net/ipv4/netfilter/ip_tables.c

   1 /*
   2  * Packet matching code.
   3  *
   4  * Copyright (C) 1999 Paul `Rusty' Russell & Michael J. Neuling
   5  * Copyright (C) 2000-2005 Netfilter Core Team <coreteam@netfilter.org>
   6  * Copyright (C) 2006-2010 Patrick McHardy <kaber@trash.net>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2 as
  10  * published by the Free Software Foundation.
  11  */
  12 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
  13 #include <linux/cache.h>
  14 #include <linux/capability.h>
  15 #include <linux/skbuff.h>
  16 #include <linux/kmod.h>
  17 #include <linux/vmalloc.h>
  18 #include <linux/netdevice.h>
  19 #include <linux/module.h>
  20 #include <linux/icmp.h>
  21 #include <net/ip.h>
  22 #include <net/compat.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/mutex.h>
  25 #include <linux/proc_fs.h>
  26 #include <linux/err.h>
  27 #include <linux/cpumask.h>
  28
  29 #include <linux/netfilter/x_tables.h>
  30 #include <linux/netfilter_ipv4/ip_tables.h>
  31 #include <net/netfilter/nf_log.h>
  32 #include "../../netfilter/xt_repldata.h"
  33
  34 MODULE_LICENSE("GPL");
  35 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36 MODULE_DESCRIPTION("IPv4 packet filter");
  37
  38 /*#define DEBUG_IP_FIREWALL*/
  39 /*#define DEBUG_ALLOW_ALL*/ /* Useful for remote debugging */
  40 /*#define DEBUG_IP_FIREWALL_USER*/
  41
  42 #ifdef DEBUG_IP_FIREWALL
  43 #define dprintf(format, args...) pr_info(format , ## args)
  44 #else
  45 #define dprintf(format, args...)
  46 #endif
  47
  48 #ifdef DEBUG_IP_FIREWALL_USER
  49 #define duprintf(format, args...) pr_info(format , ## args)
  50 #else
  51 #define duprintf(format, args...)
  52 #endif
  53
  54 #ifdef CONFIG_NETFILTER_DEBUG
  55 #define IP_NF_ASSERT(x)         WARN_ON(!(x))
  56 #else
  57 #define IP_NF_ASSERT(x)
  58 #endif
  59
  60 #if 0
  61 /* All the better to debug you with... */
  62 #define static
  63 #define inline
  64 #endif
  65
  66 void *ipt_alloc_initial_table(const struct xt_table *info)
  67 {
  68         return xt_alloc_initial_table(ipt, IPT);
  69 }
  70 EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  71
  72 /* Returns whether matches rule or not. */
  73 /* Performance critical - called for every packet */
  74 static inline bool
  75 ip_packet_match(const struct iphdr *ip,
  76                 const char *indev,
  77                 const char *outdev,
  78                 const struct ipt_ip *ipinfo,
  79                 int isfrag)
  80 {
  81         unsigned long ret;
  82
  83 #define FWINV(bool, invflg) ((bool) ^ !!(ipinfo->invflags & (invflg)))
  84
  85         if (FWINV((ip->saddr&ipinfo->smsk.s_addr) != ipinfo->src.s_addr,
  86                   IPT_INV_SRCIP) ||
  87             FWINV((ip->daddr&ipinfo->dmsk.s_addr) != ipinfo->dst.s_addr,
  88                   IPT_INV_DSTIP)) {
  89                 dprintf("Source or dest mismatch.\n");
  90
  91                 dprintf("SRC: %pI4. Mask: %pI4. Target: %pI4.%s\n",
  92                         &ip->saddr, &ipinfo->smsk.s_addr, &ipinfo->src.s_addr,
  93                         ipinfo->invflags & IPT_INV_SRCIP ? " (INV)" : "");
  94                 dprintf("DST: %pI4 Mask: %pI4 Target: %pI4.%s\n",
  95                         &ip->daddr, &ipinfo->dmsk.s_addr, &ipinfo->dst.s_addr,
  96                         ipinfo->invflags & IPT_INV_DSTIP ? " (INV)" : "");
  97                 return false;
  98         }
  99
 100         ret = ifname_compare_aligned(indev, ipinfo->iniface, ipinfo->iniface_mask);
 101
 102         if (FWINV(ret != 0, IPT_INV_VIA_IN)) {
 103                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 104                         indev, ipinfo->iniface,
 105                         ipinfo->invflags & IPT_INV_VIA_IN ? " (INV)" : "");
 106                 return false;
 107         }
 108
 109         ret = ifname_compare_aligned(outdev, ipinfo->outiface, ipinfo->outiface_mask);
 110
 111         if (FWINV(ret != 0, IPT_INV_VIA_OUT)) {
 112                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 113                         outdev, ipinfo->outiface,
 114                         ipinfo->invflags & IPT_INV_VIA_OUT ? " (INV)" : "");
 115                 return false;
 116         }
 117
 118         /* Check specific protocol */
 119         if (ipinfo->proto &&
 120             FWINV(ip->protocol != ipinfo->proto, IPT_INV_PROTO)) {
 121                 dprintf("Packet protocol %hi does not match %hi.%s\n",
 122                         ip->protocol, ipinfo->proto,
 123                         ipinfo->invflags & IPT_INV_PROTO ? " (INV)" : "");
 124                 return false;
 125         }
 126
 127         /* If we have a fragment rule but the packet is not a fragment
 128          * then we return zero */
 129         if (FWINV((ipinfo->flags&IPT_F_FRAG) && !isfrag, IPT_INV_FRAG)) {
 130                 dprintf("Fragment rule but not fragment.%s\n",
 131                         ipinfo->invflags & IPT_INV_FRAG ? " (INV)" : "");
 132                 return false;
 133         }
 134
 135         return true;
 136 }
 137
 138 static bool
 139 ip_checkentry(const struct ipt_ip *ip)
 140 {
 141         if (ip->flags & ~IPT_F_MASK) {
 142                 duprintf("Unknown flag bits set: %08X\n",
 143                          ip->flags & ~IPT_F_MASK);
 144                 return false;
 145         }
 146         if (ip->invflags & ~IPT_INV_MASK) {
 147                 duprintf("Unknown invflag bits set: %08X\n",
 148                          ip->invflags & ~IPT_INV_MASK);
 149                 return false;
 150         }
 151         return true;
 152 }
 153
 154 static unsigned int
 155 ipt_error(struct sk_buff *skb, const struct xt_action_param *par)
 156 {
 157         net_info_ratelimited("error: `%s'\n", (const char *)par->targinfo);
 158
 159         return NF_DROP;
 160 }
 161
 162 /* Performance critical */
 163 static inline struct ipt_entry *
 164 get_entry(const void *base, unsigned int offset)
 165 {
 166         return (struct ipt_entry *)(base + offset);
 167 }
 168
 169 /* All zeroes == unconditional rule. */
 170 /* Mildly perf critical (only if packet tracing is on) */
 171 static inline bool unconditional(const struct ipt_entry *e)
 172 {
 173         static const struct ipt_ip uncond;
 174
 175         return e->target_offset == sizeof(struct ipt_entry) &&
 176                memcmp(&e->ip, &uncond, sizeof(uncond)) == 0;
 177 #undef FWINV
 178 }
 179
 180 /* for const-correctness */
 181 static inline const struct xt_entry_target *
 182 ipt_get_target_c(const struct ipt_entry *e)
 183 {
 184         return ipt_get_target((struct ipt_entry *)e);
 185 }
 186
 187 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 188 static const char *const hooknames[] = {
 189         [NF_INET_PRE_ROUTING]           = "PREROUTING",
 190         [NF_INET_LOCAL_IN]              = "INPUT",
 191         [NF_INET_FORWARD]               = "FORWARD",
 192         [NF_INET_LOCAL_OUT]             = "OUTPUT",
 193         [NF_INET_POST_ROUTING]          = "POSTROUTING",
 194 };
 195
 196 enum nf_ip_trace_comments {
 197         NF_IP_TRACE_COMMENT_RULE,
 198         NF_IP_TRACE_COMMENT_RETURN,
 199         NF_IP_TRACE_COMMENT_POLICY,
 200 };
 201
 202 static const char *const comments[] = {
 203         [NF_IP_TRACE_COMMENT_RULE]      = "rule",
 204         [NF_IP_TRACE_COMMENT_RETURN]    = "return",
 205         [NF_IP_TRACE_COMMENT_POLICY]    = "policy",
 206 };
 207
 208 static struct nf_loginfo trace_loginfo = {
 209         .type = NF_LOG_TYPE_LOG,
 210         .u = {
 211                 .log = {
 212                         .level = 4,
 213                         .logflags = NF_LOG_MASK,
 214                 },
 215         },
 216 };
 217
 218 /* Mildly perf critical (only if packet tracing is on) */
 219 static inline int
 220 get_chainname_rulenum(const struct ipt_entry *s, const struct ipt_entry *e,
 221                       const char *hookname, const char **chainname,
 222                       const char **comment, unsigned int *rulenum)
 223 {
 224         const struct xt_standard_target *t = (void *)ipt_get_target_c(s);
 225
 226         if (strcmp(t->target.u.kernel.target->name, XT_ERROR_TARGET) == 0) {
 227                 /* Head of user chain: ERROR target with chainname */
 228                 *chainname = t->target.data;
 229                 (*rulenum) = 0;
 230         } else if (s == e) {
 231                 (*rulenum)++;
 232
 233                 if (unconditional(s) &&
 234                     strcmp(t->target.u.kernel.target->name,
 235                            XT_STANDARD_TARGET) == 0 &&
 236                    t->verdict < 0) {
 237                         /* Tail of chains: STANDARD target (return/policy) */
 238                         *comment = *chainname == hookname
 239                                 ? comments[NF_IP_TRACE_COMMENT_POLICY]
 240                                 : comments[NF_IP_TRACE_COMMENT_RETURN];
 241                 }
 242                 return 1;
 243         } else
 244                 (*rulenum)++;
 245
 246         return 0;
 247 }
 248
 249 static void trace_packet(struct net *net,
 250                          const struct sk_buff *skb,
 251                          unsigned int hook,
 252                          const struct net_device *in,
 253                          const struct net_device *out,
 254                          const char *tablename,
 255                          const struct xt_table_info *private,
 256                          const struct ipt_entry *e)
 257 {
 258         const struct ipt_entry *root;
 259         const char *hookname, *chainname, *comment;
 260         const struct ipt_entry *iter;
 261         unsigned int rulenum = 0;
 262
 263         root = get_entry(private->entries, private->hook_entry[hook]);
 264
 265         hookname = chainname = hooknames[hook];
 266         comment = comments[NF_IP_TRACE_COMMENT_RULE];
 267
 268         xt_entry_foreach(iter, root, private->size - private->hook_entry[hook])
 269                 if (get_chainname_rulenum(iter, e, hookname,
 270                     &chainname, &comment, &rulenum) != 0)
 271                         break;
 272
 273         nf_log_trace(net, AF_INET, hook, skb, in, out, &trace_loginfo,
 274                      "TRACE: %s:%s:%s:%u ",
 275                      tablename, chainname, comment, rulenum);
 276 }
 277 #endif
 278
 279 static inline
 280 struct ipt_entry *ipt_next_entry(const struct ipt_entry *entry)
 281 {
 282         return (void *)entry + entry->next_offset;
 283 }
 284
 285 /* Returns one of the generic firewall policies, like NF_ACCEPT. */
 286 unsigned int
 287 ipt_do_table(struct sk_buff *skb,
 288              const struct nf_hook_state *state,
 289              struct xt_table *table)
 290 {
 291         unsigned int hook = state->hook;
 292         static const char nulldevname[IFNAMSIZ] __attribute__((aligned(sizeof(long))));
 293         const struct iphdr *ip;
 294         /* Initializing verdict to NF_DROP keeps gcc happy. */
 295         unsigned int verdict = NF_DROP;
 296         const char *indev, *outdev;
 297         const void *table_base;
 298         struct ipt_entry *e, **jumpstack;
 299         unsigned int stackidx, cpu;
 300         const struct xt_table_info *private;
 301         struct xt_action_param acpar;
 302         unsigned int addend;
 303
 304         /* Initialization */
 305         stackidx = 0;
 306         ip = ip_hdr(skb);
