net/ipv4/netfilter/arp_tables.c

   1 /*
   2  * Packet matching code for ARP packets.
   3  *
   4  * Based heavily, if not almost entirely, upon ip_tables.c framework.
   5  *
   6  * Some ARP specific bits are:
   7  *
   8  * Copyright (C) 2002 David S. Miller (davem@redhat.com)
   9  *
  10  */
  11
  12 #include <linux/kernel.h>
  13 #include <linux/skbuff.h>
  14 #include <linux/netdevice.h>
  15 #include <linux/capability.h>
  16 #include <linux/if_arp.h>
  17 #include <linux/kmod.h>
  18 #include <linux/vmalloc.h>
  19 #include <linux/proc_fs.h>
  20 #include <linux/module.h>
  21 #include <linux/init.h>
  22
  23 #include <asm/uaccess.h>
  24 #include <linux/mutex.h>
  25
  26 #include <linux/netfilter/x_tables.h>
  27 #include <linux/netfilter_arp/arp_tables.h>
  28
  29 MODULE_LICENSE("GPL");
  30 MODULE_AUTHOR("David S. Miller <davem@redhat.com>");
  31 MODULE_DESCRIPTION("arptables core");
  32
  33 /*#define DEBUG_ARP_TABLES*/
  34 /*#define DEBUG_ARP_TABLES_USER*/
  35
  36 #ifdef DEBUG_ARP_TABLES
  37 #define dprintf(format, args...)  printk(format , ## args)
  38 #else
  39 #define dprintf(format, args...)
  40 #endif
  41
  42 #ifdef DEBUG_ARP_TABLES_USER
  43 #define duprintf(format, args...) printk(format , ## args)
  44 #else
  45 #define duprintf(format, args...)
  46 #endif
  47
  48 #ifdef CONFIG_NETFILTER_DEBUG
  49 #define ARP_NF_ASSERT(x)                                        \
  50 do {                                                            \
  51         if (!(x))                                               \
  52                 printk("ARP_NF_ASSERT: %s:%s:%u\n",             \
  53                        __FUNCTION__, __FILE__, __LINE__);       \
  54 } while(0)
  55 #else
  56 #define ARP_NF_ASSERT(x)
  57 #endif
  58
  59 #include <linux/netfilter_ipv4/listhelp.h>
  60
  61 static inline int arp_devaddr_compare(const struct arpt_devaddr_info *ap,
  62                                       char *hdr_addr, int len)
  63 {
  64         int i, ret;
  65
  66         if (len > ARPT_DEV_ADDR_LEN_MAX)
  67                 len = ARPT_DEV_ADDR_LEN_MAX;
  68
  69         ret = 0;
  70         for (i = 0; i < len; i++)
  71                 ret |= (hdr_addr[i] ^ ap->addr[i]) & ap->mask[i];
  72
  73         return (ret != 0);
  74 }
  75
  76 /* Returns whether packet matches rule or not. */
  77 static inline int arp_packet_match(const struct arphdr *arphdr,
  78                                    struct net_device *dev,
  79                                    const char *indev,
  80                                    const char *outdev,
  81                                    const struct arpt_arp *arpinfo)
  82 {
  83         char *arpptr = (char *)(arphdr + 1);
  84         char *src_devaddr, *tgt_devaddr;
  85         u32 src_ipaddr, tgt_ipaddr;
  86         int i, ret;
  87
  88 #define FWINV(bool,invflg) ((bool) ^ !!(arpinfo->invflags & invflg))
  89
  90         if (FWINV((arphdr->ar_op & arpinfo->arpop_mask) != arpinfo->arpop,
  91                   ARPT_INV_ARPOP)) {
  92                 dprintf("ARP operation field mismatch.\n");
  93                 dprintf("ar_op: %04x info->arpop: %04x info->arpop_mask: %04x\n",
  94                         arphdr->ar_op, arpinfo->arpop, arpinfo->arpop_mask);
  95                 return 0;
  96         }
  97
  98         if (FWINV((arphdr->ar_hrd & arpinfo->arhrd_mask) != arpinfo->arhrd,
  99                   ARPT_INV_ARPHRD)) {
 100                 dprintf("ARP hardware address format mismatch.\n");
 101                 dprintf("ar_hrd: %04x info->arhrd: %04x info->arhrd_mask: %04x\n",
 102                         arphdr->ar_hrd, arpinfo->arhrd, arpinfo->arhrd_mask);
 103                 return 0;
 104         }
 105
 106         if (FWINV((arphdr->ar_pro & arpinfo->arpro_mask) != arpinfo->arpro,
 107                   ARPT_INV_ARPPRO)) {
 108                 dprintf("ARP protocol address format mismatch.\n");
 109                 dprintf("ar_pro: %04x info->arpro: %04x info->arpro_mask: %04x\n",
 110                         arphdr->ar_pro, arpinfo->arpro, arpinfo->arpro_mask);
 111                 return 0;
 112         }
 113
 114         if (FWINV((arphdr->ar_hln & arpinfo->arhln_mask) != arpinfo->arhln,
 115                   ARPT_INV_ARPHLN)) {
 116                 dprintf("ARP hardware address length mismatch.\n");
 117                 dprintf("ar_hln: %02x info->arhln: %02x info->arhln_mask: %02x\n",
 118                         arphdr->ar_hln, arpinfo->arhln, arpinfo->arhln_mask);
 119                 return 0;
 120         }
 121
 122         src_devaddr = arpptr;
 123         arpptr += dev->addr_len;
 124         memcpy(&src_ipaddr, arpptr, sizeof(u32));
 125         arpptr += sizeof(u32);
 126         tgt_devaddr = arpptr;
 127         arpptr += dev->addr_len;
 128         memcpy(&tgt_ipaddr, arpptr, sizeof(u32));
 129
 130         if (FWINV(arp_devaddr_compare(&arpinfo->src_devaddr, src_devaddr, dev->addr_len),
 131                   ARPT_INV_SRCDEVADDR) ||
 132             FWINV(arp_devaddr_compare(&arpinfo->tgt_devaddr, tgt_devaddr, dev->addr_len),
 133                   ARPT_INV_TGTDEVADDR)) {
 134                 dprintf("Source or target device address mismatch.\n");
 135
 136                 return 0;
 137         }
 138
 139         if (FWINV((src_ipaddr & arpinfo->smsk.s_addr) != arpinfo->src.s_addr,
 140                   ARPT_INV_SRCIP) ||
 141             FWINV(((tgt_ipaddr & arpinfo->tmsk.s_addr) != arpinfo->tgt.s_addr),
 142                   ARPT_INV_TGTIP)) {
 143                 dprintf("Source or target IP address mismatch.\n");
 144
