[linux-dvb] Status of TT3200 driver development? Sharing might be the answer

Manu Abraham abraham.manu at gmail.com
Wed Sep 12 08:56:10 CEST 2007


Mario Smit wrote:

> Yes, if you go against their wishes you will certainly have problems
> later getting help from them. 
> 
> Unfortunately this creates the situation that only selected people have
> a little bit information that they are not allowed to share. So, we can
> only test the results of their works (reactive) and not work together
> (proactive) creating the best driver possible. 

More than that STM wants to see the best driver. Currently it is not yet
complete (complete in the sense, some bugs do remain in the driver.
Other than that when you say a third party hardware, it is a mix of
stuff from different vendors, so it is just not the demodulator driver
alone that always matter) as you might be aware (WIP).

The release what i had was based on a Satelco (KNC1 OEM) card using the
STB0899. There have been a couple of folks who put a lot of effort, as
well as to get additional hardware supported such as the TT S2 3200.
These are just coarse patches, not even applied to the tree. Crying out
frustration after applying these patches do not help anyone, but just a
waste of time for everyone. Such patches are meant only for the brave,
not for the people who just keep crying around.

Also i saw from one of the users I/O errors. This are due to I2C probes
which are crappy, due to a SAA7113 getting attached. (The SAA7113
doesn't have any video function at all but is used for a very different
reason) The demodulator will not respond after that utter crap.
Basically V4L drivers try to do I2C probing which is something like you
are being called suddenly by a lot of people suddenly, you get confused,
the same happens to the hardware.

Additionally STM has provided lot of support as well for getting a very
good driver. In fact it is a very complex device, because it doesn't
hide it's internals in any firmware. For the same reason due to many
patents and IP involved, they would like to keep the information far
from prying eyes.

Just with the specifications/datasheets alone (supposing you had them)
it is far from reality, that you can make anything out of it, due to the
complexity. Hardware these days are getting complex and even more
complex, not simpler.

In fact, i had to get back to STM at various stages for help, which they
really helped by putting me in touch with the relevant people involved
in the development of the various stages. Not to mention about missing
registers in the datasheets etc.

So even if you get the datasheets, it is quite pointless, unless you
understand the innards of the device.

On top of this, you will need to have a really good idea how a DVB-S2
demodulator will work to understand the same. So it is quite a lot of
effort to get a driver for such a beast going from basic bare register
specifications with around 800 registers with thousands of bitfields.

(In contrast you can take a look at drivers were specifications are
generally available and how they behave, you can see the complaints from
the users if you just see the posts on the ML.)

If you look at drivers generally, even if you have the specs, the
drivers aren't normally great unless it is well tested and certified by
the vendors. For the same reason, once it is complete STM wishes to take
the driver a do certification program for selected hardware.

This is the best that can ever happen. You get completely certified well
written OSS drivers. User tests can of course make things better.




More information about the linux-dvb mailing list