net/ipv4/netfilter/ip_tables.c

   1 /*
   2  * Packet matching code.
   3  *
   4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   6  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13 #include <linux/cache.h>
  14 #include <linux/capability.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/kmod.h>
  17 #include <linux/vmalloc.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/module.h>
  20 #include <linux/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/compat.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/mutex.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/err.h>
  27 #include <linux/cpumask.h>
  28
  29 #include <linux/netfilter/x_tables.h>
  30 #include <linux/netfilter_ipv4/ip_tables.h>
  31 #include <net/netfilter/nf_log.h>
  32 #include "../../netfilter/xt_repldata.h"
  33
  34 MODULE_LICENSE("GPL");
  35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36 MODULE_DESCRIPTION("IPv4 packet filter");
  37
  38 /*#define DEBUG_IP_FIREWALL*/
  39 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  40 /*#define DEBUG_IP_FIREWALL_USER*/
  41
  42 #ifdef DEBUG_IP_FIREWALL
  43 #define dprintf(format, args...) pr_info(format , ## args)
  44 #else
  45 #define dprintf(format, args...)
  46 #endif
  47
  48 #ifdef DEBUG_IP_FIREWALL_USER
  49 #define duprintf(format, args...) pr_info(format , ## args)
  50 #else
  51 #define duprintf(format, args...)
  52 #endif
  53
  54 #ifdef CONFIG_NETFILTER_DEBUG
  55 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  56 #else
  57 #define IP_NF_ASSERT(x)
  58 #endif
  59
  60 #if 0
  61 /* All the better to debug you with... */
  62 #define static
  63 #define inline
  64 #endif
  65
  66 void *ipt_alloc_initial_table(const struct xt_table *info)
  67 {
  68         return xt_alloc_initial_table(ipt, IPT);
  69 }
  70 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  71
  72 /* Returns whether matches rule or not. */
  73 /* Performance critical - called for every packet */
  74 static inline bool
  75 ip_packet_match(const struct iphdr *ip,
  76                 const char *indev,
  77                 const char *outdev,
  78                 const struct ipt_ip *ipinfo,
  79                 int isfrag)
  80 {
  81         unsigned long ret;
  82
  83 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  84
  85         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  86                   IPT_INV_SRCIP) ||
  87             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  88                   IPT_INV_DSTIP)) {
  89                 dprintf("Source or dest mismatch.\n");
  90
  91                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  92                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  93                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  94                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  95                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  96                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  97                 return false;
  98         }
  99
 100         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 101
 102         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 103                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 104                         indev, ipinfo->iniface,
 105                         ipinfo->invflags & IPT_INV_VIA_IN ? " (INV)" : "");
 106                 return false;
 107         }
 108
 109         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 110
 111         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 112                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 113                         outdev, ipinfo->outiface,
 114                         ipinfo->invflags & IPT_INV_VIA_OUT ? " (INV)" : "");
 115                 return false;
 116         }
 117
 118         /* Check specific protocol */
 119         if (ipinfo->proto &&
 120             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 121                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 122                         ip->protocol, ipinfo->proto,
 123                         ipinfo->invflags & IPT_INV_PROTO ? " (INV)" : "");
 124                 return false;
 125         }
 126
 127         /* If we have a fragment rule but the packet is not a fragment
 128          * then we return zero */
 129         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 130                 dprintf("Fragment rule but not fragment.%s\n",
 131                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 132                 return false;
 133         }
 134
 135         return true;
 136 }
 137
 138 static bool
 139 ip_checkentry(const struct ipt_ip *ip)
 140 {
 141         if (ip->flags & ~IPT_F_MASK) {
 142                 duprintf("Unknown flag bits set: %08X\n",
 143                          ip->flags & ~IPT_F_MASK);
 144                 return false;
 145         }
 146         if (ip->invflags & ~IPT_INV_MASK) {
 147                 duprintf("Unknown invflag bits set: %08X\n",
 148                          ip->invflags & ~IPT_INV_MASK);
 149                 return false;
 150         }
 151         return true;
 152 }
 153
 154 static unsigned int
 155 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 156 {
 157         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 158
 159         return NF_DROP;
 160 }
 161
 162 /* Performance critical */
 163 static inline struct ipt_entry *
 164 get_entry(const void *base, unsigned int offset)
 165 {
 166         return (struct ipt_entry *)(base + offset);
 167 }
 168
 169 /* All zeroes == unconditional rule. */
 170 /* Mildly perf critical (only if packet tracing is on) */
 171 static inline bool unconditional(const struct ipt_entry *e)
 172 {
 173         static const struct ipt_ip uncond;
 174
 175         return e->target_offset == sizeof(struct ipt_entry) &&
 176                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
 177 #undef FWINV
 178 }
 179
 180 /* for const-correctness */
 181 static inline const struct xt_entry_target *
 182 ipt_get_target_c(const struct ipt_entry *e)
 183 {
 184         return ipt_get_target((struct ipt_entry *)e);
 185 }
 186
 187 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 188 static const char *const hooknames[] = {
 189         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 190         [NF_INET_LOCAL_IN]              = "INPUT",
 191         [NF_INET_FORWARD]               = "FORWARD",
 192         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 193         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 194 };
 195
 196 enum nf_ip_trace_comments {
 197         NF_IP_TRACE_COMMENT_RULE,
 198         NF_IP_TRACE_COMMENT_RETURN,
 199         NF_IP_TRACE_COMMENT_POLICY,
 200 };
 201
 202 static const char *const comments[] = {
 203         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 204         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 205         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 206 };
 207
 208 static struct nf_loginfo trace_loginfo = {
 209         .type = NF_LOG_TYPE_LOG,
 210         .u = {
 211                 .log = {
 212                         .level = 4,
 213                         .logflags = NF_LOG_MASK,
 214                 },
 215         },
 216 };
 217
 218 /* Mildly perf critical (only if packet tracing is on) */
 219 static inline int
 220 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 221                       const char *hookname, const char **chainname,
 222                       const char **comment, unsigned int *rulenum)
 223 {
 224         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 225
 226         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 227                 /* Head of user chain: ERROR target with chainname */
 228                 *chainname = t->target.data;
 229                 (*rulenum) = 0;
 230         } else if (s == e) {
 231                 (*rulenum)++;
 232
 233                 if (unconditional(s) &&
 234                     strcmp(t->target.u.kernel.target->name,
 235                            XT_STANDARD_TARGET) == 0 &&
 236                    t->verdict < 0) {
 237                         /* Tail of chains: STANDARD target (return/policy) */
 238                         *comment = *chainname == hookname
 239                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 240                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 241                 }
 242                 return 1;
 243         } else
 244                 (*rulenum)++;
 245
 246         return 0;
 247 }
 248
 249 static void trace_packet(struct net *net,
 250                          const struct sk_buff *skb,
 251                          unsigned int hook,
 252                          const struct net_device *in,
 253                          const struct net_device *out,
 254                          const char *tablename,
 255                          const struct xt_table_info *private,
 256                          const struct ipt_entry *e)
 257 {
 258         const struct ipt_entry *root;
 259         const char *hookname, *chainname, *comment;
 260         const struct ipt_entry *iter;
 261         unsigned int rulenum = 0;
 262
 263         root = get_entry(private->entries, private->hook_entry[hook]);
 264
 265         hookname = chainname = hooknames[hook];
 266         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 267
 268         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 269                 if (get_chainname_rulenum(iter, e, hookname,
 270                     &chainname, &comment, &rulenum) != 0)
 271                         break;
 272
 273         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 274                      "TRACE: %s:%s:%s:%u ",
 275                      tablename, chainname, comment, rulenum);
 276 }
 277 #endif
 278
 279 static inline
 280 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 281 {
 282         return (void *)entry + entry->next_offset;
 283 }
 284
 285 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 286 unsigned int
 287 ipt_do_table(struct sk_buff *skb,
 288              const struct nf_hook_state *state,
 289              struct xt_table *table)
 290 {
 291         unsigned int hook = state->hook;
 292         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 293         const struct iphdr *ip;
 294         /* Initializing verdict to NF_DROP keeps gcc happy. */
 295         unsigned int verdict = NF_DROP;
 296         const char *indev, *outdev;
 297         const void *table_base;
 298         struct ipt_entry *e, **jumpstack;
 299         unsigned int stackidx, cpu;
 300         const struct xt_table_info *private;
 301         struct xt_action_param acpar;
 302         unsigned int addend;
 303
 304         /* Initialization */
 305         stackidx = 0;
 306         ip = ip_hdr(skb);
 307         indev = state->in ? state->in->name : nulldevname;
 308         outdev = state->out ? state->out->name : nulldevname;
 309         /* We handle fragments by dealing with the first fragment as
 310          * if it was a normal packet.  All other fragments are treated
 311          * normally, except that they will NEVER match rules that ask
 312          * things we don't know, ie. tcp syn flag or ports).  If the
 313          * rule is also a fragment-specific rule, non-fragments won't
 314          * match it. */
 315         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 316         acpar.thoff   = ip_hdrlen(skb);
 317         acpar.hotdrop = false;
 318         acpar.net     = state->net;
 319         acpar.in      = state->in;
 320         acpar.out     = state->out;
 321         acpar.family  = NFPROTO_IPV4;
 322         acpar.hooknum = hook;
 323
 324         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 325         local_bh_disable();
 326         addend = xt_write_recseq_begin();
 327         private = table->private;
 328         cpu        = smp_processor_id();
 329         /*
 330          * Ensure we load private-> members after we've fetched the base
 331          * pointer.
