/*
 *      NET3:   Implementation of the ICMP protocol layer. 
 *      
 *              Alan Cox, <alan@cymru.net>
 *
 *      This program is free software; you can redistribute it and/or
 *      modify it under the terms of the GNU General Public License
 *      as published by the Free Software Foundation; either version
 *      2 of the License, or (at your option) any later version.
 *
 *      Some of the function names and the icmp unreach table for this
 *      module were derived from [icmp.c 1.0.11 06/02/93] by
 *      Ross Biro, Fred N. van Kempen, Mark Evans, Alan Cox, Gerhard Koerting.
 *      Other than that this module is a complete rewrite.
 *
 *      Fixes:
 *              Mike Shaver     :       RFC1122 checks.
 *              Alan Cox        :       Multicast ping reply as self.
 *              Alan Cox        :       Fix atomicity lockup in ip_build_xmit 
 *                                      call.
 *              Alan Cox        :       Added 216,128 byte paths to the MTU 
 *                                      code.
 *              Martin Mares    :       RFC1812 checks.
 *              Martin Mares    :       Can be configured to follow redirects 
 *                                      if acting as a router _without_ a
 *                                      routing protocol (RFC 1812).
 *              Martin Mares    :       Echo requests may be configured to 
 *                                      be ignored (RFC 1812).
 *              Martin Mares    :       Limitation of ICMP error message 
 *                                      transmit rate (RFC 1812).
 *              Martin Mares    :       TOS and Precedence set correctly 
 *                                      (RFC 1812).
 *              Martin Mares    :       Now copying as much data from the 
 *                                      original packet as we can without
 *                                      exceeding 576 bytes (RFC 1812).
 *      Willy Konynenberg       :       Transparent proxying support.
 *              Keith Owens     :       RFC1191 correction for 4.2BSD based 
 *                                      path MTU bug.
 *
 *
 * RFC1122 (Host Requirements -- Comm. Layer) Status:
 * (boy, are there a lot of rules for ICMP)
 *  3.2.2 (Generic ICMP stuff)
 *   MUST discard messages of unknown type. (OK)
 *   MUST copy at least the first 8 bytes from the offending packet
 *     when sending ICMP errors. (OBSOLETE -- see RFC1812)
 *   MUST pass received ICMP errors up to protocol level. (OK)
 *   SHOULD send ICMP errors with TOS == 0. (OBSOLETE -- see RFC1812)
 *   MUST NOT send ICMP errors in reply to:
 *     ICMP errors (OK)
 *     Broadcast/multicast datagrams (OK)
 *     MAC broadcasts (OK)
 *     Non-initial fragments (OK)
 *     Datagram with a source address that isn't a single host. (OK)
 *  3.2.2.1 (Destination Unreachable)
 *   All the rules govern the IP layer, and are dealt with in ip.c, not here.
 *  3.2.2.2 (Redirect)
 *   Host SHOULD NOT send ICMP_REDIRECTs.  (OK)
 *   MUST update routing table in response to host or network redirects.
 *     (host OK, network OBSOLETE)
 *   SHOULD drop redirects if they're not from directly connected gateway
 *     (OK -- we drop it if it's not from our old gateway, which is close
 *      enough)
 * 3.2.2.3 (Source Quench)
 *   MUST pass incoming SOURCE_QUENCHs to transport layer (OK)
 *   Other requirements are dealt with at the transport layer.
 * 3.2.2.4 (Time Exceeded)
 *   MUST pass TIME_EXCEEDED to transport layer (OK)
 *   Other requirements dealt with at IP (generating TIME_EXCEEDED).
 * 3.2.2.5 (Parameter Problem)
 *   SHOULD generate these (OK)
 *   MUST pass received PARAMPROBLEM to transport layer (NOT YET)
 *      [Solaris 2.X seems to assert EPROTO when this occurs] -- AC
 * 3.2.2.6 (Echo Request/Reply)
 *   MUST reply to ECHO_REQUEST, and give app to do ECHO stuff (OK, OK)
 *   MAY discard broadcast ECHO_REQUESTs. (We don't, but that's OK.)
 *   MUST reply using same source address as the request was sent to.
 *     We're OK for unicast ECHOs, and it doesn't say anything about
 *     how to handle broadcast ones, since it's optional.
 *   MUST copy data from REQUEST to REPLY (OK)
 *     unless it would require illegal fragmentation (OK)
 *   MUST pass REPLYs to transport/user layer (OK)
 *   MUST use any provided source route (reversed) for REPLY. (NOT YET)
 * 3.2.2.7 (Information Request/Reply)
 *   MUST NOT implement this. (I guess that means silently discard...?) (OK)
 * 3.2.2.8 (Timestamp Request/Reply)
 *   MAY implement (OK)
 *   SHOULD be in-kernel for "minimum variability" (OK)
 *   MAY discard broadcast REQUESTs.  (OK, but see source for inconsistency)
 *   MUST reply using same source address as the request was sent to. (OK)
 *   MUST reverse source route, as per ECHO (NOT YET)
 *   MUST pass REPLYs to transport/user layer (requires RAW, just like 
 *      ECHO) (OK)
 *   MUST update clock for timestamp at least 15 times/sec (OK)
 *   MUST be "correct within a few minutes" (OK)
 * 3.2.2.9 (Address Mask Request/Reply)
 *   MAY implement (OK)
 *   MUST send a broadcast REQUEST if using this system to set netmask
 *     (OK... we don't use it)
 *   MUST discard received REPLYs if not using this system (OK)
 *   MUST NOT send replies unless specifically made agent for this sort
 *     of thing. (OK)
 *
 *
 * RFC 1812 (IPv4 Router Requirements) Status (even longer):
 *  4.3.2.1 (Unknown Message Types)
 *   MUST pass messages of unknown type to ICMP user iface or silently discard
 *     them (OK)
 *  4.3.2.2 (ICMP Message TTL)
 *   MUST initialize TTL when originating an ICMP message (OK)
 *  4.3.2.3 (Original Message Header)
 *   SHOULD copy as much data from the offending packet as possible without
 *     the length of the ICMP datagram exceeding 576 bytes (OK)
 *   MUST leave original IP header of the offending packet, but we're not
 *     required to undo modifications made (OK)
 *  4.3.2.4 (Original Message Source Address)
 *   MUST use one of addresses for the interface the orig. packet arrived as
 *     source address (OK)
 *  4.3.2.5 (TOS and Precedence)
 *   SHOULD leave TOS set to the same value unless the packet would be 
 *     discarded for that reason (OK)
