net/ipv4/netfilter/ip_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 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13 #include <linux/cache.h>
  14 #include <linux/capability.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/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/compat.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/mutex.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/err.h>
  27 #include <linux/cpumask.h>
  28
  29 #include <linux/netfilter/x_tables.h>
  30 #include <linux/netfilter_ipv4/ip_tables.h>
  31 #include <net/netfilter/nf_log.h>
  32 #include "../../netfilter/xt_repldata.h"
  33
  34 MODULE_LICENSE("GPL");
  35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36 MODULE_DESCRIPTION("IPv4 packet filter");
  37
  38 /*#define DEBUG_IP_FIREWALL*/
  39 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  40 /*#define DEBUG_IP_FIREWALL_USER*/
  41
  42 #ifdef DEBUG_IP_FIREWALL
  43 #define dprintf(format, args...) pr_info(format , ## args)
  44 #else
  45 #define dprintf(format, args...)
  46 #endif
  47
  48 #ifdef DEBUG_IP_FIREWALL_USER
  49 #define duprintf(format, args...) pr_info(format , ## args)
  50 #else
  51 #define duprintf(format, args...)
  52 #endif
  53
  54 #ifdef CONFIG_NETFILTER_DEBUG
  55 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  56 #else
  57 #define IP_NF_ASSERT(x)
  58 #endif
  59
  60 #if 0
  61 /* All the better to debug you with... */
  62 #define static
  63 #define inline
  64 #endif
  65
  66 void *ipt_alloc_initial_table(const struct xt_table *info)
  67 {
  68         return xt_alloc_initial_table(ipt, IPT);
  69 }
  70 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  71
  72 /* Returns whether matches rule or not. */
  73 /* Performance critical - called for every packet */
  74 static inline bool
  75 ip_packet_match(const struct iphdr *ip,
  76                 const char *indev,
  77                 const char *outdev,
  78                 const struct ipt_ip *ipinfo,
  79                 int isfrag)
  80 {
  81         unsigned long ret;
  82
  83 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  84
  85         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  86                   IPT_INV_SRCIP) ||
  87             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  88                   IPT_INV_DSTIP)) {
  89                 dprintf("Source or dest mismatch.\n");
  90
  91                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  92                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  93                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  94                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  95                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  96                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  97                 return false;
  98         }
  99
 100         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 101
 102         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 103                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 104                         indev, ipinfo->iniface,
 105                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
 106                 return false;
 107         }
 108
 109         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 110
 111         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 112                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 113                         outdev, ipinfo->outiface,
 114                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
 115                 return false;
 116         }
 117
 118         /* Check specific protocol */
 119         if (ipinfo->proto &&
 120             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 121                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 122                         ip->protocol, ipinfo->proto,
 123                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
 124                 return false;
 125         }
 126
 127         /* If we have a fragment rule but the packet is not a fragment
 128          * then we return zero */
 129         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 130                 dprintf("Fragment rule but not fragment.%s\n",
 131                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 132                 return false;
 133         }
 134
 135         return true;
 136 }
 137
 138 static bool
 139 ip_checkentry(const struct ipt_ip *ip)
 140 {
 141         if (ip->flags & ~IPT_F_MASK) {
 142                 duprintf("Unknown flag bits set: %08X\n",
 143                          ip->flags & ~IPT_F_MASK);
 144                 return false;
 145         }
 146         if (ip->invflags & ~IPT_INV_MASK) {
 147                 duprintf("Unknown invflag bits set: %08X\n",
 148                          ip->invflags & ~IPT_INV_MASK);
 149                 return false;
 150         }
 151         return true;
 152 }
 153
 154 static unsigned int
 155 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 156 {
 157         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 158
 159         return NF_DROP;
 160 }
 161
 162 /* Performance critical */
 163 static inline struct ipt_entry *
 164 get_entry(const void *base, unsigned int offset)
 165 {
 166         return (struct ipt_entry *)(base + offset);
 167 }
 168
 169 /* All zeroes == unconditional rule. */
 170 /* Mildly perf critical (only if packet tracing is on) */
 171 static inline bool unconditional(const struct ipt_ip *ip)
 172 {
 173         static const struct ipt_ip uncond;
 174
 175         return memcmp(ip, &uncond, sizeof(uncond)) == 0;
 176 #undef FWINV
 177 }
 178
 179 /* for const-correctness */
 180 static inline const struct xt_entry_target *
 181 ipt_get_target_c(const struct ipt_entry *e)
 182 {
 183         return ipt_get_target((struct ipt_entry *)e);
 184 }
 185
 186 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 187 static const char *const hooknames[] = {
 188         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 189         [NF_INET_LOCAL_IN]              = "INPUT",
 190         [NF_INET_FORWARD]               = "FORWARD",
 191         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 192         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 193 };
 194
 195 enum nf_ip_trace_comments {
 196         NF_IP_TRACE_COMMENT_RULE,
 197         NF_IP_TRACE_COMMENT_RETURN,
 198         NF_IP_TRACE_COMMENT_POLICY,
 199 };
 200
 201 static const char *const comments[] = {
 202         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 203         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 204         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 205 };
 206
 207 static struct nf_loginfo trace_loginfo = {
 208         .type = NF_LOG_TYPE_LOG,
 209         .u = {
 210                 .log = {
 211                         .level = 4,
 212                         .logflags = NF_LOG_MASK,
 213                 },
 214         },
 215 };
 216
 217 /* Mildly perf critical (only if packet tracing is on) */
 218 static inline int
 219 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 220                       const char *hookname, const char **chainname,
 221                       const char **comment, unsigned int *rulenum)
 222 {
 223         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 224
 225         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 226                 /* Head of user chain: ERROR target with chainname */
 227                 *chainname = t->target.data;
 228                 (*rulenum) = 0;
 229         } else if (s == e) {
 230                 (*rulenum)++;
 231
 232                 if (s->target_offset == sizeof(struct ipt_entry) &&
 233                     strcmp(t->target.u.kernel.target->name,
 234                            XT_STANDARD_TARGET) == 0 &&
 235                    t->verdict < 0 &&
 236                    unconditional(&s->ip)) {
 237                         /* Tail of chains: STANDARD target (return/policy) */
 238                         *comment = *chainname == hookname
 239                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 240                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 241                 }
 242                 return 1;
 243         } else
 244                 (*rulenum)++;
 245
 246         return 0;
 247 }
 248
 249 static void trace_packet(const struct sk_buff *skb,
 250                          unsigned int hook,
 251                          const struct net_device *in,
 252                          const struct net_device *out,
 253                          const char *tablename,
 254                          const struct xt_table_info *private,
 255                          const struct ipt_entry *e)
 256 {
 257         const struct ipt_entry *root;
 258         const char *hookname, *chainname, *comment;
 259         const struct ipt_entry *iter;
 260         unsigned int rulenum = 0;
 261         struct net *net = dev_net(in ? in : out);
 262
 263         root = get_entry(private->entries, private->hook_entry[hook]);
 264
 265         hookname = chainname = hooknames[hook];
 266         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 267
 268         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 269                 if (get_chainname_rulenum(iter, e, hookname,
 270                     &chainname, &comment, &rulenum) != 0)
 271                         break;
 272
 273         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 274                      "TRACE: %s:%s:%s:%u ",
 275                      tablename, chainname, comment, rulenum);
 276 }
 277 #endif
 278
 279 static inline __pure
 280 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 281 {
 282         return (void *)entry + entry->next_offset;
 283 }
 284
 285 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 286 unsigned int
 287 ipt_do_table(struct sk_buff *skb,
 288              unsigned int hook,
 289              const struct nf_hook_state *state,
 290              struct xt_table *table)
 291 {
 292         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 293         const struct iphdr *ip;
 294         /* Initializing verdict to NF_DROP keeps gcc happy. */
 295         unsigned int verdict = NF_DROP;
 296         const char *indev, *outdev;
 297         const void *table_base;
 298         struct ipt_entry *e, **jumpstack;
 299         unsigned int *stackptr, origptr, cpu;
 300         const struct xt_table_info *private;
 301         struct xt_action_param acpar;
 302         unsigned int addend;
 303
 304         /* Initialization */
 305         ip = ip_hdr(skb);
 306         indev = state->in ? state->in->name : nulldevname;
 307         outdev = state->out ? state->out->name : nulldevname;
 308         /* We handle fragments by dealing with the first fragment as
 309          * if it was a normal packet.  All other fragments are treated
 310          * normally, except that they will NEVER match rules that ask
 311          * things we don't know, ie. tcp syn flag or ports).  If the
 312          * rule is also a fragment-specific rule, non-fragments won't
 313          * match it. */
 314         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 315         acpar.thoff   = ip_hdrlen(skb);
 316         acpar.hotdrop = false;
 317         acpar.in      = state->in;
 318         acpar.out     = state->out;
 319         acpar.family  = NFPROTO_IPV4;
 320         acpar.hooknum = hook;
 321
 322         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 323         local_bh_disable();
 324         addend = xt_write_recseq_begin();
 325         private = table->private;
 326         cpu        = smp_processor_id();
 327         /*
 328          * Ensure we load private-> members after we've fetched the base
 329          * pointer.
