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  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  */
  11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  12 #include <linux/cache.h>
  13 #include <linux/capability.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/kmod.h>
  16 #include <linux/vmalloc.h>
  17 #include <linux/netdevice.h>
  18 #include <linux/module.h>
  19 #include <linux/icmp.h>
  20 #include <net/ip.h>
  21 #include <net/compat.h>
  22 #include <asm/uaccess.h>
  23 #include <linux/mutex.h>
  24 #include <linux/proc_fs.h>
  25 #include <linux/err.h>
  26 #include <linux/cpumask.h>
  27
  28 #include <linux/netfilter/x_tables.h>
  29 #include <linux/netfilter_ipv4/ip_tables.h>
  30 #include <net/netfilter/nf_log.h>
  31 #include "../../netfilter/xt_repldata.h"
  32
  33 MODULE_LICENSE("GPL");
  34 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  35 MODULE_DESCRIPTION("IPv4 packet filter");
  36
  37 /*#define DEBUG_IP_FIREWALL*/
  38 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  39 /*#define DEBUG_IP_FIREWALL_USER*/
  40
  41 #ifdef DEBUG_IP_FIREWALL
  42 #define dprintf(format, args...) pr_info(format , ## args)
  43 #else
  44 #define dprintf(format, args...)
  45 #endif
  46
  47 #ifdef DEBUG_IP_FIREWALL_USER
  48 #define duprintf(format, args...) pr_info(format , ## args)
  49 #else
  50 #define duprintf(format, args...)
  51 #endif
  52
  53 #ifdef CONFIG_NETFILTER_DEBUG
  54 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  55 #else
  56 #define IP_NF_ASSERT(x)
  57 #endif
  58
  59 #if 0
  60 /* All the better to debug you with... */
  61 #define static
  62 #define inline
  63 #endif
  64
  65 void *ipt_alloc_initial_table(const struct xt_table *info)
  66 {
  67         return xt_alloc_initial_table(ipt, IPT);
  68 }
  69 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  70
  71 /* Returns whether matches rule or not. */
  72 /* Performance critical - called for every packet */
  73 static inline bool
  74 ip_packet_match(const struct iphdr *ip,
  75                 const char *indev,
  76                 const char *outdev,
  77                 const struct ipt_ip *ipinfo,
  78                 int isfrag)
  79 {
  80         unsigned long ret;
  81
  82 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  83
  84         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  85                   IPT_INV_SRCIP) ||
  86             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  87                   IPT_INV_DSTIP)) {
  88                 dprintf("Source or dest mismatch.\n");
  89
  90                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  91                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  92                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  93                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  94                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  95                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  96                 return false;
  97         }
  98
  99         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 100
 101         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 102                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 103                         indev, ipinfo->iniface,
 104                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
 105                 return false;
 106         }
 107
 108         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 109
 110         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 111                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 112                         outdev, ipinfo->outiface,
 113                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
 114                 return false;
 115         }
 116
 117         /* Check specific protocol */
 118         if (ipinfo->proto &&
 119             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 120                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 121                         ip->protocol, ipinfo->proto,
 122                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
 123                 return false;
 124         }
 125
 126         /* If we have a fragment rule but the packet is not a fragment
 127          * then we return zero */
 128         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 129                 dprintf("Fragment rule but not fragment.%s\n",
 130                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 131                 return false;
 132         }
 133
 134         return true;
 135 }
 136
 137 static bool
 138 ip_checkentry(const struct ipt_ip *ip)
 139 {
 140         if (ip->flags & ~IPT_F_MASK) {
 141                 duprintf("Unknown flag bits set: %08X\n",
 142                          ip->flags & ~IPT_F_MASK);
 143                 return false;
 144         }
 145         if (ip->invflags & ~IPT_INV_MASK) {
 146                 duprintf("Unknown invflag bits set: %08X\n",
 147                          ip->invflags & ~IPT_INV_MASK);
 148                 return false;
 149         }
 150         return true;
 151 }
 152
 153 static unsigned int
 154 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 155 {
 156         if (net_ratelimit())
 157                 pr_info("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 defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
 187     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
 188 static const char *const hooknames[] = {
 189         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 190         [NF_INET_LOCAL_IN]              = "INPUT",
 191         [NF_INET_FORWARD]               = "FORWARD",
 192         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 193         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 194 };
 195
 196 enum nf_ip_trace_comments {
 197         NF_IP_TRACE_COMMENT_RULE,
 198         NF_IP_TRACE_COMMENT_RETURN,
 199         NF_IP_TRACE_COMMENT_POLICY,
 200 };
 201
 202 static const char *const comments[] = {
 203         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 204         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 205         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 206 };
 207
 208 static struct nf_loginfo trace_loginfo = {
 209         .type = NF_LOG_TYPE_LOG,
 210         .u = {
 211                 .log = {
 212                         .level = 4,
 213                         .logflags = NF_LOG_MASK,
 214                 },
 215         },
 216 };
 217
 218 /* Mildly perf critical (only if packet tracing is on) */
 219 static inline int
 220 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 221                       const char *hookname, const char **chainname,
 222                       const char **comment, unsigned int *rulenum)
 223 {
 224         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 225
 226         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 227                 /* Head of user chain: ERROR target with chainname */
 228                 *chainname = t->target.data;
 229                 (*rulenum) = 0;
 230         } else if (s == e) {
 231                 (*rulenum)++;
 232
 233                 if (s->target_offset == sizeof(struct ipt_entry) &&
 234                     strcmp(t->target.u.kernel.target->name,
 235                            XT_STANDARD_TARGET) == 0 &&
 236                    t->verdict < 0 &&
 237                    unconditional(&s->ip)) {
 238                         /* Tail of chains: STANDARD target (return/policy) */
 239                         *comment = *chainname == hookname
 240                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 241                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 242                 }
 243                 return 1;
 244         } else
 245                 (*rulenum)++;
 246
 247         return 0;
 248 }
 249
 250 static void trace_packet(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 void *table_base;
 259         const struct ipt_entry *root;
 260         const char *hookname, *chainname, *comment;
 261         const struct ipt_entry *iter;
 262         unsigned int rulenum = 0;
 263
 264         table_base = private->entries[smp_processor_id()];
 265         root = get_entry(table_base, private->hook_entry[hook]);
 266
 267         hookname = chainname = hooknames[hook];
 268         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 269
 270         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 271                 if (get_chainname_rulenum(iter, e, hookname,
 272                     &chainname, &comment, &rulenum) != 0)
 273                         break;
 274
 275         nf_log_packet(AF_INET, hook, skb, in, out, &trace_loginfo,
 276                       "TRACE: %s:%s:%s:%u ",
 277                       tablename, chainname, comment, rulenum);
 278 }
 279 #endif
 280
 281 static inline __pure
 282 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 283 {
 284         return (void *)entry + entry->next_offset;
 285 }
 286
 287 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 288 unsigned int
 289 ipt_do_table(struct sk_buff *skb,
 290              unsigned int hook,
 291              const struct net_device *in,
 292              const struct net_device *out,
 293              struct xt_table *table)
 294 {
 295         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 296         const struct iphdr *ip;
 297         /* Initializing verdict to NF_DROP keeps gcc happy. */
 298         unsigned int verdict = NF_DROP;
 299         const char *indev, *outdev;
 300         const void *table_base;
 301         struct ipt_entry *e, **jumpstack;
 302         unsigned int *stackptr, origptr, cpu;
 303         const struct xt_table_info *private;
 304         struct xt_action_param acpar;
 305         unsigned int addend;
 306
 307         /* Initialization */
 308         ip = ip_hdr(skb);
 309         indev = in ? in->name : nulldevname;
 310         outdev = out ? out->name : nulldevname;
 311         /* We handle fragments by dealing with the first fragment as
 312          * if it was a normal packet.  All other fragments are treated
 313          * normally, except that they will NEVER match rules that ask
 314          * things we don't know, ie. tcp syn flag or ports).  If the
 315          * rule is also a fragment-specific rule, non-fragments won't
 316          * match it. */
 317         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 318         acpar.thoff   = ip_hdrlen(skb);
 319         acpar.hotdrop = false;
 320         acpar.in      = in;
 321         acpar.out     = out;
 322         acpar.family  = NFPROTO_IPV4;
 323         acpar.hooknum = hook;
 324
 325         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 326         local_bh_disable();
 327         addend = xt_write_recseq_begin();
 328         private = table->private;
 329         cpu        = smp_processor_id();
 330         table_base = private->entries[cpu];
 331         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 332         stackptr   = per_cpu_ptr(private->stackptr, cpu);
 333         origptr    = *stackptr;
 334
 335         e = get_entry(table_base, private->hook_entry[hook]);
 336
 337         pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
 338                  table->name, hook, origptr,
 339                  get_entry(table_base, private->underflow[hook]));
 340
 341         do {
 342                 const struct xt_entry_target *t;
 343                 const struct xt_entry_match *ematch;
 344
 345                 IP_NF_ASSERT(e);
 346                 if (!ip_packet_match(ip, indev, outdev,
 347                     &e->ip, acpar.fragoff)) {
 348  no_match:
 349                         e = ipt_next_entry(e);
 350                         continue;
 351                 }
 352
 353                 xt_ematch_foreach(ematch, e) {
