[RESEND PATCH v2] firmware/psci: Print a warning if PSCI doesn't accept PC mode

[RESEND PATCH v2] firmware/psci: Print a warning if PSCI doesn't accept PC mode

[Dmitry Baryshkov]

The function psci_pd_try_set_osi_mode() will print an error if enabling
OSI mode fails. To ease debugging PSCI issues print corresponding
message if switching to PC mode fails too.

Acked-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
---
This is a replacement for [1], now moving the warning from
psci_set_osi_mode() callers to the function iself.

The patch is resent to include soc@kernel.org

[1] https://lore.kernel.org/all/20220727182034.983727-1-dmitry.baryshkov@linaro.org/

---
 drivers/cpuidle/cpuidle-psci-domain.c | 4 +---
 drivers/firmware/psci/psci.c          | 2 ++
 2 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
index 3db4fca1172b..821984947ed9 100644
--- a/drivers/cpuidle/cpuidle-psci-domain.c
+++ b/drivers/cpuidle/cpuidle-psci-domain.c
@@ -124,10 +124,8 @@ static bool psci_pd_try_set_osi_mode(void)
 		return false;
 
 	ret = psci_set_osi_mode(true);
-	if (ret) {
-		pr_warn("failed to enable OSI mode: %d\n", ret);
+	if (ret)
 		return false;
-	}
 
 	return true;
 }
diff --git a/drivers/firmware/psci/psci.c b/drivers/firmware/psci/psci.c
index 9fdcb6bff403..42cae0ba10e2 100644
--- a/drivers/firmware/psci/psci.c
+++ b/drivers/firmware/psci/psci.c
@@ -164,6 +164,8 @@ int psci_set_osi_mode(bool enable)
 			PSCI_1_0_SUSPEND_MODE_PC;
 
 	err = invoke_psci_fn(PSCI_1_0_FN_SET_SUSPEND_MODE, suspend_mode, 0, 0);
+	if (err < 0)
+		pr_warn("failed to set %s mode: %d\n", enable ? "OSI" : "PC", err);
 	return psci_to_linux_errno(err);
 }
 
-- 
2.35.1

Re: [RESEND PATCH v2] firmware/psci: Print a warning if PSCI doesn't accept PC mode

[Ulf Hansson]

On Wed, 19 Oct 2022 at 16:14, Johan Hovold <johan@kernel.org> wrote:
>
> On Mon, Sep 26, 2022 at 02:02:49PM +0300, Dmitry Baryshkov wrote:
> > The function psci_pd_try_set_osi_mode() will print an error if enabling
> > OSI mode fails. To ease debugging PSCI issues print corresponding
> > message if switching to PC mode fails too.
> >
> > Acked-by: Mark Rutland <mark.rutland@arm.com>
> > Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
> > Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
> > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
> > ---
> > This is a replacement for [1], now moving the warning from
> > psci_set_osi_mode() callers to the function iself.
> >
> > The patch is resent to include soc@kernel.org
> >
> > [1] https://lore.kernel.org/all/20220727182034.983727-1-dmitry.baryshkov@linaro.org/
> >
> > ---
> >  drivers/cpuidle/cpuidle-psci-domain.c | 4 +---
> >  drivers/firmware/psci/psci.c          | 2 ++
> >  2 files changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
> > index 3db4fca1172b..821984947ed9 100644
> > --- a/drivers/cpuidle/cpuidle-psci-domain.c
> > +++ b/drivers/cpuidle/cpuidle-psci-domain.c
> > @@ -124,10 +124,8 @@ static bool psci_pd_try_set_osi_mode(void)
> >               return false;
> >
> >       ret = psci_set_osi_mode(true);
> > -     if (ret) {
> > -             pr_warn("failed to enable OSI mode: %d\n", ret);
> > +     if (ret)
> >               return false;
> > -     }
> >
> >       return true;
> >  }
> > diff --git a/drivers/firmware/psci/psci.c b/drivers/firmware/psci/psci.c
> > index 9fdcb6bff403..42cae0ba10e2 100644
> > --- a/drivers/firmware/psci/psci.c
> > +++ b/drivers/firmware/psci/psci.c
> > @@ -164,6 +164,8 @@ int psci_set_osi_mode(bool enable)
> >                       PSCI_1_0_SUSPEND_MODE_PC;
> >
> >       err = invoke_psci_fn(PSCI_1_0_FN_SET_SUSPEND_MODE, suspend_mode, 0, 0);
> > +     if (err < 0)
> > +             pr_warn("failed to set %s mode: %d\n", enable ? "OSI" : "PC", err);
> >       return psci_to_linux_errno(err);
> >  }
>
> When booting 6.1-rc1 I now see:
>
>         [    0.000000] psci: OSI mode supported.
>         [    0.000000] psci: failed to set PC mode: -3
>
> on every boot with sc8280xp, while later enabling OSI mode still works:
>
>         [    0.227358] CPUidle PSCI: psci_pd_try_set_osi_mode - success
>         [    0.227599] CPUidle PSCI: Initialized CPU PM domain topology
>
> Judging from Sudeep's comment on v1:
>
>         The platform must boot in PC mode, so even if it fails we ignore
>         so not sure if the logging is of much help here IMO.
>
> perhaps logging this as an error is not a good idea?

