Received-SPF: pass (google.com: domain of linux-crypto-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Uros Bizjak <ubizjak@gmail.com>
To:     linux-crypto@vger.kernel.org, x86@kernel.org,
        linux-kernel@vger.kernel.org
Cc:     Uros Bizjak <ubizjak@gmail.com>,
        Herbert Xu <herbert@gondor.apana.org.au>,
        "David S. Miller" <davem@davemloft.net>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        "H. Peter Anvin" <hpa@zytor.com>
Subject: [PATCH v2] crypto/x86: Use CRC32 mnemonic in crc32c-intel_glue.c
Date:   Wed,  5 Aug 2020 13:17:29 +0200
Message-Id: <20200805111729.359078-1-ubizjak@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-crypto-owner@vger.kernel.org
Precedence: bulk

Current minimum required version of binutils is 2.23,
which supports CRC32 instruction mnemonic.

Replace the byte-wise specification of CRC32 with this proper mnemonic.
The compiler is now able to pass memory operand to the instruction,
so there is no need for a temporary register anymore.

Some examples of the improvement:

 12a:	48 8b 08             	mov    (%rax),%rcx
 12d:	f2 48 0f 38 f1 f1    	crc32q %rcx,%rsi
 133:	48 83 c0 08          	add    $0x8,%rax
 137:	48 39 d0             	cmp    %rdx,%rax
 13a:	75 ee                	jne    12a <crc32c_intel_update+0x1a>

to:

 125:	f2 48 0f 38 f1 06    	crc32q (%rsi),%rax
 12b:	48 83 c6 08          	add    $0x8,%rsi
 12f:	48 39 d6             	cmp    %rdx,%rsi
 132:	75 f1                	jne    125 <crc32c_intel_update+0x15>

and:

 146:	0f b6 08             	movzbl (%rax),%ecx
 149:	f2 0f 38 f0 f1       	crc32b %cl,%esi
 14e:	48 83 c0 01          	add    $0x1,%rax
 152:	48 39 d0             	cmp    %rdx,%rax
 155:	75 ef                	jne    146 <crc32c_intel_update+0x36>

to:

 13b:	f2 0f 38 f0 02       	crc32b (%rdx),%eax
 140:	48 83 c2 01          	add    $0x1,%rdx
 144:	48 39 ca             	cmp    %rcx,%rdx
 147:	75 f2                	jne    13b <crc32c_intel_update+0x2b>

As the compiler has some more freedom w.r.t. register allocation,
there is also a couple of reg-reg moves removed.

There are no hidden states for CRC32 insn, so there is no need to mark
assembly as volatile.

v2: Introduce CRC32_INST define.

Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: "David S. Miller" <davem@davemloft.net>
CC: Thomas Gleixner <tglx@linutronix.de>
CC: Ingo Molnar <mingo@redhat.com>
CC: Borislav Petkov <bp@alien8.de>
CC: "H. Peter Anvin" <hpa@zytor.com>
---
 arch/x86/crypto/crc32c-intel_glue.c | 18 ++++++------------
 1 file changed, 6 insertions(+), 12 deletions(-)

diff --git a/arch/x86/crypto/crc32c-intel_glue.c b/arch/x86/crypto/crc32c-intel_glue.c
index d2d069bd459b..feccb5254c7e 100644
--- a/arch/x86/crypto/crc32c-intel_glue.c
+++ b/arch/x86/crypto/crc32c-intel_glue.c
@@ -28,9 +28,9 @@
 #define SCALE_F	sizeof(unsigned long)
 
 #ifdef CONFIG_X86_64
-#define REX_PRE "0x48, "
+#define CRC32_INST "crc32q %1, %q0"
 #else
-#define REX_PRE
+#define CRC32_INST "crc32l %1, %0"
 #endif
 
 #ifdef CONFIG_X86_64
@@ -48,11 +48,8 @@ asmlinkage unsigned int crc_pcl(const u8 *buffer, int len,
 static u32 crc32c_intel_le_hw_byte(u32 crc, unsigned char const *data, size_t length)
 {
 	while (length--) {
-		__asm__ __volatile__(
-			".byte 0xf2, 0xf, 0x38, 0xf0, 0xf1"
-			:"=S"(crc)
-			:"0"(crc), "c"(*data)
-		);
+		asm("crc32b %1, %0"
+		    : "+r" (crc) : "rm" (*data));
 		data++;
 	}
 
@@ -66,11 +63,8 @@ static u32 __pure crc32c_intel_le_hw(u32 crc, unsigned char const *p, size_t len
 	unsigned long *ptmp = (unsigned long *)p;
 
 	while (iquotient--) {
-		__asm__ __volatile__(
-			".byte 0xf2, " REX_PRE "0xf, 0x38, 0xf1, 0xf1;"
-			:"=S"(crc)
-			:"0"(crc), "c"(*ptmp)
-		);
+		asm(CRC32_INST
+		    : "+r" (crc) : "rm" (*ptmp));
 		ptmp++;
 	}
 
-- 
2.26.2