Linux Audio

Check our new training course

Loading...
v5.9
  1/*
  2 * VMAC: Message Authentication Code using Universal Hashing
  3 *
  4 * Reference: https://tools.ietf.org/html/draft-krovetz-vmac-01
  5 *
  6 * Copyright (c) 2009, Intel Corporation.
  7 * Copyright (c) 2018, Google Inc.
  8 *
  9 * This program is free software; you can redistribute it and/or modify it
 10 * under the terms and conditions of the GNU General Public License,
 11 * version 2, as published by the Free Software Foundation.
 12 *
 13 * This program is distributed in the hope it will be useful, but WITHOUT
 14 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 15 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 16 * more details.
 17 *
 18 * You should have received a copy of the GNU General Public License along with
 19 * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
 20 * Place - Suite 330, Boston, MA 02111-1307 USA.
 21 */
 22
 23/*
 24 * Derived from:
 25 *	VMAC and VHASH Implementation by Ted Krovetz (tdk@acm.org) and Wei Dai.
 26 *	This implementation is herby placed in the public domain.
 27 *	The authors offers no warranty. Use at your own risk.
 28 *	Last modified: 17 APR 08, 1700 PDT
 29 */
 30
 31#include <asm/unaligned.h>
 32#include <linux/init.h>
 33#include <linux/types.h>
 34#include <linux/crypto.h>
 35#include <linux/module.h>
 36#include <linux/scatterlist.h>
 37#include <asm/byteorder.h>
 38#include <crypto/scatterwalk.h>
 