 307         indev = state->in ? state->in->name : nulldevname;
 308         outdev = state->out ? state->out->name : nulldevname;
 309         /* We handle fragments by dealing with the first fragment as
 310          * if it was a normal packet.  All other fragments are treated
 311          * normally, except that they will NEVER match rules that ask
 312          * things we don't know, ie. tcp syn flag or ports).  If the
 313          * rule is also a fragment-specific rule, non-fragments won't
 314          * match it. */
 315         acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 316         acpar.thoff   = ip_hdrlen(skb);
 317         acpar.hotdrop = false;
 318         acpar.net     = state->net;
 319         acpar.in      = state->in;
 320         acpar.out     = state->out;
 321         acpar.family  = NFPROTO_IPV4;
 322         acpar.hooknum = hook;
 323
 324         IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 325         local_bh_disable();
 326         addend = xt_write_recseq_begin();
 327         private = table->private;
 328         cpu        = smp_processor_id();
 329         /*
 330          * Ensure we load private-> members after we've fetched the base
 331          * pointer.
 332          */
 333         smp_read_barrier_depends();
 334         table_base = private->entries;
 335         jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 336
 337         /* Switch to alternate jumpstack if we're being invoked via TEE.
 338          * TEE issues XT_CONTINUE verdict on original skb so we must not
 339          * clobber the jumpstack.
 340          *
 341          * For recursion via REJECT or SYNPROXY the stack will be clobbered
 342          * but it is no problem since absolute verdict is issued by these.
 343          */
 344         if (static_key_false(&xt_tee_enabled))
 345                 jumpstack += private->stacksize * __this_cpu_read(nf_skb_duplicated);
 346
 347         e = get_entry(table_base, private->hook_entry[hook]);
 348
 349         pr_debug("Entering %s(hook %u), UF %p\n",
 350                  table->name, hook,
 351                  get_entry(table_base, private->underflow[hook]));
 352
 353         do {
 354                 const struct xt_entry_target *t;
 355                 const struct xt_entry_match *ematch;
 356                 struct xt_counters *counter;
 357
 358                 IP_NF_ASSERT(e);
 359                 if (!ip_packet_match(ip, indev, outdev,
 360                     &e->ip, acpar.fragoff)) {
 361  no_match:
 362                         e = ipt_next_entry(e);
 363                         continue;
 364                 }
 365
 366                 xt_ematch_foreach(ematch, e) {
 367                         acpar.match     = ematch->u.kernel.match;
 368                         acpar.matchinfo = ematch->data;
 369                         if (!acpar.match->match(skb, &acpar))
 370                                 goto no_match;
 371                 }
 372
 373                 counter = xt_get_this_cpu_counter(&e->counters);
 374                 ADD_COUNTER(*counter, skb->len, 1);
 375
 376                 t = ipt_get_target(e);
 377                 IP_NF_ASSERT(t->u.kernel.target);
 378
 379 #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
 380                 /* The packet is traced: log it */
 381                 if (unlikely(skb->nf_trace))
 382                         trace_packet(state->net, skb, hook, state->in,
 383                                      state->out, table->name, private, e);
 384 #endif
 385                 /* Standard target? */
 386                 if (!t->u.kernel.target->target) {
 387                         int v;
 388
 389                         v = ((struct xt_standard_target *)t)->verdict;
 390                         if (v < 0) {
 391                                 /* Pop from stack? */
 392                                 if (v != XT_RETURN) {
 393                                         verdict = (unsigned int)(-v) - 1;
 394                                         break;
 395                                 }
 396                                 if (stackidx == 0) {
 397                                         e = get_entry(table_base,
 398                                             private->underflow[hook]);
 399                                         pr_debug("Underflow (this is normal) "
 400                                                  "to %p\n", e);
 401                                 } else {
 402                                         e = jumpstack[--stackidx];
 403                                         pr_debug("Pulled %p out from pos %u\n",
 404                                                  e, stackidx);
 405                                         e = ipt_next_entry(e);
 406                                 }
 407                                 continue;
 408                         }
 409                         if (table_base + v != ipt_next_entry(e) &&
 410                             !(e->ip.flags & IPT_F_GOTO)) {
 411                                 jumpstack[stackidx++] = e;
 412                                 pr_debug("Pushed %p into pos %u\n",
 413                                          e, stackidx - 1);
 414                         }
 415
 416                         e = get_entry(table_base, v);
 417                         continue;
 418                 }
 419
 420                 acpar.target   = t->u.kernel.target;
 421                 acpar.targinfo = t->data;
 422
 423                 verdict = t->u.kernel.target->target(skb, &acpar);
 424                 /* Target might have changed stuff. */
 425                 ip = ip_hdr(skb);
 426                 if (verdict == XT_CONTINUE)
 427                         e = ipt_next_entry(e);
 428                 else
 429                         /* Verdict */
 430                         break;
 431         } while (!acpar.hotdrop);
 432         pr_debug("Exiting %s; sp at %u\n", __func__, stackidx);
 433
 434         xt_write_recseq_end(addend);
 435         local_bh_enable();
 436
 437 #ifdef DEBUG_ALLOW_ALL
 438         return NF_ACCEPT;
 439 #else
 440         if (acpar.hotdrop)
 441                 return NF_DROP;
 442         else return verdict;
 443 #endif
 444 }
 445
 446 /* Figures out from what hook each rule can be called: returns 0 if
 447    there are loops.  Puts hook bitmask in comefrom. */
 448 static int
 449 mark_source_chains(const struct xt_table_info *newinfo,
 450                    unsigned int valid_hooks, void *entry0)
 451 {
 452         unsigned int hook;
 453
 454         /* No recursion; use packet counter to save back ptrs (reset
 455            to 0 as we leave), and comefrom to save source hook bitmask */
 456         for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 457                 unsigned int pos = newinfo->hook_entry[hook];
 458                 struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 459
 460                 if (!(valid_hooks & (1 << hook)))
 461                         continue;
 462
 463                 /* Set initial back pointer. */
 464                 e->counters.pcnt = pos;
 465
 466                 for (;;) {
 467                         const struct xt_standard_target *t
 468                                 = (void *)ipt_get_target_c(e);
 469                         int visited = e->comefrom & (1 << hook);
 470
 471                         if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 472                                 pr_err("iptables: loop hook %u pos %u %08X.\n",
 473                                        hook, pos, e->comefrom);
 474                                 return 0;
 475                         }
 476                         e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 477
 478                         /* Unconditional return/END. */
 479                         if ((unconditional(e) &&
 480                              (strcmp(t->target.u.user.name,
 481                                      XT_STANDARD_TARGET) == 0) &&
 482                              t->verdict < 0) || visited) {
 483                                 unsigned int oldpos, size;
 484
 485                                 if ((strcmp(t->target.u.user.name,
 486                                             XT_STANDARD_TARGET) == 0) &&
 487                                     t->verdict < -NF_MAX_VERDICT - 1) {
 488                                         duprintf("mark_source_chains: bad "
 489                                                 "negative verdict (%i)\n",
 490                                                                 t->verdict);
 491                                         return 0;
 492                                 }
 493
 494                                 /* Return: backtrack through the last
 495                                    big jump. */
 496                                 do {
 497                                         e->comefrom ^= (1<<NF_INET_NUMHOOKS);
 498 #ifdef DEBUG_IP_FIREWALL_USER
 499                                         if (e->comefrom
 500                                             & (1 << NF_INET_NUMHOOKS)) {
 501                                                 duprintf("Back unset "
 502                                                          "on hook %u "
 503                                                          "rule %u\n",
 504                                                          hook, pos);
 505                                         }
 506 #endif
 507                                         oldpos = pos;
 508                                         pos = e->counters.pcnt;
 509                                         e->counters.pcnt = 0;
 510
 511                                         /* We're at the start. */
 512                                         if (pos == oldpos)
 513                                                 goto next;
 514
 515                                         e = (struct ipt_entry *)
 516                                                 (entry0 + pos);
 517                                 } while (oldpos == pos + e->next_offset);
 518
 519                                 /* Move along one */
 520                                 size = e->next_offset;
 521                                 e = (struct ipt_entry *)
 522                                         (entry0 + pos + size);
 523                                 e->counters.pcnt = pos;
 524                                 pos += size;
 525                         } else {
 526                                 int newpos = t->verdict;
 527
 528                                 if (strcmp(t->target.u.user.name,
 529                                            XT_STANDARD_TARGET) == 0 &&
 530                                     newpos >= 0) {
 531                                         if (newpos > newinfo->size -
 532                                                 sizeof(struct ipt_entry)) {
 533                                                 duprintf("mark_source_chains: "
 534                                                         "bad verdict (%i)\n",
 535                                                                 newpos);
 536                                                 return 0;
 537                                         }
 538                                         /* This a jump; chase it. */
 539                                         duprintf("Jump rule %u -> %u\n",
 540                                                  pos, newpos);
