diff mbox

drm/atomic: Don't reject reflect-only rotations

Message ID 1481113099-4788-1-git-send-email-brian.starkey@arm.com (mailing list archive)
State New, archived
Headers show

Commit Message

Brian Starkey Dec. 7, 2016, 12:18 p.m. UTC
The check to reject combinations of multiple rotation angles is overly
restrictive and has the side-effect of also failing any rotation value
which consists only of reflections.

Fix this by relaxing the check to ignore values which contain no
rotation flags.

Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
Signed-off-by: Brian Starkey <brian.starkey@arm.com>
---
 drivers/gpu/drm/drm_atomic.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Comments

Gustavo Padovan Dec. 7, 2016, 1:20 p.m. UTC | #1
Hi Brian,

2016-12-07 Brian Starkey <brian.starkey@arm.com>:

> The check to reject combinations of multiple rotation angles is overly
> restrictive and has the side-effect of also failing any rotation value
> which consists only of reflections.
> 
> Fix this by relaxing the check to ignore values which contain no
> rotation flags.
> 
> Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
> Signed-off-by: Brian Starkey <brian.starkey@arm.com>
> ---
>  drivers/gpu/drm/drm_atomic.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)

Reviewed-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>

Gustavo
Chris Wilson Dec. 7, 2016, 1:32 p.m. UTC | #2
On Wed, Dec 07, 2016 at 12:18:19PM +0000, Brian Starkey wrote:
> The check to reject combinations of multiple rotation angles is overly
> restrictive and has the side-effect of also failing any rotation value
> which consists only of reflections.
> 
> Fix this by relaxing the check to ignore values which contain no
> rotation flags.
> 
> Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
> Signed-off-by: Brian Starkey <brian.starkey@arm.com>

>  drivers/gpu/drm/drm_atomic.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
> index 362e3ea..44f4030 100644
> --- a/drivers/gpu/drm/drm_atomic.c
> +++ b/drivers/gpu/drm/drm_atomic.c
> @@ -846,7 +846,8 @@ int drm_atomic_plane_set_property(struct drm_plane *plane,
>  	} else if (property == config->prop_src_h) {
>  		state->src_h = val;
>  	} else if (property == plane->rotation_property) {
> -		if (!is_power_of_2(val & DRM_ROTATE_MASK))
> +		if ((val & DRM_ROTATE_MASK) &&

If val & DRM_ROTATE_MASK is zero, val must be REFLECT_X | REFLECT_Y.
-Chris
Ville Syrjala Dec. 7, 2016, 2:12 p.m. UTC | #3
On Wed, Dec 07, 2016 at 12:18:19PM +0000, Brian Starkey wrote:
> The check to reject combinations of multiple rotation angles is overly
> restrictive and has the side-effect of also failing any rotation value
> which consists only of reflections.
> 
> Fix this by relaxing the check to ignore values which contain no
> rotation flags.
> 
> Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
> Signed-off-by: Brian Starkey <brian.starkey@arm.com>
> ---
>  drivers/gpu/drm/drm_atomic.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
> index 362e3ea..44f4030 100644
> --- a/drivers/gpu/drm/drm_atomic.c
> +++ b/drivers/gpu/drm/drm_atomic.c
> @@ -846,7 +846,8 @@ int drm_atomic_plane_set_property(struct drm_plane *plane,
>  	} else if (property == config->prop_src_h) {
>  		state->src_h = val;
>  	} else if (property == plane->rotation_property) {
> -		if (!is_power_of_2(val & DRM_ROTATE_MASK))
> +		if ((val & DRM_ROTATE_MASK) &&
> +		    !is_power_of_2(val & DRM_ROTATE_MASK))

NAK. You must always set one and only one rotation bit.

>  			return -EINVAL;
>  		state->rotation = val;
>  	} else if (property == plane->zpos_property) {
> -- 
> 1.7.9.5
Brian Starkey Dec. 7, 2016, 2:48 p.m. UTC | #4
On Wed, Dec 07, 2016 at 04:12:07PM +0200, Ville Syrjälä wrote:
>On Wed, Dec 07, 2016 at 12:18:19PM +0000, Brian Starkey wrote:
>> The check to reject combinations of multiple rotation angles is overly
>> restrictive and has the side-effect of also failing any rotation value
>> which consists only of reflections.
>>
>> Fix this by relaxing the check to ignore values which contain no
>> rotation flags.
>>
>> Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
>> Signed-off-by: Brian Starkey <brian.starkey@arm.com>
>> ---
>>  drivers/gpu/drm/drm_atomic.c |    3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
>> index 362e3ea..44f4030 100644
>> --- a/drivers/gpu/drm/drm_atomic.c
>> +++ b/drivers/gpu/drm/drm_atomic.c
>> @@ -846,7 +846,8 @@ int drm_atomic_plane_set_property(struct drm_plane *plane,
>>  	} else if (property == config->prop_src_h) {
>>  		state->src_h = val;
>>  	} else if (property == plane->rotation_property) {
>> -		if (!is_power_of_2(val & DRM_ROTATE_MASK))
>> +		if ((val & DRM_ROTATE_MASK) &&
>> +		    !is_power_of_2(val & DRM_ROTATE_MASK))
>
>NAK. You must always set one and only one rotation bit.
>

How do you do (DRM_ROTATE_90 | DRM_REFLECT_X) then?

