2001-02-17 12:28:32

by Igor Yanover

[permalink] [raw]
Subject: More on IO-APIC trouble

Recently I came across two more things, that are possibly related to
IO-APIC problems:
1)http://xfree86.org/pipermail/xpert/2001January/004751.html
Someone with SMP that has problem with interrupt delivery (stuck
interrupt). Only in SMP mode and this is not NE2000 related.
2)http://developer.intel.com/software/idap/media/pdf/copy.pdf ( Page 8
footer)
It turns out, that there's an errata in early Pentium III revisions,
that could corrupt data written to IO-APIC. ( Only if SSE write is
followed by an APIC write)
Igor


2001-02-19 14:50:55

by Maciej W. Rozycki

[permalink] [raw]
Subject: Re: More on IO-APIC trouble

On Sat, 17 Feb 2001, Igor Yanover wrote:

> Recently I came across two more things, that are possibly related to
> IO-APIC problems:
> 1)http://xfree86.org/pipermail/xpert/2001January/004751.html
> Someone with SMP that has problem with interrupt delivery (stuck
> interrupt). Only in SMP mode and this is not NE2000 related.

I've already talked to the guy. Some software reprograms the 8254 timer
to a weird mode (one-shot, IIRC; I have full details if anyone wants to
work on it). I suspect XFree86 to be at fault.

> 2)http://developer.intel.com/software/idap/media/pdf/copy.pdf ( Page 8
> footer)
> It turns out, that there's an errata in early Pentium III revisions,
> that could corrupt data written to IO-APIC. ( Only if SSE write is
> followed by an APIC write)

We don't do SSE writes before APIC writes (they must mean the local APIC
-- I can't check the doc at the moment).

--
+ Maciej W. Rozycki, Technical University of Gdansk, Poland +
+--------------------------------------------------------------+
+ e-mail: [email protected], PGP key available +