Issues with Ethernet corruption on xSTORM at max CPU frequency

[Greg Kogut]

I've reported this before, but I've now replicated it on two different Storm Overos (IronSTORM/FireSTORM) using two different OS images (both Sakoman console images built using mkcard.sh).  

When I set the mpurate to the maximum 1Ghz I start to get periodic Ethernet errors.  During an SSH session with the Overo, this shows up most often as a "Corrupted MAC on input" message and a dropped session.  Sometimes the SSH session just drops with no message.     New connections are accepted, but the drops occur frequently (several per minute).

Setting the mpurate down to 800Mhz (or lower) corrects the problem.

Any ideas here, on how to solve this at the full 1Ghz?

Regards,

Greg

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
gumstix-users mailing list
gumstix-users@...
https://lists.sourceforge.net/lists/listinfo/gumstix-users

[sakoman]

On Wed, Apr 25, 2012 at 2:26 PM, Greg Kogut <gkogut@...> wrote:

> I've reported this before, but I've now replicated it on two different Storm
> Overos (IronSTORM/FireSTORM) using two different OS images (both Sakoman
> console images built using mkcard.sh).
>
> When I set the mpurate to the maximum 1Ghz I start to get periodic Ethernet
> errors.  During an SSH session with the Overo, this shows up most often as a
> "Corrupted MAC on input" message and a dropped session.  Sometimes the SSH
> session just drops with no message.     New connections are accepted, but
> the drops occur frequently (several per minute).
>
> Setting the mpurate down to 800Mhz (or lower) corrects the problem.
>
> Any ideas here, on how to solve this at the full 1Ghz?

Running the 37XX processors at 1Ghz requires the kernel to support
both smart reflex and adaptive body bias.

The current upstream kernel currently only has smart reflex support so
1Ghz operation is not reliable across the spectrum of hot/cold 37XX
lots.

We are all anxiously awaiting TI's submission of working adaptive body
bias (ABB) patches.

If you search the linux-omap list you will find an ABB patch
submission from last September, but this submission was not accepted
and in my testing does not work.

Steve

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
gumstix-users mailing list
gumstix-users@...
https://lists.sourceforge.net/lists/listinfo/gumstix-users

[MiLo-14]

I patched the kernel to remove the 1GHz setting completely. That solved a lot
of unexplained crashes and problems. Of 3 boards, ordered in a single batch,
one would run for 10 minutes at 100% cpu load at 1GHz (and then crash),
while another didn't even boot properly at 1GHz. All boards run fine at
800MHz, so until TI submits a working patch, I suggest sticking to 800MHz.

--
View this message in context: http://gumstix.8.n6.nabble.com/Issues-with-Ethernet-corruption-on-xSTORM-at-max-CPU-frequency-tp4918105p4933902.html
Sent from the Gumstix mailing list archive at Nabble.com.

------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
gumstix-users mailing list
gumstix-users@...
https://lists.sourceforge.net/lists/listinfo/gumstix-users