kernel_optimize_test/arch/arm64/crypto/chacha20-neon-core.S
Ard Biesheuvel b7171ce9eb crypto: arm64/chacha20 - implement NEON version based on SSE3 code
This is a straight port to arm64/NEON of the x86 SSE3 implementation
of the ChaCha20 stream cipher. It uses the new skcipher walksize
attribute to process the input in strides of 4x the block size.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2017-01-13 00:26:48 +08:00

451 lines
11 KiB
ArmAsm

/*
* ChaCha20 256-bit cipher algorithm, RFC7539, arm64 NEON functions
*
* Copyright (C) 2016 Linaro, Ltd. <ard.biesheuvel@linaro.org>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*
* Based on:
* ChaCha20 256-bit cipher algorithm, RFC7539, x64 SSSE3 functions
*
* Copyright (C) 2015 Martin Willi
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*/
#include <linux/linkage.h>
.text
.align 6
ENTRY(chacha20_block_xor_neon)
// x0: Input state matrix, s
// x1: 1 data block output, o
// x2: 1 data block input, i
//
// This function encrypts one ChaCha20 block by loading the state matrix
// in four NEON registers. It performs matrix operation on four words in
// parallel, but requires shuffling to rearrange the words after each
// round.
//
// x0..3 = s0..3
adr x3, ROT8
ld1 {v0.4s-v3.4s}, [x0]
ld1 {v8.4s-v11.4s}, [x0]
ld1 {v12.4s}, [x3]
mov x3, #10
.Ldoubleround:
// x0 += x1, x3 = rotl32(x3 ^ x0, 16)
add v0.4s, v0.4s, v1.4s
eor v3.16b, v3.16b, v0.16b
rev32 v3.8h, v3.8h
// x2 += x3, x1 = rotl32(x1 ^ x2, 12)
add v2.4s, v2.4s, v3.4s
eor v4.16b, v1.16b, v2.16b
shl v1.4s, v4.4s, #12
sri v1.4s, v4.4s, #20
// x0 += x1, x3 = rotl32(x3 ^ x0, 8)
add v0.4s, v0.4s, v1.4s
eor v3.16b, v3.16b, v0.16b
tbl v3.16b, {v3.16b}, v12.16b
// x2 += x3, x1 = rotl32(x1 ^ x2, 7)
add v2.4s, v2.4s, v3.4s
eor v4.16b, v1.16b, v2.16b
shl v1.4s, v4.4s, #7
sri v1.4s, v4.4s, #25
// x1 = shuffle32(x1, MASK(0, 3, 2, 1))
ext v1.16b, v1.16b, v1.16b, #4
// x2 = shuffle32(x2, MASK(1, 0, 3, 2))
ext v2.16b, v2.16b, v2.16b, #8
// x3 = shuffle32(x3, MASK(2, 1, 0, 3))
ext v3.16b, v3.16b, v3.16b, #12
// x0 += x1, x3 = rotl32(x3 ^ x0, 16)
add v0.4s, v0.4s, v1.4s
eor v3.16b, v3.16b, v0.16b
rev32 v3.8h, v3.8h
// x2 += x3, x1 = rotl32(x1 ^ x2, 12)
add v2.4s, v2.4s, v3.4s
eor v4.16b, v1.16b, v2.16b
shl v1.4s, v4.4s, #12
sri v1.4s, v4.4s, #20
// x0 += x1, x3 = rotl32(x3 ^ x0, 8)
add v0.4s, v0.4s, v1.4s
eor v3.16b, v3.16b, v0.16b
tbl v3.16b, {v3.16b}, v12.16b
// x2 += x3, x1 = rotl32(x1 ^ x2, 7)
add v2.4s, v2.4s, v3.4s
eor v4.16b, v1.16b, v2.16b
shl v1.4s, v4.4s, #7
sri v1.4s, v4.4s, #25
// x1 = shuffle32(x1, MASK(2, 1, 0, 3))
ext v1.16b, v1.16b, v1.16b, #12
// x2 = shuffle32(x2, MASK(1, 0, 3, 2))
ext v2.16b, v2.16b, v2.16b, #8
// x3 = shuffle32(x3, MASK(0, 3, 2, 1))
ext v3.16b, v3.16b, v3.16b, #4
subs x3, x3, #1
b.ne .Ldoubleround
ld1 {v4.16b-v7.16b}, [x2]
// o0 = i0 ^ (x0 + s0)
add v0.4s, v0.4s, v8.4s
eor v0.16b, v0.16b, v4.16b
// o1 = i1 ^ (x1 + s1)
add v1.4s, v1.4s, v9.4s
eor v1.16b, v1.16b, v5.16b
// o2 = i2 ^ (x2 + s2)
add v2.4s, v2.4s, v10.4s
eor v2.16b, v2.16b, v6.16b
// o3 = i3 ^ (x3 + s3)
add v3.4s, v3.4s, v11.4s
eor v3.16b, v3.16b, v7.16b
st1 {v0.16b-v3.16b}, [x1]
ret
ENDPROC(chacha20_block_xor_neon)
.align 6
ENTRY(chacha20_4block_xor_neon)
// x0: Input state matrix, s
// x1: 4 data blocks output, o
// x2: 4 data blocks input, i
//
// This function encrypts four consecutive ChaCha20 blocks by loading
// the state matrix in NEON registers four times. The algorithm performs
// each operation on the corresponding word of each state matrix, hence
// requires no word shuffling. For final XORing step we transpose the
// matrix by interleaving 32- and then 64-bit words, which allows us to
// do XOR in NEON registers.
