csky: atomic: Add custom atomic.h implementation

The generic atomic.h used cmpxchg to implement the atomic
operations, it will cause daul loop to reduce the forward
guarantee. The patch implement csky custom atomic operations with
ldex/stex instructions for the best performance.

Important comment by Rutland:
8e86f0b4 ("arm64: atomics: fix use of acquire + release for
full barrier semantics")

Link: https://lore.kernel.org/linux-riscv/CAJF2gTSAxpAi=LbAdu7jntZRUa=-dJwL0VfmDfBV5MHB=rcZ-w@mail.gmail.com/T/#m27a0f1342995deae49ce1d0e1f2683f8a181d6c3Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Guo Ren <guoren@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>

arch/csky/include/asm/atomic.h

+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ASM_CSKY_ATOMIC_H
+#define __ASM_CSKY_ATOMIC_H
+
+#ifdef CONFIG_SMP
+#include <asm-generic/atomic64.h>
+
+#include <asm/cmpxchg.h>
+#include <asm/barrier.h>
+
+#define __atomic_acquire_fence()	__bar_brarw()
+
+#define __atomic_release_fence()	__bar_brwaw()
+
+static __always_inline int arch_atomic_read(const atomic_t *v)
+{
+	return READ_ONCE(v->counter);
+}
+static __always_inline void arch_atomic_set(atomic_t *v, int i)
+{
+	WRITE_ONCE(v->counter, i);
+}
+
+#define ATOMIC_OP(op)							\
+static __always_inline							\
+void arch_atomic_##op(int i, atomic_t *v)				\
+{									\
+	unsigned long tmp;						\
+	__asm__ __volatile__ (						\
+	"1:	ldex.w		%0, (%2)	\n"			\
+	"	" #op "		%0, %1		\n"			\
+	"	stex.w		%0, (%2)	\n"			\
+	"	bez		%0, 1b		\n"			\
+	: "=&r" (tmp)							\
+	: "r" (i), "r" (&v->counter)					\
+	: "memory");							\
+}
+
+ATOMIC_OP(add)
+ATOMIC_OP(sub)
+ATOMIC_OP(and)
+ATOMIC_OP( or)
+ATOMIC_OP(xor)
+
+#undef ATOMIC_OP
+
+#define ATOMIC_FETCH_OP(op)						\
+static __always_inline							\
+int arch_atomic_fetch_##op##_relaxed(int i, atomic_t *v)		\
+{									\
+	register int ret, tmp;						\
+	__asm__ __volatile__ (						\
+	"1:	ldex.w		%0, (%3) \n"				\
+	"	mov		%1, %0   \n"				\
+	"	" #op "		%0, %2   \n"				\
+	"	stex.w		%0, (%3) \n"				\
+	"	bez		%0, 1b   \n"				\
+		: "=&r" (tmp), "=&r" (ret)				\
+		: "r" (i), "r"(&v->counter) 				\
+		: "memory");						\
+	return ret;							\
+}
+
+#define ATOMIC_OP_RETURN(op, c_op)					\
+static __always_inline							\
+int arch_atomic_##op##_return_relaxed(int i, atomic_t *v)		\
+{									\
+	return arch_atomic_fetch_##op##_relaxed(i, v) c_op i;		\
+}
+
+#define ATOMIC_OPS(op, c_op)						\
+	ATOMIC_FETCH_OP(op)						\
+	ATOMIC_OP_RETURN(op, c_op)
+
+ATOMIC_OPS(add, +)
+ATOMIC_OPS(sub, -)
+
+#define arch_atomic_fetch_add_relaxed	arch_atomic_fetch_add_relaxed
+#define arch_atomic_fetch_sub_relaxed	arch_atomic_fetch_sub_relaxed
+
+#define arch_atomic_add_return_relaxed	arch_atomic_add_return_relaxed
+#define arch_atomic_sub_return_relaxed	arch_atomic_sub_return_relaxed
+
+#undef ATOMIC_OPS
+#undef ATOMIC_OP_RETURN
+
+#define ATOMIC_OPS(op)							\
+	ATOMIC_FETCH_OP(op)
+
+ATOMIC_OPS(and)
+ATOMIC_OPS( or)
+ATOMIC_OPS(xor)
+
+#define arch_atomic_fetch_and_relaxed	arch_atomic_fetch_and_relaxed
+#define arch_atomic_fetch_or_relaxed	arch_atomic_fetch_or_relaxed
+#define arch_atomic_fetch_xor_relaxed	arch_atomic_fetch_xor_relaxed
+
+#undef ATOMIC_OPS
+
+#undef ATOMIC_FETCH_OP
+
+#define ATOMIC_OP()							\
+static __always_inline							\
+int arch_atomic_xchg_relaxed(atomic_t *v, int n)			\
+{									\
+	return __xchg_relaxed(n, &(v->counter), 4);			\
+}									\
+static __always_inline							\
+int arch_atomic_cmpxchg_relaxed(atomic_t *v, int o, int n)		\
+{									\
+	return __cmpxchg_relaxed(&(v->counter), o, n, 4);		\
+}									\
+static __always_inline							\
+int arch_atomic_cmpxchg_acquire(atomic_t *v, int o, int n)		\
+{									\
+	return __cmpxchg_acquire(&(v->counter), o, n, 4);		\
+}									\
+static __always_inline							\
+int arch_atomic_cmpxchg(atomic_t *v, int o, int n)			\
+{									\
+	return __cmpxchg(&(v->counter), o, n, 4);			\
+}
+
+#define ATOMIC_OPS()							\
+	ATOMIC_OP()
+
+ATOMIC_OPS()
+
+#define arch_atomic_xchg_relaxed	arch_atomic_xchg_relaxed
+#define arch_atomic_cmpxchg_relaxed	arch_atomic_cmpxchg_relaxed
+#define arch_atomic_cmpxchg_acquire	arch_atomic_cmpxchg_acquire
+#define arch_atomic_cmpxchg		arch_atomic_cmpxchg
+
+#undef ATOMIC_OPS
+#undef ATOMIC_OP
+
+#else
+#include <asm-generic/atomic.h>
+#endif
+
+#endif /* __ASM_CSKY_ATOMIC_H */