net/ipv4/netfilter/ip_tables.c

   1 /*
   2  * Packet matching code.
   3  *
   4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   6  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13 #include <linux/cache.h>
  14 #include <linux/capability.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/kmod.h>
  17 #include <linux/vmalloc.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/module.h>
  20 #include <linux/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/compat.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/mutex.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/err.h>
  27 #include <linux/cpumask.h>
  28
  29 #include <linux/netfilter/x_tables.h>
  30 #include <linux/netfilter_ipv4/ip_tables.h>
  31 #include <net/netfilter/nf_log.h>
  32 #include "../../netfilter/xt_repldata.h"
  33
  34 MODULE_LICENSE("GPL");
  35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36 MODULE_DESCRIPTION("IPv4 packet filter");
  37
  38 /*#define DEBUG_IP_FIREWALL*/
  39 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  40 /*#define DEBUG_IP_FIREWALL_USER*/
  41
  42 #ifdef DEBUG_IP_FIREWALL
  43 #define dprintf(format, args...) pr_info(format , ## args)
  44 #else
  45 #define dprintf(format, args...)
  46 #endif
  47
  48 #ifdef DEBUG_IP_FIREWALL_USER
  49 #define duprintf(format, args...) pr_info(format , ## args)
  50 #else
  51 #define duprintf(format, args...)
  52 #endif
  53
  54 #ifdef CONFIG_NETFILTER_DEBUG
  55 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  56 #else
  57 #define IP_NF_ASSERT(x)
  58 #endif
  59
  60 #if 0
  61 /* All the better to debug you with... */
  62 #define static
  63 #define inline
  64 #endif
  65
  66 void *ipt_alloc_initial_table(const struct xt_table *info)
  67 {
  68         return xt_alloc_initial_table(ipt, IPT);
  69 }
  70 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  71
  72 /* Returns whether matches rule or not. */
  73 /* Performance critical - called for every packet */
  74 static inline bool
  75 ip_packet_match(const struct iphdr *ip,
  76                 const char *indev,
  77                 const char *outdev,
  78                 const struct ipt_ip *ipinfo,
  79                 int isfrag)
  80 {
  81         unsigned long ret;
  82
  83 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  84
  85         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  86                   IPT_INV_SRCIP) ||
  87             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  88                   IPT_INV_DSTIP)) {
  89                 dprintf("Source or dest mismatch.\n");
  90
  91                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  92                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  93                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  94                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  95                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  96                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  97                 return false;
  98         }
  99
 100         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 101
 102         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 103                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 104                         indev, ipinfo->iniface,
 105                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
 106                 return false;
 107         }
 108
 109         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 110
 111         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 112                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 113                         outdev, ipinfo->outiface,
 114                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
 115                 return false;
 116         }
 117
 118         /* Check specific protocol */
 119         if (ipinfo->proto &&
 120             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 121                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 122                         ip->protocol, ipinfo->proto,
 123                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
 124                 return false;
 125         }
 126
 127         /* If we have a fragment rule but the packet is not a fragment
 128          * then we return zero */
 129         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 130                 dprintf("Fragment rule but not fragment.%s\n",
 131                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 132                 return false;
 133         }
 134
 135         return true;
 136 }
 137
 138 static bool
 139 ip_checkentry(const struct ipt_ip *ip)
 140 {
 141         if (ip->flags & ~IPT_F_MASK) {
 142                 duprintf("Unknown flag bits set: %08X\n",
 143                          ip->flags & ~IPT_F_MASK);
 144                 return false;
 145         }
 146         if (ip->invflags & ~IPT_INV_MASK) {
 147                 duprintf("Unknown invflag bits set: %08X\n",
 148                          ip->invflags & ~IPT_INV_MASK);
 149                 return false;
 150         }
 151         return true;
 152 }
 153
 154 static unsigned int
 155 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 156 {
 157         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 158
 159         return NF_DROP;
 160 }
 161
 162 /* Performance critical */
 163 static inline struct ipt_entry *
 164 get_entry(const void *base, unsigned int offset)
 165 {
 166         return (struct ipt_entry *)(base + offset);
 167 }
 168
 169 /* All zeroes == unconditional rule. */
 170 /* Mildly perf critical (only if packet tracing is on) */
 171 static inline bool unconditional(const struct ipt_ip *ip)
 172 {
 173         static const struct ipt_ip uncond;
 174
 175         return memcmp(ip, &uncond, sizeof(uncond)) == 0;
 176 #undef FWINV
 177 }
 178
 179 /* for const-correctness */
 180 static inline const struct xt_entry_target *
 181 ipt_get_target_c(const struct ipt_entry *e)
 182 {
 183         return ipt_get_target((struct ipt_entry *)e);
 184 }
 185
 186 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 187 static const char *const hooknames[] = {
 188         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 189         [NF_INET_LOCAL_IN]              = "INPUT",
 190         [NF_INET_FORWARD]               = "FORWARD",
 191         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 192         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 193 };
 194
 195 enum nf_ip_trace_comments {
 196         NF_IP_TRACE_COMMENT_RULE,
 197         NF_IP_TRACE_COMMENT_RETURN,
 198         NF_IP_TRACE_COMMENT_POLICY,
 199 };
 200
 201 static const char *const comments[] = {
 202         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 203         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 204         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 205 };
 206
 207 static struct nf_loginfo trace_loginfo = {
 208         .type = NF_LOG_TYPE_LOG,
 209         .u = {
 210                 .log = {
 211                         .level = 4,
 212                         .logflags = NF_LOG_MASK,
 213                 },
 214         },
 215 };
 216
 217 /* Mildly perf critical (only if packet tracing is on) */
 218 static inline int
 219 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 220                       const char *hookname, const char **chainname,
 221                       const char **comment, unsigned int *rulenum)
 222 {
 223         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 224
 225         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 226                 /* Head of user chain: ERROR target with chainname */
 227                 *chainname = t->target.data;
 228                 (*rulenum) = 0;
 229         } else if (s == e) {
 230                 (*rulenum)++;
 231
 232                 if (s->target_offset == sizeof(struct ipt_entry) &&
 233                     strcmp(t->target.u.kernel.target->name,
 234                            XT_STANDARD_TARGET) == 0 &&
 235                    t->verdict < 0 &&
 236                    unconditional(&s->ip)) {
 237                         /* Tail of chains: STANDARD target (return/policy) */
 238                         *comment = *chainname == hookname
 239                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 240                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 241                 }
 242                 return 1;
 243         } else
 244                 (*rulenum)++;
 245
 246         return 0;
 247 }
 248
 249 static void trace_packet(const struct sk_buff *skb,
 250                          unsigned int hook,
 251                          const struct net_device *in,
 252                          const struct net_device *out,
 253                          const char *tablename,
 254                          const struct xt_table_info *private,
 255                          const struct ipt_entry *e)
 256 {
 257         const struct ipt_entry *root;
 258         const char *hookname, *chainname, *comment;
 259         const struct ipt_entry *iter;
 260         unsigned int rulenum = 0;
 261         struct net *net = dev_net(in ? in : out);
 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 __pure
 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              unsigned int hook,
 289              const struct nf_hook_state *state,
 290              struct xt_table *table)
 291 {
 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.in      = state->in;
 319         acpar.out     = state->out;
 320         acpar.family  = NFPROTO_IPV4;
 321         acpar.hooknum = hook;
 322
 323         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 324         local_bh_disable();
 325         addend = xt_write_recseq_begin();
 326         private = table->private;
 327         cpu        = smp_processor_id();
 328         /*
 329          * Ensure we load private-> members after we've fetched the base
 330          * pointer.
