1 /* $OpenBSD: rijndael.c,v 1.6 2000/12/09 13:48:31 markus Exp $ */
3 /* This is an independent implementation of the encryption algorithm: */
5 /* RIJNDAEL by Joan Daemen and Vincent Rijmen */
7 /* which is a candidate algorithm in the Advanced Encryption Standard */
8 /* programme of the US National Institute of Standards and Technology. */
10 /* Copyright in this implementation is held by Dr B R Gladman but I */
11 /* hereby give permission for its free direct or derivative use subject */
12 /* to acknowledgment of its origin and compliance with any conditions */
13 /* that the originators of the algorithm place on its exploitation. */
15 /* Dr Brian Gladman (gladman@seven77.demon.co.uk) 14th January 1999 */
17 /* Timing data for Rijndael (rijndael.c)
19 Algorithm: rijndael (rijndael.c)
22 Key Setup: 305/1389 cycles (encrypt/decrypt)
23 Encrypt: 374 cycles = 68.4 mbits/sec
24 Decrypt: 352 cycles = 72.7 mbits/sec
25 Mean: 363 cycles = 70.5 mbits/sec
28 Key Setup: 277/1595 cycles (encrypt/decrypt)
29 Encrypt: 439 cycles = 58.3 mbits/sec
30 Decrypt: 425 cycles = 60.2 mbits/sec
31 Mean: 432 cycles = 59.3 mbits/sec
34 Key Setup: 374/1960 cycles (encrypt/decrypt)
35 Encrypt: 502 cycles = 51.0 mbits/sec
36 Decrypt: 498 cycles = 51.4 mbits/sec
37 Mean: 500 cycles = 51.2 mbits/sec
44 void gen_tabs __P((void));
46 /* 3. Basic macros for speeding up generic operations */
48 /* Circular rotate of 32 bit values */
50 #define rotr(x,n) (((x) >> ((int)(n))) | ((x) << (32 - (int)(n))))
51 #define rotl(x,n) (((x) << ((int)(n))) | ((x) >> (32 - (int)(n))))
53 /* Invert byte order in a 32 bit variable */
55 #define bswap(x) ((rotl(x, 8) & 0x00ff00ff) | (rotr(x, 8) & 0xff00ff00))
57 /* Extract byte from a 32 bit quantity (little endian notation) */
59 #define byte(x,n) ((u1byte)((x) >> (8 * n)))
61 #if BYTE_ORDER != LITTLE_ENDIAN
66 #define io_swap(x) bswap(x)
68 #define io_swap(x) (x)
78 u4byte ft_tab[4][256];
79 u4byte it_tab[4][256];
82 u4byte fl_tab[4][256];
83 u4byte il_tab[4][256];
88 #define ff_mult(a,b) (a && b ? pow_tab[(log_tab[a] + log_tab[b]) % 255] : 0)
90 #define f_rn(bo, bi, n, k) \
91 bo[n] = ft_tab[0][byte(bi[n],0)] ^ \
92 ft_tab[1][byte(bi[(n + 1) & 3],1)] ^ \
93 ft_tab[2][byte(bi[(n + 2) & 3],2)] ^ \
94 ft_tab[3][byte(bi[(n + 3) & 3],3)] ^ *(k + n)
96 #define i_rn(bo, bi, n, k) \
97 bo[n] = it_tab[0][byte(bi[n],0)] ^ \
98 it_tab[1][byte(bi[(n + 3) & 3],1)] ^ \
99 it_tab[2][byte(bi[(n + 2) & 3],2)] ^ \
