Quoting Charles Keepax (2018-12-21 05:50:37)

On Thu, Nov 29, 2018 at 11:53:51PM -0800, Stephen Boyd wrote:

quoted

Quoting Charles Keepax (2018-11-20 06:16:30)

quoted

quoted

+               break;
+       default:
+               dev_err(priv->dev, "Unknown Lochnagar type: %d\n", priv->type);
+               return -EINVAL;
+       }
+
+       if (!priv->parents)
+               return -ENOMEM;
+
+       for (i = 0; i < priv->nparents; i++) {
+               j = of_property_match_string(np, "clock-names",
+                                            priv->parents[i]);
+               if (j >= 0)
+                       priv->parents[i] = of_clk_get_parent_name(np, j);

Isn't this of_clk_parent_fill()? But there are holes or something?

I guess rather than a fill, this is perhaps more of a name and
replace. I could make this a core function if you prefer? I think
there are a couple of other drivers that could also use it,
although might be worth doing that as a separate series rather
than holding this one up.

No worries. I'm actively changing the parent_names design anyway so this
is fine.

quoted

quoted

+static struct platform_driver lochnagar_clk_driver = {
+       .driver = {
+               .name = "lochnagar-clk",
+               .of_match_table = of_match_ptr(lochnagar_of_match),

I suspect of_match_ptr() makes the build complain about unused match
table when CONFIG_OF=N. Can you try building it that way?

The driver depends on the MFD which in turn depends on CONFIG_OF
so that shouldn't be a problem.

Ok. Then it's really not needed and it's better to drop the usage of
of_match_ptr() then.