---
 arch/x86/include/asm/percpu.h |   66 ++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 64 insertions(+), 2 deletions(-)

Index: linux-2.6/arch/x86/include/asm/percpu.h
===================================================================
--- linux-2.6.orig/arch/x86/include/asm/percpu.h	2009-09-16 16:21:14.000000000 -0500
+++ linux-2.6/arch/x86/include/asm/percpu.h	2009-09-16 17:59:32.000000000 -0500
@@ -153,8 +153,70 @@ do {							\
 #define percpu_or(var, val)	percpu_to_op("or", per_cpu__##var, val)
 #define percpu_xor(var, val)	percpu_to_op("xor", per_cpu__##var, val)
 
-#define __this_cpu_read(pcp)		percpu_from_op("mov", pcp)
-#define __this_cpu_write(pcp, val)	percpu_to_op("mov", (pcp), val)
+#define __percpu_to_op(op, var, val)			\
+do {							\
+	typedef typeof(var) T__;			\
+	if (0) {					\
+		T__ tmp__;				\
+		tmp__ = (val);				\
+	}						\
+	asm(op " %1,"__percpu_arg(0)			\
+		    : "+m" (var)			\
+		    : "ri" ((T__)(val)));		\
+} while (0)
+
+#define __percpu_from_op(op, var, constraint)		\
+({							\
+	typeof(var) ret__;				\
+	asm(op " " __percpu_arg(1) ",%0"		\
+		    : "=r" (ret__)			\
+		    : constraint);			\
+	ret__;						\
+})
+
+#define __this_cpu_read_1(pcp)	__percpu_from_op("movb", (pcp), "m" (pcp))
+#define __this_cpu_read_2(pcp)	__percpu_from_op("movw", (pcp), "m" (pcp))
+#define __this_cpu_read_4(pcp)	__percpu_from_op("movl", (pcp), "m" (pcp))
+#define __this_cpu_read_8(pcp)	__percpu_from_op("movq", (pcp), "m" (pcp))
+// #define __this_cpu_read_q(pcp)	(*__this_cpu_ptr(&(pcp)))
+
+#define __this_cpu_write_1(pcp, val)	__percpu_to_op("movb", (pcp), val)
+#define __this_cpu_write_2(pcp, val)	__percpu_to_op("movw", (pcp), val)
+#define __this_cpu_write_4(pcp, val)	__percpu_to_op("movl", (pcp), val)
+#define __this_cpu_write_8(pcp, val)	__percpu_to_op("movq", (pcp), val)
+
+#define __this_branch_read(identifier, a)		\
+({							\
+	typeof(a) ret__;				\
+	switch(sizeof(a)) {				\
+	case 1: ret__ = identifier##1(a);break;		\
+	case 2: ret__ = identifier##2(a);break;		\
+	case 4: ret__ = identifier##4(a);break;		\
+	case 8: ret__ = identifier##8(a);break;		\
+	default:					\
+		__bad_percpu_size();			\
+		ret__ = 0;				\
+	};						\
+	ret__;						\
+})
+
+#define __this_branch_modify(identifier, a, value)	\
+do {							\
+	switch(sizeof(a)) {				\
+	case 1: identifier##1(a, value);break;		\
+	case 2: identifier##2(a, value);break;		\
+	case 4: identifier##4(a, value);break;		\
+	case 8: identifier##8(a, value);break;		\
+	default:					\
+		__bad_percpu_size();			\
+	}						\
+} while (0)
+
+#define __this_cpu_read(pcp)		__this_branch_read(__this_cpu_read_, (pcp))
+#define __this_cpu_write(pcp, val)	__this_branch_modify(__this_cpu_write_, (pcp), (val))
+
+// #define __this_cpu_read(pcp)		percpu_from_op("mov", pcp)
+// #define __this_cpu_write(pcp, val)	percpu_to_op("mov", (pcp), val)
 #define __this_cpu_add(pcp, val)	percpu_to_op("add", (pcp), val)
 #define __this_cpu_sub(pcp, val)	percpu_to_op("sub", (pcp), val)
 #define __this_cpu_and(pcp, val)	percpu_to_op("and", (pcp), val)