Re: APIC error interrupt

[Rogier Wolff]

Alan Cox wrote:
> > Someone on this list said once that the ABit BP6 boards are known to have some
> > APIC propblems. On the hand this might well be but on the other why are 2.2.x
> > kernels not suffering from them?
> 
> Seperate two things here
> 
> 1.	APIC CS warnings
> 
> 2.	Lockups
> 
> #1 is happening in both but only 2.4test logs it. APIC cs warnings cause
> retries so they upset your real time performance and slow things a bit but
> are basically harmless

What I still don't know is how large the checksums are that are used
to detect the "transmission errors". If the checksum is only 8 bits,
it makes sense that your system can lockup without warning say once
every 256 times a checksum error is reported. So if you have a
checksum warning 3 times every hour, then it is likely that your
machine will have an undetected error once every 80 hours or so...

And can such an undetected error cause a lockup?

Alan, maybe you didn't know this yet, but there is a voltage regulator
on the board, where they put in a larger model on some of the boards.
Turns out that the larger model also requires larger capacitors to
remain stable!

Boards can be RMA-ed, if you suffer from this problem.

On the other hand, in my case my system is now reasonably stable,
after some software changes (bios, kernel). I'm running "noapic" by
the way. (And I had an odd crash this morning: Everything worked, but
just the window manager was not-working or just VERY VERY slow. killed
the window manager (no that doesn't kill my session) but that didn't
help much. Looked like MASSIVE VM problems, but running a 100M process
was reasonably quick and situation didn't improve afterwards. I
then rebooted.)

				Roger.

-- 
** [e-mail] ** http://www.BitWizard.nl/ ** [phone] **
*-- BitWizard writes Linux device drivers for any device you may have! --*
*       Common sense is the collection of                                *
******  prejudices acquired by age eighteen.   -- Albert Einstein ********