Have you tested with only one drive connected at a time, to the new IDE cable, on the other IDE controller? (If your mobo has two). Can you move the bad drive to another computer and test it again there? Can you induce the errors while booted off the Knoppix CD? (maybe dd if=/dev/hda of=/dev/null ?)

I use xmbmon/mbmon and display the results in GKrellM. You could also try lm_sensors, but I've always had trouble setting it up correctly.

SOLUTION: I've sort of fixed it. I've moved all the usual partitions to this WD drive (now hda) and have done . So now the dma doesn't get turned off at boot and I don't seem to have any of those errors that I had about LBAsector or whatever - of course this is just hiding the problem by not accessing that Seagate disk.

I do now have a stack of udev errors, but I think they're probably not related (I updated udev to -current yesterday).

As for using knoppix I'm not sure I didn't notice anything spurious ... I don't have time to go back and check at the moment; ditto with the IDE controllers (2 on this mobo).

Thanks.

PS, Speck, I don't think my AMD Athlon 1.1G has the requisite sensors to get the voltage info in the first place, but I'll check into this a bit further.

Not wanting to open this all up again really, but was looking back at a few things:

http://lkml.org/lkml/2006/7/31/157

and other posts that I've found strongly suggest a mistake in the partition table is causing a read beyond the end of the disk and that the kernel retrys this lots of times before it decides it can't read it. As a precaution, it seems, DMA is then turned off.

Still not got a solution but considering going hex and altering the partition table to show the correct last block. Perhaps this is a BIOS error??!