CentOS - Jun 2015 - Try II: selinux, xfs, and CentOS 6 and 5 issue

Tried just the selinux list yesterday, no answers, so I'm trying again.

I partitioned GPT, and formatted, as xfs,  a large (3TB) drive on a CentOS
6 system, which has selinux in permissive mode. I then moved the drive to
a CentOS 5 system. When we run a copy (it mirror-copies from another
system), we get a ton of errors. I discovered that the CentOS 5 system was
enforcing. I changed it to permissive, I labelled the directories and
files w/ semanage, did a restorecon, and even did a fixfiles, and *then* I
tried /.autorelabel and rebooted, and we still get a ton of errors:
Jun  1 17:01:32 <server> kernel: inode_doinit_with_dentry: 
context_to_sid(unconfined_u:object_r:file_t:s0) returned 22 for dev=sdd1
ino=2151541032

I had to reboot to disabled to get it to shut up.

So: is there something that selinux does in CentOS 6 that is in the
labelling on the xfs filesystem that I can do something about on the
CentOS 5 system, or do I just have to leave selinux disabled (until, maybe
in the next year, we can rebuild to 7....)?

       mark

On 06/02/2015 11:30 AM, m.roth at 5-cent.us wrote:
> Tried just the selinux list yesterday, no answers, so I'm trying again.
>
> I partitioned GPT, and formatted, as xfs,  a large (3TB) drive on a CentOS
> 6 system, which has selinux in permissive mode. I then moved the drive to
> a CentOS 5 system. When we run a copy (it mirror-copies from another
> system), we get a ton of errors. I discovered that the CentOS 5 system was
> enforcing. I changed it to permissive, I labelled the directories and
> files w/ semanage, did a restorecon, and even did a fixfiles, and *then* I
> tried /.autorelabel and rebooted, and we still get a ton of errors:
> Jun  1 17:01:32 <server> kernel: inode_doinit_with_dentry: 
> context_to_sid(unconfined_u:object_r:file_t:s0) returned 22 for dev=sdd1
> ino=2151541032
>
> I had to reboot to disabled to get it to shut up.
>
> So: is there something that selinux does in CentOS 6 that is in the
> labelling on the xfs filesystem that I can do something about on the
> CentOS 5 system, or do I just have to leave selinux disabled (until, maybe
> in the next year, we can rebuild to 7....)?
>
>        mark
>
> --
> selinux mailing list
> selinux at lists.fedoraproject.org
> https://admin.fedoraproject.org/mailman/listinfo/selinux
SELinux on RHEL5 did not have a MLS field in the label, so the directory
can not be used by both rhel5 and RHEL6 easily.

If all of the content on the device is going to be labeled the same,
then just use a context mount option

context="system_u:object_r:usr_t:s0"  for example.

On 06/02/2015 08:30 AM, m.roth at 5-cent.us wrote:
>
> I partitioned GPT, and formatted, as xfs,  a large (3TB) drive on a CentOS
> 6 system, which has selinux in permissive mode. I then moved the drive to
> a CentOS 5 system. When we run a copy ...  we still get a ton of errors:
> Jun  1 17:01:32 <server> kernel: inode_doinit_with_dentry:
> context_to_sid(unconfined_u:object_r:file_t:s0) returned 22 for dev=sdd1
> ino=2151541032
Maybe:
https://www.redhat.com/archives/fedora-selinux-list/2005-October/msg00135.html

It sounds like the new system is writing data into the filesystem that 
the older libraries cannot read, or cannot meaningfully interpret.

The SELinux contexts don't mean anything to the old 
system/libraries/policy, so disabling SELinux is probably the best 
option.  (I did not expect to ever advise disabling SELinux).

On 06/02/2015 10:20 AM, Daniel J Walsh wrote:
> If all of the content on the device is going to be labeled the same,
> then just use a context mount option
Thanks, Dan.  I'd misread the man page for mount, and thought that the 
context= option was "only" useful when mounting filesystems that
didn't
support SELinux attributes.