 *   MUST use TOS=0 if not possible to leave original value (OK)
 *   MUST leave IP Precedence for Source Quench messages (OK -- not sent 
 *      at all)
 *   SHOULD use IP Precedence = 6 (Internetwork Control) or 7 (Network Control)
 *     for all other error messages (OK, we use 6)
 *   MAY allow configuration of IP Precedence (OK -- not done)
 *   MUST leave IP Precedence and TOS for reply messages (OK)
 *  4.3.2.6 (Source Route)
 *   SHOULD use reverse source route UNLESS sending Parameter Problem on source
 *     routing and UNLESS the packet would be immediately discarded (NOT YET)
 *  4.3.2.7 (When Not to Send ICMP Errors)
 *   MUST NOT send ICMP errors in reply to:
 *     ICMP errors (OK)
 *     Packets failing IP header validation tests unless otherwise noted (OK)
 *     Broadcast/multicast datagrams (OK)
 *     MAC broadcasts (OK)
 *     Non-initial fragments (OK)
 *     Datagram with a source address that isn't a single host. (OK)
 *  4.3.2.8 (Rate Limiting)
 *   SHOULD be able to limit error message rate (OK)
 *   SHOULD allow setting of rate limits (OK, in the source)
 *  4.3.3.1 (Destination Unreachable)
 *   All the rules govern the IP layer, and are dealt with in ip.c, not here.
 *  4.3.3.2 (Redirect)
 *   MAY ignore ICMP Redirects if running a routing protocol or if forwarding
 *     is enabled on the interface (OK -- ignores)
 *  4.3.3.3 (Source Quench)
 *   SHOULD NOT originate SQ messages (OK)
 *   MUST be able to limit SQ rate if originates them (OK as we don't 
 *      send them)
 *   MAY ignore SQ messages it receives (OK -- we don't)
 *  4.3.3.4 (Time Exceeded)
 *   Requirements dealt with at IP (generating TIME_EXCEEDED).
 *  4.3.3.5 (Parameter Problem)
 *   MUST generate these for all errors not covered by other messages (OK)
 *   MUST include original value of the value pointed by (OK)
 *  4.3.3.6 (Echo Request)
 *   MUST implement echo server function (OK)
 *   MUST process at ER of at least max(576, MTU) (OK)
 *   MAY reject broadcast/multicast ER's (We don't, but that's OK)
 *   SHOULD have a config option for silently ignoring ER's (OK)
 *   MUST have a default value for the above switch = NO (OK)
 *   MUST have application layer interface for Echo Request/Reply (OK)
 *   MUST reply using same source address as the request was sent to.
 *     We're OK for unicast ECHOs, and it doesn't say anything about
 *     how to handle broadcast ones, since it's optional.
 *   MUST copy data from Request to Reply (OK)
 *   SHOULD update Record Route / Timestamp options (??)
 *   MUST use reversed Source Route for Reply if possible (NOT YET)
 *  4.3.3.7 (Information Request/Reply)
 *   SHOULD NOT originate or respond to these (OK)
 *  4.3.3.8 (Timestamp / Timestamp Reply)
 *   MAY implement (OK)
 *   MUST reply to every Timestamp message received (OK)
 *   MAY discard broadcast REQUESTs.  (OK, but see source for inconsistency)
 *   MUST reply using same source address as the request was sent to. (OK)
 *   MUST use reversed Source Route if possible (NOT YET)
 *   SHOULD update Record Route / Timestamp options (??)
 *   MUST pass REPLYs to transport/user layer (requires RAW, just like 
 *      ECHO) (OK)
 *   MUST update clock for timestamp at least 16 times/sec (OK)
 *   MUST be "correct within a few minutes" (OK)
 * 4.3.3.9 (Address Mask Request/Reply)
 *   MUST have support for receiving AMRq and responding with AMRe (OK, 
 *      but only as a compile-time option)
 *   SHOULD have option for each interface for AMRe's, MUST default to 
 *      NO (NOT YET)
 *   MUST NOT reply to AMRq before knows the correct AM (OK)
 *   MUST NOT respond to AMRq with source address 0.0.0.0 on physical
 *      interfaces having multiple logical i-faces with different masks
 *      (NOT YET)
 *   SHOULD examine all AMRe's it receives and check them (NOT YET)
 *   SHOULD log invalid AMRe's (AM+sender) (NOT YET)
 *   MUST NOT use contents of AMRe to determine correct AM (OK)
 *   MAY broadcast AMRe's after having configured address masks (OK -- doesn't)
 *   MUST NOT do broadcast AMRe's if not set by extra option (OK, no option)
 *   MUST use the { <NetPrefix>, -1 } form of broadcast addresses (OK)
 * 4.3.3.10 (Router Advertisement and Solicitations)
 *   MUST support router part of Router Discovery Protocol on all networks we
 *     support broadcast or multicast addressing. (OK -- done by gated)
 *   MUST have all config parameters with the respective defaults (OK)
 * 5.2.7.1 (Destination Unreachable)
 *   MUST generate DU's (OK)
 *   SHOULD choose a best-match response code (OK)
 *   SHOULD NOT generate Host Isolated codes (OK)
 *   SHOULD use Communication Administratively Prohibited when administratively
 *     filtering packets (NOT YET -- bug-to-bug compatibility)
 *   MAY include config option for not generating the above and silently
 *      discard the packets instead (OK)
 *   MAY include config option for not generating Precedence Violation and
 *     Precedence Cutoff messages (OK as we don't generate them at all)
 *   MUST use Host Unreachable or Dest. Host Unknown codes whenever other hosts
 *     on the same network might be reachable (OK -- no net unreach's at all)
 *   MUST use new form of Fragmentation Needed and DF Set messages (OK)
 * 5.2.7.2 (Redirect)
 *   MUST NOT generate network redirects (OK)
 *   MUST be able to generate host redirects (OK)
 *   SHOULD be able to generate Host+TOS redirects (NO as we don't use TOS)
 *   MUST have an option to use Host redirects instead of Host+TOS ones (OK as
 *     no Host+TOS Redirects are used)
 *   MUST NOT generate redirects unless forwarding to the same i-face and the
 *     dest. address is on the same subnet as the src. address and no source
 *     routing is in use. (OK)
 *   MUST NOT follow redirects when using a routing protocol (OK)
 *   MAY use redirects if not using a routing protocol (OK, compile-time option)
 *   MUST comply to Host Requirements when not acting as a router (OK)
 *  5.2.7.3 (Time Exceeded)
 *   MUST generate Time Exceeded Code 0 when discarding packet due to TTL=0 (OK)
 *   MAY have a per-interface option to disable origination of TE messages, but
 *     it MUST default to "originate" (OK -- we don't support it)
 */

