {
 425                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 426                                         oldpos = pos;
 427                                         pos = e->counters.pcnt;
 428                                         e->counters.pcnt = 0;
 429
 430                                         /* We're at the start. */
 431                                         if (pos == oldpos)
 432                                                 goto next;
 433
 434                                         e = entry0 + pos;
 435                                 } while (oldpos == pos + e->next_offset);
 436
 437                                 /* Move along one */
 438                                 size = e->next_offset;
 439                                 e = entry0 + pos + size;
 440                                 if (pos + size >= newinfo->size)
 441                                         return 0;
 442                                 e->counters.pcnt = pos;
 443                                 pos += size;
 444                         } else {
 445                                 int newpos = t->verdict;
 446
 447                                 if (strcmp(t->target.u.user.name,
 448                                            XT_STANDARD_TARGET) == 0 &&
 449                                     newpos >= 0) {
 450                                         /* This a jump; chase it. */
 451                                         if (!xt_find_jump_offset(offsets, newpos,
 452                                                                  newinfo->number))
 453                                                 return 0;
 454                                 } else {
 455                                         /* ... this is a fallthru */
 456                                         newpos = pos + e->next_offset;
 457                                         if (newpos >= newinfo->size)
 458                                                 return 0;
 459                                 }
 460                                 e = entry0 + newpos;
 461                                 e->counters.pcnt = pos;
 462                                 pos = newpos;
 463                         }
 464                 }
 465 next:           ;
 466         }
 467         return 1;
 468 }
 469
 470 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 471 {
 472         struct xt_mtdtor_param par;
 473
 474         par.net       = net;
 475         par.match     = m->u.kernel.match;
 476         par.matchinfo = m->data;
 477         par.family    = NFPROTO_IPV6;
 478         if (par.match->destroy != NULL)
 479                 par.match->destroy(&par);
 480         module_put(par.match->me);
 481 }
 482
 483 static int check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 484 {
 485         const struct ip6t_ip6 *ipv6 = par->entryinfo;
 486
 487         par->match     = m->u.kernel.match;
 488         par->matchinfo = m->data;
 489
 490         return xt_check_match(par, m->u.match_size - sizeof(*m),
 491                               ipv6->proto, ipv6->invflags & IP6T_INV_PROTO);
 492 }
 493
 494 static int
 495 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 496 {
 497         struct xt_match *match;
 498         int ret;
 499
 500         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
 501                                       m->u.user.revision);
 502         if (IS_ERR(match))
 503                 return PTR_ERR(match);
 504
 505         m->u.kernel.match = match;
 506
 507         ret = check_match(m, par);
 508         if (ret)
 509                 goto err;
 510
 511         return 0;
 512 err:
 513         module_put(m->u.kernel.match->me);
 514         return ret;
 515 }
 516
 517 static int check_target(struct ip6t_entry *e, struct net *net, const char *name)
 518 {
 519         struct xt_entry_target *t = ip6t_get_target(e);
 520         struct xt_tgchk_param par = {
 521                 .net       = net,
 522                 .table     = name,
 523                 .entryinfo = e,
 524                 .target    = t->u.kernel.target,
 525                 .targinfo  = t->data,
 526                 .hook_mask = e->comefrom,
 527                 .family    = NFPROTO_IPV6,
 528         };
 529
 530         return xt_check_target(&par, t->u.target_size - sizeof(*t),
 531                                e->ipv6.proto,
 532                                e->ipv6.invflags & IP6T_INV_PROTO);
 533 }
 534
 535 static int
 536 find_check_entry(struct ip6t_entry *e, struct net *net, const char *name,
 537                  unsigned int size,
 538                  struct xt_percpu_counter_alloc_state *alloc_state)
 539 {
 540         struct xt_entry_target *t;
 541         struct xt_target *target;
 542         int ret;
 543         unsigned int j;
 544         struct xt_mtchk_param mtpar;
 545         struct xt_entry_match *ematch;
 546
 547         if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 548                 return -ENOMEM;
 549
 550         j = 0;
 551         memset(&mtpar, 0, sizeof(mtpar));
 552         mtpar.net       = net;
 553         mtpar.table     = name;
 554         mtpar.entryinfo = &e->ipv6;
 555         mtpar.hook_mask = e->comefrom;
 556         mtpar.family    = NFPROTO_IPV6;
 557         xt_ematch_foreach(ematch, e) {
 558                 ret = find_check_match(ematch, &mtpar);
 559                 if (ret != 0)
 560                         goto cleanup_matches;
 561                 ++j;
 562         }
 563
 564         t = ip6t_get_target(e);
 565         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
 566                                         t->u.user.revision);
 567         if (IS_ERR(target)) {
 568                 ret = PTR_ERR(target);
 569                 goto cleanup_matches;
 570         }
 571         t->u.kernel.target = target;
 572
 573         ret = check_target(e, net, name);
 574         if (ret)
 575                 goto err;
 576         return 0;
 577  err:
 578         module_put(t->u.kernel.target->me);
 579  cleanup_matches:
 580         xt_ematch_foreach(ematch, e) {
 581                 if (j-- == 0)
 582                         break;
 583                 cleanup_match(ematch, net);
 584         }
 585
 586         xt_percpu_counter_free(&e->counters);
 587
 588         return ret;
 589 }
 590
 591 static bool check_underflow(const struct ip6t_entry *e)
 592 {
 593         const struct xt_entry_target *t;
 594         unsigned int verdict;
 595
 596         if (!unconditional(e))
 597                 return false;
 598         t = ip6t_get_target_c(e);
 599         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 600                 return false;
 601         verdict = ((struct xt_standard_target *)t)->verdict;
 602         verdict = -verdict - 1;
 603         return verdict == NF_DROP || verdict == NF_ACCEPT;
 604 }
 605
 606 static int
 607 check_entry_size_and_hooks(struct ip6t_entry *e,
 608                            struct xt_table_info *newinfo,
 609                            const unsigned char *base,
