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         /*
 331          * Ensure we load private-> members after we've fetched the base
 332          * pointer.
 333          */
 334         smp_read_barrier_depends();
 335         table_base = private->entries[cpu];
 336         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 337         stackptr   = per_cpu_ptr(private->stackptr, cpu);
 338         origptr    = *stackptr;
 339
 340         e = get_entry(table_base, private->hook_entry[hook]);
 341
 342         pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
 343                  table->name, hook, origptr,
 344                  get_entry(table_base, private->underflow[hook]));
 345
 346         do {
 347                 const struct xt_entry_target *t;
 348                 const struct xt_entry_match *ematch;
 349
 350                 IP_NF_ASSERT(e);
 351                 if (!ip_packet_match(ip, indev, outdev,
 352                     &e->ip, acpar.fragoff)) {
 353  no_match:
 354                         e = ipt_next_entry(e);
 355                         continue;
 356                 }
 357
 358                 xt_ematch_foreach(ematch, e) {
 359                         acpar.match     = ematch->u.kernel.match;
 360                         acpar.matchinfo = ematch->data;
 361                         if (!acpar.match->match(skb, &acpar))
 362                                 goto no_match;
 363                 }
 364
 365                 ADD_COUNTER(e->counters, skb->len, 1);
 366
 367                 t = ipt_get_target(e);
 368                 IP_NF_ASSERT(t->u.kernel.target);
 369
 370 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 371                 /* The packet is traced: log it */
 372                 if (unlikely(skb->nf_trace))
 373                         trace_packet(skb, hook, in, out,
 374                                      table->name, private, e);
 375 #endif
 376                 /* Standard target? */
 377                 if (!t->u.kernel.target->target) {
 378                         int v;
 379
 380                         v = ((struct xt_standard_target *)t)->verdict;
 381                         if (v < 0) {
 382                                 /* Pop from stack? */
 383                                 if (v != XT_RETURN) {
 384                                         verdict = (unsigned int)(-v) - 1;
 385                                         break;
 386                                 }
 387                                 if (*stackptr <= origptr) {
 388                                         e = get_entry(table_base,
 389                                             private->underflow[hook]);
 390                                         pr_debug("Underflow (this is normal) "
 391                                                  "to %p\n", e);
 392                                 } else {
 393                                         e = jumpstack[--*stackptr];
 394                                         pr_debug("Pulled %p out from pos %u\n",
 395                                                  e, *stackptr);
 396                                         e = ipt_next_entry(e);
 397                                 }
 398                                 continue;
 399                         }
 400                         if (table_base + v != ipt_next_entry(e) &&
 401                             !(e->ip.flags & IPT_F_GOTO)) {
 402                                 if (*stackptr >= private->stacksize) {
 403                                         verdict = NF_DROP;
 404                                         break;
 405                                 }
 406                                 jumpstack[(*stackptr)++] = e;
 407                                 pr_debug("Pushed %p into pos %u\n",
 408                                          e, *stackptr - 1);
 409                         }
 410
 411                         e = get_entry(table_base, v);
 412                         continue;
 413                 }
 414
 415                 acpar.target   = t->u.kernel.target;
 416                 acpar.targinfo = t->data;
 417
 418                 verdict = t->u.kernel.target->target(skb, &acpar);
 419                 /* Target might have changed stuff. */
 420                 ip = ip_hdr(skb);
 421                 if (verdict == XT_CONTINUE)
 422                         e = ipt_next_entry(e);
 423                 else
 424                         /* Verdict */
 425                         break;
 426         } while (!acpar.hotdrop);
 427         pr_debug("Exiting %s; resetting sp from %u to %u\n",
 428                  __func__, *stackptr, origptr);
 429         *stackptr = origptr;
 430         xt_write_recseq_end(addend);
 431         local_bh_enable();
 432
 433 #ifdef DEBUG_ALLOW_ALL
 434         return NF_ACCEPT;
 435 #else
 436         if (acpar.hotdrop)
 437                 return NF_DROP;
 438         else return verdict;
 439 #endif
 440 }
 441
 442 /* Figures out from what hook each rule can be called: returns 0 if
 443    there are loops.  Puts hook bitmask in comefrom. */
 444 static int
 445 mark_source_chains(const struct xt_table_info *newinfo,
 446                    unsigned int valid_hooks, void *entry0)
 447 {
 448         unsigned int hook;
 449
 450         /* No recursion; use packet counter to save back ptrs (reset
 451            to 0 as we leave), and comefrom to save source hook bitmask */
 452         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 453                 unsigned int pos = newinfo->hook_entry[hook];
 454                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 455
 456                 if (!(valid_hooks & (1 << hook)))
 457                         continue;
 458
 459                 /* Set initial back pointer. */
 460                 e->counters.pcnt = pos;
 461
 462                 for (;;) {
 463                         const struct xt_standard_target *t
 464                                 = (void *)ipt_get_target_c(e);
 465                         int visited = e->comefrom & (1 << hook);
 466
 467                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 468                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 469                                        hook, pos, e->comefrom);
 470                                 return 0;
 471                         }
 472                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 473
 474                         /* Unconditional return/END. */
 475                         if ((e->target_offset == sizeof(struct ipt_entry) &&
 476                              (strcmp(t->target.u.user.name,
 477                                      XT_STANDARD_TARGET) == 0) &&
 478                              t->verdict < 0 && unconditional(&e->ip)) ||
 479                             visited) {
 480                                 unsigned int oldpos, size;
 481
 482                                 if ((strcmp(t->target.u.user.name,
 483                                             XT_STANDARD_TARGET) == 0) &&
 484                                     t->verdict < -NF_MAX_VERDICT - 1) {
 485                                         duprintf("mark_source_chains: bad "
 486                                                 "negative verdict (%i)\n",
 487                                                                 t->verdict);
 488                                         return 0;
 489                                 }
 490
 491                                 /* Return: backtrack through the last
 492                                    big jump. */
 493                                 do {
 494                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 495 #ifdef DEBUG_IP_FIREWALL_USER
 496                                         if (e->comefrom
 497                                             & (1 << NF_INET_NUMHOOKS)) {
 498                                                 duprintf("Back unset "
 499                                                          "on hook %u "
 500                                                          "rule %u\n",
 501                                                          hook, pos);
 502                                         }
 503 #endif
 504                                         oldpos = pos;
 505                                         pos = e->counters.pcnt;
 506                                         e->counters.pcnt = 0;
 507
 508                                         /* We're at the start. */
 509                                         if (pos == oldpos)
 510                                                 goto next;
 511
 512                                         e = (struct ipt_entry *)
 513                                                 (entry0 + pos);
 514                                 } while (oldpos == pos + e->next_offset);
 515
 516                                 /* Move along one */
 517                                 size = e->next_offset;
 518                                 e = (struct ipt_entry *)
 519                                         (entry0 + pos + size);
 520                                 e->counters.pcnt = pos;
 521                                 pos += size;
 522                         } else {
 523                                 int newpos = t->verdict;
 524
 525                                 if (strcmp(t->target.u.user.name,
