7 * \brief Implementation of the CAST5 (aka CAST-128) cipher algorithm as described in RFC 2144
23 #include "cast5-sbox.h"
27 #define S5(x) pgm_read_dword(&s5[(x)])
28 #define S6(x) pgm_read_dword(&s6[(x)])
29 #define S7(x) pgm_read_dword(&s7[(x)])
30 #define S8(x) pgm_read_dword(&s8[(x)])
33 void cast5_init_A(uint8_t *dest, uint8_t *src, bool bmode){
34 uint8_t mask = bmode?0x8:0;
35 *((uint32_t*)(&dest[0x0])) = *((uint32_t*)(&src[0x0^mask])) ^ S5(src[0xD^mask]) ^ S6(src[0xF^mask]) ^ S7(src[0xC^mask]) ^ S8(src[0xE^mask]) ^ S7(src[0x8^mask]);
36 *((uint32_t*)(&dest[0x4])) = *((uint32_t*)(&src[0x8^mask])) ^ S5(dest[0x0]) ^ S6(dest[0x2]) ^ S7(dest[0x1]) ^ S8(dest[0x3]) ^ S8(src[0xA^mask]);
37 *((uint32_t*)(&dest[0x8])) = *((uint32_t*)(&src[0xC^mask])) ^ S5(dest[0x7]) ^ S6(dest[0x6]) ^ S7(dest[0x5]) ^ S8(dest[0x4]) ^ S5(src[0x9^mask]);
38 *((uint32_t*)(&dest[0xC])) = *((uint32_t*)(&src[0x4^mask])) ^ S5(dest[0xA]) ^ S6(dest[0x9]) ^ S7(dest[0xB]) ^ S8(dest[0x8]) ^ S6(src[0xB^mask]);
41 void cast5_init_M(uint8_t *dest, uint8_t *src, bool nmode, bool xmode){
42 uint8_t nmt[] = {0xB, 0xA, 0x9, 0x8, 0xF, 0xE, 0xD, 0xC, 0x3, 0x2, 0x1, 0x0, 0x7, 0x6, 0x5, 0x4}; /* nmode table */
43 uint8_t xmt[4][4] = {{0x2, 0x6, 0x9, 0xC}, {0x8, 0xD, 0x3, 0x7}, {0x3, 0x7, 0x8, 0xD}, {0x9, 0xC, 0x2, 0x6}};
44 #define NMT(x) (src[nmode?nmt[(x)]:(x)])
45 #define XMT(x) (src[xmt[(xmode<<1) + nmode][(x)]])
46 *((uint32_t*)(&dest[0x0])) = S5(NMT(0x8)) ^ S6(NMT(0x9)) ^ S7(NMT(0x7)) ^ S8(NMT(0x6)) ^ S5(XMT(0));
47 *((uint32_t*)(&dest[0x4])) = S5(NMT(0xA)) ^ S6(NMT(0xB)) ^ S7(NMT(0x5)) ^ S8(NMT(0x4)) ^ S6(XMT(1));
48 *((uint32_t*)(&dest[0x8])) = S5(NMT(0xC)) ^ S6(NMT(0xD)) ^ S7(NMT(0x3)) ^ S8(NMT(0x2)) ^ S7(XMT(2));
49 *((uint32_t*)(&dest[0xC])) = S5(NMT(0xE)) ^ S6(NMT(0xF)) ^ S7(NMT(0x1)) ^ S8(NMT(0x0)) ^ S8(XMT(3));
52 #define S5B(x) pgm_read_byte(3+(uint8_t*)(&s5[(x)]))
53 #define S6B(x) pgm_read_byte(3+(uint8_t*)(&s6[(x)]))
54 #define S7B(x) pgm_read_byte(3+(uint8_t*)(&s7[(x)]))
55 #define S8B(x) pgm_read_byte(3+(uint8_t*)(&s8[(x)]))
57 void cast5_init_rM(uint8_t *klo, uint8_t *khi, uint8_t offset, uint8_t *src, bool nmode, bool xmode){
58 uint8_t nmt[] = {0xB, 0xA, 0x9, 0x8, 0xF, 0xE, 0xD, 0xC, 0x3, 0x2, 0x1, 0x0, 0x7, 0x6, 0x5, 0x4}; /* nmode table */
59 uint8_t xmt[4][4] = {{0x2, 0x6, 0x9, 0xC}, {0x8, 0xD, 0x3, 0x7}, {0x3, 0x7, 0x8, 0xD}, {0x9, 0xC, 0x2, 0x6}};
