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  *
   7  * This program is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License version 2 as
   9  * published by the Free Software Foundation.
  10  */
  11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  12 #include <linux/cache.h>
  13 #include <linux/capability.h>
  14 #include <linux/skbuff.h>
  15 #include <linux/kmod.h>
  16 #include <linux/vmalloc.h>
  17 #include <linux/netdevice.h>
  18 #include <linux/module.h>
  19 #include <linux/icmp.h>
  20 #include <net/ip.h>
  21 #include <net/compat.h>
  22 #include <asm/uaccess.h>
  23 #include <linux/mutex.h>
  24 #include <linux/proc_fs.h>
  25 #include <linux/err.h>
  26 #include <linux/cpumask.h>
  27
  28 #include <linux/netfilter/x_tables.h>
  29 #include <linux/netfilter_ipv4/ip_tables.h>
  30 #include <net/netfilter/nf_log.h>
  31 #include "../../netfilter/xt_repldata.h"
  32
  33 MODULE_LICENSE("GPL");
  34 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  35 MODULE_DESCRIPTION("IPv4 packet filter");
  36
  37 /*#define DEBUG_IP_FIREWALL*/
  38 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  39 /*#define DEBUG_IP_FIREWALL_USER*/
  40
  41 #ifdef DEBUG_IP_FIREWALL
  42 #define dprintf(format, args...) pr_info(format , ## args)
  43 #else
  44 #define dprintf(format, args...)
  45 #endif
  46
  47 #ifdef DEBUG_IP_FIREWALL_USER
  48 #define duprintf(format, args...) pr_info(format , ## args)
  49 #else
  50 #define duprintf(format, args...)
  51 #endif
  52
  53 #ifdef CONFIG_NETFILTER_DEBUG
  54 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  55 #else
  56 #define IP_NF_ASSERT(x)
  57 #endif
  58
  59 #if 0
  60 /* All the better to debug you with... */
  61 #define static
  62 #define inline
  63 #endif
  64
  65 void *ipt_alloc_initial_table(const struct xt_table *info)
  66 {
  67         return xt_alloc_initial_table(ipt, IPT);
  68 }
  69 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  70
  71 /* Returns whether matches rule or not. */
  72 /* Performance critical - called for every packet */
  73 static inline bool
  74 ip_packet_match(const struct iphdr *ip,
  75                 const char *indev,
  76                 const char *outdev,
  77                 const struct ipt_ip *ipinfo,
  78                 int isfrag)
  79 {
  80         unsigned long ret;
  81
  82 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  83
  84         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  85                   IPT_INV_SRCIP) ||
  86             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  87                   IPT_INV_DSTIP)) {
  88                 dprintf("Source or dest mismatch.\n");
  89
  90                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  91                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  92                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  93                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  94                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  95                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  96                 return false;
  97         }
  98
  99         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 100
 101         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 102                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 103                         indev, ipinfo->iniface,
 104                         ipinfo->invflags&IPT_INV_VIA_IN ?" (INV)":"");
 105                 return false;
 106         }
 107
 108         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 109
 110         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 111                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 112                         outdev, ipinfo->outiface,
 113                         ipinfo->invflags&IPT_INV_VIA_OUT ?" (INV)":"");
 114                 return false;
 115         }
 116
 117         /* Check specific protocol */
 118         if (ipinfo->proto &&
 119             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 120                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 121                         ip->protocol, ipinfo->proto,
 122                         ipinfo->invflags&IPT_INV_PROTO ? " (INV)":"");
 123                 return false;
 124         }
 125
 126         /* If we have a fragment rule but the packet is not a fragment
 127          * then we return zero */
 128         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 129                 dprintf("Fragment rule but not fragment.%s\n",
 130                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 131                 return false;
 132         }
 133
 134         return true;
 135 }
 136
 137 static bool
 138 ip_checkentry(const struct ipt_ip *ip)
 139 {
 140         if (ip->flags & ~IPT_F_MASK) {
 141                 duprintf("Unknown flag bits set: %08X\n",
 142                          ip->flags & ~IPT_F_MASK);
 143                 return false;
 144         }
 145         if (ip->invflags & ~IPT_INV_MASK) {
 146                 duprintf("Unknown invflag bits set: %08X\n",
 147                          ip->invflags & ~IPT_INV_MASK);
 148                 return false;
 149         }
 150         return true;
 151 }
 152
 153 static unsigned int
 154 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 155 {
 156         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 157
 158         return NF_DROP;
 159 }
 160
 161 /* Performance critical */
 162 static inline struct ipt_entry *
 163 get_entry(const void *base, unsigned int offset)
 164 {
 165         return (struct ipt_entry *)(base + offset);
 166 }
 167
 168 /* All zeroes == unconditional rule. */
 169 /* Mildly perf critical (only if packet tracing is on) */
 170 static inline bool unconditional(const struct ipt_ip *ip)
 171 {
 172         static const struct ipt_ip uncond;
 173
 174         return memcmp(ip, &uncond, sizeof(uncond)) == 0;
 175 #undef FWINV
 176 }
 177
 178 /* for const-correctness */
 179 static inline const struct xt_entry_target *
 180 ipt_get_target_c(const struct ipt_entry *e)
 181 {
 182         return ipt_get_target((struct ipt_entry *)e);
 183 }
 184
 185 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 186 static const char *const hooknames[] = {
 187         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 188         [NF_INET_LOCAL_IN]              = "INPUT",
 189         [NF_INET_FORWARD]               = "FORWARD",
 190         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 191         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 192 };
 193
 194 enum nf_ip_trace_comments {
 195         NF_IP_TRACE_COMMENT_RULE,
 196         NF_IP_TRACE_COMMENT_RETURN,
 197         NF_IP_TRACE_COMMENT_POLICY,
 198 };
 199
 200 static const char *const comments[] = {
 201         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 202         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 203         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 204 };
 205
 206 static struct nf_loginfo trace_loginfo = {
 207         .type = NF_LOG_TYPE_LOG,
 208         .u = {
 209                 .log = {
 210                         .level = 4,
 211                         .logflags = NF_LOG_MASK,
 212                 },
 213         },
 214 };
 215
 216 /* Mildly perf critical (only if packet tracing is on) */
 217 static inline int
 218 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 219                       const char *hookname, const char **chainname,
 220                       const char **comment, unsigned int *rulenum)
 221 {
 222         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 223
 224         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 225                 /* Head of user chain: ERROR target with chainname */
 226                 *chainname = t->target.data;
 227                 (*rulenum) = 0;
 228         } else if (s == e) {
 229                 (*rulenum)++;
 230
 231                 if (s->target_offset == sizeof(struct ipt_entry) &&
 232                     strcmp(t->target.u.kernel.target->name,
 233                            XT_STANDARD_TARGET) == 0 &&
 234                    t->verdict < 0 &&
 235                    unconditional(&s->ip)) {
 236                         /* Tail of chains: STANDARD target (return/policy) */
 237                         *comment = *chainname == hookname
 238                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 239                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 240                 }
 241                 return 1;
 242         } else
 243                 (*rulenum)++;
 244
 245         return 0;
 246 }
 247
 248 static void trace_packet(const struct sk_buff *skb,
 249                          unsigned int hook,
 250                          const struct net_device *in,
 251                          const struct net_device *out,
 252                          const char *tablename,
 253                          const struct xt_table_info *private,
 254                          const struct ipt_entry *e)
 255 {
 256         const void *table_base;
 257         const struct ipt_entry *root;
 258         const char *hookname, *chainname, *comment;
 259         const struct ipt_entry *iter;
 260         unsigned int rulenum = 0;
 261         struct net *net = dev_net(in ? in : out);
 262
 263         table_base = private->entries[smp_processor_id()];
 264         root = get_entry(table_base, private->hook_entry[hook]);
 265
 266         hookname = chainname = hooknames[hook];
 267         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 268
 269         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 270                 if (get_chainname_rulenum(iter, e, hookname,
 271                     &chainname, &comment, &rulenum) != 0)
 272                         break;
 273
 274         nf_log_packet(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 275                       "TRACE: %s:%s:%s:%u ",
 276                       tablename, chainname, comment, rulenum);
 277 }
 278 #endif
 279
 280 static inline __pure
 281 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 282 {
 283         return (void *)entry + entry->next_offset;
 284 }
 285
 286 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 287 unsigned int
 288 ipt_do_table(struct sk_buff *skb,
 289              unsigned int hook,
 290              const struct net_device *in,
 291              const struct net_device *out,
 292              struct xt_table *table)
 293 {
 294         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 295         const struct iphdr *ip;
 296         /* Initializing verdict to NF_DROP keeps gcc happy. */
 297         unsigned int verdict = NF_DROP;
 298         const char *indev, *outdev;
 299         const void *table_base;
 300         struct ipt_entry *e, **jumpstack;
 301         unsigned int *stackptr, origptr, cpu;
 302         const struct xt_table_info *private;
 303         struct xt_action_param acpar;
 304         unsigned int addend;
 305
 306         /* Initialization */
 307         ip = ip_hdr(skb);
 308         indev = in ? in->name : nulldevname;
 309         outdev = out ? out->name : nulldevname;
