On Sat, Sep 22, 2001 at 11:21:11PM +1200, Dave Sainty wrote:
> Well of course it might not be drive damage.  The operation that is
> failing is a read operation, it may be that the drive has written
> garbage onto the platter - and that could be driver related.

Hum, the driver only sends command, and I'm not aware of IDE commands designed
to cause this :)
If some command or command sequence cause this, then it's a bug in the
drive's firmware.

> 
> Indeed, the other day my server hung, and on a hard reset one of the
> disks refused to fully revive until a power-cycle (it probed, but
> wouldn't otherwise talk it seemed).  Which was surprising and new, and
> makes me wonder whether some change has started issuing "funny"
> commands to drives, causing them to malfunction in unusual ways (like
> failing to complete sector writes, leaving them unreadable).

I've seen this too. The drive didn't even honnor properly the
reset signal (which should reset the firmware), so I suspect it's the
IDE interface or firmware processor which was hung. This repeated several
times.
After some investigation I found the IDE cable was damaged, I remplaced it
and since then the machine is stable.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--