 332          */
 333         smp_read_barrier_depends();
 334         table_base = private->entries;
 335         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 336
 337         /* Switch to alternate jumpstack if we're being invoked via TEE.
 338          * TEE issues XT_CONTINUE verdict on original skb so we must not
 339          * clobber the jumpstack.
 340          *
 341          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 342          * but it is no problem since absolute verdict is issued by these.
 343          */
 344         if (static_key_false(&xt_tee_enabled))
 345                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 346
 347         e = get_entry(table_base, private->hook_entry[hook]);
 348
 349         pr_debug("Entering %s(hook %u), UF %p\n",
 350                  table->name, hook,
 351                  get_entry(table_base, private->underflow[hook]));
 352
 353         do {
 354                 const struct xt_entry_target *t;
 355                 const struct xt_entry_match *ematch;
 356                 struct xt_counters *counter;
 357
 358                 IP_NF_ASSERT(e);
 359                 if (!ip_packet_match(ip, indev, outdev,
 360                     &e->ip, acpar.fragoff)) {
 361  no_match:
 362                         e = ipt_next_entry(e);
 363                         continue;
 364                 }
 365
 366                 xt_ematch_foreach(ematch, e) {
 367                         acpar.match     = ematch->u.kernel.match;
 368                         acpar.matchinfo = ematch->data;
 369                         if (!acpar.match->match(skb, &acpar))
 370                                 goto no_match;
 371                 }
 372
 373                 counter = xt_get_this_cpu_counter(&e->counters);
 374                 ADD_COUNTER(*counter, skb->len, 1);
 375
 376                 t = ipt_get_target(e);
 377                 IP_NF_ASSERT(t->u.kernel.target);
 378
 379 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 380                 /* The packet is traced: log it */
 381                 if (unlikely(skb->nf_trace))
 382                         trace_packet(state->net, skb, hook, state->in,
 383                                      state->out, table->name, private, e);
 384 #endif
 385                 /* Standard target? */
 386                 if (!t->u.kernel.target->target) {
 387                         int v;
 388
 389                         v = ((struct xt_standard_target *)t)->verdict;
 390                         if (v < 0) {
 391                                 /* Pop from stack? */
 392                                 if (v != XT_RETURN) {
 393                                         verdict = (unsigned int)(-v) - 1;
 394                                         break;
 395                                 }
 396                                 if (stackidx == 0) {
 397                                         e = get_entry(table_base,
 398                                             private->underflow[hook]);
 399                                         pr_debug("Underflow (this is normal) "
 400                                                  "to %p\n", e);
 401                                 } else {
 402                                         e = jumpstack[--stackidx];
 403                                         pr_debug("Pulled %p out from pos %u\n",
 404                                                  e, stackidx);
 405                                         e = ipt_next_entry(e);
 406                                 }
 407                                 continue;
 408                         }
 409                         if (table_base + v != ipt_next_entry(e) &&
 410                             !(e->ip.flags & IPT_F_GOTO)) {
 411                                 jumpstack[stackidx++] = e;
 412                                 pr_debug("Pushed %p into pos %u\n",
 413                                          e, stackidx - 1);
 414                         }
 415
 416                         e = get_entry(table_base, v);
 417                         continue;
 418                 }
 419
 420                 acpar.target   = t->u.kernel.target;
 421                 acpar.targinfo = t->data;
 422
 423                 verdict = t->u.kernel.target->target(skb, &acpar);
 424                 /* Target might have changed stuff. */
 425                 ip = ip_hdr(skb);
 426                 if (verdict == XT_CONTINUE)
 427                         e = ipt_next_entry(e);
 428                 else
 429                         /* Verdict */
 430                         break;
 431         } while (!acpar.hotdrop);
 432         pr_debug("Exiting %s; sp at %u\n", __func__, stackidx);
 433
 434         xt_write_recseq_end(addend);
 435         local_bh_enable();
 436
 437 #ifdef DEBUG_ALLOW_ALL
 438         return NF_ACCEPT;
 439 #else
 440         if (acpar.hotdrop)
 441                 return NF_DROP;
 442         else return verdict;
 443 #endif
 444 }
 445
 446 static bool find_jump_target(const struct xt_table_info *t,
 447                              const struct ipt_entry *target)
 448 {
 449         struct ipt_entry *iter;
 450
 451         xt_entry_foreach(iter, t->entries, t->size) {
 452                  if (iter == target)
 453                         return true;
 454         }
 455         return false;
 456 }
 457
 458 /* Figures out from what hook each rule can be called: returns 0 if
 459    there are loops.  Puts hook bitmask in comefrom. */
 460 static int
 461 mark_source_chains(const struct xt_table_info *newinfo,
 462                    unsigned int valid_hooks, void *entry0)
 463 {
 464         unsigned int hook;
 465
 466         /* No recursion; use packet counter to save back ptrs (reset
 467            to 0 as we leave), and comefrom to save source hook bitmask */
 468         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 469                 unsigned int pos = newinfo->hook_entry[hook];
 470                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 471
 472                 if (!(valid_hooks & (1 << hook)))
 473                         continue;
 474
 475                 /* Set initial back pointer. */
 476                 e->counters.pcnt = pos;
 477
 478                 for (;;) {
 479                         const struct xt_standard_target *t
 480                                 = (void *)ipt_get_target_c(e);
 481                         int visited = e->comefrom & (1 << hook);
 482
 483                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 484                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 485                                        hook, pos, e->comefrom);
 486                                 return 0;
 487                         }
 488                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 489
 490                         /* Unconditional return/END. */
 491                         if ((unconditional(e) &&
 492                              (strcmp(t->target.u.user.name,
 493                                      XT_STANDARD_TARGET) == 0) &&
 494                              t->verdict < 0) || visited) {
 495                                 unsigned int oldpos, size;
 496
 497                                 if ((strcmp(t->target.u.user.name,
 498                                             XT_STANDARD_TARGET) == 0) &&
 499                                     t->verdict < -NF_MAX_VERDICT - 1) {
 500                                         duprintf("mark_source_chains: bad "
 501                                                 "negative verdict (%i)\n",
 502                                                                 t->verdict);
 503                                         return 0;
 504                                 }
 505
 506                                 /* Return: backtrack through the last
 507                                    big jump. */
 508                                 do {
 509                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 510 #ifdef DEBUG_IP_FIREWALL_USER
 511                                         if (e->comefrom
 512                                             & (1 << NF_INET_NUMHOOKS)) {
 513                                                 duprintf("Back unset "
 514                                                          "on hook %u "
 515                                                          "rule %u\n",
 516                                                          hook, pos);
 517                                         }
 518 #endif
 519                                         oldpos = pos;
 520                                         pos = e->counters.pcnt;
 521                                         e->counters.pcnt = 0;
 522
 523                                         /* We're at the start. */
