On 10/10/19 16:42, Aaron Lewis wrote: >>> Hoist support for IA32_XSS so it can be used for both AMD and Intel, >>> instead of for just Intel. >>> >>> AMD has no equivalent of Intel's "Enable XSAVES/XRSTORS" VM-execution >>> control. Instead, XSAVES is always available to the guest when supported >>> on the host. >> >> You could add that implement the XSAVES check based on host's features >> and move the MSR_IA32_XSS msr R/W from Intel only code to the common >> code. > > Isn't this covered by my comments? I mention that we are hoisting > IA32_XSS to common code in the first comment, then in the second > comment I say that XSAVES is available in the guest when supported on > the host. Yes, I agree. Perhaps you can add something like "Fortunately, right now Linux does not use MSR_IA32_XSS so the guest's usage of XSAVES will be a glorified XSAVEC, and cannot bypass vmexits for MSR loads and stores". Paolo
