xref: /freebsd/contrib/bearssl/src/symcipher/des_ct.c (revision 0957b409)

/*
 * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org>
 *
 * Permission is hereby granted, free of charge, to any person obtaining
 * a copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sublicense, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice shall be
 * included in all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#include "inner.h"

/*
 * During key schedule, we need to apply bit extraction PC-2 then permute
 * things into our bitslice representation. PC-2 extracts 48 bits out
 * of two 28-bit words (kl and kr), and we store these bits into two
 * 32-bit words sk0 and sk1.
 *
 *  -- bit 16+x of sk0 comes from bit QL0[x] of kl
 *  -- bit x of sk0 comes from bit QR0[x] of kr
 *  -- bit 16+x of sk1 comes from bit QL1[x] of kl
 *  -- bit x of sk1 comes from bit QR1[x] of kr
 */

static const unsigned char QL0[] = {
	17,  4, 27, 23, 13, 22,  7, 18,
	16, 24,  2, 20,  1,  8, 15, 26
};

static const unsigned char QR0[] = {
	25, 19,  9,  1,  5, 11, 23,  8,
	17,  0, 22,  3,  6, 20, 27, 24
};

static const unsigned char QL1[] = {
	28, 28, 14, 11, 28, 28, 25,  0,
	28, 28,  5,  9, 28, 28, 12, 21
};

static const unsigned char QR1[] = {
	28, 28, 15,  4, 28, 28, 26, 16,
	28, 28, 12,  7, 28, 28, 10, 14
};

/*
 * 32-bit rotation. The C compiler is supposed to recognize it as a
 * rotation and use the local architecture rotation opcode (if available).
 */
static inline uint32_t
rotl(uint32_t x, int n)
{
	return (x << n) | (x >> (32 - n));
}

/*
 * Compute key schedule for 8 key bytes (produces 32 subkey words).
 */
static void
keysched_unit(uint32_t *skey, const void *key)
{
	int i;

	br_des_keysched_unit(skey, key);

	/*
	 * Apply PC-2 + bitslicing.
	 */
	for (i = 0; i < 16; i ++) {
		uint32_t kl, kr, sk0, sk1;
		int j;

		kl = skey[(i << 1) + 0];
		kr = skey[(i << 1) + 1];
		sk0 = 0;
		sk1 = 0;
		for (j = 0; j < 16; j ++) {
			sk0 <<= 1;
			sk1 <<= 1;
			sk0 |= ((kl >> QL0[j]) & (uint32_t)1) << 16;
			sk0 |= (kr >> QR0[j]) & (uint32_t)1;
			sk1 |= ((kl >> QL1[j]) & (uint32_t)1) << 16;
			sk1 |= (kr >> QR1[j]) & (uint32_t)1;
		}

		skey[(i << 1) + 0] = sk0;
		skey[(i << 1) + 1] = sk1;
	}

#if 0
		/*
		 * Speed-optimized version for PC-2 + bitslicing.
		 * (Unused. Kept for reference only.)
		 */
		sk0 = kl & (uint32_t)0x00100000;
		sk0 |= (kl & (uint32_t)0x08008000) << 2;
		sk0 |= (kl & (uint32_t)0x00400000) << 4;
		sk0 |= (kl & (uint32_t)0x00800000) << 5;
		sk0 |= (kl & (uint32_t)0x00040000) << 6;
		sk0 |= (kl & (uint32_t)0x00010000) << 7;
		sk0 |= (kl & (uint32_t)0x00000100) << 10;
		sk0 |= (kl & (uint32_t)0x00022000) << 14;
		sk0 |= (kl & (uint32_t)0x00000082) << 18;
		sk0 |= (kl & (uint32_t)0x00000004) << 19;
		sk0 |= (kl & (uint32_t)0x04000000) >> 10;
		sk0 |= (kl & (uint32_t)0x00000010) << 26;
		sk0 |= (kl & (uint32_t)0x01000000) >> 2;

		sk0 |= kr & (uint32_t)0x00000100;
		sk0 |= (kr & (uint32_t)0x00000008) << 1;
		sk0 |= (kr & (uint32_t)0x00000200) << 4;
		sk0 |= rotl(kr & (uint32_t)0x08000021, 6);
		sk0 |= (kr & (uint32_t)0x01000000) >> 24;
		sk0 |= (kr & (uint32_t)0x00000002) << 11;
		sk0 |= (kr & (uint32_t)0x00100000) >> 18;
		sk0 |= (kr & (uint32_t)0x00400000) >> 17;
		sk0 |= (kr & (uint32_t)0x00800000) >> 14;
		sk0 |= (kr & (uint32_t)0x02020000) >> 10;
		sk0 |= (kr & (uint32_t)0x00080000) >> 5;
		sk0 |= (kr & (uint32_t)0x00000040) >> 3;
		sk0 |= (kr & (uint32_t)0x00000800) >> 1;

