target/linux/generic-2.4/patches/608-netfilter_ipset.patch

   1 --- a/Documentation/Configure.help
   2 +++ b/Documentation/Configure.help
   3 @@ -3220,6 +3220,75 @@ CONFIG_IP_NF_TARGET_LOG
   4    If you want to compile it as a module, say M here and read
   5    <file:Documentation/modules.txt>.  If unsure, say `N'.
   6
   7 +IP set support
   8 +CONFIG_IP_NF_SET
   9 +  This option adds IP set support to the kernel.
  10 +
  11 +  In order to define and use sets, you need userlevel utilities: an
  12 +  iptables binary which knows about IP sets and the program ipset(8),
  13 +  by which you can define and setup the sets themselves.
  14 +
  15 +  If you want to compile it as a module, say M here and read
  16 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  17 +
  18 +set match support
  19 +CONFIG_IP_NF_MATCH_SET
  20 +  This option adds IP set match support.
  21 +  You need the ipset utility to create and set up the sets.
  22 +
  23 +  If you want to compile it as a module, say M here and read
  24 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  25 +
  26 +SET target support
  27 +CONFIG_IP_NF_TARGET_SET
  28 +  This option adds IP set target support.
  29 +  You need the ipset utility to create and set up the sets.
  30 +
  31 +  If you want to compile it as a module, say M here and read
  32 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  33 +
  34 +ipmap set type support
  35 +CONFIG_IP_NF_SET_IPMAP
  36 +  This option adds the ipmap set type support.
  37 +
  38 +  If you want to compile it as a module, say M here and read
  39 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  40 +
  41 +macipmap set type support
  42 +CONFIG_IP_NF_SET_MACIPMAP
  43 +  This option adds the macipmap set type support.
  44 +
  45 +  If you want to compile it as a module, say M here and read
  46 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  47 +
  48 +portmap set type support
  49 +CONFIG_IP_NF_SET_PORTMAP
  50 +  This option adds the portmap set type support.
  51 +
  52 +  If you want to compile it as a module, say M here and read
  53 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  54 +
  55 +iphash set type support
  56 +CONFIG_IP_NF_SET_IPHASH
  57 +  This option adds the iphash set type support.
  58 +
  59 +  If you want to compile it as a module, say M here and read
  60 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  61 +
  62 +nethash set type support
  63 +CONFIG_IP_NF_SET_NETHASH
  64 +  This option adds the nethash set type support.
  65 +
  66 +  If you want to compile it as a module, say M here and read
  67 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  68 +
  69 +iptree set type support
  70 +CONFIG_IP_NF_SET_IPTREE
  71 +  This option adds the iptree set type support.
  72 +
  73 +  If you want to compile it as a module, say M here and read
  74 +  <file:Documentation/modules.txt>.  If unsure, say `N'.
  75 +
  76  TTL target support
  77  CONFIG_IP_NF_TARGET_TTL
  78    This option adds a `TTL' target, which enables the user to set
  79 --- /dev/null
  80 +++ b/include/linux/netfilter_ipv4/ip_set.h
  81 @@ -0,0 +1,489 @@
  82 +#ifndef _IP_SET_H
  83 +#define _IP_SET_H
  84 +
  85 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
  86 + *                         Patrick Schaaf <bof@bof.de>
  87 + *                         Martin Josefsson <gandalf@wlug.westbo.se>
  88 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
  89 + *
  90 + * This program is free software; you can redistribute it and/or modify
  91 + * it under the terms of the GNU General Public License version 2 as
  92 + * published by the Free Software Foundation.
  93 + */
  94 +
  95 +/*
  96 + * A sockopt of such quality has hardly ever been seen before on the open
  97 + * market!  This little beauty, hardly ever used: above 64, so it's
  98 + * traditionally used for firewalling, not touched (even once!) by the
  99 + * 2.0, 2.2 and 2.4 kernels!
 100 + *
 101 + * Comes with its own certificate of authenticity, valid anywhere in the
 102 + * Free world!
 103 + *
 104 + * Rusty, 19.4.2000
 105 + */
 106 +#define SO_IP_SET              83
 107 +
 108 +/*
 109 + * Heavily modify by Joakim Axelsson 08.03.2002
 110 + * - Made it more modulebased
 111 + *
 112 + * Additional heavy modifications by Jozsef Kadlecsik 22.02.2004
 113 + * - bindings added
 114 + * - in order to "deal with" backward compatibility, renamed to ipset
 115 + */
 116 +
 117 +/*
 118 + * Used so that the kernel module and ipset-binary can match their versions
 119 + */
 120 +#define IP_SET_PROTOCOL_VERSION 2
 121 +
 122 +#define IP_SET_MAXNAMELEN 32   /* set names and set typenames */
 123 +
 124 +/* Lets work with our own typedef for representing an IP address.
 125 + * We hope to make the code more portable, possibly to IPv6...
 126 + *
 127 + * The representation works in HOST byte order, because most set types
 128 + * will perform arithmetic operations and compare operations.
 129 + *
 130 + * For now the type is an uint32_t.
 131 + *
 132 + * Make sure to ONLY use the functions when translating and parsing
 133 + * in order to keep the host byte order and make it more portable:
 134 + *  parse_ip()
 135 + *  parse_mask()
 136 + *  parse_ipandmask()
 137 + *  ip_tostring()
 138 + * (Joakim: where are they???)
 139 + */
 140 +
 141 +typedef uint32_t ip_set_ip_t;
 142 +
 143 +/* Sets are identified by an id in kernel space. Tweak with ip_set_id_t
 144 + * and IP_SET_INVALID_ID if you want to increase the max number of sets.
 145 + */
 146 +typedef uint16_t ip_set_id_t;
 147 +
 148 +#define IP_SET_INVALID_ID      65535
 149 +
 150 +/* How deep we follow bindings  */
 151 +#define IP_SET_MAX_BINDINGS    6
 152 +
 153 +/*
 154 + * Option flags for kernel operations (ipt_set_info)
 155 + */
 156 +#define IPSET_SRC              0x01    /* Source match/add */
 157 +#define IPSET_DST              0x02    /* Destination match/add */
 158 +#define IPSET_MATCH_INV                0x04    /* Inverse matching */
 159 +
 160 +/*
 161 + * Set types (flavours)
 162 + */
 163 +#define IPSET_TYPE_IP          0       /* IP address type of set */
 164 +#define IPSET_TYPE_PORT                1       /* Port type of set */
 165 +
 166 +/* Reserved keywords */
 167 +#define IPSET_TOKEN_DEFAULT    ":default:"
 168 +#define IPSET_TOKEN_ALL                ":all:"
 169 +
 170 +/* SO_IP_SET operation constants, and their request struct types.
 171 + *
 172 + * Operation ids:
 173 + *       0-99:  commands with version checking
 174 + *     100-199: add/del/test/bind/unbind
 175 + *     200-299: list, save, restore
 176 + */
 177 +
 178 +/* Single shot operations:
 179 + * version, create, destroy, flush, rename and swap
 180 + *
 181 + * Sets are identified by name.
 182 + */
 183 +
 184 +#define IP_SET_REQ_STD         \
 185 +       unsigned op;            \
 186 +       unsigned version;       \
 187 +       char name[IP_SET_MAXNAMELEN]
 188 +
 189 +#define IP_SET_OP_CREATE       0x00000001      /* Create a new (empty) set */
 190 +struct ip_set_req_create {
 191 +       IP_SET_REQ_STD;
 192 +       char typename[IP_SET_MAXNAMELEN];
 193 +};
 194 +
 195 +#define IP_SET_OP_DESTROY      0x00000002      /* Remove a (empty) set */
 196 +struct ip_set_req_std {
 197 +       IP_SET_REQ_STD;
 198 +};
 199 +
 200 +#define IP_SET_OP_FLUSH                0x00000003      /* Remove all IPs in a set */
 201 +/* Uses ip_set_req_std */
 202 +
 203 +#define IP_SET_OP_RENAME       0x00000004      /* Rename a set */
 204 +/* Uses ip_set_req_create */
 205 +
 206 +#define IP_SET_OP_SWAP         0x00000005      /* Swap two sets */
 207 +/* Uses ip_set_req_create */
 208 +
 209 +union ip_set_name_index {
 210 +       char name[IP_SET_MAXNAMELEN];
 211 +       ip_set_id_t index;
 212 +};
 213 +
 214 +#define IP_SET_OP_GET_BYNAME   0x00000006      /* Get set index by name */
 215 +struct ip_set_req_get_set {
 216 +       unsigned op;
 217 +       unsigned version;
 218 +       union ip_set_name_index set;
 219 +};
 220 +
 221 +#define IP_SET_OP_GET_BYINDEX  0x00000007      /* Get set name by index */
 222 +/* Uses ip_set_req_get_set */
 223 +
 224 +#define IP_SET_OP_VERSION      0x00000100      /* Ask kernel version */
 225 +struct ip_set_req_version {
 226 +       unsigned op;
 227 +       unsigned version;
 228 +};
 229 +
 230 +/* Double shots operations:
 231 + * add, del, test, bind and unbind.
 232 + *
 233 + * First we query the kernel to get the index and type of the target set,
 234 + * then issue the command. Validity of IP is checked in kernel in order
 235 + * to minimalize sockopt operations.
 236 + */
 237 +
 238 +/* Get minimal set data for add/del/test/bind/unbind IP */
 239 +#define IP_SET_OP_ADT_GET      0x00000010      /* Get set and type */
 240 +struct ip_set_req_adt_get {
 241 +       unsigned op;
 242 +       unsigned version;
 243 +       union ip_set_name_index set;
 244 +       char typename[IP_SET_MAXNAMELEN];
 245 +};
 246 +
 247 +#define IP_SET_REQ_BYINDEX     \
 248 +       unsigned op;            \
 249 +       ip_set_id_t index;
 250 +
 251 +struct ip_set_req_adt {
 252 +       IP_SET_REQ_BYINDEX;
 253 +};
 254 +
 255 +#define IP_SET_OP_ADD_IP       0x00000101      /* Add an IP to a set */
 256 +/* Uses ip_set_req_adt, with type specific addage */
 257 +
 258 +#define IP_SET_OP_DEL_IP       0x00000102      /* Remove an IP from a set */
 259 +/* Uses ip_set_req_adt, with type specific addage */
 260 +
 261 +#define IP_SET_OP_TEST_IP      0x00000103      /* Test an IP in a set */
 262 +/* Uses ip_set_req_adt, with type specific addage */
 263 +
 264 +#define IP_SET_OP_BIND_SET     0x00000104      /* Bind an IP to a set */
 265 +/* Uses ip_set_req_bind, with type specific addage */
 266 +struct ip_set_req_bind {
 267 +       IP_SET_REQ_BYINDEX;
 268 +       char binding[IP_SET_MAXNAMELEN];
 269 +};
 270 +
 271 +#define IP_SET_OP_UNBIND_SET   0x00000105      /* Unbind an IP from a set */
 272 +/* Uses ip_set_req_bind, with type speficic addage
 273 + * index = 0 means unbinding for all sets */
 274 +
 275 +#define IP_SET_OP_TEST_BIND_SET        0x00000106      /* Test binding an IP to a set */
 276 +/* Uses ip_set_req_bind, with type specific addage */
 277 +
 278 +/* Multiple shots operations: list, save, restore.
 279 + *
 280 + * - check kernel version and query the max number of sets
 281 + * - get the basic information on all sets
 282 + *   and size required for the next step
 283 + * - get actual set data: header, data, bindings
 284 + */
 285 +
 286 +/* Get max_sets and the index of a queried set
 287 + */
 288 +#define IP_SET_OP_MAX_SETS     0x00000020
 289 +struct ip_set_req_max_sets {
 290 +       unsigned op;
 291 +       unsigned version;
 292 +       ip_set_id_t max_sets;           /* max_sets */
 293 +       ip_set_id_t sets;               /* real number of sets */
 294 +       union ip_set_name_index set;    /* index of set if name used */
 295 +};
 296 +
 297 +/* Get the id and name of the sets plus size for next step */
 298 +#define IP_SET_OP_LIST_SIZE    0x00000201
 299 +#define IP_SET_OP_SAVE_SIZE    0x00000202
 300 +struct ip_set_req_setnames {
 301 +       unsigned op;
 302 +       ip_set_id_t index;              /* set to list/save */
 303 +       size_t size;                    /* size to get setdata/bindings */
 304 +       /* followed by sets number of struct ip_set_name_list */
 305 +};
 306 +
 307 +struct ip_set_name_list {
 308 +       char name[IP_SET_MAXNAMELEN];
 309 +       char typename[IP_SET_MAXNAMELEN];
 310 +       ip_set_id_t index;
 311 +       ip_set_id_t id;
 312 +};
 313 +
 314 +/* The actual list operation */
 315 +#define IP_SET_OP_LIST         0x00000203
 316 +struct ip_set_req_list {
 317 +       IP_SET_REQ_BYINDEX;
 318 +       /* sets number of struct ip_set_list in reply */
 319 +};
 320 +
 321 +struct ip_set_list {
 322 +       ip_set_id_t index;
 323 +       ip_set_id_t binding;
 324 +       u_int32_t ref;
 325 +       size_t header_size;     /* Set header data of header_size */
 326 +       size_t members_size;    /* Set members data of members_size */
 327 +       size_t bindings_size;   /* Set bindings data of bindings_size */
 328 +};
 329 +
 330 +struct ip_set_hash_list {
 331 +       ip_set_ip_t ip;
 332 +       ip_set_id_t binding;
 333 +};
 334 +
 335 +/* The save operation */
 336 +#define IP_SET_OP_SAVE         0x00000204
 337 +/* Uses ip_set_req_list, in the reply replaced by
 338 + * sets number of struct ip_set_save plus a marker
 339 + * ip_set_save followed by ip_set_hash_save structures.
 340 + */
 341 +struct ip_set_save {
 342 +       ip_set_id_t index;
 343 +       ip_set_id_t binding;
 344 +       size_t header_size;     /* Set header data of header_size */
 345 +       size_t members_size;    /* Set members data of members_size */
 346 +};
 347 +
 348 +/* At restoring, ip == 0 means default binding for the given set: */
 349 +struct ip_set_hash_save {
 350 +       ip_set_ip_t ip;
 351 +       ip_set_id_t id;
 352 +       ip_set_id_t binding;
 353 +};
 354 +
 355 +/* The restore operation */
 356 +#define IP_SET_OP_RESTORE      0x00000205
 357 +/* Uses ip_set_req_setnames followed by ip_set_restore structures
 358 + * plus a marker ip_set_restore, followed by ip_set_hash_save
 359 + * structures.
 360 + */
 361 +struct ip_set_restore {
 362 +       char name[IP_SET_MAXNAMELEN];
 363 +       char typename[IP_SET_MAXNAMELEN];
 364 +       ip_set_id_t index;
 365 +       size_t header_size;     /* Create data of header_size */
 366 +       size_t members_size;    /* Set members data of members_size */
 367 +};
 368 +
 369 +static inline int bitmap_bytes(ip_set_ip_t a, ip_set_ip_t b)
 370 +{
 371 +       return 4 * ((((b - a + 8) / 8) + 3) / 4);
 372 +}
 373 +
 374 +#ifdef __KERNEL__
 375 +
 376 +#define ip_set_printk(format, args...)                         \
 377 +       do {                                                    \
 378 +               printk("%s: %s: ", __FILE__, __FUNCTION__);     \
 379 +               printk(format "\n" , ## args);                  \
 380 +       } while (0)
 381 +
 382 +#if defined(IP_SET_DEBUG)
 383 +#define DP(format, args...)                                    \
 384 +       do {                                                    \
 385 +               printk("%s: %s (DBG): ", __FILE__, __FUNCTION__);\
 386 +               printk(format "\n" , ## args);                  \
 387 +       } while (0)
 388 +#define IP_SET_ASSERT(x)                                       \
 389 +       do {                                                    \
 390 +               if (!(x))                                       \
 391 +                       printk("IP_SET_ASSERT: %s:%i(%s)\n",    \
 392 +                               __FILE__, __LINE__, __FUNCTION__); \
 393 +       } while (0)
 394 +#else
 395 +#define DP(format, args...)
 396 +#define IP_SET_ASSERT(x)
 397 +#endif
 398 +
 399 +struct ip_set;
 400 +
 401 +/*
 402 + * The ip_set_type definition - one per set type, e.g. "ipmap".
 403 + *
 404 + * Each individual set has a pointer, set->type, going to one
 405 + * of these structures. Function pointers inside the structure implement
 406 + * the real behaviour of the sets.
 407 + *
 408 + * If not mentioned differently, the implementation behind the function
 409 + * pointers of a set_type, is expected to return 0 if ok, and a negative
 410 + * errno (e.g. -EINVAL) on error.
 411 + */
 412 +struct ip_set_type {
 413 +       struct list_head list;  /* next in list of set types */
 414 +
 415 +       /* test for IP in set (kernel: iptables -m set src|dst)
 416 +        * return 0 if not in set, 1 if in set.
 417 +        */
 418 +       int (*testip_kernel) (struct ip_set *set,
 419 +                             const struct sk_buff * skb,
 420 +                             u_int32_t flags,
 421 +                             ip_set_ip_t *ip);
 422 +
 423 +       /* test for IP in set (userspace: ipset -T set IP)
 424 +        * return 0 if not in set, 1 if in set.
 425 +        */
 426 +       int (*testip) (struct ip_set *set,
 427 +                      const void *data, size_t size,
 428 +                      ip_set_ip_t *ip);
 429 +
 430 +       /*
 431 +        * Size of the data structure passed by when
 432 +        * adding/deletin/testing an entry.
 433 +        */
 434 +       size_t reqsize;
 435 +
 436 +       /* Add IP into set (userspace: ipset -A set IP)
 437 +        * Return -EEXIST if the address is already in the set,
 438 +        * and -ERANGE if the address lies outside the set bounds.
 439 +        * If the address was not already in the set, 0 is returned.
 440 +        */
 441 +       int (*addip) (struct ip_set *set,
 442 +                     const void *data, size_t size,
 443 +                     ip_set_ip_t *ip);
 444 +
 445 +       /* Add IP into set (kernel: iptables ... -j SET set src|dst)
 446 +        * Return -EEXIST if the address is already in the set,
 447 +        * and -ERANGE if the address lies outside the set bounds.
 448 +        * If the address was not already in the set, 0 is returned.
 449 +        */
 450 +       int (*addip_kernel) (struct ip_set *set,
 451 +                            const struct sk_buff * skb,
 452 +                            u_int32_t flags,
 453 +                            ip_set_ip_t *ip);
 454 +
 455 +       /* remove IP from set (userspace: ipset -D set --entry x)
 456 +        * Return -EEXIST if the address is NOT in the set,
 457 +        * and -ERANGE if the address lies outside the set bounds.
 458 +        * If the address really was in the set, 0 is returned.
 459 +        */
 460 +       int (*delip) (struct ip_set *set,
 461 +                     const void *data, size_t size,
 462 +                     ip_set_ip_t *ip);
 463 +
 464 +       /* remove IP from set (kernel: iptables ... -j SET --entry x)
 465 +        * Return -EEXIST if the address is NOT in the set,
 466 +        * and -ERANGE if the address lies outside the set bounds.
 467 +        * If the address really was in the set, 0 is returned.
 468 +        */
 469 +       int (*delip_kernel) (struct ip_set *set,
 470 +                            const struct sk_buff * skb,
 471 +                            u_int32_t flags,
 472 +                            ip_set_ip_t *ip);
 473 +
 474 +       /* new set creation - allocated type specific items
 475 +        */
 476 +       int (*create) (struct ip_set *set,
 477 +                      const void *data, size_t size);
 478 +
 479 +       /* retry the operation after successfully tweaking the set
 480 +        */
 481 +       int (*retry) (struct ip_set *set);
 482 +
 483 +       /* set destruction - free type specific items
 484 +        * There is no return value.
 485 +        * Can be called only when child sets are destroyed.
 486 +        */
 487 +       void (*destroy) (struct ip_set *set);
 488 +
 489 +       /* set flushing - reset all bits in the set, or something similar.
 490 +        * There is no return value.
 491 +        */
 492 +       void (*flush) (struct ip_set *set);
 493 +
 494 +       /* Listing: size needed for header
 495 +        */
 496 +       size_t header_size;
 497 +
 498 +       /* Listing: Get the header
 499 +        *
 500 +        * Fill in the information in "data".
 501 +        * This function is always run after list_header_size() under a
 502 +        * writelock on the set. Therefor is the length of "data" always
 503 +        * correct.
 504 +        */
 505 +       void (*list_header) (const struct ip_set *set,
 506 +                            void *data);
 507 +
 508 +       /* Listing: Get the size for the set members
 509 +        */
 510 +       int (*list_members_size) (const struct ip_set *set);
 511 +
 512 +       /* Listing: Get the set members
 513 +        *
 514 +        * Fill in the information in "data".
 515 +        * This function is always run after list_member_size() under a
 516 +        * writelock on the set. Therefor is the length of "data" always
 517 +        * correct.
 518 +        */
 519 +       void (*list_members) (const struct ip_set *set,
 520 +                             void *data);
 521 +
 522 +       char typename[IP_SET_MAXNAMELEN];
 523 +       char typecode;
 524 +       int protocol_version;
 525 +
 526 +       /* Set this to THIS_MODULE if you are a module, otherwise NULL */
 527 +       struct module *me;
 528 +};
 529 +
 530 +extern int ip_set_register_set_type(struct ip_set_type *set_type);
 531 +extern void ip_set_unregister_set_type(struct ip_set_type *set_type);
 532 +
 533 +/* A generic ipset */
 534 +struct ip_set {
 535 +       char name[IP_SET_MAXNAMELEN];   /* the name of the set */
 536 +       rwlock_t lock;                  /* lock for concurrency control */
 537 +       ip_set_id_t id;                 /* set id for swapping */
 538 +       ip_set_id_t binding;            /* default binding for the set */
 539 +       atomic_t ref;                   /* in kernel and in hash references */
 540 +       struct ip_set_type *type;       /* the set types */
 541 +       void *data;                     /* pooltype specific data */
 542 +};
 543 +
 544 +/* Structure to bind set elements to sets */
 545 +struct ip_set_hash {
 546 +       struct list_head list;          /* list of clashing entries in hash */
 547 +       ip_set_ip_t ip;                 /* ip from set */
 548 +       ip_set_id_t id;                 /* set id */
 549 +       ip_set_id_t binding;            /* set we bind the element to */
 550 +};
 551 +
 552 +/* register and unregister set references */
 553 +extern ip_set_id_t ip_set_get_byname(const char name[IP_SET_MAXNAMELEN]);
 554 +extern ip_set_id_t ip_set_get_byindex(ip_set_id_t id);
 555 +extern void ip_set_put(ip_set_id_t id);
 556 +
 557 +/* API for iptables set match, and SET target */
 558 +extern void ip_set_addip_kernel(ip_set_id_t id,
 559 +                               const struct sk_buff *skb,
 560 +                               const u_int32_t *flags);
 561 +extern void ip_set_delip_kernel(ip_set_id_t id,
 562 +                               const struct sk_buff *skb,
 563 +                               const u_int32_t *flags);
 564 +extern int ip_set_testip_kernel(ip_set_id_t id,
 565 +                               const struct sk_buff *skb,
 566 +                               const u_int32_t *flags);
 567 +
 568 +#endif                         /* __KERNEL__ */
 569 +
 570 +#endif /*_IP_SET_H*/
 571 --- /dev/null
 572 +++ b/include/linux/netfilter_ipv4/ip_set_iphash.h
 573 @@ -0,0 +1,30 @@
 574 +#ifndef __IP_SET_IPHASH_H
 575 +#define __IP_SET_IPHASH_H
 576 +
 577 +#include <linux/netfilter_ipv4/ip_set.h>
 578 +
 579 +#define SETTYPE_NAME "iphash"
 580 +#define MAX_RANGE 0x0000FFFF
 581 +
 582 +struct ip_set_iphash {
 583 +       ip_set_ip_t *members;           /* the iphash proper */
 584 +       uint32_t initval;               /* initval for jhash_1word */
 585 +       uint32_t prime;                 /* prime for double hashing */
 586 +       uint32_t hashsize;              /* hash size */
 587 +       uint16_t probes;                /* max number of probes  */
 588 +       uint16_t resize;                /* resize factor in percent */
 589 +       ip_set_ip_t netmask;            /* netmask */
 590 +};
 591 +
 592 +struct ip_set_req_iphash_create {
 593 +       uint32_t hashsize;
 594 +       uint16_t probes;
 595 +       uint16_t resize;
 596 +       ip_set_ip_t netmask;
 597 +};
 598 +
 599 +struct ip_set_req_iphash {
 600 +       ip_set_ip_t ip;
 601 +};
 602 +
 603 +#endif /* __IP_SET_IPHASH_H */
 604 --- /dev/null
 605 +++ b/include/linux/netfilter_ipv4/ip_set_ipmap.h
 606 @@ -0,0 +1,56 @@
 607 +#ifndef __IP_SET_IPMAP_H
 608 +#define __IP_SET_IPMAP_H
 609 +
 610 +#include <linux/netfilter_ipv4/ip_set.h>
 611 +
 612 +#define SETTYPE_NAME "ipmap"
 613 +#define MAX_RANGE 0x0000FFFF
 614 +
 615 +struct ip_set_ipmap {
 616 +       void *members;                  /* the ipmap proper */
 617 +       ip_set_ip_t first_ip;           /* host byte order, included in range */
 618 +       ip_set_ip_t last_ip;            /* host byte order, included in range */
 619 +       ip_set_ip_t netmask;            /* subnet netmask */
 620 +       ip_set_ip_t sizeid;             /* size of set in IPs */
 621 +       u_int16_t hosts;                /* number of hosts in a subnet */
 622 +};
 623 +
 624 +struct ip_set_req_ipmap_create {
 625 +       ip_set_ip_t from;
 626 +       ip_set_ip_t to;
 627 +       ip_set_ip_t netmask;
 628 +};
 629 +
 630 +struct ip_set_req_ipmap {
 631 +       ip_set_ip_t ip;
 632 +};
 633 +
 634 +unsigned int
 635 +mask_to_bits(ip_set_ip_t mask)
 636 +{
 637 +       unsigned int bits = 32;
 638 +       ip_set_ip_t maskaddr;
 639 +
 640 +       if (mask == 0xFFFFFFFF)
 641 +               return bits;
 642 +
 643 +       maskaddr = 0xFFFFFFFE;
 644 +       while (--bits >= 0 && maskaddr != mask)
 645 +               maskaddr <<= 1;
 646 +
 647 +       return bits;
 648 +}
 649 +
 650 +ip_set_ip_t
 651 +range_to_mask(ip_set_ip_t from, ip_set_ip_t to, unsigned int *bits)
 652 +{
 653 +       ip_set_ip_t mask = 0xFFFFFFFE;
 654 +
 655 +       *bits = 32;
 656 +       while (--(*bits) >= 0 && mask && (to & mask) != from)
 657 +               mask <<= 1;
 658 +
 659 +       return mask;
 660 +}
 661 +
 662 +#endif /* __IP_SET_IPMAP_H */
 663 --- /dev/null
 664 +++ b/include/linux/netfilter_ipv4/ip_set_iptree.h
 665 @@ -0,0 +1,39 @@
 666 +#ifndef __IP_SET_IPTREE_H
 667 +#define __IP_SET_IPTREE_H
 668 +
 669 +#include <linux/netfilter_ipv4/ip_set.h>
 670 +
 671 +#define SETTYPE_NAME "iptree"
 672 +#define MAX_RANGE 0x0000FFFF
 673 +
 674 +struct ip_set_iptreed {
 675 +       unsigned long expires[255];             /* x.x.x.ADDR */
 676 +};
 677 +
 678 +struct ip_set_iptreec {
 679 +       struct ip_set_iptreed *tree[255];       /* x.x.ADDR.* */
 680 +};
 681 +
 682 +struct ip_set_iptreeb {
 683 +       struct ip_set_iptreec *tree[255];       /* x.ADDR.*.* */
 684 +};
 685 +
 686 +struct ip_set_iptree {
 687 +       unsigned int timeout;
 688 +       unsigned int gc_interval;
 689 +#ifdef __KERNEL__
 690 +       struct timer_list gc;
 691 +       struct ip_set_iptreeb *tree[255];       /* ADDR.*.*.* */
 692 +#endif
 693 +};
 694 +
 695 +struct ip_set_req_iptree_create {
 696 +       unsigned int timeout;
 697 +};
 698 +
 699 +struct ip_set_req_iptree {
 700 +       ip_set_ip_t ip;
 701 +       unsigned int timeout;
 702 +};
 703 +
 704 +#endif /* __IP_SET_IPTREE_H */
 705 --- /dev/null
 706 +++ b/include/linux/netfilter_ipv4/ip_set_jhash.h
 707 @@ -0,0 +1,148 @@
 708 +#ifndef _LINUX_IPSET_JHASH_H
 709 +#define _LINUX_IPSET_JHASH_H
 710 +
 711 +/* This is a copy of linux/jhash.h but the types u32/u8 are changed
 712 + * to __u32/__u8 so that the header file can be included into
 713 + * userspace code as well. Jozsef Kadlecsik (kadlec@blackhole.kfki.hu)
 714 + */
 715 +
 716 +/* jhash.h: Jenkins hash support.
