diff mbox series

drm/modes: Support video parameters with only reflect option

Message ID 20191209183254.211428-1-stephan@gerhold.net (mailing list archive)
State New, archived
Headers show
Series drm/modes: Support video parameters with only reflect option | expand

Commit Message

Stephan Gerhold Dec. 9, 2019, 6:32 p.m. UTC
At the moment, video mode parameters like video=540x960,reflect_x,
(without rotation set) are silently ignored.

One of the reasons for this is that the calculation that
combines the panel_orientation with cmdline->rotation_reflection
does not handle the case when cmdline->rotation_reflection does
not have any rotation set.
(i.e. cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK == 0)

Example:
  *rotation = DRM_MODE_ROTATE_0 (no panel_orientation)
  cmdline->rotation_reflection = DRM_MODE_REFLECT_X (video=MODE,reflect_x)

The current code does:
  panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
  cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
  sum_rot = (panel_rot + cmdline_rot) % 4;

and therefore:
  panel_rot = ilog2(DRM_MODE_ROTATE_0) = ilog2(1) = 0
  cmdline_rot = ilog2(0) = -1
  sum_rot = (0 + -1) % 4 = -1 % 4 = 3
   ...
  *rotation = DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X

So we incorrectly generate DRM_MODE_ROTATE_270 in this case.
To prevent cmdline_rot from becoming -1, we need to treat
the rotation as DRM_MODE_ROTATE_0.

On the other hand, there is no need to go through that calculation
at all if no rotation is set in rotation_reflection.
A simple XOR is enough to combine the reflections.

Finally, also allow DRM_MODE_ROTATE_0 in the if statement below.
DRM_MODE_ROTATE_0 means "no rotation" and should therefore not
require any special handling (e.g. specific tiling format).

This makes video parameters with only reflect option work correctly.

Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
---
 drivers/gpu/drm/drm_client_modeset.c | 27 ++++++++++++++++-----------
 1 file changed, 16 insertions(+), 11 deletions(-)

Comments

mripard@kernel.org Dec. 10, 2019, 10:20 a.m. UTC | #1
Hi,

On Mon, Dec 09, 2019 at 07:32:54PM +0100, Stephan Gerhold wrote:
> At the moment, video mode parameters like video=540x960,reflect_x,
> (without rotation set) are silently ignored.
>
> One of the reasons for this is that the calculation that
> combines the panel_orientation with cmdline->rotation_reflection
> does not handle the case when cmdline->rotation_reflection does
> not have any rotation set.
> (i.e. cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK == 0)
>
> Example:
>   *rotation = DRM_MODE_ROTATE_0 (no panel_orientation)
>   cmdline->rotation_reflection = DRM_MODE_REFLECT_X (video=MODE,reflect_x)
>
> The current code does:
>   panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
>   cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
>   sum_rot = (panel_rot + cmdline_rot) % 4;
>
> and therefore:
>   panel_rot = ilog2(DRM_MODE_ROTATE_0) = ilog2(1) = 0
>   cmdline_rot = ilog2(0) = -1
>   sum_rot = (0 + -1) % 4 = -1 % 4 = 3
>    ...
>   *rotation = DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X
>
> So we incorrectly generate DRM_MODE_ROTATE_270 in this case.
> To prevent cmdline_rot from becoming -1, we need to treat
> the rotation as DRM_MODE_ROTATE_0.
>
> On the other hand, there is no need to go through that calculation
> at all if no rotation is set in rotation_reflection.
> A simple XOR is enough to combine the reflections.
>
> Finally, also allow DRM_MODE_ROTATE_0 in the if statement below.
> DRM_MODE_ROTATE_0 means "no rotation" and should therefore not
> require any special handling (e.g. specific tiling format).
>
> This makes video parameters with only reflect option work correctly.
>
> Signed-off-by: Stephan Gerhold <stephan@gerhold.net>

Thanks for that commit message.

Can you also add a selftest to make sure we don't get a regression in
the future?