61 t = S5B(NMT(0x8)) ^ S6B(NMT(0x9)) ^ S7B(NMT(0x7)) ^ S8B(NMT(0x6)) ^ S5B(XMT(0));
62 klo[offset*2] |= (t & 0x0f);
64 t = S5B(NMT(0xA)) ^ S6B(NMT(0xB)) ^ S7B(NMT(0x5)) ^ S8B(NMT(0x4)) ^ S6B(XMT(1));
65 klo[offset*2] |= (t<<4) & 0xf0;
67 t = S5B(NMT(0xC)) ^ S6B(NMT(0xD)) ^ S7B(NMT(0x3)) ^ S8B(NMT(0x2)) ^ S7B(XMT(2));
68 klo[offset*2+1] |= t&0xf;
70 t = S5B(NMT(0xE)) ^ S6B(NMT(0xF)) ^ S7B(NMT(0x1)) ^ S8B(NMT(0x0)) ^ S8B(XMT(3));
71 klo[offset*2+1] |= t<<4;
74 uart_putstr("\r\n\t h="); uart_hexdump(&h,1);
76 khi[offset>>1] |= h<<((offset&0x1)?4:0);
79 #define S_5X(s) pgm_read_dword(&s5[BPX[(s)]])
80 #define S_6X(s) pgm_read_dword(&s6[BPX[(s)]])
81 #define S_7X(s) pgm_read_dword(&s7[BPX[(s)]])
82 #define S_8X(s) pgm_read_dword(&s8[BPX[(s)]])
84 #define S_5Z(s) pgm_read_dword(&s5[BPZ[(s)]])
85 #define S_6Z(s) pgm_read_dword(&s6[BPZ[(s)]])
86 #define S_7Z(s) pgm_read_dword(&s7[BPZ[(s)]])
87 #define S_8Z(s) pgm_read_dword(&s8[BPZ[(s)]])
92 * \brief sets up round keys (context) for cast5 en/decryption.
93 * @param s Pointer to cast5 context.
94 * @param key Pointer to binary key.
95 * @param keylength length of keydata in bits.
97 void cast5_init(cast5_ctx_t* s, uint8_t* key, uint8_t keylength){
98 /* we migth return if the key is valid and if setup was sucessfull */
100 #define BPX ((uint8_t*)&(x[0]))
101 #define BPZ ((uint8_t*)&(z[0]))
102 s->shortkey = (keylength<=80);
103 /* littel endian only! */
104 memset(&(x[0]), 0 ,16); /* set x to zero */
107 memcpy(&(x[0]), key, (keylength+7)/8);
110 /* todo: merge a and b and compress the whole stuff */
112 cast5_init_A((uint8_t*)(&z[0]), (uint8_t*)(&x[0]), false);
114 cast5_init_M((uint8_t*)(&(s->mask[0])), (uint8_t*)(&z[0]), false, false);
116 cast5_init_A((uint8_t*)(&x[0]), (uint8_t*)(&z[0]), true);
118 cast5_init_M((uint8_t*)(&(s->mask[4])), (uint8_t*)(&x[0]), true, false);
120 cast5_init_A((uint8_t*)(&z[0]), (uint8_t*)(&x[0]), false);
122 cast5_init_M((uint8_t*)(&(s->mask[8])), (uint8_t*)(&z[0]), true, true);
124 cast5_init_A((uint8_t*)(&x[0]), (uint8_t*)(&z[0]), true);
126 cast5_init_M((uint8_t*)(&(s->mask[12])), (uint8_t*)(&x[0]), false, true);
128 /* that were the masking keys, now the rotation keys */
129 /* set the keys to zero */
130 memset(&(s->rotl[0]),0,8);
131 s->roth[0]=s->roth[1]=0;
133 cast5_init_A((uint8_t*)(&z[0]), (uint8_t*)(&x[0]), false);