 541                                 } else {
 542                                         /* ... this is a fallthru */
 543                                         newpos = pos + e->next_offset;
 544                                 }
 545                                 e = (struct ipt_entry *)
 546                                         (entry0 + newpos);
 547                                 e->counters.pcnt = pos;
 548                                 pos = newpos;
 549                         }
 550                 }
 551 next:
 552                 duprintf("Finished chain %u\n", hook);
 553         }
 554         return 1;
 555 }
 556
 557 static void cleanup_match(struct xt_entry_match *m, struct net *net)
 558 {
 559         struct xt_mtdtor_param par;
 560
 561         par.net       = net;
 562         par.match     = m->u.kernel.match;
 563         par.matchinfo = m->data;
 564         par.family    = NFPROTO_IPV4;
 565         if (par.match->destroy != NULL)
 566                 par.match->destroy(&par);
 567         module_put(par.match->me);
 568 }
 569
 570 static int
 571 check_entry(const struct ipt_entry *e)
 572 {
 573         const struct xt_entry_target *t;
 574
 575         if (!ip_checkentry(&e->ip))
 576                 return -EINVAL;
 577
 578         if (e->target_offset + sizeof(struct xt_entry_target) >
 579             e->next_offset)
 580                 return -EINVAL;
 581
 582         t = ipt_get_target_c(e);
 583         if (e->target_offset + t->u.target_size > e->next_offset)
 584                 return -EINVAL;
 585
 586         return 0;
 587 }
 588
 589 static int
 590 check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 591 {
 592         const struct ipt_ip *ip = par->entryinfo;
 593         int ret;
 594
 595         par->match     = m->u.kernel.match;
 596         par->matchinfo = m->data;
 597
 598         ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 599               ip->proto, ip->invflags & IPT_INV_PROTO);
 600         if (ret < 0) {
 601                 duprintf("check failed for `%s'.\n", par->match->name);
 602                 return ret;
 603         }
 604         return 0;
 605 }
 606
 607 static int
 608 find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 609 {
 610         struct xt_match *match;
 611         int ret;
 612
 613         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 614                                       m->u.user.revision);
 615         if (IS_ERR(match)) {
 616                 duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 617                 return PTR_ERR(match);
 618         }
 619         m->u.kernel.match = match;
 620
 621         ret = check_match(m, par);
 622         if (ret)
 623                 goto err;
 624
 625         return 0;
 626 err:
 627         module_put(m->u.kernel.match->me);
 628         return ret;
 629 }
 630
 631 static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 632 {
 633         struct xt_entry_target *t = ipt_get_target(e);
 634         struct xt_tgchk_param par = {
 635                 .net       = net,
 636                 .table     = name,
 637                 .entryinfo = e,
 638                 .target    = t->u.kernel.target,
 639                 .targinfo  = t->data,
 640                 .hook_mask = e->comefrom,
 641                 .family    = NFPROTO_IPV4,
 642         };
 643         int ret;
 644
 645         ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 646               e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 647         if (ret < 0) {
 648                 duprintf("check failed for `%s'.\n",
 649                          t->u.kernel.target->name);
 650                 return ret;
 651         }
 652         return 0;
 653 }
 654
 655 static int
 656 find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 657                  unsigned int size)
 658 {
 659         struct xt_entry_target *t;
 660         struct xt_target *target;
 661         int ret;
 662         unsigned int j;
 663         struct xt_mtchk_param mtpar;
 664         struct xt_entry_match *ematch;
 665
 666         e->counters.pcnt = xt_percpu_counter_alloc();
 667         if (IS_ERR_VALUE(e->counters.pcnt))
 668                 return -ENOMEM;
 669
 670         j = 0;
 671         mtpar.net       = net;
 672         mtpar.table     = name;
 673         mtpar.entryinfo = &e->ip;
 674         mtpar.hook_mask = e->comefrom;
 675         mtpar.family    = NFPROTO_IPV4;
 676         xt_ematch_foreach(ematch, e) {
 677                 ret = find_check_match(ematch, &mtpar);
 678                 if (ret != 0)
 679                         goto cleanup_matches;
 680                 ++j;
 681         }
 682
 683         t = ipt_get_target(e);
 684         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 685                                         t->u.user.revision);
 686         if (IS_ERR(target)) {
 687                 duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 688                 ret = PTR_ERR(target);
 689                 goto cleanup_matches;
 690         }
 691         t->u.kernel.target = target;
 692
 693         ret = check_target(e, net, name);
 694         if (ret)
 695                 goto err;
 696
 697         return 0;
 698  err:
 699         module_put(t->u.kernel.target->me);
 700  cleanup_matches:
 701         xt_ematch_foreach(ematch, e) {
 702                 if (j-- == 0)
 703                         break;
 704                 cleanup_match(ematch, net);
 705         }
 706
 707         xt_percpu_counter_free(e->counters.pcnt);
 708
 709         return ret;
 710 }
 711
 712 static bool check_underflow(const struct ipt_entry *e)
 713 {
 714         const struct xt_entry_target *t;
 715         unsigned int verdict;
 716
 717         if (!unconditional(e))
 718                 return false;
 719         t = ipt_get_target_c(e);
 720         if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 721                 return false;
 722         verdict = ((struct xt_standard_target *)t)->verdict;
 723         verdict = -verdict - 1;
 724         return verdict == NF_DROP || verdict == NF_ACCEPT;
 725 }
 726
 727 static int
 728 check_entry_size_and_hooks(struct ipt_entry *e,
 729                            struct xt_table_info *newinfo,
 730                            const unsigned char *base,
 731                            const unsigned char *limit,
 732                            const unsigned int *hook_entries,
 733                            const unsigned int *underflows,
 734                            unsigned int valid_hooks)
 735 {
 736         unsigned int h;
 737         int err;
 738
 739         if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 740             (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 741             (unsigned char *)e + e->next_offset > limit) {
 742                 duprintf("Bad offset %p\n", e);
 743                 return -EINVAL;
 744         }
 745
 746         if (e->next_offset
 747             < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 748                 duprintf("checking: element %p size %u\n",
 749                          e, e->next_offset);
 750                 return -EINVAL;
 751         }
 752
 753         err = check_entry(e);
 754         if (err)
 755                 return err;
 756
 757         /* Check hooks & underflows */
 758         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 759                 if (!(valid_hooks & (1 << h)))
 760                         continue;
 761                 if ((unsigned char *)e - base == hook_entries[h])
 762                         newinfo->hook_entry[h] = hook_entries[h];
 763                 if ((unsigned char *)e - base == underflows[h]) {
 764                         if (!check_underflow(e)) {
 765                                 pr_debug("Underflows must be unconditional and "
 766                                          "use the STANDARD target with "
 767                                          "ACCEPT/DROP\n");
 768                                 return -EINVAL;
 769                         }
 770                         newinfo->underflow[h] = underflows[h];
 771                 }
 772         }
 773
 774         /* Clear counters and comefrom */
 775         e->counters = ((struct xt_counters) { 0, 0 });
 776         e->comefrom = 0;
 777         return 0;
 778 }
 779
 780 static void
 781 cleanup_entry(struct ipt_entry *e, struct net *net)
 782 {
 783         struct xt_tgdtor_param par;
 784         struct xt_entry_target *t;
 785         struct xt_entry_match *ematch;
 786
 787         /* Cleanup all matches */
 788         xt_ematch_foreach(ematch, e)
 789                 cleanup_match(ematch, net);
 790         t = ipt_get_target(e);
 791
 792         par.net      = net;
 793         par.target   = t->u.kernel.target;
 794         par.targinfo = t->data;
 795         par.family   = NFPROTO_IPV4;
 796         if (par.target->destroy != NULL)
 797                 par.target->destroy(&par);
 798         module_put(par.target->me);
 799         xt_percpu_counter_free(e->counters.pcnt);
 800 }
 801
 802 /* Checks and translates the user-supplied table segment (held in
 803    newinfo) */
 804 static int
 805 translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 806                 const struct ipt_replace *repl)
 807 {
 808         struct ipt_entry *iter;
 809         unsigned int i;
 810         int ret = 0;
 811
 812         newinfo->size = repl->size;
 813         newinfo->number = repl->num_entries;
 814
 815         /* Init all hooks to impossible value. */
 816         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 817                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 818                 newinfo->underflow[i] = 0xFFFFFFFF;
 819         }
 820
 821         duprintf("translate_table: size %u\n", newinfo->size);
 822         i = 0;
 823         /* Walk through entries, checking offsets. */
 824         xt_entry_foreach(iter, entry0, newinfo->size) {
 825                 ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 826                                                  entry0 + repl->size,
 827                                                  repl->hook_entry,
 828                                                  repl->underflow,
 829                                                  repl->valid_hooks);
 830                 if (ret != 0)
 831                         return ret;
 832                 ++i;
 833                 if (strcmp(ipt_get_target(iter)->u.user.name,
 834                     XT_ERROR_TARGET) == 0)
 835                         ++newinfo->stacksize;
 836         }
 837
 838         if (i != repl->num_entries) {
 839                 duprintf("translate_table: %u not %u entries\n",
 840                          i, repl->num_entries);
 841                 return -EINVAL;
 842         }
 843
 844         /* Check hooks all assigned */
 845         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 846                 /* Only hooks which are valid */
 847                 if (!(repl->valid_hooks & (1 << i)))
 848                         continue;
 849                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 850                         duprintf("Invalid hook entry %u %u\n",
 851                                  i, repl->hook_entry[i]);
 852                         return -EINVAL;
 853                 }