 526                                            XT_STANDARD_TARGET) == 0 &&
 527                                     newpos >= 0) {
 528                                         if (newpos > newinfo->size -
 529                                                 sizeof(struct ipt_entry)) {
 530                                                 duprintf("mark_source_chains: "
 531                                                         "bad verdict (%i)\n",
 532                                                                 newpos);
 533                                                 return 0;
 534                                         }
 535                                         /* This a jump; chase it. */
 536                                         duprintf("Jump rule %u -> %u\n",
 537                                                  pos, newpos);
 538                                 } else {
 539                                         /* ... this is a fallthru */
 540                                         newpos = pos + e->next_offset;
 541                                 }
 542                                 e = (struct ipt_entry *)
 543                                         (entry0 + newpos);
 544                                 e->counters.pcnt = pos;
 545                                 pos = newpos;
 546                         }
 547                 }
 548                 next:
 549                 duprintf("Finished chain %u\n", hook);
 550         }
 551         return 1;
 552 }
 553
 554 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 555 {
 556         struct xt_mtdtor_param par;
 557
 558         par.net       = net;
 559         par.match     = m->u.kernel.match;
 560         par.matchinfo = m->data;
 561         par.family    = NFPROTO_IPV4;
 562         if (par.match->destroy != NULL)
 563                 par.match->destroy(&par);
 564         module_put(par.match->me);
 565 }
 566
 567 static int
 568 check_entry(const struct ipt_entry *e, const char *name)
 569 {
 570         const struct xt_entry_target *t;
 571
 572         if (!ip_checkentry(&e->ip)) {
 573                 duprintf("ip check failed %p %s.\n", e, name);
 574                 return -EINVAL;
 575         }
 576
 577         if (e->target_offset + sizeof(struct xt_entry_target) >
 578             e->next_offset)
 579                 return -EINVAL;
 580
 581         t = ipt_get_target_c(e);
 582         if (e->target_offset + t->u.target_size > e->next_offset)
 583                 return -EINVAL;
 584
 585         return 0;
 586 }
 587
 588 static int
 589 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 590 {
 591         const struct ipt_ip *ip = par->entryinfo;
 592         int ret;
 593
 594         par->match     = m->u.kernel.match;
 595         par->matchinfo = m->data;
 596
 597         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 598               ip->proto, ip->invflags & IPT_INV_PROTO);
 599         if (ret < 0) {
 600                 duprintf("check failed for `%s'.\n", par->match->name);
 601                 return ret;
 602         }
 603         return 0;
 604 }
 605
 606 static int
 607 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 608 {
 609         struct xt_match *match;
 610         int ret;
 611
 612         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 613                                       m->u.user.revision);
 614         if (IS_ERR(match)) {
 615                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 616                 return PTR_ERR(match);
 617         }
 618         m->u.kernel.match = match;
 619
 620         ret = check_match(m, par);
 621         if (ret)
 622                 goto err;
 623
 624         return 0;
 625 err:
 626         module_put(m->u.kernel.match->me);
 627         return ret;
 628 }
 629
 630 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 631 {
 632         struct xt_entry_target *t = ipt_get_target(e);
 633         struct xt_tgchk_param par = {
 634                 .net       = net,
 635                 .table     = name,
 636                 .entryinfo = e,
 637                 .target    = t->u.kernel.target,
 638                 .targinfo  = t->data,
 639                 .hook_mask = e->comefrom,
 640                 .family    = NFPROTO_IPV4,
 641         };
 642         int ret;
 643
 644         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 645               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 646         if (ret < 0) {
 647                 duprintf("check failed for `%s'.\n",
 648                          t->u.kernel.target->name);
 649                 return ret;
 650         }
 651         return 0;
 652 }
 653
 654 static int
 655 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 656                  unsigned int size)
 657 {
 658         struct xt_entry_target *t;
 659         struct xt_target *target;
 660         int ret;
 661         unsigned int j;
 662         struct xt_mtchk_param mtpar;
 663         struct xt_entry_match *ematch;
 664
 665         ret = check_entry(e, name);
 666         if (ret)
 667                 return ret;
 668
 669         j = 0;
 670         mtpar.net       = net;
 671         mtpar.table     = name;
 672         mtpar.entryinfo = &e->ip;
 673         mtpar.hook_mask = e->comefrom;
 674         mtpar.family    = NFPROTO_IPV4;
 675         xt_ematch_foreach(ematch, e) {
 676                 ret = find_check_match(ematch, &mtpar);
 677                 if (ret != 0)
 678                         goto cleanup_matches;
 679                 ++j;
 680         }
 681
 682         t = ipt_get_target(e);
 683         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 684                                         t->u.user.revision);
 685         if (IS_ERR(target)) {
 686                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 687                 ret = PTR_ERR(target);
 688                 goto cleanup_matches;
 689         }
 690         t->u.kernel.target = target;
 691
 692         ret = check_target(e, net, name);
 693         if (ret)
 694                 goto err;
 695         return 0;
 696  err:
 697         module_put(t->u.kernel.target->me);
 698  cleanup_matches:
 699         xt_ematch_foreach(ematch, e) {
 700                 if (j-- == 0)
 701                         break;
 702                 cleanup_match(ematch, net);
 703         }
 704         return ret;
 705 }
 706
 707 static bool check_underflow(const struct ipt_entry *e)
 708 {
 709         const struct xt_entry_target *t;
 710         unsigned int verdict;
 711
 712         if (!unconditional(&e->ip))
 713                 return false;
 714         t = ipt_get_target_c(e);
 715         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 716                 return false;
 717         verdict = ((struct xt_standard_target *)t)->verdict;
 718         verdict = -verdict - 1;
 719         return verdict == NF_DROP || verdict == NF_ACCEPT;
 720 }
 721
 722 static int
 723 check_entry_size_and_hooks(struct ipt_entry *e,
 724                            struct xt_table_info *newinfo,
 725                            const unsigned char *base,
 726                            const unsigned char *limit,
 727                            const unsigned int *hook_entries,
 728                            const unsigned int *underflows,
 729                            unsigned int valid_hooks)
 730 {
 731         unsigned int h;
 732
 733         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 734             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 735                 duprintf("Bad offset %p\n", e);
 736                 return -EINVAL;
 737         }
 738
 739         if (e->next_offset
 740             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 741                 duprintf("checking: element %p size %u\n",
 742                          e, e->next_offset);
 743                 return -EINVAL;
 744         }
 745
 746         /* Check hooks & underflows */
 747         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 748                 if (!(valid_hooks & (1 << h)))
 749                         continue;
 750                 if ((unsigned char *)e - base == hook_entries[h])
 751                         newinfo->hook_entry[h] = hook_entries[h];
 752                 if ((unsigned char *)e - base == underflows[h]) {
 753                         if (!check_underflow(e)) {
 754                                 pr_err("Underflows must be unconditional and "
 755                                        "use the STANDARD target with "
 756                                        "ACCEPT/DROP\n");
 757                                 return -EINVAL;
 758                         }
 759                         newinfo->underflow[h] = underflows[h];
 760                 }
 761         }
 762
 763         /* Clear counters and comefrom */
 764         e->counters = ((struct xt_counters) { 0, 0 });
 765         e->comefrom = 0;
 766         return 0;
 767 }
 768
 769 static void
 770 cleanup_entry(struct ipt_entry *e, struct net *net)
 771 {
 772         struct xt_tgdtor_param par;
 773         struct xt_entry_target *t;
 774         struct xt_entry_match *ematch;
 775
 776         /* Cleanup all matches */
 777         xt_ematch_foreach(ematch, e)
 778                 cleanup_match(ematch, net);
