[PATCH v4] LoongArch/percpu: Simplify _percpu_read() and _percpu_write()

[Uros Bizjak]

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

The secondary effect of this change is to avoid explicit __percpu
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.

Patch is compile tested only.

[1] https://lore.kernel.org/lkml/20240812115945.484051-4-ubizjak@xxxxxxxxx/

Signed-off-by: Uros Bizjak <ubizjak@xxxxxxxxx>
Cc: Huacai Chen <chenhuacai@xxxxxxxxxx>
Cc: WANG Xuerui <kernel@xxxxxxxxxx>
Cc: Xi Ruoyao <xry111@xxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
---
v2: Add a missing cast and a missing coma in the asm template,
    reported by kernel test robot. Some formatting changes.
v3: Check the type of _val in _percpu_write().
v4: Remove type ckecking of _val. Remove unnecessary zero-extensions
    and truncations.
---
 arch/loongarch/include/asm/percpu.h | 124 ++++++++--------------------
 1 file changed, 35 insertions(+), 89 deletions(-)

diff --git a/arch/loongarch/include/asm/percpu.h b/arch/loongarch/include/asm/percpu.h
index 8f290e5546cf..87be9b14e9da 100644
--- a/arch/loongarch/include/asm/percpu.h
+++ b/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)
-- 
2.46.0