Compatibility of Nvidia NVNET driver license with GPL

Compatibility of Nvidia NVNET driver license with GPL

[Karel Kulhavý]

Hello

I faintly remember reading some article on the Net from Linus Torvalds stating
something like if a piece of code is written specifically for Linux kernel, it
must be under GPL.

I have got an nVidia NForce2 board and downloaded their Ethernet driver (nvnet)
and they say in README: "the network driver provided by NVIDIA is subject to
the NVIDIA software license". How is with compatibility of such a behaviour
with GPL of the kernel sources?

Cl<

Re: Compatibility of Nvidia NVNET driver license with GPL

[Jeff Garzik]

Karel Kulhavý wrote:
> Hello
> 
> I faintly remember reading some article on the Net from Linus Torvalds stating
> something like if a piece of code is written specifically for Linux kernel, it
> must be under GPL.
> 
> I have got an nVidia NForce2 board and downloaded their Ethernet driver (nvnet)
> and they say in README: "the network driver provided by NVIDIA is subject to
> the NVIDIA software license". How is with compatibility of such a behaviour
> with GPL of the kernel sources?


Since I am not a lawyer, my engineering suggestion would be to sidestep 
legal issues by using "forcedeth" driver, to drive your nForce NIC. 
It's fully GPL'd, and full open source.

	Jeff

Re: Compatibility of Nvidia NVNET driver license with GPL

[Karel Kulhavý]

On Wed, Dec 31, 2003 at 01:36:58AM -0500, Jeff Garzik wrote:
> Karel Kulhavý wrote:
> > Hello
> > 
> > I faintly remember reading some article on the Net from Linus Torvalds stating
> > something like if a piece of code is written specifically for Linux kernel, it
> > must be under GPL.
> > 
> > I have got an nVidia NForce2 board and downloaded their Ethernet driver (nvnet)
> > and they say in README: "the network driver provided by NVIDIA is subject to
> > the NVIDIA software license". How is with compatibility of such a behaviour
> > with GPL of the kernel sources?
> 
> 
> Since I am not a lawyer, my engineering suggestion would be to sidestep 
> legal issues by using "forcedeth" driver, to drive your nForce NIC. 
> It's fully GPL'd, and full open source.

Suppose we would like to overcome these perpetual problems with misbehaving
manufacturers by designing a PCI network card from scratch in the soul of
free source hardware.

What are the requirements of the kernel for such a card to be cool instead of
piece of shit? How should such a card behave from PCI point of view, should it
support scatter, gather, how should interrupts be handled, what should be
programmable and what not? How should be busmastering implemented?

I am seriously thinking about designing something like that (tailored specially
for Linux) because of developping Ronja - if I included native Ronja support in
"my" network card, I could remove the superfluous circuits that implement
"bureaucracy" linke autonegotiation, TP link integrity etc. and concentrate on
raw performance. Also multiple ports could be on one card (say 4) which would
make the whole thing more nifty.

I have made to work the whole design chain from schematic design to production
of raw files for PCB manufacturers. Also seen a design of video capture
board for IDE connector so I judge implementing something on a PCI should'n
be a pain in the ass higher than moderate.

Is it possible to obtain some PCI specs without NDA's and such bullshit?
Is here anyone who has taken the PCI specs and rewritten them in their
own words?

Cl<
> 
> 	Jeff
> 
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: Compatibility of Nvidia NVNET driver license with GPL

[Jeff Garzik]

