[vdr] Re: final solution to the signalless operation problemfound!! -> testers+RFC required

[Guido Fiala <gfiala@s.netic.de>]

Am Wednesday 24 March 2004 21:22 schrieb Klaus Schmidinger:
> I don't see why VDR should be the culprit here.
> What's wrong with opening the frontend and not actually using it?
> Shouldn't the driver (or whoever is responsible) be stable enough
> to handle this?
>
> Of course we can add an option that makes VDR not open the frontend
> in case the user knows that there is no antenna cable connected - but
> that still leaves the case where the signal gets lost for other reasons.
> And then what happens if the user forgets to set that option?

Just to state it correctly here - my current analysis shows vdr _needs_ to 
open the frontend and call  -if proposed option is set - FE_SLEEP instantly.

Not opening the device at all (including Andreas' patch) did not have the 
stabilizing effect.

> After all, your computer doesn't crash just because a program opens
> a file and then doesn't read from it... ;-)

Exactly - i suppose, that in case the frontend runs without or with bad signal 
one of the following cases produces the problem:
1.the chip outputs blocks with IEI-Bit not set but erronous data
2.the frontend generates a lot of (unwanted) interrupts if no/bad signal, 
confusing the ARM
3.the ARM get's correctly informed by IEI-Bit but fails in correct handling of 
uncorrect blocks
4.by chance the frontend "correctly" generates correct blocks out of the noise 
which are actually just noise no data -> confusing the ARM

cases 1,2 assume a hardware implementation error, cases 3 or 4 a potential 
weakness in the firmware. Unfortunately one can only decide what is happening 
with a logic analyser connected to the data-path between the 2 chips.

Guido


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