A more complete log about my disk access errors

[Kirk Wood]

On Sun, Jun 03, 2001 at 12:14:09AM +0200, Victor Tsaran wrote:
> OK, if it turns out that I was fooled by Windows during all this time,
> do you know of any parameter which I could use to tell Linux to fall to
> PIO mode right away?

One way to keep it from ever going into UDMA mode is to remove the support
in the kernel. If you don't compile it in, it won't kick in. (Yes, it can
work on the obvious.) Regardless of the settings, the standard macroslop
and intel have pushed is that the drive starts in PIO mode at boot and
switches when the 32 bit drivers send the signal.

The better cables could be a large factor. I would recomend going with the
shortest possible and if you only use one drive, get a cable without the
connector for the second drive. The second conductor is there to help keep
the good signal in the cable and the bad signal out of the cable. 

When looking at the performance issues, the reality is that you will only
see a morginal increase in speed. While UDMA raises the speed of data
transfer across the wire considerably, this will only happen if the data
requested is in the hardware cache of the drive (and all drives have
this). If there is a miss, or sustained transfer the rate will max out
arround 7 MB per second. (This lower rate is the rate of transfer to and
from the medium.) This is why RAID can increas speeds so dramatically. It
will read the data from multiple drives at once.

=======
Kirk Wood
Cpt.Kirk at 1tree.net

Nothing is hard if you know the answer or are used to doing it.