1 /* bmw_large_speed.c */
3 This file is part of the ARM-Crypto-Lib.
4 Copyright (C) 2006-2010 Daniel Otte (daniel.otte@rub.de)
6 This program is free software: you can redistribute it and/or modify
7 it under the terms of the GNU General Public License as published by
8 the Free Software Foundation, either version 3 of the License, or
9 (at your option) any later version.
11 This program is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 GNU General Public License for more details.
16 You should have received a copy of the GNU General Public License
17 along with this program. If not, see <http://www.gnu.org/licenses/>.
22 * \email daniel.otte@rub.de
24 * \license GPLv3 or later
30 #include "bmw_large.h"
32 #define SHL64(a,n) ((a)<<(n))
33 #define SHR64(a,n) ((a)>>(n))
34 #define ROTL64(a,n) (((a)<<(n))|((a)>>(64-(n))))
35 #define ROTR64(a,n) (((a)>>(n))|((a)<<(64-(n))))
46 void ctx_dump(const bmw_large_ctx_t* ctx){
48 cli_putstr("\r\n==== ctx dump ====");
50 cli_putstr("\r\n h[");
53 cli_hexdump_rev(&(ctx->h[i]), 8);
55 cli_putstr("\r\n counter = ");
56 cli_hexdump(&(ctx->counter), 4);
59 void dump_x(const uint64_t* q, uint8_t elements, char x){
61 cli_putstr("\r\n==== ");
63 cli_putstr(" dump ====");
64 for(i=0; i<elements;++i){
70 cli_hexdump_rev(&(q[i]), 8);
78 #define S64_0(x) ( (SHR64((x), 1)) ^ \
83 #define S64_1(x) ( (SHR64((x), 1)) ^ \
88 #define S64_2(x) ( (SHR64((x), 2)) ^ \
93 #define S64_3(x) ( (SHR64((x), 2)) ^ \
98 #define S64_4(x) ( (SHR64((x), 1)) ^ x)
100 #define S64_5(x) ( (SHR64((x), 2)) ^ x)
102 #define R64_1(x) (ROTL64((x), 5))
103 #define R64_2(x) (ROTL64((x), 11))
104 #define R64_3(x) (ROTL64((x), 27))
105 #define R64_4(x) (ROTL64((x), 32))
106 #define R64_5(x) (ROTR64((x), 27))
107 #define R64_6(x) (ROTR64((x), 21))
108 #define R64_7(x) (ROTR64((x), 11))
111 #define K 0x0555555555555555LL
112 #define MASK 0xFFFFFFFFFFFFFFFFLL
114 uint64_t k_lut[] PROGMEM = {
115 16LL*K, 17LL*K, 18LL*K, 19LL*K,
116 20LL*K, 21LL*K, 22LL*K, 23LL*K,
117 24LL*K, 25LL*K, 26LL*K, 27LL*K,
118 28LL*K, 29LL*K, 30LL*K, 31LL*K };
120 /* the same as above but precomputed to avoid compiler warnings */
123 0x5555555555555550LL, 0x5aaaaaaaaaaaaaa5LL, 0x5ffffffffffffffaLL,
124 0x655555555555554fLL, 0x6aaaaaaaaaaaaaa4LL, 0x6ffffffffffffff9LL,
125 0x755555555555554eLL, 0x7aaaaaaaaaaaaaa3LL, 0x7ffffffffffffff8LL,
126 0x855555555555554dLL, 0x8aaaaaaaaaaaaaa2LL, 0x8ffffffffffffff7LL,
127 0x955555555555554cLL, 0x9aaaaaaaaaaaaaa1LL, 0x9ffffffffffffff6LL,
128 0xa55555555555554bLL };
131 uint64_t bmw_large_expand1(uint8_t j, const uint64_t* q, const void* m, const void* h){
133 /* r = 0x0555555555555555LL*(j+16); */
134 r = ( ROTL64(((uint64_t*)m)[(j)&0xf], ((j+ 0)&0xf)+1)