#include <linux/config.h>
#include <linux/types.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/fcntl.h>
#include <linux/socket.h>
#include <linux/in.h>
#include <linux/inet.h>
#include <linux/netdevice.h>
#include <linux/string.h>
#include <net/snmp.h>
#include <net/ip.h>
#include <net/route.h>
#include <net/protocol.h>
#include <net/icmp.h>
#include <net/tcp.h>
#include <net/udp.h>
#include <net/snmp.h>
#include <linux/skbuff.h>
#include <net/sock.h>
#include <linux/errno.h>
#include <linux/timer.h>
#include <asm/system.h>
#include <asm/segment.h>
#include <net/checksum.h>

#define min(a,b)        ((a)<(b)?(a):(b))

/*
 *      Statistics
 */
 
struct icmp_mib icmp_statistics;

/* An array of errno for error messages from dest unreach. */
/* RFC 1122: 3.2.2.1 States that NET_UNREACH, HOS_UNREACH and SR_FAIELD MUST be considered 'transient errs'. */

struct icmp_err icmp_err_convert[] = {
  { ENETUNREACH,        0 },    /*      ICMP_NET_UNREACH        */
  { EHOSTUNREACH,       0 },    /*      ICMP_HOST_UNREACH       */
  { ENOPROTOOPT,        1 },    /*      ICMP_PROT_UNREACH       */
  { ECONNREFUSED,       1 },    /*      ICMP_PORT_UNREACH       */
  { EOPNOTSUPP,         0 },    /*      ICMP_FRAG_NEEDED        */
  { EOPNOTSUPP,         0 },    /*      ICMP_SR_FAILED          */
  { ENETUNREACH,        1 },    /*      ICMP_NET_UNKNOWN        */
  { EHOSTDOWN,          1 },    /*      ICMP_HOST_UNKNOWN       */
  { ENONET,             1 },    /*      ICMP_HOST_ISOLATED      */
  { ENETUNREACH,        1 },    /*      ICMP_NET_ANO            */
  { EHOSTUNREACH,       1 },    /*      ICMP_HOST_ANO           */
  { EOPNOTSUPP,         0 },    /*      ICMP_NET_UNR_TOS        */
  { EOPNOTSUPP,         0 }     /*      ICMP_HOST_UNR_TOS       */
};

/*
 *      A spare long used to speed up statistics updating
 */
 
unsigned long dummy;

/*
 *      ICMP transmit rate limit control structures. We use a relatively simple
 *      approach to the problem: For each type of ICMP message with rate limit
 *      we count the number of messages sent during some time quantum. If this
 *      count exceeds given maximal value, we ignore all messages not separated
 *      from the last message sent at least by specified time.
 */

#define XRLIM_CACHE_SIZE 16             /* How many destination hosts do we cache */

struct icmp_xrl_cache                   /* One entry of the ICMP rate cache */
{
        __u32 daddr;                    /* Destination address */
        unsigned long counter;          /* Message counter */
        unsigned long next_reset;       /* Time of next reset of the counter */
        unsigned long last_access;      /* Time of last access to this entry (LRU) */
        unsigned int restricted;        /* Set if we're in restricted mode */
        unsigned long next_packet;      /* When we'll allow a next packet if restricted */
};

struct icmp_xrlim
{
        unsigned long timeout;          /* Time quantum for rate measuring */
        unsigned long limit;            /* Maximal number of messages per time quantum allowed */
        unsigned long delay;            /* How long we wait between packets when restricting */
        struct icmp_xrl_cache cache[XRLIM_CACHE_SIZE];  /* Rate cache */
};

/*
 *      ICMP control array. This specifies what to do with each ICMP.
 */
 
struct icmp_control
{
        unsigned long *output;          /* Address to increment on output */
        unsigned long *input;           /* Address to increment on input */
        void (*handler)(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 saddr, __u32 daddr, int len);
        unsigned long error;            /* This ICMP is classed as an error message */
        struct icmp_xrlim *xrlim;       /* Transmit rate limit control structure or NULL for no limits */
};

static struct icmp_control icmp_pointers[19];

/*
 *      Build xmit assembly blocks
 */

struct icmp_bxm
{
        void *data_ptr;
        int data_len;
        struct icmphdr icmph;
        unsigned long csum;
        struct options replyopts;
        unsigned char  optbuf[40];
};

