diff mbox series

[2/2] cpuidle: psci: Enable s2idle when using OSI with the PM domain topology

Message ID 20200901082707.106860-3-ulf.hansson@linaro.org
State New
Headers show
Series cpuidle: psci: Enable s2idle when using PSCI OSI | expand

Commit Message

Ulf Hansson Sept. 1, 2020, 8:27 a.m. UTC
To select domain idle states for cpuidle-psci, the PM domains via genpd are
being managed by using runtime PM. This works fine for the regular idle
path, but it doesn't when doing s2idle.

More precisely, the domain idle states becomes temporarily disabled, which
is because the PM core disables runtime PM for devices during system
suspend. Even if genpd tries to power off the PM domain in the
suspend_noirq phase, that doesn't help to properly select a domain idle
state, as this needs to be done on per CPU basis.

Let's address the issue by enabling the syscore flag for the attached CPU
devices. This prevents genpd from trying to power off the corresponding PM
domains in the suspend_noirq phase. Moreover, let's assign a specific
->enter_s2idle() callback for the corresponding domain idle state and let
it invoke pm_genpd_syscore_poweroff|poweron(), rather than using runtime
PM.

Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>

---
 drivers/cpuidle/cpuidle-psci-domain.c |  2 ++
 drivers/cpuidle/cpuidle-psci.c        | 30 +++++++++++++++++++++++----
 2 files changed, 28 insertions(+), 4 deletions(-)

-- 
2.25.1

Comments

Sudeep Holla Oct. 1, 2020, 10:17 a.m. UTC | #1
On Tue, Sep 01, 2020 at 10:27:07AM +0200, Ulf Hansson wrote:
> To select domain idle states for cpuidle-psci, the PM domains via genpd are
> being managed by using runtime PM. This works fine for the regular idle
> path, but it doesn't when doing s2idle.
>
> More precisely, the domain idle states becomes temporarily disabled, which
> is because the PM core disables runtime PM for devices during system
> suspend.

When you refer system suspend above, you mean both S2R and S2I ?

> Even if genpd tries to power off the PM domain in the
> suspend_noirq phase, that doesn't help to properly select a domain idle
> state, as this needs to be done on per CPU basis.
>

And what prevents doing per CPU basis ?

> Let's address the issue by enabling the syscore flag for the attached CPU
> devices. This prevents genpd from trying to power off the corresponding PM
> domains in the suspend_noirq phase. Moreover, let's assign a specific
> ->enter_s2idle() callback for the corresponding domain idle state and let
> it invoke pm_genpd_syscore_poweroff|poweron(), rather than using runtime
> PM.
>

The syscore_suspend is not executed for S2I and using syscore APIs here
is bit confusing IMO. If Rafael is fine, I have no objections.

> Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> ---
>  drivers/cpuidle/cpuidle-psci-domain.c |  2 ++
>  drivers/cpuidle/cpuidle-psci.c        | 30 +++++++++++++++++++++++----
>  2 files changed, 28 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
> index b6e9649ab0da..65437ba5fa78 100644
> --- a/drivers/cpuidle/cpuidle-psci-domain.c
> +++ b/drivers/cpuidle/cpuidle-psci-domain.c
> @@ -324,6 +324,8 @@ struct device *psci_dt_attach_cpu(int cpu)
>  	if (cpu_online(cpu))
>  		pm_runtime_get_sync(dev);
>
> +	dev_pm_syscore_device(dev, true);
> +
>  	return dev;
>  }
>
> diff --git a/drivers/cpuidle/cpuidle-psci.c b/drivers/cpuidle/cpuidle-psci.c
> index 74463841805f..6322d55a0a7d 100644
> --- a/drivers/cpuidle/cpuidle-psci.c
> +++ b/drivers/cpuidle/cpuidle-psci.c
> @@ -19,6 +19,7 @@
>  #include <linux/of_device.h>
>  #include <linux/platform_device.h>
>  #include <linux/psci.h>
> +#include <linux/pm_domain.h>
>  #include <linux/pm_runtime.h>
>  #include <linux/slab.h>
>  #include <linux/string.h>
> @@ -52,8 +53,9 @@ static inline int psci_enter_state(int idx, u32 state)
>  	return CPU_PM_CPU_IDLE_ENTER_PARAM(psci_cpu_suspend_enter, idx, state);
>  }
>
> -static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
> -					struct cpuidle_driver *drv, int idx)
> +static int __psci_enter_domain_idle_state(struct cpuidle_device *dev,
> +					  struct cpuidle_driver *drv, int idx,
> +					  bool s2idle)
>  {
>  	struct psci_cpuidle_data *data = this_cpu_ptr(&psci_cpuidle_data);
>  	u32 *states = data->psci_states;
> @@ -66,7 +68,10 @@ static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
>  		return -1;
>
>  	/* Do runtime PM to manage a hierarchical CPU toplogy. */
> -	pm_runtime_put_sync_suspend(pd_dev);
> +	if (s2idle)
> +		pm_genpd_syscore_poweroff(pd_dev);
> +	else
> +		pm_runtime_put_sync_suspend(pd_dev);