//
adr x3, CTRINC // ... and ROT8
ld1 {v30.4s-v31.4s}, [x3]
// x0..15[0-3] = s0..3[0..3]
mov x4, x0
ld4r { v0.4s- v3.4s}, [x4], #16
ld4r { v4.4s- v7.4s}, [x4], #16
ld4r { v8.4s-v11.4s}, [x4], #16
ld4r {v12.4s-v15.4s}, [x4]
// x12 += counter values 0-3
add v12.4s, v12.4s, v30.4s
mov x3, #10
.Ldoubleround4:
// x0 += x4, x12 = rotl32(x12 ^ x0, 16)
// x1 += x5, x13 = rotl32(x13 ^ x1, 16)
// x2 += x6, x14 = rotl32(x14 ^ x2, 16)
// x3 += x7, x15 = rotl32(x15 ^ x3, 16)
add v0.4s, v0.4s, v4.4s
add v1.4s, v1.4s, v5.4s
add v2.4s, v2.4s, v6.4s
add v3.4s, v3.4s, v7.4s
eor v12.16b, v12.16b, v0.16b
eor v13.16b, v13.16b, v1.16b
eor v14.16b, v14.16b, v2.16b
eor v15.16b, v15.16b, v3.16b
rev32 v12.8h, v12.8h
rev32 v13.8h, v13.8h
rev32 v14.8h, v14.8h
rev32 v15.8h, v15.8h
// x8 += x12, x4 = rotl32(x4 ^ x8, 12)
// x9 += x13, x5 = rotl32(x5 ^ x9, 12)
// x10 += x14, x6 = rotl32(x6 ^ x10, 12)
// x11 += x15, x7 = rotl32(x7 ^ x11, 12)
add v8.4s, v8.4s, v12.4s
add v9.4s, v9.4s, v13.4s
add v10.4s, v10.4s, v14.4s
add v11.4s, v11.4s, v15.4s
eor v16.16b, v4.16b, v8.16b
eor v17.16b, v5.16b, v9.16b
eor v18.16b, v6.16b, v10.16b
eor v19.16b, v7.16b, v11.16b
shl v4.4s, v16.4s, #12
shl v5.4s, v17.4s, #12
shl v6.4s, v18.4s, #12
shl v7.4s, v19.4s, #12
sri v4.4s, v16.4s, #20
sri v5.4s, v17.4s, #20
sri v6.4s, v18.4s, #20
sri v7.4s, v19.4s, #20
// x0 += x4, x12 = rotl32(x12 ^ x0, 8)
// x1 += x5, x13 = rotl32(x13 ^ x1, 8)
// x2 += x6, x14 = rotl32(x14 ^ x2, 8)
// x3 += x7, x15 = rotl32(x15 ^ x3, 8)
add v0.4s, v0.4s, v4.4s
add v1.4s, v1.4s, v5.4s
add v2.4s, v2.4s, v6.4s
add v3.4s, v3.4s, v7.4s
eor v12.16b, v12.16b, v0.16b
eor v13.16b, v13.16b, v1.16b
eor v14.16b, v14.16b, v2.16b
eor v15.16b, v15.16b, v3.16b
tbl v12.16b, {v12.16b}, v31.16b
tbl v13.16b, {v13.16b}, v31.16b
tbl v14.16b, {v14.16b}, v31.16b
tbl v15.16b, {v15.16b}, v31.16b
// x8 += x12, x4 = rotl32(x4 ^ x8, 7)
// x9 += x13, x5 = rotl32(x5 ^ x9, 7)
// x10 += x14, x6 = rotl32(x6 ^ x10, 7)
// x11 += x15, x7 = rotl32(x7 ^ x11, 7)
add v8.4s, v8.4s, v12.4s
add v9.4s, v9.4s, v13.4s
add v10.4s, v10.4s, v14.4s
add v11.4s, v11.4s, v15.4s
eor v16.16b, v4.16b, v8.16b
eor v17.16b, v5.16b, v9.16b
eor v18.16b, v6.16b, v10.16b
eor v19.16b, v7.16b, v11.16b
shl v4.4s, v16.4s, #7
shl v5.4s, v17.4s, #7
shl v6.4s, v18.4s, #7
shl v7.4s, v19.4s, #7
sri v4.4s, v16.4s, #25
sri v5.4s, v17.4s, #25
sri v6.4s, v18.4s, #25
sri v7.4s, v19.4s, #25
// x0 += x5, x15 = rotl32(x15 ^ x0, 16)
