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(const struct sk_buff *skb,
 250                          unsigned int hook,
 251                          const struct net_device *in,
 252                          const struct net_device *out,
 253                          const char *tablename,
 254                          const struct xt_table_info *private,
 255                          const struct ipt_entry *e)
 256 {
 257         const void *table_base;
 258         const struct ipt_entry *root;
 259         const char *hookname, *chainname, *comment;
 260         const struct ipt_entry *iter;
 261         unsigned int rulenum = 0;
 262         struct net *net = dev_net(in ? in : out);
 263
 264         table_base = private->entries[smp_processor_id()];
 265         root = get_entry(table_base, private->hook_entry[hook]);
 266
 267         hookname = chainname = hooknames[hook];
 268         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 269
 270         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 271                 if (get_chainname_rulenum(iter, e, hookname,
 272                     &chainname, &comment, &rulenum) != 0)
 273                         break;
 274
 275         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 276                      "TRACE: %s:%s:%s:%u ",
 277                      tablename, chainname, comment, rulenum);
 278 }
 279 #endif
 280
 281 static inline __pure
 282 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 283 {
 284         return (void *)entry + entry->next_offset;
 285 }
 286
 287 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 288 unsigned int
 289 ipt_do_table(struct sk_buff *skb,
 290              unsigned int hook,
 291              const struct nf_hook_state *state,
 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 = state->in ? state->in->name : nulldevname;
 309         outdev = state->out ? state->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      = 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[cpu];
 335         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 336         stackptr   = per_cpu_ptr(private->stackptr, cpu);
 337         origptr    = *stackptr;
 338
 339         e = get_entry(table_base, private->hook_entry[hook]);
 340
 341         pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
 342                  table->name, hook, origptr,
 343                  get_entry(table_base, private->underflow[hook]));
 344
 345         do {
 346                 const struct xt_entry_target *t;
 347                 const struct xt_entry_match *ematch;
 348
 349                 IP_NF_ASSERT(e);
 350                 if (!ip_packet_match(ip, indev, outdev,
 351                     &e->ip, acpar.fragoff)) {
 352  no_match:
 353                         e = ipt_next_entry(e);
 354                         continue;
 355                 }
 356
 357                 xt_ematch_foreach(ematch, e) {
 358                         acpar.match     = ematch->u.kernel.match;
 359                         acpar.matchinfo = ematch->data;
 360                         if (!acpar.match->match(skb, &acpar))
 361                                 goto no_match;
 362                 }
 363
 364                 ADD_COUNTER(e->counters, skb->len, 1);
 365
 366                 t = ipt_get_target(e);
 367                 IP_NF_ASSERT(t->u.kernel.target);
 368
 369 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 370                 /* The packet is traced: log it */
 371                 if (unlikely(skb->nf_trace))
 372                         trace_packet(skb, hook, state->in, state->out,
 373                                      table->name, private, e);
 374 #endif
 375                 /* Standard target? */
 376                 if (!t->u.kernel.target->target) {
 377                         int v;
 378
 379                         v = ((struct xt_standard_target *)t)->verdict;
 380                         if (v < 0) {
 381                                 /* Pop from stack? */
 382                                 if (v != XT_RETURN) {
 383                                         verdict = (unsigned int)(-v) - 1;
 384                                         break;
 385                                 }
 386                                 if (*stackptr <= origptr) {
 387                                         e = get_entry(table_base,
 388                                             private->underflow[hook]);
 389                                         pr_debug("Underflow (this is normal) "
 390                                                  "to %p\n", e);
 391                                 } else {
 392                                         e = jumpstack[--*stackptr];
 393                                         pr_debug("Pulled %p out from pos %u\n",
 394                                                  e, *stackptr);
 395                                         e = ipt_next_entry(e);
 396                                 }
 397                                 continue;
 398                         }
 399                         if (table_base + v != ipt_next_entry(e) &&
 400                             !(e->ip.flags & IPT_F_GOTO)) {
 401                                 if (*stackptr >= private->stacksize) {
 402                                         verdict = NF_DROP;
 403                                         break;
 404                                 }
 405                                 jumpstack[(*stackptr)++] = e;
 406                                 pr_debug("Pushed %p into pos %u\n",
 407                                          e, *stackptr - 1);
 408                         }
 409
 410                         e = get_entry(table_base, v);
 411                         continue;
 412                 }
 413
 414                 acpar.target   = t->u.kernel.target;
 415                 acpar.targinfo = t->data;
 416
 417                 verdict = t->u.kernel.target->target(skb, &acpar);
 418                 /* Target might have changed stuff. */
 419                 ip = ip_hdr(skb);
 420                 if (verdict == XT_CONTINUE)
 421                         e = ipt_next_entry(e);
 422                 else
 423                         /* Verdict */
 424                         break;
 425         } while (!acpar.hotdrop);
 426         pr_debug("Exiting %s; resetting sp from %u to %u\n",
 427                  __func__, *stackptr, origptr);
 428         *stackptr = origptr;
 429         xt_write_recseq_end(addend);
 430         local_bh_enable();
 431
 432 #ifdef DEBUG_ALLOW_ALL
 433         return NF_ACCEPT;
 434 #else
 435         if (acpar.hotdrop)
 436                 return NF_DROP;
 437         else return verdict;
 438 #endif
 439 }
 440
 441 /* Figures out from what hook each rule can be called: returns 0 if
 442    there are loops.  Puts hook bitmask in comefrom. */
 443 static int
 444 mark_source_chains(const struct xt_table_info *newinfo,
 445                    unsigned int valid_hooks, void *entry0)
 446 {
 447         unsigned int hook;
 448
 449         /* No recursion; use packet counter to save back ptrs (reset
 450            to 0 as we leave), and comefrom to save source hook bitmask */
 451         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 452                 unsigned int pos = newinfo->hook_entry[hook];
 453                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 454
 455                 if (!(valid_hooks & (1 << hook)))
 456                         continue;
 457
 458                 /* Set initial back pointer. */
 459                 e->counters.pcnt = pos;
 460
 461                 for (;;) {
 462                         const struct xt_standard_target *t
 463                                 = (void *)ipt_get_target_c(e);
 464                         int visited = e->comefrom & (1 << hook);
 465
 466                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 467                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 468                                        hook, pos, e->comefrom);
 469                                 return 0;
 470                         }
 471                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 472
 473                         /* Unconditional return/END. */
 474                         if ((e->target_offset == sizeof(struct ipt_entry) &&
 475                              (strcmp(t->target.u.user.name,
 476                                      XT_STANDARD_TARGET) == 0) &&
 477                              t->verdict < 0 && unconditional(&e->ip)) ||
 478                             visited) {
 479                                 unsigned int oldpos, size;
 480
 481                                 if ((strcmp(t->target.u.user.name,
 482                                             XT_STANDARD_TARGET) == 0) &&
 483                                     t->verdict < -NF_MAX_VERDICT - 1) {
 484                                         duprintf("mark_source_chains: bad "
