A system-wide LVM configuration change was made with F39. The vg* utilities now only use those devices that are listed in the devices file /etc/lvm/devices/system.devices. Up to F38, the vg* Utilities used all devices that were connected to the system at the time of the invocation. 

This change affects arm-image-installer if the script is used on a host that uses LVM. In any case, this always applies to Fedora Server. 

The script installs a bootable file system on an mSD adapter, which is temporarily connected to the host (and later to the ARM computer). Fedora Server on the ARM system also uses LVM and arm-image-installer needs to access the local VG and root LV for the final configuration and uses vgchange to temporarily mount the root filesystem and vgrename to resolve name conflicts. This worked smoothly until F38. Since F39 it no longer works, as the host with vgchange / vgrename only accesses the devices that are configured in its devices file. The temporary device with the mSD card is of course not included. 

For this reason, the script, which worked perfectly up to F38, has failed since F39.

For details about the change see https://bugzilla.redhat.com/show_bug.cgi?id=2258764

Reproducible: Always

Steps to Reproduce:
1. Install arm-image-installer on Fedora Server 39 or 40
2. Execute arm-image-installer to create a bootable file system for any device
3.
Actual Results:  
Script fails with 
No U-Boot files found for <YOUR_DEVICE>

Expected Results:  
U-Boot files installed

Available solutions

Previous attempts to find a solution (see https://bugzilla.redhat.com/show_bug.cgi?id=2246871 and https://bugzilla.redhat.com/show_bug.cgi?id=2247872#c17) considered deleting the devices file. Deleting it in the image file does not solve the problem. The host on which vg* is executed naturally looks in its own devices file, not in the one on a temporary device. And even if so, the LV is not accessible on the temporary device without vgchange, and vgchange is failing. And deleting this file on the host is of course taboo.

According to current knowledge, 2 options are actually available:

1. 
In arm-image-installer adding "--devices /dev/<mSD_ADAPTER>" to each vg* command. This should include the device in the respective vg* operation. That would be the best option.

2. 
Before the first execution of a vg* command execute "lvmdevices --adddev  /dev/<mSD_ADAPTER>". This would add the device to the host's devices file. Before finishing the script must get removed using "lvmdevices --deldev  /dev/<mSD_ADAPTER>". However, if the script terminates prematurely, the host configuration would be compromised. That would be a second-best solution at best.