 717 + *
 718 + * Copyright (C) 1996 Bob Jenkins (bob_jenkins@burtleburtle.net)
 719 + *
 720 + * http://burtleburtle.net/bob/hash/
 721 + *
 722 + * These are the credits from Bob's sources:
 723 + *
 724 + * lookup2.c, by Bob Jenkins, December 1996, Public Domain.
 725 + * hash(), hash2(), hash3, and mix() are externally useful functions.
 726 + * Routines to test the hash are included if SELF_TEST is defined.
 727 + * You can use this free for any purpose.  It has no warranty.
 728 + *
 729 + * Copyright (C) 2003 David S. Miller (davem@redhat.com)
 730 + *
 731 + * I've modified Bob's hash to be useful in the Linux kernel, and
 732 + * any bugs present are surely my fault.  -DaveM
 733 + */
 734 +
 735 +/* NOTE: Arguments are modified. */
 736 +#define __jhash_mix(a, b, c) \
 737 +{ \
 738 +  a -= b; a -= c; a ^= (c>>13); \
 739 +  b -= c; b -= a; b ^= (a<<8); \
 740 +  c -= a; c -= b; c ^= (b>>13); \
 741 +  a -= b; a -= c; a ^= (c>>12);  \
 742 +  b -= c; b -= a; b ^= (a<<16); \
 743 +  c -= a; c -= b; c ^= (b>>5); \
 744 +  a -= b; a -= c; a ^= (c>>3);  \
 745 +  b -= c; b -= a; b ^= (a<<10); \
 746 +  c -= a; c -= b; c ^= (b>>15); \
 747 +}
 748 +
 749 +/* The golden ration: an arbitrary value */
 750 +#define JHASH_GOLDEN_RATIO     0x9e3779b9
 751 +
 752 +/* The most generic version, hashes an arbitrary sequence
 753 + * of bytes.  No alignment or length assumptions are made about
 754 + * the input key.
 755 + */
 756 +static inline __u32 jhash(void *key, __u32 length, __u32 initval)
 757 +{
 758 +       __u32 a, b, c, len;
 759 +       __u8 *k = key;
 760 +
 761 +       len = length;
 762 +       a = b = JHASH_GOLDEN_RATIO;
 763 +       c = initval;
 764 +
 765 +       while (len >= 12) {
 766 +               a += (k[0] +((__u32)k[1]<<8) +((__u32)k[2]<<16) +((__u32)k[3]<<24));
 767 +               b += (k[4] +((__u32)k[5]<<8) +((__u32)k[6]<<16) +((__u32)k[7]<<24));
 768 +               c += (k[8] +((__u32)k[9]<<8) +((__u32)k[10]<<16)+((__u32)k[11]<<24));
 769 +
 770 +               __jhash_mix(a,b,c);
 771 +
 772 +               k += 12;
 773 +               len -= 12;
 774 +       }
 775 +
 776 +       c += length;
 777 +       switch (len) {
 778 +       case 11: c += ((__u32)k[10]<<24);
 779 +       case 10: c += ((__u32)k[9]<<16);
 780 +       case 9 : c += ((__u32)k[8]<<8);
 781 +       case 8 : b += ((__u32)k[7]<<24);
 782 +       case 7 : b += ((__u32)k[6]<<16);
 783 +       case 6 : b += ((__u32)k[5]<<8);
 784 +       case 5 : b += k[4];
 785 +       case 4 : a += ((__u32)k[3]<<24);
 786 +       case 3 : a += ((__u32)k[2]<<16);
 787 +       case 2 : a += ((__u32)k[1]<<8);
 788 +       case 1 : a += k[0];
 789 +       };
 790 +
 791 +       __jhash_mix(a,b,c);
 792 +
 793 +       return c;
 794 +}
 795 +
 796 +/* A special optimized version that handles 1 or more of __u32s.
 797 + * The length parameter here is the number of __u32s in the key.
 798 + */
 799 +static inline __u32 jhash2(__u32 *k, __u32 length, __u32 initval)
 800 +{
 801 +       __u32 a, b, c, len;
 802 +
 803 +       a = b = JHASH_GOLDEN_RATIO;
 804 +       c = initval;
 805 +       len = length;
 806 +
 807 +       while (len >= 3) {
 808 +               a += k[0];
 809 +               b += k[1];
 810 +               c += k[2];
 811 +               __jhash_mix(a, b, c);
 812 +               k += 3; len -= 3;
 813 +       }
 814 +
 815 +       c += length * 4;
 816 +
 817 +       switch (len) {
 818 +       case 2 : b += k[1];
 819 +       case 1 : a += k[0];
 820 +       };
 821 +
 822 +       __jhash_mix(a,b,c);
 823 +
 824 +       return c;
 825 +}
 826 +
 827 +
 828 +/* A special ultra-optimized versions that knows they are hashing exactly
 829 + * 3, 2 or 1 word(s).
 830 + *
 831 + * NOTE: In partilar the "c += length; __jhash_mix(a,b,c);" normally
 832 + *       done at the end is not done here.
 833 + */
 834 +static inline __u32 jhash_3words(__u32 a, __u32 b, __u32 c, __u32 initval)
 835 +{
 836 +       a += JHASH_GOLDEN_RATIO;
 837 +       b += JHASH_GOLDEN_RATIO;
 838 +       c += initval;
 839 +
 840 +       __jhash_mix(a, b, c);
 841 +
 842 +       return c;
 843 +}
 844 +
 845 +static inline __u32 jhash_2words(__u32 a, __u32 b, __u32 initval)
 846 +{
 847 +       return jhash_3words(a, b, 0, initval);
 848 +}
 849 +
 850 +static inline __u32 jhash_1word(__u32 a, __u32 initval)
 851 +{
 852 +       return jhash_3words(a, 0, 0, initval);
 853 +}
 854 +
 855 +#endif /* _LINUX_IPSET_JHASH_H */
 856 --- /dev/null
 857 +++ b/include/linux/netfilter_ipv4/ip_set_macipmap.h
 858 @@ -0,0 +1,38 @@
 859 +#ifndef __IP_SET_MACIPMAP_H
 860 +#define __IP_SET_MACIPMAP_H
 861 +
 862 +#include <linux/netfilter_ipv4/ip_set.h>
 863 +
 864 +#define SETTYPE_NAME "macipmap"
 865 +#define MAX_RANGE 0x0000FFFF
 866 +
 867 +/* general flags */
 868 +#define IPSET_MACIP_MATCHUNSET 1
 869 +
 870 +/* per ip flags */
 871 +#define IPSET_MACIP_ISSET      1
 872 +
 873 +struct ip_set_macipmap {
 874 +       void *members;                  /* the macipmap proper */
 875 +       ip_set_ip_t first_ip;           /* host byte order, included in range */
 876 +       ip_set_ip_t last_ip;            /* host byte order, included in range */
 877 +       u_int32_t flags;
 878 +};
 879 +
 880 +struct ip_set_req_macipmap_create {
 881 +       ip_set_ip_t from;
 882 +       ip_set_ip_t to;
 883 +       u_int32_t flags;
 884 +};
 885 +
 886 +struct ip_set_req_macipmap {
 887 +       ip_set_ip_t ip;
 888 +       unsigned char ethernet[ETH_ALEN];
 889 +};
 890 +
 891 +struct ip_set_macip {
 892 +       unsigned short flags;
 893 +       unsigned char ethernet[ETH_ALEN];
 894 +};
 895 +
 896 +#endif /* __IP_SET_MACIPMAP_H */
 897 --- /dev/null
 898 +++ b/include/linux/netfilter_ipv4/ip_set_malloc.h
 899 @@ -0,0 +1,27 @@
 900 +#ifndef _IP_SET_MALLOC_H
 901 +#define _IP_SET_MALLOC_H
 902 +
 903 +#ifdef __KERNEL__
 904 +
 905 +/* Memory allocation and deallocation */
 906 +static size_t max_malloc_size = 131072;                /* Guaranteed: slab.c */
 907 +
 908 +static inline void * ip_set_malloc(size_t bytes)
 909 +{
 910 +       if (bytes > max_malloc_size)
 911 +               return vmalloc(bytes);
 912 +       else
 913 +               return kmalloc(bytes, GFP_KERNEL);
 914 +}
 915 +
 916 +static inline void ip_set_free(void * data, size_t bytes)
 917 +{
 918 +       if (bytes > max_malloc_size)
 919 +               vfree(data);
 920 +       else
 921 +               kfree(data);
 922 +}
 923 +
 924 +#endif                         /* __KERNEL__ */
 925 +
 926 +#endif /*_IP_SET_MALLOC_H*/
 927 --- /dev/null
 928 +++ b/include/linux/netfilter_ipv4/ip_set_nethash.h
 929 @@ -0,0 +1,55 @@
 930 +#ifndef __IP_SET_NETHASH_H
 931 +#define __IP_SET_NETHASH_H
 932 +
 933 +#include <linux/netfilter_ipv4/ip_set.h>
 934 +
 935 +#define SETTYPE_NAME "nethash"
 936 +#define MAX_RANGE 0x0000FFFF
 937 +
 938 +struct ip_set_nethash {
 939 +       ip_set_ip_t *members;           /* the nethash proper */
 940 +       uint32_t initval;               /* initval for jhash_1word */
 941 +       uint32_t prime;                 /* prime for double hashing */
 942 +       uint32_t hashsize;              /* hash size */
 943 +       uint16_t probes;                /* max number of probes  */
 944 +       uint16_t resize;                /* resize factor in percent */
 945 +       unsigned char cidr[30];         /* CIDR sizes */
 946 +};
 947 +
 948 +struct ip_set_req_nethash_create {
 949 +       uint32_t hashsize;
 950 +       uint16_t probes;
 951 +       uint16_t resize;
 952 +};
 953 +
 954 +struct ip_set_req_nethash {
 955 +       ip_set_ip_t ip;
 956 +       unsigned char cidr;
 957 +};
 958 +
 959 +static unsigned char shifts[] = {255, 253, 249, 241, 225, 193, 129, 1};
 960 +
 961 +static inline ip_set_ip_t
 962 +pack(ip_set_ip_t ip, unsigned char cidr)
 963 +{
 964 +       ip_set_ip_t addr, *paddr = &addr;
 965 +       unsigned char n, t, *a;
 966 +
 967 +       addr = htonl(ip & (0xFFFFFFFF << (32 - (cidr))));
 968 +#ifdef __KERNEL__
 969 +       DP("ip:%u.%u.%u.%u/%u", NIPQUAD(addr), cidr);
 970 +#endif
 971 +       n = cidr / 8;
 972 +       t = cidr % 8;
 973 +       a = &((unsigned char *)paddr)[n];
 974 +       *a = *a /(1 << (8 - t)) + shifts[t];
 975 +#ifdef __KERNEL__
 976 +       DP("n: %u, t: %u, a: %u", n, t, *a);
 977 +       DP("ip:%u.%u.%u.%u/%u, %u.%u.%u.%u",
 978 +          HIPQUAD(ip), cidr, NIPQUAD(addr));
 979 +#endif
 980 +
 981 +       return ntohl(addr);
 982 +}
 983 +
 984 +#endif /* __IP_SET_NETHASH_H */
 985 --- /dev/null
 986 +++ b/include/linux/netfilter_ipv4/ip_set_portmap.h
 987 @@ -0,0 +1,25 @@
 988 +#ifndef __IP_SET_PORTMAP_H
 989 +#define __IP_SET_PORTMAP_H
 990 +
 991 +#include <linux/netfilter_ipv4/ip_set.h>
 992 +
 993 +#define SETTYPE_NAME   "portmap"
 994 +#define MAX_RANGE      0x0000FFFF
 995 +#define INVALID_PORT   (MAX_RANGE + 1)
 996 +
 997 +struct ip_set_portmap {
 998 +       void *members;                  /* the portmap proper */
 999 +       ip_set_ip_t first_port;         /* host byte order, included in range */
1000 +       ip_set_ip_t last_port;          /* host byte order, included in range */
1001 +};
1002 +
1003 +struct ip_set_req_portmap_create {
1004 +       ip_set_ip_t from;
1005 +       ip_set_ip_t to;
1006 +};
1007 +
1008 +struct ip_set_req_portmap {
1009 +       ip_set_ip_t port;
1010 +};
1011 +
1012 +#endif /* __IP_SET_PORTMAP_H */
1013 --- /dev/null
1014 +++ b/include/linux/netfilter_ipv4/ip_set_prime.h
1015 @@ -0,0 +1,34 @@
1016 +#ifndef __IP_SET_PRIME_H
1017 +#define __IP_SET_PRIME_H
1018 +
1019 +static inline unsigned make_prime_bound(unsigned nr)
1020 +{
1021 +       unsigned long long nr64 = nr;
1022 +       unsigned long long x = 1;
1023 +       nr = 1;
1024 +       while (x <= nr64) { x <<= 2; nr <<= 1; }
1025 +       return nr;
1026 +}
1027 +
1028 +static inline int make_prime_check(unsigned nr)
1029 +{
1030 +       unsigned x = 3;
1031 +       unsigned b = make_prime_bound(nr);
1032 +       while (x <= b) {
1033 +               if (0 == (nr % x)) return 0;
1034 +               x += 2;
1035 +       }
1036 +       return 1;
1037 +}
1038 +
1039 +static unsigned make_prime(unsigned nr)
1040 +{
1041 +       if (0 == (nr & 1)) nr--;
1042 +       while (nr > 1) {
1043 +               if (make_prime_check(nr)) return nr;
1044 +               nr -= 2;
1045 +       }
1046 +       return 2;
1047 +}
1048 +
1049 +#endif /* __IP_SET_PRIME_H */
1050 --- /dev/null
1051 +++ b/include/linux/netfilter_ipv4/ipt_set.h
1052 @@ -0,0 +1,21 @@
1053 +#ifndef _IPT_SET_H
1054 +#define _IPT_SET_H
1055 +
1056 +#include <linux/netfilter_ipv4/ip_set.h>
1057 +
1058 +struct ipt_set_info {
1059 +       ip_set_id_t index;
1060 +       u_int32_t flags[IP_SET_MAX_BINDINGS + 1];
1061 +};
1062 +
1063 +/* match info */
1064 +struct ipt_set_info_match {
1065 +       struct ipt_set_info match_set;
1066 +};
1067 +
1068 +struct ipt_set_info_target {
1069 +       struct ipt_set_info add_set;
1070 +       struct ipt_set_info del_set;
1071 +};
1072 +
1073 +#endif /*_IPT_SET_H*/
1074 --- a/net/ipv4/netfilter/Config.in
1075 +++ b/net/ipv4/netfilter/Config.in
1076 @@ -22,6 +22,20 @@ tristate 'IP tables support (required fo
1077  if [ "$CONFIG_IP_NF_IPTABLES" != "n" ]; then
1078  # The simple matches.
1079    dep_tristate '  limit match support' CONFIG_IP_NF_MATCH_LIMIT $CONFIG_IP_NF_IPTABLES
1080 +
1081 +  dep_tristate '  IP set support' CONFIG_IP_NF_SET $CONFIG_IP_NF_IPTABLES
1082 +  if [ "$CONFIG_IP_NF_SET" != "n" ]; then
1083 +    int '     Maximum number of sets' CONFIG_IP_NF_SET_MAX 256
1084 +    int '     Hash size for bindings of IP sets' CONFIG_IP_NF_SET_HASHSIZE 1024
1085 +    dep_tristate '    set match support' CONFIG_IP_NF_MATCH_SET $CONFIG_IP_NF_SET
1086 +    dep_tristate '    SET target support' CONFIG_IP_NF_TARGET_SET $CONFIG_IP_NF_SET
1087 +    dep_tristate '    ipmap set type support' CONFIG_IP_NF_SET_IPMAP $CONFIG_IP_NF_SET
1088 +    dep_tristate '    portmap set type support' CONFIG_IP_NF_SET_PORTMAP $CONFIG_IP_NF_SET
1089 +    dep_tristate '    macipmap set type support' CONFIG_IP_NF_SET_MACIPMAP $CONFIG_IP_NF_SET
1090 +    dep_tristate '    iphash set type support' CONFIG_IP_NF_SET_IPHASH $CONFIG_IP_NF_SET
1091 +    dep_tristate '    nethash set type support' CONFIG_IP_NF_SET_NETHASH $CONFIG_IP_NF_SET
1092 +    dep_tristate '    iptree set type support' CONFIG_IP_NF_SET_IPTREE $CONFIG_IP_NF_SET
1093 +  fi
1094    dep_tristate '  MAC address match support' CONFIG_IP_NF_MATCH_MAC $CONFIG_IP_NF_IPTABLES
1095    dep_tristate '  Packet type match support' CONFIG_IP_NF_MATCH_PKTTYPE $CONFIG_IP_NF_IPTABLES
1096    dep_tristate '  netfilter MARK match support' CONFIG_IP_NF_MATCH_MARK $CONFIG_IP_NF_IPTABLES
1097 --- /dev/null
1098 +++ b/net/ipv4/netfilter/ip_set.c
1099 @@ -0,0 +1,2002 @@
1100 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
1101 + *                         Patrick Schaaf <bof@bof.de>
1102 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
1103 + *
1104 + * This program is free software; you can redistribute it and/or modify
1105 + * it under the terms of the GNU General Public License version 2 as
1106 + * published by the Free Software Foundation.
1107 + */
1108 +
1109 +/* Kernel module for IP set management */
1110 +
1111 +#include <linux/config.h>
1112 +#include <linux/module.h>
1113 +#include <linux/kmod.h>
1114 +#include <linux/ip.h>
1115 +#include <linux/skbuff.h>
1116 +#include <linux/random.h>
1117 +#include <linux/jhash.h>
1118 +#include <linux/netfilter_ipv4/ip_tables.h>
1119 +#include <linux/errno.h>
1120 +#include <asm/uaccess.h>
1121 +#include <asm/bitops.h>
1122 +#include <asm/softirq.h>
1123 +#include <asm/semaphore.h>
1124 +#include <linux/spinlock.h>
1125 +#include <linux/vmalloc.h>
1126 +
1127 +#define ASSERT_READ_LOCK(x)    /* dont use that */
1128 +#define ASSERT_WRITE_LOCK(x)
1129 +#include <linux/netfilter_ipv4/listhelp.h>
1130 +#include <linux/netfilter_ipv4/ip_set.h>
1131 +
1132 +static struct list_head set_type_list;         /* all registered sets */
1133 +static struct ip_set **ip_set_list;            /* all individual sets */
1134 +static DECLARE_RWLOCK(ip_set_lock);            /* protects the lists and the hash */
1135 +static DECLARE_MUTEX(ip_set_app_mutex);                /* serializes user access */
1136 +static ip_set_id_t ip_set_max = CONFIG_IP_NF_SET_MAX;
1137 +static ip_set_id_t ip_set_bindings_hash_size =  CONFIG_IP_NF_SET_HASHSIZE;
1138 +static struct list_head *ip_set_hash;          /* hash of bindings */
1139 +static unsigned int ip_set_hash_random;                /* random seed */
1140 +
1141 +/* Arrgh */
1142 +#ifdef MODULE
1143 +#define __MOD_INC(foo)                 __MOD_INC_USE_COUNT(foo)
1144 +#define __MOD_DEC(foo)                 __MOD_DEC_USE_COUNT(foo)
1145 +#define __MOD_INC_SELF         MOD_INC_USE_COUNT
1146 +#define __MOD_DEC_SELF         MOD_DEC_USE_COUNT
1147 +#else
1148 +#define __MOD_INC(foo)
1149 +#define __MOD_DEC(foo)
1150 +#define __MOD_INC_SELF
1151 +#define __MOD_DEC_SELF
1152 +#endif
1153 +
1154 +/*
1155 + * Sets are identified either by the index in ip_set_list or by id.
1156 + * The id never changes and is used to find a key in the hash.
1157 + * The index may change by swapping and used at all other places
1158 + * (set/SET netfilter modules, binding value, etc.)
1159 + *
1160 + * Userspace requests are serialized by ip_set_mutex and sets can
1161 + * be deleted only from userspace. Therefore ip_set_list locking
1162 + * must obey the following rules:
1163 + *
1164 + * - kernel requests: read and write locking mandatory
1165 + * - user requests: read locking optional, write locking mandatory
1166 + */
1167 +
1168 +static inline void
1169 +__ip_set_get(ip_set_id_t index)
1170 +{
1171 +       atomic_inc(&ip_set_list[index]->ref);
1172 +}
1173 +
1174 +static inline void
1175 +__ip_set_put(ip_set_id_t index)
1176 +{
1177 +       atomic_dec(&ip_set_list[index]->ref);
1178 +}
1179 +
1180 +/*
1181 + * Binding routines
1182 + */
1183 +
1184 +static inline int
1185 +ip_hash_cmp(const struct ip_set_hash *set_hash,
1186 +           ip_set_id_t id, ip_set_ip_t ip)
1187 +{
1188 +       return set_hash->id == id && set_hash->ip == ip;
1189 +}
1190 +
1191 +static ip_set_id_t
1192 +ip_set_find_in_hash(ip_set_id_t id, ip_set_ip_t ip)
1193 +{
1194 +       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random)
1195 +                               % ip_set_bindings_hash_size;
1196 +       struct ip_set_hash *set_hash;
1197 +
1198 +       MUST_BE_READ_LOCKED(&ip_set_lock);
1199 +       IP_SET_ASSERT(ip_set_list[id]);
1200 +       DP("set: %s, ip: %u.%u.%u.%u", ip_set_list[id]->name, HIPQUAD(ip));
1201 +
1202 +       set_hash = LIST_FIND(&ip_set_hash[key], ip_hash_cmp,
1203 +                            struct ip_set_hash *, id, ip);
1204 +
1205 +       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name,
1206 +          HIPQUAD(ip),
1207 +          set_hash != NULL ? ip_set_list[set_hash->binding]->name : "");
1208 +
1209 +       return (set_hash != NULL ? set_hash->binding : IP_SET_INVALID_ID);
1210 +}
1211 +
1212 +static inline void
1213 +__set_hash_del(struct ip_set_hash *set_hash)
1214 +{
1215 +       MUST_BE_WRITE_LOCKED(&ip_set_lock);
1216 +       IP_SET_ASSERT(ip_set_list[set_hash->binding]);
1217 +
1218 +       __ip_set_put(set_hash->binding);
1219 +       list_del(&set_hash->list);
1220 +       kfree(set_hash);
1221 +}
1222 +
1223 +static int
1224 +ip_set_hash_del(ip_set_id_t id, ip_set_ip_t ip)
1225 +{
1226 +       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random)
1227 +                               % ip_set_bindings_hash_size;
1228 +       struct ip_set_hash *set_hash;
1229 +
1230 +       IP_SET_ASSERT(ip_set_list[id]);
1231 +       DP("set: %s, ip: %u.%u.%u.%u", ip_set_list[id]->name, HIPQUAD(ip));
1232 +       WRITE_LOCK(&ip_set_lock);
1233 +       set_hash = LIST_FIND(&ip_set_hash[key], ip_hash_cmp,
1234 +                            struct ip_set_hash *, id, ip);
1235 +       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name,
1236 +          HIPQUAD(ip),
1237 +          set_hash != NULL ? ip_set_list[set_hash->binding]->name : "");
1238 +
1239 +       if (set_hash != NULL)
1240 +               __set_hash_del(set_hash);
1241 +       WRITE_UNLOCK(&ip_set_lock);
1242 +       return 0;
1243 +}
1244 +
1245 +static int
1246 +ip_set_hash_add(ip_set_id_t id, ip_set_ip_t ip, ip_set_id_t binding)
1247 +{
1248 +       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random)
1249 +                               % ip_set_bindings_hash_size;
1250 +       struct ip_set_hash *set_hash;
1251 +       int ret = 0;
1252 +
1253 +       IP_SET_ASSERT(ip_set_list[id]);
1254 +       IP_SET_ASSERT(ip_set_list[binding]);
1255 +       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name,
1256 +          HIPQUAD(ip), ip_set_list[binding]->name);
1257 +       WRITE_LOCK(&ip_set_lock);
1258 +       set_hash = LIST_FIND(&ip_set_hash[key], ip_hash_cmp,
1259 +                            struct ip_set_hash *, id, ip);
1260 +       if (!set_hash) {
1261 +               set_hash = kmalloc(sizeof(struct ip_set_hash), GFP_KERNEL);
1262 +               if (!set_hash) {
1263 +                       ret = -ENOMEM;
1264 +                       goto unlock;
1265 +               }
1266 +               INIT_LIST_HEAD(&set_hash->list);
1267 +               set_hash->id = id;
1268 +               set_hash->ip = ip;
1269 +               list_add(&ip_set_hash[key], &set_hash->list);
1270 +       } else {
1271 +               IP_SET_ASSERT(ip_set_list[set_hash->binding]);
1272 +               DP("overwrite binding: %s",
1273 +                  ip_set_list[set_hash->binding]->name);
1274 +               __ip_set_put(set_hash->binding);
1275 +       }
1276 +       set_hash->binding = binding;
1277 +       __ip_set_get(set_hash->binding);
1278 +    unlock:
1279 +       WRITE_UNLOCK(&ip_set_lock);
1280 +       return ret;
1281 +}
1282 +
1283 +#define FOREACH_HASH_DO(fn, args...)                                           \
1284 +({                                                                             \
1285 +       ip_set_id_t __key;                                                      \
1286 +       struct ip_set_hash *__set_hash;                                         \
1287 +                                                                               \
1288 +       for (__key = 0; __key < ip_set_bindings_hash_size; __key++) {           \
1289 +               list_for_each_entry(__set_hash, &ip_set_hash[__key], list)      \
1290 +                       fn(__set_hash , ## args);                               \
1291 +       }                                                                       \
1292 +})
1293 +
1294 +#define FOREACH_HASH_RW_DO(fn, args...)                                                \
1295 +({                                                                             \
1296 +       ip_set_id_t __key;                                                      \
1297 +       struct ip_set_hash *__set_hash, *__n;                                   \
1298 +                                                                               \
1299 +       MUST_BE_WRITE_LOCKED(&ip_set_lock);                                     \
1300 +       for (__key = 0; __key < ip_set_bindings_hash_size; __key++) {           \
1301 +               list_for_each_entry_safe(__set_hash, __n, &ip_set_hash[__key], list)\
1302 +                       fn(__set_hash , ## args);                               \
1303 +       }                                                                       \
1304 +})
1305 +
1306 +/* Add, del and test set entries from kernel */
1307 +
1308 +#define follow_bindings(index, set, ip)                                        \
1309 +((index = ip_set_find_in_hash((set)->id, ip)) != IP_SET_INVALID_ID     \
1310 + || (index = (set)->binding) != IP_SET_INVALID_ID)
1311 +
1312 +int
1313 +ip_set_testip_kernel(ip_set_id_t index,
1314 +                    const struct sk_buff *skb,
1315 +                    const u_int32_t *flags)
1316 +{
1317 +       struct ip_set *set;
1318 +       ip_set_ip_t ip;
1319 +       int res, i = 0;
1320 +
1321 +       IP_SET_ASSERT(flags[i]);
1322 +       READ_LOCK(&ip_set_lock);
1323 +       do {
1324 +               set = ip_set_list[index];
1325 +               IP_SET_ASSERT(set);
1326 +               DP("set %s, index %u", set->name, index);
1327 +               read_lock_bh(&set->lock);
1328 +               res = set->type->testip_kernel(set, skb, flags[i], &ip);
1329 +               read_unlock_bh(&set->lock);
1330 +       } while (res > 0
1331 +                && flags[++i]
1332 +                && follow_bindings(index, set, ip));
1333 +       READ_UNLOCK(&ip_set_lock);
1334 +
1335 +       return res;
1336 +}
1337 +
1338 +void
1339 +ip_set_addip_kernel(ip_set_id_t index,
1340 +                   const struct sk_buff *skb,
1341 +                   const u_int32_t *flags)
1342 +{
1343 +       struct ip_set *set;
1344 +       ip_set_ip_t ip;
1345 +       int res, i= 0;
1346 +
1347 +       IP_SET_ASSERT(flags[i]);
1348 +    retry:
1349 +       READ_LOCK(&ip_set_lock);
1350 +       do {
1351 +               set = ip_set_list[index];
1352 +               IP_SET_ASSERT(set);
1353 +               DP("set %s, index %u", set->name, index);
1354 +               write_lock_bh(&set->lock);
1355 +               res = set->type->addip_kernel(set, skb, flags[i], &ip);
1356 +               write_unlock_bh(&set->lock);
1357 +       } while ((res == 0 || res == -EEXIST)
1358 +                && flags[++i]
1359 +                && follow_bindings(index, set, ip));
1360 +       READ_UNLOCK(&ip_set_lock);
1361 +
1362 +       if (res == -EAGAIN
1363 +           && set->type->retry
1364 +           && (res = set->type->retry(set)) == 0)
1365 +               goto retry;
1366 +}
1367 +
1368 +void
1369 +ip_set_delip_kernel(ip_set_id_t index,
1370 +                   const struct sk_buff *skb,
1371 +                   const u_int32_t *flags)
1372 +{
1373 +       struct ip_set *set;
1374 +       ip_set_ip_t ip;
1375 +       int res, i = 0;
1376 +
1377 +       IP_SET_ASSERT(flags[i]);
1378 +       READ_LOCK(&ip_set_lock);
1379 +       do {
1380 +               set = ip_set_list[index];
1381 +               IP_SET_ASSERT(set);
1382 +               DP("set %s, index %u", set->name, index);
1383 +               write_lock_bh(&set->lock);
1384 +               res = set->type->delip_kernel(set, skb, flags[i], &ip);
1385 +               write_unlock_bh(&set->lock);
1386 +       } while ((res == 0 || res == -EEXIST)
1387 +                && flags[++i]
1388 +                && follow_bindings(index, set, ip));
1389 +       READ_UNLOCK(&ip_set_lock);
1390 +}
1391 +
1392 +/* Register and deregister settype */
1393 +
1394 +static inline int
1395 +set_type_equal(const struct ip_set_type *set_type, const char *str2)
1396 +{
1397 +       return !strncmp(set_type->typename, str2, IP_SET_MAXNAMELEN - 1);
1398 +}
1399 +
1400 +static inline struct ip_set_type *
1401 +find_set_type(const char *name)
1402 +{
1403 +       return LIST_FIND(&set_type_list,
1404 +                        set_type_equal,
1405 +                        struct ip_set_type *,
1406 +                        name);
1407 +}
1408 +
1409 +int
1410 +ip_set_register_set_type(struct ip_set_type *set_type)
1411 +{
1412 +       int ret = 0;
1413 +
1414 +       if (set_type->protocol_version != IP_SET_PROTOCOL_VERSION) {
1415 +               ip_set_printk("'%s' uses wrong protocol version %u (want %u)",
1416 +                             set_type->typename,
1417 +                             set_type->protocol_version,
1418 +                             IP_SET_PROTOCOL_VERSION);
1419 +               return -EINVAL;
1420 +       }
1421 +
1422 +       WRITE_LOCK(&ip_set_lock);
1423 +       if (find_set_type(set_type->typename)) {
1424 +               /* Duplicate! */
1425 +               ip_set_printk("'%s' already registered!",
1426 +                             set_type->typename);
1427 +               ret = -EINVAL;
1428 +               goto unlock;
1429 +       }
1430 +       __MOD_INC_SELF;
1431 +       list_append(&set_type_list, set_type);
1432 +       DP("'%s' registered.", set_type->typename);
1433 +   unlock:
1434 +       WRITE_UNLOCK(&ip_set_lock);
1435 +       return ret;
1436 +}
1437 +
1438 +void
1439 +ip_set_unregister_set_type(struct ip_set_type *set_type)
1440 +{
1441 +       WRITE_LOCK(&ip_set_lock);
1442 +       if (!find_set_type(set_type->typename)) {
1443 +               ip_set_printk("'%s' not registered?",
1444 +                             set_type->typename);
1445 +               goto unlock;
1446 +       }
1447 +       LIST_DELETE(&set_type_list, set_type);
1448 +       __MOD_DEC_SELF;
1449 +       DP("'%s' unregistered.", set_type->typename);
1450 +   unlock:
1451 +       WRITE_UNLOCK(&ip_set_lock);
1452 +
1453 +}
1454 +
1455 +/*
1456 + * Userspace routines
1457 + */
1458 +
1459 +/*
1460 + * Find set by name, reference it once. The reference makes sure the
1461 + * thing pointed to, does not go away under our feet. Drop the reference
1462 + * later, using ip_set_put().
