These two are both mtk_composite arrays, one dependent on another, but
that's something that the clock framework is supposed to sort out and
anyway registering them separately isn't going to ease the framework's
job in checking dependencies.

Put the contents of top_adj_divs in top_muxes to join them together
and register them in one shot.

Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
---
 drivers/clk/mediatek/clk-mt8192.c | 13 ++-----------
 1 file changed, 2 insertions(+), 11 deletions(-)

diff --git a/drivers/clk/mediatek/clk-mt8192.c b/drivers/clk/mediatek/clk-mt8192.c
index ec9dc7fe848e..702770326286 100644
--- a/drivers/clk/mediatek/clk-mt8192.c
+++ b/drivers/clk/mediatek/clk-mt8192.c
@@ -698,9 +698,7 @@ static struct mtk_composite top_muxes[] = {
 	MUX(CLK_TOP_APLL_I2S7_M_SEL, "apll_i2s7_m_sel", apll_i2s_m_parents, 0x320, 23, 1),
 	MUX(CLK_TOP_APLL_I2S8_M_SEL, "apll_i2s8_m_sel", apll_i2s_m_parents, 0x320, 24, 1),
 	MUX(CLK_TOP_APLL_I2S9_M_SEL, "apll_i2s9_m_sel", apll_i2s_m_parents, 0x320, 25, 1),
-};
-
-static const struct mtk_composite top_adj_divs[] = {
+	/* APLL_DIV */
 	DIV_GATE(CLK_TOP_APLL12_DIV0, "apll12_div0", "apll_i2s0_m_sel", 0x320, 0, 0x328, 8, 0),
 	DIV_GATE(CLK_TOP_APLL12_DIV1, "apll12_div1", "apll_i2s1_m_sel", 0x320, 1, 0x328, 8, 8),
 	DIV_GATE(CLK_TOP_APLL12_DIV2, "apll12_div2", "apll_i2s2_m_sel", 0x320, 2, 0x328, 8, 16),
@@ -1099,15 +1097,10 @@ static int clk_mt8192_top_probe(struct platform_device *pdev)
 	if (r)
 		goto unregister_muxes;
 
-	r = mtk_clk_register_composites(top_adj_divs, ARRAY_SIZE(top_adj_divs), base,
-					&mt8192_clk_lock, top_clk_data, &pdev->dev);
-	if (r)
-		goto unregister_top_composites;
-
 	r = mtk_clk_register_gates(node, top_clks, ARRAY_SIZE(top_clks),
 					    top_clk_data, &pdev->dev);
 	if (r)
-		goto unregister_adj_divs_composites;
+		goto unregister_top_composites;
 
 	r = clk_mt8192_reg_mfg_mux_notifier(&pdev->dev,
 					    top_clk_data->hws[CLK_TOP_MFG_PLL_SEL]->clk);
@@ -1119,8 +1112,6 @@ static int clk_mt8192_top_probe(struct platform_device *pdev)
 
 unregister_gates:
 	mtk_clk_unregister_gates(top_clks, ARRAY_SIZE(top_clks), top_clk_data);
-unregister_adj_divs_composites:
-	mtk_clk_unregister_composites(top_adj_divs, ARRAY_SIZE(top_adj_divs), top_clk_data);
 unregister_top_composites:
 	mtk_clk_unregister_composites(top_muxes, ARRAY_SIZE(top_muxes), top_clk_data);
 unregister_muxes:
-- 
2.39.0

On Fri, Dec 23, 2022 at 5:43 PM AngeloGioacchino Del Regno
<angelogioacchino.delregno@collabora.com> wrote:
>
> These two are both mtk_composite arrays, one dependent on another, but
> that's something that the clock framework is supposed to sort out and
> anyway registering them separately isn't going to ease the framework's
> job in checking dependencies.
>
> Put the contents of top_adj_divs in top_muxes to join them together
> and register them in one shot.
>
> Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
> ---
>  drivers/clk/mediatek/clk-mt8192.c | 13 ++-----------
>  1 file changed, 2 insertions(+), 11 deletions(-)
>
> diff --git a/drivers/clk/mediatek/clk-mt8192.c b/drivers/clk/mediatek/clk-mt8192.c
> index ec9dc7fe848e..702770326286 100644
> --- a/drivers/clk/mediatek/clk-mt8192.c
> +++ b/drivers/clk/mediatek/clk-mt8192.c
> @@ -698,9 +698,7 @@ static struct mtk_composite top_muxes[] = {
>         MUX(CLK_TOP_APLL_I2S7_M_SEL, "apll_i2s7_m_sel", apll_i2s_m_parents, 0x320, 23, 1),
>         MUX(CLK_TOP_APLL_I2S8_M_SEL, "apll_i2s8_m_sel", apll_i2s_m_parents, 0x320, 24, 1),
>         MUX(CLK_TOP_APLL_I2S9_M_SEL, "apll_i2s9_m_sel", apll_i2s_m_parents, 0x320, 25, 1),
> -};
> -
> -static const struct mtk_composite top_adj_divs[] = {
> +       /* APLL_DIV */

