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_entry *e)
 172 {
 173         static const struct ipt_ip uncond;
 174
 175         return e->target_offset == sizeof(struct ipt_entry) &&
 176                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
 177 #undef FWINV
 178 }
 179
 180 /* for const-correctness */
 181 static inline const struct xt_entry_target *
 182 ipt_get_target_c(const struct ipt_entry *e)
 183 {
 184         return ipt_get_target((struct ipt_entry *)e);
 185 }
 186
 187 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 188 static const char *const hooknames[] = {
 189         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 190         [NF_INET_LOCAL_IN]              = "INPUT",
 191         [NF_INET_FORWARD]               = "FORWARD",
 192         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 193         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 194 };
 195
 196 enum nf_ip_trace_comments {
 197         NF_IP_TRACE_COMMENT_RULE,
 198         NF_IP_TRACE_COMMENT_RETURN,
 199         NF_IP_TRACE_COMMENT_POLICY,
 200 };
 201
 202 static const char *const comments[] = {
 203         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 204         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 205         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 206 };
 207
 208 static struct nf_loginfo trace_loginfo = {
 209         .type = NF_LOG_TYPE_LOG,
 210         .u = {
 211                 .log = {
 212                         .level = 4,
 213                         .logflags = NF_LOG_MASK,
 214                 },
 215         },
 216 };
 217
 218 /* Mildly perf critical (only if packet tracing is on) */
 219 static inline int
 220 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 221                       const char *hookname, const char **chainname,
 222                       const char **comment, unsigned int *rulenum)
 223 {
 224         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 225
 226         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 227                 /* Head of user chain: ERROR target with chainname */
 228                 *chainname = t->target.data;
 229                 (*rulenum) = 0;
 230         } else if (s == e) {
 231                 (*rulenum)++;
 232
 233                 if (unconditional(s) &&
 234                     strcmp(t->target.u.kernel.target->name,
 235                            XT_STANDARD_TARGET) == 0 &&
 236                    t->verdict < 0) {
 237                         /* Tail of chains: STANDARD target (return/policy) */
 238                         *comment = *chainname == hookname
 239                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 240                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 241                 }
 242                 return 1;
 243         } else
 244                 (*rulenum)++;
 245
 246         return 0;
 247 }
 248
 249 static void trace_packet(struct net *net,
 250                          const struct sk_buff *skb,
 251                          unsigned int hook,
 252                          const struct net_device *in,
 253                          const struct net_device *out,
 254                          const char *tablename,
 255                          const struct xt_table_info *private,
 256                          const struct ipt_entry *e)
 257 {
 258         const struct ipt_entry *root;
 259         const char *hookname, *chainname, *comment;
 260         const struct ipt_entry *iter;
 261         unsigned int rulenum = 0;
 262
 263         root = get_entry(private->entries, private->hook_entry[hook]);
 264
 265         hookname = chainname = hooknames[hook];
 266         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 267
 268         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 269                 if (get_chainname_rulenum(iter, e, hookname,
 270                     &chainname, &comment, &rulenum) != 0)
 271                         break;
 272
 273         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 274                      "TRACE: %s:%s:%s:%u ",
 275                      tablename, chainname, comment, rulenum);
 276 }
 277 #endif
 278
 279 static inline
 280 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 281 {
 282         return (void *)entry + entry->next_offset;
 283 }
 284
 285 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 286 unsigned int
 287 ipt_do_table(struct sk_buff *skb,
 288              const struct nf_hook_state *state,
 289              struct xt_table *table)
 290 {
 291         unsigned int hook = state->hook;
 292         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 293         const struct iphdr *ip;
 294         /* Initializing verdict to NF_DROP keeps gcc happy. */
 295         unsigned int verdict = NF_DROP;
 296         const char *indev, *outdev;
 297         const void *table_base;
 298         struct ipt_entry *e, **jumpstack;
 299         unsigned int stackidx, cpu;
 300         const struct xt_table_info *private;
 301         struct xt_action_param acpar;
 302         unsigned int addend;
 303
 304         /* Initialization */
 305         stackidx = 0;
 306         ip = ip_hdr(skb);
 307         indev = state->in ? state->in->name : nulldevname;
 308         outdev = state->out ? state->out->name : nulldevname;
 309         /* We handle fragments by dealing with the first fragment as
 310          * if it was a normal packet.  All other fragments are treated
 311          * normally, except that they will NEVER match rules that ask
 312          * things we don't know, ie. tcp syn flag or ports).  If the
 313          * rule is also a fragment-specific rule, non-fragments won't
 314          * match it. */
 315         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 316         acpar.thoff   = ip_hdrlen(skb);
 317         acpar.hotdrop = false;
 318         acpar.net     = state->net;
 319         acpar.in      = state->in;
 320         acpar.out     = state->out;
 321         acpar.family  = NFPROTO_IPV4;
 322         acpar.hooknum = hook;
 323
 324         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 325         local_bh_disable();
 326         addend = xt_write_recseq_begin();
 327         private = table->private;
 328         cpu        = smp_processor_id();
 329         /*
 330          * Ensure we load private-> members after we've fetched the base
 331          * pointer.
 332          */
 333         smp_read_barrier_depends();
 334         table_base = private->entries;
 335         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 336
 337         /* Switch to alternate jumpstack if we're being invoked via TEE.
 338          * TEE issues XT_CONTINUE verdict on original skb so we must not
 339          * clobber the jumpstack.
 340          *
 341          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 342          * but it is no problem since absolute verdict is issued by these.
 343          */
 344         if (static_key_false(&xt_tee_enabled))
 345                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 346
 347         e = get_entry(table_base, private->hook_entry[hook]);
 348
 349         pr_debug("Entering %s(hook %u), UF %p\n",
 350                  table->name, hook,
 351                  get_entry(table_base, private->underflow[hook]));
 352
 353         do {
 354                 const struct xt_entry_target *t;
 355                 const struct xt_entry_match *ematch;
 356                 struct xt_counters *counter;
 357
 358                 IP_NF_ASSERT(e);
 359                 if (!ip_packet_match(ip, indev, outdev,
 360                     &e->ip, acpar.fragoff)) {
 361  no_match:
 362                         e = ipt_next_entry(e);
 363                         continue;
 364                 }
 365
 366                 xt_ematch_foreach(ematch, e) {
 367                         acpar.match     = ematch->u.kernel.match;
 368                         acpar.matchinfo = ematch->data;
 369                         if (!acpar.match->match(skb, &acpar))
 370                                 goto no_match;
 371                 }
 372
 373                 counter = xt_get_this_cpu_counter(&e->counters);
 374                 ADD_COUNTER(*counter, skb->len, 1);
 375
 376                 t = ipt_get_target(e);
 377                 IP_NF_ASSERT(t->u.kernel.target);
 378
 379 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 380                 /* The packet is traced: log it */
 381                 if (unlikely(skb->nf_trace))
 382                         trace_packet(state->net, skb, hook, state->in,
 383                                      state->out, table->name, private, e);
 384 #endif
 385                 /* Standard target? */
 386                 if (!t->u.kernel.target->target) {
 387                         int v;
 388
 389                         v = ((struct xt_standard_target *)t)->verdict;
 390                         if (v < 0) {
 391                                 /* Pop from stack? */
 392                                 if (v != XT_RETURN) {
 393                                         verdict = (unsigned int)(-v) - 1;
 394                                         break;
 395                                 }
 396                                 if (stackidx == 0) {
 397                                         e = get_entry(table_base,
 398                                             private->underflow[hook]);
 399                                         pr_debug("Underflow (this is normal) "
 400                                                  "to %p\n", e);
 401                                 } else {
 402                                         e = jumpstack[--stackidx];
 403                                         pr_debug("Pulled %p out from pos %u\n",
 404                                                  e, stackidx);
 405                                         e = ipt_next_entry(e);
 406                                 }
 407                                 continue;
 408                         }
 409                         if (table_base + v != ipt_next_entry(e) &&
 410                             !(e->ip.flags & IPT_F_GOTO)) {
 411                                 jumpstack[stackidx++] = e;
 412                                 pr_debug("Pushed %p into pos %u\n",
 413                                          e, stackidx - 1);
 414                         }
 415
 416                         e = get_entry(table_base, v);
