Merge branch 'this_cpu_ops' into for-2.6.38

2 files changed, 153 insertions, 37 deletions

diff --git a/arch/x86/Kconfig.cpu b/arch/x86/Kconfig.cpu
index 2ac9069890cd..15588a0ef466 100644
--- a/arch/x86/Kconfig.cpu
+++ b/arch/x86/Kconfig.cpu
@@ -310,6 +310,9 @@ config X86_INTERNODE_CACHE_SHIFT
 config X86_CMPXCHG
 	def_bool X86_64 || (X86_32 && !M386)
 
+config CMPXCHG_LOCAL
+	def_bool X86_64 || (X86_32 && !M386)
+
 config X86_L1_CACHE_SHIFT
 	int
 	default "7" if MPENTIUM4 || MPSC
diff --git a/arch/x86/include/asm/percpu.h b/arch/x86/include/asm/percpu.h
index 38f9e965ff96..8ee45167e817 100644
--- a/arch/x86/include/asm/percpu.h
+++ b/arch/x86/include/asm/percpu.h
@@ -177,39 +177,6 @@ do {									\
 	}								\
 } while (0)
 
-/*
- * Add return operation
- */
-#define percpu_add_return_op(var, val)					\
-({									\
-	typeof(var) paro_ret__ = val;					\
-	switch (sizeof(var)) {						\
-	case 1:								\
-		asm("xaddb %0, "__percpu_arg(1)				\
-			    : "+q" (paro_ret__), "+m" (var)		\
-			    : : "memory");				\
-		break;							\
-	case 2:								\
-		asm("xaddw %0, "__percpu_arg(1)				\
-			    : "+r" (paro_ret__), "+m" (var)		\
-			    : : "memory");				\
-		break;							\
-	case 4:								\
-		asm("xaddl %0, "__percpu_arg(1)				\
-			    : "+r" (paro_ret__), "+m" (var)		\
-			    : : "memory");				\
-		break;							\
-	case 8:								\
-		asm("xaddq %0, "__percpu_arg(1)				\
-			    : "+re" (paro_ret__), "+m" (var)		\
-			    : : "memory");				\
-		break;							\
-	default: __bad_percpu_size();					\
-	}								\
-	paro_ret__ += val;						\
-	paro_ret__;							\
-})
-
 #define percpu_from_op(op, var, constraint)		\
 ({							\
 	typeof(var) pfo_ret__;				\
@@ -263,6 +230,125 @@ do {									\
 })
 
