You are not logged in.
Also what do you mean "worked by accident"? Do you think I never had enough space and it was being assigned anyway??
That would be the alternative to a bug, yes.
that first commit's archive
You're not using git but fetch zips from github?
Are you replacing/rebuilding the entire kernel or only drivers/pci?
The offending commit being in that path is plausible, but by no means guaranteed.
https://github.com/archlinux/linux/comm … .13-arch1/ has 4 commits on top of the 6.18.13 base, largely dealing w/ (irrelevant on your system) amdgpu.
You're always using https://gitlab.archlinux.org/archlinux/ … fig.x86_64 ?
Offline
You're not using git but fetch zips from github?
Yes, e.g. for the last build i set the source url to https://github.com/gregkh/linux/archive … 32d.tar.gz
I was rebuilding the kernel (although not cleanly, rerunning makepkg with existing cache on a clean docker instance), then installing the linux-lts and linux-lts-headers pkg.tar.zsts via pacman and rebuilding with nvidia-open-dkms
I can try building the whole thing from scratch from that commit again, alternatively I can try building 6.18.12 and maybe between .12 and .13 it got magically fixed again (since that pci commit is in .12 as well).
I use the linux 6.18.13 PKGBUILD as a base but I renamed the package to -lts since I want the package not to be named linux and mess up my current working kernel. I can try building 6.18.13 with the patches removed and see if that breaks it...
Offline
Oh. Nevermind, I thought I was building from the commit archive but I never changed the srcname...
Offline
This commit actually does work https://github.com/gregkh/linux/commit/ … 560cbdb770, I'll check the other ones tomorrow
Offline
Ok, so the breaking commit does in fact seem to be this one https://github.com/archlinux/linux/comm … f0d684b83d, I can try reverting it and trying to build a newer kernel version with that, not sure what else to do
Offline
Test the theory by only reverting the patch.
If you can demonstrate that it breaks your setup, you can (preferably) reply to the lkml patch, https://lore.kernel.org/all/20251219174 … tel.com/#R or contact the author directly (who will most likely take this to the lkml) and bring up your case.
I really can't say whether this is even potentially a bug or actually prevents you from undefined (but also any) behavior.
Offline