net/ipv6/netfilter/ip6_tables.c

   1 /*
   2  * Packet matching code.
   3  *
   4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   6  * Copyright (c) 2006-2010 Patrick McHardy <kaber@trash.net>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12
  13 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  14
  15 #include <linux/kernel.h>
  16 #include <linux/capability.h>
  17 #include <linux/in.h>
  18 #include <linux/skbuff.h>
  19 #include <linux/kmod.h>
  20 #include <linux/vmalloc.h>
  21 #include <linux/netdevice.h>
  22 #include <linux/module.h>
  23 #include <linux/poison.h>
  24 #include <linux/icmpv6.h>
  25 #include <net/ipv6.h>
  26 #include <net/compat.h>
  27 #include <asm/uaccess.h>
  28 #include <linux/mutex.h>
  29 #include <linux/proc_fs.h>
  30 #include <linux/err.h>
  31 #include <linux/cpumask.h>
  32
  33 #include <linux/netfilter_ipv6/ip6_tables.h>
  34 #include <linux/netfilter/x_tables.h>
  35 #include <net/netfilter/nf_log.h>
  36 #include "../../netfilter/xt_repldata.h"
  37
  38 MODULE_LICENSE("GPL");
  39 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  40 MODULE_DESCRIPTION("IPv6 packet filter");
  41
  42 /*#define DEBUG_IP_FIREWALL*/
  43 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  44 /*#define DEBUG_IP_FIREWALL_USER*/
  45
  46 #ifdef DEBUG_IP_FIREWALL
  47 #define dprintf(format, args...) pr_info(format , ## args)
  48 #else
  49 #define dprintf(format, args...)
  50 #endif
  51
  52 #ifdef DEBUG_IP_FIREWALL_USER
  53 #define duprintf(format, args...) pr_info(format , ## args)
  54 #else
  55 #define duprintf(format, args...)
  56 #endif
  57
  58 #ifdef CONFIG_NETFILTER_DEBUG
  59 #define IP_NF_ASSERT(x) WARN_ON(!(x))
  60 #else
  61 #define IP_NF_ASSERT(x)
  62 #endif
  63
  64 #if 0
  65 /* All the better to debug you with... */
  66 #define static
  67 #define inline
  68 #endif
  69
  70 void *ip6t_alloc_initial_table(const struct xt_table *info)
  71 {
  72         return xt_alloc_initial_table(ip6t, IP6T);
  73 }
  74 EXPORT_SYMBOL_GPL(ip6t_alloc_initial_table);
  75
  76 /*
  77    We keep a set of rules for each CPU, so we can avoid write-locking
  78    them in the softirq when updating the counters and therefore
  79    only need to read-lock in the softirq; doing a write_lock_bh() in user
  80    context stops packets coming through and allows user context to read
  81    the counters or update the rules.
  82
  83    Hence the start of any table is given by get_table() below.  */
  84
  85 /* Returns whether matches rule or not. */
  86 /* Performance critical - called for every packet */
  87 static inline bool
  88 ip6_packet_match(const struct sk_buff *skb,
  89                  const char *indev,
  90                  const char *outdev,
  91                  const struct ip6t_ip6 *ip6info,
  92                  unsigned int *protoff,
  93                  int *fragoff, bool *hotdrop)
  94 {
  95         unsigned long ret;
  96         const struct ipv6hdr *ipv6 = ipv6_hdr(skb);
  97
  98 #define FWINV(bool, invflg) ((bool) ^ !!(ip6info->invflags & (invflg)))
  99
 100         if (FWINV(ipv6_masked_addr_cmp(&ipv6->saddr, &ip6info->smsk,
 101                                        &ip6info->src), IP6T_INV_SRCIP) ||
 102             FWINV(ipv6_masked_addr_cmp(&ipv6->daddr, &ip6info->dmsk,
 103                                        &ip6info->dst), IP6T_INV_DSTIP)) {
 104                 dprintf("Source or dest mismatch.\n");
 105 /*
 106                 dprintf("SRC: %u. Mask: %u. Target: %u.%s\n", ip->saddr,
 107                         ipinfo->smsk.s_addr, ipinfo->src.s_addr,
 108                         ipinfo->invflags & IP6T_INV_SRCIP ? " (INV)" : "");
 109                 dprintf("DST: %u. Mask: %u. Target: %u.%s\n", ip->daddr,
 110                         ipinfo->dmsk.s_addr, ipinfo->dst.s_addr,
 111                         ipinfo->invflags & IP6T_INV_DSTIP ? " (INV)" : "");*/
 112                 return false;
 113         }
 114
 115         ret = ifname_compare_aligned(indev, ip6info->iniface, ip6info->iniface_mask);
 116
 117         if (FWINV(ret != 0, IP6T_INV_VIA_IN)) {
 118                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 119                         indev, ip6info->iniface,
 120                         ip6info->invflags & IP6T_INV_VIA_IN ? " (INV)" : "");
 121                 return false;
 122         }
 123
 124         ret = ifname_compare_aligned(outdev, ip6info->outiface, ip6info->outiface_mask);
 125
 126         if (FWINV(ret != 0, IP6T_INV_VIA_OUT)) {
 127                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 128                         outdev, ip6info->outiface,
 129                         ip6info->invflags & IP6T_INV_VIA_OUT ? " (INV)" : "");
 130                 return false;
 131         }
 132
 133 /* ... might want to do something with class and flowlabel here ... */
 134
 135         /* look for the desired protocol header */
 136         if (ip6info->flags & IP6T_F_PROTO) {
 137                 int protohdr;
 138                 unsigned short _frag_off;
 139
 140                 protohdr = ipv6_find_hdr(skb, protoff, -1, &_frag_off, NULL);
 141                 if (protohdr < 0) {
 142                         if (_frag_off == 0)
 143                                 *hotdrop = true;
 144                         return false;
 145                 }
 146                 *fragoff = _frag_off;
 147
 148                 dprintf("Packet protocol %hi ?= %s%hi.\n",
 149                                 protohdr,
 150                                 ip6info->invflags & IP6T_INV_PROTO ? "!":"",
 151                                 ip6info->proto);
 152
 153                 if (ip6info->proto == protohdr) {
 154                         if (ip6info->invflags & IP6T_INV_PROTO)
 155                                 return false;
 156
 157                         return true;
 158                 }
 159
 160                 /* We need match for the '-p all', too! */
 161                 if ((ip6info->proto != 0) &&
 162                         !(ip6info->invflags & IP6T_INV_PROTO))
 163                         return false;
 164         }
 165         return true;
 166 }
 167
 168 /* should be ip6 safe */
 169 static bool
 170 ip6_checkentry(const struct ip6t_ip6 *ipv6)
 171 {
 172         if (ipv6->flags & ~IP6T_F_MASK) {
 173                 duprintf("Unknown flag bits set: %08X\n",
 174                          ipv6->flags & ~IP6T_F_MASK);
 175                 return false;
 176         }
 177         if (ipv6->invflags & ~IP6T_INV_MASK) {
 178                 duprintf("Unknown invflag bits set: %08X\n",
 179                          ipv6->invflags & ~IP6T_INV_MASK);
 180                 return false;
 181         }
 182         return true;
 183 }
 184
 185 static unsigned int
 186 ip6t_error(struct sk_buff *skb, const struct xt_action_param *par)
 187 {
 188         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 189
 190         return NF_DROP;
 191 }
 192
 193 static inline struct ip6t_entry *
 194 get_entry(const void *base, unsigned int offset)
 195 {
 196         return (struct ip6t_entry *)(base + offset);
 197 }
 198
 199 /* All zeroes == unconditional rule. */
 200 /* Mildly perf critical (only if packet tracing is on) */
 201 static inline bool unconditional(const struct ip6t_entry *e)
 202 {
 203         static const struct ip6t_ip6 uncond;
 204
 205         return e->target_offset == sizeof(struct ip6t_entry) &&
 206                memcmp(&e->ipv6, &uncond, sizeof(uncond)) == 0;
 207 }
 208
 209 static inline const struct xt_entry_target *
 210 ip6t_get_target_c(const struct ip6t_entry *e)
 211 {
 212         return ip6t_get_target((struct ip6t_entry *)e);
 213 }
 214
 215 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 216 /* This cries for unification! */
 217 static const char *const hooknames[] = {
 218         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 219         [NF_INET_LOCAL_IN]              = "INPUT",
 220         [NF_INET_FORWARD]               = "FORWARD",
 221         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 222         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 223 };
 224
 225 enum nf_ip_trace_comments {
 226         NF_IP6_TRACE_COMMENT_RULE,
 227         NF_IP6_TRACE_COMMENT_RETURN,
 228         NF_IP6_TRACE_COMMENT_POLICY,
 229 };
 230
 231 static const char *const comments[] = {
 232         [NF_IP6_TRACE_COMMENT_RULE]     = "rule",
 233         [NF_IP6_TRACE_COMMENT_RETURN]   = "return",
 234         [NF_IP6_TRACE_COMMENT_POLICY]   = "policy",
 235 };
 236
 237 static struct nf_loginfo trace_loginfo = {
 238         .type = NF_LOG_TYPE_LOG,
 239         .u = {
 240                 .log = {
 241                         .level = LOGLEVEL_WARNING,
 242                         .logflags = NF_LOG_MASK,
 243                 },
 244         },
 245 };
 246
 247 /* Mildly perf critical (only if packet tracing is on) */
 248 static inline int
 249 get_chainname_rulenum(const struct ip6t_entry *s, const struct ip6t_entry *e,
 250                       const char *hookname, const char **chainname,
 251                       const char **comment, unsigned int *rulenum)
 252 {
 253         const struct xt_standard_target *t = (void *)ip6t_get_target_c(s);
 254
 255         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 256                 /* Head of user chain: ERROR target with chainname */
 257                 *chainname = t->target.data;
 258                 (*rulenum) = 0;
 259         } else if (s == e) {
 260                 (*rulenum)++;
 261
 262                 if (unconditional(s) &&
 263                     strcmp(t->target.u.kernel.target->name,
 264                            XT_STANDARD_TARGET) == 0 &&
 265                     t->verdict < 0) {
 266                         /* Tail of chains: STANDARD target (return/policy) */
 267                         *comment = *chainname == hookname
 268                                 ? comments[NF_IP6_TRACE_COMMENT_POLICY]
 269                                 : comments[NF_IP6_TRACE_COMMENT_RETURN];
 270                 }
 271                 return 1;
 272         } else
 273                 (*rulenum)++;
 274
 275         return 0;
 276 }
 277
 278 static void trace_packet(struct net *net,
 279                          const struct sk_buff *skb,
 280                          unsigned int hook,
 281                          const struct net_device *in,
 282                          const struct net_device *out,
 283                          const char *tablename,
 284                          const struct xt_table_info *private,
 285                          const struct ip6t_entry *e)
 286 {
 287         const struct ip6t_entry *root;
 288         const char *hookname, *chainname, *comment;
 289         const struct ip6t_entry *iter;
 290         unsigned int rulenum = 0;
 291
 292         root = get_entry(private->entries, private->hook_entry[hook]);
 293
 294         hookname = chainname = hooknames[hook];
