[PATCH 10/12] mtd: nand: davinci: don't set timings if AEMIF is used

[Sekhar Nori]

On Friday 22 November 2013 11:08 PM, Santosh Shilimkar wrote:
> On Thursday 21 November 2013 12:07 PM, Nori, Sekhar wrote:
>> On 11/14/2013 8:06 PM, Santosh Shilimkar wrote:
>>> On Thursday 14 November 2013 05:53 AM, Nori, Sekhar wrote:
>>>> On Wednesday 13 November 2013 07:44 PM, Santosh Shilimkar wrote:
>>>>> On Wednesday 13 November 2013 12:02 AM, Sekhar Nori wrote:
>>>>>> On Monday 11 November 2013 10:40 PM, Khoronzhuk, Ivan wrote:
>>>>>>> If Davinci AEMIF is used we don't need to set timings and bus width.
>>>>>>> It is done by AEMIF driver (drivers/memory/davinci-aemfi.c).
>>>>>>>
>>>>>>> Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk at ti.com>
>>>>>>> ---
>>>>>>>  drivers/mtd/nand/davinci_nand.c |   22 +++++++++++++++-------
>>>>>>>  1 file changed, 15 insertions(+), 7 deletions(-)
>>>>>>>
>>>>>>> diff --git a/drivers/mtd/nand/davinci_nand.c b/drivers/mtd/nand/davinci_nand.c
>>>>>>> index 4705214..879e915 100644
>>>>>>> --- a/drivers/mtd/nand/davinci_nand.c
>>>>>>> +++ b/drivers/mtd/nand/davinci_nand.c
>>>>>>> @@ -742,27 +742,35 @@ static int __init nand_davinci_probe(struct platform_device *pdev)
>>>>>>>                 goto err_clk_enable;
>>>>>>>         }
>>>>>>>
>>>>>>> +#if !IS_ENABLED(CONFIG_TI_DAVINCI_AEMIF)
>>>>>>
>>>>>> This is a hack! Just because AEMIF driver is enabled, it does not
>>>>>> guarantee that the timings have been setup by it. Instead of configuring
>>>>>> timings in two drivers, why not just convert everyone over to use the
>>>>>> new driver. Dont worry about breaking old platforms, I will help test
>>>>>> and ack them.
>>>>>>
>>>>> How about you take a stab at and convert the DaVinci code over to make
>>>>> use of new driver. We are trying to re-use as much as possible from the
>>>>> common drivers and also making an option so that the DaVinci arch can
>>>>> move over to these drivers if they want to.
>>>>
>>>> Sure I could.
>>>>
>>> Thanks
>>>
>>>> Ivan,
>>>>
>>>> The AEMIF driver does not use platfrom_data currently. Is that something
>>>> you can add or you want me to take a stab at that as well?
>>>>
>>> The AEMIF new driver is device tree only and thats the direction. So the
>>> better thing would be to convert Davinci to DT and then make use of this
>>> new driver. Thats how most of the new drivers used on arm socs are
>>> moving. Adding platform data to new driver is a step in backward
>>> direction so I would want to avoid that.
>>
>> Yes, that would be the ideal way. But the reality is that there is close
>> to zero chance of all the DaVinci platforms using AEMIF ever getting
>> converted to DT. This means that we will never be able to get rid of
>> this piece of code.
>>
> Well that means some of those platforms support will get deprecated
> soon considering other basic frameworks like irq, timer wheel etc
> is assuming the DT direction. But its your call because if you
> want those platforms to live, you need to convert them to DT.
>  
>> I do agree platform data is not the direction you want to take on ARM
>> but at the same time its not really a deprecated mechanism as far as the
>> broader kernel is concerned.
>>
>> So, can we consider adding platform data mechanism to the AEMIF driver?
>>
> Sorry but answer is "NO". If the above hack is really bothering you,
> we can actually rip out the above code in question and move it
> to mach-davinci. Infact thats the best direction to make the
> mtd nandn driver independent of any platform quirks(dt, pdata etc)
> 
> Ivan already has a patch to do that and he is going to post
> that patch. With that patch, we won't need $subject patch.
> 

Okay I will wait for the patch.

Thanks,
Sekhar