You are not logged in.

#1 2018-02-06 16:14:52

marcs
Member
From: Italy
Registered: 2007-09-07
Posts: 63

[SOLVED] Update to nvidia-dkms-390.25-4 can't use CUDA anymore

I updated to the lastest 4.15 kernel and the latest nvidia-dkms version, module builds fine, but now I can't see my GTX 980 in clinfo, as the nvidia-utils package suggested I ran nvidia-modprobe but I still can't see my CUDA device on clinfo.

Can someone suggest a way to investigate this problem?

Before the update I was using linux-mainline 4.15 with nvidia-dkms 390.12 (custom package).

My nvidia card is a secondary card, I have an RX 580 as a primary video card.

I'm also using the opencl-amd package.

hwinfo output:

cpu:                                                            
                       AMD Ryzen 7 1700 Eight-Core Processor, 1342 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1550 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 2880 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 2850 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1284 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1294 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1304 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1315 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1366 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1366 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1371 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1366 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1362 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1363 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1733 MHz
                       AMD Ryzen 7 1700 Eight-Core Processor, 1638 MHz
keyboard:
  /dev/input/event2    Microsoft Natural Ergonomic Keyboard 4000 V1.0
  /dev/input/event5    Holtek Portable Keyboard
mouse:
  /dev/input/mice      Holtek Portable Keyboard
  /dev/input/mice      Evoluent VerticalMouse 4
monitor:
                       DELL U2414H
                       DELL U2414H
                       LG ELECTRONICS LG ULTRAWIDE
graphics card:
                       nVidia GM204 [GeForce GTX 980]
                       ATI Ellesmere [Radeon RX 470/480]
sound:
                       nVidia GM204 High Definition Audio Controller
                       ATI Audio device
                       AMD Audio device
storage:
                       AMD FCH SATA Controller [AHCI mode]
                       AMD SATA controller
network:
  enp6s0               Intel I211 Gigabit Network Connection
network interface:
  [...]
disk:
  [...]
partition:
  [...]
cdrom:
  [...]
usb controller:
                       AMD USB Controller
                       AMD USB Controller
bios:
                       BIOS
bridge:
                       AMD Host bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD PCI bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD Host bridge
                       AMD Host bridge
                       AMD Host bridge
                       AMD Host bridge
                       AMD PCI bridge
                       AMD FCH LPC Bridge
                       AMD Host bridge
hub:
                       Texas Instruments Hub
                       Hub
                       Linux Foundation 2.0 root hub
                       Texas Instruments Hub
                       Texas Instruments Hub
                       Linux Foundation 2.0 root hub
                       Linux Foundation 3.0 root hub
                       Texas Instruments Hub
                       Linux Foundation 3.0 root hub
memory:
                       Main Memory
unknown:
                       FPU
                       DMA controller
                       PIC
                       Keyboard controller
                       AMD IOMMU
                       AMD Non-Essential Instrumentation
                       AMD Encryption controller
                       AMD Non-Essential Instrumentation
                       AMD FCH SMBus Controller
                       Serial controller
                       Oculus VR,Inc Camera DK2
                       Leap Motion Controller
                       Oculus VR Rift DK2
  /dev/input/event7    Holtek Portable Keyboard
  /dev/input/event3    Microsoft Natural Ergonomic Keyboard 4000 V1.0

Edit forgot to add my dmesg log, this is what happens when i run nvidia-modprobe:

[   80.080942] NVRM: The NVIDIA probe routine was not called for 1 device(s).
[   80.080942] NVRM: This can occur when a driver such as: 
               NVRM: nouveau, rivafb, nvidiafb or rivatv 
               NVRM: was loaded and obtained ownership of the NVIDIA device(s).
[   80.080943] NVRM: Try unloading the conflicting kernel module (and/or
               NVRM: reconfigure your kernel without the conflicting
               NVRM: driver(s)), then try loading the NVIDIA kernel module
               NVRM: again.
[   80.080943] NVRM: No NVIDIA graphics adapter probed!
[   80.081015] nvidia-nvlink: Unregistered the Nvlink Core, major device number 237

Last edited by marcs (2018-02-06 16:51:08)

Offline

#2 2018-02-06 16:41:08

marcs
Member
From: Italy
Registered: 2007-09-07
Posts: 63

Re: [SOLVED] Update to nvidia-dkms-390.25-4 can't use CUDA anymore

Ok I panicked without reason I'm using vfio for using my nvidia card with PCI-passthrough, that was what nvidia-modprobe was nagging about.

Removing the modprobe configuration for vfio e rerunning mkinitcpio fixed it, also FYI if you are running both an AMD card and an nvidia card you have to disable the nvidia glx libs, i made this script for convenience:

#!/bin/sh

if [ "$#" -lt 1 ]
then
	echo "USAGE: $0 <on|off>"
	exit 1
fi

MODE=$1
NV_PATH="/usr/lib/nvidia/xorg"


case $MODE in
	"off")
		sudo mv $NV_PATH/libglx.so $NV_PATH/libglx.so_disabled
		sudo mv $NV_PATH/libglx.so.1 $NV_PATH/libglx.so.1_disabled
		sudo mv $NV_PATH/libglx.so.390.25 $NV_PATH/libglx.so.390.25_disabled
		;;
	"on")
		sudo mv $NV_PATH/libglx.so_disabled $NV_PATH/libglx.so
		sudo mv $NV_PATH/libglx.so.1_disabled $NV_PATH/libglx.so.1
		sudo mv $NV_PATH/libglx.so.390.25_disabled $NV_PATH/libglx.so.390.25
		;;
esac

# check

echo "status:"

ls -l $NV_PATH

Offline

Board footer

Powered by FluxBB