 331          */
 332         smp_read_barrier_depends();
 333         table_base = private->entries;
 334         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 335
 336         /* Switch to alternate jumpstack if we're being invoked via TEE.
 337          * TEE issues XT_CONTINUE verdict on original skb so we must not
 338          * clobber the jumpstack.
 339          *
 340          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 341          * but it is no problem since absolute verdict is issued by these.
 342          */
 343         jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 344
 345         e = get_entry(table_base, private->hook_entry[hook]);
 346
 347         pr_debug("Entering %s(hook %u), UF %p\n",
 348                  table->name, hook,
 349                  get_entry(table_base, private->underflow[hook]));
 350
 351         do {
 352                 const struct xt_entry_target *t;
 353                 const struct xt_entry_match *ematch;
 354                 struct xt_counters *counter;
 355
 356                 IP_NF_ASSERT(e);
 357                 if (!ip_packet_match(ip, indev, outdev,
 358                     &e->ip, acpar.fragoff)) {
 359  no_match:
 360                         e = ipt_next_entry(e);
 361                         continue;
 362                 }
 363
 364                 xt_ematch_foreach(ematch, e) {
 365                         acpar.match     = ematch->u.kernel.match;
 366                         acpar.matchinfo = ematch->data;
 367                         if (!acpar.match->match(skb, &acpar))
 368                                 goto no_match;
 369                 }
 370
 371                 counter = xt_get_this_cpu_counter(&e->counters);
 372                 ADD_COUNTER(*counter, skb->len, 1);
 373
 374                 t = ipt_get_target(e);
 375                 IP_NF_ASSERT(t->u.kernel.target);
 376
 377 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 378                 /* The packet is traced: log it */
 379                 if (unlikely(skb->nf_trace))
 380                         trace_packet(skb, hook, state->in, state->out,
 381                                      table->name, private, e);
 382 #endif
 383                 /* Standard target? */
 384                 if (!t->u.kernel.target->target) {
 385                         int v;
 386
 387                         v = ((struct xt_standard_target *)t)->verdict;
 388                         if (v < 0) {
 389                                 /* Pop from stack? */
 390                                 if (v != XT_RETURN) {
 391                                         verdict = (unsigned int)(-v) - 1;
 392                                         break;
 393                                 }
 394                                 if (stackidx == 0) {
 395                                         e = get_entry(table_base,
 396                                             private->underflow[hook]);
 397                                         pr_debug("Underflow (this is normal) "
 398                                                  "to %p\n", e);
 399                                 } else {
 400                                         e = jumpstack[--stackidx];
 401                                         pr_debug("Pulled %p out from pos %u\n",
 402                                                  e, stackidx);
 403                                         e = ipt_next_entry(e);
 404                                 }
 405                                 continue;
 406                         }
 407                         if (table_base + v != ipt_next_entry(e) &&
 408                             !(e->ip.flags & IPT_F_GOTO)) {
 409                                 jumpstack[stackidx++] = e;
 410                                 pr_debug("Pushed %p into pos %u\n",
 411                                          e, stackidx - 1);
 412                         }
 413
 414                         e = get_entry(table_base, v);
 415                         continue;
 416                 }
 417
 418                 acpar.target   = t->u.kernel.target;
 419                 acpar.targinfo = t->data;
 420
 421                 verdict = t->u.kernel.target->target(skb, &acpar);
 422                 /* Target might have changed stuff. */
 423                 ip = ip_hdr(skb);
 424                 if (verdict == XT_CONTINUE)
 425                         e = ipt_next_entry(e);
 426                 else
 427                         /* Verdict */
 428                         break;
 429         } while (!acpar.hotdrop);
 430         pr_debug("Exiting %s; sp at %u\n", __func__, stackidx);
 431
 432         xt_write_recseq_end(addend);
 433         local_bh_enable();
 434
 435 #ifdef DEBUG_ALLOW_ALL
 436         return NF_ACCEPT;
 437 #else
 438         if (acpar.hotdrop)
 439                 return NF_DROP;
 440         else return verdict;
 441 #endif
 442 }
 443
 444 /* Figures out from what hook each rule can be called: returns 0 if
 445  * there are loops.  Puts hook bitmask in comefrom.
 446  *
 447  * Keeps track of largest call depth seen and stores it in newinfo->stacksize.
 448  */
 449 static int
 450 mark_source_chains(struct xt_table_info *newinfo,
 451                    unsigned int valid_hooks, void *entry0)
 452 {
 453         unsigned int calldepth, max_calldepth = 0;
 454         unsigned int hook;
 455
 456         /* No recursion; use packet counter to save back ptrs (reset
 457            to 0 as we leave), and comefrom to save source hook bitmask */
 458         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 459                 unsigned int pos = newinfo->hook_entry[hook];
 460                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 461
 462                 if (!(valid_hooks & (1 << hook)))
 463                         continue;
 464
 465                 /* Set initial back pointer. */
 466                 e->counters.pcnt = pos;
 467                 calldepth = 0;
 468
 469                 for (;;) {
 470                         const struct xt_standard_target *t
 471                                 = (void *)ipt_get_target_c(e);
 472                         int visited = e->comefrom & (1 << hook);
 473
 474                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 475                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 476                                        hook, pos, e->comefrom);
 477                                 return 0;
 478                         }
 479                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 480
 481                         /* Unconditional return/END. */
 482                         if ((e->target_offset == sizeof(struct ipt_entry) &&
 483                              (strcmp(t->target.u.user.name,
 484                                      XT_STANDARD_TARGET) == 0) &&
 485                              t->verdict < 0 && unconditional(&e->ip)) ||
 486                             visited) {
 487                                 unsigned int oldpos, size;
 488
 489                                 if ((strcmp(t->target.u.user.name,
 490                                             XT_STANDARD_TARGET) == 0) &&
 491                                     t->verdict < -NF_MAX_VERDICT - 1) {
 492                                         duprintf("mark_source_chains: bad "
 493                                                 "negative verdict (%i)\n",
 494                                                                 t->verdict);
 495                                         return 0;
 496                                 }
 497
 498                                 /* Return: backtrack through the last
 499                                    big jump. */
 500                                 do {
 501                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 502 #ifdef DEBUG_IP_FIREWALL_USER
 503                                         if (e->comefrom
 504                                             & (1 << NF_INET_NUMHOOKS)) {
 505                                                 duprintf("Back unset "
 506                                                          "on hook %u "
 507                                                          "rule %u\n",
 508                                                          hook, pos);
 509                                         }
 510 #endif
 511                                         oldpos = pos;
 512                                         pos = e->counters.pcnt;
 513                                         e->counters.pcnt = 0;
 514
 515                                         /* We're at the start. */
 516                                         if (pos == oldpos)
 517                                                 goto next;
 518
