#include <avr/pgmspace.h>
#include "bmw_large.h"
-#define SHL64(a,n) ((a)<<(n))
-#define SHR64(a,n) ((a)>>(n))
-#define ROTL64(a,n) (((a)<<(n))|((a)>>(64-(n))))
-#define ROTR64(a,n) (((a)>>(n))|((a)<<(64-(n))))
+#define SHL64(a,n) shiftl64(a,n)
+#define SHR64(a,n) shiftr64(a,n)
+#define ROTL64(a,n) rotl64(a,n)
+#define ROTR64(a,n) rotr64(a,n)
#define TWEAK 1
#define BUG24 0
-#define F0_HACK 1
+#define F0_HACK 2
#define DEBUG 0
#if DEBUG
#define dump_x(a,b,c)
#endif
+static
+uint64_t rotl64(uint64_t a, uint8_t r){
+ return (a<<r)|(a>>(64-r));
+}
+
+static
+uint64_t rotr64(uint64_t a, uint8_t r){
+ return (a>>r)|(a<<(64-r));
+}
+
+static
+uint64_t shiftl64(uint64_t a, uint8_t r){
+ return (a<<r);
+}
+
+static
+uint64_t shiftr64(uint64_t a, uint8_t r){
+ return (a>>r);
+}
+
+static
uint64_t bmw_large_s0(uint64_t x){
uint64_t r;
r = SHR64(x, 1)
return r;
}
+static
uint64_t bmw_large_s1(uint64_t x){
uint64_t r;
r = SHR64(x, 1)
return r;
}
+static
uint64_t bmw_large_s2(uint64_t x){
uint64_t r;
r = SHR64(x, 2)
return r;
}
+static
uint64_t bmw_large_s3(uint64_t x){
uint64_t r;
r = SHR64(x, 2)
return r;
}
+static
uint64_t bmw_large_s4(uint64_t x){
uint64_t r;
r = SHR64(x, 1)
return r;
}
+static
uint64_t bmw_large_s5(uint64_t x){
uint64_t r;
r = SHR64(x, 2)
return r;
}
+static
uint64_t bmw_large_r1(uint64_t x){
uint64_t r;
r = ROTL64(x, 5);
return r;
}
+static
uint64_t bmw_large_r2(uint64_t x){
uint64_t r;
r = ROTL64(x, 11);
return r;
}
+static
uint64_t bmw_large_r3(uint64_t x){
uint64_t r;
r = ROTL64(x, 27);
return r;
}
+static
uint64_t bmw_large_r4(uint64_t x){
uint64_t r;
r = ROTL64(x, 32);
return r;
}
+static
uint64_t bmw_large_r5(uint64_t x){
uint64_t r;
r = ROTR64(x, 64-37);
return r;
}
+static
uint64_t bmw_large_r6(uint64_t x){
uint64_t r;
r = ROTR64(x, 64-43);
return r;
}
+static
uint64_t bmw_large_r7(uint64_t x){
uint64_t r;
r = ROTR64(x, 64-53);
#define K 0x0555555555555555LL
#define MASK 0xFFFFFFFFFFFFFFFFLL
static
-uint64_t k_lut[] PROGMEM = {
+const uint64_t k_lut[] PROGMEM = {
16LL*K, 17LL*K, 18LL*K, 19LL*K,
20LL*K, 21LL*K, 22LL*K, 23LL*K,
24LL*K, 25LL*K, 26LL*K, 27LL*K,
*/
/* the same as above but precomputed to avoid compiler warnings */
static
-uint64_t k_lut[] PROGMEM = {
+const uint64_t k_lut[] PROGMEM = {
0x5555555555555550LL, 0x5aaaaaaaaaaaaaa5LL, 0x5ffffffffffffffaLL,
0x655555555555554fLL, 0x6aaaaaaaaaaaaaa4LL, 0x6ffffffffffffff9LL,
0x755555555555554eLL, 0x7aaaaaaaaaaaaaa3LL, 0x7ffffffffffffff8LL,
0x955555555555554cLL, 0x9aaaaaaaaaaaaaa1LL, 0x9ffffffffffffff6LL,
0xa55555555555554bLL };
+static
uint64_t bmw_large_expand1(uint8_t j, const uint64_t* q, const void* m, const void* h){
uint64_t(*s[])(uint64_t) = {bmw_large_s1, bmw_large_s2, bmw_large_s3, bmw_large_s0};
uint64_t a = 0;
return a;
}
+static
uint64_t bmw_large_expand2(uint8_t j, const uint64_t* q, const void* m, const void* h){
uint64_t(*rf[])(uint64_t) = {bmw_large_r1, bmw_large_r2, bmw_large_r3,
bmw_large_r4, bmw_large_r5, bmw_large_r6,
return a;
}
-#if F0_HACK
+#if F0_HACK==2
+/* to understand this implementation take a look at f0-opt-table.txt */
+static const uint16_t hack_table[5] PROGMEM = { 0x0311, 0xDDB3, 0x2A79, 0x07AA, 0x51C2 };
+static const uint8_t offset_table[5] PROGMEM = { 4+16, 6+16, 9+16, 12+16, 13+16 };
+
+
+static
+void bmw_large_f0(uint64_t* q, const uint64_t* h, const void* m){
