Re: Security issues with local filesystem caching

[David Howells <dhowells@redhat.com>]

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

> >      (c) A security label that defines the context under which the module
> >          operates when accessing the cache.  This allows the module, when
> >          accessing the cache, to only operate within the bounds of the
> >          cache.
> 
> Well, only if the module is well-behaved in the first place, since a
> kernel module can naturally bypass SELinux at will.  What drives this
> approach vs. exempting the module from SELinux checking via a task flag
> that it raises and lowers around the access (vs. setting and resetting
> the sid around the access to the per-cache module context)?

Christoph objected very strongly to my bypassing of vfs_mkdir() and co, and Al
wasn't to happy about it either.  This should allow me, for example, to call
vfs_mkdir() rather than calling the inode op directly as the reason I wasn't
was that I was having to avoid the security checks it made.

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

> >  (*) The module will obtain label (c) by reading label (b) from the
> >      cachefilesd process when it opens the cachefiles control chardev and
> >      then passing it through security_change_sid() to ask the security
> >      policy to for label (c).
> 
> Do you mean security_transition_sid()?  security_change_sid() doesn't
> seem suited to that purpose

That's what Karl said to use.

> What would you use as the target SID and class?

I've no idea.  I tried to find out how to use this function from Karl, but he
said I should ask on the list.

> >      (3) current->security->sid will be set to label (c) so that
> >          vfs_mkdir(), vfs_create() and lookup ops will check for the
> >          correct labels.
> 
> I think you would want this to be a new ->fssid field instead, and
> adjust SELinux to use it if set for permission checking (which could
> also be leveraged by NFS later).

I could do that.  Does it actually gain anything?  Or are there good reasons
for not altering current->security->sid?  For instance, does that affect the
label seen on /proc/pid/ files?

> Or just use a task flag to disable checking on the module internal accesses.

I could do that too.

> >      Point (3) shouldn't cause a cross-thread race as it would appear that
> >      the security label can only be changed on single-threaded processes.
> >      Attempts to do so on multi-threaded processes are rejected.
> 
> I don't quite follow this.

Sorry, I meant that a process can only change its own security label if it's a
single-threaded process.  A kernel module can, of course, change the security
label at any time.

> But mutating ->sid could yield unfortunate behavior if e.g. another process
> happens to be sending that task a signal at the same time, so if you go this
> route, you want a ->fssid.

Okay... that seems like a good reason to do use the ->fssid approach.  How do I
tell if ->fssid is set?  Is zero usable as 'unset'?  Alternatively, would it be
reasonable to have ->fssid track ->sid when the latter changes?

David

--
This message was distributed to subscribers of the selinux mailing list.
If you no longer wish to subscribe, send mail to majordomo@tycho.nsa.gov with
the words "unsubscribe selinux" without quotes as the message.