This commit was manufactured by cvs2svn to create branch 'vserver'.
authorPlanet-Lab Support <support@planet-lab.org>
Thu, 27 Jul 2006 21:05:48 +0000 (21:05 +0000)
committerPlanet-Lab Support <support@planet-lab.org>
Thu, 27 Jul 2006 21:05:48 +0000 (21:05 +0000)
22 files changed:
include/linux/netfilter_ipv4/ip_set.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_iphash.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_ipmap.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_ipporthash.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_iptree.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_jhash.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_macipmap.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_malloc.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_nethash.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ip_set_portmap.h [new file with mode: 0644]
include/linux/netfilter_ipv4/ipt_set.h [new file with mode: 0644]
include/linux/vserver/limit_int.h [new file with mode: 0644]
net/ipv4/netfilter/ip_set.c [new file with mode: 0644]
net/ipv4/netfilter/ip_set_iphash.c [new file with mode: 0644]
net/ipv4/netfilter/ip_set_ipmap.c [new file with mode: 0644]
net/ipv4/netfilter/ip_set_ipporthash.c [new file with mode: 0644]
net/ipv4/netfilter/ip_set_iptree.c [new file with mode: 0644]
net/ipv4/netfilter/ip_set_macipmap.c [new file with mode: 0644]
net/ipv4/netfilter/ip_set_nethash.c [new file with mode: 0644]
net/ipv4/netfilter/ip_set_portmap.c [new file with mode: 0644]
net/ipv4/netfilter/ipt_SET.c [new file with mode: 0644]
net/ipv4/netfilter/ipt_set.c [new file with mode: 0644]

diff --git a/include/linux/netfilter_ipv4/ip_set.h b/include/linux/netfilter_ipv4/ip_set.h
new file mode 100644 (file)
index 0000000..92a746e
--- /dev/null
@@ -0,0 +1,498 @@
+#ifndef _IP_SET_H
+#define _IP_SET_H
+
+/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
+ *                         Patrick Schaaf <bof@bof.de>
+ *                         Martin Josefsson <gandalf@wlug.westbo.se>
+ * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+#if 0
+#define IP_SET_DEBUG
+#endif
+
+/*
+ * A sockopt of such quality has hardly ever been seen before on the open
+ * market!  This little beauty, hardly ever used: above 64, so it's
+ * traditionally used for firewalling, not touched (even once!) by the
+ * 2.0, 2.2 and 2.4 kernels!
+ *
+ * Comes with its own certificate of authenticity, valid anywhere in the
+ * Free world!
+ *
+ * Rusty, 19.4.2000
+ */
+#define SO_IP_SET              83
+
+/*
+ * Heavily modify by Joakim Axelsson 08.03.2002
+ * - Made it more modulebased
+ *
+ * Additional heavy modifications by Jozsef Kadlecsik 22.02.2004
+ * - bindings added
+ * - in order to "deal with" backward compatibility, renamed to ipset
+ */
+
+/* 
+ * Used so that the kernel module and ipset-binary can match their versions 
+ */
+#define IP_SET_PROTOCOL_VERSION 2
+
+#define IP_SET_MAXNAMELEN 32   /* set names and set typenames */
+
+/* Lets work with our own typedef for representing an IP address.
+ * We hope to make the code more portable, possibly to IPv6...
+ *
+ * The representation works in HOST byte order, because most set types
+ * will perform arithmetic operations and compare operations.
+ * 
+ * For now the type is an uint32_t.
+ *
+ * Make sure to ONLY use the functions when translating and parsing
+ * in order to keep the host byte order and make it more portable:
+ *  parse_ip()
+ *  parse_mask()
+ *  parse_ipandmask()
+ *  ip_tostring()
+ * (Joakim: where are they???)
+ */
+
+typedef uint32_t ip_set_ip_t;
+
+/* Sets are identified by an id in kernel space. Tweak with ip_set_id_t
+ * and IP_SET_INVALID_ID if you want to increase the max number of sets.
+ */
+typedef uint16_t ip_set_id_t;
+
+#define IP_SET_INVALID_ID      65535
+
+/* How deep we follow bindings */
+#define IP_SET_MAX_BINDINGS    6
+
+/*
+ * Option flags for kernel operations (ipt_set_info)
+ */
+#define IPSET_SRC              0x01    /* Source match/add */
+#define IPSET_DST              0x02    /* Destination match/add */
+#define IPSET_MATCH_INV                0x04    /* Inverse matching */
+
+/*
+ * Set features
+ */
+#define IPSET_TYPE_IP          0x01    /* IP address type of set */
+#define IPSET_TYPE_PORT                0x02    /* Port type of set */
+#define IPSET_DATA_SINGLE      0x04    /* Single data storage */
+#define IPSET_DATA_DOUBLE      0x08    /* Double data storage */
+
+/* Reserved keywords */
+#define IPSET_TOKEN_DEFAULT    ":default:"
+#define IPSET_TOKEN_ALL                ":all:"
+
+/* SO_IP_SET operation constants, and their request struct types.
+ *
+ * Operation ids:
+ *       0-99:  commands with version checking
+ *     100-199: add/del/test/bind/unbind
+ *     200-299: list, save, restore
+ */
+
+/* Single shot operations: 
+ * version, create, destroy, flush, rename and swap 
+ *
+ * Sets are identified by name.
+ */
+
+#define IP_SET_REQ_STD         \
+       unsigned op;            \
+       unsigned version;       \
+       char name[IP_SET_MAXNAMELEN]
+
+#define IP_SET_OP_CREATE       0x00000001      /* Create a new (empty) set */
+struct ip_set_req_create {
+       IP_SET_REQ_STD;
+       char typename[IP_SET_MAXNAMELEN];
+};
+
+#define IP_SET_OP_DESTROY      0x00000002      /* Remove a (empty) set */
+struct ip_set_req_std {
+       IP_SET_REQ_STD;
+};
+
+#define IP_SET_OP_FLUSH                0x00000003      /* Remove all IPs in a set */
+/* Uses ip_set_req_std */
+
+#define IP_SET_OP_RENAME       0x00000004      /* Rename a set */
+/* Uses ip_set_req_create */
+
+#define IP_SET_OP_SWAP         0x00000005      /* Swap two sets */
+/* Uses ip_set_req_create */
+
+union ip_set_name_index {
+       char name[IP_SET_MAXNAMELEN];
+       ip_set_id_t index;
+};
+
+#define IP_SET_OP_GET_BYNAME   0x00000006      /* Get set index by name */
+struct ip_set_req_get_set {
+       unsigned op;
+       unsigned version;
+       union ip_set_name_index set;
+};
+
+#define IP_SET_OP_GET_BYINDEX  0x00000007      /* Get set name by index */
+/* Uses ip_set_req_get_set */
+
+#define IP_SET_OP_VERSION      0x00000100      /* Ask kernel version */
+struct ip_set_req_version {
+       unsigned op;
+       unsigned version;
+};
+
+/* Double shots operations: 
+ * add, del, test, bind and unbind.
+ *
+ * First we query the kernel to get the index and type of the target set,
+ * then issue the command. Validity of IP is checked in kernel in order
+ * to minimalize sockopt operations.
+ */
+
+/* Get minimal set data for add/del/test/bind/unbind IP */
+#define IP_SET_OP_ADT_GET      0x00000010      /* Get set and type */
+struct ip_set_req_adt_get {
+       unsigned op;
+       unsigned version;
+       union ip_set_name_index set;
+       char typename[IP_SET_MAXNAMELEN];
+};
+
+#define IP_SET_REQ_BYINDEX     \
+       unsigned op;            \
+       ip_set_id_t index;
+
+struct ip_set_req_adt {
+       IP_SET_REQ_BYINDEX;
+};
+
+#define IP_SET_OP_ADD_IP       0x00000101      /* Add an IP to a set */
+/* Uses ip_set_req_adt, with type specific addage */
+
+#define IP_SET_OP_DEL_IP       0x00000102      /* Remove an IP from a set */
+/* Uses ip_set_req_adt, with type specific addage */
+
+#define IP_SET_OP_TEST_IP      0x00000103      /* Test an IP in a set */
+/* Uses ip_set_req_adt, with type specific addage */
+
+#define IP_SET_OP_BIND_SET     0x00000104      /* Bind an IP to a set */
+/* Uses ip_set_req_bind, with type specific addage */
+struct ip_set_req_bind {
+       IP_SET_REQ_BYINDEX;
+       char binding[IP_SET_MAXNAMELEN];
+};
+
+#define IP_SET_OP_UNBIND_SET   0x00000105      /* Unbind an IP from a set */
+/* Uses ip_set_req_bind, with type speficic addage 
+ * index = 0 means unbinding for all sets */
+
+#define IP_SET_OP_TEST_BIND_SET        0x00000106      /* Test binding an IP to a set */
+/* Uses ip_set_req_bind, with type specific addage */
+
+/* Multiple shots operations: list, save, restore.
+ *
+ * - check kernel version and query the max number of sets
+ * - get the basic information on all sets
+ *   and size required for the next step
+ * - get actual set data: header, data, bindings
+ */
+
+/* Get max_sets and the index of a queried set
+ */
+#define IP_SET_OP_MAX_SETS     0x00000020
+struct ip_set_req_max_sets {
+       unsigned op;
+       unsigned version;
+       ip_set_id_t max_sets;           /* max_sets */
+       ip_set_id_t sets;               /* real number of sets */
+       union ip_set_name_index set;    /* index of set if name used */
+};
+
+/* Get the id and name of the sets plus size for next step */
+#define IP_SET_OP_LIST_SIZE    0x00000201
+#define IP_SET_OP_SAVE_SIZE    0x00000202
+struct ip_set_req_setnames {
+       unsigned op;
+       ip_set_id_t index;              /* set to list/save */
+       size_t size;                    /* size to get setdata/bindings */
+       /* followed by sets number of struct ip_set_name_list */
+};
+
+struct ip_set_name_list {
+       char name[IP_SET_MAXNAMELEN];
+       char typename[IP_SET_MAXNAMELEN];
+       ip_set_id_t index;
+       ip_set_id_t id;
+};
+
+/* The actual list operation */
+#define IP_SET_OP_LIST         0x00000203
+struct ip_set_req_list {
+       IP_SET_REQ_BYINDEX;
+       /* sets number of struct ip_set_list in reply */ 
+};
+
+struct ip_set_list {
+       ip_set_id_t index;
+       ip_set_id_t binding;
+       u_int32_t ref;
+       size_t header_size;     /* Set header data of header_size */
+       size_t members_size;    /* Set members data of members_size */
+       size_t bindings_size;   /* Set bindings data of bindings_size */
+};
+
+struct ip_set_hash_list {
+       ip_set_ip_t ip;
+       ip_set_id_t binding;
+};
+
+/* The save operation */
+#define IP_SET_OP_SAVE         0x00000204
+/* Uses ip_set_req_list, in the reply replaced by
+ * sets number of struct ip_set_save plus a marker
+ * ip_set_save followed by ip_set_hash_save structures.
+ */
+struct ip_set_save {
+       ip_set_id_t index;
+       ip_set_id_t binding;
+       size_t header_size;     /* Set header data of header_size */
+       size_t members_size;    /* Set members data of members_size */
+};
+
+/* At restoring, ip == 0 means default binding for the given set: */
+struct ip_set_hash_save {
+       ip_set_ip_t ip;
+       ip_set_id_t id;
+       ip_set_id_t binding;
+};
+
+/* The restore operation */
+#define IP_SET_OP_RESTORE      0x00000205
+/* Uses ip_set_req_setnames followed by ip_set_restore structures
+ * plus a marker ip_set_restore, followed by ip_set_hash_save 
+ * structures.
+ */
+struct ip_set_restore {
+       char name[IP_SET_MAXNAMELEN];
+       char typename[IP_SET_MAXNAMELEN];
+       ip_set_id_t index;
+       size_t header_size;     /* Create data of header_size */
+       size_t members_size;    /* Set members data of members_size */
+};
+
+static inline int bitmap_bytes(ip_set_ip_t a, ip_set_ip_t b)
+{
+       return 4 * ((((b - a + 8) / 8) + 3) / 4);
+}
+
+#ifdef __KERNEL__
+
+#define ip_set_printk(format, args...)                         \
+       do {                                                    \
+               printk("%s: %s: ", __FILE__, __FUNCTION__);     \
+               printk(format "\n" , ## args);                  \
+       } while (0)
+
+#if defined(IP_SET_DEBUG)
+#define DP(format, args...)                                    \
+       do {                                                    \
+               printk("%s: %s (DBG): ", __FILE__, __FUNCTION__);\
+               printk(format "\n" , ## args);                  \
+       } while (0)
+#define IP_SET_ASSERT(x)                                       \
+       do {                                                    \
+               if (!(x))                                       \
+                       printk("IP_SET_ASSERT: %s:%i(%s)\n",    \
+                               __FILE__, __LINE__, __FUNCTION__); \
+       } while (0)
+#else
+#define DP(format, args...)
+#define IP_SET_ASSERT(x)
+#endif
+
+struct ip_set;
+
+/*
+ * The ip_set_type definition - one per set type, e.g. "ipmap".
+ *
+ * Each individual set has a pointer, set->type, going to one
+ * of these structures. Function pointers inside the structure implement
+ * the real behaviour of the sets.
+ *
+ * If not mentioned differently, the implementation behind the function
+ * pointers of a set_type, is expected to return 0 if ok, and a negative
+ * errno (e.g. -EINVAL) on error.
+ */
+struct ip_set_type {
+       struct list_head list;  /* next in list of set types */
+
+       /* test for IP in set (kernel: iptables -m set src|dst)
+        * return 0 if not in set, 1 if in set.
+        */
+       int (*testip_kernel) (struct ip_set *set,
+                             const struct sk_buff * skb, 
+                             ip_set_ip_t *ip,
+                             const u_int32_t *flags,
+                             unsigned char index);
+
+       /* test for IP in set (userspace: ipset -T set IP)
+        * return 0 if not in set, 1 if in set.
+        */
+       int (*testip) (struct ip_set *set,
+                      const void *data, size_t size,
+                      ip_set_ip_t *ip);
+
+       /*
+        * Size of the data structure passed by when
+        * adding/deletin/testing an entry.
+        */
+       size_t reqsize;
+
+       /* Add IP into set (userspace: ipset -A set IP)
+        * Return -EEXIST if the address is already in the set,
+        * and -ERANGE if the address lies outside the set bounds.
+        * If the address was not already in the set, 0 is returned.
+        */
+       int (*addip) (struct ip_set *set, 
+                     const void *data, size_t size,
+                     ip_set_ip_t *ip);
+
+       /* Add IP into set (kernel: iptables ... -j SET set src|dst)
+        * Return -EEXIST if the address is already in the set,
+        * and -ERANGE if the address lies outside the set bounds.
+        * If the address was not already in the set, 0 is returned.
+        */
+       int (*addip_kernel) (struct ip_set *set,
+                            const struct sk_buff * skb, 
+                            ip_set_ip_t *ip,
+                            const u_int32_t *flags,
+                            unsigned char index);
+
+       /* remove IP from set (userspace: ipset -D set --entry x)
+        * Return -EEXIST if the address is NOT in the set,
+        * and -ERANGE if the address lies outside the set bounds.
+        * If the address really was in the set, 0 is returned.
+        */
+       int (*delip) (struct ip_set *set, 
+                     const void *data, size_t size,
+                     ip_set_ip_t *ip);
+
+       /* remove IP from set (kernel: iptables ... -j SET --entry x)
+        * Return -EEXIST if the address is NOT in the set,
+        * and -ERANGE if the address lies outside the set bounds.
+        * If the address really was in the set, 0 is returned.
+        */
+       int (*delip_kernel) (struct ip_set *set,
+                            const struct sk_buff * skb, 
+                            ip_set_ip_t *ip,
+                            const u_int32_t *flags,
+                            unsigned char index);
+
+       /* new set creation - allocated type specific items
+        */
+       int (*create) (struct ip_set *set,
+                      const void *data, size_t size);
+
+       /* retry the operation after successfully tweaking the set
+        */
+       int (*retry) (struct ip_set *set);
+
+       /* set destruction - free type specific items
+        * There is no return value.
+        * Can be called only when child sets are destroyed.
+        */
+       void (*destroy) (struct ip_set *set);
+
+       /* set flushing - reset all bits in the set, or something similar.
+        * There is no return value.
+        */
+       void (*flush) (struct ip_set *set);
+
+       /* Listing: size needed for header
+        */
+       size_t header_size;
+
+       /* Listing: Get the header
+        *
+        * Fill in the information in "data".
+        * This function is always run after list_header_size() under a 
+        * writelock on the set. Therefor is the length of "data" always 
+        * correct. 
+        */
+       void (*list_header) (const struct ip_set *set, 
+                            void *data);
+
+       /* Listing: Get the size for the set members
+        */
+       int (*list_members_size) (const struct ip_set *set);
+
+       /* Listing: Get the set members
+        *
+        * Fill in the information in "data".
+        * This function is always run after list_member_size() under a 
+        * writelock on the set. Therefor is the length of "data" always 
+        * correct. 
+        */
+       void (*list_members) (const struct ip_set *set,
+                             void *data);
+
+       char typename[IP_SET_MAXNAMELEN];
+       unsigned char features;
+       int protocol_version;
+
+       /* Set this to THIS_MODULE if you are a module, otherwise NULL */
+       struct module *me;
+};
+
+extern int ip_set_register_set_type(struct ip_set_type *set_type);
+extern void ip_set_unregister_set_type(struct ip_set_type *set_type);
+
+/* A generic ipset */
+struct ip_set {
+       char name[IP_SET_MAXNAMELEN];   /* the name of the set */
+       rwlock_t lock;                  /* lock for concurrency control */
+       ip_set_id_t id;                 /* set id for swapping */
+       ip_set_id_t binding;            /* default binding for the set */
+       atomic_t ref;                   /* in kernel and in hash references */
+       struct ip_set_type *type;       /* the set types */
+       void *data;                     /* pooltype specific data */
+};
+
+/* Structure to bind set elements to sets */
+struct ip_set_hash {
+       struct list_head list;          /* list of clashing entries in hash */
+       ip_set_ip_t ip;                 /* ip from set */
+       ip_set_id_t id;                 /* set id */
+       ip_set_id_t binding;            /* set we bind the element to */
+};
+
+/* register and unregister set references */
+extern ip_set_id_t ip_set_get_byname(const char name[IP_SET_MAXNAMELEN]);
+extern ip_set_id_t ip_set_get_byindex(ip_set_id_t id);
+extern void ip_set_put(ip_set_id_t id);
+
+/* API for iptables set match, and SET target */
+extern void ip_set_addip_kernel(ip_set_id_t id,
+                               const struct sk_buff *skb,
+                               const u_int32_t *flags);
+extern void ip_set_delip_kernel(ip_set_id_t id,
+                               const struct sk_buff *skb,
+                               const u_int32_t *flags);
+extern int ip_set_testip_kernel(ip_set_id_t id,
+                               const struct sk_buff *skb,
+                               const u_int32_t *flags);
+
+#endif                         /* __KERNEL__ */
+
+#endif /*_IP_SET_H*/
diff --git a/include/linux/netfilter_ipv4/ip_set_iphash.h b/include/linux/netfilter_ipv4/ip_set_iphash.h
new file mode 100644 (file)
index 0000000..65eb2be
--- /dev/null
@@ -0,0 +1,29 @@
+#ifndef __IP_SET_IPHASH_H
+#define __IP_SET_IPHASH_H
+
+#include <linux/netfilter_ipv4/ip_set.h>
+
+#define SETTYPE_NAME "iphash"
+#define MAX_RANGE 0x0000FFFF
+
+struct ip_set_iphash {
+       ip_set_ip_t *members;           /* the iphash proper */
+       uint32_t hashsize;              /* hash size */
+       uint16_t probes;                /* max number of probes  */
+       uint16_t resize;                /* resize factor in percent */
+       ip_set_ip_t netmask;            /* netmask */
+       void *initval[0];               /* initvals for jhash_1word */
+};
+
+struct ip_set_req_iphash_create {
+       uint32_t hashsize;
+       uint16_t probes;
+       uint16_t resize;
+       ip_set_ip_t netmask;
+};
+
+struct ip_set_req_iphash {
+       ip_set_ip_t ip;
+};
+
+#endif /* __IP_SET_IPHASH_H */
diff --git a/include/linux/netfilter_ipv4/ip_set_ipmap.h b/include/linux/netfilter_ipv4/ip_set_ipmap.h
new file mode 100644 (file)
index 0000000..2435102
--- /dev/null
@@ -0,0 +1,56 @@
+#ifndef __IP_SET_IPMAP_H
+#define __IP_SET_IPMAP_H
+
+#include <linux/netfilter_ipv4/ip_set.h>
+
+#define SETTYPE_NAME "ipmap"
+#define MAX_RANGE 0x0000FFFF
+
+struct ip_set_ipmap {
+       void *members;                  /* the ipmap proper */
+       ip_set_ip_t first_ip;           /* host byte order, included in range */
+       ip_set_ip_t last_ip;            /* host byte order, included in range */
+       ip_set_ip_t netmask;            /* subnet netmask */
+       ip_set_ip_t sizeid;             /* size of set in IPs */
+       ip_set_ip_t hosts;              /* number of hosts in a subnet */
+};
+
+struct ip_set_req_ipmap_create {
+       ip_set_ip_t from;
+       ip_set_ip_t to;
+       ip_set_ip_t netmask;
+};
+
+struct ip_set_req_ipmap {
+       ip_set_ip_t ip;
+};
+
+unsigned int
+mask_to_bits(ip_set_ip_t mask)
+{
+       unsigned int bits = 32;
+       ip_set_ip_t maskaddr;
+       
+       if (mask == 0xFFFFFFFF)
+               return bits;
+       
+       maskaddr = 0xFFFFFFFE;
+       while (--bits >= 0 && maskaddr != mask)
+               maskaddr <<= 1;
+       
+       return bits;
+}
+
+ip_set_ip_t
+range_to_mask(ip_set_ip_t from, ip_set_ip_t to, unsigned int *bits)
+{
+       ip_set_ip_t mask = 0xFFFFFFFE;
+       
+       *bits = 32;
+       while (--(*bits) >= 0 && mask && (to & mask) != from)
+               mask <<= 1;
+               
+       return mask;
+}
+       
+#endif /* __IP_SET_IPMAP_H */
diff --git a/include/linux/netfilter_ipv4/ip_set_ipporthash.h b/include/linux/netfilter_ipv4/ip_set_ipporthash.h
new file mode 100644 (file)
index 0000000..0e422f5
--- /dev/null
@@ -0,0 +1,33 @@
+#ifndef __IP_SET_IPPORTHASH_H
+#define __IP_SET_IPPORTHASH_H
+
+#include <linux/netfilter_ipv4/ip_set.h>
+
+#define SETTYPE_NAME "ipporthash"
+#define MAX_RANGE 0x0000FFFF
+#define INVALID_PORT   (MAX_RANGE + 1)
+
+struct ip_set_ipporthash {
+       ip_set_ip_t *members;           /* the ipporthash proper */
+       uint32_t hashsize;              /* hash size */
+       uint16_t probes;                /* max number of probes  */
+       uint16_t resize;                /* resize factor in percent */
+       ip_set_ip_t first_ip;           /* host byte order, included in range */
+       ip_set_ip_t last_ip;            /* host byte order, included in range */
+       void *initval[0];               /* initvals for jhash_1word */
+};
+
+struct ip_set_req_ipporthash_create {
+       uint32_t hashsize;
+       uint16_t probes;
+       uint16_t resize;
+       ip_set_ip_t from;
+       ip_set_ip_t to;
+};
+
+struct ip_set_req_ipporthash {
+       ip_set_ip_t ip;
+       ip_set_ip_t port;
+};
+
+#endif /* __IP_SET_IPPORTHASH_H */
diff --git a/include/linux/netfilter_ipv4/ip_set_iptree.h b/include/linux/netfilter_ipv4/ip_set_iptree.h
new file mode 100644 (file)
index 0000000..54142a7
--- /dev/null
@@ -0,0 +1,39 @@
+#ifndef __IP_SET_IPTREE_H
+#define __IP_SET_IPTREE_H
+
+#include <linux/netfilter_ipv4/ip_set.h>
+
+#define SETTYPE_NAME "iptree"
+#define MAX_RANGE 0x0000FFFF
+
+struct ip_set_iptreed {
+       unsigned long expires[256];             /* x.x.x.ADDR */
+};
+
+struct ip_set_iptreec {
+       struct ip_set_iptreed *tree[256];       /* x.x.ADDR.* */
+};
+
+struct ip_set_iptreeb {
+       struct ip_set_iptreec *tree[256];       /* x.ADDR.*.* */
+};
+
+struct ip_set_iptree {
+       unsigned int timeout;
+       unsigned int gc_interval;
+#ifdef __KERNEL__
+       struct timer_list gc;
+       struct ip_set_iptreeb *tree[256];       /* ADDR.*.*.* */
+#endif
+};
+
+struct ip_set_req_iptree_create {
+       unsigned int timeout;
+};
+
+struct ip_set_req_iptree {
+       ip_set_ip_t ip;
+       unsigned int timeout;
+};
+
+#endif /* __IP_SET_IPTREE_H */
diff --git a/include/linux/netfilter_ipv4/ip_set_jhash.h b/include/linux/netfilter_ipv4/ip_set_jhash.h
new file mode 100644 (file)
index 0000000..25c6b97
--- /dev/null
@@ -0,0 +1,148 @@
+#ifndef _LINUX_IPSET_JHASH_H
+#define _LINUX_IPSET_JHASH_H
+
+/* This is a copy of linux/jhash.h but the types u32/u8 are changed
+ * to __u32/__u8 so that the header file can be included into
+ * userspace code as well. Jozsef Kadlecsik (kadlec@blackhole.kfki.hu)
+ */
+
+/* jhash.h: Jenkins hash support.
