On Mon, Jan 07, 2019 at 05:54:27PM +0000, Will Deacon wrote:

On Wed, Jan 02, 2019 at 03:57:54PM -0500, Michael S. Tsirkin wrote:

quoted

We don't really care whether the variable is in-register
or in-memory. Relax the constraint accordingly.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
---
 include/linux/compiler.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/compiler.h b/include/linux/compiler.h
index 1ad367b4cd8d..6601d39e8c48 100644
--- a/include/linux/compiler.h
+++ b/include/linux/compiler.h

@@ -154,7 +154,7 @@ void ftrace_likely_update(struct ftrace_likely_data *f, int val,
 #ifndef OPTIMIZER_HIDE_VAR
 /* Make the optimizer believe the variable can be manipulated arbitrarily. */
 #define OPTIMIZER_HIDE_VAR(var)						\
-	__asm__ ("" : "=r" (var) : "0" (var))
+	__asm__ ("" : "=rm" (var) : "0" (var))
 #endif

I think this can break for architectures with write-back addressing modes
such as arm, where the "m" constraint is assumed to be evaluated precisely
once in the asm block.

Will

Thanks, I'll drop this patch.