Re: [PATCH] stack: replace "o" output with "r" input constraint

[Kees Cook]

On Wed, May 05, 2021 at 02:12:32PM -0700, Nathan Chancellor wrote:
> On Thu, Apr 22, 2021 at 06:02:27PM -0700, Nathan Chancellor wrote:
> > On Mon, Apr 19, 2021 at 04:17:41PM -0700, Kees Cook wrote:
> > > From: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> > > 
> > > "o" isn't a common asm() constraint to use; it triggers an assertion in
> > > assert-enabled builds of LLVM that it's not recognized when targeting
> > > aarch64 (though it appears to fall back to "m"). I've fixed this in LLVM
> > > 13 now, but there isn't really a good reason to be using "o" in particular
> > > here. To avoid causing build issues for those using assert-enabled builds
> > > of earlier LLVM versions, the constraint needs changing.
> > > 
> > > Instead, if the point is to retain the __builtin_alloca(), we can make ptr
> > > appear to "escape" via being an input to an empty inline asm block. This
> > > is preferable anyways, since otherwise this looks like a dead store.
> > > 
> > > While the use of "r" was considered in
> > > https://lore.kernel.org/lkml/202104011447.2E7F543@keescook/
> > > it was only tested as an output (which looks like a dead store, and
> > > wasn't sufficient). Use "r" as an input constraint instead, which
> > > behaves correctly across compilers and architectures:
> > > https://godbolt.org/z/E9cd411ob
> > > 
> > > Link: https://reviews.llvm.org/D100412
> > > Link: https://bugs.llvm.org/show_bug.cgi?id=49956
> > > Signed-off-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
> > > Tested-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > > Fixes: 39218ff4c625 ("stack: Optionally randomize kernel stack offset each syscall")
> 
> Kees, were you planning on taking this to Linus or someone else? It
> would be nice to have this in for -rc1 (although I understand it might
> be too late), if not, by -rc2.

I assumed Thomas would pick this up. Thomas, shall I send this directly
to Linus?

Thanks!

-Kees

> 
> Cheers,
> Nathan
> 
> > > Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > 
> > I built arm64 defconfig with and without
> > CONFIG_RANDOMIZE_KSTACK_OFFSET_DEFAULT with LLVM 12 (which does not have
> > Nick's LLVM fix) without any issues and did a quick boot test in QEMU,
> > nothing exploded.
> > 
> > Reviewed-by: Nathan Chancellor <nathan@xxxxxxxxxx>
> > Tested-by: Nathan Chancellor <nathan@xxxxxxxxxx>
> > 
> > > ---
> > >  include/linux/randomize_kstack.h | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/include/linux/randomize_kstack.h b/include/linux/randomize_kstack.h
> > > index fd80fab663a9..bebc911161b6 100644
> > > --- a/include/linux/randomize_kstack.h
> > > +++ b/include/linux/randomize_kstack.h
> > > @@ -38,7 +38,7 @@ void *__builtin_alloca(size_t size);
> > >  		u32 offset = raw_cpu_read(kstack_offset);		\
> > >  		u8 *ptr = __builtin_alloca(KSTACK_OFFSET_MAX(offset));	\
> > >  		/* Keep allocation even after "ptr" loses scope. */	\
> > > -		asm volatile("" : "=o"(*ptr) :: "memory");		\
> > > +		asm volatile("" :: "r"(ptr) : "memory");		\
> > >  	}								\
> > >  } while (0)
> > >  
> > > -- 
> > > 2.25.1
> > > 

-- 
Kees Cook