All of lore.kernel.org
 help / color / mirror / Atom feed
* new ipt_ACCOUNT version
@ 2004-05-29 13:27 Thomas Jarosch
  2004-06-06 22:40 ` Patrick McHardy
                   ` (2 more replies)
  0 siblings, 3 replies; 8+ messages in thread
From: Thomas Jarosch @ 2004-05-29 13:27 UTC (permalink / raw)
  To: netfilter-devel

[-- Attachment #1: Type: text/plain, Size: 434 bytes --]

Hi Patrick,

> I'm still not sure what to do about those long (?) > 80 chars lines.
> My "prefix" ipt_account_ uses a lot of space, maybe I should
> change it to something like "acc_". Other ideas?

I now fixed the too long lines issue. Attached is the new version.
Is it CVS ready?

The website (http://www.intra2net.com/opensource/ipt_ACCOUNT/) contains
the updated userspace counterparts (with short line lenghts, too ;-))

Thomas

[-- Attachment #2: ipt_ACCOUNT.h --]
[-- Type: text/x-chdr, Size: 3900 bytes --]

/***************************************************************************
 *   Copyright (C) 2004 by Intra2net AG                                    *
 *   opensource@intra2net.com                                              *
 *                                                                         *
 *   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 ACCOUNT_MAX_TABLES 32
#define ACCOUNT_TABLE_NAME_LEN 32
#define ACCOUNT_MAX_HANDLES 10

/* Structure for the userspace part of ipt_ACCOUNT */
struct ipt_acc_info {
    u_int32_t net_ip;
    u_int32_t net_mask;
    char table_name[ACCOUNT_TABLE_NAME_LEN];
    int32_t table_nr;
};

/* Internal table structure, generated by check_entry() */
struct ipt_acc_table {
    char name[ACCOUNT_TABLE_NAME_LEN];        /* name of the table */
    unsigned int ip;                          /* base IP of network */
    unsigned int netmask;                     /* netmask of the network */
    unsigned char depth;                      /* size of network: 
                                                 0: 8 bit, 1: 16bit, 2: 24 bit */
    unsigned int refcount;                    /* refcount of this table. 
                                                 if zero, destroy it */
    unsigned int itemcount;                   /* number of IPs in this table */
    void *data;                               /* pointer to the actual data, 
                                                 depending on netmask */
};

/* Internal handle structure */
struct ipt_acc_handle {
    unsigned int ip;                          /* base IP of network. Used for 
                                                 caculating the final IP during
                                                 get_data() */
    unsigned char depth;                      /* size of network. See above for 
                                                 details */
    unsigned int itemcount;                   /* number of IPs in this table */
    void *data;                               /* pointer to the actual data, 
                                                 depending on size */
};

/* Handle structure for communication with the userspace library */
struct ipt_acc_handle_sockopt {
    unsigned int handle_nr;                   /* Used for HANDLE_FREE */
    char name[ACCOUNT_TABLE_NAME_LEN];        /* Used for HANDLE_PREPARE_READ/
                                                 HANDLE_READ_FLUSH */
    unsigned int itemcount;                   /* Used for HANDLE_PREPARE_READ/
                                                 HANDLE_READ_FLUSH */
};

/* Used for every IP entry 
   Size is 16 bytes so that 256 (class C network) * 16 
   fits in one kernel (zero) page */
struct ipt_acc_ip {
    unsigned int src_packets;
    unsigned int src_bytes;
    unsigned int dst_packets;
    unsigned int dst_bytes;
};

/*
    Used for every IP when returning data
*/
struct ipt_acc_handle_ip {
    unsigned int ip;
    unsigned int src_packets;
    unsigned int src_bytes;
    unsigned int dst_packets;
    unsigned int dst_bytes;
};

/*
    The IPs are organized as an array so that direct slot
    calculations are possible.
    Only 8 bit networks are preallocated, 16/24 bit networks
    allocate their slots when needed -> very efficent.
*/
struct ipt_acc_mask_24 {
    struct ipt_acc_ip ip[256];
};

struct ipt_acc_mask_16 {
    struct ipt_acc_mask_24 *mask_24[256];
};

struct ipt_acc_mask_8 {
    struct ipt_acc_mask_16 *mask_16[256];
};

#endif /*_IPT_ACCOUNT_H*/

[-- Attachment #3: ipt_ACCOUNT.c --]
[-- Type: text/x-csrc, Size: 38826 bytes --]

/***************************************************************************
 *   This is a module which is used for counting packets.                  *
 *   See http://www.intra2net.com/opensource/ipt_account                   *
 *   for further information                                               *
 *                                                                         * 
 *   Copyright (C) 2004 by Intra2net AG                                    *
 *   opensource@intra2net.com                                              *
 *                                                                         *
 *   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/module.h>
#include <linux/skbuff.h>
#include <linux/ip.h>
#include <linux/spinlock.h>
#include <net/icmp.h>
#include <net/udp.h>
#include <net/tcp.h>
#include <linux/netfilter_ipv4/ip_tables.h>

struct in_device;
#include <net/route.h>
#include <linux/netfilter_ipv4/ipt_ACCOUNT.h>

#if 0
#define DEBUGP printk
#else
#define DEBUGP(format, args...)
#endif

#if (PAGE_SIZE < 4096)
#error "ipt_ACCOUNT needs at least a PAGE_SIZE of 4096"
#endif

struct ipt_acc_table *ipt_acc_tables = NULL;
struct ipt_acc_handle *ipt_acc_handles = NULL;
void *ipt_acc_tmpbuf = NULL;

/* Spinlock used for manipulating the current accounting tables/data */
static spinlock_t ipt_acc_lock = SPIN_LOCK_UNLOCKED;
/* Spinlock used for manipulating userspace handles/snapshot data */
static spinlock_t ipt_acc_userspace_lock = SPIN_LOCK_UNLOCKED;


/* Recursive free of all data structures */
void ipt_acc_data_free(void *data, unsigned char depth)
{
    /* Empty data set */
    if (!data)
        return;

    /* Free for 8 bit network */
    if (depth == 0) {
        free_page((unsigned long)data);
        data = NULL;
        return;
    }

    /* Free for 16 bit network */
    if (depth == 1) {
        struct ipt_acc_mask_16 *mask_16 = (struct ipt_acc_mask_16 *)data;
        unsigned int b;
        for (b=0; b <= 255; b++) {
            if (mask_16->mask_24[b] != 0) {
                free_page((unsigned long)mask_16->mask_24[b]);
                mask_16->mask_24[b] = NULL;
            }
        }
        free_page((unsigned long)data);
        data = NULL;
        return;
    }

    /* Free for 24 bit network */
    if (depth == 2) {
        unsigned int a, b;
        for (a=0; a <= 255; a++) {
            if (((struct ipt_acc_mask_8 *)data)->mask_16[a]) {
                struct ipt_acc_mask_16 *mask_16 = (struct ipt_acc_mask_16*)
                                   ((struct ipt_acc_mask_8 *)data)->mask_16[a];
                
                for (b=0; b <= 255; b++) {
                    if (mask_16->mask_24[b]) {
                        free_page((unsigned long)mask_16->mask_24[b]);
                        mask_16->mask_24[b] = NULL;
                    }
                }
                free_page((unsigned long)mask_16);
                mask_16 = NULL;
            }
        }
        free_page((unsigned long)data);
        data = NULL;
        return;
    }

    printk("ACCOUNT: ipt_acc_data_free called with unknown depth: %d\n", 
           depth);
    return;
}

/* Look for existing table / insert new one. 
   Return internal ID or -1 on error */
int ipt_acc_table_insert(char *name, unsigned int ip, unsigned int netmask)
{
    unsigned int i;

    DEBUGP("ACCOUNT: ipt_acc_table_insert: %s, %u.%u.%u.%u/%u.%u.%u.%u\n",
                                         name, NIPQUAD(ip), NIPQUAD(netmask));

    /* Look for existing table */
    for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
        if (strncmp(ipt_acc_tables[i].name, name, 
                    ACCOUNT_TABLE_NAME_LEN) == 0) {
            DEBUGP("ACCOUNT: Found existing slot: %d - "
                   "%u.%u.%u.%u/%u.%u.%u.%u\n", i, 
                   NIPQUAD(ipt_acc_tables[i].ip), 
                   NIPQUAD(ipt_acc_tables[i].netmask));

            if (ipt_acc_tables[i].ip != ip 
                || ipt_acc_tables[i].netmask != netmask) {
                printk("ACCOUNT: Table %s found, but IP/netmask mismatch. "
                       "IP/netmask found: %u.%u.%u.%u/%u.%u.%u.%u\n",
                       name, NIPQUAD(ipt_acc_tables[i].ip), 
                       NIPQUAD(ipt_acc_tables[i].netmask));
                return -1;
            }

            ipt_acc_tables[i].refcount++;
            DEBUGP("ACCOUNT: Refcount: %d\n", ipt_acc_tables[i].refcount);
            return i;
        }
    }

    /* Insert new table */
    for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
        /* Found free slot */
        if (ipt_acc_tables[i].name[0] == 0) {
            DEBUGP("ACCOUNT: Found free slot: %d\n", i);

            strncpy (ipt_acc_tables[i].name, name, ACCOUNT_TABLE_NAME_LEN-1);

            ipt_acc_tables[i].ip = ip;
            ipt_acc_tables[i].netmask = netmask;

            /* Calculate netsize */
            unsigned int j, calc_mask, netsize=0;
            calc_mask = htonl(netmask);
            for (j = 31; j > 0; j--) {
                if (calc_mask&(1<<j))
                    netsize++;
                else
                    break;
            }

            /* Calculate depth from netsize */
            if (netsize >= 24)
                ipt_acc_tables[i].depth = 0;
            else if (netsize >= 16)
                ipt_acc_tables[i].depth = 1;
            else if(netsize >= 8)
                ipt_acc_tables[i].depth = 2;

            DEBUGP("ACCOUNT: calculated netsize: %u -> "
                   "ipt_acc_table depth %u\n", netsize, 
                   ipt_acc_tables[i].depth);

            ipt_acc_tables[i].refcount++;
            if ((ipt_acc_tables[i].data
                = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                printk("ACCOUNT: out of memory for data of table: %s\n", name);
                memset(&ipt_acc_tables[i], 0, 
                       sizeof(struct ipt_acc_table));
                return -1;
            }

            return i;
        }
    }

    /* No free slot found */
    printk("ACCOUNT: No free table slot found (max: %d). "
           "Please increase ACCOUNT_MAX_TABLES.\n", ACCOUNT_MAX_TABLES);
    return -1;
}

static int ipt_acc_checkentry(const char *tablename,
                                  const struct ipt_entry *e,
                                  void *targinfo,
                                  unsigned int targinfosize,
                                  unsigned int hook_mask)
{
    struct ipt_acc_info *info = targinfo;

    if (targinfosize != IPT_ALIGN(sizeof(struct ipt_acc_info))) {
        DEBUGP("ACCOUNT: targinfosize %u != %u\n",
               targinfosize, IPT_ALIGN(sizeof(struct ipt_acc_info)));
        return 0;
    }

    spin_lock_bh(&ipt_acc_lock);
    int table_nr = ipt_acc_table_insert(info->table_name, info->net_ip,
                                            info->net_mask);
    if (table_nr == -1) {
        printk("ACCOUNT: Table insert problem. Aborting\n");
        spin_unlock_bh(&ipt_acc_lock);
        return 0;
    }
    /* Table nr caching so we don't have to do an extra string compare 
       for every packet */
    info->table_nr = table_nr;

    spin_unlock_bh(&ipt_acc_lock);

    return 1;
}

void ipt_acc_deleteentry(void *targinfo, unsigned int targinfosize)
{
    unsigned int i;
    struct ipt_acc_info *info = targinfo;

    if (targinfosize != IPT_ALIGN(sizeof(struct ipt_acc_info))) {
        DEBUGP("ACCOUNT: targinfosize %u != %u\n",
               targinfosize, IPT_ALIGN(sizeof(struct ipt_acc_info)));
    }

    spin_lock_bh(&ipt_acc_lock);

    DEBUGP("ACCOUNT: ipt_acc_deleteentry called for table: %s (#%d)\n", 
           info->table_name, info->table_nr);

    info->table_nr = -1;    /* Set back to original state */

    /* Look for table */
    for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
        if (strncmp(ipt_acc_tables[i].name, info->table_name, 
                    ACCOUNT_TABLE_NAME_LEN) == 0) {
            DEBUGP("ACCOUNT: Found table at slot: %d\n", i);

            ipt_acc_tables[i].refcount--;
            DEBUGP("ACCOUNT: Refcount left: %d\n", 
                   ipt_acc_tables[i].refcount);

            /* Table not needed anymore? */
            if (ipt_acc_tables[i].refcount == 0) {
                DEBUGP("ACCOUNT: Destroying table at slot: %d\n", i);
                ipt_acc_data_free(ipt_acc_tables[i].data, 
                                      ipt_acc_tables[i].depth);
                memset(&ipt_acc_tables[i], 0, 
                       sizeof(struct ipt_acc_table));
            }

            spin_unlock_bh(&ipt_acc_lock);
            return;
        }
    }

    /* Table not found */
    printk("ACCOUNT: Table %s not found for destroy\n", info->table_name);
    spin_unlock_bh(&ipt_acc_lock);
}

void ipt_acc_depth0_insert(struct ipt_acc_mask_24 *mask_24,
                               unsigned int net_ip, unsigned int netmask,
                               unsigned int src_ip, unsigned int dst_ip,
                               unsigned int size, unsigned int *itemcount)
{
    unsigned char is_src = 0, is_dst = 0;

    DEBUGP("ACCOUNT: ipt_acc_depth0_insert: %u.%u.%u.%u/%u.%u.%u.%u "
           "for net %u.%u.%u.%u/%u.%u.%u.%u, size: %u\n", NIPQUAD(src_ip), 
           NIPQUAD(dst_ip), NIPQUAD(net_ip), NIPQUAD(netmask), size);

    /* Check if src/dst is inside our network. */
    /* Special: net_ip = 0.0.0.0/0 gets stored as src in slot 0 */
    if (!netmask)
        src_ip = 0;
    if ((net_ip&netmask) == (src_ip&netmask))
        is_src = 1;
    if ((net_ip&netmask) == (dst_ip&netmask) && netmask)
        is_dst = 1;

    if (!is_src && !is_dst) {
        DEBUGP("ACCOUNT: Skipping packet %u.%u.%u.%u/%u.%u.%u.%u "
               "for net %u.%u.%u.%u/%u.%u.%u.%u\n", NIPQUAD(src_ip), 
               NIPQUAD(dst_ip), NIPQUAD(net_ip), NIPQUAD(netmask));
        return;
    }

    /* Check if this entry is new */
    char is_src_new_ip = 0, is_dst_new_ip = 0;

    /* Calculate array positions */
    unsigned char src_slot = (unsigned char)((src_ip&0xFF000000) >> 24);
    unsigned char dst_slot = (unsigned char)((dst_ip&0xFF000000) >> 24);

    /* Increase size counters */
    if (is_src) {
        /* Calculate network slot */
        DEBUGP("ACCOUNT: Calculated SRC 8 bit network slot: %d\n", src_slot);
        if (!mask_24->ip[src_slot].src_packets 
            && !mask_24->ip[src_slot].dst_packets)
            is_src_new_ip = 1;

        mask_24->ip[src_slot].src_packets++;
        mask_24->ip[src_slot].src_bytes+=size;
    }
    if (is_dst) {
        DEBUGP("ACCOUNT: Calculated DST 8 bit network slot: %d\n", dst_slot);
        if (!mask_24->ip[dst_slot].src_packets 
            && !mask_24->ip[dst_slot].dst_packets)
            is_dst_new_ip = 1;

        mask_24->ip[dst_slot].dst_packets++;
        mask_24->ip[dst_slot].dst_bytes+=size;
    }

    /* Increase itemcounter */
    DEBUGP("ACCOUNT: Itemcounter before: %d\n", *itemcount);
    if (src_slot == dst_slot) {
        if (is_src_new_ip || is_dst_new_ip) {
            DEBUGP("ACCOUNT: src_slot == dst_slot: %d, %d\n", 
                   is_src_new_ip, is_dst_new_ip);
            (*itemcount)++;
        }
    } else {
        if (is_src_new_ip) {
            DEBUGP("ACCOUNT: New src_ip: %u.%u.%u.%u\n", NIPQUAD(src_ip));
            (*itemcount)++;
        }
        if (is_dst_new_ip) {
            DEBUGP("ACCOUNT: New dst_ip: %u.%u.%u.%u\n", NIPQUAD(dst_ip));
            (*itemcount)++;
        }
    }
    DEBUGP("ACCOUNT: Itemcounter after: %d\n", *itemcount);
}

void ipt_acc_depth1_insert(struct ipt_acc_mask_16 *mask_16, 
                               unsigned int net_ip, unsigned int netmask, 
                               unsigned int src_ip, unsigned int dst_ip,
                               unsigned int size, unsigned int *itemcount)
{
    /* Do we need to process src IP? */
    if ((net_ip&netmask) == (src_ip&netmask)) {
        unsigned char slot = (unsigned char)((src_ip&0x00FF0000) >> 16);
        DEBUGP("ACCOUNT: Calculated SRC 16 bit network slot: %d\n", slot);

        /* Do we need to create a new mask_24 bucket? */
        if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] = 
             (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
            printk("ACCOUNT: Can't process packet because out of memory!\n");
            return;
        }

        ipt_acc_depth0_insert((struct ipt_acc_mask_24 *)mask_16->mask_24[slot],
                                  net_ip, netmask, src_ip, 0, size, itemcount);
    }

    /* Do we need to process dst IP? */
    if ((net_ip&netmask) == (dst_ip&netmask)) {
        unsigned char slot = (unsigned char)((dst_ip&0x00FF0000) >> 16);
        DEBUGP("ACCOUNT: Calculated DST 16 bit network slot: %d\n", slot);

        /* Do we need to create a new mask_24 bucket? */
        if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] 
            = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
            printk("ACCOUT: Can't process packet because out of memory!\n");
            return;
        }

        ipt_acc_depth0_insert((struct ipt_acc_mask_24 *)mask_16->mask_24[slot],
                                  net_ip, netmask, 0, dst_ip, size, itemcount);
    }
}

void ipt_acc_depth2_insert(struct ipt_acc_mask_8 *mask_8, 
                               unsigned int net_ip, unsigned int netmask,
                               unsigned int src_ip, unsigned int dst_ip,
                               unsigned int size, unsigned int *itemcount)
{
    /* Do we need to process src IP? */
    if ((net_ip&netmask) == (src_ip&netmask)) {
        unsigned char slot = (unsigned char)((src_ip&0x0000FF00) >> 8);
        DEBUGP("ACCOUNT: Calculated SRC 24 bit network slot: %d\n", slot);

        /* Do we need to create a new mask_24 bucket? */
        if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot] 
            = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
            printk("ACCOUNT: Can't process packet because out of memory!\n");
            return;
        }

        ipt_acc_depth1_insert((struct ipt_acc_mask_16 *)mask_8->mask_16[slot],
                                  net_ip, netmask, src_ip, 0, size, itemcount);
    }

    /* Do we need to process dst IP? */
    if ((net_ip&netmask) == (dst_ip&netmask)) {
        unsigned char slot = (unsigned char)((dst_ip&0x0000FF00) >> 8);
        DEBUGP("ACCOUNT: Calculated DST 24 bit network slot: %d\n", slot);

        /* Do we need to create a new mask_24 bucket? */
        if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot] 
            = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
            printk("ACCOUNT: Can't process packet because out of memory!\n");
            return;
        }

        ipt_acc_depth1_insert((struct ipt_acc_mask_16 *)mask_8->mask_16[slot],
                                  net_ip, netmask, 0, dst_ip, size, itemcount);
    }
}

static unsigned int ipt_acc_target(struct sk_buff **pskb,
                                       unsigned int hooknum,
                                       const struct net_device *in,
                                       const struct net_device *out,
                                       const void *targinfo,
                                       void *userinfo)
{
    const struct ipt_acc_info *info = 
        (const struct ipt_acc_info *)targinfo;
    unsigned int src_ip = (*pskb)->nh.iph->saddr;
    unsigned int dst_ip = (*pskb)->nh.iph->daddr;
    unsigned int size = ntohs((*pskb)->nh.iph->tot_len);

    spin_lock_bh(&ipt_acc_lock);

    if (ipt_acc_tables[info->table_nr].name[0] == 0) {
        printk("ACCOUNT: ipt_acc_target: Invalid table id %u. "
               "IPs %u.%u.%u.%u/%u.%u.%u.%u\n", info->table_nr, 
               NIPQUAD(src_ip), NIPQUAD(dst_ip));
        spin_unlock_bh(&ipt_acc_lock);
        return IPT_CONTINUE;
    }

    /* 8 bit network or "any" network */
    if (ipt_acc_tables[info->table_nr].depth == 0) {
        /* Count packet and check if the IP is new */
        ipt_acc_depth0_insert(
            (struct ipt_acc_mask_24 *)ipt_acc_tables[info->table_nr].data,
            ipt_acc_tables[info->table_nr].ip, 
            ipt_acc_tables[info->table_nr].netmask,
            src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
        spin_unlock_bh(&ipt_acc_lock);
        return IPT_CONTINUE;
    }

    /* 16 bit network */
    if (ipt_acc_tables[info->table_nr].depth == 1) {
        ipt_acc_depth1_insert(
            (struct ipt_acc_mask_16 *)ipt_acc_tables[info->table_nr].data,
            ipt_acc_tables[info->table_nr].ip, 
            ipt_acc_tables[info->table_nr].netmask,
            src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
        spin_unlock_bh(&ipt_acc_lock);
        return IPT_CONTINUE;
    }

    /* 24 bit network */
    if (ipt_acc_tables[info->table_nr].depth == 2) {
        ipt_acc_depth2_insert(
            (struct ipt_acc_mask_8 *)ipt_acc_tables[info->table_nr].data,
            ipt_acc_tables[info->table_nr].ip, 
            ipt_acc_tables[info->table_nr].netmask,
            src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
        spin_unlock_bh(&ipt_acc_lock);
        return IPT_CONTINUE;
    }

    printk("ACCOUNT: ipt_acc_target: Unable to process packet. "
           "Table id %u. IPs %u.%u.%u.%u/%u.%u.%u.%u\n", 
           info->table_nr, NIPQUAD(src_ip), NIPQUAD(dst_ip));

    spin_unlock_bh(&ipt_acc_lock);
    return IPT_CONTINUE;
}

/*
    Functions dealing with "handles":
    Handles are snapshots of a accounting state.
    
    read snapshots are only for debugging the code
    and are very expensive concerning speed/memory
    compared to read_and_flush.
    
