Re: [PATCH v2 10/19] dma: imx-sdma: Add multi fifo support

[Vinod Koul <vkoul@kernel.org>]

On 31-03-22, 09:58, Sascha Hauer wrote:
> On Thu, Mar 31, 2022 at 12:24:46PM +0530, Vinod Koul wrote:
> > On 31-03-22, 08:49, Sascha Hauer wrote:
> > > On Thu, Mar 31, 2022 at 10:56:09AM +0530, Vinod Koul wrote:
> > > > On 28-03-22, 13:27, Sascha Hauer wrote:
> > > > > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> > > > 
> > > > it is dmaengine: xxx
> > > 
> > > Ok.
> > > 
> > > > 
> > > > Also is this patch dependent on rest of the series, if not consider
> > > > sending separately
> > > 
> > > The rest of this series indeed depends on this patch.
> > > 
> > > > 
> > > > > diff --git a/include/linux/platform_data/dma-imx.h b/include/linux/platform_data/dma-imx.h
> > > > > index 281adbb26e6bd..4a43a048e1b4d 100644
> > > > > --- a/include/linux/platform_data/dma-imx.h
> > > > > +++ b/include/linux/platform_data/dma-imx.h
> > > > > @@ -39,6 +39,7 @@ enum sdma_peripheral_type {
> > > > >  	IMX_DMATYPE_SSI_DUAL,	/* SSI Dual FIFO */
> > > > >  	IMX_DMATYPE_ASRC_SP,	/* Shared ASRC */
> > > > >  	IMX_DMATYPE_SAI,	/* SAI */
> > > > > +	IMX_DMATYPE_MULTI_SAI,	/* MULTI FIFOs For Audio */
> > > > >  };
> > > > >  
> > > > >  enum imx_dma_prio {
> > > > > @@ -65,4 +66,10 @@ static inline int imx_dma_is_general_purpose(struct dma_chan *chan)
> > > > >  		!strcmp(chan->device->dev->driver->name, "imx-dma");
> > > > >  }
> > > > >  
> > > > > +struct sdma_peripheral_config {
> > > > > +	int n_fifos_src;
> > > > > +	int n_fifos_dst;
> > > > > +	bool sw_done;
> > > > > +};
> > > > 
> > > > Not more platform data :(
> > > 
> > > I'm not sure what you are referring to as platform_data. This is not the
> > > classical platform_data that is attached to a platform_device to
> > > configure behaviour of that device. It is rather data that needs to be
> > > communicated from the clients of the SDMA engine to the SDMA engine.
> > > 
> > > I have put this into include/linux/platform_data/dma-imx.h because
> > > that's the only existing include file that is available. I could move
> > > this to a new file if you like that better.
> > 
> > Lets move to include/linux/dma/
> 
> Ok.
> 
> > 
> > > 
> > > > 
> > > > Can you explain this structure and why this is required? What do these
> > > > fields refer to..?
> > > 
> > > The reasoning for this structure is described in the commit message that
> > > I have forgotten:
> > > 
> > >     The i.MX SDMA engine can read from / write to multiple successive
> > >     hardware FIFO registers, referred to as "Multi FIFO support". This is
> > >     needed for the micfil driver and certain configurations of the SAI
> > >     driver. This patch adds support for this feature.
> > > 
> > >     The number of FIFOs to read from / write to must be communicated from
> > >     the client driver to the SDMA engine. For this the struct
> > >     dma_slave_config::peripheral_config field is used.
> > > 
> > > I can describe the individual fields of struct sdma_peripheral_config in
> > > the header file if that's your point.
> > 
> > So you need to know the number of fifo right, what does sw_done imply?
> 
> Honestly I don't know. Setting sw_done results in the DONE_SEL0 bit in
> the SDMA engine being set. This is described in the reference manual
> as:
> 
> DONE_SEL0 Select Done from SW or HW for channel 0
>           0 HW
>           1 SW
> 
> I can only assume that the signaling when a channel has transferred
> enough data (the generation of the channel done interrupt?) can either
> be done in hardware or in software in the SDMA engine. What I can tell
> for sure is that I need this bit set ;)

That does sound like a mechanism to tell that transfer is done aka
handshaking aka interrupt line.

-- 
~Vinod