net/ipv4/netfilter/ip_tables.c

   1 /*
   2  * Packet matching code.
   3  *
   4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   6  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13 #include <linux/cache.h>
  14 #include <linux/capability.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/kmod.h>
  17 #include <linux/vmalloc.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/module.h>
  20 #include <linux/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/compat.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/mutex.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/err.h>
  27 #include <linux/cpumask.h>
  28
  29 #include <linux/netfilter/x_tables.h>
  30 #include <linux/netfilter_ipv4/ip_tables.h>
  31 #include <net/netfilter/nf_log.h>
  32 #include "../../netfilter/xt_repldata.h"
  33
  34 MODULE_LICENSE("GPL");
  35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36 MODULE_DESCRIPTION("IPv4 packet filter");
  37
  38 /*#define DEBUG_IP_FIREWALL*/
  39 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  40 /*#define DEBUG_IP_FIREWALL_USER*/
  41
  42 #ifdef DEBUG_IP_FIREWALL
  43 #define dprintf(format, args...) pr_info(format , ## args)
  44 #else
  45 #define dprintf(format, args...)
  46 #endif
  47
  48 #ifdef DEBUG_IP_FIREWALL_USER
  49 #define duprintf(format, args...) pr_info(format , ## args)
  50 #else
  51 #define duprintf(format, args...)
  52 #endif
  53
  54 #ifdef CONFIG_NETFILTER_DEBUG
  55 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  56 #else
  57 #define IP_NF_ASSERT(x)
  58 #endif
  59
  60 #if 0
  61 /* All the better to debug you with... */
  62 #define static
  63 #define inline
  64 #endif
  65
  66 void *ipt_alloc_initial_table(const struct xt_table *info)
  67 {
  68         return xt_alloc_initial_table(ipt, IPT);
  69 }
  70 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  71
  72 /* Returns whether matches rule or not. */
  73 /* Performance critical - called for every packet */
  74 static inline bool
  75 ip_packet_match(const struct iphdr *ip,
  76                 const char *indev,
  77                 const char *outdev,
  78                 const struct ipt_ip *ipinfo,
  79                 int isfrag)
  80 {
  81         unsigned long ret;
  82
  83 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  84
  85         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  86                   IPT_INV_SRCIP) ||
  87             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  88                   IPT_INV_DSTIP)) {
  89                 dprintf("Source or dest mismatch.\n");
  90
  91                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  92                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  93                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  94                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  95                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  96                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  97                 return false;
  98         }
  99
 100         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 101
 102         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 103                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 104                         indev, ipinfo->iniface,
 105                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
 106                 return false;
 107         }
 108
 109         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 110
 111         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 112                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 113                         outdev, ipinfo->outiface,
 114                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
 115                 return false;
 116         }
 117
 118         /* Check specific protocol */
 119         if (ipinfo->proto &&
 120             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 121                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 122                         ip->protocol, ipinfo->proto,
 123                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
 124                 return false;
 125         }
 126
 127         /* If we have a fragment rule but the packet is not a fragment
 128          * then we return zero */
 129         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 130                 dprintf("Fragment rule but not fragment.%s\n",
 131                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 132                 return false;
 133         }
 134
 135         return true;
 136 }
 137
 138 static bool
 139 ip_checkentry(const struct ipt_ip *ip)
 140 {
 141         if (ip->flags & ~IPT_F_MASK) {
 142                 duprintf("Unknown flag bits set: %08X\n",
 143                          ip->flags & ~IPT_F_MASK);
 144                 return false;
 145         }
 146         if (ip->invflags & ~IPT_INV_MASK) {
 147                 duprintf("Unknown invflag bits set: %08X\n",
 148                          ip->invflags & ~IPT_INV_MASK);
 149                 return false;
 150         }
 151         return true;
 152 }
 153
 154 static unsigned int
 155 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 156 {
 157         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 158
 159         return NF_DROP;
 160 }
 161
 162 /* Performance critical */
 163 static inline struct ipt_entry *
 164 get_entry(const void *base, unsigned int offset)
 165 {
 166         return (struct ipt_entry *)(base + offset);
 167 }
 168
 169 /* All zeroes == unconditional rule. */
 170 /* Mildly perf critical (only if packet tracing is on) */
 171 static inline bool unconditional(const struct ipt_ip *ip)
 172 {
 173         static const struct ipt_ip uncond;
 174
 175         return memcmp(ip, &uncond, sizeof(uncond)) == 0;
 176 #undef FWINV
 177 }
 178
 179 /* for const-correctness */
 180 static inline const struct xt_entry_target *
 181 ipt_get_target_c(const struct ipt_entry *e)
 182 {
 183         return ipt_get_target((struct ipt_entry *)e);
 184 }
 185
 186 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 187 static const char *const hooknames[] = {
 188         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 189         [NF_INET_LOCAL_IN]              = "INPUT",
 190         [NF_INET_FORWARD]               = "FORWARD",
 191         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 192         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 193 };
 194
 195 enum nf_ip_trace_comments {
 196         NF_IP_TRACE_COMMENT_RULE,
 197         NF_IP_TRACE_COMMENT_RETURN,
 198         NF_IP_TRACE_COMMENT_POLICY,
 199 };
 200
 201 static const char *const comments[] = {
 202         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 203         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 204         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 205 };
 206
 207 static struct nf_loginfo trace_loginfo = {
 208         .type = NF_LOG_TYPE_LOG,
 209         .u = {
 210                 .log = {
 211                         .level = 4,
 212                         .logflags = NF_LOG_MASK,
 213                 },
 214         },
 215 };
 216
 217 /* Mildly perf critical (only if packet tracing is on) */
 218 static inline int
 219 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 220                       const char *hookname, const char **chainname,
 221                       const char **comment, unsigned int *rulenum)
 222 {
 223         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 224
 225         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 226                 /* Head of user chain: ERROR target with chainname */
 227                 *chainname = t->target.data;
 228                 (*rulenum) = 0;
 229         } else if (s == e) {
 230                 (*rulenum)++;
 231
 232                 if (s->target_offset == sizeof(struct ipt_entry) &&
 233                     strcmp(t->target.u.kernel.target->name,
 234                            XT_STANDARD_TARGET) == 0 &&
 235                    t->verdict < 0 &&
 236                    unconditional(&s->ip)) {
 237                         /* Tail of chains: STANDARD target (return/policy) */
 238                         *comment = *chainname == hookname
 239                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 240                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 241                 }
 242                 return 1;
 243         } else
 244                 (*rulenum)++;
 245
 246         return 0;
 247 }
 248
 249 static void trace_packet(const struct sk_buff *skb,
 250                          unsigned int hook,
 251                          const struct net_device *in,
 252                          const struct net_device *out,
 253                          const char *tablename,
 254                          const struct xt_table_info *private,
 255                          const struct ipt_entry *e)
 256 {
 257         const void *table_base;
 258         const struct ipt_entry *root;
 259         const char *hookname, *chainname, *comment;
 260         const struct ipt_entry *iter;
 261         unsigned int rulenum = 0;
 262         struct net *net = dev_net(in ? in : out);
 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(net, 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 IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 366                 /* The packet is traced: log it */
 367                 if (unlikely(skb->nf_trace))
 368                         trace_packet(skb, hook, in, out,
 369                                      table->name, private, e);
 370 #endif
 371                 /* Standard target? */
 372                 if (!t->u.kernel.target->target) {
 373                         int v;
 374
 375                         v = ((struct xt_standard_target *)t)->verdict;
 376                         if (v < 0) {
 377                                 /* Pop from stack? */
 378                                 if (v != XT_RETURN) {
 379                                         verdict = (unsigned int)(-v) - 1;
 380                                         break;
 381                                 }
 382                                 if (*stackptr <= origptr) {
 383                                         e = get_entry(table_base,
 384                                             private->underflow[hook]);
 385                                         pr_debug("Underflow (this is normal) "
 386                                                  "to %p\n", e);
 387                                 } else {
 388                                         e = jumpstack[--*stackptr];
 389                                         pr_debug("Pulled %p out from pos %u\n",
 390                                                  e, *stackptr);
 391                                         e = ipt_next_entry(e);
