Subject: Re: followup to ATA flash newfs problem (WAS: newfs problem: "cg 0: bad magic number")
To: <>
From: David Laight <david@l8s.co.uk>
List: tech-kern
Date: 07/12/2002 00:30:50
On Thu, Jul 11, 2002 at 11:20:22PM +0200, Wojciech Puchar wrote:
> >
> > I don't know about other CF manufacturers, but without wear-level balancing,
> > their cards would die pretty soon when used with msdosfs, for example.
> 
> yes - even faster than with ffs.
> 

Not necessarily!  I read an old document from AMD that said that they
'guaranteed' 10 years retention on the sectors that had been written 10^6
(or whatever the quated limit is) times.  It then pointed out that you
typically don't need that much retention on the sectors you are writing
frequently.

Why cards 'wear out' is another matter entirely, I doubt that
electron 'leakage' due to wear of the insulation layer caused
by excessive quantum tunnelling of electrons would show the
catastrophic errors that are seen.
So is it software????

OTOH if the filesystem issued 'sector erase' commands to the CF
when a sector was no longer required, then the CF firmware would
have a better chance at doing writes without having to do an
expensive erase cycle.

(FYIW I've seen a CF card so corrupt that the FW must have got
it badly wrong!)

	David

-- 
David Laight: david@l8s.co.uk