 295         comment = comments[NF_IP6_TRACE_COMMENT_RULE];
 296
 297         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 298                 if (get_chainname_rulenum(iter, e, hookname,
 299                     &chainname, &comment, &rulenum) != 0)
 300                         break;
 301
 302         nf_log_trace(net, AF_INET6, hook, skb, in, out, &trace_loginfo,
 303                      "TRACE: %s:%s:%s:%u ",
 304                      tablename, chainname, comment, rulenum);
 305 }
 306 #endif
 307
 308 static inline struct ip6t_entry *
 309 ip6t_next_entry(const struct ip6t_entry *entry)
 310 {
 311         return (void *)entry + entry->next_offset;
 312 }
 313
 314 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 315 unsigned int
 316 ip6t_do_table(struct sk_buff *skb,
 317               const struct nf_hook_state *state,
 318               struct xt_table *table)
 319 {
 320         unsigned int hook = state->hook;
 321         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 322         /* Initializing verdict to NF_DROP keeps gcc happy. */
 323         unsigned int verdict = NF_DROP;
 324         const char *indev, *outdev;
 325         const void *table_base;
 326         struct ip6t_entry *e, **jumpstack;
 327         unsigned int stackidx, cpu;
 328         const struct xt_table_info *private;
 329         struct xt_action_param acpar;
 330         unsigned int addend;
 331
 332         /* Initialization */
 333         stackidx = 0;
 334         indev = state->in ? state->in->name : nulldevname;
 335         outdev = state->out ? state->out->name : nulldevname;
 336         /* We handle fragments by dealing with the first fragment as
 337          * if it was a normal packet.  All other fragments are treated
 338          * normally, except that they will NEVER match rules that ask
 339          * things we don't know, ie. tcp syn flag or ports).  If the
 340          * rule is also a fragment-specific rule, non-fragments won't
 341          * match it. */
 342         acpar.hotdrop = false;
 343         acpar.net     = state->net;
 344         acpar.in      = state->in;
 345         acpar.out     = state->out;
 346         acpar.family  = NFPROTO_IPV6;
 347         acpar.hooknum = hook;
 348
 349         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 350
 351         local_bh_disable();
 352         addend = xt_write_recseq_begin();
 353         private = table->private;
 354         /*
 355          * Ensure we load private-> members after we've fetched the base
 356          * pointer.
 357          */
 358         smp_read_barrier_depends();
 359         cpu        = smp_processor_id();
 360         table_base = private->entries;
 361         jumpstack  = (struct ip6t_entry **)private->jumpstack[cpu];
 362
 363         /* Switch to alternate jumpstack if we're being invoked via TEE.
 364          * TEE issues XT_CONTINUE verdict on original skb so we must not
 365          * clobber the jumpstack.
 366          *
 367          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 368          * but it is no problem since absolute verdict is issued by these.
 369          */
 370         if (static_key_false(&xt_tee_enabled))
 371                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 372
 373         e = get_entry(table_base, private->hook_entry[hook]);
 374
 375         do {
 376                 const struct xt_entry_target *t;
 377                 const struct xt_entry_match *ematch;
 378                 struct xt_counters *counter;
 379
 380                 IP_NF_ASSERT(e);
 381                 acpar.thoff = 0;
 382                 if (!ip6_packet_match(skb, indev, outdev, &e->ipv6,
 383                     &acpar.thoff, &acpar.fragoff, &acpar.hotdrop)) {
 384  no_match:
 385                         e = ip6t_next_entry(e);
 386                         continue;
 387                 }
 388
 389                 xt_ematch_foreach(ematch, e) {
 390                         acpar.match     = ematch->u.kernel.match;
 391                         acpar.matchinfo = ematch->data;
 392                         if (!acpar.match->match(skb, &acpar))
 393                                 goto no_match;
 394                 }
 395
 396                 counter = xt_get_this_cpu_counter(&e->counters);
 397                 ADD_COUNTER(*counter, skb->len, 1);
 398
 399                 t = ip6t_get_target_c(e);
 400                 IP_NF_ASSERT(t->u.kernel.target);
 401
 402 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 403                 /* The packet is traced: log it */
 404                 if (unlikely(skb->nf_trace))
 405                         trace_packet(state->net, skb, hook, state->in,
 406                                      state->out, table->name, private, e);
 407 #endif
 408                 /* Standard target? */
 409                 if (!t->u.kernel.target->target) {
 410                         int v;
 411
 412                         v = ((struct xt_standard_target *)t)->verdict;
 413                         if (v < 0) {
 414                                 /* Pop from stack? */
 415                                 if (v != XT_RETURN) {
 416                                         verdict = (unsigned int)(-v) - 1;
 417                                         break;
 418                                 }
 419                                 if (stackidx == 0)
 420                                         e = get_entry(table_base,
 421                                             private->underflow[hook]);
 422                                 else
 423                                         e = ip6t_next_entry(jumpstack[--stackidx]);
 424                                 continue;
 425                         }
 426                         if (table_base + v != ip6t_next_entry(e) &&
 427                             !(e->ipv6.flags & IP6T_F_GOTO)) {
 428                                 jumpstack[stackidx++] = e;
 429                         }
 430
 431                         e = get_entry(table_base, v);
 432                         continue;
 433                 }
 434
 435                 acpar.target   = t->u.kernel.target;
 436                 acpar.targinfo = t->data;
 437
 438                 verdict = t->u.kernel.target->target(skb, &acpar);
 439                 if (verdict == XT_CONTINUE)
 440                         e = ip6t_next_entry(e);
 441                 else
 442                         /* Verdict */
 443                         break;
 444         } while (!acpar.hotdrop);
 445
 446         xt_write_recseq_end(addend);
 447         local_bh_enable();
 448
 449 #ifdef DEBUG_ALLOW_ALL
 450         return NF_ACCEPT;
 451 #else
 452         if (acpar.hotdrop)
 453                 return NF_DROP;
 454         else return verdict;
 455 #endif
 456 }
 457
 458 static bool find_jump_target(const struct xt_table_info *t,
 459                              const struct ip6t_entry *target)
 460 {
 461         struct ip6t_entry *iter;
 462
 463         xt_entry_foreach(iter, t->entries, t->size) {
 464                  if (iter == target)
 465                         return true;
 466         }
 467         return false;
 468 }
 469
 470 /* Figures out from what hook each rule can be called: returns 0 if
 471    there are loops.  Puts hook bitmask in comefrom. */
 472 static int
 473 mark_source_chains(const struct xt_table_info *newinfo,
 474                    unsigned int valid_hooks, void *entry0)
 475 {
 476         unsigned int hook;
 477
 478         /* No recursion; use packet counter to save back ptrs (reset
 479            to 0 as we leave), and comefrom to save source hook bitmask */
 480         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 481                 unsigned int pos = newinfo->hook_entry[hook];
 482                 struct ip6t_entry *e = (struct ip6t_entry *)(entry0 + pos);
 483
 484                 if (!(valid_hooks & (1 << hook)))
 485                         continue;
 486
 487                 /* Set initial back pointer. */
 488                 e->counters.pcnt = pos;
 489
 490                 for (;;) {
 491                         const struct xt_standard_target *t
 492                                 = (void *)ip6t_get_target_c(e);
 493                         int visited = e->comefrom & (1 << hook);
 494
 495                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 496                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 497                                        hook, pos, e->comefrom);
 498                                 return 0;
 499                         }
 500                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 501
 502                         /* Unconditional return/END. */
 503                         if ((unconditional(e) &&
 504                              (strcmp(t->target.u.user.name,
 505                                      XT_STANDARD_TARGET) == 0) &&
 506                              t->verdict < 0) || visited) {
 507                                 unsigned int oldpos, size;
 508
 509                                 if ((strcmp(t->target.u.user.name,
 510                                             XT_STANDARD_TARGET) == 0) &&
 511                                     t->verdict < -NF_MAX_VERDICT - 1) {
 512                                         duprintf("mark_source_chains: bad "
 513                                                 "negative verdict (%i)\n",
 514                                                                 t->verdict);
 515                                         return 0;
 516                                 }
 517
 518                                 /* Return: backtrack through the last
 519                                    big jump. */
 520                                 do {
 521                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 522 #ifdef DEBUG_IP_FIREWALL_USER
 523                                         if (e->comefrom
 524                                             & (1 << NF_INET_NUMHOOKS)) {
 525                                                 duprintf("Back unset "
 526                                                          "on hook %u "
 527                                                          "rule %u\n",
 528                                                          hook, pos);
 529                                         }
 530 #endif
 531                                         oldpos = pos;
 532                                         pos = e->counters.pcnt;
 533                                         e->counters.pcnt = 0;
 534
 535                                         /* We're at the start. */
 536                                         if (pos == oldpos)
 537                                                 goto next;
 538
 539                                         e = (struct ip6t_entry *)
 540                                                 (entry0 + pos);
 541                                 } while (oldpos == pos + e->next_offset);
