target/linux/generic-2.6/patches-2.6.21/130-netfilter_ipset.patch

   1 --- a/include/linux/netfilter_ipv4/Kbuild
   2 +++ b/include/linux/netfilter_ipv4/Kbuild
   3 @@ -59,3 +59,14 @@
   4  unifdef-y += ip_nat_rule.h
   5  unifdef-y += ip_queue.h
   6  unifdef-y += ip_tables.h
   7 +
   8 +unifdef-y += ip_set.h
   9 +header-y  += ip_set_iphash.h
  10 +header-y  += ip_set_ipmap.h
  11 +header-y  += ip_set_ipporthash.h
  12 +unifdef-y += ip_set_iptree.h
  13 +unifdef-y += ip_set_iptreemap.h
  14 +header-y  += ip_set_jhash.h
  15 +header-y  += ip_set_macipmap.h
  16 +unifdef-y += ip_set_nethash.h
  17 +header-y  += ip_set_portmap.h
  18 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set.h
  19 ===================================================================
  20 --- /dev/null
  21 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set.h
  22 @@ -0,0 +1,498 @@
  23 +#ifndef _IP_SET_H
  24 +#define _IP_SET_H
  25 +
  26 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
  27 + *                         Patrick Schaaf <bof@bof.de>
  28 + *                         Martin Josefsson <gandalf@wlug.westbo.se>
  29 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
  30 + *
  31 + * This program is free software; you can redistribute it and/or modify
  32 + * it under the terms of the GNU General Public License version 2 as
  33 + * published by the Free Software Foundation.
  34 + */
  35 +
  36 +#if 0
  37 +#define IP_SET_DEBUG
  38 +#endif
  39 +
  40 +/*
  41 + * A sockopt of such quality has hardly ever been seen before on the open
  42 + * market!  This little beauty, hardly ever used: above 64, so it's
  43 + * traditionally used for firewalling, not touched (even once!) by the
  44 + * 2.0, 2.2 and 2.4 kernels!
  45 + *
  46 + * Comes with its own certificate of authenticity, valid anywhere in the
  47 + * Free world!
  48 + *
  49 + * Rusty, 19.4.2000
  50 + */
  51 +#define SO_IP_SET              83
  52 +
  53 +/*
  54 + * Heavily modify by Joakim Axelsson 08.03.2002
  55 + * - Made it more modulebased
  56 + *
  57 + * Additional heavy modifications by Jozsef Kadlecsik 22.02.2004
  58 + * - bindings added
  59 + * - in order to "deal with" backward compatibility, renamed to ipset
  60 + */
  61 +
  62 +/*
  63 + * Used so that the kernel module and ipset-binary can match their versions
  64 + */
  65 +#define IP_SET_PROTOCOL_VERSION 2
  66 +
  67 +#define IP_SET_MAXNAMELEN 32   /* set names and set typenames */
  68 +
  69 +/* Lets work with our own typedef for representing an IP address.
  70 + * We hope to make the code more portable, possibly to IPv6...
  71 + *
  72 + * The representation works in HOST byte order, because most set types
  73 + * will perform arithmetic operations and compare operations.
  74 + *
  75 + * For now the type is an uint32_t.
  76 + *
  77 + * Make sure to ONLY use the functions when translating and parsing
  78 + * in order to keep the host byte order and make it more portable:
  79 + *  parse_ip()
  80 + *  parse_mask()
  81 + *  parse_ipandmask()
  82 + *  ip_tostring()
  83 + * (Joakim: where are they???)
  84 + */
  85 +
  86 +typedef uint32_t ip_set_ip_t;
  87 +
  88 +/* Sets are identified by an id in kernel space. Tweak with ip_set_id_t
  89 + * and IP_SET_INVALID_ID if you want to increase the max number of sets.
  90 + */
  91 +typedef uint16_t ip_set_id_t;
  92 +
  93 +#define IP_SET_INVALID_ID      65535
  94 +
  95 +/* How deep we follow bindings */
  96 +#define IP_SET_MAX_BINDINGS    6
  97 +
  98 +/*
  99 + * Option flags for kernel operations (ipt_set_info)
 100 + */
 101 +#define IPSET_SRC              0x01    /* Source match/add */
 102 +#define IPSET_DST              0x02    /* Destination match/add */
 103 +#define IPSET_MATCH_INV                0x04    /* Inverse matching */
 104 +
 105 +/*
 106 + * Set features
 107 + */
 108 +#define IPSET_TYPE_IP          0x01    /* IP address type of set */
 109 +#define IPSET_TYPE_PORT                0x02    /* Port type of set */
 110 +#define IPSET_DATA_SINGLE      0x04    /* Single data storage */
 111 +#define IPSET_DATA_DOUBLE      0x08    /* Double data storage */
 112 +
 113 +/* Reserved keywords */
 114 +#define IPSET_TOKEN_DEFAULT    ":default:"
 115 +#define IPSET_TOKEN_ALL                ":all:"
 116 +
 117 +/* SO_IP_SET operation constants, and their request struct types.
 118 + *
 119 + * Operation ids:
 120 + *       0-99:  commands with version checking
 121 + *     100-199: add/del/test/bind/unbind
 122 + *     200-299: list, save, restore
 123 + */
 124 +
 125 +/* Single shot operations:
 126 + * version, create, destroy, flush, rename and swap
 127 + *
 128 + * Sets are identified by name.
 129 + */
 130 +
 131 +#define IP_SET_REQ_STD         \
 132 +       unsigned op;            \
 133 +       unsigned version;       \
 134 +       char name[IP_SET_MAXNAMELEN]
 135 +
 136 +#define IP_SET_OP_CREATE       0x00000001      /* Create a new (empty) set */
 137 +struct ip_set_req_create {
 138 +       IP_SET_REQ_STD;
 139 +       char typename[IP_SET_MAXNAMELEN];
 140 +};
 141 +
 142 +#define IP_SET_OP_DESTROY      0x00000002      /* Remove a (empty) set */
 143 +struct ip_set_req_std {
 144 +       IP_SET_REQ_STD;
 145 +};
 146 +
 147 +#define IP_SET_OP_FLUSH                0x00000003      /* Remove all IPs in a set */
 148 +/* Uses ip_set_req_std */
 149 +
 150 +#define IP_SET_OP_RENAME       0x00000004      /* Rename a set */
 151 +/* Uses ip_set_req_create */
 152 +
 153 +#define IP_SET_OP_SWAP         0x00000005      /* Swap two sets */
 154 +/* Uses ip_set_req_create */
 155 +
 156 +union ip_set_name_index {
 157 +       char name[IP_SET_MAXNAMELEN];
 158 +       ip_set_id_t index;
 159 +};
 160 +
 161 +#define IP_SET_OP_GET_BYNAME   0x00000006      /* Get set index by name */
 162 +struct ip_set_req_get_set {
 163 +       unsigned op;
 164 +       unsigned version;
 165 +       union ip_set_name_index set;
 166 +};
 167 +
 168 +#define IP_SET_OP_GET_BYINDEX  0x00000007      /* Get set name by index */
 169 +/* Uses ip_set_req_get_set */
 170 +
 171 +#define IP_SET_OP_VERSION      0x00000100      /* Ask kernel version */
 172 +struct ip_set_req_version {
 173 +       unsigned op;
 174 +       unsigned version;
 175 +};
 176 +
 177 +/* Double shots operations:
 178 + * add, del, test, bind and unbind.
 179 + *
 180 + * First we query the kernel to get the index and type of the target set,
 181 + * then issue the command. Validity of IP is checked in kernel in order
 182 + * to minimalize sockopt operations.
 183 + */
 184 +
 185 +/* Get minimal set data for add/del/test/bind/unbind IP */
 186 +#define IP_SET_OP_ADT_GET      0x00000010      /* Get set and type */
 187 +struct ip_set_req_adt_get {
 188 +       unsigned op;
 189 +       unsigned version;
 190 +       union ip_set_name_index set;
 191 +       char typename[IP_SET_MAXNAMELEN];
 192 +};
 193 +
 194 +#define IP_SET_REQ_BYINDEX     \
 195 +       unsigned op;            \
 196 +       ip_set_id_t index;
 197 +
 198 +struct ip_set_req_adt {
 199 +       IP_SET_REQ_BYINDEX;
 200 +};
 201 +
 202 +#define IP_SET_OP_ADD_IP       0x00000101      /* Add an IP to a set */
 203 +/* Uses ip_set_req_adt, with type specific addage */
 204 +
 205 +#define IP_SET_OP_DEL_IP       0x00000102      /* Remove an IP from a set */
 206 +/* Uses ip_set_req_adt, with type specific addage */
 207 +
 208 +#define IP_SET_OP_TEST_IP      0x00000103      /* Test an IP in a set */
 209 +/* Uses ip_set_req_adt, with type specific addage */
 210 +
 211 +#define IP_SET_OP_BIND_SET     0x00000104      /* Bind an IP to a set */
 212 +/* Uses ip_set_req_bind, with type specific addage */
 213 +struct ip_set_req_bind {
 214 +       IP_SET_REQ_BYINDEX;
 215 +       char binding[IP_SET_MAXNAMELEN];
 216 +};
 217 +
 218 +#define IP_SET_OP_UNBIND_SET   0x00000105      /* Unbind an IP from a set */
 219 +/* Uses ip_set_req_bind, with type speficic addage
 220 + * index = 0 means unbinding for all sets */
 221 +
 222 +#define IP_SET_OP_TEST_BIND_SET        0x00000106      /* Test binding an IP to a set */
 223 +/* Uses ip_set_req_bind, with type specific addage */
 224 +
 225 +/* Multiple shots operations: list, save, restore.
 226 + *
 227 + * - check kernel version and query the max number of sets
 228 + * - get the basic information on all sets
 229 + *   and size required for the next step
 230 + * - get actual set data: header, data, bindings
 231 + */
 232 +
 233 +/* Get max_sets and the index of a queried set
 234 + */
 235 +#define IP_SET_OP_MAX_SETS     0x00000020
 236 +struct ip_set_req_max_sets {
 237 +       unsigned op;
 238 +       unsigned version;
 239 +       ip_set_id_t max_sets;           /* max_sets */
 240 +       ip_set_id_t sets;               /* real number of sets */
 241 +       union ip_set_name_index set;    /* index of set if name used */
 242 +};
 243 +
 244 +/* Get the id and name of the sets plus size for next step */
 245 +#define IP_SET_OP_LIST_SIZE    0x00000201
 246 +#define IP_SET_OP_SAVE_SIZE    0x00000202
 247 +struct ip_set_req_setnames {
 248 +       unsigned op;
 249 +       ip_set_id_t index;              /* set to list/save */
 250 +       size_t size;                    /* size to get setdata/bindings */
 251 +       /* followed by sets number of struct ip_set_name_list */
 252 +};
 253 +
 254 +struct ip_set_name_list {
 255 +       char name[IP_SET_MAXNAMELEN];
 256 +       char typename[IP_SET_MAXNAMELEN];
 257 +       ip_set_id_t index;
 258 +       ip_set_id_t id;
 259 +};
 260 +
 261 +/* The actual list operation */
 262 +#define IP_SET_OP_LIST         0x00000203
 263 +struct ip_set_req_list {
 264 +       IP_SET_REQ_BYINDEX;
 265 +       /* sets number of struct ip_set_list in reply */
 266 +};
 267 +
 268 +struct ip_set_list {
 269 +       ip_set_id_t index;
 270 +       ip_set_id_t binding;
 271 +       u_int32_t ref;
 272 +       size_t header_size;     /* Set header data of header_size */
 273 +       size_t members_size;    /* Set members data of members_size */
 274 +       size_t bindings_size;   /* Set bindings data of bindings_size */
 275 +};
 276 +
 277 +struct ip_set_hash_list {
 278 +       ip_set_ip_t ip;
 279 +       ip_set_id_t binding;
 280 +};
 281 +
 282 +/* The save operation */
 283 +#define IP_SET_OP_SAVE         0x00000204
 284 +/* Uses ip_set_req_list, in the reply replaced by
 285 + * sets number of struct ip_set_save plus a marker
 286 + * ip_set_save followed by ip_set_hash_save structures.
 287 + */
 288 +struct ip_set_save {
 289 +       ip_set_id_t index;
 290 +       ip_set_id_t binding;
 291 +       size_t header_size;     /* Set header data of header_size */
 292 +       size_t members_size;    /* Set members data of members_size */
 293 +};
 294 +
 295 +/* At restoring, ip == 0 means default binding for the given set: */
 296 +struct ip_set_hash_save {
 297 +       ip_set_ip_t ip;
 298 +       ip_set_id_t id;
 299 +       ip_set_id_t binding;
 300 +};
 301 +
 302 +/* The restore operation */
 303 +#define IP_SET_OP_RESTORE      0x00000205
 304 +/* Uses ip_set_req_setnames followed by ip_set_restore structures
 305 + * plus a marker ip_set_restore, followed by ip_set_hash_save
 306 + * structures.
 307 + */
 308 +struct ip_set_restore {
 309 +       char name[IP_SET_MAXNAMELEN];
 310 +       char typename[IP_SET_MAXNAMELEN];
 311 +       ip_set_id_t index;
 312 +       size_t header_size;     /* Create data of header_size */
 313 +       size_t members_size;    /* Set members data of members_size */
 314 +};
 315 +
 316 +static inline int bitmap_bytes(ip_set_ip_t a, ip_set_ip_t b)
 317 +{
 318 +       return 4 * ((((b - a + 8) / 8) + 3) / 4);
 319 +}
 320 +
 321 +#ifdef __KERNEL__
 322 +
 323 +#define ip_set_printk(format, args...)                         \
 324 +       do {                                                    \
 325 +               printk("%s: %s: ", __FILE__, __FUNCTION__);     \
 326 +               printk(format "\n" , ## args);                  \
 327 +       } while (0)
 328 +
 329 +#if defined(IP_SET_DEBUG)
 330 +#define DP(format, args...)                                    \
 331 +       do {                                                    \
 332 +               printk("%s: %s (DBG): ", __FILE__, __FUNCTION__);\
 333 +               printk(format "\n" , ## args);                  \
 334 +       } while (0)
 335 +#define IP_SET_ASSERT(x)                                       \
 336 +       do {                                                    \
 337 +               if (!(x))                                       \
 338 +                       printk("IP_SET_ASSERT: %s:%i(%s)\n",    \
 339 +                               __FILE__, __LINE__, __FUNCTION__); \
 340 +       } while (0)
 341 +#else
 342 +#define DP(format, args...)
 343 +#define IP_SET_ASSERT(x)
 344 +#endif
 345 +
 346 +struct ip_set;
 347 +
 348 +/*
 349 + * The ip_set_type definition - one per set type, e.g. "ipmap".
 350 + *
 351 + * Each individual set has a pointer, set->type, going to one
 352 + * of these structures. Function pointers inside the structure implement
 353 + * the real behaviour of the sets.
 354 + *
 355 + * If not mentioned differently, the implementation behind the function
 356 + * pointers of a set_type, is expected to return 0 if ok, and a negative
 357 + * errno (e.g. -EINVAL) on error.
 358 + */
 359 +struct ip_set_type {
 360 +       struct list_head list;  /* next in list of set types */
 361 +
 362 +       /* test for IP in set (kernel: iptables -m set src|dst)
 363 +        * return 0 if not in set, 1 if in set.
 364 +        */
 365 +       int (*testip_kernel) (struct ip_set *set,
 366 +                             const struct sk_buff * skb,
 367 +                             ip_set_ip_t *ip,
 368 +                             const u_int32_t *flags,
 369 +                             unsigned char index);
 370 +
 371 +       /* test for IP in set (userspace: ipset -T set IP)
 372 +        * return 0 if not in set, 1 if in set.
 373 +        */
 374 +       int (*testip) (struct ip_set *set,
 375 +                      const void *data, size_t size,
 376 +                      ip_set_ip_t *ip);
 377 +
 378 +       /*
 379 +        * Size of the data structure passed by when
 380 +        * adding/deletin/testing an entry.
 381 +        */
 382 +       size_t reqsize;
 383 +
 384 +       /* Add IP into set (userspace: ipset -A set IP)
 385 +        * Return -EEXIST if the address is already in the set,
 386 +        * and -ERANGE if the address lies outside the set bounds.
 387 +        * If the address was not already in the set, 0 is returned.
 388 +        */
 389 +       int (*addip) (struct ip_set *set,
 390 +                     const void *data, size_t size,
 391 +                     ip_set_ip_t *ip);
 392 +
 393 +       /* Add IP into set (kernel: iptables ... -j SET set src|dst)
 394 +        * Return -EEXIST if the address is already in the set,
 395 +        * and -ERANGE if the address lies outside the set bounds.
 396 +        * If the address was not already in the set, 0 is returned.
 397 +        */
 398 +       int (*addip_kernel) (struct ip_set *set,
 399 +                            const struct sk_buff * skb,
 400 +                            ip_set_ip_t *ip,
 401 +                            const u_int32_t *flags,
 402 +                            unsigned char index);
 403 +
 404 +       /* remove IP from set (userspace: ipset -D set --entry x)
 405 +        * Return -EEXIST if the address is NOT in the set,
 406 +        * and -ERANGE if the address lies outside the set bounds.
 407 +        * If the address really was in the set, 0 is returned.
 408 +        */
 409 +       int (*delip) (struct ip_set *set,
 410 +                     const void *data, size_t size,
 411 +                     ip_set_ip_t *ip);
 412 +
 413 +       /* remove IP from set (kernel: iptables ... -j SET --entry x)
 414 +        * Return -EEXIST if the address is NOT in the set,
 415 +        * and -ERANGE if the address lies outside the set bounds.
 416 +        * If the address really was in the set, 0 is returned.
 417 +        */
 418 +       int (*delip_kernel) (struct ip_set *set,
 419 +                            const struct sk_buff * skb,
 420 +                            ip_set_ip_t *ip,
 421 +                            const u_int32_t *flags,
 422 +                            unsigned char index);
 423 +
 424 +       /* new set creation - allocated type specific items
 425 +        */
 426 +       int (*create) (struct ip_set *set,
 427 +                      const void *data, size_t size);
 428 +
 429 +       /* retry the operation after successfully tweaking the set
 430 +        */
 431 +       int (*retry) (struct ip_set *set);
 432 +
 433 +       /* set destruction - free type specific items
 434 +        * There is no return value.
 435 +        * Can be called only when child sets are destroyed.
 436 +        */
 437 +       void (*destroy) (struct ip_set *set);
 438 +
 439 +       /* set flushing - reset all bits in the set, or something similar.
 440 +        * There is no return value.
 441 +        */
 442 +       void (*flush) (struct ip_set *set);
 443 +
 444 +       /* Listing: size needed for header
 445 +        */
 446 +       size_t header_size;
 447 +
 448 +       /* Listing: Get the header
 449 +        *
 450 +        * Fill in the information in "data".
 451 +        * This function is always run after list_header_size() under a
 452 +        * writelock on the set. Therefor is the length of "data" always
 453 +        * correct.
 454 +        */
 455 +       void (*list_header) (const struct ip_set *set,
 456 +                            void *data);
 457 +
 458 +       /* Listing: Get the size for the set members
 459 +        */
 460 +       int (*list_members_size) (const struct ip_set *set);
 461 +
 462 +       /* Listing: Get the set members
 463 +        *
 464 +        * Fill in the information in "data".
 465 +        * This function is always run after list_member_size() under a
 466 +        * writelock on the set. Therefor is the length of "data" always
 467 +        * correct.
 468 +        */
 469 +       void (*list_members) (const struct ip_set *set,
 470 +                             void *data);
 471 +
 472 +       char typename[IP_SET_MAXNAMELEN];
 473 +       unsigned char features;
 474 +       int protocol_version;
 475 +
 476 +       /* Set this to THIS_MODULE if you are a module, otherwise NULL */
 477 +       struct module *me;
 478 +};
 479 +
 480 +extern int ip_set_register_set_type(struct ip_set_type *set_type);
 481 +extern void ip_set_unregister_set_type(struct ip_set_type *set_type);
 482 +
 483 +/* A generic ipset */
 484 +struct ip_set {
 485 +       char name[IP_SET_MAXNAMELEN];   /* the name of the set */
 486 +       rwlock_t lock;                  /* lock for concurrency control */
 487 +       ip_set_id_t id;                 /* set id for swapping */
 488 +       ip_set_id_t binding;            /* default binding for the set */
 489 +       atomic_t ref;                   /* in kernel and in hash references */
 490 +       struct ip_set_type *type;       /* the set types */
 491 +       void *data;                     /* pooltype specific data */
 492 +};
 493 +
 494 +/* Structure to bind set elements to sets */
 495 +struct ip_set_hash {
 496 +       struct list_head list;          /* list of clashing entries in hash */
 497 +       ip_set_ip_t ip;                 /* ip from set */
 498 +       ip_set_id_t id;                 /* set id */
 499 +       ip_set_id_t binding;            /* set we bind the element to */
 500 +};
 501 +
 502 +/* register and unregister set references */
 503 +extern ip_set_id_t ip_set_get_byname(const char name[IP_SET_MAXNAMELEN]);
 504 +extern ip_set_id_t ip_set_get_byindex(ip_set_id_t id);
 505 +extern void ip_set_put(ip_set_id_t id);
 506 +
 507 +/* API for iptables set match, and SET target */
 508 +extern void ip_set_addip_kernel(ip_set_id_t id,
 509 +                               const struct sk_buff *skb,
 510 +                               const u_int32_t *flags);
 511 +extern void ip_set_delip_kernel(ip_set_id_t id,
 512 +                               const struct sk_buff *skb,
 513 +                               const u_int32_t *flags);
 514 +extern int ip_set_testip_kernel(ip_set_id_t id,
 515 +                               const struct sk_buff *skb,
 516 +                               const u_int32_t *flags);
 517 +
 518 +#endif                         /* __KERNEL__ */
 519 +
 520 +#endif /*_IP_SET_H*/
 521 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_iphash.h
 522 ===================================================================
 523 --- /dev/null
 524 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_iphash.h
 525 @@ -0,0 +1,30 @@
 526 +#ifndef __IP_SET_IPHASH_H
 527 +#define __IP_SET_IPHASH_H
 528 +
 529 +#include <linux/netfilter_ipv4/ip_set.h>
 530 +
 531 +#define SETTYPE_NAME "iphash"
 532 +#define MAX_RANGE 0x0000FFFF
 533 +
 534 +struct ip_set_iphash {
 535 +       ip_set_ip_t *members;           /* the iphash proper */
 536 +       uint32_t elements;              /* number of elements */
 537 +       uint32_t hashsize;              /* hash size */
 538 +       uint16_t probes;                /* max number of probes  */
 539 +       uint16_t resize;                /* resize factor in percent */
 540 +       ip_set_ip_t netmask;            /* netmask */
 541 +       void *initval[0];               /* initvals for jhash_1word */
 542 +};
 543 +
 544 +struct ip_set_req_iphash_create {
 545 +       uint32_t hashsize;
 546 +       uint16_t probes;
 547 +       uint16_t resize;
 548 +       ip_set_ip_t netmask;
 549 +};
 550 +
 551 +struct ip_set_req_iphash {
 552 +       ip_set_ip_t ip;
 553 +};
 554 +
 555 +#endif /* __IP_SET_IPHASH_H */
 556 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_ipmap.h
 557 ===================================================================
 558 --- /dev/null
 559 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_ipmap.h
 560 @@ -0,0 +1,56 @@
 561 +#ifndef __IP_SET_IPMAP_H
 562 +#define __IP_SET_IPMAP_H
 563 +
 564 +#include <linux/netfilter_ipv4/ip_set.h>
 565 +
 566 +#define SETTYPE_NAME "ipmap"
 567 +#define MAX_RANGE 0x0000FFFF
 568 +
 569 +struct ip_set_ipmap {
 570 +       void *members;                  /* the ipmap proper */
 571 +       ip_set_ip_t first_ip;           /* host byte order, included in range */
 572 +       ip_set_ip_t last_ip;            /* host byte order, included in range */
 573 +       ip_set_ip_t netmask;            /* subnet netmask */
 574 +       ip_set_ip_t sizeid;             /* size of set in IPs */
 575 +       ip_set_ip_t hosts;              /* number of hosts in a subnet */
 576 +};
 577 +
 578 +struct ip_set_req_ipmap_create {
 579 +       ip_set_ip_t from;
 580 +       ip_set_ip_t to;
 581 +       ip_set_ip_t netmask;
 582 +};
 583 +
 584 +struct ip_set_req_ipmap {
 585 +       ip_set_ip_t ip;
 586 +};
 587 +
 588 +unsigned int
 589 +mask_to_bits(ip_set_ip_t mask)
 590 +{
 591 +       unsigned int bits = 32;
 592 +       ip_set_ip_t maskaddr;
 593 +
 594 +       if (mask == 0xFFFFFFFF)
 595 +               return bits;
 596 +
 597 +       maskaddr = 0xFFFFFFFE;
 598 +       while (--bits >= 0 && maskaddr != mask)
 599 +               maskaddr <<= 1;
 600 +
 601 +       return bits;
 602 +}
 603 +
 604 +ip_set_ip_t
 605 +range_to_mask(ip_set_ip_t from, ip_set_ip_t to, unsigned int *bits)
 606 +{
 607 +       ip_set_ip_t mask = 0xFFFFFFFE;
 608 +
 609 +       *bits = 32;
 610 +       while (--(*bits) >= 0 && mask && (to & mask) != from)
 611 +               mask <<= 1;
 612 +
 613 +       return mask;
 614 +}
 615 +
 616 +#endif /* __IP_SET_IPMAP_H */
 617 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_ipporthash.h
 618 ===================================================================
 619 --- /dev/null
 620 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_ipporthash.h
 621 @@ -0,0 +1,34 @@
 622 +#ifndef __IP_SET_IPPORTHASH_H
 623 +#define __IP_SET_IPPORTHASH_H
 624 +
 625 +#include <linux/netfilter_ipv4/ip_set.h>
 626 +
 627 +#define SETTYPE_NAME "ipporthash"
 628 +#define MAX_RANGE 0x0000FFFF
 629 +#define INVALID_PORT   (MAX_RANGE + 1)
 630 +
 631 +struct ip_set_ipporthash {
 632 +       ip_set_ip_t *members;           /* the ipporthash proper */
 633 +       uint32_t elements;              /* number of elements */
 634 +       uint32_t hashsize;              /* hash size */
 635 +       uint16_t probes;                /* max number of probes  */
 636 +       uint16_t resize;                /* resize factor in percent */
 637 +       ip_set_ip_t first_ip;           /* host byte order, included in range */
 638 +       ip_set_ip_t last_ip;            /* host byte order, included in range */
 639 +       void *initval[0];               /* initvals for jhash_1word */
 640 +};
 641 +
 642 +struct ip_set_req_ipporthash_create {
 643 +       uint32_t hashsize;
 644 +       uint16_t probes;
 645 +       uint16_t resize;
 646 +       ip_set_ip_t from;
 647 +       ip_set_ip_t to;
 648 +};
 649 +
 650 +struct ip_set_req_ipporthash {
 651 +       ip_set_ip_t ip;
 652 +       ip_set_ip_t port;
 653 +};
 654 +
 655 +#endif /* __IP_SET_IPPORTHASH_H */
 656 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_iptree.h
 657 ===================================================================
 658 --- /dev/null
 659 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_iptree.h
 660 @@ -0,0 +1,40 @@
 661 +#ifndef __IP_SET_IPTREE_H
 662 +#define __IP_SET_IPTREE_H
 663 +
 664 +#include <linux/netfilter_ipv4/ip_set.h>
 665 +
 666 +#define SETTYPE_NAME "iptree"
 667 +#define MAX_RANGE 0x0000FFFF
 668 +
 669 +struct ip_set_iptreed {
 670 +       unsigned long expires[256];             /* x.x.x.ADDR */
 671 +};
 672 +
 673 +struct ip_set_iptreec {
 674 +       struct ip_set_iptreed *tree[256];       /* x.x.ADDR.* */
 675 +};
 676 +
 677 +struct ip_set_iptreeb {
 678 +       struct ip_set_iptreec *tree[256];       /* x.ADDR.*.* */
 679 +};
 680 +
 681 +struct ip_set_iptree {
 682 +       unsigned int timeout;
 683 +       unsigned int gc_interval;
 684 +#ifdef __KERNEL__
 685 +       uint32_t elements;              /* number of elements */
 686 +       struct timer_list gc;
 687 +       struct ip_set_iptreeb *tree[256];       /* ADDR.*.*.* */
 688 +#endif
 689 +};
 690 +
 691 +struct ip_set_req_iptree_create {
 692 +       unsigned int timeout;
 693 +};
 694 +
 695 +struct ip_set_req_iptree {
 696 +       ip_set_ip_t ip;
 697 +       unsigned int timeout;
 698 +};
 699 +
 700 +#endif /* __IP_SET_IPTREE_H */
 701 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_iptreemap.h
 702 ===================================================================
 703 --- /dev/null
 704 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_iptreemap.h
 705 @@ -0,0 +1,40 @@
 706 +#ifndef __IP_SET_IPTREEMAP_H
 707 +#define __IP_SET_IPTREEMAP_H
 708 +
 709 +#include <linux/netfilter_ipv4/ip_set.h>
 710 +
 711 +#define SETTYPE_NAME "iptreemap"
 712 +
 713 +#ifdef __KERNEL__
 714 +struct ip_set_iptreemap_d {
 715 +       unsigned char bitmap[32]; /* x.x.x.y */
 716 +};
 717 +
 718 +struct ip_set_iptreemap_c {
 719 +       struct ip_set_iptreemap_d *tree[256]; /* x.x.y.x */
 720 +};
 721 +
 722 +struct ip_set_iptreemap_b {
 723 +       struct ip_set_iptreemap_c *tree[256]; /* x.y.x.x */
 724 +       unsigned char dirty[32];
 725 +};
 726 +#endif
 727 +
 728 +struct ip_set_iptreemap {
 729 +       unsigned int gc_interval;
 730 +#ifdef __KERNEL__
 731 +       struct timer_list gc;
 732 +       struct ip_set_iptreemap_b *tree[256]; /* y.x.x.x */
 733 +#endif
 734 +};
 735 +
 736 +struct ip_set_req_iptreemap_create {
 737 +       unsigned int gc_interval;
 738 +};
 739 +
 740 +struct ip_set_req_iptreemap {
 741 +       ip_set_ip_t start;
 742 +       ip_set_ip_t end;
 743 +};
 744 +
 745 +#endif /* __IP_SET_IPTREEMAP_H */
 746 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_jhash.h
 747 ===================================================================
 748 --- /dev/null
 749 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_jhash.h
 750 @@ -0,0 +1,148 @@
 751 +#ifndef _LINUX_IPSET_JHASH_H
 752 +#define _LINUX_IPSET_JHASH_H
 753 +
 754 +/* This is a copy of linux/jhash.h but the types u32/u8 are changed
 755 + * to __u32/__u8 so that the header file can be included into
 756 + * userspace code as well. Jozsef Kadlecsik (kadlec@blackhole.kfki.hu)
 757 + */
 758 +
 759 +/* jhash.h: Jenkins hash support.
 760 + *
 761 + * Copyright (C) 1996 Bob Jenkins (bob_jenkins@burtleburtle.net)
 762 + *
 763 + * http://burtleburtle.net/bob/hash/
 764 + *
 765 + * These are the credits from Bob's sources:
 766 + *
 767 + * lookup2.c, by Bob Jenkins, December 1996, Public Domain.
 768 + * hash(), hash2(), hash3, and mix() are externally useful functions.
 769 + * Routines to test the hash are included if SELF_TEST is defined.
 770 + * You can use this free for any purpose.  It has no warranty.
 771 + *
 772 + * Copyright (C) 2003 David S. Miller (davem@redhat.com)
 773 + *
 774 + * I've modified Bob's hash to be useful in the Linux kernel, and
 775 + * any bugs present are surely my fault.  -DaveM
 776 + */
 777 +
 778 +/* NOTE: Arguments are modified. */
 779 +#define __jhash_mix(a, b, c) \
 780 +{ \
 781 +  a -= b; a -= c; a ^= (c>>13); \
 782 +  b -= c; b -= a; b ^= (a<<8); \
 783 +  c -= a; c -= b; c ^= (b>>13); \
 784 +  a -= b; a -= c; a ^= (c>>12);  \
 785 +  b -= c; b -= a; b ^= (a<<16); \
 786 +  c -= a; c -= b; c ^= (b>>5); \
 787 +  a -= b; a -= c; a ^= (c>>3);  \
 788 +  b -= c; b -= a; b ^= (a<<10); \
 789 +  c -= a; c -= b; c ^= (b>>15); \
 790 +}
 791 +
 792 +/* The golden ration: an arbitrary value */
 793 +#define JHASH_GOLDEN_RATIO     0x9e3779b9
 794 +
 795 +/* The most generic version, hashes an arbitrary sequence
 796 + * of bytes.  No alignment or length assumptions are made about
 797 + * the input key.
 798 + */
 799 +static inline __u32 jhash(void *key, __u32 length, __u32 initval)
 800 +{
 801 +       __u32 a, b, c, len;
 802 +       __u8 *k = key;
 803 +
 804 +       len = length;
 805 +       a = b = JHASH_GOLDEN_RATIO;
 806 +       c = initval;
 807 +
 808 +       while (len >= 12) {
 809 +               a += (k[0] +((__u32)k[1]<<8) +((__u32)k[2]<<16) +((__u32)k[3]<<24));
 810 +               b += (k[4] +((__u32)k[5]<<8) +((__u32)k[6]<<16) +((__u32)k[7]<<24));
 811 +               c += (k[8] +((__u32)k[9]<<8) +((__u32)k[10]<<16)+((__u32)k[11]<<24));
 812 +
 813 +               __jhash_mix(a,b,c);
 814 +
 815 +               k += 12;
 816 +               len -= 12;
 817 +       }
 818 +
 819 +       c += length;
 820 +       switch (len) {
 821 +       case 11: c += ((__u32)k[10]<<24);
 822 +       case 10: c += ((__u32)k[9]<<16);
 823 +       case 9 : c += ((__u32)k[8]<<8);
 824 +       case 8 : b += ((__u32)k[7]<<24);
 825 +       case 7 : b += ((__u32)k[6]<<16);
 826 +       case 6 : b += ((__u32)k[5]<<8);
 827 +       case 5 : b += k[4];
 828 +       case 4 : a += ((__u32)k[3]<<24);
 829 +       case 3 : a += ((__u32)k[2]<<16);
 830 +       case 2 : a += ((__u32)k[1]<<8);
 831 +       case 1 : a += k[0];
 832 +       };
 833 +
 834 +       __jhash_mix(a,b,c);
 835 +
 836 +       return c;
 837 +}
 838 +
 839 +/* A special optimized version that handles 1 or more of __u32s.
 840 + * The length parameter here is the number of __u32s in the key.
 841 + */
 842 +static inline __u32 jhash2(__u32 *k, __u32 length, __u32 initval)
 843 +{
 844 +       __u32 a, b, c, len;
 845 +
 846 +       a = b = JHASH_GOLDEN_RATIO;
 847 +       c = initval;
 848 +       len = length;
 849 +
 850 +       while (len >= 3) {
 851 +               a += k[0];
 852 +               b += k[1];
 853 +               c += k[2];
 854 +               __jhash_mix(a, b, c);
 855 +               k += 3; len -= 3;
 856 +       }
 857 +
 858 +       c += length * 4;
 859 +
 860 +       switch (len) {
 861 +       case 2 : b += k[1];
 862 +       case 1 : a += k[0];
 863 +       };
 864 +
 865 +       __jhash_mix(a,b,c);
 866 +
 867 +       return c;
 868 +}
 869 +
 870 +
 871 +/* A special ultra-optimized versions that knows they are hashing exactly
 872 + * 3, 2 or 1 word(s).
 873 + *
 874 + * NOTE: In partilar the "c += length; __jhash_mix(a,b,c);" normally
 875 + *       done at the end is not done here.
 876 + */
 877 +static inline __u32 jhash_3words(__u32 a, __u32 b, __u32 c, __u32 initval)
 878 +{
 879 +       a += JHASH_GOLDEN_RATIO;
 880 +       b += JHASH_GOLDEN_RATIO;
 881 +       c += initval;
 882 +
 883 +       __jhash_mix(a, b, c);
 884 +
 885 +       return c;
 886 +}
 887 +
 888 +static inline __u32 jhash_2words(__u32 a, __u32 b, __u32 initval)
 889 +{
 890 +       return jhash_3words(a, b, 0, initval);
 891 +}
 892 +
 893 +static inline __u32 jhash_1word(__u32 a, __u32 initval)
 894 +{
 895 +       return jhash_3words(a, 0, 0, initval);
 896 +}
 897 +
 898 +#endif /* _LINUX_IPSET_JHASH_H */
 899 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_macipmap.h
 900 ===================================================================
 901 --- /dev/null
 902 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_macipmap.h
 903 @@ -0,0 +1,38 @@
 904 +#ifndef __IP_SET_MACIPMAP_H
 905 +#define __IP_SET_MACIPMAP_H
 906 +
 907 +#include <linux/netfilter_ipv4/ip_set.h>
 908 +
 909 +#define SETTYPE_NAME "macipmap"
 910 +#define MAX_RANGE 0x0000FFFF
 911 +
 912 +/* general flags */
 913 +#define IPSET_MACIP_MATCHUNSET 1
 914 +
 915 +/* per ip flags */
 916 +#define IPSET_MACIP_ISSET      1
 917 +
 918 +struct ip_set_macipmap {
 919 +       void *members;                  /* the macipmap proper */
 920 +       ip_set_ip_t first_ip;           /* host byte order, included in range */
 921 +       ip_set_ip_t last_ip;            /* host byte order, included in range */
 922 +       u_int32_t flags;
 923 +};
 924 +
 925 +struct ip_set_req_macipmap_create {
 926 +       ip_set_ip_t from;
 927 +       ip_set_ip_t to;
 928 +       u_int32_t flags;
 929 +};
 930 +
 931 +struct ip_set_req_macipmap {
 932 +       ip_set_ip_t ip;
 933 +       unsigned char ethernet[ETH_ALEN];
 934 +};
 935 +
 936 +struct ip_set_macip {
 937 +       unsigned short flags;
 938 +       unsigned char ethernet[ETH_ALEN];
 939 +};
 940 +
 941 +#endif /* __IP_SET_MACIPMAP_H */
 942 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_malloc.h
 943 ===================================================================
 944 --- /dev/null
 945 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_malloc.h
 946 @@ -0,0 +1,116 @@
 947 +#ifndef _IP_SET_MALLOC_H
 948 +#define _IP_SET_MALLOC_H
 949 +
 950 +#ifdef __KERNEL__
 951 +
 952 +/* Memory allocation and deallocation */
 953 +static size_t max_malloc_size = 0;
 954 +
 955 +static inline void init_max_malloc_size(void)
 956 +{
 957 +#define CACHE(x) max_malloc_size = x;
 958 +#include <linux/kmalloc_sizes.h>
 959 +#undef CACHE
 960 +}
 961 +
 962 +static inline void * ip_set_malloc(size_t bytes)
 963 +{
 964 +       if (bytes > max_malloc_size)
 965 +               return vmalloc(bytes);
 966 +       else
 967 +               return kmalloc(bytes, GFP_KERNEL);
 968 +}
 969 +
 970 +static inline void ip_set_free(void * data, size_t bytes)
 971 +{
 972 +       if (bytes > max_malloc_size)
 973 +               vfree(data);
 974 +       else
 975 +               kfree(data);
 976 +}
 977 +
 978 +struct harray {
 979 +       size_t max_elements;
 980 +       void *arrays[0];
 981 +};
 982 +
 983 +static inline void *
 984 +harray_malloc(size_t hashsize, size_t typesize, int flags)
 985 +{
 986 +       struct harray *harray;
 987 +       size_t max_elements, size, i, j;
 988 +
 989 +       if (!max_malloc_size)
 990 +               init_max_malloc_size();
 991 +
 992 +       if (typesize > max_malloc_size)
 993 +               return NULL;
 994 +
 995 +       max_elements = max_malloc_size/typesize;
 996 +       size = hashsize/max_elements;
 997 +       if (hashsize % max_elements)
 998 +               size++;
 999 +
1000 +       /* Last pointer signals end of arrays */
1001 +       harray = kmalloc(sizeof(struct harray) + (size + 1) * sizeof(void *),
1002 +                        flags);
1003 +
1004 +       if (!harray)
1005 +               return NULL;
1006 +
1007 +       for (i = 0; i < size - 1; i++) {
1008 +               harray->arrays[i] = kmalloc(max_elements * typesize, flags);
1009 +               if (!harray->arrays[i])
1010 +                       goto undo;
1011 +               memset(harray->arrays[i], 0, max_elements * typesize);
1012 +       }
1013 +       harray->arrays[i] = kmalloc((hashsize - i * max_elements) * typesize,
1014 +                                   flags);
1015 +       if (!harray->arrays[i])
1016 +               goto undo;
1017 +       memset(harray->arrays[i], 0, (hashsize - i * max_elements) * typesize);
1018 +
1019 +       harray->max_elements = max_elements;
1020 +       harray->arrays[size] = NULL;
1021 +
1022 +       return (void *)harray;
1023 +
1024 +    undo:
1025 +       for (j = 0; j < i; j++) {
1026 +               kfree(harray->arrays[j]);
1027 +       }
1028 +       kfree(harray);
1029 +       return NULL;
1030 +}
1031 +
1032 +static inline void harray_free(void *h)
1033 +{
1034 +       struct harray *harray = (struct harray *) h;
1035 +       size_t i;
1036 +
1037 +       for (i = 0; harray->arrays[i] != NULL; i++)
1038 +               kfree(harray->arrays[i]);
1039 +       kfree(harray);
1040 +}
1041 +
1042 +static inline void harray_flush(void *h, size_t hashsize, size_t typesize)
1043 +{
1044 +       struct harray *harray = (struct harray *) h;
1045 +       size_t i;
1046 +
1047 +       for (i = 0; harray->arrays[i+1] != NULL; i++)
1048 +               memset(harray->arrays[i], 0, harray->max_elements * typesize);
1049 +       memset(harray->arrays[i], 0,
1050 +              (hashsize - i * harray->max_elements) * typesize);
1051 +}
1052 +
1053 +#define HARRAY_ELEM(h, type, which)                            \
1054 +({                                                             \
1055 +       struct harray *__h = (struct harray *)(h);              \
1056 +       ((type)((__h)->arrays[(which)/(__h)->max_elements])     \
1057 +               + (which)%(__h)->max_elements);                 \
1058 +})
1059 +
1060 +#endif                         /* __KERNEL__ */
1061 +
1062 +#endif /*_IP_SET_MALLOC_H*/
1063 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_nethash.h
1064 ===================================================================
1065 --- /dev/null
1066 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_nethash.h
1067 @@ -0,0 +1,55 @@
1068 +#ifndef __IP_SET_NETHASH_H
1069 +#define __IP_SET_NETHASH_H
1070 +
1071 +#include <linux/netfilter_ipv4/ip_set.h>
1072 +
1073 +#define SETTYPE_NAME "nethash"
1074 +#define MAX_RANGE 0x0000FFFF
1075 +
1076 +struct ip_set_nethash {
1077 +       ip_set_ip_t *members;           /* the nethash proper */
1078 +       uint32_t elements;              /* number of elements */
1079 +       uint32_t hashsize;              /* hash size */
1080 +       uint16_t probes;                /* max number of probes  */
1081 +       uint16_t resize;                /* resize factor in percent */
1082 +       unsigned char cidr[30];         /* CIDR sizes */
1083 +       void *initval[0];               /* initvals for jhash_1word */
1084 +};
1085 +
1086 +struct ip_set_req_nethash_create {
1087 +       uint32_t hashsize;
1088 +       uint16_t probes;
1089 +       uint16_t resize;
1090 +};
1091 +
1092 +struct ip_set_req_nethash {
1093 +       ip_set_ip_t ip;
1094 +       unsigned char cidr;
1095 +};
1096 +
1097 +static unsigned char shifts[] = {255, 253, 249, 241, 225, 193, 129, 1};
1098 +
1099 +static inline ip_set_ip_t
1100 +pack(ip_set_ip_t ip, unsigned char cidr)
1101 +{
1102 +       ip_set_ip_t addr, *paddr = &addr;
1103 +       unsigned char n, t, *a;
1104 +
1105 +       addr = htonl(ip & (0xFFFFFFFF << (32 - (cidr))));
1106 +#ifdef __KERNEL__
1107 +       DP("ip:%u.%u.%u.%u/%u", NIPQUAD(addr), cidr);
1108 +#endif
1109 +       n = cidr / 8;
1110 +       t = cidr % 8;
1111 +       a = &((unsigned char *)paddr)[n];
1112 +       *a = *a /(1 << (8 - t)) + shifts[t];
1113 +#ifdef __KERNEL__
1114 +       DP("n: %u, t: %u, a: %u", n, t, *a);
1115 +       DP("ip:%u.%u.%u.%u/%u, %u.%u.%u.%u",
1116 +          HIPQUAD(ip), cidr, NIPQUAD(addr));
1117 +#endif
1118 +
1119 +       return ntohl(addr);
1120 +}
1121 +
1122 +#endif /* __IP_SET_NETHASH_H */
1123 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_portmap.h
1124 ===================================================================
1125 --- /dev/null
1126 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ip_set_portmap.h
1127 @@ -0,0 +1,25 @@
1128 +#ifndef __IP_SET_PORTMAP_H
1129 +#define __IP_SET_PORTMAP_H
1130 +
1131 +#include <linux/netfilter_ipv4/ip_set.h>
1132 +
1133 +#define SETTYPE_NAME   "portmap"
1134 +#define MAX_RANGE      0x0000FFFF
1135 +#define INVALID_PORT   (MAX_RANGE + 1)
1136 +
1137 +struct ip_set_portmap {
1138 +       void *members;                  /* the portmap proper */
1139 +       ip_set_ip_t first_port;         /* host byte order, included in range */
1140 +       ip_set_ip_t last_port;          /* host byte order, included in range */
1141 +};
1142 +
1143 +struct ip_set_req_portmap_create {
1144 +       ip_set_ip_t from;
1145 +       ip_set_ip_t to;
1146 +};
1147 +
1148 +struct ip_set_req_portmap {
1149 +       ip_set_ip_t port;
1150 +};
1151 +
1152 +#endif /* __IP_SET_PORTMAP_H */
1153 Index: linux-2.6.21.7/include/linux/netfilter_ipv4/ipt_set.h
1154 ===================================================================
1155 --- /dev/null
1156 +++ linux-2.6.21.7/include/linux/netfilter_ipv4/ipt_set.h
1157 @@ -0,0 +1,21 @@
1158 +#ifndef _IPT_SET_H
1159 +#define _IPT_SET_H
1160 +
1161 +#include <linux/netfilter_ipv4/ip_set.h>
1162 +
1163 +struct ipt_set_info {
1164 +       ip_set_id_t index;
1165 +       u_int32_t flags[IP_SET_MAX_BINDINGS + 1];
1166 +};
1167 +
1168 +/* match info */
1169 +struct ipt_set_info_match {
1170 +       struct ipt_set_info match_set;
1171 +};
1172 +
1173 +struct ipt_set_info_target {
1174 +       struct ipt_set_info add_set;
1175 +       struct ipt_set_info del_set;
1176 +};
1177 +
1178 +#endif /*_IPT_SET_H*/
1179 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set.c
1180 ===================================================================
1181 --- /dev/null
1182 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set.c
1183 @@ -0,0 +1,2003 @@
1184 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
1185 + *                         Patrick Schaaf <bof@bof.de>
1186 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
1187 + *
1188 + * This program is free software; you can redistribute it and/or modify
1189 + * it under the terms of the GNU General Public License version 2 as
1190 + * published by the Free Software Foundation.
1191 + */
1192 +
1193 +/* Kernel module for IP set management */
1194 +
1195 +#include <linux/version.h>
1196 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
1197 +#include <linux/config.h>
1198 +#endif
1199 +#include <linux/module.h>
1200 +#include <linux/moduleparam.h>
1201 +#include <linux/kmod.h>
1202 +#include <linux/ip.h>
1203 +#include <linux/skbuff.h>
1204 +#include <linux/random.h>
1205 +#include <linux/jhash.h>
1206 +#include <linux/netfilter_ipv4/ip_tables.h>
1207 +#include <linux/errno.h>
1208 +#include <asm/uaccess.h>
1209 +#include <asm/bitops.h>
1210 +#include <asm/semaphore.h>
1211 +#include <linux/spinlock.h>
1212 +#include <linux/vmalloc.h>
1213 +
1214 +#define ASSERT_READ_LOCK(x)
1215 +#define ASSERT_WRITE_LOCK(x)
1216 +#include <linux/netfilter_ipv4/ip_set.h>
1217 +
1218 +static struct list_head set_type_list;         /* all registered sets */
1219 +static struct ip_set **ip_set_list;            /* all individual sets */
1220 +static DEFINE_RWLOCK(ip_set_lock);             /* protects the lists and the hash */
1221 +static DECLARE_MUTEX(ip_set_app_mutex);                /* serializes user access */
1222 +static ip_set_id_t ip_set_max = CONFIG_IP_NF_SET_MAX;
1223 +static ip_set_id_t ip_set_bindings_hash_size =  CONFIG_IP_NF_SET_HASHSIZE;
1224 +static struct list_head *ip_set_hash;          /* hash of bindings */
1225 +static unsigned int ip_set_hash_random;                /* random seed */
1226 +
1227 +/*
1228 + * Sets are identified either by the index in ip_set_list or by id.
1229 + * The id never changes and is used to find a key in the hash.
1230 + * The index may change by swapping and used at all other places
1231 + * (set/SET netfilter modules, binding value, etc.)
1232 + *
1233 + * Userspace requests are serialized by ip_set_mutex and sets can
1234 + * be deleted only from userspace. Therefore ip_set_list locking
1235 + * must obey the following rules:
1236 + *
1237 + * - kernel requests: read and write locking mandatory
1238 + * - user requests: read locking optional, write locking mandatory
1239 + */
1240 +
1241 +static inline void
1242 +__ip_set_get(ip_set_id_t index)
1243 +{
1244 +       atomic_inc(&ip_set_list[index]->ref);
1245 +}
1246 +
1247 +static inline void
1248 +__ip_set_put(ip_set_id_t index)
1249 +{
1250 +       atomic_dec(&ip_set_list[index]->ref);
1251 +}
1252 +
1253 +/*
1254 + * Binding routines
1255 + */
1256 +
1257 +static inline struct ip_set_hash *
1258 +__ip_set_find(u_int32_t key, ip_set_id_t id, ip_set_ip_t ip)
1259 +{
1260 +       struct ip_set_hash *set_hash;
1261 +
1262 +       list_for_each_entry(set_hash, &ip_set_hash[key], list)
1263 +               if (set_hash->id == id && set_hash->ip == ip)
1264 +                       return set_hash;
1265 +
1266 +       return NULL;
1267 +}
1268 +
1269 +static ip_set_id_t
1270 +ip_set_find_in_hash(ip_set_id_t id, ip_set_ip_t ip)
1271 +{
1272 +       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random)
1273 +                               % ip_set_bindings_hash_size;
1274 +       struct ip_set_hash *set_hash;
1275 +
1276 +       ASSERT_READ_LOCK(&ip_set_lock);
1277 +       IP_SET_ASSERT(ip_set_list[id]);
1278 +       DP("set: %s, ip: %u.%u.%u.%u", ip_set_list[id]->name, HIPQUAD(ip));
1279 +
1280 +       set_hash = __ip_set_find(key, id, ip);
1281 +
1282 +       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name,
1283 +          HIPQUAD(ip),
1284 +          set_hash != NULL ? ip_set_list[set_hash->binding]->name : "");
1285 +
1286 +       return (set_hash != NULL ? set_hash->binding : IP_SET_INVALID_ID);
1287 +}
1288 +
1289 +static inline void
1290 +__set_hash_del(struct ip_set_hash *set_hash)
1291 +{
1292 +       ASSERT_WRITE_LOCK(&ip_set_lock);
1293 +       IP_SET_ASSERT(ip_set_list[set_hash->binding]);
1294 +
1295 +       __ip_set_put(set_hash->binding);
1296 +       list_del(&set_hash->list);
1297 +       kfree(set_hash);
1298 +}
1299 +
1300 +static int
1301 +ip_set_hash_del(ip_set_id_t id, ip_set_ip_t ip)
1302 +{
1303 +       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random)
1304 +                               % ip_set_bindings_hash_size;
1305 +       struct ip_set_hash *set_hash;
1306 +
1307 +       IP_SET_ASSERT(ip_set_list[id]);
1308 +       DP("set: %s, ip: %u.%u.%u.%u", ip_set_list[id]->name, HIPQUAD(ip));
1309 +       write_lock_bh(&ip_set_lock);
1310 +       set_hash = __ip_set_find(key, id, ip);
1311 +       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name,
1312 +          HIPQUAD(ip),
1313 +          set_hash != NULL ? ip_set_list[set_hash->binding]->name : "");
1314 +
1315 +       if (set_hash != NULL)
1316 +               __set_hash_del(set_hash);
1317 +       write_unlock_bh(&ip_set_lock);
1318 +       return 0;
1319 +}
1320 +
1321 +static int
1322 +ip_set_hash_add(ip_set_id_t id, ip_set_ip_t ip, ip_set_id_t binding)
1323 +{
1324 +       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random)
1325 +                               % ip_set_bindings_hash_size;
1326 +       struct ip_set_hash *set_hash;
1327 +       int ret = 0;
1328 +
1329 +       IP_SET_ASSERT(ip_set_list[id]);
1330 +       IP_SET_ASSERT(ip_set_list[binding]);
1331 +       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name,
1332 +          HIPQUAD(ip), ip_set_list[binding]->name);
1333 +       write_lock_bh(&ip_set_lock);
1334 +       set_hash = __ip_set_find(key, id, ip);
1335 +       if (!set_hash) {
1336 +               set_hash = kmalloc(sizeof(struct ip_set_hash), GFP_ATOMIC);
1337 +               if (!set_hash) {
1338 +                       ret = -ENOMEM;
1339 +                       goto unlock;
1340 +               }
1341 +               INIT_LIST_HEAD(&set_hash->list);
1342 +               set_hash->id = id;
1343 +               set_hash->ip = ip;
1344 +               list_add(&set_hash->list, &ip_set_hash[key]);
1345 +       } else {
1346 +               IP_SET_ASSERT(ip_set_list[set_hash->binding]);
1347 +               DP("overwrite binding: %s",
1348 +                  ip_set_list[set_hash->binding]->name);
1349 +               __ip_set_put(set_hash->binding);
1350 +       }
1351 +       set_hash->binding = binding;
1352 +       __ip_set_get(set_hash->binding);
1353 +       DP("stored: key %u, id %u (%s), ip %u.%u.%u.%u, binding %u (%s)",
1354 +          key, id, ip_set_list[id]->name,
1355 +          HIPQUAD(ip), binding, ip_set_list[binding]->name);
1356 +    unlock:
1357 +       write_unlock_bh(&ip_set_lock);
1358 +       return ret;
1359 +}
1360 +
1361 +#define FOREACH_HASH_DO(fn, args...)                                           \
1362 +({                                                                             \
1363 +       ip_set_id_t __key;                                                      \
1364 +       struct ip_set_hash *__set_hash;                                         \
1365 +                                                                               \
1366 +       for (__key = 0; __key < ip_set_bindings_hash_size; __key++) {           \
1367 +               list_for_each_entry(__set_hash, &ip_set_hash[__key], list)      \
1368 +                       fn(__set_hash , ## args);                               \
1369 +       }                                                                       \
1370 +})
1371 +
1372 +#define FOREACH_HASH_RW_DO(fn, args...)                                                \
1373 +({                                                                             \
1374 +       ip_set_id_t __key;                                                      \
1375 +       struct ip_set_hash *__set_hash, *__n;                                   \
1376 +                                                                               \
1377 +       ASSERT_WRITE_LOCK(&ip_set_lock);                                        \
1378 +       for (__key = 0; __key < ip_set_bindings_hash_size; __key++) {           \
1379 +               list_for_each_entry_safe(__set_hash, __n, &ip_set_hash[__key], list)\
1380 +                       fn(__set_hash , ## args);                               \
1381 +       }                                                                       \
1382 +})
1383 +
1384 +/* Add, del and test set entries from kernel */
1385 +
1386 +#define follow_bindings(index, set, ip)                                        \
1387 +((index = ip_set_find_in_hash((set)->id, ip)) != IP_SET_INVALID_ID     \
1388 + || (index = (set)->binding) != IP_SET_INVALID_ID)
1389 +
1390 +int
1391 +ip_set_testip_kernel(ip_set_id_t index,
1392 +                    const struct sk_buff *skb,
1393 +                    const u_int32_t *flags)
1394 +{
1395 +       struct ip_set *set;
1396 +       ip_set_ip_t ip;
1397 +       int res;
1398 +       unsigned char i = 0;
1399 +
1400 +       IP_SET_ASSERT(flags[i]);
1401 +       read_lock_bh(&ip_set_lock);
1402 +       do {
1403 +               set = ip_set_list[index];
1404 +               IP_SET_ASSERT(set);
1405 +               DP("set %s, index %u", set->name, index);
1406 +               read_lock_bh(&set->lock);
1407 +               res = set->type->testip_kernel(set, skb, &ip, flags, i++);
1408 +               read_unlock_bh(&set->lock);
1409 +               i += !!(set->type->features & IPSET_DATA_DOUBLE);
1410 +       } while (res > 0
1411 +                && flags[i]
1412 +                && follow_bindings(index, set, ip));
1413 +       read_unlock_bh(&ip_set_lock);
1414 +
1415 +       return res;
1416 +}
1417 +
1418 +void
1419 +ip_set_addip_kernel(ip_set_id_t index,
1420 +                   const struct sk_buff *skb,
1421 +                   const u_int32_t *flags)
1422 +{
1423 +       struct ip_set *set;
1424 +       ip_set_ip_t ip;
1425 +       int res;
1426 +       unsigned char i = 0;
1427 +
1428 +       IP_SET_ASSERT(flags[i]);
1429 +   retry:
1430 +       read_lock_bh(&ip_set_lock);
1431 +       do {
1432 +               set = ip_set_list[index];
1433 +               IP_SET_ASSERT(set);
1434 +               DP("set %s, index %u", set->name, index);
1435 +               write_lock_bh(&set->lock);
1436 +               res = set->type->addip_kernel(set, skb, &ip, flags, i++);
1437 +               write_unlock_bh(&set->lock);
1438 +               i += !!(set->type->features & IPSET_DATA_DOUBLE);
1439 +       } while ((res == 0 || res == -EEXIST)
1440 +                && flags[i]
1441 +                && follow_bindings(index, set, ip));
1442 +       read_unlock_bh(&ip_set_lock);
1443 +
1444 +       if (res == -EAGAIN
1445 +           && set->type->retry
1446 +           && (res = set->type->retry(set)) == 0)
1447 +               goto retry;
1448 +}
1449 +
1450 +void
1451 +ip_set_delip_kernel(ip_set_id_t index,
1452 +                   const struct sk_buff *skb,
1453 +                   const u_int32_t *flags)
1454 +{
1455 +       struct ip_set *set;
1456 +       ip_set_ip_t ip;
1457 +       int res;
1458 +       unsigned char i = 0;
1459 +
1460 +       IP_SET_ASSERT(flags[i]);
1461 +       read_lock_bh(&ip_set_lock);
1462 +       do {
1463 +               set = ip_set_list[index];
1464 +               IP_SET_ASSERT(set);
1465 +               DP("set %s, index %u", set->name, index);
1466 +               write_lock_bh(&set->lock);
1467 +               res = set->type->delip_kernel(set, skb, &ip, flags, i++);
1468 +               write_unlock_bh(&set->lock);
1469 +               i += !!(set->type->features & IPSET_DATA_DOUBLE);
1470 +       } while ((res == 0 || res == -EEXIST)
1471 +                && flags[i]
1472 +                && follow_bindings(index, set, ip));
1473 +       read_unlock_bh(&ip_set_lock);
1474 +}
1475 +
1476 +/* Register and deregister settype */
1477 +
1478 +static inline struct ip_set_type *
1479 +find_set_type(const char *name)
1480 +{
1481 +       struct ip_set_type *set_type;
1482 +
1483 +       list_for_each_entry(set_type, &set_type_list, list)
1484 +               if (!strncmp(set_type->typename, name, IP_SET_MAXNAMELEN - 1))
1485 +                       return set_type;
1486 +       return NULL;
1487 +}
1488 +
1489 +int
1490 +ip_set_register_set_type(struct ip_set_type *set_type)
1491 +{
1492 +       int ret = 0;
1493 +
1494 +       if (set_type->protocol_version != IP_SET_PROTOCOL_VERSION) {
1495 +               ip_set_printk("'%s' uses wrong protocol version %u (want %u)",
1496 +                             set_type->typename,
1497 +                             set_type->protocol_version,
1498 +                             IP_SET_PROTOCOL_VERSION);
1499 +               return -EINVAL;
1500 +       }
1501 +
1502 +       write_lock_bh(&ip_set_lock);
1503 +       if (find_set_type(set_type->typename)) {
1504 +               /* Duplicate! */
1505 +               ip_set_printk("'%s' already registered!",
1506 +                             set_type->typename);
1507 +               ret = -EINVAL;
1508 +               goto unlock;
1509 +       }
1510 +       if (!try_module_get(THIS_MODULE)) {
1511 +               ret = -EFAULT;
1512 +               goto unlock;
1513 +       }
1514 +       list_add(&set_type->list, &set_type_list);
1515 +       DP("'%s' registered.", set_type->typename);
1516 +   unlock:
1517 +       write_unlock_bh(&ip_set_lock);
1518 +       return ret;
1519 +}
1520 +
1521 +void
1522 +ip_set_unregister_set_type(struct ip_set_type *set_type)
1523 +{
1524 +       write_lock_bh(&ip_set_lock);
1525 +       if (!find_set_type(set_type->typename)) {
1526 +               ip_set_printk("'%s' not registered?",
1527 +                             set_type->typename);
1528 +               goto unlock;
1529 +       }
1530 +       list_del(&set_type->list);
1531 +       module_put(THIS_MODULE);
1532 +       DP("'%s' unregistered.", set_type->typename);
1533 +   unlock:
1534 +       write_unlock_bh(&ip_set_lock);
1535 +
1536 +}
1537 +
1538 +/*
1539 + * Userspace routines
1540 + */
1541 +
1542 +/*
1543 + * Find set by name, reference it once. The reference makes sure the
1544 + * thing pointed to, does not go away under our feet. Drop the reference
1545 + * later, using ip_set_put().
1546 + */
1547 +ip_set_id_t
1548 +ip_set_get_byname(const char *name)
1549 +{
1550 +       ip_set_id_t i, index = IP_SET_INVALID_ID;
1551 +
1552 +       down(&ip_set_app_mutex);
1553 +       for (i = 0; i < ip_set_max; i++) {
1554 +               if (ip_set_list[i] != NULL
1555 +                   && strcmp(ip_set_list[i]->name, name) == 0) {
1556 +                       __ip_set_get(i);
1557 +                       index = i;
1558 +                       break;
1559 +               }
1560 +       }
1561 +       up(&ip_set_app_mutex);
1562 +       return index;
1563 +}
1564 +
1565 +/*
1566 + * Find set by index, reference it once. The reference makes sure the
1567 + * thing pointed to, does not go away under our feet. Drop the reference
1568 + * later, using ip_set_put().
1569 + */
1570 +ip_set_id_t
1571 +ip_set_get_byindex(ip_set_id_t index)
1572 +{
1573 +       down(&ip_set_app_mutex);
1574 +
1575 +       if (index >= ip_set_max)
1576 +               return IP_SET_INVALID_ID;
1577 +
1578 +       if (ip_set_list[index])
1579 +               __ip_set_get(index);
1580 +       else
1581 +               index = IP_SET_INVALID_ID;
1582 +
1583 +       up(&ip_set_app_mutex);
1584 +       return index;
1585 +}
1586 +
1587 +/*
1588 + * If the given set pointer points to a valid set, decrement
1589 + * reference count by 1. The caller shall not assume the index
1590 + * to be valid, after calling this function.
1591 + */
1592 +void ip_set_put(ip_set_id_t index)
1593 +{
1594 +       down(&ip_set_app_mutex);
1595 +       if (ip_set_list[index])
1596 +               __ip_set_put(index);
1597 +       up(&ip_set_app_mutex);
1598 +}
1599 +
1600 +/* Find a set by name or index */
1601 +static ip_set_id_t
1602 +ip_set_find_byname(const char *name)
1603 +{
1604 +       ip_set_id_t i, index = IP_SET_INVALID_ID;
1605 +
1606 +       for (i = 0; i < ip_set_max; i++) {
1607 +               if (ip_set_list[i] != NULL
1608 +                   && strcmp(ip_set_list[i]->name, name) == 0) {
1609 +                       index = i;
1610 +                       break;
1611 +               }
1612 +       }
1613 +       return index;
1614 +}
1615 +
1616 +static ip_set_id_t
1617 +ip_set_find_byindex(ip_set_id_t index)
1618 +{
1619 +       if (index >= ip_set_max || ip_set_list[index] == NULL)
1620 +               index = IP_SET_INVALID_ID;
1621 +
1622 +       return index;
1623 +}
1624 +
1625 +/*
1626 + * Add, del, test, bind and unbind
1627 + */
1628 +
1629 +static inline int
1630 +__ip_set_testip(struct ip_set *set,
1631 +               const void *data,
1632 +               size_t size,
1633 +               ip_set_ip_t *ip)
1634 +{
1635 +       int res;
1636 +
1637 +       read_lock_bh(&set->lock);
1638 +       res = set->type->testip(set, data, size, ip);
1639 +       read_unlock_bh(&set->lock);
1640 +
1641 +       return res;
1642 +}
1643 +
1644 +static int
1645 +__ip_set_addip(ip_set_id_t index,
1646 +              const void *data,
1647 +              size_t size)
1648 +{
1649 +       struct ip_set *set = ip_set_list[index];
1650 +       ip_set_ip_t ip;
1651 +       int res;
1652 +
1653 +       IP_SET_ASSERT(set);
1654 +       do {
1655 +               write_lock_bh(&set->lock);
1656 +               res = set->type->addip(set, data, size, &ip);
1657 +               write_unlock_bh(&set->lock);
1658 +       } while (res == -EAGAIN
1659 +                && set->type->retry
1660 +                && (res = set->type->retry(set)) == 0);
1661 +
1662 +       return res;
1663 +}
1664 +
1665 +static int
1666 +ip_set_addip(ip_set_id_t index,
1667 +            const void *data,
1668 +            size_t size)
1669 +{
1670 +
1671 +       return __ip_set_addip(index,
1672 +                             data + sizeof(struct ip_set_req_adt),
1673 +                             size - sizeof(struct ip_set_req_adt));
1674 +}
1675 +
1676 +static int
1677 +ip_set_delip(ip_set_id_t index,
1678 +            const void *data,
1679 +            size_t size)
1680 +{
1681 +       struct ip_set *set = ip_set_list[index];
1682 +       ip_set_ip_t ip;
1683 +       int res;
1684 +
1685 +       IP_SET_ASSERT(set);
1686 +       write_lock_bh(&set->lock);
1687 +       res = set->type->delip(set,
1688 +                              data + sizeof(struct ip_set_req_adt),
1689 +                              size - sizeof(struct ip_set_req_adt),
1690 +                              &ip);
1691 +       write_unlock_bh(&set->lock);
1692 +
1693 +       return res;
1694 +}
1695 +
1696 +static int
1697 +ip_set_testip(ip_set_id_t index,
1698 +             const void *data,
1699 +             size_t size)
1700 +{
1701 +       struct ip_set *set = ip_set_list[index];
1702 +       ip_set_ip_t ip;
1703 +       int res;
1704 +
1705 +       IP_SET_ASSERT(set);
1706 +       res = __ip_set_testip(set,
1707 +                             data + sizeof(struct ip_set_req_adt),
1708 +                             size - sizeof(struct ip_set_req_adt),
1709 +                             &ip);
1710 +
1711 +       return (res > 0 ? -EEXIST : res);
1712 +}
1713 +
1714 +static int
1715 +ip_set_bindip(ip_set_id_t index,
1716 +             const void *data,
1717 +             size_t size)
1718 +{
1719 +       struct ip_set *set = ip_set_list[index];
1720 +       struct ip_set_req_bind *req_bind;
1721 +       ip_set_id_t binding;
1722 +       ip_set_ip_t ip;
1723 +       int res;
1724 +
1725 +       IP_SET_ASSERT(set);
1726 +       if (size < sizeof(struct ip_set_req_bind))
1727 +               return -EINVAL;
1728 +
1729 +       req_bind = (struct ip_set_req_bind *) data;
1730 +       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
1731 +
1732 +       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
1733 +               /* Default binding of a set */
1734 +               char *binding_name;
1735 +
1736 +               if (size != sizeof(struct ip_set_req_bind) + IP_SET_MAXNAMELEN)
1737 +                       return -EINVAL;
1738 +
1739 +               binding_name = (char *)(data + sizeof(struct ip_set_req_bind));
1740 +               binding_name[IP_SET_MAXNAMELEN - 1] = '\0';
1741 +
1742 +               binding = ip_set_find_byname(binding_name);
1743 +               if (binding == IP_SET_INVALID_ID)
1744 +                       return -ENOENT;
1745 +
1746 +               write_lock_bh(&ip_set_lock);
1747 +               /* Sets as binding values are referenced */
1748 +               if (set->binding != IP_SET_INVALID_ID)
1749 +                       __ip_set_put(set->binding);
1750 +               set->binding = binding;
1751 +               __ip_set_get(set->binding);
1752 +               write_unlock_bh(&ip_set_lock);
1753 +
1754 +               return 0;
1755 +       }
1756 +       binding = ip_set_find_byname(req_bind->binding);
1757 +       if (binding == IP_SET_INVALID_ID)
1758 +               return -ENOENT;
1759 +
1760 +       res = __ip_set_testip(set,
1761 +                             data + sizeof(struct ip_set_req_bind),
1762 +                             size - sizeof(struct ip_set_req_bind),
1763 +                             &ip);
1764 +       DP("set %s, ip: %u.%u.%u.%u, binding %s",
1765 +          set->name, HIPQUAD(ip), ip_set_list[binding]->name);
1766 +
1767 +       if (res >= 0)
1768 +               res = ip_set_hash_add(set->id, ip, binding);
1769 +
1770 +       return res;
1771 +}
1772 +
1773 +#define FOREACH_SET_DO(fn, args...)                            \
1774 +({                                                             \
1775 +       ip_set_id_t __i;                                        \
1776 +       struct ip_set *__set;                                   \
1777 +                                                               \
1778 +       for (__i = 0; __i < ip_set_max; __i++) {                \
1779 +               __set = ip_set_list[__i];                       \
1780 +               if (__set != NULL)                              \
1781 +                       fn(__set , ##args);                     \
1782 +       }                                                       \
1783 +})
1784 +
1785 +static inline void
1786 +__set_hash_del_byid(struct ip_set_hash *set_hash, ip_set_id_t id)
1787 +{
1788 +       if (set_hash->id == id)
1789 +               __set_hash_del(set_hash);
1790 +}
1791 +
1792 +static inline void
1793 +__unbind_default(struct ip_set *set)
1794 +{
1795 +       if (set->binding != IP_SET_INVALID_ID) {
1796 +               /* Sets as binding values are referenced */
1797 +               __ip_set_put(set->binding);
1798 +               set->binding = IP_SET_INVALID_ID;
1799 +       }
1800 +}
1801 +
1802 +static int
1803 +ip_set_unbindip(ip_set_id_t index,
1804 +               const void *data,
1805 +               size_t size)
1806 +{
1807 +       struct ip_set *set;
1808 +       struct ip_set_req_bind *req_bind;
1809 +       ip_set_ip_t ip;
1810 +       int res;
1811 +
1812 +       DP("");
1813 +       if (size < sizeof(struct ip_set_req_bind))
1814 +               return -EINVAL;
1815 +
1816 +       req_bind = (struct ip_set_req_bind *) data;
1817 +       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
1818 +
1819 +       DP("%u %s", index, req_bind->binding);
1820 +       if (index == IP_SET_INVALID_ID) {
1821 +               /* unbind :all: */
1822 +               if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
1823 +                       /* Default binding of sets */
1824 +                       write_lock_bh(&ip_set_lock);
1825 +                       FOREACH_SET_DO(__unbind_default);
1826 +                       write_unlock_bh(&ip_set_lock);
1827 +                       return 0;
1828 +               } else if (strcmp(req_bind->binding, IPSET_TOKEN_ALL) == 0) {
1829 +                       /* Flush all bindings of all sets*/
1830 +                       write_lock_bh(&ip_set_lock);
1831 +                       FOREACH_HASH_RW_DO(__set_hash_del);
1832 +                       write_unlock_bh(&ip_set_lock);
1833 +                       return 0;
1834 +               }
1835 +               DP("unreachable reached!");
1836 +               return -EINVAL;
1837 +       }
1838 +
1839 +       set = ip_set_list[index];
1840 +       IP_SET_ASSERT(set);
1841 +       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
1842 +               /* Default binding of set */
1843 +               ip_set_id_t binding = ip_set_find_byindex(set->binding);
1844 +
1845 +               if (binding == IP_SET_INVALID_ID)
1846 +                       return -ENOENT;
1847 +
1848 +               write_lock_bh(&ip_set_lock);
1849 +               /* Sets in hash values are referenced */
1850 +               __ip_set_put(set->binding);
1851 +               set->binding = IP_SET_INVALID_ID;
1852 +               write_unlock_bh(&ip_set_lock);
1853 +
1854 +               return 0;
1855 +       } else if (strcmp(req_bind->binding, IPSET_TOKEN_ALL) == 0) {
1856 +               /* Flush all bindings */
1857 +
1858 +               write_lock_bh(&ip_set_lock);
1859 +               FOREACH_HASH_RW_DO(__set_hash_del_byid, set->id);
1860 +               write_unlock_bh(&ip_set_lock);
1861 +               return 0;
1862 +       }
1863 +
1864 +       res = __ip_set_testip(set,
1865 +                             data + sizeof(struct ip_set_req_bind),
1866 +                             size - sizeof(struct ip_set_req_bind),
1867 +                             &ip);
1868 +
1869 +       DP("set %s, ip: %u.%u.%u.%u", set->name, HIPQUAD(ip));
1870 +       if (res >= 0)
1871 +               res = ip_set_hash_del(set->id, ip);
1872 +
1873 +       return res;
1874 +}
1875 +
1876 +static int
1877 +ip_set_testbind(ip_set_id_t index,
1878 +               const void *data,
1879 +               size_t size)
1880 +{
1881 +       struct ip_set *set = ip_set_list[index];
1882 +       struct ip_set_req_bind *req_bind;
1883 +       ip_set_id_t binding;
1884 +       ip_set_ip_t ip;
1885 +       int res;
1886 +
1887 +       IP_SET_ASSERT(set);
1888 +       if (size < sizeof(struct ip_set_req_bind))
1889 +               return -EINVAL;
1890 +
1891 +       req_bind = (struct ip_set_req_bind *) data;
1892 +       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
1893 +
1894 +       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
1895 +               /* Default binding of set */
1896 +               char *binding_name;
1897 +
1898 +               if (size != sizeof(struct ip_set_req_bind) + IP_SET_MAXNAMELEN)
1899 +                       return -EINVAL;
1900 +
1901 +               binding_name = (char *)(data + sizeof(struct ip_set_req_bind));
1902 +               binding_name[IP_SET_MAXNAMELEN - 1] = '\0';
1903 +
1904 +               binding = ip_set_find_byname(binding_name);
1905 +               if (binding == IP_SET_INVALID_ID)
1906 +                       return -ENOENT;
1907 +
1908 +               res = (set->binding == binding) ? -EEXIST : 0;
1909 +
1910 +               return res;
1911 +       }
1912 +       binding = ip_set_find_byname(req_bind->binding);
1913 +       if (binding == IP_SET_INVALID_ID)
1914 +               return -ENOENT;
1915 +
1916 +
1917 +       res = __ip_set_testip(set,
1918 +                             data + sizeof(struct ip_set_req_bind),
1919 +                             size - sizeof(struct ip_set_req_bind),
1920 +                             &ip);
1921 +       DP("set %s, ip: %u.%u.%u.%u, binding %s",
1922 +          set->name, HIPQUAD(ip), ip_set_list[binding]->name);
1923 +
1924 +       if (res >= 0)
1925 +               res = (ip_set_find_in_hash(set->id, ip) == binding)
1926 +                       ? -EEXIST : 0;
1927 +
1928 +       return res;
1929 +}
1930 +
1931 +static struct ip_set_type *
1932 +find_set_type_rlock(const char *typename)
1933 +{
1934 +       struct ip_set_type *type;
1935 +
1936 +       read_lock_bh(&ip_set_lock);
1937 +       type = find_set_type(typename);
1938 +       if (type == NULL)
1939 +               read_unlock_bh(&ip_set_lock);
1940 +
1941 +       return type;
1942 +}
1943 +
1944 +static int
1945 +find_free_id(const char *name,
1946 +            ip_set_id_t *index,
1947 +            ip_set_id_t *id)
1948 +{
1949 +       ip_set_id_t i;
1950 +
1951 +       *id = IP_SET_INVALID_ID;
1952 +       for (i = 0;  i < ip_set_max; i++) {
1953 +               if (ip_set_list[i] == NULL) {
1954 +                       if (*id == IP_SET_INVALID_ID)
1955 +                               *id = *index = i;
1956 +               } else if (strcmp(name, ip_set_list[i]->name) == 0)
1957 +                       /* Name clash */
1958 +                       return -EEXIST;
1959 +       }
1960 +       if (*id == IP_SET_INVALID_ID)
1961 +               /* No free slot remained */
1962 +               return -ERANGE;
1963 +       /* Check that index is usable as id (swapping) */
1964 +    check:
1965 +       for (i = 0;  i < ip_set_max; i++) {
1966 +               if (ip_set_list[i] != NULL
1967 +                   && ip_set_list[i]->id == *id) {
1968 +                   *id = i;
1969 +                   goto check;
1970 +               }
1971 +       }
1972 +       return 0;
1973 +}
1974 +
1975 +/*
1976 + * Create a set
1977 + */
1978 +static int
1979 +ip_set_create(const char *name,
1980 +             const char *typename,
1981 +             ip_set_id_t restore,
1982 +             const void *data,
1983 +             size_t size)
1984 +{
1985 +       struct ip_set *set;
1986 +       ip_set_id_t index = 0, id;
1987 +       int res = 0;
1988 +
1989 +       DP("setname: %s, typename: %s, id: %u", name, typename, restore);
1990 +       /*
1991 +        * First, and without any locks, allocate and initialize
1992 +        * a normal base set structure.
1993 +        */
1994 +       set = kmalloc(sizeof(struct ip_set), GFP_KERNEL);
1995 +       if (!set)
1996 +               return -ENOMEM;
1997 +       set->lock = RW_LOCK_UNLOCKED;
1998 +       strncpy(set->name, name, IP_SET_MAXNAMELEN);
1999 +       set->binding = IP_SET_INVALID_ID;
2000 +       atomic_set(&set->ref, 0);
2001 +
2002 +       /*
2003 +        * Next, take the &ip_set_lock, check that we know the type,
2004 +        * and take a reference on the type, to make sure it
2005 +        * stays available while constructing our new set.
2006 +        *
2007 +        * After referencing the type, we drop the &ip_set_lock,
2008 +        * and let the new set construction run without locks.
2009 +        */
2010 +       set->type = find_set_type_rlock(typename);
2011 +       if (set->type == NULL) {
2012 +               /* Try loading the module */
2013 +               char modulename[IP_SET_MAXNAMELEN + strlen("ip_set_") + 1];
2014 +               strcpy(modulename, "ip_set_");
2015 +               strcat(modulename, typename);
2016 +               DP("try to load %s", modulename);
2017 +               request_module(modulename);
2018 +               set->type = find_set_type_rlock(typename);
2019 +       }
2020 +       if (set->type == NULL) {
2021 +               ip_set_printk("no set type '%s', set '%s' not created",
2022 +                             typename, name);
2023 +               res = -ENOENT;
2024 +               goto out;
2025 +       }
2026 +       if (!try_module_get(set->type->me)) {
2027 +               read_unlock_bh(&ip_set_lock);
2028 +               res = -EFAULT;
2029 +               goto out;
2030 +       }
2031 +       read_unlock_bh(&ip_set_lock);
2032 +
2033 +       /*
2034 +        * Without holding any locks, create private part.
2035 +        */
2036 +       res = set->type->create(set, data, size);
2037 +       if (res != 0)
2038 +               goto put_out;
2039 +
2040 +       /* BTW, res==0 here. */
2041 +
2042 +       /*
2043 +        * Here, we have a valid, constructed set. &ip_set_lock again,
2044 +        * find free id/index and check that it is not already in
2045 +        * ip_set_list.
2046 +        */
2047 +       write_lock_bh(&ip_set_lock);
2048 +       if ((res = find_free_id(set->name, &index, &id)) != 0) {
2049 +               DP("no free id!");
2050 +               goto cleanup;
2051 +       }
2052 +
2053 +       /* Make sure restore gets the same index */
2054 +       if (restore != IP_SET_INVALID_ID && index != restore) {
2055 +               DP("Can't restore, sets are screwed up");
2056 +               res = -ERANGE;
2057 +               goto cleanup;
2058 +       }
2059 +
2060 +       /*
2061 +        * Finally! Add our shiny new set to the list, and be done.
2062 +        */
2063 +       DP("create: '%s' created with index %u, id %u!", set->name, index, id);
2064 +       set->id = id;
2065 +       ip_set_list[index] = set;
2066 +       write_unlock_bh(&ip_set_lock);
2067 +       return res;
2068 +
2069 +    cleanup:
2070 +       write_unlock_bh(&ip_set_lock);
2071 +       set->type->destroy(set);
2072 +    put_out:
2073 +       module_put(set->type->me);
2074 +    out:
2075 +       kfree(set);
2076 +       return res;
2077 +}
2078 +
2079 +/*
2080 + * Destroy a given existing set
2081 + */
2082 +static void
2083 +ip_set_destroy_set(ip_set_id_t index)
2084 +{
2085 +       struct ip_set *set = ip_set_list[index];
2086 +
2087 +       IP_SET_ASSERT(set);
2088 +       DP("set: %s",  set->name);
2089 +       write_lock_bh(&ip_set_lock);
2090 +       FOREACH_HASH_RW_DO(__set_hash_del_byid, set->id);
2091 +       if (set->binding != IP_SET_INVALID_ID)
2092 +               __ip_set_put(set->binding);
2093 +       ip_set_list[index] = NULL;
2094 +       write_unlock_bh(&ip_set_lock);
2095 +
2096 +       /* Must call it without holding any lock */
2097 +       set->type->destroy(set);
2098 +       module_put(set->type->me);
2099 +       kfree(set);
2100 +}
2101 +
2102 +/*
2103 + * Destroy a set - or all sets
2104 + * Sets must not be referenced/used.
2105 + */
2106 +static int
2107 +ip_set_destroy(ip_set_id_t index)
2108 +{
2109 +       ip_set_id_t i;
2110 +
2111 +       /* ref modification always protected by the mutex */
2112 +       if (index != IP_SET_INVALID_ID) {
2113 +               if (atomic_read(&ip_set_list[index]->ref))
2114 +                       return -EBUSY;
2115 +               ip_set_destroy_set(index);
2116 +       } else {
2117 +               for (i = 0; i < ip_set_max; i++) {
2118 +                       if (ip_set_list[i] != NULL
2119 +                           && (atomic_read(&ip_set_list[i]->ref)))
2120 +                               return -EBUSY;
2121 +               }
2122 +
2123 +               for (i = 0; i < ip_set_max; i++) {
2124 +                       if (ip_set_list[i] != NULL)
2125 +                               ip_set_destroy_set(i);
2126 +               }
2127 +       }
2128 +       return 0;
2129 +}
2130 +
2131 +static void
2132 +ip_set_flush_set(struct ip_set *set)
2133 +{
2134 +       DP("set: %s %u",  set->name, set->id);
2135 +
2136 +       write_lock_bh(&set->lock);
2137 +       set->type->flush(set);
2138 +       write_unlock_bh(&set->lock);
2139 +}
2140 +
2141 +/*
2142 + * Flush data in a set - or in all sets
2143 + */
2144 +static int
2145 +ip_set_flush(ip_set_id_t index)
2146 +{
2147 +       if (index != IP_SET_INVALID_ID) {
2148 +               IP_SET_ASSERT(ip_set_list[index]);
2149 +               ip_set_flush_set(ip_set_list[index]);
2150 +       } else
2151 +               FOREACH_SET_DO(ip_set_flush_set);
2152 +
2153 +       return 0;
2154 +}
2155 +
2156 +/* Rename a set */
2157 +static int
2158 +ip_set_rename(ip_set_id_t index, const char *name)
2159 +{
2160 +       struct ip_set *set = ip_set_list[index];
2161 +       ip_set_id_t i;
2162 +       int res = 0;
2163 +
2164 +       DP("set: %s to %s",  set->name, name);
2165 +       write_lock_bh(&ip_set_lock);
2166 +       for (i = 0; i < ip_set_max; i++) {
2167 +               if (ip_set_list[i] != NULL
2168 +                   && strncmp(ip_set_list[i]->name,
2169 +                              name,
2170 +                              IP_SET_MAXNAMELEN - 1) == 0) {
2171 +                       res = -EEXIST;
2172 +                       goto unlock;
2173 +               }
2174 +       }
2175 +       strncpy(set->name, name, IP_SET_MAXNAMELEN);
2176 +    unlock:
2177 +       write_unlock_bh(&ip_set_lock);
2178 +       return res;
2179 +}
2180 +
2181 +/*
2182 + * Swap two sets so that name/index points to the other.
2183 + * References are also swapped.
2184 + */
2185 +static int
2186 +ip_set_swap(ip_set_id_t from_index, ip_set_id_t to_index)
2187 +{
2188 +       struct ip_set *from = ip_set_list[from_index];
2189 +       struct ip_set *to = ip_set_list[to_index];
2190 +       char from_name[IP_SET_MAXNAMELEN];
2191 +       u_int32_t from_ref;
2192 +
2193 +       DP("set: %s to %s",  from->name, to->name);
2194 +       /* Features must not change. Artifical restriction. */
2195 +       if (from->type->features != to->type->features)
2196 +               return -ENOEXEC;
2197 +
2198 +       /* No magic here: ref munging protected by the mutex */
2199 +       write_lock_bh(&ip_set_lock);
2200 +       strncpy(from_name, from->name, IP_SET_MAXNAMELEN);
2201 +       from_ref = atomic_read(&from->ref);
2202 +
2203 +       strncpy(from->name, to->name, IP_SET_MAXNAMELEN);
2204 +       atomic_set(&from->ref, atomic_read(&to->ref));
2205 +       strncpy(to->name, from_name, IP_SET_MAXNAMELEN);
2206 +       atomic_set(&to->ref, from_ref);
2207 +
2208 +       ip_set_list[from_index] = to;
2209 +       ip_set_list[to_index] = from;
2210 +
2211 +       write_unlock_bh(&ip_set_lock);
2212 +       return 0;
2213 +}
2214 +
2215 +/*
2216 + * List set data
2217 + */
2218 +
2219 +static inline void
2220 +__set_hash_bindings_size_list(struct ip_set_hash *set_hash,
2221 +                             ip_set_id_t id, size_t *size)
2222 +{
2223 +       if (set_hash->id == id)
2224 +               *size += sizeof(struct ip_set_hash_list);
2225 +}
2226 +
2227 +static inline void
2228 +__set_hash_bindings_size_save(struct ip_set_hash *set_hash,
2229 +                             ip_set_id_t id, size_t *size)
2230 +{
2231 +       if (set_hash->id == id)
2232 +               *size += sizeof(struct ip_set_hash_save);
2233 +}
2234 +
2235 +static inline void
2236 +__set_hash_bindings(struct ip_set_hash *set_hash,
2237 +                   ip_set_id_t id, void *data, int *used)
2238 +{
2239 +       if (set_hash->id == id) {
2240 +               struct ip_set_hash_list *hash_list =
2241 +                       (struct ip_set_hash_list *)(data + *used);
2242 +
2243 +               hash_list->ip = set_hash->ip;
2244 +               hash_list->binding = set_hash->binding;
2245 +               *used += sizeof(struct ip_set_hash_list);
2246 +       }
2247 +}
2248 +
2249 +static int ip_set_list_set(ip_set_id_t index,
2250 +                          void *data,
2251 +                          int *used,
2252 +                          int len)
2253 +{
2254 +       struct ip_set *set = ip_set_list[index];
2255 +       struct ip_set_list *set_list;
2256 +
2257 +       /* Pointer to our header */
2258 +       set_list = (struct ip_set_list *) (data + *used);
2259 +
2260 +       DP("set: %s, used: %d %p %p", set->name, *used, data, data + *used);
2261 +
2262 +       /* Get and ensure header size */
2263 +       if (*used + sizeof(struct ip_set_list) > len)
2264 +               goto not_enough_mem;
2265 +       *used += sizeof(struct ip_set_list);
2266 +
2267 +       read_lock_bh(&set->lock);
2268 +       /* Get and ensure set specific header size */
2269 +       set_list->header_size = set->type->header_size;
2270 +       if (*used + set_list->header_size > len)
2271 +               goto unlock_set;
2272 +
2273 +       /* Fill in the header */
2274 +       set_list->index = index;
2275 +       set_list->binding = set->binding;
2276 +       set_list->ref = atomic_read(&set->ref);
2277 +
2278 +       /* Fill in set spefific header data */
2279 +       set->type->list_header(set, data + *used);
2280 +       *used += set_list->header_size;
2281 +
2282 +       /* Get and ensure set specific members size */
2283 +       set_list->members_size = set->type->list_members_size(set);
2284 +       if (*used + set_list->members_size > len)
2285 +               goto unlock_set;
2286 +
2287 +       /* Fill in set spefific members data */
2288 +       set->type->list_members(set, data + *used);
2289 +       *used += set_list->members_size;
2290 +       read_unlock_bh(&set->lock);
2291 +
2292 +       /* Bindings */
2293 +
2294 +       /* Get and ensure set specific bindings size */
2295 +       set_list->bindings_size = 0;
2296 +       FOREACH_HASH_DO(__set_hash_bindings_size_list,
2297 +                       set->id, &set_list->bindings_size);
2298 +       if (*used + set_list->bindings_size > len)
2299 +               goto not_enough_mem;
2300 +
2301 +       /* Fill in set spefific bindings data */
2302 +       FOREACH_HASH_DO(__set_hash_bindings, set->id, data, used);
2303 +
2304 +       return 0;
2305 +
2306 +    unlock_set:
2307 +       read_unlock_bh(&set->lock);
2308 +    not_enough_mem:
2309 +       DP("not enough mem, try again");
2310 +       return -EAGAIN;
2311 +}
2312 +
2313 +/*
2314 + * Save sets
2315 + */
2316 +static int ip_set_save_set(ip_set_id_t index,
2317 +                          void *data,
2318 +                          int *used,
2319 +                          int len)
2320 +{
2321 +       struct ip_set *set;
2322 +       struct ip_set_save *set_save;
2323 +
2324 +       /* Pointer to our header */
2325 +       set_save = (struct ip_set_save *) (data + *used);
2326 +
2327 +       /* Get and ensure header size */
2328 +       if (*used + sizeof(struct ip_set_save) > len)
2329 +               goto not_enough_mem;
2330 +       *used += sizeof(struct ip_set_save);
2331 +
2332 +       set = ip_set_list[index];
2333 +       DP("set: %s, used: %u(%u) %p %p", set->name, *used, len,
2334 +          data, data + *used);
2335 +
2336 +       read_lock_bh(&set->lock);
2337 +       /* Get and ensure set specific header size */
2338 +       set_save->header_size = set->type->header_size;
2339 +       if (*used + set_save->header_size > len)
2340 +               goto unlock_set;
2341 +
2342 +       /* Fill in the header */
2343 +       set_save->index = index;
2344 +       set_save->binding = set->binding;
2345 +
2346 +       /* Fill in set spefific header data */
2347 +       set->type->list_header(set, data + *used);
2348 +       *used += set_save->header_size;
2349 +
2350 +       DP("set header filled: %s, used: %u(%u) %p %p", set->name, *used,
2351 +          set_save->header_size, data, data + *used);
2352 +       /* Get and ensure set specific members size */
2353 +       set_save->members_size = set->type->list_members_size(set);
2354 +       if (*used + set_save->members_size > len)
2355 +               goto unlock_set;
2356 +
2357 +       /* Fill in set spefific members data */
2358 +       set->type->list_members(set, data + *used);
2359 +       *used += set_save->members_size;
2360 +       read_unlock_bh(&set->lock);
2361 +       DP("set members filled: %s, used: %u(%u) %p %p", set->name, *used,
2362 +          set_save->members_size, data, data + *used);
2363 +       return 0;
2364 +
2365 +    unlock_set:
2366 +       read_unlock_bh(&set->lock);
2367 +    not_enough_mem:
2368 +       DP("not enough mem, try again");
2369 +       return -EAGAIN;
2370 +}
2371 +
2372 +static inline void
2373 +__set_hash_save_bindings(struct ip_set_hash *set_hash,
2374 +                        ip_set_id_t id,
2375 +                        void *data,
2376 +                        int *used,
2377 +                        int len,
2378 +                        int *res)
2379 +{
2380 +       if (*res == 0
2381 +           && (id == IP_SET_INVALID_ID || set_hash->id == id)) {
2382 +               struct ip_set_hash_save *hash_save =
2383 +                       (struct ip_set_hash_save *)(data + *used);
2384 +               /* Ensure bindings size */
2385 +               if (*used + sizeof(struct ip_set_hash_save) > len) {
2386 +                       *res = -ENOMEM;
2387 +                       return;
2388 +               }
2389 +               hash_save->id = set_hash->id;
2390 +               hash_save->ip = set_hash->ip;
2391 +               hash_save->binding = set_hash->binding;
2392 +               *used += sizeof(struct ip_set_hash_save);
2393 +       }
2394 +}
2395 +
2396 +static int ip_set_save_bindings(ip_set_id_t index,
2397 +                               void *data,
2398 +                               int *used,
2399 +                               int len)
2400 +{
2401 +       int res = 0;
2402 +       struct ip_set_save *set_save;
2403 +
2404 +       DP("used %u, len %u", *used, len);
2405 +       /* Get and ensure header size */
2406 +       if (*used + sizeof(struct ip_set_save) > len)
2407 +               return -ENOMEM;
2408 +
2409 +       /* Marker */
2410 +       set_save = (struct ip_set_save *) (data + *used);
2411 +       set_save->index = IP_SET_INVALID_ID;
2412 +       set_save->header_size = 0;
2413 +       set_save->members_size = 0;
2414 +       *used += sizeof(struct ip_set_save);
2415 +
2416 +       DP("marker added used %u, len %u", *used, len);
2417 +       /* Fill in bindings data */
2418 +       if (index != IP_SET_INVALID_ID)
2419 +               /* Sets are identified by id in hash */
2420 +               index = ip_set_list[index]->id;
2421 +       FOREACH_HASH_DO(__set_hash_save_bindings, index, data, used, len, &res);
2422 +
2423 +       return res;
2424 +}
2425 +
2426 +/*
2427 + * Restore sets
2428 + */
2429 +static int ip_set_restore(void *data,
2430 +                         int len)
2431 +{
2432 +       int res = 0;
2433 +       int line = 0, used = 0, members_size;
2434 +       struct ip_set *set;
2435 +       struct ip_set_hash_save *hash_save;
2436 +       struct ip_set_restore *set_restore;
2437 +       ip_set_id_t index;
2438 +
2439 +       /* Loop to restore sets */
2440 +       while (1) {
2441 +               line++;
2442 +
2443 +               DP("%u %u %u", used, sizeof(struct ip_set_restore), len);
2444 +               /* Get and ensure header size */
2445 +               if (used + sizeof(struct ip_set_restore) > len)
2446 +                       return line;
2447 +               set_restore = (struct ip_set_restore *) (data + used);
2448 +               used += sizeof(struct ip_set_restore);
2449 +
2450 +               /* Ensure data size */
2451 +               if (used
2452 +                   + set_restore->header_size
2453 +                   + set_restore->members_size > len)
2454 +                       return line;
2455 +
2456 +               /* Check marker */
2457 +               if (set_restore->index == IP_SET_INVALID_ID) {
2458 +                       line--;
2459 +                       goto bindings;
2460 +               }
2461 +
2462 +               /* Try to create the set */
2463 +               DP("restore %s %s", set_restore->name, set_restore->typename);
2464 +               res = ip_set_create(set_restore->name,
2465 +                                   set_restore->typename,
2466 +                                   set_restore->index,
2467 +                                   data + used,
2468 +                                   set_restore->header_size);
2469 +
2470 +               if (res != 0)
2471 +                       return line;
2472 +               used += set_restore->header_size;
2473 +
2474 +               index = ip_set_find_byindex(set_restore->index);
2475 +               DP("index %u, restore_index %u", index, set_restore->index);
2476 +               if (index != set_restore->index)
2477 +                       return line;
2478 +               /* Try to restore members data */
2479 +               set = ip_set_list[index];
2480 +               members_size = 0;
2481 +               DP("members_size %u reqsize %u",
2482 +                  set_restore->members_size, set->type->reqsize);
2483 +               while (members_size + set->type->reqsize <=
2484 +                      set_restore->members_size) {
2485 +                       line++;
2486 +                       DP("members: %u, line %u", members_size, line);
2487 +                       res = __ip_set_addip(index,
2488 +                                          data + used + members_size,
2489 +                                          set->type->reqsize);
2490 +                       if (!(res == 0 || res == -EEXIST))
2491 +                               return line;
2492 +                       members_size += set->type->reqsize;
2493 +               }
2494 +
2495 +               DP("members_size %u  %u",
2496 +                  set_restore->members_size, members_size);
2497 +               if (members_size != set_restore->members_size)
2498 +                       return line++;
2499 +               used += set_restore->members_size;
2500 +       }
2501 +
2502 +   bindings:
2503 +       /* Loop to restore bindings */
2504 +       while (used < len) {
2505 +               line++;
2506 +
2507 +               DP("restore binding, line %u", line);
2508 +               /* Get and ensure size */
2509 +               if (used + sizeof(struct ip_set_hash_save) > len)
2510 +                       return line;
2511 +               hash_save = (struct ip_set_hash_save *) (data + used);
2512 +               used += sizeof(struct ip_set_hash_save);
2513 +
2514 +               /* hash_save->id is used to store the index */
2515 +               index = ip_set_find_byindex(hash_save->id);
2516 +               DP("restore binding index %u, id %u, %u -> %u",
2517 +                  index, hash_save->id, hash_save->ip, hash_save->binding);
2518 +               if (index != hash_save->id)
2519 +                       return line;
2520 +               if (ip_set_find_byindex(hash_save->binding) == IP_SET_INVALID_ID) {
2521 +                       DP("corrupt binding set index %u", hash_save->binding);
2522 +                       return line;
2523 +               }
2524 +               set = ip_set_list[hash_save->id];
2525 +               /* Null valued IP means default binding */
2526 +               if (hash_save->ip)
2527 +                       res = ip_set_hash_add(set->id,
2528 +                                             hash_save->ip,
2529 +                                             hash_save->binding);
2530 +               else {
2531 +                       IP_SET_ASSERT(set->binding == IP_SET_INVALID_ID);
2532 +                       write_lock_bh(&ip_set_lock);
2533 +                       set->binding = hash_save->binding;
2534 +                       __ip_set_get(set->binding);
2535 +                       write_unlock_bh(&ip_set_lock);
2536 +                       DP("default binding: %u", set->binding);
2537 +               }
2538 +               if (res != 0)
2539 +                       return line;
2540 +       }
2541 +       if (used != len)
2542 +               return line;
2543 +
2544 +       return 0;
2545 +}
2546 +
2547 +static int
2548 +ip_set_sockfn_set(struct sock *sk, int optval, void *user, unsigned int len)
2549 +{
2550 +       void *data;
2551 +       int res = 0;            /* Assume OK */
2552 +       unsigned *op;
2553 +       struct ip_set_req_adt *req_adt;
2554 +       ip_set_id_t index = IP_SET_INVALID_ID;
2555 +       int (*adtfn)(ip_set_id_t index,
2556 +                    const void *data, size_t size);
2557 +       struct fn_table {
2558 +               int (*fn)(ip_set_id_t index,
2559 +                         const void *data, size_t size);
2560 +       } adtfn_table[] =
2561 +       { { ip_set_addip }, { ip_set_delip }, { ip_set_testip},
2562 +         { ip_set_bindip}, { ip_set_unbindip }, { ip_set_testbind },
2563 +       };
2564 +
2565 +       DP("optval=%d, user=%p, len=%d", optval, user, len);
2566 +       if (!capable(CAP_NET_ADMIN))
2567 +               return -EPERM;
2568 +       if (optval != SO_IP_SET)
2569 +               return -EBADF;
2570 +       if (len <= sizeof(unsigned)) {
2571 +               ip_set_printk("short userdata (want >%zu, got %u)",
2572 +                             sizeof(unsigned), len);
2573 +               return -EINVAL;
2574 +       }
2575 +       data = vmalloc(len);
2576 +       if (!data) {
2577 +               DP("out of mem for %u bytes", len);
2578 +               return -ENOMEM;
2579 +       }
2580 +       if (copy_from_user(data, user, len) != 0) {
2581 +               res = -EFAULT;
2582 +               goto done;
2583 +       }
2584 +       if (down_interruptible(&ip_set_app_mutex)) {
2585 +               res = -EINTR;
2586 +               goto done;
2587 +       }
2588 +
2589 +       op = (unsigned *)data;
2590 +       DP("op=%x", *op);
2591 +
2592 +       if (*op < IP_SET_OP_VERSION) {
2593 +               /* Check the version at the beginning of operations */
2594 +               struct ip_set_req_version *req_version =
2595 +                       (struct ip_set_req_version *) data;
2596 +               if (req_version->version != IP_SET_PROTOCOL_VERSION) {
2597 +                       res = -EPROTO;
2598 +                       goto done;
2599 +               }
2600 +       }
2601 +
2602 +       switch (*op) {
2603 +       case IP_SET_OP_CREATE:{
2604 +               struct ip_set_req_create *req_create
2605 +                       = (struct ip_set_req_create *) data;
2606 +
2607 +               if (len < sizeof(struct ip_set_req_create)) {
2608 +                       ip_set_printk("short CREATE data (want >=%zu, got %u)",
2609 +                                     sizeof(struct ip_set_req_create), len);
2610 +                       res = -EINVAL;
2611 +                       goto done;
2612 +               }
2613 +               req_create->name[IP_SET_MAXNAMELEN - 1] = '\0';
2614 +               req_create->typename[IP_SET_MAXNAMELEN - 1] = '\0';
2615 +               res = ip_set_create(req_create->name,
2616 +                                   req_create->typename,
2617 +                                   IP_SET_INVALID_ID,
2618 +                                   data + sizeof(struct ip_set_req_create),
2619 +                                   len - sizeof(struct ip_set_req_create));
2620 +               goto done;
2621 +       }
2622 +       case IP_SET_OP_DESTROY:{
2623 +               struct ip_set_req_std *req_destroy
2624 +                       = (struct ip_set_req_std *) data;
2625 +
2626 +               if (len != sizeof(struct ip_set_req_std)) {
2627 +                       ip_set_printk("invalid DESTROY data (want %zu, got %u)",
2628 +                                     sizeof(struct ip_set_req_std), len);
2629 +                       res = -EINVAL;
2630 +                       goto done;
2631 +               }
2632 +               if (strcmp(req_destroy->name, IPSET_TOKEN_ALL) == 0) {
2633 +                       /* Destroy all sets */
2634 +                       index = IP_SET_INVALID_ID;
2635 +               } else {
2636 +                       req_destroy->name[IP_SET_MAXNAMELEN - 1] = '\0';
2637 +                       index = ip_set_find_byname(req_destroy->name);
2638 +
2639 +                       if (index == IP_SET_INVALID_ID) {
2640 +                               res = -ENOENT;
2641 +                               goto done;
2642 +                       }
2643 +               }
2644 +
2645 +               res = ip_set_destroy(index);
2646 +               goto done;
2647 +       }
2648 +       case IP_SET_OP_FLUSH:{
2649 +               struct ip_set_req_std *req_flush =
2650 +                       (struct ip_set_req_std *) data;
2651 +
2652 +               if (len != sizeof(struct ip_set_req_std)) {
2653 +                       ip_set_printk("invalid FLUSH data (want %zu, got %u)",
2654 +                                     sizeof(struct ip_set_req_std), len);
2655 +                       res = -EINVAL;
2656 +                       goto done;
2657 +               }
2658 +               if (strcmp(req_flush->name, IPSET_TOKEN_ALL) == 0) {
2659 +                       /* Flush all sets */
2660 +                       index = IP_SET_INVALID_ID;
2661 +               } else {
2662 +                       req_flush->name[IP_SET_MAXNAMELEN - 1] = '\0';
2663 +                       index = ip_set_find_byname(req_flush->name);
2664 +
2665 +                       if (index == IP_SET_INVALID_ID) {
2666 +                               res = -ENOENT;
2667 +                               goto done;
2668 +                       }
2669 +               }
2670 +               res = ip_set_flush(index);
2671 +               goto done;
2672 +       }
2673 +       case IP_SET_OP_RENAME:{
2674 +               struct ip_set_req_create *req_rename
2675 +                       = (struct ip_set_req_create *) data;
2676 +
2677 +               if (len != sizeof(struct ip_set_req_create)) {
2678 +                       ip_set_printk("invalid RENAME data (want %zu, got %u)",
2679 +                                     sizeof(struct ip_set_req_create), len);
2680 +                       res = -EINVAL;
2681 +                       goto done;
2682 +               }
2683 +
2684 +               req_rename->name[IP_SET_MAXNAMELEN - 1] = '\0';
2685 +               req_rename->typename[IP_SET_MAXNAMELEN - 1] = '\0';
2686 +
2687 +               index = ip_set_find_byname(req_rename->name);
2688 +               if (index == IP_SET_INVALID_ID) {
2689 +                       res = -ENOENT;
2690 +                       goto done;
2691 +               }
2692 +               res = ip_set_rename(index, req_rename->typename);
2693 +               goto done;
2694 +       }
2695 +       case IP_SET_OP_SWAP:{
2696 +               struct ip_set_req_create *req_swap
2697 +                       = (struct ip_set_req_create *) data;
2698 +               ip_set_id_t to_index;
2699 +
2700 +               if (len != sizeof(struct ip_set_req_create)) {
2701 +                       ip_set_printk("invalid SWAP data (want %zu, got %u)",
2702 +                                     sizeof(struct ip_set_req_create), len);
2703 +                       res = -EINVAL;
2704 +                       goto done;
2705 +               }
2706 +
2707 +               req_swap->name[IP_SET_MAXNAMELEN - 1] = '\0';
2708 +               req_swap->typename[IP_SET_MAXNAMELEN - 1] = '\0';
2709 +
2710 +               index = ip_set_find_byname(req_swap->name);
2711 +               if (index == IP_SET_INVALID_ID) {
2712 +                       res = -ENOENT;
2713 +                       goto done;
2714 +               }
2715 +               to_index = ip_set_find_byname(req_swap->typename);
2716 +               if (to_index == IP_SET_INVALID_ID) {
2717 +                       res = -ENOENT;
2718 +                       goto done;
2719 +               }
2720 +               res = ip_set_swap(index, to_index);
2721 +               goto done;
2722 +       }
2723 +       default:
2724 +               break;  /* Set identified by id */
2725 +       }
2726 +
2727 +       /* There we may have add/del/test/bind/unbind/test_bind operations */
2728 +       if (*op < IP_SET_OP_ADD_IP || *op > IP_SET_OP_TEST_BIND_SET) {
2729 +               res = -EBADMSG;
2730 +               goto done;
2731 +       }
2732 +       adtfn = adtfn_table[*op - IP_SET_OP_ADD_IP].fn;
2733 +
2734 +       if (len < sizeof(struct ip_set_req_adt)) {
2735 +               ip_set_printk("short data in adt request (want >=%zu, got %u)",
2736 +                             sizeof(struct ip_set_req_adt), len);
2737 +               res = -EINVAL;
2738 +               goto done;
2739 +       }
2740 +       req_adt = (struct ip_set_req_adt *) data;
2741 +
2742 +       /* -U :all: :all:|:default: uses IP_SET_INVALID_ID */
2743 +       if (!(*op == IP_SET_OP_UNBIND_SET
2744 +             && req_adt->index == IP_SET_INVALID_ID)) {
2745 +               index = ip_set_find_byindex(req_adt->index);
2746 +               if (index == IP_SET_INVALID_ID) {
2747 +                       res = -ENOENT;
2748 +                       goto done;
2749 +               }
2750 +       }
2751 +       res = adtfn(index, data, len);
2752 +
2753 +    done:
2754 +       up(&ip_set_app_mutex);
2755 +       vfree(data);
2756 +       if (res > 0)
2757 +               res = 0;
2758 +       DP("final result %d", res);
2759 +       return res;
2760 +}
2761 +
2762 +static int
2763 +ip_set_sockfn_get(struct sock *sk, int optval, void *user, int *len)
2764 +{
2765 +       int res = 0;
2766 +       unsigned *op;
2767 +       ip_set_id_t index = IP_SET_INVALID_ID;
2768 +       void *data;
2769 +       int copylen = *len;
2770 +
2771 +       DP("optval=%d, user=%p, len=%d", optval, user, *len);
2772 +       if (!capable(CAP_NET_ADMIN))
2773 +               return -EPERM;
2774 +       if (optval != SO_IP_SET)
2775 +               return -EBADF;
2776 +       if (*len < sizeof(unsigned)) {
2777 +               ip_set_printk("short userdata (want >=%zu, got %d)",
2778 +                             sizeof(unsigned), *len);
2779 +               return -EINVAL;
2780 +       }
2781 +       data = vmalloc(*len);
2782 +       if (!data) {
2783 +               DP("out of mem for %d bytes", *len);
2784 +               return -ENOMEM;
2785 +       }
2786 +       if (copy_from_user(data, user, *len) != 0) {
2787 +               res = -EFAULT;
2788 +               goto done;
2789 +       }
2790 +       if (down_interruptible(&ip_set_app_mutex)) {
2791 +               res = -EINTR;
2792 +               goto done;
2793 +       }
2794 +
2795 +       op = (unsigned *) data;
2796 +       DP("op=%x", *op);
2797 +
2798 +       if (*op < IP_SET_OP_VERSION) {
2799 +               /* Check the version at the beginning of operations */
2800 +               struct ip_set_req_version *req_version =
2801 +                       (struct ip_set_req_version *) data;
2802 +               if (req_version->version != IP_SET_PROTOCOL_VERSION) {
2803 +                       res = -EPROTO;
2804 +                       goto done;
2805 +               }
2806 +       }
2807 +
2808 +       switch (*op) {
2809 +       case IP_SET_OP_VERSION: {
2810 +               struct ip_set_req_version *req_version =
2811 +                   (struct ip_set_req_version *) data;
2812 +
2813 +               if (*len != sizeof(struct ip_set_req_version)) {
2814 +                       ip_set_printk("invalid VERSION (want %zu, got %d)",
2815 +                                     sizeof(struct ip_set_req_version),
2816 +                                     *len);
2817 +                       res = -EINVAL;
2818 +                       goto done;
2819 +               }
2820 +
2821 +               req_version->version = IP_SET_PROTOCOL_VERSION;
2822 +               res = copy_to_user(user, req_version,
2823 +                                  sizeof(struct ip_set_req_version));
2824 +               goto done;
2825 +       }
2826 +       case IP_SET_OP_GET_BYNAME: {
2827 +               struct ip_set_req_get_set *req_get
2828 +                       = (struct ip_set_req_get_set *) data;
2829 +
2830 +               if (*len != sizeof(struct ip_set_req_get_set)) {
2831 +                       ip_set_printk("invalid GET_BYNAME (want %zu, got %d)",
2832 +                                     sizeof(struct ip_set_req_get_set), *len);
2833 +                       res = -EINVAL;
2834 +                       goto done;
2835 +               }
2836 +               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
2837 +               index = ip_set_find_byname(req_get->set.name);
2838 +               req_get->set.index = index;
2839 +               goto copy;
2840 +       }
2841 +       case IP_SET_OP_GET_BYINDEX: {
2842 +               struct ip_set_req_get_set *req_get
2843 +                       = (struct ip_set_req_get_set *) data;
2844 +
2845 +               if (*len != sizeof(struct ip_set_req_get_set)) {
2846 +                       ip_set_printk("invalid GET_BYINDEX (want %zu, got %d)",
2847 +                                     sizeof(struct ip_set_req_get_set), *len);
2848 +                       res = -EINVAL;
2849 +                       goto done;
2850 +               }
2851 +               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
2852 +               index = ip_set_find_byindex(req_get->set.index);
2853 +               strncpy(req_get->set.name,
2854 +                       index == IP_SET_INVALID_ID ? ""
2855 +                       : ip_set_list[index]->name, IP_SET_MAXNAMELEN);
2856 +               goto copy;
2857 +       }
2858 +       case IP_SET_OP_ADT_GET: {
2859 +               struct ip_set_req_adt_get *req_get
2860 +                       = (struct ip_set_req_adt_get *) data;
2861 +
2862 +               if (*len != sizeof(struct ip_set_req_adt_get)) {
2863 +                       ip_set_printk("invalid ADT_GET (want %zu, got %d)",
2864 +                                     sizeof(struct ip_set_req_adt_get), *len);
2865 +                       res = -EINVAL;
2866 +                       goto done;
2867 +               }
2868 +               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
2869 +               index = ip_set_find_byname(req_get->set.name);
2870 +               if (index != IP_SET_INVALID_ID) {
2871 +                       req_get->set.index = index;
2872 +                       strncpy(req_get->typename,
2873 +                               ip_set_list[index]->type->typename,
2874 +                               IP_SET_MAXNAMELEN - 1);
2875 +               } else {
2876 +                       res = -ENOENT;
2877 +                       goto done;
2878 +               }
2879 +               goto copy;
2880 +       }
2881 +       case IP_SET_OP_MAX_SETS: {
2882 +               struct ip_set_req_max_sets *req_max_sets
2883 +                       = (struct ip_set_req_max_sets *) data;
2884 +               ip_set_id_t i;
2885 +
2886 +               if (*len != sizeof(struct ip_set_req_max_sets)) {
2887 +                       ip_set_printk("invalid MAX_SETS (want %zu, got %d)",
2888 +                                     sizeof(struct ip_set_req_max_sets), *len);
2889 +                       res = -EINVAL;
2890 +                       goto done;
2891 +               }
2892 +
2893 +               if (strcmp(req_max_sets->set.name, IPSET_TOKEN_ALL) == 0) {
2894 +                       req_max_sets->set.index = IP_SET_INVALID_ID;
2895 +               } else {
2896 +                       req_max_sets->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
2897 +                       req_max_sets->set.index =
2898 +                               ip_set_find_byname(req_max_sets->set.name);
2899 +                       if (req_max_sets->set.index == IP_SET_INVALID_ID) {
2900 +                               res = -ENOENT;
2901 +                               goto done;
2902 +                       }
2903 +               }
2904 +               req_max_sets->max_sets = ip_set_max;
2905 +               req_max_sets->sets = 0;
2906 +               for (i = 0; i < ip_set_max; i++) {
2907 +                       if (ip_set_list[i] != NULL)
2908 +                               req_max_sets->sets++;
2909 +               }
2910 +               goto copy;
2911 +       }
2912 +       case IP_SET_OP_LIST_SIZE:
2913 +       case IP_SET_OP_SAVE_SIZE: {
2914 +               struct ip_set_req_setnames *req_setnames
2915 +                       = (struct ip_set_req_setnames *) data;
2916 +               struct ip_set_name_list *name_list;
2917 +               struct ip_set *set;
2918 +               ip_set_id_t i;
2919 +               int used;
2920 +
2921 +               if (*len < sizeof(struct ip_set_req_setnames)) {
2922 +                       ip_set_printk("short LIST_SIZE (want >=%zu, got %d)",
2923 +                                     sizeof(struct ip_set_req_setnames), *len);
2924 +                       res = -EINVAL;
2925 +                       goto done;
2926 +               }
2927 +
2928 +               req_setnames->size = 0;
2929 +               used = sizeof(struct ip_set_req_setnames);
2930 +               for (i = 0; i < ip_set_max; i++) {
2931 +                       if (ip_set_list[i] == NULL)
2932 +                               continue;
2933 +                       name_list = (struct ip_set_name_list *)
2934 +                               (data + used);
2935 +                       used += sizeof(struct ip_set_name_list);
2936 +                       if (used > copylen) {
2937 +                               res = -EAGAIN;
2938 +                               goto done;
2939 +                       }
2940 +                       set = ip_set_list[i];
2941 +                       /* Fill in index, name, etc. */
2942 +                       name_list->index = i;
2943 +                       name_list->id = set->id;
2944 +                       strncpy(name_list->name,
2945 +                               set->name,
2946 +                               IP_SET_MAXNAMELEN - 1);
2947 +                       strncpy(name_list->typename,
2948 +                               set->type->typename,
