LKM NVidia Drivers Now Available For NetBSD

Dan writes "Quentin Garnier has made a loadable kernel module (LKM) version of the NVidia drivers on NetBSD. This release is very preliminary, rough and mostly meant to test the installation procedure. You will need a NetBSD-current system but the downloadable drivers code itself should be quite backward compatible with some caveats. For example, you need 'options KVM86' in your kernel config. His NVidia drivers on NetBSD page indicates that known working hardware includes RIVA TNT2 Model 64 (PCI), GeForce2 MX/MX 400, Vanta(AGP) and more!"

The answer is DRI!

[motown]

As I already pointed out in this post [slashdot.org], it would be a lot easier for both NVIDIA and others if NVIDIA just based its driver model on DRI.

Re:The answer is DRI!

[ctr2sprt]

Probably so, but back when NVIDIA first released their drivers it really was reasonable to stay away from DRI, do their own implementation of AGP, etc. Now I suspect a lot of it is inertia: if they wanted to rewrite their drivers to use DRI, they'd probably have to start nearly from scratch. This would take a long time, and the end result would likely be worse than what we have now (at first, anyway), so nobody would use it.

I'm not saying you're wrong. I think this is the direction NVIDIA needs to take in the future, and I expect they will take it (eventually). But it's not something that can be done in even a couple weeks.

(Actually, I am kind of hoping that this is the reason behind the delay in FreeBSD drivers: the *BSD and Linux teams are trying to work on a much more portable driver, which might well mean DRI. Of course, there are lots of other things to worry about in the FreeBSD drivers since nearly all OpenGL games are going to be run in the Linux emulator, so maybe it's not that simple after all.)

Re:The answer is DRI!

[X-Guy]

Why? Does the DRI support multiple screens yet? Multiple simultaneous cards? Does it support accelerated indirect rendering yet? What about workstation features like overlays or window IDs? Quad-buffered stereo? NVIDIA's drivers do all these things and the DRI still sucks like it did three years ago. NVIDIA was wise to stay away from it.

What about OpenBSD

[motown]

I checked this guy's site, and it appears he really did some impressive work hacking around those binary FreeBSD drivers and somehow getting them to work with NetBSD (even though there are currently some serious issues left, but it looks like he already completed the hard part of the work).

With this feat, I wonder if a similar hack could now also be made for OpenBSD... Although probably nobody would be interested in this. I mean, come on: why modify an otherwise stable and highly secure (mostly server) OS through the use of experimental patches, combined with binary code originally meant for another OS, which would only be beneficial to 3d graphics support? ;) But still I'm wondering, purely from a theoretical and technical point of view, how much more difficult it would be. Not much, I reckon, since the BSD's (especially NetBSD and OpenBSD) are very closely related and therefore share a lot of code and architecture.

I guess that the next (sufficiently interesting) step would be Darwin (the x86-port ofcourse). Since Darwin is more distantly related from the BSD family than the rest of the BSD's (it's based on a FreeBSD-like layer on top of a microkernel), this might prove to be more of a challenge. :)

Aaaargh! So many operating systems to potentially support! Do you see now why I am such a proponent of DRI? ;)

Re:What about OpenBSD

[mirabilos]

I'd like that too, and maybe I'll try it out the
next days.
I've got a friend who's only prevented from using
OpenBSD (or MirBSD) as his main OS because there
{we,a}re no nVidia "drivers" (though I dislike to
speak of "drivers" in a unix OS).

Unloadable kernel modules are obsolete now?

[Rares Marian]

Sorry I'm in redundancy nazi mode now. Redundancy is most common cause of pointless acronyms.

Re:Unloadable kernel modules are obsolete now?

[Rares Marian]

I call as I see em.

Just publish the spec

[Animats]

It's really annoying that NVidia keeps their board spec a secret. QNX doesn't support NVidia for that reason.

NVidia does this partly to protect their overpriced "pro" Quadro line, which is basically the same as the GEForce line but costs about 3x as much. GeForce boards are crippled in software to keep them from doing a few things the Quadro boards do. Given the dinky market for "pro" graphics boards, I'm surprised they still bother.