x86/asm: Fix inline asm call constraints for Clang
authorJosh Poimboeuf <jpoimboe@redhat.com>
Wed, 20 Sep 2017 21:24:33 +0000 (16:24 -0500)
committerIngo Molnar <mingo@kernel.org>
Sat, 23 Sep 2017 13:06:20 +0000 (15:06 +0200)
commitf5caf621ee357279e759c0911daf6d55c7d36f03
tree429270fc4ccb2c9fe7881229640858c53dfca5e2
parent0d0970eef3b03ef08b19da5bc3044410731cf38f
x86/asm: Fix inline asm call constraints for Clang

For inline asm statements which have a CALL instruction, we list the
stack pointer as a constraint to convince GCC to ensure the frame
pointer is set up first:

  static inline void foo()
  {
register void *__sp asm(_ASM_SP);
asm("call bar" : "+r" (__sp))
  }

Unfortunately, that pattern causes Clang to corrupt the stack pointer.

The fix is easy: convert the stack pointer register variable to a global
variable.

It should be noted that the end result is different based on the GCC
version.  With GCC 6.4, this patch has exactly the same result as
before:

defconfig defconfig-nofp distro distro-nofp
 before 9820389 9491555 8816046 8516940
 after 9820389 9491555 8816046 8516940

With GCC 7.2, however, GCC's behavior has changed.  It now changes its
behavior based on the conversion of the register variable to a global.
That somehow convinces it to *always* set up the frame pointer before
inserting *any* inline asm.  (Therefore, listing the variable as an
output constraint is a no-op and is no longer necessary.)  It's a bit
overkill, but the performance impact should be negligible.  And in fact,
there's a nice improvement with frame pointers disabled:

defconfig defconfig-nofp distro distro-nofp
 before 9796316 9468236 9076191 8790305
 after 9796957 9464267 9076381 8785949

So in summary, while listing the stack pointer as an output constraint
is no longer necessary for newer versions of GCC, it's still needed for
older versions.

Suggested-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reported-by: Matthias Kaehlcke <mka@chromium.org>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Dmitriy Vyukov <dvyukov@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Miguel Bernal Marin <miguel.bernal.marin@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/3db862e970c432ae823cf515c52b54fec8270e0e.1505942196.git.jpoimboe@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
13 files changed:
arch/x86/include/asm/alternative.h
arch/x86/include/asm/asm.h
arch/x86/include/asm/mshyperv.h
arch/x86/include/asm/paravirt_types.h
arch/x86/include/asm/preempt.h
arch/x86/include/asm/processor.h
arch/x86/include/asm/rwsem.h
arch/x86/include/asm/uaccess.h
arch/x86/include/asm/xen/hypercall.h
arch/x86/kvm/emulate.c
arch/x86/kvm/vmx.c
arch/x86/mm/fault.c
tools/objtool/Documentation/stack-validation.txt