[PATCH] arm64: crypto: Add ARM64 CRC32 hw accelerated module

[Yazen Ghannam <yazen.ghannam@linaro.org>]

This module registers a crc32 algorithm and a crc32c algorithm
that use the optional CRC32 and CRC32C instructions in ARMv8.

Tested on AMD Seattle.

Improvement compared to crc32c-generic algorithm:
TCRYPT CRC32C speed test shows ~450% speedup.
Simple dd write tests to btrfs filesystem show ~30% speedup.

Signed-off-by: Yazen Ghannam <yazen.ghannam@linaro.org>
Acked-by: Steve Capper <steve.capper@linaro.org>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
---
 arch/arm64/crypto/Kconfig       |   4 +
 arch/arm64/crypto/Makefile      |   4 +
 arch/arm64/crypto/crc32-arm64.c | 274 ++++++++++++++++++++++++++++++++++++++++
 3 files changed, 282 insertions(+)
 create mode 100644 arch/arm64/crypto/crc32-arm64.c

diff --git a/arch/arm64/crypto/Kconfig b/arch/arm64/crypto/Kconfig
index 5562652..c1a0468 100644
--- a/arch/arm64/crypto/Kconfig
+++ b/arch/arm64/crypto/Kconfig
@@ -50,4 +50,8 @@ config CRYPTO_AES_ARM64_NEON_BLK
 	select CRYPTO_AES
 	select CRYPTO_ABLK_HELPER
 
+config CRYPTO_CRC32_ARM64
+	tristate "CRC32 and CRC32C using optional ARMv8 instructions"
+	depends on ARM64
+	select CRYPTO_HASH
 endif
diff --git a/arch/arm64/crypto/Makefile b/arch/arm64/crypto/Makefile
index a3f935f..5720608 100644
--- a/arch/arm64/crypto/Makefile
+++ b/arch/arm64/crypto/Makefile
@@ -34,5 +34,9 @@ AFLAGS_aes-neon.o	:= -DINTERLEAVE=4
 
 CFLAGS_aes-glue-ce.o	:= -DUSE_V8_CRYPTO_EXTENSIONS
 
+obj-$(CONFIG_CRYPTO_CRC32_ARM64) += crc32-arm64.o
+
+CFLAGS_crc32-arm64.o	:= -mcpu=generic+crc
+
 $(obj)/aes-glue-%.o: $(src)/aes-glue.c FORCE
 	$(call if_changed_rule,cc_o_c)
diff --git a/arch/arm64/crypto/crc32-arm64.c b/arch/arm64/crypto/crc32-arm64.c
new file mode 100644
index 0000000..9499199
--- /dev/null
+++ b/arch/arm64/crypto/crc32-arm64.c
@@ -0,0 +1,274 @@
+/*
+ * crc32-arm64.c - CRC32 and CRC32C using optional ARMv8 instructions
+ *
+ * Module based on crypto/crc32c_generic.c
+ *
+ * CRC32 loop taken from Ed Nevill's Hadoop CRC patch
+ * http://mail-archives.apache.org/mod_mbox/hadoop-common-dev/201406.mbox/%3C1403687030.3355.19.camel%40localhost.localdomain%3E
+ *
+ * Using inline assembly instead of intrinsics in order to be backwards
+ * compatible with older compilers.
+ *
+ * Copyright (C) 2014 Linaro Ltd <yazen.ghannam@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/unaligned/access_ok.h>
+#include <linux/cpufeature.h>
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/string.h>
+
+#include <crypto/internal/hash.h>
+
+MODULE_AUTHOR("Yazen Ghannam <yazen.ghannam@linaro.org>");
+MODULE_DESCRIPTION("CRC32 and CRC32C using optional ARMv8 instructions");
+MODULE_LICENSE("GPL v2");
+
+#define CRC32X(crc, value) __asm__("crc32x %w[c], %w[c], %x[v]":[c]"+r"(crc):[v]"r"(value))
+#define CRC32W(crc, value) __asm__("crc32w %w[c], %w[c], %w[v]":[c]"+r"(crc):[v]"r"(value))
+#define CRC32H(crc, value) __asm__("crc32h %w[c], %w[c], %w[v]":[c]"+r"(crc):[v]"r"(value))
+#define CRC32B(crc, value) __asm__("crc32b %w[c], %w[c], %w[v]":[c]"+r"(crc):[v]"r"(value))
+#define CRC32CX(crc, value) __asm__("crc32cx %w[c], %w[c], %x[v]":[c]"+r"(crc):[v]"r"(value))
+#define CRC32CW(crc, value) __asm__("crc32cw %w[c], %w[c], %w[v]":[c]"+r"(crc):[v]"r"(value))
+#define CRC32CH(crc, value) __asm__("crc32ch %w[c], %w[c], %w[v]":[c]"+r"(crc):[v]"r"(value))
+#define CRC32CB(crc, value) __asm__("crc32cb %w[c], %w[c], %w[v]":[c]"+r"(crc):[v]"r"(value))
+
+static u32 crc32_arm64_le_hw(u32 crc, const u8 *p, unsigned int len)
+{
+	s64 length = len;
+
+	while ((length -= sizeof(u64)) >= 0) {
+		CRC32X(crc, get_unaligned_le64(p));
+		p += sizeof(u64);
+	}
+
+	/* The following is more efficient than the straight loop */
+	if (length & sizeof(u32)) {
+		CRC32W(crc, get_unaligned_le32(p));
+		p += sizeof(u32);
+	}
+	if (length & sizeof(u16)) {
+		CRC32H(crc, get_unaligned_le16(p));
+		p += sizeof(u16);
+	}
+	if (length & sizeof(u8))
+		CRC32B(crc, *p);
+
+	return crc;
+}
+
+static u32 crc32c_arm64_le_hw(u32 crc, const u8 *p, unsigned int len)
+{
+	s64 length = len;
+
+	while ((length -= sizeof(u64)) >= 0) {
+		CRC32CX(crc, get_unaligned_le64(p));
+		p += sizeof(u64);
+	}
+
+	/* The following is more efficient than the straight loop */
+	if (length & sizeof(u32)) {
+		CRC32CW(crc, get_unaligned_le32(p));
+		p += sizeof(u32);
+	}
+	if (length & sizeof(u16)) {
+		CRC32CH(crc, get_unaligned_le16(p));
+		p += sizeof(u16);
+	}
+	if (length & sizeof(u8))
+		CRC32CB(crc, *p);
+
+	return crc;
+}
+
+#define CHKSUM_BLOCK_SIZE	1
+#define CHKSUM_DIGEST_SIZE	4
+
+struct chksum_ctx {
+	u32 key;
+};
+
+struct chksum_desc_ctx {
+	u32 crc;
+};
+
+static int chksum_init(struct shash_desc *desc)
+{
+	struct chksum_ctx *mctx = crypto_shash_ctx(desc->tfm);
+	struct chksum_desc_ctx *ctx = shash_desc_ctx(desc);
+
+	ctx->crc = mctx->key;
+
+	return 0;
+}
+
+/*
+ * Setting the seed allows arbitrary accumulators and flexible XOR policy
+ * If your algorithm starts with ~0, then XOR with ~0 before you set
+ * the seed.
+ */
+static int chksum_setkey(struct crypto_shash *tfm, const u8 *key,
+			 unsigned int keylen)
+{
+	struct chksum_ctx *mctx = crypto_shash_ctx(tfm);
+
+	if (keylen != sizeof(mctx->key)) {
+		crypto_shash_set_flags(tfm, CRYPTO_TFM_RES_BAD_KEY_LEN);
+		return -EINVAL;
+	}
+	mctx->key = get_unaligned_le32(key);
+	return 0;
+}
+
+static int chksum_update(struct shash_desc *desc, const u8 *data,
+			 unsigned int length)
+{
+	struct chksum_desc_ctx *ctx = shash_desc_ctx(desc);
+
+	ctx->crc = crc32_arm64_le_hw(ctx->crc, data, length);
+	return 0;
+}
+
+static int chksumc_update(struct shash_desc *desc, const u8 *data,
+			 unsigned int length)
+{
+	struct chksum_desc_ctx *ctx = shash_desc_ctx(desc);
+
+	ctx->crc = crc32c_arm64_le_hw(ctx->crc, data, length);
+	return 0;
+}
+
+static int chksum_final(struct shash_desc *desc, u8 *out)
+{
+	struct chksum_desc_ctx *ctx = shash_desc_ctx(desc);
+
+	put_unaligned_le32(~ctx->crc, out);
+	return 0;
+}
+
+static int __chksum_finup(u32 crc, const u8 *data, unsigned int len, u8 *out)
+{
+	put_unaligned_le32(~crc32_arm64_le_hw(crc, data, len), out);
+	return 0;
+}
+
+static int __chksumc_finup(u32 crc, const u8 *data, unsigned int len, u8 *out)
+{
+	put_unaligned_le32(~crc32c_arm64_le_hw(crc, data, len), out);
+	return 0;
+}
+
+static int chksum_finup(struct shash_desc *desc, const u8 *data,
+			unsigned int len, u8 *out)
+{
+	struct chksum_desc_ctx *ctx = shash_desc_ctx(desc);
+
+	return __chksum_finup(ctx->crc, data, len, out);
+}
+
+static int chksumc_finup(struct shash_desc *desc, const u8 *data,
+			unsigned int len, u8 *out)
+{
+	struct chksum_desc_ctx *ctx = shash_desc_ctx(desc);
+
+	return __chksumc_finup(ctx->crc, data, len, out);
+}
+
+static int chksum_digest(struct shash_desc *desc, const u8 *data,
+			 unsigned int length, u8 *out)
+{
+	struct chksum_ctx *mctx = crypto_shash_ctx(desc->tfm);
+
+	return __chksum_finup(mctx->key, data, length, out);
+}
+
+static int chksumc_digest(struct shash_desc *desc, const u8 *data,
+			 unsigned int length, u8 *out)
+{
+	struct chksum_ctx *mctx = crypto_shash_ctx(desc->tfm);
+
+	return __chksumc_finup(mctx->key, data, length, out);
+}
+
+static int crc32_cra_init(struct crypto_tfm *tfm)
+{
+	struct chksum_ctx *mctx = crypto_tfm_ctx(tfm);
+
+	mctx->key = ~0;
+	return 0;
+}
+
+static struct shash_alg crc32_alg = {
+	.digestsize		=	CHKSUM_DIGEST_SIZE,
+	.setkey			=	chksum_setkey,
+	.init			=	chksum_init,
+	.update			=	chksum_update,
+	.final			=	chksum_final,
+	.finup			=	chksum_finup,
+	.digest			=	chksum_digest,
+	.descsize		=	sizeof(struct chksum_desc_ctx),
+	.base			=	{
+		.cra_name		=	"crc32",
+		.cra_driver_name	=	"crc32-arm64-hw",
+		.cra_priority		=	300,
+		.cra_blocksize		=	CHKSUM_BLOCK_SIZE,
+		.cra_alignmask		=	0,
+		.cra_ctxsize		=	sizeof(struct chksum_ctx),
+		.cra_module		=	THIS_MODULE,
+		.cra_init		=	crc32_cra_init,
+	}
+};
+
+static struct shash_alg crc32c_alg = {
+	.digestsize		=	CHKSUM_DIGEST_SIZE,
+	.setkey			=	chksum_setkey,
+	.init			=	chksum_init,
+	.update			=	chksumc_update,
+	.final			=	chksum_final,
+	.finup			=	chksumc_finup,
+	.digest			=	chksumc_digest,
+	.descsize		=	sizeof(struct chksum_desc_ctx),
+	.base			=	{
+		.cra_name		=	"crc32c",
+		.cra_driver_name	=	"crc32c-arm64-hw",
+		.cra_priority		=	300,
+		.cra_blocksize		=	CHKSUM_BLOCK_SIZE,
+		.cra_alignmask		=	0,
+		.cra_ctxsize		=	sizeof(struct chksum_ctx),
+		.cra_module		=	THIS_MODULE,
+		.cra_init		=	crc32_cra_init,
+	}
+};
+
+static int __init crc32_mod_init(void)
+{
+	int err;
+
+	err = crypto_register_shash(&crc32_alg);
+
+	if (err)
+		return err;
+
+	err = crypto_register_shash(&crc32c_alg);
+
+	if (err) {
+		crypto_unregister_shash(&crc32_alg);
+		return err;
+	}
+
+	return 0;
+}
+
+static void __exit crc32_mod_exit(void)
+{
+	crypto_unregister_shash(&crc32_alg);
+	crypto_unregister_shash(&crc32c_alg);
+}
+
+module_cpu_feature_match(CRC32, crc32_mod_init);
+module_exit(crc32_mod_exit);
-- 
2.1.0