1463 + */
1464 +ip_set_id_t
1465 +ip_set_get_byname(const char *name)
1466 +{
1467 +       ip_set_id_t i, index = IP_SET_INVALID_ID;
1468 +
1469 +       down(&ip_set_app_mutex);
1470 +       for (i = 0; i < ip_set_max; i++) {
1471 +               if (ip_set_list[i] != NULL
1472 +                   && strcmp(ip_set_list[i]->name, name) == 0) {
1473 +                       __ip_set_get(i);
1474 +                       index = i;
1475 +                       break;
1476 +               }
1477 +       }
1478 +       up(&ip_set_app_mutex);
1479 +       return index;
1480 +}
1481 +
1482 +/*
1483 + * Find set by index, reference it once. The reference makes sure the
1484 + * thing pointed to, does not go away under our feet. Drop the reference
1485 + * later, using ip_set_put().
1486 + */
1487 +ip_set_id_t
1488 +ip_set_get_byindex(ip_set_id_t index)
1489 +{
1490 +       down(&ip_set_app_mutex);
1491 +
1492 +       if (index >= ip_set_max)
1493 +               return IP_SET_INVALID_ID;
1494 +
1495 +       if (ip_set_list[index])
1496 +               __ip_set_get(index);
1497 +       else
1498 +               index = IP_SET_INVALID_ID;
1499 +
1500 +       up(&ip_set_app_mutex);
1501 +       return index;
1502 +}
1503 +
1504 +/*
1505 + * If the given set pointer points to a valid set, decrement
1506 + * reference count by 1. The caller shall not assume the index
1507 + * to be valid, after calling this function.
1508 + */
1509 +void ip_set_put(ip_set_id_t index)
1510 +{
1511 +       down(&ip_set_app_mutex);
1512 +       if (ip_set_list[index])
1513 +               __ip_set_put(index);
1514 +       up(&ip_set_app_mutex);
1515 +}
1516 +
1517 +/* Find a set by name or index */
1518 +static ip_set_id_t
1519 +ip_set_find_byname(const char *name)
1520 +{
1521 +       ip_set_id_t i, index = IP_SET_INVALID_ID;
1522 +
1523 +       for (i = 0; i < ip_set_max; i++) {
1524 +               if (ip_set_list[i] != NULL
1525 +                   && strcmp(ip_set_list[i]->name, name) == 0) {
1526 +                       index = i;
1527 +                       break;
1528 +               }
1529 +       }
1530 +       return index;
1531 +}
1532 +
1533 +static ip_set_id_t
1534 +ip_set_find_byindex(ip_set_id_t index)
1535 +{
1536 +       if (index >= ip_set_max || ip_set_list[index] == NULL)
1537 +               index = IP_SET_INVALID_ID;
1538 +
1539 +       return index;
1540 +}
1541 +
1542 +/*
1543 + * Add, del, test, bind and unbind
1544 + */
1545 +
1546 +static inline int
1547 +__ip_set_testip(struct ip_set *set,
1548 +               const void *data,
1549 +               size_t size,
1550 +               ip_set_ip_t *ip)
1551 +{
1552 +       int res;
1553 +
1554 +       read_lock_bh(&set->lock);
1555 +       res = set->type->testip(set, data, size, ip);
1556 +       read_unlock_bh(&set->lock);
1557 +
1558 +       return res;
1559 +}
1560 +
1561 +static int
1562 +__ip_set_addip(ip_set_id_t index,
1563 +              const void *data,
1564 +              size_t size)
1565 +{
1566 +       struct ip_set *set = ip_set_list[index];
1567 +       ip_set_ip_t ip;
1568 +       int res;
1569 +
1570 +       IP_SET_ASSERT(set);
1571 +       do {
1572 +               write_lock_bh(&set->lock);
1573 +               res = set->type->addip(set, data, size, &ip);
1574 +               write_unlock_bh(&set->lock);
1575 +       } while (res == -EAGAIN
1576 +                && set->type->retry
1577 +                && (res = set->type->retry(set)) == 0);
1578 +
1579 +       return res;
1580 +}
1581 +
1582 +static int
1583 +ip_set_addip(ip_set_id_t index,
1584 +            const void *data,
1585 +            size_t size)
1586 +{
1587 +
1588 +       return __ip_set_addip(index,
1589 +                             data + sizeof(struct ip_set_req_adt),
1590 +                             size - sizeof(struct ip_set_req_adt));
1591 +}
1592 +
1593 +static int
1594 +ip_set_delip(ip_set_id_t index,
1595 +            const void *data,
1596 +            size_t size)
1597 +{
1598 +       struct ip_set *set = ip_set_list[index];
1599 +       ip_set_ip_t ip;
1600 +       int res;
1601 +
1602 +       IP_SET_ASSERT(set);
1603 +       write_lock_bh(&set->lock);
1604 +       res = set->type->delip(set,
1605 +                              data + sizeof(struct ip_set_req_adt),
1606 +                              size - sizeof(struct ip_set_req_adt),
1607 +                              &ip);
1608 +       write_unlock_bh(&set->lock);
1609 +
1610 +       return res;
1611 +}
1612 +
1613 +static int
1614 +ip_set_testip(ip_set_id_t index,
1615 +             const void *data,
1616 +             size_t size)
1617 +{
1618 +       struct ip_set *set = ip_set_list[index];
1619 +       ip_set_ip_t ip;
1620 +       int res;
1621 +
1622 +       IP_SET_ASSERT(set);
1623 +       res = __ip_set_testip(set,
1624 +                             data + sizeof(struct ip_set_req_adt),
1625 +                             size - sizeof(struct ip_set_req_adt),
1626 +                             &ip);
1627 +
1628 +       return (res > 0 ? -EEXIST : res);
1629 +}
1630 +
1631 +static int
1632 +ip_set_bindip(ip_set_id_t index,
1633 +             const void *data,
1634 +             size_t size)
1635 +{
1636 +       struct ip_set *set = ip_set_list[index];
1637 +       struct ip_set_req_bind *req_bind;
1638 +       ip_set_id_t binding;
1639 +       ip_set_ip_t ip;
1640 +       int res;
1641 +
1642 +       IP_SET_ASSERT(set);
1643 +       if (size < sizeof(struct ip_set_req_bind))
1644 +               return -EINVAL;
1645 +
1646 +       req_bind = (struct ip_set_req_bind *) data;
1647 +       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
1648 +
1649 +       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
1650 +               /* Default binding of a set */
1651 +               char *binding_name;
1652 +
1653 +               if (size != sizeof(struct ip_set_req_bind) + IP_SET_MAXNAMELEN)
1654 +                       return -EINVAL;
1655 +
1656 +               binding_name = (char *)(data + sizeof(struct ip_set_req_bind));
1657 +               binding_name[IP_SET_MAXNAMELEN - 1] = '\0';
1658 +
1659 +               binding = ip_set_find_byname(binding_name);
1660 +               if (binding == IP_SET_INVALID_ID)
1661 +                       return -ENOENT;
1662 +
1663 +               WRITE_LOCK(&ip_set_lock);
1664 +               /* Sets as binding values are referenced */
1665 +               if (set->binding != IP_SET_INVALID_ID)
1666 +                       __ip_set_put(set->binding);
1667 +               set->binding = binding;
1668 +               __ip_set_get(set->binding);
1669 +               WRITE_UNLOCK(&ip_set_lock);
1670 +
1671 +               return 0;
1672 +       }
1673 +       binding = ip_set_find_byname(req_bind->binding);
1674 +       if (binding == IP_SET_INVALID_ID)
1675 +               return -ENOENT;
1676 +
1677 +       res = __ip_set_testip(set,
1678 +                             data + sizeof(struct ip_set_req_bind),
1679 +                             size - sizeof(struct ip_set_req_bind),
1680 +                             &ip);
1681 +       DP("set %s, ip: %u.%u.%u.%u, binding %s",
1682 +          set->name, HIPQUAD(ip), ip_set_list[binding]->name);
1683 +
1684 +       if (res >= 0)
1685 +               res = ip_set_hash_add(set->id, ip, binding);
1686 +
1687 +       return res;
1688 +}
1689 +
1690 +#define FOREACH_SET_DO(fn, args...)                            \
1691 +({                                                             \
1692 +       ip_set_id_t __i;                                        \
1693 +       struct ip_set *__set;                                   \
1694 +                                                               \
1695 +       for (__i = 0; __i < ip_set_max; __i++) {                \
1696 +               __set = ip_set_list[__i];                       \
1697 +               if (__set != NULL)                              \
1698 +                       fn(__set , ##args);                     \
1699 +       }                                                       \
1700 +})
1701 +
1702 +static inline void
1703 +__set_hash_del_byid(struct ip_set_hash *set_hash, ip_set_id_t id)
1704 +{
1705 +       if (set_hash->id == id)
1706 +               __set_hash_del(set_hash);
1707 +}
1708 +
1709 +static inline void
1710 +__unbind_default(struct ip_set *set)
1711 +{
1712 +       if (set->binding != IP_SET_INVALID_ID) {
1713 +               /* Sets as binding values are referenced */
1714 +               __ip_set_put(set->binding);
1715 +               set->binding = IP_SET_INVALID_ID;
1716 +       }
1717 +}
1718 +
1719 +static int
1720 +ip_set_unbindip(ip_set_id_t index,
1721 +               const void *data,
1722 +               size_t size)
1723 +{
1724 +       struct ip_set *set;
1725 +       struct ip_set_req_bind *req_bind;
1726 +       ip_set_ip_t ip;
1727 +       int res;
1728 +
1729 +       DP("");
1730 +       if (size < sizeof(struct ip_set_req_bind))
1731 +               return -EINVAL;
1732 +
1733 +       req_bind = (struct ip_set_req_bind *) data;
1734 +       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
1735 +
1736 +       DP("%u %s", index, req_bind->binding);
1737 +       if (index == IP_SET_INVALID_ID) {
1738 +               /* unbind :all: */
1739 +               if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
1740 +                       /* Default binding of sets */
1741 +                       WRITE_LOCK(&ip_set_lock);
1742 +                       FOREACH_SET_DO(__unbind_default);
1743 +                       WRITE_UNLOCK(&ip_set_lock);
1744 +                       return 0;
1745 +               } else if (strcmp(req_bind->binding, IPSET_TOKEN_ALL) == 0) {
1746 +                       /* Flush all bindings of all sets*/
1747 +                       WRITE_LOCK(&ip_set_lock);
1748 +                       FOREACH_HASH_RW_DO(__set_hash_del);
1749 +                       WRITE_UNLOCK(&ip_set_lock);
1750 +                       return 0;
1751 +               }
1752 +               DP("unreachable reached!");
1753 +               return -EINVAL;
1754 +       }
1755 +
1756 +       set = ip_set_list[index];
1757 +       IP_SET_ASSERT(set);
1758 +       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
1759 +               /* Default binding of set */
1760 +               ip_set_id_t binding = ip_set_find_byindex(set->binding);
1761 +
1762 +               if (binding == IP_SET_INVALID_ID)
1763 +                       return -ENOENT;
1764 +
1765 +               WRITE_LOCK(&ip_set_lock);
1766 +               /* Sets in hash values are referenced */
1767 +               __ip_set_put(set->binding);
1768 +               set->binding = IP_SET_INVALID_ID;
1769 +               WRITE_UNLOCK(&ip_set_lock);
1770 +
1771 +               return 0;
1772 +       } else if (strcmp(req_bind->binding, IPSET_TOKEN_ALL) == 0) {
1773 +               /* Flush all bindings */
1774 +
1775 +               WRITE_LOCK(&ip_set_lock);
1776 +               FOREACH_HASH_RW_DO(__set_hash_del_byid, set->id);
1777 +               WRITE_UNLOCK(&ip_set_lock);
1778 +               return 0;
1779 +       }
1780 +
1781 +       res = __ip_set_testip(set,
1782 +                             data + sizeof(struct ip_set_req_bind),
1783 +                             size - sizeof(struct ip_set_req_bind),
1784 +                             &ip);
1785 +
1786 +       DP("set %s, ip: %u.%u.%u.%u", set->name, HIPQUAD(ip));
1787 +       if (res >= 0)
1788 +               res = ip_set_hash_del(set->id, ip);
1789 +
1790 +       return res;
1791 +}
1792 +
1793 +static int
1794 +ip_set_testbind(ip_set_id_t index,
1795 +               const void *data,
1796 +               size_t size)
1797 +{
1798 +       struct ip_set *set = ip_set_list[index];
1799 +       struct ip_set_req_bind *req_bind;
1800 +       ip_set_id_t binding;
1801 +       ip_set_ip_t ip;
1802 +       int res;
1803 +
1804 +       IP_SET_ASSERT(set);
1805 +       if (size < sizeof(struct ip_set_req_bind))
1806 +               return -EINVAL;
1807 +
1808 +       req_bind = (struct ip_set_req_bind *) data;
1809 +       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
1810 +
1811 +       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
1812 +               /* Default binding of set */
1813 +               char *binding_name;
1814 +
1815 +               if (size != sizeof(struct ip_set_req_bind) + IP_SET_MAXNAMELEN)
1816 +                       return -EINVAL;
1817 +
1818 +               binding_name = (char *)(data + sizeof(struct ip_set_req_bind));
1819 +               binding_name[IP_SET_MAXNAMELEN - 1] = '\0';
1820 +
1821 +               binding = ip_set_find_byname(binding_name);
1822 +               if (binding == IP_SET_INVALID_ID)
1823 +                       return -ENOENT;
1824 +
1825 +               res = (set->binding == binding) ? -EEXIST : 0;
1826 +
1827 +               return res;
1828 +       }
1829 +       binding = ip_set_find_byname(req_bind->binding);
1830 +       if (binding == IP_SET_INVALID_ID)
1831 +               return -ENOENT;
1832 +
1833 +
1834 +       res = __ip_set_testip(set,
1835 +                             data + sizeof(struct ip_set_req_bind),
1836 +                             size - sizeof(struct ip_set_req_bind),
1837 +                             &ip);
1838 +       DP("set %s, ip: %u.%u.%u.%u, binding %s",
1839 +          set->name, HIPQUAD(ip), ip_set_list[binding]->name);
1840 +
1841 +       if (res >= 0)
1842 +               res = (ip_set_find_in_hash(set->id, ip) == binding)
1843 +                       ? -EEXIST : 0;
1844 +
1845 +       return res;
1846 +}
1847 +
1848 +static struct ip_set_type *
1849 +find_set_type_rlock(const char *typename)
1850 +{
1851 +       struct ip_set_type *type;
1852 +
1853 +       READ_LOCK(&ip_set_lock);
1854 +       type = find_set_type(typename);
1855 +       if (type == NULL)
1856 +               READ_UNLOCK(&ip_set_lock);
1857 +
1858 +       return type;
1859 +}
1860 +
1861 +static int
1862 +find_free_id(const char *name,
1863 +            ip_set_id_t *index,
1864 +            ip_set_id_t *id)
1865 +{
1866 +       ip_set_id_t i;
1867 +
1868 +       *id = IP_SET_INVALID_ID;
1869 +       for (i = 0;  i < ip_set_max; i++) {
1870 +               if (ip_set_list[i] == NULL) {
1871 +                       if (*id == IP_SET_INVALID_ID)
1872 +                               *id = *index = i;
1873 +               } else if (strcmp(name, ip_set_list[i]->name) == 0)
1874 +                       /* Name clash */
1875 +                       return -EEXIST;
1876 +       }
1877 +       if (*id == IP_SET_INVALID_ID)
1878 +               /* No free slot remained */
1879 +               return -ERANGE;
1880 +       /* Check that index is usable as id (swapping) */
1881 +    check:
1882 +       for (i = 0;  i < ip_set_max; i++) {
1883 +               if (ip_set_list[i] != NULL
1884 +                   && ip_set_list[i]->id == *id) {
1885 +                   *id = i;
1886 +                   goto check;
1887 +               }
1888 +       }
1889 +       return 0;
1890 +}
1891 +
1892 +/*
1893 + * Create a set
1894 + */
1895 +static int
1896 +ip_set_create(const char *name,
1897 +             const char *typename,
1898 +             ip_set_id_t restore,
1899 +             const void *data,
1900 +             size_t size)
1901 +{
1902 +       struct ip_set *set;
1903 +       ip_set_id_t index, id;
1904 +       int res = 0;
1905 +
1906 +       DP("setname: %s, typename: %s, id: %u", name, typename, restore);
1907 +       /*
1908 +        * First, and without any locks, allocate and initialize
1909 +        * a normal base set structure.
1910 +        */
1911 +       set = kmalloc(sizeof(struct ip_set), GFP_KERNEL);
1912 +       if (!set)
1913 +               return -ENOMEM;
1914 +       set->lock = RW_LOCK_UNLOCKED;
1915 +       strncpy(set->name, name, IP_SET_MAXNAMELEN);
1916 +       set->binding = IP_SET_INVALID_ID;
1917 +       atomic_set(&set->ref, 0);
1918 +
1919 +       /*
1920 +        * Next, take the &ip_set_lock, check that we know the type,
1921 +        * and take a reference on the type, to make sure it
1922 +        * stays available while constructing our new set.
1923 +        *
1924 +        * After referencing the type, we drop the &ip_set_lock,
1925 +        * and let the new set construction run without locks.
1926 +        */
1927 +       set->type = find_set_type_rlock(typename);
1928 +       if (set->type == NULL) {
1929 +               /* Try loading the module */
1930 +               char modulename[IP_SET_MAXNAMELEN + strlen("ip_set_") + 1];
1931 +               strcpy(modulename, "ip_set_");
1932 +               strcat(modulename, typename);
1933 +               DP("try to load %s", modulename);
1934 +               request_module(modulename);
1935 +               set->type = find_set_type_rlock(typename);
1936 +       }
1937 +       if (set->type == NULL) {
1938 +               ip_set_printk("no set type '%s', set '%s' not created",
1939 +                             typename, name);
1940 +               kfree(set);
1941 +               return -ENOENT;
1942 +       }
1943 +       __MOD_INC(set->type->me);
1944 +       READ_UNLOCK(&ip_set_lock);
1945 +
1946 +       /*
1947 +        * Without holding any locks, create private part.
1948 +        */
1949 +       res = set->type->create(set, data, size);
1950 +       if (res != 0) {
1951 +               __MOD_DEC(set->type->me);
1952 +               kfree(set);
1953 +               return res;
1954 +       }
1955 +
1956 +       /* BTW, res==0 here. */
1957 +
1958 +       /*
1959 +        * Here, we have a valid, constructed set. &ip_set_lock again,
1960 +        * find free id/index and check that it is not already in
1961 +        * ip_set_list.
1962 +        */
1963 +       WRITE_LOCK(&ip_set_lock);
1964 +       if ((res = find_free_id(set->name, &index, &id)) != 0) {
1965 +               DP("no free id!");
1966 +               goto cleanup;
1967 +       }
1968 +
1969 +       /* Make sure restore gets the same index */
1970 +       if (restore != IP_SET_INVALID_ID && index != restore) {
1971 +               DP("Can't restore, sets are screwed up");
1972 +               res = -ERANGE;
1973 +               goto cleanup;
1974 +       }
1975 +
1976 +       /*
1977 +        * Finally! Add our shiny new set to the list, and be done.
1978 +        */
1979 +       DP("create: '%s' created with index %u, id %u!", set->name, index, id);
1980 +       set->id = id;
1981 +       ip_set_list[index] = set;
1982 +       WRITE_UNLOCK(&ip_set_lock);
1983 +       return res;
1984 +
1985 +    cleanup:
1986 +       WRITE_UNLOCK(&ip_set_lock);
1987 +       set->type->destroy(set);
1988 +       __MOD_DEC(set->type->me);
1989 +       kfree(set);
1990 +       return res;
1991 +}
1992 +
1993 +/*
1994 + * Destroy a given existing set
1995 + */
1996 +static void
1997 +ip_set_destroy_set(ip_set_id_t index)
1998 +{
1999 +       struct ip_set *set = ip_set_list[index];
2000 +
2001 +       IP_SET_ASSERT(set);
2002 +       DP("set: %s",  set->name);
2003 +       WRITE_LOCK(&ip_set_lock);
2004 +       FOREACH_HASH_RW_DO(__set_hash_del_byid, set->id);
2005 +       if (set->binding != IP_SET_INVALID_ID)
2006 +               __ip_set_put(set->binding);
2007 +       ip_set_list[index] = NULL;
2008 +       WRITE_UNLOCK(&ip_set_lock);
2009 +
2010 +       /* Must call it without holding any lock */
2011 +       set->type->destroy(set);
2012 +       __MOD_DEC(set->type->me);
2013 +       kfree(set);
2014 +}
2015 +
2016 +/*
2017 + * Destroy a set - or all sets
2018 + * Sets must not be referenced/used.
2019 + */
2020 +static int
2021 +ip_set_destroy(ip_set_id_t index)
2022 +{
2023 +       ip_set_id_t i;
2024 +
2025 +       /* ref modification always protected by the mutex */
2026 +       if (index != IP_SET_INVALID_ID) {
2027 +               if (atomic_read(&ip_set_list[index]->ref))
2028 +                       return -EBUSY;
2029 +               ip_set_destroy_set(index);
2030 +       } else {
2031 +               for (i = 0; i < ip_set_max; i++) {
2032 +                       if (ip_set_list[i] != NULL
2033 +                           && (atomic_read(&ip_set_list[i]->ref)))
2034 +                               return -EBUSY;
2035 +               }
2036 +
2037 +               for (i = 0; i < ip_set_max; i++) {
2038 +                       if (ip_set_list[i] != NULL)
2039 +                               ip_set_destroy_set(i);
2040 +               }
2041 +       }
2042 +       return 0;
2043 +}
2044 +
2045 +static void
2046 +ip_set_flush_set(struct ip_set *set)
2047 +{
2048 +       DP("set: %s %u",  set->name, set->id);
2049 +
2050 +       write_lock_bh(&set->lock);
2051 +       set->type->flush(set);
2052 +       write_unlock_bh(&set->lock);
2053 +}
2054 +
2055 +/*
2056 + * Flush data in a set - or in all sets
2057 + */
2058 +static int
2059 +ip_set_flush(ip_set_id_t index)
2060 +{
2061 +       if (index != IP_SET_INVALID_ID) {
2062 +               IP_SET_ASSERT(ip_set_list[index]);
2063 +               ip_set_flush_set(ip_set_list[index]);
2064 +       } else
2065 +               FOREACH_SET_DO(ip_set_flush_set);
2066 +
2067 +       return 0;
2068 +}
2069 +
2070 +/* Rename a set */
2071 +static int
2072 +ip_set_rename(ip_set_id_t index, const char *name)
2073 +{
2074 +       struct ip_set *set = ip_set_list[index];
2075 +       ip_set_id_t i;
2076 +       int res = 0;
2077 +
2078 +       DP("set: %s to %s",  set->name, name);
2079 +       WRITE_LOCK(&ip_set_lock);
2080 +       for (i = 0; i < ip_set_max; i++) {
2081 +               if (ip_set_list[i] != NULL
2082 +                   && strncmp(ip_set_list[i]->name,
2083 +                              name,
2084 +                              IP_SET_MAXNAMELEN - 1) == 0) {
2085 +                       res = -EEXIST;
2086 +                       goto unlock;
2087 +               }
2088 +       }
2089 +       strncpy(set->name, name, IP_SET_MAXNAMELEN);
2090 +    unlock:
2091 +       WRITE_UNLOCK(&ip_set_lock);
2092 +       return res;
2093 +}
2094 +
2095 +/*
2096 + * Swap two sets so that name/index points to the other.
