aes.c 13 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378
  1. // SPDX-License-Identifier: GPL-2.0
  2. /*
  3. * Copyright (C) 2017-2019 Linaro Ltd <ard.biesheuvel@linaro.org>
  4. */
  5. #include <crypto/aes.h>
  6. #include <linux/crypto.h>
  7. #include <linux/module.h>
  8. #include <asm/unaligned.h>
  9. #include <trace/hooks/fips140.h>
  10. /*
  11. * Emit the sbox as volatile const to prevent the compiler from doing
  12. * constant folding on sbox references involving fixed indexes.
  13. */
  14. static volatile const u8 __cacheline_aligned aes_sbox[] = {
  15. 0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5,
  16. 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76,
  17. 0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0,
  18. 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0,
  19. 0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc,
  20. 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15,
  21. 0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a,
  22. 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75,
  23. 0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0,
  24. 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84,
  25. 0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b,
  26. 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf,
  27. 0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85,
  28. 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8,
  29. 0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5,
  30. 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2,
  31. 0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17,
  32. 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73,
  33. 0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88,
  34. 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb,
  35. 0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c,
  36. 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79,
  37. 0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9,
  38. 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08,
  39. 0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6,
  40. 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a,
  41. 0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e,
  42. 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e,
  43. 0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94,
  44. 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf,
  45. 0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68,
  46. 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16,
  47. };
  48. static volatile const u8 __cacheline_aligned aes_inv_sbox[] = {
  49. 0x52, 0x09, 0x6a, 0xd5, 0x30, 0x36, 0xa5, 0x38,
  50. 0xbf, 0x40, 0xa3, 0x9e, 0x81, 0xf3, 0xd7, 0xfb,
  51. 0x7c, 0xe3, 0x39, 0x82, 0x9b, 0x2f, 0xff, 0x87,
  52. 0x34, 0x8e, 0x43, 0x44, 0xc4, 0xde, 0xe9, 0xcb,
  53. 0x54, 0x7b, 0x94, 0x32, 0xa6, 0xc2, 0x23, 0x3d,
  54. 0xee, 0x4c, 0x95, 0x0b, 0x42, 0xfa, 0xc3, 0x4e,
  55. 0x08, 0x2e, 0xa1, 0x66, 0x28, 0xd9, 0x24, 0xb2,
  56. 0x76, 0x5b, 0xa2, 0x49, 0x6d, 0x8b, 0xd1, 0x25,
  57. 0x72, 0xf8, 0xf6, 0x64, 0x86, 0x68, 0x98, 0x16,
  58. 0xd4, 0xa4, 0x5c, 0xcc, 0x5d, 0x65, 0xb6, 0x92,
  59. 0x6c, 0x70, 0x48, 0x50, 0xfd, 0xed, 0xb9, 0xda,
  60. 0x5e, 0x15, 0x46, 0x57, 0xa7, 0x8d, 0x9d, 0x84,
  61. 0x90, 0xd8, 0xab, 0x00, 0x8c, 0xbc, 0xd3, 0x0a,
  62. 0xf7, 0xe4, 0x58, 0x05, 0xb8, 0xb3, 0x45, 0x06,
  63. 0xd0, 0x2c, 0x1e, 0x8f, 0xca, 0x3f, 0x0f, 0x02,
  64. 0xc1, 0xaf, 0xbd, 0x03, 0x01, 0x13, 0x8a, 0x6b,
  65. 0x3a, 0x91, 0x11, 0x41, 0x4f, 0x67, 0xdc, 0xea,