+ *
+ * Copyright (C) 1996 Bob Jenkins (bob_jenkins@burtleburtle.net)
+ *
+ * http://burtleburtle.net/bob/hash/
+ *
+ * These are the credits from Bob's sources:
+ *
+ * lookup2.c, by Bob Jenkins, December 1996, Public Domain.
+ * hash(), hash2(), hash3, and mix() are externally useful functions.
+ * Routines to test the hash are included if SELF_TEST is defined.
+ * You can use this free for any purpose.  It has no warranty.
+ *
+ * Copyright (C) 2003 David S. Miller (davem@redhat.com)
+ *
+ * I've modified Bob's hash to be useful in the Linux kernel, and
+ * any bugs present are surely my fault.  -DaveM
+ */
+
+/* NOTE: Arguments are modified. */
+#define __jhash_mix(a, b, c) \
+{ \
+  a -= b; a -= c; a ^= (c>>13); \
+  b -= c; b -= a; b ^= (a<<8); \
+  c -= a; c -= b; c ^= (b>>13); \
+  a -= b; a -= c; a ^= (c>>12);  \
+  b -= c; b -= a; b ^= (a<<16); \
+  c -= a; c -= b; c ^= (b>>5); \
+  a -= b; a -= c; a ^= (c>>3);  \
+  b -= c; b -= a; b ^= (a<<10); \
+  c -= a; c -= b; c ^= (b>>15); \
+}
+
+/* The golden ration: an arbitrary value */
+#define JHASH_GOLDEN_RATIO     0x9e3779b9
+
+/* The most generic version, hashes an arbitrary sequence
+ * of bytes.  No alignment or length assumptions are made about
+ * the input key.
+ */
+static inline __u32 jhash(void *key, __u32 length, __u32 initval)
+{
+       __u32 a, b, c, len;
+       __u8 *k = key;
+
+       len = length;
+       a = b = JHASH_GOLDEN_RATIO;
+       c = initval;
+
+       while (len >= 12) {
+               a += (k[0] +((__u32)k[1]<<8) +((__u32)k[2]<<16) +((__u32)k[3]<<24));
+               b += (k[4] +((__u32)k[5]<<8) +((__u32)k[6]<<16) +((__u32)k[7]<<24));
+               c += (k[8] +((__u32)k[9]<<8) +((__u32)k[10]<<16)+((__u32)k[11]<<24));
+
+               __jhash_mix(a,b,c);
+
+               k += 12;
+               len -= 12;
+       }
+
+       c += length;
+       switch (len) {
+       case 11: c += ((__u32)k[10]<<24);
+       case 10: c += ((__u32)k[9]<<16);
+       case 9 : c += ((__u32)k[8]<<8);
+       case 8 : b += ((__u32)k[7]<<24);
+       case 7 : b += ((__u32)k[6]<<16);
+       case 6 : b += ((__u32)k[5]<<8);
+       case 5 : b += k[4];
+       case 4 : a += ((__u32)k[3]<<24);
+       case 3 : a += ((__u32)k[2]<<16);
+       case 2 : a += ((__u32)k[1]<<8);
+       case 1 : a += k[0];
+       };
+
+       __jhash_mix(a,b,c);
+
+       return c;
+}
+
+/* A special optimized version that handles 1 or more of __u32s.
+ * The length parameter here is the number of __u32s in the key.
+ */
+static inline __u32 jhash2(__u32 *k, __u32 length, __u32 initval)
+{
+       __u32 a, b, c, len;
+
+       a = b = JHASH_GOLDEN_RATIO;
+       c = initval;
+       len = length;
+
+       while (len >= 3) {
+               a += k[0];
+               b += k[1];
+               c += k[2];
+               __jhash_mix(a, b, c);
+               k += 3; len -= 3;
+       }
+
+       c += length * 4;
+
+       switch (len) {
+       case 2 : b += k[1];
+       case 1 : a += k[0];
+       };
+
+       __jhash_mix(a,b,c);
+
+       return c;
+}
+
+
+/* A special ultra-optimized versions that knows they are hashing exactly
+ * 3, 2 or 1 word(s).
+ *
+ * NOTE: In partilar the "c += length; __jhash_mix(a,b,c);" normally
+ *       done at the end is not done here.
+ */
+static inline __u32 jhash_3words(__u32 a, __u32 b, __u32 c, __u32 initval)
+{
+       a += JHASH_GOLDEN_RATIO;
+       b += JHASH_GOLDEN_RATIO;
+       c += initval;
+
+       __jhash_mix(a, b, c);
+
+       return c;
+}
+
+static inline __u32 jhash_2words(__u32 a, __u32 b, __u32 initval)
+{
+       return jhash_3words(a, b, 0, initval);
+}
+
+static inline __u32 jhash_1word(__u32 a, __u32 initval)
+{
+       return jhash_3words(a, 0, 0, initval);
+}
+
+#endif /* _LINUX_IPSET_JHASH_H */
diff --git a/include/linux/netfilter_ipv4/ip_set_macipmap.h b/include/linux/netfilter_ipv4/ip_set_macipmap.h
new file mode 100644 (file)
index 0000000..ee34c9b
--- /dev/null
@@ -0,0 +1,38 @@
+#ifndef __IP_SET_MACIPMAP_H
+#define __IP_SET_MACIPMAP_H
+
+#include <linux/netfilter_ipv4/ip_set.h>
+
+#define SETTYPE_NAME "macipmap"
+#define MAX_RANGE 0x0000FFFF
+
+/* general flags */
+#define IPSET_MACIP_MATCHUNSET 1
+
+/* per ip flags */
+#define IPSET_MACIP_ISSET      1
+
+struct ip_set_macipmap {
+       void *members;                  /* the macipmap proper */
+       ip_set_ip_t first_ip;           /* host byte order, included in range */
+       ip_set_ip_t last_ip;            /* host byte order, included in range */
+       u_int32_t flags;
+};
+
+struct ip_set_req_macipmap_create {
+       ip_set_ip_t from;
+       ip_set_ip_t to;
+       u_int32_t flags;
+};
+
+struct ip_set_req_macipmap {
+       ip_set_ip_t ip;
+       unsigned char ethernet[ETH_ALEN];
+};
+
+struct ip_set_macip {
+       unsigned short flags;
+       unsigned char ethernet[ETH_ALEN];
+};
+
+#endif /* __IP_SET_MACIPMAP_H */
diff --git a/include/linux/netfilter_ipv4/ip_set_malloc.h b/include/linux/netfilter_ipv4/ip_set_malloc.h
new file mode 100644 (file)
index 0000000..ab97e14
--- /dev/null
@@ -0,0 +1,116 @@
+#ifndef _IP_SET_MALLOC_H
+#define _IP_SET_MALLOC_H
+
+#ifdef __KERNEL__
+
+/* Memory allocation and deallocation */
+static size_t max_malloc_size = 0;
+
+static inline void init_max_malloc_size(void)
+{
+#define CACHE(x) max_malloc_size = x;
+#include <linux/kmalloc_sizes.h>
+#undef CACHE
+}
+
+static inline void * ip_set_malloc(size_t bytes)
+{
+       if (bytes > max_malloc_size)
+               return vmalloc(bytes);
+       else
+               return kmalloc(bytes, GFP_KERNEL);
+}
+
+static inline void ip_set_free(void * data, size_t bytes)
+{
+       if (bytes > max_malloc_size)
+               vfree(data);
+       else
+               kfree(data);
+}
+
+struct harray {
+       size_t max_elements;
+       void *arrays[0];
+};
+
+static inline void * 
+harray_malloc(size_t hashsize, size_t typesize, int flags)
+{
+       struct harray *harray;
+       size_t max_elements, size, i, j;
+
+       if (!max_malloc_size)
+               init_max_malloc_size();
+
+       if (typesize > max_malloc_size)
+               return NULL;
+
+       max_elements = max_malloc_size/typesize;
+       size = hashsize/max_elements;
+       if (hashsize % max_elements)
+               size++;
+       
+       /* Last pointer signals end of arrays */
+       harray = kmalloc(sizeof(struct harray) + (size + 1) * sizeof(void *),
+                        flags);
+
+       if (!harray)
+               return NULL;
+       
+       for (i = 0; i < size - 1; i++) {
+               harray->arrays[i] = kmalloc(max_elements * typesize, flags);
+               if (!harray->arrays[i])
+                       goto undo;
+               memset(harray->arrays[i], 0, max_elements * typesize);
+       }
+       harray->arrays[i] = kmalloc((hashsize - i * max_elements) * typesize, 
+                                   flags);
+       if (!harray->arrays[i])
+               goto undo;
+       memset(harray->arrays[i], 0, (hashsize - i * max_elements) * typesize);
+
+       harray->max_elements = max_elements;
+       harray->arrays[size] = NULL;
+       
+       return (void *)harray;
+
+    undo:
+       for (j = 0; j < i; j++) {
+               kfree(harray->arrays[j]);
+       }
+       kfree(harray);
+       return NULL;
+}
+
+static inline void harray_free(void *h)
+{
+       struct harray *harray = (struct harray *) h;
+       size_t i;
+       
+       for (i = 0; harray->arrays[i] != NULL; i++)
+               kfree(harray->arrays[i]);
+       kfree(harray);
+}
+
+static inline void harray_flush(void *h, size_t hashsize, size_t typesize)
+{
+       struct harray *harray = (struct harray *) h;
+       size_t i;
+       
+       for (i = 0; harray->arrays[i+1] != NULL; i++)
+               memset(harray->arrays[i], 0, harray->max_elements * typesize);
+       memset(harray->arrays[i], 0, 
+              (hashsize - i * harray->max_elements) * typesize);
+}
+
+#define HARRAY_ELEM(h, type, which)                            \
+({                                                             \
+       struct harray *__h = (struct harray *)(h);              \
+       ((type)((__h)->arrays[(which)/(__h)->max_elements])     \
+               + (which)%(__h)->max_elements);                 \
+})
+
+#endif                         /* __KERNEL__ */
+
+#endif /*_IP_SET_MALLOC_H*/
diff --git a/include/linux/netfilter_ipv4/ip_set_nethash.h b/include/linux/netfilter_ipv4/ip_set_nethash.h
new file mode 100644 (file)
index 0000000..6823641
--- /dev/null
@@ -0,0 +1,54 @@
+#ifndef __IP_SET_NETHASH_H
+#define __IP_SET_NETHASH_H
+
+#include <linux/netfilter_ipv4/ip_set.h>
+
+#define SETTYPE_NAME "nethash"
+#define MAX_RANGE 0x0000FFFF
+
+struct ip_set_nethash {
+       ip_set_ip_t *members;           /* the nethash proper */
+       uint32_t hashsize;              /* hash size */
+       uint16_t probes;                /* max number of probes  */
+       uint16_t resize;                /* resize factor in percent */
+       unsigned char cidr[30];         /* CIDR sizes */
+       void *initval[0];               /* initvals for jhash_1word */
+};
+
+struct ip_set_req_nethash_create {
+       uint32_t hashsize;
+       uint16_t probes;
+       uint16_t resize;
+};
+
+struct ip_set_req_nethash {
+       ip_set_ip_t ip;
+       unsigned char cidr;
+};
+
+static unsigned char shifts[] = {255, 253, 249, 241, 225, 193, 129, 1};
+
+static inline ip_set_ip_t 
+pack(ip_set_ip_t ip, unsigned char cidr)
+{
+       ip_set_ip_t addr, *paddr = &addr;
+       unsigned char n, t, *a;
+
+       addr = htonl(ip & (0xFFFFFFFF << (32 - (cidr))));
+#ifdef __KERNEL__
+       DP("ip:%u.%u.%u.%u/%u", NIPQUAD(addr), cidr);
+#endif
+       n = cidr / 8;
+       t = cidr % 8;   
+       a = &((unsigned char *)paddr)[n];
+       *a = *a /(1 << (8 - t)) + shifts[t];
+#ifdef __KERNEL__
+       DP("n: %u, t: %u, a: %u", n, t, *a);
+       DP("ip:%u.%u.%u.%u/%u, %u.%u.%u.%u",
+          HIPQUAD(ip), cidr, NIPQUAD(addr));
+#endif
+
+       return ntohl(addr);
+}
+
+#endif /* __IP_SET_NETHASH_H */
diff --git a/include/linux/netfilter_ipv4/ip_set_portmap.h b/include/linux/netfilter_ipv4/ip_set_portmap.h
new file mode 100644 (file)
index 0000000..c17165c
--- /dev/null
@@ -0,0 +1,25 @@
+#ifndef __IP_SET_PORTMAP_H
+#define __IP_SET_PORTMAP_H
+
+#include <linux/netfilter_ipv4/ip_set.h>
+
+#define SETTYPE_NAME   "portmap"
+#define MAX_RANGE      0x0000FFFF
+#define INVALID_PORT   (MAX_RANGE + 1)
+
+struct ip_set_portmap {
+       void *members;                  /* the portmap proper */
+       ip_set_ip_t first_port;         /* host byte order, included in range */
+       ip_set_ip_t last_port;          /* host byte order, included in range */
+};
+
+struct ip_set_req_portmap_create {
+       ip_set_ip_t from;
+       ip_set_ip_t to;
+};
+
+struct ip_set_req_portmap {
+       ip_set_ip_t port;
+};
+
+#endif /* __IP_SET_PORTMAP_H */
diff --git a/include/linux/netfilter_ipv4/ipt_set.h b/include/linux/netfilter_ipv4/ipt_set.h
new file mode 100644 (file)
index 0000000..2a18b93
--- /dev/null
@@ -0,0 +1,21 @@
+#ifndef _IPT_SET_H
+#define _IPT_SET_H
+
+#include <linux/netfilter_ipv4/ip_set.h>
+
+struct ipt_set_info {
+       ip_set_id_t index;
+       u_int32_t flags[IP_SET_MAX_BINDINGS + 1];
+};
+
+/* match info */
+struct ipt_set_info_match {
+       struct ipt_set_info match_set;
+};
+
+struct ipt_set_info_target {
+       struct ipt_set_info add_set;
+       struct ipt_set_info del_set;
+};
+
+#endif /*_IPT_SET_H*/
diff --git a/include/linux/vserver/limit_int.h b/include/linux/vserver/limit_int.h
new file mode 100644 (file)
index 0000000..a48d242
--- /dev/null
@@ -0,0 +1,76 @@
+#ifndef _VX_LIMIT_INT_H
+#define _VX_LIMIT_INT_H
+
+
+#ifdef __KERNEL__
+
+#define VXD_RCRES(r)   VXD_CBIT(cres, (r))
+#define VXD_RLIMIT(r)  VXD_CBIT(limit, (r))
+
+extern const char *vlimit_name[NUM_LIMITS];
+
+static inline void __vx_acc_cres(struct vx_info *vxi,
+       int res, int dir, void *_data, char *_file, int _line)
+{
+       if (VXD_RCRES(res))
+               vxlprintk(1, "vx_acc_cres[%5d,%s,%2d]: %5d%s (%p)",
+                       (vxi ? vxi->vx_id : -1), vlimit_name[res], res,
+                       (vxi ? atomic_read(&vxi->limit.rcur[res]) : 0),
+                       (dir > 0) ? "++" : "--", _data, _file, _line);
+       if (!vxi)
+               return;
+
+       if (dir > 0)
+               atomic_inc(&vxi->limit.rcur[res]);
+       else
+               atomic_dec(&vxi->limit.rcur[res]);
+}
+
+static inline void __vx_add_cres(struct vx_info *vxi,
+       int res, int amount, void *_data, char *_file, int _line)
+{
+       if (VXD_RCRES(res))
+               vxlprintk(1, "vx_add_cres[%5d,%s,%2d]: %5d += %5d (%p)",
+                       (vxi ? vxi->vx_id : -1), vlimit_name[res], res,
+                       (vxi ? atomic_read(&vxi->limit.rcur[res]) : 0),
+                       amount, _data, _file, _line);
+       if (amount == 0)
+               return;
+       if (!vxi)
+               return;
+       atomic_add(amount, &vxi->limit.rcur[res]);
+}
+
+static inline int __vx_cres_avail(struct vx_info *vxi,
+               int res, int num, char *_file, int _line)
+{
+       unsigned long value;
+
+       if (VXD_RLIMIT(res))
+               vxlprintk(1, "vx_cres_avail[%5d,%s,%2d]: %5ld > %5d + %5d",
+                       (vxi ? vxi->vx_id : -1), vlimit_name[res], res,
+                       (vxi ? vxi->limit.rlim[res] : 1),
+                       (vxi ? atomic_read(&vxi->limit.rcur[res]) : 0),
+                       num, _file, _line);
+       if (num == 0)
+               return 1;
+       if (!vxi)
+               return 1;
+
+       value = atomic_read(&vxi->limit.rcur[res]);
+
+       if (value > vxi->limit.rmax[res])
+               vxi->limit.rmax[res] = value;
+
+       if (vxi->limit.rlim[res] == RLIM_INFINITY)
+               return 1;
+
+       if (value + num <= vxi->limit.rlim[res])
+               return 1;
+
+       atomic_inc(&vxi->limit.lhit[res]);
+       return 0;
+}
+
+#endif /* __KERNEL__ */
+#endif /* _VX_LIMIT_H */
diff --git a/net/ipv4/netfilter/ip_set.c b/net/ipv4/netfilter/ip_set.c
new file mode 100644 (file)
index 0000000..ff30f7c
--- /dev/null
@@ -0,0 +1,1992 @@
+/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
+ *                         Patrick Schaaf <bof@bof.de>
+ * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module for IP set management */
+
+#include <linux/config.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/kmod.h>
+#include <linux/ip.h>
+#include <linux/skbuff.h>
+#include <linux/random.h>
+#include <linux/jhash.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/errno.h>
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <asm/semaphore.h>
+#include <linux/spinlock.h>
+#include <linux/vmalloc.h>
+
+#define ASSERT_READ_LOCK(x)    /* dont use that */
+#define ASSERT_WRITE_LOCK(x)
+#include <linux/netfilter_ipv4/listhelp.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+
+static struct list_head set_type_list;         /* all registered sets */
+static struct ip_set **ip_set_list;            /* all individual sets */
+static DEFINE_RWLOCK(ip_set_lock);             /* protects the lists and the hash */
+static DECLARE_MUTEX(ip_set_app_mutex);                /* serializes user access */
+static ip_set_id_t ip_set_max = CONFIG_IP_NF_SET_MAX;
+static ip_set_id_t ip_set_bindings_hash_size =  CONFIG_IP_NF_SET_HASHSIZE;
+static struct list_head *ip_set_hash;          /* hash of bindings */
+static unsigned int ip_set_hash_random;                /* random seed */
+
+/*
+ * Sets are identified either by the index in ip_set_list or by id.
+ * The id never changes and is used to find a key in the hash. 
+ * The index may change by swapping and used at all other places 
+ * (set/SET netfilter modules, binding value, etc.)