 519                                         e = (struct ipt_entry *)
 520                                                 (entry0 + pos);
 521                                 } while (oldpos == pos + e->next_offset);
 522
 523                                 /* Move along one */
 524                                 size = e->next_offset;
 525                                 e = (struct ipt_entry *)
 526                                         (entry0 + pos + size);
 527                                 e->counters.pcnt = pos;
 528                                 pos += size;
 529                                 WARN_ON_ONCE(calldepth == 0);
 530                                 if (calldepth > 0)
 531                                         --calldepth;
 532                         } else {
 533                                 int newpos = t->verdict;
 534
 535                                 if (strcmp(t->target.u.user.name,
 536                                            XT_STANDARD_TARGET) == 0 &&
 537                                     newpos >= 0) {
 538                                         if (newpos > newinfo->size -
 539                                                 sizeof(struct ipt_entry)) {
 540                                                 duprintf("mark_source_chains: "
 541                                                         "bad verdict (%i)\n",
 542                                                                 newpos);
 543                                                 return 0;
 544                                         }
 545                                         if (entry0 + newpos != ipt_next_entry(e) &&
 546                                             !(e->ip.flags & IPT_F_GOTO) &&
 547                                             ++calldepth > max_calldepth)
 548                                                 max_calldepth = calldepth;
 549
 550                                         /* This a jump; chase it. */
 551                                         duprintf("Jump rule %u -> %u, calldepth %d\n",
 552                                                  pos, newpos, calldepth);
 553                                 } else {
 554                                         /* ... this is a fallthru */
 555                                         newpos = pos + e->next_offset;
 556                                 }
 557                                 e = (struct ipt_entry *)
 558                                         (entry0 + newpos);
 559                                 e->counters.pcnt = pos;
 560                                 pos = newpos;
 561                         }
 562                 }
 563                 next:
 564                 duprintf("Finished chain %u\n", hook);
 565         }
 566         newinfo->stacksize = max_calldepth;
 567         return 1;
 568 }
 569
 570 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 571 {
 572         struct xt_mtdtor_param par;
 573
 574         par.net       = net;
 575         par.match     = m->u.kernel.match;
 576         par.matchinfo = m->data;
 577         par.family    = NFPROTO_IPV4;
 578         if (par.match->destroy != NULL)
 579                 par.match->destroy(&par);
 580         module_put(par.match->me);
 581 }
 582
 583 static int
 584 check_entry(const struct ipt_entry *e, const char *name)
 585 {
 586         const struct xt_entry_target *t;
 587
 588         if (!ip_checkentry(&e->ip)) {
 589                 duprintf("ip check failed %p %s.\n", e, name);
 590                 return -EINVAL;
 591         }
 592
 593         if (e->target_offset + sizeof(struct xt_entry_target) >
 594             e->next_offset)
 595                 return -EINVAL;
 596
 597         t = ipt_get_target_c(e);
 598         if (e->target_offset + t->u.target_size > e->next_offset)
 599                 return -EINVAL;
 600
 601         return 0;
 602 }
 603
 604 static int
 605 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 606 {
 607         const struct ipt_ip *ip = par->entryinfo;
 608         int ret;
 609
 610         par->match     = m->u.kernel.match;
 611         par->matchinfo = m->data;
 612
 613         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 614               ip->proto, ip->invflags & IPT_INV_PROTO);
 615         if (ret < 0) {
 616                 duprintf("check failed for `%s'.\n", par->match->name);
 617                 return ret;
 618         }
 619         return 0;
 620 }
 621
 622 static int
 623 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 624 {
 625         struct xt_match *match;
 626         int ret;
 627
 628         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 629                                       m->u.user.revision);
 630         if (IS_ERR(match)) {
 631                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 632                 return PTR_ERR(match);
 633         }
 634         m->u.kernel.match = match;
 635
 636         ret = check_match(m, par);
 637         if (ret)
 638                 goto err;
 639
 640         return 0;
 641 err:
 642         module_put(m->u.kernel.match->me);
 643         return ret;
 644 }
 645
 646 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 647 {
 648         struct xt_entry_target *t = ipt_get_target(e);
 649         struct xt_tgchk_param par = {
 650                 .net       = net,
 651                 .table     = name,
 652                 .entryinfo = e,
 653                 .target    = t->u.kernel.target,
 654                 .targinfo  = t->data,
 655                 .hook_mask = e->comefrom,
 656                 .family    = NFPROTO_IPV4,
 657         };
 658         int ret;
 659
 660         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 661               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 662         if (ret < 0) {
 663                 duprintf("check failed for `%s'.\n",
 664                          t->u.kernel.target->name);
 665                 return ret;
 666         }
 667         return 0;
 668 }
 669
 670 static int
 671 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 672                  unsigned int size)
 673 {
 674         struct xt_entry_target *t;
 675         struct xt_target *target;
 676         int ret;
 677         unsigned int j;
 678         struct xt_mtchk_param mtpar;
 679         struct xt_entry_match *ematch;
 680
 681         ret = check_entry(e, name);
 682         if (ret)
 683                 return ret;
 684
 685         e->counters.pcnt = xt_percpu_counter_alloc();
 686         if (IS_ERR_VALUE(e->counters.pcnt))
 687                 return -ENOMEM;
 688
 689         j = 0;
 690         mtpar.net       = net;
 691         mtpar.table     = name;
 692         mtpar.entryinfo = &e->ip;
 693         mtpar.hook_mask = e->comefrom;
 694         mtpar.family    = NFPROTO_IPV4;
 695         xt_ematch_foreach(ematch, e) {
 696                 ret = find_check_match(ematch, &mtpar);
 697                 if (ret != 0)
 698                         goto cleanup_matches;
 699                 ++j;
 700         }
 701
 702         t = ipt_get_target(e);
 703         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 704                                         t->u.user.revision);
 705         if (IS_ERR(target)) {
 706                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 707                 ret = PTR_ERR(target);
 708                 goto cleanup_matches;
 709         }
 710         t->u.kernel.target = target;
 711
 712         ret = check_target(e, net, name);
 713         if (ret)
 714                 goto err;
 715
 716         return 0;
 717  err:
 718         module_put(t->u.kernel.target->me);
 719  cleanup_matches:
 720         xt_ematch_foreach(ematch, e) {
 721                 if (j-- == 0)
 722                         break;
 723                 cleanup_match(ematch, net);
 724         }
 725
 726         xt_percpu_counter_free(e->counters.pcnt);
 727
 728         return ret;
 729 }
 730
 731 static bool check_underflow(const struct ipt_entry *e)
 732 {
 733         const struct xt_entry_target *t;
 734         unsigned int verdict;
 735
 736         if (!unconditional(&e->ip))
 737                 return false;
 738         t = ipt_get_target_c(e);
 739         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 740                 return false;
 741         verdict = ((struct xt_standard_target *)t)->verdict;
 742         verdict = -verdict - 1;
 743         return verdict == NF_DROP || verdict == NF_ACCEPT;
 744 }
 745
 746 static int
 747 check_entry_size_and_hooks(struct ipt_entry *e,
 748                            struct xt_table_info *newinfo,
 749                            const unsigned char *base,
 750                            const unsigned char *limit,
 751                            const unsigned int *hook_entries,
