[PATCH] Augment __naked__ function attribute with __noinline__ and __noclone__

[Mikael Pettersson]

Mikael Pettersson writes:
 > The asm() bodies of naked functions often depend on standard calling
 > conventions, therefore they must be noinline and noclone.  GCC 4.[56]
 > currently fail to enforce this, see GCC PR44290.  The effect is that
 > most arch/arm/mm/copypage-*.c modules get miscompiled, resulting in
 > kernel crashes during bootup.
 > 
 > Part of the fix is to augment the __naked attribute to imply noinline
 > and noclone.  This patch implements that, and has been verified to fix
 > the failure to boot a gcc-4.5 compiled 2.6.34 kernel on a Thecus n2100
 > (XScale-based mach-iop32x).

Now also tested with a gcc-4.5 compiled 2.6.35-rc1 kernel on a ds101
(XScale-based mach-ixp4xx).

 >
 > Signed-off-by: Mikael Pettersson <mikpe at it.uu.se>
 > Signed-off-by: Khem Raj <raj.khem at gmail.com>
 > ---
 >  include/linux/compiler-gcc.h  |   11 ++++++++++-
 >  include/linux/compiler-gcc4.h |    4 ++++
 >  2 files changed, 14 insertions(+), 1 deletions(-)
 > 
 > diff --git a/include/linux/compiler-gcc.h b/include/linux/compiler-gcc.h
 > index 73dcf80..83bd87c 100644
 > --- a/include/linux/compiler-gcc.h
 > +++ b/include/linux/compiler-gcc.h
 > @@ -58,8 +58,12 @@
 >   * naked functions because then mcount is called without stack and frame pointer
 >   * being set up and there is no chance to restore the lr register to the value
 >   * before mcount was called.
 > + *
 > + * The asm() bodies of naked functions often depend on standard calling conventions,
 > + * therefore they must be noinline and noclone.  GCC 4.[56] currently fail to enforce
 > + * this, so we must do so ourselves.  See GCC PR44290.
 >   */
 > -#define __naked				__attribute__((naked)) notrace
 > +#define __naked				__attribute__((naked)) noinline __noclone notrace
 >  
 >  #define __noreturn			__attribute__((noreturn))
 >  
 > @@ -85,3 +89,8 @@
 >  #define _gcc_header(x) __gcc_header(linux/compiler-gcc##x.h)
 >  #define gcc_header(x) _gcc_header(x)
 >  #include gcc_header(__GNUC__)
 > +
 > +#if !defined(__noclone)
 > +#define __noclone	/* not needed */
 > +#endif
 > +
 > diff --git a/include/linux/compiler-gcc4.h b/include/linux/compiler-gcc4.h
 > index 94dea3f..fcfa5b9 100644
 > --- a/include/linux/compiler-gcc4.h
 > +++ b/include/linux/compiler-gcc4.h
 > @@ -48,6 +48,10 @@
 >   * unreleased.  Really, we need to have autoconf for the kernel.
 >   */
 >  #define unreachable() __builtin_unreachable()
 > +
 > +/* Mark a function definition as prohibited from being cloned. */
 > +#define __noclone	__attribute__((__noclone__))
 > +
 >  #endif
 >  
 >  #endif
 > -- 
 > 1.7.0.4
 >