1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (C) 2017-2019 Linaro Ltd <ard.biesheuvel@linaro.org>
4 */
5
6 #include <crypto/aes.h>
7 #include <linux/crypto.h>
8 #include <linux/module.h>
9 #include <asm/unaligned.h>
10
11 /*
12 * Emit the sbox as volatile const to prevent the compiler from doing
13 * constant folding on sbox references involving fixed indexes.
14 */
15 static volatile const u8 __cacheline_aligned aes_sbox[] = {
16 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5,
17 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
18 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0,
19 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
20 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc,
21 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
22 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a,
23 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
24 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0,
25 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
26 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b,
27 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
28 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85,
29 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
30 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5,
31 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
32 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17,
33 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
34 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88,
35 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
36 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c,
37 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
38 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9,
39 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
40 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6,
41 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
42 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e,
43 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
44 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94,
45 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
46 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68,
47 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16,
48 };
49
50 static volatile const u8 __cacheline_aligned aes_inv_sbox[] = {
51 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38,
52 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb,
53 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87,
54 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb,
55 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d,
56 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e,
57 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2,
58 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25,
59 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16,
60 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92,
61 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda,
62 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84,
63 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a,
64 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06,
65 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02,
66 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b,
67 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea,
68 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73,
69 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85,
70 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e,
71 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89,
72 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b,
73 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20,
74 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4,
75 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31,
76 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f,
77 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d,
78 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef,
79 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0,
80 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61,
81 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26,
82 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d,
