s390/atomic: Provide arch_atomic_*_and_test() implementations

Provide arch_atomic_*_and_test() implementations which make use of flag
output constraints, and allow the compiler to generate slightly better
code.

Reviewed-by: Juergen Christ <jchrist@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>

2 files changed, 109 insertions, 0 deletions

diff --git a/arch/s390/include/asm/atomic.h b/arch/s390/include/asm/atomic.h
index 92d6f33b6046..b36dd6a1d652 100644
--- a/arch/s390/include/asm/atomic.h
+++ b/arch/s390/include/asm/atomic.h
@@ -57,6 +57,24 @@ static __always_inline void arch_atomic_dec(atomic_t *v)
 }
 #define arch_atomic_dec arch_atomic_dec
 
+static __always_inline bool arch_atomic_sub_and_test(int i, atomic_t *v)
+{
+	return __atomic_add_and_test_barrier(-i, &v->counter);
+}
+#define arch_atomic_sub_and_test arch_atomic_sub_and_test
+
+static __always_inline bool arch_atomic_dec_and_test(atomic_t *v)
+{
+	return __atomic_add_const_and_test_barrier(-1, &v->counter);
+}
+#define arch_atomic_dec_and_test arch_atomic_dec_and_test
+
+static __always_inline bool arch_atomic_inc_and_test(atomic_t *v)
+{
+	return __atomic_add_const_and_test_barrier(1, &v->counter);
+}
+#define arch_atomic_inc_and_test arch_atomic_inc_and_test
+
 #define arch_atomic_sub(_i, _v)		arch_atomic_add(-(int)(_i), _v)
 #define arch_atomic_sub_return(_i, _v)	arch_atomic_add_return(-(int)(_i), _v)
 #define arch_atomic_fetch_sub(_i, _v)	arch_atomic_fetch_add(-(int)(_i), _v)
@@ -146,6 +164,24 @@ static __always_inline void arch_atomic64_dec(atomic64_t *v)
 }
 #define arch_atomic64_dec arch_atomic64_dec
 
+static __always_inline bool arch_atomic64_sub_and_test(s64 i, atomic64_t *v)
+{
+	return __atomic64_add_and_test_barrier(-i, (long *)&v->counter);
+}
+#define arch_atomic64_sub_and_test arch_atomic64_sub_and_test
+
+static __always_inline bool arch_atomic64_dec_and_test(atomic64_t *v)
+{
+	return __atomic64_add_const_and_test_barrier(-1, (long *)&v->counter);
+}
+#define arch_atomic64_dec_and_test arch_atomic64_dec_and_test
+
+static __always_inline bool arch_atomic64_inc_and_test(atomic64_t *v)
+{
+	return __atomic64_add_const_and_test_barrier(1, (long *)&v->counter);
+}
+#define arch_atomic64_inc_and_test arch_atomic64_inc_and_test
+
 static __always_inline s64 arch_atomic64_xchg(atomic64_t *v, s64 new)
 {
 	return arch_xchg(&v->counter, new);
diff --git a/arch/s390/include/asm/atomic_ops.h b/arch/s390/include/asm/atomic_ops.h
index 021503d81ed1..585678bbcd7a 100644
--- a/arch/s390/include/asm/atomic_ops.h
+++ b/arch/s390/include/asm/atomic_ops.h
@@ -10,6 +10,7 @@
 
 #include <linux/limits.h>
 #include <asm/march.h>
+#include <asm/asm.h>
 
 static __always_inline int __atomic_read(const int *ptr)
 {
@@ -169,4 +170,76 @@ __ATOMIC64_OPS(__atomic64_xor, "xgr")
 