 752                            const unsigned int *underflows,
 753                            unsigned int valid_hooks)
 754 {
 755         unsigned int h;
 756
 757         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 758             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 759                 duprintf("Bad offset %p\n", e);
 760                 return -EINVAL;
 761         }
 762
 763         if (e->next_offset
 764             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 765                 duprintf("checking: element %p size %u\n",
 766                          e, e->next_offset);
 767                 return -EINVAL;
 768         }
 769
 770         /* Check hooks & underflows */
 771         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 772                 if (!(valid_hooks & (1 << h)))
 773                         continue;
 774                 if ((unsigned char *)e - base == hook_entries[h])
 775                         newinfo->hook_entry[h] = hook_entries[h];
 776                 if ((unsigned char *)e - base == underflows[h]) {
 777                         if (!check_underflow(e)) {
 778                                 pr_err("Underflows must be unconditional and "
 779                                        "use the STANDARD target with "
 780                                        "ACCEPT/DROP\n");
 781                                 return -EINVAL;
 782                         }
 783                         newinfo->underflow[h] = underflows[h];
 784                 }
 785         }
 786
 787         /* Clear counters and comefrom */
 788         e->counters = ((struct xt_counters) { 0, 0 });
 789         e->comefrom = 0;
 790         return 0;
 791 }
 792
 793 static void
 794 cleanup_entry(struct ipt_entry *e, struct net *net)
 795 {
 796         struct xt_tgdtor_param par;
 797         struct xt_entry_target *t;
 798         struct xt_entry_match *ematch;
 799
 800         /* Cleanup all matches */
 801         xt_ematch_foreach(ematch, e)
 802                 cleanup_match(ematch, net);
 803         t = ipt_get_target(e);
 804
 805         par.net      = net;
 806         par.target   = t->u.kernel.target;
 807         par.targinfo = t->data;
 808         par.family   = NFPROTO_IPV4;
 809         if (par.target->destroy != NULL)
 810                 par.target->destroy(&par);
 811         module_put(par.target->me);
 812         xt_percpu_counter_free(e->counters.pcnt);
 813 }
 814
 815 /* Checks and translates the user-supplied table segment (held in
 816    newinfo) */
 817 static int
 818 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 819                 const struct ipt_replace *repl)
 820 {
 821         struct ipt_entry *iter;
 822         unsigned int i;
 823         int ret = 0;
 824
 825         newinfo->size = repl->size;
 826         newinfo->number = repl->num_entries;
 827
 828         /* Init all hooks to impossible value. */
 829         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 830                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 831                 newinfo->underflow[i] = 0xFFFFFFFF;
 832         }
 833
 834         duprintf("translate_table: size %u\n", newinfo->size);
 835         i = 0;
 836         /* Walk through entries, checking offsets. */
 837         xt_entry_foreach(iter, entry0, newinfo->size) {
 838                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 839                                                  entry0 + repl->size,
 840                                                  repl->hook_entry,
 841                                                  repl->underflow,
 842                                                  repl->valid_hooks);
 843                 if (ret != 0)
 844                         return ret;
 845                 ++i;
 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
1170         t = xt_find_table_lock(net, AF_INET, get.name);
1171         if (!IS_ERR_OR_NULL(t)) {
1172                 const struct xt_table_info *private = t->private;
1173                 duprintf("t->private->number = %u\n", private->number);
1174                 if (get.size == private->size)
1175                         ret = copy_entries_to_user(private->size,
1176                                                    t, uptr->entrytable);
1177                 else {
1178                         duprintf("get_entries: I've got %u not %u!\n",
1179                                  private->size, get.size);
1180                         ret = -EAGAIN;
1181                 }
1182                 module_put(t->me);
1183                 xt_table_unlock(t);
1184         } else
1185                 ret = t ? PTR_ERR(t) : -ENOENT;
1186
1187         return ret;
1188 }
1189
1190 static int
1191 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1192              struct xt_table_info *newinfo, unsigned int num_counters,
1193              void __user *counters_ptr)
1194 {
1195         int ret;
1196         struct xt_table *t;
1197         struct xt_table_info *oldinfo;
1198         struct xt_counters *counters;
1199         struct ipt_entry *iter;
1200
1201         ret = 0;
1202         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1203         if (!counters) {
1204                 ret = -ENOMEM;
1205                 goto out;
1206         }
1207
1208         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1209                                     "iptable_%s", name);
1210         if (IS_ERR_OR_NULL(t)) {
1211                 ret = t ? PTR_ERR(t) : -ENOENT;
1212                 goto free_newinfo_counters_untrans;
1213         }
1214
1215         /* You lied! */
1216         if (valid_hooks != t->valid_hooks) {
1217                 duprintf("Valid hook crap: %08X vs %08X\n",
1218                          valid_hooks, t->valid_hooks);
1219                 ret = -EINVAL;
1220                 goto put_module;
1221         }
1222
1223         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1224         if (!oldinfo)
1225                 goto put_module;
1226
1227         /* Update module usage count based on number of rules */
1228         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1229                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1230         if ((oldinfo->number > oldinfo->initial_entries) ||
1231             (newinfo->number <= oldinfo->initial_entries))
1232                 module_put(t->me);
1233         if ((oldinfo->number > oldinfo->initial_entries) &&
1234             (newinfo->number <= oldinfo->initial_entries))
1235                 module_put(t->me);
1236
1237         /* Get the old counters, and synchronize with replace */
1238         get_counters(oldinfo, counters);
1239
1240         /* Decrease module usage counts and free resource */
1241         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1242                 cleanup_entry(iter, net);
1243
1244         xt_free_table_info(oldinfo);
1245         if (copy_to_user(counters_ptr, counters,
1246                          sizeof(struct xt_counters) * num_counters) != 0) {
1247                 /* Silent error, can't fail, new table is already in place */
1248                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1249         }
1250         vfree(counters);
1251         xt_table_unlock(t);
1252         return ret;
1253
1254  put_module:
1255         module_put(t->me);
1256         xt_table_unlock(t);
1257  free_newinfo_counters_untrans:
1258         vfree(counters);
1259  out:
1260         return ret;
1261 }
1262
1263 static int
1264 do_replace(struct net *net, const void __user *user, unsigned int len)
1265 {
1266         int ret;
1267         struct ipt_replace tmp;
1268         struct xt_table_info *newinfo;
1269         void *loc_cpu_entry;
1270         struct ipt_entry *iter;
1271
1272         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1273                 return -EFAULT;
1274
1275         /* overflow check */
1276         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1277                 return -ENOMEM;
1278         if (tmp.num_counters == 0)
1279                 return -EINVAL;
1280
1281         tmp.name[sizeof(tmp.name)-1] = 0;
1282
1283         newinfo = xt_alloc_table_info(tmp.size);
1284         if (!newinfo)
1285                 return -ENOMEM;
1286
1287         loc_cpu_entry = newinfo->entries;
1288         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1289                            tmp.size) != 0) {
1290                 ret = -EFAULT;
1291                 goto free_newinfo;
1292         }
1293
1294         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1295         if (ret != 0)
1296                 goto free_newinfo;
1297
1298         duprintf("Translated table\n");
1299
