Re: [PATCH] PM: domains: not update genpd status when power on fail

Re: [PATCH] PM: domains: not update genpd status when power on fail

[Ulf Hansson]

On Tue, 6 Jul 2021 at 11:31, <peng.fan@oss.nxp.com> wrote:
>
> From: Peng Fan <peng.fan@nxp.com>
>
> When _genpd_power_on fail, the generic power domain status
> should not be changed to GENPD_STATE_ON.
>
> Signed-off-by: Peng Fan <peng.fan@nxp.com>
> ---
>  drivers/base/power/domain.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
> index ab0b740cc0f1..754a5d384479 100644
> --- a/drivers/base/power/domain.c
> +++ b/drivers/base/power/domain.c
> @@ -1101,6 +1101,7 @@ static void genpd_sync_power_on(struct generic_pm_domain *genpd, bool use_lock,
>                                 unsigned int depth)
>  {
>         struct gpd_link *link;
> +       int ret;
>
>         if (genpd_status_on(genpd))
>                 return;
> @@ -1117,8 +1118,9 @@ static void genpd_sync_power_on(struct generic_pm_domain *genpd, bool use_lock,
>                         genpd_unlock(link->parent);
>         }
>
> -       _genpd_power_on(genpd, false);
> -       genpd->status = GENPD_STATE_ON;
> +       ret = _genpd_power_on(genpd, false);
> +       if (!ret)

To support this error path, a lot more code needs to be added. Just
have a look at the callers of genpd_sync_power_on() and compare it
with genpd_power_on().

In my opinion, supporting this error path isn't really worth it, at
least until someone comes with a valid use case. Do you have one - or
did you send the $subject patch based solely on code inspection?

> +               genpd->status = GENPD_STATE_ON;
>  }
>
>  /**
> --
> 2.30.0
>

Kind regards
Uffe

RE: [PATCH] PM: domains: not update genpd status when power on fail

[Peng Fan (OSS)]

Hi Ulf,

> Subject: Re: [PATCH] PM: domains: not update genpd status when power on
> fail
> 
> On Tue, 6 Jul 2021 at 11:31, <peng.fan@oss.nxp.com> wrote:
> >
> > From: Peng Fan <peng.fan@nxp.com>
> >
> > When _genpd_power_on fail, the generic power domain status should not
> > be changed to GENPD_STATE_ON.
> >
> > Signed-off-by: Peng Fan <peng.fan@nxp.com>
> > ---
> >  drivers/base/power/domain.c | 6 ++++--
> >  1 file changed, 4 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
> > index ab0b740cc0f1..754a5d384479 100644
> > --- a/drivers/base/power/domain.c
> > +++ b/drivers/base/power/domain.c
> > @@ -1101,6 +1101,7 @@ static void genpd_sync_power_on(struct
> generic_pm_domain *genpd, bool use_lock,
> >                                 unsigned int depth)  {
> >         struct gpd_link *link;
> > +       int ret;
> >
> >         if (genpd_status_on(genpd))
> >                 return;
> > @@ -1117,8 +1118,9 @@ static void genpd_sync_power_on(struct
> generic_pm_domain *genpd, bool use_lock,
> >                         genpd_unlock(link->parent);
> >         }
> >
> > -       _genpd_power_on(genpd, false);
> > -       genpd->status = GENPD_STATE_ON;
> > +       ret = _genpd_power_on(genpd, false);
> > +       if (!ret)
> 
> To support this error path, a lot more code needs to be added. Just
> have a look at the callers of genpd_sync_power_on() and compare it
> with genpd_power_on().
> 
> In my opinion, supporting this error path isn't really worth it, at
> least until someone comes with a valid use case. Do you have one - or
> did you send the $subject patch based solely on code inspection?

This is two patchset:
https://patchwork.kernel.org/project/linux-arm-kernel/patch/20210604111005.6804-1-peng.fan@oss.nxp.com/
https://patchwork.kernel.org/project/linux-arm-kernel/cover/20210629072941.7980-1-peng.fan@oss.nxp.com/

I collected in github with a few fixes:
https://github.com/MrVan/linux/tree/linux-next-master-628-blk-ctl-test


In my test, when suspend/resume, I see
pgc_vpu_h1/g1/g2 has status on, but actually they are not on in hardware.
Because vpumix_blk_ctl bus domain is not on, pgc_vpu_h1/g1/g2 runtime
power off will fail.

Thanks,
Peng.