2097 + * References are also swapped.
2098 + */
2099 +static int
2100 +ip_set_swap(ip_set_id_t from_index, ip_set_id_t to_index)
2101 +{
2102 +       struct ip_set *from = ip_set_list[from_index];
2103 +       struct ip_set *to = ip_set_list[to_index];
2104 +       char from_name[IP_SET_MAXNAMELEN];
2105 +       u_int32_t from_ref;
2106 +
2107 +       DP("set: %s to %s",  from->name, to->name);
2108 +       /* Type can't be changed. Artifical restriction. */
2109 +       if (from->type->typecode != to->type->typecode)
2110 +               return -ENOEXEC;
2111 +
2112 +       /* No magic here: ref munging protected by the mutex */
2113 +       WRITE_LOCK(&ip_set_lock);
2114 +       strncpy(from_name, from->name, IP_SET_MAXNAMELEN);
2115 +       from_ref = atomic_read(&from->ref);
2116 +
2117 +       strncpy(from->name, to->name, IP_SET_MAXNAMELEN);
2118 +       atomic_set(&from->ref, atomic_read(&to->ref));
2119 +       strncpy(to->name, from_name, IP_SET_MAXNAMELEN);
2120 +       atomic_set(&to->ref, from_ref);
2121 +
2122 +       ip_set_list[from_index] = to;
2123 +       ip_set_list[to_index] = from;
2124 +
2125 +       WRITE_UNLOCK(&ip_set_lock);
2126 +       return 0;
2127 +}
2128 +
2129 +/*
2130 + * List set data
2131 + */
2132 +
2133 +static inline void
2134 +__set_hash_bindings_size_list(struct ip_set_hash *set_hash,
2135 +                             ip_set_id_t id, size_t *size)
2136 +{
2137 +       if (set_hash->id == id)
2138 +               *size += sizeof(struct ip_set_hash_list);
2139 +}
2140 +
2141 +static inline void
2142 +__set_hash_bindings_size_save(struct ip_set_hash *set_hash,
2143 +                             ip_set_id_t id, size_t *size)
2144 +{
2145 +       if (set_hash->id == id)
2146 +               *size += sizeof(struct ip_set_hash_save);
2147 +}
2148 +
2149 +static inline void
2150 +__set_hash_bindings(struct ip_set_hash *set_hash,
2151 +                   ip_set_id_t id, void *data, int *used)
2152 +{
2153 +       if (set_hash->id == id) {
2154 +               struct ip_set_hash_list *hash_list =
2155 +                       (struct ip_set_hash_list *)(data + *used);
2156 +
2157 +               hash_list->ip = set_hash->ip;
2158 +               hash_list->binding = set_hash->binding;
2159 +               *used += sizeof(struct ip_set_hash_list);
2160 +       }
2161 +}
2162 +
2163 +static int ip_set_list_set(ip_set_id_t index,
2164 +                          void *data,
2165 +                          int *used,
2166 +                          int len)
2167 +{
2168 +       struct ip_set *set = ip_set_list[index];
2169 +       struct ip_set_list *set_list;
2170 +
2171 +       /* Pointer to our header */
2172 +       set_list = (struct ip_set_list *) (data + *used);
2173 +
2174 +       DP("set: %s, used: %d %p %p", set->name, *used, data, data + *used);
2175 +
2176 +       /* Get and ensure header size */
2177 +       if (*used + sizeof(struct ip_set_list) > len)
2178 +               goto not_enough_mem;
2179 +       *used += sizeof(struct ip_set_list);
2180 +
2181 +       read_lock_bh(&set->lock);
2182 +       /* Get and ensure set specific header size */
2183 +       set_list->header_size = set->type->header_size;
2184 +       if (*used + set_list->header_size > len)
2185 +               goto unlock_set;
2186 +
2187 +       /* Fill in the header */
2188 +       set_list->index = index;
2189 +       set_list->binding = set->binding;
2190 +       set_list->ref = atomic_read(&set->ref);
2191 +
2192 +       /* Fill in set spefific header data */
2193 +       DP("call list_header");
2194 +       set->type->list_header(set, data + *used);
2195 +       DP("call list_header, done");
2196 +       *used += set_list->header_size;
2197 +
2198 +       /* Get and ensure set specific members size */
2199 +       DP("call list_members_size");
2200 +       set_list->members_size = set->type->list_members_size(set);
2201 +       DP("call list_members_size, done");
2202 +       if (*used + set_list->members_size > len)
2203 +               goto unlock_set;
2204 +
2205 +       /* Fill in set spefific members data */
2206 +       DP("call list_members");
2207 +       set->type->list_members(set, data + *used);
2208 +       DP("call list_members, done");
2209 +       *used += set_list->members_size;
2210 +       read_unlock_bh(&set->lock);
2211 +
2212 +       /* Bindings */
2213 +
2214 +       /* Get and ensure set specific bindings size */
2215 +       set_list->bindings_size = 0;
2216 +       FOREACH_HASH_DO(__set_hash_bindings_size_list,
2217 +                       set->id, &set_list->bindings_size);
2218 +       if (*used + set_list->bindings_size > len)
2219 +               goto not_enough_mem;
2220 +
2221 +       /* Fill in set spefific bindings data */
2222 +       FOREACH_HASH_DO(__set_hash_bindings, set->id, data, used);
2223 +
2224 +       return 0;
2225 +
2226 +    unlock_set:
2227 +       read_unlock_bh(&set->lock);
2228 +    not_enough_mem:
2229 +       DP("not enough mem, try again");
2230 +       return -EAGAIN;
2231 +}
2232 +
2233 +/*
2234 + * Save sets
2235 + */
2236 +static int ip_set_save_set(ip_set_id_t index,
2237 +                          void *data,
2238 +                          int *used,
2239 +                          int len)
2240 +{
2241 +       struct ip_set *set;
2242 +       struct ip_set_save *set_save;
2243 +
2244 +       /* Pointer to our header */
2245 +       set_save = (struct ip_set_save *) (data + *used);
2246 +
2247 +       /* Get and ensure header size */
2248 +       if (*used + sizeof(struct ip_set_save) > len)
2249 +               goto not_enough_mem;
2250 +       *used += sizeof(struct ip_set_save);
2251 +
2252 +       set = ip_set_list[index];
2253 +       DP("set: %s, used: %u(%u) %p %p", set->name, *used, len,
2254 +          data, data + *used);
2255 +
2256 +       read_lock_bh(&set->lock);
2257 +       /* Get and ensure set specific header size */
2258 +       set_save->header_size = set->type->header_size;
2259 +       if (*used + set_save->header_size > len)
2260 +               goto unlock_set;
2261 +
2262 +       /* Fill in the header */
2263 +       set_save->index = index;
2264 +       set_save->binding = set->binding;
2265 +
2266 +       /* Fill in set spefific header data */
2267 +       set->type->list_header(set, data + *used);
2268 +       *used += set_save->header_size;
2269 +
2270 +       DP("set header filled: %s, used: %u %p %p", set->name, *used,
2271 +          data, data + *used);
2272 +       /* Get and ensure set specific members size */
2273 +       set_save->members_size = set->type->list_members_size(set);
2274 +       if (*used + set_save->members_size > len)
2275 +               goto unlock_set;
2276 +
2277 +       /* Fill in set spefific members data */
2278 +       set->type->list_members(set, data + *used);
2279 +       *used += set_save->members_size;
2280 +       read_unlock_bh(&set->lock);
2281 +       DP("set members filled: %s, used: %u %p %p", set->name, *used,
2282 +          data, data + *used);
2283 +       return 0;
2284 +
2285 +    unlock_set:
2286 +       read_unlock_bh(&set->lock);
2287 +    not_enough_mem:
2288 +       DP("not enough mem, try again");
2289 +       return -EAGAIN;
2290 +}
2291 +
2292 +static inline void
2293 +__set_hash_save_bindings(struct ip_set_hash *set_hash,
2294 +                        ip_set_id_t id,
2295 +                        void *data,
2296 +                        int *used,
2297 +                        int len,
2298 +                        int *res)
2299 +{
2300 +       if (*res == 0
2301 +           && (id == IP_SET_INVALID_ID || set_hash->id == id)) {
2302 +               struct ip_set_hash_save *hash_save =
2303 +                       (struct ip_set_hash_save *)(data + *used);
2304 +               /* Ensure bindings size */
2305 +               if (*used + sizeof(struct ip_set_hash_save) > len) {
2306 +                       *res = -ENOMEM;
2307 +                       return;
2308 +               }
2309 +               hash_save->id = set_hash->id;
2310 +               hash_save->ip = set_hash->ip;
2311 +               hash_save->binding = set_hash->binding;
2312 +               *used += sizeof(struct ip_set_hash_save);
2313 +       }
2314 +}
2315 +
2316 +static int ip_set_save_bindings(ip_set_id_t index,
2317 +                               void *data,
2318 +                               int *used,
2319 +                               int len)
2320 +{
2321 +       int res = 0;
2322 +       struct ip_set_save *set_save;
2323 +
2324 +       DP("used %u, len %u", *used, len);
2325 +       /* Get and ensure header size */
2326 +       if (*used + sizeof(struct ip_set_save) > len)
2327 +               return -ENOMEM;
2328 +
2329 +       /* Marker */
2330 +       set_save = (struct ip_set_save *) (data + *used);
2331 +       set_save->index = IP_SET_INVALID_ID;
2332 +       *used += sizeof(struct ip_set_save);
2333 +
2334 +       DP("marker added used %u, len %u", *used, len);
2335 +       /* Fill in bindings data */
2336 +       if (index != IP_SET_INVALID_ID)
2337 +               /* Sets are identified by id in hash */
2338 +               index = ip_set_list[index]->id;
2339 +       FOREACH_HASH_DO(__set_hash_save_bindings, index, data, used, len, &res);
2340 +
2341 +       return res;
2342 +}
2343 +
2344 +/*
2345 + * Restore sets
2346 + */
2347 +static int ip_set_restore(void *data,
2348 +                         int len)
2349 +{
2350 +       int res = 0;
2351 +       int line = 0, used = 0, members_size;
2352 +       struct ip_set *set;
2353 +       struct ip_set_hash_save *hash_save;
2354 +       struct ip_set_restore *set_restore;
2355 +       ip_set_id_t index;
2356 +
2357 +       /* Loop to restore sets */
2358 +       while (1) {
2359 +               line++;
2360 +
2361 +               DP("%u %u %u", used, sizeof(struct ip_set_restore), len);
2362 +               /* Get and ensure header size */
2363 +               if (used + sizeof(struct ip_set_restore) > len)
2364 +                       return line;
2365 +               set_restore = (struct ip_set_restore *) (data + used);
2366 +               used += sizeof(struct ip_set_restore);
2367 +
2368 +               /* Ensure data size */
2369 +               if (used
2370 +                   + set_restore->header_size
2371 +                   + set_restore->members_size > len)
2372 +                       return line;
2373 +
2374 +               /* Check marker */
2375 +               if (set_restore->index == IP_SET_INVALID_ID) {
2376 +                       line--;
2377 +                       goto bindings;
2378 +               }
2379 +
2380 +               /* Try to create the set */
2381 +               DP("restore %s %s", set_restore->name, set_restore->typename);
2382 +               res = ip_set_create(set_restore->name,
2383 +                                   set_restore->typename,
2384 +                                   set_restore->index,
2385 +                                   data + used,
2386 +                                   set_restore->header_size);
2387 +
2388 +               if (res != 0)
2389 +                       return line;
2390 +               used += set_restore->header_size;
2391 +
2392 +               index = ip_set_find_byindex(set_restore->index);
2393 +               DP("index %u, restore_index %u", index, set_restore->index);
2394 +               if (index != set_restore->index)
2395 +                       return line;
2396 +               /* Try to restore members data */
2397 +               set = ip_set_list[index];
2398 +               members_size = 0;
2399 +               DP("members_size %u reqsize %u",
2400 +                  set_restore->members_size, set->type->reqsize);
2401 +               while (members_size + set->type->reqsize <=
2402 +                      set_restore->members_size) {
2403 +                       line++;
2404 +                       DP("members: %u, line %u", members_size, line);
2405 +                       res = __ip_set_addip(index,
2406 +                                          data + used + members_size,
2407 +                                          set->type->reqsize);
2408 +                       if (!(res == 0 || res == -EEXIST))
2409 +                               return line;
2410 +                       members_size += set->type->reqsize;
2411 +               }
2412 +
2413 +               DP("members_size %u  %u",
2414 +                  set_restore->members_size, members_size);
2415 +               if (members_size != set_restore->members_size)
2416 +                       return line++;
2417 +               used += set_restore->members_size;
2418 +       }
2419 +
2420 +   bindings:
2421 +       /* Loop to restore bindings */
2422 +       while (used < len) {
2423 +               line++;
2424 +
2425 +               DP("restore binding, line %u", line);
2426 +               /* Get and ensure size */
2427 +               if (used + sizeof(struct ip_set_hash_save) > len)
2428 +                       return line;
2429 +               hash_save = (struct ip_set_hash_save *) (data + used);
2430 +               used += sizeof(struct ip_set_hash_save);
2431 +
2432 +               /* hash_save->id is used to store the index */
2433 +               index = ip_set_find_byindex(hash_save->id);
2434 +               DP("restore binding index %u, id %u, %u -> %u",
2435 +                  index, hash_save->id, hash_save->ip, hash_save->binding);
2436 +               if (index != hash_save->id)
2437 +                       return line;
2438 +
2439 +               set = ip_set_list[hash_save->id];
2440 +               /* Null valued IP means default binding */
2441 +               if (hash_save->ip)
2442 +                       res = ip_set_hash_add(set->id,
2443 +                                             hash_save->ip,
2444 +                                             hash_save->binding);
2445 +               else {
2446 +                       IP_SET_ASSERT(set->binding == IP_SET_INVALID_ID);
2447 +                       WRITE_LOCK(&ip_set_lock);
2448 +                       set->binding = hash_save->binding;
2449 +                       __ip_set_get(set->binding);
2450 +                       WRITE_UNLOCK(&ip_set_lock);
2451 +                       DP("default binding: %u", set->binding);
2452 +               }
2453 +               if (res != 0)
2454 +                       return line;
2455 +       }
2456 +       if (used != len)
2457 +               return line;
2458 +
2459 +       return 0;
2460 +}
2461 +
2462 +static int
2463 +ip_set_sockfn_set(struct sock *sk, int optval, void *user, unsigned int len)
2464 +{
2465 +       void *data;
2466 +       int res = 0;            /* Assume OK */
2467 +       unsigned *op;
2468 +       struct ip_set_req_adt *req_adt;
2469 +       ip_set_id_t index = IP_SET_INVALID_ID;
2470 +       int (*adtfn)(ip_set_id_t index,
2471 +                    const void *data, size_t size);
2472 +       struct fn_table {
2473 +               int (*fn)(ip_set_id_t index,
2474 +                         const void *data, size_t size);
2475 +       } adtfn_table[] =
2476 +       { { ip_set_addip }, { ip_set_delip }, { ip_set_testip},
2477 +         { ip_set_bindip}, { ip_set_unbindip }, { ip_set_testbind },
2478 +       };
2479 +
2480 +       DP("optval=%d, user=%p, len=%d", optval, user, len);
2481 +       if (!capable(CAP_NET_ADMIN))
2482 +               return -EPERM;
2483 +       if (optval != SO_IP_SET)
2484 +               return -EBADF;
2485 +       if (len <= sizeof(unsigned)) {
2486 +               ip_set_printk("short userdata (want >%zu, got %u)",
2487 +                             sizeof(unsigned), len);
2488 +               return -EINVAL;
2489 +       }
2490 +       data = vmalloc(len);
2491 +       if (!data) {
2492 +               DP("out of mem for %u bytes", len);
2493 +               return -ENOMEM;
2494 +       }
2495 +       if (copy_from_user(data, user, len) != 0) {
2496 +               res = -EFAULT;
2497 +               goto done;
2498 +       }
2499 +       if (down_interruptible(&ip_set_app_mutex)) {
2500 +               res = -EINTR;
2501 +               goto done;
2502 +       }
2503 +
2504 +       op = (unsigned *)data;
2505 +       DP("op=%x", *op);
2506 +
2507 +       if (*op < IP_SET_OP_VERSION) {
2508 +               /* Check the version at the beginning of operations */
2509 +               struct ip_set_req_version *req_version =
2510 +                       (struct ip_set_req_version *) data;
2511 +               if (req_version->version != IP_SET_PROTOCOL_VERSION) {
2512 +                       res = -EPROTO;
2513 +                       goto done;
2514 +               }
2515 +       }
2516 +
2517 +       switch (*op) {
2518 +       case IP_SET_OP_CREATE:{
2519 +               struct ip_set_req_create *req_create
2520 +                       = (struct ip_set_req_create *) data;
2521 +
2522 +               if (len <= sizeof(struct ip_set_req_create)) {
2523 +                       ip_set_printk("short CREATE data (want >%zu, got %u)",
2524 +                                     sizeof(struct ip_set_req_create), len);
2525 +                       res = -EINVAL;
2526 +                       goto done;
2527 +               }
2528 +               req_create->name[IP_SET_MAXNAMELEN - 1] = '\0';
2529 +               req_create->typename[IP_SET_MAXNAMELEN - 1] = '\0';
2530 +               res = ip_set_create(req_create->name,
2531 +                                   req_create->typename,
2532 +                                   IP_SET_INVALID_ID,
2533 +                                   data + sizeof(struct ip_set_req_create),
2534 +                                   len - sizeof(struct ip_set_req_create));
2535 +               goto done;
2536 +       }
2537 +       case IP_SET_OP_DESTROY:{
2538 +               struct ip_set_req_std *req_destroy
2539 +                       = (struct ip_set_req_std *) data;
2540 +
2541 +               if (len != sizeof(struct ip_set_req_std)) {
2542 +                       ip_set_printk("invalid DESTROY data (want %zu, got %u)",
2543 +                                     sizeof(struct ip_set_req_std), len);
2544 +                       res = -EINVAL;
2545 +                       goto done;
2546 +               }
2547 +               if (strcmp(req_destroy->name, IPSET_TOKEN_ALL) == 0) {
2548 +                       /* Destroy all sets */
2549 +                       index = IP_SET_INVALID_ID;
2550 +               } else {
2551 +                       req_destroy->name[IP_SET_MAXNAMELEN - 1] = '\0';
2552 +                       index = ip_set_find_byname(req_destroy->name);
2553 +
2554 +                       if (index == IP_SET_INVALID_ID) {
2555 +                               res = -ENOENT;
2556 +                               goto done;
2557 +                       }
2558 +               }
2559 +
2560 +               res = ip_set_destroy(index);
2561 +               goto done;
2562 +       }
2563 +       case IP_SET_OP_FLUSH:{
2564 +               struct ip_set_req_std *req_flush =
2565 +                       (struct ip_set_req_std *) data;
2566 +
2567 +               if (len != sizeof(struct ip_set_req_std)) {
2568 +                       ip_set_printk("invalid FLUSH data (want %zu, got %u)",
2569 +                                     sizeof(struct ip_set_req_std), len);
2570 +                       res = -EINVAL;
2571 +                       goto done;
2572 +               }
2573 +               if (strcmp(req_flush->name, IPSET_TOKEN_ALL) == 0) {
2574 +                       /* Flush all sets */
2575 +                       index = IP_SET_INVALID_ID;
2576 +               } else {
2577 +                       req_flush->name[IP_SET_MAXNAMELEN - 1] = '\0';
2578 +                       index = ip_set_find_byname(req_flush->name);
2579 +
2580 +                       if (index == IP_SET_INVALID_ID) {
2581 +                               res = -ENOENT;
2582 +                               goto done;
2583 +                       }
2584 +               }
2585 +               res = ip_set_flush(index);
2586 +               goto done;
2587 +       }
2588 +       case IP_SET_OP_RENAME:{
2589 +               struct ip_set_req_create *req_rename
2590 +                       = (struct ip_set_req_create *) data;
2591 +
2592 +               if (len != sizeof(struct ip_set_req_create)) {
2593 +                       ip_set_printk("invalid RENAME data (want %zu, got %u)",
2594 +                                     sizeof(struct ip_set_req_create), len);
2595 +                       res = -EINVAL;
2596 +                       goto done;
2597 +               }
2598 +
2599 +               req_rename->name[IP_SET_MAXNAMELEN - 1] = '\0';
2600 +               req_rename->typename[IP_SET_MAXNAMELEN - 1] = '\0';
2601 +
2602 +               index = ip_set_find_byname(req_rename->name);
2603 +               if (index == IP_SET_INVALID_ID) {
2604 +                       res = -ENOENT;
2605 +                       goto done;
2606 +               }
2607 +               res = ip_set_rename(index, req_rename->typename);
2608 +               goto done;
2609 +       }
2610 +       case IP_SET_OP_SWAP:{
2611 +               struct ip_set_req_create *req_swap
2612 +                       = (struct ip_set_req_create *) data;
2613 +               ip_set_id_t to_index;
2614 +
2615 +               if (len != sizeof(struct ip_set_req_create)) {
2616 +                       ip_set_printk("invalid SWAP data (want %zu, got %u)",
2617 +                                     sizeof(struct ip_set_req_create), len);
2618 +                       res = -EINVAL;
2619 +                       goto done;
2620 +               }
2621 +
2622 +               req_swap->name[IP_SET_MAXNAMELEN - 1] = '\0';
2623 +               req_swap->typename[IP_SET_MAXNAMELEN - 1] = '\0';
2624 +
2625 +               index = ip_set_find_byname(req_swap->name);
2626 +               if (index == IP_SET_INVALID_ID) {
2627 +                       res = -ENOENT;
2628 +                       goto done;
2629 +               }
2630 +               to_index = ip_set_find_byname(req_swap->typename);
2631 +               if (to_index == IP_SET_INVALID_ID) {
2632 +                       res = -ENOENT;
2633 +                       goto done;
2634 +               }
2635 +               res = ip_set_swap(index, to_index);
2636 +               goto done;
2637 +       }
2638 +       default:
2639 +               break;  /* Set identified by id */
2640 +       }
2641 +
2642 +       /* There we may have add/del/test/bind/unbind/test_bind operations */
2643 +       if (*op < IP_SET_OP_ADD_IP || *op > IP_SET_OP_TEST_BIND_SET) {
2644 +               res = -EBADMSG;
2645 +               goto done;
2646 +       }
2647 +       adtfn = adtfn_table[*op - IP_SET_OP_ADD_IP].fn;
2648 +
2649 +       if (len < sizeof(struct ip_set_req_adt)) {
2650 +               ip_set_printk("short data in adt request (want >=%zu, got %u)",
2651 +                             sizeof(struct ip_set_req_adt), len);
2652 +               res = -EINVAL;
2653 +               goto done;
2654 +       }
2655 +       req_adt = (struct ip_set_req_adt *) data;
2656 +
2657 +       /* -U :all: :all:|:default: uses IP_SET_INVALID_ID */
2658 +       if (!(*op == IP_SET_OP_UNBIND_SET
2659 +             && req_adt->index == IP_SET_INVALID_ID)) {
2660 +               index = ip_set_find_byindex(req_adt->index);
2661 +               if (index == IP_SET_INVALID_ID) {
2662 +                       res = -ENOENT;
2663 +                       goto done;
2664 +               }
2665 +       }
2666 +       res = adtfn(index, data, len);
2667 +
2668 +    done:
2669 +       up(&ip_set_app_mutex);
2670 +       vfree(data);
2671 +       if (res > 0)
2672 +               res = 0;
2673 +       DP("final result %d", res);
2674 +       return res;
2675 +}
2676 +
2677 +static int
2678 +ip_set_sockfn_get(struct sock *sk, int optval, void *user, int *len)
2679 +{
2680 +       int res = 0;
2681 +       unsigned *op;
2682 +       ip_set_id_t index = IP_SET_INVALID_ID;
2683 +       void *data;
2684 +       int copylen = *len;
2685 +
2686 +       DP("optval=%d, user=%p, len=%d", optval, user, *len);
2687 +       if (!capable(CAP_NET_ADMIN))
2688 +               return -EPERM;
2689 +       if (optval != SO_IP_SET)
2690 +               return -EBADF;
2691 +       if (*len < sizeof(unsigned)) {
2692 +               ip_set_printk("short userdata (want >=%zu, got %d)",
2693 +                             sizeof(unsigned), *len);
2694 +               return -EINVAL;
2695 +       }
2696 +       data = vmalloc(*len);
2697 +       if (!data) {
2698 +               DP("out of mem for %d bytes", *len);
2699 +               return -ENOMEM;
2700 +       }
2701 +       if (copy_from_user(data, user, *len) != 0) {
2702 +               res = -EFAULT;
2703 +               goto done;
2704 +       }
2705 +       if (down_interruptible(&ip_set_app_mutex)) {
2706 +               res = -EINTR;
2707 +               goto done;
2708 +       }
2709 +
2710 +       op = (unsigned *) data;
2711 +       DP("op=%x", *op);
2712 +
2713 +       if (*op < IP_SET_OP_VERSION) {
2714 +               /* Check the version at the beginning of operations */
2715 +               struct ip_set_req_version *req_version =
2716 +                       (struct ip_set_req_version *) data;
2717 +               if (req_version->version != IP_SET_PROTOCOL_VERSION) {
2718 +                       res = -EPROTO;
2719 +                       goto done;
2720 +               }
2721 +       }
2722 +
2723 +       switch (*op) {
2724 +       case IP_SET_OP_VERSION: {
2725 +               struct ip_set_req_version *req_version =
2726 +                   (struct ip_set_req_version *) data;
2727 +
2728 +               if (*len != sizeof(struct ip_set_req_version)) {
2729 +                       ip_set_printk("invalid VERSION (want %zu, got %d)",
2730 +                                     sizeof(struct ip_set_req_version),
2731 +                                     *len);
2732 +                       res = -EINVAL;
2733 +                       goto done;
2734 +               }
2735 +
2736 +               req_version->version = IP_SET_PROTOCOL_VERSION;
2737 +               res = copy_to_user(user, req_version,
2738 +                                  sizeof(struct ip_set_req_version));
2739 +               goto done;
2740 +       }
2741 +       case IP_SET_OP_GET_BYNAME: {
2742 +               struct ip_set_req_get_set *req_get
2743 +                       = (struct ip_set_req_get_set *) data;
2744 +
2745 +               if (*len != sizeof(struct ip_set_req_get_set)) {
2746 +                       ip_set_printk("invalid GET_BYNAME (want %zu, got %d)",
2747 +                                     sizeof(struct ip_set_req_get_set), *len);
2748 +                       res = -EINVAL;
2749 +                       goto done;
2750 +               }
2751 +               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
2752 +               index = ip_set_find_byname(req_get->set.name);
2753 +               req_get->set.index = index;
2754 +               goto copy;
2755 +       }
2756 +       case IP_SET_OP_GET_BYINDEX: {
2757 +               struct ip_set_req_get_set *req_get
2758 +                       = (struct ip_set_req_get_set *) data;
2759 +
2760 +               if (*len != sizeof(struct ip_set_req_get_set)) {
2761 +                       ip_set_printk("invalid GET_BYINDEX (want %zu, got %d)",
2762 +                                     sizeof(struct ip_set_req_get_set), *len);
2763 +                       res = -EINVAL;
2764 +                       goto done;
2765 +               }
2766 +               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
2767 +               index = ip_set_find_byindex(req_get->set.index);
2768 +               strncpy(req_get->set.name,
2769 +                       index == IP_SET_INVALID_ID ? ""
2770 +                       : ip_set_list[index]->name, IP_SET_MAXNAMELEN);
2771 +               goto copy;
2772 +       }
2773 +       case IP_SET_OP_ADT_GET: {
2774 +               struct ip_set_req_adt_get *req_get
2775 +                       = (struct ip_set_req_adt_get *) data;
2776 +
2777 +               if (*len != sizeof(struct ip_set_req_adt_get)) {
2778 +                       ip_set_printk("invalid ADT_GET (want %zu, got %d)",
2779 +                                     sizeof(struct ip_set_req_adt_get), *len);
2780 +                       res = -EINVAL;
2781 +                       goto done;
2782 +               }
