> -----Original Message-----
> From: Hilman, Kevin
> Sent: Friday, March 04, 2011 6:53 AM
> To: DebBarma, Tarun Kanti
> Cc: linux-omap@vger.kernel.org; Basak, Partha
> Subject: Re: [PATCH v11 7/8] OMAP: dmtimer: pm_runtime support
> 
> Tarun Kanti DebBarma <tarun.ka...@ti.com> writes:
> 
> > Add pm_runtime support to dmtimer. Since dmtimer is used during
> > early boot before pm_runtime is initialized completely there are
> > provisions to enable/disable clocks directly in the code during
> > early boot.
> 
> I'm still not crazy about the duplicate logic (both early & normal) in
> all the enable/disable functions.
> 
> As I've suggested in the past, why not just do a clk_get, clk_enable in
> when the early timers are initialized, then do a clk_disable, clk_put()
> as soon as the "normal" device is ready and PM runtime is enabled.
> 
> That will greatly simplify the code and eliminate the unnecessary checks
> for ->is_early_device which will always be false except for in early
> boot (when these functions are not likely to be called anyways.)
I will do the cleanup.

> 
> > Signed-off-by: Tarun Kanti DebBarma <tarun.ka...@ti.com>
> > [p-bas...@ti.com: added pm_runtime logic in probe()]
> > Signed-off-by: Partha Basak <p-bas...@ti.com>
> > Reviewed-by: Varadarajan, Charulatha <ch...@ti.com>
> > Acked-by: Cousson, Benoit <b-cous...@ti.com>
> > ---
> >  arch/arm/plat-omap/dmtimer.c |   63
> ++++++++++++++++++++++++++++++++++++------
> >  1 files changed, 54 insertions(+), 9 deletions(-)
> >
> > diff --git a/arch/arm/plat-omap/dmtimer.c b/arch/arm/plat-omap/dmtimer.c
> > index 53f5205..fcac422 100644
> > --- a/arch/arm/plat-omap/dmtimer.c
> > +++ b/arch/arm/plat-omap/dmtimer.c
> > @@ -39,6 +39,7 @@
> >  #include <linux/delay.h>
> >  #include <linux/io.h>
> >  #include <linux/slab.h>
> > +#include <linux/pm_runtime.h>
> >  #include <linux/err.h>
> >  #include <linux/platform_device.h>
> >  #include <plat/dmtimer.h>
> > @@ -211,13 +212,16 @@ static void omap_dm_timer_prepare(struct
> omap_dm_timer *timer)
> >  {
> >     struct dmtimer_platform_data *pdata = timer->pdev->dev.platform_data;
> >
> > -   timer->fclk = clk_get(&timer->pdev->dev, "fck");
> > -   if (WARN_ON_ONCE(IS_ERR_OR_NULL(timer->fclk))) {
> > -           dev_err(&timer->pdev->dev, ": No fclk handle.\n");
> > -           return;
> > +   if (!pdata->is_omap16xx) {
> 
> Is this 'is_omap16xx' really needed here?  If the clk_get fails, set
> timer->fclk to NULL, and any other users of fclk should just check the
> validity of timer->fclk.
OK.

> 
> > +           timer->fclk = clk_get(&timer->pdev->dev, "fck");
> > +           if (WARN_ON_ONCE(IS_ERR_OR_NULL(timer->fclk))) {
> > +                   dev_err(&timer->pdev->dev, ": No fclk handle.\n");
> > +                   return;
> > +           }
> >     }
> >
> > -   omap_dm_timer_enable(timer);
> > +   if (!pdata->is_omap16xx)
> > +           omap_dm_timer_enable(timer);
> 
> I don't think this if is needed, as runtime PM calls will essentially be
> nops on OMAP1 anyways.
Right, I will remove the check.