 779         t = ipt_get_target(e);
 780
 781         par.net      = net;
 782         par.target   = t->u.kernel.target;
 783         par.targinfo = t->data;
 784         par.family   = NFPROTO_IPV4;
 785         if (par.target->destroy != NULL)
 786                 par.target->destroy(&par);
 787         module_put(par.target->me);
 788 }
 789
 790 /* Checks and translates the user-supplied table segment (held in
 791    newinfo) */
 792 static int
 793 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 794                 const struct ipt_replace *repl)
 795 {
 796         struct ipt_entry *iter;
 797         unsigned int i;
 798         int ret = 0;
 799
 800         newinfo->size = repl->size;
 801         newinfo->number = repl->num_entries;
 802
 803         /* Init all hooks to impossible value. */
 804         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 805                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 806                 newinfo->underflow[i] = 0xFFFFFFFF;
 807         }
 808
 809         duprintf("translate_table: size %u\n", newinfo->size);
 810         i = 0;
 811         /* Walk through entries, checking offsets. */
 812         xt_entry_foreach(iter, entry0, newinfo->size) {
 813                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 814                                                  entry0 + repl->size,
 815                                                  repl->hook_entry,
 816                                                  repl->underflow,
 817                                                  repl->valid_hooks);
 818                 if (ret != 0)
 819                         return ret;
 820                 ++i;
 821                 if (strcmp(ipt_get_target(iter)->u.user.name,
 822                     XT_ERROR_TARGET) == 0)
 823                         ++newinfo->stacksize;
 824         }
 825
 826         if (i != repl->num_entries) {
 827                 duprintf("translate_table: %u not %u entries\n",
 828                          i, repl->num_entries);
 829                 return -EINVAL;
 830         }
 831
 832         /* Check hooks all assigned */
 833         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 834                 /* Only hooks which are valid */
 835                 if (!(repl->valid_hooks & (1 << i)))
 836                         continue;
 837                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 838                         duprintf("Invalid hook entry %u %u\n",
 839                                  i, repl->hook_entry[i]);
 840                         return -EINVAL;
 841                 }
 842                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 843                         duprintf("Invalid underflow %u %u\n",
 844                                  i, repl->underflow[i]);
 845                         return -EINVAL;
 846                 }
 847         }
 848
 849         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 850                 return -ELOOP;
 851
 852         /* Finally, each sanity check must pass */
 853         i = 0;
 854         xt_entry_foreach(iter, entry0, newinfo->size) {
 855                 ret = find_check_entry(iter, net, repl->name, repl->size);
 856                 if (ret != 0)
 857                         break;
 858                 ++i;
 859         }
 860
 861         if (ret != 0) {
 862                 xt_entry_foreach(iter, entry0, newinfo->size) {
 863                         if (i-- == 0)
 864                                 break;
 865                         cleanup_entry(iter, net);
 866                 }
 867                 return ret;
 868         }
 869
 870         /* And one copy for every other CPU */
 871         for_each_possible_cpu(i) {
 872                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
 873                         memcpy(newinfo->entries[i], entry0, newinfo->size);
 874         }
 875
 876         return ret;
 877 }
 878
 879 static void
 880 get_counters(const struct xt_table_info *t,
 881              struct xt_counters counters[])
 882 {
 883         struct ipt_entry *iter;
 884         unsigned int cpu;
 885         unsigned int i;
 886
 887         for_each_possible_cpu(cpu) {
 888                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 889
 890                 i = 0;
 891                 xt_entry_foreach(iter, t->entries[cpu], t->size) {
 892                         u64 bcnt, pcnt;
 893                         unsigned int start;
 894
 895                         do {
 896                                 start = read_seqcount_begin(s);
 897                                 bcnt = iter->counters.bcnt;
 898                                 pcnt = iter->counters.pcnt;
 899                         } while (read_seqcount_retry(s, start));
 900
 901                         ADD_COUNTER(counters[i], bcnt, pcnt);
 902                         ++i; /* macro does multi eval of i */
 903                 }
 904         }
 905 }
 906
 907 static struct xt_counters *alloc_counters(const struct xt_table *table)
 908 {
 909         unsigned int countersize;
 910         struct xt_counters *counters;
 911         const struct xt_table_info *private = table->private;
 912
 913         /* We need atomic snapshot of counters: rest doesn't change
 914            (other than comefrom, which userspace doesn't care
 915            about). */
 916         countersize = sizeof(struct xt_counters) * private->number;
 917         counters = vzalloc(countersize);
 918
 919         if (counters == NULL)
 920                 return ERR_PTR(-ENOMEM);
 921
 922         get_counters(private, counters);
 923
 924         return counters;
 925 }
 926
 927 static int
 928 copy_entries_to_user(unsigned int total_size,
 929                      const struct xt_table *table,
 930                      void __user *userptr)
 931 {
 932         unsigned int off, num;
 933         const struct ipt_entry *e;
 934         struct xt_counters *counters;
 935         const struct xt_table_info *private = table->private;
 936         int ret = 0;
 937         const void *loc_cpu_entry;
 938
 939         counters = alloc_counters(table);
 940         if (IS_ERR(counters))
 941                 return PTR_ERR(counters);
 942
 943         /* choose the copy that is on our node/cpu, ...
 944          * This choice is lazy (because current thread is
 945          * allowed to migrate to another cpu)
 946          */
 947         loc_cpu_entry = private->entries[raw_smp_processor_id()];
 948         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 949                 ret = -EFAULT;
 950                 goto free_counters;
 951         }
 952
 953         /* FIXME: use iterator macros --RR */
 954         /* ... then go back and fix counters and names */
 955         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 956                 unsigned int i;
 957                 const struct xt_entry_match *m;
 958                 const struct xt_entry_target *t;
 959
 960                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 961                 if (copy_to_user(userptr + off
 962                                  + offsetof(struct ipt_entry, counters),
 963                                  &counters[num],
 964                                  sizeof(counters[num])) != 0) {
 965                         ret = -EFAULT;
 966                         goto free_counters;
 967                 }
 968
 969                 for (i = sizeof(struct ipt_entry);
 970                      i < e->target_offset;
 971                      i += m->u.match_size) {
 972                         m = (void *)e + i;
 973
 974                         if (copy_to_user(userptr + off + i
 975                                          + offsetof(struct xt_entry_match,
 976                                                     u.user.name),
 977                                          m->u.kernel.match->name,
 978                                          strlen(m->u.kernel.match->name)+1)
 979                             != 0) {
 980                                 ret = -EFAULT;
 981                                 goto free_counters;
 982                         }
 983                 }
 984
 985                 t = ipt_get_target_c(e);
 986                 if (copy_to_user(userptr + off + e->target_offset
 987                                  + offsetof(struct xt_entry_target,
 988                                             u.user.name),
 989                                  t->u.kernel.target->name,
 990                                  strlen(t->u.kernel.target->name)+1) != 0) {
 991                         ret = -EFAULT;
 992                         goto free_counters;
 993                 }
 994         }
 995
 996  free_counters:
 997         vfree(counters);
 998         return ret;
 999 }
1000
1001 #ifdef CONFIG_COMPAT
1002 static void compat_standard_from_user(void *dst, const void *src)
1003 {
1004         int v = *(compat_int_t *)src;
1005
1006         if (v > 0)
1007                 v += xt_compat_calc_jump(AF_INET, v);
1008         memcpy(dst, &v, sizeof(v));
1009 }
1010
1011 static int compat_standard_to_user(void __user *dst, const void *src)
1012 {
1013         compat_int_t cv = *(int *)src;
1014
1015         if (cv > 0)
1016                 cv -= xt_compat_calc_jump(AF_INET, cv);