 310         /* We handle fragments by dealing with the first fragment as
 311          * if it was a normal packet.  All other fragments are treated
 312          * normally, except that they will NEVER match rules that ask
 313          * things we don't know, ie. tcp syn flag or ports).  If the
 314          * rule is also a fragment-specific rule, non-fragments won't
 315          * match it. */
 316         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 317         acpar.thoff   = ip_hdrlen(skb);
 318         acpar.hotdrop = false;
 319         acpar.in      = in;
 320         acpar.out     = 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         table_base = private->entries[cpu];
 330         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 331         stackptr   = per_cpu_ptr(private->stackptr, cpu);
 332         origptr    = *stackptr;
 333
 334         e = get_entry(table_base, private->hook_entry[hook]);
 335
 336         pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
 337                  table->name, hook, origptr,
 338                  get_entry(table_base, private->underflow[hook]));
 339
 340         do {
 341                 const struct xt_entry_target *t;
 342                 const struct xt_entry_match *ematch;
 343
 344                 IP_NF_ASSERT(e);
 345                 if (!ip_packet_match(ip, indev, outdev,
 346                     &e->ip, acpar.fragoff)) {
 347  no_match:
 348                         e = ipt_next_entry(e);
 349                         continue;
 350                 }
 351
 352                 xt_ematch_foreach(ematch, e) {
 353                         acpar.match     = ematch->u.kernel.match;
 354                         acpar.matchinfo = ematch->data;
 355                         if (!acpar.match->match(skb, &acpar))
 356                                 goto no_match;
 357                 }
 358
 359                 ADD_COUNTER(e->counters, skb->len, 1);
 360
 361                 t = ipt_get_target(e);
 362                 IP_NF_ASSERT(t->u.kernel.target);
 363
 364 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 365                 /* The packet is traced: log it */
 366                 if (unlikely(skb->nf_trace))
 367                         trace_packet(skb, hook, in, out,
 368                                      table->name, private, e);
 369 #endif
 370                 /* Standard target? */
 371                 if (!t->u.kernel.target->target) {
 372                         int v;
 373
 374                         v = ((struct xt_standard_target *)t)->verdict;
 375                         if (v < 0) {
 376                                 /* Pop from stack? */
 377                                 if (v != XT_RETURN) {
 378                                         verdict = (unsigned int)(-v) - 1;
 379                                         break;
 380                                 }
 381                                 if (*stackptr <= origptr) {
 382                                         e = get_entry(table_base,
 383                                             private->underflow[hook]);
 384                                         pr_debug("Underflow (this is normal) "
 385                                                  "to %p\n", e);
 386                                 } else {
 387                                         e = jumpstack[--*stackptr];
 388                                         pr_debug("Pulled %p out from pos %u\n",
 389                                                  e, *stackptr);
 390                                         e = ipt_next_entry(e);
 391                                 }
 392                                 continue;
 393                         }
 394                         if (table_base + v != ipt_next_entry(e) &&
 395                             !(e->ip.flags & IPT_F_GOTO)) {
 396                                 if (*stackptr >= private->stacksize) {
 397                                         verdict = NF_DROP;
 398                                         break;
 399                                 }
 400                                 jumpstack[(*stackptr)++] = e;
 401                                 pr_debug("Pushed %p into pos %u\n",
 402                                          e, *stackptr - 1);
 403                         }
 404
 405                         e = get_entry(table_base, v);
 406                         continue;
 407                 }
 408
 409                 acpar.target   = t->u.kernel.target;
 410                 acpar.targinfo = t->data;
 411
 412                 verdict = t->u.kernel.target->target(skb, &acpar);
 413                 /* Target might have changed stuff. */
 414                 ip = ip_hdr(skb);
 415                 if (verdict == XT_CONTINUE)
 416                         e = ipt_next_entry(e);
 417                 else
 418                         /* Verdict */
 419                         break;
 420         } while (!acpar.hotdrop);
 421         pr_debug("Exiting %s; resetting sp from %u to %u\n",
 422                  __func__, *stackptr, origptr);
 423         *stackptr = origptr;
 424         xt_write_recseq_end(addend);
 425         local_bh_enable();
 426
 427 #ifdef DEBUG_ALLOW_ALL
 428         return NF_ACCEPT;
 429 #else
 430         if (acpar.hotdrop)
 431                 return NF_DROP;
 432         else return verdict;
 433 #endif
 434 }
 435
 436 /* Figures out from what hook each rule can be called: returns 0 if
 437    there are loops.  Puts hook bitmask in comefrom. */
 438 static int
 439 mark_source_chains(const struct xt_table_info *newinfo,
 440                    unsigned int valid_hooks, void *entry0)
 441 {
 442         unsigned int hook;
 443
 444         /* No recursion; use packet counter to save back ptrs (reset
 445            to 0 as we leave), and comefrom to save source hook bitmask */
 446         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 447                 unsigned int pos = newinfo->hook_entry[hook];
 448                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 449
 450                 if (!(valid_hooks & (1 << hook)))
 451                         continue;
 452
 453                 /* Set initial back pointer. */
 454                 e->counters.pcnt = pos;
 455
 456                 for (;;) {
 457                         const struct xt_standard_target *t
 458                                 = (void *)ipt_get_target_c(e);
 459                         int visited = e->comefrom & (1 << hook);
 460
 461                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 462                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 463                                        hook, pos, e->comefrom);
 464                                 return 0;
 465                         }
 466                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 467
 468                         /* Unconditional return/END. */
 469                         if ((e->target_offset == sizeof(struct ipt_entry) &&
 470                              (strcmp(t->target.u.user.name,
 471                                      XT_STANDARD_TARGET) == 0) &&
 472                              t->verdict < 0 && unconditional(&e->ip)) ||
 473                             visited) {
 474                                 unsigned int oldpos, size;
 475
 476                                 if ((strcmp(t->target.u.user.name,
 477                                             XT_STANDARD_TARGET) == 0) &&
 478                                     t->verdict < -NF_MAX_VERDICT - 1) {
 479                                         duprintf("mark_source_chains: bad "
 480                                                 "negative verdict (%i)\n",
 481                                                                 t->verdict);
 482                                         return 0;
 483                                 }
 484
 485                                 /* Return: backtrack through the last
 486                                    big jump. */
 487                                 do {
 488                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 489 #ifdef DEBUG_IP_FIREWALL_USER
 490                                         if (e->comefrom
 491                                             & (1 << NF_INET_NUMHOOKS)) {
 492                                                 duprintf("Back unset "
 493                                                          "on hook %u "
 494                                                          "rule %u\n",
 495                                                          hook, pos);
 496                                         }
 497 #endif
 498                                         oldpos = pos;
 499                                         pos = e->counters.pcnt;
 500                                         e->counters.pcnt = 0;
 501
 502                                         /* We're at the start. */
 503                                         if (pos == oldpos)
 504                                                 goto next;
 505
 506                                         e = (struct ipt_entry *)
 507                                                 (entry0 + pos);
 508                                 } while (oldpos == pos + e->next_offset);
 509
 510                                 /* Move along one */
 511                                 size = e->next_offset;
 512                                 e = (struct ipt_entry *)
 513                                         (entry0 + pos + size);
 514                                 e->counters.pcnt = pos;
 515                                 pos += size;
 516                         } else {
 517                                 int newpos = t->verdict;
 518
 519                                 if (strcmp(t->target.u.user.name,
 520                                            XT_STANDARD_TARGET) == 0 &&
 521                                     newpos >= 0) {
 522                                         if (newpos > newinfo->size -
 523                                                 sizeof(struct ipt_entry)) {
 524                                                 duprintf("mark_source_chains: "
 525                                                         "bad verdict (%i)\n",
 526                                                                 newpos);
 527                                                 return 0;
 528                                         }
 529                                         /* This a jump; chase it. */
 530                                         duprintf("Jump rule %u -> %u\n",
 531                                                  pos, newpos);
 532                                 } else {
 533                                         /* ... this is a fallthru */
 534                                         newpos = pos + e->next_offset;
 535                                 }
 536                                 e = (struct ipt_entry *)
 537                                         (entry0 + newpos);
 538                                 e->counters.pcnt = pos;
 539                                 pos = newpos;
 540                         }
 541                 }
 542                 next:
 543                 duprintf("Finished chain %u\n", hook);
 544         }