+ *
+ * Userspace requests are serialized by ip_set_mutex and sets can
+ * be deleted only from userspace. Therefore ip_set_list locking 
+ * must obey the following rules:
+ *
+ * - kernel requests: read and write locking mandatory
+ * - user requests: read locking optional, write locking mandatory
+ */
+
+static inline void
+__ip_set_get(ip_set_id_t index)
+{
+       atomic_inc(&ip_set_list[index]->ref);
+}
+
+static inline void
+__ip_set_put(ip_set_id_t index)
+{
+       atomic_dec(&ip_set_list[index]->ref);
+}
+
+/*
+ * Binding routines
+ */
+
+static inline int
+ip_hash_cmp(const struct ip_set_hash *set_hash,
+           ip_set_id_t id, ip_set_ip_t ip)
+{
+       return set_hash->id == id && set_hash->ip == ip;
+}
+
+static ip_set_id_t
+ip_set_find_in_hash(ip_set_id_t id, ip_set_ip_t ip)
+{
+       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random) 
+                               % ip_set_bindings_hash_size;
+       struct ip_set_hash *set_hash;
+
+       ASSERT_READ_LOCK(&ip_set_lock);
+       IP_SET_ASSERT(ip_set_list[id]);
+       DP("set: %s, ip: %u.%u.%u.%u", ip_set_list[id]->name, HIPQUAD(ip));     
+       
+       set_hash = LIST_FIND(&ip_set_hash[key], ip_hash_cmp,
+                            struct ip_set_hash *, id, ip);
+       
+       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name, 
+          HIPQUAD(ip),
+          set_hash != NULL ? ip_set_list[set_hash->binding]->name : "");
+
+       return (set_hash != NULL ? set_hash->binding : IP_SET_INVALID_ID);
+}
+
+static inline void 
+__set_hash_del(struct ip_set_hash *set_hash)
+{
+       ASSERT_WRITE_LOCK(&ip_set_lock);
+       IP_SET_ASSERT(ip_set_list[set_hash->binding]);  
+
+       __ip_set_put(set_hash->binding);
+       list_del(&set_hash->list);
+       kfree(set_hash);
+}
+
+static int
+ip_set_hash_del(ip_set_id_t id, ip_set_ip_t ip)
+{
+       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random)
+                               % ip_set_bindings_hash_size;
+       struct ip_set_hash *set_hash;
+       
+       IP_SET_ASSERT(ip_set_list[id]);
+       DP("set: %s, ip: %u.%u.%u.%u", ip_set_list[id]->name, HIPQUAD(ip));     
+       write_lock_bh(&ip_set_lock);
+       set_hash = LIST_FIND(&ip_set_hash[key], ip_hash_cmp,
+                            struct ip_set_hash *, id, ip);
+       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name,
+          HIPQUAD(ip),
+          set_hash != NULL ? ip_set_list[set_hash->binding]->name : "");
+
+       if (set_hash != NULL)
+               __set_hash_del(set_hash);
+       write_unlock_bh(&ip_set_lock);
+       return 0;
+}
+
+static int 
+ip_set_hash_add(ip_set_id_t id, ip_set_ip_t ip, ip_set_id_t binding)
+{
+       u_int32_t key = jhash_2words(id, ip, ip_set_hash_random)
+                               % ip_set_bindings_hash_size;
+       struct ip_set_hash *set_hash;
+       int ret = 0;
+       
+       IP_SET_ASSERT(ip_set_list[id]);
+       IP_SET_ASSERT(ip_set_list[binding]);
+       DP("set: %s, ip: %u.%u.%u.%u, binding: %s", ip_set_list[id]->name, 
+          HIPQUAD(ip), ip_set_list[binding]->name);
+       write_lock_bh(&ip_set_lock);
+       set_hash = LIST_FIND(&ip_set_hash[key], ip_hash_cmp,
+                            struct ip_set_hash *, id, ip);
+       if (!set_hash) {
+               set_hash = kmalloc(sizeof(struct ip_set_hash), GFP_KERNEL);
+               if (!set_hash) {
+                       ret = -ENOMEM;
+                       goto unlock;
+               }
+               INIT_LIST_HEAD(&set_hash->list);
+               set_hash->id = id;
+               set_hash->ip = ip;
+               list_add(&ip_set_hash[key], &set_hash->list);
+       } else {
+               IP_SET_ASSERT(ip_set_list[set_hash->binding]);  
+               DP("overwrite binding: %s",
+                  ip_set_list[set_hash->binding]->name);
+               __ip_set_put(set_hash->binding);
+       }
+       set_hash->binding = binding;
+       __ip_set_get(set_hash->binding);
+    unlock:
+       write_unlock_bh(&ip_set_lock);
+       return ret;
+}
+
+#define FOREACH_HASH_DO(fn, args...)                                           \
+({                                                                             \
+       ip_set_id_t __key;                                                      \
+       struct ip_set_hash *__set_hash;                                         \
+                                                                               \
+       for (__key = 0; __key < ip_set_bindings_hash_size; __key++) {           \
+               list_for_each_entry(__set_hash, &ip_set_hash[__key], list)      \
+                       fn(__set_hash , ## args);                               \
+       }                                                                       \
+})
+
+#define FOREACH_HASH_RW_DO(fn, args...)                                                \
+({                                                                             \
+       ip_set_id_t __key;                                                      \
+       struct ip_set_hash *__set_hash, *__n;                                   \
+                                                                               \
+       ASSERT_WRITE_LOCK(&ip_set_lock);                                        \
+       for (__key = 0; __key < ip_set_bindings_hash_size; __key++) {           \
+               list_for_each_entry_safe(__set_hash, __n, &ip_set_hash[__key], list)\
+                       fn(__set_hash , ## args);                               \
+       }                                                                       \
+})
+
+/* Add, del and test set entries from kernel */
+
+#define follow_bindings(index, set, ip)                                        \
+((index = ip_set_find_in_hash((set)->id, ip)) != IP_SET_INVALID_ID     \
+ || (index = (set)->binding) != IP_SET_INVALID_ID)
+
+int
+ip_set_testip_kernel(ip_set_id_t index,
+                    const struct sk_buff *skb,
+                    const u_int32_t *flags)
+{
+       struct ip_set *set;
+       ip_set_ip_t ip;
+       int res;
+       unsigned char i = 0;
+       
+       IP_SET_ASSERT(flags[i]);
+       read_lock_bh(&ip_set_lock);
+       do {
+               set = ip_set_list[index];
+               IP_SET_ASSERT(set);
+               DP("set %s, index %u", set->name, index);
+               read_lock_bh(&set->lock);
+               res = set->type->testip_kernel(set, skb, &ip, flags, i++);
+               read_unlock_bh(&set->lock);
+               i += !!(set->type->features & IPSET_DATA_DOUBLE);
+       } while (res > 0 
+                && flags[i] 
+                && follow_bindings(index, set, ip));
+       read_unlock_bh(&ip_set_lock);
+
+       return res;
+}
+
+void
+ip_set_addip_kernel(ip_set_id_t index,
+                   const struct sk_buff *skb,
+                   const u_int32_t *flags)
+{
+       struct ip_set *set;
+       ip_set_ip_t ip;
+       int res;
+       unsigned char i = 0;
+
+       IP_SET_ASSERT(flags[i]);
+   retry:
+       read_lock_bh(&ip_set_lock);
+       do {
+               set = ip_set_list[index];
+               IP_SET_ASSERT(set);
+               DP("set %s, index %u", set->name, index);
+               write_lock_bh(&set->lock);
+               res = set->type->addip_kernel(set, skb, &ip, flags, i++);
+               write_unlock_bh(&set->lock);
+               i += !!(set->type->features & IPSET_DATA_DOUBLE);
+       } while ((res == 0 || res == -EEXIST)
+                && flags[i] 
+                && follow_bindings(index, set, ip));
+       read_unlock_bh(&ip_set_lock);
+
+       if (res == -EAGAIN
+           && set->type->retry
+           && (res = set->type->retry(set)) == 0)
+               goto retry;
+}
+
+void
+ip_set_delip_kernel(ip_set_id_t index,
+                   const struct sk_buff *skb,
+                   const u_int32_t *flags)
+{
+       struct ip_set *set;
+       ip_set_ip_t ip;
+       int res;
+       unsigned char i = 0;
+
+       IP_SET_ASSERT(flags[i]);
+       read_lock_bh(&ip_set_lock);
+       do {
+               set = ip_set_list[index];
+               IP_SET_ASSERT(set);
+               DP("set %s, index %u", set->name, index);
+               write_lock_bh(&set->lock);
+               res = set->type->delip_kernel(set, skb, &ip, flags, i++);
+               write_unlock_bh(&set->lock);
+               i += !!(set->type->features & IPSET_DATA_DOUBLE);
+       } while ((res == 0 || res == -EEXIST)
+                && flags[i] 
+                && follow_bindings(index, set, ip));
+       read_unlock_bh(&ip_set_lock);
+}
+
+/* Register and deregister settype */
+
+static inline int
+set_type_equal(const struct ip_set_type *set_type, const char *str2)
+{
+       return !strncmp(set_type->typename, str2, IP_SET_MAXNAMELEN - 1);
+}
+
+static inline struct ip_set_type *
+find_set_type(const char *name)
+{
+       return LIST_FIND(&set_type_list,
+                        set_type_equal,
+                        struct ip_set_type *,
+                        name);
+}
+
+int 
+ip_set_register_set_type(struct ip_set_type *set_type)
+{
+       int ret = 0;
+       
+       if (set_type->protocol_version != IP_SET_PROTOCOL_VERSION) {
+               ip_set_printk("'%s' uses wrong protocol version %u (want %u)",
+                             set_type->typename,
+                             set_type->protocol_version,
+                             IP_SET_PROTOCOL_VERSION);
+               return -EINVAL;
+       }
+
+       write_lock_bh(&ip_set_lock);
+       if (find_set_type(set_type->typename)) {
+               /* Duplicate! */
+               ip_set_printk("'%s' already registered!", 
+                             set_type->typename);
+               ret = -EINVAL;
+               goto unlock;
+       }
+       if (!try_module_get(THIS_MODULE)) {
+               ret = -EFAULT;
+               goto unlock;
+       }
+       list_append(&set_type_list, set_type);
+       DP("'%s' registered.", set_type->typename);
+   unlock:
+       write_unlock_bh(&ip_set_lock);
+       return ret;
+}
+
+void
+ip_set_unregister_set_type(struct ip_set_type *set_type)
+{
+       write_lock_bh(&ip_set_lock);
+       if (!find_set_type(set_type->typename)) {
+               ip_set_printk("'%s' not registered?",
+                             set_type->typename);
+               goto unlock;
+       }
+       LIST_DELETE(&set_type_list, set_type);
+       module_put(THIS_MODULE);
+       DP("'%s' unregistered.", set_type->typename);
+   unlock:
+       write_unlock_bh(&ip_set_lock);
+
+}
+
+/*
+ * Userspace routines
+ */
+
+/*
+ * Find set by name, reference it once. The reference makes sure the
+ * thing pointed to, does not go away under our feet. Drop the reference
+ * later, using ip_set_put().
+ */
+ip_set_id_t
+ip_set_get_byname(const char *name)
+{
+       ip_set_id_t i, index = IP_SET_INVALID_ID;
+       
+       down(&ip_set_app_mutex);
+       for (i = 0; i < ip_set_max; i++) {
+               if (ip_set_list[i] != NULL
+                   && strcmp(ip_set_list[i]->name, name) == 0) {
+                       __ip_set_get(i);
+                       index = i;
+                       break;
+               }
+       }
+       up(&ip_set_app_mutex);
+       return index;
+}
+
+/*
+ * Find set by index, reference it once. The reference makes sure the
+ * thing pointed to, does not go away under our feet. Drop the reference
+ * later, using ip_set_put().
+ */
+ip_set_id_t
+ip_set_get_byindex(ip_set_id_t index)
+{
+       down(&ip_set_app_mutex);
+
+       if (index >= ip_set_max)
+               return IP_SET_INVALID_ID;
+       
+       if (ip_set_list[index])
+               __ip_set_get(index);
+       else
+               index = IP_SET_INVALID_ID;
+               
+       up(&ip_set_app_mutex);
+       return index;
+}
+
+/*
+ * If the given set pointer points to a valid set, decrement
+ * reference count by 1. The caller shall not assume the index
+ * to be valid, after calling this function.
+ */
+void ip_set_put(ip_set_id_t index)
+{
+       down(&ip_set_app_mutex);
+       if (ip_set_list[index])
+               __ip_set_put(index);
+       up(&ip_set_app_mutex);
+}
+
+/* Find a set by name or index */
+static ip_set_id_t
+ip_set_find_byname(const char *name)
+{
+       ip_set_id_t i, index = IP_SET_INVALID_ID;
+       
+       for (i = 0; i < ip_set_max; i++) {
+               if (ip_set_list[i] != NULL
+                   && strcmp(ip_set_list[i]->name, name) == 0) {
+                       index = i;
+                       break;
+               }
+       }
+       return index;
+}
+
+static ip_set_id_t
+ip_set_find_byindex(ip_set_id_t index)
+{
+       if (index >= ip_set_max || ip_set_list[index] == NULL)
+               index = IP_SET_INVALID_ID;
+       
+       return index;
+}
+
+/*
+ * Add, del, test, bind and unbind
+ */
+
+static inline int
+__ip_set_testip(struct ip_set *set,
+               const void *data,
+               size_t size,
+               ip_set_ip_t *ip)
+{
+       int res;
+
+       read_lock_bh(&set->lock);
+       res = set->type->testip(set, data, size, ip);
+       read_unlock_bh(&set->lock);
+
+       return res;
+}
+
+static int
+__ip_set_addip(ip_set_id_t index,
+              const void *data,
+              size_t size)
+{
+       struct ip_set *set = ip_set_list[index];
+       ip_set_ip_t ip;
+       int res;
+       
+       IP_SET_ASSERT(set);
+       do {
+               write_lock_bh(&set->lock);
+               res = set->type->addip(set, data, size, &ip);
+               write_unlock_bh(&set->lock);
+       } while (res == -EAGAIN
+                && set->type->retry
+                && (res = set->type->retry(set)) == 0);
+
+       return res;
+}
+
+static int
+ip_set_addip(ip_set_id_t index,
+            const void *data,
+            size_t size)
+{
+
+       return __ip_set_addip(index,
+                             data + sizeof(struct ip_set_req_adt),
+                             size - sizeof(struct ip_set_req_adt));
+}
+
+static int
+ip_set_delip(ip_set_id_t index,
+            const void *data,
+            size_t size)
+{
+       struct ip_set *set = ip_set_list[index];
+       ip_set_ip_t ip;
+       int res;
+       
+       IP_SET_ASSERT(set);
+       write_lock_bh(&set->lock);
+       res = set->type->delip(set,
+                              data + sizeof(struct ip_set_req_adt),
+                              size - sizeof(struct ip_set_req_adt),
+                              &ip);
+       write_unlock_bh(&set->lock);
+
+       return res;
+}
+
+static int
+ip_set_testip(ip_set_id_t index,
+             const void *data,
+             size_t size)
+{
+       struct ip_set *set = ip_set_list[index];
+       ip_set_ip_t ip;
+       int res;
+
+       IP_SET_ASSERT(set);
+       res = __ip_set_testip(set,
+                             data + sizeof(struct ip_set_req_adt),
+                             size - sizeof(struct ip_set_req_adt),
+                             &ip);
+
+       return (res > 0 ? -EEXIST : res);
+}
+
+static int
+ip_set_bindip(ip_set_id_t index,
+             const void *data,
+             size_t size)
+{
+       struct ip_set *set = ip_set_list[index];
+       struct ip_set_req_bind *req_bind;
+       ip_set_id_t binding;
+       ip_set_ip_t ip;
+       int res;
+
+       IP_SET_ASSERT(set);
+       if (size < sizeof(struct ip_set_req_bind))
+               return -EINVAL;
+               
+       req_bind = (struct ip_set_req_bind *) data;
+       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
+
+       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
+               /* Default binding of a set */
+               char *binding_name;
+               
+               if (size != sizeof(struct ip_set_req_bind) + IP_SET_MAXNAMELEN)
+                       return -EINVAL;
+
+               binding_name = (char *)(data + sizeof(struct ip_set_req_bind)); 
+               binding_name[IP_SET_MAXNAMELEN - 1] = '\0';
+
+               binding = ip_set_find_byname(binding_name);
+               if (binding == IP_SET_INVALID_ID)
+                       return -ENOENT;
+
+               write_lock_bh(&ip_set_lock);
+               /* Sets as binding values are referenced */
+               if (set->binding != IP_SET_INVALID_ID)
+                       __ip_set_put(set->binding);
+               set->binding = binding;
+               __ip_set_get(set->binding);
+               write_unlock_bh(&ip_set_lock);
+
+               return 0;
+       }
+       binding = ip_set_find_byname(req_bind->binding);
+       if (binding == IP_SET_INVALID_ID)
+               return -ENOENT;
+
+       res = __ip_set_testip(set,
+                             data + sizeof(struct ip_set_req_bind),
+                             size - sizeof(struct ip_set_req_bind),
+                             &ip);
+       DP("set %s, ip: %u.%u.%u.%u, binding %s",
+          set->name, HIPQUAD(ip), ip_set_list[binding]->name);
+       
+       if (res >= 0)
+               res = ip_set_hash_add(set->id, ip, binding);
+
+       return res;
+}
+
+#define FOREACH_SET_DO(fn, args...)                            \
+({                                                             \
+       ip_set_id_t __i;                                        \
+       struct ip_set *__set;                                   \
+                                                               \
+       for (__i = 0; __i < ip_set_max; __i++) {                \
+               __set = ip_set_list[__i];                       \
+               if (__set != NULL)                              \
+                       fn(__set , ##args);                     \
+       }                                                       \
+})
+
+static inline void
+__set_hash_del_byid(struct ip_set_hash *set_hash, ip_set_id_t id)
+{
+       if (set_hash->id == id)
+               __set_hash_del(set_hash);
+}
+
+static inline void
+__unbind_default(struct ip_set *set)
+{
+       if (set->binding != IP_SET_INVALID_ID) {
+               /* Sets as binding values are referenced */
+               __ip_set_put(set->binding);
+               set->binding = IP_SET_INVALID_ID;
+       }
+}
+
+static int
+ip_set_unbindip(ip_set_id_t index,
+               const void *data,
+               size_t size)
+{
+       struct ip_set *set;
+       struct ip_set_req_bind *req_bind;
+       ip_set_ip_t ip;
+       int res;
+
+       DP("");
+       if (size < sizeof(struct ip_set_req_bind))
+               return -EINVAL;
+               
+       req_bind = (struct ip_set_req_bind *) data;
+       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
+       
+       DP("%u %s", index, req_bind->binding);
+       if (index == IP_SET_INVALID_ID) {
+               /* unbind :all: */
+               if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
+                       /* Default binding of sets */
+                       write_lock_bh(&ip_set_lock);
+                       FOREACH_SET_DO(__unbind_default);
+                       write_unlock_bh(&ip_set_lock);
+                       return 0;
+               } else if (strcmp(req_bind->binding, IPSET_TOKEN_ALL) == 0) {
+                       /* Flush all bindings of all sets*/
+                       write_lock_bh(&ip_set_lock);
+                       FOREACH_HASH_RW_DO(__set_hash_del);
+                       write_unlock_bh(&ip_set_lock);
+                       return 0;
+               }
+               DP("unreachable reached!");
+               return -EINVAL;
+       }
+       
+       set = ip_set_list[index];
+       IP_SET_ASSERT(set);
+       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
+               /* Default binding of set */
+               ip_set_id_t binding = ip_set_find_byindex(set->binding);
+
+               if (binding == IP_SET_INVALID_ID)
+                       return -ENOENT;
+                       
+               write_lock_bh(&ip_set_lock);
+               /* Sets in hash values are referenced */
+               __ip_set_put(set->binding);
+               set->binding = IP_SET_INVALID_ID;
+               write_unlock_bh(&ip_set_lock);
+
+               return 0;
+       } else if (strcmp(req_bind->binding, IPSET_TOKEN_ALL) == 0) {
+               /* Flush all bindings */
+
+               write_lock_bh(&ip_set_lock);
+               FOREACH_HASH_RW_DO(__set_hash_del_byid, set->id);
+               write_unlock_bh(&ip_set_lock);
+               return 0;
+       }
+       
+       res = __ip_set_testip(set,
+                             data + sizeof(struct ip_set_req_bind),
+                             size - sizeof(struct ip_set_req_bind),
+                             &ip);
+
+       DP("set %s, ip: %u.%u.%u.%u", set->name, HIPQUAD(ip));
+       if (res >= 0)
+               res = ip_set_hash_del(set->id, ip);
+
+       return res;
+}
+
+static int
+ip_set_testbind(ip_set_id_t index,
+               const void *data,
+               size_t size)
+{
+       struct ip_set *set = ip_set_list[index];
+       struct ip_set_req_bind *req_bind;
+       ip_set_id_t binding;
+       ip_set_ip_t ip;
+       int res;
+
+       IP_SET_ASSERT(set);
+       if (size < sizeof(struct ip_set_req_bind))
+               return -EINVAL;
+               
+       req_bind = (struct ip_set_req_bind *) data;
+       req_bind->binding[IP_SET_MAXNAMELEN - 1] = '\0';
+
+       if (strcmp(req_bind->binding, IPSET_TOKEN_DEFAULT) == 0) {
+               /* Default binding of set */
+               char *binding_name;
+               
+               if (size != sizeof(struct ip_set_req_bind) + IP_SET_MAXNAMELEN)
+                       return -EINVAL;
+
+               binding_name = (char *)(data + sizeof(struct ip_set_req_bind)); 
+               binding_name[IP_SET_MAXNAMELEN - 1] = '\0';
+
+               binding = ip_set_find_byname(binding_name);
+               if (binding == IP_SET_INVALID_ID)
+                       return -ENOENT;
+               
+               res = (set->binding == binding) ? -EEXIST : 0;
+
+               return res;
+       }
+       binding = ip_set_find_byname(req_bind->binding);
+       if (binding == IP_SET_INVALID_ID)
+               return -ENOENT;
+               
+       
+       res = __ip_set_testip(set,
+                             data + sizeof(struct ip_set_req_bind),
+                             size - sizeof(struct ip_set_req_bind),
+                             &ip);
+       DP("set %s, ip: %u.%u.%u.%u, binding %s",
+          set->name, HIPQUAD(ip), ip_set_list[binding]->name);
+          
+       if (res >= 0)
+               res = (ip_set_find_in_hash(set->id, ip) == binding)
+                       ? -EEXIST : 0;
+
+       return res;
+}
+
+static struct ip_set_type *
+find_set_type_rlock(const char *typename)
+{
+       struct ip_set_type *type;
+       
+       read_lock_bh(&ip_set_lock);
+       type = find_set_type(typename);
+       if (type == NULL)
+               read_unlock_bh(&ip_set_lock);
+
+       return type;
+}
+
+static int
+find_free_id(const char *name,
+            ip_set_id_t *index,
+            ip_set_id_t *id)
+{
+       ip_set_id_t i;
+
+       *id = IP_SET_INVALID_ID;
+       for (i = 0;  i < ip_set_max; i++) {
+               if (ip_set_list[i] == NULL) {
+                       if (*id == IP_SET_INVALID_ID)
+                               *id = *index = i;
+               } else if (strcmp(name, ip_set_list[i]->name) == 0)
+                       /* Name clash */
+                       return -EEXIST;
+       }
+       if (*id == IP_SET_INVALID_ID)
+               /* No free slot remained */
+               return -ERANGE;
+       /* Check that index is usable as id (swapping) */
+    check:     
+       for (i = 0;  i < ip_set_max; i++) {
+               if (ip_set_list[i] != NULL
+                   && ip_set_list[i]->id == *id) {
+                   *id = i;
+                   goto check;
+               }
+       }
+       return 0;
+}
+
+/*
+ * Create a set
+ */
+static int
+ip_set_create(const char *name,
+             const char *typename,
+             ip_set_id_t restore,
+             const void *data,
+             size_t size)
+{
+       struct ip_set *set;
+       ip_set_id_t index, id;
+       int res = 0;
+
+       DP("setname: %s, typename: %s, id: %u", name, typename, restore);
+       /*
+        * First, and without any locks, allocate and initialize
+        * a normal base set structure.
+        */
+       set = kmalloc(sizeof(struct ip_set), GFP_KERNEL);
+       if (!set)
+               return -ENOMEM;
+       set->lock = RW_LOCK_UNLOCKED;
+       strncpy(set->name, name, IP_SET_MAXNAMELEN);
+       set->binding = IP_SET_INVALID_ID;
+       atomic_set(&set->ref, 0);
+
+       /*
+        * Next, take the &ip_set_lock, check that we know the type,
+        * and take a reference on the type, to make sure it
+        * stays available while constructing our new set.
+        *
+        * After referencing the type, we drop the &ip_set_lock,
+        * and let the new set construction run without locks.
+        */
+       set->type = find_set_type_rlock(typename);
+       if (set->type == NULL) {
+               /* Try loading the module */
+               char modulename[IP_SET_MAXNAMELEN + strlen("ip_set_") + 1];
+               strcpy(modulename, "ip_set_");
+               strcat(modulename, typename);
+               DP("try to load %s", modulename);
+               request_module(modulename);
+               set->type = find_set_type_rlock(typename);
+       }
+       if (set->type == NULL) {
+               ip_set_printk("no set type '%s', set '%s' not created",
+                             typename, name);
+               res = -ENOENT;
+               goto out;
+       }
+       if (!try_module_get(set->type->me)) {
+               read_unlock_bh(&ip_set_lock);
+               res = -EFAULT;
+               goto out;
+       }
+       read_unlock_bh(&ip_set_lock);
+
+       /*
+        * Without holding any locks, create private part.
+        */
+       res = set->type->create(set, data, size);
+       if (res != 0)
+               goto put_out;
+
+       /* BTW, res==0 here. */
+
+       /*
+        * Here, we have a valid, constructed set. &ip_set_lock again,
+        * find free id/index and check that it is not already in 
+        * ip_set_list.
+        */
+       write_lock_bh(&ip_set_lock);
+       if ((res = find_free_id(set->name, &index, &id)) != 0) {
+               DP("no free id!");
+               goto cleanup;
+       }
+
+       /* Make sure restore gets the same index */
+       if (restore != IP_SET_INVALID_ID && index != restore) {
+               DP("Can't restore, sets are screwed up");
+               res = -ERANGE;
+               goto cleanup;
+       }
+        
+       /*
+        * Finally! Add our shiny new set to the list, and be done.
+        */
+       DP("create: '%s' created with index %u, id %u!", set->name, index, id);
+       set->id = id;
+       ip_set_list[index] = set;
+       write_unlock_bh(&ip_set_lock);
+       return res;
+       
+    cleanup:
+       write_unlock_bh(&ip_set_lock);
+       set->type->destroy(set);
+    put_out:
+       module_put(set->type->me);
+    out:
+       kfree(set);
+       return res;
+}
+
+/*
+ * Destroy a given existing set
+ */
+static void
+ip_set_destroy_set(ip_set_id_t index)
+{
+       struct ip_set *set = ip_set_list[index];
+
+       IP_SET_ASSERT(set);
+       DP("set: %s",  set->name);
+       write_lock_bh(&ip_set_lock);
+       FOREACH_HASH_RW_DO(__set_hash_del_byid, set->id);
+       if (set->binding != IP_SET_INVALID_ID)
+               __ip_set_put(set->binding);
+       ip_set_list[index] = NULL;
+       write_unlock_bh(&ip_set_lock);
+
+       /* Must call it without holding any lock */
+       set->type->destroy(set);
+       module_put(set->type->me);
+       kfree(set);
+}
+
+/*
+ * Destroy a set - or all sets
+ * Sets must not be referenced/used.
