On Mon, Dec 14, 2015 at 10:52 AM, Roberts, William C <
[email protected]> wrote:

>
>
> > -----Original Message-----
> > From: Roberts, William C
> > Sent: Monday, December 14, 2015 10:38 AM
> > To: 'Stephen Smalley' <[email protected]>; [email protected]
> > Subject: RE: mac_override: What does ignore mean?
> >
> > > >> On 12/14/2015 11:57 AM, Roberts, William C wrote:
> > > >>> According to:
> > > >>> http://selinuxproject.org/page/ObjectClassesPerms#capability2,
> > > >>> mac_override is ignored. What does that actually mean? Is it
> > > >>> always denied (my guess) or always allowed?
> > > >>
> > > >> It is never checked by SELinux, only by Smack.
> > > >>
> > > >
> > > > What does that entail exactly? The messages printed to dmesg are
> > > > "avc denied". Does the "is capable" checks call into SE Linux and
> > > > EPERM is always
> > > returned?
> > > >
> > > > I ask this in the context of an out of tree driver that is currently
> > > > and incorrectly
> > > coded with a capable(MAC_OVERRIDE) check.
> > >
> > > No, the logic performed by the capable hook is not specific to any
> > > capability; it just checks whether that permission bit is set in the
> corresponding
> > access vector.
> > > So you can allow it in policy and it should be fine.  But it is wrong
> > > for the driver to be using that capability...
> >
> > That's what I thought based on looking at the code. I advised the driver
> team that
> > they Should be doing some other type of is_capable() check, likely
> SYS_ADMIN
> > for their needs.
> >
> > Thanks, I just wanted to confirm.
>
> FYI more details:
>
> Here is the code:
>
> https://android.googlesource.com/kernel/x86_64.git/+/android-x86_64-fugu-3.10-marshmallow/drivers/staging/sep54/sepfs.c
>
> Line 240.
>
> They have a UID/GID access list for each command, if the process has cap
> MAC_OVERRIDE, the check is skipped. I don't know why
> They even need this capable check, they should just add the privileged
> components into their ACL and drop this. Better yet, if they
> need finer access controls, they could do an implementation and out of
> tree patch ala binder (would like to avoid this).


The best part is, their commented out with not supported on Android L
elsewhere in the file.




> _______________________________________________
> Seandroid-list mailing list
> [email protected]
> To unsubscribe, send email to [email protected].
> To get help, send an email containing "help" to
> [email protected].
>



-- 
Respectfully,

William C Roberts
_______________________________________________
Seandroid-list mailing list
[email protected]
To unsubscribe, send email to [email protected].
To get help, send an email containing "help" to 
[email protected].

Reply via email to