 39#include <crypto/internal/hash.h>
 40
 41/*
 42 * User definable settings.
 43 */
 44#define VMAC_TAG_LEN	64
 45#define VMAC_KEY_SIZE	128/* Must be 128, 192 or 256			*/
 46#define VMAC_KEY_LEN	(VMAC_KEY_SIZE/8)
 47#define VMAC_NHBYTES	128/* Must 2^i for any 3 < i < 13 Standard = 128*/
 48#define VMAC_NONCEBYTES	16
 49
 50/* per-transform (per-key) context */
 51struct vmac_tfm_ctx {
 52	struct crypto_cipher *cipher;
 53	u64 nhkey[(VMAC_NHBYTES/8)+2*(VMAC_TAG_LEN/64-1)];
 54	u64 polykey[2*VMAC_TAG_LEN/64];
 55	u64 l3key[2*VMAC_TAG_LEN/64];
 56};
 57
 58/* per-request context */
 59struct vmac_desc_ctx {
 60	union {
 61		u8 partial[VMAC_NHBYTES];	/* partial block */
 62		__le64 partial_words[VMAC_NHBYTES / 8];
 63	};
 64	unsigned int partial_size;	/* size of the partial block */
 65	bool first_block_processed;
 66	u64 polytmp[2*VMAC_TAG_LEN/64];	/* running total of L2-hash */
 67	union {
 68		u8 bytes[VMAC_NONCEBYTES];
 69		__be64 pads[VMAC_NONCEBYTES / 8];
 70	} nonce;
 71	unsigned int nonce_size; /* nonce bytes filled so far */
 72};
 73
 74/*
 75 * Constants and masks
 76 */
 77#define UINT64_C(x) x##ULL
 78static const u64 p64   = UINT64_C(0xfffffffffffffeff);	/* 2^64 - 257 prime  */
 79static const u64 m62   = UINT64_C(0x3fffffffffffffff);	/* 62-bit mask       */
 80static const u64 m63   = UINT64_C(0x7fffffffffffffff);	/* 63-bit mask       */
 81static const u64 m64   = UINT64_C(0xffffffffffffffff);	/* 64-bit mask       */
 82static const u64 mpoly = UINT64_C(0x1fffffff1fffffff);	/* Poly key mask     */
 83
 84#define pe64_to_cpup le64_to_cpup		/* Prefer little endian */
 85
 86#ifdef __LITTLE_ENDIAN
 87#define INDEX_HIGH 1
 88#define INDEX_LOW 0
 89#else
 90#define INDEX_HIGH 0
 91#define INDEX_LOW 1
 92#endif
 93
 94/*
 95 * The following routines are used in this implementation. They are
 96 * written via macros to simulate zero-overhead call-by-reference.
 97 *
 98 * MUL64: 64x64->128-bit multiplication
 99 * PMUL64: assumes top bits cleared on inputs
100 * ADD128: 128x128->128-bit addition
101 */
102
103#define ADD128(rh, rl, ih, il)						\
104	do {								\
105		u64 _il = (il);						\
106		(rl) += (_il);						\
107		if ((rl) < (_il))					\
108			(rh)++;						\
109		(rh) += (ih);						\
110	} while (0)
111
112#define MUL32(i1, i2)	((u64)(u32)(i1)*(u32)(i2))
113
114#define PMUL64(rh, rl, i1, i2)	/* Assumes m doesn't overflow */	\
115	do {								\
116		u64 _i1 = (i1), _i2 = (i2);				\
117		u64 m = MUL32(_i1, _i2>>32) + MUL32(_i1>>32, _i2);	\
118		rh = MUL32(_i1>>32, _i2>>32);				\
119		rl = MUL32(_i1, _i2);					\
120		ADD128(rh, rl, (m >> 32), (m << 32));			\
121	} while (0)
122
123#define MUL64(rh, rl, i1, i2)						\
124	do {								\
125		u64 _i1 = (i1), _i2 = (i2);				\
126		u64 m1 = MUL32(_i1, _i2>>32);				\
127		u64 m2 = MUL32(_i1>>32, _i2);				\
128		rh = MUL32(_i1>>32, _i2>>32);				\
129		rl = MUL32(_i1, _i2);					\
130		ADD128(rh, rl, (m1 >> 32), (m1 << 32));			\
131		ADD128(rh, rl, (m2 >> 32), (m2 << 32));			\
132	} while (0)
133
134/*
135 * For highest performance the L1 NH and L2 polynomial hashes should be
136 * carefully implemented to take advantage of one's target architecture.
137 * Here these two hash functions are defined multiple time; once for
138 * 64-bit architectures, once for 32-bit SSE2 architectures, and once
139 * for the rest (32-bit) architectures.
140 * For each, nh_16 *must* be defined (works on multiples of 16 bytes).
141 * Optionally, nh_vmac_nhbytes can be defined (for multiples of
142 * VMAC_NHBYTES), and nh_16_2 and nh_vmac_nhbytes_2 (versions that do two
143 * NH computations at once).
144 */
145
146#ifdef CONFIG_64BIT
147
148#define nh_16(mp, kp, nw, rh, rl)					\
149	do {								\
150		int i; u64 th, tl;					\
151		rh = rl = 0;						\
152		for (i = 0; i < nw; i += 2) {				\
153			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
154				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
155			ADD128(rh, rl, th, tl);				\
156		}							\
157	} while (0)
158
159#define nh_16_2(mp, kp, nw, rh, rl, rh1, rl1)				\
160	do {								\
161		int i; u64 th, tl;					\
162		rh1 = rl1 = rh = rl = 0;				\
163		for (i = 0; i < nw; i += 2) {				\
164			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
165				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
166			ADD128(rh, rl, th, tl);				\
167			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2],	\
168				pe64_to_cpup((mp)+i+1)+(kp)[i+3]);	\
169			ADD128(rh1, rl1, th, tl);			\
170		}							\
171	} while (0)
172
173#if (VMAC_NHBYTES >= 64) /* These versions do 64-bytes of message at a time */
174#define nh_vmac_nhbytes(mp, kp, nw, rh, rl)				\
175	do {								\
176		int i; u64 th, tl;					\
177		rh = rl = 0;						\
178		for (i = 0; i < nw; i += 8) {				\
179			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
180				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
181			ADD128(rh, rl, th, tl);				\
182			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2],	\
183				pe64_to_cpup((mp)+i+3)+(kp)[i+3]);	\
184			ADD128(rh, rl, th, tl);				\
185			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4],	\
186				pe64_to_cpup((mp)+i+5)+(kp)[i+5]);	\
187			ADD128(rh, rl, th, tl);				\
188			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6],	\
