[PATCH v3 06/15] ARM: mxs: Add timer support

[u.kleine-koenig@pengutronix.de (Uwe Kleine-König)]

On Wed, Dec 08, 2010 at 01:58:17PM +0800, Shawn Guo wrote:
> Hi Uwe,
> 
> 2010/12/8 Uwe Kleine-K?nig <u.kleine-koenig@pengutronix.de>:
> > On Wed, Dec 08, 2010 at 12:31:56AM +0800, Shawn Guo wrote:
> [...]
> >> +static int timrotv2_set_next_event(unsigned long evt,
> >> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? struct clock_event_device *dev)
> >> +{
> >> + ? ? unsigned long match;
> >> +
> >> + ? ? /* timrot decrements the count */
> >> + ? ? match = __raw_readl(timer_base + HW_TIMROT_RUNNING_COUNTn(0))
> >> + ? ? ? ? ? ? ? ? ? ? - evt;
> >> + ? ? __raw_writel(match, timer_base + HW_TIMROT_MATCH_COUNTn(0));
> >> +
> >> + ? ? return (int)(match - __raw_readl(timer_base +
> >> + ? ? ? ? ? ? ? ? ? ? HW_TIMROT_RUNNING_COUNTn(0))) > 0 ? -ETIME : 0;
> > Wouldn't it be easier to use the same approach as in timrotv1? ?(i.e.
> > use two timers and then having a simpler set_next_event.)
> >
> Using two timrotv1 timers is not to make code look simpler. Instead,
> we have to use two, because timrotv1 has no match function.  One most
> significant improvement of v2 over v1 is adding match function, so
> that one v2 timer can serve clock_event and clocksource without
> problem.  I would keep the implementation to utilize v2 capability,
> save timer resource and show the difference between v1 and v2.
Is there an advantage of keeping one additional timer unused?  IMHO it
doesn't make much sense to use the match support just because it exists
if it doesn't make things easier.  Here the main impact is that it
complicates the set_next_event callback.

> >> +}
> >> +
> >> +static irqreturn_t mxs_timer_interrupt(int irq, void *dev_id)
> >> +{
> >> + ? ? struct clock_event_device *evt = &clockevent_mxs;
> > I assume dev_id is &clockevent_mxs
> >
> The assumption is only true if we assign &clockevent_mxs to dev_id.
> So you want to see the following change?
> 
>  static irqreturn_t mxs_timer_interrupt(int irq, void *dev_id)
>  {
> -       struct clock_event_device *evt = &clockevent_mxs;
> +       struct clock_event_device *evt = dev_id;
This might save an instruction or two.

>         timrot_irq_acknowledge();
>         evt->event_handler(evt);
> 
>  static struct irqaction mxs_timer_irq = {
>         .name           = "MXS Timer Tick",
> +       .dev_id         = &clockevent_mxs,
>         .flags          = IRQF_TIMER | IRQF_IRQPOLL,
>         .handler        = mxs_timer_interrupt,
>  };
This looks more correct at least (IMHO).

Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-K?nig            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |