nvidia issue
In an attempt to configure CUDA on my fedora21 machine, I followed these instructions: http://docs.nvidia.com/cuda/cuda-gett...
I verified that I had a CUDA-capable GPU and that I was using the correct versions of the kernel, gcc, and glibc before starting. I chose the "local package installer" for Fedora 21 (cuda-repo-fedora21-7-0-local-7.0-28.x86_64.rpm).
I continued step by step until 6.2.2.3 of the recommended post-installation actions (running the binaries - deviceQuery). deviceQuery failed with a fatal error. I made the mistake of rebooting and found myself in more trouble.
When the system started up again, I got the "Oh no! Something has gone wrong." error. It said I could press enter to logout and try again, but that only led to a black screen.
I then hit control-alt-f2, logged in as root, and issued the following command:
lsmod | grep nouveau
nouveau appeared in the output. If you follow the Fedora instructions for the link I posted, you will see that step 8 of 3.3 says to proceed to the post-installation instructions. The "package manager installation" does not include steps for manually disabling nouveau.
I thought that nouveau might be the problem so I looked at the file /usr/lib/modprobe.d/blacklist-nouveau.conf. It contained:
# RPM Fusion blacklist for nouveau driver - you need to run as root:
# dracut -f /boot/initramfs-$(uname -r).img $(uname -r)
# if nouveau is loaded despite this file.
blacklist nouveau
I followed the advice and issued the following command as root:
dracut -f /boot/initramfs-$(uname -r).img $(uname -r)
Now the entire system is unbootable. I get a kernel panic saying that it can't mount the root fs.
How can I recover from this? How can I prevent this from happening again should I choose to reintall Fedora 21 and start over?