 354                         acpar.match     = ematch->u.kernel.match;
 355                         acpar.matchinfo = ematch->data;
 356                         if (!acpar.match->match(skb, &acpar))
 357                                 goto no_match;
 358                 }
 359
 360                 ADD_COUNTER(e->counters, skb->len, 1);
 361
 362                 t = ipt_get_target(e);
 363                 IP_NF_ASSERT(t->u.kernel.target);
 364
 365 #if defined(CONFIG_NETFILTER_XT_TARGET_TRACE) || \
 366     defined(CONFIG_NETFILTER_XT_TARGET_TRACE_MODULE)
 367                 /* The packet is traced: log it */
 368                 if (unlikely(skb->nf_trace))
 369                         trace_packet(skb, hook, in, out,
 370                                      table->name, private, e);
 371 #endif
 372                 /* Standard target? */
 373                 if (!t->u.kernel.target->target) {
 374                         int v;
 375
 376                         v = ((struct xt_standard_target *)t)->verdict;
 377                         if (v < 0) {
 378                                 /* Pop from stack? */
 379                                 if (v != XT_RETURN) {
 380                                         verdict = (unsigned)(-v) - 1;
 381                                         break;
 382                                 }
 383                                 if (*stackptr <= origptr) {
 384                                         e = get_entry(table_base,
 385                                             private->underflow[hook]);
 386                                         pr_debug("Underflow (this is normal) "
 387                                                  "to %p\n", e);
 388                                 } else {
 389                                         e = jumpstack[--*stackptr];
 390                                         pr_debug("Pulled %p out from pos %u\n",
 391                                                  e, *stackptr);
 392                                         e = ipt_next_entry(e);
 393                                 }
 394                                 continue;
 395                         }
 396                         if (table_base + v != ipt_next_entry(e) &&
 397                             !(e->ip.flags & IPT_F_GOTO)) {
 398                                 if (*stackptr >= private->stacksize) {
 399                                         verdict = NF_DROP;
 400                                         break;
 401                                 }
 402                                 jumpstack[(*stackptr)++] = e;
 403                                 pr_debug("Pushed %p into pos %u\n",
 404                                          e, *stackptr - 1);
 405                         }
 406
 407                         e = get_entry(table_base, v);
 408                         continue;
 409                 }
 410
 411                 acpar.target   = t->u.kernel.target;
 412                 acpar.targinfo = t->data;
 413
 414                 verdict = t->u.kernel.target->target(skb, &acpar);
 415                 /* Target might have changed stuff. */
 416                 ip = ip_hdr(skb);
 417                 if (verdict == XT_CONTINUE)
 418                         e = ipt_next_entry(e);
 419                 else
 420                         /* Verdict */
 421                         break;
 422         } while (!acpar.hotdrop);
 423         pr_debug("Exiting %s; resetting sp from %u to %u\n",
 424                  __func__, *stackptr, origptr);
 425         *stackptr = origptr;
 426         xt_write_recseq_end(addend);
 427         local_bh_enable();
 428
 429 #ifdef DEBUG_ALLOW_ALL
 430         return NF_ACCEPT;
 431 #else
 432         if (acpar.hotdrop)
 433                 return NF_DROP;
 434         else return verdict;
 435 #endif
 436 }
 437
 438 /* Figures out from what hook each rule can be called: returns 0 if
 439    there are loops.  Puts hook bitmask in comefrom. */
 440 static int
 441 mark_source_chains(const struct xt_table_info *newinfo,
 442                    unsigned int valid_hooks, void *entry0)
 443 {
 444         unsigned int hook;
 445
 446         /* No recursion; use packet counter to save back ptrs (reset
 447            to 0 as we leave), and comefrom to save source hook bitmask */
 448         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 449                 unsigned int pos = newinfo->hook_entry[hook];
 450                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 451
 452                 if (!(valid_hooks & (1 << hook)))
 453                         continue;
 454
 455                 /* Set initial back pointer. */
 456                 e->counters.pcnt = pos;
 457
 458                 for (;;) {
 459                         const struct xt_standard_target *t
 460                                 = (void *)ipt_get_target_c(e);
 461                         int visited = e->comefrom & (1 << hook);
 462
 463                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 464                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 465                                        hook, pos, e->comefrom);
 466                                 return 0;
 467                         }
 468                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 469
 470                         /* Unconditional return/END. */
 471                         if ((e->target_offset == sizeof(struct ipt_entry) &&
 472                              (strcmp(t->target.u.user.name,
 473                                      XT_STANDARD_TARGET) == 0) &&
 474                              t->verdict < 0 && unconditional(&e->ip)) ||
 475                             visited) {
 476                                 unsigned int oldpos, size;
 477
 478                                 if ((strcmp(t->target.u.user.name,
 479                                             XT_STANDARD_TARGET) == 0) &&
 480                                     t->verdict < -NF_MAX_VERDICT - 1) {
 481                                         duprintf("mark_source_chains: bad "
 482                                                 "negative verdict (%i)\n",
 483                                                                 t->verdict);
 484                                         return 0;
 485                                 }
 486
 487                                 /* Return: backtrack through the last
 488                                    big jump. */
 489                                 do {
 490                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 491 #ifdef DEBUG_IP_FIREWALL_USER
 492                                         if (e->comefrom
 493                                             & (1 << NF_INET_NUMHOOKS)) {
 494                                                 duprintf("Back unset "
 495                                                          "on hook %u "
 496                                                          "rule %u\n",
 497                                                          hook, pos);
 498                                         }
 499 #endif
 500                                         oldpos = pos;
 501                                         pos = e->counters.pcnt;
 502                                         e->counters.pcnt = 0;
 503
 504                                         /* We're at the start. */
 505                                         if (pos == oldpos)
 506                                                 goto next;
 507
 508                                         e = (struct ipt_entry *)
 509                                                 (entry0 + pos);
 510                                 } while (oldpos == pos + e->next_offset);
 511
 512                                 /* Move along one */
 513                                 size = e->next_offset;
 514                                 e = (struct ipt_entry *)
 515                                         (entry0 + pos + size);
 516                                 e->counters.pcnt = pos;
 517                                 pos += size;
 518                         } else {
 519                                 int newpos = t->verdict;
 520
 521                                 if (strcmp(t->target.u.user.name,
 522                                            XT_STANDARD_TARGET) == 0 &&
 523                                     newpos >= 0) {
 524                                         if (newpos > newinfo->size -
 525                                                 sizeof(struct ipt_entry)) {
 526                                                 duprintf("mark_source_chains: "
 527                                                         "bad verdict (%i)\n",
 528                                                                 newpos);
 529                                                 return 0;
 530                                         }
 531                                         /* This a jump; chase it. */
 532                                         duprintf("Jump rule %u -> %u\n",
 533                                                  pos, newpos);
 534                                 } else {
 535                                         /* ... this is a fallthru */
 536                                         newpos = pos + e->next_offset;
 537                                 }
 538                                 e = (struct ipt_entry *)
 539                                         (entry0 + newpos);
 540                                 e->counters.pcnt = pos;
 541                                 pos = newpos;
 542                         }
 543                 }
 544                 next:
 545                 duprintf("Finished chain %u\n", hook);
 546         }
 547         return 1;
 548 }
 549
 550 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 551 {
 552         struct xt_mtdtor_param par;
 553
 554         par.net       = net;
 555         par.match     = m->u.kernel.match;
 556         par.matchinfo = m->data;
 557         par.family    = NFPROTO_IPV4;
 558         if (par.match->destroy != NULL)
 559                 par.match->destroy(&par);
 560         module_put(par.match->me);
 561 }
 562
 563 static int
 564 check_entry(const struct ipt_entry *e)
 565 {
 566         const struct xt_entry_target *t;
 567
 568         if (!ip_checkentry(&e->ip))
 569                 return -EINVAL;
 570
 571         if (e->target_offset + sizeof(struct xt_entry_target) >
 572             e->next_offset)
 573                 return -EINVAL;
 574
 575         t = ipt_get_target_c(e);
 576         if (e->target_offset + t->u.target_size > e->next_offset)
 577                 return -EINVAL;
 578
 579         return 0;
 580 }
 581
 582 static int
 583 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 584 {
 585         const struct ipt_ip *ip = par->entryinfo;
 586         int ret;
 587
 588         par->match     = m->u.kernel.match;
 589         par->matchinfo = m->data;
 590
 591         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 592               ip->proto, ip->invflags & IPT_INV_PROTO);
 593         if (ret < 0) {
 594                 duprintf("check failed for `%s'.\n", par->match->name);
 595                 return ret;
 596         }
 597         return 0;
 598 }
 599
 600 static int
 601 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 602 {
 603         struct xt_match *match;
 604         int ret;
 605
 606         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 607                                       m->u.user.revision);
 608         if (IS_ERR(match)) {
 609                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 610                 return PTR_ERR(match);
 611         }
 612         m->u.kernel.match = match;
 613
 614         ret = check_match(m, par);
 615         if (ret)
 616                 goto err;
 617
 618         return 0;
 619 err:
 620         module_put(m->u.kernel.match->me);
