Re: [PATCH v2 1/2] clk: qcom: add support for SM8350 DISPCC

[Stephen Boyd <sboyd@kernel.org>]

Quoting Jonathan Marek (2021-05-18 17:18:01)
> diff --git a/drivers/clk/qcom/dispcc-sm8250.c b/drivers/clk/qcom/dispcc-sm8250.c
> index de09cd5c209f..5f22a715e2f0 100644
> --- a/drivers/clk/qcom/dispcc-sm8250.c
> +++ b/drivers/clk/qcom/dispcc-sm8250.c
> @@ -36,6 +36,10 @@ static struct pll_vco vco_table[] = {
>         { 249600000, 2000000000, 0 },
>  };
>  
> +static struct pll_vco lucid_5lpe_vco[] = {

const

> +       { 249600000, 1750000000, 0 },
> +};
> +
>  static struct alpha_pll_config disp_cc_pll0_config = {
>         .l = 0x47,
>         .alpha = 0xE000,
> @@ -1039,6 +1043,7 @@ static const struct qcom_cc_desc disp_cc_sm8250_desc = {
>  static const struct of_device_id disp_cc_sm8250_match_table[] = {
>         { .compatible = "qcom,sm8150-dispcc" },
>         { .compatible = "qcom,sm8250-dispcc" },
> +       { .compatible = "qcom,sm8350-dispcc" },
>         { }
>  };
>  MODULE_DEVICE_TABLE(of, disp_cc_sm8250_match_table);
> @@ -1051,7 +1056,7 @@ static int disp_cc_sm8250_probe(struct platform_device *pdev)
>         if (IS_ERR(regmap))
>                 return PTR_ERR(regmap);
>  
> -       /* note: trion == lucid, except for the prepare() op */
> +       /* Apply differences for SM8150 and SM8350 */
>         BUILD_BUG_ON(CLK_ALPHA_PLL_TYPE_TRION != CLK_ALPHA_PLL_TYPE_LUCID);
>         if (of_device_is_compatible(pdev->dev.of_node, "qcom,sm8150-dispcc")) {
>                 disp_cc_pll0_config.config_ctl_hi_val = 0x00002267;
> @@ -1062,8 +1067,62 @@ static int disp_cc_sm8250_probe(struct platform_device *pdev)
>                 disp_cc_pll1_config.config_ctl_hi1_val = 0x00000024;
>                 disp_cc_pll1_config.user_ctl_hi1_val = 0x000000D0;
>                 disp_cc_pll1_init.ops = &clk_alpha_pll_trion_ops;
> +       } else if (of_device_is_compatible(pdev->dev.of_node, "qcom,sm8350-dispcc")) {
> +               static struct clk_rcg2* const rcgs[] = {

Please move this to global scope and give it a name.

> +                       &disp_cc_mdss_byte0_clk_src,
> +                       &disp_cc_mdss_byte1_clk_src,
> +                       &disp_cc_mdss_dp_aux1_clk_src,
> +                       &disp_cc_mdss_dp_aux_clk_src,
> +                       &disp_cc_mdss_dp_link1_clk_src,
> +                       &disp_cc_mdss_dp_link_clk_src,
> +                       &disp_cc_mdss_dp_pixel1_clk_src,
> +                       &disp_cc_mdss_dp_pixel2_clk_src,
> +                       &disp_cc_mdss_dp_pixel_clk_src,
> +                       &disp_cc_mdss_esc0_clk_src,
> +                       &disp_cc_mdss_mdp_clk_src,
> +                       &disp_cc_mdss_pclk0_clk_src,
> +                       &disp_cc_mdss_pclk1_clk_src,
> +                       &disp_cc_mdss_rot_clk_src,
> +                       &disp_cc_mdss_vsync_clk_src,
> +               };
> +               static struct clk_regmap_div* const divs[] = {

Move global.

> +                       &disp_cc_mdss_byte0_div_clk_src,
> +                       &disp_cc_mdss_byte1_div_clk_src,
> +                       &disp_cc_mdss_dp_link1_div_clk_src,
> +                       &disp_cc_mdss_dp_link_div_clk_src,
> +               };
> +               unsigned i;

int i? I doubt being unsigned helps.

> +               static bool offset_applied = false;
> +
> +               /* only apply the offsets once (in case of deferred probe) */
> +               if (!offset_applied) {

Maybe instead of doing this in probe it can be done when the driver is
added in module init? It would mean searching the DT for the compatible
string and then if it is present running the subtraction code, but at
least we would only do it once and the code would be thrown away after
init.

> +                       for (i = 0; i < ARRAY_SIZE(rcgs); i++)
> +                               rcgs[i]->cmd_rcgr -= 4;
> +
> +                       for (i = 0; i < ARRAY_SIZE(divs); i++) {
> +                               divs[i]->reg -= 4;
> +                               divs[i]->width = 4;
> +                       }
> +
> +                       disp_cc_mdss_ahb_clk.halt_reg -= 4;
> +                       disp_cc_mdss_ahb_clk.clkr.enable_reg -= 4;
> +
> +                       offset_applied = true;
> +               }
> +
> +               disp_cc_mdss_ahb_clk_src.cmd_rcgr = 0x22a0;
> +
> +               disp_cc_pll0_config.config_ctl_hi1_val = 0x2A9A699C;
> +               disp_cc_pll0_config.test_ctl_hi1_val = 0x01800000;
> +               disp_cc_pll0_init.ops = &clk_alpha_pll_lucid_5lpe_ops;
> +               disp_cc_pll0.vco_table = lucid_5lpe_vco;
> +               disp_cc_pll1_config.config_ctl_hi1_val = 0x2A9A699C;

Lowercase hex please.

> +               disp_cc_pll1_config.test_ctl_hi1_val = 0x01800000;
> +               disp_cc_pll1_init.ops = &clk_alpha_pll_lucid_5lpe_ops;
> +               disp_cc_pll1.vco_table = lucid_5lpe_vco;
>         }
>  
> +       /* note for SM8350: downstream lucid_5lpe configure differs slightly */

Is this a TODO?

>         clk_lucid_pll_configure(&disp_cc_pll0, regmap, &disp_cc_pll0_config);
>         clk_lucid_pll_configure(&disp_cc_pll1, regmap, &disp_cc_pll1_config);
>