2949 +                               IP_SET_MAXNAMELEN - 1);
2950 +                       DP("filled %s of type %s, index %u\n",
2951 +                          name_list->name, name_list->typename,
2952 +                          name_list->index);
2953 +                       if (!(req_setnames->index == IP_SET_INVALID_ID
2954 +                             || req_setnames->index == i))
2955 +                             continue;
2956 +                       /* Update size */
2957 +                       switch (*op) {
2958 +                       case IP_SET_OP_LIST_SIZE: {
2959 +                               req_setnames->size += sizeof(struct ip_set_list)
2960 +                                       + set->type->header_size
2961 +                                       + set->type->list_members_size(set);
2962 +                               /* Sets are identified by id in the hash */
2963 +                               FOREACH_HASH_DO(__set_hash_bindings_size_list,
2964 +                                               set->id, &req_setnames->size);
2965 +                               break;
2966 +                       }
2967 +                       case IP_SET_OP_SAVE_SIZE: {
2968 +                               req_setnames->size += sizeof(struct ip_set_save)
2969 +                                       + set->type->header_size
2970 +                                       + set->type->list_members_size(set);
2971 +                               FOREACH_HASH_DO(__set_hash_bindings_size_save,
2972 +                                               set->id, &req_setnames->size);
2973 +                               break;
2974 +                       }
2975 +                       default:
2976 +                               break;
2977 +                       }
2978 +               }
2979 +               if (copylen != used) {
2980 +                       res = -EAGAIN;
2981 +                       goto done;
2982 +               }
2983 +               goto copy;
2984 +       }
2985 +       case IP_SET_OP_LIST: {
2986 +               struct ip_set_req_list *req_list
2987 +                       = (struct ip_set_req_list *) data;
2988 +               ip_set_id_t i;
2989 +               int used;
2990 +
2991 +               if (*len < sizeof(struct ip_set_req_list)) {
2992 +                       ip_set_printk("short LIST (want >=%zu, got %d)",
2993 +                                     sizeof(struct ip_set_req_list), *len);
2994 +                       res = -EINVAL;
2995 +                       goto done;
2996 +               }
2997 +               index = req_list->index;
2998 +               if (index != IP_SET_INVALID_ID
2999 +                   && ip_set_find_byindex(index) != index) {
3000 +                       res = -ENOENT;
3001 +                       goto done;
3002 +               }
3003 +               used = 0;
3004 +               if (index == IP_SET_INVALID_ID) {
3005 +                       /* List all sets */
3006 +                       for (i = 0; i < ip_set_max && res == 0; i++) {
3007 +                               if (ip_set_list[i] != NULL)
3008 +                                       res = ip_set_list_set(i, data, &used, *len);
3009 +                       }
3010 +               } else {
3011 +                       /* List an individual set */
3012 +                       res = ip_set_list_set(index, data, &used, *len);
3013 +               }
3014 +               if (res != 0)
3015 +                       goto done;
3016 +               else if (copylen != used) {
3017 +                       res = -EAGAIN;
3018 +                       goto done;
3019 +               }
3020 +               goto copy;
3021 +       }
3022 +       case IP_SET_OP_SAVE: {
3023 +               struct ip_set_req_list *req_save
3024 +                       = (struct ip_set_req_list *) data;
3025 +               ip_set_id_t i;
3026 +               int used;
3027 +
3028 +               if (*len < sizeof(struct ip_set_req_list)) {
3029 +                       ip_set_printk("short SAVE (want >=%zu, got %d)",
3030 +                                     sizeof(struct ip_set_req_list), *len);
3031 +                       res = -EINVAL;
3032 +                       goto done;
3033 +               }
3034 +               index = req_save->index;
3035 +               if (index != IP_SET_INVALID_ID
3036 +                   && ip_set_find_byindex(index) != index) {
3037 +                       res = -ENOENT;
3038 +                       goto done;
3039 +               }
3040 +               used = 0;
3041 +               if (index == IP_SET_INVALID_ID) {
3042 +                       /* Save all sets */
3043 +                       for (i = 0; i < ip_set_max && res == 0; i++) {
3044 +                               if (ip_set_list[i] != NULL)
3045 +                                       res = ip_set_save_set(i, data, &used, *len);
3046 +                       }
3047 +               } else {
3048 +                       /* Save an individual set */
3049 +                       res = ip_set_save_set(index, data, &used, *len);
3050 +               }
3051 +               if (res == 0)
3052 +                       res = ip_set_save_bindings(index, data, &used, *len);
3053 +
3054 +               if (res != 0)
3055 +                       goto done;
3056 +               else if (copylen != used) {
3057 +                       res = -EAGAIN;
3058 +                       goto done;
3059 +               }
3060 +               goto copy;
3061 +       }
3062 +       case IP_SET_OP_RESTORE: {
3063 +               struct ip_set_req_setnames *req_restore
3064 +                       = (struct ip_set_req_setnames *) data;
3065 +               int line;
3066 +
3067 +               if (*len < sizeof(struct ip_set_req_setnames)
3068 +                   || *len != req_restore->size) {
3069 +                       ip_set_printk("invalid RESTORE (want =%zu, got %d)",
3070 +                                     req_restore->size, *len);
3071 +                       res = -EINVAL;
3072 +                       goto done;
3073 +               }
3074 +               line = ip_set_restore(data + sizeof(struct ip_set_req_setnames),
3075 +                                     req_restore->size - sizeof(struct ip_set_req_setnames));
3076 +               DP("ip_set_restore: %u", line);
3077 +               if (line != 0) {
3078 +                       res = -EAGAIN;
3079 +                       req_restore->size = line;
3080 +                       copylen = sizeof(struct ip_set_req_setnames);
3081 +                       goto copy;
3082 +               }
3083 +               goto done;
3084 +       }
3085 +       default:
3086 +               res = -EBADMSG;
3087 +               goto done;
3088 +       }       /* end of switch(op) */
3089 +
3090 +    copy:
3091 +       DP("set %s, copylen %u", index != IP_SET_INVALID_ID
3092 +                                && ip_set_list[index]
3093 +                    ? ip_set_list[index]->name
3094 +                    : ":all:", copylen);
3095 +       res = copy_to_user(user, data, copylen);
3096 +
3097 +    done:
3098 +       up(&ip_set_app_mutex);
3099 +       vfree(data);
3100 +       if (res > 0)
3101 +               res = 0;
3102 +       DP("final result %d", res);
3103 +       return res;
3104 +}
3105 +
3106 +static struct nf_sockopt_ops so_set = {
3107 +       .pf             = PF_INET,
3108 +       .set_optmin     = SO_IP_SET,
3109 +       .set_optmax     = SO_IP_SET + 1,
3110 +       .set            = &ip_set_sockfn_set,
3111 +       .get_optmin     = SO_IP_SET,
3112 +       .get_optmax     = SO_IP_SET + 1,
3113 +       .get            = &ip_set_sockfn_get,
3114 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
3115 +       .owner          = THIS_MODULE,
3116 +#endif
3117 +};
3118 +
3119 +static int max_sets, hash_size;
3120 +module_param(max_sets, int, 0600);
3121 +MODULE_PARM_DESC(max_sets, "maximal number of sets");
3122 +module_param(hash_size, int, 0600);
3123 +MODULE_PARM_DESC(hash_size, "hash size for bindings");
3124 +MODULE_LICENSE("GPL");
3125 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
3126 +MODULE_DESCRIPTION("module implementing core IP set support");
3127 +
3128 +static int __init ip_set_init(void)
3129 +{
3130 +       int res;
3131 +       ip_set_id_t i;
3132 +
3133 +       get_random_bytes(&ip_set_hash_random, 4);
3134 +       if (max_sets)
3135 +               ip_set_max = max_sets;
3136 +       ip_set_list = vmalloc(sizeof(struct ip_set *) * ip_set_max);
3137 +       if (!ip_set_list) {
3138 +               printk(KERN_ERR "Unable to create ip_set_list\n");
3139 +               return -ENOMEM;
3140 +       }
3141 +       memset(ip_set_list, 0, sizeof(struct ip_set *) * ip_set_max);
3142 +       if (hash_size)
3143 +               ip_set_bindings_hash_size = hash_size;
3144 +       ip_set_hash = vmalloc(sizeof(struct list_head) * ip_set_bindings_hash_size);
3145 +       if (!ip_set_hash) {
3146 +               printk(KERN_ERR "Unable to create ip_set_hash\n");
3147 +               vfree(ip_set_list);
3148 +               return -ENOMEM;
3149 +       }
3150 +       for (i = 0; i < ip_set_bindings_hash_size; i++)
3151 +               INIT_LIST_HEAD(&ip_set_hash[i]);
3152 +
3153 +       INIT_LIST_HEAD(&set_type_list);
3154 +
3155 +       res = nf_register_sockopt(&so_set);
3156 +       if (res != 0) {
3157 +               ip_set_printk("SO_SET registry failed: %d", res);
3158 +               vfree(ip_set_list);
3159 +               vfree(ip_set_hash);
3160 +               return res;
3161 +       }
3162 +       return 0;
3163 +}
3164 +
3165 +static void __exit ip_set_fini(void)
3166 +{
3167 +       /* There can't be any existing set or binding */
3168 +       nf_unregister_sockopt(&so_set);
3169 +       vfree(ip_set_list);
3170 +       vfree(ip_set_hash);
3171 +       DP("these are the famous last words");
3172 +}
3173 +
3174 +EXPORT_SYMBOL(ip_set_register_set_type);
3175 +EXPORT_SYMBOL(ip_set_unregister_set_type);
3176 +
3177 +EXPORT_SYMBOL(ip_set_get_byname);
3178 +EXPORT_SYMBOL(ip_set_get_byindex);
3179 +EXPORT_SYMBOL(ip_set_put);
3180 +
3181 +EXPORT_SYMBOL(ip_set_addip_kernel);
3182 +EXPORT_SYMBOL(ip_set_delip_kernel);
3183 +EXPORT_SYMBOL(ip_set_testip_kernel);
3184 +
3185 +module_init(ip_set_init);
3186 +module_exit(ip_set_fini);
3187 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set_iphash.c
3188 ===================================================================
3189 --- /dev/null
3190 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set_iphash.c
3191 @@ -0,0 +1,429 @@
3192 +/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
3193 + *
3194 + * This program is free software; you can redistribute it and/or modify
3195 + * it under the terms of the GNU General Public License version 2 as
3196 + * published by the Free Software Foundation.
3197 + */
3198 +
3199 +/* Kernel module implementing an ip hash set */
3200 +
3201 +#include <linux/module.h>
3202 +#include <linux/ip.h>
3203 +#include <linux/skbuff.h>
3204 +#include <linux/version.h>
3205 +#include <linux/jhash.h>
3206 +#include <linux/netfilter_ipv4/ip_tables.h>
3207 +#include <linux/netfilter_ipv4/ip_set.h>
3208 +#include <linux/errno.h>
3209 +#include <asm/uaccess.h>
3210 +#include <asm/bitops.h>
3211 +#include <linux/spinlock.h>
3212 +#include <linux/vmalloc.h>
3213 +#include <linux/random.h>
3214 +
3215 +#include <net/ip.h>
3216 +
3217 +#include <linux/netfilter_ipv4/ip_set_malloc.h>
3218 +#include <linux/netfilter_ipv4/ip_set_iphash.h>
3219 +
3220 +static int limit = MAX_RANGE;
3221 +
3222 +static inline __u32
3223 +jhash_ip(const struct ip_set_iphash *map, uint16_t i, ip_set_ip_t ip)
3224 +{
3225 +       return jhash_1word(ip, *(((uint32_t *) map->initval) + i));
3226 +}
3227 +
3228 +static inline __u32
3229 +hash_id(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3230 +{
3231 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3232 +       __u32 id;
3233 +       u_int16_t i;
3234 +       ip_set_ip_t *elem;
3235 +
3236 +       *hash_ip = ip & map->netmask;
3237 +       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u, %u.%u.%u.%u",
3238 +          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip), HIPQUAD(map->netmask));
3239 +
3240 +       for (i = 0; i < map->probes; i++) {
3241 +               id = jhash_ip(map, i, *hash_ip) % map->hashsize;
3242 +               DP("hash key: %u", id);
3243 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
3244 +               if (*elem == *hash_ip)
3245 +                       return id;
3246 +               /* No shortcut at testing - there can be deleted
3247 +                * entries. */
3248 +       }
3249 +       return UINT_MAX;
3250 +}
3251 +
3252 +static inline int
3253 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3254 +{
3255 +       return (ip && hash_id(set, ip, hash_ip) != UINT_MAX);
3256 +}
3257 +
3258 +static int
3259 +testip(struct ip_set *set, const void *data, size_t size,
3260 +       ip_set_ip_t *hash_ip)
3261 +{
3262 +       struct ip_set_req_iphash *req =
3263 +           (struct ip_set_req_iphash *) data;
3264 +
3265 +       if (size != sizeof(struct ip_set_req_iphash)) {
3266 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3267 +                             sizeof(struct ip_set_req_iphash),
3268 +                             size);
3269 +               return -EINVAL;
3270 +       }
3271 +       return __testip(set, req->ip, hash_ip);
3272 +}
3273 +
3274 +static int
3275 +testip_kernel(struct ip_set *set,
3276 +             const struct sk_buff *skb,
3277 +             ip_set_ip_t *hash_ip,
3278 +             const u_int32_t *flags,
3279 +             unsigned char index)
3280 +{
3281 +       return __testip(set,
3282 +                       ntohl(flags[index] & IPSET_SRC
3283 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
3284 +                               ? ip_hdr(skb)->saddr
3285 +                               : ip_hdr(skb)->daddr),
3286 +#else
3287 +                               ? skb->nh.iph->saddr
3288 +                               : skb->nh.iph->daddr),
3289 +#endif
3290 +                       hash_ip);
3291 +}
3292 +
3293 +static inline int
3294 +__addip(struct ip_set_iphash *map, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3295 +{
3296 +       __u32 probe;
3297 +       u_int16_t i;
3298 +       ip_set_ip_t *elem;
3299 +
3300 +       if (!ip || map->elements >= limit)
3301 +               return -ERANGE;
3302 +
3303 +       *hash_ip = ip & map->netmask;
3304 +
3305 +       for (i = 0; i < map->probes; i++) {
3306 +               probe = jhash_ip(map, i, *hash_ip) % map->hashsize;
3307 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, probe);
3308 +               if (*elem == *hash_ip)
3309 +                       return -EEXIST;
3310 +               if (!*elem) {
3311 +                       *elem = *hash_ip;
3312 +                       map->elements++;
3313 +                       return 0;
3314 +               }
3315 +       }
3316 +       /* Trigger rehashing */
3317 +       return -EAGAIN;
3318 +}
3319 +
3320 +static int
3321 +addip(struct ip_set *set, const void *data, size_t size,
3322 +        ip_set_ip_t *hash_ip)
3323 +{
3324 +       struct ip_set_req_iphash *req =
3325 +           (struct ip_set_req_iphash *) data;
3326 +
3327 +       if (size != sizeof(struct ip_set_req_iphash)) {
3328 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3329 +                             sizeof(struct ip_set_req_iphash),
3330 +                             size);
3331 +               return -EINVAL;
3332 +       }
3333 +       return __addip((struct ip_set_iphash *) set->data, req->ip, hash_ip);
3334 +}
3335 +
3336 +static int
3337 +addip_kernel(struct ip_set *set,
3338 +            const struct sk_buff *skb,
3339 +            ip_set_ip_t *hash_ip,
3340 +            const u_int32_t *flags,
3341 +            unsigned char index)
3342 +{
3343 +       return __addip((struct ip_set_iphash *) set->data,
3344 +                      ntohl(flags[index] & IPSET_SRC
3345 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
3346 +                               ? ip_hdr(skb)->saddr
3347 +                               : ip_hdr(skb)->daddr),
3348 +#else
3349 +                               ? skb->nh.iph->saddr
3350 +                               : skb->nh.iph->daddr),
3351 +#endif
3352 +                      hash_ip);
3353 +}
3354 +
3355 +static int retry(struct ip_set *set)
3356 +{
3357 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3358 +       ip_set_ip_t hash_ip, *elem;
3359 +       void *members;
3360 +       u_int32_t i, hashsize = map->hashsize;
3361 +       int res;
3362 +       struct ip_set_iphash *tmp;
3363 +
3364 +       if (map->resize == 0)
3365 +               return -ERANGE;
3366 +
3367 +    again:
3368 +       res = 0;
3369 +
3370 +       /* Calculate new hash size */
3371 +       hashsize += (hashsize * map->resize)/100;
3372 +       if (hashsize == map->hashsize)
3373 +               hashsize++;
3374 +
3375 +       ip_set_printk("rehashing of set %s triggered: "
3376 +                     "hashsize grows from %u to %u",
3377 +                     set->name, map->hashsize, hashsize);
3378 +
3379 +       tmp = kmalloc(sizeof(struct ip_set_iphash)
3380 +                     + map->probes * sizeof(uint32_t), GFP_ATOMIC);
3381 +       if (!tmp) {
3382 +               DP("out of memory for %d bytes",
3383 +                  sizeof(struct ip_set_iphash)
3384 +                  + map->probes * sizeof(uint32_t));
3385 +               return -ENOMEM;
3386 +       }
3387 +       tmp->members = harray_malloc(hashsize, sizeof(ip_set_ip_t), GFP_ATOMIC);
3388 +       if (!tmp->members) {
3389 +               DP("out of memory for %d bytes", hashsize * sizeof(ip_set_ip_t));
3390 +               kfree(tmp);
3391 +               return -ENOMEM;
3392 +       }
3393 +       tmp->hashsize = hashsize;
3394 +       tmp->elements = 0;
3395 +       tmp->probes = map->probes;
3396 +       tmp->resize = map->resize;
3397 +       tmp->netmask = map->netmask;
3398 +       memcpy(tmp->initval, map->initval, map->probes * sizeof(uint32_t));
3399 +
3400 +       write_lock_bh(&set->lock);
3401 +       map = (struct ip_set_iphash *) set->data; /* Play safe */
3402 +       for (i = 0; i < map->hashsize && res == 0; i++) {
3403 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);
3404 +               if (*elem)
3405 +                       res = __addip(tmp, *elem, &hash_ip);
3406 +       }
3407 +       if (res) {
3408 +               /* Failure, try again */
3409 +               write_unlock_bh(&set->lock);
3410 +               harray_free(tmp->members);
3411 +               kfree(tmp);
3412 +               goto again;
3413 +       }
3414 +
3415 +       /* Success at resizing! */
3416 +       members = map->members;
3417 +
3418 +       map->hashsize = tmp->hashsize;
3419 +       map->members = tmp->members;
3420 +       write_unlock_bh(&set->lock);
3421 +
3422 +       harray_free(members);
3423 +       kfree(tmp);
3424 +
3425 +       return 0;
3426 +}
3427 +
3428 +static inline int
3429 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3430 +{
3431 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3432 +       ip_set_ip_t id, *elem;
3433 +
3434 +       if (!ip)
3435 +               return -ERANGE;
3436 +
3437 +       id = hash_id(set, ip, hash_ip);
3438 +       if (id == UINT_MAX)
3439 +               return -EEXIST;
3440 +
3441 +       elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
3442 +       *elem = 0;
3443 +       map->elements--;
3444 +
3445 +       return 0;
3446 +}
3447 +
3448 +static int
3449 +delip(struct ip_set *set, const void *data, size_t size,
3450 +        ip_set_ip_t *hash_ip)
3451 +{
3452 +       struct ip_set_req_iphash *req =
3453 +           (struct ip_set_req_iphash *) data;
3454 +
3455 +       if (size != sizeof(struct ip_set_req_iphash)) {
3456 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3457 +                             sizeof(struct ip_set_req_iphash),
3458 +                             size);
3459 +               return -EINVAL;
3460 +       }
3461 +       return __delip(set, req->ip, hash_ip);
3462 +}
3463 +
3464 +static int
3465 +delip_kernel(struct ip_set *set,
3466 +            const struct sk_buff *skb,
3467 +            ip_set_ip_t *hash_ip,
3468 +            const u_int32_t *flags,
3469 +            unsigned char index)
3470 +{
3471 +       return __delip(set,
3472 +                      ntohl(flags[index] & IPSET_SRC
3473 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
3474 +                               ? ip_hdr(skb)->saddr
3475 +                               : ip_hdr(skb)->daddr),
3476 +#else
3477 +                               ? skb->nh.iph->saddr
3478 +                               : skb->nh.iph->daddr),
3479 +#endif
3480 +                      hash_ip);
3481 +}
3482 +
3483 +static int create(struct ip_set *set, const void *data, size_t size)
3484 +{
3485 +       struct ip_set_req_iphash_create *req =
3486 +           (struct ip_set_req_iphash_create *) data;
3487 +       struct ip_set_iphash *map;
3488 +       uint16_t i;
3489 +
3490 +       if (size != sizeof(struct ip_set_req_iphash_create)) {
3491 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3492 +                              sizeof(struct ip_set_req_iphash_create),
3493 +                              size);
3494 +               return -EINVAL;
3495 +       }
3496 +
3497 +       if (req->hashsize < 1) {
3498 +               ip_set_printk("hashsize too small");
3499 +               return -ENOEXEC;
3500 +       }
3501 +
3502 +       if (req->probes < 1) {
3503 +               ip_set_printk("probes too small");
3504 +               return -ENOEXEC;
3505 +       }
3506 +
3507 +       map = kmalloc(sizeof(struct ip_set_iphash)
3508 +                     + req->probes * sizeof(uint32_t), GFP_KERNEL);
3509 +       if (!map) {
3510 +               DP("out of memory for %d bytes",
3511 +                  sizeof(struct ip_set_iphash)
3512 +                  + req->probes * sizeof(uint32_t));
3513 +               return -ENOMEM;
3514 +       }
3515 +       for (i = 0; i < req->probes; i++)
3516 +               get_random_bytes(((uint32_t *) map->initval)+i, 4);
3517 +       map->elements = 0;
3518 +       map->hashsize = req->hashsize;
3519 +       map->probes = req->probes;
3520 +       map->resize = req->resize;
3521 +       map->netmask = req->netmask;
3522 +       map->members = harray_malloc(map->hashsize, sizeof(ip_set_ip_t), GFP_KERNEL);
3523 +       if (!map->members) {
3524 +               DP("out of memory for %d bytes", map->hashsize * sizeof(ip_set_ip_t));
3525 +               kfree(map);
3526 +               return -ENOMEM;
3527 +       }
3528 +
3529 +       set->data = map;
3530 +       return 0;
3531 +}
3532 +
3533 +static void destroy(struct ip_set *set)
3534 +{
3535 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3536 +
3537 +       harray_free(map->members);
3538 +       kfree(map);
3539 +
3540 +       set->data = NULL;
3541 +}
3542 +
3543 +static void flush(struct ip_set *set)
3544 +{
3545 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3546 +       harray_flush(map->members, map->hashsize, sizeof(ip_set_ip_t));
3547 +       map->elements = 0;
3548 +}
3549 +
3550 +static void list_header(const struct ip_set *set, void *data)
3551 +{
3552 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3553 +       struct ip_set_req_iphash_create *header =
3554 +           (struct ip_set_req_iphash_create *) data;
3555 +
3556 +       header->hashsize = map->hashsize;
3557 +       header->probes = map->probes;
3558 +       header->resize = map->resize;
3559 +       header->netmask = map->netmask;
3560 +}
3561 +
3562 +static int list_members_size(const struct ip_set *set)
3563 +{
3564 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3565 +
3566 +       return (map->hashsize * sizeof(ip_set_ip_t));
3567 +}
3568 +
3569 +static void list_members(const struct ip_set *set, void *data)
3570 +{
3571 +       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
3572 +       ip_set_ip_t i, *elem;
3573 +
3574 +       for (i = 0; i < map->hashsize; i++) {
3575 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);
3576 +               ((ip_set_ip_t *)data)[i] = *elem;
3577 +       }
3578 +}
3579 +
3580 +static struct ip_set_type ip_set_iphash = {
3581 +       .typename               = SETTYPE_NAME,
3582 +       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
3583 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
3584 +       .create                 = &create,
3585 +       .destroy                = &destroy,
3586 +       .flush                  = &flush,
3587 +       .reqsize                = sizeof(struct ip_set_req_iphash),
3588 +       .addip                  = &addip,
3589 +       .addip_kernel           = &addip_kernel,
3590 +       .retry                  = &retry,
3591 +       .delip                  = &delip,
3592 +       .delip_kernel           = &delip_kernel,
3593 +       .testip                 = &testip,
3594 +       .testip_kernel          = &testip_kernel,
3595 +       .header_size            = sizeof(struct ip_set_req_iphash_create),
3596 +       .list_header            = &list_header,
3597 +       .list_members_size      = &list_members_size,
3598 +       .list_members           = &list_members,
3599 +       .me                     = THIS_MODULE,
3600 +};
3601 +
3602 +MODULE_LICENSE("GPL");
3603 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
3604 +MODULE_DESCRIPTION("iphash type of IP sets");
3605 +module_param(limit, int, 0600);
3606 +MODULE_PARM_DESC(limit, "maximal number of elements stored in the sets");
3607 +
3608 +static int __init ip_set_iphash_init(void)
3609 +{
3610 +       return ip_set_register_set_type(&ip_set_iphash);
3611 +}
3612 +
3613 +static void __exit ip_set_iphash_fini(void)
3614 +{
3615 +       /* FIXME: possible race with ip_set_create() */
3616 +       ip_set_unregister_set_type(&ip_set_iphash);
3617 +}
3618 +
3619 +module_init(ip_set_iphash_init);
3620 +module_exit(ip_set_iphash_fini);
3621 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set_ipmap.c
3622 ===================================================================
3623 --- /dev/null
3624 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set_ipmap.c
3625 @@ -0,0 +1,336 @@
3626 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
3627 + *                         Patrick Schaaf <bof@bof.de>
3628 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
3629 + *
3630 + * This program is free software; you can redistribute it and/or modify
3631 + * it under the terms of the GNU General Public License version 2 as
3632 + * published by the Free Software Foundation.
3633 + */
3634 +
3635 +/* Kernel module implementing an IP set type: the single bitmap type */
3636 +
3637 +#include <linux/module.h>
3638 +#include <linux/ip.h>
3639 +#include <linux/skbuff.h>
3640 +#include <linux/version.h>
3641 +#include <linux/netfilter_ipv4/ip_tables.h>
3642 +#include <linux/netfilter_ipv4/ip_set.h>
3643 +#include <linux/errno.h>
3644 +#include <asm/uaccess.h>
3645 +#include <asm/bitops.h>
3646 +#include <linux/spinlock.h>
3647 +
3648 +#include <linux/netfilter_ipv4/ip_set_ipmap.h>
3649 +
3650 +static inline ip_set_ip_t
3651 +ip_to_id(const struct ip_set_ipmap *map, ip_set_ip_t ip)
3652 +{
3653 +       return (ip - map->first_ip)/map->hosts;
3654 +}
3655 +
3656 +static inline int
3657 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3658 +{
3659 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3660 +
3661 +       if (ip < map->first_ip || ip > map->last_ip)
3662 +               return -ERANGE;
3663 +
3664 +       *hash_ip = ip & map->netmask;
3665 +       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
3666 +          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip));
3667 +       return !!test_bit(ip_to_id(map, *hash_ip), map->members);
3668 +}
3669 +
3670 +static int
3671 +testip(struct ip_set *set, const void *data, size_t size,
3672 +       ip_set_ip_t *hash_ip)
3673 +{
3674 +       struct ip_set_req_ipmap *req =
3675 +           (struct ip_set_req_ipmap *) data;
3676 +
3677 +       if (size != sizeof(struct ip_set_req_ipmap)) {
3678 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3679 +                             sizeof(struct ip_set_req_ipmap),
3680 +                             size);
3681 +               return -EINVAL;
3682 +       }
3683 +       return __testip(set, req->ip, hash_ip);
3684 +}
3685 +
3686 +static int
3687 +testip_kernel(struct ip_set *set,
3688 +             const struct sk_buff *skb,
3689 +             ip_set_ip_t *hash_ip,
3690 +             const u_int32_t *flags,
3691 +             unsigned char index)
3692 +{
3693 +       int res =  __testip(set,
3694 +                       ntohl(flags[index] & IPSET_SRC
3695 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
3696 +                               ? ip_hdr(skb)->saddr
3697 +                               : ip_hdr(skb)->daddr),
3698 +#else
3699 +                               ? skb->nh.iph->saddr
3700 +                               : skb->nh.iph->daddr),
3701 +#endif
3702 +                       hash_ip);
3703 +       return (res < 0 ? 0 : res);
3704 +}
3705 +
3706 +static inline int
3707 +__addip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3708 +{
3709 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3710 +
3711 +       if (ip < map->first_ip || ip > map->last_ip)
3712 +               return -ERANGE;
3713 +
3714 +       *hash_ip = ip & map->netmask;
3715 +       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
3716 +       if (test_and_set_bit(ip_to_id(map, *hash_ip), map->members))
3717 +               return -EEXIST;
3718 +
3719 +       return 0;
3720 +}
3721 +
3722 +static int
3723 +addip(struct ip_set *set, const void *data, size_t size,
3724 +      ip_set_ip_t *hash_ip)
3725 +{
3726 +       struct ip_set_req_ipmap *req =
3727 +           (struct ip_set_req_ipmap *) data;
3728 +
3729 +       if (size != sizeof(struct ip_set_req_ipmap)) {
3730 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3731 +                             sizeof(struct ip_set_req_ipmap),
3732 +                             size);
3733 +               return -EINVAL;
3734 +       }
3735 +       DP("%u.%u.%u.%u", HIPQUAD(req->ip));
3736 +       return __addip(set, req->ip, hash_ip);
3737 +}
3738 +
3739 +static int
3740 +addip_kernel(struct ip_set *set,
3741 +            const struct sk_buff *skb,
3742 +            ip_set_ip_t *hash_ip,
3743 +            const u_int32_t *flags,
3744 +            unsigned char index)
3745 +{
3746 +       return __addip(set,
3747 +                      ntohl(flags[index] & IPSET_SRC
3748 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
3749 +                               ? ip_hdr(skb)->saddr
3750 +                               : ip_hdr(skb)->daddr),
3751 +#else
3752 +                               ? skb->nh.iph->saddr
3753 +                               : skb->nh.iph->daddr),
3754 +#endif
3755 +                      hash_ip);
3756 +}
3757 +
3758 +static inline int
3759 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
3760 +{
3761 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3762 +
3763 +       if (ip < map->first_ip || ip > map->last_ip)
3764 +               return -ERANGE;
3765 +
3766 +       *hash_ip = ip & map->netmask;
3767 +       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
3768 +       if (!test_and_clear_bit(ip_to_id(map, *hash_ip), map->members))
3769 +               return -EEXIST;
3770 +
3771 +       return 0;
3772 +}
3773 +
3774 +static int
3775 +delip(struct ip_set *set, const void *data, size_t size,
3776 +      ip_set_ip_t *hash_ip)
3777 +{
3778 +       struct ip_set_req_ipmap *req =
3779 +           (struct ip_set_req_ipmap *) data;
3780 +
3781 +       if (size != sizeof(struct ip_set_req_ipmap)) {
3782 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3783 +                             sizeof(struct ip_set_req_ipmap),
3784 +                             size);
3785 +               return -EINVAL;
3786 +       }
3787 +       return __delip(set, req->ip, hash_ip);
3788 +}
3789 +
3790 +static int
3791 +delip_kernel(struct ip_set *set,
3792 +            const struct sk_buff *skb,
3793 +            ip_set_ip_t *hash_ip,
3794 +            const u_int32_t *flags,
3795 +            unsigned char index)
3796 +{
3797 +       return __delip(set,
3798 +                      ntohl(flags[index] & IPSET_SRC
3799 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
3800 +                               ? ip_hdr(skb)->saddr
3801 +                               : ip_hdr(skb)->daddr),
3802 +#else
3803 +                               ? skb->nh.iph->saddr
3804 +                               : skb->nh.iph->daddr),
3805 +#endif
3806 +                      hash_ip);
3807 +}
3808 +
3809 +static int create(struct ip_set *set, const void *data, size_t size)
3810 +{
3811 +       int newbytes;
3812 +       struct ip_set_req_ipmap_create *req =
3813 +           (struct ip_set_req_ipmap_create *) data;
3814 +       struct ip_set_ipmap *map;
3815 +
3816 +       if (size != sizeof(struct ip_set_req_ipmap_create)) {
3817 +               ip_set_printk("data length wrong (want %zu, have %zu)",
3818 +                             sizeof(struct ip_set_req_ipmap_create),
3819 +                             size);
3820 +               return -EINVAL;
3821 +       }
3822 +
3823 +       DP("from %u.%u.%u.%u to %u.%u.%u.%u",
3824 +          HIPQUAD(req->from), HIPQUAD(req->to));
3825 +
3826 +       if (req->from > req->to) {
3827 +               DP("bad ip range");
3828 +               return -ENOEXEC;
3829 +       }
3830 +
3831 +       map = kmalloc(sizeof(struct ip_set_ipmap), GFP_KERNEL);
3832 +       if (!map) {
3833 +               DP("out of memory for %d bytes",
3834 +                  sizeof(struct ip_set_ipmap));
3835 +               return -ENOMEM;
3836 +       }
3837 +       map->first_ip = req->from;
3838 +       map->last_ip = req->to;
3839 +       map->netmask = req->netmask;
3840 +
3841 +       if (req->netmask == 0xFFFFFFFF) {
3842 +               map->hosts = 1;
3843 +               map->sizeid = map->last_ip - map->first_ip + 1;
3844 +       } else {
3845 +               unsigned int mask_bits, netmask_bits;
3846 +               ip_set_ip_t mask;
3847 +
3848 +               map->first_ip &= map->netmask;  /* Should we better bark? */
3849 +
3850 +               mask = range_to_mask(map->first_ip, map->last_ip, &mask_bits);
3851 +               netmask_bits = mask_to_bits(map->netmask);
3852 +
3853 +               if ((!mask && (map->first_ip || map->last_ip != 0xFFFFFFFF))
3854 +                   || netmask_bits <= mask_bits)
3855 +                       return -ENOEXEC;
3856 +
3857 +               DP("mask_bits %u, netmask_bits %u",
3858 +                  mask_bits, netmask_bits);
3859 +               map->hosts = 2 << (32 - netmask_bits - 1);
3860 +               map->sizeid = 2 << (netmask_bits - mask_bits - 1);
3861 +       }
3862 +       if (map->sizeid > MAX_RANGE + 1) {
3863 +               ip_set_printk("range too big (max %d addresses)",
3864 +                              MAX_RANGE+1);
3865 +               kfree(map);
3866 +               return -ENOEXEC;
3867 +       }
3868 +       DP("hosts %u, sizeid %u", map->hosts, map->sizeid);
3869 +       newbytes = bitmap_bytes(0, map->sizeid - 1);
3870 +       map->members = kmalloc(newbytes, GFP_KERNEL);
3871 +       if (!map->members) {
3872 +               DP("out of memory for %d bytes", newbytes);
3873 +               kfree(map);
3874 +               return -ENOMEM;
3875 +       }
3876 +       memset(map->members, 0, newbytes);
3877 +
3878 +       set->data = map;
3879 +       return 0;
3880 +}
3881 +
3882 +static void destroy(struct ip_set *set)
3883 +{
3884 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3885 +
3886 +       kfree(map->members);
3887 +       kfree(map);
3888 +
3889 +       set->data = NULL;
3890 +}
3891 +
3892 +static void flush(struct ip_set *set)
3893 +{
3894 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3895 +       memset(map->members, 0, bitmap_bytes(0, map->sizeid - 1));
3896 +}
3897 +
3898 +static void list_header(const struct ip_set *set, void *data)
3899 +{
3900 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3901 +       struct ip_set_req_ipmap_create *header =
3902 +           (struct ip_set_req_ipmap_create *) data;
3903 +
3904 +       header->from = map->first_ip;
3905 +       header->to = map->last_ip;
3906 +       header->netmask = map->netmask;
3907 +}
3908 +
3909 +static int list_members_size(const struct ip_set *set)
3910 +{
3911 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3912 +
3913 +       return bitmap_bytes(0, map->sizeid - 1);
3914 +}
3915 +
3916 +static void list_members(const struct ip_set *set, void *data)
3917 +{
3918 +       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
3919 +       int bytes = bitmap_bytes(0, map->sizeid - 1);
3920 +
3921 +       memcpy(data, map->members, bytes);
3922 +}
3923 +
3924 +static struct ip_set_type ip_set_ipmap = {
3925 +       .typename               = SETTYPE_NAME,
3926 +       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
3927 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
3928 +       .create                 = &create,
3929 +       .destroy                = &destroy,
3930 +       .flush                  = &flush,
3931 +       .reqsize                = sizeof(struct ip_set_req_ipmap),
3932 +       .addip                  = &addip,
3933 +       .addip_kernel           = &addip_kernel,
3934 +       .delip                  = &delip,
3935 +       .delip_kernel           = &delip_kernel,
3936 +       .testip                 = &testip,
3937 +       .testip_kernel          = &testip_kernel,
3938 +       .header_size            = sizeof(struct ip_set_req_ipmap_create),
3939 +       .list_header            = &list_header,
3940 +       .list_members_size      = &list_members_size,
3941 +       .list_members           = &list_members,
3942 +       .me                     = THIS_MODULE,
3943 +};
3944 +
3945 +MODULE_LICENSE("GPL");
3946 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
3947 +MODULE_DESCRIPTION("ipmap type of IP sets");
3948 +
3949 +static int __init ip_set_ipmap_init(void)
3950 +{
3951 +       return ip_set_register_set_type(&ip_set_ipmap);
3952 +}
3953 +
3954 +static void __exit ip_set_ipmap_fini(void)
3955 +{
3956 +       /* FIXME: possible race with ip_set_create() */
3957 +       ip_set_unregister_set_type(&ip_set_ipmap);
3958 +}
3959 +
3960 +module_init(ip_set_ipmap_init);
3961 +module_exit(ip_set_ipmap_fini);
3962 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set_ipporthash.c
3963 ===================================================================
3964 --- /dev/null
3965 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set_ipporthash.c
3966 @@ -0,0 +1,581 @@
3967 +/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
3968 + *
3969 + * This program is free software; you can redistribute it and/or modify
3970 + * it under the terms of the GNU General Public License version 2 as
3971 + * published by the Free Software Foundation.
