[1147] in linux-scsi channel archive
Re: medium error, what next?
daemon@ATHENA.MIT.EDU (Gerard Roudier)
Sat Dec 21 02:54:21 1996
Date: Sat, 21 Dec 1996 09:52:19 +0000 (GMT)
From: Gerard Roudier <groudier@club-internet.fr>
To: "Stephen R. van den Berg" <srb@cuci.nl>, linux-scsi@vger.rutgers.edu,
ncr53c810@colorado.edu
In-Reply-To: <Pine.LNX.3.91.961221011639.213A-100000@localhost>
Stefen,
Linux does not support raw io. So it seems quasi-impossible to find the
number of the wrong sector, even with a dichotomy method.
It is perhaps possible to read or write through the cache, but I donnot
know how.
I donnot know if dd does some read-ahead, but I know linux does
read-ahead for raw-like devices.
The sector number you got in error message is very probably the number of
the first sector of a faulty IO chunk.
So, DONNOT TRY TO WRITE WITH AWRE or switch temporarily to another O/S that
support real raw io.
Under Linux, the only way that seems to have chance to work is to set ARRE
AWRE and a large value for retry count and to cross fingers.
Gerard.
On Sat, 21 Dec 1996, Gerard Roudier wrote:
>
> On Fri, 20 Dec 1996, Gerard Roudier wrote:
>
> >
> > Use scsi-config as suggested by Leonard to change the Read-Write Recovery
> > Page.
> > Set ARRE to 1 as suggested by Stefan.
> > If EER bit change is supported by your drive set it to 1
>
> I just received my post. Obviously EER must be set to 0 (not to 1).
> Sorry.
>
> > If DCR bit ................................. set it to 0
>
> Etc..
>
> Gerard.
>
>
