[PATCH v3 0/7] ARM timer clock api support

[linux@arm.linux.org.uk (Russell King - ARM Linux)]

On Fri, Mar 11, 2011 at 07:48:48AM -0600, Rob Herring wrote:
> Russell,
>
> On 03/11/2011 02:16 AM, Russell King - ARM Linux wrote:
>> On Tue, Mar 08, 2011 at 05:34:30PM -0600, Rob Herring wrote:
>>> From: Rob Herring<rob.herring@calxeda.com>
>>>
>>> This patch series converts ARM sp804 timer and smp_twd timer to use the clock
>>> api.
>>
>> This is what I came up with - I've still not sorted these patches out,
>> the second is a superset of the first.
>
> Any comments on the smp_twd patch? Colin plans to base his cpufreq  
> patches on it.
>
>> +	clk = clk_get_sys(name, NULL);
>> +	if (IS_ERR(clk)) {
>> +		pr_err("sp804: %s clock not found, defaulting to 1MHz: %d\n",
>> +			name, ERR_PTR(clk));
>> +	} else {
>> +		int err = clk_enable(clk);
>> +		if (err != 0) {
>> +			pr_err("sp804: %s clock failed to enable: %d\n",
>> +				name, err);
>> +			clk_put(clk);
>> +		} else {
>> +			rate = clk_get_rate(clk);
>> +		}
>
> You previously said this should fully move to using the clock api.  
> Presumably, that meant no default freq of 1MHz.
>
> Also, my patch allows for clksrc and clkevent timers to have different  
> clocks.

So will mine.  Note that 'name' is passed in, which is the clocksource
name, and is also used for clk_sys_get().  The same thing will happen
with clockevents too.

> Currently, no h/w needs that, but it could happen. Even if they  
> are the same clk, you are introducing a requirement on the calling order  
> of sp804_clocksource_init and sp804_clockevents_init.

No I'm not, and I've no idea where you got that idea from.  You'll notice
that nothing to do with the clk is stored in a global variable.