[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference 7.286::fddi

Title:	FDDI - The Next Generation

Moderator:	NETCAD::STEFANI

Created:	Thu Apr 27 1989
Last Modified:	Thu Jun 05 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	2259
Total number of notes:	8590

1171.0. "FYI Solaris and IP Frag" by CTOAVX::BEAULIEU () Fri Dec 03 1993 13:33

    Hi All,
    
    Just wanted to post a warning that SUN has problems with the Ip Frag
    bit on ethernet with Solaris OS 2.2. This can potentially cause a
    DECbridge 620 to stop forwarding packets to the SUN device if IP frag is
    turned on in the bridge. I also hear it causes Kalpana's Ethernet switch
    to go belly up!
    
    SUN knows about the problem and is working on a fix.
    
    Later, Mike Beaulieu

T.R	Title	User	Personal Name	Date	Lines
1171.1		KONING::KONING	Paul Koning, B-16504	`Fri Dec 03 1993 14:31`	13
	Huh? Would you explain a bit more what you mean? The fragmentation done by the bridge is the same as a router would do. Are you saying that Sun doesn't know how to talk to routers? As for Kalpana, I don't understand this either. PLEASE be explicit. We're currently building more products that include the same sort of function. I'm very interested in any information about problems, but I HAVE to have details in order to do anything. paul
1171.2	IP Datagram Discard No Fragment??	TKTVFS::IDO	Naoki Ido, CSC/TOKYO, EWB, DTN 680-2456	`Mon Dec 06 1993 00:04`	9
	Re0: I saw similar problem that long FDDI packets were discarded at DECbridge because SUN set a "do not fragment" bit in the IP header. This causes that DECbridge discards some FDDI packets larger than ethernet max size and increments a one of the FDDI line counter called "IP Datagram Discarded No Fragment" Is the problem you are seeing this? Naoki
1171.3	Cisco can handle this	COMICS::REYNOLDS	Mad Dogs and Englishmen	`Tue Jan 11 1994 12:10`	15
	I also have a customer experiencing this problem. However, he reports that the Solaris system works through a Cisco bridge router because the router sends back an ICMP message reporting it's inability to fragment due to the 'don't fragment' bit being set. It looks like the Solaris responds to this by unsetting the bit. This ICMP message is a requirement of IP gateways under rfc 1009. But since the Decbridge doesn't claim to be a gateway, I guess it's reasonable that this is not implemented. John Reynolds, UK CSC Comms Support.
1171.4	Decbridge isn't a gateway but...	COMICS::REYNOLDS	Mad Dogs and Englishmen	`Tue Jan 11 1994 12:23`	11
	re last reply - I suppose I should say we wouldn't necessarily follow the rfc! Paul, can you confirm regarding the return of ICMPs - contrary to .0 , SUN have told my customer that their setting the bit was deliberate and not a bug -they expect gateways to tell them if they can't fragment, John.
1171.5		KONING::KONING	Paul Koning, B-16504	`Tue Jan 11 1994 13:46`	11
	DECbridge 900-MX will send the ICMP message. The interesting question is why the DF bit is being set. A possible reason is for MTU discovery (per RFC 1191). Since the 620 doesn't send the ICMP message, MTU discovery done that way won't work. The most obvious fix is to set the MTU on the Sun system manually to 1500. Again, the 900 does support RFC 1191 and things should work better with that device. paul
1171.6	ok	COMICS::REYNOLDS	Mad Dogs and Englishmen	`Tue Jan 11 1994 15:15`	11
	Paul, thanks for your fast response. I think this customer will have problems since they are using UDP so therefore cant negotiate an MTU. But at least I have a definitive answer on the 620 operation (and a possible sales opportunity for the 900MX!) thanks again, John.
1171.7		KONING::KONING	Paul Koning, B-16504	`Tue Jan 11 1994 17:32`	4
	I've heard that in general MTU is settable on a fairly global basis. That's not ideally what you'd want to do, but it would be a workaround for the moment. paul