[PATCH] PM / devfreq: Fix build issues with devfreq disabled

[PATCH] PM / devfreq: Fix build issues with devfreq disabled

[Rob Clark]

From: Rob Clark <robdclark@chromium.org>

The existing no-op shims for when PM_DEVFREQ (or an individual governor)
only do half the job.  The governor specific config/tuning structs need
to be available to avoid compile errors in drivers using devfreq.

Fixes: 6563f60f14cb ("drm/msm/gpu: Add devfreq tuning debugfs")
Signed-off-by: Rob Clark <robdclark@chromium.org>
---
Assuming this doesn't conflict with anything else landing via another
tree, an a-b to land this via drm/msm-next would let us un-break builds.
(And also start removing "select DEVFREQ_GOV_SIMPLE_ONDEMAND"s added in
various places to try to work around this issue.)

 include/linux/devfreq.h | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/include/linux/devfreq.h b/include/linux/devfreq.h
index 4dc7cda4fd46..7fd704bb8f3d 100644
--- a/include/linux/devfreq.h
+++ b/include/linux/devfreq.h
@@ -273,8 +273,8 @@ void devm_devfreq_unregister_notifier(struct device *dev,
 struct devfreq *devfreq_get_devfreq_by_node(struct device_node *node);
 struct devfreq *devfreq_get_devfreq_by_phandle(struct device *dev,
 				const char *phandle_name, int index);
+#endif /* CONFIG_PM_DEVFREQ */
 
-#if IS_ENABLED(CONFIG_DEVFREQ_GOV_SIMPLE_ONDEMAND)
 /**
  * struct devfreq_simple_ondemand_data - ``void *data`` fed to struct devfreq
  *	and devfreq_add_device
@@ -292,9 +292,7 @@ struct devfreq_simple_ondemand_data {
 	unsigned int upthreshold;
 	unsigned int downdifferential;
 };
-#endif
 
-#if IS_ENABLED(CONFIG_DEVFREQ_GOV_PASSIVE)
 enum devfreq_parent_dev_type {
 	DEVFREQ_PARENT_DEV,
 	CPUFREQ_PARENT_DEV,
@@ -337,9 +335,8 @@ struct devfreq_passive_data {
 	struct notifier_block nb;
 	struct list_head cpu_data_list;
 };
-#endif
 
-#else /* !CONFIG_PM_DEVFREQ */
+#if !defined(CONFIG_PM_DEVFREQ)
 static inline struct devfreq *devfreq_add_device(struct device *dev,
 					struct devfreq_dev_profile *profile,
 					const char *governor_name,
-- 
2.38.1

RE: [PATCH] PM / devfreq: Fix build issues with devfreq disabled

[MyungJoo Ham]

>Sender : Rob Clark <robdclark@gmail.com>
>Date : 2023-01-24 00:37 (GMT+9)
>Title : [PATCH] PM / devfreq: Fix build issues with devfreq disabled
> 
>From: Rob Clark <robdclark@chromium.org>
>
>The existing no-op shims for when PM_DEVFREQ (or an individual governor)
>only do half the job.  The governor specific config/tuning structs need
>to be available to avoid compile errors in drivers using devfreq.
>
>Fixes: 6563f60f14cb ("drm/msm/gpu: Add devfreq tuning debugfs")
>Signed-off-by: Rob Clark <robdclark@chromium.org>

Doesn't this imply that DRM_MSM should depend on PM_DEVFREQ ?

It appears that gpu/drm/msm/DRM_MSM uses PM_DEVFREQ without actually
declaring the dependency on PM_DEVFREQ.
You cannot use SIMPLE_ONDEMAND without DEVFREQ.