 524                                         if (pos == oldpos)
 525                                                 goto next;
 526
 527                                         e = (struct ipt_entry *)
 528                                                 (entry0 + pos);
 529                                 } while (oldpos == pos + e->next_offset);
 530
 531                                 /* Move along one */
 532                                 size = e->next_offset;
 533                                 e = (struct ipt_entry *)
 534                                         (entry0 + pos + size);
 535                                 if (pos + size >= newinfo->size)
 536                                         return 0;
 537                                 e->counters.pcnt = pos;
 538                                 pos += size;
 539                         } else {
 540                                 int newpos = t->verdict;
 541
 542                                 if (strcmp(t->target.u.user.name,
 543                                            XT_STANDARD_TARGET) == 0 &&
 544                                     newpos >= 0) {
 545                                         if (newpos > newinfo->size -
 546                                                 sizeof(struct ipt_entry)) {
 547                                                 duprintf("mark_source_chains: "
 548                                                         "bad verdict (%i)\n",
 549                                                                 newpos);
 550                                                 return 0;
 551                                         }
 552                                         /* This a jump; chase it. */
 553                                         duprintf("Jump rule %u -> %u\n",
 554                                                  pos, newpos);
 555                                         e = (struct ipt_entry *)
 556                                                 (entry0 + newpos);
 557                                         if (!find_jump_target(newinfo, e))
 558                                                 return 0;
 559                                 } else {
 560                                         /* ... this is a fallthru */
 561                                         newpos = pos + e->next_offset;
 562                                         if (newpos >= newinfo->size)
 563                                                 return 0;
 564                                 }
 565                                 e = (struct ipt_entry *)
 566                                         (entry0 + newpos);
 567                                 e->counters.pcnt = pos;
 568                                 pos = newpos;
 569                         }
 570                 }
 571 next:
 572                 duprintf("Finished chain %u\n", hook);
 573         }
 574         return 1;
 575 }
 576
 577 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 578 {
 579         struct xt_mtdtor_param par;
 580
 581         par.net       = net;
 582         par.match     = m->u.kernel.match;
 583         par.matchinfo = m->data;
 584         par.family    = NFPROTO_IPV4;
 585         if (par.match->destroy != NULL)
 586                 par.match->destroy(&par);
 587         module_put(par.match->me);
 588 }
 589
 590 static int
 591 check_entry(const struct ipt_entry *e)
 592 {
 593         if (!ip_checkentry(&e->ip))
 594                 return -EINVAL;
 595
 596         return xt_check_entry_offsets(e, e->target_offset, e->next_offset);
 597 }
 598
 599 static int
 600 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 601 {
 602         const struct ipt_ip *ip = par->entryinfo;
 603         int ret;
 604
 605         par->match     = m->u.kernel.match;
 606         par->matchinfo = m->data;
 607
 608         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 609               ip->proto, ip->invflags & IPT_INV_PROTO);
 610         if (ret < 0) {
 611                 duprintf("check failed for `%s'.\n", par->match->name);
 612                 return ret;
 613         }
 614         return 0;
 615 }
 616
 617 static int
 618 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 619 {
 620         struct xt_match *match;
 621         int ret;
 622
 623         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 624                                       m->u.user.revision);
 625         if (IS_ERR(match)) {
 626                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 627                 return PTR_ERR(match);
 628         }
 629         m->u.kernel.match = match;
 630
 631         ret = check_match(m, par);
 632         if (ret)
 633                 goto err;
 634
 635         return 0;
 636 err:
 637         module_put(m->u.kernel.match->me);
 638         return ret;
 639 }
 640
 641 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 642 {
 643         struct xt_entry_target *t = ipt_get_target(e);
 644         struct xt_tgchk_param par = {
 645                 .net       = net,
 646                 .table     = name,
 647                 .entryinfo = e,
 648                 .target    = t->u.kernel.target,
 649                 .targinfo  = t->data,
 650                 .hook_mask = e->comefrom,
 651                 .family    = NFPROTO_IPV4,
 652         };
 653         int ret;
 654
 655         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 656               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 657         if (ret < 0) {
 658                 duprintf("check failed for `%s'.\n",
 659                          t->u.kernel.target->name);
 660                 return ret;
 661         }
 662         return 0;
 663 }
 664
 665 static int
 666 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 667                  unsigned int size)
 668 {
 669         struct xt_entry_target *t;
 670         struct xt_target *target;
 671         int ret;
 672         unsigned int j;
 673         struct xt_mtchk_param mtpar;
 674         struct xt_entry_match *ematch;
 675
 676         e->counters.pcnt = xt_percpu_counter_alloc();
 677         if (IS_ERR_VALUE(e->counters.pcnt))
 678                 return -ENOMEM;
 679
 680         j = 0;
 681         mtpar.net       = net;
 682         mtpar.table     = name;
 683         mtpar.entryinfo = &e->ip;
 684         mtpar.hook_mask = e->comefrom;
 685         mtpar.family    = NFPROTO_IPV4;
 686         xt_ematch_foreach(ematch, e) {
 687                 ret = find_check_match(ematch, &mtpar);
 688                 if (ret != 0)
 689                         goto cleanup_matches;
 690                 ++j;
 691         }
 692
 693         t = ipt_get_target(e);
 694         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 695                                         t->u.user.revision);
 696         if (IS_ERR(target)) {
 697                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 698                 ret = PTR_ERR(target);
 699                 goto cleanup_matches;
 700         }
 701         t->u.kernel.target = target;
 702
 703         ret = check_target(e, net, name);
 704         if (ret)
 705                 goto err;
 706
 707         return 0;
 708  err:
 709         module_put(t->u.kernel.target->me);
 710  cleanup_matches:
 711         xt_ematch_foreach(ematch, e) {
 712                 if (j-- == 0)
 713                         break;
 714                 cleanup_match(ematch, net);
 715         }
 716
 717         xt_percpu_counter_free(e->counters.pcnt);
 718
 719         return ret;
 720 }
 721
 722 static bool check_underflow(const struct ipt_entry *e)
 723 {
 724         const struct xt_entry_target *t;
 725         unsigned int verdict;
 726
 727         if (!unconditional(e))
 728                 return false;
 729         t = ipt_get_target_c(e);
 730         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 731                 return false;
 732         verdict = ((struct xt_standard_target *)t)->verdict;
 733         verdict = -verdict - 1;
 734         return verdict == NF_DROP || verdict == NF_ACCEPT;
 735 }
 736
 737 static int
 738 check_entry_size_and_hooks(struct ipt_entry *e,
 739                            struct xt_table_info *newinfo,
 740                            const unsigned char *base,
 741                            const unsigned char *limit,
 742                            const unsigned int *hook_entries,
 743                            const unsigned int *underflows,
 744                            unsigned int valid_hooks)
 745 {
 746         unsigned int h;
 747         int err;
 748
 749         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 750             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 751             (unsigned char *)e + e->next_offset > limit) {
 752                 duprintf("Bad offset %p\n", e);
 753                 return -EINVAL;