189				pe64_to_cpup((mp)+i+7)+(kp)[i+7]);	\
190			ADD128(rh, rl, th, tl);				\
191		}							\
192	} while (0)
193
194#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh1, rl1)			\
195	do {								\
196		int i; u64 th, tl;					\
197		rh1 = rl1 = rh = rl = 0;				\
198		for (i = 0; i < nw; i += 8) {				\
199			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
200				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
201			ADD128(rh, rl, th, tl);				\
202			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2],	\
203				pe64_to_cpup((mp)+i+1)+(kp)[i+3]);	\
204			ADD128(rh1, rl1, th, tl);			\
205			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2],	\
206				pe64_to_cpup((mp)+i+3)+(kp)[i+3]);	\
207			ADD128(rh, rl, th, tl);				\
208			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+4],	\
209				pe64_to_cpup((mp)+i+3)+(kp)[i+5]);	\
210			ADD128(rh1, rl1, th, tl);			\
211			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4],	\
212				pe64_to_cpup((mp)+i+5)+(kp)[i+5]);	\
213			ADD128(rh, rl, th, tl);				\
214			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+6],	\
215				pe64_to_cpup((mp)+i+5)+(kp)[i+7]);	\
216			ADD128(rh1, rl1, th, tl);			\
217			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6],	\
218				pe64_to_cpup((mp)+i+7)+(kp)[i+7]);	\
219			ADD128(rh, rl, th, tl);				\
220			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+8],	\
221				pe64_to_cpup((mp)+i+7)+(kp)[i+9]);	\
222			ADD128(rh1, rl1, th, tl);			\
223		}							\
224	} while (0)
225#endif
226
227#define poly_step(ah, al, kh, kl, mh, ml)				\
228	do {								\
229		u64 t1h, t1l, t2h, t2l, t3h, t3l, z = 0;		\
230		/* compute ab*cd, put bd into result registers */	\
231		PMUL64(t3h, t3l, al, kh);				\
232		PMUL64(t2h, t2l, ah, kl);				\
233		PMUL64(t1h, t1l, ah, 2*kh);				\
234		PMUL64(ah, al, al, kl);					\
235		/* add 2 * ac to result */				\
236		ADD128(ah, al, t1h, t1l);				\
237		/* add together ad + bc */				\
238		ADD128(t2h, t2l, t3h, t3l);				\
239		/* now (ah,al), (t2l,2*t2h) need summing */		\
240		/* first add the high registers, carrying into t2h */	\
241		ADD128(t2h, ah, z, t2l);				\
242		/* double t2h and add top bit of ah */			\
243		t2h = 2 * t2h + (ah >> 63);				\
244		ah &= m63;						\
245		/* now add the low registers */				\
246		ADD128(ah, al, mh, ml);					\
247		ADD128(ah, al, z, t2h);					\
248	} while (0)
249
250#else /* ! CONFIG_64BIT */
251
252#ifndef nh_16
253#define nh_16(mp, kp, nw, rh, rl)					\
254	do {								\
255		u64 t1, t2, m1, m2, t;					\
256		int i;							\
257		rh = rl = t = 0;					\
258		for (i = 0; i < nw; i += 2)  {				\
259			t1 = pe64_to_cpup(mp+i) + kp[i];		\
260			t2 = pe64_to_cpup(mp+i+1) + kp[i+1];		\
261			m2 = MUL32(t1 >> 32, t2);			\
262			m1 = MUL32(t1, t2 >> 32);			\
263			ADD128(rh, rl, MUL32(t1 >> 32, t2 >> 32),	\
264				MUL32(t1, t2));				\
265			rh += (u64)(u32)(m1 >> 32)			\
266				+ (u32)(m2 >> 32);			\
267			t += (u64)(u32)m1 + (u32)m2;			\
268		}							\
269		ADD128(rh, rl, (t >> 32), (t << 32));			\
270	} while (0)
271#endif
272
273static void poly_step_func(u64 *ahi, u64 *alo,
274			const u64 *kh, const u64 *kl,
275			const u64 *mh, const u64 *ml)
276{
277#define a0 (*(((u32 *)alo)+INDEX_LOW))
278#define a1 (*(((u32 *)alo)+INDEX_HIGH))
279#define a2 (*(((u32 *)ahi)+INDEX_LOW))
280#define a3 (*(((u32 *)ahi)+INDEX_HIGH))
281#define k0 (*(((u32 *)kl)+INDEX_LOW))
282#define k1 (*(((u32 *)kl)+INDEX_HIGH))
283#define k2 (*(((u32 *)kh)+INDEX_LOW))
284#define k3 (*(((u32 *)kh)+INDEX_HIGH))
285
286	u64 p, q, t;
287	u32 t2;
288
289	p = MUL32(a3, k3);
290	p += p;
291	p += *(u64 *)mh;
292	p += MUL32(a0, k2);
293	p += MUL32(a1, k1);
294	p += MUL32(a2, k0);
295	t = (u32)(p);
296	p >>= 32;
297	p += MUL32(a0, k3);
298	p += MUL32(a1, k2);
299	p += MUL32(a2, k1);
300	p += MUL32(a3, k0);
301	t |= ((u64)((u32)p & 0x7fffffff)) << 32;
302	p >>= 31;
303	p += (u64)(((u32 *)ml)[INDEX_LOW]);
304	p += MUL32(a0, k0);
305	q =  MUL32(a1, k3);
306	q += MUL32(a2, k2);
307	q += MUL32(a3, k1);
308	q += q;
309	p += q;
310	t2 = (u32)(p);
311	p >>= 32;
312	p += (u64)(((u32 *)ml)[INDEX_HIGH]);
313	p += MUL32(a0, k1);
314	p += MUL32(a1, k0);
315	q =  MUL32(a2, k3);
316	q += MUL32(a3, k2);
317	q += q;
318	p += q;
319	*(u64 *)(alo) = (p << 32) | t2;
320	p >>= 32;
321	*(u64 *)(ahi) = p + t;
322
323#undef a0
324#undef a1
325#undef a2
326#undef a3
327#undef k0
328#undef k1
329#undef k2
330#undef k3
331}
332
333#define poly_step(ah, al, kh, kl, mh, ml)				\
334	poly_step_func(&(ah), &(al), &(kh), &(kl), &(mh), &(ml))
335
336#endif  /* end of specialized NH and poly definitions */
337
338/* At least nh_16 is defined. Defined others as needed here */
339#ifndef nh_16_2
340#define nh_16_2(mp, kp, nw, rh, rl, rh2, rl2)				\
341	do { 								\
342		nh_16(mp, kp, nw, rh, rl);				\
343		nh_16(mp, ((kp)+2), nw, rh2, rl2);			\
344	} while (0)
345#endif
346#ifndef nh_vmac_nhbytes
347#define nh_vmac_nhbytes(mp, kp, nw, rh, rl)				\
348	nh_16(mp, kp, nw, rh, rl)
349#endif
350#ifndef nh_vmac_nhbytes_2
351#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh2, rl2)			\
352	do {								\
353		nh_vmac_nhbytes(mp, kp, nw, rh, rl);			\
354		nh_vmac_nhbytes(mp, ((kp)+2), nw, rh2, rl2);		\
355	} while (0)
356#endif
357
 
