Issues configuring NVIDIA drivers on ASUS ROG STRIX G15 (RmInitAdapter Failed)

[burnt garnet]

Hey there, relatively new to linux in general, this is the third distro in which I'm attempting to figure out why nvidia-smi yields No devices were found.

I'm on a fresh install, only installing discord and a browser to open this server, followed the online guides to install the nvidia drivers as well as asusctl and supergfxctl. I believe only the AMD integrated graphics card is accessible despite the system recognising the dgpu and its driver (lspci -k:

    Subsystem: ASUSTeK Computer Inc. Device 1fff
    Kernel driver in use: nvidia
--
05:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Renoir [Radeon Vega Series / Radeon Vega Mobile Series] (rev c6)
    Subsystem: ASUSTeK Computer Inc. Device 1fff
    Kernel modules: amdgpu
[james@archmaster ~]$ sudo journalctl -b | curl -F 'file=@-' 0x0.st```

[Here](http://0x0.st/KuOU.txt) is the journalctl, am I missing anything?

[sharp igloo]

i'd first of all try with nvidia-open if you're not already

[burnt garnet]

I understood from the wiki that you need the proprietary drivers for a hybrid laptop setup?

[carmine sky]

-open is proprietary

[sharp igloo]

nvidia-open is the proprietary driver, it's just open kernel modules

[burnt garnet]

ah right

[sharp igloo]

it's what nvidia recommends u use if ur GPU supports it which urs does

[burnt garnet]

I'll try that and update

[burnt garnet]

didn't seem to do anything

[carmine sky]

@burnt garnet Try prime-run https://wiki.archlinux.org/title/PRIME#PRIME_render_offload

So install nvidia-prime and glxinfo

and run prime-run glxinfo | grep "OpenGL renderer"

[burnt garnet]

X Error of failed request: BadValue (integer parameter out of range for operation)
Major opcode of failed request: 150 (GLX)
Minor opcode of failed request: 24 (X_GLXCreateNewContext)
Value in failed request: 0x0
Serial number of failed request: 67
Current serial number in output stream: 68

[carmine sky]

and without prime-run?

[burnt garnet]

OpenGL renderer string: AMD Radeon Graphics (radeonsi, renoir, ACO, DRM 3.64, 6.17.1-arch1-1)

could be that I switched drivers without changing much of the config, I just uninstalled nvidia and installed nvidia-open in its stead, the wiki says the utils should be the same though

nvidia-smi still yields No devices were found, nvidia-settings shows nothing empty menu besides application profiles and nvidia-settings configuration

and dmesg still has a thousand ``NVRM: RmInitAdapter failed" type messages

[carmine sky]

You rebooted right?

[burnt garnet]

yeah

[carmine sky]

journalctl -b -0 | curl -F 'file=@-' 0x0.st

Upload the system log

nvm you already did

my bad

Did you add the nomodeset kernel param yourself?

[burnt garnet]

yeah it would be stuck on the bootloader otherwise

I can try removing it now to be safe

[carmine sky]

You can press e on the entry in the boot loader to edit it

remove all the nvidia options and nomodeset

[burnt garnet]

alright see you then

[burnt garnet]

pretty sure installing the new driver reset the params so it was clean

do you want me to regenerate the journal?

[magic marsh]

what kernel u use

lts or non lts

[carmine sky]

nah the params are a part of the bootloader config

[carmine sky]

ok

[burnt garnet]

I didn't switch journals it's what came installed

kernels*

[magic marsh]

give me uname -r

[burnt garnet]

http://0x0.st/KuVu.txt

[burnt garnet]

6.17.1-arch1-1

[magic marsh]

okay ur non lts

do u have nvidia package installed

[burnt garnet]

nvidia-open yeah

[magic marsh]

and u are on a 30 series. not even 40.

[burnt garnet]

true

3050 mobile

[magic marsh]

after all that, open nvidia drivers give 15 fps less on my experience.