2783 +               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
2784 +               index = ip_set_find_byname(req_get->set.name);
2785 +               if (index != IP_SET_INVALID_ID) {
2786 +                       req_get->set.index = index;
2787 +                       strncpy(req_get->typename,
2788 +                               ip_set_list[index]->type->typename,
2789 +                               IP_SET_MAXNAMELEN - 1);
2790 +               } else {
2791 +                       res = -ENOENT;
2792 +                       goto done;
2793 +               }
2794 +               goto copy;
2795 +       }
2796 +       case IP_SET_OP_MAX_SETS: {
2797 +               struct ip_set_req_max_sets *req_max_sets
2798 +                       = (struct ip_set_req_max_sets *) data;
2799 +               ip_set_id_t i;
2800 +
2801 +               if (*len != sizeof(struct ip_set_req_max_sets)) {
2802 +                       ip_set_printk("invalid MAX_SETS (want %zu, got %d)",
2803 +                                     sizeof(struct ip_set_req_max_sets), *len);
2804 +                       res = -EINVAL;
2805 +                       goto done;
2806 +               }
2807 +
2808 +               if (strcmp(req_max_sets->set.name, IPSET_TOKEN_ALL) == 0) {
2809 +                       req_max_sets->set.index = IP_SET_INVALID_ID;
2810 +               } else {
2811 +                       req_max_sets->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
2812 +                       req_max_sets->set.index =
2813 +                               ip_set_find_byname(req_max_sets->set.name);
2814 +                       if (req_max_sets->set.index == IP_SET_INVALID_ID) {
2815 +                               res = -ENOENT;
2816 +                               goto done;
2817 +                       }
2818 +               }
2819 +               req_max_sets->max_sets = ip_set_max;
2820 +               req_max_sets->sets = 0;
2821 +               for (i = 0; i < ip_set_max; i++) {
2822 +                       if (ip_set_list[i] != NULL)
2823 +                               req_max_sets->sets++;
2824 +               }
2825 +               goto copy;
2826 +       }
2827 +       case IP_SET_OP_LIST_SIZE:
2828 +       case IP_SET_OP_SAVE_SIZE: {
2829 +               struct ip_set_req_setnames *req_setnames
2830 +                       = (struct ip_set_req_setnames *) data;
2831 +               struct ip_set_name_list *name_list;
2832 +               struct ip_set *set;
2833 +               ip_set_id_t i;
2834 +               int used;
2835 +
2836 +               if (*len < sizeof(struct ip_set_req_setnames)) {
2837 +                       ip_set_printk("short LIST_SIZE (want >=%zu, got %d)",
2838 +                                     sizeof(struct ip_set_req_setnames), *len);
2839 +                       res = -EINVAL;
2840 +                       goto done;
2841 +               }
2842 +
2843 +               req_setnames->size = 0;
2844 +               used = sizeof(struct ip_set_req_setnames);
2845 +               for (i = 0; i < ip_set_max; i++) {
2846 +                       if (ip_set_list[i] == NULL)
2847 +                               continue;
2848 +                       name_list = (struct ip_set_name_list *)
2849 +                               (data + used);
2850 +                       used += sizeof(struct ip_set_name_list);
2851 +                       if (used > copylen) {
2852 +                               res = -EAGAIN;
2853 +                               goto done;
2854 +                       }
2855 +                       set = ip_set_list[i];
2856 +                       /* Fill in index, name, etc. */
2857 +                       name_list->index = i;
2858 +                       name_list->id = set->id;
2859 +                       strncpy(name_list->name,
2860 +                               set->name,
2861 +                               IP_SET_MAXNAMELEN - 1);
2862 +                       strncpy(name_list->typename,
2863 +                               set->type->typename,
2864 +                               IP_SET_MAXNAMELEN - 1);
2865 +                       DP("filled %s of type %s, index %u\n",
2866 +                          name_list->name, name_list->typename,
2867 +                          name_list->index);
2868 +                       if (!(req_setnames->index == IP_SET_INVALID_ID
2869 +                             || req_setnames->index == i))
2870 +                             continue;
2871 +                       /* Update size */
2872 +                       switch (*op) {
2873 +                       case IP_SET_OP_LIST_SIZE: {
2874 +                               req_setnames->size += sizeof(struct ip_set_list)
2875 +                                       + set->type->header_size
2876 +                                       + set->type->list_members_size(set);
2877 +                               FOREACH_HASH_DO(__set_hash_bindings_size_list,
2878 +                                               i, &req_setnames->size);
2879 +                               break;
2880 +                       }
2881 +                       case IP_SET_OP_SAVE_SIZE: {
2882 +                               req_setnames->size += sizeof(struct ip_set_save)
2883 +                                       + set->type->header_size
2884 +                                       + set->type->list_members_size(set);
2885 +                               FOREACH_HASH_DO(__set_hash_bindings_size_save,
2886 +                                               i, &req_setnames->size);
2887 +                               break;
2888 +                       }
2889 +                       default:
2890 +                               break;
2891 +                       }
2892 +               }
2893 +               if (copylen != used) {
2894 +                       res = -EAGAIN;
2895 +                       goto done;
2896 +               }
2897 +               goto copy;
2898 +       }
2899 +       case IP_SET_OP_LIST: {
2900 +               struct ip_set_req_list *req_list
2901 +                       = (struct ip_set_req_list *) data;
2902 +               ip_set_id_t i;
2903 +               int used;
2904 +
2905 +               if (*len < sizeof(struct ip_set_req_list)) {
2906 +                       ip_set_printk("short LIST (want >=%zu, got %d)",
2907 +                                     sizeof(struct ip_set_req_list), *len);
2908 +                       res = -EINVAL;
2909 +                       goto done;
2910 +               }
2911 +               index = req_list->index;
2912 +               if (index != IP_SET_INVALID_ID
2913 +                   && ip_set_find_byindex(index) != index) {
2914 +                       res = -ENOENT;
2915 +                       goto done;
2916 +               }
2917 +               used = 0;
2918 +               if (index == IP_SET_INVALID_ID) {
2919 +                       /* List all sets */
2920 +                       for (i = 0; i < ip_set_max && res == 0; i++) {
2921 +                               if (ip_set_list[i] != NULL)
2922 +                                       res = ip_set_list_set(i, data, &used, *len);
2923 +                       }
2924 +               } else {
2925 +                       /* List an individual set */
2926 +                       res = ip_set_list_set(index, data, &used, *len);
2927 +               }
2928 +               if (res != 0)
2929 +                       goto done;
2930 +               else if (copylen != used) {
2931 +                       res = -EAGAIN;
2932 +                       goto done;
2933 +               }
2934 +               goto copy;
2935 +       }
2936 +       case IP_SET_OP_SAVE: {
2937 +               struct ip_set_req_list *req_save
2938 +                       = (struct ip_set_req_list *) data;
2939 +               ip_set_id_t i;
2940 +               int used;
2941 +
2942 +               if (*len < sizeof(struct ip_set_req_list)) {
2943 +                       ip_set_printk("short SAVE (want >=%zu, got %d)",
2944 +                                     sizeof(struct ip_set_req_list), *len);
2945 +                       res = -EINVAL;
2946 +                       goto done;
2947 +               }
2948 +               index = req_save->index;
2949 +               if (index != IP_SET_INVALID_ID
2950 +                   && ip_set_find_byindex(index) != index) {
2951 +                       res = -ENOENT;
2952 +                       goto done;
2953 +               }
2954 +               used = 0;
2955 +               if (index == IP_SET_INVALID_ID) {
2956 +                       /* Save all sets */
2957 +                       for (i = 0; i < ip_set_max && res == 0; i++) {
2958 +                               if (ip_set_list[i] != NULL)
2959 +                                       res = ip_set_save_set(i, data, &used, *len);
2960 +                       }
2961 +               } else {
2962 +                       /* Save an individual set */
2963 +                       res = ip_set_save_set(index, data, &used, *len);
2964 +               }
2965 +               if (res == 0)
2966 +                       res = ip_set_save_bindings(index, data, &used, *len);
2967 +
2968 +               if (res != 0)
2969 +                       goto done;
2970 +               else if (copylen != used) {
2971 +                       res = -EAGAIN;
2972 +                       goto done;
2973 +               }
2974 +               goto copy;
2975 +       }
2976 +       case IP_SET_OP_RESTORE: {
2977 +               struct ip_set_req_setnames *req_restore
2978 +                       = (struct ip_set_req_setnames *) data;
2979 +               int line;
2980 +
2981 +               if (*len < sizeof(struct ip_set_req_setnames)
2982 +                   || *len != req_restore->size) {
2983 +                       ip_set_printk("invalid RESTORE (want =%zu, got %d)",
2984 +                                     req_restore->size, *len);
2985 +                       res = -EINVAL;
2986 +                       goto done;
2987 +               }
2988 +               line = ip_set_restore(data + sizeof(struct ip_set_req_setnames),
2989 +                                     req_restore->size - sizeof(struct ip_set_req_setnames));
2990 +               DP("ip_set_restore: %u", line);
2991 +               if (line != 0) {
2992 +                       res = -EAGAIN;
2993 +                       req_restore->size = line;
2994 +                       copylen = sizeof(struct ip_set_req_setnames);
2995 +                       goto copy;
2996 +               }
2997 +               goto done;
2998 +       }
2999 +       default:
3000 +               res = -EBADMSG;
3001 +               goto done;
3002 +       }       /* end of switch(op) */
3003 +
3004 +    copy:
3005 +       DP("set %s, copylen %u", index != IP_SET_INVALID_ID
3006 +                                && ip_set_list[index]
3007 +                    ? ip_set_list[index]->name
3008 +                    : ":all:", copylen);
3009 +       if (res == 0)
3010 +               res = copy_to_user(user, data, copylen);
3011 +       else
3012 +               copy_to_user(user, data, copylen);
3013 +
3014 +    done:
3015 +       up(&ip_set_app_mutex);
3016 +       vfree(data);
3017 +       if (res > 0)
3018 +               res = 0;
3019 +       DP("final result %d", res);
3020 +       return res;
3021 +}
3022 +
3023 +static struct nf_sockopt_ops so_set = {
3024 +       .pf             = PF_INET,
3025 +       .set_optmin     = SO_IP_SET,
3026 +       .set_optmax     = SO_IP_SET + 1,
3027 +       .set            = &ip_set_sockfn_set,
3028 +       .get_optmin     = SO_IP_SET,
3029 +       .get_optmax     = SO_IP_SET + 1,
3030 +       .get            = &ip_set_sockfn_get,
3031 +       .use            = 0
3032 +};
3033 +
3034 +static int max_sets, hash_size;
3035 +MODULE_PARM(max_sets, "i");
3036 +MODULE_PARM_DESC(max_sets, "maximal number of sets");
3037 +MODULE_PARM(hash_size, "i");
3038 +MODULE_PARM_DESC(hash_size, "hash size for bindings");
3039 +MODULE_LICENSE("GPL");
3040 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
3041 +MODULE_DESCRIPTION("module implementing core IP set support");
3042 +
3043 +static int __init init(void)
3044 +{
3045 +       int res;
3046 +       ip_set_id_t i;
3047 +
3048 +       get_random_bytes(&ip_set_hash_random, 4);
3049 +       if (max_sets)
3050 +               ip_set_max = max_sets;
3051 +       ip_set_list = vmalloc(sizeof(struct ip_set *) * ip_set_max);
3052 +       if (!ip_set_list) {
3053 +               printk(KERN_ERR "Unable to create ip_set_list\n");
3054 +               return -ENOMEM;
3055 +       }
3056 +       memset(ip_set_list, 0, sizeof(struct ip_set *) * ip_set_max);
3057 +       if (hash_size)
3058 +               ip_set_bindings_hash_size = hash_size;
3059 +       ip_set_hash = vmalloc(sizeof(struct list_head) * ip_set_bindings_hash_size);
3060 +       if (!ip_set_hash) {
3061 +               printk(KERN_ERR "Unable to create ip_set_hash\n");
3062 +               vfree(ip_set_list);
3063 +               return -ENOMEM;
3064 +       }
3065 +       for (i = 0; i < ip_set_bindings_hash_size; i++)
3066 +               INIT_LIST_HEAD(&ip_set_hash[i]);
3067 +
3068 +       INIT_LIST_HEAD(&set_type_list);
3069 +
3070 +       res = nf_register_sockopt(&so_set);
3071 +       if (res != 0) {
3072 +               ip_set_printk("SO_SET registry failed: %d", res);
3073 +               vfree(ip_set_list);
3074 +               vfree(ip_set_hash);
3075 +               return res;
3076 +       }
3077 +       return 0;
3078 +}
3079 +
3080 +static void __exit fini(void)
3081 +{
3082 +       /* There can't be any existing set or binding. Racy. */
3083 +       nf_unregister_sockopt(&so_set);
3084 +       vfree(ip_set_list);
3085 +       vfree(ip_set_hash);
3086 +       DP("these are the famous last words");
3087 +}
3088 +
3089 +EXPORT_SYMBOL(ip_set_register_set_type);
3090 +EXPORT_SYMBOL(ip_set_unregister_set_type);
3091 +
3092 +EXPORT_SYMBOL(ip_set_get_byname);
3093 +EXPORT_SYMBOL(ip_set_get_byindex);
3094 +EXPORT_SYMBOL(ip_set_put);
3095 +
3096 +EXPORT_SYMBOL(ip_set_addip_kernel);
3097 +EXPORT_SYMBOL(ip_set_delip_kernel);
3098 +EXPORT_SYMBOL(ip_set_testip_kernel);
3099 +
3100 +module_init(init);
3101 +module_exit(fini);
3102 --- /dev/null
3103 +++ b/net/ipv4/netfilter/ip_set_iphash.c
3104 @@ -0,0 +1,379 @@
3105 +/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
3106 + *
3107 + * This program is free software; you can redistribute it and/or modify
3108 + * it under the terms of the GNU General Public License version 2 as
3109 + * published by the Free Software Foundation.
3110 + */
3111 +
3112 +/* Kernel module implementing an ip hash set */
3113 +
3114 +#include <linux/module.h>
3115 +#include <linux/ip.h>
3116 +#include <linux/skbuff.h>
3117 +#include <linux/netfilter_ipv4/ip_tables.h>
3118 +#include <linux/netfilter_ipv4/ip_set.h>
3119 +#include <linux/errno.h>
3120 +#include <asm/uaccess.h>
3121 +#include <asm/bitops.h>
3122 +#include <asm/softirq.h>
3123 +#include <linux/spinlock.h>
3124 +#include <linux/vmalloc.h>
3125 +#include <linux/random.h>
3126 +
3127 +#include <net/ip.h>
3128 +
3129 +#include <linux/netfilter_ipv4/ip_set_malloc.h>
3130 +#include <linux/netfilter_ipv4/ip_set_iphash.h>
3131 +#include <linux/netfilter_ipv4/ip_set_jhash.h>
3132 +#include <linux/netfilter_ipv4/ip_set_prime.h>
3133 +
3134 +static inline __u32
3135 +jhash_ip(const struct ip_set_iphash *map, ip_set_ip_t ip)
3136 +{
3137 +       return jhash_1word(ip, map->initval);
3138 +}
3139 +
3140 +static inline __u32
3141 +randhash_ip(const struct ip_set_iphash *map, ip_set_ip_t ip)
3142 +{
3143 +       return (1 + ip % map->prime);
3144 +}
3145 +
3146 +static inline __u32
3147 +hash_id(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3148 +{
3149 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3150 +       __u32 jhash, randhash, id;
3151 +       u_int16_t i;
3152 +
3153 +       *hash_ip = ip & map->netmask;
3154 +       jhash = jhash_ip(map, *hash_ip);
3155 +       randhash = randhash_ip(map, *hash_ip);
3156 +       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u, %u.%u.%u.%u",
3157 +          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip), HIPQUAD(map->netmask));
3158 +
3159 +       for (i = 0; i < map->probes; i++) {
3160 +               id = (jhash + i * randhash) % map->hashsize;
3161 +               DP("hash key: %u", id);
3162 +               if (map->members[id] == *hash_ip)
3163 +                       return id;
3164 +               /* No shortcut at testing - there can be deleted
3165 +                * entries. */
3166 +       }
3167 +       return UINT_MAX;
3168 +}
3169 +
3170 +static inline int
3171 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3172 +{
3173 +       return (hash_id(set, ip, hash_ip) != UINT_MAX);
3174 +}
3175 +
3176 +static int
3177 +testip(struct ip_set *set, const void *data, size_t size,
3178 +       ip_set_ip_t *hash_ip)
3179 +{
3180 +       struct ip_set_req_iphash *req =
3181 +           (struct ip_set_req_iphash *) data;
3182 +
3183 +       if (size != sizeof(struct ip_set_req_iphash)) {
3184 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3185 +                             sizeof(struct ip_set_req_iphash),
3186 +                             size);
3187 +               return -EINVAL;
3188 +       }
3189 +       return __testip(set, req->ip, hash_ip);
3190 +}
3191 +
3192 +static int
3193 +testip_kernel(struct ip_set *set, const struct sk_buff *skb,
3194 +               u_int32_t flags, ip_set_ip_t *hash_ip)
3195 +{
3196 +       return __testip(set,
3197 +                       ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
3198 +                                               : skb->nh.iph->daddr),
3199 +                       hash_ip);
3200 +}
3201 +
3202 +static inline int
3203 +__addip(struct ip_set_iphash *map, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3204 +{
3205 +       __u32 jhash, randhash, probe;
3206 +       u_int16_t i;
3207 +
3208 +       *hash_ip = ip & map->netmask;
3209 +       jhash = jhash_ip(map, *hash_ip);
3210 +       randhash = randhash_ip(map, *hash_ip);
3211 +
3212 +       for (i = 0; i < map->probes; i++) {
3213 +               probe = (jhash + i * randhash) % map->hashsize;
3214 +               if (map->members[probe] == *hash_ip)
3215 +                       return -EEXIST;
3216 +               if (!map->members[probe]) {
3217 +                       map->members[probe] = *hash_ip;
3218 +                       return 0;
3219 +               }
3220 +       }
3221 +       /* Trigger rehashing */
3222 +       return -EAGAIN;
3223 +}
3224 +
3225 +static int
3226 +addip(struct ip_set *set, const void *data, size_t size,
3227 +        ip_set_ip_t *hash_ip)
3228 +{
3229 +       struct ip_set_req_iphash *req =
3230 +           (struct ip_set_req_iphash *) data;
3231 +
3232 +       if (size != sizeof(struct ip_set_req_iphash)) {
3233 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3234 +                             sizeof(struct ip_set_req_iphash),
3235 +                             size);
3236 +               return -EINVAL;
3237 +       }
3238 +       return __addip((struct ip_set_iphash *) set->data, req->ip, hash_ip);
3239 +}
3240 +
3241 +static int
3242 +addip_kernel(struct ip_set *set, const struct sk_buff *skb,
3243 +            u_int32_t flags, ip_set_ip_t *hash_ip)
3244 +{
3245 +       return __addip((struct ip_set_iphash *) set->data,
3246 +                      ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
3247 +                                              : skb->nh.iph->daddr),
3248 +                      hash_ip);
3249 +}
3250 +
3251 +static int retry(struct ip_set *set)
3252 +{
3253 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3254 +       ip_set_ip_t hash_ip, *members;
3255 +       u_int32_t i, hashsize;
3256 +       unsigned newbytes;
3257 +       int res;
3258 +       struct ip_set_iphash tmp = {
3259 +               .hashsize = map->hashsize,
3260 +               .probes = map->probes,
3261 +               .resize = map->resize,
3262 +               .netmask = map->netmask,
3263 +       };
3264 +
3265 +       if (map->resize == 0)
3266 +               return -ERANGE;
3267 +
3268 +    again:
3269 +       res = 0;
3270 +
3271 +       /* Calculate new parameters */
3272 +       get_random_bytes(&tmp.initval, 4);
3273 +       hashsize = tmp.hashsize + (tmp.hashsize * map->resize)/100;
3274 +       if (hashsize == tmp.hashsize)
3275 +               hashsize++;
3276 +       tmp.prime = make_prime(hashsize);
3277 +
3278 +       ip_set_printk("rehashing of set %s triggered: "
3279 +                     "hashsize grows from %u to %u",
3280 +                     set->name, tmp.hashsize, hashsize);
3281 +       tmp.hashsize = hashsize;
3282 +
3283 +       newbytes = hashsize * sizeof(ip_set_ip_t);
3284 +       tmp.members = ip_set_malloc(newbytes);
3285 +       if (!tmp.members) {
3286 +               DP("out of memory for %d bytes", newbytes);
3287 +               return -ENOMEM;
3288 +       }
3289 +       memset(tmp.members, 0, newbytes);
3290 +
3291 +       write_lock_bh(&set->lock);
3292 +       map = (struct ip_set_iphash *) set->data; /* Play safe */
3293 +       for (i = 0; i < map->hashsize && res == 0; i++) {
3294 +               if (map->members[i])
3295 +                       res = __addip(&tmp, map->members[i], &hash_ip);
3296 +       }
3297 +       if (res) {
3298 +               /* Failure, try again */
3299 +               write_unlock_bh(&set->lock);
3300 +               ip_set_free(tmp.members, newbytes);
3301 +               goto again;
3302 +       }
3303 +
3304 +       /* Success at resizing! */
3305 +       members = map->members;
3306 +       hashsize = map->hashsize;
3307 +
3308 +       map->initval = tmp.initval;
3309 +       map->prime = tmp.prime;
3310 +       map->hashsize = tmp.hashsize;
3311 +       map->members = tmp.members;
3312 +       write_unlock_bh(&set->lock);
3313 +
3314 +       ip_set_free(members, hashsize * sizeof(ip_set_ip_t));
3315 +
3316 +       return 0;
3317 +}
3318 +
3319 +static inline int
3320 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3321 +{
3322 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3323 +       ip_set_ip_t id = hash_id(set, ip, hash_ip);
3324 +
3325 +       if (id == UINT_MAX)
3326 +               return -EEXIST;
3327 +
3328 +       map->members[id] = 0;
3329 +       return 0;
3330 +}
3331 +
3332 +static int
3333 +delip(struct ip_set *set, const void *data, size_t size,
3334 +        ip_set_ip_t *hash_ip)
3335 +{
3336 +       struct ip_set_req_iphash *req =
3337 +           (struct ip_set_req_iphash *) data;
3338 +
3339 +       if (size != sizeof(struct ip_set_req_iphash)) {
3340 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3341 +                             sizeof(struct ip_set_req_iphash),
3342 +                             size);
3343 +               return -EINVAL;
3344 +       }
3345 +       return __delip(set, req->ip, hash_ip);
3346 +}
3347 +
3348 +static int
3349 +delip_kernel(struct ip_set *set, const struct sk_buff *skb,
3350 +              u_int32_t flags, ip_set_ip_t *hash_ip)
3351 +{
3352 +       return __delip(set,
3353 +                      ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
3354 +                                              : skb->nh.iph->daddr),
3355 +                      hash_ip);
3356 +}
3357 +
3358 +static int create(struct ip_set *set, const void *data, size_t size)
3359 +{
3360 +       unsigned newbytes;
3361 +       struct ip_set_req_iphash_create *req =
3362 +           (struct ip_set_req_iphash_create *) data;
3363 +       struct ip_set_iphash *map;
3364 +
3365 +       if (size != sizeof(struct ip_set_req_iphash_create)) {
3366 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3367 +                              sizeof(struct ip_set_req_iphash_create),
3368 +                              size);
3369 +               return -EINVAL;
3370 +       }
3371 +
3372 +       if (req->hashsize < 1) {
3373 +               ip_set_printk("hashsize too small");
3374 +               return -ENOEXEC;
3375 +       }
3376 +
3377 +       map = kmalloc(sizeof(struct ip_set_iphash), GFP_KERNEL);
3378 +       if (!map) {
3379 +               DP("out of memory for %d bytes",
3380 +                  sizeof(struct ip_set_iphash));
3381 +               return -ENOMEM;
3382 +       }
3383 +       get_random_bytes(&map->initval, 4);
3384 +       map->prime = make_prime(req->hashsize);
3385 +       map->hashsize = req->hashsize;
3386 +       map->probes = req->probes;
3387 +       map->resize = req->resize;
3388 +       map->netmask = req->netmask;
3389 +       newbytes = map->hashsize * sizeof(ip_set_ip_t);
3390 +       map->members = ip_set_malloc(newbytes);
3391 +       if (!map->members) {
3392 +               DP("out of memory for %d bytes", newbytes);
3393 +               kfree(map);
3394 +               return -ENOMEM;
3395 +       }
3396 +       memset(map->members, 0, newbytes);
3397 +
3398 +       set->data = map;
3399 +       return 0;
3400 +}
3401 +
3402 +static void destroy(struct ip_set *set)
3403 +{
3404 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3405 +
3406 +       ip_set_free(map->members, map->hashsize * sizeof(ip_set_ip_t));
3407 +       kfree(map);
3408 +
3409 +       set->data = NULL;
3410 +}
3411 +
3412 +static void flush(struct ip_set *set)
3413 +{
3414 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3415 +       memset(map->members, 0, map->hashsize * sizeof(ip_set_ip_t));
3416 +}
3417 +
3418 +static void list_header(const struct ip_set *set, void *data)
3419 +{
3420 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3421 +       struct ip_set_req_iphash_create *header =
3422 +           (struct ip_set_req_iphash_create *) data;
3423 +
3424 +       header->hashsize = map->hashsize;
3425 +       header->probes = map->probes;
3426 +       header->resize = map->resize;
3427 +       header->netmask = map->netmask;
3428 +}
3429 +
3430 +static int list_members_size(const struct ip_set *set)
3431 +{
3432 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3433 +
3434 +       return (map->hashsize * sizeof(ip_set_ip_t));
3435 +}
3436 +
3437 +static void list_members(const struct ip_set *set, void *data)
3438 +{
3439 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3440 +       int bytes = map->hashsize * sizeof(ip_set_ip_t);
3441 +
3442 +       memcpy(data, map->members, bytes);
3443 +}
3444 +
3445 +static struct ip_set_type ip_set_iphash = {
3446 +       .typename               = SETTYPE_NAME,
3447 +       .typecode               = IPSET_TYPE_IP,
3448 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
3449 +       .create                 = &create,
3450 +       .destroy                = &destroy,
3451 +       .flush                  = &flush,
3452 +       .reqsize                = sizeof(struct ip_set_req_iphash),
3453 +       .addip                  = &addip,
3454 +       .addip_kernel           = &addip_kernel,
3455 +       .retry                  = &retry,
3456 +       .delip                  = &delip,
3457 +       .delip_kernel           = &delip_kernel,
3458 +       .testip                 = &testip,
3459 +       .testip_kernel          = &testip_kernel,
3460 +       .header_size            = sizeof(struct ip_set_req_iphash_create),
3461 +       .list_header            = &list_header,
3462 +       .list_members_size      = &list_members_size,
3463 +       .list_members           = &list_members,
3464 +       .me                     = THIS_MODULE,
3465 +};
3466 +
3467 +MODULE_LICENSE("GPL");
3468 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
3469 +MODULE_DESCRIPTION("iphash type of IP sets");
3470 +
3471 +static int __init init(void)
3472 +{
3473 +       return ip_set_register_set_type(&ip_set_iphash);
3474 +}
3475 +
3476 +static void __exit fini(void)
3477 +{
3478 +       /* FIXME: possible race with ip_set_create() */
3479 +       ip_set_unregister_set_type(&ip_set_iphash);
3480 +}
3481 +
3482 +module_init(init);
3483 +module_exit(fini);
3484 --- /dev/null
3485 +++ b/net/ipv4/netfilter/ip_set_ipmap.c
3486 @@ -0,0 +1,314 @@
3487 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
3488 + *                         Patrick Schaaf <bof@bof.de>
3489 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
3490 + *
3491 + * This program is free software; you can redistribute it and/or modify
3492 + * it under the terms of the GNU General Public License version 2 as
3493 + * published by the Free Software Foundation.