 145                 dprintf("SRC: %u.%u.%u.%u. Mask: %u.%u.%u.%u. Target: %u.%u.%u.%u.%s\n",
 146                         NIPQUAD(src_ipaddr),
 147                         NIPQUAD(arpinfo->smsk.s_addr),
 148                         NIPQUAD(arpinfo->src.s_addr),
 149                         arpinfo->invflags & ARPT_INV_SRCIP ? " (INV)" : "");
 150                 dprintf("TGT: %u.%u.%u.%u Mask: %u.%u.%u.%u Target: %u.%u.%u.%u.%s\n",
 151                         NIPQUAD(tgt_ipaddr),
 152                         NIPQUAD(arpinfo->tmsk.s_addr),
 153                         NIPQUAD(arpinfo->tgt.s_addr),
 154                         arpinfo->invflags & ARPT_INV_TGTIP ? " (INV)" : "");
 155                 return 0;
 156         }
 157
 158         /* Look for ifname matches.  */
 159         for (i = 0, ret = 0; i < IFNAMSIZ; i++) {
 160                 ret |= (indev[i] ^ arpinfo->iniface[i])
 161                         & arpinfo->iniface_mask[i];
 162         }
 163
 164         if (FWINV(ret != 0, ARPT_INV_VIA_IN)) {
 165                 dprintf("VIA in mismatch (%s vs %s).%s\n",
 166                         indev, arpinfo->iniface,
 167                         arpinfo->invflags&ARPT_INV_VIA_IN ?" (INV)":"");
 168                 return 0;
 169         }
 170
 171         for (i = 0, ret = 0; i < IFNAMSIZ/sizeof(unsigned long); i++) {
 172                 unsigned long odev;
 173                 memcpy(&odev, outdev + i*sizeof(unsigned long),
 174                        sizeof(unsigned long));
 175                 ret |= (odev
 176                         ^ ((const unsigned long *)arpinfo->outiface)[i])
 177                         & ((const unsigned long *)arpinfo->outiface_mask)[i];
 178         }
 179
 180         if (FWINV(ret != 0, ARPT_INV_VIA_OUT)) {
 181                 dprintf("VIA out mismatch (%s vs %s).%s\n",
 182                         outdev, arpinfo->outiface,
 183                         arpinfo->invflags&ARPT_INV_VIA_OUT ?" (INV)":"");
 184                 return 0;
 185         }
 186
 187         return 1;
 188 }
 189
 190 static inline int arp_checkentry(const struct arpt_arp *arp)
 191 {
 192         if (arp->flags & ~ARPT_F_MASK) {
 193                 duprintf("Unknown flag bits set: %08X\n",
 194                          arp->flags & ~ARPT_F_MASK);
 195                 return 0;
 196         }
 197         if (arp->invflags & ~ARPT_INV_MASK) {
 198                 duprintf("Unknown invflag bits set: %08X\n",
 199                          arp->invflags & ~ARPT_INV_MASK);
 200                 return 0;
 201         }
 202
 203         return 1;
 204 }
 205
 206 static unsigned int arpt_error(struct sk_buff **pskb,
 207                                const struct net_device *in,
 208                                const struct net_device *out,
 209                                unsigned int hooknum,
 210                                const struct xt_target *target,
 211                                const void *targinfo,
 212                                void *userinfo)
 213 {
 214         if (net_ratelimit())
 215                 printk("arp_tables: error: '%s'\n", (char *)targinfo);
 216
 217         return NF_DROP;
 218 }
 219
 220 static inline struct arpt_entry *get_entry(void *base, unsigned int offset)
 221 {
 222         return (struct arpt_entry *)(base + offset);
 223 }
 224
 225 unsigned int arpt_do_table(struct sk_buff **pskb,
 226                            unsigned int hook,
 227                            const struct net_device *in,
 228                            const struct net_device *out,
 229                            struct arpt_table *table,
 230                            void *userdata)
 231 {
 232         static const char nulldevname[IFNAMSIZ];
 233         unsigned int verdict = NF_DROP;
 234         struct arphdr *arp;
 235         int hotdrop = 0;
 236         struct arpt_entry *e, *back;
 237         const char *indev, *outdev;
 238         void *table_base;
 239         struct xt_table_info *private;
 240
 241         /* ARP header, plus 2 device addresses, plus 2 IP addresses.  */
 242         if (!pskb_may_pull((*pskb), (sizeof(struct arphdr) +
 243                                      (2 * (*pskb)->dev->addr_len) +
 244                                      (2 * sizeof(u32)))))
 245                 return NF_DROP;
 246
 247         indev = in ? in->name : nulldevname;
 248         outdev = out ? out->name : nulldevname;
 249
 250         read_lock_bh(&table->lock);
 251         private = table->private;
 252         table_base = (void *)private->entries[smp_processor_id()];
 253         e = get_entry(table_base, private->hook_entry[hook]);
 254         back = get_entry(table_base, private->underflow[hook]);
 255
 256         arp = (*pskb)->nh.arph;
 257         do {
 258                 if (arp_packet_match(arp, (*pskb)->dev, indev, outdev, &e->arp)) {
 259                         struct arpt_entry_target *t;
 260                         int hdr_len;
 261
 262                         hdr_len = sizeof(*arp) + (2 * sizeof(struct in_addr)) +
 263                                 (2 * (*pskb)->dev->addr_len);
 264                         ADD_COUNTER(e->counters, hdr_len, 1);
 265
 266                         t = arpt_get_target(e);
 267
 268                         /* Standard target? */
 269                         if (!t->u.kernel.target->target) {
 270                                 int v;
 271
 272                                 v = ((struct arpt_standard_target *)t)->verdict;
 273                                 if (v < 0) {
 274                                         /* Pop from stack? */
 275                                         if (v != ARPT_RETURN) {
