[mythtv-users] DVB-S2 PCI-express cards

[Lawrence Rust]

On Fri, 2011-09-30 at 16:18 +0100, Simon Hobson wrote:
> Emmanuel wrote:
> 
> >But I thought they would release the source code for the frontend blob,
> >but this has yet to happen.
> 
> And is very unlikely to happen.
> 
> It's a standard problem that somewhere up the line someone won't 
> release information in an open manner - typically it will be some 
> level of hardware manufacturer. The normal situation is that to get 
> the info you need to write good drivers (or even any drivers) you 
> have to agree to non-disclosure restrictions. In other words, the (I 
> believe in this case, chip) manufacturer will tell you the 
> information but you aren't allowed to tell anyone else.
> 
> Thus it's not possible for TBS to release open drivers - hence the 
> binary blobs.

I don't believe this is totally true.  I'm having tuning problems with
my tbs6981.  The LNB, DiSEqC and the tuner FE are somehow incompatible
and prevent the LNB from changing bands reliably.

I made some patches to the driver to disable DiSEqC and change the way
the 22kHz tone is switched to get around my problems.  In the process I
made a disassembly of the driver FE binary to see if I could spot a bug
and/or solution.

>From the disassembly it's clear that there is a layer that loads
encrypted firmware into the cx24117 tuner and handles the various v4l
tuning operations, writing to the tuner registers in the process.  IMHO
it would be quite feasible to open source the wrapper and provide the
encrypted firmware separately.  This would permit kernel integration and
much easier maintenance.

>  Or rather, they could - but then they'd get sued by the 
> chip manufacturer, and well screw themselves by not being given 
> access to such information in the future.
> 
> I have mixed views on this.
> 
> The idealist in me says this is a bad thing - the chip manufacturer 
> shouldn't keep such information private, and the card vendor should 
> make their drivers open.

The hardware manufacturer has a vested interest in selling as many
pieces as possible.  This is made possible by cheap software.  Problem
is the software writer doesn't want to 'give it away' and wants a per
part royalty.

> The pragmatist in me says that it's better to have such binary 
> drivers (provided it's done in a sensible way) than have no drivers 
> at all.

There's no problem with binary drivers if the binary doesn't depend on
Linux kernel APIs.  Wrapping and loading firmware is fair game.  What is
problematic is shrouding parts of the driver that operate within Linux.
But even this can be handled by using DKMS and obfuscated sources.

> I've less sympathy for the chip manufacturers since you can be fairly 
> certain their competitors will get the programming information one 
> way or another. But for TBS, there is also the issue that they don't 
> want to do all the work in making an open driver - only for another 
> manufacturer to come along, stick the same chips on a card, and 
> profit from selling cards which use the open drivers but without the 
> overheads (costs) of having written the drivers.

To be honest the amount of software that TBS provide for Linux that
isn't open wouldn't fill more than a few thousand lines.  I believe TBS
are being held to ransom by their software provider.

-- 
Lawrence