Van Snyder composed on 2025-02-27 19:06 (UTC-0800):

While running at level 3 in Debian 12.5, I got the following messages:

mce: {Hardware Error]: CPU: 8 Machine Check: 0 Bank 0: 8000004000040005
mce: {Hardware Error]: TSC 1838aa435b6d
mce: {Hardware Error]: PROCESSOR 0: b0671 TIME 140710368 SOCKET 0 APIC
20 microcode 12b

Motherboard is Micro-Star International Z790 GAMING PLUS WIFI (MS-7E06)

Processor is Intel(R) Core(TM) i9-14900K model 183 with 0x12b microcode

RAM is Micron Technology Part Number: CP16G56C46U5.C8D

The MB and CPU and RAM are only a few weeks old. Should I try to get
warranty replacement of something?

Your kernel is older than your CPU by about a year, so likely doesn't have enough
backporting to fully support it properly. A newer kernel could be all it takes to
make those MCEs go away.
--
Evolution as taught in public schools is, like religion,
based on faith, not based on science.

Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi,

On Fri, Feb 28, 2025 at 11:27:40AM -0800, Van Snyder wrote:

What's "mce?"

https://en.wikipedia.org/wiki/Machine-check_exception

"apt update" says everything is up to date, but the kernel is 6.1.0-18.
I believe there are several newer ones, maybe up to 6.1.0-31?

You can try a kernel from bookworm-backports or even download the source
of the latest Linux kernel and build a deb package for it.

Thanks,
Andy

--
https://bitfolk.com/ -- No-nonsense VPS hosting

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Fri, Feb 28, 2025 at 11:27:40 -0800, Van Snyder wrote:

"apt update" says everything is up to date, but the kernel is 6.1.0-18.
I believe there are several newer ones, maybe up to 6.1.0-31?

That's correct. You're probably missing the metapackage that brings
in new kernels automatically. For an amd64 machine, that metapackage
is named "linux-image-amd64". (If you use DKMS kernel modules, you'll
also want the corresponding linux-headers-* metapackage.)

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Fri, Feb 28, 2025 at 12:46:35 -0800, Van Snyder wrote:

On Fri, 2025-02-28 at 14:34 -0500, Greg Wooledge wrote:

On Fri, Feb 28, 2025 at 11:27:40 -0800, Van Snyder wrote:

"apt update" says everything is up to date, but the kernel is
6.1.0-18.
I believe there are several newer ones, maybe up to 6.1.0-31?

That's correct.  You're probably missing the metapackage that brings
in new kernels automatically.  For an amd64 machine, that metapackage
is named "linux-image-amd64".  (If you use DKMS kernel modules,
you'll
also want the corresponding linux-headers-* metapackage.)

The NVidia 570 driver is a kernel package. What's the name of the corresponding linux-headers-* metapackage?

Whatever your linux-image-* metapackage is called, you replace "image"
with "headers" to get the other metapackage name.

For amd64 they would be linux-image-amd64 and linux-headers-amd64.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Van Snyder composed on 2025-02-28 11:27 (UTC-0800):

On Thu, 2025-02-27 at 22:35 -0500, Felix Miata wrote:

Your kernel is older than your CPU by about a year, so likely doesn't
have enough
backporting to fully support it properly. A newer kernel could be all
it takes to
make those MCEs go away.

What's "mce?"

The part you failed to quote:

mce: {Hardware Error]: CPU: 8 Machine Check: 0 Bank 0: 8000004000040005
mce: {Hardware Error]: TSC 1838aa435b6d
mce: {Hardware Error]: PROCESSOR 0: b0671 TIME 140710368 SOCKET 0 APIC
20 microcode 12b

Machine Check (Exception) is a "Hardware Error", typically caused not by a hardware error, but by attempting to use hardware newer than the software, the kernel in particular.

"apt update" says everything is up to date, but the kernel is 6.1.0-18.
I believe there are several newer ones, maybe up to 6.1.0-31?

"Newer" kernels do include those newer than your 6.1.0-18 and it's "update" derivative 6.1.0-31. Updates are newer by date, due to the security and bug fixes
they contain. But, little if any support for newer hardware is added to update kernels, so their support for newer hardware is generally limited or absent.

"Newer" in another sense, means 6.2.x and up. Newer kernels for this sense of the
word support newer hardware than that generally available around the time of the
version selected for provision with the original release of the Debian 12 operating system you are using.