 542
 543                                 /* Move along one */
 544                                 size = e->next_offset;
 545                                 e = (struct ip6t_entry *)
 546                                         (entry0 + pos + size);
 547                                 if (pos + size >= newinfo->size)
 548                                         return 0;
 549                                 e->counters.pcnt = pos;
 550                                 pos += size;
 551                         } else {
 552                                 int newpos = t->verdict;
 553
 554                                 if (strcmp(t->target.u.user.name,
 555                                            XT_STANDARD_TARGET) == 0 &&
 556                                     newpos >= 0) {
 557                                         if (newpos > newinfo->size -
 558                                                 sizeof(struct ip6t_entry)) {
 559                                                 duprintf("mark_source_chains: "
 560                                                         "bad verdict (%i)\n",
 561                                                                 newpos);
 562                                                 return 0;
 563                                         }
 564                                         /* This a jump; chase it. */
 565                                         duprintf("Jump rule %u -> %u\n",
 566                                                  pos, newpos);
 567                                         e = (struct ip6t_entry *)
 568                                                 (entry0 + newpos);
 569                                         if (!find_jump_target(newinfo, e))
 570                                                 return 0;
 571                                 } else {
 572                                         /* ... this is a fallthru */
 573                                         newpos = pos + e->next_offset;
 574                                         if (newpos >= newinfo->size)
 575                                                 return 0;
 576                                 }
 577                                 e = (struct ip6t_entry *)
 578                                         (entry0 + newpos);
 579                                 e->counters.pcnt = pos;
 580                                 pos = newpos;
 581                         }
 582                 }
 583 next:
 584                 duprintf("Finished chain %u\n", hook);
 585         }
 586         return 1;
 587 }
 588
 589 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 590 {
 591         struct xt_mtdtor_param par;
 592
 593         par.net       = net;
 594         par.match     = m->u.kernel.match;
 595         par.matchinfo = m->data;
 596         par.family    = NFPROTO_IPV6;
 597         if (par.match->destroy != NULL)
 598                 par.match->destroy(&par);
 599         module_put(par.match->me);
 600 }
 601
 602 static int check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 603 {
 604         const struct ip6t_ip6 *ipv6 = par->entryinfo;
 605         int ret;
 606
 607         par->match     = m->u.kernel.match;
 608         par->matchinfo = m->data;
 609
 610         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 611                              ipv6->proto, ipv6->invflags & IP6T_INV_PROTO);
 612         if (ret < 0) {
 613                 duprintf("ip_tables: check failed for `%s'.\n",
 614                          par.match->name);
 615                 return ret;
 616         }
 617         return 0;
 618 }
 619
 620 static int
 621 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 622 {
 623         struct xt_match *match;
 624         int ret;
 625
 626         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
 627                                       m->u.user.revision);
 628         if (IS_ERR(match)) {
 629                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 630                 return PTR_ERR(match);
 631         }
 632         m->u.kernel.match = match;
 633
 634         ret = check_match(m, par);
 635         if (ret)
 636                 goto err;
 637
 638         return 0;
 639 err:
 640         module_put(m->u.kernel.match->me);
 641         return ret;
 642 }
 643
 644 static int check_target(struct ip6t_entry *e, struct net *net, const char *name)
 645 {
 646         struct xt_entry_target *t = ip6t_get_target(e);
 647         struct xt_tgchk_param par = {
 648                 .net       = net,
 649                 .table     = name,
 650                 .entryinfo = e,
 651                 .target    = t->u.kernel.target,
 652                 .targinfo  = t->data,
 653                 .hook_mask = e->comefrom,
 654                 .family    = NFPROTO_IPV6,
 655         };
 656         int ret;
 657
 658         t = ip6t_get_target(e);
 659         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 660               e->ipv6.proto, e->ipv6.invflags & IP6T_INV_PROTO);
 661         if (ret < 0) {
 662                 duprintf("ip_tables: check failed for `%s'.\n",
 663                          t->u.kernel.target->name);
 664                 return ret;
 665         }
 666         return 0;
 667 }
 668
 669 static int
 670 find_check_entry(struct ip6t_entry *e, struct net *net, const char *name,
 671                  unsigned int size)
 672 {
 673         struct xt_entry_target *t;
 674         struct xt_target *target;
 675         int ret;
 676         unsigned int j;
 677         struct xt_mtchk_param mtpar;
 678         struct xt_entry_match *ematch;
 679
 680         e->counters.pcnt = xt_percpu_counter_alloc();
 681         if (IS_ERR_VALUE(e->counters.pcnt))
 682                 return -ENOMEM;
 683
 684         j = 0;
 685         mtpar.net       = net;
 686         mtpar.table     = name;
 687         mtpar.entryinfo = &e->ipv6;
 688         mtpar.hook_mask = e->comefrom;
 689         mtpar.family    = NFPROTO_IPV6;
 690         xt_ematch_foreach(ematch, e) {
 691                 ret = find_check_match(ematch, &mtpar);
 692                 if (ret != 0)
 693                         goto cleanup_matches;
 694                 ++j;
 695         }
 696
 697         t = ip6t_get_target(e);
 698         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
 699                                         t->u.user.revision);
 700         if (IS_ERR(target)) {
 701                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 702                 ret = PTR_ERR(target);
 703                 goto cleanup_matches;
 704         }
 705         t->u.kernel.target = target;
 706
 707         ret = check_target(e, net, name);
 708         if (ret)
 709                 goto err;
 710         return 0;
 711  err:
 712         module_put(t->u.kernel.target->me);
 713  cleanup_matches:
 714         xt_ematch_foreach(ematch, e) {
 715                 if (j-- == 0)
 716                         break;
 717                 cleanup_match(ematch, net);
 718         }
 719
 720         xt_percpu_counter_free(e->counters.pcnt);
 721
 722         return ret;
 723 }
 724
 725 static bool check_underflow(const struct ip6t_entry *e)
 726 {
 727         const struct xt_entry_target *t;
 728         unsigned int verdict;
 729
 730         if (!unconditional(e))
 731                 return false;
 732         t = ip6t_get_target_c(e);
 733         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 734                 return false;
 735         verdict = ((struct xt_standard_target *)t)->verdict;
 736         verdict = -verdict - 1;
 737         return verdict == NF_DROP || verdict == NF_ACCEPT;
 738 }
 739
 740 static int
 741 check_entry_size_and_hooks(struct ip6t_entry *e,
 742                            struct xt_table_info *newinfo,
 743                            const unsigned char *base,
 744                            const unsigned char *limit,
 745                            const unsigned int *hook_entries,
 746                            const unsigned int *underflows,
 747                            unsigned int valid_hooks)
 748 {
 749         unsigned int h;
 750         int err;
 751
 752         if ((unsigned long)e % __alignof__(struct ip6t_entry) != 0 ||
 753             (unsigned char *)e + sizeof(struct ip6t_entry) >= limit ||
 754             (unsigned char *)e + e->next_offset > limit) {
 755                 duprintf("Bad offset %p\n", e);
 756                 return -EINVAL;
 757         }
 758
 759         if (e->next_offset
 760             < sizeof(struct ip6t_entry) + sizeof(struct xt_entry_target)) {
 761                 duprintf("checking: element %p size %u\n",
 762                          e, e->next_offset);
 763                 return -EINVAL;
 764         }
 765
 766         if (!ip6_checkentry(&e->ipv6))
 767                 return -EINVAL;
 768
 769         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 770                                      e->next_offset);
 771         if (err)
 772                 return err;
 773
 774         /* Check hooks & underflows */
 775         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 776                 if (!(valid_hooks & (1 << h)))
 777                         continue;
 778                 if ((unsigned char *)e - base == hook_entries[h])
 779                         newinfo->hook_entry[h] = hook_entries[h];
 780                 if ((unsigned char *)e - base == underflows[h]) {
 781                         if (!check_underflow(e)) {
 782                                 pr_debug("Underflows must be unconditional and "
 783                                          "use the STANDARD target with "
 784                                          "ACCEPT/DROP\n");
 785                                 return -EINVAL;
 786                         }
 787                         newinfo->underflow[h] = underflows[h];
 788                 }
 789         }
 790
 791         /* Clear counters and comefrom */
 792         e->counters = ((struct xt_counters) { 0, 0 });
 793         e->comefrom = 0;
 794         return 0;
 795 }
 796
 797 static void cleanup_entry(struct ip6t_entry *e, struct net *net)
 798 {
 799         struct xt_tgdtor_param par;
 800         struct xt_entry_target *t;
 801         struct xt_entry_match *ematch;
 802
 803         /* Cleanup all matches */
 804         xt_ematch_foreach(ematch, e)
 805                 cleanup_match(ematch, net);
 806         t = ip6t_get_target(e);
 807
 808         par.net      = net;
 809         par.target   = t->u.kernel.target;
 810         par.targinfo = t->data;
 811         par.family   = NFPROTO_IPV6;
 812         if (par.target->destroy != NULL)
 813                 par.target->destroy(&par);
 814         module_put(par.target->me);
 815
 816         xt_percpu_counter_free(e->counters.pcnt);
 817 }
 818
 819 /* Checks and translates the user-supplied table segment (held in
 820    newinfo) */
 821 static int