 
 
 
 
 
 
358static u64 l3hash(u64 p1, u64 p2, u64 k1, u64 k2, u64 len)
359{
360	u64 rh, rl, t, z = 0;
361
362	/* fully reduce (p1,p2)+(len,0) mod p127 */
363	t = p1 >> 63;
364	p1 &= m63;
365	ADD128(p1, p2, len, t);
366	/* At this point, (p1,p2) is at most 2^127+(len<<64) */
367	t = (p1 > m63) + ((p1 == m63) && (p2 == m64));
368	ADD128(p1, p2, z, t);
369	p1 &= m63;
370
371	/* compute (p1,p2)/(2^64-2^32) and (p1,p2)%(2^64-2^32) */
372	t = p1 + (p2 >> 32);
373	t += (t >> 32);
374	t += (u32)t > 0xfffffffeu;
375	p1 += (t >> 32);
376	p2 += (p1 << 32);
377
378	/* compute (p1+k1)%p64 and (p2+k2)%p64 */
379	p1 += k1;
380	p1 += (0 - (p1 < k1)) & 257;
381	p2 += k2;
382	p2 += (0 - (p2 < k2)) & 257;
383
384	/* compute (p1+k1)*(p2+k2)%p64 */
385	MUL64(rh, rl, p1, p2);
386	t = rh >> 56;
387	ADD128(t, rl, z, rh);
388	rh <<= 8;
389	ADD128(t, rl, z, rh);
390	t += t << 8;
391	rl += t;
392	rl += (0 - (rl < t)) & 257;
393	rl += (0 - (rl > p64-1)) & 257;
394	return rl;
395}
396
397/* L1 and L2-hash one or more VMAC_NHBYTES-byte blocks */
398static void vhash_blocks(const struct vmac_tfm_ctx *tctx,
399			 struct vmac_desc_ctx *dctx,
400			 const __le64 *mptr, unsigned int blocks)
401{
402	const u64 *kptr = tctx->nhkey;
403	const u64 pkh = tctx->polykey[0];
404	const u64 pkl = tctx->polykey[1];
405	u64 ch = dctx->polytmp[0];
406	u64 cl = dctx->polytmp[1];
407	u64 rh, rl;
 
 
 
 
 
 
 
408
409	if (!dctx->first_block_processed) {
410		dctx->first_block_processed = true;
 
 
 
411		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
412		rh &= m62;
413		ADD128(ch, cl, rh, rl);
414		mptr += (VMAC_NHBYTES/sizeof(u64));
415		blocks--;
416	}
417
418	while (blocks--) {
419		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
420		rh &= m62;
421		poly_step(ch, cl, pkh, pkl, rh, rl);
422		mptr += (VMAC_NHBYTES/sizeof(u64));
423	}
424
425	dctx->polytmp[0] = ch;
426	dctx->polytmp[1] = cl;
427}
428
429static int vmac_setkey(struct crypto_shash *tfm,
430		       const u8 *key, unsigned int keylen)
431{
432	struct vmac_tfm_ctx *tctx = crypto_shash_ctx(tfm);
433	__be64 out[2];
434	u8 in[16] = { 0 };
435	unsigned int i;
436	int err;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
437
438	if (keylen != VMAC_KEY_LEN)
439		return -EINVAL;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
440
441	err = crypto_cipher_setkey(tctx->cipher, key, keylen);
 
 
 
 
 
 
 
 
 
 
 
 
 
442	if (err)
443		return err;
444
445	/* Fill nh key */
446	in[0] = 0x80;
447	for (i = 0; i < ARRAY_SIZE(tctx->nhkey); i += 2) {
448		crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
449		tctx->nhkey[i] = be64_to_cpu(out[0]);
450		tctx->nhkey[i+1] = be64_to_cpu(out[1]);
451		in[15]++;
 
