Re: Re: Re: Re: [PATCH] drm/sun4i: Fix blend registers corruption for DE2.0/DE3.0
ср, 25 мая 2022 г. в 18:14, Jernej Škrabec :
>
> Dne sreda, 25. maj 2022 ob 16:55:56 CEST je Roman Stratiienko napisal(a):
> > вт, 24 мая 2022 г. в 22:14, Jernej Škrabec :
> > >
> > > Dne torek, 24. maj 2022 ob 19:14:35 CEST je Roman Stratiienko napisal(a):
> > > > By the way, not related to this issue:
> > > >
> > > > I cherry-picked
> > > > https://patchwork.kernel.org/project/dri-devel/patch/
> 20220424162633.12369-9-sam...@sholland.org/
> > > > and got a blank FB console on OPI3.
> > > > Can you check it please?
> > >
> > > Reply to that patch and we'll talk.
> >
> > Despite the fact that I am the original author of the patches I'm not
> > even in CC, so I can respond to this thread.
>
> Actually, many people come up with similar idea.
>
> > Is there any other way to respond to the message where you're not in CC?
>
> Of course, I download patch as mbox and import it in my e-mail client. A reply
> to it will continue the thread.
Nevermind, it was an issue from my side. I haven't done rmmod/insmod
for sun4i_tcon.ko after recompiling.
> Best regards,
> Jernej
>
> >
> > >
> > > Best regards,
> > > Jernej
> > >
> > > >
> > > > Regards,
> > > > Roman
> > > >
> > > >
> > > >
> > > > вт, 24 мая 2022 г. в 20:10, Roman Stratiienko :
> > > > >
> > > > > Please draft a test for the zpos issue you're mentioning.
> > > > >
> > > > > It's very easy to do with kmsxx using python wrapper.
> > > > >
> > > > > Or explain steps to reproduce here, I will write it by myself.
> > > > >
> > > > > Thanks.
> > > > > Regards
> > > > > Roman
> > > > >
> > > > > вт, 24 мая 2022 г. в 19:21, Jernej Škrabec :
> > > > > >
> > > > > > Dne torek, 24. maj 2022 ob 17:31:13 CEST je Roman Stratiienko
> > > napisal(a):
> > > > > > > NAK for this. Further testing showed such an approach is not
> reliable
> > > > > > > due to .atomic_update() callback called only in case planes have
> some
> > > > > > > changes.
> > > > > >
> > > > > > Additionally, I think it would be better to fix underlaying zpos
> issue
> > > first
> > > > > > (attempted many times) and then worry about blending.
> > > > > >
> > > > > > Best regards,
> > > > > > Jernej
> > > > > >
> > > > > > >
> > > > > > > вт, 24 мая 2022 г. в 16:52, Roman Stratiienko
> > > :
> > > > > > > >
> > > > > > > > Corruption happens when plane zpos is updated
> > > > > > > >
> > > > > > > > Example scenario:
> > > > > > > >
> > > > > > > > Initial frame blender state:
> > > > > > > > PLANE_ZPOS = {0, 1, 2, 3}
> > > > > > > > BLD_ROUTE = {0, 1, 2, 0}
> > > > > > > > BLD_EN = {1, 1, 1, 0}
> > > > > > > >
> > > > > > > > New frame commit (Only ZPOS has been changed):
> > > > > > > >
> > > > > > > > PLANE_ZPOS = {0->2, 1->0, 2->1, 3}
> > > > > > > >
> > > > > > > > Expected results after plane state update:
> > > > > > > > Z0 Z1 Z2 Z3
> > > > > > > > BLD_ROUTE = {1, 2, 0, 0}
> > > > > > > > BLD_EN= {1, 1, 1, 0}
> > > > > > > >
> > > > > > > > What is currently happening:
> > > > > > > >
> > > > > > > > 1. sun8i_vi_layer_enable(enabled=true, zpos=2, old_zpos=0):
> > > > > > > > BLD_ROUTE = {1->0, 1, 2->0, 0}
> > > > > > > > BLD_EN= {1->0, 1, 1->1, 0}
> > > > > > > >
> > > > > > > > 2. sun8i_ui_layer_enable(enabled=true, zpos=0, old_zpos=1):
> > > > > > > > BLD_ROUTE = {0->1, 1->0, 0, 0}
> > > > > > > > BLD_EN= {0->1, 1->0, 1, 0}
> > > > > > > >
> > > > > > > > 3. sun8i_ui_layer_enable(enabled=true, zpos=1, old_zpos=2):
> > > > > > > > BLD_ROUTE = {1, 0->2, 0->0, 0}
> > > > > > > > BLD_EN= {1, 0->2, 1->0, 0}
> > > > > > > >
> > > > > > > > After updating of all the planes we are ending up with
> BLD_EN[2]=0,
> > > > > > > > which makes this channel disabled.
