[PATCH 4/4] dmaengine: edma: Add support for DMA filter mapping to slave devices

[Peter Ujfalusi]

On 12/03/2015 05:46 PM, Peter Ujfalusi wrote:
> On 12/03/2015 05:38 PM, Arnd Bergmann wrote:
>> On Thursday 03 December 2015 16:33:12 Peter Ujfalusi wrote:
>>> diff --git a/drivers/dma/edma.c b/drivers/dma/edma.c
>>> index 0675e268d577..46b305ea0d21 100644
>>> --- a/drivers/dma/edma.c
>>> +++ b/drivers/dma/edma.c
>>> @@ -2297,6 +2297,12 @@ static int edma_probe(struct platform_device *pdev)
>>>                 edma_set_chmap(&ecc->slave_chans[i], ecc->dummy_slot);
>>>         }
>>>  
>>> +       if (info->slave_map) {
>>> +               ecc->dma_slave.filter_map.map = info->slave_map;
>>> +               ecc->dma_slave.filter_map.mapcnt = info->slavecnt;
>>> +               ecc->dma_slave.filter_map.filter_fn = edma_filter_fn;
>>> +       }
>>> +
>>>
>>
>> Just a minor comment here: I think all three assignments can be done
>> unconditionally.
> 
> True.
> 
>> As I mentioned before, I'd also remove 'struct dma_filter'
>> and put the three members in struct dma_device directly. In fact, the
>> filter function can go with the other function pointers for consistency.
> 
> I just like to keep things in one place ;)
> I don't have strong stand on keeping the intermediate 'struct dma_filter'
> Let's hear from Vinod regarding to this


One remaining design issue is on how and where to place the filter related
variables/pointers:

Keep it separated as it was in the RFC and v01 series:

struct dma_slave_map {
        const char *devname;
        const char *slave;
        void *param;
};

struct dma_filter {
        dma_filter_fn fn;
        int mapcnt;
        const struct dma_slave_map *map;
};


struct dma_device {
	...
	struct dma_filter filter;
	...
};

Or to have them under the dma_device directly:

struct dma_device {
	...
	int filter_mapcnt;
	const struct dma_slave_map *filter_map;
	...
	dma_filter_fn filter_fn;
	...
};

Vinod: what is your preference for this?

Thanks,
Péter