 /*
+ * Add return operation
+ */
+#define percpu_add_return_op(var, val)					\
+({									\
+	typeof(var) paro_ret__ = val;					\
+	switch (sizeof(var)) {						\
+	case 1:								\
+		asm("xaddb %0, "__percpu_arg(1)				\
+			    : "+q" (paro_ret__), "+m" (var)		\
+			    : : "memory");				\
+		break;							\
+	case 2:								\
+		asm("xaddw %0, "__percpu_arg(1)				\
+			    : "+r" (paro_ret__), "+m" (var)		\
+			    : : "memory");				\
+		break;							\
+	case 4:								\
+		asm("xaddl %0, "__percpu_arg(1)				\
+			    : "+r" (paro_ret__), "+m" (var)		\
+			    : : "memory");				\
+		break;							\
+	case 8:								\
+		asm("xaddq %0, "__percpu_arg(1)				\
+			    : "+re" (paro_ret__), "+m" (var)		\
+			    : : "memory");				\
+		break;							\
+	default: __bad_percpu_size();					\
+	}								\
+	paro_ret__ += val;						\
+	paro_ret__;							\
+})
+
+/*
+ * xchg is implemented using cmpxchg without a lock prefix. xchg is
+ * expensive due to the implied lock prefix.  The processor cannot prefetch
+ * cachelines if xchg is used.
+ */
+#define percpu_xchg_op(var, nval)					\
+({									\
+	typeof(var) pxo_ret__;						\
+	typeof(var) pxo_new__ = (nval);					\
+	switch (sizeof(var)) {						\
+	case 1:								\
+		asm("\n1:mov "__percpu_arg(1)",%%al"			\
+		    "\n\tcmpxchgb %2, "__percpu_arg(1)			\
+		    "\n\tjnz 1b"					\
+			    : "=a" (pxo_ret__), "+m" (var)		\
+			    : "q" (pxo_new__)				\
+			    : "memory");				\
+		break;							\
+	case 2:								\
+		asm("\n1:mov "__percpu_arg(1)",%%ax"			\
+		    "\n\tcmpxchgw %2, "__percpu_arg(1)			\
+		    "\n\tjnz 1b"					\
+			    : "=a" (pxo_ret__), "+m" (var)		\
+			    : "r" (pxo_new__)				\
+			    : "memory");				\
+		break;							\
+	case 4:								\
+		asm("\n1:mov "__percpu_arg(1)",%%eax"			\
+		    "\n\tcmpxchgl %2, "__percpu_arg(1)			\
+		    "\n\tjnz 1b"					\
+			    : "=a" (pxo_ret__), "+m" (var)		\
+			    : "r" (pxo_new__)				\
+			    : "memory");				\
+		break;							\
+	case 8:								\
+		asm("\n1:mov "__percpu_arg(1)",%%rax"			\
+		    "\n\tcmpxchgq %2, "__percpu_arg(1)			\
+		    "\n\tjnz 1b"					\
+			    : "=a" (pxo_ret__), "+m" (var)		\
+			    : "r" (pxo_new__)				\
+			    : "memory");				\
+		break;							\
+	default: __bad_percpu_size();					\
+	}								\
+	pxo_ret__;							\
+})
+
+/*
+ * cmpxchg has no such implied lock semantics as a result it is much
+ * more efficient for cpu local operations.
+ */
+#define percpu_cmpxchg_op(var, oval, nval)				\
+({									\
+	typeof(var) pco_ret__;						\
+	typeof(var) pco_old__ = (oval);					\
+	typeof(var) pco_new__ = (nval);					\
+	switch (sizeof(var)) {						\
+	case 1:								\
+		asm("cmpxchgb %2, "__percpu_arg(1)			\
+			    : "=a" (pco_ret__), "+m" (var)		\
+			    : "q" (pco_new__), "0" (pco_old__)		\
+			    : "memory");				\
+		break;							\
+	case 2:								\
+		asm("cmpxchgw %2, "__percpu_arg(1)			\
+			    : "=a" (pco_ret__), "+m" (var)		\
+			    : "r" (pco_new__), "0" (pco_old__)		\
+			    : "memory");				\
+		break;							\
+	case 4:								\
+		asm("cmpxchgl %2, "__percpu_arg(1)			\
+			    : "=a" (pco_ret__), "+m" (var)		\
+			    : "r" (pco_new__), "0" (pco_old__)		\
+			    : "memory");				\
+		break;							\
+	case 8:								\
+		asm("cmpxchgq %2, "__percpu_arg(1)			\
+			    : "=a" (pco_ret__), "+m" (var)		\
+			    : "r" (pco_new__), "0" (pco_old__)		\
+			    : "memory");				\
+		break;							\
+	default: __bad_percpu_size();					\
+	}								\
+	pco_ret__;							\
+})
+
+/*
  * percpu_read() makes gcc load the percpu variable every time it is
  * accessed while percpu_read_stable() allows the value to be cached.
  * percpu_read_stable() is more efficient and can be used if its value
@@ -300,6 +386,12 @@ do {									\
 #define __this_cpu_xor_1(pcp, val)	percpu_to_op("xor", (pcp), val)
 #define __this_cpu_xor_2(pcp, val)	percpu_to_op("xor", (pcp), val)
 #define __this_cpu_xor_4(pcp, val)	percpu_to_op("xor", (pcp), val)
+/*
+ * Generic fallback operations for __this_cpu_xchg_[1-4] are okay and much
+ * faster than an xchg with forced lock semantics.
+ */
+#define __this_cpu_xchg_8(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define __this_cpu_cmpxchg_8(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
 
 #define this_cpu_read_1(pcp)		percpu_from_op("mov", (pcp), "m"(pcp))
 #define this_cpu_read_2(pcp)		percpu_from_op("mov", (pcp), "m"(pcp))
@@ -319,6 +411,11 @@ do {									\
 #define this_cpu_xor_1(pcp, val)	percpu_to_op("xor", (pcp), val)
 #define this_cpu_xor_2(pcp, val)	percpu_to_op("xor", (pcp), val)
 #define this_cpu_xor_4(pcp, val)	percpu_to_op("xor", (pcp), val)
+#define this_cpu_xchg_1(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define this_cpu_xchg_2(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define this_cpu_xchg_4(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define this_cpu_xchg_8(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define this_cpu_cmpxchg_8(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
 
