[linux-dvb] Lost lock problems over extended periods of time

[Manu Abraham]

Andrew de Quincey wrote:
> On Wednesday 08 June 2005 10:13, Manu Abraham wrote:
>>Andrew de Quincey wrote:
>>>>>That is sort of what I was thinking in desperation and my first hack :)
>>>>>The only thing it didn't do is set the tone/voltage settings again -
>>>>>which may explain why the first hack didn't retune - thats what I'm
>>>>>trying now - a combination of set tone/voltage plus reset & FE settings.
>>>>But i was thinking whether we would treating the symptom in that case ..
>>>>I was wondering why the lock goes off at all in the first case ..
>>>Oh yeah, I agree the hack is doing stuff with the symptoms - I'm just
>>>trying to identify what fixes it so we can track down the cause.
>>>
>>>Do the multiswitches where you saw the problem support anything about
>>>tone/voltage switching BTW?
>>No switches.. connected directly..
>>DVB-Card-->Splitter-->Amplifier-->LNB.
>>
>>The amplifier was there because of a long RG-11 cable.. The Amplifier
>>was near the LNB (approx 10m away, to reduce input noise)
> 
> Hi, I've managed to get it to auto-retune. However it is very weird.... All 
> I've done is the following in the else clause of "if (s & FE_HAS_LOCK)":
> 

Originally as per the DVB API the default behaviour was to auto tune if 
there was a loss of lock (?)

I think somebody (Johannes ?) a while stated that i was a bug that 
caused it not to retune..

So i believe you hit that bug instead.. Sometimes it so happens that 
searching for something else you find something else..


> if (fe->frontend->ops->set_voltage)
> 	fe->frontend->ops->set_voltage(fe->frontend, 0);
> if (fe->frontend->ops->set_tone)
> 	fe->frontend->ops->set_tone(fe->frontend, 0);
> mdelay(100);
> fe->frontend->ops->read_status(fe->frontend, &s);
> if (s & FE_HAS_LOCK) {
> 	fe->state = FESTATE_TUNED;
> 	continue;
> }
> 
> -- if it hasn't relocked I then kick off a retune - but this code wasn't run 
> (I've got printks in everywhere so I know for sure).
> 
> The channel in question uses voltage/tone SEC_VOLTAGE_13/SEC_TONE_ON (i.e. 
> 0/0) - I just hardcoded it to those above for quickness.
> 
> So I'm not resending the tone burst. I'm not even telling the frontend to 
> retune!
> 
> I'm going see if only one of those is needed to fix the problem now. 

I believe that will be a temporary fix for FTA.. But what if tuned to a 
scrambled transponder and decrypting it ? Sending CA_PMT again does not 
look clean ..

> Incidentally, we use nagios to monitor our servers, so I've checked the logs 
> going back a year - there is no pattern as to when the channels die.
> 

When i used to look at the logs a while back, could not even identify a 
weeks logs, imagine a year !

Manu