I think it's useful to get information that the FW doesn't fully
conform to the spec (it should support PC mode), but that doesn't
really mean that it can't work.

To me, it sounds like we should move the logging to the debug level!?

>
> Johan

Br
Uffe

Re: [RESEND PATCH v2] firmware/psci: Print a warning if PSCI doesn't accept PC mode

[Dmitry Baryshkov]

On Wed, 19 Oct 2022 at 17:37, Ulf Hansson <ulf.hansson@linaro.org> wrote:
>
> On Wed, 19 Oct 2022 at 16:14, Johan Hovold <johan@kernel.org> wrote:
> >
> > On Mon, Sep 26, 2022 at 02:02:49PM +0300, Dmitry Baryshkov wrote:
> > > The function psci_pd_try_set_osi_mode() will print an error if enabling
> > > OSI mode fails. To ease debugging PSCI issues print corresponding
> > > message if switching to PC mode fails too.
> > >
> > > Acked-by: Mark Rutland <mark.rutland@arm.com>
> > > Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
> > > Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
> > > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
> > > ---
> > > This is a replacement for [1], now moving the warning from
> > > psci_set_osi_mode() callers to the function iself.
> > >
> > > The patch is resent to include soc@kernel.org
> > >
> > > [1] https://lore.kernel.org/all/20220727182034.983727-1-dmitry.baryshkov@linaro.org/
> > >
> > > ---
> > >  drivers/cpuidle/cpuidle-psci-domain.c | 4 +---
> > >  drivers/firmware/psci/psci.c          | 2 ++
> > >  2 files changed, 3 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
> > > index 3db4fca1172b..821984947ed9 100644
> > > --- a/drivers/cpuidle/cpuidle-psci-domain.c
> > > +++ b/drivers/cpuidle/cpuidle-psci-domain.c
> > > @@ -124,10 +124,8 @@ static bool psci_pd_try_set_osi_mode(void)
> > >               return false;
> > >
> > >       ret = psci_set_osi_mode(true);
> > > -     if (ret) {
> > > -             pr_warn("failed to enable OSI mode: %d\n", ret);
> > > +     if (ret)
> > >               return false;
> > > -     }
> > >
> > >       return true;
> > >  }
> > > diff --git a/drivers/firmware/psci/psci.c b/drivers/firmware/psci/psci.c
> > > index 9fdcb6bff403..42cae0ba10e2 100644
> > > --- a/drivers/firmware/psci/psci.c
> > > +++ b/drivers/firmware/psci/psci.c
> > > @@ -164,6 +164,8 @@ int psci_set_osi_mode(bool enable)
> > >                       PSCI_1_0_SUSPEND_MODE_PC;
> > >
> > >       err = invoke_psci_fn(PSCI_1_0_FN_SET_SUSPEND_MODE, suspend_mode, 0, 0);
> > > +     if (err < 0)
> > > +             pr_warn("failed to set %s mode: %d\n", enable ? "OSI" : "PC", err);
> > >       return psci_to_linux_errno(err);
> > >  }
> >
> > When booting 6.1-rc1 I now see:
> >
> >         [    0.000000] psci: OSI mode supported.
> >         [    0.000000] psci: failed to set PC mode: -3
> >
> > on every boot with sc8280xp, while later enabling OSI mode still works:
> >
> >         [    0.227358] CPUidle PSCI: psci_pd_try_set_osi_mode - success
> >         [    0.227599] CPUidle PSCI: Initialized CPU PM domain topology
> >
> > Judging from Sudeep's comment on v1:
> >
> >         The platform must boot in PC mode, so even if it fails we ignore
> >         so not sure if the logging is of much help here IMO.
> >
> > perhaps logging this as an error is not a good idea?

