Re: [GIT] Security subsystem updates for 3.9

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

On Thu, 2013-02-21 at 08:26 -0800, Linus Torvalds wrote:
> On Thu, Feb 21, 2013 at 6:03 AM, James Morris <jmorris@namei.org> wrote:
> > This is basically a maintenance update for the TPM driver and EVM/IMA.
> 
> Hmm. There were conflicts in lib/digsig.c and ima_main.c. The digsig
> one was pretty trivial, but I'd like people to take a look at the IMA
> one.
> 
> And that's not because the conflict itself was all that complex - I'm
> pretty sure I resolved it correctly. But I do want to make sure that
> everybody agrees on the whole module integrity checking thing. I
> resolved it to be like the semantics in Mimi's commit a7f2a366f623,
> which means that for non-file modules, IMA does:
> 
> Only fail the non-file module if
>    (a) IMA_APPRAISE_MODULES was set

Almost, and enforcing file integrity is enabled.  The merged result
should look like what's contained in
linux-integrity/next-upstreamed-patches:

int ima_module_check(struct file *file)
{
        if (!file) {
                if ((ima_appraise & IMA_APPRAISE_MODULES) &&
                    (ima_appraise & IMA_APPRAISE_ENFORCE)) {
#ifndef CONFIG_MODULE_SIG_FORCE
                        return -EACCES; /* INTEGRITY_UNKNOWN */
#endif
                }
                return 0;
        }
        return process_measurement(file, file->f_dentry->d_name.name,
                                   MAY_EXEC, MODULE_CHECK);
}

thanks,

Mimi

> *and*
>    (b) CONFIG_MODULE_SIG_FORCE is not set.

> If CONFIG_MODULE_SIG_FORCE it ignores IMA_APPRAISE_MODULES entirely,
> and the module signature checking overrides everything. And if
> IMA_APPRAISE_MODULES is not set, we say "whatever". So it makes sense,
> but I wanted people to just be aware of it and agree on it, since the
> security tree modified this part without apparently being aware of the
> changed semantics.