Planes without an alpha property, using __drm_atomic_helper_plane_reset
will have their plane state alpha initialised as zero, which represents
a transparent alpha.
If this value is then used for the plane, it may not be visible by
default, and thus doesn't represent a good initialisation state.
Update the default state->alpha value to DRM_BLEND_ALPHA_OPAQUE
unconditionally when the plane is reset.
Signed-off-by: Kieran Bingham <[email protected]>
---
drivers/gpu/drm/drm_atomic_helper.c | 4 +---
1 file changed, 1 insertion(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
index 3cf1aa132778..e49b22381048 100644
--- a/drivers/gpu/drm/drm_atomic_helper.c
+++ b/drivers/gpu/drm/drm_atomic_helper.c
@@ -3569,9 +3569,7 @@ void __drm_atomic_helper_plane_reset(struct drm_plane *plane,
state->plane = plane;
state->rotation = DRM_MODE_ROTATE_0;
- /* Reset the alpha value to fully opaque if it matters */
- if (plane->alpha_property)
- state->alpha = plane->alpha_property->values[1];
+ state->alpha = DRM_BLEND_ALPHA_OPAQUE;
state->pixel_blend_mode = DRM_MODE_BLEND_PREMULTI;
plane->state = state;
--
2.17.1
On Wed, Sep 19, 2018 at 04:56:58PM +0100, Kieran Bingham wrote:
> Planes without an alpha property, using __drm_atomic_helper_plane_reset
> will have their plane state alpha initialised as zero, which represents
> a transparent alpha.
>
> If this value is then used for the plane, it may not be visible by
> default, and thus doesn't represent a good initialisation state.
>
> Update the default state->alpha value to DRM_BLEND_ALPHA_OPAQUE
> unconditionally when the plane is reset.
>
> Signed-off-by: Kieran Bingham <[email protected]>
> ---
> drivers/gpu/drm/drm_atomic_helper.c | 4 +---
> 1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
> index 3cf1aa132778..e49b22381048 100644
> --- a/drivers/gpu/drm/drm_atomic_helper.c
> +++ b/drivers/gpu/drm/drm_atomic_helper.c
> @@ -3569,9 +3569,7 @@ void __drm_atomic_helper_plane_reset(struct drm_plane *plane,
> state->plane = plane;
> state->rotation = DRM_MODE_ROTATE_0;
>
> - /* Reset the alpha value to fully opaque if it matters */
> - if (plane->alpha_property)
> - state->alpha = plane->alpha_property->values[1];
> + state->alpha = DRM_BLEND_ALPHA_OPAQUE;
I can't come up with a solid excuse for not initializing it always.
Reviewed-by: Ville Syrj?l? <[email protected]>
> state->pixel_blend_mode = DRM_MODE_BLEND_PREMULTI;
>
> plane->state = state;
> --
> 2.17.1
>
> _______________________________________________
> dri-devel mailing list
> [email protected]
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
--
Ville Syrj?l?
Intel
Hi Kieran,
On Wed, Sep 19, 2018 at 07:15:45PM +0300, Ville Syrj?l? wrote:
> On Wed, Sep 19, 2018 at 04:56:58PM +0100, Kieran Bingham wrote:
> > Planes without an alpha property, using __drm_atomic_helper_plane_reset
> > will have their plane state alpha initialised as zero, which represents
> > a transparent alpha.
> >
> > If this value is then used for the plane, it may not be visible by
> > default, and thus doesn't represent a good initialisation state.
> >
> > Update the default state->alpha value to DRM_BLEND_ALPHA_OPAQUE
> > unconditionally when the plane is reset.
> >
> > Signed-off-by: Kieran Bingham <[email protected]>
> > ---
> > drivers/gpu/drm/drm_atomic_helper.c | 4 +---
> > 1 file changed, 1 insertion(+), 3 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
> > index 3cf1aa132778..e49b22381048 100644
> > --- a/drivers/gpu/drm/drm_atomic_helper.c
> > +++ b/drivers/gpu/drm/drm_atomic_helper.c
> > @@ -3569,9 +3569,7 @@ void __drm_atomic_helper_plane_reset(struct drm_plane *plane,
> > state->plane = plane;
> > state->rotation = DRM_MODE_ROTATE_0;
> >
> > - /* Reset the alpha value to fully opaque if it matters */
> > - if (plane->alpha_property)
> > - state->alpha = plane->alpha_property->values[1];
> > + state->alpha = DRM_BLEND_ALPHA_OPAQUE;
>
> I can't come up with a solid excuse for not initializing it always.
>
> Reviewed-by: Ville Syrj?l? <[email protected]>
Neither do I, so:
Reviewed-by: Alexandru Gheorghe <[email protected]>
And thanks again.
I plan to push it tomorrow to drm-misc-next.
