Re: [RESEND PATCH v6 5/7] clk: starfive: jh7110-sys: Add PLL clocks source from DTS

From: Emil Renner Berthing
Date: Fri Jul 14 2023 - 05:36:50 EST


On Fri, 14 Jul 2023 at 10:05, Xingyu Wu <xingyu.wu@xxxxxxxxxxxxxxxx> wrote:
>
> On 2023/7/13 21:15, Emil Renner Berthing wrote:
> > On Tue, 4 Jul 2023 at 08:49, Xingyu Wu <xingyu.wu@xxxxxxxxxxxxxxxx> wrote:
> >>
> >> Modify PLL clocks source to be got from DTS or
> >> the fixed factor clocks.
> >>
> >> Signed-off-by: Xingyu Wu <xingyu.wu@xxxxxxxxxxxxxxxx>
> >> ---
> >> drivers/clk/starfive/Kconfig | 1 +
> >> .../clk/starfive/clk-starfive-jh7110-sys.c | 45 +++++++++++--------
> >> 2 files changed, 28 insertions(+), 18 deletions(-)
> >>
> >> diff --git a/drivers/clk/starfive/Kconfig b/drivers/clk/starfive/Kconfig
> >> index 5195f7be5213..978b78ec08b1 100644
> >> --- a/drivers/clk/starfive/Kconfig
> >> +++ b/drivers/clk/starfive/Kconfig
> >> @@ -35,6 +35,7 @@ config CLK_STARFIVE_JH7110_SYS
> >> select AUXILIARY_BUS
> >> select CLK_STARFIVE_JH71X0
> >> select RESET_STARFIVE_JH7110 if RESET_CONTROLLER
> >> + select CLK_STARFIVE_JH7110_PLL
> >> default ARCH_STARFIVE
> >> help
> >> Say yes here to support the system clock controller on the
> >> diff --git a/drivers/clk/starfive/clk-starfive-jh7110-sys.c b/drivers/clk/starfive/clk-starfive-jh7110-sys.c
> >> index e6031345ef05..d56f48013388 100644
> >> --- a/drivers/clk/starfive/clk-starfive-jh7110-sys.c
> >> +++ b/drivers/clk/starfive/clk-starfive-jh7110-sys.c
> >> @@ -7,6 +7,7 @@
> >> */
> >>
> >> #include <linux/auxiliary_bus.h>
> >> +#include <linux/clk.h>
> >> #include <linux/clk-provider.h>
> >> #include <linux/init.h>
> >> #include <linux/io.h>
> >> @@ -386,6 +387,7 @@ EXPORT_SYMBOL_GPL(jh7110_reset_controller_register);
> >>
> >> static int __init jh7110_syscrg_probe(struct platform_device *pdev)
> >> {
> >> + bool use_fixed_pll = true; /* PLL clocks use fixed factor clocks or PLL driver */
> >
> > nit: reverse christmas tree ordering, eg. move this below priv
> >
> >> struct jh71x0_clk_priv *priv;
> >> unsigned int idx;
> >> int ret;
> >> @@ -402,28 +404,29 @@ static int __init jh7110_syscrg_probe(struct platform_device *pdev)
> >> if (IS_ERR(priv->base))
> >> return PTR_ERR(priv->base);
> >>
> >> - /*
> >> - * These PLL clocks are not actually fixed factor clocks and can be
> >> - * controlled by the syscon registers of JH7110. They will be dropped
> >> - * and registered in the PLL clock driver instead.
> >> - */
> >> + if (!IS_ERR(devm_clk_get(priv->dev, "pll0_out")))
> >> + use_fixed_pll = false; /* can get pll clocks from PLL driver */
> >
> > The devm_clk_get() variant will allocate memory for a callback to call
> > clk_put() when the driver is unloaded, but proper references
> > associated with the consumers of the pll0_out clock are already taken
> > below. So unless we find a better way to detect if the pll references
> > are specified in the device tree or not, maybe something like this
> > instead:
> >
>
> Thanks. It looks more reasonable. I will follow it in next version.
>
> > priv->pll[0] = clk_get(priv->dev, "pll0_out);
>
> The priv->pll[] are clk_hw* struct no clk* struct and this could be failed
> when building. So maybe use a temporary clk* struct.