 545         return 1;
 546 }
 547
 548 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 549 {
 550         struct xt_mtdtor_param par;
 551
 552         par.net       = net;
 553         par.match     = m->u.kernel.match;
 554         par.matchinfo = m->data;
 555         par.family    = NFPROTO_IPV4;
 556         if (par.match->destroy != NULL)
 557                 par.match->destroy(&par);
 558         module_put(par.match->me);
 559 }
 560
 561 static int
 562 check_entry(const struct ipt_entry *e, const char *name)
 563 {
 564         const struct xt_entry_target *t;
 565
 566         if (!ip_checkentry(&e->ip)) {
 567                 duprintf("ip check failed %p %s.\n", e, name);
 568                 return -EINVAL;
 569         }
 570
 571         if (e->target_offset + sizeof(struct xt_entry_target) >
 572             e->next_offset)
 573                 return -EINVAL;
 574
 575         t = ipt_get_target_c(e);
 576         if (e->target_offset + t->u.target_size > e->next_offset)
 577                 return -EINVAL;
 578
 579         return 0;
 580 }
 581
 582 static int
 583 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 584 {
 585         const struct ipt_ip *ip = par->entryinfo;
 586         int ret;
 587
 588         par->match     = m->u.kernel.match;
 589         par->matchinfo = m->data;
 590
 591         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 592               ip->proto, ip->invflags & IPT_INV_PROTO);
 593         if (ret < 0) {
 594                 duprintf("check failed for `%s'.\n", par->match->name);
 595                 return ret;
 596         }
 597         return 0;
 598 }
 599
 600 static int
 601 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 602 {
 603         struct xt_match *match;
 604         int ret;
 605
 606         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 607                                       m->u.user.revision);
 608         if (IS_ERR(match)) {
 609                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 610                 return PTR_ERR(match);
 611         }
 612         m->u.kernel.match = match;
 613
 614         ret = check_match(m, par);
 615         if (ret)
 616                 goto err;
 617
 618         return 0;
 619 err:
 620         module_put(m->u.kernel.match->me);
 621         return ret;
 622 }
 623
 624 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 625 {
 626         struct xt_entry_target *t = ipt_get_target(e);
 627         struct xt_tgchk_param par = {
 628                 .net       = net,
 629                 .table     = name,
 630                 .entryinfo = e,
 631                 .target    = t->u.kernel.target,
 632                 .targinfo  = t->data,
 633                 .hook_mask = e->comefrom,
 634                 .family    = NFPROTO_IPV4,
 635         };
 636         int ret;
 637
 638         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 639               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 640         if (ret < 0) {
 641                 duprintf("check failed for `%s'.\n",
 642                          t->u.kernel.target->name);
 643                 return ret;
 644         }
 645         return 0;
 646 }
 647
 648 static int
 649 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 650                  unsigned int size)
 651 {
 652         struct xt_entry_target *t;
 653         struct xt_target *target;
 654         int ret;
 655         unsigned int j;
 656         struct xt_mtchk_param mtpar;
 657         struct xt_entry_match *ematch;
 658
 659         ret = check_entry(e, name);
 660         if (ret)
 661                 return ret;
 662
 663         j = 0;
 664         mtpar.net       = net;
 665         mtpar.table     = name;
 666         mtpar.entryinfo = &e->ip;
 667         mtpar.hook_mask = e->comefrom;
 668         mtpar.family    = NFPROTO_IPV4;
 669         xt_ematch_foreach(ematch, e) {
 670                 ret = find_check_match(ematch, &mtpar);
 671                 if (ret != 0)
 672                         goto cleanup_matches;
 673                 ++j;
 674         }
 675
 676         t = ipt_get_target(e);
 677         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 678                                         t->u.user.revision);
 679         if (IS_ERR(target)) {
 680                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 681                 ret = PTR_ERR(target);
 682                 goto cleanup_matches;
 683         }
 684         t->u.kernel.target = target;
 685
 686         ret = check_target(e, net, name);
 687         if (ret)
 688                 goto err;
 689         return 0;
 690  err:
 691         module_put(t->u.kernel.target->me);
 692  cleanup_matches:
 693         xt_ematch_foreach(ematch, e) {
 694                 if (j-- == 0)
 695                         break;
 696                 cleanup_match(ematch, net);
 697         }
 698         return ret;
 699 }
 700
 701 static bool check_underflow(const struct ipt_entry *e)
 702 {
 703         const struct xt_entry_target *t;
 704         unsigned int verdict;
 705
 706         if (!unconditional(&e->ip))
 707                 return false;
 708         t = ipt_get_target_c(e);
 709         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 710                 return false;
 711         verdict = ((struct xt_standard_target *)t)->verdict;
 712         verdict = -verdict - 1;
 713         return verdict == NF_DROP || verdict == NF_ACCEPT;
 714 }
 715
 716 static int
 717 check_entry_size_and_hooks(struct ipt_entry *e,
 718                            struct xt_table_info *newinfo,
 719                            const unsigned char *base,
 720                            const unsigned char *limit,
 721                            const unsigned int *hook_entries,
 722                            const unsigned int *underflows,
 723                            unsigned int valid_hooks)
 724 {
 725         unsigned int h;
 726
 727         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 728             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 729                 duprintf("Bad offset %p\n", e);
 730                 return -EINVAL;
 731         }
 732
 733         if (e->next_offset
 734             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 735                 duprintf("checking: element %p size %u\n",
 736                          e, e->next_offset);
 737                 return -EINVAL;
 738         }
 739
 740         /* Check hooks & underflows */
 741         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 742                 if (!(valid_hooks & (1 << h)))
 743                         continue;
 744                 if ((unsigned char *)e - base == hook_entries[h])
 745                         newinfo->hook_entry[h] = hook_entries[h];
 746                 if ((unsigned char *)e - base == underflows[h]) {
 747                         if (!check_underflow(e)) {
 748                                 pr_err("Underflows must be unconditional and "
 749                                        "use the STANDARD target with "
 750                                        "ACCEPT/DROP\n");
 751                                 return -EINVAL;
 752                         }
 753                         newinfo->underflow[h] = underflows[h];
 754                 }
 755         }
 756
 757         /* Clear counters and comefrom */
 758         e->counters = ((struct xt_counters) { 0, 0 });
 759         e->comefrom = 0;
 760         return 0;
 761 }
 762
 763 static void
 764 cleanup_entry(struct ipt_entry *e, struct net *net)
 765 {
 766         struct xt_tgdtor_param par;
 767         struct xt_entry_target *t;
 768         struct xt_entry_match *ematch;
 769
 770         /* Cleanup all matches */
 771         xt_ematch_foreach(ematch, e)
 772                 cleanup_match(ematch, net);
 773         t = ipt_get_target(e);
 774
 775         par.net      = net;
 776         par.target   = t->u.kernel.target;
 777         par.targinfo = t->data;
 778         par.family   = NFPROTO_IPV4;
 779         if (par.target->destroy != NULL)
 780                 par.target->destroy(&par);
 781         module_put(par.target->me);
 782 }
 783
 784 /* Checks and translates the user-supplied table segment (held in
 785    newinfo) */
 786 static int
 787 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 788                 const struct ipt_replace *repl)
 789 {
 790         struct ipt_entry *iter;
 791         unsigned int i;
 792         int ret = 0;
 793
 794         newinfo->size = repl->size;
 795         newinfo->number = repl->num_entries;
 796
 797         /* Init all hooks to impossible value. */
 798         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 799                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 800                 newinfo->underflow[i] = 0xFFFFFFFF;
 801         }
 802
 803         duprintf("translate_table: size %u\n", newinfo->size);
 804         i = 0;
 805         /* Walk through entries, checking offsets. */
 806         xt_entry_foreach(iter, entry0, newinfo->size) {
 807                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 808                                                  entry0 + repl->size,
 809                                                  repl->hook_entry,
 810                                                  repl->underflow,
 811                                                  repl->valid_hooks);
 812                 if (ret != 0)
 813                         return ret;
 814                 ++i;
 815                 if (strcmp(ipt_get_target(iter)->u.user.name,
 816                     XT_ERROR_TARGET) == 0)
 817                         ++newinfo->stacksize;
 818         }
 819
 820         if (i != repl->num_entries) {
 821                 duprintf("translate_table: %u not %u entries\n",
 822                          i, repl->num_entries);
 823                 return -EINVAL;
 824         }
 825
 826         /* Check hooks all assigned */
 827         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 828                 /* Only hooks which are valid */
 829                 if (!(repl->valid_hooks & (1 << i)))
 830                         continue;
 831                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 832                         duprintf("Invalid hook entry %u %u\n",
 833                                  i, repl->hook_entry[i]);
 834                         return -EINVAL;
 835                 }
 836                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 837                         duprintf("Invalid underflow %u %u\n",
 838                                  i, repl->underflow[i]);
 839                         return -EINVAL;
 840                 }
 841         }
 842
 843         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 844                 return -ELOOP;
 845
 846         /* Finally, each sanity check must pass */
 847         i = 0;
 848         xt_entry_foreach(iter, entry0, newinfo->size) {
 849                 ret = find_check_entry(iter, net, repl->name, repl->size);
 850                 if (ret != 0)
 851                         break;
 852                 ++i;
 853         }
 854
 855         if (ret != 0) {
 856                 xt_entry_foreach(iter, entry0, newinfo->size) {
