On 07/19/2017 01:52 AM, Stephen Boyd wrote:

On 07/18, Vladimir Zapolskiy wrote:

quoted

On 07/18/2017 10:53 AM, gabriel.fernandez at st.com wrote:

quoted

From: Gabriel Fernandez <redacted>
 }
+EXPORT_SYMBOL_GPL(clk_gate_is_enabled);
 
 const struct clk_ops clk_gate_ops = {
 	.enable = clk_gate_enable,

diff --git a/include/linux/clk-provider.h b/include/linux/clk-provider.h
index c59c625..e9587ab 100644
--- a/include/linux/clk-provider.h
+++ b/include/linux/clk-provider.h

@@ -343,6 +343,7 @@ struct clk_hw *clk_hw_register_gate(struct device *dev, const char *name,
 		u8 clk_gate_flags, spinlock_t *lock);
 void clk_unregister_gate(struct clk *clk);
 void clk_hw_unregister_gate(struct clk_hw *hw);
+int clk_gate_is_enabled(struct clk_hw *hw);

Here the prefix does not reflect the type of its argument, it might be
acceptable for a veiled function, but it is not wanted for the exported
function. Something like clk_hw_gate_is_enabled() is expected here, but
again, let's firstly come to an agreement, that the export is needed.

I'd prefer clk_gate_is_enabled() as it's not a struct clk_hw_gate,
it's a struct clk_gate

Formally it's a 'struct clk_hw', and 'struct clk_hw_gate' does not exist.

and there isn't any requirement for function names to reflect the type
of the argument. That's what we have type analysis for.

Sure, a function name can be selected arbitrary, however because in this
particular case type analysis operates on 'struct clk_hw', it would fail
to separate 'struct clk_gate' from 'struct clk_divider', a little hint
in the naming may be helpful.

I don't insist on my preference, of course your acceptance is sufficient.

--
With best wishes,
Vladimir