s390/percpu: make use of interlocked-access facility 1 instructions

Optimize this_cpu_* functions for 64 bit by making use of new instructions
that came with the interlocked-access facility 1 (load-and-*) and the
general-instructions-extension facility (asi, agsi).
That way we get rid of the compare-and-swap loop in most cases.
Code size reduction (defconfig, -march=z196): 11,555 bytes.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/include/asm/percpu.h

@@ -21,7 +21,11 @@
 #define ARCH_NEEDS_WEAK_PER_CPU
 #endif
 
-#define arch_this_cpu_to_op(pcp, val, op)				\
+/*
+ * We use a compare-and-swap loop since that uses less cpu cycles than
+ * disabling and enabling interrupts like the generic variant would do.
+ */
+#define arch_this_cpu_to_op_simple(pcp, val, op)			\
 ({									\
 	typedef typeof(pcp) pcp_op_T__;					\
 	pcp_op_T__ old__, new__, prev__;				\
@@ -38,30 +42,101 @@
 	new__;								\
 })
 
-#define this_cpu_add_1(pcp, val) arch_this_cpu_to_op(pcp, val, +)
-#define this_cpu_add_2(pcp, val) arch_this_cpu_to_op(pcp, val, +)
-#define this_cpu_add_4(pcp, val) arch_this_cpu_to_op(pcp, val, +)
-#define this_cpu_add_8(pcp, val) arch_this_cpu_to_op(pcp, val, +)
-
-#define this_cpu_add_return_1(pcp, val) arch_this_cpu_to_op(pcp, val, +)
-#define this_cpu_add_return_2(pcp, val) arch_this_cpu_to_op(pcp, val, +)
-#define this_cpu_add_return_4(pcp, val) arch_this_cpu_to_op(pcp, val, +)
-#define this_cpu_add_return_8(pcp, val) arch_this_cpu_to_op(pcp, val, +)
-
-#define this_cpu_and_1(pcp, val) arch_this_cpu_to_op(pcp, val, &)
-#define this_cpu_and_2(pcp, val) arch_this_cpu_to_op(pcp, val, &)
-#define this_cpu_and_4(pcp, val) arch_this_cpu_to_op(pcp, val, &)
-#define this_cpu_and_8(pcp, val) arch_this_cpu_to_op(pcp, val, &)
-
-#define this_cpu_or_1(pcp, val) arch_this_cpu_to_op(pcp, val, |)
-#define this_cpu_or_2(pcp, val) arch_this_cpu_to_op(pcp, val, |)
-#define this_cpu_or_4(pcp, val) arch_this_cpu_to_op(pcp, val, |)
-#define this_cpu_or_8(pcp, val) arch_this_cpu_to_op(pcp, val, |)
-
-#define this_cpu_xor_1(pcp, val) arch_this_cpu_to_op(pcp, val, ^)
-#define this_cpu_xor_2(pcp, val) arch_this_cpu_to_op(pcp, val, ^)
-#define this_cpu_xor_4(pcp, val) arch_this_cpu_to_op(pcp, val, ^)
-#define this_cpu_xor_8(pcp, val) arch_this_cpu_to_op(pcp, val, ^)
+#define this_cpu_add_1(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, +)
+#define this_cpu_add_2(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, +)
+#define this_cpu_add_return_1(pcp, val) arch_this_cpu_to_op_simple(pcp, val, +)
+#define this_cpu_add_return_2(pcp, val) arch_this_cpu_to_op_simple(pcp, val, +)
+#define this_cpu_and_1(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, &)
+#define this_cpu_and_2(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, &)
+#define this_cpu_or_1(pcp, val)		arch_this_cpu_to_op_simple(pcp, val, |)
+#define this_cpu_or_2(pcp, val)		arch_this_cpu_to_op_simple(pcp, val, |)
+#define this_cpu_xor_1(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, ^)
+#define this_cpu_xor_2(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, ^)
+
+#ifndef CONFIG_HAVE_MARCH_Z196_FEATURES
+
+#define this_cpu_add_4(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, +)
