vmac.c 18.8 KB
Newer Older
1
/*
2 3 4 5
 * VMAC: Message Authentication Code using Universal Hashing
 *
 * Reference: https://tools.ietf.org/html/draft-krovetz-vmac-01
 *
6
 * Copyright (c) 2009, Intel Corporation.
7
 * Copyright (c) 2018, Google Inc.
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms and conditions of the GNU General Public License,
 * version 2, as published by the Free Software Foundation.
 *
 * This program is distributed in the hope it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along with
 * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
 * Place - Suite 330, Boston, MA 02111-1307 USA.
 */

23 24 25 26 27 28 29
/*
 * Derived from:
 *	VMAC and VHASH Implementation by Ted Krovetz ([email protected]) and Wei Dai.
 *	This implementation is herby placed in the public domain.
 *	The authors offers no warranty. Use at your own risk.
 *	Last modified: 17 APR 08, 1700 PDT
 */
30

31
#include <asm/unaligned.h>
32 33 34
#include <linux/init.h>
#include <linux/types.h>
#include <linux/crypto.h>
35
#include <linux/module.h>
36 37 38 39 40
#include <linux/scatterlist.h>
#include <asm/byteorder.h>
#include <crypto/scatterwalk.h>
#include <crypto/internal/hash.h>

41 42 43 44 45 46 47
/*
 * User definable settings.
 */
#define VMAC_TAG_LEN	64
#define VMAC_KEY_SIZE	128/* Must be 128, 192 or 256			*/
#define VMAC_KEY_LEN	(VMAC_KEY_SIZE/8)
#define VMAC_NHBYTES	128/* Must 2^i for any 3 < i < 13 Standard = 128*/
48
#define VMAC_NONCEBYTES	16
49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66

/* per-transform (per-key) context */
struct vmac_tfm_ctx {
	struct crypto_cipher *cipher;
	u64 nhkey[(VMAC_NHBYTES/8)+2*(VMAC_TAG_LEN/64-1)];
	u64 polykey[2*VMAC_TAG_LEN/64];
	u64 l3key[2*VMAC_TAG_LEN/64];
};

/* per-request context */
struct vmac_desc_ctx {
	union {
		u8 partial[VMAC_NHBYTES];	/* partial block */
		__le64 partial_words[VMAC_NHBYTES / 8];
	};
	unsigned int partial_size;	/* size of the partial block */
	bool first_block_processed;
	u64 polytmp[2*VMAC_TAG_LEN/64];	/* running total of L2-hash */
67 68 69 70 71
	union {
		u8 bytes[VMAC_NONCEBYTES];
		__be64 pads[VMAC_NONCEBYTES / 8];
	} nonce;
	unsigned int nonce_size; /* nonce bytes filled so far */
72 73
};

74 75 76 77
/*
 * Constants and masks
 */
#define UINT64_C(x) x##ULL
78 79 80 81 82
static const u64 p64   = UINT64_C(0xfffffffffffffeff);	/* 2^64 - 257 prime  */
static const u64 m62   = UINT64_C(0x3fffffffffffffff);	/* 62-bit mask       */
static const u64 m63   = UINT64_C(0x7fffffffffffffff);	/* 63-bit mask       */
static const u64 m64   = UINT64_C(0xffffffffffffffff);	/* 64-bit mask       */
static const u64 mpoly = UINT64_C(0x1fffffff1fffffff);	/* Poly key mask     */
83

84 85
#define pe64_to_cpup le64_to_cpup		/* Prefer little endian */

86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
#ifdef __LITTLE_ENDIAN
#define INDEX_HIGH 1
#define INDEX_LOW 0
#else
#define INDEX_HIGH 0
#define INDEX_LOW 1
#endif

/*
 * The following routines are used in this implementation. They are
 * written via macros to simulate zero-overhead call-by-reference.
 *
 * MUL64: 64x64->128-bit multiplication
 * PMUL64: assumes top bits cleared on inputs
 * ADD128: 128x128->128-bit addition
 */