1300         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1301                            tmp.num_counters, tmp.counters);
1302         if (ret)
1303                 goto free_newinfo_untrans;
1304         return 0;
1305
1306  free_newinfo_untrans:
1307         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1308                 cleanup_entry(iter, net);
1309  free_newinfo:
1310         xt_free_table_info(newinfo);
1311         return ret;
1312 }
1313
1314 static int
1315 do_add_counters(struct net *net, const void __user *user,
1316                 unsigned int len, int compat)
1317 {
1318         unsigned int i;
1319         struct xt_counters_info tmp;
1320         struct xt_counters *paddc;
1321         unsigned int num_counters;
1322         const char *name;
1323         int size;
1324         void *ptmp;
1325         struct xt_table *t;
1326         const struct xt_table_info *private;
1327         int ret = 0;
1328         struct ipt_entry *iter;
1329         unsigned int addend;
1330 #ifdef CONFIG_COMPAT
1331         struct compat_xt_counters_info compat_tmp;
1332
1333         if (compat) {
1334                 ptmp = &compat_tmp;
1335                 size = sizeof(struct compat_xt_counters_info);
1336         } else
1337 #endif
1338         {
1339                 ptmp = &tmp;
1340                 size = sizeof(struct xt_counters_info);
1341         }
1342
1343         if (copy_from_user(ptmp, user, size) != 0)
1344                 return -EFAULT;
1345
1346 #ifdef CONFIG_COMPAT
1347         if (compat) {
1348                 num_counters = compat_tmp.num_counters;
1349                 name = compat_tmp.name;
1350         } else
1351 #endif
1352         {
1353                 num_counters = tmp.num_counters;
1354                 name = tmp.name;
1355         }
1356
1357         if (len != size + num_counters * sizeof(struct xt_counters))
1358                 return -EINVAL;
1359
1360         paddc = vmalloc(len - size);
1361         if (!paddc)
1362                 return -ENOMEM;
1363
1364         if (copy_from_user(paddc, user + size, len - size) != 0) {
1365                 ret = -EFAULT;
1366                 goto free;
1367         }
1368
1369         t = xt_find_table_lock(net, AF_INET, name);
1370         if (IS_ERR_OR_NULL(t)) {
1371                 ret = t ? PTR_ERR(t) : -ENOENT;
1372                 goto free;
1373         }
1374
1375         local_bh_disable();
1376         private = t->private;
1377         if (private->number != num_counters) {
1378                 ret = -EINVAL;
1379                 goto unlock_up_free;
1380         }
1381
1382         i = 0;
1383         addend = xt_write_recseq_begin();
1384         xt_entry_foreach(iter, private->entries, private->size) {
1385                 struct xt_counters *tmp;
1386
1387                 tmp = xt_get_this_cpu_counter(&iter->counters);
1388                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1389                 ++i;
1390         }
1391         xt_write_recseq_end(addend);
1392  unlock_up_free:
1393         local_bh_enable();
1394         xt_table_unlock(t);
1395         module_put(t->me);
1396  free:
1397         vfree(paddc);
1398
1399         return ret;
1400 }
1401
1402 #ifdef CONFIG_COMPAT
1403 struct compat_ipt_replace {
1404         char                    name[XT_TABLE_MAXNAMELEN];
1405         u32                     valid_hooks;
1406         u32                     num_entries;
1407         u32                     size;
1408         u32                     hook_entry[NF_INET_NUMHOOKS];
1409         u32                     underflow[NF_INET_NUMHOOKS];
1410         u32                     num_counters;
1411         compat_uptr_t           counters;       /* struct xt_counters * */
1412         struct compat_ipt_entry entries[0];
1413 };
1414
1415 static int
1416 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1417                           unsigned int *size, struct xt_counters *counters,
1418                           unsigned int i)
1419 {
1420         struct xt_entry_target *t;
1421         struct compat_ipt_entry __user *ce;
1422         u_int16_t target_offset, next_offset;
1423         compat_uint_t origsize;
1424         const struct xt_entry_match *ematch;
1425         int ret = 0;
1426
1427         origsize = *size;
1428         ce = (struct compat_ipt_entry __user *)*dstptr;
1429         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1430             copy_to_user(&ce->counters, &counters[i],
1431             sizeof(counters[i])) != 0)
1432                 return -EFAULT;
1433
1434         *dstptr += sizeof(struct compat_ipt_entry);
1435         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1436
1437         xt_ematch_foreach(ematch, e) {
1438                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1439                 if (ret != 0)
1440                         return ret;
1441         }
1442         target_offset = e->target_offset - (origsize - *size);
1443         t = ipt_get_target(e);
1444         ret = xt_compat_target_to_user(t, dstptr, size);
1445         if (ret)
1446                 return ret;
1447         next_offset = e->next_offset - (origsize - *size);
1448         if (put_user(target_offset, &ce->target_offset) != 0 ||
1449             put_user(next_offset, &ce->next_offset) != 0)
1450                 return -EFAULT;
1451         return 0;
1452 }
1453
1454 static int
1455 compat_find_calc_match(struct xt_entry_match *m,
1456                        const char *name,
1457                        const struct ipt_ip *ip,
1458                        int *size)
1459 {
1460         struct xt_match *match;
1461
1462         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1463                                       m->u.user.revision);
1464         if (IS_ERR(match)) {
1465                 duprintf("compat_check_calc_match: `%s' not found\n",
1466                          m->u.user.name);
1467                 return PTR_ERR(match);
1468         }
1469         m->u.kernel.match = match;
1470         *size += xt_compat_match_offset(match);
1471         return 0;
1472 }
1473
1474 static void compat_release_entry(struct compat_ipt_entry *e)
1475 {
1476         struct xt_entry_target *t;
1477         struct xt_entry_match *ematch;
1478
1479         /* Cleanup all matches */
1480         xt_ematch_foreach(ematch, e)
1481                 module_put(ematch->u.kernel.match->me);
1482         t = compat_ipt_get_target(e);
1483         module_put(t->u.kernel.target->me);
1484 }
1485
1486 static int
1487 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1488                                   struct xt_table_info *newinfo,
1489                                   unsigned int *size,
1490                                   const unsigned char *base,
1491                                   const unsigned char *limit,
1492                                   const unsigned int *hook_entries,
1493                                   const unsigned int *underflows,
1494                                   const char *name)
1495 {
1496         struct xt_entry_match *ematch;
1497         struct xt_entry_target *t;
1498         struct xt_target *target;
1499         unsigned int entry_offset;
1500         unsigned int j;
1501         int ret, off, h;
1502
1503         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1504         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1505             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1506                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1507                 return -EINVAL;
1508         }
1509
1510         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1511                              sizeof(struct compat_xt_entry_target)) {
1512                 duprintf("checking: element %p size %u\n",
1513                          e, e->next_offset);
1514                 return -EINVAL;
1515         }
1516
1517         /* For purposes of check_entry casting the compat entry is fine */
1518         ret = check_entry((struct ipt_entry *)e, name);
1519         if (ret)
1520                 return ret;
1521
1522         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1523         entry_offset = (void *)e - (void *)base;
1524         j = 0;
1525         xt_ematch_foreach(ematch, e) {
1526                 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1527                 if (ret != 0)
1528                         goto release_matches;
1529                 ++j;
1530         }
1531
1532         t = compat_ipt_get_target(e);
1533         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1534                                         t->u.user.revision);
1535         if (IS_ERR(target)) {
1536                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1537                          t->u.user.name);