    The functions aren't protected by spinlocks themselves
    as this is done in the ioctl part of the code.
*/

/*
    Find a free handle slot. Normally only one should be used,
    but there could be two or more applications accessing the data
    at the same time.
*/
int ipt_acc_handle_find_slot(void)
{
    unsigned int i;
    /* Insert new table */
    for (i = 0; i < ACCOUNT_MAX_HANDLES; i++) {
        /* Found free slot */
        if (ipt_acc_handles[i].data == NULL) {
            /* Don't "mark" data as used as we are protected by a spinlock 
               by the calling function. handle_find_slot() is only a function
               to prevent code duplication. */
            return i;
        }
    }

    /* No free slot found */
    printk("ACCOUNT: No free handle slot found (max: %u). "
           "Please increase ACCOUNT_MAX_HANDLES.\n", ACCOUNT_MAX_HANDLES);
    return -1;
}

int ipt_acc_handle_free(unsigned int handle)
{
    if (handle >= ACCOUNT_MAX_HANDLES) {
        printk("ACCOUNT: Invalid handle for ipt_acc_handle_free() specified:"
               " %u\n", handle);
        return -EINVAL;
    }

    ipt_acc_data_free(ipt_acc_handles[handle].data, 
                          ipt_acc_handles[handle].depth);
    memset (&ipt_acc_handles[handle], 0, sizeof (struct ipt_acc_handle));
    return 0;
}

/* Prepare data for read without flush. Use only for debugging!
   Real applications should use read&flush as it's way more efficent */
int ipt_acc_handle_prepare_read(char *tablename, unsigned int *count)
{
    int handle, i, table_nr=-1;

    for (i = 0; i < ACCOUNT_MAX_TABLES; i++)
        if (strncmp(ipt_acc_tables[i].name, tablename, 
            ACCOUNT_TABLE_NAME_LEN) == 0) {
            table_nr = i;
            break;
        }

    if (table_nr == -1) {
        printk("ACCOUNT: ipt_acc_handle_prepare_read(): "
               "Table %s not found\n", tablename);
        return -1;
    }

    /* Can't find a free handle slot? */
    if ((handle = ipt_acc_handle_find_slot()) == -1)
        return -1;

    /* Fill up handle structure */
    ipt_acc_handles[handle].ip = ipt_acc_tables[table_nr].ip;
    ipt_acc_handles[handle].depth = ipt_acc_tables[table_nr].depth;
    ipt_acc_handles[handle].itemcount = ipt_acc_tables[table_nr].itemcount;

    /* allocate "root" table */
    if ((ipt_acc_handles[handle].data = 
         (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
        printk("ACCOUNT: out of memory for root table "
               "in ipt_acc_handle_prepare_read()\n");
        memset (&ipt_acc_handles[handle], 0, 
                sizeof(struct ipt_acc_handle));
        return -1;
    }

    /* Recursive copy of complete data structure */
    unsigned int depth = ipt_acc_handles[handle].depth;
    if (depth == 0) {
        memcpy(ipt_acc_handles[handle].data, 
               ipt_acc_tables[table_nr].data, 
               sizeof(struct ipt_acc_mask_24));
    } else if (depth == 1) {
        struct ipt_acc_mask_16 *src_16 = 
            (struct ipt_acc_mask_16 *)ipt_acc_tables[table_nr].data;
        struct ipt_acc_mask_16 *network_16 =
            (struct ipt_acc_mask_16 *)ipt_acc_handles[handle].data;
        unsigned int b;

        for (b = 0; b <= 255; b++) {
            if (src_16->mask_24[b]) {
                if ((network_16->mask_24[b] = 
                     (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                    printk("ACCOUNT: out of memory during copy of 16 bit "
                           "network in ipt_acc_handle_prepare_read()\n");
                    ipt_acc_data_free(ipt_acc_handles[handle].data, depth);
                    memset (&ipt_acc_handles[handle], 0, 
                            sizeof(struct ipt_acc_handle));
                    return -1;
                }

                memcpy(network_16->mask_24[b], src_16->mask_24[b], 
                       sizeof(struct ipt_acc_mask_24));
            }
        }
    } else if(depth == 2) {
        struct ipt_acc_mask_8 *src_8 = 
            (struct ipt_acc_mask_8 *)ipt_acc_tables[table_nr].data;
        struct ipt_acc_mask_8 *network_8 = 
            (struct ipt_acc_mask_8 *)ipt_acc_handles[handle].data;
        unsigned int a;

        for (a = 0; a <= 255; a++) {
            if (src_8->mask_16[a]) {
                if ((network_8->mask_16[a] = 
                     (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                    printk("ACCOUNT: out of memory during copy of 24 bit network"
                           " in ipt_acc_handle_prepare_read()\n");
                    ipt_acc_data_free(ipt_acc_handles[handle].data, depth);
                    memset (&ipt_acc_handles[handle], 0, 
                            sizeof(struct ipt_acc_handle));
                    return -1;
                }

                memcpy(network_8->mask_16[a], src_8->mask_16[a], 
                       sizeof(struct ipt_acc_mask_16));

                struct ipt_acc_mask_16 *src_16 = src_8->mask_16[a];
                struct ipt_acc_mask_16 *network_16 = network_8->mask_16[a];
                unsigned int b;

                for (b = 0; b <= 255; b++) {
                    if (src_16->mask_24[b]) {
                        if ((network_16->mask_24[b] = 
                             (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                            printk("ACCOUNT: out of memory during copy of 16 bit"
                                   " network in ipt_acc_handle_prepare_read()\n");
                            ipt_acc_data_free(ipt_acc_handles[handle].data,
                                                  depth);
                            memset (&ipt_acc_handles[handle], 0, 
                                    sizeof(struct ipt_acc_handle));
                            return -1;
                        }

                        memcpy(network_16->mask_24[b], src_16->mask_24[b], 
                               sizeof(struct ipt_acc_mask_24));
                    }
                }
            }
        }
    }

    *count = ipt_acc_tables[table_nr].itemcount;
    return handle;
}

/* Prepare data for read and flush it */
int ipt_acc_handle_prepare_read_flush(char *tablename, unsigned int *count)
{
    int handle, i, table_nr=-1;

    for (i = 0; i < ACCOUNT_MAX_TABLES; i++)
        if (strncmp(ipt_acc_tables[i].name, tablename, 
                    ACCOUNT_TABLE_NAME_LEN) == 0) {
            table_nr = i;
            break;
        }

    if (table_nr == -1) {
        printk("ACCOUNT: ipt_acc_handle_prepare_read_flush(): "
               "Table %s not found\n", tablename);
        return -1;
    }

    /* Can't find a free handle slot? */
    if ((handle = ipt_acc_handle_find_slot()) == -1)
        return -1;

    /* Try to allocate memory */
    void *new_data_page = (void*)get_zeroed_page(GFP_ATOMIC);
    if (!new_data_page)
    {
        printk("ACCOUNT: ipt_acc_handle_prepare_read_flush(): "
               "Out of memory!\n");
        return -1;
    }

    /* Fill up handle structure */
    ipt_acc_handles[handle].ip = ipt_acc_tables[table_nr].ip;
    ipt_acc_handles[handle].depth = ipt_acc_tables[table_nr].depth;
    ipt_acc_handles[handle].itemcount = ipt_acc_tables[table_nr].itemcount;
    ipt_acc_handles[handle].data = ipt_acc_tables[table_nr].data;
    *count = ipt_acc_tables[table_nr].itemcount;

    /* "Flush" table data */
    ipt_acc_tables[table_nr].data = new_data_page;
    ipt_acc_tables[table_nr].itemcount = 0;

    return handle;
}

/* Copy 8 bit network data into a prepared buffer.
   We only copy entries != 0 to increase performance.
*/
void ipt_acc_handle_copy_data(void *to_user, int *pos, 
                                  struct ipt_acc_mask_24 *data,
                                  unsigned int net_ip, unsigned int net_OR_mask)
{
    struct ipt_acc_handle_ip handle_ip;
    unsigned int handle_ip_size = sizeof (struct ipt_acc_handle_ip);
    
    unsigned int i;
    
    for (i = 0; i <= 255; i++) {
        if (data->ip[i].src_packets || data->ip[i].dst_packets) {
            handle_ip.ip = net_ip | net_OR_mask | (i<<24);
            
            handle_ip.src_packets = data->ip[i].src_packets;
            handle_ip.src_bytes = data->ip[i].src_bytes;
            handle_ip.dst_packets = data->ip[i].dst_packets;
            handle_ip.dst_bytes = data->ip[i].dst_bytes;

            /* Temporary buffer full? Flush to userspace */
            if (*pos+handle_ip_size >= PAGE_SIZE) {
                copy_to_user(to_user, ipt_acc_tmpbuf, *pos);
                *pos = 0;
            }
            memcpy(ipt_acc_tmpbuf+*pos, &handle_ip, handle_ip_size);
            *pos += handle_ip_size;
        }
    }
}
   
/* Copy the data from our internal structure 
   We only copy entries != 0 to increase performance.
   Overwrites ipt_acc_tmpbuf.
*/
int ipt_acc_handle_get_data(unsigned int handle, void *to_user)
{
    unsigned int tmpbuf_pos=0;

    if (handle >= ACCOUNT_MAX_HANDLES) {
        printk("ACCOUNT: invalid handle for ipt_acc_handle_get_data() "
               "specified: %u\n", handle);
        return -1;
    }

    if (ipt_acc_handles[handle].data == NULL) {
        printk("ACCOUNT: handle %u is BROKEN: Contains no data\n", handle);
        return -1;
    }

    unsigned int net_ip = ipt_acc_handles[handle].ip;
    unsigned int depth = ipt_acc_handles[handle].depth;

    /* 8 bit network */
    if (depth == 0) {
        struct ipt_acc_mask_24 *network = 
            (struct ipt_acc_mask_24*)ipt_acc_handles[handle].data;
        ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network, net_ip, 0);
        
        /* Flush remaining data to userspace */
        if (tmpbuf_pos)
            copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos);

        return 0;
    }

    /* 16 bit network */
    if (depth == 1) {
        struct ipt_acc_mask_16 *network_16 = 
            (struct ipt_acc_mask_16*)ipt_acc_handles[handle].data;
        unsigned int b;
        for (b = 0; b <= 255; b++) {
            if (network_16->mask_24[b]) {
                struct ipt_acc_mask_24 *network = 
                    (struct ipt_acc_mask_24*)network_16->mask_24[b];
                ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network, 
                                             net_ip, (b << 16));
            }
        }

        /* Flush remaining data to userspace */
        if (tmpbuf_pos)
            copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos);

        return 0;
    }

    /* 24 bit network */
    if (depth == 2) {
        struct ipt_acc_mask_8 *network_8 = 
            (struct ipt_acc_mask_8*)ipt_acc_handles[handle].data;
        unsigned int a, b;
        for (a = 0; a <= 255; a++) {
            if (network_8->mask_16[a]) {
                struct ipt_acc_mask_16 *network_16 = 
                    (struct ipt_acc_mask_16*)network_8->mask_16[a];
                for (b = 0; b <= 255; b++) {
                    if (network_16->mask_24[b]) {
                        struct ipt_acc_mask_24 *network = 
                            (struct ipt_acc_mask_24*)network_16->mask_24[b];
                        ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network,
                                                 net_ip, (a << 8) | (b << 16));
                    }
                }
            }
        }

        /* Flush remaining data to userspace */
        if (tmpbuf_pos)
            copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos);

        return 0;
    }
    
    return -1;
}

static int ipt_acc_set_ctl(struct sock *sk, int cmd, 
                               void *user, unsigned int len)
{
    struct ipt_acc_handle_sockopt handle;
    int ret = -EINVAL;

    if (!capable(CAP_NET_ADMIN))
        return -EPERM;

    switch (cmd) {
    case IPT_SO_SET_ACCOUNT_HANDLE_FREE:
        if (len != sizeof(struct ipt_acc_handle_sockopt)) {
            printk("ACCOUNT: ipt_acc_set_ctl: wrong data size (%u != %u) "
                   "for IPT_SO_SET_HANDLE_FREE\n", 
                   len, sizeof(struct ipt_acc_handle_sockopt));
            break;
        }

        if (copy_from_user (&handle, user, len)) {
            printk("ACCOUNT: ipt_acc_set_ctl: copy_from_user failed for "
                   "IPT_SO_SET_HANDLE_FREE\n");
            break;
        }

        spin_lock_bh(&ipt_acc_userspace_lock);
        ret = ipt_acc_handle_free(handle.handle_nr);
        spin_unlock_bh(&ipt_acc_userspace_lock);
        break;
    case IPT_SO_SET_ACCOUNT_HANDLE_FREE_ALL: {
            unsigned int i;
            spin_lock_bh(&ipt_acc_userspace_lock);
            for (i = 0; i < ACCOUNT_MAX_HANDLES; i++)
                ipt_acc_handle_free(i);
            spin_unlock_bh(&ipt_acc_userspace_lock);
            ret = 0;
            break;
        }
    default:
        printk("ACCOUNT: ipt_acc_set_ctl: unknown request %i\n", cmd);
    }

    return ret;
}

static int ipt_acc_get_ctl(struct sock *sk, int cmd, void *user, int *len)
{
    struct ipt_acc_handle_sockopt handle;
    int ret = -EINVAL;

    if (!capable(CAP_NET_ADMIN))
        return -EPERM;

    switch (cmd) {
    case IPT_SO_GET_ACCOUNT_PREPARE_READ_FLUSH:
    case IPT_SO_GET_ACCOUNT_PREPARE_READ:
        if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
            printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u) "
                   "for IPT_SO_GET_ACCOUNT_PREPARE_READ/READ_FLUSH\n",
                   *len, sizeof(struct ipt_acc_handle_sockopt));
            break;
        }

        if (copy_from_user (&handle, user, 
                            sizeof(struct ipt_acc_handle_sockopt))) {
            return -EFAULT;
            break;
        }

        spin_lock_bh(&ipt_acc_lock);
        spin_lock_bh(&ipt_acc_userspace_lock);
        if (cmd == IPT_SO_GET_ACCOUNT_PREPARE_READ_FLUSH)
            handle.handle_nr = ipt_acc_handle_prepare_read_flush(
                                   handle.name, &handle.itemcount);
        else
            handle.handle_nr = ipt_acc_handle_prepare_read(
                                   handle.name, &handle.itemcount);
        spin_unlock_bh(&ipt_acc_userspace_lock);
        spin_unlock_bh(&ipt_acc_lock);

        if (handle.handle_nr == -1) {
            return -EINVAL;
            break;
        }

        if (copy_to_user(user, &handle, 
                         sizeof(struct ipt_acc_handle_sockopt))) {
            return -EFAULT;
            break;
        }
        ret = 0;
        break;
    case IPT_SO_GET_ACCOUNT_GET_DATA:
        if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
            printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u)"
                   " for IPT_SO_GET_ACCOUNT_PREPARE_READ/READ_FLUSH\n",
                   *len, sizeof(struct ipt_acc_handle_sockopt));
            break;
        }

        if (copy_from_user (&handle, user, 
                            sizeof(struct ipt_acc_handle_sockopt))) {
            return -EFAULT;
            break;
        }

        if (handle.handle_nr >= ACCOUNT_MAX_HANDLES) {
            return -EINVAL;
            break;
        }

        if (*len < ipt_acc_handles[handle.handle_nr].itemcount
                   * sizeof(struct ipt_acc_handle_ip)) {
            printk("ACCOUNT: ipt_acc_get_ctl: not enough space (%u < %u)"
                   " to store data from IPT_SO_GET_ACCOUNT_GET_DATA\n",
                   *len, ipt_acc_handles[handle.handle_nr].itemcount
                   * sizeof(struct ipt_acc_handle_ip));
            ret = -ENOMEM;
            break;
        }

        spin_lock_bh(&ipt_acc_userspace_lock);
        ret = ipt_acc_handle_get_data(handle.handle_nr, user);
        spin_unlock_bh(&ipt_acc_userspace_lock);
        if (ret) {
            printk("ACCOUNT: ipt_acc_get_ctl: ipt_acc_handle_get_data"
                   " failed for handle %u\n", handle.handle_nr);
            break;
        }

        ret = 0;
        break;
    case IPT_SO_GET_ACCOUNT_GET_HANDLE_USAGE: {
            if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
                printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u)"
                       " for IPT_SO_GET_ACCOUNT_GET_HANDLE_USAGE\n",
                       *len, sizeof(struct ipt_acc_handle_sockopt));
                break;
            }

            /* Find out how many handles are in use */
            unsigned int i;
            handle.itemcount = 0;
            spin_lock_bh(&ipt_acc_userspace_lock);
            for (i = 0; i < ACCOUNT_MAX_HANDLES; i++)
                if (ipt_acc_handles[i].data)
                    handle.itemcount++;
            spin_unlock_bh(&ipt_acc_userspace_lock);

            if (copy_to_user(user, &handle, 
                             sizeof(struct ipt_acc_handle_sockopt))) {
                return -EFAULT;
                break;
            }
            ret = 0;
            break;
        }
    case IPT_SO_GET_ACCOUNT_GET_TABLE_NAMES: {
            spin_lock_bh(&ipt_acc_lock);

            /* Determine size of table names */
            unsigned int size = 0, i;
            for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
                if (ipt_acc_tables[i].name[0] != 0)
                    size += strlen (ipt_acc_tables[i].name) + 1;
            }
            size += 1;    /* Terminating NULL character */

            if (*len < size) {
                spin_unlock_bh(&ipt_acc_lock);
                printk("ACCOUNT: ipt_acc_get_ctl: not enough space (%u < %u)"
                       " to store table names\n", *len, size);
                ret = -ENOMEM;
                break;
            }
            /* Copy table names to userspace */
            char *tnames = user;
            for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
                if (ipt_acc_tables[i].name[0] != 0) {
                    int len = strlen (ipt_acc_tables[i].name) + 1;
                    /* copy string + terminating zero */
                    copy_to_user(tnames, ipt_acc_tables[i].name, len);
                    tnames += len;
                }
            }
            /* Append terminating zero */
            i = 0;
            copy_to_user(tnames, &i, 1);
            spin_unlock_bh(&ipt_acc_lock);
            ret = 0;
            break;
        }
    default:
        printk("ACCOUNT: ipt_acc_get_ctl: unknown request %i\n", cmd);
    }

    return ret;
}

static struct ipt_target ipt_acc_reg = {
    {
        NULL, NULL
    },
    "ACCOUNT",
    ipt_acc_target,
    ipt_acc_checkentry,
    ipt_acc_deleteentry,
    THIS_MODULE
};

static struct nf_sockopt_ops ipt_acc_sockopts = {
    {
        NULL, NULL
    },
    PF_INET,
    IPT_SO_SET_ACCOUNT_HANDLE_FREE,
    IPT_SO_SET_ACCOUNT_MAX+1,
    ipt_acc_set_ctl,
    IPT_SO_GET_ACCOUNT_PREPARE_READ,
    IPT_SO_GET_ACCOUNT_MAX+1,
    ipt_acc_get_ctl,
    0,
    NULL
};

static int __init init(void)
{
    if ((ipt_acc_tables = 
         kmalloc(ACCOUNT_MAX_TABLES * 
                 sizeof(struct ipt_acc_table), GFP_KERNEL)) == NULL) {
        printk("ACCOUNT: Out of memory allocating account_tables structure");
        return -EINVAL;
    }
    memset(ipt_acc_tables, 0, 
           ACCOUNT_MAX_TABLES * sizeof(struct ipt_acc_table));

    if ((ipt_acc_handles = 
         kmalloc(ACCOUNT_MAX_HANDLES * 
                 sizeof(struct ipt_acc_handle), GFP_KERNEL)) == NULL) {
        printk("ACCOUNT: Out of memory allocating account_handles structure");
        kfree (ipt_acc_tables);
        ipt_acc_tables = NULL;
        return -EINVAL;
    }
    memset(ipt_acc_handles, 0, 
           ACCOUNT_MAX_HANDLES * sizeof(struct ipt_acc_handle));

    /* Allocate one page as temporary storage */
    if ((ipt_acc_tmpbuf = (void*)__get_free_page(GFP_KERNEL)) == NULL) {
        printk("ACCOUNT: Out of memory for temporary buffer page\n");
        kfree(ipt_acc_tables);
        kfree(ipt_acc_handles);
        ipt_acc_tables = NULL;
        ipt_acc_handles = NULL;
        return -EINVAL;
    }

    /* Register setsockopt */
    if (nf_register_sockopt(&ipt_acc_sockopts) < 0) {
        printk("ACCOUNT: Can't register sockopts. Aborting\n");

        kfree(ipt_acc_tables);
        kfree(ipt_acc_handles);
        free_page((unsigned long)ipt_acc_tmpbuf);
        ipt_acc_tables = NULL;
        ipt_acc_handles = NULL;
        ipt_acc_tmpbuf = NULL;

        return -EINVAL;
    }

    if (ipt_register_target(&ipt_acc_reg))
        return -EINVAL;

    return 0;
}

static void __exit fini(void)
{
    ipt_unregister_target(&ipt_acc_reg);

    nf_unregister_sockopt(&ipt_acc_sockopts);

    kfree(ipt_acc_tables);
    kfree(ipt_acc_handles);
    free_page((unsigned long)ipt_acc_tmpbuf);

    ipt_acc_tables = NULL;
    ipt_acc_handles = NULL;
    ipt_acc_tmpbuf = NULL;
}

module_init(init);
module_exit(fini);
MODULE_LICENSE("GPL");

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: new ipt_ACCOUNT version
  2004-05-29 13:27 new ipt_ACCOUNT version Thomas Jarosch
@ 2004-06-06 22:40 ` Patrick McHardy
  2004-06-07 16:11 ` Patrick McHardy
       [not found] ` <200406091133.59850.thomas.jarosch@intra2net.com>
  2 siblings, 0 replies; 8+ messages in thread
From: Patrick McHardy @ 2004-06-06 22:40 UTC (permalink / raw)
  To: Thomas Jarosch; +Cc: netfilter-devel

Hi Thomas,

sorry for the late reply, I missed your email. Please also put
me in the receipients when writing to me, I don't always read
all mails on the list.