#define ADD128(rh, rl, ih, il)						\
	do {								\
		u64 _il = (il);						\
		(rl) += (_il);						\
		if ((rl) < (_il))					\
			(rh)++;						\
		(rh) += (ih);						\
	} while (0)

#define MUL32(i1, i2)	((u64)(u32)(i1)*(u32)(i2))

#define PMUL64(rh, rl, i1, i2)	/* Assumes m doesn't overflow */	\
	do {								\
		u64 _i1 = (i1), _i2 = (i2);				\
		u64 m = MUL32(_i1, _i2>>32) + MUL32(_i1>>32, _i2);	\
		rh = MUL32(_i1>>32, _i2>>32);				\
		rl = MUL32(_i1, _i2);					\
		ADD128(rh, rl, (m >> 32), (m << 32));			\
	} while (0)

#define MUL64(rh, rl, i1, i2)						\
	do {								\
		u64 _i1 = (i1), _i2 = (i2);				\
		u64 m1 = MUL32(_i1, _i2>>32);				\
		u64 m2 = MUL32(_i1>>32, _i2);				\
		rh = MUL32(_i1>>32, _i2>>32);				\
		rl = MUL32(_i1, _i2);					\
		ADD128(rh, rl, (m1 >> 32), (m1 << 32));			\
		ADD128(rh, rl, (m2 >> 32), (m2 << 32));			\
	} while (0)

/*
 * For highest performance the L1 NH and L2 polynomial hashes should be
Lucas De Marchi's avatar
Lucas De Marchi committed
136
 * carefully implemented to take advantage of one's target architecture.
137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152
 * Here these two hash functions are defined multiple time; once for
 * 64-bit architectures, once for 32-bit SSE2 architectures, and once
 * for the rest (32-bit) architectures.
 * For each, nh_16 *must* be defined (works on multiples of 16 bytes).
 * Optionally, nh_vmac_nhbytes can be defined (for multiples of
 * VMAC_NHBYTES), and nh_16_2 and nh_vmac_nhbytes_2 (versions that do two
 * NH computations at once).
 */

#ifdef CONFIG_64BIT

#define nh_16(mp, kp, nw, rh, rl)					\
	do {								\
		int i; u64 th, tl;					\
		rh = rl = 0;						\
		for (i = 0; i < nw; i += 2) {				\
153 154
			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
155 156 157 158 159 160 161 162 163
			ADD128(rh, rl, th, tl);				\
		}							\
	} while (0)

#define nh_16_2(mp, kp, nw, rh, rl, rh1, rl1)				\
	do {								\
		int i; u64 th, tl;					\
		rh1 = rl1 = rh = rl = 0;				\
		for (i = 0; i < nw; i += 2) {				\
164 165
			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
166
			ADD128(rh, rl, th, tl);				\
167 168
			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2],	\
				pe64_to_cpup((mp)+i+1)+(kp)[i+3]);	\
169 170 171 172 173 174 175 176 177 178
			ADD128(rh1, rl1, th, tl);			\
		}							\
	} while (0)

#if (VMAC_NHBYTES >= 64) /* These versions do 64-bytes of message at a time */
#define nh_vmac_nhbytes(mp, kp, nw, rh, rl)				\
	do {								\
		int i; u64 th, tl;					\
		rh = rl = 0;						\
		for (i = 0; i < nw; i += 8) {				\
179 180
			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
181
			ADD128(rh, rl, th, tl);				\
182 183
			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2],	\
				pe64_to_cpup((mp)+i+3)+(kp)[i+3]);	\
184
			ADD128(rh, rl, th, tl);				\
185 186
			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4],	\
				pe64_to_cpup((mp)+i+5)+(kp)[i+5]);	\
187
			ADD128(rh, rl, th, tl);				\
188 189
			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6],	\
				pe64_to_cpup((mp)+i+7)+(kp)[i+7]);	\
190 191 192 193 194 195 196 197 198
			ADD128(rh, rl, th, tl);				\
		}							\
	} while (0)