100 it_tab[3][byte(bi[(n + 1) & 3],3)] ^ *(k + n)
105 ( fl_tab[0][byte(x, 0)] ^ \
106 fl_tab[1][byte(x, 1)] ^ \
107 fl_tab[2][byte(x, 2)] ^ \
108 fl_tab[3][byte(x, 3)] )
110 #define f_rl(bo, bi, n, k) \
111 bo[n] = fl_tab[0][byte(bi[n],0)] ^ \
112 fl_tab[1][byte(bi[(n + 1) & 3],1)] ^ \
113 fl_tab[2][byte(bi[(n + 2) & 3],2)] ^ \
114 fl_tab[3][byte(bi[(n + 3) & 3],3)] ^ *(k + n)
116 #define i_rl(bo, bi, n, k) \
117 bo[n] = il_tab[0][byte(bi[n],0)] ^ \
118 il_tab[1][byte(bi[(n + 3) & 3],1)] ^ \
119 il_tab[2][byte(bi[(n + 2) & 3],2)] ^ \
120 il_tab[3][byte(bi[(n + 1) & 3],3)] ^ *(k + n)
125 ((u4byte)sbx_tab[byte(x, 0)] << 0) ^ \
126 ((u4byte)sbx_tab[byte(x, 1)] << 8) ^ \
127 ((u4byte)sbx_tab[byte(x, 2)] << 16) ^ \
128 ((u4byte)sbx_tab[byte(x, 3)] << 24)
130 #define f_rl(bo, bi, n, k) \
131 bo[n] = (u4byte)sbx_tab[byte(bi[n],0)] ^ \
132 rotl(((u4byte)sbx_tab[byte(bi[(n + 1) & 3],1)]), 8) ^ \
133 rotl(((u4byte)sbx_tab[byte(bi[(n + 2) & 3],2)]), 16) ^ \
134 rotl(((u4byte)sbx_tab[byte(bi[(n + 3) & 3],3)]), 24) ^ *(k + n)
136 #define i_rl(bo, bi, n, k) \
137 bo[n] = (u4byte)isb_tab[byte(bi[n],0)] ^ \
138 rotl(((u4byte)isb_tab[byte(bi[(n + 3) & 3],1)]), 8) ^ \
139 rotl(((u4byte)isb_tab[byte(bi[(n + 2) & 3],2)]), 16) ^ \
140 rotl(((u4byte)isb_tab[byte(bi[(n + 1) & 3],3)]), 24) ^ *(k + n)
150 /* log and power tables for GF(2**8) finite field with */
151 /* 0x11b as modular polynomial - the simplest prmitive */
152 /* root is 0x11, used here to generate the tables */
154 for(i = 0,p = 1; i < 256; ++i) {
155 pow_tab[i] = (u1byte)p; log_tab[p] = (u1byte)i;
157 p = p ^ (p << 1) ^ (p & 0x80 ? 0x01b : 0);
160 log_tab[1] = 0; p = 1;
162 for(i = 0; i < 10; ++i) {
165 p = (p << 1) ^ (p & 0x80 ? 0x1b : 0);
168 /* note that the affine byte transformation matrix in */
169 /* rijndael specification is in big endian format with */
170 /* bit 0 as the most significant bit. In the remainder */
171 /* of the specification the bits are numbered from the */
172 /* least significant end of a byte. */
174 for(i = 0; i < 256; ++i) {
175 p = (i ? pow_tab[255 - log_tab[i]] : 0); q = p;
176 q = (q >> 7) | (q << 1); p ^= q;
177 q = (q >> 7) | (q << 1); p ^= q;
178 q = (q >> 7) | (q << 1); p ^= q;
179 q = (q >> 7) | (q << 1); p ^= q ^ 0x63;
180 sbx_tab[i] = (u1byte)p; isb_tab[p] = (u1byte)i;
183 for(i = 0; i < 256; ++i) {
188 t = p; fl_tab[0][i] = t;
189 fl_tab[1][i] = rotl(t, 8);
190 fl_tab[2][i] = rotl(t, 16);
191 fl_tab[3][i] = rotl(t, 24);
193 t = ((u4byte)ff_mult(2, p)) |