 392                                 }
 393                                 continue;
 394                         }
 395                         if (table_base + v != ipt_next_entry(e) &&
 396                             !(e->ip.flags & IPT_F_GOTO)) {
 397                                 if (*stackptr >= private->stacksize) {
 398                                         verdict = NF_DROP;
 399                                         break;
 400                                 }
 401                                 jumpstack[(*stackptr)++] = e;
 402                                 pr_debug("Pushed %p into pos %u\n",
 403                                          e, *stackptr - 1);
 404                         }
 405
 406                         e = get_entry(table_base, v);
 407                         continue;
 408                 }
 409
 410                 acpar.target   = t->u.kernel.target;
 411                 acpar.targinfo = t->data;
 412
 413                 verdict = t->u.kernel.target->target(skb, &acpar);
 414                 /* Target might have changed stuff. */
 415                 ip = ip_hdr(skb);
 416                 if (verdict == XT_CONTINUE)
 417                         e = ipt_next_entry(e);
 418                 else
 419                         /* Verdict */
 420                         break;
 421         } while (!acpar.hotdrop);
 422         pr_debug("Exiting %s; resetting sp from %u to %u\n",
 423                  __func__, *stackptr, origptr);
 424         *stackptr = origptr;
 425         xt_write_recseq_end(addend);
 426         local_bh_enable();
 427
 428 #ifdef DEBUG_ALLOW_ALL
 429         return NF_ACCEPT;
 430 #else
 431         if (acpar.hotdrop)
 432                 return NF_DROP;
 433         else return verdict;
 434 #endif
 435 }
 436
 437 /* Figures out from what hook each rule can be called: returns 0 if
 438    there are loops.  Puts hook bitmask in comefrom. */
 439 static int
 440 mark_source_chains(const struct xt_table_info *newinfo,
 441                    unsigned int valid_hooks, void *entry0)
 442 {
 443         unsigned int hook;
 444
 445         /* No recursion; use packet counter to save back ptrs (reset
 446            to 0 as we leave), and comefrom to save source hook bitmask */
 447         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 448                 unsigned int pos = newinfo->hook_entry[hook];
 449                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 450
 451                 if (!(valid_hooks & (1 << hook)))
 452                         continue;
 453
 454                 /* Set initial back pointer. */
 455                 e->counters.pcnt = pos;
 456
 457                 for (;;) {
 458                         const struct xt_standard_target *t
 459                                 = (void *)ipt_get_target_c(e);
 460                         int visited = e->comefrom & (1 << hook);
 461
 462                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 463                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 464                                        hook, pos, e->comefrom);
 465                                 return 0;
 466                         }
 467                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 468
 469                         /* Unconditional return/END. */
 470                         if ((e->target_offset == sizeof(struct ipt_entry) &&
 471                              (strcmp(t->target.u.user.name,
 472                                      XT_STANDARD_TARGET) == 0) &&
 473                              t->verdict < 0 && unconditional(&e->ip)) ||
 474                             visited) {
 475                                 unsigned int oldpos, size;
 476
 477                                 if ((strcmp(t->target.u.user.name,
 478                                             XT_STANDARD_TARGET) == 0) &&
 479                                     t->verdict < -NF_MAX_VERDICT - 1) {
 480                                         duprintf("mark_source_chains: bad "
 481                                                 "negative verdict (%i)\n",
 482                                                                 t->verdict);
 483                                         return 0;
 484                                 }
 485
 486                                 /* Return: backtrack through the last
 487                                    big jump. */
 488                                 do {
 489                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 490 #ifdef DEBUG_IP_FIREWALL_USER
 491                                         if (e->comefrom
 492                                             & (1 << NF_INET_NUMHOOKS)) {
 493                                                 duprintf("Back unset "
 494                                                          "on hook %u "
 495                                                          "rule %u\n",
 496                                                          hook, pos);
 497                                         }
 498 #endif
 499                                         oldpos = pos;
 500                                         pos = e->counters.pcnt;
 501                                         e->counters.pcnt = 0;
 502
 503                                         /* We're at the start. */
 504                                         if (pos == oldpos)
 505                                                 goto next;
 506
 507                                         e = (struct ipt_entry *)
 508                                                 (entry0 + pos);
 509                                 } while (oldpos == pos + e->next_offset);
 510
 511                                 /* Move along one */
 512                                 size = e->next_offset;
 513                                 e = (struct ipt_entry *)
 514                                         (entry0 + pos + size);
 515                                 e->counters.pcnt = pos;
 516                                 pos += size;
 517                         } else {
 518                                 int newpos = t->verdict;
 519
 520                                 if (strcmp(t->target.u.user.name,
 521                                            XT_STANDARD_TARGET) == 0 &&
 522                                     newpos >= 0) {
 523                                         if (newpos > newinfo->size -
 524                                                 sizeof(struct ipt_entry)) {
 525                                                 duprintf("mark_source_chains: "
 526                                                         "bad verdict (%i)\n",
 527                                                                 newpos);
 528                                                 return 0;
 529                                         }
 530                                         /* This a jump; chase it. */
 531                                         duprintf("Jump rule %u -> %u\n",
 532                                                  pos, newpos);
 533                                 } else {
 534                                         /* ... this is a fallthru */
 535                                         newpos = pos + e->next_offset;
 536                                 }
 537                                 e = (struct ipt_entry *)
 538                                         (entry0 + newpos);
 539                                 e->counters.pcnt = pos;
 540                                 pos = newpos;
 541                         }
 542                 }
 543                 next:
 544                 duprintf("Finished chain %u\n", hook);
 545         }
 546         return 1;
 547 }
 548
 549 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 550 {
 551         struct xt_mtdtor_param par;
 552
 553         par.net       = net;
 554         par.match     = m->u.kernel.match;
 555         par.matchinfo = m->data;
 556         par.family    = NFPROTO_IPV4;
 557         if (par.match->destroy != NULL)
 558                 par.match->destroy(&par);
 559         module_put(par.match->me);
 560 }
 561
 562 static int
 563 check_entry(const struct ipt_entry *e, const char *name)
 564 {
 565         const struct xt_entry_target *t;
 566
 567         if (!ip_checkentry(&e->ip)) {
 568                 duprintf("ip check failed %p %s.\n", e, name);
 569                 return -EINVAL;
 570         }
 571
 572         if (e->target_offset + sizeof(struct xt_entry_target) >
 573             e->next_offset)
 574                 return -EINVAL;
 575
 576         t = ipt_get_target_c(e);
 577         if (e->target_offset + t->u.target_size > e->next_offset)
 578                 return -EINVAL;
 579
 580         return 0;
 581 }
 582
 583 static int
 584 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 585 {
 586         const struct ipt_ip *ip = par->entryinfo;
 587         int ret;
 588
 589         par->match     = m->u.kernel.match;
 590         par->matchinfo = m->data;
 591
 592         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 593               ip->proto, ip->invflags & IPT_INV_PROTO);
 594         if (ret < 0) {
 595                 duprintf("check failed for `%s'.\n", par->match->name);
 596                 return ret;
 597         }
 598         return 0;
 599 }
 600
 601 static int
 602 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 603 {
 604         struct xt_match *match;
 605         int ret;
 606
 607         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 608                                       m->u.user.revision);
 609         if (IS_ERR(match)) {
 610                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 611                 return PTR_ERR(match);
 612         }
 613         m->u.kernel.match = match;
 614
 615         ret = check_match(m, par);
 616         if (ret)
 617                 goto err;
 618
 619         return 0;
 620 err:
 621         module_put(m->u.kernel.match->me);
 622         return ret;
 623 }
 624
 625 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 626 {
 627         struct xt_entry_target *t = ipt_get_target(e);
 628         struct xt_tgchk_param par = {
 629                 .net       = net,
 630                 .table     = name,
 631                 .entryinfo = e,
 632                 .target    = t->u.kernel.target,
 633                 .targinfo  = t->data,
 634                 .hook_mask = e->comefrom,
 635                 .family    = NFPROTO_IPV4,
 636         };
 637         int ret;
 638
 639         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 640               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 641         if (ret < 0) {
 642                 duprintf("check failed for `%s'.\n",
 643                          t->u.kernel.target->name);
 644                 return ret;
 645         }
 646         return 0;
 647 }
 648
 649 static int