#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh1, rl1)			\
	do {								\
		int i; u64 th, tl;					\
		rh1 = rl1 = rh = rl = 0;				\
		for (i = 0; i < nw; i += 8) {				\
199 200
			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
201
			ADD128(rh, rl, th, tl);				\
202 203
			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2],	\
				pe64_to_cpup((mp)+i+1)+(kp)[i+3]);	\
204
			ADD128(rh1, rl1, th, tl);			\
205 206
			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2],	\
				pe64_to_cpup((mp)+i+3)+(kp)[i+3]);	\
207
			ADD128(rh, rl, th, tl);				\
208 209
			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+4],	\
				pe64_to_cpup((mp)+i+3)+(kp)[i+5]);	\
210
			ADD128(rh1, rl1, th, tl);			\
211 212
			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4],	\
				pe64_to_cpup((mp)+i+5)+(kp)[i+5]);	\
213
			ADD128(rh, rl, th, tl);				\
214 215
			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+6],	\
				pe64_to_cpup((mp)+i+5)+(kp)[i+7]);	\
216
			ADD128(rh1, rl1, th, tl);			\
217 218
			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6],	\
				pe64_to_cpup((mp)+i+7)+(kp)[i+7]);	\
219
			ADD128(rh, rl, th, tl);				\
220 221
			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+8],	\
				pe64_to_cpup((mp)+i+7)+(kp)[i+9]);	\
222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258
			ADD128(rh1, rl1, th, tl);			\
		}							\
	} while (0)
#endif

#define poly_step(ah, al, kh, kl, mh, ml)				\
	do {								\
		u64 t1h, t1l, t2h, t2l, t3h, t3l, z = 0;		\
		/* compute ab*cd, put bd into result registers */	\
		PMUL64(t3h, t3l, al, kh);				\
		PMUL64(t2h, t2l, ah, kl);				\
		PMUL64(t1h, t1l, ah, 2*kh);				\
		PMUL64(ah, al, al, kl);					\
		/* add 2 * ac to result */				\
		ADD128(ah, al, t1h, t1l);				\
		/* add together ad + bc */				\
		ADD128(t2h, t2l, t3h, t3l);				\
		/* now (ah,al), (t2l,2*t2h) need summing */		\
		/* first add the high registers, carrying into t2h */	\
		ADD128(t2h, ah, z, t2l);				\
		/* double t2h and add top bit of ah */			\
		t2h = 2 * t2h + (ah >> 63);				\
		ah &= m63;						\
		/* now add the low registers */				\
		ADD128(ah, al, mh, ml);					\
		ADD128(ah, al, z, t2h);					\
	} while (0)

#else /* ! CONFIG_64BIT */

#ifndef nh_16
#define nh_16(mp, kp, nw, rh, rl)					\
	do {								\
		u64 t1, t2, m1, m2, t;					\
		int i;							\
		rh = rl = t = 0;					\
		for (i = 0; i < nw; i += 2)  {				\
259 260
			t1 = pe64_to_cpup(mp+i) + kp[i];		\
			t2 = pe64_to_cpup(mp+i+1) + kp[i+1];		\
261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
			m2 = MUL32(t1 >> 32, t2);			\
			m1 = MUL32(t1, t2 >> 32);			\
			ADD128(rh, rl, MUL32(t1 >> 32, t2 >> 32),	\
				MUL32(t1, t2));				\
			rh += (u64)(u32)(m1 >> 32)			\
				+ (u32)(m2 >> 32);			\
			t += (u64)(u32)m1 + (u32)m2;			\
		}							\
		ADD128(rh, rl, (t >> 32), (t << 32));			\
	} while (0)
#endif

static void poly_step_func(u64 *ahi, u64 *alo,
			const u64 *kh, const u64 *kl,
			const u64 *mh, const u64 *ml)
{
#define a0 (*(((u32 *)alo)+INDEX_LOW))
#define a1 (*(((u32 *)alo)+INDEX_HIGH))
#define a2 (*(((u32 *)ahi)+INDEX_LOW))
#define a3 (*(((u32 *)ahi)+INDEX_HIGH))
#define k0 (*(((u32 *)kl)+INDEX_LOW))
#define k1 (*(((u32 *)kl)+INDEX_HIGH))
#define k2 (*(((u32 *)kh)+INDEX_LOW))
#define k3 (*(((u32 *)kh)+INDEX_HIGH))

