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_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 592 {
 593         const struct ipt_ip *ip = par->entryinfo;
 594         int ret;
 595
 596         par->match     = m->u.kernel.match;
 597         par->matchinfo = m->data;
 598
 599         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 600               ip->proto, ip->invflags & IPT_INV_PROTO);
 601         if (ret < 0) {
 602                 duprintf("check failed for `%s'.\n", par->match->name);
 603                 return ret;
 604         }
 605         return 0;
 606 }
 607
 608 static int
 609 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 610 {
 611         struct xt_match *match;
 612         int ret;
 613
 614         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 615                                       m->u.user.revision);
 616         if (IS_ERR(match)) {
 617                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 618                 return PTR_ERR(match);
 619         }
 620         m->u.kernel.match = match;
 621
 622         ret = check_match(m, par);
 623         if (ret)
 624                 goto err;
 625
 626         return 0;
 627 err:
 628         module_put(m->u.kernel.match->me);
 629         return ret;
 630 }
 631
 632 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 633 {
 634         struct xt_entry_target *t = ipt_get_target(e);
 635         struct xt_tgchk_param par = {
 636                 .net       = net,
 637                 .table     = name,
 638                 .entryinfo = e,
 639                 .target    = t->u.kernel.target,
 640                 .targinfo  = t->data,
 641                 .hook_mask = e->comefrom,
 642                 .family    = NFPROTO_IPV4,
 643         };
 644         int ret;
 645
 646         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 647               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 648         if (ret < 0) {
 649                 duprintf("check failed for `%s'.\n",
 650                          t->u.kernel.target->name);
 651                 return ret;
 652         }
 653         return 0;
 654 }
 655
 656 static int
 657 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 658                  unsigned int size)
 659 {
 660         struct xt_entry_target *t;
 661         struct xt_target *target;
 662         int ret;
 663         unsigned int j;
 664         struct xt_mtchk_param mtpar;
 665         struct xt_entry_match *ematch;
 666
 667         e->counters.pcnt = xt_percpu_counter_alloc();
 668         if (IS_ERR_VALUE(e->counters.pcnt))
 669                 return -ENOMEM;
 670
 671         j = 0;
 672         mtpar.net       = net;
 673         mtpar.table     = name;
 674         mtpar.entryinfo = &e->ip;
 675         mtpar.hook_mask = e->comefrom;
 676         mtpar.family    = NFPROTO_IPV4;
 677         xt_ematch_foreach(ematch, e) {
 678                 ret = find_check_match(ematch, &mtpar);
 679                 if (ret != 0)
 680                         goto cleanup_matches;
 681                 ++j;
 682         }
 683
 684         t = ipt_get_target(e);
 685         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 686                                         t->u.user.revision);
 687         if (IS_ERR(target)) {
 688                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 689                 ret = PTR_ERR(target);
 690                 goto cleanup_matches;
 691         }
 692         t->u.kernel.target = target;
 693
 694         ret = check_target(e, net, name);
 695         if (ret)
 696                 goto err;
 697
 698         return 0;
 699  err:
 700         module_put(t->u.kernel.target->me);
 701  cleanup_matches:
 702         xt_ematch_foreach(ematch, e) {
 703                 if (j-- == 0)
 704                         break;
 705                 cleanup_match(ematch, net);
 706         }
 707
 708         xt_percpu_counter_free(e->counters.pcnt);
 709
 710         return ret;
 711 }
 712
 713 static bool check_underflow(const struct ipt_entry *e)
 714 {
 715         const struct xt_entry_target *t;
 716         unsigned int verdict;
 717
 718         if (!unconditional(e))
 719                 return false;
 720         t = ipt_get_target_c(e);
 721         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 722                 return false;
 723         verdict = ((struct xt_standard_target *)t)->verdict;
 724         verdict = -verdict - 1;
 725         return verdict == NF_DROP || verdict == NF_ACCEPT;
 726 }
 727
 728 static int
 729 check_entry_size_and_hooks(struct ipt_entry *e,
 730                            struct xt_table_info *newinfo,
 731                            const unsigned char *base,
 732                            const unsigned char *limit,
 733                            const unsigned int *hook_entries,
 734                            const unsigned int *underflows,
 735                            unsigned int valid_hooks)
 736 {
 737         unsigned int h;
 738         int err;
 739
 740         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 741             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 742             (unsigned char *)e + e->next_offset > limit) {
 743                 duprintf("Bad offset %p\n", e);
 744                 return -EINVAL;
 745         }
 746
 747         if (e->next_offset
 748             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 749                 duprintf("checking: element %p size %u\n",
 750                          e, e->next_offset);
 751                 return -EINVAL;
 752         }
 753
 754         if (!ip_checkentry(&e->ip))
 755                 return -EINVAL;
 756
 757         err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 758                                      e->next_offset);
 759         if (err)
 760                 return err;
 761
 762         /* Check hooks & underflows */
 763         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 764                 if (!(valid_hooks & (1 << h)))
 765                         continue;
 766                 if ((unsigned char *)e - base == hook_entries[h])
 767                         newinfo->hook_entry[h] = hook_entries[h];
 768                 if ((unsigned char *)e - base == underflows[h]) {
 769                         if (!check_underflow(e)) {
 770                                 pr_debug("Underflows must be unconditional and "
 771                                          "use the STANDARD target with "
 772                                          "ACCEPT/DROP\n");
 773                                 return -EINVAL;
 774                         }
 775                         newinfo->underflow[h] = underflows[h];
 776                 }
 777         }
 778
 779         /* Clear counters and comefrom */
 780         e->counters = ((struct xt_counters) { 0, 0 });
 781         e->comefrom = 0;
 782         return 0;
 783 }
 784
 785 static void
 786 cleanup_entry(struct ipt_entry *e, struct net *net)
 787 {
 788         struct xt_tgdtor_param par;
 789         struct xt_entry_target *t;
 790         struct xt_entry_match *ematch;
 791
 792         /* Cleanup all matches */
 793         xt_ematch_foreach(ematch, e)
 794                 cleanup_match(ematch, net);
 795         t = ipt_get_target(e);
 796
 797         par.net      = net;
 798         par.target   = t->u.kernel.target;
 799         par.targinfo = t->data;
 800         par.family   = NFPROTO_IPV4;
 801         if (par.target->destroy != NULL)
 802                 par.target->destroy(&par);
 803         module_put(par.target->me);
 804         xt_percpu_counter_free(e->counters.pcnt);
 805 }
 806
 807 /* Checks and translates the user-supplied table segment (held in