As far as I understand, some of the Qualcomm platforms booted in the
OSI mode, no matter what the spec says. I'm not sure about the
sc8280xp, but it definitely was the case previously. This was the
primary reason for adding a log here.

> I think it's useful to get information that the FW doesn't fully
> conform to the spec (it should support PC mode), but that doesn't
> really mean that it can't work.
>
> To me, it sounds like we should move the logging to the debug level!?

I'd say, either ignore the error, or move the warning to the info or
debug level.

-- 
With best wishes
Dmitry

Re: [RESEND PATCH v2] firmware/psci: Print a warning if PSCI doesn't accept PC mode

[Ulf Hansson]

On Thu, 20 Oct 2022 at 08:58, Johan Hovold <johan@kernel.org> wrote:
>
> On Wed, Oct 19, 2022 at 06:02:37PM +0300, Dmitry Baryshkov wrote:
> > On Wed, 19 Oct 2022 at 17:37, Ulf Hansson <ulf.hansson@linaro.org> wrote:
> > > On Wed, 19 Oct 2022 at 16:14, Johan Hovold <johan@kernel.org> wrote:
> > > > On Mon, Sep 26, 2022 at 02:02:49PM +0300, Dmitry Baryshkov wrote:
> > > > > The function psci_pd_try_set_osi_mode() will print an error if enabling
> > > > > OSI mode fails. To ease debugging PSCI issues print corresponding
> > > > > message if switching to PC mode fails too.
> > > > >
> > > > > Acked-by: Mark Rutland <mark.rutland@arm.com>
> > > > > Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
> > > > > Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
> > > > > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
>
> > > > > diff --git a/drivers/firmware/psci/psci.c b/drivers/firmware/psci/psci.c
> > > > > index 9fdcb6bff403..42cae0ba10e2 100644
> > > > > --- a/drivers/firmware/psci/psci.c
> > > > > +++ b/drivers/firmware/psci/psci.c
> > > > > @@ -164,6 +164,8 @@ int psci_set_osi_mode(bool enable)
> > > > >                       PSCI_1_0_SUSPEND_MODE_PC;
> > > > >
> > > > >       err = invoke_psci_fn(PSCI_1_0_FN_SET_SUSPEND_MODE, suspend_mode, 0, 0);
> > > > > +     if (err < 0)
> > > > > +             pr_warn("failed to set %s mode: %d\n", enable ? "OSI" : "PC", err);
> > > > >       return psci_to_linux_errno(err);
> > > > >  }
> > > >
> > > > When booting 6.1-rc1 I now see:
> > > >
> > > >         [    0.000000] psci: OSI mode supported.
> > > >         [    0.000000] psci: failed to set PC mode: -3
> > > >
> > > > on every boot with sc8280xp, while later enabling OSI mode still works:
> > > >
> > > >         [    0.227358] CPUidle PSCI: psci_pd_try_set_osi_mode - success
> > > >         [    0.227599] CPUidle PSCI: Initialized CPU PM domain topology
> > > >
> > > > Judging from Sudeep's comment on v1:
> > > >
> > > >         The platform must boot in PC mode, so even if it fails we ignore
> > > >         so not sure if the logging is of much help here IMO.
> > > >
> > > > perhaps logging this as an error is not a good idea?
> >
> > As far as I understand, some of the Qualcomm platforms booted in the
> > OSI mode, no matter what the spec says. I'm not sure about the
> > sc8280xp, but it definitely was the case previously. This was the
> > primary reason for adding a log here.
> >
> > > I think it's useful to get information that the FW doesn't fully
> > > conform to the spec (it should support PC mode), but that doesn't
> > > really mean that it can't work.
> > >
> > > To me, it sounds like we should move the logging to the debug level!?
> >
> > I'd say, either ignore the error, or move the warning to the info or
> > debug level.
>
> "Ignore the error" as in removing the warning or as in telling every
> owner of an X13s Thinkpad to ignore it?
>
> If there's nothing that can be done about it and this is nothing to
> worry about, then moving to debug level would seem like the right thing
> to do here. At least for the PC case.