 276                                                 verdict = (unsigned)(-v) - 1;
 277                                                 break;
 278                                         }
 279                                         e = back;
 280                                         back = get_entry(table_base,
 281                                                          back->comefrom);
 282                                         continue;
 283                                 }
 284                                 if (table_base + v
 285                                     != (void *)e + e->next_offset) {
 286                                         /* Save old back ptr in next entry */
 287                                         struct arpt_entry *next
 288                                                 = (void *)e + e->next_offset;
 289                                         next->comefrom =
 290                                                 (void *)back - table_base;
 291
 292                                         /* set back pointer to next entry */
 293                                         back = next;
 294                                 }
 295
 296                                 e = get_entry(table_base, v);
 297                         } else {
 298                                 /* Targets which reenter must return
 299                                  * abs. verdicts
 300                                  */
 301                                 verdict = t->u.kernel.target->target(pskb,
 302                                                                      in, out,
 303                                                                      hook,
 304                                                                      t->u.kernel.target,
 305                                                                      t->data,
 306                                                                      userdata);
 307
 308                                 /* Target might have changed stuff. */
 309                                 arp = (*pskb)->nh.arph;
 310
 311                                 if (verdict == ARPT_CONTINUE)
 312                                         e = (void *)e + e->next_offset;
 313                                 else
 314                                         /* Verdict */
 315                                         break;
 316                         }
 317                 } else {
 318                         e = (void *)e + e->next_offset;
 319                 }
 320         } while (!hotdrop);
 321         read_unlock_bh(&table->lock);
 322
 323         if (hotdrop)
 324                 return NF_DROP;
 325         else
 326                 return verdict;
 327 }
 328
 329 /* All zeroes == unconditional rule. */
 330 static inline int unconditional(const struct arpt_arp *arp)
 331 {
 332         unsigned int i;
 333
 334         for (i = 0; i < sizeof(*arp)/sizeof(__u32); i++)
 335                 if (((__u32 *)arp)[i])
 336                         return 0;
 337
 338         return 1;
 339 }
 340
 341 /* Figures out from what hook each rule can be called: returns 0 if
 342  * there are loops.  Puts hook bitmask in comefrom.
 343  */
 344 static int mark_source_chains(struct xt_table_info *newinfo,
 345                               unsigned int valid_hooks, void *entry0)
 346 {
 347         unsigned int hook;
 348
 349         /* No recursion; use packet counter to save back ptrs (reset
 350          * to 0 as we leave), and comefrom to save source hook bitmask.
 351          */
 352         for (hook = 0; hook < NF_ARP_NUMHOOKS; hook++) {
 353                 unsigned int pos = newinfo->hook_entry[hook];
 354                 struct arpt_entry *e
 355                         = (struct arpt_entry *)(entry0 + pos);
 356
 357                 if (!(valid_hooks & (1 << hook)))
 358                         continue;
 359
 360                 /* Set initial back pointer. */
 361                 e->counters.pcnt = pos;
 362
 363                 for (;;) {
 364                         struct arpt_standard_target *t
 365                                 = (void *)arpt_get_target(e);
 366
 367                         if (e->comefrom & (1 << NF_ARP_NUMHOOKS)) {
 368                                 printk("arptables: loop hook %u pos %u %08X.\n",
 369                                        hook, pos, e->comefrom);
 370                                 return 0;
 371                         }
 372                         e->comefrom
 373                                 |= ((1 << hook) | (1 << NF_ARP_NUMHOOKS));
 374
 375                         /* Unconditional return/END. */
 376                         if (e->target_offset == sizeof(struct arpt_entry)
 377                             && (strcmp(t->target.u.user.name,
 378                                        ARPT_STANDARD_TARGET) == 0)
 379                             && t->verdict < 0
 380                             && unconditional(&e->arp)) {
 381                                 unsigned int oldpos, size;
 382
 383                                 if (t->verdict < -NF_MAX_VERDICT - 1) {
 384                                         duprintf("mark_source_chains: bad "
 385                                                 "negative verdict (%i)\n",
 386                                                                 t->verdict);
 387                                         return 0;
 388                                 }
 389
 390                                 /* Return: backtrack through the last
 391                                  * big jump.
