Re: [PATCH 6/7] dt-bindings: dmaengine: at_xdmac: add optional microchip,m2m property

[Vinod Koul]

Hi Eugen,

On 16-10-20, 06:45, Eugen.Hristev@xxxxxxxxxxxxx wrote:
> On 23.09.2020 02:33, Rob Herring wrote:
> 
> > On Mon, Sep 14, 2020 at 05:09:55PM +0300, Eugen Hristev wrote:
> >> Add optional microchip,m2m property that specifies if a controller is
> >> dedicated to memory to memory operations only.
> >>
> >> Signed-off-by: Eugen Hristev <eugen.hristev@xxxxxxxxxxxxx>
> >> ---
> >>   Documentation/devicetree/bindings/dma/atmel-xdma.txt | 6 ++++++
> >>   1 file changed, 6 insertions(+)
> >>
> >> diff --git a/Documentation/devicetree/bindings/dma/atmel-xdma.txt b/Documentation/devicetree/bindings/dma/atmel-xdma.txt
> >> index 510b7f25ba24..642da6b95a29 100644
> >> --- a/Documentation/devicetree/bindings/dma/atmel-xdma.txt
> >> +++ b/Documentation/devicetree/bindings/dma/atmel-xdma.txt
> >> @@ -15,6 +15,12 @@ the dmas property of client devices.
> >>       interface identifier,
> >>       - bit 30-24: PERID, peripheral identifier.
> >>
> >> +Optional properties:
> >> +- microchip,m2m: this controller is connected on AXI only to memory and it's
> >> +     dedicated to memory to memory DMA operations. If this option is
> >> +     missing, it's assumed that the DMA controller is connected to
> >> +     peripherals, thus it's a per2mem and mem2per.
> > 
> > Wouldn't 'dma-requests = <0>' cover this case?
> > 
> > Rob
> > 
> 
> Hi Rob,
> 
> I do not think so. With requests = 0, it means that actually the DMA 
> controller is unusable ?
> Since you suggest requests = 0, it means that it cannot take requests at 
> all ?
> I do not find another example in current DT with this property set to zero.

Not really, dma-requests implies "request signals supported" which are
used for peripheral cases. m2m does not need request signals, so it is
very reasonable to conclude that dma-requests = <0> would imply no
peripheral support and only m2m support.

Thanks
-- 
~Vinod