196 ((u4byte)ff_mult(3, p) << 24);
199 ft_tab[1][i] = rotl(t, 8);
200 ft_tab[2][i] = rotl(t, 16);
201 ft_tab[3][i] = rotl(t, 24);
207 t = p; il_tab[0][i] = t;
208 il_tab[1][i] = rotl(t, 8);
209 il_tab[2][i] = rotl(t, 16);
210 il_tab[3][i] = rotl(t, 24);
212 t = ((u4byte)ff_mult(14, p)) |
213 ((u4byte)ff_mult( 9, p) << 8) |
214 ((u4byte)ff_mult(13, p) << 16) |
215 ((u4byte)ff_mult(11, p) << 24);
218 it_tab[1][i] = rotl(t, 8);
219 it_tab[2][i] = rotl(t, 16);
220 it_tab[3][i] = rotl(t, 24);
226 #define star_x(x) (((x) & 0x7f7f7f7f) << 1) ^ ((((x) & 0x80808080) >> 7) * 0x1b)
228 #define imix_col(y,x) \
234 (y) ^= rotr(u ^ t, 8) ^ \
238 /* initialise the key schedule from the user supplied key */
241 { t = ls_box(rotr(t, 8)) ^ rco_tab[i]; \
242 t ^= e_key[4 * i]; e_key[4 * i + 4] = t; \
243 t ^= e_key[4 * i + 1]; e_key[4 * i + 5] = t; \
244 t ^= e_key[4 * i + 2]; e_key[4 * i + 6] = t; \
245 t ^= e_key[4 * i + 3]; e_key[4 * i + 7] = t; \
249 { t = ls_box(rotr(t, 8)) ^ rco_tab[i]; \
250 t ^= e_key[6 * i]; e_key[6 * i + 6] = t; \
251 t ^= e_key[6 * i + 1]; e_key[6 * i + 7] = t; \
252 t ^= e_key[6 * i + 2]; e_key[6 * i + 8] = t; \
253 t ^= e_key[6 * i + 3]; e_key[6 * i + 9] = t; \
254 t ^= e_key[6 * i + 4]; e_key[6 * i + 10] = t; \
255 t ^= e_key[6 * i + 5]; e_key[6 * i + 11] = t; \
259 { t = ls_box(rotr(t, 8)) ^ rco_tab[i]; \
260 t ^= e_key[8 * i]; e_key[8 * i + 8] = t; \
261 t ^= e_key[8 * i + 1]; e_key[8 * i + 9] = t; \
262 t ^= e_key[8 * i + 2]; e_key[8 * i + 10] = t; \
263 t ^= e_key[8 * i + 3]; e_key[8 * i + 11] = t; \
264 t = e_key[8 * i + 4] ^ ls_box(t); \
265 e_key[8 * i + 12] = t; \
266 t ^= e_key[8 * i + 5]; e_key[8 * i + 13] = t; \
267 t ^= e_key[8 * i + 6]; e_key[8 * i + 14] = t; \
268 t ^= e_key[8 * i + 7]; e_key[8 * i + 15] = t; \
272 rijndael_set_key(rijndael_ctx *ctx, const u4byte *in_key, const u4byte key_len,
275 u4byte i, t, u, v, w;
276 u4byte *e_key = ctx->e_key;
277 u4byte *d_key = ctx->d_key;
279 ctx->decrypt = !encrypt;
284 ctx->k_len = (key_len + 31) / 32;
286 e_key[0] = io_swap(in_key[0]); e_key[1] = io_swap(in_key[1]);
287 e_key[2] = io_swap(in_key[2]); e_key[3] = io_swap(in_key[3]);
290 case 4: t = e_key[3];
291 for(i = 0; i < 10; ++i)
295 case 6: e_key[4] = io_swap(in_key[4]); t = e_key[5] = io_swap(in_key[5]);
296 for(i = 0; i < 8; ++i)
300 case 8: e_key[4] = io_swap(in_key[4]); e_key[5] = io_swap(in_key[5]);
301 e_key[6] = io_swap(in_key[6]); t = e_key[7] = io_swap(in_key[7]);
302 for(i = 0; i < 7; ++i)