1017         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1018 }
1019
1020 static int compat_calc_entry(const struct ipt_entry *e,
1021                              const struct xt_table_info *info,
1022                              const void *base, struct xt_table_info *newinfo)
1023 {
1024         const struct xt_entry_match *ematch;
1025         const struct xt_entry_target *t;
1026         unsigned int entry_offset;
1027         int off, i, ret;
1028
1029         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1030         entry_offset = (void *)e - base;
1031         xt_ematch_foreach(ematch, e)
1032                 off += xt_compat_match_offset(ematch->u.kernel.match);
1033         t = ipt_get_target_c(e);
1034         off += xt_compat_target_offset(t->u.kernel.target);
1035         newinfo->size -= off;
1036         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1037         if (ret)
1038                 return ret;
1039
1040         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1041                 if (info->hook_entry[i] &&
1042                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1043                         newinfo->hook_entry[i] -= off;
1044                 if (info->underflow[i] &&
1045                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1046                         newinfo->underflow[i] -= off;
1047         }
1048         return 0;
1049 }
1050
1051 static int compat_table_info(const struct xt_table_info *info,
1052                              struct xt_table_info *newinfo)
1053 {
1054         struct ipt_entry *iter;
1055         void *loc_cpu_entry;
1056         int ret;
1057
1058         if (!newinfo || !info)
1059                 return -EINVAL;
1060
1061         /* we dont care about newinfo->entries[] */
1062         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1063         newinfo->initial_entries = 0;
1064         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1065         xt_compat_init_offsets(AF_INET, info->number);
1066         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1067                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1068                 if (ret != 0)
1069                         return ret;
1070         }
1071         return 0;
1072 }
1073 #endif
1074
1075 static int get_info(struct net *net, void __user *user,
1076                     const int *len, int compat)
1077 {
1078         char name[XT_TABLE_MAXNAMELEN];
1079         struct xt_table *t;
1080         int ret;
1081
1082         if (*len != sizeof(struct ipt_getinfo)) {
1083                 duprintf("length %u != %zu\n", *len,
1084                          sizeof(struct ipt_getinfo));
1085                 return -EINVAL;
1086         }
1087
1088         if (copy_from_user(name, user, sizeof(name)) != 0)
1089                 return -EFAULT;
1090
1091         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1092 #ifdef CONFIG_COMPAT
1093         if (compat)
1094                 xt_compat_lock(AF_INET);
1095 #endif
1096         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1097                                     "iptable_%s", name);
1098         if (!IS_ERR_OR_NULL(t)) {
1099                 struct ipt_getinfo info;
1100                 const struct xt_table_info *private = t->private;
1101 #ifdef CONFIG_COMPAT
1102                 struct xt_table_info tmp;
1103
1104                 if (compat) {
1105                         ret = compat_table_info(private, &tmp);
1106                         xt_compat_flush_offsets(AF_INET);
1107                         private = &tmp;
1108                 }
1109 #endif
1110                 memset(&info, 0, sizeof(info));
1111                 info.valid_hooks = t->valid_hooks;
1112                 memcpy(info.hook_entry, private->hook_entry,
1113                        sizeof(info.hook_entry));
1114                 memcpy(info.underflow, private->underflow,
1115                        sizeof(info.underflow));
1116                 info.num_entries = private->number;
1117                 info.size = private->size;
1118                 strcpy(info.name, name);
1119
1120                 if (copy_to_user(user, &info, *len) != 0)
1121                         ret = -EFAULT;
1122                 else
1123                         ret = 0;
1124
1125                 xt_table_unlock(t);
1126                 module_put(t->me);
1127         } else
1128                 ret = t ? PTR_ERR(t) : -ENOENT;
1129 #ifdef CONFIG_COMPAT
1130         if (compat)
1131                 xt_compat_unlock(AF_INET);
1132 #endif
1133         return ret;
1134 }
1135
1136 static int
1137 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1138             const int *len)
1139 {
1140         int ret;
1141         struct ipt_get_entries get;
1142         struct xt_table *t;
1143
1144         if (*len < sizeof(get)) {
1145                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1146                 return -EINVAL;
1147         }
1148         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1149                 return -EFAULT;
1150         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1151                 duprintf("get_entries: %u != %zu\n",
1152                          *len, sizeof(get) + get.size);
1153                 return -EINVAL;
1154         }
1155
1156         t = xt_find_table_lock(net, AF_INET, get.name);
1157         if (!IS_ERR_OR_NULL(t)) {
1158                 const struct xt_table_info *private = t->private;
1159                 duprintf("t->private->number = %u\n", private->number);
1160                 if (get.size == private->size)
1161                         ret = copy_entries_to_user(private->size,
1162                                                    t, uptr->entrytable);
1163                 else {
1164                         duprintf("get_entries: I've got %u not %u!\n",
1165                                  private->size, get.size);
1166                         ret = -EAGAIN;
1167                 }
1168                 module_put(t->me);
1169                 xt_table_unlock(t);
1170         } else
1171                 ret = t ? PTR_ERR(t) : -ENOENT;
1172
1173         return ret;
1174 }
1175
1176 static int
1177 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1178              struct xt_table_info *newinfo, unsigned int num_counters,
1179              void __user *counters_ptr)
1180 {
1181         int ret;
1182         struct xt_table *t;
1183         struct xt_table_info *oldinfo;
1184         struct xt_counters *counters;
1185         void *loc_cpu_old_entry;
1186         struct ipt_entry *iter;
1187
1188         ret = 0;
1189         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1190         if (!counters) {
1191                 ret = -ENOMEM;
1192                 goto out;
1193         }
1194
1195         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1196                                     "iptable_%s", name);
1197         if (IS_ERR_OR_NULL(t)) {
1198                 ret = t ? PTR_ERR(t) : -ENOENT;
1199                 goto free_newinfo_counters_untrans;
1200         }
1201
1202         /* You lied! */
1203         if (valid_hooks != t->valid_hooks) {
1204                 duprintf("Valid hook crap: %08X vs %08X\n",
1205                          valid_hooks, t->valid_hooks);
1206                 ret = -EINVAL;
1207                 goto put_module;
1208         }
1209
1210         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1211         if (!oldinfo)
1212                 goto put_module;
1213
1214         /* Update module usage count based on number of rules */
1215         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1216                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1217         if ((oldinfo->number > oldinfo->initial_entries) ||
1218             (newinfo->number <= oldinfo->initial_entries))
1219                 module_put(t->me);
1220         if ((oldinfo->number > oldinfo->initial_entries) &&
1221             (newinfo->number <= oldinfo->initial_entries))
1222                 module_put(t->me);
1223
1224         /* Get the old counters, and synchronize with replace */
1225         get_counters(oldinfo, counters);
1226
1227         /* Decrease module usage counts and free resource */
1228         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1229         xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1230                 cleanup_entry(iter, net);
1231
1232         xt_free_table_info(oldinfo);
1233         if (copy_to_user(counters_ptr, counters,
1234                          sizeof(struct xt_counters) * num_counters) != 0) {
1235                 /* Silent error, can't fail, new table is already in place */
1236                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1237         }
1238         vfree(counters);
1239         xt_table_unlock(t);
1240         return ret;
1241
1242  put_module:
1243         module_put(t->me);
1244         xt_table_unlock(t);
1245  free_newinfo_counters_untrans:
1246         vfree(counters);
1247  out:
1248         return ret;
1249 }
1250
1251 static int
1252 do_replace(struct net *net, const void __user *user, unsigned int len)
1253 {
1254         int ret;
1255         struct ipt_replace tmp;
1256         struct xt_table_info *newinfo;
1257         void *loc_cpu_entry;
1258         struct ipt_entry *iter;
