Re: [PATCH v3] drm/mediatek: ovl: Add fmt_convert function pointer to driver data

["Jason-JH Lin (林睿祥)" <Jason-JH.Lin@mediatek.com>]

On Mon, 2024-09-23 at 03:51 +0000, CK Hu (胡俊光) wrote:
> Hi, Jason:
> 
> On Wed, 2024-09-18 at 00:44 +0800, Jason-JH.Lin wrote:
> > OVL_CON_CLRFMT_MAN is a configuration for extending color format
> > settings of DISP_REG_OVL_CON(n).
> > It will change some of the original color format settings.
> > 
> > Take the settings of (3 << 12) for example.
> > - If OVL_CON_CLRFMT_MAN = 0 means OVL_CON_CLRFMT_RGBA8888.
> > - If OVL_CON_CLRFMT_MAN = 1 means OVL_CON_CLRFMT_PARGB8888.
> > 
> > Since OVL_CON_CLRFMT_MAN is not supported on previous SoCs,
> > It breaks the OVL color format setting of MT8173.
> > 
> > Therefore, the fmt_convert function pointer is added to the driver
> > data
> > and mtk_ovl_fmt_convert_with_blend is implemented for MT8192 and
> > MT8195
> > that support OVL_CON_CLRFMT_MAN, and mtk_ovl_fmt_convert is
> > implemented
> > for other SoCs that do not support it to solve the degradation
> > problem.
> > 
> > Fixes: a3f7f7ef4bfe ("drm/mediatek: Support "Pre-multiplied"
> > blending in OVL")
> > Signed-off-by: Jason-JH.Lin <jason-jh.lin@mediatek.com>
> > Tested-by: Alper Nebi Yasak <alpernebiyasak@gmail.com>
> > ---
> >  drivers/gpu/drm/mediatek/mtk_disp_ovl.c | 63
> > ++++++++++++++++++++++---
> >  1 file changed, 56 insertions(+), 7 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/mediatek/mtk_disp_ovl.c
> > b/drivers/gpu/drm/mediatek/mtk_disp_ovl.c
> > index 89b439dcf3a6..4948f269fb81 100644
> > --- a/drivers/gpu/drm/mediatek/mtk_disp_ovl.c
> > +++ b/drivers/gpu/drm/mediatek/mtk_disp_ovl.c
> > @@ -143,6 +143,7 @@ struct mtk_disp_ovl_data {
> >  	unsigned int addr;
> >  	unsigned int gmc_bits;
> >  	unsigned int layer_nr;
> > +	unsigned int (*fmt_convert)(struct device *dev, struct
> > mtk_plane_state *state);
> >  	bool fmt_rgb565_is_0;
> >  	bool smi_id_en;
> >  	bool supports_afbc;
> > @@ -386,13 +387,54 @@ void mtk_ovl_layer_off(struct device *dev,
> > unsigned int idx,
> >  		      DISP_REG_OVL_RDMA_CTRL(idx));
> >  }
> >  
> > -static unsigned int ovl_fmt_convert(struct mtk_disp_ovl *ovl,
> > unsigned int fmt,
> > -				    unsigned int blend_mode)
> > +static unsigned int mtk_ovl_fmt_convert(struct device *dev, struct
> > mtk_plane_state *state)
> >  {
> > -	/* The return value in switch "MEM_MODE_INPUT_FORMAT_XXX"
> > -	 * is defined in mediatek HW data sheet.
> > -	 * The alphabet order in XXX is no relation to data
> > -	 * arrangement in memory.
> > +	struct mtk_disp_ovl *ovl = dev_get_drvdata(dev);
> > +	unsigned int fmt = state->pending.format;
> > +
> > +	switch (fmt) {
> > +	default:
> > +	case DRM_FORMAT_RGB565:
> > +		return OVL_CON_CLRFMT_RGB565(ovl);
> > +	case DRM_FORMAT_BGR565:
> > +		return OVL_CON_CLRFMT_RGB565(ovl) | OVL_CON_BYTE_SWAP;
> > +	case DRM_FORMAT_RGB888:
> > +		return OVL_CON_CLRFMT_RGB888(ovl);
> > +	case DRM_FORMAT_BGR888:
> > +		return OVL_CON_CLRFMT_RGB888(ovl) | OVL_CON_BYTE_SWAP;
> > +	case DRM_FORMAT_RGBX8888:
> > +	case DRM_FORMAT_RGBA8888:
> > +	case DRM_FORMAT_RGBX1010102:
> > +	case DRM_FORMAT_RGBA1010102:
> > +		return OVL_CON_CLRFMT_RGBA8888;
> 
> Comparing mtk_ovl_fmt_convert() with
> mtk_ovl_fmt_convert_with_blend(), some pixel formats are supported,
> such as BGRA8888.
> But in kernel v6.10, these formats are supported.
> Is the code wrong in v6.10?
> 

Yes, I lost those parts...

I'll add the cases:

case BGRX8888:
case BGRA8888:
case BGRX1010102:
case BGRA1010102:
    return OVL_CON_CLRFMT_BGRA8888;

and send the v4 soon.

Regards,
Jason-JH.Lin

> Regards,
> CK