Hi Tomasz,

On Fri, Nov 16, 2018 at 04:04:40PM +0900, Tomasz Figa wrote:
> > I've been using the controls currently integrated into ChromeOS that
> > have a working version of this particular setup. However, these
> > controls have a number of shortcomings and inconsistencies with other
> > decoding API. I've worked with libva so far, but I've noticed already
> > that:
> >   - The kernel UAPI expects to have the nal_ref_idc variable, while
> >     libva only exposes whether that frame is a reference frame or
> >     not. I've looked at the rockchip driver in the ChromeOS tree, and
> >     our own driver, and they both need only the information about
> >     whether the frame is a reference one or not, so maybe we should
> >     change this?
> 
> Since this is something that is actually present in the stream and the
> problem is that libva doesn't convey the information properly, I
> believe you can workaround it in the libva backend using this API by
> just setting it to 0 and some arbitrary non-zero value in a binary
> fashion.

That could work yes, thanks for the suggestion!

> >   - The H264 bitstream exposes the picture default reference list (for
> >     both list 0 and list 1), the slice reference list and an override
> >     flag. The libva will only pass the reference list to be used (so
> >     either the picture default's or the slice's) depending on the
> >     override flag. The kernel UAPI wants the picture default reference
> >     list and the slice reference list, but doesn't expose the override
> >     flag, which prevents us from configuring properly the
> >     hardware. Our video decoding engine needs the three information,
> >     but we can easily adapt to having only one. However, having two
> >     doesn't really work for us.
> >
> 
> From what I can see in the H.264 Slice header, there are 3 related data:
>  - num_ref_idx_active_override_flag - affects the number of reference
> indices for the slice,
>  - ref_list_l{0,1}_modifications - modifications for the reference lists,
>  - ref_pic_list_modification_flag_l{0,1} - selects whether the
> modifications are applied.
> 
> The reference lists inside the v4l2_ctrl_h264_slice_param are expected
> to already take all the above into account and be the final reference
> lists to be used for the slice. For reference, the H.264 specification
> refers to those final reference lists as RefPicList0 and RefPicList1
> and so the names of the fields in the struct.
> 
> There is some interesting background here, though. The Rockchip VPU
> parses the slice headers itself and handles the above data on its own.
> This means that it needs to be programmed with the unmodified
> reference lists, as in v4l2_ctrl_h264_decode_param.
> 
> Given that, it sounds like we need to have both. Your driver would
> always use the lists in v4l2_ctrl_h264_slice_param, while the Rockchip
> VPU would ignore them, use the ones in v4l2_ctrl_h264_decode_param and
> perform the per-slice modifications on its own.

I guess that would work, yep

Thanks!
Maxime

-- 
Maxime Ripard, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com