[mythtv-users] VIDIOCMCAPTURE0: Device or resource busy

[Andrew M. Bishop]

"David Madsen" <david.madsen at gmail.com> writes:

>> File to patch is libs/libmythtv/NuppelVideoRecorder.cpp
>>>
>>> Final follow-up
>>>
>>> Problem solved.
>>>
>>> First, this does appear to be a kernel issue with bttv and v4l1:
>>> http://kerneltrap.org/mailarchive/linux-kernel/2008/8/4/2806884
>>>
>>> Using kernel version 2.6.24 works fine.
>>>
>>> In addition, patching mythtv to use v4l2 for bttv cards allows it to
>>> work with kernels >= 2.6.25.  This also reduced my CPU time while
>>> recording from 40% to 25% (Sempron 2400+)
>>>
>>> 932c932
>>> <         if (driver == "cx8800" || driver == "go7007" || driver == "em28xx")
>>> ---
>>> >         if (driver == "cx8800" || driver == "go7007" || driver == "em28xx" || driver == "bttv")
>>>
>>> I have not yet experienced any channel changing errors as documented
>>> here:
>>> http://www.mythtv.org/pipermail/mythtv-dev/2005-February/031229.html
>
> I filed a bug for this issue a few weeks ago, but the changes haven't
> been made in SVN yet.  Feel free to add any relevant info to the bug
> report.
>
> http://svn.mythtv.org/trac/ticket/5608

My experiences with this patch are not good, but they are only one
data point.  Perhaps it works perfectly well for other people.

I have a bttv card and experienced the problem with kernel 2.6.25 and
2.6.26 which stopped me using them with MythTV.  Because of this I
switched back to 2.6.24 which was fine.  When this patch came out I
tried it and at first it seemed to work; I could upgrade to kernel
2.6.26 and record with MythTV.

After a while though I realised that the recordings were poor quality.
During playback there were frames displayed at the wrong time, these
are most easily seen when editing a recording.  For example I saw that
one interlaced field in a frame should actually have been about 7
frames earlier (estimating from a fast moving picture).  This occurs
every second or so (it isn't regular) and only causes a visual
disturbance with fast moving scenes.  The problem was definitely in
the recorded data, not playback.

Since switching back to kernel 2.6.24 and backing out the patch
everything seems to be back to normal.

-- 
Andrew.
----------------------------------------------------------------------
Andrew M. Bishop                             amb at gedanken.demon.co.uk
                                      http://www.gedanken.demon.co.uk/