 485                                                 "negative verdict (%i)\n",
 486                                                                 t->verdict);
 487                                         return 0;
 488                                 }
 489
 490                                 /* Return: backtrack through the last
 491                                    big jump. */
 492                                 do {
 493                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 494 #ifdef DEBUG_IP_FIREWALL_USER
 495                                         if (e->comefrom
 496                                             & (1 << NF_INET_NUMHOOKS)) {
 497                                                 duprintf("Back unset "
 498                                                          "on hook %u "
 499                                                          "rule %u\n",
 500                                                          hook, pos);
 501                                         }
 502 #endif
 503                                         oldpos = pos;
 504                                         pos = e->counters.pcnt;
 505                                         e->counters.pcnt = 0;
 506
 507                                         /* We're at the start. */
 508                                         if (pos == oldpos)
 509                                                 goto next;
 510
 511                                         e = (struct ipt_entry *)
 512                                                 (entry0 + pos);
 513                                 } while (oldpos == pos + e->next_offset);
 514
 515                                 /* Move along one */
 516                                 size = e->next_offset;
 517                                 e = (struct ipt_entry *)
 518                                         (entry0 + pos + size);
 519                                 e->counters.pcnt = pos;
 520                                 pos += size;
 521                         } else {
 522                                 int newpos = t->verdict;
 523
 524                                 if (strcmp(t->target.u.user.name,
 525                                            XT_STANDARD_TARGET) == 0 &&
 526                                     newpos >= 0) {
 527                                         if (newpos > newinfo->size -
 528                                                 sizeof(struct ipt_entry)) {
 529                                                 duprintf("mark_source_chains: "
 530                                                         "bad verdict (%i)\n",
 531                                                                 newpos);
 532                                                 return 0;
 533                                         }
 534                                         /* This a jump; chase it. */
 535                                         duprintf("Jump rule %u -> %u\n",
 536                                                  pos, newpos);
 537                                 } else {
 538                                         /* ... this is a fallthru */
 539                                         newpos = pos + e->next_offset;
 540                                 }
 541                                 e = (struct ipt_entry *)
 542                                         (entry0 + newpos);
 543                                 e->counters.pcnt = pos;
 544                                 pos = newpos;
 545                         }
 546                 }
 547                 next:
 548                 duprintf("Finished chain %u\n", hook);
 549         }
 550         return 1;
 551 }
 552
 553 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 554 {
 555         struct xt_mtdtor_param par;
 556
 557         par.net       = net;
 558         par.match     = m->u.kernel.match;
 559         par.matchinfo = m->data;
 560         par.family    = NFPROTO_IPV4;
 561         if (par.match->destroy != NULL)
 562                 par.match->destroy(&par);
 563         module_put(par.match->me);
 564 }
 565
 566 static int
 567 check_entry(const struct ipt_entry *e, const char *name)
 568 {
 569         const struct xt_entry_target *t;
 570
 571         if (!ip_checkentry(&e->ip)) {
 572                 duprintf("ip check failed %p %s.\n", e, name);
 573                 return -EINVAL;
 574         }
 575
 576         if (e->target_offset + sizeof(struct xt_entry_target) >
 577             e->next_offset)
 578                 return -EINVAL;
 579
 580         t = ipt_get_target_c(e);
 581         if (e->target_offset + t->u.target_size > e->next_offset)
 582                 return -EINVAL;
 583
 584         return 0;
 585 }
 586
 587 static int
 588 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 589 {
 590         const struct ipt_ip *ip = par->entryinfo;
 591         int ret;
 592
 593         par->match     = m->u.kernel.match;
 594         par->matchinfo = m->data;
 595
 596         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 597               ip->proto, ip->invflags & IPT_INV_PROTO);
 598         if (ret < 0) {
 599                 duprintf("check failed for `%s'.\n", par->match->name);
 600                 return ret;
 601         }
 602         return 0;
 603 }
 604
 605 static int
 606 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 607 {
 608         struct xt_match *match;
 609         int ret;
 610
 611         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 612                                       m->u.user.revision);
 613         if (IS_ERR(match)) {
 614                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 615                 return PTR_ERR(match);
 616         }
 617         m->u.kernel.match = match;
 618
 619         ret = check_match(m, par);
 620         if (ret)
 621                 goto err;
 622
 623         return 0;
 624 err:
 625         module_put(m->u.kernel.match->me);
 626         return ret;
 627 }
 628
 629 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 630 {
 631         struct xt_entry_target *t = ipt_get_target(e);
 632         struct xt_tgchk_param par = {
 633                 .net       = net,
 634                 .table     = name,
 635                 .entryinfo = e,
 636                 .target    = t->u.kernel.target,
 637                 .targinfo  = t->data,
 638                 .hook_mask = e->comefrom,
 639                 .family    = NFPROTO_IPV4,
 640         };
 641         int ret;
 642
 643         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 644               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 645         if (ret < 0) {
 646                 duprintf("check failed for `%s'.\n",
 647                          t->u.kernel.target->name);
 648                 return ret;
 649         }
 650         return 0;
 651 }
 652
 653 static int
 654 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 655                  unsigned int size)
 656 {
 657         struct xt_entry_target *t;
 658         struct xt_target *target;
 659         int ret;
 660         unsigned int j;
 661         struct xt_mtchk_param mtpar;
 662         struct xt_entry_match *ematch;
 663
 664         ret = check_entry(e, name);
 665         if (ret)
 666                 return ret;
 667
 668         j = 0;
 669         mtpar.net       = net;
 670         mtpar.table     = name;
 671         mtpar.entryinfo = &e->ip;
 672         mtpar.hook_mask = e->comefrom;
 673         mtpar.family    = NFPROTO_IPV4;
 674         xt_ematch_foreach(ematch, e) {
 675                 ret = find_check_match(ematch, &mtpar);
 676                 if (ret != 0)
 677                         goto cleanup_matches;
 678                 ++j;
 679         }
 680
 681         t = ipt_get_target(e);
 682         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 683                                         t->u.user.revision);
 684         if (IS_ERR(target)) {
 685                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 686                 ret = PTR_ERR(target);
 687                 goto cleanup_matches;
 688         }
 689         t->u.kernel.target = target;
 690
 691         ret = check_target(e, net, name);
 692         if (ret)
 693                 goto err;
 694         return 0;
 695  err:
 696         module_put(t->u.kernel.target->me);
 697  cleanup_matches:
 698         xt_ematch_foreach(ematch, e) {
 699                 if (j-- == 0)
 700                         break;
 701                 cleanup_match(ematch, net);
 702         }
 703         return ret;
 704 }
 705
 706 static bool check_underflow(const struct ipt_entry *e)
 707 {
 708         const struct xt_entry_target *t;
 709         unsigned int verdict;
 710
 711         if (!unconditional(&e->ip))
 712                 return false;
 713         t = ipt_get_target_c(e);