1538                 ret = PTR_ERR(target);
1539                 goto release_matches;
1540         }
1541         t->u.kernel.target = target;
1542
1543         off += xt_compat_target_offset(target);
1544         *size += off;
1545         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1546         if (ret)
1547                 goto out;
1548
1549         /* Check hooks & underflows */
1550         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1551                 if ((unsigned char *)e - base == hook_entries[h])
1552                         newinfo->hook_entry[h] = hook_entries[h];
1553                 if ((unsigned char *)e - base == underflows[h])
1554                         newinfo->underflow[h] = underflows[h];
1555         }
1556
1557         /* Clear counters and comefrom */
1558         memset(&e->counters, 0, sizeof(e->counters));
1559         e->comefrom = 0;
1560         return 0;
1561
1562 out:
1563         module_put(t->u.kernel.target->me);
1564 release_matches:
1565         xt_ematch_foreach(ematch, e) {
1566                 if (j-- == 0)
1567                         break;
1568                 module_put(ematch->u.kernel.match->me);
1569         }
1570         return ret;
1571 }
1572
1573 static int
1574 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1575                             unsigned int *size, const char *name,
1576                             struct xt_table_info *newinfo, unsigned char *base)
1577 {
1578         struct xt_entry_target *t;
1579         struct xt_target *target;
1580         struct ipt_entry *de;
1581         unsigned int origsize;
1582         int ret, h;
1583         struct xt_entry_match *ematch;
1584
1585         ret = 0;
1586         origsize = *size;
1587         de = (struct ipt_entry *)*dstptr;
1588         memcpy(de, e, sizeof(struct ipt_entry));
1589         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1590
1591         *dstptr += sizeof(struct ipt_entry);
1592         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1593
1594         xt_ematch_foreach(ematch, e) {
1595                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1596                 if (ret != 0)
1597                         return ret;
1598         }
1599         de->target_offset = e->target_offset - (origsize - *size);
1600         t = compat_ipt_get_target(e);
1601         target = t->u.kernel.target;
1602         xt_compat_target_from_user(t, dstptr, size);
1603
1604         de->next_offset = e->next_offset - (origsize - *size);
1605         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1606                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1607                         newinfo->hook_entry[h] -= origsize - *size;
1608                 if ((unsigned char *)de - base < newinfo->underflow[h])
1609                         newinfo->underflow[h] -= origsize - *size;
1610         }
1611         return ret;
1612 }
1613
1614 static int
1615 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1616 {
1617         struct xt_entry_match *ematch;
1618         struct xt_mtchk_param mtpar;
1619         unsigned int j;
1620         int ret = 0;
1621
1622         e->counters.pcnt = xt_percpu_counter_alloc();
1623         if (IS_ERR_VALUE(e->counters.pcnt))
1624                 return -ENOMEM;
1625
1626         j = 0;
1627         mtpar.net       = net;
1628         mtpar.table     = name;
1629         mtpar.entryinfo = &e->ip;
1630         mtpar.hook_mask = e->comefrom;
1631         mtpar.family    = NFPROTO_IPV4;
1632         xt_ematch_foreach(ematch, e) {
1633                 ret = check_match(ematch, &mtpar);
1634                 if (ret != 0)
1635                         goto cleanup_matches;
1636                 ++j;
1637         }
1638
1639         ret = check_target(e, net, name);
1640         if (ret)
1641                 goto cleanup_matches;
1642         return 0;
1643
1644  cleanup_matches:
1645         xt_ematch_foreach(ematch, e) {
1646                 if (j-- == 0)
1647                         break;
1648                 cleanup_match(ematch, net);
1649         }
1650
1651         xt_percpu_counter_free(e->counters.pcnt);
1652
1653         return ret;
1654 }
1655
1656 static int
1657 translate_compat_table(struct net *net,
1658                        const char *name,
1659                        unsigned int valid_hooks,
1660                        struct xt_table_info **pinfo,
1661                        void **pentry0,
1662                        unsigned int total_size,
1663                        unsigned int number,
1664                        unsigned int *hook_entries,
1665                        unsigned int *underflows)
1666 {
1667         unsigned int i, j;
1668         struct xt_table_info *newinfo, *info;
1669         void *pos, *entry0, *entry1;
1670         struct compat_ipt_entry *iter0;
1671         struct ipt_entry *iter1;
1672         unsigned int size;
1673         int ret;
1674
1675         info = *pinfo;
1676         entry0 = *pentry0;
1677         size = total_size;
1678         info->number = number;
1679
1680         /* Init all hooks to impossible value. */
1681         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1682                 info->hook_entry[i] = 0xFFFFFFFF;
1683                 info->underflow[i] = 0xFFFFFFFF;
1684         }
1685
1686         duprintf("translate_compat_table: size %u\n", info->size);
1687         j = 0;
1688         xt_compat_lock(AF_INET);
1689         xt_compat_init_offsets(AF_INET, number);
1690         /* Walk through entries, checking offsets. */
1691         xt_entry_foreach(iter0, entry0, total_size) {
1692                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1693                                                         entry0,
1694                                                         entry0 + total_size,
1695                                                         hook_entries,
1696                                                         underflows,
1697                                                         name);
1698                 if (ret != 0)
1699                         goto out_unlock;
1700                 ++j;
1701         }
1702
1703         ret = -EINVAL;
1704         if (j != number) {
1705                 duprintf("translate_compat_table: %u not %u entries\n",
1706                          j, number);
1707                 goto out_unlock;
1708         }
1709
1710         /* Check hooks all assigned */
1711         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1712                 /* Only hooks which are valid */
1713                 if (!(valid_hooks & (1 << i)))
1714                         continue;
1715                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1716                         duprintf("Invalid hook entry %u %u\n",
1717                                  i, hook_entries[i]);
1718                         goto out_unlock;
1719                 }
1720                 if (info->underflow[i] == 0xFFFFFFFF) {
1721                         duprintf("Invalid underflow %u %u\n",
1722                                  i, underflows[i]);
1723                         goto out_unlock;
1724                 }
1725         }
1726
1727         ret = -ENOMEM;
1728         newinfo = xt_alloc_table_info(size);
1729         if (!newinfo)
1730                 goto out_unlock;
1731
1732         newinfo->number = number;
1733         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1734                 newinfo->hook_entry[i] = info->hook_entry[i];
1735                 newinfo->underflow[i] = info->underflow[i];
1736         }
1737         entry1 = newinfo->entries;
1738         pos = entry1;
1739         size = total_size;
1740         xt_entry_foreach(iter0, entry0, total_size) {
1741                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1742                                                   name, newinfo, entry1);
1743                 if (ret != 0)
1744                         break;
1745         }
1746         xt_compat_flush_offsets(AF_INET);
1747         xt_compat_unlock(AF_INET);
1748         if (ret)
1749                 goto free_newinfo;
1750
1751         ret = -ELOOP;
1752         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1753                 goto free_newinfo;
1754
1755         i = 0;
1756         xt_entry_foreach(iter1, entry1, newinfo->size) {
1757                 ret = compat_check_entry(iter1, net, name);
1758                 if (ret != 0)
1759                         break;
1760                 ++i;
1761         }
1762         if (ret) {
1763                 /*
1764                  * The first i matches need cleanup_entry (calls ->destroy)
1765                  * because they had called ->check already. The other j-i
1766                  * entries need only release.
