X-Git-Url: https://git.cryptolib.org/?a=blobdiff_plain;f=bmw%2Fbmw_small_speed.c;h=ab03c6716c60bb5171dfcd88c72ed72a25d59b52;hb=c88d99752182184f54d3b6897705d47cea6a9562;hp=982cb13e83964dec3f011772e5420537585b3e14;hpb=5f46191d2615ebe2caa6e111b478031a34f20b9e;p=arm-crypto-lib.git diff --git a/bmw/bmw_small_speed.c b/bmw/bmw_small_speed.c index 982cb13..ab03c67 100644 --- a/bmw/bmw_small_speed.c +++ b/bmw/bmw_small_speed.c @@ -28,7 +28,6 @@ #include #include #include "bmw_small.h" -#include "memxor.h" #define SHL32(a,n) ((a)<<(n)) #define SHR32(a,n) ((a)>>(n)) @@ -105,159 +104,115 @@ #define R32_5(x) (ROTR32((x), 13)) #define R32_6(x) (ROTR32((x), 9)) #define R32_7(x) (ROTR32((x), 5)) -/* -#define K 0x05555555L -static -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[] = { - 0x55555550L, 0x5aaaaaa5L, 0x5ffffffaL, - 0x6555554fL, 0x6aaaaaa4L, 0x6ffffff9L, - 0x7555554eL, 0x7aaaaaa3L, 0x7ffffff8L, - 0x8555554dL, 0x8aaaaaa2L, 0x8ffffff7L, - 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 r; - /* r = 0x05555555*(j+16); */ - 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 ) - + k_lut[j] - ) ^ ((uint32_t*)h)[(j+7)&0xf]; - r += S32_1(q[j+ 0]) + S32_2(q[j+ 1]) + S32_3(q[j+ 2]) + S32_0(q[j+ 3]) + - S32_1(q[j+ 4]) + S32_2(q[j+ 5]) + S32_3(q[j+ 6]) + S32_0(q[j+ 7]) + - S32_1(q[j+ 8]) + S32_2(q[j+ 9]) + S32_3(q[j+10]) + S32_0(q[j+11]) + - S32_1(q[j+12]) + S32_2(q[j+13]) + S32_3(q[j+14]) + S32_0(q[j+15]); - - return r; -} -static -uint32_t bmw_small_expand2(uint8_t j, const uint32_t* q, const void* m, const void* h){ - uint32_t r; - 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 ) - + k_lut[j] - ) ^ ((uint32_t*)h)[(j+7)&0xf]; - r += (q[j+ 0]) + R32_1(q[j+ 1]) + (q[j+ 2]) + R32_2(q[j+ 3]) + - (q[j+ 4]) + R32_3(q[j+ 5]) + (q[j+ 6]) + R32_4(q[j+ 7]) + - (q[j+ 8]) + R32_5(q[j+ 9]) + (q[j+10]) + R32_6(q[j+11]) + - (q[j+12]) + R32_7(q[j+13]) + S32_4(q[j+14]) + S32_5(q[j+15]); - return r; -} +#include "f1_autogen.i" + +/* static inline */ +void bmw_small_f0(uint32_t* q, uint32_t* h, const uint32_t* m){ + uint32_t t[16]; + uint32_t tr0, tr1, tr2, tmp; + t[ 0] = h[ 0] ^ m[ 0]; + t[ 1] = h[ 1] ^ m[ 1]; + t[ 2] = h[ 2] ^ m[ 2]; + t[ 3] = h[ 3] ^ m[ 3]; + t[ 4] = h[ 4] ^ m[ 4]; + t[ 6] = h[ 6] ^ m[ 6]; + t[ 