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(struct net *net,
 250                          const struct sk_buff *skb,
 251                          unsigned int hook,
 252                          const struct net_device *in,
 253                          const struct net_device *out,
 254                          const char *tablename,
 255                          const struct xt_table_info *private,
 256                          const struct ipt_entry *e)
 257 {
 258         const struct ipt_entry *root;
 259         const char *hookname, *chainname, *comment;
 260         const struct ipt_entry *iter;
 261         unsigned int rulenum = 0;
 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
 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              const struct nf_hook_state *state,
 289              struct xt_table *table)
 290 {
 291         unsigned int hook = state->hook;
 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 stackidx, cpu;
 300         const struct xt_table_info *private;
 301         struct xt_action_param acpar;
 302         unsigned int addend;
 303
 304         /* Initialization */
 305         stackidx = 0;
 306         ip = ip_hdr(skb);
 307         indev = state->in ? state->in->name : nulldevname;
 308         outdev = state->out ? state->out->name : nulldevname;
 309         /* We handle fragments by dealing with the first fragment as
 310          * if it was a normal packet.  All other fragments are treated
 311          * normally, except that they will NEVER match rules that ask
 312          * things we don't know, ie. tcp syn flag or ports).  If the
 313          * rule is also a fragment-specific rule, non-fragments won't
 314          * match it. */
 315         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 316         acpar.thoff   = ip_hdrlen(skb);
 317         acpar.hotdrop = false;
 318         acpar.net     = state->net;
 319         acpar.in      = state->in;
 320         acpar.out     = state->out;
 321         acpar.family  = NFPROTO_IPV4;
 322         acpar.hooknum = hook;
 323
 324         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 325         local_bh_disable();
 326         addend = xt_write_recseq_begin();
 327         private = table->private;
 328         cpu        = smp_processor_id();
 329         /*
 330          * Ensure we load private-> members after we've fetched the base
 331          * pointer.
 332          */
 333         smp_read_barrier_depends();
 334         table_base = private->entries;
 335         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 336
 337         /* Switch to alternate jumpstack if we're being invoked via TEE.
 338          * TEE issues XT_CONTINUE verdict on original skb so we must not
 339          * clobber the jumpstack.
 340          *
 341          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 342          * but it is no problem since absolute verdict is issued by these.
 343          */
 344         if (static_key_false(&xt_tee_enabled))
 345                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 346
 347         e = get_entry(table_base, private->hook_entry[hook]);
 348
 349         pr_debug("Entering %s(hook %u), UF %p\n",
 350                  table->name, hook,
 351                  get_entry(table_base, private->underflow[hook]));
 352
 353         do {
 354                 const struct xt_entry_target *t;
 355                 const struct xt_entry_match *ematch;
 356                 struct xt_counters *counter;
 357
 358                 IP_NF_ASSERT(e);
 359                 if (!ip_packet_match(ip, indev, outdev,
 360                     &e->ip, acpar.fragoff)) {
 361  no_match:
 362                         e = ipt_next_entry(e);
 363                         continue;
 364                 }
 365
 366                 xt_ematch_foreach(ematch, e) {
 367                         acpar.match     = ematch->u.kernel.match;
 368                         acpar.matchinfo = ematch->data;
 369                         if (!acpar.match->match(skb, &acpar))
 370                                 goto no_match;
 371                 }
 372
 373                 counter = xt_get_this_cpu_counter(&e->counters);
 374                 ADD_COUNTER(*counter, skb->len, 1);
 375
 376                 t = ipt_get_target(e);
 377                 IP_NF_ASSERT(t->u.kernel.target);
 378
 379 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 380                 /* The packet is traced: log it */
 381                 if (unlikely(skb->nf_trace))
 382                         trace_packet(state->net, skb, hook, state->in,
 383                                      state->out, table->name, private, e);
 384 #endif
 385                 /* Standard target? */
 386                 if (!t->u.kernel.target->target) {
 387                         int v;
 388
 389                         v = ((struct xt_standard_target *)t)->verdict;
 390                         if (v < 0) {
 391                                 /* Pop from stack? */
 392                                 if (v != XT_RETURN) {
 393                                         verdict = (unsigned int)(-v) - 1;
 394                                         break;
 395                                 }
 396                                 if (stackidx == 0) {
 397                                         e = get_entry(table_base,
 398                                             private->underflow[hook]);
 399                                         pr_debug("Underflow (this is normal) "
 400                                                  "to %p\n", e);
 401                                 } else {
 402                                         e = jumpstack[--stackidx];
 403                                         pr_debug("Pulled %p out from pos %u\n",
 404                                                  e, stackidx);
 405                                         e = ipt_next_entry(e);
 406                                 }
 407                                 continue;
 408                         }
 409                         if (table_base + v != ipt_next_entry(e) &&
 410                             !(e->ip.flags & IPT_F_GOTO)) {
 411                                 jumpstack[stackidx++] = e;
 412                                 pr_debug("Pushed %p into pos %u\n",
 413                                          e, stackidx - 1);
 414                         }
 415
 416                         e = get_entry(table_base, v);
 417                         continue;
 418                 }
 419
 420                 acpar.target   = t->u.kernel.target;
 421                 acpar.targinfo = t->data;
 422
 423                 verdict = t->u.kernel.target->target(skb, &acpar);
 424                 /* Target might have changed stuff. */
 425                 ip = ip_hdr(skb);
 426                 if (verdict == XT_CONTINUE)
 427                         e = ipt_next_entry(e);
 428                 else
 429                         /* Verdict */
 430                         break;
 431         } while (!acpar.hotdrop);
 432         pr_debug("Exiting %s; sp at %u\n", __func__, stackidx);
 433
 434         xt_write_recseq_end(addend);
 435         local_bh_enable();
 436
 437 #ifdef DEBUG_ALLOW_ALL
 438         return NF_ACCEPT;
 439 #else
 440         if (acpar.hotdrop)
 441                 return NF_DROP;
 442         else return verdict;
 443 #endif
 444 }
 445
 446 /* Figures out from what hook each rule can be called: returns 0 if
 447    there are loops.  Puts hook bitmask in comefrom. */
 448 static int
 449 mark_source_chains(const struct xt_table_info *newinfo,
 450                    unsigned int valid_hooks, void *entry0)
 451 {
 452         unsigned int hook;
 453
 454         /* No recursion; use packet counter to save back ptrs (reset
 455            to 0 as we leave), and comefrom to save source hook bitmask */
 456         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 457                 unsigned int pos = newinfo->hook_entry[hook];
 458                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 459
 460                 if (!(valid_hooks & (1 << hook)))
 461                         continue;
 462
 463                 /* Set initial back pointer. */
 464                 e->counters.pcnt = pos;
 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                         } else {
 527                                 int newpos = t->verdict;
 528
 529                                 if (strcmp(t->target.u.user.name,
 530                                            XT_STANDARD_TARGET) == 0 &&
 531                                     newpos >= 0) {
 532                                         if (newpos > newinfo->size -
 533                                                 sizeof(struct ipt_entry)) {
 534                                                 duprintf("mark_source_chains: "
 535                                                         "bad verdict (%i)\n",
 536                                                                 newpos);
 537                                                 return 0;
 538                                         }
