diff options
Diffstat (limited to 'arch/riscv/crypto')
| -rw-r--r-- | arch/riscv/crypto/Kconfig | 93 | ||||
| -rw-r--r-- | arch/riscv/crypto/Makefile | 23 | ||||
| -rw-r--r-- | arch/riscv/crypto/aes-macros.S | 156 | ||||
| -rw-r--r-- | arch/riscv/crypto/aes-riscv64-glue.c | 637 | ||||
| -rw-r--r-- | arch/riscv/crypto/aes-riscv64-zvkned-zvbb-zvkg.S | 312 | ||||
| -rw-r--r-- | arch/riscv/crypto/aes-riscv64-zvkned-zvkb.S | 146 | ||||
| -rw-r--r-- | arch/riscv/crypto/aes-riscv64-zvkned.S | 339 | ||||
| -rw-r--r-- | arch/riscv/crypto/chacha-riscv64-glue.c | 101 | ||||
| -rw-r--r-- | arch/riscv/crypto/chacha-riscv64-zvkb.S | 294 | ||||
| -rw-r--r-- | arch/riscv/crypto/ghash-riscv64-glue.c | 168 | ||||
| -rw-r--r-- | arch/riscv/crypto/ghash-riscv64-zvkg.S | 72 | ||||
| -rw-r--r-- | arch/riscv/crypto/sha256-riscv64-glue.c | 137 | ||||
| -rw-r--r-- | arch/riscv/crypto/sha256-riscv64-zvknha_or_zvknhb-zvkb.S | 225 | ||||
| -rw-r--r-- | arch/riscv/crypto/sha512-riscv64-glue.c | 133 | ||||
| -rw-r--r-- | arch/riscv/crypto/sha512-riscv64-zvknhb-zvkb.S | 203 | ||||
| -rw-r--r-- | arch/riscv/crypto/sm3-riscv64-glue.c | 112 | ||||
| -rw-r--r-- | arch/riscv/crypto/sm3-riscv64-zvksh-zvkb.S | 123 | ||||
| -rw-r--r-- | arch/riscv/crypto/sm4-riscv64-glue.c | 107 | ||||
| -rw-r--r-- | arch/riscv/crypto/sm4-riscv64-zvksed-zvkb.S | 117 |
19 files changed, 3498 insertions, 0 deletions
diff --git a/arch/riscv/crypto/Kconfig b/arch/riscv/crypto/Kconfig new file mode 100644 index 0000000000..ad58dad9a5 --- /dev/null +++ b/arch/riscv/crypto/Kconfig @@ -0,0 +1,93 @@ +# SPDX-License-Identifier: GPL-2.0 + +menu "Accelerated Cryptographic Algorithms for CPU (riscv)" + +config CRYPTO_AES_RISCV64 + tristate "Ciphers: AES, modes: ECB, CBC, CTS, CTR, XTS" + depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO + select CRYPTO_ALGAPI + select CRYPTO_LIB_AES + select CRYPTO_SKCIPHER + help + Block cipher: AES cipher algorithms + Length-preserving ciphers: AES with ECB, CBC, CTS, 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 + +config CRYPTO_GHASH_RISCV64 + tristate "Hash functions: GHASH" + depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO + select CRYPTO_GCM + help + GCM GHASH function (NIST SP 800-38D) + + Architecture: riscv64 using: + - Zvkg vector crypto extension + +config CRYPTO_SHA256_RISCV64 + tristate "Hash functions: SHA-224 and SHA-256" + depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO + select CRYPTO_SHA256 + help + SHA-224 and SHA-256 secure hash algorithm (FIPS 180) + + Architecture: riscv64 using: + - Zvknha or Zvknhb vector crypto extensions + - Zvkb vector crypto extension + +config CRYPTO_SHA512_RISCV64 + tristate "Hash functions: SHA-384 and SHA-512" + depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO + select CRYPTO_SHA512 + help + SHA-384 and SHA-512 secure hash algorithm (FIPS 180) + + Architecture: riscv64 using: + - Zvknhb vector crypto extension + - Zvkb vector crypto extension + +config CRYPTO_SM3_RISCV64 + tristate "Hash functions: SM3 (ShangMi 3)" + depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO + select CRYPTO_HASH + select CRYPTO_SM3 + help + SM3 (ShangMi 3) secure hash function (OSCCA GM/T 0004-2012) + + Architecture: riscv64 using: + - Zvksh vector crypto extension + - Zvkb vector crypto extension + +config CRYPTO_SM4_RISCV64 + tristate "Ciphers: SM4 (ShangMi 4)" + depends on 64BIT && RISCV_ISA_V && TOOLCHAIN_HAS_VECTOR_CRYPTO + select CRYPTO_ALGAPI + select CRYPTO_SM4 + help + SM4 block cipher algorithm (OSCCA GB/T 32907-2016, + ISO/IEC 18033-3:2010/Amd 1:2021) + + SM4 (GBT.32907-2016) is a cryptographic standard issued by the + Organization of State Commercial Administration of China (OSCCA) + as an authorized cryptographic algorithm for use within China. + + Architecture: riscv64 using: + - Zvksed vector crypto extension + - Zvkb vector crypto extension + +endmenu diff --git a/arch/riscv/crypto/Makefile b/arch/riscv/crypto/Makefile new file mode 100644 index 0000000000..247c7bc728 --- /dev/null +++ b/arch/riscv/crypto/Makefile @@ -0,0 +1,23 @@ +# 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 + +obj-$(CONFIG_CRYPTO_GHASH_RISCV64) += ghash-riscv64.o +ghash-riscv64-y := ghash-riscv64-glue.o ghash-riscv64-zvkg.o + +obj-$(CONFIG_CRYPTO_SHA256_RISCV64) += sha256-riscv64.o +sha256-riscv64-y := sha256-riscv64-glue.o sha256-riscv64-zvknha_or_zvknhb-zvkb.o + +obj-$(CONFIG_CRYPTO_SHA512_RISCV64) += sha512-riscv64.o +sha512-riscv64-y := sha512-riscv64-glue.o sha512-riscv64-zvknhb-zvkb.o + +obj-$(CONFIG_CRYPTO_SM3_RISCV64) += sm3-riscv64.o +sm3-riscv64-y := sm3-riscv64-glue.o sm3-riscv64-zvksh-zvkb.o + +obj-$(CONFIG_CRYPTO_SM4_RISCV64) += sm4-riscv64.o +sm4-riscv64-y := sm4-riscv64-glue.o sm4-riscv64-zvksed-zvkb.o diff --git a/arch/riscv/crypto/aes-macros.S b/arch/riscv/crypto/aes-macros.S new file mode 100644 index 0000000000..d1a258d04b --- /dev/null +++ b/arch/riscv/crypto/aes-macros.S @@ -0,0 +1,156 @@ +/* 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, Christoph Müllner <christoph.muellner@vrull.eu> +// Copyright (c) 2023, Phoebe Chen <phoebe.chen@sifive.com> +// Copyright (c) 2023, Jerry Shih <jerry.shih@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. + +// This file contains macros that are shared by the other aes-*.S files. The +// generated code of these macros depends on the following RISC-V extensions: +// - RV64I +// - RISC-V Vector ('V') with VLEN >= 128 +// - RISC-V Vector AES block cipher extension ('Zvkned') + +// Loads the AES round keys from \keyp into vector registers and jumps to code +// specific to the length of the key. Specifically: +// - If AES-128, loads round keys into v1-v11 and jumps to \label128. +// - If AES-192, loads round keys into v1-v13 and jumps to \label192. +// - If AES-256, loads round keys into v1-v15 and continues onwards. +// +// Also sets vl=4 and vtype=e32,m1,ta,ma. Clobbers t0 and t1. +.macro aes_begin keyp, label128, label192 + lwu t0, 480(\keyp) // t0 = key length in bytes + li t1, 24 // t1 = key length for AES-192 + vsetivli zero, 4, e32, m1, ta, ma + vle32.v v1, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v2, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v3, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v4, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v5, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v6, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v7, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v8, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v9, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v10, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v11, (\keyp) + blt t0, t1, \label128 // If AES-128, goto label128. + addi \keyp, \keyp, 16 + vle32.v v12, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v13, (\keyp) + beq t0, t1, \label192 // If AES-192, goto label192. + // Else, it's AES-256. + addi \keyp, \keyp, 16 + vle32.v v14, (\keyp) + addi \keyp, \keyp, 16 + vle32.v v15, (\keyp) +.endm + +// Encrypts \data using zvkned instructions, using the round keys loaded into +// v1-v11 (for AES-128), v1-v13 (for AES-192), or v1-v15 (for AES-256). \keylen +// is the AES key length in bits. vl and vtype must already be set +// appropriately. Note that if vl > 4, multiple blocks are encrypted. +.macro aes_encrypt data, keylen + vaesz.vs \data, v1 + vaesem.vs \data, v2 + vaesem.vs \data, v3 + vaesem.vs \data, v4 + vaesem.vs \data, v5 + vaesem.vs \data, v6 + vaesem.vs \data, v7 + vaesem.vs \data, v8 + vaesem.vs \data, v9 + vaesem.vs \data, v10 +.if \keylen == 128 + vaesef.vs \data, v11 +.elseif \keylen == 192 + vaesem.vs \data, v11 + vaesem.vs \data, v12 + vaesef.vs \data, v13 +.else + vaesem.vs \data, v11 + vaesem.vs \data, v12 + vaesem.vs \data, v13 + vaesem.vs \data, v14 + vaesef.vs \data, v15 +.endif +.endm + +// Same as aes_encrypt, but decrypts instead of encrypts. +.macro aes_decrypt data, keylen +.if \keylen == 128 + vaesz.vs \data, v11 +.elseif \keylen == 192 + vaesz.vs \data, v13 + vaesdm.vs \data, v12 + vaesdm.vs \data, v11 +.else + vaesz.vs \data, v15 + vaesdm.vs \data, v14 + vaesdm.vs \data, v13 + vaesdm.vs \data, v12 + vaesdm.vs \data, v11 +.endif + vaesdm.vs \data, v10 + vaesdm.vs \data, v9 + vaesdm.vs \data, v8 + vaesdm.vs \data, v7 + vaesdm.vs \data, v6 + vaesdm.vs \data, v5 + vaesdm.vs \data, v4 + vaesdm.vs \data, v3 + vaesdm.vs \data, v2 + vaesdf.vs \data, v1 +.endm + +// Expands to aes_encrypt or aes_decrypt according to \enc, which is 1 or 0. +.macro aes_crypt data, enc, keylen +.if \enc + aes_encrypt \data, \keylen +.else + aes_decrypt \data, \keylen +.endif +.endm diff --git a/arch/riscv/crypto/aes-riscv64-glue.c b/arch/riscv/crypto/aes-riscv64-glue.c new file mode 100644 index 0000000000..f814ee0485 --- /dev/null +++ b/arch/riscv/crypto/aes-riscv64-glue.c @@ -0,0 +1,637 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * AES using the RISC-V vector crypto extensions. Includes the bare block + * cipher and the ECB, CBC, CBC-CTS, CTR, and XTS modes. + * + * Copyright (C) 2023 VRULL GmbH + * Author: Heiko Stuebner <heiko.stuebner@vrull.eu> + * + * Copyright (C) 2023 SiFive, Inc. + * Author: Jerry Shih <jerry.shih@sifive.com> + * + * Copyright 2024 Google LLC + */ + +#include <asm/simd.h> +#include <asm/vector.h> +#include <crypto/aes.h> +#include <crypto/internal/cipher.h> +#include <crypto/internal/simd.h> +#include <crypto/internal/skcipher.h> +#include <crypto/scatterwalk.h> +#include <crypto/xts.h> +#include <linux/linkage.h> +#include <linux/module.h> + +asmlinkage void aes_encrypt_zvkned(const struct crypto_aes_ctx *key, + const u8 in[AES_BLOCK_SIZE], + u8 out[AES_BLOCK_SIZE]); +asmlinkage void aes_decrypt_zvkned(const struct crypto_aes_ctx *key, + const u8 in[AES_BLOCK_SIZE], + u8 out[AES_BLOCK_SIZE]); + +asmlinkage void aes_ecb_encrypt_zvkned(const struct crypto_aes_ctx *key, + const u8 *in, u8 *out, size_t len); +asmlinkage void aes_ecb_decrypt_zvkned(const struct crypto_aes_ctx *key, + const u8 *in, u8 *out, size_t len); + +asmlinkage void aes_cbc_encrypt_zvkned(const struct crypto_aes_ctx *key, + const u8 *in, u8 *out, size_t len, + u8 iv[AES_BLOCK_SIZE]); +asmlinkage void aes_cbc_decrypt_zvkned(const struct crypto_aes_ctx *key, + const u8 *in, u8 *out, size_t len, + u8 iv[AES_BLOCK_SIZE]); + +asmlinkage void aes_cbc_cts_crypt_zvkned(const struct crypto_aes_ctx *key, + const u8 *in, u8 *out, size_t len, + const u8 iv[AES_BLOCK_SIZE], bool enc); + +asmlinkage void aes_ctr32_crypt_zvkned_zvkb(const struct crypto_aes_ctx *key, + const u8 *in, u8 *out, size_t len, + u8 iv[AES_BLOCK_SIZE]); + +asmlinkage void aes_xts_encrypt_zvkned_zvbb_zvkg( + const struct crypto_aes_ctx *key, + const u8 *in, u8 *out, size_t len, + u8 tweak[AES_BLOCK_SIZE]); + +asmlinkage void aes_xts_decrypt_zvkned_zvbb_zvkg( + const