I tried to GPGPU passthrough on KVM and tried to use GPGPU on CentOS 6 Cloud image(CentOS-6-x86_64-GenericCloud.qcow2) but I failed to install cuda packages using rpm network installation.I checking /etc/x11/xorg.conf.d/99-nvidia.conf I might correct the ‘ModulePath’ settings below
Section “Files”
ModulePath “/usr/lib64/nvidia/xorg”
ModulePath “/usr/lib64/xorg/modules”
ModulePath “/usr/lib64/xorg/modules/drivers” #Adding the Path
EndSection
And checking the /var/lib/dkms/nvidia/build/make.log I found it need perl to build the cuda package.
/bin/sh: perl: command not found
make[3]: *** [/var/lib/dkms/nvidia/352.79/build/nv-frontend.o] Error 127
make[2]: *** [module/var/lib/dkms/nvidia/352.79/build] Error 2
make[1]: *** [sub-make] Error 2
make[1]: Leaving directory `/usr/src/kernels/2.6.32-573.18.1.el6.x86_64’
NVIDIA: left KBUILD.
nvidia.ko failed to build!
make: *** [module] Error 1
So I installed perl packages with yum
[root@centos build]# yum install perl
And compilation was failed though ‘yum install cuda’ had been completed so I reboot the system to recognized it.
[root@gpu-centos build]# yum install cuda
Loaded plugins: fastestmirror, presto
Setting up Install Process
Loading mirror speeds from cached hostfile
- base: ftp.iij.ad.jp
- epel: epel.mirror.net.in
- extras: centos.usonyx.net
- updates: ftp.iij.ad.jp
Package cuda-7.5-18.x86_64 already installed and latest version
Nothing to do
Perhaps cuda rpm package need ‘Requires:perl’ in the spec file…
-masafumi