- up to 2.6.17.28

[packages/kernel.git] / kernel-ipt_account.patch

diff -uNrp linux/net/ipv4/netfilter/ipt_account.c linux/net/ipv4/netfilter/ipt_account.c
--- linux/net/ipv4/netfilter/ipt_account.c	1970-01-01 01:00:00.000000000 +0100
+++ linux/net/ipv4/netfilter/ipt_account.c	2007-08-04 16:22:15.000000000 +0200
@@ -0,0 +1,973 @@
+/* Copyright (c) 2004-2007 Piotr 'QuakeR' Gasidlo <quaker@barbara.eu.org>
+ *
+ * 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.
+ */
+
+#include <linux/version.h>
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/vmalloc.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#include <linux/time.h>
+#include <linux/ip.h>
+#include <linux/in.h>
+
+#define IPT_ACCOUNT_VERSION "0.1.21"
+
+//#define DEBUG_IPT_ACCOUNT
+
+MODULE_AUTHOR("Piotr Gasidlo <quaker@barbara.eu.org>");
+MODULE_DESCRIPTION("Traffic accounting module");
+MODULE_LICENSE("GPL");
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)  
+#include <linux/netfilter/x_tables.h>
+#else
+#include <linux/netfilter_ipv4/ip_tables.h>
+#endif
+#include <linux/netfilter_ipv4/ipt_account.h>
+
+/* defaults, can be overriden */
+static unsigned int netmask = 16; /* Safe netmask, if you try to create table
+                                     for larger netblock you will get error. 
+                                     Increase by command line only when you
+                                     known what are you doing. */
+
+#ifdef DEBUG_IPT_ACCOUNT
+static int debug = 0;
+#endif
+module_param(netmask, uint, 0400);
+
+MODULE_PARM_DESC(netmask,"maximum *save* netmask");
+#ifdef DEBUG_IPT_ACCOUNT
+module_param(debug, bool, 0600);
+MODULE_PARM_DESC(debug,"enable debugging output");
+#endif
+
+/* structure with statistics counter, used when table is created without --ashort switch */
+struct t_ipt_account_stat_long {
+  u_int64_t b_all, b_tcp, b_udp, b_icmp, b_other;
+  u_int64_t p_all, p_tcp, p_udp, p_icmp, p_other;
+};
+
+/* same as above, for tables created with --ashort switch */
+struct t_ipt_account_stat_short {
+  u_int64_t b_all;
+  u_int64_t p_all;
+};
+
+/* structure holding to/from statistics for single ip when table is created without --ashort switch */
+struct t_ipt_account_stats_long {
+  struct t_ipt_account_stat_long src, dst;
+  struct timespec time; /* time, when statistics was last modified */
+};
+
+/* same as above, for tables created with --ashort switch */
+struct t_ipt_account_stats_short {
+  struct t_ipt_account_stat_short src, dst;
+  struct timespec time;
+};
+
+/* defines for "show" table option */
+#define SHOW_ANY 0
+#define SHOW_SRC 1
+#define SHOW_DST 2
+#define SHOW_SRC_OR_DST 3
+#define SHOW_SRC_AND_DST 4
+
+/* structure describing single table */
+struct t_ipt_account_table {
+  struct list_head list;
+  atomic_t use; /* use counter, the number of rules which points to this table */
+
+  char name[IPT_ACCOUNT_NAME_LEN + 1]; /* table name ( = filename in /proc/net/ipt_account/) */
+  u_int32_t network, netmask, count; /* network/netmask/hosts count coverted by table */
+
+  int shortlisting:1; /* gather only total statistics (set for tables created with --ashort switch) */
+  int timesrc:1; /* update time when accounting outgoing traffic */
+  int timedst:1; /* update time when accounting incomming traffic */
+  int resetonread:1; /* reset statistics after reading it via proc */
+  int show; /* show with entries */
+
+  /* FIXME: why int show:3 results in 'warning: comparison is always 0 due to width of bit-field' in ipt_account_seq_show
+   * gcc -v: gcc version 3.4.6 */
+  
+  union { /* statistics for each ip in network/netmask */
+    struct t_ipt_account_stats_long *l;
+    struct t_ipt_account_stats_short *s;
+  } stats;
+  rwlock_t stats_lock; /* lock, to assure that above union can be safely modified */
+
+  struct proc_dir_entry *pde; /* handle to proc entry */
+};
+
+static LIST_HEAD(ipt_account_tables);
+static rwlock_t ipt_account_lock = RW_LOCK_UNLOCKED; /* lock, to assure that table list can be safely modified */
+static DECLARE_MUTEX(ipt_account_mutex); /* additional checkentry protection */
+
+static struct file_operations ipt_account_proc_fops;
+static struct proc_dir_entry *ipt_account_procdir;
+
+/*
+ * Function creates new table and inserts it into linked list.
+ */
+static struct t_ipt_account_table *
+ipt_account_table_init(struct t_ipt_account_info *info)
+{ 
+  struct t_ipt_account_table *table;
+
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [ipt_account_table_init]: name = %s\n", info->name);
+#endif
+
+  /*
+   * Allocate memory for table.
