Re: [PATCH 3/3] mtd: spi-nand: add devicetree binding

[Ezequiel Garcia <ezequiel.garcia@imgtec.com>]

On 01/11/2015 10:57 PM, Peter Pan 潘栋 (peterpandong) wrote:
>> On 01/08/2015 02:04 AM, Peter Pan 潘栋 (peterpandong) wrote:
>>>>> This commit adds the devicetree binding document that specifies the
>>>>> spi nand devices support.
>>>>>
>>>>> Signed-off-by: Peter Pan <peterpandong@micron.com>
>>>>> ---
>>>>>  Documentation/devicetree/bindings/mtd/spi-nand.txt | 22
>>>> ++++++++++++++++++++++
>>>>>  1 file changed, 22 insertions(+)
>>>>>  create mode 100644 Documentation/devicetree/bindings/mtd/spi-
>>>> nand.txt
>>>>>
>>>>> diff --git a/Documentation/devicetree/bindings/mtd/spi-nand.txt
>>>> b/Documentation/devicetree/bindings/mtd/spi-nand.txt
>>>>> new file mode 100644
>>>>> index 0000000..9dd3efd
>>>>> --- /dev/null
>>>>> +++ b/Documentation/devicetree/bindings/mtd/spi-nand.txt
>>>>> @@ -0,0 +1,22 @@
>>>>> +* NAND driver for MT29F, GD5F and similar SPI NAND flash chips
>>>>> +
>>>>> +Required properties:
>>>>> +- #address-cells, #size-cells : Must be present if the device has
>>>> sub-nodes
>>>>> +  representing partitions.
>>>>> +- compatible : Should be the manufacturer and the name of the chip.
>>>> Bear in mind
>>>>>
>>>>
>>>> Unless I'm mistaken, we don't need the chip ID here, as SPI NAND
>> allows
>>>> to autodetect the device. Any reason why we can't just use a generic
>>>> compatible "spi-nand" here?
>>>> --
>>>> Ezequiel
>>>
>>> In fact, I don't know how to autodetect the SPI NAND device. Micron
>> device and
>>> Gigadevice device have different read ID functions. The Chip ID here
>> is used to
>>> determine which function to use.
>>>
>>
>> Isn't the difference between the Read ID very minor? One of the vendor
>> needs a 2-byte ID read, and the other one needs a 3-byte ID read.
>>
>> So you can just try with 2-byte, and if that fails (no vendor ID is
>> found on the first byte), you can try with the 3-byte command.
>>
>> It's not the most elegant solution, but it's not super awful either.
>> --
>> Ezequiel
> 
> I agree with your idea about try different ways to read ID. It is a better way
> to detect the chip if it can work properly.
> 
> The difference between the Read ID is that some chips need to send a dummy byte
> or address after opcode then read 2 byte, some chips needn't send any data after
> opcode but read 3 byte.
> 
> I'm concerned maybe chip will get into unknown or vendor specified state if we
> send the wrong sequence.
> 

Have you seen this happpening? Or this a potential/future problem?

For the latter, we can always fix it when the issue appears.
-- 
Ezequiel