Merge patch series "riscv/barrier: tidying up barrier-related macro"

Eric Chan <ericchancf@google.com> says:

This series makes barrier-related macro more neat and clear.
This is a follow-up to [0-3], change to multiple patches,
for readability, create new message thread.

[0](v1/v2) https://lore.kernel.org/lkml/20240209125048.4078639-1-ericchancf@google.com/
[1] (v3) https://lore.kernel.org/lkml/20240213142856.2416073-1-ericchancf@google.com/
[2] (v4) https://lore.kernel.org/lkml/20240213200923.2547570-1-ericchancf@google.com/
[4] (v5) https://lore.kernel.org/lkml/20240213223810.2595804-1-ericchancf@google.com/

* b4-shazam-merge:
  riscv/barrier: Add missing space after ','
  riscv/barrier: Consolidate fence definitions
  riscv/barrier: Define RISCV_FULL_BARRIER
  riscv/barrier: Define __{mb,rmb,wmb}

Link: https://lore.kernel.org/r/20240217131206.3667544-1-ericchancf@google.comSigned-off-by: Palmer Dabbelt <palmer@rivosinc.com>

arch/riscv/include/asm/atomic.h

@@ -17,7 +17,6 @@
 #endif
 
 #include <asm/cmpxchg.h>
-#include <asm/barrier.h>
 
 #define __atomic_acquire_fence()					\
 	__asm__ __volatile__(RISCV_ACQUIRE_BARRIER "" ::: "memory")
@@ -207,7 +206,7 @@ static __always_inline int arch_atomic_fetch_add_unless(atomic_t *v, int a, int
 		"	add      %[rc], %[p], %[a]\n"
 		"	sc.w.rl  %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		: [a]"r" (a), [u]"r" (u)
@@ -228,7 +227,7 @@ static __always_inline s64 arch_atomic64_fetch_add_unless(atomic64_t *v, s64 a,
 		"	add      %[rc], %[p], %[a]\n"
 		"	sc.d.rl  %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		: [a]"r" (a), [u]"r" (u)
@@ -248,7 +247,7 @@ static __always_inline bool arch_atomic_inc_unless_negative(atomic_t *v)
 		"	addi      %[rc], %[p], 1\n"
 		"	sc.w.rl   %[rc], %[rc], %[c]\n"
 		"	bnez      %[rc], 0b\n"
-		"	fence     rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -268,7 +267,7 @@ static __always_inline bool arch_atomic_dec_unless_positive(atomic_t *v)
 		"	addi      %[rc], %[p], -1\n"
 		"	sc.w.rl   %[rc], %[rc], %[c]\n"
 		"	bnez      %[rc], 0b\n"
-		"	fence     rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -288,7 +287,7 @@ static __always_inline int arch_atomic_dec_if_positive(atomic_t *v)
 		"	bltz     %[rc], 1f\n"
 		"	sc.w.rl  %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -310,7 +309,7 @@ static __always_inline bool arch_atomic64_inc_unless_negative(atomic64_t *v)
 		"	addi      %[rc], %[p], 1\n"
 		"	sc.d.rl   %[rc], %[rc], %[c]\n"
 		"	bnez      %[rc], 0b\n"
-		"	fence     rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -331,7 +330,7 @@ static __always_inline bool arch_atomic64_dec_unless_positive(atomic64_t *v)
 		"	addi      %[rc], %[p], -1\n"
 		"	sc.d.rl   %[rc], %[rc], %[c]\n"
 		"	bnez      %[rc], 0b\n"
-		"	fence     rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:
@@ -352,7 +351,7 @@ static __always_inline s64 arch_atomic64_dec_if_positive(atomic64_t *v)
 		"	bltz     %[rc], 1f\n"
 		"	sc.d.rl  %[rc], %[rc], %[c]\n"
 		"	bnez     %[rc], 0b\n"
-		"	fence    rw, rw\n"
+		RISCV_FULL_BARRIER
 		"1:\n"
 		: [p]"=&r" (prev), [rc]"=&r" (rc), [c]"+A" (v->counter)
 		:

arch/riscv/include/asm/barrier.h

@@ -11,28 +11,27 @@
 #define _ASM_RISCV_BARRIER_H
 
 #ifndef __ASSEMBLY__
+#include <asm/fence.h>
 
 #define nop()		__asm__ __volatile__ ("nop")
 #define __nops(n)	".rept	" #n "\nnop\n.endr\n"
 #define nops(n)		__asm__ __volatile__ (__nops(n))
 
-#define RISCV_FENCE(p, s) \
-	__asm__ __volatile__ ("fence " #p "," #s : : : "memory")
 
 /* These barriers need to enforce ordering on both devices or memory. */
-#define mb()		RISCV_FENCE(iorw,iorw)
-#define rmb()		RISCV_FENCE(ir,ir)
-#define wmb()		RISCV_FENCE(ow,ow)
+#define __mb()		RISCV_FENCE(iorw, iorw)
+#define __rmb()		RISCV_FENCE(ir, ir)
+#define __wmb()		RISCV_FENCE(ow, ow)
 
 /* These barriers do not need to enforce ordering on devices, just memory. */
-#define __smp_mb()	RISCV_FENCE(rw,rw)
-#define __smp_rmb()	RISCV_FENCE(r,r)
-#define __smp_wmb()	RISCV_FENCE(w,w)
+#define __smp_mb()	RISCV_FENCE(rw, rw)
+#define __smp_rmb()	RISCV_FENCE(r, r)
+#define __smp_wmb()	RISCV_FENCE(w, w)
 