 857                         if (i-- == 0)
 858                                 break;
 859                         cleanup_entry(iter, net);
 860                 }
 861                 return ret;
 862         }
 863
 864         /* And one copy for every other CPU */
 865         for_each_possible_cpu(i) {
 866                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
 867                         memcpy(newinfo->entries[i], entry0, newinfo->size);
 868         }
 869
 870         return ret;
 871 }
 872
 873 static void
 874 get_counters(const struct xt_table_info *t,
 875              struct xt_counters counters[])
 876 {
 877         struct ipt_entry *iter;
 878         unsigned int cpu;
 879         unsigned int i;
 880
 881         for_each_possible_cpu(cpu) {
 882                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 883
 884                 i = 0;
 885                 xt_entry_foreach(iter, t->entries[cpu], t->size) {
 886                         u64 bcnt, pcnt;
 887                         unsigned int start;
 888
 889                         do {
 890                                 start = read_seqcount_begin(s);
 891                                 bcnt = iter->counters.bcnt;
 892                                 pcnt = iter->counters.pcnt;
 893                         } while (read_seqcount_retry(s, start));
 894
 895                         ADD_COUNTER(counters[i], bcnt, pcnt);
 896                         ++i; /* macro does multi eval of i */
 897                 }
 898         }
 899 }
 900
 901 static struct xt_counters *alloc_counters(const struct xt_table *table)
 902 {
 903         unsigned int countersize;
 904         struct xt_counters *counters;
 905         const struct xt_table_info *private = table->private;
 906
 907         /* We need atomic snapshot of counters: rest doesn't change
 908            (other than comefrom, which userspace doesn't care
 909            about). */
 910         countersize = sizeof(struct xt_counters) * private->number;
 911         counters = vzalloc(countersize);
 912
 913         if (counters == NULL)
 914                 return ERR_PTR(-ENOMEM);
 915
 916         get_counters(private, counters);
 917
 918         return counters;
 919 }
 920
 921 static int
 922 copy_entries_to_user(unsigned int total_size,
 923                      const struct xt_table *table,
 924                      void __user *userptr)
 925 {
 926         unsigned int off, num;
 927         const struct ipt_entry *e;
 928         struct xt_counters *counters;
 929         const struct xt_table_info *private = table->private;
 930         int ret = 0;
 931         const void *loc_cpu_entry;
 932
 933         counters = alloc_counters(table);
 934         if (IS_ERR(counters))
 935                 return PTR_ERR(counters);
 936
 937         /* choose the copy that is on our node/cpu, ...
 938          * This choice is lazy (because current thread is
 939          * allowed to migrate to another cpu)
 940          */
 941         loc_cpu_entry = private->entries[raw_smp_processor_id()];
 942         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 943                 ret = -EFAULT;
 944                 goto free_counters;
 945         }
 946
 947         /* FIXME: use iterator macros --RR */
 948         /* ... then go back and fix counters and names */
 949         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 950                 unsigned int i;
 951                 const struct xt_entry_match *m;
 952                 const struct xt_entry_target *t;
 953
 954                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 955                 if (copy_to_user(userptr + off
 956                                  + offsetof(struct ipt_entry, counters),
 957                                  &counters[num],
 958                                  sizeof(counters[num])) != 0) {
 959                         ret = -EFAULT;
 960                         goto free_counters;
 961                 }
 962
 963                 for (i = sizeof(struct ipt_entry);
 964                      i < e->target_offset;
 965                      i += m->u.match_size) {
 966                         m = (void *)e + i;
 967
 968                         if (copy_to_user(userptr + off + i
 969                                          + offsetof(struct xt_entry_match,
 970                                                     u.user.name),
 971                                          m->u.kernel.match->name,
 972                                          strlen(m->u.kernel.match->name)+1)
 973                             != 0) {
 974                                 ret = -EFAULT;
 975                                 goto free_counters;
 976                         }
 977                 }
 978
 979                 t = ipt_get_target_c(e);
 980                 if (copy_to_user(userptr + off + e->target_offset
 981                                  + offsetof(struct xt_entry_target,
 982                                             u.user.name),
 983                                  t->u.kernel.target->name,
 984                                  strlen(t->u.kernel.target->name)+1) != 0) {
 985                         ret = -EFAULT;
 986                         goto free_counters;
 987                 }
 988         }
 989
 990  free_counters:
 991         vfree(counters);
 992         return ret;
 993 }
 994
 995 #ifdef CONFIG_COMPAT
 996 static void compat_standard_from_user(void *dst, const void *src)
 997 {
 998         int v = *(compat_int_t *)src;
 999
1000         if (v > 0)
1001                 v += xt_compat_calc_jump(AF_INET, v);
1002         memcpy(dst, &v, sizeof(v));
1003 }
1004
1005 static int compat_standard_to_user(void __user *dst, const void *src)
1006 {
1007         compat_int_t cv = *(int *)src;
1008
1009         if (cv > 0)
1010                 cv -= xt_compat_calc_jump(AF_INET, cv);
1011         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1012 }
1013
1014 static int compat_calc_entry(const struct ipt_entry *e,
1015                              const struct xt_table_info *info,
1016                              const void *base, struct xt_table_info *newinfo)
1017 {
1018         const struct xt_entry_match *ematch;
1019         const struct xt_entry_target *t;
1020         unsigned int entry_offset;
1021         int off, i, ret;
1022
1023         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1024         entry_offset = (void *)e - base;
1025         xt_ematch_foreach(ematch, e)
1026                 off += xt_compat_match_offset(ematch->u.kernel.match);
1027         t = ipt_get_target_c(e);
1028         off += xt_compat_target_offset(t->u.kernel.target);
1029         newinfo->size -= off;
1030         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1031         if (ret)
1032                 return ret;
1033
1034         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1035                 if (info->hook_entry[i] &&
1036                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1037                         newinfo->hook_entry[i] -= off;
1038                 if (info->underflow[i] &&
1039                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1040                         newinfo->underflow[i] -= off;
1041         }
1042         return 0;
1043 }
1044
1045 static int compat_table_info(const struct xt_table_info *info,
1046                              struct xt_table_info *newinfo)
1047 {
1048         struct ipt_entry *iter;
1049         void *loc_cpu_entry;
1050         int ret;
1051
1052         if (!newinfo || !info)
1053                 return -EINVAL;
1054
1055         /* we dont care about newinfo->entries[] */
1056         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1057         newinfo->initial_entries = 0;
1058         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1059         xt_compat_init_offsets(AF_INET, info->number);
1060         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1061                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1062                 if (ret != 0)
1063                         return ret;
1064         }
1065         return 0;
1066 }
1067 #endif
1068
1069 static int get_info(struct net *net, void __user *user,
1070                     const int *len, int compat)
1071 {
1072         char name[XT_TABLE_MAXNAMELEN];
1073         struct xt_table *t;
1074         int ret;
1075
1076         if (*len != sizeof(struct ipt_getinfo)) {
1077                 duprintf("length %u != %zu\n", *len,
1078                          sizeof(struct ipt_getinfo));
1079                 return -EINVAL;
1080         }
1081
1082         if (copy_from_user(name, user, sizeof(name)) != 0)
1083                 return -EFAULT;
1084
1085         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1086 #ifdef CONFIG_COMPAT
1087         if (compat)
1088                 xt_compat_lock(AF_INET);
1089 #endif
1090         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1091                                     "iptable_%s", name);
1092         if (!IS_ERR_OR_NULL(t)) {
1093                 struct ipt_getinfo info;
1094                 const struct xt_table_info *private = t->private;
1095 #ifdef CONFIG_COMPAT
1096                 struct xt_table_info tmp;
1097
1098                 if (compat) {
1099                         ret = compat_table_info(private, &tmp);
1100                         xt_compat_flush_offsets(AF_INET);
1101                         private = &tmp;
1102                 }
1103 #endif
1104                 memset(&info, 0, sizeof(info));
1105                 info.valid_hooks = t->valid_hooks;
1106                 memcpy(info.hook_entry, private->hook_entry,
1107                        sizeof(info.hook_entry));
1108                 memcpy(info.underflow, private->underflow,
1109                        sizeof(info.underflow));
1110                 info.num_entries = private->number;
1111                 info.size = private->size;
1112                 strcpy(info.name, name);
1113
1114                 if (copy_to_user(user, &info, *len) != 0)
1115                         ret = -EFAULT;
1116                 else
1117                         ret = 0;
1118
1119                 xt_table_unlock(t);
1120                 module_put(t->me);
1121         } else
1122                 ret = t ? PTR_ERR(t) : -ENOENT;
1123 #ifdef CONFIG_COMPAT
1124         if (compat)
1125                 xt_compat_unlock(AF_INET);
1126 #endif
1127         return ret;
1128 }
1129
1130 static int
1131 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1132             const int *len)
1133 {
1134         int ret;
1135         struct ipt_get_entries get;
1136         struct xt_table *t;
1137
1138         if (*len < sizeof(get)) {
1139                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1140                 return -EINVAL;
1141         }
1142         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1143                 return -EFAULT;
1144         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1145                 duprintf("get_entries: %u != %zu\n",
1146                          *len, sizeof(get) + get.size);
1147                 return -EINVAL;
1148         }
1149
1150         t = xt_find_table_lock(net, AF_INET, get.name);