 #define irqsafe_cpu_add_1(pcp, val)	percpu_add_op((pcp), val)
 #define irqsafe_cpu_add_2(pcp, val)	percpu_add_op((pcp), val)
@@ -332,15 +429,32 @@ do {									\
 #define irqsafe_cpu_xor_1(pcp, val)	percpu_to_op("xor", (pcp), val)
 #define irqsafe_cpu_xor_2(pcp, val)	percpu_to_op("xor", (pcp), val)
 #define irqsafe_cpu_xor_4(pcp, val)	percpu_to_op("xor", (pcp), val)
+#define irqsafe_cpu_xchg_1(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define irqsafe_cpu_xchg_2(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define irqsafe_cpu_xchg_4(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define irqsafe_cpu_xchg_8(pcp, nval)	percpu_xchg_op(pcp, nval)
+#define irqsafe_cpu_cmpxchg_8(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
 
 #ifndef CONFIG_M386
 #define __this_cpu_add_return_1(pcp, val) percpu_add_return_op(pcp, val)
 #define __this_cpu_add_return_2(pcp, val) percpu_add_return_op(pcp, val)
 #define __this_cpu_add_return_4(pcp, val) percpu_add_return_op(pcp, val)
+#define __this_cpu_cmpxchg_1(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+#define __this_cpu_cmpxchg_2(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+#define __this_cpu_cmpxchg_4(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+
 #define this_cpu_add_return_1(pcp, val)	percpu_add_return_op(pcp, val)
 #define this_cpu_add_return_2(pcp, val)	percpu_add_return_op(pcp, val)
 #define this_cpu_add_return_4(pcp, val)	percpu_add_return_op(pcp, val)
-#endif
+#define this_cpu_cmpxchg_1(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+#define this_cpu_cmpxchg_2(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+#define this_cpu_cmpxchg_4(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+
+#define irqsafe_cpu_cmpxchg_1(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+#define irqsafe_cpu_cmpxchg_2(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+#define irqsafe_cpu_cmpxchg_4(pcp, oval, nval)	percpu_cmpxchg_op(pcp, oval, nval)
+#endif /* !CONFIG_M386 */
+
 /*
  * Per cpu atomic 64 bit operations are only available under 64 bit.
  * 32 bit must fall back to generic operations.
@@ -352,6 +466,7 @@ do {									\
 #define __this_cpu_and_8(pcp, val)	percpu_to_op("and", (pcp), val)
 #define __this_cpu_or_8(pcp, val)	percpu_to_op("or", (pcp), val)
 #define __this_cpu_xor_8(pcp, val)	percpu_to_op("xor", (pcp), val)
+#define __this_cpu_add_return_8(pcp, val) percpu_add_return_op(pcp, val)
 
 #define this_cpu_read_8(pcp)		percpu_from_op("mov", (pcp), "m"(pcp))
 #define this_cpu_write_8(pcp, val)	percpu_to_op("mov", (pcp), val)
@@ -359,14 +474,12 @@ do {									\
 #define this_cpu_and_8(pcp, val)	percpu_to_op("and", (pcp), val)
 #define this_cpu_or_8(pcp, val)		percpu_to_op("or", (pcp), val)
 #define this_cpu_xor_8(pcp, val)	percpu_to_op("xor", (pcp), val)
+#define this_cpu_add_return_8(pcp, val)	percpu_add_return_op(pcp, val)
 
 #define irqsafe_cpu_add_8(pcp, val)	percpu_add_op((pcp), val)
 #define irqsafe_cpu_and_8(pcp, val)	percpu_to_op("and", (pcp), val)
 #define irqsafe_cpu_or_8(pcp, val)	percpu_to_op("or", (pcp), val)
 #define irqsafe_cpu_xor_8(pcp, val)	percpu_to_op("xor", (pcp), val)
-
-#define __this_cpu_add_return_8(pcp, val) percpu_add_return_op(pcp, val)
-#define this_cpu_add_return_8(pcp, val)	percpu_add_return_op(pcp, val)
 #endif
 
 /* This is not atomic against other CPUs -- CPU preemption needs to be off */