 417                         continue;
 418                 }
 419
 420                 acpar.target   = t->u.kernel.target;
 421                 acpar.targinfo = t->data;
 422
 423                 verdict = t->u.kernel.target->target(skb, &acpar);
 424                 /* Target might have changed stuff. */
 425                 ip = ip_hdr(skb);
 426                 if (verdict == XT_CONTINUE)
 427                         e = ipt_next_entry(e);
 428                 else
 429                         /* Verdict */
 430                         break;
 431         } while (!acpar.hotdrop);
 432         pr_debug("Exiting %s; sp at %u\n", __func__, stackidx);
 433
 434         xt_write_recseq_end(addend);
 435         local_bh_enable();
 436
 437 #ifdef DEBUG_ALLOW_ALL
 438         return NF_ACCEPT;
 439 #else
 440         if (acpar.hotdrop)
 441                 return NF_DROP;
 442         else return verdict;
 443 #endif
 444 }
 445
 446 static bool find_jump_target(const struct xt_table_info *t,
 447                              const struct ipt_entry *target)
 448 {
 449         struct ipt_entry *iter;
 450
 451         xt_entry_foreach(iter, t->entries, t->size) {
 452                  if (iter == target)
 453                         return true;
 454         }
 455         return false;
 456 }
 457
 458 /* Figures out from what hook each rule can be called: returns 0 if
 459    there are loops.  Puts hook bitmask in comefrom. */
 460 static int
 461 mark_source_chains(const struct xt_table_info *newinfo,
 462                    unsigned int valid_hooks, void *entry0)
 463 {
 464         unsigned int hook;
 465
 466         /* No recursion; use packet counter to save back ptrs (reset
 467            to 0 as we leave), and comefrom to save source hook bitmask */
 468         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 469                 unsigned int pos = newinfo->hook_entry[hook];
 470                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 471
 472                 if (!(valid_hooks & (1 << hook)))
 473                         continue;
 474
 475                 /* Set initial back pointer. */
 476                 e->counters.pcnt = pos;
 477
 478                 for (;;) {
 479                         const struct xt_standard_target *t
 480                                 = (void *)ipt_get_target_c(e);
 481                         int visited = e->comefrom & (1 << hook);
 482
 483                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 484                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 485                                        hook, pos, e->comefrom);
 486                                 return 0;
 487                         }
 488                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 489
 490                         /* Unconditional return/END. */
 491                         if ((unconditional(e) &&
 492                              (strcmp(t->target.u.user.name,
 493                                      XT_STANDARD_TARGET) == 0) &&
 494                              t->verdict < 0) || visited) {
 495                                 unsigned int oldpos, size;
 496
 497                                 if ((strcmp(t->target.u.user.name,
 498                                             XT_STANDARD_TARGET) == 0) &&
 499                                     t->verdict < -NF_MAX_VERDICT - 1) {
 500                                         duprintf("mark_source_chains: bad "
 501                                                 "negative verdict (%i)\n",
 502                                                                 t->verdict);
 503                                         return 0;
 504                                 }
 505
 506                                 /* Return: backtrack through the last
 507                                    big jump. */
 508                                 do {
 509                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 510 #ifdef DEBUG_IP_FIREWALL_USER
 511                                         if (e->comefrom
 512                                             & (1 << NF_INET_NUMHOOKS)) {
 513                                                 duprintf("Back unset "
 514                                                          "on hook %u "
 515                                                          "rule %u\n",
 516                                                          hook, pos);
 517                                         }
 518 #endif
 519                                         oldpos = pos;
 520                                         pos = e->counters.pcnt;
 521                                         e->counters.pcnt = 0;
 522
 523                                         /* We're at the start. */
 524                                         if (pos == oldpos)
 525                                                 goto next;
 526
 527                                         e = (struct ipt_entry *)
 528                                                 (entry0 + pos);
 529                                 } while (oldpos == pos + e->next_offset);
 530
 531                                 /* Move along one */
 532                                 size = e->next_offset;
 533                                 e = (struct ipt_entry *)
 534                                         (entry0 + pos + size);
 535                                 if (pos + size >= newinfo->size)
 536                                         return 0;
 537                                 e->counters.pcnt = pos;
 538                                 pos += size;
 539                         } else {
 540                                 int newpos = t->verdict;
 541
 542                                 if (strcmp(t->target.u.user.name,
 543                                            XT_STANDARD_TARGET) == 0 &&
 544                                     newpos >= 0) {
 545                                         if (newpos > newinfo->size -
 546                                                 sizeof(struct ipt_entry)) {
 547                                                 duprintf("mark_source_chains: "
 548                                                         "bad verdict (%i)\n",
 549                                                                 newpos);
 550                                                 return 0;
 551                                         }
 552                                         /* This a jump; chase it. */
 553                                         duprintf("Jump rule %u -> %u\n",
 554                                                  pos, newpos);
 555                                         e = (struct ipt_entry *)
 556                                                 (entry0 + newpos);
 557                                         if (!find_jump_target(newinfo, e))
 558                                                 return 0;
 559                                 } else {
 560                                         /* ... this is a fallthru */
 561                                         newpos = pos + e->next_offset;
 562                                         if (newpos >= newinfo->size)
 563                                                 return 0;
 564                                 }
 565                                 e = (struct ipt_entry *)
 566                                         (entry0 + newpos);
 567                                 e->counters.pcnt = pos;
 568                                 pos = newpos;
 569                         }
 570                 }
 571 next:
 572                 duprintf("Finished chain %u\n", hook);
 573         }
 574         return 1;
 575 }
 576
 577 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 578 {
 579         struct xt_mtdtor_param par;
 580
 581         par.net       = net;
 582         par.match     = m->u.kernel.match;
 583         par.matchinfo = m->data;
 584         par.family    = NFPROTO_IPV4;
 585         if (par.match->destroy != NULL)
 586                 par.match->destroy(&par);
 587         module_put(par.match->me);
 588 }
 589
 590 static int
 591 check_entry(const struct ipt_entry *e)
 592 {
 593         if (!ip_checkentry(&e->ip))
 594                 return -EINVAL;
 595
 596         return xt_check_entry_offsets(e, e->target_offset, e->next_offset);
 597 }
 598
 599 static int
 600 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 601 {
 602         const struct ipt_ip *ip = par->entryinfo;
 603         int ret;
 604
 605         par->match     = m->u.kernel.match;
 606         par->matchinfo = m->data;
 607
 608         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 609               ip->proto, ip->invflags & IPT_INV_PROTO);
 610         if (ret < 0) {
 611                 duprintf("check failed for `%s'.\n", par->match->name);
 612                 return ret;
 613         }
 614         return 0;
 615 }
 616
 617 static int
 618 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 619 {
 620         struct xt_match *match;
 621         int ret;
 622
 623         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 624                                       m->u.user.revision);
 625         if (IS_ERR(match)) {
 626                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 627                 return PTR_ERR(match);
 628         }
 629         m->u.kernel.match = match;
 630
 631         ret = check_match(m, par);
 632         if (ret)
 633                 goto err;
 634
 635         return 0;
 636 err:
 637         module_put(m->u.kernel.match->me);
 638         return ret;
 639 }
 640
 641 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 642 {
 643         struct xt_entry_target *t = ipt_get_target(e);
 644         struct xt_tgchk_param par = {
 645                 .net       = net,
 646                 .table     = name,
 647                 .entryinfo = e,
 648                 .target    = t->u.kernel.target,
 649                 .targinfo  = t->data,
 650                 .hook_mask = e->comefrom,
 651                 .family    = NFPROTO_IPV4,
 652         };
 653         int ret;
 654
 655         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 656               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 657         if (ret < 0) {
 658                 duprintf("check failed for `%s'.\n",
 659                          t->u.kernel.target->name);
 660                 return ret;
 661         }
 662         return 0;
 663 }
 664
 665 static int
