Ubuntu 20.04, 3 machines, 3 different kernel versions, often missing headers and module-extra

Polymerase

3/11/23, 3:09 AM

Using Ubuntu 20.04 on 3 different desktops with rather old CPU (Core2 Q9650, AMD FX-8320, Xeon E5-2678 v3).

The installation was made from Ubuntu Desktop ISO. The machines are updated regularly. I didn't remember doing anything special. But the kernel on these machines are upgraded from 5.4, then 5.8 and now currently Ubuntu 20.04.3 with kernel 5.11. This is OK as long as the machines run stable.

However uname -rs show different kernel versions on each machine, even if they are all up to date

Linux 5.11.0-27-generic # on Xeon machine
Linux 5.11.0-38-generic # on Core2 machine
Linux 5.11.0-40-generic # on AMD-FX machine

In particular, each time the minor version of kernel is updated, eg from 5.11.0-aa to 5.11.0-bb. The Core2 and AMD-FX machines almost always failed to get the corresponding linux-headers and linux-modules-extra. Resulting in a loss of network when rebooted. In such a case, I reboot to Grub menu, select to start with previous kernel version. Then sudo apt install the missing parts. Reboot and the machine is back to normal operation.

The Xeon machine seems to stick with 5.11.0-27-generic regardless of sudo apt upgrade I run from times to times. And this whether with secure boot enabled or not.

QUESTION 1: What is the criteria Ubuntu uses to decide which kernel version to install?

QUESTION 2: Any reason Ubuntu updater consistently failed to fetch linux-headers and linux-modules-extra when the kernel minor version is updated?

Thanks

241

0 + 0

upgrade

kernel

guiverc

3/11/23, 3:18 AM

Ubuntu LTS releases offer two kernel stack choices. (1) the GA or general stack which is most *stable* and default for Server installs, plus *flavors* where 20.04 or 20.04.1 media is used to install, or (2) HWE or hardware enablement stack which upgrades during the life of the product; ie. used 5.4 GA then 5.8 from 20.10 (20.04.2), then 5.11 from 21.04 (20.04.3) and next 5.13 from 21.10 (20.04.4), before it's final... HWE is the default stack for *flavors* with 20.04.2 or later media installs, and Ubuntu 20.04 LTS desktop installs. Server install ISOs let you choose the stack at install time

guiverc

3/11/23, 3:24 AM

Installation media will dictate defaults (with 20.04 & 20.04.1, vs. 20.04.2 & later as per prior comment making a difference). It can be selected at install time (GA or HWE stack) on some ISOs (Ubuntu Server ISOs with `subiquity`), though Ubuntu Desktop ISOs using `ubiquity` can also opt to use OEM kernels if detected to be better for your hardware which change what I've already mentioned... All of this can be overruled by user-choice too - https://wiki.ubuntu.com/Kernel/LTSEnablementStack but that doesn't impact -27, -38 or -40 you also mention.

Polymerase

3/11/23, 3:46 AM

@guiverc thanks for the explanation. All 3 machines were installed using the Ubuntu Desktop ISO. The Xeon machine uses UEFI boot, the 2 other machines use Legacy boot. SO I suppose this is why the kernel is updated every 20.04.x release. Doing `dpkg -l 'linux*' | grep hwe ` I see `linux-hwe-5.11-headers-5.11.0-xx` exists on those machines. How can I determine with certainty if the kernel is of HWE or OEM flavour?

guiverc

3/11/23, 3:48 AM

You're not using OEM kernels as per your paste; I see only HWE kernels (5.11) and not GA (5.4), and importantly spot the word *generic* (not OEM)

Elon Musk

I sit in a Tesla and translated this thread with Ai:

EN: Ubuntu 20.04, 3 machines, 3 different kernel versions, often missing headers and module-extra

TH: Ubuntu 20.04, 3 machines, 3 different kernel versions, often missing headers and module-extra

RO: Ubuntu 20.04, 3 machines, 3 different kernel versions, often missing headers and module-extra

RU: Ubuntu 20.04, 3 machines, 3 different kernel versions, often missing headers and module-extra

VI: Ubuntu 20.04, 3 machines, 3 different kernel versions, often missing headers and module-extra

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.