Received-SPF: pass (google.com: domain of linux-kernel+bounces-64472-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Tue, 13 Feb 2024 22:39:41 +0000
In-Reply-To: <20240213223810.2595804-1-ericchancf@google.com>
Precedence: bulk
Mime-Version: 1.0
References: <20240213223810.2595804-1-ericchancf@google.com>
Message-ID: <20240213223941.2597526-1-ericchancf@google.com>
Subject: [PATCH v5 3/4] riscv/barrier: Consolidate fence definitions
From: Eric Chan <ericchancf@google.com>
To: paul.walmsley@sifive.com, palmer@dabbelt.com, aou@eecs.berkeley.edu
Cc: linux-riscv@lists.infradead.org, linux-kernel@vger.kernel.org, 
	ericchancf@google.com
Content-Type: text/plain; charset="UTF-8"

Disparate fence implementations are consolidated into fence.h.
Also introduce RISCV_FENCE_ASM to make fence macro more reusable.

Signed-off-by: Eric Chan <ericchancf@google.com>
---
v4 -> v5: keep  __atomic_acquire_fence and __atomic_release_fence
omit-the-fence-on-uniprocessor optimization, and fix the typo of RISCV_RELEASE_BARRIER
when spliting the patch in v3.

 arch/riscv/include/asm/atomic.h  |  1 -
 arch/riscv/include/asm/barrier.h |  3 +--
 arch/riscv/include/asm/cmpxchg.h |  1 -
 arch/riscv/include/asm/fence.h   | 10 +++++++---
 arch/riscv/include/asm/io.h      |  8 ++++----
 arch/riscv/include/asm/mmio.h    |  5 +++--
 arch/riscv/include/asm/mmiowb.h  |  2 +-
 7 files changed, 16 insertions(+), 14 deletions(-)

diff --git a/arch/riscv/include/asm/atomic.h b/arch/riscv/include/asm/atomic.h
index 31e6e2e7cc18..0e0522e588ca 100644
--- a/arch/riscv/include/asm/atomic.h
+++ b/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")
diff --git a/arch/riscv/include/asm/barrier.h b/arch/riscv/include/asm/barrier.h
index 4c49a8ff2c68..4f4743d7440d 100644
--- a/arch/riscv/include/asm/barrier.h
+++ b/arch/riscv/include/asm/barrier.h
@@ -11,13 +11,12 @@
 #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)
diff --git a/arch/riscv/include/asm/cmpxchg.h b/arch/riscv/include/asm/cmpxchg.h
index a608e4d1a0a4..2fee65cc8443 100644
--- a/arch/riscv/include/asm/cmpxchg.h
+++ b/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)					\
diff --git a/arch/riscv/include/asm/fence.h b/arch/riscv/include/asm/fence.h
index 6c26c44dfcd6..6bcd80325dfc 100644
--- a/arch/riscv/include/asm/fence.h
+++ b/arch/riscv/include/asm/fence.h
@@ -1,10 +1,14 @@
 #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_FULL_BARRIER		"\tfence rw, rw\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
diff --git a/arch/riscv/include/asm/io.h b/arch/riscv/include/asm/io.h
index 42497d487a17..1c5c641075d2 100644
--- a/arch/riscv/include/asm/io.h
+++ b/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
diff --git a/arch/riscv/include/asm/mmio.h b/arch/riscv/include/asm/mmio.h
index 4c58ee7f95ec..06cadfd7a237 100644
--- a/arch/riscv/include/asm/mmio.h
+++ b/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; })
diff --git a/arch/riscv/include/asm/mmiowb.h b/arch/riscv/include/asm/mmiowb.h
index 0b2333e71fdc..52ce4a399d9b 100644
--- a/arch/riscv/include/asm/mmiowb.h
+++ b/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>
--
2.43.0.687.g38aa6559b0-goog