public-mirrors/linux
1
0
Fork 0
mirror of git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git synced 2025-08-05 16:54:27 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
linux/lib/crypto/riscv/chacha-riscv64-zvkb.S

298 lines
8.2 KiB
ArmAsm
Raw Permalink Normal View History

crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
/* 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
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
#define STATEP a0
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
#define INP a1
#define OUTP a2
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
#define NBLOCKS a3
#define NROUNDS a4
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
#define CONSTS0 a5
#define CONSTS1 a6
#define CONSTS2 a7
#define CONSTS3 t0
#define TMP t1
#define VL t2
#define STRIDE t3
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
#define ROUND_CTR t4
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
#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
crypto: lib/chacha - strongly type the ChaCha state The ChaCha state matrix is 16 32-bit words. Currently it is represented in the code as a raw u32 array, or even just a pointer to u32. This weak typing is error-prone. Instead, introduce struct chacha_state: struct chacha_state { u32 x[16]; }; Convert all ChaCha and HChaCha functions to use struct chacha_state. No functional changes. Signed-off-by: Eric Biggers <ebiggers@google.com> Acked-by: Kent Overstreet <kent.overstreet@linux.dev> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-05-05 11:18:21 -07:00
// void chacha_zvkb(struct chacha_state *state, const u8 *in, u8 *out,
// size_t nblocks, int nrounds);
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
//
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
// |nblocks| is the number of 64-byte blocks to process, and must be nonzero.
//
// |state| gives the ChaCha state matrix, including the 32-bit counter in
crypto: lib/chacha - strongly type the ChaCha state The ChaCha state matrix is 16 32-bit words. Currently it is represented in the code as a raw u32 array, or even just a pointer to u32. This weak typing is error-prone. Instead, introduce struct chacha_state: struct chacha_state { u32 x[16]; }; Convert all ChaCha and HChaCha functions to use struct chacha_state. No functional changes. Signed-off-by: Eric Biggers <ebiggers@google.com> Acked-by: Kent Overstreet <kent.overstreet@linux.dev> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-05-05 11:18:21 -07:00
// state->x[12] following the RFC7539 convention; note that this differs from
// the original Salsa20 paper which uses a 64-bit counter in state->x[12..13].
// The updated 32-bit counter is written back to state->x[12] before returning.
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
SYM_FUNC_START(chacha_zvkb)
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
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.
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
lw CONSTS0, 0(STATEP)
lw CONSTS1, 4(STATEP)
lw CONSTS2, 8(STATEP)
lw CONSTS3, 12(STATEP)
lw KEY0, 16(STATEP)
lw KEY1, 20(STATEP)
lw KEY2, 24(STATEP)
lw KEY3, 28(STATEP)
lw KEY4, 32(STATEP)
lw KEY5, 36(STATEP)
lw KEY6, 40(STATEP)
lw KEY7, 44(STATEP)
lw COUNTER, 48(STATEP)
lw NONCE0, 52(STATEP)
lw NONCE1, 56(STATEP)
lw NONCE2, 60(STATEP)
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
.Lblock_loop:
// Set vl to the number of blocks to process in this iteration.
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
vsetvli VL, NBLOCKS, e32, m1, ta, ma
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
// 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
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
mv ROUND_CTR, NROUNDS
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
.Lnext_doubleround:
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
addi ROUND_CTR, ROUND_CTR, -2
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
// 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
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
bnez ROUND_CTR, .Lnext_doubleround
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
// 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
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
sub NBLOCKS, NBLOCKS, VL
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
slli TMP, VL, 6
add OUTP, OUTP, TMP
add INP, INP, TMP
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
bnez NBLOCKS, .Lblock_loop
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
sw COUNTER, 48(STATEP)
crypto: riscv - add vector crypto accelerated ChaCha20 Add an implementation of ChaCha20 using the Zvkb extension. The assembly code is derived from OpenSSL code (openssl/openssl#21923) that was dual-licensed so that it could be reused in the kernel. Nevertheless, the assembly has been significantly reworked for integration with the kernel, for example by using a regular .S file instead of the so-called perlasm, using the assembler instead of bare '.inst', and reducing code duplication. Signed-off-by: Jerry Shih <jerry.shih@sifive.com> Co-developed-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Eric Biggers <ebiggers@google.com> Link: https://lore.kernel.org/r/20240122002024.27477-6-ebiggers@kernel.org Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
2024-01-21 16:19:16 -08:00
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
crypto: riscv/chacha - implement library instead of skcipher Currently the RISC-V optimized ChaCha20 is only wired up to the crypto_skcipher API, which makes it unavailable to users of the library API. The crypto_skcipher API for ChaCha20 is going to change to be implemented on top of the library API, so the library API needs to be supported. And of course it's needed anyway to serve the library users. Therefore, change the RISC-V ChaCha20 code to implement the library API instead of the crypto_skcipher API. The library functions take the ChaCha state matrix directly (instead of key and IV) and support both ChaCha20 and ChaCha12. To make the RISC-V code work properly for that, change the assembly code to take the state matrix directly and add a nrounds parameter. Signed-off-by: Eric Biggers <ebiggers@google.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2025-04-05 11:26:01 -07:00
SYM_FUNC_END(chacha_zvkb)
Reference in a new issue Copy permalink