Thanks!
Maxime
Stephan Gerhold Dec. 10, 2019, 10:42 a.m. UTC | #2
On Tue, Dec 10, 2019 at 11:20:46AM +0100, Maxime Ripard wrote:
> Hi,
> 
> On Mon, Dec 09, 2019 at 07:32:54PM +0100, Stephan Gerhold wrote:
> > At the moment, video mode parameters like video=540x960,reflect_x,
> > (without rotation set) are silently ignored.
> >
> > One of the reasons for this is that the calculation that
> > combines the panel_orientation with cmdline->rotation_reflection
> > does not handle the case when cmdline->rotation_reflection does
> > not have any rotation set.
> > (i.e. cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK == 0)
> >
> > Example:
> >   *rotation = DRM_MODE_ROTATE_0 (no panel_orientation)
> >   cmdline->rotation_reflection = DRM_MODE_REFLECT_X (video=MODE,reflect_x)
> >
> > The current code does:
> >   panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
> >   cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
> >   sum_rot = (panel_rot + cmdline_rot) % 4;
> >
> > and therefore:
> >   panel_rot = ilog2(DRM_MODE_ROTATE_0) = ilog2(1) = 0
> >   cmdline_rot = ilog2(0) = -1
> >   sum_rot = (0 + -1) % 4 = -1 % 4 = 3
> >    ...
> >   *rotation = DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X
> >
> > So we incorrectly generate DRM_MODE_ROTATE_270 in this case.
> > To prevent cmdline_rot from becoming -1, we need to treat
> > the rotation as DRM_MODE_ROTATE_0.
> >
> > On the other hand, there is no need to go through that calculation
> > at all if no rotation is set in rotation_reflection.
> > A simple XOR is enough to combine the reflections.
> >
> > Finally, also allow DRM_MODE_ROTATE_0 in the if statement below.
> > DRM_MODE_ROTATE_0 means "no rotation" and should therefore not
> > require any special handling (e.g. specific tiling format).
> >
> > This makes video parameters with only reflect option work correctly.
> >
> > Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
> 
> Thanks for that commit message.
> 
> Can you also add a selftest to make sure we don't get a regression in
> the future?

Can you explain how/where I would add a test for drm_client_rotation()
in drm_client_modeset.c? I'm not familiar with selftests to be honest.

I found test-drm_cmdline_parser.c but that seems to cover only the
cmdline parsing (which is working correctly already).

Thanks,
Stephan
mripard@kernel.org Dec. 11, 2019, 6:10 p.m. UTC | #3
Hi Stephan,

On Tue, Dec 10, 2019 at 11:42:37AM +0100, Stephan Gerhold wrote:
> On Tue, Dec 10, 2019 at 11:20:46AM +0100, Maxime Ripard wrote:
> > Hi,
> >
> > On Mon, Dec 09, 2019 at 07:32:54PM +0100, Stephan Gerhold wrote:
> > > At the moment, video mode parameters like video=540x960,reflect_x,
> > > (without rotation set) are silently ignored.
> > >
> > > One of the reasons for this is that the calculation that
> > > combines the panel_orientation with cmdline->rotation_reflection
> > > does not handle the case when cmdline->rotation_reflection does
> > > not have any rotation set.
> > > (i.e. cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK == 0)
> > >
> > > Example:
> > >   *rotation = DRM_MODE_ROTATE_0 (no panel_orientation)
> > >   cmdline->rotation_reflection = DRM_MODE_REFLECT_X (video=MODE,reflect_x)
> > >
> > > The current code does:
> > >   panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
> > >   cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
> > >   sum_rot = (panel_rot + cmdline_rot) % 4;
> > >
> > > and therefore:
> > >   panel_rot = ilog2(DRM_MODE_ROTATE_0) = ilog2(1) = 0
> > >   cmdline_rot = ilog2(0) = -1
> > >   sum_rot = (0 + -1) % 4 = -1 % 4 = 3
> > >    ...
> > >   *rotation = DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X
> > >
> > > So we incorrectly generate DRM_MODE_ROTATE_270 in this case.
> > > To prevent cmdline_rot from becoming -1, we need to treat
> > > the rotation as DRM_MODE_ROTATE_0.
> > >
> > > On the other hand, there is no need to go through that calculation
> > > at all if no rotation is set in rotation_reflection.
> > > A simple XOR is enough to combine the reflections.
> > >
> > > Finally, also allow DRM_MODE_ROTATE_0 in the if statement below.
> > > DRM_MODE_ROTATE_0 means "no rotation" and should therefore not
> > > require any special handling (e.g. specific tiling format).
> > >
> > > This makes video parameters with only reflect option work correctly.
> > >
> > > Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
> >
> > Thanks for that commit message.
> >
> > Can you also add a selftest to make sure we don't get a regression in
> > the future?
>
> Can you explain how/where I would add a test for drm_client_rotation()
> in drm_client_modeset.c? I'm not familiar with selftests to be honest.
>
> I found test-drm_cmdline_parser.c but that seems to cover only the
> cmdline parsing (which is working correctly already).

