Re: [PATCH v23 02/23] LSM: Create and manage the lsmblob data structure.

From: Mimi Zohar <zohar@linux.ibm.com>
To: Casey Schaufler <casey@schaufler-ca.com>,
	casey.schaufler@intel.com, jmorris@namei.org,
	linux-security-module@vger.kernel.org, selinux@vger.kernel.org
Cc: john.johansen@canonical.com, linux-kernel@vger.kernel.org,
	linux-audit@redhat.com, bpf@vger.kernel.org, sds@tycho.nsa.gov
Subject: Re: [PATCH v23 02/23] LSM: Create and manage the lsmblob data structure.
Date: Mon, 28 Dec 2020 17:14:48 -0500	[thread overview]
Message-ID: <07784164969d0c31debd9defaedb46d89409ad78.camel@linux.ibm.com> (raw)
In-Reply-To: <886fcd04-6a08-d78c-dc82-301c991e5ad8@schaufler-ca.com>

On Mon, 2020-12-28 at 12:06 -0800, Casey Schaufler wrote:
> On 12/28/2020 11:24 AM, Mimi Zohar wrote:
> > Hi Casey,
> >
> > On Fri, 2020-11-20 at 12:14 -0800, Casey Schaufler wrote:
> >> diff --git a/security/security.c b/security/security.c
> >> index 5da8b3643680..d01363cb0082 100644
> >> --- a/security/security.c
> >> +++ b/security/security.c
> >>
> >> @@ -2510,7 +2526,24 @@ int security_key_getsecurity(struct key *key, char **_buffer)
> >>
> >>  int security_audit_rule_init(u32 field, u32 op, char *rulestr, void **lsmrule)
> >>  {
> >> -       return call_int_hook(audit_rule_init, 0, field, op, rulestr, lsmrule);
> >> +       struct security_hook_list *hp;
> >> +       bool one_is_good = false;
> >> +       int rc = 0;
> >> +       int trc;
> >> +
> >> +       hlist_for_each_entry(hp, &security_hook_heads.audit_rule_init, list) {
> >> +               if (WARN_ON(hp->lsmid->slot < 0 || hp->lsmid->slot >= lsm_slot))
> >> +                       continue;
> >> +               trc = hp->hook.audit_rule_init(field, op, rulestr,
> >> +                                              &lsmrule[hp->lsmid->slot]);
> >> +               if (trc == 0)
> >> +                       one_is_good = true;
> >> +               else
> >> +                       rc = trc;
> >> +       }
> >> +       if (one_is_good)
> >> +               return 0;
> >> +       return rc;
> >>  }
> > So the same string may be defined by multiple LSMs.
> 
> Yes. Any legal AppArmor label would also be a legal Smack label.
> 
> >>  int security_audit_rule_known(struct audit_krule *krule)
> >> @@ -2518,14 +2551,31 @@ int security_audit_rule_known(struct audit_krule *krule)
> >>         return call_int_hook(audit_rule_known, 0, krule);
> >>  }
> >>
> >> -void security_audit_rule_free(void *lsmrule)
> >> +void security_audit_rule_free(void **lsmrule)
> >>  {
> >> -       call_void_hook(audit_rule_free, lsmrule);
> >> +       struct security_hook_list *hp;
> >> +
> >> +       hlist_for_each_entry(hp, &security_hook_heads.audit_rule_free, list) {
> >> +               if (WARN_ON(hp->lsmid->slot < 0 || hp->lsmid->slot >= lsm_slot))
> >> +                       continue;
> >> +               hp->hook.audit_rule_free(lsmrule[hp->lsmid->slot]);
> >> +       }
> >>  }
> >>
> > If one LSM frees the string, then the string is deleted from all LSMs. 
> > I don't understand how this safe.
> 
> The audit system doesn't have a way to specify which LSM
> a watched label is associated with. Even if we added one,
> we'd still have to address the current behavior. Assigning
> the watch to all modules means that seeing the string
> in any module is sufficient to generate the event.

I originally thought loading a new LSM policy could not delete existing
LSM labels, but that isn't true.  If LSM labels can come and go based
on policy, with this code, could loading a new policy for one LSM
result in deleting labels of another LSM?

> 
> >
> >> -int security_audit_rule_match(u32 secid, u32 field, u32 op, void *lsmrule)
> >> +int security_audit_rule_match(u32 secid, u32 field, u32 op, void **lsmrule)
> >>  {
> >> -       return call_int_hook(audit_rule_match, 0, secid, field, op, lsmrule);
> >> +       struct security_hook_list *hp;
> >> +       int rc;
> >> +
> >> +       hlist_for_each_entry(hp, &security_hook_heads.audit_rule_match, list) {
> >> +               if (WARN_ON(hp->lsmid->slot < 0 || hp->lsmid->slot >= lsm_slot))
> >> +                       continue;
> >> +               rc = hp->hook.audit_rule_match(secid, field, op,
> >> +                                              &lsmrule[hp->lsmid->slot]);
> >> +               if (rc)
> >> +                       return rc;
> > Suppose that there is an IMA dont_measure or dont_appraise rule, if one
> > LSM matches, then this returns true, causing any measurement or
> > integrity verification to be skipped.
> 
> Yes, that is correct. Like the audit system, you're doing a string based
> lookup, which pretty well has to work this way. I have proposed compound
> label specifications in the past, but even if we accepted something like
> "apparmor=dates,selinux=figs" we'd still have to be compatible with the
> old style inputs.
> 
> >
> > Sample policy rules:
> > dont_measure obj_type=foo_log
> > dont_appraise obj_type=foo_log

IMA could extend the existing policy rules like "lsm=[selinux] |
[smack] | [apparmor]", but that assumes that the underlying
infrastructure supports it.

> >
> > Are there any plans to prevent label collisions or at least notify of a
> > label collision?
> 
> What would that look like? You can't say that Smack isn't allowed
> to use valid AppArmor labels. How would Smack know? If the label is
> valid to both, how would you decide which LSM gets to use it?

As this is a runtime issue, when loading a new policy at least flag the
collision.  When removing the label, when it is defined by multiple
LSMs, at least flag the removal.

> 
> >
> >> +       }
> >> +       return 0;
> >>  }
> >>  #endif /* CONFIG_AUDIT */
> 

--
Linux-audit mailing list
Linux-audit@redhat.com
https://www.redhat.com/mailman/listinfo/linux-audit