On Thu, Jul 25, 2019 at 10:50 PM Jerome Brunet [off-list ref] wrote:

On Mon 22 Jul 2019 at 11:54, Alexandre Mergnat [off-list ref] wrote:

quoted

@@ -1592,13 +1737,29 @@ static struct clk_regmap gxbb_vid_pll_div = {
      .hw.init = &(struct clk_init_data) {
              .name = "vid_pll_div",
              .ops = &meson_vid_pll_div_ro_ops,
-             .parent_names = (const char *[]){ "hdmi_pll" },
+             .parent_data = &(const struct clk_parent_data) {
+                     /*
+                      * This clock is declared here for GXL and GXBB SoC, so
+                      * we must use string name to set this parent to avoid
+                      * pointer issue.
+                      */

I don't really get the issue with this comment.

How about:

/*
 * Note:
 * gxl and gxbb have different hdmi_plls (with different struct clk_hw).
 * We fallback to the global naming string mechanism so vid_pll_div picks
 * up the appropriate one.
 */

If you're sticking to global names for now, you could just skip converting
this clock altogether. I suspect .parent_names will be around for some time.

On the other hand, if you really want to get rid of global clock name based
parenting, you could use clk_hw pointers, and have the probe function fix
up this one based on the compatible string. That's what I did.

Just my two cents.

ChenYu

quoted

+                     .name = "hdmi_pll",
+                     .index = -1,
+             },
              .num_parents = 1,
              .flags = CLK_SET_RATE_PARENT | CLK_GET_RATE_NOCACHE,
      },
 };

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel