Re: RED State Exception on E3500

[Julian Coleman <jdc%coris.org.uk@localhost>]

Hi,

> I'll try swapping out boards and see if I can get it stable.

I installed Solaris 10 and ran build.sh there, to see if that would: a) run;
and b) find any hardware errors.  That showed up an occasional ECC error on
one of the DIMM's, so I replaced that DIMM.  After that, I managed to run
build.sh -j 12 to completion a couple of times (took about 4.5 hours each and
needed bash as the shell and a couple of fixes for a Solaris host), as well
as some fairly large java code.

However, when I tried NetBSD again, I saw:

  panic: fpusave_lwp ipi didn't
  cpu2: data fault: pc=11f7b30 addr=ffffffffffff8000
  kernel trap 30: data access exception

and then a Makefile with 8 bytes of nulls inserted, and third time, the same
RED State Exception as before.

The CPU's in this are part number X2580A (501-6009, 501-6619, 501-6622 and
501-6624), which have in their description:

  "Mirrored Data SRAM Mirrored Tag SRAM"

Do we need to do something special to handle these?

Thanks,

J

-- 
  My other computer also runs NetBSD    /        Sailing at Newbiggin
        http://www.netbsd.org/        /   http://www.newbigginsailingclub.org/