Ah yes, you're right. You'll need a local struct clk *clk for that then.

> > if (IS_ERR(priv->pll[0])) {
> > /* 24MHZ -> 1000.0MHz */
> > priv->pll[0] = ...
> > ...
> >
> > } else {
> > clk_put(priv->pll[0]);
> > priv->pll[0] = NULL;
>
> >
> >> + /* Use fixed factor clocks if can not get the PLL clocks from DTS */
> >> + if (use_fixed_pll) {
> >> /* 24MHz -> 1000.0MHz */
> >
> > These comments are not indented with the code, which just looks weird.
>
> Will fix.
>
> >
> >> - priv->pll[0] = devm_clk_hw_register_fixed_factor(priv->dev, "pll0_out",
> >> - "osc", 0, 125, 3);
> >> - if (IS_ERR(priv->pll[0]))
> >> - return PTR_ERR(priv->pll[0]);
> >> + priv->pll[0] = devm_clk_hw_register_fixed_factor(priv->dev, "pll0_out",
> >> + "osc", 0, 125, 3);
> >> + if (IS_ERR(priv->pll[0]))
> >> + return PTR_ERR(priv->pll[0]);
> >>
> >> /* 24MHz -> 1066.0MHz */
> >> - priv->pll[1] = devm_clk_hw_register_fixed_factor(priv->dev, "pll1_out",
> >> - "osc", 0, 533, 12);
> >> - if (IS_ERR(priv->pll[1]))
> >> - return PTR_ERR(priv->pll[1]);
> >> + priv->pll[1] = devm_clk_hw_register_fixed_factor(priv->dev, "pll1_out",
> >> + "osc", 0, 533, 12);
> >> + if (IS_ERR(priv->pll[1]))
> >> + return PTR_ERR(priv->pll[1]);
> >>
> >> /* 24MHz -> 1188.0MHz */
> >> - priv->pll[2] = devm_clk_hw_register_fixed_factor(priv->dev, "pll2_out",
> >> - "osc", 0, 99, 2);
> >> - if (IS_ERR(priv->pll[2]))
> >> - return PTR_ERR(priv->pll[2]);
> >> + priv->pll[2] = devm_clk_hw_register_fixed_factor(priv->dev, "pll2_out",
> >> + "osc", 0, 99, 2);
> >> + if (IS_ERR(priv->pll[2]))
> >> + return PTR_ERR(priv->pll[2]);
> >> + }
> >>
> >> for (idx = 0; idx < JH7110_SYSCLK_END; idx++) {
> >> u32 max = jh7110_sysclk_data[idx].max;
> >> @@ -462,8 +465,14 @@ static int __init jh7110_syscrg_probe(struct platform_device *pdev)
> >> parents[i].fw_name = "tdm_ext";
> >> else if (pidx == JH7110_SYSCLK_MCLK_EXT)
> >> parents[i].fw_name = "mclk_ext";
> >> - else
> >> + else if (use_fixed_pll)
> >
> > else if (priv->pll[0])
>
> Will change.
>
> >
> >> parents[i].hw = priv->pll[pidx - JH7110_SYSCLK_PLL0_OUT];
> >> + else if (pidx == JH7110_SYSCLK_PLL0_OUT)
> >> + parents[i].fw_name = "pll0_out";
> >> + else if (pidx == JH7110_SYSCLK_PLL1_OUT)
> >> + parents[i].fw_name = "pll1_out";
> >> + else if (pidx == JH7110_SYSCLK_PLL2_OUT)
> >> + parents[i].fw_name = "pll2_out";
> >> }
> >>
> >> clk->hw.init = &init;
> >> --
> >> 2.25.1
> >>
>
> Best regards,
> Xingyu Wu