 621         return ret;
 622 }
 623
 624 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 625 {
 626         struct xt_entry_target *t = ipt_get_target(e);
 627         struct xt_tgchk_param par = {
 628                 .net       = net,
 629                 .table     = name,
 630                 .entryinfo = e,
 631                 .target    = t->u.kernel.target,
 632                 .targinfo  = t->data,
 633                 .hook_mask = e->comefrom,
 634                 .family    = NFPROTO_IPV4,
 635         };
 636         int ret;
 637
 638         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 639               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 640         if (ret < 0) {
 641                 duprintf("check failed for `%s'.\n",
 642                          t->u.kernel.target->name);
 643                 return ret;
 644         }
 645         return 0;
 646 }
 647
 648 static int
 649 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 650                  unsigned int size)
 651 {
 652         struct xt_entry_target *t;
 653         struct xt_target *target;
 654         int ret;
 655         unsigned int j;
 656         struct xt_mtchk_param mtpar;
 657         struct xt_entry_match *ematch;
 658
 659         j = 0;
 660         mtpar.net       = net;
 661         mtpar.table     = name;
 662         mtpar.entryinfo = &e->ip;
 663         mtpar.hook_mask = e->comefrom;
 664         mtpar.family    = NFPROTO_IPV4;
 665         xt_ematch_foreach(ematch, e) {
 666                 ret = find_check_match(ematch, &mtpar);
 667                 if (ret != 0)
 668                         goto cleanup_matches;
 669                 ++j;
 670         }
 671
 672         t = ipt_get_target(e);
 673         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 674                                         t->u.user.revision);
 675         if (IS_ERR(target)) {
 676                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 677                 ret = PTR_ERR(target);
 678                 goto cleanup_matches;
 679         }
 680         t->u.kernel.target = target;
 681
 682         ret = check_target(e, net, name);
 683         if (ret)
 684                 goto err;
 685         return 0;
 686  err:
 687         module_put(t->u.kernel.target->me);
 688  cleanup_matches:
 689         xt_ematch_foreach(ematch, e) {
 690                 if (j-- == 0)
 691                         break;
 692                 cleanup_match(ematch, net);
 693         }
 694         return ret;
 695 }
 696
 697 static bool check_underflow(const struct ipt_entry *e)
 698 {
 699         const struct xt_entry_target *t;
 700         unsigned int verdict;
 701
 702         if (!unconditional(&e->ip))
 703                 return false;
 704         t = ipt_get_target_c(e);
 705         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 706                 return false;
 707         verdict = ((struct xt_standard_target *)t)->verdict;
 708         verdict = -verdict - 1;
 709         return verdict == NF_DROP || verdict == NF_ACCEPT;
 710 }
 711
 712 static int
 713 check_entry_size_and_hooks(struct ipt_entry *e,
 714                            struct xt_table_info *newinfo,
 715                            const unsigned char *base,
 716                            const unsigned char *limit,
 717                            const unsigned int *hook_entries,
 718                            const unsigned int *underflows,
 719                            unsigned int valid_hooks)
 720 {
 721         unsigned int h;
 722         int err;
 723
 724         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 725             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 726             (unsigned char *)e + e->next_offset > limit) {
 727                 duprintf("Bad offset %p\n", e);
 728                 return -EINVAL;
 729         }
 730
 731         if (e->next_offset
 732             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 733                 duprintf("checking: element %p size %u\n",
 734                          e, e->next_offset);
 735                 return -EINVAL;
 736         }
 737
 738         err = check_entry(e);
 739         if (err)
 740                 return err;
 741
 742         /* Check hooks & underflows */
 743         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 744                 if (!(valid_hooks & (1 << h)))
 745                         continue;
 746                 if ((unsigned char *)e - base == hook_entries[h])
 747                         newinfo->hook_entry[h] = hook_entries[h];
 748                 if ((unsigned char *)e - base == underflows[h]) {
 749                         if (!check_underflow(e)) {
 750                                 pr_err("Underflows must be unconditional and "
 751                                        "use the STANDARD target with "
 752                                        "ACCEPT/DROP\n");
 753                                 return -EINVAL;
 754                         }
 755                         newinfo->underflow[h] = underflows[h];
 756                 }
 757         }
 758
 759         /* Clear counters and comefrom */
 760         e->counters = ((struct xt_counters) { 0, 0 });
 761         e->comefrom = 0;
 762         return 0;
 763 }
 764
 765 static void
 766 cleanup_entry(struct ipt_entry *e, struct net *net)
 767 {
 768         struct xt_tgdtor_param par;
 769         struct xt_entry_target *t;
 770         struct xt_entry_match *ematch;
 771
 772         /* Cleanup all matches */
 773         xt_ematch_foreach(ematch, e)
 774                 cleanup_match(ematch, net);
 775         t = ipt_get_target(e);
 776
 777         par.net      = net;
 778         par.target   = t->u.kernel.target;
 779         par.targinfo = t->data;
 780         par.family   = NFPROTO_IPV4;
 781         if (par.target->destroy != NULL)
 782                 par.target->destroy(&par);
 783         module_put(par.target->me);
 784 }
 785
 786 /* Checks and translates the user-supplied table segment (held in
 787    newinfo) */
 788 static int
 789 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 790                 const struct ipt_replace *repl)
 791 {
 792         struct ipt_entry *iter;
 793         unsigned int i;
 794         int ret = 0;
 795
 796         newinfo->size = repl->size;
 797         newinfo->number = repl->num_entries;
 798
 799         /* Init all hooks to impossible value. */
 800         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 801                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 802                 newinfo->underflow[i] = 0xFFFFFFFF;
 803         }
 804
 805         duprintf("translate_table: size %u\n", newinfo->size);
 806         i = 0;
 807         /* Walk through entries, checking offsets. */
 808         xt_entry_foreach(iter, entry0, newinfo->size) {
 809                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 810                                                  entry0 + repl->size,
 811                                                  repl->hook_entry,
 812                                                  repl->underflow,
 813                                                  repl->valid_hooks);
 814                 if (ret != 0)
 815                         return ret;
 816                 ++i;
 817                 if (strcmp(ipt_get_target(iter)->u.user.name,
 818                     XT_ERROR_TARGET) == 0)
 819                         ++newinfo->stacksize;
 820         }
 821
 822         if (i != repl->num_entries) {
 823                 duprintf("translate_table: %u not %u entries\n",
 824                          i, repl->num_entries);
 825                 return -EINVAL;
 826         }
 827
 828         /* Check hooks all assigned */
 829         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 830                 /* Only hooks which are valid */
 831                 if (!(repl->valid_hooks & (1 << i)))
 832                         continue;
 833                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 834                         duprintf("Invalid hook entry %u %u\n",
 835                                  i, repl->hook_entry[i]);
 836                         return -EINVAL;
 837                 }
 838                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 839                         duprintf("Invalid underflow %u %u\n",
 840                                  i, repl->underflow[i]);
 841                         return -EINVAL;
 842                 }
 843         }
 844
 845         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 846                 return -ELOOP;
 847
 848         /* Finally, each sanity check must pass */
 849         i = 0;
 850         xt_entry_foreach(iter, entry0, newinfo->size) {
 851                 ret = find_check_entry(iter, net, repl->name, repl->size);
 852                 if (ret != 0)
 853                         break;
 854                 ++i;
 855         }
 856
 857         if (ret != 0) {
 858                 xt_entry_foreach(iter, entry0, newinfo->size) {
 859                         if (i-- == 0)
 860                                 break;
 861                         cleanup_entry(iter, net);
 862                 }
 863                 return ret;
 864         }
 865
 866         /* And one copy for every other CPU */
 867         for_each_possible_cpu(i) {
 868                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
 869                         memcpy(newinfo->entries[i], entry0, newinfo->size);
 870         }
 871
 872         return ret;
 873 }
 874
 875 static void
 876 get_counters(const struct xt_table_info *t,
 877              struct xt_counters counters[])
 878 {
 879         struct ipt_entry *iter;
 880         unsigned int cpu;
 881         unsigned int i;
 882
 883         for_each_possible_cpu(cpu) {
 884                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 885
 886                 i = 0;
 887                 xt_entry_foreach(iter, t->entries[cpu], t->size) {
 888                         u64 bcnt, pcnt;
 889                         unsigned int start;
 890
 891                         do {
 892                                 start = read_seqcount_begin(s);
 893                                 bcnt = iter->counters.bcnt;
 894                                 pcnt = iter->counters.pcnt;
 895                         } while (read_seqcount_retry(s, start));
 896
 897                         ADD_COUNTER(counters[i], bcnt, pcnt);
 898                         ++i; /* macro does multi eval of i */
 899                 }
 900         }
 901 }
 902
 903 static struct xt_counters *alloc_counters(const struct xt_table *table)
 904 {
 905         unsigned int countersize;
 906         struct xt_counters *counters;
 907         const struct xt_table_info *private = table->private;
 908
 909         /* We need atomic snapshot of counters: rest doesn't change
 910            (other than comefrom, which userspace doesn't care
 911            about). */
 912         countersize = sizeof(struct xt_counters) * private->number;
 913         counters = vzalloc(countersize);
 914
 915         if (counters == NULL)
 916                 return ERR_PTR(-ENOMEM);
 917
 918         get_counters(private, counters);
 919
 920         return counters;
 921 }
 922
 923 static int
 924 copy_entries_to_user(unsigned int total_size,
 925                      const struct xt_table *table,
 926                      void __user *userptr)
 927 {
 928         unsigned int off, num;
 929         const struct ipt_entry *e;
 930         struct xt_counters *counters;
 931         const struct xt_table_info *private = table->private;
 932         int ret = 0;
 933         const void *loc_cpu_entry;
 934
 935         counters = alloc_counters(table);
 936         if (IS_ERR(counters))
 937                 return PTR_ERR(counters);
 938
 939         /* choose the copy that is on our node/cpu, ...
