Re: [PATCH v3 2/3] clk: qcom: lpass-sc7180: Disentangle the two clock devices

[Doug Anderson]

Hi,

On Wed, Oct 14, 2020 at 3:10 PM Stephen Boyd <sboyd@xxxxxxxxxx> wrote:
>
> Quoting Douglas Anderson (2020-10-14 14:05:22)
> > diff --git a/drivers/clk/qcom/lpasscorecc-sc7180.c b/drivers/clk/qcom/lpasscorecc-sc7180.c
> > index abcf36006926..48d370e2108e 100644
> > --- a/drivers/clk/qcom/lpasscorecc-sc7180.c
> > +++ b/drivers/clk/qcom/lpasscorecc-sc7180.c
> > @@ -356,12 +356,48 @@ static const struct qcom_cc_desc lpass_audio_hm_sc7180_desc = {
> >         .num_gdscs = ARRAY_SIZE(lpass_audio_hm_sc7180_gdscs),
> >  };
> >
> > +static void lpass_pm_runtime_disable(void *data)
> > +{
> > +       pm_runtime_disable(data);
> > +}
> > +
> > +static void lapss_pm_clk_destroy(void *data)
> > +{
> > +       pm_clk_destroy(data);
> > +}
>
> Why are these helpers added again? And do we even need them? Can't we
> just pass pm_runtime_disable or pm_clk_destroy to the
> devm_add_action_or_reset() second parameter?

Unfortunately, we can't due to the C specification.  Take a look at
all the other users of devm_add_action_or_reset() and they all have
pretty much the same stupid thing.  This is a pretty great grep, for
instance:

git grep devm_add_action_or_reset.*clk_disable_unprepare

I did a quick Google search and the top hit was a stack overflow that
explained it:

https://stackoverflow.com/questions/559581/casting-a-function-pointer-to-another-type

The net-net is that the answer there says:

> Hence, since a void* is not compatible with a struct my_struct*, a
> function pointer of type void (*)(void*) is not compatible with a
> function pointer of type void (*)(struct my_struct*), so this
> casting of function pointers is technically undefined behavior.

I suppose I could try to add devm variants of these functions
somewhere more general if you think it's a good idea, though there it
seems like there's not a huge need since these two greps are zero:

git grep devm_add_action_or_reset.*runtime_disable
git grep devm_add_action_or_reset.*pm_clk_destroy

...actually, do we even need the runtime_disable in the error path?
When the dev goes away does it matter if you left pm_runtime enabled
on it?

-Doug