Re: [PATCH] clk: Constify struct clk_init_data

[Saravana Kannan]

On 05/15/2012 09:42 AM, Saravana Kannan wrote:
> On 05/15/2012 12:00 AM, Sascha Hauer wrote:
> > On Mon, May 14, 2012 at 06:19:16PM -0700, Saravana Kannan wrote:
> > > On 05/14/2012 02:53 PM, Turquette, Mike wrote:
> > > > On Mon, May 14, 2012 at 7:12 AM, Mark
> > > > Brown<broonie@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote:
> > > > > Allow drivers to declare their clk_init_data const, the framework
> > > > > really
> > > > > shouldn't be modifying the data.
> > > > >
> > > > > Signed-off-by: Mark Brown<broonie@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
> > > >
> > > > +interested parties
> > > >
> > > > > diff --git a/include/linux/clk-provider.h
> > > > > b/include/linux/clk-provider.h
> > > > > index c1c23b9..fc43ea6 100644
> > > > > --- a/include/linux/clk-provider.h
> > > > > +++ b/include/linux/clk-provider.h
> > > > > @@ -143,7 +143,7 @@ struct clk_init_data {
> > > > > */
> > > > > struct clk_hw {
> > > > > struct clk *clk;
> > > > > - struct clk_init_data *init;
> > > > > + const struct clk_init_data *init;
> > >
> > > Oh, wait. This won't work for the case where the clock registration
> > > is completely dynamic. Say, created from device tree or thru some
> > > PCI/USB device probe, etc. That's why I didn't add it before.
> >
> > Why not? In this case clk_init_data is also only used in clk_register.
> > Given that Mark has posted the patch I assume it actually works.
> >
> > Sascha
>
> It's used in __clk_register() though. Which I added as a "as close as
> clk_register() but allows static init" function.

Sorry about my rushed responses earlier. I don't think the const will 
work even when purely using clk_register().

Say I use device tree (you can extend this to HW probing/detection) to 
know that there is a fixed factor clock. Now, if I want to dynamically 
create and register it, I will have to allocate struct clk_fixed_factor 
and struct clk_init_data. Then populate the fields of those structs and 
finally point clk_fixed_factor.hw.init to the allocated clk_init_data. 
Then call clk_register on clk_fixed_factor.hw. I'm not sure the 
assigning of the init field above will be possible if we mark it as const.

Really though, not marking it as const shouldn't really have a huge 
impact. What matters is that the common clock framework shouldn't try to 
use it after clk_register() and we can do that by following what Russell 
suggested (set hw.init to NULL at the end of clk_register).

Does that make sense?

Thanks,
Saravana

--
Sent by an employee of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
--
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/