In another sense of the word "newer", that which is most likely applicable to your
environment, is a version newer than your i9-14900K CPU, which was released to market in the third quarter of 2023, when the upstream newest available kernel was
6.8.x, 7 major versions newer than Bookworm's 6.1.x.
--
Evolution as taught in public schools is, like religion,
based on faith, not based on science.

Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Fri, 28 Feb 2025 22:00:41 -0800
Van Snyder <van.snyder@sbcglobal.net> wrote:

On Fri, 2025-02-28 at 12:46 -0800, Van Snyder wrote:

On Fri, 2025-02-28 at 14:34 -0500, Greg Wooledge wrote:

On Fri, Feb 28, 2025 at 11:27:40 -0800, Van Snyder wrote:

"apt update" says everything is up to date, but the kernel is
6.1.0-18.
I believe there are several newer ones, maybe up to 6.1.0-31?

That's correct.  You're probably missing the metapackage that
brings
in new kernels automatically.  For an amd64 machine, that
metapackage
is named "linux-image-amd64".  (If you use DKMS kernel modules,
you'll
also want the corresponding linux-headers-* metapackage.)

The NVidia 570 driver is a kernel package. What's the name of the corresponding linux-headers-* metapackage?

The NVidia 570 driver is a kernel package. It's installed by
downloading and running an bash script that requires the kernel
headers. If I get the metapackage linux-image-amd64 will I need to
rebuild the NVidia driver every time it loads a new kernel?

If I get a new kernel by way of "apt update" it leaves a line in grub
to load older kernels. Will the metapackage do that so I can at least
boot something until I rebuild the NVidia 570 driver to go with a new
kernel?

It will normally install a new kernel and keep the previous one. I
can't remember if it removes earlier ones or whether I do that, as I
run apt autoremove regularly. Autoremove certainly leaves the last
kernel installed as well as the current one.

--
Joe

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Sat, Mar 01, 2025 at 04:06:37 -0500, Timothy M Butterworth wrote:

On Sat, Mar 1, 2025 at 4:04 AM Joe <joe@jretrading.com> wrote:

On Fri, 28 Feb 2025 11:27:40 -0800
Van Snyder <van.snyder@sbcglobal.net> wrote:

On Thu, 2025-02-27 at 22:35 -0500, Felix Miata wrote:

Your kernel is older than your CPU by about a year, so likely
doesn't have enough
backporting to fully support it properly. A newer kernel could be
all it takes to
make those MCEs go away.

What's "mce?"

"apt update" says everything is up to date, but the kernel is
6.1.0-18. I believe there are several newer ones, maybe up to
6.1.0-31?

When is the last time you rebooted? Debian does not provide live kernel patching. If apt says you are all up to date but uname shows an older
kernel then it is time to reboot.

That's only true if the linux-image-$ARCH metapackage is installed,
and if the user has been using "apt update" + "apt upgrade" (or
equivalents) on a regular basis.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Am Freitag, 28. Februar 2025, 21:46:35 CET schrieb Van Snyder:

On Fri, 2025-02-28 at 14:34 -0500, Greg Wooledge wrote:

On Fri, Feb 28, 2025 at 11:27:40 -0800, Van Snyder wrote:

"apt update" says everything is up to date, but the kernel is
6.1.0-18.
I believe there are several newer ones, maybe up to 6.1.0-31?

That's correct. You're probably missing the metapackage that brings
in new kernels automatically. For an amd64 machine, that metapackage
is named "linux-image-amd64". (If you use DKMS kernel modules,
you'll
also want the corresponding linux-headers-* metapackage.)

The NVidia 570 driver is a kernel package. What's the name of the corresponding linux-headers-* metapackage?

if the kernel is 6.1.0-31-amd64, the kernel-header-package is

linux-headers-6.1.0-31-amd64

If in doubt, install the "module-assistant" package, then start it with the command "m-a", and then over the ncurses-gui you can easily prepare your compiling environmen for building kernel-modules. It is self explainary and it will install all packages you need for the actual running kernel.

Hope this helps.


Best

Hans

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Am Samstag, 1. März 2025, 20:27:25 CET schrieb Van Snyder:

On Fri, 2025-02-28 at 22:00 -0800, Van Snyder wrote:

That's correct. You're probably missing the metapackage that brings
in new kernels automatically. For an amd64 machine, that metapackage
is named "linux-image-amd64". (If you use DKMS kernel modules,
you'll
also want the corresponding linux-headers-* metapackage.)

The NVidia 570 driver is a kernel package that's installed by running
an bash script that requires the kernel headers.

If I get the metapackages linux-image-amd64 and limux-headers-amd64,
will I need to rebuild the NVidia driver every time it loads a new
kernel?

