Re: [PATCH v4 8/8] x86/fpu/xstate: Warn if guest-only supervisor states are detected in normal fpstate

[Dave Hansen]

On 4/3/25 17:02, Chang S. Bae wrote:
> On 4/2/2025 7:30 AM, Chao Gao wrote:
>> The goal is to ensure that guest-only _supervisor_ features are
>> not enabled in non-guest FPUs.
> 
> I think the common XSAVE path matters here. The other XSAVE paths — 
> signal delivery and saving the LBR state — already handle supervisor 
> states properly. Signal delivery excludes all of supervisor states
> from the RFBM, while LBR state saving triggers a warning if any non-
> LBR state is set in the RFBM. Given this, the guard seems good
> enough unless missing something.
This patch is looking more and more optional. There's no broad consensus
on the need for a warning or exactly what it should warn on.

I think it should be dropped.