135 cast5_init_rM(&(s->rotl[0]), &(s->roth[0]), 0, (uint8_t*)(&z[0]), false, false);
137 cast5_init_A((uint8_t*)(&x[0]), (uint8_t*)(&z[0]), true);
139 cast5_init_rM(&(s->rotl[0]), &(s->roth[0]), 1, (uint8_t*)(&x[0]), true, false);
141 cast5_init_A((uint8_t*)(&z[0]), (uint8_t*)(&x[0]), false);
143 cast5_init_rM(&(s->rotl[0]), &(s->roth[0]), 2, (uint8_t*)(&z[0]), true, true);
145 cast5_init_A((uint8_t*)(&x[0]), (uint8_t*)(&z[0]), true);
147 cast5_init_rM(&(s->rotl[0]), &(s->roth[0]), 3, (uint8_t*)(&x[0]), false, true);
153 /********************************************************************************************************/
155 #define ROTL32(a,n) ((a)<<(n) | (a)>>(32-(n)))
156 #define CHANGE_ENDIAN32(x) ((x)<<24 | (x)>>24 | ((x)&0xff00)<<8 | ((x)&0xff0000)>>8 )
158 typedef uint32_t cast5_f_t(uint32_t,uint32_t,uint8_t);
166 uint32_t cast5_f1(uint32_t d, uint32_t m, uint8_t r){
168 t = ROTL32((d + m),r);
170 uint32_t ia,ib,ic,id;
171 uart_putstr("\r\n f1("); uart_hexdump(&d, 4); uart_putc(',');
172 uart_hexdump(&m , 4); uart_putc(','); uart_hexdump(&r, 1);uart_putstr("): I=");
174 ia = pgm_read_dword(&s1[((uint8_t*)&t)[IA]] );
175 ib = pgm_read_dword(&s2[((uint8_t*)&t)[IB]] );
176 ic = pgm_read_dword(&s3[((uint8_t*)&t)[IC]] );
177 id = pgm_read_dword(&s4[((uint8_t*)&t)[ID]] );
178 uart_putstr("\r\n\tIA="); uart_hexdump(&ia, 4);
179 uart_putstr("\r\n\tIB="); uart_hexdump(&ib, 4);
180 uart_putstr("\r\n\tIC="); uart_hexdump(&ic, 4);
181 uart_putstr("\r\n\tID="); uart_hexdump(&id, 4);
183 return (((ia ^ ib) - ic) + id);
187 return (((pgm_read_dword(&s1[((uint8_t*)&t)[IA]] ) ^ pgm_read_dword(&s2[((uint8_t*)&t)[IB]] ))
188 - pgm_read_dword(&s3[((uint8_t*)&t)[IC]] )) + pgm_read_dword(&s4[((uint8_t*)&t)[ID]]));
194 uint32_t cast5_f2(uint32_t d, uint32_t m, uint8_t r){
196 t = ROTL32((d ^ m),r);
198 uint32_t ia,ib,ic,id;
199 uart_putstr("\r\n f2("); uart_hexdump(&d, 4); uart_putc(',');
200 uart_hexdump(&m , 4); uart_putc(','); uart_hexdump(&r, 1);uart_putstr("): I=");
203 ia = pgm_read_dword(&s1[((uint8_t*)&t)[IA]] );
204 ib = pgm_read_dword(&s2[((uint8_t*)&t)[IB]] );
205 ic = pgm_read_dword(&s3[((uint8_t*)&t)[IC]] );
206 id = pgm_read_dword(&s4[((uint8_t*)&t)[ID]] );
208 uart_putstr("\r\n\tIA="); uart_hexdump(&ia, 4);
209 uart_putstr("\r\n\tIB="); uart_hexdump(&ib, 4);
210 uart_putstr("\r\n\tIC="); uart_hexdump(&ic, 4);
