https://www.youtube.com/watch?v=CkyQhfhf048&t=628s
For point releases, some users were reporting that their older hardware was incompatible with the newer kernels provided by the HWE enabled via default in the point releases.
Perhaps, we should include a kernel manager for those who wish to upgrade to the HWE or stay on the LTS kernels for older system stability vs HWE/bleeding edge.
I confirm that new kernels often create problems on old hardware
I have the problem on my Dell Inspiron3793: NO problem with Kernel 5.19 but problem with 6.x Kernels
Regressions have been reported after every kernel release for the entire life of Ubuntu. This seems daily-routine for the Kernel Team, so I wonder if this topic should be in the Kernel section instead of Desktop.
When folks reporting regressions show up in support venues, routine advice is to 1) File a bug report against the linux package detailing the regression, and 2) Select their last working kernel at GRUB as a workaround. This works for new kernels at point releases, too. The older kernel is still there, typically for a couple weeks.
There are also already several AskUbuntu questions clearly detailing the process to switch from the HWE metapackage to the GA metapackage. It is one apt command, reboot, two more apt commands to clean up and avoid running out of /boot space (another routine support issue). The same can be done in GUI using Synaptic.
I worry that a “kernel manager” application may result in 1) Fewer regressions getting reported (and thus actually fixed), and 2) More users breaking their systems, and correspondingly greater support volume, by trying inappropriate kernels made easy (“Well, I thought I’d try the -azure kernel, and now my hardware won’t boot anymore. I’m going to rant that Ubuntu is confusing and unreliable!”)
I still would’ve expected all LTS releases to default to the LTS kernel instead of HWE. Then again, that’s why there are other distributions that stay on the LTS kernel cycle (with the option of bleeding edge ISO images) like Linux Mint.
For newer hardware, it may not be as big of a deal, but even some newer hardware could have issues with a interim release kernel.
I remember back when I had an AMD 5500U that snowballed (white noise screen) after a 5.13 update and I had to Grub into the previous kernel and remove that kernel, so complications can occur.
I always recommend sticking with an LTS kernel unless you absolutely need a rolling release kernel.
There is quite a difference between “bleeding edge” or “rolling release” kernels, though, and what Ubuntu offers with the HWE kernel.
Unless I’m misunderstanding, the current Ubuntu HWE kernel is based on 6.2, while the most recently released “stable” kernel is 6.6. Without significant manual intervention, I’d feel fairly confident saying no Ubuntu user is going to get pushed a “just-released”, untested kernel. Not to say there still can’t be issues introduced (as someone with an AMD processor and an fTPM chip that was never patched by the manufacturer, I experienced the fun of all that mess), but ultimately I’d think it’s a worse experience if the average user can’t pick up a 1-year-old machine and have the core components work “out of the box” - and that’s a real risk with kernels that are as old as 5.15 is now.
Exactly.
There used to be a common support problem: User installed older Desktop LTS onto newer hardware (“for the stability!”), then complained loudly when their new/fancy/expensive hardware didn’t work (“Ubuntu stinks! I’m going rant everywhere!”).
The change to Desktop LTS defaulting to HWE kernels has reduced those to a trickle of corner-cases. From a support perspective, it seems a success.
Clarification:
True… just a couple of commands – assuming that the HWE kernel doesn’t brick the older machine before the changes can happen lol. I think it would be a good idea to keep the LTS kernel in there as an option to Grub into if a catastrophic issue occurs.
In my humble opinion HWE should be in the backports repository instead of installed by default
If somebody is encountering regressions often enough to be wary of new kernels, that suggests too few folks with that hardware are participating in testing.
That’s easy enough to fix:
Grab a pre-release .iso, make a LiveUSB, boot it, enter the “Try Ubuntu” environment, and test that hardware thoroughly against the proposed kernel…without touching the working, installed system.
Reporting regression bugs (well, any bugs!) during the six months before a release is, by far, the best time to do so. That’s when developer attention (and kernel engineer attention) is focused on bugs for that release.
Widespread community testing before release has been a bedrock of Ubuntu development, and a vital role for community participation, since the beginning (18 years). Debian, too (30 years).
True, this is why all of us use different systems for testing.
Old and new architectures, with different types of processors.
I moved this discussion to Kernel since this really does revolve around the kernel and not so much the desktop.
What you might not know is the kernel team backports a lot of fixes via patches to the kernel from later kernel versions, so keep that in mind.
There is site https://linux-hardware.org/ where you can find if your computer is tested by some user and on his/her test result it will show if there is some hardware incompatibility or problem.
Such tests are not automatically run for every user only if manually tested and maybe Ubuntu will somehow subscribe to this or something and will process such results to gain market traction and make a difference to this abandoned world of hardware shit but only if desired user…
I think an app similar to Firmware Updater one would help all users manage the kernel based on the hardware available. Especially the non-expert user.
It seems that 6.* kernels are incompatible with some old nvidia drivers