LoongArch: Simplify _percpu_read() and _percpu_write()

Now _percpu_read() and _percpu_write() macros call __percpu_read()
and __percpu_write() static inline functions that result in a single
assembly instruction. However, percpu infrastructure expects its leaf
definitions to encode the size of their percpu variable, so the patch
merges all the asm clauses from the static inline function into the
corresponding leaf macros.

The secondary effect of this change is to avoid explicit __percpu
annotations for function arguments. Currently, __percpu macro is defined
in include/linux/compiler_types.h, but with proposed patch [1], __percpu
definition will need macros from include/asm-generic/percpu.h, creating
forward dependency loop.

The proposed solution is the same as x86 architecture uses.

[1] https://lore.kernel.org/lkml/[email protected]/

Tested-by: Xi Ruoyao <[email protected]>
Signed-off-by: Uros Bizjak <[email protected]>
Signed-off-by: Huacai Chen <[email protected]>

Author: ubizjak

arch/loongarch/include/asm/percpu.h

@@ -68,75 +68,6 @@ PERCPU_OP(and, and, &)
 PERCPU_OP(or, or, |)
 #undef PERCPU_OP
 
-static __always_inline unsigned long __percpu_read(void __percpu *ptr, int size)
-{
-	unsigned long ret;
-
-	switch (size) {
-	case 1:
-		__asm__ __volatile__ ("ldx.b %[ret], $r21, %[ptr]	\n"
-		: [ret] "=&r"(ret)
-		: [ptr] "r"(ptr)
-		: "memory");
-		break;
-	case 2:
-		__asm__ __volatile__ ("ldx.h %[ret], $r21, %[ptr]	\n"
-		: [ret] "=&r"(ret)
-		: [ptr] "r"(ptr)
-		: "memory");
-		break;
-	case 4:
-		__asm__ __volatile__ ("ldx.w %[ret], $r21, %[ptr]	\n"
-		: [ret] "=&r"(ret)
-		: [ptr] "r"(ptr)
-		: "memory");
-		break;
-	case 8:
-		__asm__ __volatile__ ("ldx.d %[ret], $r21, %[ptr]	\n"
-		: [ret] "=&r"(ret)
-		: [ptr] "r"(ptr)
-		: "memory");
-		break;
-	default:
-		ret = 0;
-		BUILD_BUG();
-	}
-
-	return ret;
-}
-
-static __always_inline void __percpu_write(void __percpu *ptr, unsigned long val, int size)
-{
-	switch (size) {
-	case 1:
-		__asm__ __volatile__("stx.b %[val], $r21, %[ptr]	\n"
-		:
-		: [val] "r" (val), [ptr] "r" (ptr)
-		: "memory");
-		break;
-	case 2:
-		__asm__ __volatile__("stx.h %[val], $r21, %[ptr]	\n"
-		:
-		: [val] "r" (val), [ptr] "r" (ptr)
-		: "memory");
-		break;
-	case 4:
-		__asm__ __volatile__("stx.w %[val], $r21, %[ptr]	\n"
-		:
-		: [val] "r" (val), [ptr] "r" (ptr)
-		: "memory");
-		break;
-	case 8:
-		__asm__ __volatile__("stx.d %[val], $r21, %[ptr]	\n"
-		:
-		: [val] "r" (val), [ptr] "r" (ptr)
-		: "memory");
-		break;
-	default:
-		BUILD_BUG();
-	}
-}
-
 static __always_inline unsigned long __percpu_xchg(void *ptr, unsigned long val, int size)
 {
 	switch (size) {
@@ -157,6 +88,33 @@ static __always_inline unsigned long __percpu_xchg(void *ptr, unsigned long val,
 	return 0;
 }
 
+#define __pcpu_op_1(op)		op ".b "
+#define __pcpu_op_2(op)		op ".h "
+#define __pcpu_op_4(op)		op ".w "
+#define __pcpu_op_8(op)		op ".d "
+
+#define _percpu_read(size, _pcp)					\
+({									\
+	typeof(_pcp) __pcp_ret;						\
+									\
+	__asm__ __volatile__(						\
+		__pcpu_op_##size("ldx") "%[ret], $r21, %[ptr]	\n"	\
+		: [ret] "=&r"(__pcp_ret)				\
+		: [ptr] "r"(&(_pcp))					\
+		: "memory");						\
+									\
+	__pcp_ret;							\
+})
+
+#define _percpu_write(size, _pcp, _val)					\
+do {									\
+	__asm__ __volatile__(						\
+		__pcpu_op_##size("stx") "%[val], $r21, %[ptr]	\n"	\
+		:							\
+		: [val] "r"(_val), [ptr] "r"(&(_pcp))			\
+		: "memory");						\
+} while (0)
+
 /* this_cpu_cmpxchg */
 #define _protect_cmpxchg_local(pcp, o, n)			\
 ({								\
@@ -167,18 +125,6 @@ static __always_inline unsigned long __percpu_xchg(void *ptr, unsigned long val,
 	__ret;							\
 })
 
-#define _percpu_read(pcp)						\
-({									\
-	typeof(pcp) __retval;						\
-	__retval = (typeof(pcp))__percpu_read(&(pcp), sizeof(pcp));	\
-	__retval;							\
-})
-
-#define _percpu_write(pcp, val)						\
-do {									\
-	__percpu_write(&(pcp), (unsigned long)(val), sizeof(pcp));	\
-} while (0)								\
-
 #define _pcp_protect(operation, pcp, val)			\
 ({								\
 	typeof(pcp) __retval;					\
@@ -215,15 +161,15 @@ do {									\
 #define this_cpu_or_4(pcp, val) _percpu_or(pcp, val)
 #define this_cpu_or_8(pcp, val) _percpu_or(pcp, val)
 
-#define this_cpu_read_1(pcp) _percpu_read(pcp)
-#define this_cpu_read_2(pcp) _percpu_read(pcp)
-#define this_cpu_read_4(pcp) _percpu_read(pcp)
-#define this_cpu_read_8(pcp) _percpu_read(pcp)
+#define this_cpu_read_1(pcp) _percpu_read(1, pcp)
+#define this_cpu_read_2(pcp) _percpu_read(2, pcp)
+#define this_cpu_read_4(pcp) _percpu_read(4, pcp)
+#define this_cpu_read_8(pcp) _percpu_read(8, pcp)
 
-#define this_cpu_write_1(pcp, val) _percpu_write(pcp, val)
-#define this_cpu_write_2(pcp, val) _percpu_write(pcp, val)
-#define this_cpu_write_4(pcp, val) _percpu_write(pcp, val)
-#define this_cpu_write_8(pcp, val) _percpu_write(pcp, val)
+#define this_cpu_write_1(pcp, val) _percpu_write(1, pcp, val)
+#define this_cpu_write_2(pcp, val) _percpu_write(2, pcp, val)
+#define this_cpu_write_4(pcp, val) _percpu_write(4, pcp, val)
+#define this_cpu_write_8(pcp, val) _percpu_write(8, pcp, val)
 
 #define this_cpu_xchg_1(pcp, val) _percpu_xchg(pcp, val)
 #define this_cpu_xchg_2(pcp, val) _percpu_xchg(pcp, val)