// x1 += x6, x12 = rotl32(x12 ^ x1, 16)
// x2 += x7, x13 = rotl32(x13 ^ x2, 16)
// x3 += x4, x14 = rotl32(x14 ^ x3, 16)
add v0.4s, v0.4s, v5.4s
add v1.4s, v1.4s, v6.4s
add v2.4s, v2.4s, v7.4s
add v3.4s, v3.4s, v4.4s
eor v15.16b, v15.16b, v0.16b
eor v12.16b, v12.16b, v1.16b
eor v13.16b, v13.16b, v2.16b
eor v14.16b, v14.16b, v3.16b
rev32 v15.8h, v15.8h
rev32 v12.8h, v12.8h
rev32 v13.8h, v13.8h
rev32 v14.8h, v14.8h
// x10 += x15, x5 = rotl32(x5 ^ x10, 12)
// x11 += x12, x6 = rotl32(x6 ^ x11, 12)
// x8 += x13, x7 = rotl32(x7 ^ x8, 12)
// x9 += x14, x4 = rotl32(x4 ^ x9, 12)
add v10.4s, v10.4s, v15.4s
add v11.4s, v11.4s, v12.4s
add v8.4s, v8.4s, v13.4s
add v9.4s, v9.4s, v14.4s
eor v16.16b, v5.16b, v10.16b
eor v17.16b, v6.16b, v11.16b
eor v18.16b, v7.16b, v8.16b
eor v19.16b, v4.16b, v9.16b
shl v5.4s, v16.4s, #12
shl v6.4s, v17.4s, #12
shl v7.4s, v18.4s, #12
shl v4.4s, v19.4s, #12
sri v5.4s, v16.4s, #20
sri v6.4s, v17.4s, #20
sri v7.4s, v18.4s, #20
sri v4.4s, v19.4s, #20
// x0 += x5, x15 = rotl32(x15 ^ x0, 8)
// x1 += x6, x12 = rotl32(x12 ^ x1, 8)
// x2 += x7, x13 = rotl32(x13 ^ x2, 8)
// x3 += x4, x14 = rotl32(x14 ^ x3, 8)
add v0.4s, v0.4s, v5.4s
add v1.4s, v1.4s, v6.4s
add v2.4s, v2.4s, v7.4s
add v3.4s, v3.4s, v4.4s
eor v15.16b, v15.16b, v0.16b
eor v12.16b, v12.16b, v1.16b
eor v13.16b, v13.16b, v2.16b
eor v14.16b, v14.16b, v3.16b
tbl v15.16b, {v15.16b}, v31.16b
tbl v12.16b, {v12.16b}, v31.16b
tbl v13.16b, {v13.16b}, v31.16b
tbl v14.16b, {v14.16b}, v31.16b
// x10 += x15, x5 = rotl32(x5 ^ x10, 7)
// x11 += x12, x6 = rotl32(x6 ^ x11, 7)
// x8 += x13, x7 = rotl32(x7 ^ x8, 7)
// x9 += x14, x4 = rotl32(x4 ^ x9, 7)
add v10.4s, v10.4s, v15.4s
add v11.4s, v11.4s, v12.4s
add v8.4s, v8.4s, v13.4s
add v9.4s, v9.4s, v14.4s
eor v16.16b, v5.16b, v10.16b
eor v17.16b, v6.16b, v11.16b
eor v18.16b, v7.16b, v8.16b
eor v19.16b, v4.16b, v9.16b
shl v5.4s, v16.4s, #7
shl v6.4s, v17.4s, #7
shl v7.4s, v18.4s, #7
shl v4.4s, v19.4s, #7
sri v5.4s, v16.4s, #25
sri v6.4s, v17.4s, #25
sri v7.4s, v18.4s, #25
sri v4.4s, v19.4s, #25
subs x3, x3, #1
b.ne .Ldoubleround4
ld4r {v16.4s-v19.4s}, [x0], #16
ld4r {v20.4s-v23.4s}, [x0], #16
// x12 += counter values 0-3
add v12.4s, v12.4s, v30.4s
// x0[0-3] += s0[0]
// x1[0-3] += s0[1]
// x2[0-3] += s0[2]
// x3[0-3] += s0[3]
add v0.4s, v0.4s, v16.4s
add v1.4s, v1.4s, v17.4s
add v2.4s, v2.4s, v18.4s
add v3.4s, v3.4s, v19.4s
ld4r {v24.4s-v27.4s}, [x0], #16
ld4r {v28.4s-v31.4s}, [x0]
