Esempio n. 1
0
void skein_512_process_block(struct skein_512_ctx *ctx, const u8 *blk_ptr,
			     size_t blk_cnt, size_t byte_cnt_add)
{ /* do it in C */
	enum {
		WCNT = SKEIN_512_STATE_WORDS
	};
#undef  RCNT
#define RCNT  (SKEIN_512_ROUNDS_TOTAL/8)

#ifdef SKEIN_LOOP /* configure how much to unroll the loop */
#define SKEIN_UNROLL_512 (((SKEIN_LOOP)/10)%10)
#else
#define SKEIN_UNROLL_512 (0)
#endif

#if SKEIN_UNROLL_512
#if (RCNT % SKEIN_UNROLL_512)
#error "Invalid SKEIN_UNROLL_512" /* sanity check on unroll count */
#endif
	size_t  r;
	u64  kw[WCNT+4+RCNT*2]; /* key sched: chaining vars + tweak + "rot"*/
#else
	u64  kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
#endif
	u64  X0, X1, X2, X3, X4, X5, X6, X7; /* local copies, for speed */
	u64  w[WCNT]; /* local copy of input block */
#ifdef SKEIN_DEBUG
	const u64 *X_ptr[8]; /* use for debugging (help cc put Xn in regs) */

	X_ptr[0] = &X0;  X_ptr[1] = &X1;  X_ptr[2] = &X2;  X_ptr[3] = &X3;
	X_ptr[4] = &X4;  X_ptr[5] = &X5;  X_ptr[6] = &X6;  X_ptr[7] = &X7;
#endif

	skein_assert(blk_cnt != 0); /* never call with blk_cnt == 0! */
	ts[0] = ctx->h.tweak[0];
	ts[1] = ctx->h.tweak[1];
	do  {
		/*
		 * this implementation only supports 2**64 input bytes
		 * (no carry out here)
		 */
		ts[0] += byte_cnt_add; /* update processed length */

		/* precompute the key schedule for this block */
		ks[0] = ctx->x[0];
		ks[1] = ctx->x[1];
		ks[2] = ctx->x[2];
		ks[3] = ctx->x[3];
		ks[4] = ctx->x[4];
		ks[5] = ctx->x[5];
		ks[6] = ctx->x[6];
		ks[7] = ctx->x[7];
		ks[8] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
			ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^ SKEIN_KS_PARITY;

		ts[2] = ts[0] ^ ts[1];

		/* get input block in little-endian format */
		skein_get64_lsb_first(w, blk_ptr, WCNT);
		debug_save_tweak(ctx);
		skein_show_block(BLK_BITS, &ctx->h, ctx->x, blk_ptr, w, ks, ts);

		X0   = w[0] + ks[0]; /* do the first full key injection */
		X1   = w[1] + ks[1];
		X2   = w[2] + ks[2];
		X3   = w[3] + ks[3];
		X4   = w[4] + ks[4];
		X5   = w[5] + ks[5] + ts[0];
		X6   = w[6] + ks[6] + ts[1];
		X7   = w[7] + ks[7];

		blk_ptr += SKEIN_512_BLOCK_BYTES;

		skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INITIAL,
				 X_ptr);
		/* run the rounds */
#define ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) \
do { \
	X##p0 += X##p1; X##p1 = rotl_64(X##p1, ROT##_0); X##p1 ^= X##p0; \
	X##p2 += X##p3; X##p3 = rotl_64(X##p3, ROT##_1); X##p3 ^= X##p2; \
	X##p4 += X##p5; X##p5 = rotl_64(X##p5, ROT##_2); X##p5 ^= X##p4; \
	X##p6 += X##p7; X##p7 = rotl_64(X##p7, ROT##_3); X##p7 ^= X##p6; \
} while (0)

#if SKEIN_UNROLL_512 == 0
#define R512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) /* unrolled */ \
do { \
	ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) \
	skein_show_r_ptr(BLK_BITS, &ctx->h, r_num, X_ptr); \
} while (0)

#define I512(R) \
do { \
	/* inject the key schedule value */ \
	X0   += ks[((R) + 1) % 9]; \
	X1   += ks[((R) + 2) % 9]; \
	X2   += ks[((R) + 3) % 9]; \
	X3   += ks[((R) + 4) % 9]; \
	X4   += ks[((R) + 5) % 9]; \
	X5   += ks[((R) + 6) % 9] + ts[((R) + 1) % 3]; \
	X6   += ks[((R) + 7) % 9] + ts[((R) + 2) % 3]; \
	X7   += ks[((R) + 8) % 9] +     (R) + 1;       \
	skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, X_ptr); \
} while (0)
#else /* looping version */
#define R512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) \
do { \
	ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num); \
	skein_show_r_ptr(BLK_BITS, &ctx->h, 4 * (r - 1) + r_num, X_ptr); \
} while (0)

