On Mon, Feb 10, 2014 at 11:12 AM, Steve Grubb <sgrubb(a)redhat.com> wrote:
On Monday, February 10, 2014 11:01:36 AM Andy Lutomirski wrote:
> >> And I still think this needs more changes. Once again, I do not think
> >> that, say, __audit_log_bprm_fcaps() should populate context->aux if
> >> !TIF_SYSCALL_AUDIT, this list can grow indefinitely. Or
> >> __audit_signal_info()...
> >>
> >> Perhaps __audit_syscall_exit() should also set context->dummy?
> >
> > That would work.
> >
> > I'm still torn between trying to make it possible for things like
> > __audit_log_bprm_fcaps to start a syscall audit record in the middle
> > of a syscall or to just try to tighten up the current approach to the
> > point where it will work correctly.
>
> This is worse than I thought. Things like signal auditing can enter
> the audit system from outside of a syscall. I don't think there's
> currently any way to tell whether you're in a syscall (when
> TIF_SYSCALL_AUDIT is clear) so getting this to work right would
> require arch help.
>
> I'll ask what people on the Fedora list think about just changing the
> default to -t task,never.
I can't recall ever seeing the task filter used in real life. But assuming you
wanted to audit no tasks, what is the difference between using that filter and
never setting audit_enable in the first place?
Two possible differences:
1. You can toggle it both ways at runtime. Setting audit_enabled is a
one-way street (at least as far TIF_SYSCALL_AUDIT is concerned).
2. Do AVC denial messages still get logged if audit_enable == 0? If
not, then audit_enable is a non-starter.
--Andy