  66. 0x97, 0xf2, 0xcf, 0xce, 0xf0, 0xb4, 0xe6, 0x73,
  67. 0x96, 0xac, 0x74, 0x22, 0xe7, 0xad, 0x35, 0x85,
  68. 0xe2, 0xf9, 0x37, 0xe8, 0x1c, 0x75, 0xdf, 0x6e,
  69. 0x47, 0xf1, 0x1a, 0x71, 0x1d, 0x29, 0xc5, 0x89,
  70. 0x6f, 0xb7, 0x62, 0x0e, 0xaa, 0x18, 0xbe, 0x1b,
  71. 0xfc, 0x56, 0x3e, 0x4b, 0xc6, 0xd2, 0x79, 0x20,
  72. 0x9a, 0xdb, 0xc0, 0xfe, 0x78, 0xcd, 0x5a, 0xf4,
  73. 0x1f, 0xdd, 0xa8, 0x33, 0x88, 0x07, 0xc7, 0x31,
  74. 0xb1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xec, 0x5f,
  75. 0x60, 0x51, 0x7f, 0xa9, 0x19, 0xb5, 0x4a, 0x0d,
  76. 0x2d, 0xe5, 0x7a, 0x9f, 0x93, 0xc9, 0x9c, 0xef,
  77. 0xa0, 0xe0, 0x3b, 0x4d, 0xae, 0x2a, 0xf5, 0xb0,
  78. 0xc8, 0xeb, 0xbb, 0x3c, 0x83, 0x53, 0x99, 0x61,
  79. 0x17, 0x2b, 0x04, 0x7e, 0xba, 0x77, 0xd6, 0x26,
  80. 0xe1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0c, 0x7d,
  81. };
  82. extern const u8 crypto_aes_sbox[256] __alias(aes_sbox);
  83. extern const u8 crypto_aes_inv_sbox[256] __alias(aes_inv_sbox);
  84. EXPORT_SYMBOL(crypto_aes_sbox);
  85. EXPORT_SYMBOL(crypto_aes_inv_sbox);
  86. static u32 mul_by_x(u32 w)
  87. {
  88. u32 x = w & 0x7f7f7f7f;
  89. u32 y = w & 0x80808080;
  90. /* multiply by polynomial 'x' (0b10) in GF(2^8) */
  91. return (x << 1) ^ (y >> 7) * 0x1b;
  92. }
  93. static u32 mul_by_x2(u32 w)
  94. {
  95. u32 x = w & 0x3f3f3f3f;
  96. u32 y = w & 0x80808080;
  97. u32 z = w & 0x40404040;
  98. /* multiply by polynomial 'x^2' (0b100) in GF(2^8) */
  99. return (x << 2) ^ (y >> 7) * 0x36 ^ (z >> 6) * 0x1b;
  100. }
  101. static u32 mix_columns(u32 x)
  102. {
  103. /*
  104. * Perform the following matrix multiplication in GF(2^8)
  105. *
  106. * | 0x2 0x3 0x1 0x1 | | x[0] |
  107. * | 0x1 0x2 0x3 0x1 | | x[1] |
  108. * | 0x1 0x1 0x2 0x3 | x | x[2] |
  109. * | 0x3 0x1 0x1 0x2 | | x[3] |
  110. */
  111. u32 y = mul_by_x(x) ^ ror32(x, 16);
  112. return y ^ ror32(x ^ y, 8);
  113. }
  114. static u32 inv_mix_columns(u32 x)
  115. {
  116. /*
  117. * Perform the following matrix multiplication in GF(2^8)
  118. *
  119. * | 0xe 0xb 0xd 0x9 | | x[0] |
  120. * | 0x9 0xe 0xb 0xd | | x[1] |
  121. * | 0xd 0x9 0xe 0xb | x | x[2] |
  122. * | 0xb 0xd 0x9 0xe | | x[3] |
  123. *
  124. * which can conveniently be reduced to
  125. *
  126. * | 0x2 0x3 0x1 0x1 | | 0x5 0x0 0x4 0x0 | | x[0] |
  127. * | 0x1 0x2 0x3 0x1 | | 0x0 0x5 0x0 0x4 | | x[1] |
  128. * | 0x1 0x1 0x2 0x3 | x | 0x4 0x0 0x5 0x0 | x | x[2] |
  129. * | 0x3 0x1 0x1 0x2 | | 0x0 0x4 0x0 0x5 | | x[3] |
  130. */
  131. u32 y = mul_by_x2(x);
  132. return mix_columns(x ^ y ^ ror32(y, 16));
  133. }
  134. static __always_inline u32 subshift(u32 in[], int pos)
  135. {
  136. return (aes_sbox[in[pos] & 0xff]) ^
  137. (aes_sbox[(in[(pos + 1) % 4] >> 8) & 0xff] << 8) ^
  138. (aes_sbox[(in[(pos + 2) % 4] >> 16) & 0xff] << 16) ^
  139. (aes_sbox[(in[(pos + 3) % 4] >> 24) & 0xff] << 24);
  140. }
  141. static __always_inline u32 inv_subshift(u32 in[], int pos)
  142. {
  143. return (aes_inv_sbox[in[pos] & 0xff]) ^
  144. (aes_inv_sbox[(in[(pos + 3) % 4] >> 8) & 0xff] << 8) ^
  145. (aes_inv_sbox[(in[(pos + 2) % 4] >> 16) & 0xff] << 16) ^
  146. (aes_inv_sbox[(in[(pos + 1) % 4] >> 24) & 0xff] << 24);
  147. }
  148. static u32 subw(u32 in)
  149. {
  150. return (aes_sbox[in & 0xff]) ^
  151. (aes_sbox[(in >> 8) & 0xff] << 8) ^
  152. (aes_sbox[(in >> 16) & 0xff] << 16) ^
  153. (aes_sbox[(in >> 24) & 0xff] << 24);
  154. }
  155. /**
  156. * aes_expandkey - Expands the AES key as described in FIPS-197
  157. * @ctx: The location where the computed key will be stored.
