[vortex] Re: [eepro100] MTU more then 1500

Donald Becker becker@scyld.com
Thu Dec 5 13:30:33 2002

On Thu, 5 Dec 2002, Ben Greear wrote:

> > The eepro100 patch just removes the length limit.  That opens up huge
> > holes for buffer overflows, denial of service and crashing the machine.
> > 
> > Go ahead and apply it to your own machine, but please never distribute
> > the patches or the resulting drivers.
> > 
> > [[ Yes, this upsets me: irresponsible patches like these make it into
> > the kernel all of the time.  Especially with 100 people saying "it works
> > for me".  If changing that single bit in the configuration table was a
> > good idea, don't you think that I would have done it?  ]]
> It upsets me that I hear how evil this patch is over and over, but no
> one, especially the people that Should Quite Understand,
> ever offers a better one.

Broken is broken -- if it cannot be done safely, it shouldn't be done.
This is just like replacing a fuse with a penny, and then saying you had
to do it because I refused make a fuse that wouldn't burn out.

Microsoft is notorious for making programs that assume they will only be
receiving well-formed objects from programs they control. 
"Yes, it crashes and passes viruses, but now your email can sing to
you!  And tap out Morse code on your inkjet.  And overwrite your boot

The original question was about arbitrary >1500 byte packets.  For the
case Ben cares about, 1504 byte VLAN packets, there is a solution.  The
newer eepro100 chips do support VLANs, but we cannot get the current
documentation from Intel.  The solution is to get the documentation and
errata list, and make the likely-trivial driver changes.

Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Scyld Beowulf cluster system
Annapolis MD 21403			410-990-9993