3494 + */
3495 +
3496 +/* Kernel module implementing an IP set type: the single bitmap type */
3497 +
3498 +#include <linux/module.h>
3499 +#include <linux/ip.h>
3500 +#include <linux/skbuff.h>
3501 +#include <linux/netfilter_ipv4/ip_tables.h>
3502 +#include <linux/netfilter_ipv4/ip_set.h>
3503 +#include <linux/errno.h>
3504 +#include <asm/uaccess.h>
3505 +#include <asm/bitops.h>
3506 +#include <asm/softirq.h>
3507 +#include <linux/spinlock.h>
3508 +
3509 +#include <linux/netfilter_ipv4/ip_set_ipmap.h>
3510 +
3511 +static inline ip_set_ip_t
3512 +ip_to_id(const struct ip_set_ipmap *map, ip_set_ip_t ip)
3513 +{
3514 +       return (ip - map->first_ip)/map->hosts;
3515 +}
3516 +
3517 +static inline int
3518 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3519 +{
3520 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3521 +
3522 +       if (ip < map->first_ip || ip > map->last_ip)
3523 +               return -ERANGE;
3524 +
3525 +       *hash_ip = ip & map->netmask;
3526 +       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
3527 +          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip));
3528 +       return !!test_bit(ip_to_id(map, *hash_ip), map->members);
3529 +}
3530 +
3531 +static int
3532 +testip(struct ip_set *set, const void *data, size_t size,
3533 +       ip_set_ip_t *hash_ip)
3534 +{
3535 +       struct ip_set_req_ipmap *req =
3536 +           (struct ip_set_req_ipmap *) data;
3537 +
3538 +       if (size != sizeof(struct ip_set_req_ipmap)) {
3539 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3540 +                             sizeof(struct ip_set_req_ipmap),
3541 +                             size);
3542 +               return -EINVAL;
3543 +       }
3544 +       return __testip(set, req->ip, hash_ip);
3545 +}
3546 +
3547 +static int
3548 +testip_kernel(struct ip_set *set,
3549 +             const struct sk_buff *skb,
3550 +             u_int32_t flags,
3551 +             ip_set_ip_t *hash_ip)
3552 +{
3553 +       int res;
3554 +
3555 +       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
3556 +          flags & IPSET_SRC ? "SRC" : "DST",
3557 +          NIPQUAD(skb->nh.iph->saddr),
3558 +          NIPQUAD(skb->nh.iph->daddr));
3559 +
3560 +       res =  __testip(set,
3561 +                       ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
3562 +                                               : skb->nh.iph->daddr),
3563 +                       hash_ip);
3564 +       return (res < 0 ? 0 : res);
3565 +}
3566 +
3567 +static inline int
3568 +__addip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3569 +{
3570 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3571 +
3572 +       if (ip < map->first_ip || ip > map->last_ip)
3573 +               return -ERANGE;
3574 +
3575 +       *hash_ip = ip & map->netmask;
3576 +       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
3577 +       if (test_and_set_bit(ip_to_id(map, *hash_ip), map->members))
3578 +               return -EEXIST;
3579 +
3580 +       return 0;
3581 +}
3582 +
3583 +static int
3584 +addip(struct ip_set *set, const void *data, size_t size,
3585 +      ip_set_ip_t *hash_ip)
3586 +{
3587 +       struct ip_set_req_ipmap *req =
3588 +           (struct ip_set_req_ipmap *) data;
3589 +
3590 +       if (size != sizeof(struct ip_set_req_ipmap)) {
3591 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3592 +                             sizeof(struct ip_set_req_ipmap),
3593 +                             size);
3594 +               return -EINVAL;
3595 +       }
3596 +       DP("%u.%u.%u.%u", HIPQUAD(req->ip));
3597 +       return __addip(set, req->ip, hash_ip);
3598 +}
3599 +
3600 +static int
3601 +addip_kernel(struct ip_set *set, const struct sk_buff *skb,
3602 +            u_int32_t flags, ip_set_ip_t *hash_ip)
3603 +{
3604 +       return __addip(set,
3605 +                      ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
3606 +                                              : skb->nh.iph->daddr),
3607 +                      hash_ip);
3608 +}
3609 +
3610 +static inline int
3611 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3612 +{
3613 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3614 +
3615 +       if (ip < map->first_ip || ip > map->last_ip)
3616 +               return -ERANGE;
3617 +
3618 +       *hash_ip = ip & map->netmask;
3619 +       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
3620 +       if (!test_and_clear_bit(ip_to_id(map, *hash_ip), map->members))
3621 +               return -EEXIST;
3622 +
3623 +       return 0;
3624 +}
3625 +
3626 +static int
3627 +delip(struct ip_set *set, const void *data, size_t size,
3628 +      ip_set_ip_t *hash_ip)
3629 +{
3630 +       struct ip_set_req_ipmap *req =
3631 +           (struct ip_set_req_ipmap *) data;
3632 +
3633 +       if (size != sizeof(struct ip_set_req_ipmap)) {
3634 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3635 +                             sizeof(struct ip_set_req_ipmap),
3636 +                             size);
3637 +               return -EINVAL;
3638 +       }
3639 +       return __delip(set, req->ip, hash_ip);
3640 +}
3641 +
3642 +static int
3643 +delip_kernel(struct ip_set *set, const struct sk_buff *skb,
3644 +            u_int32_t flags, ip_set_ip_t *hash_ip)
3645 +{
3646 +       return __delip(set,
3647 +                      ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
3648 +                                              : skb->nh.iph->daddr),
3649 +                      hash_ip);
3650 +}
3651 +
3652 +static int create(struct ip_set *set, const void *data, size_t size)
3653 +{
3654 +       int newbytes;
3655 +       struct ip_set_req_ipmap_create *req =
3656 +           (struct ip_set_req_ipmap_create *) data;
3657 +       struct ip_set_ipmap *map;
3658 +
3659 +       if (size != sizeof(struct ip_set_req_ipmap_create)) {
3660 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3661 +                             sizeof(struct ip_set_req_ipmap_create),
3662 +                             size);
3663 +               return -EINVAL;
3664 +       }
3665 +
3666 +       DP("from %u.%u.%u.%u to %u.%u.%u.%u",
3667 +          HIPQUAD(req->from), HIPQUAD(req->to));
3668 +
3669 +       if (req->from > req->to) {
3670 +               DP("bad ip range");
3671 +               return -ENOEXEC;
3672 +       }
3673 +
3674 +       if (req->to - req->from > MAX_RANGE) {
3675 +               ip_set_printk("range too big (max %d addresses)",
3676 +                              MAX_RANGE);
3677 +               return -ENOEXEC;
3678 +       }
3679 +
3680 +       map = kmalloc(sizeof(struct ip_set_ipmap), GFP_KERNEL);
3681 +       if (!map) {
3682 +               DP("out of memory for %d bytes",
3683 +                  sizeof(struct ip_set_ipmap));
3684 +               return -ENOMEM;
3685 +       }
3686 +       map->first_ip = req->from;
3687 +       map->last_ip = req->to;
3688 +       map->netmask = req->netmask;
3689 +
3690 +       if (req->netmask == 0xFFFFFFFF) {
3691 +               map->hosts = 1;
3692 +               map->sizeid = map->last_ip - map->first_ip + 1;
3693 +       } else {
3694 +               unsigned int mask_bits, netmask_bits;
3695 +               ip_set_ip_t mask;
3696 +
3697 +               map->first_ip &= map->netmask;  /* Should we better bark? */
3698 +
3699 +               mask = range_to_mask(map->first_ip, map->last_ip, &mask_bits);
3700 +               netmask_bits = mask_to_bits(map->netmask);
3701 +
3702 +               if (!mask || netmask_bits <= mask_bits)
3703 +                       return -ENOEXEC;
3704 +
3705 +               map->hosts = 2 << (32 - netmask_bits - 1);
3706 +               map->sizeid = 2 << (netmask_bits - mask_bits - 1);
3707 +       }
3708 +       newbytes = bitmap_bytes(0, map->sizeid - 1);
3709 +       map->members = kmalloc(newbytes, GFP_KERNEL);
3710 +       if (!map->members) {
3711 +               DP("out of memory for %d bytes", newbytes);
3712 +               kfree(map);
3713 +               return -ENOMEM;
3714 +       }
3715 +       memset(map->members, 0, newbytes);
3716 +
3717 +       set->data = map;
3718 +       return 0;
3719 +}
3720 +
3721 +static void destroy(struct ip_set *set)
3722 +{
3723 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3724 +
3725 +       kfree(map->members);
3726 +       kfree(map);
3727 +
3728 +       set->data = NULL;
3729 +}
3730 +
3731 +static void flush(struct ip_set *set)
3732 +{
3733 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3734 +       memset(map->members, 0, bitmap_bytes(0, map->sizeid - 1));
3735 +}
3736 +
3737 +static void list_header(const struct ip_set *set, void *data)
3738 +{
3739 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3740 +       struct ip_set_req_ipmap_create *header =
3741 +           (struct ip_set_req_ipmap_create *) data;
3742 +
3743 +       header->from = map->first_ip;
3744 +       header->to = map->last_ip;
3745 +       header->netmask = map->netmask;
3746 +}
3747 +
3748 +static int list_members_size(const struct ip_set *set)
3749 +{
3750 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3751 +
3752 +       return bitmap_bytes(0, map->sizeid - 1);
3753 +}
3754 +
3755 +static void list_members(const struct ip_set *set, void *data)
3756 +{
3757 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3758 +       int bytes = bitmap_bytes(0, map->sizeid - 1);
3759 +
3760 +       memcpy(data, map->members, bytes);
3761 +}
3762 +
3763 +static struct ip_set_type ip_set_ipmap = {
3764 +       .typename               = SETTYPE_NAME,
3765 +       .typecode               = IPSET_TYPE_IP,
3766 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
3767 +       .create                 = &create,
3768 +       .destroy                = &destroy,
3769 +       .flush                  = &flush,
3770 +       .reqsize                = sizeof(struct ip_set_req_ipmap),
3771 +       .addip                  = &addip,
3772 +       .addip_kernel           = &addip_kernel,
3773 +       .delip                  = &delip,
3774 +       .delip_kernel           = &delip_kernel,
3775 +       .testip                 = &testip,
3776 +       .testip_kernel          = &testip_kernel,
3777 +       .header_size            = sizeof(struct ip_set_req_ipmap_create),
3778 +       .list_header            = &list_header,
3779 +       .list_members_size      = &list_members_size,
3780 +       .list_members           = &list_members,
3781 +       .me                     = THIS_MODULE,
3782 +};
3783 +
3784 +MODULE_LICENSE("GPL");
3785 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
3786 +MODULE_DESCRIPTION("ipmap type of IP sets");
3787 +
3788 +static int __init init(void)
3789 +{
3790 +       return ip_set_register_set_type(&ip_set_ipmap);
3791 +}
3792 +
3793 +static void __exit fini(void)
3794 +{
3795 +       /* FIXME: possible race with ip_set_create() */
3796 +       ip_set_unregister_set_type(&ip_set_ipmap);
3797 +}
3798 +
3799 +module_init(init);
3800 +module_exit(fini);
3801 --- /dev/null
3802 +++ b/net/ipv4/netfilter/ip_set_iptree.c
3803 @@ -0,0 +1,511 @@
3804 +/* Copyright (C) 2005 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
3805 + *
3806 + * This program is free software; you can redistribute it and/or modify
3807 + * it under the terms of the GNU General Public License version 2 as
3808 + * published by the Free Software Foundation.
3809 + */
3810 +
3811 +/* Kernel module implementing an IP set type: the iptree type */
3812 +
3813 +#include <linux/module.h>
3814 +#include <linux/ip.h>
3815 +#include <linux/skbuff.h>
3816 +#include <linux/slab.h>
3817 +#include <linux/delay.h>
3818 +#include <linux/netfilter_ipv4/ip_tables.h>
3819 +#include <linux/netfilter_ipv4/ip_set.h>
3820 +#include <linux/errno.h>
3821 +#include <asm/uaccess.h>
3822 +#include <asm/bitops.h>
3823 +#include <linux/interrupt.h>
3824 +#include <linux/spinlock.h>
3825 +
3826 +#include <linux/netfilter_ipv4/ip_set_iptree.h>
3827 +
3828 +/* Garbage collection interval in seconds: */
3829 +#define IPTREE_GC_TIME         5*60
3830 +/* Sleep so many milliseconds before trying again
3831 + * to delete the gc timer at destroying a set */
3832 +#define IPTREE_DESTROY_SLEEP   100
3833 +
3834 +static kmem_cache_t *branch_cachep;
3835 +static kmem_cache_t *leaf_cachep;
3836 +
3837 +#define ABCD(a,b,c,d,addrp) do {               \
3838 +       a = ((unsigned char *)addrp)[3];        \
3839 +       b = ((unsigned char *)addrp)[2];        \
3840 +       c = ((unsigned char *)addrp)[1];        \
3841 +       d = ((unsigned char *)addrp)[0];        \
3842 +} while (0)
3843 +
3844 +#define TESTIP_WALK(map, elem, branch) do {    \
3845 +       if ((map)->tree[elem]) {                \
3846 +               branch = (map)->tree[elem];     \
3847 +       } else                                  \
3848 +               return 0;                       \
3849 +} while (0)
3850 +
3851 +static inline int
3852 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3853 +{
3854 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
3855 +       struct ip_set_iptreeb *btree;
3856 +       struct ip_set_iptreec *ctree;
3857 +       struct ip_set_iptreed *dtree;
3858 +       unsigned char a,b,c,d;
3859 +
3860 +       *hash_ip = ip;
3861 +       ABCD(a, b, c, d, hash_ip);
3862 +       DP("%u %u %u %u timeout %u", a, b, c, d, map->timeout);
3863 +       TESTIP_WALK(map, a, btree);
3864 +       TESTIP_WALK(btree, b, ctree);
3865 +       TESTIP_WALK(ctree, c, dtree);
3866 +       DP("%lu %lu", dtree->expires[d], jiffies);
3867 +       return !!(map->timeout ? (time_after(dtree->expires[d], jiffies))
3868 +                              : dtree->expires[d]);
3869 +}
3870 +
3871 +static int
3872 +testip(struct ip_set *set, const void *data, size_t size,
3873 +       ip_set_ip_t *hash_ip)
3874 +{
3875 +       struct ip_set_req_iptree *req =
3876 +           (struct ip_set_req_iptree *) data;
3877 +
3878 +       if (size != sizeof(struct ip_set_req_iptree)) {
3879 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3880 +                             sizeof(struct ip_set_req_iptree),
3881 +                             size);
3882 +               return -EINVAL;
3883 +       }
3884 +       return __testip(set, req->ip, hash_ip);
3885 +}
3886 +
3887 +static int
3888 +testip_kernel(struct ip_set *set,
3889 +             const struct sk_buff *skb,
3890 +             u_int32_t flags,
3891 +             ip_set_ip_t *hash_ip)
3892 +{
3893 +       int res;
3894 +
3895 +       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
3896 +          flags & IPSET_SRC ? "SRC" : "DST",
3897 +          NIPQUAD(skb->nh.iph->saddr),
3898 +          NIPQUAD(skb->nh.iph->daddr));
3899 +
3900 +       res =  __testip(set,
3901 +                       ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
3902 +                                               : skb->nh.iph->daddr),
3903 +                       hash_ip);
3904 +       return (res < 0 ? 0 : res);
3905 +}
3906 +
3907 +#define ADDIP_WALK(map, elem, branch, type, cachep) do {       \
3908 +       if ((map)->tree[elem]) {                                \
3909 +               DP("found %u", elem);                           \
3910 +               branch = (map)->tree[elem];                     \
3911 +       } else {                                                \
3912 +               branch = (type *)                               \
3913 +                       kmem_cache_alloc(cachep, GFP_KERNEL);   \
3914 +               if (branch == NULL)                             \
3915 +                       return -ENOMEM;                         \
3916 +               memset(branch, 0, sizeof(*branch));             \
3917 +               (map)->tree[elem] = branch;                     \
3918 +               DP("alloc %u", elem);                           \
3919 +       }                                                       \
3920 +} while (0)
3921 +
3922 +static inline int
3923 +__addip(struct ip_set *set, ip_set_ip_t ip, unsigned int timeout,
3924 +       ip_set_ip_t *hash_ip)
3925 +{
3926 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
3927 +       struct ip_set_iptreeb *btree;
3928 +       struct ip_set_iptreec *ctree;
3929 +       struct ip_set_iptreed *dtree;
3930 +       unsigned char a,b,c,d;
3931 +       int ret = 0;
3932 +
3933 +       *hash_ip = ip;
3934 +       ABCD(a, b, c, d, hash_ip);
3935 +       DP("%u %u %u %u timeout %u", a, b, c, d, timeout);
3936 +       ADDIP_WALK(map, a, btree, struct ip_set_iptreeb, branch_cachep);
3937 +       ADDIP_WALK(btree, b, ctree, struct ip_set_iptreec, branch_cachep);
3938 +       ADDIP_WALK(ctree, c, dtree, struct ip_set_iptreed, leaf_cachep);
3939 +       if (dtree->expires[d]
3940 +           && (!map->timeout || time_after(dtree->expires[d], jiffies)))
3941 +               ret = -EEXIST;
3942 +       dtree->expires[d] = map->timeout ? (timeout * HZ + jiffies) : 1;
3943 +       DP("%u %lu", d, dtree->expires[d]);
3944 +       return ret;
3945 +}
3946 +
3947 +static int
3948 +addip(struct ip_set *set, const void *data, size_t size,
3949 +      ip_set_ip_t *hash_ip)
3950 +{
3951 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
3952 +       struct ip_set_req_iptree *req =
3953 +               (struct ip_set_req_iptree *) data;
3954 +
3955 +       if (size != sizeof(struct ip_set_req_iptree)) {
3956 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3957 +                             sizeof(struct ip_set_req_iptree),
3958 +                             size);
3959 +               return -EINVAL;
3960 +       }
3961 +       DP("%u.%u.%u.%u %u", HIPQUAD(req->ip), req->timeout);
3962 +       return __addip(set, req->ip,
3963 +                      req->timeout ? req->timeout : map->timeout,
3964 +                      hash_ip);
3965 +}
3966 +
3967 +static int
3968 +addip_kernel(struct ip_set *set, const struct sk_buff *skb,
3969 +            u_int32_t flags, ip_set_ip_t *hash_ip)
3970 +{
3971 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
3972 +
3973 +       return __addip(set,
3974 +                      ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
3975 +                                              : skb->nh.iph->daddr),
3976 +                      map->timeout,
3977 +                      hash_ip);
3978 +}
3979 +
3980 +#define DELIP_WALK(map, elem, branch) do {     \
3981 +       if ((map)->tree[elem]) {                \
3982 +               branch = (map)->tree[elem];     \
3983 +       } else                                  \
3984 +               return -EEXIST;                 \
3985 +} while (0)
3986 +
3987 +static inline int
3988 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3989 +{
3990 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
3991 +       struct ip_set_iptreeb *btree;
3992 +       struct ip_set_iptreec *ctree;
3993 +       struct ip_set_iptreed *dtree;
3994 +       unsigned char a,b,c,d;
3995 +
3996 +       *hash_ip = ip;
3997 +       ABCD(a, b, c, d, hash_ip);
3998 +       DELIP_WALK(map, a, btree);
3999 +       DELIP_WALK(btree, b, ctree);
4000 +       DELIP_WALK(ctree, c, dtree);
4001 +
4002 +       if (dtree->expires[d]) {
4003 +               dtree->expires[d] = 0;
4004 +               return 0;
4005 +       }
4006 +       return -EEXIST;
4007 +}
4008 +
4009 +static int
4010 +delip(struct ip_set *set, const void *data, size_t size,
4011 +      ip_set_ip_t *hash_ip)
4012 +{
4013 +       struct ip_set_req_iptree *req =
4014 +           (struct ip_set_req_iptree *) data;
4015 +
4016 +       if (size != sizeof(struct ip_set_req_iptree)) {
4017 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4018 +                             sizeof(struct ip_set_req_iptree),
4019 +                             size);
4020 +               return -EINVAL;
4021 +       }
4022 +       return __delip(set, req->ip, hash_ip);
4023 +}
4024 +
4025 +static int
4026 +delip_kernel(struct ip_set *set, const struct sk_buff *skb,
4027 +            u_int32_t flags, ip_set_ip_t *hash_ip)
4028 +{
4029 +       return __delip(set,
4030 +                      ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
4031 +                                              : skb->nh.iph->daddr),
4032 +                      hash_ip);
4033 +}
4034 +
4035 +#define LOOP_WALK_BEGIN(map, i, branch) \
4036 +       for (i = 0; i < 255; i++) {     \
4037 +               if (!(map)->tree[i])    \
4038 +                       continue;       \
4039 +               branch = (map)->tree[i]
4040 +
4041 +#define LOOP_WALK_END }
4042 +
4043 +static void ip_tree_gc(unsigned long ul_set)
4044 +{
4045 +       struct ip_set *set = (void *) ul_set;
4046 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4047 +       struct ip_set_iptreeb *btree;
4048 +       struct ip_set_iptreec *ctree;
4049 +       struct ip_set_iptreed *dtree;
4050 +       unsigned char a,b,c,d;
4051 +       unsigned char i,j,k;
4052 +
4053 +       i = j = k = 0;
4054 +       DP("gc: %s", set->name);
4055 +       write_lock_bh(&set->lock);
4056 +       LOOP_WALK_BEGIN(map, a, btree);
4057 +       LOOP_WALK_BEGIN(btree, b, ctree);
4058 +       LOOP_WALK_BEGIN(ctree, c, dtree);
4059 +       for (d = 0; d < 255; d++) {
4060 +               if (dtree->expires[d]) {
4061 +                       DP("gc: %u %u %u %u: expires %lu jiffies %lu",
4062 +                           a, b, c, d,
4063 +                           dtree->expires[d], jiffies);
4064 +                       if (map->timeout
4065 +                           && time_before(dtree->expires[d], jiffies))
4066 +                               dtree->expires[d] = 0;
4067 +                       else
4068 +                               k = 1;
4069 +               }
4070 +       }
4071 +       if (k == 0) {
4072 +               DP("gc: %s: leaf %u %u %u empty",
4073 +                   set->name, a, b, c);
4074 +               kmem_cache_free(leaf_cachep, dtree);
4075 +               ctree->tree[c] = NULL;
4076 +       } else {
4077 +               DP("gc: %s: leaf %u %u %u not empty",
4078 +                   set->name, a, b, c);
4079 +               j = 1;
4080 +               k = 0;
4081 +       }
4082 +       LOOP_WALK_END;
4083 +       if (j == 0) {
4084 +               DP("gc: %s: branch %u %u empty",
4085 +                   set->name, a, b);
4086 +               kmem_cache_free(branch_cachep, ctree);
4087 +               btree->tree[b] = NULL;
4088 +       } else {
4089 +               DP("gc: %s: branch %u %u not empty",
4090 +                   set->name, a, b);
4091 +               i = 1;
4092 +               j = k = 0;
4093 +       }
4094 +       LOOP_WALK_END;
4095 +       if (i == 0) {
4096 +               DP("gc: %s: branch %u empty",
4097 +                   set->name, a);
4098 +               kmem_cache_free(branch_cachep, btree);
4099 +               map->tree[a] = NULL;
4100 +       } else {
4101 +               DP("gc: %s: branch %u not empty",
4102 +                   set->name, a);
4103 +               i = j = k = 0;
4104 +       }
4105 +       LOOP_WALK_END;
4106 +       write_unlock_bh(&set->lock);
4107 +
4108 +       map->gc.expires = jiffies + map->gc_interval * HZ;
4109 +       add_timer(&map->gc);
4110 +}
4111 +
4112 +static int create(struct ip_set *set, const void *data, size_t size)
4113 +{
4114 +       struct ip_set_req_iptree_create *req =
4115 +           (struct ip_set_req_iptree_create *) data;
4116 +       struct ip_set_iptree *map;
4117 +
4118 +       if (size != sizeof(struct ip_set_req_iptree_create)) {
4119 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4120 +                             sizeof(struct ip_set_req_iptree_create),
4121 +                             size);
4122 +               return -EINVAL;
4123 +       }
4124 +
4125 +       map = kmalloc(sizeof(struct ip_set_iptree), GFP_KERNEL);
4126 +       if (!map) {
4127 +               DP("out of memory for %d bytes",
4128 +                  sizeof(struct ip_set_iptree));
4129 +               return -ENOMEM;
4130 +       }
4131 +       memset(map, 0, sizeof(*map));
4132 +       map->timeout = req->timeout;
4133 +       set->data = map;
4134 +
4135 +       /* If there is no timeout for the entries,
4136 +        * we still have to call gc because delete
4137 +        * do not clean up empty branches */
4138 +       map->gc_interval = IPTREE_GC_TIME;
4139 +       init_timer(&map->gc);
4140 +       map->gc.data = (unsigned long) set;
4141 +       map->gc.function = ip_tree_gc;
4142 +       map->gc.expires = jiffies + map->gc_interval * HZ;
4143 +       add_timer(&map->gc);
4144 +
4145 +       return 0;
4146 +}
4147 +
4148 +static void __flush(struct ip_set_iptree *map)
4149 +{
4150 +       struct ip_set_iptreeb *btree;
4151 +       struct ip_set_iptreec *ctree;
4152 +       struct ip_set_iptreed *dtree;
4153 +       unsigned int a,b,c;
4154 +
4155 +       LOOP_WALK_BEGIN(map, a, btree);
4156 +       LOOP_WALK_BEGIN(btree, b, ctree);
4157 +       LOOP_WALK_BEGIN(ctree, c, dtree);
4158 +       kmem_cache_free(leaf_cachep, dtree);
4159 +       LOOP_WALK_END;
4160 +       kmem_cache_free(branch_cachep, ctree);
4161 +       LOOP_WALK_END;
4162 +       kmem_cache_free(branch_cachep, btree);
4163 +       LOOP_WALK_END;
4164 +}
4165 +
4166 +static void destroy(struct ip_set *set)
4167 +{
4168 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4169 +
4170 +       while (!del_timer(&map->gc))
4171 +               msleep(IPTREE_DESTROY_SLEEP);
4172 +       __flush(map);
4173 +       kfree(map);
4174 +       set->data = NULL;
4175 +}
4176 +
4177 +static void flush(struct ip_set *set)
4178 +{
4179 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4180 +       unsigned int timeout = map->timeout;
4181 +
4182 +       __flush(map);
4183 +       memset(map, 0, sizeof(*map));
4184 +       map->timeout = timeout;
4185 +}
4186 +
4187 +static void list_header(const struct ip_set *set, void *data)
4188 +{
4189 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4190 +       struct ip_set_req_iptree_create *header =
4191 +           (struct ip_set_req_iptree_create *) data;
4192 +
4193 +       header->timeout = map->timeout;
4194 +}
4195 +
4196 +static int list_members_size(const struct ip_set *set)
4197 +{
4198 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4199 +       struct ip_set_iptreeb *btree;
4200 +       struct ip_set_iptreec *ctree;
4201 +       struct ip_set_iptreed *dtree;
4202 +       unsigned char a,b,c,d;
4203 +       unsigned int count = 0;
4204 +
4205 +       LOOP_WALK_BEGIN(map, a, btree);
4206 +       LOOP_WALK_BEGIN(btree, b, ctree);
4207 +       LOOP_WALK_BEGIN(ctree, c, dtree);
4208 +       for (d = 0; d < 255; d++) {
4209 +               if (dtree->expires[d]
4210 +                   && (!map->timeout || time_after(dtree->expires[d], jiffies)))
4211 +                       count++;
4212 +       }
4213 +       LOOP_WALK_END;
4214 +       LOOP_WALK_END;
4215 +       LOOP_WALK_END;
4216 +
4217 +       DP("members %u", count);
4218 +       return (count * sizeof(struct ip_set_req_iptree));
4219 +}
4220 +
4221 +static void list_members(const struct ip_set *set, void *data)
4222 +{
4223 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4224 +       struct ip_set_iptreeb *btree;
4225 +       struct ip_set_iptreec *ctree;
4226 +       struct ip_set_iptreed *dtree;
4227 +       unsigned char a,b,c,d;
4228 +       size_t offset = 0;
4229 +       struct ip_set_req_iptree *entry;
4230 +
4231 +       LOOP_WALK_BEGIN(map, a, btree);
4232 +       LOOP_WALK_BEGIN(btree, b, ctree);
4233 +       LOOP_WALK_BEGIN(ctree, c, dtree);
4234 +       for (d = 0; d < 255; d++) {
4235 +               if (dtree->expires[d]
4236 +                   && (!map->timeout || time_after(dtree->expires[d], jiffies))) {
4237 +                       entry = (struct ip_set_req_iptree *)(data + offset);
4238 +                       entry->ip = ((a << 24) | (b << 16) | (c << 8) | d);
4239 +                       entry->timeout = !map->timeout ? 0
4240 +                               : (dtree->expires[d] - jiffies)/HZ;
4241 +                       offset += sizeof(struct ip_set_req_iptree);
4242 +               }
4243 +       }
4244 +       LOOP_WALK_END;
4245 +       LOOP_WALK_END;
4246 +       LOOP_WALK_END;
4247 +}
4248 +
4249 +static struct ip_set_type ip_set_iptree = {
4250 +       .typename               = SETTYPE_NAME,
4251 +       .typecode               = IPSET_TYPE_IP,
4252 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
4253 +       .create                 = &create,
4254 +       .destroy                = &destroy,
4255 +       .flush                  = &flush,
4256 +       .reqsize                = sizeof(struct ip_set_req_iptree),
4257 +       .addip                  = &addip,
4258 +       .addip_kernel           = &addip_kernel,
4259 +       .delip                  = &delip,
4260 +       .delip_kernel           = &delip_kernel,
4261 +       .testip                 = &testip,
4262 +       .testip_kernel          = &testip_kernel,
4263 +       .header_size            = sizeof(struct ip_set_req_iptree_create),
4264 +       .list_header            = &list_header,
4265 +       .list_members_size      = &list_members_size,
4266 +       .list_members           = &list_members,
4267 +       .me                     = THIS_MODULE,
4268 +};
4269 +
4270 +MODULE_LICENSE("GPL");
4271 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
4272 +MODULE_DESCRIPTION("iptree type of IP sets");
4273 +
4274 +static int __init init(void)
4275 +{
4276 +       int ret;
4277 +
4278 +       branch_cachep = kmem_cache_create("ip_set_iptreeb",
4279 +                               sizeof(struct ip_set_iptreeb),
4280 +                               0, 0, NULL, NULL);
4281 +       if (!branch_cachep) {
4282 +               printk(KERN_ERR "Unable to create ip_set_iptreeb slab cache\n");
4283 +               ret = -ENOMEM;
4284 +               goto out;
4285 +       }
4286 +       leaf_cachep = kmem_cache_create("ip_set_iptreed",
4287 +                               sizeof(struct ip_set_iptreed),
4288 +                               0, 0, NULL, NULL);
4289 +       if (!leaf_cachep) {
4290 +               printk(KERN_ERR "Unable to create ip_set_iptreed slab cache\n");
4291 +               ret = -ENOMEM;
4292 +               goto free_branch;
4293 +       }
4294 +       ret = ip_set_register_set_type(&ip_set_iptree);
4295 +       if (ret == 0)
4296 +               goto out;
4297 +
4298 +       kmem_cache_destroy(leaf_cachep);
4299 +    free_branch:
4300 +       kmem_cache_destroy(branch_cachep);
4301 +    out:
4302 +       return ret;
4303 +}
4304 +
4305 +static void __exit fini(void)
4306 +{
4307 +       /* FIXME: possible race with ip_set_create() */
4308 +       ip_set_unregister_set_type(&ip_set_iptree);
4309 +       kmem_cache_destroy(leaf_cachep);
4310 +       kmem_cache_destroy(branch_cachep);
4311 +}
4312 +
4313 +module_init(init);
4314 +module_exit(fini);
4315 --- /dev/null
4316 +++ b/net/ipv4/netfilter/ip_set_macipmap.c
4317 @@ -0,0 +1,340 @@
4318 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
4319 + *                         Patrick Schaaf <bof@bof.de>
4320 + *                         Martin Josefsson <gandalf@wlug.westbo.se>
4321 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
4322 + *
4323 + * This program is free software; you can redistribute it and/or modify
4324 + * it under the terms of the GNU General Public License version 2 as
4325 + * published by the Free Software Foundation.