  158. * @in_key: The supplied key.
  159. * @key_len: The length of the supplied key.
  160. *
  161. * Returns 0 on success. The function fails only if an invalid key size (or
  162. * pointer) is supplied.
  163. * The expanded key size is 240 bytes (max of 14 rounds with a unique 16 bytes
  164. * key schedule plus a 16 bytes key which is used before the first round).
  165. * The decryption key is prepared for the "Equivalent Inverse Cipher" as
  166. * described in FIPS-197. The first slot (16 bytes) of each key (enc or dec) is
  167. * for the initial combination, the second slot for the first round and so on.
  168. */
  169. int aes_expandkey(struct crypto_aes_ctx *ctx, const u8 *in_key,
  170. unsigned int key_len)
  171. {
  172. u32 kwords = key_len / sizeof(u32);
  173. u32 rc, i, j;
  174. int err;
  175. #if defined(CONFIG_CRYPTO_FIPS140) && !defined(BUILD_FIPS140_KO)
  176. err = -(MAX_ERRNO + 1);
  177. trace_android_vh_aes_expandkey(ctx, in_key, key_len, &err);
  178. if (err != -(MAX_ERRNO + 1))
  179. return err;
  180. #endif
  181. err = aes_check_keylen(key_len);
  182. if (err)
  183. return err;
  184. ctx->key_length = key_len;
  185. for (i = 0; i < kwords; i++)
  186. ctx->key_enc[i] = get_unaligned_le32(in_key + i * sizeof(u32));
  187. for (i = 0, rc = 1; i < 10; i++, rc = mul_by_x(rc)) {
  188. u32 *rki = ctx->key_enc + (i * kwords);
  189. u32 *rko = rki + kwords;
  190. rko[0] = ror32(subw(rki[kwords - 1]), 8) ^ rc ^ rki[0];
  191. rko[1] = rko[0] ^ rki[1];
  192. rko[2] = rko[1] ^ rki[2];
  193. rko[3] = rko[2] ^ rki[3];
  194. if (key_len == AES_KEYSIZE_192) {
  195. if (i >= 7)
  196. break;
  197. rko[4] = rko[3] ^ rki[4];
  198. rko[5] = rko[4] ^ rki[5];
  199. } else if (key_len == AES_KEYSIZE_256) {
  200. if (i >= 6)
  201. break;
  202. rko[4] = subw(rko[3]) ^ rki[4];
  203. rko[5] = rko[4] ^ rki[5];
  204. rko[6] = rko[5] ^ rki[6];
  205. rko[7] = rko[6] ^ rki[7];
  206. }
  207. }
  208. /*
  209. * Generate the decryption keys for the Equivalent Inverse Cipher.
  210. * This involves reversing the order of the round keys, and applying
  211. * the Inverse Mix Columns transformation to all but the first and
  212. * the last one.