+   */
+  table = vmalloc(sizeof(struct t_ipt_account_table));
+  if (!table) {
+    printk(KERN_ERR "ipt_account [ipt_account_table_init]: table = vmalloc(sizeof(struct t_ipt_account_table)) failed.\n");
+    goto cleanup_none;
+  }
+  memset(table, 0, sizeof(struct t_ipt_account_table));
+
+  /*
+   * Table attributes.
+   */
+  strncpy(table->name, info->name, IPT_ACCOUNT_NAME_LEN); 
+  table->name[IPT_ACCOUNT_NAME_LEN] = '\0';
+
+  table->network = info->network;
+  table->netmask = info->netmask;
+  table->count = (0xffffffff ^ table->netmask) + 1;
+  
+  /*
+   * Table properties.
+   */
+  table->shortlisting = info->shortlisting;
+  table->timesrc = 1;
+  table->timedst = 1;
+  table->resetonread = 0; 
+  table->show = SHOW_ANY;
+
+  /*
+   * Initialize use counter.
+   */
+  atomic_set(&table->use, 1);
+
+  /*
+   * Allocate memory for statistic counters.
+   */
+  if (table->shortlisting) {
+    table->stats.s = vmalloc(sizeof(struct t_ipt_account_stats_short) * table->count);
+    if (!table->stats.s) {
+      printk(KERN_ERR "ipt_account [ipt_account_table_init]: table->stats.s = vmalloc(sizeof(struct t_ipt_account_stats_short) * table->count) failed.\n");
+      goto cleanup_table;
+    }
+    memset(table->stats.s, 0, sizeof(struct t_ipt_account_stats_short) * table->count);
+  } else {
+    table->stats.l = vmalloc(sizeof(struct t_ipt_account_stats_long) * table->count);
+    if (!table->stats.l) {
+      printk(KERN_ERR "ipt_account [ipt_account_table_init]: table->stats.l = vmalloc(sizeof(struct t_ipt_account_stats_long) * table->count) failed.\n");
+      goto cleanup_table;
+    }
+    memset(table->stats.l, 0, sizeof(struct t_ipt_account_stats_long) * table->count);
+  }
+  
+  /*
+   * Reset locks.
+   */
+  table->stats_lock = RW_LOCK_UNLOCKED;
+  
+  /*
+   * Create /proc/ipt_account/name entry.
+   */
+  table->pde = create_proc_entry(table->name, S_IWUSR | S_IRUSR, ipt_account_procdir);
+  if (!table->pde) {
+    goto cleanup_stats;
+  }
+  table->pde->proc_fops = &ipt_account_proc_fops;
+  table->pde->data = table;
+  
+  /*
+   * Insert table into list.
+   */
+  write_lock_bh(&ipt_account_lock);
+  list_add(&table->list, &ipt_account_tables);
+  write_unlock_bh(&ipt_account_lock);
+  
+  return table;
+
+  /*
+   * If something goes wrong we end here.
+   */
+cleanup_stats:
+  if (table->shortlisting)
+    vfree(table->stats.s);
+  else
+    vfree(table->stats.l);
+  
+cleanup_table:
+  vfree(table);
+cleanup_none:
+  return NULL;
+  
+}
+
+/*
+ * Function destroys table. Table *must* be already unlinked.
+ */
+static void
+ipt_account_table_destroy(struct t_ipt_account_table *table)
+{
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [ipt_account_table_destory]: name = %s\n", table->name);
+#endif  
+  remove_proc_entry(table->pde->name, table->pde->parent);
+  if (table->shortlisting)
+    vfree(table->stats.s);
+  else
+    vfree(table->stats.l);
+  vfree(table);
+}
+
+/*
+ * Function increments use counter for table.
+ */
+static inline void
+ipt_account_table_get(struct t_ipt_account_table *table) 
+{
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [ipt_account_table_get]: name = %s\n", table->name);
+#endif    
+  atomic_inc(&table->use);
+}
+
+/*
+ * Function decrements use counter for table. If use counter drops to zero,
+ * table is removed from linked list and destroyed.
+ */
+static inline void
+ipt_account_table_put(struct t_ipt_account_table *table) 
+{
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [ipt_account_table_put]: name = %s\n", table->name);  
+#endif  
+  if (atomic_dec_and_test(&table->use)) {
+    write_lock_bh(&ipt_account_lock);
+    list_del(&table->list);
+    write_unlock_bh(&ipt_account_lock);
+    ipt_account_table_destroy(table);
+  }
+}
+
+/*
+ * Helper function, which returns a structure pointer to a table with
+ * specified name.
+ */
+static struct t_ipt_account_table *
+__ipt_account_table_find(char *name) 
+{
+  struct list_head *pos;
+  list_for_each(pos, &ipt_account_tables) {
+    struct t_ipt_account_table *table = list_entry(pos,
+        struct t_ipt_account_table, list);
+    if (!strncmp(table->name, name, IPT_ACCOUNT_NAME_LEN))
+      return table;
+  }
+  return NULL;
+}
+
+/*
+ * Function, which returns a structure pointer to a table with
+ * specified name. When such table is found its use coutner
+ * is incremented.
