[linux-dvb] [RFC] merged tree with newer patch series

Markus Rechberger mrechberger at gmail.com
Tue Oct 30 12:01:49 CET 2007


On 10/30/07, Mauro Carvalho Chehab <mchehab at infradead.org> wrote:
> > Looking at tuner-xc2028.c, there seem to be some issues:
> >
> > a) priv->count is initialized to zero by the first user. This would
> > result in priv either being freed prematurely or never freed at all.
> > b) If priv is freed, it doesn't seem to be removed from the global
> > xc2028_list list, meaning that this list is left containing dangling
> > pointers that will be followed the next time xc2028_attach is called.
> >
> > Of course, unless there's more than one user of the same tuner, these
> > would mostly cancel each other, leaving nothing but a slight memory
> > leak.
>
> Thanks for the finding, Aidan! I've wrote a patch fixing it.
>
> I've already merged all the patch series, including the fix, at v4l-dvb.
>
> A good news is that I had some time to make HVR-950 (Analog part) to
> work with tuner-xc2028.

I don't like the idea of using a reverse engineered xc2028 driver with
the em28xx driver.

* any reversed xc2028/3028 driver will not work properly for all
countries (this includes my current driver which is available on
mcentral.de)
* as for the chip reset which is slightly different between certain
firmware versions having that part reversed is not good either (eg.
power consumption might have a negative impact - as it's the case at
the moment).
* radio support is far away from the xc2028 for many devices with such
a driver (especially because of the firmware <-> driver hw/API)

Markus



More information about the linux-dvb mailing list