 940          * This choice is lazy (because current thread is
 941          * allowed to migrate to another cpu)
 942          */
 943         loc_cpu_entry = private->entries[raw_smp_processor_id()];
 944         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 945                 ret = -EFAULT;
 946                 goto free_counters;
 947         }
 948
 949         /* FIXME: use iterator macros --RR */
 950         /* ... then go back and fix counters and names */
 951         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 952                 unsigned int i;
 953                 const struct xt_entry_match *m;
 954                 const struct xt_entry_target *t;
 955
 956                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 957                 if (copy_to_user(userptr + off
 958                                  + offsetof(struct ipt_entry, counters),
 959                                  &counters[num],
 960                                  sizeof(counters[num])) != 0) {
 961                         ret = -EFAULT;
 962                         goto free_counters;
 963                 }
 964
 965                 for (i = sizeof(struct ipt_entry);
 966                      i < e->target_offset;
 967                      i += m->u.match_size) {
 968                         m = (void *)e + i;
 969
 970                         if (copy_to_user(userptr + off + i
 971                                          + offsetof(struct xt_entry_match,
 972                                                     u.user.name),
 973                                          m->u.kernel.match->name,
 974                                          strlen(m->u.kernel.match->name)+1)
 975                             != 0) {
 976                                 ret = -EFAULT;
 977                                 goto free_counters;
 978                         }
 979                 }
 980
 981                 t = ipt_get_target_c(e);
 982                 if (copy_to_user(userptr + off + e->target_offset
 983                                  + offsetof(struct xt_entry_target,
 984                                             u.user.name),
 985                                  t->u.kernel.target->name,
 986                                  strlen(t->u.kernel.target->name)+1) != 0) {
 987                         ret = -EFAULT;
 988                         goto free_counters;
 989                 }
 990         }
 991
 992  free_counters:
 993         vfree(counters);
 994         return ret;
 995 }
 996
 997 #ifdef CONFIG_COMPAT
 998 static void compat_standard_from_user(void *dst, const void *src)
 999 {
1000         int v = *(compat_int_t *)src;
1001
1002         if (v > 0)
1003                 v += xt_compat_calc_jump(AF_INET, v);
1004         memcpy(dst, &v, sizeof(v));
1005 }
1006
1007 static int compat_standard_to_user(void __user *dst, const void *src)
1008 {
1009         compat_int_t cv = *(int *)src;
1010
1011         if (cv > 0)
1012                 cv -= xt_compat_calc_jump(AF_INET, cv);
1013         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1014 }
1015
1016 static int compat_calc_entry(const struct ipt_entry *e,
1017                              const struct xt_table_info *info,
1018                              const void *base, struct xt_table_info *newinfo)
1019 {
1020         const struct xt_entry_match *ematch;
1021         const struct xt_entry_target *t;
1022         unsigned int entry_offset;
1023         int off, i, ret;
1024
1025         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1026         entry_offset = (void *)e - base;
1027         xt_ematch_foreach(ematch, e)
1028                 off += xt_compat_match_offset(ematch->u.kernel.match);
1029         t = ipt_get_target_c(e);
1030         off += xt_compat_target_offset(t->u.kernel.target);
1031         newinfo->size -= off;
1032         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1033         if (ret)
1034                 return ret;
1035
1036         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1037                 if (info->hook_entry[i] &&
1038                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1039                         newinfo->hook_entry[i] -= off;
1040                 if (info->underflow[i] &&
1041                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1042                         newinfo->underflow[i] -= off;
1043         }
1044         return 0;
1045 }
1046
1047 static int compat_table_info(const struct xt_table_info *info,
1048                              struct xt_table_info *newinfo)
1049 {
1050         struct ipt_entry *iter;
1051         void *loc_cpu_entry;
1052         int ret;
1053
1054         if (!newinfo || !info)
1055                 return -EINVAL;
1056
1057         /* we dont care about newinfo->entries[] */
1058         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1059         newinfo->initial_entries = 0;
1060         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1061         xt_compat_init_offsets(AF_INET, info->number);
1062         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1063                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1064                 if (ret != 0)
1065                         return ret;
1066         }
1067         return 0;
1068 }
1069 #endif
1070
1071 static int get_info(struct net *net, void __user *user,
1072                     const int *len, int compat)
1073 {
1074         char name[XT_TABLE_MAXNAMELEN];
1075         struct xt_table *t;
1076         int ret;
1077
1078         if (*len != sizeof(struct ipt_getinfo)) {
1079                 duprintf("length %u != %zu\n", *len,
1080                          sizeof(struct ipt_getinfo));
1081                 return -EINVAL;
1082         }
1083
1084         if (copy_from_user(name, user, sizeof(name)) != 0)
1085                 return -EFAULT;
1086
1087         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1088 #ifdef CONFIG_COMPAT
1089         if (compat)
1090                 xt_compat_lock(AF_INET);
1091 #endif
1092         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1093                                     "iptable_%s", name);
1094         if (t && !IS_ERR(t)) {
1095                 struct ipt_getinfo info;
1096                 const struct xt_table_info *private = t->private;
1097 #ifdef CONFIG_COMPAT
1098                 struct xt_table_info tmp;
1099
1100                 if (compat) {
1101                         ret = compat_table_info(private, &tmp);
1102                         xt_compat_flush_offsets(AF_INET);
1103                         private = &tmp;
1104                 }
1105 #endif
1106                 memset(&info, 0, sizeof(info));
1107                 info.valid_hooks = t->valid_hooks;
1108                 memcpy(info.hook_entry, private->hook_entry,
1109                        sizeof(info.hook_entry));
1110                 memcpy(info.underflow, private->underflow,
1111                        sizeof(info.underflow));
1112                 info.num_entries = private->number;
1113                 info.size = private->size;
1114                 strcpy(info.name, name);
1115
1116                 if (copy_to_user(user, &info, *len) != 0)
1117                         ret = -EFAULT;
1118                 else
1119                         ret = 0;
1120
1121                 xt_table_unlock(t);
1122                 module_put(t->me);
1123         } else
1124                 ret = t ? PTR_ERR(t) : -ENOENT;
1125 #ifdef CONFIG_COMPAT
1126         if (compat)
1127                 xt_compat_unlock(AF_INET);
1128 #endif
1129         return ret;
1130 }
1131
1132 static int
1133 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1134             const int *len)
1135 {
1136         int ret;
1137         struct ipt_get_entries get;
1138         struct xt_table *t;
1139
1140         if (*len < sizeof(get)) {
1141                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1142                 return -EINVAL;
1143         }
1144         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1145                 return -EFAULT;
1146         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1147                 duprintf("get_entries: %u != %zu\n",
1148                          *len, sizeof(get) + get.size);
1149                 return -EINVAL;
1150         }
1151
1152         t = xt_find_table_lock(net, AF_INET, get.name);
1153         if (t && !IS_ERR(t)) {
1154                 const struct xt_table_info *private = t->private;
1155                 duprintf("t->private->number = %u\n", private->number);
1156                 if (get.size == private->size)
1157                         ret = copy_entries_to_user(private->size,
1158                                                    t, uptr->entrytable);
1159                 else {
1160                         duprintf("get_entries: I've got %u not %u!\n",
1161                                  private->size, get.size);
1162                         ret = -EAGAIN;
1163                 }
1164                 module_put(t->me);
1165                 xt_table_unlock(t);
1166         } else
1167                 ret = t ? PTR_ERR(t) : -ENOENT;
1168
1169         return ret;
1170 }
1171
1172 static int
1173 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1174              struct xt_table_info *newinfo, unsigned int num_counters,
1175              void __user *counters_ptr)
1176 {
1177         int ret;
1178         struct xt_table *t;
1179         struct xt_table_info *oldinfo;
1180         struct xt_counters *counters;
1181         void *loc_cpu_old_entry;
1182         struct ipt_entry *iter;
1183
1184         ret = 0;
1185         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1186         if (!counters) {
1187                 ret = -ENOMEM;
1188                 goto out;
1189         }
1190
1191         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1192                                     "iptable_%s", name);
1193         if (!t || IS_ERR(t)) {
1194                 ret = t ? PTR_ERR(t) : -ENOENT;
1195                 goto free_newinfo_counters_untrans;
1196         }
1197
1198         /* You lied! */
1199         if (valid_hooks != t->valid_hooks) {
1200                 duprintf("Valid hook crap: %08X vs %08X\n",
1201                          valid_hooks, t->valid_hooks);
1202                 ret = -EINVAL;
