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