|
1 | | -//! Software emulation support for CLMUL hardware intrinsics. |
2 | | -//! |
3 | | -//! WARNING: Not constant time! Should be made constant-time or disabled by default. |
4 | | -
|
5 | | -// TODO(tarcieri): performance-oriented constant-time implementation |
6 | | -// See: <https://bearssl.org/gitweb/?p=BearSSL;a=blob;f=src/hash/ghash_ctmul64.c> |
| 1 | +//! Constant-time software implementation of POLYVAL |
| 2 | +
|
| 3 | +// Adapted from BearSSL's `ghash_ctmul64.c` |
| 4 | +// <https://bearssl.org/gitweb/?p=BearSSL;a=blob;f=src/hash/ghash_ctmul64.c;hb=4b6046412bf927d6424f20fc7ee495bb96dbd227> |
| 5 | +// |
| 6 | +// Copyright (c) 2016 Thomas Pornin <pornin@bolet.org> |
| 7 | +// |
| 8 | +// Permission is hereby granted, free of charge, to any person obtaining |
| 9 | +// a copy of this software and associated documentation files (the |
| 10 | +// "Software"), to deal in the Software without restriction, including |
| 11 | +// without limitation the rights to use, copy, modify, merge, publish, |
| 12 | +// distribute, sublicense, and/or sell copies of the Software, and to |
| 13 | +// permit persons to whom the Software is furnished to do so, subject to |
| 14 | +// the following conditions: |
| 15 | +// |
| 16 | +// The above copyright notice and this permission notice shall be |
| 17 | +// included in all copies or substantial portions of the Software. |
| 18 | +// |
| 19 | +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, |
| 20 | +// EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
| 21 | +// MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
| 22 | +// NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS |
| 23 | +// BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN |
| 24 | +// ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN |
| 25 | +// CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 26 | +// SOFTWARE. |
| 27 | + |
| 28 | +#![allow(missing_docs)] |
7 | 29 |
|
8 | 30 | use super::Backend; |
9 | 31 | use crate::field::Block; |
10 | | -use core::{convert::TryInto, ops::Add}; |
| 32 | +use core::{ |
| 33 | + convert::TryInto, |
| 34 | + ops::{Add, Mul}, |
| 35 | +}; |
11 | 36 |
|
12 | | -/// 2 x `u64` values emulating an XMM register |
| 37 | +/// 2 x `u64` values |
13 | 38 | #[derive(Copy, Clone, Debug, Eq, PartialEq)] |
14 | 39 | pub struct U64x2(u64, u64); |
15 | 40 |
|
| 41 | +impl Backend for U64x2 {} |
| 42 | + |
16 | 43 | impl From<Block> for U64x2 { |
17 | 44 | fn from(bytes: Block) -> U64x2 { |
18 | 45 | U64x2( |
@@ -43,53 +70,104 @@ impl From<U64x2> for u128 { |
43 | 70 | } |
44 | 71 | } |
45 | 72 |
|
| 73 | +#[allow(clippy::suspicious_arithmetic_impl)] |
46 | 74 | impl Add for U64x2 { |
47 | 75 | type Output = Self; |
48 | 76 |
|
49 | 77 | /// Adds two POLYVAL field elements. |
50 | | - fn add(self, rhs: Self) -> Self { |
| 78 | + fn add(self, rhs: Self) -> Self::Output { |
51 | 79 | U64x2(self.0 ^ rhs.0, self.1 ^ rhs.1) |
52 | 80 | } |
53 | 81 | } |
54 | 82 |
|
55 | | -impl Backend for U64x2 { |
56 | | - fn clmul(self, other: Self, imm: u8) -> Self { |
57 | | - let (a, b) = match imm.into() { |
58 | | - 0x00 => (self.0, other.0), |
59 | | - 0x01 => (self.1, other.0), |
60 | | - 0x10 => (self.0, other.1), |
61 | | - 0x11 => (self.1, other.1), |
62 | | - _ => unreachable!(), |
63 | | - }; |
| 83 | +#[allow(clippy::suspicious_arithmetic_impl)] |
| 84 | +impl Mul for U64x2 { |
| 85 | + type Output = Self; |
64 | 86 |
|
65 | | - let mut result = U64x2(0, 0); |
| 87 | + /// Computes POLYVAL multiplication over GF(2^128). |
| 88 | + // TODO(tarcieri): actually adapt the arithmetic below from GHASH |
| 89 | + fn mul(self, rhs: Self) -> Self { |