/*
 *      The ICMP socket. This is the most convenient way to flow control
 *      our ICMP output as well as maintain a clean interface throughout
 *      all layers. All Socketless IP sends will soon be gone.
 */
        
struct socket icmp_socket;

/*
 *      Send an ICMP frame.
 */


/*
 *      Initialize the transmit rate limitation mechanism.
 */

#ifndef CONFIG_NO_ICMP_LIMIT

static void xrlim_init(void)
{
        int type, entry;
        struct icmp_xrlim *xr;

        for (type=0; type<=18; type++) {
                xr = icmp_pointers[type].xrlim;
                if (xr) {
                        for (entry=0; entry<XRLIM_CACHE_SIZE; entry++)
                                xr->cache[entry].daddr = INADDR_NONE;
                }
        }
}

/*
 *      Check transmit rate limitation for given message.
 *
 *      RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
 *                        SHOULD allow setting of rate limits (we allow 
 *                        in the source)
 */

static int xrlim_allow(int type, __u32 addr)
{
        struct icmp_xrlim *r;
        struct icmp_xrl_cache *c;
        unsigned long now;

        if (type > 18)                  /* No time limit present */
                return 1;
        r = icmp_pointers[type].xrlim;
        if (!r)
                return 1;

        for (c = r->cache; c < &r->cache[XRLIM_CACHE_SIZE]; c++)        
          /* Cache lookup */
                if (c->daddr == addr)
                        break;

        now = jiffies;          /* Cache current time (saves accesses to volatile variable) */

        if (c == &r->cache[XRLIM_CACHE_SIZE]) {         /* Cache miss */
                unsigned long oldest = now;             /* Find the oldest entry to replace */
                struct icmp_xrl_cache *d;
                c = r->cache;
                for (d = r->cache; d < &r->cache[XRLIM_CACHE_SIZE]; d++)
                        if (!d->daddr) {                /* Unused entry */
                                c = d;
                                break;
                        } else if (d->last_access < oldest) {
                                oldest = d->last_access;
                                c = d;
                        }
                c->last_access = now;                   /* Fill the entry with new data */
                c->daddr = addr;
                c->counter = 1;
                c->next_reset = now + r->timeout;
                c->restricted = 0;
                return 1;
        }

        c->last_access = now;
        if (c->next_reset > now) {                      /* Let's increment the counter */
                c->counter++;
                if (c->counter == r->limit) {           /* Limit exceeded, start restrictions */
                        c->restricted = 1;
                        c->next_packet = now + r->delay;
                        return 0;
                }
                if (c->restricted) {                    /* Any restrictions pending? */
                        if (c->next_packet > now)
                                return 0;
                        c->next_packet = now + r->delay;
                        return 1;
                }
        } else {                                        /* Reset the counter */
                if (c->counter < r->limit)              /* Switch off all restrictions */
                        c->restricted = 0;
                c->next_reset = now + r->timeout;
                c->counter = 0;
        }

        return 1;                                       /* Send the packet */
}

#endif /* CONFIG_NO_ICMP_LIMIT */

/*
 *      Maintain the counters used in the SNMP statistics for outgoing ICMP
 */
 
static void icmp_out_count(int type)
{
        if(type>18)
                return;
        (*icmp_pointers[type].output)++;
        icmp_statistics.IcmpOutMsgs++;
}
 
/*
 *      Checksum each fragment, and on the first include the headers and final checksum.
 */
 
static void icmp_glue_bits(const void *p, __u32 saddr, char *to, unsigned int offset, unsigned int fraglen)
{
        struct icmp_bxm *icmp_param = (struct icmp_bxm *)p;
        struct icmphdr *icmph;
        unsigned long csum;

        if (offset) {
                icmp_param->csum=csum_partial_copy(icmp_param->data_ptr+offset-sizeof(struct icmphdr), 
                                to, fraglen,icmp_param->csum);
                return;
        }

        /*
         *      First fragment includes header. Note that we've done
         *      the other fragments first, so that we get the checksum
         *      for the whole packet here.
         */
        csum = csum_partial_copy((void *)&icmp_param->icmph,
                to, sizeof(struct icmphdr), 
                icmp_param->csum);
        csum = csum_partial_copy(icmp_param->data_ptr,
                to+sizeof(struct icmphdr),
                fraglen-sizeof(struct icmphdr), csum);
        icmph=(struct icmphdr *)to;
        icmph->checksum = csum_fold(csum);
}
 
/*
 *      Driving logic for building and sending ICMP messages.
 */

static void icmp_build_xmit(struct icmp_bxm *icmp_param, __u32 saddr, __u32 daddr, __u8 tos)
{
        struct sock *sk=icmp_socket.data;
        icmp_param->icmph.checksum=0;
        icmp_param->csum=0;
        icmp_out_count(icmp_param->icmph.type);
        sk->ip_tos = tos;
        ip_build_xmit(sk, icmp_glue_bits, icmp_param, 
                icmp_param->data_len+sizeof(struct icmphdr),
                daddr, saddr, &icmp_param->replyopts, 0, IPPROTO_ICMP, 1);
}


/*
 *      Send an ICMP message in response to a situation
 *
 *      RFC 1122: 3.2.2 MUST send at least the IP header and 8 bytes of header. MAY send more (we do).
 *                      MUST NOT change this header information.
 *                      MUST NOT reply to a multicast/broadcast IP address.
 *                      MUST NOT reply to a multicast/broadcast MAC address.
 *                      MUST reply to only the first fragment.
 */

void icmp_send(struct sk_buff *skb_in, int type, int code, unsigned long info, struct device *dev)
{
        struct iphdr *iph;
        struct icmphdr *icmph;
        int atype, room;
        struct icmp_bxm icmp_param;
        __u32 saddr;
        
        /*
         *      Find the original header
         */
         
        iph = skb_in->ip_hdr;
        
