linux/net/ipv4/netfilter/ip_tables.c
<<
>>
Prefs
   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
  34MODULE_LICENSE("GPL");
  35MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
  36MODULE_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
  66void *ipt_alloc_initial_table(const struct xt_table *info)
  67{
  68        return xt_alloc_initial_table(ipt, IPT);
  69}
  70EXPORT_SYMBOL_GPL(ipt_alloc_initial_table);
  71
  72/* Returns whether matches rule or not. */
  73/* Performance critical - called for every packet */
  74static inline bool
  75ip_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
 138static bool
 139ip_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
 154static unsigned int
 155ipt_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 */
 163static inline struct ipt_entry *
 164get_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) */
 171static 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 */
 181static inline const struct xt_entry_target *
 182ipt_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)
 188static 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
 196enum nf_ip_trace_comments {
 197        NF_IP_TRACE_COMMENT_RULE,
 198        NF_IP_TRACE_COMMENT_RETURN,
 199        NF_IP_TRACE_COMMENT_POLICY,
 200};
 201
 202static 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
 208static 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) */
 219static inline int
 220get_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
 249static 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 struct ipt_entry *root;
 258        const char *hookname, *chainname, *comment;
 259        const struct ipt_entry *iter;
 260        unsigned int rulenum = 0;
 261        struct net *net = dev_net(in ? in : out);
 262
 263        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
 279static inline __pure
 280struct 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. */
 286unsigned int
 287ipt_do_table(struct sk_buff *skb,
 288             unsigned int hook,
 289             const struct nf_hook_state *state,
 290             struct xt_table *table)
 291{
 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 *stackptr, origptr, cpu;
 300        const struct xt_table_info *private;
 301        struct xt_action_param acpar;
 302        unsigned int addend;
 303
 304        /* Initialization */
 305        ip = ip_hdr(skb);
 306        indev = state->in ? state->in->name : nulldevname;
 307        outdev = state->out ? state->out->name : nulldevname;
 308        /* We handle fragments by dealing with the first fragment as
 309         * if it was a normal packet.  All other fragments are treated
 310         * normally, except that they will NEVER match rules that ask
 311         * things we don't know, ie. tcp syn flag or ports).  If the
 312         * rule is also a fragment-specific rule, non-fragments won't
 313         * match it. */
 314        acpar.fragoff = ntohs(ip->frag_off) & IP_OFFSET;
 315        acpar.thoff   = ip_hdrlen(skb);
 316        acpar.hotdrop = false;
 317        acpar.in      = state->in;
 318        acpar.out     = state->out;
 319        acpar.family  = NFPROTO_IPV4;
 320        acpar.hooknum = hook;
 321
 322        IP_NF_ASSERT(table->valid_hooks & (1 << hook));
 323        local_bh_disable();
 324        addend = xt_write_recseq_begin();
 325        private = table->private;
 326        cpu        = smp_processor_id();
 327        /*
 328         * Ensure we load private-> members after we've fetched the base
 329         * pointer.
 330         */
 331        smp_read_barrier_depends();
 332        table_base = private->entries;
 333        jumpstack  = (struct ipt_entry **)private->jumpstack[cpu];
 334        stackptr   = per_cpu_ptr(private->stackptr, cpu);
 335        origptr    = *stackptr;
 336
 337        e = get_entry(table_base, private->hook_entry[hook]);
 338
 339        pr_debug("Entering %s(hook %u); sp at %u (UF %p)\n",
 340                 table->name, hook, origptr,
 341                 get_entry(table_base, private->underflow[hook]));
 342
 343        do {
 344                const struct xt_entry_target *t;
 345                const struct xt_entry_match *ematch;
 346                struct xt_counters *counter;
 347
 348                IP_NF_ASSERT(e);
 349                if (!ip_packet_match(ip, indev, outdev,
 350                    &e->ip, acpar.fragoff)) {
 351 no_match:
 352                        e = ipt_next_entry(e);
 353                        continue;
 354                }
 355
 356                xt_ematch_foreach(ematch, e) {
 357                        acpar.match     = ematch->u.kernel.match;
 358                        acpar.matchinfo = ematch->data;
 359                        if (!acpar.match->match(skb, &acpar))
 360                                goto no_match;
 361                }
 362
 363                counter = xt_get_this_cpu_counter(&e->counters);
 364                ADD_COUNTER(*counter, 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. */
 443static int
 444mark_source_chains(const struct xt_table_info *newinfo,
 445                   unsigned int valid_hooks, void *entry0,
 446                   unsigned int *offsets)
 447{
 448        unsigned int hook;
 449
 450        /* No recursion; use packet counter to save back ptrs (reset
 451           to 0 as we leave), and comefrom to save source hook bitmask */
 452        for (hook = 0; hook < NF_INET_NUMHOOKS; hook++) {
 453                unsigned int pos = newinfo->hook_entry[hook];
 454                struct ipt_entry *e = (struct ipt_entry *)(entry0 + pos);
 455
 456                if (!(valid_hooks & (1 << hook)))
 457                        continue;
 458
 459                /* Set initial back pointer. */
 460                e->counters.pcnt = pos;
 461
 462                for (;;) {
 463                        const struct xt_standard_target *t
 464                                = (void *)ipt_get_target_c(e);
 465                        int visited = e->comefrom & (1 << hook);
 466
 467                        if (e->comefrom & (1 << NF_INET_NUMHOOKS)) {
 468                                pr_err("iptables: loop hook %u pos %u %08X.\n",
 469                                       hook, pos, e->comefrom);
 470                                return 0;
 471                        }
 472                        e->comefrom |= ((1 << hook) | (1 << NF_INET_NUMHOOKS));
 473
 474                        /* Unconditional return/END. */
 475                        if ((unconditional(e) &&
 476                             (strcmp(t->target.u.user.name,
 477                                     XT_STANDARD_TARGET) == 0) &&
 478                             t->verdict < 0) || 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                                if (pos + size >= newinfo->size)
 520                                        return 0;
 521                                e->counters.pcnt = pos;
 522                                pos += size;
 523                        } else {
 524                                int newpos = t->verdict;
 525
 526                                if (strcmp(t->target.u.user.name,
 527                                           XT_STANDARD_TARGET) == 0 &&
 528                                    newpos >= 0) {
 529                                        if (newpos > newinfo->size -
 530                                                sizeof(struct ipt_entry)) {
 531                                                duprintf("mark_source_chains: "
 532                                                        "bad verdict (%i)\n",
 533                                                                newpos);
 534                                                return 0;
 535                                        }
 536                                        /* This a jump; chase it. */
 537                                        duprintf("Jump rule %u -> %u\n",
 538                                                 pos, newpos);
 539                                        if (!xt_find_jump_offset(offsets, newpos,
 540                                                                 newinfo->number))
