[linux-dvb] Re: New firmware with more video buffer

[Gregor Lawatscheck <gpel@mpex.net>]

At 02:40 04/06/2003, you wrote:
> I haven't had a single arm crash for weeks. All arm crash problems
> disappeared when the OSD was fixed in March.
Do you use transfer mode much? I believe replaying / transfermode with 
recordings / programms that require a high video memory buffer over around 
4 hours breaks things over here. It's almost as if there is some kind of 
residual effect on memory not being cleared properly or something, and 
eventually it breaks. There is no single piece of data which crashes the 
ARM, it kind of accumulates over time. If you replayed one of the old RTL 
recordings continuously over a couple of hours you'd probably get the same 
result.

> > I don't think it's
> > the size video buffer on its own. What else could have changed,
> > especially on 30th August 2002 in the firmware concerning buffer
> > handling which appears to make it lock up in continuous replay /
> > transfer mode?
>
> Around this date the time-shifting-with-one-card firmware appeared for
> the first time.
It would be great if this feature could be switched off by a seperate 
command on loading the drivers. If that's too complicated I'd really like 
to test a firmware which was stripped down of new features introduced in 
Aug 2002 to track down where the bug may be.

>  For details see
> http://linuxtv.org/mailinglists/linux-dvb/2002/09-2002/msg00011.html
"By rewriting some of the core components of the cards software we were
able to simplify and speed up the data transfer inside the AV711X.
This was combined with the improved buffer handling in the
communication between AV711X and PC, which we already introduced
some time ago, and the software filters in the Linux DVB drivers."

It seems some bug was introduced in this process which makes the ARM crash 
eventually on continuous replay. Could there even be "memory leaks" or is 
this sort of thing out of the question. If not, what else could it be...

From the posting:
"
To activate the new filtering mode you only have to get this new firmware
and change one line in the driver in the file dvb.c or av7110.c (depending
on the driver branch).
In the function StartHWFilter(), change

        buf[3] = mode;
into

        buf[3] = 0xb96a;
"

can this still be toggled in newer firmwares? Apart from that using 
hw_section = 0 - i.e. use software filters doesn't seem to make a 
difference whether or not the arm crashes.


At 02:40 04/06/2003, you wrote:
> I haven't had a single arm crash for weeks. All arm crash problems
> disappeared when the OSD was fixed in March.
Do you use transfer mode much? I believe replaying / transfermode with 
recordings / programms that require a high video memory buffer over around 
4 hours breaks things over here. It's almost as if there is some kind of 
residual effect on memory not being cleared properly or something, and 
eventually it breaks. There is no single piece of data which crashes the 
ARM, it kind of accumulates over time. If you replayed one of the old RTL 
recordings continuously over a couple of hours you'd probably get the same 
result.

> > I don't think it's
> > the size video buffer on its own. What else could have changed,
> > especially on 30th August 2002 in the firmware concerning buffer
> > handling which appears to make it lock up in continuous replay /
> > transfer mode?
>
> Around this date the time-shifting-with-one-card firmware appeared for
> the first time.
It would be great if this feature could be switched off by a seperate 
command on loading the drivers. If that's too complicated I'd really like 
to test a firmware which was stripped down of new features introduced in 
Aug 2002 to track down where the bug may be.

>  For details see
> http://linuxtv.org/mailinglists/linux-dvb/2002/09-2002/msg00011.html
"By rewriting some of the core components of the cards software we were
able to simplify and speed up the data transfer inside the AV711X.
This was combined with the improved buffer handling in the
communication between AV711X and PC, which we already introduced
some time ago, and the software filters in the Linux DVB drivers."

It seems some bug was introduced in this process which makes the ARM crash 
eventually on continuous replay. Could there even be "memory leaks" or is 
this sort of thing out of the question. If not, what else could it be...

From the posting:
"
To activate the new filtering mode you only have to get this new firmware
and change one line in the driver in the file dvb.c or av7110.c (depending
on the driver branch).
In the function StartHWFilter(), change

        buf[3] = mode;
into

        buf[3] = 0xb96a;
"

can this still be toggled in newer firmwares? Apart from that using 
hw_section = 0 - i.e. use software filters doesn't seem to make a 
difference whether or not the ARM crashes.

- Gregor 


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