 666 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 667                  unsigned int size)
 668 {
 669         struct xt_entry_target *t;
 670         struct xt_target *target;
 671         int ret;
 672         unsigned int j;
 673         struct xt_mtchk_param mtpar;
 674         struct xt_entry_match *ematch;
 675
 676         e->counters.pcnt = xt_percpu_counter_alloc();
 677         if (IS_ERR_VALUE(e->counters.pcnt))
 678                 return -ENOMEM;
 679
 680         j = 0;
 681         mtpar.net       = net;
 682         mtpar.table     = name;
 683         mtpar.entryinfo = &e->ip;
 684         mtpar.hook_mask = e->comefrom;
 685         mtpar.family    = NFPROTO_IPV4;
 686         xt_ematch_foreach(ematch, e) {
 687                 ret = find_check_match(ematch, &mtpar);
 688                 if (ret != 0)
 689                         goto cleanup_matches;
 690                 ++j;
 691         }
 692
 693         t = ipt_get_target(e);
 694         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 695                                         t->u.user.revision);
 696         if (IS_ERR(target)) {
 697                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 698                 ret = PTR_ERR(target);
 699                 goto cleanup_matches;
 700         }
 701         t->u.kernel.target = target;
 702
 703         ret = check_target(e, net, name);
 704         if (ret)
 705                 goto err;
 706
 707         return 0;
 708  err:
 709         module_put(t->u.kernel.target->me);
 710  cleanup_matches:
 711         xt_ematch_foreach(ematch, e) {
 712                 if (j-- == 0)
 713                         break;
 714                 cleanup_match(ematch, net);
 715         }
 716
 717         xt_percpu_counter_free(e->counters.pcnt);
 718
 719         return ret;
 720 }
 721
 722 static bool check_underflow(const struct ipt_entry *e)
 723 {
 724         const struct xt_entry_target *t;
 725         unsigned int verdict;
 726
 727         if (!unconditional(e))
 728                 return false;
 729         t = ipt_get_target_c(e);
 730         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 731                 return false;
 732         verdict = ((struct xt_standard_target *)t)->verdict;
 733         verdict = -verdict - 1;
 734         return verdict == NF_DROP || verdict == NF_ACCEPT;
 735 }
 736
 737 static int
 738 check_entry_size_and_hooks(struct ipt_entry *e,
 739                            struct xt_table_info *newinfo,
 740                            const unsigned char *base,
 741                            const unsigned char *limit,
 742                            const unsigned int *hook_entries,
 743                            const unsigned int *underflows,
 744                            unsigned int valid_hooks)
 745 {
 746         unsigned int h;
 747         int err;
 748
 749         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 750             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 751             (unsigned char *)e + e->next_offset > limit) {
 752                 duprintf("Bad offset %p\n", e);
 753                 return -EINVAL;
 754         }
 755
 756         if (e->next_offset
 757             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 758                 duprintf("checking: element %p size %u\n",
 759                          e, e->next_offset);
 760                 return -EINVAL;
 761         }
 762
 763         err = check_entry(e);
 764         if (err)
 765                 return err;
 766
 767         /* Check hooks & underflows */
 768         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 769                 if (!(valid_hooks & (1 << h)))
 770                         continue;
 771                 if ((unsigned char *)e - base == hook_entries[h])
 772                         newinfo->hook_entry[h] = hook_entries[h];
 773                 if ((unsigned char *)e - base == underflows[h]) {
 774                         if (!check_underflow(e)) {
 775                                 pr_debug("Underflows must be unconditional and "
 776                                          "use the STANDARD target with "
 777                                          "ACCEPT/DROP\n");
 778                                 return -EINVAL;
 779                         }
 780                         newinfo->underflow[h] = underflows[h];
 781                 }
 782         }
 783
 784         /* Clear counters and comefrom */
 785         e->counters = ((struct xt_counters) { 0, 0 });
 786         e->comefrom = 0;
 787         return 0;
 788 }
 789
 790 static void
 791 cleanup_entry(struct ipt_entry *e, struct net *net)
 792 {
 793         struct xt_tgdtor_param par;
 794         struct xt_entry_target *t;
 795         struct xt_entry_match *ematch;
 796
 797         /* Cleanup all matches */
 798         xt_ematch_foreach(ematch, e)
 799                 cleanup_match(ematch, net);
 800         t = ipt_get_target(e);
 801
 802         par.net      = net;
 803         par.target   = t->u.kernel.target;
 804         par.targinfo = t->data;
 805         par.family   = NFPROTO_IPV4;
 806         if (par.target->destroy != NULL)
 807                 par.target->destroy(&par);
 808         module_put(par.target->me);
 809         xt_percpu_counter_free(e->counters.pcnt);
 810 }
 811
 812 /* Checks and translates the user-supplied table segment (held in
 813    newinfo) */
 814 static int
 815 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 816                 const struct ipt_replace *repl)
 817 {
 818         struct ipt_entry *iter;
 819         unsigned int i;
 820         int ret = 0;
 821
 822         newinfo->size = repl->size;
 823         newinfo->number = repl->num_entries;
 824
 825         /* Init all hooks to impossible value. */
 826         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 827                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 828                 newinfo->underflow[i] = 0xFFFFFFFF;
 829         }
 830
 831         duprintf("translate_table: size %u\n", newinfo->size);
 832         i = 0;
 833         /* Walk through entries, checking offsets. */
 834         xt_entry_foreach(iter, entry0, newinfo->size) {
 835                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 836                                                  entry0 + repl->size,
 837                                                  repl->hook_entry,
 838                                                  repl->underflow,
 839                                                  repl->valid_hooks);
 840                 if (ret != 0)
 841                         return ret;
 842                 ++i;
 843                 if (strcmp(ipt_get_target(iter)->u.user.name,
 844                     XT_ERROR_TARGET) == 0)
 845                         ++newinfo->stacksize;
 846         }
 847
 848         if (i != repl->num_entries) {
 849                 duprintf("translate_table: %u not %u entries\n",
 850                          i, repl->num_entries);
 851                 return -EINVAL;
 852         }
 853
 854         /* Check hooks all assigned */
 855         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 856                 /* Only hooks which are valid */
 857                 if (!(repl->valid_hooks & (1 << i)))
 858                         continue;
 859                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 860                         duprintf("Invalid hook entry %u %u\n",
 861                                  i, repl->hook_entry[i]);
 862                         return -EINVAL;
 863                 }
 864                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 865                         duprintf("Invalid underflow %u %u\n",
 866                                  i, repl->underflow[i]);
 867                         return -EINVAL;
 868                 }
 869         }
 870
 871         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 872                 return -ELOOP;
 873
 874         /* Finally, each sanity check must pass */
 875         i = 0;
 876         xt_entry_foreach(iter, entry0, newinfo->size) {
 877                 ret = find_check_entry(iter, net, repl->name, repl->size);
 878                 if (ret != 0)
 879                         break;
 880                 ++i;
 881         }
 882
 883         if (ret != 0) {
 884                 xt_entry_foreach(iter, entry0, newinfo->size) {
 885                         if (i-- == 0)
 886                                 break;
 887                         cleanup_entry(iter, net);
 888                 }
 889                 return ret;
 890         }
 891
 892         return ret;
 893 }
 894
 895 static void
 896 get_counters(const struct xt_table_info *t,
 897              struct xt_counters counters[])
 898 {
 899         struct ipt_entry *iter;
 900         unsigned int cpu;
 901         unsigned int i;
 902
 903         for_each_possible_cpu(cpu) {
 904                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 905
 906                 i = 0;
 907                 xt_entry_foreach(iter, t->entries, t->size) {
 908                         struct xt_counters *tmp;
 909                         u64 bcnt, pcnt;
 910                         unsigned int start;
 911
 912                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 913                         do {
 914                                 start = read_seqcount_begin(s);
 915                                 bcnt = tmp->bcnt;
 916                                 pcnt = tmp->pcnt;
 917                         } while (read_seqcount_retry(s, start));
 918
 919                         ADD_COUNTER(counters[i], bcnt, pcnt);
 920                         ++i; /* macro does multi eval of i */
 921                 }
 922         }
 923 }
 924
 925 static struct xt_counters *alloc_counters(const struct xt_table *table)
 926 {
 927         unsigned int countersize;
 928         struct xt_counters *counters;
 929         const struct xt_table_info *private = table->private;
 930
 931         /* We need atomic snapshot of counters: rest doesn't change
 932            (other than comefrom, which userspace doesn't care
 933            about). */
 934         countersize = sizeof(struct xt_counters) * private->number;
 935         counters = vzalloc(countersize);
 936
 937         if (counters == NULL)
 938                 return ERR_PTR(-ENOMEM);
 939
 940         get_counters(private, counters);
 941
 942         return counters;
