[SOLVED] Is new kernel faulty ??

[stuarte]

Hi all,

Last week, in response to a system tray notification, I carried out a system update. This included updating the kernel from (IIRC) 5.0.15-70 to 5.0.15-71. Since that time, the (few) errors that I was getting have significantly increased in number and severity. The system has now reached the stage where the boot up process does not complete and drops into some sort of "emergency mode". Here all that can be done is to examine some sort of log using (IIRC) "journalctl -xd". Some of the entries therein are displayed in either pink or brown text, this before reaching the "journal started" for which the entries are shown in green.

I would now like to "roll back" the kernel to 5.0.15-70 to see if this helps reduce the problem, given that I had only three ACPI and six USB problems with this previous kernel. As I can no longer log in to this system I am going to have to use an older version of SystemRescue DVD.

[ASIDE: The ACPI problems seemed not to affect the overall performance of the system as they have been in existence for quite some time and I have always been able to log in without any problems. Also, online research has indicated that the ACPI subsystem does not always receive the same level of attention as that lavished on other subsystems.]

Does anyone know of a decent tutorial that uses that distro and walks someone through the process of rolling back the system to a previous kernel ?


Stuart

[claydoh]

You can select previous kernels from the Grub boot menu. You should have about three versions unless this is a very fresh install.

There are manual methods to boot to a specific kernel by default, but there is the utility Grub Customizer you can install that makes this much easier.

[Schwarzer Kater]

1. Boot and hit [Shift] and [Esc] to get the GRUB boot menu.
2. Select "(K)Ubuntu, with Linux 5.15.0-70-generic" to boot with the 5.15.0-70 kernel.

Now you can for example:

• do this until a newer kernel which works is released or modify your GRUB to always boot into 5.15.0-70 (make a backup of system files before altering them)
• remove the 5.15.0-71 kernel (best including the header files) and ignore new kernels/hold the 5.15.0-70 kernel
• try if the 5.19 kernel works by installing it with sudo apt install --install-recommends linux-generic-hwe-22.04

[stuarte]

Hi claydoh,

Thank you for such a swift response.

As I stated in my OP,

The system has now reached the stage where the boot up process does not complete and drops into some sort of "emergency mode".

Consequently, the boot up process does not reach the Ububtu/Memtest86/etc screen. The black screen now just shows the previously noted failing ACPI whatevers and then drops into some sort of emergency mode. So I cannot see(!) the grub boot menu let alone Grub Customizer (actually installed, IIRC) which uses a GUI.

Also, I have only two kernels on the system, 5.0.15-70 and 5.0.15-71 (the new one which may be causing/agrivating the problem). This is the reason that I think I may(!!) need to use SystemRescueCD from the optical drive (first in boot drive order in BIOS). I need to find out :-
a) how to mount the target hdd/partition;
b) which file(s) to edit to boot the previous kernel.

Stuart

[claydoh]

Consequently, the boot up process does not reach the Ububtu/Memtest86/etc screen

If it isn't getting to grub, then it most definitely is not the kernel update, since it isn't getting far enough to bein loading a kernel at all.

boot up process does not complete and drops into some sort of "emergency mode".

A bit more description of exactly what this means, and what you see, might be useful to see where it actually fails.


"Memtest86" indicates you are booting a "legacy" BIOS/MBR setup, so your MBR could be damaged, maybe. It happens.
I have forgotten most of this old-ish skool stuff, so I am not sure of the exact process to fix this from a live OS session.

[stuarte]

Hi Schwarzer Kater,

Thank you for such a swift response and your advice. I will try them first thing tomorrow morning, as soon as this library opens. However I have a couple of questions arising from your response.

First, with respect to

...hold the 5.15.0-70 kernel

, how do I hold a kernel ?
[ASIDE: I have never "held" a kernel before.]

Second, what is an "hwe" kernel ?

