Re: [PATCH 5/7 v2] dma: sh: use an integer slave ID to improve APIcompatibility

From: Guennadi Liakhovetski
Date: Mon Jul 16 2012 - 03:13:36 EST


On Mon, 16 Jul 2012, Vinod Koul wrote:

> On Mon, 2012-07-16 at 08:37 +0200, Guennadi Liakhovetski wrote:
> > On Mon, 16 Jul 2012, Vinod Koul wrote:
> >
> > > On Thu, 2012-07-05 at 12:29 +0200, Guennadi Liakhovetski wrote:
> > > > diff --git a/include/linux/sh_dma.h b/include/linux/sh_dma.h
> > > > index a79f10a..4e83f3e 100644
> > > > --- a/include/linux/sh_dma.h
> > > > +++ b/include/linux/sh_dma.h
> > > > @@ -27,10 +27,10 @@ struct sh_dmae_slave {
> > > > * a certain peripheral
> > > > */
> > > > struct sh_dmae_slave_config {
> > > > - unsigned int slave_id;
> > > > - dma_addr_t addr;
> > > > - u32 chcr;
> > > > - char mid_rid;
> > > > + int slave_id;
> > > > + dma_addr_t addr;
> > > > + u32 chcr;
> > > > + char mid_rid;
> > > > };
> > > why would you want to keep your own slave_id and addr fields when they
> > > are already provided in struct dma_slave_config.
> >
> > Modifying the driver's API would break all its clients.
> rightly so, they need to be moved too.

Sure, that's why patch 6/7 is "providing a migration path" for them.

> > > Also while at it, what
> > > would be the use of last tow fields, what do they describe?
> >
> > They tell the driver how the channel has to be configured to support this
> > specific client. They are values of two specific registers. In fact, CHCR
> > means exactly that - CHannel Control Register.
> what exactly does the channel control register do in shdma? Should shdma
> driver deduce this value rather than client giving it?
> Same question for mid_rid?

See, e.g., arch/arm/mach-shmobile/setup-sh7372.c::sh7372_dmae_slaves[].
Platforms are supplying these values with shdma driver platform data,
together with slave IDs. Them when slaves request channels and supply
their slave IDs, the driver searches the above array, looking for the
matching slave ID, then it uses the rest of the information in those
structs to configure the channel for this client. Again, this is nothing
new, this is how the driver has been functioning since a long time, this
driver is not modifying anything there. Any changes to this procedure,
like providing all thig information from clients themselves instead of
keeping it with DMACs, requires these patches to be committed first.

Thanks
Guennadi
---
Guennadi Liakhovetski, Ph.D.
Freelance Open-Source Software Developer
http://www.open-technology.de/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/