Re: [PATCH 04/10] x86/fpu: Document reasoning of FX-only fallback

On Tue, Jun 30, 2026 at 12:23=E2=80=AFPM Chang S. Bae <chang.seok.bae@intel=
.com> wrote:
>
> On 6/15/2026 12:37 PM, Andrei Vagin wrote:
> >
> > This fallback is dangerous as it can trigger silent corruptions of
> > user-space state by resetting extended registers if the process was
> > using them but the frame metadata was malformed.
> >
> > XSAVE was introduced 15 years ago, we may need to consider removing thi=
s
> > fallback entirely or introducing a sysctl to enable/disable it.
>
> This reads more like a separate problem statement. I don't see any
> follow-up implementing either option here.
>
> Without additional context, it isn't entirely clear to me why this
> should be in this changelog at least.

Fair point, Chang. I included it as a note on the risks of the legacy
fallback, based on my experience debugging various issues where this
fallback masked bugs. We may leave this context for anyone else
who might be working on this code in the future. If you think it should not
be here, I will remote it.

Thanks,
Andrei