Re: [RFC PATCH] media/doc: Allow sizeimage to be set by v4l clients

[Stanimir Varbanov <stanimir.varbanov@linaro.org>]

On 4/10/19 5:39 PM, Hans Verkuil wrote:
> On 4/10/19 4:21 PM, Stanimir Varbanov wrote:
>> Hi Hans,
>>
>> On 3/14/19 3:11 PM, Hans Verkuil wrote:
>>> On 1/21/19 11:48 AM, Stanimir Varbanov wrote:
>>>> Hi Hans,
>>>>
>>>> On 1/18/19 11:13 AM, Hans Verkuil wrote:
>>>>> On 1/16/19 1:37 PM, Stanimir Varbanov wrote:
>>>>>> This changes v4l2_pix_format and v4l2_plane_pix_format sizeimage
>>>>>> field description to allow v4l clients to set bigger image size
>>>>>> in case of variable length compressed data.
>>>>>>
>>>>>> Signed-off-by: Stanimir Varbanov <stanimir.varbanov@linaro.org>
>>>>>> ---
>>>>>>  Documentation/media/uapi/v4l/pixfmt-v4l2-mplane.rst | 5 ++++-
>>>>>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst        | 3 ++-
>>>>>>  2 files changed, 6 insertions(+), 2 deletions(-)
>>>>>>
>>>>>> diff --git a/Documentation/media/uapi/v4l/pixfmt-v4l2-mplane.rst b/Documentation/media/uapi/v4l/pixfmt-v4l2-mplane.rst
>>>>>> index 7f82dad9013a..dbe0b74e9ba4 100644
>>>>>> --- a/Documentation/media/uapi/v4l/pixfmt-v4l2-mplane.rst
>>>>>> +++ b/Documentation/media/uapi/v4l/pixfmt-v4l2-mplane.rst
>>>>>> @@ -30,7 +30,10 @@ describing all planes of that format.
>>>>>>  
>>>>>>      * - __u32
>>>>>>        - ``sizeimage``
>>>>>> -      - Maximum size in bytes required for image data in this plane.
>>>>>> +      - Maximum size in bytes required for image data in this plane,
>>>>>> +        set by the driver. When the image consists of variable length
>>>>>> +        compressed data this is the maximum number of bytes required
>>>>>> +        to hold an image, and it is allowed to be set by the client.
>>>>>>      * - __u32
>>>>>>        - ``bytesperline``
>>>>>>        - Distance in bytes between the leftmost pixels in two adjacent
>>>>>> diff --git a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
>>>>>> index 71eebfc6d853..54b6d2b67bd7 100644
>>>>>> --- a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
>>>>>> +++ b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
>>>>>> @@ -89,7 +89,8 @@ Single-planar format structure
>>>>>>        - Size in bytes of the buffer to hold a complete image, set by the
>>>>>>  	driver. Usually this is ``bytesperline`` times ``height``. When
>>>>>>  	the image consists of variable length compressed data this is the
>>>>>> -	maximum number of bytes required to hold an image.
>>>>>> +	maximum number of bytes required to hold an image, and it is
>>>>>> +	allowed to be set by the client.
>>>>>>      * - __u32
>>>>>>        - ``colorspace``
>>>>>>        - Image colorspace, from enum :c:type:`v4l2_colorspace`.
>>>>>>
>>>>>
>>>>> Hmm. "maximum number of bytes required to hold an image": that's not actually true
>>>>> for bitstream formats like MPEG. It's just the size of the buffer used to store the
>>>>> bitstream, i.e. one buffer may actually contain multiple compressed images, or a
>>>>> compressed image is split over multiple buffers.
>>>>>
>>>>
>>>> Do you want me to change something in the current documentation, i.e.
>>>> the quoted above?
>>>
>>> Hmm, it looks like this discussion stalled (i.e. I forgot to reply).
>>>
>>> How about this:
>>>
>>> "When the image consists of variable length compressed data this is the
>>> number of bytes required by the encoder to support the worst-case
>>
>> I don't think 'encoder' is the right word here:
>>
>> s/encoder/encoder or decoder
> 
> or perhaps just 'codec'? Either is fine by me.

OK, then just 'codec'.

> 
>>
>>> compression scenario. Clients are allowed to set this field. However,
>>> drivers may ignore the value or modify it."
>>
>> Can we rephrase to:
>>
>> "Clients are allowed to set sizeimage field, but however drivers my
>> ignore the value or modify it."
>>
> 
> How about this:
> 
> "Clients are allowed to set the sizeimage field, but drivers may
>  modify it."

I'm fine with that.

I'll send with those changes and as regular patch. Thanks!

> 
> 'ignore the value' isn't right, since that suggests that the sizeimage
> value that is returned may not be in actual use. Since sizeimage can now
> be set by userspace, this means that a sanity check is needed in drivers
> as well (i.e. what if userspace sets sizeimage to ~0?)
> 
> This should be tested in v4l2-compliance as well.
> 
> Regards,
> 
> 	Hans
> 

-- 
regards,
Stan