1151         if (!IS_ERR_OR_NULL(t)) {
1152                 const struct xt_table_info *private = t->private;
1153                 duprintf("t->private->number = %u\n", private->number);
1154                 if (get.size == private->size)
1155                         ret = copy_entries_to_user(private->size,
1156                                                    t, uptr->entrytable);
1157                 else {
1158                         duprintf("get_entries: I've got %u not %u!\n",
1159                                  private->size, get.size);
1160                         ret = -EAGAIN;
1161                 }
1162                 module_put(t->me);
1163                 xt_table_unlock(t);
1164         } else
1165                 ret = t ? PTR_ERR(t) : -ENOENT;
1166
1167         return ret;
1168 }
1169
1170 static int
1171 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1172              struct xt_table_info *newinfo, unsigned int num_counters,
1173              void __user *counters_ptr)
1174 {
1175         int ret;
1176         struct xt_table *t;
1177         struct xt_table_info *oldinfo;
1178         struct xt_counters *counters;
1179         void *loc_cpu_old_entry;
1180         struct ipt_entry *iter;
1181
1182         ret = 0;
1183         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1184         if (!counters) {
1185                 ret = -ENOMEM;
1186                 goto out;
1187         }
1188
1189         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1190                                     "iptable_%s", name);
1191         if (IS_ERR_OR_NULL(t)) {
1192                 ret = t ? PTR_ERR(t) : -ENOENT;
1193                 goto free_newinfo_counters_untrans;
1194         }
1195
1196         /* You lied! */
1197         if (valid_hooks != t->valid_hooks) {
1198                 duprintf("Valid hook crap: %08X vs %08X\n",
1199                          valid_hooks, t->valid_hooks);
1200                 ret = -EINVAL;
1201                 goto put_module;
1202         }
1203
1204         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1205         if (!oldinfo)
1206                 goto put_module;
1207
1208         /* Update module usage count based on number of rules */
1209         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1210                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1211         if ((oldinfo->number > oldinfo->initial_entries) ||
1212             (newinfo->number <= oldinfo->initial_entries))
1213                 module_put(t->me);
1214         if ((oldinfo->number > oldinfo->initial_entries) &&
1215             (newinfo->number <= oldinfo->initial_entries))
1216                 module_put(t->me);
1217
1218         /* Get the old counters, and synchronize with replace */
1219         get_counters(oldinfo, counters);
1220
1221         /* Decrease module usage counts and free resource */
1222         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1223         xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1224                 cleanup_entry(iter, net);
1225
1226         xt_free_table_info(oldinfo);
1227         if (copy_to_user(counters_ptr, counters,
1228                          sizeof(struct xt_counters) * num_counters) != 0)
1229                 ret = -EFAULT;
1230         vfree(counters);
1231         xt_table_unlock(t);
1232         return ret;
1233
1234  put_module:
1235         module_put(t->me);
1236         xt_table_unlock(t);
1237  free_newinfo_counters_untrans:
1238         vfree(counters);
1239  out:
1240         return ret;
1241 }
1242
1243 static int
1244 do_replace(struct net *net, const void __user *user, unsigned int len)
1245 {
1246         int ret;
1247         struct ipt_replace tmp;
1248         struct xt_table_info *newinfo;
1249         void *loc_cpu_entry;
1250         struct ipt_entry *iter;
1251
1252         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1253                 return -EFAULT;
1254
1255         /* overflow check */
1256         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1257                 return -ENOMEM;
1258         tmp.name[sizeof(tmp.name)-1] = 0;
1259
1260         newinfo = xt_alloc_table_info(tmp.size);
1261         if (!newinfo)
1262                 return -ENOMEM;
1263
1264         /* choose the copy that is on our node/cpu */
1265         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1266         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1267                            tmp.size) != 0) {
1268                 ret = -EFAULT;
1269                 goto free_newinfo;
1270         }
1271
1272         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1273         if (ret != 0)
1274                 goto free_newinfo;
1275
1276         duprintf("Translated table\n");
1277
1278         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1279                            tmp.num_counters, tmp.counters);
1280         if (ret)
1281                 goto free_newinfo_untrans;
1282         return 0;
1283
1284  free_newinfo_untrans:
1285         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1286                 cleanup_entry(iter, net);
1287  free_newinfo:
1288         xt_free_table_info(newinfo);
1289         return ret;
1290 }
1291
1292 static int
1293 do_add_counters(struct net *net, const void __user *user,
1294                 unsigned int len, int compat)
1295 {
1296         unsigned int i, curcpu;
1297         struct xt_counters_info tmp;
1298         struct xt_counters *paddc;
1299         unsigned int num_counters;
1300         const char *name;
1301         int size;
1302         void *ptmp;
1303         struct xt_table *t;
1304         const struct xt_table_info *private;
1305         int ret = 0;
1306         void *loc_cpu_entry;
1307         struct ipt_entry *iter;
1308         unsigned int addend;
1309 #ifdef CONFIG_COMPAT
1310         struct compat_xt_counters_info compat_tmp;
1311
1312         if (compat) {
1313                 ptmp = &compat_tmp;
1314                 size = sizeof(struct compat_xt_counters_info);
1315         } else
1316 #endif
1317         {
1318                 ptmp = &tmp;
1319                 size = sizeof(struct xt_counters_info);
1320         }
1321
1322         if (copy_from_user(ptmp, user, size) != 0)
1323                 return -EFAULT;
1324
1325 #ifdef CONFIG_COMPAT
1326         if (compat) {
1327                 num_counters = compat_tmp.num_counters;
1328                 name = compat_tmp.name;
1329         } else
1330 #endif
1331         {
1332                 num_counters = tmp.num_counters;
1333                 name = tmp.name;
1334         }
1335
1336         if (len != size + num_counters * sizeof(struct xt_counters))
1337                 return -EINVAL;
1338
1339         paddc = vmalloc(len - size);
1340         if (!paddc)
1341                 return -ENOMEM;
1342
1343         if (copy_from_user(paddc, user + size, len - size) != 0) {
1344                 ret = -EFAULT;
1345                 goto free;
1346         }
1347
1348         t = xt_find_table_lock(net, AF_INET, name);
1349         if (IS_ERR_OR_NULL(t)) {
1350                 ret = t ? PTR_ERR(t) : -ENOENT;
1351                 goto free;
1352         }
1353
1354         local_bh_disable();
1355         private = t->private;
1356         if (private->number != num_counters) {
1357                 ret = -EINVAL;
1358                 goto unlock_up_free;
1359         }
1360
1361         i = 0;
1362         /* Choose the copy that is on our node */
1363         curcpu = smp_processor_id();
1364         loc_cpu_entry = private->entries[curcpu];
1365         addend = xt_write_recseq_begin();
1366         xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1367                 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1368                 ++i;
1369         }
1370         xt_write_recseq_end(addend);
1371  unlock_up_free:
1372         local_bh_enable();
1373         xt_table_unlock(t);
1374         module_put(t->me);
1375  free:
1376         vfree(paddc);
1377
1378         return ret;
1379 }
1380
1381 #ifdef CONFIG_COMPAT
1382 struct compat_ipt_replace {
1383         char                    name[XT_TABLE_MAXNAMELEN];
1384         u32                     valid_hooks;
1385         u32                     num_entries;
1386         u32                     size;
1387         u32                     hook_entry[NF_INET_NUMHOOKS];
1388         u32                     underflow[NF_INET_NUMHOOKS];
1389         u32                     num_counters;
1390         compat_uptr_t           counters;       /* struct xt_counters * */
1391         struct compat_ipt_entry entries[0];
1392 };
1393
1394 static int
1395 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1396                           unsigned int *size, struct xt_counters *counters,
1397                           unsigned int i)
1398 {
1399         struct xt_entry_target *t;
1400         struct compat_ipt_entry __user *ce;
1401         u_int16_t target_offset, next_offset;
1402         compat_uint_t origsize;
1403         const struct xt_entry_match *ematch;
1404         int ret = 0;
1405
1406         origsize = *size;
1407         ce = (struct compat_ipt_entry __user *)*dstptr;
1408         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1409             copy_to_user(&ce->counters, &counters[i],
1410             sizeof(counters[i])) != 0)
1411                 return -EFAULT;
1412
1413         *dstptr += sizeof(struct compat_ipt_entry);
1414         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1415
1416         xt_ematch_foreach(ematch, e) {
1417                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1418                 if (ret != 0)
1419                         return ret;
1420         }
1421         target_offset = e->target_offset - (origsize - *size);
1422         t = ipt_get_target(e);
1423         ret = xt_compat_target_to_user(t, dstptr, size);
1424         if (ret)
1425                 return ret;
1426         next_offset = e->next_offset - (origsize - *size);
1427         if (put_user(target_offset, &ce->target_offset) != 0 ||
1428             put_user(next_offset, &ce->next_offset) != 0)
1429                 return -EFAULT;
1430         return 0;
1431 }
1432
1433 static int
1434 compat_find_calc_match(struct xt_entry_match *m,
1435                        const char *name,
1436                        const struct ipt_ip *ip,
1437                        unsigned int hookmask,
1438                        int *size)
1439 {
1440         struct xt_match *match;
1441
1442         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1443                                       m->u.user.revision);
1444         if (IS_ERR(match)) {
1445                 duprintf("compat_check_calc_match: `%s' not found\n",
1446                          m->u.user.name);
1447                 return PTR_ERR(match);
1448         }
1449         m->u.kernel.match = match;
1450         *size += xt_compat_match_offset(match);
1451         return 0;
1452 }
1453
1454 static void compat_release_entry(struct compat_ipt_entry *e)
1455 {
1456         struct xt_entry_target *t;
1457         struct xt_entry_match *ematch;
1458
1459         /* Cleanup all matches */
1460         xt_ematch_foreach(ematch, e)