 754         }
 755
 756         if (e->next_offset
 757             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 758                 duprintf("checking: element %p size %u\n",
 759                          e, e->next_offset);
 760                 return -EINVAL;
 761         }
 762
 763         err = check_entry(e);
 764         if (err)
 765                 return err;
 766
 767         /* Check hooks & underflows */
 768         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 769                 if (!(valid_hooks & (1 << h)))
 770                         continue;
 771                 if ((unsigned char *)e - base == hook_entries[h])
 772                         newinfo->hook_entry[h] = hook_entries[h];
 773                 if ((unsigned char *)e - base == underflows[h]) {
 774                         if (!check_underflow(e)) {
 775                                 pr_debug("Underflows must be unconditional and "
 776                                          "use the STANDARD target with "
 777                                          "ACCEPT/DROP\n");
 778                                 return -EINVAL;
 779                         }
 780                         newinfo->underflow[h] = underflows[h];
 781                 }
 782         }
 783
 784         /* Clear counters and comefrom */
 785         e->counters = ((struct xt_counters) { 0, 0 });
 786         e->comefrom = 0;
 787         return 0;
 788 }
 789
 790 static void
 791 cleanup_entry(struct ipt_entry *e, struct net *net)
 792 {
 793         struct xt_tgdtor_param par;
 794         struct xt_entry_target *t;
 795         struct xt_entry_match *ematch;
 796
 797         /* Cleanup all matches */
 798         xt_ematch_foreach(ematch, e)
 799                 cleanup_match(ematch, net);
 800         t = ipt_get_target(e);
 801
 802         par.net      = net;
 803         par.target   = t->u.kernel.target;
 804         par.targinfo = t->data;
 805         par.family   = NFPROTO_IPV4;
 806         if (par.target->destroy != NULL)
 807                 par.target->destroy(&par);
 808         module_put(par.target->me);
 809         xt_percpu_counter_free(e->counters.pcnt);
 810 }
 811
 812 /* Checks and translates the user-supplied table segment (held in
 813    newinfo) */
 814 static int
 815 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 816                 const struct ipt_replace *repl)
 817 {
 818         struct ipt_entry *iter;
 819         unsigned int i;
 820         int ret = 0;
 821
 822         newinfo->size = repl->size;
 823         newinfo->number = repl->num_entries;
 824
 825         /* Init all hooks to impossible value. */
 826         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 827                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 828                 newinfo->underflow[i] = 0xFFFFFFFF;
 829         }
 830
 831         duprintf("translate_table: size %u\n", newinfo->size);
 832         i = 0;
 833         /* Walk through entries, checking offsets. */
 834         xt_entry_foreach(iter, entry0, newinfo->size) {
 835                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 836                                                  entry0 + repl->size,
 837                                                  repl->hook_entry,
 838                                                  repl->underflow,
 839                                                  repl->valid_hooks);
 840                 if (ret != 0)
 841                         return ret;
 842                 ++i;
 843                 if (strcmp(ipt_get_target(iter)->u.user.name,
 844                     XT_ERROR_TARGET) == 0)
 845                         ++newinfo->stacksize;
 846         }
 847
 848         if (i != repl->num_entries) {
 849                 duprintf("translate_table: %u not %u entries\n",
 850                          i, repl->num_entries);
 851                 return -EINVAL;
 852         }
 853
 854         /* Check hooks all assigned */
 855         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 856                 /* Only hooks which are valid */
 857                 if (!(repl->valid_hooks & (1 << i)))
 858                         continue;
 859                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 860                         duprintf("Invalid hook entry %u %u\n",
 861                                  i, repl->hook_entry[i]);
 862                         return -EINVAL;
 863                 }
 864                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 865                         duprintf("Invalid underflow %u %u\n",
 866                                  i, repl->underflow[i]);
 867                         return -EINVAL;
 868                 }
 869         }
 870
 871         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 872                 return -ELOOP;
 873
 874         /* Finally, each sanity check must pass */
 875         i = 0;
 876         xt_entry_foreach(iter, entry0, newinfo->size) {
 877                 ret = find_check_entry(iter, net, repl->name, repl->size);
 878                 if (ret != 0)
 879                         break;
 880                 ++i;
 881         }
 882
 883         if (ret != 0) {
 884                 xt_entry_foreach(iter, entry0, newinfo->size) {
 885                         if (i-- == 0)
 886                                 break;
 887                         cleanup_entry(iter, net);
 888                 }
 889                 return ret;
 890         }
 891
 892         return ret;
 893 }
 894
 895 static void
 896 get_counters(const struct xt_table_info *t,
 897              struct xt_counters counters[])
 898 {
 899         struct ipt_entry *iter;
 900         unsigned int cpu;
 901         unsigned int i;
 902
 903         for_each_possible_cpu(cpu) {
 904                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 905
 906                 i = 0;
 907                 xt_entry_foreach(iter, t->entries, t->size) {
 908                         struct xt_counters *tmp;
 909                         u64 bcnt, pcnt;
 910                         unsigned int start;
 911
 912                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 913                         do {
 914                                 start = read_seqcount_begin(s);
 915                                 bcnt = tmp->bcnt;
 916                                 pcnt = tmp->pcnt;
 917                         } while (read_seqcount_retry(s, start));
 918
 919                         ADD_COUNTER(counters[i], bcnt, pcnt);
 920                         ++i; /* macro does multi eval of i */
 921                 }
 922         }
 923 }
 924
 925 static struct xt_counters *alloc_counters(const struct xt_table *table)
 926 {
 927         unsigned int countersize;
 928         struct xt_counters *counters;
 929         const struct xt_table_info *private = table->private;
 930
 931         /* We need atomic snapshot of counters: rest doesn't change
 932            (other than comefrom, which userspace doesn't care
 933            about). */
 934         countersize = sizeof(struct xt_counters) * private->number;
 935         counters = vzalloc(countersize);
 936
 937         if (counters == NULL)
 938                 return ERR_PTR(-ENOMEM);
 939
 940         get_counters(private, counters);
 941
 942         return counters;
 943 }
 944
 945 static int
 946 copy_entries_to_user(unsigned int total_size,
 947                      const struct xt_table *table,
 948                      void __user *userptr)
 949 {
 950         unsigned int off, num;
 951         const struct ipt_entry *e;
 952         struct xt_counters *counters;
 953         const struct xt_table_info *private = table->private;
 954         int ret = 0;
 955         const void *loc_cpu_entry;
 956
 957         counters = alloc_counters(table);
 958         if (IS_ERR(counters))
 959                 return PTR_ERR(counters);
 960
 961         loc_cpu_entry = private->entries;
 962         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 963                 ret = -EFAULT;
 964                 goto free_counters;
 965         }
 966
 967         /* FIXME: use iterator macros --RR */
 968         /* ... then go back and fix counters and names */
 969         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 970                 unsigned int i;
 971                 const struct xt_entry_match *m;
 972                 const struct xt_entry_target *t;
 973
 974                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 975                 if (copy_to_user(userptr + off
 976                                  + offsetof(struct ipt_entry, counters),
 977                                  &counters[num],
 978                                  sizeof(counters[num])) != 0) {
 979                         ret = -EFAULT;