Thomas Jarosch wrote:
> Hi Patrick,
> 
> I now fixed the too long lines issue. Attached is the new version.
> Is it CVS ready?

It looks a lot better, thanks. I'm going to look at it more deeply
tomorrow and let you know.

Regards
Patrick

> 
> The website (http://www.intra2net.com/opensource/ipt_ACCOUNT/) contains
> the updated userspace counterparts (with short line lenghts, too ;-))
> 
> Thomas
> 
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: new ipt_ACCOUNT version
  2004-05-29 13:27 new ipt_ACCOUNT version Thomas Jarosch
  2004-06-06 22:40 ` Patrick McHardy
@ 2004-06-07 16:11 ` Patrick McHardy
  2004-06-07 19:48   ` Thomas Jarosch
       [not found] ` <200406091133.59850.thomas.jarosch@intra2net.com>
  2 siblings, 1 reply; 8+ messages in thread
From: Patrick McHardy @ 2004-06-07 16:11 UTC (permalink / raw)
  To: Thomas Jarosch; +Cc: netfilter-devel

Thomas Jarosch wrote:
> I now fixed the too long lines issue. Attached is the new version.
> Is it CVS ready?

There are still some things that should to be fixed, see my comments
below.

Regards
Patrick

> 
> The website (http://www.intra2net.com/opensource/ipt_ACCOUNT/) contains
> the updated userspace counterparts (with short line lenghts, too ;-))
> 
> Thomas
> 
> 
> ------------------------------------------------------------------------
> 
> /***************************************************************************
>  *   Copyright (C) 2004 by Intra2net AG                                    *
>  *   opensource@intra2net.com                                              *
>  *                                                                         *
>  *   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 ACCOUNT_MAX_TABLES 32
> #define ACCOUNT_TABLE_NAME_LEN 32
> #define ACCOUNT_MAX_HANDLES 10
> 
> /* Structure for the userspace part of ipt_ACCOUNT */
> struct ipt_acc_info {
>     u_int32_t net_ip;
>     u_int32_t net_mask;
>     char table_name[ACCOUNT_TABLE_NAME_LEN];
>     int32_t table_nr;
> };
> 
> /* Internal table structure, generated by check_entry() */
> struct ipt_acc_table {
>     char name[ACCOUNT_TABLE_NAME_LEN];        /* name of the table */
>     unsigned int ip;                          /* base IP of network */
>     unsigned int netmask;                     /* netmask of the network */
^^^ use u_int32_t

>     unsigned char depth;                      /* size of network: 
>                                                  0: 8 bit, 1: 16bit, 2: 24 bit */
>     unsigned int refcount;                    /* refcount of this table. 
>                                                  if zero, destroy it */
>     unsigned int itemcount;                   /* number of IPs in this table */
>     void *data;                               /* pointer to the actual data, 
>                                                  depending on netmask */
> };
> 
> /* Internal handle structure */
> struct ipt_acc_handle {
>     unsigned int ip;                          /* base IP of network. Used for 
>                                                  caculating the final IP during
>                                                  get_data() */
^^^ also here
>     unsigned char depth;                      /* size of network. See above for 
>                                                  details */
>     unsigned int itemcount;                   /* number of IPs in this table */
>     void *data;                               /* pointer to the actual data, 
>                                                  depending on size */
> };
> 
> /* Handle structure for communication with the userspace library */
> struct ipt_acc_handle_sockopt {
>     unsigned int handle_nr;                   /* Used for HANDLE_FREE */
>     char name[ACCOUNT_TABLE_NAME_LEN];        /* Used for HANDLE_PREPARE_READ/
>                                                  HANDLE_READ_FLUSH */
>     unsigned int itemcount;                   /* Used for HANDLE_PREPARE_READ/
>                                                  HANDLE_READ_FLUSH */
> };
> 
> /* Used for every IP entry 
>    Size is 16 bytes so that 256 (class C network) * 16 
>    fits in one kernel (zero) page */
> struct ipt_acc_ip {
>     unsigned int src_packets;
>     unsigned int src_bytes;
>     unsigned int dst_packets;
>     unsigned int dst_bytes;
> };
> 
> /*
>     Used for every IP when returning data
> */
> struct ipt_acc_handle_ip {
>     unsigned int ip;
>     unsigned int src_packets;
>     unsigned int src_bytes;
>     unsigned int dst_packets;
>     unsigned int dst_bytes;
> };

^^^ It's probably better to use fixed size types for communication with
userspace to avoid problems on 64bit machines with 32bit binaries.

> 
> /*
>     The IPs are organized as an array so that direct slot
>     calculations are possible.
>     Only 8 bit networks are preallocated, 16/24 bit networks
>     allocate their slots when needed -> very efficent.
> */
> struct ipt_acc_mask_24 {
>     struct ipt_acc_ip ip[256];
> };
> 
> struct ipt_acc_mask_16 {
>     struct ipt_acc_mask_24 *mask_24[256];
> };
> 
> struct ipt_acc_mask_8 {
>     struct ipt_acc_mask_16 *mask_16[256];
> };
> 
> #endif /*_IPT_ACCOUNT_H*/
> 
> 
> ------------------------------------------------------------------------
> 
> /***************************************************************************
>  *   This is a module which is used for counting packets.                  *
>  *   See http://www.intra2net.com/opensource/ipt_account                   *
>  *   for further information                                               *
>  *                                                                         * 
>  *   Copyright (C) 2004 by Intra2net AG                                    *
>  *   opensource@intra2net.com                                              *
>  *                                                                         *
>  *   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/module.h>
> #include <linux/skbuff.h>
> #include <linux/ip.h>
> #include <linux/spinlock.h>
> #include <net/icmp.h>
> #include <net/udp.h>
> #include <net/tcp.h>
> #include <linux/netfilter_ipv4/ip_tables.h>

Looks like some are missing:

linux/kernel.h printk, NIPQUAD
linux/gfp.h get_zeroed_page/free_page
linux/string.h strncpy, memset, ..
asm/uaccess.h copy_from/to_user

> 
> struct in_device;
> #include <net/route.h>
> #include <linux/netfilter_ipv4/ipt_ACCOUNT.h>
> 
> #if 0
> #define DEBUGP printk
> #else
> #define DEBUGP(format, args...)
> #endif
> 
> #if (PAGE_SIZE < 4096)
> #error "ipt_ACCOUNT needs at least a PAGE_SIZE of 4096"
> #endif
> 
> struct ipt_acc_table *ipt_acc_tables = NULL;
> struct ipt_acc_handle *ipt_acc_handles = NULL;
> void *ipt_acc_tmpbuf = NULL;
> 
> /* Spinlock used for manipulating the current accounting tables/data */
> static spinlock_t ipt_acc_lock = SPIN_LOCK_UNLOCKED;
> /* Spinlock used for manipulating userspace handles/snapshot data */
> static spinlock_t ipt_acc_userspace_lock = SPIN_LOCK_UNLOCKED;
> 
> 
> /* Recursive free of all data structures */
> void ipt_acc_data_free(void *data, unsigned char depth)
> {
>     /* Empty data set */
>     if (!data)
>         return;
> 
>     /* Free for 8 bit network */
>     if (depth == 0) {
>         free_page((unsigned long)data);
>         data = NULL;
^^ unneccessary, some more below

>         return;
>     }
> 
>     /* Free for 16 bit network */
>     if (depth == 1) {
>         struct ipt_acc_mask_16 *mask_16 = (struct ipt_acc_mask_16 *)data;
>         unsigned int b;
>         for (b=0; b <= 255; b++) {
>             if (mask_16->mask_24[b] != 0) {
>                 free_page((unsigned long)mask_16->mask_24[b]);
>                 mask_16->mask_24[b] = NULL;
^^^

>             }
>         }
>         free_page((unsigned long)data);
>         data = NULL;
^^^

>         return;
>     }
> 
>     /* Free for 24 bit network */
>     if (depth == 2) {
>         unsigned int a, b;
>         for (a=0; a <= 255; a++) {
>             if (((struct ipt_acc_mask_8 *)data)->mask_16[a]) {
>                 struct ipt_acc_mask_16 *mask_16 = (struct ipt_acc_mask_16*)
>                                    ((struct ipt_acc_mask_8 *)data)->mask_16[a];
>                 
>                 for (b=0; b <= 255; b++) {
>                     if (mask_16->mask_24[b]) {
>                         free_page((unsigned long)mask_16->mask_24[b]);
>                         mask_16->mask_24[b] = NULL;
^^^

>                     }
>                 }
>                 free_page((unsigned long)mask_16);
>                 mask_16 = NULL;
^^^

>             }
>         }
>         free_page((unsigned long)data);
>         data = NULL;
^^^

>         return;
>     }
> 
>     printk("ACCOUNT: ipt_acc_data_free called with unknown depth: %d\n", 
>            depth);
>     return;
> }
> 
> /* Look for existing table / insert new one. 
>    Return internal ID or -1 on error */
> int ipt_acc_table_insert(char *name, unsigned int ip, unsigned int netmask)
> {
>     unsigned int i;
> 
>     DEBUGP("ACCOUNT: ipt_acc_table_insert: %s, %u.%u.%u.%u/%u.%u.%u.%u\n",
>                                          name, NIPQUAD(ip), NIPQUAD(netmask));
> 
>     /* Look for existing table */
>     for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
>         if (strncmp(ipt_acc_tables[i].name, name, 
>                     ACCOUNT_TABLE_NAME_LEN) == 0) {
>             DEBUGP("ACCOUNT: Found existing slot: %d - "
>                    "%u.%u.%u.%u/%u.%u.%u.%u\n", i, 
>                    NIPQUAD(ipt_acc_tables[i].ip), 
>                    NIPQUAD(ipt_acc_tables[i].netmask));
> 
>             if (ipt_acc_tables[i].ip != ip 
>                 || ipt_acc_tables[i].netmask != netmask) {
>                 printk("ACCOUNT: Table %s found, but IP/netmask mismatch. "
>                        "IP/netmask found: %u.%u.%u.%u/%u.%u.%u.%u\n",
>                        name, NIPQUAD(ipt_acc_tables[i].ip), 
>                        NIPQUAD(ipt_acc_tables[i].netmask));
>                 return -1;
>             }
> 
>             ipt_acc_tables[i].refcount++;
>             DEBUGP("ACCOUNT: Refcount: %d\n", ipt_acc_tables[i].refcount);
>             return i;
>         }
>     }
> 
>     /* Insert new table */
>     for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
>         /* Found free slot */
>         if (ipt_acc_tables[i].name[0] == 0) {
>             DEBUGP("ACCOUNT: Found free slot: %d\n", i);
> 
>             strncpy (ipt_acc_tables[i].name, name, ACCOUNT_TABLE_NAME_LEN-1);
> 
>             ipt_acc_tables[i].ip = ip;
>             ipt_acc_tables[i].netmask = netmask;
> 


>             /* Calculate netsize */
>             unsigned int j, calc_mask, netsize=0;
>             calc_mask = htonl(netmask);
>             for (j = 31; j > 0; j--) {
>                 if (calc_mask&(1<<j))
>                     netsize++;
>                 else
>                     break;
>             }
^^^
should be >= 0, for 255.255.255.255 it results in netsize = 31.
Works anyway, but why not fix it ..

