Red Hat Bugzilla – Bug 270821
Doesn't suspend after a suspend action went wrong
Last modified: 2015-03-04 20:19:07 EST
Description of problem: If a suspend action goes wrong, it may leave the /.suspended file on the next boot which inhibits subsequent suspend/hibernate actions due to take_suspend_lock() in /usr/lib/pm-utils/functions. Version-Release number of selected component (if applicable): pm-utils-0.99.3-6.fc7 How reproducible: It depends. Steps to Reproduce: 1. On a machine with broken suspend/hibernate (one of those not working should be sufficient), run pm-suspend or pm-hibernate (whatever is broken, e.g. with an old version of pm-utils with broken quirks) 2. After the action has sufficiently blown up in your face, reboot the machine 3. Run pm-suspend/-hibernate (whatever works, e.g. a new version of pm-utils with fixed quirks Actual results: Machine switches to a VT and nothing else Expected results: Machine suspends/hibernates, stale /.suspended files get cleaned up some way. Additional info: A stale /.suspended file doesn't even give an error message, that's really bad. It should at least give that so the user had an idea what goes wrong. Then, stale /.suspended should be detected so it doesn't inhibit subsequent suspend/hibernate actions.
This appears to be a duplicate of bug 270841.
There is already some code in take_suspend_lock() that checks, whether a process with the id stored in the lockfile is still running. I do not know a better way to implement a lockfile in bash.
Sorry, this will be only fixed in Fedora 9, it is currently fixed in Fedora Rawhide.