Re: [PATCH 03/18] clk: meson: migrate a1 clock drivers out of hw_onecell_data to drop NR_CLKS

[Neil Armstrong]

Hi,

On 09/06/2023 13:48, Dmitry Rokosov wrote:
> Hello Neil,
>
> On Thu, Jun 08, 2023 at 02:53:50PM +0200, Neil Armstrong wrote:
> > On 08/06/2023 14:45, Jerome Brunet wrote:
> > > > +struct meson_a1_pll_clks {
> > > > +	struct clk_hw **hw_clks;
> > > > +	unsigned int hw_clk_num;
> > > > +};
> > > > +
> > > > +static struct meson_a1_pll_clks a1_pll_clks = {
> > > > +	.hw_clks = a1_pll_hw_clks,
> > > > +	.hw_clk_num = ARRAY_SIZE(a1_pll_hw_clks),
> > > > +};
> > > > +
> > > > +static struct clk_hw *meson_a1_pll_hw_get(struct of_phandle_args *clkspec, void *clk_data)
> > > > +{
> > > > +	const struct meson_a1_pll_clks *data = clk_data;
> > > > +	unsigned int idx = clkspec->args[0];
> > > > +
> > > > +	if (idx >= data->hw_clk_num) {
> > > > +		pr_err("%s: invalid index %u\n", __func__, idx);
> > > > +		return ERR_PTR(-EINVAL);
> > > > +	}
> > > > +
> > > > +	return data->hw_clks[idx];
> > > > +}
> > >
> > > I'd prefer to have a single struct type and and single custom
> > > callback for the different SoC please.
> >
> > Sure, I've written a common code for that, but I have a hard time finding
> > a proper naming for it... so I choosed meson-clkc since it could have
> > more common helper code for duplicated code over the clk driver:
> >
> > ===================================><============================================================================
> > diff --git a/drivers/clk/meson/Kconfig b/drivers/clk/meson/Kconfig
> > index 8ce846fdbe43..9070dcfd9e71 100644
> > --- a/drivers/clk/meson/Kconfig
> > +++ b/drivers/clk/meson/Kconfig
> > @@ -30,6 +30,9 @@ config COMMON_CLK_MESON_VID_PLL_DIV
> >   	tristate
> >   	select COMMON_CLK_MESON_REGMAP
> >
> > +config COMMON_CLK_MESON_CLKC
> > +	tristate
> > +
> >   config COMMON_CLK_MESON_AO_CLKC
> >   	tristate
> >   	select COMMON_CLK_MESON_REGMAP
> > diff --git a/drivers/clk/meson/Makefile b/drivers/clk/meson/Makefile
> > index d5288662881d..13c6db466986 100644
> > --- a/drivers/clk/meson/Makefile
> > +++ b/drivers/clk/meson/Makefile
> > @@ -1,6 +1,7 @@
> >   # SPDX-License-Identifier: GPL-2.0-only
> >   # Amlogic clock drivers
> >
> > +obj-$(CONFIG_COMMON_CLK_MESON_CLKC) += meson-clkc.o
> >   obj-$(CONFIG_COMMON_CLK_MESON_AO_CLKC) += meson-aoclk.o
> >   obj-$(CONFIG_COMMON_CLK_MESON_CPU_DYNDIV) += clk-cpu-dyndiv.o
> >   obj-$(CONFIG_COMMON_CLK_MESON_DUALDIV) += clk-dualdiv.o
> > diff --git a/drivers/clk/meson/meson-clkc.c b/drivers/clk/meson/meson-clkc.c
> > new file mode 100644
> > index 000000000000..fa98b9d09011
> > --- /dev/null
> > +++ b/drivers/clk/meson/meson-clkc.c
> > @@ -0,0 +1,25 @@
> > +// SPDX-License-Identifier: GPL-2.0+
> > +/*
> > + * Copyright (c) 2023 Neil Armstrong <neil.armstrong@xxxxxxxxxx>
> > + */
> > +
> > +#include <linux/of_device.h>
> > +#include <linux/clk-provider.h>
> > +#include <linux/module.h>
> > +#include "meson-clkc.h"
> > +
> > +struct clk_hw *meson_clk_hw_get(struct of_phandle_args *clkspec, void *clk_hw_data)
> > +{
> > +	const struct meson_clk_hw_data *data = clk_hw_data;
> > +	unsigned int idx = clkspec->args[0];
> > +
> > +	if (idx >= data->num) {
> > +		pr_err("%s: invalid index %u\n", __func__, idx);
> > +		return ERR_PTR(-EINVAL);
> > +	}
> > +
> > +	return data->hws[idx];
> > +}
> > +EXPORT_SYMBOL_GPL(meson_clk_hw_get);
> > +
> > +MODULE_LICENSE("GPL v2");
> > diff --git a/drivers/clk/meson/meson-clkc.h b/drivers/clk/meson/meson-clkc.h
> > new file mode 100644
> > index 000000000000..e3bad2aa17eb
> > --- /dev/null
> > +++ b/drivers/clk/meson/meson-clkc.h
> > @@ -0,0 +1,19 @@
> > +/* SPDX-License-Identifier: (GPL-2.0+ OR MIT) */
> > +/*
> > + * Copyright (c) 2023 Neil Armstrong <neil.armstrong@xxxxxxxxxx>
> > + */
> > +
> > +#ifndef __MESON_HW_CLKC_H__
> > +#define __MESON_HW_CLKC_H__
> > +
> > +#include <linux/of_device.h>
> > +#include <linux/clk-provider.h>
> > +
> > +struct meson_clk_hw_data {
> > +	struct clk_hw	**hws;
> > +	unsigned int	num;
> > +};
> > +
> > +struct clk_hw *meson_clk_hw_get(struct of_phandle_args *clkspec, void *clk_hw_data);
> > +
> > +#endif
> > ===================================><============================================================================
> >
> > If it's ok I'll send a v2 using this.
> >
> > Thanks,
> > Neil
>
> In addition, I propose consolidating the probe() routine of the a1
> clocks into a common part, which can be utilized for s4 and other
> similar clocks. This solution was presented in the early a1 review
> stages of this patch set.
>
> https://lore.kernel.org/linux-amlogic/20221201225703.6507-7-ddrokosov@xxxxxxxxxxxxxx/
>
> Maybe, it can be generalized for the all meson clock controller drivers.

Sure, the goal of that is to get rid of NR_CLKS & private IDS only,
then yes afterwards any common clkc function will be welcome !

Neil