Re: [PATCH v10 1/5] PM / Runtime: Allow accessing irq_safe if no PM_RUNTIME

From: Krzysztof Kozlowski
Date: Fri Nov 07 2014 - 03:06:55 EST


On czw, 2014-11-06 at 23:51 +0100, Rafael J. Wysocki wrote:
> On Thursday, November 06, 2014 09:36:46 AM Krzysztof Kozlowski wrote:
> > Some drivers (e.g. bus drivers) may want to check if power.irq_safe was
> > called by child driver, regardless of CONFIG_PM_RUNTIME.
> >
> > An example scenario is amba/bus.c and dma/pl330.c drivers. The runtime
> > suspend/resume callbacks in amba bus driver act differently if irq_safe
> > was set by child driver (in irq_safe mode bus clock is only disabled).
> >
> > The pl330 driver sets irq_safe and assumes that amba bus driver will
> > only disable the clock in runtime PM. So in system sleep suspend
> > callback the pl330 driver unprepares the clock after calling
> > pm_runtime_force_suspend().
> >
> > However inconsistency would appear if CONFIG_PM_RUNTIME is not set and
> > child drivers do not want the irq_safe runtime PM. In such case amba bus
> > driver still has to know whether child driver wanted irq_safe - by
> > looking at dev->power.irq_safe field.
> >
> > Signed-off-by: Krzysztof Kozlowski <k.kozlowski@xxxxxxxxxxx>
> > ---
> > include/linux/pm.h | 2 +-
> > include/linux/pm_runtime.h | 5 ++++-
> > 2 files changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/include/linux/pm.h b/include/linux/pm.h
> > index 383fd68aaee1..b05fa954f50d 100644
> > --- a/include/linux/pm.h
> > +++ b/include/linux/pm.h
> > @@ -566,6 +566,7 @@ struct dev_pm_info {
> > bool ignore_children:1;
> > bool early_init:1; /* Owned by the PM core */
> > bool direct_complete:1; /* Owned by the PM core */
> > + unsigned int irq_safe:1; /* PM runtime */
> > spinlock_t lock;
> > #ifdef CONFIG_PM_SLEEP
> > struct list_head entry;
> > @@ -590,7 +591,6 @@ struct dev_pm_info {
> > unsigned int run_wake:1;
> > unsigned int runtime_auto:1;
> > unsigned int no_callbacks:1;
> > - unsigned int irq_safe:1;
> > unsigned int use_autosuspend:1;
> > unsigned int timer_autosuspends:1;
> > unsigned int memalloc_noio:1;
>
> Well, that is a good reason to introduce a wrapper around power.irq_safe in my
> view.
>
> And define the wrapper so that it always returns false for CONFIG_PM_RUNTIME
> unset.
>
> This way not only you wouldn't need to move the flag from under the #ifdef,
> but also you would make the compiler skip the relevant pieces of code
> entiretly for CONFIG_PM_RUNTIME unset.

Few days ago I would be happy with your opinion :), but know I think
this is better solution than wrapper. Consider case:
1. PM_RUNTIME unset.
2. System suspends.
3. The pl330 in its suspend callback calls force_runtime_suspend which
leads us to amba/bus.
4. The amba/bus.c in runtime suspend checks for irq_safe (it is FALSE),
so it disables and unprepares the clock.
5. The pl330 in probe requested irq_safe so it assumes amba/bus will
only disable the clock. So the pl330 unprepares the clock. Again.

Best regards,
Krzysztof


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/