+ */
+static inline struct t_ipt_account_table *
+ipt_account_table_find_get(char *name) 
+{
+  struct t_ipt_account_table *table;
+  
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [ipt_account_table_find_get]: name = %s\n", name);  
+#endif  
+  read_lock_bh(&ipt_account_lock);
+  table = __ipt_account_table_find(name);
+  if (!table) {
+    read_unlock_bh(&ipt_account_lock);
+    return NULL;
+  }
+  atomic_inc(&table->use);
+  read_unlock_bh(&ipt_account_lock);
+  return table;
+} 
+
+/*
+ * Helper function, with updates statistics for specified IP. It's only
+ * used for tables created without --ashort switch.
+ */
+static inline void
+__account_long(struct t_ipt_account_stat_long *stat, const struct sk_buff *skb) 
+{
+  stat->b_all += skb->len;
+  stat->p_all++;
+  
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)  
+  switch (ip_hdr(skb)->protocol) {
+#else
+  switch (skb->nh.iph->protocol) {
+#endif    
+    case IPPROTO_TCP:
+      stat->b_tcp += skb->len;
+      stat->p_tcp++;
+      break;
+    case IPPROTO_UDP:
+      stat->b_udp += skb->len;
+      stat->p_udp++;
+      break;
+    case IPPROTO_ICMP:
+      stat->b_icmp += skb->len;
+      stat->p_icmp++;
+      break;
+    default:
+      stat->b_other += skb->len;
+      stat->p_other++;
+  }
+}
+
+/*
+ * Same as above, but used for tables created with --ashort switch.
+ */
+static inline void
+__account_short(struct t_ipt_account_stat_short *stat, const struct sk_buff *skb)
+{
+  stat->b_all += skb->len;
+  stat->p_all++;
+}
+
+/*
+ * Match function. Here we do accounting stuff.
+ */
+static bool
+match(const struct sk_buff *skb,
+      const struct net_device *in,
+      const struct net_device *out,
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
+      const struct xt_match *match,
+#endif            
+      const void *matchinfo,
+      int offset,
+      unsigned int protoff,
+      bool *hotdrop)
+{
+  struct t_ipt_account_info *info = (struct t_ipt_account_info *)matchinfo;
+  struct t_ipt_account_table *table = info->table;
+  u_int32_t address;  
+  /* Get current time. */
+  struct timespec now = CURRENT_TIME_SEC;
+  /* Default we assume no match. */
+  bool ret = false;
+    
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [match]: name = %s\n", table->name);
+#endif  
+  /* Check whether traffic from source ip address ... */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)  
+  address = ntohl(ip_hdr(skb)->saddr);
+#else
+  address = ntohl(skb->nh.iph->saddr);
+#endif  
+  /* ... is being accounted by this table. */ 
+  if (address && ((u_int32_t)(address & table->netmask) == (u_int32_t)table->network)) {    
+    write_lock_bh(&table->stats_lock);
+    /* Yes, account this packet. */
+#ifdef DEBUG_IPT_ACCOUNT  
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)  
+    if (debug) printk(KERN_DEBUG "ipt_account: [match]: accounting packet src = %u.%u.%u.%u, proto = %u.\n", HIPQUAD(address), ip_hdr(skb)->protocol);
+#else
+    if (debug) printk(KERN_DEBUG "ipt_account: [match]: accounting packet src = %u.%u.%u.%u, proto = %u.\n", HIPQUAD(address), skb->nh.iph->protocol);
+#endif  
+#endif  
+    /* Update counters this host. */
+    if (!table->shortlisting) {
+      __account_long(&table->stats.l[address - table->network].src, skb);
+      if (table->timesrc)
+        table->stats.l[address - table->network].time = now;
+      /* Update also counters for all hosts in this table (network address) */
+      if (table->count > 1) {
+        __account_long(&table->stats.l[0].src, skb);
+        table->stats.l[0].time = now;
+      }
+    } else {
+      __account_short(&table->stats.s[address - table->network].src, skb);
+      if (table->timedst)
+        table->stats.s[address - table->network].time = now;
+      if (table->count > 1) {
+        __account_short(&table->stats.s[0].src, skb);
+        table->stats.s[0].time = now;
+      }
+    }
+    write_unlock_bh(&table->stats_lock);
+    /* Yes, it's a match. */
+    ret = true;
+  }
+  
+  /* Do the same thing with destination ip address. */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)  
+  address = ntohl(ip_hdr(skb)->daddr);
+#else
+  address = ntohl(skb->nh.iph->daddr);
+#endif  
+  if (address && ((u_int32_t)(address & table->netmask) == (u_int32_t)table->network)) {
+    write_lock_bh(&table->stats_lock);
+#ifdef DEBUG_IPT_ACCOUNT 
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,22)  
+    if (debug) printk(KERN_DEBUG "ipt_account: [match]: accounting packet dst = %u.%u.%u.%u, proto = %u.\n", HIPQUAD(address), ip_hdr(skb)->protocol);
+#else    
+    if (debug) printk(KERN_DEBUG "ipt_account: [match]: accounting packet dst = %u.%u.%u.%u, proto = %u.\n", HIPQUAD(address), skb->nh.iph->protocol);
+#endif  
+#endif    
+    if (!table->shortlisting) {
+      __account_long(&table->stats.l[address - table->network].dst, skb);
+      table->stats.l[address - table->network].time = now;
+      if (table->count > 1) {
+        __account_long(&table->stats.l[0].dst, skb);
+        table->stats.l[0].time = now;
+      }
+    } else {
+      __account_short(&table->stats.s[address - table->network].dst, skb);
+      table->stats.s[address - table->network].time = now;
+      if (table->count > 1) {
+        __account_short(&table->stats.s[0].dst, skb);
+        table->stats.s[0].time = now;
+      }
+    }
+    write_unlock_bh(&table->stats_lock);
+    ret = true;
+  }
+  
+  return ret;
+}
+
+/*
+ * Checkentry function.
