| xj | b04a402 | 2021-11-25 15:01:52 +0800 | [diff] [blame] | 1 | /* |
| 2 | * ARM NEON accelerated ChaCha and XChaCha stream ciphers, |
| 3 | * including ChaCha20 (RFC7539) |
| 4 | * |
| 5 | * Copyright (C) 2016 Linaro, Ltd. <ard.biesheuvel@linaro.org> |
| 6 | * |
| 7 | * This program is free software; you can redistribute it and/or modify |
| 8 | * it under the terms of the GNU General Public License version 2 as |
| 9 | * published by the Free Software Foundation. |
| 10 | * |
| 11 | * Based on: |
| 12 | * ChaCha20 256-bit cipher algorithm, RFC7539, SIMD glue code |
| 13 | * |
| 14 | * Copyright (C) 2015 Martin Willi |
| 15 | * |
| 16 | * This program is free software; you can redistribute it and/or modify |
| 17 | * it under the terms of the GNU General Public License as published by |
| 18 | * the Free Software Foundation; either version 2 of the License, or |
| 19 | * (at your option) any later version. |
| 20 | */ |
| 21 | |
| 22 | #include <crypto/algapi.h> |
| 23 | #include <crypto/chacha.h> |
| 24 | #include <crypto/internal/skcipher.h> |
| 25 | #include <linux/kernel.h> |
| 26 | #include <linux/module.h> |
| 27 | |
| 28 | #include <asm/hwcap.h> |
| 29 | #include <asm/neon.h> |
| 30 | #include <asm/simd.h> |
| 31 | |
| 32 | asmlinkage void chacha_block_xor_neon(const u32 *state, u8 *dst, const u8 *src, |
| 33 | int nrounds); |
| 34 | asmlinkage void chacha_4block_xor_neon(const u32 *state, u8 *dst, const u8 *src, |
| 35 | int nrounds); |
| 36 | asmlinkage void hchacha_block_neon(const u32 *state, u32 *out, int nrounds); |
| 37 | |
| 38 | static void chacha_doneon(u32 *state, u8 *dst, const u8 *src, |
| 39 | unsigned int bytes, int nrounds) |
| 40 | { |
| 41 | u8 buf[CHACHA_BLOCK_SIZE]; |
| 42 | |
| 43 | while (bytes >= CHACHA_BLOCK_SIZE * 4) { |
| 44 | chacha_4block_xor_neon(state, dst, src, nrounds); |
| 45 | bytes -= CHACHA_BLOCK_SIZE * 4; |
| 46 | src += CHACHA_BLOCK_SIZE * 4; |
| 47 | dst += CHACHA_BLOCK_SIZE * 4; |
| 48 | state[12] += 4; |
| 49 | } |
| 50 | while (bytes >= CHACHA_BLOCK_SIZE) { |
| 51 | chacha_block_xor_neon(state, dst, src, nrounds); |
| 52 | bytes -= CHACHA_BLOCK_SIZE; |
| 53 | src += CHACHA_BLOCK_SIZE; |
| 54 | dst += CHACHA_BLOCK_SIZE; |
| 55 | state[12]++; |
| 56 | } |
| 57 | if (bytes) { |
| 58 | memcpy(buf, src, bytes); |
| 59 | chacha_block_xor_neon(state, buf, buf, nrounds); |
| 60 | memcpy(dst, buf, bytes); |
| 61 | } |
| 62 | } |
| 63 | |
| 64 | static int chacha_neon_stream_xor(struct skcipher_request *req, |
| 65 | struct chacha_ctx *ctx, u8 *iv) |
| 66 | { |
| 67 | struct skcipher_walk walk; |
| 68 | u32 state[16]; |
| 69 | int err; |
| 70 | |
| 71 | err = skcipher_walk_virt(&walk, req, false); |
| 72 | |
| 73 | crypto_chacha_init(state, ctx, iv); |
| 74 | |
| 75 | while (walk.nbytes > 0) { |
| 76 | unsigned int nbytes = walk.nbytes; |
| 77 | |
| 78 | if (nbytes < walk.total) |
| 79 | nbytes = round_down(nbytes, walk.stride); |
| 80 | |
| 81 | kernel_neon_begin(); |
| 82 | chacha_doneon(state, walk.dst.virt.addr, walk.src.virt.addr, |
| 83 | nbytes, ctx->nrounds); |
| 84 | kernel_neon_end(); |
| 85 | err = skcipher_walk_done(&walk, walk.nbytes - nbytes); |
| 86 | } |
| 87 | |
| 88 | return err; |
| 89 | } |
| 90 | |
| 91 | static int chacha_neon(struct skcipher_request *req) |
| 92 | { |
| 93 | struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); |
| 94 | struct chacha_ctx *ctx = crypto_skcipher_ctx(tfm); |
| 95 | |
| 96 | if (req->cryptlen <= CHACHA_BLOCK_SIZE || !may_use_simd()) |
| 97 | return crypto_chacha_crypt(req); |
| 98 | |
| 99 | return chacha_neon_stream_xor(req, ctx, req->iv); |
| 100 | } |
| 101 | |
| 102 | static int xchacha_neon(struct skcipher_request *req) |
