Re: [PATCH] Make math_state_restore() save and restore the interrupt flag

[Linus Torvalds]

On Sat, Feb 1, 2014 at 5:57 PM, H. Peter Anvin <hpa@xxxxxxxxx> wrote:
>
> Twiddling CR0.TS is pretty slow if we're not taking advantage of it.

Immaterial.

We *already* avoid twiddling TS if it's not needed.

It is true that we used to twiddle it at every context switch (and
then twiddle it *again* if we decided that we'd want to pre-load the
FPU state anyway, and avoid the extra fault).

But that was fixed, and if we switch from a task that had math state
to another task that has math state, we leave TS alone.

But Suresh apparently hits on the real issue:

> not all the state under xsave adhers to cr0.TS/DNA rules

which if so is sad but yes, makes CR0.TS no longer sufficient.

             Linus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/