308 d_key[0] = e_key[0]; d_key[1] = e_key[1];
309 d_key[2] = e_key[2]; d_key[3] = e_key[3];
311 for(i = 4; i < 4 * ctx->k_len + 24; ++i) {
312 imix_col(d_key[i], e_key[i]);
319 /* encrypt a block of text */
321 #define f_nround(bo, bi, k) \
322 f_rn(bo, bi, 0, k); \
323 f_rn(bo, bi, 1, k); \
324 f_rn(bo, bi, 2, k); \
325 f_rn(bo, bi, 3, k); \
328 #define f_lround(bo, bi, k) \
329 f_rl(bo, bi, 0, k); \
330 f_rl(bo, bi, 1, k); \
331 f_rl(bo, bi, 2, k); \
335 rijndael_encrypt(rijndael_ctx *ctx, const u4byte *in_blk, u4byte *out_blk)
337 u4byte k_len = ctx->k_len;
338 u4byte *e_key = ctx->e_key;
339 u4byte b0[4], b1[4], *kp;
341 b0[0] = io_swap(in_blk[0]) ^ e_key[0];
342 b0[1] = io_swap(in_blk[1]) ^ e_key[1];
343 b0[2] = io_swap(in_blk[2]) ^ e_key[2];
344 b0[3] = io_swap(in_blk[3]) ^ e_key[3];
349 f_nround(b1, b0, kp); f_nround(b0, b1, kp);
353 f_nround(b1, b0, kp); f_nround(b0, b1, kp);
356 f_nround(b1, b0, kp); f_nround(b0, b1, kp);
357 f_nround(b1, b0, kp); f_nround(b0, b1, kp);
358 f_nround(b1, b0, kp); f_nround(b0, b1, kp);
359 f_nround(b1, b0, kp); f_nround(b0, b1, kp);
360 f_nround(b1, b0, kp); f_lround(b0, b1, kp);
362 out_blk[0] = io_swap(b0[0]); out_blk[1] = io_swap(b0[1]);
363 out_blk[2] = io_swap(b0[2]); out_blk[3] = io_swap(b0[3]);
366 /* decrypt a block of text */
368 #define i_nround(bo, bi, k) \
369 i_rn(bo, bi, 0, k); \
370 i_rn(bo, bi, 1, k); \
371 i_rn(bo, bi, 2, k); \
372 i_rn(bo, bi, 3, k); \
375 #define i_lround(bo, bi, k) \
376 i_rl(bo, bi, 0, k); \
377 i_rl(bo, bi, 1, k); \
378 i_rl(bo, bi, 2, k); \
382 rijndael_decrypt(rijndael_ctx *ctx, const u4byte *in_blk, u4byte *out_blk)
384 u4byte b0[4], b1[4], *kp;
385 u4byte k_len = ctx->k_len;
386 u4byte *e_key = ctx->e_key;
387 u4byte *d_key = ctx->d_key;
389 b0[0] = io_swap(in_blk[0]) ^ e_key[4 * k_len + 24];
390 b0[1] = io_swap(in_blk[1]) ^ e_key[4 * k_len + 25];
391 b0[2] = io_swap(in_blk[2]) ^ e_key[4 * k_len + 26];
392 b0[3] = io_swap(in_blk[3]) ^ e_key[4 * k_len + 27];
394 kp = d_key + 4 * (k_len + 5);
397 i_nround(b1, b0, kp); i_nround(b0, b1, kp);
401 i_nround(b1, b0, kp); i_nround(b0, b1, kp);
404 i_nround(b1, b0, kp); i_nround(b0, b1, kp);
405 i_nround(b1, b0, kp); i_nround(b0, b1, kp);
406 i_nround(b1, b0, kp); i_nround(b0, b1, kp);
407 i_nround(b1, b0, kp); i_nround(b0, b1, kp);
408 i_nround(b1, b0, kp); i_lround(b0, b1, kp);
410 out_blk[0] = io_swap(b0[0]); out_blk[1] = io_swap(b0[1]);
411 out_blk[2] = io_swap(b0[2]); out_blk[3] = io_swap(b0[3]);