 714         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 715                 return false;
 716         verdict = ((struct xt_standard_target *)t)->verdict;
 717         verdict = -verdict - 1;
 718         return verdict == NF_DROP || verdict == NF_ACCEPT;
 719 }
 720
 721 static int
 722 check_entry_size_and_hooks(struct ipt_entry *e,
 723                            struct xt_table_info *newinfo,
 724                            const unsigned char *base,
 725                            const unsigned char *limit,
 726                            const unsigned int *hook_entries,
 727                            const unsigned int *underflows,
 728                            unsigned int valid_hooks)
 729 {
 730         unsigned int h;
 731
 732         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 733             (unsigned char *)e + sizeof(struct ipt_entry) >= limit) {
 734                 duprintf("Bad offset %p\n", e);
 735                 return -EINVAL;
 736         }
 737
 738         if (e->next_offset
 739             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 740                 duprintf("checking: element %p size %u\n",
 741                          e, e->next_offset);
 742                 return -EINVAL;
 743         }
 744
 745         /* Check hooks & underflows */
 746         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 747                 if (!(valid_hooks & (1 << h)))
 748                         continue;
 749                 if ((unsigned char *)e - base == hook_entries[h])
 750                         newinfo->hook_entry[h] = hook_entries[h];
 751                 if ((unsigned char *)e - base == underflows[h]) {
 752                         if (!check_underflow(e)) {
 753                                 pr_err("Underflows must be unconditional and "
 754                                        "use the STANDARD target with "
 755                                        "ACCEPT/DROP\n");
 756                                 return -EINVAL;
 757                         }
 758                         newinfo->underflow[h] = underflows[h];
 759                 }
 760         }
 761
 762         /* Clear counters and comefrom */
 763         e->counters = ((struct xt_counters) { 0, 0 });
 764         e->comefrom = 0;
 765         return 0;
 766 }
 767
 768 static void
 769 cleanup_entry(struct ipt_entry *e, struct net *net)
 770 {
 771         struct xt_tgdtor_param par;
 772         struct xt_entry_target *t;
 773         struct xt_entry_match *ematch;
 774
 775         /* Cleanup all matches */
 776         xt_ematch_foreach(ematch, e)
 777                 cleanup_match(ematch, net);
 778         t = ipt_get_target(e);
 779
 780         par.net      = net;
 781         par.target   = t->u.kernel.target;
 782         par.targinfo = t->data;
 783         par.family   = NFPROTO_IPV4;
 784         if (par.target->destroy != NULL)
 785                 par.target->destroy(&par);
 786         module_put(par.target->me);
 787 }
 788
 789 /* Checks and translates the user-supplied table segment (held in
 790    newinfo) */
 791 static int
 792 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 793                 const struct ipt_replace *repl)
 794 {
 795         struct ipt_entry *iter;
 796         unsigned int i;
 797         int ret = 0;
 798
 799         newinfo->size = repl->size;
 800         newinfo->number = repl->num_entries;
 801
 802         /* Init all hooks to impossible value. */
 803         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 804                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 805                 newinfo->underflow[i] = 0xFFFFFFFF;
 806         }
 807
 808         duprintf("translate_table: size %u\n", newinfo->size);
 809         i = 0;
 810         /* Walk through entries, checking offsets. */
 811         xt_entry_foreach(iter, entry0, newinfo->size) {
 812                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 813                                                  entry0 + repl->size,
 814                                                  repl->hook_entry,
 815                                                  repl->underflow,
 816                                                  repl->valid_hooks);
 817                 if (ret != 0)
 818                         return ret;
 819                 ++i;
 820                 if (strcmp(ipt_get_target(iter)->u.user.name,
 821                     XT_ERROR_TARGET) == 0)
 822                         ++newinfo->stacksize;
 823         }
 824
 825         if (i != repl->num_entries) {
 826                 duprintf("translate_table: %u not %u entries\n",
 827                          i, repl->num_entries);
 828                 return -EINVAL;
 829         }
 830
 831         /* Check hooks all assigned */
 832         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 833                 /* Only hooks which are valid */
 834                 if (!(repl->valid_hooks & (1 << i)))
 835                         continue;
 836                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 837                         duprintf("Invalid hook entry %u %u\n",
 838                                  i, repl->hook_entry[i]);
 839                         return -EINVAL;
 840                 }
 841                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 842                         duprintf("Invalid underflow %u %u\n",
 843                                  i, repl->underflow[i]);
 844                         return -EINVAL;
 845                 }
 846         }
 847
 848         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 849                 return -ELOOP;
 850
 851         /* Finally, each sanity check must pass */
 852         i = 0;
 853         xt_entry_foreach(iter, entry0, newinfo->size) {
 854                 ret = find_check_entry(iter, net, repl->name, repl->size);
 855                 if (ret != 0)
 856                         break;
 857                 ++i;
 858         }
 859
 860         if (ret != 0) {
 861                 xt_entry_foreach(iter, entry0, newinfo->size) {
 862                         if (i-- == 0)
 863                                 break;
 864                         cleanup_entry(iter, net);
 865                 }
 866                 return ret;
 867         }
 868
 869         /* And one copy for every other CPU */
 870         for_each_possible_cpu(i) {
 871                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
 872                         memcpy(newinfo->entries[i], entry0, newinfo->size);
 873         }
 874
 875         return ret;
 876 }
 877
 878 static void
 879 get_counters(const struct xt_table_info *t,
 880              struct xt_counters counters[])
 881 {
 882         struct ipt_entry *iter;
 883         unsigned int cpu;
 884         unsigned int i;
 885
 886         for_each_possible_cpu(cpu) {
 887                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 888
 889                 i = 0;
 890                 xt_entry_foreach(iter, t->entries[cpu], t->size) {
 891                         u64 bcnt, pcnt;
 892                         unsigned int start;
 893
 894                         do {
 895                                 start = read_seqcount_begin(s);
 896                                 bcnt = iter->counters.bcnt;
 897                                 pcnt = iter->counters.pcnt;
 898                         } while (read_seqcount_retry(s, start));
 899
 900                         ADD_COUNTER(counters[i], bcnt, pcnt);
 901                         ++i; /* macro does multi eval of i */
 902                 }
 903         }
 904 }
 905
 906 static struct xt_counters *alloc_counters(const struct xt_table *table)
 907 {
 908         unsigned int countersize;
 909         struct xt_counters *counters;
 910         const struct xt_table_info *private = table->private;
 911
 912         /* We need atomic snapshot of counters: rest doesn't change
 913            (other than comefrom, which userspace doesn't care
 914            about). */
 915         countersize = sizeof(struct xt_counters) * private->number;
 916         counters = vzalloc(countersize);
 917
 918         if (counters == NULL)
 919                 return ERR_PTR(-ENOMEM);
 920
 921         get_counters(private, counters);
 922
 923         return counters;
 924 }
 925
 926 static int
 927 copy_entries_to_user(unsigned int total_size,
 928                      const struct xt_table *table,
 929                      void __user *userptr)
 930 {
 931         unsigned int off, num;
 932         const struct ipt_entry *e;
 933         struct xt_counters *counters;
 934         const struct xt_table_info *private = table->private;
 935         int ret = 0;
 936         const void *loc_cpu_entry;
 937
 938         counters = alloc_counters(table);
 939         if (IS_ERR(counters))
 940                 return PTR_ERR(counters);
 941
 942         /* choose the copy that is on our node/cpu, ...