> 
>             /* Calculate depth from netsize */
>             if (netsize >= 24)
>                 ipt_acc_tables[i].depth = 0;
>             else if (netsize >= 16)
>                 ipt_acc_tables[i].depth = 1;
>             else if(netsize >= 8)
>                 ipt_acc_tables[i].depth = 2;
> 
>             DEBUGP("ACCOUNT: calculated netsize: %u -> "
>                    "ipt_acc_table depth %u\n", netsize, 
>                    ipt_acc_tables[i].depth);
> 
>             ipt_acc_tables[i].refcount++;
>             if ((ipt_acc_tables[i].data
>                 = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>                 printk("ACCOUNT: out of memory for data of table: %s\n", name);
>                 memset(&ipt_acc_tables[i], 0, 
>                        sizeof(struct ipt_acc_table));
>                 return -1;
>             }
> 
>             return i;
>         }
>     }
> 
>     /* No free slot found */
>     printk("ACCOUNT: No free table slot found (max: %d). "
>            "Please increase ACCOUNT_MAX_TABLES.\n", ACCOUNT_MAX_TABLES);
>     return -1;
> }
> 
> static int ipt_acc_checkentry(const char *tablename,
>                                   const struct ipt_entry *e,
>                                   void *targinfo,
>                                   unsigned int targinfosize,
>                                   unsigned int hook_mask)
> {
>     struct ipt_acc_info *info = targinfo;
> 
>     if (targinfosize != IPT_ALIGN(sizeof(struct ipt_acc_info))) {
>         DEBUGP("ACCOUNT: targinfosize %u != %u\n",
>                targinfosize, IPT_ALIGN(sizeof(struct ipt_acc_info)));
>         return 0;
>     }
> 
>     spin_lock_bh(&ipt_acc_lock);
>     int table_nr = ipt_acc_table_insert(info->table_name, info->net_ip,
>                                             info->net_mask);
>     if (table_nr == -1) {
>         printk("ACCOUNT: Table insert problem. Aborting\n");
>         spin_unlock_bh(&ipt_acc_lock);
>         return 0;
>     }
>     /* Table nr caching so we don't have to do an extra string compare 
>        for every packet */
>     info->table_nr = table_nr;
> 
>     spin_unlock_bh(&ipt_acc_lock);
> 
>     return 1;
> }
> 
> void ipt_acc_deleteentry(void *targinfo, unsigned int targinfosize)
> {
>     unsigned int i;
>     struct ipt_acc_info *info = targinfo;
> 
>     if (targinfosize != IPT_ALIGN(sizeof(struct ipt_acc_info))) {
>         DEBUGP("ACCOUNT: targinfosize %u != %u\n",
>                targinfosize, IPT_ALIGN(sizeof(struct ipt_acc_info)));
>     }
> 
>     spin_lock_bh(&ipt_acc_lock);
> 
>     DEBUGP("ACCOUNT: ipt_acc_deleteentry called for table: %s (#%d)\n", 
>            info->table_name, info->table_nr);
> 
>     info->table_nr = -1;    /* Set back to original state */
> 
>     /* Look for table */
>     for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
>         if (strncmp(ipt_acc_tables[i].name, info->table_name, 
>                     ACCOUNT_TABLE_NAME_LEN) == 0) {
>             DEBUGP("ACCOUNT: Found table at slot: %d\n", i);
> 
>             ipt_acc_tables[i].refcount--;
>             DEBUGP("ACCOUNT: Refcount left: %d\n", 
>                    ipt_acc_tables[i].refcount);
> 
>             /* Table not needed anymore? */
>             if (ipt_acc_tables[i].refcount == 0) {
>                 DEBUGP("ACCOUNT: Destroying table at slot: %d\n", i);
>                 ipt_acc_data_free(ipt_acc_tables[i].data, 
>                                       ipt_acc_tables[i].depth);
>                 memset(&ipt_acc_tables[i], 0, 
>                        sizeof(struct ipt_acc_table));
>             }
> 
>             spin_unlock_bh(&ipt_acc_lock);
>             return;
>         }
>     }
> 
>     /* Table not found */
>     printk("ACCOUNT: Table %s not found for destroy\n", info->table_name);
>     spin_unlock_bh(&ipt_acc_lock);
> }
> 
> void ipt_acc_depth0_insert(struct ipt_acc_mask_24 *mask_24,
>                                unsigned int net_ip, unsigned int netmask,
>                                unsigned int src_ip, unsigned int dst_ip,
>                                unsigned int size, unsigned int *itemcount)
> {
>     unsigned char is_src = 0, is_dst = 0;
> 
>     DEBUGP("ACCOUNT: ipt_acc_depth0_insert: %u.%u.%u.%u/%u.%u.%u.%u "
>            "for net %u.%u.%u.%u/%u.%u.%u.%u, size: %u\n", NIPQUAD(src_ip), 
>            NIPQUAD(dst_ip), NIPQUAD(net_ip), NIPQUAD(netmask), size);
> 
>     /* Check if src/dst is inside our network. */
>     /* Special: net_ip = 0.0.0.0/0 gets stored as src in slot 0 */
>     if (!netmask)
>         src_ip = 0;
>     if ((net_ip&netmask) == (src_ip&netmask))
>         is_src = 1;
>     if ((net_ip&netmask) == (dst_ip&netmask) && netmask)
>         is_dst = 1;
> 
>     if (!is_src && !is_dst) {
>         DEBUGP("ACCOUNT: Skipping packet %u.%u.%u.%u/%u.%u.%u.%u "
>                "for net %u.%u.%u.%u/%u.%u.%u.%u\n", NIPQUAD(src_ip), 
>                NIPQUAD(dst_ip), NIPQUAD(net_ip), NIPQUAD(netmask));
>         return;
>     }
> 
>     /* Check if this entry is new */
>     char is_src_new_ip = 0, is_dst_new_ip = 0;
> 
>     /* Calculate array positions */
>     unsigned char src_slot = (unsigned char)((src_ip&0xFF000000) >> 24);
>     unsigned char dst_slot = (unsigned char)((dst_ip&0xFF000000) >> 24);
> 
>     /* Increase size counters */
>     if (is_src) {
>         /* Calculate network slot */
>         DEBUGP("ACCOUNT: Calculated SRC 8 bit network slot: %d\n", src_slot);
>         if (!mask_24->ip[src_slot].src_packets 
>             && !mask_24->ip[src_slot].dst_packets)
>             is_src_new_ip = 1;
> 
>         mask_24->ip[src_slot].src_packets++;
>         mask_24->ip[src_slot].src_bytes+=size;
>     }
>     if (is_dst) {
>         DEBUGP("ACCOUNT: Calculated DST 8 bit network slot: %d\n", dst_slot);
>         if (!mask_24->ip[dst_slot].src_packets 
>             && !mask_24->ip[dst_slot].dst_packets)
>             is_dst_new_ip = 1;
> 
>         mask_24->ip[dst_slot].dst_packets++;
>         mask_24->ip[dst_slot].dst_bytes+=size;
>     }
> 
>     /* Increase itemcounter */
>     DEBUGP("ACCOUNT: Itemcounter before: %d\n", *itemcount);
>     if (src_slot == dst_slot) {
>         if (is_src_new_ip || is_dst_new_ip) {
>             DEBUGP("ACCOUNT: src_slot == dst_slot: %d, %d\n", 
>                    is_src_new_ip, is_dst_new_ip);
>             (*itemcount)++;
>         }
>     } else {
>         if (is_src_new_ip) {
>             DEBUGP("ACCOUNT: New src_ip: %u.%u.%u.%u\n", NIPQUAD(src_ip));
>             (*itemcount)++;
>         }
>         if (is_dst_new_ip) {
>             DEBUGP("ACCOUNT: New dst_ip: %u.%u.%u.%u\n", NIPQUAD(dst_ip));
>             (*itemcount)++;
>         }
>     }
>     DEBUGP("ACCOUNT: Itemcounter after: %d\n", *itemcount);
> }
> 
> void ipt_acc_depth1_insert(struct ipt_acc_mask_16 *mask_16, 
>                                unsigned int net_ip, unsigned int netmask, 
>                                unsigned int src_ip, unsigned int dst_ip,
>                                unsigned int size, unsigned int *itemcount)
> {
>     /* Do we need to process src IP? */
>     if ((net_ip&netmask) == (src_ip&netmask)) {
>         unsigned char slot = (unsigned char)((src_ip&0x00FF0000) >> 16);
>         DEBUGP("ACCOUNT: Calculated SRC 16 bit network slot: %d\n", slot);
> 
>         /* Do we need to create a new mask_24 bucket? */
>         if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] = 
>              (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>             printk("ACCOUNT: Can't process packet because out of memory!\n");
>             return;
>         }
> 
>         ipt_acc_depth0_insert((struct ipt_acc_mask_24 *)mask_16->mask_24[slot],
>                                   net_ip, netmask, src_ip, 0, size, itemcount);
>     }
> 
>     /* Do we need to process dst IP? */
>     if ((net_ip&netmask) == (dst_ip&netmask)) {
>         unsigned char slot = (unsigned char)((dst_ip&0x00FF0000) >> 16);
>         DEBUGP("ACCOUNT: Calculated DST 16 bit network slot: %d\n", slot);
> 
>         /* Do we need to create a new mask_24 bucket? */
>         if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] 
>             = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>             printk("ACCOUT: Can't process packet because out of memory!\n");
>             return;
>         }
> 
>         ipt_acc_depth0_insert((struct ipt_acc_mask_24 *)mask_16->mask_24[slot],
>                                   net_ip, netmask, 0, dst_ip, size, itemcount);
>     }
> }
> 
> void ipt_acc_depth2_insert(struct ipt_acc_mask_8 *mask_8, 
>                                unsigned int net_ip, unsigned int netmask,
>                                unsigned int src_ip, unsigned int dst_ip,
>                                unsigned int size, unsigned int *itemcount)
> {
>     /* Do we need to process src IP? */
>     if ((net_ip&netmask) == (src_ip&netmask)) {
>         unsigned char slot = (unsigned char)((src_ip&0x0000FF00) >> 8);
>         DEBUGP("ACCOUNT: Calculated SRC 24 bit network slot: %d\n", slot);
> 
>         /* Do we need to create a new mask_24 bucket? */
>         if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot] 
>             = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>             printk("ACCOUNT: Can't process packet because out of memory!\n");
>             return;
>         }
> 
>         ipt_acc_depth1_insert((struct ipt_acc_mask_16 *)mask_8->mask_16[slot],
>                                   net_ip, netmask, src_ip, 0, size, itemcount);
>     }
> 
>     /* Do we need to process dst IP? */
>     if ((net_ip&netmask) == (dst_ip&netmask)) {
>         unsigned char slot = (unsigned char)((dst_ip&0x0000FF00) >> 8);
>         DEBUGP("ACCOUNT: Calculated DST 24 bit network slot: %d\n", slot);
> 
>         /* Do we need to create a new mask_24 bucket? */
>         if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot] 
>             = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>             printk("ACCOUNT: Can't process packet because out of memory!\n");
>             return;
>         }
> 
>         ipt_acc_depth1_insert((struct ipt_acc_mask_16 *)mask_8->mask_16[slot],
>                                   net_ip, netmask, 0, dst_ip, size, itemcount);
>     }
> }
> 
> static unsigned int ipt_acc_target(struct sk_buff **pskb,
>                                        unsigned int hooknum,
>                                        const struct net_device *in,
>                                        const struct net_device *out,
>                                        const void *targinfo,
>                                        void *userinfo)
> {
>     const struct ipt_acc_info *info = 
>         (const struct ipt_acc_info *)targinfo;
>     unsigned int src_ip = (*pskb)->nh.iph->saddr;
>     unsigned int dst_ip = (*pskb)->nh.iph->daddr;
>     unsigned int size = ntohs((*pskb)->nh.iph->tot_len);
> 
>     spin_lock_bh(&ipt_acc_lock);
> 
>     if (ipt_acc_tables[info->table_nr].name[0] == 0) {
>         printk("ACCOUNT: ipt_acc_target: Invalid table id %u. "
>                "IPs %u.%u.%u.%u/%u.%u.%u.%u\n", info->table_nr, 
>                NIPQUAD(src_ip), NIPQUAD(dst_ip));
>         spin_unlock_bh(&ipt_acc_lock);
>         return IPT_CONTINUE;
>     }
> 
>     /* 8 bit network or "any" network */
>     if (ipt_acc_tables[info->table_nr].depth == 0) {
>         /* Count packet and check if the IP is new */
>         ipt_acc_depth0_insert(
>             (struct ipt_acc_mask_24 *)ipt_acc_tables[info->table_nr].data,
>             ipt_acc_tables[info->table_nr].ip, 
>             ipt_acc_tables[info->table_nr].netmask,
>             src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
>         spin_unlock_bh(&ipt_acc_lock);
>         return IPT_CONTINUE;
>     }
> 
>     /* 16 bit network */
>     if (ipt_acc_tables[info->table_nr].depth == 1) {
>         ipt_acc_depth1_insert(
>             (struct ipt_acc_mask_16 *)ipt_acc_tables[info->table_nr].data,
>             ipt_acc_tables[info->table_nr].ip, 
>             ipt_acc_tables[info->table_nr].netmask,
>             src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
>         spin_unlock_bh(&ipt_acc_lock);
>         return IPT_CONTINUE;
>     }
> 
>     /* 24 bit network */
>     if (ipt_acc_tables[info->table_nr].depth == 2) {
>         ipt_acc_depth2_insert(
>             (struct ipt_acc_mask_8 *)ipt_acc_tables[info->table_nr].data,
>             ipt_acc_tables[info->table_nr].ip, 
>             ipt_acc_tables[info->table_nr].netmask,
>             src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
>         spin_unlock_bh(&ipt_acc_lock);
>         return IPT_CONTINUE;
>     }
> 
>     printk("ACCOUNT: ipt_acc_target: Unable to process packet. "
>            "Table id %u. IPs %u.%u.%u.%u/%u.%u.%u.%u\n", 
>            info->table_nr, NIPQUAD(src_ip), NIPQUAD(dst_ip));
> 
>     spin_unlock_bh(&ipt_acc_lock);
>     return IPT_CONTINUE;
> }
> 
> /*
>     Functions dealing with "handles":
>     Handles are snapshots of a accounting state.
>     
>     read snapshots are only for debugging the code
>     and are very expensive concerning speed/memory
>     compared to read_and_flush.
>     
>     The functions aren't protected by spinlocks themselves
>     as this is done in the ioctl part of the code.
> */
> 
> /*
>     Find a free handle slot. Normally only one should be used,
>     but there could be two or more applications accessing the data
>     at the same time.
> */
> int ipt_acc_handle_find_slot(void)
> {
>     unsigned int i;
>     /* Insert new table */
>     for (i = 0; i < ACCOUNT_MAX_HANDLES; i++) {
>         /* Found free slot */
>         if (ipt_acc_handles[i].data == NULL) {
>             /* Don't "mark" data as used as we are protected by a spinlock 
>                by the calling function. handle_find_slot() is only a function
>                to prevent code duplication. */
>             return i;
>         }
>     }
> 
>     /* No free slot found */
>     printk("ACCOUNT: No free handle slot found (max: %u). "
>            "Please increase ACCOUNT_MAX_HANDLES.\n", ACCOUNT_MAX_HANDLES);
>     return -1;
> }
> 
> int ipt_acc_handle_free(unsigned int handle)
> {
>     if (handle >= ACCOUNT_MAX_HANDLES) {
>         printk("ACCOUNT: Invalid handle for ipt_acc_handle_free() specified:"
>                " %u\n", handle);
>         return -EINVAL;
>     }
> 
>     ipt_acc_data_free(ipt_acc_handles[handle].data, 
>                           ipt_acc_handles[handle].depth);
>     memset (&ipt_acc_handles[handle], 0, sizeof (struct ipt_acc_handle));
>     return 0;
> }
> 
> /* Prepare data for read without flush. Use only for debugging!
>    Real applications should use read&flush as it's way more efficent */
> int ipt_acc_handle_prepare_read(char *tablename, unsigned int *count)
> {
>     int handle, i, table_nr=-1;
> 
>     for (i = 0; i < ACCOUNT_MAX_TABLES; i++)
>         if (strncmp(ipt_acc_tables[i].name, tablename, 
>             ACCOUNT_TABLE_NAME_LEN) == 0) {
>             table_nr = i;
>             break;
>         }
> 
>     if (table_nr == -1) {
>         printk("ACCOUNT: ipt_acc_handle_prepare_read(): "
>                "Table %s not found\n", tablename);
>         return -1;
>     }
> 
>     /* Can't find a free handle slot? */
>     if ((handle = ipt_acc_handle_find_slot()) == -1)
>         return -1;
> 
>     /* Fill up handle structure */
>     ipt_acc_handles[handle].ip = ipt_acc_tables[table_nr].ip;
>     ipt_acc_handles[handle].depth = ipt_acc_tables[table_nr].depth;
>     ipt_acc_handles[handle].itemcount = ipt_acc_tables[table_nr].itemcount;
> 
>     /* allocate "root" table */
>     if ((ipt_acc_handles[handle].data = 
>          (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>         printk("ACCOUNT: out of memory for root table "
>                "in ipt_acc_handle_prepare_read()\n");
>         memset (&ipt_acc_handles[handle], 0, 
>                 sizeof(struct ipt_acc_handle));
>         return -1;
>     }
> 
>     /* Recursive copy of complete data structure */
>     unsigned int depth = ipt_acc_handles[handle].depth;
>     if (depth == 0) {
>         memcpy(ipt_acc_handles[handle].data, 
>                ipt_acc_tables[table_nr].data, 
>                sizeof(struct ipt_acc_mask_24));
>     } else if (depth == 1) {
>         struct ipt_acc_mask_16 *src_16 = 
>             (struct ipt_acc_mask_16 *)ipt_acc_tables[table_nr].data;
>         struct ipt_acc_mask_16 *network_16 =
>             (struct ipt_acc_mask_16 *)ipt_acc_handles[handle].data;
>         unsigned int b;
> 
>         for (b = 0; b <= 255; b++) {
>             if (src_16->mask_24[b]) {
>                 if ((network_16->mask_24[b] = 
>                      (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>                     printk("ACCOUNT: out of memory during copy of 16 bit "
>                            "network in ipt_acc_handle_prepare_read()\n");
>                     ipt_acc_data_free(ipt_acc_handles[handle].data, depth);
>                     memset (&ipt_acc_handles[handle], 0, 
>                             sizeof(struct ipt_acc_handle));
>                     return -1;
>                 }
> 
>                 memcpy(network_16->mask_24[b], src_16->mask_24[b], 
>                        sizeof(struct ipt_acc_mask_24));
>             }
>         }
>     } else if(depth == 2) {
>         struct ipt_acc_mask_8 *src_8 = 
>             (struct ipt_acc_mask_8 *)ipt_acc_tables[table_nr].data;
>         struct ipt_acc_mask_8 *network_8 = 
>             (struct ipt_acc_mask_8 *)ipt_acc_handles[handle].data;
>         unsigned int a;
> 
>         for (a = 0; a <= 255; a++) {
>             if (src_8->mask_16[a]) {
>                 if ((network_8->mask_16[a] = 
>                      (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>                     printk("ACCOUNT: out of memory during copy of 24 bit network"
>                            " in ipt_acc_handle_prepare_read()\n");
>                     ipt_acc_data_free(ipt_acc_handles[handle].data, depth);
>                     memset (&ipt_acc_handles[handle], 0, 
>                             sizeof(struct ipt_acc_handle));
>                     return -1;
>                 }
> 
>                 memcpy(network_8->mask_16[a], src_8->mask_16[a], 
>                        sizeof(struct ipt_acc_mask_16));
> 
>                 struct ipt_acc_mask_16 *src_16 = src_8->mask_16[a];
>                 struct ipt_acc_mask_16 *network_16 = network_8->mask_16[a];
>                 unsigned int b;
> 
>                 for (b = 0; b <= 255; b++) {
>                     if (src_16->mask_24[b]) {
>                         if ((network_16->mask_24[b] = 
>                              (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
>                             printk("ACCOUNT: out of memory during copy of 16 bit"
>                                    " network in ipt_acc_handle_prepare_read()\n");
>                             ipt_acc_data_free(ipt_acc_handles[handle].data,
>                                                   depth);
>                             memset (&ipt_acc_handles[handle], 0, 
>                                     sizeof(struct ipt_acc_handle));
>                             return -1;
>                         }
> 
>                         memcpy(network_16->mask_24[b], src_16->mask_24[b], 
>                                sizeof(struct ipt_acc_mask_24));
>                     }
>                 }
>             }
>         }
>     }
> 
>     *count = ipt_acc_tables[table_nr].itemcount;
>     return handle;
> }
> 
> /* Prepare data for read and flush it */
> int ipt_acc_handle_prepare_read_flush(char *tablename, unsigned int *count)
> {
>     int handle, i, table_nr=-1;
> 
>     for (i = 0; i < ACCOUNT_MAX_TABLES; i++)
>         if (strncmp(ipt_acc_tables[i].name, tablename, 
>                     ACCOUNT_TABLE_NAME_LEN) == 0) {
>             table_nr = i;
>             break;
>         }
> 
>     if (table_nr == -1) {
>         printk("ACCOUNT: ipt_acc_handle_prepare_read_flush(): "
>                "Table %s not found\n", tablename);
>         return -1;
>     }
> 
>     /* Can't find a free handle slot? */
>     if ((handle = ipt_acc_handle_find_slot()) == -1)
>         return -1;
> 
>     /* Try to allocate memory */
>     void *new_data_page = (void*)get_zeroed_page(GFP_ATOMIC);
>     if (!new_data_page)
>     {
>         printk("ACCOUNT: ipt_acc_handle_prepare_read_flush(): "
>                "Out of memory!\n");
>         return -1;
>     }
> 
>     /* Fill up handle structure */
>     ipt_acc_handles[handle].ip = ipt_acc_tables[table_nr].ip;
>     ipt_acc_handles[handle].depth = ipt_acc_tables[table_nr].depth;
>     ipt_acc_handles[handle].itemcount = ipt_acc_tables[table_nr].itemcount;
>     ipt_acc_handles[handle].data = ipt_acc_tables[table_nr].data;
>     *count = ipt_acc_tables[table_nr].itemcount;
> 
>     /* "Flush" table data */
>     ipt_acc_tables[table_nr].data = new_data_page;
>     ipt_acc_tables[table_nr].itemcount = 0;
> 
>     return handle;
> }
> 
> /* Copy 8 bit network data into a prepared buffer.
>    We only copy entries != 0 to increase performance.
> */
> void ipt_acc_handle_copy_data(void *to_user, int *pos, 
>                                   struct ipt_acc_mask_24 *data,
>                                   unsigned int net_ip, unsigned int net_OR_mask)
> {
>     struct ipt_acc_handle_ip handle_ip;
>     unsigned int handle_ip_size = sizeof (struct ipt_acc_handle_ip);
>     
>     unsigned int i;
>     
>     for (i = 0; i <= 255; i++) {
>         if (data->ip[i].src_packets || data->ip[i].dst_packets) {
>             handle_ip.ip = net_ip | net_OR_mask | (i<<24);
>             
>             handle_ip.src_packets = data->ip[i].src_packets;
>             handle_ip.src_bytes = data->ip[i].src_bytes;
>             handle_ip.dst_packets = data->ip[i].dst_packets;
>             handle_ip.dst_bytes = data->ip[i].dst_bytes;
> 
>             /* Temporary buffer full? Flush to userspace */
>             if (*pos+handle_ip_size >= PAGE_SIZE) {
>                 copy_to_user(to_user, ipt_acc_tmpbuf, *pos);
>                 *pos = 0;
>             }
>             memcpy(ipt_acc_tmpbuf+*pos, &handle_ip, handle_ip_size);
>             *pos += handle_ip_size;
>         }
>     }
> }
>    
> /* Copy the data from our internal structure 
>    We only copy entries != 0 to increase performance.
>    Overwrites ipt_acc_tmpbuf.
> */
> int ipt_acc_handle_get_data(unsigned int handle, void *to_user)
> {
>     unsigned int tmpbuf_pos=0;
> 
>     if (handle >= ACCOUNT_MAX_HANDLES) {
>         printk("ACCOUNT: invalid handle for ipt_acc_handle_get_data() "
>                "specified: %u\n", handle);
>         return -1;
>     }
> 
>     if (ipt_acc_handles[handle].data == NULL) {
>         printk("ACCOUNT: handle %u is BROKEN: Contains no data\n", handle);
>         return -1;
>     }
> 
>     unsigned int net_ip = ipt_acc_handles[handle].ip;
>     unsigned int depth = ipt_acc_handles[handle].depth;
> 
>     /* 8 bit network */
>     if (depth == 0) {
>         struct ipt_acc_mask_24 *network = 
>             (struct ipt_acc_mask_24*)ipt_acc_handles[handle].data;
>         ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network, net_ip, 0);
>         
>         /* Flush remaining data to userspace */
>         if (tmpbuf_pos)
>             copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos);
^^^ missing check for failure
> 
>         return 0;
>     }
> 
>     /* 16 bit network */
>     if (depth == 1) {
>         struct ipt_acc_mask_16 *network_16 = 
>             (struct ipt_acc_mask_16*)ipt_acc_handles[handle].data;
>         unsigned int b;
>         for (b = 0; b <= 255; b++) {
>             if (network_16->mask_24[b]) {
>                 struct ipt_acc_mask_24 *network = 
>                     (struct ipt_acc_mask_24*)network_16->mask_24[b];
>                 ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network, 
>                                              net_ip, (b << 16));
>             }
>         }
> 
>         /* Flush remaining data to userspace */
>         if (tmpbuf_pos)
>             copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos);
^^^ missing check
> 
>         return 0;
>     }
> 
>     /* 24 bit network */
>     if (depth == 2) {
>         struct ipt_acc_mask_8 *network_8 = 
>             (struct ipt_acc_mask_8*)ipt_acc_handles[handle].data;
>         unsigned int a, b;
>         for (a = 0; a <= 255; a++) {
>             if (network_8->mask_16[a]) {
>                 struct ipt_acc_mask_16 *network_16 = 
>                     (struct ipt_acc_mask_16*)network_8->mask_16[a];
>                 for (b = 0; b <= 255; b++) {
>                     if (network_16->mask_24[b]) {
>                         struct ipt_acc_mask_24 *network = 
>                             (struct ipt_acc_mask_24*)network_16->mask_24[b];
>                         ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network,
>                                                  net_ip, (a << 8) | (b << 16));
>                     }
>                 }
>             }
>         }
> 
>         /* Flush remaining data to userspace */
>         if (tmpbuf_pos)
>             copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos);
^^^ missing check for failure
> 
>         return 0;
>     }
>     
>     return -1;
> }
> 
> static int ipt_acc_set_ctl(struct sock *sk, int cmd, 
>                                void *user, unsigned int len)
> {
>     struct ipt_acc_handle_sockopt handle;
>     int ret = -EINVAL;
> 
>     if (!capable(CAP_NET_ADMIN))
>         return -EPERM;
> 
>     switch (cmd) {
>     case IPT_SO_SET_ACCOUNT_HANDLE_FREE:
>         if (len != sizeof(struct ipt_acc_handle_sockopt)) {
>             printk("ACCOUNT: ipt_acc_set_ctl: wrong data size (%u != %u) "
>                    "for IPT_SO_SET_HANDLE_FREE\n", 
>                    len, sizeof(struct ipt_acc_handle_sockopt));
>             break;
>         }
> 
>         if (copy_from_user (&handle, user, len)) {
>             printk("ACCOUNT: ipt_acc_set_ctl: copy_from_user failed for "
>                    "IPT_SO_SET_HANDLE_FREE\n");
>             break;
>         }
> 
>         spin_lock_bh(&ipt_acc_userspace_lock);
>         ret = ipt_acc_handle_free(handle.handle_nr);
>         spin_unlock_bh(&ipt_acc_userspace_lock);
>         break;
>     case IPT_SO_SET_ACCOUNT_HANDLE_FREE_ALL: {
>             unsigned int i;
>             spin_lock_bh(&ipt_acc_userspace_lock);
>             for (i = 0; i < ACCOUNT_MAX_HANDLES; i++)
>                 ipt_acc_handle_free(i);
>             spin_unlock_bh(&ipt_acc_userspace_lock);
>             ret = 0;
>             break;
>         }
>     default:
>         printk("ACCOUNT: ipt_acc_set_ctl: unknown request %i\n", cmd);
>     }
> 
>     return ret;
> }
> 
> static int ipt_acc_get_ctl(struct sock *sk, int cmd, void *user, int *len)
> {
>     struct ipt_acc_handle_sockopt handle;
>     int ret = -EINVAL;
> 
>     if (!capable(CAP_NET_ADMIN))
>         return -EPERM;
> 
>     switch (cmd) {
>     case IPT_SO_GET_ACCOUNT_PREPARE_READ_FLUSH:
>     case IPT_SO_GET_ACCOUNT_PREPARE_READ:
>         if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
>             printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u) "
>                    "for IPT_SO_GET_ACCOUNT_PREPARE_READ/READ_FLUSH\n",
>                    *len, sizeof(struct ipt_acc_handle_sockopt));
>             break;
>         }
> 
>         if (copy_from_user (&handle, user, 
>                             sizeof(struct ipt_acc_handle_sockopt))) {
>             return -EFAULT;
>             break;
>         }
> 
>         spin_lock_bh(&ipt_acc_lock);
>         spin_lock_bh(&ipt_acc_userspace_lock);
>         if (cmd == IPT_SO_GET_ACCOUNT_PREPARE_READ_FLUSH)
>             handle.handle_nr = ipt_acc_handle_prepare_read_flush(
>                                    handle.name, &handle.itemcount);
>         else
>             handle.handle_nr = ipt_acc_handle_prepare_read(
>                                    handle.name, &handle.itemcount);
>         spin_unlock_bh(&ipt_acc_userspace_lock);
>         spin_unlock_bh(&ipt_acc_lock);
> 
>         if (handle.handle_nr == -1) {
>             return -EINVAL;
>             break;
>         }
> 
>         if (copy_to_user(user, &handle, 
>                          sizeof(struct ipt_acc_handle_sockopt))) {
>             return -EFAULT;
>             break;
>         }
>         ret = 0;
>         break;
>     case IPT_SO_GET_ACCOUNT_GET_DATA:
>         if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
>             printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u)"
>                    " for IPT_SO_GET_ACCOUNT_PREPARE_READ/READ_FLUSH\n",
>                    *len, sizeof(struct ipt_acc_handle_sockopt));
>             break;
>         }
> 
>         if (copy_from_user (&handle, user, 
>                             sizeof(struct ipt_acc_handle_sockopt))) {
>             return -EFAULT;
>             break;
>         }
> 
>         if (handle.handle_nr >= ACCOUNT_MAX_HANDLES) {
>             return -EINVAL;
>             break;
>         }
> 
>         if (*len < ipt_acc_handles[handle.handle_nr].itemcount
>                    * sizeof(struct ipt_acc_handle_ip)) {
>             printk("ACCOUNT: ipt_acc_get_ctl: not enough space (%u < %u)"
>                    " to store data from IPT_SO_GET_ACCOUNT_GET_DATA\n",
>                    *len, ipt_acc_handles[handle.handle_nr].itemcount
>                    * sizeof(struct ipt_acc_handle_ip));
>             ret = -ENOMEM;
>             break;
>         }
> 
>         spin_lock_bh(&ipt_acc_userspace_lock);
>         ret = ipt_acc_handle_get_data(handle.handle_nr, user);
>         spin_unlock_bh(&ipt_acc_userspace_lock);
>         if (ret) {
>             printk("ACCOUNT: ipt_acc_get_ctl: ipt_acc_handle_get_data"
>                    " failed for handle %u\n", handle.handle_nr);
>             break;
>         }
> 
>         ret = 0;
>         break;
>     case IPT_SO_GET_ACCOUNT_GET_HANDLE_USAGE: {
>             if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
>                 printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u)"
>                        " for IPT_SO_GET_ACCOUNT_GET_HANDLE_USAGE\n",
>                        *len, sizeof(struct ipt_acc_handle_sockopt));
>                 break;
>             }
> 
>             /* Find out how many handles are in use */
>             unsigned int i;
^^^ Does not compile with older gccs

