On quattro (reproduced at 268bac1), pulling the install medium once the installer says it is done — at the "Installed Omarchy in …" / Reboot Now screen — leaves the machine unable to reboot, scrolling kernel errors until it is power-cycled.
Reproduce
- Boot the ISO and run a normal full-disk install to the finish screen. (QEMU: ISO as an ejectable cdrom device; real hardware: a USB stick.)
- At the Reboot Now screen, remove the medium (QMP
eject / pull the stick). Nothing complains — the screen sits there looking finished.
- Press Reboot Now.
Observed
The graceful reboot's stop jobs immediately fault pages from the medium that is no longer there:
I/O error, dev sr0, sector 2501728 op 0x0:(READ) ...
I/O error, dev loop0, sector 1972766 op 0x0:(READ) ...
SQUASHFS error: Failed to read block 0x3c343fc1: -5
SQUASHFS error: Failed to read block 0x3c343fc1: -5
...
The last line repeats ~100 times a second, indefinitely — after 2.5 minutes the kernel was still failing on the same block. The machine never reboots; the only exit is the power button.
It is worse than a scary screen, because of what is still held at that moment. At the finish screen the target is not released:
/mnt /dev/mapper/root[/@] btrfs rw,...
/mnt/home /dev/mapper/root[/@home] btrfs rw,...
/mnt/boot /dev/vda1 vfat rw,...
/mnt/swap/swapfile file 4000508 ... (ACTIVE swap)
/dev/mapper/root (LUKS mapper open)
So the forced power-off this ends in also takes the freshly installed filesystem down rw-mounted with live swap on it.
Why
The live root pages from the install medium on demand (copytoram=n), and a graceful reboot is the worst case for that: its stop jobs page in unit files and ExecStop binaries. The same failure needs no button press on real hardware — during PR #113's A/B testing on a ThinkPad X200, yanking the stick at this screen segfaulted the live system the same way.
Nothing tells the user the medium is still needed: the natural reading of "Installed Omarchy in 0m 55s" is that it is safe to take the stick.
On
quattro(reproduced at268bac1), pulling the install medium once the installer says it is done — at the "Installed Omarchy in …" / Reboot Now screen — leaves the machine unable to reboot, scrolling kernel errors until it is power-cycled.Reproduce
eject/ pull the stick). Nothing complains — the screen sits there looking finished.Observed
The graceful reboot's stop jobs immediately fault pages from the medium that is no longer there:
The last line repeats ~100 times a second, indefinitely — after 2.5 minutes the kernel was still failing on the same block. The machine never reboots; the only exit is the power button.
It is worse than a scary screen, because of what is still held at that moment. At the finish screen the target is not released:
So the forced power-off this ends in also takes the freshly installed filesystem down rw-mounted with live swap on it.
Why
The live root pages from the install medium on demand (
copytoram=n), and a graceful reboot is the worst case for that: its stop jobs page in unit files and ExecStop binaries. The same failure needs no button press on real hardware — during PR #113's A/B testing on a ThinkPad X200, yanking the stick at this screen segfaulted the live system the same way.Nothing tells the user the medium is still needed: the natural reading of "Installed Omarchy in 0m 55s" is that it is safe to take the stick.