+ */
+static bool
+checkentry(const char *tablename,
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,16)    
+     const void *ip,
+#else
+     const struct ipt_entry *ip,
+#endif
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)     
+     const struct xt_match *match,
+#endif     
+     void *matchinfo,
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)
+     unsigned int matchsize,
+#endif     
+     unsigned int hook_mask)
+{
+  struct t_ipt_account_info *info = matchinfo;
+  struct t_ipt_account_table *table;
+
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [checkentry]: name = %s\n", info->name);
+#endif  
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17)
+  if (matchsize != IPT_ALIGN(sizeof(struct t_ipt_account_info))) {
+#ifdef DEBUG_IPT_ACCOUNT  
+    if (debug) printk(KERN_DEBUG "ipt_account [checkentry]: matchsize %u != %u\n", matchsize, IPT_ALIGN(sizeof(struct t_ipt_account_info)));
+#endif    
+    return false;
+  }
+#endif
+
+  /* 
+   * Sanity checks. 
+   */
+  if (info->netmask < ((~0L << (32 - netmask)) & 0xffffffff)) {
+    printk(KERN_ERR "ipt_account[checkentry]: too big netmask (increase module 'netmask' parameter).\n");
+    return false;
+  }
+  if ((info->network & info->netmask) != info->network) {
+    printk(KERN_ERR "ipt_account[checkentry]: wrong network/netmask.\n");
+    return false;
+  }
+  if (info->name[0] == '\0') {
+    printk(KERN_ERR "ipt_account[checkentry]: wrong table name.\n");
+    return false;
+  }
+
+  /*
+   * We got new rule. Try to find table with the same name as given in info structure.
+   * Mutex magic based on xt_hashlimit.c.
+   */
+  down(&ipt_account_mutex);
+  table = ipt_account_table_find_get(info->name);
+  if (table) {
+    if (info->table != NULL) {
+      if (info->table != table) {
+        printk(KERN_ERR "ipt_account[checkentry]: reloaded rule has invalid table pointer.\n");
+        up(&ipt_account_mutex);
+        return false;
+      }
+      up(&ipt_account_mutex);
+      return true;
+    } else {
+#ifdef DEBUG_IPT_ACCOUNT  
+      if (debug) printk(KERN_DEBUG "ipt_account [checkentry]: table found, checking.\n");
+#endif  
+      /* 
+       * Table exists, but whether rule network/netmask/shortlisting matches 
+       * table network/netmask/shortlisting. Failure on missmatch. 
+       */   
+      if (table->network != info->network || table->netmask != info->netmask || table->shortlisting != info->shortlisting) {
+        printk(KERN_ERR "ipt_account [checkentry]: table found, rule network/netmask/shortlisting not match table network/netmask/shortlisting.\n");
+        /*
+         * Remember to release table usage counter.
+         */
+        ipt_account_table_put(table);
+        up(&ipt_account_mutex);
+        return false;
+      }
+#ifdef DEBUG_IPT_ACCOUNT  
+      if (debug) printk(KERN_DEBUG "ipt_account [checkentry]: table found, reusing.\n");
+#endif  
+      /*
+       * Link rule with table.
+       */ 
+      info->table = table;
+    }
+  } else {
+#ifdef DEBUG_IPT_ACCOUNT  
+    if (debug) printk(KERN_DEBUG "ipt_account [checkentry]: table not found, creating new one.\n");
+#endif  
+    /*
+     * Table not exist, create new one.
+     */
+    info->table = table = ipt_account_table_init(info);
+    if (!table) {
+      up(&ipt_account_mutex);
+      return false;
+    }
+  }
+  up(&ipt_account_mutex);
+  return true;
+}
+
+/*
+ * Destroy function.
