Re: [PATCH] integrity: fix IMA inode leak

[Mimi Zohar <zohar@linux.vnet.ibm.com>]

On Mon, 2009-06-08 at 09:15 -0700, Linus Torvalds wrote:
> 
> On Mon, 8 Jun 2009, Mimi Zohar wrote:
> > 
> > Today the security calls are synomymous with MAC.  If I understand
> > correctly, you're suggesting we need to have a single security layer,
> > which, depending on the hook, calls either MAC or integrity, or both.
> 
> I don't think we need a single security layer per se.
> 
> But I do think that we _already_ hide IMA conceptually under the 
> "security/" subdirectory, and that the VFS layer shouldn't need to care 
> about whatever internal details.
> 
> We should not have generic code end up having to know about all the 
> details, when we already have a conceptual nesting. It would be much 
> better for generic code to just have to worry about one security hook that 
> then encompasses all the models, than having several different hooks for 
> each detail.
> 
> 		Linus

Ok, so instead of having a full fledge single security layer, only add
the security layer for those places where both the LSM hooks and IMA
co-exist: security_file_mmap, security_bprm_check, security_inode_alloc,
security_inode_free, and security_file_free. As the LSM hooks are called
'security_XXXX', the call would look something like:

security_all_inode_free() {
        ima_inode_free()
        security_inode_free()
}

Mimi Zohar