Red Hat Bugzilla – Bug 722814
/var/lock/lockdev is lost after reboot - add to tmpfiles.d
Last modified: 2011-07-19 05:05:14 EDT
Description of problem:
If the /var/lock/lockdev directory is missing, minicom will error out with "Device /dev/ttyUSB0 lock failed: Operation not permitted."
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Plug a serial adapter on /dev/ttyUSB0
2. minicom -s USB0 to configure it
3. minicom USB0
Installing the lockdev package _or_ creating the /var/lock/lockdev directory fixed the problem.
A simple fix for this would be to add a lockdev dependency to minicom.
Actually - after a bit more diagnosis, the dependency is there, and lockdev is installed in the problem machines.
The issue is with /var/lock/lockdev -- it is lost after reboot. Should be declared in a /etc/tmpfiles.d/ directory. Something like
$ cat /etc/tmpdiles.d/lockdev.conf
d /var/lock/lockdev 0775 root lock
Issue exists in rawhide as well.
I had dropped /etc/tmpfiles.d/lockdev.conf from lockdev package because of bug #692714. Moving this BZ to systemd which should ship this file in F15.
Note that this issue affects F14 as well (and this impacts OLPC OS releases).
Even if systemd gets this fixed on F15/rawhide, F14 needs /etc/tmpfiles.d/lockdev.conf
Hmm, /usr/lib/tmpfiles.d/legacy.conf creates that directory, so it should be there. Are you suggesting this file is not existant for you?
Actually, you are right, and my repro test on F15 was wrong, so systemd is in the clear. Reassigning to lockdev.
The problem repros on F14 -- the F14 package in fedpkg needs to get some bits from commit 76efe301a85adf8470f72569c6a90d8fd4c2a8e5 .
In other words - the change was applied for F15 and then removed in favour of the systemd fix. That initial change needs to be applied to F14.
is F15 feature so the /var/lock/lockdev should survive reboot on F14.
Adding /etc/tmpfiles.d/lockdev.conf to F14 doesn't change anything.
Or am I wrong, Lennart ?
Gaah. Peter Robinson just pointed out that I see this on F14 because at OLPC we have backported var-run-tmpfs to our custom F14.
So this bug is invalid (in Fedora), and I apologize for having bothered everyone with it. On the upside, you have helped me diagnose a bug in OLPC's build, which is used by >1.5M children around the world .
Thanks for your patience.