 808    newinfo) */
 809 static int
 810 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 811                 const struct ipt_replace *repl)
 812 {
 813         struct ipt_entry *iter;
 814         unsigned int i;
 815         int ret = 0;
 816
 817         newinfo->size = repl->size;
 818         newinfo->number = repl->num_entries;
 819
 820         /* Init all hooks to impossible value. */
 821         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 822                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 823                 newinfo->underflow[i] = 0xFFFFFFFF;
 824         }
 825
 826         duprintf("translate_table: size %u\n", newinfo->size);
 827         i = 0;
 828         /* Walk through entries, checking offsets. */
 829         xt_entry_foreach(iter, entry0, newinfo->size) {
 830                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 831                                                  entry0 + repl->size,
 832                                                  repl->hook_entry,
 833                                                  repl->underflow,
 834                                                  repl->valid_hooks);
 835                 if (ret != 0)
 836                         return ret;
 837                 ++i;
 838                 if (strcmp(ipt_get_target(iter)->u.user.name,
 839                     XT_ERROR_TARGET) == 0)
 840                         ++newinfo->stacksize;
 841         }
 842
 843         if (i != repl->num_entries) {
 844                 duprintf("translate_table: %u not %u entries\n",
 845                          i, repl->num_entries);
 846                 return -EINVAL;
 847         }
 848
 849         /* Check hooks all assigned */
 850         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 851                 /* Only hooks which are valid */
 852                 if (!(repl->valid_hooks & (1 << i)))
 853                         continue;
 854                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 855                         duprintf("Invalid hook entry %u %u\n",
 856                                  i, repl->hook_entry[i]);
 857                         return -EINVAL;
 858                 }
 859                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 860                         duprintf("Invalid underflow %u %u\n",
 861                                  i, repl->underflow[i]);
 862                         return -EINVAL;
 863                 }
 864         }
 865
 866         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 867                 return -ELOOP;
 868
 869         /* Finally, each sanity check must pass */
 870         i = 0;
 871         xt_entry_foreach(iter, entry0, newinfo->size) {
 872                 ret = find_check_entry(iter, net, repl->name, repl->size);
 873                 if (ret != 0)
 874                         break;
 875                 ++i;
 876         }
 877
 878         if (ret != 0) {
 879                 xt_entry_foreach(iter, entry0, newinfo->size) {
 880                         if (i-- == 0)
 881                                 break;
 882                         cleanup_entry(iter, net);
 883                 }
 884                 return ret;
 885         }
 886
 887         return ret;
 888 }
 889
 890 static void
 891 get_counters(const struct xt_table_info *t,
 892              struct xt_counters counters[])
 893 {
 894         struct ipt_entry *iter;
 895         unsigned int cpu;
 896         unsigned int i;
 897
 898         for_each_possible_cpu(cpu) {
 899                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 900
 901                 i = 0;
 902                 xt_entry_foreach(iter, t->entries, t->size) {
 903                         struct xt_counters *tmp;
 904                         u64 bcnt, pcnt;
 905                         unsigned int start;
 906
 907                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 908                         do {
 909                                 start = read_seqcount_begin(s);
 910                                 bcnt = tmp->bcnt;
 911                                 pcnt = tmp->pcnt;
 912                         } while (read_seqcount_retry(s, start));
 913
 914                         ADD_COUNTER(counters[i], bcnt, pcnt);
 915                         ++i; /* macro does multi eval of i */
 916                 }
 917         }
 918 }
 919
 920 static struct xt_counters *alloc_counters(const struct xt_table *table)
 921 {
 922         unsigned int countersize;
 923         struct xt_counters *counters;
 924         const struct xt_table_info *private = table->private;
 925
 926         /* We need atomic snapshot of counters: rest doesn't change
 927            (other than comefrom, which userspace doesn't care
 928            about). */
 929         countersize = sizeof(struct xt_counters) * private->number;
 930         counters = vzalloc(countersize);
 931
 932         if (counters == NULL)
 933                 return ERR_PTR(-ENOMEM);
 934
 935         get_counters(private, counters);
 936
 937         return counters;
 938 }
 939
 940 static int
 941 copy_entries_to_user(unsigned int total_size,
 942                      const struct xt_table *table,
 943                      void __user *userptr)
 944 {
 945         unsigned int off, num;
 946         const struct ipt_entry *e;
 947         struct xt_counters *counters;
 948         const struct xt_table_info *private = table->private;
 949         int ret = 0;
 950         const void *loc_cpu_entry;
 951
 952         counters = alloc_counters(table);
 953         if (IS_ERR(counters))
 954                 return PTR_ERR(counters);
 955
 956         loc_cpu_entry = private->entries;
 957         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 958                 ret = -EFAULT;
 959                 goto free_counters;
 960         }
 961
 962         /* FIXME: use iterator macros --RR */
 963         /* ... then go back and fix counters and names */
 964         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 965                 unsigned int i;
 966                 const struct xt_entry_match *m;
 967                 const struct xt_entry_target *t;
 968
 969                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 970                 if (copy_to_user(userptr + off
 971                                  + offsetof(struct ipt_entry, counters),
 972                                  &counters[num],
 973                                  sizeof(counters[num])) != 0) {
 974                         ret = -EFAULT;
 975                         goto free_counters;
 976                 }
 977
 978                 for (i = sizeof(struct ipt_entry);
 979                      i < e->target_offset;
 980                      i += m->u.match_size) {
 981                         m = (void *)e + i;
 982
 983                         if (copy_to_user(userptr + off + i
 984                                          + offsetof(struct xt_entry_match,
 985                                                     u.user.name),
 986                                          m->u.kernel.match->name,
 987                                          strlen(m->u.kernel.match->name)+1)
 988                             != 0) {
 989                                 ret = -EFAULT;
 990                                 goto free_counters;
 991                         }
 992                 }
 993
 994                 t = ipt_get_target_c(e);
 995                 if (copy_to_user(userptr + off + e->target_offset
 996                                  + offsetof(struct xt_entry_target,
 997                                             u.user.name),
 998                                  t->u.kernel.target->name,