 650 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 651                  unsigned int size)
 652 {
 653         struct xt_entry_target *t;
 654         struct xt_target *target;
 655         int ret;
 656         unsigned int j;
 657         struct xt_mtchk_param mtpar;
 658         struct xt_entry_match *ematch;
 659
 660         ret = check_entry(e, name);
 661         if (ret)
 662                 return ret;
 663
 664         j = 0;
 665         mtpar.net       = net;
 666         mtpar.table     = name;
 667         mtpar.entryinfo = &e->ip;
 668         mtpar.hook_mask = e->comefrom;
 669         mtpar.family    = NFPROTO_IPV4;
 670         xt_ematch_foreach(ematch, e) {
 671                 ret = find_check_match(ematch, &mtpar);
 672                 if (ret != 0)
 673                         goto cleanup_matches;
 674                 ++j;
 675         }
 676
 677         t = ipt_get_target(e);
 678         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 679                                         t->u.user.revision);
 680         if (IS_ERR(target)) {
 681                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 682                 ret = PTR_ERR(target);
 683                 goto cleanup_matches;
 684         }
 685         t->u.kernel.target = target;
 686
 687         ret = check_target(e, net, name);
 688         if (ret)
 689                 goto err;
 690         return 0;
 691  err:
 692         module_put(t->u.kernel.target->me);
 693  cleanup_matches:
 694         xt_ematch_foreach(ematch, e) {
 695                 if (j-- == 0)
 696                         break;
 697                 cleanup_match(ematch, net);
 698         }
 699         return ret;
 700 }
 701
 702 static bool check_underflow(const struct ipt_entry *e)
 703 {
 704         const struct xt_entry_target *t;
 705         unsigned int verdict;
 706
 707         if (!unconditional(&e->ip))
 708                 return false;
 709         t = ipt_get_target_c(e);
 710         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 711                 return false;
 712         verdict = ((struct xt_standard_target *)t)->verdict;
 713         verdict = -verdict - 1;
 714         return verdict == NF_DROP || verdict == NF_ACCEPT;
 715 }
 716
 717 static int
 718 check_entry_size_and_hooks(struct ipt_entry *e,
 719                            struct xt_table_info *newinfo,
 720                            const unsigned char *base,
 721                            const unsigned char *limit,
 722                            const unsigned int *hook_entries,
 723                            const unsigned int *underflows,
 724                            unsigned int valid_hooks)
 725 {
 726         unsigned int h;
 727
 728         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 729             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 730                 duprintf("Bad offset %p\n", e);
 731                 return -EINVAL;
 732         }
 733
 734         if (e->next_offset
 735             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 736                 duprintf("checking: element %p size %u\n",
 737                          e, e->next_offset);
 738                 return -EINVAL;
 739         }
 740
 741         /* Check hooks & underflows */
 742         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 743                 if (!(valid_hooks & (1 << h)))
 744                         continue;
 745                 if ((unsigned char *)e - base == hook_entries[h])
 746                         newinfo->hook_entry[h] = hook_entries[h];
 747                 if ((unsigned char *)e - base == underflows[h]) {
 748                         if (!check_underflow(e)) {
 749                                 pr_err("Underflows must be unconditional and "
 750                                        "use the STANDARD target with "
 751                                        "ACCEPT/DROP\n");
 752                                 return -EINVAL;
 753                         }
 754                         newinfo->underflow[h] = underflows[h];
 755                 }
 756         }
 757
 758         /* Clear counters and comefrom */
 759         e->counters = ((struct xt_counters) { 0, 0 });
 760         e->comefrom = 0;
 761         return 0;
 762 }
 763
 764 static void
 765 cleanup_entry(struct ipt_entry *e, struct net *net)
 766 {
 767         struct xt_tgdtor_param par;
 768         struct xt_entry_target *t;
 769         struct xt_entry_match *ematch;
 770
 771         /* Cleanup all matches */
 772         xt_ematch_foreach(ematch, e)
 773                 cleanup_match(ematch, net);
 774         t = ipt_get_target(e);
 775
 776         par.net      = net;
 777         par.target   = t->u.kernel.target;
 778         par.targinfo = t->data;
 779         par.family   = NFPROTO_IPV4;
 780         if (par.target->destroy != NULL)
 781                 par.target->destroy(&par);
 782         module_put(par.target->me);
 783 }
 784
 785 /* Checks and translates the user-supplied table segment (held in
 786    newinfo) */
 787 static int
 788 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 789                 const struct ipt_replace *repl)
 790 {
 791         struct ipt_entry *iter;
 792         unsigned int i;
 793         int ret = 0;
 794
 795         newinfo->size = repl->size;
 796         newinfo->number = repl->num_entries;
 797
 798         /* Init all hooks to impossible value. */
 799         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 800                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 801                 newinfo->underflow[i] = 0xFFFFFFFF;
 802         }
 803
 804         duprintf("translate_table: size %u\n", newinfo->size);
 805         i = 0;
 806         /* Walk through entries, checking offsets. */
 807         xt_entry_foreach(iter, entry0, newinfo->size) {
 808                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 809                                                  entry0 + repl->size,
 810                                                  repl->hook_entry,
 811                                                  repl->underflow,
 812                                                  repl->valid_hooks);
 813                 if (ret != 0)
 814                         return ret;
 815                 ++i;
 816                 if (strcmp(ipt_get_target(iter)->u.user.name,
 817                     XT_ERROR_TARGET) == 0)
 818                         ++newinfo->stacksize;
 819         }
 820
 821         if (i != repl->num_entries) {
 822                 duprintf("translate_table: %u not %u entries\n",
 823                          i, repl->num_entries);
 824                 return -EINVAL;
 825         }
 826
 827         /* Check hooks all assigned */
 828         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 829                 /* Only hooks which are valid */
 830                 if (!(repl->valid_hooks & (1 << i)))
 831                         continue;
 832                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 833                         duprintf("Invalid hook entry %u %u\n",
 834                                  i, repl->hook_entry[i]);
 835                         return -EINVAL;
 836                 }
 837                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 838                         duprintf("Invalid underflow %u %u\n",
 839                                  i, repl->underflow[i]);
 840                         return -EINVAL;
 841                 }
 842         }
 843
 844         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 845                 return -ELOOP;
 846
 847         /* Finally, each sanity check must pass */
 848         i = 0;
 849         xt_entry_foreach(iter, entry0, newinfo->size) {
 850                 ret = find_check_entry(iter, net, repl->name, repl->size);
 851                 if (ret != 0)
 852                         break;
 853                 ++i;
 854         }
 855
 856         if (ret != 0) {
 857                 xt_entry_foreach(iter, entry0, newinfo->size) {
 858                         if (i-- == 0)
 859                                 break;
 860                         cleanup_entry(iter, net);
 861                 }
 862                 return ret;
 863         }
 864
 865         /* And one copy for every other CPU */
 866         for_each_possible_cpu(i) {
 867                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
 868                         memcpy(newinfo->entries[i], entry0, newinfo->size);
 869         }
 870
 871         return ret;
 872 }
 873
 874 static void
 875 get_counters(const struct xt_table_info *t,
 876              struct xt_counters counters[])
 877 {
 878         struct ipt_entry *iter;
 879         unsigned int cpu;
 880         unsigned int i;
 881
 882         for_each_possible_cpu(cpu) {
 883                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 884
 885                 i = 0;
 886                 xt_entry_foreach(iter, t->entries[cpu], t->size) {
 887                         u64 bcnt, pcnt;
 888                         unsigned int start;
 889
 890                         do {
 891                                 start = read_seqcount_begin(s);
 892                                 bcnt = iter->counters.bcnt;
 893                                 pcnt = iter->counters.pcnt;
 894                         } while (read_seqcount_retry(s, start));
 895
 896                         ADD_COUNTER(counters[i], bcnt, pcnt);
 897                         ++i; /* macro does multi eval of i */
 898                 }
 899         }
 900 }
 901
 902 static struct xt_counters *alloc_counters(const struct xt_table *table)
 903 {
 904         unsigned int countersize;
 905         struct xt_counters *counters;
 906         const struct xt_table_info *private = table->private;
 907
 908         /* We need atomic snapshot of counters: rest doesn't change
 909            (other than comefrom, which userspace doesn't care
 910            about). */
 911         countersize = sizeof(struct xt_counters) * private->number;
 912         counters = vzalloc(countersize);
 913
 914         if (counters == NULL)
 915                 return ERR_PTR(-ENOMEM);
 916
 917         get_counters(private, counters);
 918
 919         return counters;
 920 }
 921
 922 static int
 923 copy_entries_to_user(unsigned int total_size,
 924                      const struct xt_table *table,
 925                      void __user *userptr)
 926 {
 927         unsigned int off, num;
 928         const struct ipt_entry *e;
 929         struct xt_counters *counters;
 930         const struct xt_table_info *private = table->private;
 931         int ret = 0;
 932         const void *loc_cpu_entry;
 933
 934         counters = alloc_counters(table);
 935         if (IS_ERR(counters))
 936                 return PTR_ERR(counters);
 937
 938         /* choose the copy that is on our node/cpu, ...