8] = h[ 8] ^ m[ 8]; + t[ 9] = h[ 9] ^ m[ 9]; + t[11] = h[11] ^ m[11]; + t[12] = h[12] ^ m[12]; + t[15] = h[15] ^ m[15]; + t[ 5] = h[ 5] ^ m[ 5]; + t[10] = h[10] ^ m[10]; + t[13] = h[13] ^ m[13]; + t[ 7] = h[ 7] ^ m[ 7]; + t[14] = h[14] ^ m[14]; + + dump_x(t, 16, 'T'); + /* + q[ 0] = (t[ 5] - t[ 7] + t[10] + t[13] + t[14]); + q[ 3] = (t[ 0] - t[ 1] + t[ 8] - t[10] + t[13]); + q[ 6] = (t[ 4] - t[ 0] - t[ 3] - t[11] + t[13]); + q[ 9] = (t[ 0] - t[ 3] + t[ 6] - t[ 7] + t[14]); + q[12] = (t[ 1] + t[ 3] - t[ 6] - t[ 9] + t[10]); + q[15] = (t[12] - t[ 4] - t[ 6] - t[ 9] + t[13]); + q[ 2] = (t[ 0] + t[ 7] + t[ 9] - t[12] + t[15]); + q[ 5] = (t[ 3] - t[ 2] + t[10] - t[12] + t[15]); + q[ 8] = (t[ 2] - t[ 5] - t[ 6] + t[13] - t[15]); + q[11] = (t[ 8] - t[ 0] - t[ 2] - t[ 5] + t[ 9]); + q[14] = (t[ 3] - t[ 5] + t[ 8] - t[11] - t[12]); + q[ 1] = (t[ 6] - t[ 8] + t[11] + t[14] - t[15]); + q[ 4] = (t[ 1] + t[ 2] + t[ 9] - t[11] - t[14]); + q[ 7] = (t[ 1] - t[ 4] - t[ 5] - t[12] - t[14]); + q[10] = (t[ 8] - t[ 1] - t[ 4] - t[ 7] + t[15]); + q[13] = (t[ 2] + t[ 4] + t[ 7] + t[10] + t[11]); + */ + tmp = +t[ 5] +t[10] +t[13] +(tr1=-t[ 7]+t[14]) ; + q[ 0] = S32_0(tmp) + h[ 1]; + tmp = +t[ 8] +t[13] +t[ 0] -(tr2=+t[ 1]+t[10]) ; + q[ 3] = S32_3(tmp) + h[ 4]; + tmp = -t[11] +t[13] -t[ 0] -t[ 3] +t[ 4] ; + q[ 6] = S32_1(tmp) + h[ 7]; + tmp = +t[ 0] +(tr0=-t[ 3]+t[ 6]) +(tr1) ; + q[ 9] = S32_4(tmp) + h[10]; + tmp = -t[ 9] -(tr0) +(tr2) ; + q[12] = S32_2(tmp) + h[13]; + tmp = -t[ 4] +(tr0=-t[ 9]+t[12]) +(tr1=-t[ 6]+t[13]) ; + q[15] = S32_0(tmp) + h[ 0]; + tmp = +t[ 7] +t[15] +t[ 0] -(tr0) ; + q[ 2] = S32_2(tmp) + h[ 3]; + tmp = +t[10] +(tr0=-t[ 2]+t[15]) +(tr2=+t[ 3]-t[12]) ; + q[ 5] = S32_0(tmp) + h[ 6]; + tmp = -t[ 5] -(tr0) +(tr1) ; + q[ 8] = S32_3(tmp) + h[ 9]; + tmp = -t[ 0] -t[ 2] +t[ 9] +(tr0=-t[ 5]+t[ 8]) ; + q[11] = S32_1(tmp) + h[12]; + tmp = -t[11] +(tr0) +(tr2) ; + q[14] = S32_4(tmp) + h[15]; + tmp = +t[ 6] +(tr0=+t[11]+t[14]) -(tr1=+t[ 8]+t[15]) ; + q[ 1] = S32_1(tmp) + h[ 2]; + tmp = +t[ 9] +t[ 1] +t[ 2] -(tr0) ; + q[ 4] = S32_4(tmp) + h[ 5]; + tmp = -t[12] -t[14] +t[ 1] -t[ 4] -t[ 5] ; + q[ 7] = S32_2(tmp) + h[ 8]; + tmp = -t[ 1] -(tr0=+t[ 4]+t[ 7]) +(tr1) ; + q[10] = S32_0(tmp) + h[11]; + tmp = +t[ 