+ */
+static int
+ip_set_destroy(ip_set_id_t index)
+{
+       ip_set_id_t i;
+
+       /* ref modification always protected by the mutex */
+       if (index != IP_SET_INVALID_ID) {
+               if (atomic_read(&ip_set_list[index]->ref))
+                       return -EBUSY;
+               ip_set_destroy_set(index);
+       } else {
+               for (i = 0; i < ip_set_max; i++) {
+                       if (ip_set_list[i] != NULL 
+                           && (atomic_read(&ip_set_list[i]->ref)))
+                               return -EBUSY;
+               }
+
+               for (i = 0; i < ip_set_max; i++) {
+                       if (ip_set_list[i] != NULL)
+                               ip_set_destroy_set(i);
+               }
+       }
+       return 0;
+}
+
+static void
+ip_set_flush_set(struct ip_set *set)
+{
+       DP("set: %s %u",  set->name, set->id);
+
+       write_lock_bh(&set->lock);
+       set->type->flush(set);
+       write_unlock_bh(&set->lock);
+}
+
+/* 
+ * Flush data in a set - or in all sets
+ */
+static int
+ip_set_flush(ip_set_id_t index)
+{
+       if (index != IP_SET_INVALID_ID) {
+               IP_SET_ASSERT(ip_set_list[index]);
+               ip_set_flush_set(ip_set_list[index]);
+       } else
+               FOREACH_SET_DO(ip_set_flush_set);
+
+       return 0;
+}
+
+/* Rename a set */
+static int
+ip_set_rename(ip_set_id_t index, const char *name)
+{
+       struct ip_set *set = ip_set_list[index];
+       ip_set_id_t i;
+       int res = 0;
+
+       DP("set: %s to %s",  set->name, name);
+       write_lock_bh(&ip_set_lock);
+       for (i = 0; i < ip_set_max; i++) {
+               if (ip_set_list[i] != NULL
+                   && strncmp(ip_set_list[i]->name, 
+                              name,
+                              IP_SET_MAXNAMELEN - 1) == 0) {
+                       res = -EEXIST;
+                       goto unlock;
+               }
+       }
+       strncpy(set->name, name, IP_SET_MAXNAMELEN);
+    unlock:
+       write_unlock_bh(&ip_set_lock);
+       return res;
+}
+
+/*
+ * Swap two sets so that name/index points to the other.
+ * References are also swapped.
+ */
+static int
+ip_set_swap(ip_set_id_t from_index, ip_set_id_t to_index)
+{
+       struct ip_set *from = ip_set_list[from_index];
+       struct ip_set *to = ip_set_list[to_index];
+       char from_name[IP_SET_MAXNAMELEN];
+       u_int32_t from_ref;
+
+       DP("set: %s to %s",  from->name, to->name);
+       /* Features must not change. Artifical restriction. */
+       if (from->type->features != to->type->features)
+               return -ENOEXEC;
+
+       /* No magic here: ref munging protected by the mutex */ 
+       write_lock_bh(&ip_set_lock);
+       strncpy(from_name, from->name, IP_SET_MAXNAMELEN);
+       from_ref = atomic_read(&from->ref);
+
+       strncpy(from->name, to->name, IP_SET_MAXNAMELEN);
+       atomic_set(&from->ref, atomic_read(&to->ref));
+       strncpy(to->name, from_name, IP_SET_MAXNAMELEN);
+       atomic_set(&to->ref, from_ref);
+       
+       ip_set_list[from_index] = to;
+       ip_set_list[to_index] = from;
+       
+       write_unlock_bh(&ip_set_lock);
+       return 0;
+}
+
+/*
+ * List set data
+ */
+
+static inline void
+__set_hash_bindings_size_list(struct ip_set_hash *set_hash,
+                             ip_set_id_t id, size_t *size)
+{
+       if (set_hash->id == id)
+               *size += sizeof(struct ip_set_hash_list);
+}
+
+static inline void
+__set_hash_bindings_size_save(struct ip_set_hash *set_hash,
+                             ip_set_id_t id, size_t *size)
+{
+       if (set_hash->id == id)
+               *size += sizeof(struct ip_set_hash_save);
+}
+
+static inline void
+__set_hash_bindings(struct ip_set_hash *set_hash,
+                   ip_set_id_t id, void *data, int *used)
+{
+       if (set_hash->id == id) {
+               struct ip_set_hash_list *hash_list = 
+                       (struct ip_set_hash_list *)(data + *used);
+
+               hash_list->ip = set_hash->ip;
+               hash_list->binding = set_hash->binding;
+               *used += sizeof(struct ip_set_hash_list);
+       }
+}
+
+static int ip_set_list_set(ip_set_id_t index,
+                          void *data,
+                          int *used,
+                          int len)
+{
+       struct ip_set *set = ip_set_list[index];
+       struct ip_set_list *set_list;
+
+       /* Pointer to our header */
+       set_list = (struct ip_set_list *) (data + *used);
+
+       DP("set: %s, used: %d %p %p", set->name, *used, data, data + *used);
+
+       /* Get and ensure header size */
+       if (*used + sizeof(struct ip_set_list) > len)
+               goto not_enough_mem;
+       *used += sizeof(struct ip_set_list);
+
+       read_lock_bh(&set->lock);
+       /* Get and ensure set specific header size */
+       set_list->header_size = set->type->header_size;
+       if (*used + set_list->header_size > len)
+               goto unlock_set;
+
+       /* Fill in the header */
+       set_list->index = index;
+       set_list->binding = set->binding;
+       set_list->ref = atomic_read(&set->ref);
+
+       /* Fill in set spefific header data */
+       set->type->list_header(set, data + *used);
+       *used += set_list->header_size;
+
+       /* Get and ensure set specific members size */
+       set_list->members_size = set->type->list_members_size(set);
+       if (*used + set_list->members_size > len)
+               goto unlock_set;
+
+       /* Fill in set spefific members data */
+       set->type->list_members(set, data + *used);
+       *used += set_list->members_size;
+       read_unlock_bh(&set->lock);
+
+       /* Bindings */
+
+       /* Get and ensure set specific bindings size */
+       set_list->bindings_size = 0;
+       FOREACH_HASH_DO(__set_hash_bindings_size_list,
+                       set->id, &set_list->bindings_size);
+       if (*used + set_list->bindings_size > len)
+               goto not_enough_mem;
+
+       /* Fill in set spefific bindings data */
+       FOREACH_HASH_DO(__set_hash_bindings, set->id, data, used);
+       
+       return 0;
+
+    unlock_set:
+       read_unlock_bh(&set->lock);
+    not_enough_mem:
+       DP("not enough mem, try again");
+       return -EAGAIN;
+}
+
+/*
+ * Save sets
+ */
+static int ip_set_save_set(ip_set_id_t index,
+                          void *data,
+                          int *used,
+                          int len)
+{
+       struct ip_set *set;
+       struct ip_set_save *set_save;
+
+       /* Pointer to our header */
+       set_save = (struct ip_set_save *) (data + *used);
+
+       /* Get and ensure header size */
+       if (*used + sizeof(struct ip_set_save) > len)
+               goto not_enough_mem;
+       *used += sizeof(struct ip_set_save);
+
+       set = ip_set_list[index];
+       DP("set: %s, used: %u(%u) %p %p", set->name, *used, len, 
+          data, data + *used);
+
+       read_lock_bh(&set->lock);
+       /* Get and ensure set specific header size */
+       set_save->header_size = set->type->header_size;
+       if (*used + set_save->header_size > len)
+               goto unlock_set;
+
+       /* Fill in the header */
+       set_save->index = index;
+       set_save->binding = set->binding;
+
+       /* Fill in set spefific header data */
+       set->type->list_header(set, data + *used);
+       *used += set_save->header_size;
+
+       DP("set header filled: %s, used: %u %p %p", set->name, *used,
+          data, data + *used);
+       /* Get and ensure set specific members size */
+       set_save->members_size = set->type->list_members_size(set);
+       if (*used + set_save->members_size > len)
+               goto unlock_set;
+
+       /* Fill in set spefific members data */
+       set->type->list_members(set, data + *used);
+       *used += set_save->members_size;
+       read_unlock_bh(&set->lock);
+       DP("set members filled: %s, used: %u %p %p", set->name, *used,
+          data, data + *used);
+       return 0;
+
+    unlock_set:
+       read_unlock_bh(&set->lock);
+    not_enough_mem:
+       DP("not enough mem, try again");
+       return -EAGAIN;
+}
+
+static inline void
+__set_hash_save_bindings(struct ip_set_hash *set_hash,
+                        ip_set_id_t id,
+                        void *data,
+                        int *used,
+                        int len,
+                        int *res)
+{
+       if (*res == 0
+           && (id == IP_SET_INVALID_ID || set_hash->id == id)) {
+               struct ip_set_hash_save *hash_save = 
+                       (struct ip_set_hash_save *)(data + *used);
+               /* Ensure bindings size */
+               if (*used + sizeof(struct ip_set_hash_save) > len) {
+                       *res = -ENOMEM;
+                       return;
+               }
+               hash_save->id = set_hash->id;
+               hash_save->ip = set_hash->ip;
+               hash_save->binding = set_hash->binding;
+               *used += sizeof(struct ip_set_hash_save);
+       }
+}
+
+static int ip_set_save_bindings(ip_set_id_t index,
+                               void *data,
+                               int *used,
+                               int len)
+{
+       int res = 0;
+       struct ip_set_save *set_save;
+
+       DP("used %u, len %u", *used, len);
+       /* Get and ensure header size */
+       if (*used + sizeof(struct ip_set_save) > len)
+               return -ENOMEM;
+
+       /* Marker */
+       set_save = (struct ip_set_save *) (data + *used);
+       set_save->index = IP_SET_INVALID_ID;
+       *used += sizeof(struct ip_set_save);
+
+       DP("marker added used %u, len %u", *used, len);
+       /* Fill in bindings data */
+       if (index != IP_SET_INVALID_ID)
+               /* Sets are identified by id in hash */
+               index = ip_set_list[index]->id;
+       FOREACH_HASH_DO(__set_hash_save_bindings, index, data, used, len, &res);
+
+       return res;     
+}
+
+/*
+ * Restore sets
+ */
+static int ip_set_restore(void *data,
+                         int len)
+{
+       int res = 0;
+       int line = 0, used = 0, members_size;
+       struct ip_set *set;
+       struct ip_set_hash_save *hash_save;
+       struct ip_set_restore *set_restore;
+       ip_set_id_t index;
+
+       /* Loop to restore sets */
+       while (1) {
+               line++;
+               
+               DP("%u %u %u", used, sizeof(struct ip_set_restore), len);
+               /* Get and ensure header size */
+               if (used + sizeof(struct ip_set_restore) > len)
+                       return line;
+               set_restore = (struct ip_set_restore *) (data + used);
+               used += sizeof(struct ip_set_restore);
+
+               /* Ensure data size */
+               if (used 
+                   + set_restore->header_size 
+                   + set_restore->members_size > len)
+                       return line;
+
+               /* Check marker */
+               if (set_restore->index == IP_SET_INVALID_ID) {
+                       line--;
+                       goto bindings;
+               }
+               
+               /* Try to create the set */
+               DP("restore %s %s", set_restore->name, set_restore->typename);
+               res = ip_set_create(set_restore->name,
+                                   set_restore->typename,
+                                   set_restore->index,
+                                   data + used,
+                                   set_restore->header_size);
+               
+               if (res != 0)
+                       return line;
+               used += set_restore->header_size;
+
+               index = ip_set_find_byindex(set_restore->index);
+               DP("index %u, restore_index %u", index, set_restore->index);
+               if (index != set_restore->index)
+                       return line;
+               /* Try to restore members data */
+               set = ip_set_list[index];
+               members_size = 0;
+               DP("members_size %u reqsize %u",
+                  set_restore->members_size, set->type->reqsize);
+               while (members_size + set->type->reqsize <=
+                      set_restore->members_size) {
+                       line++;
+                       DP("members: %u, line %u", members_size, line);
+                       res = __ip_set_addip(index,
+                                          data + used + members_size,
+                                          set->type->reqsize);
+                       if (!(res == 0 || res == -EEXIST)) 
+                               return line;
+                       members_size += set->type->reqsize;
+               }
+
+               DP("members_size %u  %u",
+                  set_restore->members_size, members_size);
+               if (members_size != set_restore->members_size)
+                       return line++;
+               used += set_restore->members_size;              
+       }
+       
+   bindings:
+       /* Loop to restore bindings */
+       while (used < len) {
+               line++;
+
+               DP("restore binding, line %u", line);           
+               /* Get and ensure size */
+               if (used + sizeof(struct ip_set_hash_save) > len)
+                       return line;
+               hash_save = (struct ip_set_hash_save *) (data + used);
+               used += sizeof(struct ip_set_hash_save);
+               
+               /* hash_save->id is used to store the index */
+               index = ip_set_find_byindex(hash_save->id);
+               DP("restore binding index %u, id %u, %u -> %u",
+                  index, hash_save->id, hash_save->ip, hash_save->binding);            
+               if (index != hash_save->id)
+                       return line;
+                       
+               set = ip_set_list[hash_save->id];
+               /* Null valued IP means default binding */
+               if (hash_save->ip)
+                       res = ip_set_hash_add(set->id, 
+                                             hash_save->ip,
+                                             hash_save->binding);
+               else {
+                       IP_SET_ASSERT(set->binding == IP_SET_INVALID_ID);
+                       write_lock_bh(&ip_set_lock);
+                       set->binding = hash_save->binding;
+                       __ip_set_get(set->binding);
+                       write_unlock_bh(&ip_set_lock);
+                       DP("default binding: %u", set->binding);
+               }
+               if (res != 0)
+                       return line;
+       }
+       if (used != len)
+               return line;
+       
+       return 0;       
+}
+
+static int
+ip_set_sockfn_set(struct sock *sk, int optval, void *user, unsigned int len)
+{
+       void *data;
+       int res = 0;            /* Assume OK */
+       unsigned *op;
+       struct ip_set_req_adt *req_adt;
+       ip_set_id_t index = IP_SET_INVALID_ID;
+       int (*adtfn)(ip_set_id_t index,
+                    const void *data, size_t size);
+       struct fn_table {
+               int (*fn)(ip_set_id_t index,
+                         const void *data, size_t size);
+       } adtfn_table[] =
+       { { ip_set_addip }, { ip_set_delip }, { ip_set_testip},
+         { ip_set_bindip}, { ip_set_unbindip }, { ip_set_testbind },
+       };
+
+       DP("optval=%d, user=%p, len=%d", optval, user, len);
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
+       if (optval != SO_IP_SET)
+               return -EBADF;
+       if (len <= sizeof(unsigned)) {
+               ip_set_printk("short userdata (want >%zu, got %u)",
+                             sizeof(unsigned), len);
+               return -EINVAL;
+       }
+       data = vmalloc(len);
+       if (!data) {
+               DP("out of mem for %u bytes", len);
+               return -ENOMEM;
+       }
+       if (copy_from_user(data, user, len) != 0) {
+               res = -EFAULT;
+               goto done;
+       }
+       if (down_interruptible(&ip_set_app_mutex)) {
+               res = -EINTR;
+               goto done;
+       }
+
+       op = (unsigned *)data;
+       DP("op=%x", *op);
+       
+       if (*op < IP_SET_OP_VERSION) {
+               /* Check the version at the beginning of operations */
+               struct ip_set_req_version *req_version =
+                       (struct ip_set_req_version *) data;
+               if (req_version->version != IP_SET_PROTOCOL_VERSION) {
+                       res = -EPROTO;
+                       goto done;
+               }
+       }
+
+       switch (*op) {
+       case IP_SET_OP_CREATE:{
+               struct ip_set_req_create *req_create
+                       = (struct ip_set_req_create *) data;
+               
+               if (len <= sizeof(struct ip_set_req_create)) {
+                       ip_set_printk("short CREATE data (want >%zu, got %u)",
+                                     sizeof(struct ip_set_req_create), len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               req_create->name[IP_SET_MAXNAMELEN - 1] = '\0';
+               req_create->typename[IP_SET_MAXNAMELEN - 1] = '\0';
+               res = ip_set_create(req_create->name,
+                                   req_create->typename,
+                                   IP_SET_INVALID_ID,
+                                   data + sizeof(struct ip_set_req_create),
+                                   len - sizeof(struct ip_set_req_create));
+               goto done;
+       }
+       case IP_SET_OP_DESTROY:{
+               struct ip_set_req_std *req_destroy
+                       = (struct ip_set_req_std *) data;
+               
+               if (len != sizeof(struct ip_set_req_std)) {
+                       ip_set_printk("invalid DESTROY data (want %zu, got %u)",
+                                     sizeof(struct ip_set_req_std), len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               if (strcmp(req_destroy->name, IPSET_TOKEN_ALL) == 0) {
+                       /* Destroy all sets */
+                       index = IP_SET_INVALID_ID;
+               } else {
+                       req_destroy->name[IP_SET_MAXNAMELEN - 1] = '\0';
+                       index = ip_set_find_byname(req_destroy->name);
+
+                       if (index == IP_SET_INVALID_ID) {
+                               res = -ENOENT;
+                               goto done;
+                       }
+               }
+                       
+               res = ip_set_destroy(index);
+               goto done;
+       }
+       case IP_SET_OP_FLUSH:{
+               struct ip_set_req_std *req_flush =
+                       (struct ip_set_req_std *) data;
+
+               if (len != sizeof(struct ip_set_req_std)) {
+                       ip_set_printk("invalid FLUSH data (want %zu, got %u)",
+                                     sizeof(struct ip_set_req_std), len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               if (strcmp(req_flush->name, IPSET_TOKEN_ALL) == 0) {
+                       /* Flush all sets */
+                       index = IP_SET_INVALID_ID;
+               } else {
+                       req_flush->name[IP_SET_MAXNAMELEN - 1] = '\0';
+                       index = ip_set_find_byname(req_flush->name);
+
+                       if (index == IP_SET_INVALID_ID) {
+                               res = -ENOENT;
+                               goto done;
+                       }
+               }
+               res = ip_set_flush(index);
+               goto done;
+       }
+       case IP_SET_OP_RENAME:{
+               struct ip_set_req_create *req_rename
+                       = (struct ip_set_req_create *) data;
+
+               if (len != sizeof(struct ip_set_req_create)) {
+                       ip_set_printk("invalid RENAME data (want %zu, got %u)",
+                                     sizeof(struct ip_set_req_create), len);
+                       res = -EINVAL;
+                       goto done;
+               }
+
+               req_rename->name[IP_SET_MAXNAMELEN - 1] = '\0';
+               req_rename->typename[IP_SET_MAXNAMELEN - 1] = '\0';
+                       
+               index = ip_set_find_byname(req_rename->name);
+               if (index == IP_SET_INVALID_ID) {
+                       res = -ENOENT;
+                       goto done;
+               }
+               res = ip_set_rename(index, req_rename->typename);
+               goto done;
+       }
+       case IP_SET_OP_SWAP:{
+               struct ip_set_req_create *req_swap
+                       = (struct ip_set_req_create *) data;
+               ip_set_id_t to_index;
+
+               if (len != sizeof(struct ip_set_req_create)) {
+                       ip_set_printk("invalid SWAP data (want %zu, got %u)",
+                                     sizeof(struct ip_set_req_create), len);
+                       res = -EINVAL;
+                       goto done;
+               }
+
+               req_swap->name[IP_SET_MAXNAMELEN - 1] = '\0';
+               req_swap->typename[IP_SET_MAXNAMELEN - 1] = '\0';
+
+               index = ip_set_find_byname(req_swap->name);
+               if (index == IP_SET_INVALID_ID) {
+                       res = -ENOENT;
+                       goto done;
+               }
+               to_index = ip_set_find_byname(req_swap->typename);
+               if (to_index == IP_SET_INVALID_ID) {
+                       res = -ENOENT;
+                       goto done;
+               }
+               res = ip_set_swap(index, to_index);
+               goto done;
+       }
+       default: 
+               break;  /* Set identified by id */
+       }
+       
+       /* There we may have add/del/test/bind/unbind/test_bind operations */
+       if (*op < IP_SET_OP_ADD_IP || *op > IP_SET_OP_TEST_BIND_SET) {
+               res = -EBADMSG;
+               goto done;
+       }
+       adtfn = adtfn_table[*op - IP_SET_OP_ADD_IP].fn;
+
+       if (len < sizeof(struct ip_set_req_adt)) {
+               ip_set_printk("short data in adt request (want >=%zu, got %u)",
+                             sizeof(struct ip_set_req_adt), len);
+               res = -EINVAL;
+               goto done;
+       }
+       req_adt = (struct ip_set_req_adt *) data;
+
+       /* -U :all: :all:|:default: uses IP_SET_INVALID_ID */
+       if (!(*op == IP_SET_OP_UNBIND_SET 
+             && req_adt->index == IP_SET_INVALID_ID)) {
+               index = ip_set_find_byindex(req_adt->index);
+               if (index == IP_SET_INVALID_ID) {
+                       res = -ENOENT;
+                       goto done;
+               }
+       }
+       res = adtfn(index, data, len);
+
+    done:
+       up(&ip_set_app_mutex);
+       vfree(data);
+       if (res > 0)
+               res = 0;
+       DP("final result %d", res);
+       return res;
+}
+
+static int 
+ip_set_sockfn_get(struct sock *sk, int optval, void *user, int *len)
+{
+       int res = 0;
+       unsigned *op;
+       ip_set_id_t index = IP_SET_INVALID_ID;
+       void *data;
+       int copylen = *len;
+
+       DP("optval=%d, user=%p, len=%d", optval, user, *len);
+       if (!capable(CAP_NET_ADMIN))
+               return -EPERM;
+       if (optval != SO_IP_SET)
+               return -EBADF;
+       if (*len < sizeof(unsigned)) {
+               ip_set_printk("short userdata (want >=%zu, got %d)",
+                             sizeof(unsigned), *len);
+               return -EINVAL;
+       }
+       data = vmalloc(*len);
+       if (!data) {
+               DP("out of mem for %d bytes", *len);
+               return -ENOMEM;
+       }
+       if (copy_from_user(data, user, *len) != 0) {
+               res = -EFAULT;
+               goto done;
+       }
+       if (down_interruptible(&ip_set_app_mutex)) {
+               res = -EINTR;
+               goto done;
+       }
+
+       op = (unsigned *) data;
+       DP("op=%x", *op);
+
+       if (*op < IP_SET_OP_VERSION) {
+               /* Check the version at the beginning of operations */
+               struct ip_set_req_version *req_version =
+                       (struct ip_set_req_version *) data;
+               if (req_version->version != IP_SET_PROTOCOL_VERSION) {
+                       res = -EPROTO;
+                       goto done;
+               }
+       }
+
+       switch (*op) {
+       case IP_SET_OP_VERSION: {
+               struct ip_set_req_version *req_version =
+                   (struct ip_set_req_version *) data;
+
+               if (*len != sizeof(struct ip_set_req_version)) {
+                       ip_set_printk("invalid VERSION (want %zu, got %d)",
+                                     sizeof(struct ip_set_req_version),
+                                     *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+
+               req_version->version = IP_SET_PROTOCOL_VERSION;
+               res = copy_to_user(user, req_version,
+                                  sizeof(struct ip_set_req_version));
+               goto done;
+       }
+       case IP_SET_OP_GET_BYNAME: {
+               struct ip_set_req_get_set *req_get
+                       = (struct ip_set_req_get_set *) data;
+
+               if (*len != sizeof(struct ip_set_req_get_set)) {
+                       ip_set_printk("invalid GET_BYNAME (want %zu, got %d)",
+                                     sizeof(struct ip_set_req_get_set), *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
+               index = ip_set_find_byname(req_get->set.name);
+               req_get->set.index = index;
+               goto copy;
+       }
+       case IP_SET_OP_GET_BYINDEX: {
+               struct ip_set_req_get_set *req_get
+                       = (struct ip_set_req_get_set *) data;
+
+               if (*len != sizeof(struct ip_set_req_get_set)) {
+                       ip_set_printk("invalid GET_BYINDEX (want %zu, got %d)",
+                                     sizeof(struct ip_set_req_get_set), *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
+               index = ip_set_find_byindex(req_get->set.index);
+               strncpy(req_get->set.name,
+                       index == IP_SET_INVALID_ID ? ""
+                       : ip_set_list[index]->name, IP_SET_MAXNAMELEN);
+               goto copy;
+       }
+       case IP_SET_OP_ADT_GET: {
+               struct ip_set_req_adt_get *req_get
+                       = (struct ip_set_req_adt_get *) data;
+
+               if (*len != sizeof(struct ip_set_req_adt_get)) {
+                       ip_set_printk("invalid ADT_GET (want %zu, got %d)",
+                                     sizeof(struct ip_set_req_adt_get), *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               req_get->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
+               index = ip_set_find_byname(req_get->set.name);
+               if (index != IP_SET_INVALID_ID) {
+                       req_get->set.index = index;
+                       strncpy(req_get->typename,
+                               ip_set_list[index]->type->typename,
+                               IP_SET_MAXNAMELEN - 1);
+               } else {
+                       res = -ENOENT;
+                       goto done;
+               }
+               goto copy;
+       }
+       case IP_SET_OP_MAX_SETS: {
+               struct ip_set_req_max_sets *req_max_sets
+                       = (struct ip_set_req_max_sets *) data;
+               ip_set_id_t i;
+
+               if (*len != sizeof(struct ip_set_req_max_sets)) {
+                       ip_set_printk("invalid MAX_SETS (want %zu, got %d)",
+                                     sizeof(struct ip_set_req_max_sets), *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+
+               if (strcmp(req_max_sets->set.name, IPSET_TOKEN_ALL) == 0) {
+                       req_max_sets->set.index = IP_SET_INVALID_ID;
+               } else {
+                       req_max_sets->set.name[IP_SET_MAXNAMELEN - 1] = '\0';
+                       req_max_sets->set.index = 
+                               ip_set_find_byname(req_max_sets->set.name);
+                       if (req_max_sets->set.index == IP_SET_INVALID_ID) {
+                               res = -ENOENT;
+                               goto done;
+                       }
+               }
+               req_max_sets->max_sets = ip_set_max;
+               req_max_sets->sets = 0;
+               for (i = 0; i < ip_set_max; i++) {
+                       if (ip_set_list[i] != NULL)
+                               req_max_sets->sets++;
+               }
+               goto copy;
+       }
+       case IP_SET_OP_LIST_SIZE: 
+       case IP_SET_OP_SAVE_SIZE: {
+               struct ip_set_req_setnames *req_setnames
+                       = (struct ip_set_req_setnames *) data;
+               struct ip_set_name_list *name_list;
+               struct ip_set *set;
+               ip_set_id_t i;
+               int used;
+
+               if (*len < sizeof(struct ip_set_req_setnames)) {
+                       ip_set_printk("short LIST_SIZE (want >=%zu, got %d)",
+                                     sizeof(struct ip_set_req_setnames), *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+
+               req_setnames->size = 0;
+               used = sizeof(struct ip_set_req_setnames);
+               for (i = 0; i < ip_set_max; i++) {
+                       if (ip_set_list[i] == NULL)
+                               continue;
+                       name_list = (struct ip_set_name_list *) 
+                               (data + used);
+                       used += sizeof(struct ip_set_name_list);
+                       if (used > copylen) {
+                               res = -EAGAIN;
+                               goto done;
+                       }
+                       set = ip_set_list[i];
+                       /* Fill in index, name, etc. */
+                       name_list->index = i;
+                       name_list->id = set->id;
+                       strncpy(name_list->name,
+                               set->name,
+                               IP_SET_MAXNAMELEN - 1);
+                       strncpy(name_list->typename,
+                               set->type->typename,