 943          * This choice is lazy (because current thread is
 944          * allowed to migrate to another cpu)
 945          */
 946         loc_cpu_entry = private->entries[raw_smp_processor_id()];
 947         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 948                 ret = -EFAULT;
 949                 goto free_counters;
 950         }
 951
 952         /* FIXME: use iterator macros --RR */
 953         /* ... then go back and fix counters and names */
 954         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 955                 unsigned int i;
 956                 const struct xt_entry_match *m;
 957                 const struct xt_entry_target *t;
 958
 959                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 960                 if (copy_to_user(userptr + off
 961                                  + offsetof(struct ipt_entry, counters),
 962                                  &counters[num],
 963                                  sizeof(counters[num])) != 0) {
 964                         ret = -EFAULT;
 965                         goto free_counters;
 966                 }
 967
 968                 for (i = sizeof(struct ipt_entry);
 969                      i < e->target_offset;
 970                      i += m->u.match_size) {
 971                         m = (void *)e + i;
 972
 973                         if (copy_to_user(userptr + off + i
 974                                          + offsetof(struct xt_entry_match,
 975                                                     u.user.name),
 976                                          m->u.kernel.match->name,
 977                                          strlen(m->u.kernel.match->name)+1)
 978                             != 0) {
 979                                 ret = -EFAULT;
 980                                 goto free_counters;
 981                         }
 982                 }
 983
 984                 t = ipt_get_target_c(e);
 985                 if (copy_to_user(userptr + off + e->target_offset
 986                                  + offsetof(struct xt_entry_target,
 987                                             u.user.name),
 988                                  t->u.kernel.target->name,
 989                                  strlen(t->u.kernel.target->name)+1) != 0) {
 990                         ret = -EFAULT;
 991                         goto free_counters;
 992                 }
 993         }
 994
 995  free_counters:
 996         vfree(counters);
 997         return ret;
 998 }
 999
1000 #ifdef CONFIG_COMPAT
1001 static void compat_standard_from_user(void *dst, const void *src)
1002 {
1003         int v = *(compat_int_t *)src;
1004
1005         if (v > 0)
1006                 v += xt_compat_calc_jump(AF_INET, v);
1007         memcpy(dst, &v, sizeof(v));
1008 }
1009
1010 static int compat_standard_to_user(void __user *dst, const void *src)
1011 {
1012         compat_int_t cv = *(int *)src;
1013
1014         if (cv > 0)
1015                 cv -= xt_compat_calc_jump(AF_INET, cv);
1016         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1017 }
1018
1019 static int compat_calc_entry(const struct ipt_entry *e,
1020                              const struct xt_table_info *info,
1021                              const void *base, struct xt_table_info *newinfo)
1022 {
1023         const struct xt_entry_match *ematch;
1024         const struct xt_entry_target *t;
1025         unsigned int entry_offset;
1026         int off, i, ret;
1027
1028         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1029         entry_offset = (void *)e - base;
1030         xt_ematch_foreach(ematch, e)
1031                 off += xt_compat_match_offset(ematch->u.kernel.match);
1032         t = ipt_get_target_c(e);
1033         off += xt_compat_target_offset(t->u.kernel.target);
1034         newinfo->size -= off;
1035         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1036         if (ret)
1037                 return ret;
1038
1039         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1040                 if (info->hook_entry[i] &&
1041                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1042                         newinfo->hook_entry[i] -= off;
1043                 if (info->underflow[i] &&
1044                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1045                         newinfo->underflow[i] -= off;
1046         }
1047         return 0;
1048 }
1049
1050 static int compat_table_info(const struct xt_table_info *info,
1051                              struct xt_table_info *newinfo)
1052 {
1053         struct ipt_entry *iter;
1054         void *loc_cpu_entry;
1055         int ret;
1056
1057         if (!newinfo || !info)
1058                 return -EINVAL;
1059
1060         /* we dont care about newinfo->entries[] */
1061         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1062         newinfo->initial_entries = 0;
1063         loc_cpu_entry = info->entries[raw_smp_processor_id()];
1064         xt_compat_init_offsets(AF_INET, info->number);
1065         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1066                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1067                 if (ret != 0)
1068                         return ret;
1069         }
1070         return 0;
1071 }
1072 #endif
1073
1074 static int get_info(struct net *net, void __user *user,
1075                     const int *len, int compat)
1076 {
1077         char name[XT_TABLE_MAXNAMELEN];
1078         struct xt_table *t;
1079         int ret;
1080
1081         if (*len != sizeof(struct ipt_getinfo)) {
1082                 duprintf("length %u != %zu\n", *len,
1083                          sizeof(struct ipt_getinfo));
1084                 return -EINVAL;
1085         }
1086
1087         if (copy_from_user(name, user, sizeof(name)) != 0)
1088                 return -EFAULT;
1089
1090         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1091 #ifdef CONFIG_COMPAT
1092         if (compat)
1093                 xt_compat_lock(AF_INET);
1094 #endif
1095         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1096                                     "iptable_%s", name);
1097         if (!IS_ERR_OR_NULL(t)) {
1098                 struct ipt_getinfo info;
1099                 const struct xt_table_info *private = t->private;
1100 #ifdef CONFIG_COMPAT
1101                 struct xt_table_info tmp;
1102
1103                 if (compat) {
1104                         ret = compat_table_info(private, &tmp);
1105                         xt_compat_flush_offsets(AF_INET);
1106                         private = &tmp;
1107                 }
1108 #endif
1109                 memset(&info, 0, sizeof(info));
1110                 info.valid_hooks = t->valid_hooks;
1111                 memcpy(info.hook_entry, private->hook_entry,
1112                        sizeof(info.hook_entry));
1113                 memcpy(info.underflow, private->underflow,
1114                        sizeof(info.underflow));
1115                 info.num_entries = private->number;
1116                 info.size = private->size;
1117                 strcpy(info.name, name);
1118
1119                 if (copy_to_user(user, &info, *len) != 0)
1120                         ret = -EFAULT;
1121                 else
1122                         ret = 0;
1123
1124                 xt_table_unlock(t);
1125                 module_put(t->me);
1126         } else
1127                 ret = t ? PTR_ERR(t) : -ENOENT;
1128 #ifdef CONFIG_COMPAT
1129         if (compat)
1130                 xt_compat_unlock(AF_INET);
1131 #endif
1132         return ret;
1133 }
1134
1135 static int
1136 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1137             const int *len)
1138 {
1139         int ret;
1140         struct ipt_get_entries get;
1141         struct xt_table *t;
1142
1143         if (*len < sizeof(get)) {
1144                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1145                 return -EINVAL;
1146         }
1147         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1148                 return -EFAULT;
1149         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1150                 duprintf("get_entries: %u != %zu\n",
1151                          *len, sizeof(get) + get.size);
1152                 return -EINVAL;
1153         }
1154
1155         t = xt_find_table_lock(net, AF_INET, get.name);
1156         if (!IS_ERR_OR_NULL(t)) {
1157                 const struct xt_table_info *private = t->private;
1158                 duprintf("t->private->number = %u\n", private->number);
1159                 if (get.size == private->size)
1160                         ret = copy_entries_to_user(private->size,
1161                                                    t, uptr->entrytable);
1162                 else {
1163                         duprintf("get_entries: I've got %u not %u!\n",
1164                                  private->size, get.size);
1165                         ret = -EAGAIN;
1166                 }
1167                 module_put(t->me);
1168                 xt_table_unlock(t);
1169         } else
1170                 ret = t ? PTR_ERR(t) : -ENOENT;
1171
1172         return ret;