 541                                                return 0;
 542                                        e = (struct ipt_entry *)
 543                                                (entry0 + newpos);
 544                                } else {
 545                                        /* ... this is a fallthru */
 546                                        newpos = pos + e->next_offset;
 547                                        if (newpos >= newinfo->size)
 548                                                return 0;
 549                                }
 550                                e = (struct ipt_entry *)
 551                                        (entry0 + newpos);
 552                                e->counters.pcnt = pos;
 553                                pos = newpos;
 554                        }
 555                }
 556                next:
 557                duprintf("Finished chain %u\n", hook);
 558        }
 559        return 1;
 560}
 561
 562static void cleanup_match(struct xt_entry_match *m, struct net *net)
 563{
 564        struct xt_mtdtor_param par;
 565
 566        par.net       = net;
 567        par.match     = m->u.kernel.match;
 568        par.matchinfo = m->data;
 569        par.family    = NFPROTO_IPV4;
 570        if (par.match->destroy != NULL)
 571                par.match->destroy(&par);
 572        module_put(par.match->me);
 573}
 574
 575static int
 576check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 577{
 578        const struct ipt_ip *ip = par->entryinfo;
 579        int ret;
 580
 581        par->match     = m->u.kernel.match;
 582        par->matchinfo = m->data;
 583
 584        ret = xt_check_match(par, m->u.match_size - sizeof(*m),
 585              ip->proto, ip->invflags & IPT_INV_PROTO);
 586        if (ret < 0) {
 587                duprintf("check failed for `%s'.\n", par->match->name);
 588                return ret;
 589        }
 590        return 0;
 591}
 592
 593static int
 594find_check_match(struct xt_entry_match *m, struct xt_mtchk_param *par)
 595{
 596        struct xt_match *match;
 597        int ret;
 598
 599        match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
 600                                      m->u.user.revision);
 601        if (IS_ERR(match)) {
 602                duprintf("find_check_match: `%s' not found\n", m->u.user.name);
 603                return PTR_ERR(match);
 604        }
 605        m->u.kernel.match = match;
 606
 607        ret = check_match(m, par);
 608        if (ret)
 609                goto err;
 610
 611        return 0;
 612err:
 613        module_put(m->u.kernel.match->me);
 614        return ret;
 615}
 616
 617static int check_target(struct ipt_entry *e, struct net *net, const char *name)
 618{
 619        struct xt_entry_target *t = ipt_get_target(e);
 620        struct xt_tgchk_param par = {
 621                .net       = net,
 622                .table     = name,
 623                .entryinfo = e,
 624                .target    = t->u.kernel.target,
 625                .targinfo  = t->data,
 626                .hook_mask = e->comefrom,
 627                .family    = NFPROTO_IPV4,
 628        };
 629        int ret;
 630
 631        ret = xt_check_target(&par, t->u.target_size - sizeof(*t),
 632              e->ip.proto, e->ip.invflags & IPT_INV_PROTO);
 633        if (ret < 0) {
 634                duprintf("check failed for `%s'.\n",
 635                         t->u.kernel.target->name);
 636                return ret;
 637        }
 638        return 0;
 639}
 640
 641static int
 642find_check_entry(struct ipt_entry *e, struct net *net, const char *name,
 643                 unsigned int size,
 644                 struct xt_percpu_counter_alloc_state *alloc_state)
 645{
 646        struct xt_entry_target *t;
 647        struct xt_target *target;
 648        int ret;
 649        unsigned int j;
 650        struct xt_mtchk_param mtpar;
 651        struct xt_entry_match *ematch;
 652
 653        if (!xt_percpu_counter_alloc(alloc_state, &e->counters))
 654                return -ENOMEM;
 655
 656        j = 0;
 657        mtpar.net       = net;
 658        mtpar.table     = name;
 659        mtpar.entryinfo = &e->ip;
 660        mtpar.hook_mask = e->comefrom;
 661        mtpar.family    = NFPROTO_IPV4;
 662        xt_ematch_foreach(ematch, e) {
 663                ret = find_check_match(ematch, &mtpar);
 664                if (ret != 0)
 665                        goto cleanup_matches;
 666                ++j;
 667        }
 668
 669        t = ipt_get_target(e);
 670        target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
 671                                        t->u.user.revision);
 672        if (IS_ERR(target)) {
 673                duprintf("find_check_entry: `%s' not found\n", t->u.user.name);
 674                ret = PTR_ERR(target);
 675                goto cleanup_matches;
 676        }
 677        t->u.kernel.target = target;
 678
 679        ret = check_target(e, net, name);
 680        if (ret)
 681                goto err;
 682
 683        return 0;
 684 err:
 685        module_put(t->u.kernel.target->me);
 686 cleanup_matches:
 687        xt_ematch_foreach(ematch, e) {
 688                if (j-- == 0)
 689                        break;
 690                cleanup_match(ematch, net);
 691        }
 692
 693        xt_percpu_counter_free(&e->counters);
 694
 695        return ret;
 696}
 697
 698static bool check_underflow(const struct ipt_entry *e)
 699{
 700        const struct xt_entry_target *t;
 701        unsigned int verdict;
 702
 703        if (!unconditional(e))
 704                return false;
 705        t = ipt_get_target_c(e);
 706        if (strcmp(t->u.user.name, XT_STANDARD_TARGET) != 0)
 707                return false;
 708        verdict = ((struct xt_standard_target *)t)->verdict;
 709        verdict = -verdict - 1;
 710        return verdict == NF_DROP || verdict == NF_ACCEPT;
 711}
 712
 713static int
 714check_entry_size_and_hooks(struct ipt_entry *e,
 715                           struct xt_table_info *newinfo,
 716                           const unsigned char *base,
 717                           const unsigned char *limit,
 718                           const unsigned int *hook_entries,
 719                           const unsigned int *underflows,
 720                           unsigned int valid_hooks)
 721{
 722        unsigned int h;
 723        int err;
 724
 725        if ((unsigned long)e % __alignof__(struct ipt_entry) != 0 ||
 726            (unsigned char *)e + sizeof(struct ipt_entry) >= limit ||
 727            (unsigned char *)e + e->next_offset > limit) {
 728                duprintf("Bad offset %p\n", e);
 729                return -EINVAL;
 730        }
 731
 732        if (e->next_offset
 733            < sizeof(struct ipt_entry) + sizeof(struct xt_entry_target)) {
 734                duprintf("checking: element %p size %u\n",
 735                         e, e->next_offset);
 736                return -EINVAL;
 737        }
 738
 739        if (!ip_checkentry(&e->ip))
 740                return -EINVAL;
 741
 742        err = xt_check_entry_offsets(e, e->elems, e->target_offset,
 743                                     e->next_offset);
 744        if (err)
 745                return err;
 746
 747        /* Check hooks & underflows */
 748        for (h = 0; h < NF_INET_NUMHOOKS; h++) {
 749                if (!(valid_hooks & (1 << h)))
 750                        continue;
 751                if ((unsigned char *)e - base == hook_entries[h])
 752                        newinfo->hook_entry[h] = hook_entries[h];
 753                if ((unsigned char *)e - base == underflows[h]) {
 754                        if (!check_underflow(e)) {
 755                                pr_debug("Underflows must be unconditional and "
 756                                         "use the STANDARD target with "
 757                                         "ACCEPT/DROP\n");
 758                                return -EINVAL;
 759                        }
 760                        newinfo->underflow[h] = underflows[h];
 761                }
 762        }
 763
 764        /* Clear counters and comefrom */
 765        e->counters = ((struct xt_counters) { 0, 0 });