 980                         goto free_counters;
 981                 }
 982
 983                 for (i = sizeof(struct ipt_entry);
 984                      i < e->target_offset;
 985                      i += m->u.match_size) {
 986                         m = (void *)e + i;
 987
 988                         if (copy_to_user(userptr + off + i
 989                                          + offsetof(struct xt_entry_match,
 990                                                     u.user.name),
 991                                          m->u.kernel.match->name,
 992                                          strlen(m->u.kernel.match->name)+1)
 993                             != 0) {
 994                                 ret = -EFAULT;
 995                                 goto free_counters;
 996                         }
 997                 }
 998
 999                 t = ipt_get_target_c(e);
1000                 if (copy_to_user(userptr + off + e->target_offset
1001                                  + offsetof(struct xt_entry_target,
1002                                             u.user.name),
1003                                  t->u.kernel.target->name,
1004                                  strlen(t->u.kernel.target->name)+1) != 0) {
1005                         ret = -EFAULT;
1006                         goto free_counters;
1007                 }
1008         }
1009
1010  free_counters:
1011         vfree(counters);
1012         return ret;
1013 }
1014
1015 #ifdef CONFIG_COMPAT
1016 static void compat_standard_from_user(void *dst, const void *src)
1017 {
1018         int v = *(compat_int_t *)src;
1019
1020         if (v > 0)
1021                 v += xt_compat_calc_jump(AF_INET, v);
1022         memcpy(dst, &v, sizeof(v));
1023 }
1024
1025 static int compat_standard_to_user(void __user *dst, const void *src)
1026 {
1027         compat_int_t cv = *(int *)src;
1028
1029         if (cv > 0)
1030                 cv -= xt_compat_calc_jump(AF_INET, cv);
1031         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1032 }
1033
1034 static int compat_calc_entry(const struct ipt_entry *e,
1035                              const struct xt_table_info *info,
1036                              const void *base, struct xt_table_info *newinfo)
1037 {
1038         const struct xt_entry_match *ematch;
1039         const struct xt_entry_target *t;
1040         unsigned int entry_offset;
1041         int off, i, ret;
1042
1043         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1044         entry_offset = (void *)e - base;
1045         xt_ematch_foreach(ematch, e)
1046                 off += xt_compat_match_offset(ematch->u.kernel.match);
1047         t = ipt_get_target_c(e);
1048         off += xt_compat_target_offset(t->u.kernel.target);
1049         newinfo->size -= off;
1050         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1051         if (ret)
1052                 return ret;
1053
1054         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1055                 if (info->hook_entry[i] &&
1056                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1057                         newinfo->hook_entry[i] -= off;
1058                 if (info->underflow[i] &&
1059                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1060                         newinfo->underflow[i] -= off;
1061         }
1062         return 0;
1063 }
1064
1065 static int compat_table_info(const struct xt_table_info *info,
1066                              struct xt_table_info *newinfo)
1067 {
1068         struct ipt_entry *iter;
1069         const void *loc_cpu_entry;
1070         int ret;
1071
1072         if (!newinfo || !info)
1073                 return -EINVAL;
1074
1075         /* we dont care about newinfo->entries */
1076         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1077         newinfo->initial_entries = 0;
1078         loc_cpu_entry = info->entries;
1079         xt_compat_init_offsets(AF_INET, info->number);
1080         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1081                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1082                 if (ret != 0)
1083                         return ret;
1084         }
1085         return 0;
1086 }
1087 #endif
1088
1089 static int get_info(struct net *net, void __user *user,
1090                     const int *len, int compat)
1091 {
1092         char name[XT_TABLE_MAXNAMELEN];
1093         struct xt_table *t;
1094         int ret;
1095
1096         if (*len != sizeof(struct ipt_getinfo)) {
1097                 duprintf("length %u != %zu\n", *len,
1098                          sizeof(struct ipt_getinfo));
1099                 return -EINVAL;
1100         }
1101
1102         if (copy_from_user(name, user, sizeof(name)) != 0)
1103                 return -EFAULT;
1104
1105         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1106 #ifdef CONFIG_COMPAT
1107         if (compat)
1108                 xt_compat_lock(AF_INET);
1109 #endif
1110         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1111                                     "iptable_%s", name);
1112         if (!IS_ERR_OR_NULL(t)) {
1113                 struct ipt_getinfo info;
1114                 const struct xt_table_info *private = t->private;
1115 #ifdef CONFIG_COMPAT
1116                 struct xt_table_info tmp;
1117
1118                 if (compat) {
1119                         ret = compat_table_info(private, &tmp);
1120                         xt_compat_flush_offsets(AF_INET);
1121                         private = &tmp;
1122                 }
1123 #endif
1124                 memset(&info, 0, sizeof(info));
1125                 info.valid_hooks = t->valid_hooks;
1126                 memcpy(info.hook_entry, private->hook_entry,
1127                        sizeof(info.hook_entry));
1128                 memcpy(info.underflow, private->underflow,
1129                        sizeof(info.underflow));
1130                 info.num_entries = private->number;
1131                 info.size = private->size;
1132                 strcpy(info.name, name);
1133
1134                 if (copy_to_user(user, &info, *len) != 0)
1135                         ret = -EFAULT;
1136                 else
1137                         ret = 0;
1138
1139                 xt_table_unlock(t);
1140                 module_put(t->me);
1141         } else
1142                 ret = t ? PTR_ERR(t) : -ENOENT;
1143 #ifdef CONFIG_COMPAT
1144         if (compat)
1145                 xt_compat_unlock(AF_INET);
1146 #endif
1147         return ret;
1148 }
1149
1150 static int
1151 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1152             const int *len)
1153 {
1154         int ret;
1155         struct ipt_get_entries get;
1156         struct xt_table *t;
1157
1158         if (*len < sizeof(get)) {
1159                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1160                 return -EINVAL;
1161         }
1162         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1163                 return -EFAULT;
1164         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1165                 duprintf("get_entries: %u != %zu\n",
1166                          *len, sizeof(get) + get.size);
1167                 return -EINVAL;
1168         }
1169
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             (unsigned char *)e + e->next_offset > limit) {
1507                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1508                 return -EINVAL;
1509         }
1510
1511         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1512                              sizeof(struct compat_xt_entry_target)) {
1513                 duprintf("checking: element %p size %u\n",
1514                          e, e->next_offset);
1515                 return -EINVAL;
1516         }
1517
1518         /* For purposes of check_entry casting the compat entry is fine */
1519         ret = check_entry((struct ipt_entry *)e);
1520         if (ret)
1521                 return ret;
1522
1523         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1524         entry_offset = (void *)e - (void *)base;
1525         j = 0;
1526         xt_ematch_foreach(ematch, e) {
1527                 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1528                 if (ret != 0)
1529                         goto release_matches;
1530                 ++j;
1531         }
1532
1533         t = compat_ipt_get_target(e);
1534         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1535                                         t->u.user.revision);
1536         if (IS_ERR(target)) {
1537                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1538                          t->u.user.name);
1539                 ret = PTR_ERR(target);
1540                 goto release_matches;
1541         }