211 uart_putstr("\r\n\tID="); uart_hexdump(&id, 4);
213 return (((ia - ib) + ic) ^ id);
216 return (((pgm_read_dword(&s1[((uint8_t*)&t)[IA]] ) - pgm_read_dword(&s2[((uint8_t*)&t)[IB]] ))
217 + pgm_read_dword(&s3[((uint8_t*)&t)[IC]] )) ^ pgm_read_dword(&s4[((uint8_t*)&t)[ID]]));
222 uint32_t cast5_f3(uint32_t d, uint32_t m, uint8_t r){
224 t = ROTL32((m - d),r);
227 uint32_t ia,ib,ic,id;
229 uart_putstr("\r\n f3("); uart_hexdump(&d, 4); uart_putc(',');
230 uart_hexdump(&m , 4); uart_putc(','); uart_hexdump(&r, 1);uart_putstr("): I=");
233 ia = pgm_read_dword(&s1[((uint8_t*)&t)[IA]] );
234 ib = pgm_read_dword(&s2[((uint8_t*)&t)[IB]] );
235 ic = pgm_read_dword(&s3[((uint8_t*)&t)[IC]] );
236 id = pgm_read_dword(&s4[((uint8_t*)&t)[ID]] );
238 uart_putstr("\r\n\tIA="); uart_hexdump(&ia, 4);
239 uart_putstr("\r\n\tIB="); uart_hexdump(&ib, 4);
240 uart_putstr("\r\n\tIC="); uart_hexdump(&ic, 4);
241 uart_putstr("\r\n\tID="); uart_hexdump(&id, 4);
242 return (((ia + ib) ^ ic) - id);
244 return ((pgm_read_dword(&s1[((uint8_t*)&t)[IA]] ) + pgm_read_dword(&s2[((uint8_t*)&t)[IB]] ))
245 ^ pgm_read_dword(&s3[((uint8_t*)&t)[IC]] )) - pgm_read_dword(&s4[((uint8_t*)&t)[ID]] );
250 /*************************************************************************/
253 * \brief encrypts a datablock with cast5
254 * @param s Pointer to cast5 roundkeys (context)
255 * @param block Pointer to datablock
257 void cast5_enc(cast5_ctx_t *s, void* block){
260 cast5_f_t* f[]={cast5_f1,cast5_f2,cast5_f3};
261 l=((uint32_t*)block)[0];
262 r=((uint32_t*)block)[1];
263 // uart_putstr("\r\n round[-1] = ");
264 // uart_hexdump(&r, 4);
265 for (i=0;i<(s->shortkey?12:16);++i){
267 y = (f[i%3])(CHANGE_ENDIAN32(r), CHANGE_ENDIAN32(s->mask[i]),
268 (((s->roth[i>>3]) & (1<<(i&0x7)))?0x10:0x00)
269 + ( ((s->rotl[i>>1])>>((i&1)?4:0)) & 0x0f) );
270 r = l ^ CHANGE_ENDIAN32(y);
271 // uart_putstr("\r\n round["); DEBUG_B(i); uart_putstr("] = ");
272 // uart_hexdump(&r, 4);
275 ((uint32_t*)block)[0]=r;
276 ((uint32_t*)block)[1]=l;
279 /*************************************************************************/
282 * \brief decrypts a datablock with cast5
283 * @param s Pointer to cast5 roundkeys (context)
284 * @param block Pointer to datablock
286 void cast5_dec(cast5_ctx_t *s, void* block){
289 cast5_f_t* f[]={cast5_f1,cast5_f2,cast5_f3};
290 l=((uint32_t*)block)[0];
291 r=((uint32_t*)block)[1];
