On Tue, Jan 10, 2023 at 02:31:40PM +0100, AngeloGioacchino Del Regno wrote:
> Il 30/12/22 06:12, Chen-Yu Tsai ha scritto:
> > On Fri, Dec 23, 2022 at 5:43 PM AngeloGioacchino Del Regno
> > <angelogioacchino.delregno@xxxxxxxxxxxxx> wrote:
> > >
> > > mtk_clk_simple_probe() is a function that registers mtk gate clocks
> > > and, if reset data is present, a reset controller and across all of
> > > the MTK clock drivers, such a function is duplicated many times:
> > > switch to the common mtk_clk_simple_probe() function for all of the
> > > clock drivers that are registering as platform drivers.
> > >
> > > Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@xxxxxxxxxxxxx>
> > > ---
> > > drivers/clk/mediatek/clk-mt2701-aud.c | 26 +++----
> > > drivers/clk/mediatek/clk-mt2701-eth.c | 34 +++------
> > > drivers/clk/mediatek/clk-mt2701-g3d.c | 56 +++-----------
> > > drivers/clk/mediatek/clk-mt2701-hif.c | 36 +++------
> > > drivers/clk/mediatek/clk-mt2712.c | 83 ++++++++-------------
> > > drivers/clk/mediatek/clk-mt6779.c | 42 ++++++-----
> > > drivers/clk/mediatek/clk-mt7622-aud.c | 49 +++----------
> > > drivers/clk/mediatek/clk-mt7622-eth.c | 82 ++++-----------------
> > > drivers/clk/mediatek/clk-mt7622-hif.c | 85 ++++-----------------
> > > drivers/clk/mediatek/clk-mt7629-hif.c | 85 ++++-----------------
> > > drivers/clk/mediatek/clk-mt8183-audio.c | 19 +++--
> > > drivers/clk/mediatek/clk-mt8183.c | 75 ++++++++-----------
> > > drivers/clk/mediatek/clk-mt8192-aud.c | 25 +++----
> > > drivers/clk/mediatek/clk-mt8192.c | 98 ++++++++-----------------
> > > 14 files changed, 236 insertions(+), 559 deletions(-)
> >
> > This looks mostly good, however ...
> >
> > > diff --git a/drivers/clk/mediatek/clk-mt2701-aud.c b/drivers/clk/mediatek/clk-mt2701-aud.c
> > > index ab13ab618fb5..1fd6d96b34dc 100644
> > > --- a/drivers/clk/mediatek/clk-mt2701-aud.c
> > > +++ b/drivers/clk/mediatek/clk-mt2701-aud.c
> > > @@ -76,6 +76,7 @@ static const struct mtk_gate_regs audio3_cg_regs = {
> > > };
> > >
> > > static const struct mtk_gate audio_clks[] = {
> > > + GATE_DUMMY(CLK_DUMMY, "aud_dummy"),
> > > /* AUDIO0 */
> > > GATE_AUDIO0(CLK_AUD_AFE, "audio_afe", "aud_intbus_sel", 2),
> > > GATE_AUDIO0(CLK_AUD_HDMI, "audio_hdmi", "audpll_sel", 20),
> > > @@ -138,29 +139,26 @@ static const struct mtk_gate audio_clks[] = {
> > > GATE_AUDIO3(CLK_AUD_MEM_ASRC5, "audio_mem_asrc5", "asm_h_sel", 14),
> > > };
> > >
> > > +static const struct mtk_clk_desc audio_desc = {
> > > + .clks = audio_clks,
> > > + .num_clks = ARRAY_SIZE(audio_clks),
> > > +};
> > > +
> > > static const struct of_device_id of_match_clk_mt2701_aud[] = {
> > > - { .compatible = "mediatek,mt2701-audsys", },
> > > - {}
> > > + { .compatible = "mediatek,mt2701-audsys", .data = &audio_desc },
> > > + { /* sentinel */ }
> > > };
> > >
> > > static int clk_mt2701_aud_probe(struct platform_device *pdev)
> > > {
> > > - struct clk_hw_onecell_data *clk_data;
> > > - struct device_node *node = pdev->dev.of_node;
> > > int r;
> > >
> > > - clk_data = mtk_alloc_clk_data(CLK_AUD_NR);
> > > -
> > > - mtk_clk_register_gates(node, audio_clks, ARRAY_SIZE(audio_clks),
> > > - clk_data, &pdev->dev);
> > > -
> > > - r = of_clk_add_hw_provider(node, of_clk_hw_onecell_get, clk_data);
> > > + r = mtk_clk_simple_probe(pdev);
> > > if (r) {
> > > dev_err(&pdev->dev,
> > > "could not register clock provider: %s: %d\n",
> > > pdev->name, r);
> > > -
> > > - goto err_clk_provider;
> > > + return r;
> > > }
> > >
> > > r = devm_of_platform_populate(&pdev->dev);
> > > @@ -170,13 +168,13 @@ static int clk_mt2701_aud_probe(struct platform_device *pdev)
> > > return 0;
> > >
> > > err_plat_populate:
> > > - of_clk_del_provider(node);
> > > -err_clk_provider:
> > > + mtk_clk_simple_remove(pdev);
> > > return r;
> > > }
> > >
> > > static struct platform_driver clk_mt2701_aud_drv = {
> > > .probe = clk_mt2701_aud_probe,
> > > + .remove = mtk_clk_simple_remove,
> >
> > I'm not a big fan of mixing devres and non-devres teardown code. Automatic
> > devres teardown happens after the remove callback returns, so in this
> > case you could have child devices being unregistered that touch clocks
> > or resets that have already been unregistered and freed in the remove
> > callback.
> >
>
> I'll add a custom remove function that calls of_platform_depopulate() and
> tears everything down manually in the correct order.
>
> Thanks for catching this one, I didn't notice at all!
>
> > > .driver = {
> > > .name = "clk-mt2701-aud",
> > > .of_match_table = of_match_clk_mt2701_aud,
> >
> > [...]
> >
> > > --- a/drivers/clk/mediatek/clk-mt2712.c
> > > +++ b/drivers/clk/mediatek/clk-mt2712.c
> >
> > [...]
> >
> > > @@ -1482,7 +1459,11 @@ static struct platform_driver clk_mt2712_drv = {
> > >
> > > static int __init clk_mt2712_init(void)
> > > {
> > > - return platform_driver_register(&clk_mt2712_drv);
> > > + int ret = platform_driver_register(&clk_mt2712_drv);
> > > +
> > > + if (ret)
> > > + return ret;
> > > + return platform_driver_register(&clk_mt2712_simple_drv);
> > > }
> > >
> > > arch_initcall(clk_mt2712_init);
> >
> > Would this get cleaned up even more? I.e. have just one driver left and
> > we could have the nice *_platform_driver() macros.
> >
>
> In the future, yes - granted that I find someone that can help with the testing,
> as I don't have any MT2712 hardware here.
>
> Not in this series though (please!).
Got it.
Maybe Miles has access to some EVBs, or knows someone who does.
ChenYu