 822 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 823                 const struct ip6t_replace *repl)
 824 {
 825         struct ip6t_entry *iter;
 826         unsigned int i;
 827         int ret = 0;
 828
 829         newinfo->size = repl->size;
 830         newinfo->number = repl->num_entries;
 831
 832         /* Init all hooks to impossible value. */
 833         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 834                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 835                 newinfo->underflow[i] = 0xFFFFFFFF;
 836         }
 837
 838         duprintf("translate_table: size %u\n", newinfo->size);
 839         i = 0;
 840         /* Walk through entries, checking offsets. */
 841         xt_entry_foreach(iter, entry0, newinfo->size) {
 842                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 843                                                  entry0 + repl->size,
 844                                                  repl->hook_entry,
 845                                                  repl->underflow,
 846                                                  repl->valid_hooks);
 847                 if (ret != 0)
 848                         return ret;
 849                 ++i;
 850                 if (strcmp(ip6t_get_target(iter)->u.user.name,
 851                     XT_ERROR_TARGET) == 0)
 852                         ++newinfo->stacksize;
 853         }
 854
 855         if (i != repl->num_entries) {
 856                 duprintf("translate_table: %u not %u entries\n",
 857                          i, repl->num_entries);
 858                 return -EINVAL;
 859         }
 860
 861         /* Check hooks all assigned */
 862         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 863                 /* Only hooks which are valid */
 864                 if (!(repl->valid_hooks & (1 << i)))
 865                         continue;
 866                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 867                         duprintf("Invalid hook entry %u %u\n",
 868                                  i, repl->hook_entry[i]);
 869                         return -EINVAL;
 870                 }
 871                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 872                         duprintf("Invalid underflow %u %u\n",
 873                                  i, repl->underflow[i]);
 874                         return -EINVAL;
 875                 }
 876         }
 877
 878         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 879                 return -ELOOP;
 880
 881         /* Finally, each sanity check must pass */
 882         i = 0;
 883         xt_entry_foreach(iter, entry0, newinfo->size) {
 884                 ret = find_check_entry(iter, net, repl->name, repl->size);
 885                 if (ret != 0)
 886                         break;
 887                 ++i;
 888         }
 889
 890         if (ret != 0) {
 891                 xt_entry_foreach(iter, entry0, newinfo->size) {
 892                         if (i-- == 0)
 893                                 break;
 894                         cleanup_entry(iter, net);
 895                 }
 896                 return ret;
 897         }
 898
 899         return ret;
 900 }
 901
 902 static void
 903 get_counters(const struct xt_table_info *t,
 904              struct xt_counters counters[])
 905 {
 906         struct ip6t_entry *iter;
 907         unsigned int cpu;
 908         unsigned int i;
 909
 910         for_each_possible_cpu(cpu) {
 911                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 912
 913                 i = 0;
 914                 xt_entry_foreach(iter, t->entries, t->size) {
 915                         struct xt_counters *tmp;
 916                         u64 bcnt, pcnt;
 917                         unsigned int start;
 918
 919                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 920                         do {
 921                                 start = read_seqcount_begin(s);
 922                                 bcnt = tmp->bcnt;
 923                                 pcnt = tmp->pcnt;
 924                         } while (read_seqcount_retry(s, start));
 925
 926                         ADD_COUNTER(counters[i], bcnt, pcnt);
 927                         ++i;
 928                 }
 929         }
 930 }
 931
 932 static struct xt_counters *alloc_counters(const struct xt_table *table)
 933 {
 934         unsigned int countersize;
 935         struct xt_counters *counters;
 936         const struct xt_table_info *private = table->private;
 937
 938         /* We need atomic snapshot of counters: rest doesn't change
 939            (other than comefrom, which userspace doesn't care
 940            about). */
 941         countersize = sizeof(struct xt_counters) * private->number;
 942         counters = vzalloc(countersize);
 943
 944         if (counters == NULL)
 945                 return ERR_PTR(-ENOMEM);
 946
 947         get_counters(private, counters);
 948
 949         return counters;
 950 }
 951
 952 static int
 953 copy_entries_to_user(unsigned int total_size,
 954                      const struct xt_table *table,
 955                      void __user *userptr)
 956 {
 957         unsigned int off, num;
 958         const struct ip6t_entry *e;
 959         struct xt_counters *counters;
 960         const struct xt_table_info *private = table->private;
 961         int ret = 0;
 962         const void *loc_cpu_entry;
 963
 964         counters = alloc_counters(table);
 965         if (IS_ERR(counters))
 966                 return PTR_ERR(counters);
 967
 968         loc_cpu_entry = private->entries;
 969         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 970                 ret = -EFAULT;
 971                 goto free_counters;
 972         }
 973
 974         /* FIXME: use iterator macros --RR */
 975         /* ... then go back and fix counters and names */
 976         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 977                 unsigned int i;
 978                 const struct xt_entry_match *m;
 979                 const struct xt_entry_target *t;
 980
 981                 e = (struct ip6t_entry *)(loc_cpu_entry + off);
 982                 if (copy_to_user(userptr + off
 983                                  + offsetof(struct ip6t_entry, counters),
 984                                  &counters[num],
 985                                  sizeof(counters[num])) != 0) {
 986                         ret = -EFAULT;
 987                         goto free_counters;
 988                 }
 989
 990                 for (i = sizeof(struct ip6t_entry);
 991                      i < e->target_offset;
 992                      i += m->u.match_size) {
 993                         m = (void *)e + i;
 994
 995                         if (copy_to_user(userptr + off + i
 996                                          + offsetof(struct xt_entry_match,
 997                                                     u.user.name),
 998                                          m->u.kernel.match->name,
 999                                          strlen(m->u.kernel.match->name)+1)
1000                             != 0) {
1001                                 ret = -EFAULT;
1002                                 goto free_counters;
1003                         }
1004                 }
1005
1006                 t = ip6t_get_target_c(e);
1007                 if (copy_to_user(userptr + off + e->target_offset
1008                                  + offsetof(struct xt_entry_target,
1009                                             u.user.name),
1010                                  t->u.kernel.target->name,
1011                                  strlen(t->u.kernel.target->name)+1) != 0) {
1012                         ret = -EFAULT;
1013                         goto free_counters;
1014                 }
1015         }
1016
1017  free_counters:
1018         vfree(counters);
1019         return ret;
1020 }
1021
1022 #ifdef CONFIG_COMPAT
1023 static void compat_standard_from_user(void *dst, const void *src)
1024 {
1025         int v = *(compat_int_t *)src;
1026
1027         if (v > 0)
1028                 v += xt_compat_calc_jump(AF_INET6, v);
1029         memcpy(dst, &v, sizeof(v));
1030 }
1031
1032 static int compat_standard_to_user(void __user *dst, const void *src)
1033 {
1034         compat_int_t cv = *(int *)src;
1035
1036         if (cv > 0)
1037                 cv -= xt_compat_calc_jump(AF_INET6, cv);
1038         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1039 }
1040
1041 static int compat_calc_entry(const struct ip6t_entry *e,
1042                              const struct xt_table_info *info,
1043                              const void *base, struct xt_table_info *newinfo)
1044 {
1045         const struct xt_entry_match *ematch;
1046         const struct xt_entry_target *t;
1047         unsigned int entry_offset;
1048         int off, i, ret;
1049
1050         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1051         entry_offset = (void *)e - base;
1052         xt_ematch_foreach(ematch, e)
1053                 off += xt_compat_match_offset(ematch->u.kernel.match);
1054         t = ip6t_get_target_c(e);
1055         off += xt_compat_target_offset(t->u.kernel.target);
1056         newinfo->size -= off;
1057         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1058         if (ret)
1059                 return ret;
1060
1061         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1062                 if (info->hook_entry[i] &&
1063                     (e < (struct ip6t_entry *)(base + info->hook_entry[i])))
1064                         newinfo->hook_entry[i] -= off;
1065                 if (info->underflow[i] &&
1066                     (e < (struct ip6t_entry *)(base + info->underflow[i])))
1067                         newinfo->underflow[i] -= off;
1068         }
1069         return 0;
1070 }
1071
1072 static int compat_table_info(const struct xt_table_info *info,
1073                              struct xt_table_info *newinfo)
1074 {
1075         struct ip6t_entry *iter;
1076         const void *loc_cpu_entry;
1077         int ret;
1078
1079         if (!newinfo || !info)
1080                 return -EINVAL;
1081
1082         /* we dont care about newinfo->entries */
1083         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1084         newinfo->initial_entries = 0;
1085         loc_cpu_entry = info->entries;
1086         xt_compat_init_offsets(AF_INET6, info->number);
1087         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1088                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1089                 if (ret != 0)
1090                         return ret;
1091         }
1092         return 0;
1093 }
1094 #endif
1095
1096 static int get_info(struct net *net, void __user *user,
1097                     const int *len, int compat)
1098 {
1099         char name[XT_TABLE_MAXNAMELEN];
1100         struct xt_table *t;
