Re: [PATCH v5 7/7] v4l: Document timestamp buffer flag behaviour

[Hans Verkuil <hverkuil@xs4all.nl>]

On 02/25/2014 06:08 PM, Sakari Ailus wrote:
> Hi Hans,
> 
> On Sun, Feb 23, 2014 at 12:45:28PM +0100, Hans Verkuil wrote:
>> On 02/15/2014 09:53 PM, Sakari Ailus wrote:
>>> Timestamp buffer flags are constant at the moment. Document them so that 1)
>>> they're always valid and 2) not changed by the drivers. This leaves room to
>>> extend the functionality later on if needed.
>>>
>>> Signed-off-by: Sakari Ailus <sakari.ailus@iki.fi>
>>> ---
>>>  Documentation/DocBook/media/v4l/io.xml |   10 ++++++++++
>>>  1 file changed, 10 insertions(+)
>>>
>>> diff --git a/Documentation/DocBook/media/v4l/io.xml b/Documentation/DocBook/media/v4l/io.xml
>>> index fbd0c6e..4f76565 100644
>>> --- a/Documentation/DocBook/media/v4l/io.xml
>>> +++ b/Documentation/DocBook/media/v4l/io.xml
>>> @@ -653,6 +653,16 @@ plane, are stored in struct <structname>v4l2_plane</structname> instead.
>>>  In that case, struct <structname>v4l2_buffer</structname> contains an array of
>>>  plane structures.</para>
>>>  
>>> +    <para>Dequeued video buffers come with timestamps. These
>>> +    timestamps can be taken from different clocks and at different
>>> +    part of the frame, depending on the driver. Please see flags in
>>> +    the masks <constant>V4L2_BUF_FLAG_TIMESTAMP_MASK</constant> and
>>> +    <constant>V4L2_BUF_FLAG_TSTAMP_SRC_MASK</constant> in <xref
>>> +    linkend="buffer-flags">. These flags are guaranteed to be always
>>> +    valid and will not be changed by the driver autonomously. Changes
>>> +    in these flags may take place due as a side effect of
>>> +    &VIDIOC-S-INPUT; or &VIDIOC-S-OUTPUT; however.</para>
>>
>> There is one exception to this: if the timestamps are copied from the output
>> buffer to the capture buffer (TIMESTAMP_COPY), then it can change theoretically
>> for every buffer since it entirely depends on what is being sent to it. The
>> value comes from userspace and you simply don't have any control over that.
> 
> Yes; I agree.
> 
> And a good point as well --- the timestamp source flags currently come from
> __fill_v4l2_buffer() which takes them from q->timestamp. This isn't right
> for m2m devices.
> 
> I'll fix and resend (3rd patch most likely).

You'll want to reference this patch I posted today:

[RFCv1 PATCH 16/20] vb2: fix timecode and flags handling for output buffers

Also available in this git repo:

http://git.linuxtv.org/hverkuil/media_tree.git/shortlog/refs/heads/vb2-part4

The current implementation in vb2 is actually broken (which is one of the
things fixed by this patch): if you prepare a buffer (VIDIOC_PREPARE_BUF)
and only then call VIDIOC_QBUF with a timestamp, that timestamp will be
lost since it will use the one set by PREPARE_BUF (either that or it is
zeroed, I've forgotten which of the two it was).

If you want to take that patch and add your own changes to it, then that's
fine by me. It should be pretty much standalone.

Regards,

	Hans