1173 }
1174
1175 static int
1176 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1177              struct xt_table_info *newinfo, unsigned int num_counters,
1178              void __user *counters_ptr)
1179 {
1180         int ret;
1181         struct xt_table *t;
1182         struct xt_table_info *oldinfo;
1183         struct xt_counters *counters;
1184         void *loc_cpu_old_entry;
1185         struct ipt_entry *iter;
1186
1187         ret = 0;
1188         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1189         if (!counters) {
1190                 ret = -ENOMEM;
1191                 goto out;
1192         }
1193
1194         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1195                                     "iptable_%s", name);
1196         if (IS_ERR_OR_NULL(t)) {
1197                 ret = t ? PTR_ERR(t) : -ENOENT;
1198                 goto free_newinfo_counters_untrans;
1199         }
1200
1201         /* You lied! */
1202         if (valid_hooks != t->valid_hooks) {
1203                 duprintf("Valid hook crap: %08X vs %08X\n",
1204                          valid_hooks, t->valid_hooks);
1205                 ret = -EINVAL;
1206                 goto put_module;
1207         }
1208
1209         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1210         if (!oldinfo)
1211                 goto put_module;
1212
1213         /* Update module usage count based on number of rules */
1214         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1215                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1216         if ((oldinfo->number > oldinfo->initial_entries) ||
1217             (newinfo->number <= oldinfo->initial_entries))
1218                 module_put(t->me);
1219         if ((oldinfo->number > oldinfo->initial_entries) &&
1220             (newinfo->number <= oldinfo->initial_entries))
1221                 module_put(t->me);
1222
1223         /* Get the old counters, and synchronize with replace */
1224         get_counters(oldinfo, counters);
1225
1226         /* Decrease module usage counts and free resource */
1227         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
1228         xt_entry_foreach(iter, loc_cpu_old_entry, oldinfo->size)
1229                 cleanup_entry(iter, net);
1230
1231         xt_free_table_info(oldinfo);
1232         if (copy_to_user(counters_ptr, counters,
1233                          sizeof(struct xt_counters) * num_counters) != 0) {
1234                 /* Silent error, can't fail, new table is already in place */
1235                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1236         }
1237         vfree(counters);
1238         xt_table_unlock(t);
1239         return ret;
1240
1241  put_module:
1242         module_put(t->me);
1243         xt_table_unlock(t);
1244  free_newinfo_counters_untrans:
1245         vfree(counters);
1246  out:
1247         return ret;
1248 }
1249
1250 static int
1251 do_replace(struct net *net, const void __user *user, unsigned int len)
1252 {
1253         int ret;
1254         struct ipt_replace tmp;
1255         struct xt_table_info *newinfo;
1256         void *loc_cpu_entry;
1257         struct ipt_entry *iter;
1258
1259         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1260                 return -EFAULT;
1261
1262         /* overflow check */
1263         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1264                 return -ENOMEM;
1265         if (tmp.num_counters == 0)
1266                 return -EINVAL;
1267
1268         tmp.name[sizeof(tmp.name)-1] = 0;
1269
1270         newinfo = xt_alloc_table_info(tmp.size);
1271         if (!newinfo)
1272                 return -ENOMEM;
1273
1274         /* choose the copy that is on our node/cpu */
1275         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1276         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1277                            tmp.size) != 0) {
1278                 ret = -EFAULT;
1279                 goto free_newinfo;
1280         }
1281
1282         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1283         if (ret != 0)
1284                 goto free_newinfo;
1285
1286         duprintf("Translated table\n");
1287
1288         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1289                            tmp.num_counters, tmp.counters);
1290         if (ret)
1291                 goto free_newinfo_untrans;
1292         return 0;
1293
1294  free_newinfo_untrans:
1295         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1296                 cleanup_entry(iter, net);
1297  free_newinfo:
1298         xt_free_table_info(newinfo);
1299         return ret;
1300 }
1301
1302 static int
1303 do_add_counters(struct net *net, const void __user *user,
1304                 unsigned int len, int compat)
1305 {
1306         unsigned int i, curcpu;
1307         struct xt_counters_info tmp;
1308         struct xt_counters *paddc;
1309         unsigned int num_counters;
1310         const char *name;
1311         int size;
1312         void *ptmp;
1313         struct xt_table *t;
1314         const struct xt_table_info *private;
1315         int ret = 0;
1316         void *loc_cpu_entry;
1317         struct ipt_entry *iter;
1318         unsigned int addend;
1319 #ifdef CONFIG_COMPAT
1320         struct compat_xt_counters_info compat_tmp;
1321
1322         if (compat) {
1323                 ptmp = &compat_tmp;
1324                 size = sizeof(struct compat_xt_counters_info);
1325         } else
1326 #endif
1327         {
1328                 ptmp = &tmp;
1329                 size = sizeof(struct xt_counters_info);
1330         }
1331
1332         if (copy_from_user(ptmp, user, size) != 0)
1333                 return -EFAULT;
1334
1335 #ifdef CONFIG_COMPAT
1336         if (compat) {
1337                 num_counters = compat_tmp.num_counters;
1338                 name = compat_tmp.name;
1339         } else
1340 #endif
1341         {
1342                 num_counters = tmp.num_counters;
1343                 name = tmp.name;
1344         }
1345
1346         if (len != size + num_counters * sizeof(struct xt_counters))
1347                 return -EINVAL;
1348
1349         paddc = vmalloc(len - size);
1350         if (!paddc)
1351                 return -ENOMEM;
1352
1353         if (copy_from_user(paddc, user + size, len - size) != 0) {
1354                 ret = -EFAULT;
1355                 goto free;
1356         }
1357
1358         t = xt_find_table_lock(net, AF_INET, name);
1359         if (IS_ERR_OR_NULL(t)) {
1360                 ret = t ? PTR_ERR(t) : -ENOENT;
1361                 goto free;
1362         }
1363
1364         local_bh_disable();
1365         private = t->private;
1366         if (private->number != num_counters) {
1367                 ret = -EINVAL;
1368                 goto unlock_up_free;
1369         }
1370
1371         i = 0;
1372         /* Choose the copy that is on our node */
1373         curcpu = smp_processor_id();
1374         loc_cpu_entry = private->entries[curcpu];
1375         addend = xt_write_recseq_begin();
1376         xt_entry_foreach(iter, loc_cpu_entry, private->size) {
1377                 ADD_COUNTER(iter->counters, paddc[i].bcnt, paddc[i].pcnt);
1378                 ++i;
1379         }
1380         xt_write_recseq_end(addend);
1381  unlock_up_free:
1382         local_bh_enable();
1383         xt_table_unlock(t);
1384         module_put(t->me);
1385  free:
1386         vfree(paddc);
1387
1388         return ret;
1389 }
1390
1391 #ifdef CONFIG_COMPAT
1392 struct compat_ipt_replace {
1393         char                    name[XT_TABLE_MAXNAMELEN];
1394         u32                     valid_hooks;
1395         u32                     num_entries;
1396         u32                     size;
1397         u32                     hook_entry[NF_INET_NUMHOOKS];
1398         u32                     underflow[NF_INET_NUMHOOKS];
1399         u32                     num_counters;
1400         compat_uptr_t           counters;       /* struct xt_counters * */
1401         struct compat_ipt_entry entries[0];
1402 };
1403
1404 static int
1405 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1406                           unsigned int *size, struct xt_counters *counters,
1407                           unsigned int i)
1408 {
1409         struct xt_entry_target *t;
1410         struct compat_ipt_entry __user *ce;
1411         u_int16_t target_offset, next_offset;
1412         compat_uint_t origsize;
1413         const struct xt_entry_match *ematch;
1414         int ret = 0;
1415
1416         origsize = *size;
1417         ce = (struct compat_ipt_entry __user *)*dstptr;
1418         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1419             copy_to_user(&ce->counters, &counters[i],
1420             sizeof(counters[i])) != 0)
1421                 return -EFAULT;
1422
1423         *dstptr += sizeof(struct compat_ipt_entry);
1424         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1425
1426         xt_ematch_foreach(ematch, e) {
1427                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1428                 if (ret != 0)
1429                         return ret;
1430         }
1431         target_offset = e->target_offset - (origsize - *size);