        /*
         *      No replies to physical multicast/broadcast
         */
         
        if(skb_in->pkt_type!=PACKET_HOST)
                return;
                
        /*
         *      Now check at the protocol level
         */
         
        atype=ip_chk_addr(iph->daddr);
        if(atype==IS_BROADCAST||atype==IS_MULTICAST)
                return;
                
        /*
         *      Only reply to fragment 0. We byte re-order the constant
         *      mask for efficiency.
         */
         
        if(iph->frag_off&htons(IP_OFFSET))
                return;
                
        /* 
         *      If we send an ICMP error to an ICMP error a mess would result..
         */
         
        if(icmp_pointers[type].error)
        {
                /*
                 *      We are an error, check if we are replying to an ICMP error
                 */
                 
                if(iph->protocol==IPPROTO_ICMP)
                {
                        icmph = (struct icmphdr *)((char *)iph + (iph->ihl<<2));
                        /*
                         *      Assume any unknown ICMP type is an error. This isn't
                         *      specified by the RFC, but think about it..
                         */
                        if(icmph->type>18 || icmp_pointers[icmph->type].error)
                                return;
                }
        }

        /*
         *      Check the rate limit
         */

#ifndef CONFIG_NO_ICMP_LIMIT
        if (!xrlim_allow(type, iph->saddr))
                return;
#endif  

        /*
         *      Construct source address and options.
         */
         
        saddr=iph->daddr;
        if(saddr!=dev->pa_addr && ip_chk_addr(saddr)!=IS_MYADDR)
                saddr=dev->pa_addr;
        if(ip_options_echo(&icmp_param.replyopts, NULL, saddr, iph->saddr, skb_in))
                return;

        /*
         *      Prepare data for ICMP header.
         */

        icmp_param.icmph.type=type;
        icmp_param.icmph.code=code;
        icmp_param.icmph.un.gateway = info;
        icmp_param.data_ptr=iph;
        room = 576 - sizeof(struct iphdr) - icmp_param.replyopts.optlen;
        icmp_param.data_len=(iph->ihl<<2)+skb_in->len;  /* RFC says return as much as we can without exceeding 576 bytes */
        if (icmp_param.data_len > room)
                icmp_param.data_len = room;
        
        /*
         *      Build and send the packet.
         */

        icmp_build_xmit(&icmp_param, saddr, iph->saddr, ((iph->tos & 0x38) | 6));
}


/* 
 *      Handle ICMP_DEST_UNREACH, ICMP_TIME_EXCEED, and ICMP_QUENCH. 
 */
 
static void icmp_unreach(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 saddr, __u32 daddr, int len)
{
        struct iphdr *iph;
        int hash;
        struct inet_protocol *ipprot;
        unsigned char *dp;      
        
        iph = (struct iphdr *) (icmph + 1);
        
        dp= ((unsigned char *)iph)+(iph->ihl<<2);
        
        if(icmph->type==ICMP_DEST_UNREACH)
        {
                switch(icmph->code & 15)
                {
                        case ICMP_NET_UNREACH:
                                break;
                        case ICMP_HOST_UNREACH:
                                break;
                        case ICMP_PROT_UNREACH:
/*                              printk(KERN_INFO "ICMP: %s:%d: protocol unreachable.\n",
                                        in_ntoa(iph->daddr), (int)iph->protocol);*/
                                break;
                        case ICMP_PORT_UNREACH:
                                break;
                        case ICMP_FRAG_NEEDED:
#ifdef CONFIG_NO_PATH_MTU_DISCOVERY
                                printk(KERN_INFO "ICMP: %s: fragmentation needed and DF set.\n",
                                                                in_ntoa(iph->daddr));
                                break;
#else
                        {
                                unsigned short old_mtu = ntohs(iph->tot_len);
                                unsigned short new_mtu = ntohs(icmph->un.echo.sequence);

                                /*
                                 * RFC1191 5.  4.2BSD based router can return incorrect
                                 * Total Length.  If current mtu is unknown or old_mtu
                                 * is not less than current mtu, reduce old_mtu by 4 times
                                 * the header length.
                                 */

                                if (skb->sk == NULL /* can this happen? */
                                        || skb->sk->ip_route_cache == NULL
                                        || skb->sk->ip_route_cache->rt_mtu <= old_mtu)
                                {
                                        NETDEBUG(printk(KERN_INFO "4.2BSD based fragmenting router between here and %s, mtu corrected from %d", in_ntoa(iph->daddr), old_mtu));
                                        old_mtu -= 4 * iph->ihl;
                                        NETDEBUG(printk(" to %d\n", old_mtu));
                                }

                                if (new_mtu < 68 || new_mtu >= old_mtu)
                                {
                                        /*
                                         *      It is either dumb router, which does not
                                         *      understand Path MTU Disc. protocol
                                         *      or broken (f.e. Linux<=1.3.37 8) router.
                                         *      Try to guess...
                                         *      The table is taken from RFC-1191.
                                         */
                                        if (old_mtu > 32000)
                                                new_mtu = 32000;
                                        else if (old_mtu > 17914)
                                                new_mtu = 17914;
                                        else if (old_mtu > 8166)
                                                new_mtu = 8166;
                                        else if (old_mtu > 4352)
                                                new_mtu = 4352;
                                        else if (old_mtu > 2002)
                                                new_mtu = 2002;
                                        else if (old_mtu > 1492)
                                                new_mtu = 1492;
                                        else if (old_mtu > 576)
                                                new_mtu = 576;
                                        else if (old_mtu > 296)
                                                new_mtu = 296;
                                        /*
                                         *      These two are not from the RFC but
                                         *      are needed for AMPRnet AX.25 paths.
                                         */
                                        else if (old_mtu > 216)
                                                new_mtu = 216;
                                        else if (old_mtu > 128)
                                                new_mtu = 128;
                                        else
                                        /*
                                         *      Despair..
                                         */
                                                new_mtu = 68;
                                }
                                /*
                                 * Ugly trick to pass MTU to protocol layer.
                                 * Really we should add argument "info" to error handler.
                                 */
                                iph->id = htons(new_mtu);
                                break;
                        }
#endif
                        case ICMP_SR_FAILED:
                                printk(KERN_INFO "ICMP: %s: Source Route Failed.\n", in_ntoa(iph->daddr));
                                break;
                        default:
                                break;
                }
                if(icmph->code>12)      /* Invalid type */
                        return;
        }
        
