On Mon, Nov 04, 2013 at 10:43:03PM +0100, Martin Sperl wrote:

> again: first the dma-engine part would need to be upstream before I assume
> you would accept to merge the driver, as the functionality would NOT work
> otherwise. So we talk about 3.15 at the very earliest...

Well, v3.14 should be possible - we're just coming up to the merge
window for v3.13 so v3.14 is still wide open.

> So is the presented idea an approach to move forward on?
> Or do you have a better approach which we could follow to get us out of this 
> tie?

I think the current approach seems fine.

> /* generic wrapper functions for device drivers to use */
> static int bcm2835dma_spi_prepare_message(struct spi_device *spi,
> 	struct spi_message *msg)
> {
> 	if(spi->master->prepare_message) {
> 		return spi->master->prepare_message(spi,msg);
> 	} else {
> 		return 0;
> 	}
> }

Like I said earlier I'm concerned about the idea of drivers calling
these functions directly, this should all be factored out of drivers
since there's nothing really driver specific here.