1259
1260         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1261                 return -EFAULT;
1262
1263         /* overflow check */
1264         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1265                 return -ENOMEM;
1266         tmp.name[sizeof(tmp.name)-1] = 0;
1267
1268         newinfo = xt_alloc_table_info(tmp.size);
1269         if (!newinfo)
1270                 return -ENOMEM;
1271
1272         /* choose the copy that is on our node/cpu */
1273         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1274         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1275                            tmp.size) != 0) {
1276                 ret = -EFAULT;
1277                 goto free_newinfo;
1278         }
1279
1280         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1281         if (ret != 0)
1282                 goto free_newinfo;
1283
1284         duprintf("Translated table\n");
1285
1286         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1287                            tmp.num_counters, tmp.counters);
1288         if (ret)
1289                 goto free_newinfo_untrans;
1290         return 0;
1291
1292  free_newinfo_untrans:
1293         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1294                 cleanup_entry(iter, net);
1295  free_newinfo:
1296         xt_free_table_info(newinfo);
1297         return ret;
1298 }
1299
1300 static int
1301 do_add_counters(struct net *net, const void __user *user,
1302                 unsigned int len, int compat)
1303 {
1304         unsigned int i, curcpu;
1305         struct xt_counters_info tmp;
1306         struct xt_counters *paddc;
1307         unsigned int num_counters;
1308         const char *name;
1309         int size;
1310         void *ptmp;
1311         struct xt_table *t;
1312         const struct xt_table_info *private;
1313         int ret = 0;
1314         void *loc_cpu_entry;
1315         struct ipt_entry *iter;
1316         unsigned int addend;
1317 #ifdef CONFIG_COMPAT
1318         struct compat_xt_counters_info compat_tmp;
1319
1320         if (compat) {
1321                 ptmp = &compat_tmp;
1322                 size = sizeof(struct compat_xt_counters_info);
1323         } else
1324 #endif
1325         {
1326                 ptmp = &tmp;
1327                 size = sizeof(struct xt_counters_info);
1328         }
1329
1330         if (copy_from_user(ptmp, user, size) != 0)
1331                 return -EFAULT;
1332
1333 #ifdef CONFIG_COMPAT
1334         if (compat) {
1335                 num_counters = compat_tmp.num_counters;
1336                 name = compat_tmp.name;
1337         } else
1338 #endif
1339         {
1340                 num_counters = tmp.num_counters;
1341                 name = tmp.name;
1342         }
1343
1344         if (len != size + num_counters * sizeof(struct xt_counters))
1345                 return -EINVAL;
1346
1347         paddc = vmalloc(len - size);
1348         if (!paddc)
1349                 return -ENOMEM;
1350
1351         if (copy_from_user(paddc, user + size, len - size) != 0) {
1352                 ret = -EFAULT;
1353                 goto free;
1354         }
1355
1356         t = xt_find_table_lock(net, AF_INET, name);
1357         if (IS_ERR_OR_NULL(t)) {
1358                 ret = t ? PTR_ERR(t) : -ENOENT;
1359                 goto free;
1360         }
1361
1362         local_bh_disable();
1363         private = t->private;
1364         if (private->number != num_counters) {
1365                 ret = -EINVAL;
1366                 goto unlock_up_free;
1367         }
1368
1369         i = 0;
1370         /* Choose the copy that is on our node */
1371         curcpu = smp_processor_id();
1372         loc_cpu_entry = private->entries[curcpu];
1373         addend = xt_write_recseq_begin();
1374         xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1375                 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1376                 ++i;
1377         }
1378         xt_write_recseq_end(addend);
1379  unlock_up_free:
1380         local_bh_enable();
1381         xt_table_unlock(t);
1382         module_put(t->me);
1383  free:
1384         vfree(paddc);
1385
1386         return ret;
1387 }
1388
1389 #ifdef CONFIG_COMPAT
1390 struct compat_ipt_replace {
1391         char                    name[XT_TABLE_MAXNAMELEN];
1392         u32                     valid_hooks;
1393         u32                     num_entries;
1394         u32                     size;
1395         u32                     hook_entry[NF_INET_NUMHOOKS];
1396         u32                     underflow[NF_INET_NUMHOOKS];
1397         u32                     num_counters;
1398         compat_uptr_t           counters;       /* struct xt_counters * */
1399         struct compat_ipt_entry entries[0];
1400 };
1401
1402 static int
1403 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1404                           unsigned int *size, struct xt_counters *counters,
1405                           unsigned int i)
1406 {
1407         struct xt_entry_target *t;
1408         struct compat_ipt_entry __user *ce;
1409         u_int16_t target_offset, next_offset;
1410         compat_uint_t origsize;
1411         const struct xt_entry_match *ematch;
1412         int ret = 0;
1413
1414         origsize = *size;
1415         ce = (struct compat_ipt_entry __user *)*dstptr;
1416         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1417             copy_to_user(&ce->counters, &counters[i],
1418             sizeof(counters[i])) != 0)
1419                 return -EFAULT;
1420
1421         *dstptr += sizeof(struct compat_ipt_entry);
1422         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1423
1424         xt_ematch_foreach(ematch, e) {
1425                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1426                 if (ret != 0)
1427                         return ret;
1428         }
1429         target_offset = e->target_offset - (origsize - *size);
1430         t = ipt_get_target(e);
1431         ret = xt_compat_target_to_user(t, dstptr, size);
1432         if (ret)
1433                 return ret;
1434         next_offset = e->next_offset - (origsize - *size);
1435         if (put_user(target_offset, &ce->target_offset) != 0 ||
1436             put_user(next_offset, &ce->next_offset) != 0)
1437                 return -EFAULT;
1438         return 0;
1439 }
1440
1441 static int
1442 compat_find_calc_match(struct xt_entry_match *m,
1443                        const char *name,
1444                        const struct ipt_ip *ip,
1445                        unsigned int hookmask,
1446                        int *size)
1447 {
1448         struct xt_match *match;
1449
1450         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1451                                       m->u.user.revision);
1452         if (IS_ERR(match)) {
1453                 duprintf("compat_check_calc_match: `%s' not found\n",
1454                          m->u.user.name);
1455                 return PTR_ERR(match);
1456         }
1457         m->u.kernel.match = match;
1458         *size += xt_compat_match_offset(match);
1459         return 0;
1460 }
1461
1462 static void compat_release_entry(struct compat_ipt_entry *e)
1463 {
1464         struct xt_entry_target *t;
1465         struct xt_entry_match *ematch;
1466
1467         /* Cleanup all matches */
1468         xt_ematch_foreach(ematch, e)
1469                 module_put(ematch->u.kernel.match->me);
1470         t = compat_ipt_get_target(e);
1471         module_put(t->u.kernel.target->me);
1472 }
1473
1474 static int
1475 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1476                                   struct xt_table_info *newinfo,
1477                                   unsigned int *size,
1478                                   const unsigned char *base,
1479                                   const unsigned char *limit,
1480                                   const unsigned int *hook_entries,
1481                                   const unsigned int *underflows,
1482                                   const char *name)
1483 {
1484         struct xt_entry_match *ematch;
1485         struct xt_entry_target *t;
1486         struct xt_target *target;
1487         unsigned int entry_offset;
1488         unsigned int j;
1489         int ret, off, h;
1490
1491         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1492         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1493             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1494                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1495                 return -EINVAL;
1496         }
1497
1498         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1499                              sizeof(struct compat_xt_entry_target)) {
1500                 duprintf("checking: element %p size %u\n",
1501                          e, e->next_offset);
1502                 return -EINVAL;
1503         }
1504
1505         /* For purposes of check_entry casting the compat entry is fine */
1506         ret = check_entry((struct ipt_entry *)e, name);
1507         if (ret)
1508                 return ret;
1509
1510         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1511         entry_offset = (void *)e - (void *)base;
1512         j = 0;
1513         xt_ematch_foreach(ematch, e) {
