umountnfs.sh is called improperly, and fails, stopping a machine rebooting
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
sysvinit (Ubuntu) |
New
|
Undecided
|
Unassigned |
Bug Description
Binary package hint: initscripts
umountnfs.sh is marked "start on runlevels 0 and 6", it should be "stop on runlevels 0 and 6" ("defaults" works acceptably). The script is set to ignore requests to start since it doesn't do any starting. Alternatively the script should be changed to interpret "start" as "do your unmounting thing" and ignore requests to stop (umountfs and umountroot are marked "start on 0 and 6", so I assume they work this way)
What makes this particularly nasty is that because the NFS share is still mounted, the system refuses to shut down, getting stuck at ~90% -- thus the box is effectively bricked until someone with physical access comes and pulls the power by hand...
There is a bug here, but your assertion that it should interpret "start" as "unmount" is incorrect -- from sysv-rc README.runlevels (same applies for rc0.d):
"Then the /etc/rc6.d/SXXxxxx scripts will be executed alphabetically with "stop" as the first argument as well. The reason is that there is nothing to start anymore at this point - all scripts that are run are meant to bring the system down."