[mythtv] Live TV issue

Maaltan Naatlam maaltan at gmail.com
Fri Nov 9 04:30:09 UTC 2007


I am having issues with using the new live tv method in 20.2.  I upgraded
from 18.1 so i can use schedules direct.  Unfortunately Live TV now locks up
the entire system.  I am at a loss because it (mostly) appears to random an
never leaves any log anywhere.  The entire thing freezes up (even ping
fails).  There is no blink codes on the keyboard and unfortunately i never
got around to fixing the console display while using ivtv drivers(maybe this
is a good time, does anyone know where a howto exists?).  Watching a
recording (even one currently recording) seems to work fine.

I am using Debian Sarge (2.4.16 kernel.not sure about subrevision number) on
a celeron 700 512mzh ram with a PVR350 and PVR150.

I suspect system bus saturation causing a kernel panic.  I have had this
issue before when trying to do to much (changing channels rapidly, doing
expensive SQL queries, etc).  Another possible reason is the autoexpire job
is trimming the live tv buffer too close.  I have also seen freeze issues
when corrupt/incomplete mpeg2 frames are passed to pvr-350 decoder.   I have
seen it trim down to about 7 seconds.  if i FF or RW closer than 3 seconds
to either end of the recording, i can get a pretty predictable system
freeze.

Yet another idea ive had is that the FS cannot delete the files fast enough
(ext3) and the FS runs out of space.  i tried moving the autoexpire
threshold to 4gb (about an hour of recording at my bitrate).  I am still
seeing freezes.  The old ringbuffer partition uses a fast deleting FS (JFS i
think) which is brought me to this hypothesis

Tonights test involves using the other encoder for live tv (my pvr-150).  If
the video is still playing tomarrow morning, at least we have progress(up to
almost an hour .. one of the longest runs so far... cross your fingers).

If I cannot solve this issue, How hard would it be to backout just the new
live tv method (create a semi-persistant file in the tv directory) and
replace it with the old method (transient ringbugger on a dedicated
partition)?  I dont have a problem with hardcoding paths (since the db/UI
doesnt support the buffer anymore). This is just an idea and wanted to
bounce it off some people familiar with the code.

Barring that, how hard would it be to port the "schedules direct" code back
to 18.1?

Thanks for your help. (and sorry if this post is a bit hard to follow, its a
bit late)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mythtv.org/pipermail/mythtv-dev/attachments/20071108/fc972ed8/attachment.htm 


More information about the mythtv-dev mailing list