3972 + */
3973 +
3974 +/* Kernel module implementing an ip+port hash set */
3975 +
3976 +#include <linux/module.h>
3977 +#include <linux/ip.h>
3978 +#include <linux/tcp.h>
3979 +#include <linux/udp.h>
3980 +#include <linux/skbuff.h>
3981 +#include <linux/version.h>
3982 +#include <linux/jhash.h>
3983 +#include <linux/netfilter_ipv4/ip_tables.h>
3984 +#include <linux/netfilter_ipv4/ip_set.h>
3985 +#include <linux/errno.h>
3986 +#include <asm/uaccess.h>
3987 +#include <asm/bitops.h>
3988 +#include <linux/spinlock.h>
3989 +#include <linux/vmalloc.h>
3990 +#include <linux/random.h>
3991 +
3992 +#include <net/ip.h>
3993 +
3994 +#include <linux/netfilter_ipv4/ip_set_malloc.h>
3995 +#include <linux/netfilter_ipv4/ip_set_ipporthash.h>
3996 +
3997 +static int limit = MAX_RANGE;
3998 +
3999 +/* We must handle non-linear skbs */
4000 +static inline ip_set_ip_t
4001 +get_port(const struct sk_buff *skb, u_int32_t flags)
4002 +{
4003 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4004 +       struct iphdr *iph = ip_hdr(skb);
4005 +#else
4006 +       struct iphdr *iph = skb->nh.iph;
4007 +#endif
4008 +       u_int16_t offset = ntohs(iph->frag_off) & IP_OFFSET;
4009 +
4010 +       switch (iph->protocol) {
4011 +       case IPPROTO_TCP: {
4012 +               struct tcphdr tcph;
4013 +
4014 +               /* See comments at tcp_match in ip_tables.c */
4015 +               if (offset)
4016 +                       return INVALID_PORT;
4017 +
4018 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4019 +               if (skb_copy_bits(skb, ip_hdr(skb)->ihl*4, &tcph, sizeof(tcph)) < 0)
4020 +#else
4021 +               if (skb_copy_bits(skb, skb->nh.iph->ihl*4, &tcph, sizeof(tcph)) < 0)
4022 +#endif
4023 +                       /* No choice either */
4024 +                       return INVALID_PORT;
4025 +
4026 +               return ntohs(flags & IPSET_SRC ?
4027 +                            tcph.source : tcph.dest);
4028 +           }
4029 +       case IPPROTO_UDP: {
4030 +               struct udphdr udph;
4031 +
4032 +               if (offset)
4033 +                       return INVALID_PORT;
4034 +
4035 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4036 +               if (skb_copy_bits(skb, ip_hdr(skb)->ihl*4, &udph, sizeof(udph)) < 0)
4037 +#else
4038 +               if (skb_copy_bits(skb, skb->nh.iph->ihl*4, &udph, sizeof(udph)) < 0)
4039 +#endif
4040 +                       /* No choice either */
4041 +                       return INVALID_PORT;
4042 +
4043 +               return ntohs(flags & IPSET_SRC ?
4044 +                            udph.source : udph.dest);
4045 +           }
4046 +       default:
4047 +               return INVALID_PORT;
4048 +       }
4049 +}
4050 +
4051 +static inline __u32
4052 +jhash_ip(const struct ip_set_ipporthash *map, uint16_t i, ip_set_ip_t ip)
4053 +{
4054 +       return jhash_1word(ip, *(((uint32_t *) map->initval) + i));
4055 +}
4056 +
4057 +#define HASH_IP(map, ip, port) (port + ((ip - ((map)->first_ip)) << 16))
4058 +
4059 +static inline __u32
4060 +hash_id(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t port,
4061 +       ip_set_ip_t *hash_ip)
4062 +{
4063 +       struct ip_set_ipporthash *map =
4064 +               (struct ip_set_ipporthash *) set->data;
4065 +       __u32 id;
4066 +       u_int16_t i;
4067 +       ip_set_ip_t *elem;
4068 +
4069 +       *hash_ip = HASH_IP(map, ip, port);
4070 +       DP("set: %s, ipport:%u.%u.%u.%u:%u, %u.%u.%u.%u",
4071 +          set->name, HIPQUAD(ip), port, HIPQUAD(*hash_ip));
4072 +
4073 +       for (i = 0; i < map->probes; i++) {
4074 +               id = jhash_ip(map, i, *hash_ip) % map->hashsize;
4075 +               DP("hash key: %u", id);
4076 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
4077 +               if (*elem == *hash_ip)
4078 +                       return id;
4079 +               /* No shortcut at testing - there can be deleted
4080 +                * entries. */
4081 +       }
4082 +       return UINT_MAX;
4083 +}
4084 +
4085 +static inline int
4086 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t port,
4087 +        ip_set_ip_t *hash_ip)
4088 +{
4089 +       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
4090 +
4091 +       if (ip < map->first_ip || ip > map->last_ip)
4092 +               return -ERANGE;
4093 +
4094 +       return (hash_id(set, ip, port, hash_ip) != UINT_MAX);
4095 +}
4096 +
4097 +static int
4098 +testip(struct ip_set *set, const void *data, size_t size,
4099 +       ip_set_ip_t *hash_ip)
4100 +{
4101 +       struct ip_set_req_ipporthash *req =
4102 +           (struct ip_set_req_ipporthash *) data;
4103 +
4104 +       if (size != sizeof(struct ip_set_req_ipporthash)) {
4105 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4106 +                             sizeof(struct ip_set_req_ipporthash),
4107 +                             size);
4108 +               return -EINVAL;
4109 +       }
4110 +       return __testip(set, req->ip, req->port, hash_ip);
4111 +}
4112 +
4113 +static int
4114 +testip_kernel(struct ip_set *set,
4115 +             const struct sk_buff *skb,
4116 +             ip_set_ip_t *hash_ip,
4117 +             const u_int32_t *flags,
4118 +             unsigned char index)
4119 +{
4120 +       ip_set_ip_t port;
4121 +       int res;
4122 +
4123 +       if (flags[index+1] == 0)
4124 +               return 0;
4125 +
4126 +       port = get_port(skb, flags[index+1]);
4127 +
4128 +       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
4129 +          flags[index] & IPSET_SRC ? "SRC" : "DST",
4130 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4131 +          NIPQUAD(ip_hdr(skb)->saddr),
4132 +          NIPQUAD(ip_hdr(skb)->daddr));
4133 +#else
4134 +          NIPQUAD(skb->nh.iph->saddr),
4135 +          NIPQUAD(skb->nh.iph->daddr));
4136 +#endif
4137 +       DP("flag %s port %u",
4138 +          flags[index+1] & IPSET_SRC ? "SRC" : "DST",
4139 +          port);
4140 +       if (port == INVALID_PORT)
4141 +               return 0;
4142 +
4143 +       res =  __testip(set,
4144 +                       ntohl(flags[index] & IPSET_SRC
4145 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4146 +                                       ? ip_hdr(skb)->saddr
4147 +                                       : ip_hdr(skb)->daddr),
4148 +#else
4149 +                                       ? skb->nh.iph->saddr
4150 +                                       : skb->nh.iph->daddr),
4151 +#endif
4152 +                       port,
4153 +                       hash_ip);
4154 +       return (res < 0 ? 0 : res);
4155 +
4156 +}
4157 +
4158 +static inline int
4159 +__add_haship(struct ip_set_ipporthash *map, ip_set_ip_t hash_ip)
4160 +{
4161 +       __u32 probe;
4162 +       u_int16_t i;
4163 +       ip_set_ip_t *elem;
4164 +
4165 +       for (i = 0; i < map->probes; i++) {
4166 +               probe = jhash_ip(map, i, hash_ip) % map->hashsize;
4167 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, probe);
4168 +               if (*elem == hash_ip)
4169 +                       return -EEXIST;
4170 +               if (!*elem) {
4171 +                       *elem = hash_ip;
4172 +                       map->elements++;
4173 +                       return 0;
4174 +               }
4175 +       }
4176 +       /* Trigger rehashing */
4177 +       return -EAGAIN;
4178 +}
4179 +
4180 +static inline int
4181 +__addip(struct ip_set_ipporthash *map, ip_set_ip_t ip, ip_set_ip_t port,
4182 +       ip_set_ip_t *hash_ip)
4183 +{
4184 +       if (map->elements > limit)
4185 +               return -ERANGE;
4186 +       if (ip < map->first_ip || ip > map->last_ip)
4187 +               return -ERANGE;
4188 +
4189 +       *hash_ip = HASH_IP(map, ip, port);
4190 +
4191 +       return __add_haship(map, *hash_ip);
4192 +}
4193 +
4194 +static int
4195 +addip(struct ip_set *set, const void *data, size_t size,
4196 +        ip_set_ip_t *hash_ip)
4197 +{
4198 +       struct ip_set_req_ipporthash *req =
4199 +           (struct ip_set_req_ipporthash *) data;
4200 +
4201 +       if (size != sizeof(struct ip_set_req_ipporthash)) {
4202 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4203 +                             sizeof(struct ip_set_req_ipporthash),
4204 +                             size);
4205 +               return -EINVAL;
4206 +       }
4207 +       return __addip((struct ip_set_ipporthash *) set->data,
4208 +                       req->ip, req->port, hash_ip);
4209 +}
4210 +
4211 +static int
4212 +addip_kernel(struct ip_set *set,
4213 +            const struct sk_buff *skb,
4214 +            ip_set_ip_t *hash_ip,
4215 +            const u_int32_t *flags,
4216 +            unsigned char index)
4217 +{
4218 +       ip_set_ip_t port;
4219 +
4220 +       if (flags[index+1] == 0)
4221 +               return -EINVAL;
4222 +
4223 +       port = get_port(skb, flags[index+1]);
4224 +
4225 +       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
4226 +          flags[index] & IPSET_SRC ? "SRC" : "DST",
4227 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4228 +          NIPQUAD(ip_hdr(skb)->saddr),
4229 +          NIPQUAD(ip_hdr(skb)->daddr));
4230 +#else
4231 +          NIPQUAD(skb->nh.iph->saddr),
4232 +          NIPQUAD(skb->nh.iph->daddr));
4233 +#endif
4234 +       DP("flag %s port %u",
4235 +          flags[index+1] & IPSET_SRC ? "SRC" : "DST",
4236 +          port);
4237 +       if (port == INVALID_PORT)
4238 +               return -EINVAL;
4239 +
4240 +       return __addip((struct ip_set_ipporthash *) set->data,
4241 +                      ntohl(flags[index] & IPSET_SRC
4242 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4243 +                               ? ip_hdr(skb)->saddr
4244 +                               : ip_hdr(skb)->daddr),
4245 +#else
4246 +                               ? skb->nh.iph->saddr
4247 +                               : skb->nh.iph->daddr),
4248 +#endif
4249 +                      port,
4250 +                      hash_ip);
4251 +}
4252 +
4253 +static int retry(struct ip_set *set)
4254 +{
4255 +       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
4256 +       ip_set_ip_t *elem;
4257 +       void *members;
4258 +       u_int32_t i, hashsize = map->hashsize;
4259 +       int res;
4260 +       struct ip_set_ipporthash *tmp;
4261 +
4262 +       if (map->resize == 0)
4263 +               return -ERANGE;
4264 +
4265 +    again:
4266 +       res = 0;
4267 +
4268 +       /* Calculate new hash size */
4269 +       hashsize += (hashsize * map->resize)/100;
4270 +       if (hashsize == map->hashsize)
4271 +               hashsize++;
4272 +
4273 +       ip_set_printk("rehashing of set %s triggered: "
4274 +                     "hashsize grows from %u to %u",
4275 +                     set->name, map->hashsize, hashsize);
4276 +
4277 +       tmp = kmalloc(sizeof(struct ip_set_ipporthash)
4278 +                     + map->probes * sizeof(uint32_t), GFP_ATOMIC);
4279 +       if (!tmp) {
4280 +               DP("out of memory for %d bytes",
4281 +                  sizeof(struct ip_set_ipporthash)
4282 +                  + map->probes * sizeof(uint32_t));
4283 +               return -ENOMEM;
4284 +       }
4285 +       tmp->members = harray_malloc(hashsize, sizeof(ip_set_ip_t), GFP_ATOMIC);
4286 +       if (!tmp->members) {
4287 +               DP("out of memory for %d bytes", hashsize * sizeof(ip_set_ip_t));
4288 +               kfree(tmp);
4289 +               return -ENOMEM;
4290 +       }
4291 +       tmp->hashsize = hashsize;
4292 +       tmp->elements = 0;
4293 +       tmp->probes = map->probes;
4294 +       tmp->resize = map->resize;
4295 +       tmp->first_ip = map->first_ip;
4296 +       tmp->last_ip = map->last_ip;
4297 +       memcpy(tmp->initval, map->initval, map->probes * sizeof(uint32_t));
4298 +
4299 +       write_lock_bh(&set->lock);
4300 +       map = (struct ip_set_ipporthash *) set->data; /* Play safe */
4301 +       for (i = 0; i < map->hashsize && res == 0; i++) {
4302 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);
4303 +               if (*elem)
4304 +                       res = __add_haship(tmp, *elem);
4305 +       }
4306 +       if (res) {
4307 +               /* Failure, try again */
4308 +               write_unlock_bh(&set->lock);
4309 +               harray_free(tmp->members);
4310 +               kfree(tmp);
4311 +               goto again;
4312 +       }
4313 +
4314 +       /* Success at resizing! */
4315 +       members = map->members;
4316 +
4317 +       map->hashsize = tmp->hashsize;
4318 +       map->members = tmp->members;
4319 +       write_unlock_bh(&set->lock);
4320 +
4321 +       harray_free(members);
4322 +       kfree(tmp);
4323 +
4324 +       return 0;
4325 +}
4326 +
4327 +static inline int
4328 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t port,
4329 +       ip_set_ip_t *hash_ip)
4330 +{
4331 +       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
4332 +       ip_set_ip_t id;
4333 +       ip_set_ip_t *elem;
4334 +
4335 +       if (ip < map->first_ip || ip > map->last_ip)
4336 +               return -ERANGE;
4337 +
4338 +       id = hash_id(set, ip, port, hash_ip);
4339 +
4340 +       if (id == UINT_MAX)
4341 +               return -EEXIST;
4342 +
4343 +       elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
4344 +       *elem = 0;
4345 +       map->elements--;
4346 +
4347 +       return 0;
4348 +}
4349 +
4350 +static int
4351 +delip(struct ip_set *set, const void *data, size_t size,
4352 +        ip_set_ip_t *hash_ip)
4353 +{
4354 +       struct ip_set_req_ipporthash *req =
4355 +           (struct ip_set_req_ipporthash *) data;
4356 +
4357 +       if (size != sizeof(struct ip_set_req_ipporthash)) {
4358 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4359 +                             sizeof(struct ip_set_req_ipporthash),
4360 +                             size);
4361 +               return -EINVAL;
4362 +       }
4363 +       return __delip(set, req->ip, req->port, hash_ip);
4364 +}
4365 +
4366 +static int
4367 +delip_kernel(struct ip_set *set,
4368 +            const struct sk_buff *skb,
4369 +            ip_set_ip_t *hash_ip,
4370 +            const u_int32_t *flags,
4371 +            unsigned char index)
4372 +{
4373 +       ip_set_ip_t port;
4374 +
4375 +       if (flags[index+1] == 0)
4376 +               return -EINVAL;
4377 +
4378 +       port = get_port(skb, flags[index+1]);
4379 +
4380 +       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
4381 +          flags[index] & IPSET_SRC ? "SRC" : "DST",
4382 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4383 +          NIPQUAD(ip_hdr(skb)->saddr),
4384 +          NIPQUAD(ip_hdr(skb)->daddr));
4385 +#else
4386 +          NIPQUAD(skb->nh.iph->saddr),
4387 +          NIPQUAD(skb->nh.iph->daddr));
4388 +#endif
4389 +       DP("flag %s port %u",
4390 +          flags[index+1] & IPSET_SRC ? "SRC" : "DST",
4391 +          port);
4392 +       if (port == INVALID_PORT)
4393 +               return -EINVAL;
4394 +
4395 +       return __delip(set,
4396 +                      ntohl(flags[index] & IPSET_SRC
4397 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4398 +                               ? ip_hdr(skb)->saddr
4399 +                               : ip_hdr(skb)->daddr),
4400 +#else
4401 +                               ? skb->nh.iph->saddr
4402 +                               : skb->nh.iph->daddr),
4403 +#endif
4404 +                      port,
4405 +                      hash_ip);
4406 +}
4407 +
4408 +static int create(struct ip_set *set, const void *data, size_t size)
4409 +{
4410 +       struct ip_set_req_ipporthash_create *req =
4411 +           (struct ip_set_req_ipporthash_create *) data;
4412 +       struct ip_set_ipporthash *map;
4413 +       uint16_t i;
4414 +
4415 +       if (size != sizeof(struct ip_set_req_ipporthash_create)) {
4416 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4417 +                              sizeof(struct ip_set_req_ipporthash_create),
4418 +                              size);
4419 +               return -EINVAL;
4420 +       }
4421 +
4422 +       if (req->hashsize < 1) {
4423 +               ip_set_printk("hashsize too small");
4424 +               return -ENOEXEC;
4425 +       }
4426 +
4427 +       if (req->probes < 1) {
4428 +               ip_set_printk("probes too small");
4429 +               return -ENOEXEC;
4430 +       }
4431 +
4432 +       map = kmalloc(sizeof(struct ip_set_ipporthash)
4433 +                     + req->probes * sizeof(uint32_t), GFP_KERNEL);
4434 +       if (!map) {
4435 +               DP("out of memory for %d bytes",
4436 +                  sizeof(struct ip_set_ipporthash)
4437 +                  + req->probes * sizeof(uint32_t));
4438 +               return -ENOMEM;
4439 +       }
4440 +       for (i = 0; i < req->probes; i++)
4441 +               get_random_bytes(((uint32_t *) map->initval)+i, 4);
4442 +       map->elements = 0;
4443 +       map->hashsize = req->hashsize;
4444 +       map->probes = req->probes;
4445 +       map->resize = req->resize;
4446 +       map->first_ip = req->from;
4447 +       map->last_ip = req->to;
4448 +       map->members = harray_malloc(map->hashsize, sizeof(ip_set_ip_t), GFP_KERNEL);
4449 +       if (!map->members) {
4450 +               DP("out of memory for %d bytes", map->hashsize * sizeof(ip_set_ip_t));
4451 +               kfree(map);
4452 +               return -ENOMEM;
4453 +       }
4454 +
4455 +       set->data = map;
4456 +       return 0;
4457 +}
4458 +
4459 +static void destroy(struct ip_set *set)
4460 +{
4461 +       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
4462 +
4463 +       harray_free(map->members);
4464 +       kfree(map);
4465 +
4466 +       set->data = NULL;
4467 +}
4468 +
4469 +static void flush(struct ip_set *set)
4470 +{
4471 +       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
4472 +       harray_flush(map->members, map->hashsize, sizeof(ip_set_ip_t));
4473 +       map->elements = 0;
4474 +}
4475 +
4476 +static void list_header(const struct ip_set *set, void *data)
4477 +{
4478 +       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
4479 +       struct ip_set_req_ipporthash_create *header =
4480 +           (struct ip_set_req_ipporthash_create *) data;
4481 +
4482 +       header->hashsize = map->hashsize;
4483 +       header->probes = map->probes;
4484 +       header->resize = map->resize;
4485 +       header->from = map->first_ip;
4486 +       header->to = map->last_ip;
4487 +}
4488 +
4489 +static int list_members_size(const struct ip_set *set)
4490 +{
4491 +       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
4492 +
4493 +       return (map->hashsize * sizeof(ip_set_ip_t));
4494 +}
4495 +
4496 +static void list_members(const struct ip_set *set, void *data)
4497 +{
4498 +       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
4499 +       ip_set_ip_t i, *elem;
4500 +
4501 +       for (i = 0; i < map->hashsize; i++) {
4502 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);
4503 +               ((ip_set_ip_t *)data)[i] = *elem;
4504 +       }
4505 +}
4506 +
4507 +static struct ip_set_type ip_set_ipporthash = {
4508 +       .typename               = SETTYPE_NAME,
4509 +       .features               = IPSET_TYPE_IP | IPSET_TYPE_PORT | IPSET_DATA_DOUBLE,
4510 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
4511 +       .create                 = &create,
4512 +       .destroy                = &destroy,
4513 +       .flush                  = &flush,
4514 +       .reqsize                = sizeof(struct ip_set_req_ipporthash),
4515 +       .addip                  = &addip,
4516 +       .addip_kernel           = &addip_kernel,
4517 +       .retry                  = &retry,
4518 +       .delip                  = &delip,
4519 +       .delip_kernel           = &delip_kernel,
4520 +       .testip                 = &testip,
4521 +       .testip_kernel          = &testip_kernel,
4522 +       .header_size            = sizeof(struct ip_set_req_ipporthash_create),
4523 +       .list_header            = &list_header,
4524 +       .list_members_size      = &list_members_size,
4525 +       .list_members           = &list_members,
4526 +       .me                     = THIS_MODULE,
4527 +};
4528 +
4529 +MODULE_LICENSE("GPL");
4530 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
4531 +MODULE_DESCRIPTION("ipporthash type of IP sets");
4532 +module_param(limit, int, 0600);
4533 +MODULE_PARM_DESC(limit, "maximal number of elements stored in the sets");
4534 +
4535 +static int __init ip_set_ipporthash_init(void)
4536 +{
4537 +       return ip_set_register_set_type(&ip_set_ipporthash);
4538 +}
4539 +
4540 +static void __exit ip_set_ipporthash_fini(void)
4541 +{
4542 +       /* FIXME: possible race with ip_set_create() */
4543 +       ip_set_unregister_set_type(&ip_set_ipporthash);
4544 +}
4545 +
4546 +module_init(ip_set_ipporthash_init);
4547 +module_exit(ip_set_ipporthash_fini);
4548 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set_iptree.c
4549 ===================================================================
4550 --- /dev/null
4551 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set_iptree.c
4552 @@ -0,0 +1,612 @@
4553 +/* Copyright (C) 2005 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
4554 + *
4555 + * This program is free software; you can redistribute it and/or modify
4556 + * it under the terms of the GNU General Public License version 2 as
4557 + * published by the Free Software Foundation.
4558 + */
4559 +
4560 +/* Kernel module implementing an IP set type: the iptree type */
4561 +
4562 +#include <linux/version.h>
4563 +#include <linux/module.h>
4564 +#include <linux/ip.h>
4565 +#include <linux/skbuff.h>
4566 +#include <linux/slab.h>
4567 +#include <linux/delay.h>
4568 +#include <linux/netfilter_ipv4/ip_tables.h>
4569 +#include <linux/netfilter_ipv4/ip_set.h>
4570 +#include <linux/errno.h>
4571 +#include <asm/uaccess.h>
4572 +#include <asm/bitops.h>
4573 +#include <linux/spinlock.h>
4574 +
4575 +/* Backward compatibility */
4576 +#ifndef __nocast
4577 +#define __nocast
4578 +#endif
4579 +
4580 +#include <linux/netfilter_ipv4/ip_set_iptree.h>
4581 +
4582 +static int limit = MAX_RANGE;
4583 +
4584 +/* Garbage collection interval in seconds: */
4585 +#define IPTREE_GC_TIME         5*60
4586 +/* Sleep so many milliseconds before trying again
4587 + * to delete the gc timer at destroying/flushing a set */
4588 +#define IPTREE_DESTROY_SLEEP   100
4589 +
4590 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
4591 +static struct kmem_cache *branch_cachep;
4592 +static struct kmem_cache *leaf_cachep;
4593 +#else
4594 +static kmem_cache_t *branch_cachep;
4595 +static kmem_cache_t *leaf_cachep;
4596 +#endif
4597 +
4598 +#if defined(__LITTLE_ENDIAN)
4599 +#define ABCD(a,b,c,d,addrp) do {               \
4600 +       a = ((unsigned char *)addrp)[3];        \
4601 +       b = ((unsigned char *)addrp)[2];        \
4602 +       c = ((unsigned char *)addrp)[1];        \
4603 +       d = ((unsigned char *)addrp)[0];        \
4604 +} while (0)
4605 +#elif defined(__BIG_ENDIAN)
4606 +#define ABCD(a,b,c,d,addrp) do {               \
4607 +       a = ((unsigned char *)addrp)[0];        \
4608 +       b = ((unsigned char *)addrp)[1];        \
4609 +       c = ((unsigned char *)addrp)[2];        \
4610 +       d = ((unsigned char *)addrp)[3];        \
4611 +} while (0)
4612 +#else
4613 +#error "Please fix asm/byteorder.h"
4614 +#endif /* __LITTLE_ENDIAN */
4615 +
4616 +#define TESTIP_WALK(map, elem, branch) do {    \
4617 +       if ((map)->tree[elem]) {                \
4618 +               branch = (map)->tree[elem];     \
4619 +       } else                                  \
4620 +               return 0;                       \
4621 +} while (0)
4622 +
4623 +static inline int
4624 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
4625 +{
4626 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4627 +       struct ip_set_iptreeb *btree;
4628 +       struct ip_set_iptreec *ctree;
4629 +       struct ip_set_iptreed *dtree;
4630 +       unsigned char a,b,c,d;
4631 +
4632 +       if (!ip)
4633 +               return -ERANGE;
4634 +
4635 +       *hash_ip = ip;
4636 +       ABCD(a, b, c, d, hash_ip);
4637 +       DP("%u %u %u %u timeout %u", a, b, c, d, map->timeout);
4638 +       TESTIP_WALK(map, a, btree);
4639 +       TESTIP_WALK(btree, b, ctree);
4640 +       TESTIP_WALK(ctree, c, dtree);
4641 +       DP("%lu %lu", dtree->expires[d], jiffies);
4642 +       return dtree->expires[d]
4643 +              && (!map->timeout
4644 +                  || time_after(dtree->expires[d], jiffies));
4645 +}
4646 +
4647 +static int
4648 +testip(struct ip_set *set, const void *data, size_t size,
4649 +       ip_set_ip_t *hash_ip)
4650 +{
4651 +       struct ip_set_req_iptree *req =
4652 +           (struct ip_set_req_iptree *) data;
4653 +
4654 +       if (size != sizeof(struct ip_set_req_iptree)) {
4655 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4656 +                             sizeof(struct ip_set_req_iptree),
4657 +                             size);
4658 +               return -EINVAL;
4659 +       }
4660 +       return __testip(set, req->ip, hash_ip);
4661 +}
4662 +
4663 +static int
4664 +testip_kernel(struct ip_set *set,
4665 +             const struct sk_buff *skb,
4666 +             ip_set_ip_t *hash_ip,
4667 +             const u_int32_t *flags,
4668 +             unsigned char index)
4669 +{
4670 +       int res;
4671 +
4672 +       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
4673 +          flags[index] & IPSET_SRC ? "SRC" : "DST",
4674 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4675 +          NIPQUAD(ip_hdr(skb)->saddr),
4676 +          NIPQUAD(ip_hdr(skb)->daddr));
4677 +#else
4678 +          NIPQUAD(skb->nh.iph->saddr),
4679 +          NIPQUAD(skb->nh.iph->daddr));
4680 +#endif
4681 +
4682 +       res =  __testip(set,
4683 +                       ntohl(flags[index] & IPSET_SRC
4684 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4685 +                               ? ip_hdr(skb)->saddr
4686 +                               : ip_hdr(skb)->daddr),
4687 +#else
4688 +                               ? skb->nh.iph->saddr
4689 +                               : skb->nh.iph->daddr),
4690 +#endif
4691 +                       hash_ip);
4692 +       return (res < 0 ? 0 : res);
4693 +}
4694 +
4695 +#define ADDIP_WALK(map, elem, branch, type, cachep) do {       \
4696 +       if ((map)->tree[elem]) {                                \
4697 +               DP("found %u", elem);                           \
4698 +               branch = (map)->tree[elem];                     \
4699 +       } else {                                                \
4700 +               branch = (type *)                               \
4701 +                       kmem_cache_alloc(cachep, GFP_ATOMIC);   \
4702 +               if (branch == NULL)                             \
4703 +                       return -ENOMEM;                         \
4704 +               memset(branch, 0, sizeof(*branch));             \
4705 +               (map)->tree[elem] = branch;                     \
4706 +               DP("alloc %u", elem);                           \
4707 +       }                                                       \
4708 +} while (0)
4709 +
4710 +static inline int
4711 +__addip(struct ip_set *set, ip_set_ip_t ip, unsigned int timeout,
4712 +       ip_set_ip_t *hash_ip)
4713 +{
4714 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4715 +       struct ip_set_iptreeb *btree;
4716 +       struct ip_set_iptreec *ctree;
4717 +       struct ip_set_iptreed *dtree;
4718 +       unsigned char a,b,c,d;
4719 +       int ret = 0;
4720 +
4721 +       if (!ip || map->elements >= limit)
4722 +               /* We could call the garbage collector
4723 +                * but it's probably overkill */
4724 +               return -ERANGE;
4725 +
4726 +       *hash_ip = ip;
4727 +       ABCD(a, b, c, d, hash_ip);
4728 +       DP("%u %u %u %u timeout %u", a, b, c, d, timeout);
4729 +       ADDIP_WALK(map, a, btree, struct ip_set_iptreeb, branch_cachep);
4730 +       ADDIP_WALK(btree, b, ctree, struct ip_set_iptreec, branch_cachep);
4731 +       ADDIP_WALK(ctree, c, dtree, struct ip_set_iptreed, leaf_cachep);
4732 +       if (dtree->expires[d]
4733 +           && (!map->timeout || time_after(dtree->expires[d], jiffies)))
4734 +               ret = -EEXIST;
4735 +       dtree->expires[d] = map->timeout ? (timeout * HZ + jiffies) : 1;
4736 +       /* Lottery: I won! */
4737 +       if (dtree->expires[d] == 0)
4738 +               dtree->expires[d] = 1;
4739 +       DP("%u %lu", d, dtree->expires[d]);
4740 +       if (ret == 0)
4741 +               map->elements++;
4742 +       return ret;
4743 +}
4744 +
4745 +static int
4746 +addip(struct ip_set *set, const void *data, size_t size,
4747 +      ip_set_ip_t *hash_ip)
4748 +{
4749 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4750 +       struct ip_set_req_iptree *req =
4751 +               (struct ip_set_req_iptree *) data;
4752 +
4753 +       if (size != sizeof(struct ip_set_req_iptree)) {
4754 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4755 +                             sizeof(struct ip_set_req_iptree),
4756 +                             size);
4757 +               return -EINVAL;
4758 +       }
4759 +       DP("%u.%u.%u.%u %u", HIPQUAD(req->ip), req->timeout);
4760 +       return __addip(set, req->ip,
4761 +                      req->timeout ? req->timeout : map->timeout,
4762 +                      hash_ip);
4763 +}
4764 +
4765 +static int
4766 +addip_kernel(struct ip_set *set,
4767 +            const struct sk_buff *skb,
4768 +            ip_set_ip_t *hash_ip,
4769 +            const u_int32_t *flags,
4770 +            unsigned char index)
4771 +{
4772 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4773 +
4774 +       return __addip(set,
4775 +                      ntohl(flags[index] & IPSET_SRC
4776 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4777 +                               ? ip_hdr(skb)->saddr
4778 +                               : ip_hdr(skb)->daddr),
4779 +#else
4780 +                               ? skb->nh.iph->saddr
4781 +                               : skb->nh.iph->daddr),
4782 +#endif
4783 +                      map->timeout,
4784 +                      hash_ip);
4785 +}
4786 +
4787 +#define DELIP_WALK(map, elem, branch) do {     \
4788 +       if ((map)->tree[elem]) {                \
4789 +               branch = (map)->tree[elem];     \
4790 +       } else                                  \
4791 +               return -EEXIST;                 \
4792 +} while (0)
4793 +
4794 +static inline int
4795 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
4796 +{
4797 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4798 +       struct ip_set_iptreeb *btree;
4799 +       struct ip_set_iptreec *ctree;
4800 +       struct ip_set_iptreed *dtree;
4801 +       unsigned char a,b,c,d;
4802 +
4803 +       if (!ip)
4804 +               return -ERANGE;
4805 +
4806 +       *hash_ip = ip;
4807 +       ABCD(a, b, c, d, hash_ip);
4808 +       DELIP_WALK(map, a, btree);
4809 +       DELIP_WALK(btree, b, ctree);
4810 +       DELIP_WALK(ctree, c, dtree);
4811 +
4812 +       if (dtree->expires[d]) {
4813 +               dtree->expires[d] = 0;
4814 +               map->elements--;
4815 +               return 0;
4816 +       }
4817 +       return -EEXIST;
4818 +}
4819 +
4820 +static int
4821 +delip(struct ip_set *set, const void *data, size_t size,
4822 +      ip_set_ip_t *hash_ip)
4823 +{
4824 +       struct ip_set_req_iptree *req =
4825 +           (struct ip_set_req_iptree *) data;
4826 +
4827 +       if (size != sizeof(struct ip_set_req_iptree)) {
4828 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4829 +                             sizeof(struct ip_set_req_iptree),
4830 +                             size);
4831 +               return -EINVAL;
4832 +       }
4833 +       return __delip(set, req->ip, hash_ip);
4834 +}
4835 +
4836 +static int
4837 +delip_kernel(struct ip_set *set,
4838 +            const struct sk_buff *skb,
4839 +            ip_set_ip_t *hash_ip,
4840 +            const u_int32_t *flags,
4841 +            unsigned char index)
4842 +{
4843 +       return __delip(set,
4844 +                      ntohl(flags[index] & IPSET_SRC
4845 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
4846 +                               ? ip_hdr(skb)->saddr
4847 +                               : ip_hdr(skb)->daddr),
4848 +#else
4849 +                               ? skb->nh.iph->saddr
4850 +                               : skb->nh.iph->daddr),
4851 +#endif
4852 +                      hash_ip);
4853 +}
4854 +
4855 +#define LOOP_WALK_BEGIN(map, i, branch) \
4856 +       for (i = 0; i < 256; i++) {     \
4857 +               if (!(map)->tree[i])    \
4858 +                       continue;       \
4859 +               branch = (map)->tree[i]
4860 +
4861 +#define LOOP_WALK_END }
4862 +
4863 +static void ip_tree_gc(unsigned long ul_set)
4864 +{
4865 +       struct ip_set *set = (void *) ul_set;
4866 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4867 +       struct ip_set_iptreeb *btree;
4868 +       struct ip_set_iptreec *ctree;
4869 +       struct ip_set_iptreed *dtree;
4870 +       unsigned int a,b,c,d;
4871 +       unsigned char i,j,k;
4872 +
4873 +       i = j = k = 0;
4874 +       DP("gc: %s", set->name);
4875 +       write_lock_bh(&set->lock);
4876 +       LOOP_WALK_BEGIN(map, a, btree);
4877 +       LOOP_WALK_BEGIN(btree, b, ctree);
4878 +       LOOP_WALK_BEGIN(ctree, c, dtree);
4879 +       for (d = 0; d < 256; d++) {
4880 +               if (dtree->expires[d]) {
4881 +                       DP("gc: %u %u %u %u: expires %lu jiffies %lu",
4882 +                           a, b, c, d,
4883 +                           dtree->expires[d], jiffies);
4884 +                       if (map->timeout
4885 +                           && time_before(dtree->expires[d], jiffies)) {
4886 +                               dtree->expires[d] = 0;
4887 +                               map->elements--;
4888 +                       } else
4889 +                               k = 1;
4890 +               }
4891 +       }
4892 +       if (k == 0) {
4893 +               DP("gc: %s: leaf %u %u %u empty",
4894 +                   set->name, a, b, c);
4895 +               kmem_cache_free(leaf_cachep, dtree);
4896 +               ctree->tree[c] = NULL;
4897 +       } else {
4898 +               DP("gc: %s: leaf %u %u %u not empty",
4899 +                   set->name, a, b, c);
4900 +               j = 1;
4901 +               k = 0;
4902 +       }
4903 +       LOOP_WALK_END;
4904 +       if (j == 0) {
4905 +               DP("gc: %s: branch %u %u empty",
4906 +                   set->name, a, b);
4907 +               kmem_cache_free(branch_cachep, ctree);
4908 +               btree->tree[b] = NULL;
4909 +       } else {
4910 +               DP("gc: %s: branch %u %u not empty",
4911 +                   set->name, a, b);
4912 +               i = 1;
4913 +               j = k = 0;
4914 +       }
4915 +       LOOP_WALK_END;
4916 +       if (i == 0) {
4917 +               DP("gc: %s: branch %u empty",
4918 +                   set->name, a);
4919 +               kmem_cache_free(branch_cachep, btree);
4920 +               map->tree[a] = NULL;
4921 +       } else {
4922 +               DP("gc: %s: branch %u not empty",
4923 +                   set->name, a);
4924 +               i = j = k = 0;
4925 +       }
4926 +       LOOP_WALK_END;
4927 +       write_unlock_bh(&set->lock);
4928 +
4929 +       map->gc.expires = jiffies + map->gc_interval * HZ;
4930 +       add_timer(&map->gc);
4931 +}
4932 +
4933 +static inline void init_gc_timer(struct ip_set *set)
4934 +{
4935 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4936 +
4937 +       /* Even if there is no timeout for the entries,
4938 +        * we still have to call gc because delete
4939 +        * do not clean up empty branches */
4940 +       map->gc_interval = IPTREE_GC_TIME;
4941 +       init_timer(&map->gc);
4942 +       map->gc.data = (unsigned long) set;
4943 +       map->gc.function = ip_tree_gc;
4944 +       map->gc.expires = jiffies + map->gc_interval * HZ;
4945 +       add_timer(&map->gc);
4946 +}
4947 +
4948 +static int create(struct ip_set *set, const void *data, size_t size)
4949 +{
4950 +       struct ip_set_req_iptree_create *req =
4951 +           (struct ip_set_req_iptree_create *) data;
4952 +       struct ip_set_iptree *map;
4953 +
4954 +       if (size != sizeof(struct ip_set_req_iptree_create)) {
4955 +               ip_set_printk("data length wrong (want %zu, have %zu)",
4956 +                             sizeof(struct ip_set_req_iptree_create),
4957 +                             size);
4958 +               return -EINVAL;
4959 +       }
4960 +
4961 +       map = kmalloc(sizeof(struct ip_set_iptree), GFP_KERNEL);
4962 +       if (!map) {
4963 +               DP("out of memory for %d bytes",
4964 +                  sizeof(struct ip_set_iptree));
4965 +               return -ENOMEM;
4966 +       }
4967 +       memset(map, 0, sizeof(*map));
4968 +       map->timeout = req->timeout;
4969 +       map->elements = 0;
4970 +       set->data = map;
4971 +
4972 +       init_gc_timer(set);
4973 +
4974 +       return 0;
4975 +}
4976 +
4977 +static void __flush(struct ip_set_iptree *map)
4978 +{
4979 +       struct ip_set_iptreeb *btree;
4980 +       struct ip_set_iptreec *ctree;
4981 +       struct ip_set_iptreed *dtree;
4982 +       unsigned int a,b,c;
4983 +
4984 +       LOOP_WALK_BEGIN(map, a, btree);
4985 +       LOOP_WALK_BEGIN(btree, b, ctree);
4986 +       LOOP_WALK_BEGIN(ctree, c, dtree);
4987 +       kmem_cache_free(leaf_cachep, dtree);
4988 +       LOOP_WALK_END;
4989 +       kmem_cache_free(branch_cachep, ctree);
4990 +       LOOP_WALK_END;
4991 +       kmem_cache_free(branch_cachep, btree);
4992 +       LOOP_WALK_END;
4993 +       map->elements = 0;
4994 +}
4995 +
4996 +static void destroy(struct ip_set *set)
4997 +{
4998 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
4999 +
5000 +       /* gc might be running */
5001 +       while (!del_timer(&map->gc))
5002 +               msleep(IPTREE_DESTROY_SLEEP);
5003 +       __flush(map);
5004 +       kfree(map);
5005 +       set->data = NULL;
5006 +}
5007 +
5008 +static void flush(struct ip_set *set)
5009 +{
5010 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
5011 +       unsigned int timeout = map->timeout;
5012 +
5013 +       /* gc might be running */
5014 +       while (!del_timer(&map->gc))
5015 +               msleep(IPTREE_DESTROY_SLEEP);
5016 +       __flush(map);
5017 +       memset(map, 0, sizeof(*map));
5018 +       map->timeout = timeout;
5019 +
5020 +       init_gc_timer(set);
5021 +}
5022 +
5023 +static void list_header(const struct ip_set *set, void *data)
5024 +{
5025 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
5026 +       struct ip_set_req_iptree_create *header =
5027 +           (struct ip_set_req_iptree_create *) data;
5028 +
5029 +       header->timeout = map->timeout;
5030 +}
5031 +
5032 +static int list_members_size(const struct ip_set *set)
5033 +{
5034 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
5035 +       struct ip_set_iptreeb *btree;
5036 +       struct ip_set_iptreec *ctree;
5037 +       struct ip_set_iptreed *dtree;
5038 +       unsigned int a,b,c,d;
5039 +       unsigned int count = 0;
5040 +
5041 +       LOOP_WALK_BEGIN(map, a, btree);
5042 +       LOOP_WALK_BEGIN(btree, b, ctree);
5043 +       LOOP_WALK_BEGIN(ctree, c, dtree);
