[CRIU] Cannot run zdtm.py script, flock problem

[Rodrigo Bruno]

Hi,

On Mon, 25 Jan 2016 12:38:14 +0300
Pavel Emelyanov <xemul at parallels.com> wrote:

> On 01/23/2016 11:09 PM, Rodrigo Bruno wrote:
> > Hi,
> > 
> > I was trying to run tests using the zdtm.py script.
> > 
> > However, when running "./zdtm.py run -a":
> > 
> > flock: ./zdtm_mount_cgroups: Text file busy
> > Traceback (most recent call last):
> >   File "./zdtm.py", line 1403, in <module>
> >     tst.available()
> >   File "./zdtm.py", line 436, in available
> >     subprocess.check_call(["flock", "zdtm_mount_cgroups", "./zdtm_mount_cgroups"])
> >   File "/usr/lib/python2.7/subprocess.py", line 540, in check_call
> >     raise CalledProcessError(retcode, cmd)
> > subprocess.CalledProcessError: Command '['flock', 'zdtm_mount_cgroups', './zdtm_mount_cgroups']' returned non-zero exit status 69
> > 
> > This problem seems to happen because flock is trying to lock the script it is going
> > to run.
> > 
> > Has anyone got the same problem?
> > 
> > Is this lock important?
> 
> Yes, this is needed to pre-mount cgroups only once when we run several zdtm.py-s
> in parallel.
> 
> > I'm able to run all tests I if remove the flock command.
> 
> OK, so while you don't run zdtm.py-s in parallel, you can move forward without
> one. But once time permits, can you investigate why flock behaves like that?

I found an old discussion about this:
http://comments.gmane.org/gmane.linux.utilities.util-linux-ng/4997

Is seems that the flock version that is distributed in all Ubuntu and Ubuntu-based 
systems (2.20.1) does not allow locking the script it is going to execute.

I cloned and compiled util-linux (https://github.com/karelzak/util-linux) and the
resultant flock already works.

A possible solution, if you think this is a problem, is to create a lock file 
(not using the script file as lock).

> 
> > cheers,
> > 
> > rodrigo
> > 
> 
> -- Pavel


-- 
Rodrigo Bruno <rbruno at gsd.inesc-id.pt>