On 28/12/2018 12:39, Andrew Cooper wrote: > c/s fd32dcfe4c "x86/vmx: Don't leak EFER.NXE into guest context" had an > unintended consequence on Harpertown cores which, as it turns out, don't > load MSR_EFER fully from the MSR Load List - on reentry to the guest, > EFER.SCE is clear irrespective of the value in load list. > > This, being catastrophic to 64bit guests, is far worse than the EFER.NXE > leakage which was trying to be fixed. > > Introduce cpu_bug_msr_ll_efer_sce to encapsulate this partial revert. > Avoid adding MSR_EFER to the Load Lists on impacted hardware, and > reintroduce the logic to use the guests EFER.SCE setting. > > In the common case of running 64bit HVM guests, these extra adjustments > to EFER should only be hit during guest boot. > > Signed-off-by: Andrew Cooper <[email protected]> > --- > CC: Jan Beulich <[email protected]> > CC: Wei Liu <[email protected]> > CC: Roger Pau Monné <[email protected]> > CC: Jun Nakajima <[email protected]> > CC: Kevin Tian <[email protected]> > > This is RFC at the moment, because the test lab is full at the moment > and I don't have a Harpertown CPU to hand. I'm fairly sure the change > is complete and will test when it becomes available, but I don't expect > to make any code changes.
Sadly testing says no. I'll try and figure out what is going on here. ~Andrew _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
