Re: [Yaffs] bad block management

On Wed, Aug 6, 2014 at 7:26 PM, bpqw <bpqw@micron.com> wrote:
> Hi Clarles,
> We recommended if the bitflip over threshold we just need to refresh the block but not retire it.
> So we doubt is it reasonable just according to the bitflips over
> mtd->bitflip_threshold over three times to judge the block as bad block?
>

Hello White Ding

I certainly understand where you are coming from here.

The concern this raises is that we then lose the safety net of
retiring blocks before they go bad.

What we really need is two thresholds in mtd: refresh only, refresh
and apply retiring logic.

I suppose we could use bitflip_threshold and bitflip_strength for
those, or maybe something slightly different to give more margin.
Unfortunately the threshold and strength are often set to be the same.

So I would like to use something that can be made up called, say, retire_limit.
Where
retire_limit = (refresh_threshold + ecc_strength + 1) /2.


For that to work Yaffs needs another level of ECC error.

How about this:
enum yaffs_ecc_result {
    YAFFS_ECC_RESULT_UNKNOWN,
    YAFFS_ECC_RESULT_NO_ERROR,
    YAFFS_ECC_RESULT_REFRESH,
    YAFFS_ECC_RESULT_FIXED,
    YAFFS_ECC_RESULT_UNFIXED
};

Then we can have something like, say:

if errors < bitflip_threshold --> NO_ERROR
else if errors < retire_limit --> REFRESH
else if not corrupted --> FIXED
else --> UNFIXED

So for a concrete example, let us say we have something where:
ecc_strength = 8, bitflip_threshold = 4 then we would have retire_limit = 6

So we would get the following behaviour:

* 0 to 3 errors --> NO_ERROR
* 4 to 5 errors --> REFRESH (refresh only)
* 6 to 8 errors -->FIXED (refresh and increment chunk_error_strikes)
* more than that --> UNFIXED.

Does that make sense?

Charles