On Tuesday, 26 October 2021 4:01:08 PM ACDT Andrei Borzenkov wrote:
On 26.10.2021 06:14, Bernhard M. Wiedemann wrote:
On 26/10/2021 02.54, Larry Finger wrote:
With a recent Tumbleweed snapshot (probably 20211023), the utility installkernel was removed from /usr/sbin. Without this code, a kernel 'make install' returns "Cannot find LILO.". I could not tell if I was in a decades-long time warp, or if Linda Walsh had gained control. Copying
This is default fallback in arch/x86/boot/install.sh if installkernel is not present.
the one from my Leap 15.3 partition restored the functionality.
In Leap 15.3, this file is in the dracut package, but definitey missing in Tumbleweed.
https://code.opensuse.org/package/dracut/c/c7b5d678046f286e9c963d05e7706dd be05b5a31?branch=master
has
remove /sbin/installkernel script from dracut package
That came in via https://build.opensuse.org/request/show/923450 but unfortunately, not with a good git-style commit message that explains the "Why?"
https://github.com/openSUSE/dracut/commit/8c93cf30948522308bfd7bd02aa882eb09 e79609
revert: remove /sbin/installkernel script from dracut package Nowadays, /sbin/installkernel is only used by developers who compile their own kernels on the target system. It doesn't need to be part of dracut any more. This commit partially reverts dc98785 ("Add suse specific files").
BTW subject is certainly misleading. "Revert: remove /sbin/installkernel" implies that this commit reverts *removal* of /sbin/installkernel thus restoring this file.
I wonder if a symlink of installkernel to /usr/bin/true would also work.
It will shut up the warning, but it will not make new kernel available which is the purpose of this script. Removed script generated initrd and called update-bootloader, making new kernel available for booting.
This would explain why I ended up with a non-bootable system after re-building my kernel to include an out-of-tree I2C driver that's for some reason not yet included in the stock kernel (required for OpenRGB to talk to NCT6775 I2C RGB controllers on some motherboards). Building kernel 5.14.11-2 is the first time I've ever had a failure of 'make install", and now I know why. I had to manually copy vmlinuz-5.14.11-2-default to vmlinuz-5.14.11-2-default.old (along with its associated system.map file), rename 'vmlinuz' (apparently localversion was ignored during the build) to vmlinuz-5.14.11-2-default and create a symlink for vmlinuz pointing to that kernel image. This was all previously done by 'make install'. -- ================================================================================================================== Rodney Baker rodney.baker@iinet.net.au ==================================================================================================================