On Thu, Oct 4, 2018 at 9:47 AM Jiri Kosina [off-list ref] wrote:

On Thu, 27 Sep 2018, Jann Horn wrote:

quoted

quoted

Yes. Since the PTRACE_MODE_NOAUDIT was in PTRACE_MODE_IBPB in Jiri's
previous patch set and not in PTRACE_MODE_SCHED in this one I assumed
that there was a good reason for it.

Jiri, was there a good reason for it, and if so, what was it?

[ FWIW PTRACE_MODE_NOAUDIT being in PTRACE_MODE_IBPB goes back to original
  Tim's pre-CRD patchset ]

Well, we can't really call out into audit from scheduler code, and the
previous versions of the patchsets didn't have PTRACE_MODE_SCHED, so it
had to be included in PTRACE_MODE_IBPB in order to make sure we're not
calling into audit from context switch code.

Or did I misunderstand the question?

If I understand Casey correctly, he is saying that your patch
(https://lore.kernel.org/lkml/nycvar.YFH.7.76.1809251437340.15880@cbobk.fhfr.pm/ (local))
doesn't include PTRACE_MODE_NOAUDIT for IBPB, but the previous v6 of
your patch (https://lore.kernel.org/lkml/nycvar.YFH.7.76.1809121105330.15880@cbobk.fhfr.pm/ (local))
did include it, and therefore Casey thinks that there is a specific
reason why you removed PTRACE_MODE_NOAUDIT, and therefore Casey is
adding special-case logic for PTRACE_MODE_SCHED to Smack when simply
using PTRACE_MODE_NOAUDIT would also work.

I think that Casey should change ptrace_may_access_sched() to use
"mode | PTRACE_MODE_SCHED | PTRACE_MODE_NOAUDIT".