1542         t->u.kernel.target = target;
1543
1544         off += xt_compat_target_offset(target);
1545         *size += off;
1546         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1547         if (ret)
1548                 goto out;
1549
1550         /* Check hooks & underflows */
1551         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1552                 if ((unsigned char *)e - base == hook_entries[h])
1553                         newinfo->hook_entry[h] = hook_entries[h];
1554                 if ((unsigned char *)e - base == underflows[h])
1555                         newinfo->underflow[h] = underflows[h];
1556         }
1557
1558         /* Clear counters and comefrom */
1559         memset(&e->counters, 0, sizeof(e->counters));
1560         e->comefrom = 0;
1561         return 0;
1562
1563 out:
1564         module_put(t->u.kernel.target->me);
1565 release_matches:
1566         xt_ematch_foreach(ematch, e) {
1567                 if (j-- == 0)
1568                         break;
1569                 module_put(ematch->u.kernel.match->me);
1570         }
1571         return ret;
1572 }
1573
1574 static int
1575 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1576                             unsigned int *size, const char *name,
1577                             struct xt_table_info *newinfo, unsigned char *base)
1578 {
1579         struct xt_entry_target *t;
1580         struct xt_target *target;
1581         struct ipt_entry *de;
1582         unsigned int origsize;
1583         int ret, h;
1584         struct xt_entry_match *ematch;
1585
1586         ret = 0;
1587         origsize = *size;
1588         de = (struct ipt_entry *)*dstptr;
1589         memcpy(de, e, sizeof(struct ipt_entry));
1590         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1591
1592         *dstptr += sizeof(struct ipt_entry);
1593         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1594
1595         xt_ematch_foreach(ematch, e) {
1596                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1597                 if (ret != 0)
1598                         return ret;
1599         }
1600         de->target_offset = e->target_offset - (origsize - *size);
1601         t = compat_ipt_get_target(e);
1602         target = t->u.kernel.target;
1603         xt_compat_target_from_user(t, dstptr, size);
1604
1605         de->next_offset = e->next_offset - (origsize - *size);
1606         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1607                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1608                         newinfo->hook_entry[h] -= origsize - *size;
1609                 if ((unsigned char *)de - base < newinfo->underflow[h])
1610                         newinfo->underflow[h] -= origsize - *size;
1611         }
1612         return ret;
1613 }
1614
1615 static int
1616 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1617 {
1618         struct xt_entry_match *ematch;
1619         struct xt_mtchk_param mtpar;
1620         unsigned int j;
1621         int ret = 0;
1622
1623         e->counters.pcnt = xt_percpu_counter_alloc();
1624         if (IS_ERR_VALUE(e->counters.pcnt))
1625                 return -ENOMEM;
1626
1627         j = 0;
1628         mtpar.net       = net;
1629         mtpar.table     = name;
1630         mtpar.entryinfo = &e->ip;
1631         mtpar.hook_mask = e->comefrom;
1632         mtpar.family    = NFPROTO_IPV4;
1633         xt_ematch_foreach(ematch, e) {
1634                 ret = check_match(ematch, &mtpar);
1635                 if (ret != 0)
1636                         goto cleanup_matches;
1637                 ++j;
1638         }
1639
1640         ret = check_target(e, net, name);
1641         if (ret)
1642                 goto cleanup_matches;
1643         return 0;
1644
1645  cleanup_matches:
1646         xt_ematch_foreach(ematch, e) {
1647                 if (j-- == 0)
1648                         break;
1649                 cleanup_match(ematch, net);
1650         }
1651
1652         xt_percpu_counter_free(e->counters.pcnt);
1653
1654         return ret;
1655 }
1656
1657 static int
1658 translate_compat_table(struct net *net,
1659                        const char *name,
1660                        unsigned int valid_hooks,
1661                        struct xt_table_info **pinfo,
1662                        void **pentry0,
1663                        unsigned int total_size,
1664                        unsigned int number,
1665                        unsigned int *hook_entries,
1666                        unsigned int *underflows)
1667 {
1668         unsigned int i, j;
1669         struct xt_table_info *newinfo, *info;
1670         void *pos, *entry0, *entry1;
1671         struct compat_ipt_entry *iter0;
1672         struct ipt_entry *iter1;
1673         unsigned int size;
1674         int ret;
1675
1676         info = *pinfo;
1677         entry0 = *pentry0;
1678         size = total_size;
1679         info->number = number;
1680
1681         /* Init all hooks to impossible value. */
1682         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1683                 info->hook_entry[i] = 0xFFFFFFFF;
1684                 info->underflow[i] = 0xFFFFFFFF;
1685         }
1686
1687         duprintf("translate_compat_table: size %u\n", info->size);
1688         j = 0;
1689         xt_compat_lock(AF_INET);
1690         xt_compat_init_offsets(AF_INET, number);
1691         /* Walk through entries, checking offsets. */
1692         xt_entry_foreach(iter0, entry0, total_size) {
1693                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1694                                                         entry0,
1695                                                         entry0 + total_size,
1696                                                         hook_entries,
1697                                                         underflows,
1698                                                         name);
1699                 if (ret != 0)
1700                         goto out_unlock;
1701                 ++j;
1702         }
1703
1704         ret = -EINVAL;
1705         if (j != number) {
1706                 duprintf("translate_compat_table: %u not %u entries\n",
1707                          j, number);
1708                 goto out_unlock;
1709         }
1710
1711         /* Check hooks all assigned */
1712         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1713                 /* Only hooks which are valid */
1714                 if (!(valid_hooks & (1 << i)))
1715                         continue;
1716                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1717                         duprintf("Invalid hook entry %u %u\n",
1718                                  i, hook_entries[i]);
1719                         goto out_unlock;
1720                 }
1721                 if (info->underflow[i] == 0xFFFFFFFF) {
1722                         duprintf("Invalid underflow %u %u\n",
1723                                  i, underflows[i]);
1724                         goto out_unlock;
1725                 }
1726         }
1727
1728         ret = -ENOMEM;
1729         newinfo = xt_alloc_table_info(size);
1730         if (!newinfo)
1731                 goto out_unlock;
1732
1733         newinfo->number = number;
1734         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1735                 newinfo->hook_entry[i] = info->hook_entry[i];
1736                 newinfo->underflow[i] = info->underflow[i];
1737         }
1738         entry1 = newinfo->entries;
1739         pos = entry1;
1740         size = total_size;
1741         xt_entry_foreach(iter0, entry0, total_size) {
1742                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1743                                                   name, newinfo, entry1);
1744                 if (ret != 0)
1745                         break;
1746         }
1747         xt_compat_flush_offsets(AF_INET);
1748         xt_compat_unlock(AF_INET);
1749         if (ret)
1750                 goto free_newinfo;
1751
1752         ret = -ELOOP;
1753         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1754                 goto free_newinfo;
1755
1756         i = 0;
1757         xt_entry_foreach(iter1, entry1, newinfo->size) {
1758                 ret = compat_check_entry(iter1, net, name);
1759                 if (ret != 0)
1760                         break;
1761                 ++i;
1762                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1763                     XT_ERROR_TARGET) == 0)
1764                         ++newinfo->stacksize;
1765         }
1766         if (ret) {
1767                 /*
1768                  * The first i matches need cleanup_entry (calls ->destroy)
1769                  * because they had called ->check already. The other j-i
1770                  * entries need only release.