 943 }
 944
 945 static int
 946 copy_entries_to_user(unsigned int total_size,
 947                      const struct xt_table *table,
 948                      void __user *userptr)
 949 {
 950         unsigned int off, num;
 951         const struct ipt_entry *e;
 952         struct xt_counters *counters;
 953         const struct xt_table_info *private = table->private;
 954         int ret = 0;
 955         const void *loc_cpu_entry;
 956
 957         counters = alloc_counters(table);
 958         if (IS_ERR(counters))
 959                 return PTR_ERR(counters);
 960
 961         loc_cpu_entry = private->entries;
 962         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 963                 ret = -EFAULT;
 964                 goto free_counters;
 965         }
 966
 967         /* FIXME: use iterator macros --RR */
 968         /* ... then go back and fix counters and names */
 969         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 970                 unsigned int i;
 971                 const struct xt_entry_match *m;
 972                 const struct xt_entry_target *t;
 973
 974                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 975                 if (copy_to_user(userptr + off
 976                                  + offsetof(struct ipt_entry, counters),
 977                                  &counters[num],
 978                                  sizeof(counters[num])) != 0) {
 979                         ret = -EFAULT;
 980                         goto free_counters;
 981                 }
 982
 983                 for (i = sizeof(struct ipt_entry);
 984                      i < e->target_offset;
 985                      i += m->u.match_size) {
 986                         m = (void *)e + i;
 987
 988                         if (copy_to_user(userptr + off + i
 989                                          + offsetof(struct xt_entry_match,
 990                                                     u.user.name),
 991                                          m->u.kernel.match->name,
 992                                          strlen(m->u.kernel.match->name)+1)
 993                             != 0) {
 994                                 ret = -EFAULT;
 995                                 goto free_counters;
 996                         }
 997                 }
 998
 999                 t = ipt_get_target_c(e);
1000                 if (copy_to_user(userptr + off + e->target_offset
1001                                  + offsetof(struct xt_entry_target,
1002                                             u.user.name),
1003                                  t->u.kernel.target->name,
1004                                  strlen(t->u.kernel.target->name)+1) != 0) {
1005                         ret = -EFAULT;
1006                         goto free_counters;
1007                 }
1008         }
1009
1010  free_counters:
1011         vfree(counters);
1012         return ret;
1013 }
1014
1015 #ifdef CONFIG_COMPAT
1016 static void compat_standard_from_user(void *dst, const void *src)
1017 {
1018         int v = *(compat_int_t *)src;
1019
1020         if (v > 0)
1021                 v += xt_compat_calc_jump(AF_INET, v);
1022         memcpy(dst, &v, sizeof(v));
1023 }
1024
1025 static int compat_standard_to_user(void __user *dst, const void *src)
1026 {
1027         compat_int_t cv = *(int *)src;
1028
1029         if (cv > 0)
1030                 cv -= xt_compat_calc_jump(AF_INET, cv);
1031         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1032 }
1033
1034 static int compat_calc_entry(const struct ipt_entry *e,
1035                              const struct xt_table_info *info,
1036                              const void *base, struct xt_table_info *newinfo)
1037 {
1038         const struct xt_entry_match *ematch;
1039         const struct xt_entry_target *t;
1040         unsigned int entry_offset;
1041         int off, i, ret;
1042
1043         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1044         entry_offset = (void *)e - base;
1045         xt_ematch_foreach(ematch, e)
1046                 off += xt_compat_match_offset(ematch->u.kernel.match);
1047         t = ipt_get_target_c(e);
1048         off += xt_compat_target_offset(t->u.kernel.target);
1049         newinfo->size -= off;
1050         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1051         if (ret)
1052                 return ret;
1053
1054         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1055                 if (info->hook_entry[i] &&
1056                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1057                         newinfo->hook_entry[i] -= off;
1058                 if (info->underflow[i] &&
1059                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1060                         newinfo->underflow[i] -= off;
1061         }
1062         return 0;
1063 }
1064
1065 static int compat_table_info(const struct xt_table_info *info,
1066                              struct xt_table_info *newinfo)
1067 {
1068         struct ipt_entry *iter;
1069         const void *loc_cpu_entry;
1070         int ret;
1071
1072         if (!newinfo || !info)
1073                 return -EINVAL;
1074
1075         /* we dont care about newinfo->entries */
1076         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1077         newinfo->initial_entries = 0;
1078         loc_cpu_entry = info->entries;
1079         xt_compat_init_offsets(AF_INET, info->number);
1080         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1081                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1082                 if (ret != 0)
1083                         return ret;
1084         }
1085         return 0;
1086 }
1087 #endif
1088
1089 static int get_info(struct net *net, void __user *user,
1090                     const int *len, int compat)
1091 {
1092         char name[XT_TABLE_MAXNAMELEN];
1093         struct xt_table *t;
1094         int ret;
1095
1096         if (*len != sizeof(struct ipt_getinfo)) {
1097                 duprintf("length %u != %zu\n", *len,
1098                          sizeof(struct ipt_getinfo));
1099                 return -EINVAL;
1100         }
1101
1102         if (copy_from_user(name, user, sizeof(name)) != 0)
1103                 return -EFAULT;
1104
1105         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1106 #ifdef CONFIG_COMPAT
1107         if (compat)
1108                 xt_compat_lock(AF_INET);
1109 #endif
1110         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1111                                     "iptable_%s", name);
1112         if (!IS_ERR_OR_NULL(t)) {
1113                 struct ipt_getinfo info;
1114                 const struct xt_table_info *private = t->private;
1115 #ifdef CONFIG_COMPAT
1116                 struct xt_table_info tmp;
1117
1118                 if (compat) {
1119                         ret = compat_table_info(private, &tmp);
1120                         xt_compat_flush_offsets(AF_INET);
1121                         private = &tmp;
1122                 }
1123 #endif
1124                 memset(&info, 0, sizeof(info));
1125                 info.valid_hooks = t->valid_hooks;
1126                 memcpy(info.hook_entry, private->hook_entry,
1127                        sizeof(info.hook_entry));
1128                 memcpy(info.underflow, private->underflow,
1129                        sizeof(info.underflow));
1130                 info.num_entries = private->number;
1131                 info.size = private->size;
1132                 strcpy(info.name, name);
1133
1134                 if (copy_to_user(user, &info, *len) != 0)
1135                         ret = -EFAULT;
1136                 else
1137                         ret = 0;
1138
1139                 xt_table_unlock(t);
1140                 module_put(t->me);
1141         } else
1142                 ret = t ? PTR_ERR(t) : -ENOENT;
1143 #ifdef CONFIG_COMPAT
1144         if (compat)
1145                 xt_compat_unlock(AF_INET);
1146 #endif
1147         return ret;
1148 }
1149
1150 static int
1151 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1152             const int *len)
1153 {
1154         int ret;
1155         struct ipt_get_entries get;
1156         struct xt_table *t;
1157
1158         if (*len < sizeof(get)) {
1159                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1160                 return -EINVAL;
1161         }
1162         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1163                 return -EFAULT;
1164         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1165                 duprintf("get_entries: %u != %zu\n",
1166                          *len, sizeof(get) + get.size);
1167                 return -EINVAL;
1168         }
1169         get.name[sizeof(get.name) - 1] = '\0';
1170
1171         t = xt_find_table_lock(net, AF_INET, get.name);
1172         if (!IS_ERR_OR_NULL(t)) {
1173                 const struct xt_table_info *private = t->private;
1174                 duprintf("t->private->number = %u\n", private->number);
1175                 if (get.size == private->size)
1176                         ret = copy_entries_to_user(private->size,
1177                                                    t, uptr->entrytable);
1178                 else {
1179                         duprintf("get_entries: I've got %u not %u!\n",
1180                                  private->size, get.size);
1181                         ret = -EAGAIN;
1182                 }
1183                 module_put(t->me);
1184                 xt_table_unlock(t);
1185         } else
1186                 ret = t ? PTR_ERR(t) : -ENOENT;
1187
1188         return ret;
1189 }
1190
1191 static int
1192 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1193              struct xt_table_info *newinfo, unsigned int num_counters,
1194              void __user *counters_ptr)
1195 {
1196         int ret;
1197         struct xt_table *t;
1198         struct xt_table_info *oldinfo;
1199         struct xt_counters *counters;
1200         struct ipt_entry *iter;
1201
1202         ret = 0;
1203         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1204         if (!counters) {
1205                 ret = -ENOMEM;
1206                 goto out;
1207         }
1208
1209         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1210                                     "iptable_%s", name);
1211         if (IS_ERR_OR_NULL(t)) {
1212                 ret = t ? PTR_ERR(t) : -ENOENT;
