Re: [PATCH/RFC] dma-mapping: Provide dummy set_dma_ops() for NO_DMA=y

From: Arnd Bergmann
Date: Mon Jul 10 2017 - 03:55:39 EST

On Sun, Jul 9, 2017 at 9:33 PM, Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> Adding a dummy for set_dma_ops() allows to compile (sub)drivers that
> don't actually use the DMA API, but propagate DMA ops configuration to a
> second driver that may or may not use the DMA API. Of course the second
> driver does have to depend on HAS_DMA if it uses the DMA API.
> An example is commit 5567e989198b5a8d ("fsl/fman: propagate dma_ops").
> This allows to revert commit 85688d9adf685572 ("fsl/fman: add dependency
> on HAS_DMA").
> Signed-off-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>

I can't think of any correct use case for this new helper. IMO a device
driver should never set the dma_map_ops for any device, and we should
instead for the fman driver correctly.