 392                                  */
 393                                 do {
 394                                         e->comefrom ^= (1<<NF_ARP_NUMHOOKS);
 395                                         oldpos = pos;
 396                                         pos = e->counters.pcnt;
 397                                         e->counters.pcnt = 0;
 398
 399                                         /* We're at the start. */
 400                                         if (pos == oldpos)
 401                                                 goto next;
 402
 403                                         e = (struct arpt_entry *)
 404                                                 (entry0 + pos);
 405                                 } while (oldpos == pos + e->next_offset);
 406
 407                                 /* Move along one */
 408                                 size = e->next_offset;
 409                                 e = (struct arpt_entry *)
 410                                         (entry0 + pos + size);
 411                                 e->counters.pcnt = pos;
 412                                 pos += size;
 413                         } else {
 414                                 int newpos = t->verdict;
 415
 416                                 if (strcmp(t->target.u.user.name,
 417                                            ARPT_STANDARD_TARGET) == 0
 418                                     && newpos >= 0) {
 419                                         if (newpos > newinfo->size -
 420                                                 sizeof(struct arpt_entry)) {
 421                                                 duprintf("mark_source_chains: "
 422                                                         "bad verdict (%i)\n",
 423                                                                 newpos);
 424                                                 return 0;
 425                                         }
 426
 427                                         /* This a jump; chase it. */
 428                                         duprintf("Jump rule %u -> %u\n",
 429                                                  pos, newpos);
 430                                 } else {
 431                                         /* ... this is a fallthru */
 432                                         newpos = pos + e->next_offset;
 433                                 }
 434                                 e = (struct arpt_entry *)
 435                                         (entry0 + newpos);
 436                                 e->counters.pcnt = pos;
 437                                 pos = newpos;
 438                         }
 439                 }
 440                 next:
 441                 duprintf("Finished chain %u\n", hook);
 442         }
 443         return 1;
 444 }
 445
 446 static inline int standard_check(const struct arpt_entry_target *t,
 447                                  unsigned int max_offset)
 448 {
 449         /* Check standard info. */
 450         if (t->u.target_size
 451             != ARPT_ALIGN(sizeof(struct arpt_standard_target))) {
 452                 duprintf("arpt_standard_check: target size %u != %Zu\n",
 453                          t->u.target_size,
 454                          ARPT_ALIGN(sizeof(struct arpt_standard_target)));
 455                 return 0;
 456         }
 457
 458         return 1;
 459 }
 460
 461 static struct arpt_target arpt_standard_target;
 462
 463 static inline int check_entry(struct arpt_entry *e, const char *name, unsigned int size,
 464                               unsigned int *i)
 465 {
 466         struct arpt_entry_target *t;
 467         struct arpt_target *target;
 468         int ret;
 469
 470         if (!arp_checkentry(&e->arp)) {
 471                 duprintf("arp_tables: arp check failed %p %s.\n", e, name);
 472                 return -EINVAL;
 473         }
 474
 475         if (e->target_offset + sizeof(struct arpt_entry_target) > e->next_offset)
 476                 return -EINVAL;
 477
 478         t = arpt_get_target(e);
 479         if (e->target_offset + t->u.target_size > e->next_offset)
 480                 return -EINVAL;
 481
 482         target = try_then_request_module(xt_find_target(NF_ARP, t->u.user.name,
 483                                                         t->u.user.revision),
 484                                          "arpt_%s", t->u.user.name);
 485         if (IS_ERR(target) || !target) {
 486                 duprintf("check_entry: `%s' not found\n", t->u.user.name);
 487                 ret = target ? PTR_ERR(target) : -ENOENT;
 488                 goto out;
 489         }
 490         t->u.kernel.target = target;
 491
 492         ret = xt_check_target(target, NF_ARP, t->u.target_size - sizeof(*t),
 493                               name, e->comefrom, 0, 0);
 494         if (ret)
 495                 goto err;
 496
 497         if (t->u.kernel.target == &arpt_standard_target) {
 498                 if (!standard_check(t, size)) {
 499                         ret = -EINVAL;
 500                         goto out;
 501                 }
 502         } else if (t->u.kernel.target->checkentry
 503                    && !t->u.kernel.target->checkentry(name, e, target, t->data,
 504                                                       t->u.target_size
 505                                                       - sizeof(*t),
 506                                                       e->comefrom)) {
 507                 duprintf("arp_tables: check failed for `%s'.\n",
 508                          t->u.kernel.target->name);
 509                 ret = -EINVAL;
 510                 goto err;
 511         }
 512
 513         (*i)++;
 514         return 0;
 515 err:
 516         module_put(t->u.kernel.target->me);
 517 out:
 518         return ret;
 519 }
 520
 521 static inline int check_entry_size_and_hooks(struct arpt_entry *e,
 522                                              struct xt_table_info *newinfo,
 523                                              unsigned char *base,
 524                                              unsigned char *limit,
 525                                              const unsigned int *hook_entries,
 526                                              const unsigned int *underflows,
 527                                              unsigned int *i)
 528 {
 529         unsigned int h;
 530
 531         if ((unsigned long)e % __alignof__(struct arpt_entry) != 0
 532             || (unsigned char *)e + sizeof(struct arpt_entry) >= limit) {
 533                 duprintf("Bad offset %p\n", e);
 534                 return -EINVAL;
 535         }
 536
 537         if (e->next_offset
 538             < sizeof(struct arpt_entry) + sizeof(struct arpt_entry_target)) {
 539                 duprintf("checking: element %p size %u\n",
 540                          e, e->next_offset);
 541                 return -EINVAL;
 542         }
 543
 544         /* Check hooks & underflows */
 545         for (h = 0; h < NF_ARP_NUMHOOKS; h++) {
 546                 if ((unsigned char *)e - base == hook_entries[h])
 547                         newinfo->hook_entry[h] = hook_entries[h];
 548                 if ((unsigned char *)e - base == underflows[h])
 549                         newinfo->underflow[h] = underflows[h];
 550         }
 551
 552         /* FIXME: underflows must be unconditional, standard verdicts
 553            < 0 (not ARPT_RETURN). --RR */
 554
 555         /* Clear counters and comefrom */
 556         e->counters = ((struct xt_counters) { 0, 0 });
 557         e->comefrom = 0;
 558
 559         (*i)++;
 560         return 0;
 561 }
 562
 563 static inline int cleanup_entry(struct arpt_entry *e, unsigned int *i)
 564 {
 565         struct arpt_entry_target *t;
 566
 567         if (i && (*i)-- == 0)
 568                 return 1;
 569
 570         t = arpt_get_target(e);
 571         if (t->u.kernel.target->destroy)
 572                 t->u.kernel.target->destroy(t->u.kernel.target, t->data,
 573                                             t->u.target_size - sizeof(*t));
 574         module_put(t->u.kernel.target->me);
 575         return 0;
 576 }
 577
 578 /* Checks and translates the user-supplied table segment (held in
 579  * newinfo).
