On 1/30/2015 5:44 PM, Matthew Gillen wrote:
Looking at that command on my fedora 20 box, I see the following: ldd -r /usr/bin/condor_status shows that libselinux.so is explicitly linked in to the binary. So it will always try to load it. Interestingly, there is no libsepol.so that gets loaded if I run it as a user or root (although that file does exist in /lib on my system, and SELinux is enabled).
That's the behavior that I'm seeing on the boxes that segfault: runs fine as root with no libsepol linked, segfaults as me after libsepol is loaded. What's extra weird is that only three of the six boxes do that; the other three run normally with the same binaries and SELinux likewise disabled.
Perhaps libsepol is only loaded by libselinux under certain conditions (i.e. using explicit calls to dlopen instead of relying on the startup-linking), and your user has some environment var set that creates those conditions?
Unlikely. I used the same tarball to install and configure Condor on all six nodes and I'm running from my AFS home directory on all six so my environment is constant.
Since I couldn't figure out the cause I tried a newer Condor tarball from U-Wisc. More strangeness: the binaries in that newer tarball work correctly on all six nodes. Makes me think that there's a problem with the older binaries from U-Wisc.
-- Rich P. _______________________________________________ Discuss mailing list [email protected] http://lists.blu.org/mailman/listinfo/discuss