> 
> >     if (pdata->dm_timer_reset)
> >             pdata->dm_timer_reset(timer);
> > @@ -292,10 +296,22 @@ EXPORT_SYMBOL_GPL(omap_dm_timer_free);
> >
> >  void omap_dm_timer_enable(struct omap_dm_timer *timer)
> >  {
> > +   struct dmtimer_platform_data *pdata = timer->pdev->dev.platform_data;
> > +
> >     if (timer->enabled)
> >             return;
> >
> > -   clk_enable(timer->fclk);
> > +   if (unlikely(pdata->is_early_init)) {
> > +           clk_enable(timer->fclk);
> > +           timer->enabled = 1;
> > +           return;
> > +   }
> > +
> > +   if (pm_runtime_get_sync(&timer->pdev->dev) < 0) {
> > +           dev_err(&timer->pdev->dev, "%s: pm_runtime_get_sync() FAILED\n",
> > +                   __func__);
> > +           return;
> > +   }
> >
> >     timer->enabled = 1;
> >  }
> 
> Taking my above suggestion, this can be simplified to:
> 
> void omap_dm_timer_enable(struct omap_dm_timer *timer)
> {
>       pm_runtime_get_sync(&timer->pdev->dev);
> }
> 
> Note that with runtime PM, the ->enabled flag is no longer needed
> either, as you can simply check pm_runtime_suspended(dev) instead.
Sounds good!

> 
> > @@ -303,10 +319,22 @@ EXPORT_SYMBOL_GPL(omap_dm_timer_enable);
> >
> >  void omap_dm_timer_disable(struct omap_dm_timer *timer)
> >  {
> > +   struct dmtimer_platform_data *pdata = timer->pdev->dev.platform_data;
> > +
> >     if (!timer->enabled)
> >             return;
> >
> > -   clk_disable(timer->fclk);
> > +   if (unlikely(pdata->is_early_init)) {
> > +           clk_disable(timer->fclk);
> > +           timer->enabled = 0;
> > +           return;
> > +   }
> > +
> > +   if (pm_runtime_put_sync(&timer->pdev->dev) < 0) {
> > +           dev_err(&timer->pdev->dev, "%s: pm_runtime_put_sync() FAILED\n",
> > +                   __func__);
> > +           return;
> > +   }
> >
> >     timer->enabled = 0;
> >  }
> 
> Likewise, this becomes
> 
> void omap_dm_timer_disable(struct omap_dm_timer *timer)
> {
>       pm_runtime_put_sync(&timer->pdev->dev);
> }
Yes.

> 
> > @@ -425,10 +453,12 @@ int omap_dm_timer_set_source(struct omap_dm_timer
> *timer, int source)
> >     if (source < 0 || source >= 3)
> >             return -EINVAL;
> >
> > -   omap_dm_timer_disable(timer);
> > +   if (!pdata->is_omap16xx)
> 
> drop the if
Yes.

> 
> > +           omap_dm_timer_disable(timer);
> >     /* change the timer clock source */
> >     ret = pdata->set_timer_src(timer->pdev, source);
> > -   omap_dm_timer_enable(timer);
> > +   if (!pdata->is_omap16xx)
> 
> and this one,
> 
> and with that, there are no users of is_omap16xx in this patch.
Right!

> 
> Kevin
> 
> 
> > +           omap_dm_timer_enable(timer);
> >
> >     /*
> >      * When the functional clock disappears, too quick writes seem
> > @@ -600,11 +630,26 @@ static int __devinit omap_dm_timer_probe(struct
> platform_device *pdev)
> >             return -ENODEV;
> >     }
> >
> > +   /*
> > +    * OMAP2+
> > +    * Early timers are already registered and in list.
> > +    * What we need to do during second phase of probe
> > +    * is to assign the newly allocated/configured pdev
> > +    * to timer->pdev. We also call pm_runtime_enable()
> > +    * for each device because it could not be called
> > +    * during early boot because pm_runtime framework
> > +    * was not yet up and running.
> > +    */
> >     spin_lock_irqsave(&dm_timer_lock, flags);
> >     list_for_each_entry(timer, &omap_timer_list, node)
> >             if (!pdata->is_early_init && timer->id == pdev->id) {
> >                     timer->pdev = pdev;
> >                     spin_unlock_irqrestore(&dm_timer_lock, flags);
> > +                   pm_runtime_enable(&pdev->dev);
> > +                   /* update PM runtime for active early timers */
> > +                   if (timer->reserved)
> > +                           pm_runtime_get_sync(&pdev->dev);
> > +
> >                     dev_dbg(&pdev->dev, "Regular Probed\n");
> >                     return 0;
> >             }
--
To unsubscribe from this list: send the line "unsubscribe linux-omap" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to