On 06/01/11 12:15, Jamie Iles wrote:
> On Thu, Jan 06, 2011 at 10:51:10AM +0100, John Crispin wrote:
>
>> On 06/01/11 00:49, Jamie Iles wrote:
>>
>>> I think you need a clk_put() here too to balance the clk_get() in the
>>> probe method so you'll need to keep a reference to the clk.
>>>
>>>
>>>
>> Hi Jamie,
>>
>> i will fold your suggestions into the series.
>>
>> the clk.c/h implementation on the lantiq target is very simple. it only
>> allows to read the static rates of the 3 clocks. clk_put is implemented
>> as follows
>>
>> void
>> clk_put(struct clk *clk)
>> {
>> /* not used */
>> }
>> EXPORT_SYMBOL(clk_put);
>>
>> so in theory you are right and we should call that function, however as
>> it is only a stub and the driver is only used by the lantiq target i
>> think it is save to leave out the clk_put(); call. we could however put
>> a commet in the code to make this clear (same as with the clk_enable()
>> not being needed as the clocks are always running)
>>
> Could that ever change for future devices that share the same watchdog
> block? If so, then adding in that clk_put() and clk_enable() might be
> worth it as it doesn't cost much.
>
> Jamie
>
>
agreed, i will add both calls.
