mtd_oobtest fails with GPMI-NAND

[Vikram Narayanan]

On 1/30/2013 7:57 AM, Huang Shijie wrote:
> 于 2013年01月30日 00:28, Vikram Narayanan 写道:
>> On 1/29/2013 8:06 AM, Huang Shijie wrote:
>>>> May I know in what way the driver is restricted for using with other
>>>> boards?
>>> The devicetree. Please check the dts file in arch/arm/boot/dts/.
>>> If there is no dts file for your board, the gpmi will not works very
>>> well for your board.
>>> for example, there is arch/arm/boot/dts/imx6q-arm2.dts for the MX6Q-ARM2
>>> board.
>>
>> I've a valid dts file for my board which is good enough to probe the
>> Toshiba NAND on-board and mount a UBIFS filesystem.
> could you tell the dts file for your board?
> are you sure the pinmux settings are right?

I'm using the same pinumx settings mentioned in the file
"arch/arm/boot/dts/imx6q.dtsi". Hope that is right.

>>
>>>> You're right and we knew it earlier as the Kernel detects the OOB size
>>>> as 128 bytes. So, some remaining bytes become unused at the end of
>>>> each page. i.e., (224-128). Instead of supporting ecc strength as ECC9
>>>> we end up using ECC8. Should this be a cause for the -74 error?
>>> no. this is not the cause. even you get the wrong geometry of the nand
>>> chip. the gpmi still can works.
>>>
>>> I think the root cause is the devicetree issue as above.
>>
>> Something else is causing the issue. Can you give some other pointers
>> which can potentially cause this -74 error while mounting?
> the -74 error is the EBADMSG, it's caused by the uncorrectable ECC failure.
> There are many reasons can cause this error:
>     [1] the pinmux setting is not right, the gpmi does not works well.

As said above, my pinmux settings are right. GPMI detects my NAND, does 
read/write/erase/mount. At times I'm seeing this -74 issue. Sometimes 
this error goes off in the next reboot.

>     [2] the wrong setting of BCH. for example, the page size is 8K, you
> set 4k to BCH.
>         so you can set the correct nand geometry parameters for your
> nand chip firstly, and then do the debug.

 From the driver I understood that the geometry are determined according 
to the data from the read_id bytes. mtd layer detects my NAND as 4K 
page-sized. So, this shouldn't be a problem I guess.

>     [3] timings.

Should I tweak something specific according to the NAND device that I 
have? Can you give more hints please?

Thanks,
Vikram