Getting nvnet specs

[Bodhi]

Does anyone have any suggestions on how to get nvidia to release the specs for nvnet?
What is nvidia's stance on it in the first place?
I mean I don't see them gaining anything from keeping the specs from us(unlike for their GPU), except complaints :P

[connyosis]

Yeah, this has been discussed before and as far as I know we never got a statement from any nVidia employee about it. It's true that everyone would benefit from open specs, so I dont see much point in not opening them.

[whit]

With the nForce chips getting great reviews and getting included on high-end motherboards, which Linux users tend to acquire, it would make sense for nVidia to open as much as it can to community support - the stuff will become more dependable and easier to install, and nVidia will save staff time in the long run as the community takes over. Especially for a NIC, what can they be protecting? I saw a 10/100 NIC for $5 in a store yesterday. It's just not a big part of the value of what nVidia's chips provide, especially if it takes more than $5 of extra puzzlement for the user to get the drivers installed, as compared to open drivers for just about any other NIC, which are already included with the kernel.

[blueworm]

I kind of wish that board manufacturers would make the boards without lan and sound. that way us linux users can use what has best support.

[connyosis]

Well we still can. Just because there is onboard audio/video/lan doesnt mean we have to use them. We're not forced to use the integrated features, its just a possibility.

[whit]

We're not forced to use the onboard features, unless we've built an X PC that only has one PCI and one AGP expansion slot. It's really the onboard features that make those systems possible and popular.

[erwos]

IIRC, the LAN chip was developed using other companies' IP. It's not so much an issue of "want" so much as "can".

I would also add that USB can take care of your networking needs for the most part. I've been using a wireless USB adapter on my RH9 system, and it works very well. I'd imagine that it would not be much more difficult to find a wired adapter, or even a modem. A bit more expensive, of course, but you aren't cluttering up PCI slots.

-Erwos

[whit]

The advantage to the other company of making public the specs to the NIC would be the same as the advantage to nVidia - happier end users because of better OS support of the hardware.

As for hanging stuff off USB - the point of having a small case (like the Shuttle that's currently available with nForce chips) is in having the system be compact as possible. An extraneous external device defeats that goal. There is such an animal though (see http://www.tomshardware.com/network/20021122/) but it only works in Windows (not DOS or Linux).

If you look at the NIC drivers included in the kernel code, it's pretty much everyone in the industry aside from nVidia (and these USB NICs). This doesn't seem to have hurt 3com's business, for instance, has it?

[geowiz]

First, I'd like to say that it's good that Nvidia has provided network drivers for the Linux community. It works well in the 2.4.X series, even if it causes install problems for some users.
I also understand Nvidia's position on not GLPing the current drivers since they are apparently based on 3rd party non-GLP code.

But the current nvnet drivers cause problems for the community. Problems include:
They are difficult to install for less experienced users
They taint the kernel making it difficult to debug kernel issues
Don't keep pace with the kernel (e.g. the current ones do not work with the 2.6.0 series)
Make it impossible for distributions to deliver configured solutions out of the box due to the license
Reduces hardware use by other Open Source projects - FreeBSD, Hurd etc.

What could Nvidia do to help resolve the issue?
Communicate to the Open Source community that Nvidia is/is not actively working with the 3rd party to release the drivers under the GPL. If negotiation is ongoing an indication of timing and a communication of that information.
Inform the community who the 3rd party is and the chip base used so the community could reverse engineer a GPL driver. If Nvidia is restricted from releasing the vendors name, to work with the 3rd party to resolve that issue. If not restricted from releasing the 3rd party's name, then the community could help influence the 3rd party to release the required information.
Release the api to the chip without divulging the 3rd party IP. A couple of current kernel netorking modules load ok but they don't work due to invalid data returns (byte offsets?)
Provide a patch to an existing network driver that would allow that kernel driver to work with the current chip.

Benefits to Nvidia
Shows further support to the Open Source community
Reduces Nvidia's costs by having the community provide support
Makes it easier for users to utilize the on-board networking
Makes kernel debugging easier (no tainted modules)
Allows the driver to keep pace with kernel development amd vendor releases
Provides a solution to distribution vendors so that the network card works out of the box.
Reduces users problems in getting the network working thus improving their experience and decreasing their perception that the boards are "broken" and the Linux is a pain to use
Opens up the hardware to other Open Source projects to use the Nvidia based boards (FreeBSD, Hurd, OpenBSD, etc) (Increased market )

So, can we expect an official or semi-official response to the issues of the non-GPL network drivers?
What can we do to help you to help us?

Thanks - Geowiz

[whit]

Originally posted by geowiz
First, I'd like to say that it's good that Nvidia has provided network drivers for the Linux community.

Very well argued geowiz. Thanks.

To improve some other functions, I've switched to a 2.6 kernel, and that means I have to go without the NIC and soundsystem for now. This is absurd - high end hardware that's rendered incompatible with high end software because it has the only NIC or sound board in current production whose interface isn't openly specified.

I'm waiting to see what nVidia does to improve this. If they fix the problem soon I'll become their biggest fan. If not I'll be one of those people passing the word to avoid nforce products. I'd rather be the first guy - being positive is almost always more satisfying.

At the least, with the 2.6 kernel almost at official release, I hope they'll be quick in supporting it. If the interfaces were already publicly specified there's no doubt the 2.6 kernel would already support this hardware. Turning down free outside labor hardly seems feduciary. It's their free decision, but how do they explain to investors in the company that they're taking the course that requires them to maintain extra staff to do the work that otherwise would be done for free? This isn't just public relations; it's investor relations.

[Bodhi]

This patch applies to the nforce-1.0-0261 package(Not my creation), it will let you compile nvnet under 2.6, as for the sound chip you can use alsa(snd-intel8x0)

[whit]

Originally posted by Bodhi
This patch applies to the nforce-1.0-0261 package(Not my creation), it will let you compile nvnet under 2.6, as for the sound chip you can use alsa(snd-intel8x0)

That patch works perfectly. Patch, compile, install, insmod and I can bring the NIC right up.

Thanks!

[micha_b]

... a patch is fine, but it doesn't change anything like geowiz wrote.
Is there a petition anywhere?
Why do we have to have petitions all the time?
Why does no nVidia official respond?

Micha

[Bodhi]

http://www.petitiononline.com/nforce2/petition.html

1916 Signatures, hmm =D
Although i'm afraid i've signed it twice now, once a long time ago I think :P

[geowiz]

I'd have to say that the simple fact that the community (well, some of the talented ones) have "hacked" the nvnet and the NVIDIA_kernel module code to work with the 2.6.0 release and there is no apparent beta or official release from Nvidia that works with 2.6.0 supports some of our arguments.