2] +t[10] +t[11] +(tr0) ; + q[13] = S32_3(tmp) + h[14]; -static -void bmw_small_f0(uint32_t* q, uint32_t* h, const void* m){ - uint8_t i; - i=15; - do{ - ((uint32_t*)h)[i] ^= ((uint32_t*)m)[i]; - }while(i--); - dump_x(h, 16, 'T'); - q[ 0] = (h[ 5] - h[ 7] + h[10] + h[13] + h[14]); - q[ 1] = (h[ 6] - h[ 8] + h[11] + h[14] - h[15]); - q[ 2] = (h[ 0] + h[ 7] + h[ 9] - h[12] + h[15]); - q[ 3] = (h[ 0] - h[ 1] + h[ 8] - h[10] + h[13]); - q[ 4] = (h[ 1] + h[ 2] + h[ 9] - h[11] - h[14]); - q[ 5] = (h[ 3] - h[ 2] + h[10] - h[12] + h[15]); - q[ 6] = (h[ 4] - h[ 0] - h[ 3] - h[11] + h[13]); - q[ 7] = (h[ 1] - h[ 4] - h[ 5] - h[12] - h[14]); - q[ 8] = (h[ 2] - h[ 5] - h[ 6] + h[13] - h[15]); - q[ 9] = (h[ 0] - h[ 3] + h[ 6] - h[ 7] + h[14]); - q[10] = (h[ 8] - h[ 1] - h[ 4] - h[ 7] + h[15]); - q[11] = (h[ 8] - h[ 0] - h[ 2] - h[ 5] + h[ 9]); - q[12] = (h[ 1] + h[ 3] - h[ 6] - h[ 9] + h[10]); - q[13] = (h[ 2] + h[ 4] + h[ 7] + h[10] + h[11]); - q[14] = (h[ 3] - h[ 5] + h[ 8] - h[11] - h[12]); - q[15] = (h[12] - h[ 4] - h[ 6] - h[ 9] + h[13]); dump_x(q, 16, 'W'); - q[ 0] = S32_0(q[ 0]); q[ 1] = S32_1(q[ 1]); q[ 2] = S32_2(q[ 2]); q[ 3] = S32_3(q[ 3]); q[ 4] = S32_4(q[ 4]); - q[ 5] = S32_0(q[ 5]); q[ 6] = S32_1(q[ 6]); q[ 7] = S32_2(q[ 7]); q[ 8] = S32_3(q[ 8]); q[ 9] = S32_4(q[ 9]); - q[10] = S32_0(q[10]); q[11] = S32_1(q[11]); q[12] = S32_2(q[12]); q[13] = S32_3(q[13]); q[14] = S32_4(q[14]); - q[15] = S32_0(q[15]); - i=15; - do{ - q[(i+15)&15] += ((uint32_t*)h)[i] ^= ((uint32_t*)m)[i]; - }while(i--); -} - -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); - q[17] = bmw_small_expand1(1, q, m, h); - for(i=2; i<16; ++i){ - q[16+i] = bmw_small_expand2(i, q, m, h); - } } -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){ - xl ^= q[i]; - } - xh = xl; - for(i=24;i<32;++i){ - xh ^= q[i]; - } +static inline +void bmw_small_f2(uint32_t* h, uint32_t* q, const uint32_t* m){ + uint32_t xl, xh; + xl = q[16] ^ q[17] ^ q[18] ^ q[19] ^ q[20] ^ q[21] ^ q[22] ^ q[23]; + xh = xl ^ q[24] ^ q[25] ^ q[26] ^ q[27] ^ q[28] ^ q[29] ^ q[30] ^ q[31]; #if DEBUG cli_putstr("\r\n XL = "); cli_hexdump_rev(&xl, 4); cli_putstr("\r\n XH = "); cli_hexdump_rev(&xh, 4); #endif - memcpy(h, m, 16*4); - h[0] ^= SHL32(xh, 5) ^ SHR32(q[16], 5); - h[1] ^= SHR32(xh, 7) ^ SHL32(q[17], 8); - h[2] ^= SHR32(xh, 5) ^ SHL32(q[18], 5); - h[3] ^= SHR32(xh, 1) ^ SHL32(q[19], 5); - h[4] ^= SHR32(xh, 3) ^ q[20]; - h[5] ^= SHL32(xh, 6) ^ SHR32(q[21], 6); - h[6] ^= SHR32(xh, 4) ^ SHL32(q[22], 6); - h[7] ^= SHR32(xh,11) ^ SHL32(q[23], 2); - for(i=0; i<8; ++i){ - h[i] += xl ^ q[24+i] ^ q[i]; - } - for(i=0; i<8; ++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[11] += SHL32(xl, 4) ^ q[18] ^ q[11]; - h[12] += SHR32(xl, 3) ^ q[19] ^ q[12]; - 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]; -*/ - i=6; - do{ - q[9+i] ^= q[16+i]; - }while(i--); - 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]; + + h[0] = (SHL32(xh, 5) ^ SHR32(q[16], 5) ^ m[ 0]) + (xl ^ q[24] ^ q[ 0]); + h[1] = (SHR32(xh, 7) ^ SHL32(q[17], 8) ^ m[ 1]) + (xl ^ q[25] ^ q[ 1]); + h[2] = (SHR32(xh, 5) ^ SHL32(q[18], 5) ^ m[ 2]) + (xl ^ q[26] ^ q[ 2]); + h[3] = (SHR32(xh, 1) ^ SHL32(q[19], 5) ^ m[ 3]) + (xl ^ q[27] ^ q[ 3]); + h[4] = (SHR32(xh, 3) ^ q[20] ^ m[ 4]) + (xl ^ q[28] ^ q[ 4]); + h[5] = (SHL32(xh, 6) ^ SHR32(q[21], 6) ^ m[ 5]) + (xl ^ q[29] ^ q[ 5]); + h[6] = (SHR32(xh, 4) ^ SHL32(q[22], 6) ^ m[ 6]) + (xl ^ q[30] ^ q[ 6]); + h[7] = (SHR32(xh,11) ^ SHL32(q[23], 2) ^ m[ 7]) + (xl ^ q[31] ^ q[ 7]); + + h[ 8] = ROTL32(h[4], 9) + (xh ^ q[24] ^ m[ 8]) + (SHL32(xl, 8) ^ q[23] ^ q[ 8]); + h[ 9] = ROTL32(h[5], 10) + (xh ^ q[25] ^ m[ 9]) + (SHR32(xl, 6) ^ q[16] ^ q[ 9]); + h[10] = ROTL32(h[6], 11) + (xh ^ q[26] ^ m[10]) + (SHL32(xl, 6) ^ q[17] ^ q[10]); + h[11] = ROTL32(h[7], 12) + (xh ^ q[27] ^ m[11]) + (SHL32(xl, 4) ^ q[18] ^ q[11]); + h[12] = ROTL32(h[0], 13) + (xh ^ q[28] ^ m[12]) + (SHR32(xl, 3) ^ q[19] ^ q[12]); + h[13] = ROTL32(h[1], 14) + (xh ^ q[29] ^ m[13]) + (SHR32(xl, 4) ^ q[20] ^ q[13]); + h[14] = ROTL32(h[2], 15) + (xh ^ q[30] ^ m[14]) + (SHR32(xl, 7) ^ q[21] ^ q[14]); + h[15] = ROTL32(h[3], 16) + (xh ^ q[31] ^ m[15]) + (SHR32(xl, 2) ^ q[22] ^ q[15]); } void bmw_small_nextBlock(bmw_small_ctx_t* ctx, const void* block){ @@ -266,7 +221,7 @@ void bmw_small_nextBlock(bmw_small_ctx_t* ctx, const void* block){ bmw_small_f0(q, ctx->h, block); dump_x(q, 16, 'Q'); bmw_small_f1(q, block, ctx->h); - dump_x(q, 32, 'Q'); + dump_x(q+16, 16, 'Q'); bmw_small_f2(ctx->h, q, block); ctx->counter += 1; ctx_dump(ctx);