Yes.

For the OSI case, I think a better option is to extend the print of
"Initialized CPU PM domain topology\n", in psci_cpuidle_domain_probe()
with information about whether OSI is being used or not.

Kind regards
Uffe

Re: [RESEND PATCH v2] firmware/psci: Print a warning if PSCI doesn't accept PC mode

[Dmitry Baryshkov]

On 20/10/2022 09:58, Johan Hovold wrote:
> On Wed, Oct 19, 2022 at 06:02:37PM +0300, Dmitry Baryshkov wrote:
>> On Wed, 19 Oct 2022 at 17:37, Ulf Hansson <ulf.hansson@linaro.org> wrote:
>>> On Wed, 19 Oct 2022 at 16:14, Johan Hovold <johan@kernel.org> wrote:
>>>> On Mon, Sep 26, 2022 at 02:02:49PM +0300, Dmitry Baryshkov wrote:
>>>>> The function psci_pd_try_set_osi_mode() will print an error if enabling
>>>>> OSI mode fails. To ease debugging PSCI issues print corresponding
>>>>> message if switching to PC mode fails too.
>>>>>
>>>>> Acked-by: Mark Rutland <mark.rutland@arm.com>
>>>>> Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
>>>>> Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
>>>>> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
> 
>>>>> diff --git a/drivers/firmware/psci/psci.c b/drivers/firmware/psci/psci.c
>>>>> index 9fdcb6bff403..42cae0ba10e2 100644
>>>>> --- a/drivers/firmware/psci/psci.c
>>>>> +++ b/drivers/firmware/psci/psci.c
>>>>> @@ -164,6 +164,8 @@ int psci_set_osi_mode(bool enable)
>>>>>                        PSCI_1_0_SUSPEND_MODE_PC;
>>>>>
>>>>>        err = invoke_psci_fn(PSCI_1_0_FN_SET_SUSPEND_MODE, suspend_mode, 0, 0);
>>>>> +     if (err < 0)
>>>>> +             pr_warn("failed to set %s mode: %d\n", enable ? "OSI" : "PC", err);
>>>>>        return psci_to_linux_errno(err);
>>>>>   }
>>>>
>>>> When booting 6.1-rc1 I now see:
>>>>
>>>>          [    0.000000] psci: OSI mode supported.
>>>>          [    0.000000] psci: failed to set PC mode: -3
>>>>
>>>> on every boot with sc8280xp, while later enabling OSI mode still works:
>>>>
>>>>          [    0.227358] CPUidle PSCI: psci_pd_try_set_osi_mode - success
>>>>          [    0.227599] CPUidle PSCI: Initialized CPU PM domain topology
>>>>
>>>> Judging from Sudeep's comment on v1:
>>>>
>>>>          The platform must boot in PC mode, so even if it fails we ignore
>>>>          so not sure if the logging is of much help here IMO.
>>>>
>>>> perhaps logging this as an error is not a good idea?
>>
>> As far as I understand, some of the Qualcomm platforms booted in the
>> OSI mode, no matter what the spec says. I'm not sure about the
>> sc8280xp, but it definitely was the case previously. This was the
>> primary reason for adding a log here.
>>
>>> I think it's useful to get information that the FW doesn't fully
>>> conform to the spec (it should support PC mode), but that doesn't
>>> really mean that it can't work.
>>>
>>> To me, it sounds like we should move the logging to the debug level!?
>>
>> I'd say, either ignore the error, or move the warning to the info or
>> debug level.
> 
> "Ignore the error" as in removing the warning or as in telling every
> owner of an X13s Thinkpad to ignore it?
> 
> If there's nothing that can be done about it and this is nothing to
> worry about, then moving to debug level would seem like the right thing
> to do here. At least for the PC case.

