]> git.ipfire.org Git - thirdparty/systemd.git/commit - units/systemd-tmpfiles-clean.service
units: make sure systemd-tmpfiles-{setup,clean} don't survive switch-root
authorZbigniew Jędrzejewski-Szmek <zbyszek@in.waw.pl>
Thu, 5 Aug 2021 11:22:58 +0000 (13:22 +0200)
committerLuca Boccassi <luca.boccassi@gmail.com>
Fri, 6 Aug 2021 10:11:14 +0000 (11:11 +0100)
commit2525682565b372b9b83c848bfe89c025fed47a1d
treec3851d2e95a62fff3fc5dd8e068205d7b4e8533b
parent32a2ee2bb4fa265577c883403748c909cd6784dd
units: make sure systemd-tmpfiles-{setup,clean} don't survive switch-root

Normally, these services are killed because we run isolate. But I booted into
emergency mode (because of a futher bug with us timing out improperly on the
luks password prompt), and then continuted to the host system by running
'systemctl start systemd-switch-root.service'. My error, but the results are
confusing and bad: systemd in the host sees 'systemd-tmpfiles-setup.service'
as started successfully, and doesn't restart it, so the setup for /tmp/.X11 is
not done and gdm.service fails. So while we wouldn't encounter this during
normal successful boot, I think it's good to make this more robust.

The dep is added to systemd-tmpfiles-{setup,clean}, because /tmp is not
propagated over switch-root. /dev is, so I didn't touch
systemd-tmpfiles-setup-dev.service.
units/systemd-tmpfiles-clean.service
units/systemd-tmpfiles-setup.service