struct crypto_aes_ctx *key, + const u8 *in, u8 *out, size_t len, + u8 tweak[AES_BLOCK_SIZE]); + +static int riscv64_aes_setkey(struct crypto_aes_ctx *ctx, + const u8 *key, unsigned int keylen) +{ + /* + * For now we just use the generic key expansion, for these reasons: + * + * - zvkned's key expansion instructions don't support AES-192. + * So, non-zvkned fallback code would be needed anyway. + * + * - Users of AES in Linux usually don't change keys frequently. + * So, key expansion isn't performance-critical. + * + * - For single-block AES exposed as a "cipher" algorithm, it's + * necessary to use struct crypto_aes_ctx and initialize its 'key_dec' + * field with the round keys for the Equivalent Inverse Cipher. This + * is because with "cipher", decryption can be requested from a + * context where the vector unit isn't usable, necessitating a + * fallback to aes_decrypt(). But, zvkned can only generate and use + * the normal round keys. Of course, it's preferable to not have + * special code just for "cipher", as e.g. XTS also uses a + * single-block AES encryption. It's simplest to just use + * struct crypto_aes_ctx and aes_expandkey() everywhere. + */ + return aes_expandkey(ctx, key, keylen); +} + +static int riscv64_aes_setkey_cipher(struct crypto_tfm *tfm, + const u8 *key, unsigned int keylen) +{ + struct crypto_aes_ctx *ctx = crypto_tfm_ctx(tfm); + + return riscv64_aes_setkey(ctx, key, keylen); +} + +static int riscv64_aes_setkey_skcipher(struct crypto_skcipher *tfm, + const u8 *key, unsigned int keylen) +{ + struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm); + + return riscv64_aes_setkey(ctx, key, keylen); +} + +/* Bare AES, without a mode of operation */ + +static void riscv64_aes_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src) +{ + const struct crypto_aes_ctx *ctx = crypto_tfm_ctx(tfm); + + if (crypto_simd_usable()) { + kernel_vector_begin(); + aes_encrypt_zvkned(ctx, src, dst); + kernel_vector_end(); + } else { + aes_encrypt(ctx, dst, src); + } +} + +static void riscv64_aes_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src) +{ + const struct crypto_aes_ctx *ctx = crypto_tfm_ctx(tfm); + + if (crypto_simd_usable()) { + kernel_vector_begin(); + aes_decrypt_zvkned(ctx, src, dst); + kernel_vector_end(); + } else { + aes_decrypt(ctx, dst, src); + } +} + +/* AES-ECB */ + +static inline int riscv64_aes_ecb_crypt(struct skcipher_request *req, bool enc) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); + const struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm); + struct skcipher_walk walk; + unsigned int nbytes; + int err; + + err = skcipher_walk_virt(&walk, req, false); + while ((nbytes = walk.nbytes) != 0) { + kernel_vector_begin(); + if (enc) + aes_ecb_encrypt_zvkned(ctx, walk.src.virt.addr, + walk.dst.virt.addr, + nbytes & ~(AES_BLOCK_SIZE - 1)); + else + aes_ecb_decrypt_zvkned(ctx, walk.src.virt.addr, + walk.dst.virt.addr, + nbytes & ~(AES_BLOCK_SIZE - 1)); + kernel_vector_end(); + err = skcipher_walk_done(&walk, nbytes & (AES_BLOCK_SIZE - 1)); + } + + return err; +} + +static int riscv64_aes_ecb_encrypt(struct skcipher_request *req) +{ + return riscv64_aes_ecb_crypt(req, true); +} + +static int riscv64_aes_ecb_decrypt(struct skcipher_request *req) +{ + return riscv64_aes_ecb_crypt(req, false); +} + +/* AES-CBC */ + +static int riscv64_aes_cbc_crypt(struct skcipher_request *req, bool enc) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); + const struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm); + struct skcipher_walk walk; + unsigned int nbytes; + int err; + + err = skcipher_walk_virt(&walk, req, false); + while ((nbytes = walk.nbytes) != 0) { + kernel_vector_begin(); + if (enc) + aes_cbc_encrypt_zvkned(ctx, walk.src.virt.addr, + walk.dst.virt.addr, + nbytes & ~(AES_BLOCK_SIZE - 1), + walk.iv); + else + aes_cbc_decrypt_zvkned(ctx, walk.src.virt.addr, + walk.dst.virt.addr, + nbytes & ~(AES_BLOCK_SIZE - 1), + walk.iv); + kernel_vector_end(); + err = skcipher_walk_done(&walk, nbytes & (AES_BLOCK_SIZE - 1)); + } + + return err; +} + +static int riscv64_aes_cbc_encrypt(struct skcipher_request *req) +{ + return riscv64_aes_cbc_crypt(req, true); +} + +static int riscv64_aes_cbc_decrypt(struct skcipher_request *req) +{ + return riscv64_aes_cbc_crypt(req, false); +} + +/* AES-CBC-CTS */ + +static int riscv64_aes_cbc_cts_crypt(struct skcipher_request *req, bool enc) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); + const struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm); + struct scatterlist sg_src[2], sg_dst[2]; + struct skcipher_request subreq; + struct scatterlist *src, *dst; + struct skcipher_walk walk; + unsigned int cbc_len; + int err; + + if (req->cryptlen < AES_BLOCK_SIZE) + return -EINVAL; + + err = skcipher_walk_virt(&walk, req, false); + if (err) + return err; + /* + * If the full message is available in one step, decrypt it in one call + * to the CBC-CTS assembly function. This reduces overhead, especially + * on short messages. Otherwise, fall back to doing CBC up to the last + * two blocks, then invoke CTS just for the ciphertext stealing. + */ + if (unlikely(walk.nbytes != req->cryptlen)) { + cbc_len = round_down(req->cryptlen - AES_BLOCK_SIZE - 1, + AES_BLOCK_SIZE); + skcipher_walk_abort(&walk); + skcipher_request_set_tfm(&subreq, tfm); + skcipher_request_set_callback(&subreq, + skcipher_request_flags(req), + NULL, NULL); + skcipher_request_set_crypt(&subreq, req->src, req->dst, + cbc_len, req->iv); + err = riscv64_aes_cbc_crypt(&subreq, enc); + if (err) + return err; + dst = src = scatterwalk_ffwd(sg_src, req->src, cbc_len); + if (req->dst != req->src) + dst = scatterwalk_ffwd(sg_dst, req->dst, cbc_len); + skcipher_request_set_crypt(&subreq, src, dst, + req->cryptlen - cbc_len, req->iv); + err = skcipher_walk_virt(&walk, &subreq, false); + if (err) + return err; + } + kernel_vector_begin(); + aes_cbc_cts_crypt_zvkned(ctx, walk.src.virt.addr, walk.dst.virt.addr, + walk.nbytes, req->iv, enc); + kernel_vector_end(); + return skcipher_walk_done(&walk, 0); +} + +static int riscv64_aes_cbc_cts_encrypt(struct skcipher_request *req) +{ + return riscv64_aes_cbc_cts_crypt(req, true); +} + +static int riscv64_aes_cbc_cts_decrypt(struct skcipher_request *req) +{ + return riscv64_aes_cbc_cts_crypt(req, false); +} + +/* AES-CTR */ + +static int riscv64_aes_ctr_crypt(struct skcipher_request *req) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); + const struct crypto_aes_ctx *ctx = crypto_skcipher_ctx(tfm); + unsigned int nbytes, p1_nbytes; + struct skcipher_walk walk; + u32 ctr32, nblocks; + int err; + + /* Get the low 32-bit word of the 128-bit big endian counter. */ + ctr32 = get_unaligned_be32(req->iv + 12); + + err = skcipher_walk_virt(&walk, req, false); + while ((nbytes = walk.nbytes) != 0) { + if (nbytes < walk.total) { + /* Not the end yet, so keep the length block-aligned. */ + nbytes = round_down(nbytes, AES_BLOCK_SIZE); + nblocks = nbytes / AES_BLOCK_SIZE; + } else { + /* It's the end, so include any final partial block. */ + nblocks = DIV_ROUND_UP(nbytes, AES_BLOCK_SIZE); + } + ctr32 += nblocks; + + kernel_vector_begin(); + if (ctr32 >= nblocks) { + /* The low 32-bit word of the counter won't overflow. */ + aes_ctr32_crypt_zvkned_zvkb(ctx, walk.src.virt.addr, + walk.dst.virt.addr, nbytes, + req->iv); + } else { + /* + * The low 32-bit word of the counter will overflow. + * The assembly doesn't handle this case, so split the + * operation into two at the point where the overflow + * will occur. After the first part, add the carry bit. + */ + p1_nbytes = min_t(unsigned int, nbytes, + (nblocks - ctr32) * AES_BLOCK_SIZE); + aes_ctr32_crypt_zvkned_zvkb(ctx, walk.src.virt.addr, + walk.dst.virt.addr, + p1_nbytes, req->iv); + crypto_inc(req->iv, 12); + + if (ctr32) { + aes_ctr32_crypt_zvkned_zvkb( + ctx, + walk.src.virt.addr + p1_nbytes, + walk.dst.virt.addr + p1_nbytes, + nbytes - p1_nbytes, req->iv); + } + } + kernel_vector_end(); + + err = skcipher_walk_done(&walk, walk.nbytes - nbytes); + } + + return err; +} + +/* AES-XTS */ + +struct riscv64_aes_xts_ctx { + struct crypto_aes_ctx ctx1; + struct crypto_aes_ctx ctx2; +}; + +static int riscv64_aes_xts_setkey(struct crypto_skcipher *tfm, const u8 *key, + unsigned int keylen) +{ + struct riscv64_aes_xts_ctx *ctx = crypto_skcipher_ctx(tfm); + + return xts_verify_key(tfm, key, keylen) ?: + riscv64_aes_setkey(&ctx->ctx1, key, keylen / 2) ?: + riscv64_aes_setkey(&ctx->ctx2, key + keylen / 2, keylen / 2); +} + +static int riscv64_aes_xts_crypt(struct skcipher_request *req, bool enc) +{ + struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); + const struct riscv64_aes_xts_ctx *ctx = crypto_skcipher_ctx(tfm); + int tail = req->cryptlen % AES_BLOCK_SIZE; + struct scatterlist sg_src[2], sg_dst[2]; + struct skcipher_request subreq; + struct scatterlist *src, *dst; + struct skcipher_walk walk; + int err; + + if (req->cryptlen < AES_BLOCK_SIZE) + return -EINVAL; + + /* Encrypt the IV with the tweak key to get the first tweak. */ + kernel_vector_begin(); + aes_encrypt_zvkned(&ctx->ctx2, req->iv, req->iv); + kernel_vector_end(); + + err = skcipher_walk_virt(&walk, req, false); + + /* + * If the message length isn't divisible by the AES block size and the + * full message isn't available in one step of the scatterlist walk, + * then separate off the last full block and the partial block. This + * ensures that they are processed in the same call to the assembly + * function, which is required for ciphertext stealing. + */ + if (unlikely(tail > 0 && walk.nbytes < walk.total)) { + skcipher_walk_abort(&walk); + + skcipher_request_set_tfm(&subreq, tfm); + skcipher_request_set_callback(&subreq, + skcipher_request_flags(req), + NULL, NULL); + skcipher_request_set_crypt(&subreq, req->src, req->dst, + req->cryptlen - tail - AES_BLOCK_SIZE, + req->iv); + req = &subreq; + err = skcipher_walk_virt(&walk, req, false); + } else { + tail = 0; + } + + while (walk.nbytes) { + unsigned int nbytes = walk.nbytes; + + if (nbytes < walk.total) + nbytes = round_down(nbytes, AES_BLOCK_SIZE); + + kernel_vector_begin(); + if (enc) + aes_xts_encrypt_zvkned_zvbb_zvkg( + &ctx->ctx1, walk.src.virt.addr, + walk.dst.virt.addr, nbytes, req->iv); + else + aes_xts_decrypt_zvkned_zvbb_zvkg( + &ctx->ctx1, walk.src.virt.addr, + walk.dst.virt.addr, nbytes, req->iv); + kernel_vector_end(); + err = skcipher_walk_done(&walk, walk.nbytes - nbytes); + } + + if (err || likely(!tail)) + return err; + + /* Do ciphertext stealing with the last full block and partial block. */ + + dst = src = scatterwalk_ffwd(sg_src, req->src, req->cryptlen); + if (req->dst != req->src) + dst = scatterwalk_ffwd(sg_dst, req->dst, req->cryptlen); + + skcipher_request_set_crypt(req, src, dst, AES_BLOCK_SIZE + tail, + req->iv); + + err = skcipher_walk_virt(&walk, req, false); + if (err) + return err; + + kernel_vector_begin(); + if (enc) + aes_xts_encrypt_zvkned_zvbb_zvkg( + &ctx->ctx1, walk.src.virt.addr, + walk.dst.virt.addr, walk.nbytes, req->iv); + else + aes_xts_decrypt_zvkned_zvbb_zvkg( + &ctx->ctx1, walk.src.virt.addr, + walk.dst.virt.addr, walk.nbytes, req->iv); + kernel_vector_end(); + + return skcipher_walk_done(&walk, 0); +} + +static int riscv64_aes_xts_encrypt(struct skcipher_request *req) +{ + return riscv64_aes_xts_crypt(req, true); +} + +static int riscv64_aes_xts_decrypt(struct skcipher_request *req) +{ + return riscv64_aes_xts_crypt(req, false); +} + +/* Algorithm definitions */ + +static struct crypto_alg riscv64_zvkned_aes_cipher_alg = { + .cra_flags = CRYPTO_ALG_TYPE_CIPHER, + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct crypto_aes_ctx), + .cra_priority = 300, + .cra_name = "aes", + .cra_driver_name = "aes-riscv64-zvkned", + .cra_cipher = { + .cia_min_keysize = AES_MIN_KEY_SIZE, + .cia_max_keysize = AES_MAX_KEY_SIZE, + .cia_setkey = riscv64_aes_setkey_cipher, + .cia_encrypt = riscv64_aes_encrypt, + .cia_decrypt = riscv64_aes_decrypt, + }, + .cra_module = THIS_MODULE, +}; + +static struct skcipher_alg riscv64_zvkned_aes_skcipher_algs[] = { + { + .setkey = riscv64_aes_setkey_skcipher, + .encrypt = riscv64_aes_ecb_encrypt, + .decrypt = riscv64_aes_ecb_decrypt, + .min_keysize = AES_MIN_KEY_SIZE, + .max_keysize = AES_MAX_KEY_SIZE, + .walksize = 8 * AES_BLOCK_SIZE, /* matches LMUL=8 */ + .base = { + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct crypto_aes_ctx), + .cra_priority = 300, + .cra_name = "ecb(aes)", + .cra_driver_name = "ecb-aes-riscv64-zvkned", + .cra_module = THIS_MODULE, + }, + }, { + .setkey = riscv64_aes_setkey_skcipher, + .encrypt = riscv64_aes_cbc_encrypt, + .decrypt = riscv64_aes_cbc_decrypt, + .min_keysize = AES_MIN_KEY_SIZE, + .max_keysize = AES_MAX_KEY_SIZE, + .ivsize = AES_BLOCK_SIZE, + .base = { + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct crypto_aes_ctx), + .cra_priority = 300, + .cra_name = "cbc(aes)", + .cra_driver_name = "cbc-aes-riscv64-zvkned", + .cra_module = THIS_MODULE, + }, + }, { + .setkey = riscv64_aes_setkey_skcipher, + .encrypt = riscv64_aes_cbc_cts_encrypt, + .decrypt = riscv64_aes_cbc_cts_decrypt, + .min_keysize = AES_MIN_KEY_SIZE, + .max_keysize = AES_MAX_KEY_SIZE, + .ivsize = AES_BLOCK_SIZE, + .walksize = 4 * AES_BLOCK_SIZE, /* matches LMUL=4 */ + .base = { + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct crypto_aes_ctx), + .cra_priority = 300, + .cra_name = "cts(cbc(aes))", + .cra_driver_name = "cts-cbc-aes-riscv64-zvkned", + .cra_module = THIS_MODULE, + }, + } +}; + +static struct skcipher_alg riscv64_zvkned_zvkb_aes_skcipher_alg = { + .setkey = riscv64_aes_setkey_skcipher, + .encrypt = riscv64_aes_ctr_crypt, + .decrypt = riscv64_aes_ctr_crypt, + .min_keysize = AES_MIN_KEY_SIZE, + .max_keysize = AES_MAX_KEY_SIZE, + .ivsize = AES_BLOCK_SIZE, + .chunksize = AES_BLOCK_SIZE, + .walksize = 4 * AES_BLOCK_SIZE, /* matches LMUL=4 */ + .base = { + .cra_blocksize = 1, + .cra_ctxsize = sizeof(struct crypto_aes_ctx), + .cra_priority = 300, + .cra_name = "ctr(aes)", + .cra_driver_name = "ctr-aes-riscv64-zvkned-zvkb", + .cra_module = THIS_MODULE, + }, +}; + +static struct skcipher_alg riscv64_zvkned_zvbb_zvkg_aes_skcipher_alg = { + .setkey = riscv64_aes_xts_setkey, + .encrypt = riscv64_aes_xts_encrypt, + .decrypt = riscv64_aes_xts_decrypt, + .min_keysize = 2 * AES_MIN_KEY_SIZE, + .max_keysize = 2 * AES_MAX_KEY_SIZE, + .ivsize = AES_BLOCK_SIZE, + .chunksize = AES_BLOCK_SIZE, + .walksize = 4 * AES_BLOCK_SIZE, /* matches LMUL=4 */ + .base = { + .cra_blocksize = AES_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct riscv64_aes_xts_ctx), + .cra_priority = 300, + .cra_name = "xts(aes)", + .cra_driver_name = "xts-aes-riscv64-zvkned-zvbb-zvkg", + .cra_module = THIS_MODULE, + }, +}; + +static inline bool riscv64_aes_xts_supported(void) +{ + return riscv_isa_extension_available(NULL, ZVBB) && + riscv_isa_extension_available(NULL, ZVKG) && + riscv_vector_vlen() < 2048 /* Implementation limitation */; +} + +static int __init riscv64_aes_mod_init(void) +{ + int err = -ENODEV; + + if (riscv_isa_extension_available(NULL, ZVKNED) && + riscv_vector_vlen() >= 128) { + err = crypto_register_alg(&riscv64_zvkned_aes_cipher_alg); + if (err) + return err; + + err = crypto_register_skciphers( + riscv64_zvkned_aes_skcipher_algs, + ARRAY_SIZE(riscv64_zvkned_aes_skcipher_algs)); + if (err) + goto unregister_zvkned_cipher_alg; + + if (riscv_isa_extension_available(NULL, ZVKB)) { + err = crypto_register_skcipher( + &riscv64_zvkned_zvkb_aes_skcipher_alg); + if (err) + goto unregister_zvkned_skcipher_algs; + } + + if (riscv64_aes_xts_supported()) { + err = crypto_register_skcipher( + &riscv64_zvkned_zvbb_zvkg_aes_skcipher_alg); + if (err) + goto unregister_zvkned_zvkb_skcipher_alg; + } + } + + return err; + +unregister_zvkned_zvkb_skcipher_alg: + if (riscv_isa_extension_available(NULL, ZVKB)) + crypto_unregister_skcipher(&riscv64_zvkned_zvkb_aes_skcipher_alg); +unregister_zvkned_skcipher_algs: + crypto_unregister_skciphers(riscv64_zvkned_aes_skcipher_algs, + ARRAY_SIZE(riscv64_zvkned_aes_skcipher_algs)); +unregister_zvkned_cipher_alg: + crypto_unregister_alg(&riscv64_zvkned_aes_cipher_alg); + return err; +} + +static void __exit riscv64_aes_mod_exit(void) +{ + if (riscv64_aes_xts_supported()) + crypto_unregister_skcipher(&riscv64_zvkned_zvbb_zvkg_aes_skcipher_alg); + if (riscv_isa_extension_available(NULL, ZVKB)) + crypto_unregister_skcipher(&riscv64_zvkned_zvkb_aes_skcipher_alg); + crypto_unregister_skciphers(riscv64_zvkned_aes_skcipher_algs, + ARRAY_SIZE(riscv64_zvkned_aes_skcipher_algs)); + crypto_unregister_alg(&riscv64_zvkned_aes_cipher_alg); +} + +module_init(riscv64_aes_mod_init); +module_exit(riscv64_aes_mod_exit); + +MODULE_DESCRIPTION("AES-ECB/CBC/CTS/CTR/XTS (RISC-V accelerated)"); +MODULE_AUTHOR("Jerry Shih <jerry.shih@sifive.com>"); +MODULE_LICENSE("GPL"); +MODULE_ALIAS_CRYPTO("aes"); +MODULE_ALIAS_CRYPTO("ecb(aes)"); +MODULE_ALIAS_CRYPTO("cbc(aes)"); +MODULE_ALIAS_CRYPTO("cts(cbc(aes))"); +MODULE_ALIAS_CRYPTO("ctr(aes)"); +MODULE_ALIAS_CRYPTO("xts(aes)"); diff --git a/arch/riscv/crypto/aes-riscv64-zvkned-zvbb-zvkg.S b/arch/riscv/crypto/aes-riscv64-zvkned-zvbb-zvkg.S new file mode 100644 index 0000000000..146fc9cfb2 --- /dev/null +++ b/arch/riscv/crypto/aes-riscv64-zvkned-zvbb-zvkg.S @@ -0,0 +1,312 @@ +/* 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@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 && VLEN < 2048 +// - RISC-V Vector AES block cipher extension ('Zvkned') +// - RISC-V Vector Bit-manipulation extension ('Zvbb') +// - RISC-V Vector GCM/GMAC extension ('Zvkg') + +#include <linux/linkage.h> + +.text +.option arch, +zvkned, +zvbb, +zvkg + +#include "aes-macros.S" + +#define KEYP a0 +#define INP a1 +#define OUTP a2 +#define LEN a3 +#define TWEAKP a4 + +#define LEN32 a5 +#define TAIL_LEN a6 +#define VL a7 +#define VLMAX t4 + +// v1-v15 contain the AES round keys, but they are used for temporaries before +// the AES round keys have been loaded. +#define TWEAKS v16 // LMUL=4 (most of the time) +#define TWEAKS_BREV v20 // LMUL=4 (most of the time) +#define MULTS_BREV v24 // LMUL=4 (most of the time) +#define TMP0 v28 +#define TMP1 v29 +#define TMP2 v30 +#define TMP3 v31 + +// xts_init initializes the following values: +// +// TWEAKS: N 128-bit tweaks T*(x^i) for i in 0..(N - 1) +// TWEAKS_BREV: same as TWEAKS, but bit-reversed +// MULTS_BREV: N 128-bit values x^N, bit-reversed. Only if N > 1. +// +// N is the maximum number of blocks that will be processed per loop iteration, +// computed using vsetvli. +// +// The field convention used by XTS is the same as that of GHASH, but with the +// bits reversed within each byte. The zvkg extension provides the vgmul +// instruction which does multiplication in this field. Therefore, for tweak +// computation we use vgmul to do multiplications in parallel, instead of +// serially multiplying by x using shifting+xoring. Note that for this to work, +// the inputs and outputs to vgmul must be bit-reversed (we do it with vbrev8). +.macro xts_init + + // Load the first tweak T. + vsetivli zero, 4, e32, m1, ta, ma + vle32.v TWEAKS, (TWEAKP) + + // If there's only one block (or no blocks at all), then skip the tweak + // sequence computation because (at most) T itself is needed. + li t0, 16 + ble LEN, t0, .Linit_single_block\@ + + // Save a copy of T bit-reversed in v12. + vbrev8.v v12, TWEAKS + + // + // Generate x^i for i in 0..(N - 1), i.e. 128-bit values 1 << i assuming + // that N <= 128. Though, this code actually requires N < 64 (or + // equivalently VLEN < 2048) due to the use of 64-bit intermediate + // values here and in the x^N computation later. + // + vsetvli VL, LEN32, e32, m4, ta, ma + srli t0, VL, 2 // t0 = N (num blocks) + // Generate two sequences, each with N 32-bit values: + // v0=[1, 1, 1, ...] and v1=[0, 1, 2, ...]. + vsetvli zero, t0, e32, m1, ta, ma + vmv.v.i v0, 1 + vid.v v1 + // Use vzext to zero-extend the sequences to 64 bits. Reinterpret them + // as two sequences, each with 2*N 32-bit values: + // v2=[1, 0, 1, 0, 1, 0, ...] and v4=[0, 0, 1, 0, 2, 0, ...]. + vsetvli zero, t0, e64, m2, ta, ma + vzext.vf2 v2, v0 + vzext.vf2 v4, v1 + slli t1, t0, 1 // t1 = 2*N + vsetvli zero, t1, e32, m2, ta, ma + // Use vwsll to compute [1<<0, 0<<0, 1<<1, 0<<0, 1<<2, 0<<0, ...], + // widening to 64 bits per element. When reinterpreted as N 128-bit + // values, this is the needed sequence of 128-bit values 1 << i (x^i). + vwsll.vv v8, v2, v4 + + // Copy the bit-reversed T to all N elements of TWEAKS_BREV, then + // multiply by x^i. This gives the sequence T*(x^i), bit-reversed. + vsetvli zero, LEN32, e32, m4, ta, ma + vmv.v.i TWEAKS_BREV, 0 + vaesz.vs TWEAKS_BREV, v12 + vbrev8.v v8, v8 + vgmul.vv TWEAKS_BREV, v8 + + // Save a copy of the sequence T*(x^i) with the bit reversal undone. + vbrev8.v TWEAKS, TWEAKS_BREV + + // Generate N copies of x^N, i.e. 128-bit values 1 << N, bit-reversed. + li t1, 1 + sll t1, t1, t0 // t1 = 1 << N + vsetivli zero, 2, e64, m1, ta, ma + vmv.v.i v0, 0 + vsetivli zero, 1, e64, m1, tu, ma + vmv.v.x v0, t1 + vbrev8.v v0, v0 + vsetvli zero, LEN32, e32, m4, ta, ma + vmv.v.i MULTS_BREV, 0 + vaesz.vs MULTS_BREV, v0 + + j .Linit_done\@ + +.Linit_single_block\@: + vbrev8.v TWEAKS_BREV, TWEAKS +.Linit_done\@: +.endm + +// Set the first 128 bits of MULTS_BREV to 0x40, i.e. 