 939          * This choice is lazy (because current thread is
 940          * allowed to migrate to another cpu)
 941          */
 942         loc_cpu_entry = private->entries[raw_smp_processor_id()];
 943         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 944                 ret = -EFAULT;
 945                 goto free_counters;
 946         }
 947
 948         /* FIXME: use iterator macros --RR */
 949         /* ... then go back and fix counters and names */
 950         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 951                 unsigned int i;
 952                 const struct xt_entry_match *m;
 953                 const struct xt_entry_target *t;
 954
 955                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 956                 if (copy_to_user(userptr + off
 957                                  + offsetof(struct ipt_entry, counters),
 958                                  &counters[num],
 959                                  sizeof(counters[num])) != 0) {
 960                         ret = -EFAULT;
 961                         goto free_counters;
 962                 }
 963
 964                 for (i = sizeof(struct ipt_entry);
 965                      i < e->target_offset;
 966                      i += m->u.match_size) {
 967                         m = (void *)e + i;
 968
 969                         if (copy_to_user(userptr + off + i
 970                                          + offsetof(struct xt_entry_match,
 971                                                     u.user.name),
 972                                          m->u.kernel.match->name,
 973                                          strlen(m->u.kernel.match->name)+1)
 974                             != 0) {
 975                                 ret = -EFAULT;
 976                                 goto free_counters;
 977                         }
 978                 }
 979
 980                 t = ipt_get_target_c(e);
 981                 if (copy_to_user(userptr + off + e->target_offset
 982                                  + offsetof(struct xt_entry_target,
 983                                             u.user.name),
 984                                  t->u.kernel.target->name,
 985                                  strlen(t->u.kernel.target->name)+1) != 0) {
 986                         ret = -EFAULT;
 987                         goto free_counters;
 988                 }
 989         }
 990
 991  free_counters:
 992         vfree(counters);
 993         return ret;
 994 }
 995
 996 #ifdef CONFIG_COMPAT
 997 static void compat_standard_from_user(void *dst, const void *src)
 998 {
 999         int v = *(compat_int_t *)src;
1000
1001         if (v > 0)
1002                 v += xt_compat_calc_jump(AF_INET, v);
1003         memcpy(dst, &v, sizeof(v));
1004 }
1005
1006 static int compat_standard_to_user(void __user *dst, const void *src)
1007 {
1008         compat_int_t cv = *(int *)src;
1009
1010         if (cv > 0)
1011                 cv -= xt_compat_calc_jump(AF_INET, cv);
1012         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1013 }
1014
1015 static int compat_calc_entry(const struct ipt_entry *e,
1016                              const struct xt_table_info *info,
1017                              const void *base, struct xt_table_info *newinfo)
1018 {
1019         const struct xt_entry_match *ematch;
1020         const struct xt_entry_target *t;
1021         unsigned int entry_offset;
1022         int off, i, ret;
1023
1024         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1025         entry_offset = (void *)e - base;
1026         xt_ematch_foreach(ematch, e)
1027                 off += xt_compat_match_offset(ematch->u.kernel.match);
1028         t = ipt_get_target_c(e);
1029         off += xt_compat_target_offset(t->u.kernel.target);
1030         newinfo->size -= off;
1031         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1032         if (ret)
1033                 return ret;
1034
1035         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1036                 if (info->hook_entry[i] &&
1037                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1038                         newinfo->hook_entry[i] -= off;
1039                 if (info->underflow[i] &&
1040                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1041                         newinfo->underflow[i] -= off;
1042         }
1043         return 0;
1044 }
1045
1046 static int compat_table_info(const struct xt_table_info *info,
1047                              struct xt_table_info *newinfo)
1048 {
1049         struct ipt_entry *iter;
1050         void *loc_cpu_entry;
1051         int ret;
1052
1053         if (!newinfo || !info)
1054                 return -EINVAL;
1055
1056         /* we dont care about newinfo->entries[] */
1057         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1058         newinfo->initial_entries = 0;
1059         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1060         xt_compat_init_offsets(AF_INET, info->number);
1061         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1062                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1063                 if (ret != 0)
1064                         return ret;
1065         }
1066         return 0;
1067 }
1068 #endif
1069
1070 static int get_info(struct net *net, void __user *user,
1071                     const int *len, int compat)
1072 {
1073         char name[XT_TABLE_MAXNAMELEN];
1074         struct xt_table *t;
1075         int ret;
1076
1077         if (*len != sizeof(struct ipt_getinfo)) {
1078                 duprintf("length %u != %zu\n", *len,
1079                          sizeof(struct ipt_getinfo));
1080                 return -EINVAL;
1081         }
1082
1083         if (copy_from_user(name, user, sizeof(name)) != 0)
1084                 return -EFAULT;
1085
1086         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1087 #ifdef CONFIG_COMPAT
1088         if (compat)
1089                 xt_compat_lock(AF_INET);
1090 #endif
1091         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1092                                     "iptable_%s", name);
1093         if (!IS_ERR_OR_NULL(t)) {
1094                 struct ipt_getinfo info;
1095                 const struct xt_table_info *private = t->private;
1096 #ifdef CONFIG_COMPAT
1097                 struct xt_table_info tmp;
1098
1099                 if (compat) {
1100                         ret = compat_table_info(private, &tmp);
1101                         xt_compat_flush_offsets(AF_INET);
1102                         private = &tmp;
1103                 }
1104 #endif
1105                 memset(&info, 0, sizeof(info));
1106                 info.valid_hooks = t->valid_hooks;
1107                 memcpy(info.hook_entry, private->hook_entry,
1108                        sizeof(info.hook_entry));
1109                 memcpy(info.underflow, private->underflow,
1110                        sizeof(info.underflow));
1111                 info.num_entries = private->number;
1112                 info.size = private->size;
1113                 strcpy(info.name, name);
1114
1115                 if (copy_to_user(user, &info, *len) != 0)
1116                         ret = -EFAULT;
1117                 else
1118                         ret = 0;
1119
1120                 xt_table_unlock(t);
1121                 module_put(t->me);
1122         } else
1123                 ret = t ? PTR_ERR(t) : -ENOENT;
1124 #ifdef CONFIG_COMPAT
1125         if (compat)
1126                 xt_compat_unlock(AF_INET);
1127 #endif
1128         return ret;
1129 }
1130
1131 static int
1132 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1133             const int *len)
1134 {
1135         int ret;
1136         struct ipt_get_entries get;
1137         struct xt_table *t;
1138
1139         if (*len < sizeof(get)) {
1140                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1141                 return -EINVAL;
1142         }
1143         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1144                 return -EFAULT;
1145         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1146                 duprintf("get_entries: %u != %zu\n",
1147                          *len, sizeof(get) + get.size);
1148                 return -EINVAL;
1149         }
1150
1151         t = xt_find_table_lock(net, AF_INET, get.name);
1152         if (!IS_ERR_OR_NULL(t)) {
1153                 const struct xt_table_info *private = t->private;
1154                 duprintf("t->private->number = %u\n", private->number);
1155                 if (get.size == private->size)
1156                         ret = copy_entries_to_user(private->size,
1157                                                    t, uptr->entrytable);
1158                 else {
1159                         duprintf("get_entries: I've got %u not %u!\n",
1160                                  private->size, get.size);
1161                         ret = -EAGAIN;
1162                 }
1163                 module_put(t->me);
1164                 xt_table_unlock(t);
1165         } else
1166                 ret = t ? PTR_ERR(t) : -ENOENT;
1167
1168         return ret;
1169 }
1170
1171 static int
1172 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1173              struct xt_table_info *newinfo, unsigned int num_counters,
1174              void __user *counters_ptr)
1175 {
1176         int ret;
1177         struct xt_table *t;
1178         struct xt_table_info *oldinfo;
1179         struct xt_counters *counters;
1180         void *loc_cpu_old_entry;
1181         struct ipt_entry *iter;
1182
1183         ret = 0;
1184         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1185         if (!counters) {
1186                 ret = -ENOMEM;
1187                 goto out;
1188         }
1189
1190         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1191                                     "iptable_%s", name);
1192         if (IS_ERR_OR_NULL(t)) {
1193                 ret = t ? PTR_ERR(t) : -ENOENT;
1194                 goto free_newinfo_counters_untrans;
1195         }
1196
1197         /* You lied! */
1198         if (valid_hooks != t->valid_hooks) {
1199                 duprintf("Valid hook crap: %08X vs %08X\n",
1200                          valid_hooks, t->valid_hooks);
1201                 ret = -EINVAL;
