suggestion to extend life of ATA IDE flash drive

[Charles Manning]

On Saturday 15 November 2003 21:43, David Woodhouse wrote:
> On Thu, 2003-11-13 at 15:12 -0800, Andre Sodermans wrote:
> > I did read the note By David on the very bottom of the
> > mtd homepage about compactflash IDE drives, but I have
> > been hearing different stories about this - somebody
> > mentioned that there exists an mtd-emulation. Is this
> > really true?
>
> Yes, â?? such an emulation. It's the 'blkmtd' device, which uses any block
> device (not just CF) as backing store for a faked MTD device.
>
> You can use JFFS2 on that. It's not wonderfully efficient, since every
> time we write a node to JFFS2 we end up rewriting a whole 512-byte
> 'sector' (or more). We should use the code we already _have_ for NAND
> flash, which batches writes to the medium (but handles fsync() etc.
> properly).
>
> Left as an exercise for the reader...

This surely does not get around the issue that the most vulnerable part of a 
CF system is the logical to physical mappings etc in the CF card. Thus, you 
might gain some wear levelling (maybe), but I don't think you're going to 
improve the robustness.

No magic bullet.