	u64 p, q, t;
	u32 t2;

	p = MUL32(a3, k3);
	p += p;
	p += *(u64 *)mh;
	p += MUL32(a0, k2);
	p += MUL32(a1, k1);
	p += MUL32(a2, k0);
	t = (u32)(p);
	p >>= 32;
	p += MUL32(a0, k3);
	p += MUL32(a1, k2);
	p += MUL32(a2, k1);
	p += MUL32(a3, k0);
	t |= ((u64)((u32)p & 0x7fffffff)) << 32;
	p >>= 31;
	p += (u64)(((u32 *)ml)[INDEX_LOW]);
	p += MUL32(a0, k0);
	q =  MUL32(a1, k3);
	q += MUL32(a2, k2);
	q += MUL32(a3, k1);
	q += q;
	p += q;
	t2 = (u32)(p);
	p >>= 32;
	p += (u64)(((u32 *)ml)[INDEX_HIGH]);
	p += MUL32(a0, k1);
	p += MUL32(a1, k0);
	q =  MUL32(a2, k3);
	q += MUL32(a3, k2);
	q += q;
	p += q;
	*(u64 *)(alo) = (p << 32) | t2;
	p >>= 32;
	*(u64 *)(ahi) = p + t;

#undef a0
#undef a1
#undef a2
#undef a3
#undef k0
#undef k1
#undef k2
#undef k3
}

#define poly_step(ah, al, kh, kl, mh, ml)				\
	poly_step_func(&(ah), &(al), &(kh), &(kl), &(mh), &(ml))

#endif  /* end of specialized NH and poly definitions */

/* At least nh_16 is defined. Defined others as needed here */
#ifndef nh_16_2
#define nh_16_2(mp, kp, nw, rh, rl, rh2, rl2)				\
	do { 								\
		nh_16(mp, kp, nw, rh, rl);				\
		nh_16(mp, ((kp)+2), nw, rh2, rl2);			\
	} while (0)
#endif
#ifndef nh_vmac_nhbytes
#define nh_vmac_nhbytes(mp, kp, nw, rh, rl)				\
	nh_16(mp, kp, nw, rh, rl)
#endif
#ifndef nh_vmac_nhbytes_2
#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh2, rl2)			\
	do {								\
		nh_vmac_nhbytes(mp, kp, nw, rh, rl);			\
		nh_vmac_nhbytes(mp, ((kp)+2), nw, rh2, rl2);		\
	} while (0)
#endif

358
static u64 l3hash(u64 p1, u64 p2, u64 k1, u64 k2, u64 len)
359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396
{
	u64 rh, rl, t, z = 0;

	/* fully reduce (p1,p2)+(len,0) mod p127 */
	t = p1 >> 63;
	p1 &= m63;
	ADD128(p1, p2, len, t);
	/* At this point, (p1,p2) is at most 2^127+(len<<64) */
	t = (p1 > m63) + ((p1 == m63) && (p2 == m64));
	ADD128(p1, p2, z, t);
	p1 &= m63;

	/* compute (p1,p2)/(2^64-2^32) and (p1,p2)%(2^64-2^32) */
	t = p1 + (p2 >> 32);
	t += (t >> 32);
	t += (u32)t > 0xfffffffeu;
	p1 += (t >> 32);
	p2 += (p1 << 32);

	/* compute (p1+k1)%p64 and (p2+k2)%p64 */
	p1 += k1;
	p1 += (0 - (p1 < k1)) & 257;
	p2 += k2;
	p2 += (0 - (p2 < k2)) & 257;