Now, I've seen the plane_reset patches in the pull request for drm-next
4.20, I wonder if someone could tell me what should I do to get this
patch on that train.
>
> > state->pixel_blend_mode = DRM_MODE_BLEND_PREMULTI;
> >
> > plane->state = state;
> > --
> > 2.17.1
> >
> > _______________________________________________
> > dri-devel mailing list
> > [email protected]
> > https://lists.freedesktop.org/mailman/listinfo/dri-devel
>
> --
> Ville Syrj?l?
> Intel
--
Cheers,
Alex G
Hi Alexandru,
On 19/09/18 17:43, Alexandru-Cosmin Gheorghe wrote:
> Hi Kieran,
>
>
> On Wed, Sep 19, 2018 at 07:15:45PM +0300, Ville Syrjälä wrote:
>> On Wed, Sep 19, 2018 at 04:56:58PM +0100, Kieran Bingham wrote:
>>> Planes without an alpha property, using __drm_atomic_helper_plane_reset
>>> will have their plane state alpha initialised as zero, which represents
>>> a transparent alpha.
>>>
>>> If this value is then used for the plane, it may not be visible by
>>> default, and thus doesn't represent a good initialisation state.
>>>
>>> Update the default state->alpha value to DRM_BLEND_ALPHA_OPAQUE
>>> unconditionally when the plane is reset.
>>>
>>> Signed-off-by: Kieran Bingham <[email protected]>
>>> ---
>>> drivers/gpu/drm/drm_atomic_helper.c | 4 +---
>>> 1 file changed, 1 insertion(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
>>> index 3cf1aa132778..e49b22381048 100644
>>> --- a/drivers/gpu/drm/drm_atomic_helper.c
>>> +++ b/drivers/gpu/drm/drm_atomic_helper.c
>>> @@ -3569,9 +3569,7 @@ void __drm_atomic_helper_plane_reset(struct drm_plane *plane,
>>> state->plane = plane;
>>> state->rotation = DRM_MODE_ROTATE_0;
>>>
>>> - /* Reset the alpha value to fully opaque if it matters */
>>> - if (plane->alpha_property)
>>> - state->alpha = plane->alpha_property->values[1];
>>> + state->alpha = DRM_BLEND_ALPHA_OPAQUE;
>>
>> I can't come up with a solid excuse for not initializing it always.
>>
>> Reviewed-by: Ville Syrjälä <[email protected]>
>
> Neither do I, so:
> Reviewed-by: Alexandru Gheorghe <[email protected]>
>
> And thanks again.
>
> I plan to push it tomorrow to drm-misc-next.
>
> Now, I've seen the plane_reset patches in the pull request for drm-next
> 4.20, I wonder if someone could tell me what should I do to get this
> patch on that train.
I've submitted a separate patch for the rcar-du which enables the alpha
property for the primary plane - and it incorporates a "Fixes:
161ad653d6c9 ("drm: rcar-du: Use __drm_atomic_helper_plane_reset
instead of copying the logic")" tag.
Technically that should be sufficient to get that fix into v4.19 I
believe ...
But if you feel that this patch should also be included - we could add
the same tag to this patch, and get it queued up for v4.19 fixes?
--
Regards
Kieran
>
>>
>>> state->pixel_blend_mode = DRM_MODE_BLEND_PREMULTI;
>>>
>>> plane->state = state;
>>> --
>>> 2.17.1
>>>
>>> _______________________________________________
>>> dri-devel mailing list
>>> [email protected]
>>> https://lists.freedesktop.org/mailman/listinfo/dri-devel
>>
>> --
>> Ville Syrjälä
>> Intel
>
On Thu, Sep 20, 2018 at 11:03:12AM +0100, Kieran Bingham wrote:
> Hi Alexandru,
>
> On 19/09/18 17:43, Alexandru-Cosmin Gheorghe wrote:
> > Hi Kieran,
> >
> >
> > On Wed, Sep 19, 2018 at 07:15:45PM +0300, Ville Syrj?l? wrote:
> >> On Wed, Sep 19, 2018 at 04:56:58PM +0100, Kieran Bingham wrote:
> >>> Planes without an alpha property, using __drm_atomic_helper_plane_reset
> >>> will have their plane state alpha initialised as zero, which represents
> >>> a transparent alpha.
> >>>
> >>> If this value is then used for the plane, it may not be visible by
> >>> default, and thus doesn't represent a good initialisation state.
> >>>
> >>> Update the default state->alpha value to DRM_BLEND_ALPHA_OPAQUE
> >>> unconditionally when the plane is reset.
