You are not logged in.

#1 2023-08-17 07:37:28

dr1fter
Member
From: Germany
Registered: 2022-01-21
Posts: 58

[solved] Updating from 6.4.8-arch1-1 -> 6.4.10.arch1-1 does not work

Description:

I recently updated - among other packages - from linux-6.4.8-arch1-1 -> linux-6.4.10.arch1-1

After reboot, X11 startup failed. According to dmesg, this was due to an inconsistency of nvidia package versions: installed version 535.98 of client differed from installed kernel-module (535.86.05).

I then compared `pacman -Q linux` (6.4.10.arch1-1) and `uname -r` (6.4.8-arch1-1). So it seems that - even though pacman tried to install linux-6.4.10, the kernel used for actually booting was not updated.

I re-installed both nvidia and linux, resulting in recreation of initramfs, but this did not change anything (reboot yielded the same error + command-output). After downgrading nvidia, X11 would at least startup again.


Steps to reproduce:

On my machine (before my workaround):

pacman -Syu && reboot

Any help very apreciated.

excerpt from /var/log/Xorg.0.log.old

6.458] (EE) NVIDIA: Failed to initialize the NVIDIA kernel module. Please see the
6.458] (EE) NVIDIA:     system's kernel log for additional error messages and
6.458] (EE) NVIDIA:     consult the NVIDIA README for details.
6.458] (EE) NVIDIA: Failed to initialize the NVIDIA kernel module. Please see the
6.458] (EE) NVIDIA:     system's kernel log for additional error messages and
6.458] (EE) NVIDIA:     consult the NVIDIA README for details.
6.458] (EE) NVIDIA: Failed to initialize the NVIDIA kernel module. Please see the
6.458] (EE) NVIDIA:     system's kernel log for additional error messages and
6.458] (EE) NVIDIA:     consult the NVIDIA README for details.
6.458] (EE) NVIDIA: Failed to initialize the NVIDIA kernel module. Please see the
6.458] (EE) NVIDIA:     system's kernel log for additional error messages and
6.458] (EE) NVIDIA:     consult the NVIDIA README for details.
6.458] (EE) No devices detected.
6.458] (EE) 
server error:
6.458] (EE) no screens found(EE) 
6.458] (EE) 

excerpt from dmesg:

Aug 16 21:40:27.039560 arch kernel: NVRM: API mismatch: the client has the version 535.98, but
1229                                     NVRM: this kernel module has the version 535.86.05.  Please
1230                                     NVRM: make sure that this kernel module and all NVIDIA driver
1231                                     NVRM: components have the same version.
1232 Aug 16 21:40:27.039630 arch kernel: NVRM: API mismatch: the client has the version 535.98, but
1233                                     NVRM: this kernel module has the version 535.86.05.  Please
1234                                     NVRM: make sure that this kernel module and all NVIDIA driver
1235                                     NVRM: components have the same version.
1236 Aug 16 21:40:27.039645 arch kernel: NVRM: API mismatch: the client has the version 535.98, but
1237                                     NVRM: this kernel module has the version 535.86.05.  Please
1238                                     NVRM: make sure that this kernel module and all NVIDIA driver
1239                                     NVRM: components have the same version.
... (the same error is repeated several times)

Last edited by dr1fter (2023-08-17 15:02:02)

Offline

#2 2023-08-17 10:43:03

astralc
Member
Registered: 2022-09-17
Posts: 127

Re: [solved] Updating from 6.4.8-arch1-1 -> 6.4.10.arch1-1 does not work

Maybe you didn't mount /boot, or did something that cause your bootloader to load older kernel instead? check that /boot mounted correctly if it separate partition and reinstall 'linux'

Offline

#3 2023-08-17 12:30:56

dr1fter
Member
From: Germany
Registered: 2022-01-21
Posts: 58

Re: [solved] Updating from 6.4.8-arch1-1 -> 6.4.10.arch1-1 does not work

when I checked in "rescue-shell" (because X11 did not start), according to `lsblk`, `/boot` was mounted correctly + the timestamps from initramfs-linux.img and vmlinuz-linux were pretty recent (same days vs. uname -a telling me kernel being from two weeks ago). Is there a means to check (after boot) from where kernel was loaded?

Offline

#4 2023-08-17 12:38:53

dr1fter
Member
From: Germany
Registered: 2022-01-21
Posts: 58

Re: [solved] Updating from 6.4.8-arch1-1 -> 6.4.10.arch1-1 does not work

however, I have a suspicion. some months ago, I added a new SSD (because my old one - Samsung 980 Pro - was affected by a firmware-bug, and I failed to update the firmware), and replicated all partitions (including boot/efi and root) to the new one. I assigned new partition-UUIDs and configured a new boot-entry. however, I could not reset the UUID for the efi-partition (it needs to be FAT32, maybe this is why).

I just checked: the timestamps for the old boot-partition (which was not mounted during boot, however) somewhat correspond to the kernel's timestamp. so might it be that by chance, during update, initramfs was written to one of the two boot-partitions, and during boot, the other one was (by bad luck) mounted? I suppose this should be fairly easy to find out (and fix, at least as a workaround). thanks! will try and post an update

Offline

#5 2023-08-17 15:01:28

dr1fter
Member
From: Germany
Registered: 2022-01-21
Posts: 58

Re: [solved] Updating from 6.4.8-arch1-1 -> 6.4.10.arch1-1 does not work

I can confirm that after manually replicating initramfs + vmlinuz-liunx to the "old" boot-partiation + again updating all packages (pacman -Syu), rebooting succeeded, and the expected kernel-version is loaded (according to uname -r)

Now it is for me to figure out how to ensure boot-partitions will no longer be - sporadically - be mixed-up. Thank you so much for your quick help

Offline

Board footer

Powered by FluxBB