[magic marsh]

i have a 4050

[burnt garnet]

this laptop is about 3 years old?

ish

[magic marsh]

its ok

[carmine sky]

Oct 09 14:33:30 archmaster kernel: NVRM: testIfDsmSubFunctionEnabled: GPS ACPI DSM called before _acpiDsmSupportedFuncCacheInit subfunction = 11.
Oct 09 14:33:30 archmaster kernel: NVRM: gpuWaitForGfwBootComplete_TU102: failed to wait for GFW_BOOT: (progress 0x1)
Oct 09 14:33:30 archmaster kernel: NVRM: kgspWaitForGfwBootOk_TU102: failed to wait for GFW boot complete: 0x55 VBIOS version 94.07.3A.00.FB
Oct 09 14:33:30 archmaster kernel: NVRM: kgspWaitForGfwBootOk_TU102: (the GPU may be in a bad state and may need to be reset)
Oct 09 14:33:30 archmaster kernel: NVRM: nvCheckOkFailedNoLog: Check failed: Generic Error: Not ready [NV_ERR_NOT_READY] (0x00000055) returned from kgspWaitForGfwBootOk_HAL(pGpu, pKernelGsp) @ kernel_gsp.c:3874
Oct 09 14:33:30 archmaster kernel: NVRM: RmInitAdapter: Cannot initialize GSP firmware RM
Oct 09 14:33:30 archmaster kernel: NVRM: GPU 0000:01:00.0: RmInitAdapter failed! (0x62:0x55:2015)
Oct 09 14:33:30 archmaster kernel: NVRM: GPU 0000:01:00.0: rm_init_adapter failed, device minor number 0
Oct 09 14:33:30 archmaster kernel: [drm:nv_drm_dev_load [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to allocate NvKmsKapiDevice

[magic marsh]

..

[olive cave]

Too few arguments for command addnote!
Usage: .addnote <name> <text>

[magic marsh]

kernel issue

[burnt garnet]

huh

should I try switching to lts then or something?

[magic marsh]

no

it says firmware failed to boot

[carmine sky]

make sure you got linux-firmware-nvidia

[magic marsh]

this

install it

[burnt garnet]

pacman says it's up to date (reinstalling)

[magic marsh]

give me pacman -Qs nvidia

[burnt garnet]

    The GBM EGL external platform library
local/egl-wayland 4:1.1.20-1
    EGLStream-based Wayland external platform
local/egl-x11 1.0.3-1
    NVIDIA XLib and XCB EGL Platform Library
local/lib32-nvidia-utils 580.95.05-1
    NVIDIA drivers utilities (32-bit)
local/libvdpau 1.5-3
    Nvidia VDPAU library
local/libxnvctrl 580.95.05-1
    NVIDIA NV-CONTROL X extension
local/linux-firmware-nvidia 20250917-1
    Firmware files for Linux - Firmware for NVIDIA GPUs and SoCs
local/nvidia-open 580.95.05-2
    NVIDIA open kernel modules
local/nvidia-prime 1.0-5
    NVIDIA Prime Render Offload configuration and utilities
local/nvidia-settings 580.95.05-1
    Tool for configuring the NVIDIA graphics driver
local/nvidia-utils 580.95.05-1
    NVIDIA drivers utilities
local/supergfxctl 5.2.7-2
    A utility for Linux graphics switching on Intel/AMD iGPU + nVidia dGPU laptops```

[magic marsh]

im not sure, should he try nvidia-dkms

[carmine sky]

nah, the driver woudn't try to load at all if it was the wrong kernel

Add options nvidia NVreg_EnableGpuFirmware=0 to /etc/modprobe.d/nvidia.conf @burnt garnet
and run sudo mkinitcpio -P

and reboot

[burnt garnet]

did that, still No devices were found

in smi

[carmine sky]

send journal

[burnt garnet]

http://0x0.st/KuVV.txt

[carmine sky]

is cat /sys/module/nvidia/parameters/NVreg_EnableGpuFirmware 1 0r 0

[magic marsh]

uhh.. ur disabling vbios?

dude

wait

[carmine sky]

The driver should take over

When it is enabled driver uses GSP which is RISC-V processor included with recent (Turing and newer) NVIDIA cards. It is responsible for doing things that were previously handled by driver itself (for example power management). In theory it should improve performance but due to some bugs it is currently slower for some people. When you disable it driver doesn’t use it and handles things by itself like before.

