#include <stdint.h>
#include <string.h>
#include <avr/pgmspace.h>
+#include "memxor.h"
#include "bmw_small.h"
# define BUG24 1
#endif
-#define F0_HACK 1
+#define F0_HACK 2
#define DEBUG 0
+#ifndef F0_HACK
+# define F0_HACK 0
+#endif
+
#if DEBUG
#include "cli.h"
#define dump_x(a,b,c)
#endif
+static
uint32_t bmw_small_s0(uint32_t x){
uint32_t r;
r = SHR32(x, 1)
return r;
}
+static
uint32_t bmw_small_s1(uint32_t x){
uint32_t r;
r = SHR32(x, 1)
return r;
}
+static
uint32_t bmw_small_s2(uint32_t x){
uint32_t r;
r = SHR32(x, 2)
return r;
}
+static
uint32_t bmw_small_s3(uint32_t x){
uint32_t r;
r = SHR32(x, 2)
return r;
}
+static
uint32_t bmw_small_s4(uint32_t x){
uint32_t r;
r = SHR32(x, 1)
return r;
}
+static
uint32_t bmw_small_s5(uint32_t x){
uint32_t r;
r = SHR32(x, 2)
return r;
}
+static
uint32_t bmw_small_r1(uint32_t x){
uint32_t r;
r = ROTL32(x, 3);
return r;
}
+static
uint32_t bmw_small_r2(uint32_t x){
uint32_t r;
r = ROTL32(x, 7);
return r;
}
+static
uint32_t bmw_small_r3(uint32_t x){
uint32_t r;
r = ROTL32(x, 13);
return r;
}
+static
uint32_t bmw_small_r4(uint32_t x){
uint32_t r;
r = ROTL32(x, 16);
return r;
}
+static
uint32_t bmw_small_r5(uint32_t x){
uint32_t r;
r = ROTR32(x, 13);
return r;
}
+static
uint32_t bmw_small_r6(uint32_t x){
uint32_t r;
r = ROTR32(x, 9);
return r;
}
+static
uint32_t bmw_small_r7(uint32_t x){
uint32_t r;
r = ROTR32(x, 5);
/*
#define K 0x05555555L
static
-uint32_t k_lut[] PROGMEM = {
+const uint32_t k_lut[] PROGMEM = {
16L*K, 17L*K, 18L*K, 19L*K, 20L*K, 21L*K, 22L*K, 23L*K,
24L*K, 25L*K, 26L*K, 27L*K, 28L*K, 29L*K, 30L*K, 31L*K
};
/* same as above but precomputed to avoid compiler warnings */
static
-uint32_t k_lut[] PROGMEM = {
+const uint32_t k_lut[] PROGMEM = {
0x55555550L, 0x5aaaaaa5L, 0x5ffffffaL,
0x6555554fL, 0x6aaaaaa4L, 0x6ffffff9L,
0x7555554eL, 0x7aaaaaa3L, 0x7ffffff8L,
0x9555554cL, 0x9aaaaaa1L, 0x9ffffff6L,
0xa555554bL };
-
+static
uint32_t bmw_small_expand1(uint8_t j, const uint32_t* q, const void* m, const void* h){
uint32_t(*s[])(uint32_t) = {bmw_small_s1, bmw_small_s2, bmw_small_s3, bmw_small_s0};
- uint32_t r=0;
+ uint32_t r;
uint8_t i;
/* r = 0x05555555*(j+16); */
- for(i=0; i<16; ++i){
- r += s[i%4](q[j+i]);
- }
+
#if TWEAK
- r += ( ROTL32(((uint32_t*)m)[j&0xf], ((j+0)&0xf)+1 )
+ r = ( ROTL32(((uint32_t*)m)[j&0xf], ((j+0)&0xf)+1 )
+ ROTL32(((uint32_t*)m)[(j+3)&0xf], ((j+3)&0xf)+1 )
- ROTL32(((uint32_t*)m)[(j+10)&0xf], ((j+10)&0xf)+1 )
+ pgm_read_dword(k_lut+j)
) ^ ((uint32_t*)h)[(j+7)&0xf];
#else
- r += pgm_read_dword(k_lut+j);
+ r = pgm_read_dword(k_lut+j);
r += ((uint32_t*)m)[j&0xf];
r += ((uint32_t*)m)[(j+3)&0xf];
r -= ((uint32_t*)m)[(j+10)&0xf];
#endif
+ for(i=0; i<16; ++i){
+ r += s[i%4](q[j+i]);
+ }
return r;
}
+static
uint32_t bmw_small_expand2(uint8_t j, const uint32_t* q, const void* m, const void* h){
uint32_t(*rf[])(uint32_t) = {bmw_small_r1, bmw_small_r2, bmw_small_r3,
bmw_small_r4, bmw_small_r5, bmw_small_r6,
return r;
}
-#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
-uint8_t f0_lut[] PROGMEM = {
+void bmw_small_f0(uint32_t* q, uint32_t* h, const void* m){
+ uint16_t hack_reg;
+ uint8_t c,i,j;
+ uint32_t(*s[])(uint32_t)={ bmw_small_s0, bmw_small_s1, bmw_small_s2,
+ bmw_small_s3, bmw_small_s4 };
+ for(i=0; i<16; ++i){
+ ((uint32_t*)h)[i] ^= ((uint32_t*)m)[i];
+ }
+ dump_x(h, 16, 'T');
