Re: [PATCH 0/1] Compact interface for Device-Tree

[Arnd Bergmann <arnd@arndb.de>]

On Friday 31 October 2014 23:53:28 Rafael J. Wysocki wrote:
> On Saturday, November 01, 2014 05:13:45 AM Rob Herring wrote:
> > On Fri, Oct 31, 2014 at 6:59 AM, Gilad Avidov <gavidov@codeaurora.org> wrote:
> > >
> > > Device-Tree compact API
> > > ------------------------
> > >
> > > Common code seen in driver’s probe reads device tree values and handling
> > > erroneous return codes from all those of_property_read_xxx()  APIs. This
> > > common code is factored out by the of_property_map module which allows
> > > driver’s probe to replace that (often lengthy) code with a concise table:
> > >
> > > struct of_prop_map map[] = {
> > >     {"i2c",            &dev->id,        OF_REQ,  OF_ID,  -1},
> > >     {"qcom,clk-freq-out",    &dev->clk_freq_out,    OF_REQ,  OF_U32,  0},
> > >     {"qcom,clk-freq-in",    &dev->clk_freq_in,    OF_REQ,  OF_U32,  0},
> > >     {"qcom,disable-dma",    &dev->disable_dma,    OF_OPT,  OF_BOOL, 0},
> > >     {"qcom,master-id",    &dev->mstr_id,        OF_SGST, OF_U32,  0},
> > >     {NULL,            NULL,            0,       0,       0},
> > > };
> > >
> > > Then call populate to read the values into the device’s variables:
> > >
> > > ret = of_prop_populate(dev, dev->of_node, map);
> > 
> > Interesting idea. The main concern I have with this is there has been
> > on-going discussions about how to generalize property handling across
> > DT and ACPI to make drivers more agnostic, so I'm copying a few folks
> > involved in that. That may be a bit orthogonal to what this is doing,
> > but we may want some coordination here.
> 
> Agreed.
> 
> We actually have a patchset adding a unified device property API in 
> linux-next (http://git.kernel.org/cgit/linux/kernel/git/rafael/linux-pm.git/log/?h=device-properties)
> and I'd prefer to see the "compactization" to happen at that level, if possible,
> rather that for of_ only.

Agreed, this should definitely use the new generalized API.
I have prototyped a similar concept last year, which actually went much
further and also abstracted high-level properties such as interrupts,
gpios, pwm, dma-engine, etc. I still think we should do something
like that, but I've never had the time to follow up and nobody else
picked up my work from back then.

Would others like to see that?

	Arnd