The cmdline here is the kernel command line. You were mentionning in
your commit log that video=540x960,reflect_x was broken?

But yeah, I meant this one.

Maxime
Stephan Gerhold Dec. 11, 2019, 7:08 p.m. UTC | #4
On Wed, Dec 11, 2019 at 07:10:46PM +0100, Maxime Ripard wrote:
> Hi Stephan,
> 
> On Tue, Dec 10, 2019 at 11:42:37AM +0100, Stephan Gerhold wrote:
> > On Tue, Dec 10, 2019 at 11:20:46AM +0100, Maxime Ripard wrote:
> > > Hi,
> > >
> > > On Mon, Dec 09, 2019 at 07:32:54PM +0100, Stephan Gerhold wrote:
> > > > At the moment, video mode parameters like video=540x960,reflect_x,
> > > > (without rotation set) are silently ignored.
> > > >
> > > > One of the reasons for this is that the calculation that
> > > > combines the panel_orientation with cmdline->rotation_reflection
> > > > does not handle the case when cmdline->rotation_reflection does
> > > > not have any rotation set.
> > > > (i.e. cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK == 0)
> > > >
> > > > Example:
> > > >   *rotation = DRM_MODE_ROTATE_0 (no panel_orientation)
> > > >   cmdline->rotation_reflection = DRM_MODE_REFLECT_X (video=MODE,reflect_x)
> > > >
> > > > The current code does:
> > > >   panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
> > > >   cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
> > > >   sum_rot = (panel_rot + cmdline_rot) % 4;
> > > >
> > > > and therefore:
> > > >   panel_rot = ilog2(DRM_MODE_ROTATE_0) = ilog2(1) = 0
> > > >   cmdline_rot = ilog2(0) = -1
> > > >   sum_rot = (0 + -1) % 4 = -1 % 4 = 3
> > > >    ...
> > > >   *rotation = DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X
> > > >
> > > > So we incorrectly generate DRM_MODE_ROTATE_270 in this case.
> > > > To prevent cmdline_rot from becoming -1, we need to treat
> > > > the rotation as DRM_MODE_ROTATE_0.
> > > >
> > > > On the other hand, there is no need to go through that calculation
> > > > at all if no rotation is set in rotation_reflection.
> > > > A simple XOR is enough to combine the reflections.
> > > >
> > > > Finally, also allow DRM_MODE_ROTATE_0 in the if statement below.
> > > > DRM_MODE_ROTATE_0 means "no rotation" and should therefore not
> > > > require any special handling (e.g. specific tiling format).
> > > >
> > > > This makes video parameters with only reflect option work correctly.
> > > >
> > > > Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
> > >
> > > Thanks for that commit message.
> > >
> > > Can you also add a selftest to make sure we don't get a regression in
> > > the future?
> >
> > Can you explain how/where I would add a test for drm_client_rotation()
> > in drm_client_modeset.c? I'm not familiar with selftests to be honest.
> >
> > I found test-drm_cmdline_parser.c but that seems to cover only the
> > cmdline parsing (which is working correctly already).
> 
> The cmdline here is the kernel command line. You were mentionning in
> your commit log that video=540x960,reflect_x was broken?
> 

The parameter is parsed correctly and placed into connector->cmdline_mode.
Therefore, not the *parsing* is broken, only the way we try to apply
and merge them with the panel orientation in drm_client_modeset.c.