> > > > > > > >
> > > > > > > > To fix this issue, clear BLEND registers before updating the
> planes
> > > > > > > > and do not clear the old state while processing every plane.
> > > > > > > >
> > > > > > > > Signed-off-by: Roman Stratiienko
> > >
> > > > > > > > ---
> > > > > > > > drivers/gpu/drm/sun4i/sun8i_mixer.c| 16 +++
> > > > > > > > drivers/gpu/drm/sun4i/sun8i_ui_layer.c | 28 +++
> > > +--
> > > > > > > > drivers/gpu/drm/sun4i/sun8i_vi_layer.c | 28 +++
> > > +--
> > > > > > > > 3 files changed, 24 insertions(+), 48 deletions(-)
> > > > > > > >
> > > > > > > > diff --git a/drivers/gpu/drm/sun4i/sun8i_mixer.c b/drivers/gpu/
> drm/
> > > sun4i/
> > > > > > sun8i_mixer.c
> > > > > > > > index f5e8aeaa3cdf..004377a000fc 100644
> > > > > > > > --- a/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > > > > +++ b/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > > > > @@ -248,6 +248,21 @@ int sun8i_mixer_drm_format_to_hw(u32
> format,
> > > u32
> > > > > > *hw_format)
> > > > > > > > return -EINVAL;
> > > > > > > > }
> > > > > > > >
> > > > > > > > +static void sun8i_atomic_begin(struct sunxi_engine *engine,
> > > > > > > > + struct drm_crtc_state *old_state)
> > > > >
Re: Re: Re: Re: [PATCH] drm/sun4i: Fix blend registers corruption for DE2.0/DE3.0
Dne sreda, 25. maj 2022 ob 16:55:56 CEST je Roman Stratiienko napisal(a):
> вт, 24 мая 2022 г. в 22:14, Jernej Škrabec :
> >
> > Dne torek, 24. maj 2022 ob 19:14:35 CEST je Roman Stratiienko napisal(a):
> > > By the way, not related to this issue:
> > >
> > > I cherry-picked
> > > https://patchwork.kernel.org/project/dri-devel/patch/
20220424162633.12369-9-sam...@sholland.org/
> > > and got a blank FB console on OPI3.
> > > Can you check it please?
> >
> > Reply to that patch and we'll talk.
>
> Despite the fact that I am the original author of the patches I'm not
> even in CC, so I can respond to this thread.
Actually, many people come up with similar idea.
> Is there any other way to respond to the message where you're not in CC?
Of course, I download patch as mbox and import it in my e-mail client. A reply
to it will continue the thread.
Best regards,
Jernej
>
> >
> > Best regards,
> > Jernej
> >
> > >
> > > Regards,
> > > Roman
> > >
> > >
> > >
> > > вт, 24 мая 2022 г. в 20:10, Roman Stratiienko :
> > > >
> > > > Please draft a test for the zpos issue you're mentioning.
> > > >
> > > > It's very easy to do with kmsxx using python wrapper.
> > > >
> > > > Or explain steps to reproduce here, I will write it by myself.
> > > >
> > > > Thanks.
> > > > Regards
> > > > Roman
> > > >
> > > > вт, 24 мая 2022 г. в 19:21, Jernej Škrabec :
> > > > >
> > > > > Dne torek, 24. maj 2022 ob 17:31:13 CEST je Roman Stratiienko
> > napisal(a):
> > > > > > NAK for this. Further testing showed such an approach is not
reliable
> > > > > > due to .atomic_update() callback called only in case planes have
some
> > > > > > changes.
> > > > >
> > > > > Additionally, I think it would be better to fix underlaying zpos
issue
> > first
> > > > > (attempted many times) and then worry about blending.
