Re: [patch V2 02/52] x86/fpu: Fix copy_xstate_to_kernel() gap handling

[Borislav Petkov]

On Tue, Jun 15, 2021 at 02:47:14PM +0200, Thomas Gleixner wrote:
> I'll never learn that. /me goes to write some elisp.

You could also say, this is a way to keep your reviewers awake. :-P

> Nah. The wonder of membug_write() is that it does not write behind the

membug - Freudian slip, yeah, I know which is the word you've been
writing the most, lately. :-P

> end of the buffer which is designed to allow partial reads w/o checking
> a gazillion times for return values etc.

Yeah, right. I'm just being overly paranoid here, as most of the time.

> The point is that this gives us the proper init.mxcsr value when SSE and
> YMM are not set.

Oh sure - what I mean is, this could be a simple assignment into those
mxcsr and mxcsr_mask things but if you add the whole conditional code
around it, it'll become hard to read too and it will be the only copy
which doesn't call copy_feature() and that would throw off people
looking for the same pattern of calling copy_feature() in that whole
function.

And the destination @to would need casting...

Fget about it. :-)

Thx.

-- 
Regards/Gruss,
    Boris.

SUSE Software Solutions Germany GmbH, GF: Felix Imendörffer, HRB 36809, AG Nürnberg