Yes, when you get a new kernel-version, let`s say, 6.1.0-32-am64, then the nvidia kernel module has to be built.

So, after an upgrade and a new kernel, let the old kernel installed. When the nvidia build fails and X is not starting, you can reboot and start with the older kernel.

After the new kernel is starting well and X is loaded and the nvidia-module and drivers are working well, you can deinstall the older kernel.

With an upgrade the build of the nvidia-kernel-module should run automatically.

Hans

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

On Sat, 01 Mar 2025 11:27:25 -0800
Van Snyder <van.snyder@sbcglobal.net> wrote:

If I get the metapackages linux-image-amd64 and limux-headers-amd64,
will I need to rebuild the NVidia driver every time it loads a new
kernel?

Probably yes. If you install a fixed kernel version, you'll probably
need only the headers for exactly this kernel version. But an outdated
kernel version might have security issues.
--
kind regards
Frank

-----BEGIN PGP SIGNATURE-----

iQGzBAEBCgAdFiEE86z15c6qwvuAkhy+zDIN/uu9BloFAmfDaFkACgkQzDIN/uu9 Blq38Qv/RFpb3qFqpXrf3ml/hEK5v5J42xdTwKMaE71+mK3igJqyB6YA/Dxe5Rhb Ol3tL3084+3n68iDeeCMTjGPr8+mpgHZU0Du6NoTRJYXD/niU+UP8NFFjbRzdP1R cA+dJBVFRd3/9SwSez5Uz3Johgnb57naqXrXI0NHAY/fqoIOxcUiIME6e17WmgDx L8r4DQaZq6N9oRhRUaBsJj0yNac0BIFdldaW4HHtVdaCabW6tiMuEb33mHFFMYdE vizEBFMRu5UAAfvEyTks/4nXpX0r2MHujXX4fWESf5WPyWJRlUzr+JnJImMw/mhS FiII5yv4A+A9J85WiW/Lllw+fnz6V5r5aYX87RO2cpqcMfTvtsT0hHfOd0hZmZFH TAQpP+LAPJcUhCSaORU+dDlrjiMkrOl8WqDkBoq0dcTTI8t/2LcJCo6Ouk9jxRYD 9jzvRyXy8otkno29ZSSnJuy3/sJ9m/CS0WnTH2IHFQzb1J9jbHrdFcv5MKAlT+/5
6pmBTL+a
=cwap
-----END PGP SIGNATURE-----

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Hi,

On Sat, Mar 01, 2025 at 02:24:04PM -0800, Van Snyder wrote:

The NVidia kernel module is built by running a bash script. It's not a
.deb package.

Will it still be automatigically rebuilt?

If it's a DKMS, which is what my nvidia driver is, then it will try to
be built for any kernel install and should work as long as you have
headers installed. Though there have been times that things have changed
and its build is broken.

If you need a newer kernel to properly support your CPU or motherboard
then my gut feeling is that you'll need something newer than the 6.1.x
provided by Debian stable packages, as these tend to be a long term
stable kernel with only bug fix backported, not new features or device
support. So you may have to see what is in backports or experimental.

Thanks,
Andy

--
https://bitfolk.com/ -- No-nonsense VPS hosting

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Van Snyder <van.snyder@sbcglobal.net> writes:

I install the driver by running the NVIDIA-Linux-x86_64-570.124.04.run script at level 3, then rebooting.

Why?

Is that DKMS?

To be clear, it's a manual installation of drivers from the
manufacturer. Definitely not DKMS. You want install stuff manually, you
get to update it yourself.

Was there some reason to go with this instead of just installing the nvidia-driver package from non-free? Even the free nouveau driver seems
to have fairly good support for your GPU.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

The nvidia-driver package from non-free apparently doesn't work with a
Quadro K2200. NVidia recommends the 570 driver.

When I first installed the system, the left-hand pane of Evolution
would spontaneously scroll, even if a different window had keyboard and
mouse focus, and the mouse cursor disappeared when I put it into a
window title bar or boundary.

The nouveau driver would occasionally lock up.

Those problems were cured by installing the NVidia driver.

I believe, for the Quadro K2200 the kernel driver is the "nvidia-open-kernel" package. This is for the newest graphics card.

Did you try this one?

Best

Hans

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)

Van Snyder <van.snyder@sbcglobal.net> writes:

The nvidia-driver package from non-free apparently doesn't work with a Quadro K2200.

But you didn't check? The release notes tell a different story.

NVidia recommends the 570 driver.

They always recommend the latest.

--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)