4

Based on my recent experience with the Nvidia proprietary driver released by Ubuntu (NOT directly from Nvidia), I am curious how drivers are checked for production LTS updates. Ubuntu states that production release LTS represents the "best work" from Canonical Ubuntu. When I just updated the production kernel on 20.04.1 LTS, as urged by the stock Ubunutu GUI application Software Updater, the Nvidia proprietary driver failed. Under a separate posting/thread ( 20.04.1 LTS kernel 5.8.0-34 nvidia build fails ), a partial answer was posted by Terrance that was I able to use to find a working solution. I fully understand that third party repos (such as ElRepo for Enterprise Linux, not Ubuntu) have difficulty with always staying current on the latest production updates, particularly kernels or some basic pervasive systems environments outside the kernel. However, this issue arose with a LTS production deployment update from Ubuntu. I fully understand that Ubuntu paid professional staff have only limited resources, and things can "fall through the cracks". Is this common? If so, after a new major system update is suggested by, say, Software Updater, how long should one wait? Several weeks and check Ask Ubuntu to see if issues have arisen? Note that Ask Ubuntu is not LTS specific, but includes enthusiast Ubuntu releases (non LTS). Thank you for any insights into the Ubuntu LTS deployment and development processes.

Yasha Karant
  • 961
  • I regret to confirm that it quite common. Almost every time when a new HWE kernel is rolled out these sort of issues with different drivers happen. But I am voting to close this question because it invites for a discussion. – Pilot6 Jan 09 '21 at 20:38
  • The problem is not that something is wrong with the drivers. They work with the 5.8 kernel on Ubuntu 20.10 release. The problem is that someone forgot to upgrade them in 20.04 repos. – Pilot6 Jan 09 '21 at 20:41

1 Answers1

3

I can explain what really happened.

In October 2020 Ubuntu 20.10 was released with the 5.8 Linux kernel. All drivers were tested and there are no problems with them.

For those who installed Ubuntu 20.04 from the 20.04.1 image, linux-image-hwe-20.04 meta package was installed. It was pointing to the same kernel as linux-generic (5.4). That was an initial mistake IMHO. Before the HWE meta appeared only with xx.04.2 releases. And it didn't cause so much trouble.

People who wanted to test a new kernel could install HWE before but at their own risk. I did it many times. But this time all of a sudden...

Recently the linux-image-hwe-20.04 was switched to the 5.8.0-34 kernel, but drivers stayed the same in the repos. Some of them are incompatible with the 5.8 kernel. This is a second mistake.

Kernels got an update to 5.8, and a lot of people is having a lot of trouble.

This can be more or less easily fixed by installing drivers from 20.10 repos.

So this is not a matter of driver testing. It is disappointing that every time a HWE kernel is released some kind of trouble with drivers happens. This time it is worse than ever. But I wouldn't complain very much about it.

Pilot6
  • 90,100
  • 91
  • 213
  • 324
  • From: https://maas.io/docs/ubuntu-kernels#:~:text=The%20acronym%20HWE%20stands%20for%20%E2%80%9CHardWare%20Enablement.%E2%80%9D%20You,the%20Ubuntu%20host%20is%20running%20a%20desktop%20environment. New hardware gets released all the time. If an Ubuntu host runs an older kernel, it’s unlikely that MAAS can support the hardware. Canonical does make every effort to backport more recent kernels enabling more hardware. The acronym HWE stands for “HardWare Enablement.”

    You also gain kernel improvements and new features when installing an HWE kernel. End quote. Due to limitations, more to follow

     – Yasha Karant Jan 09 '21 at 21:51
  • This was a good idea. But the procedures are not really polished. Someone by mistake submitter the hwe to 20.04.1 where it was not supposed to be. You don't need to post it here. We all know what HWE is. – Pilot6 Jan 09 '21 at 21:55
  • Continued. I am not complaining too much. However, production "enterprise" releases -- LTS -- "best work" -- present an issue when this sort of thing is "broken". Also, Pilot6 mentions 20.04.2 LTS. Doing updates with Software Updater, is the transition to x.2 automatic from x.1, or does this require the specific "upgrade" mechanism? In most cases, a sub-sub major production release is an update, not upgrade -- that is, 20.04.0 to 20.04.1 to 20.04.2 is an update, whereas, say, 20.04 to 20.10 would be an upgrade, and 20.x to 22.x would be a major upgrade. Upgrade in place works in LTS. – Yasha Karant Jan 09 '21 at 21:57
  • Upgrade to .2 doesn't require a special procedure. But a confusing thing that if you initially install 20.04 without a point, or xx.04.1 before the 20, you would never get a HWE kernel without a special command. That was the idea. But when you installed from .2, .3, etc you already had a new major kernel. It was supposed to be tested at that time. So a small mistake caused this mess. – Pilot6 Jan 09 '21 at 22:00
  • HWE and other specific acronyms may be well known to Ubuntu professionals, but those who may be professionals from other environments or distros do not know it. Most importantly, production HWE kernels (including for LTS) add more functionality in terms of hardware support (e.g.,a HWE production kernel release may solve the microphone problem on a Dell touchscreen enterprise laptop, otherwise only available in the non-public Dell Ubuntu driver). – Yasha Karant Jan 09 '21 at 22:02
  • @YashaKarant - This is an excellent, knowledgeable answer for the question that was asked. I think further discussion would be more appropriate at https://discourse.ubuntu.com, where you can (politely, professionally) discuss the issue directly with the Ubuntu Desktop Team. – user535733 Jan 09 '21 at 22:53