 #endif /* MARCH_HAS_Z196_FEATURES */
 
+#if defined(MARCH_HAS_Z196_FEATURES) && defined(__HAVE_ASM_FLAG_OUTPUTS__)
+
+#define __ATOMIC_TEST_OP(op_name, op_type, op_string, op_barrier)	\
+static __always_inline bool op_name(op_type val, op_type *ptr)		\
+{									\
+	op_type tmp;							\
+	int cc;								\
+									\
+	asm volatile(							\
+		op_string "	%[tmp],%[val],%[ptr]\n"			\
+		op_barrier						\
+		: "=@cc" (cc), [tmp] "=d" (tmp), [ptr] "+QS" (*ptr)	\
+		: [val] "d" (val)					\
+		: "memory");						\
+	return (cc == 0) || (cc == 2);					\
+}									\
+
+#define __ATOMIC_TEST_OPS(op_name, op_type, op_string)			\
+	__ATOMIC_TEST_OP(op_name, op_type, op_string, "")		\
+	__ATOMIC_TEST_OP(op_name##_barrier, op_type, op_string, "bcr 14,0\n")
+
+__ATOMIC_TEST_OPS(__atomic_add_and_test, int, "laal")
+__ATOMIC_TEST_OPS(__atomic64_add_and_test, long, "laalg")
+
+#undef __ATOMIC_TEST_OPS
+#undef __ATOMIC_TEST_OP
+
+#define __ATOMIC_CONST_TEST_OP(op_name, op_type, op_string, op_barrier)	\
+static __always_inline bool op_name(op_type val, op_type *ptr)		\
+{									\
+	int cc;								\
+									\
+	asm volatile(							\
+		op_string "	%[ptr],%[val]\n"			\
+		op_barrier						\
+		: "=@cc" (cc), [ptr] "+QS" (*ptr)			\
+		: [val] "i" (val)					\
+		: "memory");						\
+	return (cc == 0) || (cc == 2);					\
+}
+
+#define __ATOMIC_CONST_TEST_OPS(op_name, op_type, op_string)		\
+	__ATOMIC_CONST_TEST_OP(op_name, op_type, op_string, "")		\
+	__ATOMIC_CONST_TEST_OP(op_name##_barrier, op_type, op_string, "bcr 14,0\n")
+
+__ATOMIC_CONST_TEST_OPS(__atomic_add_const_and_test, int, "alsi")
+__ATOMIC_CONST_TEST_OPS(__atomic64_add_const_and_test, long, "algsi")
+
+#undef __ATOMIC_CONST_TEST_OPS
+#undef __ATOMIC_CONST_TEST_OP
+
+#else /* defined(MARCH_HAS_Z196_FEATURES) && defined(__HAVE_ASM_FLAG_OUTPUTS__) */
+
+#define __ATOMIC_TEST_OP(op_name, op_func, op_type)			\
+static __always_inline bool op_name(op_type val, op_type *ptr)		\
+{									\
+	return op_func(val, ptr) == -val;				\
+}
+
+__ATOMIC_TEST_OP(__atomic_add_and_test,			__atomic_add,		int)
+__ATOMIC_TEST_OP(__atomic_add_and_test_barrier,		__atomic_add_barrier,	int)
+__ATOMIC_TEST_OP(__atomic_add_const_and_test,		__atomic_add,		int)
+__ATOMIC_TEST_OP(__atomic_add_const_and_test_barrier,	__atomic_add_barrier,	int)
+__ATOMIC_TEST_OP(__atomic64_add_and_test,		__atomic64_add,		long)
+__ATOMIC_TEST_OP(__atomic64_add_and_test_barrier,	__atomic64_add_barrier, long)
+__ATOMIC_TEST_OP(__atomic64_add_const_and_test,		__atomic64_add,		long)
+__ATOMIC_TEST_OP(__atomic64_add_const_and_test_barrier,	__atomic64_add_barrier,	long)
+
+#undef __ATOMIC_TEST_OP
+
+#endif /* defined(MARCH_HAS_Z196_FEATURES) && defined(__HAVE_ASM_FLAG_OUTPUTS__) */
+
 #endif /* __ARCH_S390_ATOMIC_OPS__  */