linux-kernel - Re: [PATCH v10 6/9] media: uapi: Add a control for HANTRO driver

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <815a4bd6-599b-cfb8-9ddc-efa4b7092c23@xs4all.nl>
Date:   Thu, 6 May 2021 14:50:18 +0200
From:   Hans Verkuil <hverkuil-cisco@...all.nl>
To:     Benjamin Gaignard <benjamin.gaignard@...labora.com>,
        ezequiel@...labora.com, p.zabel@...gutronix.de, mchehab@...nel.org,
        robh+dt@...nel.org, shawnguo@...nel.org, s.hauer@...gutronix.de,
        festevam@...il.com, lee.jones@...aro.org,
        gregkh@...uxfoundation.org, mripard@...nel.org,
        paul.kocialkowski@...tlin.com, wens@...e.org,
        jernej.skrabec@...l.net, emil.l.velikov@...il.com
Cc:     kernel@...gutronix.de, linux-imx@....com,
        linux-media@...r.kernel.org, linux-rockchip@...ts.infradead.org,
        devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        linux-kernel@...r.kernel.org, devel@...verdev.osuosl.org,
        kernel@...labora.com, cphealy@...il.com
Subject: Re: [PATCH v10 6/9] media: uapi: Add a control for HANTRO driver

On 05/05/2021 17:20, Benjamin Gaignard wrote:
> 
> Le 05/05/2021 à 16:55, Hans Verkuil a écrit :
>> On 20/04/2021 14:10, Benjamin Gaignard wrote:
>>> The HEVC HANTRO driver needs to know the number of bits to skip at
>>> the beginning of the slice header.
>>> That is a hardware specific requirement so create a dedicated control
>>> for this purpose.
>>>
>>> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@...labora.com>
>>> ---
>>>   .../userspace-api/media/drivers/hantro.rst    | 19 +++++++++++++++++++
>>>   .../userspace-api/media/drivers/index.rst     |  1 +
>>>   include/media/hevc-ctrls.h                    | 13 +++++++++++++
>>>   3 files changed, 33 insertions(+)
>>>   create mode 100644 Documentation/userspace-api/media/drivers/hantro.rst
>>>
>>> diff --git a/Documentation/userspace-api/media/drivers/hantro.rst b/Documentation/userspace-api/media/drivers/hantro.rst
>>> new file mode 100644
>>> index 000000000000..cd9754b4e005
>>> --- /dev/null
>>> +++ b/Documentation/userspace-api/media/drivers/hantro.rst
>>> @@ -0,0 +1,19 @@
>>> +.. SPDX-License-Identifier: GPL-2.0
>>> +
>>> +Hantro video decoder driver
>>> +===========================
>>> +
>>> +The Hantro video decoder driver implements the following driver-specific controls:
>>> +
>>> +``V4L2_CID_HANTRO_HEVC_SLICE_HEADER_SKIP (integer)``
>>> +    Specifies to Hantro HEVC video decoder driver the number of data (in bits) to
>>> +    skip in the slice segment header.
>>> +    If non-IDR, the bits to be skipped go from syntax element "pic_output_flag"
>>> +    to before syntax element "slice_temporal_mvp_enabled_flag".
>>> +    If IDR, the skipped bits are just "pic_output_flag"
>>> +    (separate_colour_plane_flag is not supported).
>> I'm not very keen on this. Without this information the video data cannot be
>> decoded, or will it just be suboptimal?
> 
> Without that information the video can't be decoded.
> 
>>
>> The problem is that a generic decoder would have to know that the HW is a hantro,
>> and then call this control. If they don't (and are testing on non-hantro HW), then
>> it won't work, thus defeating the purpose of the HW independent decoder API.
>>
>> Since hantro is widely used, and if there is no other way to do this beside explitely
>> setting this control, then perhaps this should be part of the standard HEVC API.
>> Non-hantro drivers that do not need this can just skip it.
> 
> Even if I put this parameter in decode_params structure that would means that a generic
> userland decoder will have to know how the compute this value for hantro HW since it
> isn't something that could be done on kernel side.

But since hantro is very common, any userland decoder will need to calculate this anyway.
So perhaps it is better to have this as part of the decode_params?

I'd like to know what others think about this.

Regards,

	Hans

> 
> 
> Regards,
> Benjamin
> 
>>
>> Regards,
>>
>> 	Hans
>>
>>> +
>>> +.. note::
>>> +
>>> +        This control is not yet part of the public kernel API and
>>> +        it is expected to change.
>>> diff --git a/Documentation/userspace-api/media/drivers/index.rst b/Documentation/userspace-api/media/drivers/index.rst
>>> index 1a9038f5f9fa..12e3c512d718 100644
>>> --- a/Documentation/userspace-api/media/drivers/index.rst
>>> +++ b/Documentation/userspace-api/media/drivers/index.rst
>>> @@ -33,6 +33,7 @@ For more details see the file COPYING in the source distribution of Linux.
>>>   
>>>   	ccs
>>>   	cx2341x-uapi
>>> +        hantro
>>>   	imx-uapi
>>>   	max2175
>>>   	meye-uapi
>>> diff --git a/include/media/hevc-ctrls.h b/include/media/hevc-ctrls.h
>>> index 8e0109eea454..b713eeed1915 100644
>>> --- a/include/media/hevc-ctrls.h
>>> +++ b/include/media/hevc-ctrls.h
>>> @@ -224,4 +224,17 @@ struct v4l2_ctrl_hevc_decode_params {
>>>   	__u64	flags;
>>>   };
>>>   
>>> +/*  MPEG-class control IDs specific to the Hantro driver as defined by V4L2 */
>>> +#define V4L2_CID_CODEC_HANTRO_BASE				(V4L2_CTRL_CLASS_CODEC | 0x1200)
>>> +/*
>>> + * V4L2_CID_HANTRO_HEVC_SLICE_HEADER_SKIP -
>>> + * the number of data (in bits) to skip in the
>>> + * slice segment header.
>>> + * If non-IDR, the bits to be skipped go from syntax element "pic_output_flag"
>>> + * to before syntax element "slice_temporal_mvp_enabled_flag".
>>> + * If IDR, the skipped bits are just "pic_output_flag"
>>> + * (separate_colour_plane_flag is not supported).
>>> + */
>>> +#define V4L2_CID_HANTRO_HEVC_SLICE_HEADER_SKIP	(V4L2_CID_CODEC_HANTRO_BASE + 0)
>>> +
>>>   #endif
>>>
>>