4326 + */
4327 +
4328 +/* Kernel module implementing an IP set type: the macipmap type */
4329 +
4330 +#include <linux/module.h>
4331 +#include <linux/ip.h>
4332 +#include <linux/skbuff.h>
4333 +#include <linux/netfilter_ipv4/ip_tables.h>
4334 +#include <linux/netfilter_ipv4/ip_set.h>
4335 +#include <linux/errno.h>
4336 +#include <asm/uaccess.h>
4337 +#include <asm/bitops.h>
4338 +#include <asm/softirq.h>
4339 +#include <linux/spinlock.h>
4340 +#include <linux/if_ether.h>
4341 +#include <linux/vmalloc.h>
4342 +
4343 +#include <linux/netfilter_ipv4/ip_set_malloc.h>
4344 +#include <linux/netfilter_ipv4/ip_set_macipmap.h>
4345 +
4346 +static int
4347 +testip(struct ip_set *set, const void *data, size_t size, ip_set_ip_t *hash_ip)
4348 +{
4349 +       struct ip_set_macipmap *map = (struct ip_set_macipmap *) set->data;
4350 +       struct ip_set_macip *table = (struct ip_set_macip *) map->members;
4351 +       struct ip_set_req_macipmap *req = (struct ip_set_req_macipmap *) data;
4352 +
4353 +       if (size != sizeof(struct ip_set_req_macipmap)) {
4354 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4355 +                             sizeof(struct ip_set_req_macipmap),
4356 +                             size);
4357 +               return -EINVAL;
4358 +       }
4359 +
4360 +       if (req->ip < map->first_ip || req->ip > map->last_ip)
4361 +               return -ERANGE;
4362 +
4363 +       *hash_ip = req->ip;
4364 +       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
4365 +          set->name, HIPQUAD(req->ip), HIPQUAD(*hash_ip));
4366 +       if (test_bit(IPSET_MACIP_ISSET,
4367 +                    &table[req->ip - map->first_ip].flags)) {
4368 +               return (memcmp(req->ethernet,
4369 +                              &table[req->ip - map->first_ip].ethernet,
4370 +                              ETH_ALEN) == 0);
4371 +       } else {
4372 +               return (map->flags & IPSET_MACIP_MATCHUNSET ? 1 : 0);
4373 +       }
4374 +}
4375 +
4376 +static int
4377 +testip_kernel(struct ip_set *set, const struct sk_buff *skb,
4378 +             u_int32_t flags, ip_set_ip_t *hash_ip)
4379 +{
4380 +       struct ip_set_macipmap *map =
4381 +           (struct ip_set_macipmap *) set->data;
4382 +       struct ip_set_macip *table =
4383 +           (struct ip_set_macip *) map->members;
4384 +       ip_set_ip_t ip;
4385 +
4386 +       ip = ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
4387 +                                    : skb->nh.iph->daddr);
4388 +       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
4389 +          flags & IPSET_SRC ? "SRC" : "DST",
4390 +          NIPQUAD(skb->nh.iph->saddr),
4391 +          NIPQUAD(skb->nh.iph->daddr));
4392 +
4393 +       if (ip < map->first_ip || ip > map->last_ip)
4394 +               return 0;
4395 +
4396 +       *hash_ip = ip;
4397 +       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
4398 +          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip));
4399 +       if (test_bit(IPSET_MACIP_ISSET, &table[ip - map->first_ip].flags)) {
4400 +               /* Is mac pointer valid?
4401 +                * If so, compare... */
4402 +               return (skb->mac.raw >= skb->head
4403 +                       && (skb->mac.raw + ETH_HLEN) <= skb->data
4404 +                       && (memcmp(skb->mac.ethernet->h_source,
4405 +                                  &table[ip - map->first_ip].ethernet,
4406 +                                  ETH_ALEN) == 0));
4407 +       } else {
4408 +               return (map->flags & IPSET_MACIP_MATCHUNSET ? 1 : 0);
4409 +       }
4410 +}
4411 +
4412 +/* returns 0 on success */
4413 +static inline int
4414 +__addip(struct ip_set *set,
4415 +       ip_set_ip_t ip, unsigned char *ethernet, ip_set_ip_t *hash_ip)
4416 +{
4417 +       struct ip_set_macipmap *map =
4418 +           (struct ip_set_macipmap *) set->data;
4419 +       struct ip_set_macip *table =
4420 +           (struct ip_set_macip *) map->members;
4421 +
4422 +       if (ip < map->first_ip || ip > map->last_ip)
4423 +               return -ERANGE;
4424 +       if (test_and_set_bit(IPSET_MACIP_ISSET,
4425 +                            &table[ip - map->first_ip].flags))
4426 +               return -EEXIST;
4427 +
4428 +       *hash_ip = ip;
4429 +       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
4430 +       memcpy(&table[ip - map->first_ip].ethernet, ethernet, ETH_ALEN);
4431 +       return 0;
4432 +}
4433 +
4434 +static int
4435 +addip(struct ip_set *set, const void *data, size_t size,
4436 +      ip_set_ip_t *hash_ip)
4437 +{
4438 +       struct ip_set_req_macipmap *req =
4439 +           (struct ip_set_req_macipmap *) data;
4440 +
4441 +       if (size != sizeof(struct ip_set_req_macipmap)) {
4442 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4443 +                             sizeof(struct ip_set_req_macipmap),
4444 +                             size);
4445 +               return -EINVAL;
4446 +       }
4447 +       return __addip(set, req->ip, req->ethernet, hash_ip);
4448 +}
4449 +
4450 +static int
4451 +addip_kernel(struct ip_set *set, const struct sk_buff *skb,
4452 +            u_int32_t flags, ip_set_ip_t *hash_ip)
4453 +{
4454 +       ip_set_ip_t ip;
4455 +
4456 +       ip = ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
4457 +                                    : skb->nh.iph->daddr);
4458 +
4459 +       if (!(skb->mac.raw >= skb->head
4460 +             && (skb->mac.raw + ETH_HLEN) <= skb->data))
4461 +               return -EINVAL;
4462 +
4463 +       return __addip(set, ip, skb->mac.ethernet->h_source, hash_ip);
4464 +}
4465 +
4466 +static inline int
4467 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
4468 +{
4469 +       struct ip_set_macipmap *map =
4470 +           (struct ip_set_macipmap *) set->data;
4471 +       struct ip_set_macip *table =
4472 +           (struct ip_set_macip *) map->members;
4473 +
4474 +       if (ip < map->first_ip || ip > map->last_ip)
4475 +               return -ERANGE;
4476 +       if (!test_and_clear_bit(IPSET_MACIP_ISSET,
4477 +                               &table[ip - map->first_ip].flags))
4478 +               return -EEXIST;
4479 +
4480 +       *hash_ip = ip;
4481 +       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
4482 +       return 0;
4483 +}
4484 +
4485 +static int
4486 +delip(struct ip_set *set, const void *data, size_t size,
4487 +     ip_set_ip_t *hash_ip)
4488 +{
4489 +       struct ip_set_req_macipmap *req =
4490 +           (struct ip_set_req_macipmap *) data;
4491 +
4492 +       if (size != sizeof(struct ip_set_req_macipmap)) {
4493 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4494 +                             sizeof(struct ip_set_req_macipmap),
4495 +                             size);
4496 +               return -EINVAL;
4497 +       }
4498 +       return __delip(set, req->ip, hash_ip);
4499 +}
4500 +
4501 +static int
4502 +delip_kernel(struct ip_set *set, const struct sk_buff *skb,
4503 +            u_int32_t flags, ip_set_ip_t *hash_ip)
4504 +{
4505 +       return __delip(set,
4506 +                      ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
4507 +                                              : skb->nh.iph->daddr),
4508 +                      hash_ip);
4509 +}
4510 +
4511 +static inline size_t members_size(ip_set_id_t from, ip_set_id_t to)
4512 +{
4513 +       return (size_t)((to - from + 1) * sizeof(struct ip_set_macip));
4514 +}
4515 +
4516 +static int create(struct ip_set *set, const void *data, size_t size)
4517 +{
4518 +       int newbytes;
4519 +       struct ip_set_req_macipmap_create *req =
4520 +           (struct ip_set_req_macipmap_create *) data;
4521 +       struct ip_set_macipmap *map;
4522 +
4523 +       if (size != sizeof(struct ip_set_req_macipmap_create)) {
4524 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4525 +                             sizeof(struct ip_set_req_macipmap_create),
4526 +                             size);
4527 +               return -EINVAL;
4528 +       }
4529 +
4530 +       DP("from %u.%u.%u.%u to %u.%u.%u.%u",
4531 +          HIPQUAD(req->from), HIPQUAD(req->to));
4532 +
4533 +       if (req->from > req->to) {
4534 +               DP("bad ip range");
4535 +               return -ENOEXEC;
4536 +       }
4537 +
4538 +       if (req->to - req->from > MAX_RANGE) {
4539 +               ip_set_printk("range too big (max %d addresses)",
4540 +                              MAX_RANGE);
4541 +               return -ENOEXEC;
4542 +       }
4543 +
4544 +       map = kmalloc(sizeof(struct ip_set_macipmap), GFP_KERNEL);
4545 +       if (!map) {
4546 +               DP("out of memory for %d bytes",
4547 +                  sizeof(struct ip_set_macipmap));
4548 +               return -ENOMEM;
4549 +       }
4550 +       map->flags = req->flags;
4551 +       map->first_ip = req->from;
4552 +       map->last_ip = req->to;
4553 +       newbytes = members_size(map->first_ip, map->last_ip);
4554 +       map->members = ip_set_malloc(newbytes);
4555 +       DP("members: %u %p", newbytes, map->members);
4556 +       if (!map->members) {
4557 +               DP("out of memory for %d bytes", newbytes);
4558 +               kfree(map);
4559 +               return -ENOMEM;
4560 +       }
4561 +       memset(map->members, 0, newbytes);
4562 +
4563 +       set->data = map;
4564 +       return 0;
4565 +}
4566 +
4567 +static void destroy(struct ip_set *set)
4568 +{
4569 +       struct ip_set_macipmap *map =
4570 +           (struct ip_set_macipmap *) set->data;
4571 +
4572 +       ip_set_free(map->members, members_size(map->first_ip, map->last_ip));
4573 +       kfree(map);
4574 +
4575 +       set->data = NULL;
4576 +}
4577 +
4578 +static void flush(struct ip_set *set)
4579 +{
4580 +       struct ip_set_macipmap *map =
4581 +           (struct ip_set_macipmap *) set->data;
4582 +       memset(map->members, 0, members_size(map->first_ip, map->last_ip));
4583 +}
4584 +
4585 +static void list_header(const struct ip_set *set, void *data)
4586 +{
4587 +       struct ip_set_macipmap *map =
4588 +           (struct ip_set_macipmap *) set->data;
4589 +       struct ip_set_req_macipmap_create *header =
4590 +           (struct ip_set_req_macipmap_create *) data;
4591 +
4592 +       DP("list_header %x %x %u", map->first_ip, map->last_ip,
4593 +          map->flags);
4594 +
4595 +       header->from = map->first_ip;
4596 +       header->to = map->last_ip;
4597 +       header->flags = map->flags;
4598 +}
4599 +
4600 +static int list_members_size(const struct ip_set *set)
4601 +{
4602 +       struct ip_set_macipmap *map =
4603 +           (struct ip_set_macipmap *) set->data;
4604 +
4605 +       DP("%u", members_size(map->first_ip, map->last_ip));
4606 +       return members_size(map->first_ip, map->last_ip);
4607 +}
4608 +
4609 +static void list_members(const struct ip_set *set, void *data)
4610 +{
4611 +       struct ip_set_macipmap *map =
4612 +           (struct ip_set_macipmap *) set->data;
4613 +
4614 +       int bytes = members_size(map->first_ip, map->last_ip);
4615 +
4616 +       DP("members: %u %p", bytes, map->members);
4617 +       memcpy(data, map->members, bytes);
4618 +}
4619 +
4620 +static struct ip_set_type ip_set_macipmap = {
4621 +       .typename               = SETTYPE_NAME,
4622 +       .typecode               = IPSET_TYPE_IP,
4623 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
4624 +       .create                 = &create,
4625 +       .destroy                = &destroy,
4626 +       .flush                  = &flush,
4627 +       .reqsize                = sizeof(struct ip_set_req_macipmap),
4628 +       .addip                  = &addip,
4629 +       .addip_kernel           = &addip_kernel,
4630 +       .delip                  = &delip,
4631 +       .delip_kernel           = &delip_kernel,
4632 +       .testip                 = &testip,
4633 +       .testip_kernel          = &testip_kernel,
4634 +       .header_size            = sizeof(struct ip_set_req_macipmap_create),
4635 +       .list_header            = &list_header,
4636 +       .list_members_size      = &list_members_size,
4637 +       .list_members           = &list_members,
4638 +       .me                     = THIS_MODULE,
4639 +};
4640 +
4641 +MODULE_LICENSE("GPL");
4642 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
4643 +MODULE_DESCRIPTION("macipmap type of IP sets");
4644 +
4645 +static int __init init(void)
4646 +{
4647 +       return ip_set_register_set_type(&ip_set_macipmap);
4648 +}
4649 +
4650 +static void __exit fini(void)
4651 +{
4652 +       /* FIXME: possible race with ip_set_create() */
4653 +       ip_set_unregister_set_type(&ip_set_macipmap);
4654 +}
4655 +
4656 +module_init(init);
4657 +module_exit(fini);
4658 --- /dev/null
4659 +++ b/net/ipv4/netfilter/ip_set_nethash.c
4660 @@ -0,0 +1,450 @@
4661 +/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
4662 + *
4663 + * This program is free software; you can redistribute it and/or modify
4664 + * it under the terms of the GNU General Public License version 2 as
4665 + * published by the Free Software Foundation.
4666 + */
4667 +
4668 +/* Kernel module implementing a cidr nethash set */
4669 +
4670 +#include <linux/module.h>
4671 +#include <linux/ip.h>
4672 +#include <linux/skbuff.h>
4673 +#include <linux/netfilter_ipv4/ip_tables.h>
4674 +#include <linux/netfilter_ipv4/ip_set.h>
4675 +#include <linux/errno.h>
4676 +#include <asm/uaccess.h>
4677 +#include <asm/bitops.h>
4678 +#include <asm/softirq.h>
4679 +#include <linux/spinlock.h>
4680 +#include <linux/vmalloc.h>
4681 +#include <linux/random.h>
4682 +
4683 +#include <net/ip.h>
4684 +
4685 +#include <linux/netfilter_ipv4/ip_set_malloc.h>
4686 +#include <linux/netfilter_ipv4/ip_set_nethash.h>
4687 +#include <linux/netfilter_ipv4/ip_set_jhash.h>
4688 +#include <linux/netfilter_ipv4/ip_set_prime.h>
4689 +
4690 +static inline __u32
4691 +jhash_ip(const struct ip_set_nethash *map, ip_set_ip_t ip)
4692 +{
4693 +       return jhash_1word(ip, map->initval);
4694 +}
4695 +
4696 +static inline __u32
4697 +randhash_ip(const struct ip_set_nethash *map, ip_set_ip_t ip)
4698 +{
4699 +       return (1 + ip % map->prime);
4700 +}
4701 +
4702 +static inline __u32
4703 +hash_id_cidr(struct ip_set_nethash *map,
4704 +            ip_set_ip_t ip,
4705 +            unsigned char cidr,
4706 +            ip_set_ip_t *hash_ip)
4707 +{
4708 +       __u32 jhash, randhash, id;
4709 +       u_int16_t i;
4710 +
4711 +       *hash_ip = pack(ip, cidr);
4712 +       jhash = jhash_ip(map, *hash_ip);
4713 +       randhash = randhash_ip(map, *hash_ip);
4714 +
4715 +       for (i = 0; i < map->probes; i++) {
4716 +               id = (jhash + i * randhash) % map->hashsize;
4717 +               DP("hash key: %u", id);
4718 +               if (map->members[id] == *hash_ip)
4719 +                       return id;
4720 +       }
4721 +       return UINT_MAX;
4722 +}
4723 +
4724 +static inline __u32
4725 +hash_id(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
4726 +{
4727 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
4728 +       __u32 id = UINT_MAX;
4729 +       int i;
4730 +
4731 +       for (i = 0; i < 30 && map->cidr[i]; i++) {
4732 +               id = hash_id_cidr(map, ip, map->cidr[i], hash_ip);
4733 +               if (id != UINT_MAX)
4734 +                       break;
4735 +       }
4736 +       return id;
4737 +}
4738 +
4739 +static inline int
4740 +__testip_cidr(struct ip_set *set, ip_set_ip_t ip, unsigned char cidr,
4741 +             ip_set_ip_t *hash_ip)
4742 +{
4743 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
4744 +
4745 +       return (hash_id_cidr(map, ip, cidr, hash_ip) != UINT_MAX);
4746 +}
4747 +
4748 +static inline int
4749 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
4750 +{
4751 +       return (hash_id(set, ip, hash_ip) != UINT_MAX);
4752 +}
4753 +
4754 +static int
4755 +testip(struct ip_set *set, const void *data, size_t size,
4756 +       ip_set_ip_t *hash_ip)
4757 +{
4758 +       struct ip_set_req_nethash *req =
4759 +           (struct ip_set_req_nethash *) data;
4760 +
4761 +       if (size != sizeof(struct ip_set_req_nethash)) {
4762 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4763 +                             sizeof(struct ip_set_req_nethash),
4764 +                             size);
4765 +               return -EINVAL;
4766 +       }
4767 +       return (req->cidr == 32 ? __testip(set, req->ip, hash_ip)
4768 +               : __testip_cidr(set, req->ip, req->cidr, hash_ip));
4769 +}
4770 +
4771 +static int
4772 +testip_kernel(struct ip_set *set, const struct sk_buff *skb,
4773 +               u_int32_t flags, ip_set_ip_t *hash_ip)
4774 +{
4775 +       return __testip(set,
4776 +                       ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
4777 +                                               : skb->nh.iph->daddr),
4778 +                       hash_ip);
4779 +}
4780 +
4781 +static inline int
4782 +__addip_base(struct ip_set_nethash *map, ip_set_ip_t ip)
4783 +{
4784 +       __u32 jhash, randhash, probe;
4785 +       u_int16_t i;
4786 +
4787 +       jhash = jhash_ip(map, ip);
4788 +       randhash = randhash_ip(map, ip);
4789 +
4790 +       for (i = 0; i < map->probes; i++) {
4791 +               probe = (jhash + i * randhash) % map->hashsize;
4792 +               if (map->members[probe] == ip)
4793 +                       return -EEXIST;
4794 +               if (!map->members[probe]) {
4795 +                       map->members[probe] = ip;
4796 +                       return 0;
4797 +               }
4798 +       }
4799 +       /* Trigger rehashing */
4800 +       return -EAGAIN;
4801 +}
4802 +
4803 +static inline int
4804 +__addip(struct ip_set_nethash *map, ip_set_ip_t ip, unsigned char cidr,
4805 +       ip_set_ip_t *hash_ip)
4806 +{
4807 +       *hash_ip = pack(ip, cidr);
4808 +       DP("%u.%u.%u.%u/%u, %u.%u.%u.%u", HIPQUAD(ip), cidr, HIPQUAD(*hash_ip));
4809 +
4810 +       return __addip_base(map, *hash_ip);
4811 +}
4812 +
4813 +static void
4814 +update_cidr_sizes(struct ip_set_nethash *map, unsigned char cidr)
4815 +{
4816 +       unsigned char next;
4817 +       int i;
4818 +
4819 +       for (i = 0; i < 30 && map->cidr[i]; i++) {
4820 +               if (map->cidr[i] == cidr) {
4821 +                       return;
4822 +               } else if (map->cidr[i] < cidr) {
4823 +                       next = map->cidr[i];
4824 +                       map->cidr[i] = cidr;
4825 +                       cidr = next;
4826 +               }
4827 +       }
4828 +       if (i < 30)
4829 +               map->cidr[i] = cidr;
4830 +}
4831 +
4832 +static int
4833 +addip(struct ip_set *set, const void *data, size_t size,
4834 +        ip_set_ip_t *hash_ip)
4835 +{
4836 +       struct ip_set_req_nethash *req =
4837 +           (struct ip_set_req_nethash *) data;
4838 +       int ret;
4839 +
4840 +       if (size != sizeof(struct ip_set_req_nethash)) {
4841 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4842 +                             sizeof(struct ip_set_req_nethash),
4843 +                             size);
4844 +               return -EINVAL;
4845 +       }
4846 +       ret = __addip((struct ip_set_nethash *) set->data,
4847 +                     req->ip, req->cidr, hash_ip);
4848 +
4849 +       if (ret == 0)
4850 +               update_cidr_sizes((struct ip_set_nethash *) set->data,
4851 +                                 req->cidr);
4852 +
4853 +       return ret;
4854 +}
4855 +
4856 +static int
4857 +addip_kernel(struct ip_set *set, const struct sk_buff *skb,
4858 +            u_int32_t flags, ip_set_ip_t *hash_ip)
4859 +{
4860 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
4861 +       int ret = -ERANGE;
4862 +       ip_set_ip_t ip = ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
4863 +                                                : skb->nh.iph->daddr);
4864 +
4865 +       if (map->cidr[0])
4866 +               ret = __addip(map, ip, map->cidr[0], hash_ip);
4867 +
4868 +       return ret;
4869 +}
4870 +
4871 +static int retry(struct ip_set *set)
4872 +{
4873 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
4874 +       ip_set_ip_t *members;
4875 +       u_int32_t i, hashsize;
4876 +       unsigned newbytes;
4877 +       int res;
4878 +       struct ip_set_nethash tmp = {
4879 +               .hashsize = map->hashsize,
4880 +               .probes = map->probes,
4881 +               .resize = map->resize
4882 +       };
4883 +
4884 +       if (map->resize == 0)
4885 +               return -ERANGE;
4886 +
4887 +       memcpy(tmp.cidr, map->cidr, 30 * sizeof(unsigned char));
4888 +    again:
4889 +       res = 0;
4890 +
4891 +       /* Calculate new parameters */
4892 +       get_random_bytes(&tmp.initval, 4);
4893 +       hashsize = tmp.hashsize + (tmp.hashsize * map->resize)/100;
4894 +       if (hashsize == tmp.hashsize)
4895 +               hashsize++;
4896 +       tmp.prime = make_prime(hashsize);
4897 +
4898 +       ip_set_printk("rehashing of set %s triggered: "
4899 +                     "hashsize grows from %u to %u",
4900 +                     set->name, tmp.hashsize, hashsize);
4901 +       tmp.hashsize = hashsize;
4902 +
4903 +       newbytes = hashsize * sizeof(ip_set_ip_t);
4904 +       tmp.members = ip_set_malloc(newbytes);
4905 +       if (!tmp.members) {
4906 +               DP("out of memory for %d bytes", newbytes);
4907 +               return -ENOMEM;
4908 +       }
4909 +       memset(tmp.members, 0, newbytes);
4910 +
4911 +       write_lock_bh(&set->lock);
4912 +       map = (struct ip_set_nethash *) set->data; /* Play safe */
4913 +       for (i = 0; i < map->hashsize && res == 0; i++) {
4914 +               if (map->members[i])
4915 +                       res = __addip_base(&tmp, map->members[i]);
4916 +       }
4917 +       if (res) {
4918 +               /* Failure, try again */
4919 +               write_unlock_bh(&set->lock);
4920 +               ip_set_free(tmp.members, newbytes);
4921 +               goto again;
4922 +       }
4923 +
4924 +       /* Success at resizing! */
4925 +       members = map->members;
4926 +       hashsize = map->hashsize;
4927 +
4928 +       map->initval = tmp.initval;
4929 +       map->prime = tmp.prime;
4930 +       map->hashsize = tmp.hashsize;
4931 +       map->members = tmp.members;
4932 +       write_unlock_bh(&set->lock);
4933 +
4934 +       ip_set_free(members, hashsize * sizeof(ip_set_ip_t));
4935 +
4936 +       return 0;
4937 +}
4938 +
4939 +static inline int
4940 +__delip(struct ip_set_nethash *map, ip_set_ip_t ip, unsigned char cidr,
4941 +       ip_set_ip_t *hash_ip)
4942 +{
4943 +       ip_set_ip_t id = hash_id_cidr(map, ip, cidr, hash_ip);
4944 +
4945 +       if (id == UINT_MAX)
4946 +               return -EEXIST;
4947 +
4948 +       map->members[id] = 0;
4949 +       return 0;
4950 +}
4951 +
4952 +static int
4953 +delip(struct ip_set *set, const void *data, size_t size,
4954 +        ip_set_ip_t *hash_ip)
4955 +{
4956 +       struct ip_set_req_nethash *req =
4957 +           (struct ip_set_req_nethash *) data;
4958 +
4959 +       if (size != sizeof(struct ip_set_req_nethash)) {
4960 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4961 +                             sizeof(struct ip_set_req_nethash),
4962 +                             size);
4963 +               return -EINVAL;
4964 +       }
4965 +       /* TODO: no garbage collection in map->cidr */
4966 +       return __delip((struct ip_set_nethash *) set->data,
4967 +                      req->ip, req->cidr, hash_ip);
4968 +}
4969 +
4970 +static int
4971 +delip_kernel(struct ip_set *set, const struct sk_buff *skb,
4972 +              u_int32_t flags, ip_set_ip_t *hash_ip)
4973 +{
4974 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
4975 +       int ret = -ERANGE;
4976 +       ip_set_ip_t ip = ntohl(flags & IPSET_SRC ? skb->nh.iph->saddr
4977 +                                                : skb->nh.iph->daddr);
4978 +
4979 +       if (map->cidr[0])
4980 +               ret = __delip(map, ip, map->cidr[0], hash_ip);
4981 +
4982 +       return ret;
4983 +}
4984 +
4985 +static int create(struct ip_set *set, const void *data, size_t size)
4986 +{
4987 +       unsigned newbytes;
4988 +       struct ip_set_req_nethash_create *req =
4989 +           (struct ip_set_req_nethash_create *) data;
4990 +       struct ip_set_nethash *map;
4991 +
4992 +       if (size != sizeof(struct ip_set_req_nethash_create)) {
4993 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4994 +                              sizeof(struct ip_set_req_nethash_create),
4995 +                              size);
4996 +               return -EINVAL;
4997 +       }
4998 +
4999 +       if (req->hashsize < 1) {
5000 +               ip_set_printk("hashsize too small");
5001 +               return -ENOEXEC;
5002 +       }
5003 +
5004 +       map = kmalloc(sizeof(struct ip_set_nethash), GFP_KERNEL);
5005 +       if (!map) {
5006 +               DP("out of memory for %d bytes",
5007 +                  sizeof(struct ip_set_nethash));
5008 +               return -ENOMEM;
5009 +       }
5010 +       get_random_bytes(&map->initval, 4);
5011 +       map->prime = make_prime(req->hashsize);
5012 +       map->hashsize = req->hashsize;
5013 +       map->probes = req->probes;
5014 +       map->resize = req->resize;
5015 +       memset(map->cidr, 0, 30 * sizeof(unsigned char));
5016 +       newbytes = map->hashsize * sizeof(ip_set_ip_t);
5017 +       map->members = ip_set_malloc(newbytes);
5018 +       if (!map->members) {
5019 +               DP("out of memory for %d bytes", newbytes);
5020 +               kfree(map);
5021 +               return -ENOMEM;
5022 +       }
5023 +       memset(map->members, 0, newbytes);
5024 +
5025 +       set->data = map;
5026 +       return 0;
5027 +}
5028 +
5029 +static void destroy(struct ip_set *set)
5030 +{
5031 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
5032 +
5033 +       ip_set_free(map->members, map->hashsize * sizeof(ip_set_ip_t));
5034 +       kfree(map);
5035 +
5036 +       set->data = NULL;
5037 +}
5038 +
5039 +static void flush(struct ip_set *set)
5040 +{
5041 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
5042 +       memset(map->members, 0, map->hashsize * sizeof(ip_set_ip_t));
5043 +       memset(map->cidr, 0, 30 * sizeof(unsigned char));
5044 +}
5045 +
5046 +static void list_header(const struct ip_set *set, void *data)
5047 +{
5048 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
5049 +       struct ip_set_req_nethash_create *header =
5050 +           (struct ip_set_req_nethash_create *) data;
5051 +
5052 +       header->hashsize = map->hashsize;
5053 +       header->probes = map->probes;
5054 +       header->resize = map->resize;
5055 +}
5056 +
5057 +static int list_members_size(const struct ip_set *set)
5058 +{
5059 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
5060 +
5061 +       return (map->hashsize * sizeof(ip_set_ip_t));
5062 +}
5063 +
5064 +static void list_members(const struct ip_set *set, void *data)
5065 +{
5066 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
5067 +       int bytes = map->hashsize * sizeof(ip_set_ip_t);
5068 +
5069 +       memcpy(data, map->members, bytes);
5070 +}
5071 +
5072 +static struct ip_set_type ip_set_nethash = {
5073 +       .typename               = SETTYPE_NAME,
5074 +       .typecode               = IPSET_TYPE_IP,
5075 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
5076 +       .create                 = &create,
5077 +       .destroy                = &destroy,
5078 +       .flush                  = &flush,
5079 +       .reqsize                = sizeof(struct ip_set_req_nethash),
5080 +       .addip                  = &addip,
5081 +       .addip_kernel           = &addip_kernel,
5082 +       .retry                  = &retry,
5083 +       .delip                  = &delip,
5084 +       .delip_kernel           = &delip_kernel,
5085 +       .testip                 = &testip,
5086 +       .testip_kernel          = &testip_kernel,
5087 +       .header_size            = sizeof(struct ip_set_req_nethash_create),
5088 +       .list_header            = &list_header,
5089 +       .list_members_size      = &list_members_size,
5090 +       .list_members           = &list_members,
5091 +       .me                     = THIS_MODULE,
5092 +};
5093 +
5094 +MODULE_LICENSE("GPL");
5095 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
5096 +MODULE_DESCRIPTION("nethash type of IP sets");
5097 +
5098 +static int __init init(void)
5099 +{
5100 +       return ip_set_register_set_type(&ip_set_nethash);
5101 +}
5102 +
5103 +static void __exit fini(void)
5104 +{
5105 +       /* FIXME: possible race with ip_set_create() */
5106 +       ip_set_unregister_set_type(&ip_set_nethash);
5107 +}
5108 +
5109 +module_init(init);
5110 +module_exit(fini);
5111 --- /dev/null
5112 +++ b/net/ipv4/netfilter/ip_set_portmap.c
5113 @@ -0,0 +1,321 @@
5114 +/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
5115 + *
5116 + * This program is free software; you can redistribute it and/or modify
5117 + * it under the terms of the GNU General Public License version 2 as
5118 + * published by the Free Software Foundation.
