Re: [PATCH 08/28] SECURITY: Allow kernel services to override LSM settings for task actions [try #2]

[Casey Schaufler <casey@schaufler-ca.com>]

--- Stephen Smalley <sds@tycho.nsa.gov> wrote:

> 
> > I am much more concerned with the interfaces used to pass the
> > information into the kernel. I would expect that to be LSM
> > independent, not a call into libselinux that resolves into a
> > selinuxfs operation, or it's netlink equivilant. It would be
> > unfortunate if the userland/kernel interface became an obstacle
> > to cachefiles being adopted.
> 
> That wasn't the issue.  The interface to the cachefiles module would
> just consist of cachefilesd writing a string label to some pseudo file
> tell cachefiles what label to apply as the acting label for operations
> performed by cachefiles.  Which isn't SELinux-specific at all.

Calm down. Just checking.

> David was asking though how cachefilesd (the userspace agent) would
> obtain such a label to use.  And that may very well be LSM-specific, and
> as there is no LSM userspace API, it makes sense for him to invoke a
> libselinux function at present.  If a liblsm is later created and
> provides a common front-end API (internally dlopen'ing the right shared
> library based on some configuration, whether libselinux or libsmack or
> whatever), then cachefilesd can instead call the liblsm interface, but
> that doesn't exist today.

I am certainly not in favor of adding such complexity.
I suggest that cachefilesd get the context it wants using
whatever scheme works. I think there should be a cachefiles
specific (LSM independent) scheme for getting it into the
kernel, and a LSM hooks for setting it, if that's what he
really wants to do.



Casey Schaufler
casey@schaufler-ca.com