 539                                         /* This a jump; chase it. */
 540                                         duprintf("Jump rule %u -> %u\n",
 541                                                  pos, newpos);
 542                                 } else {
 543                                         /* ... this is a fallthru */
 544                                         newpos = pos + e->next_offset;
 545                                 }
 546                                 e = (struct ipt_entry *)
 547                                         (entry0 + newpos);
 548                                 e->counters.pcnt = pos;
 549                                 pos = newpos;
 550                         }
 551                 }
 552 next:
 553                 duprintf("Finished chain %u\n", hook);
 554         }
 555         return 1;
 556 }
 557
 558 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 559 {
 560         struct xt_mtdtor_param par;
 561
 562         par.net       = net;
 563         par.match     = m->u.kernel.match;
 564         par.matchinfo = m->data;
 565         par.family    = NFPROTO_IPV4;
 566         if (par.match->destroy != NULL)
 567                 par.match->destroy(&par);
 568         module_put(par.match->me);
 569 }
 570
 571 static int
 572 check_entry(const struct ipt_entry *e)
 573 {
 574         const struct xt_entry_target *t;
 575
 576         if (!ip_checkentry(&e->ip))
 577                 return -EINVAL;
 578
 579         if (e->target_offset + sizeof(struct xt_entry_target) >
 580             e->next_offset)
 581                 return -EINVAL;
 582
 583         t = ipt_get_target_c(e);
 584         if (e->target_offset + t->u.target_size > e->next_offset)
 585                 return -EINVAL;
 586
 587         return 0;
 588 }
 589
 590 static int
 591 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 592 {
 593         const struct ipt_ip *ip = par->entryinfo;
 594         int ret;
 595
 596         par->match     = m->u.kernel.match;
 597         par->matchinfo = m->data;
 598
 599         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 600               ip->proto, ip->invflags & IPT_INV_PROTO);
 601         if (ret < 0) {
 602                 duprintf("check failed for `%s'.\n", par->match->name);
 603                 return ret;
 604         }
 605         return 0;
 606 }
 607
 608 static int
 609 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 610 {
 611         struct xt_match *match;
 612         int ret;
 613
 614         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 615                                       m->u.user.revision);
 616         if (IS_ERR(match)) {
 617                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 618                 return PTR_ERR(match);
 619         }
 620         m->u.kernel.match = match;
 621
 622         ret = check_match(m, par);
 623         if (ret)
 624                 goto err;
 625
 626         return 0;
 627 err:
 628         module_put(m->u.kernel.match->me);
 629         return ret;
 630 }
 631
 632 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 633 {
 634         struct xt_entry_target *t = ipt_get_target(e);
 635         struct xt_tgchk_param par = {
 636                 .net       = net,
 637                 .table     = name,
 638                 .entryinfo = e,
 639                 .target    = t->u.kernel.target,
 640                 .targinfo  = t->data,
 641                 .hook_mask = e->comefrom,
 642                 .family    = NFPROTO_IPV4,
 643         };
 644         int ret;
 645
 646         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 647               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 648         if (ret < 0) {
 649                 duprintf("check failed for `%s'.\n",
 650                          t->u.kernel.target->name);
 651                 return ret;
 652         }
 653         return 0;
 654 }
 655
 656 static int
 657 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 658                  unsigned int size)
 659 {
 660         struct xt_entry_target *t;
 661         struct xt_target *target;
 662         int ret;
 663         unsigned int j;
 664         struct xt_mtchk_param mtpar;
 665         struct xt_entry_match *ematch;
 666
 667         e->counters.pcnt = xt_percpu_counter_alloc();
 668         if (IS_ERR_VALUE(e->counters.pcnt))
 669                 return -ENOMEM;
 670
 671         j = 0;
 672         mtpar.net       = net;
 673         mtpar.table     = name;
 674         mtpar.entryinfo = &e->ip;
 675         mtpar.hook_mask = e->comefrom;
 676         mtpar.family    = NFPROTO_IPV4;
 677         xt_ematch_foreach(ematch, e) {
 678                 ret = find_check_match(ematch, &mtpar);
 679                 if (ret != 0)
 680                         goto cleanup_matches;
 681                 ++j;
 682         }
 683
 684         t = ipt_get_target(e);
 685         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 686                                         t->u.user.revision);
 687         if (IS_ERR(target)) {
 688                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 689                 ret = PTR_ERR(target);
 690                 goto cleanup_matches;
 691         }
 692         t->u.kernel.target = target;
 693
 694         ret = check_target(e, net, name);
 695         if (ret)
 696                 goto err;
 697
 698         return 0;
 699  err:
 700         module_put(t->u.kernel.target->me);
 701  cleanup_matches:
 702         xt_ematch_foreach(ematch, e) {
 703                 if (j-- == 0)
 704                         break;
 705                 cleanup_match(ematch, net);
 706         }
 707
 708         xt_percpu_counter_free(e->counters.pcnt);
 709
 710         return ret;
 711 }
 712
 713 static bool check_underflow(const struct ipt_entry *e)
 714 {
 715         const struct xt_entry_target *t;
 716         unsigned int verdict;
 717
 718         if (!unconditional(&e->ip))
 719                 return false;
 720         t = ipt_get_target_c(e);
 721         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 722                 return false;
 723         verdict = ((struct xt_standard_target *)t)->verdict;
 724         verdict = -verdict - 1;
 725         return verdict == NF_DROP || verdict == NF_ACCEPT;
 726 }
 727
 728 static int
 729 check_entry_size_and_hooks(struct ipt_entry *e,
 730                            struct xt_table_info *newinfo,
 731                            const unsigned char *base,
 732                            const unsigned char *limit,
 733                            const unsigned int *hook_entries,
 734                            const unsigned int *underflows,
 735                            unsigned int valid_hooks)
 736 {
 737         unsigned int h;
 738         int err;
 739
 740         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 741             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 742                 duprintf("Bad offset %p\n", e);
 743                 return -EINVAL;
 744         }
 745
 746         if (e->next_offset
 747             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 748                 duprintf("checking: element %p size %u\n",
 749                          e, e->next_offset);
 750                 return -EINVAL;
 751         }
 752
 753         err = check_entry(e);
 754         if (err)
 755                 return err;
 756
 757         /* Check hooks & underflows */
 758         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 759                 if (!(valid_hooks & (1 << h)))
 760                         continue;
 761                 if ((unsigned char *)e - base == hook_entries[h])
 762                         newinfo->hook_entry[h] = hook_entries[h];
 763                 if ((unsigned char *)e - base == underflows[h]) {
 764                         if (!check_underflow(e)) {
 765                                 pr_err("Underflows must be unconditional and "
 766                                        "use the STANDARD target with "
 767                                        "ACCEPT/DROP\n");
 768                                 return -EINVAL;
 769                         }
 770                         newinfo->underflow[h] = underflows[h];
 771                 }
 772         }
 773
 774         /* Clear counters and comefrom */
 775         e->counters = ((struct xt_counters) { 0, 0 });
 776         e->comefrom = 0;
 777         return 0;
 778 }
 779
 780 static void
 781 cleanup_entry(struct ipt_entry *e, struct net *net)
 782 {
 783         struct xt_tgdtor_param par;
 784         struct xt_entry_target *t;
 785         struct xt_entry_match *ematch;
 786
 787         /* Cleanup all matches */
 788         xt_ematch_foreach(ematch, e)
 789                 cleanup_match(ematch, net);
 790         t = ipt_get_target(e);
 791
 792         par.net      = net;
 793         par.target   = t->u.kernel.target;
 794         par.targinfo = t->data;
 795         par.family   = NFPROTO_IPV4;
 796         if (par.target->destroy != NULL)
 797                 par.target->destroy(&par);
 798         module_put(par.target->me);