292 rounds = (s->shortkey?12:16);
293 for (i=rounds-1; i>=0 ;--i){
295 y = (f[i%3])(CHANGE_ENDIAN32(r), CHANGE_ENDIAN32(s->mask[i]),
296 (((s->roth[i>>3]) & (1<<(i&0x7)))?0x10:0x00)
297 + ( ((s->rotl[i>>1])>>((i&1)?4:0)) & 0x0f) );
298 r = l ^ CHANGE_ENDIAN32(y);
301 ((uint32_t*)block)[0]=r;
302 ((uint32_t*)block)[1]=l;
306 /*********************************************************************************************************/
307 /*********************************************************************************************************/
308 /*********************************************************************************************************/
312 void cast5_old_init(cast5_ctx_t* s, uint8_t* key, uint8_t keylength){
313 /* we migth return if the key is valid and if setup was sucessfull */
314 uint32_t x[4], z[4], t;
315 #define BPX ((uint8_t*)&(x[0]))
316 #define BPZ ((uint8_t*)&(z[0]))
317 s->shortkey = (keylength<=80);
318 /* littel endian only! */
319 memset(&(x[0]), 0 ,16); /* set x to zero */
320 memcpy(&(x[0]), key, keylength/8);
323 /* todo: merge a and b and compress the whole stuff */
325 z[0] = x[0] ^ S_5X(0xD) ^ S_6X(0xF) ^ S_7X(0xC) ^ S_8X(0xE) ^ S_7X(0x8);
326 z[1] = x[2] ^ S_5Z(0x0) ^ S_6Z(0x2) ^ S_7Z(0x1) ^ S_8Z(0x3) ^ S_8X(0xA);
327 z[2] = x[3] ^ S_5Z(0x7) ^ S_6Z(0x6) ^ S_7Z(0x5) ^ S_8Z(0x4) ^ S_5X(0x9);
328 z[3] = x[1] ^ S_5Z(0xA) ^ S_6Z(0x9) ^ S_7Z(0xB) ^ S_8Z(0x8) ^ S_6X(0xB);
330 s->mask[0] = S_5Z(0x8) ^ S_6Z(0x9) ^ S_7Z(0x7) ^ S_8Z(0x6) ^ S_5Z(0x2);
331 s->mask[1] = S_5Z(0xA) ^ S_6Z(0xB) ^ S_7Z(0x5) ^ S_8Z(0x4) ^ S_6Z(0x6);
332 s->mask[2] = S_5Z(0xC) ^ S_6Z(0xD) ^ S_7Z(0x3) ^ S_8Z(0x2) ^ S_7Z(0x9);
333 s->mask[3] = S_5Z(0xE) ^ S_6Z(0xF) ^ S_7Z(0x1) ^ S_8Z(0x0) ^ S_8Z(0xC);
335 x[0] = z[2] ^ S_5Z(0x5) ^ S_6Z(0x7) ^ S_7Z(0x4) ^ S_8Z(0x6) ^ S_7Z(0x0);
336 x[1] = z[0] ^ S_5X(0x0) ^ S_6X(0x2) ^ S_7X(0x1) ^ S_8X(0x3) ^ S_8Z(0x2);
337 x[2] = z[1] ^ S_5X(0x7) ^ S_6X(0x6) ^ S_7X(0x5) ^ S_8X(0x4) ^ S_5Z(0x1);
338 x[3] = z[3] ^ S_5X(0xA) ^ S_6X(0x9) ^ S_7X(0xB) ^ S_8X(0x8) ^ S_6Z(0x3);
340 s->mask[4] = S_5X(0x3) ^ S_6X(0x2) ^ S_7X(0xC) ^ S_8X(0xD) ^ S_5X(0x8);
341 s->mask[5] = S_5X(0x1) ^ S_6X(0x0) ^ S_7X(0xE) ^ S_8X(0xF) ^ S_6X(0xD);
342 s->mask[6] = S_5X(0x7) ^ S_6X(0x6) ^ S_7X(0x8) ^ S_8X(0x9) ^ S_7X(0x3);
343 s->mask[7] = S_5X(0x5) ^ S_6X(0x4) ^ S_7X(0xA) ^ S_8X(0xB) ^ S_8X(0x7);