 330          */
 331         smp_read_barrier_depends();
 332         table_base = private->entries;
 333         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 334         stackptr   = per_cpu_ptr(private->stackptr, cpu);
 335         origptr    = *stackptr;
 336
 337         e = get_entry(table_base, private->hook_entry[hook]);
 338
 339         pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
 340                  table->name, hook, origptr,
 341                  get_entry(table_base, private->underflow[hook]));
 342
 343         do {
 344                 const struct xt_entry_target *t;
 345                 const struct xt_entry_match *ematch;
 346                 struct xt_counters *counter;
 347
 348                 IP_NF_ASSERT(e);
 349                 if (!ip_packet_match(ip, indev, outdev,
 350                     &e->ip, acpar.fragoff)) {
 351  no_match:
 352                         e = ipt_next_entry(e);
 353                         continue;
 354                 }
 355
 356                 xt_ematch_foreach(ematch, e) {
 357                         acpar.match     = ematch->u.kernel.match;
 358                         acpar.matchinfo = ematch->data;
 359                         if (!acpar.match->match(skb, &acpar))
 360                                 goto no_match;
 361                 }
 362
 363                 counter = xt_get_this_cpu_counter(&e->counters);
 364                 ADD_COUNTER(*counter, skb->len, 1);
 365
 366                 t = ipt_get_target(e);
 367                 IP_NF_ASSERT(t->u.kernel.target);
 368
 369 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 370                 /* The packet is traced: log it */
 371                 if (unlikely(skb->nf_trace))
 372                         trace_packet(skb, hook, state->in, state->out,
 373                                      table->name, private, e);
 374 #endif
 375                 /* Standard target? */
 376                 if (!t->u.kernel.target->target) {
 377                         int v;
 378
 379                         v = ((struct xt_standard_target *)t)->verdict;
 380                         if (v < 0) {
 381                                 /* Pop from stack? */
 382                                 if (v != XT_RETURN) {
 383                                         verdict = (unsigned int)(-v) - 1;
 384                                         break;
 385                                 }
 386                                 if (*stackptr <= origptr) {
 387                                         e = get_entry(table_base,
 388                                             private->underflow[hook]);
 389                                         pr_debug("Underflow (this is normal) "
 390                                                  "to %p\n", e);
 391                                 } else {
 392                                         e = jumpstack[--*stackptr];
 393                                         pr_debug("Pulled %p out from pos %u\n",
 394                                                  e, *stackptr);
 395                                         e = ipt_next_entry(e);
 396                                 }
 397                                 continue;
 398                         }
 399                         if (table_base + v != ipt_next_entry(e) &&
 400                             !(e->ip.flags & IPT_F_GOTO)) {
 401                                 if (*stackptr >= private->stacksize) {
 402                                         verdict = NF_DROP;
 403                                         break;
 404                                 }
 405                                 jumpstack[(*stackptr)++] = e;
 406                                 pr_debug("Pushed %p into pos %u\n",
 407                                          e, *stackptr - 1);
 408                         }
 409
 410                         e = get_entry(table_base, v);
 411                         continue;
 412                 }
 413
 414                 acpar.target   = t->u.kernel.target;
 415                 acpar.targinfo = t->data;
 416
 417                 verdict = t->u.kernel.target->target(skb, &acpar);
 418                 /* Target might have changed stuff. */
 419                 ip = ip_hdr(skb);
 420                 if (verdict == XT_CONTINUE)
 421                         e = ipt_next_entry(e);
 422                 else
 423                         /* Verdict */
 424                         break;
 425         } while (!acpar.hotdrop);
 426         pr_debug("Exiting %s; resetting sp from %u to %u\n",
 427                  __func__, *stackptr, origptr);
 428         *stackptr = origptr;
 429         xt_write_recseq_end(addend);
 430         local_bh_enable();
 431
 432 #ifdef DEBUG_ALLOW_ALL
 433         return NF_ACCEPT;
 434 #else
 435         if (acpar.hotdrop)
 436                 return NF_DROP;
 437         else return verdict;
 438 #endif
 439 }
 440
 441 /* Figures out from what hook each rule can be called: returns 0 if
 442  * there are loops.  Puts hook bitmask in comefrom.
 443  *
 444  * Keeps track of largest call depth seen and stores it in newinfo->stacksize.
 445  */
 446 static int
 447 mark_source_chains(struct xt_table_info *newinfo,
 448                    unsigned int valid_hooks, void *entry0)
 449 {
 450         unsigned int calldepth, max_calldepth = 0;
 451         unsigned int hook;
 452
 453         /* No recursion; use packet counter to save back ptrs (reset
 454            to 0 as we leave), and comefrom to save source hook bitmask */
 455         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 456                 unsigned int pos = newinfo->hook_entry[hook];
 457                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 458
 459                 if (!(valid_hooks & (1 << hook)))
 460                         continue;
 461
 462                 /* Set initial back pointer. */
 463                 e->counters.pcnt = pos;
 464                 calldepth = 0;
 465
 466                 for (;;) {
 467                         const struct xt_standard_target *t
 468                                 = (void *)ipt_get_target_c(e);
 469                         int visited = e->comefrom & (1 << hook);
 470
 471                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 472                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 473                                        hook, pos, e->comefrom);
 474                                 return 0;
 475                         }
 476                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 477
 478                         /* Unconditional return/END. */
 479                         if ((e->target_offset == sizeof(struct ipt_entry) &&
 480                              (strcmp(t->target.u.user.name,
 481                                      XT_STANDARD_TARGET) == 0) &&
 482                              t->verdict < 0 && unconditional(&e->ip)) ||
 483                             visited) {
 484                                 unsigned int oldpos, size;
 485
 486                                 if ((strcmp(t->target.u.user.name,
 487                                             XT_STANDARD_TARGET) == 0) &&
 488                                     t->verdict < -NF_MAX_VERDICT - 1) {
 489                                         duprintf("mark_source_chains: bad "
 490                                                 "negative verdict (%i)\n",
 491                                                                 t->verdict);
 492                                         return 0;
 493                                 }
 494
 495                                 /* Return: backtrack through the last
 496                                    big jump. */
 497                                 do {
 498                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 499 #ifdef DEBUG_IP_FIREWALL_USER
 500                                         if (e->comefrom
 501                                             & (1 << NF_INET_NUMHOOKS)) {
 502                                                 duprintf("Back unset "
 503                                                          "on hook %u "
 504                                                          "rule %u\n",
 505                                                          hook, pos);
 506                                         }
 507 #endif
 508                                         oldpos = pos;
 509                                         pos = e->counters.pcnt;
 510                                         e->counters.pcnt = 0;
 511
 512                                         /* We're at the start. */
 513                                         if (pos == oldpos)
 514                                                 goto next;
 515
 516                                         e = (struct ipt_entry *)
 517                                                 (entry0 + pos);
 518                                 } while (oldpos == pos + e->next_offset);
 519
 520                                 /* Move along one */
 521                                 size = e->next_offset;
 522                                 e = (struct ipt_entry *)
 523                                         (entry0 + pos + size);
 524                                 e->counters.pcnt = pos;
 525                                 pos += size;
 526                                 WARN_ON_ONCE(calldepth == 0);
 527                                 if (calldepth > 0)
 528                                         --calldepth;
 529                         } else {
 530                                 int newpos = t->verdict;
 531
 532                                 if (strcmp(t->target.u.user.name,
 533                                            XT_STANDARD_TARGET) == 0 &&
 534                                     newpos >= 0) {
 535                                         if (newpos > newinfo->size -
 536                                                 sizeof(struct ipt_entry)) {
 537                                                 duprintf("mark_source_chains: "
 538                                                         "bad verdict (%i)\n",
 539                                                                 newpos);
 540                                                 return 0;
 541                                         }
 542                                         if (entry0 + newpos != ipt_next_entry(e) &&
 543                                             !(e->ip.flags & IPT_F_GOTO) &&
 544                                             ++calldepth > max_calldepth)
 545                                                 max_calldepth = calldepth;
 546
 547                                         /* This a jump; chase it. */
 548                                         duprintf("Jump rule %u -> %u, calldepth %d\n",
 549                                                  pos, newpos, calldepth);
 550                                 } else {
 551                                         /* ... this is a fallthru */
 552                                         newpos = pos + e->next_offset;
 553                                 }
 554                                 e = (struct ipt_entry *)