 610                            const unsigned char *limit,
 611                            const unsigned int *hook_entries,
 612                            const unsigned int *underflows,
 613                            unsigned int valid_hooks)
 614 {
 615         unsigned int h;
 616         int err;
 617
 618         if ((unsigned long)e % __alignof__(struct ip6t_entry) != 0 ||
 619             (unsigned char *)e + sizeof(struct ip6t_entry) >= limit ||
 620             (unsigned char *)e + e->next_offset > limit)
 621                 return -EINVAL;
 622
 623         if (e->next_offset
 624             < sizeof(struct ip6t_entry) + sizeof(struct xt_entry_target))
 625                 return -EINVAL;
 626
 627         if (!ip6_checkentry(&e->ipv6))
 628                 return -EINVAL;
 629
 630         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 631                                      e->next_offset);
 632         if (err)
 633                 return err;
 634
 635         /* Check hooks & underflows */
 636         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 637                 if (!(valid_hooks & (1 << h)))
 638                         continue;
 639                 if ((unsigned char *)e - base == hook_entries[h])
 640                         newinfo->hook_entry[h] = hook_entries[h];
 641                 if ((unsigned char *)e - base == underflows[h]) {
 642                         if (!check_underflow(e))
 643                                 return -EINVAL;
 644
 645                         newinfo->underflow[h] = underflows[h];
 646                 }
 647         }
 648
 649         /* Clear counters and comefrom */
 650         e->counters = ((struct xt_counters) { 0, 0 });
 651         e->comefrom = 0;
 652         return 0;
 653 }
 654
 655 static void cleanup_entry(struct ip6t_entry *e, struct net *net)
 656 {
 657         struct xt_tgdtor_param par;
 658         struct xt_entry_target *t;
 659         struct xt_entry_match *ematch;
 660
 661         /* Cleanup all matches */
 662         xt_ematch_foreach(ematch, e)
 663                 cleanup_match(ematch, net);
 664         t = ip6t_get_target(e);
 665
 666         par.net      = net;
 667         par.target   = t->u.kernel.target;
 668         par.targinfo = t->data;
 669         par.family   = NFPROTO_IPV6;
 670         if (par.target->destroy != NULL)
 671                 par.target->destroy(&par);
 672         module_put(par.target->me);
 673         xt_percpu_counter_free(&e->counters);
 674 }
 675
 676 /* Checks and translates the user-supplied table segment (held in
 677    newinfo) */
 678 static int
 679 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 680                 const struct ip6t_replace *repl)
 681 {
 682         struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 683         struct ip6t_entry *iter;
 684         unsigned int *offsets;
 685         unsigned int i;
 686         int ret = 0;
 687
 688         newinfo->size = repl->size;
 689         newinfo->number = repl->num_entries;
 690
 691         /* Init all hooks to impossible value. */
 692         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 693                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 694                 newinfo->underflow[i] = 0xFFFFFFFF;
 695         }
 696
 697         offsets = xt_alloc_entry_offsets(newinfo->number);
 698         if (!offsets)
 699                 return -ENOMEM;
 700         i = 0;
 701         /* Walk through entries, checking offsets. */
 702         xt_entry_foreach(iter, entry0, newinfo->size) {
 703                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 704                                                  entry0 + repl->size,
 705                                                  repl->hook_entry,
 706                                                  repl->underflow,
 707                                                  repl->valid_hooks);
 708                 if (ret != 0)
 709                         goto out_free;
 710                 if (i < repl->num_entries)
 711                         offsets[i] = (void *)iter - entry0;
 712                 ++i;
 713                 if (strcmp(ip6t_get_target(iter)->u.user.name,
 714                     XT_ERROR_TARGET) == 0)
 715                         ++newinfo->stacksize;
 716         }
 717
 718         ret = -EINVAL;
 719         if (i != repl->num_entries)
 720                 goto out_free;
 721
 722         ret = xt_check_table_hooks(newinfo, repl->valid_hooks);
 723         if (ret)
 724                 goto out_free;
 725
 726         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 727                 ret = -ELOOP;
 728                 goto out_free;
 729         }
 730         kvfree(offsets);
 731
 732         /* Finally, each sanity check must pass */
 733         i = 0;
 734         xt_entry_foreach(iter, entry0, newinfo->size) {
 735                 ret = find_check_entry(iter, net, repl->name, repl->size,
 736                                        &alloc_state);
 737                 if (ret != 0)
 738                         break;
 739                 ++i;
 740         }
 741
 742         if (ret != 0) {
 743                 xt_entry_foreach(iter, entry0, newinfo->size) {
 744                         if (i-- == 0)
 745                                 break;
 746                         cleanup_entry(iter, net);
 747                 }
 748                 return ret;
 749         }
 750
 751         return ret;
 752  out_free:
 753         kvfree(offsets);
 754         return ret;
 755 }
 756
 757 static void
 758 get_counters(const struct xt_table_info *t,
 759              struct xt_counters counters[])
 760 {
 761         struct ip6t_entry *iter;
 762         unsigned int cpu;
 763         unsigned int i;
 764
 765         for_each_possible_cpu(cpu) {
 766                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 767
 768                 i = 0;
 769                 xt_entry_foreach(iter, t->entries, t->size) {
 770                         struct xt_counters *tmp;
 771                         u64 bcnt, pcnt;
 772                         unsigned int start;
 773
 774                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 775                         do {
 776                                 start = read_seqcount_begin(s);
 777                                 bcnt = tmp->bcnt;
 778                                 pcnt = tmp->pcnt;
 779                         } while (read_seqcount_retry(s, start));
 780
 781                         ADD_COUNTER(counters[i], bcnt, pcnt);
 782                         ++i;
 783                         cond_resched();
 784                 }
 785         }
 786 }
 787
 788 static void get_old_counters(const struct xt_table_info *t,
 789                              struct xt_counters counters[])
 790 {
 791         struct ip6t_entry *iter;
 792         unsigned int cpu, i;
 793
 794         for_each_possible_cpu(cpu) {
 795                 i = 0;
 796                 xt_entry_foreach(iter, t->entries, t->size) {
 797                         const struct xt_counters *tmp;