There are existing test cases for the parsing of parameters similar to
video=540x960,reflect_x, see drm_cmdline_test_hmirror()
in the aforementioned test file.

Maybe my commit message was not as clear as I hoped :)

Thanks,
Stephan
mripard@kernel.org Dec. 13, 2019, 9:39 a.m. UTC | #5
Hi Stephan,

On Wed, Dec 11, 2019 at 08:08:39PM +0100, Stephan Gerhold wrote:
> On Wed, Dec 11, 2019 at 07:10:46PM +0100, Maxime Ripard wrote:
> > On Tue, Dec 10, 2019 at 11:42:37AM +0100, Stephan Gerhold wrote:
> > > On Tue, Dec 10, 2019 at 11:20:46AM +0100, Maxime Ripard wrote:
> > > > Hi,
> > > >
> > > > On Mon, Dec 09, 2019 at 07:32:54PM +0100, Stephan Gerhold wrote:
> > > > > At the moment, video mode parameters like video=540x960,reflect_x,
> > > > > (without rotation set) are silently ignored.
> > > > >
> > > > > One of the reasons for this is that the calculation that
> > > > > combines the panel_orientation with cmdline->rotation_reflection
> > > > > does not handle the case when cmdline->rotation_reflection does
> > > > > not have any rotation set.
> > > > > (i.e. cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK == 0)
> > > > >
> > > > > Example:
> > > > >   *rotation = DRM_MODE_ROTATE_0 (no panel_orientation)
> > > > >   cmdline->rotation_reflection = DRM_MODE_REFLECT_X (video=MODE,reflect_x)
> > > > >
> > > > > The current code does:
> > > > >   panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
> > > > >   cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
> > > > >   sum_rot = (panel_rot + cmdline_rot) % 4;
> > > > >
> > > > > and therefore:
> > > > >   panel_rot = ilog2(DRM_MODE_ROTATE_0) = ilog2(1) = 0
> > > > >   cmdline_rot = ilog2(0) = -1
> > > > >   sum_rot = (0 + -1) % 4 = -1 % 4 = 3
> > > > >    ...
> > > > >   *rotation = DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X
> > > > >
> > > > > So we incorrectly generate DRM_MODE_ROTATE_270 in this case.
> > > > > To prevent cmdline_rot from becoming -1, we need to treat
> > > > > the rotation as DRM_MODE_ROTATE_0.
> > > > >
> > > > > On the other hand, there is no need to go through that calculation
> > > > > at all if no rotation is set in rotation_reflection.
> > > > > A simple XOR is enough to combine the reflections.
> > > > >
> > > > > Finally, also allow DRM_MODE_ROTATE_0 in the if statement below.
> > > > > DRM_MODE_ROTATE_0 means "no rotation" and should therefore not
> > > > > require any special handling (e.g. specific tiling format).
> > > > >
> > > > > This makes video parameters with only reflect option work correctly.
> > > > >
> > > > > Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
> > > >
> > > > Thanks for that commit message.
> > > >
> > > > Can you also add a selftest to make sure we don't get a regression in
> > > > the future?
> > >
> > > Can you explain how/where I would add a test for drm_client_rotation()
> > > in drm_client_modeset.c? I'm not familiar with selftests to be honest.
> > >
> > > I found test-drm_cmdline_parser.c but that seems to cover only the
> > > cmdline parsing (which is working correctly already).
> >
> > The cmdline here is the kernel command line. You were mentionning in
> > your commit log that video=540x960,reflect_x was broken?
> >
>
> The parameter is parsed correctly and placed into connector->cmdline_mode.
> Therefore, not the *parsing* is broken, only the way we try to apply
> and merge them with the panel orientation in drm_client_modeset.c.
>
> There are existing test cases for the parsing of parameters similar to
> video=540x960,reflect_x, see drm_cmdline_test_hmirror()
> in the aforementioned test file.
>
> Maybe my commit message was not as clear as I hoped :)

Ah, I see what you meant by "silently ignored" now :)

Yeah, maybe we can change that by "not taken into account when
applying the mode" or something like that?