        /*
         *      Throw it at our lower layers
         *
         *      RFC 1122: 3.2.2 MUST extract the protocol ID from the passed header.
         *      RFC 1122: 3.2.2.1 MUST pass ICMP unreach messages to the transport layer.
         *      RFC 1122: 3.2.2.2 MUST pass ICMP time expired messages to transport layer.
         */

        /*
         *      Get the protocol(s). 
         */
         
        hash = iph->protocol & (MAX_INET_PROTOS -1);

        /*
         *      This can't change while we are doing it. 
         *
         *      FIXME: Deliver to appropriate raw sockets too.
         */
         
        ipprot = (struct inet_protocol *) inet_protos[hash];
        while(ipprot != NULL) 
        {
                struct inet_protocol *nextip;

                nextip = (struct inet_protocol *) ipprot->next;
        
                /* 
                 *      Pass it off to everyone who wants it. 
                 */

                /* RFC1122: OK. Passes appropriate ICMP errors to the */
                /* appropriate protocol layer (MUST), as per 3.2.2. */

                if (iph->protocol == ipprot->protocol && ipprot->err_handler) 
                {
                        ipprot->err_handler(icmph->type, icmph->code, dp,
                                            iph->daddr, iph->saddr, ipprot);
                }

                ipprot = nextip;
        }
        kfree_skb(skb, FREE_READ);
}


/*
 *      Handle ICMP_REDIRECT. 
 */

static void icmp_redirect(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 source, __u32 daddr, int len)
{
        struct iphdr *iph;
        unsigned long ip;

        /*
         *      Get the copied header of the packet that caused the redirect
         */
         
        iph = (struct iphdr *) (icmph + 1);
        ip = iph->daddr;

        /*
         *      If we are a router and we run a routing protocol, we MUST NOT follow redirects.
         *      When using no routing protocol, we MAY follow redirects. (RFC 1812, 5.2.7.2)
         */

#if defined(CONFIG_IP_FORWARD) && !defined(CONFIG_IP_DUMB_ROUTER)
        NETDEBUG(printk(KERN_INFO "icmp: ICMP redirect ignored. dest = %lX, "
               "orig gw = %lX, \"new\" gw = %lX, device = %s.\n", ntohl(ip),
                ntohl(source), ntohl(icmph->un.gateway), dev->name));
#else   
        switch(icmph->code & 7) 
        {
                case ICMP_REDIR_NET:
                        /*
                         *      This causes a problem with subnetted networks. What we should do
                         *      is use ICMP_ADDRESS to get the subnet mask of the problem route
                         *      and set both. But we don't.. [RFC1812 says routers MUST NOT
                         *      generate Network Redirects]
                         */
#ifdef not_a_good_idea
                        ip_rt_add((RTF_DYNAMIC | RTF_MODIFIED | RTF_GATEWAY),
                                ip, 0, icmph->un.gateway, dev,0, 0, 0);
#endif
                        /*
                         *      As per RFC recommendations now handle it as
                         *      a host redirect.
                         */
                         
                case ICMP_REDIR_HOST:
                        /*
                         *      Add better route to host.
                         *      But first check that the redirect
                         *      comes from the old gateway..
                         *      And make sure it's an ok host address
                         *      (not some confused thing sending our
                         *      address)
                         */
                        printk(KERN_INFO "ICMP redirect from %s\n", in_ntoa(source));
                        ip_rt_redirect(source, ip, icmph->un.gateway, dev);
                        break;
                case ICMP_REDIR_NETTOS:
                case ICMP_REDIR_HOSTTOS:
                        printk(KERN_INFO "ICMP: cannot handle TOS redirects yet!\n");
                        break;
                default:
                        break;
        }
#endif          
        /*
         *      Discard the original packet
         */
         
        kfree_skb(skb, FREE_READ);
}

/*
 *      Handle ICMP_ECHO ("ping") requests. 
 *
 *      RFC 1122: 3.2.2.6 MUST have an echo server that answers ICMP echo requests.
 *      RFC 1122: 3.2.2.6 Data received in the ICMP_ECHO request MUST be included in the reply.
 *      RFC 1812: 4.3.3.6 SHOULD have a config option for silently ignoring echo requests, MUST have default=NOT.
 *      See also WRT handling of options once they are done and working.
 */
 
static void icmp_echo(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 saddr, __u32 daddr, int len)
{
#ifndef CONFIG_IP_IGNORE_ECHO_REQUESTS
        struct icmp_bxm icmp_param;
        icmp_param.icmph=*icmph;
        icmp_param.icmph.type=ICMP_ECHOREPLY;
        icmp_param.data_ptr=(icmph+1);
        icmp_param.data_len=len;
        if (ip_options_echo(&icmp_param.replyopts, NULL, daddr, saddr, skb)==0)
                icmp_build_xmit(&icmp_param, daddr, saddr, skb->ip_hdr->tos);
#endif
        kfree_skb(skb, FREE_READ);
}

/*
 *      Handle ICMP Timestamp requests. 
 *      RFC 1122: 3.2.2.8 MAY implement ICMP timestamp requests.
 *                SHOULD be in the kernel for minimum random latency.
 *                MUST be accurate to a few minutes.
 *                MUST be updated at least at 15Hz.
 */
 
static void icmp_timestamp(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 saddr, __u32 daddr, int len)
{
        __u32 times[3];         /* So the new timestamp works on ALPHA's.. */
        struct icmp_bxm icmp_param;
        