1461                 module_put(ematch->u.kernel.match->me);
1462         t = compat_ipt_get_target(e);
1463         module_put(t->u.kernel.target->me);
1464 }
1465
1466 static int
1467 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1468                                   struct xt_table_info *newinfo,
1469                                   unsigned int *size,
1470                                   const unsigned char *base,
1471                                   const unsigned char *limit,
1472                                   const unsigned int *hook_entries,
1473                                   const unsigned int *underflows,
1474                                   const char *name)
1475 {
1476         struct xt_entry_match *ematch;
1477         struct xt_entry_target *t;
1478         struct xt_target *target;
1479         unsigned int entry_offset;
1480         unsigned int j;
1481         int ret, off, h;
1482
1483         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1484         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1485             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1486                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1487                 return -EINVAL;
1488         }
1489
1490         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1491                              sizeof(struct compat_xt_entry_target)) {
1492                 duprintf("checking: element %p size %u\n",
1493                          e, e->next_offset);
1494                 return -EINVAL;
1495         }
1496
1497         /* For purposes of check_entry casting the compat entry is fine */
1498         ret = check_entry((struct ipt_entry *)e, name);
1499         if (ret)
1500                 return ret;
1501
1502         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1503         entry_offset = (void *)e - (void *)base;
1504         j = 0;
1505         xt_ematch_foreach(ematch, e) {
1506                 ret = compat_find_calc_match(ematch, name,
1507                                              &e->ip, e->comefrom, &off);
1508                 if (ret != 0)
1509                         goto release_matches;
1510                 ++j;
1511         }
1512
1513         t = compat_ipt_get_target(e);
1514         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1515                                         t->u.user.revision);
1516         if (IS_ERR(target)) {
1517                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1518                          t->u.user.name);
1519                 ret = PTR_ERR(target);
1520                 goto release_matches;
1521         }
1522         t->u.kernel.target = target;
1523
1524         off += xt_compat_target_offset(target);
1525         *size += off;
1526         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1527         if (ret)
1528                 goto out;
1529
1530         /* Check hooks & underflows */
1531         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1532                 if ((unsigned char *)e - base == hook_entries[h])
1533                         newinfo->hook_entry[h] = hook_entries[h];
1534                 if ((unsigned char *)e - base == underflows[h])
1535                         newinfo->underflow[h] = underflows[h];
1536         }
1537
1538         /* Clear counters and comefrom */
1539         memset(&e->counters, 0, sizeof(e->counters));
1540         e->comefrom = 0;
1541         return 0;
1542
1543 out:
1544         module_put(t->u.kernel.target->me);
1545 release_matches:
1546         xt_ematch_foreach(ematch, e) {
1547                 if (j-- == 0)
1548                         break;
1549                 module_put(ematch->u.kernel.match->me);
1550         }
1551         return ret;
1552 }
1553
1554 static int
1555 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1556                             unsigned int *size, const char *name,
1557                             struct xt_table_info *newinfo, unsigned char *base)
1558 {
1559         struct xt_entry_target *t;
1560         struct xt_target *target;
1561         struct ipt_entry *de;
1562         unsigned int origsize;
1563         int ret, h;
1564         struct xt_entry_match *ematch;
1565
1566         ret = 0;
1567         origsize = *size;
1568         de = (struct ipt_entry *)*dstptr;
1569         memcpy(de, e, sizeof(struct ipt_entry));
1570         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1571
1572         *dstptr += sizeof(struct ipt_entry);
1573         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1574
1575         xt_ematch_foreach(ematch, e) {
1576                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1577                 if (ret != 0)
1578                         return ret;
1579         }
1580         de->target_offset = e->target_offset - (origsize - *size);
1581         t = compat_ipt_get_target(e);
1582         target = t->u.kernel.target;
1583         xt_compat_target_from_user(t, dstptr, size);
1584
1585         de->next_offset = e->next_offset - (origsize - *size);
1586         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1587                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1588                         newinfo->hook_entry[h] -= origsize - *size;
1589                 if ((unsigned char *)de - base < newinfo->underflow[h])
1590                         newinfo->underflow[h] -= origsize - *size;
1591         }
1592         return ret;
1593 }
1594
1595 static int
1596 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1597 {
1598         struct xt_entry_match *ematch;
1599         struct xt_mtchk_param mtpar;
1600         unsigned int j;
1601         int ret = 0;
1602
1603         j = 0;
1604         mtpar.net       = net;
1605         mtpar.table     = name;
1606         mtpar.entryinfo = &e->ip;
1607         mtpar.hook_mask = e->comefrom;
1608         mtpar.family    = NFPROTO_IPV4;
1609         xt_ematch_foreach(ematch, e) {
1610                 ret = check_match(ematch, &mtpar);
1611                 if (ret != 0)
1612                         goto cleanup_matches;
1613                 ++j;
1614         }
1615
1616         ret = check_target(e, net, name);
1617         if (ret)
1618                 goto cleanup_matches;
1619         return 0;
1620
1621  cleanup_matches:
1622         xt_ematch_foreach(ematch, e) {
1623                 if (j-- == 0)
1624                         break;
1625                 cleanup_match(ematch, net);
1626         }
1627         return ret;
1628 }
1629
1630 static int
1631 translate_compat_table(struct net *net,
1632                        const char *name,
1633                        unsigned int valid_hooks,
1634                        struct xt_table_info **pinfo,
1635                        void **pentry0,
1636                        unsigned int total_size,
1637                        unsigned int number,
1638                        unsigned int *hook_entries,
1639                        unsigned int *underflows)
1640 {
1641         unsigned int i, j;
1642         struct xt_table_info *newinfo, *info;
1643         void *pos, *entry0, *entry1;
1644         struct compat_ipt_entry *iter0;
1645         struct ipt_entry *iter1;
1646         unsigned int size;
1647         int ret;
1648
1649         info = *pinfo;
1650         entry0 = *pentry0;
1651         size = total_size;
1652         info->number = number;
1653
1654         /* Init all hooks to impossible value. */
1655         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1656                 info->hook_entry[i] = 0xFFFFFFFF;
1657                 info->underflow[i] = 0xFFFFFFFF;
1658         }
1659
1660         duprintf("translate_compat_table: size %u\n", info->size);
1661         j = 0;
1662         xt_compat_lock(AF_INET);
1663         xt_compat_init_offsets(AF_INET, number);
1664         /* Walk through entries, checking offsets. */
1665         xt_entry_foreach(iter0, entry0, total_size) {
1666                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1667                                                         entry0,
1668                                                         entry0 + total_size,
1669                                                         hook_entries,
1670                                                         underflows,
1671                                                         name);
1672                 if (ret != 0)
1673                         goto out_unlock;
1674                 ++j;
1675         }
1676
1677         ret = -EINVAL;
1678         if (j != number) {
1679                 duprintf("translate_compat_table: %u not %u entries\n",
1680                          j, number);
1681                 goto out_unlock;
1682         }
1683
1684         /* Check hooks all assigned */
1685         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1686                 /* Only hooks which are valid */
1687                 if (!(valid_hooks & (1 << i)))
1688                         continue;
1689                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1690                         duprintf("Invalid hook entry %u %u\n",
1691                                  i, hook_entries[i]);
1692                         goto out_unlock;
1693                 }
1694                 if (info->underflow[i] == 0xFFFFFFFF) {
1695                         duprintf("Invalid underflow %u %u\n",
1696                                  i, underflows[i]);
1697                         goto out_unlock;
1698                 }
1699         }
1700
1701         ret = -ENOMEM;
1702         newinfo = xt_alloc_table_info(size);
1703         if (!newinfo)
1704                 goto out_unlock;
1705
1706         newinfo->number = number;
1707         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1708                 newinfo->hook_entry[i] = info->hook_entry[i];
1709                 newinfo->underflow[i] = info->underflow[i];
1710         }
1711         entry1 = newinfo->entries[raw_smp_processor_id()];
1712         pos = entry1;
1713         size = total_size;
1714         xt_entry_foreach(iter0, entry0, total_size) {
1715                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1716                                                   name, newinfo, entry1);
1717                 if (ret != 0)
1718                         break;
1719         }
1720         xt_compat_flush_offsets(AF_INET);
1721         xt_compat_unlock(AF_INET);
1722         if (ret)
1723                 goto free_newinfo;
1724
1725         ret = -ELOOP;
1726         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1727                 goto free_newinfo;
1728
1729         i = 0;
1730         xt_entry_foreach(iter1, entry1, newinfo->size) {
1731                 ret = compat_check_entry(iter1, net, name);
1732                 if (ret != 0)
1733                         break;
1734                 ++i;
1735                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1736                     XT_ERROR_TARGET) == 0)
1737                         ++newinfo->stacksize;
1738         }
1739         if (ret) {
1740                 /*
1741                  * The first i matches need cleanup_entry (calls ->destroy)
1742                  * because they had called ->check already. The other j-i
1743                  * entries need only release.