 555                                         (entry0 + newpos);
 556                                 e->counters.pcnt = pos;
 557                                 pos = newpos;
 558                         }
 559                 }
 560                 next:
 561                 duprintf("Finished chain %u\n", hook);
 562         }
 563         newinfo->stacksize = max_calldepth;
 564         return 1;
 565 }
 566
 567 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 568 {
 569         struct xt_mtdtor_param par;
 570
 571         par.net       = net;
 572         par.match     = m->u.kernel.match;
 573         par.matchinfo = m->data;
 574         par.family    = NFPROTO_IPV4;
 575         if (par.match->destroy != NULL)
 576                 par.match->destroy(&par);
 577         module_put(par.match->me);
 578 }
 579
 580 static int
 581 check_entry(const struct ipt_entry *e, const char *name)
 582 {
 583         const struct xt_entry_target *t;
 584
 585         if (!ip_checkentry(&e->ip)) {
 586                 duprintf("ip check failed %p %s.\n", e, name);
 587                 return -EINVAL;
 588         }
 589
 590         if (e->target_offset + sizeof(struct xt_entry_target) >
 591             e->next_offset)
 592                 return -EINVAL;
 593
 594         t = ipt_get_target_c(e);
 595         if (e->target_offset + t->u.target_size > e->next_offset)
 596                 return -EINVAL;
 597
 598         return 0;
 599 }
 600
 601 static int
 602 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 603 {
 604         const struct ipt_ip *ip = 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               ip->proto, ip->invflags & IPT_INV_PROTO);
 612         if (ret < 0) {
 613                 duprintf("check failed for `%s'.\n", par->match->name);
 614                 return ret;
 615         }
 616         return 0;
 617 }
 618
 619 static int
 620 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 621 {
 622         struct xt_match *match;
 623         int ret;
 624
 625         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 626                                       m->u.user.revision);
 627         if (IS_ERR(match)) {
 628                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 629                 return PTR_ERR(match);
 630         }
 631         m->u.kernel.match = match;
 632
 633         ret = check_match(m, par);
 634         if (ret)
 635                 goto err;
 636
 637         return 0;
 638 err:
 639         module_put(m->u.kernel.match->me);
 640         return ret;
 641 }
 642
 643 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 644 {
 645         struct xt_entry_target *t = ipt_get_target(e);
 646         struct xt_tgchk_param par = {
 647                 .net       = net,
 648                 .table     = name,
 649                 .entryinfo = e,
 650                 .target    = t->u.kernel.target,
 651                 .targinfo  = t->data,
 652                 .hook_mask = e->comefrom,
 653                 .family    = NFPROTO_IPV4,
 654         };
 655         int ret;
 656
 657         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 658               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 659         if (ret < 0) {
 660                 duprintf("check failed for `%s'.\n",
 661                          t->u.kernel.target->name);
 662                 return ret;
 663         }
 664         return 0;
 665 }
 666
 667 static int
 668 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 669                  unsigned int size)
 670 {
 671         struct xt_entry_target *t;
 672         struct xt_target *target;
 673         int ret;
 674         unsigned int j;
 675         struct xt_mtchk_param mtpar;
 676         struct xt_entry_match *ematch;
 677
 678         ret = check_entry(e, name);
 679         if (ret)
 680                 return ret;
 681
 682         e->counters.pcnt = xt_percpu_counter_alloc();
 683         if (IS_ERR_VALUE(e->counters.pcnt))
 684                 return -ENOMEM;
 685
 686         j = 0;
 687         mtpar.net       = net;
 688         mtpar.table     = name;
 689         mtpar.entryinfo = &e->ip;
 690         mtpar.hook_mask = e->comefrom;
 691         mtpar.family    = NFPROTO_IPV4;
 692         xt_ematch_foreach(ematch, e) {
 693                 ret = find_check_match(ematch, &mtpar);
 694                 if (ret != 0)
 695                         goto cleanup_matches;
 696                 ++j;
 697         }
 698
 699         t = ipt_get_target(e);
 700         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 701                                         t->u.user.revision);
 702         if (IS_ERR(target)) {
 703                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 704                 ret = PTR_ERR(target);
 705                 goto cleanup_matches;
 706         }
 707         t->u.kernel.target = target;
 708
 709         ret = check_target(e, net, name);
 710         if (ret)
 711                 goto err;
 712
 713         return 0;
 714  err:
 715         module_put(t->u.kernel.target->me);
 716  cleanup_matches:
 717         xt_ematch_foreach(ematch, e) {
 718                 if (j-- == 0)
 719                         break;
 720                 cleanup_match(ematch, net);
 721         }
 722
 723         xt_percpu_counter_free(e->counters.pcnt);
 724
 725         return ret;
 726 }
 727
 728 static bool check_underflow(const struct ipt_entry *e)
 729 {
 730         const struct xt_entry_target *t;
 731         unsigned int verdict;
 732
 733         if (!unconditional(&e->ip))
 734                 return false;
 735         t = ipt_get_target_c(e);
 736         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 737                 return false;
 738         verdict = ((struct xt_standard_target *)t)->verdict;
 739         verdict = -verdict - 1;
 740         return verdict == NF_DROP || verdict == NF_ACCEPT;
 741 }
 742
 743 static int
 744 check_entry_size_and_hooks(struct ipt_entry *e,
 745                            struct xt_table_info *newinfo,
 746                            const unsigned char *base,
 747                            const unsigned char *limit,
 748                            const unsigned int *hook_entries,
 749                            const unsigned int *underflows,
 750                            unsigned int valid_hooks)
 751 {
 752         unsigned int h;
 753
 754         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 755             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 756                 duprintf("Bad offset %p\n", e);
 757                 return -EINVAL;
 758         }
 759
 760         if (e->next_offset
 761             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 762                 duprintf("checking: element %p size %u\n",
 763                          e, e->next_offset);
 764                 return -EINVAL;
 765         }
 766
 767         /* Check hooks & underflows */
 768         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 769                 if (!(valid_hooks & (1 << h)))
 770                         continue;
 771                 if ((unsigned char *)e - base == hook_entries[h])
 772                         newinfo->hook_entry[h] = hook_entries[h];
 773                 if ((unsigned char *)e - base == underflows[h]) {
 774                         if (!check_underflow(e)) {
 775                                 pr_err("Underflows must be unconditional and "
 776                                        "use the STANDARD target with "
 777                                        "ACCEPT/DROP\n");
 778                                 return -EINVAL;
 779                         }
 780                         newinfo->underflow[h] = underflows[h];
 781                 }
 782         }
 783
 784         /* Clear counters and comefrom */
 785         e->counters = ((struct xt_counters) { 0, 0 });
 786         e->comefrom = 0;
 787         return 0;
 788 }
 789
 790 static void
 791 cleanup_entry(struct ipt_entry *e, struct net *net)
 792 {
 793         struct xt_tgdtor_param par;
 794         struct xt_entry_target *t;
 795         struct xt_entry_match *ematch;
 796
 797         /* Cleanup all matches */
 798         xt_ematch_foreach(ematch, e)
 799                 cleanup_match(ematch, net);
 800         t = ipt_get_target(e);
 801
 802         par.net      = net;
 803         par.target   = t->u.kernel.target;
 804         par.targinfo = t->data;
 805         par.family   = NFPROTO_IPV4;
 806         if (par.target->destroy != NULL)
 807                 par.target->destroy(&par);
 808         module_put(par.target->me);
 809         xt_percpu_counter_free(e->counters.pcnt);
 810 }
 811
 812 /* Checks and translates the user-supplied table segment (held in
 813    newinfo) */
 814 static int
 815 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 816                 const struct ipt_replace *repl)
 817 {
 818         struct ipt_entry *iter;
 819         unsigned int i;
 820         int ret = 0;
 821
 822         newinfo->size = repl->size;
 823         newinfo->number = repl->num_entries;
 824
 825         /* Init all hooks to impossible value. */
 826         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 827                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 828                 newinfo->underflow[i] = 0xFFFFFFFF;
 829         }
 830
 831         duprintf("translate_table: size %u\n", newinfo->size);
 832         i = 0;
 833         /* Walk through entries, checking offsets. */
 834         xt_entry_foreach(iter, entry0, newinfo->size) {
 835                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 836                                                  entry0 + repl->size,
 837                                                  repl->hook_entry,
 838                                                  repl->underflow,
 839                                                  repl->valid_hooks);
 840                 if (ret != 0)
 841                         return ret;
 842                 ++i;
 843         }
 844
 845         if (i != repl->num_entries) {
 846                 duprintf("translate_table: %u not %u entries\n",
 847                          i, repl->num_entries);
 848                 return -EINVAL;
 849         }
 850
 851         /* Check hooks all assigned */
 852         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 853                 /* Only hooks which are valid */
 854                 if (!(repl->valid_hooks & (1 << i)))
 855                         continue;
 856                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 857                         duprintf("Invalid hook entry %u %u\n",
 858                                  i, repl->hook_entry[i]);
 859                         return -EINVAL;
 860                 }
 861                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 862                         duprintf("Invalid underflow %u %u\n",
 863                                  i, repl->underflow[i]);
 864                         return -EINVAL;
 865                 }
 866         }
 867
 868         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 869                 return -ELOOP;
 870
 871         /* Finally, each sanity check must pass */