 999                                  strlen(t->u.kernel.target->name)+1) != 0) {
1000                         ret = -EFAULT;
1001                         goto free_counters;
1002                 }
1003         }
1004
1005  free_counters:
1006         vfree(counters);
1007         return ret;
1008 }
1009
1010 #ifdef CONFIG_COMPAT
1011 static void compat_standard_from_user(void *dst, const void *src)
1012 {
1013         int v = *(compat_int_t *)src;
1014
1015         if (v > 0)
1016                 v += xt_compat_calc_jump(AF_INET, v);
1017         memcpy(dst, &v, sizeof(v));
1018 }
1019
1020 static int compat_standard_to_user(void __user *dst, const void *src)
1021 {
1022         compat_int_t cv = *(int *)src;
1023
1024         if (cv > 0)
1025                 cv -= xt_compat_calc_jump(AF_INET, cv);
1026         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1027 }
1028
1029 static int compat_calc_entry(const struct ipt_entry *e,
1030                              const struct xt_table_info *info,
1031                              const void *base, struct xt_table_info *newinfo)
1032 {
1033         const struct xt_entry_match *ematch;
1034         const struct xt_entry_target *t;
1035         unsigned int entry_offset;
1036         int off, i, ret;
1037
1038         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1039         entry_offset = (void *)e - base;
1040         xt_ematch_foreach(ematch, e)
1041                 off += xt_compat_match_offset(ematch->u.kernel.match);
1042         t = ipt_get_target_c(e);
1043         off += xt_compat_target_offset(t->u.kernel.target);
1044         newinfo->size -= off;
1045         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1046         if (ret)
1047                 return ret;
1048
1049         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1050                 if (info->hook_entry[i] &&
1051                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1052                         newinfo->hook_entry[i] -= off;
1053                 if (info->underflow[i] &&
1054                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1055                         newinfo->underflow[i] -= off;
1056         }
1057         return 0;
1058 }
1059
1060 static int compat_table_info(const struct xt_table_info *info,
1061                              struct xt_table_info *newinfo)
1062 {
1063         struct ipt_entry *iter;
1064         const void *loc_cpu_entry;
1065         int ret;
1066
1067         if (!newinfo || !info)
1068                 return -EINVAL;
1069
1070         /* we dont care about newinfo->entries */
1071         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1072         newinfo->initial_entries = 0;
1073         loc_cpu_entry = info->entries;
1074         xt_compat_init_offsets(AF_INET, info->number);
1075         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1076                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1077                 if (ret != 0)
1078                         return ret;
1079         }
1080         return 0;
1081 }
1082 #endif
1083
1084 static int get_info(struct net *net, void __user *user,
1085                     const int *len, int compat)
1086 {
1087         char name[XT_TABLE_MAXNAMELEN];
1088         struct xt_table *t;
1089         int ret;
1090
1091         if (*len != sizeof(struct ipt_getinfo)) {
1092                 duprintf("length %u != %zu\n", *len,
1093                          sizeof(struct ipt_getinfo));
1094                 return -EINVAL;
1095         }
1096
1097         if (copy_from_user(name, user, sizeof(name)) != 0)
1098                 return -EFAULT;
1099
1100         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1101 #ifdef CONFIG_COMPAT
1102         if (compat)
1103                 xt_compat_lock(AF_INET);
1104 #endif
1105         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1106                                     "iptable_%s", name);
1107         if (!IS_ERR_OR_NULL(t)) {
1108                 struct ipt_getinfo info;
1109                 const struct xt_table_info *private = t->private;
1110 #ifdef CONFIG_COMPAT
1111                 struct xt_table_info tmp;
1112
1113                 if (compat) {
1114                         ret = compat_table_info(private, &tmp);
1115                         xt_compat_flush_offsets(AF_INET);
1116                         private = &tmp;
1117                 }
1118 #endif
1119                 memset(&info, 0, sizeof(info));
1120                 info.valid_hooks = t->valid_hooks;
1121                 memcpy(info.hook_entry, private->hook_entry,
1122                        sizeof(info.hook_entry));
1123                 memcpy(info.underflow, private->underflow,
1124                        sizeof(info.underflow));
1125                 info.num_entries = private->number;
1126                 info.size = private->size;
1127                 strcpy(info.name, name);
1128
1129                 if (copy_to_user(user, &info, *len) != 0)
1130                         ret = -EFAULT;
1131                 else
1132                         ret = 0;
1133
1134                 xt_table_unlock(t);
1135                 module_put(t->me);
1136         } else
1137                 ret = t ? PTR_ERR(t) : -ENOENT;
1138 #ifdef CONFIG_COMPAT
1139         if (compat)
1140                 xt_compat_unlock(AF_INET);
1141 #endif
1142         return ret;
1143 }
1144
1145 static int
1146 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1147             const int *len)
1148 {
1149         int ret;
1150         struct ipt_get_entries get;
1151         struct xt_table *t;
1152
1153         if (*len < sizeof(get)) {
1154                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1155                 return -EINVAL;
1156         }
1157         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1158                 return -EFAULT;
1159         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1160                 duprintf("get_entries: %u != %zu\n",
1161                          *len, sizeof(get) + get.size);
1162                 return -EINVAL;
1163         }
1164
1165         t = xt_find_table_lock(net, AF_INET, get.name);
1166         if (!IS_ERR_OR_NULL(t)) {
1167                 const struct xt_table_info *private = t->private;
1168                 duprintf("t->private->number = %u\n", private->number);
1169                 if (get.size == private->size)
1170                         ret = copy_entries_to_user(private->size,
1171                                                    t, uptr->entrytable);
1172                 else {
1173                         duprintf("get_entries: I've got %u not %u!\n",
1174                                  private->size, get.size);
1175                         ret = -EAGAIN;
1176                 }
1177                 module_put(t->me);
1178                 xt_table_unlock(t);
1179         } else
1180                 ret = t ? PTR_ERR(t) : -ENOENT;
1181
1182         return ret;
1183 }
1184
1185 static int
1186 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1187              struct xt_table_info *newinfo, unsigned int num_counters,
1188              void __user *counters_ptr)
1189 {
1190         int ret;
1191         struct xt_table *t;
1192         struct xt_table_info *oldinfo;
1193         struct xt_counters *counters;
1194         struct ipt_entry *iter;
1195
1196         ret = 0;
1197         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1198         if (!counters) {
1199                 ret = -ENOMEM;
1200                 goto out;
1201         }
1202
1203         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1204                                     "iptable_%s", name);
1205         if (IS_ERR_OR_NULL(t)) {
1206                 ret = t ? PTR_ERR(t) : -ENOENT;