1203                 goto put_module;
1204         }
1205
1206         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1207         if (!oldinfo)
1208                 goto put_module;
1209
1210         /* Update module usage count based on number of rules */
1211         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1212                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1213         if ((oldinfo->number > oldinfo->initial_entries) ||
1214             (newinfo->number <= oldinfo->initial_entries))
1215                 module_put(t->me);
1216         if ((oldinfo->number > oldinfo->initial_entries) &&
1217             (newinfo->number <= oldinfo->initial_entries))
1218                 module_put(t->me);
1219
1220         /* Get the old counters, and synchronize with replace */
1221         get_counters(oldinfo, counters);
1222
1223         /* Decrease module usage counts and free resource */
1224         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1225         xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1226                 cleanup_entry(iter, net);
1227
1228         xt_free_table_info(oldinfo);
1229         if (copy_to_user(counters_ptr, counters,
1230                          sizeof(struct xt_counters) * num_counters) != 0) {
1231                 /* Silent error, can't fail, new table is already in place */
1232                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1233         }
1234         vfree(counters);
1235         xt_table_unlock(t);
1236         return ret;
1237
1238  put_module:
1239         module_put(t->me);
1240         xt_table_unlock(t);
1241  free_newinfo_counters_untrans:
1242         vfree(counters);
1243  out:
1244         return ret;
1245 }
1246
1247 static int
1248 do_replace(struct net *net, const void __user *user, unsigned int len)
1249 {
1250         int ret;
1251         struct ipt_replace tmp;
1252         struct xt_table_info *newinfo;
1253         void *loc_cpu_entry;
1254         struct ipt_entry *iter;
1255
1256         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1257                 return -EFAULT;
1258
1259         /* overflow check */
1260         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1261                 return -ENOMEM;
1262         tmp.name[sizeof(tmp.name)-1] = 0;
1263
1264         newinfo = xt_alloc_table_info(tmp.size);
1265         if (!newinfo)
1266                 return -ENOMEM;
1267
1268         /* choose the copy that is on our node/cpu */
1269         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1270         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1271                            tmp.size) != 0) {
1272                 ret = -EFAULT;
1273                 goto free_newinfo;
1274         }
1275
1276         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1277         if (ret != 0)
1278                 goto free_newinfo;
1279
1280         duprintf("Translated table\n");
1281
1282         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1283                            tmp.num_counters, tmp.counters);
1284         if (ret)
1285                 goto free_newinfo_untrans;
1286         return 0;
1287
1288  free_newinfo_untrans:
1289         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1290                 cleanup_entry(iter, net);
1291  free_newinfo:
1292         xt_free_table_info(newinfo);
1293         return ret;
1294 }
1295
1296 static int
1297 do_add_counters(struct net *net, const void __user *user,
1298                 unsigned int len, int compat)
1299 {
1300         unsigned int i, curcpu;
1301         struct xt_counters_info tmp;
1302         struct xt_counters *paddc;
1303         unsigned int num_counters;
1304         const char *name;
1305         int size;
1306         void *ptmp;
1307         struct xt_table *t;
1308         const struct xt_table_info *private;
1309         int ret = 0;
1310         void *loc_cpu_entry;
1311         struct ipt_entry *iter;
1312         unsigned int addend;
1313 #ifdef CONFIG_COMPAT
1314         struct compat_xt_counters_info compat_tmp;
1315
1316         if (compat) {
1317                 ptmp = &compat_tmp;
1318                 size = sizeof(struct compat_xt_counters_info);
1319         } else
1320 #endif
1321         {
1322                 ptmp = &tmp;
1323                 size = sizeof(struct xt_counters_info);
1324         }
1325
1326         if (copy_from_user(ptmp, user, size) != 0)
1327                 return -EFAULT;
1328
1329 #ifdef CONFIG_COMPAT
1330         if (compat) {
1331                 num_counters = compat_tmp.num_counters;
1332                 name = compat_tmp.name;
1333         } else
1334 #endif
1335         {
1336                 num_counters = tmp.num_counters;
1337                 name = tmp.name;
1338         }
1339
1340         if (len != size + num_counters * sizeof(struct xt_counters))
1341                 return -EINVAL;
1342
1343         paddc = vmalloc(len - size);
1344         if (!paddc)
1345                 return -ENOMEM;
1346
1347         if (copy_from_user(paddc, user + size, len - size) != 0) {
1348                 ret = -EFAULT;
1349                 goto free;
1350         }
1351
1352         t = xt_find_table_lock(net, AF_INET, name);
1353         if (!t || IS_ERR(t)) {
1354                 ret = t ? PTR_ERR(t) : -ENOENT;
1355                 goto free;
1356         }
1357
1358         local_bh_disable();
1359         private = t->private;
1360         if (private->number != num_counters) {
1361                 ret = -EINVAL;
1362                 goto unlock_up_free;
1363         }
1364
1365         i = 0;
1366         /* Choose the copy that is on our node */
1367         curcpu = smp_processor_id();
1368         loc_cpu_entry = private->entries[curcpu];
1369         addend = xt_write_recseq_begin();
1370         xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1371                 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1372                 ++i;
1373         }
1374         xt_write_recseq_end(addend);
1375  unlock_up_free:
1376         local_bh_enable();
1377         xt_table_unlock(t);
1378         module_put(t->me);
1379  free:
1380         vfree(paddc);
1381
1382         return ret;
1383 }
1384
1385 #ifdef CONFIG_COMPAT
1386 struct compat_ipt_replace {
1387         char                    name[XT_TABLE_MAXNAMELEN];
1388         u32                     valid_hooks;
1389         u32                     num_entries;
1390         u32                     size;
1391         u32                     hook_entry[NF_INET_NUMHOOKS];
1392         u32                     underflow[NF_INET_NUMHOOKS];
1393         u32                     num_counters;
1394         compat_uptr_t           counters;       /* struct xt_counters * */
1395         struct compat_ipt_entry entries[0];
1396 };
1397
1398 static int
1399 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1400                           unsigned int *size, struct xt_counters *counters,
1401                           unsigned int i)
1402 {
1403         struct xt_entry_target *t;
1404         struct compat_ipt_entry __user *ce;
1405         u_int16_t target_offset, next_offset;
1406         compat_uint_t origsize;
1407         const struct xt_entry_match *ematch;
1408         int ret = 0;
1409
1410         origsize = *size;
1411         ce = (struct compat_ipt_entry __user *)*dstptr;
1412         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1413             copy_to_user(&ce->counters, &counters[i],
1414             sizeof(counters[i])) != 0)
1415                 return -EFAULT;
1416
1417         *dstptr += sizeof(struct compat_ipt_entry);
1418         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1419
1420         xt_ematch_foreach(ematch, e) {
1421                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1422                 if (ret != 0)
1423                         return ret;
1424         }
1425         target_offset = e->target_offset - (origsize - *size);
1426         t = ipt_get_target(e);
1427         ret = xt_compat_target_to_user(t, dstptr, size);
1428         if (ret)
1429                 return ret;
1430         next_offset = e->next_offset - (origsize - *size);
1431         if (put_user(target_offset, &ce->target_offset) != 0 ||
1432             put_user(next_offset, &ce->next_offset) != 0)
1433                 return -EFAULT;
1434         return 0;
1435 }
1436
1437 static int
1438 compat_find_calc_match(struct xt_entry_match *m,
1439                        const char *name,
1440                        const struct ipt_ip *ip,
1441                        unsigned int hookmask,
1442                        int *size)
1443 {
1444         struct xt_match *match;
1445
1446         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1447                                       m->u.user.revision);
1448         if (IS_ERR(match)) {
1449                 duprintf("compat_check_calc_match: `%s' not found\n",
1450                          m->u.user.name);
1451                 return PTR_ERR(match);
1452         }
1453         m->u.kernel.match = match;
1454         *size += xt_compat_match_offset(match);
1455         return 0;
1456 }
1457
1458 static void compat_release_entry(struct compat_ipt_entry *e)
1459 {
1460         struct xt_entry_target *t;
1461         struct xt_entry_match *ematch;
1462
1463         /* Cleanup all matches */
1464         xt_ematch_foreach(ematch, e)
1465                 module_put(ematch->u.kernel.match->me);
1466         t = compat_ipt_get_target(e);
1467         module_put(t->u.kernel.target->me);
1468 }
1469
1470 static int
1471 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1472                                   struct xt_table_info *newinfo,
1473                                   unsigned int *size,
1474                                   const unsigned char *base,
1475                                   const unsigned char *limit,
1476                                   const unsigned int *hook_entries,
1477                                   const unsigned int *underflows,
1478                                   const char *name)
1479 {
1480         struct xt_entry_match *ematch;
1481         struct xt_entry_target *t;
1482         struct xt_target *target;
1483         unsigned int entry_offset;
1484         unsigned int j;
1485         int ret, off, h;
1486