1432         t = ipt_get_target(e);
1433         ret = xt_compat_target_to_user(t, dstptr, size);
1434         if (ret)
1435                 return ret;
1436         next_offset = e->next_offset - (origsize - *size);
1437         if (put_user(target_offset, &ce->target_offset) != 0 ||
1438             put_user(next_offset, &ce->next_offset) != 0)
1439                 return -EFAULT;
1440         return 0;
1441 }
1442
1443 static int
1444 compat_find_calc_match(struct xt_entry_match *m,
1445                        const char *name,
1446                        const struct ipt_ip *ip,
1447                        unsigned int hookmask,
1448                        int *size)
1449 {
1450         struct xt_match *match;
1451
1452         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1453                                       m->u.user.revision);
1454         if (IS_ERR(match)) {
1455                 duprintf("compat_check_calc_match: `%s' not found\n",
1456                          m->u.user.name);
1457                 return PTR_ERR(match);
1458         }
1459         m->u.kernel.match = match;
1460         *size += xt_compat_match_offset(match);
1461         return 0;
1462 }
1463
1464 static void compat_release_entry(struct compat_ipt_entry *e)
1465 {
1466         struct xt_entry_target *t;
1467         struct xt_entry_match *ematch;
1468
1469         /* Cleanup all matches */
1470         xt_ematch_foreach(ematch, e)
1471                 module_put(ematch->u.kernel.match->me);
1472         t = compat_ipt_get_target(e);
1473         module_put(t->u.kernel.target->me);
1474 }
1475
1476 static int
1477 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1478                                   struct xt_table_info *newinfo,
1479                                   unsigned int *size,
1480                                   const unsigned char *base,
1481                                   const unsigned char *limit,
1482                                   const unsigned int *hook_entries,
1483                                   const unsigned int *underflows,
1484                                   const char *name)
1485 {
1486         struct xt_entry_match *ematch;
1487         struct xt_entry_target *t;
1488         struct xt_target *target;
1489         unsigned int entry_offset;
1490         unsigned int j;
1491         int ret, off, h;
1492
1493         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1494         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1495             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit) {
1496                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1497                 return -EINVAL;
1498         }
1499
1500         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1501                              sizeof(struct compat_xt_entry_target)) {
1502                 duprintf("checking: element %p size %u\n",
1503                          e, e->next_offset);
1504                 return -EINVAL;
1505         }
1506
1507         /* For purposes of check_entry casting the compat entry is fine */
1508         ret = check_entry((struct ipt_entry *)e, name);
1509         if (ret)
1510                 return ret;
1511
1512         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1513         entry_offset = (void *)e - (void *)base;
1514         j = 0;
1515         xt_ematch_foreach(ematch, e) {
1516                 ret = compat_find_calc_match(ematch, name,
1517                                              &e->ip, e->comefrom, &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         j = 0;
1614         mtpar.net       = net;
1615         mtpar.table     = name;
1616         mtpar.entryinfo = &e->ip;
1617         mtpar.hook_mask = e->comefrom;
1618         mtpar.family    = NFPROTO_IPV4;
1619         xt_ematch_foreach(ematch, e) {
1620                 ret = check_match(ematch, &mtpar);
1621                 if (ret != 0)
1622                         goto cleanup_matches;
1623                 ++j;
1624         }
1625
1626         ret = check_target(e, net, name);
1627         if (ret)
1628                 goto cleanup_matches;
1629         return 0;
1630
1631  cleanup_matches:
1632         xt_ematch_foreach(ematch, e) {
1633                 if (j-- == 0)
1634                         break;
1635                 cleanup_match(ematch, net);
1636         }
1637         return ret;
1638 }
1639
1640 static int
1641 translate_compat_table(struct net *net,
1642                        const char *name,
1643                        unsigned int valid_hooks,
1644                        struct xt_table_info **pinfo,
1645                        void **pentry0,
1646                        unsigned int total_size,
1647                        unsigned int number,
1648                        unsigned int *hook_entries,
1649                        unsigned int *underflows)
1650 {
1651         unsigned int i, j;
1652         struct xt_table_info *newinfo, *info;
1653         void *pos, *entry0, *entry1;
1654         struct compat_ipt_entry *iter0;
1655         struct ipt_entry *iter1;
1656         unsigned int size;
1657         int ret;
1658
1659         info = *pinfo;
1660         entry0 = *pentry0;
1661         size = total_size;
1662         info->number = number;
1663
1664         /* Init all hooks to impossible value. */
1665         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1666                 info->hook_entry[i] = 0xFFFFFFFF;
1667                 info->underflow[i] = 0xFFFFFFFF;
1668         }
1669
1670         duprintf("translate_compat_table: size %u\n", info->size);
1671         j = 0;
1672         xt_compat_lock(AF_INET);
1673         xt_compat_init_offsets(AF_INET, number);
1674         /* Walk through entries, checking offsets. */
1675         xt_entry_foreach(iter0, entry0, total_size) {
1676                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1677                                                         entry0,
1678                                                         entry0 + total_size,
1679                                                         hook_entries,
1680                                                         underflows,
1681                                                         name);
1682                 if (ret != 0)
1683                         goto out_unlock;
1684                 ++j;
1685         }
1686
1687         ret = -EINVAL;
1688         if (j != number) {
1689                 duprintf("translate_compat_table: %u not %u entries\n",
1690                          j, number);
1691                 goto out_unlock;
1692         }
1693
1694         /* Check hooks all assigned */
1695         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1696                 /* Only hooks which are valid */
1697                 if (!(valid_hooks & (1 << i)))
1698                         continue;
1699                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1700                         duprintf("Invalid hook entry %u %u\n",
1701                                  i, hook_entries[i]);
1702                         goto out_unlock;
1703                 }
1704                 if (info->underflow[i] == 0xFFFFFFFF) {
1705                         duprintf("Invalid underflow %u %u\n",
1706                                  i, underflows[i]);
1707                         goto out_unlock;
1708                 }
1709         }
1710
1711         ret = -ENOMEM;
1712         newinfo = xt_alloc_table_info(size);
1713         if (!newinfo)
1714                 goto out_unlock;
1715
1716         newinfo->number = number;
1717         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1718                 newinfo->hook_entry[i] = info->hook_entry[i];
1719                 newinfo->underflow[i] = info->underflow[i];
1720         }
1721         entry1 = newinfo->entries[raw_smp_processor_id()];
1722         pos = entry1;
1723         size = total_size;
1724         xt_entry_foreach(iter0, entry0, total_size) {
1725                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1726                                                   name, newinfo, entry1);
1727                 if (ret != 0)
1728                         break;
1729         }
1730         xt_compat_flush_offsets(AF_INET);
1731         xt_compat_unlock(AF_INET);
1732         if (ret)
1733                 goto free_newinfo;
1734
1735         ret = -ELOOP;
1736         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1737                 goto free_newinfo;
1738
1739         i = 0;
1740         xt_entry_foreach(iter1, entry1, newinfo->size) {
1741                 ret = compat_check_entry(iter1, net, name);
1742                 if (ret != 0)
1743                         break;
1744                 ++i;
1745                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1746                     XT_ERROR_TARGET) == 0)
1747                         ++newinfo->stacksize;
1748         }
1749         if (ret) {
1750                 /*
1751                  * The first i matches need cleanup_entry (calls ->destroy)
1752                  * because they had called ->check already. The other j-i
1753                  * entries need only release.