+                               IP_SET_MAXNAMELEN - 1);
+                       DP("filled %s of type %s, index %u\n",
+                          name_list->name, name_list->typename,
+                          name_list->index);
+                       if (!(req_setnames->index == IP_SET_INVALID_ID
+                             || req_setnames->index == i))
+                             continue;
+                       /* Update size */
+                       switch (*op) {
+                       case IP_SET_OP_LIST_SIZE: {
+                               req_setnames->size += sizeof(struct ip_set_list)
+                                       + set->type->header_size
+                                       + set->type->list_members_size(set);
+                               FOREACH_HASH_DO(__set_hash_bindings_size_list, 
+                                               i, &req_setnames->size);
+                               break;
+                       }
+                       case IP_SET_OP_SAVE_SIZE: {
+                               req_setnames->size += sizeof(struct ip_set_save)
+                                       + set->type->header_size
+                                       + set->type->list_members_size(set);
+                               FOREACH_HASH_DO(__set_hash_bindings_size_save,
+                                               i, &req_setnames->size);
+                               break;
+                       }
+                       default:
+                               break;
+                       }
+               }
+               if (copylen != used) {
+                       res = -EAGAIN;
+                       goto done;
+               }
+               goto copy;
+       }
+       case IP_SET_OP_LIST: {
+               struct ip_set_req_list *req_list
+                       = (struct ip_set_req_list *) data;
+               ip_set_id_t i;
+               int used;
+
+               if (*len < sizeof(struct ip_set_req_list)) {
+                       ip_set_printk("short LIST (want >=%zu, got %d)",
+                                     sizeof(struct ip_set_req_list), *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               index = req_list->index;
+               if (index != IP_SET_INVALID_ID
+                   && ip_set_find_byindex(index) != index) {
+                       res = -ENOENT;
+                       goto done;
+               }
+               used = 0;
+               if (index == IP_SET_INVALID_ID) {
+                       /* List all sets */
+                       for (i = 0; i < ip_set_max && res == 0; i++) {
+                               if (ip_set_list[i] != NULL)
+                                       res = ip_set_list_set(i, data, &used, *len);
+                       }
+               } else {
+                       /* List an individual set */
+                       res = ip_set_list_set(index, data, &used, *len);
+               }
+               if (res != 0)
+                       goto done;
+               else if (copylen != used) {
+                       res = -EAGAIN;
+                       goto done;
+               }
+               goto copy;
+       }
+       case IP_SET_OP_SAVE: {
+               struct ip_set_req_list *req_save
+                       = (struct ip_set_req_list *) data;
+               ip_set_id_t i;
+               int used;
+
+               if (*len < sizeof(struct ip_set_req_list)) {
+                       ip_set_printk("short SAVE (want >=%zu, got %d)",
+                                     sizeof(struct ip_set_req_list), *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               index = req_save->index;
+               if (index != IP_SET_INVALID_ID
+                   && ip_set_find_byindex(index) != index) {
+                       res = -ENOENT;
+                       goto done;
+               }
+               used = 0;
+               if (index == IP_SET_INVALID_ID) {
+                       /* Save all sets */
+                       for (i = 0; i < ip_set_max && res == 0; i++) {
+                               if (ip_set_list[i] != NULL)
+                                       res = ip_set_save_set(i, data, &used, *len);
+                       }
+               } else {
+                       /* Save an individual set */
+                       res = ip_set_save_set(index, data, &used, *len);
+               }
+               if (res == 0)
+                       res = ip_set_save_bindings(index, data, &used, *len);
+                       
+               if (res != 0)
+                       goto done;
+               else if (copylen != used) {
+                       res = -EAGAIN;
+                       goto done;
+               }
+               goto copy;
+       }
+       case IP_SET_OP_RESTORE: {
+               struct ip_set_req_setnames *req_restore
+                       = (struct ip_set_req_setnames *) data;
+               int line;
+
+               if (*len < sizeof(struct ip_set_req_setnames)
+                   || *len != req_restore->size) {
+                       ip_set_printk("invalid RESTORE (want =%zu, got %d)",
+                                     req_restore->size, *len);
+                       res = -EINVAL;
+                       goto done;
+               }
+               line = ip_set_restore(data + sizeof(struct ip_set_req_setnames),
+                                     req_restore->size - sizeof(struct ip_set_req_setnames));
+               DP("ip_set_restore: %u", line);
+               if (line != 0) {
+                       res = -EAGAIN;
+                       req_restore->size = line;
+                       copylen = sizeof(struct ip_set_req_setnames);
+                       goto copy;
+               }
+               goto done;
+       }
+       default:
+               res = -EBADMSG;
+               goto done;
+       }       /* end of switch(op) */
+
+    copy:
+       DP("set %s, copylen %u", index != IP_SET_INVALID_ID
+                                && ip_set_list[index]
+                    ? ip_set_list[index]->name
+                    : ":all:", copylen);
+       res = copy_to_user(user, data, copylen);
+       
+    done:
+       up(&ip_set_app_mutex);
+       vfree(data);
+       if (res > 0)
+               res = 0;
+       DP("final result %d", res);
+       return res;
+}
+
+static struct nf_sockopt_ops so_set = {
+       .pf             = PF_INET,
+       .set_optmin     = SO_IP_SET,
+       .set_optmax     = SO_IP_SET + 1,
+       .set            = &ip_set_sockfn_set,
+       .get_optmin     = SO_IP_SET,
+       .get_optmax     = SO_IP_SET + 1,
+       .get            = &ip_set_sockfn_get,
+       .use            = 0
+};
+
+static int max_sets, hash_size;
+module_param(max_sets, int, 0600);
+MODULE_PARM_DESC(max_sets, "maximal number of sets");
+module_param(hash_size, int, 0600);
+MODULE_PARM_DESC(hash_size, "hash size for bindings");
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("module implementing core IP set support");
+
+static int __init init(void)
+{
+       int res;
+       ip_set_id_t i;
+
+       get_random_bytes(&ip_set_hash_random, 4);
+       if (max_sets)
+               ip_set_max = max_sets;
+       ip_set_list = vmalloc(sizeof(struct ip_set *) * ip_set_max);
+       if (!ip_set_list) {
+               printk(KERN_ERR "Unable to create ip_set_list\n");
+               return -ENOMEM;
+       }
+       memset(ip_set_list, 0, sizeof(struct ip_set *) * ip_set_max);
+       if (hash_size)
+               ip_set_bindings_hash_size = hash_size;
+       ip_set_hash = vmalloc(sizeof(struct list_head) * ip_set_bindings_hash_size);
+       if (!ip_set_hash) {
+               printk(KERN_ERR "Unable to create ip_set_hash\n");
+               vfree(ip_set_list);
+               return -ENOMEM;
+       }
+       for (i = 0; i < ip_set_bindings_hash_size; i++)
+               INIT_LIST_HEAD(&ip_set_hash[i]);
+
+       INIT_LIST_HEAD(&set_type_list);
+
+       res = nf_register_sockopt(&so_set);
+       if (res != 0) {
+               ip_set_printk("SO_SET registry failed: %d", res);
+               vfree(ip_set_list);
+               vfree(ip_set_hash);
+               return res;
+       }
+       return 0;
+}
+
+static void __exit fini(void)
+{
+       /* There can't be any existing set or binding */
+       nf_unregister_sockopt(&so_set);
+       vfree(ip_set_list);
+       vfree(ip_set_hash);
+       DP("these are the famous last words");
+}
+
+EXPORT_SYMBOL(ip_set_register_set_type);
+EXPORT_SYMBOL(ip_set_unregister_set_type);
+
+EXPORT_SYMBOL(ip_set_get_byname);
+EXPORT_SYMBOL(ip_set_get_byindex);
+EXPORT_SYMBOL(ip_set_put);
+
+EXPORT_SYMBOL(ip_set_addip_kernel);
+EXPORT_SYMBOL(ip_set_delip_kernel);
+EXPORT_SYMBOL(ip_set_testip_kernel);
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ip_set_iphash.c b/net/ipv4/netfilter/ip_set_iphash.c
new file mode 100644 (file)
index 0000000..ef6dfa3
--- /dev/null
@@ -0,0 +1,398 @@
+/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module implementing an ip hash set */
+
+#include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/skbuff.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+#include <linux/errno.h>
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <linux/spinlock.h>
+#include <linux/vmalloc.h>
+#include <linux/random.h>
+
+#include <net/ip.h>
+
+#include <linux/netfilter_ipv4/ip_set_malloc.h>
+#include <linux/netfilter_ipv4/ip_set_iphash.h>
+#include <linux/netfilter_ipv4/ip_set_jhash.h>
+
+static inline __u32
+jhash_ip(const struct ip_set_iphash *map, uint16_t i, ip_set_ip_t ip)
+{
+       return jhash_1word(ip, *(((uint32_t *) map->initval) + i));
+}
+
+static inline __u32
+hash_id(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
+       __u32 id;
+       u_int16_t i;
+       ip_set_ip_t *elem;
+
+       *hash_ip = ip & map->netmask;
+       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u, %u.%u.%u.%u",
+          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip), HIPQUAD(map->netmask));
+       
+       for (i = 0; i < map->probes; i++) {
+               id = jhash_ip(map, i, *hash_ip) % map->hashsize;
+               DP("hash key: %u", id);
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
+               if (*elem == *hash_ip)
+                       return id;
+               /* No shortcut at testing - there can be deleted
+                * entries. */
+       }
+       return UINT_MAX;
+}
+
+static inline int
+__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       return (hash_id(set, ip, hash_ip) != UINT_MAX);
+}
+
+static int
+testip(struct ip_set *set, const void *data, size_t size,
+       ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_iphash *req = 
+           (struct ip_set_req_iphash *) data;
+
+       if (size != sizeof(struct ip_set_req_iphash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_iphash),
+                             size);
+               return -EINVAL;
+       }
+       return __testip(set, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       return __testip(set,
+                       ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                       hash_ip);
+}
+
+static inline int
+__addip(struct ip_set_iphash *map, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       __u32 probe;
+       u_int16_t i;
+       ip_set_ip_t *elem;
+
+       *hash_ip = ip & map->netmask;
+       
+       for (i = 0; i < map->probes; i++) {
+               probe = jhash_ip(map, i, *hash_ip) % map->hashsize;
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, probe);
+               if (*elem == *hash_ip)
+                       return -EEXIST;
+               if (!*elem) {
+                       *elem = *hash_ip;
+                       return 0;
+               }
+       }
+       /* Trigger rehashing */
+       return -EAGAIN;
+}
+
+static int
+addip(struct ip_set *set, const void *data, size_t size,
+        ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_iphash *req = 
+           (struct ip_set_req_iphash *) data;
+
+       if (size != sizeof(struct ip_set_req_iphash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_iphash),
+                             size);
+               return -EINVAL;
+       }
+       return __addip((struct ip_set_iphash *) set->data, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       return __addip((struct ip_set_iphash *) set->data,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      hash_ip);
+}
+
+static int retry(struct ip_set *set)
+{
+       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
+       ip_set_ip_t hash_ip, *elem;
+       void *members;
+       u_int32_t i, hashsize = map->hashsize;
+       int res;
+       struct ip_set_iphash *tmp;
+       
+       if (map->resize == 0)
+               return -ERANGE;
+
+    again:
+       res = 0;
+       
+       /* Calculate new hash size */
+       hashsize += (hashsize * map->resize)/100;
+       if (hashsize == map->hashsize)
+               hashsize++;
+       
+       ip_set_printk("rehashing of set %s triggered: "
+                     "hashsize grows from %u to %u",
+                     set->name, map->hashsize, hashsize);
+
+       tmp = kmalloc(sizeof(struct ip_set_iphash) 
+                     + map->probes * sizeof(uint32_t), GFP_ATOMIC);
+       if (!tmp) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_iphash)
+                  + map->probes * sizeof(uint32_t));
+               return -ENOMEM;
+       }
+       tmp->members = harray_malloc(hashsize, sizeof(ip_set_ip_t), GFP_ATOMIC);
+       if (!tmp->members) {
+               DP("out of memory for %d bytes", hashsize * sizeof(ip_set_ip_t));
+               kfree(tmp);
+               return -ENOMEM;
+       }
+       tmp->hashsize = hashsize;
+       tmp->probes = map->probes;
+       tmp->resize = map->resize;
+       tmp->netmask = map->netmask;
+       memcpy(tmp->initval, map->initval, map->probes * sizeof(uint32_t));
+       
+       write_lock_bh(&set->lock);
+       map = (struct ip_set_iphash *) set->data; /* Play safe */
+       for (i = 0; i < map->hashsize && res == 0; i++) {
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);     
+               if (*elem)
+                       res = __addip(tmp, *elem, &hash_ip);
+       }
+       if (res) {
+               /* Failure, try again */
+               write_unlock_bh(&set->lock);
+               harray_free(tmp->members);
+               kfree(tmp);
+               goto again;
+       }
+       
+       /* Success at resizing! */
+       members = map->members;
+
+       map->hashsize = tmp->hashsize;
+       map->members = tmp->members;
+       write_unlock_bh(&set->lock);
+
+       harray_free(members);
+       kfree(tmp);
+
+       return 0;
+}
+
+static inline int
+__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
+       ip_set_ip_t id = hash_id(set, ip, hash_ip);
+       ip_set_ip_t *elem;
+
+       if (id == UINT_MAX)
+               return -EEXIST;
+               
+       elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
+       *elem = 0;
+
+       return 0;
+}
+
+static int
+delip(struct ip_set *set, const void *data, size_t size,
+        ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_iphash *req =
+           (struct ip_set_req_iphash *) data;
+
+       if (size != sizeof(struct ip_set_req_iphash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_iphash),
+                             size);
+               return -EINVAL;
+       }
+       return __delip(set, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       return __delip(set,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      hash_ip);
+}
+
+static int create(struct ip_set *set, const void *data, size_t size)
+{
+       struct ip_set_req_iphash_create *req =
+           (struct ip_set_req_iphash_create *) data;
+       struct ip_set_iphash *map;
+       uint16_t i;
+
+       if (size != sizeof(struct ip_set_req_iphash_create)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                              sizeof(struct ip_set_req_iphash_create),
+                              size);
+               return -EINVAL;
+       }
+
+       if (req->hashsize < 1) {
+               ip_set_printk("hashsize too small");
+               return -ENOEXEC;
+       }
+
+       if (req->probes < 1) {
+               ip_set_printk("probes too small");
+               return -ENOEXEC;
+       }
+
+       map = kmalloc(sizeof(struct ip_set_iphash) 
+                     + req->probes * sizeof(uint32_t), GFP_KERNEL);
+       if (!map) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_iphash)
+                  + req->probes * sizeof(uint32_t));
+               return -ENOMEM;
+       }
+       for (i = 0; i < req->probes; i++)
+               get_random_bytes(((uint32_t *) map->initval)+i, 4);
+       map->hashsize = req->hashsize;
+       map->probes = req->probes;
+       map->resize = req->resize;
+       map->netmask = req->netmask;
+       map->members = harray_malloc(map->hashsize, sizeof(ip_set_ip_t), GFP_KERNEL);
+       if (!map->members) {
+               DP("out of memory for %d bytes", map->hashsize * sizeof(ip_set_ip_t));
+               kfree(map);
+               return -ENOMEM;
+       }
+
+       set->data = map;
+       return 0;
+}
+
+static void destroy(struct ip_set *set)
+{
+       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
+
+       harray_free(map->members);
+       kfree(map);
+
+       set->data = NULL;
+}
+
+static void flush(struct ip_set *set)
+{
+       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
+       harray_flush(map->members, map->hashsize, sizeof(ip_set_ip_t));
+}
+
+static void list_header(const struct ip_set *set, void *data)
+{
+       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
+       struct ip_set_req_iphash_create *header =
+           (struct ip_set_req_iphash_create *) data;
+
+       header->hashsize = map->hashsize;
+       header->probes = map->probes;
+       header->resize = map->resize;
+       header->netmask = map->netmask;
+}
+
+static int list_members_size(const struct ip_set *set)
+{
+       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
+
+       return (map->hashsize * sizeof(ip_set_ip_t));
+}
+
+static void list_members(const struct ip_set *set, void *data)
+{
+       struct ip_set_iphash *map = (struct ip_set_iphash *) set->data;
+       ip_set_ip_t i, *elem;
+
+       for (i = 0; i < map->hashsize; i++) {
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);     
+               ((ip_set_ip_t *)data)[i] = *elem;
+       }
+}
+
+static struct ip_set_type ip_set_iphash = {
+       .typename               = SETTYPE_NAME,
+       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
+       .protocol_version       = IP_SET_PROTOCOL_VERSION,
+       .create                 = &create,
+       .destroy                = &destroy,
+       .flush                  = &flush,
+       .reqsize                = sizeof(struct ip_set_req_iphash),
+       .addip                  = &addip,
+       .addip_kernel           = &addip_kernel,
+       .retry                  = &retry,
+       .delip                  = &delip,
+       .delip_kernel           = &delip_kernel,
+       .testip                 = &testip,
+       .testip_kernel          = &testip_kernel,
+       .header_size            = sizeof(struct ip_set_req_iphash_create),
+       .list_header            = &list_header,
+       .list_members_size      = &list_members_size,
+       .list_members           = &list_members,
+       .me                     = THIS_MODULE,
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("iphash type of IP sets");
+
+static int __init init(void)
+{
+       return ip_set_register_set_type(&ip_set_iphash);
+}
+
+static void __exit fini(void)
+{
+       /* FIXME: possible race with ip_set_create() */
+       ip_set_unregister_set_type(&ip_set_iphash);
+}
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ip_set_ipmap.c b/net/ipv4/netfilter/ip_set_ipmap.c
new file mode 100644 (file)
index 0000000..a173b0e
--- /dev/null
@@ -0,0 +1,327 @@
+/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
+ *                         Patrick Schaaf <bof@bof.de>
+ * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module implementing an IP set type: the single bitmap type */
+
+#include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/skbuff.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+#include <linux/errno.h>
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <linux/spinlock.h>
+
+#include <linux/netfilter_ipv4/ip_set_ipmap.h>
+
+static inline ip_set_ip_t
+ip_to_id(const struct ip_set_ipmap *map, ip_set_ip_t ip)
+{
+       return (ip - map->first_ip)/map->hosts;
+}
+
+static inline int
+__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
+       
+       if (ip < map->first_ip || ip > map->last_ip)
+               return -ERANGE;
+
+       *hash_ip = ip & map->netmask;
+       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
+          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip));
+       return !!test_bit(ip_to_id(map, *hash_ip), map->members);
+}
+
+static int
+testip(struct ip_set *set, const void *data, size_t size,
+       ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_ipmap *req = 
+           (struct ip_set_req_ipmap *) data;
+
+       if (size != sizeof(struct ip_set_req_ipmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_ipmap),
+                             size);
+               return -EINVAL;
+       }
+       return __testip(set, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       int res;
+       
+       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
+          flags[index] & IPSET_SRC ? "SRC" : "DST",
+          NIPQUAD(skb->nh.iph->saddr),
+          NIPQUAD(skb->nh.iph->daddr));
+
+       res =  __testip(set,
+                       ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                       hash_ip);
+       return (res < 0 ? 0 : res);
+}
+
+static inline int
+__addip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
+
+       if (ip < map->first_ip || ip > map->last_ip)
+               return -ERANGE;
+
+       *hash_ip = ip & map->netmask;
+       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
+       if (test_and_set_bit(ip_to_id(map, *hash_ip), map->members))
+               return -EEXIST;
+
+       return 0;
+}
+
+static int
+addip(struct ip_set *set, const void *data, size_t size,
+      ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_ipmap *req = 
+           (struct ip_set_req_ipmap *) data;
+
+       if (size != sizeof(struct ip_set_req_ipmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_ipmap),
+                             size);
+               return -EINVAL;
+       }
+       DP("%u.%u.%u.%u", HIPQUAD(req->ip));
+       return __addip(set, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       return __addip(set,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      hash_ip);
+}
+
+static inline int 
+__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
+
+       if (ip < map->first_ip || ip > map->last_ip)
+               return -ERANGE;
+
+       *hash_ip = ip & map->netmask;
+       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
+       if (!test_and_clear_bit(ip_to_id(map, *hash_ip), map->members))
+               return -EEXIST;
+       
+       return 0;
+}
+
+static int
+delip(struct ip_set *set, const void *data, size_t size,
+      ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_ipmap *req =
+           (struct ip_set_req_ipmap *) data;
+
+       if (size != sizeof(struct ip_set_req_ipmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_ipmap),
+                             size);
+               return -EINVAL;
+       }
+       return __delip(set, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       return __delip(set,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      hash_ip);
+}
+
+static int create(struct ip_set *set, const void *data, size_t size)
+{
+       int newbytes;
+       struct ip_set_req_ipmap_create *req =
+           (struct ip_set_req_ipmap_create *) data;
+       struct ip_set_ipmap *map;
+
+       if (size != sizeof(struct ip_set_req_ipmap_create)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_ipmap_create),
+                             size);
+               return -EINVAL;
+       }
+
+       DP("from %u.%u.%u.%u to %u.%u.%u.%u",
+          HIPQUAD(req->from), HIPQUAD(req->to));
+
+       if (req->from > req->to) {
+               DP("bad ip range");
+               return -ENOEXEC;
+       }
+
+       map = kmalloc(sizeof(struct ip_set_ipmap), GFP_KERNEL);
+       if (!map) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_ipmap));
+               return -ENOMEM;
+       }
+       map->first_ip = req->from;
+       map->last_ip = req->to;
+       map->netmask = req->netmask;
+
+       if (req->netmask == 0xFFFFFFFF) {
+               map->hosts = 1;
+               map->sizeid = map->last_ip - map->first_ip + 1;
+       } else {
+               unsigned int mask_bits, netmask_bits;
+               ip_set_ip_t mask;
+               
+               map->first_ip &= map->netmask;  /* Should we better bark? */
+               
+               mask = range_to_mask(map->first_ip, map->last_ip, &mask_bits);
+               netmask_bits = mask_to_bits(map->netmask);
+               
+               if ((!mask && (map->first_ip || map->last_ip != 0xFFFFFFFF))
+                   || netmask_bits <= mask_bits)
+                       return -ENOEXEC;
+
+               DP("mask_bits %u, netmask_bits %u",
+                  mask_bits, netmask_bits);
+               map->hosts = 2 << (32 - netmask_bits - 1);
+               map->sizeid = 2 << (netmask_bits - mask_bits - 1);
+       }
+       if (map->sizeid > MAX_RANGE + 1) {
+               ip_set_printk("range too big (max %d addresses)",
+                              MAX_RANGE+1);
+               kfree(map);
+               return -ENOEXEC;
+       }
+       DP("hosts %u, sizeid %u", map->hosts, map->sizeid);
+       newbytes = bitmap_bytes(0, map->sizeid - 1);
+       map->members = kmalloc(newbytes, GFP_KERNEL);
+       if (!map->members) {
+               DP("out of memory for %d bytes", newbytes);
+               kfree(map);
+               return -ENOMEM;
+       }
+       memset(map->members, 0, newbytes);
+       
+       set->data = map;
+       return 0;
+}
+
+static void destroy(struct ip_set *set)
+{
+       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
+       
+       kfree(map->members);
+       kfree(map);
+       
+       set->data = NULL;
+}
+
+static void flush(struct ip_set *set)
+{
+       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
+       memset(map->members, 0, bitmap_bytes(0, map->sizeid - 1));
+}
+
+static void list_header(const struct ip_set *set, void *data)
+{
+       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
+       struct ip_set_req_ipmap_create *header =
+           (struct ip_set_req_ipmap_create *) data;
+
+       header->from = map->first_ip;
+       header->to = map->last_ip;
+       header->netmask = map->netmask;
+}
+
+static int list_members_size(const struct ip_set *set)
+{
+       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
+
+       return bitmap_bytes(0, map->sizeid - 1);
+}
+
+static void list_members(const struct ip_set *set, void *data)
+{
+       struct ip_set_ipmap *map = (struct ip_set_ipmap *) set->data;
+       int bytes = bitmap_bytes(0, map->sizeid - 1);
+
+       memcpy(data, map->members, bytes);
+}
+
+static struct ip_set_type ip_set_ipmap = {
+       .typename               = SETTYPE_NAME,
+       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
+       .protocol_version       = IP_SET_PROTOCOL_VERSION,
+       .create                 = &create,
+       .destroy                = &destroy,
+       .flush                  = &flush,
+       .reqsize                = sizeof(struct ip_set_req_ipmap),
+       .addip                  = &addip,
+       .addip_kernel           = &addip_kernel,
+       .delip                  = &delip,
+       .delip_kernel           = &delip_kernel,
+       .testip                 = &testip,
+       .testip_kernel          = &testip_kernel,
+       .header_size            = sizeof(struct ip_set_req_ipmap_create),
+       .list_header            = &list_header,
+       .list_members_size      = &list_members_size,
+       .list_members           = &list_members,
+       .me                     = THIS_MODULE,
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("ipmap type of IP sets");
+
+static int __init init(void)
+{
+       return ip_set_register_set_type(&ip_set_ipmap);
+}
+
+static void __exit fini(void)
+{
+       /* FIXME: possible race with ip_set_create() */
+       ip_set_unregister_set_type(&ip_set_ipmap);
+}
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ip_set_ipporthash.c b/net/ipv4/netfilter/ip_set_ipporthash.c
new file mode 100644 (file)
index 0000000..86a4c6c
--- /dev/null
@@ -0,0 +1,524 @@
+/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module implementing an ip+port hash set */
+
+#include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/udp.h>
+#include <linux/skbuff.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+#include <linux/errno.h>
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <linux/spinlock.h>
+#include <linux/vmalloc.h>
+#include <linux/random.h>
+
+#include <net/ip.h>
+
+#include <linux/netfilter_ipv4/ip_set_malloc.h>
+#include <linux/netfilter_ipv4/ip_set_ipporthash.h>
+#include <linux/netfilter_ipv4/ip_set_jhash.h>
+
+/* We must handle non-linear skbs */
+static inline ip_set_ip_t
+get_port(const struct sk_buff *skb, u_int32_t flags)
+{
+       struct iphdr *iph = skb->nh.iph;
+       u_int16_t offset = ntohs(iph->frag_off) & IP_OFFSET;
+
+       switch (iph->protocol) {
+       case IPPROTO_TCP: {
+               struct tcphdr tcph;
+               
+               /* See comments at tcp_match in ip_tables.c */
+               if (offset)
+                       return INVALID_PORT;
+
+               if (skb_copy_bits(skb, skb->nh.iph->ihl*4, &tcph, sizeof(tcph)) < 0)
+                       /* No choice either */
+                       return INVALID_PORT;
+               
+               return ntohs(flags & IPSET_SRC ?