452	}
453
454	/* Fill poly key */
455	in[0] = 0xC0;
456	in[15] = 0;
457	for (i = 0; i < ARRAY_SIZE(tctx->polykey); i += 2) {
458		crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
459		tctx->polykey[i] = be64_to_cpu(out[0]) & mpoly;
460		tctx->polykey[i+1] = be64_to_cpu(out[1]) & mpoly;
461		in[15]++;
 
 
 
 
 
462	}
463
464	/* Fill ip key */
465	in[0] = 0xE0;
466	in[15] = 0;
467	for (i = 0; i < ARRAY_SIZE(tctx->l3key); i += 2) {
468		do {
469			crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
470			tctx->l3key[i] = be64_to_cpu(out[0]);
471			tctx->l3key[i+1] = be64_to_cpu(out[1]);
472			in[15]++;
473		} while (tctx->l3key[i] >= p64 || tctx->l3key[i+1] >= p64);
 
 
474	}
475
476	return 0;
 
 
 
 
 
477}
478
479static int vmac_init(struct shash_desc *desc)
 
480{
481	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
482	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
483
484	dctx->partial_size = 0;
485	dctx->first_block_processed = false;
486	memcpy(dctx->polytmp, tctx->polykey, sizeof(dctx->polytmp));
487	dctx->nonce_size = 0;
488	return 0;
 
489}
490
491static int vmac_update(struct shash_desc *desc, const u8 *p, unsigned int len)
492{
493	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
494	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
495	unsigned int n;
496
497	/* Nonce is passed as first VMAC_NONCEBYTES bytes of data */
498	if (dctx->nonce_size < VMAC_NONCEBYTES) {
499		n = min(len, VMAC_NONCEBYTES - dctx->nonce_size);
500		memcpy(&dctx->nonce.bytes[dctx->nonce_size], p, n);
501		dctx->nonce_size += n;
502		p += n;
503		len -= n;
504	}
505
506	if (dctx->partial_size) {
507		n = min(len, VMAC_NHBYTES - dctx->partial_size);
508		memcpy(&dctx->partial[dctx->partial_size], p, n);
509		dctx->partial_size += n;
510		p += n;
511		len -= n;
512		if (dctx->partial_size == VMAC_NHBYTES) {
513			vhash_blocks(tctx, dctx, dctx->partial_words, 1);
514			dctx->partial_size = 0;
515		}
516	}
517
518	if (len >= VMAC_NHBYTES) {
519		n = round_down(len, VMAC_NHBYTES);
520		/* TODO: 'p' may be misaligned here */
521		vhash_blocks(tctx, dctx, (const __le64 *)p, n / VMAC_NHBYTES);
522		p += n;
523		len -= n;
524	}
525
526	if (len) {
527		memcpy(dctx->partial, p, len);
528		dctx->partial_size = len;
529	}
530
531	return 0;
532}
533
534static u64 vhash_final(const struct vmac_tfm_ctx *tctx,
535		       struct vmac_desc_ctx *dctx)
536{
537	unsigned int partial = dctx->partial_size;
538	u64 ch = dctx->polytmp[0];
539	u64 cl = dctx->polytmp[1];
540
541	/* L1 and L2-hash the final block if needed */
542	if (partial) {
543		/* Zero-pad to next 128-bit boundary */
544		unsigned int n = round_up(partial, 16);
545		u64 rh, rl;
546
547		memset(&dctx->partial[partial], 0, n - partial);
548		nh_16(dctx->partial_words, tctx->nhkey, n / 8, rh, rl);
549		rh &= m62;
550		if (dctx->first_block_processed)
551			poly_step(ch, cl, tctx->polykey[0], tctx->polykey[1],
552				  rh, rl);
553		else
554			ADD128(ch, cl, rh, rl);
555	}
556
557	/* L3-hash the 128-bit output of L2-hash */
558	return l3hash(ch, cl, tctx->l3key[0], tctx->l3key[1], partial * 8);
559}
560
561static int vmac_final(struct shash_desc *desc, u8 *out)
562{
563	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
564	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
565	int index;
566	u64 hash, pad;
567
568	if (dctx->nonce_size != VMAC_NONCEBYTES)
569		return -EINVAL;
570
571	/*
572	 * The VMAC specification requires a nonce at least 1 bit shorter than
573	 * the block cipher's block length, so we actually only accept a 127-bit
574	 * nonce.  We define the unused bit to be the first one and require that
575	 * it be 0, so the needed prepending of a 0 bit is implicit.
576	 */
577	if (dctx->nonce.bytes[0] & 0x80)
578		return -EINVAL;
579
580	/* Finish calculating the VHASH of the message */
581	hash = vhash_final(tctx, dctx);
 
 
 
 
 
