On Fri, 2 Aug 2019 at 06:14, Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> wrote:
>
> Hi Stephen:
>
> On Fri, Aug 02, 2019 at 10:20:19AM +1000, Stephen Rothwell wrote:
> >
> > It might be time to revert all this series and try again. The
> > implementation seems to have not been well thought through from a kernel
> > building point of view. For a start the two commits
> >
> > 7cdc0ddbf74a ("crypto: aegis128 - add support for SIMD acceleration")
> > ecc8bc81f2fb ("crypto: aegis128 - provide a SIMD implementation based on NEON intrinsics")
>
> I think the idea was that it would get optimised out if the
> implementation is absent which is why it was meant to work in
> this order. But oviously as we have found out this didn't work.
>
> Ard, I think relying on the compiler to optimise something out based
> on an assignment within an if statement is just too error-prone.
> We'll need a different mechanism for this.
>
Indeed. This is definitely something I tested, and it appears to be
dependent on the GCC version.
> For now I'm going to back out those two specific patches as the
> rest seem to be valid by themselves.
>
OK. I will adopt this mechanism [0] after all and resubmit, once I get
confirmation from either Voldis or Heiko that this makes the issue go
away (given that my local GCC does not reproduce the issue)
[0] https://lore.kernel.org/linux-crypto/20190729074434.21064-1-ard.biesheuvel@xxxxxxxxxx/
