[Arm-netbook] building a device driver

[lkcl luke]

On Mon, Jun 18, 2012 at 11:16 PM,  <simon at koala.ie> wrote:
> On Mon, 18 Jun 2012 21:35:41 +0100, lkcl luke <luke.leighton at gmail.com>
> wrote:
>>  yeaahhh, you shouldn't have bought a device that didn't have the full
>> kernel source code, should you? slapped wrist :)
>>
>
> i was so focused on getting a tablet with an a10 in it.
> it never occurred to me that they would not provide source.

 yeah.  i learned that lesson 2 years ago when the arm-netbooks list
was first set up.  it was so embarrassing: 15 free software people
bought the CT-PC89E laptops... and the factory didn't f*****g provide
the source code.  so i was forced to go into overdrive and
reverse-engineer the device, which took about 3 weeks, with frans
pop's help.

 well, it turned out that the factory didn't *have* the source code.
as we pursued things for several months, back through the chain, it
turned out that the hardware R&D company didn't have the source code
either: it was a company in naijing that, after a lot of thought and
some discrete enquiries turned out to be PRC-govt funded.

 we believe they were trying to create a "locked-down" laptop (badly,
by relying on quotes security quotes through obscurity).  when we put
a laptop in front of the hardware engineers with a reverse-engineered
kernel and debian 5.0 on it they went veeery quiet :)   anyway, they
learned their lesson: don't do a crap job of trying to secure hardware
through software-only.

 the second lesson was from AMLogic.  we bought samples at a cost of
about $1800 from a company recommended by AMLogic.  we asked for the
source code: they refused, citing an NDA signed with AMLogic over the
GPL Linux Kernel and u-boot source code (in direct violation of the
GPL).  this cost us a client because we'd been recommending the
samples to a UK firm, who was expecting to be able to put their own
firmware onto them.

 you simply don't know what the situation and/or the chain is: they
may have paid $10k for a binary-only image and a hardware reference
design and not even *have* an agreement in place or even have thought
that they would need to have an agreement in place to receive the
source code.

 the bottom line is: don't expect them to even *have* the source code;
don't expect *their* suppliers to have the source code; don't even
expect them to be able to understand the request; don't expect them to
be able to honour a GPL kernel request, nor expect their suppliers to
be able to, nor expect the supplier's suppliers to be able to.

 it really is that ridiculous.  if this was microsoft, IBM or HP who
wrote all the GPL source code, the situation would be easy: microsoft,
IBM or HP would enforce an agreement, right down the chain, right
across the board.  but it doesn't work that way: there's no-one
*actually* putting their neck on the line to enforce the GPL - not
properly - because there's so many people involved, and it's all
"open" source, right?

 so this is why the EOMA project was started - to end this ridiculous
situation, by encouraging very large mass-volume factories to work
with software (libre) developers on mass-volume products; get enough
traction behind the standards so that other companies join in because
they can see the benefits of working with hybrid hardware-software
open standards (*1) and because they're getting their arses kicked on
sales volumes and on time-to-market (*2)

l.

(*1) that's why the patents are so important - to put real teeth
behind the project.  a small project with 50k unit sales would not
need patents.

(*2) yeah yeah all right the A10 CPU card's taking time to be
developed - we know why.  fortunately as it's the first, and now that
all the connectors have been sourced, subsequent cards will be a lot
quicker.