1514                 ret = compat_find_calc_match(ematch, name,
1515                                              &e->ip, e->comefrom, &off);
1516                 if (ret != 0)
1517                         goto release_matches;
1518                 ++j;
1519         }
1520
1521         t = compat_ipt_get_target(e);
1522         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1523                                         t->u.user.revision);
1524         if (IS_ERR(target)) {
1525                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1526                          t->u.user.name);
1527                 ret = PTR_ERR(target);
1528                 goto release_matches;
1529         }
1530         t->u.kernel.target = target;
1531
1532         off += xt_compat_target_offset(target);
1533         *size += off;
1534         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1535         if (ret)
1536                 goto out;
1537
1538         /* Check hooks & underflows */
1539         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1540                 if ((unsigned char *)e - base == hook_entries[h])
1541                         newinfo->hook_entry[h] = hook_entries[h];
1542                 if ((unsigned char *)e - base == underflows[h])
1543                         newinfo->underflow[h] = underflows[h];
1544         }
1545
1546         /* Clear counters and comefrom */
1547         memset(&e->counters, 0, sizeof(e->counters));
1548         e->comefrom = 0;
1549         return 0;
1550
1551 out:
1552         module_put(t->u.kernel.target->me);
1553 release_matches:
1554         xt_ematch_foreach(ematch, e) {
1555                 if (j-- == 0)
1556                         break;
1557                 module_put(ematch->u.kernel.match->me);
1558         }
1559         return ret;
1560 }
1561
1562 static int
1563 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1564                             unsigned int *size, const char *name,
1565                             struct xt_table_info *newinfo, unsigned char *base)
1566 {
1567         struct xt_entry_target *t;
1568         struct xt_target *target;
1569         struct ipt_entry *de;
1570         unsigned int origsize;
1571         int ret, h;
1572         struct xt_entry_match *ematch;
1573
1574         ret = 0;
1575         origsize = *size;
1576         de = (struct ipt_entry *)*dstptr;
1577         memcpy(de, e, sizeof(struct ipt_entry));
1578         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1579
1580         *dstptr += sizeof(struct ipt_entry);
1581         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1582
1583         xt_ematch_foreach(ematch, e) {
1584                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1585                 if (ret != 0)
1586                         return ret;
1587         }
1588         de->target_offset = e->target_offset - (origsize - *size);
1589         t = compat_ipt_get_target(e);
1590         target = t->u.kernel.target;
1591         xt_compat_target_from_user(t, dstptr, size);
1592
1593         de->next_offset = e->next_offset - (origsize - *size);
1594         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1595                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1596                         newinfo->hook_entry[h] -= origsize - *size;
1597                 if ((unsigned char *)de - base < newinfo->underflow[h])
1598                         newinfo->underflow[h] -= origsize - *size;
1599         }
1600         return ret;
1601 }
1602
1603 static int
1604 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1605 {
1606         struct xt_entry_match *ematch;
1607         struct xt_mtchk_param mtpar;
1608         unsigned int j;
1609         int ret = 0;
1610
1611         j = 0;
1612         mtpar.net       = net;
1613         mtpar.table     = name;
1614         mtpar.entryinfo = &e->ip;
1615         mtpar.hook_mask = e->comefrom;
1616         mtpar.family    = NFPROTO_IPV4;
1617         xt_ematch_foreach(ematch, e) {
1618                 ret = check_match(ematch, &mtpar);
1619                 if (ret != 0)
1620                         goto cleanup_matches;
1621                 ++j;
1622         }
1623
1624         ret = check_target(e, net, name);
1625         if (ret)
1626                 goto cleanup_matches;
1627         return 0;
1628
1629  cleanup_matches:
1630         xt_ematch_foreach(ematch, e) {
1631                 if (j-- == 0)
1632                         break;
1633                 cleanup_match(ematch, net);
1634         }
1635         return ret;
1636 }
1637
1638 static int
1639 translate_compat_table(struct net *net,
1640                        const char *name,
1641                        unsigned int valid_hooks,
1642                        struct xt_table_info **pinfo,
1643                        void **pentry0,
1644                        unsigned int total_size,
1645                        unsigned int number,
1646                        unsigned int *hook_entries,
1647                        unsigned int *underflows)
1648 {
1649         unsigned int i, j;
1650         struct xt_table_info *newinfo, *info;
1651         void *pos, *entry0, *entry1;
1652         struct compat_ipt_entry *iter0;
1653         struct ipt_entry *iter1;
1654         unsigned int size;
1655         int ret;
1656
1657         info = *pinfo;
1658         entry0 = *pentry0;
1659         size = total_size;
1660         info->number = number;
1661
1662         /* Init all hooks to impossible value. */
1663         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1664                 info->hook_entry[i] = 0xFFFFFFFF;
1665                 info->underflow[i] = 0xFFFFFFFF;
1666         }
1667
1668         duprintf("translate_compat_table: size %u\n", info->size);
1669         j = 0;
1670         xt_compat_lock(AF_INET);
1671         xt_compat_init_offsets(AF_INET, number);
1672         /* Walk through entries, checking offsets. */
1673         xt_entry_foreach(iter0, entry0, total_size) {
1674                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1675                                                         entry0,
1676                                                         entry0 + total_size,
1677                                                         hook_entries,
1678                                                         underflows,
1679                                                         name);
1680                 if (ret != 0)
1681                         goto out_unlock;
1682                 ++j;
1683         }
1684
1685         ret = -EINVAL;
1686         if (j != number) {
1687                 duprintf("translate_compat_table: %u not %u entries\n",
1688                          j, number);
1689                 goto out_unlock;
1690         }
1691
1692         /* Check hooks all assigned */
1693         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1694                 /* Only hooks which are valid */
1695                 if (!(valid_hooks & (1 << i)))
1696                         continue;
1697                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1698                         duprintf("Invalid hook entry %u %u\n",
1699                                  i, hook_entries[i]);
1700                         goto out_unlock;
1701                 }
1702                 if (info->underflow[i] == 0xFFFFFFFF) {
1703                         duprintf("Invalid underflow %u %u\n",
1704                                  i, underflows[i]);
1705                         goto out_unlock;
1706                 }
1707         }
1708
1709         ret = -ENOMEM;
1710         newinfo = xt_alloc_table_info(size);
1711         if (!newinfo)
1712                 goto out_unlock;
1713
1714         newinfo->number = number;
1715         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1716                 newinfo->hook_entry[i] = info->hook_entry[i];
1717                 newinfo->underflow[i] = info->underflow[i];
1718         }
1719         entry1 = newinfo->entries[raw_smp_processor_id()];
1720         pos = entry1;
1721         size = total_size;
1722         xt_entry_foreach(iter0, entry0, total_size) {
1723                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1724                                                   name, newinfo, entry1);
1725                 if (ret != 0)
1726                         break;
1727         }
1728         xt_compat_flush_offsets(AF_INET);
1729         xt_compat_unlock(AF_INET);
1730         if (ret)
1731                 goto free_newinfo;
1732
1733         ret = -ELOOP;
1734         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1735                 goto free_newinfo;
1736
1737         i = 0;
1738         xt_entry_foreach(iter1, entry1, newinfo->size) {
1739                 ret = compat_check_entry(iter1, net, name);
1740                 if (ret != 0)
1741                         break;
1742                 ++i;
1743                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1744                     XT_ERROR_TARGET) == 0)
1745                         ++newinfo->stacksize;
1746         }
1747         if (ret) {
1748                 /*
1749                  * The first i matches need cleanup_entry (calls ->destroy)
1750                  * because they had called ->check already. The other j-i
1751                  * entries need only release.
