I did today some research and in short it looks like "Remember of 3xR
Rule": Retry, Reboot, Reinstall (which is especially true when we are
talking about 9x, but I also had success applying it to NT :-). Below is
long version of forensics in my poor english for archival purposes.
> Does this system use the default IDE driver esdi_560.pdr? In device
> manger, the entries for Primary/Secondary IDE,
> Properties->Driver->Driver File Details should list "esdi_560.pdr" and
> "ios.vxd". The version numbers should be 4.90.3000
I checked that and it looks like this:
First have correct version and identical to one on install CD (I assume
you mistyped 506 as 560 because I haven't found such file anywhere).
Second was a surprise. There was no ios.vxd file on hard disk (but I
actually found it on CD !) Vmm32.vxd have no version information. I have
extracted copy from CD and that one also have no version. But It was
twice as small, so I though I should try it, and so I did. But with it,
system repeatedly rebooted itself after two seconds of loading.
I have started looking for something else and found in "add/remove
programs" Intel chipset and video drivers uninstall option (a year ago
this system was running on Intel motherboard - until it died, and was
moved to AMD/SIS; but I thought that Windows was then reinstalled from
scratch). I kindly used it, then moved smartvsd into proper place, then
rebooted and then first success: no blue screen. So, as you suggested,
third party driver was interfering with something. But so much for joy,
as smartctl says that can't read smart status. One more deeper look, and
I have deleted intel* files from iosubsys. After reboot still no blue
screens, no smart and additionally system started to behave bad. It was
becoming unresponsible because of spool32 appearing process.
And then I thought that this is proper moment for last step: "In-Place
Reinstallation". After 20 minutes system was up again and so far
everything works including smartmontools. Note: vmm32.vxd is still in
"big" version (so probably there are two or more versions on CD).
My conclusion: primary/secondary channel devices are not proper (or not
only) places to look for conflicts.