[magic marsh]

does your laptop have a mux switch

[burnt garnet]

no such file or directory

[magic marsh]

did u disable gpu in bios

[burnt garnet]

I have not

[magic marsh]

check it please

[burnt garnet]

the bios doesn't even give you an option to do that

[carmine sky]

The gpu is visible to the OS

[magic marsh]

mine does lol

[magic marsh]

idk its a connection error as it seems

[burnt garnet]

mine doesn't I tried that too

again this is the third distro I'm trying this on

[magic marsh]

bro

hmm

qhat distros u tried

[burnt garnet]

nvidia/parameters doesn't seem to exist

[magic marsh]

installed gentoo with full desktop and nvidia/hybrid desktop.
i installed nvidia drivers with proper hybrid graphics support across
ubuntu
fedora
arch
cachyos
gentoo
debian
openSUSE (this the most pmo)
freebsd
even cuda works with ai stuff

i mean

[magic marsh]

did u fry your damn gpu?

[burnt garnet]

maybe I misspelled the param?

[carmine sky]

cat /sys/bus/pci/devices/0000:01:00.0/power/runtime_status

[burnt garnet]

hell who knows, this just came back from the technician a few weeks ago because of gpu issues

[magic marsh]

anyway im leaving this to hm this is too firmware relatef

[magic marsh]

and we have our answer.

[burnt garnet]

suspended

the issue was a faulty cable, they replaced it

and I think that was even from the igpu so I doubt that was related

[carmine sky]

echo on | sudo tee /sys/bus/pci/devices/0000:01:00.0/power/control
sudo rmmod nvidia_drm nvidia_modeset nvidia
sudo modprobe nvidia_drm

[magic marsh]

what cable

[carmine sky]

Also is this Y or N cat /sys/module/nvidia_drm/parameters/modeset

[magic marsh]

why your laptop has cable for gpu

i set nvidia-drm.modeset=1 in linux parameters every distro for good meadure

[carmine sky]

No longer necessary since 560.35.03-5

[magic marsh]

oh

[burnt garnet]

Y

[magic marsh]

wait

[carmine sky]

now do the other commands I sent

[magic marsh]

i dont have nvidia instslled on my void this is awesome

anyway good luck gang

[burnt garnet]

thanks mate

yeah if I get another laptop I'll be sure to avoid nvidia shit

[magic marsh]

no.

the issue isnt nvidia.

[burnt garnet]

nvidia_modeset: command not found

[magic marsh]

the issue isnt u either

do the rmmod one by one

[burnt garnet]

yeah I did

for some reason bash doesn't recognise the nvidia_modeset command

[magic marsh]

its not a command

[carmine sky]

It's not a command

[magic marsh]

its a kernel module anong the others

the command is to unload them

[burnt garnet]

ah right right

[magic marsh]

do rmmod <module> one by one

[burnt garnet]

rmmod: ERROR: Module nvidia is in use by:rmmod: ERROR: nvidia_uvmrmmod: ERROR:

besides that everything else is unloaded and I modprobed drm

[carmine sky]

hm

[carmine sky]

is in use by:

lol

empty

[burnt garnet]

yeah

[carmine sky]

stop the various gpu services you started

[magic marsh]

lsof +D nvidia.ko

[carmine sky]

maybe it's those

[burnt garnet]

lsof: WARNING: can't stat(nvidia.ko): No such file or directory

[magic marsh]

uh..

[carmine sky]

you have to use a absolute path

or be there

modinfo nvidia

that will have the path

[burnt garnet]

there's a .ko.zst that's fine right

[carmine sky]

yeah

[magic marsh]

sudo fuser /dev/nvidia*

[burnt garnet]

lsof: WARNING: not a directory: /usr/lib/modules/6.17.1-arch1-1/extramodules/nvidia.ko.zst
lsof 4.99.5

[carmine sky]

I haven't used lsof to check what's using a kernel module so I may be wrong

[magic marsh]

maybe sudo lsof /dev/nvidia

that can also work

or..