1202                 goto put_module;
1203         }
1204
1205         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1206         if (!oldinfo)
1207                 goto put_module;
1208
1209         /* Update module usage count based on number of rules */
1210         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1211                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1212         if ((oldinfo->number > oldinfo->initial_entries) ||
1213             (newinfo->number <= oldinfo->initial_entries))
1214                 module_put(t->me);
1215         if ((oldinfo->number > oldinfo->initial_entries) &&
1216             (newinfo->number <= oldinfo->initial_entries))
1217                 module_put(t->me);
1218
1219         /* Get the old counters, and synchronize with replace */
1220         get_counters(oldinfo, counters);
1221
1222         /* Decrease module usage counts and free resource */
1223         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1224         xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1225                 cleanup_entry(iter, net);
1226
1227         xt_free_table_info(oldinfo);
1228         if (copy_to_user(counters_ptr, counters,
1229                          sizeof(struct xt_counters) * num_counters) != 0) {
1230                 /* Silent error, can't fail, new table is already in place */
1231                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1232         }
1233         vfree(counters);
1234         xt_table_unlock(t);
1235         return ret;
1236
1237  put_module:
1238         module_put(t->me);
1239         xt_table_unlock(t);
1240  free_newinfo_counters_untrans:
1241         vfree(counters);
1242  out:
1243         return ret;
1244 }
1245
1246 static int
1247 do_replace(struct net *net, const void __user *user, unsigned int len)
1248 {
1249         int ret;
1250         struct ipt_replace tmp;
1251         struct xt_table_info *newinfo;
1252         void *loc_cpu_entry;
1253         struct ipt_entry *iter;
1254
1255         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1256                 return -EFAULT;
1257
1258         /* overflow check */
1259         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1260                 return -ENOMEM;
1261         tmp.name[sizeof(tmp.name)-1] = 0;
1262
1263         newinfo = xt_alloc_table_info(tmp.size);
1264         if (!newinfo)
1265                 return -ENOMEM;
1266
1267         /* choose the copy that is on our node/cpu */
1268         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1269         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1270                            tmp.size) != 0) {
1271                 ret = -EFAULT;
1272                 goto free_newinfo;
1273         }
1274
1275         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1276         if (ret != 0)
1277                 goto free_newinfo;
1278
1279         duprintf("Translated table\n");
1280
1281         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1282                            tmp.num_counters, tmp.counters);
1283         if (ret)
1284                 goto free_newinfo_untrans;
1285         return 0;
1286
1287  free_newinfo_untrans:
1288         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1289                 cleanup_entry(iter, net);
1290  free_newinfo:
1291         xt_free_table_info(newinfo);
1292         return ret;
1293 }
1294
1295 static int
1296 do_add_counters(struct net *net, const void __user *user,
1297                 unsigned int len, int compat)
1298 {
1299         unsigned int i, curcpu;
1300         struct xt_counters_info tmp;
1301         struct xt_counters *paddc;
1302         unsigned int num_counters;
1303         const char *name;
1304         int size;
1305         void *ptmp;
1306         struct xt_table *t;
1307         const struct xt_table_info *private;
1308         int ret = 0;
1309         void *loc_cpu_entry;
1310         struct ipt_entry *iter;
1311         unsigned int addend;
1312 #ifdef CONFIG_COMPAT
1313         struct compat_xt_counters_info compat_tmp;
1314
1315         if (compat) {
1316                 ptmp = &compat_tmp;
1317                 size = sizeof(struct compat_xt_counters_info);
1318         } else
1319 #endif
1320         {
1321                 ptmp = &tmp;
1322                 size = sizeof(struct xt_counters_info);
1323         }
1324
1325         if (copy_from_user(ptmp, user, size) != 0)
1326                 return -EFAULT;
1327
1328 #ifdef CONFIG_COMPAT
1329         if (compat) {
1330                 num_counters = compat_tmp.num_counters;
1331                 name = compat_tmp.name;
1332         } else
1333 #endif
1334         {
1335                 num_counters = tmp.num_counters;
1336                 name = tmp.name;
1337         }
1338
1339         if (len != size + num_counters * sizeof(struct xt_counters))
1340                 return -EINVAL;
1341
1342         paddc = vmalloc(len - size);
1343         if (!paddc)
1344                 return -ENOMEM;
1345
1346         if (copy_from_user(paddc, user + size, len - size) != 0) {
1347                 ret = -EFAULT;
1348                 goto free;
1349         }
1350
1351         t = xt_find_table_lock(net, AF_INET, name);
1352         if (IS_ERR_OR_NULL(t)) {
1353                 ret = t ? PTR_ERR(t) : -ENOENT;
1354                 goto free;
1355         }
1356
1357         local_bh_disable();
1358         private = t->private;
1359         if (private->number != num_counters) {
1360                 ret = -EINVAL;
1361                 goto unlock_up_free;
1362         }
1363
1364         i = 0;
1365         /* Choose the copy that is on our node */
1366         curcpu = smp_processor_id();
1367         loc_cpu_entry = private->entries[curcpu];
1368         addend = xt_write_recseq_begin();
1369         xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1370                 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1371                 ++i;
1372         }
1373         xt_write_recseq_end(addend);
1374  unlock_up_free:
1375         local_bh_enable();
1376         xt_table_unlock(t);
1377         module_put(t->me);
1378  free:
1379         vfree(paddc);
1380
1381         return ret;
1382 }
1383
1384 #ifdef CONFIG_COMPAT
1385 struct compat_ipt_replace {
1386         char                    name[XT_TABLE_MAXNAMELEN];
1387         u32                     valid_hooks;
1388         u32                     num_entries;
1389         u32                     size;
1390         u32                     hook_entry[NF_INET_NUMHOOKS];
1391         u32                     underflow[NF_INET_NUMHOOKS];
1392         u32                     num_counters;
1393         compat_uptr_t           counters;       /* struct xt_counters * */
1394         struct compat_ipt_entry entries[0];
1395 };
1396
1397 static int
1398 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1399                           unsigned int *size, struct xt_counters *counters,
1400                           unsigned int i)
1401 {
1402         struct xt_entry_target *t;
1403         struct compat_ipt_entry __user *ce;
1404         u_int16_t target_offset, next_offset;
1405         compat_uint_t origsize;
1406         const struct xt_entry_match *ematch;
1407         int ret = 0;
1408
1409         origsize = *size;
1410         ce = (struct compat_ipt_entry __user *)*dstptr;
1411         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1412             copy_to_user(&ce->counters, &counters[i],
1413             sizeof(counters[i])) != 0)
1414                 return -EFAULT;
1415
1416         *dstptr += sizeof(struct compat_ipt_entry);
1417         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1418
1419         xt_ematch_foreach(ematch, e) {
1420                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1421                 if (ret != 0)
1422                         return ret;
1423         }
1424         target_offset = e->target_offset - (origsize - *size);
1425         t = ipt_get_target(e);
1426         ret = xt_compat_target_to_user(t, dstptr, size);
1427         if (ret)
1428                 return ret;
1429         next_offset = e->next_offset - (origsize - *size);
1430         if (put_user(target_offset, &ce->target_offset) != 0 ||
1431             put_user(next_offset, &ce->next_offset) != 0)
1432                 return -EFAULT;
1433         return 0;
1434 }
1435
1436 static int
1437 compat_find_calc_match(struct xt_entry_match *m,
1438                        const char *name,
1439                        const struct ipt_ip *ip,
1440                        unsigned int hookmask,
1441                        int *size)
1442 {
1443         struct xt_match *match;
1444
1445         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1446                                       m->u.user.revision);
1447         if (IS_ERR(match)) {
1448                 duprintf("compat_check_calc_match: `%s' not found\n",
1449                          m->u.user.name);
1450                 return PTR_ERR(match);
1451         }
1452         m->u.kernel.match = match;
1453         *size += xt_compat_match_offset(match);
1454         return 0;
1455 }
1456
1457 static void compat_release_entry(struct compat_ipt_entry *e)
1458 {
1459         struct xt_entry_target *t;
1460         struct xt_entry_match *ematch;
1461
1462         /* Cleanup all matches */
1463         xt_ematch_foreach(ematch, e)
1464                 module_put(ematch->u.kernel.match->me);
1465         t = compat_ipt_get_target(e);
1466         module_put(t->u.kernel.target->me);
1467 }
1468
1469 static int
1470 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1471                                   struct xt_table_info *newinfo,
1472                                   unsigned int *size,
1473                                   const unsigned char *base,
1474                                   const unsigned char *limit,
1475                                   const unsigned int *hook_entries,
1476                                   const unsigned int *underflows,
1477                                   const char *name)
1478 {
1479         struct xt_entry_match *ematch;
1480         struct xt_entry_target *t;
1481         struct xt_target *target;
1482         unsigned int entry_offset;
1483         unsigned int j;
1484         int ret, off, h;
1485
