Re: [cryptodev:master 130/134] aes_generic.c:undefined reference to `_restgpr_31_x'

[Segher Boessenkool <segher@kernel.crashing.org>]

On Fri, Jan 12, 2018 at 08:43:21PM +0100, Arnd Bergmann wrote:
> On Fri, Jan 12, 2018 at 5:39 PM, Segher Boessenkool
> <segher@kernel.crashing.org> wrote:
> 
> >> or why the aes_generic implementation needs this on
> >> powerpc when built with 'gcc -Os'. FWIW, the -Os change was needed
> >> to work around a possible kernel stack overflow that can happen with
> >> gcc-7.2, see https://patchwork.kernel.org/patch/10143607/ and
> >> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83356
> >
> > The _x versions are smaller but slower; that's why they are used with -Os.
> > Apparently nothing else was built with -Os (and the other needed flags)
> > before.
> 
> Ah, that explains it, the definition is in arch/powerpc/lib/crtsavres.S,
> but inside of #ifdef CONFIG_CC_OPTIMIZE_FOR_SIZE.

Ah ok.  Right.

> We could theoretically work around it by turning that into
> "#if defined(CONFIG_CC_OPTIMIZE_FOR_SIZE) ||
> defined(CONFIG_CRYPTO_AES)", but that seems rather ugly.
> 
> My earlier patch already tried to be more specific, turning very
> specific optimizations off rather than moving from -O2 to -Os,
> but that turned out to lead to significantly worse performance,
> where -Os improved performance slightly. Is there a way
> to ask powerpc compilers to use mostly -Os but not the
> specific thing that makes it link to _restgpr_31_x?

There is no such thing, sorry.  Would be very hard to implement, and
older compilers will never get it, so it won't help you anyway :-(

Maybe for now just enable it in crtsavres.S always, with a comment?
That -Os workaround is hopefully not going to live long either...


Segher