 
582
583	/* Generate pseudorandom pad by encrypting the nonce */
584	BUILD_BUG_ON(VMAC_NONCEBYTES != 2 * (VMAC_TAG_LEN / 8));
585	index = dctx->nonce.bytes[VMAC_NONCEBYTES - 1] & 1;
586	dctx->nonce.bytes[VMAC_NONCEBYTES - 1] &= ~1;
587	crypto_cipher_encrypt_one(tctx->cipher, dctx->nonce.bytes,
588				  dctx->nonce.bytes);
589	pad = be64_to_cpu(dctx->nonce.pads[index]);
590
591	/* The VMAC is the sum of VHASH and the pseudorandom pad */
592	put_unaligned_be64(hash + pad, out);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
593	return 0;
594}
595
596static int vmac_init_tfm(struct crypto_tfm *tfm)
597{
598	struct crypto_instance *inst = crypto_tfm_alg_instance(tfm);
599	struct crypto_cipher_spawn *spawn = crypto_instance_ctx(inst);
600	struct vmac_tfm_ctx *tctx = crypto_tfm_ctx(tfm);
601	struct crypto_cipher *cipher;
 
 
 
602
603	cipher = crypto_spawn_cipher(spawn);
604	if (IS_ERR(cipher))
605		return PTR_ERR(cipher);
606
607	tctx->cipher = cipher;
608	return 0;
609}
610
611static void vmac_exit_tfm(struct crypto_tfm *tfm)
612{
613	struct vmac_tfm_ctx *tctx = crypto_tfm_ctx(tfm);
614
615	crypto_free_cipher(tctx->cipher);
616}
617
618static int vmac_create(struct crypto_template *tmpl, struct rtattr **tb)
619{
620	struct shash_instance *inst;
621	struct crypto_cipher_spawn *spawn;
622	struct crypto_alg *alg;
623	u32 mask;
624	int err;
625
626	err = crypto_check_attr_type(tb, CRYPTO_ALG_TYPE_SHASH, &mask);
627	if (err)
628		return err;
629
630	inst = kzalloc(sizeof(*inst) + sizeof(*spawn), GFP_KERNEL);
631	if (!inst)
632		return -ENOMEM;
633	spawn = shash_instance_ctx(inst);
634
635	err = crypto_grab_cipher(spawn, shash_crypto_instance(inst),
636				 crypto_attr_alg_name(tb[1]), 0, mask);
637	if (err)
638		goto err_free_inst;
639	alg = crypto_spawn_cipher_alg(spawn);
640
641	err = -EINVAL;
642	if (alg->cra_blocksize != VMAC_NONCEBYTES)
643		goto err_free_inst;
644
645	err = crypto_inst_setname(shash_crypto_instance(inst), tmpl->name, alg);
646	if (err)
647		goto err_free_inst;
648
649	inst->alg.base.cra_priority = alg->cra_priority;
650	inst->alg.base.cra_blocksize = alg->cra_blocksize;
651	inst->alg.base.cra_alignmask = alg->cra_alignmask;
652
653	inst->alg.base.cra_ctxsize = sizeof(struct vmac_tfm_ctx);
 
654	inst->alg.base.cra_init = vmac_init_tfm;
655	inst->alg.base.cra_exit = vmac_exit_tfm;
656
657	inst->alg.descsize = sizeof(struct vmac_desc_ctx);
658	inst->alg.digestsize = VMAC_TAG_LEN / 8;
659	inst->alg.init = vmac_init;
660	inst->alg.update = vmac_update;
661	inst->alg.final = vmac_final;
662	inst->alg.setkey = vmac_setkey;
663
664	inst->free = shash_free_singlespawn_instance;
665
666	err = shash_register_instance(tmpl, inst);
667	if (err) {
668err_free_inst:
669		shash_free_singlespawn_instance(inst);
670	}
 
 
 
671	return err;
672}
673
674static struct crypto_template vmac64_tmpl = {
675	.name = "vmac64",
676	.create = vmac_create,
 
677	.module = THIS_MODULE,
678};
679
680static int __init vmac_module_init(void)
681{
682	return crypto_register_template(&vmac64_tmpl);
683}
684
685static void __exit vmac_module_exit(void)
686{
687	crypto_unregister_template(&vmac64_tmpl);
688}
689
690subsys_initcall(vmac_module_init);
691module_exit(vmac_module_exit);
692
693MODULE_LICENSE("GPL");
694MODULE_DESCRIPTION("VMAC hash algorithm");
695MODULE_ALIAS_CRYPTO("vmac64");
v4.17
  1/*
  2 * Modified to interface to the Linux kernel
 
 
 
  3 * Copyright (c) 2009, Intel Corporation.
 
