RE: [EXT] Re: [PATCH] media: amphion: only insert the first sequence startcode for vc1l format

[Ming Qian]

>>From: Ming Qian
>>Sent: 2022年7月5日 10:00
>>To: Nicolas Dufresne <nicolas@xxxxxxxxxxxx>; mchehab@xxxxxxxxxx;
>>hverkuil-cisco@xxxxxxxxx
>>Cc: shawnguo@xxxxxxxxxx; robh+dt@xxxxxxxxxx; s.hauer@xxxxxxxxxxxxxx;
>>kernel@xxxxxxxxxxxxxx; festevam@xxxxxxxxx; dl-linux-imx
>><linux-imx@xxxxxxx>; linux-media@xxxxxxxxxxxxxxx;
>>linux-kernel@xxxxxxxxxxxxxxx; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
>>Subject: RE: [EXT] Re: [PATCH] media: amphion: only insert the first sequence
>>startcode for vc1l format
>>
>>> From: Nicolas Dufresne <nicolas@xxxxxxxxxxxx>
>>> Sent: 2022年7月5日 0:06
>>> To: Ming Qian <ming.qian@xxxxxxx>; mchehab@xxxxxxxxxx;
>>> hverkuil-cisco@xxxxxxxxx
>>> Cc: shawnguo@xxxxxxxxxx; robh+dt@xxxxxxxxxx; s.hauer@xxxxxxxxxxxxxx;
>>> kernel@xxxxxxxxxxxxxx; festevam@xxxxxxxxx; dl-linux-imx
>>> <linux-imx@xxxxxxx>; linux-media@xxxxxxxxxxxxxxx;
>>> linux-kernel@xxxxxxxxxxxxxxx; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
>>> Subject: [EXT] Re: [PATCH] media: amphion: only insert the first
>>> sequence startcode for vc1l format
>>>
>>> Caution: EXT Email
>>>
>>> Hi Ming,
>>>
>>> Le mardi 28 juin 2022 à 13:20 +0800, Ming Qian a écrit :
>>> > For some formats, the amphion vpu requires startcode before sequence
>>> > and frame, such as vc1, vp8.
>>>
>>> I'm not sure why VP8 is being mentioned here. There is no specified
>>> sartcode for VP8, and no split of headers either.
>>>
>>
>>Hi Nicolas,
>>    This patch has nothing to do with vp8, and indeed there is no specified
>>startcode for VP8.
>>But amphion vpu requires driver to help insert some custom startcode for vp8
>>and vc1.
>>It's custom behavior.
>>
> Instead of exposing the custom format, you should use data_offset like VENUS driver do.  They vp8/9 codec in VENUS have the frame prefixed with an IVF header, the data_offset let the userland skip it.

Hi Nicolas,
    There is a stream buffer, and driver will copy the coded frame data to the stream buffer, driver can decide to insert custom startcode or not, userland won't know about it, so I don't think it's necessary to use data_offset to let userland skip something.
    Currently , driver will insert startcode for format vp8 and vc1. This is transparent to the user.

