The nvidiafb driver seems to conflict with the nvidia driver that this guide shows the user how to install. There should probably be a note either inline or in the troubleshooting section that tells the user to ensure the nvidiafb driver isn't compiled into the kernel or loaded as a module. Reproducible: Always Steps to Reproduce: 1. Enable nvidiafb as a module or compile it into the kernel. 2. Emerge the nvidia driver as described in the guide. 3. nVidia driver no workee. Disable nvidiafb to fix problem. Gentoo 2.6.12-r9
I'll check this one, does this issue exist with other kernel sources? like vanilla for example? Unless nvidiafb is autoloaded and then causes a conflict, I don't see how it being compliled as a module is an issue...
I'll need to get my graphics card back before I can test...so this will be delayed. Would it be possible for you to test with different kernel sources and let me know? What sources do you use currently?
I had a friend tell me this happened on his system also. But can say at least on my amd64 it is enabled and not a problem. Seems like an x86 issue.
Checking on an x86 system that the nvidia devices aren't created on, it doesn't have nvidiafb enabled, but yet with the 2.6.14-gentoo-r2 is still fails to create the nodes in /dev . Added creating them to local.start and system loads up fine. At first he thought it was udev, since the kernel wanted udev-073, but that didn't help either. I looked and can't find a reason they aren't being created either. So on one x86 system, the problem was nvidiafb, not sure the kernel used. On the other system, nvidiafb not enabled and using 2.6.14-gentoo-r2 and udev-071 and 073 but not creating nvidia devices.
Does this problem still exist with the later kernels and drivers?
(In reply to comment #5) > Does this problem still exist with the later kernels and drivers? Dunno about that, but I know that for the 2.6.11 and 2.6.12 kernels, nvidiafb conflicts with the binary nvidia driver, and must not be compiled into the kernel at all.
this is something that will always exist until the nvidia peeps opensource their driver ... and since that will never happen, this will always be an issue
(In reply to comment #7) > this is something that will always exist until the nvidia peeps opensource > their > driver ... and since that will never happen, this will always be an issue > Right, which is why that point should be made clearly in the guide: it's a one-or-the-other, but not both choice.
Created attachment 77089 [details, diff] nvidia-guide.xml.patch Okay, this implements a paragraph and code example for removing the nvidiafb drivers from x86 kernels only. In the future (if other arches develop issues as well), then I suppose this could be expanded into its own <section>, maybe something like "Arch-specific notes".
Created attachment 77091 [details, diff] nvidia-guide.xml.patch As per fox2mike's suggestion, added a bit about selecting an alternative framebuffer. vesafb is known to work well with nvidia cards; it's up to the user to decide between vesafb and vesafb-tng. Changed the whole thing into a <section>Arch-specific notes</section> for clarity's sake. This way, users can more easily skip it if it doesn't apply.
I noticed in the patch that you said this bug only affects x86 systems, but I'm the original reporter and I experienced this bug on an AMD64 (ia64?) install. I sincerely apologize for not having chosen the correct option from the hardware dropdown list.
(In reply to comment #11) > I noticed in the patch that you said this bug only affects x86 systems, but I'm > the original reporter and I experienced this bug on an AMD64 (ia64?) install. I > sincerely apologize for not having chosen the correct option from the hardware > dropdown list. > Which is it, though? ia64 is Itanium. EM64T (or EMT64) is Intel's Pentium4 and derivatives with 64-bit extensions and some of the 64-bit instruction set. AMD64 is a fully 64-bit processor. Which do you have? I'll include the appropriate change in the patch, but first I need to know your arch. :) Other AMD64 users have not reported such issues.
(In reply to comment #12) > Which is it, though? ia64 is Itanium. EM64T (or EMT64) is Intel's Pentium4 and > derivatives with 64-bit extensions and some of the 64-bit instruction set. > > AMD64 is a fully 64-bit processor. Which do you have? I'll include the > appropriate change in the patch, but first I need to know your arch. :) Other > AMD64 users have not reported such issues. Ah, I'm sorry, I had crazy incorrect notions in my head. The processor is AMD64, and the kernel goes into arch/x86_64 when I compile. Thanks for your patience. ;)
on a side note, the nvidia module could equally do this check (linux-mod gives you functionality for this) and warn/error whatever.
Created attachment 80825 [details, diff] nvidia-guide.xml.patch A new, revised patch. Updated to include AMD64, as the users here have requested (and I've seen lots of similar stuff on the forums for that arch). Tightened the patch up, so it's all ready to go unless there are any more requests.
Fixed in CVS after adding a small section on fb documentation in the kernel src dir. Thanks everyone for their inputs and Josh for his patch.