1754                  */
1755                 int skip = i;
1756                 j -= i;
1757                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1758                         if (skip-- > 0)
1759                                 continue;
1760                         if (j-- == 0)
1761                                 break;
1762                         compat_release_entry(iter0);
1763                 }
1764                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1765                         if (i-- == 0)
1766                                 break;
1767                         cleanup_entry(iter1, net);
1768                 }
1769                 xt_free_table_info(newinfo);
1770                 return ret;
1771         }
1772
1773         /* And one copy for every other CPU */
1774         for_each_possible_cpu(i)
1775                 if (newinfo->entries[i] && newinfo->entries[i] != entry1)
1776                         memcpy(newinfo->entries[i], entry1, newinfo->size);
1777
1778         *pinfo = newinfo;
1779         *pentry0 = entry1;
1780         xt_free_table_info(info);
1781         return 0;
1782
1783 free_newinfo:
1784         xt_free_table_info(newinfo);
1785 out:
1786         xt_entry_foreach(iter0, entry0, total_size) {
1787                 if (j-- == 0)
1788                         break;
1789                 compat_release_entry(iter0);
1790         }
1791         return ret;
1792 out_unlock:
1793         xt_compat_flush_offsets(AF_INET);
1794         xt_compat_unlock(AF_INET);
1795         goto out;
1796 }
1797
1798 static int
1799 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1800 {
1801         int ret;
1802         struct compat_ipt_replace tmp;
1803         struct xt_table_info *newinfo;
1804         void *loc_cpu_entry;
1805         struct ipt_entry *iter;
1806
1807         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1808                 return -EFAULT;
1809
1810         /* overflow check */
1811         if (tmp.size >= INT_MAX / num_possible_cpus())
1812                 return -ENOMEM;
1813         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1814                 return -ENOMEM;
1815         if (tmp.num_counters == 0)
1816                 return -EINVAL;
1817
1818         tmp.name[sizeof(tmp.name)-1] = 0;
1819
1820         newinfo = xt_alloc_table_info(tmp.size);
1821         if (!newinfo)
1822                 return -ENOMEM;
1823
1824         /* choose the copy that is on our node/cpu */
1825         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1826         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1827                            tmp.size) != 0) {
1828                 ret = -EFAULT;
1829                 goto free_newinfo;
1830         }
1831
1832         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1833                                      &newinfo, &loc_cpu_entry, tmp.size,
1834                                      tmp.num_entries, tmp.hook_entry,
1835                                      tmp.underflow);
1836         if (ret != 0)
1837                 goto free_newinfo;
1838
1839         duprintf("compat_do_replace: Translated table\n");
1840
1841         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1842                            tmp.num_counters, compat_ptr(tmp.counters));
1843         if (ret)
1844                 goto free_newinfo_untrans;
1845         return 0;
1846
1847  free_newinfo_untrans:
1848         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1849                 cleanup_entry(iter, net);
1850  free_newinfo:
1851         xt_free_table_info(newinfo);
1852         return ret;
1853 }
1854
1855 static int
1856 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1857                       unsigned int len)
1858 {
1859         int ret;
1860
1861         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1862                 return -EPERM;
1863
1864         switch (cmd) {
1865         case IPT_SO_SET_REPLACE:
1866                 ret = compat_do_replace(sock_net(sk), user, len);
1867                 break;
1868
1869         case IPT_SO_SET_ADD_COUNTERS:
1870                 ret = do_add_counters(sock_net(sk), user, len, 1);
1871                 break;
1872
1873         default:
1874                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1875                 ret = -EINVAL;
1876         }
1877
1878         return ret;
1879 }
1880
1881 struct compat_ipt_get_entries {
1882         char name[XT_TABLE_MAXNAMELEN];
1883         compat_uint_t size;
1884         struct compat_ipt_entry entrytable[0];
1885 };
1886
1887 static int
1888 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1889                             void __user *userptr)
1890 {
1891         struct xt_counters *counters;
1892         const struct xt_table_info *private = table->private;
1893         void __user *pos;
1894         unsigned int size;
1895         int ret = 0;
1896         const void *loc_cpu_entry;
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         /* choose the copy that is on our node/cpu, ...
1905          * This choice is lazy (because current thread is
1906          * allowed to migrate to another cpu)
1907          */
1908         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1909         pos = userptr;
1910         size = total_size;
1911         xt_entry_foreach(iter, loc_cpu_entry, total_size) {
1912                 ret = compat_copy_entry_to_user(iter, &pos,
1913                                                 &size, counters, i++);
1914                 if (ret != 0)
1915                         break;
1916         }
1917
1918         vfree(counters);
1919         return ret;
1920 }
1921
1922 static int
1923 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1924                    int *len)
1925 {
1926         int ret;
1927         struct compat_ipt_get_entries get;
1928         struct xt_table *t;
1929
1930         if (*len < sizeof(get)) {
1931                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1932                 return -EINVAL;
1933         }
1934
1935         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1936                 return -EFAULT;
1937
1938         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1939                 duprintf("compat_get_entries: %u != %zu\n",
1940                          *len, sizeof(get) + get.size);
1941                 return -EINVAL;
1942         }
1943
1944         xt_compat_lock(AF_INET);
1945         t = xt_find_table_lock(net, AF_INET, get.name);
1946         if (!IS_ERR_OR_NULL(t)) {
1947                 const struct xt_table_info *private = t->private;
1948                 struct xt_table_info info;
1949                 duprintf("t->private->number = %u\n", private->number);
1950                 ret = compat_table_info(private, &info);
1951                 if (!ret && get.size == info.size) {
1952                         ret = compat_copy_entries_to_user(private->size,
1953                                                           t, uptr->entrytable);
1954                 } else if (!ret) {
1955                         duprintf("compat_get_entries: I've got %u not %u!\n",
1956                                  private->size, get.size);
1957                         ret = -EAGAIN;
1958                 }
1959                 xt_compat_flush_offsets(AF_INET);
1960                 module_put(t->me);
1961                 xt_table_unlock(t);
1962         } else
1963                 ret = t ? PTR_ERR(t) : -ENOENT;
1964
1965         xt_compat_unlock(AF_INET);
1966         return ret;
1967 }
1968
1969 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1970
1971 static int
1972 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1973 {
1974         int ret;
1975
1976         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1977                 return -EPERM;
1978
1979         switch (cmd) {
1980         case IPT_SO_GET_INFO:
1981                 ret = get_info(sock_net(sk), user, len, 1);
1982                 break;
1983         case IPT_SO_GET_ENTRIES:
1984                 ret = compat_get_entries(sock_net(sk), user, len);
1985                 break;
1986         default:
1987                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1988         }
1989         return ret;
1990 }
1991 #endif
1992
1993 static int
1994 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1995 {
1996         int ret;
1997
1998         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1999                 return -EPERM;
2000
2001         switch (cmd) {
2002         case IPT_SO_SET_REPLACE:
2003                 ret = do_replace(sock_net(sk), user, len);
2004                 break;
2005
2006         case IPT_SO_SET_ADD_COUNTERS:
2007                 ret = do_add_counters(sock_net(sk), user, len, 0);
2008                 break;
2009
2010         default:
2011                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2012                 ret = -EINVAL;
2013         }
2014
2015         return ret;
2016 }
2017
2018 static int
2019 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2020 {
2021         int ret;
2022