83 };
84
85 extern const u8 crypto_aes_sbox[256] __alias(aes_sbox);
86 extern const u8 crypto_aes_inv_sbox[256] __alias(aes_inv_sbox);
87
88 EXPORT_SYMBOL(crypto_aes_sbox);
89 EXPORT_SYMBOL(crypto_aes_inv_sbox);
90
91 static u32 mul_by_x(u32 w)
92 {
93 u32 x = w & 0x7f7f7f7f;
94 u32 y = w & 0x80808080;
95
96 /* multiply by polynomial 'x' (0b10) in GF(2^8) */
97 return (x << 1) ^ (y >> 7) * 0x1b;
98 }
99
100 static u32 mul_by_x2(u32 w)
101 {
102 u32 x = w & 0x3f3f3f3f;
103 u32 y = w & 0x80808080;
104 u32 z = w & 0x40404040;
105
106 /* multiply by polynomial 'x^2' (0b100) in GF(2^8) */
107 return (x << 2) ^ (y >> 7) * 0x36 ^ (z >> 6) * 0x1b;
108 }
109
110 static u32 mix_columns(u32 x)
111 {
112 /*
113 * Perform the following matrix multiplication in GF(2^8)
114 *
115 * | 0x2 0x3 0x1 0x1 | | x[0] |
116 * | 0x1 0x2 0x3 0x1 | | x[1] |
117 * | 0x1 0x1 0x2 0x3 | x | x[2] |
118 * | 0x3 0x1 0x1 0x2 | | x[3] |
119 */
120 u32 y = mul_by_x(x) ^ ror32(x, 16);
121
122 return y ^ ror32(x ^ y, 8);
123 }
124
125 static u32 inv_mix_columns(u32 x)
126 {
127 /*
128 * Perform the following matrix multiplication in GF(2^8)
129 *
130 * | 0xe 0xb 0xd 0x9 | | x[0] |
131 * | 0x9 0xe 0xb 0xd | | x[1] |
132 * | 0xd 0x9 0xe 0xb | x | x[2] |
133 * | 0xb 0xd 0x9 0xe | | x[3] |
134 *
135 * which can conveniently be reduced to
136 *
137 * | 0x2 0x3 0x1 0x1 | | 0x5 0x0 0x4 0x0 | | x[0] |
138 * | 0x1 0x2 0x3 0x1 | | 0x0 0x5 0x0 0x4 | | x[1] |
139 * | 0x1 0x1 0x2 0x3 | x | 0x4 0x0 0x5 0x0 | x | x[2] |
140 * | 0x3 0x1 0x1 0x2 | | 0x0 0x4 0x0 0x5 | | x[3] |
141 */
142 u32 y = mul_by_x2(x);
143
144 return mix_columns(x ^ y ^ ror32(y, 16));
145 }
146
147 static __always_inline u32 subshift(u32 in[], int pos)
148 {
149 return (aes_sbox[in[pos] & 0xff]) ^
150 (aes_sbox[(in[(pos + 1) % 4] >> 8) & 0xff] << 8) ^
151 (aes_sbox[(in[(pos + 2) % 4] >> 16) & 0xff] << 16) ^
152 (aes_sbox[(in[(pos + 3) % 4] >> 24) & 0xff] << 24);
153 }
154
155 static __always_inline u32 inv_subshift(u32 in[], int pos)
156 {
157 return (aes_inv_sbox[in[pos] & 0xff]) ^
158 (aes_inv_sbox[(in[(pos + 3) % 4] >> 8) & 0xff] << 8) ^
159 (aes_inv_sbox[(in[(pos + 2) % 4] >> 16) & 0xff] << 16) ^
160 (aes_inv_sbox[(in[(pos + 1) % 4] >> 24) & 0xff] << 24);
161 }
162
163 static u32 subw(u32 in)
164 {
165 return (aes_sbox[in & 0xff]) ^
166 (aes_sbox[(in >> 8) & 0xff] << 8) ^
167 (aes_sbox[(in >> 16) & 0xff] << 16) ^
168 (aes_sbox[(in >> 24) & 0xff] << 24);
169 }
170
171 /**
172 * aes_expandkey - Expands the AES key as described in FIPS-197
173 * @ctx: The location where the computed key will be stored.
174 * @in_key: The supplied key.
175 * @key_len: The length of the supplied key.
176 *
177 * Returns 0 on success. The function fails only if an invalid key size (or
178 * pointer) is supplied.
179 * The expanded key size is 240 bytes (max of 14 rounds with a unique 16 bytes
180 * key schedule plus a 16 bytes key which is used before the first round).
181 * The decryption key is prepared for the "Equivalent Inverse Cipher" as
182 * described in FIPS-197. The first slot (16 bytes) of each key (enc or dec) is
183 * for the initial combination, the second slot for the first round and so on.
184 */
185 int aes_expandkey(struct crypto_aes_ctx *ctx, const u8 *in_key,
186 unsigned int key_len)
187 {
188 u32 kwords = key_len / sizeof(u32);
189 u32 rc, i, j;
190 int err;
191
192 err = aes_check_keylen(key_len);
193 if (err)
194 return err;
195
196 ctx->key_length = key_len;
197
198 for (i = 0; i < kwords; i++)
199 ctx->key_enc[i] = get_unaligned_le32(in_key + i * sizeof(u32));
200
201 for (i = 0, rc = 1; i < 10; i++, rc = mul_by_x(rc)) {
202 u32 *rki = ctx->key_enc + (i * kwords);
203 u32 *rko = rki + kwords;
204
205 rko[0] = ror32(subw(rki[kwords - 1]), 8) ^ rc ^ rki[0];
206 rko[1] = rko[0] ^ rki[1];
207 rko[2] = rko[1] ^ rki[2];
208 rko[3] = rko[2] ^ rki[3];
209
210 if (key_len == AES_KEYSIZE_192) {
211 if (i >= 7)
212 break;
213 rko[4] = rko[3] ^ rki[4];
214 rko[5] = rko[4] ^ rki[5];
215 } else if (key_len == AES_KEYSIZE_256) {
216 if (i >= 6)
217 break;
218 rko[4] = subw(rko[3]) ^ rki[4];
219 rko[5] = rko[4] ^ rki[5];
220 rko[6] = rko[5] ^ rki[6];
221 rko[7] = rko[6] ^ rki[7];
222 }
223 }
224
225 /*
226 * Generate the decryption keys for the Equivalent Inverse Cipher.
