> Shane Wegner wrote on Wed, 17 Nov 1999 16:44:33 PST
> >
> >I am indeed.
> >
>
> The problem seems to occur primarily under heavy load for me
> and has to do with a FIFO status flag set prematurely on the
> controller and data corrupted as a result.
>
> I plopped in an older 2940UW, rebuilt the raid device and the
> problem was gone. A fix for the U2W was proposed by Justin
> Gibbs, and I have a patch from Tor Egge. Doug Ledford posted a
> more recent patch in the kernel group (search for FIFOEMP in
> dejanews).
>
> Sorry to say, I'd be suspicious of all data on /dev/md0 at
> this point.
>
I just got over this bug too. I am now running 2.2.13 + the
aforementioned patch, plus one or too patches selected from the 2.2.14pre
series just for paranoia (ext2 race fix by Andrea Arcangeli). I really,
really hope my problem goes away now. It hasn't recurred in a week of
uptime.
My data was NOT (for the record) corrupted. It seemed to be an INPUT
error as far as I could tell, but then again most of the files on the
device were somewhat transitory in nature.
If you want the patch, I have it in my mailbox and can forward it. Or
also search for messages from the last few weeks with 'LOTS OF BAD STUFF'
in the title.
David
--
/==============================\
| David Mansfield |
| [EMAIL PROTECTED] |
\==============================/