+ */
+static void
+destroy(
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)    
+    const struct xt_match *match,
+#endif    
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19)
+    void *matchinfo
+#else
+    void *matchinfo,
+    unsigned int matchsize
+#endif    
+)
+{
+  struct t_ipt_account_info *info = matchinfo;
+  
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [destroy]: name = %s\n", info->name);
+#endif  
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17)
+  if (matchsize != IPT_ALIGN(sizeof(struct t_ipt_account_info))) {
+#ifdef DEBUG_IPT_ACCOUNT  
+    if (debug) printk(KERN_DEBUG "ipt_account [checkentry]: matchsize %u != %u\n", matchsize, IPT_ALIGN(sizeof(struct t_ipt_account_info)));
+#endif    
+    return;
+  }
+#endif
+
+  /*
+   * Release table, by decreasing its usage counter. When
+   * counter hits zero, memory used by table structure is
+   * released and table is removed from list.
+   */
+  ipt_account_table_put(info->table);
+  return;
+}
+
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
+static struct xt_match account_match = { 
+#else
+static struct ipt_match account_match = { 
+#endif  
+  .name = "account", 
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
+  .family = AF_INET,
+#endif  
+  .match = &match, 
+  .checkentry = &checkentry, 
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,17)
+  .matchsize = sizeof(struct t_ipt_account_info),
+#endif  
+  .destroy = &destroy, 
+  .me = THIS_MODULE
+};
+
+/*
+ * Below functions (ipt_account_seq_start, ipt_account_seq_next, 
+ * ipt_account_seq_stop, ipt_account_seq_show, ipt_account_proc_write) 
+ * are used to implement proc stuff.
+ */
+static void *ipt_account_seq_start(struct seq_file *sf, loff_t *pos)
+{
+  struct proc_dir_entry *pde = sf->private;
+  struct t_ipt_account_table *table = pde->data;
+  unsigned int *i;
+
+  if (table->resetonread) {
+    /* When we reset entries after read we must have exclusive lock. */
+    write_lock_bh(&table->stats_lock);
+  } else {
+    read_lock_bh(&table->stats_lock);
+  }
+  if (*pos >= table->count)
+    return NULL;
+  i = kmalloc(sizeof(unsigned int), GFP_ATOMIC);
+  if (!i)
+    return ERR_PTR(-ENOMEM);
+  *i = *pos;
+  return i;
+}
+
+static void *ipt_account_seq_next(struct seq_file *sf, void *v, loff_t *pos)
+{
+  struct proc_dir_entry *pde = sf->private;
+  struct t_ipt_account_table *table = pde->data;
+  unsigned int *i = (unsigned int *)v;
+
+  *pos = ++(*i);
+  if (*i >= table->count) {
+    kfree(v);
+    return NULL;
+  }
+  return i;
+}
+
+static void ipt_account_seq_stop(struct seq_file *sf, void *v)
+{
+  struct proc_dir_entry *pde = sf->private;
+  struct t_ipt_account_table *table = pde->data;
+  kfree(v);
+  if (table->resetonread) {   
+    write_unlock_bh(&table->stats_lock);
+  } else {
+    read_unlock_bh(&table->stats_lock);
+  }
+}
+
+static int ipt_account_seq_show(struct seq_file *sf, void *v)
+{
+  struct proc_dir_entry *pde = sf->private;
+  struct t_ipt_account_table *table = pde->data;
+  unsigned int *i = (unsigned int *)v;
+  
+  struct timespec now = CURRENT_TIME_SEC;
+  
+  u_int32_t address = table->network + *i;
+
+  if (!table->shortlisting) {
+    struct t_ipt_account_stats_long *l = &table->stats.l[*i];
+    /* Don't list rows not matching show requirements. */
+    if (
+        ((table->show == SHOW_SRC) && (l->src.p_all == 0)) ||
+        ((table->show == SHOW_DST) && (l->dst.p_all == 0)) ||
+        ((table->show == SHOW_SRC_OR_DST) && ((l->src.p_all == 0) && (l->dst.p_all == 0))) ||
+        ((table->show == SHOW_SRC_AND_DST) && ((l->src.p_all == 0) || (l->dst.p_all == 0)))
+       ) {
+      return 0;
+    }
+    seq_printf(sf,
+        "ip = %u.%u.%u.%u bytes_src = %llu %llu %llu %llu %llu packets_src = %llu %llu %llu %llu %llu bytes_dst = %llu %llu %llu %llu %llu packets_dst = %llu %llu %llu %llu %llu time = %lu\n",
