1 /* 2 * Copyright (c) 2016 Thomas Pornin <pornin (at) bolet.org> 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining 5 * a copy of this software and associated documentation files (the 6 * "Software"), to deal in the Software without restriction, including 7 * without limitation the rights to use, copy, modify, merge, publish, 8 * distribute, sublicense, and/or sell copies of the Software, and to 9 * permit persons to whom the Software is furnished to do so, subject to 10 * the following conditions: 11 * 12 * The above copyright notice and this permission notice shall be 13 * included in all copies or substantial portions of the Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 16 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 17 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 18 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 19 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 20 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 21 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 22 * SOFTWARE. 23 */ 24 25 #include <sys/cdefs.h> 26 __KERNEL_RCSID(1, "$NetBSD: aes_sse2_dec.c,v 1.1 2020/06/29 23:47:54 riastradh Exp $"); 27 28 #include <sys/types.h> 29 30 #include "aes_sse2_impl.h" 31 32 /* see inner.h */ 33 void 34 aes_sse2_bitslice_invSbox(__m128i q[static 4]) 35 { 36 /* 37 * See br_aes_ct_bitslice_invSbox(). This is the natural extension 38 * to 64-bit registers. 39 */ 40 __m128i q0, q1, q2, q3, q4, q5, q6, q7; 41 42 q0 = ~q[0]; 43 q1 = ~q[1]; 44 q2 = q[2]; 45 q3 = q[3]; 46 q4 = _mm_shuffle_epi32(q[0], 0x0e); 47 q5 = _mm_shuffle_epi32(~q[1], 0x0e); 48 q6 = _mm_shuffle_epi32(~q[2], 0x0e); 49 q7 = _mm_shuffle_epi32(q[3], 0x0e); 50 51 q[3] = _mm_unpacklo_epi64(q5 ^ q0 ^ q2, q1 ^ q4 ^ q6); 52 q[2] = _mm_unpacklo_epi64(q4 ^ q7 ^ q1, q0 ^ q3 ^ q5); 53 q[1] = _mm_unpacklo_epi64(q3 ^ q6 ^ q0, q7 ^ q2 ^ q4); 54 q[0] = _mm_unpacklo_epi64(q2 ^ q5 ^ q7, q6 ^ q1 ^ q3); 55 56 aes_sse2_bitslice_Sbox(q); 57 58 q0 = ~q[0]; 59 q1 = ~q[1]; 60 q2 = q[2]; 61 q3 = q[3]; 62 q4 = _mm_shuffle_epi32(q[0], 0x0e); 63 q5 = _mm_shuffle_epi32(~q[1], 0x0e); 64 q6 = _mm_shuffle_epi32(~q[2], 0x0e); 65 q7 = _mm_shuffle_epi32(q[3], 0x0e); 66 67 q[3] = _mm_unpacklo_epi64(q5 ^ q0 ^ q2, q1 ^ q4 ^ q6); 68 q[2] = _mm_unpacklo_epi64(q4 ^ q7 ^ q1, q0 ^ q3 ^ q5); 69 q[1] = _mm_unpacklo_epi64(q3 ^ q6 ^ q0, q7 ^ q2 ^ q4); 70 q[0] = _mm_unpacklo_epi64(q2 ^ q5 ^ q7, q6 ^ q1 ^ q3); 71 } 72 73 static inline void 74 add_round_key(__m128i q[static 4], const uint64_t sk[static 8]) 75 { 76 q[0] ^= _mm_set_epi64x(sk[4], sk[0]); 77 q[1] ^= _mm_set_epi64x(sk[5], sk[1]); 78 q[2] ^= _mm_set_epi64x(sk[6], sk[2]); 79 q[3] ^= _mm_set_epi64x(sk[7], sk[3]); 80 } 81 82 static inline __m128i 83 inv_shift_row(__m128i q) 84 { 85 __m128i x, y0, y1, y2, y3, y4, y5, y6; 86 87 x = q; 88 y0 = x & _mm_set1_epi64x(0x000000000000FFFF); 89 y1 = x & _mm_set1_epi64x(0x000000000FFF0000); 90 y2 = x & _mm_set1_epi64x(0x00000000F0000000); 91 y3 = x & _mm_set1_epi64x(0x000000FF00000000); 92 y4 = x & _mm_set1_epi64x(0x0000FF0000000000); 93 y5 = x & _mm_set1_epi64x(0x000F000000000000); 94 y6 = x & _mm_set1_epi64x(0xFFF0000000000000); 95 y1 = _mm_slli_epi64(y1, 4); 96 y2 = _mm_srli_epi64(y2, 12); 97 y3 = _mm_slli_epi64(y3, 8); 98 y4 = _mm_srli_epi64(y4, 8); 99 y5 = _mm_slli_epi64(y5, 12); 100 y6 = _mm_srli_epi64(y6, 4); 101 return y0 | y1 | y2 | y3 | y4 | y5 | y6; 102 } 103 104 static inline void 105 inv_shift_rows(__m128i q[static 4]) 106 { 107 108 q[0] = inv_shift_row(q[0]); 109 q[1] = inv_shift_row(q[1]); 110 q[2] = inv_shift_row(q[2]); 111 q[3] = inv_shift_row(q[3]); 112 } 113 114 static inline __m128i 115 rotr32(__m128i x) 116 { 117 return _mm_slli_epi64(x, 32) | _mm_srli_epi64(x, 32); 118 } 119 120 static inline void 121 inv_mix_columns(__m128i q[4]) 122 { 123 __m128i q0, q1, q2, q3, q4, q5, q6, q7; 124 __m128i r0, r1, r2, r3, r4, r5, r6, r7; 125 __m128i s0, s1, s2, s3, s4, s5, s6, s7; 126 127 q0 = q[0]; 128 q1 = q[1]; 129 q2 = q[2]; 130 q3 = q[3]; 131 r0 = _mm_srli_epi64(q0, 16) | _mm_slli_epi64(q0, 48); 132 r1 = _mm_srli_epi64(q1, 16) | _mm_slli_epi64(q1, 48); 133 r2 = _mm_srli_epi64(q2, 16) | _mm_slli_epi64(q2, 48); 134 r3 = _mm_srli_epi64(q3, 16) | _mm_slli_epi64(q3, 48); 135 136 q7 = _mm_shuffle_epi32(q3, 0x0e); 137 q6 = _mm_shuffle_epi32(q2, 0x0e); 138 q5 = _mm_shuffle_epi32(q1, 0x0e); 139 q4 = _mm_shuffle_epi32(q0, 0x0e); 140 141 r7 = _mm_shuffle_epi32(r3, 0x0e); 142 r6 = _mm_shuffle_epi32(r2, 0x0e); 143 r5 = _mm_shuffle_epi32(r1, 0x0e); 144 r4 = _mm_shuffle_epi32(r0, 0x0e); 145 146 s0 = q5 ^ q6 ^ q7 ^ r0 ^ r5 ^ r7 ^ rotr32(q0 ^ q5 ^ q6 ^ r0 ^ r5); 147 s1 = q0 ^ q5 ^ r0 ^ r1 ^ r5 ^ r6 ^ r7 ^ rotr32(q1 ^ q5 ^ q7 ^ r1 ^ r5 ^ r6); 148 s2 = q0 ^ q1 ^ q6 ^ r1 ^ r2 ^ r6 ^ r7 ^ rotr32(q0 ^ q2 ^ q6 ^ r2 ^ r6 ^ r7); 149 s3 = q0 ^ q1 ^ q2 ^ q5 ^ q6 ^ r0 ^ r2 ^ r3 ^ r5 ^ rotr32(q0 ^ q1 ^ q3 ^ q5 ^ q6 ^ q7 ^ r0 ^ r3 ^ r5 ^ r7); 150 s4 = q1 ^ q2 ^ q3 ^ q5 ^ r1 ^ r3 ^ r4 ^ r5 ^ r6 ^ r7 ^ rotr32(q1 ^ q2 ^ q4 ^ q5 ^ q7 ^ r1 ^ r4 ^ r5 ^ r6); 151 s5 = q2 ^ q3 ^ q4 ^ q6 ^ r2 ^ r4 ^ r5 ^ r6 ^ r7 ^ rotr32(q2 ^ q3 ^ q5 ^ q6 ^ r2 ^ r5 ^ r6 ^ r7); 152 s6 = q3 ^ q4 ^ q5 ^ q7 ^ r3 ^ r5 ^ r6 ^ r7 ^ rotr32(q3 ^ q4 ^ q6 ^ q7 ^ r3 ^ r6 ^ r7); 153 s7 = q4 ^ q5 ^ q6 ^ r4 ^ r6 ^ r7 ^ rotr32(q4 ^ q5 ^ q7 ^ r4 ^ r7); 154 155 q[0] = _mm_unpacklo_epi64(s0, s4); 156 q[1] = _mm_unpacklo_epi64(s1, s5); 157 q[2] = _mm_unpacklo_epi64(s2, s6); 158 q[3] = _mm_unpacklo_epi64(s3, s7); 159 } 160 161 /* see inner.h */ 162 void 163 aes_sse2_bitslice_decrypt(unsigned num_rounds, 164 const uint64_t *skey, __m128i q[static 4]) 165 { 166 unsigned u; 167 168 add_round_key(q, skey + (num_rounds << 3)); 169 for (u = num_rounds - 1; u > 0; u --) { 170 inv_shift_rows(q); 171 aes_sse2_bitslice_invSbox(q); 172 add_round_key(q, skey + (u << 3)); 173 inv_mix_columns(q); 174 } 175 inv_shift_rows(q); 176 aes_sse2_bitslice_invSbox(q); 177 add_round_key(q, skey); 178 } 179
Indexes created Sat Sep 27 21:09:56 GMT 2025