Since CPU genpd is special and handled so in core, can this be moved to core ?
I mean pm_runtime_put_sync_suspend handle that based genpd_is_cpu_domain.

--
Regards,
Sudeep
Rafael J. Wysocki Oct. 1, 2020, 11:24 a.m. UTC | #2
On Thu, Oct 1, 2020 at 12:18 PM Sudeep Holla <sudeep.holla@arm.com> wrote:
>
> On Tue, Sep 01, 2020 at 10:27:07AM +0200, Ulf Hansson wrote:
> > To select domain idle states for cpuidle-psci, the PM domains via genpd are
> > being managed by using runtime PM. This works fine for the regular idle
> > path, but it doesn't when doing s2idle.
> >
> > More precisely, the domain idle states becomes temporarily disabled, which
> > is because the PM core disables runtime PM for devices during system
> > suspend.
>
> When you refer system suspend above, you mean both S2R and S2I ?
>
> > Even if genpd tries to power off the PM domain in the
> > suspend_noirq phase, that doesn't help to properly select a domain idle
> > state, as this needs to be done on per CPU basis.
> >
>
> And what prevents doing per CPU basis ?
>
> > Let's address the issue by enabling the syscore flag for the attached CPU
> > devices. This prevents genpd from trying to power off the corresponding PM
> > domains in the suspend_noirq phase. Moreover, let's assign a specific
> > ->enter_s2idle() callback for the corresponding domain idle state and let
> > it invoke pm_genpd_syscore_poweroff|poweron(), rather than using runtime
> > PM.
> >
>
> The syscore_suspend is not executed for S2I and using syscore APIs here
> is bit confusing IMO.

Right.

> If Rafael is fine,

No, I'm not.

Cheers!