 799         xt_percpu_counter_free(e->counters.pcnt);
 800 }
 801
 802 /* Checks and translates the user-supplied table segment (held in
 803    newinfo) */
 804 static int
 805 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 806                 const struct ipt_replace *repl)
 807 {
 808         struct ipt_entry *iter;
 809         unsigned int i;
 810         int ret = 0;
 811
 812         newinfo->size = repl->size;
 813         newinfo->number = repl->num_entries;
 814
 815         /* Init all hooks to impossible value. */
 816         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 817                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 818                 newinfo->underflow[i] = 0xFFFFFFFF;
 819         }
 820
 821         duprintf("translate_table: size %u\n", newinfo->size);
 822         i = 0;
 823         /* Walk through entries, checking offsets. */
 824         xt_entry_foreach(iter, entry0, newinfo->size) {
 825                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 826                                                  entry0 + repl->size,
 827                                                  repl->hook_entry,
 828                                                  repl->underflow,
 829                                                  repl->valid_hooks);
 830                 if (ret != 0)
 831                         return ret;
 832                 ++i;
 833                 if (strcmp(ipt_get_target(iter)->u.user.name,
 834                     XT_ERROR_TARGET) == 0)
 835                         ++newinfo->stacksize;
 836         }
 837
 838         if (i != repl->num_entries) {
 839                 duprintf("translate_table: %u not %u entries\n",
 840                          i, repl->num_entries);
 841                 return -EINVAL;
 842         }
 843
 844         /* Check hooks all assigned */
 845         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 846                 /* Only hooks which are valid */
 847                 if (!(repl->valid_hooks & (1 << i)))
 848                         continue;
 849                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 850                         duprintf("Invalid hook entry %u %u\n",
 851                                  i, repl->hook_entry[i]);
 852                         return -EINVAL;
 853                 }
 854                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 855                         duprintf("Invalid underflow %u %u\n",
 856                                  i, repl->underflow[i]);
 857                         return -EINVAL;
 858                 }
 859         }
 860
 861         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 862                 return -ELOOP;
 863
 864         /* Finally, each sanity check must pass */
 865         i = 0;
 866         xt_entry_foreach(iter, entry0, newinfo->size) {
 867                 ret = find_check_entry(iter, net, repl->name, repl->size);
 868                 if (ret != 0)
 869                         break;
 870                 ++i;
 871         }
 872
 873         if (ret != 0) {
 874                 xt_entry_foreach(iter, entry0, newinfo->size) {
 875                         if (i-- == 0)
 876                                 break;
 877                         cleanup_entry(iter, net);
 878                 }
 879                 return ret;
 880         }
 881
 882         return ret;
 883 }
 884
 885 static void
 886 get_counters(const struct xt_table_info *t,
 887              struct xt_counters counters[])
 888 {
 889         struct ipt_entry *iter;
 890         unsigned int cpu;
 891         unsigned int i;
 892
 893         for_each_possible_cpu(cpu) {
 894                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 895
 896                 i = 0;
 897                 xt_entry_foreach(iter, t->entries, t->size) {
 898                         struct xt_counters *tmp;
 899                         u64 bcnt, pcnt;
 900                         unsigned int start;
 901
 902                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 903                         do {
 904                                 start = read_seqcount_begin(s);
 905                                 bcnt = tmp->bcnt;
 906                                 pcnt = tmp->pcnt;
 907                         } while (read_seqcount_retry(s, start));
 908
 909                         ADD_COUNTER(counters[i], bcnt, pcnt);
 910                         ++i; /* macro does multi eval of i */
 911                 }
 912         }
 913 }
 914
 915 static struct xt_counters *alloc_counters(const struct xt_table *table)
 916 {
 917         unsigned int countersize;
 918         struct xt_counters *counters;
 919         const struct xt_table_info *private = table->private;
 920
 921         /* We need atomic snapshot of counters: rest doesn't change
 922            (other than comefrom, which userspace doesn't care
 923            about). */
 924         countersize = sizeof(struct xt_counters) * private->number;
 925         counters = vzalloc(countersize);
 926
 927         if (counters == NULL)
 928                 return ERR_PTR(-ENOMEM);
 929
 930         get_counters(private, counters);
 931
 932         return counters;
 933 }
 934
 935 static int
 936 copy_entries_to_user(unsigned int total_size,
 937                      const struct xt_table *table,
 938                      void __user *userptr)
 939 {
 940         unsigned int off, num;
 941         const struct ipt_entry *e;
 942         struct xt_counters *counters;
 943         const struct xt_table_info *private = table->private;
 944         int ret = 0;
 945         const void *loc_cpu_entry;
 946
 947         counters = alloc_counters(table);
 948         if (IS_ERR(counters))
 949                 return PTR_ERR(counters);
 950
 951         loc_cpu_entry = private->entries;
 952         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 953                 ret = -EFAULT;
 954                 goto free_counters;
 955         }
 956
 957         /* FIXME: use iterator macros --RR */
 958         /* ... then go back and fix counters and names */
 959         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 960                 unsigned int i;
 961                 const struct xt_entry_match *m;
 962                 const struct xt_entry_target *t;
 963
 964                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 965                 if (copy_to_user(userptr + off
 966                                  + offsetof(struct ipt_entry, counters),
 967                                  &counters[num],
 968                                  sizeof(counters[num])) != 0) {
 969                         ret = -EFAULT;
 970                         goto free_counters;
 971                 }
 972
 973                 for (i = sizeof(struct ipt_entry);
 974                      i < e->target_offset;
 975                      i += m->u.match_size) {
 976                         m = (void *)e + i;
 977
 978                         if (copy_to_user(userptr + off + i
 979                                          + offsetof(struct xt_entry_match,
 980                                                     u.user.name),
 981                                          m->u.kernel.match->name,
 982                                          strlen(m->u.kernel.match->name)+1)
 983                             != 0) {
 984                                 ret = -EFAULT;
 985                                 goto free_counters;
 986                         }
 987                 }
 988
 989                 t = ipt_get_target_c(e);
 990                 if (copy_to_user(userptr + off + e->target_offset
 991                                  + offsetof(struct xt_entry_target,
 992                                             u.user.name),
 993                                  t->u.kernel.target->name,
 994                                  strlen(t->u.kernel.target->name)+1) != 0) {
 995                         ret = -EFAULT;
 996                         goto free_counters;
 997                 }
 998         }
 999
1000  free_counters:
1001         vfree(counters);
1002         return ret;
1003 }
1004
1005 #ifdef CONFIG_COMPAT
1006 static void compat_standard_from_user(void *dst, const void *src)
1007 {
1008         int v = *(compat_int_t *)src;
1009
1010         if (v > 0)
1011                 v += xt_compat_calc_jump(AF_INET, v);
1012         memcpy(dst, &v, sizeof(v));
1013 }
1014
1015 static int compat_standard_to_user(void __user *dst, const void *src)
1016 {
1017         compat_int_t cv = *(int *)src;
1018
1019         if (cv > 0)
1020                 cv -= xt_compat_calc_jump(AF_INET, cv);
1021         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1022 }
1023
1024 static int compat_calc_entry(const struct ipt_entry *e,
1025                              const struct xt_table_info *info,
1026                              const void *base, struct xt_table_info *newinfo)
1027 {
1028         const struct xt_entry_match *ematch;
1029         const struct xt_entry_target *t;
1030         unsigned int entry_offset;
1031         int off, i, ret;
1032