+        HIPQUAD(address),
+        l->src.b_all,
+        l->src.b_tcp,
+        l->src.b_udp,
+        l->src.b_icmp,
+        l->src.b_other,
+        l->src.p_all,
+        l->src.p_tcp,
+        l->src.p_udp,
+        l->src.p_icmp,
+        l->src.p_other,
+        l->dst.b_all,
+        l->dst.b_tcp,
+        l->dst.b_udp,
+        l->dst.b_icmp,
+        l->dst.b_other,       
+        l->dst.p_all,
+        l->dst.p_tcp,
+        l->dst.p_udp,
+        l->dst.p_icmp,
+        l->dst.p_other,
+        now.tv_sec - l->time.tv_sec
+      );
+    if (table->resetonread)
+      memset(l, 0, sizeof(struct t_ipt_account_stats_long));
+    
+  } else {
+    struct t_ipt_account_stats_short *s = &table->stats.s[*i];
+    if (
+        ((table->show == SHOW_SRC) && (s->src.p_all == 0)) ||
+        ((table->show == SHOW_DST) && (s->dst.p_all == 0)) ||
+        ((table->show == SHOW_SRC_OR_DST) && ((s->src.p_all == 0) && (s->dst.p_all == 0))) ||
+        ((table->show == SHOW_SRC_AND_DST) && ((s->src.p_all == 0) || (s->dst.p_all == 0)))
+       ) {
+      return 0;
+    }
+    seq_printf(sf,
+        "ip = %u.%u.%u.%u bytes_src = %llu packets_src = %llu bytes_dst = %llu packets_dst = %llu time = %lu\n",
+        HIPQUAD(address),
+        s->src.b_all,
+        s->src.p_all,
+        s->dst.b_all,
+        s->dst.p_all,
+        now.tv_sec - s->time.tv_sec
+      );
+    if (table->resetonread)
+      memset(s, 0, sizeof(struct t_ipt_account_stats_short));
+  }
+
+  return 0;
+}
+
+static struct seq_operations ipt_account_seq_ops = {
+  .start = ipt_account_seq_start,
+  .next = ipt_account_seq_next,
+  .stop = ipt_account_seq_stop,
+  .show = ipt_account_seq_show
+};
+
+static ssize_t ipt_account_proc_write(struct file *file, const char __user *input, size_t size, loff_t *ofs)
+{
+  char buffer[1024];
+  struct proc_dir_entry *pde = PDE(file->f_dentry->d_inode);
+  struct t_ipt_account_table *table = pde->data;
+
+  u_int32_t o[4], ip;
+  struct t_ipt_account_stats_long l;
+  struct t_ipt_account_stats_short s;
+
+#ifdef DEBUG_IPT_ACCOUNT  
+  if (debug) printk(KERN_DEBUG "ipt_account [ipt_account_proc_write]: name = %s.\n", table->name);
+#endif  
+  if (copy_from_user(buffer, input, 1024))
+    return -EFAULT;
+  buffer[1023] = '\0';
+
+  if (!strncmp(buffer, "reset\n", 6)) {
+    /*
+     * User requested to clear all table. Ignorant, does
+     * he known how match time it took us to fill it? ;-)
+     */
+    write_lock_bh(&table->stats_lock);
+    if (table->shortlisting)
+      memset(table->stats.s, 0, sizeof(struct t_ipt_account_stats_short) * table->count);
+    else
+      memset(table->stats.l, 0, sizeof(struct t_ipt_account_stats_long) * table->count);
+    write_unlock_bh(&table->stats_lock);    
+  } else if (!strncmp(buffer, "reset-on-read=yes\n", 18)) {
+    /*
+     * We must be sure that ipt_account_seq_* is not running now. This option
+     * changes lock type which is taken in ipt_account_seq_{start|stop}. When
+     * we change this option without this lock, and ipt_account_seq_start is
+     * already run (but not ipt_account_seq_stop) there is possibility that
+     * we execute wrong "unlock" function.
+     */
+    write_lock_bh(&table->stats_lock);
+    table->resetonread = 1;
+    write_unlock_bh(&table->stats_lock);
+  } else if (!strncmp(buffer, "reset-on-read=no\n", 17)) {
+    write_lock_bh(&table->stats_lock);
+    table->resetonread = 0; 
+    write_unlock_bh(&table->stats_lock);
+  } else if (!strncmp(buffer, "reset-on-read\n", 14)) {
+    write_lock_bh(&table->stats_lock);
+    table->resetonread = 1;
+    write_unlock_bh(&table->stats_lock);
+  } else if (!strncmp(buffer, "show=any\n", 9)) {
+    /*
+     * Here we should lock but we don't have to. So we don't lock. We only get
+     * wrong results on already run ipt_account_seq_show, but we won't crush
+     * the system.