+#define this_cpu_add_8(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, +)
+#define this_cpu_add_return_4(pcp, val) arch_this_cpu_to_op_simple(pcp, val, +)
+#define this_cpu_add_return_8(pcp, val) arch_this_cpu_to_op_simple(pcp, val, +)
+#define this_cpu_and_4(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, &)
+#define this_cpu_and_8(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, &)
+#define this_cpu_or_4(pcp, val)		arch_this_cpu_to_op_simple(pcp, val, |)
+#define this_cpu_or_8(pcp, val)		arch_this_cpu_to_op_simple(pcp, val, |)
+#define this_cpu_xor_4(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, ^)
+#define this_cpu_xor_8(pcp, val)	arch_this_cpu_to_op_simple(pcp, val, ^)
+
+#else /* CONFIG_HAVE_MARCH_Z196_FEATURES */
+
+#define arch_this_cpu_add(pcp, val, op1, op2, szcast)			\
+{									\
+	typedef typeof(pcp) pcp_op_T__; 				\
+	pcp_op_T__ val__ = (val);					\
+	pcp_op_T__ old__, *ptr__;					\
+	preempt_disable();						\
+	ptr__ = __this_cpu_ptr(&(pcp)); 				\
+	if (__builtin_constant_p(val__) &&				\
+	    ((szcast)val__ > -129) && ((szcast)val__ < 128)) {		\
+		asm volatile(						\
+			op2 "   %[ptr__],%[val__]\n"			\
+			: [ptr__] "+Q" (*ptr__) 			\
+			: [val__] "i" ((szcast)val__)			\
+			: "cc");					\
+	} else {							\
+		asm volatile(						\
+			op1 "   %[old__],%[val__],%[ptr__]\n"		\
+			: [old__] "=d" (old__), [ptr__] "+Q" (*ptr__)	\
+			: [val__] "d" (val__)				\
+			: "cc");					\
+	}								\
+	preempt_enable();						\
+}
+
+#define this_cpu_add_4(pcp, val) arch_this_cpu_add(pcp, val, "laa", "asi", int)
+#define this_cpu_add_8(pcp, val) arch_this_cpu_add(pcp, val, "laag", "agsi", long)
+
+#define arch_this_cpu_add_return(pcp, val, op)				\
+({									\
+	typedef typeof(pcp) pcp_op_T__; 				\
+	pcp_op_T__ val__ = (val);					\
+	pcp_op_T__ old__, *ptr__;					\
+	preempt_disable();						\
+	ptr__ = __this_cpu_ptr(&(pcp)); 				\
+	asm volatile(							\
+		op "    %[old__],%[val__],%[ptr__]\n"			\
+		: [old__] "=d" (old__), [ptr__] "+Q" (*ptr__)		\
+		: [val__] "d" (val__)					\
+		: "cc");						\
+	preempt_enable();						\
+	old__ + val__;							\
+})
+
+#define this_cpu_add_return_4(pcp, val) arch_this_cpu_add_return(pcp, val, "laa")
+#define this_cpu_add_return_8(pcp, val) arch_this_cpu_add_return(pcp, val, "laag")
+
+#define arch_this_cpu_to_op(pcp, val, op)				\
+{									\
+	typedef typeof(pcp) pcp_op_T__; 				\
+	pcp_op_T__ val__ = (val);					\
+	pcp_op_T__ old__, *ptr__;					\
+	preempt_disable();						\
+	ptr__ = __this_cpu_ptr(&(pcp)); 				\
+	asm volatile(							\
+		op "    %[old__],%[val__],%[ptr__]\n"			\
+		: [old__] "=d" (old__), [ptr__] "+Q" (*ptr__)		\
+		: [val__] "d" (val__)					\
+		: "cc");						\
+	preempt_enable();						\
+}
+
+#define this_cpu_and_4(pcp, val)	arch_this_cpu_to_op(pcp, val, "lan")
+#define this_cpu_and_8(pcp, val)	arch_this_cpu_to_op(pcp, val, "lang")
+#define this_cpu_or_4(pcp, val)		arch_this_cpu_to_op(pcp, val, "lao")
+#define this_cpu_or_8(pcp, val)		arch_this_cpu_to_op(pcp, val, "laog")
+#define this_cpu_xor_4(pcp, val)	arch_this_cpu_to_op(pcp, val, "lax")
+#define this_cpu_xor_8(pcp, val)	arch_this_cpu_to_op(pcp, val, "laxg")
+
+#endif /* CONFIG_HAVE_MARCH_Z196_FEATURES */
 
 #define arch_this_cpu_cmpxchg(pcp, oval, nval)				\
 ({									\