> >>>
> >>> Signed-off-by: Kieran Bingham <[email protected]>
> >>> ---
> >>> drivers/gpu/drm/drm_atomic_helper.c | 4 +---
> >>> 1 file changed, 1 insertion(+), 3 deletions(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/drm_atomic_helper.c b/drivers/gpu/drm/drm_atomic_helper.c
> >>> index 3cf1aa132778..e49b22381048 100644
> >>> --- a/drivers/gpu/drm/drm_atomic_helper.c
> >>> +++ b/drivers/gpu/drm/drm_atomic_helper.c
> >>> @@ -3569,9 +3569,7 @@ void __drm_atomic_helper_plane_reset(struct drm_plane *plane,
> >>> state->plane = plane;
> >>> state->rotation = DRM_MODE_ROTATE_0;
> >>>
> >>> - /* Reset the alpha value to fully opaque if it matters */
> >>> - if (plane->alpha_property)
> >>> - state->alpha = plane->alpha_property->values[1];
> >>> + state->alpha = DRM_BLEND_ALPHA_OPAQUE;
> >>
> >> I can't come up with a solid excuse for not initializing it always.
> >>
> >> Reviewed-by: Ville Syrj?l? <[email protected]>
> >
> > Neither do I, so:
> > Reviewed-by: Alexandru Gheorghe <[email protected]>
> >
> > And thanks again.
> >
> > I plan to push it tomorrow to drm-misc-next.
> >
> > Now, I've seen the plane_reset patches in the pull request for drm-next
> > 4.20, I wonder if someone could tell me what should I do to get this
> > patch on that train.
>
> I've submitted a separate patch for the rcar-du which enables the alpha
> property for the primary plane - and it incorporates a "Fixes:
> 161ad653d6c9 ("drm: rcar-du: Use __drm_atomic_helper_plane_reset
> instead of copying the logic")" tag.
>
> Technically that should be sufficient to get that fix into v4.19 I
> believe ...
plane_reset patches are not in v4.19, but in drm-next which as far as
I know will be sent for v4.20
>
> But if you feel that this patch should also be included - we could add
> the same tag to this patch, and get it queued up for v4.19 fixes?
Looking here
https://01.org/linuxgraphics/gfx-docs/maintainer-tools/repositories.html#drm-misc-next
It seems that the last pull request for v4.20 of drm-misc-next will be
sent around rc6, so I think putting it in drm-misc-next will be enough
to get this with the rest of the plane_reset patch in v4.20, but just
to make sure I will double check with Sean on IRC.
>
> --
> Regards
>
> Kieran
>
>
> >
> >>
> >>> state->pixel_blend_mode = DRM_MODE_BLEND_PREMULTI;
> >>>
> >>> plane->state = state;
> >>> --
> >>> 2.17.1
> >>>
> >>> _______________________________________________
> >>> dri-devel mailing list
> >>> [email protected]
> >>> https://lists.freedesktop.org/mailman/listinfo/dri-devel
> >>
> >> --
> >> Ville Syrj?l?
> >> Intel
> >
--
Cheers,
Alex G
Hi Kieran,
Thank you for the patch.
On Wednesday, 19 September 2018 18:56:58 EEST Kieran Bingham wrote:
> Planes without an alpha property, using __drm_atomic_helper_plane_reset
> will have their plane state alpha initialised as zero, which represents
> a transparent alpha.
>
> If this value is then used for the plane, it may not be visible by
> default, and thus doesn't represent a good initialisation state.
>
> Update the default state->alpha value to DRM_BLEND_ALPHA_OPAQUE
> unconditionally when the plane is reset.
>
> Signed-off-by: Kieran Bingham <[email protected]>
I believe the decision to use plane->alpha_property->values[1] instead of
hardcoding DRM_BLEND_ALPHA_OPAQUE comes from earlier versions of the alpha
patch series that supported driver-specific ranges for the alpha value. The
current implementation uses DRM_BLEND_ALPHA_OPAQUE unconditionally, and no
driver modifies the maximum value behind the scene, so
Reviewed-by: Laurent Pinchart <[email protected]>
> ---
> drivers/gpu/drm/drm_atomic_helper.c | 4 +---
> 1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_atomic_helper.c
> b/drivers/gpu/drm/drm_atomic_helper.c index 3cf1aa132778..e49b22381048
> 100644
> --- a/drivers/gpu/drm/drm_atomic_helper.c
> +++ b/drivers/gpu/drm/drm_atomic_helper.c
> @@ -3569,9 +3569,7 @@ void __drm_atomic_helper_plane_reset(struct drm_plane
> *plane, state->plane = plane;
> state->rotation = DRM_MODE_ROTATE_0;
>
> - /* Reset the alpha value to fully opaque if it matters */
> - if (plane->alpha_property)
> - state->alpha = plane->alpha_property->values[1];
> + state->alpha = DRM_BLEND_ALPHA_OPAQUE;
> state->pixel_blend_mode = DRM_MODE_BLEND_PREMULTI;
>
> plane->state = state;
--
Regards,
Laurent Pinchart