		sk1 = kl & (uint32_t)0x02000000;
		sk1 |= (kl & (uint32_t)0x00001000) << 5;
		sk1 |= (kl & (uint32_t)0x00000200) << 11;
		sk1 |= (kl & (uint32_t)0x00004000) << 15;
		sk1 |= (kl & (uint32_t)0x00000020) << 16;
		sk1 |= (kl & (uint32_t)0x00000800) << 17;
		sk1 |= (kl & (uint32_t)0x00000001) << 24;
		sk1 |= (kl & (uint32_t)0x00200000) >> 5;

		sk1 |= (kr & (uint32_t)0x00000010) << 8;
		sk1 |= (kr & (uint32_t)0x04000000) >> 17;
		sk1 |= (kr & (uint32_t)0x00004000) >> 14;
		sk1 |= (kr & (uint32_t)0x00000400) >> 9;
		sk1 |= (kr & (uint32_t)0x00010000) >> 8;
		sk1 |= (kr & (uint32_t)0x00001000) >> 7;
		sk1 |= (kr & (uint32_t)0x00000080) >> 3;
		sk1 |= (kr & (uint32_t)0x00008000) >> 2;
#endif
}

/* see inner.h */
unsigned
br_des_ct_keysched(uint32_t *skey, const void *key, size_t key_len)
{
	switch (key_len) {
	case 8:
		keysched_unit(skey, key);
		return 1;
	case 16:
		keysched_unit(skey, key);
		keysched_unit(skey + 32, (const unsigned char *)key + 8);
		br_des_rev_skey(skey + 32);
		memcpy(skey + 64, skey, 32 * sizeof *skey);
		return 3;
	default:
		keysched_unit(skey, key);
		keysched_unit(skey + 32, (const unsigned char *)key + 8);
		br_des_rev_skey(skey + 32);
		keysched_unit(skey + 64, (const unsigned char *)key + 16);
		return 3;
	}
}

/*
 * DES confusion function. This function performs expansion E (32 to
 * 48 bits), XOR with subkey, S-boxes, and permutation P.
 */
static inline uint32_t
Fconf(uint32_t r0, const uint32_t *sk)
{
	/*
	 * Each 6->4 S-box is virtually turned into four 6->1 boxes; we
	 * thus end up with 32 boxes that we call "T-boxes" here. We will
	 * evaluate them with bitslice code.
	 *
	 * Each T-box is a circuit of multiplexers (sort of) and thus
	 * takes 70 inputs: the 6 actual T-box inputs, and 64 constants
	 * that describe the T-box output for all combinations of the
	 * 6 inputs. With this model, all T-boxes are identical (with
	 * distinct inputs) and thus can be executed in parallel with
	 * bitslice code.
	 *
	 * T-boxes are numbered from 0 to 31, in least-to-most
	 * significant order. Thus, S-box S1 corresponds to T-boxes 31,
	 * 30, 29 and 28, in that order. T-box 'n' is computed with the
	 * bits at rank 'n' in the 32-bit words.
	 *
	 * Words x0 to x5 contain the T-box inputs 0 to 5.
	 */
	uint32_t x0, x1, x2, x3, x4, x5, z0;
	uint32_t y0, y1, y2, y3, y4, y5, y6, y7, y8, y9;
	uint32_t y10, y11, y12, y13, y14, y15, y16, y17, y18, y19;
	uint32_t y20, y21, y22, y23, y24, y25, y26, y27, y28, y29;
	uint32_t y30;

	/*
	 * Spread input bits over the 6 input words x*.
	 */
	x1 = r0 & (uint32_t)0x11111111;
	x2 = (r0 >> 1) & (uint32_t)0x11111111;
	x3 = (r0 >> 2) & (uint32_t)0x11111111;
	x4 = (r0 >> 3) & (uint32_t)0x11111111;
	x1 = (x1 << 4) - x1;
	x2 = (x2 << 4) - x2;
	x3 = (x3 << 4) - x3;
	x4 = (x4 << 4) - x4;
	x0 = (x4 << 4) | (x4 >> 28);
	x5 = (x1 >> 4) | (x1 << 28);

	/*
	 * XOR with the subkey for this round.
	 */
	x0 ^= sk[0];
	x1 ^= sk[1];
	x2 ^= sk[2];
	x3 ^= sk[3];
	x4 ^= sk[4];
	x5 ^= sk[5];