345 z[0] = x[0] ^ S_5X(0xD) ^ S_6X(0xF) ^ S_7X(0xC) ^ S_8X(0xE) ^ S_7X(0x8);
346 z[1] = x[2] ^ S_5Z(0x0) ^ S_6Z(0x2) ^ S_7Z(0x1) ^ S_8Z(0x3) ^ S_8X(0xA);
347 z[2] = x[3] ^ S_5Z(0x7) ^ S_6Z(0x6) ^ S_7Z(0x5) ^ S_8Z(0x4) ^ S_5X(0x9);
348 z[3] = x[1] ^ S_5Z(0xA) ^ S_6Z(0x9) ^ S_7Z(0xB) ^ S_8Z(0x8) ^ S_6X(0xB);
350 s->mask[8] = S_5Z(0x3) ^ S_6Z(0x2) ^ S_7Z(0xC) ^ S_8Z(0xD) ^ S_5Z(0x9);
351 s->mask[9] = S_5Z(0x1) ^ S_6Z(0x0) ^ S_7Z(0xE) ^ S_8Z(0xF) ^ S_6Z(0xC);
352 s->mask[10] = S_5Z(0x7) ^ S_6Z(0x6) ^ S_7Z(0x8) ^ S_8Z(0x9) ^ S_7Z(0x2);
353 s->mask[11] = S_5Z(0x5) ^ S_6Z(0x4) ^ S_7Z(0xA) ^ S_8Z(0xB) ^ S_8Z(0x6);
355 x[0] = z[2] ^ S_5Z(0x5) ^ S_6Z(0x7) ^ S_7Z(0x4) ^ S_8Z(0x6) ^ S_7Z(0x0);
356 x[1] = z[0] ^ S_5X(0x0) ^ S_6X(0x2) ^ S_7X(0x1) ^ S_8X(0x3) ^ S_8Z(0x2);
357 x[2] = z[1] ^ S_5X(0x7) ^ S_6X(0x6) ^ S_7X(0x5) ^ S_8X(0x4) ^ S_5Z(0x1);
358 x[3] = z[3] ^ S_5X(0xA) ^ S_6X(0x9) ^ S_7X(0xB) ^ S_8X(0x8) ^ S_6Z(0x3);
360 s->mask[12] = S_5X(0x8) ^ S_6X(0x9) ^ S_7X(0x7) ^ S_8X(0x6) ^ S_5X(0x3);
361 s->mask[13] = S_5X(0xA) ^ S_6X(0xB) ^ S_7X(0x5) ^ S_8X(0x4) ^ S_6X(0x7);
362 s->mask[14] = S_5X(0xC) ^ S_6X(0xD) ^ S_7X(0x3) ^ S_8X(0x2) ^ S_7X(0x8);
363 s->mask[15] = S_5X(0xE) ^ S_6X(0xF) ^ S_7X(0x1) ^ S_8X(0x0) ^ S_8X(0xD);
365 /* that were the masking keys, now the rotation keys */
366 /* set the keys to zero */
367 memset(&(s->rotl[0]),0,8);
368 s->roth[0]=s->roth[1]=0;
370 z[0] = x[0] ^ S_5X(0xD) ^ S_6X(0xF) ^ S_7X(0xC) ^ S_8X(0xE) ^ S_7X(0x8);
371 z[1] = x[2] ^ S_5Z(0x0) ^ S_6Z(0x2) ^ S_7Z(0x1) ^ S_8Z(0x3) ^ S_8X(0xA);
372 z[2] = x[3] ^ S_5Z(0x7) ^ S_6Z(0x6) ^ S_7Z(0x5) ^ S_8Z(0x4) ^ S_5X(0x9);
373 z[3] = x[1] ^ S_5Z(0xA) ^ S_6Z(0x9) ^ S_7Z(0xB) ^ S_8Z(0x8) ^ S_6X(0xB);
375 t = S_5Z(0x8) ^ S_6Z(0x9) ^ S_7Z(0x7) ^ S_8Z(0x6) ^ S_5Z(0x2);
377 s->rotl[0] |= t & 0x0f;
378 s->roth[0] |= (t >> 4) & (1<<0);
379 t = S_5Z(0xA) ^ S_6Z(0xB) ^ S_7Z(0x5) ^ S_8Z(0x4) ^ S_6Z(0x6);
381 s->rotl[0] |= (t<<4) & 0xf0;
382 s->roth[0] |= (t >> 3) & (1<<1);
383 t = S_5Z(0xC) ^ S_6Z(0xD) ^ S_7Z(0x3) ^ S_8Z(0x2) ^ S_7Z(0x9);
385 s->rotl[1] |= t & 0x0f;
386 s->roth[0] |= (t >> 2) & (1<<2);
387 t = S_5Z(0xE) ^ S_6Z(0xF) ^ S_7Z(0x1) ^ S_8Z(0x0) ^ S_8Z(0xC);
389 s->rotl[1] |= (t<<4) & 0xf0;
390 s->roth[0] |= (t >> 1) & (1<<3);
392 x[0] = z[2] ^ S_5Z(0x5) ^ S_6Z(0x7) ^ S_7Z(0x4) ^ S_8Z(0x6) ^ S_7Z(0x0);
393 x[1] = z[0] ^ S_5X(0x0) ^ S_6X(0x2) ^ S_7X(0x1) ^ S_8X(0x3) ^ S_8Z(0x2);