Cheers,
MyungJoo

>---
>Assuming this doesn't conflict with anything else landing via another
>tree, an a-b to land this via drm/msm-next would let us un-break builds.
>(And also start removing "select DEVFREQ_GOV_SIMPLE_ONDEMAND"s added in
>various places to try to work around this issue.)
>
> include/linux/devfreq.h | 7 ++-----
> 1 file changed, 2 insertions(+), 5 deletions(-)
>
>diff --git a/include/linux/devfreq.h b/include/linux/devfreq.h
>index 4dc7cda4fd46..7fd704bb8f3d 100644
>--- a/include/linux/devfreq.h
>+++ b/include/linux/devfreq.h
>@@ -273,8 +273,8 @@ void devm_devfreq_unregister_notifier(struct device *dev,
> struct devfreq *devfreq_get_devfreq_by_node(struct device_node *node);
> struct devfreq *devfreq_get_devfreq_by_phandle(struct device *dev,
>                                 const char *phandle_name, int index);
>+#endif /* CONFIG_PM_DEVFREQ */
> 
>-#if IS_ENABLED(CONFIG_DEVFREQ_GOV_SIMPLE_ONDEMAND)
> /**
>  * struct devfreq_simple_ondemand_data - ``void *data`` fed to struct devfreq
>  *        and devfreq_add_device
>@@ -292,9 +292,7 @@ struct devfreq_simple_ondemand_data {
>         unsigned int upthreshold;
>         unsigned int downdifferential;
> };
>-#endif
> 
>-#if IS_ENABLED(CONFIG_DEVFREQ_GOV_PASSIVE)
> enum devfreq_parent_dev_type {
>         DEVFREQ_PARENT_DEV,
>         CPUFREQ_PARENT_DEV,
>@@ -337,9 +335,8 @@ struct devfreq_passive_data {
>         struct notifier_block nb;
>         struct list_head cpu_data_list;
> };
>-#endif
> 
>-#else /* !CONFIG_PM_DEVFREQ */
>+#if !defined(CONFIG_PM_DEVFREQ)
> static inline struct devfreq *devfreq_add_device(struct device *dev,
>                                         struct devfreq_dev_profile *profile,
>                                         const char *governor_name,
>-- 
>2.38.1
>

Re: [PATCH] PM / devfreq: Fix build issues with devfreq disabled

[Rob Clark]

On Tue, Jan 24, 2023 at 8:04 PM MyungJoo Ham <myungjoo.ham@samsung.com> wrote:
>
> >Sender : Rob Clark <robdclark@gmail.com>
> >Date : 2023-01-24 00:37 (GMT+9)
> >Title : [PATCH] PM / devfreq: Fix build issues with devfreq disabled
> >
> >From: Rob Clark <robdclark@chromium.org>
> >
> >The existing no-op shims for when PM_DEVFREQ (or an individual governor)
> >only do half the job.  The governor specific config/tuning structs need
> >to be available to avoid compile errors in drivers using devfreq.
> >
> >Fixes: 6563f60f14cb ("drm/msm/gpu: Add devfreq tuning debugfs")
> >Signed-off-by: Rob Clark <robdclark@chromium.org>
>
> Doesn't this imply that DRM_MSM should depend on PM_DEVFREQ ?
>
> It appears that gpu/drm/msm/DRM_MSM uses PM_DEVFREQ without actually
> declaring the dependency on PM_DEVFREQ.
> You cannot use SIMPLE_ONDEMAND without DEVFREQ.

Possibly that would resolve some issues, and that might have been my
mistake in assuming that depending on SIMPLE_ONDEMAND implied a
dependency on DEFREQ (which seems like a reasonable assumption, IMHO)

But AFAICT some kernel configs that could otherwise use DRM_MSM don't
support PM_DEVFREQ.. either way, lets solve this properly and remove
needless dependencies on devfreq.

BR,
-R


> Cheers,
> MyungJoo
>
> >---
> >Assuming this doesn't conflict with anything else landing via another
> >tree, an a-b to land this via drm/msm-next would let us un-break builds.
> >(And also start removing "select DEVFREQ_GOV_SIMPLE_ONDEMAND"s added in
> >various places to try to work around this issue.)
> >
> > include/linux/devfreq.h | 7 ++-----
> > 1 file changed, 2 insertions(+), 5 deletions(-)
> >
> >diff --git a/include/linux/devfreq.h b/include/linux/devfreq.h
> >index 4dc7cda4fd46..7fd704bb8f3d 100644
> >--- a/include/linux/devfreq.h
> >+++ b/include/linux/devfreq.h
> >@@ -273,8 +273,8 @@ void devm_devfreq_unregister_notifier(struct device *dev,
> > struct devfreq *devfreq_get_devfreq_by_node(struct device_node *node);
> > struct devfreq *devfreq_get_devfreq_by_phandle(struct device *dev,
> >                                 const char *phandle_name, int index);
> >+#endif /* CONFIG_PM_DEVFREQ */
> >
> >-#if IS_ENABLED(CONFIG_DEVFREQ_GOV_SIMPLE_ONDEMAND)
> > /**
> >  * struct devfreq_simple_ondemand_data - ``void *data`` fed to struct devfreq
> >  *        and devfreq_add_device
> >@@ -292,9 +292,7 @@ struct devfreq_simple_ondemand_data {
> >         unsigned int upthreshold;
> >         unsigned int downdifferential;
> > };
> >-#endif
> >
> >-#if IS_ENABLED(CONFIG_DEVFREQ_GOV_PASSIVE)
> > enum devfreq_parent_dev_type {
> >         DEVFREQ_PARENT_DEV,
> >         CPUFREQ_PARENT_DEV,
> >@@ -337,9 +335,8 @@ struct devfreq_passive_data {
> >         struct notifier_block nb;
> >         struct list_head cpu_data_list;
> > };
> >-#endif
> >
> >-#else /* !CONFIG_PM_DEVFREQ */
> >+#if !defined(CONFIG_PM_DEVFREQ)
> > static inline struct devfreq *devfreq_add_device(struct device *dev,
> >                                         struct devfreq_dev_profile *profile,
> >                                         const char *governor_name,
> >--
> >2.38.1
> >
>

