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