 798
 799                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 800                         ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
 801                         ++i;
 802                 }
 803                 cond_resched();
 804         }
 805 }
 806
 807 static struct xt_counters *alloc_counters(const struct xt_table *table)
 808 {
 809         unsigned int countersize;
 810         struct xt_counters *counters;
 811         const struct xt_table_info *private = table->private;
 812
 813         /* We need atomic snapshot of counters: rest doesn't change
 814            (other than comefrom, which userspace doesn't care
 815            about). */
 816         countersize = sizeof(struct xt_counters) * private->number;
 817         counters = vzalloc(countersize);
 818
 819         if (counters == NULL)
 820                 return ERR_PTR(-ENOMEM);
 821
 822         get_counters(private, counters);
 823
 824         return counters;
 825 }
 826
 827 static int
 828 copy_entries_to_user(unsigned int total_size,
 829                      const struct xt_table *table,
 830                      void __user *userptr)
 831 {
 832         unsigned int off, num;
 833         const struct ip6t_entry *e;
 834         struct xt_counters *counters;
 835         const struct xt_table_info *private = table->private;
 836         int ret = 0;
 837         const void *loc_cpu_entry;
 838
 839         counters = alloc_counters(table);
 840         if (IS_ERR(counters))
 841                 return PTR_ERR(counters);
 842
 843         loc_cpu_entry = private->entries;
 844
 845         /* FIXME: use iterator macros --RR */
 846         /* ... then go back and fix counters and names */
 847         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 848                 unsigned int i;
 849                 const struct xt_entry_match *m;
 850                 const struct xt_entry_target *t;
 851
 852                 e = loc_cpu_entry + off;
 853                 if (copy_to_user(userptr + off, e, sizeof(*e))) {
 854                         ret = -EFAULT;
 855                         goto free_counters;
 856                 }
 857                 if (copy_to_user(userptr + off
 858                                  + offsetof(struct ip6t_entry, counters),
 859                                  &counters[num],
 860                                  sizeof(counters[num])) != 0) {
 861                         ret = -EFAULT;
 862                         goto free_counters;
 863                 }
 864
 865                 for (i = sizeof(struct ip6t_entry);
 866                      i < e->target_offset;
 867                      i += m->u.match_size) {
 868                         m = (void *)e + i;
 869
 870                         if (xt_match_to_user(m, userptr + off + i)) {
 871                                 ret = -EFAULT;
 872                                 goto free_counters;
 873                         }
 874                 }
 875
 876                 t = ip6t_get_target_c(e);
 877                 if (xt_target_to_user(t, userptr + off + e->target_offset)) {
 878                         ret = -EFAULT;
 879                         goto free_counters;
 880                 }
 881         }
 882
 883  free_counters:
 884         vfree(counters);
 885         return ret;
 886 }
 887
 888 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
 889 static void compat_standard_from_user(void *dst, const void *src)
 890 {
 891         int v = *(compat_int_t *)src;
 892
 893         if (v > 0)
 894                 v += xt_compat_calc_jump(AF_INET6, v);
 895         memcpy(dst, &v, sizeof(v));
 896 }
 897
 898 static int compat_standard_to_user(void __user *dst, const void *src)
 899 {
 900         compat_int_t cv = *(int *)src;
 901
 902         if (cv > 0)
 903                 cv -= xt_compat_calc_jump(AF_INET6, cv);
 904         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
 905 }
 906
 907 static int compat_calc_entry(const struct ip6t_entry *e,
 908                              const struct xt_table_info *info,
 909                              const void *base, struct xt_table_info *newinfo)
 910 {
 911         const struct xt_entry_match *ematch;
 912         const struct xt_entry_target *t;
 913         unsigned int entry_offset;
 914         int off, i, ret;
 915
 916         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
 917         entry_offset = (void *)e - base;
 918         xt_ematch_foreach(ematch, e)
 919                 off += xt_compat_match_offset(ematch->u.kernel.match);
 920         t = ip6t_get_target_c(e);
 921         off += xt_compat_target_offset(t->u.kernel.target);
 922         newinfo->size -= off;
 923         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
 924         if (ret)
 925                 return ret;
 926
 927         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 928                 if (info->hook_entry[i] &&
 929                     (e < (struct ip6t_entry *)(base + info->hook_entry[i])))
 930                         newinfo->hook_entry[i] -= off;
 931                 if (info->underflow[i] &&
 932                     (e < (struct ip6t_entry *)(base + info->underflow[i])))
 933                         newinfo->underflow[i] -= off;
 934         }
 935         return 0;
 936 }
 937
 938 static int compat_table_info(const struct xt_table_info *info,
 939                              struct xt_table_info *newinfo)
 940 {
 941         struct ip6t_entry *iter;
 942         const void *loc_cpu_entry;
 943         int ret;
 944
 945         if (!newinfo || !info)
 946                 return -EINVAL;
 947
 948         /* we dont care about newinfo->entries */
 949         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
 950         newinfo->initial_entries = 0;
 951         loc_cpu_entry = info->entries;
 952         ret = xt_compat_init_offsets(AF_INET6, info->number);
 953         if (ret)
 954                 return ret;
 955         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
 956                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
 957                 if (ret != 0)
 958                         return ret;
 959         }
 960         return 0;
 961 }
 962 #endif
 963
 964 static int get_info(struct net *net, void __user *user, const int *len)
 965 {
 966         char name[XT_TABLE_MAXNAMELEN];
 967         struct xt_table *t;
 968         int ret;
 969
 970         if (*len != sizeof(struct ip6t_getinfo))
 971                 return -EINVAL;
 972
 973         if (copy_from_user(name, user, sizeof(name)) != 0)
 974                 return -EFAULT;
 975
 976         name[XT_TABLE_MAXNAMELEN-1] = '\0';
 977 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
 978         if (in_compat_syscall())
 979                 xt_compat_lock(AF_INET6);
 980 #endif
 981         t = xt_request_find_table_lock(net, AF_INET6, name);
 982         if (!IS_ERR(t)) {
 983                 struct ip6t_getinfo info;
 984                 const struct xt_table_info *private = t->private;
 985 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