1213                 goto free_newinfo_counters_untrans;
1214         }
1215
1216         /* You lied! */
1217         if (valid_hooks != t->valid_hooks) {
1218                 duprintf("Valid hook crap: %08X vs %08X\n",
1219                          valid_hooks, t->valid_hooks);
1220                 ret = -EINVAL;
1221                 goto put_module;
1222         }
1223
1224         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1225         if (!oldinfo)
1226                 goto put_module;
1227
1228         /* Update module usage count based on number of rules */
1229         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1230                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1231         if ((oldinfo->number > oldinfo->initial_entries) ||
1232             (newinfo->number <= oldinfo->initial_entries))
1233                 module_put(t->me);
1234         if ((oldinfo->number > oldinfo->initial_entries) &&
1235             (newinfo->number <= oldinfo->initial_entries))
1236                 module_put(t->me);
1237
1238         /* Get the old counters, and synchronize with replace */
1239         get_counters(oldinfo, counters);
1240
1241         /* Decrease module usage counts and free resource */
1242         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1243                 cleanup_entry(iter, net);
1244
1245         xt_free_table_info(oldinfo);
1246         if (copy_to_user(counters_ptr, counters,
1247                          sizeof(struct xt_counters) * num_counters) != 0) {
1248                 /* Silent error, can't fail, new table is already in place */
1249                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1250         }
1251         vfree(counters);
1252         xt_table_unlock(t);
1253         return ret;
1254
1255  put_module:
1256         module_put(t->me);
1257         xt_table_unlock(t);
1258  free_newinfo_counters_untrans:
1259         vfree(counters);
1260  out:
1261         return ret;
1262 }
1263
1264 static int
1265 do_replace(struct net *net, const void __user *user, unsigned int len)
1266 {
1267         int ret;
1268         struct ipt_replace tmp;
1269         struct xt_table_info *newinfo;
1270         void *loc_cpu_entry;
1271         struct ipt_entry *iter;
1272
1273         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1274                 return -EFAULT;
1275
1276         /* overflow check */
1277         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1278                 return -ENOMEM;
1279         if (tmp.num_counters == 0)
1280                 return -EINVAL;
1281
1282         tmp.name[sizeof(tmp.name)-1] = 0;
1283
1284         newinfo = xt_alloc_table_info(tmp.size);
1285         if (!newinfo)
1286                 return -ENOMEM;
1287
1288         loc_cpu_entry = newinfo->entries;
1289         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1290                            tmp.size) != 0) {
1291                 ret = -EFAULT;
1292                 goto free_newinfo;
1293         }
1294
1295         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1296         if (ret != 0)
1297                 goto free_newinfo;
1298
1299         duprintf("Translated table\n");
1300
1301         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1302                            tmp.num_counters, tmp.counters);
1303         if (ret)
1304                 goto free_newinfo_untrans;
1305         return 0;
1306
1307  free_newinfo_untrans:
1308         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1309                 cleanup_entry(iter, net);
1310  free_newinfo:
1311         xt_free_table_info(newinfo);
1312         return ret;
1313 }
1314
1315 static int
1316 do_add_counters(struct net *net, const void __user *user,
1317                 unsigned int len, int compat)
1318 {
1319         unsigned int i;
1320         struct xt_counters_info tmp;
1321         struct xt_counters *paddc;
1322         unsigned int num_counters;
1323         const char *name;
1324         int size;
1325         void *ptmp;
1326         struct xt_table *t;
1327         const struct xt_table_info *private;
1328         int ret = 0;
1329         struct ipt_entry *iter;
1330         unsigned int addend;
1331 #ifdef CONFIG_COMPAT
1332         struct compat_xt_counters_info compat_tmp;
1333
1334         if (compat) {
1335                 ptmp = &compat_tmp;
1336                 size = sizeof(struct compat_xt_counters_info);
1337         } else
1338 #endif
1339         {
1340                 ptmp = &tmp;
1341                 size = sizeof(struct xt_counters_info);
1342         }
1343
1344         if (copy_from_user(ptmp, user, size) != 0)
1345                 return -EFAULT;
1346
1347 #ifdef CONFIG_COMPAT
1348         if (compat) {
1349                 num_counters = compat_tmp.num_counters;
1350                 name = compat_tmp.name;
1351         } else
1352 #endif
1353         {
1354                 num_counters = tmp.num_counters;
1355                 name = tmp.name;
1356         }
1357
1358         if (len != size + num_counters * sizeof(struct xt_counters))
1359                 return -EINVAL;
1360
1361         paddc = vmalloc(len - size);
1362         if (!paddc)
1363                 return -ENOMEM;
1364
1365         if (copy_from_user(paddc, user + size, len - size) != 0) {
1366                 ret = -EFAULT;
1367                 goto free;
1368         }
1369
1370         t = xt_find_table_lock(net, AF_INET, name);
1371         if (IS_ERR_OR_NULL(t)) {
1372                 ret = t ? PTR_ERR(t) : -ENOENT;
1373                 goto free;
1374         }
1375
1376         local_bh_disable();
1377         private = t->private;
1378         if (private->number != num_counters) {
1379                 ret = -EINVAL;
1380                 goto unlock_up_free;
1381         }
1382
1383         i = 0;
1384         addend = xt_write_recseq_begin();
1385         xt_entry_foreach(iter, private->entries, private->size) {
1386                 struct xt_counters *tmp;
1387
1388                 tmp = xt_get_this_cpu_counter(&iter->counters);
1389                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1390                 ++i;
1391         }
1392         xt_write_recseq_end(addend);
1393  unlock_up_free:
1394         local_bh_enable();
1395         xt_table_unlock(t);
1396         module_put(t->me);
1397  free:
1398         vfree(paddc);
1399
1400         return ret;
1401 }
1402
1403 #ifdef CONFIG_COMPAT
1404 struct compat_ipt_replace {
1405         char                    name[XT_TABLE_MAXNAMELEN];
1406         u32                     valid_hooks;
1407         u32                     num_entries;
1408         u32                     size;
1409         u32                     hook_entry[NF_INET_NUMHOOKS];
1410         u32                     underflow[NF_INET_NUMHOOKS];
1411         u32                     num_counters;
1412         compat_uptr_t           counters;       /* struct xt_counters * */
1413         struct compat_ipt_entry entries[0];
1414 };
1415
1416 static int
1417 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1418                           unsigned int *size, struct xt_counters *counters,
1419                           unsigned int i)
1420 {
1421         struct xt_entry_target *t;
1422         struct compat_ipt_entry __user *ce;
1423         u_int16_t target_offset, next_offset;
1424         compat_uint_t origsize;
1425         const struct xt_entry_match *ematch;
1426         int ret = 0;
1427
1428         origsize = *size;
1429         ce = (struct compat_ipt_entry __user *)*dstptr;
1430         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1431             copy_to_user(&ce->counters, &counters[i],
1432             sizeof(counters[i])) != 0)
1433                 return -EFAULT;
1434
1435         *dstptr += sizeof(struct compat_ipt_entry);
1436         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1437
1438         xt_ematch_foreach(ematch, e) {
1439                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1440                 if (ret != 0)
1441                         return ret;
1442         }
1443         target_offset = e->target_offset - (origsize - *size);
1444         t = ipt_get_target(e);
1445         ret = xt_compat_target_to_user(t, dstptr, size);
1446         if (ret)
1447                 return ret;
1448         next_offset = e->next_offset - (origsize - *size);
1449         if (put_user(target_offset, &ce->target_offset) != 0 ||
1450             put_user(next_offset, &ce->next_offset) != 0)
1451                 return -EFAULT;
1452         return 0;
1453 }
1454
1455 static int
1456 compat_find_calc_match(struct xt_entry_match *m,
1457                        const char *name,
1458                        const struct ipt_ip *ip,
1459                        int *size)
1460 {
1461         struct xt_match *match;
1462
1463         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1464                                       m->u.user.revision);
1465         if (IS_ERR(match)) {
1466                 duprintf("compat_check_calc_match: `%s' not found\n",
1467                          m->u.user.name);
1468                 return PTR_ERR(match);
1469         }
1470         m->u.kernel.match = match;
1471         *size += xt_compat_match_offset(match);
1472         return 0;
1473 }
1474
1475 static void compat_release_entry(struct compat_ipt_entry *e)
1476 {
1477         struct xt_entry_target *t;
1478         struct xt_entry_match *ematch;
1479
1480         /* Cleanup all matches */
1481         xt_ematch_foreach(ematch, e)
1482                 module_put(ematch->u.kernel.match->me);
1483         t = compat_ipt_get_target(e);
1484         module_put(t->u.kernel.target->me);
1485 }
1486
1487 static int
1488 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1489                                   struct xt_table_info *newinfo,
1490                                   unsigned int *size,
1491                                   const unsigned char *base,
1492                                   const unsigned char *limit,
1493                                   const unsigned int *hook_entries,
1494                                   const unsigned int *underflows,
1495                                   const char *name)
1496 {
1497         struct xt_entry_match *ematch;
1498         struct xt_entry_target *t;
1499         struct xt_target *target;
1500         unsigned int entry_offset;
1501         unsigned int j;
1502         int ret, off, h;
1503