1744                  */
1745                 int skip = i;
1746                 j -= i;
1747                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1748                         if (skip-- > 0)
1749                                 continue;
1750                         if (j-- == 0)
1751                                 break;
1752                         compat_release_entry(iter0);
1753                 }
1754                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1755                         if (i-- == 0)
1756                                 break;
1757                         cleanup_entry(iter1, net);
1758                 }
1759                 xt_free_table_info(newinfo);
1760                 return ret;
1761         }
1762
1763         /* And one copy for every other CPU */
1764         for_each_possible_cpu(i)
1765                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1766                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1767
1768         *pinfo = newinfo;
1769         *pentry0 = entry1;
1770         xt_free_table_info(info);
1771         return 0;
1772
1773 free_newinfo:
1774         xt_free_table_info(newinfo);
1775 out:
1776         xt_entry_foreach(iter0, entry0, total_size) {
1777                 if (j-- == 0)
1778                         break;
1779                 compat_release_entry(iter0);
1780         }
1781         return ret;
1782 out_unlock:
1783         xt_compat_flush_offsets(AF_INET);
1784         xt_compat_unlock(AF_INET);
1785         goto out;
1786 }
1787
1788 static int
1789 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1790 {
1791         int ret;
1792         struct compat_ipt_replace tmp;
1793         struct xt_table_info *newinfo;
1794         void *loc_cpu_entry;
1795         struct ipt_entry *iter;
1796
1797         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1798                 return -EFAULT;
1799
1800         /* overflow check */
1801         if (tmp.size >= INT_MAX / num_possible_cpus())
1802                 return -ENOMEM;
1803         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1804                 return -ENOMEM;
1805         tmp.name[sizeof(tmp.name)-1] = 0;
1806
1807         newinfo = xt_alloc_table_info(tmp.size);
1808         if (!newinfo)
1809                 return -ENOMEM;
1810
1811         /* choose the copy that is on our node/cpu */
1812         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1813         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1814                            tmp.size) != 0) {
1815                 ret = -EFAULT;
1816                 goto free_newinfo;
1817         }
1818
1819         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1820                                      &newinfo, &loc_cpu_entry, tmp.size,
1821                                      tmp.num_entries, tmp.hook_entry,
1822                                      tmp.underflow);
1823         if (ret != 0)
1824                 goto free_newinfo;
1825
1826         duprintf("compat_do_replace: Translated table\n");
1827
1828         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1829                            tmp.num_counters, compat_ptr(tmp.counters));
1830         if (ret)
1831                 goto free_newinfo_untrans;
1832         return 0;
1833
1834  free_newinfo_untrans:
1835         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1836                 cleanup_entry(iter, net);
1837  free_newinfo:
1838         xt_free_table_info(newinfo);
1839         return ret;
1840 }
1841
1842 static int
1843 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1844                       unsigned int len)
1845 {
1846         int ret;
1847
1848         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1849                 return -EPERM;
1850
1851         switch (cmd) {
1852         case IPT_SO_SET_REPLACE:
1853                 ret = compat_do_replace(sock_net(sk), user, len);
1854                 break;
1855
1856         case IPT_SO_SET_ADD_COUNTERS:
1857                 ret = do_add_counters(sock_net(sk), user, len, 1);
1858                 break;
1859
1860         default:
1861                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1862                 ret = -EINVAL;
1863         }
1864
1865         return ret;
1866 }
1867
1868 struct compat_ipt_get_entries {
1869         char name[XT_TABLE_MAXNAMELEN];
1870         compat_uint_t size;
1871         struct compat_ipt_entry entrytable[0];
1872 };
1873
1874 static int
1875 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1876                             void __user *userptr)
1877 {
1878         struct xt_counters *counters;
1879         const struct xt_table_info *private = table->private;
1880         void __user *pos;
1881         unsigned int size;
1882         int ret = 0;
1883         const void *loc_cpu_entry;
1884         unsigned int i = 0;
1885         struct ipt_entry *iter;
1886
1887         counters = alloc_counters(table);
1888         if (IS_ERR(counters))
1889                 return PTR_ERR(counters);
1890
1891         /* choose the copy that is on our node/cpu, ...