Maxime
Stephan Gerhold Dec. 13, 2019, 12:16 p.m. UTC | #6
On Fri, Dec 13, 2019 at 10:39:50AM +0100, Maxime Ripard wrote:
> Hi Stephan,
> 
> On Wed, Dec 11, 2019 at 08:08:39PM +0100, Stephan Gerhold wrote:
> > On Wed, Dec 11, 2019 at 07:10:46PM +0100, Maxime Ripard wrote:
> > > On Tue, Dec 10, 2019 at 11:42:37AM +0100, Stephan Gerhold wrote:
> > > > On Tue, Dec 10, 2019 at 11:20:46AM +0100, Maxime Ripard wrote:
> > > > > Hi,
> > > > >
> > > > > On Mon, Dec 09, 2019 at 07:32:54PM +0100, Stephan Gerhold wrote:
> > > > > > At the moment, video mode parameters like video=540x960,reflect_x,
> > > > > > (without rotation set) are silently ignored.
> > > > > >
> > > > > > One of the reasons for this is that the calculation that
> > > > > > combines the panel_orientation with cmdline->rotation_reflection
> > > > > > does not handle the case when cmdline->rotation_reflection does
> > > > > > not have any rotation set.
> > > > > > (i.e. cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK == 0)
> > > > > >
> > > > > > Example:
> > > > > >   *rotation = DRM_MODE_ROTATE_0 (no panel_orientation)
> > > > > >   cmdline->rotation_reflection = DRM_MODE_REFLECT_X (video=MODE,reflect_x)
> > > > > >
> > > > > > The current code does:
> > > > > >   panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
> > > > > >   cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
> > > > > >   sum_rot = (panel_rot + cmdline_rot) % 4;
> > > > > >
> > > > > > and therefore:
> > > > > >   panel_rot = ilog2(DRM_MODE_ROTATE_0) = ilog2(1) = 0
> > > > > >   cmdline_rot = ilog2(0) = -1
> > > > > >   sum_rot = (0 + -1) % 4 = -1 % 4 = 3
> > > > > >    ...
> > > > > >   *rotation = DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X
> > > > > >
> > > > > > So we incorrectly generate DRM_MODE_ROTATE_270 in this case.
> > > > > > To prevent cmdline_rot from becoming -1, we need to treat
> > > > > > the rotation as DRM_MODE_ROTATE_0.
> > > > > >
> > > > > > On the other hand, there is no need to go through that calculation
> > > > > > at all if no rotation is set in rotation_reflection.
> > > > > > A simple XOR is enough to combine the reflections.
> > > > > >
> > > > > > Finally, also allow DRM_MODE_ROTATE_0 in the if statement below.
> > > > > > DRM_MODE_ROTATE_0 means "no rotation" and should therefore not
> > > > > > require any special handling (e.g. specific tiling format).
> > > > > >
> > > > > > This makes video parameters with only reflect option work correctly.
> > > > > >
> > > > > > Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
> > > > >
> > > > > Thanks for that commit message.
> > > > >
> > > > > Can you also add a selftest to make sure we don't get a regression in
> > > > > the future?
> > > >
> > > > Can you explain how/where I would add a test for drm_client_rotation()
> > > > in drm_client_modeset.c? I'm not familiar with selftests to be honest.
> > > >
> > > > I found test-drm_cmdline_parser.c but that seems to cover only the
> > > > cmdline parsing (which is working correctly already).
> > >
> > > The cmdline here is the kernel command line. You were mentionning in
> > > your commit log that video=540x960,reflect_x was broken?
> > >
> >
> > The parameter is parsed correctly and placed into connector->cmdline_mode.
> > Therefore, not the *parsing* is broken, only the way we try to apply
> > and merge them with the panel orientation in drm_client_modeset.c.
> >
> > There are existing test cases for the parsing of parameters similar to
> > video=540x960,reflect_x, see drm_cmdline_test_hmirror()
> > in the aforementioned test file.
> >
> > Maybe my commit message was not as clear as I hoped :)
> 
> Ah, I see what you meant by "silently ignored" now :)
> 
> Yeah, maybe we can change that by "not taken into account when
> applying the mode" or something like that?

