Re: [PATCH 4/6] drm/format-helper: Support the AB24 format

[Thierry Reding <thierry.reding@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 3384 bytes --]

On Wed, Sep 07, 2022 at 09:23:01AM +0200, Thomas Zimmermann wrote:
> Hi
> 
> Am 05.09.22 um 18:32 schrieb Thierry Reding:
> > From: Thierry Reding <treding@nvidia.com>
> > 
> > Add a conversion helper for the AB24 format to use in drm_fb_blit().
> > 
> > Signed-off-by: Thierry Reding <treding@nvidia.com>
> > ---
> >   drivers/gpu/drm/drm_format_helper.c | 35 +++++++++++++++++++++++++++++
> >   1 file changed, 35 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/drm_format_helper.c b/drivers/gpu/drm/drm_format_helper.c
> > index 56642816fdff..d564412a816b 100644
> > --- a/drivers/gpu/drm/drm_format_helper.c
> > +++ b/drivers/gpu/drm/drm_format_helper.c
> > @@ -503,6 +503,36 @@ static void drm_fb_rgb888_to_xrgb8888(struct iosys_map *dst, const unsigned int
> >   		    drm_fb_rgb888_to_xrgb8888_line);
> >   }
> > +static void drm_fb_xrgb8888_to_abgr8888_line(void *dbuf, const void *sbuf, unsigned int pixels)
> > +{
> > +	__le32 *dbuf32 = dbuf;
> > +	const __le32 *sbuf32 = sbuf;
> > +	unsigned int x;
> > +	u32 pix;
> > +
> > +	for (x = 0; x < pixels; x++) {
> > +		pix = le32_to_cpu(sbuf32[x]);
> > +		pix = ((pix & 0xff000000) >> 24) << 24 |
> > +		      ((pix & 0x00ff0000) >> 16) <<  0 |
> > +		      ((pix & 0x0000ff00) >>  8) <<  8 |
> > +		      ((pix & 0x000000ff) >>  0) << 16;
> > +		*dbuf32++ = cpu_to_le32(pix);
> > +	}
> > +}
> 
> What does the Jetson device do with these alpha bits?
> 
> AFAIK the X channel's content is undefined. Shifting the bits into the A
> channel might result in wrong results in the general case. Better just set
> the alpha to 0xff unconditionally.

I'm not exactly sure, so I'd have to do some digging. I suspect that
perhaps this might actually be XBGR rather than ABGR, so this may end up
doing nothing. EFI FB reports this as ABGR, at least if I read the
output right, so that's what I went with for the DT implementation.

> > +
> > +static void drm_fb_xrgb8888_to_abgr8888(struct iosys_map *dst, const unsigned int *dst_pitch,
> > +					const struct iosys_map *src,
> > +					const struct drm_framebuffer *fb,
> > +					const struct drm_rect *clip)
> > +{
> > +	static const u8 dst_pixsize[DRM_FORMAT_MAX_PLANES] = {
> > +		4,
> > +	};
> > +
> > +	drm_fb_xfrm(dst, dst_pitch, dst_pixsize, src, fb, clip, false,
> > +		    drm_fb_xrgb8888_to_abgr8888_line);
> > +}
> > +
> >   static void drm_fb_xrgb8888_to_xrgb2101010_line(void *dbuf, const void *sbuf, unsigned int pixels)
> >   {
> >   	__le32 *dbuf32 = dbuf;
> > @@ -672,6 +702,11 @@ int drm_fb_blit(struct iosys_map *dst, const unsigned int *dst_pitch, uint32_t d
> >   			drm_fb_rgb565_to_xrgb8888(dst, dst_pitch, src, fb, clip);
> >   			return 0;
> >   		}
> > +	} else if (dst_format == DRM_FORMAT_ABGR8888) {
> > +		if (fb_format == DRM_FORMAT_XRGB8888) {
> > +			drm_fb_xrgb8888_to_abgr8888(dst, dst_pitch, src, fb, clip);
> > +			return 0;
> > +		}
> 
> For the other alpha-containing formats, we treat them like non-alpha formats
> (see the top of this function). Maybe just do the same here and then
> implement the conversion as drm_fb_xrgb8888_to_xbgr8888() helpers?

Yeah, that's probably for the best. None of the boot splash uses alpha
as far as I can tell and neither does the kernel framebuffer console, so
there's really little point in trying to support it.

Thierry

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]