[En-Nut-Discussion] TCP/IP problems with httpd

[Bernd Walter]

On Wed, Jan 12, 2011 at 05:42:05AM -0800, uprinz wrote:
> 
> I might have to correct myself...
> 
> We checked a bunch of AT91SAM7X boards over here now and identified boards
> that seem to have a defective PHY/EMAC. At least I guess that it must be the
> cause for the non functioning ping/DHCP as four  boards work without a flew.
> 
> My colleague is now reverting some minor modifications to at91_emac and
> at91_efc to check if it works out of the box. 
> 
> In the meantime:
> ping -A [ip] returned
> 1209150 packets transmitted, 1209088 received, 0% packet loss, time 560094ms
> rtt min/avg/max/mdev = 0.376/0.418/137.244/0.323 ms, pipe 14, ipg/ewma
> 0.463/0.418 ms
> 
> There is still a small problem that sometimes DHCP doesn't come up, but none
> of the boards are crashing anymore under heavy traffic.

I wonder how a PHY problem can crash a board unless it produces glitches
on the power lines.
Either a correct packet reaches the MAC or not and in the incorrect
case the MAC should filter it out - and if this fails the IP checksum
should do.
Or is it that the PHY itself (it has a link related state machine)
or it's oscillator crashes?
In that case the MAC might even be unable to send packets because it
gets it's transmit clock from the PHY and probably Nut/OS hangs
in any transmit routine.
I can even imagine that a bufer full hang wouldn't happen under normal
conditions and if can't resolve such a condition it even can't got on
after the PHY has recovered from the faulty condition.

-- 
B.Walter <bernd at bwct.de> http://www.bwct.de
Modbus/TCP Ethernet I/O Baugruppen, ARM basierte FreeBSD Rechner uvm.