+                            tcph.source : tcph.dest);
+           }
+       case IPPROTO_UDP: {
+               struct udphdr udph;
+
+               if (offset)
+                       return INVALID_PORT;
+
+               if (skb_copy_bits(skb, skb->nh.iph->ihl*4, &udph, sizeof(udph)) < 0)
+                       /* No choice either */
+                       return INVALID_PORT;
+               
+               return ntohs(flags & IPSET_SRC ?
+                            udph.source : udph.dest);
+           }
+       default:
+               return INVALID_PORT;
+       }
+}
+
+static inline __u32
+jhash_ip(const struct ip_set_ipporthash *map, uint16_t i, ip_set_ip_t ip)
+{
+       return jhash_1word(ip, *(((uint32_t *) map->initval) + i));
+}
+
+#define HASH_IP(map, ip, port) (port + ((ip - ((map)->first_ip)) << 16))
+
+static inline __u32
+hash_id(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t port,
+       ip_set_ip_t *hash_ip)
+{
+       struct ip_set_ipporthash *map = 
+               (struct ip_set_ipporthash *) set->data;
+       __u32 id;
+       u_int16_t i;
+       ip_set_ip_t *elem;
+
+       *hash_ip = HASH_IP(map, ip, port);
+       DP("set: %s, ipport:%u.%u.%u.%u:%u, %u.%u.%u.%u",
+          set->name, HIPQUAD(ip), port, HIPQUAD(*hash_ip));
+       
+       for (i = 0; i < map->probes; i++) {
+               id = jhash_ip(map, i, *hash_ip) % map->hashsize;
+               DP("hash key: %u", id);
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
+               if (*elem == *hash_ip)
+                       return id;
+               /* No shortcut at testing - there can be deleted
+                * entries. */
+       }
+       return UINT_MAX;
+}
+
+static inline int
+__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t port,
+        ip_set_ip_t *hash_ip)
+{
+       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
+       
+       if (ip < map->first_ip || ip > map->last_ip)
+               return -ERANGE;
+
+       return (hash_id(set, ip, port, hash_ip) != UINT_MAX);
+}
+
+static int
+testip(struct ip_set *set, const void *data, size_t size,
+       ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_ipporthash *req = 
+           (struct ip_set_req_ipporthash *) data;
+
+       if (size != sizeof(struct ip_set_req_ipporthash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_ipporthash),
+                             size);
+               return -EINVAL;
+       }
+       return __testip(set, req->ip, req->port, hash_ip);
+}
+
+static int
+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)
+{
+       ip_set_ip_t port;
+
+       if (flags[index+1] == 0)
+               return -EINVAL;
+               
+       port = get_port(skb, flags[index+1]);
+
+       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
+          flags[index] & IPSET_SRC ? "SRC" : "DST",
+          NIPQUAD(skb->nh.iph->saddr),
+          NIPQUAD(skb->nh.iph->daddr));
+       DP("flag %s port %u",
+          flags[index+1] & IPSET_SRC ? "SRC" : "DST", 
+          port);       
+       if (port == INVALID_PORT)
+               return 0;       
+
+       return __testip(set,
+                       ntohl(flags[index] & IPSET_SRC 
+                                       ? skb->nh.iph->saddr 
+                                       : skb->nh.iph->daddr),
+                       port,
+                       hash_ip);
+}
+
+static inline int
+__add_haship(struct ip_set_ipporthash *map, ip_set_ip_t hash_ip)
+{
+       __u32 probe;
+       u_int16_t i;
+       ip_set_ip_t *elem;
+
+       for (i = 0; i < map->probes; i++) {
+               probe = jhash_ip(map, i, hash_ip) % map->hashsize;
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, probe);
+               if (*elem == hash_ip)
+                       return -EEXIST;
+               if (!*elem) {
+                       *elem = hash_ip;
+                       return 0;
+               }
+       }
+       /* Trigger rehashing */
+       return -EAGAIN;
+}
+
+static inline int
+__addip(struct ip_set_ipporthash *map, ip_set_ip_t ip, ip_set_ip_t port,
+       ip_set_ip_t *hash_ip)
+{
+       if (ip < map->first_ip || ip > map->last_ip)
+               return -ERANGE;
+
+       *hash_ip = HASH_IP(map, ip, port);
+       
+       return __add_haship(map, *hash_ip);
+}
+
+static int
+addip(struct ip_set *set, const void *data, size_t size,
+        ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_ipporthash *req = 
+           (struct ip_set_req_ipporthash *) data;
+
+       if (size != sizeof(struct ip_set_req_ipporthash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_ipporthash),
+                             size);
+               return -EINVAL;
+       }
+       return __addip((struct ip_set_ipporthash *) set->data, 
+                       req->ip, req->port, hash_ip);
+}
+
+static int
+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)
+{
+       ip_set_ip_t port;
+
+       if (flags[index+1] == 0)
+               return -EINVAL;
+               
+       port = get_port(skb, flags[index+1]);
+
+       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
+          flags[index] & IPSET_SRC ? "SRC" : "DST",
+          NIPQUAD(skb->nh.iph->saddr),
+          NIPQUAD(skb->nh.iph->daddr));
+       DP("flag %s port %u", 
+          flags[index+1] & IPSET_SRC ? "SRC" : "DST", 
+          port);       
+       if (port == INVALID_PORT)
+               return -EINVAL; 
+
+       return __addip((struct ip_set_ipporthash *) set->data,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      port,
+                      hash_ip);
+}
+
+static int retry(struct ip_set *set)
+{
+       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
+       ip_set_ip_t *elem;
+       void *members;
+       u_int32_t i, hashsize = map->hashsize;
+       int res;
+       struct ip_set_ipporthash *tmp;
+       
+       if (map->resize == 0)
+               return -ERANGE;
+
+    again:
+       res = 0;
+       
+       /* Calculate new hash size */
+       hashsize += (hashsize * map->resize)/100;
+       if (hashsize == map->hashsize)
+               hashsize++;
+       
+       ip_set_printk("rehashing of set %s triggered: "
+                     "hashsize grows from %u to %u",
+                     set->name, map->hashsize, hashsize);
+
+       tmp = kmalloc(sizeof(struct ip_set_ipporthash) 
+                     + map->probes * sizeof(uint32_t), GFP_ATOMIC);
+       if (!tmp) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_ipporthash)
+                  + map->probes * sizeof(uint32_t));
+               return -ENOMEM;
+       }
+       tmp->members = harray_malloc(hashsize, sizeof(ip_set_ip_t), GFP_ATOMIC);
+       if (!tmp->members) {
+               DP("out of memory for %d bytes", hashsize * sizeof(ip_set_ip_t));
+               kfree(tmp);
+               return -ENOMEM;
+       }
+       tmp->hashsize = hashsize;
+       tmp->probes = map->probes;
+       tmp->resize = map->resize;
+       tmp->first_ip = map->first_ip;
+       tmp->last_ip = map->last_ip;
+       memcpy(tmp->initval, map->initval, map->probes * sizeof(uint32_t));
+       
+       write_lock_bh(&set->lock);
+       map = (struct ip_set_ipporthash *) set->data; /* Play safe */
+       for (i = 0; i < map->hashsize && res == 0; i++) {
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);     
+               if (*elem)
+                       res = __add_haship(tmp, *elem);
+       }
+       if (res) {
+               /* Failure, try again */
+               write_unlock_bh(&set->lock);
+               harray_free(tmp->members);
+               kfree(tmp);
+               goto again;
+       }
+       
+       /* Success at resizing! */
+       members = map->members;
+
+       map->hashsize = tmp->hashsize;
+       map->members = tmp->members;
+       write_unlock_bh(&set->lock);
+
+       harray_free(members);
+       kfree(tmp);
+
+       return 0;
+}
+
+static inline int
+__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t port,
+       ip_set_ip_t *hash_ip)
+{
+       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
+       ip_set_ip_t id;
+       ip_set_ip_t *elem;
+
+       if (ip < map->first_ip || ip > map->last_ip)
+               return -ERANGE;
+
+       id = hash_id(set, ip, port, hash_ip);
+
+       if (id == UINT_MAX)
+               return -EEXIST;
+               
+       elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
+       *elem = 0;
+
+       return 0;
+}
+
+static int
+delip(struct ip_set *set, const void *data, size_t size,
+        ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_ipporthash *req =
+           (struct ip_set_req_ipporthash *) data;
+
+       if (size != sizeof(struct ip_set_req_ipporthash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_ipporthash),
+                             size);
+               return -EINVAL;
+       }
+       return __delip(set, req->ip, req->port, hash_ip);
+}
+
+static int
+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)
+{
+       ip_set_ip_t port;
+
+       if (flags[index+1] == 0)
+               return -EINVAL;
+               
+       port = get_port(skb, flags[index+1]);
+
+       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
+          flags[index] & IPSET_SRC ? "SRC" : "DST",
+          NIPQUAD(skb->nh.iph->saddr),
+          NIPQUAD(skb->nh.iph->daddr));
+       DP("flag %s port %u",
+          flags[index+1] & IPSET_SRC ? "SRC" : "DST", 
+          port);       
+       if (port == INVALID_PORT)
+               return -EINVAL; 
+
+       return __delip(set,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      port,
+                      hash_ip);
+}
+
+static int create(struct ip_set *set, const void *data, size_t size)
+{
+       struct ip_set_req_ipporthash_create *req =
+           (struct ip_set_req_ipporthash_create *) data;
+       struct ip_set_ipporthash *map;
+       uint16_t i;
+
+       if (size != sizeof(struct ip_set_req_ipporthash_create)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                              sizeof(struct ip_set_req_ipporthash_create),
+                              size);
+               return -EINVAL;
+       }
+
+       if (req->hashsize < 1) {
+               ip_set_printk("hashsize too small");
+               return -ENOEXEC;
+       }
+
+       if (req->probes < 1) {
+               ip_set_printk("probes too small");
+               return -ENOEXEC;
+       }
+
+       map = kmalloc(sizeof(struct ip_set_ipporthash) 
+                     + req->probes * sizeof(uint32_t), GFP_KERNEL);
+       if (!map) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_ipporthash)
+                  + req->probes * sizeof(uint32_t));
+               return -ENOMEM;
+       }
+       for (i = 0; i < req->probes; i++)
+               get_random_bytes(((uint32_t *) map->initval)+i, 4);
+       map->hashsize = req->hashsize;
+       map->probes = req->probes;
+       map->resize = req->resize;
+       map->first_ip = req->from;
+       map->last_ip = req->to;
+       map->members = harray_malloc(map->hashsize, sizeof(ip_set_ip_t), GFP_KERNEL);
+       if (!map->members) {
+               DP("out of memory for %d bytes", map->hashsize * sizeof(ip_set_ip_t));
+               kfree(map);
+               return -ENOMEM;
+       }
+
+       set->data = map;
+       return 0;
+}
+
+static void destroy(struct ip_set *set)
+{
+       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
+
+       harray_free(map->members);
+       kfree(map);
+
+       set->data = NULL;
+}
+
+static void flush(struct ip_set *set)
+{
+       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
+       harray_flush(map->members, map->hashsize, sizeof(ip_set_ip_t));
+}
+
+static void list_header(const struct ip_set *set, void *data)
+{
+       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
+       struct ip_set_req_ipporthash_create *header =
+           (struct ip_set_req_ipporthash_create *) data;
+
+       header->hashsize = map->hashsize;
+       header->probes = map->probes;
+       header->resize = map->resize;
+       header->from = map->first_ip;
+       header->to = map->last_ip;
+}
+
+static int list_members_size(const struct ip_set *set)
+{
+       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
+
+       return (map->hashsize * sizeof(ip_set_ip_t));
+}
+
+static void list_members(const struct ip_set *set, void *data)
+{
+       struct ip_set_ipporthash *map = (struct ip_set_ipporthash *) set->data;
+       ip_set_ip_t i, *elem;
+
+       for (i = 0; i < map->hashsize; i++) {
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);     
+               ((ip_set_ip_t *)data)[i] = *elem;
+       }
+}
+
+static struct ip_set_type ip_set_ipporthash = {
+       .typename               = SETTYPE_NAME,
+       .features               = IPSET_TYPE_IP | IPSET_TYPE_PORT | IPSET_DATA_DOUBLE,
+       .protocol_version       = IP_SET_PROTOCOL_VERSION,
+       .create                 = &create,
+       .destroy                = &destroy,
+       .flush                  = &flush,
+       .reqsize                = sizeof(struct ip_set_req_ipporthash),
+       .addip                  = &addip,
+       .addip_kernel           = &addip_kernel,
+       .retry                  = &retry,
+       .delip                  = &delip,
+       .delip_kernel           = &delip_kernel,
+       .testip                 = &testip,
+       .testip_kernel          = &testip_kernel,
+       .header_size            = sizeof(struct ip_set_req_ipporthash_create),
+       .list_header            = &list_header,
+       .list_members_size      = &list_members_size,
+       .list_members           = &list_members,
+       .me                     = THIS_MODULE,
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("ipporthash type of IP sets");
+
+static int __init init(void)
+{
+       return ip_set_register_set_type(&ip_set_ipporthash);
+}
+
+static void __exit fini(void)
+{
+       /* FIXME: possible race with ip_set_create() */
+       ip_set_unregister_set_type(&ip_set_ipporthash);
+}
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ip_set_iptree.c b/net/ipv4/netfilter/ip_set_iptree.c
new file mode 100644 (file)
index 0000000..0d195e3
--- /dev/null
@@ -0,0 +1,544 @@
+/* Copyright (C) 2005 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module implementing an IP set type: the iptree type */
+
+#include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/skbuff.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+#include <linux/errno.h>
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <linux/spinlock.h>
+
+/* Backward compatibility */
+#ifndef __nocast
+#define __nocast
+#endif
+
+#include <linux/netfilter_ipv4/ip_set_iptree.h>
+
+/* Garbage collection interval in seconds: */
+#define IPTREE_GC_TIME         5*60
+/* Sleep so many milliseconds before trying again 
+ * to delete the gc timer at destroying/flushing a set */ 
+#define IPTREE_DESTROY_SLEEP   100
+
+static kmem_cache_t *branch_cachep;
+static kmem_cache_t *leaf_cachep;
+
+#define ABCD(a,b,c,d,addrp) do {               \
+       a = ((unsigned char *)addrp)[3];        \
+       b = ((unsigned char *)addrp)[2];        \
+       c = ((unsigned char *)addrp)[1];        \
+       d = ((unsigned char *)addrp)[0];        \
+} while (0)
+
+#define TESTIP_WALK(map, elem, branch) do {    \
+       if ((map)->tree[elem]) {                \
+               branch = (map)->tree[elem];     \
+       } else                                  \
+               return 0;                       \
+} while (0)
+
+static inline int
+__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       struct ip_set_iptreeb *btree;
+       struct ip_set_iptreec *ctree;
+       struct ip_set_iptreed *dtree;
+       unsigned char a,b,c,d;
+       
+       *hash_ip = ip;
+       ABCD(a, b, c, d, hash_ip);
+       DP("%u %u %u %u timeout %u", a, b, c, d, map->timeout);
+       TESTIP_WALK(map, a, btree);
+       TESTIP_WALK(btree, b, ctree);
+       TESTIP_WALK(ctree, c, dtree);
+       DP("%lu %lu", dtree->expires[d], jiffies);
+       return !!(map->timeout ? (time_after(dtree->expires[d], jiffies))
+                              : dtree->expires[d]);
+}
+
+static int
+testip(struct ip_set *set, const void *data, size_t size,
+       ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_iptree *req = 
+           (struct ip_set_req_iptree *) data;
+
+       if (size != sizeof(struct ip_set_req_iptree)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_iptree),
+                             size);
+               return -EINVAL;
+       }
+       return __testip(set, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       int res;
+       
+       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
+          flags[index] & IPSET_SRC ? "SRC" : "DST",
+          NIPQUAD(skb->nh.iph->saddr),
+          NIPQUAD(skb->nh.iph->daddr));
+
+       res =  __testip(set,
+                       ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                       hash_ip);
+       return (res < 0 ? 0 : res);
+}
+
+#define ADDIP_WALK(map, elem, branch, type, cachep, flags) do {        \
+       if ((map)->tree[elem]) {                                \
+               DP("found %u", elem);                           \
+               branch = (map)->tree[elem];                     \
+       } else {                                                \
+               branch = (type *)                               \
+                       kmem_cache_alloc(cachep, flags);        \
+               if (branch == NULL)                             \
+                       return -ENOMEM;                         \
+               memset(branch, 0, sizeof(*branch));             \
+               (map)->tree[elem] = branch;                     \
+               DP("alloc %u", elem);                           \
+       }                                                       \
+} while (0)    
+
+static inline int
+__addip(struct ip_set *set, ip_set_ip_t ip, unsigned int timeout,
+       ip_set_ip_t *hash_ip,
+       unsigned int __nocast flags)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       struct ip_set_iptreeb *btree;
+       struct ip_set_iptreec *ctree;
+       struct ip_set_iptreed *dtree;
+       unsigned char a,b,c,d;
+       int ret = 0;
+       
+       *hash_ip = ip;
+       ABCD(a, b, c, d, hash_ip);
+       DP("%u %u %u %u timeout %u", a, b, c, d, timeout);
+       ADDIP_WALK(map, a, btree, struct ip_set_iptreeb, branch_cachep, flags);
+       ADDIP_WALK(btree, b, ctree, struct ip_set_iptreec, branch_cachep, flags);
+       ADDIP_WALK(ctree, c, dtree, struct ip_set_iptreed, leaf_cachep, flags);
+       if (dtree->expires[d]
+           && (!map->timeout || time_after(dtree->expires[d], jiffies)))
+               ret = -EEXIST;
+       dtree->expires[d] = map->timeout ? (timeout * HZ + jiffies) : 1;
+       /* Lottery */
+       if (dtree->expires[d] == 0)
+               dtree->expires[d] = 1;
+       DP("%u %lu", d, dtree->expires[d]);
+       return ret;
+}
+
+static int
+addip(struct ip_set *set, const void *data, size_t size,
+      ip_set_ip_t *hash_ip)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       struct ip_set_req_iptree *req = 
+               (struct ip_set_req_iptree *) data;
+
+       if (size != sizeof(struct ip_set_req_iptree)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_iptree),
+                             size);
+               return -EINVAL;
+       }
+       DP("%u.%u.%u.%u %u", HIPQUAD(req->ip), req->timeout);
+       return __addip(set, req->ip,
+                      req->timeout ? req->timeout : map->timeout,
+                      hash_ip,
+                      GFP_ATOMIC);
+}
+
+static int
+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)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+
+       return __addip(set,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      map->timeout,
+                      hash_ip,
+                      GFP_ATOMIC);
+}
+
+#define DELIP_WALK(map, elem, branch) do {     \
+       if ((map)->tree[elem]) {                \
+               branch = (map)->tree[elem];     \
+       } else                                  \
+               return -EEXIST;                 \
+} while (0)
+
+static inline int 
+__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       struct ip_set_iptreeb *btree;
+       struct ip_set_iptreec *ctree;
+       struct ip_set_iptreed *dtree;
+       unsigned char a,b,c,d;
+       
+       *hash_ip = ip;
+       ABCD(a, b, c, d, hash_ip);
+       DELIP_WALK(map, a, btree);
+       DELIP_WALK(btree, b, ctree);
+       DELIP_WALK(ctree, c, dtree);
+
+       if (dtree->expires[d]) {
+               dtree->expires[d] = 0;
+               return 0;
+       }
+       return -EEXIST;
+}
+
+static int
+delip(struct ip_set *set, const void *data, size_t size,
+      ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_iptree *req =
+           (struct ip_set_req_iptree *) data;
+
+       if (size != sizeof(struct ip_set_req_iptree)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_iptree),
+                             size);
+               return -EINVAL;
+       }
+       return __delip(set, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       return __delip(set,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      hash_ip);
+}
+
+#define LOOP_WALK_BEGIN(map, i, branch) \
+       for (i = 0; i < 256; i++) {     \
+               if (!(map)->tree[i])    \
+                       continue;       \
+               branch = (map)->tree[i]
+
+#define LOOP_WALK_END }
+
+static void ip_tree_gc(unsigned long ul_set)
+{
+       struct ip_set *set = (void *) ul_set;
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       struct ip_set_iptreeb *btree;
+       struct ip_set_iptreec *ctree;
+       struct ip_set_iptreed *dtree;
+       unsigned int a,b,c,d;
+       unsigned char i,j,k;
+
+       i = j = k = 0;
+       DP("gc: %s", set->name);
+       write_lock_bh(&set->lock);
+       LOOP_WALK_BEGIN(map, a, btree);
+       LOOP_WALK_BEGIN(btree, b, ctree);
+       LOOP_WALK_BEGIN(ctree, c, dtree);
+       for (d = 0; d < 256; d++) {
+               if (dtree->expires[d]) {
+                       DP("gc: %u %u %u %u: expires %lu jiffies %lu",
+                           a, b, c, d,
+                           dtree->expires[d], jiffies);
+                       if (map->timeout
+                           && time_before(dtree->expires[d], jiffies))
+                               dtree->expires[d] = 0;
+                       else
+                               k = 1;
+               }
+       }
+       if (k == 0) {
+               DP("gc: %s: leaf %u %u %u empty",
+                   set->name, a, b, c);
+               kmem_cache_free(leaf_cachep, dtree);
+               ctree->tree[c] = NULL;
+       } else {
+               DP("gc: %s: leaf %u %u %u not empty",
+                   set->name, a, b, c);
+               j = 1;
+               k = 0;
+       }
+       LOOP_WALK_END;
+       if (j == 0) {
+               DP("gc: %s: branch %u %u empty",
+                   set->name, a, b);
+               kmem_cache_free(branch_cachep, ctree);
+               btree->tree[b] = NULL;
+       } else {
+               DP("gc: %s: branch %u %u not empty",
+                   set->name, a, b);
+               i = 1;
+               j = k = 0;
+       }
+       LOOP_WALK_END;
+       if (i == 0) {
+               DP("gc: %s: branch %u empty",
+                   set->name, a);
+               kmem_cache_free(branch_cachep, btree);
+               map->tree[a] = NULL;
+       } else {
+               DP("gc: %s: branch %u not empty",
+                   set->name, a);
+               i = j = k = 0;
+       }
+       LOOP_WALK_END;
+       write_unlock_bh(&set->lock);
+       
+       map->gc.expires = jiffies + map->gc_interval * HZ;
+       add_timer(&map->gc);
+}
+
+static inline void init_gc_timer(struct ip_set *set)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+
+       /* Even if there is no timeout for the entries,
+        * we still have to call gc because delete
+        * do not clean up empty branches */
+       map->gc_interval = IPTREE_GC_TIME;