1752                  */
1753                 int skip = i;
1754                 j -= i;
1755                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1756                         if (skip-- > 0)
1757                                 continue;
1758                         if (j-- == 0)
1759                                 break;
1760                         compat_release_entry(iter0);
1761                 }
1762                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1763                         if (i-- == 0)
1764                                 break;
1765                         cleanup_entry(iter1, net);
1766                 }
1767                 xt_free_table_info(newinfo);
1768                 return ret;
1769         }
1770
1771         /* And one copy for every other CPU */
1772         for_each_possible_cpu(i)
1773                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1774                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1775
1776         *pinfo = newinfo;
1777         *pentry0 = entry1;
1778         xt_free_table_info(info);
1779         return 0;
1780
1781 free_newinfo:
1782         xt_free_table_info(newinfo);
1783 out:
1784         xt_entry_foreach(iter0, entry0, total_size) {
1785                 if (j-- == 0)
1786                         break;
1787                 compat_release_entry(iter0);
1788         }
1789         return ret;
1790 out_unlock:
1791         xt_compat_flush_offsets(AF_INET);
1792         xt_compat_unlock(AF_INET);
1793         goto out;
1794 }
1795
1796 static int
1797 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1798 {
1799         int ret;
1800         struct compat_ipt_replace tmp;
1801         struct xt_table_info *newinfo;
1802         void *loc_cpu_entry;
1803         struct ipt_entry *iter;
1804
1805         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1806                 return -EFAULT;
1807
1808         /* overflow check */
1809         if (tmp.size >= INT_MAX / num_possible_cpus())
1810                 return -ENOMEM;
1811         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1812                 return -ENOMEM;
1813         tmp.name[sizeof(tmp.name)-1] = 0;
1814
1815         newinfo = xt_alloc_table_info(tmp.size);
1816         if (!newinfo)
1817                 return -ENOMEM;
1818
1819         /* choose the copy that is on our node/cpu */
1820         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1821         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1822                            tmp.size) != 0) {
1823                 ret = -EFAULT;
1824                 goto free_newinfo;
1825         }
1826
1827         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1828                                      &newinfo, &loc_cpu_entry, tmp.size,
1829                                      tmp.num_entries, tmp.hook_entry,
1830                                      tmp.underflow);
1831         if (ret != 0)
1832                 goto free_newinfo;
1833
1834         duprintf("compat_do_replace: Translated table\n");
1835
1836         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1837                            tmp.num_counters, compat_ptr(tmp.counters));
1838         if (ret)
1839                 goto free_newinfo_untrans;
1840         return 0;
1841
1842  free_newinfo_untrans:
1843         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1844                 cleanup_entry(iter, net);
1845  free_newinfo:
1846         xt_free_table_info(newinfo);
1847         return ret;
1848 }
1849
1850 static int
1851 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1852                       unsigned int len)
1853 {
1854         int ret;
1855
1856         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1857                 return -EPERM;
1858
1859         switch (cmd) {
1860         case IPT_SO_SET_REPLACE:
1861                 ret = compat_do_replace(sock_net(sk), user, len);
1862                 break;
1863
1864         case IPT_SO_SET_ADD_COUNTERS:
1865                 ret = do_add_counters(sock_net(sk), user, len, 1);
1866                 break;
1867
1868         default:
1869                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1870                 ret = -EINVAL;
1871         }
1872
1873         return ret;
1874 }
1875
1876 struct compat_ipt_get_entries {
1877         char name[XT_TABLE_MAXNAMELEN];
1878         compat_uint_t size;
1879         struct compat_ipt_entry entrytable[0];
1880 };
1881
1882 static int
1883 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1884                             void __user *userptr)
1885 {
1886         struct xt_counters *counters;
1887         const struct xt_table_info *private = table->private;
1888         void __user *pos;
1889         unsigned int size;
1890         int ret = 0;
1891         const void *loc_cpu_entry;
1892         unsigned int i = 0;
1893         struct ipt_entry *iter;
1894
1895         counters = alloc_counters(table);
1896         if (IS_ERR(counters))
1897                 return PTR_ERR(counters);
1898
1899         /* choose the copy that is on our node/cpu, ...
1900          * This choice is lazy (because current thread is
1901          * allowed to migrate to another cpu)
1902          */
1903         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1904         pos = userptr;
1905         size = total_size;
1906         xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1907                 ret = compat_copy_entry_to_user(iter, &pos,
1908                                                 &size, counters, i++);
1909                 if (ret != 0)
1910                         break;
1911         }
1912
1913         vfree(counters);
1914         return ret;
1915 }
1916
1917 static int
1918 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1919                    int *len)
1920 {
1921         int ret;
1922         struct compat_ipt_get_entries get;
1923         struct xt_table *t;
1924
1925         if (*len < sizeof(get)) {
1926                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1927                 return -EINVAL;
1928         }
1929
1930         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1931                 return -EFAULT;
1932
1933         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1934                 duprintf("compat_get_entries: %u != %zu\n",
1935                          *len, sizeof(get) + get.size);
1936                 return -EINVAL;
1937         }
1938
1939         xt_compat_lock(AF_INET);
1940         t = xt_find_table_lock(net, AF_INET, get.name);
1941         if (!IS_ERR_OR_NULL(t)) {
1942                 const struct xt_table_info *private = t->private;
1943                 struct xt_table_info info;
1944                 duprintf("t->private->number = %u\n", private->number);
1945                 ret = compat_table_info(private, &info);
1946                 if (!ret && get.size == info.size) {
1947                         ret = compat_copy_entries_to_user(private->size,
1948                                                           t, uptr->entrytable);
1949                 } else if (!ret) {
1950                         duprintf("compat_get_entries: I've got %u not %u!\n",
1951                                  private->size, get.size);
1952                         ret = -EAGAIN;
1953                 }
1954                 xt_compat_flush_offsets(AF_INET);
1955                 module_put(t->me);
1956                 xt_table_unlock(t);
1957         } else
1958                 ret = t ? PTR_ERR(t) : -ENOENT;
1959
1960         xt_compat_unlock(AF_INET);
1961         return ret;
1962 }
1963
1964 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1965
1966 static int
1967 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1968 {
1969         int ret;
1970
1971         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1972                 return -EPERM;
1973
1974         switch (cmd) {
1975         case IPT_SO_GET_INFO:
1976                 ret = get_info(sock_net(sk), user, len, 1);
1977                 break;
1978         case IPT_SO_GET_ENTRIES:
1979                 ret = compat_get_entries(sock_net(sk), user, len);
1980                 break;
1981         default:
1982                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1983         }
1984         return ret;
1985 }
1986 #endif
1987
1988 static int
1989 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1990 {
1991         int ret;
1992
1993         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1994                 return -EPERM;
1995
1996         switch (cmd) {
1997         case IPT_SO_SET_REPLACE:
1998                 ret = do_replace(sock_net(sk), user, len);
1999                 break;
2000
2001         case IPT_SO_SET_ADD_COUNTERS:
2002                 ret = do_add_counters(sock_net(sk), user, len, 0);
2003                 break;
2004
2005         default:
2006                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2007                 ret = -EINVAL;
2008         }
2009
2010         return ret;
2011 }
2012
2013 static int
2014 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2015 {