> I have no objections.
>
> > Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> > ---
> >  drivers/cpuidle/cpuidle-psci-domain.c |  2 ++
> >  drivers/cpuidle/cpuidle-psci.c        | 30 +++++++++++++++++++++++----
> >  2 files changed, 28 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
> > index b6e9649ab0da..65437ba5fa78 100644
> > --- a/drivers/cpuidle/cpuidle-psci-domain.c
> > +++ b/drivers/cpuidle/cpuidle-psci-domain.c
> > @@ -324,6 +324,8 @@ struct device *psci_dt_attach_cpu(int cpu)
> >       if (cpu_online(cpu))
> >               pm_runtime_get_sync(dev);
> >
> > +     dev_pm_syscore_device(dev, true);
> > +
> >       return dev;
> >  }
> >
> > diff --git a/drivers/cpuidle/cpuidle-psci.c b/drivers/cpuidle/cpuidle-psci.c
> > index 74463841805f..6322d55a0a7d 100644
> > --- a/drivers/cpuidle/cpuidle-psci.c
> > +++ b/drivers/cpuidle/cpuidle-psci.c
> > @@ -19,6 +19,7 @@
> >  #include <linux/of_device.h>
> >  #include <linux/platform_device.h>
> >  #include <linux/psci.h>
> > +#include <linux/pm_domain.h>
> >  #include <linux/pm_runtime.h>
> >  #include <linux/slab.h>
> >  #include <linux/string.h>
> > @@ -52,8 +53,9 @@ static inline int psci_enter_state(int idx, u32 state)
> >       return CPU_PM_CPU_IDLE_ENTER_PARAM(psci_cpu_suspend_enter, idx, state);
> >  }
> >
> > -static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
> > -                                     struct cpuidle_driver *drv, int idx)
> > +static int __psci_enter_domain_idle_state(struct cpuidle_device *dev,
> > +                                       struct cpuidle_driver *drv, int idx,
> > +                                       bool s2idle)
> >  {
> >       struct psci_cpuidle_data *data = this_cpu_ptr(&psci_cpuidle_data);
> >       u32 *states = data->psci_states;
> > @@ -66,7 +68,10 @@ static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
> >               return -1;
> >
> >       /* Do runtime PM to manage a hierarchical CPU toplogy. */
> > -     pm_runtime_put_sync_suspend(pd_dev);
> > +     if (s2idle)
> > +             pm_genpd_syscore_poweroff(pd_dev);
> > +     else
> > +             pm_runtime_put_sync_suspend(pd_dev);
>
> Since CPU genpd is special and handled so in core, can this be moved to core ?
> I mean pm_runtime_put_sync_suspend handle that based genpd_is_cpu_domain.
>
> --
> Regards,
> Sudeep
Ulf Hansson Oct. 1, 2020, 11:32 a.m. UTC | #3
On Thu, 1 Oct 2020 at 12:18, Sudeep Holla <sudeep.holla@arm.com> wrote:
>
> On Tue, Sep 01, 2020 at 10:27:07AM +0200, Ulf Hansson wrote:
> > To select domain idle states for cpuidle-psci, the PM domains via genpd are
> > being managed by using runtime PM. This works fine for the regular idle
> > path, but it doesn't when doing s2idle.
> >
> > More precisely, the domain idle states becomes temporarily disabled, which
> > is because the PM core disables runtime PM for devices during system
> > suspend.
>
> When you refer system suspend above, you mean both S2R and S2I ?

Correct.

Although, there is no problem with S2R to reach the proper idlestate,
because of the way we offline all but the boot CPU.

>
> > Even if genpd tries to power off the PM domain in the
> > suspend_noirq phase, that doesn't help to properly select a domain idle
> > state, as this needs to be done on per CPU basis.
> >
>
> And what prevents doing per CPU basis ?

The PM core doesn't execute the system suspend callbacks on a per CPU basis.

>
> > Let's address the issue by enabling the syscore flag for the attached CPU
> > devices. This prevents genpd from trying to power off the corresponding PM
> > domains in the suspend_noirq phase. Moreover, let's assign a specific
> > ->enter_s2idle() callback for the corresponding domain idle state and let
> > it invoke pm_genpd_syscore_poweroff|poweron(), rather than using runtime
> > PM.
> >
>
> The syscore_suspend is not executed for S2I and using syscore APIs here
> is bit confusing IMO. If Rafael is fine, I have no objections.

That's correct, the syscore phase doesn't exist in the S2I path.

However, in some cases the same functions that are being called in the
syscore phase, are also being called for S2I. For example, have a look
at timekeeping_suspend(), which is being called from both paths.

In the end, I think the confusing part is the name of the genpd functions.

Maybe we should rename pm_genpd_syscore_poweroff|poweron() to
pm_genpd_suspend|resume() - or something along those lines.

