[vdr] Re: [linux-dvb] "Video datastream brocken" - origin (maybe) found

[Helmut Auer <vdr@helmutauer.de>]

Andreas Share wrote:

> Hi,
>
> after tons of debug-logs i have found (maybe) the reason for the "Video
> datastream broken" errors (and maybe UPT errors too).
>
> The problems, at least with my skystar2 cards, are caused by an inconsistend
> timing (different runtime) between frontend and demux ioctl´s.
>
> With an unmodified vdr and driver i have seen that opening/closing filters
> is almost done during the frontend tuning phase. This gave me an runtime
> beween 30 and 180min with my skystar2 cards. So i have tried to "sync" the
> different ioctls (inspired by the 0.9.4 driver witch use only one ioctl and
> distribute the different calls to the different driver part) a little bit.
> At the moment i have done following changes, but i think some are not
> nessasary:
>
> In dvb_frontend_ioctl (dvb_frontend.c):
>
> 1. delayed the first tuning command by dvb_delay(50) to give the demux a
> chance to close all filters.
>
> case FE_SET_TONE:
>  if (fe->status)
>   dvb_call_frontend_notifiers (fe, 0);
> +  if (fe->state & FESTATE_TUNED)
> +   dvb_delay(50);
>  fe->state = FESTATE_DISEQC;
>  dvb_frontend_internal_ioctl (&fe->frontend, cmd, parg);
>  break;
>
> 2. insert FE_READ_STATUS to override the frontend status, to make shure i
> could not get any "HAS_LOCK" during tuning.
>
> case FE_GET_EVENT:
>  err = dvb_frontend_get_event (fe, parg, file->f_flags);
>  break;
>
> +   case FE_READ_STATUS:
> +  if (fe->state != FESTATE_TUNED){
> +    fe->status = 0;
> +     }
> +  else {
> +          err = dvb_frontend_internal_ioctl (&fe->frontend, cmd, parg);
> +             }
> +  break;
>
> case FE_GET_FRONTEND:
>
> 1. In vdr 1.3.11 void cDvbTuner::Action(void) (dvbdevice.c): To loop more
> often, (inspired by Stefan Meyknecht)
>
> -       newSet.TimedWait(mutex, (ciHandler && (time(NULL) - startTime < 20))
> ? 100 : 1000);
>
> +         bool fast = (tunerStatus == tsTuned) || (ciHandler &&(time(NULL) -
> startTime < 20));
> +         newSet.TimedWait(mutex, fast ? 100 : 1000);
>        }
> }
>
> 2. bool cDvbTuner::SetFrontend(void) : Disable diseqc "caching" allways the
> same tuning sequence (and timing)
>
>                             break;
>                        }
>                      }
>                 - diseqc->Commands();
>                 +  diseqcCommands = NULL;
>                  }
>               frequency -= diseqc->Lof();
>               }
>            else {
>               esyslog("ERROR: no DiSEqC parameters found for channel %d",
> channel.Number());
>               return false;
>               }
>
> 3. using a little modified WAIT_FOR_LOCK_AFTER_TUNING loop
>
> #ifdef WAIT_FOR_LOCK_AFTER_TUNING
>  //XXX TODO preliminary fix for the "Unknown picture type" error
>  time_t t0 = time(NULL);
>  while (!dvbTuner->Locked() && time(NULL) - t0 < 5)
>        usleep(100);
> #endif
>
> +  time_t t0 = time(NULL);
> +  while (!dvbTuner->Locked() && time(NULL) - t0 < 3){
> +        usleep(500);
> + }
> +
> +  if (!dvbTuner->Locked()){
> + TurnOnLivePIDs = false;
> +  }
>
> With this modifikations my both skystar2 cards now run for 24h the epg-scan,
> without hangs. I´m shure some (or the most) of this modifications are not
> nessasary, but im not a programmer; this was only on of my tries to find out
> the reason for hanging cards. In the next day i will try out my other cards
> (ff 1.3 + 1.5 & Nova-CI), but i´m very shure all cards problems like video
> datastream errors and UPT errors (and maybe outcom-errors) are caused by the
> to early fitler setting. And, i´m think this modifications could only be a
> workaround at the moment, because this slow down tuning a little bit.
>
>
> Andreas
>  
Hello,

Are the DVB-driver developers aware of this problem ?
I am surprised to see no comments here. For some VDR-users the UPT- or 
datastream broken error is a great pain.

--
Helmut Auer, helmut@helmutauer.de