1767                  */
1768                 int skip = i;
1769                 j -= i;
1770                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1771                         if (skip-- > 0)
1772                                 continue;
1773                         if (j-- == 0)
1774                                 break;
1775                         compat_release_entry(iter0);
1776                 }
1777                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1778                         if (i-- == 0)
1779                                 break;
1780                         cleanup_entry(iter1, net);
1781                 }
1782                 xt_free_table_info(newinfo);
1783                 return ret;
1784         }
1785
1786         *pinfo = newinfo;
1787         *pentry0 = entry1;
1788         xt_free_table_info(info);
1789         return 0;
1790
1791 free_newinfo:
1792         xt_free_table_info(newinfo);
1793 out:
1794         xt_entry_foreach(iter0, entry0, total_size) {
1795                 if (j-- == 0)
1796                         break;
1797                 compat_release_entry(iter0);
1798         }
1799         return ret;
1800 out_unlock:
1801         xt_compat_flush_offsets(AF_INET);
1802         xt_compat_unlock(AF_INET);
1803         goto out;
1804 }
1805
1806 static int
1807 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1808 {
1809         int ret;
1810         struct compat_ipt_replace tmp;
1811         struct xt_table_info *newinfo;
1812         void *loc_cpu_entry;
1813         struct ipt_entry *iter;
1814
1815         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1816                 return -EFAULT;
1817
1818         /* overflow check */
1819         if (tmp.size >= INT_MAX / num_possible_cpus())
1820                 return -ENOMEM;
1821         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1822                 return -ENOMEM;
1823         if (tmp.num_counters == 0)
1824                 return -EINVAL;
1825
1826         tmp.name[sizeof(tmp.name)-1] = 0;
1827
1828         newinfo = xt_alloc_table_info(tmp.size);
1829         if (!newinfo)
1830                 return -ENOMEM;
1831
1832         loc_cpu_entry = newinfo->entries;
1833         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1834                            tmp.size) != 0) {
1835                 ret = -EFAULT;
1836                 goto free_newinfo;
1837         }
1838
1839         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1840                                      &newinfo, &loc_cpu_entry, tmp.size,
1841                                      tmp.num_entries, tmp.hook_entry,
1842                                      tmp.underflow);
1843         if (ret != 0)
1844                 goto free_newinfo;
1845
1846         duprintf("compat_do_replace: Translated table\n");
1847
1848         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1849                            tmp.num_counters, compat_ptr(tmp.counters));
1850         if (ret)
1851                 goto free_newinfo_untrans;
1852         return 0;
1853
1854  free_newinfo_untrans:
1855         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1856                 cleanup_entry(iter, net);
1857  free_newinfo:
1858         xt_free_table_info(newinfo);
1859         return ret;
1860 }
1861
1862 static int
1863 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1864                       unsigned int len)
1865 {
1866         int ret;
1867
1868         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1869                 return -EPERM;
1870
1871         switch (cmd) {
1872         case IPT_SO_SET_REPLACE:
1873                 ret = compat_do_replace(sock_net(sk), user, len);
1874                 break;
1875
1876         case IPT_SO_SET_ADD_COUNTERS:
1877                 ret = do_add_counters(sock_net(sk), user, len, 1);
1878                 break;
1879
1880         default:
1881                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1882                 ret = -EINVAL;
1883         }
1884
1885         return ret;
1886 }
1887
1888 struct compat_ipt_get_entries {
1889         char name[XT_TABLE_MAXNAMELEN];
1890         compat_uint_t size;
1891         struct compat_ipt_entry entrytable[0];
1892 };
1893
1894 static int
1895 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1896                             void __user *userptr)
1897 {
1898         struct xt_counters *counters;
1899         const struct xt_table_info *private = table->private;
1900         void __user *pos;
1901         unsigned int size;
1902         int ret = 0;
1903         unsigned int i = 0;
1904         struct ipt_entry *iter;
1905
1906         counters = alloc_counters(table);
1907         if (IS_ERR(counters))
1908                 return PTR_ERR(counters);
1909
1910         pos = userptr;
1911         size = total_size;
1912         xt_entry_foreach(iter, private->entries, total_size) {
1913                 ret = compat_copy_entry_to_user(iter, &pos,
1914                                                 &size, counters, i++);
1915                 if (ret != 0)
1916                         break;
1917         }
1918
1919         vfree(counters);
1920         return ret;
1921 }
1922
1923 static int
1924 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1925                    int *len)
1926 {
1927         int ret;
1928         struct compat_ipt_get_entries get;
1929         struct xt_table *t;
1930
1931         if (*len < sizeof(get)) {
1932                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1933                 return -EINVAL;
1934         }
1935
1936         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1937                 return -EFAULT;
1938
1939         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1940                 duprintf("compat_get_entries: %u != %zu\n",
1941                          *len, sizeof(get) + get.size);
1942                 return -EINVAL;
1943         }
1944
1945         xt_compat_lock(AF_INET);
1946         t = xt_find_table_lock(net, AF_INET, get.name);
1947         if (!IS_ERR_OR_NULL(t)) {
1948                 const struct xt_table_info *private = t->private;
1949                 struct xt_table_info info;
1950                 duprintf("t->private->number = %u\n", private->number);
1951                 ret = compat_table_info(private, &info);
1952                 if (!ret && get.size == info.size) {
1953                         ret = compat_copy_entries_to_user(private->size,
1954                                                           t, uptr->entrytable);
1955                 } else if (!ret) {
1956                         duprintf("compat_get_entries: I've got %u not %u!\n",
1957                                  private->size, get.size);
1958                         ret = -EAGAIN;
1959                 }
1960                 xt_compat_flush_offsets(AF_INET);
1961                 module_put(t->me);
1962                 xt_table_unlock(t);
1963         } else
1964                 ret = t ? PTR_ERR(t) : -ENOENT;
1965
1966         xt_compat_unlock(AF_INET);
1967         return ret;
1968 }
1969
1970 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1971
1972 static int
1973 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1974 {
1975         int ret;
1976
1977         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1978                 return -EPERM;
1979
1980         switch (cmd) {
1981         case IPT_SO_GET_INFO:
1982                 ret = get_info(sock_net(sk), user, len, 1);
1983                 break;
1984         case IPT_SO_GET_ENTRIES:
1985                 ret = compat_get_entries(sock_net(sk), user, len);
1986                 break;
1987         default:
1988                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1989         }
1990         return ret;
1991 }
1992 #endif
1993
1994 static int
1995 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1996 {
1997         int ret;
1998
1999         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2000                 return -EPERM;
2001
2002         switch (cmd) {
2003         case IPT_SO_SET_REPLACE:
2004                 ret = do_replace(sock_net(sk), user, len);
2005                 break;
2006
2007         case IPT_SO_SET_ADD_COUNTERS:
2008                 ret = do_add_counters(sock_net(sk), user, len, 0);
2009                 break;
2010
2011         default:
2012                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2013                 ret = -EINVAL;
2014         }
2015
2016         return ret;
2017 }
2018
2019 static int
2020 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2021 {
2022         int ret;
2023
2024         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2025                 return -EPERM;
2026
2027         switch (cmd) {
2028         case IPT_SO_GET_INFO:
