[PATCH 0/5 v16] KASan for Arm

[Ard Biesheuvel]

On Thu, 5 Nov 2020 at 01:38, Nick Desaulniers <ndesaulniers at google.com> wrote:
>
> On Wed, Nov 4, 2020 at 4:30 PM Fāng-ruì Sòng <maskray at google.com> wrote:
> >
> > About "ARM: Replace string mem* functions for KASan", it added .weak
> > memcpy to arch/arm/lib/memcpy.S
> > However, the convention is to change .weak + ENTRY(...) to WEAK(...)
> >
> > In particular, clang 12 will reject `.weak memcpy; .global memcpy`
> > (they had different semantics in GNU as and LLVM integrated assembler
> > for a long time AND binutils maintainer agreed that .weak and .global
> > are strange and should not be used - though binutils will not make the
> > change to cause disruption)
> >
> > Please see https://github.com/ClangBuiltLinux/linux/issues/1190
>
> It's straightforward for us to fixup; Fangrui has sent patches for
> x86, perf, and 64b arm already which have been accepted.  I forget
> though what is the convention for "Fixes:" tags for patches in -next?
> The SHAs change when pulled, right?

No, only when rebasing. It depends on the maintainer how likely that
is to happen for changes that are in -next, but usually the commit IDs
are stable once pulled into -next.