[PATCH 1/2] [MTD] [JFFS2] MLC NAND support

[Amit Kumar Sharma]

Thanks Artem for pre information , your information I hope 
will be useful for us and we will consider your points 
during UBIFS with OneNAND-FlexOneNAND.
----- Original Message ----- 
From: "Artem Bityutskiy" <dedekind at infradead.org>
To: "Amit Kumar Sharma" <amitsharma.9 at samsung.com>
Cc: "Jason Liu" <liu.h.jason at gmail.com>; "apgmoorthy" 
<moorthy.apg at samsung.com>; "Kyungmin Park" 
<kmpark at infradead.org>; <David.Woodhouse at intel.com>; 
<linux-mtd at lists.infradead.org>
Sent: Wednesday, October 01, 2008 3:22 PM
Subject: Re: [PATCH 1/2] [MTD] [JFFS2] MLC NAND support


>
> On Wed, 2008-10-01 at 15:14 +0530, Amit Kumar Sharma 
> wrote:
>> no we have started working for UBIFS , very soon we will
>> start working for UBIFS , let u know results.
>
> Note, JFFS2 is not really suitable for MLC because it has
> random wear-leveling algorithm, and it may erase one 
> eraseblock
> thousands of times, while erasing the other eraseblock 
> only hundreds of
> times.
>
> In UBI you may set the WL threshold to control this. Also, 
> JFFS2
> does not really help with read-disturb problem, but this 
> is fixable.
>
> UBI refreshes data when it meets a bit-flip.
>
> However, UBIFS is not really ready to handle this paired 
> page problem.
> I mean, in MLC, if you interrupt power during write 
> operation, this
> may result in corruptions in one of the previous NAND 
> page. The only
> simple way to fix this is to make UBIFS assume that min. 
> I/O unit size
> is not one NAND page, but several pages, so that both 
> paired NAND pages
> were in one min. I/O unit.
>
> -- 
> Best regards,
> Artem Bityutskiy (???????? ?????)
>
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
>