1892          * This choice is lazy (because current thread is
1893          * allowed to migrate to another cpu)
1894          */
1895         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1896         pos = userptr;
1897         size = total_size;
1898         xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1899                 ret = compat_copy_entry_to_user(iter, &pos,
1900                                                 &size, counters, i++);
1901                 if (ret != 0)
1902                         break;
1903         }
1904
1905         vfree(counters);
1906         return ret;
1907 }
1908
1909 static int
1910 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1911                    int *len)
1912 {
1913         int ret;
1914         struct compat_ipt_get_entries get;
1915         struct xt_table *t;
1916
1917         if (*len < sizeof(get)) {
1918                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1919                 return -EINVAL;
1920         }
1921
1922         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1923                 return -EFAULT;
1924
1925         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1926                 duprintf("compat_get_entries: %u != %zu\n",
1927                          *len, sizeof(get) + get.size);
1928                 return -EINVAL;
1929         }
1930
1931         xt_compat_lock(AF_INET);
1932         t = xt_find_table_lock(net, AF_INET, get.name);
1933         if (!IS_ERR_OR_NULL(t)) {
1934                 const struct xt_table_info *private = t->private;
1935                 struct xt_table_info info;
1936                 duprintf("t->private->number = %u\n", private->number);
1937                 ret = compat_table_info(private, &info);
1938                 if (!ret && get.size == info.size) {
1939                         ret = compat_copy_entries_to_user(private->size,
1940                                                           t, uptr->entrytable);
1941                 } else if (!ret) {
1942                         duprintf("compat_get_entries: I've got %u not %u!\n",
1943                                  private->size, get.size);
1944                         ret = -EAGAIN;
1945                 }
1946                 xt_compat_flush_offsets(AF_INET);
1947                 module_put(t->me);
1948                 xt_table_unlock(t);
1949         } else
1950                 ret = t ? PTR_ERR(t) : -ENOENT;
1951
1952         xt_compat_unlock(AF_INET);
1953         return ret;
1954 }
1955
1956 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1957
1958 static int
1959 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1960 {
1961         int ret;
1962
1963         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1964                 return -EPERM;
1965
1966         switch (cmd) {
1967         case IPT_SO_GET_INFO:
1968                 ret = get_info(sock_net(sk), user, len, 1);
1969                 break;
1970         case IPT_SO_GET_ENTRIES:
1971                 ret = compat_get_entries(sock_net(sk), user, len);
1972                 break;
1973         default:
1974                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1975         }
1976         return ret;
1977 }
1978 #endif
1979
1980 static int
1981 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1982 {
1983         int ret;
1984
1985         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1986                 return -EPERM;
1987
1988         switch (cmd) {
1989         case IPT_SO_SET_REPLACE:
1990                 ret = do_replace(sock_net(sk), user, len);
1991                 break;
1992
1993         case IPT_SO_SET_ADD_COUNTERS:
1994                 ret = do_add_counters(sock_net(sk), user, len, 0);
1995                 break;
1996
1997         default:
1998                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1999                 ret = -EINVAL;
2000         }
2001
2002         return ret;
2003 }
2004
2005 static int
2006 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2007 {
2008         int ret;
2009
2010         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2011                 return -EPERM;
2012
2013         switch (cmd) {
2014         case IPT_SO_GET_INFO:
2015                 ret = get_info(sock_net(sk), user, len, 0);
2016                 break;
2017
2018         case IPT_SO_GET_ENTRIES:
2019                 ret = get_entries(sock_net(sk), user, len);
2020                 break;
2021
2022         case IPT_SO_GET_REVISION_MATCH:
2023         case IPT_SO_GET_REVISION_TARGET: {
2024                 struct xt_get_revision rev;
2025                 int target;
2026
2027                 if (*len != sizeof(rev)) {
2028                         ret = -EINVAL;
2029                         break;
2030                 }
2031                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2032                         ret = -EFAULT;
2033                         break;
2034                 }
2035                 rev.name[sizeof(rev.name)-1] = 0;
2036
2037                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2038                         target = 1;
2039                 else
2040                         target = 0;
2041
2042                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2043                                                          rev.revision,
2044                                                          target, &ret),
2045                                         "ipt_%s", rev.name);
2046                 break;
2047         }
2048
2049         default:
2050                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2051                 ret = -EINVAL;
2052         }
2053
2054         return ret;
2055 }
2056
2057 struct xt_table *ipt_register_table(struct net *net,
2058                                     const struct xt_table *table,
2059                                     const struct ipt_replace *repl)
2060 {
2061         int ret;
2062         struct xt_table_info *newinfo;
2063         struct xt_table_info bootstrap = {0};
2064         void *loc_cpu_entry;
2065         struct xt_table *new_table;
2066
2067         newinfo = xt_alloc_table_info(repl->size);
2068         if (!newinfo) {
2069                 ret = -ENOMEM;
2070                 goto out;
2071         }
2072
2073         /* choose the copy on our node/cpu, but dont care about preemption */
2074         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2075         memcpy(loc_cpu_entry, repl->entries, repl->size);
2076
2077         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2078         if (ret != 0)
2079                 goto out_free;
2080
2081         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2082         if (IS_ERR(new_table)) {
2083                 ret = PTR_ERR(new_table);
2084                 goto out_free;
2085         }
2086
2087         return new_table;
2088
2089 out_free:
2090         xt_free_table_info(newinfo);
2091 out:
2092         return ERR_PTR(ret);
2093 }
2094
2095 void ipt_unregister_table(struct net *net, struct xt_table *table)
2096 {
2097         struct xt_table_info *private;
2098         void *loc_cpu_entry;
2099         struct module *table_owner = table->me;
2100         struct ipt_entry *iter;
2101
2102         private = xt_unregister_table(table);
2103
2104         /* Decrease module usage counts and free resources */
2105         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2106         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2107                 cleanup_entry(iter, net);
2108         if (private->number > private->initial_entries)
2109                 module_put(table_owner);
2110         xt_free_table_info(private);
2111 }
2112
2113 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2114 static inline bool
2115 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2116                      u_int8_t type, u_int8_t code,
2117                      bool invert)
2118 {
2119         return ((test_type == 0xFF) ||
2120                 (type == test_type && code >= min_code && code <= max_code))
2121                 ^ invert;
2122 }
2123
2124 static bool
2125 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2126 {
2127         const struct icmphdr *ic;
2128         struct icmphdr _icmph;
2129         const struct ipt_icmp *icmpinfo = par->matchinfo;
2130
2131         /* Must not be a fragment. */
2132         if (par->fragoff != 0)
2133                 return false;
2134
2135         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2136         if (ic == NULL) {
2137                 /* We've been asked to examine this packet, and we
2138                  * can't.  Hence, no choice but to drop.
2139                  */
2140                 duprintf("Dropping evil ICMP tinygram.\n");
2141                 par->hotdrop = true;
2142                 return false;
2143         }
2144
2145         return icmp_type_code_match(icmpinfo->type,
2146                                     icmpinfo->code[0],
2147                                     icmpinfo->code[1],
2148                                     ic->type, ic->code,
2149                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2150 }
2151
2152 static int icmp_checkentry(const struct xt_mtchk_param *par)
2153 {
2154         const struct ipt_icmp *icmpinfo = par->matchinfo;
2155
2156         /* Must specify no unknown invflags */
2157         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2158 }
2159
2160 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2161         {
2162                 .name             = XT_STANDARD_TARGET,
2163                 .targetsize       = sizeof(int),
2164                 .family           = NFPROTO_IPV4,
2165 #ifdef CONFIG_COMPAT
2166                 .compatsize       = sizeof(compat_int_t),
2167                 .compat_from_user = compat_standard_from_user,
2168                 .compat_to_user   = compat_standard_to_user,
2169 #endif
2170         },
2171         {
2172                 .name             = XT_ERROR_TARGET,
2173                 .target           = ipt_error,
2174                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2175                 .family           = NFPROTO_IPV4,
2176         },
2177 };
2178
2179 static struct nf_sockopt_ops ipt_sockopts = {
2180         .pf             = PF_INET,
2181         .set_optmin     = IPT_BASE_CTL,
2182         .set_optmax     = IPT_SO_SET_MAX+1,
2183         .set            = do_ipt_set_ctl,
2184 #ifdef CONFIG_COMPAT
2185         .compat_set     = compat_do_ipt_set_ctl,
2186 #endif
2187         .get_optmin     = IPT_BASE_CTL,
2188         .get_optmax     = IPT_SO_GET_MAX+1,
2189         .get            = do_ipt_get_ctl,
2190 #ifdef CONFIG_COMPAT
2191         .compat_get     = compat_do_ipt_get_ctl,
2192 #endif
2193         .owner          = THIS_MODULE,
2194 };
2195
2196 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2197         {
2198                 .name       = "icmp",
2199                 .match      = icmp_match,
2200                 .matchsize  = sizeof(struct ipt_icmp),
2201                 .checkentry = icmp_checkentry,
2202                 .proto      = IPPROTO_ICMP,
2203                 .family     = NFPROTO_IPV4,
2204         },
2205 };
2206
2207 static int __net_init ip_tables_net_init(struct net *net)
2208 {
2209         return xt_proto_init(net, NFPROTO_IPV4);
2210 }
2211
2212 static void __net_exit ip_tables_net_exit(struct net *net)
2213 {
2214         xt_proto_fini(net, NFPROTO_IPV4);
2215 }
2216
2217 static struct pernet_operations ip_tables_net_ops = {
2218         .init = ip_tables_net_init,
2219         .exit = ip_tables_net_exit,
2220 };
2221
2222 static int __init ip_tables_init(void)
2223 {
2224         int ret;
2225
2226         ret = register_pernet_subsys(&ip_tables_net_ops);
2227         if (ret < 0)
2228                 goto err1;
2229
2230         /* No one else will be downing sem now, so we won't sleep */
2231         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2232         if (ret < 0)
2233                 goto err2;
2234         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2235         if (ret < 0)
2236                 goto err4;
2237
2238         /* Register setsockopt */
2239         ret = nf_register_sockopt(&ipt_sockopts);
2240         if (ret < 0)
2241                 goto err5;
2242
2243         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2244         return 0;
2245
2246 err5:
2247         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2248 err4:
2249         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2250 err2:
2251         unregister_pernet_subsys(&ip_tables_net_ops);
2252 err1:
2253         return ret;
2254 }
2255
2256 static void __exit ip_tables_fini(void)
2257 {
2258         nf_unregister_sockopt(&ipt_sockopts);
2259
2260         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2261         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2262         unregister_pernet_subsys(&ip_tables_net_ops);
2263 }
2264
2265 EXPORT_SYMBOL(ipt_register_table);
2266 EXPORT_SYMBOL(ipt_unregister_table);
2267 EXPORT_SYMBOL(ipt_do_table);
2268 module_init(ip_tables_init);
2269 module_exit(ip_tables_fini);