 766        e->comefrom = 0;
 767        return 0;
 768}
 769
 770static void
 771cleanup_entry(struct ipt_entry *e, struct net *net)
 772{
 773        struct xt_tgdtor_param par;
 774        struct xt_entry_target *t;
 775        struct xt_entry_match *ematch;
 776
 777        /* Cleanup all matches */
 778        xt_ematch_foreach(ematch, e)
 779                cleanup_match(ematch, net);
 780        t = ipt_get_target(e);
 781
 782        par.net      = net;
 783        par.target   = t->u.kernel.target;
 784        par.targinfo = t->data;
 785        par.family   = NFPROTO_IPV4;
 786        if (par.target->destroy != NULL)
 787                par.target->destroy(&par);
 788        module_put(par.target->me);
 789        xt_percpu_counter_free(&e->counters);
 790}
 791
 792/* Checks and translates the user-supplied table segment (held in
 793   newinfo) */
 794static int
 795translate_table(struct net *net, struct xt_table_info *newinfo, void *entry0,
 796                const struct ipt_replace *repl)
 797{
 798        struct xt_percpu_counter_alloc_state alloc_state = { 0 };
 799        struct ipt_entry *iter;
 800        unsigned int *offsets;
 801        unsigned int i;
 802        int ret = 0;
 803
 804        newinfo->size = repl->size;
 805        newinfo->number = repl->num_entries;
 806
 807        /* Init all hooks to impossible value. */
 808        for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 809                newinfo->hook_entry[i] = 0xFFFFFFFF;
 810                newinfo->underflow[i] = 0xFFFFFFFF;
 811        }
 812
 813        duprintf("translate_table: size %u\n", newinfo->size);
 814        offsets = xt_alloc_entry_offsets(newinfo->number);
 815        if (!offsets)
 816                return -ENOMEM;
 817        i = 0;
 818        /* Walk through entries, checking offsets. */
 819        xt_entry_foreach(iter, entry0, newinfo->size) {
 820                ret = check_entry_size_and_hooks(iter, newinfo, entry0,
 821                                                 entry0 + repl->size,
 822                                                 repl->hook_entry,
 823                                                 repl->underflow,
 824                                                 repl->valid_hooks);
 825                if (ret != 0)
 826                        goto out_free;
 827                if (i < repl->num_entries)
 828                        offsets[i] = (void *)iter - entry0;
 829                ++i;
 830                if (strcmp(ipt_get_target(iter)->u.user.name,
 831                    XT_ERROR_TARGET) == 0)
 832                        ++newinfo->stacksize;
 833        }
 834
 835        ret = -EINVAL;
 836        if (i != repl->num_entries) {
 837                duprintf("translate_table: %u not %u entries\n",
 838                         i, repl->num_entries);
 839                goto out_free;
 840        }
 841
 842        /* Check hooks all assigned */
 843        for (i = 0; i < NF_INET_NUMHOOKS; i++) {
 844                /* Only hooks which are valid */
 845                if (!(repl->valid_hooks & (1 << i)))
 846                        continue;
 847                if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 848                        duprintf("Invalid hook entry %u %u\n",
 849                                 i, repl->hook_entry[i]);
 850                        goto out_free;
 851                }
 852                if (newinfo->underflow[i] == 0xFFFFFFFF) {
 853                        duprintf("Invalid underflow %u %u\n",
 854                                 i, repl->underflow[i]);
 855                        goto out_free;
 856                }
 857        }
 858
 859        if (!mark_source_chains(newinfo, repl->valid_hooks, entry0, offsets)) {
 860                ret = -ELOOP;
 861                goto out_free;
 862        }
 863        kvfree(offsets);
 864
 865        /* Finally, each sanity check must pass */
 866        i = 0;
 867        xt_entry_foreach(iter, entry0, newinfo->size) {
 868                ret = find_check_entry(iter, net, repl->name, repl->size,
 869                                       &alloc_state);
 870                if (ret != 0)
 871                        break;
 872                ++i;
 873        }
 874
 875        if (ret != 0) {
 876                xt_entry_foreach(iter, entry0, newinfo->size) {
 877                        if (i-- == 0)
 878                                break;
 879                        cleanup_entry(iter, net);
 880                }
 881                return ret;
 882        }
 883
 884        return ret;
 885 out_free:
 886        kvfree(offsets);
 887        return ret;
 888}
 889
 890static void
 891get_counters(const struct xt_table_info *t,
 892             struct xt_counters counters[])
 893{
 894        struct ipt_entry *iter;
 895        unsigned int cpu;
 896        unsigned int i;
 897
 898        for_each_possible_cpu(cpu) {
 899                seqcount_t *s = &per_cpu(xt_recseq, cpu);
 900
 901                i = 0;
 902                xt_entry_foreach(iter, t->entries, t->size) {
 903                        struct xt_counters *tmp;
 904                        u64 bcnt, pcnt;
 905                        unsigned int start;
 906
 907                        tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 908                        do {
 909                                start = read_seqcount_begin(s);
 910                                bcnt = tmp->bcnt;
 911                                pcnt = tmp->pcnt;
 912                        } while (read_seqcount_retry(s, start));
 913
 914                        ADD_COUNTER(counters[i], bcnt, pcnt);
 915                        ++i; /* macro does multi eval of i */
 916                        cond_resched();
 917                }
 918        }
 919}
 920
 921static void get_old_counters(const struct xt_table_info *t,
 922                             struct xt_counters counters[])
 923{
 924        struct ipt_entry *iter;
 925        unsigned int cpu, i;
 926
 927        for_each_possible_cpu(cpu) {
 928                i = 0;
 929                xt_entry_foreach(iter, t->entries, t->size) {
 930                        const struct xt_counters *tmp;
 931
 932                        tmp = xt_get_per_cpu_counter(&iter->counters, cpu);
 933                        ADD_COUNTER(counters[i], tmp->bcnt, tmp->pcnt);
 934                        ++i; /* macro does multi eval of i */
 935                }
 936
 937                cond_resched();
 938        }
 939}
 940
 941static struct xt_counters *alloc_counters(const struct xt_table *table)
 942{
 943        unsigned int countersize;
 944        struct xt_counters *counters;
 945        const struct xt_table_info *private = table->private;
 946
 947        /* We need atomic snapshot of counters: rest doesn't change
 948           (other than comefrom, which userspace doesn't care
 949           about). */
 950        countersize = sizeof(struct xt_counters) * private->number;
 951        counters = vzalloc(countersize);
 952
 953        if (counters == NULL)
 954                return ERR_PTR(-ENOMEM);
 955
 956        get_counters(private, counters);
 957
 958        return counters;
 959}
 960
 961static int
 962copy_entries_to_user(unsigned int total_size,
 963                     const struct xt_table *table,
 964                     void __user *userptr)
 965{
 966        unsigned int off, num;
 967        const struct ipt_entry *e;
 968        struct xt_counters *counters;
 969        const struct xt_table_info *private = table->private;
 970        int ret = 0;
 971        const void *loc_cpu_entry;
 972
 973        counters = alloc_counters(table);
 974        if (IS_ERR(counters))
 975                return PTR_ERR(counters);
 976
 977        loc_cpu_entry = private->entries;
 978        if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 979                ret = -EFAULT;
 980                goto free_counters;
 981        }
 982
 983        /* FIXME: use iterator macros --RR */
 984        /* ... then go back and fix counters and names */
 985        for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 986                unsigned int i;
 987                const struct xt_entry_match *m;
