linux-kernel - Re: [PATCH 0/3] Handling of reduced FPS in V4L2

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]

Date:   Tue, 28 Mar 2017 12:07:00 +0200
From:   Hans Verkuil <hverkuil@...all.nl>
To:     Jose Abreu <Jose.Abreu@...opsys.com>,
        Hans Verkuil <hansverk@...co.com>, linux-media@...r.kernel.org
Cc:     Carlos Palminha <CARLOS.PALMINHA@...opsys.com>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Hans Verkuil <hans.verkuil@...co.com>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 0/3] Handling of reduced FPS in V4L2

On 27/03/17 13:58, Jose Abreu wrote:
> Hi Hans,
> 
> 
> On 24-03-2017 12:28, Hans Verkuil wrote:
>> On 03/24/17 13:21, Jose Abreu wrote:
>>> Hi Hans,
>>>
>>>
>>> On 24-03-2017 12:12, Hans Verkuil wrote:
>>>> On 03/24/17 12:52, Jose Abreu wrote:
>>>>> Hi Hans,
>>>>>
>>>>>
>>>>>>> Can you please review this series, when possible? And if you
>>>>>>> could test it on cobalt it would be great :)
>>>>>> Hopefully next week. 
>>>>> Thanks :)
>>>>>
>>>>>> Did you have some real-world numbers w.r.t. measured
>>>>>> pixelclock frequencies and 60 vs 59.94 Hz and 24 vs 23.976 Hz?
>>>>> I did make some measurements but I'm afraid I didn't yet test
>>>>> with many sources (I mostly tested with signal generators which
>>>>> should have a higher precision clock than real sources). I have a
>>>>> bunch of players here, I will test them as soon as I can.
>>>>> Regarding precision: for our controller is theoretically and
>>>>> effectively enough: The worst case is for 640x480, and even in
>>>>> that case the difference between 60Hz and 59.94Hz is > 1 unit of
>>>>> the measuring register. This still doesn't solve the problem of
>>>>> having a bad source with a bad clock, but I don't know if we can
>>>>> do much more about that.
>>>> I would really like to see a table with different sources sending
>>>> these different framerates and the value that your HW detects.
>>>>
>>>> If there is an obvious and clear difference, then this feature makes
>>>> sense. If it is all over the place, then I need to think about this
>>>> some more.
>>>>
>>>> To be honest, I expect that you will see 'an obvious and clear'
>>>> difference, but that is no more than a gut feeling at the moment and
>>>> I would like to see some proper test results.
>>> Ok, I will make a table. The test procedure will be like this:
>>>     - Measure pixel clock value using certified HDMI analyzer
>>>     - Measure pixel clock using our controller
>>>     - Compare the values obtained from analyzer, controller and
>>> the values that the source is telling to send (the value
>>> displayed in source menu for example [though, some of them may
>>> not discriminate the exact frame rate, thats why analyzer should
>>> be used also]).
>>>
>>> Seems ok? I will need some time, something like a week because my
>>> setup was "borrowed".
>> That sounds good. Sorry for adding to your workload, but there is no
>> point to have a flag that in practice is meaningless.
>>
>> I'm actually very curious about the results!
> 
> I managed to do the tests but unfortunately I can't publish the
> full results (at least until I get approval).
> 
> I can say that the results look good. As you expected we have
> some sources with a bad clock but this is correctly detected by
> the controller (and also by the HDMI analyzer).
> 
> Using the v4l2_calc_framerate function I managed to get this:
> 
> | Source       | Resolution                  | v4l2_calc_framerate()
> ------------------------------------------------------------------------------
> | Analyzer 1 | 640x480@...94         | 59.92
> | Analyzer 1 | 640x480@60              | 60
> | Analyzer 1 | 1920x1080@60          | 60
> | Player 1     | 1920x1080@...94     | 59.94
> | Player 2     | 1920x1080@...94     | 59.93
> | Player 3     | 3840x2160@...94     | 59.94
> | Player 4     | 1920x1080@...94     | 59.94
> | Player 5     | 1920x1080@...94     | 59.93
> | Player 6     | 1280x720@50            | 50
> | Player 7     | 1920x1080@...94     | 59.93
> | Player 8     | 1920x1080@60          | 60
> | Analyzer 2 | 720x480@...94         | 59.94
> | Analyzer 2 | 720x480@60              | 60
> | Analyzer 2 | 1920x1080@...94     | 59.93
> | Analyzer 2 | 1920x180@60            | 60
> | Analyzer 2 | 3840x2160@...98     | 23.97
> | Analyzer 2 | 3840x2160@24          | 24
> | Analyzer 2 | 3840x2160@...97     | 29.96
> | Analyzer 2 | 3840x2160@30          | 30
> | Analyzer 2 | 3840x2160@...94     | 59.93
> | Analyzer 2 | 3840x2160@60          | 60

Nice!

Are the sources with a bad clock included in these results? I only see deviations
of 0.02 at most, so I don't think so.

> ------------------------------------------------------------------------------
> 
> What do you think? Shall we continue integrating this new patch
> or drop it?

Yes, we can continue. This is what I wanted to know :-)
Thank you for testing this, much appreciated.

Regards,

	Hans