[carmine sky]

sudo lsof /dev | grep nvidia

[burnt garnet]

COMMAND PID  USER FD   TYPE  DEVICE SIZE/OFF NODE NAME
sway    807 james 13u   CHR 195,255      0t0  889 /dev/nvidiactl```

nvidiactl being the only result of fuser/dev/nvidi*

[magic marsh]

sudo pkill xserver-xorg
sudo pkill egl-wayland

[magic marsh]

sudo kill -9 sway

[carmine sky]

nah dont go around killing everyithing lol

[magic marsh]

or sudo kill -9 807

[magic marsh]

LOL

[burnt garnet]

maybe reboot and try the modprobe thing again in tty?

[carmine sky]

Stop the service

[burnt garnet]

still haven't configured the initrc so it shouldn't start any display server

[magic marsh]

sway disagrees

lmao just kill the process

[carmine sky]

Is nvidiactl a service?

[burnt garnet]

well still on reboot I get sent to tty it doesn't open sway unprompted

[magic marsh]

... uhh no?

i dont think so

[burnt garnet]

it's a process

[magic marsh]

i think its a wants.target by some process

[burnt garnet]

doing that closed sway

[magic marsh]

yes

[burnt garnet]

I'll stay in the cozy tty for now

[carmine sky]

sudo lsof /dev/nvidiactl

ah sorry

[magic marsh]

bro u were in a display manager and expecting nvidis to be not used

[carmine sky]

you already did that

[burnt garnet]

You're right you're right

[carmine sky]

yeah sway but shouldn't sway have been running on the igpu :P

[carmine sky]

do these now

[magic marsh]

if he has nvidia-open sway ahs no problem with nvidia

[carmine sky]

then after you run them succesfully, upload the log

and you can try the prime-run glxinfo

[burnt garnet]

same issue as last time...

now nvidiactl is not on

lemme try rebooting

[magic marsh]

im dumb

lsof +D makes lsof search for a directory

no non +D

[burnt garnet]

same error

[carmine sky]

stop supergfxd

sudo systemcl stop supergfxd

[burnt garnet]

still nothing

[magic marsh]

alright thats it

sudo systemctl stop display-manager
sudo reboot
grub>e> put 3 in linux cmdline (runlevel3, no graphics target)
control x

[carmine sky]

also stop

nvidia-powerd

[burnt garnet]

runlevel 3 is already what's configured in the grub command line iirc

it doesn't start any display manager on boot

[magic marsh]

what

[burnt garnet]

Unit display-manager.service not loaded.

[magic marsh]

do this with nomodeset

[carmine sky]

Hold on there's a different way to do it with systemd

[burnt garnet]

I'm on grub

[carmine sky]

systemd.unit=multi-user.target

this cmdline

graphical.target is the dm

[magic marsh]

single user target?

[burnt garnet]

yeah loglevel=3 in cmdline

just checked

[carmine sky]

loglevel is different

[magic marsh]

... its not the loglevel

just put 3 in the cmdline

[burnt garnet]

also grub is making my pc scream for some reason

[carmine sky]

do this

[magic marsh]

isnt that the default

[burnt garnet]

also in the linux Line in grub's commands?

[carmine sky]

graphical is default right

[magic marsh]

idk

[burnt garnet]

the linux line right now is loglevel=3 quiet splash, I'll be adding systemd.unit=multi-user.target 3 as well

[carmine sky]

graphical is the default because if you got a display manager you don't need to change the target yourself

[carmine sky]

no need for the 3

[burnt garnet]

alright

[magic marsh]

bro the 3 is the same as the systemd setting

but firmer

[burnt garnet]

ok I'm in

[magic marsh]

i kinda think he should do this instead tbh

and nomodeset

[burnt garnet]

ah shit

[magic marsh]

to ban all graph driver from loading

[carmine sky]

nvidia needs modeset

[burnt garnet]

I'm guessing those commands are one-time considering I didn't see my changes when I went back in

or is there a way to save them and I'm stupid

[carmine sky]

the kernel cmdline?

