Newer kernels fail to boot

rickj · Posted: Sat Apr 11, 2020 12:07 am Post subject: Newer kernels fail to boot

On a Pentium4 machine, using gentoo-sources, kernels up to 5.1.10 boot via grub with no problems. Kernels 5.5.9 and 5.6.2 fail to boot. The boot process stops without any error message at a random point between 0.5 and 1.5 seconds. I have tried compiling kernels with no new options, and with all security options disabled, without improvement. As far as I can see the only change is that the compiler is now gcc-9.2.0. On several other machines, both x86 and amd-64, I have had no trouble with updates.

Can anyone see what I am missing here?

Zucca · Posted: Sat Apr 11, 2020 2:09 pm Post subject: Re: Newer kernels fail to boot

DONAHUE · Posted: Sat Apr 11, 2020 3:25 pm Post subject:

when per linux kernel database:
https://cateee.net/lkddb/web-lkddb/RANDOM_TRUST_BOOTLOADER.html says
option found in kernel versions 4.19-4.20 then reappeared in 5.4 until present
https://cateee.net/lkddb/web-lkddb/RANDOM_TRUST_BOOTLOADER.html says
option found in kernel version 5.4 until present

Theory looking good.
_________________
Defund the FCC.

rickj · Posted: Sat Apr 11, 2020 4:25 pm Post subject:

Unfortunately, it seems that no amount of key pressing during boot prevents the stall.

This happens well before init has started. At present the last lines on the screen are:

rickj · Posted: Sat Apr 11, 2020 4:56 pm Post subject:

With both CONFIG_RANDOM_TRUST_CPU and CONFIG_RANDOM_TRUST_BOOTLOADER set, the situation is unchanged. No amount of key pressing seems to help before or after setting these options.

rickj · Posted: Sat Apr 11, 2020 5:11 pm Post subject:

I have installed haveged, but it does not help. It is hard to see how it can be started so early as 0.5 seconds into the boot.

rickj · Posted: Fri Apr 17, 2020 3:59 pm Post subject:

On the basis of "if in doubt, thrash about" I tried removing all modules and module support from the kernel. This had absolutely no effect.
Memtest 86+ runs 20 passes on the system with no errors.

Zucca · Posted: Fri Apr 17, 2020 7:14 pm Post subject:

I didn't quite get rid of the booting problems, so I downgraded to latest stable gentoo-sources, which is 5.4.28.
It has been working ever since without problems to boot or any crashes.
One thing that did stopped working, however, was sway's screensaver, swayidle. I doubt this is any kernel related problem...

Have you tried 5.4 -series kernel or earlier?

Also you might want to add debug to the kernel command line in hopes to get more detailed error messages.
_________________
..: Zucca :..

rickj · Posted: Sat Apr 18, 2020 2:41 am Post subject:

I have now tried the 5.4.28 kernel, and it fails to boot. The failure is later, at a little over 1.5s:

rickj · Posted: Sat Apr 18, 2020 6:42 pm Post subject:

I have also tried the 5.4.28 kernel, compiled with gcc-8.2.0. It still does not boot, although it fails in a different place.

rickj · Posted: Thu Apr 30, 2020 2:37 am Post subject:

In the hope that the trouble was in the boot manager, grub, I just installed and configured Lilo.

This has no effect, booting fails in the same way, so the problem is somehow in the kernel.

fturco · Veteran Joined: 08 Dec 2010 Posts: 1181

@rickj: You may try to do a kernel git bisect: https://wiki.gentoo.org/wiki/Kernel_git-bisect
You just need to know the last good kernel version and the first bad one, and use the same config file.