[SOLVED] CPU interfaces dead for kernels > 6.18.2

Tojan · 2026-07-26 16:11:03

Hi all,

I will keep it brief: Since a few month each kernel update kills CPU based interfaces like audio or networking, requiring me to manually go back to the last version which still seems to be working fine. I was hoping it was some kind of packaging error which would solve itself with a next version but that doesn't seem to be the case. Looking into the logs I have not seen anything jumping out, but I also lack the knowledge to really understand most of the stuff happening on kernel level. The last working kernel seems to be linux-6.18.2.arch2-1-x86_64 and I have tried 6.19.8, 6.19.10, 6.19.11, 6.19.13, 6.19.14, 7.0.3, 7.0.10, 7.0.11, 7.0.13, 7.0.14, 7.1.2, 7.1.5, all of them show the same behavior. I suspect a broken dependency or package causing all this but I have no idea where to look anymore.

Hopefully you can give me some pointers to what may cause this and if any specific logs are required, let me know.

Edit: Solution was a broken fstab which resulted in the boot partition not being mounted by default. manually mounting the boot partition before doing the kernel update or having a working fstab will solve the problem of an incomplete kernel update.

Last edited by Tojan (2026-07-26 20:22:06)

cryptearth · 2026-07-26 16:46:53

ok, first of all: what you mean by "cpu interfaces"? either you have really old hardware - i'm talking about the kind before linux was even created - or you mis-understand something really low-level ...

... as both sound and network in x86 where never directly controlled by the cpu but were always connected via some bus (today on-board sound usually goes cpu -> chipset -> lpc -> sound) and network is done via pci-e

anyway: 6.18 is lts - so have you tried current lts 6.18.40? you only mention 6.18.2 and then 6.19.8 - haven't you updated in between?

as for information: at least a base line like cpu and board so we have at least a chance ("i have hardware issue" but listing hardware is blind guessing)

which linux-firmware packages have you installed?

ewaller · 2026-07-26 18:15:45

This sounds to me like the kernel cannot find its loadable modules. This often happens if the boot volume is not mounted on /boot when pacman updates the kernel. What is the output of

mount | grep boot

?

Last edited by ewaller (2026-07-26 18:16:09)

Tojan · 2026-07-26 19:16:52

cryptearth wrote:

ok, first of all: what you mean by "cpu interfaces"? either you have really old hardware - i'm talking about the kind before linux was even created - or you mis-understand something really low-level ...
... as both sound and network in x86 where never directly controlled by the cpu but were always connected via some bus (today on-board sound usually goes cpu -> chipset -> lpc -> sound) and network is done via pci-e

well this was my try with my limited computer science knowledge to differentiate between things. What I wanted to say is: things like the GPU, USB controller etc. are working as expected.

cryptearth wrote:

anyway: 6.18 is lts - so have you tried current lts 6.18.40? you only mention 6.18.2 and then 6.19.8 - haven't you updated in between?

I have not updated in between, the PC lay dormant for a while.

cryptearth wrote:

as for information: at least a base line like cpu and board so we have at least a chance ("i have hardware issue" but listing hardware is blind guessing)

Processors: 16 × AMD Ryzen 7 9800X3D 8-Core Processor
Memory: 32 GB of RAM (30,4 GB usable)
Graphics Processor 1: AMD Radeon RX 6800 XT
Manufacturer: Micro-Star International Co., Ltd.
Product Name: MS-7E47
Mainboard: MSI X870 GAMING PLUS WIFI

cryptearth wrote:

which linux-firmware packages have you installed?

the latest available I think, specifically:

linux-firmware-20260622-1-any
linux-firmware-amdgpu-20260622-1-any
linux-firmware-atheros-20260622-1-any
linux-firmware-broadcom-20260622-1-any
linux-firmware-cirrus-20260622-1-any
linux-firmware-intel-20260622-1-any
linux-firmware-mediatek-20260622-1-any
linux-firmware-nvidea-20260622-1-any
linux-firmware-other-20260622-1-any
linux-firmware-radeon-20260622-1-any
linux-firmware-realtek-20260622-1-any
linux-firmware-whence-20260622-1-any

Last edited by Tojan (2026-07-26 19:19:46)

Tojan · 2026-07-26 19:19:11

ewaller wrote:

This sounds to me like the kernel cannot find its loadable modules. This often happens if the boot volume is not mounted on /boot when pacman updates the kernel. What is the output of
mount | grep boot
?

