Kernel 2054 breaks nvidia.ko loading

Arjan van de Ven arjan at
Fri Mar 17 21:27:20 UTC 2006

(off list because I'm not interested in a flamewar either)
On Fri, 2006-03-17 at 22:07 +0200, Gilboa Davara wrote:
> On Fri, 2006-03-17 at 19:25 +0100, Arjan van de Ven wrote:
> > > Other then that, I'm being employed write a certain kernel level
> > > services that are to be deployed on RHEL.
> > > Due to specific reasons I cannot disclose (and as a matter of law) my
> > > company cannot release the code under GPL (though I might get a
> > > permission to release small unrelated parts of it). If indeed you
> > > represent the official Linux-kernel-dev-line, my employer should halt
> > > all Linux development and switch to BSD/Solaris/what-ever, right?
> > 
> > your companies lawyers should have read the GPL before, read
> > COPYING.modules from the RHEL4 kernel tree (or RHEL3) which has Linus'
> > position on this. And question themselves how on earth they can obey by
> > both clause 2 and clause 3 of the GPL... (for example: if you're going
> > to ship RHEL4 or the kernel together with your module, how on earth can
> > you reasonably argue that the compiled module is independent of the
> > kernel as clause 2 demands) And how much they like that you are
> > including kernel code into the binary you're going to distribute (via
> > the headers for example). It's your companies lawyers that need to
> > decide that, because it's their responsibility to make sure your company
> > obeys the law, and to defend it in court if they get chalanged.
> > 
> I'm not looking to start a GPL flame-war. Far from it.
> My company's layers have vent through the GPL and it's up for them to
> decide which parts of the code must be released.
> I assume that all the parts that interface directly with the kernel
> source/headers will be released as required.

I'm sure your lawyers will do what they think is the right thing, and
you should listen to them more than to me ;)

> BTW, last time I checked, nVidia is doing exactly the same. All the
> kernel dependent code is open and released to the public while all the
> kernel in-dependent is closed.

the lawyers I talked to actually said this was not ok.
Either that part needs to be open, or it doesn't. BUT here is the catch:
if it does have to be open it HAS to be GPL. (there is no scenario where
it has to be open but doesn't have to be GPL)
and once it has to be GPL, clause 2 says the entire module has to be

> > If that means that they conclude the same as the lawyers I talked to
> > ("it can't be done except in <totally unrealistic way>") and if that
> > means that the answer then is "then we don't do a Linux driver", then
> > you have the answer...
> >  
> I fear that this view will demolish the fledgling Linux ecosystem that
> is slowly being built.

"view" or "facts"... the license is what it is.

> Without high performance 3D card and drivers the chances of Linux
> gaining market share on Vista are a pipe-dream.

well from the bottom others are comming up. Intel and Via for example.
These have open drivers. Today Intel isn't there. The next generation is
apparently 3 steps further (according to the register), and you can see
the trend... nvidia and ati are scared of it. (so scared that they went
into chipsets exactly for this reason)

>  Heck, using open source
> drivers only I doubt that 10% of the users here will be able to run
> Xgl/Bling/etc.

to some degree 3D-to-the-desktop is a bit (misguided/mistimed ;(

> Even worse, with the kernel devs going into RMS mode, 

That is a mischaracterisation; just because you don't agree with someone
doesn't mean he is a fundamentalist.
Linux got where it is today BECAUSE it requires drivers to be open.
That's just about the ONLY difference that made it that linux and not
BSD because this leading open source OS.

> large Enterprise
> companies will shy away from using Linux and go back to Microsoft.
> Guess Microsoft needn't worry about Linux; we are perfectly capable of
> finding fresh new ways to shoot ourselves in the foot.

well to some degree.. if you need closed drivers you're giving away the
biggest gain you got with open source (flexibility and freedom), so why
use open source then? (you can say the same about using enterprise
distributions fwiw)

> Don't get me wrong, I do understand the "Software must be GPL'ed" view..
> In a perfect world I'd be marching right beside.
> Problem is, in a perfect world you don't have El-Quida people sifting
> through your code, trying to find ways to disable a nuclear reactor
> emergency coolers and/or missile guidance system.

yeah, in that world those people just go over the disassembly instead ;)

> Back to the original subject: I shelled out 400$ for my GF6800GT card
> and I'm willing to spend twice that amount on a fully open graphics card
> that will perform the same. Can you offer me such as option?

today? No. Why? because too many people accepted a binary driver as
normal, so much so that ati stopped supporting open 3D. 
sad but true. Note that being in the kernel is fully NVIdia's choice.
All other 3D drivers have the part that nvidia considers IP in
userspace. (well ATI is now copying nvidia on this one)
NVidia doesn't want that mostly because they do it in the kernel in
windows, and they think they will loose 0.5 to 1% performance (due to
paging overhead). But it's their CHOICE. 

More information about the fedora-test-list mailing list