[burnt garnet]

yeah

[carmine sky]

https://wiki.archlinux.org/title/Kernel_parameters#GRUB

[burnt garnet]

ah I see so the boot is one-time

fair enough

[carmine sky]

modprobe.blacklist=nvidia,nvidia_drm,nvidia_modeset

boot with this

[burnt garnet]

doesn't matter even with nomodeset I get the same rmmod error

for module nvidia

[carmine sky]

then lsmod | grep nvidia

[burnt garnet]

alright

so I'm thinking maybe I misdid all these commands because I see it does say Module nvidia_open is not currently loaded

though it says that also for keysmashes so probably not

[carmine sky]

no need to rmmod since its unloaded

continue

[burnt garnet]

it still does this no matter what

but yeah sure I did lsmod now I see the modules

[carmine sky]

there's no such modules as nvidia_open btw

[burnt garnet]

yeah I figured

[carmine sky]

sudo rmmod nvidia_drm nvidia_modeset nvidia nvidia_uvm

Try this, I added nvidia_uvm

[burnt garnet]

ok that seemed to work

forgot about that one

[carmine sky]

now continue

[burnt garnet]

ok modprobed drm

[carmine sky]

send journal

and try prime-run

[burnt garnet]

prime-run nvidia-smi still yields No devices were found

one sec for the journal

[carmine sky]

cat /sys/bus/pci/devices/0000:01:00.0/power/runtime_status

[burnt garnet]

http://0x0.st/KvVd.txt

[burnt garnet]

suspended

[carmine sky]

you ran the on right?

How about running rmmod, then echo on ... and then modprobe

[burnt garnet]

still yields suspended

[carmine sky]

cat /sys/module/nvidia/parameters/NVreg_DynamicPowerManagement

[burnt garnet]

I did rmmod nvidia_drm nvidia_modeset nvidia_uvm nvidia which executed without complaint

then I did echo mod | sudo tee etcetc

which yielded on and returned to prompt quickly

then I modprobed without complaint

[burnt garnet]

the parameters directory doesn't exist

[carmine sky]

ok

[carmine sky]

doesn't work

mistyped

[burnt garnet]

try http://0x0.st/KuW_.txt

[magic marsh]

sudo nano /etc/default/grub
GRUB_COMMANDLINE_LINUX_DEFAULT="3 nomodeset"

then

enter

fix your gpu

[carmine sky]

it was that way at the start too

Anyways I gotta go

we can continue later

I recommend feeding chatgpt or some other LLM the nvidia related logs

LLMs are good at troubleshooting with you

just verify their output by googling the commands and such

add quotes around parts of the command to see if it's a real thing

[burnt garnet]

damn it

[burnt garnet]

well guess I'll just ask the technician to take a look if they indeed tampered with the nvidia gpu

unfortunately it's technically holidays now soI gota wait till sunday

[magic marsh]

"archmaster kernel"

[burnt garnet]

that's just the hostname

[magic marsh]

can u

sudo pacman -S linux linux-lts linux-firmware linux-firmware-nvidia
grub mkconfig -o /boot/grub/grub.cfg

[burnt garnet]

and have both the standard and lts kernel installed?

[magic marsh]

shh

just do it

[burnt garnet]

what

eh sure whatever

linux-nvidia-firmware not found

[magic marsh]

...

[magic marsh]

edited

retry

[burnt garnet]

ok

I have a feeling that rebooting now would cause kernel panic but you're the boss it's done

[magic marsh]

remove the weird parameters from nano /etc/default/grub

leave quiet splash

reboot

[burnt garnet]

and loglevel=3 I'm guessing

[magic marsh]

yes

then sudo grub-mkconfig -o /boot/grub/grub.cfg

[burnt garnet]

grub-mkconfig again after reboot?

anyways nvidia-smi now yields NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

happened once before on a different distro

uname-r reveals I am now using the lts kernel

I guess I'll try installing the lts drivers

[magic marsh]

no before

[burnt garnet]

ah yeah I did that then

I'm trying to install nvidia open lts now

should I remove the previous driver first?

alright I made initcpio let's see what a reboot does

after reboot: nvidia-smi: No devices were found.

