[linux-dvb] Re: Problems with WinTV NOVA-CI

[Christian Schuld <chris@sonnengesicht.de>]

Andrew,

thanks for your work on this topic. In my *very* humble opinion the bug is in 
dvb-core. I have a setup with 2 FF, 1 Nova, 1 Skystar2 and in happens quite 
often, that rmmod refuses to unload dvb-core after a problem. rmmod -f 
sometimes works, but once in a week i have to reboot.

The problems occurs more often with bad reception, e.g. if I zap on a weak 
satellite. So perhaps missing error checks in softdemux cause dvb-core to go 
wild? Just a guess.

I really appreciate your work on this topic!! 

On Wednesday 15 December 2004 03:27, Andrew de Quincey wrote:
> > > If it were in an upper layer (i.e. not tied to specific hardware),
> > > reloading the drivers would fix the saa7146 based budget cards. It does
> > > not. Therefore it is highly unlikely to be a bug in the upper driver
> > > layers. I've been through this line of thought several times already :(
> >
> > Thanks. But:
> > I'm afraid I don't get your logic: "If it were in an upper layer..."
> > And what if if the upper layer is doing wrong things?
>
> What I mean is - there aren't really many upper (driver) layers involved
> here: basically tuning,  softdemux, and en50221.
>
> Tuning we know always works OK.
>
> People see this problem without en50221, so it can be ignored.
>
> Softdemux works as follows:
> * The demodulator delivers the entire transport stream to the saa7146.
> * The saa7146 sends the entire transport stream across the PCI bus.
> * The saa7146 driver gets an interrupt when data is ready, pulls the data
> from the card, and sends it into the soft demuxer.
> * The soft demuxer extracts the PIDs the user wants and discards the rest.
>
> The only thing that could go wrong is if the soft demux failed to call the
> card specific start_feed()/stop_feed() implementations. These calls setup
> the saa7146 to start sending data across the PCI bus, and stop sending it
> respectively. I've just realised I'd assumed they worked fine as I've never
> seen them fail - totally stupid mistake now I come to think of it.
>
> However, this is a possibility - it would be good if you could check this,
> as well as someone with the VDR/FF+budget problem. Basically set the
> "budget_debug" parameter to 3 on the budget-core module. You should be able
> to check if the calls to budget_start_feed()/budget_stop_feed() are
> happening in dmesg.
>
> Another thing to check would be to add some debugging printks into
> budget-core.c/start_ts_capture() and budget-core.c/stop_ts_capture() to
> ensure it is getting all the way through those functions - also that the
> correct paths are being taken.
>
> Michael, is there any way we can check if the saa7146 is receiving data
> from the demodulator, even though it may not be sending data over the PCI
> bus for some reason?
>
> BTW: sorry if I sound annoyed, I'm just really frustrated with this
> problem! :)

-- 
Viele Grüße
Christian