1486         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1487         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1488             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1489                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1490                 return -EINVAL;
1491         }
1492
1493         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1494                              sizeof(struct compat_xt_entry_target)) {
1495                 duprintf("checking: element %p size %u\n",
1496                          e, e->next_offset);
1497                 return -EINVAL;
1498         }
1499
1500         /* For purposes of check_entry casting the compat entry is fine */
1501         ret = check_entry((struct ipt_entry *)e, name);
1502         if (ret)
1503                 return ret;
1504
1505         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1506         entry_offset = (void *)e - (void *)base;
1507         j = 0;
1508         xt_ematch_foreach(ematch, e) {
1509                 ret = compat_find_calc_match(ematch, name,
1510                                              &e->ip, e->comefrom, &off);
1511                 if (ret != 0)
1512                         goto release_matches;
1513                 ++j;
1514         }
1515
1516         t = compat_ipt_get_target(e);
1517         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1518                                         t->u.user.revision);
1519         if (IS_ERR(target)) {
1520                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1521                          t->u.user.name);
1522                 ret = PTR_ERR(target);
1523                 goto release_matches;
1524         }
1525         t->u.kernel.target = target;
1526
1527         off += xt_compat_target_offset(target);
1528         *size += off;
1529         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1530         if (ret)
1531                 goto out;
1532
1533         /* Check hooks & underflows */
1534         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1535                 if ((unsigned char *)e - base == hook_entries[h])
1536                         newinfo->hook_entry[h] = hook_entries[h];
1537                 if ((unsigned char *)e - base == underflows[h])
1538                         newinfo->underflow[h] = underflows[h];
1539         }
1540
1541         /* Clear counters and comefrom */
1542         memset(&e->counters, 0, sizeof(e->counters));
1543         e->comefrom = 0;
1544         return 0;
1545
1546 out:
1547         module_put(t->u.kernel.target->me);
1548 release_matches:
1549         xt_ematch_foreach(ematch, e) {
1550                 if (j-- == 0)
1551                         break;
1552                 module_put(ematch->u.kernel.match->me);
1553         }
1554         return ret;
1555 }
1556
1557 static int
1558 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1559                             unsigned int *size, const char *name,
1560                             struct xt_table_info *newinfo, unsigned char *base)
1561 {
1562         struct xt_entry_target *t;
1563         struct xt_target *target;
1564         struct ipt_entry *de;
1565         unsigned int origsize;
1566         int ret, h;
1567         struct xt_entry_match *ematch;
1568
1569         ret = 0;
1570         origsize = *size;
1571         de = (struct ipt_entry *)*dstptr;
1572         memcpy(de, e, sizeof(struct ipt_entry));
1573         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1574
1575         *dstptr += sizeof(struct ipt_entry);
1576         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1577
1578         xt_ematch_foreach(ematch, e) {
1579                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1580                 if (ret != 0)
1581                         return ret;
1582         }
1583         de->target_offset = e->target_offset - (origsize - *size);
1584         t = compat_ipt_get_target(e);
1585         target = t->u.kernel.target;
1586         xt_compat_target_from_user(t, dstptr, size);
1587
1588         de->next_offset = e->next_offset - (origsize - *size);
1589         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1590                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1591                         newinfo->hook_entry[h] -= origsize - *size;
1592                 if ((unsigned char *)de - base < newinfo->underflow[h])
1593                         newinfo->underflow[h] -= origsize - *size;
1594         }
1595         return ret;
1596 }
1597
1598 static int
1599 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1600 {
1601         struct xt_entry_match *ematch;
1602         struct xt_mtchk_param mtpar;
1603         unsigned int j;
1604         int ret = 0;
1605
1606         j = 0;
1607         mtpar.net       = net;
1608         mtpar.table     = name;
1609         mtpar.entryinfo = &e->ip;
1610         mtpar.hook_mask = e->comefrom;
1611         mtpar.family    = NFPROTO_IPV4;
1612         xt_ematch_foreach(ematch, e) {
1613                 ret = check_match(ematch, &mtpar);
1614                 if (ret != 0)
1615                         goto cleanup_matches;
1616                 ++j;
1617         }
1618
1619         ret = check_target(e, net, name);
1620         if (ret)
1621                 goto cleanup_matches;
1622         return 0;
1623
1624  cleanup_matches:
1625         xt_ematch_foreach(ematch, e) {
1626                 if (j-- == 0)
1627                         break;
1628                 cleanup_match(ematch, net);
1629         }
1630         return ret;
1631 }
1632
1633 static int
1634 translate_compat_table(struct net *net,
1635                        const char *name,
1636                        unsigned int valid_hooks,
1637                        struct xt_table_info **pinfo,
1638                        void **pentry0,
1639                        unsigned int total_size,
1640                        unsigned int number,
1641                        unsigned int *hook_entries,
1642                        unsigned int *underflows)
1643 {
1644         unsigned int i, j;
1645         struct xt_table_info *newinfo, *info;
1646         void *pos, *entry0, *entry1;
1647         struct compat_ipt_entry *iter0;
1648         struct ipt_entry *iter1;
1649         unsigned int size;
1650         int ret;
1651
1652         info = *pinfo;
1653         entry0 = *pentry0;
1654         size = total_size;
1655         info->number = number;
1656
1657         /* Init all hooks to impossible value. */
1658         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1659                 info->hook_entry[i] = 0xFFFFFFFF;
1660                 info->underflow[i] = 0xFFFFFFFF;
1661         }
1662
1663         duprintf("translate_compat_table: size %u\n", info->size);
1664         j = 0;
1665         xt_compat_lock(AF_INET);
1666         xt_compat_init_offsets(AF_INET, number);
1667         /* Walk through entries, checking offsets. */
1668         xt_entry_foreach(iter0, entry0, total_size) {
1669                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1670                                                         entry0,
1671                                                         entry0 + total_size,
1672                                                         hook_entries,
1673                                                         underflows,
1674                                                         name);
1675                 if (ret != 0)
1676                         goto out_unlock;
1677                 ++j;
1678         }
1679
1680         ret = -EINVAL;
1681         if (j != number) {
1682                 duprintf("translate_compat_table: %u not %u entries\n",
1683                          j, number);
1684                 goto out_unlock;
1685         }
1686
1687         /* Check hooks all assigned */
1688         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1689                 /* Only hooks which are valid */
1690                 if (!(valid_hooks & (1 << i)))
1691                         continue;
1692                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1693                         duprintf("Invalid hook entry %u %u\n",
1694                                  i, hook_entries[i]);
1695                         goto out_unlock;
1696                 }
1697                 if (info->underflow[i] == 0xFFFFFFFF) {
1698                         duprintf("Invalid underflow %u %u\n",
1699                                  i, underflows[i]);
1700                         goto out_unlock;
1701                 }
1702         }
1703
1704         ret = -ENOMEM;
1705         newinfo = xt_alloc_table_info(size);
1706         if (!newinfo)
1707                 goto out_unlock;
1708
1709         newinfo->number = number;
1710         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1711                 newinfo->hook_entry[i] = info->hook_entry[i];
1712                 newinfo->underflow[i] = info->underflow[i];
1713         }
1714         entry1 = newinfo->entries[raw_smp_processor_id()];
1715         pos = entry1;
1716         size = total_size;
1717         xt_entry_foreach(iter0, entry0, total_size) {
1718                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1719                                                   name, newinfo, entry1);
1720                 if (ret != 0)
1721                         break;
1722         }
1723         xt_compat_flush_offsets(AF_INET);
1724         xt_compat_unlock(AF_INET);
1725         if (ret)
1726                 goto free_newinfo;
1727
1728         ret = -ELOOP;
1729         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1730                 goto free_newinfo;
1731
1732         i = 0;
1733         xt_entry_foreach(iter1, entry1, newinfo->size) {
1734                 ret = compat_check_entry(iter1, net, name);
1735                 if (ret != 0)
1736                         break;
1737                 ++i;
1738                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1739                     XT_ERROR_TARGET) == 0)
1740                         ++newinfo->stacksize;
1741         }
1742         if (ret) {
1743                 /*
1744                  * The first i matches need cleanup_entry (calls ->destroy)
1745                  * because they had called ->check already. The other j-i
1746                  * entries need only release.
