On Tue, 07 Mar 2000, you wrote:
> Civileme wrote:
> > We are obviously dealing with something very subtle at the
> > edge--the bleeding edge it would seem. What is different about
> > Seagate drives? I tried Fujitsu, Maxtor, IBM, and Quantum in the
> > same position and none showed the problem. It seems to take
> > three conditions
> >
> > 1. Large Seagate Drive (8.4G or bigger)
> > 2. VIA MVP3 or MVP4 chipset Super7 motherboard with IDT, AMD,
> > INtel, or Cyrix processor
> > 3. Linux-Mandrake 6.0 6.1(6.5MacMillan) or 7.0
>
>
> Civileme, please have a look at this (if you didn't already):
> http://kt.linuxcare.com/kt20000214_54.epl (the 2nd item)
>
> While it is not exactly the same problem, it has also to do
> with "lost interrupt" and it seems to be a pure hardware problem.
>
Yep, flaky disk drive timing. The strangeness was that I was able to make
something happen positively by swapping it away then back to /dev/hda
I never had a drive on the same IDE channel with the offensive Seagate.
It is interesting that the signal reflection mentioned there is WD drives withj
PII and PIII processors while I am observing this with Seagate and Super7
processors. Also, it appears, the problem relates to timing requirements
becoming more strict as we progress upward in the processors we compile for.
FreeBSD and Win98 showed NO errors and are 386 compiled. L-M 6.0, 6.1, and 7.0
are 586 compiled and obviously do not permit the sloppiness built into the
Seagate, just as kernels built for the 686 and 586 gaVE EXTREMELY nasty
performance with WDC and Maxtor drives on the same IDE channel for PII and PIII
processors.
And yes, it appears UDMA/66 is not ready for prime time unless you have a 386
compiled kernel
Thanks for the heads-up. I'll be following that one, but it appears time for a
convocation of Disk drive manufacturers with topics interoperability and
quality.
And I won't be putting any Maxtor master/WD slave combos up at all. Looks like
that is downright dangerous to your data and temperament.
Civileme
Now on the Seagate I am seeing the lost interrupt on the OTHER channel and I am
having trouble with the initial command set, but I am willing to bet it is an
unwelcome signal reflection.
Thus far, the IBM branded drives seem to be above this .
> -- > Jean-Louis
Debert [EMAIL PROTECTED] > 74 Annemasse France > old Linux fan
