mbox series

[0/2] s2idle fixes for systems without cpuidle

Message ID 20230711-cpuidle-v1-0-f391224b3140@gmail.com
Headers show
Series s2idle fixes for systems without cpuidle | expand

Message

Kazuki Hashimoto July 11, 2023, 5:54 a.m. UTC
Changes in V2:
Redo implementation following input from Hector Martin

Changes in V3:
Rewrite commit message following input from Rafael J. Wysocki

---
Kazuki Hashimoto (2):
      cpuidle: Don't pass any values to cpuidle_not_available
      PM: s2idle: Fully prevent the system from entering s2idle when cpuidle isn't supported

 drivers/cpuidle/cpuidle.c |  6 ++++--
 include/linux/cpuidle.h   |  6 ++----
 kernel/power/main.c       | 12 +++++++++---
 kernel/power/suspend.c    |  5 +++++
 kernel/sched/idle.c       |  2 +-
 5 files changed, 21 insertions(+), 10 deletions(-)
---
base-commit: 8fc3b8f082cc2f5faa6eae315b938bc5e79c332e
change-id: 20230709-cpuidle-8c5469788f77

Best regards,

Comments

Kazuki Hashimoto July 11, 2023, 6:48 p.m. UTC | #1
on oue, Jul 11, 2023 at 09:42:31AM +0200, Peter Zijlstra wrote:
> On Tue, Jul 11, 2023 at 02:54:21PM +0900, Kazuki Hashimoto wrote:
> > There's no reason to pass any values to cpuidle_not_available() as the
> > function works standalone. Since we're planning to use the function in
> > other places, make it so to avoid code duplication.
> > 
> > Signed-off-by: Kazuki Hashimoto <kazukih0205@gmail.com>
> > ---
> >  drivers/cpuidle/cpuidle.c | 6 ++++--
> >  include/linux/cpuidle.h   | 6 ++----
> >  kernel/sched/idle.c       | 2 +-
> >  3 files changed, 7 insertions(+), 7 deletions(-)
> > 
> > diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
> > index 737a026ef58a..c9ba51e0fa38 100644
> > --- a/drivers/cpuidle/cpuidle.c
> > +++ b/drivers/cpuidle/cpuidle.c
> > @@ -49,9 +49,11 @@ void disable_cpuidle(void)
> >  	off = 1;
> >  }
> >  
> > -bool cpuidle_not_available(struct cpuidle_driver *drv,
> > -			   struct cpuidle_device *dev)
> > +bool cpuidle_not_available(void)
> >  {
> > +	struct cpuidle_device *dev = cpuidle_get_device();
> > +	struct cpuidle_driver *drv = cpuidle_get_cpu_driver(dev);
> > +
> >  	return off || !initialized || !drv || !dev || !dev->enabled;
> >  }
> 
> It appears to me these are a lot of conditions to check *every* time we
> go idle -- especially since they hardly, if ever, change.
> 
> Can't cpuidle track all this in a single global variable, preferably as
> a static_key ?
I don't think so? I'll drop this one though since it adds unnecesary
overhead.

Thanks,
Kazuki