RE: [PATCH v5 1/2] x86/speculation: apply IBPB more strictly to avoid cross-process data leak

[Schaufler, Casey]

> -----Original Message-----
> From: Jiri Kosina [mailto:jikos@xxxxxxxxxx]
> Sent: Monday, September 10, 2018 12:14 PM
> To: Schaufler, Casey <casey.schaufler@xxxxxxxxx>
> Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>; Ingo Molnar <mingo@xxxxxxxxxx>;
> Peter Zijlstra <peterz@xxxxxxxxxxxxx>; Josh Poimboeuf
> <jpoimboe@xxxxxxxxxx>; Andrea Arcangeli <aarcange@xxxxxxxxxx>;
> Woodhouse, David <dwmw@xxxxxxxxxxxx>; Andi Kleen <ak@xxxxxxxxxxxxxxx>;
> Tim Chen <tim.c.chen@xxxxxxxxxxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx;
> x86@xxxxxxxxxx
> Subject: RE: [PATCH v5 1/2] x86/speculation: apply IBPB more strictly to avoid
> cross-process data leak
> 
> On Mon, 10 Sep 2018, Schaufler, Casey wrote:
> 
> > Why are you dropping the LSM check here, when in v4 you fixed the
> > SELinux audit locking issue? We can avoid introducing an LSM hook
> > and all the baggage around it if you can do the
> security_ptrace_access_check()
> > here.
> 
> So what guarantees that none of the hooks that
> security_ptrace_access_check() is invoking will not be taking locks (from
> scheduler context in this case)?

The locking issue in the security modules is the same regardless of
whether the call of security_ptrace_access_check() comes from the
__ptrace_access_check() you're adding here or from a new security
hook (I have proposed security_task_safe_sidechannel) that gets added
in the same place later on. Adding a new hook results in duplication,
because there now has to be code that does exactly the same thing as
__ptrace_access_check() but without the new NOACCESS_CHECK mode.

Yes, It would require that this patch be tested against all the existing
security modules that provide a ptrace_access_check hook. It's not like
the security module writers don't have a bunch of locking issues to deal with. 

> Thanks,
> 
> --
> Jiri Kosina
> SUSE Labs