1504         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1505         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1506             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1507             (unsigned char *)e + e->next_offset > limit) {
1508                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1509                 return -EINVAL;
1510         }
1511
1512         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1513                              sizeof(struct compat_xt_entry_target)) {
1514                 duprintf("checking: element %p size %u\n",
1515                          e, e->next_offset);
1516                 return -EINVAL;
1517         }
1518
1519         /* For purposes of check_entry casting the compat entry is fine */
1520         ret = check_entry((struct ipt_entry *)e);
1521         if (ret)
1522                 return ret;
1523
1524         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1525         entry_offset = (void *)e - (void *)base;
1526         j = 0;
1527         xt_ematch_foreach(ematch, e) {
1528                 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1529                 if (ret != 0)
1530                         goto release_matches;
1531                 ++j;
1532         }
1533
1534         t = compat_ipt_get_target(e);
1535         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1536                                         t->u.user.revision);
1537         if (IS_ERR(target)) {
1538                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1539                          t->u.user.name);
1540                 ret = PTR_ERR(target);
1541                 goto release_matches;
1542         }
1543         t->u.kernel.target = target;
1544
1545         off += xt_compat_target_offset(target);
1546         *size += off;
1547         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1548         if (ret)
1549                 goto out;
1550
1551         /* Check hooks & underflows */
1552         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1553                 if ((unsigned char *)e - base == hook_entries[h])
1554                         newinfo->hook_entry[h] = hook_entries[h];
1555                 if ((unsigned char *)e - base == underflows[h])
1556                         newinfo->underflow[h] = underflows[h];
1557         }
1558
1559         /* Clear counters and comefrom */
1560         memset(&e->counters, 0, sizeof(e->counters));
1561         e->comefrom = 0;
1562         return 0;
1563
1564 out:
1565         module_put(t->u.kernel.target->me);
1566 release_matches:
1567         xt_ematch_foreach(ematch, e) {
1568                 if (j-- == 0)
1569                         break;
1570                 module_put(ematch->u.kernel.match->me);
1571         }
1572         return ret;
1573 }
1574
1575 static int
1576 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1577                             unsigned int *size, const char *name,
1578                             struct xt_table_info *newinfo, unsigned char *base)
1579 {
1580         struct xt_entry_target *t;
1581         struct xt_target *target;
1582         struct ipt_entry *de;
1583         unsigned int origsize;
1584         int ret, h;
1585         struct xt_entry_match *ematch;
1586
1587         ret = 0;
1588         origsize = *size;
1589         de = (struct ipt_entry *)*dstptr;
1590         memcpy(de, e, sizeof(struct ipt_entry));
1591         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1592
1593         *dstptr += sizeof(struct ipt_entry);
1594         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1595
1596         xt_ematch_foreach(ematch, e) {
1597                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1598                 if (ret != 0)
1599                         return ret;
1600         }
1601         de->target_offset = e->target_offset - (origsize - *size);
1602         t = compat_ipt_get_target(e);
1603         target = t->u.kernel.target;
1604         xt_compat_target_from_user(t, dstptr, size);
1605
1606         de->next_offset = e->next_offset - (origsize - *size);
1607         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1608                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1609                         newinfo->hook_entry[h] -= origsize - *size;
1610                 if ((unsigned char *)de - base < newinfo->underflow[h])
1611                         newinfo->underflow[h] -= origsize - *size;
1612         }
1613         return ret;
1614 }
1615
1616 static int
1617 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1618 {
1619         struct xt_entry_match *ematch;
1620         struct xt_mtchk_param mtpar;
1621         unsigned int j;
1622         int ret = 0;
1623
1624         e->counters.pcnt = xt_percpu_counter_alloc();
1625         if (IS_ERR_VALUE(e->counters.pcnt))
1626                 return -ENOMEM;
1627
1628         j = 0;
1629         mtpar.net       = net;
1630         mtpar.table     = name;
1631         mtpar.entryinfo = &e->ip;
1632         mtpar.hook_mask = e->comefrom;
1633         mtpar.family    = NFPROTO_IPV4;
1634         xt_ematch_foreach(ematch, e) {
1635                 ret = check_match(ematch, &mtpar);
1636                 if (ret != 0)
1637                         goto cleanup_matches;
1638                 ++j;
1639         }
1640
1641         ret = check_target(e, net, name);
1642         if (ret)
1643                 goto cleanup_matches;
1644         return 0;
1645
1646  cleanup_matches:
1647         xt_ematch_foreach(ematch, e) {
1648                 if (j-- == 0)
1649                         break;
1650                 cleanup_match(ematch, net);
1651         }
1652
1653         xt_percpu_counter_free(e->counters.pcnt);
1654
1655         return ret;
1656 }
1657
1658 static int
1659 translate_compat_table(struct net *net,
1660                        const char *name,
1661                        unsigned int valid_hooks,
1662                        struct xt_table_info **pinfo,
1663                        void **pentry0,
1664                        unsigned int total_size,
1665                        unsigned int number,
1666                        unsigned int *hook_entries,
1667                        unsigned int *underflows)
1668 {
1669         unsigned int i, j;
1670         struct xt_table_info *newinfo, *info;
1671         void *pos, *entry0, *entry1;
1672         struct compat_ipt_entry *iter0;
1673         struct ipt_entry *iter1;
1674         unsigned int size;
1675         int ret;
1676
1677         info = *pinfo;
1678         entry0 = *pentry0;
1679         size = total_size;
1680         info->number = number;
1681
1682         /* Init all hooks to impossible value. */
1683         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1684                 info->hook_entry[i] = 0xFFFFFFFF;
1685                 info->underflow[i] = 0xFFFFFFFF;
1686         }
1687
1688         duprintf("translate_compat_table: size %u\n", info->size);
1689         j = 0;
1690         xt_compat_lock(AF_INET);
1691         xt_compat_init_offsets(AF_INET, number);
1692         /* Walk through entries, checking offsets. */
1693         xt_entry_foreach(iter0, entry0, total_size) {
1694                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1695                                                         entry0,
1696                                                         entry0 + total_size,
1697                                                         hook_entries,
1698                                                         underflows,
1699                                                         name);
1700                 if (ret != 0)
1701                         goto out_unlock;
1702                 ++j;
1703         }
1704
1705         ret = -EINVAL;
1706         if (j != number) {
1707                 duprintf("translate_compat_table: %u not %u entries\n",
1708                          j, number);
1709                 goto out_unlock;
1710         }
1711
1712         /* Check hooks all assigned */
1713         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1714                 /* Only hooks which are valid */
1715                 if (!(valid_hooks & (1 << i)))
1716                         continue;
1717                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1718                         duprintf("Invalid hook entry %u %u\n",
1719                                  i, hook_entries[i]);
1720                         goto out_unlock;
1721                 }
1722                 if (info->underflow[i] == 0xFFFFFFFF) {
1723                         duprintf("Invalid underflow %u %u\n",
1724                                  i, underflows[i]);
1725                         goto out_unlock;
1726                 }
1727         }
1728
1729         ret = -ENOMEM;
1730         newinfo = xt_alloc_table_info(size);
1731         if (!newinfo)
1732                 goto out_unlock;
1733
1734         newinfo->number = number;
1735         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1736                 newinfo->hook_entry[i] = info->hook_entry[i];
1737                 newinfo->underflow[i] = info->underflow[i];
1738         }
1739         entry1 = newinfo->entries;
1740         pos = entry1;
1741         size = total_size;
1742         xt_entry_foreach(iter0, entry0, total_size) {
1743                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1744                                                   name, newinfo, entry1);
1745                 if (ret != 0)
1746                         break;
1747         }
1748         xt_compat_flush_offsets(AF_INET);
1749         xt_compat_unlock(AF_INET);
1750         if (ret)
1751                 goto free_newinfo;
1752
1753         ret = -ELOOP;
1754         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1755                 goto free_newinfo;
1756
1757         i = 0;
1758         xt_entry_foreach(iter1, entry1, newinfo->size) {
1759                 ret = compat_check_entry(iter1, net, name);
1760                 if (ret != 0)
1761                         break;
1762                 ++i;
1763                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1764                     XT_ERROR_TARGET) == 0)
1765                         ++newinfo->stacksize;
1766         }
1767         if (ret) {
1768                 /*
1769                  * The first i matches need cleanup_entry (calls ->destroy)
1770                  * because they had called ->check already. The other j-i
1771                  * entries need only release.