Feel free to send the patch. I'd maybe use dev_info rather than dev_dbg, 
to let the user have a chance of seeing the warning.

Regarding telling every user to ignore the warning. This is not 
something new. E.g. each time I boot my ThinkPad, I see two or three 
warning messages related to TPM, despite having 'quiet' in kernel boot args.

> 
> Johan

-- 
With best wishes
Dmitry

Re: [RESEND PATCH v2] firmware/psci: Print a warning if PSCI doesn't accept PC mode

[Sudeep Holla]

On Wed, Oct 19, 2022 at 04:14:04PM +0200, Johan Hovold wrote:
> On Mon, Sep 26, 2022 at 02:02:49PM +0300, Dmitry Baryshkov wrote:
> > The function psci_pd_try_set_osi_mode() will print an error if enabling
> > OSI mode fails. To ease debugging PSCI issues print corresponding
> > message if switching to PC mode fails too.
> >
> > Acked-by: Mark Rutland <mark.rutland@arm.com>
> > Reviewed-by: Sudeep Holla <sudeep.holla@arm.com>
> > Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
> > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
> > ---
> > This is a replacement for [1], now moving the warning from
> > psci_set_osi_mode() callers to the function iself.
> >
> > The patch is resent to include soc@kernel.org
> >
> > [1] https://lore.kernel.org/all/20220727182034.983727-1-dmitry.baryshkov@linaro.org/
> >
> > ---
> >  drivers/cpuidle/cpuidle-psci-domain.c | 4 +---
> >  drivers/firmware/psci/psci.c          | 2 ++
> >  2 files changed, 3 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/cpuidle/cpuidle-psci-domain.c b/drivers/cpuidle/cpuidle-psci-domain.c
> > index 3db4fca1172b..821984947ed9 100644
> > --- a/drivers/cpuidle/cpuidle-psci-domain.c
> > +++ b/drivers/cpuidle/cpuidle-psci-domain.c
> > @@ -124,10 +124,8 @@ static bool psci_pd_try_set_osi_mode(void)
> >  		return false;
> >
> >  	ret = psci_set_osi_mode(true);
> > -	if (ret) {
> > -		pr_warn("failed to enable OSI mode: %d\n", ret);
> > +	if (ret)
> >  		return false;
> > -	}
> >
> >  	return true;
> >  }
> > diff --git a/drivers/firmware/psci/psci.c b/drivers/firmware/psci/psci.c
> > index 9fdcb6bff403..42cae0ba10e2 100644
> > --- a/drivers/firmware/psci/psci.c
> > +++ b/drivers/firmware/psci/psci.c
> > @@ -164,6 +164,8 @@ int psci_set_osi_mode(bool enable)
> >  			PSCI_1_0_SUSPEND_MODE_PC;
> >
> >  	err = invoke_psci_fn(PSCI_1_0_FN_SET_SUSPEND_MODE, suspend_mode, 0, 0);
> > +	if (err < 0)
> > +		pr_warn("failed to set %s mode: %d\n", enable ? "OSI" : "PC", err);
> >  	return psci_to_linux_errno(err);
> >  }
>
> When booting 6.1-rc1 I now see:
>
> 	[    0.000000] psci: OSI mode supported.
> 	[    0.000000] psci: failed to set PC mode: -3
>
> on every boot with sc8280xp, while later enabling OSI mode still works:
>
> 	[    0.227358] CPUidle PSCI: psci_pd_try_set_osi_mode - success
> 	[    0.227599] CPUidle PSCI: Initialized CPU PM domain topology
>
> Judging from Sudeep's comment on v1:
>
> 	The platform must boot in PC mode, so even if it fails we ignore
> 	so not sure if the logging is of much help here IMO.
>
> perhaps logging this as an error is not a good idea?
>

Sorry I have forgotten about the discussion on this already. I may contradict
myself, but it the platform is misbehaving, why is it not a good idea to
log it as error. It can be ignored right if it is not an issue on the
platform ? My main worry is no one will turn on debug unless they are
debugging and new platforms with same issues may go unnoticed while we
want them to be fixed, right ?

--
Regards,
Sudeep