[mythtv-users] Problem with a second DVB card, help needed

[Otto Kolsi]

>> With all the other apps (dvbscan, dvbstream, kaffeine, mplayer) I can
>> successfully tune to channels, and watch programs with both cards. With
>> mplayer I can succesfully watch two different programs from separate
>> cards at the same time.
>>
>> Problem is that when I try to use these two cards *at the same time*
>> with Myth, the second card is not working. First card records program
>> fine but although Myth starts to record a program, the recording is
>> never made with the second card (the file does not exist).
>>
>> And more strangely, after this kind of failure, if I stop the backend
>> I'm not anymore able to use any other apps with the failed card. Tuning
>> doesn't work etc. After reboot everything is okay again. How is this
>> possible? Could apps using DVB API misuse the card somehow so that it is
>> leaved to non-working state? Or does this clearly show that this is
>> kernel / driver issue?
> 
> What are your logfiles (mythbackend particularly) showing?  Is there an
> error message when the second card tries to start a recording? What size is
> your ringbuffer setting? What does /proc/meminfo show for free/used vmalloc
> space? I have just been through something similar (perhaps - I didn't notice
> that particular backend lockup) with 3 DVB cards, the 3rd one wouldn't work
> at the same time as the other two. All three are able to record
> simultaneously now though, with a larger vmalloc kernel parameter.
> http://www.gossamer-threads.com/lists/mythtv/users/223469
> 
> Henry

Now I managed to reproduce the problem outside Myth. If I run mplayer 
for both cards at the same time and then also run "updatedb", the second 
mplayer either freezes or dies. Somehow the HD activity interferes with 
the card.

So this is not a Myth issue. I've checked that cards are not sharing 
interrupts with other devices. Any other hints, what to check or change? 
Cards have PCI latency 64, I think I'll try to play with those to see if 
anything changes.
-- 
   Otto