        /*
         *      Too short.
         */
         
        if(len<12)
        {
                icmp_statistics.IcmpInErrors++;
                kfree_skb(skb, FREE_READ);
                return;
        }
        
        /*
         *      Fill in the current time as ms since midnight UT: 
         */
         
        {
                struct timeval tv;
                do_gettimeofday(&tv);
                times[1] = htonl((tv.tv_sec % 86400) * 1000 + tv.tv_usec / 1000);
        }
        times[2] = times[1];
        memcpy((void *)×[0], icmph+1, 4);               /* Incoming stamp */
        icmp_param.icmph=*icmph;
        icmp_param.icmph.type=ICMP_TIMESTAMPREPLY;
        icmp_param.icmph.code=0;
        icmp_param.data_ptr=×
        icmp_param.data_len=12;
        if (ip_options_echo(&icmp_param.replyopts, NULL, daddr, saddr, skb)==0)
                icmp_build_xmit(&icmp_param, daddr, saddr, skb->ip_hdr->tos);
        kfree_skb(skb,FREE_READ);
}


/* 
 *      Handle ICMP_ADDRESS_MASK requests.  (RFC950)
 *
 * RFC1122 (3.2.2.9).  A host MUST only send replies to 
 * ADDRESS_MASK requests if it's been configured as an address mask 
 * agent.  Receiving a request doesn't constitute implicit permission to 
 * act as one. Of course, implementing this correctly requires (SHOULD) 
 * a way to turn the functionality on and off.  Another one for sysctl(), 
 * I guess. -- MS 
 * Botched with a CONFIG option for now - Linus add scts sysctl please.. 
 */
 
static void icmp_address(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 saddr, __u32 daddr, int len)
{
#ifdef CONFIG_IP_ADDR_AGENT     /* Don't use, broken */
        struct icmp_bxm icmp_param;
        icmp_param.icmph.type=ICMP_ADDRESSREPLY;
        icmp_param.icmph.code=0;
        icmp_param.icmph.un.echo.id = icmph->un.echo.id;
        icmp_param.icmph.un.echo.sequence = icmph->un.echo.sequence;
        icmp_param.data_ptr=&dev->pa_mask;
        icmp_param.data_len=4;
        if (ip_options_echo(&icmp_param.replyopts, NULL, daddr, saddr, skb)==0)
                icmp_build_xmit(&icmp_param, daddr, saddr, skb->iph->tos);
#endif  
        kfree_skb(skb, FREE_READ);      
}

static void icmp_discard(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 saddr, __u32 daddr, int len)
{
        kfree_skb(skb, FREE_READ);
}

#ifdef CONFIG_IP_TRANSPARENT_PROXY
/*
 *      Check incoming icmp packets not addressed locally, to check whether
 *      they relate to a (proxying) socket on our system.
 *      Needed for transparent proxying.
 *
 *      This code is presently ugly and needs cleanup.
 *      Probably should add a chkaddr entry to ipprot to call a chk routine
 *      in udp.c or tcp.c...
 */

int icmp_chkaddr(struct sk_buff *skb)
{
        struct icmphdr *icmph=(struct icmphdr *)(skb->h.raw + skb->h.iph->ihl*4);
        struct iphdr *iph = (struct iphdr *) (icmph + 1);
        void (*handler)(struct icmphdr *icmph, struct sk_buff *skb, struct device *dev, __u32 saddr, __u32 daddr, int len) = icmp_pointers[icmph->type].handler;

        if (handler == icmp_unreach || handler == icmp_redirect) {
                struct sock *sk;

                switch (iph->protocol) {
                case IPPROTO_TCP:
                        {
                        struct tcphdr *th = (struct tcphdr *)(((unsigned char *)iph)+(iph->ihl<<2));

                        sk = get_sock(&tcp_prot, th->source, iph->daddr,
                                                th->dest, iph->saddr, 0, 0);
                        if (!sk) return 0;
                        if (sk->saddr != iph->saddr) return 0;
                        if (sk->daddr != iph->daddr) return 0;
                        if (sk->dummy_th.dest != th->dest) return 0;
                        /*
                         * This packet came from us.
                         */
                        return 1;
                        }
                case IPPROTO_UDP:
                        {
                        struct udphdr *uh = (struct udphdr *)(((unsigned char *)iph)+(iph->ihl<<2));

                        sk = get_sock(&udp_prot, uh->source, iph->daddr,
                                                uh->dest, iph->saddr, 0, 0);
                        if (!sk) return 0;
                        if (sk->saddr != iph->saddr && ip_chk_addr(iph->saddr) != IS_MYADDR)
                                return 0;
                        /*
                         * This packet may have come from us.
                         * Assume it did.
                         */
                        return 1;
                        }
                }
        }
        return 0;
}

#endif
/* 
 *      Deal with incoming ICMP packets. 
 */
 
int icmp_rcv(struct sk_buff *skb, struct device *dev, struct options *opt,
         __u32 daddr, unsigned short len,
         __u32 saddr, int redo, struct inet_protocol *protocol)
{
        struct icmphdr *icmph=(void *)skb->h.raw;
#ifdef CONFIG_IP_TRANSPARENT_PROXY
        int r;
#endif
        icmp_statistics.IcmpInMsgs++;
        
        /*
         *      Validate the packet
         */
        
        if (ip_compute_csum((unsigned char *) icmph, len)) 
        {
                /* Failed checksum! */
                icmp_statistics.IcmpInErrors++;
                printk(KERN_INFO "ICMP: failed checksum from %s!\n", in_ntoa(saddr));
                kfree_skb(skb, FREE_READ);
                return(0);
        }
        
