icmp.c

GNU Hurd 源代码

/*
 *	NET3:	Implementation of the ICMP protocol layer. 
 *	
 *		Alan Cox, <[e-mail]>
 *
 *	Version: $Id: icmp.c,v 1.52.2.4 1999/11/16 02:28:40 davem Exp $
 *
 *	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.
 *		Thomas Quinot	:	ICMP Dest Unreach codes up to 15 are
 *					valid (RFC 1812).
 *		Andi Kleen	:	Check all packet lengths properly
 *					and moved all kfree_skb() up to
 *					icmp_rcv.
 *		Andi Kleen	:	Move the rate limit bookkeeping
 *					into the dest entry and use a token
 *					bucket filter (thanks to ANK). Make
 *					the rates sysctl configurable.
 *		Yu Tianli	:	Fixed two ugly bugs in icmp_send
 *					- IP option length was accounted wrongly
 *					- ICMP header length was not accounted at all.
 *              Tristan Greaves :       Added sysctl option to ignore bogus broadcast
 *                                      responses from broken routers.
 *
 * To Fix:
 *
 *	- Should use skb_pull() instead of all the manual checking.
 *	  This would also greatly simply some upper layer error handlers. --AK
 *
 * 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. (Configurable with a sysctl.)
 *   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/raw.h>
#include <net/snmp.h>
#include <linux/skbuff.h>
#include <net/sock.h>
#include <linux/errno.h>
#include <linux/timer.h>
#include <linux/init.h>
#include <asm/system.h>
#include <asm/uaccess.h>
#include <net/checksum.h>

#ifdef CONFIG_IP_MASQUERADE
#include <net/ip_masq.h>
#endif

#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	*/
  { EMSGSIZE,		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		*/
  { ENETUNREACH,	0 },	/*	ICMP_NET_UNR_TOS	*/
  { EHOSTUNREACH,	0 },	/*	ICMP_HOST_UNR_TOS	*/
  { EHOSTUNREACH,	1 },	/*	ICMP_PKT_FILTERED	*/
  { EHOSTUNREACH,	1 },	/*	ICMP_PREC_VIOLATION	*/
  { EHOSTUNREACH,	1 }	/*	ICMP_PREC_CUTOFF	*/
};

/* Control parameters for ECHO relies. */
int sysctl_icmp_echo_ignore_all = 0;
int sysctl_icmp_echo_ignore_broadcasts = 0;

/* Control parameter - ignore bogus broadcast responses? */
int sysctl_icmp_ignore_bogus_error_responses =0;

extern int sysctl_ip_always_defrag;

/*
 *	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, int len);
	short	error;		/* This ICMP is classed as an error message */
	int *timeout; /* Rate limit */
};

static struct icmp_control icmp_pointers[NR_ICMP_TYPES+1];

/*
 *	Build xmit assembly blocks
 */

struct icmp_bxm
{
	void *data_ptr;
	int data_len;
	struct icmphdr icmph;
	unsigned long csum;
	struct ip_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 inode icmp_inode;
struct socket *icmp_socket=&icmp_inode.u.socket_i;

/*
 *	Send an ICMP frame.
 */

/*
 *	Check transmit rate limitation for given message.
 *	The rate information is held in the destination cache now.
 *	This function is generic and could be used for other purposes
 *	too. It uses a Token bucket filter as suggested by Alexey Kuznetsov.
 *
 *	Note that the same dst_entry fields are modified by functions in 
 *	route.c too, but these work for packet destinations while xrlim_allow
 *	works for icmp destinations. This means the rate limiting information
 *	for one "ip object" is shared.
 *
 *	Note that the same dst_entry fields are modified by functions in 
 *	route.c too, but these work for packet destinations while xrlim_allow
 *	works for icmp destinations. This means the rate limiting information
 *	for one "ip object" is shared - and these ICMPs are twice limited:
 *	by source and by destination.
 *
 *	RFC 1812: 4.3.2.8 SHOULD be able to limit error message rate
 *			  SHOULD allow setting of rate limits 
 *