1771                  */
1772                 int skip = i;
1773                 j -= i;
1774                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1775                         if (skip-- > 0)
1776                                 continue;
1777                         if (j-- == 0)
1778                                 break;
1779                         compat_release_entry(iter0);
1780                 }
1781                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1782                         if (i-- == 0)
1783                                 break;
1784                         cleanup_entry(iter1, net);
1785                 }
1786                 xt_free_table_info(newinfo);
1787                 return ret;
1788         }
1789
1790         *pinfo = newinfo;
1791         *pentry0 = entry1;
1792         xt_free_table_info(info);
1793         return 0;
1794
1795 free_newinfo:
1796         xt_free_table_info(newinfo);
1797 out:
1798         xt_entry_foreach(iter0, entry0, total_size) {
1799                 if (j-- == 0)
1800                         break;
1801                 compat_release_entry(iter0);
1802         }
1803         return ret;
1804 out_unlock:
1805         xt_compat_flush_offsets(AF_INET);
1806         xt_compat_unlock(AF_INET);
1807         goto out;
1808 }
1809
1810 static int
1811 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1812 {
1813         int ret;
1814         struct compat_ipt_replace tmp;
1815         struct xt_table_info *newinfo;
1816         void *loc_cpu_entry;
1817         struct ipt_entry *iter;
1818
1819         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1820                 return -EFAULT;
1821
1822         /* overflow check */
1823         if (tmp.size >= INT_MAX / num_possible_cpus())
1824                 return -ENOMEM;
1825         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1826                 return -ENOMEM;
1827         if (tmp.num_counters == 0)
1828                 return -EINVAL;
1829
1830         tmp.name[sizeof(tmp.name)-1] = 0;
1831
1832         newinfo = xt_alloc_table_info(tmp.size);
1833         if (!newinfo)
1834                 return -ENOMEM;
1835
1836         loc_cpu_entry = newinfo->entries;
1837         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1838                            tmp.size) != 0) {
1839                 ret = -EFAULT;
1840                 goto free_newinfo;
1841         }
1842
1843         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1844                                      &newinfo, &loc_cpu_entry, tmp.size,
1845                                      tmp.num_entries, tmp.hook_entry,
1846                                      tmp.underflow);
1847         if (ret != 0)
1848                 goto free_newinfo;
1849
1850         duprintf("compat_do_replace: Translated table\n");
1851
1852         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1853                            tmp.num_counters, compat_ptr(tmp.counters));
1854         if (ret)
1855                 goto free_newinfo_untrans;
1856         return 0;
1857
1858  free_newinfo_untrans:
1859         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1860                 cleanup_entry(iter, net);
1861  free_newinfo:
1862         xt_free_table_info(newinfo);
1863         return ret;
1864 }
1865
1866 static int
1867 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1868                       unsigned int len)
1869 {
1870         int ret;
1871
1872         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1873                 return -EPERM;
1874
1875         switch (cmd) {
1876         case IPT_SO_SET_REPLACE:
1877                 ret = compat_do_replace(sock_net(sk), user, len);
1878                 break;
1879
1880         case IPT_SO_SET_ADD_COUNTERS:
1881                 ret = do_add_counters(sock_net(sk), user, len, 1);
1882                 break;
1883
1884         default:
1885                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1886                 ret = -EINVAL;
1887         }
1888
1889         return ret;
1890 }
1891
1892 struct compat_ipt_get_entries {
1893         char name[XT_TABLE_MAXNAMELEN];
1894         compat_uint_t size;
1895         struct compat_ipt_entry entrytable[0];
1896 };
1897
1898 static int
1899 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1900                             void __user *userptr)
1901 {
1902         struct xt_counters *counters;
1903         const struct xt_table_info *private = table->private;
1904         void __user *pos;
1905         unsigned int size;
1906         int ret = 0;
1907         unsigned int i = 0;
1908         struct ipt_entry *iter;
1909
1910         counters = alloc_counters(table);
1911         if (IS_ERR(counters))
1912                 return PTR_ERR(counters);
1913
1914         pos = userptr;
1915         size = total_size;
1916         xt_entry_foreach(iter, private->entries, total_size) {
1917                 ret = compat_copy_entry_to_user(iter, &pos,
1918                                                 &size, counters, i++);
1919                 if (ret != 0)
1920                         break;
1921         }
1922
1923         vfree(counters);
1924         return ret;
1925 }
1926
1927 static int
1928 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1929                    int *len)
1930 {
1931         int ret;
1932         struct compat_ipt_get_entries get;
1933         struct xt_table *t;
1934
1935         if (*len < sizeof(get)) {
1936                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1937                 return -EINVAL;
1938         }
1939
1940         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1941                 return -EFAULT;
1942
1943         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1944                 duprintf("compat_get_entries: %u != %zu\n",
1945                          *len, sizeof(get) + get.size);
1946                 return -EINVAL;
1947         }
1948
1949         xt_compat_lock(AF_INET);
1950         t = xt_find_table_lock(net, AF_INET, get.name);
1951         if (!IS_ERR_OR_NULL(t)) {
1952                 const struct xt_table_info *private = t->private;
1953                 struct xt_table_info info;
1954                 duprintf("t->private->number = %u\n", private->number);
1955                 ret = compat_table_info(private, &info);
1956                 if (!ret && get.size == info.size) {
1957                         ret = compat_copy_entries_to_user(private->size,
1958                                                           t, uptr->entrytable);
1959                 } else if (!ret) {
1960                         duprintf("compat_get_entries: I've got %u not %u!\n",
1961                                  private->size, get.size);
1962                         ret = -EAGAIN;
1963                 }
1964                 xt_compat_flush_offsets(AF_INET);
1965                 module_put(t->me);
1966                 xt_table_unlock(t);
1967         } else
1968                 ret = t ? PTR_ERR(t) : -ENOENT;
1969
1970         xt_compat_unlock(AF_INET);
1971         return ret;
1972 }
1973
1974 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1975
1976 static int
1977 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1978 {
1979         int ret;
1980
1981         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1982                 return -EPERM;
1983
1984         switch (cmd) {
1985         case IPT_SO_GET_INFO:
1986                 ret = get_info(sock_net(sk), user, len, 1);
1987                 break;
1988         case IPT_SO_GET_ENTRIES:
1989                 ret = compat_get_entries(sock_net(sk), user, len);
1990                 break;
1991         default:
1992                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1993         }
1994         return ret;
1995 }
1996 #endif
1997
1998 static int
1999 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
2000 {
2001         int ret;
2002
2003         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2004                 return -EPERM;
2005
2006         switch (cmd) {
2007         case IPT_SO_SET_REPLACE:
2008                 ret = do_replace(sock_net(sk), user, len);
2009                 break;
2010
2011         case IPT_SO_SET_ADD_COUNTERS:
2012                 ret = do_add_counters(sock_net(sk), user, len, 0);
2013                 break;
2014
2015         default:
2016                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2017                 ret = -EINVAL;
2018         }
2019
2020         return ret;
2021 }
2022
2023 static int
2024 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2025 {
2026         int ret;
2027
2028         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2029                 return -EPERM;
2030
2031         switch (cmd) {
2032         case IPT_SO_GET_INFO:
