On Tue, 6 Apr 2021 11:32:11 +0200 Borislav Petkov wrote: > On Tue, Apr 06, 2021 at 09:24:24PM +1200, Kai Huang wrote: > > Such invalid input has already been handled in handle_encls_xx() before calling > > the two helpers in this patch. KVM returns to Qemu and let it decide whether to > > kill or not. The access_ok()s here are trying to catch KVM bug. > > Whatever they try to do, you cannot continue creating an enclave using > invalid input, no matter whether you've warned or not. People do not > stare at dmesg all the time. > > > If so we'd better inject an exception to guest (and return 1) in KVM so guest > > can continue to run. Otherwise basically KVM will return to Qemu and let it > > decide (and basically it will kill guest). > > > > I think killing guest is also OK. KVM part patches needs to be updated, though, > > anyway. > > Ok, I'll make the changes and you can redo the KVM rest ontop. > Thank you!