	/* compute (p1+k1)*(p2+k2)%p64 */
	MUL64(rh, rl, p1, p2);
	t = rh >> 56;
	ADD128(t, rl, z, rh);
	rh <<= 8;
	ADD128(t, rl, z, rh);
	t += t << 8;
	rl += t;
	rl += (0 - (rl < t)) & 257;
	rl += (0 - (rl > p64-1)) & 257;
	return rl;
}

397 398 399 400
/* L1 and L2-hash one or more VMAC_NHBYTES-byte blocks */
static void vhash_blocks(const struct vmac_tfm_ctx *tctx,
			 struct vmac_desc_ctx *dctx,
			 const __le64 *mptr, unsigned int blocks)
401
{
402 403 404 405 406 407 408 409 410
	const u64 *kptr = tctx->nhkey;
	const u64 pkh = tctx->polykey[0];
	const u64 pkl = tctx->polykey[1];
	u64 ch = dctx->polytmp[0];
	u64 cl = dctx->polytmp[1];
	u64 rh, rl;

	if (!dctx->first_block_processed) {
		dctx->first_block_processed = true;
411 412 413 414
		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
		rh &= m62;
		ADD128(ch, cl, rh, rl);
		mptr += (VMAC_NHBYTES/sizeof(u64));
415
		blocks--;
416 417
	}

418
	while (blocks--) {
419 420 421 422 423 424
		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
		rh &= m62;
		poly_step(ch, cl, pkh, pkl, rh, rl);
		mptr += (VMAC_NHBYTES/sizeof(u64));
	}

425 426
	dctx->polytmp[0] = ch;
	dctx->polytmp[1] = cl;
427 428
}

429 430
static int vmac_setkey(struct crypto_shash *tfm,
		       const u8 *key, unsigned int keylen)
431
{
432 433 434 435 436
	struct vmac_tfm_ctx *tctx = crypto_shash_ctx(tfm);
	__be64 out[2];
	u8 in[16] = { 0 };
	unsigned int i;
	int err;
437

438 439 440
	if (keylen != VMAC_KEY_LEN) {
		crypto_shash_set_flags(tfm, CRYPTO_TFM_RES_BAD_KEY_LEN);
		return -EINVAL;
441 442
	}

443
	err = crypto_cipher_setkey(tctx->cipher, key, keylen);
444 445 446 447
	if (err)
		return err;

	/* Fill nh key */
448 449 450 451 452 453
	in[0] = 0x80;
	for (i = 0; i < ARRAY_SIZE(tctx->nhkey); i += 2) {
		crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
		tctx->nhkey[i] = be64_to_cpu(out[0]);
		tctx->nhkey[i+1] = be64_to_cpu(out[1]);
		in[15]++;
454 455 456
	}

	/* Fill poly key */
457 458 459 460 461 462 463
	in[0] = 0xC0;
	in[15] = 0;
	for (i = 0; i < ARRAY_SIZE(tctx->polykey); i += 2) {
		crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
		tctx->polykey[i] = be64_to_cpu(out[0]) & mpoly;
		tctx->polykey[i+1] = be64_to_cpu(out[1]) & mpoly;
		in[15]++;
464 465 466
	}

	/* Fill ip key */
467 468 469
	in[0] = 0xE0;
	in[15] = 0;
	for (i = 0; i < ARRAY_SIZE(tctx->l3key); i += 2) {
470
		do {
471 472 473 474 475
			crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
			tctx->l3key[i] = be64_to_cpu(out[0]);
			tctx->l3key[i+1] = be64_to_cpu(out[1]);
			in[15]++;
		} while (tctx->l3key[i] >= p64 || tctx->l3key[i+1] >= p64);
476 477
	}

478
	return 0;
479 480
}

481
static int vmac_init(struct shash_desc *desc)
482
{
483 484
	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
485

486 487 488
	dctx->partial_size = 0;
	dctx->first_block_processed = false;
	memcpy(dctx->polytmp, tctx->polykey, sizeof(dctx->polytmp));
489 490 491 492
	dctx->nonce_size = 0;
	return 0;
}