2033                 ret = get_info(sock_net(sk), user, len, 0);
2034                 break;
2035
2036         case IPT_SO_GET_ENTRIES:
2037                 ret = get_entries(sock_net(sk), user, len);
2038                 break;
2039
2040         case IPT_SO_GET_REVISION_MATCH:
2041         case IPT_SO_GET_REVISION_TARGET: {
2042                 struct xt_get_revision rev;
2043                 int target;
2044
2045                 if (*len != sizeof(rev)) {
2046                         ret = -EINVAL;
2047                         break;
2048                 }
2049                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2050                         ret = -EFAULT;
2051                         break;
2052                 }
2053                 rev.name[sizeof(rev.name)-1] = 0;
2054
2055                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2056                         target = 1;
2057                 else
2058                         target = 0;
2059
2060                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2061                                                          rev.revision,
2062                                                          target, &ret),
2063                                         "ipt_%s", rev.name);
2064                 break;
2065         }
2066
2067         default:
2068                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2069                 ret = -EINVAL;
2070         }
2071
2072         return ret;
2073 }
2074
2075 struct xt_table *ipt_register_table(struct net *net,
2076                                     const struct xt_table *table,
2077                                     const struct ipt_replace *repl)
2078 {
2079         int ret;
2080         struct xt_table_info *newinfo;
2081         struct xt_table_info bootstrap = {0};
2082         void *loc_cpu_entry;
2083         struct xt_table *new_table;
2084
2085         newinfo = xt_alloc_table_info(repl->size);
2086         if (!newinfo) {
2087                 ret = -ENOMEM;
2088                 goto out;
2089         }
2090
2091         loc_cpu_entry = newinfo->entries;
2092         memcpy(loc_cpu_entry, repl->entries, repl->size);
2093
2094         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2095         if (ret != 0)
2096                 goto out_free;
2097
2098         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2099         if (IS_ERR(new_table)) {
2100                 ret = PTR_ERR(new_table);
2101                 goto out_free;
2102         }
2103
2104         return new_table;
2105
2106 out_free:
2107         xt_free_table_info(newinfo);
2108 out:
2109         return ERR_PTR(ret);
2110 }
2111
2112 void ipt_unregister_table(struct net *net, struct xt_table *table)
2113 {
2114         struct xt_table_info *private;
2115         void *loc_cpu_entry;
2116         struct module *table_owner = table->me;
2117         struct ipt_entry *iter;
2118
2119         private = xt_unregister_table(table);
2120
2121         /* Decrease module usage counts and free resources */
2122         loc_cpu_entry = private->entries;
2123         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2124                 cleanup_entry(iter, net);
2125         if (private->number > private->initial_entries)
2126                 module_put(table_owner);
2127         xt_free_table_info(private);
2128 }
2129
2130 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2131 static inline bool
2132 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2133                      u_int8_t type, u_int8_t code,
2134                      bool invert)
2135 {
2136         return ((test_type == 0xFF) ||
2137                 (type == test_type && code >= min_code && code <= max_code))
2138                 ^ invert;
2139 }
2140
2141 static bool
2142 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2143 {
2144         const struct icmphdr *ic;
2145         struct icmphdr _icmph;
2146         const struct ipt_icmp *icmpinfo = par->matchinfo;
2147
2148         /* Must not be a fragment. */
2149         if (par->fragoff != 0)
2150                 return false;
2151
2152         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2153         if (ic == NULL) {
2154                 /* We've been asked to examine this packet, and we
2155                  * can't.  Hence, no choice but to drop.
2156                  */
2157                 duprintf("Dropping evil ICMP tinygram.\n");
2158                 par->hotdrop = true;
2159                 return false;
2160         }
2161
2162         return icmp_type_code_match(icmpinfo->type,
2163                                     icmpinfo->code[0],
2164                                     icmpinfo->code[1],
2165                                     ic->type, ic->code,
2166                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2167 }
2168
2169 static int icmp_checkentry(const struct xt_mtchk_param *par)
2170 {
2171         const struct ipt_icmp *icmpinfo = par->matchinfo;
2172
2173         /* Must specify no unknown invflags */
2174         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2175 }
2176
2177 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2178         {
2179                 .name             = XT_STANDARD_TARGET,
2180                 .targetsize       = sizeof(int),
2181                 .family           = NFPROTO_IPV4,
2182 #ifdef CONFIG_COMPAT
2183                 .compatsize       = sizeof(compat_int_t),
2184                 .compat_from_user = compat_standard_from_user,
2185                 .compat_to_user   = compat_standard_to_user,
2186 #endif
2187         },
2188         {
2189                 .name             = XT_ERROR_TARGET,
2190                 .target           = ipt_error,
2191                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2192                 .family           = NFPROTO_IPV4,
2193         },
2194 };
2195
2196 static struct nf_sockopt_ops ipt_sockopts = {
2197         .pf             = PF_INET,
2198         .set_optmin     = IPT_BASE_CTL,
2199         .set_optmax     = IPT_SO_SET_MAX+1,
2200         .set            = do_ipt_set_ctl,
2201 #ifdef CONFIG_COMPAT
2202         .compat_set     = compat_do_ipt_set_ctl,
2203 #endif
2204         .get_optmin     = IPT_BASE_CTL,
2205         .get_optmax     = IPT_SO_GET_MAX+1,
2206         .get            = do_ipt_get_ctl,
2207 #ifdef CONFIG_COMPAT
2208         .compat_get     = compat_do_ipt_get_ctl,
2209 #endif
2210         .owner          = THIS_MODULE,
2211 };
2212
2213 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2214         {
2215                 .name       = "icmp",
2216                 .match      = icmp_match,
2217                 .matchsize  = sizeof(struct ipt_icmp),
2218                 .checkentry = icmp_checkentry,
2219                 .proto      = IPPROTO_ICMP,
2220                 .family     = NFPROTO_IPV4,
2221         },
2222 };
2223
2224 static int __net_init ip_tables_net_init(struct net *net)
2225 {
2226         return xt_proto_init(net, NFPROTO_IPV4);
2227 }
2228
2229 static void __net_exit ip_tables_net_exit(struct net *net)
2230 {
2231         xt_proto_fini(net, NFPROTO_IPV4);
2232 }
2233
2234 static struct pernet_operations ip_tables_net_ops = {
2235         .init = ip_tables_net_init,
2236         .exit = ip_tables_net_exit,
2237 };
2238
2239 static int __init ip_tables_init(void)
2240 {
2241         int ret;
2242
2243         ret = register_pernet_subsys(&ip_tables_net_ops);
2244         if (ret < 0)
2245                 goto err1;
2246
2247         /* No one else will be downing sem now, so we won't sleep */
2248         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2249         if (ret < 0)
2250                 goto err2;
2251         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2252         if (ret < 0)
2253                 goto err4;
2254
2255         /* Register setsockopt */
2256         ret = nf_register_sockopt(&ipt_sockopts);
2257         if (ret < 0)
2258                 goto err5;
2259
2260         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2261         return 0;
2262
2263 err5:
2264         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2265 err4:
2266         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2267 err2:
2268         unregister_pernet_subsys(&ip_tables_net_ops);
2269 err1:
2270         return ret;
2271 }
2272
2273 static void __exit ip_tables_fini(void)
2274 {
2275         nf_unregister_sockopt(&ipt_sockopts);
2276
2277         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2278         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2279         unregister_pernet_subsys(&ip_tables_net_ops);
2280 }
2281
2282 EXPORT_SYMBOL(ipt_register_table);
2283 EXPORT_SYMBOL(ipt_unregister_table);
2284 EXPORT_SYMBOL(ipt_do_table);
2285 module_init(ip_tables_init);
2286 module_exit(ip_tables_fini);