Re: [Yaffs] Failure to write to an erased flash due to yaff…

Top Page
Attachments:
Message as email
+ (text/plain)
Delete this message
Reply to this message
Author: Eugene Surovegin
Date:  
To: Max Rayskiy
CC: 'yaffs@lists.aleph1.co.uk'
Subject: Re: [Yaffs] Failure to write to an erased flash due to yaffs_CheckChunkErased failure
On Thu, Jul 01, 2010 at 10:50:36AM -0700, Max Rayskiy wrote:
> Eugene,
>
> Thanks for the tip. It is a pity I have to verify content because our hardware cannot help me here, but it may be not a big deal since I have to do it only in case of an uncorrectable error.
>
> One question though - are you saying yaffs may decide to retire a block based on _corrected_ errors reported by mtd? I know nand layer (I am on 2.6.26.8) returns -EUCLEAN on corrected error, but I see that yaffs treats it as any other error. Our hardware has pretty strong ecc block - up to 6 bit errors per 512 byte chunk. I would hate to waste that capability.


Yes, correctable errors will cause eventual "retirements". I think
it's one of many MTD API deficiencies that you have to play games like
this.

> So where is the threshold where I need to report correctable errors? Maybe I should not do it at all?


That's the answer you'll have to determine yourself :). In your case I
think 1 or 2 bit errors per chunk are safe to "hide" from upper
levels.

--
Eugene