 854                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 855                         duprintf("Invalid underflow %u %u\n",
 856                                  i, repl->underflow[i]);
 857                         return -EINVAL;
 858                 }
 859         }
 860
 861         if (!mark_source_chains(newinfo, repl->valid_hooks, entry0))
 862                 return -ELOOP;
 863
 864         /* Finally, each sanity check must pass */
 865         i = 0;
 866         xt_entry_foreach(iter, entry0, newinfo->size) {
 867                 ret = find_check_entry(iter, net, repl->name, repl->size);
 868                 if (ret != 0)
 869                         break;
 870                 ++i;
 871         }
 872
 873         if (ret != 0) {
 874                 xt_entry_foreach(iter, entry0, newinfo->size) {
 875                         if (i-- == 0)
 876                                 break;
 877                         cleanup_entry(iter, net);
 878                 }
 879                 return ret;
 880         }
 881
 882         return ret;
 883 }
 884
 885 static void
 886 get_counters(const struct xt_table_info *t,
 887              struct xt_counters counters[])
 888 {
 889         struct ipt_entry *iter;
 890         unsigned int cpu;
 891         unsigned int i;
 892
 893         for_each_possible_cpu(cpu) {
 894                 seqcount_t *s = &per_cpu(xt_recseq, cpu);
 895
 896                 i = 0;
 897                 xt_entry_foreach(iter, t->entries, t->size) {
 898                         struct xt_counters *tmp;
 899                         u64 bcnt, pcnt;
 900                         unsigned int start;
 901
 902                         tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 903                         do {
 904                                 start = read_seqcount_begin(s);
 905                                 bcnt = tmp->bcnt;
 906                                 pcnt = tmp->pcnt;
 907                         } while (read_seqcount_retry(s, start));
 908
 909                         ADD_COUNTER(counters[i], bcnt, pcnt);
 910                         ++i; /* macro does multi eval of i */
 911                 }
 912         }
 913 }
 914
 915 static struct xt_counters *alloc_counters(const struct xt_table *table)
 916 {
 917         unsigned int countersize;
 918         struct xt_counters *counters;
 919         const struct xt_table_info *private = table->private;
 920
 921         /* We need atomic snapshot of counters: rest doesn't change
 922            (other than comefrom, which userspace doesn't care
 923            about). */
 924         countersize = sizeof(struct xt_counters) * private->number;
 925         counters = vzalloc(countersize);
 926
 927         if (counters == NULL)
 928                 return ERR_PTR(-ENOMEM);
 929
 930         get_counters(private, counters);
 931
 932         return counters;
 933 }
 934
 935 static int
 936 copy_entries_to_user(unsigned int total_size,
 937                      const struct xt_table *table,
 938                      void __user *userptr)
 939 {
 940         unsigned int off, num;
 941         const struct ipt_entry *e;
 942         struct xt_counters *counters;
 943         const struct xt_table_info *private = table->private;
 944         int ret = 0;
 945         const void *loc_cpu_entry;
 946
 947         counters = alloc_counters(table);
 948         if (IS_ERR(counters))
 949                 return PTR_ERR(counters);
 950
 951         loc_cpu_entry = private->entries;
 952         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 953                 ret = -EFAULT;
 954                 goto free_counters;
 955         }
 956
 957         /* FIXME: use iterator macros --RR */
 958         /* ... then go back and fix counters and names */
 959         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 960                 unsigned int i;
 961                 const struct xt_entry_match *m;
 962                 const struct xt_entry_target *t;
 963
 964                 e = (struct ipt_entry *)(loc_cpu_entry + off);
 965                 if (copy_to_user(userptr + off
 966                                  + offsetof(struct ipt_entry, counters),
 967                                  &counters[num],
 968                                  sizeof(counters[num])) != 0) {
 969                         ret = -EFAULT;
 970                         goto free_counters;
 971                 }
 972
 973                 for (i = sizeof(struct ipt_entry);
 974                      i < e->target_offset;
 975                      i += m->u.match_size) {
 976                         m = (void *)e + i;
 977
 978                         if (copy_to_user(userptr + off + i
 979                                          + offsetof(struct xt_entry_match,
 980                                                     u.user.name),
 981                                          m->u.kernel.match->name,
 982                                          strlen(m->u.kernel.match->name)+1)
 983                             != 0) {
 984                                 ret = -EFAULT;
 985                                 goto free_counters;
 986                         }
 987                 }
 988
 989                 t = ipt_get_target_c(e);
 990                 if (copy_to_user(userptr + off + e->target_offset
 991                                  + offsetof(struct xt_entry_target,
 992                                             u.user.name),
 993                                  t->u.kernel.target->name,
 994                                  strlen(t->u.kernel.target->name)+1) != 0) {
 995                         ret = -EFAULT;
 996                         goto free_counters;
 997                 }
 998         }
 999
1000  free_counters:
1001         vfree(counters);
1002         return ret;
1003 }
1004
1005 #ifdef CONFIG_COMPAT
1006 static void compat_standard_from_user(void *dst, const void *src)
1007 {
1008         int v = *(compat_int_t *)src;
1009
1010         if (v > 0)
1011                 v += xt_compat_calc_jump(AF_INET, v);
1012         memcpy(dst, &v, sizeof(v));
1013 }
1014
1015 static int compat_standard_to_user(void __user *dst, const void *src)
1016 {
1017         compat_int_t cv = *(int *)src;
1018
1019         if (cv > 0)
1020                 cv -= xt_compat_calc_jump(AF_INET, cv);
1021         return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1022 }
1023
1024 static int compat_calc_entry(const struct ipt_entry *e,
1025                              const struct xt_table_info *info,
1026                              const void *base, struct xt_table_info *newinfo)
1027 {
1028         const struct xt_entry_match *ematch;
1029         const struct xt_entry_target *t;
1030         unsigned int entry_offset;
1031         int off, i, ret;
1032
1033         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1034         entry_offset = (void *)e - base;
1035         xt_ematch_foreach(ematch, e)
1036                 off += xt_compat_match_offset(ematch->u.kernel.match);
1037         t = ipt_get_target_c(e);
1038         off += xt_compat_target_offset(t->u.kernel.target);
1039         newinfo->size -= off;
1040         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1041         if (ret)
1042                 return ret;
1043
1044         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1045                 if (info->hook_entry[i] &&
1046                     (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1047                         newinfo->hook_entry[i] -= off;
1048                 if (info->underflow[i] &&
1049                     (e < (struct ipt_entry *)(base + info->underflow[i])))
1050                         newinfo->underflow[i] -= off;
1051         }
1052         return 0;
1053 }
1054
1055 static int compat_table_info(const struct xt_table_info *info,
1056                              struct xt_table_info *newinfo)
1057 {
1058         struct ipt_entry *iter;
1059         const void *loc_cpu_entry;
1060         int ret;
1061
1062         if (!newinfo || !info)
1063                 return -EINVAL;
1064
1065         /* we dont care about newinfo->entries */
1066         memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1067         newinfo->initial_entries = 0;
1068         loc_cpu_entry = info->entries;
1069         xt_compat_init_offsets(AF_INET, info->number);
1070         xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1071                 ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1072                 if (ret != 0)
1073                         return ret;
1074         }
1075         return 0;
1076 }
1077 #endif
1078
1079 static int get_info(struct net *net, void __user *user,
1080                     const int *len, int compat)
1081 {
1082         char name[XT_TABLE_MAXNAMELEN];
1083         struct xt_table *t;
1084         int ret;
1085
1086         if (*len != sizeof(struct ipt_getinfo)) {
1087                 duprintf("length %u != %zu\n", *len,
1088                          sizeof(struct ipt_getinfo));
1089                 return -EINVAL;
1090         }
1091
1092         if (copy_from_user(name, user, sizeof(name)) != 0)
1093                 return -EFAULT;
1094
1095         name[XT_TABLE_MAXNAMELEN-1] = '\0';
1096 #ifdef CONFIG_COMPAT
1097         if (compat)
1098                 xt_compat_lock(AF_INET);
1099 #endif
1100         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1101                                     "iptable_%s", name);
1102         if (!IS_ERR_OR_NULL(t)) {
1103                 struct ipt_getinfo info;
1104                 const struct xt_table_info *private = t->private;
1105 #ifdef CONFIG_COMPAT
1106                 struct xt_table_info tmp;
1107
1108                 if (compat) {
1109                         ret = compat_table_info(private, &tmp);
1110                         xt_compat_flush_offsets(AF_INET);
1111                         private = &tmp;
1112                 }
1113 #endif
1114                 memset(&info, 0, sizeof(info));
1115                 info.valid_hooks = t->valid_hooks;
1116                 memcpy(info.hook_entry, private->hook_entry,
1117                        sizeof(info.hook_entry));
1118                 memcpy(info.underflow, private->underflow,
1119                        sizeof(info.underflow));
1120                 info.num_entries = private->number;
1121                 info.size = private->size;
1122                 strcpy(info.name, name);
1123
1124                 if (copy_to_user(user, &info, *len) != 0)
1125                         ret = -EFAULT;
1126                 else
1127                         ret = 0;
1128
1129                 xt_table_unlock(t);
1130                 module_put(t->me);
1131         } else
1132                 ret = t ? PTR_ERR(t) : -ENOENT;
1133 #ifdef CONFIG_COMPAT
1134         if (compat)
1135                 xt_compat_unlock(AF_INET);
1136 #endif
1137         return ret;
1138 }
1139
1140 static int
1141 get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1142             const int *len)
1143 {
1144         int ret;
1145         struct ipt_get_entries get;
1146         struct xt_table *t;
1147
1148         if (*len < sizeof(get)) {
1149                 duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1150                 return -EINVAL;
1151         }
1152         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1153                 return -EFAULT;
1154         if (*len != sizeof(struct ipt_get_entries) + get.size) {
1155                 duprintf("get_entries: %u != %zu\n",
