Text 3234, 75 rader
Skriven 2010-11-28 03:46:26 av Doug Bissett
Ärende: Re: Strange trap in OS2KRNL on boot - solved ? Maybe ...
================================================================
mail-complaints-to="news_AT_tioat_DOT_net"
From: "Doug Bissett" <dougb007!SPAM@Use-Author-Supplied-Address.invalid>

On Sat, 27 Nov 2010 19:58:26 UTC, Dave Yeo <dave.r.yeo@gmail.com> 
wrote:

> Doug Bissett wrote:
> > On Sat, 27 Nov 2010 11:57:17 UTC, Lars Erdmann<lars.erdmann@arcor.de>
> > wrote:
> >
> >> Doug Bissett wrote:
> >>> On Fri, 26 Nov 2010 07:28:07 UTC, Lars Erdmann<lars.erdmann@arcor.de>
> >>> wrote:
> >>>
> >>>> It is so cheap, I think it does
> >>>> not even have the ECC feature.
> >>>
> >>> 99% of the memory that exists, in modern hardware, doesn't have the
> >>> ECC feature. Memory, today, is reliable enough that it is not needed,
> >>> and the only reason to actually use ECC memory is for "mission
> >>> critical" applications (life support systems etc.). The really cheap
> >>> memory would not have the parity bit, but memory is cheap enough now,
> >>> that I doubt if you could find such a thing.
> >> Ok, did not know that.
> >>
> >>> On the other hand, I am somewhat surprised that a machine would serve
> >>> up a parity error, because the BIOS startup, even without the extended
> >>> memory test, is supposed to write a pattern to the memory, before it
> >>> boots.
> >>>
> >> Yes, but if "Quick Power On Self Test" is enabled it will only do that
> >> for the DOS relevant memory region of<= 1MB.
> >> If you disable, it will do that for the whole memory (supposedly).
> >>
> >>
> >> Lars
> >
> > I am pretty sure that the BIOS is supposed to write good data through
> > all of memory, even if you do the quick test. The reason for doing
> > that, is to prevent exactly what you are describing. There may be many
> > reasons why the BIOS may not clean all memory, or, you may have a bad
> > memory chip in there somewhere. Then it is possible that the
> > manufacturer was attempting to speed up boot, and dropped that from
> > the BIOS (poor decision, if they did).
> >
> > I am also under the impression that OS/2 itself initializes all of
> > memory during boot, so you should never see that sort of thing anyway.
> >
> > I think I would find a memory test program, and see what it has to say
> > about the problem.
> >
> 
> I tried a bad SIMM a while ago. The BIOS accepted it fine. Both OS/2 and 
> Win2k just ignored it and Linux crashed. Memtest showed one stuck bit. I 
> think I was just using the quick memory test so the BIOS can miss bad 
> memory.
> Dave

BIOS, using the extended memory test, can miss bad memory as well. 
Windows doesn't care if it has bad memory, and will usually work, 
unless the bad memory happens to be in some core part of windows. Most
of the time, OS/2 will trap, if it finds a bad memory bit, but that 
depends on where it is. If it is in a part of memory that rarely gets 
used, OS/2 might miss it. 

MEMTEST is really the only good test of memory.

-- 
From the eComStation of Doug Bissett
dougb007 at telus dot net
(Please make the obvious changes, to e-mail me)


--- Internet Rex 2.31
 * Origin: The gateway at Omicron Theta (1:261/20.999)