ATA_DMA errors (and fs corruption!)

[Bob Bishop]

At 12:12 20/06/2005, Tony Byrne wrote:
>Hello Bob,
>
>
> >>can be hardware reasons for timeouts such as a dying disk or cable,
> >>but I think we've eliminated these in our case. [etc]
>
>BB> Don't ignore the possibility of failing controller hardware. We had
>BB> comparable mysterious problems on a client system, causing a lot of
>BB> head-scratching. Eventually the failure went hard and we had to 
>replace the
>BB> motherboard.
>
>I hear ya!  However, moving back to an older kernel changes the
>severity of the problem from a timeout every 2 to three minutes during
>heavy activity to about 4 or 5 in a 24 hour period.  That doesn't
>sound like hardware to me.

It didn't to me either. Note the use of 'mysterious' :-)
I'd eliminated drives and cables, and then did it all over again when the 
failure went hard, leaving the controller (or something else on the mobo). 
With a new mobo all the annoying timeouts which I'd put down to driver 
misbehaviour just went away.

--
Bob Bishop		    +44 (0)118 940 1243
rb at gid.co.uk		fax +44 (0)118 940 1295