135 + ROTL64(((uint64_t*)m)[(j+3)&0xf], ((j+ 3)&0xf)+1)
137 - ROTL64(((uint64_t*)m)[(j+10)&0xf],((j+10)&0xf)+1)
138 ) ^ ((uint64_t*)h)[(j+7)&0xf];
139 r += S64_1(q[j+ 0]) + S64_2(q[j+ 1]) + S64_3(q[j+ 2]) + S64_0(q[j+ 3]) +
140 S64_1(q[j+ 4]) + S64_2(q[j+ 5]) + S64_3(q[j+ 6]) + S64_0(q[j+ 7]) +
141 S64_1(q[j+ 8]) + S64_2(q[j+ 9]) + S64_3(q[j+10]) + S64_0(q[j+11]) +
142 S64_1(q[j+12]) + S64_2(q[j+13]) + S64_3(q[j+14]) + S64_0(q[j+15]);
148 uint64_t bmw_large_expand2(uint8_t j, const uint64_t* q, const void* m, const void* h){
150 r = ( ROTL64(((uint64_t*)m)[(j)&0xf], ((j+ 0)&0xf)+1)
151 + ROTL64(((uint64_t*)m)[(j+3)&0xf], ((j+ 3)&0xf)+1)
153 - ROTL64(((uint64_t*)m)[(j+10)&0xf],((j+10)&0xf)+1)
154 ) ^ ((uint64_t*)h)[(j+7)&0xf];
155 r += (q[j+ 0]) + R64_1(q[j+ 1]) + (q[j+ 2]) + R64_2(q[j+ 3]) +
156 (q[j+ 4]) + R64_3(q[j+ 5]) + (q[j+ 6]) + R64_4(q[j+ 7]) +
157 (q[j+ 8]) + R64_5(q[j+ 9]) + (q[j+10]) + R64_6(q[j+11]) +
158 (q[j+12]) + R64_7(q[j+13]) + S64_4(q[j+14]) + S64_5(q[j+15]);
164 /* to understand this implementation take a look at f0-opt-table.txt */
165 static uint16_t hack_table[5] = { 0x0311, 0xDDB3, 0x2A79, 0x07AA, 0x51C2 };
166 static uint8_t offset_table[5] = { 4+16, 6+16, 9+16, 12+16, 13+16 };
170 void bmw_large_f0(uint64_t* q, const uint64_t* h, const void* m){
173 uint64_t(*s[])(uint64_t)={ bmw_large_s0, bmw_large_s1, bmw_large_s2,
174 bmw_large_s3, bmw_large_s4 };
176 ((uint64_t*)h)[i] ^= ((uint64_t*)m)[i];
184 hack_reg = hack_table[c];
201 ((uint64_t*)h)[i] ^= ((uint64_t*)m)[i];
204 q[i] += h[(i+1)&0xf];
208 #endif /* F0_HACK==2 */
212 uint8_t f0_lut[] PROGMEM ={
213 5<<1, ( 7<<1)+1, (10<<1)+0, (13<<1)+0, (14<<1)+0,
214 6<<1, ( 8<<1)+1, (11<<1)+0, (14<<1)+0, (15<<1)+1,
215 0<<1, ( 7<<1)+0, ( 9<<1)+0, (12<<1)+1, (15<<1)+0,
216 0<<1, ( 1<<1)+1, ( 8<<1)+0, (10<<1)+1, (13<<1)+0,
217 1<<1, ( 2<<1)+0, ( 9<<1)+0, (11<<1)+1, (14<<1)+1,
218 3<<1, ( 2<<1)+1, (10<<1)+0, (12<<1)+1, (15<<1)+0,
219 4<<1, ( 0<<1)+1, ( 3<<1)+1, (11<<1)+1, (13<<1)+0,
220 1<<1, ( 4<<1)+1, ( 5<<1)+1, (12<<1)+1, (14<<1)+1,
221 2<<1, ( 5<<1)+1, ( 6<<1)+1, (13<<1)+0, (15<<1)+1,
222 0<<1, ( 3<<1)+1, ( 6<<1)+0, ( 7<<1)+1, (14<<1)+0,
223 8<<1, ( 1<<1)+1, ( 4<<1)+1, ( 7<<1)+1, (15<<1)+0,
224 8<<1, ( 0<<1)+1, ( 2<<1)+1, ( 5<<1)+1, ( 9<<1)+0,
225 1<<1, ( 3<<1)+0, ( 6<<1)+1, ( 9<<1)+1, (10<<1)+0,
226 2<<1, ( 4<<1)+0, ( 7<<1)+0, (10<<1)+0, (11<<1)+0,
227 3<<1, ( 5<<1)+1, ( 8<<1)+0, (11<<1)+1, (12<<1)+1,
228 12<<1, ( 4<<1)+1, ( 6<<1)+1, ( 9<<1)+1, (13<<1)+0
232 void bmw_large_f0(uint64_t* q, const uint64_t* h, const void* m){
233 uint8_t i,j=-1,v,sign,l=0;
234 uint64_t(*s[])(uint64_t)={ bmw_large_s0, bmw_large_s1, bmw_large_s2,
235 bmw_large_s3, bmw_large_s4 };
237 ((uint64_t*)h)[i] ^= ((uint64_t*)m)[i];
