[PATCH v3 05/10] crypto: riscv - add vector crypto accelerated ChaCha20
Eric Biggers
ebiggers at kernel.org
Sun Jan 21 16:19:16 PST 2024
From: Jerry Shih <jerry.shih at sifive.com>
Add an implementation of ChaCha20 using the Zvkb extension. The
assembly code is derived from OpenSSL code (openssl/openssl#21923) that
was dual-licensed so that it could be reused in the kernel.
Nevertheless, the assembly has been significantly reworked for
integration with the kernel, for example by using a regular .S file
instead of the so-called perlasm, using the assembler instead of bare
'.inst', and reducing code duplication.
Signed-off-by: Jerry Shih <jerry.shih at sifive.com>
Co-developed-by: Eric Biggers <ebiggers at google.com>
Signed-off-by: Eric Biggers <ebiggers at google.com>
---
arch/riscv/crypto/Kconfig | 11 +
arch/riscv/crypto/Makefile | 3 +
arch/riscv/crypto/chacha-riscv64-glue.c | 101 ++++++++
arch/riscv/crypto/chacha-riscv64-zvkb.S | 294 ++++++++++++++++++++++++
4 files changed, 409 insertions(+)
create mode 100644 arch/riscv/crypto/chacha-riscv64-glue.c
create mode 100644 arch/riscv/crypto/chacha-riscv64-zvkb.S
diff --git a/arch/riscv/crypto/Kconfig b/arch/riscv/crypto/Kconfig
index ebe805fa3f5f7..cb59e1d954952 100644
--- a/arch/riscv/crypto/Kconfig
+++ b/arch/riscv/crypto/Kconfig
@@ -11,11 +11,22 @@ config CRYPTO_AES_RISCV64
help
Block cipher: AES cipher algorithms
Length-preserving ciphers: AES with ECB, CBC, CTR, XTS
Architecture: riscv64 using:
- Zvkned vector crypto extension
- Zvbb vector extension (XTS)
- Zvkb vector crypto extension (CTR)
- Zvkg vector crypto extension (XTS)
+config CRYPTO_CHACHA_RISCV64
+ tristate "Ciphers: ChaCha"
+ depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO
+ select CRYPTO_SKCIPHER
+ select CRYPTO_LIB_CHACHA_GENERIC
+ help
+ Length-preserving ciphers: ChaCha20 stream cipher algorithm
+
+ Architecture: riscv64 using:
+ - Zvkb vector crypto extension
+
endmenu
diff --git a/arch/riscv/crypto/Makefile b/arch/riscv/crypto/Makefile
index 44922df7d182f..ee994c8e65509 100644
--- a/arch/riscv/crypto/Makefile
+++ b/arch/riscv/crypto/Makefile
@@ -1,5 +1,8 @@
# SPDX-License-Identifier: GPL-2.0-only
obj-$(CONFIG_CRYPTO_AES_RISCV64) += aes-riscv64.o
aes-riscv64-y := aes-riscv64-glue.o aes-riscv64-zvkned.o \
aes-riscv64-zvkned-zvbb-zvkg.o aes-riscv64-zvkned-zvkb.o
+
+obj-$(CONFIG_CRYPTO_CHACHA_RISCV64) += chacha-riscv64.o
+chacha-riscv64-y := chacha-riscv64-glue.o chacha-riscv64-zvkb.o
diff --git a/arch/riscv/crypto/chacha-riscv64-glue.c b/arch/riscv/crypto/chacha-riscv64-glue.c
new file mode 100644
index 0000000000000..10b46f36375af
--- /dev/null
+++ b/arch/riscv/crypto/chacha-riscv64-glue.c
@@ -0,0 +1,101 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * ChaCha20 using the RISC-V vector crypto extensions
+ *
+ * Copyright (C) 2023 SiFive, Inc.
