[PATCH 01/17] of: add dma-mask binding

[robherring2@gmail.com (Rob Herring)]

On 11/14/2012 12:00 AM, Jean-Christophe PLAGNIOL-VILLARD wrote:
> On 21:55 Tue 13 Nov     , Rob Herring wrote:
>> On 11/12/2012 02:52 AM, Wenyou Yang wrote:
>>> From: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
>>>
>>> This will allow each device to specify its dma-mask for this we use the
>>> coherent_dma_mask as pointer. By default the dma-mask will be set to
>>> DMA_BIT_MASK(32).
>>
>> Do you really have a use case this is not DMA_BIT_MASK(32)?
> yes as exmample on 64bit platfrom it will be 64 on x86 it's also 24, on power
> pc 40
>>
>>> The microblaze architecture hook is drop
>>>
>>> Signed-off-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
>>> Cc: grant.likely at secretlab.ca
>>> Cc: rob.herring at calxeda.com
>>> Cc: devicetree-discuss at lists.ozlabs.org
>>> ---
>>>  drivers/of/platform.c |   23 ++++++++++++++++-------
>>>  1 file changed, 16 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/drivers/of/platform.c b/drivers/of/platform.c
>>> index b80891b..31ed405 100644
>>> --- a/drivers/of/platform.c
>>> +++ b/drivers/of/platform.c
>>> @@ -130,6 +130,21 @@ void of_device_make_bus_id(struct device *dev)
>>>  	dev_set_name(dev, "%s.%d", node->name, magic - 1);
>>>  }
>>>  
>>> +static void of_get_dma_mask(struct device *dev, struct device_node *np)
>>> +{
>>> +	const __be32 *prop;
>>> +	int len;
>>> +
>>> +	prop = of_get_property(np, "dma-mask", &len);
>>
>> dma-ranges may work for this purpose.
>>
>>> +
>>> +	dev->dma_mask = &dev->coherent_dma_mask;
>>
>> I don't really know, but I suspect this is wrong.
> no this is correct was suggest by Russell
> we already do so on other ARM or SH as example
> 
> the dma expect a pointer for dma_mask but the value is the same as coherent

Okay. Then perhaps this part should be a separate patch as that is
useful on its own for 32-bit machines with no DMA address restrictions
(most modern ARM h/w).

My comment on using dma-ranges still stands though. The form is
<parent-address child-address size>. Normally, parent and child would be
the same. I think the mask would be "parent address + size - 1". The
simple case is <0 0 0> for all of 32-bit memory. I'm using 0 for full
32-bit size here as #size-cells is typically 1 and I can't imagine that
0 size would be useful.

Rob

>>
>>> +
>>> +	if (!prop)
>>> +		dev->coherent_dma_mask = DMA_BIT_MASK(32);
>>> +	else
>>> +		dev->coherent_dma_mask = of_read_number(prop, len / 4);
>>> +}
>>> +
>>>  /**
>>>   * of_device_alloc - Allocate and initialize an of_device
>>>   * @np: device node to assign to device
>>> @@ -171,10 +186,8 @@ struct platform_device *of_device_alloc(struct device_node *np,
>>>  		WARN_ON(of_irq_to_resource_table(np, res, num_irq) != num_irq);
>>>  	}
>>>  
>>> +	of_get_dma_mask(&dev->dev, np);
>>>  	dev->dev.of_node = of_node_get(np);
>>> -#if defined(CONFIG_MICROBLAZE)
>>> -	dev->dev.dma_mask = &dev->archdata.dma_mask;
>>> -#endif
>>>  	dev->dev.parent = parent;
>>>  
>>>  	if (bus_id)
>>> @@ -211,10 +224,6 @@ struct platform_device *of_platform_device_create_pdata(
>>>  	if (!dev)
>>>  		return NULL;
>>>  
>>> -#if defined(CONFIG_MICROBLAZE)
>>> -	dev->archdata.dma_mask = 0xffffffffUL;
>>> -#endif
>>> -	dev->dev.coherent_dma_mask = DMA_BIT_MASK(32);
>>>  	dev->dev.bus = &platform_bus_type;
>>>  	dev->dev.platform_data = platform_data;
>>>  
>>>