Ming

>>    I'm sorry that my description include some confusion
>>
>>Ming
>>
>>> >
>>> > But for V4L2_PIX_FMT_VC1_ANNEX_L, only the first sequence startcode
>>> > is needed, the extra startcode will cause decoding error.
>>> > So after seek, we don't need to insert the sequence startcode.
>>> >
>>> > In other words, for V4L2_PIX_FMT_VC1_ANNEX_L, the vpu doesn't
>>> > support dynamic resolution change.
>>> >
>>> > Fixes: 145e936380edb ("media: amphion: implement malone decoder rpc
>>> > interface")
>>> > Signed-off-by: Ming Qian <ming.qian@xxxxxxx>
>>> > ---
>>> >  drivers/media/platform/amphion/vdec.c       | 1 +
>>> >  drivers/media/platform/amphion/vpu.h        | 1 +
>>> >  drivers/media/platform/amphion/vpu_malone.c | 2 ++
>>> >  drivers/media/platform/amphion/vpu_rpc.h    | 7 ++++++-
>>> >  4 files changed, 10 insertions(+), 1 deletion(-)
>>> >
>>> > diff --git a/drivers/media/platform/amphion/vdec.c
>>> > b/drivers/media/platform/amphion/vdec.c
>>> > index 09d4f27970ec..51218a41a5ac 100644
>>> > --- a/drivers/media/platform/amphion/vdec.c
>>> > +++ b/drivers/media/platform/amphion/vdec.c
>>> > @@ -731,6 +731,7 @@ static void vdec_stop_done(struct vpu_inst *inst)
>>> >       vdec->eos_received = 0;
>>> >       vdec->is_source_changed = false;
>>> >       vdec->source_change = 0;
>>> > +     inst->total_input_count = 0;
>>> >       vpu_inst_unlock(inst);
>>> >  }
>>> >
>>> > diff --git a/drivers/media/platform/amphion/vpu.h
>>> > b/drivers/media/platform/amphion/vpu.h
>>> > index e56b96a7e5d3..f914de6ed81e 100644
>>> > --- a/drivers/media/platform/amphion/vpu.h
>>> > +++ b/drivers/media/platform/amphion/vpu.h
>>> > @@ -258,6 +258,7 @@ struct vpu_inst {
>>> >       struct vpu_format cap_format;
>>> >       u32 min_buffer_cap;
>>> >       u32 min_buffer_out;
>>> > +     u32 total_input_count;
>>> >
>>> >       struct v4l2_rect crop;
>>> >       u32 colorspace;
>>> > diff --git a/drivers/media/platform/amphion/vpu_malone.c
>>> > b/drivers/media/platform/amphion/vpu_malone.c
>>> > index c62b49e85060..f4a488bf9880 100644
>>> > --- a/drivers/media/platform/amphion/vpu_malone.c
>>> > +++ b/drivers/media/platform/amphion/vpu_malone.c
>>> > @@ -1314,6 +1314,8 @@ static int
>>> vpu_malone_insert_scode_vc1_l_seq(struct malone_scode_t *scode)
>>> >       int size = 0;
>>> >       u8 rcv_seqhdr[MALONE_VC1_RCV_SEQ_HEADER_LEN];
>>> >
>>> > +     if (scode->inst->total_input_count)
>>> > +             return 0;
>>> >       scode->need_data = 0;
>>> >
>>> >       ret = vpu_malone_insert_scode_seq(scode,
>>> > MALONE_CODEC_ID_VC1_SIMPLE, sizeof(rcv_seqhdr)); diff --git
>>> > a/drivers/media/platform/amphion/vpu_rpc.h
>>> > b/drivers/media/platform/amphion/vpu_rpc.h
>>> > index 25119e5e807e..7eb6f01e6ab5 100644
>>> > --- a/drivers/media/platform/amphion/vpu_rpc.h
>>> > +++ b/drivers/media/platform/amphion/vpu_rpc.h
>>> > @@ -312,11 +312,16 @@ static inline int vpu_iface_input_frame(struct
>>> vpu_inst *inst,
>>> >                                       struct vb2_buffer *vb)  {
>>> >       struct vpu_iface_ops *ops = vpu_core_get_iface(inst->core);
>>> > +     int ret;
>>> >
>>> >       if (!ops || !ops->input_frame)
>>> >               return -EINVAL;
>>> >
>>> > -     return ops->input_frame(inst->core->iface, inst, vb);
>>> > +     ret = ops->input_frame(inst->core->iface, inst, vb);
>>> > +     if (ret < 0)
>>> > +             return ret;
>>> > +     inst->total_input_count++;
>>> > +     return ret;
>>> >  }
>>> >
>>> >  static inline int vpu_iface_config_memory_resource(struct vpu_inst
>>> > *inst,