>             handle.itemcount = 0;
>             spin_lock_bh(&ipt_acc_userspace_lock);
>             for (i = 0; i < ACCOUNT_MAX_HANDLES; i++)
>                 if (ipt_acc_handles[i].data)
>                     handle.itemcount++;
>             spin_unlock_bh(&ipt_acc_userspace_lock);
> 
>             if (copy_to_user(user, &handle, 
>                              sizeof(struct ipt_acc_handle_sockopt))) {
>                 return -EFAULT;
>                 break;
>             }
>             ret = 0;
>             break;
>         }
>     case IPT_SO_GET_ACCOUNT_GET_TABLE_NAMES: {
>             spin_lock_bh(&ipt_acc_lock);
> 
>             /* Determine size of table names */
>             unsigned int size = 0, i;
>             for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
>                 if (ipt_acc_tables[i].name[0] != 0)
>                     size += strlen (ipt_acc_tables[i].name) + 1;
>             }
>             size += 1;    /* Terminating NULL character */
> 
>             if (*len < size) {
>                 spin_unlock_bh(&ipt_acc_lock);
>                 printk("ACCOUNT: ipt_acc_get_ctl: not enough space (%u < %u)"
>                        " to store table names\n", *len, size);
>                 ret = -ENOMEM;
>                 break;
>             }
>             /* Copy table names to userspace */
>             char *tnames = user;
>             for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
>                 if (ipt_acc_tables[i].name[0] != 0) {
>                     int len = strlen (ipt_acc_tables[i].name) + 1;
>                     /* copy string + terminating zero */
>                     copy_to_user(tnames, ipt_acc_tables[i].name, len);
^^^ no check for failure


>                     tnames += len;
>                 }
>             }
>             /* Append terminating zero */
>             i = 0;
>             copy_to_user(tnames, &i, 1);
>             spin_unlock_bh(&ipt_acc_lock);
>             ret = 0;
>             break;
>         }
>     default:
>         printk("ACCOUNT: ipt_acc_get_ctl: unknown request %i\n", cmd);
>     }
> 
>     return ret;
> }
> 
> static struct ipt_target ipt_acc_reg = {
>     {
>         NULL, NULL
>     },
>     "ACCOUNT",
>     ipt_acc_target,
>     ipt_acc_checkentry,
>     ipt_acc_deleteentry,
>     THIS_MODULE
> };
> 
> static struct nf_sockopt_ops ipt_acc_sockopts = {
>     {
>         NULL, NULL
>     },
>     PF_INET,
>     IPT_SO_SET_ACCOUNT_HANDLE_FREE,
>     IPT_SO_SET_ACCOUNT_MAX+1,
>     ipt_acc_set_ctl,
>     IPT_SO_GET_ACCOUNT_PREPARE_READ,
>     IPT_SO_GET_ACCOUNT_MAX+1,
>     ipt_acc_get_ctl,
>     0,
>     NULL
> };

^^ Use C99 initializers, than you can avoid the 0 and NULL
initializations

> 
> static int __init init(void)
> {
>     if ((ipt_acc_tables = 
>          kmalloc(ACCOUNT_MAX_TABLES * 
>                  sizeof(struct ipt_acc_table), GFP_KERNEL)) == NULL) {
>         printk("ACCOUNT: Out of memory allocating account_tables structure");
>         return -EINVAL;
>     }
>     memset(ipt_acc_tables, 0, 
>            ACCOUNT_MAX_TABLES * sizeof(struct ipt_acc_table));
> 
>     if ((ipt_acc_handles = 
>          kmalloc(ACCOUNT_MAX_HANDLES * 
>                  sizeof(struct ipt_acc_handle), GFP_KERNEL)) == NULL) {
>         printk("ACCOUNT: Out of memory allocating account_handles structure");
>         kfree (ipt_acc_tables);
>         ipt_acc_tables = NULL;
>         return -EINVAL;
>     }
>     memset(ipt_acc_handles, 0, 
>            ACCOUNT_MAX_HANDLES * sizeof(struct ipt_acc_handle));
> 
>     /* Allocate one page as temporary storage */
>     if ((ipt_acc_tmpbuf = (void*)__get_free_page(GFP_KERNEL)) == NULL) {
>         printk("ACCOUNT: Out of memory for temporary buffer page\n");
>         kfree(ipt_acc_tables);
>         kfree(ipt_acc_handles);
>         ipt_acc_tables = NULL;
>         ipt_acc_handles = NULL;
>         return -EINVAL;
>     }
> 
>     /* Register setsockopt */
>     if (nf_register_sockopt(&ipt_acc_sockopts) < 0) {
>         printk("ACCOUNT: Can't register sockopts. Aborting\n");
> 
>         kfree(ipt_acc_tables);
>         kfree(ipt_acc_handles);
>         free_page((unsigned long)ipt_acc_tmpbuf);
>         ipt_acc_tables = NULL;
>         ipt_acc_handles = NULL;
>         ipt_acc_tmpbuf = NULL;
> 
>         return -EINVAL;
>     }
> 
>     if (ipt_register_target(&ipt_acc_reg))

^^^ leaking memory here. Move the cleanup-stuff down and
use goto's.


>         return -EINVAL;
> 
>     return 0;
> }
> 
> static void __exit fini(void)
> {
>     ipt_unregister_target(&ipt_acc_reg);
> 
>     nf_unregister_sockopt(&ipt_acc_sockopts);
> 
>     kfree(ipt_acc_tables);
>     kfree(ipt_acc_handles);
>     free_page((unsigned long)ipt_acc_tmpbuf);
> 
>     ipt_acc_tables = NULL;
>     ipt_acc_handles = NULL;
>     ipt_acc_tmpbuf = NULL;
^^^ Useless

> }
> 
> module_init(init);
> module_exit(fini);
> MODULE_LICENSE("GPL");

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: new ipt_ACCOUNT version
  2004-06-07 16:11 ` Patrick McHardy
@ 2004-06-07 19:48   ` Thomas Jarosch
  2004-06-09  8:39     ` Patrick McHardy
  0 siblings, 1 reply; 8+ messages in thread
From: Thomas Jarosch @ 2004-06-07 19:48 UTC (permalink / raw)
  To: Patrick McHardy; +Cc: netfilter-devel

[-- Attachment #1: Type: text/plain, Size: 403 bytes --]

Hi Patrick,

> > I now fixed the too long lines issue. Attached is the new version.
> > Is it CVS ready?
>
> There are still some things that should to be fixed, see my comments
> below.

Thanks for the in-depth review! I was amazed you found
the off-by-one "bug" in the netmask calculation code :-)

Attached is a new version with all the fixes.
I did some more old-gcc-compile fixing.

Cheers,
Thomas

[-- Attachment #2: ipt_ACCOUNT.h --]
[-- Type: text/x-chdr, Size: 3831 bytes --]

/***************************************************************************
 *   Copyright (C) 2004 by Intra2net AG                                    *
 *   opensource@intra2net.com                                              *
 *                                                                         *
 *   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 ACCOUNT_MAX_TABLES 32
#define ACCOUNT_TABLE_NAME_LEN 32
#define ACCOUNT_MAX_HANDLES 10

/* Structure for the userspace part of ipt_ACCOUNT */
struct ipt_acc_info {
    u_int32_t net_ip;
    u_int32_t net_mask;
    char table_name[ACCOUNT_TABLE_NAME_LEN];
    int32_t table_nr;
};

/* Internal table structure, generated by check_entry() */
struct ipt_acc_table {
    char name[ACCOUNT_TABLE_NAME_LEN];     /* name of the table */
    u_int32_t ip;                          /* base IP of network */
    u_int32_t netmask;                     /* netmask of the network */
    unsigned char depth;                   /* size of network: 
                                                 0: 8 bit, 1: 16bit, 2: 24 bit */
    u_int32_t refcount;                    /* refcount of this table. 
                                                 if zero, destroy it */
    u_int32_t itemcount;                   /* number of IPs in this table */
    void *data;                            /* pointer to the actual data, 
                                                 depending on netmask */
};

/* Internal handle structure */
struct ipt_acc_handle {
    u_int32_t ip;                          /* base IP of network. Used for 
                                                 caculating the final IP during
                                                 get_data() */
    unsigned char depth;                   /* size of network. See above for 
                                                 details */
    u_int32_t itemcount;                   /* number of IPs in this table */
    void *data;                            /* pointer to the actual data, 
                                                 depending on size */
};

/* Handle structure for communication with the userspace library */
struct ipt_acc_handle_sockopt {
    u_int32_t handle_nr;                   /* Used for HANDLE_FREE */
    char name[ACCOUNT_TABLE_NAME_LEN];     /* Used for HANDLE_PREPARE_READ/
                                                 HANDLE_READ_FLUSH */
    u_int32_t itemcount;                   /* Used for HANDLE_PREPARE_READ/
                                                 HANDLE_READ_FLUSH */
};

/* Used for every IP entry 
   Size is 16 bytes so that 256 (class C network) * 16 
   fits in one kernel (zero) page */
struct ipt_acc_ip {
    u_int32_t src_packets;
    u_int32_t src_bytes;
    u_int32_t dst_packets;
    u_int32_t dst_bytes;
};

/*
    Used for every IP when returning data
*/
struct ipt_acc_handle_ip {
    u_int32_t ip;
    u_int32_t src_packets;
    u_int32_t src_bytes;
    u_int32_t dst_packets;
    u_int32_t dst_bytes;
};

/*
    The IPs are organized as an array so that direct slot
    calculations are possible.
    Only 8 bit networks are preallocated, 16/24 bit networks
    allocate their slots when needed -> very efficent.
*/
struct ipt_acc_mask_24 {
    struct ipt_acc_ip ip[256];
};

struct ipt_acc_mask_16 {
    struct ipt_acc_mask_24 *mask_24[256];
};

struct ipt_acc_mask_8 {
    struct ipt_acc_mask_16 *mask_16[256];
};

#endif /*_IPT_ACCOUNT_H*/

[-- Attachment #3: ipt_ACCOUNT.c --]
[-- Type: text/x-csrc, Size: 38919 bytes --]

/***************************************************************************
 *   This is a module which is used for counting packets.                  *
 *   See http://www.intra2net.com/opensource/ipt_account                   *
 *   for further information                                               *
 *                                                                         * 
 *   Copyright (C) 2004 by Intra2net AG                                    *
 *   opensource@intra2net.com                                              *
 *                                                                         *
 *   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/module.h>
#include <linux/skbuff.h>
#include <linux/ip.h>
#include <linux/spinlock.h>
#include <net/icmp.h>
#include <net/udp.h>
#include <net/tcp.h>
#include <linux/netfilter_ipv4/ip_tables.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <asm/uaccess.h>

struct in_device;
#include <net/route.h>
#include <linux/netfilter_ipv4/ipt_ACCOUNT.h>

#if 0
#define DEBUGP printk
#else
#define DEBUGP(format, args...)
#endif

#if (PAGE_SIZE < 4096)
#error "ipt_ACCOUNT needs at least a PAGE_SIZE of 4096"
#endif

struct ipt_acc_table *ipt_acc_tables = NULL;
struct ipt_acc_handle *ipt_acc_handles = NULL;
void *ipt_acc_tmpbuf = NULL;

/* Spinlock used for manipulating the current accounting tables/data */
static spinlock_t ipt_acc_lock = SPIN_LOCK_UNLOCKED;
/* Spinlock used for manipulating userspace handles/snapshot data */
static spinlock_t ipt_acc_userspace_lock = SPIN_LOCK_UNLOCKED;


/* Recursive free of all data structures */
void ipt_acc_data_free(void *data, unsigned char depth)
{
    /* Empty data set */
    if (!data)
        return;

    /* Free for 8 bit network */
    if (depth == 0) {
        free_page((unsigned long)data);
        return;
    }

    /* Free for 16 bit network */
    if (depth == 1) {
        struct ipt_acc_mask_16 *mask_16 = (struct ipt_acc_mask_16 *)data;
        u_int32_t b;
        for (b=0; b <= 255; b++) {
            if (mask_16->mask_24[b] != 0) {
                free_page((unsigned long)mask_16->mask_24[b]);
            }
        }
        free_page((unsigned long)data);
        return;
    }

    /* Free for 24 bit network */
    if (depth == 2) {
        u_int32_t a, b;
        for (a=0; a <= 255; a++) {
            if (((struct ipt_acc_mask_8 *)data)->mask_16[a]) {
                struct ipt_acc_mask_16 *mask_16 = (struct ipt_acc_mask_16*)
                                   ((struct ipt_acc_mask_8 *)data)->mask_16[a];
                
                for (b=0; b <= 255; b++) {
                    if (mask_16->mask_24[b]) {
                        free_page((unsigned long)mask_16->mask_24[b]);
                    }
                }
                free_page((unsigned long)mask_16);
            }
        }
        free_page((unsigned long)data);
        return;
    }

    printk("ACCOUNT: ipt_acc_data_free called with unknown depth: %d\n", 
           depth);
    return;
}

/* Look for existing table / insert new one. 
   Return internal ID or -1 on error */
int ipt_acc_table_insert(char *name, u_int32_t ip, u_int32_t netmask)
{
    u_int32_t i;

    DEBUGP("ACCOUNT: ipt_acc_table_insert: %s, %u.%u.%u.%u/%u.%u.%u.%u\n",
                                         name, NIPQUAD(ip), NIPQUAD(netmask));

    /* Look for existing table */
    for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
        if (strncmp(ipt_acc_tables[i].name, name, 
                    ACCOUNT_TABLE_NAME_LEN) == 0) {
            DEBUGP("ACCOUNT: Found existing slot: %d - "
                   "%u.%u.%u.%u/%u.%u.%u.%u\n", i, 
                   NIPQUAD(ipt_acc_tables[i].ip), 
                   NIPQUAD(ipt_acc_tables[i].netmask));

            if (ipt_acc_tables[i].ip != ip 
                || ipt_acc_tables[i].netmask != netmask) {
                printk("ACCOUNT: Table %s found, but IP/netmask mismatch. "
                       "IP/netmask found: %u.%u.%u.%u/%u.%u.%u.%u\n",
                       name, NIPQUAD(ipt_acc_tables[i].ip), 
                       NIPQUAD(ipt_acc_tables[i].netmask));
                return -1;
            }

            ipt_acc_tables[i].refcount++;
            DEBUGP("ACCOUNT: Refcount: %d\n", ipt_acc_tables[i].refcount);
            return i;
        }
    }

    /* Insert new table */
    for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
        /* Found free slot */
        if (ipt_acc_tables[i].name[0] == 0) {
            u_int32_t j, calc_mask, netsize=0;
            
            DEBUGP("ACCOUNT: Found free slot: %d\n", i);
            strncpy (ipt_acc_tables[i].name, name, ACCOUNT_TABLE_NAME_LEN-1);

            ipt_acc_tables[i].ip = ip;
            ipt_acc_tables[i].netmask = netmask;

            /* Calculate netsize */
            calc_mask = htonl(netmask);
            for (j = 31; j >= 0; j--) {
                if (calc_mask&(1<<j))
                    netsize++;
                else
                    break;
            }

            /* Calculate depth from netsize */
            if (netsize >= 24)
                ipt_acc_tables[i].depth = 0;
            else if (netsize >= 16)
                ipt_acc_tables[i].depth = 1;
            else if(netsize >= 8)
                ipt_acc_tables[i].depth = 2;

            DEBUGP("ACCOUNT: calculated netsize: %u -> "
                   "ipt_acc_table depth %u\n", netsize, 
                   ipt_acc_tables[i].depth);

            ipt_acc_tables[i].refcount++;
            if ((ipt_acc_tables[i].data
                = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                printk("ACCOUNT: out of memory for data of table: %s\n", name);
                memset(&ipt_acc_tables[i], 0, 
                       sizeof(struct ipt_acc_table));
                return -1;
            }

            return i;
        }
    }

    /* No free slot found */
    printk("ACCOUNT: No free table slot found (max: %d). "
           "Please increase ACCOUNT_MAX_TABLES.\n", ACCOUNT_MAX_TABLES);
    return -1;
}

static int ipt_acc_checkentry(const char *tablename,
                                  const struct ipt_entry *e,
                                  void *targinfo,
                                  unsigned int targinfosize,
                                  unsigned int hook_mask)
{
    struct ipt_acc_info *info = targinfo;
    int table_nr;

    if (targinfosize != IPT_ALIGN(sizeof(struct ipt_acc_info))) {
        DEBUGP("ACCOUNT: targinfosize %u != %u\n",
               targinfosize, IPT_ALIGN(sizeof(struct ipt_acc_info)));
        return 0;
    }

    spin_lock_bh(&ipt_acc_lock);
    table_nr = ipt_acc_table_insert(info->table_name, info->net_ip,
                                                      info->net_mask);
    if (table_nr == -1) {
        printk("ACCOUNT: Table insert problem. Aborting\n");
        spin_unlock_bh(&ipt_acc_lock);
        return 0;
    }
    /* Table nr caching so we don't have to do an extra string compare 
       for every packet */
    info->table_nr = table_nr;

    spin_unlock_bh(&ipt_acc_lock);

    return 1;
}

void ipt_acc_deleteentry(void *targinfo, unsigned int targinfosize)
{
    u_int32_t i;
    struct ipt_acc_info *info = targinfo;

    if (targinfosize != IPT_ALIGN(sizeof(struct ipt_acc_info))) {
        DEBUGP("ACCOUNT: targinfosize %u != %u\n",
               targinfosize, IPT_ALIGN(sizeof(struct ipt_acc_info)));
    }

    spin_lock_bh(&ipt_acc_lock);

    DEBUGP("ACCOUNT: ipt_acc_deleteentry called for table: %s (#%d)\n", 
           info->table_name, info->table_nr);

    info->table_nr = -1;    /* Set back to original state */

    /* Look for table */
    for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
        if (strncmp(ipt_acc_tables[i].name, info->table_name, 
                    ACCOUNT_TABLE_NAME_LEN) == 0) {
            DEBUGP("ACCOUNT: Found table at slot: %d\n", i);

            ipt_acc_tables[i].refcount--;
            DEBUGP("ACCOUNT: Refcount left: %d\n", 
                   ipt_acc_tables[i].refcount);

            /* Table not needed anymore? */
            if (ipt_acc_tables[i].refcount == 0) {
                DEBUGP("ACCOUNT: Destroying table at slot: %d\n", i);
                ipt_acc_data_free(ipt_acc_tables[i].data, 
                                      ipt_acc_tables[i].depth);
                memset(&ipt_acc_tables[i], 0, 
                       sizeof(struct ipt_acc_table));
            }

            spin_unlock_bh(&ipt_acc_lock);
            return;
        }
    }

    /* Table not found */
    printk("ACCOUNT: Table %s not found for destroy\n", info->table_name);
    spin_unlock_bh(&ipt_acc_lock);
}

void ipt_acc_depth0_insert(struct ipt_acc_mask_24 *mask_24,
                               u_int32_t net_ip, u_int32_t netmask,
                               u_int32_t src_ip, u_int32_t dst_ip,
                               u_int32_t size, u_int32_t *itemcount)
{
    unsigned char is_src = 0, is_dst = 0, src_slot, dst_slot;
    char is_src_new_ip = 0, is_dst_new_ip = 0; /* Check if this entry is new */

    DEBUGP("ACCOUNT: ipt_acc_depth0_insert: %u.%u.%u.%u/%u.%u.%u.%u "
           "for net %u.%u.%u.%u/%u.%u.%u.%u, size: %u\n", NIPQUAD(src_ip), 
           NIPQUAD(dst_ip), NIPQUAD(net_ip), NIPQUAD(netmask), size);

    /* Check if src/dst is inside our network. */
    /* Special: net_ip = 0.0.0.0/0 gets stored as src in slot 0 */
    if (!netmask)
        src_ip = 0;
    if ((net_ip&netmask) == (src_ip&netmask))
        is_src = 1;
    if ((net_ip&netmask) == (dst_ip&netmask) && netmask)
        is_dst = 1;

    if (!is_src && !is_dst) {
        DEBUGP("ACCOUNT: Skipping packet %u.%u.%u.%u/%u.%u.%u.%u "
               "for net %u.%u.%u.%u/%u.%u.%u.%u\n", NIPQUAD(src_ip), 
               NIPQUAD(dst_ip), NIPQUAD(net_ip), NIPQUAD(netmask));
        return;
    }

    /* Calculate array positions */
    src_slot = (unsigned char)((src_ip&0xFF000000) >> 24);
    dst_slot = (unsigned char)((dst_ip&0xFF000000) >> 24);

    /* Increase size counters */
    if (is_src) {
        /* Calculate network slot */
        DEBUGP("ACCOUNT: Calculated SRC 8 bit network slot: %d\n", src_slot);
        if (!mask_24->ip[src_slot].src_packets 
            && !mask_24->ip[src_slot].dst_packets)
            is_src_new_ip = 1;

        mask_24->ip[src_slot].src_packets++;
        mask_24->ip[src_slot].src_bytes+=size;
    }
    if (is_dst) {
        DEBUGP("ACCOUNT: Calculated DST 8 bit network slot: %d\n", dst_slot);
        if (!mask_24->ip[dst_slot].src_packets 
            && !mask_24->ip[dst_slot].dst_packets)
            is_dst_new_ip = 1;

        mask_24->ip[dst_slot].dst_packets++;
        mask_24->ip[dst_slot].dst_bytes+=size;
    }

    /* Increase itemcounter */
    DEBUGP("ACCOUNT: Itemcounter before: %d\n", *itemcount);
    if (src_slot == dst_slot) {
        if (is_src_new_ip || is_dst_new_ip) {
            DEBUGP("ACCOUNT: src_slot == dst_slot: %d, %d\n", 
                   is_src_new_ip, is_dst_new_ip);
            (*itemcount)++;
        }
    } else {
        if (is_src_new_ip) {
            DEBUGP("ACCOUNT: New src_ip: %u.%u.%u.%u\n", NIPQUAD(src_ip));
            (*itemcount)++;
        }
        if (is_dst_new_ip) {
            DEBUGP("ACCOUNT: New dst_ip: %u.%u.%u.%u\n", NIPQUAD(dst_ip));
            (*itemcount)++;
        }
    }
    DEBUGP("ACCOUNT: Itemcounter after: %d\n", *itemcount);
}

void ipt_acc_depth1_insert(struct ipt_acc_mask_16 *mask_16, 
                               u_int32_t net_ip, u_int32_t netmask, 
                               u_int32_t src_ip, u_int32_t dst_ip,
                               u_int32_t size, u_int32_t *itemcount)
{
    /* Do we need to process src IP? */
    if ((net_ip&netmask) == (src_ip&netmask)) {
        unsigned char slot = (unsigned char)((src_ip&0x00FF0000) >> 16);
        DEBUGP("ACCOUNT: Calculated SRC 16 bit network slot: %d\n", slot);