> > > > >
> > > > > Best regards,
> > > > > Jernej
> > > > >
> > > > > >
> > > > > > вт, 24 мая 2022 г. в 16:52, Roman Stratiienko
> > :
> > > > > > >
> > > > > > > Corruption happens when plane zpos is updated
> > > > > > >
> > > > > > > Example scenario:
> > > > > > >
> > > > > > > Initial frame blender state:
> > > > > > > PLANE_ZPOS = {0, 1, 2, 3}
> > > > > > > BLD_ROUTE = {0, 1, 2, 0}
> > > > > > > BLD_EN = {1, 1, 1, 0}
> > > > > > >
> > > > > > > New frame commit (Only ZPOS has been changed):
> > > > > > >
> > > > > > > PLANE_ZPOS = {0->2, 1->0, 2->1, 3}
> > > > > > >
> > > > > > > Expected results after plane state update:
> > > > > > > Z0 Z1 Z2 Z3
> > > > > > > BLD_ROUTE = {1, 2, 0, 0}
> > > > > > > BLD_EN= {1, 1, 1, 0}
> > > > > > >
> > > > > > > What is currently happening:
> > > > > > >
> > > > > > > 1. sun8i_vi_layer_enable(enabled=true, zpos=2, old_zpos=0):
> > > > > > > BLD_ROUTE = {1->0, 1, 2->0, 0}
> > > > > > > BLD_EN= {1->0, 1, 1->1, 0}
> > > > > > >
> > > > > > > 2. sun8i_ui_layer_enable(enabled=true, zpos=0, old_zpos=1):
> > > > > > > BLD_ROUTE = {0->1, 1->0, 0, 0}
> > > > > > > BLD_EN= {0->1, 1->0, 1, 0}
> > > > > > >
> > > > > > > 3. sun8i_ui_layer_enable(enabled=true, zpos=1, old_zpos=2):
> > > > > > > BLD_ROUTE = {1, 0->2, 0->0, 0}
> > > > > > > BLD_EN= {1, 0->2, 1->0, 0}
> > > > > > >
> > > > > > > After updating of all the planes we are ending up with
BLD_EN[2]=0,
> > > > > > > which makes this channel disabled.
> > > > > > >
> > > > > > > To fix this issue, clear BLEND registers before updating the
planes
> > > > > > > and do not clear the old state while processing every plane.
> > > > > > >
> > > > > > > Signed-off-by: Roman Stratiienko
> >
> > > > > > > ---
> > > > > > > drivers/gpu/drm/sun4i/sun8i_mixer.c| 16 +++
> > > > > > > drivers/gpu/drm/sun4i/sun8i_ui_layer.c | 28 +++
> > +--
> > > > > > > drivers/gpu/drm/sun4i/sun8i_vi_layer.c | 28 +++
> > +--
> > > > > > > 3 files changed, 24 insertions(+), 48 deletions(-)
> > > > > > >
> > > > > > > diff --git a/drivers/gpu/drm/sun4i/sun8i_mixer.c b/drivers/gpu/
drm/
> > sun4i/
> > > > > sun8i_mixer.c
> > > > > > > index f5e8aeaa3cdf..004377a000fc 100644
> > > > > > > --- a/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > > > +++ b/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > > > @@ -248,6 +248,21 @@ int sun8i_mixer_drm_format_to_hw(u32
format,
> > u32
> > > > > *hw_format)
> > > > > > > return -EINVAL;
> > > > > > > }
> > > > > > >
> > > > > > > +static void sun8i_atomic_begin(struct sunxi_engine *engine,
> > > > > > > + struct drm_crtc_state *old_state)
> > > > > > > +{
> > > > > > > + struct sun8i_mixer *mixer =
engine_to_sun8i_mixer(engine);
> > > > > > > + u32 bld_base = sun8i_blender_base(mixer);
> > > > > > > +
> > > > > > > + regmap_write(engine->regs,
> > > > > > > +SUN8I_MIXER_BLEND_PIPE_CTL(bld_base),
> > > > > > > +0);
> > > > > > > +
> > > > > > > + regmap_write(engine->regs,
> > > > > > > +
Re: Re: Re: [PATCH] drm/sun4i: Fix blend registers corruption for DE2.0/DE3.0
вт, 24 мая 2022 г. в 22:14, Jernej Škrabec :
>
> Dne torek, 24. maj 2022 ob 19:14:35 CEST je Roman Stratiienko napisal(a):
> > By the way, not related to this issue:
> >
> > I cherry-picked
> > https://patchwork.kernel.org/project/dri-devel/patch/20220424162633.12369-9-sam...@sholland.org/
> > and got a blank FB console on OPI3.
> > Can you check it please?
>
> Reply to that patch and we'll talk.
Despite the fact that I am the original author of the patches I'm not
even in CC, so I can respond to this thread.
Is there any other way to respond to the message where you're not in CC?
>
> Best regards,
> Jernej
>
> >
> > Regards,
> > Roman
> >
> >
> >
> > вт, 24 мая 2022 г. в 20:10, Roman Stratiienko :
> > >
> > > Please draft a test for the zpos issue you're mentioning.
> > >
> > > It's very easy to do with kmsxx using python wrapper.
> > >
> > > Or explain steps to reproduce here, I will write it by myself.
> > >
> > > Thanks.
> > > Regards
> > > Roman
> > >
> > > вт, 24 мая 2022 г. в 19:21, Jernej Škrabec :
> > > >
> > > > Dne torek, 24. maj 2022 ob 17:31:13 CEST je Roman Stratiienko
> napisal(a):
> > > > > NAK for this. Further testing showed such an approach is not reliable
> > > > > due to .atomic_update() callback called only in case planes have some
> > > > > changes.