+       init_timer(&map->gc);
+       map->gc.data = (unsigned long) set;
+       map->gc.function = ip_tree_gc;
+       map->gc.expires = jiffies + map->gc_interval * HZ;
+       add_timer(&map->gc);
+}
+
+static int create(struct ip_set *set, const void *data, size_t size)
+{
+       struct ip_set_req_iptree_create *req =
+           (struct ip_set_req_iptree_create *) data;
+       struct ip_set_iptree *map;
+
+       if (size != sizeof(struct ip_set_req_iptree_create)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_iptree_create),
+                             size);
+               return -EINVAL;
+       }
+
+       map = kmalloc(sizeof(struct ip_set_iptree), GFP_KERNEL);
+       if (!map) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_iptree));
+               return -ENOMEM;
+       }
+       memset(map, 0, sizeof(*map));
+       map->timeout = req->timeout;
+       set->data = map;
+
+       init_gc_timer(set);
+
+       return 0;
+}
+
+static void __flush(struct ip_set_iptree *map)
+{
+       struct ip_set_iptreeb *btree;
+       struct ip_set_iptreec *ctree;
+       struct ip_set_iptreed *dtree;
+       unsigned int a,b,c;
+
+       LOOP_WALK_BEGIN(map, a, btree);
+       LOOP_WALK_BEGIN(btree, b, ctree);
+       LOOP_WALK_BEGIN(ctree, c, dtree);
+       kmem_cache_free(leaf_cachep, dtree);
+       LOOP_WALK_END;
+       kmem_cache_free(branch_cachep, ctree);
+       LOOP_WALK_END;
+       kmem_cache_free(branch_cachep, btree);
+       LOOP_WALK_END;
+}
+
+static void destroy(struct ip_set *set)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+
+       /* gc might be running */
+       while (!del_timer(&map->gc))
+               msleep(IPTREE_DESTROY_SLEEP);
+       __flush(map);
+       kfree(map);
+       set->data = NULL;
+}
+
+static void flush(struct ip_set *set)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       unsigned int timeout = map->timeout;
+       
+       /* gc might be running */
+       while (!del_timer(&map->gc))
+               msleep(IPTREE_DESTROY_SLEEP);
+       __flush(map);
+       memset(map, 0, sizeof(*map));
+       map->timeout = timeout;
+
+       init_gc_timer(set);
+}
+
+static void list_header(const struct ip_set *set, void *data)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       struct ip_set_req_iptree_create *header =
+           (struct ip_set_req_iptree_create *) data;
+
+       header->timeout = map->timeout;
+}
+
+static int list_members_size(const struct ip_set *set)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       struct ip_set_iptreeb *btree;
+       struct ip_set_iptreec *ctree;
+       struct ip_set_iptreed *dtree;
+       unsigned int a,b,c,d;
+       unsigned int count = 0;
+
+       LOOP_WALK_BEGIN(map, a, btree);
+       LOOP_WALK_BEGIN(btree, b, ctree);
+       LOOP_WALK_BEGIN(ctree, c, dtree);
+       for (d = 0; d < 256; d++) {
+               if (dtree->expires[d]
+                   && (!map->timeout || time_after(dtree->expires[d], jiffies)))
+                       count++;
+       }
+       LOOP_WALK_END;
+       LOOP_WALK_END;
+       LOOP_WALK_END;
+
+       DP("members %u", count);
+       return (count * sizeof(struct ip_set_req_iptree));
+}
+
+static void list_members(const struct ip_set *set, void *data)
+{
+       struct ip_set_iptree *map = (struct ip_set_iptree *) set->data;
+       struct ip_set_iptreeb *btree;
+       struct ip_set_iptreec *ctree;
+       struct ip_set_iptreed *dtree;
+       unsigned int a,b,c,d;
+       size_t offset = 0;
+       struct ip_set_req_iptree *entry;
+
+       LOOP_WALK_BEGIN(map, a, btree);
+       LOOP_WALK_BEGIN(btree, b, ctree);
+       LOOP_WALK_BEGIN(ctree, c, dtree);
+       for (d = 0; d < 256; d++) {
+               if (dtree->expires[d]
+                   && (!map->timeout || time_after(dtree->expires[d], jiffies))) {
+                       entry = (struct ip_set_req_iptree *)(data + offset);
+                       entry->ip = ((a << 24) | (b << 16) | (c << 8) | d);
+                       entry->timeout = !map->timeout ? 0 
+                               : (dtree->expires[d] - jiffies)/HZ;
+                       offset += sizeof(struct ip_set_req_iptree);
+               }
+       }
+       LOOP_WALK_END;
+       LOOP_WALK_END;
+       LOOP_WALK_END;
+}
+
+static struct ip_set_type ip_set_iptree = {
+       .typename               = SETTYPE_NAME,
+       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
+       .protocol_version       = IP_SET_PROTOCOL_VERSION,
+       .create                 = &create,
+       .destroy                = &destroy,
+       .flush                  = &flush,
+       .reqsize                = sizeof(struct ip_set_req_iptree),
+       .addip                  = &addip,
+       .addip_kernel           = &addip_kernel,
+       .delip                  = &delip,
+       .delip_kernel           = &delip_kernel,
+       .testip                 = &testip,
+       .testip_kernel          = &testip_kernel,
+       .header_size            = sizeof(struct ip_set_req_iptree_create),
+       .list_header            = &list_header,
+       .list_members_size      = &list_members_size,
+       .list_members           = &list_members,
+       .me                     = THIS_MODULE,
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("iptree type of IP sets");
+
+static int __init init(void)
+{
+       int ret;
+       
+       branch_cachep = kmem_cache_create("ip_set_iptreeb",
+                               sizeof(struct ip_set_iptreeb),
+                               0, 0, NULL, NULL);
+       if (!branch_cachep) {
+               printk(KERN_ERR "Unable to create ip_set_iptreeb slab cache\n");
+               ret = -ENOMEM;
+               goto out;
+       }
+       leaf_cachep = kmem_cache_create("ip_set_iptreed",
+                               sizeof(struct ip_set_iptreed),
+                               0, 0, NULL, NULL);
+       if (!leaf_cachep) {
+               printk(KERN_ERR "Unable to create ip_set_iptreed slab cache\n");
+               ret = -ENOMEM;
+               goto free_branch;
+       }
+       ret = ip_set_register_set_type(&ip_set_iptree);
+       if (ret == 0)
+               goto out;
+
+       kmem_cache_destroy(leaf_cachep);
+    free_branch:       
+       kmem_cache_destroy(branch_cachep);
+    out:
+       return ret;
+}
+
+static void __exit fini(void)
+{
+       /* FIXME: possible race with ip_set_create() */
+       ip_set_unregister_set_type(&ip_set_iptree);
+       kmem_cache_destroy(leaf_cachep);
+       kmem_cache_destroy(branch_cachep);
+}
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ip_set_macipmap.c b/net/ipv4/netfilter/ip_set_macipmap.c
new file mode 100644 (file)
index 0000000..0558e8c
--- /dev/null
@@ -0,0 +1,353 @@
+/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
+ *                         Patrick Schaaf <bof@bof.de>
+ *                         Martin Josefsson <gandalf@wlug.westbo.se>
+ * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module implementing an IP set type: the macipmap type */
+
+#include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/skbuff.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+#include <linux/errno.h>
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <linux/spinlock.h>
+#include <linux/if_ether.h>
+#include <linux/vmalloc.h>
+
+#include <linux/netfilter_ipv4/ip_set_malloc.h>
+#include <linux/netfilter_ipv4/ip_set_macipmap.h>
+
+static int
+testip(struct ip_set *set, const void *data, size_t size, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_macipmap *map = (struct ip_set_macipmap *) set->data;
+       struct ip_set_macip *table = (struct ip_set_macip *) map->members;      
+       struct ip_set_req_macipmap *req = (struct ip_set_req_macipmap *) data;
+
+       if (size != sizeof(struct ip_set_req_macipmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_macipmap),
+                             size);
+               return -EINVAL;
+       }
+
+       if (req->ip < map->first_ip || req->ip > map->last_ip)
+               return -ERANGE;
+
+       *hash_ip = req->ip;
+       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
+          set->name, HIPQUAD(req->ip), HIPQUAD(*hash_ip));             
+       if (test_bit(IPSET_MACIP_ISSET,
+                    (void *) &table[req->ip - map->first_ip].flags)) {
+               return (memcmp(req->ethernet,
+                              &table[req->ip - map->first_ip].ethernet,
+                              ETH_ALEN) == 0);
+       } else {
+               return (map->flags & IPSET_MACIP_MATCHUNSET ? 1 : 0);
+       }
+}
+
+static int
+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)
+{
+       struct ip_set_macipmap *map =
+           (struct ip_set_macipmap *) set->data;
+       struct ip_set_macip *table =
+           (struct ip_set_macip *) map->members;
+       ip_set_ip_t ip;
+       
+       ip = ntohl(flags[index] & IPSET_SRC
+                       ? skb->nh.iph->saddr
+                       : skb->nh.iph->daddr);
+       DP("flag: %s src: %u.%u.%u.%u dst: %u.%u.%u.%u",
+          flags[index] & IPSET_SRC ? "SRC" : "DST",
+          NIPQUAD(skb->nh.iph->saddr),
+          NIPQUAD(skb->nh.iph->daddr));
+
+       if (ip < map->first_ip || ip > map->last_ip)
+               return 0;
+
+       *hash_ip = ip;  
+       DP("set: %s, ip:%u.%u.%u.%u, %u.%u.%u.%u",
+          set->name, HIPQUAD(ip), HIPQUAD(*hash_ip));          
+       if (test_bit(IPSET_MACIP_ISSET,
+           (void *) &table[ip - map->first_ip].flags)) {
+               /* Is mac pointer valid?
+                * If so, compare... */
+               return (skb->mac.raw >= skb->head
+                       && (skb->mac.raw + ETH_HLEN) <= skb->data
+                       && (memcmp(eth_hdr(skb)->h_source,
+                                  &table[ip - map->first_ip].ethernet,
+                                  ETH_ALEN) == 0));
+       } else {
+               return (map->flags & IPSET_MACIP_MATCHUNSET ? 1 : 0);
+       }
+}
+
+/* returns 0 on success */
+static inline int
+__addip(struct ip_set *set, 
+       ip_set_ip_t ip, unsigned char *ethernet, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_macipmap *map =
+           (struct ip_set_macipmap *) set->data;
+       struct ip_set_macip *table =
+           (struct ip_set_macip *) map->members;
+
+       if (ip < map->first_ip || ip > map->last_ip)
+               return -ERANGE;
+       if (test_and_set_bit(IPSET_MACIP_ISSET, 
+                            (void *) &table[ip - map->first_ip].flags))
+               return -EEXIST;
+
+       *hash_ip = ip;
+       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
+       memcpy(&table[ip - map->first_ip].ethernet, ethernet, ETH_ALEN);
+       return 0;
+}
+
+static int
+addip(struct ip_set *set, const void *data, size_t size,
+      ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_macipmap *req =
+           (struct ip_set_req_macipmap *) data;
+
+       if (size != sizeof(struct ip_set_req_macipmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_macipmap),
+                             size);
+               return -EINVAL;
+       }
+       return __addip(set, req->ip, req->ethernet, hash_ip);
+}
+
+static int
+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)
+{
+       ip_set_ip_t ip;
+       
+       ip = ntohl(flags[index] & IPSET_SRC
+                       ? skb->nh.iph->saddr
+                       : skb->nh.iph->daddr);
+
+       if (!(skb->mac.raw >= skb->head
+             && (skb->mac.raw + ETH_HLEN) <= skb->data))
+               return -EINVAL;
+
+       return __addip(set, ip, eth_hdr(skb)->h_source, hash_ip);
+}
+
+static inline int
+__delip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_macipmap *map =
+           (struct ip_set_macipmap *) set->data;
+       struct ip_set_macip *table =
+           (struct ip_set_macip *) map->members;
+
+       if (ip < map->first_ip || ip > map->last_ip)
+               return -ERANGE;
+       if (!test_and_clear_bit(IPSET_MACIP_ISSET, 
+                               (void *)&table[ip - map->first_ip].flags))
+               return -EEXIST;
+
+       *hash_ip = ip;
+       DP("%u.%u.%u.%u, %u.%u.%u.%u", HIPQUAD(ip), HIPQUAD(*hash_ip));
+       return 0;
+}
+
+static int
+delip(struct ip_set *set, const void *data, size_t size,
+     ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_macipmap *req =
+           (struct ip_set_req_macipmap *) data;
+
+       if (size != sizeof(struct ip_set_req_macipmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_macipmap),
+                             size);
+               return -EINVAL;
+       }
+       return __delip(set, req->ip, hash_ip);
+}
+
+static int
+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)
+{
+       return __delip(set,
+                      ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                      hash_ip);
+}
+
+static inline size_t members_size(ip_set_id_t from, ip_set_id_t to)
+{
+       return (size_t)((to - from + 1) * sizeof(struct ip_set_macip));
+}
+
+static int create(struct ip_set *set, const void *data, size_t size)
+{
+       int newbytes;
+       struct ip_set_req_macipmap_create *req =
+           (struct ip_set_req_macipmap_create *) data;
+       struct ip_set_macipmap *map;
+
+       if (size != sizeof(struct ip_set_req_macipmap_create)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_macipmap_create),
+                             size);
+               return -EINVAL;
+       }
+
+       DP("from %u.%u.%u.%u to %u.%u.%u.%u",
+          HIPQUAD(req->from), HIPQUAD(req->to));
+
+       if (req->from > req->to) {
+               DP("bad ip range");
+               return -ENOEXEC;
+       }
+
+       if (req->to - req->from > MAX_RANGE) {
+               ip_set_printk("range too big (max %d addresses)",
+                              MAX_RANGE+1);
+               return -ENOEXEC;
+       }
+
+       map = kmalloc(sizeof(struct ip_set_macipmap), GFP_KERNEL);
+       if (!map) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_macipmap));
+               return -ENOMEM;
+       }
+       map->flags = req->flags;
+       map->first_ip = req->from;
+       map->last_ip = req->to;
+       newbytes = members_size(map->first_ip, map->last_ip);
+       map->members = ip_set_malloc(newbytes);
+       DP("members: %u %p", newbytes, map->members);
+       if (!map->members) {
+               DP("out of memory for %d bytes", newbytes);
+               kfree(map);
+               return -ENOMEM;
+       }
+       memset(map->members, 0, newbytes);
+       
+       set->data = map;
+       return 0;
+}
+
+static void destroy(struct ip_set *set)
+{
+       struct ip_set_macipmap *map =
+           (struct ip_set_macipmap *) set->data;
+
+       ip_set_free(map->members, members_size(map->first_ip, map->last_ip));
+       kfree(map);
+
+       set->data = NULL;
+}
+
+static void flush(struct ip_set *set)
+{
+       struct ip_set_macipmap *map =
+           (struct ip_set_macipmap *) set->data;
+       memset(map->members, 0, members_size(map->first_ip, map->last_ip));
+}
+
+static void list_header(const struct ip_set *set, void *data)
+{
+       struct ip_set_macipmap *map =
+           (struct ip_set_macipmap *) set->data;
+       struct ip_set_req_macipmap_create *header =
+           (struct ip_set_req_macipmap_create *) data;
+
+       DP("list_header %x %x %u", map->first_ip, map->last_ip,
+          map->flags);
+
+       header->from = map->first_ip;
+       header->to = map->last_ip;
+       header->flags = map->flags;
+}
+
+static int list_members_size(const struct ip_set *set)
+{
+       struct ip_set_macipmap *map =
+           (struct ip_set_macipmap *) set->data;
+
+       DP("%u", members_size(map->first_ip, map->last_ip));
+       return members_size(map->first_ip, map->last_ip);
+}
+
+static void list_members(const struct ip_set *set, void *data)
+{
+       struct ip_set_macipmap *map =
+           (struct ip_set_macipmap *) set->data;
+
+       int bytes = members_size(map->first_ip, map->last_ip);
+
+       DP("members: %u %p", bytes, map->members);
+       memcpy(data, map->members, bytes);
+}
+
+static struct ip_set_type ip_set_macipmap = {
+       .typename               = SETTYPE_NAME,
+       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
+       .protocol_version       = IP_SET_PROTOCOL_VERSION,
+       .create                 = &create,
+       .destroy                = &destroy,
+       .flush                  = &flush,
+       .reqsize                = sizeof(struct ip_set_req_macipmap),
+       .addip                  = &addip,
+       .addip_kernel           = &addip_kernel,
+       .delip                  = &delip,
+       .delip_kernel           = &delip_kernel,
+       .testip                 = &testip,
+       .testip_kernel          = &testip_kernel,
+       .header_size            = sizeof(struct ip_set_req_macipmap_create),
+       .list_header            = &list_header,
+       .list_members_size      = &list_members_size,
+       .list_members           = &list_members,
+       .me                     = THIS_MODULE,
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("macipmap type of IP sets");
+
+static int __init init(void)
+{
+       init_max_malloc_size();
+       return ip_set_register_set_type(&ip_set_macipmap);
+}
+
+static void __exit fini(void)
+{
+       /* FIXME: possible race with ip_set_create() */
+       ip_set_unregister_set_type(&ip_set_macipmap);
+}
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ip_set_nethash.c b/net/ipv4/netfilter/ip_set_nethash.c
new file mode 100644 (file)
index 0000000..b78988f
--- /dev/null
@@ -0,0 +1,466 @@
+/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module implementing a cidr nethash set */
+
+#include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/skbuff.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+#include <linux/errno.h>
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <linux/spinlock.h>
+#include <linux/vmalloc.h>
+#include <linux/random.h>
+
+#include <net/ip.h>
+
+#include <linux/netfilter_ipv4/ip_set_malloc.h>
+#include <linux/netfilter_ipv4/ip_set_nethash.h>
+#include <linux/netfilter_ipv4/ip_set_jhash.h>
+
+static inline __u32
+jhash_ip(const struct ip_set_nethash *map, uint16_t i, ip_set_ip_t ip)
+{
+       return jhash_1word(ip, *(((uint32_t *) map->initval) + i));
+}
+
+static inline __u32
+hash_id_cidr(struct ip_set_nethash *map,
+            ip_set_ip_t ip,
+            unsigned char cidr,
+            ip_set_ip_t *hash_ip)
+{
+       __u32 id;
+       u_int16_t i;
+       ip_set_ip_t *elem;
+
+       *hash_ip = pack(ip, cidr);
+       
+       for (i = 0; i < map->probes; i++) {
+               id = jhash_ip(map, i, *hash_ip) % map->hashsize;
+               DP("hash key: %u", id);
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
+               if (*elem == *hash_ip)
+                       return id;
+       }
+       return UINT_MAX;
+}
+
+static inline __u32
+hash_id(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+       __u32 id = UINT_MAX;
+       int i;
+
+       for (i = 0; i < 30 && map->cidr[i]; i++) {
+               id = hash_id_cidr(map, ip, map->cidr[i], hash_ip);
+               if (id != UINT_MAX)
+                       break;
+       }
+       return id;
+}
+
+static inline int
+__testip_cidr(struct ip_set *set, ip_set_ip_t ip, unsigned char cidr,
+             ip_set_ip_t *hash_ip)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+
+       return (hash_id_cidr(map, ip, cidr, hash_ip) != UINT_MAX);
+}
+
+static inline int
+__testip(struct ip_set *set, ip_set_ip_t ip, ip_set_ip_t *hash_ip)
+{
+       return (hash_id(set, ip, hash_ip) != UINT_MAX);
+}
+
+static int
+testip(struct ip_set *set, const void *data, size_t size,
+       ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_nethash *req = 
+           (struct ip_set_req_nethash *) data;
+
+       if (size != sizeof(struct ip_set_req_nethash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_nethash),
+                             size);
+               return -EINVAL;
+       }
+       return (req->cidr == 32 ? __testip(set, req->ip, hash_ip)
+               : __testip_cidr(set, req->ip, req->cidr, hash_ip));
+}
+
+static int
+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)
+{
+       return __testip(set,
+                       ntohl(flags[index] & IPSET_SRC 
+                               ? skb->nh.iph->saddr 
+                               : skb->nh.iph->daddr),
+                       hash_ip);
+}
+
+static inline int
+__addip_base(struct ip_set_nethash *map, ip_set_ip_t ip)
+{
+       __u32 probe;
+       u_int16_t i;
+       ip_set_ip_t *elem;
+       
+       for (i = 0; i < map->probes; i++) {
+               probe = jhash_ip(map, i, ip) % map->hashsize;
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, probe);
+               if (*elem == ip)
+                       return -EEXIST;
+               if (!*elem) {
+                       *elem = ip;
+                       return 0;
+               }
+       }
+       /* Trigger rehashing */
+       return -EAGAIN;
+}
+
+static inline int
+__addip(struct ip_set_nethash *map, ip_set_ip_t ip, unsigned char cidr,
+       ip_set_ip_t *hash_ip)
+{
+       *hash_ip = pack(ip, cidr);
+       DP("%u.%u.%u.%u/%u, %u.%u.%u.%u", HIPQUAD(ip), cidr, HIPQUAD(*hash_ip));
+       
+       return __addip_base(map, *hash_ip);
+}
+
+static void
+update_cidr_sizes(struct ip_set_nethash *map, unsigned char cidr)
+{
+       unsigned char next;
+       int i;
+       
+       for (i = 0; i < 30 && map->cidr[i]; i++) {
+               if (map->cidr[i] == cidr) {
+                       return;
+               } else if (map->cidr[i] < cidr) {
+                       next = map->cidr[i];
+                       map->cidr[i] = cidr;
+                       cidr = next;
+               }
+       }
+       if (i < 30)
+               map->cidr[i] = cidr;
+}
+
+static int
+addip(struct ip_set *set, const void *data, size_t size,
+        ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_nethash *req = 
+           (struct ip_set_req_nethash *) data;
+       int ret;
+
+       if (size != sizeof(struct ip_set_req_nethash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_nethash),
+                             size);
+               return -EINVAL;
+       }
+       ret = __addip((struct ip_set_nethash *) set->data, 
+                     req->ip, req->cidr, hash_ip);
+       
+       if (ret == 0)
+               update_cidr_sizes((struct ip_set_nethash *) set->data,
+                                 req->cidr);
+       
+       return ret;
+}
+
+static int
+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)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+       int ret = -ERANGE;
+       ip_set_ip_t ip = ntohl(flags[index] & IPSET_SRC 
+                                       ? skb->nh.iph->saddr
+                                       : skb->nh.iph->daddr);
+       
+       if (map->cidr[0])
+               ret = __addip(map, ip, map->cidr[0], hash_ip);
+               
+       return ret;
+}
+
+static int retry(struct ip_set *set)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+       ip_set_ip_t *elem;
+       void *members;
+       u_int32_t i, hashsize = map->hashsize;
+       int res;
+       struct ip_set_nethash *tmp;
+       
+       if (map->resize == 0)
+               return -ERANGE;
+
+    again:
+       res = 0;
+       
+       /* Calculate new parameters */
+       hashsize += (hashsize * map->resize)/100;
+       if (hashsize == map->hashsize)
+               hashsize++;
+       
+       ip_set_printk("rehashing of set %s triggered: "
+                     "hashsize grows from %u to %u",
+                     set->name, map->hashsize, hashsize);
+
+       tmp = kmalloc(sizeof(struct ip_set_nethash) 
+                     + map->probes * sizeof(uint32_t), GFP_ATOMIC);
+       if (!tmp) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_nethash)
+                  + map->probes * sizeof(uint32_t));
+               return -ENOMEM;
+       }
+       tmp->members = harray_malloc(hashsize, sizeof(ip_set_ip_t), GFP_ATOMIC);
+       if (!tmp->members) {