        /* Do we need to create a new mask_24 bucket? */
        if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] = 
             (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
            printk("ACCOUNT: Can't process packet because out of memory!\n");
            return;
        }

        ipt_acc_depth0_insert((struct ipt_acc_mask_24 *)mask_16->mask_24[slot],
                                  net_ip, netmask, src_ip, 0, size, itemcount);
    }

    /* Do we need to process dst IP? */
    if ((net_ip&netmask) == (dst_ip&netmask)) {
        unsigned char slot = (unsigned char)((dst_ip&0x00FF0000) >> 16);
        DEBUGP("ACCOUNT: Calculated DST 16 bit network slot: %d\n", slot);

        /* Do we need to create a new mask_24 bucket? */
        if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] 
            = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
            printk("ACCOUT: Can't process packet because out of memory!\n");
            return;
        }

        ipt_acc_depth0_insert((struct ipt_acc_mask_24 *)mask_16->mask_24[slot],
                                  net_ip, netmask, 0, dst_ip, size, itemcount);
    }
}

void ipt_acc_depth2_insert(struct ipt_acc_mask_8 *mask_8, 
                               u_int32_t net_ip, u_int32_t netmask,
                               u_int32_t src_ip, u_int32_t dst_ip,
                               u_int32_t size, u_int32_t *itemcount)
{
    /* Do we need to process src IP? */
    if ((net_ip&netmask) == (src_ip&netmask)) {
        unsigned char slot = (unsigned char)((src_ip&0x0000FF00) >> 8);
        DEBUGP("ACCOUNT: Calculated SRC 24 bit network slot: %d\n", slot);

        /* Do we need to create a new mask_24 bucket? */
        if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot] 
            = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
            printk("ACCOUNT: Can't process packet because out of memory!\n");
            return;
        }

        ipt_acc_depth1_insert((struct ipt_acc_mask_16 *)mask_8->mask_16[slot],
                                  net_ip, netmask, src_ip, 0, size, itemcount);
    }

    /* Do we need to process dst IP? */
    if ((net_ip&netmask) == (dst_ip&netmask)) {
        unsigned char slot = (unsigned char)((dst_ip&0x0000FF00) >> 8);
        DEBUGP("ACCOUNT: Calculated DST 24 bit network slot: %d\n", slot);

        /* Do we need to create a new mask_24 bucket? */
        if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot] 
            = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
            printk("ACCOUNT: Can't process packet because out of memory!\n");
            return;
        }

        ipt_acc_depth1_insert((struct ipt_acc_mask_16 *)mask_8->mask_16[slot],
                                  net_ip, netmask, 0, dst_ip, size, itemcount);
    }
}

static unsigned int ipt_acc_target(struct sk_buff **pskb,
                                       unsigned int hooknum,
                                       const struct net_device *in,
                                       const struct net_device *out,
                                       const void *targinfo,
                                       void *userinfo)
{
    const struct ipt_acc_info *info = 
        (const struct ipt_acc_info *)targinfo;
    u_int32_t src_ip = (*pskb)->nh.iph->saddr;
    u_int32_t dst_ip = (*pskb)->nh.iph->daddr;
    u_int32_t size = ntohs((*pskb)->nh.iph->tot_len);

    spin_lock_bh(&ipt_acc_lock);

    if (ipt_acc_tables[info->table_nr].name[0] == 0) {
        printk("ACCOUNT: ipt_acc_target: Invalid table id %u. "
               "IPs %u.%u.%u.%u/%u.%u.%u.%u\n", info->table_nr, 
               NIPQUAD(src_ip), NIPQUAD(dst_ip));
        spin_unlock_bh(&ipt_acc_lock);
        return IPT_CONTINUE;
    }

    /* 8 bit network or "any" network */
    if (ipt_acc_tables[info->table_nr].depth == 0) {
        /* Count packet and check if the IP is new */
        ipt_acc_depth0_insert(
            (struct ipt_acc_mask_24 *)ipt_acc_tables[info->table_nr].data,
            ipt_acc_tables[info->table_nr].ip, 
            ipt_acc_tables[info->table_nr].netmask,
            src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
        spin_unlock_bh(&ipt_acc_lock);
        return IPT_CONTINUE;
    }

    /* 16 bit network */
    if (ipt_acc_tables[info->table_nr].depth == 1) {
        ipt_acc_depth1_insert(
            (struct ipt_acc_mask_16 *)ipt_acc_tables[info->table_nr].data,
            ipt_acc_tables[info->table_nr].ip, 
            ipt_acc_tables[info->table_nr].netmask,
            src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
        spin_unlock_bh(&ipt_acc_lock);
        return IPT_CONTINUE;
    }

    /* 24 bit network */
    if (ipt_acc_tables[info->table_nr].depth == 2) {
        ipt_acc_depth2_insert(
            (struct ipt_acc_mask_8 *)ipt_acc_tables[info->table_nr].data,
            ipt_acc_tables[info->table_nr].ip, 
            ipt_acc_tables[info->table_nr].netmask,
            src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
        spin_unlock_bh(&ipt_acc_lock);
        return IPT_CONTINUE;
    }

    printk("ACCOUNT: ipt_acc_target: Unable to process packet. "
           "Table id %u. IPs %u.%u.%u.%u/%u.%u.%u.%u\n", 
           info->table_nr, NIPQUAD(src_ip), NIPQUAD(dst_ip));

    spin_unlock_bh(&ipt_acc_lock);
    return IPT_CONTINUE;
}

/*
    Functions dealing with "handles":
    Handles are snapshots of a accounting state.
    
    read snapshots are only for debugging the code
    and are very expensive concerning speed/memory
    compared to read_and_flush.
    
    The functions aren't protected by spinlocks themselves
    as this is done in the ioctl part of the code.
*/

/*
    Find a free handle slot. Normally only one should be used,
    but there could be two or more applications accessing the data
    at the same time.
*/
int ipt_acc_handle_find_slot(void)
{
    u_int32_t i;
    /* Insert new table */
    for (i = 0; i < ACCOUNT_MAX_HANDLES; i++) {
        /* Found free slot */
        if (ipt_acc_handles[i].data == NULL) {
            /* Don't "mark" data as used as we are protected by a spinlock 
               by the calling function. handle_find_slot() is only a function
               to prevent code duplication. */
            return i;
        }
    }

    /* No free slot found */
    printk("ACCOUNT: No free handle slot found (max: %u). "
           "Please increase ACCOUNT_MAX_HANDLES.\n", ACCOUNT_MAX_HANDLES);
    return -1;
}

int ipt_acc_handle_free(u_int32_t handle)
{
    if (handle >= ACCOUNT_MAX_HANDLES) {
        printk("ACCOUNT: Invalid handle for ipt_acc_handle_free() specified:"
               " %u\n", handle);
        return -EINVAL;
    }

    ipt_acc_data_free(ipt_acc_handles[handle].data, 
                          ipt_acc_handles[handle].depth);
    memset (&ipt_acc_handles[handle], 0, sizeof (struct ipt_acc_handle));
    return 0;
}

/* Prepare data for read without flush. Use only for debugging!
   Real applications should use read&flush as it's way more efficent */
int ipt_acc_handle_prepare_read(char *tablename, u_int32_t *count)
{
    int handle, i, table_nr=-1;
    unsigned char depth;

    for (i = 0; i < ACCOUNT_MAX_TABLES; i++)
        if (strncmp(ipt_acc_tables[i].name, tablename, 
            ACCOUNT_TABLE_NAME_LEN) == 0) {
            table_nr = i;
            break;
        }

    if (table_nr == -1) {
        printk("ACCOUNT: ipt_acc_handle_prepare_read(): "
               "Table %s not found\n", tablename);
        return -1;
    }

    /* Can't find a free handle slot? */
    if ((handle = ipt_acc_handle_find_slot()) == -1)
        return -1;

    /* Fill up handle structure */
    ipt_acc_handles[handle].ip = ipt_acc_tables[table_nr].ip;
    ipt_acc_handles[handle].depth = ipt_acc_tables[table_nr].depth;
    ipt_acc_handles[handle].itemcount = ipt_acc_tables[table_nr].itemcount;

    /* allocate "root" table */
    if ((ipt_acc_handles[handle].data = 
         (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
        printk("ACCOUNT: out of memory for root table "
               "in ipt_acc_handle_prepare_read()\n");
        memset (&ipt_acc_handles[handle], 0, 
                sizeof(struct ipt_acc_handle));
        return -1;
    }

    /* Recursive copy of complete data structure */
    depth = ipt_acc_handles[handle].depth;
    if (depth == 0) {
        memcpy(ipt_acc_handles[handle].data, 
               ipt_acc_tables[table_nr].data, 
               sizeof(struct ipt_acc_mask_24));
    } else if (depth == 1) {
        struct ipt_acc_mask_16 *src_16 = 
            (struct ipt_acc_mask_16 *)ipt_acc_tables[table_nr].data;
        struct ipt_acc_mask_16 *network_16 =
            (struct ipt_acc_mask_16 *)ipt_acc_handles[handle].data;
        u_int32_t b;

        for (b = 0; b <= 255; b++) {
            if (src_16->mask_24[b]) {
                if ((network_16->mask_24[b] = 
                     (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                    printk("ACCOUNT: out of memory during copy of 16 bit "
                           "network in ipt_acc_handle_prepare_read()\n");
                    ipt_acc_data_free(ipt_acc_handles[handle].data, depth);
                    memset (&ipt_acc_handles[handle], 0, 
                            sizeof(struct ipt_acc_handle));
                    return -1;
                }

                memcpy(network_16->mask_24[b], src_16->mask_24[b], 
                       sizeof(struct ipt_acc_mask_24));
            }
        }
    } else if(depth == 2) {
        struct ipt_acc_mask_8 *src_8 = 
            (struct ipt_acc_mask_8 *)ipt_acc_tables[table_nr].data;
        struct ipt_acc_mask_8 *network_8 = 
            (struct ipt_acc_mask_8 *)ipt_acc_handles[handle].data;
        u_int32_t a;

        for (a = 0; a <= 255; a++) {
            if (src_8->mask_16[a]) {
                if ((network_8->mask_16[a] = 
                     (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                    printk("ACCOUNT: out of memory during copy of 24 bit network"
                           " in ipt_acc_handle_prepare_read()\n");
                    ipt_acc_data_free(ipt_acc_handles[handle].data, depth);
                    memset (&ipt_acc_handles[handle], 0, 
                            sizeof(struct ipt_acc_handle));
                    return -1;
                }

                memcpy(network_8->mask_16[a], src_8->mask_16[a], 
                       sizeof(struct ipt_acc_mask_16));

                struct ipt_acc_mask_16 *src_16 = src_8->mask_16[a];
                struct ipt_acc_mask_16 *network_16 = network_8->mask_16[a];
                u_int32_t b;

                for (b = 0; b <= 255; b++) {
                    if (src_16->mask_24[b]) {
                        if ((network_16->mask_24[b] = 
                             (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                            printk("ACCOUNT: out of memory during copy of 16 bit"
                                   " network in ipt_acc_handle_prepare_read()\n");
                            ipt_acc_data_free(ipt_acc_handles[handle].data,
                                                  depth);
                            memset (&ipt_acc_handles[handle], 0, 
                                    sizeof(struct ipt_acc_handle));
                            return -1;
                        }

                        memcpy(network_16->mask_24[b], src_16->mask_24[b], 
                               sizeof(struct ipt_acc_mask_24));
                    }
                }
            }
        }
    }

    *count = ipt_acc_tables[table_nr].itemcount;
    return handle;
}

/* Prepare data for read and flush it */
int ipt_acc_handle_prepare_read_flush(char *tablename, u_int32_t *count)
{
    int handle, i, table_nr=-1;
    void *new_data_page;

    for (i = 0; i < ACCOUNT_MAX_TABLES; i++)
        if (strncmp(ipt_acc_tables[i].name, tablename, 
                    ACCOUNT_TABLE_NAME_LEN) == 0) {
            table_nr = i;
            break;
        }

    if (table_nr == -1) {
        printk("ACCOUNT: ipt_acc_handle_prepare_read_flush(): "
               "Table %s not found\n", tablename);
        return -1;
    }

    /* Can't find a free handle slot? */
    if ((handle = ipt_acc_handle_find_slot()) == -1)
        return -1;

    /* Try to allocate memory */
    if (!(new_data_page = (void*)get_zeroed_page(GFP_ATOMIC))) {
        printk("ACCOUNT: ipt_acc_handle_prepare_read_flush(): "
               "Out of memory!\n");
        return -1;
    }

    /* Fill up handle structure */
    ipt_acc_handles[handle].ip = ipt_acc_tables[table_nr].ip;
    ipt_acc_handles[handle].depth = ipt_acc_tables[table_nr].depth;
    ipt_acc_handles[handle].itemcount = ipt_acc_tables[table_nr].itemcount;
    ipt_acc_handles[handle].data = ipt_acc_tables[table_nr].data;
    *count = ipt_acc_tables[table_nr].itemcount;

    /* "Flush" table data */
    ipt_acc_tables[table_nr].data = new_data_page;
    ipt_acc_tables[table_nr].itemcount = 0;

    return handle;
}

/* Copy 8 bit network data into a prepared buffer.
   We only copy entries != 0 to increase performance.
*/
int ipt_acc_handle_copy_data(void *to_user, int *pos, 
                                  struct ipt_acc_mask_24 *data,
                                  u_int32_t net_ip, u_int32_t net_OR_mask)
{
    struct ipt_acc_handle_ip handle_ip;
    u_int32_t handle_ip_size = sizeof (struct ipt_acc_handle_ip);
    u_int32_t i;
    
    for (i = 0; i <= 255; i++) {
        if (data->ip[i].src_packets || data->ip[i].dst_packets) {
            handle_ip.ip = net_ip | net_OR_mask | (i<<24);
            
            handle_ip.src_packets = data->ip[i].src_packets;
            handle_ip.src_bytes = data->ip[i].src_bytes;
            handle_ip.dst_packets = data->ip[i].dst_packets;
            handle_ip.dst_bytes = data->ip[i].dst_bytes;

            /* Temporary buffer full? Flush to userspace */
            if (*pos+handle_ip_size >= PAGE_SIZE) {
                *pos = 0;
                if (copy_to_user(to_user, ipt_acc_tmpbuf, *pos))
                    return -EFAULT;
            }
            memcpy(ipt_acc_tmpbuf+*pos, &handle_ip, handle_ip_size);
            *pos += handle_ip_size;
        }
    }
    
    return 0;
}
   
/* Copy the data from our internal structure 
   We only copy entries != 0 to increase performance.
   Overwrites ipt_acc_tmpbuf.
*/
int ipt_acc_handle_get_data(u_int32_t handle, void *to_user)
{
    u_int32_t tmpbuf_pos=0, net_ip;
    unsigned char depth;

    if (handle >= ACCOUNT_MAX_HANDLES) {
        printk("ACCOUNT: invalid handle for ipt_acc_handle_get_data() "
               "specified: %u\n", handle);
        return -1;
    }

    if (ipt_acc_handles[handle].data == NULL) {
        printk("ACCOUNT: handle %u is BROKEN: Contains no data\n", handle);
        return -1;
    }

    net_ip = ipt_acc_handles[handle].ip;
    depth = ipt_acc_handles[handle].depth;

    /* 8 bit network */
    if (depth == 0) {
        struct ipt_acc_mask_24 *network = 
            (struct ipt_acc_mask_24*)ipt_acc_handles[handle].data;
        if (ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network, net_ip, 0))
            return -1;
        
        /* Flush remaining data to userspace */
        if (tmpbuf_pos)
            if (copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos))
                return -1;

        return 0;
    }

    /* 16 bit network */
    if (depth == 1) {
        struct ipt_acc_mask_16 *network_16 = 
            (struct ipt_acc_mask_16*)ipt_acc_handles[handle].data;
        u_int32_t b;
        for (b = 0; b <= 255; b++) {
            if (network_16->mask_24[b]) {
                struct ipt_acc_mask_24 *network = 
                    (struct ipt_acc_mask_24*)network_16->mask_24[b];
                if (ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network, 
                                             net_ip, (b << 16)))
                    return -1;
            }
        }

        /* Flush remaining data to userspace */
        if (tmpbuf_pos)
            if (copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos))
                return -1;

        return 0;
    }

    /* 24 bit network */
    if (depth == 2) {
        struct ipt_acc_mask_8 *network_8 = 
            (struct ipt_acc_mask_8*)ipt_acc_handles[handle].data;
        u_int32_t a, b;
        for (a = 0; a <= 255; a++) {
            if (network_8->mask_16[a]) {
                struct ipt_acc_mask_16 *network_16 = 
                    (struct ipt_acc_mask_16*)network_8->mask_16[a];
                for (b = 0; b <= 255; b++) {
                    if (network_16->mask_24[b]) {
                        struct ipt_acc_mask_24 *network = 
                            (struct ipt_acc_mask_24*)network_16->mask_24[b];
                        if (ipt_acc_handle_copy_data(to_user, &tmpbuf_pos,
                                        network, net_ip, (a << 8) | (b << 16)))
                            return -1;
                    }
                }
            }
        }

        /* Flush remaining data to userspace */
        if (tmpbuf_pos)
            if (copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos))
                return -1;

        return 0;
    }
    
    return -1;
}

static int ipt_acc_set_ctl(struct sock *sk, int cmd, 
                               void *user, u_int32_t len)
{
    struct ipt_acc_handle_sockopt handle;
    int ret = -EINVAL;

    if (!capable(CAP_NET_ADMIN))
        return -EPERM;

    switch (cmd) {
    case IPT_SO_SET_ACCOUNT_HANDLE_FREE:
        if (len != sizeof(struct ipt_acc_handle_sockopt)) {
            printk("ACCOUNT: ipt_acc_set_ctl: wrong data size (%u != %u) "
                   "for IPT_SO_SET_HANDLE_FREE\n", 
                   len, sizeof(struct ipt_acc_handle_sockopt));
            break;
        }

        if (copy_from_user (&handle, user, len)) {
            printk("ACCOUNT: ipt_acc_set_ctl: copy_from_user failed for "
                   "IPT_SO_SET_HANDLE_FREE\n");
            break;
        }

        spin_lock_bh(&ipt_acc_userspace_lock);
        ret = ipt_acc_handle_free(handle.handle_nr);
        spin_unlock_bh(&ipt_acc_userspace_lock);
        break;
    case IPT_SO_SET_ACCOUNT_HANDLE_FREE_ALL: {
            u_int32_t i;
            spin_lock_bh(&ipt_acc_userspace_lock);
            for (i = 0; i < ACCOUNT_MAX_HANDLES; i++)
                ipt_acc_handle_free(i);
            spin_unlock_bh(&ipt_acc_userspace_lock);
            ret = 0;
            break;
        }
    default:
        printk("ACCOUNT: ipt_acc_set_ctl: unknown request %i\n", cmd);
    }

    return ret;
}

static int ipt_acc_get_ctl(struct sock *sk, int cmd, void *user, int *len)
{
    struct ipt_acc_handle_sockopt handle;
    int ret = -EINVAL;

    if (!capable(CAP_NET_ADMIN))
        return -EPERM;

    switch (cmd) {
    case IPT_SO_GET_ACCOUNT_PREPARE_READ_FLUSH:
    case IPT_SO_GET_ACCOUNT_PREPARE_READ:
        if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
            printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u) "
                   "for IPT_SO_GET_ACCOUNT_PREPARE_READ/READ_FLUSH\n",
                   *len, sizeof(struct ipt_acc_handle_sockopt));
            break;
        }

        if (copy_from_user (&handle, user, 
                            sizeof(struct ipt_acc_handle_sockopt))) {
            return -EFAULT;
            break;
        }

        spin_lock_bh(&ipt_acc_lock);
        spin_lock_bh(&ipt_acc_userspace_lock);
        if (cmd == IPT_SO_GET_ACCOUNT_PREPARE_READ_FLUSH)
            handle.handle_nr = ipt_acc_handle_prepare_read_flush(
                                   handle.name, &handle.itemcount);
        else
            handle.handle_nr = ipt_acc_handle_prepare_read(
                                   handle.name, &handle.itemcount);
        spin_unlock_bh(&ipt_acc_userspace_lock);
        spin_unlock_bh(&ipt_acc_lock);

        if (handle.handle_nr == -1) {
            return -EINVAL;
            break;
        }

        if (copy_to_user(user, &handle, 
                         sizeof(struct ipt_acc_handle_sockopt))) {
            return -EFAULT;
            break;
        }
        ret = 0;
        break;
    case IPT_SO_GET_ACCOUNT_GET_DATA:
        if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
            printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u)"
                   " for IPT_SO_GET_ACCOUNT_PREPARE_READ/READ_FLUSH\n",
                   *len, sizeof(struct ipt_acc_handle_sockopt));
            break;
        }

        if (copy_from_user (&handle, user, 
                            sizeof(struct ipt_acc_handle_sockopt))) {
            return -EFAULT;
            break;
        }

        if (handle.handle_nr >= ACCOUNT_MAX_HANDLES) {
            return -EINVAL;
            break;
        }

        if (*len < ipt_acc_handles[handle.handle_nr].itemcount
                   * sizeof(struct ipt_acc_handle_ip)) {
            printk("ACCOUNT: ipt_acc_get_ctl: not enough space (%u < %u)"
                   " to store data from IPT_SO_GET_ACCOUNT_GET_DATA\n",
                   *len, ipt_acc_handles[handle.handle_nr].itemcount
                   * sizeof(struct ipt_acc_handle_ip));
            ret = -ENOMEM;
            break;
        }

        spin_lock_bh(&ipt_acc_userspace_lock);
        ret = ipt_acc_handle_get_data(handle.handle_nr, user);
        spin_unlock_bh(&ipt_acc_userspace_lock);
        if (ret) {
            printk("ACCOUNT: ipt_acc_get_ctl: ipt_acc_handle_get_data"
                   " failed for handle %u\n", handle.handle_nr);
            break;
        }

        ret = 0;
        break;
    case IPT_SO_GET_ACCOUNT_GET_HANDLE_USAGE: {
            u_int32_t i;
            if (*len < sizeof(struct ipt_acc_handle_sockopt)) {
                printk("ACCOUNT: ipt_acc_get_ctl: wrong data size (%u != %u)"
                       " for IPT_SO_GET_ACCOUNT_GET_HANDLE_USAGE\n",
                       *len, sizeof(struct ipt_acc_handle_sockopt));
                break;
            }