 988                const struct xt_entry_target *t;
 989
 990                e = (struct ipt_entry *)(loc_cpu_entry + off);
 991                if (copy_to_user(userptr + off
 992                                 + offsetof(struct ipt_entry, counters),
 993                                 &counters[num],
 994                                 sizeof(counters[num])) != 0) {
 995                        ret = -EFAULT;
 996                        goto free_counters;
 997                }
 998
 999                for (i = sizeof(struct ipt_entry);
1000                     i < e->target_offset;
1001                     i += m->u.match_size) {
1002                        m = (void *)e + i;
1003
1004                        if (copy_to_user(userptr + off + i
1005                                         + offsetof(struct xt_entry_match,
1006                                                    u.user.name),
1007                                         m->u.kernel.match->name,
1008                                         strlen(m->u.kernel.match->name)+1)
1009                            != 0) {
1010                                ret = -EFAULT;
1011                                goto free_counters;
1012                        }
1013                }
1014
1015                t = ipt_get_target_c(e);
1016                if (copy_to_user(userptr + off + e->target_offset
1017                                 + offsetof(struct xt_entry_target,
1018                                            u.user.name),
1019                                 t->u.kernel.target->name,
1020                                 strlen(t->u.kernel.target->name)+1) != 0) {
1021                        ret = -EFAULT;
1022                        goto free_counters;
1023                }
1024        }
1025
1026 free_counters:
1027        vfree(counters);
1028        return ret;
1029}
1030
1031#ifdef CONFIG_COMPAT
1032static void compat_standard_from_user(void *dst, const void *src)
1033{
1034        int v = *(compat_int_t *)src;
1035
1036        if (v > 0)
1037                v += xt_compat_calc_jump(AF_INET, v);
1038        memcpy(dst, &v, sizeof(v));
1039}
1040
1041static int compat_standard_to_user(void __user *dst, const void *src)
1042{
1043        compat_int_t cv = *(int *)src;
1044
1045        if (cv > 0)
1046                cv -= xt_compat_calc_jump(AF_INET, cv);
1047        return copy_to_user(dst, &cv, sizeof(cv)) ? -EFAULT : 0;
1048}
1049
1050static int compat_calc_entry(const struct ipt_entry *e,
1051                             const struct xt_table_info *info,
1052                             const void *base, struct xt_table_info *newinfo)
1053{
1054        const struct xt_entry_match *ematch;
1055        const struct xt_entry_target *t;
1056        unsigned int entry_offset;
1057        int off, i, ret;
1058
1059        off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1060        entry_offset = (void *)e - base;
1061        xt_ematch_foreach(ematch, e)
1062                off += xt_compat_match_offset(ematch->u.kernel.match);
1063        t = ipt_get_target_c(e);
1064        off += xt_compat_target_offset(t->u.kernel.target);
1065        newinfo->size -= off;
1066        ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1067        if (ret)
1068                return ret;
1069
1070        for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1071                if (info->hook_entry[i] &&
1072                    (e < (struct ipt_entry *)(base + info->hook_entry[i])))
1073                        newinfo->hook_entry[i] -= off;
1074                if (info->underflow[i] &&
1075                    (e < (struct ipt_entry *)(base + info->underflow[i])))
1076                        newinfo->underflow[i] -= off;
1077        }
1078        return 0;
1079}
1080
1081static int compat_table_info(const struct xt_table_info *info,
1082                             struct xt_table_info *newinfo)
1083{
1084        struct ipt_entry *iter;
1085        const void *loc_cpu_entry;
1086        int ret;
1087
1088        if (!newinfo || !info)
1089                return -EINVAL;
1090
1091        /* we dont care about newinfo->entries */
1092        memcpy(newinfo, info, offsetof(struct xt_table_info, entries));
1093        newinfo->initial_entries = 0;
1094        loc_cpu_entry = info->entries;
1095        xt_compat_init_offsets(AF_INET, info->number);
1096        xt_entry_foreach(iter, loc_cpu_entry, info->size) {
1097                ret = compat_calc_entry(iter, info, loc_cpu_entry, newinfo);
1098                if (ret != 0)
1099                        return ret;
1100        }
1101        return 0;
1102}
1103#endif
1104
1105static int get_info(struct net *net, void __user *user,
1106                    const int *len, int compat)
1107{
1108        char name[XT_TABLE_MAXNAMELEN];
1109        struct xt_table *t;
1110        int ret;
1111
1112        if (*len != sizeof(struct ipt_getinfo)) {
1113                duprintf("length %u != %zu\n", *len,
1114                         sizeof(struct ipt_getinfo));
1115                return -EINVAL;
1116        }
1117
1118        if (copy_from_user(name, user, sizeof(name)) != 0)
1119                return -EFAULT;
1120
1121        name[XT_TABLE_MAXNAMELEN-1] = '\0';
1122#ifdef CONFIG_COMPAT
1123        if (compat)
1124                xt_compat_lock(AF_INET);
1125#endif
1126        t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1127                                    "iptable_%s", name);
1128        if (!IS_ERR_OR_NULL(t)) {
1129                struct ipt_getinfo info;
1130                const struct xt_table_info *private = t->private;
1131#ifdef CONFIG_COMPAT
1132                struct xt_table_info tmp;
1133
1134                if (compat) {
1135                        ret = compat_table_info(private, &tmp);
1136                        xt_compat_flush_offsets(AF_INET);
1137                        private = &tmp;
1138                }
1139#endif
1140                memset(&info, 0, sizeof(info));
1141                info.valid_hooks = t->valid_hooks;
1142                memcpy(info.hook_entry, private->hook_entry,
1143                       sizeof(info.hook_entry));
1144                memcpy(info.underflow, private->underflow,
1145                       sizeof(info.underflow));
1146                info.num_entries = private->number;
1147                info.size = private->size;
1148                strcpy(info.name, name);
1149
1150                if (copy_to_user(user, &info, *len) != 0)
1151                        ret = -EFAULT;
1152                else
1153                        ret = 0;
1154
1155                xt_table_unlock(t);
1156                module_put(t->me);
1157        } else
1158                ret = t ? PTR_ERR(t) : -ENOENT;
1159#ifdef CONFIG_COMPAT
1160        if (compat)
1161                xt_compat_unlock(AF_INET);
1162#endif
1163        return ret;
1164}
1165
1166static int
1167get_entries(struct net *net, struct ipt_get_entries __user *uptr,
1168            const int *len)
1169{
1170        int ret;
1171        struct ipt_get_entries get;
1172        struct xt_table *t;
1173
1174        if (*len < sizeof(get)) {
1175                duprintf("get_entries: %u < %zu\n", *len, sizeof(get));
1176                return -EINVAL;
1177        }
1178        if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1179                return -EFAULT;
1180        if (*len != sizeof(struct ipt_get_entries) + get.size) {
1181                duprintf("get_entries: %u != %zu\n",
1182                         *len, sizeof(get) + get.size);
1183                return -EINVAL;
1184        }
1185        get.name[sizeof(get.name) - 1] = '\0';
1186
1187        t = xt_find_table_lock(net, AF_INET, get.name);
1188        if (!IS_ERR_OR_NULL(t)) {
1189                const struct xt_table_info *private = t->private;
1190                duprintf("t->private->number = %u\n", private->number);
1191                if (get.size == private->size)
1192                        ret = copy_entries_to_user(private->size,
1193                                                   t, uptr->entrytable);
1194                else {
1195                        duprintf("get_entries: I've got %u not %u!\n",
1196                                 private->size, get.size);