	/*
	 * The T-boxes are done in parallel, since they all use a
	 * "tree of multiplexer". We use "fake multiplexers":
	 *
	 *   y = a ^ (x & b)
	 *
	 * computes y as either 'a' (if x == 0) or 'a ^ b' (if x == 1).
	 */
	y0 = (uint32_t)0xEFA72C4D ^ (x0 & (uint32_t)0xEC7AC69C);
	y1 = (uint32_t)0xAEAAEDFF ^ (x0 & (uint32_t)0x500FB821);
	y2 = (uint32_t)0x37396665 ^ (x0 & (uint32_t)0x40EFA809);
	y3 = (uint32_t)0x68D7B833 ^ (x0 & (uint32_t)0xA5EC0B28);
	y4 = (uint32_t)0xC9C755BB ^ (x0 & (uint32_t)0x252CF820);
	y5 = (uint32_t)0x73FC3606 ^ (x0 & (uint32_t)0x40205801);
	y6 = (uint32_t)0xA2A0A918 ^ (x0 & (uint32_t)0xE220F929);
	y7 = (uint32_t)0x8222BD90 ^ (x0 & (uint32_t)0x44A3F9E1);
	y8 = (uint32_t)0xD6B6AC77 ^ (x0 & (uint32_t)0x794F104A);
	y9 = (uint32_t)0x3069300C ^ (x0 & (uint32_t)0x026F320B);
	y10 = (uint32_t)0x6CE0D5CC ^ (x0 & (uint32_t)0x7640B01A);
	y11 = (uint32_t)0x59A9A22D ^ (x0 & (uint32_t)0x238F1572);
	y12 = (uint32_t)0xAC6D0BD4 ^ (x0 & (uint32_t)0x7A63C083);
	y13 = (uint32_t)0x21C83200 ^ (x0 & (uint32_t)0x11CCA000);
	y14 = (uint32_t)0xA0E62188 ^ (x0 & (uint32_t)0x202F69AA);
	/* y15 = (uint32_t)0x00000000 ^ (x0 & (uint32_t)0x00000000); */
	y16 = (uint32_t)0xAF7D655A ^ (x0 & (uint32_t)0x51B33BE9);
	y17 = (uint32_t)0xF0168AA3 ^ (x0 & (uint32_t)0x3B0FE8AE);
	y18 = (uint32_t)0x90AA30C6 ^ (x0 & (uint32_t)0x90BF8816);
	y19 = (uint32_t)0x5AB2750A ^ (x0 & (uint32_t)0x09E34F9B);
	y20 = (uint32_t)0x5391BE65 ^ (x0 & (uint32_t)0x0103BE88);
	y21 = (uint32_t)0x93372BAF ^ (x0 & (uint32_t)0x49AC8E25);
	y22 = (uint32_t)0xF288210C ^ (x0 & (uint32_t)0x922C313D);
	y23 = (uint32_t)0x920AF5C0 ^ (x0 & (uint32_t)0x70EF31B0);
	y24 = (uint32_t)0x63D312C0 ^ (x0 & (uint32_t)0x6A707100);
	y25 = (uint32_t)0x537B3006 ^ (x0 & (uint32_t)0xB97C9011);
	y26 = (uint32_t)0xA2EFB0A5 ^ (x0 & (uint32_t)0xA320C959);
	y27 = (uint32_t)0xBC8F96A5 ^ (x0 & (uint32_t)0x6EA0AB4A);
	y28 = (uint32_t)0xFAD176A5 ^ (x0 & (uint32_t)0x6953DDF8);
	y29 = (uint32_t)0x665A14A3 ^ (x0 & (uint32_t)0xF74F3E2B);
	y30 = (uint32_t)0xF2EFF0CC ^ (x0 & (uint32_t)0xF0306CAD);
	/* y31 = (uint32_t)0x00000000 ^ (x0 & (uint32_t)0x00000000); */

	y0 = y0 ^ (x1 & y1);
	y1 = y2 ^ (x1 & y3);
	y2 = y4 ^ (x1 & y5);
	y3 = y6 ^ (x1 & y7);
	y4 = y8 ^ (x1 & y9);
	y5 = y10 ^ (x1 & y11);
	y6 = y12 ^ (x1 & y13);
	y7 = y14; /* was: y14 ^ (x1 & y15) */
	y8 = y16 ^ (x1 & y17);
	y9 = y18 ^ (x1 & y19);
	y10 = y20 ^ (x1 & y21);
	y11 = y22 ^ (x1 & y23);
	y12 = y24 ^ (x1 & y25);
	y13 = y26 ^ (x1 & y27);
	y14 = y28 ^ (x1 & y29);
	y15 = y30; /* was: y30 ^ (x1 & y31) */