1101         int ret;
1102
1103         if (*len != sizeof(struct ip6t_getinfo)) {
1104                 duprintf("length %u != %zu\n", *len,
1105                          sizeof(struct ip6t_getinfo));
1106                 return -EINVAL;
1107         }
1108
1109         if (copy_from_user(name, user, sizeof(name)) != 0)
1110                 return -EFAULT;
1111
1112         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1113 #ifdef CONFIG_COMPAT
1114         if (compat)
1115                 xt_compat_lock(AF_INET6);
1116 #endif
1117         t = try_then_request_module(xt_find_table_lock(net, AF_INET6, name),
1118                                     "ip6table_%s", name);
1119         if (!IS_ERR_OR_NULL(t)) {
1120                 struct ip6t_getinfo info;
1121                 const struct xt_table_info *private = t->private;
1122 #ifdef CONFIG_COMPAT
1123                 struct xt_table_info tmp;
1124
1125                 if (compat) {
1126                         ret = compat_table_info(private, &tmp);
1127                         xt_compat_flush_offsets(AF_INET6);
1128                         private = &tmp;
1129                 }
1130 #endif
1131                 memset(&info, 0, sizeof(info));
1132                 info.valid_hooks = t->valid_hooks;
1133                 memcpy(info.hook_entry, private->hook_entry,
1134                        sizeof(info.hook_entry));
1135                 memcpy(info.underflow, private->underflow,
1136                        sizeof(info.underflow));
1137                 info.num_entries = private->number;
1138                 info.size = private->size;
1139                 strcpy(info.name, name);
1140
1141                 if (copy_to_user(user, &info, *len) != 0)
1142                         ret = -EFAULT;
1143                 else
1144                         ret = 0;
1145
1146                 xt_table_unlock(t);
1147                 module_put(t->me);
1148         } else
1149                 ret = t ? PTR_ERR(t) : -ENOENT;
1150 #ifdef CONFIG_COMPAT
1151         if (compat)
1152                 xt_compat_unlock(AF_INET6);
1153 #endif
1154         return ret;
1155 }
1156
1157 static int
1158 get_entries(struct net *net, struct ip6t_get_entries __user *uptr,
1159             const int *len)
1160 {
1161         int ret;
1162         struct ip6t_get_entries get;
1163         struct xt_table *t;
1164
1165         if (*len < sizeof(get)) {
1166                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1167                 return -EINVAL;
1168         }
1169         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1170                 return -EFAULT;
1171         if (*len != sizeof(struct ip6t_get_entries) + get.size) {
1172                 duprintf("get_entries: %u != %zu\n",
1173                          *len, sizeof(get) + get.size);
1174                 return -EINVAL;
1175         }
1176
1177         t = xt_find_table_lock(net, AF_INET6, get.name);
1178         if (!IS_ERR_OR_NULL(t)) {
1179                 struct xt_table_info *private = t->private;
1180                 duprintf("t->private->number = %u\n", private->number);
1181                 if (get.size == private->size)
1182                         ret = copy_entries_to_user(private->size,
1183                                                    t, uptr->entrytable);
1184                 else {
1185                         duprintf("get_entries: I've got %u not %u!\n",
1186                                  private->size, get.size);
1187                         ret = -EAGAIN;
1188                 }
1189                 module_put(t->me);
1190                 xt_table_unlock(t);
1191         } else
1192                 ret = t ? PTR_ERR(t) : -ENOENT;
1193
1194         return ret;
1195 }
1196
1197 static int
1198 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1199              struct xt_table_info *newinfo, unsigned int num_counters,
1200              void __user *counters_ptr)
1201 {
1202         int ret;
1203         struct xt_table *t;
1204         struct xt_table_info *oldinfo;
1205         struct xt_counters *counters;
1206         struct ip6t_entry *iter;
1207
1208         ret = 0;
1209         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1210         if (!counters) {
1211                 ret = -ENOMEM;
1212                 goto out;
1213         }
1214
1215         t = try_then_request_module(xt_find_table_lock(net, AF_INET6, name),
1216                                     "ip6table_%s", name);
1217         if (IS_ERR_OR_NULL(t)) {
1218                 ret = t ? PTR_ERR(t) : -ENOENT;
1219                 goto free_newinfo_counters_untrans;
1220         }
1221
1222         /* You lied! */
1223         if (valid_hooks != t->valid_hooks) {
1224                 duprintf("Valid hook crap: %08X vs %08X\n",
1225                          valid_hooks, t->valid_hooks);
1226                 ret = -EINVAL;
1227                 goto put_module;
1228         }
1229
1230         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1231         if (!oldinfo)
1232                 goto put_module;
1233
1234         /* Update module usage count based on number of rules */
1235         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1236                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1237         if ((oldinfo->number > oldinfo->initial_entries) ||
1238             (newinfo->number <= oldinfo->initial_entries))
1239                 module_put(t->me);
1240         if ((oldinfo->number > oldinfo->initial_entries) &&
1241             (newinfo->number <= oldinfo->initial_entries))
1242                 module_put(t->me);
1243
1244         /* Get the old counters, and synchronize with replace */
1245         get_counters(oldinfo, counters);
1246
1247         /* Decrease module usage counts and free resource */
1248         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1249                 cleanup_entry(iter, net);
1250
1251         xt_free_table_info(oldinfo);
1252         if (copy_to_user(counters_ptr, counters,
1253                          sizeof(struct xt_counters) * num_counters) != 0) {
1254                 /* Silent error, can't fail, new table is already in place */
1255                 net_warn_ratelimited("ip6tables: counters copy to user failed while replacing table\n");
1256         }
1257         vfree(counters);
1258         xt_table_unlock(t);
1259         return ret;
1260
1261  put_module:
1262         module_put(t->me);
1263         xt_table_unlock(t);
1264  free_newinfo_counters_untrans:
1265         vfree(counters);
1266  out:
1267         return ret;
1268 }
1269
1270 static int
1271 do_replace(struct net *net, const void __user *user, unsigned int len)
1272 {
1273         int ret;
1274         struct ip6t_replace tmp;
1275         struct xt_table_info *newinfo;
1276         void *loc_cpu_entry;
1277         struct ip6t_entry *iter;
1278
1279         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1280                 return -EFAULT;
1281
1282         /* overflow check */
1283         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1284                 return -ENOMEM;
1285         if (tmp.num_counters == 0)
1286                 return -EINVAL;
1287
1288         tmp.name[sizeof(tmp.name)-1] = 0;
1289
1290         newinfo = xt_alloc_table_info(tmp.size);
1291         if (!newinfo)
1292                 return -ENOMEM;
1293
1294         loc_cpu_entry = newinfo->entries;
1295         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1296                            tmp.size) != 0) {
1297                 ret = -EFAULT;
1298                 goto free_newinfo;
1299         }
1300
1301         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1302         if (ret != 0)
1303                 goto free_newinfo;
1304
1305         duprintf("ip_tables: Translated table\n");
1306
1307         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1308                            tmp.num_counters, tmp.counters);
1309         if (ret)
1310                 goto free_newinfo_untrans;
1311         return 0;
1312
1313  free_newinfo_untrans:
1314         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1315                 cleanup_entry(iter, net);
1316  free_newinfo:
1317         xt_free_table_info(newinfo);
1318         return ret;
1319 }
1320
1321 static int
1322 do_add_counters(struct net *net, const void __user *user, unsigned int len,
1323                 int compat)
1324 {
1325         unsigned int i;
1326         struct xt_counters_info tmp;
1327         struct xt_counters *paddc;
1328         unsigned int num_counters;
1329         char *name;
1330         int size;
1331         void *ptmp;
1332         struct xt_table *t;
1333         const struct xt_table_info *private;
1334         int ret = 0;
1335         struct ip6t_entry *iter;
1336         unsigned int addend;
1337 #ifdef CONFIG_COMPAT
1338         struct compat_xt_counters_info compat_tmp;
1339
1340         if (compat) {
1341                 ptmp = &compat_tmp;
1342                 size = sizeof(struct compat_xt_counters_info);
1343         } else
1344 #endif
1345         {
1346                 ptmp = &tmp;
1347                 size = sizeof(struct xt_counters_info);
1348         }
1349
1350         if (copy_from_user(ptmp, user, size) != 0)
1351                 return -EFAULT;
1352
1353 #ifdef CONFIG_COMPAT
1354         if (compat) {
1355                 num_counters = compat_tmp.num_counters;
1356                 name = compat_tmp.name;
1357         } else
1358 #endif
1359         {
1360                 num_counters = tmp.num_counters;
1361                 name = tmp.name;
1362         }
1363
1364         if (len != size + num_counters * sizeof(struct xt_counters))
1365                 return -EINVAL;
1366
1367         paddc = vmalloc(len - size);
1368         if (!paddc)
1369                 return -ENOMEM;
1370
1371         if (copy_from_user(paddc, user + size, len - size) != 0) {
1372                 ret = -EFAULT;
1373                 goto free;
1374         }
1375
1376         t = xt_find_table_lock(net, AF_INET6, name);
1377         if (IS_ERR_OR_NULL(t)) {
1378                 ret = t ? PTR_ERR(t) : -ENOENT;
1379                 goto free;
1380         }
1381
1382         local_bh_disable();
1383         private = t->private;
1384         if (private->number != num_counters) {
1385                 ret = -EINVAL;
1386                 goto unlock_up_free;
1387         }
1388
1389         i = 0;
1390         addend = xt_write_recseq_begin();
1391         xt_entry_foreach(iter, private->entries, private->size) {
1392                 struct xt_counters *tmp;
1393
1394                 tmp = xt_get_this_cpu_counter(&iter->counters);
1395                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1396                 ++i;
1397         }
1398         xt_write_recseq_end(addend);
1399  unlock_up_free:
1400         local_bh_enable();
1401         xt_table_unlock(t);
1402         module_put(t->me);
1403  free:
1404         vfree(paddc);
1405