1197                        ret = -EAGAIN;
1198                }
1199                module_put(t->me);
1200                xt_table_unlock(t);
1201        } else
1202                ret = t ? PTR_ERR(t) : -ENOENT;
1203
1204        return ret;
1205}
1206
1207static int
1208__do_replace(struct net *net, const char *name, unsigned int valid_hooks,
1209             struct xt_table_info *newinfo, unsigned int num_counters,
1210             void __user *counters_ptr)
1211{
1212        int ret;
1213        struct xt_table *t;
1214        struct xt_table_info *oldinfo;
1215        struct xt_counters *counters;
1216        struct ipt_entry *iter;
1217
1218        ret = 0;
1219        counters = vzalloc(num_counters * sizeof(struct xt_counters));
1220        if (!counters) {
1221                ret = -ENOMEM;
1222                goto out;
1223        }
1224
1225        t = try_then_request_module(xt_find_table_lock(net, AF_INET, name),
1226                                    "iptable_%s", name);
1227        if (IS_ERR_OR_NULL(t)) {
1228                ret = t ? PTR_ERR(t) : -ENOENT;
1229                goto free_newinfo_counters_untrans;
1230        }
1231
1232        /* You lied! */
1233        if (valid_hooks != t->valid_hooks) {
1234                duprintf("Valid hook crap: %08X vs %08X\n",
1235                         valid_hooks, t->valid_hooks);
1236                ret = -EINVAL;
1237                goto put_module;
1238        }
1239
1240        oldinfo = xt_replace_table(t, num_counters, newinfo, &ret);
1241        if (!oldinfo)
1242                goto put_module;
1243
1244        /* Update module usage count based on number of rules */
1245        duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
1246                oldinfo->number, oldinfo->initial_entries, newinfo->number);
1247        if ((oldinfo->number > oldinfo->initial_entries) ||
1248            (newinfo->number <= oldinfo->initial_entries))
1249                module_put(t->me);
1250        if ((oldinfo->number > oldinfo->initial_entries) &&
1251            (newinfo->number <= oldinfo->initial_entries))
1252                module_put(t->me);
1253
1254        get_old_counters(oldinfo, counters);
1255
1256        /* Decrease module usage counts and free resource */
1257        xt_entry_foreach(iter, oldinfo->entries, oldinfo->size)
1258                cleanup_entry(iter, net);
1259
1260        xt_free_table_info(oldinfo);
1261        if (copy_to_user(counters_ptr, counters,
1262                         sizeof(struct xt_counters) * num_counters) != 0) {
1263                /* Silent error, can't fail, new table is already in place */
1264                net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
1265        }
1266        vfree(counters);
1267        xt_table_unlock(t);
1268        return ret;
1269
1270 put_module:
1271        module_put(t->me);
1272        xt_table_unlock(t);
1273 free_newinfo_counters_untrans:
1274        vfree(counters);
1275 out:
1276        return ret;
1277}
1278
1279static int
1280do_replace(struct net *net, const void __user *user, unsigned int len)
1281{
1282        int ret;
1283        struct ipt_replace tmp;
1284        struct xt_table_info *newinfo;
1285        void *loc_cpu_entry;
1286        struct ipt_entry *iter;
1287
1288        if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1289                return -EFAULT;
1290
1291        /* overflow check */
1292        if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1293                return -ENOMEM;
1294        tmp.name[sizeof(tmp.name)-1] = 0;
1295
1296        newinfo = xt_alloc_table_info(tmp.size);
1297        if (!newinfo)
1298                return -ENOMEM;
1299
1300        loc_cpu_entry = newinfo->entries;
1301        if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1302                           tmp.size) != 0) {
1303                ret = -EFAULT;
1304                goto free_newinfo;
1305        }
1306
1307        ret = translate_table(net, newinfo, loc_cpu_entry, &tmp);
1308        if (ret != 0)
1309                goto free_newinfo;
1310
1311        duprintf("Translated table\n");
1312
1313        ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1314                           tmp.num_counters, tmp.counters);
1315        if (ret)
1316                goto free_newinfo_untrans;
1317        return 0;
1318
1319 free_newinfo_untrans:
1320        xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1321                cleanup_entry(iter, net);
1322 free_newinfo:
1323        xt_free_table_info(newinfo);
1324        return ret;
1325}
1326
1327static int
1328do_add_counters(struct net *net, const void __user *user,
1329                unsigned int len, int compat)
1330{
1331        unsigned int i;
1332        struct xt_counters_info tmp;
1333        struct xt_counters *paddc;
1334        struct xt_table *t;
1335        const struct xt_table_info *private;
1336        int ret = 0;
1337        struct ipt_entry *iter;
1338        unsigned int addend;
1339
1340        paddc = xt_copy_counters_from_user(user, len, &tmp, compat);
1341        if (IS_ERR(paddc))
1342                return PTR_ERR(paddc);
1343
1344        t = xt_find_table_lock(net, AF_INET, tmp.name);
1345        if (IS_ERR_OR_NULL(t)) {
1346                ret = t ? PTR_ERR(t) : -ENOENT;
1347                goto free;
1348        }
1349
1350        local_bh_disable();
1351        private = t->private;
1352        if (private->number != tmp.num_counters) {
1353                ret = -EINVAL;
1354                goto unlock_up_free;
1355        }
1356
1357        i = 0;
1358        addend = xt_write_recseq_begin();
1359        xt_entry_foreach(iter, private->entries, private->size) {
1360                struct xt_counters *tmp;
1361
1362                tmp = xt_get_this_cpu_counter(&iter->counters);
1363                ADD_COUNTER(*tmp, paddc[i].bcnt, paddc[i].pcnt);
1364                ++i;
1365        }
1366        xt_write_recseq_end(addend);
1367 unlock_up_free:
1368        local_bh_enable();
1369        xt_table_unlock(t);
1370        module_put(t->me);
1371 free:
1372        vfree(paddc);
1373
1374        return ret;
1375}
1376
1377#ifdef CONFIG_COMPAT
1378struct compat_ipt_replace {
1379        char                    name[XT_TABLE_MAXNAMELEN];
1380        u32                     valid_hooks;
1381        u32                     num_entries;
1382        u32                     size;
1383        u32                     hook_entry[NF_INET_NUMHOOKS];
1384        u32                     underflow[NF_INET_NUMHOOKS];
1385        u32                     num_counters;
1386        compat_uptr_t           counters;       /* struct xt_counters * */
1387        struct compat_ipt_entry entries[0];
1388};
1389
1390static int
1391compat_copy_entry_to_user(struct ipt_entry *e, void __user **dstptr,
1392                          unsigned int *size, struct xt_counters *counters,
1393                          unsigned int i)
1394{
1395        struct xt_entry_target *t;
1396        struct compat_ipt_entry __user *ce;
1397        u_int16_t target_offset, next_offset;
1398        compat_uint_t origsize;
1399        const struct xt_entry_match *ematch;
1400        int ret = 0;
1401
1402        origsize = *size;
1403        ce = (struct compat_ipt_entry __user *)*dstptr;
1404        if (copy_to_user(ce, e, sizeof(struct ipt_entry)) != 0 ||
1405            copy_to_user(&ce->counters, &counters[i],
1406            sizeof(counters[i])) != 0)
1407                return -EFAULT;
1408
1409        *dstptr += sizeof(struct compat_ipt_entry);
1410        *size -= sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1411
1412        xt_ematch_foreach(ematch, e) {
1413                ret = xt_compat_match_to_user(ematch, dstptr, size);
1414                if (ret != 0)
1415                        return ret;
1416        }
1417        target_offset = e->target_offset - (origsize - *size);
1418        t = ipt_get_target(e);
1419        ret = xt_compat_target_to_user(t, dstptr, size);
1420        if (ret)
1421                return ret;
1422        next_offset = e->next_offset - (origsize - *size);
1423        if (put_user(target_offset, &ce->target_offset) != 0 ||
1424            put_user(next_offset, &ce->next_offset) != 0)
