Enabling persistence in Buster leads to issues at shutdown

Originally created by @CyrilBrulebois on #16312 (Redmine)

Seen with the first scenario of persistence.feature: once the test suite has been patched enough that persistence gets successfully enabled, the shutdown step contains many errors, even if it manages to stop the machine in the end.

• Screenshot 1: initial error messages popping up.

• Screenshot 2: example of systemd units going haywire before finally timing out.

• Screenshot 3: last display before black screen; many units are stopped between the second and third screenshots.

I’m reproducing (parts of) the initial & final messages below to make this ticket searchable:

print_req_error: I/O error, dev sr0
print_req_error: I/O error, dev loop0
SQUASHFS error: squashfs_read_data failed to read block
SQUASHFS error: Unable to read data cache entry
SQUASHFS error: Unable to read page, block
SQUASHFS error: Unable to read fragment cache entry

[…]

[FAILED] Failed unmounting /lib/live/mount/overlay.
[FAILED] Failed unmounting /lib/live/mount/medium.
[FAILED] Failed unmounting /lib/live/mount/rootfs/filesystem.squashfs.

[…]

Failed to unmount /lib/live/mount/overlay: Device or resource busy
Failed to unmount /lib/live/mount/rootfs/filesystem.squashfs: Device or resource busy
Failed to unmount /lib/live/mount/medium: Device or resource busy

Attachments

• tails-persistence-shutdown-errors-1.png
• tails-persistence-shutdown-errors-2.png
• tails-persistence-shutdown-errors-3.png

Related issues

• Blocked by #16097 (closed)
• Blocked by #16324 (closed)