        /*
         *      18 is the highest 'known' ICMP type. Anything else is a mystery
         *
         *      RFC 1122: 3.2.2  Unknown ICMP messages types MUST be silently discarded.
         */
         
        if(icmph->type > 18)
        {
                icmp_statistics.IcmpInErrors++;         /* Is this right - or do we ignore ? */
                kfree_skb(skb,FREE_READ);
                return(0);
        }
        
        /*
         *      Parse the ICMP message 
         */

#ifdef CONFIG_IP_TRANSPARENT_PROXY
        /*
         *      We may get non-local addresses and still want to handle them
         *      locally, due to transparent proxying.
         *      Thus, narrow down the test to what is really meant.
         */
        if (daddr!=dev->pa_addr && ((r = ip_chk_addr(daddr)) == IS_BROADCAST || r == IS_MULTICAST))
#else
        if (daddr!=dev->pa_addr && ip_chk_addr(daddr) != IS_MYADDR)
#endif
        {
                /*
                 *      RFC 1122: 3.2.2.6 An ICMP_ECHO to broadcast MAY be silently ignored (we don't as it is used
                 *      by some network mapping tools).
                 *      RFC 1122: 3.2.2.8 An ICMP_TIMESTAMP MAY be silently discarded if to broadcast/multicast.
                 */
                if (icmph->type != ICMP_ECHO) 
                {
                        icmp_statistics.IcmpInErrors++;
                        kfree_skb(skb, FREE_READ);
                        return(0);
                }
                /*
                 *      Reply the multicast/broadcast using a legal
                 *      interface - in this case the device we got
                 *      it from.
                 */
                daddr=dev->pa_addr;
        }
        
        len-=sizeof(struct icmphdr);
        (*icmp_pointers[icmph->type].input)++;
        (icmp_pointers[icmph->type].handler)(icmph,skb,skb->dev,saddr,daddr,len);
        return 0;
}

/*
 *      This table defined limits of ICMP sending rate for various ICMP messages.
 */

static struct icmp_xrlim
        xrl_unreach = { 4*HZ, 80, HZ/4 },               /* Host Unreachable */
        xrl_redirect = { 2*HZ, 10, HZ/2 },              /* Redirect */
        xrl_generic = { 3*HZ, 30, HZ/4 };               /* All other errors */

/*
 *      This table is the definition of how we handle ICMP.
 */
 
static struct icmp_control icmp_pointers[19] = {
/* ECHO REPLY (0) */
 { &icmp_statistics.IcmpOutEchoReps, &icmp_statistics.IcmpInEchoReps, icmp_discard, 0, NULL },
 { &dummy, &icmp_statistics.IcmpInErrors, icmp_discard, 1, NULL },
 { &dummy, &icmp_statistics.IcmpInErrors, icmp_discard, 1, NULL },
/* DEST UNREACH (3) */
 { &icmp_statistics.IcmpOutDestUnreachs, &icmp_statistics.IcmpInDestUnreachs, icmp_unreach, 1, &xrl_unreach },
/* SOURCE QUENCH (4) */
 { &icmp_statistics.IcmpOutSrcQuenchs, &icmp_statistics.IcmpInSrcQuenchs, icmp_unreach, 1, NULL },
/* REDIRECT (5) */
 { &icmp_statistics.IcmpOutRedirects, &icmp_statistics.IcmpInRedirects, icmp_redirect, 1, &xrl_redirect },
 { &dummy, &icmp_statistics.IcmpInErrors, icmp_discard, 1, NULL },
 { &dummy, &icmp_statistics.IcmpInErrors, icmp_discard, 1, NULL },
/* ECHO (8) */
 { &icmp_statistics.IcmpOutEchos, &icmp_statistics.IcmpInEchos, icmp_echo, 0, NULL },
 { &dummy, &icmp_statistics.IcmpInErrors, icmp_discard, 1, NULL },
 { &dummy, &icmp_statistics.IcmpInErrors, icmp_discard, 1, NULL },
/* TIME EXCEEDED (11) */
 { &icmp_statistics.IcmpOutTimeExcds, &icmp_statistics.IcmpInTimeExcds, icmp_unreach, 1, &xrl_generic },
/* PARAMETER PROBLEM (12) */
/* FIXME: RFC1122 3.2.2.5 - MUST pass PARAM_PROB messages to transport layer */
 { &icmp_statistics.IcmpOutParmProbs, &icmp_statistics.IcmpInParmProbs, icmp_discard, 1, &xrl_generic },
/* TIMESTAMP (13) */
 { &icmp_statistics.IcmpOutTimestamps, &icmp_statistics.IcmpInTimestamps, icmp_timestamp, 0, NULL },
/* TIMESTAMP REPLY (14) */
 { &icmp_statistics.IcmpOutTimestampReps, &icmp_statistics.IcmpInTimestampReps, icmp_discard, 0, NULL },
/* INFO (15) */
 { &dummy, &dummy, icmp_discard, 0, NULL },
/* INFO REPLY (16) */
 { &dummy, &dummy, icmp_discard, 0, NULL },
/* ADDR MASK (17) */
 { &icmp_statistics.IcmpOutAddrMasks, &icmp_statistics.IcmpInAddrMasks, icmp_address, 0, NULL },
/* ADDR MASK REPLY (18) */
 { &icmp_statistics.IcmpOutAddrMaskReps, &icmp_statistics.IcmpInAddrMaskReps, icmp_discard, 0, NULL }
};

void icmp_init(struct proto_ops *ops)
{
        struct sock *sk;
        int err;
        icmp_socket.type=SOCK_RAW;
        icmp_socket.ops=ops;
        if((err=ops->create(&icmp_socket, IPPROTO_ICMP))<0)
                panic("Failed to create the ICMP control socket.\n");
        sk=icmp_socket.data;
        sk->allocation=GFP_ATOMIC;
        sk->num = 256;                  /* Don't receive any data */
#ifndef CONFIG_NO_ICMP_LIMIT
        xrlim_init();
#endif
}