1747                  */
1748                 int skip = i;
1749                 j -= i;
1750                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1751                         if (skip-- > 0)
1752                                 continue;
1753                         if (j-- == 0)
1754                                 break;
1755                         compat_release_entry(iter0);
1756                 }
1757                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1758                         if (i-- == 0)
1759                                 break;
1760                         cleanup_entry(iter1, net);
1761                 }
1762                 xt_free_table_info(newinfo);
1763                 return ret;
1764         }
1765
1766         /* And one copy for every other CPU */
1767         for_each_possible_cpu(i)
1768                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1769                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1770
1771         *pinfo = newinfo;
1772         *pentry0 = entry1;
1773         xt_free_table_info(info);
1774         return 0;
1775
1776 free_newinfo:
1777         xt_free_table_info(newinfo);
1778 out:
1779         xt_entry_foreach(iter0, entry0, total_size) {
1780                 if (j-- == 0)
1781                         break;
1782                 compat_release_entry(iter0);
1783         }
1784         return ret;
1785 out_unlock:
1786         xt_compat_flush_offsets(AF_INET);
1787         xt_compat_unlock(AF_INET);
1788         goto out;
1789 }
1790
1791 static int
1792 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1793 {
1794         int ret;
1795         struct compat_ipt_replace tmp;
1796         struct xt_table_info *newinfo;
1797         void *loc_cpu_entry;
1798         struct ipt_entry *iter;
1799
1800         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1801                 return -EFAULT;
1802
1803         /* overflow check */
1804         if (tmp.size >= INT_MAX / num_possible_cpus())
1805                 return -ENOMEM;
1806         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1807                 return -ENOMEM;
1808         tmp.name[sizeof(tmp.name)-1] = 0;
1809
1810         newinfo = xt_alloc_table_info(tmp.size);
1811         if (!newinfo)
1812                 return -ENOMEM;
1813
1814         /* choose the copy that is on our node/cpu */
1815         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1816         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1817                            tmp.size) != 0) {
1818                 ret = -EFAULT;
1819                 goto free_newinfo;
1820         }
1821
1822         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1823                                      &newinfo, &loc_cpu_entry, tmp.size,
1824                                      tmp.num_entries, tmp.hook_entry,
1825                                      tmp.underflow);
1826         if (ret != 0)
1827                 goto free_newinfo;
1828
1829         duprintf("compat_do_replace: Translated table\n");
1830
1831         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1832                            tmp.num_counters, compat_ptr(tmp.counters));
1833         if (ret)
1834                 goto free_newinfo_untrans;
1835         return 0;
1836
1837  free_newinfo_untrans:
1838         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1839                 cleanup_entry(iter, net);
1840  free_newinfo:
1841         xt_free_table_info(newinfo);
1842         return ret;
1843 }
1844
1845 static int
1846 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1847                       unsigned int len)
1848 {
1849         int ret;
1850
1851         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1852                 return -EPERM;
1853
1854         switch (cmd) {
1855         case IPT_SO_SET_REPLACE:
1856                 ret = compat_do_replace(sock_net(sk), user, len);
1857                 break;
1858
1859         case IPT_SO_SET_ADD_COUNTERS:
1860                 ret = do_add_counters(sock_net(sk), user, len, 1);
1861                 break;
1862
1863         default:
1864                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1865                 ret = -EINVAL;
1866         }
1867
1868         return ret;
1869 }
1870
1871 struct compat_ipt_get_entries {
1872         char name[XT_TABLE_MAXNAMELEN];
1873         compat_uint_t size;
1874         struct compat_ipt_entry entrytable[0];
1875 };
1876
1877 static int
1878 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1879                             void __user *userptr)
1880 {
1881         struct xt_counters *counters;
1882         const struct xt_table_info *private = table->private;
1883         void __user *pos;
1884         unsigned int size;
1885         int ret = 0;
1886         const void *loc_cpu_entry;
1887         unsigned int i = 0;
1888         struct ipt_entry *iter;
1889
1890         counters = alloc_counters(table);
1891         if (IS_ERR(counters))
1892                 return PTR_ERR(counters);
1893
1894         /* choose the copy that is on our node/cpu, ...
1895          * This choice is lazy (because current thread is
1896          * allowed to migrate to another cpu)
1897          */
1898         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1899         pos = userptr;
1900         size = total_size;
1901         xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1902                 ret = compat_copy_entry_to_user(iter, &pos,
1903                                                 &size, counters, i++);
1904                 if (ret != 0)
1905                         break;
1906         }
1907
1908         vfree(counters);
1909         return ret;
1910 }
1911
1912 static int
1913 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1914                    int *len)
1915 {
1916         int ret;
1917         struct compat_ipt_get_entries get;
1918         struct xt_table *t;
1919
1920         if (*len < sizeof(get)) {
1921                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1922                 return -EINVAL;
1923         }
1924
1925         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1926                 return -EFAULT;
1927
1928         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1929                 duprintf("compat_get_entries: %u != %zu\n",
1930                          *len, sizeof(get) + get.size);
1931                 return -EINVAL;
1932         }
1933
1934         xt_compat_lock(AF_INET);
1935         t = xt_find_table_lock(net, AF_INET, get.name);
1936         if (!IS_ERR_OR_NULL(t)) {
1937                 const struct xt_table_info *private = t->private;
1938                 struct xt_table_info info;
1939                 duprintf("t->private->number = %u\n", private->number);
1940                 ret = compat_table_info(private, &info);
1941                 if (!ret && get.size == info.size) {
1942                         ret = compat_copy_entries_to_user(private->size,
1943                                                           t, uptr->entrytable);
1944                 } else if (!ret) {
1945                         duprintf("compat_get_entries: I've got %u not %u!\n",
1946                                  private->size, get.size);
1947                         ret = -EAGAIN;
1948                 }
1949                 xt_compat_flush_offsets(AF_INET);
1950                 module_put(t->me);
1951                 xt_table_unlock(t);
1952         } else
1953                 ret = t ? PTR_ERR(t) : -ENOENT;
1954
1955         xt_compat_unlock(AF_INET);
1956         return ret;
1957 }
1958
1959 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1960
1961 static int
1962 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1963 {
1964         int ret;
1965
1966         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1967                 return -EPERM;
1968
1969         switch (cmd) {
1970         case IPT_SO_GET_INFO:
1971                 ret = get_info(sock_net(sk), user, len, 1);
1972                 break;
1973         case IPT_SO_GET_ENTRIES:
1974                 ret = compat_get_entries(sock_net(sk), user, len);
1975                 break;
1976         default:
1977                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1978         }
1979         return ret;
1980 }
1981 #endif
1982
1983 static int
1984 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1985 {
1986         int ret;
1987
1988         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1989                 return -EPERM;
1990
1991         switch (cmd) {
1992         case IPT_SO_SET_REPLACE:
1993                 ret = do_replace(sock_net(sk), user, len);
1994                 break;
1995
1996         case IPT_SO_SET_ADD_COUNTERS:
1997                 ret = do_add_counters(sock_net(sk), user, len, 0);
1998                 break;
1999
2000         default:
2001                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2002                 ret = -EINVAL;
2003         }
2004
2005         return ret;
2006 }
2007
2008 static int
2009 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2010 {
