On Mon, Apr 10, 2017 at 9:04 PM, Casey Schaufler [off-list ref] wrote:

On 4/10/2017 11:27 AM, Djalal Harouni wrote:

quoted

On Mon, Apr 10, 2017 at 5:42 PM, Casey Schaufler [off-list ref] wrote:

quoted

On 4/9/2017 3:42 AM, Djalal Harouni wrote:

[...]

quoted

quoted

quoted

--- a/security/security.c
+++ b/security/security.c

@@ -70,6 +70,7 @@ int __init security_init(void)
      capability_add_hooks();
      yama_add_hooks();
      loadpin_add_hooks();
+     modautorestrict_init();

This should be modautorestrict_add_hooks() if this were
a "minor" module, but as it's using a blob it is a "major"
module. Either way, this is not right.

Do you mean that if I'm using a blob, it should go with the rest LSMs
in do_security_initcalls() ?

Right. Today you have coincidental non-interference because
no one else is using the task blob. As you're aware, TOMOYO
is going to start using it, and I believe the AppArmor has
plans for it as well. There are parts of the Smack cred blob
that should probably go in the task blob as they aren't used
in access decisions. I haven't looked closely enough, but that's
possible for SELinux, too. So even though it's a new blob, the
major/minor rules apply.

Ok, point taken.

Thanks!

-- 
tixxdz