#define I512(R) \
do { \
	/* inject the key schedule value */ \
	X0   += ks[r + (R) + 0]; \
	X1   += ks[r + (R) + 1]; \
	X2   += ks[r + (R) + 2]; \
	X3   += ks[r + (R) + 3]; \
	X4   += ks[r + (R) + 4]; \
	X5   += ks[r + (R) + 5] + ts[r + (R) + 0]; \
	X6   += ks[r + (R) + 6] + ts[r + (R) + 1]; \
	X7   += ks[r + (R) + 7] +         r + (R); \
	/* rotate key schedule */ \
	ks[r +         (R) + 8] = ks[r + (R) - 1]; \
	ts[r +         (R) + 2] = ts[r + (R) - 1]; \
	skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, X_ptr); \
} while (0)

		for (r = 1; r < 2 * RCNT; r += 2 * SKEIN_UNROLL_512)
#endif /* end of looped code definitions */
		{
#define R512_8_ROUNDS(R)  /* do 8 full rounds */  \
do { \
		R512(0, 1, 2, 3, 4, 5, 6, 7, R_512_0, 8 * (R) + 1);   \
		R512(2, 1, 4, 7, 6, 5, 0, 3, R_512_1, 8 * (R) + 2);   \
		R512(4, 1, 6, 3, 0, 5, 2, 7, R_512_2, 8 * (R) + 3);   \
		R512(6, 1, 0, 7, 2, 5, 4, 3, R_512_3, 8 * (R) + 4);   \
		I512(2 * (R));                              \
		R512(0, 1, 2, 3, 4, 5, 6, 7, R_512_4, 8 * (R) + 5);   \
		R512(2, 1, 4, 7, 6, 5, 0, 3, R_512_5, 8 * (R) + 6);   \
		R512(4, 1, 6, 3, 0, 5, 2, 7, R_512_6, 8 * (R) + 7);   \
		R512(6, 1, 0, 7, 2, 5, 4, 3, R_512_7, 8 * (R) + 8);   \
		I512(2 * (R) + 1);        /* and key injection */ \
} while (0)

			R512_8_ROUNDS(0);

#define R512_UNROLL_R(NN) \
		((SKEIN_UNROLL_512 == 0 && \
		  SKEIN_512_ROUNDS_TOTAL/8 > (NN)) || \
		 (SKEIN_UNROLL_512 > (NN)))

	#if   R512_UNROLL_R(1)
			R512_8_ROUNDS(1);
	#endif
	#if   R512_UNROLL_R(2)
			R512_8_ROUNDS(2);
	#endif
	#if   R512_UNROLL_R(3)
			R512_8_ROUNDS(3);
	#endif
	#if   R512_UNROLL_R(4)
			R512_8_ROUNDS(4);
	#endif
	#if   R512_UNROLL_R(5)
			R512_8_ROUNDS(5);
	#endif
	#if   R512_UNROLL_R(6)
			R512_8_ROUNDS(6);
	#endif
	#if   R512_UNROLL_R(7)
			R512_8_ROUNDS(7);
	#endif
	#if   R512_UNROLL_R(8)
			R512_8_ROUNDS(8);
	#endif
	#if   R512_UNROLL_R(9)
			R512_8_ROUNDS(9);
	#endif
	#if   R512_UNROLL_R(10)
			R512_8_ROUNDS(10);
	#endif
	#if   R512_UNROLL_R(11)
			R512_8_ROUNDS(11);
	#endif
	#if   R512_UNROLL_R(12)
			R512_8_ROUNDS(12);
	#endif
	#if   R512_UNROLL_R(13)
			R512_8_ROUNDS(13);
	#endif
	#if   R512_UNROLL_R(14)
			R512_8_ROUNDS(14);
	#endif
	#if  (SKEIN_UNROLL_512 > 14)
#error  "need more unrolling in skein_512_process_block"
	#endif
		}

		/* do the final "feedforward" xor, update context chaining */
		ctx->x[0] = X0 ^ w[0];
		ctx->x[1] = X1 ^ w[1];
		ctx->x[2] = X2 ^ w[2];
		ctx->x[3] = X3 ^ w[3];
		ctx->x[4] = X4 ^ w[4];
		ctx->x[5] = X5 ^ w[5];
		ctx->x[6] = X6 ^ w[6];
		ctx->x[7] = X7 ^ w[7];
		skein_show_round(BLK_BITS, &ctx->h, SKEIN_RND_FEED_FWD, ctx->x);