> > > >
> > > > Additionally, I think it would be better to fix underlaying zpos issue
> first
> > > > (attempted many times) and then worry about blending.
> > > >
> > > > Best regards,
> > > > Jernej
> > > >
> > > > >
> > > > > вт, 24 мая 2022 г. в 16:52, Roman Stratiienko
> :
> > > > > >
> > > > > > Corruption happens when plane zpos is updated
> > > > > >
> > > > > > Example scenario:
> > > > > >
> > > > > > Initial frame blender state:
> > > > > > PLANE_ZPOS = {0, 1, 2, 3}
> > > > > > BLD_ROUTE = {0, 1, 2, 0}
> > > > > > BLD_EN = {1, 1, 1, 0}
> > > > > >
> > > > > > New frame commit (Only ZPOS has been changed):
> > > > > >
> > > > > > PLANE_ZPOS = {0->2, 1->0, 2->1, 3}
> > > > > >
> > > > > > Expected results after plane state update:
> > > > > > Z0 Z1 Z2 Z3
> > > > > > BLD_ROUTE = {1, 2, 0, 0}
> > > > > > BLD_EN= {1, 1, 1, 0}
> > > > > >
> > > > > > What is currently happening:
> > > > > >
> > > > > > 1. sun8i_vi_layer_enable(enabled=true, zpos=2, old_zpos=0):
> > > > > > BLD_ROUTE = {1->0, 1, 2->0, 0}
> > > > > > BLD_EN= {1->0, 1, 1->1, 0}
> > > > > >
> > > > > > 2. sun8i_ui_layer_enable(enabled=true, zpos=0, old_zpos=1):
> > > > > > BLD_ROUTE = {0->1, 1->0, 0, 0}
> > > > > > BLD_EN= {0->1, 1->0, 1, 0}
> > > > > >
> > > > > > 3. sun8i_ui_layer_enable(enabled=true, zpos=1, old_zpos=2):
> > > > > > BLD_ROUTE = {1, 0->2, 0->0, 0}
> > > > > > BLD_EN= {1, 0->2, 1->0, 0}
> > > > > >
> > > > > > After updating of all the planes we are ending up with BLD_EN[2]=0,
> > > > > > which makes this channel disabled.
> > > > > >
> > > > > > To fix this issue, clear BLEND registers before updating the planes
> > > > > > and do not clear the old state while processing every plane.
> > > > > >
> > > > > > Signed-off-by: Roman Stratiienko
>
> > > > > > ---
> > > > > > drivers/gpu/drm/sun4i/sun8i_mixer.c| 16 +++
> > > > > > drivers/gpu/drm/sun4i/sun8i_ui_layer.c | 28 +++
> +--
> > > > > > drivers/gpu/drm/sun4i/sun8i_vi_layer.c | 28 +++
> +--
> > > > > > 3 files changed, 24 insertions(+), 48 deletions(-)
> > > > > >
> > > > > > diff --git a/drivers/gpu/drm/sun4i/sun8i_mixer.c b/drivers/gpu/drm/
> sun4i/
> > > > sun8i_mixer.c
> > > > > > index f5e8aeaa3cdf..004377a000fc 100644
> > > > > > --- a/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > > +++ b/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > > @@ -248,6 +248,21 @@ int sun8i_mixer_drm_format_to_hw(u32 format,
> u32
> > > > *hw_format)
> > > > > > return -EINVAL;
> > > > > > }
> > > > > >
> > > > > > +static void sun8i_atomic_begin(struct sunxi_engine *engine,
> > > > > > + struct drm_crtc_state *old_state)
> > > > > > +{
> > > > > > + struct sun8i_mixer *mixer = engine_to_sun8i_mixer(engine);
> > > > > > + u32 bld_base = sun8i_blender_base(mixer);
> > > > > > +
> > > > > > + regmap_write(engine->regs,
> > > > > > +SUN8I_MIXER_BLEND_PIPE_CTL(bld_base),
> > > > > > +0);
> > > > > > +
> > > > > > + regmap_write(engine->regs,
> > > > > > +SUN8I_MIXER_BLEND_ROUTE(bld_base),
> > > > > > +0);
> > > > > > +}
> > > > > > +
> > > > > > static void sun8i_mixer_commit(struct sunxi_engine *engine)
> > > > > > {
> > > > > > DRM_DEBUG_DRIVER("Committing changes\n");
> > > > > > @@ -299,6 +314,7 @@ static struct drm_plane
> **sun8i_layers_init(struct
> > > > drm_device *drm,
> > > > > > }
> > > > > >
> > > > > > static const struct sunxi_engine_ops sun8i_engine_ops = {
> > > > > > + .atomic_begin =
Re: Re: Re: [PATCH] drm/sun4i: Fix blend registers corruption for DE2.0/DE3.0
вт, 24 мая 2022 г. в 22:13, Jernej Škrabec :
>
> Don't top post, it's annoying and against rules.