+ memset(q, 0, 4*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){
+ ((uint32_t*)h)[i] ^= ((uint32_t*)m)[i];
+ }
+ for(i=0; i<16; ++i){
+ q[i] += h[(i+1)&0xf];
+ }
+#endif
+}
+#endif /* F0_HACK==2*/
+
+#if F0_HACK==1
+static
+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_small_f0(uint32_t* q, uint32_t* h, const void* m){
uint8_t i,j=-1,v,sign,l=0;
uint32_t(*s[])(uint32_t)={ bmw_small_s0, bmw_small_s1, bmw_small_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_small_f0(uint32_t* q, uint32_t* h, const void* m){
uint8_t i;
uint32_t(*s[])(uint32_t)={ bmw_small_s0, bmw_small_s1, bmw_small_s2,
for(i=0; i<16; ++i){
q[i] += h[(i+1)&0xf];
}
-#endif
+#endif /* TWEAK */
}
-#endif
+#endif /* F0_HACK==0 */
+static
void bmw_small_f1(uint32_t* q, const void* m, const void* h){
uint8_t i;
q[16] = bmw_small_expand1(0, q, m, h);
}
}
-void bmw_small_f2(uint32_t* h, const uint32_t* q, const void* m){
+static
+void bmw_small_f2(uint32_t* h, uint32_t* q, const void* m){
uint32_t xl=0, xh;
uint8_t i;
for(i=16;i<24;++i){
h[8+i] ^= xh ^ q[24+i];
h[8+i] += ROTL32(h[(4+i)%8],i+9);
}
+/*
h[ 8] += SHL32(xl, 8) ^ q[23] ^ q[ 8];
h[ 9] += SHR32(xl, 6) ^ q[16] ^ q[ 9];
h[10] += SHL32(xl, 6) ^ q[17] ^ q[10];
h[13] += SHR32(xl, 4) ^ q[20] ^ q[13];
h[14] += SHR32(xl, 7) ^ q[21] ^ q[14];
h[15] += SHR32(xl, 2) ^ q[22] ^ q[15];
+*/
+ memxor(q+9, q+16, 7*4);
+ q[8] ^= q[23];
+ h[ 8] += SHL32(xl, 8) ^ q[ 8];
+ h[ 9] += SHR32(xl, 6) ^ q[ 9];
+ h[10] += SHL32(xl, 6) ^ q[10];
+ h[11] += SHL32(xl, 4) ^ q[11];
+ h[12] += SHR32(xl, 3) ^ q[12];
+ h[13] += SHR32(xl, 4) ^ q[13];
+ h[14] += SHR32(xl, 7) ^ q[14];
+ h[15] += SHR32(xl, 2) ^ q[15];
+
}
void bmw_small_nextBlock(bmw_small_ctx_t* ctx, const void* block){
}
void bmw_small_lastBlock(bmw_small_ctx_t* ctx, const void* block, uint16_t length_b){
- uint8_t buffer[64];
+ union {
+ uint8_t v8[64];
+ uint32_t v32[16];
+ uint64_t v64[ 8];
+ } buffer;
while(length_b >= BMW_SMALL_BLOCKSIZE){
bmw_small_nextBlock(ctx, block);
length_b -= BMW_SMALL_BLOCKSIZE;
block = (uint8_t*)block + BMW_SMALL_BLOCKSIZE_B;
}
- memset(buffer, 0, 64);
- memcpy(buffer, block, (length_b+7)/8);
- buffer[length_b>>3] |= 0x80 >> (length_b&0x07);
+ memset(buffer.v8, 0, 64);
+ memcpy(buffer.v8, block, (length_b+7)/8);
+ buffer.v8[length_b>>3] |= 0x80 >> (length_b&0x07);
if(length_b+1>64*8-64){
- bmw_small_nextBlock(ctx, buffer);
- memset(buffer, 0, 64-8);
+ bmw_small_nextBlock(ctx, buffer.v8);
+ memset(buffer.v8, 0, 64-8);
ctx->counter -= 1;
}
- *((uint64_t*)&(buffer[64-8])) = (uint64_t)(ctx->counter*512LL)+(uint64_t)length_b;
- bmw_small_nextBlock(ctx, buffer);
+ buffer.v64[7] = (uint64_t)(ctx->counter*512LL)+(uint64_t)length_b;
+ bmw_small_nextBlock(ctx, buffer.v8);
#if TWEAK
uint8_t i;
uint32_t q[32];
- memset(buffer, 0xaa, 64);
+ memset(buffer.v8, 0xaa, 64);
for(i=0; i<16;++i){
- buffer[i*4] = i+0xa0;
+ buffer.v8[i*4] = i+0xa0;
}
-// dump_x(buffer, 16, 'A');
+// dump_x(buffer.v8, 16, 'A');
dump_x(ctx->h, 16, 'M');
- bmw_small_f0(q, (uint32_t*)buffer, ctx->h);
- dump_x(buffer, 16, 'a');
+ bmw_small_f0(q, buffer.v32, ctx->h);
+ dump_x(buffer.v8, 16, 'a');
dump_x(q, 16, 'Q');
- bmw_small_f1(q, ctx->h, (uint32_t*)buffer);
+ bmw_small_f1(q, ctx->h, buffer.v32);
dump_x(q, 32, 'Q');
- bmw_small_f2((uint32_t*)buffer, q, ctx->h);
- memcpy(ctx->h, buffer, 64);
+ bmw_small_f2(buffer.v32, q, ctx->h);
+ memcpy(ctx->h, buffer.v8, 64);
#endif
}