mountall: long, silent hangs on boot from 'find' while clearing tmpfs

[DaveHansen]

I thought my system was hanging at boot. It turns out that a lot of stuff was left in /tmp from a benchmark that was running and then crashed. I instrumented /etc/init/mounted-tmp.conf and found out that the 'find's to clear out /tmp were taking about 20 minutes to complete:

[ 125.732385] init: startpar-bridge (mountall-net--stopped) goal changed from start to stop
[ 125.732478] init: startpar-bridge (mountall-net--stopped) state changed from running to stopping
[ 125.732582] init: Handling stopping event
[ 125.732707] init: startpar-bridge (mountall-net--stopped) state changed from stopping to killed
[ 125.732779] init: startpar-bridge (mountall-net--stopped) state changed from killed to post-stop
[ 125.732839] init: startpar-bridge (mountall-net--stopped) state changed from post-stop to waiting
[ 125.733068] init: Handling stopped event
[ 126.076006] init: Connection from private client
[ 134.712191] init: Connection from private client
[ 1269.704606] init: mounted-tmp main process (1210) exited normally
[ 1269.704737] init: mounted-tmp goal changed from start to stop
[ 1269.704879] init: mounted-tmp state changed from running to stopping
[ 1269.704960] init: Handling stopping event
...

That's a bit too long, and _really_ makes it look like the system has hung for 20 minutes. Is there a chance we could get some messages spit out to the console somehow during this process? Or, perhaps, time out after a certain amount of time and fall back to the tmpfs that happens later.

I ran in to this on 13.10, but confirmed that the same code exists in the 14.04 and 14.10 bzr branches of mountall.

[Steve Langasek (vorlon)]

Yes, it would be appropriate to have the mounted-tmp job display a message via plymouth in this case.