5044 +       for (d = 0; d < 256; d++) {
5045 +               if (dtree->expires[d]
5046 +                   && (!map->timeout || time_after(dtree->expires[d], jiffies)))
5047 +                       count++;
5048 +       }
5049 +       LOOP_WALK_END;
5050 +       LOOP_WALK_END;
5051 +       LOOP_WALK_END;
5052 +
5053 +       DP("members %u", count);
5054 +       return (count * sizeof(struct ip_set_req_iptree));
5055 +}
5056 +
5057 +static void list_members(const struct ip_set *set, void *data)
5058 +{
5059 +       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
5060 +       struct ip_set_iptreeb *btree;
5061 +       struct ip_set_iptreec *ctree;
5062 +       struct ip_set_iptreed *dtree;
5063 +       unsigned int a,b,c,d;
5064 +       size_t offset = 0;
5065 +       struct ip_set_req_iptree *entry;
5066 +
5067 +       LOOP_WALK_BEGIN(map, a, btree);
5068 +       LOOP_WALK_BEGIN(btree, b, ctree);
5069 +       LOOP_WALK_BEGIN(ctree, c, dtree);
5070 +       for (d = 0; d < 256; d++) {
5071 +               if (dtree->expires[d]
5072 +                   && (!map->timeout || time_after(dtree->expires[d], jiffies))) {
5073 +                       entry = (struct ip_set_req_iptree *)(data + offset);
5074 +                       entry->ip = ((a << 24) | (b << 16) | (c << 8) | d);
5075 +                       entry->timeout = !map->timeout ? 0
5076 +                               : (dtree->expires[d] - jiffies)/HZ;
5077 +                       offset += sizeof(struct ip_set_req_iptree);
5078 +               }
5079 +       }
5080 +       LOOP_WALK_END;
5081 +       LOOP_WALK_END;
5082 +       LOOP_WALK_END;
5083 +}
5084 +
5085 +static struct ip_set_type ip_set_iptree = {
5086 +       .typename               = SETTYPE_NAME,
5087 +       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
5088 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
5089 +       .create                 = &create,
5090 +       .destroy                = &destroy,
5091 +       .flush                  = &flush,
5092 +       .reqsize                = sizeof(struct ip_set_req_iptree),
5093 +       .addip                  = &addip,
5094 +       .addip_kernel           = &addip_kernel,
5095 +       .delip                  = &delip,
5096 +       .delip_kernel           = &delip_kernel,
5097 +       .testip                 = &testip,
5098 +       .testip_kernel          = &testip_kernel,
5099 +       .header_size            = sizeof(struct ip_set_req_iptree_create),
5100 +       .list_header            = &list_header,
5101 +       .list_members_size      = &list_members_size,
5102 +       .list_members           = &list_members,
5103 +       .me                     = THIS_MODULE,
5104 +};
5105 +
5106 +MODULE_LICENSE("GPL");
5107 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
5108 +MODULE_DESCRIPTION("iptree type of IP sets");
5109 +module_param(limit, int, 0600);
5110 +MODULE_PARM_DESC(limit, "maximal number of elements stored in the sets");
5111 +
5112 +static int __init ip_set_iptree_init(void)
5113 +{
5114 +       int ret;
5115 +
5116 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
5117 +       branch_cachep = kmem_cache_create("ip_set_iptreeb",
5118 +                               sizeof(struct ip_set_iptreeb),
5119 +                               0, 0, NULL);
5120 +#else
5121 +       branch_cachep = kmem_cache_create("ip_set_iptreeb",
5122 +                               sizeof(struct ip_set_iptreeb),
5123 +                               0, 0, NULL, NULL);
5124 +#endif
5125 +       if (!branch_cachep) {
5126 +               printk(KERN_ERR "Unable to create ip_set_iptreeb slab cache\n");
5127 +               ret = -ENOMEM;
5128 +               goto out;
5129 +       }
5130 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
5131 +       leaf_cachep = kmem_cache_create("ip_set_iptreed",
5132 +                               sizeof(struct ip_set_iptreed),
5133 +                               0, 0, NULL);
5134 +#else
5135 +       leaf_cachep = kmem_cache_create("ip_set_iptreed",
5136 +                               sizeof(struct ip_set_iptreed),
5137 +                               0, 0, NULL, NULL);
5138 +#endif
5139 +       if (!leaf_cachep) {
5140 +               printk(KERN_ERR "Unable to create ip_set_iptreed slab cache\n");
5141 +               ret = -ENOMEM;
5142 +               goto free_branch;
5143 +       }
5144 +       ret = ip_set_register_set_type(&ip_set_iptree);
5145 +       if (ret == 0)
5146 +               goto out;
5147 +
5148 +       kmem_cache_destroy(leaf_cachep);
5149 +    free_branch:
5150 +       kmem_cache_destroy(branch_cachep);
5151 +    out:
5152 +       return ret;
5153 +}
5154 +
5155 +static void __exit ip_set_iptree_fini(void)
5156 +{
5157 +       /* FIXME: possible race with ip_set_create() */
5158 +       ip_set_unregister_set_type(&ip_set_iptree);
5159 +       kmem_cache_destroy(leaf_cachep);
5160 +       kmem_cache_destroy(branch_cachep);
5161 +}
5162 +
5163 +module_init(ip_set_iptree_init);
5164 +module_exit(ip_set_iptree_fini);
5165 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set_iptreemap.c
5166 ===================================================================
5167 --- /dev/null
5168 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set_iptreemap.c
5169 @@ -0,0 +1,829 @@
5170 +/* Copyright (C) 2007 Sven Wegener <sven.wegener@stealer.net>
5171 + *
5172 + * This program is free software; you can redistribute it and/or modify it
5173 + * under the terms of the GNU General Public License version 2 as published by
5174 + * the Free Software Foundation.
5175 + */
5176 +
5177 +/* This modules implements the iptreemap ipset type. It uses bitmaps to
5178 + * represent every single IPv4 address as a single bit. The bitmaps are managed
5179 + * in a tree structure, where the first three octets of an addresses are used
5180 + * as an index to find the bitmap and the last octet is used as the bit number.
5181 + */
5182 +
5183 +#include <linux/version.h>
5184 +#include <linux/module.h>
5185 +#include <linux/ip.h>
5186 +#include <linux/skbuff.h>
5187 +#include <linux/slab.h>
5188 +#include <linux/delay.h>
5189 +#include <linux/netfilter_ipv4/ip_tables.h>
5190 +#include <linux/netfilter_ipv4/ip_set.h>
5191 +#include <linux/errno.h>
5192 +#include <asm/uaccess.h>
5193 +#include <asm/bitops.h>
5194 +#include <linux/spinlock.h>
5195 +
5196 +#include <linux/netfilter_ipv4/ip_set_iptreemap.h>
5197 +
5198 +#define IPTREEMAP_DEFAULT_GC_TIME (5 * 60)
5199 +#define IPTREEMAP_DESTROY_SLEEP (100)
5200 +
5201 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
5202 +static struct kmem_cache *cachep_b;
5203 +static struct kmem_cache *cachep_c;
5204 +static struct kmem_cache *cachep_d;
5205 +#else
5206 +static kmem_cache_t *cachep_b;
5207 +static kmem_cache_t *cachep_c;
5208 +static kmem_cache_t *cachep_d;
5209 +#endif
5210 +
5211 +static struct ip_set_iptreemap_d *fullbitmap_d;
5212 +static struct ip_set_iptreemap_c *fullbitmap_c;
5213 +static struct ip_set_iptreemap_b *fullbitmap_b;
5214 +
5215 +#if defined(__LITTLE_ENDIAN)
5216 +#define ABCD(a, b, c, d, addr) \
5217 +       do { \
5218 +               a = ((unsigned char *)addr)[3]; \
5219 +               b = ((unsigned char *)addr)[2]; \
5220 +               c = ((unsigned char *)addr)[1]; \
5221 +               d = ((unsigned char *)addr)[0]; \
5222 +       } while (0)
5223 +#elif defined(__BIG_ENDIAN)
5224 +#define ABCD(a,b,c,d,addrp) do {               \
5225 +       a = ((unsigned char *)addrp)[0];        \
5226 +       b = ((unsigned char *)addrp)[1];        \
5227 +       c = ((unsigned char *)addrp)[2];        \
5228 +       d = ((unsigned char *)addrp)[3];        \
5229 +} while (0)
5230 +#else
5231 +#error "Please fix asm/byteorder.h"
5232 +#endif /* __LITTLE_ENDIAN */
5233 +
5234 +#define TESTIP_WALK(map, elem, branch, full) \
5235 +       do { \
5236 +               branch = (map)->tree[elem]; \
5237 +               if (!branch) \
5238 +                       return 0; \
5239 +               else if (branch == full) \
5240 +                       return 1; \
5241 +       } while (0)
5242 +
5243 +#define ADDIP_WALK(map, elem, branch, type, cachep, full) \
5244 +       do { \
5245 +               branch = (map)->tree[elem]; \
5246 +               if (!branch) { \
5247 +                       branch = (type *) kmem_cache_alloc(cachep, GFP_ATOMIC); \
5248 +                       if (!branch) \
5249 +                               return -ENOMEM; \
5250 +                       memset(branch, 0, sizeof(*branch)); \
5251 +                       (map)->tree[elem] = branch; \
5252 +               } else if (branch == full) { \
5253 +                       return -EEXIST; \
5254 +               } \
5255 +       } while (0)
5256 +
5257 +#define ADDIP_RANGE_LOOP(map, a, a1, a2, hint, branch, full, cachep, free) \
5258 +       for (a = a1; a <= a2; a++) { \
5259 +               branch = (map)->tree[a]; \
5260 +               if (branch != full) { \
5261 +                       if ((a > a1 && a < a2) || (hint)) { \
5262 +                               if (branch) \
5263 +                                       free(branch); \
5264 +                               (map)->tree[a] = full; \
5265 +                               continue; \
5266 +                       } else if (!branch) { \
5267 +                               branch = kmem_cache_alloc(cachep, GFP_ATOMIC); \
5268 +                               if (!branch) \
5269 +                                       return -ENOMEM; \
5270 +                               memset(branch, 0, sizeof(*branch)); \
5271 +                               (map)->tree[a] = branch; \
5272 +                       }
5273 +
5274 +#define ADDIP_RANGE_LOOP_END() \
5275 +               } \
5276 +       }
5277 +
5278 +#define DELIP_WALK(map, elem, branch, cachep, full, flags) \
5279 +       do { \
5280 +               branch = (map)->tree[elem]; \
5281 +               if (!branch) { \
5282 +                       return -EEXIST; \
5283 +               } else if (branch == full) { \
5284 +                       branch = kmem_cache_alloc(cachep, flags); \
5285 +                       if (!branch) \
5286 +                               return -ENOMEM; \
5287 +                       memcpy(branch, full, sizeof(*full)); \
5288 +                       (map)->tree[elem] = branch; \
5289 +               } \
5290 +       } while (0)
5291 +
5292 +#define DELIP_RANGE_LOOP(map, a, a1, a2, hint, branch, full, cachep, free, flags) \
5293 +       for (a = a1; a <= a2; a++) { \
5294 +               branch = (map)->tree[a]; \
5295 +               if (branch) { \
5296 +                       if ((a > a1 && a < a2) || (hint)) { \
5297 +                               if (branch != full) \
5298 +                                       free(branch); \
5299 +                               (map)->tree[a] = NULL; \
5300 +                               continue; \
5301 +                       } else if (branch == full) { \
5302 +                               branch = kmem_cache_alloc(cachep, flags); \
5303 +                               if (!branch) \
5304 +                                       return -ENOMEM; \
5305 +                               memcpy(branch, full, sizeof(*branch)); \
5306 +                               (map)->tree[a] = branch; \
5307 +                       }
5308 +
5309 +#define DELIP_RANGE_LOOP_END() \
5310 +               } \
5311 +       }
5312 +
5313 +#define LOOP_WALK_BEGIN(map, i, branch) \
5314 +       for (i = 0; i < 256; i++) { \
5315 +               branch = (map)->tree[i]; \
5316 +               if (likely(!branch)) \
5317 +                       continue;
5318 +
5319 +#define LOOP_WALK_END() \
5320 +       }
5321 +
5322 +#define LOOP_WALK_BEGIN_GC(map, i, branch, full, cachep, count) \
5323 +       count = -256; \
5324 +       for (i = 0; i < 256; i++) { \
5325 +               branch = (map)->tree[i]; \
5326 +               if (likely(!branch)) \
5327 +                       continue; \
5328 +               count++; \
5329 +               if (branch == full) { \
5330 +                       count++; \
5331 +                       continue; \
5332 +               }
5333 +
5334 +#define LOOP_WALK_END_GC(map, i, branch, full, cachep, count) \
5335 +               if (-256 == count) { \
5336 +                       kmem_cache_free(cachep, branch); \
5337 +                       (map)->tree[i] = NULL; \
5338 +               } else if (256 == count) { \
5339 +                       kmem_cache_free(cachep, branch); \
5340 +                       (map)->tree[i] = full; \
5341 +               } \
5342 +       }
5343 +
5344 +#define LOOP_WALK_BEGIN_COUNT(map, i, branch, inrange, count) \
5345 +       for (i = 0; i < 256; i++) { \
5346 +               if (!(map)->tree[i]) { \
5347 +                       if (inrange) { \
5348 +                               count++; \
5349 +                               inrange = 0; \
5350 +                       } \
5351 +                       continue; \
5352 +               } \
5353 +               branch = (map)->tree[i];
5354 +
5355 +#define LOOP_WALK_END_COUNT() \
5356 +       }
5357 +
5358 +#define MIN(a, b) (a < b ? a : b)
5359 +#define MAX(a, b) (a > b ? a : b)
5360 +
5361 +#define GETVALUE1(a, a1, b1, r) \
5362 +       (a == a1 ? b1 : r)
5363 +
5364 +#define GETVALUE2(a, b, a1, b1, c1, r) \
5365 +       (a == a1 && b == b1 ? c1 : r)
5366 +
5367 +#define GETVALUE3(a, b, c, a1, b1, c1, d1, r) \
5368 +       (a == a1 && b == b1 && c == c1 ? d1 : r)
5369 +
5370 +#define CHECK1(a, a1, a2, b1, b2, c1, c2, d1, d2) \
5371 +       ( \
5372 +               GETVALUE1(a, a1, b1, 0) == 0 \
5373 +               && GETVALUE1(a, a2, b2, 255) == 255 \
5374 +               && c1 == 0 \
5375 +               && c2 == 255 \
5376 +               && d1 == 0 \
5377 +               && d2 == 255 \
5378 +       )
5379 +
5380 +#define CHECK2(a, b, a1, a2, b1, b2, c1, c2, d1, d2) \
5381 +       ( \
5382 +               GETVALUE2(a, b, a1, b1, c1, 0) == 0 \
5383 +               && GETVALUE2(a, b, a2, b2, c2, 255) == 255 \
5384 +               && d1 == 0 \
5385 +               && d2 == 255 \
5386 +       )
5387 +
5388 +#define CHECK3(a, b, c, a1, a2, b1, b2, c1, c2, d1, d2) \
5389 +       ( \
5390 +               GETVALUE3(a, b, c, a1, b1, c1, d1, 0) == 0 \
5391 +               && GETVALUE3(a, b, c, a2, b2, c2, d2, 255) == 255 \
5392 +       )
5393 +
5394 +
5395 +static inline void
5396 +free_d(struct ip_set_iptreemap_d *map)
5397 +{
5398 +       kmem_cache_free(cachep_d, map);
5399 +}
5400 +
5401 +static inline void
5402 +free_c(struct ip_set_iptreemap_c *map)
5403 +{
5404 +       struct ip_set_iptreemap_d *dtree;
5405 +       unsigned int i;
5406 +
5407 +       LOOP_WALK_BEGIN(map, i, dtree) {
5408 +               if (dtree != fullbitmap_d)
5409 +                       free_d(dtree);
5410 +       } LOOP_WALK_END();
5411 +
5412 +       kmem_cache_free(cachep_c, map);
5413 +}
5414 +
5415 +static inline void
5416 +free_b(struct ip_set_iptreemap_b *map)
5417 +{
5418 +       struct ip_set_iptreemap_c *ctree;
5419 +       unsigned int i;
5420 +
5421 +       LOOP_WALK_BEGIN(map, i, ctree) {
5422 +               if (ctree != fullbitmap_c)
5423 +                       free_c(ctree);
5424 +       } LOOP_WALK_END();
5425 +
5426 +       kmem_cache_free(cachep_b, map);
5427 +}
5428 +
5429 +static inline int
5430 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
5431 +{
5432 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5433 +       struct ip_set_iptreemap_b *btree;
5434 +       struct ip_set_iptreemap_c *ctree;
5435 +       struct ip_set_iptreemap_d *dtree;
5436 +       unsigned char a, b, c, d;
5437 +
5438 +       *hash_ip = ip;
5439 +
5440 +       ABCD(a, b, c, d, hash_ip);
5441 +
5442 +       TESTIP_WALK(map, a, btree, fullbitmap_b);
5443 +       TESTIP_WALK(btree, b, ctree, fullbitmap_c);
5444 +       TESTIP_WALK(ctree, c, dtree, fullbitmap_d);
5445 +
5446 +       return !!test_bit(d, (void *) dtree->bitmap);
5447 +}
5448 +
5449 +static int
5450 +testip(struct ip_set *set, const void *data, size_t size, ip_set_ip_t *hash_ip)
5451 +{
5452 +       struct ip_set_req_iptreemap *req = (struct ip_set_req_iptreemap *) data;
5453 +
5454 +       if (size != sizeof(struct ip_set_req_iptreemap)) {
5455 +               ip_set_printk("data length wrong (want %zu, have %zu)", sizeof(struct ip_set_req_iptreemap), size);
5456 +               return -EINVAL;
5457 +       }
5458 +
5459 +       return __testip(set, req->start, hash_ip);
5460 +}
5461 +
5462 +static int
5463 +testip_kernel(struct ip_set *set, const struct sk_buff *skb, ip_set_ip_t *hash_ip, const u_int32_t *flags, unsigned char index)
5464 +{
5465 +       int res;
5466 +
5467 +       res = __testip(set,
5468 +                      ntohl(flags[index] & IPSET_SRC
5469 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
5470 +                               ? ip_hdr(skb)->saddr
5471 +                               : ip_hdr(skb)->daddr),
5472 +#else
5473 +                               ? skb->nh.iph->saddr
5474 +                               : skb->nh.iph->daddr),
5475 +#endif
5476 +                      hash_ip);
5477 +
5478 +       return (res < 0 ? 0 : res);
5479 +}
5480 +
5481 +static inline int
5482 +__addip_single(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
5483 +{
5484 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5485 +       struct ip_set_iptreemap_b *btree;
5486 +       struct ip_set_iptreemap_c *ctree;
5487 +       struct ip_set_iptreemap_d *dtree;
5488 +       unsigned char a, b, c, d;
5489 +
5490 +       *hash_ip = ip;
5491 +
5492 +       ABCD(a, b, c, d, hash_ip);
5493 +
5494 +       ADDIP_WALK(map, a, btree, struct ip_set_iptreemap_b, cachep_b, fullbitmap_b);
5495 +       ADDIP_WALK(btree, b, ctree, struct ip_set_iptreemap_c, cachep_c, fullbitmap_c);
5496 +       ADDIP_WALK(ctree, c, dtree, struct ip_set_iptreemap_d, cachep_d, fullbitmap_d);
5497 +
5498 +       if (test_and_set_bit(d, (void *) dtree->bitmap))
5499 +               return -EEXIST;
5500 +
5501 +       set_bit(b, (void *) btree->dirty);
5502 +
5503 +       return 0;
5504 +}
5505 +
5506 +static inline int
5507 +__addip_range(struct ip_set *set, ip_set_ip_t start, ip_set_ip_t end, ip_set_ip_t *hash_ip)
5508 +{
5509 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5510 +       struct ip_set_iptreemap_b *btree;
5511 +       struct ip_set_iptreemap_c *ctree;
5512 +       struct ip_set_iptreemap_d *dtree;
5513 +       unsigned int a, b, c, d;
5514 +       unsigned char a1, b1, c1, d1;
5515 +       unsigned char a2, b2, c2, d2;
5516 +
5517 +       if (start == end)
5518 +               return __addip_single(set, start, hash_ip);
5519 +
5520 +       *hash_ip = start;
5521 +
5522 +       ABCD(a1, b1, c1, d1, &start);
5523 +       ABCD(a2, b2, c2, d2, &end);
5524 +
5525 +       /* This is sooo ugly... */
5526 +       ADDIP_RANGE_LOOP(map, a, a1, a2, CHECK1(a, a1, a2, b1, b2, c1, c2, d1, d2), btree, fullbitmap_b, cachep_b, free_b) {
5527 +               ADDIP_RANGE_LOOP(btree, b, GETVALUE1(a, a1, b1, 0), GETVALUE1(a, a2, b2, 255), CHECK2(a, b, a1, a2, b1, b2, c1, c2, d1, d2), ctree, fullbitmap_c, cachep_c, free_c) {
5528 +                       ADDIP_RANGE_LOOP(ctree, c, GETVALUE2(a, b, a1, b1, c1, 0), GETVALUE2(a, b, a2, b2, c2, 255), CHECK3(a, b, c, a1, a2, b1, b2, c1, c2, d1, d2), dtree, fullbitmap_d, cachep_d, free_d) {
5529 +                               for (d = GETVALUE3(a, b, c, a1, b1, c1, d1, 0); d <= GETVALUE3(a, b, c, a2, b2, c2, d2, 255); d++)
5530 +                                       set_bit(d, (void *) dtree->bitmap);
5531 +                               set_bit(b, (void *) btree->dirty);
5532 +                       } ADDIP_RANGE_LOOP_END();
5533 +               } ADDIP_RANGE_LOOP_END();
5534 +       } ADDIP_RANGE_LOOP_END();
5535 +
5536 +       return 0;
5537 +}
5538 +
5539 +static int
5540 +addip(struct ip_set *set, const void *data, size_t size, ip_set_ip_t *hash_ip)
5541 +{
5542 +       struct ip_set_req_iptreemap *req = (struct ip_set_req_iptreemap *) data;
5543 +
5544 +       if (size != sizeof(struct ip_set_req_iptreemap)) {
5545 +               ip_set_printk("data length wrong (want %zu, have %zu)", sizeof(struct ip_set_req_iptreemap), size);
5546 +               return -EINVAL;
5547 +       }
5548 +
5549 +       return __addip_range(set, MIN(req->start, req->end), MAX(req->start, req->end), hash_ip);
5550 +}
5551 +
5552 +static int
5553 +addip_kernel(struct ip_set *set, const struct sk_buff *skb, ip_set_ip_t *hash_ip, const u_int32_t *flags, unsigned char index)
5554 +{
5555 +
5556 +       return __addip_single(set,
5557 +                       ntohl(flags[index] & IPSET_SRC
5558 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
5559 +                               ? ip_hdr(skb)->saddr
5560 +                               : ip_hdr(skb)->daddr),
5561 +#else
5562 +                               ? skb->nh.iph->saddr
5563 +                               : skb->nh.iph->daddr),
5564 +#endif
5565 +                       hash_ip);
5566 +}
5567 +
5568 +static inline int
5569 +__delip_single(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip, unsigned int __nocast flags)
5570 +{
5571 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5572 +       struct ip_set_iptreemap_b *btree;
5573 +       struct ip_set_iptreemap_c *ctree;
5574 +       struct ip_set_iptreemap_d *dtree;
5575 +       unsigned char a,b,c,d;
5576 +
5577 +       *hash_ip = ip;
5578 +
5579 +       ABCD(a, b, c, d, hash_ip);
5580 +
5581 +       DELIP_WALK(map, a, btree, cachep_b, fullbitmap_b, flags);
5582 +       DELIP_WALK(btree, b, ctree, cachep_c, fullbitmap_c, flags);
5583 +       DELIP_WALK(ctree, c, dtree, cachep_d, fullbitmap_d, flags);
5584 +
5585 +       if (!test_and_clear_bit(d, (void *) dtree->bitmap))
5586 +               return -EEXIST;
5587 +
5588 +       set_bit(b, (void *) btree->dirty);
5589 +
5590 +       return 0;
5591 +}
5592 +
5593 +static inline int
5594 +__delip_range(struct ip_set *set, ip_set_ip_t start, ip_set_ip_t end, ip_set_ip_t *hash_ip, unsigned int __nocast flags)
5595 +{
5596 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5597 +       struct ip_set_iptreemap_b *btree;
5598 +       struct ip_set_iptreemap_c *ctree;
5599 +       struct ip_set_iptreemap_d *dtree;
5600 +       unsigned int a, b, c, d;
5601 +       unsigned char a1, b1, c1, d1;
5602 +       unsigned char a2, b2, c2, d2;
5603 +
5604 +       if (start == end)
5605 +               return __delip_single(set, start, hash_ip, flags);
5606 +
5607 +       *hash_ip = start;
5608 +
5609 +       ABCD(a1, b1, c1, d1, &start);
5610 +       ABCD(a2, b2, c2, d2, &end);
5611 +
5612 +       /* This is sooo ugly... */
5613 +       DELIP_RANGE_LOOP(map, a, a1, a2, CHECK1(a, a1, a2, b1, b2, c1, c2, d1, d2), btree, fullbitmap_b, cachep_b, free_b, flags) {
5614 +               DELIP_RANGE_LOOP(btree, b, GETVALUE1(a, a1, b1, 0), GETVALUE1(a, a2, b2, 255), CHECK2(a, b, a1, a2, b1, b2, c1, c2, d1, d2), ctree, fullbitmap_c, cachep_c, free_c, flags) {
5615 +                       DELIP_RANGE_LOOP(ctree, c, GETVALUE2(a, b, a1, b1, c1, 0), GETVALUE2(a, b, a2, b2, c2, 255), CHECK3(a, b, c, a1, a2, b1, b2, c1, c2, d1, d2), dtree, fullbitmap_d, cachep_d, free_d, flags) {
5616 +                               for (d = GETVALUE3(a, b, c, a1, b1, c1, d1, 0); d <= GETVALUE3(a, b, c, a2, b2, c2, d2, 255); d++)
5617 +                                       clear_bit(d, (void *) dtree->bitmap);
5618 +                               set_bit(b, (void *) btree->dirty);
5619 +                       } DELIP_RANGE_LOOP_END();
5620 +               } DELIP_RANGE_LOOP_END();
5621 +       } DELIP_RANGE_LOOP_END();
5622 +
5623 +       return 0;
5624 +}
5625 +
5626 +static int
5627 +delip(struct ip_set *set, const void *data, size_t size, ip_set_ip_t *hash_ip)
5628 +{
5629 +       struct ip_set_req_iptreemap *req = (struct ip_set_req_iptreemap *) data;
5630 +
5631 +       if (size != sizeof(struct ip_set_req_iptreemap)) {
5632 +               ip_set_printk("data length wrong (want %zu, have %zu)", sizeof(struct ip_set_req_iptreemap), size);
5633 +               return -EINVAL;
5634 +       }
5635 +
5636 +       return __delip_range(set, MIN(req->start, req->end), MAX(req->start, req->end), hash_ip, GFP_KERNEL);
5637 +}
5638 +
5639 +static int
5640 +delip_kernel(struct ip_set *set, const struct sk_buff *skb, ip_set_ip_t *hash_ip, const u_int32_t *flags, unsigned char index)
5641 +{
5642 +       return __delip_single(set,
5643 +                       ntohl(flags[index] & IPSET_SRC
5644 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
5645 +                               ? ip_hdr(skb)->saddr
5646 +                               : ip_hdr(skb)->daddr),
5647 +#else
5648 +                               ? skb->nh.iph->saddr
5649 +                               : skb->nh.iph->daddr),
5650 +#endif
5651 +                       hash_ip,
5652 +                       GFP_ATOMIC);
5653 +}
5654 +
5655 +/* Check the status of the bitmap
5656 + * -1 == all bits cleared
5657 + *  1 == all bits set
5658 + *  0 == anything else
5659 + */
5660 +static inline int
5661 +bitmap_status(struct ip_set_iptreemap_d *dtree)
5662 +{
5663 +       unsigned char first = dtree->bitmap[0];
5664 +       int a;
5665 +
5666 +       for (a = 1; a < 32; a++)
5667 +               if (dtree->bitmap[a] != first)
5668 +                       return 0;
5669 +
5670 +       return (first == 0 ? -1 : (first == 255 ? 1 : 0));
5671 +}
5672 +
5673 +static void
5674 +gc(unsigned long addr)
5675 +{
5676 +       struct ip_set *set = (struct ip_set *) addr;
5677 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5678 +       struct ip_set_iptreemap_b *btree;
5679 +       struct ip_set_iptreemap_c *ctree;
5680 +       struct ip_set_iptreemap_d *dtree;
5681 +       unsigned int a, b, c;
5682 +       int i, j, k;
5683 +
5684 +       write_lock_bh(&set->lock);
5685 +
5686 +       LOOP_WALK_BEGIN_GC(map, a, btree, fullbitmap_b, cachep_b, i) {
5687 +               LOOP_WALK_BEGIN_GC(btree, b, ctree, fullbitmap_c, cachep_c, j) {
5688 +                       if (!test_and_clear_bit(b, (void *) btree->dirty))
5689 +                               continue;
5690 +                       LOOP_WALK_BEGIN_GC(ctree, c, dtree, fullbitmap_d, cachep_d, k) {
5691 +                               switch (bitmap_status(dtree)) {
5692 +                                       case -1:
5693 +                                               kmem_cache_free(cachep_d, dtree);
5694 +                                               ctree->tree[c] = NULL;
5695 +                                               k--;
5696 +                                       break;
5697 +                                       case 1:
5698 +                                               kmem_cache_free(cachep_d, dtree);
5699 +                                               ctree->tree[c] = fullbitmap_d;
5700 +                                               k++;
5701 +                                       break;
5702 +                               }
5703 +                       } LOOP_WALK_END();
5704 +               } LOOP_WALK_END_GC(btree, b, ctree, fullbitmap_c, cachep_c, k);
5705 +       } LOOP_WALK_END_GC(map, a, btree, fullbitmap_b, cachep_b, j);
5706 +
5707 +       write_unlock_bh(&set->lock);
5708 +
5709 +       map->gc.expires = jiffies + map->gc_interval * HZ;
5710 +       add_timer(&map->gc);
5711 +}
5712 +
5713 +static inline void
5714 +init_gc_timer(struct ip_set *set)
5715 +{
5716 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5717 +
5718 +       init_timer(&map->gc);
5719 +       map->gc.data = (unsigned long) set;
5720 +       map->gc.function = gc;
5721 +       map->gc.expires = jiffies + map->gc_interval * HZ;
5722 +       add_timer(&map->gc);
5723 +}
5724 +
5725 +static int create(struct ip_set *set, const void *data, size_t size)
5726 +{
5727 +       struct ip_set_req_iptreemap_create *req = (struct ip_set_req_iptreemap_create *) data;
5728 +       struct ip_set_iptreemap *map;
5729 +
5730 +       if (size != sizeof(struct ip_set_req_iptreemap_create)) {
5731 +               ip_set_printk("data length wrong (want %zu, have %zu)", sizeof(struct ip_set_req_iptreemap_create), size);
5732 +               return -EINVAL;
5733 +       }
5734 +
5735 +       map = kzalloc(sizeof(*map), GFP_KERNEL);
5736 +       if (!map)
5737 +               return -ENOMEM;
5738 +
5739 +       map->gc_interval = req->gc_interval ? req->gc_interval : IPTREEMAP_DEFAULT_GC_TIME;
5740 +       set->data = map;
5741 +
5742 +       init_gc_timer(set);
5743 +
5744 +       return 0;
5745 +}
5746 +
5747 +static inline void __flush(struct ip_set_iptreemap *map)
5748 +{
5749 +       struct ip_set_iptreemap_b *btree;
5750 +       unsigned int a;
5751 +
5752 +       LOOP_WALK_BEGIN(map, a, btree);
5753 +               if (btree != fullbitmap_b)
5754 +                       free_b(btree);
5755 +       LOOP_WALK_END();
5756 +}
5757 +
5758 +static void destroy(struct ip_set *set)
5759 +{
5760 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5761 +
5762 +       while (!del_timer(&map->gc))
5763 +               msleep(IPTREEMAP_DESTROY_SLEEP);
5764 +
5765 +       __flush(map);
5766 +       kfree(map);
5767 +
5768 +       set->data = NULL;
5769 +}
5770 +
5771 +static void flush(struct ip_set *set)
5772 +{
5773 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5774 +
5775 +       while (!del_timer(&map->gc))
5776 +               msleep(IPTREEMAP_DESTROY_SLEEP);
5777 +
5778 +       __flush(map);
5779 +
5780 +       memset(map, 0, sizeof(*map));
5781 +
5782 +       init_gc_timer(set);
5783 +}
5784 +
5785 +static void list_header(const struct ip_set *set, void *data)
5786 +{
5787 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5788 +       struct ip_set_req_iptreemap_create *header = (struct ip_set_req_iptreemap_create *) data;
5789 +
5790 +       header->gc_interval = map->gc_interval;
5791 +}
5792 +
5793 +static int list_members_size(const struct ip_set *set)
5794 +{
5795 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5796 +       struct ip_set_iptreemap_b *btree;
5797 +       struct ip_set_iptreemap_c *ctree;
5798 +       struct ip_set_iptreemap_d *dtree;
5799 +       unsigned int a, b, c, d, inrange = 0, count = 0;
5800 +
5801 +       LOOP_WALK_BEGIN_COUNT(map, a, btree, inrange, count) {
5802 +               LOOP_WALK_BEGIN_COUNT(btree, b, ctree, inrange, count) {
5803 +                       LOOP_WALK_BEGIN_COUNT(ctree, c, dtree, inrange, count) {
5804 +                               for (d = 0; d < 256; d++) {
5805 +                                       if (test_bit(d, (void *) dtree->bitmap)) {
5806 +                                               inrange = 1;
5807 +                                       } else if (inrange) {
5808 +                                               count++;
5809 +                                               inrange = 0;
5810 +                                       }
5811 +                               }
5812 +                       } LOOP_WALK_END_COUNT();
5813 +               } LOOP_WALK_END_COUNT();
5814 +       } LOOP_WALK_END_COUNT();
5815 +
5816 +       if (inrange)
5817 +               count++;
5818 +
5819 +       return (count * sizeof(struct ip_set_req_iptreemap));
5820 +}
5821 +
5822 +static inline size_t add_member(void *data, size_t offset, ip_set_ip_t start, ip_set_ip_t end)
5823 +{
5824 +       struct ip_set_req_iptreemap *entry = (struct ip_set_req_iptreemap *) (data + offset);
5825 +
5826 +       entry->start = start;
5827 +       entry->end = end;
5828 +
5829 +       return sizeof(*entry);
5830 +}
5831 +
5832 +static void list_members(const struct ip_set *set, void *data)
5833 +{
5834 +       struct ip_set_iptreemap *map = (struct ip_set_iptreemap *) set->data;
5835 +       struct ip_set_iptreemap_b *btree;
5836 +       struct ip_set_iptreemap_c *ctree;
5837 +       struct ip_set_iptreemap_d *dtree;
5838 +       unsigned int a, b, c, d, inrange = 0;
5839 +       size_t offset = 0;
5840 +       ip_set_ip_t start = 0, end = 0, ip;
5841 +
5842 +       LOOP_WALK_BEGIN(map, a, btree) {
5843 +               LOOP_WALK_BEGIN(btree, b, ctree) {
5844 +                       LOOP_WALK_BEGIN(ctree, c, dtree) {
5845 +                               for (d = 0; d < 256; d++) {
5846 +                                       if (test_bit(d, (void *) dtree->bitmap)) {
5847 +                                               ip = ((a << 24) | (b << 16) | (c << 8) | d);
5848 +                                               if (!inrange) {
5849 +                                                       inrange = 1;
5850 +                                                       start = ip;
5851 +                                               } else if (end < ip - 1) {
5852 +                                                       offset += add_member(data, offset, start, end);
5853 +                                                       start = ip;
5854 +                                               }
5855 +                                               end = ip;
5856 +                                       } else if (inrange) {
5857 +                                               offset += add_member(data, offset, start, end);
5858 +                                               inrange = 0;
5859 +                                       }
5860 +                               }
5861 +                       } LOOP_WALK_END();
5862 +               } LOOP_WALK_END();
5863 +       } LOOP_WALK_END();
5864 +
5865 +       if (inrange)
5866 +               add_member(data, offset, start, end);
5867 +}
5868 +
5869 +static struct ip_set_type ip_set_iptreemap = {
5870 +       .typename               = SETTYPE_NAME,
5871 +       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
5872 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
5873 +       .create                 = create,
5874 +       .destroy                = destroy,
5875 +       .flush                  = flush,
5876 +       .reqsize                = sizeof(struct ip_set_req_iptreemap),
5877 +       .addip                  = addip,
5878 +       .addip_kernel           = addip_kernel,
5879 +       .delip                  = delip,
5880 +       .delip_kernel           = delip_kernel,
5881 +       .testip                 = testip,
5882 +       .testip_kernel          = testip_kernel,
5883 +       .header_size            = sizeof(struct ip_set_req_iptreemap_create),
5884 +       .list_header            = list_header,
5885 +       .list_members_size      = list_members_size,
5886 +       .list_members           = list_members,
5887 +       .me                     = THIS_MODULE,
5888 +};
5889 +
5890 +MODULE_LICENSE("GPL");
5891 +MODULE_AUTHOR("Sven Wegener <sven.wegener@stealer.net>");
5892 +MODULE_DESCRIPTION("iptreemap type of IP sets");
5893 +
5894 +static int __init ip_set_iptreemap_init(void)
5895 +{
5896 +       int ret = -ENOMEM;
5897 +       int a;
5898 +
5899 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
5900 +       cachep_b = kmem_cache_create("ip_set_iptreemap_b",
5901 +                                    sizeof(struct ip_set_iptreemap_b),
5902 +                                    0, 0, NULL);
5903 +#else
5904 +       cachep_b = kmem_cache_create("ip_set_iptreemap_b",
5905 +                                    sizeof(struct ip_set_iptreemap_b),
5906 +                                    0, 0, NULL, NULL);
5907 +#endif
5908 +       if (!cachep_b) {
5909 +               ip_set_printk("Unable to create ip_set_iptreemap_b slab cache");
5910 +               goto out;
5911 +       }
5912 +
5913 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
5914 +       cachep_c = kmem_cache_create("ip_set_iptreemap_c",
5915 +                                    sizeof(struct ip_set_iptreemap_c),
5916 +                                    0, 0, NULL);
5917 +#else
5918 +       cachep_c = kmem_cache_create("ip_set_iptreemap_c",
5919 +                                    sizeof(struct ip_set_iptreemap_c),
5920 +                                    0, 0, NULL, NULL);
5921 +#endif
5922 +       if (!cachep_c) {
5923 +               ip_set_printk("Unable to create ip_set_iptreemap_c slab cache");
5924 +               goto outb;
5925 +       }
5926 +
5927 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
5928 +       cachep_d = kmem_cache_create("ip_set_iptreemap_d",
5929 +                                    sizeof(struct ip_set_iptreemap_d),
5930 +                                    0, 0, NULL);
5931 +#else
5932 +       cachep_d = kmem_cache_create("ip_set_iptreemap_d",
5933 +                                    sizeof(struct ip_set_iptreemap_d),
5934 +                                    0, 0, NULL, NULL);
5935 +#endif
5936 +       if (!cachep_d) {
5937 +               ip_set_printk("Unable to create ip_set_iptreemap_d slab cache");
5938 +               goto outc;
5939 +       }
5940 +
5941 +       fullbitmap_d = kmem_cache_alloc(cachep_d, GFP_KERNEL);
5942 +       if (!fullbitmap_d)
5943 +               goto outd;
5944 +
5945 +       fullbitmap_c = kmem_cache_alloc(cachep_c, GFP_KERNEL);
5946 +       if (!fullbitmap_c)
5947 +               goto outbitmapd;
5948 +
5949 +       fullbitmap_b = kmem_cache_alloc(cachep_b, GFP_KERNEL);
5950 +       if (!fullbitmap_b)
5951 +               goto outbitmapc;
5952 +
5953 +       ret = ip_set_register_set_type(&ip_set_iptreemap);
5954 +       if (0 > ret)
5955 +               goto outbitmapb;
5956 +
5957 +       /* Now init our global bitmaps */
5958 +       memset(fullbitmap_d->bitmap, 0xff, sizeof(fullbitmap_d->bitmap));
5959 +
5960 +       for (a = 0; a < 256; a++)
5961 +               fullbitmap_c->tree[a] = fullbitmap_d;
5962 +
5963 +       for (a = 0; a < 256; a++)
5964 +               fullbitmap_b->tree[a] = fullbitmap_c;
5965 +       memset(fullbitmap_b->dirty, 0, sizeof(fullbitmap_b->dirty));
5966 +
5967 +       return 0;
5968 +
5969 +outbitmapb:
5970 +       kmem_cache_free(cachep_b, fullbitmap_b);
5971 +outbitmapc:
5972 +       kmem_cache_free(cachep_c, fullbitmap_c);
5973 +outbitmapd:
5974 +       kmem_cache_free(cachep_d, fullbitmap_d);
5975 +outd:
5976 +       kmem_cache_destroy(cachep_d);
5977 +outc:
5978 +       kmem_cache_destroy(cachep_c);
5979 +outb:
5980 +       kmem_cache_destroy(cachep_b);
5981 +out:
5982 +
5983 +       return ret;
5984 +}
5985 +
5986 +static void __exit ip_set_iptreemap_fini(void)
5987 +{
5988 +       ip_set_unregister_set_type(&ip_set_iptreemap);
5989 +       kmem_cache_free(cachep_d, fullbitmap_d);
5990 +       kmem_cache_free(cachep_c, fullbitmap_c);
5991 +       kmem_cache_free(cachep_b, fullbitmap_b);
5992 +       kmem_cache_destroy(cachep_d);
5993 +       kmem_cache_destroy(cachep_c);
5994 +       kmem_cache_destroy(cachep_b);
5995 +}
5996 +
5997 +module_init(ip_set_iptreemap_init);
5998 +module_exit(ip_set_iptreemap_fini);
5999 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set_macipmap.c
6000 ===================================================================
6001 --- /dev/null
6002 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set_macipmap.c
6003 @@ -0,0 +1,375 @@
6004 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
6005 + *                         Patrick Schaaf <bof@bof.de>
6006 + *                         Martin Josefsson <gandalf@wlug.westbo.se>
6007 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
6008 + *
6009 + * This program is free software; you can redistribute it and/or modify
6010 + * it under the terms of the GNU General Public License version 2 as
6011 + * published by the Free Software Foundation.
