RE: [PATCH 1/3] drivers: clk: Add clk_get_children support

[Jolly Shah]

Hi Stephan,

> -----Original Message-----
> From: Stephen Boyd [mailto:sboyd@xxxxxxxxxx]
> Sent: Monday, March 19, 2018 11:22 AM
> To: Jolly Shah <JOLLYS@xxxxxxxxxx>; linux-clk@xxxxxxxxxxxxxxx;
> mark.rutland@xxxxxxx; michal.simek@xxxxxxxxxx; mturquette@xxxxxxxxxxxx;
> robh+dt@xxxxxxxxxx; sboyd@xxxxxxxxxxxxxx
> Cc: Rajan Vaja <RAJANV@xxxxxxxxxx>; devicetree@xxxxxxxxxxxxxxx; linux-arm-
> kernel@xxxxxxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; Jolly Shah
> <JOLLYS@xxxxxxxxxx>; Jolly Shah <JOLLYS@xxxxxxxxxx>; Tejas Patel
> <TEJASP@xxxxxxxxxx>; Shubhrajyoti Datta <shubhraj@xxxxxxxxxx>
> Subject: Re: [PATCH 1/3] drivers: clk: Add clk_get_children support
> 
> Quoting Jolly Shah (2018-02-28 14:27:39)
> > From: Jolly Shah <jolly.shah@xxxxxxxxxx>
> >
> > This API helps to determine the users for any clock.
> 
> Ok, but why do you need it?

As you suggested in other patch, we will move children validation in FW.

> 
> >
> > Signed-off-by: Jolly Shah <jollys@xxxxxxxxxx>
> > Signed-off-by: Tejas Patel <tejasp@xxxxxxxxxx>
> > Signed-off-by: Shubhrajyoti Datta <shubhrajyoti.datta@xxxxxxxxxx>
> > ---
> >  drivers/clk/clk.c            | 28 ++++++++++++++++++++++++++++
> >  include/linux/clk-provider.h |  1 +
> >  2 files changed, 29 insertions(+)
> >
> > diff --git a/drivers/clk/clk.c b/drivers/clk/clk.c index
> > 0f686a9..947a18b 100644
> > --- a/drivers/clk/clk.c
> > +++ b/drivers/clk/clk.c
> > @@ -274,6 +274,34 @@ struct clk_hw *clk_hw_get_parent(const struct
> > clk_hw *hw)  }  EXPORT_SYMBOL_GPL(clk_hw_get_parent);
> >
> > +static unsigned int sibling;
> 
> Looks very thread unsafe!
> 
> > +
> > +static void clk_show_subtree(struct clk_core *c,
> > +                            int level) {
> > +       struct clk_core *child;
> > +
> > +       if (!c)
> > +               return;
> > +
> > +       if (level == 1)
> > +               sibling++;
> > +
> > +       hlist_for_each_entry(child, &c->children, child_node)
> > +               clk_show_subtree(child, level + 1); }
> > +
> > +unsigned int clk_get_children(char *name) {
> > +       struct clk_core *core;
> > +       struct clk *pclk = __clk_lookup(name);
> > +
> > +       sibling = 0;
> > +       core = pclk->core;
> > +       clk_show_subtree(core, 0);
> > +       return sibling;
> > +}
> > +
> >  static struct clk_core *__clk_lookup_subtree(const char *name,
> >                                              struct clk_core *core)  {
> > diff --git a/include/linux/clk-provider.h
> > b/include/linux/clk-provider.h index f711be6..e94dfb2 100644
> > --- a/include/linux/clk-provider.h
> > +++ b/include/linux/clk-provider.h
> > @@ -745,6 +745,7 @@ unsigned int __clk_get_enable_count(struct clk
> > *clk);  unsigned long clk_hw_get_rate(const struct clk_hw *hw);
> > unsigned long __clk_get_flags(struct clk *clk);  unsigned long
> > clk_hw_get_flags(const struct clk_hw *hw);
> > +unsigned int clk_get_children(char *name);
> 
> And uses a string lookup instead of having the clk_hw pointer in hand.
> No thanks.