1156                          *len, sizeof(get) + get.size);
1157                 return -EINVAL;
1158         }
1159         get.name[sizeof(get.name) - 1] = '\0';
1160
1161         t = xt_find_table_lock(net, AF_INET, get.name);
1162         if (!IS_ERR_OR_NULL(t)) {
1163                 const struct xt_table_info *private = t->private;
1164                 duprintf("t->private->number = %u\n", private->number);
1165                 if (get.size == private->size)
1166                         ret = copy_entries_to_user(private->size,
1167                                                    t, uptr->entrytable);
1168                 else {
1169                         duprintf("get_entries: I've got %u not %u!\n",
1170                                  private->size, get.size);
1171                         ret = -EAGAIN;
1172                 }
1173                 module_put(t->me);
1174                 xt_table_unlock(t);
1175         } else
1176                 ret = t ? PTR_ERR(t) : -ENOENT;
1177
1178         return ret;
1179 }
1180
1181 static int
1182 __do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1183              struct xt_table_info *newinfo, unsigned int num_counters,
1184              void __user *counters_ptr)
1185 {
1186         int ret;
1187         struct xt_table *t;
1188         struct xt_table_info *oldinfo;
1189         struct xt_counters *counters;
1190         struct ipt_entry *iter;
1191
1192         ret = 0;
1193         counters = vzalloc(num_counters * sizeof(struct xt_counters));
1194         if (!counters) {
1195                 ret = -ENOMEM;
1196                 goto out;
1197         }
1198
1199         t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1200                                     "iptable_%s", name);
1201         if (IS_ERR_OR_NULL(t)) {
1202                 ret = t ? PTR_ERR(t) : -ENOENT;
1203                 goto free_newinfo_counters_untrans;
1204         }
1205
1206         /* You lied! */
1207         if (valid_hooks != t->valid_hooks) {
1208                 duprintf("Valid hook crap: %08X vs %08X\n",
1209                          valid_hooks, t->valid_hooks);
1210                 ret = -EINVAL;
1211                 goto put_module;
1212         }
1213
1214         oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1215         if (!oldinfo)
1216                 goto put_module;
1217
1218         /* Update module usage count based on number of rules */
1219         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1220                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
1221         if ((oldinfo->number > oldinfo->initial_entries) ||
1222             (newinfo->number <= oldinfo->initial_entries))
1223                 module_put(t->me);
1224         if ((oldinfo->number > oldinfo->initial_entries) &&
1225             (newinfo->number <= oldinfo->initial_entries))
1226                 module_put(t->me);
1227
1228         /* Get the old counters, and synchronize with replace */
1229         get_counters(oldinfo, counters);
1230
1231         /* Decrease module usage counts and free resource */
1232         xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1233                 cleanup_entry(iter, net);
1234
1235         xt_free_table_info(oldinfo);
1236         if (copy_to_user(counters_ptr, counters,
1237                          sizeof(struct xt_counters) * num_counters) != 0) {
1238                 /* Silent error, can't fail, new table is already in place */
1239                 net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1240         }
1241         vfree(counters);
1242         xt_table_unlock(t);
1243         return ret;
1244
1245  put_module:
1246         module_put(t->me);
1247         xt_table_unlock(t);
1248  free_newinfo_counters_untrans:
1249         vfree(counters);
1250  out:
1251         return ret;
1252 }
1253
1254 static int
1255 do_replace(struct net *net, const void __user *user, unsigned int len)
1256 {
1257         int ret;
1258         struct ipt_replace tmp;
1259         struct xt_table_info *newinfo;
1260         void *loc_cpu_entry;
1261         struct ipt_entry *iter;
1262
1263         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1264                 return -EFAULT;
1265
1266         /* overflow check */
1267         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1268                 return -ENOMEM;
1269         if (tmp.num_counters == 0)
1270                 return -EINVAL;
1271
1272         tmp.name[sizeof(tmp.name)-1] = 0;
1273
1274         newinfo = xt_alloc_table_info(tmp.size);
1275         if (!newinfo)
1276                 return -ENOMEM;
1277
1278         loc_cpu_entry = newinfo->entries;
1279         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1280                            tmp.size) != 0) {
1281                 ret = -EFAULT;
1282                 goto free_newinfo;
1283         }
1284
1285         ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1286         if (ret != 0)
1287                 goto free_newinfo;
1288
1289         duprintf("Translated table\n");
1290
1291         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1292                            tmp.num_counters, tmp.counters);
1293         if (ret)
1294                 goto free_newinfo_untrans;
1295         return 0;
1296
1297  free_newinfo_untrans:
1298         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1299                 cleanup_entry(iter, net);
1300  free_newinfo:
1301         xt_free_table_info(newinfo);
1302         return ret;
1303 }
1304
1305 static int
1306 do_add_counters(struct net *net, const void __user *user,
1307                 unsigned int len, int compat)
1308 {
1309         unsigned int i;
1310         struct xt_counters_info tmp;
1311         struct xt_counters *paddc;
1312         unsigned int num_counters;
1313         const char *name;
1314         int size;
1315         void *ptmp;
1316         struct xt_table *t;
1317         const struct xt_table_info *private;
1318         int ret = 0;
1319         struct ipt_entry *iter;
1320         unsigned int addend;
1321 #ifdef CONFIG_COMPAT
1322         struct compat_xt_counters_info compat_tmp;
1323
1324         if (compat) {
1325                 ptmp = &compat_tmp;
1326                 size = sizeof(struct compat_xt_counters_info);
1327         } else
1328 #endif
1329         {
1330                 ptmp = &tmp;
1331                 size = sizeof(struct xt_counters_info);
1332         }
1333
1334         if (copy_from_user(ptmp, user, size) != 0)
1335                 return -EFAULT;
1336
1337 #ifdef CONFIG_COMPAT
1338         if (compat) {
1339                 num_counters = compat_tmp.num_counters;
1340                 name = compat_tmp.name;
1341         } else
1342 #endif
1343         {
1344                 num_counters = tmp.num_counters;
1345                 name = tmp.name;
1346         }
1347
1348         if (len != size + num_counters * sizeof(struct xt_counters))
1349                 return -EINVAL;
1350
1351         paddc = vmalloc(len - size);
1352         if (!paddc)
1353                 return -ENOMEM;
1354
1355         if (copy_from_user(paddc, user + size, len - size) != 0) {
1356                 ret = -EFAULT;
1357                 goto free;
1358         }
1359
1360         t = xt_find_table_lock(net, AF_INET, name);
1361         if (IS_ERR_OR_NULL(t)) {
1362                 ret = t ? PTR_ERR(t) : -ENOENT;
1363                 goto free;
1364         }
1365
1366         local_bh_disable();
1367         private = t->private;
1368         if (private->number != num_counters) {
1369                 ret = -EINVAL;
1370                 goto unlock_up_free;
1371         }
1372
1373         i = 0;
1374         addend = xt_write_recseq_begin();
1375         xt_entry_foreach(iter, private->entries, private->size) {
1376                 struct xt_counters *tmp;
1377
1378                 tmp = xt_get_this_cpu_counter(&iter->counters);
1379                 ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1380                 ++i;
1381         }
1382         xt_write_recseq_end(addend);
1383  unlock_up_free:
1384         local_bh_enable();
1385         xt_table_unlock(t);
1386         module_put(t->me);
1387  free:
1388         vfree(paddc);
1389
1390         return ret;
1391 }
1392
1393 #ifdef CONFIG_COMPAT
1394 struct compat_ipt_replace {
1395         char                    name[XT_TABLE_MAXNAMELEN];
1396         u32                     valid_hooks;
1397         u32                     num_entries;
1398         u32                     size;
1399         u32                     hook_entry[NF_INET_NUMHOOKS];
1400         u32                     underflow[NF_INET_NUMHOOKS];
1401         u32                     num_counters;
1402         compat_uptr_t           counters;       /* struct xt_counters * */
1403         struct compat_ipt_entry entries[0];
1404 };
1405
1406 static int
1407 compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1408                           unsigned int *size, struct xt_counters *counters,
1409                           unsigned int i)
1410 {
1411         struct xt_entry_target *t;
1412         struct compat_ipt_entry __user *ce;
1413         u_int16_t target_offset, next_offset;
1414         compat_uint_t origsize;
1415         const struct xt_entry_match *ematch;
1416         int ret = 0;
1417
1418         origsize = *size;
1419         ce = (struct compat_ipt_entry __user *)*dstptr;
1420         if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1421             copy_to_user(&ce->counters, &counters[i],
1422             sizeof(counters[i])) != 0)
1423                 return -EFAULT;
1424
1425         *dstptr += sizeof(struct compat_ipt_entry);
1426         *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1427
1428         xt_ematch_foreach(ematch, e) {
1429                 ret = xt_compat_match_to_user(ematch, dstptr, size);
1430                 if (ret != 0)
1431                         return ret;
1432         }
1433         target_offset = e->target_offset - (origsize - *size);
1434         t = ipt_get_target(e);
1435         ret = xt_compat_target_to_user(t, dstptr, size);
1436         if (ret)
1437                 return ret;
1438         next_offset = e->next_offset - (origsize - *size);
1439         if (put_user(target_offset, &ce->target_offset) != 0 ||
1440             put_user(next_offset, &ce->next_offset) != 0)
1441                 return -EFAULT;
1442         return 0;
1443 }
1444
1445 static int
1446 compat_find_calc_match(struct xt_entry_match *m,
1447                        const char *name,
1448                        const struct ipt_ip *ip,
1449                        int *size)
1450 {
1451         struct xt_match *match;
1452
1453         match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1454                                       m->u.user.revision);
1455         if (IS_ERR(match)) {
1456                 duprintf("compat_check_calc_match: `%s' not found\n",
1457                          m->u.user.name);
1458                 return PTR_ERR(match);
1459         }
1460         m->u.kernel.match = match;
1461         *size += xt_compat_match_offset(match);
1462         return 0;
1463 }
1464
1465 static void compat_release_entry(struct compat_ipt_entry *e)
1466 {
1467         struct xt_entry_target *t;
1468         struct xt_entry_match *ematch;
1469
1470         /* Cleanup all matches */
1471         xt_ematch_foreach(ematch, e)
1472                 module_put(ematch->u.kernel.match->me);
