home bbs files messages ]

Forums before death by AOL, social media and spammers... "We can't have nice things"

   comp.protocols.tcp-ip      TCP and IP network protocols.      14,669 messages   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]

   Message 14,023 of 14,669   
   rahul.dev.agg@gmail.com to All   
   Re: icmp echo to a host with smaller mtu   
   12 Jun 13 10:38:10   
   
   > > Thanks for your reply. I am just doing it out of curiosity. I am   
   >   
   > > trying to understand how path mtu discovery works.   
   >   
   >   
   >   
   > > Can we have two hosts with different mtu connected to the same link   
   >   
   > > ? Ethernet supports both 1500 and jumbo frames (9000). So, can we   
   >   
   > > have one host with mtu = 1500 and other with mtu = 9000 on the same   
   >   
   > > link.   
   >   
   >   
   >   
   > > If yes, what would be the pmtu for the path between these nodes ?   
   >   
   > > Can node A find the mtu of node B on the same link ?   
   >   
   >   
   >   
   > The starting point is this, in pseudo IEEEspeak (which I've probably   
   >   
   > botched) with a smattering of IETF style:   
   >   
   >   
   >   
   >     All stations in the same broadcast domain MUST have the same MTU.   
   >   
   >   
   >   
   > Ethernet has no way to communicate frame size between peers.  If a   
   >   
   > frame larger than the station is prepared to receive arrives, that   
   >   
   > frame will be dropped.   
   >   
   >   
   >   
   > Path MTU is up at the IP layer and uses ICMP messages to communicate   
   >   
   > MTU between hosts.  The PathMTU logic is run when a router looks to   
   >   
   > forward the IP datagram - when it goes to send it.  That means it must   
   >   
   > have received it in the first place.  For IP to "receive" the datagram   
   >   
   > it must first be received at the layer below it - in this case   
   >   
   > Ethernet.   
   >   
   >   
   >   
   > However, if the interface on which it was going to receive the   
   >   
   > datagram has an MTU/framesize smaller than the size of the datagram   
   >   
   > sent, the datagram won't be "received" by the IP layer so it cannot be   
   >   
   > resent, so the Path MTU logic cannot trigger.   
   >   
   >   
   >   
   > Thus the reason why all stations (hosts, systems, what you will) in   
   >   
   > the broadcast domain (everything joined at layer 2 eg ethernet) MUST   
   >   
   > have the same MTU.   
   >   
   >   
   >   
   > Now, if you have a router (a device making forwarding decisions at   
   >   
   > layer three - eg IP) it will have a foot in two different broadcast   
   >   
   > domains.  So long as its feet are the correct size for each broadcast   
   >   
   > domain, you can have different MTUs on each side of the router.  Then,   
   >   
   > PathMTU discovery will be able to do its thing.  But when all there   
   >   
   > are between the two hosts are switches (a device making forwarding   
   >   
   > decisions at layer 2 - eg Ethernet) there is no way for PathMTU   
   >   
   > discovery to become invovled in the first place.   
   >   
   >   
   Thanks a lot Rick for such an elaborate reply. This is really useful.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]


(c) 1994,  bbs@darkrealms.ca