6012 + */
6013 +
6014 +/* Kernel module implementing an IP set type: the macipmap type */
6015 +
6016 +#include <linux/module.h>
6017 +#include <linux/ip.h>
6018 +#include <linux/skbuff.h>
6019 +#include <linux/version.h>
6020 +#include <linux/netfilter_ipv4/ip_tables.h>
6021 +#include <linux/netfilter_ipv4/ip_set.h>
6022 +#include <linux/errno.h>
6023 +#include <asm/uaccess.h>
6024 +#include <asm/bitops.h>
6025 +#include <linux/spinlock.h>
6026 +#include <linux/if_ether.h>
6027 +#include <linux/vmalloc.h>
6028 +
6029 +#include <linux/netfilter_ipv4/ip_set_malloc.h>
6030 +#include <linux/netfilter_ipv4/ip_set_macipmap.h>
6031 +
6032 +static int
6033 +testip(struct ip_set *set, const void *data, size_t size, ip_set_ip_t *hash_ip)
6034 +{
6035 +       struct ip_set_macipmap *map = (struct ip_set_macipmap *) set->data;
6036 +       struct ip_set_macip *table = (struct ip_set_macip *) map->members;
6037 +       struct ip_set_req_macipmap *req = (struct ip_set_req_macipmap *) data;
6038 +
6039 +       if (size != sizeof(struct ip_set_req_macipmap)) {
6040 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6041 +                             sizeof(struct ip_set_req_macipmap),
6042 +                             size);
6043 +               return -EINVAL;
6044 +       }
6045 +
6046 +       if (req->ip < map->first_ip || req->ip > map->last_ip)
6047 +               return -ERANGE;
6048 +
6049 +       *hash_ip = req->ip;
6050 +       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
6051 +          set->name, HIPQUAD(req->ip), HIPQUAD(*hash_ip));
6052 +       if (test_bit(IPSET_MACIP_ISSET,
6053 +                    (void *) &table[req->ip - map->first_ip].flags)) {
6054 +               return (memcmp(req->ethernet,
6055 +                              &table[req->ip - map->first_ip].ethernet,
6056 +                              ETH_ALEN) == 0);
6057 +       } else {
6058 +               return (map->flags & IPSET_MACIP_MATCHUNSET ? 1 : 0);
6059 +       }
6060 +}
6061 +
6062 +static int
6063 +testip_kernel(struct ip_set *set,
6064 +             const struct sk_buff *skb,
6065 +             ip_set_ip_t *hash_ip,
6066 +             const u_int32_t *flags,
6067 +             unsigned char index)
6068 +{
6069 +       struct ip_set_macipmap *map =
6070 +           (struct ip_set_macipmap *) set->data;
6071 +       struct ip_set_macip *table =
6072 +           (struct ip_set_macip *) map->members;
6073 +       ip_set_ip_t ip;
6074 +
6075 +       ip = ntohl(flags[index] & IPSET_SRC
6076 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6077 +                       ? ip_hdr(skb)->saddr
6078 +                       : ip_hdr(skb)->daddr);
6079 +#else
6080 +                       ? skb->nh.iph->saddr
6081 +                       : skb->nh.iph->daddr);
6082 +#endif
6083 +
6084 +       if (ip < map->first_ip || ip > map->last_ip)
6085 +               return 0;
6086 +
6087 +       *hash_ip = ip;
6088 +       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
6089 +          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip));
6090 +       if (test_bit(IPSET_MACIP_ISSET,
6091 +           (void *) &table[ip - map->first_ip].flags)) {
6092 +               /* Is mac pointer valid?
6093 +                * If so, compare... */
6094 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6095 +               return (skb_mac_header(skb) >= skb->head
6096 +                       && (skb_mac_header(skb) + ETH_HLEN) <= skb->data
6097 +#else
6098 +               return (skb->mac.raw >= skb->head
6099 +                       && (skb->mac.raw + ETH_HLEN) <= skb->data
6100 +#endif
6101 +                       && (memcmp(eth_hdr(skb)->h_source,
6102 +                                  &table[ip - map->first_ip].ethernet,
6103 +                                  ETH_ALEN) == 0));
6104 +       } else {
6105 +               return (map->flags & IPSET_MACIP_MATCHUNSET ? 1 : 0);
6106 +       }
6107 +}
6108 +
6109 +/* returns 0 on success */
6110 +static inline int
6111 +__addip(struct ip_set *set,
6112 +       ip_set_ip_t ip, unsigned char *ethernet, ip_set_ip_t *hash_ip)
6113 +{
6114 +       struct ip_set_macipmap *map =
6115 +           (struct ip_set_macipmap *) set->data;
6116 +       struct ip_set_macip *table =
6117 +           (struct ip_set_macip *) map->members;
6118 +
6119 +       if (ip < map->first_ip || ip > map->last_ip)
6120 +               return -ERANGE;
6121 +       if (test_and_set_bit(IPSET_MACIP_ISSET,
6122 +                            (void *) &table[ip - map->first_ip].flags))
6123 +               return -EEXIST;
6124 +
6125 +       *hash_ip = ip;
6126 +       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
6127 +       memcpy(&table[ip - map->first_ip].ethernet, ethernet, ETH_ALEN);
6128 +       return 0;
6129 +}
6130 +
6131 +static int
6132 +addip(struct ip_set *set, const void *data, size_t size,
6133 +      ip_set_ip_t *hash_ip)
6134 +{
6135 +       struct ip_set_req_macipmap *req =
6136 +           (struct ip_set_req_macipmap *) data;
6137 +
6138 +       if (size != sizeof(struct ip_set_req_macipmap)) {
6139 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6140 +                             sizeof(struct ip_set_req_macipmap),
6141 +                             size);
6142 +               return -EINVAL;
6143 +       }
6144 +       return __addip(set, req->ip, req->ethernet, hash_ip);
6145 +}
6146 +
6147 +static int
6148 +addip_kernel(struct ip_set *set,
6149 +            const struct sk_buff *skb,
6150 +            ip_set_ip_t *hash_ip,
6151 +            const u_int32_t *flags,
6152 +            unsigned char index)
6153 +{
6154 +       ip_set_ip_t ip;
6155 +
6156 +       ip = ntohl(flags[index] & IPSET_SRC
6157 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6158 +                       ? ip_hdr(skb)->saddr
6159 +                       : ip_hdr(skb)->daddr);
6160 +#else
6161 +                       ? skb->nh.iph->saddr
6162 +                       : skb->nh.iph->daddr);
6163 +#endif
6164 +
6165 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6166 +       if (!(skb_mac_header(skb) >= skb->head
6167 +             && (skb_mac_header(skb) + ETH_HLEN) <= skb->data))
6168 +#else
6169 +       if (!(skb->mac.raw >= skb->head
6170 +             && (skb->mac.raw + ETH_HLEN) <= skb->data))
6171 +#endif
6172 +               return -EINVAL;
6173 +
6174 +       return __addip(set, ip, eth_hdr(skb)->h_source, hash_ip);
6175 +}
6176 +
6177 +static inline int
6178 +__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
6179 +{
6180 +       struct ip_set_macipmap *map =
6181 +           (struct ip_set_macipmap *) set->data;
6182 +       struct ip_set_macip *table =
6183 +           (struct ip_set_macip *) map->members;
6184 +
6185 +       if (ip < map->first_ip || ip > map->last_ip)
6186 +               return -ERANGE;
6187 +       if (!test_and_clear_bit(IPSET_MACIP_ISSET,
6188 +                               (void *)&table[ip - map->first_ip].flags))
6189 +               return -EEXIST;
6190 +
6191 +       *hash_ip = ip;
6192 +       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
6193 +       return 0;
6194 +}
6195 +
6196 +static int
6197 +delip(struct ip_set *set, const void *data, size_t size,
6198 +     ip_set_ip_t *hash_ip)
6199 +{
6200 +       struct ip_set_req_macipmap *req =
6201 +           (struct ip_set_req_macipmap *) data;
6202 +
6203 +       if (size != sizeof(struct ip_set_req_macipmap)) {
6204 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6205 +                             sizeof(struct ip_set_req_macipmap),
6206 +                             size);
6207 +               return -EINVAL;
6208 +       }
6209 +       return __delip(set, req->ip, hash_ip);
6210 +}
6211 +
6212 +static int
6213 +delip_kernel(struct ip_set *set,
6214 +            const struct sk_buff *skb,
6215 +            ip_set_ip_t *hash_ip,
6216 +            const u_int32_t *flags,
6217 +            unsigned char index)
6218 +{
6219 +       return __delip(set,
6220 +                      ntohl(flags[index] & IPSET_SRC
6221 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6222 +                               ? ip_hdr(skb)->saddr
6223 +                               : ip_hdr(skb)->daddr),
6224 +#else
6225 +                               ? skb->nh.iph->saddr
6226 +                               : skb->nh.iph->daddr),
6227 +#endif
6228 +                      hash_ip);
6229 +}
6230 +
6231 +static inline size_t members_size(ip_set_id_t from, ip_set_id_t to)
6232 +{
6233 +       return (size_t)((to - from + 1) * sizeof(struct ip_set_macip));
6234 +}
6235 +
6236 +static int create(struct ip_set *set, const void *data, size_t size)
6237 +{
6238 +       int newbytes;
6239 +       struct ip_set_req_macipmap_create *req =
6240 +           (struct ip_set_req_macipmap_create *) data;
6241 +       struct ip_set_macipmap *map;
6242 +
6243 +       if (size != sizeof(struct ip_set_req_macipmap_create)) {
6244 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6245 +                             sizeof(struct ip_set_req_macipmap_create),
6246 +                             size);
6247 +               return -EINVAL;
6248 +       }
6249 +
6250 +       DP("from %u.%u.%u.%u to %u.%u.%u.%u",
6251 +          HIPQUAD(req->from), HIPQUAD(req->to));
6252 +
6253 +       if (req->from > req->to) {
6254 +               DP("bad ip range");
6255 +               return -ENOEXEC;
6256 +       }
6257 +
6258 +       if (req->to - req->from > MAX_RANGE) {
6259 +               ip_set_printk("range too big (max %d addresses)",
6260 +                              MAX_RANGE+1);
6261 +               return -ENOEXEC;
6262 +       }
6263 +
6264 +       map = kmalloc(sizeof(struct ip_set_macipmap), GFP_KERNEL);
6265 +       if (!map) {
6266 +               DP("out of memory for %d bytes",
6267 +                  sizeof(struct ip_set_macipmap));
6268 +               return -ENOMEM;
6269 +       }
6270 +       map->flags = req->flags;
6271 +       map->first_ip = req->from;
6272 +       map->last_ip = req->to;
6273 +       newbytes = members_size(map->first_ip, map->last_ip);
6274 +       map->members = ip_set_malloc(newbytes);
6275 +       DP("members: %u %p", newbytes, map->members);
6276 +       if (!map->members) {
6277 +               DP("out of memory for %d bytes", newbytes);
6278 +               kfree(map);
6279 +               return -ENOMEM;
6280 +       }
6281 +       memset(map->members, 0, newbytes);
6282 +
6283 +       set->data = map;
6284 +       return 0;
6285 +}
6286 +
6287 +static void destroy(struct ip_set *set)
6288 +{
6289 +       struct ip_set_macipmap *map =
6290 +           (struct ip_set_macipmap *) set->data;
6291 +
6292 +       ip_set_free(map->members, members_size(map->first_ip, map->last_ip));
6293 +       kfree(map);
6294 +
6295 +       set->data = NULL;
6296 +}
6297 +
6298 +static void flush(struct ip_set *set)
6299 +{
6300 +       struct ip_set_macipmap *map =
6301 +           (struct ip_set_macipmap *) set->data;
6302 +       memset(map->members, 0, members_size(map->first_ip, map->last_ip));
6303 +}
6304 +
6305 +static void list_header(const struct ip_set *set, void *data)
6306 +{
6307 +       struct ip_set_macipmap *map =
6308 +           (struct ip_set_macipmap *) set->data;
6309 +       struct ip_set_req_macipmap_create *header =
6310 +           (struct ip_set_req_macipmap_create *) data;
6311 +
6312 +       DP("list_header %x %x %u", map->first_ip, map->last_ip,
6313 +          map->flags);
6314 +
6315 +       header->from = map->first_ip;
6316 +       header->to = map->last_ip;
6317 +       header->flags = map->flags;
6318 +}
6319 +
6320 +static int list_members_size(const struct ip_set *set)
6321 +{
6322 +       struct ip_set_macipmap *map =
6323 +           (struct ip_set_macipmap *) set->data;
6324 +
6325 +       DP("%u", members_size(map->first_ip, map->last_ip));
6326 +       return members_size(map->first_ip, map->last_ip);
6327 +}
6328 +
6329 +static void list_members(const struct ip_set *set, void *data)
6330 +{
6331 +       struct ip_set_macipmap *map =
6332 +           (struct ip_set_macipmap *) set->data;
6333 +
6334 +       int bytes = members_size(map->first_ip, map->last_ip);
6335 +
6336 +       DP("members: %u %p", bytes, map->members);
6337 +       memcpy(data, map->members, bytes);
6338 +}
6339 +
6340 +static struct ip_set_type ip_set_macipmap = {
6341 +       .typename               = SETTYPE_NAME,
6342 +       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
6343 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
6344 +       .create                 = &create,
6345 +       .destroy                = &destroy,
6346 +       .flush                  = &flush,
6347 +       .reqsize                = sizeof(struct ip_set_req_macipmap),
6348 +       .addip                  = &addip,
6349 +       .addip_kernel           = &addip_kernel,
6350 +       .delip                  = &delip,
6351 +       .delip_kernel           = &delip_kernel,
6352 +       .testip                 = &testip,
6353 +       .testip_kernel          = &testip_kernel,
6354 +       .header_size            = sizeof(struct ip_set_req_macipmap_create),
6355 +       .list_header            = &list_header,
6356 +       .list_members_size      = &list_members_size,
6357 +       .list_members           = &list_members,
6358 +       .me                     = THIS_MODULE,
6359 +};
6360 +
6361 +MODULE_LICENSE("GPL");
6362 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
6363 +MODULE_DESCRIPTION("macipmap type of IP sets");
6364 +
6365 +static int __init ip_set_macipmap_init(void)
6366 +{
6367 +       init_max_malloc_size();
6368 +       return ip_set_register_set_type(&ip_set_macipmap);
6369 +}
6370 +
6371 +static void __exit ip_set_macipmap_fini(void)
6372 +{
6373 +       /* FIXME: possible race with ip_set_create() */
6374 +       ip_set_unregister_set_type(&ip_set_macipmap);
6375 +}
6376 +
6377 +module_init(ip_set_macipmap_init);
6378 +module_exit(ip_set_macipmap_fini);
6379 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set_nethash.c
6380 ===================================================================
6381 --- /dev/null
6382 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set_nethash.c
6383 @@ -0,0 +1,497 @@
6384 +/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
6385 + *
6386 + * This program is free software; you can redistribute it and/or modify
6387 + * it under the terms of the GNU General Public License version 2 as
6388 + * published by the Free Software Foundation.
6389 + */
6390 +
6391 +/* Kernel module implementing a cidr nethash set */
6392 +
6393 +#include <linux/module.h>
6394 +#include <linux/ip.h>
6395 +#include <linux/skbuff.h>
6396 +#include <linux/version.h>
6397 +#include <linux/jhash.h>
6398 +#include <linux/netfilter_ipv4/ip_tables.h>
6399 +#include <linux/netfilter_ipv4/ip_set.h>
6400 +#include <linux/errno.h>
6401 +#include <asm/uaccess.h>
6402 +#include <asm/bitops.h>
6403 +#include <linux/spinlock.h>
6404 +#include <linux/vmalloc.h>
6405 +#include <linux/random.h>
6406 +
6407 +#include <net/ip.h>
6408 +
6409 +#include <linux/netfilter_ipv4/ip_set_malloc.h>
6410 +#include <linux/netfilter_ipv4/ip_set_nethash.h>
6411 +
6412 +static int limit = MAX_RANGE;
6413 +
6414 +static inline __u32
6415 +jhash_ip(const struct ip_set_nethash *map, uint16_t i, ip_set_ip_t ip)
6416 +{
6417 +       return jhash_1word(ip, *(((uint32_t *) map->initval) + i));
6418 +}
6419 +
6420 +static inline __u32
6421 +hash_id_cidr(struct ip_set_nethash *map,
6422 +            ip_set_ip_t ip,
6423 +            unsigned char cidr,
6424 +            ip_set_ip_t *hash_ip)
6425 +{
6426 +       __u32 id;
6427 +       u_int16_t i;
6428 +       ip_set_ip_t *elem;
6429 +
6430 +       *hash_ip = pack(ip, cidr);
6431 +
6432 +       for (i = 0; i < map->probes; i++) {
6433 +               id = jhash_ip(map, i, *hash_ip) % map->hashsize;
6434 +               DP("hash key: %u", id);
6435 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
6436 +               if (*elem == *hash_ip)
6437 +                       return id;
6438 +       }
6439 +       return UINT_MAX;
6440 +}
6441 +
6442 +static inline __u32
6443 +hash_id(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
6444 +{
6445 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6446 +       __u32 id = UINT_MAX;
6447 +       int i;
6448 +
6449 +       for (i = 0; i < 30 && map->cidr[i]; i++) {
6450 +               id = hash_id_cidr(map, ip, map->cidr[i], hash_ip);
6451 +               if (id != UINT_MAX)
6452 +                       break;
6453 +       }
6454 +       return id;
6455 +}
6456 +
6457 +static inline int
6458 +__testip_cidr(struct ip_set *set, ip_set_ip_t ip, unsigned char cidr,
6459 +             ip_set_ip_t *hash_ip)
6460 +{
6461 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6462 +
6463 +       return (ip && hash_id_cidr(map, ip, cidr, hash_ip) != UINT_MAX);
6464 +}
6465 +
6466 +static inline int
6467 +__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
6468 +{
6469 +       return (ip && hash_id(set, ip, hash_ip) != UINT_MAX);
6470 +}
6471 +
6472 +static int
6473 +testip(struct ip_set *set, const void *data, size_t size,
6474 +       ip_set_ip_t *hash_ip)
6475 +{
6476 +       struct ip_set_req_nethash *req =
6477 +           (struct ip_set_req_nethash *) data;
6478 +
6479 +       if (size != sizeof(struct ip_set_req_nethash)) {
6480 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6481 +                             sizeof(struct ip_set_req_nethash),
6482 +                             size);
6483 +               return -EINVAL;
6484 +       }
6485 +       return (req->cidr == 32 ? __testip(set, req->ip, hash_ip)
6486 +               : __testip_cidr(set, req->ip, req->cidr, hash_ip));
6487 +}
6488 +
6489 +static int
6490 +testip_kernel(struct ip_set *set,
6491 +             const struct sk_buff *skb,
6492 +             ip_set_ip_t *hash_ip,
6493 +             const u_int32_t *flags,
6494 +             unsigned char index)
6495 +{
6496 +       return __testip(set,
6497 +                       ntohl(flags[index] & IPSET_SRC
6498 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6499 +                               ? ip_hdr(skb)->saddr
6500 +                               : ip_hdr(skb)->daddr),
6501 +#else
6502 +                               ? skb->nh.iph->saddr
6503 +                               : skb->nh.iph->daddr),
6504 +#endif
6505 +                       hash_ip);
6506 +}
6507 +
6508 +static inline int
6509 +__addip_base(struct ip_set_nethash *map, ip_set_ip_t ip)
6510 +{
6511 +       __u32 probe;
6512 +       u_int16_t i;
6513 +       ip_set_ip_t *elem;
6514 +
6515 +       for (i = 0; i < map->probes; i++) {
6516 +               probe = jhash_ip(map, i, ip) % map->hashsize;
6517 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, probe);
6518 +               if (*elem == ip)
6519 +                       return -EEXIST;
6520 +               if (!*elem) {
6521 +                       *elem = ip;
6522 +                       map->elements++;
6523 +                       return 0;
6524 +               }
6525 +       }
6526 +       /* Trigger rehashing */
6527 +       return -EAGAIN;
6528 +}
6529 +
6530 +static inline int
6531 +__addip(struct ip_set_nethash *map, ip_set_ip_t ip, unsigned char cidr,
6532 +       ip_set_ip_t *hash_ip)
6533 +{
6534 +       if (!ip || map->elements >= limit)
6535 +               return -ERANGE;
6536 +
6537 +       *hash_ip = pack(ip, cidr);
6538 +       DP("%u.%u.%u.%u/%u, %u.%u.%u.%u", HIPQUAD(ip), cidr, HIPQUAD(*hash_ip));
6539 +
6540 +       return __addip_base(map, *hash_ip);
6541 +}
6542 +
6543 +static void
6544 +update_cidr_sizes(struct ip_set_nethash *map, unsigned char cidr)
6545 +{
6546 +       unsigned char next;
6547 +       int i;
6548 +
6549 +       for (i = 0; i < 30 && map->cidr[i]; i++) {
6550 +               if (map->cidr[i] == cidr) {
6551 +                       return;
6552 +               } else if (map->cidr[i] < cidr) {
6553 +                       next = map->cidr[i];
6554 +                       map->cidr[i] = cidr;
6555 +                       cidr = next;
6556 +               }
6557 +       }
6558 +       if (i < 30)
6559 +               map->cidr[i] = cidr;
6560 +}
6561 +
6562 +static int
6563 +addip(struct ip_set *set, const void *data, size_t size,
6564 +        ip_set_ip_t *hash_ip)
6565 +{
6566 +       struct ip_set_req_nethash *req =
6567 +           (struct ip_set_req_nethash *) data;
6568 +       int ret;
6569 +
6570 +       if (size != sizeof(struct ip_set_req_nethash)) {
6571 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6572 +                             sizeof(struct ip_set_req_nethash),
6573 +                             size);
6574 +               return -EINVAL;
6575 +       }
6576 +       ret = __addip((struct ip_set_nethash *) set->data,
6577 +                     req->ip, req->cidr, hash_ip);
6578 +
6579 +       if (ret == 0)
6580 +               update_cidr_sizes((struct ip_set_nethash *) set->data,
6581 +                                 req->cidr);
6582 +
6583 +       return ret;
6584 +}
6585 +
6586 +static int
6587 +addip_kernel(struct ip_set *set,
6588 +            const struct sk_buff *skb,
6589 +            ip_set_ip_t *hash_ip,
6590 +            const u_int32_t *flags,
6591 +            unsigned char index)
6592 +{
6593 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6594 +       int ret = -ERANGE;
6595 +       ip_set_ip_t ip = ntohl(flags[index] & IPSET_SRC
6596 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6597 +                                       ? ip_hdr(skb)->saddr
6598 +                                       : ip_hdr(skb)->daddr);
6599 +#else
6600 +                                       ? skb->nh.iph->saddr
6601 +                                       : skb->nh.iph->daddr);
6602 +#endif
6603 +
6604 +       if (map->cidr[0])
6605 +               ret = __addip(map, ip, map->cidr[0], hash_ip);
6606 +
6607 +       return ret;
6608 +}
6609 +
6610 +static int retry(struct ip_set *set)
6611 +{
6612 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6613 +       ip_set_ip_t *elem;
6614 +       void *members;
6615 +       u_int32_t i, hashsize = map->hashsize;
6616 +       int res;
6617 +       struct ip_set_nethash *tmp;
6618 +
6619 +       if (map->resize == 0)
6620 +               return -ERANGE;
6621 +
6622 +    again:
6623 +       res = 0;
6624 +
6625 +       /* Calculate new parameters */
6626 +       hashsize += (hashsize * map->resize)/100;
6627 +       if (hashsize == map->hashsize)
6628 +               hashsize++;
6629 +
6630 +       ip_set_printk("rehashing of set %s triggered: "
6631 +                     "hashsize grows from %u to %u",
6632 +                     set->name, map->hashsize, hashsize);
6633 +
6634 +       tmp = kmalloc(sizeof(struct ip_set_nethash)
6635 +                     + map->probes * sizeof(uint32_t), GFP_ATOMIC);
6636 +       if (!tmp) {
6637 +               DP("out of memory for %d bytes",
6638 +                  sizeof(struct ip_set_nethash)
6639 +                  + map->probes * sizeof(uint32_t));
6640 +               return -ENOMEM;
6641 +       }
6642 +       tmp->members = harray_malloc(hashsize, sizeof(ip_set_ip_t), GFP_ATOMIC);
6643 +       if (!tmp->members) {
6644 +               DP("out of memory for %d bytes", hashsize * sizeof(ip_set_ip_t));
6645 +               kfree(tmp);
6646 +               return -ENOMEM;
6647 +       }
6648 +       tmp->hashsize = hashsize;
6649 +       tmp->elements = 0;
6650 +       tmp->probes = map->probes;
6651 +       tmp->resize = map->resize;
6652 +       memcpy(tmp->initval, map->initval, map->probes * sizeof(uint32_t));
6653 +       memcpy(tmp->cidr, map->cidr, 30 * sizeof(unsigned char));
6654 +
6655 +       write_lock_bh(&set->lock);
6656 +       map = (struct ip_set_nethash *) set->data; /* Play safe */
6657 +       for (i = 0; i < map->hashsize && res == 0; i++) {
6658 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);
6659 +               if (*elem)
6660 +                       res = __addip_base(tmp, *elem);
6661 +       }
6662 +       if (res) {
6663 +               /* Failure, try again */
6664 +               write_unlock_bh(&set->lock);
6665 +               harray_free(tmp->members);
6666 +               kfree(tmp);
6667 +               goto again;
6668 +       }
6669 +
6670 +       /* Success at resizing! */
6671 +       members = map->members;
6672 +
6673 +       map->hashsize = tmp->hashsize;
6674 +       map->members = tmp->members;
6675 +       write_unlock_bh(&set->lock);
6676 +
6677 +       harray_free(members);
6678 +       kfree(tmp);
6679 +
6680 +       return 0;
6681 +}
6682 +
6683 +static inline int
6684 +__delip(struct ip_set_nethash *map, ip_set_ip_t ip, unsigned char cidr,
6685 +       ip_set_ip_t *hash_ip)
6686 +{
6687 +       ip_set_ip_t id, *elem;
6688 +
6689 +       if (!ip)
6690 +               return -ERANGE;
6691 +
6692 +       id = hash_id_cidr(map, ip, cidr, hash_ip);
6693 +       if (id == UINT_MAX)
6694 +               return -EEXIST;
6695 +
6696 +       elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
6697 +       *elem = 0;
6698 +       map->elements--;
6699 +       return 0;
6700 +}
6701 +
6702 +static int
6703 +delip(struct ip_set *set, const void *data, size_t size,
6704 +        ip_set_ip_t *hash_ip)
6705 +{
6706 +       struct ip_set_req_nethash *req =
6707 +           (struct ip_set_req_nethash *) data;
6708 +
6709 +       if (size != sizeof(struct ip_set_req_nethash)) {
6710 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6711 +                             sizeof(struct ip_set_req_nethash),
6712 +                             size);
6713 +               return -EINVAL;
6714 +       }
6715 +       /* TODO: no garbage collection in map->cidr */
6716 +       return __delip((struct ip_set_nethash *) set->data,
6717 +                      req->ip, req->cidr, hash_ip);
6718 +}
6719 +
6720 +static int
6721 +delip_kernel(struct ip_set *set,
6722 +            const struct sk_buff *skb,
6723 +            ip_set_ip_t *hash_ip,
6724 +            const u_int32_t *flags,
6725 +            unsigned char index)
6726 +{
6727 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6728 +       int ret = -ERANGE;
6729 +       ip_set_ip_t ip = ntohl(flags[index] & IPSET_SRC
6730 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6731 +                                       ? ip_hdr(skb)->saddr
6732 +                                       : ip_hdr(skb)->daddr);
6733 +#else
6734 +                                       ? skb->nh.iph->saddr
6735 +                                       : skb->nh.iph->daddr);
6736 +#endif
6737 +
6738 +       if (map->cidr[0])
6739 +               ret = __delip(map, ip, map->cidr[0], hash_ip);
6740 +
6741 +       return ret;
6742 +}
6743 +
6744 +static int create(struct ip_set *set, const void *data, size_t size)
6745 +{
6746 +       struct ip_set_req_nethash_create *req =
6747 +           (struct ip_set_req_nethash_create *) data;
6748 +       struct ip_set_nethash *map;
6749 +       uint16_t i;
6750 +
6751 +       if (size != sizeof(struct ip_set_req_nethash_create)) {
6752 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6753 +                              sizeof(struct ip_set_req_nethash_create),
6754 +                              size);
6755 +               return -EINVAL;
6756 +       }
6757 +
6758 +       if (req->hashsize < 1) {
6759 +               ip_set_printk("hashsize too small");
6760 +               return -ENOEXEC;
6761 +       }
6762 +       if (req->probes < 1) {
6763 +               ip_set_printk("probes too small");
6764 +               return -ENOEXEC;
6765 +       }
6766 +
6767 +       map = kmalloc(sizeof(struct ip_set_nethash)
6768 +                     + req->probes * sizeof(uint32_t), GFP_KERNEL);
6769 +       if (!map) {
6770 +               DP("out of memory for %d bytes",
6771 +                  sizeof(struct ip_set_nethash)
6772 +                  + req->probes * sizeof(uint32_t));
6773 +               return -ENOMEM;
6774 +       }
6775 +       for (i = 0; i < req->probes; i++)
6776 +               get_random_bytes(((uint32_t *) map->initval)+i, 4);
6777 +       map->elements = 0;
6778 +       map->hashsize = req->hashsize;
6779 +       map->probes = req->probes;
6780 +       map->resize = req->resize;
6781 +       memset(map->cidr, 0, 30 * sizeof(unsigned char));
6782 +       map->members = harray_malloc(map->hashsize, sizeof(ip_set_ip_t), GFP_KERNEL);
6783 +       if (!map->members) {
6784 +               DP("out of memory for %d bytes", map->hashsize * sizeof(ip_set_ip_t));
6785 +               kfree(map);
6786 +               return -ENOMEM;
6787 +       }
6788 +
6789 +       set->data = map;
6790 +       return 0;
6791 +}
6792 +
6793 +static void destroy(struct ip_set *set)
6794 +{
6795 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6796 +
6797 +       harray_free(map->members);
6798 +       kfree(map);
6799 +
6800 +       set->data = NULL;
6801 +}
6802 +
6803 +static void flush(struct ip_set *set)
6804 +{
6805 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6806 +       harray_flush(map->members, map->hashsize, sizeof(ip_set_ip_t));
6807 +       memset(map->cidr, 0, 30 * sizeof(unsigned char));
6808 +       map->elements = 0;
6809 +}
6810 +
6811 +static void list_header(const struct ip_set *set, void *data)
6812 +{
6813 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6814 +       struct ip_set_req_nethash_create *header =
6815 +           (struct ip_set_req_nethash_create *) data;
6816 +
6817 +       header->hashsize = map->hashsize;
6818 +       header->probes = map->probes;
6819 +       header->resize = map->resize;
6820 +}
6821 +
6822 +static int list_members_size(const struct ip_set *set)
6823 +{
6824 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6825 +
6826 +       return (map->hashsize * sizeof(ip_set_ip_t));
6827 +}
6828 +
6829 +static void list_members(const struct ip_set *set, void *data)
6830 +{
6831 +       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
6832 +       ip_set_ip_t i, *elem;
6833 +
6834 +       for (i = 0; i < map->hashsize; i++) {
6835 +               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);
6836 +               ((ip_set_ip_t *)data)[i] = *elem;
6837 +       }
6838 +}
6839 +
6840 +static struct ip_set_type ip_set_nethash = {
6841 +       .typename               = SETTYPE_NAME,
6842 +       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
6843 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
6844 +       .create                 = &create,
6845 +       .destroy                = &destroy,
6846 +       .flush                  = &flush,
6847 +       .reqsize                = sizeof(struct ip_set_req_nethash),
6848 +       .addip                  = &addip,
6849 +       .addip_kernel           = &addip_kernel,
6850 +       .retry                  = &retry,
6851 +       .delip                  = &delip,
6852 +       .delip_kernel           = &delip_kernel,
6853 +       .testip                 = &testip,
6854 +       .testip_kernel          = &testip_kernel,
6855 +       .header_size            = sizeof(struct ip_set_req_nethash_create),
6856 +       .list_header            = &list_header,
6857 +       .list_members_size      = &list_members_size,
6858 +       .list_members           = &list_members,
6859 +       .me                     = THIS_MODULE,
6860 +};
6861 +
6862 +MODULE_LICENSE("GPL");
6863 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
6864 +MODULE_DESCRIPTION("nethash type of IP sets");
6865 +module_param(limit, int, 0600);
6866 +MODULE_PARM_DESC(limit, "maximal number of elements stored in the sets");
6867 +
6868 +static int __init ip_set_nethash_init(void)
6869 +{
6870 +       return ip_set_register_set_type(&ip_set_nethash);
6871 +}
6872 +
6873 +static void __exit ip_set_nethash_fini(void)
6874 +{
6875 +       /* FIXME: possible race with ip_set_create() */
6876 +       ip_set_unregister_set_type(&ip_set_nethash);
6877 +}
6878 +
6879 +module_init(ip_set_nethash_init);
6880 +module_exit(ip_set_nethash_fini);
6881 Index: linux-2.6.21.7/net/ipv4/netfilter/ip_set_portmap.c
6882 ===================================================================
6883 --- /dev/null
6884 +++ linux-2.6.21.7/net/ipv4/netfilter/ip_set_portmap.c
6885 @@ -0,0 +1,346 @@
6886 +/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
6887 + *
6888 + * This program is free software; you can redistribute it and/or modify
6889 + * it under the terms of the GNU General Public License version 2 as
6890 + * published by the Free Software Foundation.