Sure, I will try to clarify it for v2.

Anything else I should change?

Thanks,
Stephan
mripard@kernel.org Dec. 16, 2019, 4:26 p.m. UTC | #7
On Fri, Dec 13, 2019 at 01:16:19PM +0100, Stephan Gerhold wrote:
> On Fri, Dec 13, 2019 at 10:39:50AM +0100, Maxime Ripard wrote:
> > Hi Stephan,
> >
> > On Wed, Dec 11, 2019 at 08:08:39PM +0100, Stephan Gerhold wrote:
> > > On Wed, Dec 11, 2019 at 07:10:46PM +0100, Maxime Ripard wrote:
> > > > On Tue, Dec 10, 2019 at 11:42:37AM +0100, Stephan Gerhold wrote:
> > > > > On Tue, Dec 10, 2019 at 11:20:46AM +0100, Maxime Ripard wrote:
> > > > > > Hi,
> > > > > >
> > > > > > On Mon, Dec 09, 2019 at 07:32:54PM +0100, Stephan Gerhold wrote:
> > > > > > > At the moment, video mode parameters like video=540x960,reflect_x,
> > > > > > > (without rotation set) are silently ignored.
> > > > > > >
> > > > > > > One of the reasons for this is that the calculation that
> > > > > > > combines the panel_orientation with cmdline->rotation_reflection
> > > > > > > does not handle the case when cmdline->rotation_reflection does
> > > > > > > not have any rotation set.
> > > > > > > (i.e. cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK == 0)
> > > > > > >
> > > > > > > Example:
> > > > > > >   *rotation = DRM_MODE_ROTATE_0 (no panel_orientation)
> > > > > > >   cmdline->rotation_reflection = DRM_MODE_REFLECT_X (video=MODE,reflect_x)
> > > > > > >
> > > > > > > The current code does:
> > > > > > >   panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
> > > > > > >   cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
> > > > > > >   sum_rot = (panel_rot + cmdline_rot) % 4;
> > > > > > >
> > > > > > > and therefore:
> > > > > > >   panel_rot = ilog2(DRM_MODE_ROTATE_0) = ilog2(1) = 0
> > > > > > >   cmdline_rot = ilog2(0) = -1
> > > > > > >   sum_rot = (0 + -1) % 4 = -1 % 4 = 3
> > > > > > >    ...
> > > > > > >   *rotation = DRM_MODE_ROTATE_270 | DRM_MODE_REFLECT_X
> > > > > > >
> > > > > > > So we incorrectly generate DRM_MODE_ROTATE_270 in this case.
> > > > > > > To prevent cmdline_rot from becoming -1, we need to treat
> > > > > > > the rotation as DRM_MODE_ROTATE_0.
> > > > > > >
> > > > > > > On the other hand, there is no need to go through that calculation
> > > > > > > at all if no rotation is set in rotation_reflection.
> > > > > > > A simple XOR is enough to combine the reflections.
> > > > > > >
> > > > > > > Finally, also allow DRM_MODE_ROTATE_0 in the if statement below.
> > > > > > > DRM_MODE_ROTATE_0 means "no rotation" and should therefore not
> > > > > > > require any special handling (e.g. specific tiling format).
> > > > > > >
> > > > > > > This makes video parameters with only reflect option work correctly.
> > > > > > >
> > > > > > > Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
> > > > > >
> > > > > > Thanks for that commit message.
> > > > > >
> > > > > > Can you also add a selftest to make sure we don't get a regression in
> > > > > > the future?
> > > > >
> > > > > Can you explain how/where I would add a test for drm_client_rotation()
> > > > > in drm_client_modeset.c? I'm not familiar with selftests to be honest.
> > > > >
> > > > > I found test-drm_cmdline_parser.c but that seems to cover only the
> > > > > cmdline parsing (which is working correctly already).
> > > >
> > > > The cmdline here is the kernel command line. You were mentionning in
> > > > your commit log that video=540x960,reflect_x was broken?
> > > >
> > >
> > > The parameter is parsed correctly and placed into connector->cmdline_mode.
> > > Therefore, not the *parsing* is broken, only the way we try to apply
> > > and merge them with the panel orientation in drm_client_modeset.c.
> > >
> > > There are existing test cases for the parsing of parameters similar to
> > > video=540x960,reflect_x, see drm_cmdline_test_hmirror()
> > > in the aforementioned test file.
> > >
> > > Maybe my commit message was not as clear as I hoped :)
> >
> > Ah, I see what you meant by "silently ignored" now :)
> >
> > Yeah, maybe we can change that by "not taken into account when
> > applying the mode" or something like that?
>
> Sure, I will try to clarify it for v2.
>
> Anything else I should change?

