V2 [PATCH] x86: Support GNU_PROPERTY_X86_ISA_1_V[234] marker:

H.J. Lu hjl.tools@gmail.com
Wed Oct 28 13:36:13 GMT 2020


On Wed, Oct 28, 2020 at 4:09 AM Florian Weimer <fweimer@redhat.com> wrote:
>
> * H. J. Lu:
>
> > On Tue, Oct 13, 2020 at 4:29 AM Florian Weimer <fweimer@redhat.com> wrote:
> >>
> >> * H. J. Lu:
> >>
> >> > On Tue, Oct 13, 2020 at 3:48 AM Florian Weimer <fweimer@redhat.com> wrote:
> >> >>
> >> >> * H. J. Lu via Binutils:
> >> >>
> >> >> > Here is the updated patch with -z x86-64-v[234] linker command line
> >> >> > option to mark x86-64-v[234] ISA level as needed.
> >> >>
> >> >> I think this needs a marker for the baseline as well.  Currently, a
> >> >> missing marker can either mean “the programmer did not specify a marker”
> >> >> or “the object uses the baseline ABI”.
> >> >>
> >> >
> >> > “the object uses the baseline ABI” and “the programmer did not specify
> >> > a marker” mean the same thing to ld.so since a baseline or
> >> > x86-64-v[234] objects may contain ANY instructions.
> >>
> >> ld.so won't be the sole consumer of such annotations.  For example, an
> >> install-like tool could use the annotations to select the appropriate
> >> installation path for an ELF shared object.
> >>
> >
> > They should treat “the object uses the baseline ABI” the same way as
> > “the programmer did not specify a marker” since the ISA marker specifies
> > the minimum ISA level, not the maximum.
>
> I don't think this will work for us in our downstream binary analysis,
> sorry.  Please add a marker for the baseline ABI.  Thanks.

Done:

https://gitlab.com/x86-psABIs/x86-64-ABI/-/merge_requests/13
https://gitlab.com/x86-psABIs/x86-64-ABI/-/wikis/Micro-architecture-levels

-- 
H.J.


More information about the Binutils mailing list