Re: [PATCH v2 3/3] Kconfig: add ARCH_WANTS_NO_INSTR+CC_HAS_NO_PROFILE_FN_ATTR, depend on for GCOV and PGO

[Marco Elver]

On Tue, 22 Jun 2021 at 09:26, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
>
> On Mon, Jun 21, 2021 at 04:18:22PM -0700, Nick Desaulniers wrote:
> > We don't want compiler instrumentation to touch noinstr functions, which
> > are annotated with the no_profile_instrument_function function
> > attribute. Add a Kconfig test for this and make PGO and GCOV depend on
> > it.
> >
> > If an architecture is using noinstr, it should denote that via this
> > Kconfig value. That makes Kconfigs that depend on noinstr able to
> > express dependencies in an architecturally agnostic way.
> >
> > Cc: Masahiro Yamada <masahiroy@xxxxxxxxxx>
> > Cc: Peter Oberparleiter <oberpar@xxxxxxxxxxxxx>
> > Link: https://lore.kernel.org/lkml/YMTn9yjuemKFLbws@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/
> > Link: https://lore.kernel.org/lkml/YMcssV%2Fn5IBGv4f0@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx/
> > Suggested-by: Nathan Chancellor <nathan@xxxxxxxxxx>
> > Suggested-by: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> > Signed-off-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> > ---
> > Changes V1 -> V2:
> > * Add ARCH_WANTS_NO_INSTR
> > * Change depdendencies to be !ARCH_WANTS_NO_INSTR || CC_HAS_NO_PROFILE_FN_ATTR
> >   rather than list architectures explicitly, as per Nathan.
> > * s/no_profile/no_profile_instrument_function/
> >
> >  arch/Kconfig        | 7 +++++++
> >  arch/arm64/Kconfig  | 1 +
> >  arch/s390/Kconfig   | 1 +
> >  arch/x86/Kconfig    | 1 +
> >  init/Kconfig        | 3 +++
> >  kernel/gcov/Kconfig | 1 +
> >  kernel/pgo/Kconfig  | 3 ++-
> >  7 files changed, 16 insertions(+), 1 deletion(-)
> >
> > diff --git a/arch/Kconfig b/arch/Kconfig
> > index 2b4109b0edee..2113c6b3b801 100644
> > --- a/arch/Kconfig
> > +++ b/arch/Kconfig
> > @@ -285,6 +285,13 @@ config ARCH_THREAD_STACK_ALLOCATOR
> >  config ARCH_WANTS_DYNAMIC_TASK_STRUCT
> >       bool
> >
> > +config ARCH_WANTS_NO_INSTR
> > +     bool
> > +     help
> > +       An architecure should select this if the noinstr macro is being used on
> > +       functions to denote that the toolchain should avoid instrumenting such
> > +       functions and is required for correctness.
> > +
> >  config ARCH_32BIT_OFF_T
> >       bool
> >       depends on !64BIT
>
> There's also CC_HAS_WORKING_NOSANITIZE_ADDRESS in lib/Kconfig.kasan that
> might want to be hooked into this, but that can be done separately I
> suppose.

KASAN already depends on this for all compiler instrumentation modes,
not just for 'noinstr' but also to avoid false positives. So it's not
just for noinstr's benefit, and we should not weaken the requirement
there.