Early boot freeze on newer kernels

mavrix · 2024-07-06 10:16:25

I'm trying to sort out a problem I've had since end of version 6.6.9 which is the last one that allows me to boot on my Dell XPS 13. After that, I still can select the systemd-boot loader entry but then the computer freezes with a black screen, where not even sysrq works anymore. The same change was apparently introduced in the lts versions between 6.6.17-1-lts and 6.6.18-1-lts where the former still works fine, whereas the latter also freezes. Newer versions work fine on my other computers with similar, but not exactly same setup: btrfs, luks, lvm... (The main difference is that the other computers have a plymouth bootsplash).
Any ideas how to proceed with troubleshooting or what the changes may have been, that were introduced between the above mentioned versions? Thanks.

cryptearth · 2024-07-06 11:15:43

just out of the blue: maybe change to bootloader from systemd to some other and remove the "quite" option from the kernel command line
by this I figured out a "hang" which was caused by a long-timeout due to a messed up fstab
maybe you can get some additional output

gromit · 2024-07-06 12:52:16

Were you able to verify that downgrading the kernel package actually helps with the issue?

Also this sounds like a possible regression, which would need to be bisected and reported upstream to the linux kernel developers

If you want I can help you doing this by supplying you prebuilt packages for each version to test!

Last edited by gromit (2024-07-06 12:52:28)

mavrix · 2024-07-06 14:11:07

Downgrading fixes the issue, yes. I can flipflop between the two versions - one works, the other one won't. Some prebuilt packages would be much appreciated; would likely take ages on my ageing laptop otherwise.

gromit · 2024-07-06 15:07:53

Is there a specific reason you are using the linux-lts package over the linux package? Could you see if the issue also is present in the latest mainline release?

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-mainline-6.10rc6-1-x86_64.pkg.tar.zst

I'll supply you the first images once I got my setup for the lts kernels, now I gotta go for a birthday party

gromit · 2024-07-06 19:17:29

Please test the first package:

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-lts-v6.6.17.r169.gb476ae1-1-x86_64.pkg.tar.zst

Apparently we have around 7 images to test:

$ git bisect bad v6.6.18
status: waiting for good commit(s), bad commit known
$ git bisect good v6.6.17
Bisecting: 169 revisions left to test after this (roughly 7 steps)
[b476ae1da1369bc56e090d0a55bbd55bf0d8d409] drm/amdgpu/soc21: update VCN 4 max HEVC encoding resolution

mavrix · 2024-07-06 19:51:56

That was a short birthday party

I do primarily use the linux package; that's the one I was referring to in my first sentence about 6.6.9 being the last working version. I've tried the lazy approach so far, adding linux to IgnorePkg and waiting to see if things would sort themselves out over time - as they have for other issues in the past . Whenever I got around to it, I then tried to corner the problem (e.g. doing the usual searches online, trying the zen kernel - which also doesn't work - or fiddling with the kernel boot parameters etc) with not much to show for all the work, unfortunately. linux-lts is usually my fallback but ended up not working anymore as well.

Trying your first package now.

mavrix · 2024-07-06 20:00:18

So far so good.

$ uname -r
6.6.17-1-lts-00169-gb476ae1da136-dirty

mavrix · 2024-07-06 20:08:29

Mainline freezes as well btw.

gromit · 2024-07-06 21:03:08

Please try

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-lts-v6.6.17.r254.gb93a6756-1-x86_64.pkg.tar.zst

mavrix · 2024-07-06 21:19:25

Still fine.

$ uname -r
6.6.17-1-lts-00254-gb93a6756a01f-dirty

gromit · 2024-07-07 09:27:25

Please try

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-lts-v6.6.17.r296.gf3f41f4-1-x86_64.pkg.tar.zst

mavrix · 2024-07-07 15:29:51

Hmm. Still works.

gromit · 2024-07-07 15:45:54

Hm, thats a bit weird Anyways lets keep on testing and if we test good all the way through well need to put some more thought into this

Please test the following:

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-lts-v6.6.17.r317.ge17a8bf-1-x86_64.pkg.tar.zst

mavrix · 2024-07-07 16:06:51

Ok. That was the first bad one now.