 580  */
 581 static int translate_table(const char *name,
 582                            unsigned int valid_hooks,
 583                            struct xt_table_info *newinfo,
 584                            void *entry0,
 585                            unsigned int size,
 586                            unsigned int number,
 587                            const unsigned int *hook_entries,
 588                            const unsigned int *underflows)
 589 {
 590         unsigned int i;
 591         int ret;
 592
 593         newinfo->size = size;
 594         newinfo->number = number;
 595
 596         /* Init all hooks to impossible value. */
 597         for (i = 0; i < NF_ARP_NUMHOOKS; i++) {
 598                 newinfo->hook_entry[i] = 0xFFFFFFFF;
 599                 newinfo->underflow[i] = 0xFFFFFFFF;
 600         }
 601
 602         duprintf("translate_table: size %u\n", newinfo->size);
 603         i = 0;
 604
 605         /* Walk through entries, checking offsets. */
 606         ret = ARPT_ENTRY_ITERATE(entry0, newinfo->size,
 607                                  check_entry_size_and_hooks,
 608                                  newinfo,
 609                                  entry0,
 610                                  entry0 + size,
 611                                  hook_entries, underflows, &i);
 612         duprintf("translate_table: ARPT_ENTRY_ITERATE gives %d\n", ret);
 613         if (ret != 0)
 614                 return ret;
 615
 616         if (i != number) {
 617                 duprintf("translate_table: %u not %u entries\n",
 618                          i, number);
 619                 return -EINVAL;
 620         }
 621
 622         /* Check hooks all assigned */
 623         for (i = 0; i < NF_ARP_NUMHOOKS; i++) {
 624                 /* Only hooks which are valid */
 625                 if (!(valid_hooks & (1 << i)))
 626                         continue;
 627                 if (newinfo->hook_entry[i] == 0xFFFFFFFF) {
 628                         duprintf("Invalid hook entry %u %u\n",
 629                                  i, hook_entries[i]);
 630                         return -EINVAL;
 631                 }
 632                 if (newinfo->underflow[i] == 0xFFFFFFFF) {
 633                         duprintf("Invalid underflow %u %u\n",
 634                                  i, underflows[i]);
 635                         return -EINVAL;
 636                 }
 637         }
 638
 639         if (!mark_source_chains(newinfo, valid_hooks, entry0)) {
 640                 duprintf("Looping hook\n");
 641                 return -ELOOP;
 642         }
 643
 644         /* Finally, each sanity check must pass */
 645         i = 0;
 646         ret = ARPT_ENTRY_ITERATE(entry0, newinfo->size,
 647                                  check_entry, name, size, &i);
 648
 649         if (ret != 0) {
 650                 ARPT_ENTRY_ITERATE(entry0, newinfo->size,
 651                                 cleanup_entry, &i);
 652                 return ret;
 653         }
 654
 655         /* And one copy for every other CPU */
 656         for_each_possible_cpu(i) {
 657                 if (newinfo->entries[i] && newinfo->entries[i] != entry0)
 658                         memcpy(newinfo->entries[i], entry0, newinfo->size);
 659         }
 660
 661         return ret;
 662 }
 663
 664 /* Gets counters. */
 665 static inline int add_entry_to_counter(const struct arpt_entry *e,
 666                                        struct xt_counters total[],
 667                                        unsigned int *i)
 668 {
 669         ADD_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
 670
 671         (*i)++;
 672         return 0;
 673 }
 674
 675 static inline int set_entry_to_counter(const struct arpt_entry *e,
 676                                        struct xt_counters total[],
 677                                        unsigned int *i)
 678 {
 679         SET_COUNTER(total[*i], e->counters.bcnt, e->counters.pcnt);
 680
 681         (*i)++;
 682         return 0;
 683 }
 684
 685 static void get_counters(const struct xt_table_info *t,
 686                          struct xt_counters counters[])
 687 {
 688         unsigned int cpu;
 689         unsigned int i;
 690         unsigned int curcpu;
 691
 692         /* Instead of clearing (by a previous call to memset())
 693          * the counters and using adds, we set the counters
 694          * with data used by 'current' CPU
 695          * We dont care about preemption here.
 696          */
 697         curcpu = raw_smp_processor_id();
 698
 699         i = 0;
 700         ARPT_ENTRY_ITERATE(t->entries[curcpu],
 701                            t->size,
 702                            set_entry_to_counter,
 703                            counters,
 704                            &i);
 705
 706         for_each_possible_cpu(cpu) {
 707                 if (cpu == curcpu)
 708                         continue;
 709                 i = 0;
 710                 ARPT_ENTRY_ITERATE(t->entries[cpu],
 711                                    t->size,
 712                                    add_entry_to_counter,
 713                                    counters,
 714                                    &i);
 715         }
 716 }
 717
 718 static int copy_entries_to_user(unsigned int total_size,
 719                                 struct arpt_table *table,
 720                                 void __user *userptr)
 721 {
 722         unsigned int off, num, countersize;
 723         struct arpt_entry *e;
 724         struct xt_counters *counters;
 725         struct xt_table_info *private = table->private;
 726         int ret = 0;
 727         void *loc_cpu_entry;
 728
 729         /* We need atomic snapshot of counters: rest doesn't change
 730          * (other than comefrom, which userspace doesn't care
 731          * about).