+     */
+    table->show = SHOW_ANY;
+  } else if (!strncmp(buffer, "show=src\n", 9)) {
+    table->show = SHOW_SRC;
+  } else if (!strncmp(buffer, "show=dst\n", 9)) {
+    table->show = SHOW_DST;
+  } else if (!strncmp(buffer, "show=src-or-dst\n", 16) || !strncmp(buffer, "show=dst-or-src\n", 16)) {
+    table->show = SHOW_SRC_OR_DST;
+  } else if (!strncmp(buffer, "show=src-and-dst\n", 17) || !strncmp(buffer, "show=dst-and-src\n", 17)) {
+    table->show = SHOW_SRC_AND_DST;
+  } else if (!strncmp(buffer, "time=any\n", 9)) {
+    table->timesrc = table->timedst = 1;
+  } else if (!strncmp(buffer, "time=src\n", 9)) {
+    table->timesrc = 1;
+    table->timedst = 0;
+  } else if (!strncmp(buffer, "time=dst\n", 9)) {
+    table->timesrc = 0;
+    table->timedst = 1;
+  } else if (!table->shortlisting && sscanf(buffer, "ip = %u.%u.%u.%u bytes_src = %llu %llu %llu %llu %llu packets_src = %llu %llu %llu %llu %llu bytes_dst = %llu %llu %llu %llu %llu packets_dst = %llu %llu %llu %llu %llu time = %lu",        
+        &o[0], &o[1], &o[2], &o[3],
+        &l.src.b_all, &l.src.b_tcp, &l.src.b_udp, &l.src.b_icmp, &l.src.b_other,
+        &l.src.p_all, &l.src.p_tcp, &l.src.p_udp, &l.src.p_icmp, &l.src.p_other,
+        &l.dst.b_all, &l.dst.b_tcp, &l.dst.b_udp, &l.dst.b_icmp, &l.dst.b_other,
+        &l.dst.p_all, &l.dst.p_tcp, &l.dst.p_udp, &l.dst.p_icmp, &l.dst.p_other,
+        &l.time.tv_sec) == 25 ) {
+    /*
+     * We got line formated like long listing row. We have to
+     * check, if IP is accounted by table. If so, we
+     * simply replace row with user's one.
+     */
+    ip = o[0] << 24 | o[1] << 16 | o[2] << 8 | o[3];
+    if ((u_int32_t)(ip & table->netmask) == (u_int32_t)table->network) {
+      /*
+       * Ignore user input time. Set current time.
+       */
+      l.time = CURRENT_TIME_SEC;
+      write_lock_bh(&table->stats_lock);
+      table->stats.l[ip - table->network] = l;
+      write_unlock_bh(&table->stats_lock);
+    }
+  } else if (table->shortlisting && sscanf(buffer, "ip = %u.%u.%u.%u bytes_src = %llu packets_src = %llu bytes_dst = %llu packets_dst = %llu time = %lu\n", 
+        &o[0], &o[1], &o[2], &o[3], 
+        &s.src.b_all, 
+        &s.src.p_all, 
+        &s.dst.b_all, 
+        &s.dst.p_all, 
+        &s.time.tv_sec) == 9) {   
+    /*
+     * We got line formated like short listing row. Do the
+     * same action like above.
+     */
+    ip = o[0] << 24 | o[1] << 16 | o[2] << 8 | o[3];    
+    if ((u_int32_t)(ip & table->netmask) == (u_int32_t)table->network) {
+      s.time = CURRENT_TIME_SEC;
+      write_lock_bh(&table->stats_lock);
+      table->stats.s[ip - table->network] = s;
+      write_unlock_bh(&table->stats_lock);
+    }
+  } else {
+    /*
+     * We don't understand what user have just wrote.
+     */
+    return -EIO;
+  }
+
+  return size;
+}
+
+static int ipt_account_proc_open(struct inode *inode, struct file *file)
+{
+  int ret = seq_open(file, &ipt_account_seq_ops);
+  if (!ret) {
+    struct seq_file *sf = file->private_data;
+    struct proc_dir_entry *pde = PDE(inode);
+    struct t_ipt_account_table *table = pde->data;
+    
+    sf->private = pde;
+
+    ipt_account_table_get(table);
+  }
+  return ret;
+}
+
+static int ipt_account_proc_release(struct inode *inode, struct file *file)
+{
+  struct proc_dir_entry *pde = PDE(inode);
+  struct t_ipt_account_table *table = pde->data;
+  int ret;
+
+  ret = seq_release(inode, file);
+
+  if (!ret)
+    ipt_account_table_put(table);
+  
+  return ret;
+}
+
+static struct file_operations ipt_account_proc_fops = {
+  .owner = THIS_MODULE,
+  .open = ipt_account_proc_open,
+  .read = seq_read,
+  .write = ipt_account_proc_write,
+  .llseek = seq_lseek,
+  .release = ipt_account_proc_release
+};
+
+/*
+ * Module init function.
+ */
+static int __init init(void)
+{
+  int ret = 0;
+
+  printk(KERN_INFO "ipt_account %s : Piotr Gasidlo <quaker@barbara.eu.org>, http://www.barbara.eu.org/~quaker/ipt_account/\n", IPT_ACCOUNT_VERSION);
+
+  /* Check module parameters. */
+  if (netmask > 32 || netmask < 0) {
+    printk(KERN_ERR "ipt_account[__init]: Wrong netmask given as parameter (%i). Valid is 32 to 0.\n", netmask);
+    ret = -EINVAL;
+    goto cleanup_none;
+  }
+  
+  /* Register match. */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
+  if (xt_register_match(&account_match)) {
+#else    
+  if (ipt_register_match(&account_match)) {
+#endif  
+    ret = -EINVAL;
+    goto cleanup_none;
+  }
+
+  /* Create /proc/net/ipt_account/ entry. */
+  ipt_account_procdir = proc_mkdir("ipt_account", init_net.proc_net);
+  if (!ipt_account_procdir) {   
+    printk(KERN_ERR "ipt_account [__init]: ipt_account_procdir = proc_mkdir(\"ipt_account\", init_net.proc_net) failed.\n");
+    ret = -ENOMEM;
+    goto cleanup_match;
+  }
+  
+  return ret;
+
+  /* If something goes wrong we end here. */
+cleanup_match:
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
+  xt_unregister_match(&account_match);
+#else  
+  ipt_unregister_match(&account_match);
+#endif  
+cleanup_none:
+  return ret;
+}
+
+/*
+ * Module exit function.
