I've continued to run into selinux issues where this message pops out: /opt/intel/composerxe-2011.3.174/compiler/lib/intel64/libcilkrts.so.5: cannot restore segment prot after reloc: Permission denied We typically fiddle around with selinux settings to make this go away. My big concern is that my customers will run into this problem, and our customers will not want to fiddle with selinux settings. I don't really understand what's going on here, but I have the feeling that if I could ship staticall-linked binaries, the problem wouldn't arise. What are the best practices for an ISV to ship binaries that use cilk? (For us, one of the "best practices" has always been static compilation of anything that's not in the standard linux libraries.) -Bradley
selinux issues
selinux issues
Nähere Informationen zur Compiler-Optimierung finden Sie in unserem Optimierungshinweis.
Hi Bradley,
So can you confirm that if the SELinux settings are set to permissive that this problem goes away? What OS are you seeing this on and what are the SELinux settings when it fails?
Brandon Hewitt
Technical Consulting Engineer
Tools Knowledge Base: http://software.intel.com/en-us/articles/tools
Software Product support info: http://www.intel.com/software/support
Hello Brandon,
I work at Tokutek with Bradley. I can confirm that SELinux settings are set to permissive, the problem goes away, but if they are set to Enforcing, then the problem persists (these are the only two settings AFAIK). The OS is CentOS 5.5.
-Zardosht
So if our customer is running in Enforcing mode, will our software fail?
It looks likely. I wasn't able to reproduce this when I posted about this at http://software.intel.com/en-us/forums/showthread.php?t=78752&o=a&s=lr, but admittedly I might not have got the right configuration. And we don't have any CentOS* systems in our lab - CentOS is not one of our supported/tested Linux* distributions. Ones that are from our Release Notes:
o Asianux* 3.0, 4.0
o Fedora* 12, 13
o Red Hat Enterprise Linux* 4, 5, 6
o SUSE LINUX Enterprise Server* 10, 11
o Ubuntu* 10.04
o Debian* 5.0
That doesn't mean that I don't consider this a problem, but it will make it trickier for me to deal with if it's CentOS-specific. I'll keep you posted on this thread.
Brandon Hewitt
Technical Consulting Engineer
Tools Knowledge Base: http://software.intel.com/en-us/articles/tools
Software Product support info: http://www.intel.com/software/support
centos 5 should be the same as rhel 5. So this is puzzling. -Bradley
So Red Hat 4.2 works fine:
$ ./linear-recurrence
the linear recurrence value is 9
time (in ms) for sequential execution: 218.913
time (in ms) for parallel execution: 241.232
$cat /selinux/enforce
0$ sudo /usr/sbin/setenforce 1
$ ./linear-recurrence
the linear recurrence value is 9
time (in ms) for sequential execution: 218.913
time (in ms) for parallel execution: 238.633
I'll see if I can try it on Red Hat EL 5.
Brandon Hewitt
Technical Consulting Engineer
Tools Knowledge Base: http://software.intel.com/en-us/articles/tools
Software Product support info: http://www.intel.com/software/support