+ * Author: Jerry Shih <jerry.shih at sifive.com>
+ */
+
+#include <asm/simd.h>
+#include <asm/vector.h>
+#include <crypto/internal/chacha.h>
+#include <crypto/internal/skcipher.h>
+#include <linux/linkage.h>
+#include <linux/module.h>
+
+asmlinkage void chacha20_zvkb(const u32 key[8], const u8 *in, u8 *out,
+ size_t len, const u32 iv[4]);
+
+static int riscv64_chacha20_crypt(struct skcipher_request *req)
+{
+ u32 iv[CHACHA_IV_SIZE / sizeof(u32)];
+ u8 block_buffer[CHACHA_BLOCK_SIZE];
+ struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+ const struct chacha_ctx *ctx = crypto_skcipher_ctx(tfm);
+ struct skcipher_walk walk;
+ unsigned int nbytes;
+ unsigned int tail_bytes;
+ int err;
+
+ iv[0] = get_unaligned_le32(req->iv);
+ iv[1] = get_unaligned_le32(req->iv + 4);
+ iv[2] = get_unaligned_le32(req->iv + 8);
+ iv[3] = get_unaligned_le32(req->iv + 12);
+
+ err = skcipher_walk_virt(&walk, req, false);
+ while (walk.nbytes) {
+ nbytes = walk.nbytes & ~(CHACHA_BLOCK_SIZE - 1);
+ tail_bytes = walk.nbytes & (CHACHA_BLOCK_SIZE - 1);
+ kernel_vector_begin();
+ if (nbytes) {
+ chacha20_zvkb(ctx->key, walk.src.virt.addr,
+ walk.dst.virt.addr, nbytes, iv);
+ iv[0] += nbytes / CHACHA_BLOCK_SIZE;
+ }
+ if (walk.nbytes == walk.total && tail_bytes > 0) {
+ memcpy(block_buffer, walk.src.virt.addr + nbytes,
+ tail_bytes);
+ chacha20_zvkb(ctx->key, block_buffer, block_buffer,
+ CHACHA_BLOCK_SIZE, iv);
+ memcpy(walk.dst.virt.addr + nbytes, block_buffer,
+ tail_bytes);
+ tail_bytes = 0;
+ }
+ kernel_vector_end();
+
+ err = skcipher_walk_done(&walk, tail_bytes);
+ }
+
+ return err;
+}
+
+static struct skcipher_alg riscv64_chacha_alg = {
+ .setkey = chacha20_setkey,
+ .encrypt = riscv64_chacha20_crypt,
+ .decrypt = riscv64_chacha20_crypt,
+ .min_keysize = CHACHA_KEY_SIZE,
+ .max_keysize = CHACHA_KEY_SIZE,
+ .ivsize = CHACHA_IV_SIZE,
+ .chunksize = CHACHA_BLOCK_SIZE,
+ .walksize = 4 * CHACHA_BLOCK_SIZE,
+ .base = {
+ .cra_blocksize = 1,
+ .cra_ctxsize = sizeof(struct chacha_ctx),
+ .cra_priority = 300,
+ .cra_name = "chacha20",
+ .cra_driver_name = "chacha20-riscv64-zvkb",
+ .cra_module = THIS_MODULE,
+ },
+};
+
+static int __init riscv64_chacha_mod_init(void)
+{
+ if (riscv_isa_extension_available(NULL, ZVKB) &&
+ riscv_vector_vlen() >= 128)
+ return crypto_register_skcipher(&riscv64_chacha_alg);
+
+ return -ENODEV;
+}
+
+static void __exit riscv64_chacha_mod_exit(void)
+{
+ crypto_unregister_skcipher(&riscv64_chacha_alg);
+}
+
+module_init(riscv64_chacha_mod_init);
+module_exit(riscv64_chacha_mod_exit);
+
+MODULE_DESCRIPTION("ChaCha20 (RISC-V accelerated)");
+MODULE_AUTHOR("Jerry Shih <jerry.shih at sifive.com>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS_CRYPTO("chacha20");
diff --git a/arch/riscv/crypto/chacha-riscv64-zvkb.S b/arch/riscv/crypto/chacha-riscv64-zvkb.S
new file mode 100644
index 0000000000000..bf057737ac693
--- /dev/null
+++ b/arch/riscv/crypto/chacha-riscv64-zvkb.S
@@ -0,0 +1,294 @@
+/* SPDX-License-Identifier: Apache-2.0 OR BSD-2-Clause */
+//
+// This file is dual-licensed, meaning that you can use it under your
+// choice of either of the following two licenses:
+//
+// Copyright 2023 The OpenSSL Project Authors. All Rights Reserved.
+//
+// Licensed under the Apache License 2.0 (the "License"). You can obtain
+// a copy in the file LICENSE in the source distribution or at
+// https://www.openssl.org/source/license.html
+//
+// or
+//
+// Copyright (c) 2023, Jerry Shih <jerry.shih at sifive.com>
+// Copyright 2024 Google LLC
+// All rights reserved.
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions
+// are met:
+// 1. Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+// 2. Redistributions in binary form must reproduce the above copyright
+// notice, this list of conditions and the following disclaimer in the
+// documentation and/or other materials provided with the distribution.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// The generated code of this file depends on the following RISC-V extensions:
+// - RV64I
+// - RISC-V Vector ('V') with VLEN >= 128
+// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb')
+
+#include <linux/linkage.h>
+
+.text
+.option arch, +zvkb
+
+#define KEYP a0
+#define INP a1
+#define OUTP a2
+#define LEN a3
+#define IVP a4
+
+#define CONSTS0 a5
+#define CONSTS1 a6
+#define CONSTS2 a7
+#define CONSTS3 t0
+#define TMP t1
+#define VL t2
+#define STRIDE t3
+#define NROUNDS t4
+#define KEY0 s0
+#define KEY1 s1
+#define KEY2 s2
+#define KEY3 s3
+#define KEY4 s4
+#define KEY5 s5
+#define KEY6 s6
+#define KEY7 s7
+#define COUNTER s8
+#define NONCE0 s9
+#define NONCE1 s10
+#define NONCE2 s11
+
+.macro chacha_round a0, b0, c0, d0, a1, b1, c1, d1, \
+ a2, b2, c2, d2, a3, b3, c3, d3
+ // a += b; d ^= a; d = rol(d, 16);
+ vadd.vv \a0, \a0, \b0
+ vadd.vv \a1, \a1, \b1
+ vadd.vv \a2, \a2, \b2
+ vadd.vv \a3, \a3, \b3
+ vxor.vv \d0, \d0, \a0
+ vxor.vv \d1, \d1, \a1
+ vxor.vv \d2, \d2, \a2
+ vxor.vv \d3, \d3, \a3
+ vror.vi \d0, \d0, 32 - 16
+ vror.vi \d1, \d1, 32 - 16
+ vror.vi \d2, \d2, 32 - 16
+ vror.vi \d3, \d3, 32 - 16
+
+ // c += d; b ^= c; b = rol(b, 12);
+ vadd.vv \c0, \c0, \d0
+ vadd.vv \c1, \c1, \d1
+ vadd.vv \c2, \c2, \d2
+ vadd.vv \c3, \c3, \d3
+ vxor.vv \b0, \b0, \c0
+ vxor.vv \b1, \b1, \c1
+ vxor.vv \b2, \b2, \c2
+ vxor.vv \b3, \b3, \c3
+ vror.vi \b0, \b0, 32 - 12
+ vror.vi \b1, \b1, 32 - 12
+ vror.vi \b2, \b2, 32 - 12
+ vror.vi \b3, \b3, 32 - 12
+
+ // a += b; d ^= a; d = rol(d, 8);
+ vadd.vv \a0, \a0, \b0
+ vadd.vv \a1, \a1, \b1
+ vadd.vv \a2, \a2, \b2
+ vadd.vv \a3, \a3, \b3
