Re: [PATCH] clk: Add clk_set_parent debugfs node

[Geert Uytterhoeven]

Hi Sam,

On Tue, Oct 5, 2021 at 12:11 PM Sam Protsenko
<semen.protsenko@xxxxxxxxxx> wrote:
> On Tue, 14 Sept 2021 at 17:19, Sam Protsenko <semen.protsenko@xxxxxxxxxx> wrote:
> > Useful for testing mux clocks. One can write the index of the parent to
> > set into clk_set_parent node, starting from 0. Example
> >
> >     # cat clk_possible_parrents
> >       dout_shared0_div4 dout_shared1_div4
> >     # cat clk_parent
> >       dout_shared0_div4
> >     # echo 1 > clk_set_parent
> >     # cat clk_parent
> >       dout_shared1_div4
> >
> > Define CLOCK_ALLOW_WRITE_DEBUGFS in drivers/clk/clk.c in order to use
> > this feature.
> >
> > Signed-off-by: Sam Protsenko <semen.protsenko@xxxxxxxxxx>
> > ---
>
> + Adding more folks for review
>
> Guys, can you please review this one?

Thanks for your patch!

> > --- a/drivers/clk/clk.c
> > +++ b/drivers/clk/clk.c
> > @@ -3214,6 +3214,30 @@ static int current_parent_show(struct seq_file *s, void *data)
> >  }
> >  DEFINE_SHOW_ATTRIBUTE(current_parent);
> >
> > +#ifdef CLOCK_ALLOW_WRITE_DEBUGFS
> > +static int clk_set_parent_set(void *data, u64 val)

u64 is overkill, num_parents is u8.

> > +{
> > +       struct clk_core *core = data, *parent;
> > +       int ret;
> > +
> > +       if (val >= core->num_parents)
> > +               return -EINVAL;

clk_core_get_parent_by_index() called below already checks this.

> > +
> > +       parent = clk_core_get_parent_by_index(core, val);
> > +       if (IS_ERR_OR_NULL(parent))
> > +               return PTR_ERR(parent);
> > +
> > +       clk_prepare_lock();
> > +       ret = clk_core_set_parent_nolock(core, parent);
> > +       clk_prepare_unlock();
> > +
> > +       return ret;
> > +}
> > +
> > +DEFINE_DEBUGFS_ATTRIBUTE(clk_set_parent_fops, NULL, clk_set_parent_set,
> > +                        "%llu\n");
> > +#endif
> > +
> >  static int clk_duty_cycle_show(struct seq_file *s, void *data)
> >  {
> >         struct clk_core *core = s->private;
> > @@ -3285,9 +3309,14 @@ static void clk_debug_create_one(struct clk_core *core, struct dentry *pdentry)
> >                 debugfs_create_file("clk_parent", 0444, root, core,
> >                                     &current_parent_fops);
> >
> > -       if (core->num_parents > 1)
> > +       if (core->num_parents > 1) {
> >                 debugfs_create_file("clk_possible_parents", 0444, root, core,
> >                                     &possible_parents_fops);
> > +#ifdef CLOCK_ALLOW_WRITE_DEBUGFS
> > +               debugfs_create_file("clk_set_parent", 0200, root, core,
> > +                                   &clk_set_parent_fops);
> > +#endif

Why add a new file, instead of making the existing "clk_parent" file
writable, like is done for "clk_rate"?
Yes, "clk_parent" prints a name, while you use a parent number, but
I guess that can be fixed? Or even both can be accepted?

> > +       }
> >
> >         if (core->ops->debug_init)
> >                 core->ops->debug_init(core->hw, core->dentry);

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds