NAND ECC problem w/ EZSDK 5.05 & DM8148

Joel,

No guarantees on the below, I am not an expert here.

Here is an answer from the developers again:

"

Q:

In that formula:

http://processors.wiki.ti.com/index.php/Raw_NAND_ECC#Is_it_possible_to_use_any_ECC_algorithm_for_any_NAND.3F


How many bytes should be "B" for Hamming [1 bit]?

Is this use case OK: ECC Hamming [1 bit] + JFFS2,  if the NAND supports the 1 bit ECC and it has enough OOB area?
That use case is not officially verified, I guess?

A:

B = 3 for Hamming 1-bit ECC. (as in given equation)
Its not recommended to use 1-bit ECC, multi bit-flips are commonly seen in some scenarios.
 - Its highly recommended to upgrade to UBIFS+BCH8, as ECC correction capability is also good with BCH8, and UBIFS has implicit power-failure tolerances.
 - UBIFS+BCH8 works seamlessly in most production environments.

"

***************

Thank you.

BR

Vladimir

Hi Vladimir,

Sorry, I was not clear.  I am seeing these failures using the EZSDK 5.05.02 default ECC: BCH8.   UBIFS/JFFS2 do not come in to play here, as this is a test on the MTD device, and no filesystem is in use.  

From what I see so far, it appears that BCH8 ECC for NAND is 'broken' for me in EZSDK 5.05.02.  I don't know the root cause or the details, but what I do know is that EZSDK 5.05.02 uses PSP 4.04.00.01, but since the EZSDK was released there has been a PSP 4.04.00.02 release.  I back-ported all the NAND-related changes from PSP 4.04.00.02 to EZSDK 5.05.02, and it seems to solve all my ECC/NAND issues - at least it is much better.

I'd still appreciate any comments from anyone out there using EZSDK 5.05.02 with a 2K page NAND with BCH8 ECC.  Does it work for you?

Thanks,
Joel

Joel,

Well, I know that there are some issues in PSP04.04.00.01 from that forum thread:

Booting from NAND causes UBI error in EZSDK 5.04.00.11 using BCH8

http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/716/t/211477.aspx

I understand that the NAND tests do not involve particular fs, the commits discussed there seem to me not UBIFS specific.

AFAIK, EZSDK5.05.02.00 is validated only with PSP04.04.00.01 and not with PSP04.04.00.02, but just like you, I would have tried to apply (only) NAND related patches from the arago PSP git repository. I have not seen the NAND test issue before, don't know whether it is related to the other forum thread, but I am glad that applying the patches works for you. Generally, I think you could always read through the arago PSP recent NAND commits, checking what's new, especially when trying to resolve something. But test even more carefully your system, after applying changes. This would be outside EZSDK release.

Thank you.

BR
Vladimir

Joel,

Are you still looking for a solution? Can you share more logs of the issue that you are facing?

All my problems appear to be fixed in yet-to-be-released patches from beyond PSP 4.04.00.02.   If anyone else is experiencing NAND problems,  please check out my other thread at:

http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/716/p/269235/945846.aspx#945846

Thanks,

Joel