 986                 struct xt_table_info tmp;
 987
 988                 if (in_compat_syscall()) {
 989                         ret = compat_table_info(private, &tmp);
 990                         xt_compat_flush_offsets(AF_INET6);
 991                         private = &tmp;
 992                 }
 993 #endif
 994                 memset(&info, 0, sizeof(info));
 995                 info.valid_hooks = t->valid_hooks;
 996                 memcpy(info.hook_entry, private->hook_entry,
 997                        sizeof(info.hook_entry));
 998                 memcpy(info.underflow, private->underflow,
 999                        sizeof(info.underflow));
1000                 info.num_entries = private->number;
1001                 info.size = private->size;
1002                 strcpy(info.name, name);
1003
1004                 if (copy_to_user(user, &info, *len) != 0)
1005                         ret = -EFAULT;
1006                 else
1007                         ret = 0;
1008
1009                 xt_table_unlock(t);
1010                 module_put(t->me);
1011         } else
1012                 ret = PTR_ERR(t);
1013 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1014         if (in_compat_syscall())
1015                 xt_compat_unlock(AF_INET6);
1016 #endif
1017         return ret;
1018 }
1019
1020 static int
1021 get_entries(struct net *net, struct ip6t_get_entries __user *uptr,
1022             const int *len)
1023 {
1024         int ret;
1025         struct ip6t_get_entries get;
1026         struct xt_table *t;
1027
1028         if (*len < sizeof(get))
1029                 return -EINVAL;
1030         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1031                 return -EFAULT;
1032         if (*len != sizeof(struct ip6t_get_entries) + get.size)
1033                 return -EINVAL;
1034
1035         get.name[sizeof(get.name) - 1] = '\0';
1036
1037         t = xt_find_table_lock(net, AF_INET6, get.name);
1038         if (!IS_ERR(t)) {
1039                 struct xt_table_info *private = t->private;
1040                 if (get.size == private->size)
1041                         ret = copy_entries_to_user(private->size,
1042                                                    t, uptr->entrytable);
1043                 else
1044                         ret = -EAGAIN;
1045
1046                 module_put(t->me);
1047                 xt_table_unlock(t);
1048         } else
1049                 ret = PTR_ERR(t);
1050
1051         return ret;
1052 }
1053
1054 static int
1055 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1056              struct xt_table_info *newinfo, unsigned int num_counters,
1057              void __user *counters_ptr)
1058 {
1059         int ret;
1060         struct xt_table *t;
1061         struct xt_table_info *oldinfo;
1062         struct xt_counters *counters;
1063         struct ip6t_entry *iter;
1064
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 0;
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 (len < sizeof(tmp))
1131                 return -EINVAL;
1132         if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0)
1133                 return -EFAULT;
1134
1135         /* overflow check */
1136         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1137                 return -ENOMEM;
1138         if (tmp.num_counters == 0)
1139                 return -EINVAL;
1140         if ((u64)len < (u64)tmp.size + sizeof(tmp))
1141                 return -EINVAL;
1142
1143         tmp.name[sizeof(tmp.name)-1] = 0;
1144
1145         newinfo = xt_alloc_table_info(tmp.size);
1146         if (!newinfo)
1147                 return -ENOMEM;
1148
1149         loc_cpu_entry = newinfo->entries;
1150         if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp),
1151                         tmp.size) != 0) {
1152                 ret = -EFAULT;
1153                 goto free_newinfo;
1154         }
1155
1156         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1157         if (ret != 0)
1158                 goto free_newinfo;
1159
1160         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1161                            tmp.num_counters, tmp.counters);
1162         if (ret)
1163                 goto free_newinfo_untrans;
1164         return 0;
1165
1166  free_newinfo_untrans:
1167         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1168                 cleanup_entry(iter, net);
1169  free_newinfo:
1170         xt_free_table_info(newinfo);
1171         return ret;
1172 }
1173
1174 static int
1175 do_add_counters(struct net *net, sockptr_t arg, unsigned int len)
1176 {
1177         unsigned int i;
1178         struct xt_counters_info tmp;
1179         struct xt_counters *paddc;
1180         struct xt_table *t;
1181         const struct xt_table_info *private;
1182         int ret = 0;
1183         struct ip6t_entry *iter;
1184         unsigned int addend;
1185
1186         paddc = xt_copy_counters(arg, len, &tmp);
1187         if (IS_ERR(paddc))
1188                 return PTR_ERR(paddc);
1189         t = xt_find_table_lock(net, AF_INET6, tmp.name);
1190         if (IS_ERR(t)) {
1191                 ret = PTR_ERR(t);
1192                 goto free;
1193         }
1194
1195         local_bh_disable();
1196         private = t->private;
1197         if (private->number != tmp.num_counters) {
1198                 ret = -EINVAL;
1199                 goto unlock_up_free;
1200         }
1201
1202         i = 0;
1203         addend = xt_write_recseq_begin();
1204         xt_entry_foreach(iter, private->entries, private->size) {
1205                 struct xt_counters *tmp;
1206
1207                 tmp = xt_get_this_cpu_counter(&iter->counters);
1208                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1209                 ++i;
1210         }
1211         xt_write_recseq_end(addend);
1212  unlock_up_free:
1213         local_bh_enable();
1214         xt_table_unlock(t);
1215         module_put(t->me);
1216  free:
1217         vfree(paddc);
1218
1219         return ret;
1220 }
1221
1222 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1223 struct compat_ip6t_replace {
1224         char                    name[XT_TABLE_MAXNAMELEN];
1225         u32                     valid_hooks;
1226         u32                     num_entries;
1227         u32                     size;
1228         u32                     hook_entry[NF_INET_NUMHOOKS];
1229         u32                     underflow[NF_INET_NUMHOOKS];
1230         u32                     num_counters;
1231         compat_uptr_t           counters;       /* struct xt_counters * */
1232         struct compat_ip6t_entry entries[];
1233 };
1234
1235 static int
1236 compat_copy_entry_to_user(struct ip6t_entry *e, void __user **dstptr,
1237                           unsigned int *size, struct xt_counters *counters,
1238                           unsigned int i)
1239 {
1240         struct xt_entry_target *t;
1241         struct compat_ip6t_entry __user *ce;
1242         u_int16_t target_offset, next_offset;
1243         compat_uint_t origsize;