RE: Re: [PATCH] PM / devfreq: Fix build issues with devfreq disabled

[MyungJoo Ham]

>On Tue, Jan 24, 2023 at 8:04 PM MyungJoo Ham <myungjoo.ham@samsung.com> wrote:
>>
>> >Sender : Rob Clark <robdclark@gmail.com>
>> >Date : 2023-01-24 00:37 (GMT+9)
>> >Title : [PATCH] PM / devfreq: Fix build issues with devfreq disabled
>> >
>> >From: Rob Clark <robdclark@chromium.org>
>> >
>> >The existing no-op shims for when PM_DEVFREQ (or an individual governor)
>> >only do half the job.  The governor specific config/tuning structs need
>> >to be available to avoid compile errors in drivers using devfreq.
>> >
>> >Fixes: 6563f60f14cb ("drm/msm/gpu: Add devfreq tuning debugfs")
>> >Signed-off-by: Rob Clark <robdclark@chromium.org>
>>
>> Doesn't this imply that DRM_MSM should depend on PM_DEVFREQ ?
>>
>> It appears that gpu/drm/msm/DRM_MSM uses PM_DEVFREQ without actually
>> declaring the dependency on PM_DEVFREQ.
>> You cannot use SIMPLE_ONDEMAND without DEVFREQ.
>
>Possibly that would resolve some issues, and that might have been my
>mistake in assuming that depending on SIMPLE_ONDEMAND implied a
>dependency on DEFREQ (which seems like a reasonable assumption, IMHO)
>
>But AFAICT some kernel configs that could otherwise use DRM_MSM don't
>support PM_DEVFREQ.. either way, lets solve this properly and remove
>needless dependencies on devfreq.
>
>BR,
>-R

Ok. You are enabling struct and enum only and that looks harmless.

PTAL, Chanwoo.

Acked-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Cheers,
MyungJoo.

Re: Re: [PATCH] PM / devfreq: Fix build issues with devfreq disabled

[Rob Clark]

On Tue, Jan 24, 2023 at 9:46 PM MyungJoo Ham <myungjoo.ham@samsung.com> wrote:
>
> >On Tue, Jan 24, 2023 at 8:04 PM MyungJoo Ham <myungjoo.ham@samsung.com> wrote:
> >>
> >> >Sender : Rob Clark <robdclark@gmail.com>
> >> >Date : 2023-01-24 00:37 (GMT+9)
> >> >Title : [PATCH] PM / devfreq: Fix build issues with devfreq disabled
> >> >
> >> >From: Rob Clark <robdclark@chromium.org>
> >> >
> >> >The existing no-op shims for when PM_DEVFREQ (or an individual governor)
> >> >only do half the job.  The governor specific config/tuning structs need
> >> >to be available to avoid compile errors in drivers using devfreq.
> >> >
> >> >Fixes: 6563f60f14cb ("drm/msm/gpu: Add devfreq tuning debugfs")
> >> >Signed-off-by: Rob Clark <robdclark@chromium.org>
> >>
> >> Doesn't this imply that DRM_MSM should depend on PM_DEVFREQ ?
> >>
> >> It appears that gpu/drm/msm/DRM_MSM uses PM_DEVFREQ without actually
> >> declaring the dependency on PM_DEVFREQ.
> >> You cannot use SIMPLE_ONDEMAND without DEVFREQ.
> >
> >Possibly that would resolve some issues, and that might have been my
> >mistake in assuming that depending on SIMPLE_ONDEMAND implied a
> >dependency on DEFREQ (which seems like a reasonable assumption, IMHO)
> >
> >But AFAICT some kernel configs that could otherwise use DRM_MSM don't
> >support PM_DEVFREQ.. either way, lets solve this properly and remove
> >needless dependencies on devfreq.
> >
> >BR,
> >-R
>
> Ok. You are enabling struct and enum only and that looks harmless.
>
> PTAL, Chanwoo.
>
> Acked-by: MyungJoo Ham <myungjoo.ham@samsung.com>

Thanks, if possible, an ack to land this via msm-next would avoid
build break headaches with COMPILE_TEST=y and other more obscure
setups

BR,
-R