On Wed, Jul 20, 2022 at 05:13:08PM +1200, Kai Huang <kai.huang@xxxxxxxxx> wrote: > On Tue, 2022-07-19 at 07:49 -0700, Isaku Yamahata wrote: > > On Fri, Jul 08, 2022 at 02:23:43PM +1200, > > Kai Huang <kai.huang@xxxxxxxxx> wrote: > > > > > On Mon, 2022-06-27 at 14:53 -0700, isaku.yamahata@xxxxxxxxx wrote: > > > > From: Isaku Yamahata <isaku.yamahata@xxxxxxxxx> > > > > > > > > To support TDX, KVM is enhanced to operate with #VE. For TDX, KVM programs > > > > to inject #VE conditionally and set #VE suppress bit in EPT entry. For VMX > > > > case, #VE isn't used. If #VE happens for VMX, it's a bug. To be > > > > defensive (test that VMX case isn't broken), introduce option > > > > ept_violation_ve_test and when it's set, set error. > > > > > > I don't see why we need this patch. It may be helpful during your test, but why > > > do we need this patch for formal submission? > > > > > > And for a normal guest, what prevents one vcpu from sending #VE IPI to another > > > vcpu? > > > > Paolo suggested it as follows. Maybe it should be kernel config. > > (I forgot to add suggested-by. I'll add it) > > > > https://lore.kernel.org/lkml/84d56339-4a8a-6ddb-17cb-12074588ba9c@xxxxxxxxxx/ > > > > > > > OK. But can we assume a normal guest won't sending #VE IPI? Theoretically nothing prevents that. I wouldn't way "normal". Anyway this is off by default. -- Isaku Yamahata <isaku.yamahata@xxxxxxxxx>
