FreeBSD 4-10 install, RAM parity errors that don't seem to happen inLinux.

[Charles Swiger]

On Jun 14, 2004, at 1:38 PM, Derrick MacPherson wrote:
> Derrick MacPherson <dmacpherson <at> mainframe.ca> writes:
>> I am trying to move as much of our servers as I can to FreeBSD, and
>> there's a few boxes that they have here that the RAM is a about 2 mm
>> high and requires the case (1U machines) to press on the RAM when
>> closed. These machines run RH Linux for months without a problem, yet 
>> 3
>> out of 4 I just pulled are giving RAM parity problems during FreeBSD
>> instalation.
>>
>> Does FreeBSD not allow/recover from those types of errors the same way
>> Linux does? Any solutions?
>
> I posted this last week and hadn't seen a repsonse to it, is there 
> someone that
> cares to take a poke at this?

Well, if you are using ECC, normally they will correct all single-bit 
errors and notice but not correct larger multibit errors.  This is done 
at a fairly low level in the hardware (which is why the BIOS typically 
controls the use of ECC), and is not something that is supposed to vary 
depending on which OS you are using.

That being said, Linux and FreeBSD might be using different portions of 
memory which hit different RAM chips, and so you see the errors for one 
and not the other, but if you've got failing RAM, your systems are not 
going to be stable under heavy load regardless of what OS you use.  I 
would expect you to see signs of problems under Linux, too, but 
consider running memtest86 for a day or so and see what you see:

http://www.memtest86.com/

-- 
-Chuck