[S390] percpu: implement arch specific irqsafe_cpu_ops

Implement arch specific irqsafe_cpu ops. The arch specific ops do not
disable/enable interrupts since that is an expensive operation. Instead
we disable preemption and perform a compare and swap loop.
Since on server distros (the ones we care about) preemption is disabled
the preempt_disable()/preempt_enable() pair is a nop.
In the end this code should be faster than the generic one.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/include/asm/cmpxchg.h

@@ -167,7 +167,6 @@ static inline unsigned long __cmpxchg(void *ptr, unsigned long old,
 #ifdef CONFIG_64BIT
 #define cmpxchg64(ptr, o, n)						\
 ({									\
-	BUILD_BUG_ON(sizeof(*(ptr)) != 8);				\
 	cmpxchg((ptr), (o), (n));					\
 })
 #else /* CONFIG_64BIT */

arch/s390/include/asm/percpu.h

@@ -1,6 +1,9 @@
 #ifndef __ARCH_S390_PERCPU__
 #define __ARCH_S390_PERCPU__
 
+#include <linux/preempt.h>
+#include <asm/cmpxchg.h>
+
 /*
  * s390 uses its own implementation for per cpu data, the offset of
  * the cpu local data area is cached in the cpu's lowcore memory.
@@ -16,6 +19,71 @@
 #define ARCH_NEEDS_WEAK_PER_CPU
 #endif
 
+#define arch_irqsafe_cpu_to_op(pcp, val, op)				\
+do {									\
+	typedef typeof(pcp) pcp_op_T__;					\
+	pcp_op_T__ old__, new__, prev__;				\
+	pcp_op_T__ *ptr__;						\
+	preempt_disable();						\
+	ptr__ = __this_cpu_ptr(&(pcp));					\
+	prev__ = *ptr__;						\
+	do {								\
+		old__ = prev__;						\
+		new__ = old__ op (val);					\
+		switch (sizeof(*ptr__)) {				\
+		case 8:							\
+			prev__ = cmpxchg64(ptr__, old__, new__);	\
+			break;						\
+		default:						\
+			prev__ = cmpxchg(ptr__, old__, new__);		\
+		}							\
+	} while (prev__ != old__);					\
+	preempt_enable();						\
+} while (0)
+
+#define irqsafe_cpu_add_1(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, +)
+#define irqsafe_cpu_add_2(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, +)
+#define irqsafe_cpu_add_4(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, +)
+#define irqsafe_cpu_add_8(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, +)
+
+#define irqsafe_cpu_and_1(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, &)
+#define irqsafe_cpu_and_2(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, &)
+#define irqsafe_cpu_and_4(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, &)
+#define irqsafe_cpu_and_8(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, &)
+
+#define irqsafe_cpu_or_1(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, |)
+#define irqsafe_cpu_or_2(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, |)
+#define irqsafe_cpu_or_4(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, |)
+#define irqsafe_cpu_or_8(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, |)
+
+#define irqsafe_cpu_xor_1(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, ^)
+#define irqsafe_cpu_xor_2(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, ^)
+#define irqsafe_cpu_xor_4(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, ^)
+#define irqsafe_cpu_xor_8(pcp, val) arch_irqsafe_cpu_to_op(pcp, val, ^)
+
+#define arch_irqsafe_cpu_cmpxchg(pcp, oval, nval)			\
+({									\
+	typedef typeof(pcp) pcp_op_T__;					\
+	pcp_op_T__ ret__;						\
+	pcp_op_T__ *ptr__;						\
+	preempt_disable();						\
+	ptr__ = __this_cpu_ptr(&(pcp));					\
+	switch (sizeof(*ptr__)) {					\
+	case 8:								\
+		ret__ = cmpxchg64(ptr__, oval, nval);			\
+		break;							\
+	default:							\
+		ret__ = cmpxchg(ptr__, oval, nval);			\
+	}								\
+	preempt_enable();						\
+	ret__;								\
+})
+
+#define irqsafe_cpu_cmpxchg_1(pcp, oval, nval) arch_irqsafe_cpu_cmpxchg(pcp, oval, nval)
+#define irqsafe_cpu_cmpxchg_2(pcp, oval, nval) arch_irqsafe_cpu_cmpxchg(pcp, oval, nval)
+#define irqsafe_cpu_cmpxchg_4(pcp, oval, nval) arch_irqsafe_cpu_cmpxchg(pcp, oval, nval)
+#define irqsafe_cpu_cmpxchg_8(pcp, oval, nval) arch_irqsafe_cpu_cmpxchg(pcp, oval, nval)
+
 #include <asm-generic/percpu.h>
 
 #endif /* __ARCH_S390_PERCPU__ */