[mythtv] Ringbuffer: fsync thread may equal corruption

John Patrick Poet john at BlueSkyTours.com
Mon Dec 20 07:01:33 UTC 2004


John Patrick Poet wrote:

>
> Since I started using a version of ThreadedFileWriter with a separate 
> thread for doing fsyncs, I have had some of my shows end up with 
> corruption.  I cannot *make* it happen, and cannot prove that the 
> separate fsync thread is the cause, but it is the only cause I can 
> think of.
>
> John


Scratch that.  I went back to using a version of ThreadedFileWriter 
which does not use a separate thread for fsync.  I still had a file get 
corrupt about 45 minutes in.

That time frame is about right for when this problem first started to 
appear.

About that time I also switched the hdtvrecorder ringbuffer thread to 
read directly from the ringbuffer, instead of using an intermediary 
buffer -- but switching that back also did not solve the problem.

That was also about the time that the new HD3000-HD2000-CD-rev4 driver 
came out, but switching back to rev3 also does not solve the problem.

I have not tried going back to v37 of Daniel's hdtvrecorder patch, but I 
find it very hard to believe that v38 could be the cause.

It is very weird.  A show can become corrupt within minutes of it 
starting, or not until it is over 2 hours into it, or not at all.  Most 
of the time, once a show is corrupt, it continues to be corrupt until 
the recording is over.  Sometimes, however, it will recover after 
several minutes and be okay until the end.

I cannot figure out the cause.

John



More information about the mythtv-dev mailing list