1244         const struct xt_entry_match *ematch;
1245         int ret = 0;
1246
1247         origsize = *size;
1248         ce = *dstptr;
1249         if (copy_to_user(ce, e, sizeof(struct ip6t_entry)) != 0 ||
1250             copy_to_user(&ce->counters, &counters[i],
1251             sizeof(counters[i])) != 0)
1252                 return -EFAULT;
1253
1254         *dstptr += sizeof(struct compat_ip6t_entry);
1255         *size -= sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1256
1257         xt_ematch_foreach(ematch, e) {
1258                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1259                 if (ret != 0)
1260                         return ret;
1261         }
1262         target_offset = e->target_offset - (origsize - *size);
1263         t = ip6t_get_target(e);
1264         ret = xt_compat_target_to_user(t, dstptr, size);
1265         if (ret)
1266                 return ret;
1267         next_offset = e->next_offset - (origsize - *size);
1268         if (put_user(target_offset, &ce->target_offset) != 0 ||
1269             put_user(next_offset, &ce->next_offset) != 0)
1270                 return -EFAULT;
1271         return 0;
1272 }
1273
1274 static int
1275 compat_find_calc_match(struct xt_entry_match *m,
1276                        const struct ip6t_ip6 *ipv6,
1277                        int *size)
1278 {
1279         struct xt_match *match;
1280
1281         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
1282                                       m->u.user.revision);
1283         if (IS_ERR(match))
1284                 return PTR_ERR(match);
1285
1286         m->u.kernel.match = match;
1287         *size += xt_compat_match_offset(match);
1288         return 0;
1289 }
1290
1291 static void compat_release_entry(struct compat_ip6t_entry *e)
1292 {
1293         struct xt_entry_target *t;
1294         struct xt_entry_match *ematch;
1295
1296         /* Cleanup all matches */
1297         xt_ematch_foreach(ematch, e)
1298                 module_put(ematch->u.kernel.match->me);
1299         t = compat_ip6t_get_target(e);
1300         module_put(t->u.kernel.target->me);
1301 }
1302
1303 static int
1304 check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
1305                                   struct xt_table_info *newinfo,
1306                                   unsigned int *size,
1307                                   const unsigned char *base,
1308                                   const unsigned char *limit)
1309 {
1310         struct xt_entry_match *ematch;
1311         struct xt_entry_target *t;
1312         struct xt_target *target;
1313         unsigned int entry_offset;
1314         unsigned int j;
1315         int ret, off;
1316
1317         if ((unsigned long)e % __alignof__(struct compat_ip6t_entry) != 0 ||
1318             (unsigned char *)e + sizeof(struct compat_ip6t_entry) >= limit ||
1319             (unsigned char *)e + e->next_offset > limit)
1320                 return -EINVAL;
1321
1322         if (e->next_offset < sizeof(struct compat_ip6t_entry) +
1323                              sizeof(struct compat_xt_entry_target))
1324                 return -EINVAL;
1325
1326         if (!ip6_checkentry(&e->ipv6))
1327                 return -EINVAL;
1328
1329         ret = xt_compat_check_entry_offsets(e, e->elems,
1330                                             e->target_offset, e->next_offset);
1331         if (ret)
1332                 return ret;
1333
1334         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1335         entry_offset = (void *)e - (void *)base;
1336         j = 0;
1337         xt_ematch_foreach(ematch, e) {
1338                 ret = compat_find_calc_match(ematch, &e->ipv6, &off);
1339                 if (ret != 0)
1340                         goto release_matches;
1341                 ++j;
1342         }
1343
1344         t = compat_ip6t_get_target(e);
1345         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
1346                                         t->u.user.revision);
1347         if (IS_ERR(target)) {
1348                 ret = PTR_ERR(target);
1349                 goto release_matches;
1350         }
1351         t->u.kernel.target = target;
1352
1353         off += xt_compat_target_offset(target);
1354         *size += off;
1355         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1356         if (ret)
1357                 goto out;
1358
1359         return 0;
1360
1361 out:
1362         module_put(t->u.kernel.target->me);
1363 release_matches:
1364         xt_ematch_foreach(ematch, e) {
1365                 if (j-- == 0)
1366                         break;
1367                 module_put(ematch->u.kernel.match->me);
1368         }
1369         return ret;
1370 }
1371
1372 static void
1373 compat_copy_entry_from_user(struct compat_ip6t_entry *e, void **dstptr,
1374                             unsigned int *size,
1375                             struct xt_table_info *newinfo, unsigned char *base)
1376 {
1377         struct xt_entry_target *t;
1378         struct ip6t_entry *de;
1379         unsigned int origsize;
1380         int h;
1381         struct xt_entry_match *ematch;
1382
1383         origsize = *size;
1384         de = *dstptr;
1385         memcpy(de, e, sizeof(struct ip6t_entry));
1386         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1387
1388         *dstptr += sizeof(struct ip6t_entry);
1389         *size += sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1390
1391         xt_ematch_foreach(ematch, e)
1392                 xt_compat_match_from_user(ematch, dstptr, size);
1393
1394         de->target_offset = e->target_offset - (origsize - *size);
1395         t = compat_ip6t_get_target(e);
1396         xt_compat_target_from_user(t, dstptr, size);
1397
1398         de->next_offset = e->next_offset - (origsize - *size);
1399         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1400                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1401                         newinfo->hook_entry[h] -= origsize - *size;
1402                 if ((unsigned char *)de - base < newinfo->underflow[h])
1403                         newinfo->underflow[h] -= origsize - *size;
1404         }
1405 }
1406
1407 static int
1408 translate_compat_table(struct net *net,
1409                        struct xt_table_info **pinfo,
1410                        void **pentry0,
1411                        const struct compat_ip6t_replace *compatr)
1412 {
1413         unsigned int i, j;
1414         struct xt_table_info *newinfo, *info;
1415         void *pos, *entry0, *entry1;
1416         struct compat_ip6t_entry *iter0;
1417         struct ip6t_replace repl;
1418         unsigned int size;
1419         int ret;
1420
1421         info = *pinfo;
1422         entry0 = *pentry0;
1423         size = compatr->size;
1424         info->number = compatr->num_entries;
1425
1426         j = 0;
1427         xt_compat_lock(AF_INET6);
1428         ret = xt_compat_init_offsets(AF_INET6, compatr->num_entries);
1429         if (ret)
1430                 goto out_unlock;
1431         /* Walk through entries, checking offsets. */