 732          */
 733         countersize = sizeof(struct xt_counters) * private->number;
 734         counters = vmalloc_node(countersize, numa_node_id());
 735
 736         if (counters == NULL)
 737                 return -ENOMEM;
 738
 739         /* First, sum counters... */
 740         write_lock_bh(&table->lock);
 741         get_counters(private, counters);
 742         write_unlock_bh(&table->lock);
 743
 744         loc_cpu_entry = private->entries[raw_smp_processor_id()];
 745         /* ... then copy entire thing ... */
 746         if (copy_to_user(userptr, loc_cpu_entry, total_size) != 0) {
 747                 ret = -EFAULT;
 748                 goto free_counters;
 749         }
 750
 751         /* FIXME: use iterator macros --RR */
 752         /* ... then go back and fix counters and names */
 753         for (off = 0, num = 0; off < total_size; off += e->next_offset, num++){
 754                 struct arpt_entry_target *t;
 755
 756                 e = (struct arpt_entry *)(loc_cpu_entry + off);
 757                 if (copy_to_user(userptr + off
 758                                  + offsetof(struct arpt_entry, counters),
 759                                  &counters[num],
 760                                  sizeof(counters[num])) != 0) {
 761                         ret = -EFAULT;
 762                         goto free_counters;
 763                 }
 764
 765                 t = arpt_get_target(e);
 766                 if (copy_to_user(userptr + off + e->target_offset
 767                                  + offsetof(struct arpt_entry_target,
 768                                             u.user.name),
 769                                  t->u.kernel.target->name,
 770                                  strlen(t->u.kernel.target->name)+1) != 0) {
 771                         ret = -EFAULT;
 772                         goto free_counters;
 773                 }
 774         }
 775
 776  free_counters:
 777         vfree(counters);
 778         return ret;
 779 }
 780
 781 static int get_entries(const struct arpt_get_entries *entries,
 782                        struct arpt_get_entries __user *uptr)
 783 {
 784         int ret;
 785         struct arpt_table *t;
 786
 787         t = xt_find_table_lock(NF_ARP, entries->name);
 788         if (t && !IS_ERR(t)) {
 789                 struct xt_table_info *private = t->private;
 790                 duprintf("t->private->number = %u\n",
 791                          private->number);
 792                 if (entries->size == private->size)
 793                         ret = copy_entries_to_user(private->size,
 794                                                    t, uptr->entrytable);
 795                 else {
 796                         duprintf("get_entries: I've got %u not %u!\n",
 797                                  private->size, entries->size);
 798                         ret = -EINVAL;
 799                 }
 800                 module_put(t->me);
 801                 xt_table_unlock(t);
 802         } else
 803                 ret = t ? PTR_ERR(t) : -ENOENT;
 804
 805         return ret;
 806 }
 807
 808 static int do_replace(void __user *user, unsigned int len)
 809 {
 810         int ret;
 811         struct arpt_replace tmp;
 812         struct arpt_table *t;
 813         struct xt_table_info *newinfo, *oldinfo;
 814         struct xt_counters *counters;
 815         void *loc_cpu_entry, *loc_cpu_old_entry;
 816
 817         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
 818                 return -EFAULT;
 819
 820         /* Hack: Causes ipchains to give correct error msg --RR */
 821         if (len != sizeof(tmp) + tmp.size)
 822                 return -ENOPROTOOPT;
 823
 824         /* overflow check */
 825         if (tmp.size >= (INT_MAX - sizeof(struct xt_table_info)) / NR_CPUS -
 826                         SMP_CACHE_BYTES)
 827                 return -ENOMEM;
 828         if (tmp.num_counters >= INT_MAX / sizeof(struct xt_counters))
 829                 return -ENOMEM;
 830
 831         newinfo = xt_alloc_table_info(tmp.size);
 832         if (!newinfo)
 833                 return -ENOMEM;
 834
 835         /* choose the copy that is on our node/cpu */
 836         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
 837         if (copy_from_user(loc_cpu_entry, user + sizeof(tmp),
 838                            tmp.size) != 0) {
 839                 ret = -EFAULT;
 840                 goto free_newinfo;
 841         }
 842
 843         counters = vmalloc(tmp.num_counters * sizeof(struct xt_counters));
 844         if (!counters) {
 845                 ret = -ENOMEM;
 846                 goto free_newinfo;
 847         }
 848
 849         ret = translate_table(tmp.name, tmp.valid_hooks,
 850                               newinfo, loc_cpu_entry, tmp.size, tmp.num_entries,
 851                               tmp.hook_entry, tmp.underflow);
 852         if (ret != 0)
 853                 goto free_newinfo_counters;
 854
 855         duprintf("arp_tables: Translated table\n");
 856
 857         t = try_then_request_module(xt_find_table_lock(NF_ARP, tmp.name),
 858                                     "arptable_%s", tmp.name);
 859         if (!t || IS_ERR(t)) {
 860                 ret = t ? PTR_ERR(t) : -ENOENT;
 861                 goto free_newinfo_counters_untrans;
 862         }
 863
 864         /* You lied! */
 865         if (tmp.valid_hooks != t->valid_hooks) {
 866                 duprintf("Valid hook crap: %08X vs %08X\n",
 867                          tmp.valid_hooks, t->valid_hooks);
 868                 ret = -EINVAL;
 869                 goto put_module;
 870         }
 871
 872         oldinfo = xt_replace_table(t, tmp.num_counters, newinfo, &ret);
 873         if (!oldinfo)
 874                 goto put_module;
 875
 876         /* Update module usage count based on number of rules */
 877         duprintf("do_replace: oldnum=%u, initnum=%u, newnum=%u\n",
 878                 oldinfo->number, oldinfo->initial_entries, newinfo->number);
 879         if ((oldinfo->number > oldinfo->initial_entries) ||
 880             (newinfo->number <= oldinfo->initial_entries))
 881                 module_put(t->me);
 882         if ((oldinfo->number > oldinfo->initial_entries) &&
 883             (newinfo->number <= oldinfo->initial_entries))
 884                 module_put(t->me);
 885
 886         /* Get the old counters. */
 887         get_counters(oldinfo, counters);
 888         /* Decrease module usage counts and free resource */
 889         loc_cpu_old_entry = oldinfo->entries[raw_smp_processor_id()];
 890         ARPT_ENTRY_ITERATE(loc_cpu_old_entry, oldinfo->size, cleanup_entry,NULL);
 891
 892         xt_free_table_info(oldinfo);
 893         if (copy_to_user(tmp.counters, counters,
 894                          sizeof(struct xt_counters) * tmp.num_counters) != 0)
 895                 ret = -EFAULT;
 896         vfree(counters);
 897         xt_table_unlock(t);
 898         return ret;
 899
 900  put_module:
 901         module_put(t->me);
 902         xt_table_unlock(t);
 903  free_newinfo_counters_untrans:
 904         ARPT_ENTRY_ITERATE(loc_cpu_entry, newinfo->size, cleanup_entry, NULL);
 905  free_newinfo_counters:
 906         vfree(counters);
 907  free_newinfo:
 908         xt_free_table_info(newinfo);
 909         return ret;
 910 }
 911
 912 /* We're lazy, and add to the first CPU; overflow works its fey magic
 913  * and everything is OK.