Could you suggest the correct fix? Because right now a commit which
sets rotation = "reflect-x" fails on this check

Thanks,
Brian

>>  			return -EINVAL;
>>  		state->rotation = val;
>>  	} else if (property == plane->zpos_property) {
>> --
>> 1.7.9.5
>
>-- 
>Ville Syrjälä
>Intel OTC
Daniel Vetter Dec. 7, 2016, 2:52 p.m. UTC | #5
On Wed, Dec 07, 2016 at 01:32:57PM +0000, Chris Wilson wrote:
> On Wed, Dec 07, 2016 at 12:18:19PM +0000, Brian Starkey wrote:
> > The check to reject combinations of multiple rotation angles is overly
> > restrictive and has the side-effect of also failing any rotation value
> > which consists only of reflections.
> > 
> > Fix this by relaxing the check to ignore values which contain no
> > rotation flags.
> > 
> > Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
> > Signed-off-by: Brian Starkey <brian.starkey@arm.com>
> 
> >  drivers/gpu/drm/drm_atomic.c |    3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
> > index 362e3ea..44f4030 100644
> > --- a/drivers/gpu/drm/drm_atomic.c
> > +++ b/drivers/gpu/drm/drm_atomic.c
> > @@ -846,7 +846,8 @@ int drm_atomic_plane_set_property(struct drm_plane *plane,
> >  	} else if (property == config->prop_src_h) {
> >  		state->src_h = val;
> >  	} else if (property == plane->rotation_property) {
> > -		if (!is_power_of_2(val & DRM_ROTATE_MASK))
> > +		if ((val & DRM_ROTATE_MASK) &&
> 
> If val & DRM_ROTATE_MASK is zero, val must be REFLECT_X | REFLECT_Y.

While we bikeshed this check: Validation like this should be somewhere
behind ->atomic_check, since if it's only here then you can sneak invalid
stuff in through the legacy/compat set_property ioctls.
-Daniel
Ville Syrjala Dec. 7, 2016, 3:13 p.m. UTC | #6
On Wed, Dec 07, 2016 at 03:52:29PM +0100, Daniel Vetter wrote:
> On Wed, Dec 07, 2016 at 01:32:57PM +0000, Chris Wilson wrote:
> > On Wed, Dec 07, 2016 at 12:18:19PM +0000, Brian Starkey wrote:
> > > The check to reject combinations of multiple rotation angles is overly
> > > restrictive and has the side-effect of also failing any rotation value
> > > which consists only of reflections.
> > > 
> > > Fix this by relaxing the check to ignore values which contain no
> > > rotation flags.
> > > 
> > > Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
> > > Signed-off-by: Brian Starkey <brian.starkey@arm.com>
> > 
> > >  drivers/gpu/drm/drm_atomic.c |    3 ++-
> > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
> > > index 362e3ea..44f4030 100644
> > > --- a/drivers/gpu/drm/drm_atomic.c
> > > +++ b/drivers/gpu/drm/drm_atomic.c
> > > @@ -846,7 +846,8 @@ int drm_atomic_plane_set_property(struct drm_plane *plane,
> > >  	} else if (property == config->prop_src_h) {
> > >  		state->src_h = val;
> > >  	} else if (property == plane->rotation_property) {
> > > -		if (!is_power_of_2(val & DRM_ROTATE_MASK))
> > > +		if ((val & DRM_ROTATE_MASK) &&
> > 
> > If val & DRM_ROTATE_MASK is zero, val must be REFLECT_X | REFLECT_Y.
> 
> While we bikeshed this check: Validation like this should be somewhere
> behind ->atomic_check, since if it's only here then you can sneak invalid
> stuff in through the legacy/compat set_property ioctls.