1207                 goto free_newinfo_counters_untrans;
1208         }
1209
1210         /* You lied! */
1211         if (valid_hooks != t->valid_hooks) {
1212                 duprintf("Valid hook crap: %08X vs %08X\n",
1213                          valid_hooks, t->valid_hooks);
1214                 ret = -EINVAL;
1215                 goto put_module;
1216         }
1217
1218         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1219         if (!oldinfo)
1220                 goto put_module;
1221
1222         /* Update module usage count based on number of rules */
1223         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1224                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1225         if ((oldinfo->number > oldinfo->initial_entries) ||
1226             (newinfo->number <= oldinfo->initial_entries))
1227                 module_put(t->me);
1228         if ((oldinfo->number > oldinfo->initial_entries) &&
1229             (newinfo->number <= oldinfo->initial_entries))
1230                 module_put(t->me);
1231
1232         /* Get the old counters, and synchronize with replace */
1233         get_counters(oldinfo, counters);
1234
1235         /* Decrease module usage counts and free resource */
1236         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1237                 cleanup_entry(iter, net);
1238
1239         xt_free_table_info(oldinfo);
1240         if (copy_to_user(counters_ptr, counters,
1241                          sizeof(struct xt_counters) * num_counters) != 0) {
1242                 /* Silent error, can't fail, new table is already in place */
1243                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1244         }
1245         vfree(counters);
1246         xt_table_unlock(t);
1247         return ret;
1248
1249  put_module:
1250         module_put(t->me);
1251         xt_table_unlock(t);
1252  free_newinfo_counters_untrans:
1253         vfree(counters);
1254  out:
1255         return ret;
1256 }
1257
1258 static int
1259 do_replace(struct net *net, const void __user *user, unsigned int len)
1260 {
1261         int ret;
1262         struct ipt_replace tmp;
1263         struct xt_table_info *newinfo;
1264         void *loc_cpu_entry;
1265         struct ipt_entry *iter;
1266
1267         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1268                 return -EFAULT;
1269
1270         /* overflow check */
1271         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1272                 return -ENOMEM;
1273         if (tmp.num_counters == 0)
1274                 return -EINVAL;
1275
1276         tmp.name[sizeof(tmp.name)-1] = 0;
1277
1278         newinfo = xt_alloc_table_info(tmp.size);
1279         if (!newinfo)
1280                 return -ENOMEM;
1281
1282         loc_cpu_entry = newinfo->entries;
1283         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1284                            tmp.size) != 0) {
1285                 ret = -EFAULT;
1286                 goto free_newinfo;
1287         }
1288
1289         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1290         if (ret != 0)
1291                 goto free_newinfo;
1292
1293         duprintf("Translated table\n");
1294
1295         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1296                            tmp.num_counters, tmp.counters);
1297         if (ret)
1298                 goto free_newinfo_untrans;
1299         return 0;
1300
1301  free_newinfo_untrans:
1302         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1303                 cleanup_entry(iter, net);
1304  free_newinfo:
1305         xt_free_table_info(newinfo);
1306         return ret;
1307 }
1308
1309 static int
1310 do_add_counters(struct net *net, const void __user *user,
1311                 unsigned int len, int compat)
1312 {
1313         unsigned int i;
1314         struct xt_counters_info tmp;
1315         struct xt_counters *paddc;
1316         unsigned int num_counters;
1317         const char *name;
1318         int size;
1319         void *ptmp;
1320         struct xt_table *t;
1321         const struct xt_table_info *private;
1322         int ret = 0;
1323         struct ipt_entry *iter;
1324         unsigned int addend;
1325 #ifdef CONFIG_COMPAT
1326         struct compat_xt_counters_info compat_tmp;
1327
1328         if (compat) {
1329                 ptmp = &compat_tmp;
1330                 size = sizeof(struct compat_xt_counters_info);
1331         } else
1332 #endif
1333         {
1334                 ptmp = &tmp;
1335                 size = sizeof(struct xt_counters_info);
1336         }
1337
1338         if (copy_from_user(ptmp, user, size) != 0)
1339                 return -EFAULT;
1340
1341 #ifdef CONFIG_COMPAT
1342         if (compat) {
1343                 num_counters = compat_tmp.num_counters;
1344                 name = compat_tmp.name;
1345         } else
1346 #endif
1347         {
1348                 num_counters = tmp.num_counters;
1349                 name = tmp.name;
1350         }
1351
1352         if (len != size + num_counters * sizeof(struct xt_counters))
1353                 return -EINVAL;
1354
1355         paddc = vmalloc(len - size);
1356         if (!paddc)
1357                 return -ENOMEM;
1358
1359         if (copy_from_user(paddc, user + size, len - size) != 0) {
1360                 ret = -EFAULT;
1361                 goto free;
1362         }
1363
1364         t = xt_find_table_lock(net, AF_INET, name);
1365         if (IS_ERR_OR_NULL(t)) {
1366                 ret = t ? PTR_ERR(t) : -ENOENT;
1367                 goto free;
1368         }
1369
1370         local_bh_disable();
1371         private = t->private;
1372         if (private->number != num_counters) {
1373                 ret = -EINVAL;
1374                 goto unlock_up_free;
1375         }
1376
1377         i = 0;
1378         addend = xt_write_recseq_begin();
1379         xt_entry_foreach(iter, private->entries, private->size) {
1380                 struct xt_counters *tmp;
1381
1382                 tmp = xt_get_this_cpu_counter(&iter->counters);
1383                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1384                 ++i;
1385         }
1386         xt_write_recseq_end(addend);
1387  unlock_up_free:
1388         local_bh_enable();
1389         xt_table_unlock(t);
1390         module_put(t->me);
1391  free:
1392         vfree(paddc);
1393
1394         return ret;
1395 }
1396
1397 #ifdef CONFIG_COMPAT
1398 struct compat_ipt_replace {
1399         char                    name[XT_TABLE_MAXNAMELEN];
1400         u32                     valid_hooks;
1401         u32                     num_entries;
1402         u32                     size;
1403         u32                     hook_entry[NF_INET_NUMHOOKS];
1404         u32                     underflow[NF_INET_NUMHOOKS];
1405         u32                     num_counters;
1406         compat_uptr_t           counters;       /* struct xt_counters * */
1407         struct compat_ipt_entry entries[0];
1408 };
1409
1410 static int
1411 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1412                           unsigned int *size, struct xt_counters *counters,
1413                           unsigned int i)
1414 {
1415         struct xt_entry_target *t;
1416         struct compat_ipt_entry __user *ce;
1417         u_int16_t target_offset, next_offset;
1418         compat_uint_t origsize;
1419         const struct xt_entry_match *ematch;
1420         int ret = 0;
1421
1422         origsize = *size;
1423         ce = (struct compat_ipt_entry __user *)*dstptr;
1424         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1425             copy_to_user(&ce->counters, &counters[i],
1426             sizeof(counters[i])) != 0)
1427                 return -EFAULT;
