[Sorry for the delay in responding to this.]

On Tue, Jul 22, 2014 at 12:16:53PM -0700, John Johansen wrote:
> Recently a bug was opened due to a misunderstanding of how apparmor's
> script handling and permissions work.
> 
> https://bugs.launchpad.net/apparmor/+bug/1346553
> 
> Basically the profile that a script runs under does not need r or x
> permissions on the interpreter (generally). The question was raised
> if this is the behavior that is desired, or whether a script profile
> should require access permissions to the interpreters binary.
> 
> AppArmor used to do this years ago, and it would be fairly trivial to
> add it back in (kernel change only). And it could be conditional on
> ABI versioning to maintain compatability.
> 
> So that only leaves the question of whether we should keep the
> current behavior or require explicit permissions for the interpreter
> binary.

Truthfully, this was a behavioral change I had not noticed, nor
did I expect. It seems to me that some form of permission on the
interpreter for a confined script is important; being able to adjust
the interpreter allowed based on modifying the script seems... dubious.

There are a fair number of assumptions in policy that an interpreter
will be required; for example, abstractions/base contains many
variations of ld.so with "mrix" permissions, yet that no longer seems
necessary for binfmt_elf binaries (I realize there's a difference int
he kernel's handling of scripts via binfmt_misc and how elf binaries
are handled with binfmt_elf, but they are analogous).

All that said, it strikes me as a low priority issue, although I do
have the nagging feeling there's a threat here with not requiring a
permission that I can't quite articulate.

-- 
Steve Beattie
<[email protected]>
http://NxNW.org/~steve/

Attachment: signature.asc
Description: Digital signature

-- 
AppArmor mailing list
[email protected]
Modify settings or unsubscribe at: 
https://lists.ubuntu.com/mailman/listinfo/apparmor

Reply via email to