394 x[2] = z[1] ^ S_5X(0x7) ^ S_6X(0x6) ^ S_7X(0x5) ^ S_8X(0x4) ^ S_5Z(0x1);
395 x[3] = z[3] ^ S_5X(0xA) ^ S_6X(0x9) ^ S_7X(0xB) ^ S_8X(0x8) ^ S_6Z(0x3);
397 t = S_5X(0x3) ^ S_6X(0x2) ^ S_7X(0xC) ^ S_8X(0xD) ^ S_5X(0x8);
399 s->rotl[2] |= t & 0x0f;
400 s->roth[0] |= t & (1<<4);
401 t = S_5X(0x1) ^ S_6X(0x0) ^ S_7X(0xE) ^ S_8X(0xF) ^ S_6X(0xD);
403 s->rotl[2] |= (t<<4) & 0xf0;
404 s->roth[0] |= (t<<1) & (1<<5);
405 t = S_5X(0x7) ^ S_6X(0x6) ^ S_7X(0x8) ^ S_8X(0x9) ^ S_7X(0x3);
407 s->rotl[3] |= t & 0x0f;
408 s->roth[0] |= (t<<2) & (1<<6);
409 t = S_5X(0x5) ^ S_6X(0x4) ^ S_7X(0xA) ^ S_8X(0xB) ^ S_8X(0x7);
411 s->rotl[3] |= (t<<4) & 0xf0;
412 s->roth[0] |= (t<<3) & (1<<7);
414 z[0] = x[0] ^ S_5X(0xD) ^ S_6X(0xF) ^ S_7X(0xC) ^ S_8X(0xE) ^ S_7X(0x8);
415 z[1] = x[2] ^ S_5Z(0x0) ^ S_6Z(0x2) ^ S_7Z(0x1) ^ S_8Z(0x3) ^ S_8X(0xA);
416 z[2] = x[3] ^ S_5Z(0x7) ^ S_6Z(0x6) ^ S_7Z(0x5) ^ S_8Z(0x4) ^ S_5X(0x9);
417 z[3] = x[1] ^ S_5Z(0xA) ^ S_6Z(0x9) ^ S_7Z(0xB) ^ S_8Z(0x8) ^ S_6X(0xB);
419 t = S_5Z(0x3) ^ S_6Z(0x2) ^ S_7Z(0xC) ^ S_8Z(0xD) ^ S_5Z(0x9);
421 s->rotl[4] |= t & 0x0f;
422 s->roth[1] |= (t>>4) & (1<<0);
423 t = S_5Z(0x1) ^ S_6Z(0x0) ^ S_7Z(0xE) ^ S_8Z(0xF) ^ S_6Z(0xC);
425 s->rotl[4] |= (t<<4) & 0xf0;
426 s->roth[1] |= (t>>3) & (1<<1);
427 t = S_5Z(0x7) ^ S_6Z(0x6) ^ S_7Z(0x8) ^ S_8Z(0x9) ^ S_7Z(0x2);
429 s->rotl[5] |= t & 0x0f;
430 s->roth[1] |= (t>>2) & (1<<2);
431 t = S_5Z(0x5) ^ S_6Z(0x4) ^ S_7Z(0xA) ^ S_8Z(0xB) ^ S_8Z(0x6);
433 s->rotl[5] |= (t<<4) & 0xf0;
434 s->roth[1] |= (t>>1) & (1<<3);
436 x[0] = z[2] ^ S_5Z(0x5) ^ S_6Z(0x7) ^ S_7Z(0x4) ^ S_8Z(0x6) ^ S_7Z(0x0);
437 x[1] = z[0] ^ S_5X(0x0) ^ S_6X(0x2) ^ S_7X(0x1) ^ S_8X(0x3) ^ S_8Z(0x2);
438 x[2] = z[1] ^ S_5X(0x7) ^ S_6X(0x6) ^ S_7X(0x5) ^ S_8X(0x4) ^ S_5Z(0x1);
439 x[3] = z[3] ^ S_5X(0xA) ^ S_6X(0x9) ^ S_7X(0xB) ^ S_8X(0x8) ^ S_6Z(0x3);
441 t = S_5X(0x8) ^ S_6X(0x9) ^ S_7X(0x7) ^ S_8X(0x6) ^ S_5X(0x3);
443 s->rotl[6] |= t & 0x0f;
444 s->roth[1] |= t & (1<<4);
445 t = S_5X(0xA) ^ S_6X(0xB) ^ S_7X(0x5) ^ S_8X(0x4) ^ S_6X(0x7);
447 s->rotl[6] |= (t<<4) & 0xf0;
448 s->roth[1] |= (t<<1) & (1<<5);
449 t = S_5X(0xC) ^ S_6X(0xD) ^ S_7X(0x3) ^ S_8X(0x2) ^ S_7X(0x8);
451 s->rotl[7] |= t & 0x0f;
452 s->roth[1] |= (t<<2) & (1<<6);
453 t = S_5X(0xE) ^ S_6X(0xF) ^ S_7X(0x1) ^ S_8X(0x0) ^ S_8X(0xD);
455 s->rotl[7] |= (t<<4) & 0xf0;
456 s->roth[1] |= (t<<3) & (1<<7);