1406         return ret;
1407 }
1408
1409 #ifdef CONFIG_COMPAT
1410 struct compat_ip6t_replace {
1411         char                    name[XT_TABLE_MAXNAMELEN];
1412         u32                     valid_hooks;
1413         u32                     num_entries;
1414         u32                     size;
1415         u32                     hook_entry[NF_INET_NUMHOOKS];
1416         u32                     underflow[NF_INET_NUMHOOKS];
1417         u32                     num_counters;
1418         compat_uptr_t           counters;       /* struct xt_counters * */
1419         struct compat_ip6t_entry entries[0];
1420 };
1421
1422 static int
1423 compat_copy_entry_to_user(struct ip6t_entry *e, void __user **dstptr,
1424                           unsigned int *size, struct xt_counters *counters,
1425                           unsigned int i)
1426 {
1427         struct xt_entry_target *t;
1428         struct compat_ip6t_entry __user *ce;
1429         u_int16_t target_offset, next_offset;
1430         compat_uint_t origsize;
1431         const struct xt_entry_match *ematch;
1432         int ret = 0;
1433
1434         origsize = *size;
1435         ce = (struct compat_ip6t_entry __user *)*dstptr;
1436         if (copy_to_user(ce, e, sizeof(struct ip6t_entry)) != 0 ||
1437             copy_to_user(&ce->counters, &counters[i],
1438             sizeof(counters[i])) != 0)
1439                 return -EFAULT;
1440
1441         *dstptr += sizeof(struct compat_ip6t_entry);
1442         *size -= sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1443
1444         xt_ematch_foreach(ematch, e) {
1445                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1446                 if (ret != 0)
1447                         return ret;
1448         }
1449         target_offset = e->target_offset - (origsize - *size);
1450         t = ip6t_get_target(e);
1451         ret = xt_compat_target_to_user(t, dstptr, size);
1452         if (ret)
1453                 return ret;
1454         next_offset = e->next_offset - (origsize - *size);
1455         if (put_user(target_offset, &ce->target_offset) != 0 ||
1456             put_user(next_offset, &ce->next_offset) != 0)
1457                 return -EFAULT;
1458         return 0;
1459 }
1460
1461 static int
1462 compat_find_calc_match(struct xt_entry_match *m,
1463                        const char *name,
1464                        const struct ip6t_ip6 *ipv6,
1465                        int *size)
1466 {
1467         struct xt_match *match;
1468
1469         match = xt_request_find_match(NFPROTO_IPV6, m->u.user.name,
1470                                       m->u.user.revision);
1471         if (IS_ERR(match)) {
1472                 duprintf("compat_check_calc_match: `%s' not found\n",
1473                          m->u.user.name);
1474                 return PTR_ERR(match);
1475         }
1476         m->u.kernel.match = match;
1477         *size += xt_compat_match_offset(match);
1478         return 0;
1479 }
1480
1481 static void compat_release_entry(struct compat_ip6t_entry *e)
1482 {
1483         struct xt_entry_target *t;
1484         struct xt_entry_match *ematch;
1485
1486         /* Cleanup all matches */
1487         xt_ematch_foreach(ematch, e)
1488                 module_put(ematch->u.kernel.match->me);
1489         t = compat_ip6t_get_target(e);
1490         module_put(t->u.kernel.target->me);
1491 }
1492
1493 static int
1494 check_compat_entry_size_and_hooks(struct compat_ip6t_entry *e,
1495                                   struct xt_table_info *newinfo,
1496                                   unsigned int *size,
1497                                   const unsigned char *base,
1498                                   const unsigned char *limit,
1499                                   const unsigned int *hook_entries,
1500                                   const unsigned int *underflows,
1501                                   const char *name)
1502 {
1503         struct xt_entry_match *ematch;
1504         struct xt_entry_target *t;
1505         struct xt_target *target;
1506         unsigned int entry_offset;
1507         unsigned int j;
1508         int ret, off, h;
1509
1510         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1511         if ((unsigned long)e % __alignof__(struct compat_ip6t_entry) != 0 ||
1512             (unsigned char *)e + sizeof(struct compat_ip6t_entry) >= limit ||
1513             (unsigned char *)e + e->next_offset > limit) {
1514                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1515                 return -EINVAL;
1516         }
1517
1518         if (e->next_offset < sizeof(struct compat_ip6t_entry) +
1519                              sizeof(struct compat_xt_entry_target)) {
1520                 duprintf("checking: element %p size %u\n",
1521                          e, e->next_offset);
1522                 return -EINVAL;
1523         }
1524
1525         if (!ip6_checkentry(&e->ipv6))
1526                 return -EINVAL;
1527
1528         ret = xt_compat_check_entry_offsets(e, e->elems,
1529                                             e->target_offset, e->next_offset);
1530         if (ret)
1531                 return ret;
1532
1533         off = sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1534         entry_offset = (void *)e - (void *)base;
1535         j = 0;
1536         xt_ematch_foreach(ematch, e) {
1537                 ret = compat_find_calc_match(ematch, name, &e->ipv6, &off);
1538                 if (ret != 0)
1539                         goto release_matches;
1540                 ++j;
1541         }
1542
1543         t = compat_ip6t_get_target(e);
1544         target = xt_request_find_target(NFPROTO_IPV6, t->u.user.name,
1545                                         t->u.user.revision);
1546         if (IS_ERR(target)) {
1547                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1548                          t->u.user.name);
1549                 ret = PTR_ERR(target);
1550                 goto release_matches;
1551         }
1552         t->u.kernel.target = target;
1553
1554         off += xt_compat_target_offset(target);
1555         *size += off;
1556         ret = xt_compat_add_offset(AF_INET6, entry_offset, off);
1557         if (ret)
1558                 goto out;
1559
1560         /* Check hooks & underflows */
1561         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1562                 if ((unsigned char *)e - base == hook_entries[h])
1563                         newinfo->hook_entry[h] = hook_entries[h];
1564                 if ((unsigned char *)e - base == underflows[h])
1565                         newinfo->underflow[h] = underflows[h];
1566         }
1567
1568         /* Clear counters and comefrom */
1569         memset(&e->counters, 0, sizeof(e->counters));
1570         e->comefrom = 0;
1571         return 0;
1572
1573 out:
1574         module_put(t->u.kernel.target->me);
1575 release_matches:
1576         xt_ematch_foreach(ematch, e) {
1577                 if (j-- == 0)
1578                         break;
1579                 module_put(ematch->u.kernel.match->me);
1580         }
1581         return ret;
1582 }
1583
1584 static int
1585 compat_copy_entry_from_user(struct compat_ip6t_entry *e, void **dstptr,
1586                             unsigned int *size, const char *name,
1587                             struct xt_table_info *newinfo, unsigned char *base)
1588 {
1589         struct xt_entry_target *t;
1590         struct ip6t_entry *de;
1591         unsigned int origsize;
1592         int ret, h;
1593         struct xt_entry_match *ematch;
1594
1595         ret = 0;
1596         origsize = *size;
1597         de = (struct ip6t_entry *)*dstptr;
1598         memcpy(de, e, sizeof(struct ip6t_entry));
1599         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1600
1601         *dstptr += sizeof(struct ip6t_entry);
1602         *size += sizeof(struct ip6t_entry) - sizeof(struct compat_ip6t_entry);
1603
1604         xt_ematch_foreach(ematch, e) {
1605                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1606                 if (ret != 0)
1607                         return ret;
1608         }
1609         de->target_offset = e->target_offset - (origsize - *size);
1610         t = compat_ip6t_get_target(e);
1611         xt_compat_target_from_user(t, dstptr, size);
1612
1613         de->next_offset = e->next_offset - (origsize - *size);
1614         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1615                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1616                         newinfo->hook_entry[h] -= origsize - *size;
1617                 if ((unsigned char *)de - base < newinfo->underflow[h])
1618                         newinfo->underflow[h] -= origsize - *size;
1619         }
1620         return ret;
1621 }
1622
1623 static int compat_check_entry(struct ip6t_entry *e, struct net *net,
1624                               const char *name)
1625 {
1626         unsigned int j;
1627         int ret = 0;
1628         struct xt_mtchk_param mtpar;
1629         struct xt_entry_match *ematch;
1630
1631         e->counters.pcnt = xt_percpu_counter_alloc();
1632         if (IS_ERR_VALUE(e->counters.pcnt))
1633                 return -ENOMEM;
1634         j = 0;
1635         mtpar.net       = net;
1636         mtpar.table     = name;
1637         mtpar.entryinfo = &e->ipv6;
1638         mtpar.hook_mask = e->comefrom;
1639         mtpar.family    = NFPROTO_IPV6;
1640         xt_ematch_foreach(ematch, e) {
1641                 ret = check_match(ematch, &mtpar);
1642                 if (ret != 0)
1643                         goto cleanup_matches;
1644                 ++j;
1645         }
1646
1647         ret = check_target(e, net, name);
1648         if (ret)
1649                 goto cleanup_matches;
1650         return 0;
1651
1652  cleanup_matches:
1653         xt_ematch_foreach(ematch, e) {
1654                 if (j-- == 0)
1655                         break;
1656                 cleanup_match(ematch, net);
1657         }
1658
1659         xt_percpu_counter_free(e->counters.pcnt);
1660
1661         return ret;
1662 }
1663
1664 static int
1665 translate_compat_table(struct net *net,
1666                        const char *name,
1667                        unsigned int valid_hooks,
1668                        struct xt_table_info **pinfo,
1669                        void **pentry0,
1670                        unsigned int total_size,
1671                        unsigned int number,
1672                        unsigned int *hook_entries,
1673                        unsigned int *underflows)
1674 {
1675         unsigned int i, j;
1676         struct xt_table_info *newinfo, *info;
1677         void *pos, *entry0, *entry1;
1678         struct compat_ip6t_entry *iter0;
1679         struct ip6t_entry *iter1;
1680         unsigned int size;
1681         int ret = 0;
1682
1683         info = *pinfo;
1684         entry0 = *pentry0;
1685         size = total_size;
1686         info->number = number;