2029                 ret = get_info(sock_net(sk), user, len, 0);
2030                 break;
2031
2032         case IPT_SO_GET_ENTRIES:
2033                 ret = get_entries(sock_net(sk), user, len);
2034                 break;
2035
2036         case IPT_SO_GET_REVISION_MATCH:
2037         case IPT_SO_GET_REVISION_TARGET: {
2038                 struct xt_get_revision rev;
2039                 int target;
2040
2041                 if (*len != sizeof(rev)) {
2042                         ret = -EINVAL;
2043                         break;
2044                 }
2045                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2046                         ret = -EFAULT;
2047                         break;
2048                 }
2049                 rev.name[sizeof(rev.name)-1] = 0;
2050
2051                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2052                         target = 1;
2053                 else
2054                         target = 0;
2055
2056                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2057                                                          rev.revision,
2058                                                          target, &ret),
2059                                         "ipt_%s", rev.name);
2060                 break;
2061         }
2062
2063         default:
2064                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2065                 ret = -EINVAL;
2066         }
2067
2068         return ret;
2069 }
2070
2071 struct xt_table *ipt_register_table(struct net *net,
2072                                     const struct xt_table *table,
2073                                     const struct ipt_replace *repl)
2074 {
2075         int ret;
2076         struct xt_table_info *newinfo;
2077         struct xt_table_info bootstrap = {0};
2078         void *loc_cpu_entry;
2079         struct xt_table *new_table;
2080
2081         newinfo = xt_alloc_table_info(repl->size);
2082         if (!newinfo) {
2083                 ret = -ENOMEM;
2084                 goto out;
2085         }
2086
2087         loc_cpu_entry = newinfo->entries;
2088         memcpy(loc_cpu_entry, repl->entries, repl->size);
2089
2090         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2091         if (ret != 0)
2092                 goto out_free;
2093
2094         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2095         if (IS_ERR(new_table)) {
2096                 ret = PTR_ERR(new_table);
2097                 goto out_free;
2098         }
2099
2100         return new_table;
2101
2102 out_free:
2103         xt_free_table_info(newinfo);
2104 out:
2105         return ERR_PTR(ret);
2106 }
2107
2108 void ipt_unregister_table(struct net *net, struct xt_table *table)
2109 {
2110         struct xt_table_info *private;
2111         void *loc_cpu_entry;
2112         struct module *table_owner = table->me;
2113         struct ipt_entry *iter;
2114
2115         private = xt_unregister_table(table);
2116
2117         /* Decrease module usage counts and free resources */
2118         loc_cpu_entry = private->entries;
2119         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2120                 cleanup_entry(iter, net);
2121         if (private->number > private->initial_entries)
2122                 module_put(table_owner);
2123         xt_free_table_info(private);
2124 }
2125
2126 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2127 static inline bool
2128 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2129                      u_int8_t type, u_int8_t code,
2130                      bool invert)
2131 {
2132         return ((test_type == 0xFF) ||
2133                 (type == test_type && code >= min_code && code <= max_code))
2134                 ^ invert;
2135 }
2136
2137 static bool
2138 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2139 {
2140         const struct icmphdr *ic;
2141         struct icmphdr _icmph;
2142         const struct ipt_icmp *icmpinfo = par->matchinfo;
2143
2144         /* Must not be a fragment. */
2145         if (par->fragoff != 0)
2146                 return false;
2147
2148         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2149         if (ic == NULL) {
2150                 /* We've been asked to examine this packet, and we
2151                  * can't.  Hence, no choice but to drop.
2152                  */
2153                 duprintf("Dropping evil ICMP tinygram.\n");
2154                 par->hotdrop = true;
2155                 return false;
2156         }
2157
2158         return icmp_type_code_match(icmpinfo->type,
2159                                     icmpinfo->code[0],
2160                                     icmpinfo->code[1],
2161                                     ic->type, ic->code,
2162                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2163 }
2164
2165 static int icmp_checkentry(const struct xt_mtchk_param *par)
2166 {
2167         const struct ipt_icmp *icmpinfo = par->matchinfo;
2168
2169         /* Must specify no unknown invflags */
2170         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2171 }
2172
2173 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2174         {
2175                 .name             = XT_STANDARD_TARGET,
2176                 .targetsize       = sizeof(int),
2177                 .family           = NFPROTO_IPV4,
2178 #ifdef CONFIG_COMPAT
2179                 .compatsize       = sizeof(compat_int_t),
2180                 .compat_from_user = compat_standard_from_user,
2181                 .compat_to_user   = compat_standard_to_user,
2182 #endif
2183         },
2184         {
2185                 .name             = XT_ERROR_TARGET,
2186                 .target           = ipt_error,
2187                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2188                 .family           = NFPROTO_IPV4,
2189         },
2190 };
2191
2192 static struct nf_sockopt_ops ipt_sockopts = {
2193         .pf             = PF_INET,
2194         .set_optmin     = IPT_BASE_CTL,
2195         .set_optmax     = IPT_SO_SET_MAX+1,
2196         .set            = do_ipt_set_ctl,
2197 #ifdef CONFIG_COMPAT
2198         .compat_set     = compat_do_ipt_set_ctl,
2199 #endif
2200         .get_optmin     = IPT_BASE_CTL,
2201         .get_optmax     = IPT_SO_GET_MAX+1,
2202         .get            = do_ipt_get_ctl,
2203 #ifdef CONFIG_COMPAT
2204         .compat_get     = compat_do_ipt_get_ctl,
2205 #endif
2206         .owner          = THIS_MODULE,
2207 };
2208
2209 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2210         {
2211                 .name       = "icmp",
2212                 .match      = icmp_match,
2213                 .matchsize  = sizeof(struct ipt_icmp),
2214                 .checkentry = icmp_checkentry,
2215                 .proto      = IPPROTO_ICMP,
2216                 .family     = NFPROTO_IPV4,
2217         },
2218 };
2219
2220 static int __net_init ip_tables_net_init(struct net *net)
2221 {
2222         return xt_proto_init(net, NFPROTO_IPV4);
2223 }
2224
2225 static void __net_exit ip_tables_net_exit(struct net *net)
2226 {
2227         xt_proto_fini(net, NFPROTO_IPV4);
2228 }
2229
2230 static struct pernet_operations ip_tables_net_ops = {
2231         .init = ip_tables_net_init,
2232         .exit = ip_tables_net_exit,
2233 };
2234
2235 static int __init ip_tables_init(void)
2236 {
2237         int ret;
2238
2239         ret = register_pernet_subsys(&ip_tables_net_ops);
2240         if (ret < 0)
2241                 goto err1;
2242
2243         /* No one else will be downing sem now, so we won't sleep */
2244         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2245         if (ret < 0)
2246                 goto err2;
2247         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2248         if (ret < 0)
2249                 goto err4;
2250
2251         /* Register setsockopt */
2252         ret = nf_register_sockopt(&ipt_sockopts);
2253         if (ret < 0)
2254                 goto err5;
2255
2256         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2257         return 0;
2258
2259 err5:
2260         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2261 err4:
2262         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2263 err2:
2264         unregister_pernet_subsys(&ip_tables_net_ops);
2265 err1:
2266         return ret;
2267 }
2268
2269 static void __exit ip_tables_fini(void)
2270 {
2271         nf_unregister_sockopt(&ipt_sockopts);
2272
2273         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2274         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2275         unregister_pernet_subsys(&ip_tables_net_ops);
2276 }
2277
2278 EXPORT_SYMBOL(ipt_register_table);
2279 EXPORT_SYMBOL(ipt_unregister_table);
2280 EXPORT_SYMBOL(ipt_do_table);
2281 module_init(ip_tables_init);
2282 module_exit(ip_tables_fini);