Kind regards
Uffe

>
> > Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
> > ---
> >  drivers/cpuidle/cpuidle-psci-domain.c |  2 ++
> >  drivers/cpuidle/cpuidle-psci.c        | 30 +++++++++++++++++++++++----
> >  2 files changed, 28 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
> > index b6e9649ab0da..65437ba5fa78 100644
> > --- a/drivers/cpuidle/cpuidle-psci-domain.c
> > +++ b/drivers/cpuidle/cpuidle-psci-domain.c
> > @@ -324,6 +324,8 @@ struct device *psci_dt_attach_cpu(int cpu)
> >       if (cpu_online(cpu))
> >               pm_runtime_get_sync(dev);
> >
> > +     dev_pm_syscore_device(dev, true);
> > +
> >       return dev;
> >  }
> >
> > diff --git a/drivers/cpuidle/cpuidle-psci.c b/drivers/cpuidle/cpuidle-psci.c
> > index 74463841805f..6322d55a0a7d 100644
> > --- a/drivers/cpuidle/cpuidle-psci.c
> > +++ b/drivers/cpuidle/cpuidle-psci.c
> > @@ -19,6 +19,7 @@
> >  #include <linux/of_device.h>
> >  #include <linux/platform_device.h>
> >  #include <linux/psci.h>
> > +#include <linux/pm_domain.h>
> >  #include <linux/pm_runtime.h>
> >  #include <linux/slab.h>
> >  #include <linux/string.h>
> > @@ -52,8 +53,9 @@ static inline int psci_enter_state(int idx, u32 state)
> >       return CPU_PM_CPU_IDLE_ENTER_PARAM(psci_cpu_suspend_enter, idx, state);
> >  }
> >
> > -static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
> > -                                     struct cpuidle_driver *drv, int idx)
> > +static int __psci_enter_domain_idle_state(struct cpuidle_device *dev,
> > +                                       struct cpuidle_driver *drv, int idx,
> > +                                       bool s2idle)
> >  {
> >       struct psci_cpuidle_data *data = this_cpu_ptr(&psci_cpuidle_data);
> >       u32 *states = data->psci_states;
> > @@ -66,7 +68,10 @@ static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
> >               return -1;
> >
> >       /* Do runtime PM to manage a hierarchical CPU toplogy. */
> > -     pm_runtime_put_sync_suspend(pd_dev);
> > +     if (s2idle)
> > +             pm_genpd_syscore_poweroff(pd_dev);
> > +     else
> > +             pm_runtime_put_sync_suspend(pd_dev);
>
> Since CPU genpd is special and handled so in core, can this be moved to core ?
> I mean pm_runtime_put_sync_suspend handle that based genpd_is_cpu_domain.
>
> --
> Regards,
> Sudeep
Ulf Hansson Oct. 1, 2020, 11:44 a.m. UTC | #4
On Thu, 1 Oct 2020 at 13:32, Ulf Hansson <ulf.hansson@linaro.org> wrote:
>
> On Thu, 1 Oct 2020 at 12:18, Sudeep Holla <sudeep.holla@arm.com> wrote:
> >
> > On Tue, Sep 01, 2020 at 10:27:07AM +0200, Ulf Hansson wrote:
> > > To select domain idle states for cpuidle-psci, the PM domains via genpd are
> > > being managed by using runtime PM. This works fine for the regular idle
> > > path, but it doesn't when doing s2idle.
> > >
> > > More precisely, the domain idle states becomes temporarily disabled, which
> > > is because the PM core disables runtime PM for devices during system
> > > suspend.
> >
> > When you refer system suspend above, you mean both S2R and S2I ?
>
> Correct.
>
> Although, there is no problem with S2R to reach the proper idlestate,
> because of the way we offline all but the boot CPU.
>
> >
> > > Even if genpd tries to power off the PM domain in the
> > > suspend_noirq phase, that doesn't help to properly select a domain idle
> > > state, as this needs to be done on per CPU basis.
> > >
> >
> > And what prevents doing per CPU basis ?
>
> The PM core doesn't execute the system suspend callbacks on a per CPU basis.
>
> >
> > > Let's address the issue by enabling the syscore flag for the attached CPU
> > > devices. This prevents genpd from trying to power off the corresponding PM
> > > domains in the suspend_noirq phase. Moreover, let's assign a specific
> > > ->enter_s2idle() callback for the corresponding domain idle state and let
> > > it invoke pm_genpd_syscore_poweroff|poweron(), rather than using runtime
> > > PM.
> > >
> >
> > The syscore_suspend is not executed for S2I and using syscore APIs here
> > is bit confusing IMO. If Rafael is fine, I have no objections.
>
> That's correct, the syscore phase doesn't exist in the S2I path.
>
> However, in some cases the same functions that are being called in the
> syscore phase, are also being called for S2I. For example, have a look
> at timekeeping_suspend(), which is being called from both paths.
>
> In the end, I think the confusing part is the name of the genpd functions.
>
> Maybe we should rename pm_genpd_syscore_poweroff|poweron() to
> pm_genpd_suspend|resume() - or something along those lines.
>

[...]