'x' bit-reversed. This is +// the multiplier required to advance the tweak by one. +.macro load_x + li t0, 0x40 + vsetivli zero, 4, e32, m1, ta, ma + vmv.v.i MULTS_BREV, 0 + vsetivli zero, 1, e8, m1, tu, ma + vmv.v.x MULTS_BREV, t0 +.endm + +.macro __aes_xts_crypt enc, keylen + // With 16 < len <= 31, there's no main loop, just ciphertext stealing. + beqz LEN32, .Lcts_without_main_loop\@ + + vsetvli VLMAX, zero, e32, m4, ta, ma +1: + vsetvli VL, LEN32, e32, m4, ta, ma +2: + // Encrypt or decrypt VL/4 blocks. + vle32.v TMP0, (INP) + vxor.vv TMP0, TMP0, TWEAKS + aes_crypt TMP0, \enc, \keylen + vxor.vv TMP0, TMP0, TWEAKS + vse32.v TMP0, (OUTP) + + // Update the pointers and the remaining length. + slli t0, VL, 2 + add INP, INP, t0 + add OUTP, OUTP, t0 + sub LEN32, LEN32, VL + + // Check whether more blocks remain. + beqz LEN32, .Lmain_loop_done\@ + + // Compute the next sequence of tweaks by multiplying the previous + // sequence by x^N. Store the result in both bit-reversed order and + // regular order (i.e. with the bit reversal undone). + vgmul.vv TWEAKS_BREV, MULTS_BREV + vbrev8.v TWEAKS, TWEAKS_BREV + + // Since we compute the tweak multipliers x^N in advance, we require + // that each iteration process the same length except possibly the last. + // This conflicts slightly with the behavior allowed by RISC-V Vector + // Extension, where CPUs can select a lower length for both of the last + // two iterations. E.g., vl might take the sequence of values + // [16, 16, 16, 12, 12], whereas we need [16, 16, 16, 16, 8] so that we + // can use x^4 again instead of computing x^3. Therefore, we explicitly + // keep the vl at VLMAX if there is at least VLMAX remaining. + bge LEN32, VLMAX, 2b + j 1b + +.Lmain_loop_done\@: + load_x + + // Compute the next tweak. + addi t0, VL, -4 + vsetivli zero, 4, e32, m4, ta, ma + vslidedown.vx TWEAKS_BREV, TWEAKS_BREV, t0 // Extract last tweak + vsetivli zero, 4, e32, m1, ta, ma + vgmul.vv TWEAKS_BREV, MULTS_BREV // Advance to next tweak + + bnez TAIL_LEN, .Lcts\@ + + // Update *TWEAKP to contain the next tweak. + vbrev8.v TWEAKS, TWEAKS_BREV + vse32.v TWEAKS, (TWEAKP) + ret + +.Lcts_without_main_loop\@: + load_x +.Lcts\@: + // TWEAKS_BREV now contains the next tweak. Compute the one after that. + vsetivli zero, 4, e32, m1, ta, ma + vmv.v.v TMP0, TWEAKS_BREV + vgmul.vv TMP0, MULTS_BREV + // Undo the bit reversal of the next two tweaks and store them in TMP1 + // and TMP2, such that TMP1 is the first needed and TMP2 the second. +.if \enc + vbrev8.v TMP1, TWEAKS_BREV + vbrev8.v TMP2, TMP0 +.else + vbrev8.v TMP1, TMP0 + vbrev8.v TMP2, TWEAKS_BREV +.endif + + // Encrypt/decrypt the last full block. + vle32.v TMP0, (INP) + vxor.vv TMP0, TMP0, TMP1 + aes_crypt TMP0, \enc, \keylen + vxor.vv TMP0, TMP0, TMP1 + + // Swap the first TAIL_LEN bytes of the above result with the tail. + // Note that to support in-place encryption/decryption, the load from + // the input tail must happen before the store to the output tail. + addi t0, INP, 16 + addi t1, OUTP, 16 + vmv.v.v TMP3, TMP0 + vsetvli zero, TAIL_LEN, e8, m1, tu, ma + vle8.v TMP0, (t0) + vse8.v TMP3, (t1) + + // Encrypt/decrypt again and store the last full block. + vsetivli zero, 4, e32, m1, ta, ma + vxor.vv TMP0, TMP0, TMP2 + aes_crypt TMP0, \enc, \keylen + vxor.vv TMP0, TMP0, TMP2 + vse32.v TMP0, (OUTP) + + ret +.endm + +.macro aes_xts_crypt enc + + // Check whether the length is a multiple of the AES block size. + andi TAIL_LEN, LEN, 15 + beqz TAIL_LEN, 1f + + // The length isn't a multiple of the AES block size, so ciphertext + // stealing will be required. Ciphertext stealing involves special + // handling of the partial block and the last full block, so subtract + // the length of both from the length to be processed in the main loop. + sub LEN, LEN, TAIL_LEN + addi LEN, LEN, -16 +1: + srli LEN32, LEN, 2 + // LEN and LEN32 now contain the total length of the blocks that will be + // processed in the main loop, in bytes and 32-bit words respectively. + + xts_init + aes_begin KEYP, 128f, 192f + __aes_xts_crypt \enc, 256 +128: + __aes_xts_crypt \enc, 128 +192: + __aes_xts_crypt \enc, 192 +.endm + +// void aes_xts_encrypt_zvkned_zvbb_zvkg(const struct crypto_aes_ctx *key, +// const u8 *in, u8 *out, size_t len, +// u8 tweak[16]); +// +// |key| is the data key. |tweak| contains the next tweak; the encryption of +// the original IV with the tweak key was already done. This function supports +// incremental computation, but |len| must always be >= 16 (AES_BLOCK_SIZE), and +// |len| must be a multiple of 16 except on the last call. If |len| is a +// multiple of 16, then this function updates |tweak| to contain the next tweak. +SYM_FUNC_START(aes_xts_encrypt_zvkned_zvbb_zvkg) + aes_xts_crypt 1 +SYM_FUNC_END(aes_xts_encrypt_zvkned_zvbb_zvkg) + +// Same prototype and calling convention as the encryption function +SYM_FUNC_START(aes_xts_decrypt_zvkned_zvbb_zvkg) + aes_xts_crypt 0 +SYM_FUNC_END(aes_xts_decrypt_zvkned_zvbb_zvkg) diff --git a/arch/riscv/crypto/aes-riscv64-zvkned-zvkb.S b/arch/riscv/crypto/aes-riscv64-zvkned-zvkb.S new file mode 100644 index 0000000000..9962d45005 --- /dev/null +++ b/arch/riscv/crypto/aes-riscv64-zvkned-zvkb.S @@ -0,0 +1,146 @@ +/* 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@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 AES block cipher extension ('Zvkned') +// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb') + +#include <linux/linkage.h> + +.text +.option arch, +zvkned, +zvkb + +#include "aes-macros.S" + +#define KEYP a0 +#define INP a1 +#define OUTP a2 +#define LEN a3 +#define IVP a4 + +#define LEN32 a5 +#define VL_E32 a6 +#define VL_BLOCKS a7 + +.macro aes_ctr32_crypt keylen + // LEN32 = number of blocks, rounded up, in 32-bit words. + addi t0, LEN, 15 + srli t0, t0, 4 + slli LEN32, t0, 2 + + // Create a mask that selects the last 32-bit word of each 128-bit + // block. This is the word that contains the (big-endian) counter. + li t0, 0x88 + vsetvli t1, zero, e8, m1, ta, ma + vmv.v.x v0, t0 + + // Load the IV into v31. The last 32-bit word contains the counter. + vsetivli zero, 4, e32, m1, ta, ma + vle32.v v31, (IVP) + + // Convert the big-endian counter into little-endian. + vsetivli zero, 4, e32, m1, ta, mu + vrev8.v v31, v31, v0.t + + // Splat the IV to v16 (with LMUL=4). The number of copies is the + // maximum number of blocks that will be processed per iteration. + vsetvli zero, LEN32, e32, m4, ta, ma + vmv.v.i v16, 0 + vaesz.vs v16, v31 + + // v20 = [x, x, x, 0, x, x, x, 1, ...] + viota.m v20, v0, v0.t + // v16 = [IV0, IV1, IV2, counter+0, IV0, IV1, IV2, counter+1, ...] + vsetvli VL_E32, LEN32, e32, m4, ta, mu + vadd.vv v16, v16, v20, v0.t + + j 2f +1: + // Set the number of blocks to process in this iteration. vl=VL_E32 is + // the length in 32-bit words, i.e. 4 times the number of blocks. + vsetvli VL_E32, LEN32, e32, m4, ta, mu + + // Increment the counters by the number of blocks processed in the + // previous iteration. + vadd.vx v16, v16, VL_BLOCKS, v0.t +2: + // Prepare the AES inputs into v24. + vmv.v.v v24, v16 + vrev8.v v24, v24, v0.t // Convert counters back to big-endian. + + // Encrypt the AES inputs to create the next portion of the keystream. + aes_encrypt v24, \keylen + + // XOR the data with the keystream. + vsetvli t0, LEN, e8, m4, ta, ma + vle8.v v20, (INP) + vxor.vv v20, v20, v24 + vse8.v v20, (OUTP) + + // Advance the pointers and update the remaining length. + add INP, INP, t0 + add OUTP, OUTP, t0 + sub LEN, LEN, t0 + sub LEN32, LEN32, VL_E32 + srli VL_BLOCKS, VL_E32, 2 + + // Repeat if more data remains. + bnez LEN, 1b + + // Update *IVP to contain the next counter. + vsetivli zero, 4, e32, m1, ta, mu + vadd.vx v16, v16, VL_BLOCKS, v0.t + vrev8.v v16, v16, v0.t // Convert counters back to big-endian. + vse32.v v16, (IVP) + + ret +.endm + +// void aes_ctr32_crypt_zvkned_zvkb(const struct crypto_aes_ctx *key, +// const u8 *in, u8 *out, size_t len, +// u8 iv[16]); +SYM_FUNC_START(aes_ctr32_crypt_zvkned_zvkb) + aes_begin KEYP, 128f, 192f + aes_ctr32_crypt 256 +128: + aes_ctr32_crypt 128 +192: + aes_ctr32_crypt 192 +SYM_FUNC_END(aes_ctr32_crypt_zvkned_zvkb) diff --git a/arch/riscv/crypto/aes-riscv64-zvkned.S b/arch/riscv/crypto/aes-riscv64-zvkned.S new file mode 100644 index 0000000000..23d063f94c --- /dev/null +++ b/arch/riscv/crypto/aes-riscv64-zvkned.S @@ -0,0 +1,339 @@ +/* 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, Christoph Müllner <christoph.muellner@vrull.eu> +// Copyright (c) 2023, Phoebe Chen <phoebe.chen@sifive.com> +// Copyright (c) 2023, Jerry Shih <jerry.shih@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 AES block cipher extension ('Zvkned') + +#include <linux/linkage.h> + +.text +.option arch, +zvkned + +#include "aes-macros.S" + +#define KEYP a0 +#define INP a1 +#define OUTP a2 +#define LEN a3 +#define IVP a4 + +.macro __aes_crypt_zvkned enc, keylen + vle32.v v16, (INP) + aes_crypt v16, \enc, \keylen + vse32.v v16, (OUTP) + ret +.endm + +.macro aes_crypt_zvkned enc + aes_begin KEYP, 128f, 192f + __aes_crypt_zvkned \enc, 256 +128: + __aes_crypt_zvkned \enc, 128 +192: + __aes_crypt_zvkned \enc, 192 +.endm + +// void aes_encrypt_zvkned(const struct crypto_aes_ctx *key, +// const u8 in[16], u8 out[16]); +SYM_FUNC_START(aes_encrypt_zvkned) + aes_crypt_zvkned 1 +SYM_FUNC_END(aes_encrypt_zvkned) + +// Same prototype and calling convention as the encryption function +SYM_FUNC_START(aes_decrypt_zvkned) + aes_crypt_zvkned 0 +SYM_FUNC_END(aes_decrypt_zvkned) + +.macro __aes_ecb_crypt enc, keylen + srli t0, LEN, 2 + // t0 is the remaining length in 32-bit words. It's a multiple of 4. +1: + vsetvli t1, t0, e32, m8, ta, ma + sub t0, t0, t1 // Subtract number of words processed + slli t1, t1, 2 // Words to bytes + vle32.v v16, (INP) + aes_crypt v16, \enc, \keylen + vse32.v v16, (OUTP) + add INP, INP, t1 + add OUTP, OUTP, t1 + bnez t0, 1b + + ret +.endm + +.macro aes_ecb_crypt enc + aes_begin KEYP, 128f, 192f + __aes_ecb_crypt \enc, 256 +128: + __aes_ecb_crypt \enc, 128 +192: + __aes_ecb_crypt \enc, 192 +.endm + +// void aes_ecb_encrypt_zvkned(const struct crypto_aes_ctx *key, +// const u8 *in, u8 *out, size_t len); +// +// |len| must be nonzero and a multiple of 16 (AES_BLOCK_SIZE). +SYM_FUNC_START(aes_ecb_encrypt_zvkned) + aes_ecb_crypt 1 +SYM_FUNC_END(aes_ecb_encrypt_zvkned) + +// Same prototype and calling convention as the encryption function +SYM_FUNC_START(aes_ecb_decrypt_zvkned) + aes_ecb_crypt 0 +SYM_FUNC_END(aes_ecb_decrypt_zvkned) + +.macro aes_cbc_encrypt keylen + vle32.v v16, (IVP) // Load IV +1: + vle32.v v17, (INP) // Load plaintext block + vxor.vv v16, v16, v17 // XOR with IV or prev ciphertext block + aes_encrypt v16, \keylen // Encrypt + vse32.v v16, (OUTP) // Store ciphertext block + addi INP, INP, 16 + addi OUTP, OUTP, 16 + addi LEN, LEN, -16 + bnez LEN, 1b + + vse32.v v16, (IVP) // Store next IV + ret +.endm + +.macro aes_cbc_decrypt keylen + srli LEN, LEN, 2 // Convert LEN from bytes to words + vle32.v v16, (IVP) // Load IV +1: + vsetvli t0, LEN, e32, m4, ta, ma + vle32.v v20, (INP) // Load