1473         t = compat_ipt_get_target(e);
1474         module_put(t->u.kernel.target->me);
1475 }
1476
1477 static int
1478 check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1479                                   struct xt_table_info *newinfo,
1480                                   unsigned int *size,
1481                                   const unsigned char *base,
1482                                   const unsigned char *limit,
1483                                   const unsigned int *hook_entries,
1484                                   const unsigned int *underflows,
1485                                   const char *name)
1486 {
1487         struct xt_entry_match *ematch;
1488         struct xt_entry_target *t;
1489         struct xt_target *target;
1490         unsigned int entry_offset;
1491         unsigned int j;
1492         int ret, off, h;
1493
1494         duprintf("check_compat_entry_size_and_hooks %p\n", e);
1495         if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1496             (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1497             (unsigned char *)e + e->next_offset > limit) {
1498                 duprintf("Bad offset %p, limit = %p\n", e, limit);
1499                 return -EINVAL;
1500         }
1501
1502         if (e->next_offset < sizeof(struct compat_ipt_entry) +
1503                              sizeof(struct compat_xt_entry_target)) {
1504                 duprintf("checking: element %p size %u\n",
1505                          e, e->next_offset);
1506                 return -EINVAL;
1507         }
1508
1509         /* For purposes of check_entry casting the compat entry is fine */
1510         ret = check_entry((struct ipt_entry *)e);
1511         if (ret)
1512                 return ret;
1513
1514         off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1515         entry_offset = (void *)e - (void *)base;
1516         j = 0;
1517         xt_ematch_foreach(ematch, e) {
1518                 ret = compat_find_calc_match(ematch, name, &e->ip, &off);
1519                 if (ret != 0)
1520                         goto release_matches;
1521                 ++j;
1522         }
1523
1524         t = compat_ipt_get_target(e);
1525         target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1526                                         t->u.user.revision);
1527         if (IS_ERR(target)) {
1528                 duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1529                          t->u.user.name);
1530                 ret = PTR_ERR(target);
1531                 goto release_matches;
1532         }
1533         t->u.kernel.target = target;
1534
1535         off += xt_compat_target_offset(target);
1536         *size += off;
1537         ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1538         if (ret)
1539                 goto out;
1540
1541         /* Check hooks & underflows */
1542         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1543                 if ((unsigned char *)e - base == hook_entries[h])
1544                         newinfo->hook_entry[h] = hook_entries[h];
1545                 if ((unsigned char *)e - base == underflows[h])
1546                         newinfo->underflow[h] = underflows[h];
1547         }
1548
1549         /* Clear counters and comefrom */
1550         memset(&e->counters, 0, sizeof(e->counters));
1551         e->comefrom = 0;
1552         return 0;
1553
1554 out:
1555         module_put(t->u.kernel.target->me);
1556 release_matches:
1557         xt_ematch_foreach(ematch, e) {
1558                 if (j-- == 0)
1559                         break;
1560                 module_put(ematch->u.kernel.match->me);
1561         }
1562         return ret;
1563 }
1564
1565 static int
1566 compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1567                             unsigned int *size, const char *name,
1568                             struct xt_table_info *newinfo, unsigned char *base)
1569 {
1570         struct xt_entry_target *t;
1571         struct xt_target *target;
1572         struct ipt_entry *de;
1573         unsigned int origsize;
1574         int ret, h;
1575         struct xt_entry_match *ematch;
1576
1577         ret = 0;
1578         origsize = *size;
1579         de = (struct ipt_entry *)*dstptr;
1580         memcpy(de, e, sizeof(struct ipt_entry));
1581         memcpy(&de->counters, &e->counters, sizeof(e->counters));
1582
1583         *dstptr += sizeof(struct ipt_entry);
1584         *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1585
1586         xt_ematch_foreach(ematch, e) {
1587                 ret = xt_compat_match_from_user(ematch, dstptr, size);
1588                 if (ret != 0)
1589                         return ret;
1590         }
1591         de->target_offset = e->target_offset - (origsize - *size);
1592         t = compat_ipt_get_target(e);
1593         target = t->u.kernel.target;
1594         xt_compat_target_from_user(t, dstptr, size);
1595
1596         de->next_offset = e->next_offset - (origsize - *size);
1597         for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1598                 if ((unsigned char *)de - base < newinfo->hook_entry[h])
1599                         newinfo->hook_entry[h] -= origsize - *size;
1600                 if ((unsigned char *)de - base < newinfo->underflow[h])
1601                         newinfo->underflow[h] -= origsize - *size;
1602         }
1603         return ret;
1604 }
1605
1606 static int
1607 compat_check_entry(struct ipt_entry *e, struct net *net, const char *name)
1608 {
1609         struct xt_entry_match *ematch;
1610         struct xt_mtchk_param mtpar;
1611         unsigned int j;
1612         int ret = 0;
1613
1614         e->counters.pcnt = xt_percpu_counter_alloc();
1615         if (IS_ERR_VALUE(e->counters.pcnt))
1616                 return -ENOMEM;
1617
1618         j = 0;
1619         mtpar.net       = net;
1620         mtpar.table     = name;
1621         mtpar.entryinfo = &e->ip;
1622         mtpar.hook_mask = e->comefrom;
1623         mtpar.family    = NFPROTO_IPV4;
1624         xt_ematch_foreach(ematch, e) {
1625                 ret = check_match(ematch, &mtpar);
1626                 if (ret != 0)
1627                         goto cleanup_matches;
1628                 ++j;
1629         }
1630
1631         ret = check_target(e, net, name);
1632         if (ret)
1633                 goto cleanup_matches;
1634         return 0;
1635
1636  cleanup_matches:
1637         xt_ematch_foreach(ematch, e) {
1638                 if (j-- == 0)
1639                         break;
1640                 cleanup_match(ematch, net);
1641         }
1642
1643         xt_percpu_counter_free(e->counters.pcnt);
1644
1645         return ret;
1646 }
1647
1648 static int
1649 translate_compat_table(struct net *net,
1650                        const char *name,
1651                        unsigned int valid_hooks,
1652                        struct xt_table_info **pinfo,
1653                        void **pentry0,
1654                        unsigned int total_size,
1655                        unsigned int number,
1656                        unsigned int *hook_entries,
1657                        unsigned int *underflows)
1658 {
1659         unsigned int i, j;
1660         struct xt_table_info *newinfo, *info;
1661         void *pos, *entry0, *entry1;
1662         struct compat_ipt_entry *iter0;
1663         struct ipt_entry *iter1;
1664         unsigned int size;
1665         int ret;
1666
1667         info = *pinfo;
1668         entry0 = *pentry0;
1669         size = total_size;
1670         info->number = number;
1671
1672         /* Init all hooks to impossible value. */
1673         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1674                 info->hook_entry[i] = 0xFFFFFFFF;
1675                 info->underflow[i] = 0xFFFFFFFF;
1676         }
1677
1678         duprintf("translate_compat_table: size %u\n", info->size);
1679         j = 0;
1680         xt_compat_lock(AF_INET);
1681         xt_compat_init_offsets(AF_INET, number);
1682         /* Walk through entries, checking offsets. */
1683         xt_entry_foreach(iter0, entry0, total_size) {
1684                 ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1685                                                         entry0,
1686                                                         entry0 + total_size,
1687                                                         hook_entries,
1688                                                         underflows,
1689                                                         name);
1690                 if (ret != 0)
1691                         goto out_unlock;
1692                 ++j;
1693         }
1694
1695         ret = -EINVAL;
1696         if (j != number) {
1697                 duprintf("translate_compat_table: %u not %u entries\n",
1698                          j, number);
1699                 goto out_unlock;
1700         }
1701
1702         /* Check hooks all assigned */
1703         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1704                 /* Only hooks which are valid */
1705                 if (!(valid_hooks & (1 << i)))
1706                         continue;
1707                 if (info->hook_entry[i] == 0xFFFFFFFF) {
1708                         duprintf("Invalid hook entry %u %u\n",
1709                                  i, hook_entries[i]);
1710                         goto out_unlock;
1711                 }
1712                 if (info->underflow[i] == 0xFFFFFFFF) {
1713                         duprintf("Invalid underflow %u %u\n",
1714                                  i, underflows[i]);
1715                         goto out_unlock;
1716                 }
1717         }
1718
1719         ret = -ENOMEM;
1720         newinfo = xt_alloc_table_info(size);
1721         if (!newinfo)
1722                 goto out_unlock;
1723
1724         newinfo->number = number;
1725         for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1726                 newinfo->hook_entry[i] = info->hook_entry[i];
1727                 newinfo->underflow[i] = info->underflow[i];
1728         }
1729         entry1 = newinfo->entries;
1730         pos = entry1;
1731         size = total_size;
1732         xt_entry_foreach(iter0, entry0, total_size) {
1733                 ret = compat_copy_entry_from_user(iter0, &pos, &size,
1734                                                   name, newinfo, entry1);
1735                 if (ret != 0)
1736                         break;
1737         }
1738         xt_compat_flush_offsets(AF_INET);
1739         xt_compat_unlock(AF_INET);
1740         if (ret)
1741                 goto free_newinfo;
1742
1743         ret = -ELOOP;
1744         if (!mark_source_chains(newinfo, valid_hooks, entry1))
1745                 goto free_newinfo;
1746
1747         i = 0;
1748         xt_entry_foreach(iter1, entry1, newinfo->size) {
1749                 ret = compat_check_entry(iter1, net, name);
1750                 if (ret != 0)
1751                         break;
1752                 ++i;
1753                 if (strcmp(ipt_get_target(iter1)->u.user.name,
1754                     XT_ERROR_TARGET) == 0)
1755                         ++newinfo->stacksize;
1756         }
1757         if (ret) {
1758                 /*
1759                  * The first i matches need cleanup_entry (calls ->destroy)
1760                  * because they had called ->check already. The other j-i
1761                  * entries need only release.