Karel Kulhavý wrote:
> On Wed, Dec 31, 2003 at 01:36:58AM -0500, Jeff Garzik wrote:
> 
>>Karel Kulhavý wrote:
>>
>>>Hello
>>>
>>>I faintly remember reading some article on the Net from Linus Torvalds stating
>>>something like if a piece of code is written specifically for Linux kernel, it
>>>must be under GPL.
>>>
>>>I have got an nVidia NForce2 board and downloaded their Ethernet driver (nvnet)
>>>and they say in README: "the network driver provided by NVIDIA is subject to
>>>the NVIDIA software license". How is with compatibility of such a behaviour
>>>with GPL of the kernel sources?
>>
>>
>>Since I am not a lawyer, my engineering suggestion would be to sidestep 
>>legal issues by using "forcedeth" driver, to drive your nForce NIC. 
>>It's fully GPL'd, and full open source.
> 
> 
> Suppose we would like to overcome these perpetual problems with misbehaving
> manufacturers by designing a PCI network card from scratch in the soul of
> free source hardware.
> 
> What are the requirements of the kernel for such a card to be cool instead of
> piece of shit? How should such a card behave from PCI point of view, should it
> support scatter, gather, how should interrupts be handled, what should be
> programmable and what not? How should be busmastering implemented?
> 
> I am seriously thinking about designing something like that (tailored specially
> for Linux) because of developping Ronja - if I included native Ronja support in
> "my" network card, I could remove the superfluous circuits that implement
> "bureaucracy" linke autonegotiation, TP link integrity etc. and concentrate on
> raw performance. Also multiple ports could be on one card (say 4) which would
> make the whole thing more nifty.
> 
> I have made to work the whole design chain from schematic design to production
> of raw files for PCB manufacturers. Also seen a design of video capture
> board for IDE connector so I judge implementing something on a PCI should'n
> be a pain in the ass higher than moderate.

If you are serious about this, we have tons of good ideas, and tons of 
suggestions on how to avoid bad ideas :)

OpenCores (http://www.opencores.org/) might be a good place to start, as 
they already have a 10/100 ethernet MAC which is working, and has been 
silicon'd:  http://www.opencores.org/projects/ethmac/  Full "source" for 
the MAC is available, in VHDL I think.  OpenCores also has PCI VHDL and 
other glue you may need.

You'll definitely want to implement autonegotiation.  It's a showstopper 
without that.  And if it's not gigabit ethernet, it's already outdated. 
  So it's a tough challenge.

Once you have basic gigabit ethernet working with 10/100/1000 
autonegotiation, let us know, and we'll pelt you with good suggestions :)


> Is it possible to obtain some PCI specs without NDA's and such bullshit?
> Is here anyone who has taken the PCI specs and rewritten them in their
> own words?

You don't need an NDA, but you do need to pay US$50 or so for the specs. 
  Or a nice person might just send them to you :)

	Jeff

Re: Compatibility of Nvidia NVNET driver license with GPL

[Paul Jackson]

nVidia's _video_ drivers are mostly proprietary code that is not
specific to Linux.  They provide a GPL wrapper or shim that, apparently
in the view of their lawyers, insulates their proprietary code from GPL
license terms.

Perhaps their network software is done the same way?

-- 
                          I won't rest till it's the best ...
                          Programmer, Linux Scalability
                          Paul Jackson <pj@sgi.com> 1.650.933.1373

Re: Compatibility of Nvidia NVNET driver license with GPL

[Jeff Garzik]

On Wed, Dec 31, 2003 at 09:11:01PM -0800, Paul Jackson wrote:
> nVidia's _video_ drivers are mostly proprietary code that is not
> specific to Linux.  They provide a GPL wrapper or shim that, apparently
> in the view of their lawyers, insulates their proprietary code from GPL
> license terms.
> 
> Perhaps their network software is done the same way?

It's a fair argument that nVidia has significant IP in their video
drivers...  I understand that's where they get a fair amount of their
speed.

For a network driver, nVidia will have a really tough time convincing me
there is useful IP in their NIC driver, or the NIC itself :)  There are
much more advanced NICs out there (with public docs, no less)...

	Jeff

Re: Compatibility of Nvidia NVNET driver license with GPL

[Chuck Campbell]

On Thu, Jan 01, 2004 at 12:48:45PM -0500, Jeff Garzik wrote:
> 
> For a network driver, nVidia will have a really tough time convincing me
> there is useful IP in their NIC driver, or the NIC itself :)  There are
> much more advanced NICs out there (with public docs, no less)...
> 
> 	Jeff

Where might one look to find a list of these, for system planning purposes?

