Linux Virtualization - Apr 2017 - [PATCH 2/2] drm: virtio: fix virtio_gpu_mode_dumb_create

Lookup format using virtio_gpu_translate_format()
instead of hardcoding it.  Fixes xorg display on
bigendian guests (i.e. ppc64).

Signed-off-by: Gerd Hoffmann <kraxel at redhat.com>
---
 drivers/gpu/drm/virtio/virtgpu_gem.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/virtio/virtgpu_gem.c
b/drivers/gpu/drm/virtio/virtgpu_gem.c
index 336a57f..cc025d8 100644
--- a/drivers/gpu/drm/virtio/virtgpu_gem.c
+++ b/drivers/gpu/drm/virtio/virtgpu_gem.c
@@ -88,6 +88,7 @@ int virtio_gpu_mode_dumb_create(struct drm_file *file_priv,
 	int ret;
 	uint32_t pitch;
 	uint32_t resid;
+	uint32_t format;
 
 	pitch = args->width * ((args->bpp + 1) / 8);
 	args->size = pitch * args->height;
@@ -98,9 +99,10 @@ int virtio_gpu_mode_dumb_create(struct drm_file *file_priv,
 	if (ret)
 		goto fail;
 
+	format = virtio_gpu_translate_format(DRM_FORMAT_XRGB8888);
 	virtio_gpu_resource_id_get(vgdev, &resid);
-	virtio_gpu_cmd_create_resource(vgdev, resid,
-				       2, args->width, args->height);
+	virtio_gpu_cmd_create_resource(vgdev, resid, format,
+				       args->width, args->height);
 
 	/* attach the object to the resource */
 	obj = gem_to_virtio_gpu_obj(gobj);
-- 
2.9.3

On 3 April 2017 at 17:08, Gerd Hoffmann <kraxel at redhat.com>
wrote:
> Lookup format using virtio_gpu_translate_format()
> instead of hardcoding it.  Fixes xorg display on
> bigendian guests (i.e. ppc64).
>
> Signed-off-by: Gerd Hoffmann <kraxel at redhat.com>
For the series,

Reviewed-by: Dave Airlie <airlied at redhat.com>
> ---
>  drivers/gpu/drm/virtio/virtgpu_gem.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/virtio/virtgpu_gem.c
b/drivers/gpu/drm/virtio/virtgpu_gem.c
> index 336a57f..cc025d8 100644
> --- a/drivers/gpu/drm/virtio/virtgpu_gem.c
> +++ b/drivers/gpu/drm/virtio/virtgpu_gem.c
> @@ -88,6 +88,7 @@ int virtio_gpu_mode_dumb_create(struct drm_file
*file_priv,
>         int ret;
>         uint32_t pitch;
>         uint32_t resid;
> +       uint32_t format;
>
>         pitch = args->width * ((args->bpp + 1) / 8);
>         args->size = pitch * args->height;
> @@ -98,9 +99,10 @@ int virtio_gpu_mode_dumb_create(struct drm_file
*file_priv,
>         if (ret)
>                 goto fail;
>
> +       format = virtio_gpu_translate_format(DRM_FORMAT_XRGB8888);
>         virtio_gpu_resource_id_get(vgdev, &resid);
> -       virtio_gpu_cmd_create_resource(vgdev, resid,
> -                                      2, args->width, args->height);
> +       virtio_gpu_cmd_create_resource(vgdev, resid, format,
> +                                      args->width, args->height);
>
>         /* attach the object to the resource */
>         obj = gem_to_virtio_gpu_obj(gobj);
> --
> 2.9.3
>
> _______________________________________________
> dri-devel mailing list
> dri-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel

On Mon, Apr 03, 2017 at 09:08:45AM +0200, Gerd Hoffmann
wrote:
> Lookup format using virtio_gpu_translate_format()
> instead of hardcoding it.  Fixes xorg display on
> bigendian guests (i.e. ppc64).
> 
> Signed-off-by: Gerd Hoffmann <kraxel at redhat.com>
ow ... :(

With what kind of client have you tested this? fbdev? kms native?

Should we patch drm_fourcc.h and essentially state that _all_ drm_fourcc
are in native endian, don't dare mixing things up? Would be good to send
out an rfc and discuss it with radeon/amd maintainers. Afaik they are the
only ones who also care about some big endian platforms.

On the patch itself, once you've added more details about what/how it
falls over:

Acked-by: Daniel Vetter <daniel.vetter at ffwll.ch>
-Daniel
> ---
>  drivers/gpu/drm/virtio/virtgpu_gem.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/virtio/virtgpu_gem.c
b/drivers/gpu/drm/virtio/virtgpu_gem.c
> index 336a57f..cc025d8 100644
> --- a/drivers/gpu/drm/virtio/virtgpu_gem.c
> +++ b/drivers/gpu/drm/virtio/virtgpu_gem.c
> @@ -88,6 +88,7 @@ int virtio_gpu_mode_dumb_create(struct drm_file
*file_priv,
>  	int ret;
>  	uint32_t pitch;
>  	uint32_t resid;
> +	uint32_t format;
>  
>  	pitch = args->width * ((args->bpp + 1) / 8);
>  	args->size = pitch * args->height;
> @@ -98,9 +99,10 @@ int virtio_gpu_mode_dumb_create(struct drm_file
*file_priv,
>  	if (ret)
>  		goto fail;
>  
> +	format = virtio_gpu_translate_format(DRM_FORMAT_XRGB8888);
>  	virtio_gpu_resource_id_get(vgdev, &resid);
> -	virtio_gpu_cmd_create_resource(vgdev, resid,
> -				       2, args->width, args->height);
> +	virtio_gpu_cmd_create_resource(vgdev, resid, format,
> +				       args->width, args->height);
>  
>  	/* attach the object to the resource */
>  	obj = gem_to_virtio_gpu_obj(gobj);
> -- 
> 2.9.3
> 
> _______________________________________________
> dri-devel mailing list
> dri-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch

On Mo, 2017-04-03 at 09:53 +0200, Daniel Vetter wrote:
> On Mon, Apr 03, 2017 at 09:08:45AM +0200, Gerd Hoffmann wrote:
> > Lookup format using virtio_gpu_translate_format()
> > instead of hardcoding it.  Fixes xorg display on
> > bigendian guests (i.e. ppc64).
> > 
> > Signed-off-by: Gerd Hoffmann <kraxel at redhat.com>
> 
> ow ... :(
> 
> With what kind of client have you tested this? fbdev? kms native?
text console (fbcon) is fine, xorg (modesetting driver) fails.
> Should we patch drm_fourcc.h and essentially state that _all_ drm_fourcc
> are in native endian, don't dare mixing things up?
Well, I don't think this needs clarification.  I never assumed that
fourcc is something else than native endian.  It really was just an
oversight in the virtio driver.

One code path (which creates the framebuffer for fbdev emulation) did
proper drm_fourcc -> virtio format mapping, including endianness
handling.  Patch 1/2 of this series factors this code out into a helper
function so we can easily reuse it.

The other code path (for creating dumb buffers) had the virtio format
hardcoded which happened to work correctly in the little endian case.

cheers,
  Gerd