1428
1429         *dstptr += sizeof(struct compat_ipt_entry);
1430         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1431
1432         xt_ematch_foreach(ematch, e) {
1433                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1434                 if (ret != 0)
1435                         return ret;
1436         }
1437         target_offset = e->target_offset - (origsize - *size);
1438         t = ipt_get_target(e);
1439         ret = xt_compat_target_to_user(t, dstptr, size);
1440         if (ret)
1441                 return ret;
1442         next_offset = e->next_offset - (origsize - *size);
1443         if (put_user(target_offset, &ce->target_offset) != 0 ||
1444             put_user(next_offset, &ce->next_offset) != 0)
1445                 return -EFAULT;
1446         return 0;
1447 }
1448
1449 static int
1450 compat_find_calc_match(struct xt_entry_match *m,
1451                        const char *name,
1452                        const struct ipt_ip *ip,
1453                        int *size)
1454 {
1455         struct xt_match *match;
1456
1457         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1458                                       m->u.user.revision);
1459         if (IS_ERR(match)) {
1460                 duprintf("compat_check_calc_match: `%s' not found\n",
1461                          m->u.user.name);
1462                 return PTR_ERR(match);
1463         }
1464         m->u.kernel.match = match;
1465         *size += xt_compat_match_offset(match);
1466         return 0;
1467 }
1468
1469 static void compat_release_entry(struct compat_ipt_entry *e)
1470 {
1471         struct xt_entry_target *t;
1472         struct xt_entry_match *ematch;
1473
1474         /* Cleanup all matches */
1475         xt_ematch_foreach(ematch, e)
1476                 module_put(ematch->u.kernel.match->me);
1477         t = compat_ipt_get_target(e);
1478         module_put(t->u.kernel.target->me);
1479 }
1480
1481 static int
1482 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1483                                   struct xt_table_info *newinfo,
1484                                   unsigned int *size,
1485                                   const unsigned char *base,
1486                                   const unsigned char *limit,
1487                                   const unsigned int *hook_entries,
1488                                   const unsigned int *underflows,
1489                                   const char *name)
1490 {
1491         struct xt_entry_match *ematch;
1492         struct xt_entry_target *t;
1493         struct xt_target *target;
1494         unsigned int entry_offset;
1495         unsigned int j;
1496         int ret, off, h;
1497
1498         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1499         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1500             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1501             (unsigned char *)e + e->next_offset > limit) {
1502                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1503                 return -EINVAL;
1504         }
1505
1506         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1507                              sizeof(struct compat_xt_entry_target)) {
1508                 duprintf("checking: element %p size %u\n",
1509                          e, e->next_offset);
1510                 return -EINVAL;
1511         }
1512
1513         if (!ip_checkentry(&e->ip))
1514                 return -EINVAL;
1515
1516         ret = xt_compat_check_entry_offsets(e, e->elems,
1517                                             e->target_offset, e->next_offset);
1518         if (ret)
1519                 return ret;
1520
1521         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1522         entry_offset = (void *)e - (void *)base;
1523         j = 0;
1524         xt_ematch_foreach(ematch, e) {
1525                 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1526                 if (ret != 0)
1527                         goto release_matches;
1528                 ++j;
1529         }
1530
1531         t = compat_ipt_get_target(e);
1532         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1533                                         t->u.user.revision);
1534         if (IS_ERR(target)) {
1535                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1536                          t->u.user.name);
1537                 ret = PTR_ERR(target);
1538                 goto release_matches;
1539         }
1540         t->u.kernel.target = target;
1541
1542         off += xt_compat_target_offset(target);
1543         *size += off;
1544         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1545         if (ret)
1546                 goto out;
1547
1548         /* Check hooks & underflows */
1549         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1550                 if ((unsigned char *)e - base == hook_entries[h])
1551                         newinfo->hook_entry[h] = hook_entries[h];
1552                 if ((unsigned char *)e - base == underflows[h])
1553                         newinfo->underflow[h] = underflows[h];
1554         }
1555
1556         /* Clear counters and comefrom */
1557         memset(&e->counters, 0, sizeof(e->counters));
1558         e->comefrom = 0;
1559         return 0;
1560
1561 out:
1562         module_put(t->u.kernel.target->me);
1563 release_matches:
1564         xt_ematch_foreach(ematch, e) {
1565                 if (j-- == 0)
1566                         break;
1567                 module_put(ematch->u.kernel.match->me);
1568         }
1569         return ret;
1570 }
1571
1572 static int
1573 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1574                             unsigned int *size, const char *name,
1575                             struct xt_table_info *newinfo, unsigned char *base)
1576 {
1577         struct xt_entry_target *t;
1578         struct xt_target *target;
1579         struct ipt_entry *de;
1580         unsigned int origsize;
1581         int ret, h;
1582         struct xt_entry_match *ematch;
1583
1584         ret = 0;
1585         origsize = *size;
1586         de = (struct ipt_entry *)*dstptr;
1587         memcpy(de, e, sizeof(struct ipt_entry));
1588         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1589
1590         *dstptr += sizeof(struct ipt_entry);
1591         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1592
1593         xt_ematch_foreach(ematch, e) {
1594                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1595                 if (ret != 0)
1596                         return ret;
1597         }
1598         de->target_offset = e->target_offset - (origsize - *size);
1599         t = compat_ipt_get_target(e);
1600         target = t->u.kernel.target;
1601         xt_compat_target_from_user(t, dstptr, size);
1602
1603         de->next_offset = e->next_offset - (origsize - *size);
1604         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1605                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1606                         newinfo->hook_entry[h] -= origsize - *size;
1607                 if ((unsigned char *)de - base < newinfo->underflow[h])
1608                         newinfo->underflow[h] -= origsize - *size;
1609         }
1610         return ret;
1611 }
1612
1613 static int
1614 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1615 {
1616         struct xt_entry_match *ematch;
1617         struct xt_mtchk_param mtpar;
1618         unsigned int j;
1619         int ret = 0;
1620
1621         e->counters.pcnt = xt_percpu_counter_alloc();
1622         if (IS_ERR_VALUE(e->counters.pcnt))
1623                 return -ENOMEM;
1624
1625         j = 0;
