Re: JFFS2 ignore ECC bytes in cleanmarker

[Charles Manning <manningc2@actrix.gen.nz>]

On Thursday 20 October 2005 01:00, Vitaly Wool wrote:
> Hi,
>
> I do agree that the OOB data returned should be in the same format as
> OOB data provided. The layout-based model for handling NAND reads/writes
> is capable of fixing that, however, I didn't implement that in my
> layout-based patch due to the fact that I wanted to make it compatible
> with the current drivers.

Sounds good. Fix one thing at a time.

In the mean time I (or maybe someone else) will see if that proposal patch can 
be made to work properly. If so, I think we can get useful semantics pretty 
quickly too (without impacting on your work).

I wonder whether fixing things the other way around makes your life easier for 
what you're doing now (ie. if you are not boxed in by the current oobsel 
stuff)?

>
> So, I suggest the following way to go:
>
> 1. After linux-mtd is merged into the kernel, commit layout-based patch
> and fix the problems found (if any :)).
> 2. Agree on the format of OOB data.
> 3. Implement what was agreed upon.
> 4. Get rid of redundant structures (I hope that redundant will become
> eccbytes and maybe even oobfree).
Most definitely progress. We've had about four or five iterations of these and 
keeping everyone on the same page all the time is tricky.
> 5. Update the MTD userland utilities accordingly.
>
> Does that make sense?

Makes good sense to me.

-- Charles