The output is empty on the working 6.18.2 as well as on 7.1.5, which sounds wrong? mount called alone returns a bunch of devices as expected (hard drives, sys folders etc.)

Last edited by Tojan (2026-07-26 19:23:54)

seth · 2026-07-26 19:23:44

The output is empty, which sounds wrong?

cat /proc/cmdline

if the kernel is referenced as /boot/vmlinuz-* you're not booting from a dedicated partion, if it's just /vmlinuz-* you are and should™ have it mounted for the kernel update.
=>

lsblk -f

Tojan · 2026-07-26 19:30:43

seth wrote:

The output is empty, which sounds wrong?
cat /proc/cmdline
if the kernel is referenced as /boot/vmlinuz-* you're not booting from a dedicated partion, if it's just /vmlinuz-* you are and should™ have it mounted for the kernel update.

yeah it is. my question then would be: I have not changed my boot loader or the mounting files, how come the kernel updates were no problem up to 6.18.2? The PC is running arch for a few years now.

BOOT_IMAGE=/vmlinuz-linux

seth wrote:

=>

lsblk -f

NAME        FSTYPE FSVER LABEL   UUID                                 FSAVAIL FSUSE% MOUNTPOINTS
sda                                                                                  
└─sda1      ext4   1.0           1be959b0-78a7-40f0-b349-56a4e5507487  434,1G     0% /efi
sdb                                                                                  
└─sdb1      ntfs         HDD     1C5414E25414C086                                    
sdc                                                                                  
└─sdc1      ntfs         860 Evo 50A2F59AA2F5852E                                    
nvme0n1                                                                              
├─nvme0n1p1 vfat   FAT32         9DC7-9A3E                                           
├─nvme0n1p2 swap   1             74847257-90ad-4c40-859f-f2597d965498                [SWAP]
├─nvme0n1p3 ext4   1.0           94f79692-2824-454d-bc3b-cbba4ed0e647    4,2G    93% /
└─nvme0n1p4 ext4   1.0           b00f629f-125f-4135-8062-c467afc8c9cf  153,6G    72% /run/media/tojan/NVME

seth · 2026-07-26 19:33:31

Can you

sudo mount /dev/nvme0n1p1 /boot

cat /etc/fstab

Ceterum censeo: 3rd link below. Mandatory.
Disable it (it's NOT the BIOS setting!) and reboot windows and linux twice for voodo reasons.

Tojan · 2026-07-26 19:49:47

seth wrote:

Can you

sudo mount /dev/nvme0n1p1 /boot

cat /etc/fstab

# Static information about the filesystems.
# See fstab(5) for details.

# <file system> <dir> <type> <options> <dump> <pass>
# /dev/nvme0n1p3
UUID=94f79692-2824-454d-bc3b-cbba4ed0e647       /               ext4            rw,relatime     0 1

# /dev/nvme0n1p1
UUID=9DC7-9A3E          /boot           vfat            rw,relatime,fmask=0022,dmask=0022,codepage=437,iocharset=ascii,shortname=mixed,utf8,errors=remount-ro   0>

# /dev/nvme0n1p2
UUID=74847257-90ad-4c40-859f-f2597d965498       none            swap            defaults        0 0

# UUID=50A2F59AA2F5852E LABEL=860Evo
/dev/sda1               /efi	ext4            rw,nosuid,nodev,relatime,errors=remount-ro       		0 0

#/dev/sdb1               /run/media/tojan/HDD    ntfs3           rw,nosuid,nodev,relatime,uid=1000,gid=1000,iocharset=utf8       0 0

#/dev/sdc1               /run/media/tojan/860Evo ntfs3           rw,nosuid,nodev,relatime,uid=1000,gid=1000,iocharset=utf8       0 0

# UUID=b00f629f-125f-4135-8062-c467afc8c9cf
/dev/nvme0n1p4          /run/media/tojan/NVME   ext4            rw,nosuid,nodev,relatime,errors=remount-ro      0 2

seth wrote:

Ceterum censeo: 3rd link below. Mandatory.
Disable it (it's NOT the BIOS setting!) and reboot windows and linux twice for voodo reasons.