[magic marsh]

..

[olive cave]

Too few arguments for command addnote!
Usage: .addnote <name> <text>

[magic marsh]

no devices found

your card is tempered.

[burnt garnet]

if that's the case that stinks

I have no way of knowing if it worked back when this pc was running windows eiter

maybe I should reinstall it real quick just to see if it really is a hardware issue

[magic marsh]

can u do lspci -k -vvv and send us the entirety

[magic marsh]

dmesg also plz

i will look at them attentively when arrived home

[burnt garnet]

http://0x0.st/Ku4w.txt

[carmine sky]

Tried windows?

[burnt garnet]

not yet

I just used windows 10 on this pc before trying to switch

[carmine sky]

If you want you can install windows on a external ssd/hdd

You'll just have to change the linux esp type away from esp to something else

So windows creates it's own esp on the external drive

Then the install is portable and you can use it on any computer

[burnt garnet]

hmm

[magic marsh]

no

[burnt garnet]

I have an external hard drive lying around but there's also some data on it

[magic marsh]

windows on the go doesnt work anymore

[burnt garnet]

oh

[magic marsh]

he has to do it manually

[burnt garnet]

I mean I can partition my internal nvme drive manually

[magic marsh]

which takes long and ifs confusing

u have to open insrall.esd which is hidden

[burnt garnet]

this is a clean install of arch and I don't mind reinstalling seeing how I've done that a thousand times already

[magic marsh]

only rufus can see it

[carmine sky]

What windows on the go

You just install like normal

[magic marsh]

windows doesnt let u insrall to a external bus

like a usb thumb drive or hdd

[magic marsh]

thats called windows on the go

[carmine sky]

At least installing on a internal drive and then putting it in a usb enclosure works

Win 10 at least

[magic marsh]

wait thats a fire strat lemme try that

[burnt garnet]

well I don't really need that

[magic marsh]

but doesnt the drive uuids change

[burnt garnet]

I can just quickly partition 50 or 100 gigs of my nvme and install windows on that right?

or would that cause problems

I can also just completely install windows on this computer then reinstall arch again I installed the bare minimum to see if this works

[magic marsh]

did nouveau work

[burnt garnet]

never bothered to check that actually

how would I check that

[magic marsh]

idk bro

it wont work

your gpu and firmware isnt even running

let go of the software

we could always try nvidia-dkms though

[carmine sky]

No

[magic marsh]

idk much about windows internals sorry

[burnt garnet]

that's what RmInitAdapter failed means?

[magic marsh]

yes

maybe your kernel is too new to support your hardware

[carmine sky]

Filesystem UUIDs are a part of the filesystem

GUIDs are a part of the partition table

UUID in the common context means filesystem uuid

[burnt garnet]

I would think that but it also said the same thing on slackware which uses 5.15.19

so I doubt that

[carmine sky]

Lol what

You mean too old

But the gpu is from 2021

[magic marsh]

idk i qaa just doing guessing game atp

anyway

in gonna say hardware failure

..

[olive cave]

Too few arguments for command addnote!
Usage: .addnote <name> <text>

[magic marsh]

do modprobe nvidia

gimme the error

[burnt garnet]

no error when I do that

just nothing

[magic marsh]

..

[olive cave]

Too few arguments for command addnote!
Usage: .addnote <name> <text>

[magic marsh]

wait

[magic marsh]

lsmod | grep nvidia

[burnt garnet]

nvidia_modeset       2273280  1 nvidia_drm
drm_ttm_helper         16384  3 amdgpu,nvidia_drm
nvidia_uvm           4022272  0
nvidia              15888384  6 nvidia_uvm,nvidia_modeset
video                  81920  4 asus_wmi,amdgpu,asus_nb_wmi,nvidia_modeset```

[burnt garnet]

Installed windows, looks like the Nvidia GPU displays code 43 and doesn't work there either