ubuntu重启后显卡驱动掉了

ubuntu20.04重启后，显卡驱动掉了。
nvidia-smi检测不到显卡，双屏幕，只点亮了一个屏幕。
到系统about中检查，看到显卡驱动是llvm。
说明显卡驱动真的掉了。

参考 ubuntu重启后丢失驱动

查看已安装显卡驱动

ls /usr/src | grep nvidia

显示

nvidia-525.89.02

如果没有dkms的话，安装一下

sudo apt install dkms

因为我的系统是再生龙安装的，里面已经有dkms了，就跳过安装dkms的过程了。
dmks安装nvidia

sudo dkms install -m nvidia -v 525.89.02

用dkms status查看

nvidia, 525.89.02, 5.15.0-69-generic, x86_64: installed
virtualbox, 6.1.38, 5.15.0-56-generic, x86_64: installed
virtualbox, 6.1.38, 5.15.0-69-generic, x86_64: installed

没有执行dkms install前，dkms status只有virtualbox这两行。

查看下显卡

nvidia-smi

如果还不行

如果还不行
查看显示如下

sudo dkms status
[sudo] password for udeer:
nvidia, 525.125.06, 5.15.0-75-generic, x86_64: installed (WARNING! Diff between built and installed module!) (WARNING! Diff between built and installed module!) (WARNING! Diff between built and installed module!) (WARNING! Diff between built and installed module!) (WARNING! Diff between built and installed module!)
nvidia, 525.125.06, 5.15.0-78-generic, x86_64: installed

版本不对，估计是昨天安装virtualbox时，重装了virtualbox-dkms把环境搞乱了。
解决办法：卸载重装

sudo dkms remove -m nvidia -v 525.125.06 --all
sudo dkms install -m nvidia -v 525.125.06
sudo dkms status
nvidia-smi 正常了

如果还不行，则

sudo apt install --reinstall linux-headers-$(uname -r)

then after a reboot (not sure if necesary) I retook the steps here:

sudo dkms remove nvidia/510.73.05 --all
sudo dkms install nvidia/510.73.05

记得重启

https://forums.developer.nvidia.com/t/modprobe-error-could-not-insert-nvidia-exec-format-error/215071

如果此时不能启动gpu docker，则重启docker

sudo systemctl restart docker

参考 https://github.com/NVIDIA/nvidia-docker/issues/1034#issuecomment-632977198