Re: [PATCH v2 4/4] clocksource: dw_apb_timer_of: use clocksource_of_init

["Heiko Stübner" <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>]

Am Dienstag, 4. Juni 2013, 16:21:20 schrieb Jamie Iles:
> Hi Heiko,
> 
> On Tue, Jun 04, 2013 at 11:38:42AM +0200, Heiko Stübner wrote:
> > dw_apb_timer_init used to search the devicetree for matching timer
> > devices, making calls to it from board files necessary.
> > 
> > Change the dw_apb_timer_init to work with CLOCKSOURCE_OF_DECLARE.
> > With this change the function gets called once for each timer node
> > and tracks these number of calls to attach clockevent and clocksource
> > devices to the nodes.
> > 
> > Also remove the calls to dw_apb_timer_init from all previous users, as
> > clocksource_of_init is the default for init_time now.
> > 
> > Tested on the upcoming rk3066 code.
> > 
> > Signed-off-by: Heiko Stuebner <heiko-4mtYJXux2i+zQB+pC5nmwQ@public.gmane.org>
> > Acked-by: Rob Herring <rob.herring-bsGFqQB8/DxBDgjK7y7TUQ@public.gmane.org>
> > Acked-by: Arnd Bergmann <arnd-r2nGTMty4D4@public.gmane.org>
> > ---
> 
> ...
> 
> > -void __init dw_apb_timer_init(void)
> > +static int num_called;
> > +static void __init dw_apb_timer_init(struct device_node *timer)
> > 
> >  {
> > 
> > -	struct device_node *event_timer, *source_timer;
> > -
> > -	event_timer = of_find_matching_node(NULL, osctimer_ids);
> > -	if (!event_timer)
> > -		panic("No timer for clockevent");
> > -	add_clockevent(event_timer);
> > -
> > -	source_timer = of_find_matching_node(event_timer, osctimer_ids);
> > -	if (!source_timer)
> > -		panic("No timer for clocksource");
> > -	add_clocksource(source_timer);
> > -
> > -	of_node_put(source_timer);
> > +	switch (num_called) {
> > +	case 0:
> > +		pr_debug("%s: found clockevent timer\n", __func__);
> > +		add_clockevent(timer);
> > +		of_node_put(timer);
> > +		break;
> > +	case 1:
> > +		pr_debug("%s: found clocksource timer\n", __func__);
> > +		add_clocksource(timer);
> > +		of_node_put(timer);
> > +		init_sched_clock();
> > +		break;
> > +	default:
> > +		break;
> > +	}
> > 
> > -	init_sched_clock();
> > +	num_called++;
> > 
> >  }
> > 
> > +CLOCKSOURCE_OF_DECLARE(pc3x2_timer, "picochip,pc3x2-timer",
> > dw_apb_timer_init); +CLOCKSOURCE_OF_DECLARE(apb_timer,
> > "snps,dw-apb-timer-osc", dw_apb_timer_init);
> 
> I think maybe we also want CLOCKSOURCE_OF_DECLARE() instances for the
> contents of sptimer_ids for completeness, otherwise looks good.

Hmm ... difficult.

As you can see, currently we call init_sched_clock when the second timer is 
found, which then tries to find the sptimer or uses the clocksource as 
sched_clock.

So to use CLOCKSOURCE_OF_DECLARE for the sptimer the init_sched_clock call 
would need to move to its init function. So on system without sptimer (like 
mine) this third call would never happen, and the system would be missing a 
sched_clock.

Or I just don't see the solution currently. In any case this could probably be 
handled by someone with the relevant hardware to test ;-) .


> Acked-by: Jamie Iles <jamie-wmLquQDDieKakBO8gow8eQ@public.gmane.org>

thanks