On 05/09/2016 10:19 PM, Javier Martinez Canillas wrote:

Hello Krzysztof,

On 05/08/2016 03:05 PM, Krzysztof Kozlowski wrote:

quoted

Just like clock driver for Exynos542x/5800, provide the fixed clock here
so the clock bindings and their consumers would be consistent and
similar.

However a clock named "fin_pll" is already provided by generic
fixed-clock and it is both referenced in the clock driver (by name) and
in DT (by phandle). To make the transition smooth, first introduce the
new external fixed clock here under temporary, different name and switch
internal users to it.

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
---

[snip]

quoted

 
+/* Same as in Exynos5420 */
+static const struct of_device_id ext_clk_match[] __initconst = {
+	{ .compatible = "samsung,exynos5420-oscclk", .data = (void *)0, },

Since using designated initializers, I was about to say that there is no need
to explicitly set .data to 0 since omitted fields are implicitly initialized...

quoted

+	{ },
+};
+
 /* register exynos5410 clocks */
 static void __init exynos5410_clk_init(struct device_node *np)
 {

@@ -192,6 +204,10 @@ static void __init exynos5410_clk_init(struct device_node *np)
 
 	ctx = samsung_clk_init(np, reg_base, CLK_NR_CLKS);
 
+	samsung_clk_of_register_fixed_ext(ctx, exynos5x_fixed_rate_ext_clks,
+			ARRAY_SIZE(exynos5x_fixed_rate_ext_clks),
+			ext_clk_match);
+

... but then I noticed that .data is used as exynos5x_fixed_rate_ext_clks
array index in samsung_clk_of_register_fixed_ext(). So makes sense to set
it explicitly to make the intent clear.

Yes, that is the intention here. The '(void *)0' is a hint that clock
needs frequency from DT.

Thanks for review,
Krzysztof