1487         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1488         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1489             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1490             (unsigned char *)e + e->next_offset > limit) {
1491                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1492                 return -EINVAL;
1493         }
1494
1495         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1496                              sizeof(struct compat_xt_entry_target)) {
1497                 duprintf("checking: element %p size %u\n",
1498                          e, e->next_offset);
1499                 return -EINVAL;
1500         }
1501
1502         /* For purposes of check_entry casting the compat entry is fine */
1503         ret = check_entry((struct ipt_entry *)e);
1504         if (ret)
1505                 return ret;
1506
1507         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1508         entry_offset = (void *)e - (void *)base;
1509         j = 0;
1510         xt_ematch_foreach(ematch, e) {
1511                 ret = compat_find_calc_match(ematch, name,
1512                                              &e->ip, e->comefrom, &off);
1513                 if (ret != 0)
1514                         goto release_matches;
1515                 ++j;
1516         }
1517
1518         t = compat_ipt_get_target(e);
1519         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1520                                         t->u.user.revision);
1521         if (IS_ERR(target)) {
1522                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1523                          t->u.user.name);
1524                 ret = PTR_ERR(target);
1525                 goto release_matches;
1526         }
1527         t->u.kernel.target = target;
1528
1529         off += xt_compat_target_offset(target);
1530         *size += off;
1531         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1532         if (ret)
1533                 goto out;
1534
1535         /* Check hooks & underflows */
1536         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1537                 if ((unsigned char *)e - base == hook_entries[h])
1538                         newinfo->hook_entry[h] = hook_entries[h];
1539                 if ((unsigned char *)e - base == underflows[h])
1540                         newinfo->underflow[h] = underflows[h];
1541         }
1542
1543         /* Clear counters and comefrom */
1544         memset(&e->counters, 0, sizeof(e->counters));
1545         e->comefrom = 0;
1546         return 0;
1547
1548 out:
1549         module_put(t->u.kernel.target->me);
1550 release_matches:
1551         xt_ematch_foreach(ematch, e) {
1552                 if (j-- == 0)
1553                         break;
1554                 module_put(ematch->u.kernel.match->me);
1555         }
1556         return ret;
1557 }
1558
1559 static int
1560 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1561                             unsigned int *size, const char *name,
1562                             struct xt_table_info *newinfo, unsigned char *base)
1563 {
1564         struct xt_entry_target *t;
1565         struct xt_target *target;
1566         struct ipt_entry *de;
1567         unsigned int origsize;
1568         int ret, h;
1569         struct xt_entry_match *ematch;
1570
1571         ret = 0;
1572         origsize = *size;
1573         de = (struct ipt_entry *)*dstptr;
1574         memcpy(de, e, sizeof(struct ipt_entry));
1575         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1576
1577         *dstptr += sizeof(struct ipt_entry);
1578         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1579
1580         xt_ematch_foreach(ematch, e) {
1581                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1582                 if (ret != 0)
1583                         return ret;
1584         }
1585         de->target_offset = e->target_offset - (origsize - *size);
1586         t = compat_ipt_get_target(e);
1587         target = t->u.kernel.target;
1588         xt_compat_target_from_user(t, dstptr, size);
1589
1590         de->next_offset = e->next_offset - (origsize - *size);
1591         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1592                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1593                         newinfo->hook_entry[h] -= origsize - *size;
1594                 if ((unsigned char *)de - base < newinfo->underflow[h])
1595                         newinfo->underflow[h] -= origsize - *size;
1596         }
1597         return ret;
1598 }
1599
1600 static int
1601 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1602 {
1603         struct xt_entry_match *ematch;
1604         struct xt_mtchk_param mtpar;
1605         unsigned int j;
1606         int ret = 0;
1607
1608         j = 0;
1609         mtpar.net       = net;
1610         mtpar.table     = name;
1611         mtpar.entryinfo = &e->ip;
1612         mtpar.hook_mask = e->comefrom;
1613         mtpar.family    = NFPROTO_IPV4;
1614         xt_ematch_foreach(ematch, e) {
1615                 ret = check_match(ematch, &mtpar);
1616                 if (ret != 0)
1617                         goto cleanup_matches;
1618                 ++j;
1619         }
1620
1621         ret = check_target(e, net, name);
1622         if (ret)
1623                 goto cleanup_matches;
1624         return 0;
1625
1626  cleanup_matches:
1627         xt_ematch_foreach(ematch, e) {
1628                 if (j-- == 0)
1629                         break;
1630                 cleanup_match(ematch, net);
1631         }
1632         return ret;
1633 }
1634
1635 static int
1636 translate_compat_table(struct net *net,
1637                        const char *name,
1638                        unsigned int valid_hooks,
1639                        struct xt_table_info **pinfo,
1640                        void **pentry0,
1641                        unsigned int total_size,
1642                        unsigned int number,
1643                        unsigned int *hook_entries,
1644                        unsigned int *underflows)
1645 {
1646         unsigned int i, j;
1647         struct xt_table_info *newinfo, *info;
1648         void *pos, *entry0, *entry1;
1649         struct compat_ipt_entry *iter0;
1650         struct ipt_entry *iter1;
1651         unsigned int size;
1652         int ret;
1653
1654         info = *pinfo;
1655         entry0 = *pentry0;
1656         size = total_size;
1657         info->number = number;
1658
1659         /* Init all hooks to impossible value. */
1660         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1661                 info->hook_entry[i] = 0xFFFFFFFF;
1662                 info->underflow[i] = 0xFFFFFFFF;
1663         }
1664
1665         duprintf("translate_compat_table: size %u\n", info->size);
1666         j = 0;
1667         xt_compat_lock(AF_INET);
1668         xt_compat_init_offsets(AF_INET, number);
1669         /* Walk through entries, checking offsets. */
1670         xt_entry_foreach(iter0, entry0, total_size) {
1671                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1672                                                         entry0,
1673                                                         entry0 + total_size,
1674                                                         hook_entries,
1675                                                         underflows,
1676                                                         name);
1677                 if (ret != 0)
1678                         goto out_unlock;
1679                 ++j;
1680         }
1681
1682         ret = -EINVAL;
1683         if (j != number) {
1684                 duprintf("translate_compat_table: %u not %u entries\n",
1685                          j, number);
1686                 goto out_unlock;
1687         }
1688
1689         /* Check hooks all assigned */
1690         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1691                 /* Only hooks which are valid */
1692                 if (!(valid_hooks & (1 << i)))
1693                         continue;
1694                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1695                         duprintf("Invalid hook entry %u %u\n",
1696                                  i, hook_entries[i]);
1697                         goto out_unlock;
1698                 }
1699                 if (info->underflow[i] == 0xFFFFFFFF) {
1700                         duprintf("Invalid underflow %u %u\n",
1701                                  i, underflows[i]);
1702                         goto out_unlock;
1703                 }
1704         }
1705
1706         ret = -ENOMEM;
1707         newinfo = xt_alloc_table_info(size);
1708         if (!newinfo)
1709                 goto out_unlock;
1710
1711         newinfo->number = number;
1712         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1713                 newinfo->hook_entry[i] = info->hook_entry[i];
1714                 newinfo->underflow[i] = info->underflow[i];
1715         }
1716         entry1 = newinfo->entries[raw_smp_processor_id()];
1717         pos = entry1;
1718         size = total_size;
1719         xt_entry_foreach(iter0, entry0, total_size) {
1720                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1721                                                   name, newinfo, entry1);
1722                 if (ret != 0)
1723                         break;
1724         }
1725         xt_compat_flush_offsets(AF_INET);
1726         xt_compat_unlock(AF_INET);
1727         if (ret)
1728                 goto free_newinfo;
1729
1730         ret = -ELOOP;
1731         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1732                 goto free_newinfo;
1733
1734         i = 0;
1735         xt_entry_foreach(iter1, entry1, newinfo->size) {
1736                 ret = compat_check_entry(iter1, net, name);
1737                 if (ret != 0)
1738                         break;
1739                 ++i;
1740                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1741                     XT_ERROR_TARGET) == 0)
1742                         ++newinfo->stacksize;
1743         }
1744         if (ret) {
1745                 /*
1746                  * The first i matches need cleanup_entry (calls ->destroy)
1747                  * because they had called ->check already. The other j-i
1748                  * entries need only release.
1749                  */
1750                 int skip = i;
1751                 j -= i;
1752                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1753                         if (skip-- > 0)
1754                                 continue;
1755                         if (j-- == 0)
1756                                 break;
1757                         compat_release_entry(iter0);
1758                 }
1759                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1760                         if (i-- == 0)
1761                                 break;
1762                         cleanup_entry(iter1, net);
1763                 }
1764                 xt_free_table_info(newinfo);
1765                 return ret;
1766         }
1767
1768         /* And one copy for every other CPU */
1769         for_each_possible_cpu(i)
1770                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1771                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1772
1773         *pinfo = newinfo;
1774         *pentry0 = entry1;
1775         xt_free_table_info(info);
1776         return 0;
1777
1778 free_newinfo:
1779         xt_free_table_info(newinfo);
1780 out:
1781         xt_entry_foreach(iter0, entry0, total_size) {
1782                 if (j-- == 0)
1783                         break;
1784                 compat_release_entry(iter0);
1785         }
1786         return ret;
1787 out_unlock:
1788         xt_compat_flush_offsets(AF_INET);
1789         xt_compat_unlock(AF_INET);
1790         goto out;
1791 }
1792
1793 static int
1794 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1795 {
1796         int ret;
1797         struct compat_ipt_replace tmp;
1798         struct xt_table_info *newinfo;
1799         void *loc_cpu_entry;
1800         struct ipt_entry *iter;
1801
1802         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1803                 return -EFAULT;
1804
1805         /* overflow check */
1806         if (tmp.size >= INT_MAX / num_possible_cpus())
1807                 return -ENOMEM;
1808         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1809                 return -ENOMEM;
1810         tmp.name[sizeof(tmp.name)-1] = 0;
1811
1812         newinfo = xt_alloc_table_info(tmp.size);
1813         if (!newinfo)
1814                 return -ENOMEM;
1815
1816         /* choose the copy that is on our node/cpu */
1817         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1818         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1819                            tmp.size) != 0) {
1820                 ret = -EFAULT;
1821                 goto free_newinfo;
1822         }
1823
1824         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1825                                      &newinfo, &loc_cpu_entry, tmp.size,
1826                                      tmp.num_entries, tmp.hook_entry,
1827                                      tmp.underflow);
1828         if (ret != 0)
1829                 goto free_newinfo;
1830
1831         duprintf("compat_do_replace: Translated table\n");
1832
1833         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1834                            tmp.num_counters, compat_ptr(tmp.counters));
1835         if (ret)
1836                 goto free_newinfo_untrans;
1837         return 0;
1838
1839  free_newinfo_untrans:
1840         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1841                 cleanup_entry(iter, net);
1842  free_newinfo:
1843         xt_free_table_info(newinfo);
1844         return ret;
1845 }
1846
1847 static int
1848 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1849                       unsigned int len)
1850 {
1851         int ret;
1852
1853         if (!capable(CAP_NET_ADMIN))
1854                 return -EPERM;
1855
1856         switch (cmd) {
1857         case IPT_SO_SET_REPLACE:
1858                 ret = compat_do_replace(sock_net(sk), user, len);
1859                 break;
1860
1861         case IPT_SO_SET_ADD_COUNTERS:
1862                 ret = do_add_counters(sock_net(sk), user, len, 1);
1863                 break;
1864
1865         default:
1866                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1867                 ret = -EINVAL;
1868         }
1869
1870         return ret;
1871 }
1872
1873 struct compat_ipt_get_entries {
1874         char name[XT_TABLE_MAXNAMELEN];
1875         compat_uint_t size;
1876         struct compat_ipt_entry entrytable[0];
1877 };
1878
1879 static int
1880 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1881                             void __user *userptr)
1882 {
1883         struct xt_counters *counters;
1884         const struct xt_table_info *private = table->private;
1885         void __user *pos;
1886         unsigned int size;
1887         int ret = 0;
1888         const void *loc_cpu_entry;
1889         unsigned int i = 0;
1890         struct ipt_entry *iter;
1891
1892         counters = alloc_counters(table);
1893         if (IS_ERR(counters))
1894                 return PTR_ERR(counters);
1895
1896         /* choose the copy that is on our node/cpu, ...
1897          * This choice is lazy (because current thread is
1898          * allowed to migrate to another cpu)
1899          */
1900         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1901         pos = userptr;
1902         size = total_size;
1903         xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1904                 ret = compat_copy_entry_to_user(iter, &pos,
1905                                                 &size, counters, i++);
1906                 if (ret != 0)
1907                         break;
1908         }
1909
1910         vfree(counters);
1911         return ret;
1912 }
1913
1914 static int
1915 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1916                    int *len)
1917 {
1918         int ret;
1919         struct compat_ipt_get_entries get;
1920         struct xt_table *t;
1921
1922         if (*len < sizeof(get)) {
1923                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1924                 return -EINVAL;
1925         }
1926
1927         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1928                 return -EFAULT;
1929
1930         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1931                 duprintf("compat_get_entries: %u != %zu\n",
1932                          *len, sizeof(get) + get.size);
1933                 return -EINVAL;
1934         }
1935
1936         xt_compat_lock(AF_INET);
1937         t = xt_find_table_lock(net, AF_INET, get.name);
1938         if (t && !IS_ERR(t)) {
1939                 const struct xt_table_info *private = t->private;
1940                 struct xt_table_info info;
1941                 duprintf("t->private->number = %u\n", private->number);
1942                 ret = compat_table_info(private, &info);
1943                 if (!ret && get.size == info.size) {
1944                         ret = compat_copy_entries_to_user(private->size,
1945                                                           t, uptr->entrytable);
1946                 } else if (!ret) {
1947                         duprintf("compat_get_entries: I've got %u not %u!\n",
1948                                  private->size, get.size);
1949                         ret = -EAGAIN;
1950                 }
1951                 xt_compat_flush_offsets(AF_INET);
1952                 module_put(t->me);
1953                 xt_table_unlock(t);
1954         } else
1955                 ret = t ? PTR_ERR(t) : -ENOENT;
1956
1957         xt_compat_unlock(AF_INET);
1958         return ret;
1959 }
1960
1961 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1962
1963 static int
1964 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1965 {
1966         int ret;
1967
1968         if (!capable(CAP_NET_ADMIN))
1969                 return -EPERM;
1970
1971         switch (cmd) {
1972         case IPT_SO_GET_INFO:
1973                 ret = get_info(sock_net(sk), user, len, 1);
1974                 break;
1975         case IPT_SO_GET_ENTRIES:
1976                 ret = compat_get_entries(sock_net(sk), user, len);
1977                 break;
1978         default:
1979                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1980         }
1981         return ret;
1982 }
1983 #endif
1984
1985 static int
1986 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1987 {
1988         int ret;
1989
1990         if (!capable(CAP_NET_ADMIN))
1991                 return -EPERM;
1992
1993         switch (cmd) {
1994         case IPT_SO_SET_REPLACE:
1995                 ret = do_replace(sock_net(sk), user, len);
1996                 break;
1997
1998         case IPT_SO_SET_ADD_COUNTERS:
1999                 ret = do_add_counters(sock_net(sk), user, len, 0);
2000                 break;
2001
2002         default:
2003                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2004                 ret = -EINVAL;
2005         }
2006
2007         return ret;
2008 }
2009
2010 static int
2011 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2012 {
2013         int ret;
2014
2015         if (!capable(CAP_NET_ADMIN))
2016                 return -EPERM;
2017
2018         switch (cmd) {
2019         case IPT_SO_GET_INFO:
2020                 ret = get_info(sock_net(sk), user, len, 0);
2021                 break;
2022
2023         case IPT_SO_GET_ENTRIES:
2024                 ret = get_entries(sock_net(sk), user, len);
2025                 break;
2026
2027         case IPT_SO_GET_REVISION_MATCH:
2028         case IPT_SO_GET_REVISION_TARGET: {
2029                 struct xt_get_revision rev;
2030                 int target;
2031
2032                 if (*len != sizeof(rev)) {
2033                         ret = -EINVAL;
2034                         break;
2035                 }
2036                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2037                         ret = -EFAULT;
2038                         break;
2039                 }
2040                 rev.name[sizeof(rev.name)-1] = 0;
2041
2042                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2043                         target = 1;
2044                 else
2045                         target = 0;
2046
2047                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2048                                                          rev.revision,
2049                                                          target, &ret),
2050                                         "ipt_%s", rev.name);
2051                 break;
2052         }
2053
2054         default:
2055                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2056                 ret = -EINVAL;
2057         }
2058
2059         return ret;
2060 }
2061
2062 struct xt_table *ipt_register_table(struct net *net,
2063                                     const struct xt_table *table,
2064                                     const struct ipt_replace *repl)
2065 {
2066         int ret;
2067         struct xt_table_info *newinfo;
2068         struct xt_table_info bootstrap = {0};
2069         void *loc_cpu_entry;
2070         struct xt_table *new_table;
2071
2072         newinfo = xt_alloc_table_info(repl->size);
2073         if (!newinfo) {
2074                 ret = -ENOMEM;
2075                 goto out;
2076         }
2077
2078         /* choose the copy on our node/cpu, but dont care about preemption */
2079         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2080         memcpy(loc_cpu_entry, repl->entries, repl->size);
2081
2082         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2083         if (ret != 0)
2084                 goto out_free;
2085
2086         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2087         if (IS_ERR(new_table)) {
2088                 ret = PTR_ERR(new_table);
2089                 goto out_free;
2090         }
2091
2092         return new_table;
2093
2094 out_free:
2095         xt_free_table_info(newinfo);
2096 out:
2097         return ERR_PTR(ret);
2098 }
2099
2100 void ipt_unregister_table(struct net *net, struct xt_table *table)
2101 {
2102         struct xt_table_info *private;
2103         void *loc_cpu_entry;
2104         struct module *table_owner = table->me;
2105         struct ipt_entry *iter;
2106
2107         private = xt_unregister_table(table);
2108
2109         /* Decrease module usage counts and free resources */
2110         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2111         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2112                 cleanup_entry(iter, net);
2113         if (private->number > private->initial_entries)
2114                 module_put(table_owner);
2115         xt_free_table_info(private);
2116 }
2117
2118 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2119 static inline bool
2120 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2121                      u_int8_t type, u_int8_t code,
2122                      bool invert)
2123 {
2124         return ((test_type == 0xFF) ||
2125                 (type == test_type && code >= min_code && code <= max_code))
2126                 ^ invert;
2127 }
2128
2129 static bool
2130 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2131 {
2132         const struct icmphdr *ic;
2133         struct icmphdr _icmph;
2134         const struct ipt_icmp *icmpinfo = par->matchinfo;
2135
2136         /* Must not be a fragment. */
2137         if (par->fragoff != 0)
2138                 return false;
2139
2140         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2141         if (ic == NULL) {
2142                 /* We've been asked to examine this packet, and we
2143                  * can't.  Hence, no choice but to drop.
2144                  */
2145                 duprintf("Dropping evil ICMP tinygram.\n");
2146                 par->hotdrop = true;
2147                 return false;
2148         }
2149
2150         return icmp_type_code_match(icmpinfo->type,
2151                                     icmpinfo->code[0],
2152                                     icmpinfo->code[1],
2153                                     ic->type, ic->code,
2154                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2155 }
2156
2157 static int icmp_checkentry(const struct xt_mtchk_param *par)
2158 {
2159         const struct ipt_icmp *icmpinfo = par->matchinfo;
2160
2161         /* Must specify no unknown invflags */
2162         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2163 }
2164
2165 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2166         {
2167                 .name             = XT_STANDARD_TARGET,
2168                 .targetsize       = sizeof(int),
2169                 .family           = NFPROTO_IPV4,
2170 #ifdef CONFIG_COMPAT
2171                 .compatsize       = sizeof(compat_int_t),
2172                 .compat_from_user = compat_standard_from_user,
2173                 .compat_to_user   = compat_standard_to_user,
2174 #endif
2175         },
2176         {
2177                 .name             = XT_ERROR_TARGET,
2178                 .target           = ipt_error,
2179                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2180                 .family           = NFPROTO_IPV4,
2181         },
2182 };
2183
2184 static struct nf_sockopt_ops ipt_sockopts = {
2185         .pf             = PF_INET,
2186         .set_optmin     = IPT_BASE_CTL,
2187         .set_optmax     = IPT_SO_SET_MAX+1,
2188         .set            = do_ipt_set_ctl,
2189 #ifdef CONFIG_COMPAT
2190         .compat_set     = compat_do_ipt_set_ctl,
2191 #endif
2192         .get_optmin     = IPT_BASE_CTL,
2193         .get_optmax     = IPT_SO_GET_MAX+1,
2194         .get            = do_ipt_get_ctl,
2195 #ifdef CONFIG_COMPAT
2196         .compat_get     = compat_do_ipt_get_ctl,
2197 #endif
2198         .owner          = THIS_MODULE,
2199 };
2200
2201 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2202         {
2203                 .name       = "icmp",
2204                 .match      = icmp_match,
2205                 .matchsize  = sizeof(struct ipt_icmp),
2206                 .checkentry = icmp_checkentry,
2207                 .proto      = IPPROTO_ICMP,
2208                 .family     = NFPROTO_IPV4,
2209         },
2210 };
2211
2212 static int __net_init ip_tables_net_init(struct net *net)
2213 {
2214         return xt_proto_init(net, NFPROTO_IPV4);
2215 }
2216
2217 static void __net_exit ip_tables_net_exit(struct net *net)
2218 {
2219         xt_proto_fini(net, NFPROTO_IPV4);
2220 }
2221
2222 static struct pernet_operations ip_tables_net_ops = {
2223         .init = ip_tables_net_init,
2224         .exit = ip_tables_net_exit,
2225 };
2226
2227 static int __init ip_tables_init(void)
2228 {
2229         int ret;
2230
2231         ret = register_pernet_subsys(&ip_tables_net_ops);
2232         if (ret < 0)
2233                 goto err1;
2234
2235         /* No one else will be downing sem now, so we won't sleep */
2236         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2237         if (ret < 0)
2238                 goto err2;
2239         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2240         if (ret < 0)
2241                 goto err4;
2242
2243         /* Register setsockopt */
2244         ret = nf_register_sockopt(&ipt_sockopts);
2245         if (ret < 0)
2246                 goto err5;
2247
2248         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2249         return 0;
2250
2251 err5:
2252         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2253 err4:
2254         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2255 err2:
2256         unregister_pernet_subsys(&ip_tables_net_ops);
2257 err1:
2258         return ret;
2259 }
2260
2261 static void __exit ip_tables_fini(void)
2262 {
2263         nf_unregister_sockopt(&ipt_sockopts);
2264
2265         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2266         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2267         unregister_pernet_subsys(&ip_tables_net_ops);
2268 }
2269
2270 EXPORT_SYMBOL(ipt_register_table);
2271 EXPORT_SYMBOL(ipt_unregister_table);
2272 EXPORT_SYMBOL(ipt_do_table);
2273 module_init(ip_tables_init);
2274 module_exit(ip_tables_fini);