On 02/06/2023 5:08 pm, Alejandro Vallejo wrote: > On Fri, Jun 02, 2023 at 03:22:20PM +0100, Andrew Cooper wrote: >> Linux deals with this in verify_cpu() (early asm) along with a FMS check >> protecting the access to MSR_MISC_ENABLE, rather than using rdmsr_safe() >> and catching the #GP. > On a related note, we don't use rdmsr_safe() either. We just hope it exists > on any Intel CPU. It fortunately does on any Intel CPU we care about > because it was introduced shortly before Pentium 4 (Netburst), so we're > fine since we mandate long mode.
Oh, good point. Yeah, that's fine, but only try reading it in the case that we've found LM, not NX, and GenuineIntel. There are old versions of Xen which don't emulate the MSR at all, and the only reason Xen does emulate it in all guests is for a CPUID-faulting corner case. The same assumptions are unlikely to hold for other virtualised cases. Failing with a clear "NX not available" is strictly preferable to triple faulting. ~Andrew
