Re: [PATCH] x86/asm: Replace __force_order with memory clobber

[Sedat Dilek]

On Tue, Aug 25, 2020 at 5:19 PM Arvind Sankar <nivedita@xxxxxxxxxxxx> wrote:
>
> On Mon, Aug 24, 2020 at 09:13:34PM +0200, Miguel Ojeda wrote:
> > Hi Arvind,
> >
> > On Sun, Aug 23, 2020 at 11:25 PM Arvind Sankar <nivedita@xxxxxxxxxxxx> wrote:
> > >
> > > - Using a dummy input operand with an arbitrary constant address for the
> > >   read functions, instead of a global variable. This will prevent reads
> > >   from being reordered across writes, while allowing memory loads to be
> > >   cached/reordered across CRn reads, which should be safe.
> >
> > Assuming no surprises from compilers, this looks better than dealing
> > with different code for each compiler.
> >
> > > Signed-off-by: Arvind Sankar <nivedita@xxxxxxxxxxxx>
> > > Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82602
> >
> > A lore link to the other discussion would be nice here for context.
> >
>
> Ok.
>
> > > + * The compiler should not reorder volatile asm, however older versions of GCC
> > > + * had a bug (which was fixed in 8.1, 7.3 and 6.5) where they could sometimes
> >
> > I'd mention the state of GCC 5 here.
> >
>
> Ok.
>
> > > + * reorder volatile asm. The write functions are not a problem since they have
> > > + * memory clobbers preventing reordering. To prevent reads from being reordered
> > > + * with respect to writes, use a dummy memory operand.
> > >   */
> > > -extern unsigned long __force_order;
> > > +
> >
> > Spurious newline?
> >
>
> This was intentional, but I can remove it if people don't like the extra
> whitespace.
>
> I'll wait a few days for additional review comments before sending v2.
>
> Thanks.

Thanks for taking care and your patch.

I have tested this on top of Linux v5.9-rc2 with LLVM toolchain
v11.0.0-rc2 (ThinLTO).

Tested-by: Sedat Dilek <sedat.dilek@xxxxxxxxx>

- Sedat -