Could you elaborate on what you mean it is not the BIOS setting? I disabled it in BIOS back when I installed Arch as recommended, so the setup should be the same.

Last edited by Tojan (2026-07-26 19:52:25)

seth · 2026-07-26 19:54:17

I posted that because of the ntfs partitions; users often confuse windows fast-start w/ the BIOS fast boot (the latter just disables some self checks and is irrelevant itr)

nvme0n1p1 is supposed to be mounted at /boot, I suspect this fails because you're already lacking the vfat module?

uname -a
pacman -Qs kernel
modinfo vfat

nb. the particular line is bogus but I assume the trailing ">" is a result of you copy-pasting this out of some pager (eg. less)?

Tojan · 2026-07-26 20:15:38

seth wrote:

nb. the particular line is bogus but I assume the trailing ">" is a result of you copy-pasting this out of some pager (eg. less)?

nope, that '>' was exactly the problem. I must have tempered with the fstab a while back without remembering and by accident placed it there. So summary: make sure your fstab is correct and the boot partition gets properly mounted and you can update the kernel like everyone else.

seth wrote:

I posted that because of the ntfs partitions; users often confuse windows fast-start w/ the BIOS fast boot (the latter just disables some self checks and is irrelevant itr)

Ah I see where this is coming from now. the two ntfs partitions are just a left-over from when the PC still run on windows, have not seen a reason to change them as they are not used anyway rn.

Thanks for your time and sharp eye!

Last edited by Tojan (2026-07-26 20:16:17)

cryptearth · 2026-07-26 20:22:54

Tojan wrote:

NAME        FSTYPE FSVER LABEL   UUID                                 FSAVAIL FSUSE% MOUNTPOINTS
sda                                                                                  
└─sda1      ext4   1.0           1be959b0-78a7-40f0-b349-56a4e5507487  434,1G     0% /efi

THAT is VERY wrong!
aside from an ESP has to be FAT32 by uefi spec you dedicated an entire drive to just the boot loader
these are the kinds of issues and problems that result in

Tojan wrote:

yeah it is. my question then would be: I have not changed my boot loader or the mounting files, how come the kernel updates were no problem up to 6.18.2?

// F5

oh, and that of course:

Tojan wrote:

nope, that '>' was exactly the problem. I must have tempered with the fstab a while back without remembering and by accident placed it there. So summary: make sure your fstab is correct and the boot partition gets properly mounted and you can update the kernel like everyone else.

Tojan wrote:

Edit: Solution was a broken fstab which resulted in the boot partition not being mounted by default. manually mounting the boot partition before doing the kernel update or having a working fstab will solve the problem of an incomplete kernel update.

don't get me wrong - there's nothin wrong with playin around with such stuff - but if you lack experience maybe asking before doing and verifying would be adviseable

seth · 2026-07-26 20:43:46

and by accident placed it there

<> is simply dangerously close to shift and control

The ESP concern is very valid, though mounting random stuff into /efi isn't an immediate problem any kind of efi update would fail.

find /boot /efi -iname '*.efi'

It's also generally advisable to use the UUID syntax since the enumeration of the device nodes isn't stable (eg. the UUID above would have fit the ntfs sdc1)

Arch Linux

#1 2026-07-26 16:11:03

[SOLVED] CPU interfaces dead for kernels > 6.18.2

#2 2026-07-26 16:46:53

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#3 2026-07-26 18:15:45

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#4 2026-07-26 19:16:52

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#5 2026-07-26 19:19:11

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#6 2026-07-26 19:23:44

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#7 2026-07-26 19:30:43

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#8 2026-07-26 19:33:31

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#9 2026-07-26 19:49:47

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#10 2026-07-26 19:54:17

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#11 2026-07-26 20:15:38

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#12 2026-07-26 20:22:54

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

#13 2026-07-26 20:43:46

Re: [SOLVED] CPU interfaces dead for kernels > 6.18.2

Board footer