493
static int vmac_update(struct shash_desc *desc, const u8 *p, unsigned int len)
494
{
495 496 497 498
	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
	unsigned int n;

499 500 501 502 503 504 505 506 507
	/* Nonce is passed as first VMAC_NONCEBYTES bytes of data */
	if (dctx->nonce_size < VMAC_NONCEBYTES) {
		n = min(len, VMAC_NONCEBYTES - dctx->nonce_size);
		memcpy(&dctx->nonce.bytes[dctx->nonce_size], p, n);
		dctx->nonce_size += n;
		p += n;
		len -= n;
	}

508 509 510 511 512 513 514 515 516 517 518
	if (dctx->partial_size) {
		n = min(len, VMAC_NHBYTES - dctx->partial_size);
		memcpy(&dctx->partial[dctx->partial_size], p, n);
		dctx->partial_size += n;
		p += n;
		len -= n;
		if (dctx->partial_size == VMAC_NHBYTES) {
			vhash_blocks(tctx, dctx, dctx->partial_words, 1);
			dctx->partial_size = 0;
		}
	}
519

520 521 522 523 524 525
	if (len >= VMAC_NHBYTES) {
		n = round_down(len, VMAC_NHBYTES);
		/* TODO: 'p' may be misaligned here */
		vhash_blocks(tctx, dctx, (const __le64 *)p, n / VMAC_NHBYTES);
		p += n;
		len -= n;
526 527
	}

528 529 530 531
	if (len) {
		memcpy(dctx->partial, p, len);
		dctx->partial_size = len;
	}
532 533 534 535

	return 0;
}

536 537
static u64 vhash_final(const struct vmac_tfm_ctx *tctx,
		       struct vmac_desc_ctx *dctx)
538
{
539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556
	unsigned int partial = dctx->partial_size;
	u64 ch = dctx->polytmp[0];
	u64 cl = dctx->polytmp[1];

	/* L1 and L2-hash the final block if needed */
	if (partial) {
		/* Zero-pad to next 128-bit boundary */
		unsigned int n = round_up(partial, 16);
		u64 rh, rl;

		memset(&dctx->partial[partial], 0, n - partial);
		nh_16(dctx->partial_words, tctx->nhkey, n / 8, rh, rl);
		rh &= m62;
		if (dctx->first_block_processed)
			poly_step(ch, cl, tctx->polykey[0], tctx->polykey[1],
				  rh, rl);
		else
			ADD128(ch, cl, rh, rl);
557
	}
558 559 560 561 562

	/* L3-hash the 128-bit output of L2-hash */
	return l3hash(ch, cl, tctx->l3key[0], tctx->l3key[1], partial * 8);
}

563
static int vmac_final(struct shash_desc *desc, u8 *out)
564 565 566 567 568 569
{
	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
	int index;
	u64 hash, pad;

570 571 572 573 574 575 576 577 578 579 580 581
	if (dctx->nonce_size != VMAC_NONCEBYTES)
		return -EINVAL;

	/*
	 * The VMAC specification requires a nonce at least 1 bit shorter than
	 * the block cipher's block length, so we actually only accept a 127-bit
	 * nonce.  We define the unused bit to be the first one and require that
	 * it be 0, so the needed prepending of a 0 bit is implicit.
	 */
	if (dctx->nonce.bytes[0] & 0x80)
		return -EINVAL;

582 583 584 585
	/* Finish calculating the VHASH of the message */
	hash = vhash_final(tctx, dctx);

	/* Generate pseudorandom pad by encrypting the nonce */
586 587 588 589 590 591
	BUILD_BUG_ON(VMAC_NONCEBYTES != 2 * (VMAC_TAG_LEN / 8));
	index = dctx->nonce.bytes[VMAC_NONCEBYTES - 1] & 1;
	dctx->nonce.bytes[VMAC_NONCEBYTES - 1] &= ~1;
	crypto_cipher_encrypt_one(tctx->cipher, dctx->nonce.bytes,
				  dctx->nonce.bytes);
	pad = be64_to_cpu(dctx->nonce.pads[index]);
592 593

