James Harper wrote: >>James Harper wrote: >> >>>(sorry if I double posted... I only just re-joined the list and it >>>simultaneously rejected this posting and replied with the [VLAN] in > > the > >>>subject indicating it sent it to the list) >>> >>>I am getting many (>10%) runt frames on a cisco switch on the port > > that > >>>is connected to a linux router running with vlans. Despite the > > reported > >>>errors, there are no observed problems with the network. >> >>I think this is an old cisco bug. > > > Not sure I agree with you there... two other machines trunking vlans > plugged into the same switch don't have the same problems. The only > difference there is that they are using e100 cards, not via-rhine > > >>W/regard to the 60 v/s 64 minimum pkt size, the bytes on the wire are > > 64, > >>but that is because the NIC will add a 4 byte CRC to the end of the >>60-byte payload. > > > 60 bytes is the #define in if_ether.h, and yes it states that it is > 'sans FCS', so when it says 60 it really means 64. > > What I was getting at though, is that a normal Ethernet minimum frame > size is 64 (60 + FCS). Should the minimum frame size be incremented by 4 > when there are vlan headers present? The 60/64 limit is a low-level ethernet thing, and VLANs really shouldn't matter at that level. I think the cisco bug is that it does think you need +4 for the minimum size for VLAN packets. Back (oh, 7 years ago??) when I read the VLAN spec, I don't remember any requirement to bump the minimum packet size. > Something is different about frames sent by the e100 vs the via-rhine, > to verify what I'm thinking, is there a way to check? (can tcpdump tell > me what I want to know, or is the padding invisible to that layer?) You could certainly try tcpdump or tethereal...I'm not sure whether it will show the padding or not... Ben > > Thanks > > James > > > _______________________________________________ > Vlan mailing list > Vlan@xxxxxxxxxxxxxxx > http://www.lanforge.com/mailman/listinfo/vlan > -- Ben Greear <greearb@xxxxxxxxxxxxxxx> Candela Technologies Inc http://www.candelatech.com