2016         int ret;
2017
2018         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2019                 return -EPERM;
2020
2021         switch (cmd) {
2022         case IPT_SO_GET_INFO:
2023                 ret = get_info(sock_net(sk), user, len, 0);
2024                 break;
2025
2026         case IPT_SO_GET_ENTRIES:
2027                 ret = get_entries(sock_net(sk), user, len);
2028                 break;
2029
2030         case IPT_SO_GET_REVISION_MATCH:
2031         case IPT_SO_GET_REVISION_TARGET: {
2032                 struct xt_get_revision rev;
2033                 int target;
2034
2035                 if (*len != sizeof(rev)) {
2036                         ret = -EINVAL;
2037                         break;
2038                 }
2039                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2040                         ret = -EFAULT;
2041                         break;
2042                 }
2043                 rev.name[sizeof(rev.name)-1] = 0;
2044
2045                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2046                         target = 1;
2047                 else
2048                         target = 0;
2049
2050                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2051                                                          rev.revision,
2052                                                          target, &ret),
2053                                         "ipt_%s", rev.name);
2054                 break;
2055         }
2056
2057         default:
2058                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2059                 ret = -EINVAL;
2060         }
2061
2062         return ret;
2063 }
2064
2065 struct xt_table *ipt_register_table(struct net *net,
2066                                     const struct xt_table *table,
2067                                     const struct ipt_replace *repl)
2068 {
2069         int ret;
2070         struct xt_table_info *newinfo;
2071         struct xt_table_info bootstrap = {0};
2072         void *loc_cpu_entry;
2073         struct xt_table *new_table;
2074
2075         newinfo = xt_alloc_table_info(repl->size);
2076         if (!newinfo) {
2077                 ret = -ENOMEM;
2078                 goto out;
2079         }
2080
2081         /* choose the copy on our node/cpu, but dont care about preemption */
2082         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2083         memcpy(loc_cpu_entry, repl->entries, repl->size);
2084
2085         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2086         if (ret != 0)
2087                 goto out_free;
2088
2089         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2090         if (IS_ERR(new_table)) {
2091                 ret = PTR_ERR(new_table);
2092                 goto out_free;
2093         }
2094
2095         return new_table;
2096
2097 out_free:
2098         xt_free_table_info(newinfo);
2099 out:
2100         return ERR_PTR(ret);
2101 }
2102
2103 void ipt_unregister_table(struct net *net, struct xt_table *table)
2104 {
2105         struct xt_table_info *private;
2106         void *loc_cpu_entry;
2107         struct module *table_owner = table->me;
2108         struct ipt_entry *iter;
2109
2110         private = xt_unregister_table(table);
2111
2112         /* Decrease module usage counts and free resources */
2113         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2114         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2115                 cleanup_entry(iter, net);
2116         if (private->number > private->initial_entries)
2117                 module_put(table_owner);
2118         xt_free_table_info(private);
2119 }
2120
2121 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2122 static inline bool
2123 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2124                      u_int8_t type, u_int8_t code,
2125                      bool invert)
2126 {
2127         return ((test_type == 0xFF) ||
2128                 (type == test_type && code >= min_code && code <= max_code))
2129                 ^ invert;
2130 }
2131
2132 static bool
2133 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2134 {
2135         const struct icmphdr *ic;
2136         struct icmphdr _icmph;
2137         const struct ipt_icmp *icmpinfo = par->matchinfo;
2138
2139         /* Must not be a fragment. */
2140         if (par->fragoff != 0)
2141                 return false;
2142
2143         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2144         if (ic == NULL) {
2145                 /* We've been asked to examine this packet, and we
2146                  * can't.  Hence, no choice but to drop.
2147                  */
2148                 duprintf("Dropping evil ICMP tinygram.\n");
2149                 par->hotdrop = true;
2150                 return false;
2151         }
2152
2153         return icmp_type_code_match(icmpinfo->type,
2154                                     icmpinfo->code[0],
2155                                     icmpinfo->code[1],
2156                                     ic->type, ic->code,
2157                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2158 }
2159
2160 static int icmp_checkentry(const struct xt_mtchk_param *par)
2161 {
2162         const struct ipt_icmp *icmpinfo = par->matchinfo;
2163
2164         /* Must specify no unknown invflags */
2165         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2166 }
2167
2168 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2169         {
2170                 .name             = XT_STANDARD_TARGET,
2171                 .targetsize       = sizeof(int),
2172                 .family           = NFPROTO_IPV4,
2173 #ifdef CONFIG_COMPAT
2174                 .compatsize       = sizeof(compat_int_t),
2175                 .compat_from_user = compat_standard_from_user,
2176                 .compat_to_user   = compat_standard_to_user,
2177 #endif
2178         },
2179         {
2180                 .name             = XT_ERROR_TARGET,
2181                 .target           = ipt_error,
2182                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2183                 .family           = NFPROTO_IPV4,
2184         },
2185 };
2186
2187 static struct nf_sockopt_ops ipt_sockopts = {
2188         .pf             = PF_INET,
2189         .set_optmin     = IPT_BASE_CTL,
2190         .set_optmax     = IPT_SO_SET_MAX+1,
2191         .set            = do_ipt_set_ctl,
2192 #ifdef CONFIG_COMPAT
2193         .compat_set     = compat_do_ipt_set_ctl,
2194 #endif
2195         .get_optmin     = IPT_BASE_CTL,
2196         .get_optmax     = IPT_SO_GET_MAX+1,
2197         .get            = do_ipt_get_ctl,
2198 #ifdef CONFIG_COMPAT
2199         .compat_get     = compat_do_ipt_get_ctl,
2200 #endif
2201         .owner          = THIS_MODULE,
2202 };
2203
2204 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2205         {
2206                 .name       = "icmp",
2207                 .match      = icmp_match,
2208                 .matchsize  = sizeof(struct ipt_icmp),
2209                 .checkentry = icmp_checkentry,
2210                 .proto      = IPPROTO_ICMP,
2211                 .family     = NFPROTO_IPV4,
2212         },
2213 };
2214
2215 static int __net_init ip_tables_net_init(struct net *net)
2216 {
2217         return xt_proto_init(net, NFPROTO_IPV4);
2218 }
2219
2220 static void __net_exit ip_tables_net_exit(struct net *net)
2221 {
2222         xt_proto_fini(net, NFPROTO_IPV4);
2223 }
2224
2225 static struct pernet_operations ip_tables_net_ops = {
2226         .init = ip_tables_net_init,
2227         .exit = ip_tables_net_exit,
2228 };
2229
2230 static int __init ip_tables_init(void)
2231 {
2232         int ret;
2233
2234         ret = register_pernet_subsys(&ip_tables_net_ops);
2235         if (ret < 0)
2236                 goto err1;
2237
2238         /* No one else will be downing sem now, so we won't sleep */
2239         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2240         if (ret < 0)
2241                 goto err2;
2242         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2243         if (ret < 0)
2244                 goto err4;
2245
2246         /* Register setsockopt */
2247         ret = nf_register_sockopt(&ipt_sockopts);
2248         if (ret < 0)
2249                 goto err5;
2250
2251         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2252         return 0;
2253
2254 err5:
2255         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2256 err4:
2257         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2258 err2:
2259         unregister_pernet_subsys(&ip_tables_net_ops);
2260 err1:
2261         return ret;
2262 }
2263
2264 static void __exit ip_tables_fini(void)
2265 {
2266         nf_unregister_sockopt(&ipt_sockopts);
2267
2268         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2269         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2270         unregister_pernet_subsys(&ip_tables_net_ops);
2271 }
2272
2273 EXPORT_SYMBOL(ipt_register_table);
2274 EXPORT_SYMBOL(ipt_unregister_table);
2275 EXPORT_SYMBOL(ipt_do_table);
2276 module_init(ip_tables_init);
2277 module_exit(ip_tables_fini);