227 * This involves reversing the order of the round keys, and applying
228 * the Inverse Mix Columns transformation to all but the first and
229 * the last one.
230 */
231 ctx->key_dec[0] = ctx->key_enc[key_len + 24];
232 ctx->key_dec[1] = ctx->key_enc[key_len + 25];
233 ctx->key_dec[2] = ctx->key_enc[key_len + 26];
234 ctx->key_dec[3] = ctx->key_enc[key_len + 27];
235
236 for (i = 4, j = key_len + 20; j > 0; i += 4, j -= 4) {
237 ctx->key_dec[i] = inv_mix_columns(ctx->key_enc[j]);
238 ctx->key_dec[i + 1] = inv_mix_columns(ctx->key_enc[j + 1]);
239 ctx->key_dec[i + 2] = inv_mix_columns(ctx->key_enc[j + 2]);
240 ctx->key_dec[i + 3] = inv_mix_columns(ctx->key_enc[j + 3]);
241 }
242
243 ctx->key_dec[i] = ctx->key_enc[0];
244 ctx->key_dec[i + 1] = ctx->key_enc[1];
245 ctx->key_dec[i + 2] = ctx->key_enc[2];
246 ctx->key_dec[i + 3] = ctx->key_enc[3];
247
248 return 0;
249 }
250 EXPORT_SYMBOL(aes_expandkey);
251
252 /**
253 * aes_encrypt - Encrypt a single AES block
254 * @ctx: Context struct containing the key schedule
255 * @out: Buffer to store the ciphertext
256 * @in: Buffer containing the plaintext
257 */
258 void aes_encrypt(const struct crypto_aes_ctx *ctx, u8 *out, const u8 *in)
259 {
260 const u32 *rkp = ctx->key_enc + 4;
261 int rounds = 6 + ctx->key_length / 4;
262 u32 st0[4], st1[4];
263 int round;
264
265 st0[0] = ctx->key_enc[0] ^ get_unaligned_le32(in);
266 st0[1] = ctx->key_enc[1] ^ get_unaligned_le32(in + 4);
267 st0[2] = ctx->key_enc[2] ^ get_unaligned_le32(in + 8);
268 st0[3] = ctx->key_enc[3] ^ get_unaligned_le32(in + 12);
269
270 /*
271 * Force the compiler to emit data independent Sbox references,
272 * by xoring the input with Sbox values that are known to add up
273 * to zero. This pulls the entire Sbox into the D-cache before any
274 * data dependent lookups are done.
275 */
276 st0[0] ^= aes_sbox[ 0] ^ aes_sbox[ 64] ^ aes_sbox[134] ^ aes_sbox[195];
277 st0[1] ^= aes_sbox[16] ^ aes_sbox[ 82] ^ aes_sbox[158] ^ aes_sbox[221];
278 st0[2] ^= aes_sbox[32] ^ aes_sbox[ 96] ^ aes_sbox[160] ^ aes_sbox[234];
279 st0[3] ^= aes_sbox[48] ^ aes_sbox[112] ^ aes_sbox[186] ^ aes_sbox[241];
280
281 for (round = 0;; round += 2, rkp += 8) {
282 st1[0] = mix_columns(subshift(st0, 0)) ^ rkp[0];
283 st1[1] = mix_columns(subshift(st0, 1)) ^ rkp[1];
284 st1[2] = mix_columns(subshift(st0, 2)) ^ rkp[2];
285 st1[3] = mix_columns(subshift(st0, 3)) ^ rkp[3];
286
287 if (round == rounds - 2)
288 break;
289
290 st0[0] = mix_columns(subshift(st1, 0)) ^ rkp[4];
291 st0[1] = mix_columns(subshift(st1, 1)) ^ rkp[5];
292 st0[2] = mix_columns(subshift(st1, 2)) ^ rkp[6];
293 st0[3] = mix_columns(subshift(st1, 3)) ^ rkp[7];
