Arch Linux User Repository

Then it all looks good. I have no idea what your bootup needs more then. Everything looks good from my end. Perhaps force remove, reboot to other kernel, check mentioned dirs (boot, modules) if they dont contain any old znver3 files, clear pacman cached znver3 kernel, and then reinstall the kernel from the repo? because I am out of ideas for you..

Yes eggz, there are.

[mario@asus-arch ~]$ ls -lh /usr/lib/modules/6.6.9-AMD-znver3/ total 16M drwxr-xr-x 1 root root 324 ene 1 09:40 build drwxr-xr-x 1 root root 110 ene 1 09:40 kernel -rw-r--r-- 1 root root 1,1M ene 1 09:41 modules.alias -rw-r--r-- 1 root root 1,1M ene 1 09:41 modules.alias.bin -rw-r--r-- 1 root root 5,4K ene 1 08:54 modules.builtin -rw-r--r-- 1 root root 8,5K ene 1 09:41 modules.builtin.alias.bin -rw-r--r-- 1 root root 7,4K ene 1 09:41 modules.builtin.bin -rw-r--r-- 1 root root 45K ene 1 08:54 modules.builtin.modinfo -rw-r--r-- 1 root root 524K ene 1 09:41 modules.dep -rw-r--r-- 1 root root 678K ene 1 09:41 modules.dep.bin -rw-r--r-- 1 root root 453 ene 1 09:41 modules.devname -rw-r--r-- 1 root root 139K ene 1 08:54 modules.order -rw-r--r-- 1 root root 1,1K ene 1 09:41 modules.softdep -rw-r--r-- 1 root root 595K ene 1 09:41 modules.symbols -rw-r--r-- 1 root root 719K ene 1 09:41 modules.symbols.bin -rw-r--r-- 1 root root 17 ene 1 08:54 pkgbase drwxr-xr-x 1 root root 8 ene 1 09:40 updates -rw-r--r-- 1 root root 11M ene 1 08:54 vmlinuz [mario@asus-arch ~]$ date lun 01 ene 2024 13:47:06 -05 [mario@asus-arch ~]$

Hey Mario, I did the same tests on this kernel in the mean time and I also couldn't find anything wrong too

do you have a folder /usr/lib/modules/6.6.9-AMD-znver3/ when the kernel is installed?

Ok, my problem happend with this version. :p

Oh yeah my PKGBUILDS (I reused) def go back more than 4 years. Okay, I did not know that, good to know. I will use the simplified way soon after some more testing.

As for the diffrences, im not completely convinced everything in the core kernel is automaticly better. If I wanted to make a 1:1 copy, well, im not sure what my added value is here and you might aswell just start using the core kernel with my Kconfig. But I'll take a look at them anyway.

I use git instead of tarbals because my pipelines(build,test,aur,repo,...) are custom made and I would like to be able to uniformly (and transparantly) insert gitpatches/reverts inside them when I want to make modifications/patches in the kernel code. While using tarbals in the pipelines by some extra mechanisms is not impossible (by using some git overlay or something?), it gives me alot more work to redo big parts of the pipelines and I currently dont have the time for that, I'm sorry. If I wanted to run by you every detail of my pipeline and the pros and cons, we would be here discussing for days. Lets just say it's just a choice I have made and so far im sticking by it. If you want a fast install, use the repo, if you want all the details, use the aur git build way :-)

I should thank you as well for your effors, nice work and returning my feedbacks quickly.

Looks like they added those pacman hooks 4 years ago according to the git history of core/mkinitcpio. That's might be the reason too.

And I'd like to point out a few more things.

I noticed some dependency differences between your and core/linux's PKGBUILD files as you can see here, here, here and here. You might want to take a look at those too.

And may I ask is there any reason for using git instead of tarballs from Linux's CDN? It would be nice to check its hash values too.

thx

Okay, I will take it under advisement. I believe I implemented it in the very beginning because pacman/mkinicpio was not implementing modules as expected (Or maybe it was a false positive/other problem afterall?). I will now do some testing on my test kernels to see if I can ommit al these manual steps -- to be sure that it does not bring up any other problems -- before implementing them directly. Thank you for telling me this onurmercury.

It simply detects the changes under /usr/lib/modules/*/vmlinuz

mkinitcpio has the necessary hooks (https://gitlab.archlinux.org/archlinux/mkinitcpio/mkinitcpio/-/blob/master/libalpm/hooks/90-mkinitcpio-install.hook)

And it is explained here: https://wiki.archlinux.org/title/mkinitcpio#Automated_generation and according to this article, you don't need to provide .preset file either.

dracut doesn't have necessary hooks built-in but it's automation is explained too: https://wiki.archlinux.org/title/Dracut#Generate_a_new_initramfs_on_kernel_upgrade.

(btw kernel-install automation for dracut is available too https://aur.archlinux.org/packages/kernel-install-for-dracut. (both aur packages maintained by endeavouros devs))

I have tried several kernel packages from various repositories, including AUR, and I have only saw such an approach in your packages.

yeah as of now I dont really support systemd-boot, sorry. Interesting, I did not know pacman builds the ramdisk and stuff for you. What are the conditions/triggers there? How does it know the package is a kernel? Well for now I rather include my own trigger scripts to be sure and have more control.

hi, im using systemd-boot as my boot manager. i can automatically generate initrds with kernel-install feature (and a bit help of https://aur.archlinux.org/packages/kernel-install-mkinitcpio) of systemd.

right now mkinitcpio runs twice because of entries in the .install file. can you implement some kind of check to prevent this happening? https://paste.debian.net/1302385/

(and because of it, it also generates unnecessary vmlinuz and initrd files under /boot, kernel-install already generates those under ESP(/efi, /boot, or /boot/efi)) https://paste.debian.net/1302387/

and may i ask why are we running depmod and mkinitcpio in .install file? pls correct me if im wrong but afaik pacman can take care of it (for ex. like the default linux kernel).