1626         mtpar.net       = net;
1627         mtpar.table     = name;
1628         mtpar.entryinfo = &e->ip;
1629         mtpar.hook_mask = e->comefrom;
1630         mtpar.family    = NFPROTO_IPV4;
1631         xt_ematch_foreach(ematch, e) {
1632                 ret = check_match(ematch, &mtpar);
1633                 if (ret != 0)
1634                         goto cleanup_matches;
1635                 ++j;
1636         }
1637
1638         ret = check_target(e, net, name);
1639         if (ret)
1640                 goto cleanup_matches;
1641         return 0;
1642
1643  cleanup_matches:
1644         xt_ematch_foreach(ematch, e) {
1645                 if (j-- == 0)
1646                         break;
1647                 cleanup_match(ematch, net);
1648         }
1649
1650         xt_percpu_counter_free(e->counters.pcnt);
1651
1652         return ret;
1653 }
1654
1655 static int
1656 translate_compat_table(struct net *net,
1657                        const char *name,
1658                        unsigned int valid_hooks,
1659                        struct xt_table_info **pinfo,
1660                        void **pentry0,
1661                        unsigned int total_size,
1662                        unsigned int number,
1663                        unsigned int *hook_entries,
1664                        unsigned int *underflows)
1665 {
1666         unsigned int i, j;
1667         struct xt_table_info *newinfo, *info;
1668         void *pos, *entry0, *entry1;
1669         struct compat_ipt_entry *iter0;
1670         struct ipt_entry *iter1;
1671         unsigned int size;
1672         int ret;
1673
1674         info = *pinfo;
1675         entry0 = *pentry0;
1676         size = total_size;
1677         info->number = number;
1678
1679         /* Init all hooks to impossible value. */
1680         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1681                 info->hook_entry[i] = 0xFFFFFFFF;
1682                 info->underflow[i] = 0xFFFFFFFF;
1683         }
1684
1685         duprintf("translate_compat_table: size %u\n", info->size);
1686         j = 0;
1687         xt_compat_lock(AF_INET);
1688         xt_compat_init_offsets(AF_INET, number);
1689         /* Walk through entries, checking offsets. */
1690         xt_entry_foreach(iter0, entry0, total_size) {
1691                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1692                                                         entry0,
1693                                                         entry0 + total_size,
1694                                                         hook_entries,
1695                                                         underflows,
1696                                                         name);
1697                 if (ret != 0)
1698                         goto out_unlock;
1699                 ++j;
1700         }
1701
1702         ret = -EINVAL;
1703         if (j != number) {
1704                 duprintf("translate_compat_table: %u not %u entries\n",
1705                          j, number);
1706                 goto out_unlock;
1707         }
1708
1709         /* Check hooks all assigned */
1710         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1711                 /* Only hooks which are valid */
1712                 if (!(valid_hooks & (1 << i)))
1713                         continue;
1714                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1715                         duprintf("Invalid hook entry %u %u\n",
1716                                  i, hook_entries[i]);
1717                         goto out_unlock;
1718                 }
1719                 if (info->underflow[i] == 0xFFFFFFFF) {
1720                         duprintf("Invalid underflow %u %u\n",
1721                                  i, underflows[i]);
1722                         goto out_unlock;
1723                 }
1724         }
1725
1726         ret = -ENOMEM;
1727         newinfo = xt_alloc_table_info(size);
1728         if (!newinfo)
1729                 goto out_unlock;
1730
1731         newinfo->number = number;
1732         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1733                 newinfo->hook_entry[i] = info->hook_entry[i];
1734                 newinfo->underflow[i] = info->underflow[i];
1735         }
1736         entry1 = newinfo->entries;
1737         pos = entry1;
1738         size = total_size;
1739         xt_entry_foreach(iter0, entry0, total_size) {
1740                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1741                                                   name, newinfo, entry1);
1742                 if (ret != 0)
1743                         break;
1744         }
1745         xt_compat_flush_offsets(AF_INET);
1746         xt_compat_unlock(AF_INET);
1747         if (ret)
1748                 goto free_newinfo;
1749
1750         ret = -ELOOP;
1751         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1752                 goto free_newinfo;
1753
1754         i = 0;
1755         xt_entry_foreach(iter1, entry1, newinfo->size) {
1756                 ret = compat_check_entry(iter1, net, name);
1757                 if (ret != 0)
1758                         break;
1759                 ++i;
1760                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1761                     XT_ERROR_TARGET) == 0)
1762                         ++newinfo->stacksize;
1763         }
1764         if (ret) {
1765                 /*
1766                  * The first i matches need cleanup_entry (calls ->destroy)
1767                  * because they had called ->check already. The other j-i
1768                  * entries need only release.
1769                  */
1770                 int skip = i;
1771                 j -= i;
1772                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1773                         if (skip-- > 0)
1774                                 continue;
1775                         if (j-- == 0)
1776                                 break;
1777                         compat_release_entry(iter0);
1778                 }
1779                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1780                         if (i-- == 0)
1781                                 break;
1782                         cleanup_entry(iter1, net);
1783                 }
1784                 xt_free_table_info(newinfo);
1785                 return ret;
1786         }
1787
1788         *pinfo = newinfo;
1789         *pentry0 = entry1;
1790         xt_free_table_info(info);
1791         return 0;
1792
1793 free_newinfo:
1794         xt_free_table_info(newinfo);
1795 out:
1796         xt_entry_foreach(iter0, entry0, total_size) {
1797                 if (j-- == 0)
1798                         break;
1799                 compat_release_entry(iter0);
1800         }
1801         return ret;
1802 out_unlock:
1803         xt_compat_flush_offsets(AF_INET);
1804         xt_compat_unlock(AF_INET);
1805         goto out;
1806 }
1807
1808 static int
1809 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1810 {
1811         int ret;
1812         struct compat_ipt_replace tmp;
1813         struct xt_table_info *newinfo;
1814         void *loc_cpu_entry;
1815         struct ipt_entry *iter;
1816
1817         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1818                 return -EFAULT;
1819
1820         /* overflow check */
1821         if (tmp.size >= INT_MAX / num_possible_cpus())
1822                 return -ENOMEM;
1823         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1824                 return -ENOMEM;
1825         if (tmp.num_counters == 0)
1826                 return -EINVAL;
1827
1828         tmp.name[sizeof(tmp.name)-1] = 0;
1829
1830         newinfo = xt_alloc_table_info(tmp.size);
1831         if (!newinfo)