294 }
295
296 put_unaligned_le32(subshift(st1, 0) ^ rkp[4], out);
297 put_unaligned_le32(subshift(st1, 1) ^ rkp[5], out + 4);
298 put_unaligned_le32(subshift(st1, 2) ^ rkp[6], out + 8);
299 put_unaligned_le32(subshift(st1, 3) ^ rkp[7], out + 12);
300 }
301 EXPORT_SYMBOL(aes_encrypt);
302
303 /**
304 * aes_decrypt - Decrypt a single AES block
305 * @ctx: Context struct containing the key schedule
306 * @out: Buffer to store the plaintext
307 * @in: Buffer containing the ciphertext
308 */
309 void aes_decrypt(const struct crypto_aes_ctx *ctx, u8 *out, const u8 *in)
310 {
311 const u32 *rkp = ctx->key_dec + 4;
312 int rounds = 6 + ctx->key_length / 4;
313 u32 st0[4], st1[4];
314 int round;
315
316 st0[0] = ctx->key_dec[0] ^ get_unaligned_le32(in);
317 st0[1] = ctx->key_dec[1] ^ get_unaligned_le32(in + 4);
318 st0[2] = ctx->key_dec[2] ^ get_unaligned_le32(in + 8);
319 st0[3] = ctx->key_dec[3] ^ get_unaligned_le32(in + 12);
320
321 /*
322 * Force the compiler to emit data independent Sbox references,
323 * by xoring the input with Sbox values that are known to add up
324 * to zero. This pulls the entire Sbox into the D-cache before any
325 * data dependent lookups are done.
326 */
327 st0[0] ^= aes_inv_sbox[ 0] ^ aes_inv_sbox[ 64] ^ aes_inv_sbox[129] ^ aes_inv_sbox[200];
328 st0[1] ^= aes_inv_sbox[16] ^ aes_inv_sbox[ 83] ^ aes_inv_sbox[150] ^ aes_inv_sbox[212];
329 st0[2] ^= aes_inv_sbox[32] ^ aes_inv_sbox[ 96] ^ aes_inv_sbox[160] ^ aes_inv_sbox[236];
330 st0[3] ^= aes_inv_sbox[48] ^ aes_inv_sbox[112] ^ aes_inv_sbox[187] ^ aes_inv_sbox[247];
331
332 for (round = 0;; round += 2, rkp += 8) {
333 st1[0] = inv_mix_columns(inv_subshift(st0, 0)) ^ rkp[0];
334 st1[1] = inv_mix_columns(inv_subshift(st0, 1)) ^ rkp[1];
335 st1[2] = inv_mix_columns(inv_subshift(st0, 2)) ^ rkp[2];
336 st1[3] = inv_mix_columns(inv_subshift(st0, 3)) ^ rkp[3];
337
338 if (round == rounds - 2)
339 break;
340
341 st0[0] = inv_mix_columns(inv_subshift(st1, 0)) ^ rkp[4];
342 st0[1] = inv_mix_columns(inv_subshift(st1, 1)) ^ rkp[5];
343 st0[2] = inv_mix_columns(inv_subshift(st1, 2)) ^ rkp[6];
344 st0[3] = inv_mix_columns(inv_subshift(st1, 3)) ^ rkp[7];
345 }
346
347 put_unaligned_le32(inv_subshift(st1, 0) ^ rkp[4], out);
348 put_unaligned_le32(inv_subshift(st1, 1) ^ rkp[5], out + 4);
349 put_unaligned_le32(inv_subshift(st1, 2) ^ rkp[6], out + 8);
350 put_unaligned_le32(inv_subshift(st1, 3) ^ rkp[7], out + 12);
351 }
352 EXPORT_SYMBOL(aes_decrypt);
353
354 MODULE_DESCRIPTION("Generic AES library");
355 MODULE_AUTHOR("Ard Biesheuvel <ard.biesheuvel@linaro.org>");
356 MODULE_LICENSE("GPL v2");
357
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.