> 
> > +               genpd->status = GENPD_STATE_ON;
> >  }
> >
> >  /**
> > --
> > 2.30.0
> >
> 
> Kind regards
> Uffe

[PATCH] PM: domains: not update genpd status when power on fail

[peng.fan]

From: Peng Fan <peng.fan@nxp.com>

When _genpd_power_on fail, the generic power domain status
should not be changed to GENPD_STATE_ON.

Signed-off-by: Peng Fan <peng.fan@nxp.com>
---
 drivers/base/power/domain.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
index ab0b740cc0f1..754a5d384479 100644
--- a/drivers/base/power/domain.c
+++ b/drivers/base/power/domain.c
@@ -1101,6 +1101,7 @@ static void genpd_sync_power_on(struct generic_pm_domain *genpd, bool use_lock,
 				unsigned int depth)
 {
 	struct gpd_link *link;
+	int ret;
 
 	if (genpd_status_on(genpd))
 		return;
@@ -1117,8 +1118,9 @@ static void genpd_sync_power_on(struct generic_pm_domain *genpd, bool use_lock,
 			genpd_unlock(link->parent);
 	}
 
-	_genpd_power_on(genpd, false);
-	genpd->status = GENPD_STATE_ON;
+	ret = _genpd_power_on(genpd, false);
+	if (!ret)
+		genpd->status = GENPD_STATE_ON;
 }
 
 /**
-- 
2.30.0

Re: [PATCH] PM: domains: not update genpd status when power on fail

[Ulf Hansson]

On Tue, 6 Jul 2021 at 11:52, Peng Fan (OSS) <peng.fan@oss.nxp.com> wrote:
>
> Hi Ulf,
>
> > Subject: Re: [PATCH] PM: domains: not update genpd status when power on
> > fail
> >
> > On Tue, 6 Jul 2021 at 11:31, <peng.fan@oss.nxp.com> wrote:
> > >
> > > From: Peng Fan <peng.fan@nxp.com>
> > >
> > > When _genpd_power_on fail, the generic power domain status should not
> > > be changed to GENPD_STATE_ON.
> > >
> > > Signed-off-by: Peng Fan <peng.fan@nxp.com>
> > > ---
> > >  drivers/base/power/domain.c | 6 ++++--
> > >  1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
> > > index ab0b740cc0f1..754a5d384479 100644
> > > --- a/drivers/base/power/domain.c
> > > +++ b/drivers/base/power/domain.c
> > > @@ -1101,6 +1101,7 @@ static void genpd_sync_power_on(struct
> > generic_pm_domain *genpd, bool use_lock,
> > >                                 unsigned int depth)  {
> > >         struct gpd_link *link;
> > > +       int ret;
> > >
> > >         if (genpd_status_on(genpd))
> > >                 return;
> > > @@ -1117,8 +1118,9 @@ static void genpd_sync_power_on(struct
> > generic_pm_domain *genpd, bool use_lock,
> > >                         genpd_unlock(link->parent);
> > >         }
> > >
> > > -       _genpd_power_on(genpd, false);
> > > -       genpd->status = GENPD_STATE_ON;
> > > +       ret = _genpd_power_on(genpd, false);
> > > +       if (!ret)
> >
> > To support this error path, a lot more code needs to be added. Just
> > have a look at the callers of genpd_sync_power_on() and compare it
> > with genpd_power_on().
> >
> > In my opinion, supporting this error path isn't really worth it, at
> > least until someone comes with a valid use case. Do you have one - or
> > did you send the $subject patch based solely on code inspection?
>
> This is two patchset:
> https://patchwork.kernel.org/project/linux-arm-kernel/patch/20210604111005.6804-1-peng.fan@oss.nxp.com/
> https://patchwork.kernel.org/project/linux-arm-kernel/cover/20210629072941.7980-1-peng.fan@oss.nxp.com/
>
> I collected in github with a few fixes:
> https://github.com/MrVan/linux/tree/linux-next-master-628-blk-ctl-test

I looked at some of the changes there - and I especially don't quite
understand why you enable runtime PM for the genpd->dev. That looks
wrong to me. I am referring to:
"soc: imx: blk-ctl: fix to v8 patchset".

>
>
> In my test, when suspend/resume, I see
> pgc_vpu_h1/g1/g2 has status on, but actually they are not on in hardware.

That sounds like a bug somewhere. Did genpd fail to power off the PM
domain or what happened, more exactly?

> Because vpumix_blk_ctl bus domain is not on, pgc_vpu_h1/g1/g2 runtime
> power off will fail.

What is pgc_vpu_h1/g1/g2?

I am happy to help/guide you forward, but it sounds a bit that more
in-depth debugging is needed at your side. At least I need some stack
trace or kernel logs with debug messages enabled.

[...]

Kind regards
Uffe