 872         i = 0;
 873         xt_entry_foreach(iter, entry0, newinfo->size) {
 874                 ret = find_check_entry(iter, net, repl->name, repl->size);
 875                 if (ret != 0)
 876                         break;
 877                 ++i;
 878         }
 879
 880         if (ret != 0) {
 881                 xt_entry_foreach(iter, entry0, newinfo->size) {
 882                         if (i-- == 0)
 883                                 break;
 884                         cleanup_entry(iter, net);
 885                 }
 886                 return ret;
 887         }
 888
 889         return ret;
 890 }
 891
 892 static void
 893 get_counters(const struct xt_table_info *t,
 894              struct xt_counters counters[])
 895 {
 896         struct ipt_entry *iter;
 897         unsigned int cpu;
 898         unsigned int i;
 899
 900         for_each_possible_cpu(cpu) {
 901                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 902
 903                 i = 0;
 904                 xt_entry_foreach(iter, t->entries, t->size) {
 905                         struct xt_counters *tmp;
 906                         u64 bcnt, pcnt;
 907                         unsigned int start;
 908
 909                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 910                         do {
 911                                 start = read_seqcount_begin(s);
 912                                 bcnt = tmp->bcnt;
 913                                 pcnt = tmp->pcnt;
 914                         } while (read_seqcount_retry(s, start));
 915
 916                         ADD_COUNTER(counters[i], bcnt, pcnt);
 917                         ++i; /* macro does multi eval of i */
 918                 }
 919         }
 920 }
 921
 922 static struct xt_counters *alloc_counters(const struct xt_table *table)
 923 {
 924         unsigned int countersize;
 925         struct xt_counters *counters;
 926         const struct xt_table_info *private = table->private;
 927
 928         /* We need atomic snapshot of counters: rest doesn't change
 929            (other than comefrom, which userspace doesn't care
 930            about). */
 931         countersize = sizeof(struct xt_counters) * private->number;
 932         counters = vzalloc(countersize);
 933
 934         if (counters == NULL)
 935                 return ERR_PTR(-ENOMEM);
 936
 937         get_counters(private, counters);
 938
 939         return counters;
 940 }
 941
 942 static int
 943 copy_entries_to_user(unsigned int total_size,
 944                      const struct xt_table *table,
 945                      void __user *userptr)
 946 {
 947         unsigned int off, num;
 948         const struct ipt_entry *e;
 949         struct xt_counters *counters;
 950         const struct xt_table_info *private = table->private;
 951         int ret = 0;
 952         const void *loc_cpu_entry;
 953
 954         counters = alloc_counters(table);
 955         if (IS_ERR(counters))
 956                 return PTR_ERR(counters);
 957
 958         loc_cpu_entry = private->entries;
 959         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 960                 ret = -EFAULT;
 961                 goto free_counters;
 962         }
 963
 964         /* FIXME: use iterator macros --RR */
 965         /* ... then go back and fix counters and names */
 966         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 967                 unsigned int i;
 968                 const struct xt_entry_match *m;
 969                 const struct xt_entry_target *t;
 970
 971                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 972                 if (copy_to_user(userptr + off
 973                                  + offsetof(struct ipt_entry, counters),
 974                                  &counters[num],
 975                                  sizeof(counters[num])) != 0) {
 976                         ret = -EFAULT;
 977                         goto free_counters;
 978                 }
 979
 980                 for (i = sizeof(struct ipt_entry);
 981                      i < e->target_offset;
 982                      i += m->u.match_size) {
 983                         m = (void *)e + i;
 984
 985                         if (copy_to_user(userptr + off + i
 986                                          + offsetof(struct xt_entry_match,
 987                                                     u.user.name),
 988                                          m->u.kernel.match->name,
 989                                          strlen(m->u.kernel.match->name)+1)
 990                             != 0) {
 991                                 ret = -EFAULT;
 992                                 goto free_counters;
 993                         }
 994                 }
 995
 996                 t = ipt_get_target_c(e);
 997                 if (copy_to_user(userptr + off + e->target_offset
 998                                  + offsetof(struct xt_entry_target,
 999                                             u.user.name),
1000                                  t->u.kernel.target->name,
1001                                  strlen(t->u.kernel.target->name)+1) != 0) {
1002                         ret = -EFAULT;
1003                         goto free_counters;
1004                 }
1005         }
1006
1007  free_counters:
1008         vfree(counters);
1009         return ret;
1010 }
1011
1012 #ifdef CONFIG_COMPAT
1013 static void compat_standard_from_user(void *dst, const void *src)
1014 {
1015         int v = *(compat_int_t *)src;
1016
1017         if (v > 0)
1018                 v += xt_compat_calc_jump(AF_INET, v);
1019         memcpy(dst, &v, sizeof(v));
1020 }
1021
1022 static int compat_standard_to_user(void __user *dst, const void *src)
1023 {
1024         compat_int_t cv = *(int *)src;
1025
1026         if (cv > 0)
1027                 cv -= xt_compat_calc_jump(AF_INET, cv);
1028         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1029 }
1030
1031 static int compat_calc_entry(const struct ipt_entry *e,
1032                              const struct xt_table_info *info,
1033                              const void *base, struct xt_table_info *newinfo)
1034 {
1035         const struct xt_entry_match *ematch;
1036         const struct xt_entry_target *t;
1037         unsigned int entry_offset;
1038         int off, i, ret;
1039
1040         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1041         entry_offset = (void *)e - base;
1042         xt_ematch_foreach(ematch, e)
1043                 off += xt_compat_match_offset(ematch->u.kernel.match);
1044         t = ipt_get_target_c(e);
1045         off += xt_compat_target_offset(t->u.kernel.target);
1046         newinfo->size -= off;
1047         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1048         if (ret)
1049                 return ret;
1050
1051         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1052                 if (info->hook_entry[i] &&
1053                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1054                         newinfo->hook_entry[i] -= off;
1055                 if (info->underflow[i] &&
1056                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1057                         newinfo->underflow[i] -= off;
1058         }
1059         return 0;
1060 }
1061
1062 static int compat_table_info(const struct xt_table_info *info,
1063                              struct xt_table_info *newinfo)
1064 {
1065         struct ipt_entry *iter;
1066         const void *loc_cpu_entry;
1067         int ret;
1068
1069         if (!newinfo || !info)
1070                 return -EINVAL;
1071
1072         /* we dont care about newinfo->entries */
1073         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1074         newinfo->initial_entries = 0;
1075         loc_cpu_entry = info->entries;
1076         xt_compat_init_offsets(AF_INET, info->number);
1077         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1078                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1079                 if (ret != 0)
1080                         return ret;
1081         }
1082         return 0;
1083 }
1084 #endif
1085
1086 static int get_info(struct net *net, void __user *user,
1087                     const int *len, int compat)
1088 {
1089         char name[XT_TABLE_MAXNAMELEN];
1090         struct xt_table *t;
1091         int ret;
1092
1093         if (*len != sizeof(struct ipt_getinfo)) {
1094                 duprintf("length %u != %zu\n", *len,
1095                          sizeof(struct ipt_getinfo));
1096                 return -EINVAL;
1097         }
1098
1099         if (copy_from_user(name, user, sizeof(name)) != 0)
1100                 return -EFAULT;
1101
1102         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1103 #ifdef CONFIG_COMPAT
1104         if (compat)
1105                 xt_compat_lock(AF_INET);
1106 #endif
1107         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1108                                     "iptable_%s", name);
1109         if (!IS_ERR_OR_NULL(t)) {
1110                 struct ipt_getinfo info;
1111                 const struct xt_table_info *private = t->private;
1112 #ifdef CONFIG_COMPAT
1113                 struct xt_table_info tmp;
1114
1115                 if (compat) {
1116                         ret = compat_table_info(private, &tmp);
1117                         xt_compat_flush_offsets(AF_INET);
1118                         private = &tmp;
1119                 }
1120 #endif
1121                 memset(&info, 0, sizeof(info));
1122                 info.valid_hooks = t->valid_hooks;
1123                 memcpy(info.hook_entry, private->hook_entry,
1124                        sizeof(info.hook_entry));
1125                 memcpy(info.underflow, private->underflow,
1126                        sizeof(info.underflow));
1127                 info.num_entries = private->number;
1128                 info.size = private->size;
1129                 strcpy(info.name, name);
1130
1131                 if (copy_to_user(user, &info, *len) != 0)
1132                         ret = -EFAULT;
1133                 else
1134                         ret = 0;
1135
1136                 xt_table_unlock(t);
1137                 module_put(t->me);
1138         } else
1139                 ret = t ? PTR_ERR(t) : -ENOENT;
1140 #ifdef CONFIG_COMPAT
1141         if (compat)
1142                 xt_compat_unlock(AF_INET);
1143 #endif
1144         return ret;
1145 }
1146
1147 static int
1148 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1149             const int *len)
1150 {
1151         int ret;
1152         struct ipt_get_entries get;
1153         struct xt_table *t;
1154
1155         if (*len < sizeof(get)) {
1156                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1157                 return -EINVAL;
1158         }
1159         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1160                 return -EFAULT;
1161         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1162                 duprintf("get_entries: %u != %zu\n",
1163                          *len, sizeof(get) + get.size);
1164                 return -EINVAL;
1165         }
1166
1167         t = xt_find_table_lock(net, AF_INET, get.name);
1168         if (!IS_ERR_OR_NULL(t)) {
1169                 const struct xt_table_info *private = t->private;
1170                 duprintf("t->private->number = %u\n", private->number);
1171                 if (get.size == private->size)
1172                         ret = copy_entries_to_user(private->size,
1173                                                    t, uptr->entrytable);
1174                 else {
1175                         duprintf("get_entries: I've got %u not %u!\n",
1176                                  private->size, get.size);
1177                         ret = -EAGAIN;
1178                 }
1179                 module_put(t->me);
1180                 xt_table_unlock(t);
1181         } else
1182                 ret = t ? PTR_ERR(t) : -ENOENT;
1183
1184         return ret;
1185 }
1186
1187 static int
1188 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1189              struct xt_table_info *newinfo, unsigned int num_counters,
1190              void __user *counters_ptr)
1191 {
1192         int ret;
1193         struct xt_table *t;
1194         struct xt_table_info *oldinfo;
1195         struct xt_counters *counters;
1196         struct ipt_entry *iter;
1197
1198         ret = 0;
1199         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1200         if (!counters) {
1201                 ret = -ENOMEM;
1202                 goto out;
1203         }
1204
1205         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1206                                     "iptable_%s", name);
1207         if (IS_ERR_OR_NULL(t)) {
1208                 ret = t ? PTR_ERR(t) : -ENOENT;
1209                 goto free_newinfo_counters_untrans;
1210         }
1211
1212         /* You lied! */
1213         if (valid_hooks != t->valid_hooks) {
1214                 duprintf("Valid hook crap: %08X vs %08X\n",
1215                          valid_hooks, t->valid_hooks);
1216                 ret = -EINVAL;
1217                 goto put_module;
1218         }
1219
1220         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1221         if (!oldinfo)
1222                 goto put_module;
1223
1224         /* Update module usage count based on number of rules */
1225         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1226                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1227         if ((oldinfo->number > oldinfo->initial_entries) ||
1228             (newinfo->number <= oldinfo->initial_entries))
1229                 module_put(t->me);
1230         if ((oldinfo->number > oldinfo->initial_entries) &&
1231             (newinfo->number <= oldinfo->initial_entries))
1232                 module_put(t->me);
1233
1234         /* Get the old counters, and synchronize with replace */
1235         get_counters(oldinfo, counters);
1236
1237         /* Decrease module usage counts and free resource */
1238         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1239                 cleanup_entry(iter, net);
1240
1241         xt_free_table_info(oldinfo);
1242         if (copy_to_user(counters_ptr, counters,
1243                          sizeof(struct xt_counters) * num_counters) != 0) {
1244                 /* Silent error, can't fail, new table is already in place */
1245                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1246         }
1247         vfree(counters);
1248         xt_table_unlock(t);
1249         return ret;
1250
1251  put_module:
1252         module_put(t->me);
1253         xt_table_unlock(t);
1254  free_newinfo_counters_untrans:
1255         vfree(counters);
1256  out:
1257         return ret;
1258 }
1259
1260 static int
1261 do_replace(struct net *net, const void __user *user, unsigned int len)
1262 {
1263         int ret;
1264         struct ipt_replace tmp;
1265         struct xt_table_info *newinfo;
1266         void *loc_cpu_entry;
1267         struct ipt_entry *iter;
1268
1269         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1270                 return -EFAULT;
1271
1272         /* overflow check */
1273         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1274                 return -ENOMEM;
1275         if (tmp.num_counters == 0)
1276                 return -EINVAL;
1277
1278         tmp.name[sizeof(tmp.name)-1] = 0;
1279
1280         newinfo = xt_alloc_table_info(tmp.size);
1281         if (!newinfo)
1282                 return -ENOMEM;
1283
1284         loc_cpu_entry = newinfo->entries;
1285         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1286                            tmp.size) != 0) {
1287                 ret = -EFAULT;
1288                 goto free_newinfo;
1289         }
1290
1291         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1292         if (ret != 0)
1293                 goto free_newinfo;
1294
1295         duprintf("Translated table\n");
1296
1297         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1298                            tmp.num_counters, tmp.counters);
1299         if (ret)
1300                 goto free_newinfo_untrans;
1301         return 0;
1302
1303  free_newinfo_untrans:
1304         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1305                 cleanup_entry(iter, net);
1306  free_newinfo:
1307         xt_free_table_info(newinfo);
1308         return ret;
1309 }
1310
1311 static int
1312 do_add_counters(struct net *net, const void __user *user,
1313                 unsigned int len, int compat)
1314 {
1315         unsigned int i;
1316         struct xt_counters_info tmp;
1317         struct xt_counters *paddc;
1318         unsigned int num_counters;
1319         const char *name;
1320         int size;
1321         void *ptmp;
1322         struct xt_table *t;
1323         const struct xt_table_info *private;
1324         int ret = 0;
1325         struct ipt_entry *iter;
1326         unsigned int addend;
1327 #ifdef CONFIG_COMPAT
1328         struct compat_xt_counters_info compat_tmp;
1329
1330         if (compat) {
1331                 ptmp = &compat_tmp;
1332                 size = sizeof(struct compat_xt_counters_info);
1333         } else
1334 #endif
1335         {
1336                 ptmp = &tmp;
1337                 size = sizeof(struct xt_counters_info);
1338         }
1339
1340         if (copy_from_user(ptmp, user, size) != 0)
1341                 return -EFAULT;
1342
1343 #ifdef CONFIG_COMPAT
1344         if (compat) {
1345                 num_counters = compat_tmp.num_counters;
1346                 name = compat_tmp.name;
1347         } else
1348 #endif
1349         {
1350                 num_counters = tmp.num_counters;
1351                 name = tmp.name;
1352         }
1353
1354         if (len != size + num_counters * sizeof(struct xt_counters))
1355                 return -EINVAL;
1356
1357         paddc = vmalloc(len - size);
1358         if (!paddc)
1359                 return -ENOMEM;
1360
1361         if (copy_from_user(paddc, user + size, len - size) != 0) {
1362                 ret = -EFAULT;
1363                 goto free;
1364         }
1365
1366         t = xt_find_table_lock(net, AF_INET, name);
1367         if (IS_ERR_OR_NULL(t)) {
1368                 ret = t ? PTR_ERR(t) : -ENOENT;
1369                 goto free;
1370         }
1371
1372         local_bh_disable();
1373         private = t->private;
1374         if (private->number != num_counters) {
1375                 ret = -EINVAL;
1376                 goto unlock_up_free;
1377         }
1378
1379         i = 0;
1380         addend = xt_write_recseq_begin();
1381         xt_entry_foreach(iter, private->entries, private->size) {
1382                 struct xt_counters *tmp;
1383
1384                 tmp = xt_get_this_cpu_counter(&iter->counters);
1385                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1386                 ++i;
1387         }
1388         xt_write_recseq_end(addend);
1389  unlock_up_free:
1390         local_bh_enable();
1391         xt_table_unlock(t);
1392         module_put(t->me);
1393  free:
1394         vfree(paddc);
1395
1396         return ret;
1397 }
1398
1399 #ifdef CONFIG_COMPAT
1400 struct compat_ipt_replace {
1401         char                    name[XT_TABLE_MAXNAMELEN];
1402         u32                     valid_hooks;
1403         u32                     num_entries;
1404         u32                     size;
1405         u32                     hook_entry[NF_INET_NUMHOOKS];
1406         u32                     underflow[NF_INET_NUMHOOKS];
1407         u32                     num_counters;
1408         compat_uptr_t           counters;       /* struct xt_counters * */
1409         struct compat_ipt_entry entries[0];
1410 };
1411
1412 static int
1413 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1414                           unsigned int *size, struct xt_counters *counters,
1415                           unsigned int i)
1416 {
1417         struct xt_entry_target *t;
1418         struct compat_ipt_entry __user *ce;
1419         u_int16_t target_offset, next_offset;
1420         compat_uint_t origsize;
1421         const struct xt_entry_match *ematch;
1422         int ret = 0;
1423
1424         origsize = *size;
1425         ce = (struct compat_ipt_entry __user *)*dstptr;
1426         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1427             copy_to_user(&ce->counters, &counters[i],
1428             sizeof(counters[i])) != 0)
1429                 return -EFAULT;
1430
1431         *dstptr += sizeof(struct compat_ipt_entry);
1432         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1433
1434         xt_ematch_foreach(ematch, e) {
1435                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1436                 if (ret != 0)
1437                         return ret;
1438         }
1439         target_offset = e->target_offset - (origsize - *size);
1440         t = ipt_get_target(e);
1441         ret = xt_compat_target_to_user(t, dstptr, size);
1442         if (ret)
1443                 return ret;
1444         next_offset = e->next_offset - (origsize - *size);
1445         if (put_user(target_offset, &ce->target_offset) != 0 ||
1446             put_user(next_offset, &ce->next_offset) != 0)
1447                 return -EFAULT;
1448         return 0;
1449 }
1450
1451 static int
1452 compat_find_calc_match(struct xt_entry_match *m,
1453                        const char *name,
1454                        const struct ipt_ip *ip,
1455                        int *size)
1456 {
1457         struct xt_match *match;
1458
1459         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1460                                       m->u.user.revision);
1461         if (IS_ERR(match)) {
1462                 duprintf("compat_check_calc_match: `%s' not found\n",
1463                          m->u.user.name);
1464                 return PTR_ERR(match);
1465         }
1466         m->u.kernel.match = match;
1467         *size += xt_compat_match_offset(match);
1468         return 0;
1469 }
1470
1471 static void compat_release_entry(struct compat_ipt_entry *e)
1472 {
1473         struct xt_entry_target *t;
1474         struct xt_entry_match *ematch;
1475
1476         /* Cleanup all matches */
1477         xt_ematch_foreach(ematch, e)
1478                 module_put(ematch->u.kernel.match->me);