Rafael, I understand you have objections to the approach. Would
renaming the genpd APIs be a way forward? Another option is to add two
new genpd APIs along the side of the existing, not sure what would be
the best.

Kind regards
Uffe
diff mbox series

Patch

diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
index b6e9649ab0da..65437ba5fa78 100644
--- a/drivers/cpuidle/cpuidle-psci-domain.c
+++ b/drivers/cpuidle/cpuidle-psci-domain.c
@@ -324,6 +324,8 @@  struct device *psci_dt_attach_cpu(int cpu)
 	if (cpu_online(cpu))
 		pm_runtime_get_sync(dev);
 
+	dev_pm_syscore_device(dev, true);
+
 	return dev;
 }
 
diff --git a/drivers/cpuidle/cpuidle-psci.c b/drivers/cpuidle/cpuidle-psci.c
index 74463841805f..6322d55a0a7d 100644
--- a/drivers/cpuidle/cpuidle-psci.c
+++ b/drivers/cpuidle/cpuidle-psci.c
@@ -19,6 +19,7 @@ 
 #include <linux/of_device.h>
 #include <linux/platform_device.h>
 #include <linux/psci.h>
+#include <linux/pm_domain.h>
 #include <linux/pm_runtime.h>
 #include <linux/slab.h>
 #include <linux/string.h>
@@ -52,8 +53,9 @@  static inline int psci_enter_state(int idx, u32 state)
 	return CPU_PM_CPU_IDLE_ENTER_PARAM(psci_cpu_suspend_enter, idx, state);
 }
 
-static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
-					struct cpuidle_driver *drv, int idx)
+static int __psci_enter_domain_idle_state(struct cpuidle_device *dev,
+					  struct cpuidle_driver *drv, int idx,
+					  bool s2idle)
 {
 	struct psci_cpuidle_data *data = this_cpu_ptr(&psci_cpuidle_data);
 	u32 *states = data->psci_states;
@@ -66,7 +68,10 @@  static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
 		return -1;
 
 	/* Do runtime PM to manage a hierarchical CPU toplogy. */
-	pm_runtime_put_sync_suspend(pd_dev);
+	if (s2idle)
+		pm_genpd_syscore_poweroff(pd_dev);
+	else
+		pm_runtime_put_sync_suspend(pd_dev);
 
 	state = psci_get_domain_state();
 	if (!state)
@@ -74,7 +79,10 @@  static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
 
 	ret = psci_cpu_suspend_enter(state) ? -1 : idx;
 
-	pm_runtime_get_sync(pd_dev);
+	if (s2idle)
+		pm_genpd_syscore_poweron(pd_dev);
+	else
+		pm_runtime_get_sync(pd_dev);
 
 	cpu_pm_exit();
 
@@ -83,6 +91,19 @@  static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
 	return ret;
 }
 
+static int psci_enter_domain_idle_state(struct cpuidle_device *dev,
+					struct cpuidle_driver *drv, int idx)
+{
+	return __psci_enter_domain_idle_state(dev, drv, idx, false);
+}
+
+static int psci_enter_s2idle_domain_idle_state(struct cpuidle_device *dev,
+					       struct cpuidle_driver *drv,
+					       int idx)
+{
+	return __psci_enter_domain_idle_state(dev, drv, idx, true);
+}
+
 static int psci_idle_cpuhp_up(unsigned int cpu)
 {
 	struct device *pd_dev = __this_cpu_read(psci_cpuidle_data.dev);
@@ -170,6 +191,7 @@  static int psci_dt_cpu_init_topology(struct cpuidle_driver *drv,
 	 * deeper states.
 	 */
 	drv->states[state_count - 1].enter = psci_enter_domain_idle_state;
+	drv->states[state_count - 1].enter_s2idle = psci_enter_s2idle_domain_idle_state;
 	psci_cpuidle_use_cpuhp = true;
 
 	return 0;