2011         int ret;
2012
2013         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2014                 return -EPERM;
2015
2016         switch (cmd) {
2017         case IPT_SO_GET_INFO:
2018                 ret = get_info(sock_net(sk), user, len, 0);
2019                 break;
2020
2021         case IPT_SO_GET_ENTRIES:
2022                 ret = get_entries(sock_net(sk), user, len);
2023                 break;
2024
2025         case IPT_SO_GET_REVISION_MATCH:
2026         case IPT_SO_GET_REVISION_TARGET: {
2027                 struct xt_get_revision rev;
2028                 int target;
2029
2030                 if (*len != sizeof(rev)) {
2031                         ret = -EINVAL;
2032                         break;
2033                 }
2034                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2035                         ret = -EFAULT;
2036                         break;
2037                 }
2038                 rev.name[sizeof(rev.name)-1] = 0;
2039
2040                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2041                         target = 1;
2042                 else
2043                         target = 0;
2044
2045                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2046                                                          rev.revision,
2047                                                          target, &ret),
2048                                         "ipt_%s", rev.name);
2049                 break;
2050         }
2051
2052         default:
2053                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2054                 ret = -EINVAL;
2055         }
2056
2057         return ret;
2058 }
2059
2060 struct xt_table *ipt_register_table(struct net *net,
2061                                     const struct xt_table *table,
2062                                     const struct ipt_replace *repl)
2063 {
2064         int ret;
2065         struct xt_table_info *newinfo;
2066         struct xt_table_info bootstrap = {0};
2067         void *loc_cpu_entry;
2068         struct xt_table *new_table;
2069
2070         newinfo = xt_alloc_table_info(repl->size);
2071         if (!newinfo) {
2072                 ret = -ENOMEM;
2073                 goto out;
2074         }
2075
2076         /* choose the copy on our node/cpu, but dont care about preemption */
2077         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2078         memcpy(loc_cpu_entry, repl->entries, repl->size);
2079
2080         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2081         if (ret != 0)
2082                 goto out_free;
2083
2084         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2085         if (IS_ERR(new_table)) {
2086                 ret = PTR_ERR(new_table);
2087                 goto out_free;
2088         }
2089
2090         return new_table;
2091
2092 out_free:
2093         xt_free_table_info(newinfo);
2094 out:
2095         return ERR_PTR(ret);
2096 }
2097
2098 void ipt_unregister_table(struct net *net, struct xt_table *table)
2099 {
2100         struct xt_table_info *private;
2101         void *loc_cpu_entry;
2102         struct module *table_owner = table->me;
2103         struct ipt_entry *iter;
2104
2105         private = xt_unregister_table(table);
2106
2107         /* Decrease module usage counts and free resources */
2108         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2109         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2110                 cleanup_entry(iter, net);
2111         if (private->number > private->initial_entries)
2112                 module_put(table_owner);
2113         xt_free_table_info(private);
2114 }
2115
2116 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2117 static inline bool
2118 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2119                      u_int8_t type, u_int8_t code,
2120                      bool invert)
2121 {
2122         return ((test_type == 0xFF) ||
2123                 (type == test_type && code >= min_code && code <= max_code))
2124                 ^ invert;
2125 }
2126
2127 static bool
2128 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2129 {
2130         const struct icmphdr *ic;
2131         struct icmphdr _icmph;
2132         const struct ipt_icmp *icmpinfo = par->matchinfo;
2133
2134         /* Must not be a fragment. */
2135         if (par->fragoff != 0)
2136                 return false;
2137
2138         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2139         if (ic == NULL) {
2140                 /* We've been asked to examine this packet, and we
2141                  * can't.  Hence, no choice but to drop.
2142                  */
2143                 duprintf("Dropping evil ICMP tinygram.\n");
2144                 par->hotdrop = true;
2145                 return false;
2146         }
2147
2148         return icmp_type_code_match(icmpinfo->type,
2149                                     icmpinfo->code[0],
2150                                     icmpinfo->code[1],
2151                                     ic->type, ic->code,
2152                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2153 }
2154
2155 static int icmp_checkentry(const struct xt_mtchk_param *par)
2156 {
2157         const struct ipt_icmp *icmpinfo = par->matchinfo;
2158
2159         /* Must specify no unknown invflags */
2160         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2161 }
2162
2163 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2164         {
2165                 .name             = XT_STANDARD_TARGET,
2166                 .targetsize       = sizeof(int),
2167                 .family           = NFPROTO_IPV4,
2168 #ifdef CONFIG_COMPAT
2169                 .compatsize       = sizeof(compat_int_t),
2170                 .compat_from_user = compat_standard_from_user,
2171                 .compat_to_user   = compat_standard_to_user,
2172 #endif
2173         },
2174         {
2175                 .name             = XT_ERROR_TARGET,
2176                 .target           = ipt_error,
2177                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2178                 .family           = NFPROTO_IPV4,
2179         },
2180 };
2181
2182 static struct nf_sockopt_ops ipt_sockopts = {
2183         .pf             = PF_INET,
2184         .set_optmin     = IPT_BASE_CTL,
2185         .set_optmax     = IPT_SO_SET_MAX+1,
2186         .set            = do_ipt_set_ctl,
2187 #ifdef CONFIG_COMPAT
2188         .compat_set     = compat_do_ipt_set_ctl,
2189 #endif
2190         .get_optmin     = IPT_BASE_CTL,
2191         .get_optmax     = IPT_SO_GET_MAX+1,
2192         .get            = do_ipt_get_ctl,
2193 #ifdef CONFIG_COMPAT
2194         .compat_get     = compat_do_ipt_get_ctl,
2195 #endif
2196         .owner          = THIS_MODULE,
2197 };
2198
2199 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2200         {
2201                 .name       = "icmp",
2202                 .match      = icmp_match,
2203                 .matchsize  = sizeof(struct ipt_icmp),
2204                 .checkentry = icmp_checkentry,
2205                 .proto      = IPPROTO_ICMP,
2206                 .family     = NFPROTO_IPV4,
2207         },
2208 };
2209
2210 static int __net_init ip_tables_net_init(struct net *net)
2211 {
2212         return xt_proto_init(net, NFPROTO_IPV4);
2213 }
2214
2215 static void __net_exit ip_tables_net_exit(struct net *net)
2216 {
2217         xt_proto_fini(net, NFPROTO_IPV4);
2218 }
2219
2220 static struct pernet_operations ip_tables_net_ops = {
2221         .init = ip_tables_net_init,
2222         .exit = ip_tables_net_exit,
2223 };
2224
2225 static int __init ip_tables_init(void)
2226 {
2227         int ret;
2228
2229         ret = register_pernet_subsys(&ip_tables_net_ops);
2230         if (ret < 0)
2231                 goto err1;
2232
2233         /* No one else will be downing sem now, so we won't sleep */
2234         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2235         if (ret < 0)
2236                 goto err2;
2237         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2238         if (ret < 0)
2239                 goto err4;
2240
2241         /* Register setsockopt */
2242         ret = nf_register_sockopt(&ipt_sockopts);
2243         if (ret < 0)
2244                 goto err5;
2245
2246         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2247         return 0;
2248
2249 err5:
2250         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2251 err4:
2252         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2253 err2:
2254         unregister_pernet_subsys(&ip_tables_net_ops);
2255 err1:
2256         return ret;
2257 }
2258
2259 static void __exit ip_tables_fini(void)
2260 {
2261         nf_unregister_sockopt(&ipt_sockopts);
2262
2263         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2264         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2265         unregister_pernet_subsys(&ip_tables_net_ops);
2266 }
2267
2268 EXPORT_SYMBOL(ipt_register_table);
2269 EXPORT_SYMBOL(ipt_unregister_table);
2270 EXPORT_SYMBOL(ipt_do_table);
2271 module_init(ip_tables_init);
2272 module_exit(ip_tables_fini);