[linux-dvb] Re: ARM crashing again

[Gregor Lawatscheck <gpel@mpex.net>]

Guido Fiala wrote:

> Am Donnerstag, 26. Februar 2004 10:30 schrieb Jarkko Santala:
>
> > > http://www.linuxtv.org/mailinglists/linux-dvb/2003/05-2003/msg00279.html
It turned out there were two seperate problems with the firmware at the 
time. The video buffer was too small to replay some streams (aka RTL bug 
and BBC mux problem), the other was and is the ARM crash we're talking 
about here. The patch bypassed the problem but became obsolete around 
driver 1.0.0 when the video buffer was increased slightly. Unfortunately 
it didn't solve the ARM crash problems and seems to be unrelated.

> > > Is this still the best known fix to this problem even after all this time
> > > or should I spend even more hours browsing the archive?
> > It did not solve the problem. Clues, anyone?
>
> I have the same problems, as many for years now...
Rev 1.3 and rev 1.5 FF cards have a problem with replaying only, when no 
signal is attached to the tuner! Rev 1.6 and later are fine. I believe 
the bug of this special case (backward compatibilty) is in the firmware 
somewhere and was introduced between Dpram / Root  1.1.2.6 and 1.1.2.7 
on 30 August 2002 and has something to do with the "internal buffer 
handling". That's only my best guess as other driver / firmware 
cominations seem to crash for different reasons, which doesn't exactly 
make it easy to pin-point.

The pre-newstruct driver 0.94 (works with vdr 1.1.13 which allows DVB-T 
settings) is stable with rev 1.3. cards for example and that is what I 
used for a long time before getting a rev 1.6 card. So there is hope 
that this can be fixed, however it requires help from the firmware 
wizards but they're busy people who could arguably say we should get 
cards that work properly...

> Would suggest to trace down the error to the component probably responsible by 
> doing some tests/logical exclusions:
>
> 1. is the firmware the same as under windows?
>
> if yes and the failure does not happen under windows, the error is definitely 
> in the linux-driver part...
Good idea for a test but is it even possible to use the same firmware in 
Windows and Linux? I thought this only applied to some of the newer 
DVB-T budget cards which use the Windows firmware / driver under Linux.

> 2. if no, does the error happen with video-data only or with audio-data-only?
>
> if with video-data-only it's probably in the video-data part (likely)
>
> 3. does the error happen with 100% perfect data? (e.g. send one and the same 
> sequence of frames in an endless loop) or only with broken data (simply send 
> random data to the driver)
>
> if it happens only with broken data, it's probably some problem with 
> resynchronisation in the ARM-code on healty data.
>
> 4.If it happens also with 100% perfect data it might be a) a race-condition or 
> similar timing dependend error or b) a hardware problem in the ARM itself :-(
>
> Has someone already made such tests?
Yes, not as thoroughly but I remember I replayed the same samples of 
video-only data in a loop for a couple of hours and made it crash with 
no signal attached to the FF-card. It were short samples of very likely 
100% perfect data but I wouldn't know how to verify that a sample is 
perfect. I used both mplayer and vdr with the same results.

The samples played back fine with 0.94 more than double the amount of 
time without an ARM crash. Likewise the samples played back fine for a 
long time with a 1.0.0+ driver and a signal attached. There's something 
cards with rev. 1.6 and later do differently when it comes to playback / 
checking the sat signal, but I have no idea what that would be.

- Gregor



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