1479         t = compat_ipt_get_target(e);
1480         module_put(t->u.kernel.target->me);
1481 }
1482
1483 static int
1484 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1485                                   struct xt_table_info *newinfo,
1486                                   unsigned int *size,
1487                                   const unsigned char *base,
1488                                   const unsigned char *limit,
1489                                   const unsigned int *hook_entries,
1490                                   const unsigned int *underflows,
1491                                   const char *name)
1492 {
1493         struct xt_entry_match *ematch;
1494         struct xt_entry_target *t;
1495         struct xt_target *target;
1496         unsigned int entry_offset;
1497         unsigned int j;
1498         int ret, off, h;
1499
1500         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1501         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1502             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1503                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1504                 return -EINVAL;
1505         }
1506
1507         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1508                              sizeof(struct compat_xt_entry_target)) {
1509                 duprintf("checking: element %p size %u\n",
1510                          e, e->next_offset);
1511                 return -EINVAL;
1512         }
1513
1514         /* For purposes of check_entry casting the compat entry is fine */
1515         ret = check_entry((struct ipt_entry *)e, name);
1516         if (ret)
1517                 return ret;
1518
1519         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1520         entry_offset = (void *)e - (void *)base;
1521         j = 0;
1522         xt_ematch_foreach(ematch, e) {
1523                 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1524                 if (ret != 0)
1525                         goto release_matches;
1526                 ++j;
1527         }
1528
1529         t = compat_ipt_get_target(e);
1530         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1531                                         t->u.user.revision);
1532         if (IS_ERR(target)) {
1533                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1534                          t->u.user.name);
1535                 ret = PTR_ERR(target);
1536                 goto release_matches;
1537         }
1538         t->u.kernel.target = target;
1539
1540         off += xt_compat_target_offset(target);
1541         *size += off;
1542         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1543         if (ret)
1544                 goto out;
1545
1546         /* Check hooks & underflows */
1547         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1548                 if ((unsigned char *)e - base == hook_entries[h])
1549                         newinfo->hook_entry[h] = hook_entries[h];
1550                 if ((unsigned char *)e - base == underflows[h])
1551                         newinfo->underflow[h] = underflows[h];
1552         }
1553
1554         /* Clear counters and comefrom */
1555         memset(&e->counters, 0, sizeof(e->counters));
1556         e->comefrom = 0;
1557         return 0;
1558
1559 out:
1560         module_put(t->u.kernel.target->me);
1561 release_matches:
1562         xt_ematch_foreach(ematch, e) {
1563                 if (j-- == 0)
1564                         break;
1565                 module_put(ematch->u.kernel.match->me);
1566         }
1567         return ret;
1568 }
1569
1570 static int
1571 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1572                             unsigned int *size, const char *name,
1573                             struct xt_table_info *newinfo, unsigned char *base)
1574 {
1575         struct xt_entry_target *t;
1576         struct xt_target *target;
1577         struct ipt_entry *de;
1578         unsigned int origsize;
1579         int ret, h;
1580         struct xt_entry_match *ematch;
1581
1582         ret = 0;
1583         origsize = *size;
1584         de = (struct ipt_entry *)*dstptr;
1585         memcpy(de, e, sizeof(struct ipt_entry));
1586         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1587
1588         *dstptr += sizeof(struct ipt_entry);
1589         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1590
1591         xt_ematch_foreach(ematch, e) {
1592                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1593                 if (ret != 0)
1594                         return ret;
1595         }
1596         de->target_offset = e->target_offset - (origsize - *size);
1597         t = compat_ipt_get_target(e);
1598         target = t->u.kernel.target;
1599         xt_compat_target_from_user(t, dstptr, size);
1600
1601         de->next_offset = e->next_offset - (origsize - *size);
1602         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1603                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1604                         newinfo->hook_entry[h] -= origsize - *size;
1605                 if ((unsigned char *)de - base < newinfo->underflow[h])
1606                         newinfo->underflow[h] -= origsize - *size;
1607         }
1608         return ret;
1609 }
1610
1611 static int
1612 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1613 {
1614         struct xt_entry_match *ematch;
1615         struct xt_mtchk_param mtpar;
1616         unsigned int j;
1617         int ret = 0;
1618
1619         e->counters.pcnt = xt_percpu_counter_alloc();
1620         if (IS_ERR_VALUE(e->counters.pcnt))
1621                 return -ENOMEM;
1622
1623         j = 0;
1624         mtpar.net       = net;
1625         mtpar.table     = name;
1626         mtpar.entryinfo = &e->ip;
1627         mtpar.hook_mask = e->comefrom;
1628         mtpar.family    = NFPROTO_IPV4;
1629         xt_ematch_foreach(ematch, e) {
1630                 ret = check_match(ematch, &mtpar);
1631                 if (ret != 0)
1632                         goto cleanup_matches;
1633                 ++j;
1634         }
1635
1636         ret = check_target(e, net, name);
1637         if (ret)
1638                 goto cleanup_matches;
1639         return 0;
1640
1641  cleanup_matches:
1642         xt_ematch_foreach(ematch, e) {
1643                 if (j-- == 0)
1644                         break;
1645                 cleanup_match(ematch, net);
1646         }
1647
1648         xt_percpu_counter_free(e->counters.pcnt);
1649
1650         return ret;
1651 }
1652
1653 static int
1654 translate_compat_table(struct net *net,
1655                        const char *name,
1656                        unsigned int valid_hooks,
1657                        struct xt_table_info **pinfo,
1658                        void **pentry0,
1659                        unsigned int total_size,
1660                        unsigned int number,
1661                        unsigned int *hook_entries,
1662                        unsigned int *underflows)
1663 {
1664         unsigned int i, j;
1665         struct xt_table_info *newinfo, *info;
1666         void *pos, *entry0, *entry1;
1667         struct compat_ipt_entry *iter0;
1668         struct ipt_entry *iter1;
1669         unsigned int size;
1670         int ret;
1671
1672         info = *pinfo;
1673         entry0 = *pentry0;
1674         size = total_size;
1675         info->number = number;
1676
1677         /* Init all hooks to impossible value. */
1678         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1679                 info->hook_entry[i] = 0xFFFFFFFF;
1680                 info->underflow[i] = 0xFFFFFFFF;
1681         }
1682
1683         duprintf("translate_compat_table: size %u\n", info->size);
1684         j = 0;
1685         xt_compat_lock(AF_INET);
1686         xt_compat_init_offsets(AF_INET, number);
1687         /* Walk through entries, checking offsets. */
1688         xt_entry_foreach(iter0, entry0, total_size) {
1689                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1690                                                         entry0,
1691                                                         entry0 + total_size,
1692                                                         hook_entries,
1693                                                         underflows,
1694                                                         name);
1695                 if (ret != 0)
1696                         goto out_unlock;
1697                 ++j;
1698         }
1699
1700         ret = -EINVAL;
1701         if (j != number) {
1702                 duprintf("translate_compat_table: %u not %u entries\n",
1703                          j, number);
1704                 goto out_unlock;
1705         }
1706
1707         /* Check hooks all assigned */
1708         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1709                 /* Only hooks which are valid */
1710                 if (!(valid_hooks & (1 << i)))
1711                         continue;
1712                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1713                         duprintf("Invalid hook entry %u %u\n",
1714                                  i, hook_entries[i]);
1715                         goto out_unlock;
1716                 }
1717                 if (info->underflow[i] == 0xFFFFFFFF) {
1718                         duprintf("Invalid underflow %u %u\n",
1719                                  i, underflows[i]);
1720                         goto out_unlock;
1721                 }
1722         }
1723
1724         ret = -ENOMEM;
1725         newinfo = xt_alloc_table_info(size);
1726         if (!newinfo)
1727                 goto out_unlock;
1728
1729         newinfo->number = number;
1730         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1731                 newinfo->hook_entry[i] = info->hook_entry[i];
1732                 newinfo->underflow[i] = info->underflow[i];
1733         }
1734         entry1 = newinfo->entries;
1735         pos = entry1;
1736         size = total_size;
1737         xt_entry_foreach(iter0, entry0, total_size) {
1738                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1739                                                   name, newinfo, entry1);
1740                 if (ret != 0)
1741                         break;
1742         }
1743         xt_compat_flush_offsets(AF_INET);
1744         xt_compat_unlock(AF_INET);
1745         if (ret)
1746                 goto free_newinfo;
1747
1748         ret = -ELOOP;
1749         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1750                 goto free_newinfo;
1751
1752         i = 0;
1753         xt_entry_foreach(iter1, entry1, newinfo->size) {
1754                 ret = compat_check_entry(iter1, net, name);
1755                 if (ret != 0)
1756                         break;
1757                 ++i;
1758         }
1759         if (ret) {
1760                 /*
1761                  * The first i matches need cleanup_entry (calls ->destroy)
1762                  * because they had called ->check already. The other j-i
1763                  * entries need only release.