240 // memset(q, 0, 4*16);
241 for(i=0; i<5*16; ++i){
242 v = pgm_read_byte(f0_lut+i);
263 ((uint64_t*)h)[i] ^= ((uint64_t*)m)[i];
266 q[i] += h[(i+1)&0xf];
270 #endif /* F0_HACK==1 */
274 void bmw_large_f0(uint64_t* q, const uint64_t* h, const void* m){
277 ((uint64_t*)h)[i] ^= ((uint64_t*)m)[i];
279 // dump_x(t, 16, 'T');
280 q[ 0] = (h[ 5] - h[ 7] + h[10] + h[13] + h[14]);
281 q[ 1] = (h[ 6] - h[ 8] + h[11] + h[14] - h[15]);
282 q[ 2] = (h[ 0] + h[ 7] + h[ 9] - h[12] + h[15]);
283 q[ 3] = (h[ 0] - h[ 1] + h[ 8] - h[10] + h[13]);
284 q[ 4] = (h[ 1] + h[ 2] + h[ 9] - h[11] - h[14]);
285 q[ 5] = (h[ 3] - h[ 2] + h[10] - h[12] + h[15]);
286 q[ 6] = (h[ 4] - h[ 0] - h[ 3] - h[11] + h[13]);
287 q[ 7] = (h[ 1] - h[ 4] - h[ 5] - h[12] - h[14]);
288 q[ 8] = (h[ 2] - h[ 5] - h[ 6] + h[13] - h[15]);
289 q[ 9] = (h[ 0] - h[ 3] + h[ 6] - h[ 7] + h[14]);
290 q[10] = (h[ 8] - h[ 1] - h[ 4] - h[ 7] + h[15]);
291 q[11] = (h[ 8] - h[ 0] - h[ 2] - h[ 5] + h[ 9]);
292 q[12] = (h[ 1] + h[ 3] - h[ 6] - h[ 9] + h[10]);
293 q[13] = (h[ 2] + h[ 4] + h[ 7] + h[10] + h[11]);
294 q[14] = (h[ 3] - h[ 5] + h[ 8] - h[11] - h[12]);
295 q[15] = (h[12] - h[ 4] - h[ 6] - h[ 9] + h[13]);
297 q[ 0] = S64_0(q[ 0]); q[ 1] = S64_1(q[ 1]); q[ 2] = S64_2(q[ 2]); q[ 3] = S64_3(q[ 3]); q[ 4] = S64_4(q[ 4]);
298 q[ 5] = S64_0(q[ 5]); q[ 6] = S64_1(q[ 6]); q[ 7] = S64_2(q[ 7]); q[ 8] = S64_3(q[ 8]); q[ 9] = S64_4(q[ 9]);
299 q[10] = S64_0(q[10]); q[11] = S64_1(q[11]); q[12] = S64_2(q[12]); q[13] = S64_3(q[13]); q[14] = S64_4(q[14]);
300 q[15] = S64_0(q[15]);
303 ((uint64_t*)h)[i] ^= ((uint64_t*)m)[i];
306 q[i] += h[(i+1)&0xf];
309 #endif /* F0_HACK==0 */
312 void bmw_large_f1(uint64_t* q, const void* m, const uint64_t* h){
314 q[16] = bmw_large_expand1(0, q, m, h);
315 q[17] = bmw_large_expand1(1, q, m, h);
317 q[16+i] = bmw_large_expand2(i, q, m, h);
322 void bmw_large_f2(uint64_t* h, const uint64_t* q, const void* m){
333 cli_putstr("\r\n XL = ");
334 cli_hexdump_rev(&xl, 4);
335 cli_putstr("\r\n XH = ");
336 cli_hexdump_rev(&xh, 4);
339 h[0] ^= SHL64(xh, 5) ^ SHR64(q[16], 5);
340 h[1] ^= SHR64(xh, 7) ^ SHL64(q[17], 8);
341 h[2] ^= SHR64(xh, 5) ^ SHL64(q[18], 5);
342 h[3] ^= SHR64(xh, 1) ^ SHL64(q[19], 5);
343 h[4] ^= SHR64(xh, 3) ^ q[20];
344 h[5] ^= SHL64(xh, 6) ^ SHR64(q[21], 6);
345 h[6] ^= SHR64(xh, 4) ^ SHL64(q[22], 6);
346 h[7] ^= SHR64(xh,11) ^ SHL64(q[23], 2);
348 h[i] += xl ^ q[24+i] ^ q[i];
351 h[8+i] ^= xh ^ q[24+i];
352 h[8+i] += ROTL64(h[(4+i)%8],i+9);
354 h[ 8] += SHL64(xl, 8) ^ q[23] ^ q[ 8];
355 h[ 9] += SHR64(xl, 6) ^ q[16] ^ q[ 9];
356 h[10] += SHL64(xl, 6) ^ q[17] ^ q[10];
357 h[11] += SHL64(xl, 4) ^ q[18] ^ q[11];
358 h[12] += SHR64(xl, 3) ^ q[19] ^ q[12];
359 h[13] += SHR64(xl, 4) ^ q[20] ^ q[13];