 914  */
 915 static inline int add_counter_to_entry(struct arpt_entry *e,
 916                                        const struct xt_counters addme[],
 917                                        unsigned int *i)
 918 {
 919
 920         ADD_COUNTER(e->counters, addme[*i].bcnt, addme[*i].pcnt);
 921
 922         (*i)++;
 923         return 0;
 924 }
 925
 926 static int do_add_counters(void __user *user, unsigned int len)
 927 {
 928         unsigned int i;
 929         struct xt_counters_info tmp, *paddc;
 930         struct arpt_table *t;
 931         struct xt_table_info *private;
 932         int ret = 0;
 933         void *loc_cpu_entry;
 934
 935         if (copy_from_user(&tmp, user, sizeof(tmp)) != 0)
 936                 return -EFAULT;
 937
 938         if (len != sizeof(tmp) + tmp.num_counters*sizeof(struct xt_counters))
 939                 return -EINVAL;
 940
 941         paddc = vmalloc(len);
 942         if (!paddc)
 943                 return -ENOMEM;
 944
 945         if (copy_from_user(paddc, user, len) != 0) {
 946                 ret = -EFAULT;
 947                 goto free;
 948         }
 949
 950         t = xt_find_table_lock(NF_ARP, tmp.name);
 951         if (!t || IS_ERR(t)) {
 952                 ret = t ? PTR_ERR(t) : -ENOENT;
 953                 goto free;
 954         }
 955
 956         write_lock_bh(&t->lock);
 957         private = t->private;
 958         if (private->number != tmp.num_counters) {
 959                 ret = -EINVAL;
 960                 goto unlock_up_free;
 961         }
 962
 963         i = 0;
 964         /* Choose the copy that is on our node */
 965         loc_cpu_entry = private->entries[smp_processor_id()];
 966         ARPT_ENTRY_ITERATE(loc_cpu_entry,
 967                            private->size,
 968                            add_counter_to_entry,
 969                            paddc->counters,
 970                            &i);
 971  unlock_up_free:
 972         write_unlock_bh(&t->lock);
 973         xt_table_unlock(t);
 974         module_put(t->me);
 975  free:
 976         vfree(paddc);
 977
 978         return ret;
 979 }
 980
 981 static int do_arpt_set_ctl(struct sock *sk, int cmd, void __user *user, unsigned int len)
 982 {
 983         int ret;
 984
 985         if (!capable(CAP_NET_ADMIN))
 986                 return -EPERM;
 987
 988         switch (cmd) {
 989         case ARPT_SO_SET_REPLACE:
 990                 ret = do_replace(user, len);
 991                 break;
 992
 993         case ARPT_SO_SET_ADD_COUNTERS:
 994                 ret = do_add_counters(user, len);
 995                 break;
 996
 997         default:
 998                 duprintf("do_arpt_set_ctl:  unknown request %i\n", cmd);
 999                 ret = -EINVAL;
1000         }
1001
1002         return ret;
1003 }
1004
1005 static int do_arpt_get_ctl(struct sock *sk, int cmd, void __user *user, int *len)
1006 {
1007         int ret;
1008
1009         if (!capable(CAP_NET_ADMIN))
1010                 return -EPERM;
1011
1012         switch (cmd) {
1013         case ARPT_SO_GET_INFO: {
1014                 char name[ARPT_TABLE_MAXNAMELEN];
1015                 struct arpt_table *t;
1016
1017                 if (*len != sizeof(struct arpt_getinfo)) {
1018                         duprintf("length %u != %Zu\n", *len,
1019                                  sizeof(struct arpt_getinfo));
1020                         ret = -EINVAL;
1021                         break;
1022                 }
1023
1024                 if (copy_from_user(name, user, sizeof(name)) != 0) {
1025                         ret = -EFAULT;
1026                         break;
1027                 }
1028                 name[ARPT_TABLE_MAXNAMELEN-1] = '\0';
1029
1030                 t = try_then_request_module(xt_find_table_lock(NF_ARP, name),
1031                                             "arptable_%s", name);
1032                 if (t && !IS_ERR(t)) {
1033                         struct arpt_getinfo info;
1034                         struct xt_table_info *private = t->private;
1035
1036                         info.valid_hooks = t->valid_hooks;
1037                         memcpy(info.hook_entry, private->hook_entry,
1038                                sizeof(info.hook_entry));
1039                         memcpy(info.underflow, private->underflow,
1040                                sizeof(info.underflow));
1041                         info.num_entries = private->number;
1042                         info.size = private->size;
1043                         strcpy(info.name, name);
1044
1045                         if (copy_to_user(user, &info, *len) != 0)
1046                                 ret = -EFAULT;
1047                         else
1048                                 ret = 0;
1049                         xt_table_unlock(t);
1050                         module_put(t->me);
1051                 } else
1052                         ret = t ? PTR_ERR(t) : -ENOENT;
1053         }
1054         break;
1055
1056         case ARPT_SO_GET_ENTRIES: {
1057                 struct arpt_get_entries get;
1058
1059                 if (*len < sizeof(get)) {
1060                         duprintf("get_entries: %u < %Zu\n", *len, sizeof(get));
1061                         ret = -EINVAL;
1062                 } else if (copy_from_user(&get, user, sizeof(get)) != 0) {
1063                         ret = -EFAULT;
1064                 } else if (*len != sizeof(struct arpt_get_entries) + get.size) {
