Re: [PATCH v2] media: mediatek: vcodec: Force capture queue format to MM21

[Nícolas F. R. A. Prado]

Hi,

On Tue, Feb 14, 2023 at 02:28:04AM +0000, Yunfei Dong (董云飞) wrote:
> Hi Nicolas,
> 
> Thanks for your suggestion.
> On Fri, 2023-02-10 at 10:36 -0500, Nicolas Dufresne wrote:
> > Le vendredi 10 février 2023 à 13:55 +0800, Yunfei Dong a écrit :
> > > In order to conver the format of capture queue from mediatek MM21
> > > to
> > > standard yuv420 with Libyuv, need to force capture queue format to
> > > MM21 for Libyuv can't covert mediatek MT21 format at current
> > > period.
> > 
> > Please rework this text, it is hard to understand.
> > 
> 
> Will re-write the commit message in next version as below, could you
> please help to review whether it's well to be used ?
> 
> Libyuv is one software library used to covert format. Only covert
> mediatek MM21 to standard yuv420 for MT21 is compressed mode and MM21
> is uncompressed mode at current period. Need to set capture queue
> format to MM21 in order to use Libyuv.
> 
> > > 
> > > Fixes: 7501edef6b1f ("media: mediatek: vcodec: Different codec
> > > using different capture format")
> > > Signed-off-by: Yunfei Dong <yunfei.dong@xxxxxxxxxxxx>
> > > ---
> > > changed with v1:
> > > - add Fixes tag.
> > > ---
> > >  drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec.c | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > 
> > > diff --git
> > > a/drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec.c
> > > b/drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec.c
> > > index 641f533c417f..4f5e9c20214f 100644
> > > --- a/drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec.c
> > > +++ b/drivers/media/platform/mediatek/vcodec/mtk_vcodec_dec.c
> > > @@ -41,7 +41,7 @@ static bool mtk_vdec_get_cap_fmt(struct
> > > mtk_vcodec_ctx *ctx, int format_index)
> > >  	const struct mtk_video_fmt *fmt;
> > >  	struct mtk_q_data *q_data;
> > >  	int num_frame_count = 0, i;
> > > -	bool ret = true;
> > > +	bool ret = false;

This change doesn't do anything, so I'd drop it.

> > >  
> > >  	for (i = 0; i < *dec_pdata->num_formats; i++) {
> > >  		if (dec_pdata->vdec_formats[i].type != MTK_FMT_FRAME)
> > > @@ -63,7 +63,7 @@ static bool mtk_vdec_get_cap_fmt(struct
> > > mtk_vcodec_ctx *ctx, int format_index)
> > >  	case V4L2_PIX_FMT_H264_SLICE:
> > >  	case V4L2_PIX_FMT_VP9_FRAME:
> > >  		if (fmt->fourcc == V4L2_PIX_FMT_MM21)
> > > -			ret = false;
> > > +			ret = true;
> > 
> > This makes the VP8 and the other cases identical, leaving anything
> > that touches
> > MT21 as dead code. I'm not sure, cause I cannot test it, but it
> > should in theory
> > render MT8192 unusable, unless a new firmware has been submitted to
> > linux-
> > firmware with MM21 support ?
> > 
> 
> If the firmware only support MT21 => won't exist for vp8 need to use
> MM21.

And that's the issue, the scp.img for MT8192 on linux-firmware only supports
MT21 [1]. Can you please update it to support both MM21 and MT21?

For MT8195, only MM21 is supported in scp.img [2], but since the hardware
supports both MM21 and MT21, the firmware should also support both. So please
also update it on linux-firmware.

[1] https://lore.kernel.org/all/20230112204626.ciaff4amseoidybw@notapiano/
[2] https://lore.kernel.org/all/20230112205825.wb5qcqhh5kwvyi3y@notapiano/

Thanks,
Nícolas

> 
> If the firmware only support MM21, will using MM21.
> If the firmware support MT21 & MM21, will force VP8 to use MM21,
> H264/VP9/AV1 will use MT21 before this change. 
> 
> Will force all driver to use MM21 after adding this change. 
> 
> Best Regards,
> Yunfei Dong
> > >  		break;
> > >  	default:
> > >  		ret = true;
> > 
> >