1033         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1034         entry_offset = (void *)e - base;
1035         xt_ematch_foreach(ematch, e)
1036                 off += xt_compat_match_offset(ematch->u.kernel.match);
1037         t = ipt_get_target_c(e);
1038         off += xt_compat_target_offset(t->u.kernel.target);
1039         newinfo->size -= off;
1040         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1041         if (ret)
1042                 return ret;
1043
1044         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1045                 if (info->hook_entry[i] &&
1046                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1047                         newinfo->hook_entry[i] -= off;
1048                 if (info->underflow[i] &&
1049                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1050                         newinfo->underflow[i] -= off;
1051         }
1052         return 0;
1053 }
1054
1055 static int compat_table_info(const struct xt_table_info *info,
1056                              struct xt_table_info *newinfo)
1057 {
1058         struct ipt_entry *iter;
1059         const void *loc_cpu_entry;
1060         int ret;
1061
1062         if (!newinfo || !info)
1063                 return -EINVAL;
1064
1065         /* we dont care about newinfo->entries */
1066         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1067         newinfo->initial_entries = 0;
1068         loc_cpu_entry = info->entries;
1069         xt_compat_init_offsets(AF_INET, info->number);
1070         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1071                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1072                 if (ret != 0)
1073                         return ret;
1074         }
1075         return 0;
1076 }
1077 #endif
1078
1079 static int get_info(struct net *net, void __user *user,
1080                     const int *len, int compat)
1081 {
1082         char name[XT_TABLE_MAXNAMELEN];
1083         struct xt_table *t;
1084         int ret;
1085
1086         if (*len != sizeof(struct ipt_getinfo)) {
1087                 duprintf("length %u != %zu\n", *len,
1088                          sizeof(struct ipt_getinfo));
1089                 return -EINVAL;
1090         }
1091
1092         if (copy_from_user(name, user, sizeof(name)) != 0)
1093                 return -EFAULT;
1094
1095         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1096 #ifdef CONFIG_COMPAT
1097         if (compat)
1098                 xt_compat_lock(AF_INET);
1099 #endif
1100         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1101                                     "iptable_%s", name);
1102         if (!IS_ERR_OR_NULL(t)) {
1103                 struct ipt_getinfo info;
1104                 const struct xt_table_info *private = t->private;
1105 #ifdef CONFIG_COMPAT
1106                 struct xt_table_info tmp;
1107
1108                 if (compat) {
1109                         ret = compat_table_info(private, &tmp);
1110                         xt_compat_flush_offsets(AF_INET);
1111                         private = &tmp;
1112                 }
1113 #endif
1114                 memset(&info, 0, sizeof(info));
1115                 info.valid_hooks = t->valid_hooks;
1116                 memcpy(info.hook_entry, private->hook_entry,
1117                        sizeof(info.hook_entry));
1118                 memcpy(info.underflow, private->underflow,
1119                        sizeof(info.underflow));
1120                 info.num_entries = private->number;
1121                 info.size = private->size;
1122                 strcpy(info.name, name);
1123
1124                 if (copy_to_user(user, &info, *len) != 0)
1125                         ret = -EFAULT;
1126                 else
1127                         ret = 0;
1128
1129                 xt_table_unlock(t);
1130                 module_put(t->me);
1131         } else
1132                 ret = t ? PTR_ERR(t) : -ENOENT;
1133 #ifdef CONFIG_COMPAT
1134         if (compat)
1135                 xt_compat_unlock(AF_INET);
1136 #endif
1137         return ret;
1138 }
1139
1140 static int
1141 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1142             const int *len)
1143 {
1144         int ret;
1145         struct ipt_get_entries get;
1146         struct xt_table *t;
1147
1148         if (*len < sizeof(get)) {
1149                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1150                 return -EINVAL;
1151         }
1152         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1153                 return -EFAULT;
1154         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1155                 duprintf("get_entries: %u != %zu\n",
1156                          *len, sizeof(get) + get.size);
1157                 return -EINVAL;
1158         }
1159
1160         t = xt_find_table_lock(net, AF_INET, get.name);
1161         if (!IS_ERR_OR_NULL(t)) {
1162                 const struct xt_table_info *private = t->private;
1163                 duprintf("t->private->number = %u\n", private->number);
1164                 if (get.size == private->size)
1165                         ret = copy_entries_to_user(private->size,
1166                                                    t, uptr->entrytable);
1167                 else {
1168                         duprintf("get_entries: I've got %u not %u!\n",
1169                                  private->size, get.size);
1170                         ret = -EAGAIN;
1171                 }
1172                 module_put(t->me);
1173                 xt_table_unlock(t);
1174         } else
1175                 ret = t ? PTR_ERR(t) : -ENOENT;
1176
1177         return ret;
1178 }
1179
1180 static int
1181 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1182              struct xt_table_info *newinfo, unsigned int num_counters,
1183              void __user *counters_ptr)
1184 {
1185         int ret;
1186         struct xt_table *t;
1187         struct xt_table_info *oldinfo;
1188         struct xt_counters *counters;
1189         struct ipt_entry *iter;
1190
1191         ret = 0;
1192         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1193         if (!counters) {
1194                 ret = -ENOMEM;
1195                 goto out;
1196         }
1197
1198         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1199                                     "iptable_%s", name);
1200         if (IS_ERR_OR_NULL(t)) {
1201                 ret = t ? PTR_ERR(t) : -ENOENT;
1202                 goto free_newinfo_counters_untrans;
1203         }
1204
1205         /* You lied! */
1206         if (valid_hooks != t->valid_hooks) {
1207                 duprintf("Valid hook crap: %08X vs %08X\n",
1208                          valid_hooks, t->valid_hooks);
1209                 ret = -EINVAL;
1210                 goto put_module;
1211         }
1212
1213         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1214         if (!oldinfo)
1215                 goto put_module;
1216
1217         /* Update module usage count based on number of rules */
1218         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1219                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1220         if ((oldinfo->number > oldinfo->initial_entries) ||
1221             (newinfo->number <= oldinfo->initial_entries))
1222                 module_put(t->me);
1223         if ((oldinfo->number > oldinfo->initial_entries) &&
1224             (newinfo->number <= oldinfo->initial_entries))
1225                 module_put(t->me);
1226
1227         /* Get the old counters, and synchronize with replace */
1228         get_counters(oldinfo, counters);
1229
1230         /* Decrease module usage counts and free resource */
1231         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1232                 cleanup_entry(iter, net);
1233
1234         xt_free_table_info(oldinfo);
1235         if (copy_to_user(counters_ptr, counters,
1236                          sizeof(struct xt_counters) * num_counters) != 0) {
1237                 /* Silent error, can't fail, new table is already in place */
1238                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1239         }
1240         vfree(counters);
1241         xt_table_unlock(t);
1242         return ret;
1243
1244  put_module:
1245         module_put(t->me);
1246         xt_table_unlock(t);
1247  free_newinfo_counters_untrans:
1248         vfree(counters);
1249  out:
1250         return ret;
1251 }
1252
1253 static int
1254 do_replace(struct net *net, const void __user *user, unsigned int len)
1255 {
1256         int ret;
1257         struct ipt_replace tmp;
1258         struct xt_table_info *newinfo;
1259         void *loc_cpu_entry;
1260         struct ipt_entry *iter;
1261
1262         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1263                 return -EFAULT;
1264
1265         /* overflow check */
1266         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1267                 return -ENOMEM;
1268         if (tmp.num_counters == 0)
1269                 return -EINVAL;
1270
1271         tmp.name[sizeof(tmp.name)-1] = 0;
1272