 #define __smp_store_release(p, v)					\
 do {									\
 	compiletime_assert_atomic_type(*p);				\
-	RISCV_FENCE(rw,w);						\
+	RISCV_FENCE(rw, w);						\
 	WRITE_ONCE(*p, v);						\
 } while (0)
 
@@ -40,7 +39,7 @@ do {									\
 ({									\
 	typeof(*p) ___p1 = READ_ONCE(*p);				\
 	compiletime_assert_atomic_type(*p);				\
-	RISCV_FENCE(r,rw);						\
+	RISCV_FENCE(r, rw);						\
 	___p1;								\
 })
 
@@ -69,7 +68,7 @@ do {									\
  * instances the scheduler pairs this with an mb(), so nothing is necessary on
  * the new hart.
  */
-#define smp_mb__after_spinlock()	RISCV_FENCE(iorw,iorw)
+#define smp_mb__after_spinlock()	RISCV_FENCE(iorw, iorw)
 
 #include <asm-generic/barrier.h>
 

arch/riscv/include/asm/cmpxchg.h

@@ -8,7 +8,6 @@
 
 #include <linux/bug.h>
 
-#include <asm/barrier.h>
 #include <asm/fence.h>
 
 #define __xchg_relaxed(ptr, new, size)					\
@@ -313,7 +312,7 @@
 			"	bne  %0, %z3, 1f\n"			\
 			"	sc.w.rl %1, %z4, %2\n"			\
 			"	bnez %1, 0b\n"				\
-			"	fence rw, rw\n"				\
+			RISCV_FULL_BARRIER				\
 			"1:\n"						\
 			: "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)	\
 			: "rJ" ((long)__old), "rJ" (__new)		\
@@ -325,7 +324,7 @@
 			"	bne %0, %z3, 1f\n"			\
 			"	sc.d.rl %1, %z4, %2\n"			\
 			"	bnez %1, 0b\n"				\
-			"	fence rw, rw\n"				\
+			RISCV_FULL_BARRIER				\
 			"1:\n"						\
 			: "=&r" (__ret), "=&r" (__rc), "+A" (*__ptr)	\
 			: "rJ" (__old), "rJ" (__new)			\

arch/riscv/include/asm/fence.h

 #ifndef _ASM_RISCV_FENCE_H
 #define _ASM_RISCV_FENCE_H
 
+#define RISCV_FENCE_ASM(p, s)		"\tfence " #p "," #s "\n"
+#define RISCV_FENCE(p, s) \
+	({ __asm__ __volatile__ (RISCV_FENCE_ASM(p, s) : : : "memory"); })
+
 #ifdef CONFIG_SMP
-#define RISCV_ACQUIRE_BARRIER		"\tfence r , rw\n"
-#define RISCV_RELEASE_BARRIER		"\tfence rw,  w\n"
+#define RISCV_ACQUIRE_BARRIER		RISCV_FENCE_ASM(r, rw)
+#define RISCV_RELEASE_BARRIER		RISCV_FENCE_ASM(rw, w)
+#define RISCV_FULL_BARRIER		RISCV_FENCE_ASM(rw, rw)
 #else
 #define RISCV_ACQUIRE_BARRIER
 #define RISCV_RELEASE_BARRIER
+#define RISCV_FULL_BARRIER
 #endif
 
 #endif	/* _ASM_RISCV_FENCE_H */

arch/riscv/include/asm/io.h

@@ -47,10 +47,10 @@
  * sufficient to ensure this works sanely on controllers that support I/O
  * writes.
  */
-#define __io_pbr()	__asm__ __volatile__ ("fence io,i"  : : : "memory");
-#define __io_par(v)	__asm__ __volatile__ ("fence i,ior" : : : "memory");
-#define __io_pbw()	__asm__ __volatile__ ("fence iow,o" : : : "memory");
-#define __io_paw()	__asm__ __volatile__ ("fence o,io"  : : : "memory");
+#define __io_pbr()	RISCV_FENCE(io, i)
+#define __io_par(v)	RISCV_FENCE(i, ior)
+#define __io_pbw()	RISCV_FENCE(iow, o)
+#define __io_paw()	RISCV_FENCE(o, io)
 
 /*
  * Accesses from a single hart to a single I/O address must be ordered.  This

arch/riscv/include/asm/mmio.h

@@ -12,6 +12,7 @@
 #define _ASM_RISCV_MMIO_H
 
 #include <linux/types.h>
+#include <asm/fence.h>
 #include <asm/mmiowb.h>
 
 /* Generic IO read/write.  These perform native-endian accesses. */
@@ -131,8 +132,8 @@ static inline u64 __raw_readq(const volatile void __iomem *addr)
  * doesn't define any ordering between the memory space and the I/O space.
  */
 #define __io_br()	do {} while (0)
-#define __io_ar(v)	({ __asm__ __volatile__ ("fence i,ir" : : : "memory"); })
-#define __io_bw()	({ __asm__ __volatile__ ("fence w,o" : : : "memory"); })
+#define __io_ar(v)	RISCV_FENCE(i, ir)
+#define __io_bw()	RISCV_FENCE(w, o)
 #define __io_aw()	mmiowb_set_pending()
 
 #define readb(c)	({ u8  __v; __io_br(); __v = readb_cpu(c); __io_ar(__v); __v; })

arch/riscv/include/asm/mmiowb.h

@@ -7,7 +7,7 @@
  * "o,w" is sufficient to ensure that all writes to the device have completed
  * before the write to the spinlock is allowed to commit.
  */
-#define mmiowb()	__asm__ __volatile__ ("fence o,w" : : : "memory");
+#define mmiowb()	RISCV_FENCE(o, w)
 
 #include <linux/smp.h>
 #include <asm-generic/mmiowb.h>