  4 *
  5 * This program is free software; you can redistribute it and/or modify it
  6 * under the terms and conditions of the GNU General Public License,
  7 * version 2, as published by the Free Software Foundation.
  8 *
  9 * This program is distributed in the hope it will be useful, but WITHOUT
 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 11 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
 12 * more details.
 13 *
 14 * You should have received a copy of the GNU General Public License along with
 15 * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
 16 * Place - Suite 330, Boston, MA 02111-1307 USA.
 17 */
 18
 19/* --------------------------------------------------------------------------
 20 * VMAC and VHASH Implementation by Ted Krovetz (tdk@acm.org) and Wei Dai.
 21 * This implementation is herby placed in the public domain.
 22 * The authors offers no warranty. Use at your own risk.
 23 * Please send bug reports to the authors.
 24 * Last modified: 17 APR 08, 1700 PDT
 25 * ----------------------------------------------------------------------- */
 26
 
 27#include <linux/init.h>
 28#include <linux/types.h>
 29#include <linux/crypto.h>
 30#include <linux/module.h>
 31#include <linux/scatterlist.h>
 32#include <asm/byteorder.h>
 33#include <crypto/scatterwalk.h>
 34#include <crypto/vmac.h>
 35#include <crypto/internal/hash.h>
 36
 37/*
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 38 * Constants and masks
 39 */
 40#define UINT64_C(x) x##ULL
 41static const u64 p64   = UINT64_C(0xfffffffffffffeff);	/* 2^64 - 257 prime  */
 42static const u64 m62   = UINT64_C(0x3fffffffffffffff);	/* 62-bit mask       */
 43static const u64 m63   = UINT64_C(0x7fffffffffffffff);	/* 63-bit mask       */
 44static const u64 m64   = UINT64_C(0xffffffffffffffff);	/* 64-bit mask       */
 45static const u64 mpoly = UINT64_C(0x1fffffff1fffffff);	/* Poly key mask     */
 46
 47#define pe64_to_cpup le64_to_cpup		/* Prefer little endian */
 48
 49#ifdef __LITTLE_ENDIAN
 50#define INDEX_HIGH 1
 51#define INDEX_LOW 0
 52#else
 53#define INDEX_HIGH 0
 54#define INDEX_LOW 1
 55#endif
 56
 57/*
 58 * The following routines are used in this implementation. They are
 59 * written via macros to simulate zero-overhead call-by-reference.
 60 *
 61 * MUL64: 64x64->128-bit multiplication
 62 * PMUL64: assumes top bits cleared on inputs
 63 * ADD128: 128x128->128-bit addition
 64 */
 65
 66#define ADD128(rh, rl, ih, il)						\
 67	do {								\
 68		u64 _il = (il);						\
 69		(rl) += (_il);						\
 70		if ((rl) < (_il))					\
 71			(rh)++;						\
 72		(rh) += (ih);						\
 73	} while (0)
 74
 75#define MUL32(i1, i2)	((u64)(u32)(i1)*(u32)(i2))
 76
 77#define PMUL64(rh, rl, i1, i2)	/* Assumes m doesn't overflow */	\
 78	do {								\
 79		u64 _i1 = (i1), _i2 = (i2);				\
 80		u64 m = MUL32(_i1, _i2>>32) + MUL32(_i1>>32, _i2);	\
 81		rh = MUL32(_i1>>32, _i2>>32);				\
 82		rl = MUL32(_i1, _i2);					\
 83		ADD128(rh, rl, (m >> 32), (m << 32));			\
 84	} while (0)
 85
 86#define MUL64(rh, rl, i1, i2)						\
 87	do {								\
 88		u64 _i1 = (i1), _i2 = (i2);				\
 89		u64 m1 = MUL32(_i1, _i2>>32);				\
 90		u64 m2 = MUL32(_i1>>32, _i2);				\
 91		rh = MUL32(_i1>>32, _i2>>32);				\
 92		rl = MUL32(_i1, _i2);					\
 93		ADD128(rh, rl, (m1 >> 32), (m1 << 32));			\
 94		ADD128(rh, rl, (m2 >> 32), (m2 << 32));			\
 95	} while (0)
 96
 97/*
 98 * For highest performance the L1 NH and L2 polynomial hashes should be
 99 * carefully implemented to take advantage of one's target architecture.
100 * Here these two hash functions are defined multiple time; once for
101 * 64-bit architectures, once for 32-bit SSE2 architectures, and once
102 * for the rest (32-bit) architectures.
103 * For each, nh_16 *must* be defined (works on multiples of 16 bytes).
104 * Optionally, nh_vmac_nhbytes can be defined (for multiples of
105 * VMAC_NHBYTES), and nh_16_2 and nh_vmac_nhbytes_2 (versions that do two
106 * NH computations at once).
107 */
108
109#ifdef CONFIG_64BIT
110
111#define nh_16(mp, kp, nw, rh, rl)					\
112	do {								\
113		int i; u64 th, tl;					\
114		rh = rl = 0;						\
115		for (i = 0; i < nw; i += 2) {				\
