Re: [PATCH] clk: starfive: jh7100: Use provided clocks instead of hardcoded names

[Geert Uytterhoeven]

Hi Samuel,

On Thu, Mar 14, 2024 at 3:56 PM Samuel Holland
<samuel.holland@xxxxxxxxxx> wrote:
> On 2024-03-14 9:48 AM, Geert Uytterhoeven wrote:
> > On Thu, Mar 14, 2024 at 3:32 PM Samuel Holland
> > <samuel.holland@xxxxxxxxxx> wrote:
> >> On 2024-03-14 6:05 AM, Geert Uytterhoeven wrote:
> >>> The Starfive JH7100 clock driver does not use the DT "clocks" property
> >>> to find its external input clocks, but instead relies on the names of
> >>> the actual external clock providers.  This is fragile, and caused
> >>> breakage when sanitizing clock names in DT.
> >>>
> >>> Fix this by obtaining the external input clocks through the DT "clocks"
> >>> property, and using their clk_hw objects or corresponding name.
> >>>
> >>> Fixes: f03606470886 ("riscv: dts: starfive: replace underscores in node names")
> >>> Fixes: 4210be668a09ee20 ("clk: starfive: Add JH7100 clock generator driver")
> >>> Signed-off-by: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> >
> >>> --- a/drivers/clk/starfive/clk-starfive-jh7100.c
> >>> +++ b/drivers/clk/starfive/clk-starfive-jh7100.c
> >
> >>> @@ -298,13 +311,23 @@ static int __init clk_starfive_jh7100_probe(struct platform_device *pdev)
> >>>       if (IS_ERR(priv->base))
> >>>               return PTR_ERR(priv->base);
> >>>
> >>> +     for (idx = 0; idx < EXT_NUM_CLKS; idx++) {
> >>> +             clk = devm_clk_get(&pdev->dev, jh7100_ext_clk[idx]);
> >>> +             if (IS_ERR(clk))
> >>> +                     return PTR_ERR(clk);
> >>> +
> >>> +             priv->ext[idx] = __clk_get_hw(clk);
> >>> +     }
> >>> +
> >>> +     osc_sys = clk_hw_get_name(priv->ext[EXT_CLK_OSC_SYS]);
> >>> +
> >>>       priv->pll[0] = devm_clk_hw_register_fixed_factor(priv->dev, "pll0_out",
> >>> -                                                      "osc_sys", 0, 40, 1);
> >>> +                                                      osc_sys, 0, 40, 1);
> >>>       if (IS_ERR(priv->pll[0]))
> >>>               return PTR_ERR(priv->pll[0]);
> >>>
> >>>       priv->pll[1] = devm_clk_hw_register_fixed_factor(priv->dev, "pll1_out",
> >>> -                                                      "osc_sys", 0, 64, 1);
> >>> +                                                      osc_sys, 0, 64, 1);
> >>
> >> These should use devm_clk_hw_register_fixed_factor_parent_hw(). (Or you could
> >
> > Thanks, I didn't know about that function!
> >
> >> define a devm_clk_hw_register_fixed_factor_fw_name() and drop the other changes.)
> >
> > Sorry, I don't understand what you mean here?
>
> In the loop below, the parents are already referenced via .fw_name. That means
> the string is the DT clock-names property value, not the Linux-internal clock
> name (see clk_core_get()). These two function calls are the only ones that
> depend on the internal clock name. If you change them to use .fw_name as well,
> the clk_core_get() will do the right thing, and you don't need to manually call
> devm_clk_get().

Thank you, I wasn't aware these fw_name names were actually doing the
right thing ;-)

Still, looking up by name is more expensive than just using the passed
clk_hw.  As there are multiple occurrences of the four external
clocks in the clocks table, thus involving multiple look-ups, I think
it's better to use clk_hw everywhere.

I will update for v2...

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68korg

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