>
> Dne torek, 24. maj 2022 ob 19:10:06 CEST je Roman Stratiienko napisal(a):
> > Please draft a test for the zpos issue you're mentioning.
> >
> > It's very easy to do with kmsxx using python wrapper.
> >
> > Or explain steps to reproduce here, I will write it by myself.
>
> I'm talking about the issue which you, Ondrej Jirman and me all tried to fix
> it
> in the past one way or another:
>
> https://patchwork.kernel.org/project/dri-devel/patch/20190914220337.646719-1-meg...@megous.com/
> https://patchwork.kernel.org/project/dri-devel/patch/20210106204630.1800284-1-jernej.skra...@siol.net/
This particular patch and the one I've sent today [1] aims to fix the
ZPOS issue you're talking about.
Second patch [1] fixes https://github.com/rsglobal/kmsxx/pull/3 test
and unlike this patch it does not cause additional regressions (as far
as I can test).
[1]:
https://patchwork.kernel.org/project/dri-devel/patch/20220525115445.93500-1-roman.o.stratiie...@globallogic.com/
>
> Best regards,
> Jernej
>
> >
> > Thanks.
> > Regards
> > Roman
> >
> > вт, 24 мая 2022 г. в 19:21, Jernej Škrabec :
> > >
> > > Dne torek, 24. maj 2022 ob 17:31:13 CEST je Roman Stratiienko napisal(a):
> > > > NAK for this. Further testing showed such an approach is not reliable
> > > > due to .atomic_update() callback called only in case planes have some
> > > > changes.
> > >
> > > Additionally, I think it would be better to fix underlaying zpos issue
> > > first
> > > (attempted many times) and then worry about blending.
> > >
> > > Best regards,
> > > Jernej
> > >
> > > >
> > > > вт, 24 мая 2022 г. в 16:52, Roman Stratiienko :
> > > > >
> > > > > Corruption happens when plane zpos is updated
> > > > >
> > > > > Example scenario:
> > > > >
> > > > > Initial frame blender state:
> > > > > PLANE_ZPOS = {0, 1, 2, 3}
> > > > > BLD_ROUTE = {0, 1, 2, 0}
> > > > > BLD_EN = {1, 1, 1, 0}
> > > > >
> > > > > New frame commit (Only ZPOS has been changed):
> > > > >
> > > > > PLANE_ZPOS = {0->2, 1->0, 2->1, 3}
> > > > >
> > > > > Expected results after plane state update:
> > > > > Z0 Z1 Z2 Z3
> > > > > BLD_ROUTE = {1, 2, 0, 0}
> > > > > BLD_EN= {1, 1, 1, 0}
> > > > >
> > > > > What is currently happening:
> > > > >
> > > > > 1. sun8i_vi_layer_enable(enabled=true, zpos=2, old_zpos=0):
> > > > > BLD_ROUTE = {1->0, 1, 2->0, 0}
> > > > > BLD_EN= {1->0, 1, 1->1, 0}
> > > > >
> > > > > 2. sun8i_ui_layer_enable(enabled=true, zpos=0, old_zpos=1):
> > > > > BLD_ROUTE = {0->1, 1->0, 0, 0}
> > > > > BLD_EN= {0->1, 1->0, 1, 0}
> > > > >
> > > > > 3. sun8i_ui_layer_enable(enabled=true, zpos=1, old_zpos=2):
> > > > > BLD_ROUTE = {1, 0->2, 0->0, 0}
> > > > > BLD_EN= {1, 0->2, 1->0, 0}
> > > > >
> > > > > After updating of all the planes we are ending up with BLD_EN[2]=0,
> > > > > which makes this channel disabled.
> > > > >
> > > > > To fix this issue, clear BLEND registers before updating the planes
> > > > > and do not clear the old state while processing every plane.