ciphertext blocks + vslideup.vi v16, v20, 4 // Setup prev ciphertext blocks + addi t1, t0, -4 + vslidedown.vx v24, v20, t1 // Save last ciphertext block + aes_decrypt v20, \keylen // Decrypt the blocks + vxor.vv v20, v20, v16 // XOR with prev ciphertext blocks + vse32.v v20, (OUTP) // Store plaintext blocks + vmv.v.v v16, v24 // Next "IV" is last ciphertext block + slli t1, t0, 2 // Words to bytes + add INP, INP, t1 + add OUTP, OUTP, t1 + sub LEN, LEN, t0 + bnez LEN, 1b + + vsetivli zero, 4, e32, m1, ta, ma + vse32.v v16, (IVP) // Store next IV + ret +.endm + +// void aes_cbc_encrypt_zvkned(const struct crypto_aes_ctx *key, +// const u8 *in, u8 *out, size_t len, u8 iv[16]); +// +// |len| must be nonzero and a multiple of 16 (AES_BLOCK_SIZE). +SYM_FUNC_START(aes_cbc_encrypt_zvkned) + aes_begin KEYP, 128f, 192f + aes_cbc_encrypt 256 +128: + aes_cbc_encrypt 128 +192: + aes_cbc_encrypt 192 +SYM_FUNC_END(aes_cbc_encrypt_zvkned) + +// Same prototype and calling convention as the encryption function +SYM_FUNC_START(aes_cbc_decrypt_zvkned) + aes_begin KEYP, 128f, 192f + aes_cbc_decrypt 256 +128: + aes_cbc_decrypt 128 +192: + aes_cbc_decrypt 192 +SYM_FUNC_END(aes_cbc_decrypt_zvkned) + +.macro aes_cbc_cts_encrypt keylen + + // CBC-encrypt all blocks except the last. But don't store the + // second-to-last block to the output buffer yet, since it will be + // handled specially in the ciphertext stealing step. Exception: if the + // message is single-block, still encrypt the last (and only) block. + li t0, 16 + j 2f +1: + vse32.v v16, (OUTP) // Store ciphertext block + addi OUTP, OUTP, 16 +2: + vle32.v v17, (INP) // Load plaintext block + vxor.vv v16, v16, v17 // XOR with IV or prev ciphertext block + aes_encrypt v16, \keylen // Encrypt + addi INP, INP, 16 + addi LEN, LEN, -16 + bgt LEN, t0, 1b // Repeat if more than one block remains + + // Special case: if the message is a single block, just do CBC. + beqz LEN, .Lcts_encrypt_done\@ + + // Encrypt the last two blocks using ciphertext stealing as follows: + // C[n-1] = Encrypt(Encrypt(P[n-1] ^ C[n-2]) ^ P[n]) + // C[n] = Encrypt(P[n-1] ^ C[n-2])[0..LEN] + // + // C[i] denotes the i'th ciphertext block, and likewise P[i] the i'th + // plaintext block. Block n, the last block, may be partial; its length + // is 1 <= LEN <= 16. If there are only 2 blocks, C[n-2] means the IV. + // + // v16 already contains Encrypt(P[n-1] ^ C[n-2]). + // INP points to P[n]. OUTP points to where C[n-1] should go. + // To support in-place encryption, load P[n] before storing C[n]. + addi t0, OUTP, 16 // Get pointer to where C[n] should go + vsetvli zero, LEN, e8, m1, tu, ma + vle8.v v17, (INP) // Load P[n] + vse8.v v16, (t0) // Store C[n] + vxor.vv v16, v16, v17 // v16 = Encrypt(P[n-1] ^ C[n-2]) ^ P[n] + vsetivli zero, 4, e32, m1, ta, ma + aes_encrypt v16, \keylen +.Lcts_encrypt_done\@: + vse32.v v16, (OUTP) // Store C[n-1] (or C[n] in single-block case) + ret +.endm + +#define LEN32 t4 // Length of remaining full blocks in 32-bit words +#define LEN_MOD16 t5 // Length of message in bytes mod 16 + +.macro aes_cbc_cts_decrypt keylen + andi LEN32, LEN, ~15 + srli LEN32, LEN32, 2 + andi LEN_MOD16, LEN, 15 + + // Save C[n-2] in v28 so that it's available later during the ciphertext + // stealing step. If there are fewer than three blocks, C[n-2] means + // the IV, otherwise it means the third-to-last ciphertext block. + vmv.v.v v28, v16 // IV + add t0, LEN, -33 + bltz t0, .Lcts_decrypt_loop\@ + andi t0, t0, ~15 + add t0, t0, INP + vle32.v v28, (t0) + + // CBC-decrypt all full blocks. For the last full block, or the last 2 + // full blocks if the message is block-aligned, this doesn't write the + // correct output blocks (unless the message is only a single block), + // because it XORs the wrong values with the raw AES plaintexts. But we + // fix this after this loop without redoing the AES decryptions. This + // approach allows more of the AES decryptions to be parallelized. +.Lcts_decrypt_loop\@: + vsetvli t0, LEN32, e32, m4, ta, ma + addi t1, t0, -4 + vle32.v v20, (INP) // Load next set of ciphertext blocks + vmv.v.v v24, v16 // Get IV or last ciphertext block of prev set + vslideup.vi v24, v20, 4 // Setup prev ciphertext blocks + vslidedown.vx v16, v20, t1 // Save last ciphertext block of this set + aes_decrypt v20, \keylen // Decrypt this set of blocks + vxor.vv v24, v24, v20 // XOR prev ciphertext blocks with decrypted blocks + vse32.v v24, (OUTP) // Store this set of plaintext blocks + sub LEN32, LEN32, t0 + slli t0, t0, 2 // Words to bytes + add INP, INP, t0 + add OUTP, OUTP, t0 + bnez LEN32, .Lcts_decrypt_loop\@ + + vsetivli zero, 4, e32, m4, ta, ma + vslidedown.vx v20, v20, t1 // Extract raw plaintext of last full block + addi t0, OUTP, -16 // Get pointer to last full plaintext block + bnez LEN_MOD16, .Lcts_decrypt_non_block_aligned\@ + + // Special case: if the message is a single block, just do CBC. + li t1, 16 + beq LEN, t1, .Lcts_decrypt_done\@ + + // Block-aligned message. Just fix up the last 2 blocks. We need: + // + // P[n-1] = Decrypt(C[n]) ^ C[n-2] + // P[n] = Decrypt(C[n-1]) ^ C[n] + // + // We have C[n] in v16, Decrypt(C[n]) in v20, and C[n-2] in v28. + // Together with Decrypt(C[n-1]) ^ C[n-2] from the output buffer, this + // is everything needed to fix the output without re-decrypting blocks. + addi t1, OUTP, -32 // Get pointer to where P[n-1] should go + vxor.vv v20, v20, v28 // Decrypt(C[n]) ^ C[n-2] == P[n-1] + vle32.v v24, (t1) // Decrypt(C[n-1]) ^ C[n-2] + vse32.v v20, (t1) // Store P[n-1] + vxor.vv v20, v24, v16 // Decrypt(C[n-1]) ^ C[n-2] ^ C[n] == P[n] ^ C[n-2] + j .Lcts_decrypt_finish\@ + +.Lcts_decrypt_non_block_aligned\@: + // Decrypt the last two blocks using ciphertext stealing as follows: + // + // P[n-1] = Decrypt(C[n] || Decrypt(C[n-1])[LEN_MOD16..16]) ^ C[n-2] + // P[n] = (Decrypt(C[n-1]) ^ C[n])[0..LEN_MOD16] + // + // We already have Decrypt(C[n-1]) in v20 and C[n-2] in v28. + vmv.v.v v16, v20 // v16 = Decrypt(C[n-1]) + vsetvli zero, LEN_MOD16, e8, m1, tu, ma + vle8.v v20, (INP) // v20 = C[n] || Decrypt(C[n-1])[LEN_MOD16..16] + vxor.vv v16, v16, v20 // v16 = Decrypt(C[n-1]) ^ C[n] + vse8.v v16, (OUTP) // Store P[n] + vsetivli zero, 4, e32, m1, ta, ma + aes_decrypt v20, \keylen // v20 = Decrypt(C[n] || Decrypt(C[n-1])[LEN_MOD16..16]) +.Lcts_decrypt_finish\@: + vxor.vv v20, v20, v28 // XOR with C[n-2] + vse32.v v20, (t0) // Store last full plaintext block +.Lcts_decrypt_done\@: + ret +.endm + +.macro aes_cbc_cts_crypt keylen + vle32.v v16, (IVP) // Load IV + beqz a5, .Lcts_decrypt\@ + aes_cbc_cts_encrypt \keylen +.Lcts_decrypt\@: + aes_cbc_cts_decrypt \keylen +.endm + +// void aes_cbc_cts_crypt_zvkned(const struct crypto_aes_ctx *key, +// const u8 *in, u8 *out, size_t len, +// const u8 iv[16], bool enc); +// +// Encrypts or decrypts a message with the CS3 variant of AES-CBC-CTS. +// This is the variant that unconditionally swaps the last two blocks. +SYM_FUNC_START(aes_cbc_cts_crypt_zvkned) + aes_begin KEYP, 128f, 192f + aes_cbc_cts_crypt 256 +128: + aes_cbc_cts_crypt 128 +192: + aes_cbc_cts_crypt 192 +SYM_FUNC_END(aes_cbc_cts_crypt_zvkned) diff --git a/arch/riscv/crypto/chacha-riscv64-glue.c b/arch/riscv/crypto/chacha-riscv64-glue.c new file mode 100644 index 0000000000..10b46f3637 --- /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@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@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 0000000000..bf057737ac --- /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@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) diff --git a/arch/riscv/crypto/ghash-riscv64-glue.c b/arch/riscv/crypto/ghash-riscv64-glue.c new file mode 100644 index 0000000000..312e7891fd --- /dev/null +++ b/arch/riscv/crypto/ghash-riscv64-glue.c @@ -0,0 +1,168 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * GHASH using the RISC-V vector crypto extensions + * + * Copyright (C) 2023 VRULL GmbH + * Author: Heiko Stuebner <heiko.stuebner@vrull.eu> + * + * Copyright (C) 2023 SiFive, Inc. + * Author: Jerry Shih <jerry.shih@sifive.com> + */ + +#include <asm/simd.h> +#include <asm/vector.h> +#include <crypto/ghash.h> +#include <crypto/internal/hash.h> +#include <crypto/internal/simd.h> +#include <linux/linkage.h> +#include <linux/module.h> + +asmlinkage void ghash_zvkg(be128 *accumulator, const be128 *key, const u8 *data, + size_t len); + +struct riscv64_ghash_tfm_ctx { + be128 key; +}; + +struct riscv64_ghash_desc_ctx { + be128 accumulator; + u8 buffer[GHASH_BLOCK_SIZE]; + u32 bytes; +}; + +static int riscv64_ghash_setkey(struct crypto_shash *tfm, const u8 *key, + unsigned int keylen) +{ + struct riscv64_ghash_tfm_ctx *tctx = crypto_shash_ctx(tfm); + + if (keylen != GHASH_BLOCK_SIZE) + return -EINVAL; + + memcpy(&tctx->key, key, GHASH_BLOCK_SIZE); + + return 0; +} + +static int riscv64_ghash_init(struct shash_desc *desc) +{ + struct riscv64_ghash_desc_ctx *dctx = shash_desc_ctx(desc); + + *dctx = (struct riscv64_ghash_desc_ctx){}; + + return 0; +} + +static inline void +riscv64_ghash_blocks(const struct riscv64_ghash_tfm_ctx *tctx, + struct riscv64_ghash_desc_ctx *dctx, + const u8 *src, size_t srclen) +{ + /* The srclen is nonzero and a multiple of 16. */ + if (crypto_simd_usable()) { + kernel_vector_begin(); + ghash_zvkg(&dctx->accumulator, &tctx->key, src, srclen); + kernel_vector_end(); + } else { + do { + crypto_xor((u8 *)&dctx->accumulator, src, + GHASH_BLOCK_SIZE); + gf128mul_lle(&dctx->accumulator, &tctx->key); + src += GHASH_BLOCK_SIZE; + srclen -= GHASH_BLOCK_SIZE; + } while (srclen); + } +} + +static int riscv64_ghash_update(struct shash_desc *desc, const u8 *src, + unsigned int srclen) +{ + const struct riscv64_ghash_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm); + struct riscv64_ghash_desc_ctx *dctx = shash_desc_ctx(desc); + unsigned int len; + + if (dctx->bytes) { + if (dctx->bytes + srclen < GHASH_BLOCK_SIZE) { + memcpy(dctx->buffer + dctx->bytes, src, srclen); + dctx->bytes += srclen; + return 0; + } + memcpy(dctx->buffer + dctx->bytes, src, + GHASH_BLOCK_SIZE - dctx->bytes); + riscv64_ghash_blocks(tctx, dctx, dctx->buffer, + GHASH_BLOCK_SIZE); + src += GHASH_BLOCK_SIZE - dctx->bytes; + srclen -= GHASH_BLOCK_SIZE - dctx->bytes; + dctx->bytes = 0; + } + + len = round_down(srclen, GHASH_BLOCK_SIZE); + if (len) { + riscv64_ghash_blocks(tctx, dctx, src, len); + src += len; + srclen -= len; + } + + if (srclen) { + memcpy(dctx->buffer, src, srclen); + dctx->bytes = srclen; + } + + return 0; +} + +static int riscv64_ghash_final(struct shash_desc *desc, u8 *out) +{ + const struct riscv64_ghash_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm); + struct riscv64_ghash_desc_ctx *dctx = shash_desc_ctx(desc); + int i; + + if (dctx->bytes) { + for (i = dctx->bytes; i < GHASH_BLOCK_SIZE; i++) + dctx->buffer[i] = 0; + + riscv64_ghash_blocks(tctx, dctx, dctx->buffer, + GHASH_BLOCK_SIZE); + } + + memcpy(out, &dctx->accumulator, GHASH_DIGEST_SIZE); + return 0; +} + +static struct shash_alg riscv64_ghash_alg = { + .init = riscv64_ghash_init, + .update = riscv64_ghash_update, + .final = riscv64_ghash_final, + .setkey = riscv64_ghash_setkey, + .descsize = sizeof(struct riscv64_ghash_desc_ctx), + .digestsize = GHASH_DIGEST_SIZE, + .base = { + .cra_blocksize = GHASH_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct riscv64_ghash_tfm_ctx), + .cra_priority = 300, + .cra_name = "ghash", + .cra_driver_name = "ghash-riscv64-zvkg", + .cra_module = THIS_MODULE, + }, +}; + +static int __init riscv64_ghash_mod_init(void) +{ + if (riscv_isa_extension_available(NULL, ZVKG) && + riscv_vector_vlen() >= 128) + return crypto_register_shash(&riscv64_ghash_alg); + + return -ENODEV; +} + +static void __exit riscv64_ghash_mod_exit(void) +{ + crypto_unregister_shash(&riscv64_ghash_alg); +} + +module_init(riscv64_ghash_mod_init); +module_exit(riscv64_ghash_mod_exit); + +MODULE_DESCRIPTION("GHASH (RISC-V accelerated)"); +MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>"); +MODULE_LICENSE("GPL"); +MODULE_ALIAS_CRYPTO("ghash"); diff --git a/arch/riscv/crypto/ghash-riscv64-zvkg.S b/arch/riscv/crypto/ghash-riscv64-zvkg.S new file mode 100644 index 0000000000..f2b43fb4d4 --- /dev/null +++ b/arch/riscv/crypto/ghash-riscv64-zvkg.S @@ -0,0 +1,72 @@ +/* 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, Christoph Müllner <christoph.muellner@vrull.eu> +// Copyright (c) 2023, Jerry Shih <jerry.shih@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 GCM/GMAC extension ('Zvkg') + +#include <linux/linkage.h> + +.text +.option arch, +zvkg + +#define ACCUMULATOR a0 +#define KEY a1 +#define DATA a2 +#define LEN a3 + +// void ghash_zvkg(be128 *accumulator, const be128 *key, const u8 *data, +// size_t len); +// +// |len| must be nonzero and a multiple of 16 (GHASH_BLOCK_SIZE). +SYM_FUNC_START(ghash_zvkg) + vsetivli zero, 4, e32, m1, ta, ma + vle32.v v1, (ACCUMULATOR) + vle32.v v2, (KEY) +.Lnext_block: + vle32.v v3, (DATA) + vghsh.vv v1, v2, v3 + addi DATA, DATA, 16 + addi LEN, LEN, -16 + bnez LEN, .Lnext_block + + vse32.v v1, (ACCUMULATOR) + ret +SYM_FUNC_END(ghash_zvkg) diff --git a/arch/riscv/crypto/sha256-riscv64-glue.c b/arch/riscv/crypto/sha256-riscv64-glue.c new file mode 100644 index 0000000000..71e051e40a --- /dev/null +++ b/arch/riscv/crypto/sha256-riscv64-glue.c @@ -0,0 +1,137 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * SHA-256 and SHA-224 using the RISC-V vector crypto extensions + * + * Copyright (C) 2022 VRULL GmbH + * Author: Heiko Stuebner <heiko.stuebner@vrull.eu> + * + * Copyright (C) 2023 SiFive, Inc. + * Author: Jerry Shih <jerry.shih@sifive.com> + */ + +#include <asm/simd.h> +#include <asm/vector.h> +#include <crypto/internal/hash.h> +#include <crypto/internal/simd.h> +#include <crypto/sha256_base.h> +#include <linux/linkage.h> +#include <linux/module.h> + +/* + * Note: the asm function only uses the 'state' field of struct sha256_state. + * It is assumed to be the first field. + */ +asmlinkage void sha256_transform_zvknha_or_zvknhb_zvkb( + struct sha256_state *state, const u8 *data, int num_blocks); + +static int riscv64_sha256_update(struct shash_desc *desc, const u8 *data, + unsigned int len) +{ + /* + * Ensure struct sha256_state begins directly with the SHA-256 + * 256-bit internal state, as this is what the asm function expects. + */ + BUILD_BUG_ON(offsetof(struct sha256_state, state) != 0); + + if (crypto_simd_usable()) { + kernel_vector_begin(); + sha256_base_do_update(desc, data, len, + sha256_transform_zvknha_or_zvknhb_zvkb); + kernel_vector_end(); + } else { + crypto_sha256_update(desc, data, len); + } + return 0; +} + +static int riscv64_sha256_finup(struct shash_desc *desc, const u8 *data, + unsigned int len, u8 *out) +{ + if (crypto_simd_usable()) { + kernel_vector_begin(); + if (len) + sha256_base_do_update( + desc, data, len, + sha256_transform_zvknha_or_zvknhb_zvkb); + sha256_base_do_finalize( + desc, sha256_transform_zvknha_or_zvknhb_zvkb); + kernel_vector_end(); + + return sha256_base_finish(desc, out); + } + + return crypto_sha256_finup(desc, data, len, out); +} + +static int riscv64_sha256_final(struct shash_desc *desc, u8 *out) +{ + return riscv64_sha256_finup(desc, NULL, 0, out); +} + +static int riscv64_sha256_digest(struct shash_desc *desc, const u8 *data, + unsigned int len, u8 *out) +{ + return sha256_base_init(desc) ?: + riscv64_sha256_finup(desc, data, len, out); +} + +static struct shash_alg riscv64_sha256_algs[] = { + { + .init = sha256_base_init, + .update = riscv64_sha256_update, + .final = riscv64_sha256_final, + .finup = riscv64_sha256_finup, + .digest = riscv64_sha256_digest, + .descsize = sizeof(struct sha256_state), + .digestsize = SHA256_DIGEST_SIZE, + .base = { + .cra_blocksize = SHA256_BLOCK_SIZE, + .cra_priority = 300, + .cra_name = "sha256", + .cra_driver_name = "sha256-riscv64-zvknha_or_zvknhb-zvkb", + .cra_module = THIS_MODULE, + }, + }, { + .init = sha224_base_init, + .update = riscv64_sha256_update, + .final = riscv64_sha256_final, + .finup = riscv64_sha256_finup, + .descsize = sizeof(struct sha256_state), + .digestsize = SHA224_DIGEST_SIZE, + .base = { + .cra_blocksize = SHA224_BLOCK_SIZE, + .cra_priority = 300, + .cra_name = "sha224", + .cra_driver_name = "sha224-riscv64-zvknha_or_zvknhb-zvkb", + .cra_module = THIS_MODULE, + }, + }, +}; + +static int __init riscv64_sha256_mod_init(void) +{ + /* Both zvknha and zvknhb provide the SHA-256 instructions. */ + if ((riscv_isa_extension_available(NULL, ZVKNHA) || + riscv_isa_extension_available(NULL, ZVKNHB)) && + riscv_isa_extension_available(NULL, ZVKB) && + riscv_vector_vlen() >= 128) + return crypto_register_shashes(riscv64_sha256_algs, + ARRAY_SIZE(riscv64_sha256_algs)); + + return -ENODEV; +} + +static void __exit riscv64_sha256_mod_exit(void) +{ + crypto_unregister_shashes(riscv64_sha256_algs, + ARRAY_SIZE(riscv64_sha256_algs)); +} + +module_init(riscv64_sha256_mod_init); +module_exit(riscv64_sha256_mod_exit); + +MODULE_DESCRIPTION("SHA-256 (RISC-V accelerated)"); +MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>"); +MODULE_LICENSE("GPL"); +MODULE_ALIAS_CRYPTO("sha256"); +MODULE_ALIAS_CRYPTO("sha224"); diff --git a/arch/riscv/crypto/sha256-riscv64-zvknha_or_zvknhb-zvkb.S b/arch/riscv/crypto/sha256-riscv64-zvknha_or_zvknhb-zvkb.S new file mode 100644 index 0000000000..8ebcc17de4 --- /dev/null +++ b/arch/riscv/crypto/sha256-riscv64-zvknha_or_zvknhb-zvkb.S @@ -0,0 +1,225 @@ +/* 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, Christoph Müllner <christoph.muellner@vrull.eu> +// Copyright (c) 2023, Phoebe Chen <phoebe.chen@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 SHA-2 Secure Hash extension ('Zvknha' or 'Zvknhb') +// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb') + +#include <linux/cfi_types.h> + +.text +.option arch, +zvknha, +zvkb + +#define STATEP a0 +#define DATA a1 +#define NUM_BLOCKS a2 + +#define STATEP_C a3 + +#define MASK v0 +#define INDICES v1 +#define W0 v2 +#define W1 v3 +#define W2 v4 +#define W3 v5 +#define VTMP v6 +#define FEBA v7 +#define HGDC v8 +#define K0 v10 +#define K1 v11 +#define K2 v12 +#define K3 v13 +#define K4 v14 +#define K5 v15 +#define K6 v16 +#define K7 v17 +#define K8 v18 +#define K9 v19 +#define K10 v20 +#define K11 v21 +#define K12 v22 +#define K13 v23 +#define K14 v24 +#define K15 v25 +#define PREV_FEBA v26 +#define PREV_HGDC v27 + +// Do 4 rounds of SHA-256. w0 contains the current 4 message schedule words. +// +// If not all the message schedule words have been computed yet, then this also +// computes 4 more message schedule words. w1-w3 contain the next 3 groups of 4 +// message schedule words; this macro computes the group after w3 and writes it +// to w0. This means that the next (w0, w1, w2, w3) is the current (w1, w2, w3, +// w0), so the caller must cycle through the registers accordingly. +.macro sha256_4rounds last, k, w0, w1, w2, w3 + vadd.vv VTMP, \k, \w0 + vsha2cl.vv HGDC, FEBA, VTMP + vsha2ch.vv FEBA, HGDC, VTMP +.if !\last + vmerge.vvm VTMP, \w2, \w1, MASK + vsha2ms.vv \w0, VTMP, \w3 +.endif +.endm + +.macro sha256_16rounds last, k0, k1, k2, k3 + sha256_4rounds \last, \k0, W0, W1, W2, W3 + sha256_4rounds \last, \k1, W1, W2, W3, W0 + sha256_4rounds \last, \k2, W2, W3, W0, W1 + sha256_4rounds \last, \k3, W3, W0, W1, W2 +.endm + +// void sha256_transform_zvknha_or_zvknhb_zvkb(u32 state[8], const u8 *data, +// int num_blocks); +SYM_TYPED_FUNC_START(sha256_transform_zvknha_or_zvknhb_zvkb) + + // Load the round constants into K0-K15. + vsetivli zero, 4, e32, m1, ta, ma + la t0, K256 + vle32.v K0, (t0) + addi t0, t0, 16 + vle32.v K1, (t0) + addi t0, t0, 16 + vle32.v K2, (t0) + addi t0, t0, 16 + vle32.v K3, (t0) + addi t0, t0, 16 + vle32.v K4, (t0) + addi t0, t0, 16 + vle32.v K5, (t0) + addi t0, t0, 16 + vle32.v K6, (t0) + addi t0, t0, 16 + vle32.v K7, (t0) + addi t0, t0, 16 + vle32.v K8, (t0) + addi t0, t0, 16 + vle32.v K9, (t0) + addi t0, t0, 16 + vle32.v K10, (t0) + addi t0, t0, 16 + vle32.v K11, (t0) + addi t0, t0, 16 + vle32.v K12, (t0) + addi t0, t0, 16 + vle32.v K13, (t0) + addi t0, t0, 16 + vle32.v K14, (t0) + addi t0, t0, 16 + vle32.v K15, (t0) + + // Setup mask for the vmerge to replace the first word (idx==0) in + // message scheduling. There are 4 words, so an 8-bit mask suffices. + vsetivli zero, 1, e8, m1, ta, ma + vmv.v.i MASK, 0x01 + + // Load the state. The state is stored as {a,b,c,d,e,f,g,h}, but we + // need {f,e,b,a},{h,g,d,c}. The dst vtype is e32m1 and the index vtype + // is e8mf4. We use index-load with the i8 indices {20, 16, 4, 0}, + // loaded using the 32-bit little endian value 0x00041014. + li t0, 0x00041014 + vsetivli zero, 1, e32, m1, ta, ma + vmv.v.x INDICES, t0 + addi STATEP_C, STATEP, 8 + vsetivli zero, 4, e32, m1, ta, ma + vluxei8.v FEBA, (STATEP), INDICES + vluxei8.v HGDC, (STATEP_C), INDICES + +.Lnext_block: + addi NUM_BLOCKS, NUM_BLOCKS, -1 + + // Save the previous state, as it's needed later. + vmv.v.v PREV_FEBA, FEBA + vmv.v.v PREV_HGDC, HGDC + + // Load the next 512-bit message block and endian-swap each 32-bit word. + vle32.v W0, (DATA) + vrev8.v W0, W0 + addi DATA, DATA, 16 + vle32.v W1, (DATA) + vrev8.v W1, W1 + addi DATA, DATA, 16 + vle32.v W2, (DATA) + vrev8.v W2, W2 + addi DATA, DATA, 16 + vle32.v W3, (DATA) + vrev8.v W3, W3 + addi DATA, DATA, 16 + + // Do the 64 rounds of SHA-256. + sha256_16rounds 0, K0, K1, K2, K3 + sha256_16rounds 0, K4, K5, K6, K7 + sha256_16rounds 0, K8, K9, K10, K11 + sha256_16rounds 1, K12, K13, K14, K15 + + // Add the previous state. + vadd.vv FEBA, FEBA, PREV_FEBA + vadd.vv HGDC, HGDC, PREV_HGDC + + // Repeat if more blocks remain. + bnez NUM_BLOCKS, .Lnext_block + + // Store the new state and return. + vsuxei8.v FEBA, (STATEP), INDICES + vsuxei8.v HGDC, (STATEP_C), INDICES + ret +SYM_FUNC_END(sha256_transform_zvknha_or_zvknhb_zvkb) + +.section ".rodata" +.p2align 2 +.type K256, @object +K256: + .word 0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5 + .word 0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5 + .word 0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3 + .word 0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174 + .word 0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc + .word 0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da + .word 0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7 + .word 0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967 + .word 0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13 + .word 0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85 + .word 0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3 + .word 0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070 + .word 0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5 + .word 0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3 + .word 0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208 + .word 0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2 +.size K256, . - K256 diff --git a/arch/riscv/crypto/sha512-riscv64-glue.c b/arch/riscv/crypto/sha512-riscv64-glue.c new file mode 100644 index 0000000000..43b56a08ae --- /dev/null +++ b/arch/riscv/crypto/sha512-riscv64-glue.c @@ -0,0 +1,133 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * SHA-512 and SHA-384 using the RISC-V vector crypto extensions + * + * Copyright (C) 2023 VRULL GmbH + * Author: Heiko Stuebner <heiko.stuebner@vrull.eu> + * + * Copyright (C) 2023 SiFive, Inc. + * Author: Jerry Shih <jerry.shih@sifive.com> + */ + +#include <asm/simd.h> +#include <asm/vector.h> +#include <crypto/internal/hash.h> +#include <crypto/internal/simd.h> +#include <crypto/sha512_base.h> +#include <linux/linkage.h> +#include <linux/module.h> + +/* + * Note: the asm function only uses the 'state' field of struct sha512_state. + * It is assumed to be the first field. + */ +asmlinkage void sha512_transform_zvknhb_zvkb( + struct sha512_state *state, const u8 *data, int num_blocks); + +static int riscv64_sha512_update(struct shash_desc *desc, const u8 *data, + unsigned int len) +{ + /* + * Ensure struct sha512_state begins directly with the SHA-512 + * 512-bit internal state, as this is what the asm function expects. + */ + BUILD_BUG_ON(offsetof(struct sha512_state, state) != 0); + + if (crypto_simd_usable()) { + kernel_vector_begin(); + sha512_base_do_update(desc, data, len, + sha512_transform_zvknhb_zvkb); + kernel_vector_end(); + } else { + crypto_sha512_update(desc, data, len); + } + return 0; +} + +static int riscv64_sha512_finup(struct shash_desc *desc, const u8 *data, + unsigned int len, u8 *out) +{ + if (crypto_simd_usable()) { + kernel_vector_begin(); + if (len) + sha512_base_do_update(desc, data, len, + sha512_transform_zvknhb_zvkb); + sha512_base_do_finalize(desc, sha512_transform_zvknhb_zvkb); + kernel_vector_end(); + + return sha512_base_finish(desc, out); + } + + return crypto_sha512_finup(desc, data, len, out); +} + +static int riscv64_sha512_final(struct shash_desc *desc, u8 *out) +{ + return riscv64_sha512_finup(desc, NULL, 0, out); +} + +static int riscv64_sha512_digest(struct shash_desc *desc, const u8 *data, + unsigned int len, u8 *out) +{ + return sha512_base_init(desc) ?: + riscv64_sha512_finup(desc, data, len, out); +} + +static struct shash_alg riscv64_sha512_algs[] = { + { + .init = sha512_base_init, + .update = riscv64_sha512_update, + .final = riscv64_sha512_final, + .finup = riscv64_sha512_finup, + .digest = riscv64_sha512_digest, + .descsize = sizeof(struct sha512_state), + .digestsize = SHA512_DIGEST_SIZE, + .base = { + .cra_blocksize = SHA512_BLOCK_SIZE, + .cra_priority = 300, + .cra_name = "sha512", + .cra_driver_name = "sha512-riscv64-zvknhb-zvkb", + .cra_module = THIS_MODULE, + }, + }, { + .init = sha384_base_init, + .update = riscv64_sha512_update, + .final = riscv64_sha512_final, + .finup = riscv64_sha512_finup, + .descsize = sizeof(struct sha512_state), + .digestsize = SHA384_DIGEST_SIZE, + .base = { + .cra_blocksize = SHA384_BLOCK_SIZE, + .cra_priority = 300, + .cra_name = "sha384", + .cra_driver_name = "sha384-riscv64-zvknhb-zvkb", + .cra_module = THIS_MODULE, + }, + }, +}; + +static int __init riscv64_sha512_mod_init(void) +{ + if (riscv_isa_extension_available(NULL, ZVKNHB) && + riscv_isa_extension_available(NULL, ZVKB) && + riscv_vector_vlen() >= 128) + return crypto_register_shashes(riscv64_sha512_algs, + ARRAY_SIZE(riscv64_sha512_algs)); + + return -ENODEV; +} + +static void __exit riscv64_sha512_mod_exit(void) +{ + crypto_unregister_shashes(riscv64_sha512_algs, + ARRAY_SIZE(riscv64_sha512_algs)); +} + +module_init(riscv64_sha512_mod_init); +module_exit(riscv64_sha512_mod_exit); + +MODULE_DESCRIPTION("SHA-512 (RISC-V accelerated)"); +MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>"); +MODULE_LICENSE("GPL"); +MODULE_ALIAS_CRYPTO("sha512"); +MODULE_ALIAS_CRYPTO("sha384"); diff --git a/arch/riscv/crypto/sha512-riscv64-zvknhb-zvkb.S b/arch/riscv/crypto/sha512-riscv64-zvknhb-zvkb.S new file mode 100644 index 0000000000..3a9ae210f9 --- /dev/null +++ b/arch/riscv/crypto/sha512-riscv64-zvknhb-zvkb.S @@ -0,0 +1,203 @@ +/* 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, Christoph Müllner <christoph.muellner@vrull.eu> +// Copyright (c) 2023, Phoebe Chen <phoebe.chen@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 SHA-2 Secure Hash extension ('Zvknhb') +// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb') + +#include <linux/cfi_types.h> + +.text +.option arch, +zvknhb, +zvkb + +#define STATEP a0 +#define DATA a1 +#define NUM_BLOCKS a2 + +#define STATEP_C a3 +#define K a4 + +#define MASK v0 +#define INDICES v1 +#define W0 v10 // LMUL=2 +#define W1 v12 // LMUL=2 +#define W2 v14 // LMUL=2 +#define W3 v16 // LMUL=2 +#define VTMP v20 // LMUL=2 +#define FEBA v22 // LMUL=2 +#define HGDC v24 // LMUL=2 +#define PREV_FEBA v26 // LMUL=2 +#define PREV_HGDC v28 // LMUL=2 + +// Do 4 rounds of SHA-512. w0 contains the current 4 message schedule words. +// +// If not all the message schedule words have been computed yet, then this also +// computes 4 more message schedule words. w1-w3 contain the next 3 groups of 4 +// message schedule words; this macro computes the group after w3 and writes it +// to w0. This means that the next (w0, w1, w2, w3) is the current (w1, w2, w3, +// w0), so the caller must cycle through the registers accordingly. +.macro sha512_4rounds last, w0, w1, w2, w3 + vle64.v VTMP, (K) + addi K, K, 32 + vadd.vv VTMP, VTMP, \w0 + vsha2cl.vv HGDC, FEBA, VTMP + vsha2ch.vv FEBA, HGDC, VTMP +.if !\last + vmerge.vvm VTMP, \w2, \w1, MASK + vsha2ms.vv \w0, VTMP, \w3 +.endif +.endm + +.macro sha512_16rounds last + sha512_4rounds \last, W0, W1, W2, W3 + sha512_4rounds \last, W1, W2, W3, W0 + sha512_4rounds \last, W2, W3, W0, W1 + sha512_4rounds \last, W3, W0, W1, W2 +.endm + +// void sha512_transform_zvknhb_zvkb(u64 state[8], const u8 *data, +// int num_blocks); +SYM_TYPED_FUNC_START(sha512_transform_zvknhb_zvkb) + + // Setup mask for the vmerge to replace the first word (idx==0) in + // message scheduling. There are 4 words, so an 8-bit mask suffices. + vsetivli zero, 1, e8, m1, ta, ma + vmv.v.i MASK, 0x01 + + // Load the state. The state is stored as {a,b,c,d,e,f,g,h}, but we + // need {f,e,b,a},{h,g,d,c}. The dst vtype is e64m2 and the index vtype + // is e8mf4. We use index-load with the i8 indices {40, 32, 8, 0}, + // loaded using the 32-bit little endian value 0x00082028. + li t0, 0x00082028 + vsetivli zero, 1, e32, m1, ta, ma + vmv.v.x INDICES, t0 + addi STATEP_C, STATEP, 16 + vsetivli zero, 4, e64, m2, ta, ma + vluxei8.v FEBA, (STATEP), INDICES + vluxei8.v HGDC, (STATEP_C), INDICES + +.Lnext_block: + la K, K512 + addi NUM_BLOCKS, NUM_BLOCKS, -1 + + // Save the previous state, as it's needed later. + vmv.v.v PREV_FEBA, FEBA + vmv.v.v PREV_HGDC, HGDC + + // Load the next 1024-bit message block and endian-swap each 64-bit word + vle64.v W0, (DATA) + vrev8.v W0, W0 + addi DATA, DATA, 32 + vle64.v W1, (DATA) + vrev8.v W1, W1 + addi DATA, DATA, 32 + vle64.v W2, (DATA) + vrev8.v W2, W2 + addi DATA, DATA, 32 + vle64.v W3, (DATA) + vrev8.v W3, W3 + addi DATA, DATA, 32 + + // Do the 80 rounds of SHA-512. + sha512_16rounds 0 + sha512_16rounds 0 + sha512_16rounds 0 + sha512_16rounds 0 + sha512_16rounds 1 + + // Add the previous state. + vadd.vv FEBA, FEBA, PREV_FEBA + vadd.vv HGDC, HGDC, PREV_HGDC + + // Repeat if more blocks remain. + bnez NUM_BLOCKS, .Lnext_block + + // Store the new state and return. + vsuxei8.v FEBA, (STATEP), INDICES + vsuxei8.v HGDC, (STATEP_C), INDICES + ret +SYM_FUNC_END(sha512_transform_zvknhb_zvkb) + +.section ".rodata" +.p2align 3 +.type K512, @object +K512: + .dword 0x428a2f98d728ae22, 0x7137449123ef65cd + .dword 0xb5c0fbcfec4d3b2f, 0xe9b5dba58189dbbc + .dword 0x3956c25bf348b538, 0x59f111f1b605d019 + .dword 0x923f82a4af194f9b, 0xab1c5ed5da6d8118 + .dword 0xd807aa98a3030242, 0x12835b0145706fbe + .dword 0x243185be4ee4b28c, 0x550c7dc3d5ffb4e2 + .dword 0x72be5d74f27b896f, 0x80deb1fe3b1696b1 + .dword 0x9bdc06a725c71235, 0xc19bf174cf692694 + .dword 0xe49b69c19ef14ad2, 0xefbe4786384f25e3 + .dword 0x0fc19dc68b8cd5b5, 0x240ca1cc77ac9c65 + .dword 0x2de92c6f592b0275, 0x4a7484aa6ea6e483 + .dword 0x5cb0a9dcbd41fbd4, 0x76f988da831153b5 + .dword 0x983e5152ee66dfab, 0xa831c66d2db43210 + .dword 0xb00327c898fb213f, 0xbf597fc7beef0ee4 + .dword 0xc6e00bf33da88fc2, 0xd5a79147930aa725 + .dword 0x06ca6351e003826f, 0x142929670a0e6e70 + .dword 0x27b70a8546d22ffc, 0x2e1b21385c26c926 + .dword 0x4d2c6dfc5ac42aed, 0x53380d139d95b3df + .dword 0x650a73548baf63de, 0x766a0abb3c77b2a8 + .dword 0x81c2c92e47edaee6, 0x92722c851482353b + .dword 0xa2bfe8a14cf10364, 0xa81a664bbc423001 + .dword 0xc24b8b70d0f89791, 0xc76c51a30654be30 + .dword 0xd192e819d6ef5218, 0xd69906245565a910 + .dword 0xf40e35855771202a, 0x106aa07032bbd1b8 + .dword 0x19a4c116b8d2d0c8, 0x1e376c085141ab53 + .dword 0x2748774cdf8eeb99, 0x34b0bcb5e19b48a8 + .dword 0x391c0cb3c5c95a63, 0x4ed8aa4ae3418acb + .dword 0x5b9cca4f7763e373, 0x682e6ff3d6b2b8a3 + .dword 0x748f82ee5defb2fc, 0x78a5636f43172f60 + .dword 0x84c87814a1f0ab72, 0x8cc702081a6439ec + .dword 0x90befffa23631e28, 0xa4506cebde82bde9 + .dword 0xbef9a3f7b2c67915, 0xc67178f2e372532b + .dword 0xca273eceea26619c, 0xd186b8c721c0c207 + .dword 0xeada7dd6cde0eb1e, 0xf57d4f7fee6ed178 + .dword 0x06f067aa72176fba, 0x0a637dc5a2c898a6 + .dword 0x113f9804bef90dae, 0x1b710b35131c471b + .dword 0x28db77f523047d84, 0x32caab7b40c72493 + .dword 0x3c9ebe0a15c9bebc, 