NAND ECC on a per-partition basis

NAND ECC on a per-partition basis

[Peter Barada]

I'm working with a Micron POP part for an omap3530, and I need to write
pages in the boot block using the builtin Hamming ECC(as the bootrom
expects it), but the rest of the blocks I want to use the NAND's builtin
4-bit ECC engine.

Is there any way to specify the ECC method on a per-partition basis?

If not, what would be the best way to go about modifying the MTD code to
allow such?


-- 
Peter Barada <peter.barada@gmail.com>

Re: NAND ECC on a per-partition basis

[Matthieu CASTET]

Peter Barada a écrit :
> I'm working with a Micron POP part for an omap3530, and I need to write
> pages in the boot block using the builtin Hamming ECC(as the bootrom
> expects it), but the rest of the blocks I want to use the NAND's builtin
> 4-bit ECC engine.
> 
> Is there any way to specify the ECC method on a per-partition basis?
>
No, but you can write in raw mode and compute the ecc in userspace.

  Matthieu

Re: NAND ECC on a per-partition basis

[Florian Fainelli]

On Tuesday 17 August 2010 09:31:29 Matthieu CASTET wrote:
> Peter Barada a écrit :
> > I'm working with a Micron POP part for an omap3530, and I need to write
> > pages in the boot block using the builtin Hamming ECC(as the bootrom
> > expects it), but the rest of the blocks I want to use the NAND's builtin
> > 4-bit ECC engine.
> > 
> > Is there any way to specify the ECC method on a per-partition basis?
> 
> No, but you can write in raw mode and compute the ecc in userspace.

Yes, this is a good solution. However, Linux is very likely to report bad 
blocks in your bootloader partition, due to the ECC algorithm being different.

At some point we might want to be able to be able to define the ECC layout on a 
per-partition basis anyway. I do not say this is easy, especially as we might 
be switching between software/hardware ECC layouts from one block to another.
--
Florian