            /* Find out how many handles are in use */
            handle.itemcount = 0;
            spin_lock_bh(&ipt_acc_userspace_lock);
            for (i = 0; i < ACCOUNT_MAX_HANDLES; i++)
                if (ipt_acc_handles[i].data)
                    handle.itemcount++;
            spin_unlock_bh(&ipt_acc_userspace_lock);

            if (copy_to_user(user, &handle, 
                             sizeof(struct ipt_acc_handle_sockopt))) {
                return -EFAULT;
                break;
            }
            ret = 0;
            break;
        }
    case IPT_SO_GET_ACCOUNT_GET_TABLE_NAMES: {
            u_int32_t size = 0, i;
            char *tnames;
            
            spin_lock_bh(&ipt_acc_lock);

            /* Determine size of table names */
            for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
                if (ipt_acc_tables[i].name[0] != 0)
                    size += strlen (ipt_acc_tables[i].name) + 1;
            }
            size += 1;    /* Terminating NULL character */

            if (*len < size) {
                spin_unlock_bh(&ipt_acc_lock);
                printk("ACCOUNT: ipt_acc_get_ctl: not enough space (%u < %u)"
                       " to store table names\n", *len, size);
                ret = -ENOMEM;
                break;
            }
            /* Copy table names to userspace */
            tnames = user;
            for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
                if (ipt_acc_tables[i].name[0] != 0) {
                    int len = strlen (ipt_acc_tables[i].name) + 1;
                    /* copy string + terminating zero */
                    if (copy_to_user(tnames, ipt_acc_tables[i].name, len)) {
                        spin_unlock_bh(&ipt_acc_lock);
                        return -EFAULT;
                    }
                    tnames += len;
                }
            }
            /* Append terminating zero */
            i = 0;
            ret = copy_to_user(tnames, &i, 1);
            spin_unlock_bh(&ipt_acc_lock);
            if (ret)
                return -EFAULT;
            ret = 0;
            break;
        }
    default:
        printk("ACCOUNT: ipt_acc_get_ctl: unknown request %i\n", cmd);
    }

    return ret;
}

static struct ipt_target ipt_acc_reg = {
    .name = "ACCOUNT",
    .target = ipt_acc_target,
    .checkentry = ipt_acc_checkentry,
    .destroy = ipt_acc_deleteentry,
    .me = THIS_MODULE
};

static struct nf_sockopt_ops ipt_acc_sockopts = {
    .pf = PF_INET,
    .set_optmin = IPT_SO_SET_ACCOUNT_HANDLE_FREE,
    .set_optmax = IPT_SO_SET_ACCOUNT_MAX+1,
    .set = ipt_acc_set_ctl,
    .get_optmin = IPT_SO_GET_ACCOUNT_PREPARE_READ,
    .get_optmax = IPT_SO_GET_ACCOUNT_MAX+1,
    .get = ipt_acc_get_ctl
};

static int __init init(void)
{
    if ((ipt_acc_tables = 
         kmalloc(ACCOUNT_MAX_TABLES * 
                 sizeof(struct ipt_acc_table), GFP_KERNEL)) == NULL) {
        printk("ACCOUNT: Out of memory allocating account_tables structure");
        goto error_cleanup;
    }
    memset(ipt_acc_tables, 0, 
           ACCOUNT_MAX_TABLES * sizeof(struct ipt_acc_table));

    if ((ipt_acc_handles = 
         kmalloc(ACCOUNT_MAX_HANDLES * 
                 sizeof(struct ipt_acc_handle), GFP_KERNEL)) == NULL) {
        printk("ACCOUNT: Out of memory allocating account_handles structure");
        goto error_cleanup;
    }
    memset(ipt_acc_handles, 0, 
           ACCOUNT_MAX_HANDLES * sizeof(struct ipt_acc_handle));

    /* Allocate one page as temporary storage */
    if ((ipt_acc_tmpbuf = (void*)__get_free_page(GFP_KERNEL)) == NULL) {
        printk("ACCOUNT: Out of memory for temporary buffer page\n");
        goto error_cleanup;
    }

    /* Register setsockopt */
    if (nf_register_sockopt(&ipt_acc_sockopts) < 0) {
        printk("ACCOUNT: Can't register sockopts. Aborting\n");
        goto error_cleanup;
    }

    if (ipt_register_target(&ipt_acc_reg))
        goto error_cleanup;
        
    return 0;
        
error_cleanup:
    if(ipt_acc_tables)
        kfree(ipt_acc_tables);
    if(ipt_acc_handles)
        kfree(ipt_acc_handles);
    if (ipt_acc_tmpbuf)
        free_page((unsigned long)ipt_acc_tmpbuf);
        
    return -EINVAL;
}

static void __exit fini(void)
{
    ipt_unregister_target(&ipt_acc_reg);

    nf_unregister_sockopt(&ipt_acc_sockopts);

    kfree(ipt_acc_tables);
    kfree(ipt_acc_handles);
    free_page((unsigned long)ipt_acc_tmpbuf);
}

module_init(init);
module_exit(fini);
MODULE_LICENSE("GPL");

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: new ipt_ACCOUNT version
  2004-06-07 19:48   ` Thomas Jarosch
@ 2004-06-09  8:39     ` Patrick McHardy
  0 siblings, 0 replies; 8+ messages in thread
From: Patrick McHardy @ 2004-06-09  8:39 UTC (permalink / raw)
  To: Thomas Jarosch; +Cc: netfilter-devel

Thomas Jarosch wrote:
> Hi Patrick,
> 
> Thanks for the in-depth review! I was amazed you found
> the off-by-one "bug" in the netmask calculation code :-)
> 
> Attached is a new version with all the fixes.
> I did some more old-gcc-compile fixing.

Thanks, it looks ok, we can put it in CVS and do remaining
cleanup then. Can you send me a patch (including userspace)
for pom-ng ?

Regards
Patrick

> 
> Cheers,
> Thomas
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: new ipt_ACCOUNT version
       [not found]     ` <200406101907.05352.thomas.jarosch@intra2net.com>
@ 2004-06-13 20:44       ` Patrick McHardy
  2004-06-13 21:09         ` Thomas Jarosch
  0 siblings, 1 reply; 8+ messages in thread
From: Patrick McHardy @ 2004-06-13 20:44 UTC (permalink / raw)
  To: Thomas Jarosch; +Cc: netfilter-devel

[-- Attachment #1: Type: text/plain, Size: 263 bytes --]

On Thu, 2004-06-10 at 19:07, Thomas Jarosch wrote:
> Hi Patrick,
>
> Please see the attached file. Hope the userspace patch is ok.

Applied with the attached patch on top (trailing whitespace cleanup
and missing statics).

Regards
Patrick


> 
> Cheers,
> Thomas

[-- Attachment #2: x --]
[-- Type: text/plain, Size: 32244 bytes --]

Index: include/linux/netfilter_ipv4/ipt_ACCOUNT.h
===================================================================
RCS file: /cvsroot/patch-o-matic-ng/ACCOUNT/linux/include/linux/netfilter_ipv4/ipt_ACCOUNT.h,v
retrieving revision 1.1
diff -u -r1.1 ipt_ACCOUNT.h
--- include/linux/netfilter_ipv4/ipt_ACCOUNT.h	13 Jun 2004 20:35:22 -0000	1.1
+++ include/linux/netfilter_ipv4/ipt_ACCOUNT.h	13 Jun 2004 20:40:27 -0000
@@ -28,24 +28,24 @@
     char name[ACCOUNT_TABLE_NAME_LEN];     /* name of the table */
     u_int32_t ip;                          /* base IP of network */
     u_int32_t netmask;                     /* netmask of the network */
-    unsigned char depth;                   /* size of network: 
+    unsigned char depth;                   /* size of network:
                                                  0: 8 bit, 1: 16bit, 2: 24 bit */
-    u_int32_t refcount;                    /* refcount of this table. 
+    u_int32_t refcount;                    /* refcount of this table.
                                                  if zero, destroy it */
     u_int32_t itemcount;                   /* number of IPs in this table */
-    void *data;                            /* pointer to the actual data, 
+    void *data;                            /* pointer to the actual data,
                                                  depending on netmask */
 };
 
 /* Internal handle structure */
 struct ipt_acc_handle {
-    u_int32_t ip;                          /* base IP of network. Used for 
+    u_int32_t ip;                          /* base IP of network. Used for
                                                  caculating the final IP during
                                                  get_data() */
-    unsigned char depth;                   /* size of network. See above for 
+    unsigned char depth;                   /* size of network. See above for
                                                  details */
     u_int32_t itemcount;                   /* number of IPs in this table */
-    void *data;                            /* pointer to the actual data, 
+    void *data;                            /* pointer to the actual data,
                                                  depending on size */
 };
 
@@ -58,8 +58,8 @@
                                                  HANDLE_READ_FLUSH */
 };
 
-/* Used for every IP entry 
-   Size is 16 bytes so that 256 (class C network) * 16 
+/* Used for every IP entry
+   Size is 16 bytes so that 256 (class C network) * 16
    fits in one kernel (zero) page */
 struct ipt_acc_ip {
     u_int32_t src_packets;
Index: net/ipv4/netfilter/ipt_ACCOUNT.c
===================================================================
RCS file: /cvsroot/patch-o-matic-ng/ACCOUNT/linux/net/ipv4/netfilter/ipt_ACCOUNT.c,v
retrieving revision 1.1
diff -u -r1.1 ipt_ACCOUNT.c
--- net/ipv4/netfilter/ipt_ACCOUNT.c	13 Jun 2004 20:35:22 -0000	1.1
+++ net/ipv4/netfilter/ipt_ACCOUNT.c	13 Jun 2004 20:40:29 -0000
@@ -2,7 +2,7 @@
  *   This is a module which is used for counting packets.                  *
  *   See http://www.intra2net.com/opensource/ipt_account                   *
  *   for further information                                               *
- *                                                                         * 
+ *                                                                         *
  *   Copyright (C) 2004 by Intra2net AG                                    *
  *   opensource@intra2net.com                                              *
  *                                                                         *
@@ -25,7 +25,6 @@
 #include <linux/string.h>
 #include <asm/uaccess.h>
 
-struct in_device;
 #include <net/route.h>
 #include <linux/netfilter_ipv4/ipt_ACCOUNT.h>
 
@@ -39,9 +38,9 @@
 #error "ipt_ACCOUNT needs at least a PAGE_SIZE of 4096"
 #endif
 
-struct ipt_acc_table *ipt_acc_tables = NULL;
-struct ipt_acc_handle *ipt_acc_handles = NULL;
-void *ipt_acc_tmpbuf = NULL;
+static struct ipt_acc_table *ipt_acc_tables = NULL;
+static struct ipt_acc_handle *ipt_acc_handles = NULL;
+static void *ipt_acc_tmpbuf = NULL;
 
 /* Spinlock used for manipulating the current accounting tables/data */
 static spinlock_t ipt_acc_lock = SPIN_LOCK_UNLOCKED;
@@ -50,7 +49,7 @@
 
 
 /* Recursive free of all data structures */
-void ipt_acc_data_free(void *data, unsigned char depth)
+static void ipt_acc_data_free(void *data, unsigned char depth)
 {
     /* Empty data set */
     if (!data)
@@ -82,7 +81,7 @@
             if (((struct ipt_acc_mask_8 *)data)->mask_16[a]) {
                 struct ipt_acc_mask_16 *mask_16 = (struct ipt_acc_mask_16*)
                                    ((struct ipt_acc_mask_8 *)data)->mask_16[a];
-                
+
                 for (b=0; b <= 255; b++) {
                     if (mask_16->mask_24[b]) {
                         free_page((unsigned long)mask_16->mask_24[b]);
@@ -95,14 +94,14 @@
         return;
     }
 
-    printk("ACCOUNT: ipt_acc_data_free called with unknown depth: %d\n", 
+    printk("ACCOUNT: ipt_acc_data_free called with unknown depth: %d\n",
            depth);
     return;
 }
 
-/* Look for existing table / insert new one. 
+/* Look for existing table / insert new one.
    Return internal ID or -1 on error */
-int ipt_acc_table_insert(char *name, u_int32_t ip, u_int32_t netmask)
+static int ipt_acc_table_insert(char *name, u_int32_t ip, u_int32_t netmask)
 {
     u_int32_t i;
 
@@ -111,18 +110,18 @@
 
     /* Look for existing table */
     for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
-        if (strncmp(ipt_acc_tables[i].name, name, 
+        if (strncmp(ipt_acc_tables[i].name, name,
                     ACCOUNT_TABLE_NAME_LEN) == 0) {
             DEBUGP("ACCOUNT: Found existing slot: %d - "
-                   "%u.%u.%u.%u/%u.%u.%u.%u\n", i, 
-                   NIPQUAD(ipt_acc_tables[i].ip), 
+                   "%u.%u.%u.%u/%u.%u.%u.%u\n", i,
+                   NIPQUAD(ipt_acc_tables[i].ip),
                    NIPQUAD(ipt_acc_tables[i].netmask));
 
-            if (ipt_acc_tables[i].ip != ip 
+            if (ipt_acc_tables[i].ip != ip
                 || ipt_acc_tables[i].netmask != netmask) {
                 printk("ACCOUNT: Table %s found, but IP/netmask mismatch. "
                        "IP/netmask found: %u.%u.%u.%u/%u.%u.%u.%u\n",
-                       name, NIPQUAD(ipt_acc_tables[i].ip), 
+                       name, NIPQUAD(ipt_acc_tables[i].ip),
                        NIPQUAD(ipt_acc_tables[i].netmask));
                 return -1;
             }
@@ -138,7 +137,7 @@
         /* Found free slot */
         if (ipt_acc_tables[i].name[0] == 0) {
             u_int32_t j, calc_mask, netsize=0;
-            
+
             DEBUGP("ACCOUNT: Found free slot: %d\n", i);
             strncpy (ipt_acc_tables[i].name, name, ACCOUNT_TABLE_NAME_LEN-1);
 
@@ -163,14 +162,14 @@
                 ipt_acc_tables[i].depth = 2;
 
             DEBUGP("ACCOUNT: calculated netsize: %u -> "
-                   "ipt_acc_table depth %u\n", netsize, 
+                   "ipt_acc_table depth %u\n", netsize,
                    ipt_acc_tables[i].depth);
 
             ipt_acc_tables[i].refcount++;
             if ((ipt_acc_tables[i].data
                 = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                 printk("ACCOUNT: out of memory for data of table: %s\n", name);
-                memset(&ipt_acc_tables[i], 0, 
+                memset(&ipt_acc_tables[i], 0,
                        sizeof(struct ipt_acc_table));
                 return -1;
             }
@@ -208,7 +207,7 @@
         spin_unlock_bh(&ipt_acc_lock);
         return 0;
     }
-    /* Table nr caching so we don't have to do an extra string compare 
+    /* Table nr caching so we don't have to do an extra string compare
        for every packet */
     info->table_nr = table_nr;
 
@@ -217,7 +216,7 @@
     return 1;
 }
 
-void ipt_acc_deleteentry(void *targinfo, unsigned int targinfosize)
+static void ipt_acc_deleteentry(void *targinfo, unsigned int targinfosize)
 {
     u_int32_t i;
     struct ipt_acc_info *info = targinfo;
@@ -229,27 +228,27 @@
 
     spin_lock_bh(&ipt_acc_lock);
 
-    DEBUGP("ACCOUNT: ipt_acc_deleteentry called for table: %s (#%d)\n", 
+    DEBUGP("ACCOUNT: ipt_acc_deleteentry called for table: %s (#%d)\n",
            info->table_name, info->table_nr);
 
     info->table_nr = -1;    /* Set back to original state */
 
     /* Look for table */
     for (i = 0; i < ACCOUNT_MAX_TABLES; i++) {
-        if (strncmp(ipt_acc_tables[i].name, info->table_name, 
+        if (strncmp(ipt_acc_tables[i].name, info->table_name,
                     ACCOUNT_TABLE_NAME_LEN) == 0) {
             DEBUGP("ACCOUNT: Found table at slot: %d\n", i);
 
             ipt_acc_tables[i].refcount--;
-            DEBUGP("ACCOUNT: Refcount left: %d\n", 
+            DEBUGP("ACCOUNT: Refcount left: %d\n",
                    ipt_acc_tables[i].refcount);
 
             /* Table not needed anymore? */
             if (ipt_acc_tables[i].refcount == 0) {
                 DEBUGP("ACCOUNT: Destroying table at slot: %d\n", i);
-                ipt_acc_data_free(ipt_acc_tables[i].data, 
+                ipt_acc_data_free(ipt_acc_tables[i].data,
                                       ipt_acc_tables[i].depth);
-                memset(&ipt_acc_tables[i], 0, 
+                memset(&ipt_acc_tables[i], 0,
                        sizeof(struct ipt_acc_table));
             }
 
@@ -263,16 +262,16 @@
     spin_unlock_bh(&ipt_acc_lock);
 }
 
-void ipt_acc_depth0_insert(struct ipt_acc_mask_24 *mask_24,
-                               u_int32_t net_ip, u_int32_t netmask,
-                               u_int32_t src_ip, u_int32_t dst_ip,
-                               u_int32_t size, u_int32_t *itemcount)
+static void ipt_acc_depth0_insert(struct ipt_acc_mask_24 *mask_24,
+                                  u_int32_t net_ip, u_int32_t netmask,
+                                  u_int32_t src_ip, u_int32_t dst_ip,
+                                  u_int32_t size, u_int32_t *itemcount)
 {
     unsigned char is_src = 0, is_dst = 0, src_slot, dst_slot;
     char is_src_new_ip = 0, is_dst_new_ip = 0; /* Check if this entry is new */
 
     DEBUGP("ACCOUNT: ipt_acc_depth0_insert: %u.%u.%u.%u/%u.%u.%u.%u "
-           "for net %u.%u.%u.%u/%u.%u.%u.%u, size: %u\n", NIPQUAD(src_ip), 
+           "for net %u.%u.%u.%u/%u.%u.%u.%u, size: %u\n", NIPQUAD(src_ip),
            NIPQUAD(dst_ip), NIPQUAD(net_ip), NIPQUAD(netmask), size);
 
     /* Check if src/dst is inside our network. */
@@ -286,7 +285,7 @@
 
     if (!is_src && !is_dst) {
         DEBUGP("ACCOUNT: Skipping packet %u.%u.%u.%u/%u.%u.%u.%u "
-               "for net %u.%u.%u.%u/%u.%u.%u.%u\n", NIPQUAD(src_ip), 
+               "for net %u.%u.%u.%u/%u.%u.%u.%u\n", NIPQUAD(src_ip),
                NIPQUAD(dst_ip), NIPQUAD(net_ip), NIPQUAD(netmask));
         return;
     }
@@ -299,7 +298,7 @@
     if (is_src) {
         /* Calculate network slot */
         DEBUGP("ACCOUNT: Calculated SRC 8 bit network slot: %d\n", src_slot);
-        if (!mask_24->ip[src_slot].src_packets 
+        if (!mask_24->ip[src_slot].src_packets
             && !mask_24->ip[src_slot].dst_packets)
             is_src_new_ip = 1;
 
@@ -308,7 +307,7 @@
     }
     if (is_dst) {
         DEBUGP("ACCOUNT: Calculated DST 8 bit network slot: %d\n", dst_slot);
-        if (!mask_24->ip[dst_slot].src_packets 
+        if (!mask_24->ip[dst_slot].src_packets
             && !mask_24->ip[dst_slot].dst_packets)
             is_dst_new_ip = 1;
 
@@ -320,7 +319,7 @@
     DEBUGP("ACCOUNT: Itemcounter before: %d\n", *itemcount);
     if (src_slot == dst_slot) {
         if (is_src_new_ip || is_dst_new_ip) {
-            DEBUGP("ACCOUNT: src_slot == dst_slot: %d, %d\n", 
+            DEBUGP("ACCOUNT: src_slot == dst_slot: %d, %d\n",
                    is_src_new_ip, is_dst_new_ip);
             (*itemcount)++;
         }
@@ -337,10 +336,10 @@
     DEBUGP("ACCOUNT: Itemcounter after: %d\n", *itemcount);
 }
 
-void ipt_acc_depth1_insert(struct ipt_acc_mask_16 *mask_16, 
-                               u_int32_t net_ip, u_int32_t netmask, 
-                               u_int32_t src_ip, u_int32_t dst_ip,
-                               u_int32_t size, u_int32_t *itemcount)
+static void ipt_acc_depth1_insert(struct ipt_acc_mask_16 *mask_16,
+                                  u_int32_t net_ip, u_int32_t netmask,
+                                  u_int32_t src_ip, u_int32_t dst_ip,
+                                  u_int32_t size, u_int32_t *itemcount)
 {
     /* Do we need to process src IP? */
     if ((net_ip&netmask) == (src_ip&netmask)) {
@@ -348,7 +347,7 @@
         DEBUGP("ACCOUNT: Calculated SRC 16 bit network slot: %d\n", slot);
 
         /* Do we need to create a new mask_24 bucket? */
-        if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] = 
+        if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] =
              (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
             printk("ACCOUNT: Can't process packet because out of memory!\n");
             return;
@@ -364,7 +363,7 @@
         DEBUGP("ACCOUNT: Calculated DST 16 bit network slot: %d\n", slot);
 
         /* Do we need to create a new mask_24 bucket? */
-        if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot] 
+        if (!mask_16->mask_24[slot] && (mask_16->mask_24[slot]
             = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
             printk("ACCOUT: Can't process packet because out of memory!\n");
             return;
@@ -375,10 +374,10 @@
     }
 }
 