1273         newinfo = xt_alloc_table_info(tmp.size);
1274         if (!newinfo)
1275                 return -ENOMEM;
1276
1277         loc_cpu_entry = newinfo->entries;
1278         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1279                            tmp.size) != 0) {
1280                 ret = -EFAULT;
1281                 goto free_newinfo;
1282         }
1283
1284         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1285         if (ret != 0)
1286                 goto free_newinfo;
1287
1288         duprintf("Translated table\n");
1289
1290         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1291                            tmp.num_counters, tmp.counters);
1292         if (ret)
1293                 goto free_newinfo_untrans;
1294         return 0;
1295
1296  free_newinfo_untrans:
1297         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1298                 cleanup_entry(iter, net);
1299  free_newinfo:
1300         xt_free_table_info(newinfo);
1301         return ret;
1302 }
1303
1304 static int
1305 do_add_counters(struct net *net, const void __user *user,
1306                 unsigned int len, int compat)
1307 {
1308         unsigned int i;
1309         struct xt_counters_info tmp;
1310         struct xt_counters *paddc;
1311         unsigned int num_counters;
1312         const char *name;
1313         int size;
1314         void *ptmp;
1315         struct xt_table *t;
1316         const struct xt_table_info *private;
1317         int ret = 0;
1318         struct ipt_entry *iter;
1319         unsigned int addend;
1320 #ifdef CONFIG_COMPAT
1321         struct compat_xt_counters_info compat_tmp;
1322
1323         if (compat) {
1324                 ptmp = &compat_tmp;
1325                 size = sizeof(struct compat_xt_counters_info);
1326         } else
1327 #endif
1328         {
1329                 ptmp = &tmp;
1330                 size = sizeof(struct xt_counters_info);
1331         }
1332
1333         if (copy_from_user(ptmp, user, size) != 0)
1334                 return -EFAULT;
1335
1336 #ifdef CONFIG_COMPAT
1337         if (compat) {
1338                 num_counters = compat_tmp.num_counters;
1339                 name = compat_tmp.name;
1340         } else
1341 #endif
1342         {
1343                 num_counters = tmp.num_counters;
1344                 name = tmp.name;
1345         }
1346
1347         if (len != size + num_counters * sizeof(struct xt_counters))
1348                 return -EINVAL;
1349
1350         paddc = vmalloc(len - size);
1351         if (!paddc)
1352                 return -ENOMEM;
1353
1354         if (copy_from_user(paddc, user + size, len - size) != 0) {
1355                 ret = -EFAULT;
1356                 goto free;
1357         }
1358
1359         t = xt_find_table_lock(net, AF_INET, name);
1360         if (IS_ERR_OR_NULL(t)) {
1361                 ret = t ? PTR_ERR(t) : -ENOENT;
1362                 goto free;
1363         }
1364
1365         local_bh_disable();
1366         private = t->private;
1367         if (private->number != num_counters) {
1368                 ret = -EINVAL;
1369                 goto unlock_up_free;
1370         }
1371
1372         i = 0;
1373         addend = xt_write_recseq_begin();
1374         xt_entry_foreach(iter, private->entries, private->size) {
1375                 struct xt_counters *tmp;
1376
1377                 tmp = xt_get_this_cpu_counter(&iter->counters);
1378                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1379                 ++i;
1380         }
1381         xt_write_recseq_end(addend);
1382  unlock_up_free:
1383         local_bh_enable();
1384         xt_table_unlock(t);
1385         module_put(t->me);
1386  free:
1387         vfree(paddc);
1388
1389         return ret;
1390 }
1391
1392 #ifdef CONFIG_COMPAT
1393 struct compat_ipt_replace {
1394         char                    name[XT_TABLE_MAXNAMELEN];
1395         u32                     valid_hooks;
1396         u32                     num_entries;
1397         u32                     size;
1398         u32                     hook_entry[NF_INET_NUMHOOKS];
1399         u32                     underflow[NF_INET_NUMHOOKS];
1400         u32                     num_counters;
1401         compat_uptr_t           counters;       /* struct xt_counters * */
1402         struct compat_ipt_entry entries[0];
1403 };
1404
1405 static int
1406 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1407                           unsigned int *size, struct xt_counters *counters,
1408                           unsigned int i)
1409 {
1410         struct xt_entry_target *t;
1411         struct compat_ipt_entry __user *ce;
1412         u_int16_t target_offset, next_offset;
1413         compat_uint_t origsize;
1414         const struct xt_entry_match *ematch;
1415         int ret = 0;
1416
1417         origsize = *size;
1418         ce = (struct compat_ipt_entry __user *)*dstptr;
1419         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1420             copy_to_user(&ce->counters, &counters[i],
1421             sizeof(counters[i])) != 0)
1422                 return -EFAULT;
1423
1424         *dstptr += sizeof(struct compat_ipt_entry);
1425         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1426
1427         xt_ematch_foreach(ematch, e) {
1428                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1429                 if (ret != 0)
1430                         return ret;
1431         }
1432         target_offset = e->target_offset - (origsize - *size);
1433         t = ipt_get_target(e);
1434         ret = xt_compat_target_to_user(t, dstptr, size);
1435         if (ret)
1436                 return ret;
1437         next_offset = e->next_offset - (origsize - *size);
1438         if (put_user(target_offset, &ce->target_offset) != 0 ||
1439             put_user(next_offset, &ce->next_offset) != 0)
1440                 return -EFAULT;
1441         return 0;
1442 }
1443
1444 static int
1445 compat_find_calc_match(struct xt_entry_match *m,
1446                        const char *name,
1447                        const struct ipt_ip *ip,
1448                        int *size)
1449 {
1450         struct xt_match *match;
1451
1452         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1453                                       m->u.user.revision);
1454         if (IS_ERR(match)) {
1455                 duprintf("compat_check_calc_match: `%s' not found\n",
1456                          m->u.user.name);
1457                 return PTR_ERR(match);
1458         }
1459         m->u.kernel.match = match;
1460         *size += xt_compat_match_offset(match);
1461         return 0;
1462 }
1463
1464 static void compat_release_entry(struct compat_ipt_entry *e)
1465 {
1466         struct xt_entry_target *t;
1467         struct xt_entry_match *ematch;
1468
1469         /* Cleanup all matches */
1470         xt_ematch_foreach(ematch, e)
1471                 module_put(ematch->u.kernel.match->me);
1472         t = compat_ipt_get_target(e);
1473         module_put(t->u.kernel.target->me);
1474 }
1475
1476 static int
1477 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1478                                   struct xt_table_info *newinfo,
1479                                   unsigned int *size,
1480                                   const unsigned char *base,
1481                                   const unsigned char *limit,
1482                                   const unsigned int *hook_entries,
1483                                   const unsigned int *underflows,
1484                                   const char *name)
1485 {
1486         struct xt_entry_match *ematch;
1487         struct xt_entry_target *t;
1488         struct xt_target *target;
1489         unsigned int entry_offset;
1490         unsigned int j;
1491         int ret, off, h;
1492
1493         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1494         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1495             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1496                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1497                 return -EINVAL;
1498         }
1499
1500         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1501                              sizeof(struct compat_xt_entry_target)) {
1502                 duprintf("checking: element %p size %u\n",
1503                          e, e->next_offset);
1504                 return -EINVAL;
1505         }
1506
1507         /* For purposes of check_entry casting the compat entry is fine */
1508         ret = check_entry((struct ipt_entry *)e);
1509         if (ret)
1510                 return ret;
1511
1512         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1513         entry_offset = (void *)e - (void *)base;
1514         j = 0;
1515         xt_ematch_foreach(ematch, e) {
1516                 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1517                 if (ret != 0)
1518                         goto release_matches;
1519                 ++j;
1520         }
1521
1522         t = compat_ipt_get_target(e);
1523         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1524                                         t->u.user.revision);
