Re: [PATCH 2/2] pinctrl: spacemit: add clock support for K1 SoC

[Rob Herring]

On Sat, Apr 12, 2025 at 10:34:31PM +0000, Yixun Lan wrote:
> Hi Rob,
> 
> On 13:27 Sat 12 Apr     , Rob Herring wrote:
> > On Sat, Apr 12, 2025 at 02:58:11PM +0800, Yixun Lan wrote:
> > > For SpacemiT K1 SoC's pinctrl, explicitly acquiring clocks in
> > > the driver instead of relying on bootloader or default hardware
> > > settings to enable it.
> > > 
> > > Signed-off-by: Yixun Lan <dlan@xxxxxxxxxx>
> > > ---
> > >  drivers/pinctrl/spacemit/pinctrl-k1.c | 10 ++++++++++
> > >  1 file changed, 10 insertions(+)
> > > 
> > > diff --git a/drivers/pinctrl/spacemit/pinctrl-k1.c b/drivers/pinctrl/spacemit/pinctrl-k1.c
> > > index 67e867b04a02ea1887d93aedfdea5bda037f88b1..3805fb09c1bc3b8cf2ccfc22dd25367292b397b9 100644
> > > --- a/drivers/pinctrl/spacemit/pinctrl-k1.c
> > > +++ b/drivers/pinctrl/spacemit/pinctrl-k1.c
> > > @@ -2,6 +2,7 @@
> > >  /* Copyright (c) 2024 Yixun Lan <dlan@xxxxxxxxxx> */
> > >  
> > >  #include <linux/bits.h>
> > > +#include <linux/clk.h>
> > >  #include <linux/cleanup.h>
> > >  #include <linux/io.h>
> > >  #include <linux/of.h>
> > > @@ -721,6 +722,7 @@ static int spacemit_pinctrl_probe(struct platform_device *pdev)
> > >  {
> > >  	struct device *dev = &pdev->dev;
> > >  	struct spacemit_pinctrl *pctrl;
> > > +	struct clk *func_clk, *bus_clk;
> > >  	const struct spacemit_pinctrl_data *pctrl_data;
> > >  	int ret;
> > >  
> > > @@ -739,6 +741,14 @@ static int spacemit_pinctrl_probe(struct platform_device *pdev)
> > >  	if (IS_ERR(pctrl->regs))
> > >  		return PTR_ERR(pctrl->regs);
> > >  
> > > +	func_clk = devm_clk_get_optional_enabled(dev, "func");
> > > +	if (IS_ERR(func_clk))
> > > +		return dev_err_probe(dev, PTR_ERR(func_clk), "failed to get func clock\n");
> > > +
> > > +	bus_clk = devm_clk_get_optional_enabled(dev, "bus");
> > > +	if (IS_ERR(bus_clk))
> > > +		return dev_err_probe(dev, PTR_ERR(bus_clk), "failed to get bus clock\n");
> > 
> > Do you really need these to be optional? Yes, it maintains 
> >
> Yes, the motivation here to make it optional is maintaining the 
> compatibility, to not break the case of using old dtb with new kernel,
> since the serial console device (uart0) is activated now [1]
> 
> IIUC, from the DT perspective, it's mandatory to keep this compatibility?

It's up to the platform. It's only mandatory to understand you are 
breaking compatibility when you do.

> 
> In dt-binding of patch [1/2], the clocks/clock-names has been described as 
> required property, so it's sort of mandatory from DT's view.
> 
> One lesson I learned is that the pinctrl dt node shouldn't be activated
> untill all prerequisite dependencies meet.. it's ok to push the driver,
> but should postpone the DT part..
> 
> > compatibility, but if this platform isn't stable, then do you really 
> > need that?
> > 
> I don't get what's your meaning of "isn't stable" here, the fact won't
> change for K1 SoC: the pinctrl controller requires two clocks

You required 1 clock and now you require 2. That's not stable. 
Generally, early on with platforms, their DT is not complete enough to 
maintain compatibility. There's also likely not many users or h/w 
availability for compatibility to be an issue. So requiring the DT to be 
in-sync with the kernel is not a problem.

Rob