Appears to yes , i have 3 machines with some ryzen cpus and only fail 1 that is intel.

2 boot fine (amd ryzne s) 1 laptop 1 tower.

1 fail to boot ( intel i7 7500u ) laptop.

1 members found this post helpful.

I am also using elilo.

1 members found this post helpful.

The latest Nvidia production driver-550.78, builds with the 6.9.0 kernel. I have installed it twice. Once with a "roll-your-own" kernel, then removed it, the kernel and the driver, and installed the driver again with the kernel provided in yesterday's -current updates.
Not a cough, hiccup or a sneeze. Everything is working as it should.

1 members found this post helpful.

Both of my Intel machines (DualCore & i7), boot just fine to the 6.9 kernel
and everything I have run at the cli works just fine. BUT, X will not start.

Right now underway here on this Intel DualCore machine is a compile of the 6.9.0 kernel
using using /boot/config-generic-6.6.30.x64 as a staring point.

slackpkg install k/*

root@glennmcc-server:/# slackpkg search kernel

Looking for kernel in package list. Please wait... DONE

The list below shows all packages with name matching "kernel".

[ upgrade ] - kernel-firmware-20240502_28f75f1-noarch-1 --> kernel-firmware-20240510_b9d2bf2-noarch-1
[ upgrade ] - kernel-generic-6.6.30-x86_64-1 --> kernel-generic-6.9.0-x86_64-1
[ upgrade ] - kernel-headers-6.6.30-x86-1 --> kernel-headers-6.9.0-x86-1
[ upgrade ] - kernel-modules-6.6.30-x86_64-1 --> kernel-modules-6.9.0-x86_64-1
[ installed ] - kernel-source-6.9.0-noarch-1


cp /boot/config-generic-6.6.30.x64 /usr/src/linux-6.9/.config

cd /usr/src/linux-6.9

make menuconfig

Hit 'save' and 'exit' without making any changes.

time make

Will report back later with my results.

Oh yeah, other than the kernel... this system is a fully upgraded slackware64-current
therefore this kernel is being built with gcc14

While this looks a lot like what's being reported here, that patch made it into gcc-14.1.0. But there could possibly be more issues like this lurking in gcc-14.

As is usually the case, everything is working great here. I'm getting no crashes or illegal instructions on any of my machines, ranging from old i3/i5/i7 (an older i7 than was reported as problematic here), and even the venerable Phenom II X6 that completely lacks AVX support. I've tried xfce4-terminal, mc, okular, and whatever else had crash reports.

I might take a day off from compiling anything else (i.e. quit digging that hole deeper) and see what I can find. Hints appreciated, of course.

The kernel is another question, but that's also considerably easier to back out.

2 members found this post helpful.

It seems to be code reading the hibernation image from disk. I have a swap file, not a partition, so maybe that's why everyone doesn't see this. When I now removed resume= and resume_offset= parameters from the kernel command line, even my laptop boots Pat's 6.9.0 kernel.

3 members found this post helpful.

I can try a simply test

Build 6.6.30 with gcc14 and test on my problematic machine.

If run , then the problem comes in linux 6.9
If not run , the problem comes in gcc14

Building linux 6.6.30 ... i need only arround 7 minutes to make test.

2 members found this post helpful.

GDS - Gather Data Sampling
Depends on (Intel) CPU, microcode, kernel version, boot parameters.

https://www.intel.com/content/www/us...-sampling.html
https://docs.kernel.org/admin-guide/..._sampling.html

lscpu :
Vulnerabilities:
Gather data sampling*: Vulnerable: No microcode
or "Not affected" or "Mitigation: microcode" or "Mitigation: No microcode; AVX disabled" (or something like that)

Some Intel CPUs have no mitigation by microcode (even with last version 2024-05-14), disabling AVX causes "illegal instruction" with mc.

Kernel 6.6 does not disable AVX by default if no microcode mitigation, use append="clearcpuid=avx" in lilo to disable AVX
Kernel 6.9 disables AVX by default if no microcode mitigation, use append="gather_data_sampling=off" in lilo to keep AVX instructions

JFB

9 members found this post helpful.

Hi , i finish my little test.

GCC 14 IS NOT THE CULPRIT

Is something with linux 6.9 , or bug on code , or a config issue.


Building linux 6.6.30 with gcc 14 , and runs fine on my problematic machine ...
Linux 6.9 not run on my machine.

I can do , other test , i have 15.0 with old gcc 11 if im not remember bad.

I can test build 6.9 on slackware 15.0 and see , but i imagine if is a bug ..the compiler is not relevant.

4 members found this post helpful.

Ah yes, now we're getting some problems here! I have all my machines running with mitigations=off since I completely lack paranoia over side channel attacks.

Seems like a poor kernel default. /me goes looking to see if there's a Kconfig toggle...

3 members found this post helpful.

Pff , im not sure now


slackware64-15.0

i build linux 6.9 , using gcc 11 , and problematic machine start okey.

same machine , cant start -current with linux 6.9 , its very rare.

firmwares on 15.0 are more old ...

Nothing clear.

EDIT: Downgrade firmwares no fix the issue. , sorry but i not get nothing clear here.

1 members found this post helpful.

Here is a snippet of code. I was trying to build the nvidia driver with kernel-6.9.0.

It builds find with 6.6.25 and gcc.

CC [M] /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-p2p.o
cc: error: unrecognized command-line option '-fmin-function-alignment=16'; did you mean '-flimit-function-alignment'?
CC [M] /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-pat.o
make[3]: *** [scripts/Makefile.build:244: /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv.o] Error 1
CC [M] /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-procfs.o
CC [M] /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-usermap.o
cc: error: unrecognized command-line option '-fmin-function-alignment=16'; did you mean '-flimit-function-alignment'?
cc: error: unrecognized command-line option '-fmin-function-alignment=16'; did you mean '-flimit-function-alignment'?
make[3]: *** [scripts/Makefile.build:244: /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-pci.o] Error 1
CC [M] /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-vm.o
cc: error: unrecognized command-line option '-fmin-function-alignment=16'; did you mean '-flimit-function-alignment'?
make[3]: *** [scripts/Makefile.build:244: /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-acpi.o] Error 1
cc: error: unrecognized command-line option '-fmin-function-alignment=16'; did you mean '-flimit-function-alignment'?
make[3]: *** [scripts/Makefile.build:244: /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-nano-timer.o] Error 1
cc: error: unrecognized command-line option '-fmin-function-alignment=16'; did you mean '-flimit-function-alignment'?
CC [M] /tmp/selfgz1716/NVIDIA-Linux-x86_64-550.67/kernel/nvidia/nv-vtophys.o
cc: error: unrecognized command-line option '-fmin-function-alignment=16'; did you mean '-flimit-function-alignment'?

1 members found this post helpful.

FWIW I have built a 6.9.0 kernel with gcc 14 and gcc 13.2. Both cause crashes on my -current laptop, which has an intel cpu.

I'll take a look at what @billaud suggests with mitigations too.

1 members found this post helpful.

Yep that fixes it. I added "mitigations=off" to the kernel command line et voila, sddm and plasma work again.

Thanks again for the hint. I don't tinker with the kernel congfis much but I'll take a peek. Hopefully thats enough for Pat to narrow in on a fix.

4 members found this post helpful.

I may have spoken too soon. I downgraded gcc to 13.2 and then rebuilt kernel 6.9 under that gcc, and got the exactly same problems. SDDM not booting, applications randomly crashing, "Illegal instructions" errors... so it should be the kernel or somehting else. Reverted to kernel 6.6.30 and gcc 13.2 for now. I will investigate further.

1 members found this post helpful.