-chuck

-- 

Re: Compatibility of Nvidia NVNET driver license with GPL

[Jan-Benedict Glaw]

[-- Attachment #1: Type: text/plain, Size: 1303 bytes --]

On Fri, 2004-01-02 11:04:02 -0600, Chuck Campbell <campbell@accelinc.com>
wrote in message <20040102170402.GD5731@helium.inexs.com>:
> On Thu, Jan 01, 2004 at 12:48:45PM -0500, Jeff Garzik wrote:
> > 
> > For a network driver, nVidia will have a really tough time convincing me
> > there is useful IP in their NIC driver, or the NIC itself :)  There are
> > much more advanced NICs out there (with public docs, no less)...
> 
> Where might one look to find a list of these, for system planning purposes?

Personally, I like to buy tulip based cards. One you might easily get is
the KTI KT-320 (or was it KF-320? Both exist, one is el-cheapo, the good
one is at about 40..50 EUR or US-$).

Some time ago, I also used eepro100-based cards. But since Intel has
started to put these into their chipset (as it seems with some
additional silicone bugs, which freezes the box with eepro100, but not
with working-around Intel's e100 driver) I don't use them any longer...

MfG, JBG

-- 
   Jan-Benedict Glaw       jbglaw@lug-owl.de    . +49-172-7608481
   "Eine Freie Meinung in  einem Freien Kopf    | Gegen Zensur | Gegen Krieg
    fuer einen Freien Staat voll Freier Bürger" | im Internet! |   im Irak!
   ret = do_actions((curr | FREE_SPEECH) & ~(NEW_COPYRIGHT_LAW | DRM | TCPA));

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: Compatibility of Nvidia NVNET driver license with GPL

[Pavel Machek]

Hi!

> If you are serious about this, we have tons of good ideas, and tons of 
> suggestions on how to avoid bad ideas :)
> 
> OpenCores (http://www.opencores.org/) might be a good place to start, as 
> they already have a 10/100 ethernet MAC which is working, and has been 
> silicon'd:  http://www.opencores.org/projects/ethmac/  Full "source" for 
> the MAC is available, in VHDL I think.  OpenCores also has PCI VHDL and 
> other glue you may need.
> 
> You'll definitely want to implement autonegotiation.  It's a showstopper 
> without that.  And if it's not gigabit ethernet, it's already outdated. 
>  So it's a tough challenge.

AFAIK, Clock is developing
ethernet-over-infrared-over-300meters-of-air. It knows what is at the
other end, and probably does not need autonegotiation. It is probably
not going to be gigabit, either. [Current version that works is 10mbit
over ~300meters].
								Pavel
-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]

Re: Compatibility of Nvidia NVNET driver license with GPL

[Jeff Garzik]

On Sat, Jan 03, 2004 at 07:44:34PM +0100, Pavel Machek wrote:
> Hi!
> 
> > If you are serious about this, we have tons of good ideas, and tons of 
> > suggestions on how to avoid bad ideas :)
> > 
> > OpenCores (http://www.opencores.org/) might be a good place to start, as 
> > they already have a 10/100 ethernet MAC which is working, and has been 
> > silicon'd:  http://www.opencores.org/projects/ethmac/  Full "source" for 
> > the MAC is available, in VHDL I think.  OpenCores also has PCI VHDL and 
> > other glue you may need.
> > 
> > You'll definitely want to implement autonegotiation.  It's a showstopper 
> > without that.  And if it's not gigabit ethernet, it's already outdated. 
> >  So it's a tough challenge.
> 
> AFAIK, Clock is developing
> ethernet-over-infrared-over-300meters-of-air. It knows what is at the
> other end, and probably does not need autonegotiation. It is probably
> not going to be gigabit, either. [Current version that works is 10mbit
> over ~300meters].

FWIW autonegotiation is strictly related to "the wire", so wireless
would be a totally different space.

	Jeff