Re: [PATCH v3 5/9] vcodec: mediatek: Add Mediatek V4L2 Video Decoder Driver

[Tiffany Lin <tiffany.lin@mediatek.com>]

Hi Hans,

On Mon, 2016-08-01 at 12:39 +0200, Hans Verkuil wrote:
> Hi Tiffany,
> 
> On 08/01/2016 11:38 AM, Tiffany Lin wrote:
> > Hi Hans,
> 
> <snip>
> 
> >>> +static int vidioc_vdec_g_selection(struct file *file, void *priv,
> >>> +			struct v4l2_selection *s)
> >>> +{
> >>> +	struct mtk_vcodec_ctx *ctx = fh_to_ctx(priv);
> >>> +
> >>> +	if (V4L2_TYPE_IS_OUTPUT(s->type))
> >>> +		return -EINVAL;
> >>> +
> >>> +	if (s->target != V4L2_SEL_TGT_CROP)
> >>> +		return -EINVAL;
> >>
> >> How does the cropping rectangle relate to the format size? There is no s_selection,
> >> so this doesn't make sense.
> >>
> > I want to return encoded crop information or real display region that
> > display driver could know where is padding region.
> 
> Sorry, I don't understand this.
> 
> This is a decoder, so based on the bitstream it decodes to a certain width
> and height. I assume that is what you refer to as coded_width and coded_height?
> 
> If so, then what is the 'real display region' and how does it relate to the
> coded width/height?
> 
> This is probably a terminology issue but I need to understand this before I
> can decide what should be done here.
> 
We define coded_width and coded_height as frame buffer width and height.
Coded resolution is stream resolution in pixels aligned to codec format
and hardware requirements.
And we assign bytesperline as coded_width.

g_selection is for getting visible resolution for user space.
visible resolution - stream resolution of the visible picture, in
pixels, to be used for display purposes; must be smaller or equal to
coded resolution;
visible height - height for given visible resolution
visible width - width for given visible resolution

For my understand, g_fmt should return coded width/height in
pix_mp->width, pix_mp->height.
And bytesperline is coded with.
g_selection return visible width/height.
Is this understanding correct?


best regards,
Tiffany
> Regards,
> 
> 	Hans
> 
> > User space use s_fmt/g_fmt  to set/get coded_width and coded_height, and
> > use g_crop to get real display region.
> > That's why I do not add s_selection.
> > 
> >> Alternatively, it could be that you are really returning V4L2_SEL_TGT_COMPOSE_PADDED.
> >>
> > 
> > V4L2_SEL_TGT_COMPOSE_PADDED means
> > "The active area and all padding pixels that are inserted or modified by
> > hardware."
> > But I just want to return active area to user space.
> > 
> >>> +
> >>> +	if (ctx->state < MTK_STATE_HEADER)
> >>> +		return -EINVAL;
> >>> +
> >>> +	if ((ctx->q_data[MTK_Q_DATA_SRC].fmt->fourcc == V4L2_PIX_FMT_H264) ||
> >>> +	    (ctx->q_data[MTK_Q_DATA_SRC].fmt->fourcc == V4L2_PIX_FMT_VP8) ||
> >>> +	    (ctx->q_data[MTK_Q_DATA_SRC].fmt->fourcc == V4L2_PIX_FMT_VP9)) {
> >>> +
> >>> +		if (vdec_if_get_param(ctx, GET_PARAM_CROP_INFO, &(s->r))) {
> >>> +			mtk_v4l2_debug(2,
> >>> +					"[%d]Error!! Cannot get param : GET_PARAM_CROP_INFO ERR",
> >>> +					ctx->id);
> >>> +			s->r.left = 0;
> >>> +			s->r.top = 0;
> >>> +			s->r.width = ctx->picinfo.buf_w;
> >>> +			s->r.height = ctx->picinfo.buf_h;
> >>> +		}
> >>> +		mtk_v4l2_debug(2, "Cropping info: l=%d t=%d w=%d h=%d",
> >>> +				s->r.left, s->r.top, s->r.width,
> >>> +				s->r.height);
> >>> +	} else {
> >>> +		s->r.left = 0;
> >>> +		s->r.top = 0;
> >>> +		s->r.width = ctx->picinfo.pic_w;
> >>> +		s->r.height = ctx->picinfo.pic_h;
> >>> +		mtk_v4l2_debug(2, "Cropping info: w=%d h=%d fw=%d fh=%d",
> >>> +				s->r.width, s->r.height, ctx->picinfo.buf_w,
> >>> +				ctx->picinfo.buf_h);
> >>> +	}
> >>> +	return 0;
> >>> +}