On Fri, Feb 12, 2021 at 2:39 PM Andi Kleen <ak@xxxxxxxxxxxxxxx> wrote: > > > But AFAI recursive #VE is entirely possible. The moment #VE reads that > > ve_info thing, NMIs can happen, which can trigger another #VE which then > > clobbers your stack and we're irrecoverably screwed again. > > I don't believe we have anything currently in the NMI handler that > would trigger #VE. While some operations may need TDCALL (like MSR > accesses) those should be all directly hooked. > > Also in general to avoid clobbering your stack you would just need > to make sure to adjust the IST stack before you do anything that > could cause another #VE. Except that the world contains more than just #VE. We could get #VE and then NMI and then #VE or #VE and MCE (in a future revision?), etc. The x86 exception situation is a mess.
