Re: [PATCH 02/24] x86/traps: Move arch/x86/kernel/traps.c to arch/x86/entry/

* Re: [PATCH 02/24] x86/traps: Move arch/x86/kernel/traps.c to arch/x86/entry/
       [not found]             ` <YTDS/p/bnsTFzNES@hirez.programming.kicks-ass.net>
@ 2021-09-02 17:05               ` Nick Desaulniers
  2021-09-02 17:19                 ` Miguel Ojeda
  2021-09-03  7:36                 ` Martin Liška
  0 siblings, 2 replies; 6+ messages in thread
From: Nick Desaulniers @ 2021-09-02 17:05 UTC (permalink / raw)
  To: Peter Zijlstra, Miguel Ojeda, Martin Liška
  Cc: Lai Jiangshan, Joerg Roedel, Lai Jiangshan, linux-kernel,
	Andy Lutomirski, Thomas Gleixner, Ingo Molnar, Borislav Petkov,
	x86, H. Peter Anvin, Daniel Bristot de Oliveira, Brijesh Singh,
	Andy Shevchenko, Arvind Sankar, Chester Lin, Juergen Gross,
	andrew.cooper3, linux-toolchains

On Thu, Sep 2, 2021 at 6:36 AM Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Thu, Sep 02, 2021 at 02:05:41PM +0200, Peter Zijlstra wrote:
> > On Thu, Sep 02, 2021 at 07:54:25PM +0800, Lai Jiangshan wrote:
> > > For example, stack-protector is instrumenting many noninstr functions now
> > > if the CONFIG is yes.  It is normally Ok and gcc is adding per-function control
> > > on it.
> >
> > IIRC the latest compiler have an attribute for this too, that really
> > should be added to noinstr. Lemme go find.
>
> Something like so... Nick ?
>
> diff --git a/include/linux/compiler-clang.h b/include/linux/compiler-clang.h
> index 49b0ac8b6fd3..6a15c3d8df5c 100644
> --- a/include/linux/compiler-clang.h
> +++ b/include/linux/compiler-clang.h
> @@ -62,6 +62,12 @@
>  #define __no_sanitize_coverage
>  #endif
>
> +#if defined(CONFIG_STACKPROTECTOR)
> +#define __no_stack_protector __attribute__((no_stack_protector))
> +#else
> +#define __no_stack_protector
> +#endif
> +
>  /*
>   * Not all versions of clang implement the type-generic versions
>   * of the builtin overflow checkers. Fortunately, clang implements
> diff --git a/include/linux/compiler-gcc.h b/include/linux/compiler-gcc.h
> index cb9217fc60af..7ac0a3f11ba9 100644
> --- a/include/linux/compiler-gcc.h
> +++ b/include/linux/compiler-gcc.h
> @@ -128,6 +128,12 @@
>  #define __no_sanitize_coverage
>  #endif
>
> +#if defined(CONFIG_STACKPROTECTOR) && __has_attribute__(__no_stack_protector__)
> +#define __no_stack_protector __attribute__((no_stack_protector))
> +#else
> +#define __no_stack_protector
> +#endif
> +

The above 2 hunks should go in include/linux/compiler_attributes.h,
but yes.  I'd been meaning to send such a patch; it's nice to have
finer grain function-level control over -fno-stack-protector which
significantly hurts ergonomics for things like LTO.  IIRC GCC only
added the attribute recently in the 10.X release, so it might be too
new to rely on quite yet.

>  #if GCC_VERSION >= 50100
>  #define COMPILER_HAS_GENERIC_BUILTIN_OVERFLOW 1
>  #endif
> diff --git a/include/linux/compiler_types.h b/include/linux/compiler_types.h
> index e4ea86fc584d..5ae1c08dba8d 100644
> --- a/include/linux/compiler_types.h
> +++ b/include/linux/compiler_types.h
> @@ -210,7 +210,8 @@ struct ftrace_likely_data {
>  /* Section for code which can't be instrumented at all */
>  #define noinstr                                                                \
>         noinline notrace __attribute((__section__(".noinstr.text")))    \
> -       __no_kcsan __no_sanitize_address __no_profile __no_sanitize_coverage
> +       __no_kcsan __no_sanitize_address __no_profile __no_sanitize_coverage \
> +       __no_stack_protector
>
>  #endif /* __KERNEL__ */
>

-- 
Thanks,
~Nick Desaulniers

^ permalink raw reply	[flat|nested] 6+ messages in thread