| 90 | + let h0 = self.0; |
| 91 | + let h1 = self.1; |
| 92 | + let h0r = rev64(h0); |
| 93 | + let h1r = rev64(h1); |
| 94 | + let h2 = h0 ^ h1; |
| 95 | + let h2r = h0r ^ h1r; |
| 96 | + |
| 97 | + let y0 = rhs.0; |
| 98 | + let y1 = rhs.1; |
| 99 | + let y0r = rev64(y0); |
| 100 | + let y1r = rev64(y1); |
| 101 | + let y2 = y0 ^ y1; |
| 102 | + let y2r = y0r ^ y1r; |
| 103 | + let z0 = bmul64(y0, h0); |
| 104 | + let z1 = bmul64(y1, h1); |
| 105 | + |
| 106 | + let mut z2 = bmul64(y2, h2); |
| 107 | + let mut z0h = bmul64(y0r, h0r); |
| 108 | + let mut z1h = bmul64(y1r, h1r); |
| 109 | + let mut z2h = bmul64(y2r, h2r); |
| 110 | + |
| 111 | + z2 ^= z0 ^ z1; |
| 112 | + z2h ^= z0h ^ z1h; |
| 113 | + z0h = rev64(z0h) >> 1; |
| 114 | + z1h = rev64(z1h) >> 1; |
| 115 | + z2h = rev64(z2h) >> 1; |
| 116 | + |
| 117 | + let mut v0 = z0; |
| 118 | + let mut v1 = z0h ^ z2; |
| 119 | + let mut v2 = z1 ^ z2h; |
| 120 | + let mut v3 = z1h; |
| 121 | + |
| 122 | + v3 = v3 << 1 | v2 >> 63; |
| 123 | + v2 = v2 << 1 | v1 >> 63; |
| 124 | + v1 = v1 << 1 | v0 >> 63; |
| 125 | + v0 <<= 1; |
| 126 | + |
| 127 | + v2 ^= v0 ^ v0 >> 1 ^ v0 >> 2 ^ v0 >> 7; |
| 128 | + v1 ^= v0 << 63 ^ v0 << 62 ^ v0 << 57; |
| 129 | + v3 ^= v1 ^ v1 >> 1 ^ v1 >> 2 ^ v1 >> 7; |
| 130 | + v2 ^= v1 << 63 ^ v1 << 62 ^ v1 << 57; |
| 131 | + |
| 132 | + U64x2(v2, v3) |
| 133 | + } |
| 134 | +} |
66 | 135 |
|
67 | | - for i in 0..64 { |
68 | | - if b & (1 << i) != 0 { |
69 | | - result.1 ^= a; |
70 | | - } |
| 136 | +fn rev64(mut x: u64) -> u64 { |
| 137 | + x = ((x & 0x5555_5555_5555_5555) << 1) | ((x >> 1) & 0x5555_5555_5555_5555); |
| 138 | + x = ((x & 0x3333_3333_3333_3333) << 2) | ((x >> 2) & 0x3333_3333_3333_3333); |
| 139 | + x = ((x & 0x0f0f_0f0f_0f0f_0f0f) << 4) | ((x >> 4) & 0x0f0f_0f0f_0f0f_0f0f); |
| 140 | + x = ((x & 0x00ff_00ff_00ff_00ff) << 8) | ((x >> 8) & 0x00ff_00ff_00ff_00ff); |
| 141 | + x = ((x & 0xffff_0000_ffff) << 16) | ((x >> 16) & 0xffff_0000_ffff); |
| 142 | + (x << 32) | (x >> 32) |
| 143 | +} |
71 | 144 |
|
72 | | - result.0 >>= 1; |
| 145 | +fn bmul64(x: u64, y: u64) -> u64 { |
| 146 | + let x0 = x & 0x1111_1111_1111_1111; |
| 147 | + let x1 = x & 0x2222_2222_2222_2222; |
| 148 | + let x2 = x & 0x4444_4444_4444_4444; |
| 149 | + let x3 = x & 0x8888_8888_8888_8888; |
| 150 | + let y0 = y & 0x1111_1111_1111_1111; |
| 151 | + let y1 = y & 0x2222_2222_2222_2222; |
| 152 | + let y2 = y & 0x4444_4444_4444_4444; |
| 153 | + let y3 = y & 0x8888_8888_8888_8888; |
73 | 154 |
|
74 | | - if result.1 & 1 != 0 { |
75 | | - result.0 ^= 1 << 63; |
76 | | - } |
| 155 | + let mut z0 = |
| 156 | + x0.wrapping_mul(y0) ^ x1.wrapping_mul(y3) ^ x2.wrapping_mul(y2) ^ x3.wrapping_mul(y1); |
77 | 157 |
|
78 | | - result.1 >>= 1; |
79 | | - } |
| 158 | + let mut z1 = |
| 159 | + x0.wrapping_mul(y1) ^ x1.wrapping_mul(y0) ^ x2.wrapping_mul(y3) ^ x3.wrapping_mul(y2); |
80 | 160 |
|
81 | | - result |
82 | | - } |
| 161 | + let mut z2 = |
| 162 | + x0.wrapping_mul(y2) ^ x1.wrapping_mul(y1) ^ x2.wrapping_mul(y0) ^ x3.wrapping_mul(y3); |
83 | 163 |
|
84 | | - fn shuffle(self) -> Self { |
85 | | - U64x2(self.1, self.0) |
86 | | - } |
| 164 | + let mut z3 = |
| 165 | + x0.wrapping_mul(y3) ^ x1.wrapping_mul(y2) ^ x2.wrapping_mul(y1) ^ x3.wrapping_mul(y0); |
87 | 166 |
|
88 | | - fn shl64(self) -> Self { |
89 | | - U64x2(0, self.0) |
90 | | - } |
| 167 | + z0 &= 0x1111_1111_1111_1111; |
| 168 | + z1 &= 0x2222_2222_2222_2222; |
| 169 | + z2 &= 0x4444_4444_4444_4444; |
| 170 | + z3 &= 0x8888_8888_8888_8888; |
91 | 171 |
|
92 | | - fn shr64(self) -> Self { |
93 | | - U64x2(self.1, 0) |
94 | | - } |
| 172 | + z0 | z1 | z2 | z3 |
95 | 173 | } |
0 commit comments