On Thu, 2020-05-14 at 18:20 +0200, Borislav Petkov wrote: > On Thu, May 14, 2020 at 09:15:59AM -0700, Sean Christopherson wrote: > > I'm not opposed to adding a kernel param to disable SGX. At one point > > there was a proposal to extend clearcpuid to allow disabling multiple > > feature bits, but it looks like that went the way of the dodo. > > > > Note, such a param would disable SGX entirely, e.g. clear the feature bit > > in /proc/cpuinfo and prevent any in-kernel SGX code from running. > > It is a usual practice for big features like SGX to add a > "nosgx" cmdline param to disable it in case something goes > south. We do this for all features - see all "no*" switches in > Documentation/admin-guide/kernel-parameters.txt Uh oh, should probably address this. Should I send v31 today with a "nosgx" patch added? Sorry for missing this one :-/ /Jarkko