+ uint16_t hack_reg;
+ uint8_t i,j,c;
+ uint64_t(*s[])(uint64_t)={ bmw_large_s0, bmw_large_s1, bmw_large_s2,
+ bmw_large_s3, bmw_large_s4 };
+ for(i=0; i<16; ++i){
+ ((uint64_t*)h)[i] ^= ((uint64_t*)m)[i];
+ }
+ dump_x(h, 16, 'T');
+ memset(q, 0, 8*16);
+ c=4;
+ do{
+ i=15;
+ j=pgm_read_byte(offset_table+c);
+ hack_reg=pgm_read_word(&(hack_table[c]));
+ do{
+ if(hack_reg&1){
+ q[i]-= h[j&15];
+ }else{
+ q[i]+= h[j&15];
+ }
+ --j;
+ hack_reg>>= 1;
+ }while(i--!=0);
+ }while(c--!=0);
+ dump_x(q, 16, 'W');
+ for(i=0; i<16; ++i){
+ q[i] = s[i%5](q[i]);
+ }
+#if TWEAK
+ for(i=0; i<16; ++i){
+ ((uint64_t*)h)[i] ^= ((uint64_t*)m)[i];
+ }
+ for(i=0; i<16; ++i){
+ q[i] += h[(i+1)&0xf];
+ }
+#endif /* TWEAK */
+}
+#endif /* F0_HACK==2 */
+
+#if F0_HACK==1
static
-uint8_t f0_lut[] PROGMEM ={
+const uint8_t f0_lut[] PROGMEM ={
5<<1, ( 7<<1)+1, (10<<1)+0, (13<<1)+0, (14<<1)+0,
6<<1, ( 8<<1)+1, (11<<1)+0, (14<<1)+0, (15<<1)+1,
0<<1, ( 7<<1)+0, ( 9<<1)+0, (12<<1)+1, (15<<1)+0,
12<<1, ( 4<<1)+1, ( 6<<1)+1, ( 9<<1)+1, (13<<1)+0
};
+static
void bmw_large_f0(uint64_t* q, const uint64_t* h, const void* m){
uint8_t i,j=-1,v,sign,l=0;
uint64_t(*s[])(uint64_t)={ bmw_large_s0, bmw_large_s1, bmw_large_s2,
for(i=0; i<16; ++i){
q[i] += h[(i+1)&0xf];
}
-#endif
+#endif /* TWEAK */
}
+#endif /* F0_HACK==1 */
-#else
+#if F0_HACK==0
+static
void bmw_large_f0(uint64_t* q, const uint64_t* h, const void* m){
uint8_t i;
uint64_t(*s[])(uint64_t)={ bmw_large_s0, bmw_large_s1, bmw_large_s2,
for(i=0; i<16; ++i){
q[i] += h[(i+1)&0xf];
}
-#endif
+#endif /* TWEAK */
}
-#endif
+#endif /* F0_HACK==0 */
+static
void bmw_large_f1(uint64_t* q, const void* m, const uint64_t* h){
uint8_t i;
q[16] = bmw_large_expand1(0, q, m, h);
}
}
+static
void bmw_large_f2(uint64_t* h, const uint64_t* q, const void* m){
uint64_t xl=0, xh;
uint8_t i;
}
void bmw_large_lastBlock(bmw_large_ctx_t* ctx, const void* block, uint16_t length_b){
- uint8_t buffer[128];
+ union {
+ uint8_t v8[128];
+ uint64_t v64[ 16];
+ } buffer;
while(length_b >= BMW_LARGE_BLOCKSIZE){
bmw_large_nextBlock(ctx, block);
length_b -= BMW_LARGE_BLOCKSIZE;
block = (uint8_t*)block + BMW_LARGE_BLOCKSIZE_B;
}
- memset(buffer, 0, 128);
- memcpy(buffer, block, (length_b+7)/8);
- buffer[length_b>>3] |= 0x80 >> (length_b&0x07);
+ memset(buffer.v8, 0, 128);
+ memcpy(buffer.v8, block, (length_b+7)/8);
+ buffer.v8[length_b>>3] |= 0x80 >> (length_b&0x07);
if(length_b+1>128*8-64){
- bmw_large_nextBlock(ctx, buffer);
- memset(buffer, 0, 128-8);
+ bmw_large_nextBlock(ctx, buffer.v8);
+ memset(buffer.v8, 0, 128-8);
ctx->counter -= 1;
}
- *((uint64_t*)&(buffer[128-8])) = (uint64_t)(ctx->counter*1024LL)+(uint64_t)length_b;
- bmw_large_nextBlock(ctx, buffer);
+ buffer.v64[15] = (uint64_t)(ctx->counter*1024LL)+(uint64_t)length_b;
+ bmw_large_nextBlock(ctx, buffer.v8);
#if TWEAK
uint8_t i;
uint64_t q[32];
- memset(buffer, 0xaa, 128);
+ memset(buffer.v8, 0xaa, 128);
for(i=0; i<16; ++i){
- buffer[8*i] = i + 0xa0;
+ buffer.v8[8*i] = i + 0xa0;
}
- bmw_large_f0(q, (uint64_t*)buffer, ctx->h);
- bmw_large_f1(q, ctx->h, (uint64_t*)buffer);
- bmw_large_f2((uint64_t*)buffer, q, ctx->h);
- memcpy(ctx->h, buffer, 128);
+ bmw_large_f0(q, buffer.v64, ctx->h);
+ bmw_large_f1(q, ctx->h, buffer.v64);
+ bmw_large_f2(buffer.v64, q, ctx->h);
+ memcpy(ctx->h, buffer.v8, 128);
#endif
}