On Mon, Sep 25, 2006 at 08:47:19AM -0700, Brian Buhrow wrote:
> 	The rewriting of bad blocks on IDE disks in the 3ware firmware must
> have been an additional feature which wasn't present in the cards I used

For the 9550 it's explicitely stated in the docs, from what I remember.
It can also do scrubbing.

> because it definitely wasn't working when I was using such cards.  When
> disks had bad blocks on them, the card would simply label them as failed
> disks.  Worse, when trying to rebuild such arrays, the disks would stil
> fail, even though it should have been writing over the bad blocks.

Maybe you didn't have any spare block left ? If the firmware does bad
block rewriting, it will do so silently, until the drive runs out of
spare blocks.

> 	If the feature is there, it should work in raid mode or in jbod mode,
> I should think. 

No. The firmware has to know which data to write. In raid mode, if it
gets an error on read it can use the parity disk to rebuild the sector and
rewrite it. In jbod mode the firmware doens't have any way to rebuild
the failed sector.

> If not, you can work around it in jbod mode, I believe, by
> writing to the logically inaccessible block and then using raidframe to
> rebuild your array.

Yes, just rebuilding will do it. But with arrays of several disks,
you can have bad blocks on differents disks you didn't notice before
rebuilding (e.g. blocks in unused parts of the filesystem, or belonging
to rarely-used files). In such a case, with raidframe you loose.

-- 
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
     NetBSD: 26 ans d'experience feront toujours la difference
--