-void ipt_acc_depth2_insert(struct ipt_acc_mask_8 *mask_8, 
-                               u_int32_t net_ip, u_int32_t netmask,
-                               u_int32_t src_ip, u_int32_t dst_ip,
-                               u_int32_t size, u_int32_t *itemcount)
+static void ipt_acc_depth2_insert(struct ipt_acc_mask_8 *mask_8,
+                                  u_int32_t net_ip, u_int32_t netmask,
+                                  u_int32_t src_ip, u_int32_t dst_ip,
+                                  u_int32_t size, u_int32_t *itemcount)
 {
     /* Do we need to process src IP? */
     if ((net_ip&netmask) == (src_ip&netmask)) {
@@ -386,7 +385,7 @@
         DEBUGP("ACCOUNT: Calculated SRC 24 bit network slot: %d\n", slot);
 
         /* Do we need to create a new mask_24 bucket? */
-        if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot] 
+        if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot]
             = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
             printk("ACCOUNT: Can't process packet because out of memory!\n");
             return;
@@ -402,7 +401,7 @@
         DEBUGP("ACCOUNT: Calculated DST 24 bit network slot: %d\n", slot);
 
         /* Do we need to create a new mask_24 bucket? */
-        if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot] 
+        if (!mask_8->mask_16[slot] && (mask_8->mask_16[slot]
             = (void *)get_zeroed_page(GFP_ATOMIC)) == NULL) {
             printk("ACCOUNT: Can't process packet because out of memory!\n");
             return;
@@ -420,7 +419,7 @@
                                        const void *targinfo,
                                        void *userinfo)
 {
-    const struct ipt_acc_info *info = 
+    const struct ipt_acc_info *info =
         (const struct ipt_acc_info *)targinfo;
     u_int32_t src_ip = (*pskb)->nh.iph->saddr;
     u_int32_t dst_ip = (*pskb)->nh.iph->daddr;
@@ -430,7 +429,7 @@
 
     if (ipt_acc_tables[info->table_nr].name[0] == 0) {
         printk("ACCOUNT: ipt_acc_target: Invalid table id %u. "
-               "IPs %u.%u.%u.%u/%u.%u.%u.%u\n", info->table_nr, 
+               "IPs %u.%u.%u.%u/%u.%u.%u.%u\n", info->table_nr,
                NIPQUAD(src_ip), NIPQUAD(dst_ip));
         spin_unlock_bh(&ipt_acc_lock);
         return IPT_CONTINUE;
@@ -441,7 +440,7 @@
         /* Count packet and check if the IP is new */
         ipt_acc_depth0_insert(
             (struct ipt_acc_mask_24 *)ipt_acc_tables[info->table_nr].data,
-            ipt_acc_tables[info->table_nr].ip, 
+            ipt_acc_tables[info->table_nr].ip,
             ipt_acc_tables[info->table_nr].netmask,
             src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
         spin_unlock_bh(&ipt_acc_lock);
@@ -452,7 +451,7 @@
     if (ipt_acc_tables[info->table_nr].depth == 1) {
         ipt_acc_depth1_insert(
             (struct ipt_acc_mask_16 *)ipt_acc_tables[info->table_nr].data,
-            ipt_acc_tables[info->table_nr].ip, 
+            ipt_acc_tables[info->table_nr].ip,
             ipt_acc_tables[info->table_nr].netmask,
             src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
         spin_unlock_bh(&ipt_acc_lock);
@@ -463,7 +462,7 @@
     if (ipt_acc_tables[info->table_nr].depth == 2) {
         ipt_acc_depth2_insert(
             (struct ipt_acc_mask_8 *)ipt_acc_tables[info->table_nr].data,
-            ipt_acc_tables[info->table_nr].ip, 
+            ipt_acc_tables[info->table_nr].ip,
             ipt_acc_tables[info->table_nr].netmask,
             src_ip, dst_ip, size, &ipt_acc_tables[info->table_nr].itemcount);
         spin_unlock_bh(&ipt_acc_lock);
@@ -471,7 +470,7 @@
     }
 
     printk("ACCOUNT: ipt_acc_target: Unable to process packet. "
-           "Table id %u. IPs %u.%u.%u.%u/%u.%u.%u.%u\n", 
+           "Table id %u. IPs %u.%u.%u.%u/%u.%u.%u.%u\n",
            info->table_nr, NIPQUAD(src_ip), NIPQUAD(dst_ip));
 
     spin_unlock_bh(&ipt_acc_lock);
@@ -481,11 +480,11 @@
 /*
     Functions dealing with "handles":
     Handles are snapshots of a accounting state.
-    
+
     read snapshots are only for debugging the code
     and are very expensive concerning speed/memory
     compared to read_and_flush.
-    
+
     The functions aren't protected by spinlocks themselves
     as this is done in the ioctl part of the code.
 */
@@ -495,14 +494,14 @@
     but there could be two or more applications accessing the data
     at the same time.
 */
-int ipt_acc_handle_find_slot(void)
+static int ipt_acc_handle_find_slot(void)
 {
     u_int32_t i;
     /* Insert new table */
     for (i = 0; i < ACCOUNT_MAX_HANDLES; i++) {
         /* Found free slot */
         if (ipt_acc_handles[i].data == NULL) {
-            /* Don't "mark" data as used as we are protected by a spinlock 
+            /* Don't "mark" data as used as we are protected by a spinlock
                by the calling function. handle_find_slot() is only a function
                to prevent code duplication. */
             return i;
@@ -515,7 +514,7 @@
     return -1;
 }
 
-int ipt_acc_handle_free(u_int32_t handle)
+static int ipt_acc_handle_free(u_int32_t handle)
 {
     if (handle >= ACCOUNT_MAX_HANDLES) {
         printk("ACCOUNT: Invalid handle for ipt_acc_handle_free() specified:"
@@ -523,7 +522,7 @@
         return -EINVAL;
     }
 
-    ipt_acc_data_free(ipt_acc_handles[handle].data, 
+    ipt_acc_data_free(ipt_acc_handles[handle].data,
                           ipt_acc_handles[handle].depth);
     memset (&ipt_acc_handles[handle], 0, sizeof (struct ipt_acc_handle));
     return 0;
@@ -531,13 +530,13 @@
 
 /* Prepare data for read without flush. Use only for debugging!
    Real applications should use read&flush as it's way more efficent */
-int ipt_acc_handle_prepare_read(char *tablename, u_int32_t *count)
+static int ipt_acc_handle_prepare_read(char *tablename, u_int32_t *count)
 {
     int handle, i, table_nr=-1;
     unsigned char depth;
 
     for (i = 0; i < ACCOUNT_MAX_TABLES; i++)
-        if (strncmp(ipt_acc_tables[i].name, tablename, 
+        if (strncmp(ipt_acc_tables[i].name, tablename,
             ACCOUNT_TABLE_NAME_LEN) == 0) {
             table_nr = i;
             break;
@@ -559,11 +558,11 @@
     ipt_acc_handles[handle].itemcount = ipt_acc_tables[table_nr].itemcount;
 
     /* allocate "root" table */
-    if ((ipt_acc_handles[handle].data = 
+    if ((ipt_acc_handles[handle].data =
          (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
         printk("ACCOUNT: out of memory for root table "
                "in ipt_acc_handle_prepare_read()\n");
-        memset (&ipt_acc_handles[handle], 0, 
+        memset (&ipt_acc_handles[handle], 0,
                 sizeof(struct ipt_acc_handle));
         return -1;
     }
@@ -571,11 +570,11 @@
     /* Recursive copy of complete data structure */
     depth = ipt_acc_handles[handle].depth;
     if (depth == 0) {
-        memcpy(ipt_acc_handles[handle].data, 
-               ipt_acc_tables[table_nr].data, 
+        memcpy(ipt_acc_handles[handle].data,
+               ipt_acc_tables[table_nr].data,
                sizeof(struct ipt_acc_mask_24));
     } else if (depth == 1) {
-        struct ipt_acc_mask_16 *src_16 = 
+        struct ipt_acc_mask_16 *src_16 =
             (struct ipt_acc_mask_16 *)ipt_acc_tables[table_nr].data;
         struct ipt_acc_mask_16 *network_16 =
             (struct ipt_acc_mask_16 *)ipt_acc_handles[handle].data;
@@ -583,40 +582,40 @@
 
         for (b = 0; b <= 255; b++) {
             if (src_16->mask_24[b]) {
-                if ((network_16->mask_24[b] = 
+                if ((network_16->mask_24[b] =
                      (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                     printk("ACCOUNT: out of memory during copy of 16 bit "
                            "network in ipt_acc_handle_prepare_read()\n");
                     ipt_acc_data_free(ipt_acc_handles[handle].data, depth);
-                    memset (&ipt_acc_handles[handle], 0, 
+                    memset (&ipt_acc_handles[handle], 0,
                             sizeof(struct ipt_acc_handle));
                     return -1;
                 }
 
-                memcpy(network_16->mask_24[b], src_16->mask_24[b], 
+                memcpy(network_16->mask_24[b], src_16->mask_24[b],
                        sizeof(struct ipt_acc_mask_24));
             }
         }
     } else if(depth == 2) {
-        struct ipt_acc_mask_8 *src_8 = 
+        struct ipt_acc_mask_8 *src_8 =
             (struct ipt_acc_mask_8 *)ipt_acc_tables[table_nr].data;
-        struct ipt_acc_mask_8 *network_8 = 
+        struct ipt_acc_mask_8 *network_8 =
             (struct ipt_acc_mask_8 *)ipt_acc_handles[handle].data;
         u_int32_t a;
 
         for (a = 0; a <= 255; a++) {
             if (src_8->mask_16[a]) {
-                if ((network_8->mask_16[a] = 
+                if ((network_8->mask_16[a] =
                      (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                     printk("ACCOUNT: out of memory during copy of 24 bit network"
                            " in ipt_acc_handle_prepare_read()\n");
                     ipt_acc_data_free(ipt_acc_handles[handle].data, depth);
-                    memset (&ipt_acc_handles[handle], 0, 
+                    memset (&ipt_acc_handles[handle], 0,
                             sizeof(struct ipt_acc_handle));
                     return -1;
                 }
 
-                memcpy(network_8->mask_16[a], src_8->mask_16[a], 
+                memcpy(network_8->mask_16[a], src_8->mask_16[a],
                        sizeof(struct ipt_acc_mask_16));
 
                 struct ipt_acc_mask_16 *src_16 = src_8->mask_16[a];
@@ -625,18 +624,18 @@
 
                 for (b = 0; b <= 255; b++) {
                     if (src_16->mask_24[b]) {
-                        if ((network_16->mask_24[b] = 
+                        if ((network_16->mask_24[b] =
                              (void*)get_zeroed_page(GFP_ATOMIC)) == NULL) {
                             printk("ACCOUNT: out of memory during copy of 16 bit"
                                    " network in ipt_acc_handle_prepare_read()\n");
                             ipt_acc_data_free(ipt_acc_handles[handle].data,
                                                   depth);
-                            memset (&ipt_acc_handles[handle], 0, 
+                            memset (&ipt_acc_handles[handle], 0,
                                     sizeof(struct ipt_acc_handle));
                             return -1;
                         }
 
-                        memcpy(network_16->mask_24[b], src_16->mask_24[b], 
+                        memcpy(network_16->mask_24[b], src_16->mask_24[b],
                                sizeof(struct ipt_acc_mask_24));
                     }
                 }
@@ -649,13 +648,13 @@
 }
 
 /* Prepare data for read and flush it */
-int ipt_acc_handle_prepare_read_flush(char *tablename, u_int32_t *count)
+static int ipt_acc_handle_prepare_read_flush(char *tablename, u_int32_t *count)
 {
     int handle, i, table_nr=-1;
     void *new_data_page;
 
     for (i = 0; i < ACCOUNT_MAX_TABLES; i++)
-        if (strncmp(ipt_acc_tables[i].name, tablename, 
+        if (strncmp(ipt_acc_tables[i].name, tablename,
                     ACCOUNT_TABLE_NAME_LEN) == 0) {
             table_nr = i;
             break;
@@ -695,18 +694,18 @@
 /* Copy 8 bit network data into a prepared buffer.
    We only copy entries != 0 to increase performance.
 */
-int ipt_acc_handle_copy_data(void *to_user, int *pos, 
-                                  struct ipt_acc_mask_24 *data,
-                                  u_int32_t net_ip, u_int32_t net_OR_mask)
+static int ipt_acc_handle_copy_data(void *to_user, int *pos,
+                                    struct ipt_acc_mask_24 *data,
+                                    u_int32_t net_ip, u_int32_t net_OR_mask)
 {
     struct ipt_acc_handle_ip handle_ip;
     u_int32_t handle_ip_size = sizeof (struct ipt_acc_handle_ip);
     u_int32_t i;
-    
+
     for (i = 0; i <= 255; i++) {
         if (data->ip[i].src_packets || data->ip[i].dst_packets) {
             handle_ip.ip = net_ip | net_OR_mask | (i<<24);
-            
+
             handle_ip.src_packets = data->ip[i].src_packets;
             handle_ip.src_bytes = data->ip[i].src_bytes;
             handle_ip.dst_packets = data->ip[i].dst_packets;
@@ -722,15 +721,15 @@
             *pos += handle_ip_size;
         }
     }
-    
+
     return 0;
 }
-   
-/* Copy the data from our internal structure 
+
+/* Copy the data from our internal structure
    We only copy entries != 0 to increase performance.
    Overwrites ipt_acc_tmpbuf.
 */
-int ipt_acc_handle_get_data(u_int32_t handle, void *to_user)
+static int ipt_acc_handle_get_data(u_int32_t handle, void *to_user)
 {
     u_int32_t tmpbuf_pos=0, net_ip;
     unsigned char depth;
@@ -751,11 +750,11 @@
 
     /* 8 bit network */
     if (depth == 0) {
-        struct ipt_acc_mask_24 *network = 
+        struct ipt_acc_mask_24 *network =
             (struct ipt_acc_mask_24*)ipt_acc_handles[handle].data;
         if (ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network, net_ip, 0))
             return -1;
-        
+
         /* Flush remaining data to userspace */
         if (tmpbuf_pos)
             if (copy_to_user(to_user, ipt_acc_tmpbuf, tmpbuf_pos))
@@ -766,14 +765,14 @@
 
     /* 16 bit network */
     if (depth == 1) {
-        struct ipt_acc_mask_16 *network_16 = 
+        struct ipt_acc_mask_16 *network_16 =
             (struct ipt_acc_mask_16*)ipt_acc_handles[handle].data;
         u_int32_t b;
         for (b = 0; b <= 255; b++) {
             if (network_16->mask_24[b]) {
-                struct ipt_acc_mask_24 *network = 
+                struct ipt_acc_mask_24 *network =
                     (struct ipt_acc_mask_24*)network_16->mask_24[b];
-                if (ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network, 
+                if (ipt_acc_handle_copy_data(to_user, &tmpbuf_pos, network,
                                              net_ip, (b << 16)))
                     return -1;
             }
@@ -789,16 +788,16 @@
 
     /* 24 bit network */
     if (depth == 2) {
-        struct ipt_acc_mask_8 *network_8 = 
+        struct ipt_acc_mask_8 *network_8 =
             (struct ipt_acc_mask_8*)ipt_acc_handles[handle].data;
         u_int32_t a, b;
         for (a = 0; a <= 255; a++) {
             if (network_8->mask_16[a]) {
-                struct ipt_acc_mask_16 *network_16 = 
+                struct ipt_acc_mask_16 *network_16 =
                     (struct ipt_acc_mask_16*)network_8->mask_16[a];
                 for (b = 0; b <= 255; b++) {
                     if (network_16->mask_24[b]) {
-                        struct ipt_acc_mask_24 *network = 
+                        struct ipt_acc_mask_24 *network =
                             (struct ipt_acc_mask_24*)network_16->mask_24[b];
                         if (ipt_acc_handle_copy_data(to_user, &tmpbuf_pos,
                                         network, net_ip, (a << 8) | (b << 16)))
@@ -815,11 +814,11 @@
 
         return 0;
     }
-    
+
     return -1;
 }
 
-static int ipt_acc_set_ctl(struct sock *sk, int cmd, 
+static int ipt_acc_set_ctl(struct sock *sk, int cmd,
                                void *user, u_int32_t len)
 {
     struct ipt_acc_handle_sockopt handle;
@@ -832,7 +831,7 @@
     case IPT_SO_SET_ACCOUNT_HANDLE_FREE:
         if (len != sizeof(struct ipt_acc_handle_sockopt)) {
             printk("ACCOUNT: ipt_acc_set_ctl: wrong data size (%u != %u) "
-                   "for IPT_SO_SET_HANDLE_FREE\n", 
+                   "for IPT_SO_SET_HANDLE_FREE\n",
                    len, sizeof(struct ipt_acc_handle_sockopt));
             break;
         }
@@ -881,7 +880,7 @@
             break;
         }
 
-        if (copy_from_user (&handle, user, 
+        if (copy_from_user (&handle, user,
                             sizeof(struct ipt_acc_handle_sockopt))) {
             return -EFAULT;
             break;
@@ -903,7 +902,7 @@
             break;
         }
 
-        if (copy_to_user(user, &handle, 
+        if (copy_to_user(user, &handle,
                          sizeof(struct ipt_acc_handle_sockopt))) {
             return -EFAULT;
             break;
@@ -918,7 +917,7 @@
             break;
         }
 
-        if (copy_from_user (&handle, user, 
+        if (copy_from_user (&handle, user,
                             sizeof(struct ipt_acc_handle_sockopt))) {
             return -EFAULT;
             break;
@@ -967,7 +966,7 @@
                     handle.itemcount++;
             spin_unlock_bh(&ipt_acc_userspace_lock);
 
-            if (copy_to_user(user, &handle, 
+            if (copy_to_user(user, &handle,
                              sizeof(struct ipt_acc_handle_sockopt))) {
                 return -EFAULT;
                 break;
@@ -978,7 +977,7 @@
     case IPT_SO_GET_ACCOUNT_GET_TABLE_NAMES: {
             u_int32_t size = 0, i;
             char *tnames;
-            
+
             spin_lock_bh(&ipt_acc_lock);
 
             /* Determine size of table names */
@@ -1044,22 +1043,22 @@
 
 static int __init init(void)
 {
-    if ((ipt_acc_tables = 
-         kmalloc(ACCOUNT_MAX_TABLES * 
+    if ((ipt_acc_tables =
+         kmalloc(ACCOUNT_MAX_TABLES *
                  sizeof(struct ipt_acc_table), GFP_KERNEL)) == NULL) {
         printk("ACCOUNT: Out of memory allocating account_tables structure");
         goto error_cleanup;
     }
-    memset(ipt_acc_tables, 0, 
+    memset(ipt_acc_tables, 0,
            ACCOUNT_MAX_TABLES * sizeof(struct ipt_acc_table));
 
-    if ((ipt_acc_handles = 
-         kmalloc(ACCOUNT_MAX_HANDLES * 
+    if ((ipt_acc_handles =
+         kmalloc(ACCOUNT_MAX_HANDLES *
                  sizeof(struct ipt_acc_handle), GFP_KERNEL)) == NULL) {
         printk("ACCOUNT: Out of memory allocating account_handles structure");
         goto error_cleanup;
     }
-    memset(ipt_acc_handles, 0, 
+    memset(ipt_acc_handles, 0,
            ACCOUNT_MAX_HANDLES * sizeof(struct ipt_acc_handle));
 
     /* Allocate one page as temporary storage */
@@ -1076,9 +1075,9 @@
 
     if (ipt_register_target(&ipt_acc_reg))
         goto error_cleanup;
-        
+
     return 0;
-        
+
 error_cleanup:
     if(ipt_acc_tables)
         kfree(ipt_acc_tables);
@@ -1086,7 +1085,7 @@
         kfree(ipt_acc_handles);
     if (ipt_acc_tmpbuf)
         free_page((unsigned long)ipt_acc_tmpbuf);
-        
+
     return -EINVAL;
 }
 

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: new ipt_ACCOUNT version
  2004-06-13 20:44       ` Patrick McHardy
@ 2004-06-13 21:09         ` Thomas Jarosch
  2004-06-13 21:29           ` Patrick McHardy
  0 siblings, 1 reply; 8+ messages in thread
From: Thomas Jarosch @ 2004-06-13 21:09 UTC (permalink / raw)
  To: Patrick McHardy; +Cc: netfilter-devel

> On Thu, 2004-06-10 at 19:07, Thomas Jarosch wrote:
> > Hi Patrick,
> >
> > Please see the attached file. Hope the userspace patch is ok.
>
> Applied with the attached patch on top (trailing whitespace cleanup
> and missing statics).

Nice! :-)

Out of curiosity, where did the "info", "help" or userspace stuff go to?

Thomas

^ permalink raw reply	[flat|nested] 8+ messages in thread

* Re: new ipt_ACCOUNT version
  2004-06-13 21:09         ` Thomas Jarosch
@ 2004-06-13 21:29           ` Patrick McHardy
  0 siblings, 0 replies; 8+ messages in thread
From: Patrick McHardy @ 2004-06-13 21:29 UTC (permalink / raw)
  To: Thomas Jarosch; +Cc: netfilter-devel

Thomas Jarosch wrote:
>>On Thu, 2004-06-10 at 19:07, Thomas Jarosch wrote:
>>
>>Applied with the attached patch on top (trailing whitespace cleanup
>>and missing statics).
> 
> Nice! :-)
> 
> Out of curiosity, where did the "info", "help" or userspace stuff go to?

File: help              Status: Locally Added

Thanks for catching this, I accidentally committed in the linux
directory.

Regards
Patrick


> 
> Thomas
> 

^ permalink raw reply	[flat|nested] 8+ messages in thread

end of thread, other threads:[~2004-06-13 21:29 UTC | newest]

Thread overview: 8+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2004-05-29 13:27 new ipt_ACCOUNT version Thomas Jarosch
2004-06-06 22:40 ` Patrick McHardy
2004-06-07 16:11 ` Patrick McHardy
2004-06-07 19:48   ` Thomas Jarosch
2004-06-09  8:39     ` Patrick McHardy
     [not found] ` <200406091133.59850.thomas.jarosch@intra2net.com>
     [not found]   ` <40C6DB4E.2080804@trash.net>
     [not found]     ` <200406101907.05352.thomas.jarosch@intra2net.com>
2004-06-13 20:44       ` Patrick McHardy
2004-06-13 21:09         ` Thomas Jarosch
2004-06-13 21:29           ` Patrick McHardy

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.