1432         xt_entry_foreach(iter0, entry0, compatr->size) {
1433                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1434                                                         entry0,
1435                                                         entry0 + compatr->size);
1436                 if (ret != 0)
1437                         goto out_unlock;
1438                 ++j;
1439         }
1440
1441         ret = -EINVAL;
1442         if (j != compatr->num_entries)
1443                 goto out_unlock;
1444
1445         ret = -ENOMEM;
1446         newinfo = xt_alloc_table_info(size);
1447         if (!newinfo)
1448                 goto out_unlock;
1449
1450         memset(newinfo->entries, 0, size);
1451
1452         newinfo->number = compatr->num_entries;
1453         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1454                 newinfo->hook_entry[i] = compatr->hook_entry[i];
1455                 newinfo->underflow[i] = compatr->underflow[i];
1456         }
1457         entry1 = newinfo->entries;
1458         pos = entry1;
1459         size = compatr->size;
1460         xt_entry_foreach(iter0, entry0, compatr->size)
1461                 compat_copy_entry_from_user(iter0, &pos, &size,
1462                                             newinfo, entry1);
1463
1464         /* all module references in entry0 are now gone. */
1465         xt_compat_flush_offsets(AF_INET6);
1466         xt_compat_unlock(AF_INET6);
1467
1468         memcpy(&repl, compatr, sizeof(*compatr));
1469
1470         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1471                 repl.hook_entry[i] = newinfo->hook_entry[i];
1472                 repl.underflow[i] = newinfo->underflow[i];
1473         }
1474
1475         repl.num_counters = 0;
1476         repl.counters = NULL;
1477         repl.size = newinfo->size;
1478         ret = translate_table(net, newinfo, entry1, &repl);
1479         if (ret)
1480                 goto free_newinfo;
1481
1482         *pinfo = newinfo;
1483         *pentry0 = entry1;
1484         xt_free_table_info(info);
1485         return 0;
1486
1487 free_newinfo:
1488         xt_free_table_info(newinfo);
1489         return ret;
1490 out_unlock:
1491         xt_compat_flush_offsets(AF_INET6);
1492         xt_compat_unlock(AF_INET6);
1493         xt_entry_foreach(iter0, entry0, compatr->size) {
1494                 if (j-- == 0)
1495                         break;
1496                 compat_release_entry(iter0);
1497         }
1498         return ret;
1499 }
1500
1501 static int
1502 compat_do_replace(struct net *net, sockptr_t arg, unsigned int len)
1503 {
1504         int ret;
1505         struct compat_ip6t_replace tmp;
1506         struct xt_table_info *newinfo;
1507         void *loc_cpu_entry;
1508         struct ip6t_entry *iter;
1509
1510         if (len < sizeof(tmp))
1511                 return -EINVAL;
1512         if (copy_from_sockptr(&tmp, arg, sizeof(tmp)) != 0)
1513                 return -EFAULT;
1514
1515         /* overflow check */
1516         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1517                 return -ENOMEM;
1518         if (tmp.num_counters == 0)
1519                 return -EINVAL;
1520         if ((u64)len < (u64)tmp.size + sizeof(tmp))
1521                 return -EINVAL;
1522
1523         tmp.name[sizeof(tmp.name)-1] = 0;
1524
1525         newinfo = xt_alloc_table_info(tmp.size);
1526         if (!newinfo)
1527                 return -ENOMEM;
1528
1529         loc_cpu_entry = newinfo->entries;
1530         if (copy_from_sockptr_offset(loc_cpu_entry, arg, sizeof(tmp),
1531                         tmp.size) != 0) {
1532                 ret = -EFAULT;
1533                 goto free_newinfo;
1534         }
1535
1536         ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1537         if (ret != 0)
1538                 goto free_newinfo;
1539
1540         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1541                            tmp.num_counters, compat_ptr(tmp.counters));
1542         if (ret)
1543                 goto free_newinfo_untrans;
1544         return 0;
1545
1546  free_newinfo_untrans:
1547         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1548                 cleanup_entry(iter, net);
1549  free_newinfo:
1550         xt_free_table_info(newinfo);
1551         return ret;
1552 }
1553
1554 struct compat_ip6t_get_entries {
1555         char name[XT_TABLE_MAXNAMELEN];
1556         compat_uint_t size;
1557         struct compat_ip6t_entry entrytable[];
1558 };
1559
1560 static int
1561 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1562                             void __user *userptr)
1563 {
1564         struct xt_counters *counters;
1565         const struct xt_table_info *private = table->private;
1566         void __user *pos;
1567         unsigned int size;
1568         int ret = 0;
1569         unsigned int i = 0;
1570         struct ip6t_entry *iter;
1571
1572         counters = alloc_counters(table);
1573         if (IS_ERR(counters))
1574                 return PTR_ERR(counters);
1575
1576         pos = userptr;
1577         size = total_size;
1578         xt_entry_foreach(iter, private->entries, total_size) {
1579                 ret = compat_copy_entry_to_user(iter, &pos,
1580                                                 &size, counters, i++);
1581                 if (ret != 0)
1582                         break;
1583         }
1584
1585         vfree(counters);
1586         return ret;
1587 }
1588
1589 static int
1590 compat_get_entries(struct net *net, struct compat_ip6t_get_entries __user *uptr,
1591                    int *len)
1592 {
1593         int ret;
1594         struct compat_ip6t_get_entries get;
1595         struct xt_table *t;
1596
1597         if (*len < sizeof(get))
1598                 return -EINVAL;
1599
1600         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1601                 return -EFAULT;
1602
1603         if (*len != sizeof(struct compat_ip6t_get_entries) + get.size)
1604                 return -EINVAL;
1605
1606         get.name[sizeof(get.name) - 1] = '\0';
1607
1608         xt_compat_lock(AF_INET6);
1609         t = xt_find_table_lock(net, AF_INET6, get.name);
1610         if (!IS_ERR(t)) {
1611                 const struct xt_table_info *private = t->private;
1612                 struct xt_table_info info;
1613                 ret = compat_table_info(private, &info);
1614                 if (!ret && get.size == info.size)
1615                         ret = compat_copy_entries_to_user(private->size,
1616                                                           t, uptr->entrytable);
1617                 else if (!ret)
1618                         ret = -EAGAIN;
1619
1620                 xt_compat_flush_offsets(AF_INET6);
1621                 module_put(t->me);
1622                 xt_table_unlock(t);
1623         } else
1624                 ret = PTR_ERR(t);
1625
1626         xt_compat_unlock(AF_INET6);
1627         return ret;
1628 }
1629 #endif
1630
1631 static int
1632 do_ip6t_set_ctl(struct sock *sk, int cmd, sockptr_t arg, unsigned int len)