1425                return -EFAULT;
1426        return 0;
1427}
1428
1429static int
1430compat_find_calc_match(struct xt_entry_match *m,
1431                       const struct ipt_ip *ip,
1432                       int *size)
1433{
1434        struct xt_match *match;
1435
1436        match = xt_request_find_match(NFPROTO_IPV4, m->u.user.name,
1437                                      m->u.user.revision);
1438        if (IS_ERR(match)) {
1439                duprintf("compat_check_calc_match: `%s' not found\n",
1440                         m->u.user.name);
1441                return PTR_ERR(match);
1442        }
1443        m->u.kernel.match = match;
1444        *size += xt_compat_match_offset(match);
1445        return 0;
1446}
1447
1448static void compat_release_entry(struct compat_ipt_entry *e)
1449{
1450        struct xt_entry_target *t;
1451        struct xt_entry_match *ematch;
1452
1453        /* Cleanup all matches */
1454        xt_ematch_foreach(ematch, e)
1455                module_put(ematch->u.kernel.match->me);
1456        t = compat_ipt_get_target(e);
1457        module_put(t->u.kernel.target->me);
1458}
1459
1460static int
1461check_compat_entry_size_and_hooks(struct compat_ipt_entry *e,
1462                                  struct xt_table_info *newinfo,
1463                                  unsigned int *size,
1464                                  const unsigned char *base,
1465                                  const unsigned char *limit)
1466{
1467        struct xt_entry_match *ematch;
1468        struct xt_entry_target *t;
1469        struct xt_target *target;
1470        unsigned int entry_offset;
1471        unsigned int j;
1472        int ret, off;
1473
1474        duprintf("check_compat_entry_size_and_hooks %p\n", e);
1475        if ((unsigned long)e % __alignof__(struct compat_ipt_entry) != 0 ||
1476            (unsigned char *)e + sizeof(struct compat_ipt_entry) >= limit ||
1477            (unsigned char *)e + e->next_offset > limit) {
1478                duprintf("Bad offset %p, limit = %p\n", e, limit);
1479                return -EINVAL;
1480        }
1481
1482        if (e->next_offset < sizeof(struct compat_ipt_entry) +
1483                             sizeof(struct compat_xt_entry_target)) {
1484                duprintf("checking: element %p size %u\n",
1485                         e, e->next_offset);
1486                return -EINVAL;
1487        }
1488
1489        if (!ip_checkentry(&e->ip))
1490                return -EINVAL;
1491
1492        ret = xt_compat_check_entry_offsets(e, e->elems,
1493                                            e->target_offset, e->next_offset);
1494        if (ret)
1495                return ret;
1496
1497        off = sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1498        entry_offset = (void *)e - (void *)base;
1499        j = 0;
1500        xt_ematch_foreach(ematch, e) {
1501                ret = compat_find_calc_match(ematch, &e->ip, &off);
1502                if (ret != 0)
1503                        goto release_matches;
1504                ++j;
1505        }
1506
1507        t = compat_ipt_get_target(e);
1508        target = xt_request_find_target(NFPROTO_IPV4, t->u.user.name,
1509                                        t->u.user.revision);
1510        if (IS_ERR(target)) {
1511                duprintf("check_compat_entry_size_and_hooks: `%s' not found\n",
1512                         t->u.user.name);
1513                ret = PTR_ERR(target);
1514                goto release_matches;
1515        }
1516        t->u.kernel.target = target;
1517
1518        off += xt_compat_target_offset(target);
1519        *size += off;
1520        ret = xt_compat_add_offset(AF_INET, entry_offset, off);
1521        if (ret)
1522                goto out;
1523
1524        return 0;
1525
1526out:
1527        module_put(t->u.kernel.target->me);
1528release_matches:
1529        xt_ematch_foreach(ematch, e) {
1530                if (j-- == 0)
1531                        break;
1532                module_put(ematch->u.kernel.match->me);
1533        }
1534        return ret;
1535}
1536
1537static void
1538compat_copy_entry_from_user(struct compat_ipt_entry *e, void **dstptr,
1539                            unsigned int *size,
1540                            struct xt_table_info *newinfo, unsigned char *base)
1541{
1542        struct xt_entry_target *t;
1543        struct xt_target *target;
1544        struct ipt_entry *de;
1545        unsigned int origsize;
1546        int h;
1547        struct xt_entry_match *ematch;
1548
1549        origsize = *size;
1550        de = (struct ipt_entry *)*dstptr;
1551        memcpy(de, e, sizeof(struct ipt_entry));
1552        memcpy(&de->counters, &e->counters, sizeof(e->counters));
1553
1554        *dstptr += sizeof(struct ipt_entry);
1555        *size += sizeof(struct ipt_entry) - sizeof(struct compat_ipt_entry);
1556
1557        xt_ematch_foreach(ematch, e)
1558                xt_compat_match_from_user(ematch, dstptr, size);
1559
1560        de->target_offset = e->target_offset - (origsize - *size);
1561        t = compat_ipt_get_target(e);
1562        target = t->u.kernel.target;
1563        xt_compat_target_from_user(t, dstptr, size);
1564
1565        de->next_offset = e->next_offset - (origsize - *size);
1566
1567        for (h = 0; h < NF_INET_NUMHOOKS; h++) {
1568                if ((unsigned char *)de - base < newinfo->hook_entry[h])
1569                        newinfo->hook_entry[h] -= origsize - *size;
1570                if ((unsigned char *)de - base < newinfo->underflow[h])
1571                        newinfo->underflow[h] -= origsize - *size;
1572        }
1573}
1574
1575static int
1576translate_compat_table(struct net *net,
1577                       struct xt_table_info **pinfo,
1578                       void **pentry0,
1579                       const struct compat_ipt_replace *compatr)
1580{
1581        unsigned int i, j;
1582        struct xt_table_info *newinfo, *info;
1583        void *pos, *entry0, *entry1;
1584        struct compat_ipt_entry *iter0;
1585        struct ipt_replace repl;
1586        unsigned int size;
1587        int ret;
1588
1589        info = *pinfo;
1590        entry0 = *pentry0;
1591        size = compatr->size;
1592        info->number = compatr->num_entries;
1593
1594        duprintf("translate_compat_table: size %u\n", info->size);
1595        j = 0;
1596        xt_compat_lock(AF_INET);
1597        xt_compat_init_offsets(AF_INET, compatr->num_entries);
1598        /* Walk through entries, checking offsets. */
1599        xt_entry_foreach(iter0, entry0, compatr->size) {
1600                ret = check_compat_entry_size_and_hooks(iter0, info, &size,
1601                                                        entry0,
1602                                                        entry0 + compatr->size);
1603                if (ret != 0)
1604                        goto out_unlock;
1605                ++j;
1606        }
1607
1608        ret = -EINVAL;
1609        if (j != compatr->num_entries) {
1610                duprintf("translate_compat_table: %u not %u entries\n",
1611                         j, compatr->num_entries);
1612                goto out_unlock;
1613        }
1614
1615        ret = -ENOMEM;
1616        newinfo = xt_alloc_table_info(size);
1617        if (!newinfo)
1618                goto out_unlock;
1619
1620        newinfo->number = compatr->num_entries;
1621        for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1622                newinfo->hook_entry[i] = compatr->hook_entry[i];
1623                newinfo->underflow[i] = compatr->underflow[i];
1624        }
1625        entry1 = newinfo->entries;
1626        pos = entry1;
1627        size = compatr->size;
1628        xt_entry_foreach(iter0, entry0, compatr->size)
1629                compat_copy_entry_from_user(iter0, &pos, &size,
1630                                            newinfo, entry1);
1631
1632        /* all module references in entry0 are now gone.
1633         * entry1/newinfo contains a 64bit ruleset that looks exactly as
1634         * generated by 64bit userspace.
1635         *
1636         * Call standard translate_table() to validate all hook_entrys,
1637         * underflows, check for loops, etc.