		ts[1] &= ~SKEIN_T1_FLAG_FIRST;
	} while (--blk_cnt);
	ctx->h.tweak[0] = ts[0];
	ctx->h.tweak[1] = ts[1];
}
Esempio n. 2
0
void skein_512_process_block(struct skein_512_ctx *ctx, const u8 *blk_ptr,
			     size_t blk_cnt, size_t byte_cnt_add)
{ /* do it in C */
	enum {
		WCNT = SKEIN_512_STATE_WORDS
	};
	size_t  r;
#if SKEIN_UNROLL_512
	u64  kw[WCNT+4+RCNT*2]; /* key sched: chaining vars + tweak + "rot"*/
#else
	u64  kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
#endif
	u64  X0, X1, X2, X3, X4, X5, X6, X7; /* local copies, for speed */
	u64  w[WCNT]; /* local copy of input block */
#ifdef SKEIN_DEBUG
	const u64 *X_ptr[8]; /* use for debugging (help cc put Xn in regs) */

	X_ptr[0] = &X0;
	X_ptr[1] = &X1;
	X_ptr[2] = &X2;
	X_ptr[3] = &X3;
	X_ptr[4] = &X4;
	X_ptr[5] = &X5;
	X_ptr[6] = &X6;
	X_ptr[7] = &X7;
#endif

	skein_assert(blk_cnt != 0); /* never call with blk_cnt == 0! */
	ts[0] = ctx->h.tweak[0];
	ts[1] = ctx->h.tweak[1];
	do  {
		/*
		 * this implementation only supports 2**64 input bytes
		 * (no carry out here)
		 */
		ts[0] += byte_cnt_add; /* update processed length */

		/* precompute the key schedule for this block */
		ks[0] = ctx->x[0];
		ks[1] = ctx->x[1];
		ks[2] = ctx->x[2];
		ks[3] = ctx->x[3];
		ks[4] = ctx->x[4];
		ks[5] = ctx->x[5];
		ks[6] = ctx->x[6];
		ks[7] = ctx->x[7];
		ks[8] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
			ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^ SKEIN_KS_PARITY;

		ts[2] = ts[0] ^ ts[1];

		/* get input block in little-endian format */
		skein_get64_lsb_first(w, blk_ptr, WCNT);
		debug_save_tweak(ctx);

		/* do the first full key injection */
		X0 = w[0] + ks[0];
		X1 = w[1] + ks[1];
		X2 = w[2] + ks[2];
		X3 = w[3] + ks[3];
		X4 = w[4] + ks[4];
		X5 = w[5] + ks[5] + ts[0];
		X6 = w[6] + ks[6] + ts[1];
		X7 = w[7] + ks[7];

		blk_ptr += SKEIN_512_BLOCK_BYTES;

		/* run the rounds */
		for (r = 1;
			r < (SKEIN_UNROLL_512 ? 2 * RCNT : 2);
			r += (SKEIN_UNROLL_512 ? 2 * SKEIN_UNROLL_512 : 1)) {

			R512_8_ROUNDS(0);

#if   R512_UNROLL_R(1)
			R512_8_ROUNDS(1);
#endif
#if   R512_UNROLL_R(2)
			R512_8_ROUNDS(2);
#endif
#if   R512_UNROLL_R(3)
			R512_8_ROUNDS(3);
#endif
#if   R512_UNROLL_R(4)
			R512_8_ROUNDS(4);
#endif
#if   R512_UNROLL_R(5)
			R512_8_ROUNDS(5);
#endif
#if   R512_UNROLL_R(6)
			R512_8_ROUNDS(6);
#endif
#if   R512_UNROLL_R(7)
			R512_8_ROUNDS(7);
#endif
#if   R512_UNROLL_R(8)
			R512_8_ROUNDS(8);
#endif
#if   R512_UNROLL_R(9)
			R512_8_ROUNDS(9);
#endif
#if   R512_UNROLL_R(10)
			R512_8_ROUNDS(10);
#endif
#if   R512_UNROLL_R(11)
			R512_8_ROUNDS(11);
#endif
#if   R512_UNROLL_R(12)
			R512_8_ROUNDS(12);
#endif
#if   R512_UNROLL_R(13)
			R512_8_ROUNDS(13);
#endif
#if   R512_UNROLL_R(14)
			R512_8_ROUNDS(14);
#endif
		}

		/* do the final "feedforward" xor, update context chaining */
		ctx->x[0] = X0 ^ w[0];
		ctx->x[1] = X1 ^ w[1];
		ctx->x[2] = X2 ^ w[2];
		ctx->x[3] = X3 ^ w[3];
		ctx->x[4] = X4 ^ w[4];
		ctx->x[5] = X5 ^ w[5];
		ctx->x[6] = X6 ^ w[6];
		ctx->x[7] = X7 ^ w[7];

		ts[1] &= ~SKEIN_T1_FLAG_FIRST;
	} while (--blk_cnt);
	ctx->h.tweak[0] = ts[0];
	ctx->h.tweak[1] = ts[1];
}