0x431d67c49c100d4c + .dword 0x4cc5d4becb3e42b6, 0x597f299cfc657e2a + .dword 0x5fcb6fab3ad6faec, 0x6c44198c4a475817 +.size K512, . - K512 diff --git a/arch/riscv/crypto/sm3-riscv64-glue.c b/arch/riscv/crypto/sm3-riscv64-glue.c new file mode 100644 index 0000000000..e1737a970c --- /dev/null +++ b/arch/riscv/crypto/sm3-riscv64-glue.c @@ -0,0 +1,112 @@ +// SPDX-License-Identifier: GPL-2.0-or-later +/* + * SM3 using the RISC-V vector crypto extensions + * + * Copyright (C) 2023 VRULL GmbH + * Author: Heiko Stuebner <heiko.stuebner@vrull.eu> + * + * Copyright (C) 2023 SiFive, Inc. + * Author: Jerry Shih <jerry.shih@sifive.com> + */ + +#include <asm/simd.h> +#include <asm/vector.h> +#include <crypto/internal/hash.h> +#include <crypto/internal/simd.h> +#include <crypto/sm3_base.h> +#include <linux/linkage.h> +#include <linux/module.h> + +/* + * Note: the asm function only uses the 'state' field of struct sm3_state. + * It is assumed to be the first field. + */ +asmlinkage void sm3_transform_zvksh_zvkb( + struct sm3_state *state, const u8 *data, int num_blocks); + +static int riscv64_sm3_update(struct shash_desc *desc, const u8 *data, + unsigned int len) +{ + /* + * Ensure struct sm3_state begins directly with the SM3 + * 256-bit internal state, as this is what the asm function expects. + */ + BUILD_BUG_ON(offsetof(struct sm3_state, state) != 0); + + if (crypto_simd_usable()) { + kernel_vector_begin(); + sm3_base_do_update(desc, data, len, sm3_transform_zvksh_zvkb); + kernel_vector_end(); + } else { + sm3_update(shash_desc_ctx(desc), data, len); + } + return 0; +} + +static int riscv64_sm3_finup(struct shash_desc *desc, const u8 *data, + unsigned int len, u8 *out) +{ + struct sm3_state *ctx; + + if (crypto_simd_usable()) { + kernel_vector_begin(); + if (len) + sm3_base_do_update(desc, data, len, + sm3_transform_zvksh_zvkb); + sm3_base_do_finalize(desc, sm3_transform_zvksh_zvkb); + kernel_vector_end(); + + return sm3_base_finish(desc, out); + } + + ctx = shash_desc_ctx(desc); + if (len) + sm3_update(ctx, data, len); + sm3_final(ctx, out); + + return 0; +} + +static int riscv64_sm3_final(struct shash_desc *desc, u8 *out) +{ + return riscv64_sm3_finup(desc, NULL, 0, out); +} + +static struct shash_alg riscv64_sm3_alg = { + .init = sm3_base_init, + .update = riscv64_sm3_update, + .final = riscv64_sm3_final, + .finup = riscv64_sm3_finup, + .descsize = sizeof(struct sm3_state), + .digestsize = SM3_DIGEST_SIZE, + .base = { + .cra_blocksize = SM3_BLOCK_SIZE, + .cra_priority = 300, + .cra_name = "sm3", + .cra_driver_name = "sm3-riscv64-zvksh-zvkb", + .cra_module = THIS_MODULE, + }, +}; + +static int __init riscv64_sm3_mod_init(void) +{ + if (riscv_isa_extension_available(NULL, ZVKSH) && + riscv_isa_extension_available(NULL, ZVKB) && + riscv_vector_vlen() >= 128) + return crypto_register_shash(&riscv64_sm3_alg); + + return -ENODEV; +} + +static void __exit riscv64_sm3_mod_exit(void) +{ + crypto_unregister_shash(&riscv64_sm3_alg); +} + +module_init(riscv64_sm3_mod_init); +module_exit(riscv64_sm3_mod_exit); + +MODULE_DESCRIPTION("SM3 (RISC-V accelerated)"); +MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>"); +MODULE_LICENSE("GPL"); +MODULE_ALIAS_CRYPTO("sm3"); diff --git a/arch/riscv/crypto/sm3-riscv64-zvksh-zvkb.S b/arch/riscv/crypto/sm3-riscv64-zvksh-zvkb.S new file mode 100644 index 0000000000..a2b65d961c --- /dev/null +++ b/arch/riscv/crypto/sm3-riscv64-zvksh-zvkb.S @@ -0,0 +1,123 @@ +/* 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, Christoph Müllner <christoph.muellner@vrull.eu> +// Copyright (c) 2023, Jerry Shih <jerry.shih@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 SM3 Secure Hash extension ('Zvksh') +// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb') + +#include <linux/cfi_types.h> + +.text +.option arch, +zvksh, +zvkb + +#define STATEP a0 +#define DATA a1 +#define NUM_BLOCKS a2 + +#define STATE v0 // LMUL=2 +#define PREV_STATE v2 // LMUL=2 +#define W0 v4 // LMUL=2 +#define W1 v6 // LMUL=2 +#define VTMP v8 // LMUL=2 + +.macro sm3_8rounds i, w0, w1 + // Do 4 rounds using W_{0+i}..W_{7+i}. + vsm3c.vi STATE, \w0, \i + 0 + vslidedown.vi VTMP, \w0, 2 + vsm3c.vi STATE, VTMP, \i + 1 + + // Compute W_{4+i}..W_{11+i}. + vslidedown.vi VTMP, \w0, 4 + vslideup.vi VTMP, \w1, 4 + + // Do 4 rounds using W_{4+i}..W_{11+i}. + vsm3c.vi STATE, VTMP, \i + 2 + vslidedown.vi VTMP, VTMP, 2 + vsm3c.vi STATE, VTMP, \i + 3 + +.if \i < 28 + // Compute W_{16+i}..W_{23+i}. + vsm3me.vv \w0, \w1, \w0 +.endif + // For the next 8 rounds, w0 and w1 are swapped. +.endm + +// void sm3_transform_zvksh_zvkb(u32 state[8], const u8 *data, int num_blocks); +SYM_TYPED_FUNC_START(sm3_transform_zvksh_zvkb) + + // Load the state and endian-swap each 32-bit word. + vsetivli zero, 8, e32, m2, ta, ma + vle32.v STATE, (STATEP) + vrev8.v STATE, STATE + +.Lnext_block: + addi NUM_BLOCKS, NUM_BLOCKS, -1 + + // Save the previous state, as it's needed later. + vmv.v.v PREV_STATE, STATE + + // Load the next 512-bit message block into W0-W1. + vle32.v W0, (DATA) + addi DATA, DATA, 32 + vle32.v W1, (DATA) + addi DATA, DATA, 32 + + // Do the 64 rounds of SM3. + sm3_8rounds 0, W0, W1 + sm3_8rounds 4, W1, W0 + sm3_8rounds 8, W0, W1 + sm3_8rounds 12, W1, W0 + sm3_8rounds 16, W0, W1 + sm3_8rounds 20, W1, W0 + sm3_8rounds 24, W0, W1 + sm3_8rounds 28, W1, W0 + + // XOR in the previous state. + vxor.vv STATE, STATE, PREV_STATE + + // Repeat if more blocks remain. + bnez NUM_BLOCKS, .Lnext_block + + // Store the new state and return. + vrev8.v STATE, STATE + vse32.v STATE, (STATEP) + ret +SYM_FUNC_END(sm3_transform_zvksh_zvkb) diff --git a/arch/riscv/crypto/sm4-riscv64-glue.c b/arch/riscv/crypto/sm4-riscv64-glue.c new file mode 100644 index 0000000000..47fb84ebe5 --- /dev/null +++ b/arch/riscv/crypto/sm4-riscv64-glue.c @@ -0,0 +1,107 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * SM4 using the RISC-V vector crypto extensions + * + * Copyright (C) 2023 VRULL GmbH + * Author: Heiko Stuebner <heiko.stuebner@vrull.eu> + * + * Copyright (C) 2023 SiFive, Inc. + * Author: Jerry Shih <jerry.shih@sifive.com> + */ + +#include <asm/simd.h> +#include <asm/vector.h> +#include <crypto/internal/cipher.h> +#include <crypto/internal/simd.h> +#include <crypto/sm4.h> +#include <linux/linkage.h> +#include <linux/module.h> + +asmlinkage void sm4_expandkey_zvksed_zvkb(const u8 user_key[SM4_KEY_SIZE], + u32 rkey_enc[SM4_RKEY_WORDS], + u32 rkey_dec[SM4_RKEY_WORDS]); +asmlinkage void sm4_crypt_zvksed_zvkb(const u32 rkey[SM4_RKEY_WORDS], + const u8 in[SM4_BLOCK_SIZE], + u8 out[SM4_BLOCK_SIZE]); + +static int riscv64_sm4_setkey(struct crypto_tfm *tfm, const u8 *key, + unsigned int keylen) +{ + struct sm4_ctx *ctx = crypto_tfm_ctx(tfm); + + if (crypto_simd_usable()) { + if (keylen != SM4_KEY_SIZE) + return -EINVAL; + kernel_vector_begin(); + sm4_expandkey_zvksed_zvkb(key, ctx->rkey_enc, ctx->rkey_dec); + kernel_vector_end(); + return 0; + } + return sm4_expandkey(ctx, key, keylen); +} + +static void riscv64_sm4_encrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src) +{ + const struct sm4_ctx *ctx = crypto_tfm_ctx(tfm); + + if (crypto_simd_usable()) { + kernel_vector_begin(); + sm4_crypt_zvksed_zvkb(ctx->rkey_enc, src, dst); + kernel_vector_end(); + } else { + sm4_crypt_block(ctx->rkey_enc, dst, src); + } +} + +static void riscv64_sm4_decrypt(struct crypto_tfm *tfm, u8 *dst, const u8 *src) +{ + const struct sm4_ctx *ctx = crypto_tfm_ctx(tfm); + + if (crypto_simd_usable()) { + kernel_vector_begin(); + sm4_crypt_zvksed_zvkb(ctx->rkey_dec, src, dst); + kernel_vector_end(); + } else { + sm4_crypt_block(ctx->rkey_dec, dst, src); + } +} + +static struct crypto_alg riscv64_sm4_alg = { + .cra_flags = CRYPTO_ALG_TYPE_CIPHER, + .cra_blocksize = SM4_BLOCK_SIZE, + .cra_ctxsize = sizeof(struct sm4_ctx), + .cra_priority = 300, + .cra_name = "sm4", + .cra_driver_name = "sm4-riscv64-zvksed-zvkb", + .cra_cipher = { + .cia_min_keysize = SM4_KEY_SIZE, + .cia_max_keysize = SM4_KEY_SIZE, + .cia_setkey = riscv64_sm4_setkey, + .cia_encrypt = riscv64_sm4_encrypt, + .cia_decrypt = riscv64_sm4_decrypt, + }, + .cra_module = THIS_MODULE, +}; + +static int __init riscv64_sm4_mod_init(void) +{ + if (riscv_isa_extension_available(NULL, ZVKSED) && + riscv_isa_extension_available(NULL, ZVKB) && + riscv_vector_vlen() >= 128) + return crypto_register_alg(&riscv64_sm4_alg); + + return -ENODEV; +} + +static void __exit riscv64_sm4_mod_exit(void) +{ + crypto_unregister_alg(&riscv64_sm4_alg); +} + +module_init(riscv64_sm4_mod_init); +module_exit(riscv64_sm4_mod_exit); + +MODULE_DESCRIPTION("SM4 (RISC-V accelerated)"); +MODULE_AUTHOR("Heiko Stuebner <heiko.stuebner@vrull.eu>"); +MODULE_LICENSE("GPL"); +MODULE_ALIAS_CRYPTO("sm4"); diff --git a/arch/riscv/crypto/sm4-riscv64-zvksed-zvkb.S b/arch/riscv/crypto/sm4-riscv64-zvksed-zvkb.S new file mode 100644 index 0000000000..fae62179a4 --- /dev/null +++ b/arch/riscv/crypto/sm4-riscv64-zvksed-zvkb.S @@ -0,0 +1,117 @@ +/* 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, Christoph Müllner <christoph.muellner@vrull.eu> +// Copyright (c) 2023, Jerry Shih <jerry.shih@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 SM4 Block Cipher extension ('Zvksed') +// - RISC-V Vector Cryptography Bit-manipulation extension ('Zvkb') + +#include <linux/linkage.h> + +.text +.option arch, +zvksed, +zvkb + +// void sm4_expandkey_zksed_zvkb(const u8 user_key[16], u32 rkey_enc[32], +// u32 rkey_dec[32]); +SYM_FUNC_START(sm4_expandkey_zvksed_zvkb) + vsetivli zero, 4, e32, m1, ta, ma + + // Load the user key. + vle32.v v1, (a0) + vrev8.v v1, v1 + + // XOR the user key with the family key. + la t0, FAMILY_KEY + vle32.v v2, (t0) + vxor.vv v1, v1, v2 + + // Compute the round keys. Store them in forwards order in rkey_enc + // and in reverse order in rkey_dec. + addi a2, a2, 31*4 + li t0, -4 + .set i, 0 +.rept 8 + vsm4k.vi v1, v1, i + vse32.v v1, (a1) // Store to rkey_enc. + vsse32.v v1, (a2), t0 // Store to rkey_dec. +.if i < 7 + addi a1, a1, 16 + addi a2, a2, -16 +.endif + .set i, i + 1 +.endr + + ret +SYM_FUNC_END(sm4_expandkey_zvksed_zvkb) + +// void sm4_crypt_zvksed_zvkb(const u32 rkey[32], const u8 in[16], u8 out[16]); +SYM_FUNC_START(sm4_crypt_zvksed_zvkb) + vsetivli zero, 4, e32, m1, ta, ma + + // Load the input data. + vle32.v v1, (a1) + vrev8.v v1, v1 + + // Do the 32 rounds of SM4, 4 at a time. + .set i, 0 +.rept 8 + vle32.v v2, (a0) + vsm4r.vs v1, v2 +.if i < 7 + addi a0, a0, 16 +.endif + .set i, i + 1 +.endr + + // Store the output data (in reverse element order). + vrev8.v v1, v1 + li t0, -4 + addi a2, a2, 12 + vsse32.v v1, (a2), t0 + + ret +SYM_FUNC_END(sm4_crypt_zvksed_zvkb) + +.section ".rodata" +.p2align 2 +.type FAMILY_KEY, @object +FAMILY_KEY: + .word 0xA3B1BAC6, 0x56AA3350, 0x677D9197, 0xB27022DC +.size FAMILY_KEY, . - FAMILY_KEY |