[PATCH 4/4] The control group itself

Paul Menage menage at google.com
Tue Jan 15 08:13:40 PST 2008

On Jan 15, 2008 6:44 AM, Serge E. Hallyn <serue at us.ibm.com> wrote:
> I don't think so...  Wouldn't really make sense for the cgroup
> infrastructure to presume to know what to enforce, and I don't see any
> checks around the _write functions in cgroup.c, and no capable() calls
> at all.

The cgroup filesystem can provide simple unix-level permissions on any
given file. Am I right in thinking that having an entry in the mapper
doesn't automatically give privileges for a device to the members of
the cgroup, but they also have to have sufficient privilege in their
own right? If so, that might be sufficient.

One other thought - should the parse/print routines themselves do a
translation based on the device mappings for the writer/reader's
cgroup? That way you could safely give a VE full permission to write
to its children's device maps, but it would only be able to add/remap
device targets that it could address itself.


More information about the Containers mailing list