Re: [patch 01/10] x86/fpu/signal: Clarify exception handling in restore_fpregs_from_user()
From: Jarkko Sakkinen
Date: Fri Sep 03 2021 - 02:05:59 EST
On Fri, 2021-09-03 at 09:00 +0300, Jarkko Sakkinen wrote:
> On Thu, 2021-09-02 at 16:08 +0200, Thomas Gleixner wrote:
> > On Thu, Sep 02 2021 at 16:08, Jarkko Sakkinen wrote:
> > > On Wed, 2021-09-01 at 16:47 +0000, Sean Christopherson wrote:
> > > > As for SGX consuming the trap number in general, it's correct. For non-KVM usage,
> > > > it's nice to have but not strictly necessary. Any fault except #PF on ENCLS is
> > > > guaranteed to be a kernel or hardware bug; SGX uses the trap number to WARN on a
> > > > !#PF exception, e.g. on #GP or #UD. Not having the trap number would mean losing
> > > > those sanity checks, which have been useful in the past.
> > >
> > > AFAIK, we do not consider #UD as a bug. Agree with the conclusion that SGX
> > > should never #MC, I just did not get this part. #UD is something that is
> > > useful for SGX run-time.
> > I understood that storing the trap number is useful. I was just
> > questioning the #MC angle. I.e. pretending that the #MC caused by ENCLS
> > is recoverable.
> Absolutely not.
> I mixed up #UD caused by CPU executing inside enclave and ENCLS causing
> #UD. Sorry about that.
> Because of KVM we have to catch #PF's, given that a new power cycle
> in the host resets the state of SGX protected memory in the guest.
.. catching #PF's makes also quite a lot of sense for the bare
metal case because otherwise we would have to have hook for
power state change that would have invalidate all enclaves
running in the system.