6891 + */
6892 +
6893 +/* Kernel module implementing a port set type as a bitmap */
6894 +
6895 +#include <linux/module.h>
6896 +#include <linux/ip.h>
6897 +#include <linux/tcp.h>
6898 +#include <linux/udp.h>
6899 +#include <linux/skbuff.h>
6900 +#include <linux/version.h>
6901 +#include <linux/netfilter_ipv4/ip_tables.h>
6902 +#include <linux/netfilter_ipv4/ip_set.h>
6903 +#include <linux/errno.h>
6904 +#include <asm/uaccess.h>
6905 +#include <asm/bitops.h>
6906 +#include <linux/spinlock.h>
6907 +
6908 +#include <net/ip.h>
6909 +
6910 +#include <linux/netfilter_ipv4/ip_set_portmap.h>
6911 +
6912 +/* We must handle non-linear skbs */
6913 +static inline ip_set_ip_t
6914 +get_port(const struct sk_buff *skb, u_int32_t flags)
6915 +{
6916 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6917 +       struct iphdr *iph = ip_hdr(skb);
6918 +#else
6919 +       struct iphdr *iph = skb->nh.iph;
6920 +#endif
6921 +       u_int16_t offset = ntohs(iph->frag_off) & IP_OFFSET;
6922 +       switch (iph->protocol) {
6923 +       case IPPROTO_TCP: {
6924 +               struct tcphdr tcph;
6925 +
6926 +               /* See comments at tcp_match in ip_tables.c */
6927 +               if (offset)
6928 +                       return INVALID_PORT;
6929 +
6930 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6931 +               if (skb_copy_bits(skb, ip_hdr(skb)->ihl*4, &tcph, sizeof(tcph)) < 0)
6932 +#else
6933 +               if (skb_copy_bits(skb, skb->nh.iph->ihl*4, &tcph, sizeof(tcph)) < 0)
6934 +#endif
6935 +                       /* No choice either */
6936 +                       return INVALID_PORT;
6937 +
6938 +               return ntohs(flags & IPSET_SRC ?
6939 +                            tcph.source : tcph.dest);
6940 +           }
6941 +       case IPPROTO_UDP: {
6942 +               struct udphdr udph;
6943 +
6944 +               if (offset)
6945 +                       return INVALID_PORT;
6946 +
6947 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)
6948 +               if (skb_copy_bits(skb, ip_hdr(skb)->ihl*4, &udph, sizeof(udph)) < 0)
6949 +#else
6950 +               if (skb_copy_bits(skb, skb->nh.iph->ihl*4, &udph, sizeof(udph)) < 0)
6951 +#endif
6952 +                       /* No choice either */
6953 +                       return INVALID_PORT;
6954 +
6955 +               return ntohs(flags & IPSET_SRC ?
6956 +                            udph.source : udph.dest);
6957 +           }
6958 +       default:
6959 +               return INVALID_PORT;
6960 +       }
6961 +}
6962 +
6963 +static inline int
6964 +__testport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
6965 +{
6966 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
6967 +
6968 +       if (port < map->first_port || port > map->last_port)
6969 +               return -ERANGE;
6970 +
6971 +       *hash_port = port;
6972 +       DP("set: %s, port:%u, %u", set->name, port, *hash_port);
6973 +       return !!test_bit(port - map->first_port, map->members);
6974 +}
6975 +
6976 +static int
6977 +testport(struct ip_set *set, const void *data, size_t size,
6978 +         ip_set_ip_t *hash_port)
6979 +{
6980 +       struct ip_set_req_portmap *req =
6981 +           (struct ip_set_req_portmap *) data;
6982 +
6983 +       if (size != sizeof(struct ip_set_req_portmap)) {
6984 +               ip_set_printk("data length wrong (want %zu, have %zu)",
6985 +                             sizeof(struct ip_set_req_portmap),
6986 +                             size);
6987 +               return -EINVAL;
6988 +       }
6989 +       return __testport(set, req->port, hash_port);
6990 +}
6991 +
6992 +static int
6993 +testport_kernel(struct ip_set *set,
6994 +               const struct sk_buff *skb,
6995 +               ip_set_ip_t *hash_port,
6996 +               const u_int32_t *flags,
6997 +               unsigned char index)
6998 +{
6999 +       int res;
7000 +       ip_set_ip_t port = get_port(skb, flags[index]);
7001 +
7002 +       DP("flag %s port %u", flags[index] & IPSET_SRC ? "SRC" : "DST", port);
7003 +       if (port == INVALID_PORT)
7004 +               return 0;
7005 +
7006 +       res =  __testport(set, port, hash_port);
7007 +
7008 +       return (res < 0 ? 0 : res);
7009 +}
7010 +
7011 +static inline int
7012 +__addport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
7013 +{
7014 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
7015 +
7016 +       if (port < map->first_port || port > map->last_port)
7017 +               return -ERANGE;
7018 +       if (test_and_set_bit(port - map->first_port, map->members))
7019 +               return -EEXIST;
7020 +
7021 +       *hash_port = port;
7022 +       DP("port %u", port);
7023 +       return 0;
7024 +}
7025 +
7026 +static int
7027 +addport(struct ip_set *set, const void *data, size_t size,
7028 +        ip_set_ip_t *hash_port)
7029 +{
7030 +       struct ip_set_req_portmap *req =
7031 +           (struct ip_set_req_portmap *) data;
7032 +
7033 +       if (size != sizeof(struct ip_set_req_portmap)) {
7034 +               ip_set_printk("data length wrong (want %zu, have %zu)",
7035 +                             sizeof(struct ip_set_req_portmap),
7036 +                             size);
7037 +               return -EINVAL;
7038 +       }
7039 +       return __addport(set, req->port, hash_port);
7040 +}
7041 +
7042 +static int
7043 +addport_kernel(struct ip_set *set,
7044 +              const struct sk_buff *skb,
7045 +              ip_set_ip_t *hash_port,
7046 +              const u_int32_t *flags,
7047 +              unsigned char index)
7048 +{
7049 +       ip_set_ip_t port = get_port(skb, flags[index]);
7050 +
7051 +       if (port == INVALID_PORT)
7052 +               return -EINVAL;
7053 +
7054 +       return __addport(set, port, hash_port);
7055 +}
7056 +
7057 +static inline int
7058 +__delport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
7059 +{
7060 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
7061 +
7062 +       if (port < map->first_port || port > map->last_port)
7063 +               return -ERANGE;
7064 +       if (!test_and_clear_bit(port - map->first_port, map->members))
7065 +               return -EEXIST;
7066 +
7067 +       *hash_port = port;
7068 +       DP("port %u", port);
7069 +       return 0;
7070 +}
7071 +
7072 +static int
7073 +delport(struct ip_set *set, const void *data, size_t size,
7074 +        ip_set_ip_t *hash_port)
7075 +{
7076 +       struct ip_set_req_portmap *req =
7077 +           (struct ip_set_req_portmap *) data;
7078 +
7079 +       if (size != sizeof(struct ip_set_req_portmap)) {
7080 +               ip_set_printk("data length wrong (want %zu, have %zu)",
7081 +                             sizeof(struct ip_set_req_portmap),
7082 +                             size);
7083 +               return -EINVAL;
7084 +       }
7085 +       return __delport(set, req->port, hash_port);
7086 +}
7087 +
7088 +static int
7089 +delport_kernel(struct ip_set *set,
7090 +              const struct sk_buff *skb,
7091 +              ip_set_ip_t *hash_port,
7092 +              const u_int32_t *flags,
7093 +              unsigned char index)
7094 +{
7095 +       ip_set_ip_t port = get_port(skb, flags[index]);
7096 +
7097 +       if (port == INVALID_PORT)
7098 +               return -EINVAL;
7099 +
7100 +       return __delport(set, port, hash_port);
7101 +}
7102 +
7103 +static int create(struct ip_set *set, const void *data, size_t size)
7104 +{
7105 +       int newbytes;
7106 +       struct ip_set_req_portmap_create *req =
7107 +           (struct ip_set_req_portmap_create *) data;
7108 +       struct ip_set_portmap *map;
7109 +
7110 +       if (size != sizeof(struct ip_set_req_portmap_create)) {
7111 +               ip_set_printk("data length wrong (want %zu, have %zu)",
7112 +                              sizeof(struct ip_set_req_portmap_create),
7113 +                              size);
7114 +               return -EINVAL;
7115 +       }
7116 +
7117 +       DP("from %u to %u", req->from, req->to);
7118 +
7119 +       if (req->from > req->to) {
7120 +               DP("bad port range");
7121 +               return -ENOEXEC;
7122 +       }
7123 +
7124 +       if (req->to - req->from > MAX_RANGE) {
7125 +               ip_set_printk("range too big (max %d ports)",
7126 +                              MAX_RANGE+1);
7127 +               return -ENOEXEC;
7128 +       }
7129 +
7130 +       map = kmalloc(sizeof(struct ip_set_portmap), GFP_KERNEL);
7131 +       if (!map) {
7132 +               DP("out of memory for %d bytes",
7133 +                  sizeof(struct ip_set_portmap));
7134 +               return -ENOMEM;
7135 +       }
7136 +       map->first_port = req->from;
7137 +       map->last_port = req->to;
7138 +       newbytes = bitmap_bytes(req->from, req->to);
7139 +       map->members = kmalloc(newbytes, GFP_KERNEL);
7140 +       if (!map->members) {
7141 +               DP("out of memory for %d bytes", newbytes);
7142 +               kfree(map);
7143 +               return -ENOMEM;
7144 +       }
7145 +       memset(map->members, 0, newbytes);
7146 +
7147 +       set->data = map;
7148 +       return 0;
7149 +}
7150 +
7151 +static void destroy(struct ip_set *set)
7152 +{
7153 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
7154 +
7155 +       kfree(map->members);
7156 +       kfree(map);
7157 +
7158 +       set->data = NULL;
7159 +}
7160 +
7161 +static void flush(struct ip_set *set)
7162 +{
7163 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
7164 +       memset(map->members, 0, bitmap_bytes(map->first_port, map->last_port));
7165 +}
7166 +
7167 +static void list_header(const struct ip_set *set, void *data)
7168 +{
7169 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
7170 +       struct ip_set_req_portmap_create *header =
7171 +           (struct ip_set_req_portmap_create *) data;
7172 +
7173 +       DP("list_header %u %u", map->first_port, map->last_port);
7174 +
7175 +       header->from = map->first_port;
7176 +       header->to = map->last_port;
7177 +}
7178 +
7179 +static int list_members_size(const struct ip_set *set)
7180 +{
7181 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
7182 +
7183 +       return bitmap_bytes(map->first_port, map->last_port);
7184 +}
7185 +
7186 +static void list_members(const struct ip_set *set, void *data)
7187 +{
7188 +       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
7189 +       int bytes = bitmap_bytes(map->first_port, map->last_port);
7190 +
7191 +       memcpy(data, map->members, bytes);
7192 +}
7193 +
7194 +static struct ip_set_type ip_set_portmap = {
7195 +       .typename               = SETTYPE_NAME,
7196 +       .features               = IPSET_TYPE_PORT | IPSET_DATA_SINGLE,
7197 +       .protocol_version       = IP_SET_PROTOCOL_VERSION,
7198 +       .create                 = &create,
7199 +       .destroy                = &destroy,
7200 +       .flush                  = &flush,
7201 +       .reqsize                = sizeof(struct ip_set_req_portmap),
7202 +       .addip                  = &addport,
7203 +       .addip_kernel           = &addport_kernel,
7204 +       .delip                  = &delport,
7205 +       .delip_kernel           = &delport_kernel,
7206 +       .testip                 = &testport,
7207 +       .testip_kernel          = &testport_kernel,
7208 +       .header_size            = sizeof(struct ip_set_req_portmap_create),
7209 +       .list_header            = &list_header,
7210 +       .list_members_size      = &list_members_size,
7211 +       .list_members           = &list_members,
7212 +       .me                     = THIS_MODULE,
7213 +};
7214 +
7215 +MODULE_LICENSE("GPL");
7216 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
7217 +MODULE_DESCRIPTION("portmap type of IP sets");
7218 +
7219 +static int __init ip_set_portmap_init(void)
7220 +{
7221 +       return ip_set_register_set_type(&ip_set_portmap);
7222 +}
7223 +
7224 +static void __exit ip_set_portmap_fini(void)
7225 +{
7226 +       /* FIXME: possible race with ip_set_create() */
7227 +       ip_set_unregister_set_type(&ip_set_portmap);
7228 +}
7229 +
7230 +module_init(ip_set_portmap_init);
7231 +module_exit(ip_set_portmap_fini);
7232 Index: linux-2.6.21.7/net/ipv4/netfilter/ipt_set.c
7233 ===================================================================
7234 --- /dev/null
7235 +++ linux-2.6.21.7/net/ipv4/netfilter/ipt_set.c
7236 @@ -0,0 +1,160 @@
7237 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
7238 + *                         Patrick Schaaf <bof@bof.de>
7239 + *                         Martin Josefsson <gandalf@wlug.westbo.se>
7240 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
7241 + *
7242 + * This program is free software; you can redistribute it and/or modify
7243 + * it under the terms of the GNU General Public License version 2 as
7244 + * published by the Free Software Foundation.
7245 + */
7246 +
7247 +/* Kernel module to match an IP set. */
7248 +
7249 +#include <linux/module.h>
7250 +#include <linux/ip.h>
7251 +#include <linux/skbuff.h>
7252 +#include <linux/version.h>
7253 +
7254 +#include <linux/netfilter_ipv4/ip_tables.h>
7255 +#include <linux/netfilter_ipv4/ip_set.h>
7256 +#include <linux/netfilter_ipv4/ipt_set.h>
7257 +
7258 +static inline int
7259 +match_set(const struct ipt_set_info *info,
7260 +         const struct sk_buff *skb,
7261 +         int inv)
7262 +{
7263 +       if (ip_set_testip_kernel(info->index, skb, info->flags))
7264 +               inv = !inv;
7265 +       return inv;
7266 +}
7267 +
7268 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
7269 +static bool
7270 +#else
7271 +static int
7272 +#endif
7273 +match(const struct sk_buff *skb,
7274 +      const struct net_device *in,
7275 +      const struct net_device *out,
7276 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
7277 +      const struct xt_match *match,
7278 +#endif
7279 +      const void *matchinfo,
7280 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
7281 +      int offset, unsigned int protoff, bool *hotdrop)
7282 +#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16)
7283 +      int offset, unsigned int protoff, int *hotdrop)
7284 +#else
7285 +      int offset, int *hotdrop)
7286 +#endif
7287 +{
7288 +       const struct ipt_set_info_match *info = matchinfo;
7289 +
7290 +       return match_set(&info->match_set,
7291 +                        skb,
7292 +                        info->match_set.flags[0] & IPSET_MATCH_INV);
7293 +}
7294 +
7295 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
7296 +bool
7297 +#else
7298 +static int
7299 +#endif
7300 +checkentry(const char *tablename,
7301 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16)
7302 +          const void *inf,
7303 +#else
7304 +          const struct ipt_ip *ip,
7305 +#endif
7306 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
7307 +          const struct xt_match *match,
7308 +#endif
7309 +          void *matchinfo,
7310 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7311 +          unsigned int matchsize,
7312 +#endif
7313 +          unsigned int hook_mask)
7314 +{
7315 +       struct ipt_set_info_match *info =
7316 +               (struct ipt_set_info_match *) matchinfo;
7317 +       ip_set_id_t index;
7318 +
7319 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7320 +       if (matchsize != IPT_ALIGN(sizeof(struct ipt_set_info_match))) {
7321 +               ip_set_printk("invalid matchsize %d", matchsize);
7322 +               return 0;
7323 +       }
7324 +#endif
7325 +
7326 +       index = ip_set_get_byindex(info->match_set.index);
7327 +
7328 +       if (index == IP_SET_INVALID_ID) {
7329 +               ip_set_printk("Cannot find set indentified by id %u to match",
7330 +                             info->match_set.index);
7331 +               return 0;       /* error */
7332 +       }
7333 +       if (info->match_set.flags[IP_SET_MAX_BINDINGS] != 0) {
7334 +               ip_set_printk("That's nasty!");
7335 +               return 0;       /* error */
7336 +       }
7337 +
7338 +       return 1;
7339 +}
7340 +
7341 +static void destroy(
7342 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
7343 +                   const struct xt_match *match,
7344 +#endif
7345 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7346 +                   void *matchinfo, unsigned int matchsize)
7347 +#else
7348 +                   void *matchinfo)
7349 +#endif
7350 +{
7351 +       struct ipt_set_info_match *info = matchinfo;
7352 +
7353 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7354 +       if (matchsize != IPT_ALIGN(sizeof(struct ipt_set_info_match))) {
7355 +               ip_set_printk("invalid matchsize %d", matchsize);
7356 +               return;
7357 +       }
7358 +#endif
7359 +       ip_set_put(info->match_set.index);
7360 +}
7361 +
7362 +static struct ipt_match set_match = {
7363 +       .name           = "set",
7364 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
7365 +       .family         = AF_INET,
7366 +#endif
7367 +       .match          = &match,
7368 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
7369 +       .matchsize      = sizeof(struct ipt_set_info_match),
7370 +#endif
7371 +       .checkentry     = &checkentry,
7372 +       .destroy        = &destroy,
7373 +       .me             = THIS_MODULE
7374 +};
7375 +
7376 +MODULE_LICENSE("GPL");
7377 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
7378 +MODULE_DESCRIPTION("iptables IP set match module");
7379 +
7380 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
7381 +#define ipt_register_match     xt_register_match
7382 +#define ipt_unregister_match   xt_unregister_match
7383 +#endif
7384 +
7385 +static int __init ipt_ipset_init(void)
7386 +{
7387 +       return ipt_register_match(&set_match);
7388 +}
7389 +
7390 +static void __exit ipt_ipset_fini(void)
7391 +{
7392 +       ipt_unregister_match(&set_match);
7393 +}
7394 +
7395 +module_init(ipt_ipset_init);
7396 +module_exit(ipt_ipset_fini);
7397 Index: linux-2.6.21.7/net/ipv4/netfilter/ipt_SET.c
7398 ===================================================================
7399 --- /dev/null
7400 +++ linux-2.6.21.7/net/ipv4/netfilter/ipt_SET.c
7401 @@ -0,0 +1,172 @@
7402 +/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
7403 + *                         Patrick Schaaf <bof@bof.de>
7404 + *                         Martin Josefsson <gandalf@wlug.westbo.se>
7405 + * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
7406 + *
7407 + * This program is free software; you can redistribute it and/or modify
7408 + * it under the terms of the GNU General Public License version 2 as
7409 + * published by the Free Software Foundation.
7410 + */
7411 +
7412 +/* ipt_SET.c - netfilter target to manipulate IP sets */
7413 +
7414 +#include <linux/types.h>
7415 +#include <linux/ip.h>
7416 +#include <linux/timer.h>
7417 +#include <linux/module.h>
7418 +#include <linux/netfilter.h>
7419 +#include <linux/netdevice.h>
7420 +#include <linux/if.h>
7421 +#include <linux/inetdevice.h>
7422 +#include <linux/version.h>
7423 +#include <net/protocol.h>
7424 +#include <net/checksum.h>
7425 +#include <linux/netfilter_ipv4.h>
7426 +#include <linux/netfilter_ipv4/ip_tables.h>
7427 +#include <linux/netfilter_ipv4/ipt_set.h>
7428 +
7429 +static unsigned int
7430 +target(struct sk_buff **pskb,
7431 +       const struct net_device *in,
7432 +       const struct net_device *out,
7433 +       unsigned int hooknum,
7434 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
7435 +       const struct xt_target *target,
7436 +#endif
7437 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7438 +       const void *targinfo,
7439 +       void *userinfo)
7440 +#else
7441 +       const void *targinfo)
7442 +#endif
7443 +{
7444 +       const struct ipt_set_info_target *info = targinfo;
7445 +
7446 +       if (info->add_set.index != IP_SET_INVALID_ID)
7447 +               ip_set_addip_kernel(info->add_set.index,
7448 +                                   *pskb,
7449 +                                   info->add_set.flags);
7450 +       if (info->del_set.index != IP_SET_INVALID_ID)
7451 +               ip_set_delip_kernel(info->del_set.index,
7452 +                                   *pskb,
7453 +                                   info->del_set.flags);
7454 +
7455 +       return IPT_CONTINUE;
7456 +}
7457 +
7458 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
7459 +static bool
7460 +#else
7461 +static int
7462 +#endif
7463 +checkentry(const char *tablename,
7464 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16)
7465 +          const void *e,
7466 +#else
7467 +          const struct ipt_entry *e,
7468 +#endif
7469 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
7470 +          const struct xt_target *target,
7471 +#endif
7472 +          void *targinfo,
7473 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7474 +          unsigned int targinfosize,
7475 +#endif
7476 +          unsigned int hook_mask)
7477 +{
7478 +       struct ipt_set_info_target *info =
7479 +               (struct ipt_set_info_target *) targinfo;
7480 +       ip_set_id_t index;
7481 +
7482 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7483 +       if (targinfosize != IPT_ALIGN(sizeof(*info))) {
7484 +               DP("bad target info size %u", targinfosize);
7485 +               return 0;
7486 +       }
7487 +#endif
7488 +
7489 +       if (info->add_set.index != IP_SET_INVALID_ID) {
7490 +               index = ip_set_get_byindex(info->add_set.index);
7491 +               if (index == IP_SET_INVALID_ID) {
7492 +                       ip_set_printk("cannot find add_set index %u as target",
7493 +                                     info->add_set.index);
7494 +                       return 0;       /* error */
7495 +               }
7496 +       }
7497 +
7498 +       if (info->del_set.index != IP_SET_INVALID_ID) {
7499 +               index = ip_set_get_byindex(info->del_set.index);
7500 +               if (index == IP_SET_INVALID_ID) {
7501 +                       ip_set_printk("cannot find del_set index %u as target",
7502 +                                     info->del_set.index);
7503 +                       return 0;       /* error */
7504 +               }
7505 +       }
7506 +       if (info->add_set.flags[IP_SET_MAX_BINDINGS] != 0
7507 +           || info->del_set.flags[IP_SET_MAX_BINDINGS] != 0) {
7508 +               ip_set_printk("That's nasty!");
7509 +               return 0;       /* error */
7510 +       }
7511 +
7512 +       return 1;
7513 +}
7514 +
7515 +static void destroy(
7516 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
7517 +                   const struct xt_target *target,
7518 +#endif
7519 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7520 +                   void *targetinfo, unsigned int targetsize)
7521 +#else
7522 +                   void *targetinfo)
7523 +#endif
7524 +{
7525 +       struct ipt_set_info_target *info = targetinfo;
7526 +
7527 +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
7528 +       if (targetsize != IPT_ALIGN(sizeof(struct ipt_set_info_target))) {
7529 +               ip_set_printk("invalid targetsize %d", targetsize);
7530 +               return;
7531 +       }
7532 +#endif
7533 +       if (info->add_set.index != IP_SET_INVALID_ID)
7534 +               ip_set_put(info->add_set.index);
7535 +       if (info->del_set.index != IP_SET_INVALID_ID)
7536 +               ip_set_put(info->del_set.index);
7537 +}
7538 +
7539 +static struct ipt_target SET_target = {
7540 +       .name           = "SET",
7541 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
7542 +       .family         = AF_INET,
7543 +#endif
7544 +       .target         = target,
7545 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
7546 +       .targetsize     = sizeof(struct ipt_set_info_target),
7547 +#endif
7548 +       .checkentry     = checkentry,
7549 +       .destroy        = destroy,
7550 +       .me             = THIS_MODULE
7551 +};
7552 +
7553 +MODULE_LICENSE("GPL");
7554 +MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
7555 +MODULE_DESCRIPTION("iptables IP set target module");
7556 +
7557 +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
7558 +#define ipt_register_target      xt_register_target
7559 +#define ipt_unregister_target    xt_unregister_target
7560 +#endif
7561 +
7562 +static int __init ipt_SET_init(void)
7563 +{
7564 +       return ipt_register_target(&SET_target);
7565 +}
7566 +
7567 +static void __exit ipt_SET_fini(void)
7568 +{
7569 +       ipt_unregister_target(&SET_target);
7570 +}
7571 +
7572 +module_init(ipt_SET_init);
7573 +module_exit(ipt_SET_fini);
7574 Index: linux-2.6.21.7/net/ipv4/netfilter/Kconfig
7575 ===================================================================
7576 --- linux-2.6.21.7.orig/net/ipv4/netfilter/Kconfig
7577 +++ linux-2.6.21.7/net/ipv4/netfilter/Kconfig
7578 @@ -663,5 +663,122 @@ config IP_NF_ARP_MANGLE
7579           Allows altering the ARP packet payload: source and destination
7580           hardware and network addresses.
7581
7582 +config IP_NF_SET
7583 +       tristate "IP set support"
7584 +       depends on INET && NETFILTER
7585 +       help
7586 +         This option adds IP set support to the kernel.
7587 +         In order to define and use sets, you need the userspace utility
7588 +         ipset(8).
7589 +
7590 +         To compile it as a module, choose M here.  If unsure, say N.
7591 +
7592 +config IP_NF_SET_MAX
7593 +       int "Maximum number of IP sets"
7594 +       default 256
7595 +       range 2 65534
7596 +       depends on IP_NF_SET
7597 +       help
7598 +         You can define here default value of the maximum number
7599 +         of IP sets for the kernel.
7600 +
7601 +         The value can be overriden by the 'max_sets' module
7602 +         parameter of the 'ip_set' module.
7603 +
7604 +config IP_NF_SET_HASHSIZE
7605 +       int "Hash size for bindings of IP sets"
7606 +       default 1024
7607 +       depends on IP_NF_SET
7608 +       help
7609 +         You can define here default value of the hash size for
7610 +         bindings of IP sets.
7611 +
7612 +         The value can be overriden by the 'hash_size' module
7613 +         parameter of the 'ip_set' module.
7614 +
7615 +config IP_NF_SET_IPMAP
7616 +       tristate "ipmap set support"
7617 +       depends on IP_NF_SET
7618 +       help
7619 +         This option adds the ipmap set type support.
7620 +
7621 +         To compile it as a module, choose M here.  If unsure, say N.
7622 +
7623 +config IP_NF_SET_MACIPMAP
7624 +       tristate "macipmap set support"
7625 +       depends on IP_NF_SET
7626 +       help
7627 +         This option adds the macipmap set type support.
7628 +
7629 +         To compile it as a module, choose M here.  If unsure, say N.
7630 +
7631 +config IP_NF_SET_PORTMAP
7632 +       tristate "portmap set support"
7633 +       depends on IP_NF_SET
7634 +       help
7635 +         This option adds the portmap set type support.
7636 +
7637 +         To compile it as a module, choose M here.  If unsure, say N.
7638 +
7639 +config IP_NF_SET_IPHASH
7640 +       tristate "iphash set support"
7641 +       depends on IP_NF_SET
7642 +       help
7643 +         This option adds the iphash set type support.
7644 +
7645 +         To compile it as a module, choose M here.  If unsure, say N.
7646 +
7647 +config IP_NF_SET_NETHASH
7648 +       tristate "nethash set support"
7649 +       depends on IP_NF_SET
7650 +       help
7651 +         This option adds the nethash set type support.
7652 +
7653 +         To compile it as a module, choose M here.  If unsure, say N.
7654 +
7655 +config IP_NF_SET_IPPORTHASH
7656 +       tristate "ipporthash set support"
7657 +       depends on IP_NF_SET
7658 +       help
7659 +         This option adds the ipporthash set type support.
7660 +
7661 +         To compile it as a module, choose M here.  If unsure, say N.
7662 +
7663 +config IP_NF_SET_IPTREE
7664 +       tristate "iptree set support"
7665 +       depends on IP_NF_SET
7666 +       help
7667 +         This option adds the iptree set type support.
7668 +
7669 +         To compile it as a module, choose M here.  If unsure, say N.
7670 +
7671 +config IP_NF_SET_IPTREEMAP
7672 +       tristate "iptreemap set support"
7673 +       depends on IP_NF_SET
7674 +       help
7675 +         This option adds the iptreemap set type support.
7676 +
7677 +         To compile it as a module, choose M here.  If unsure, say N.
7678 +
7679 +config IP_NF_MATCH_SET
7680 +       tristate "set match support"
7681 +       depends on IP_NF_SET
7682 +       help
7683 +         Set matching matches against given IP sets.
7684 +         You need the ipset utility to create and set up the sets.
7685 +
7686 +         To compile it as a module, choose M here.  If unsure, say N.
7687 +
7688 +config IP_NF_TARGET_SET
7689 +       tristate "SET target support"
7690 +       depends on IP_NF_SET
7691 +       help
7692 +         The SET target makes possible to add/delete entries
7693 +         in IP sets.
7694 +         You need the ipset utility to create and set up the sets.
7695 +
7696 +         To compile it as a module, choose M here.  If unsure, say N.
7697 +
7698 +
7699  endmenu
7700
7701 Index: linux-2.6.21.7/net/ipv4/netfilter/Makefile
7702 ===================================================================
7703 --- linux-2.6.21.7.orig/net/ipv4/netfilter/Makefile
7704 +++ linux-2.6.21.7/net/ipv4/netfilter/Makefile
7705 @@ -90,6 +90,7 @@ obj-$(CONFIG_IP_NF_MATCH_RECENT) += ipt_
7706  obj-$(CONFIG_IP_NF_MATCH_ECN) += ipt_ecn.o
7707  obj-$(CONFIG_IP_NF_MATCH_AH) += ipt_ah.o
7708  obj-$(CONFIG_IP_NF_MATCH_TTL) += ipt_ttl.o
7709 +obj-$(CONFIG_IP_NF_MATCH_SET) += ipt_set.o
7710  obj-$(CONFIG_IP_NF_MATCH_ADDRTYPE) += ipt_addrtype.o
7711  obj-$(CONFIG_IP_NF_MATCH_IPP2P) += ipt_ipp2p.o
7712
7713 @@ -106,6 +107,18 @@ obj-$(CONFIG_IP_NF_TARGET_LOG) += ipt_LO
7714  obj-$(CONFIG_IP_NF_TARGET_ULOG) += ipt_ULOG.o
7715  obj-$(CONFIG_IP_NF_TARGET_CLUSTERIP) += ipt_CLUSTERIP.o
7716  obj-$(CONFIG_IP_NF_TARGET_TTL) += ipt_TTL.o
7717 +obj-$(CONFIG_IP_NF_TARGET_SET) += ipt_SET.o
7718 +
7719 +# sets
7720 +obj-$(CONFIG_IP_NF_SET) += ip_set.o
7721 +obj-$(CONFIG_IP_NF_SET_IPMAP) += ip_set_ipmap.o
7722 +obj-$(CONFIG_IP_NF_SET_PORTMAP) += ip_set_portmap.o
7723 +obj-$(CONFIG_IP_NF_SET_MACIPMAP) += ip_set_macipmap.o
7724 +obj-$(CONFIG_IP_NF_SET_IPHASH) += ip_set_iphash.o
7725 +obj-$(CONFIG_IP_NF_SET_NETHASH) += ip_set_nethash.o
7726 +obj-$(CONFIG_IP_NF_SET_IPPORTHASH) += ip_set_ipporthash.o
7727 +obj-$(CONFIG_IP_NF_SET_IPTREE) += ip_set_iptree.o
7728 +obj-$(CONFIG_IP_NF_SET_IPTREEMAP) += ip_set_iptreemap.o
7729
7730  # generic ARP tables
7731  obj-$(CONFIG_IP_NF_ARPTABLES) += arp_tables.o