> > > > >
> > > > > Signed-off-by: Roman Stratiienko
> > > > > ---
> > > > > drivers/gpu/drm/sun4i/sun8i_mixer.c| 16 +++
> > > > > drivers/gpu/drm/sun4i/sun8i_ui_layer.c | 28 +++
> +--
> > > > > drivers/gpu/drm/sun4i/sun8i_vi_layer.c | 28 +++
> +--
> > > > > 3 files changed, 24 insertions(+), 48 deletions(-)
> > > > >
> > > > > diff --git a/drivers/gpu/drm/sun4i/sun8i_mixer.c b/drivers/gpu/drm/
> sun4i/
> > > sun8i_mixer.c
> > > > > index f5e8aeaa3cdf..004377a000fc 100644
> > > > > --- a/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > +++ b/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > @@ -248,6 +248,21 @@ int sun8i_mixer_drm_format_to_hw(u32 format, u32
> > > *hw_format)
> > > > > return -EINVAL;
> > > > > }
> > > > >
> > > > > +static void sun8i_atomic_begin(struct sunxi_engine *engine,
> > > > > + struct drm_crtc_state *old_state)
> > > > > +{
> > > > > + struct sun8i_mixer *mixer = engine_to_sun8i_mixer(engine);
> > > > > + u32 bld_base = sun8i_blender_base(mixer);
> > > > > +
> > > > > + regmap_write(engine->regs,
> > > > > +SUN8I_MIXER_BLEND_PIPE_CTL(bld_base),
> > > > > +0);
> > > > > +
> > > > > + regmap_write(engine->regs,
> > > > > +SUN8I_MIXER_BLEND_ROUTE(bld_base),
> > > > > +0);
> > > > > +}
> > > > > +
> > > > > static void sun8i_mixer_commit(struct sunxi_engine *engine)
> > > > > {
> > > > > DRM_DEBUG_DRIVER("Committing changes\n");
> > > > > @@ -299,6 +314,7 @@ static struct drm_plane **sun8i_layers_init(struct
> > > drm_device *drm,
> > > > > }
> > > > >
> > > > > static const struct sunxi_engine_ops sun8i_engine_ops = {
> > > > > + .atomic_begin = sun8i_atomic_begin,
> > > > >
Re: Re: Re: [PATCH] drm/sun4i: Fix blend registers corruption for DE2.0/DE3.0
Dne torek, 24. maj 2022 ob 19:14:35 CEST je Roman Stratiienko napisal(a):
> By the way, not related to this issue:
>
> I cherry-picked
> https://patchwork.kernel.org/project/dri-devel/patch/20220424162633.12369-9-sam...@sholland.org/
> and got a blank FB console on OPI3.
> Can you check it please?
Reply to that patch and we'll talk.
Best regards,
Jernej
>
> Regards,
> Roman
>
>
>
> вт, 24 мая 2022 г. в 20:10, Roman Stratiienko :
> >
> > Please draft a test for the zpos issue you're mentioning.
> >
> > It's very easy to do with kmsxx using python wrapper.
> >
> > Or explain steps to reproduce here, I will write it by myself.
> >
> > Thanks.
> > Regards
> > Roman
> >
> > вт, 24 мая 2022 г. в 19:21, Jernej Škrabec :
> > >
> > > Dne torek, 24. maj 2022 ob 17:31:13 CEST je Roman Stratiienko
napisal(a):
> > > > NAK for this. Further testing showed such an approach is not reliable
> > > > due to .atomic_update() callback called only in case planes have some
> > > > changes.
> > >
> > > Additionally, I think it would be better to fix underlaying zpos issue
first
> > > (attempted many times) and then worry about blending.
> > >
> > > Best regards,
> > > Jernej
> > >
> > > >
> > > > вт, 24 мая 2022 г. в 16:52, Roman Stratiienko
:
> > > > >
> > > > > Corruption happens when plane zpos is updated
> > > > >
> > > > > Example scenario:
> > > > >
> > > > > Initial frame blender state:
> > > > > PLANE_ZPOS = {0, 1, 2, 3}
> > > > > BLD_ROUTE = {0, 1, 2, 0}
> > > > > BLD_EN = {1, 1, 1, 0}
> > > > >
> > > > > New frame commit (Only ZPOS has been changed):
> > > > >
> > > > > PLANE_ZPOS = {0->2, 1->0, 2->1, 3}
> > > > >
> > > > > Expected results after plane state update:
> > > > > Z0 Z1 Z2 Z3
> > > > > BLD_ROUTE = {1, 2, 0, 0}
> > > > > BLD_EN= {1, 1, 1, 0}
> > > > >
> > > > > What is currently happening:
> > > > >
> > > > > 1. sun8i_vi_layer_enable(enabled=true, zpos=2, old_zpos=0):
> > > > > BLD_ROUTE = {1->0, 1, 2->0, 0}
> > > > > BLD_EN= {1->0, 1, 1->1, 0}
> > > > >
> > > > > 2. sun8i_ui_layer_enable(enabled=true, zpos=0, old_zpos=1):
> > > > > BLD_ROUTE = {0->1, 1->0, 0, 0}
> > > > > BLD_EN= {0->1, 1->0, 1, 0}
> > > > >
> > > > > 3. sun8i_ui_layer_enable(enabled=true, zpos=1, old_zpos=2):
> > > > > BLD_ROUTE = {1, 0->2, 0->0, 0}
> > > > > BLD_EN= {1, 0->2, 1->0, 0}
> > > > >
> > > > > After updating of all the planes we are ending up with BLD_EN[2]=0,
> > > > > which makes this channel disabled.