Stuart

[Schwarzer Kater]

[…]
"Memtest86" indicates you are booting a "legacy" BIOS/MBR setup, so your MBR could be damaged, maybe. It happens. […]

Actually *Ubuntu (automatically) added Memtest86+ version 6.10 to UEFI boot several weeks ago - one can test memory from the GRUB menu again with UEFI.
I think the oldest Kubuntu version that got this was 22.04.


stuarte see: https://wiki.ubuntu.com/Kernel/LTSEnablementStack

Users who installed Kubuntu 22.04 with the 22.04.0 or 22.04.1 installation medium can use this to get newer kernels. Users who installed Kubuntu with the 22.04.2 or newer installation medium get kernel 5.19 and newer anyway.
If one wants to keep the LTS kernel 5.15 one has to install Kubuntu 22.04 with the 22.04.0 or 22.04.1 installation medium


To prevent a package from updating (including the kernel and its headers) one could use apt-mark hold, here is an example:
https://connectwww.com/how-to-preven...-update/62052/

But if you don't even get to GRUB…
I usually use the rescue mode of a Debian or openSUSE installation USB stick in this case to reinstall GRUB (if this is the problem here - your description is a bit vague…), because I am mostly too lazy to chroot.


PS:

[ASIDE: I have never "held" a kernel before.]

Really? Not even when visiting the countryside / hiking in the woods? I remember lots of nature in Scotland.

[claydoh]

Actually *Ubuntu (automatically) added Memtest86+ version 6.10 to UEFI boot several weeks ago - one can test memory from the GRUB menu again with UEFI.

I don't think the version that supports EFI is in 22.04, I sure don't have that version. It is not something that would be added/upgraded after the fact to an LTS release.

[Schwarzer Kater]

OT:
claydoh : Hm, I have gotten Memtest86+ independently on three machines here at home in Kubuntu 22.04 (the oldest one in my bedroom is a mini PC from 2012 with a veeery old and basic UEFI…).
Bare metal installations.

PS: I just double-checked - it indeed could have come from other Linux installations, but no - it's on my MacBook Pro's Kubuntu 22.04 too (installed April 2022), and the only other operating system there is macOS.
I "think" you don't get it with a "minimal" installation, but this assumption could be wrong.

[Snowhog]

I don't think this is in 22.04, I sure don't have it.

https://packages.ubuntu.com/jammy/misc/memtest86+

[claydoh]

https://packages.ubuntu.com/jammy/misc/memtest86+

Wrong link , yours is for the OG version, which definitely does not support UEFI.
6.10 is in jammy-backports:

https://packages.ubuntu.com/jammy-backports/memtest86+

[claydoh]

I have gotten Memtest86+ independently on three machines here at home in Kubuntu 22.04

I am not seeing it as I have not updated grub in a bit, I assume. I definitely do NOT have it in my grub menu.

[Schwarzer Kater]

Oh, it came from backports? This could be the explanation why not everone has gotten it in 22.04 - because I have set backports to Pin-Priority: 500 on my machines.

So your thought in post # 5 is probably right and it could be MBR/BIOS.

PS: the default priority of *Ubuntu backports is 100.

[claydoh]

However:

$ apt policy memtest86+
memtest86+:
Installed: 5.31b+dfsg-4
Candidate: 5.31b+dfsg-4
Version table:
6.10-4~bpo22.04.1 100
100 http://us.archive.ubuntu.com/ubuntu jammy-backports/main amd64 Packages
*** 5.31b+dfsg-4 500
500 http://us.archive.ubuntu.com/ubuntu jammy/main amd64 Packages
100 /var/lib/dpkg/status

​

Since jammy-backports are enabled by default in *buntu, and I definitely have it enabled.
Odd (and off-topic)

My 22.04 setup is 100% stock, with very little to no extra software or PPAs installed, even.

Oh, it came from backports? This could be the explanation why not everone has gotten it in 22.04 - because I have set backports to Pin-Priority: 500 on my machines.

Yeah, makes sense, though I don't remember that one needed to do this. Though it is not a topic I would pay attention to.

[claydoh]

So your thought in post # 5 is probably right and it could be MBR/BIOS.

More off-topic:

To install something from foo-backports (NOT related to the kubuntu-backports PPA in any way):

sudo apt install -t RELEASE-backports $PACKAGENAME

Moar infoses and stuffs (a little outdated in spots): https://help.ubuntu.com/community/UbuntuBackports