Re: [PATCH v3 1/6] memory: omap-gpmc: Store handle to GPMC dev

[Roger Quadros <rogerq@ti.com>]

Franklin,

On 10/03/16 15:40, Franklin S Cooper Jr. wrote:
> 
> 
> On 03/10/2016 06:40 AM, Roger Quadros wrote:
>> Franklin,
>>
>> On 10/03/16 06:07, Franklin S Cooper Jr wrote:
>>> The dma channel information is located within the GPMC node. The NAND
>>> driver requires a handle to the GPMC's dev to properly parse the DMA
>>> properties. Therefore, store a handle to the dev so it can be referenced
>>> within the NAND driver.
>>>
>>> Signed-off-by: Franklin S Cooper Jr <fcooper@ti.com>
>>> ---
>>>  drivers/memory/omap-gpmc.c                   | 1 +
>>>  include/linux/platform_data/mtd-nand-omap2.h | 2 ++
>>>  2 files changed, 3 insertions(+)
>>>
>>> diff --git a/drivers/memory/omap-gpmc.c b/drivers/memory/omap-gpmc.c
>>> index 6515dfc..2932d13 100644
>>> --- a/drivers/memory/omap-gpmc.c
>>> +++ b/drivers/memory/omap-gpmc.c
>>> @@ -1796,6 +1796,7 @@ static int gpmc_probe_nand_child(struct platform_device *pdev,
>>>  
>>>  	gpmc_nand_data->cs = val;
>>>  	gpmc_nand_data->of_node = child;
>>> +	gpmc_nand_data->gpmc_dev = &pdev->dev;
>>>  
>>>  	/* Detect availability of ELM module */
>>>  	gpmc_nand_data->elm_of_node = of_parse_phandle(child, "ti,elm-id", 0);
>>> diff --git a/include/linux/platform_data/mtd-nand-omap2.h b/include/linux/platform_data/mtd-nand-omap2.h
>>> index 090bbab..534b984 100644
>>> --- a/include/linux/platform_data/mtd-nand-omap2.h
>>> +++ b/include/linux/platform_data/mtd-nand-omap2.h
>>> @@ -80,5 +80,7 @@ struct omap_nand_platform_data {
>>>  	/* for passing the partitions */
>>>  	struct device_node	*of_node;
>>>  	struct device_node	*elm_of_node;
>>> +
>>> +	struct device		*gpmc_dev;
>>>  };
>>>  #endif
>>>
>> Why do you need this? Can't we just use dev->parent in the omap2-nand driver?
> 
> The omap2-nand platform_device is created by
> platform_device_alloc. The created platform_device dev
> parent isn't by default set to the GPMC dev. I didn't think
> this was the right approach to manually set the parent
> property. However, taking a look at other usages of
> platform_device_alloc it seems this is actually pretty
> common. If your ok with this then I can go that route.

After my DT cleanup we no longer use platform_dev_alloc for DT cases.
As Nand node is child of GPMC node, omap2 nand's parent is guaranteed to be the
GPMC device.

But for legacy boot we still use gpmc_nand_init() which calls platform_device_add()
without setting pdev->parent and so it will set the parent to platform bus.

Maybe we could patch gpmc_nand_init() to set NAND's parent to GPMC device?

cheers,
-roger