UEFI installation always fail

[bobblkabb]

I have already worked through the instructions on 3 computers, point by point and always the encrypted version.
None of the installations worked.
At the point refind-install (I had chosen the pre-build zfsbootmenu variant) I got efi variable error messages. I then ran efivar-tester which claimed that the device was not UEFI capable.
The first two laptops have a CSM option in the bios, so I assumed that they are UEFI capable. And the third laptop (Chromebook Acer C720 with Coreboot) there is even a UEFI shell.
In this respect, I must have made a mistake when installing according to instructions. What could I have done wrong.

/sys/firmware/efi/efivars/ is empty

[Dragon-debian]

If /sys/firmware/efi/efivars/ is empty, it means that efivarfs pseudo filesystem is not mounted there.The init system usually mounts this automatically.

Try to mount it manually and see what happens.

mount -t efivarfs efivarfs /sys/firmware/efi/efivars

you can replace second efivarfs with any string you want. its only a name.

If mount said unknown filesystem type 'efivarfs', the laptop most likely is booted in legacy mode.
check the uefi settings and make sure its booting in uefi mode, not legacy support/mode or csm enabled.

There are some cases that system is booted in uefi mode but you cant access EFI vars:

1- kernel does not support EFI. void linux kernel (and most other distros) has support for this. if you are using a custom kernel, check that kernel compiled with CONFIG_EFI=y.

2- you are using a real time kernel. efivarfs is disabled on those kernels by default. boot the kernel with efi=runtime kernel command line and try again.

3- efivarfs kernel module is not loaded. void linux kernel has this module builtin, so there is no need to run modprobe efivarfs if you are using void linux kernel.

4- kernel is booted with noefi kernel command line. if this is true, reboot and remove that from kernel command line.

[bobblkabb]

Yes, the scenario about the unknown filesystem happened.
I did another installation without UEFI because I wasn't sure about the UEFI capability of the laptop but it miserably failed.
So I installed once again - with UEFI. All your talking about the linux kernel reminded me that I probably took the latest kernel (5.15) not the kernel as instructed. Thought it was outdated. Now I get a proper UEFI Boot menu and I can unlock the hard disk but still having problems to load the system:

Last accessed by <unknown> (hostid=0) at Wed Apr 27 16:23:30 2022
The pool can be imported, use 'zpool import -f' to import the pool.
dracut Warning: ZFS: Unable to import pool zroot
cannot import 'zroot': pool was previously in use from another system.

dracut Warning: Can't mount root filesystem

[bobblkabb]

Changed my password like described here, I removed a special character:

1. Booted with hrmpf-Stick and chrooted into the system
2. echo 'my new password' > /etc/zfs/zroot.key
3. zfs change-key zroot
4. xbps-reconfigure -f linux5.10

Now I get following error during boot (after password prompt):
Wrong password
zfs_mount_at() failed: encryption key not loadedEnrcypted ZFS password for zroot: Key load error: Failed to open key material file: No such file or directory

[Dragon-debian]

Is key file avaiable in initramfs?
dracut should executes an emergency shell if it cant mount the root file system.
In that shell, check that /etc/zfs/zroot.key is available or not.
If its avaiable, check its content.

cat /etc/zfs/zroot.key

If the key file is ok, maybe zpool.cache is causing problem.
Boot ZFSBootMenu and boot into boot menu.
press MOD + w to reimport all the pools rw.
MOD can be ctrl, altor ctrl + alt.

Jump into your boot environment with MOD + j.
Remove zpool.cache and create a new one.

rm /etc/zfs/zpool.cache
zpool set cachefile=/etc/zfs/zpool.cache zroot

EDIT: I added --kver.
Now, regenerate initramfs.

dracut -f --kver 5.15.34_1

Replace 5.15.34_1 with right kernel version. look in /boot to find the right one.

You can also reconfigure the kernel package to generate a new initramfs.

xbps-reconfigure -f linux5.15

Replace linux5.15 with kernel version you have installed.

Both of the above commands can be used to regenerate initramfs in void linux. I dont know their difference.

[bobblkabb]

Deleting and setting works but fails at 'dracut -f':

dracut: Cannot find module directory /lib/modules/5.10.101_1/
dracut: and --no-kernel was not specified 

/lib/modules/ shows following folders:
5.10.111_1
5.15.34_1

So I tried 'dracut -f -kver /lib/modules/5.10.111_1':

dracut: -k/--kmoddir path must contain "lib/modules" as a parent of your kernel module directory,
dracut: or modules may not be placed int the correct location inside the initramfs.
dracut: was given: ver
dracut: expected: ./lib/modules/5.10.101_1
dracut: Please move your modules into the correct directory structure and pass the new location,
dracut: or set DRACUT_KMODDIR_OVERRIDE=1 to ignore the check.

So I copied 5.10.111_1 to 5.10.101_1 and tried it again with the same message as a result.

At last I succeeded with 'dracut -f --no-kernel', followed by 'xbps-reconfigure -f linux5.10'.
'...5.10.101_1 completed successful'-like message at end of the operation.

Now I get following error:

cannot import 'zroot': pool was previously in use from another system.
Last accessed by <unknown> (hostid=0)
The pool can be imported, use 'zpool import -f' to import the pool.
dracut Warning: ZFS: Unable to import pool zroot

I end up in the debug shell.

[Dragon-debian]

I forgot something. If you dont specify --kverfor dracut, it guesses the right version from uname -r output.
The kernel version in ZFSBootMenu and your boot environment wasnt the same, so dracut failed to detect the kernel version and didnt generate an initramfs.

remove the copy you created, /lib/modules/ 5.10.101_1.

The right command that you had to execute was this:

dracut -f --kver 5.10.111_1

I think the xbps-reconfigure way is better as it passes the right kernel version to dracut and works fine in chroot.

[ahesford]

You don't have a consistent /etc/hostid in your ZBM and regular initramfs images. Either fix that or boot ZBM with the command-line options

zbm.import_policy=hostid zbm.set_hostid

[bobblkabb]

I deleted the /lib/modules/5.10.101_1 folder again.

Then I entered chroot in read/write mode:

rm /etc/zfs/zpool.cache
zpool set cachefile=/etc/zfs/zpool.cache zroot
dracut -f --kver 5.10.111_1

It finishes with:

..
Creating image file '/boot/initramfs-5.10.111_1.img'
Creating initramfs image file '/boot/initramfs-5.10.111_1.img'

Now when I press [CTRL+K ] (kernels) and switch from 5.15.34_1 to 5.10.111_1 it works!
How do I make the 5.10.111_1 as default? Or how do I make 5.15.34_1 work?

[ahesford]

Either use the menu to set the default or read the zfsbootmenu.7 man page.

[bobblkabb]

Ha! Tricked the system. Just 'dracuted' the latest kernel and it works. Good a functional system.
Thank you very much for your support! Very appreciated.