Not if you use drm_atomic_helper_plane_set_property().
Daniel Vetter Dec. 7, 2016, 3:54 p.m. UTC | #7
On Wed, Dec 07, 2016 at 05:13:24PM +0200, Ville Syrjälä wrote:
> On Wed, Dec 07, 2016 at 03:52:29PM +0100, Daniel Vetter wrote:
> > On Wed, Dec 07, 2016 at 01:32:57PM +0000, Chris Wilson wrote:
> > > On Wed, Dec 07, 2016 at 12:18:19PM +0000, Brian Starkey wrote:
> > > > The check to reject combinations of multiple rotation angles is overly
> > > > restrictive and has the side-effect of also failing any rotation value
> > > > which consists only of reflections.
> > > > 
> > > > Fix this by relaxing the check to ignore values which contain no
> > > > rotation flags.
> > > > 
> > > > Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
> > > > Signed-off-by: Brian Starkey <brian.starkey@arm.com>
> > > 
> > > >  drivers/gpu/drm/drm_atomic.c |    3 ++-
> > > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > > > 
> > > > diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
> > > > index 362e3ea..44f4030 100644
> > > > --- a/drivers/gpu/drm/drm_atomic.c
> > > > +++ b/drivers/gpu/drm/drm_atomic.c
> > > > @@ -846,7 +846,8 @@ int drm_atomic_plane_set_property(struct drm_plane *plane,
> > > >  	} else if (property == config->prop_src_h) {
> > > >  		state->src_h = val;
> > > >  	} else if (property == plane->rotation_property) {
> > > > -		if (!is_power_of_2(val & DRM_ROTATE_MASK))
> > > > +		if ((val & DRM_ROTATE_MASK) &&
> > > 
> > > If val & DRM_ROTATE_MASK is zero, val must be REFLECT_X | REFLECT_Y.
> > 
> > While we bikeshed this check: Validation like this should be somewhere
> > behind ->atomic_check, since if it's only here then you can sneak invalid
> > stuff in through the legacy/compat set_property ioctls.
> 
> Not if you use drm_atomic_helper_plane_set_property().

Oh right, I mixed up the callchain maze in my head again. I still feel
like putting all the validation code in one place might be a good idea.
But yeah not strictly needed.
-Daniel
Ville Syrjala Dec. 7, 2016, 4:02 p.m. UTC | #8
On Wed, Dec 07, 2016 at 04:54:40PM +0100, Daniel Vetter wrote:
> On Wed, Dec 07, 2016 at 05:13:24PM +0200, Ville Syrjälä wrote:
> > On Wed, Dec 07, 2016 at 03:52:29PM +0100, Daniel Vetter wrote:
> > > On Wed, Dec 07, 2016 at 01:32:57PM +0000, Chris Wilson wrote:
> > > > On Wed, Dec 07, 2016 at 12:18:19PM +0000, Brian Starkey wrote:
> > > > > The check to reject combinations of multiple rotation angles is overly
> > > > > restrictive and has the side-effect of also failing any rotation value
> > > > > which consists only of reflections.
> > > > > 
> > > > > Fix this by relaxing the check to ignore values which contain no
> > > > > rotation flags.
> > > > > 
> > > > > Fixes: 6e0c7c3358d4 ("drm/atomic: Reject attempts to use multiple rotation angles at once")
> > > > > Signed-off-by: Brian Starkey <brian.starkey@arm.com>
> > > > 
> > > > >  drivers/gpu/drm/drm_atomic.c |    3 ++-
> > > > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > > > > 
> > > > > diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
> > > > > index 362e3ea..44f4030 100644
> > > > > --- a/drivers/gpu/drm/drm_atomic.c
> > > > > +++ b/drivers/gpu/drm/drm_atomic.c
> > > > > @@ -846,7 +846,8 @@ int drm_atomic_plane_set_property(struct drm_plane *plane,
> > > > >  	} else if (property == config->prop_src_h) {
> > > > >  		state->src_h = val;
> > > > >  	} else if (property == plane->rotation_property) {
> > > > > -		if (!is_power_of_2(val & DRM_ROTATE_MASK))
> > > > > +		if ((val & DRM_ROTATE_MASK) &&
> > > > 
> > > > If val & DRM_ROTATE_MASK is zero, val must be REFLECT_X | REFLECT_Y.
> > > 
> > > While we bikeshed this check: Validation like this should be somewhere
> > > behind ->atomic_check, since if it's only here then you can sneak invalid
> > > stuff in through the legacy/compat set_property ioctls.
> > 
> > Not if you use drm_atomic_helper_plane_set_property().
> 
> Oh right, I mixed up the callchain maze in my head again. I still feel
> like putting all the validation code in one place might be a good idea.
> But yeah not strictly needed.

Well, we do quite a bit of checking already earlier. Eg.
drm_property_change_valid(), and obj ID lookups. I think this sort of
"totally invalid use of the API" check fits in pretty well with that
crowd.
diff mbox

Patch

diff --git a/drivers/gpu/drm/drm_atomic.c b/drivers/gpu/drm/drm_atomic.c
index 362e3ea..44f4030 100644
--- a/drivers/gpu/drm/drm_atomic.c
+++ b/drivers/gpu/drm/drm_atomic.c
@@ -846,7 +846,8 @@  int drm_atomic_plane_set_property(struct drm_plane *plane,
 	} else if (property == config->prop_src_h) {
 		state->src_h = val;
 	} else if (property == plane->rotation_property) {
-		if (!is_power_of_2(val & DRM_ROTATE_MASK))
+		if ((val & DRM_ROTATE_MASK) &&
+		    !is_power_of_2(val & DRM_ROTATE_MASK))
 			return -EINVAL;
 		state->rotation = val;
 	} else if (property == plane->zpos_property) {