On Sun, May 15 2022 at 10:25, Thomas Gleixner wrote:
> On Fri, May 13 2022 at 09:52, Linus Torvalds wrote:
> The code which copies the ptrace supplied state has a pile of sanity
> checks to catch invalid state, but I wouldn't bet my hat on it that it's
> 100% complete. We can be more defensive here, but I would be surprised.
>
> Something like the untested below. I'll expose it to some testing to see
> what explodes.

I was not able to get something broken past the existing sanity checks
which was caught by the trial restore. I seems the FPU units are pretty
resilent against random nonsense except for the XSTATE header and the
MXCRS content. Though I ponder to just add that trial restore due to
paranoia.

Thanks,

        tglx