1772                  */
1773                 int skip = i;
1774                 j -= i;
1775                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1776                         if (skip-- > 0)
1777                                 continue;
1778                         if (j-- == 0)
1779                                 break;
1780                         compat_release_entry(iter0);
1781                 }
1782                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1783                         if (i-- == 0)
1784                                 break;
1785                         cleanup_entry(iter1, net);
1786                 }
1787                 xt_free_table_info(newinfo);
1788                 return ret;
1789         }
1790
1791         *pinfo = newinfo;
1792         *pentry0 = entry1;
1793         xt_free_table_info(info);
1794         return 0;
1795
1796 free_newinfo:
1797         xt_free_table_info(newinfo);
1798 out:
1799         xt_entry_foreach(iter0, entry0, total_size) {
1800                 if (j-- == 0)
1801                         break;
1802                 compat_release_entry(iter0);
1803         }
1804         return ret;
1805 out_unlock:
1806         xt_compat_flush_offsets(AF_INET);
1807         xt_compat_unlock(AF_INET);
1808         goto out;
1809 }
1810
1811 static int
1812 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1813 {
1814         int ret;
1815         struct compat_ipt_replace tmp;
1816         struct xt_table_info *newinfo;
1817         void *loc_cpu_entry;
1818         struct ipt_entry *iter;
1819
1820         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1821                 return -EFAULT;
1822
1823         /* overflow check */
1824         if (tmp.size >= INT_MAX / num_possible_cpus())
1825                 return -ENOMEM;
1826         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1827                 return -ENOMEM;
1828         if (tmp.num_counters == 0)
1829                 return -EINVAL;
1830
1831         tmp.name[sizeof(tmp.name)-1] = 0;
1832
1833         newinfo = xt_alloc_table_info(tmp.size);
1834         if (!newinfo)
1835                 return -ENOMEM;
1836
1837         loc_cpu_entry = newinfo->entries;
1838         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1839                            tmp.size) != 0) {
1840                 ret = -EFAULT;
1841                 goto free_newinfo;
1842         }
1843
1844         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1845                                      &newinfo, &loc_cpu_entry, tmp.size,
1846                                      tmp.num_entries, tmp.hook_entry,
1847                                      tmp.underflow);
1848         if (ret != 0)
1849                 goto free_newinfo;
1850
1851         duprintf("compat_do_replace: Translated table\n");
1852
1853         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1854                            tmp.num_counters, compat_ptr(tmp.counters));
1855         if (ret)
1856                 goto free_newinfo_untrans;
1857         return 0;
1858
1859  free_newinfo_untrans:
1860         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1861                 cleanup_entry(iter, net);
1862  free_newinfo:
1863         xt_free_table_info(newinfo);
1864         return ret;
1865 }
1866
1867 static int
1868 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1869                       unsigned int len)
1870 {
1871         int ret;
1872
1873         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1874                 return -EPERM;
1875
1876         switch (cmd) {
1877         case IPT_SO_SET_REPLACE:
1878                 ret = compat_do_replace(sock_net(sk), user, len);
1879                 break;
1880
1881         case IPT_SO_SET_ADD_COUNTERS:
1882                 ret = do_add_counters(sock_net(sk), user, len, 1);
1883                 break;
1884
1885         default:
1886                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1887                 ret = -EINVAL;
1888         }
1889
1890         return ret;
1891 }
1892
1893 struct compat_ipt_get_entries {
1894         char name[XT_TABLE_MAXNAMELEN];
1895         compat_uint_t size;
1896         struct compat_ipt_entry entrytable[0];
1897 };
1898
1899 static int
1900 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1901                             void __user *userptr)
1902 {
1903         struct xt_counters *counters;
1904         const struct xt_table_info *private = table->private;
1905         void __user *pos;
1906         unsigned int size;
1907         int ret = 0;
1908         unsigned int i = 0;
1909         struct ipt_entry *iter;
1910
1911         counters = alloc_counters(table);
1912         if (IS_ERR(counters))
1913                 return PTR_ERR(counters);
1914
1915         pos = userptr;
1916         size = total_size;
1917         xt_entry_foreach(iter, private->entries, total_size) {
1918                 ret = compat_copy_entry_to_user(iter, &pos,
1919                                                 &size, counters, i++);
1920                 if (ret != 0)
1921                         break;
1922         }
1923
1924         vfree(counters);
1925         return ret;
1926 }
1927
1928 static int
1929 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1930                    int *len)
1931 {
1932         int ret;
1933         struct compat_ipt_get_entries get;
1934         struct xt_table *t;
1935
1936         if (*len < sizeof(get)) {
1937                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1938                 return -EINVAL;
1939         }
1940
1941         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1942                 return -EFAULT;
1943
1944         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1945                 duprintf("compat_get_entries: %u != %zu\n",
1946                          *len, sizeof(get) + get.size);
1947                 return -EINVAL;
1948         }
1949         get.name[sizeof(get.name) - 1] = '\0';
1950
1951         xt_compat_lock(AF_INET);
1952         t = xt_find_table_lock(net, AF_INET, get.name);
1953         if (!IS_ERR_OR_NULL(t)) {
1954                 const struct xt_table_info *private = t->private;
1955                 struct xt_table_info info;
1956                 duprintf("t->private->number = %u\n", private->number);
1957                 ret = compat_table_info(private, &info);
1958                 if (!ret && get.size == info.size) {
1959                         ret = compat_copy_entries_to_user(private->size,
1960                                                           t, uptr->entrytable);
1961                 } else if (!ret) {
1962                         duprintf("compat_get_entries: I've got %u not %u!\n",
1963                                  private->size, get.size);
1964                         ret = -EAGAIN;
1965                 }
1966                 xt_compat_flush_offsets(AF_INET);
1967                 module_put(t->me);
1968                 xt_table_unlock(t);
1969         } else
1970                 ret = t ? PTR_ERR(t) : -ENOENT;
1971
1972         xt_compat_unlock(AF_INET);
1973         return ret;
1974 }
1975
1976 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1977
1978 static int
1979 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1980 {
1981         int ret;
1982
1983         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1984                 return -EPERM;
1985
1986         switch (cmd) {
1987         case IPT_SO_GET_INFO:
1988                 ret = get_info(sock_net(sk), user, len, 1);
1989                 break;
1990         case IPT_SO_GET_ENTRIES:
1991                 ret = compat_get_entries(sock_net(sk), user, len);
1992                 break;
1993         default:
1994                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1995         }
1996         return ret;
1997 }
1998 #endif
1999
2000 static int
2001 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2002 {
2003         int ret;
2004
2005         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2006                 return -EPERM;
2007
2008         switch (cmd) {
2009         case IPT_SO_SET_REPLACE:
2010                 ret = do_replace(sock_net(sk), user, len);
2011                 break;
2012
2013         case IPT_SO_SET_ADD_COUNTERS:
2014                 ret = do_add_counters(sock_net(sk), user, len, 0);
2015                 break;
2016
2017         default:
2018                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2019                 ret = -EINVAL;
2020         }
2021
2022         return ret;
2023 }
2024
2025 static int
2026 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2027 {
2028         int ret;
2029
2030         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2031                 return -EPERM;
2032
2033         switch (cmd) {
2034         case IPT_SO_GET_INFO:
2035                 ret = get_info(sock_net(sk), user, len, 0);
2036                 break;
2037
2038         case IPT_SO_GET_ENTRIES:
2039                 ret = get_entries(sock_net(sk), user, len);
2040                 break;
2041