1633 {
1634         int ret;
1635
1636         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1637                 return -EPERM;
1638
1639         switch (cmd) {
1640         case IP6T_SO_SET_REPLACE:
1641 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1642                 if (in_compat_syscall())
1643                         ret = compat_do_replace(sock_net(sk), arg, len);
1644                 else
1645 #endif
1646                         ret = do_replace(sock_net(sk), arg, len);
1647                 break;
1648
1649         case IP6T_SO_SET_ADD_COUNTERS:
1650                 ret = do_add_counters(sock_net(sk), arg, len);
1651                 break;
1652
1653         default:
1654                 ret = -EINVAL;
1655         }
1656
1657         return ret;
1658 }
1659
1660 static int
1661 do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1662 {
1663         int ret;
1664
1665         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1666                 return -EPERM;
1667
1668         switch (cmd) {
1669         case IP6T_SO_GET_INFO:
1670                 ret = get_info(sock_net(sk), user, len);
1671                 break;
1672
1673         case IP6T_SO_GET_ENTRIES:
1674 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1675                 if (in_compat_syscall())
1676                         ret = compat_get_entries(sock_net(sk), user, len);
1677                 else
1678 #endif
1679                         ret = get_entries(sock_net(sk), user, len);
1680                 break;
1681
1682         case IP6T_SO_GET_REVISION_MATCH:
1683         case IP6T_SO_GET_REVISION_TARGET: {
1684                 struct xt_get_revision rev;
1685                 int target;
1686
1687                 if (*len != sizeof(rev)) {
1688                         ret = -EINVAL;
1689                         break;
1690                 }
1691                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1692                         ret = -EFAULT;
1693                         break;
1694                 }
1695                 rev.name[sizeof(rev.name)-1] = 0;
1696
1697                 if (cmd == IP6T_SO_GET_REVISION_TARGET)
1698                         target = 1;
1699                 else
1700                         target = 0;
1701
1702                 try_then_request_module(xt_find_revision(AF_INET6, rev.name,
1703                                                          rev.revision,
1704                                                          target, &ret),
1705                                         "ip6t_%s", rev.name);
1706                 break;
1707         }
1708
1709         default:
1710                 ret = -EINVAL;
1711         }
1712
1713         return ret;
1714 }
1715
1716 static void __ip6t_unregister_table(struct net *net, struct xt_table *table)
1717 {
1718         struct xt_table_info *private;
1719         void *loc_cpu_entry;
1720         struct module *table_owner = table->me;
1721         struct ip6t_entry *iter;
1722
1723         private = xt_unregister_table(table);
1724
1725         /* Decrease module usage counts and free resources */
1726         loc_cpu_entry = private->entries;
1727         xt_entry_foreach(iter, loc_cpu_entry, private->size)
1728                 cleanup_entry(iter, net);
1729         if (private->number > private->initial_entries)
1730                 module_put(table_owner);
1731         xt_free_table_info(private);
1732 }
1733
1734 int ip6t_register_table(struct net *net, const struct xt_table *table,
1735                         const struct ip6t_replace *repl,
1736                         const struct nf_hook_ops *template_ops)
1737 {
1738         struct nf_hook_ops *ops;
1739         unsigned int num_ops;
1740         int ret, i;
1741         struct xt_table_info *newinfo;
1742         struct xt_table_info bootstrap = {0};
1743         void *loc_cpu_entry;
1744         struct xt_table *new_table;
1745
1746         newinfo = xt_alloc_table_info(repl->size);
1747         if (!newinfo)
1748                 return -ENOMEM;
1749
1750         loc_cpu_entry = newinfo->entries;
1751         memcpy(loc_cpu_entry, repl->entries, repl->size);
1752
1753         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1754         if (ret != 0) {
1755                 xt_free_table_info(newinfo);
1756                 return ret;
1757         }
1758
1759         new_table = xt_register_table(net, table, &bootstrap, newinfo);
1760         if (IS_ERR(new_table)) {
1761                 struct ip6t_entry *iter;
1762
1763                 xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1764                         cleanup_entry(iter, net);
1765                 xt_free_table_info(newinfo);
1766                 return PTR_ERR(new_table);
1767         }
1768
1769         if (!template_ops)
1770                 return 0;
1771
1772         num_ops = hweight32(table->valid_hooks);
1773         if (num_ops == 0) {
1774                 ret = -EINVAL;
1775                 goto out_free;
1776         }
1777
1778         ops = kmemdup(template_ops, sizeof(*ops) * num_ops, GFP_KERNEL);
1779         if (!ops) {
1780                 ret = -ENOMEM;
1781                 goto out_free;
1782         }
1783
1784         for (i = 0; i < num_ops; i++)
1785                 ops[i].priv = new_table;
1786
1787         new_table->ops = ops;
1788
1789         ret = nf_register_net_hooks(net, ops, num_ops);
1790         if (ret != 0)
1791                 goto out_free;
1792
1793         return ret;
1794
1795 out_free:
1796         __ip6t_unregister_table(net, new_table);
1797         return ret;
1798 }
1799
1800 void ip6t_unregister_table_pre_exit(struct net *net, const char *name)
1801 {
1802         struct xt_table *table = xt_find_table(net, NFPROTO_IPV6, name);
1803
1804         if (table)
1805                 nf_unregister_net_hooks(net, table->ops, hweight32(table->valid_hooks));
1806 }
1807
1808 void ip6t_unregister_table_exit(struct net *net, const char *name)
1809 {
1810         struct xt_table *table = xt_find_table(net, NFPROTO_IPV6, name);
1811
1812         if (table)
1813                 __ip6t_unregister_table(net, table);
1814 }
1815
1816 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
1817 static inline bool
1818 icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1819                      u_int8_t type, u_int8_t code,
1820                      bool invert)
1821 {
1822         return (type == test_type && code >= min_code && code <= max_code)
1823                 ^ invert;
1824 }
1825
1826 static bool
1827 icmp6_match(const struct sk_buff *skb, struct xt_action_param *par)
1828 {
1829         const struct icmp6hdr *ic;
1830         struct icmp6hdr _icmph;
1831         const struct ip6t_icmp *icmpinfo = par->matchinfo;
1832
1833         /* Must not be a fragment. */
1834         if (par->fragoff != 0)
1835                 return false;
1836
1837         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
1838         if (ic == NULL) {
1839                 /* We've been asked to examine this packet, and we
1840                  * can't.  Hence, no choice but to drop.