1638         */
1639        xt_compat_flush_offsets(AF_INET);
1640        xt_compat_unlock(AF_INET);
1641
1642        memcpy(&repl, compatr, sizeof(*compatr));
1643
1644        for (i = 0; i < NF_INET_NUMHOOKS; i++) {
1645                repl.hook_entry[i] = newinfo->hook_entry[i];
1646                repl.underflow[i] = newinfo->underflow[i];
1647        }
1648
1649        repl.num_counters = 0;
1650        repl.counters = NULL;
1651        repl.size = newinfo->size;
1652        ret = translate_table(net, newinfo, entry1, &repl);
1653        if (ret)
1654                goto free_newinfo;
1655
1656        *pinfo = newinfo;
1657        *pentry0 = entry1;
1658        xt_free_table_info(info);
1659        return 0;
1660
1661free_newinfo:
1662        xt_free_table_info(newinfo);
1663        return ret;
1664out_unlock:
1665        xt_compat_flush_offsets(AF_INET);
1666        xt_compat_unlock(AF_INET);
1667        xt_entry_foreach(iter0, entry0, compatr->size) {
1668                if (j-- == 0)
1669                        break;
1670                compat_release_entry(iter0);
1671        }
1672        return ret;
1673}
1674
1675static int
1676compat_do_replace(struct net *net, void __user *user, unsigned int len)
1677{
1678        int ret;
1679        struct compat_ipt_replace tmp;
1680        struct xt_table_info *newinfo;
1681        void *loc_cpu_entry;
1682        struct ipt_entry *iter;
1683
1684        if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
1685                return -EFAULT;
1686
1687        /* overflow check */
1688        if (tmp.size >= INT_MAX / num_possible_cpus())
1689                return -ENOMEM;
1690        if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
1691                return -ENOMEM;
1692        tmp.name[sizeof(tmp.name)-1] = 0;
1693
1694        newinfo = xt_alloc_table_info(tmp.size);
1695        if (!newinfo)
1696                return -ENOMEM;
1697
1698        loc_cpu_entry = newinfo->entries;
1699        if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
1700                           tmp.size) != 0) {
1701                ret = -EFAULT;
1702                goto free_newinfo;
1703        }
1704
1705        ret = translate_compat_table(net, &newinfo, &loc_cpu_entry, &tmp);
1706        if (ret != 0)
1707                goto free_newinfo;
1708
1709        duprintf("compat_do_replace: Translated table\n");
1710
1711        ret = __do_replace(net, tmp.name, tmp.valid_hooks, newinfo,
1712                           tmp.num_counters, compat_ptr(tmp.counters));
1713        if (ret)
1714                goto free_newinfo_untrans;
1715        return 0;
1716
1717 free_newinfo_untrans:
1718        xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
1719                cleanup_entry(iter, net);
1720 free_newinfo:
1721        xt_free_table_info(newinfo);
1722        return ret;
1723}
1724
1725static int
1726compat_do_ipt_set_ctl(struct sock *sk,  int cmd, void __user *user,
1727                      unsigned int len)
1728{
1729        int ret;
1730
1731        if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1732                return -EPERM;
1733
1734        switch (cmd) {
1735        case IPT_SO_SET_REPLACE:
1736                ret = compat_do_replace(sock_net(sk), user, len);
1737                break;
1738
1739        case IPT_SO_SET_ADD_COUNTERS:
1740                ret = do_add_counters(sock_net(sk), user, len, 1);
1741                break;
1742
1743        default:
1744                duprintf("do_ipt_set_ctl:  unknown request %i\n", cmd);
1745                ret = -EINVAL;
1746        }
1747
1748        return ret;
1749}
1750
1751struct compat_ipt_get_entries {
1752        char name[XT_TABLE_MAXNAMELEN];
1753        compat_uint_t size;
1754        struct compat_ipt_entry entrytable[0];
1755};
1756
1757static int
1758compat_copy_entries_to_user(unsigned int total_size, struct xt_table *table,
1759                            void __user *userptr)
1760{
1761        struct xt_counters *counters;
1762        const struct xt_table_info *private = table->private;
1763        void __user *pos;
1764        unsigned int size;
1765        int ret = 0;
1766        unsigned int i = 0;
1767        struct ipt_entry *iter;
1768
1769        counters = alloc_counters(table);
1770        if (IS_ERR(counters))
1771                return PTR_ERR(counters);
1772
1773        pos = userptr;
1774        size = total_size;
1775        xt_entry_foreach(iter, private->entries, total_size) {
1776                ret = compat_copy_entry_to_user(iter, &pos,
1777                                                &size, counters, i++);
1778                if (ret != 0)
1779                        break;
1780        }
1781
1782        vfree(counters);
1783        return ret;
1784}
1785
1786static int
1787compat_get_entries(struct net *net, struct compat_ipt_get_entries __user *uptr,
1788                   int *len)
1789{
1790        int ret;
1791        struct compat_ipt_get_entries get;
1792        struct xt_table *t;
1793
1794        if (*len < sizeof(get)) {
1795                duprintf("compat_get_entries: %u < %zu\n", *len, sizeof(get));
1796                return -EINVAL;
1797        }
1798
1799        if (copy_from_user(&get, uptr, sizeof(get)) != 0)
1800                return -EFAULT;
1801
1802        if (*len != sizeof(struct compat_ipt_get_entries) + get.size) {
1803                duprintf("compat_get_entries: %u != %zu\n",
1804                         *len, sizeof(get) + get.size);
1805                return -EINVAL;
1806        }
1807        get.name[sizeof(get.name) - 1] = '\0';
1808
1809        xt_compat_lock(AF_INET);
1810        t = xt_find_table_lock(net, AF_INET, get.name);
1811        if (!IS_ERR_OR_NULL(t)) {
1812                const struct xt_table_info *private = t->private;
1813                struct xt_table_info info;
1814                duprintf("t->private->number = %u\n", private->number);
1815                ret = compat_table_info(private, &info);
1816                if (!ret && get.size == info.size) {
1817                        ret = compat_copy_entries_to_user(private->size,
1818                                                          t, uptr->entrytable);
1819                } else if (!ret) {
1820                        duprintf("compat_get_entries: I've got %u not %u!\n",
1821                                 private->size, get.size);
1822                        ret = -EAGAIN;
1823                }
1824                xt_compat_flush_offsets(AF_INET);
1825                module_put(t->me);
1826                xt_table_unlock(t);
1827        } else
1828                ret = t ? PTR_ERR(t) : -ENOENT;
1829
1830        xt_compat_unlock(AF_INET);
1831        return ret;
1832}
1833
1834static int do_ipt_get_ctl(struct sock *, int, void __user *, int *);
1835
1836static int
1837compat_do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1838{
1839        int ret;
1840
1841        if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1842                return -EPERM;
1843
1844        switch (cmd) {
1845        case IPT_SO_GET_INFO:
1846                ret = get_info(sock_net(sk), user, len, 1);
1847                break;
1848        case IPT_SO_GET_ENTRIES:
1849                ret = compat_get_entries(sock_net(sk), user, len);
1850                break;
1851        default:
1852                ret = do_ipt_get_ctl(sk, cmd, user, len);
1853        }
1854        return ret;
1855}
1856#endif
1857
1858static int
1859do_ipt_set_ctl(struct sock *sk, int cmd, void __user *user, 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 = 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, 0);
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
1883static int
1884do_ipt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1885{
1886        int ret;
1887
1888        if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1889                return -EPERM;
1890
1891        switch (cmd) {
1892        case IPT_SO_GET_INFO:
1893                ret = get_info(sock_net(sk), user, len, 0);
1894                break;
1895
1896        case IPT_SO_GET_ENTRIES:
1897                ret = get_entries(sock_net(sk), user, len);
1898                break;
1899
1900        case IPT_SO_GET_REVISION_MATCH:
1901        case IPT_SO_GET_REVISION_TARGET: {
1902                struct xt_get_revision rev;
1903                int target;
1904
1905                if (*len != sizeof(rev)) {