1762                  */
1763                 int skip = i;
1764                 j -= i;
1765                 xt_entry_foreach(iter0, entry0, newinfo->size) {
1766                         if (skip-- > 0)
1767                                 continue;
1768                         if (j-- == 0)
1769                                 break;
1770                         compat_release_entry(iter0);
1771                 }
1772                 xt_entry_foreach(iter1, entry1, newinfo->size) {
1773                         if (i-- == 0)
1774                                 break;
1775                         cleanup_entry(iter1, net);
1776                 }
1777                 xt_free_table_info(newinfo);
1778                 return ret;
1779         }
1780
1781         *pinfo = newinfo;
1782         *pentry0 = entry1;
1783         xt_free_table_info(info);
1784         return 0;
1785
1786 free_newinfo:
1787         xt_free_table_info(newinfo);
1788 out:
1789         xt_entry_foreach(iter0, entry0, total_size) {
1790                 if (j-- == 0)
1791                         break;
1792                 compat_release_entry(iter0);
1793         }
1794         return ret;
1795 out_unlock:
1796         xt_compat_flush_offsets(AF_INET);
1797         xt_compat_unlock(AF_INET);
1798         goto out;
1799 }
1800
1801 static int
1802 compat_do_replace(struct net *net, void __user *user, unsigned int len)
1803 {
1804         int ret;
1805         struct compat_ipt_replace tmp;
1806         struct xt_table_info *newinfo;
1807         void *loc_cpu_entry;
1808         struct ipt_entry *iter;
1809
1810         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1811                 return -EFAULT;
1812
1813         /* overflow check */
1814         if (tmp.size >= INT_MAX / num_possible_cpus())
1815                 return -ENOMEM;
1816         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1817                 return -ENOMEM;
1818         if (tmp.num_counters == 0)
1819                 return -EINVAL;
1820
1821         tmp.name[sizeof(tmp.name)-1] = 0;
1822
1823         newinfo = xt_alloc_table_info(tmp.size);
1824         if (!newinfo)
1825                 return -ENOMEM;
1826
1827         loc_cpu_entry = newinfo->entries;
1828         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1829                            tmp.size) != 0) {
1830                 ret = -EFAULT;
1831                 goto free_newinfo;
1832         }
1833
1834         ret = translate_compat_table(net, tmp.name, tmp.valid_hooks,
1835                                      &newinfo, &loc_cpu_entry, tmp.size,
1836                                      tmp.num_entries, tmp.hook_entry,
1837                                      tmp.underflow);
1838         if (ret != 0)
1839                 goto free_newinfo;
1840
1841         duprintf("compat_do_replace: Translated table\n");
1842
1843         ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1844                            tmp.num_counters, compat_ptr(tmp.counters));
1845         if (ret)
1846                 goto free_newinfo_untrans;
1847         return 0;
1848
1849  free_newinfo_untrans:
1850         xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1851                 cleanup_entry(iter, net);
1852  free_newinfo:
1853         xt_free_table_info(newinfo);
1854         return ret;
1855 }
1856
1857 static int
1858 compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1859                       unsigned int len)
1860 {
1861         int ret;
1862
1863         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1864                 return -EPERM;
1865
1866         switch (cmd) {
1867         case IPT_SO_SET_REPLACE:
1868                 ret = compat_do_replace(sock_net(sk), user, len);
1869                 break;
1870
1871         case IPT_SO_SET_ADD_COUNTERS:
1872                 ret = do_add_counters(sock_net(sk), user, len, 1);
1873                 break;
1874
1875         default:
1876                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1877                 ret = -EINVAL;
1878         }
1879
1880         return ret;
1881 }
1882
1883 struct compat_ipt_get_entries {
1884         char name[XT_TABLE_MAXNAMELEN];
1885         compat_uint_t size;
1886         struct compat_ipt_entry entrytable[0];
1887 };
1888
1889 static int
1890 compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1891                             void __user *userptr)
1892 {
1893         struct xt_counters *counters;
1894         const struct xt_table_info *private = table->private;
1895         void __user *pos;
1896         unsigned int size;
1897         int ret = 0;
1898         unsigned int i = 0;
1899         struct ipt_entry *iter;
1900
1901         counters = alloc_counters(table);
1902         if (IS_ERR(counters))
1903                 return PTR_ERR(counters);
1904
1905         pos = userptr;
1906         size = total_size;
1907         xt_entry_foreach(iter, private->entries, total_size) {
1908                 ret = compat_copy_entry_to_user(iter, &pos,
1909                                                 &size, counters, i++);
1910                 if (ret != 0)
1911                         break;
1912         }
1913
1914         vfree(counters);
1915         return ret;
1916 }
1917
1918 static int
1919 compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1920                    int *len)
1921 {
1922         int ret;
1923         struct compat_ipt_get_entries get;
1924         struct xt_table *t;
1925
1926         if (*len < sizeof(get)) {
1927                 duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1928                 return -EINVAL;
1929         }
1930
1931         if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1932                 return -EFAULT;
1933
1934         if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1935                 duprintf("compat_get_entries: %u != %zu\n",
1936                          *len, sizeof(get) + get.size);
1937                 return -EINVAL;
1938         }
1939         get.name[sizeof(get.name) - 1] = '\0';
1940
1941         xt_compat_lock(AF_INET);
1942         t = xt_find_table_lock(net, AF_INET, get.name);
1943         if (!IS_ERR_OR_NULL(t)) {
1944                 const struct xt_table_info *private = t->private;
1945                 struct xt_table_info info;
1946                 duprintf("t->private->number = %u\n", private->number);
1947                 ret = compat_table_info(private, &info);
1948                 if (!ret && get.size == info.size) {
1949                         ret = compat_copy_entries_to_user(private->size,
1950                                                           t, uptr->entrytable);
1951                 } else if (!ret) {
1952                         duprintf("compat_get_entries: I've got %u not %u!\n",
1953                                  private->size, get.size);
1954                         ret = -EAGAIN;
1955                 }
1956                 xt_compat_flush_offsets(AF_INET);
1957                 module_put(t->me);
1958                 xt_table_unlock(t);
1959         } else
1960                 ret = t ? PTR_ERR(t) : -ENOENT;
1961
1962         xt_compat_unlock(AF_INET);
1963         return ret;
1964 }
1965
1966 static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1967
1968 static int
1969 compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1970 {
1971         int ret;
1972
1973         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1974                 return -EPERM;
1975
1976         switch (cmd) {
1977         case IPT_SO_GET_INFO:
1978                 ret = get_info(sock_net(sk), user, len, 1);
1979                 break;
1980         case IPT_SO_GET_ENTRIES:
1981                 ret = compat_get_entries(sock_net(sk), user, len);
1982                 break;
1983         default:
1984                 ret = do_ipt_get_ctl(sk, cmd, user, len);
1985         }
1986         return ret;
1987 }
1988 #endif
1989
1990 static int
1991 do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
1992 {
1993         int ret;
1994
1995         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1996                 return -EPERM;
1997
1998         switch (cmd) {
1999         case IPT_SO_SET_REPLACE:
2000                 ret = do_replace(sock_net(sk), user, len);
2001                 break;
2002
2003         case IPT_SO_SET_ADD_COUNTERS:
2004                 ret = do_add_counters(sock_net(sk), user, len, 0);
2005                 break;
2006
2007         default:
2008                 duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
2009                 ret = -EINVAL;
2010         }
2011
2012         return ret;
2013 }
2014
2015 static int
2016 do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
2017 {
2018         int ret;
2019
2020         if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2021                 return -EPERM;
2022
2023         switch (cmd) {
2024         case IPT_SO_GET_INFO:
2025                 ret = get_info(sock_net(sk), user, len, 0);
2026                 break;
2027
2028         case IPT_SO_GET_ENTRIES:
2029                 ret = get_entries(sock_net(sk), user, len);
2030                 break;
2031
2032         case IPT_SO_GET_REVISION_MATCH:
2033         case IPT_SO_GET_REVISION_TARGET: {
2034                 struct xt_get_revision rev;
2035                 int target;
2036
2037                 if (*len != sizeof(rev)) {
2038                         ret = -EINVAL;
2039                         break;
2040                 }
2041                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
2042                         ret = -EFAULT;
2043                         break;
2044                 }
2045                 rev.name[sizeof(rev.name)-1] = 0;
2046
2047                 if (cmd == IPT_SO_GET_REVISION_TARGET)
2048                         target = 1;
2049                 else
2050                         target = 0;
2051
2052                 try_then_request_module(xt_find_revision(AF_INET, rev.name,
2053                                                          rev.revision,
2054                                                          target, &ret),
2055                                         "ipt_%s", rev.name);
2056                 break;
2057         }
2058
2059         default:
2060                 duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
2061                 ret = -EINVAL;
2062         }
2063
2064         return ret;
2065 }
2066
2067 static void __ipt_unregister_table(struct net *net, struct xt_table *table)
2068 {
2069         struct xt_table_info *private;
2070         void *loc_cpu_entry;
2071         struct module *table_owner = table->me;
2072         struct ipt_entry *iter;
2073
2074         private = xt_unregister_table(table);
2075
2076         /* Decrease module usage counts and free resources */
2077         loc_cpu_entry = private->entries;
2078         xt_entry_foreach(iter, loc_cpu_entry, private->size)
2079                 cleanup_entry(iter, net);
2080         if (private->number > private->initial_entries)
2081                 module_put(table_owner);
2082         xt_free_table_info(private);
2083 }
2084
2085 int ipt_register_table(struct net *net, const struct xt_table *table,
2086                        const struct ipt_replace *repl,
2087                        const struct nf_hook_ops *ops, struct xt_table **res)
2088 {
2089         int ret;
2090         struct xt_table_info *newinfo;
2091         struct xt_table_info bootstrap = {0};
2092         void *loc_cpu_entry;
2093         struct xt_table *new_table;
2094
2095         newinfo = xt_alloc_table_info(repl->size);
2096         if (!newinfo)
2097                 return -ENOMEM;
2098
2099         loc_cpu_entry = newinfo->entries;
2100         memcpy(loc_cpu_entry, repl->entries, repl->size);
2101
2102         ret = translate_table(net, newinfo, loc_cpu_entry, repl);
2103         if (ret != 0)
2104                 goto out_free;
2105
2106         new_table = xt_register_table(net, table, &bootstrap, newinfo);
2107         if (IS_ERR(new_table)) {
2108                 ret = PTR_ERR(new_table);
2109                 goto out_free;
2110         }
2111
2112         /* set res now, will see skbs right after nf_register_net_hooks */
2113         WRITE_ONCE(*res, new_table);
2114
2115         ret = nf_register_net_hooks(net, ops, hweight32(table->valid_hooks));
2116         if (ret != 0) {
2117                 __ipt_unregister_table(net, new_table);
2118                 *res = NULL;
2119         }
2120
2121         return ret;
2122
2123 out_free:
2124         xt_free_table_info(newinfo);
2125         return ret;
2126 }
2127
2128 void ipt_unregister_table(struct net *net, struct xt_table *table,
2129                           const struct nf_hook_ops *ops)
2130 {
2131         nf_unregister_net_hooks(net, ops, hweight32(table->valid_hooks));
2132         __ipt_unregister_table(net, table);
2133 }
2134
2135 /* Returns 1 if the type and code is matched by the range, 0 otherwise */
2136 static inline bool
2137 icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
2138                      u_int8_t type, u_int8_t code,
2139                      bool invert)
2140 {
2141         return ((test_type == 0xFF) ||
2142                 (type == test_type && code >= min_code && code <= max_code))
2143                 ^ invert;
2144 }
2145
2146 static bool
2147 icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2148 {
2149         const struct icmphdr *ic;
2150         struct icmphdr _icmph;
2151         const struct ipt_icmp *icmpinfo = par->matchinfo;
2152
2153         /* Must not be a fragment. */
2154         if (par->fragoff != 0)
2155                 return false;
2156
2157         ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2158         if (ic == NULL) {
2159                 /* We've been asked to examine this packet, and we
2160                  * can't.  Hence, no choice but to drop.
2161                  */
2162                 duprintf("Dropping evil ICMP tinygram.\n");
2163                 par->hotdrop = true;
2164                 return false;
2165         }
2166
2167         return icmp_type_code_match(icmpinfo->type,
2168                                     icmpinfo->code[0],
2169                                     icmpinfo->code[1],
2170                                     ic->type, ic->code,
2171                                     !!(icmpinfo->invflags&IPT_ICMP_INV));
2172 }
2173
2174 static int icmp_checkentry(const struct xt_mtchk_param *par)
2175 {
2176         const struct ipt_icmp *icmpinfo = par->matchinfo;
2177
2178         /* Must specify no unknown invflags */
2179         return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2180 }
2181
2182 static struct xt_target ipt_builtin_tg[] __read_mostly = {
2183         {
2184                 .name             = XT_STANDARD_TARGET,
2185                 .targetsize       = sizeof(int),
2186                 .family           = NFPROTO_IPV4,
2187 #ifdef CONFIG_COMPAT
2188                 .compatsize       = sizeof(compat_int_t),
2189                 .compat_from_user = compat_standard_from_user,
2190                 .compat_to_user   = compat_standard_to_user,
2191 #endif
2192         },
2193         {
2194                 .name             = XT_ERROR_TARGET,
2195                 .target           = ipt_error,
2196                 .targetsize       = XT_FUNCTION_MAXNAMELEN,
2197                 .family           = NFPROTO_IPV4,
2198         },
2199 };
2200
2201 static struct nf_sockopt_ops ipt_sockopts = {
2202         .pf             = PF_INET,
2203         .set_optmin     = IPT_BASE_CTL,
2204         .set_optmax     = IPT_SO_SET_MAX+1,
2205         .set            = do_ipt_set_ctl,
2206 #ifdef CONFIG_COMPAT
2207         .compat_set     = compat_do_ipt_set_ctl,
2208 #endif
2209         .get_optmin     = IPT_BASE_CTL,
2210         .get_optmax     = IPT_SO_GET_MAX+1,
2211         .get            = do_ipt_get_ctl,
2212 #ifdef CONFIG_COMPAT
2213         .compat_get     = compat_do_ipt_get_ctl,
2214 #endif
2215         .owner          = THIS_MODULE,
2216 };
2217
2218 static struct xt_match ipt_builtin_mt[] __read_mostly = {
2219         {
2220                 .name       = "icmp",
2221                 .match      = icmp_match,
2222                 .matchsize  = sizeof(struct ipt_icmp),
2223                 .checkentry = icmp_checkentry,
2224                 .proto      = IPPROTO_ICMP,
2225                 .family     = NFPROTO_IPV4,
2226         },
2227 };
2228
2229 static int __net_init ip_tables_net_init(struct net *net)
2230 {
2231         return xt_proto_init(net, NFPROTO_IPV4);
2232 }
2233
2234 static void __net_exit ip_tables_net_exit(struct net *net)
2235 {
2236         xt_proto_fini(net, NFPROTO_IPV4);
2237 }
2238
2239 static struct pernet_operations ip_tables_net_ops = {
2240         .init = ip_tables_net_init,
2241         .exit = ip_tables_net_exit,
2242 };
2243
2244 static int __init ip_tables_init(void)
2245 {
2246         int ret;
2247
2248         ret = register_pernet_subsys(&ip_tables_net_ops);
2249         if (ret < 0)
2250                 goto err1;
2251
2252         /* No one else will be downing sem now, so we won't sleep */
2253         ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2254         if (ret < 0)
2255                 goto err2;
2256         ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2257         if (ret < 0)
2258                 goto err4;
2259
2260         /* Register setsockopt */
2261         ret = nf_register_sockopt(&ipt_sockopts);
2262         if (ret < 0)
2263                 goto err5;
2264
2265         pr_info("(C) 2000-2006 Netfilter Core Team\n");
2266         return 0;
2267
2268 err5:
2269         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2270 err4:
2271         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2272 err2:
2273         unregister_pernet_subsys(&ip_tables_net_ops);
2274 err1:
2275         return ret;
2276 }
2277
2278 static void __exit ip_tables_fini(void)
2279 {
2280         nf_unregister_sockopt(&ipt_sockopts);
2281
2282         xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2283         xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2284         unregister_pernet_subsys(&ip_tables_net_ops);
2285 }
2286
2287 EXPORT_SYMBOL(ipt_register_table);
2288 EXPORT_SYMBOL(ipt_unregister_table);
2289 EXPORT_SYMBOL(ipt_do_table);
2290 module_init(ip_tables_init);
2291 module_exit(ip_tables_fini);