| 103 | { |
| 104 | struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req); |
| 105 | struct chacha_ctx *ctx = crypto_skcipher_ctx(tfm); |
| 106 | struct chacha_ctx subctx; |
| 107 | u32 state[16]; |
| 108 | u8 real_iv[16]; |
| 109 | |
| 110 | if (req->cryptlen <= CHACHA_BLOCK_SIZE || !may_use_simd()) |
| 111 | return crypto_xchacha_crypt(req); |
| 112 | |
| 113 | crypto_chacha_init(state, ctx, req->iv); |
| 114 | |
| 115 | kernel_neon_begin(); |
| 116 | hchacha_block_neon(state, subctx.key, ctx->nrounds); |
| 117 | kernel_neon_end(); |
| 118 | subctx.nrounds = ctx->nrounds; |
| 119 | |
| 120 | memcpy(&real_iv[0], req->iv + 24, 8); |
| 121 | memcpy(&real_iv[8], req->iv + 16, 8); |
| 122 | return chacha_neon_stream_xor(req, &subctx, real_iv); |
| 123 | } |
| 124 | |
| 125 | static struct skcipher_alg algs[] = { |
| 126 | { |
| 127 | .base.cra_name = "chacha20", |
| 128 | .base.cra_driver_name = "chacha20-neon", |
| 129 | .base.cra_priority = 300, |
| 130 | .base.cra_blocksize = 1, |
| 131 | .base.cra_ctxsize = sizeof(struct chacha_ctx), |
| 132 | .base.cra_module = THIS_MODULE, |
| 133 | |
| 134 | .min_keysize = CHACHA_KEY_SIZE, |
| 135 | .max_keysize = CHACHA_KEY_SIZE, |
| 136 | .ivsize = CHACHA_IV_SIZE, |
| 137 | .chunksize = CHACHA_BLOCK_SIZE, |
| 138 | .walksize = 4 * CHACHA_BLOCK_SIZE, |
| 139 | .setkey = crypto_chacha20_setkey, |
| 140 | .encrypt = chacha_neon, |
| 141 | .decrypt = chacha_neon, |
| 142 | }, { |
| 143 | .base.cra_name = "xchacha20", |
| 144 | .base.cra_driver_name = "xchacha20-neon", |
| 145 | .base.cra_priority = 300, |
| 146 | .base.cra_blocksize = 1, |
| 147 | .base.cra_ctxsize = sizeof(struct chacha_ctx), |
| 148 | .base.cra_module = THIS_MODULE, |
| 149 | |
| 150 | .min_keysize = CHACHA_KEY_SIZE, |
| 151 | .max_keysize = CHACHA_KEY_SIZE, |
| 152 | .ivsize = XCHACHA_IV_SIZE, |
| 153 | .chunksize = CHACHA_BLOCK_SIZE, |
| 154 | .walksize = 4 * CHACHA_BLOCK_SIZE, |
| 155 | .setkey = crypto_chacha20_setkey, |
| 156 | .encrypt = xchacha_neon, |
| 157 | .decrypt = xchacha_neon, |
| 158 | }, { |
| 159 | .base.cra_name = "xchacha12", |
| 160 | .base.cra_driver_name = "xchacha12-neon", |
| 161 | .base.cra_priority = 300, |
| 162 | .base.cra_blocksize = 1, |
| 163 | .base.cra_ctxsize = sizeof(struct chacha_ctx), |
| 164 | .base.cra_module = THIS_MODULE, |
| 165 | |
| 166 | .min_keysize = CHACHA_KEY_SIZE, |
| 167 | .max_keysize = CHACHA_KEY_SIZE, |
| 168 | .ivsize = XCHACHA_IV_SIZE, |
| 169 | .chunksize = CHACHA_BLOCK_SIZE, |
| 170 | .walksize = 4 * CHACHA_BLOCK_SIZE, |
| 171 | .setkey = crypto_chacha12_setkey, |
| 172 | .encrypt = xchacha_neon, |
| 173 | .decrypt = xchacha_neon, |
| 174 | } |
| 175 | }; |
| 176 | |
| 177 | static int __init chacha_simd_mod_init(void) |
| 178 | { |
| 179 | if (!(elf_hwcap & HWCAP_NEON)) |
| 180 | return -ENODEV; |
| 181 | |
| 182 | return crypto_register_skciphers(algs, ARRAY_SIZE(algs)); |
| 183 | } |
| 184 | |
| 185 | static void __exit chacha_simd_mod_fini(void) |
| 186 | { |
| 187 | crypto_unregister_skciphers(algs, ARRAY_SIZE(algs)); |
| 188 | } |
| 189 | |
| 190 | module_init(chacha_simd_mod_init); |
| 191 | module_exit(chacha_simd_mod_fini); |
| 192 | |
| 193 | MODULE_DESCRIPTION("ChaCha and XChaCha stream ciphers (NEON accelerated)"); |
| 194 | MODULE_AUTHOR("Ard Biesheuvel <ard.biesheuvel@linaro.org>"); |
| 195 | MODULE_LICENSE("GPL v2"); |
| 196 | MODULE_ALIAS_CRYPTO("chacha20"); |
| 197 | MODULE_ALIAS_CRYPTO("chacha20-neon"); |
| 198 | MODULE_ALIAS_CRYPTO("xchacha20"); |
| 199 | MODULE_ALIAS_CRYPTO("xchacha20-neon"); |
| 200 | MODULE_ALIAS_CRYPTO("xchacha12"); |
| 201 | MODULE_ALIAS_CRYPTO("xchacha12-neon"); |