[vdr] Re: Buffer overflows

[Christoph Gohle <christoph@gohle.org>]

Hi I don't know all the technical details, but i can tell you my 
experience and the workaround i am using.

> At Donnerstag, 25. September 2003 22:14 Jan Ekholm wrote:
>
> > Anyone know if increasing the various buffers within VDR is enough to get
> > rid of the buffer overflows? Is it a true solution or just band aid?
> > There was a mail a while ago that mentioned values that could be used and
> > where, but I've not tried that yet. But tonight I missed a South Park
> > episode due to an overflow, so I could give the idea a shot. :)
To my experience, it happens sometimes (at least on my system) when vdr 
goes into "transfermode" this happens always, when you start a recording 
(in a single card system). This transferthread which is responsible for 
the liveview, i believe and that's what my logfiles display, is for some 
reason somtimes unable to process the data it recieves from the card 
fast enought. As soon as it's buffer overflows the liveview will have 
hickups.

On my system this sometimes happens so frequently (especially when 
recording a encrypted channel) that i have a few ten thousand log 
messages a second. In that case the system overloads for writing log 
messages and the recorder thread in turn is not able to deliver it's 
data to disk on time and this is the point where my recording get corrupted.

The workaround i am using, is to switch of the define 
DO_MULTIPLE_RECORDINGS in dvbdevice.c, after which vdr won't go into 
transfermode anymore for a single card system. The drawback of this 
workaround is that you loose two features of vdr:
1. you can't do more than one recording of different channels on the 
same transponder and card anymore (doing overlapping recordings on the 
same channel should still work).
2. timeshift and replay during a recording doesn't work when 
DO_MULTIPLE_RECORDINGS is not defined. Maybe, this is a bug (anyone know 
that for shure?) but as i wasn't able (because i don't understand the 
interior of vdr very well) to resolve this, i decided to undefine 
DO_REC_AND_PLAY_ON_PRIMARY_DEVICE in dvbdevice.c as well.

since i did the above changes no ringbuffer overflows occured anymore 
and all recordings where successful.

My system is a P133 with dvb-s rev. 1.3 running on kernel 2.4.18. There 
is at least one other post that states that maybe the kernle might be 
the problem (Klaus sais his 2.4.10 kernel works fine) 
http://www.linuxtv.org/mailinglists/vdr/2003/08-2003/msg01166.html
and there are also others with faster machines having the same problem 
(i believe) 
http://www.linuxtv.org/mailinglists/vdr/2003/08-2003/msg01166.html
so maybe this is a bug in the newer kernel (newer than 2.4.10) since 
there are some users experiencing the problem and others that are not.

At least the workaround described above, works for me, so maybe it's 
worth a try.

Cheers,
Christoph

--
*********************************************************
*Christoph Gohle                                        *
*                                                       *
*Gailkircherstrasse 7                                   *
*81247 München                                          *
*                                                       *
*Tel.:  089  32905-266  (Büro)                          *
*                 -257  (Labor)                         *
*       089  8116932    (Privat)                        *
*       0179 6089431    (Mobil)                         *
*mail:  chef@mnet-online.de		                *
*********************************************************



--
Info:
To unsubscribe send a mail to ecartis@linuxtv.org with "unsubscribe vdr" as subject.