+ vxor.vv \d0, \d0, \a0
+ vxor.vv \d1, \d1, \a1
+ vxor.vv \d2, \d2, \a2
+ vxor.vv \d3, \d3, \a3
+ vror.vi \d0, \d0, 32 - 8
+ vror.vi \d1, \d1, 32 - 8
+ vror.vi \d2, \d2, 32 - 8
+ vror.vi \d3, \d3, 32 - 8
+
+ // c += d; b ^= c; b = rol(b, 7);
+ vadd.vv \c0, \c0, \d0
+ vadd.vv \c1, \c1, \d1
+ vadd.vv \c2, \c2, \d2
+ vadd.vv \c3, \c3, \d3
+ vxor.vv \b0, \b0, \c0
+ vxor.vv \b1, \b1, \c1
+ vxor.vv \b2, \b2, \c2
+ vxor.vv \b3, \b3, \c3
+ vror.vi \b0, \b0, 32 - 7
+ vror.vi \b1, \b1, 32 - 7
+ vror.vi \b2, \b2, 32 - 7
+ vror.vi \b3, \b3, 32 - 7
+.endm
+
+// void chacha20_zvkb(const u32 key[8], const u8 *in, u8 *out, size_t len,
+// const u32 iv[4]);
+//
+// |len| must be nonzero and a multiple of 64 (CHACHA_BLOCK_SIZE).
+// The counter is treated as 32-bit, following the RFC7539 convention.
+SYM_FUNC_START(chacha20_zvkb)
+ srli LEN, LEN, 6 // Bytes to blocks
+
+ addi sp, sp, -96
+ sd s0, 0(sp)
+ sd s1, 8(sp)
+ sd s2, 16(sp)
+ sd s3, 24(sp)
+ sd s4, 32(sp)
+ sd s5, 40(sp)
+ sd s6, 48(sp)
+ sd s7, 56(sp)
+ sd s8, 64(sp)
+ sd s9, 72(sp)
+ sd s10, 80(sp)
+ sd s11, 88(sp)
+
+ li STRIDE, 64
+
+ // Set up the initial state matrix in scalar registers.
+ li CONSTS0, 0x61707865 // "expa" little endian
+ li CONSTS1, 0x3320646e // "nd 3" little endian
+ li CONSTS2, 0x79622d32 // "2-by" little endian
+ li CONSTS3, 0x6b206574 // "te k" little endian
+ lw KEY0, 0(KEYP)
+ lw KEY1, 4(KEYP)
+ lw KEY2, 8(KEYP)
+ lw KEY3, 12(KEYP)
+ lw KEY4, 16(KEYP)
+ lw KEY5, 20(KEYP)
+ lw KEY6, 24(KEYP)
+ lw KEY7, 28(KEYP)
+ lw COUNTER, 0(IVP)
+ lw NONCE0, 4(IVP)
+ lw NONCE1, 8(IVP)
+ lw NONCE2, 12(IVP)
+
+.Lblock_loop:
+ // Set vl to the number of blocks to process in this iteration.
+ vsetvli VL, LEN, e32, m1, ta, ma
+
+ // Set up the initial state matrix for the next VL blocks in v0-v15.
+ // v{i} holds the i'th 32-bit word of the state matrix for all blocks.
+ // Note that only the counter word, at index 12, differs across blocks.
+ vmv.v.x v0, CONSTS0
+ vmv.v.x v1, CONSTS1
+ vmv.v.x v2, CONSTS2
+ vmv.v.x v3, CONSTS3
+ vmv.v.x v4, KEY0
+ vmv.v.x v5, KEY1
+ vmv.v.x v6, KEY2
+ vmv.v.x v7, KEY3
+ vmv.v.x v8, KEY4
+ vmv.v.x v9, KEY5
+ vmv.v.x v10, KEY6
+ vmv.v.x v11, KEY7
+ vid.v v12
+ vadd.vx v12, v12, COUNTER
+ vmv.v.x v13, NONCE0
+ vmv.v.x v14, NONCE1
+ vmv.v.x v15, NONCE2
+
+ // Load the first half of the input data for each block into v16-v23.
+ // v{16+i} holds the i'th 32-bit word for all blocks.
+ vlsseg8e32.v v16, (INP), STRIDE
+
+ li NROUNDS, 20
+.Lnext_doubleround:
+ addi NROUNDS, NROUNDS, -2
+ // column round
+ chacha_round v0, v4, v8, v12, v1, v5, v9, v13, \
+ v2, v6, v10, v14, v3, v7, v11, v15
+ // diagonal round
+ chacha_round v0, v5, v10, v15, v1, v6, v11, v12, \
+ v2, v7, v8, v13, v3, v4, v9, v14
+ bnez NROUNDS, .Lnext_doubleround
+
+ // Load the second half of the input data for each block into v24-v31.
+ // v{24+i} holds the {8+i}'th 32-bit word for all blocks.
+ addi TMP, INP, 32
+ vlsseg8e32.v v24, (TMP), STRIDE
+
+ // Finalize the first half of the keystream for each block.
+ vadd.vx v0, v0, CONSTS0
+ vadd.vx v1, v1, CONSTS1
+ vadd.vx v2, v2, CONSTS2
+ vadd.vx v3, v3, CONSTS3
+ vadd.vx v4, v4, KEY0
+ vadd.vx v5, v5, KEY1
+ vadd.vx v6, v6, KEY2
+ vadd.vx v7, v7, KEY3
+
+ // Encrypt/decrypt the first half of the data for each block.
+ vxor.vv v16, v16, v0
+ vxor.vv v17, v17, v1
+ vxor.vv v18, v18, v2
+ vxor.vv v19, v19, v3
+ vxor.vv v20, v20, v4
+ vxor.vv v21, v21, v5
+ vxor.vv v22, v22, v6
+ vxor.vv v23, v23, v7
+
+ // Store the first half of the output data for each block.
+ vssseg8e32.v v16, (OUTP), STRIDE
+
+ // Finalize the second half of the keystream for each block.
+ vadd.vx v8, v8, KEY4
+ vadd.vx v9, v9, KEY5
+ vadd.vx v10, v10, KEY6
+ vadd.vx v11, v11, KEY7
+ vid.v v0
+ vadd.vx v12, v12, COUNTER
+ vadd.vx v13, v13, NONCE0
+ vadd.vx v14, v14, NONCE1
+ vadd.vx v15, v15, NONCE2
+ vadd.vv v12, v12, v0
+
+ // Encrypt/decrypt the second half of the data for each block.
+ vxor.vv v24, v24, v8
+ vxor.vv v25, v25, v9
+ vxor.vv v26, v26, v10
+ vxor.vv v27, v27, v11
+ vxor.vv v29, v29, v13
+ vxor.vv v28, v28, v12
+ vxor.vv v30, v30, v14
+ vxor.vv v31, v31, v15
+
+ // Store the second half of the output data for each block.
+ addi TMP, OUTP, 32
+ vssseg8e32.v v24, (TMP), STRIDE
+
+ // Update the counter, the remaining number of blocks, and the input and
+ // output pointers according to the number of blocks processed (VL).
+ add COUNTER, COUNTER, VL
+ sub LEN, LEN, VL
+ slli TMP, VL, 6
+ add OUTP, OUTP, TMP
+ add INP, INP, TMP
+ bnez LEN, .Lblock_loop
+
+ ld s0, 0(sp)
+ ld s1, 8(sp)
+ ld s2, 16(sp)
+ ld s3, 24(sp)
+ ld s4, 32(sp)
+ ld s5, 40(sp)
+ ld s6, 48(sp)
+ ld s7, 56(sp)
+ ld s8, 64(sp)
+ ld s9, 72(sp)
+ ld s10, 80(sp)
+ ld s11, 88(sp)
+ addi sp, sp, 96
+ ret
+SYM_FUNC_END(chacha20_zvkb)
--
2.43.0
More information about the linux-riscv
mailing list