> > > > >
> > > > > To fix this issue, clear BLEND registers before updating the planes
> > > > > and do not clear the old state while processing every plane.
> > > > >
> > > > > Signed-off-by: Roman Stratiienko
> > > > > ---
> > > > > drivers/gpu/drm/sun4i/sun8i_mixer.c| 16 +++
> > > > > drivers/gpu/drm/sun4i/sun8i_ui_layer.c | 28 +++
+--
> > > > > drivers/gpu/drm/sun4i/sun8i_vi_layer.c | 28 +++
+--
> > > > > 3 files changed, 24 insertions(+), 48 deletions(-)
> > > > >
> > > > > diff --git a/drivers/gpu/drm/sun4i/sun8i_mixer.c b/drivers/gpu/drm/
sun4i/
> > > sun8i_mixer.c
> > > > > index f5e8aeaa3cdf..004377a000fc 100644
> > > > > --- a/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > +++ b/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > > @@ -248,6 +248,21 @@ int sun8i_mixer_drm_format_to_hw(u32 format,
u32
> > > *hw_format)
> > > > > return -EINVAL;
> > > > > }
> > > > >
> > > > > +static void sun8i_atomic_begin(struct sunxi_engine *engine,
> > > > > + struct drm_crtc_state *old_state)
> > > > > +{
> > > > > + struct sun8i_mixer *mixer = engine_to_sun8i_mixer(engine);
> > > > > + u32 bld_base = sun8i_blender_base(mixer);
> > > > > +
> > > > > + regmap_write(engine->regs,
> > > > > +SUN8I_MIXER_BLEND_PIPE_CTL(bld_base),
> > > > > +0);
> > > > > +
> > > > > + regmap_write(engine->regs,
> > > > > +SUN8I_MIXER_BLEND_ROUTE(bld_base),
> > > > > +0);
> > > > > +}
> > > > > +
> > > > > static void sun8i_mixer_commit(struct sunxi_engine *engine)
> > > > > {
> > > > > DRM_DEBUG_DRIVER("Committing changes\n");
> > > > > @@ -299,6 +314,7 @@ static struct drm_plane
**sun8i_layers_init(struct
> > > drm_device *drm,
> > > > > }
> > > > >
> > > > > static const struct sunxi_engine_ops sun8i_engine_ops = {
> > > > > + .atomic_begin = sun8i_atomic_begin,
> > > > > .commit = sun8i_mixer_commit,
> > > > > .layers_init= sun8i_layers_init,
> > > > > };
> > > > > diff --git a/drivers/gpu/drm/sun4i/sun8i_ui_layer.c b/drivers/gpu/
drm/
> > > sun4i/sun8i_ui_layer.c
> > > > > index 7845c2a53a7f..b294a882626a 100644
> > > > > --- a/drivers/gpu/drm/sun4i/sun8i_ui_layer.c
> > > > > +++ b/drivers/gpu/drm/sun4i/sun8i_ui_layer.c
> > > > > @@ -24,8 +24,7 @@
> > > > > #include "sun8i_ui_scaler.h"
> > > > >
> > > > >
Re: Re: Re: [PATCH] drm/sun4i: Fix blend registers corruption for DE2.0/DE3.0
Don't top post, it's annoying and against rules.
Dne torek, 24. maj 2022 ob 19:10:06 CEST je Roman Stratiienko napisal(a):
> Please draft a test for the zpos issue you're mentioning.
>
> It's very easy to do with kmsxx using python wrapper.
>
> Or explain steps to reproduce here, I will write it by myself.
I'm talking about the issue which you, Ondrej Jirman and me all tried to fix it
in the past one way or another:
https://patchwork.kernel.org/project/dri-devel/patch/20190914220337.646719-1-meg...@megous.com/
https://patchwork.kernel.org/project/dri-devel/patch/20210106204630.1800284-1-jernej.skra...@siol.net/
Best regards,
Jernej
>
> Thanks.
> Regards
> Roman
>
> вт, 24 мая 2022 г. в 19:21, Jernej Škrabec :
> >
> > Dne torek, 24. maj 2022 ob 17:31:13 CEST je Roman Stratiienko napisal(a):
> > > NAK for this. Further testing showed such an approach is not reliable
> > > due to .atomic_update() callback called only in case planes have some
> > > changes.
> >
> > Additionally, I think it would be better to fix underlaying zpos issue first
> > (attempted many times) and then worry about blending.
