386 revival - problems
emg...@gmx.net
today, I decided to revive my girlfriend's old 386 which was out of use for
quite a while...
The system has an AMD386DX40, a Cyrix "Fast Math" 387 clone, AMI BIOS,
32K cache and currently 8MB memory (will be 20MB soon). I intend to have
it running as workstation[0] / X terminal connected to my own PC.
So I went and replaced it's cute li'l 40MB IDE drive[1] with a 540MB EIDE I
had lying around and installed a CD-ROM provisionally, so I could install
RedHat 5.2. Unfortunately, I ran into several problems:
- first, the install bailed out with a crc error. Obviously, I was
suspicious of the RAM chips, but after a few experiments, I discovered
that the install would proceed, if I switched the "Timing Parameter
Selection" in the advanced section of the CMOS setup from "Extended" to
"Normal". The built-in help describes this as "DMA timing, ROM timing(?!)
and AT bus cycles(?)", with "Extended" being "worst case" and "Normal"
being "best case". Extended is default.
- later, I discovered that this very option switches back to "Extended"
whenever I reboot the machine. I suspected the CMOS battery, especially,
as this PC was off for a very long time (I had to re-program the CMOS
when I first switched it on) - but that parameter is the only "volatile"
one (well - and the clock is very slow...), all others are fine.
- after I managed to finish the install, the machine booted fine (even
with that "Extended" setting). Strangely, Linux only recognises 6MB
instead of all 8MB (according to cat /proc/meminfo and top) - never seen
something like that happening before.
- the second odd thing is that fdisk complains about not having the
geometry information for the hard drives. I discovered this when I added
another 120MB IDE drive as hdb. Both drives are set properly in CMOS, but
fdisk comes up with 0 0 0 for cylinders, headers and sectors. Very odd...
Now, if anybody out there has any ideas as to what is going on with this
machine, I'd be very grateful to hear them, as I'd really like to get that
PC going again - if possible. I'd be happy to supply more information if
necessary.
Thanks in advance,
Thomas
[0] for sufficiently low values of "work"...
[1] monster drive - twice the height of the 540MB drive... :-) ...and twice
as loud...
--
-----------------------------------------------------------------------------
Thomas Ribbrock | "You have to live on the edge of reality -
| to make your dreams come true!"
Limerick / Ireland | http://www.bigfoot.com/~kaytan
Paul Hovnanian
>
> Hi folks,
[snip]
> - later, I discovered that this very option switches back to "Extended"
> whenever I reboot the machine. I suspected the CMOS battery, especially,
> as this PC was off for a very long time (I had to re-program the CMOS
> when I first switched it on) - but that parameter is the only "volatile"
> one (well - and the clock is very slow...), all others are fine.
Probably. But also see below.
> - after I managed to finish the install, the machine booted fine (even
> with that "Extended" setting). Strangely, Linux only recognises 6MB
> instead of all 8MB (according to cat /proc/meminfo and top) - never seen
> something like that happening before.
>
> - the second odd thing is that fdisk complains about not having the
> geometry information for the hard drives. I discovered this when I added
> another 120MB IDE drive as hdb. Both drives are set properly in CMOS, but
> fdisk comes up with 0 0 0 for cylinders, headers and sectors. Very odd...
>
> Now, if anybody out there has any ideas as to what is going on with this
> machine, I'd be very grateful to hear them, as I'd really like to get that
> PC going again - if possible. I'd be happy to supply more information if
> necessary.
With old 386s (I've got one) the BIOS more often than not lies to the kernel.
I have fould that thinks like disk geometry are best set with a boot configuration
parameter in lilo (or however you boot). My old 386 did fine with 13Mb of memory,
but almost everything else I had to specify.
--
Paul Hovnanian mailto:Pa...@Hovnanian.com
---------------------------------------------------------------
Matter cannot be created or destroyed,
nor can it be returned without a receipt.
Joe Ringer
>Hi folks,
>
>today, I decided to revive my girlfriend's old 386 which was out of use for
>quite a while...
>The system has an AMD386DX40, a Cyrix "Fast Math" 387 clone, AMI BIOS,
>32K cache and currently 8MB memory (will be 20MB soon). I intend to have
>it running as workstation[0] / X terminal connected to my own PC.
>So I went and replaced it's cute li'l 40MB IDE drive[1] with a 540MB EIDE I
>had lying around and installed a CD-ROM provisionally, so I could install
>RedHat 5.2. Unfortunately, I ran into several problems:
>
> - first, the install bailed out with a crc error. Obviously, I was
> suspicious of the RAM chips, but after a few experiments, I discovered
> that the install would proceed, if I switched the "Timing Parameter
> Selection" in the advanced section of the CMOS setup from "Extended" to
> "Normal". The built-in help describes this as "DMA timing, ROM timing(?!)
> and AT bus cycles(?)", with "Extended" being "worst case" and "Normal"
> being "best case". Extended is default.
>
> whenever I reboot the machine. I suspected the CMOS battery, especially,
> as this PC was off for a very long time (I had to re-program the CMOS
> when I first switched it on) - but that parameter is the only "volatile"
> one (well - and the clock is very slow...), all others are fine.
>
> with that "Extended" setting). Strangely, Linux only recognises 6MB
> instead of all 8MB (according to cat /proc/meminfo and top) - never seen
> something like that happening before.
Do you have bios and video shadowing off?
> - the second odd thing is that fdisk complains about not having the
> geometry information for the hard drives. I discovered this when I added
> another 120MB IDE drive as hdb. Both drives are set properly in CMOS, but
> fdisk comes up with 0 0 0 for cylinders, headers and sectors. Very odd...
Can't disk geometry be passed via a command line parameter?
>Now, if anybody out there has any ideas as to what is going on with this
>machine, I'd be very grateful to hear them, as I'd really like to get that
>PC going again - if possible. I'd be happy to supply more information if
>necessary.
--
clear skies, |http://www.erols.com/jringer3/astro1.htm
Joe |
|The internet treats censorship like
|damage and routes around it.
Geoff Short
:
: - after I managed to finish the install, the machine booted fine (even
: with that "Extended" setting). Strangely, Linux only recognises 6MB
: instead of all 8MB (according to cat /proc/meminfo and top) - never seen
: something like that happening before.
I think you'll find this always happens: meminfo reports the user memory, ie
what is left free by the kernel. Look for a line in the kernel startup
which says Memory: 6000k/8000k available or whatever.
: - the second odd thing is that fdisk complains about not having the
: geometry information for the hard drives. I discovered this when I added
: another 120MB IDE drive as hdb. Both drives are set properly in CMOS, but
: fdisk comes up with 0 0 0 for cylinders, headers and sectors. Very odd...
The bios or ide controller is obviously not quite responding to a probe
from linux. You might need to specify the hard drive parameters when you
boot - check the lilo.conf man page.
Geoff
--
----------------------------------------------------------------------------
Ever sit and watch ants? They're always busy with Geoff Short
something, never stop for a moment. I just ge...@kipper.york.ac.uk
can't identify with that kind of work ethic. http://kipper.york.ac.uk/~geoff