Re: [PATCH 08/12] De-PLTize __stack_chk_fail internal calls within libc.so.

[Florian Weimer <fweimer at redhat dot com>]

On 12/15/2016 06:24 PM, Nix wrote:
> On 15 Dec 2016, nix@esperi.org.uk verbalised:
>
> > diff --git a/sysdeps/generic/symbol-hacks.h b/sysdeps/generic/symbol-hacks.h
> > index 36908b5..15ff56a 100644
> > --- a/sysdeps/generic/symbol-hacks.h
> > +++ b/sysdeps/generic/symbol-hacks.h
> > @@ -7,5 +7,7 @@ asm ("memcpy = __GI_memcpy");
> >
> >  /* -fstack-protector generates calls to __stack_chk_fail, which need
> >     similar adjustments to avoid going through the PLT.  */
> > +# if defined STACK_PROTECTOR_LEVEL && STACK_PROTECTOR_LEVEL > 0
> >  asm ("__stack_chk_fail = __stack_chk_fail_local");
> > +# endif
> >  #endif
>
> This causes (minor) problems on SPARC:
>
> Extra PLT reference: libc.so: __stack_chk_fail
>
> Whether we can disregard this, I don't know, but it does feel wrong.

Well, avoiding this is the point of __stack_chk_fail_local, isn't it? 
So we surely can't ignore it.

I think what is going on is that once a symbol has a hidden anywhere in 
a static link, all references to it are turned hidden.  Previously, this 
hidden reference occurred in the file which *defined* 
__stack_chk_fail_local, but is now gone from there.  (At the assembler 
level, the .hidden directive is markedly different from the GCC 
visibility attribute.)

Could you try this?

# if defined STACK_PROTECTOR_LEVEL && STACK_PROTECTOR_LEVEL > 0
asm (".hidden __stack_chk_fail_local");
asm ("__stack_chk_fail = __stack_chk_fail_local");
# endif

Thanks,
Florian