1764                  */
1765                 int skip = i;
1766                 j -= i;
1767                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1768                         if (skip-- > 0)
1769                                 continue;
1770                         if (j-- == 0)
1771                                 break;
1772                         compat_release_entry(iter0);
1773                 }
1774                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1775                         if (i-- == 0)
1776                                 break;
1777                         cleanup_entry(iter1, net);
1778                 }
1779                 xt_free_table_info(newinfo);
1780                 return ret;
1781         }
1782
1783         *pinfo = newinfo;
1784         *pentry0 = entry1;
1785         xt_free_table_info(info);
1786         return 0;
1787
1788 free_newinfo:
1789         xt_free_table_info(newinfo);
1790 out:
1791         xt_entry_foreach(iter0, entry0, total_size) {
1792                 if (j-- == 0)
1793                         break;
1794                 compat_release_entry(iter0);
1795         }
1796         return ret;
1797 out_unlock:
1798         xt_compat_flush_offsets(AF_INET);
1799         xt_compat_unlock(AF_INET);
1800         goto out;
1801 }
1802
1803 static int
1804 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1805 {
1806         int ret;
1807         struct compat_ipt_replace tmp;
1808         struct xt_table_info *newinfo;
1809         void *loc_cpu_entry;
1810         struct ipt_entry *iter;
1811
1812         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1813                 return -EFAULT;
1814
1815         /* overflow check */
1816         if (tmp.size >= INT_MAX / num_possible_cpus())
1817                 return -ENOMEM;
1818         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1819                 return -ENOMEM;
1820         if (tmp.num_counters == 0)
1821                 return -EINVAL;
1822
1823         tmp.name[sizeof(tmp.name)-1] = 0;
1824
1825         newinfo = xt_alloc_table_info(tmp.size);
1826         if (!newinfo)
1827                 return -ENOMEM;
1828
1829         loc_cpu_entry = newinfo->entries;
1830         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1831                            tmp.size) != 0) {
1832                 ret = -EFAULT;
1833                 goto free_newinfo;
1834         }
1835
1836         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1837                                      &newinfo, &loc_cpu_entry, tmp.size,
1838                                      tmp.num_entries, tmp.hook_entry,
1839                                      tmp.underflow);
1840         if (ret != 0)
1841                 goto free_newinfo;
1842
1843         duprintf("compat_do_replace: Translated table\n");
1844
1845         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1846                            tmp.num_counters, compat_ptr(tmp.counters));
1847         if (ret)
1848                 goto free_newinfo_untrans;
1849         return 0;
1850
1851  free_newinfo_untrans:
1852         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1853                 cleanup_entry(iter, net);
1854  free_newinfo:
1855         xt_free_table_info(newinfo);
1856         return ret;
1857 }
1858
1859 static int
1860 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1861                       unsigned int len)
1862 {
1863         int ret;
1864
1865         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1866                 return -EPERM;
1867
1868         switch (cmd) {
1869         case IPT_SO_SET_REPLACE:
1870                 ret = compat_do_replace(sock_net(sk), user, len);
1871                 break;
1872
1873         case IPT_SO_SET_ADD_COUNTERS:
1874                 ret = do_add_counters(sock_net(sk), user, len, 1);
1875                 break;
1876
1877         default:
1878                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1879                 ret = -EINVAL;
1880         }
1881
1882         return ret;
1883 }
1884
1885 struct compat_ipt_get_entries {
1886         char name[XT_TABLE_MAXNAMELEN];
1887         compat_uint_t size;
1888         struct compat_ipt_entry entrytable[0];
1889 };
1890
1891 static int
1892 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1893                             void __user *userptr)
1894 {
1895         struct xt_counters *counters;
1896         const struct xt_table_info *private = table->private;
1897         void __user *pos;
1898         unsigned int size;
1899         int ret = 0;
1900         unsigned int i = 0;
1901         struct ipt_entry *iter;
1902
1903         counters = alloc_counters(table);
1904         if (IS_ERR(counters))
1905                 return PTR_ERR(counters);
1906
1907         pos = userptr;
1908         size = total_size;
1909         xt_entry_foreach(iter, private->entries, total_size) {
1910                 ret = compat_copy_entry_to_user(iter, &pos,
1911                                                 &size, counters, i++);
1912                 if (ret != 0)
1913                         break;
1914         }
1915
1916         vfree(counters);
1917         return ret;
1918 }
1919
1920 static int
1921 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1922                    int *len)
1923 {
1924         int ret;
1925         struct compat_ipt_get_entries get;
1926         struct xt_table *t;
1927
1928         if (*len < sizeof(get)) {
1929                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1930                 return -EINVAL;
1931         }
1932
1933         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1934                 return -EFAULT;
1935
1936         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1937                 duprintf("compat_get_entries: %u != %zu\n",
1938                          *len, sizeof(get) + get.size);
1939                 return -EINVAL;
1940         }
1941
1942         xt_compat_lock(AF_INET);
1943         t = xt_find_table_lock(net, AF_INET, get.name);
1944         if (!IS_ERR_OR_NULL(t)) {
1945                 const struct xt_table_info *private = t->private;
1946                 struct xt_table_info info;
1947                 duprintf("t->private->number = %u\n", private->number);
1948                 ret = compat_table_info(private, &info);
1949                 if (!ret && get.size == info.size) {
1950                         ret = compat_copy_entries_to_user(private->size,
1951                                                           t, uptr->entrytable);
1952                 } else if (!ret) {
1953                         duprintf("compat_get_entries: I've got %u not %u!\n",
1954                                  private->size, get.size);
1955                         ret = -EAGAIN;
1956                 }
1957                 xt_compat_flush_offsets(AF_INET);
1958                 module_put(t->me);
1959                 xt_table_unlock(t);
1960         } else
1961                 ret = t ? PTR_ERR(t) : -ENOENT;
1962
1963         xt_compat_unlock(AF_INET);
1964         return ret;
1965 }
1966
1967 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1968
1969 static int
1970 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1971 {
1972         int ret;
1973
1974         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1975                 return -EPERM;
1976
1977         switch (cmd) {
1978         case IPT_SO_GET_INFO:
1979                 ret = get_info(sock_net(sk), user, len, 1);
1980                 break;
1981         case IPT_SO_GET_ENTRIES:
1982                 ret = compat_get_entries(sock_net(sk), user, len);
1983                 break;
1984         default:
1985                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1986         }
1987         return ret;
1988 }
1989 #endif
1990
1991 static int
1992 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1993 {
1994         int ret;
1995
1996         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1997                 return -EPERM;
1998
1999         switch (cmd) {
2000         case IPT_SO_SET_REPLACE:
2001                 ret = do_replace(sock_net(sk), user, len);
2002                 break;
2003
2004         case IPT_SO_SET_ADD_COUNTERS:
2005                 ret = do_add_counters(sock_net(sk), user, len, 0);
2006                 break;
2007
2008         default:
2009                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2010                 ret = -EINVAL;
2011         }
2012
2013         return ret;
2014 }
2015
2016 static int
2017 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2018 {
2019         int ret;
2020
2021         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2022                 return -EPERM;
2023
2024         switch (cmd) {
2025         case IPT_SO_GET_INFO:
2026                 ret = get_info(sock_net(sk), user, len, 0);
2027                 break;
2028
2029         case IPT_SO_GET_ENTRIES:
2030                 ret = get_entries(sock_net(sk), user, len);
2031                 break;
2032
2033         case IPT_SO_GET_REVISION_MATCH:
2034         case IPT_SO_GET_REVISION_TARGET: {
2035                 struct xt_get_revision rev;
2036                 int target;
2037
2038                 if (*len != sizeof(rev)) {
2039                         ret = -EINVAL;
2040                         break;
2041                 }
2042                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2043                         ret = -EFAULT;
2044                         break;
2045                 }
2046                 rev.name[sizeof(rev.name)-1] = 0;
2047
2048                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2049                         target = 1;
2050                 else
2051                         target = 0;
2052
2053                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2054                                                          rev.revision,
2055                                                          target, &ret),
2056                                         "ipt_%s", rev.name);
2057                 break;
2058         }
2059
2060         default:
2061                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2062                 ret = -EINVAL;
2063         }
2064
2065         return ret;
2066 }
2067
2068 struct xt_table *ipt_register_table(struct net *net,
2069                                     const struct xt_table *table,
2070                                     const struct ipt_replace *repl)
2071 {
2072         int ret;
2073         struct xt_table_info *newinfo;
2074         struct xt_table_info bootstrap = {0};
2075         void *loc_cpu_entry;
2076         struct xt_table *new_table;
2077
2078         newinfo = xt_alloc_table_info(repl->size);
2079         if (!newinfo) {
2080                 ret = -ENOMEM;
2081                 goto out;
2082         }
2083
2084         loc_cpu_entry = newinfo->entries;
2085         memcpy(loc_cpu_entry, repl->entries, repl->size);
2086
2087         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2088         if (ret != 0)
2089                 goto out_free;
2090
2091         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2092         if (IS_ERR(new_table)) {
2093                 ret = PTR_ERR(new_table);
2094                 goto out_free;
2095         }
2096
2097         return new_table;
2098
2099 out_free:
2100         xt_free_table_info(newinfo);
2101 out:
2102         return ERR_PTR(ret);
2103 }
2104
2105 void ipt_unregister_table(struct net *net, struct xt_table *table)
2106 {
2107         struct xt_table_info *private;
2108         void *loc_cpu_entry;
2109         struct module *table_owner = table->me;
2110         struct ipt_entry *iter;
2111
2112         private = xt_unregister_table(table);
2113
2114         /* Decrease module usage counts and free resources */
2115         loc_cpu_entry = private->entries;
2116         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2117                 cleanup_entry(iter, net);
2118         if (private->number > private->initial_entries)
2119                 module_put(table_owner);
2120         xt_free_table_info(private);
2121 }
2122
2123 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2124 static inline bool
2125 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2126                      u_int8_t type, u_int8_t code,
2127                      bool invert)
2128 {
2129         return ((test_type == 0xFF) ||
2130                 (type == test_type && code >= min_code && code <= max_code))
2131                 ^ invert;
2132 }
2133
2134 static bool
2135 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2136 {
2137         const struct icmphdr *ic;
2138         struct icmphdr _icmph;
2139         const struct ipt_icmp *icmpinfo = par->matchinfo;
2140
2141         /* Must not be a fragment. */
2142         if (par->fragoff != 0)
2143                 return false;
2144
2145         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2146         if (ic == NULL) {
2147                 /* We've been asked to examine this packet, and we
2148                  * can't.  Hence, no choice but to drop.
2149                  */
2150                 duprintf("Dropping evil ICMP tinygram.\n");
2151                 par->hotdrop = true;
2152                 return false;
2153         }
2154
2155         return icmp_type_code_match(icmpinfo->type,
2156                                     icmpinfo->code[0],
2157                                     icmpinfo->code[1],
2158                                     ic->type, ic->code,
2159                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2160 }
2161
2162 static int icmp_checkentry(const struct xt_mtchk_param *par)
2163 {
2164         const struct ipt_icmp *icmpinfo = par->matchinfo;
2165
2166         /* Must specify no unknown invflags */
2167         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2168 }
2169
2170 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2171         {
2172                 .name             = XT_STANDARD_TARGET,
2173                 .targetsize       = sizeof(int),
2174                 .family           = NFPROTO_IPV4,
2175 #ifdef CONFIG_COMPAT
2176                 .compatsize       = sizeof(compat_int_t),
2177                 .compat_from_user = compat_standard_from_user,
2178                 .compat_to_user   = compat_standard_to_user,
2179 #endif
2180         },
2181         {
2182                 .name             = XT_ERROR_TARGET,
2183                 .target           = ipt_error,
2184                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2185                 .family           = NFPROTO_IPV4,
2186         },
2187 };
2188
2189 static struct nf_sockopt_ops ipt_sockopts = {
2190         .pf             = PF_INET,
2191         .set_optmin     = IPT_BASE_CTL,
2192         .set_optmax     = IPT_SO_SET_MAX+1,
2193         .set            = do_ipt_set_ctl,
2194 #ifdef CONFIG_COMPAT
2195         .compat_set     = compat_do_ipt_set_ctl,
2196 #endif
2197         .get_optmin     = IPT_BASE_CTL,
2198         .get_optmax     = IPT_SO_GET_MAX+1,
2199         .get            = do_ipt_get_ctl,
2200 #ifdef CONFIG_COMPAT
2201         .compat_get     = compat_do_ipt_get_ctl,
2202 #endif
2203         .owner          = THIS_MODULE,
2204 };
2205
2206 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2207         {
2208                 .name       = "icmp",
2209                 .match      = icmp_match,
2210                 .matchsize  = sizeof(struct ipt_icmp),
2211                 .checkentry = icmp_checkentry,
2212                 .proto      = IPPROTO_ICMP,
2213                 .family     = NFPROTO_IPV4,
2214         },
2215 };
2216
2217 static int __net_init ip_tables_net_init(struct net *net)
2218 {
2219         return xt_proto_init(net, NFPROTO_IPV4);
2220 }
2221
2222 static void __net_exit ip_tables_net_exit(struct net *net)
2223 {
2224         xt_proto_fini(net, NFPROTO_IPV4);
2225 }
2226
2227 static struct pernet_operations ip_tables_net_ops = {
2228         .init = ip_tables_net_init,
2229         .exit = ip_tables_net_exit,
2230 };
2231
2232 static int __init ip_tables_init(void)
2233 {
2234         int ret;
2235
2236         ret = register_pernet_subsys(&ip_tables_net_ops);
2237         if (ret < 0)
2238                 goto err1;
2239
2240         /* No one else will be downing sem now, so we won't sleep */
2241         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2242         if (ret < 0)
2243                 goto err2;
2244         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2245         if (ret < 0)
2246                 goto err4;
2247
2248         /* Register setsockopt */
2249         ret = nf_register_sockopt(&ipt_sockopts);
2250         if (ret < 0)
2251                 goto err5;
2252
2253         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2254         return 0;
2255
2256 err5:
2257         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2258 err4:
2259         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2260 err2:
2261         unregister_pernet_subsys(&ip_tables_net_ops);
2262 err1:
2263         return ret;
2264 }
2265
2266 static void __exit ip_tables_fini(void)
2267 {
2268         nf_unregister_sockopt(&ipt_sockopts);
2269
2270         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2271         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2272         unregister_pernet_subsys(&ip_tables_net_ops);
2273 }
2274
2275 EXPORT_SYMBOL(ipt_register_table);
2276 EXPORT_SYMBOL(ipt_unregister_table);
2277 EXPORT_SYMBOL(ipt_do_table);
2278 module_init(ip_tables_init);
2279 module_exit(ip_tables_fini);