The rest looks fine to me.

Thanks!
Maxime
diff mbox series

Patch

diff --git a/drivers/gpu/drm/drm_client_modeset.c b/drivers/gpu/drm/drm_client_modeset.c
index 895b73f23079..cfebce4f19a5 100644
--- a/drivers/gpu/drm/drm_client_modeset.c
+++ b/drivers/gpu/drm/drm_client_modeset.c
@@ -859,19 +859,23 @@  bool drm_client_rotation(struct drm_mode_set *modeset, unsigned int *rotation)
 	 */
 	cmdline = &connector->cmdline_mode;
 	if (cmdline->specified && cmdline->rotation_reflection) {
-		unsigned int cmdline_rest, panel_rest;
-		unsigned int cmdline_rot, panel_rot;
-		unsigned int sum_rot, sum_rest;
+		if (cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK) {
+			unsigned int cmdline_rest, panel_rest;
+			unsigned int cmdline_rot, panel_rot;
+			unsigned int sum_rot, sum_rest;
 
-		panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
-		cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
-		sum_rot = (panel_rot + cmdline_rot) % 4;
+			panel_rot = ilog2(*rotation & DRM_MODE_ROTATE_MASK);
+			cmdline_rot = ilog2(cmdline->rotation_reflection & DRM_MODE_ROTATE_MASK);
+			sum_rot = (panel_rot + cmdline_rot) % 4;
 
-		panel_rest = *rotation & ~DRM_MODE_ROTATE_MASK;
-		cmdline_rest = cmdline->rotation_reflection & ~DRM_MODE_ROTATE_MASK;
-		sum_rest = panel_rest ^ cmdline_rest;
+			panel_rest = *rotation & ~DRM_MODE_ROTATE_MASK;
+			cmdline_rest = cmdline->rotation_reflection & ~DRM_MODE_ROTATE_MASK;
+			sum_rest = panel_rest ^ cmdline_rest;
 
-		*rotation = (1 << sum_rot) | sum_rest;
+			*rotation = (1 << sum_rot) | sum_rest;
+		} else {
+			*rotation ^= cmdline->rotation_reflection;
+		}
 	}
 
 	/*
@@ -879,7 +883,8 @@  bool drm_client_rotation(struct drm_mode_set *modeset, unsigned int *rotation)
 	 * depending on the hardware this may require the framebuffer
 	 * to be in a specific tiling format.
 	 */
-	if ((*rotation & DRM_MODE_ROTATE_MASK) != DRM_MODE_ROTATE_180 ||
+	if (((*rotation & DRM_MODE_ROTATE_MASK) != DRM_MODE_ROTATE_0 &&
+	     (*rotation & DRM_MODE_ROTATE_MASK) != DRM_MODE_ROTATE_180) ||
 	    !plane->rotation_property)
 		return false;