5119 + */
5120 +
5121 +/* Kernel module implementing a port set type as a bitmap */
5122 +
5123 +#include <linux/module.h>
5124 +#include <linux/ip.h>
5125 +#include <linux/tcp.h>
5126 +#include <linux/udp.h>
5127 +#include <linux/skbuff.h>
5128 +#include <linux/netfilter_ipv4/ip_tables.h>
5129 +#include <linux/netfilter_ipv4/ip_set.h>
5130 +#include <linux/errno.h>
5131 +#include <asm/uaccess.h>
5132 +#include <asm/bitops.h>
5133 +#include <asm/softirq.h>
5134 +#include <linux/spinlock.h>
5135 +
5136 +#include <net/ip.h>
5137 +
5138 +#include <linux/netfilter_ipv4/ip_set_portmap.h>
5139 +
5140 +static inline ip_set_ip_t
5141 +get_port(const struct sk_buff *skb, u_int32_t flags)
5142 +{
5143 +       struct iphdr *iph = skb->nh.iph;
5144 +       u_int16_t offset = ntohs(iph->frag_off) & IP_OFFSET;
5145 +
5146 +       switch (iph->protocol) {
5147 +       case IPPROTO_TCP: {
5148 +               struct tcphdr *tcph = (struct tcphdr *)((u_int32_t *)iph + iph->ihl);
5149 +
5150 +               /* See comments at tcp_match in ip_tables.c */
5151 +               if (offset != 0
5152 +                   || (offset == 0
5153 +                       && (skb->len - iph->ihl * 4) < sizeof(struct tcphdr)))
5154 +                       return INVALID_PORT;
5155 +
5156 +               return ntohs(flags & IPSET_SRC ?
5157 +                            tcph->source : tcph->dest);
5158 +           }
5159 +       case IPPROTO_UDP: {
5160 +               struct udphdr *udph = (struct udphdr *)((u_int32_t *)iph + iph->ihl);
5161 +
5162 +               if (offset != 0
5163 +                   || (offset == 0
5164 +                       && (skb->len - iph->ihl * 4) < sizeof(struct udphdr)))
5165 +                       return INVALID_PORT;
5166 +
5167 +               return ntohs(flags & IPSET_SRC ?
5168 +                            udph->source : udph->dest);
5169 +           }
5170 +       default:
5171 +               return INVALID_PORT;
5172 +       }
5173 +}
5174 +
5175 +static inline int
5176 +__testport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
5177 +{
5178 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
5179 +
5180 +       if (port < map->first_port || port > map->last_port)
5181 +               return -ERANGE;
5182 +
5183 +       *hash_port = port;
5184 +       DP("set: %s, port:%u, %u", set->name, port, *hash_port);
5185 +       return !!test_bit(port - map->first_port, map->members);
5186 +}
5187 +
5188 +static int
5189 +testport(struct ip_set *set, const void *data, size_t size,
5190 +         ip_set_ip_t *hash_port)
5191 +{
5192 +       struct ip_set_req_portmap *req =
5193 +           (struct ip_set_req_portmap *) data;
5194 +
5195 +       if (size != sizeof(struct ip_set_req_portmap)) {
5196 +               ip_set_printk("data length wrong (want %zu, have %zu)",
5197 +                             sizeof(struct ip_set_req_portmap),
5198 +                             size);
5199 +               return -EINVAL;
5200 +       }
5201 +       return __testport(set, req->port, hash_port);
5202 +}
5203 +
5204 +static int
5205 +testport_kernel(struct ip_set *set, const struct sk_buff *skb,
5206 +               u_int32_t flags, ip_set_ip_t *hash_port)
5207 +{
5208 +       int res;
5209 +       ip_set_ip_t port = get_port(skb, flags);
5210 +
5211 +       DP("flag %s port %u", flags & IPSET_SRC ? "SRC" : "DST", port);
5212 +       if (port == INVALID_PORT)
5213 +               return 0;
5214 +
5215 +       res =  __testport(set, port, hash_port);
5216 +
5217 +       return (res < 0 ? 0 : res);
5218 +}
5219 +
5220 +static inline int
5221 +__addport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
5222 +{
5223 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
5224 +
5225 +       if (port < map->first_port || port > map->last_port)
5226 +               return -ERANGE;
5227 +       if (test_and_set_bit(port - map->first_port, map->members))
5228 +               return -EEXIST;
5229 +
5230 +       *hash_port = port;
5231 +       DP("port %u", port);
5232 +       return 0;
5233 +}
5234 +
5235 +static int
5236 +addport(struct ip_set *set, const void *data, size_t size,
5237 +        ip_set_ip_t *hash_port)
5238 +{
5239 +       struct ip_set_req_portmap *req =
5240 +           (struct ip_set_req_portmap *) data;
5241 +
5242 +       if (size != sizeof(struct ip_set_req_portmap)) {
5243 +               ip_set_printk("data length wrong (want %zu, have %zu)",
5244 +                             sizeof(struct ip_set_req_portmap),
5245 +                             size);
5246 +               return -EINVAL;
5247 +       }
5248 +       return __addport(set, req->port, hash_port);
5249 +}
5250 +
5251 +static int
5252 +addport_kernel(struct ip_set *set, const struct sk_buff *skb,
5253 +              u_int32_t flags, ip_set_ip_t *hash_port)
5254 +{
5255 +       ip_set_ip_t port = get_port(skb, flags);
5256 +
5257 +       if (port == INVALID_PORT)
5258 +               return -EINVAL;
5259 +
5260 +       return __addport(set, port, hash_port);
5261 +}
5262 +
5263 +static inline int
5264 +__delport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
5265 +{
5266 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
5267 +
5268 +       if (port < map->first_port || port > map->last_port)
5269 +               return -ERANGE;
5270 +       if (!test_and_clear_bit(port - map->first_port, map->members))
5271 +               return -EEXIST;
5272 +
5273 +       *hash_port = port;
5274 +       DP("port %u", port);
5275 +       return 0;
5276 +}
5277 +
5278 +static int
5279 +delport(struct ip_set *set, const void *data, size_t size,
5280 +        ip_set_ip_t *hash_port)
5281 +{
5282 +       struct ip_set_req_portmap *req =
5283 +           (struct ip_set_req_portmap *) data;
5284 +
5285 +       if (size != sizeof(struct ip_set_req_portmap)) {
5286 +               ip_set_printk("data length wrong (want %zu, have %zu)",
5287 +                             sizeof(struct ip_set_req_portmap),
5288 +                             size);
5289 +               return -EINVAL;
5290 +       }
5291 +       return __delport(set, req->port, hash_port);
5292 +}
5293 +
5294 +static int
5295 +delport_kernel(struct ip_set *set, const struct sk_buff *skb,
5296 +              u_int32_t flags, ip_set_ip_t *hash_port)
5297 +{
5298 +       ip_set_ip_t port = get_port(skb, flags);
5299 +
5300 +       if (port == INVALID_PORT)
5301 +               return -EINVAL;
5302 +
5303 +       return __delport(set, port, hash_port);
5304 +}
5305 +
5306 +static int create(struct ip_set *set, const void *data, size_t size)
5307 +{
5308 +       int newbytes;
5309 +       struct ip_set_req_portmap_create *req =
5310 +           (struct ip_set_req_portmap_create *) data;
5311 +       struct ip_set_portmap *map;
5312 +
5313 +       if (size != sizeof(struct ip_set_req_portmap_create)) {
5314 +               ip_set_printk("data length wrong (want %zu, have %zu)",
5315 +                              sizeof(struct ip_set_req_portmap_create),
5316 +                              size);
5317 +               return -EINVAL;
5318 +       }
5319 +
5320 +       DP("from %u to %u", req->from, req->to);
5321 +
5322 +       if (req->from > req->to) {
5323 +               DP("bad port range");
5324 +               return -ENOEXEC;
5325 +       }
5326 +
5327 +       if (req->to - req->from > MAX_RANGE) {
5328 +               ip_set_printk("range too big (max %d ports)",
5329 +                              MAX_RANGE);
5330 +               return -ENOEXEC;
5331 +       }
5332 +
5333 +       map = kmalloc(sizeof(struct ip_set_portmap), GFP_KERNEL);
5334 +       if (!map) {
5335 +               DP("out of memory for %d bytes",
5336 +                  sizeof(struct ip_set_portmap));
5337 +               return -ENOMEM;
5338 +       }
5339 +       map->first_port = req->from;
5340 +       map->last_port = req->to;
5341 +       newbytes = bitmap_bytes(req->from, req->to);
5342 +       map->members = kmalloc(newbytes, GFP_KERNEL);
5343 +       if (!map->members) {
5344 +               DP("out of memory for %d bytes", newbytes);
5345 +               kfree(map);
5346 +               return -ENOMEM;
5347 +       }
5348 +       memset(map->members, 0, newbytes);
5349 +
5350 +       set->data = map;
5351 +       return 0;
5352 +}
5353 +
5354 +static void destroy(struct ip_set *set)
5355 +{
5356 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
5357 +
5358 +       kfree(map->members);
5359 +       kfree(map);
5360 +
5361 +       set->data = NULL;
5362 +}
5363 +
5364 +static void flush(struct ip_set *set)
5365 +{
5366 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
5367 +       memset(map->members, 0, bitmap_bytes(map->first_port, map->last_port));
5368 +}
5369 +
5370 +static void list_header(const struct ip_set *set, void *data)
5371 +{
5372 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
5373 +       struct ip_set_req_portmap_create *header =
5374 +           (struct ip_set_req_portmap_create *) data;
5375 +
5376 +       DP("list_header %u %u", map->first_port, map->last_port);
5377 +
5378 +       header->from = map->first_port;
5379 +       header->to = map->last_port;
5380 +}
5381 +
5382 +static int list_members_size(const struct ip_set *set)
5383 +{
5384 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
5385 +
5386 +       return bitmap_bytes(map->first_port, map->last_port);
5387 +}
5388 +
5389 +static void list_members(const struct ip_set *set, void *data)
5390 +{
5391 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
5392 +       int bytes = bitmap_bytes(map->first_port, map->last_port);
5393 +
5394 +       memcpy(data, map->members, bytes);
5395 +}
5396 +
5397 +static struct ip_set_type ip_set_portmap = {
5398 +       .typename               = SETTYPE_NAME,
5399 +       .typecode               = IPSET_TYPE_PORT,
5400 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
5401 +       .create                 = &create,
5402 +       .destroy                = &destroy,
5403 +       .flush                  = &flush,
5404 +       .reqsize                = sizeof(struct ip_set_req_portmap),
5405 +       .addip                  = &addport,
5406 +       .addip_kernel           = &addport_kernel,
5407 +       .delip                  = &delport,
5408 +       .delip_kernel           = &delport_kernel,
5409 +       .testip                 = &testport,
5410 +       .testip_kernel          = &testport_kernel,
5411 +       .header_size            = sizeof(struct ip_set_req_portmap_create),
5412 +       .list_header            = &list_header,
5413 +       .list_members_size      = &list_members_size,
5414 +       .list_members           = &list_members,
5415 +       .me                     = THIS_MODULE,
5416 +};
5417 +
5418 +MODULE_LICENSE("GPL");
5419 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
5420 +MODULE_DESCRIPTION("portmap type of IP sets");
5421 +
5422 +static int __init init(void)
5423 +{
5424 +       return ip_set_register_set_type(&ip_set_portmap);
5425 +}
5426 +
5427 +static void __exit fini(void)
5428 +{
5429 +       /* FIXME: possible race with ip_set_create() */
5430 +       ip_set_unregister_set_type(&ip_set_portmap);
5431 +}
5432 +
5433 +module_init(init);
5434 +module_exit(fini);
5435 --- /dev/null
5436 +++ b/net/ipv4/netfilter/ipt_set.c
5437 @@ -0,0 +1,114 @@
5438 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
5439 + *                         Patrick Schaaf <bof@bof.de>
5440 + *                         Martin Josefsson <gandalf@wlug.westbo.se>
5441 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
5442 + *
5443 + * This program is free software; you can redistribute it and/or modify
5444 + * it under the terms of the GNU General Public License version 2 as
5445 + * published by the Free Software Foundation.
5446 + */
5447 +
5448 +/* Kernel module to match an IP set. */
5449 +
5450 +#include <linux/module.h>
5451 +#include <linux/ip.h>
5452 +#include <linux/skbuff.h>
5453 +
5454 +#include <linux/netfilter_ipv4/ip_tables.h>
5455 +#include <linux/netfilter_ipv4/ip_set.h>
5456 +#include <linux/netfilter_ipv4/ipt_set.h>
5457 +
5458 +static inline int
5459 +match_set(const struct ipt_set_info *info,
5460 +         const struct sk_buff *skb,
5461 +         int inv)
5462 +{
5463 +       if (ip_set_testip_kernel(info->index, skb, info->flags))
5464 +               inv = !inv;
5465 +       return inv;
5466 +}
5467 +
5468 +static int
5469 +match(const struct sk_buff *skb,
5470 +      const struct net_device *in,
5471 +      const struct net_device *out,
5472 +      const void *matchinfo,
5473 +      int offset,
5474 +      const void *hdr,
5475 +      u_int16_t datalen,
5476 +      int *hotdrop)
5477 +{
5478 +       const struct ipt_set_info_match *info = matchinfo;
5479 +
5480 +       return match_set(&info->match_set,
5481 +                        skb,
5482 +                        info->match_set.flags[0] & IPSET_MATCH_INV);
5483 +}
5484 +
5485 +static int
5486 +checkentry(const char *tablename,
5487 +          const struct ipt_ip *ip,
5488 +          void *matchinfo,
5489 +          unsigned int matchsize,
5490 +          unsigned int hook_mask)
5491 +{
5492 +       struct ipt_set_info_match *info =
5493 +               (struct ipt_set_info_match *) matchinfo;
5494 +       ip_set_id_t index;
5495 +
5496 +       if (matchsize != IPT_ALIGN(sizeof(struct ipt_set_info_match))) {
5497 +               ip_set_printk("invalid matchsize %d", matchsize);
5498 +               return 0;
5499 +       }
5500 +
5501 +       index = ip_set_get_byindex(info->match_set.index);
5502 +
5503 +       if (index == IP_SET_INVALID_ID) {
5504 +               ip_set_printk("Cannot find set indentified by id %u to match",
5505 +                             info->match_set.index);
5506 +               return 0;       /* error */
5507 +       }
5508 +       if (info->match_set.flags[IP_SET_MAX_BINDINGS] != 0) {
5509 +               ip_set_printk("That's nasty!");
5510 +               return 0;       /* error */
5511 +       }
5512 +
5513 +       return 1;
5514 +}
5515 +
5516 +static void destroy(void *matchinfo, unsigned int matchsize)
5517 +{
5518 +       struct ipt_set_info_match *info = matchinfo;
5519 +
5520 +       if (matchsize != IPT_ALIGN(sizeof(struct ipt_set_info_match))) {
5521 +               ip_set_printk("invalid matchsize %d", matchsize);
5522 +               return;
5523 +       }
5524 +
5525 +       ip_set_put(info->match_set.index);
5526 +}
5527 +
5528 +static struct ipt_match set_match = {
5529 +       .name           = "set",
5530 +       .match          = &match,
5531 +       .checkentry     = &checkentry,
5532 +       .destroy        = &destroy,
5533 +       .me             = THIS_MODULE
5534 +};
5535 +
5536 +MODULE_LICENSE("GPL");
5537 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
5538 +MODULE_DESCRIPTION("iptables IP set match module");
5539 +
5540 +static int __init init(void)
5541 +{
5542 +       return ipt_register_match(&set_match);
5543 +}
5544 +
5545 +static void __exit fini(void)
5546 +{
5547 +       ipt_unregister_match(&set_match);
5548 +}
5549 +
5550 +module_init(init);
5551 +module_exit(fini);
5552 --- /dev/null
5553 +++ b/net/ipv4/netfilter/ipt_SET.c
5554 @@ -0,0 +1,127 @@
5555 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
5556 + *                         Patrick Schaaf <bof@bof.de>
5557 + *                         Martin Josefsson <gandalf@wlug.westbo.se>
5558 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
5559 + *
5560 + * This program is free software; you can redistribute it and/or modify
5561 + * it under the terms of the GNU General Public License version 2 as
5562 + * published by the Free Software Foundation.
5563 + */
5564 +
5565 +/* ipt_SET.c - netfilter target to manipulate IP sets */
5566 +
5567 +#include <linux/types.h>
5568 +#include <linux/ip.h>
5569 +#include <linux/timer.h>
5570 +#include <linux/module.h>
5571 +#include <linux/netfilter.h>
5572 +#include <linux/netdevice.h>
5573 +#include <linux/if.h>
5574 +#include <linux/inetdevice.h>
5575 +#include <net/protocol.h>
5576 +#include <net/checksum.h>
5577 +#include <linux/netfilter_ipv4.h>
5578 +#include <linux/netfilter_ipv4/ip_nat_rule.h>
5579 +#include <linux/netfilter_ipv4/ipt_set.h>
5580 +
5581 +static unsigned int
5582 +target(struct sk_buff **pskb,
5583 +       unsigned int hooknum,
5584 +       const struct net_device *in,
5585 +       const struct net_device *out,
5586 +       const void *targinfo,
5587 +       void *userinfo)
5588 +{
5589 +       const struct ipt_set_info_target *info = targinfo;
5590 +
5591 +       if (info->add_set.index != IP_SET_INVALID_ID)
5592 +               ip_set_addip_kernel(info->add_set.index,
5593 +                                   *pskb,
5594 +                                   info->add_set.flags);
5595 +       if (info->del_set.index != IP_SET_INVALID_ID)
5596 +               ip_set_delip_kernel(info->del_set.index,
5597 +                                   *pskb,
5598 +                                   info->del_set.flags);
5599 +
5600 +       return IPT_CONTINUE;
5601 +}
5602 +
5603 +static int
5604 +checkentry(const char *tablename,
5605 +          const struct ipt_entry *e,
5606 +          void *targinfo,
5607 +          unsigned int targinfosize, unsigned int hook_mask)
5608 +{
5609 +       struct ipt_set_info_target *info =
5610 +               (struct ipt_set_info_target *) targinfo;
5611 +       ip_set_id_t index;
5612 +
5613 +       if (targinfosize != IPT_ALIGN(sizeof(*info))) {
5614 +               DP("bad target info size %u", targinfosize);
5615 +               return 0;
5616 +       }
5617 +
5618 +       if (info->add_set.index != IP_SET_INVALID_ID) {
5619 +               index = ip_set_get_byindex(info->add_set.index);
5620 +               if (index == IP_SET_INVALID_ID) {
5621 +                       ip_set_printk("cannot find add_set index %u as target",
5622 +                                     info->add_set.index);
5623 +                       return 0;       /* error */
5624 +               }
5625 +       }
5626 +
5627 +       if (info->del_set.index != IP_SET_INVALID_ID) {
5628 +               index = ip_set_get_byindex(info->del_set.index);
5629 +               if (index == IP_SET_INVALID_ID) {
5630 +                       ip_set_printk("cannot find del_set index %u as target",
5631 +                                     info->del_set.index);
5632 +                       return 0;       /* error */
5633 +               }
5634 +       }
5635 +       if (info->add_set.flags[IP_SET_MAX_BINDINGS] != 0
5636 +           || info->del_set.flags[IP_SET_MAX_BINDINGS] != 0) {
5637 +               ip_set_printk("That's nasty!");
5638 +               return 0;       /* error */
5639 +       }
5640 +       return 1;
5641 +}
5642 +
5643 +static void destroy(void *targetinfo, unsigned int targetsize)
5644 +{
5645 +       struct ipt_set_info_target *info = targetinfo;
5646 +
5647 +       if (targetsize != IPT_ALIGN(sizeof(struct ipt_set_info_target))) {
5648 +               ip_set_printk("invalid targetsize %d", targetsize);
5649 +               return;
5650 +       }
5651 +
5652 +       if (info->add_set.index != IP_SET_INVALID_ID)
5653 +               ip_set_put(info->add_set.index);
5654 +       if (info->del_set.index != IP_SET_INVALID_ID)
5655 +               ip_set_put(info->del_set.index);
5656 +}
5657 +
5658 +static struct ipt_target SET_target = {
5659 +       .name           = "SET",
5660 +       .target         = target,
5661 +       .checkentry     = checkentry,
5662 +       .destroy        = destroy,
5663 +       .me             = THIS_MODULE
5664 +};
5665 +
5666 +MODULE_LICENSE("GPL");
5667 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
5668 +MODULE_DESCRIPTION("iptables IP set target module");
5669 +
5670 +static int __init init(void)
5671 +{
5672 +       return ipt_register_target(&SET_target);
5673 +}
5674 +
5675 +static void __exit fini(void)
5676 +{
5677 +       ipt_unregister_target(&SET_target);
5678 +}
5679 +
5680 +module_init(init);
5681 +module_exit(fini);
5682 --- a/net/ipv4/netfilter/Makefile
5683 +++ b/net/ipv4/netfilter/Makefile
5684 @@ -75,6 +75,18 @@ obj-$(CONFIG_IP_NF_NAT) += iptable_nat.o
5685  obj-$(CONFIG_IP_NF_MATCH_HELPER) += ipt_helper.o
5686  obj-$(CONFIG_IP_NF_MATCH_LIMIT) += ipt_limit.o
5687  obj-$(CONFIG_IP_NF_MATCH_MARK) += ipt_mark.o
5688 +obj-$(CONFIG_IP_NF_MATCH_SET) += ipt_set.o
5689 +obj-$(CONFIG_IP_NF_TARGET_SET) += ipt_SET.o
5690 +ifdef CONFIG_IP_NF_SET
5691 +       obj-$(CONFIG_IP_NF_SET) += ip_set.o
5692 +       export-objs += ip_set.o
5693 +endif
5694 +obj-$(CONFIG_IP_NF_SET_IPMAP) += ip_set_ipmap.o
5695 +obj-$(CONFIG_IP_NF_SET_PORTMAP) += ip_set_portmap.o
5696 +obj-$(CONFIG_IP_NF_SET_MACIPMAP) += ip_set_macipmap.o
5697 +obj-$(CONFIG_IP_NF_SET_IPHASH) += ip_set_iphash.o
5698 +obj-$(CONFIG_IP_NF_SET_NETHASH) += ip_set_nethash.o
5699 +obj-$(CONFIG_IP_NF_SET_IPTREE) += ip_set_iptree.o
5700  obj-$(CONFIG_IP_NF_MATCH_MAC) += ipt_mac.o
5701  obj-$(CONFIG_IP_NF_MATCH_IPP2P) += ipt_ipp2p.o
5702