This would be CLK_AUDDIV_2 ~ 4 actually. 4 dividers per register.

Otherwise,

Reviewed-by: Chen-Yu Tsai <wenst@chromium.org>

Hi,

> These two are both mtk_composite arrays, one dependent on another, but
> that's something that the clock framework is supposed to sort out and
> anyway registering them separately isn't going to ease the framework's
> job in checking dependencies.
> 
> Put the contents of top_adj_divs in top_muxes to join them together
> and register them in one shot.
> 

In mt8192, we can join top_adj_divs and top_muxes:

mtk_clk_register_composites(top_muxes, ARRAY_SIZE(top_muxes), base, &mt8192_clk_lock,
		top_clk_data);
mtk_clk_register_composites(top_adj_divs, ARRAY_SIZE(top_adj_divs), base, &mt8192_clk_lock,
		top_clk_data);

However, there are other top_adj_divs[] and top_muxes[] in different types so 
we cannot join them.

For example:
in drivers/clk/mediatek/clk-mt8167.c:mtk_topckgen_init():

mtk_clk_register_composites(top_muxes, ARRAY_SIZE(top_muxes), base,
		&mt8167_clk_lock, clk_data);
mtk_clk_register_dividers(top_adj_divs, ARRAY_SIZE(top_adj_divs),
		base, &mt8167_clk_lock, clk_data);

So we can join top_adj_divs and top_muxes in some platforms, but we
cannot join top_adj_divs and top_muxes in some other platforms.

I'm afraid that this will confuses people.


thanks,
Miles

On Wed, Dec 28, 2022 at 4:31 PM Miles Chen <miles.chen@mediatek.com> wrote:
>
> Hi,
>
> > These two are both mtk_composite arrays, one dependent on another, but
> > that's something that the clock framework is supposed to sort out and
> > anyway registering them separately isn't going to ease the framework's
> > job in checking dependencies.
> >
> > Put the contents of top_adj_divs in top_muxes to join them together
> > and register them in one shot.
> >
>
> In mt8192, we can join top_adj_divs and top_muxes:
>
> mtk_clk_register_composites(top_muxes, ARRAY_SIZE(top_muxes), base, &mt8192_clk_lock,
>                 top_clk_data);
> mtk_clk_register_composites(top_adj_divs, ARRAY_SIZE(top_adj_divs), base, &mt8192_clk_lock,
>                 top_clk_data);
>
> However, there are other top_adj_divs[] and top_muxes[] in different types so
> we cannot join them.
>
> For example:
> in drivers/clk/mediatek/clk-mt8167.c:mtk_topckgen_init():
>
> mtk_clk_register_composites(top_muxes, ARRAY_SIZE(top_muxes), base,
>                 &mt8167_clk_lock, clk_data);
> mtk_clk_register_dividers(top_adj_divs, ARRAY_SIZE(top_adj_divs),
>                 base, &mt8167_clk_lock, clk_data);
>
> So we can join top_adj_divs and top_muxes in some platforms, but we
> cannot join top_adj_divs and top_muxes in some other platforms.
>
> I'm afraid that this will confuses people.

I think the confusion comes from the macro names. It's not exactly clear that
DIV_GATE is for composite clocks, while DIV_ADJ is for divider clocks.
Doubly so for the mux related types.

Either way, using the wrong macro or the wrong type will cause the compiler
to complain, so I think it's something we can live with. If it's still not
working out, maybe we should rethink the naming.

ChenYu