2042         case IPT_SO_GET_REVISION_MATCH:
2043         case IPT_SO_GET_REVISION_TARGET: {
2044                 struct xt_get_revision rev;
2045                 int target;
2046
2047                 if (*len != sizeof(rev)) {
2048                         ret = -EINVAL;
2049                         break;
2050                 }
2051                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2052                         ret = -EFAULT;
2053                         break;
2054                 }
2055                 rev.name[sizeof(rev.name)-1] = 0;
2056
2057                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2058                         target = 1;
2059                 else
2060                         target = 0;
2061
2062                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2063                                                          rev.revision,
2064                                                          target, &ret),
2065                                         "ipt_%s", rev.name);
2066                 break;
2067         }
2068
2069         default:
2070                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2071                 ret = -EINVAL;
2072         }
2073
2074         return ret;
2075 }
2076
2077 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
2078 {
2079         struct xt_table_info *private;
2080         void *loc_cpu_entry;
2081         struct module *table_owner = table->me;
2082         struct ipt_entry *iter;
2083
2084         private = xt_unregister_table(table);
2085
2086         /* Decrease module usage counts and free resources */
2087         loc_cpu_entry = private->entries;
2088         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2089                 cleanup_entry(iter, net);
2090         if (private->number > private->initial_entries)
2091                 module_put(table_owner);
2092         xt_free_table_info(private);
2093 }
2094
2095 int ipt_register_table(struct net *net, const struct xt_table *table,
2096                        const struct ipt_replace *repl,
2097                        const struct nf_hook_ops *ops, struct xt_table **res)
2098 {
2099         int ret;
2100         struct xt_table_info *newinfo;
2101         struct xt_table_info bootstrap = {0};
2102         void *loc_cpu_entry;
2103         struct xt_table *new_table;
2104
2105         newinfo = xt_alloc_table_info(repl->size);
2106         if (!newinfo)
2107                 return -ENOMEM;
2108
2109         loc_cpu_entry = newinfo->entries;
2110         memcpy(loc_cpu_entry, repl->entries, repl->size);
2111
2112         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2113         if (ret != 0)
2114                 goto out_free;
2115
2116         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2117         if (IS_ERR(new_table)) {
2118                 ret = PTR_ERR(new_table);
2119                 goto out_free;
2120         }
2121
2122         /* set res now, will see skbs right after nf_register_net_hooks */
2123         WRITE_ONCE(*res, new_table);
2124
2125         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
2126         if (ret != 0) {
2127                 __ipt_unregister_table(net, new_table);
2128                 *res = NULL;
2129         }
2130
2131         return ret;
2132
2133 out_free:
2134         xt_free_table_info(newinfo);
2135         return ret;
2136 }
2137
2138 void ipt_unregister_table(struct net *net, struct xt_table *table,
2139                           const struct nf_hook_ops *ops)
2140 {
2141         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
2142         __ipt_unregister_table(net, table);
2143 }
2144
2145 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2146 static inline bool
2147 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2148                      u_int8_t type, u_int8_t code,
2149                      bool invert)
2150 {
2151         return ((test_type == 0xFF) ||
2152                 (type == test_type && code >= min_code && code <= max_code))
2153                 ^ invert;
2154 }
2155
2156 static bool
2157 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2158 {
2159         const struct icmphdr *ic;
2160         struct icmphdr _icmph;
2161         const struct ipt_icmp *icmpinfo = par->matchinfo;
2162
2163         /* Must not be a fragment. */
2164         if (par->fragoff != 0)
2165                 return false;
2166
2167         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2168         if (ic == NULL) {
2169                 /* We've been asked to examine this packet, and we
2170                  * can't.  Hence, no choice but to drop.
2171                  */
2172                 duprintf("Dropping evil ICMP tinygram.\n");
2173                 par->hotdrop = true;
2174                 return false;
2175         }
2176
2177         return icmp_type_code_match(icmpinfo->type,
2178                                     icmpinfo->code[0],
2179                                     icmpinfo->code[1],
2180                                     ic->type, ic->code,
2181                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2182 }
2183
2184 static int icmp_checkentry(const struct xt_mtchk_param *par)
2185 {
2186         const struct ipt_icmp *icmpinfo = par->matchinfo;
2187
2188         /* Must specify no unknown invflags */
2189         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2190 }
2191
2192 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2193         {
2194                 .name             = XT_STANDARD_TARGET,
2195                 .targetsize       = sizeof(int),
2196                 .family           = NFPROTO_IPV4,
2197 #ifdef CONFIG_COMPAT
2198                 .compatsize       = sizeof(compat_int_t),
2199                 .compat_from_user = compat_standard_from_user,
2200                 .compat_to_user   = compat_standard_to_user,
2201 #endif
2202         },
2203         {
2204                 .name             = XT_ERROR_TARGET,
2205                 .target           = ipt_error,
2206                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2207                 .family           = NFPROTO_IPV4,
2208         },
2209 };
2210
2211 static struct nf_sockopt_ops ipt_sockopts = {
2212         .pf             = PF_INET,
2213         .set_optmin     = IPT_BASE_CTL,
2214         .set_optmax     = IPT_SO_SET_MAX+1,
2215         .set            = do_ipt_set_ctl,
2216 #ifdef CONFIG_COMPAT
2217         .compat_set     = compat_do_ipt_set_ctl,
2218 #endif
2219         .get_optmin     = IPT_BASE_CTL,
2220         .get_optmax     = IPT_SO_GET_MAX+1,
2221         .get            = do_ipt_get_ctl,
2222 #ifdef CONFIG_COMPAT
2223         .compat_get     = compat_do_ipt_get_ctl,
2224 #endif
2225         .owner          = THIS_MODULE,
2226 };
2227
2228 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2229         {
2230                 .name       = "icmp",
2231                 .match      = icmp_match,
2232                 .matchsize  = sizeof(struct ipt_icmp),
2233                 .checkentry = icmp_checkentry,
2234                 .proto      = IPPROTO_ICMP,
2235                 .family     = NFPROTO_IPV4,
2236         },
2237 };
2238
2239 static int __net_init ip_tables_net_init(struct net *net)
2240 {
2241         return xt_proto_init(net, NFPROTO_IPV4);
2242 }
2243
2244 static void __net_exit ip_tables_net_exit(struct net *net)
2245 {
2246         xt_proto_fini(net, NFPROTO_IPV4);
2247 }
2248
2249 static struct pernet_operations ip_tables_net_ops = {
2250         .init = ip_tables_net_init,
2251         .exit = ip_tables_net_exit,
2252 };
2253
2254 static int __init ip_tables_init(void)
2255 {
2256         int ret;
2257
2258         ret = register_pernet_subsys(&ip_tables_net_ops);
2259         if (ret < 0)
2260                 goto err1;
2261
2262         /* No one else will be downing sem now, so we won't sleep */
2263         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2264         if (ret < 0)
2265                 goto err2;
2266         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2267         if (ret < 0)
2268                 goto err4;
2269
2270         /* Register setsockopt */
2271         ret = nf_register_sockopt(&ipt_sockopts);
2272         if (ret < 0)
2273                 goto err5;
2274
2275         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2276         return 0;
2277
2278 err5:
2279         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2280 err4:
2281         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2282 err2:
2283         unregister_pernet_subsys(&ip_tables_net_ops);
2284 err1:
2285         return ret;
2286 }
2287
2288 static void __exit ip_tables_fini(void)
2289 {
2290         nf_unregister_sockopt(&ipt_sockopts);
2291
2292         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2293         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2294         unregister_pernet_subsys(&ip_tables_net_ops);
2295 }
2296
2297 EXPORT_SYMBOL(ipt_register_table);
2298 EXPORT_SYMBOL(ipt_unregister_table);
2299 EXPORT_SYMBOL(ipt_do_table);
2300 module_init(ip_tables_init);
2301 module_exit(ip_tables_fini);