drm/vc4: Allow fb modifiers early enough to fill IN_FORMATS property

Submitted by Paul Kocialkowski on Dec. 24, 2018, 12:25 p.m.

Details

Message ID 20181224122551.9912-1-paul.kocialkowski@bootlin.com
State New
Headers show
Series "drm/vc4: Allow fb modifiers early enough to fill IN_FORMATS property" ( rev: 1 ) in DRI devel

Not browsing as part of any series.

Commit Message

Paul Kocialkowski Dec. 24, 2018, 12:25 p.m.
The KMS mode_config elements are currently configured in vc4_kms_load,
that is called after all components are binded (component_bind_all).
However, the CRTC component (for the Pixel Valve) needs to access the
allow_fb_modifiers element at bind time, when initializing its planes
through drm_universal_plane_init.

This helpers checks allow_fb_modifiers to decide whether to fill the
IN_FORMATS property. Because allow_fb_modifiers is still set to false
at this point, the property is never filled and userspace cannot
retrieve the combination of supported formats and modifiers.

Fix this by setting allow_fb_modifiers right after calling
drm_mode_config_init (which initializes the structure), before binding
the components of the driver.

Fixes: 423ad7b3cbd1 ("drm/vc4: Advertise supported modifiers for planes")
Signed-off-by: Paul Kocialkowski <paul.kocialkowski@bootlin.com>
---
 drivers/gpu/drm/vc4/vc4_drv.c | 1 +
 drivers/gpu/drm/vc4/vc4_kms.c | 1 -
 2 files changed, 1 insertion(+), 1 deletion(-)

Patch hide | download patch | download mbox

diff --git a/drivers/gpu/drm/vc4/vc4_drv.c b/drivers/gpu/drm/vc4/vc4_drv.c
index 7195a0bcceb3..5c24f80dd34e 100644
--- a/drivers/gpu/drm/vc4/vc4_drv.c
+++ b/drivers/gpu/drm/vc4/vc4_drv.c
@@ -273,6 +273,7 @@  static int vc4_drm_bind(struct device *dev)
 		goto dev_put;
 
 	drm_mode_config_init(drm);
+	drm->mode_config.allow_fb_modifiers = true;
 
 	vc4_gem_init(drm);
 
diff --git a/drivers/gpu/drm/vc4/vc4_kms.c b/drivers/gpu/drm/vc4/vc4_kms.c
index b0cbb869f659..b70931fd8cf0 100644
--- a/drivers/gpu/drm/vc4/vc4_kms.c
+++ b/drivers/gpu/drm/vc4/vc4_kms.c
@@ -530,7 +530,6 @@  int vc4_kms_load(struct drm_device *dev)
 	dev->mode_config.funcs = &vc4_mode_funcs;
 	dev->mode_config.preferred_depth = 24;
 	dev->mode_config.async_page_flip = true;
-	dev->mode_config.allow_fb_modifiers = true;
 
 	drm_modeset_lock_init(&vc4->ctm_state_lock);
 

Comments

Paul Kocialkowski <paul.kocialkowski@bootlin.com> writes:

> The KMS mode_config elements are currently configured in vc4_kms_load,
> that is called after all components are binded (component_bind_all).
> However, the CRTC component (for the Pixel Valve) needs to access the
> allow_fb_modifiers element at bind time, when initializing its planes
> through drm_universal_plane_init.
>
> This helpers checks allow_fb_modifiers to decide whether to fill the
> IN_FORMATS property. Because allow_fb_modifiers is still set to false
> at this point, the property is never filled and userspace cannot
> retrieve the combination of supported formats and modifiers.
>
> Fix this by setting allow_fb_modifiers right after calling
> drm_mode_config_init (which initializes the structure), before binding
> the components of the driver.

This makes me wonder if the flag could be removed and replaced with "did
non-NULL modifiers get supplied to plane init?"  I think I've tripped
over this flag in other KMS hacking, too.
On Thu, Dec 27, 2018 at 03:36:20PM -0800, Eric Anholt wrote:
> Paul Kocialkowski <paul.kocialkowski@bootlin.com> writes:
> 
> > The KMS mode_config elements are currently configured in vc4_kms_load,
> > that is called after all components are binded (component_bind_all).
> > However, the CRTC component (for the Pixel Valve) needs to access the
> > allow_fb_modifiers element at bind time, when initializing its planes
> > through drm_universal_plane_init.
> >
> > This helpers checks allow_fb_modifiers to decide whether to fill the
> > IN_FORMATS property. Because allow_fb_modifiers is still set to false
> > at this point, the property is never filled and userspace cannot
> > retrieve the combination of supported formats and modifiers.
> >
> > Fix this by setting allow_fb_modifiers right after calling
> > drm_mode_config_init (which initializes the structure), before binding
> > the components of the driver.
> 
> This makes me wonder if the flag could be removed and replaced with "did
> non-NULL modifiers get supplied to plane init?"  I think I've tripped
> over this flag in other KMS hacking, too.

Sounds like a good idea to me. Just setting it in plane_init should work I
think ...
-Daniel
On Fri, 2018-12-28 at 13:08 +0100, Daniel Vetter wrote:
> On Thu, Dec 27, 2018 at 03:36:20PM -0800, Eric Anholt wrote:
> > Paul Kocialkowski <paul.kocialkowski@bootlin.com> writes:
> > 
> > > The KMS mode_config elements are currently configured in vc4_kms_load,
> > > that is called after all components are binded (component_bind_all).
> > > However, the CRTC component (for the Pixel Valve) needs to access the
> > > allow_fb_modifiers element at bind time, when initializing its planes
> > > through drm_universal_plane_init.
> > > 
> > > This helpers checks allow_fb_modifiers to decide whether to fill the
> > > IN_FORMATS property. Because allow_fb_modifiers is still set to false
> > > at this point, the property is never filled and userspace cannot
> > > retrieve the combination of supported formats and modifiers.
> > > 
> > > Fix this by setting allow_fb_modifiers right after calling
> > > drm_mode_config_init (which initializes the structure), before binding
> > > the components of the driver.
> > 
> > This makes me wonder if the flag could be removed and replaced with "did
> > non-NULL modifiers get supplied to plane init?"  I think I've tripped
> > over this flag in other KMS hacking, too.
> 
> Sounds like a good idea to me. Just setting it in plane_init should work I
> think ...

Yeah, we probably need to keep the flag around since it's used in
various places, but we could totally have it auto-set as soon as a
plane is registered with a list of modifiers.

I was also thinking of allowing the core to fill-in the IN_FORMATS prop
even without this flag set, which would only list formats (but not
modifiers).

What do you think?

Cheers,

Paul