© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
73416ba
(Get the code!)
Steve, note that while this issue has been masked in Maverick & Natty by making killall5 accept any number of OMITPIDs, it is still a problem in Lucid for those who have upstart jobs that won't shut down. The symptom is a journal replay (or heaven forbid worse) after every single reboot (because killall5 crashes and doesn't kill what it's supposed to kill). It depends on how many buggy upstart jobs one has installed...
For Maverick and Natty: it may seem that PID reusal is unlikely, but it's not quite so: at boot time, a long-running upstart job gets a low PID (e.g. 425). The system stays up for a day or two. In that interval, PIDs wrap around 32767 several times. At shutdown, it's very possible for a new process to hit PID 425 and get inappropriate immunity. You may have thought that reuse only occurred if PIDs made a full wrap-around in the 10 seconds of so worth of shutdown -- but it's not so.
In any case, as I said, why even run this risk? There's no need to accumulate PIDs that once belonged to now-dead upstart jobs (I think you agreed with my analysis of the code, but I'm not quite sure, given your phrasing.)