+ */
+static void __exit fini(void)
+{
+  /* Remove /proc/net/ipt_account/ */
+  remove_proc_entry(ipt_account_procdir->name, ipt_account_procdir->parent);  
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,21)
+  xt_unregister_match(&account_match);
+#else
+  ipt_unregister_match(&account_match);
+#endif  
+}
+
+module_init(init);
+module_exit(fini);
+
diff -uNrp linux/include/linux/netfilter_ipv4/ipt_account.h linux/include/linux/netfilter_ipv4/ipt_account.h
--- linux/include/linux/netfilter_ipv4/ipt_account.h	1970-01-01 01:00:00.000000000 +0100
+++ linux/include/linux/netfilter_ipv4/ipt_account.h	2007-08-04 15:24:49.000000000 +0200
@@ -0,0 +1,24 @@
+/* Copyright (c) 2004-2007 Piotr 'QuakeR' Gasidlo <quaker@barbara.eu.org>
+ *
+ * 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.
+ */
+
+#ifndef _IPT_ACCOUNT_H_
+#define _IPT_ACCOUNT_H_
+
+#define IPT_ACCOUNT_NAME_LEN 64
+
+struct t_ipt_account_table;
+
+struct t_ipt_account_info {
+  char name[IPT_ACCOUNT_NAME_LEN + 1];
+  u_int32_t network, netmask;
+  int shortlisting:1;
+  /* pointer to the table for fast matching */
+  struct t_ipt_account_table *table;
+};
+
+#endif /* _IPT_ACCOUNT_H */
+
diff -uNrp linux/net/ipv4/netfilter/Makefile.a linux/net/ipv4/netfilter/Makefile.b
--- linux/net/ipv4/netfilter/Makefile	1971-01-01 01:00:00.000000000 +0100
+++ linux/net/ipv4/netfilter/Makefile	2007-08-15 13:23:25.375304000 +0200
@@ -0,0 +0,1 @@
+obj-$(CONFIG_IP_NF_MATCH_ACCOUNT) += ipt_account.o
diff -Nur linux/net/ipv4/netfilter/Kconfig linux/net/ipv4/netfilter/Kconfig
--- linux/net/ipv4/netfilter/Kconfig	2006-05-02 23:38:44.000000000 +0200
+++ linux/net/ipv4/netfilter/Kconfig	2006-05-04 11:23:02.000000000 +0200
@@ -606,5 +606,51 @@
 	  Allows altering the ARP packet payload: source and destination
 	  hardware and network addresses.
 
+config IP_NF_MATCH_ACCOUNT
+	tristate "account match support"
+	depends on IP_NF_IPTABLES && PROC_FS
+	help
+	  This match is used for accounting traffic for all hosts in
+	  defined network/netmask. 
+	  
+	  Features:
+	  - long (one counter per protocol TCP/UDP/IMCP/Other) and short statistics
+	  - one iptables rule for all hosts in network/netmask
+	  - loading/saving counters (by reading/writting to procfs entries)
+	  
+	  Example usage:
+	  
+	  account traffic for/to 192.168.0.0/24 network into table mynetwork:
+	  
+	  # iptables -A FORWARD -m account --aname mynetwork --aaddr 192.168.0.0/24
+	  
+	  account traffic for/to WWW serwer for 192.168.0.0/24 network into table 
+	  mywwwserver:
+	  
+	  # iptables -A INPUT -p tcp --dport 80 
+	    -m account --aname mywwwserver --aaddr 192.168.0.0/24 --ashort
+	  # iptables -A OUTPUT -p tcp --sport 80
+	    -m account --aname mywwwserver --aaddr 192.168.0.0/24 --ashort    
+	  
+	  read counters:
+	  
+	  # cat /proc/net/ipt_account/mynetwork
+	  # cat /proc/net/ipt_account/mywwwserver
+	  
+	  set counters:
+	  
+	  # echo "ip = 192.168.0.1 packets_src = 0" > /proc/net/ipt_account/mywwserver
+	  
+	  Webpage: 
+	    http://www.barbara.eu.org/~quaker/ipt_account/
+
+config IP_NF_MATCH_ACCOUNT_DEBUG
+	bool "account debugging output"
+	depends on IP_NF_MATCH_ACCOUNT
+	help
+	  Say Y to get lots of debugging output.
+	  
+
+
 endmenu