From: kf8nh@kf8nh.wariat.org (Brandon S. Allbery) Subject: Re: lpd Date: Fri, 01 Jan 1993 01:42:17 EST
Bob.Martin@f34.n343.z1.fidonet.org (Bob Martin) writes:
> Since I new the diskettes were good and the disk drive is ok (works fine
> under dos and windows) I started looking at the CMOS settings. After
> trial and error I found that if I added 1 wait state the hangups went
> away.
>
> Any body else seen this before or can confirm this, may just be my
> hardware.
I haven't seen it with Linux or anything else on my 386DX-33, but I *have*
seen something like this. It's what made me buy the 386DX-33.
My previous machine was an XT with an Intel Inboard-386 (386DX-16). It
started out with 1MB RAM. When I decided to upgrade it to be almost a real
computer :-) so I could run Borland C++ and multiple DESQview windows (the
Inboard requires special activation to enable 386 capabilities, and no
386-mode OS supports this activation, so *ix-like OSes were clearly out of
the question) I added a 2MB daughterboard and replaced my ancient MFM
controller and 20MB hard drive with a 50MB IDE setup.
Under DOS it worked. Under DESQview disk reads were often utterly corrupted.
The "utter corruption" turned out to be SuperStor, which I had been running
on the 20MB drive and installed on the 50MB drive. When I removed it, I
found that the computer was occasionally catching a byte sent by the disk
drive *twice*.
I tried again, first with an SCSI disk and an ST01 controller, then with an
MFM drive and a newer MFM controller (my old controller was so old that it
couldn't habdle a drive larger than 20MB). The SCSI was the most reliable,
but even it duplicated characters occasionally. Forcing wait states on hard
drive reads (this can be done with the Inboard configuration/386 mode
activation TSR) made it even more reliable, but it still had the problem.
By then it was obvious that it was a hardware problem between a fast hard
drive, a fast CPU, and a very slow (XT standard) bus, and bought a real 386.
I suspect hardware wait states would have worked even better, but there was
no way to get them on an XT....
Again, this only happened under DESQview. Had I been capable of booting
something like Linux, it would probably have happened there as well.
So: it is quite possible for hardware timing to be such that things work
under BIOS/DOS, but not in a multitasking environment.
++Brandon
He's BAAAACK! Brandon S. Allbery NOTE NEW ADDRESS!!! kf8nh@kf8nh.wariat.org