  213. */
  214. ctx->key_dec[0] = ctx->key_enc[key_len + 24];
  215. ctx->key_dec[1] = ctx->key_enc[key_len + 25];
  216. ctx->key_dec[2] = ctx->key_enc[key_len + 26];
  217. ctx->key_dec[3] = ctx->key_enc[key_len + 27];
  218. for (i = 4, j = key_len + 20; j > 0; i += 4, j -= 4) {
  219. ctx->key_dec[i] = inv_mix_columns(ctx->key_enc[j]);
  220. ctx->key_dec[i + 1] = inv_mix_columns(ctx->key_enc[j + 1]);
  221. ctx->key_dec[i + 2] = inv_mix_columns(ctx->key_enc[j + 2]);
  222. ctx->key_dec[i + 3] = inv_mix_columns(ctx->key_enc[j + 3]);
  223. }
  224. ctx->key_dec[i] = ctx->key_enc[0];
  225. ctx->key_dec[i + 1] = ctx->key_enc[1];
  226. ctx->key_dec[i + 2] = ctx->key_enc[2];
  227. ctx->key_dec[i + 3] = ctx->key_enc[3];
  228. return 0;
  229. }
  230. EXPORT_SYMBOL(aes_expandkey);
  231. /**
  232. * aes_encrypt - Encrypt a single AES block
  233. * @ctx: Context struct containing the key schedule
  234. * @out: Buffer to store the ciphertext
  235. * @in: Buffer containing the plaintext
  236. */
  237. void aes_encrypt(const struct crypto_aes_ctx *ctx, u8 *out, const u8 *in)
  238. {
  239. const u32 *rkp = ctx->key_enc + 4;
  240. int rounds = 6 + ctx->key_length / 4;
  241. u32 st0[4], st1[4];
  242. int round;
  243. #if defined(CONFIG_CRYPTO_FIPS140) && !defined(BUILD_FIPS140_KO)
  244. int hook_inuse = 0;
  245. trace_android_vh_aes_encrypt(ctx, out, in, &hook_inuse);
  246. if (hook_inuse)
  247. return;
  248. #endif
  249. st0[0] = ctx->key_enc[0] ^ get_unaligned_le32(in);
  250. st0[1] = ctx->key_enc[1] ^ get_unaligned_le32(in + 4);
  251. st0[2] = ctx->key_enc[2] ^ get_unaligned_le32(in + 8);
  252. st0[3] = ctx->key_enc[3] ^ get_unaligned_le32(in + 12);
  253. /*
  254. * Force the compiler to emit data independent Sbox references,
  255. * by xoring the input with Sbox values that are known to add up
  256. * to zero. This pulls the entire Sbox into the D-cache before any
  257. * data dependent lookups are done.
  258. */
  259. st0[0] ^= aes_sbox[ 0] ^ aes_sbox[ 64] ^ aes_sbox[134] ^ aes_sbox[195];
  260. st0[1] ^= aes_sbox[16] ^ aes_sbox[ 82] ^ aes_sbox[158] ^ aes_sbox[221];
  261. st0[2] ^= aes_sbox[32] ^ aes_sbox[ 96] ^ aes_sbox[160] ^ aes_sbox[234];
  262. st0[3] ^= aes_sbox[48] ^ aes_sbox[112] ^ aes_sbox[186] ^ aes_sbox[241];
  263. for (round = 0;; round += 2, rkp += 8) {
  264. st1[0] = mix_columns(subshift(st0, 0)) ^ rkp[0];
  265. st1[1] = mix_columns(subshift(st0, 1)) ^ rkp[1];
  266. st1[2] = mix_columns(subshift(st0, 2)) ^ rkp[2];
  267. st1[3] = mix_columns(subshift(st0, 3)) ^ rkp[3];
  268. if (round == rounds - 2)
  269. break;
  270. st0[0] = mix_columns(subshift(st1, 0)) ^ rkp[4];
  271. st0[1] = mix_columns(subshift(st1, 1)) ^ rkp[5];
  272. st0[2] = mix_columns(subshift(st1, 2)) ^ rkp[6];
  273. st0[3] = mix_columns(subshift(st1, 3)) ^ rkp[7];
  274. }
  275. put_unaligned_le32(subshift(st1, 0) ^ rkp[4], out);
