AMD Drivers

[amarand]

I've been getting the hang of the rapid CentOS 8 Stream, which means tons of very regular updates, and one of the problems is with my AMD Drivers.

Every time the kernel updates, I need to reinstall the proprietary AMD Drivers. Sometimes it "just works" (kernel updates, graphics still work after the reboot), sometimes it doesn't.

This week, it was a "just doesn't" week.

My kernel went from:

Code: Select all

4.18.0-338.el8.x86_64

To:

Code: Select all

4.18.0-331.rt7.112.el8.x86_64

And now, when I try to install the drivers, AMD is freaking out:

Code: Select all

Installed:
  amdgpu-versionlist-21.10-1247438.el8.noarch

Complete!
Last metadata expiration check: 0:00:09 ago on Wed 01 Sep 2021 10:47:25 AM EDT.
No match for argument: kernel-devel-4.18.0-331.rt7.112.el8.x86_64
Error: Unable to find a match: kernel-devel-4.18.0-331.rt7.112.el8.x86_64

What I really want is this to "just work." It's not rocket science. It's standard 1080p video on a well-established video card. Before, I was able to just install this, and it would work: AMD_Radeon_R9_280_15.12_x86

But that throws a different error:

Code: Select all

[root@home fglrx-15.302]# ./check.sh
Detected configuration:
Architecture: x86_64 (64-bit)
X Server: XServer 1.20.11
[root@home fglrx-15.302]# ./amd-driver-installer-15.302-x86.x86_64.run
Created directory fglrx-install.DU3d0a
Verifying archive integrity... All good.
Uncompressing AMD Proprietary Driver-15.302...........................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................
=====================================================================
 AMD  Proprietary Driver Installer/Packager
=====================================================================

error: Detected X Server version 'XServer 1.20.11_64a' is not supported. Supported versions are X.Org 6.9 or later, up to XServer 1.10 (default:v2:x86_64:lib:XServer 1.20.11_64a:none:4.18.0-331.rt7.112.el8.x86_64:)
Installation will not proceed.

Removing temporary directory: fglrx-install.DU3d0a
[root@home fglrx-15.302]#

What are my options, now that I'm command line only until I get this fixed?

CentOS 8 Stream feels like I'm back in the Wild West Fedora days when nothing was stable and everything was a chore. How is this better?

[tunk]

Don't know how to solve the problem, but:
Stream could be described as a rolling pre-release of the
next RHEL point release (some think it's a rolling beta).
I guess it's better for Red Hat as they get more people
to find and weed out bugs (and they're doing it for free).
For the rest of us, I don't think anyone would use Stream
for anything remotely critical.

[amarand]

Don't know how to solve the problem, but:
Stream could be described as a rolling pre-release of the
next RHEL point release (some think it's a rolling beta).
I guess it's better for Red Hat as they get more people
to find and weed out bugs (and they're doing it for free).
For the rest of us, I don't think anyone would use Stream
for anything remotely critical.

My choice of CentOS has always been to keep my RHEL skills up, for free. I don't use it for anything remotely critical...anymore. It used to be my "firewall server" now it's just my fiddling desktop with command line. When I turn it on. Which is less and less frequently these days. I feel like the never-ending updates are going to kill CentOS for the majority of the users who like to have a stable - and free- option over RHEL.

It's only just recently, with this weird jump from "el8" kernel to "rt7" (seems backwards, even in the version numbers), with no explanation as to why. And, of course, AMD hasn't added this kernel yet because it probably just came out, and the updates are too fast for them to keep up with.

Do I have any other options, under CentOS 8 Stream, to make a fairly common AMD video card work? Again, I'm not gaming, or doing 3D rendering, I just want a standard desktop environment, that mirrors RHEL - but for free.

[amarand]

Don't know how to solve the problem, but:
Stream could be described as a rolling pre-release of the
next RHEL point release (some think it's a rolling beta).
I guess it's better for Red Hat as they get more people
to find and weed out bugs (and they're doing it for free).
For the rest of us, I don't think anyone would use Stream
for anything remotely critical.

A few years ago (2019), someone said:

Code: Select all

CentOS 8 is bassed on Fedora 28. It has open source radeon version in kernel. If should recognise it and give default usability until you download and install AND’s driver.

Which is cool, if true. So does CentOS 8 Stream still have built-in "Open Source Radeon" drivers built into the kernel? If so, how can I switch it back over to those built-in drivers? I just need 1080p. Very basic desktop.

[tkuraku]

Just uninstall the proprietary drivers and the open source drivers will be active.

[TrevorH]

You appear to have managed to install the realtime kernel there which may be part of the problem.

Also, fglrx? That was the old old AMD driver from when they were ATI. It's been superceded by amdgpu for years.

[amarand]

1) The URL shared earlier got me to a different kernel: 5.14.0-1.el8.elrepo.x86_64 #1 SMP Sun Aug 29 19:09:35 EDT 2021 x86

I'm not sure where that puts me in the Stream of things. 5.14.0-1.el8 is far different from the kernel I started with (el8) and the kernel I ended up with (rt7), and the kernel I'm on right now (shown above).

You appear to have managed to install the realtime kernel there which may be part of the problem.

Also, fglrx? That was the old old AMD driver from when they were ATI. It's been superceded by amdgpu for years.

As an RHCE from way back, I'm used to using yum/dnf update, but over the past year or so, I've just used the GUI package manager update functionality, especially with the regularity of the CentOS 8 Stream updates, which are at least three or more times per week, no hyperbole.

So....

I am -super- confused how I "accidentally" installed the realtime kernel. It was never offered as a choice. I just did a standard GUI-based update, and bam, I was on rt7.

Can you offer any suggestions as to how I can revert back to the 4.x el8 line, or should I just stay on 5.x el8?

[amarand]

You appear to have managed to install the realtime kernel there which may be part of the problem.

Also, fglrx? That was the old old AMD driver from when they were ATI. It's been superceded by amdgpu for years.

Oh, I buried the lede here, sorry. Somewhere, somehow, I got a link to the following URL:

viewtopic.php?f=54&t=72778

That switched me over to the 5.14.0-1.el8.elrepo.x86_64 kernel, which resolved the graphics display issue.

I suspect this was a sledgehammer approach to the other recommendation:

Just uninstall the proprietary drivers and the open source drivers will be active.

It worked. But now, instead of being on my original kernel thread:

Code: Select all

4.18.0-338.el8.x86_64

And going to this kernel thread:

Code: Select all

4.18.0-331.rt7.112.el8.x86_64

I'm now on THIS kernel thread:

Code: Select all

5.14.0-1.el8.elrepo.x86_64

Which seems like a huge jump to a...development?...kernel.

Thoughts?

[amarand]

Ahh, it wasn't this thread, it was THIS thread:

viewtopic.php?t=71876

Another forum moderator recommended the steps...and the steps brought me to the development kernel.

Probably won't hurt anything...maybe? Any way to get back to the "production" kernel, or am I stuck here?

[TrevorH]

It's not a development kernel but it is not one that CentOS supply. That's a mainline linux kernel - as released by Linus - rebuilt by ELRepo for CentOS/RHEL. It's newer and more likely to have new bugs but it will support more hardware. The drawback is that it doesn't have the stable KABI that the CentOS/RHEL kernels have and also that it gets a new version about twice a week so you will be continually updating to stay current.

The CentOS/RHEL kernel is 4.18.0-305.something on CentOS Linux 8 and 4.18.0-336 or something for Stream (which also effectively has no stable KABI either).