oh, I didnt know that 640cmd is PCI related. I noticed the
same hda errors on my board ( irq timeout { BadSeek } or similar ...),
an Asus SP3, which has an SiS chipset. The kernel-config says
that the CMD640bugfix is SiS related, so ... :-)
I can reliably reproduce the bug. I have an SCSI-CDROM + aha2940, but
the root-filesystem lives on hda1. There is no other device on the SCSI
bus present. When I type:
dd if=/dev/scd0 of=/dev/null
on one VC, then switch to a different VC, log in, and do some work on
the IDE-harddisk, the system will become unusable after a few seconds,
saying "irq timeout { hda: Complete Seek Error }" ...
I'm not at home yet, when I am, I will report the chip-numbers printed
on the SiS chips.
/herp