gromit · 2024-07-07 17:07:35

Please test the following:

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-lts-v6.6.17.r306.g21faa3d-1-x86_64.pkg.tar.zst

mavrix · 2024-07-07 17:20:32

That one worked again.

gromit · 2024-07-07 17:42:54

Please test the following:

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-lts-v6.6.17.r311.ged823ca-1-x86_64.pkg.tar.zst

mavrix · 2024-07-07 17:49:23

Still good. Should be close now.

gromit · 2024-07-07 18:02:34

Please test the following:

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-lts-v6.6.17.r314.g940c919-1-x86_64.pkg.tar.zst

mavrix · 2024-07-07 18:16:06

Still good, too. Must be the commit after that then, I suppose? Any information on what it is?

$ uname -r
6.6.17-1-lts-00314-g940c919be813-dirty

loqs · 2024-07-07 18:23:29

https://git.kernel.org/pub/scm/linux/ke … 7c1f4232d0 or https://git.kernel.org/pub/scm/linux/ke … 8b6171f5fe or https://git.kernel.org/pub/scm/linux/ke … d04297758d

mavrix · 2024-07-07 18:36:00

All seem to have something to do with the EFI boot process and mostly refer to "legacy" and "no longer needed". I wonder if there's anything that has changed over the years, that just didn't get updated on my machine. I originally installed Arch maybe 5-10 years ago and it has pretty much always been running fine with the regular sync/updates. Think systemd-boot was actually still gummiboot when I installed it.

gromit · 2024-07-07 18:39:03

Please test the following:

sudo pacman -U https://pkgbuild.com/\~gromit/linux-bisection-kernels/linux-lts-v6.6.17.r316.g8117961-1-x86_64.pkg.tar.zst

The remaining commits are as @loqs said:

$ git bisect visualize --oneline
e17a8bf1fb5d6 x86/efi: Drop alignment flags from PE section headers
8117961d98fb2 (HEAD) x86/efi: Disregard setup header of loaded image
fa244085025f4 x86/efi: Drop EFI stub .bss from .data section

mavrix · 2024-07-07 18:44:26

That's a bad one again.

Arch Linux

#1 2024-07-06 10:16:25

Early boot freeze on newer kernels

#2 2024-07-06 11:15:43

Re: Early boot freeze on newer kernels

#3 2024-07-06 12:52:16

Re: Early boot freeze on newer kernels

#4 2024-07-06 14:11:07

Re: Early boot freeze on newer kernels

#5 2024-07-06 15:07:53

Re: Early boot freeze on newer kernels

#6 2024-07-06 19:17:29

Re: Early boot freeze on newer kernels

#7 2024-07-06 19:51:56

Re: Early boot freeze on newer kernels

#8 2024-07-06 20:00:18

Re: Early boot freeze on newer kernels

#9 2024-07-06 20:08:29

Re: Early boot freeze on newer kernels

#10 2024-07-06 21:03:08

Re: Early boot freeze on newer kernels

#11 2024-07-06 21:19:25

Re: Early boot freeze on newer kernels

#12 2024-07-07 09:27:25

Re: Early boot freeze on newer kernels

#13 2024-07-07 15:29:51

Re: Early boot freeze on newer kernels

#14 2024-07-07 15:45:54

Re: Early boot freeze on newer kernels

#15 2024-07-07 16:06:51

Re: Early boot freeze on newer kernels

#16 2024-07-07 17:07:35

Re: Early boot freeze on newer kernels

#17 2024-07-07 17:20:32

Re: Early boot freeze on newer kernels

#18 2024-07-07 17:42:54

Re: Early boot freeze on newer kernels

#19 2024-07-07 17:49:23

Re: Early boot freeze on newer kernels

#20 2024-07-07 18:02:34

Re: Early boot freeze on newer kernels

#21 2024-07-07 18:16:06

Re: Early boot freeze on newer kernels

#22 2024-07-07 18:23:29

Re: Early boot freeze on newer kernels

#23 2024-07-07 18:36:00

Re: Early boot freeze on newer kernels

#24 2024-07-07 18:39:03

Re: Early boot freeze on newer kernels

#25 2024-07-07 18:44:26

Re: Early boot freeze on newer kernels

Board footer