>>> On 2/11/2011 at 01:26 AM, Anthony Liguori <anth...@codemonkey.ws> wrote: > On 02/10/2011 11:54 PM, Bruce Rogers wrote: >> With recent gpxe eepro100 drivers, short packets are rejected, >> so ensure the minimum ethernet packet size. >> >> Signed-off-by: Bruce Rogers<brog...@novell.com> >> > > This doesn't make much sense. I think this is more likely a case where > we're incorrectly calculating packet size. Michael fixed an issue > similar to this in e1000 relating to FCR if I recall correctly. Maybe > eepro100 has the same problem? >
I just took a clue from the other, similar code in slirp.c (arp_input()), where it ensured a minimum of 64 bytes. I now see that that is the wrong approach, and the way this is handled is that the nic emulation adds padding to short packets in its receive handler. I'll submit a patch with that approach instead. Thanks for the review. Bruce