1687
1688         /* Init all hooks to impossible value. */
1689         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1690                 info->hook_entry[i] = 0xFFFFFFFF;
1691                 info->underflow[i] = 0xFFFFFFFF;
1692         }
1693
1694         duprintf("translate_compat_table: size %u\n", info->size);
1695         j = 0;
1696         xt_compat_lock(AF_INET6);
1697         xt_compat_init_offsets(AF_INET6, number);
1698         /* Walk through entries, checking offsets. */
1699         xt_entry_foreach(iter0, entry0, total_size) {
1700                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1701                                                         entry0,
1702                                                         entry0 + total_size,
1703                                                         hook_entries,
1704                                                         underflows,
1705                                                         name);
1706                 if (ret != 0)
1707                         goto out_unlock;
1708                 ++j;
1709         }
1710
1711         ret = -EINVAL;
1712         if (j != number) {
1713                 duprintf("translate_compat_table: %u not %u entries\n",
1714                          j, number);
1715                 goto out_unlock;
1716         }
1717
1718         /* Check hooks all assigned */
1719         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1720                 /* Only hooks which are valid */
1721                 if (!(valid_hooks & (1 << i)))
1722                         continue;
1723                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1724                         duprintf("Invalid hook entry %u %u\n",
1725                                  i, hook_entries[i]);
1726                         goto out_unlock;
1727                 }
1728                 if (info->underflow[i] == 0xFFFFFFFF) {
1729                         duprintf("Invalid underflow %u %u\n",
1730                                  i, underflows[i]);
1731                         goto out_unlock;
1732                 }
1733         }
1734
1735         ret = -ENOMEM;
1736         newinfo = xt_alloc_table_info(size);
1737         if (!newinfo)
1738                 goto out_unlock;
1739
1740         newinfo->number = number;
1741         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1742                 newinfo->hook_entry[i] = info->hook_entry[i];
1743                 newinfo->underflow[i] = info->underflow[i];
1744         }
1745         entry1 = newinfo->entries;
1746         pos = entry1;
1747         size = total_size;
1748         xt_entry_foreach(iter0, entry0, total_size) {
1749                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1750                                                   name, newinfo, entry1);
1751                 if (ret != 0)
1752                         break;
1753         }
1754         xt_compat_flush_offsets(AF_INET6);
1755         xt_compat_unlock(AF_INET6);
1756         if (ret)
1757                 goto free_newinfo;
1758
1759         ret = -ELOOP;
1760         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1761                 goto free_newinfo;
1762
1763         i = 0;
1764         xt_entry_foreach(iter1, entry1, newinfo->size) {
1765                 ret = compat_check_entry(iter1, net, name);
1766                 if (ret != 0)
1767                         break;
1768                 ++i;
1769                 if (strcmp(ip6t_get_target(iter1)->u.user.name,
1770                     XT_ERROR_TARGET) == 0)
1771                         ++newinfo->stacksize;
1772         }
1773         if (ret) {
1774                 /*
1775                  * The first i matches need cleanup_entry (calls ->destroy)
1776                  * because they had called ->check already. The other j-i
1777                  * entries need only release.
1778                  */
1779                 int skip = i;
1780                 j -= i;
1781                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1782                         if (skip-- > 0)
1783                                 continue;
1784                         if (j-- == 0)
1785                                 break;
1786                         compat_release_entry(iter0);
1787                 }
1788                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1789                         if (i-- == 0)
1790                                 break;
1791                         cleanup_entry(iter1, net);
1792                 }
1793                 xt_free_table_info(newinfo);
1794                 return ret;
1795         }
1796
1797         *pinfo = newinfo;
1798         *pentry0 = entry1;
1799         xt_free_table_info(info);
1800         return 0;
1801
1802 free_newinfo:
1803         xt_free_table_info(newinfo);
1804 out:
1805         xt_entry_foreach(iter0, entry0, total_size) {
1806                 if (j-- == 0)
1807                         break;
1808                 compat_release_entry(iter0);
1809         }
1810         return ret;
1811 out_unlock:
1812         xt_compat_flush_offsets(AF_INET6);
1813         xt_compat_unlock(AF_INET6);
1814         goto out;
1815 }
1816
1817 static int
1818 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1819 {
1820         int ret;
1821         struct compat_ip6t_replace tmp;
1822         struct xt_table_info *newinfo;
1823         void *loc_cpu_entry;
1824         struct ip6t_entry *iter;
1825
1826         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1827                 return -EFAULT;
1828
1829         /* overflow check */
1830         if (tmp.size >= INT_MAX / num_possible_cpus())
1831                 return -ENOMEM;
1832         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1833                 return -ENOMEM;
1834         if (tmp.num_counters == 0)
1835                 return -EINVAL;
1836
1837         tmp.name[sizeof(tmp.name)-1] = 0;
1838
1839         newinfo = xt_alloc_table_info(tmp.size);
1840         if (!newinfo)
1841                 return -ENOMEM;
1842
1843         loc_cpu_entry = newinfo->entries;
1844         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1845                            tmp.size) != 0) {
1846                 ret = -EFAULT;
1847                 goto free_newinfo;
1848         }
1849
1850         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1851                                      &newinfo, &loc_cpu_entry, tmp.size,
1852                                      tmp.num_entries, tmp.hook_entry,
1853                                      tmp.underflow);
1854         if (ret != 0)
1855                 goto free_newinfo;
1856
1857         duprintf("compat_do_replace: Translated table\n");
1858
1859         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1860                            tmp.num_counters, compat_ptr(tmp.counters));
1861         if (ret)
1862                 goto free_newinfo_untrans;
1863         return 0;
1864
1865  free_newinfo_untrans:
1866         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1867                 cleanup_entry(iter, net);
1868  free_newinfo:
1869         xt_free_table_info(newinfo);
1870         return ret;
1871 }
1872
1873 static int
1874 compat_do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user,
1875                        unsigned int len)
1876 {
1877         int ret;
1878
1879         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1880                 return -EPERM;
1881
1882         switch (cmd) {
1883         case IP6T_SO_SET_REPLACE:
1884                 ret = compat_do_replace(sock_net(sk), user, len);
1885                 break;
1886
1887         case IP6T_SO_SET_ADD_COUNTERS:
1888                 ret = do_add_counters(sock_net(sk), user, len, 1);
1889                 break;
1890
1891         default:
1892                 duprintf("do_ip6t_set_ctl:  unknown request %i\n", cmd);
1893                 ret = -EINVAL;
1894         }
1895
1896         return ret;
1897 }
1898
1899 struct compat_ip6t_get_entries {
1900         char name[XT_TABLE_MAXNAMELEN];
1901         compat_uint_t size;
1902         struct compat_ip6t_entry entrytable[0];
1903 };
1904
1905 static int
1906 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1907                             void __user *userptr)
1908 {
1909         struct xt_counters *counters;
1910         const struct xt_table_info *private = table->private;
1911         void __user *pos;
1912         unsigned int size;
1913         int ret = 0;
1914         unsigned int i = 0;
1915         struct ip6t_entry *iter;
1916
1917         counters = alloc_counters(table);
1918         if (IS_ERR(counters))
1919                 return PTR_ERR(counters);
1920
1921         pos = userptr;
1922         size = total_size;
1923         xt_entry_foreach(iter, private->entries, total_size) {
1924                 ret = compat_copy_entry_to_user(iter, &pos,
1925                                                 &size, counters, i++);
1926                 if (ret != 0)
1927                         break;
1928         }
1929
1930         vfree(counters);
1931         return ret;
1932 }
1933
1934 static int
1935 compat_get_entries(struct net *net, struct compat_ip6t_get_entries __user *uptr,
1936                    int *len)
1937 {
1938         int ret;
1939         struct compat_ip6t_get_entries get;
1940         struct xt_table *t;
1941
1942         if (*len < sizeof(get)) {
1943                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1944                 return -EINVAL;
1945         }
1946
1947         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1948                 return -EFAULT;
1949
1950         if (*len != sizeof(struct compat_ip6t_get_entries) + get.size) {
1951                 duprintf("compat_get_entries: %u != %zu\n",
1952                          *len, sizeof(get) + get.size);
1953                 return -EINVAL;
1954         }
1955
1956         xt_compat_lock(AF_INET6);
1957         t = xt_find_table_lock(net, AF_INET6, get.name);
1958         if (!IS_ERR_OR_NULL(t)) {
1959                 const struct xt_table_info *private = t->private;
1960                 struct xt_table_info info;
1961                 duprintf("t->private->number = %u\n", private->number);
1962                 ret = compat_table_info(private, &info);
1963                 if (!ret && get.size == info.size) {
1964                         ret = compat_copy_entries_to_user(private->size,
1965                                                           t, uptr->entrytable);
1966                 } else if (!ret) {
1967                         duprintf("compat_get_entries: I've got %u not %u!\n",
1968                                  private->size, get.size);
1969                         ret = -EAGAIN;
1970                 }
1971                 xt_compat_flush_offsets(AF_INET6);
1972                 module_put(t->me);
1973                 xt_table_unlock(t);
1974         } else