1841                  */
1842                 par->hotdrop = true;
1843                 return false;
1844         }
1845
1846         return icmp6_type_code_match(icmpinfo->type,
1847                                      icmpinfo->code[0],
1848                                      icmpinfo->code[1],
1849                                      ic->icmp6_type, ic->icmp6_code,
1850                                      !!(icmpinfo->invflags&IP6T_ICMP_INV));
1851 }
1852
1853 /* Called when user tries to insert an entry of this type. */
1854 static int icmp6_checkentry(const struct xt_mtchk_param *par)
1855 {
1856         const struct ip6t_icmp *icmpinfo = par->matchinfo;
1857
1858         /* Must specify no unknown invflags */
1859         return (icmpinfo->invflags & ~IP6T_ICMP_INV) ? -EINVAL : 0;
1860 }
1861
1862 /* The built-in targets: standard (NULL) and error. */
1863 static struct xt_target ip6t_builtin_tg[] __read_mostly = {
1864         {
1865                 .name             = XT_STANDARD_TARGET,
1866                 .targetsize       = sizeof(int),
1867                 .family           = NFPROTO_IPV6,
1868 #ifdef CONFIG_NETFILTER_XTABLES_COMPAT
1869                 .compatsize       = sizeof(compat_int_t),
1870                 .compat_from_user = compat_standard_from_user,
1871                 .compat_to_user   = compat_standard_to_user,
1872 #endif
1873         },
1874         {
1875                 .name             = XT_ERROR_TARGET,
1876                 .target           = ip6t_error,
1877                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
1878                 .family           = NFPROTO_IPV6,
1879         },
1880 };
1881
1882 static struct nf_sockopt_ops ip6t_sockopts = {
1883         .pf             = PF_INET6,
1884         .set_optmin     = IP6T_BASE_CTL,
1885         .set_optmax     = IP6T_SO_SET_MAX+1,
1886         .set            = do_ip6t_set_ctl,
1887         .get_optmin     = IP6T_BASE_CTL,
1888         .get_optmax     = IP6T_SO_GET_MAX+1,
1889         .get            = do_ip6t_get_ctl,
1890         .owner          = THIS_MODULE,
1891 };
1892
1893 static struct xt_match ip6t_builtin_mt[] __read_mostly = {
1894         {
1895                 .name       = "icmp6",
1896                 .match      = icmp6_match,
1897                 .matchsize  = sizeof(struct ip6t_icmp),
1898                 .checkentry = icmp6_checkentry,
1899                 .proto      = IPPROTO_ICMPV6,
1900                 .family     = NFPROTO_IPV6,
1901                 .me         = THIS_MODULE,
1902         },
1903 };
1904
1905 static int __net_init ip6_tables_net_init(struct net *net)
1906 {
1907         return xt_proto_init(net, NFPROTO_IPV6);
1908 }
1909
1910 static void __net_exit ip6_tables_net_exit(struct net *net)
1911 {
1912         xt_proto_fini(net, NFPROTO_IPV6);
1913 }
1914
1915 static struct pernet_operations ip6_tables_net_ops = {
1916         .init = ip6_tables_net_init,
1917         .exit = ip6_tables_net_exit,
1918 };
1919
1920 static int __init ip6_tables_init(void)
1921 {
1922         int ret;
1923
1924         ret = register_pernet_subsys(&ip6_tables_net_ops);
1925         if (ret < 0)
1926                 goto err1;
1927
1928         /* No one else will be downing sem now, so we won't sleep */
1929         ret = xt_register_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1930         if (ret < 0)
1931                 goto err2;
1932         ret = xt_register_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1933         if (ret < 0)
1934                 goto err4;
1935
1936         /* Register setsockopt */
1937         ret = nf_register_sockopt(&ip6t_sockopts);
1938         if (ret < 0)
1939                 goto err5;
1940
1941         return 0;
1942
1943 err5:
1944         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1945 err4:
1946         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1947 err2:
1948         unregister_pernet_subsys(&ip6_tables_net_ops);
1949 err1:
1950         return ret;
1951 }
1952
1953 static void __exit ip6_tables_fini(void)
1954 {
1955         nf_unregister_sockopt(&ip6t_sockopts);
1956
1957         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
1958         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
1959         unregister_pernet_subsys(&ip6_tables_net_ops);
1960 }
1961
1962 EXPORT_SYMBOL(ip6t_register_table);
1963 EXPORT_SYMBOL(ip6t_unregister_table_pre_exit);
1964 EXPORT_SYMBOL(ip6t_unregister_table_exit);
1965 EXPORT_SYMBOL(ip6t_do_table);
1966
1967 module_init(ip6_tables_init);
1968 module_exit(ip6_tables_fini);