	y0 = y0 ^ (x2 & y1);
	y1 = y2 ^ (x2 & y3);
	y2 = y4 ^ (x2 & y5);
	y3 = y6 ^ (x2 & y7);
	y4 = y8 ^ (x2 & y9);
	y5 = y10 ^ (x2 & y11);
	y6 = y12 ^ (x2 & y13);
	y7 = y14 ^ (x2 & y15);

	y0 = y0 ^ (x3 & y1);
	y1 = y2 ^ (x3 & y3);
	y2 = y4 ^ (x3 & y5);
	y3 = y6 ^ (x3 & y7);

	y0 = y0 ^ (x4 & y1);
	y1 = y2 ^ (x4 & y3);

	y0 = y0 ^ (x5 & y1);

	/*
	 * The P permutation:
	 * -- Each bit move is converted into a mask + left rotation.
	 * -- Rotations that use the same movement are coalesced together.
	 * -- Left and right shifts are used as alternatives to a rotation
	 * where appropriate (this will help architectures that do not have
	 * a rotation opcode).
	 */
	z0 = (y0 & (uint32_t)0x00000004) << 3;
	z0 |= (y0 & (uint32_t)0x00004000) << 4;
	z0 |= rotl(y0 & 0x12020120, 5);
	z0 |= (y0 & (uint32_t)0x00100000) << 6;
	z0 |= (y0 & (uint32_t)0x00008000) << 9;
	z0 |= (y0 & (uint32_t)0x04000000) >> 22;
	z0 |= (y0 & (uint32_t)0x00000001) << 11;
	z0 |= rotl(y0 & 0x20000200, 12);
	z0 |= (y0 & (uint32_t)0x00200000) >> 19;
	z0 |= (y0 & (uint32_t)0x00000040) << 14;
	z0 |= (y0 & (uint32_t)0x00010000) << 15;
	z0 |= (y0 & (uint32_t)0x00000002) << 16;
	z0 |= rotl(y0 & 0x40801800, 17);
	z0 |= (y0 & (uint32_t)0x00080000) >> 13;
	z0 |= (y0 & (uint32_t)0x00000010) << 21;
	z0 |= (y0 & (uint32_t)0x01000000) >> 10;
	z0 |= rotl(y0 & 0x88000008, 24);
	z0 |= (y0 & (uint32_t)0x00000480) >> 7;
	z0 |= (y0 & (uint32_t)0x00442000) >> 6;
	return z0;
}

/*
 * Process one block through 16 successive rounds, omitting the swap
 * in the final round.
 */
static void
process_block_unit(uint32_t *pl, uint32_t *pr, const uint32_t *sk_exp)
{
	int i;
	uint32_t l, r;

	l = *pl;
	r = *pr;
	for (i = 0; i < 16; i ++) {
		uint32_t t;

		t = l ^ Fconf(r, sk_exp);
		l = r;
		r = t;
		sk_exp += 6;
	}
	*pl = r;
	*pr = l;
}

/* see inner.h */
void
br_des_ct_process_block(unsigned num_rounds,
	const uint32_t *sk_exp, void *block)
{
	unsigned char *buf;
	uint32_t l, r;

	buf = block;
	l = br_dec32be(buf);
	r = br_dec32be(buf + 4);
	br_des_do_IP(&l, &r);
	while (num_rounds -- > 0) {
		process_block_unit(&l, &r, sk_exp);
		sk_exp += 96;
	}
	br_des_do_invIP(&l, &r);
	br_enc32be(buf, l);
	br_enc32be(buf + 4, r);
}

/* see inner.h */
void
br_des_ct_skey_expand(uint32_t *sk_exp,
	unsigned num_rounds, const uint32_t *skey)
{
	num_rounds <<= 4;
	while (num_rounds -- > 0) {
		uint32_t v, w0, w1, w2, w3;

		v = *skey ++;
		w0 = v & 0x11111111;
		w1 = (v >> 1) & 0x11111111;
		w2 = (v >> 2) & 0x11111111;
		w3 = (v >> 3) & 0x11111111;
		*sk_exp ++ = (w0 << 4) - w0;
		*sk_exp ++ = (w1 << 4) - w1;
		*sk_exp ++ = (w2 << 4) - w2;
		*sk_exp ++ = (w3 << 4) - w3;
		v = *skey ++;
		w0 = v & 0x11111111;
		w1 = (v >> 1) & 0x11111111;
		*sk_exp ++ = (w0 << 4) - w0;
		*sk_exp ++ = (w1 << 4) - w1;
	}
}