Re: [RFC PATCH] x86/arch_prctl: Add ARCH_SET_XCR0 to mask XCR0 per-thread

[Dave Hansen]

On 06/18/2018 10:22 AM, Keno Fischer wrote:
>> No, I'm saying that depending on faults is not a viable solution.  We
>> are not guaranteed to get faults in all the cases you would need to fix up.
>>
>> XSAVE*/XRSTOR* are not even *called* in some of those cases.
> Ah, my apologies, I was under the mistaken impression that xsaves also
> read xcomp_bv to inform the layout, rather than using the RFBM and then
> updating the xcomp_bv field. Let me think about this some more and see
> what I can come up with.

It's not even that.  If you do this, xstate_comp_offsets[] will
potentially need to be recalculated for each XCR0 value.  Granted, we've
pretty nicely consolidated these references in the kernel into very few
spots, but you still need to fix that up.