1832                 return -ENOMEM;
1833
1834         loc_cpu_entry = newinfo->entries;
1835         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1836                            tmp.size) != 0) {
1837                 ret = -EFAULT;
1838                 goto free_newinfo;
1839         }
1840
1841         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1842                                      &newinfo, &loc_cpu_entry, tmp.size,
1843                                      tmp.num_entries, tmp.hook_entry,
1844                                      tmp.underflow);
1845         if (ret != 0)
1846                 goto free_newinfo;
1847
1848         duprintf("compat_do_replace: Translated table\n");
1849
1850         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1851                            tmp.num_counters, compat_ptr(tmp.counters));
1852         if (ret)
1853                 goto free_newinfo_untrans;
1854         return 0;
1855
1856  free_newinfo_untrans:
1857         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1858                 cleanup_entry(iter, net);
1859  free_newinfo:
1860         xt_free_table_info(newinfo);
1861         return ret;
1862 }
1863
1864 static int
1865 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1866                       unsigned int len)
1867 {
1868         int ret;
1869
1870         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1871                 return -EPERM;
1872
1873         switch (cmd) {
1874         case IPT_SO_SET_REPLACE:
1875                 ret = compat_do_replace(sock_net(sk), user, len);
1876                 break;
1877
1878         case IPT_SO_SET_ADD_COUNTERS:
1879                 ret = do_add_counters(sock_net(sk), user, len, 1);
1880                 break;
1881
1882         default:
1883                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1884                 ret = -EINVAL;
1885         }
1886
1887         return ret;
1888 }
1889
1890 struct compat_ipt_get_entries {
1891         char name[XT_TABLE_MAXNAMELEN];
1892         compat_uint_t size;
1893         struct compat_ipt_entry entrytable[0];
1894 };
1895
1896 static int
1897 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1898                             void __user *userptr)
1899 {
1900         struct xt_counters *counters;
1901         const struct xt_table_info *private = table->private;
1902         void __user *pos;
1903         unsigned int size;
1904         int ret = 0;
1905         unsigned int i = 0;
1906         struct ipt_entry *iter;
1907
1908         counters = alloc_counters(table);
1909         if (IS_ERR(counters))
1910                 return PTR_ERR(counters);
1911
1912         pos = userptr;
1913         size = total_size;
1914         xt_entry_foreach(iter, private->entries, total_size) {
1915                 ret = compat_copy_entry_to_user(iter, &pos,
1916                                                 &size, counters, i++);
1917                 if (ret != 0)
1918                         break;
1919         }
1920
1921         vfree(counters);
1922         return ret;
1923 }
1924
1925 static int
1926 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1927                    int *len)
1928 {
1929         int ret;
1930         struct compat_ipt_get_entries get;
1931         struct xt_table *t;
1932
1933         if (*len < sizeof(get)) {
1934                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1935                 return -EINVAL;
1936         }
1937
1938         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1939                 return -EFAULT;
1940
1941         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1942                 duprintf("compat_get_entries: %u != %zu\n",
1943                          *len, sizeof(get) + get.size);
1944                 return -EINVAL;
1945         }
1946
1947         xt_compat_lock(AF_INET);
1948         t = xt_find_table_lock(net, AF_INET, get.name);
1949         if (!IS_ERR_OR_NULL(t)) {
1950                 const struct xt_table_info *private = t->private;
1951                 struct xt_table_info info;
1952                 duprintf("t->private->number = %u\n", private->number);
1953                 ret = compat_table_info(private, &info);
1954                 if (!ret && get.size == info.size) {
1955                         ret = compat_copy_entries_to_user(private->size,
1956                                                           t, uptr->entrytable);
1957                 } else if (!ret) {
1958                         duprintf("compat_get_entries: I've got %u not %u!\n",
1959                                  private->size, get.size);
1960                         ret = -EAGAIN;
1961                 }
1962                 xt_compat_flush_offsets(AF_INET);
1963                 module_put(t->me);
1964                 xt_table_unlock(t);
1965         } else
1966                 ret = t ? PTR_ERR(t) : -ENOENT;
1967
1968         xt_compat_unlock(AF_INET);
1969         return ret;
1970 }
1971
1972 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1973
1974 static int
1975 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1976 {
1977         int ret;
1978
1979         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1980                 return -EPERM;
1981
1982         switch (cmd) {
1983         case IPT_SO_GET_INFO:
1984                 ret = get_info(sock_net(sk), user, len, 1);
1985                 break;
1986         case IPT_SO_GET_ENTRIES:
1987                 ret = compat_get_entries(sock_net(sk), user, len);
1988                 break;
1989         default:
1990                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1991         }
1992         return ret;
1993 }
1994 #endif
1995
1996 static int
1997 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1998 {
1999         int ret;
2000
2001         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2002                 return -EPERM;
2003
2004         switch (cmd) {
2005         case IPT_SO_SET_REPLACE:
2006                 ret = do_replace(sock_net(sk), user, len);
2007                 break;
2008
2009         case IPT_SO_SET_ADD_COUNTERS:
2010                 ret = do_add_counters(sock_net(sk), user, len, 0);
2011                 break;
2012
2013         default:
2014                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2015                 ret = -EINVAL;
2016         }
2017
2018         return ret;
2019 }
2020
2021 static int
2022 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2023 {
2024         int ret;
2025
2026         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2027                 return -EPERM;
2028
2029         switch (cmd) {
2030         case IPT_SO_GET_INFO:
2031                 ret = get_info(sock_net(sk), user, len, 0);
2032                 break;
2033
2034         case IPT_SO_GET_ENTRIES:
2035                 ret = get_entries(sock_net(sk), user, len);
2036                 break;
2037
2038         case IPT_SO_GET_REVISION_MATCH:
2039         case IPT_SO_GET_REVISION_TARGET: {
2040                 struct xt_get_revision rev;
2041                 int target;
2042
2043                 if (*len != sizeof(rev)) {
2044                         ret = -EINVAL;
2045                         break;
2046                 }
2047                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2048                         ret = -EFAULT;
2049                         break;
2050                 }
2051                 rev.name[sizeof(rev.name)-1] = 0;
2052
2053                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2054                         target = 1;
2055                 else
2056                         target = 0;
2057
2058                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2059                                                          rev.revision,
