On 02/17/2014 09:18 AM, Shevchenko, Andriy wrote:
On Sun, 2014-02-16 at 14:21 +0100, Lars-Peter Clausen wrote:
[]
+ * If not: if it is fixed so that it be sent in static from the platform
+ * data, then prefer to do that.
This configuration is per channel, correct?
Then it sounds wrong to pass this via platform data, especially in case
when, for example, channels are equal and user doesn't care which
channel they would like to use.
There is certain static configuration options that do not belong in the DMA
client driver because he doesn't have to know about them. E.g. options that
describe the fabric between the DMA master and the DMA slave. We've already
solved this properly for devicetree where you can add additional information
to the specifier in the client node, but which is completely transparent to
the slave driver and gets parsed by the master driver. For legacy platforms
with board drivers we are still missing the DMA equivalent to something like
the GPIO descriptor API.
Doing this via platform_data for now is not perfect, since it does, as you
said, not work in all cases, but it's in my opinion better then what a lot
of drivers currently do by abusing the filter callback and assign data to
the channels private_data field in the callback.
I btw. didn't add that paragraph, just re-flowed it.
- Lars
--
To unsubscribe from this list: send the line "unsubscribe dmaengine" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html