Re: [PATCH v5] pinctrl: stm32: Add check for clk_enable()

[Antonio Borneo <antonio.borneo@xxxxxxxxxxx>]

On Mon, 2024-12-16 at 00:41 +0100, Marek Vasut wrote:
> On 12/15/24 9:40 PM, Mingwei Zheng wrote:
> 
> [...]
> 
> > @@ -1617,10 +1602,18 @@ int stm32_pctl_probe(struct platform_device *pdev)
> >                 return -EINVAL;
> >         }
> >         pctl->banks = devm_kcalloc(dev, banks, sizeof(*pctl->banks),
> > -                       GFP_KERNEL);
> > +                                  GFP_KERNEL);
> 
> Please drop this one change.
> 
> >         if (!pctl->banks)
> >                 return -ENOMEM;
> >   
> > +       pctl->clks = devm_kcalloc(dev, banks, sizeof(*pctl->clks),
> > +                                 GFP_KERNEL);
> > +       if (!pctl->clks)
> > +               return -ENOMEM;
> > +
> > +       for (i = 0; i < banks; ++i)
> > +               pctl->clks[i].id = "";
> 
> Is this ^ assignment necessary ? If so, why ?

The existing DTs don't have the 'clock-names' property, whose value is used to set this struct clk_bulk_data::id.
With this field kept at NULL, the error messages in clk_bulk_enable() and similar should print '(null)'.
This line sets it to empty string.
I would say it's not necessary, but I don't know if it's better to have:
"Failed to enable clk '': %d"
or
"Failed to enable clk '(null)': %d"

Antonio