1525         if (IS_ERR(target)) {
1526                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1527                          t->u.user.name);
1528                 ret = PTR_ERR(target);
1529                 goto release_matches;
1530         }
1531         t->u.kernel.target = target;
1532
1533         off += xt_compat_target_offset(target);
1534         *size += off;
1535         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1536         if (ret)
1537                 goto out;
1538
1539         /* Check hooks & underflows */
1540         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1541                 if ((unsigned char *)e - base == hook_entries[h])
1542                         newinfo->hook_entry[h] = hook_entries[h];
1543                 if ((unsigned char *)e - base == underflows[h])
1544                         newinfo->underflow[h] = underflows[h];
1545         }
1546
1547         /* Clear counters and comefrom */
1548         memset(&e->counters, 0, sizeof(e->counters));
1549         e->comefrom = 0;
1550         return 0;
1551
1552 out:
1553         module_put(t->u.kernel.target->me);
1554 release_matches:
1555         xt_ematch_foreach(ematch, e) {
1556                 if (j-- == 0)
1557                         break;
1558                 module_put(ematch->u.kernel.match->me);
1559         }
1560         return ret;
1561 }
1562
1563 static int
1564 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1565                             unsigned int *size, const char *name,
1566                             struct xt_table_info *newinfo, unsigned char *base)
1567 {
1568         struct xt_entry_target *t;
1569         struct xt_target *target;
1570         struct ipt_entry *de;
1571         unsigned int origsize;
1572         int ret, h;
1573         struct xt_entry_match *ematch;
1574
1575         ret = 0;
1576         origsize = *size;
1577         de = (struct ipt_entry *)*dstptr;
1578         memcpy(de, e, sizeof(struct ipt_entry));
1579         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1580
1581         *dstptr += sizeof(struct ipt_entry);
1582         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1583
1584         xt_ematch_foreach(ematch, e) {
1585                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1586                 if (ret != 0)
1587                         return ret;
1588         }
1589         de->target_offset = e->target_offset - (origsize - *size);
1590         t = compat_ipt_get_target(e);
1591         target = t->u.kernel.target;
1592         xt_compat_target_from_user(t, dstptr, size);
1593
1594         de->next_offset = e->next_offset - (origsize - *size);
1595         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1596                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1597                         newinfo->hook_entry[h] -= origsize - *size;
1598                 if ((unsigned char *)de - base < newinfo->underflow[h])
1599                         newinfo->underflow[h] -= origsize - *size;
1600         }
1601         return ret;
1602 }
1603
1604 static int
1605 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1606 {
1607         struct xt_entry_match *ematch;
1608         struct xt_mtchk_param mtpar;
1609         unsigned int j;
1610         int ret = 0;
1611
1612         e->counters.pcnt = xt_percpu_counter_alloc();
1613         if (IS_ERR_VALUE(e->counters.pcnt))
1614                 return -ENOMEM;
1615
1616         j = 0;
1617         mtpar.net       = net;
1618         mtpar.table     = name;
1619         mtpar.entryinfo = &e->ip;
1620         mtpar.hook_mask = e->comefrom;
1621         mtpar.family    = NFPROTO_IPV4;
1622         xt_ematch_foreach(ematch, e) {
1623                 ret = check_match(ematch, &mtpar);
1624                 if (ret != 0)
1625                         goto cleanup_matches;
1626                 ++j;
1627         }
1628
1629         ret = check_target(e, net, name);
1630         if (ret)
1631                 goto cleanup_matches;
1632         return 0;
1633
1634  cleanup_matches:
1635         xt_ematch_foreach(ematch, e) {
1636                 if (j-- == 0)
1637                         break;
1638                 cleanup_match(ematch, net);
1639         }
1640
1641         xt_percpu_counter_free(e->counters.pcnt);
1642
1643         return ret;
1644 }
1645
1646 static int
1647 translate_compat_table(struct net *net,
1648                        const char *name,
1649                        unsigned int valid_hooks,
1650                        struct xt_table_info **pinfo,
1651                        void **pentry0,
1652                        unsigned int total_size,
1653                        unsigned int number,
1654                        unsigned int *hook_entries,
1655                        unsigned int *underflows)
1656 {
1657         unsigned int i, j;
1658         struct xt_table_info *newinfo, *info;
1659         void *pos, *entry0, *entry1;
1660         struct compat_ipt_entry *iter0;
1661         struct ipt_entry *iter1;
1662         unsigned int size;
1663         int ret;
1664
1665         info = *pinfo;
1666         entry0 = *pentry0;
1667         size = total_size;
1668         info->number = number;
1669
1670         /* Init all hooks to impossible value. */
1671         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1672                 info->hook_entry[i] = 0xFFFFFFFF;
1673                 info->underflow[i] = 0xFFFFFFFF;
1674         }
1675
1676         duprintf("translate_compat_table: size %u\n", info->size);
1677         j = 0;
1678         xt_compat_lock(AF_INET);
1679         xt_compat_init_offsets(AF_INET, number);
1680         /* Walk through entries, checking offsets. */
1681         xt_entry_foreach(iter0, entry0, total_size) {
1682                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1683                                                         entry0,
1684                                                         entry0 + total_size,
1685                                                         hook_entries,
1686                                                         underflows,
1687                                                         name);
1688                 if (ret != 0)
1689                         goto out_unlock;
1690                 ++j;
1691         }
1692
1693         ret = -EINVAL;
1694         if (j != number) {
1695                 duprintf("translate_compat_table: %u not %u entries\n",
1696                          j, number);
1697                 goto out_unlock;
1698         }
1699
1700         /* Check hooks all assigned */
1701         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1702                 /* Only hooks which are valid */
1703                 if (!(valid_hooks & (1 << i)))
1704                         continue;
1705                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1706                         duprintf("Invalid hook entry %u %u\n",
1707                                  i, hook_entries[i]);
1708                         goto out_unlock;
1709                 }
1710                 if (info->underflow[i] == 0xFFFFFFFF) {
1711                         duprintf("Invalid underflow %u %u\n",
1712                                  i, underflows[i]);
1713                         goto out_unlock;
1714                 }
1715         }
1716
1717         ret = -ENOMEM;
1718         newinfo = xt_alloc_table_info(size);
1719         if (!newinfo)
1720                 goto out_unlock;
1721
1722         newinfo->number = number;
1723         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1724                 newinfo->hook_entry[i] = info->hook_entry[i];
1725                 newinfo->underflow[i] = info->underflow[i];
1726         }
1727         entry1 = newinfo->entries;
1728         pos = entry1;
1729         size = total_size;
1730         xt_entry_foreach(iter0, entry0, total_size) {
1731                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1732                                                   name, newinfo, entry1);
1733                 if (ret != 0)
1734                         break;
1735         }
1736         xt_compat_flush_offsets(AF_INET);
1737         xt_compat_unlock(AF_INET);
1738         if (ret)
1739                 goto free_newinfo;
1740
1741         ret = -ELOOP;
1742         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1743                 goto free_newinfo;
1744
1745         i = 0;
1746         xt_entry_foreach(iter1, entry1, newinfo->size) {
1747                 ret = compat_check_entry(iter1, net, name);
1748                 if (ret != 0)
1749                         break;
1750                 ++i;
1751                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1752                     XT_ERROR_TARGET) == 0)
1753                         ++newinfo->stacksize;
1754         }
1755         if (ret) {
1756                 /*
1757                  * The first i matches need cleanup_entry (calls ->destroy)
1758                  * because they had called ->check already. The other j-i
1759                  * entries need only release.
1760                  */
1761                 int skip = i;
1762                 j -= i;
1763                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1764                         if (skip-- > 0)
1765                                 continue;
1766                         if (j-- == 0)
1767                                 break;
1768                         compat_release_entry(iter0);
1769                 }
1770                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1771                         if (i-- == 0)
1772                                 break;
1773                         cleanup_entry(iter1, net);
1774                 }
1775                 xt_free_table_info(newinfo);
1776                 return ret;
1777         }
1778
1779         *pinfo = newinfo;
1780         *pentry0 = entry1;
1781         xt_free_table_info(info);
1782         return 0;
1783
1784 free_newinfo:
1785         xt_free_table_info(newinfo);
1786 out:
1787         xt_entry_foreach(iter0, entry0, total_size) {
1788                 if (j-- == 0)
1789                         break;
1790                 compat_release_entry(iter0);
1791         }
1792         return ret;
1793 out_unlock:
1794         xt_compat_flush_offsets(AF_INET);
1795         xt_compat_unlock(AF_INET);
1796         goto out;
1797 }
1798
1799 static int
1800 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1801 {
1802         int ret;
1803         struct compat_ipt_replace tmp;
1804         struct xt_table_info *newinfo;
1805         void *loc_cpu_entry;
1806         struct ipt_entry *iter;
1807
1808         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1809                 return -EFAULT;
1810
1811         /* overflow check */
1812         if (tmp.size >= INT_MAX / num_possible_cpus())
1813                 return -ENOMEM;
1814         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1815                 return -ENOMEM;
1816         if (tmp.num_counters == 0)
1817                 return -EINVAL;
1818
1819         tmp.name[sizeof(tmp.name)-1] = 0;
1820
1821         newinfo = xt_alloc_table_info(tmp.size);
1822         if (!newinfo)
1823                 return -ENOMEM;
1824
1825         loc_cpu_entry = newinfo->entries;
1826         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1827                            tmp.size) != 0) {
1828                 ret = -EFAULT;
1829                 goto free_newinfo;
1830         }
1831
1832         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1833                                      &newinfo, &loc_cpu_entry, tmp.size,
1834                                      tmp.num_entries, tmp.hook_entry,
1835                                      tmp.underflow);
1836         if (ret != 0)
1837                 goto free_newinfo;
1838
1839         duprintf("compat_do_replace: Translated table\n");
1840
1841         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1842                            tmp.num_counters, compat_ptr(tmp.counters));
1843         if (ret)
1844                 goto free_newinfo_untrans;
1845         return 0;
1846
1847  free_newinfo_untrans:
1848         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1849                 cleanup_entry(iter, net);
1850  free_newinfo:
1851         xt_free_table_info(newinfo);
1852         return ret;
1853 }
1854
1855 static int
1856 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1857                       unsigned int len)
1858 {
1859         int ret;
1860
1861         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1862                 return -EPERM;
1863
1864         switch (cmd) {
1865         case IPT_SO_SET_REPLACE:
1866                 ret = compat_do_replace(sock_net(sk), user, len);
1867                 break;
1868
1869         case IPT_SO_SET_ADD_COUNTERS:
1870                 ret = do_add_counters(sock_net(sk), user, len, 1);
1871                 break;
1872
1873         default:
1874                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1875                 ret = -EINVAL;
1876         }
1877
1878         return ret;
1879 }
1880
1881 struct compat_ipt_get_entries {
1882         char name[XT_TABLE_MAXNAMELEN];
1883         compat_uint_t size;
1884         struct compat_ipt_entry entrytable[0];
1885 };
1886
1887 static int
1888 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1889                             void __user *userptr)
1890 {
1891         struct xt_counters *counters;
1892         const struct xt_table_info *private = table->private;
1893         void __user *pos;
1894         unsigned int size;
1895         int ret = 0;
1896         unsigned int i = 0;
1897         struct ipt_entry *iter;
1898
1899         counters = alloc_counters(table);
1900         if (IS_ERR(counters))
1901                 return PTR_ERR(counters);
1902
1903         pos = userptr;
1904         size = total_size;
1905         xt_entry_foreach(iter, private->entries, total_size) {
1906                 ret = compat_copy_entry_to_user(iter, &pos,
1907                                                 &size, counters, i++);
1908                 if (ret != 0)
1909                         break;
1910         }
1911
1912         vfree(counters);
1913         return ret;
1914 }
1915
1916 static int
1917 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1918                    int *len)
1919 {
1920         int ret;
1921         struct compat_ipt_get_entries get;
1922         struct xt_table *t;
1923
1924         if (*len < sizeof(get)) {
1925                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1926                 return -EINVAL;
1927         }
1928
1929         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1930                 return -EFAULT;
1931
1932         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1933                 duprintf("compat_get_entries: %u != %zu\n",
1934                          *len, sizeof(get) + get.size);
1935                 return -EINVAL;
1936         }
1937
1938         xt_compat_lock(AF_INET);
1939         t = xt_find_table_lock(net, AF_INET, get.name);
1940         if (!IS_ERR_OR_NULL(t)) {
1941                 const struct xt_table_info *private = t->private;
1942                 struct xt_table_info info;
1943                 duprintf("t->private->number = %u\n", private->number);
1944                 ret = compat_table_info(private, &info);
1945                 if (!ret && get.size == info.size) {
1946                         ret = compat_copy_entries_to_user(private->size,
1947                                                           t, uptr->entrytable);
1948                 } else if (!ret) {
1949                         duprintf("compat_get_entries: I've got %u not %u!\n",
1950                                  private->size, get.size);
1951                         ret = -EAGAIN;
1952                 }
1953                 xt_compat_flush_offsets(AF_INET);
1954                 module_put(t->me);
1955                 xt_table_unlock(t);
1956         } else
1957                 ret = t ? PTR_ERR(t) : -ENOENT;
1958
1959         xt_compat_unlock(AF_INET);
1960         return ret;
1961 }
1962
1963 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1964
1965 static int
1966 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1967 {
1968         int ret;
1969
1970         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1971                 return -EPERM;
1972
1973         switch (cmd) {
1974         case IPT_SO_GET_INFO:
1975                 ret = get_info(sock_net(sk), user, len, 1);
1976                 break;
1977         case IPT_SO_GET_ENTRIES:
1978                 ret = compat_get_entries(sock_net(sk), user, len);
1979                 break;
1980         default:
1981                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1982         }
1983         return ret;
1984 }
1985 #endif
1986
1987 static int
1988 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1989 {
1990         int ret;
1991
1992         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1993                 return -EPERM;
1994
1995         switch (cmd) {
1996         case IPT_SO_SET_REPLACE:
1997                 ret = do_replace(sock_net(sk), user, len);
1998                 break;
1999
2000         case IPT_SO_SET_ADD_COUNTERS:
2001                 ret = do_add_counters(sock_net(sk), user, len, 0);
2002                 break;
2003
2004         default:
2005                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2006                 ret = -EINVAL;
2007         }
2008
2009         return ret;
2010 }
2011
2012 static int
2013 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2014 {
2015         int ret;
2016
2017         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2018                 return -EPERM;
2019
2020         switch (cmd) {
2021         case IPT_SO_GET_INFO:
