Syslinux - Feb 2002 - addendum to syslinux/ tftp-hpa docs

After a few days of troubleshooting, we finally determined why we were 
having problems.  It seems as though the tftp-hpa daemon requires the 
MTU of the interface it's running on to be >= 1500.  We had the MTU set 
at 1400 (don't ask why) and the PXE client was receiving fragmented 
packets and failing.  As soon as we set it to over 1500, it worked fine.

hpa,

Could you please update your syslinux/pxelinux and/or tftp-hpa docs so 
they reflect this requirement?

Thanks,

James Martin

James S. Martin wrote:
> After a few days of troubleshooting, we finally determined why we were
> having problems.  It seems as though the tftp-hpa daemon requires the
> MTU of the interface it's running on to be >= 1500.  We had the MTU
set
> at 1400 (don't ask why) and the PXE client was receiving fragmented
> packets and failing.  As soon as we set it to over 1500, it worked fine.
> 
> hpa,
> 
> Could you please update your syslinux/pxelinux and/or tftp-hpa docs so
> they reflect this requirement?
> 

Hmm... the problem is that what you're seeing isn't the tftp-hpa daemon
doing anything wrong, what you're seeing is a bug in your PXE stack. 
I'll
add it to the "broken hardware info" page.

	-hpa

Peter,

Is the problem that a PXE client, running on Ethernet, should not assume
that the MTU is at least 1500?  Do you think that it probably doesn't
support IP segment reassembly?

Thanks,
John
> -----Original Message-----
> From: H. Peter Anvin [mailto:hpa at zytor.com]
> Sent: Thursday, February 28, 2002 4:46 PM
> To: James S. Martin
> Cc: syslinux at zytor.com
> Subject: Re: [syslinux] addendum to syslinux/ tftp-hpa docs
> 
> 
> James S. Martin wrote:
> 
> > After a few days of troubleshooting, we finally determined 
> why we were
> > having problems.  It seems as though the tftp-hpa daemon 
> requires the
> > MTU of the interface it's running on to be >= 1500.  We had 
> the MTU set
> > at 1400 (don't ask why) and the PXE client was receiving
fragmented
> > packets and failing.  As soon as we set it to over 1500, it 
> worked fine.
> > 
> > hpa,
> > 
> > Could you please update your syslinux/pxelinux and/or 
> tftp-hpa docs so
> > they reflect this requirement?
> > 
> 
> 
> Hmm... the problem is that what you're seeing isn't the 
> tftp-hpa daemon
> doing anything wrong, what you're seeing is a bug in your PXE 
> stack.  I'll
> add it to the "broken hardware info" page.
> 
> 	-hpa