2023         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2024                 return -EPERM;
2025
2026         switch (cmd) {
2027         case IPT_SO_GET_INFO:
2028                 ret = get_info(sock_net(sk), user, len, 0);
2029                 break;
2030
2031         case IPT_SO_GET_ENTRIES:
2032                 ret = get_entries(sock_net(sk), user, len);
2033                 break;
2034
2035         case IPT_SO_GET_REVISION_MATCH:
2036         case IPT_SO_GET_REVISION_TARGET: {
2037                 struct xt_get_revision rev;
2038                 int target;
2039
2040                 if (*len != sizeof(rev)) {
2041                         ret = -EINVAL;
2042                         break;
2043                 }
2044                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2045                         ret = -EFAULT;
2046                         break;
2047                 }
2048                 rev.name[sizeof(rev.name)-1] = 0;
2049
2050                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2051                         target = 1;
2052                 else
2053                         target = 0;
2054
2055                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2056                                                          rev.revision,
2057                                                          target, &ret),
2058                                         "ipt_%s", rev.name);
2059                 break;
2060         }
2061
2062         default:
2063                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2064                 ret = -EINVAL;
2065         }
2066
2067         return ret;
2068 }
2069
2070 struct xt_table *ipt_register_table(struct net *net,
2071                                     const struct xt_table *table,
2072                                     const struct ipt_replace *repl)
2073 {
2074         int ret;
2075         struct xt_table_info *newinfo;
2076         struct xt_table_info bootstrap = {0};
2077         void *loc_cpu_entry;
2078         struct xt_table *new_table;
2079
2080         newinfo = xt_alloc_table_info(repl->size);
2081         if (!newinfo) {
2082                 ret = -ENOMEM;
2083                 goto out;
2084         }
2085
2086         /* choose the copy on our node/cpu, but dont care about preemption */
2087         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
2088         memcpy(loc_cpu_entry, repl->entries, repl->size);
2089
2090         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2091         if (ret != 0)
2092                 goto out_free;
2093
2094         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2095         if (IS_ERR(new_table)) {
2096                 ret = PTR_ERR(new_table);
2097                 goto out_free;
2098         }
2099
2100         return new_table;
2101
2102 out_free:
2103         xt_free_table_info(newinfo);
2104 out:
2105         return ERR_PTR(ret);
2106 }
2107
2108 void ipt_unregister_table(struct net *net, struct xt_table *table)
2109 {
2110         struct xt_table_info *private;
2111         void *loc_cpu_entry;
2112         struct module *table_owner = table->me;
2113         struct ipt_entry *iter;
2114
2115         private = xt_unregister_table(table);
2116
2117         /* Decrease module usage counts and free resources */
2118         loc_cpu_entry = private->entries[raw_smp_processor_id()];
2119         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2120                 cleanup_entry(iter, net);
2121         if (private->number > private->initial_entries)
2122                 module_put(table_owner);
2123         xt_free_table_info(private);
2124 }
2125
2126 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2127 static inline bool
2128 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2129                      u_int8_t type, u_int8_t code,
2130                      bool invert)
2131 {
2132         return ((test_type == 0xFF) ||
2133                 (type == test_type && code >= min_code && code <= max_code))
2134                 ^ invert;
2135 }
2136
2137 static bool
2138 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2139 {
2140         const struct icmphdr *ic;
2141         struct icmphdr _icmph;
2142         const struct ipt_icmp *icmpinfo = par->matchinfo;
2143
2144         /* Must not be a fragment. */
2145         if (par->fragoff != 0)
2146                 return false;
2147
2148         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2149         if (ic == NULL) {
2150                 /* We've been asked to examine this packet, and we
2151                  * can't.  Hence, no choice but to drop.
2152                  */
2153                 duprintf("Dropping evil ICMP tinygram.\n");
2154                 par->hotdrop = true;
2155                 return false;
2156         }
2157
2158         return icmp_type_code_match(icmpinfo->type,
2159                                     icmpinfo->code[0],
2160                                     icmpinfo->code[1],
2161                                     ic->type, ic->code,
2162                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2163 }
2164
2165 static int icmp_checkentry(const struct xt_mtchk_param *par)
2166 {
2167         const struct ipt_icmp *icmpinfo = par->matchinfo;
2168
2169         /* Must specify no unknown invflags */
2170         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2171 }
2172
2173 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2174         {
2175                 .name             = XT_STANDARD_TARGET,
2176                 .targetsize       = sizeof(int),
2177                 .family           = NFPROTO_IPV4,
2178 #ifdef CONFIG_COMPAT
2179                 .compatsize       = sizeof(compat_int_t),
2180                 .compat_from_user = compat_standard_from_user,
2181                 .compat_to_user   = compat_standard_to_user,
2182 #endif
2183         },
2184         {
2185                 .name             = XT_ERROR_TARGET,
2186                 .target           = ipt_error,
2187                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2188                 .family           = NFPROTO_IPV4,
2189         },
2190 };
2191
2192 static struct nf_sockopt_ops ipt_sockopts = {
2193         .pf             = PF_INET,
2194         .set_optmin     = IPT_BASE_CTL,
2195         .set_optmax     = IPT_SO_SET_MAX+1,
2196         .set            = do_ipt_set_ctl,
2197 #ifdef CONFIG_COMPAT
2198         .compat_set     = compat_do_ipt_set_ctl,
2199 #endif
2200         .get_optmin     = IPT_BASE_CTL,
2201         .get_optmax     = IPT_SO_GET_MAX+1,
2202         .get            = do_ipt_get_ctl,
2203 #ifdef CONFIG_COMPAT
2204         .compat_get     = compat_do_ipt_get_ctl,
2205 #endif
2206         .owner          = THIS_MODULE,
2207 };
2208
2209 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2210         {
2211                 .name       = "icmp",
2212                 .match      = icmp_match,
2213                 .matchsize  = sizeof(struct ipt_icmp),
2214                 .checkentry = icmp_checkentry,
2215                 .proto      = IPPROTO_ICMP,
2216                 .family     = NFPROTO_IPV4,
2217         },
2218 };
2219
2220 static int __net_init ip_tables_net_init(struct net *net)
2221 {
2222         return xt_proto_init(net, NFPROTO_IPV4);
2223 }
2224
2225 static void __net_exit ip_tables_net_exit(struct net *net)
2226 {
2227         xt_proto_fini(net, NFPROTO_IPV4);
2228 }
2229
2230 static struct pernet_operations ip_tables_net_ops = {
2231         .init = ip_tables_net_init,
2232         .exit = ip_tables_net_exit,
2233 };
2234
2235 static int __init ip_tables_init(void)
2236 {
2237         int ret;
2238
2239         ret = register_pernet_subsys(&ip_tables_net_ops);
2240         if (ret < 0)
2241                 goto err1;
2242
2243         /* No one else will be downing sem now, so we won't sleep */
2244         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2245         if (ret < 0)
2246                 goto err2;
2247         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2248         if (ret < 0)
2249                 goto err4;
2250
2251         /* Register setsockopt */
2252         ret = nf_register_sockopt(&ipt_sockopts);
2253         if (ret < 0)
2254                 goto err5;
2255
2256         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2257         return 0;
2258
2259 err5:
2260         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2261 err4:
2262         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2263 err2:
2264         unregister_pernet_subsys(&ip_tables_net_ops);
2265 err1:
2266         return ret;
2267 }
2268
2269 static void __exit ip_tables_fini(void)
2270 {
2271         nf_unregister_sockopt(&ipt_sockopts);
2272
2273         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2274         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2275         unregister_pernet_subsys(&ip_tables_net_ops);
2276 }
2277
2278 EXPORT_SYMBOL(ipt_register_table);
2279 EXPORT_SYMBOL(ipt_unregister_table);
2280 EXPORT_SYMBOL(ipt_do_table);
2281 module_init(ip_tables_init);
2282 module_exit(ip_tables_fini);