2022                 ret = get_info(sock_net(sk), user, len, 0);
2023                 break;
2024
2025         case IPT_SO_GET_ENTRIES:
2026                 ret = get_entries(sock_net(sk), user, len);
2027                 break;
2028
2029         case IPT_SO_GET_REVISION_MATCH:
2030         case IPT_SO_GET_REVISION_TARGET: {
2031                 struct xt_get_revision rev;
2032                 int target;
2033
2034                 if (*len != sizeof(rev)) {
2035                         ret = -EINVAL;
2036                         break;
2037                 }
2038                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2039                         ret = -EFAULT;
2040                         break;
2041                 }
2042                 rev.name[sizeof(rev.name)-1] = 0;
2043
2044                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2045                         target = 1;
2046                 else
2047                         target = 0;
2048
2049                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2050                                                          rev.revision,
2051                                                          target, &ret),
2052                                         "ipt_%s", rev.name);
2053                 break;
2054         }
2055
2056         default:
2057                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2058                 ret = -EINVAL;
2059         }
2060
2061         return ret;
2062 }
2063
2064 struct xt_table *ipt_register_table(struct net *net,
2065                                     const struct xt_table *table,
2066                                     const struct ipt_replace *repl)
2067 {
2068         int ret;
2069         struct xt_table_info *newinfo;
2070         struct xt_table_info bootstrap = {0};
2071         void *loc_cpu_entry;
2072         struct xt_table *new_table;
2073
2074         newinfo = xt_alloc_table_info(repl->size);
2075         if (!newinfo) {
2076                 ret = -ENOMEM;
2077                 goto out;
2078         }
2079
2080         loc_cpu_entry = newinfo->entries;
2081         memcpy(loc_cpu_entry, repl->entries, repl->size);
2082
2083         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2084         if (ret != 0)
2085                 goto out_free;
2086
2087         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2088         if (IS_ERR(new_table)) {
2089                 ret = PTR_ERR(new_table);
2090                 goto out_free;
2091         }
2092
2093         return new_table;
2094
2095 out_free:
2096         xt_free_table_info(newinfo);
2097 out:
2098         return ERR_PTR(ret);
2099 }
2100
2101 void ipt_unregister_table(struct net *net, struct xt_table *table)
2102 {
2103         struct xt_table_info *private;
2104         void *loc_cpu_entry;
2105         struct module *table_owner = table->me;
2106         struct ipt_entry *iter;
2107
2108         private = xt_unregister_table(table);
2109
2110         /* Decrease module usage counts and free resources */
2111         loc_cpu_entry = private->entries;
2112         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2113                 cleanup_entry(iter, net);
2114         if (private->number > private->initial_entries)
2115                 module_put(table_owner);
2116         xt_free_table_info(private);
2117 }
2118
2119 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2120 static inline bool
2121 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2122                      u_int8_t type, u_int8_t code,
2123                      bool invert)
2124 {
2125         return ((test_type == 0xFF) ||
2126                 (type == test_type && code >= min_code && code <= max_code))
2127                 ^ invert;
2128 }
2129
2130 static bool
2131 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2132 {
2133         const struct icmphdr *ic;
2134         struct icmphdr _icmph;
2135         const struct ipt_icmp *icmpinfo = par->matchinfo;
2136
2137         /* Must not be a fragment. */
2138         if (par->fragoff != 0)
2139                 return false;
2140
2141         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2142         if (ic == NULL) {
2143                 /* We've been asked to examine this packet, and we
2144                  * can't.  Hence, no choice but to drop.
2145                  */
2146                 duprintf("Dropping evil ICMP tinygram.\n");
2147                 par->hotdrop = true;
2148                 return false;
2149         }
2150
2151         return icmp_type_code_match(icmpinfo->type,
2152                                     icmpinfo->code[0],
2153                                     icmpinfo->code[1],
2154                                     ic->type, ic->code,
2155                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2156 }
2157
2158 static int icmp_checkentry(const struct xt_mtchk_param *par)
2159 {
2160         const struct ipt_icmp *icmpinfo = par->matchinfo;
2161
2162         /* Must specify no unknown invflags */
2163         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2164 }
2165
2166 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2167         {
2168                 .name             = XT_STANDARD_TARGET,
2169                 .targetsize       = sizeof(int),
2170                 .family           = NFPROTO_IPV4,
2171 #ifdef CONFIG_COMPAT
2172                 .compatsize       = sizeof(compat_int_t),
2173                 .compat_from_user = compat_standard_from_user,
2174                 .compat_to_user   = compat_standard_to_user,
2175 #endif
2176         },
2177         {
2178                 .name             = XT_ERROR_TARGET,
2179                 .target           = ipt_error,
2180                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2181                 .family           = NFPROTO_IPV4,
2182         },
2183 };
2184
2185 static struct nf_sockopt_ops ipt_sockopts = {
2186         .pf             = PF_INET,
2187         .set_optmin     = IPT_BASE_CTL,
2188         .set_optmax     = IPT_SO_SET_MAX+1,
2189         .set            = do_ipt_set_ctl,
2190 #ifdef CONFIG_COMPAT
2191         .compat_set     = compat_do_ipt_set_ctl,
2192 #endif
2193         .get_optmin     = IPT_BASE_CTL,
2194         .get_optmax     = IPT_SO_GET_MAX+1,
2195         .get            = do_ipt_get_ctl,
2196 #ifdef CONFIG_COMPAT
2197         .compat_get     = compat_do_ipt_get_ctl,
2198 #endif
2199         .owner          = THIS_MODULE,
2200 };
2201
2202 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2203         {
2204                 .name       = "icmp",
2205                 .match      = icmp_match,
2206                 .matchsize  = sizeof(struct ipt_icmp),
2207                 .checkentry = icmp_checkentry,
2208                 .proto      = IPPROTO_ICMP,
2209                 .family     = NFPROTO_IPV4,
2210         },
2211 };
2212
2213 static int __net_init ip_tables_net_init(struct net *net)
2214 {
2215         return xt_proto_init(net, NFPROTO_IPV4);
2216 }
2217
2218 static void __net_exit ip_tables_net_exit(struct net *net)
2219 {
2220         xt_proto_fini(net, NFPROTO_IPV4);
2221 }
2222
2223 static struct pernet_operations ip_tables_net_ops = {
2224         .init = ip_tables_net_init,
2225         .exit = ip_tables_net_exit,
2226 };
2227
2228 static int __init ip_tables_init(void)
2229 {
2230         int ret;
2231
2232         ret = register_pernet_subsys(&ip_tables_net_ops);
2233         if (ret < 0)
2234                 goto err1;
2235
2236         /* No one else will be downing sem now, so we won't sleep */
2237         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2238         if (ret < 0)
2239                 goto err2;
2240         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2241         if (ret < 0)
2242                 goto err4;
2243
2244         /* Register setsockopt */
2245         ret = nf_register_sockopt(&ipt_sockopts);
2246         if (ret < 0)
2247                 goto err5;
2248
2249         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2250         return 0;
2251
2252 err5:
2253         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2254 err4:
2255         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2256 err2:
2257         unregister_pernet_subsys(&ip_tables_net_ops);
2258 err1:
2259         return ret;
2260 }
2261
2262 static void __exit ip_tables_fini(void)
2263 {
2264         nf_unregister_sockopt(&ipt_sockopts);
2265
2266         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2267         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2268         unregister_pernet_subsys(&ip_tables_net_ops);
2269 }
2270
2271 EXPORT_SYMBOL(ipt_register_table);
2272 EXPORT_SYMBOL(ipt_unregister_table);
2273 EXPORT_SYMBOL(ipt_do_table);
2274 module_init(ip_tables_init);
2275 module_exit(ip_tables_fini);