Re: [PATCH v6 10/22] clk: mediatek: Add MT8192 basic clocks support

From: Weiyi Lu
Date: Wed Feb 17 2021 - 21:00:28 EST


On Wed, 2021-02-10 at 13:46 +0100, Matthias Brugger wrote:
>
> On 22/12/2020 14:09, Weiyi Lu wrote:
> > Add MT8192 basic clock providers, include topckgen, apmixedsys,
> > infracfg and pericfg.
> >
> > Signed-off-by: Weiyi Lu <weiyi.lu@xxxxxxxxxxxx>
> > ---
> > drivers/clk/mediatek/Kconfig | 8 +
> > drivers/clk/mediatek/Makefile | 1 +
> > drivers/clk/mediatek/clk-mt8192.c | 1326 +++++++++++++++++++++++++++++++++++++
> > drivers/clk/mediatek/clk-mux.h | 15 +
> > 4 files changed, 1350 insertions(+)
> > create mode 100644 drivers/clk/mediatek/clk-mt8192.c
> >
>
> [...]
>
> > +static int clk_mt8192_apmixed_probe(struct platform_device *pdev)
> > +{
> > + struct clk_onecell_data *clk_data;
> > + struct device_node *node = pdev->dev.of_node;
> > + int r;
> > +
> > + clk_data = mtk_alloc_clk_data(CLK_APMIXED_NR_CLK);
> > + if (!clk_data)
> > + return -ENOMEM;
> > +
> > + mtk_clk_register_plls(node, plls, ARRAY_SIZE(plls), clk_data);
> > + r = mtk_clk_register_gates(node, apmixed_clks, ARRAY_SIZE(apmixed_clks), clk_data);
> > + if (r)
> > + return r;
> > +
> > + return of_clk_add_provider(node, of_clk_src_onecell_get, clk_data);
> > +}
> > +
> > +static const struct of_device_id of_match_clk_mt8192[] = {
> > + {
> > + .compatible = "mediatek,mt8192-apmixedsys",
> > + .data = clk_mt8192_apmixed_probe,
> > + }, {
> > + .compatible = "mediatek,mt8192-topckgen",
> > + .data = clk_mt8192_top_probe,
> > + }, {
> > + .compatible = "mediatek,mt8192-infracfg",
> > + .data = clk_mt8192_infra_probe,
> > + }, {
> > + .compatible = "mediatek,mt8192-pericfg",
> > + .data = clk_mt8192_peri_probe,
> > + }, {
> > + /* sentinel */
> > + }
> > +};
> > +
> > +static int clk_mt8192_probe(struct platform_device *pdev)
> > +{
> > + int (*clk_probe)(struct platform_device *pdev);
> > + int r;
> > +
> > + clk_probe = of_device_get_match_data(&pdev->dev);
> > + if (!clk_probe)
> > + return -EINVAL;
> > +
> > + r = clk_probe(pdev);
> > + if (r)
> > + dev_err(&pdev->dev, "could not register clock provider: %s: %d\n", pdev->name, r);
> > +
> > + return r;
> > +}
> > +
> > +static struct platform_driver clk_mt8192_drv = {
> > + .probe = clk_mt8192_probe,
> > + .driver = {
> > + .name = "clk-mt8192",
> > + .of_match_table = of_match_clk_mt8192,
> > + },
> > +};
> > +
> > +static int __init clk_mt8192_init(void)
> > +{
> > + return platform_driver_register(&clk_mt8192_drv);
> > +}
> > +
> > +arch_initcall(clk_mt8192_init);
>
> Do we really need all these clocks that early?
> Why don't we use CLK_OF_DECLARE_DRIVER() then and why do we initialize some
> clocks CLK_OF_DECLARE_DRIVER and other with arch_initcall?
>
> I know that this is in other drivers for MediaTek SoCs, but that does not mean
> it's the right approach.
>

I guess you mean CLK_OF_DECLARE() but not CLK_OF_DECLARE_DRIVER(), am I
correct?
I saw there had some discussion[1][2] about initializing these basic
clocks by CLK_OF_DECLARE() or the current implementation by arch_init().
Could I have more of your opinion on this discussion?
[1]
https://patchwork.kernel.org/project/linux-mediatek/patch/1454665050-37776-5-git-send-email-jamesjj.liao@xxxxxxxxxxxx/
[2]
https://patchwork.kernel.org/project/linux-mediatek/patch/1460621514-65191-5-git-send-email-jamesjj.liao@xxxxxxxxxxxx/

As to CLK_OF_DECLARE_DRIVER(), we have to initialize the clocks earlier
for timer(clocksource) driver by this way.

>
> > diff --git a/drivers/clk/mediatek/clk-mux.h b/drivers/clk/mediatek/clk-mux.h
> > index f5625f4..afbc7df 100644
> > --- a/drivers/clk/mediatek/clk-mux.h
> > +++ b/drivers/clk/mediatek/clk-mux.h
> > @@ -77,6 +77,21 @@ struct mtk_mux {
> > _width, _gate, _upd_ofs, _upd, \
> > CLK_SET_RATE_PARENT)
> >
> > +#define MUX_CLR_SET_UPD_FLAGS(_id, _name, _parents, _mux_ofs, \
> > + _mux_set_ofs, _mux_clr_ofs, _shift, _width, \
> > + _upd_ofs, _upd, _flags) \
> > + GATE_CLR_SET_UPD_FLAGS(_id, _name, _parents, _mux_ofs, \
> > + _mux_set_ofs, _mux_clr_ofs, _shift, _width, \
> > + 0, _upd_ofs, _upd, _flags, \
> > + mtk_mux_clr_set_upd_ops)
> > +
> > +#define MUX_CLR_SET_UPD(_id, _name, _parents, _mux_ofs, \
> > + _mux_set_ofs, _mux_clr_ofs, _shift, _width, \
> > + _upd_ofs, _upd) \
> > + MUX_CLR_SET_UPD_FLAGS(_id, _name, _parents, \
> > + _mux_ofs, _mux_set_ofs, _mux_clr_ofs, _shift, \
> > + _width, _upd_ofs, _upd, CLK_SET_RATE_PARENT)
> > +
>
> Why can't we do something like:
>
> #define MUX_CLR_SET_UPD(_id, _name, _parents, _mux_ofs, \
> _mux_set_ofs, _mux_clr_ofs, _shift, _width, \
> _upd_ofs, _upd) \
> GATE_CLR_SET_UPD_FLAGS(_id, _name, _parents, _mux_ofs, \
> _mux_set_ofs, _mux_clr_ofs, _shift, _width, \
> 0, _upd_ofs, _upd, CLK_SET_RATE_PARENT, \
> mtk_mux_clr_set_upd_ops)
>
> > struct clk *mtk_clk_register_mux(const struct mtk_mux *mux,
> > struct regmap *regmap,
> > spinlock_t *lock);
> >

It could be, so I'll fix in next patch.

>
> _______________________________________________
> Linux-mediatek mailing list
> Linux-mediatek@xxxxxxxxxxxxxxxxxxx
> http://lists.infradead.org/mailman/listinfo/linux-mediatek