// x4[0-3] += s1[0]
// x5[0-3] += s1[1]
// x6[0-3] += s1[2]
// x7[0-3] += s1[3]
add v4.4s, v4.4s, v20.4s
add v5.4s, v5.4s, v21.4s
add v6.4s, v6.4s, v22.4s
add v7.4s, v7.4s, v23.4s
// x8[0-3] += s2[0]
// x9[0-3] += s2[1]
// x10[0-3] += s2[2]
// x11[0-3] += s2[3]
add v8.4s, v8.4s, v24.4s
add v9.4s, v9.4s, v25.4s
add v10.4s, v10.4s, v26.4s
add v11.4s, v11.4s, v27.4s
// x12[0-3] += s3[0]
// x13[0-3] += s3[1]
// x14[0-3] += s3[2]
// x15[0-3] += s3[3]
add v12.4s, v12.4s, v28.4s
add v13.4s, v13.4s, v29.4s
add v14.4s, v14.4s, v30.4s
add v15.4s, v15.4s, v31.4s
// interleave 32-bit words in state n, n+1
zip1 v16.4s, v0.4s, v1.4s
zip2 v17.4s, v0.4s, v1.4s
zip1 v18.4s, v2.4s, v3.4s
zip2 v19.4s, v2.4s, v3.4s
zip1 v20.4s, v4.4s, v5.4s
zip2 v21.4s, v4.4s, v5.4s
zip1 v22.4s, v6.4s, v7.4s
zip2 v23.4s, v6.4s, v7.4s
zip1 v24.4s, v8.4s, v9.4s
zip2 v25.4s, v8.4s, v9.4s
zip1 v26.4s, v10.4s, v11.4s
zip2 v27.4s, v10.4s, v11.4s
zip1 v28.4s, v12.4s, v13.4s
zip2 v29.4s, v12.4s, v13.4s
zip1 v30.4s, v14.4s, v15.4s
zip2 v31.4s, v14.4s, v15.4s
// interleave 64-bit words in state n, n+2
zip1 v0.2d, v16.2d, v18.2d
zip2 v4.2d, v16.2d, v18.2d
zip1 v8.2d, v17.2d, v19.2d
zip2 v12.2d, v17.2d, v19.2d
ld1 {v16.16b-v19.16b}, [x2], #64
zip1 v1.2d, v20.2d, v22.2d
zip2 v5.2d, v20.2d, v22.2d
zip1 v9.2d, v21.2d, v23.2d
zip2 v13.2d, v21.2d, v23.2d
ld1 {v20.16b-v23.16b}, [x2], #64
zip1 v2.2d, v24.2d, v26.2d
zip2 v6.2d, v24.2d, v26.2d
zip1 v10.2d, v25.2d, v27.2d
zip2 v14.2d, v25.2d, v27.2d
ld1 {v24.16b-v27.16b}, [x2], #64
zip1 v3.2d, v28.2d, v30.2d
zip2 v7.2d, v28.2d, v30.2d
zip1 v11.2d, v29.2d, v31.2d
zip2 v15.2d, v29.2d, v31.2d
ld1 {v28.16b-v31.16b}, [x2]
// xor with corresponding input, write to output
eor v16.16b, v16.16b, v0.16b
eor v17.16b, v17.16b, v1.16b
eor v18.16b, v18.16b, v2.16b
eor v19.16b, v19.16b, v3.16b
eor v20.16b, v20.16b, v4.16b
eor v21.16b, v21.16b, v5.16b
st1 {v16.16b-v19.16b}, [x1], #64
eor v22.16b, v22.16b, v6.16b
eor v23.16b, v23.16b, v7.16b
eor v24.16b, v24.16b, v8.16b
eor v25.16b, v25.16b, v9.16b
st1 {v20.16b-v23.16b}, [x1], #64
eor v26.16b, v26.16b, v10.16b
eor v27.16b, v27.16b, v11.16b
eor v28.16b, v28.16b, v12.16b
st1 {v24.16b-v27.16b}, [x1], #64
eor v29.16b, v29.16b, v13.16b
eor v30.16b, v30.16b, v14.16b
eor v31.16b, v31.16b, v15.16b
st1 {v28.16b-v31.16b}, [x1]
ret
ENDPROC(chacha20_4block_xor_neon)
CTRINC: .word 0, 1, 2, 3
ROT8: .word 0x02010003, 0x06050407, 0x0a09080b, 0x0e0d0c0f