imx6 NAND chip selects

[Hugo Villeneuve]

On 05/08/2014 11:37 AM, Huang Shijie wrote:
> On Wed, May 07, 2014 at 12:31:42PM -0400, Hugo Villeneuve wrote:
>> Hi,
>> I am trying to connect a NAND flash device (Micron MT29F8G08) to a
>> Wandboard Dual.
>>
>> Here is the pin mapping:
>>
>> i.MX6             NAND flash
>> ----------------------------
>> NANDF_CS0    -->  CE#
>> NANDF_RB0    -->  R/B#
>> NANDF_WP_B   -->  WP#
>> NANDF_CLE    -->  CLE
>> NANDF_ALE    -->  ALE
>> NANDF_WE     -->  WE#
>> NANDF_RE     -->  RE#
>> NANDF_D[0-7] -->  I/O[0-7]
>>
>> This NAND flash has a single chip-select.
>>
>> I'm testing it using Linux kernel 3.15.0-rc3.
>>
>> In the wandboard device tree file, I have added all the above gpmi
>> pins as GPMI pins.
>>
>> I am still booting from the SD card, so at this point I am only
>> trying to detect the NAND device. Unfortunately, when the kernel
>> boots, the GPMI driver fails to register (error -19).
>>
>> If I monitor the NANDF_CS0 line, I can see that it is pulsed low,
>> but I am not seing any activity on the NANDF_WE line.
>>
>> However, If I monitor the NANDF_CS1 line, I can see that it is
>> pulsed low, and during the time it is low the NANDF_WE line is also
>> going low.
>>
>> I then made sure to reconfigure the unused pins NANDF_CS[1-3] as
>> GPIOs (ALT5 function) in my device tree file. I have confirmed with
>> kernel debug traces that the iomux register for NANDF_CS1 have the
>> value 0x05, which means ALT5 function (GPIO). But I am still seeing
>> the same behavior as before on the oscilloscope.
>>
>> So how can I configure the GPMI driver to only use NANDF_CS0 and not
>> NANDF_CS1?
>>
>> And is it normal that NANDF_CS1 is still used by the GPMI driver
>> when configured as GPIO?
> It is not normal at all.
>
> Please disable the GPIO pinctrl, and only enable the pinctrl for gpmi
> and try again.
>
> I am sure that you did not configure the PADs correctly.
>
> thanks
> Huang Shijie

Hi Huang,
knowing that it was not normal to have NANDF_CS1 toggling made me think 
to look more closely at the hardware level, and I found that the 8 data 
lines are unfortunately inverted on our PCB.

I have temporarily patched the Linux GPMI driver to invert LSB and MSBs 
for read and writes, and voilà, now my NAND flash is correctly detected!

Sorry for the noise.

Hugo.