Fix for 20% loss of ICMP Packets

[mike%brl-vgr at sri-unix.UUCP]

From:      Michael John Muuss <mike at brl-vgr>

Index:	sys/netinet/ip_icmp.c 4.2BSD

Description:
	When using a "ping" program to send ICMP_ECHO packets onto
	the network, a steady 20% of them are lost, regardless of
	interface, including the software loopback driver.	
	The problem is that ip_icmp is incorrectly looking in
	the sequence number and ID field of the ICMP header,
	as if this was an "error advice" packet much like ICMP_REDIRECT.
Repeat-By:
	Ping the loopback device.  Ping program availible on request.
	(Pitty that one was not included with 4.2)
Fix:
	A little code has to be rearranged in ip_icmp.c, around line 200+.

******* ORIGINAL CODE *******
		goto free;		/* not yet implemented: ignore */
#endif

	case ICMP_REDIRECT:
	case ICMP_ECHOREPLY:
	case ICMP_TSTAMPREPLY:
	case ICMP_IREQREPLY:
		if (icmplen < ICMP_ADVLENMIN || icmplen < ICMP_ADVLEN(icp)) {
			icmpstat.icps_badlen++;
			goto free;
		}
		/*
		 * Short circuit routing redirects to force
		 * immediate change in the kernel's routing
		 * tables.  The message is also handed to anyone
		 * listening on a raw socket (e.g. the routing
		 * daemon for use in updating it's tables).
		 */
		if (icp->icmp_type == ICMP_REDIRECT) {
			icmpsrc.sin_addr = icp->icmp_ip.ip_dst;
			icmpdst.sin_addr = icp->icmp_gwaddr;
			rtredirect((struct sockaddr *)&icmpsrc,
			  (struct sockaddr *)&icmpdst);
		}
		icmpsrc.sin_addr = ip->ip_src;
		icmpdst.sin_addr = ip->ip_dst;
		raw_input(dtom(icp), &icmproto, (struct sockaddr *)&icmpsrc,
		  (struct sockaddr *)&icmpdst);
		return;

	default:
		goto free;

********** Correct code **********
		goto free;		/* not yet implemented: ignore */
#endif

	case ICMP_R