This is the beginnings of an ebuild for x86_64 nvidia drivers version 1.0-4499. I can't get it to actually install but it currently builds the driver. I do have the drivers installed by hand, on a 2.6.0-test7 kernel, and they work perfectly at 8x AGP.
Created attachment 19029 [details, diff] kernel ebuild file kernel ebuild file.
Created attachment 19030 [details, diff] patch file for 2.6 kernels this file goes in /usr/portage/media-video/nvidia-kernel/files/1.0.4499/
Created attachment 19031 [details] working nvidia-glx-1.0.4499 ebuild. This file works as is, providing the patch file I'll attach next is installed as well.
Created attachment 19032 [details, diff] defines patch, updated for 4499 Just an adaptation of the previous builds to add one define. Makes the 4499 patch work.
Created attachment 19059 [details] More complete ebuild to support 2.4 kernel and 2.6 kernel.
Created attachment 19060 [details, diff] Another possible patch that I worked on. Needs testing.
Comment on attachment 19060 [details, diff] Another possible patch that I worked on. Needs testing. this file goes in /usr/portage/media-video/nvidia-kernel/files/1.0.4499/
Augustus's ebuild and patch work perfectly for me on the 2.6-test7 kernel. -James
I retract my earlier statement. After installing the ebuild in attachments 19059 and the patch in 19060, my X server will not start on kernel 2.6.0-test7. When I rebuilt the module using my patch (attachment 19030 [details, diff]), everything started fine. -James
can you be more specific about "does not start". I'd be interested to know the exact symptoms, as I want to ensure that they're not the same symptoms augustus saw. -Brad
The server begins to start up, identifies my chipset as an "unknown nvidia chipset", continues starting, then finally logs a "NVO: unable to set up DMA mapping for client" and quits, leaving my VT in a partially usable state. It could be related to brad's problems, but it looks unrelated. I think my X gets a bit farther than his does before dying. I dunno if he has tried my patch to install the kernel modules by hand yet, but I'd be interested to know if that works for him. -James
X is now up and running for me after a complete reinstall and I don't know what's different from last time. I was having same problems with portage that I guess could have messed things up. I am working on my NVIDIA patch to get it working. I'm still trying to figure out why the resulting .ko file of each build is different. On another related note, I don't have AGP working. I'm trying to figure out why, on an x86_64 build, you don't get the usual AGP option when building the kernel. It enables it by default and even if I edit the .config manually, it has no effect.
AGP is now working. For some reason (ask a kernel developer) if "IOMMU support" is enabled in the kernel then AGP is not allowed to be compiled. I turn that off I get the options of: "/dev/agpgart (AGP Support)" "AMD Opteron/Athlon64 on-CPU GART support" I compiled those as modules and now I can modprobe them and then startx with "NvAGP" set to 2. Still working on the patch for the ebuild.
well, that is good to know. As soon as I get my nvidia GeForce card in that will work with my MB, I'll test this out and commit it. And also make some of those notes in the technotes.
Created attachment 19210 [details] Working 4499 ebuild for kernel. Supports 2.4 and 2.6 kernel. I tried it and it works now. Grab both the nvidia-1.1 and the NVIDIA_kernel-1.0-4499-2.6-20031014.diff.
Created attachment 19211 [details] nvidia modules.d file taken from 32-bit driver ebuilds Place this in the files directory.
Created attachment 19212 [details, diff] Working NVIDIA_kernel patch. Tested. Place in the files/1.0.4499 directory.
This works well for me over here. Commited both ebuilds (kernel and glx). Thanks guys :) !
