linux-kernel - [PATCH V4 3/3] csky: atomic: Add conditional atomic operations' optimization

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-Id: <20220424072918.2596899-4-guoren@kernel.org>
Date:   Sun, 24 Apr 2022 15:29:18 +0800
From:   guoren@...nel.org
To:     guoren@...nel.org, arnd@...db.de, mark.rutland@....com,
        boqun.feng@...il.com, peterz@...radead.org, will@...nel.org
Cc:     linux-arch@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-csky@...r.kernel.org, Guo Ren <guoren@...ux.alibaba.com>
Subject: [PATCH V4 3/3] csky: atomic: Add conditional atomic operations' optimization

From: Guo Ren <guoren@...ux.alibaba.com>

Add conditional atomic operations' optimization:
 - arch_atomic_fetch_add_unless
 - arch_atomic_inc_unless_negative
 - arch_atomic_dec_unless_positive
 - arch_atomic_dec_if_positive

Comments by Boqun:

FWIW, you probably need to make sure that a barrier instruction inside
an lr/sc loop is a good thing. IIUC, the execution time of a barrier
instruction is determined by the status of store buffers and invalidate
queues (and probably other stuffs), so it may increase the execution
time of the lr/sc loop, and make it unlikely to succeed. But this really
depends on how the arch executes these instructions.

Signed-off-by: Guo Ren <guoren@...ux.alibaba.com>
Signed-off-by: Guo Ren <guoren@...nel.org>
Cc: Boqun Feng <boqun.feng@...il.com>
---
 arch/csky/include/asm/atomic.h | 95 ++++++++++++++++++++++++++++++++++
 1 file changed, 95 insertions(+)

diff --git a/arch/csky/include/asm/atomic.h b/arch/csky/include/asm/atomic.h
index 56c9dc8e91b3..60406ef9c2bb 100644
--- a/arch/csky/include/asm/atomic.h
+++ b/arch/csky/include/asm/atomic.h
@@ -100,6 +100,101 @@ ATOMIC_OPS(xor)
 
 #undef ATOMIC_FETCH_OP
 
+static __always_inline int
+arch_atomic_fetch_add_unless(atomic_t *v, int a, int u)
+{
+	int prev, tmp;
+
+	__asm__ __volatile__ (
+		RELEASE_FENCE
+		"1:	ldex.w		%0, (%3)	\n"
+		"	cmpne		%0, %4		\n"
+		"	bf		2f		\n"
+		"	mov		%1, %0		\n"
+		"	add		%1, %2		\n"
+		"	stex.w		%1, (%3)	\n"
+		"	bez		%1, 1b		\n"
+		FULL_FENCE
+		"2:\n"
+		: "=&r" (prev), "=&r" (tmp)
+		: "r" (a), "r" (&v->counter), "r" (u)
+		: "memory");
+
+	return prev;
+}
+#define arch_atomic_fetch_add_unless arch_atomic_fetch_add_unless
+
+static __always_inline bool
+arch_atomic_inc_unless_negative(atomic_t *v)
+{
+	int rc, tmp;
+
+	__asm__ __volatile__ (
+		RELEASE_FENCE
+		"1:	ldex.w		%0, (%2)	\n"
+		"	movi		%1, 0		\n"
+		"	blz		%0, 2f		\n"
+		"	movi		%1, 1		\n"
+		"	addi		%0, 1		\n"
+		"	stex.w		%0, (%2)	\n"
+		"	bez		%0, 1b		\n"
+		FULL_FENCE
+		"2:\n"
+		: "=&r" (tmp), "=&r" (rc)
+		: "r" (&v->counter)
+		: "memory");
+
+	return tmp ? true : false;
+
+}
+#define arch_atomic_inc_unless_negative arch_atomic_inc_unless_negative
+
+static __always_inline bool
+arch_atomic_dec_unless_positive(atomic_t *v)
+{
+	int rc, tmp;
+
+	__asm__ __volatile__ (
+		RELEASE_FENCE
+		"1:	ldex.w		%0, (%2)	\n"
+		"	movi		%1, 0		\n"
+		"	bhz		%0, 2f		\n"
+		"	movi		%1, 1		\n"
+		"	subi		%0, 1		\n"
+		"	stex.w		%0, (%2)	\n"
+		"	bez		%0, 1b		\n"
+		FULL_FENCE
+		"2:\n"
+		: "=&r" (tmp), "=&r" (rc)
+		: "r" (&v->counter)
+		: "memory");
+
+	return tmp ? true : false;
+}
+#define arch_atomic_dec_unless_positive arch_atomic_dec_unless_positive
+
+static __always_inline int
+arch_atomic_dec_if_positive(atomic_t *v)
+{
+	int dec, tmp;
+
+	__asm__ __volatile__ (
+		RELEASE_FENCE
+		"1:	ldex.w		%0, (%2)	\n"
+		"	subi		%1, %0, 1	\n"
+		"	blz		%1, 2f		\n"
+		"	stex.w		%1, (%2)	\n"
+		"	bez		%1, 1b		\n"
+		FULL_FENCE
+		"2:\n"
+		: "=&r" (dec), "=&r" (tmp)
+		: "r" (&v->counter)
+		: "memory");
+
+	return dec - 1;
+}
+#define arch_atomic_dec_if_positive arch_atomic_dec_if_positive
+
 #define ATOMIC_OP()							\
 static __always_inline							\
 int arch_atomic_xchg_relaxed(atomic_t *v, int n)			\
-- 
2.25.1