+               DP("out of memory for %d bytes", hashsize * sizeof(ip_set_ip_t));
+               kfree(tmp);
+               return -ENOMEM;
+       }
+       tmp->hashsize = hashsize;
+       tmp->probes = map->probes;
+       tmp->resize = map->resize;
+       memcpy(tmp->initval, map->initval, map->probes * sizeof(uint32_t));
+       memcpy(tmp->cidr, map->cidr, 30 * sizeof(unsigned char));
+       
+       write_lock_bh(&set->lock);
+       map = (struct ip_set_nethash *) set->data; /* Play safe */
+       for (i = 0; i < map->hashsize && res == 0; i++) {
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);     
+               if (*elem)
+                       res = __addip_base(tmp, *elem);
+       }
+       if (res) {
+               /* Failure, try again */
+               write_unlock_bh(&set->lock);
+               harray_free(tmp->members);
+               kfree(tmp);
+               goto again;
+       }
+       
+       /* Success at resizing! */
+       members = map->members;
+       
+       map->hashsize = tmp->hashsize;
+       map->members = tmp->members;
+       write_unlock_bh(&set->lock);
+
+       harray_free(members);
+       kfree(tmp);
+
+       return 0;
+}
+
+static inline int
+__delip(struct ip_set_nethash *map, ip_set_ip_t ip, unsigned char cidr,
+       ip_set_ip_t *hash_ip)
+{
+       ip_set_ip_t id = hash_id_cidr(map, ip, cidr, hash_ip);
+       ip_set_ip_t *elem;
+
+       if (id == UINT_MAX)
+               return -EEXIST;
+               
+       elem = HARRAY_ELEM(map->members, ip_set_ip_t *, id);
+       *elem = 0;
+       return 0;
+}
+
+static int
+delip(struct ip_set *set, const void *data, size_t size,
+        ip_set_ip_t *hash_ip)
+{
+       struct ip_set_req_nethash *req =
+           (struct ip_set_req_nethash *) data;
+
+       if (size != sizeof(struct ip_set_req_nethash)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_nethash),
+                             size);
+               return -EINVAL;
+       }
+       /* TODO: no garbage collection in map->cidr */          
+       return __delip((struct ip_set_nethash *) set->data, 
+                      req->ip, req->cidr, hash_ip);
+}
+
+static int
+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)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+       int ret = -ERANGE;
+       ip_set_ip_t ip = ntohl(flags[index] & IPSET_SRC 
+                                       ? skb->nh.iph->saddr
+                                       : skb->nh.iph->daddr);
+       
+       if (map->cidr[0])
+               ret = __delip(map, ip, map->cidr[0], hash_ip);
+       
+       return ret;
+}
+
+static int create(struct ip_set *set, const void *data, size_t size)
+{
+       struct ip_set_req_nethash_create *req =
+           (struct ip_set_req_nethash_create *) data;
+       struct ip_set_nethash *map;
+       uint16_t i;
+
+       if (size != sizeof(struct ip_set_req_nethash_create)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                              sizeof(struct ip_set_req_nethash_create),
+                              size);
+               return -EINVAL;
+       }
+
+       if (req->hashsize < 1) {
+               ip_set_printk("hashsize too small");
+               return -ENOEXEC;
+       }
+       if (req->probes < 1) {
+               ip_set_printk("probes too small");
+               return -ENOEXEC;
+       }
+
+       map = kmalloc(sizeof(struct ip_set_nethash)
+                     + req->probes * sizeof(uint32_t), GFP_KERNEL);
+       if (!map) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_nethash)
+                  + req->probes * sizeof(uint32_t));
+               return -ENOMEM;
+       }
+       for (i = 0; i < req->probes; i++)
+               get_random_bytes(((uint32_t *) map->initval)+i, 4);
+       map->hashsize = req->hashsize;
+       map->probes = req->probes;
+       map->resize = req->resize;
+       memset(map->cidr, 0, 30 * sizeof(unsigned char));
+       map->members = harray_malloc(map->hashsize, sizeof(ip_set_ip_t), GFP_KERNEL);
+       if (!map->members) {
+               DP("out of memory for %d bytes", map->hashsize * sizeof(ip_set_ip_t));
+               kfree(map);
+               return -ENOMEM;
+       }
+       
+       set->data = map;
+       return 0;
+}
+
+static void destroy(struct ip_set *set)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+
+       harray_free(map->members);
+       kfree(map);
+
+       set->data = NULL;
+}
+
+static void flush(struct ip_set *set)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+       harray_flush(map->members, map->hashsize, sizeof(ip_set_ip_t));
+       memset(map->cidr, 0, 30 * sizeof(unsigned char));
+}
+
+static void list_header(const struct ip_set *set, void *data)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+       struct ip_set_req_nethash_create *header =
+           (struct ip_set_req_nethash_create *) data;
+
+       header->hashsize = map->hashsize;
+       header->probes = map->probes;
+       header->resize = map->resize;
+}
+
+static int list_members_size(const struct ip_set *set)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+
+       return (map->hashsize * sizeof(ip_set_ip_t));
+}
+
+static void list_members(const struct ip_set *set, void *data)
+{
+       struct ip_set_nethash *map = (struct ip_set_nethash *) set->data;
+       ip_set_ip_t i, *elem;
+
+       for (i = 0; i < map->hashsize; i++) {
+               elem = HARRAY_ELEM(map->members, ip_set_ip_t *, i);     
+               ((ip_set_ip_t *)data)[i] = *elem;
+       }
+}
+
+static struct ip_set_type ip_set_nethash = {
+       .typename               = SETTYPE_NAME,
+       .features               = IPSET_TYPE_IP | IPSET_DATA_SINGLE,
+       .protocol_version       = IP_SET_PROTOCOL_VERSION,
+       .create                 = &create,
+       .destroy                = &destroy,
+       .flush                  = &flush,
+       .reqsize                = sizeof(struct ip_set_req_nethash),
+       .addip                  = &addip,
+       .addip_kernel           = &addip_kernel,
+       .retry                  = &retry,
+       .delip                  = &delip,
+       .delip_kernel           = &delip_kernel,
+       .testip                 = &testip,
+       .testip_kernel          = &testip_kernel,
+       .header_size            = sizeof(struct ip_set_req_nethash_create),
+       .list_header            = &list_header,
+       .list_members_size      = &list_members_size,
+       .list_members           = &list_members,
+       .me                     = THIS_MODULE,
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("nethash type of IP sets");
+
+static int __init init(void)
+{
+       return ip_set_register_set_type(&ip_set_nethash);
+}
+
+static void __exit fini(void)
+{
+       /* FIXME: possible race with ip_set_create() */
+       ip_set_unregister_set_type(&ip_set_nethash);
+}
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ip_set_portmap.c b/net/ipv4/netfilter/ip_set_portmap.c
new file mode 100644 (file)
index 0000000..5816e9a
--- /dev/null
@@ -0,0 +1,334 @@
+/* Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module implementing a port set type as a bitmap */
+
+#include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/udp.h>
+#include <linux/skbuff.h>
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+#include <linux/errno.h>
+#include <asm/uaccess.h>
+#include <asm/bitops.h>
+#include <linux/spinlock.h>
+
+#include <net/ip.h>
+
+#include <linux/netfilter_ipv4/ip_set_portmap.h>
+
+/* We must handle non-linear skbs */
+static inline ip_set_ip_t
+get_port(const struct sk_buff *skb, u_int32_t flags)
+{
+       struct iphdr *iph = skb->nh.iph;
+       u_int16_t offset = ntohs(iph->frag_off) & IP_OFFSET;
+
+       switch (iph->protocol) {
+       case IPPROTO_TCP: {
+               struct tcphdr tcph;
+               
+               /* See comments at tcp_match in ip_tables.c */
+               if (offset)
+                       return INVALID_PORT;
+
+               if (skb_copy_bits(skb, skb->nh.iph->ihl*4, &tcph, sizeof(tcph)) < 0)
+                       /* No choice either */
+                       return INVALID_PORT;
+               
+               return ntohs(flags & IPSET_SRC ?
+                            tcph.source : tcph.dest);
+           }
+       case IPPROTO_UDP: {
+               struct udphdr udph;
+
+               if (offset)
+                       return INVALID_PORT;
+
+               if (skb_copy_bits(skb, skb->nh.iph->ihl*4, &udph, sizeof(udph)) < 0)
+                       /* No choice either */
+                       return INVALID_PORT;
+               
+               return ntohs(flags & IPSET_SRC ?
+                            udph.source : udph.dest);
+           }
+       default:
+               return INVALID_PORT;
+       }
+}
+
+static inline int
+__testport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
+{
+       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
+
+       if (port < map->first_port || port > map->last_port)
+               return -ERANGE;
+               
+       *hash_port = port;
+       DP("set: %s, port:%u, %u", set->name, port, *hash_port);
+       return !!test_bit(port - map->first_port, map->members);
+}
+
+static int
+testport(struct ip_set *set, const void *data, size_t size,
+         ip_set_ip_t *hash_port)
+{
+       struct ip_set_req_portmap *req = 
+           (struct ip_set_req_portmap *) data;
+
+       if (size != sizeof(struct ip_set_req_portmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_portmap),
+                             size);
+               return -EINVAL;
+       }
+       return __testport(set, req->port, hash_port);
+}
+
+static int
+testport_kernel(struct ip_set *set, 
+               const struct sk_buff *skb,
+               ip_set_ip_t *hash_port,
+               const u_int32_t *flags,
+               unsigned char index)
+{
+       int res;
+       ip_set_ip_t port = get_port(skb, flags[index]);
+
+       DP("flag %s port %u", flags[index] & IPSET_SRC ? "SRC" : "DST", port);  
+       if (port == INVALID_PORT)
+               return 0;       
+
+       res =  __testport(set, port, hash_port);
+       
+       return (res < 0 ? 0 : res);
+}
+
+static inline int
+__addport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
+{
+       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
+
+       if (port < map->first_port || port > map->last_port)
+               return -ERANGE;
+       if (test_and_set_bit(port - map->first_port, map->members))
+               return -EEXIST;
+               
+       *hash_port = port;
+       DP("port %u", port);
+       return 0;
+}
+
+static int
+addport(struct ip_set *set, const void *data, size_t size,
+        ip_set_ip_t *hash_port)
+{
+       struct ip_set_req_portmap *req = 
+           (struct ip_set_req_portmap *) data;
+
+       if (size != sizeof(struct ip_set_req_portmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_portmap),
+                             size);
+               return -EINVAL;
+       }
+       return __addport(set, req->port, hash_port);
+}
+
+static int
+addport_kernel(struct ip_set *set, 
+              const struct sk_buff *skb,
+              ip_set_ip_t *hash_port,
+              const u_int32_t *flags,
+              unsigned char index)
+{
+       ip_set_ip_t port = get_port(skb, flags[index]);
+       
+       if (port == INVALID_PORT)
+               return -EINVAL;
+
+       return __addport(set, port, hash_port);
+}
+
+static inline int
+__delport(struct ip_set *set, ip_set_ip_t port, ip_set_ip_t *hash_port)
+{
+       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
+
+       if (port < map->first_port || port > map->last_port)
+               return -ERANGE;
+       if (!test_and_clear_bit(port - map->first_port, map->members))
+               return -EEXIST;
+               
+       *hash_port = port;
+       DP("port %u", port);
+       return 0;
+}
+
+static int
+delport(struct ip_set *set, const void *data, size_t size,
+        ip_set_ip_t *hash_port)
+{
+       struct ip_set_req_portmap *req =
+           (struct ip_set_req_portmap *) data;
+
+       if (size != sizeof(struct ip_set_req_portmap)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                             sizeof(struct ip_set_req_portmap),
+                             size);
+               return -EINVAL;
+       }
+       return __delport(set, req->port, hash_port);
+}
+
+static int
+delport_kernel(struct ip_set *set, 
+              const struct sk_buff *skb,
+              ip_set_ip_t *hash_port,
+              const u_int32_t *flags,
+              unsigned char index)
+{
+       ip_set_ip_t port = get_port(skb, flags[index]);
+       
+       if (port == INVALID_PORT)
+               return -EINVAL;
+
+       return __delport(set, port, hash_port);
+}
+
+static int create(struct ip_set *set, const void *data, size_t size)
+{
+       int newbytes;
+       struct ip_set_req_portmap_create *req =
+           (struct ip_set_req_portmap_create *) data;
+       struct ip_set_portmap *map;
+
+       if (size != sizeof(struct ip_set_req_portmap_create)) {
+               ip_set_printk("data length wrong (want %zu, have %zu)",
+                              sizeof(struct ip_set_req_portmap_create),
+                              size);
+               return -EINVAL;
+       }
+
+       DP("from %u to %u", req->from, req->to);
+
+       if (req->from > req->to) {
+               DP("bad port range");
+               return -ENOEXEC;
+       }
+
+       if (req->to - req->from > MAX_RANGE) {
+               ip_set_printk("range too big (max %d ports)",
+                              MAX_RANGE+1);
+               return -ENOEXEC;
+       }
+
+       map = kmalloc(sizeof(struct ip_set_portmap), GFP_KERNEL);
+       if (!map) {
+               DP("out of memory for %d bytes",
+                  sizeof(struct ip_set_portmap));
+               return -ENOMEM;
+       }
+       map->first_port = req->from;
+       map->last_port = req->to;
+       newbytes = bitmap_bytes(req->from, req->to);
+       map->members = kmalloc(newbytes, GFP_KERNEL);
+       if (!map->members) {
+               DP("out of memory for %d bytes", newbytes);
+               kfree(map);
+               return -ENOMEM;
+       }
+       memset(map->members, 0, newbytes);
+
+       set->data = map;
+       return 0;
+}
+
+static void destroy(struct ip_set *set)
+{
+       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
+
+       kfree(map->members);
+       kfree(map);
+
+       set->data = NULL;
+}
+
+static void flush(struct ip_set *set)
+{
+       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
+       memset(map->members, 0, bitmap_bytes(map->first_port, map->last_port));
+}
+
+static void list_header(const struct ip_set *set, void *data)
+{
+       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
+       struct ip_set_req_portmap_create *header =
+           (struct ip_set_req_portmap_create *) data;
+
+       DP("list_header %u %u", map->first_port, map->last_port);
+
+       header->from = map->first_port;
+       header->to = map->last_port;
+}
+
+static int list_members_size(const struct ip_set *set)
+{
+       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
+
+       return bitmap_bytes(map->first_port, map->last_port);
+}
+
+static void list_members(const struct ip_set *set, void *data)
+{
+       struct ip_set_portmap *map = (struct ip_set_portmap *) set->data;
+       int bytes = bitmap_bytes(map->first_port, map->last_port);
+
+       memcpy(data, map->members, bytes);
+}
+
+static struct ip_set_type ip_set_portmap = {
+       .typename               = SETTYPE_NAME,
+       .features               = IPSET_TYPE_PORT | IPSET_DATA_SINGLE,
+       .protocol_version       = IP_SET_PROTOCOL_VERSION,
+       .create                 = &create,
+       .destroy                = &destroy,
+       .flush                  = &flush,
+       .reqsize                = sizeof(struct ip_set_req_portmap),
+       .addip                  = &addport,
+       .addip_kernel           = &addport_kernel,
+       .delip                  = &delport,
+       .delip_kernel           = &delport_kernel,
+       .testip                 = &testport,
+       .testip_kernel          = &testport_kernel,
+       .header_size            = sizeof(struct ip_set_req_portmap_create),
+       .list_header            = &list_header,
+       .list_members_size      = &list_members_size,
+       .list_members           = &list_members,
+       .me                     = THIS_MODULE,
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("portmap type of IP sets");
+
+static int __init init(void)
+{
+       return ip_set_register_set_type(&ip_set_portmap);
+}
+
+static void __exit fini(void)
+{
+       /* FIXME: possible race with ip_set_create() */
+       ip_set_unregister_set_type(&ip_set_portmap);
+}
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ipt_SET.c b/net/ipv4/netfilter/ipt_SET.c
new file mode 100644 (file)
index 0000000..37e18ec
--- /dev/null
@@ -0,0 +1,128 @@
+/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
+ *                         Patrick Schaaf <bof@bof.de>
+ *                         Martin Josefsson <gandalf@wlug.westbo.se>
+ * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* ipt_SET.c - netfilter target to manipulate IP sets */
+
+#include <linux/types.h>
+#include <linux/ip.h>
+#include <linux/timer.h>
+#include <linux/module.h>
+#include <linux/netfilter.h>
+#include <linux/netdevice.h>
+#include <linux/if.h>
+#include <linux/inetdevice.h>
+#include <net/protocol.h>
+#include <net/checksum.h>
+#include <linux/netfilter_ipv4.h>
+#include <linux/netfilter_ipv4/ip_nat_rule.h>
+#include <linux/netfilter_ipv4/ipt_set.h>
+
+static unsigned int
+target(struct sk_buff **pskb,
+       const struct net_device *in,
+       const struct net_device *out,
+       unsigned int hooknum,
+       const void *targinfo,
+       void *userinfo)
+{
+       const struct ipt_set_info_target *info = targinfo;
+       
+       if (info->add_set.index != IP_SET_INVALID_ID)
+               ip_set_addip_kernel(info->add_set.index,
+                                   *pskb,
+                                   info->add_set.flags);
+       if (info->del_set.index != IP_SET_INVALID_ID)
+               ip_set_delip_kernel(info->del_set.index,
+                                   *pskb,
+                                   info->del_set.flags);
+
+       return IPT_CONTINUE;
+}
+
+static int
+checkentry(const char *tablename,
+          const struct ipt_entry *e,
+          void *targinfo,
+          unsigned int targinfosize, unsigned int hook_mask)
+{
+       struct ipt_set_info_target *info = 
+               (struct ipt_set_info_target *) targinfo;
+       ip_set_id_t index;
+
+       if (targinfosize != IPT_ALIGN(sizeof(*info))) {
+               DP("bad target info size %u", targinfosize);
+               return 0;
+       }
+
+       if (info->add_set.index != IP_SET_INVALID_ID) {
+               index = ip_set_get_byindex(info->add_set.index);
+               if (index == IP_SET_INVALID_ID) {
+                       ip_set_printk("cannot find add_set index %u as target",
+                                     info->add_set.index);
+                       return 0;       /* error */
+               }
+       }
+
+       if (info->del_set.index != IP_SET_INVALID_ID) {
+               index = ip_set_get_byindex(info->del_set.index);
+               if (index == IP_SET_INVALID_ID) {
+                       ip_set_printk("cannot find del_set index %u as target",
+                                     info->del_set.index);
+                       return 0;       /* error */
+               }
+       }
+       if (info->add_set.flags[IP_SET_MAX_BINDINGS] != 0
+           || info->del_set.flags[IP_SET_MAX_BINDINGS] != 0) {
+               ip_set_printk("That's nasty!");
+               return 0;       /* error */
+       }
+
+       return 1;
+}
+
+static void destroy(void *targetinfo, unsigned int targetsize)
+{
+       struct ipt_set_info_target *info = targetinfo;
+
+       if (targetsize != IPT_ALIGN(sizeof(struct ipt_set_info_target))) {
+               ip_set_printk("invalid targetsize %d", targetsize);
+               return;
+       }
+
+       if (info->add_set.index != IP_SET_INVALID_ID)
+               ip_set_put(info->add_set.index);
+       if (info->del_set.index != IP_SET_INVALID_ID)
+               ip_set_put(info->del_set.index);
+}
+
+static struct ipt_target SET_target = {
+       .name           = "SET",
+       .target         = target,
+       .checkentry     = checkentry,
+       .destroy        = destroy,
+       .me             = THIS_MODULE
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("iptables IP set target module");
+
+static int __init init(void)
+{
+       return ipt_register_target(&SET_target);
+}
+
+static void __exit fini(void)
+{
+       ipt_unregister_target(&SET_target);
+}
+
+module_init(init);
+module_exit(fini);
diff --git a/net/ipv4/netfilter/ipt_set.c b/net/ipv4/netfilter/ipt_set.c
new file mode 100644 (file)
index 0000000..9e36c81
--- /dev/null
@@ -0,0 +1,112 @@
+/* Copyright (C) 2000-2002 Joakim Axelsson <gozem@linux.nu>
+ *                         Patrick Schaaf <bof@bof.de>
+ *                         Martin Josefsson <gandalf@wlug.westbo.se>
+ * Copyright (C) 2003-2004 Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.  
+ */
+
+/* Kernel module to match an IP set. */
+
+#include <linux/module.h>
+#include <linux/ip.h>
+#include <linux/skbuff.h>
+
+#include <linux/netfilter_ipv4/ip_tables.h>
+#include <linux/netfilter_ipv4/ip_set.h>
+#include <linux/netfilter_ipv4/ipt_set.h>
+
+static inline int
+match_set(const struct ipt_set_info *info,
+         const struct sk_buff *skb,
+         int inv)
+{      
+       if (ip_set_testip_kernel(info->index, skb, info->flags))
+               inv = !inv;
+       return inv;
+}
+
+static int
+match(const struct sk_buff *skb,
+      const struct net_device *in,
+      const struct net_device *out,
+      const void *matchinfo,
+      int offset,
+      int *hotdrop)
+{
+       const struct ipt_set_info_match *info = matchinfo;
+               
+       return match_set(&info->match_set,
+                        skb,
+                        info->match_set.flags[0] & IPSET_MATCH_INV);
+}
+
+static int
+checkentry(const char *tablename,
+          const struct ipt_ip *ip,
+          void *matchinfo,
+          unsigned int matchsize,
+          unsigned int hook_mask)
+{
+       struct ipt_set_info_match *info = 
+               (struct ipt_set_info_match *) matchinfo;
+       ip_set_id_t index;
+
+       if (matchsize != IPT_ALIGN(sizeof(struct ipt_set_info_match))) {
+               ip_set_printk("invalid matchsize %d", matchsize);
+               return 0;
+       }
+
+       index = ip_set_get_byindex(info->match_set.index);
+               
+       if (index == IP_SET_INVALID_ID) {
+               ip_set_printk("Cannot find set indentified by id %u to match",
+                             info->match_set.index);
+               return 0;       /* error */
+       }
+       if (info->match_set.flags[IP_SET_MAX_BINDINGS] != 0) {
+               ip_set_printk("That's nasty!");
+               return 0;       /* error */
+       }
+
+       return 1;
+}
+
+static void destroy(void *matchinfo, unsigned int matchsize)
+{
+       struct ipt_set_info_match *info = matchinfo;
+
+       if (matchsize != IPT_ALIGN(sizeof(struct ipt_set_info_match))) {
+               ip_set_printk("invalid matchsize %d", matchsize);
+               return;
+       }
+
+       ip_set_put(info->match_set.index);
+}
+
+static struct ipt_match set_match = {
+       .name           = "set",
+       .match          = &match,
+       .checkentry     = &checkentry,
+       .destroy        = &destroy,
+       .me             = THIS_MODULE
+};
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>");
+MODULE_DESCRIPTION("iptables IP set match module");
+
+static int __init init(void)
+{
+       return ipt_register_match(&set_match);
+}
+
+static void __exit fini(void)
+{
+       ipt_unregister_match(&set_match);
+}
+
+module_init(init);
+module_exit(fini);