1065                         duprintf("get_entries: %u != %Zu\n", *len,
1066                                  sizeof(struct arpt_get_entries) + get.size);
1067                         ret = -EINVAL;
1068                 } else
1069                         ret = get_entries(&get, user);
1070                 break;
1071         }
1072
1073         case ARPT_SO_GET_REVISION_TARGET: {
1074                 struct xt_get_revision rev;
1075
1076                 if (*len != sizeof(rev)) {
1077                         ret = -EINVAL;
1078                         break;
1079                 }
1080                 if (copy_from_user(&rev, user, sizeof(rev)) != 0) {
1081                         ret = -EFAULT;
1082                         break;
1083                 }
1084
1085                 try_then_request_module(xt_find_revision(NF_ARP, rev.name,
1086                                                          rev.revision, 1, &ret),
1087                                         "arpt_%s", rev.name);
1088                 break;
1089         }
1090
1091         default:
1092                 duprintf("do_arpt_get_ctl: unknown request %i\n", cmd);
1093                 ret = -EINVAL;
1094         }
1095
1096         return ret;
1097 }
1098
1099 int arpt_register_table(struct arpt_table *table,
1100                         const struct arpt_replace *repl)
1101 {
1102         int ret;
1103         struct xt_table_info *newinfo;
1104         static struct xt_table_info bootstrap
1105                 = { 0, 0, 0, { 0 }, { 0 }, { } };
1106         void *loc_cpu_entry;
1107
1108         newinfo = xt_alloc_table_info(repl->size);
1109         if (!newinfo) {
1110                 ret = -ENOMEM;
1111                 return ret;
1112         }
1113
1114         /* choose the copy on our node/cpu */
1115         loc_cpu_entry = newinfo->entries[raw_smp_processor_id()];
1116         memcpy(loc_cpu_entry, repl->entries, repl->size);
1117
1118         ret = translate_table(table->name, table->valid_hooks,
1119                               newinfo, loc_cpu_entry, repl->size,
1120                               repl->num_entries,
1121                               repl->hook_entry,
1122                               repl->underflow);
1123
1124         duprintf("arpt_register_table: translate table gives %d\n", ret);
1125         if (ret != 0) {
1126                 xt_free_table_info(newinfo);
1127                 return ret;
1128         }
1129
1130         ret = xt_register_table(table, &bootstrap, newinfo);
1131         if (ret != 0) {
1132                 xt_free_table_info(newinfo);
1133                 return ret;
1134         }
1135
1136         return 0;
1137 }
1138
1139 void arpt_unregister_table(struct arpt_table *table)
1140 {
1141         struct xt_table_info *private;
1142         void *loc_cpu_entry;
1143
1144         private = xt_unregister_table(table);
1145
1146         /* Decrease module usage counts and free resources */
1147         loc_cpu_entry = private->entries[raw_smp_processor_id()];
1148         ARPT_ENTRY_ITERATE(loc_cpu_entry, private->size,
1149                            cleanup_entry, NULL);
1150         xt_free_table_info(private);
1151 }
1152
1153 /* The built-in targets: standard (NULL) and error. */
1154 static struct arpt_target arpt_standard_target = {
1155         .name           = ARPT_STANDARD_TARGET,
1156         .targetsize     = sizeof(int),
1157         .family         = NF_ARP,
1158 };
1159
1160 static struct arpt_target arpt_error_target = {
1161         .name           = ARPT_ERROR_TARGET,
1162         .target         = arpt_error,
1163         .targetsize     = ARPT_FUNCTION_MAXNAMELEN,
1164         .family         = NF_ARP,
1165 };
1166
1167 static struct nf_sockopt_ops arpt_sockopts = {
1168         .pf             = PF_INET,
1169         .set_optmin     = ARPT_BASE_CTL,
1170         .set_optmax     = ARPT_SO_SET_MAX+1,
1171         .set            = do_arpt_set_ctl,
1172         .get_optmin     = ARPT_BASE_CTL,
1173         .get_optmax     = ARPT_SO_GET_MAX+1,
1174         .get            = do_arpt_get_ctl,
1175 };
1176
1177 static int __init arp_tables_init(void)
1178 {
1179         int ret;
1180
1181         ret = xt_proto_init(NF_ARP);
1182         if (ret < 0)
1183                 goto err1;
1184
1185         /* Noone else will be downing sem now, so we won't sleep */
1186         ret = xt_register_target(&arpt_standard_target);
1187         if (ret < 0)
1188                 goto err2;
1189         ret = xt_register_target(&arpt_error_target);
1190         if (ret < 0)
1191                 goto err3;
1192
1193         /* Register setsockopt */
1194         ret = nf_register_sockopt(&arpt_sockopts);
1195         if (ret < 0)
1196                 goto err4;
1197
1198         printk("arp_tables: (C) 2002 David S. Miller\n");
1199         return 0;
1200
1201 err4:
1202         xt_unregister_target(&arpt_error_target);
1203 err3:
1204         xt_unregister_target(&arpt_standard_target);
1205 err2:
1206         xt_proto_fini(NF_ARP);
1207 err1:
1208         return ret;
1209 }
1210
1211 static void __exit arp_tables_fini(void)
1212 {
1213         nf_unregister_sockopt(&arpt_sockopts);
1214         xt_unregister_target(&arpt_error_target);
1215         xt_unregister_target(&arpt_standard_target);
1216         xt_proto_fini(NF_ARP);
1217 }
1218
1219 EXPORT_SYMBOL(arpt_register_table);
1220 EXPORT_SYMBOL(arpt_unregister_table);
1221 EXPORT_SYMBOL(arpt_do_table);
1222
1223 module_init(arp_tables_init);
1224 module_exit(arp_tables_fini);