360 h[14] += SHR64(xl, 7) ^ q[21] ^ q[14];
361 h[15] += SHR64(xl, 2) ^ q[22] ^ q[15];
364 void bmw_large_nextBlock(bmw_large_ctx_t* ctx, const void* block){
366 dump_x(block, 16, 'M');
367 bmw_large_f0(q, ctx->h, block);
369 bmw_large_f1(q, block, ctx->h);
371 bmw_large_f2(ctx->h, q, block);
376 void bmw_large_lastBlock(bmw_large_ctx_t* ctx, const void* block, uint16_t length_b){
378 while(length_b >= BMW_LARGE_BLOCKSIZE){
379 bmw_large_nextBlock(ctx, block);
380 length_b -= BMW_LARGE_BLOCKSIZE;
381 block = (uint8_t*)block + BMW_LARGE_BLOCKSIZE_B;
383 memset(buffer, 0, 128);
384 memcpy(buffer, block, (length_b+7)/8);
385 buffer[length_b>>3] |= 0x80 >> (length_b&0x07);
386 if(length_b+1>128*8-64){
387 bmw_large_nextBlock(ctx, buffer);
388 memset(buffer, 0, 128-8);
391 *((uint64_t*)&(buffer[128-8])) = (uint64_t)(ctx->counter*1024LL)+(uint64_t)length_b;
392 bmw_large_nextBlock(ctx, buffer);
396 memset(buffer, 0xaa, 128);
398 buffer[8*i] = i + 0xa0;
400 bmw_large_f0(q, (uint64_t*)buffer, ctx->h);
401 bmw_large_f1(q, ctx->h, (uint64_t*)buffer);
402 bmw_large_f2((uint64_t*)buffer, q, ctx->h);
403 memcpy(ctx->h, buffer, 128);
407 void bmw384_init(bmw384_ctx_t* ctx){
409 ctx->h[0] = 0x0001020304050607LL;
411 ctx->h[i] = ctx->h[i-1]+ 0x0808080808080808LL;
414 ctx->h[6] = 0x3031323324353637LL;
420 void bmw512_init(bmw512_ctx_t* ctx){
422 ctx->h[0] = 0x8081828384858687LL;
424 ctx->h[i] = ctx->h[i-1]+ 0x0808080808080808LL;
430 void bmw384_nextBlock(bmw384_ctx_t* ctx, const void* block){
431 bmw_large_nextBlock(ctx, block);
434 void bmw512_nextBlock(bmw512_ctx_t* ctx, const void* block){
435 bmw_large_nextBlock(ctx, block);
438 void bmw384_lastBlock(bmw384_ctx_t* ctx, const void* block, uint16_t length_b){
439 bmw_large_lastBlock(ctx, block, length_b);
442 void bmw512_lastBlock(bmw512_ctx_t* ctx, const void* block, uint16_t length_b){
443 bmw_large_lastBlock(ctx, block, length_b);
446 void bmw384_ctx2hash(void* dest, const bmw384_ctx_t* ctx){
447 memcpy(dest, &(ctx->h[10]), 384/8);
450 void bmw512_ctx2hash(void* dest, const bmw512_ctx_t* ctx){
451 memcpy(dest, &(ctx->h[8]), 512/8);
454 void bmw384(void* dest, const void* msg, uint32_t length_b){
457 while(length_b>=BMW_LARGE_BLOCKSIZE){
458 bmw_large_nextBlock(&ctx, msg);
459 length_b -= BMW_LARGE_BLOCKSIZE;
460 msg = (uint8_t*)msg + BMW_LARGE_BLOCKSIZE_B;
462 bmw_large_lastBlock(&ctx, msg, length_b);
463 bmw384_ctx2hash(dest, &ctx);
466 void bmw512(void* dest, const void* msg, uint32_t length_b){
469 while(length_b>=BMW_LARGE_BLOCKSIZE){
470 bmw_large_nextBlock(&ctx, msg);
471 length_b -= BMW_LARGE_BLOCKSIZE;
472 msg = (uint8_t*)msg + BMW_LARGE_BLOCKSIZE_B;
474 bmw_large_lastBlock(&ctx, msg, length_b);
475 bmw512_ctx2hash(dest, &ctx);