[systemd-devel] Partially failed legacy initscripts

[Andrey Borzenkov]

I noticed that tmp unmount failed even if I rebooted immediately after
booting up. The reason was vmware initscript. Script partially failed
because no drivers were present in currently running kernel. Still
some stuff was started. But apparently because systemd believed
service has failed, it did not try to stop it:

/tmp:                 1814
UID        PID  PPID  C STIME TTY          TIME CMD
root      1814     1  0 20:41 ?        00:00:00 /usr/bin/vmware-usbarbitrator

vmware.service - LSB: Manages the services needed to run VMware software
          Loaded: loaded (/etc/rc.d/init.d/vmware)
          Active: failed since [Tue, 05 Oct 2010 20:41:41 +0400; 3min 36s ago]
         Process: 1770 (/etc/rc.d/init.d/vmware start, code=exited,
status=1/FAILURE)
          CGroup: name=systemd:/systemd-1/vmware.service
                  └ 1814 /usr/bin/vmware-usbarbitrator

Traditional sysvinit would have stopped it anyway. Notice that cgroup
is still available and contains some processes for this service. So
what would be the right behaviour in this case?

a) try to stop failed service anyway - that would be compatible with
traditional initscripts handling
b) try to stop failed service if it has some processes running