2060                                                          target, &ret),
2061                                         "ipt_%s", rev.name);
2062                 break;
2063         }
2064
2065         default:
2066                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2067                 ret = -EINVAL;
2068         }
2069
2070         return ret;
2071 }
2072
2073 struct xt_table *ipt_register_table(struct net *net,
2074                                     const struct xt_table *table,
2075                                     const struct ipt_replace *repl)
2076 {
2077         int ret;
2078         struct xt_table_info *newinfo;
2079         struct xt_table_info bootstrap = {0};
2080         void *loc_cpu_entry;
2081         struct xt_table *new_table;
2082
2083         newinfo = xt_alloc_table_info(repl->size);
2084         if (!newinfo) {
2085                 ret = -ENOMEM;
2086                 goto out;
2087         }
2088
2089         loc_cpu_entry = newinfo->entries;
2090         memcpy(loc_cpu_entry, repl->entries, repl->size);
2091
2092         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2093         if (ret != 0)
2094                 goto out_free;
2095
2096         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2097         if (IS_ERR(new_table)) {
2098                 ret = PTR_ERR(new_table);
2099                 goto out_free;
2100         }
2101
2102         return new_table;
2103
2104 out_free:
2105         xt_free_table_info(newinfo);
2106 out:
2107         return ERR_PTR(ret);
2108 }
2109
2110 void ipt_unregister_table(struct net *net, struct xt_table *table)
2111 {
2112         struct xt_table_info *private;
2113         void *loc_cpu_entry;
2114         struct module *table_owner = table->me;
2115         struct ipt_entry *iter;
2116
2117         private = xt_unregister_table(table);
2118
2119         /* Decrease module usage counts and free resources */
2120         loc_cpu_entry = private->entries;
2121         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2122                 cleanup_entry(iter, net);
2123         if (private->number > private->initial_entries)
2124                 module_put(table_owner);
2125         xt_free_table_info(private);
2126 }
2127
2128 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2129 static inline bool
2130 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2131                      u_int8_t type, u_int8_t code,
2132                      bool invert)
2133 {
2134         return ((test_type == 0xFF) ||
2135                 (type == test_type && code >= min_code && code <= max_code))
2136                 ^ invert;
2137 }
2138
2139 static bool
2140 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2141 {
2142         const struct icmphdr *ic;
2143         struct icmphdr _icmph;
2144         const struct ipt_icmp *icmpinfo = par->matchinfo;
2145
2146         /* Must not be a fragment. */
2147         if (par->fragoff != 0)
2148                 return false;
2149
2150         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2151         if (ic == NULL) {
2152                 /* We've been asked to examine this packet, and we
2153                  * can't.  Hence, no choice but to drop.
2154                  */
2155                 duprintf("Dropping evil ICMP tinygram.\n");
2156                 par->hotdrop = true;
2157                 return false;
2158         }
2159
2160         return icmp_type_code_match(icmpinfo->type,
2161                                     icmpinfo->code[0],
2162                                     icmpinfo->code[1],
2163                                     ic->type, ic->code,
2164                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2165 }
2166
2167 static int icmp_checkentry(const struct xt_mtchk_param *par)
2168 {
2169         const struct ipt_icmp *icmpinfo = par->matchinfo;
2170
2171         /* Must specify no unknown invflags */
2172         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2173 }
2174
2175 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2176         {
2177                 .name             = XT_STANDARD_TARGET,
2178                 .targetsize       = sizeof(int),
2179                 .family           = NFPROTO_IPV4,
2180 #ifdef CONFIG_COMPAT
2181                 .compatsize       = sizeof(compat_int_t),
2182                 .compat_from_user = compat_standard_from_user,
2183                 .compat_to_user   = compat_standard_to_user,
2184 #endif
2185         },
2186         {
2187                 .name             = XT_ERROR_TARGET,
2188                 .target           = ipt_error,
2189                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2190                 .family           = NFPROTO_IPV4,
2191         },
2192 };
2193
2194 static struct nf_sockopt_ops ipt_sockopts = {
2195         .pf             = PF_INET,
2196         .set_optmin     = IPT_BASE_CTL,
2197         .set_optmax     = IPT_SO_SET_MAX+1,
2198         .set            = do_ipt_set_ctl,
2199 #ifdef CONFIG_COMPAT
2200         .compat_set     = compat_do_ipt_set_ctl,
2201 #endif
2202         .get_optmin     = IPT_BASE_CTL,
2203         .get_optmax     = IPT_SO_GET_MAX+1,
2204         .get            = do_ipt_get_ctl,
2205 #ifdef CONFIG_COMPAT
2206         .compat_get     = compat_do_ipt_get_ctl,
2207 #endif
2208         .owner          = THIS_MODULE,
2209 };
2210
2211 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2212         {
2213                 .name       = "icmp",
2214                 .match      = icmp_match,
2215                 .matchsize  = sizeof(struct ipt_icmp),
2216                 .checkentry = icmp_checkentry,
2217                 .proto      = IPPROTO_ICMP,
2218                 .family     = NFPROTO_IPV4,
2219         },
2220 };
2221
2222 static int __net_init ip_tables_net_init(struct net *net)
2223 {
2224         return xt_proto_init(net, NFPROTO_IPV4);
2225 }
2226
2227 static void __net_exit ip_tables_net_exit(struct net *net)
2228 {
2229         xt_proto_fini(net, NFPROTO_IPV4);
2230 }
2231
2232 static struct pernet_operations ip_tables_net_ops = {
2233         .init = ip_tables_net_init,
2234         .exit = ip_tables_net_exit,
2235 };
2236
2237 static int __init ip_tables_init(void)
2238 {
2239         int ret;
2240
2241         ret = register_pernet_subsys(&ip_tables_net_ops);
2242         if (ret < 0)
2243                 goto err1;
2244
2245         /* No one else will be downing sem now, so we won't sleep */
2246         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2247         if (ret < 0)
2248                 goto err2;
2249         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2250         if (ret < 0)
2251                 goto err4;
2252
2253         /* Register setsockopt */
2254         ret = nf_register_sockopt(&ipt_sockopts);
2255         if (ret < 0)
2256                 goto err5;
2257
2258         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2259         return 0;
2260
2261 err5:
2262         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2263 err4:
2264         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2265 err2:
2266         unregister_pernet_subsys(&ip_tables_net_ops);
2267 err1:
2268         return ret;
2269 }
2270
2271 static void __exit ip_tables_fini(void)
2272 {
2273         nf_unregister_sockopt(&ipt_sockopts);
2274
2275         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2276         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2277         unregister_pernet_subsys(&ip_tables_net_ops);
2278 }
2279
2280 EXPORT_SYMBOL(ipt_register_table);
2281 EXPORT_SYMBOL(ipt_unregister_table);
2282 EXPORT_SYMBOL(ipt_do_table);
2283 module_init(ip_tables_init);
2284 module_exit(ip_tables_fini);