1906                        ret = -EINVAL;
1907                        break;
1908                }
1909                if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1910                        ret = -EFAULT;
1911                        break;
1912                }
1913                rev.name[sizeof(rev.name)-1] = 0;
1914
1915                if (cmd == IPT_SO_GET_REVISION_TARGET)
1916                        target = 1;
1917                else
1918                        target = 0;
1919
1920                try_then_request_module(xt_find_revision(AF_INET, rev.name,
1921                                                         rev.revision,
1922                                                         target, &ret),
1923                                        "ipt_%s", rev.name);
1924                break;
1925        }
1926
1927        default:
1928                duprintf("do_ipt_get_ctl: unknown request %i\n", cmd);
1929                ret = -EINVAL;
1930        }
1931
1932        return ret;
1933}
1934
1935struct xt_table *ipt_register_table(struct net *net,
1936                                    const struct xt_table *table,
1937                                    const struct ipt_replace *repl)
1938{
1939        int ret;
1940        struct xt_table_info *newinfo;
1941        struct xt_table_info bootstrap = {0};
1942        void *loc_cpu_entry;
1943        struct xt_table *new_table;
1944
1945        newinfo = xt_alloc_table_info(repl->size);
1946        if (!newinfo) {
1947                ret = -ENOMEM;
1948                goto out;
1949        }
1950
1951        loc_cpu_entry = newinfo->entries;
1952        memcpy(loc_cpu_entry, repl->entries, repl->size);
1953
1954        ret = translate_table(net, newinfo, loc_cpu_entry, repl);
1955        if (ret != 0)
1956                goto out_free;
1957
1958        new_table = xt_register_table(net, table, &bootstrap, newinfo);
1959        if (IS_ERR(new_table)) {
1960                ret = PTR_ERR(new_table);
1961                goto out_free;
1962        }
1963
1964        return new_table;
1965
1966out_free:
1967        xt_free_table_info(newinfo);
1968out:
1969        return ERR_PTR(ret);
1970}
1971
1972void ipt_unregister_table(struct net *net, struct xt_table *table)
1973{
1974        struct xt_table_info *private;
1975        void *loc_cpu_entry;
1976        struct module *table_owner = table->me;
1977        struct ipt_entry *iter;
1978
1979        private = xt_unregister_table(table);
1980
1981        /* Decrease module usage counts and free resources */
1982        loc_cpu_entry = private->entries;
1983        xt_entry_foreach(iter, loc_cpu_entry, private->size)
1984                cleanup_entry(iter, net);
1985        if (private->number > private->initial_entries)
1986                module_put(table_owner);
1987        xt_free_table_info(private);
1988}
1989
1990/* Returns 1 if the type and code is matched by the range, 0 otherwise */
1991static inline bool
1992icmp_type_code_match(u_int8_t test_type, u_int8_t min_code, u_int8_t max_code,
1993                     u_int8_t type, u_int8_t code,
1994                     bool invert)
1995{
1996        return ((test_type == 0xFF) ||
1997                (type == test_type && code >= min_code && code <= max_code))
1998                ^ invert;
1999}
2000
2001static bool
2002icmp_match(const struct sk_buff *skb, struct xt_action_param *par)
2003{
2004        const struct icmphdr *ic;
2005        struct icmphdr _icmph;
2006        const struct ipt_icmp *icmpinfo = par->matchinfo;
2007
2008        /* Must not be a fragment. */
2009        if (par->fragoff != 0)
2010                return false;
2011
2012        ic = skb_header_pointer(skb, par->thoff, sizeof(_icmph), &_icmph);
2013        if (ic == NULL) {
2014                /* We've been asked to examine this packet, and we
2015                 * can't.  Hence, no choice but to drop.
2016                 */
2017                duprintf("Dropping evil ICMP tinygram.\n");
2018                par->hotdrop = true;
2019                return false;
2020        }
2021
2022        return icmp_type_code_match(icmpinfo->type,
2023                                    icmpinfo->code[0],
2024                                    icmpinfo->code[1],
2025                                    ic->type, ic->code,
2026                                    !!(icmpinfo->invflags&IPT_ICMP_INV));
2027}
2028
2029static int icmp_checkentry(const struct xt_mtchk_param *par)
2030{
2031        const struct ipt_icmp *icmpinfo = par->matchinfo;
2032
2033        /* Must specify no unknown invflags */
2034        return (icmpinfo->invflags & ~IPT_ICMP_INV) ? -EINVAL : 0;
2035}
2036
2037static struct xt_target ipt_builtin_tg[] __read_mostly = {
2038        {
2039                .name             = XT_STANDARD_TARGET,
2040                .targetsize       = sizeof(int),
2041                .family           = NFPROTO_IPV4,
2042#ifdef CONFIG_COMPAT
2043                .compatsize       = sizeof(compat_int_t),
2044                .compat_from_user = compat_standard_from_user,
2045                .compat_to_user   = compat_standard_to_user,
2046#endif
2047        },
2048        {
2049                .name             = XT_ERROR_TARGET,
2050                .target           = ipt_error,
2051                .targetsize       = XT_FUNCTION_MAXNAMELEN,
2052                .family           = NFPROTO_IPV4,
2053        },
2054};
2055
2056static struct nf_sockopt_ops ipt_sockopts = {
2057        .pf             = PF_INET,
2058        .set_optmin     = IPT_BASE_CTL,
2059        .set_optmax     = IPT_SO_SET_MAX+1,
2060        .set            = do_ipt_set_ctl,
2061#ifdef CONFIG_COMPAT
2062        .compat_set     = compat_do_ipt_set_ctl,
2063#endif
2064        .get_optmin     = IPT_BASE_CTL,
2065        .get_optmax     = IPT_SO_GET_MAX+1,
2066        .get            = do_ipt_get_ctl,
2067#ifdef CONFIG_COMPAT
2068        .compat_get     = compat_do_ipt_get_ctl,
2069#endif
2070        .owner          = THIS_MODULE,
2071};
2072
2073static struct xt_match ipt_builtin_mt[] __read_mostly = {
2074        {
2075                .name       = "icmp",
2076                .match      = icmp_match,
2077                .matchsize  = sizeof(struct ipt_icmp),
2078                .checkentry = icmp_checkentry,
2079                .proto      = IPPROTO_ICMP,
2080                .family     = NFPROTO_IPV4,
2081        },
2082};
2083
2084static int __net_init ip_tables_net_init(struct net *net)
2085{
2086        return xt_proto_init(net, NFPROTO_IPV4);
2087}
2088
2089static void __net_exit ip_tables_net_exit(struct net *net)
2090{
2091        xt_proto_fini(net, NFPROTO_IPV4);
2092}
2093
2094static struct pernet_operations ip_tables_net_ops = {
2095        .init = ip_tables_net_init,
2096        .exit = ip_tables_net_exit,
2097};
2098
2099static int __init ip_tables_init(void)
2100{
2101        int ret;
2102
2103        ret = register_pernet_subsys(&ip_tables_net_ops);
2104        if (ret < 0)
2105                goto err1;
2106
2107        /* No one else will be downing sem now, so we won't sleep */
2108        ret = xt_register_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2109        if (ret < 0)
2110                goto err2;
2111        ret = xt_register_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2112        if (ret < 0)
2113                goto err4;
2114
2115        /* Register setsockopt */
2116        ret = nf_register_sockopt(&ipt_sockopts);
2117        if (ret < 0)
2118                goto err5;
2119
2120        pr_info("(C) 2000-2006 Netfilter Core Team\n");
2121        return 0;
2122
2123err5:
2124        xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2125err4:
2126        xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2127err2:
2128        unregister_pernet_subsys(&ip_tables_net_ops);
2129err1:
2130        return ret;
2131}
2132
2133static void __exit ip_tables_fini(void)
2134{
2135        nf_unregister_sockopt(&ipt_sockopts);
2136
2137        xt_unregister_matches(ipt_builtin_mt, ARRAY_SIZE(ipt_builtin_mt));
2138        xt_unregister_targets(ipt_builtin_tg, ARRAY_SIZE(ipt_builtin_tg));
2139        unregister_pernet_subsys(&ip_tables_net_ops);
2140}
2141
2142EXPORT_SYMBOL(ipt_register_table);
2143EXPORT_SYMBOL(ipt_unregister_table);
2144EXPORT_SYMBOL(ipt_do_table);
2145module_init(ip_tables_init);
2146module_exit(ip_tables_fini);
2147