	/* The VMAC is the sum of VHASH and the pseudorandom pad */
594
	put_unaligned_be64(hash + pad, out);
595 596 597 598 599
	return 0;
}

static int vmac_init_tfm(struct crypto_tfm *tfm)
{
600
	struct crypto_instance *inst = crypto_tfm_alg_instance(tfm);
601
	struct crypto_spawn *spawn = crypto_instance_ctx(inst);
602 603
	struct vmac_tfm_ctx *tctx = crypto_tfm_ctx(tfm);
	struct crypto_cipher *cipher;
604 605 606 607 608

	cipher = crypto_spawn_cipher(spawn);
	if (IS_ERR(cipher))
		return PTR_ERR(cipher);

609
	tctx->cipher = cipher;
610 611 612 613 614
	return 0;
}

static void vmac_exit_tfm(struct crypto_tfm *tfm)
{
615 616 617
	struct vmac_tfm_ctx *tctx = crypto_tfm_ctx(tfm);

	crypto_free_cipher(tctx->cipher);
618 619
}

620
static int vmac_create(struct crypto_template *tmpl, struct rtattr **tb)
621 622 623 624 625 626 627 628 629 630 631 632 633 634
{
	struct shash_instance *inst;
	struct crypto_alg *alg;
	int err;

	err = crypto_check_attr_type(tb, CRYPTO_ALG_TYPE_SHASH);
	if (err)
		return err;

	alg = crypto_get_attr_alg(tb, CRYPTO_ALG_TYPE_CIPHER,
			CRYPTO_ALG_TYPE_MASK);
	if (IS_ERR(alg))
		return PTR_ERR(alg);

635
	err = -EINVAL;
636
	if (alg->cra_blocksize != VMAC_NONCEBYTES)
637 638
		goto out_put_alg;

639
	inst = shash_alloc_instance(tmpl->name, alg);
640 641 642 643 644 645 646 647 648 649 650 651 652 653
	err = PTR_ERR(inst);
	if (IS_ERR(inst))
		goto out_put_alg;

	err = crypto_init_spawn(shash_instance_ctx(inst), alg,
			shash_crypto_instance(inst),
			CRYPTO_ALG_TYPE_MASK);
	if (err)
		goto out_free_inst;

	inst->alg.base.cra_priority = alg->cra_priority;
	inst->alg.base.cra_blocksize = alg->cra_blocksize;
	inst->alg.base.cra_alignmask = alg->cra_alignmask;

654
	inst->alg.base.cra_ctxsize = sizeof(struct vmac_tfm_ctx);
655 656 657
	inst->alg.base.cra_init = vmac_init_tfm;
	inst->alg.base.cra_exit = vmac_exit_tfm;

658 659
	inst->alg.descsize = sizeof(struct vmac_desc_ctx);
	inst->alg.digestsize = VMAC_TAG_LEN / 8;
660
	inst->alg.init = vmac_init;
661
	inst->alg.update = vmac_update;
662
	inst->alg.final = vmac_final;
663 664 665 666 667 668 669 670 671 672 673 674 675
	inst->alg.setkey = vmac_setkey;

	err = shash_register_instance(tmpl, inst);
	if (err) {
out_free_inst:
		shash_free_instance(shash_crypto_instance(inst));
	}

out_put_alg:
	crypto_mod_put(alg);
	return err;
}

676 677
static struct crypto_template vmac64_tmpl = {
	.name = "vmac64",
678
	.create = vmac_create,
679 680 681 682
	.free = shash_free_instance,
	.module = THIS_MODULE,
};

683 684
static int __init vmac_module_init(void)
{
685
	return crypto_register_template(&vmac64_tmpl);
686 687 688 689
}

static void __exit vmac_module_exit(void)
{
690
	crypto_unregister_template(&vmac64_tmpl);
691 692 693 694 695 696 697
}

module_init(vmac_module_init);
module_exit(vmac_module_exit);

MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("VMAC hash algorithm");
698
MODULE_ALIAS_CRYPTO("vmac64");