[RFC] dmaengine: add new api for preparing simple slave transfer

Raju, Sundaram sundaram at ti.com
Tue Jun 14 01:38:51 EDT 2011


Linus,

Thanks for the pointers.

> -----Original Message-----
> From: Linus Walleij [mailto:linus.walleij at linaro.org]
> Sent: Monday, June 13, 2011 7:43 PM
> To: Raju, Sundaram
> Cc: Russell King - ARM Linux; Koul, Vinod; Dan; davinci-linux-open-
> source at linux.davincidsp.com; linux-omap at vger.kernel.org; linux-
> kernel at vger.kernel.org; linux-arm-kernel at lists.infradead.org
> Subject: Re: [RFC] dmaengine: add new api for preparing simple slave transfer
> 
> On Fri, Jun 10, 2011 at 3:33 PM, Russell King - ARM Linux
> <linux at arm.linux.org.uk> wrote:
> > On Fri, Jun 10, 2011 at 05:18:46PM +0530, Raju, Sundaram wrote:
> >> Now DMACs capable of 3D transfer, do transfer of the whole 1D
> >> buffer per sync received or even whole 2D buffer per sync received
> >> (based on the sync rate programmed in the DMAC).
> >
> > The only issue which I see that we don't cover is the case where you want
> > to describe a single buffer which is organised as N bytes to be transferred,
> > M following bytes to be skipped, N bytes to be transferred, M bytes to be
> > skipped.  I doubt there are many controllers which can be programmed with
> > both 'N' and 'M' parameters directly.
> 
> Sundaram is this how your controller works?
> I mean the hardware can skip over sequences like this?
> 
> When we added the config interface to DMAengine I originally included
> a "custom config" call, but Dan wanted me to keep it out until we
> had some specific usecase for it. FSLDMA recently started
> to use it.
> 
> Notice how dmaengine_slave_config() is implemented:
> 
> static inline int dmaengine_slave_config(struct dma_chan *chan,
> 					  struct dma_slave_config *config)
> {
> 	return dmaengine_device_control(chan, DMA_SLAVE_CONFIG,
> 			(unsigned long)config);
> }
> 
> So what is passed to the driver is just an unsigned long.
> 
> This is actually modeled to be ioctl()-like so you can pass in a
> custom config to the same callback on the device driver,
> just use some other enumerator than DMA_SLAVE_CONFIG,
> say like FSLDMA already does with FSLDMA_EXTERNAL_START.
> 
> Just put some enumerator in enum dma_ctrl_cmd in
> dmaengine.h such as SDMA_TEXAS_STRIDE_CONFIG and call
> like this:
> 
> /* However that config struct needs to look, basically */
> static struct sdma_ti_stride_cgf = {
>      take = M,
>      skip = N,
> };
> 
> ret = chan->device->device_control(chan, SDMA_TEXAS_STRIDE_CONFIG,
> &sdma_ti_stride_cfg);
> 
> Or something like this.

Yes, the hardware can skip over sequences like that.
I also thought about your suggestion, at first before submitting the RFC.
But I dint pursue this because, this ioctl() call has to be made before
every single prepare call.

I may have to end up using this if we decide not to change the API
signature for prepare APIs.

I actually intend to use this for all DMAC related ioctl(). Configuring
the DMAC and programming various modes etc specific to the
DMAC. I suppose this is the only way to do it.
Let me know if there is any other way to do it.

Thanks,
Sundaram



More information about the linux-arm-kernel mailing list