  276. put_unaligned_le32(subshift(st1, 1) ^ rkp[5], out + 4);
  277. put_unaligned_le32(subshift(st1, 2) ^ rkp[6], out + 8);
  278. put_unaligned_le32(subshift(st1, 3) ^ rkp[7], out + 12);
  279. }
  280. EXPORT_SYMBOL(aes_encrypt);
  281. /**
  282. * aes_decrypt - Decrypt a single AES block
  283. * @ctx: Context struct containing the key schedule
  284. * @out: Buffer to store the plaintext
  285. * @in: Buffer containing the ciphertext
  286. */
  287. void aes_decrypt(const struct crypto_aes_ctx *ctx, u8 *out, const u8 *in)
  288. {
  289. const u32 *rkp = ctx->key_dec + 4;
  290. int rounds = 6 + ctx->key_length / 4;
  291. u32 st0[4], st1[4];
  292. int round;
  293. #if defined(CONFIG_CRYPTO_FIPS140) && !defined(BUILD_FIPS140_KO)
  294. int hook_inuse = 0;
  295. trace_android_vh_aes_decrypt(ctx, out, in, &hook_inuse);
  296. if (hook_inuse)
  297. return;
  298. #endif
  299. st0[0] = ctx->key_dec[0] ^ get_unaligned_le32(in);
  300. st0[1] = ctx->key_dec[1] ^ get_unaligned_le32(in + 4);
  301. st0[2] = ctx->key_dec[2] ^ get_unaligned_le32(in + 8);
  302. st0[3] = ctx->key_dec[3] ^ get_unaligned_le32(in + 12);
  303. /*
  304. * Force the compiler to emit data independent Sbox references,
  305. * by xoring the input with Sbox values that are known to add up
  306. * to zero. This pulls the entire Sbox into the D-cache before any
  307. * data dependent lookups are done.
  308. */
  309. st0[0] ^= aes_inv_sbox[ 0] ^ aes_inv_sbox[ 64] ^ aes_inv_sbox[129] ^ aes_inv_sbox[200];
  310. st0[1] ^= aes_inv_sbox[16] ^ aes_inv_sbox[ 83] ^ aes_inv_sbox[150] ^ aes_inv_sbox[212];
  311. st0[2] ^= aes_inv_sbox[32] ^ aes_inv_sbox[ 96] ^ aes_inv_sbox[160] ^ aes_inv_sbox[236];
  312. st0[3] ^= aes_inv_sbox[48] ^ aes_inv_sbox[112] ^ aes_inv_sbox[187] ^ aes_inv_sbox[247];
  313. for (round = 0;; round += 2, rkp += 8) {
  314. st1[0] = inv_mix_columns(inv_subshift(st0, 0)) ^ rkp[0];
  315. st1[1] = inv_mix_columns(inv_subshift(st0, 1)) ^ rkp[1];
  316. st1[2] = inv_mix_columns(inv_subshift(st0, 2)) ^ rkp[2];
  317. st1[3] = inv_mix_columns(inv_subshift(st0, 3)) ^ rkp[3];
  318. if (round == rounds - 2)
  319. break;
  320. st0[0] = inv_mix_columns(inv_subshift(st1, 0)) ^ rkp[4];
  321. st0[1] = inv_mix_columns(inv_subshift(st1, 1)) ^ rkp[5];
  322. st0[2] = inv_mix_columns(inv_subshift(st1, 2)) ^ rkp[6];
  323. st0[3] = inv_mix_columns(inv_subshift(st1, 3)) ^ rkp[7];
  324. }
  325. put_unaligned_le32(inv_subshift(st1, 0) ^ rkp[4], out);
  326. put_unaligned_le32(inv_subshift(st1, 1) ^ rkp[5], out + 4);
  327. put_unaligned_le32(inv_subshift(st1, 2) ^ rkp[6], out + 8);
  328. put_unaligned_le32(inv_subshift(st1, 3) ^ rkp[7], out + 12);
  329. }
  330. EXPORT_SYMBOL(aes_decrypt);
  331. MODULE_DESCRIPTION("Generic AES library");
  332. MODULE_AUTHOR("Ard Biesheuvel <ard.biesheuvel@linaro.org>");
  333. MODULE_LICENSE("GPL v2");