> >
> > Best regards,
> > Jernej
> >
> > >
> > > вт, 24 мая 2022 г. в 16:52, Roman Stratiienko :
> > > >
> > > > Corruption happens when plane zpos is updated
> > > >
> > > > Example scenario:
> > > >
> > > > Initial frame blender state:
> > > > PLANE_ZPOS = {0, 1, 2, 3}
> > > > BLD_ROUTE = {0, 1, 2, 0}
> > > > BLD_EN = {1, 1, 1, 0}
> > > >
> > > > New frame commit (Only ZPOS has been changed):
> > > >
> > > > PLANE_ZPOS = {0->2, 1->0, 2->1, 3}
> > > >
> > > > Expected results after plane state update:
> > > > Z0 Z1 Z2 Z3
> > > > BLD_ROUTE = {1, 2, 0, 0}
> > > > BLD_EN= {1, 1, 1, 0}
> > > >
> > > > What is currently happening:
> > > >
> > > > 1. sun8i_vi_layer_enable(enabled=true, zpos=2, old_zpos=0):
> > > > BLD_ROUTE = {1->0, 1, 2->0, 0}
> > > > BLD_EN= {1->0, 1, 1->1, 0}
> > > >
> > > > 2. sun8i_ui_layer_enable(enabled=true, zpos=0, old_zpos=1):
> > > > BLD_ROUTE = {0->1, 1->0, 0, 0}
> > > > BLD_EN= {0->1, 1->0, 1, 0}
> > > >
> > > > 3. sun8i_ui_layer_enable(enabled=true, zpos=1, old_zpos=2):
> > > > BLD_ROUTE = {1, 0->2, 0->0, 0}
> > > > BLD_EN= {1, 0->2, 1->0, 0}
> > > >
> > > > After updating of all the planes we are ending up with BLD_EN[2]=0,
> > > > which makes this channel disabled.
> > > >
> > > > To fix this issue, clear BLEND registers before updating the planes
> > > > and do not clear the old state while processing every plane.
> > > >
> > > > Signed-off-by: Roman Stratiienko
> > > > ---
> > > > drivers/gpu/drm/sun4i/sun8i_mixer.c| 16 +++
> > > > drivers/gpu/drm/sun4i/sun8i_ui_layer.c | 28 +++
+--
> > > > drivers/gpu/drm/sun4i/sun8i_vi_layer.c | 28 +++
+--
> > > > 3 files changed, 24 insertions(+), 48 deletions(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/sun4i/sun8i_mixer.c b/drivers/gpu/drm/
sun4i/
> > sun8i_mixer.c
> > > > index f5e8aeaa3cdf..004377a000fc 100644
> > > > --- a/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > +++ b/drivers/gpu/drm/sun4i/sun8i_mixer.c
> > > > @@ -248,6 +248,21 @@ int sun8i_mixer_drm_format_to_hw(u32 format, u32
> > *hw_format)
> > > > return -EINVAL;
> > > > }
> > > >
> > > > +static void sun8i_atomic_begin(struct sunxi_engine *engine,
> > > > + struct drm_crtc_state *old_state)
> > > > +{
> > > > + struct sun8i_mixer *mixer = engine_to_sun8i_mixer(engine);
> > > > + u32 bld_base = sun8i_blender_base(mixer);
> > > > +
> > > > + regmap_write(engine->regs,
> > > > +SUN8I_MIXER_BLEND_PIPE_CTL(bld_base),
> > > > +0);
> > > > +
> > > > + regmap_write(engine->regs,
> > > > +SUN8I_MIXER_BLEND_ROUTE(bld_base),
> > > > +0);
> > > > +}
> > > > +
> > > > static void sun8i_mixer_commit(struct sunxi_engine *engine)
> > > > {
> > > > DRM_DEBUG_DRIVER("Committing changes\n");
> > > > @@ -299,6 +314,7 @@ static struct drm_plane **sun8i_layers_init(struct
> > drm_device *drm,
> > > > }
> > > >
> > > > static const struct sunxi_engine_ops sun8i_engine_ops = {
> > > > + .atomic_begin = sun8i_atomic_begin,
> > > > .commit = sun8i_mixer_commit,
> > > > .layers_init= sun8i_layers_init,
> > > > };
> > > > diff --git a/drivers/gpu/drm/sun4i/sun8i_ui_layer.c b/drivers/gpu/drm/
> > sun4i/sun8i_ui_layer.c
> > > > index 7845c2a53a7f..b294a882626a 100644
> > > > --- a/drivers/gpu/drm/sun4i/sun8i_ui_layer.c
> > > > +++ b/drivers/gpu/drm/sun4i/sun8i_ui_layer.c
> > > > @@ -24,8 +24,7 @@
> > > > #include "sun8i_ui_scaler.h"
> > > >
> > > > static void sun8i_ui_layer_enable(struct sun8i_mixer *mixer, int
channel,
> > > > - int overlay, bool enable, unsigned
int
> > zpos,
> > > > - unsigned int