1975                 ret = t ? PTR_ERR(t) : -ENOENT;
1976
1977         xt_compat_unlock(AF_INET6);
1978         return ret;
1979 }
1980
1981 static int do_ip6t_get_ctl(struct sock *, int, void __user *, int *);
1982
1983 static int
1984 compat_do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1985 {
1986         int ret;
1987
1988         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1989                 return -EPERM;
1990
1991         switch (cmd) {
1992         case IP6T_SO_GET_INFO:
1993                 ret = get_info(sock_net(sk), user, len, 1);
1994                 break;
1995         case IP6T_SO_GET_ENTRIES:
1996                 ret = compat_get_entries(sock_net(sk), user, len);
1997                 break;
1998         default:
1999                 ret = do_ip6t_get_ctl(sk, cmd, user, len);
2000         }
2001         return ret;
2002 }
2003 #endif
2004
2005 static int
2006 do_ip6t_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2007 {
2008         int ret;
2009
2010         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2011                 return -EPERM;
2012
2013         switch (cmd) {
2014         case IP6T_SO_SET_REPLACE:
2015                 ret = do_replace(sock_net(sk), user, len);
2016                 break;
2017
2018         case IP6T_SO_SET_ADD_COUNTERS:
2019                 ret = do_add_counters(sock_net(sk), user, len, 0);
2020                 break;
2021
2022         default:
2023                 duprintf("do_ip6t_set_ctl:  unknown request %i\n", cmd);
2024                 ret = -EINVAL;
2025         }
2026
2027         return ret;
2028 }
2029
2030 static int
2031 do_ip6t_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2032 {
2033         int ret;
2034
2035         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2036                 return -EPERM;
2037
2038         switch (cmd) {
2039         case IP6T_SO_GET_INFO:
2040                 ret = get_info(sock_net(sk), user, len, 0);
2041                 break;
2042
2043         case IP6T_SO_GET_ENTRIES:
2044                 ret = get_entries(sock_net(sk), user, len);
2045                 break;
2046
2047         case IP6T_SO_GET_REVISION_MATCH:
2048         case IP6T_SO_GET_REVISION_TARGET: {
2049                 struct xt_get_revision rev;
2050                 int target;
2051
2052                 if (*len != sizeof(rev)) {
2053                         ret = -EINVAL;
2054                         break;
2055                 }
2056                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2057                         ret = -EFAULT;
2058                         break;
2059                 }
2060                 rev.name[sizeof(rev.name)-1] = 0;
2061
2062                 if (cmd == IP6T_SO_GET_REVISION_TARGET)
2063                         target = 1;
2064                 else
2065                         target = 0;
2066
2067                 try_then_request_module(xt_find_revision(AF_INET6, rev.name,
2068                                                          rev.revision,
2069                                                          target, &ret),
2070                                         "ip6t_%s", rev.name);
2071                 break;
2072         }
2073
2074         default:
2075                 duprintf("do_ip6t_get_ctl: unknown request %i\n", cmd);
2076                 ret = -EINVAL;
2077         }
2078
2079         return ret;
2080 }
2081
2082 struct xt_table *ip6t_register_table(struct net *net,
2083                                      const struct xt_table *table,
2084                                      const struct ip6t_replace *repl)
2085 {
2086         int ret;
2087         struct xt_table_info *newinfo;
2088         struct xt_table_info bootstrap = {0};
2089         void *loc_cpu_entry;
2090         struct xt_table *new_table;
2091
2092         newinfo = xt_alloc_table_info(repl->size);
2093         if (!newinfo) {
2094                 ret = -ENOMEM;
2095                 goto out;
2096         }
2097
2098         loc_cpu_entry = newinfo->entries;
2099         memcpy(loc_cpu_entry, repl->entries, repl->size);
2100
2101         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2102         if (ret != 0)
2103                 goto out_free;
2104
2105         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2106         if (IS_ERR(new_table)) {
2107                 ret = PTR_ERR(new_table);
2108                 goto out_free;
2109         }
2110         return new_table;
2111
2112 out_free:
2113         xt_free_table_info(newinfo);
2114 out:
2115         return ERR_PTR(ret);
2116 }
2117
2118 void ip6t_unregister_table(struct net *net, struct xt_table *table)
2119 {
2120         struct xt_table_info *private;
2121         void *loc_cpu_entry;
2122         struct module *table_owner = table->me;
2123         struct ip6t_entry *iter;
2124
2125         private = xt_unregister_table(table);
2126
2127         /* Decrease module usage counts and free resources */
2128         loc_cpu_entry = private->entries;
2129         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2130                 cleanup_entry(iter, net);
2131         if (private->number > private->initial_entries)
2132                 module_put(table_owner);
2133         xt_free_table_info(private);
2134 }
2135
2136 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2137 static inline bool
2138 icmp6_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2139                      u_int8_t type, u_int8_t code,
2140                      bool invert)
2141 {
2142         return (type == test_type && code >= min_code && code <= max_code)
2143                 ^ invert;
2144 }
2145
2146 static bool
2147 icmp6_match(const struct sk_buff *skb, struct xt_action_param *par)
2148 {
2149         const struct icmp6hdr *ic;
2150         struct icmp6hdr _icmph;
2151         const struct ip6t_icmp *icmpinfo = par->matchinfo;
2152
2153         /* Must not be a fragment. */
2154         if (par->fragoff != 0)
2155                 return false;
2156
2157         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2158         if (ic == NULL) {
2159                 /* We've been asked to examine this packet, and we
2160                  * can't.  Hence, no choice but to drop.
2161                  */
2162                 duprintf("Dropping evil ICMP tinygram.\n");
2163                 par->hotdrop = true;
2164                 return false;
2165         }
2166
2167         return icmp6_type_code_match(icmpinfo->type,
2168                                      icmpinfo->code[0],
2169                                      icmpinfo->code[1],
2170                                      ic->icmp6_type, ic->icmp6_code,
2171                                      !!(icmpinfo->invflags&IP6T_ICMP_INV));
2172 }
2173
2174 /* Called when user tries to insert an entry of this type. */
2175 static int icmp6_checkentry(const struct xt_mtchk_param *par)
2176 {
2177         const struct ip6t_icmp *icmpinfo = par->matchinfo;
2178
2179         /* Must specify no unknown invflags */
2180         return (icmpinfo->invflags & ~IP6T_ICMP_INV) ? -EINVAL : 0;
2181 }
2182
2183 /* The built-in targets: standard (NULL) and error. */
2184 static struct xt_target ip6t_builtin_tg[] __read_mostly = {
2185         {
2186                 .name             = XT_STANDARD_TARGET,
2187                 .targetsize       = sizeof(int),
2188                 .family           = NFPROTO_IPV6,
2189 #ifdef CONFIG_COMPAT
2190                 .compatsize       = sizeof(compat_int_t),
2191                 .compat_from_user = compat_standard_from_user,
2192                 .compat_to_user   = compat_standard_to_user,
2193 #endif
2194         },
2195         {
2196                 .name             = XT_ERROR_TARGET,
2197                 .target           = ip6t_error,
2198                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2199                 .family           = NFPROTO_IPV6,
2200         },
2201 };
2202
2203 static struct nf_sockopt_ops ip6t_sockopts = {
2204         .pf             = PF_INET6,
2205         .set_optmin     = IP6T_BASE_CTL,
2206         .set_optmax     = IP6T_SO_SET_MAX+1,
2207         .set            = do_ip6t_set_ctl,
2208 #ifdef CONFIG_COMPAT
2209         .compat_set     = compat_do_ip6t_set_ctl,
2210 #endif
2211         .get_optmin     = IP6T_BASE_CTL,
2212         .get_optmax     = IP6T_SO_GET_MAX+1,
2213         .get            = do_ip6t_get_ctl,
2214 #ifdef CONFIG_COMPAT
2215         .compat_get     = compat_do_ip6t_get_ctl,
2216 #endif
2217         .owner          = THIS_MODULE,
2218 };
2219
2220 static struct xt_match ip6t_builtin_mt[] __read_mostly = {
2221         {
2222                 .name       = "icmp6",
2223                 .match      = icmp6_match,
2224                 .matchsize  = sizeof(struct ip6t_icmp),
2225                 .checkentry = icmp6_checkentry,
2226                 .proto      = IPPROTO_ICMPV6,
2227                 .family     = NFPROTO_IPV6,
2228         },
2229 };
2230
2231 static int __net_init ip6_tables_net_init(struct net *net)
2232 {
2233         return xt_proto_init(net, NFPROTO_IPV6);
2234 }
2235
2236 static void __net_exit ip6_tables_net_exit(struct net *net)
2237 {
2238         xt_proto_fini(net, NFPROTO_IPV6);
2239 }
2240
2241 static struct pernet_operations ip6_tables_net_ops = {
2242         .init = ip6_tables_net_init,
2243         .exit = ip6_tables_net_exit,
2244 };
2245
2246 static int __init ip6_tables_init(void)
2247 {
2248         int ret;
2249
2250         ret = register_pernet_subsys(&ip6_tables_net_ops);
2251         if (ret < 0)
2252                 goto err1;
2253
2254         /* No one else will be downing sem now, so we won't sleep */
2255         ret = xt_register_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
2256         if (ret < 0)
2257                 goto err2;
2258         ret = xt_register_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
2259         if (ret < 0)
2260                 goto err4;
2261
2262         /* Register setsockopt */
2263         ret = nf_register_sockopt(&ip6t_sockopts);
2264         if (ret < 0)
2265                 goto err5;
2266
2267         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2268         return 0;
2269
2270 err5:
2271         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
2272 err4:
2273         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
2274 err2:
2275         unregister_pernet_subsys(&ip6_tables_net_ops);
2276 err1:
2277         return ret;
2278 }
2279
2280 static void __exit ip6_tables_fini(void)
2281 {
2282         nf_unregister_sockopt(&ip6t_sockopts);
2283
2284         xt_unregister_matches(ip6t_builtin_mt, ARRAY_SIZE(ip6t_builtin_mt));
2285         xt_unregister_targets(ip6t_builtin_tg, ARRAY_SIZE(ip6t_builtin_tg));
2286         unregister_pernet_subsys(&ip6_tables_net_ops);
2287 }
2288
2289 EXPORT_SYMBOL(ip6t_register_table);
2290 EXPORT_SYMBOL(ip6t_unregister_table);
2291 EXPORT_SYMBOL(ip6t_do_table);
2292
2293 module_init(ip6_tables_init);
2294 module_exit(ip6_tables_fini);