116			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
117				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
118			ADD128(rh, rl, th, tl);				\
119		}							\
120	} while (0)
121
122#define nh_16_2(mp, kp, nw, rh, rl, rh1, rl1)				\
123	do {								\
124		int i; u64 th, tl;					\
125		rh1 = rl1 = rh = rl = 0;				\
126		for (i = 0; i < nw; i += 2) {				\
127			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
128				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
129			ADD128(rh, rl, th, tl);				\
130			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2],	\
131				pe64_to_cpup((mp)+i+1)+(kp)[i+3]);	\
132			ADD128(rh1, rl1, th, tl);			\
133		}							\
134	} while (0)
135
136#if (VMAC_NHBYTES >= 64) /* These versions do 64-bytes of message at a time */
137#define nh_vmac_nhbytes(mp, kp, nw, rh, rl)				\
138	do {								\
139		int i; u64 th, tl;					\
140		rh = rl = 0;						\
141		for (i = 0; i < nw; i += 8) {				\
142			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
143				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
144			ADD128(rh, rl, th, tl);				\
145			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2],	\
146				pe64_to_cpup((mp)+i+3)+(kp)[i+3]);	\
147			ADD128(rh, rl, th, tl);				\
148			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4],	\
149				pe64_to_cpup((mp)+i+5)+(kp)[i+5]);	\
150			ADD128(rh, rl, th, tl);				\
151			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6],	\
152				pe64_to_cpup((mp)+i+7)+(kp)[i+7]);	\
153			ADD128(rh, rl, th, tl);				\
154		}							\
155	} while (0)
156
157#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh1, rl1)			\
158	do {								\
159		int i; u64 th, tl;					\
160		rh1 = rl1 = rh = rl = 0;				\
161		for (i = 0; i < nw; i += 8) {				\
162			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
163				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
164			ADD128(rh, rl, th, tl);				\
165			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2],	\
166				pe64_to_cpup((mp)+i+1)+(kp)[i+3]);	\
167			ADD128(rh1, rl1, th, tl);			\
168			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2],	\
169				pe64_to_cpup((mp)+i+3)+(kp)[i+3]);	\
170			ADD128(rh, rl, th, tl);				\
171			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+4],	\
172				pe64_to_cpup((mp)+i+3)+(kp)[i+5]);	\
173			ADD128(rh1, rl1, th, tl);			\
174			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4],	\
175				pe64_to_cpup((mp)+i+5)+(kp)[i+5]);	\
176			ADD128(rh, rl, th, tl);				\
177			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+6],	\
178				pe64_to_cpup((mp)+i+5)+(kp)[i+7]);	\
179			ADD128(rh1, rl1, th, tl);			\
180			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6],	\
181				pe64_to_cpup((mp)+i+7)+(kp)[i+7]);	\
182			ADD128(rh, rl, th, tl);				\
183			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+8],	\
184				pe64_to_cpup((mp)+i+7)+(kp)[i+9]);	\
185			ADD128(rh1, rl1, th, tl);			\
186		}							\
187	} while (0)
188#endif
189
190#define poly_step(ah, al, kh, kl, mh, ml)				\
191	do {								\
192		u64 t1h, t1l, t2h, t2l, t3h, t3l, z = 0;		\
193		/* compute ab*cd, put bd into result registers */	\
194		PMUL64(t3h, t3l, al, kh);				\
195		PMUL64(t2h, t2l, ah, kl);				\
196		PMUL64(t1h, t1l, ah, 2*kh);				\
197		PMUL64(ah, al, al, kl);					\
198		/* add 2 * ac to result */				\
199		ADD128(ah, al, t1h, t1l);				\
200		/* add together ad + bc */				\
201		ADD128(t2h, t2l, t3h, t3l);				\
202		/* now (ah,al), (t2l,2*t2h) need summing */		\
203		/* first add the high registers, carrying into t2h */	\
204		ADD128(t2h, ah, z, t2l);				\
205		/* double t2h and add top bit of ah */			\
206		t2h = 2 * t2h + (ah >> 63);				\
207		ah &= m63;						\
208		/* now add the low registers */				\
209		ADD128(ah, al, mh, ml);					\
210		ADD128(ah, al, z, t2h);					\
211	} while (0)
212
213#else /* ! CONFIG_64BIT */
214
215#ifndef nh_16
216#define nh_16(mp, kp, nw, rh, rl)					\
217	do {								\
218		u64 t1, t2, m1, m2, t;					\
219		int i;							\
220		rh = rl = t = 0;					\
221		for (i = 0; i < nw; i += 2)  {				\
222			t1 = pe64_to_cpup(mp+i) + kp[i];		\
223			t2 = pe64_to_cpup(mp+i+1) + kp[i+1];		\
224			m2 = MUL32(t1 >> 32, t2);			\
225			m1 = MUL32(t1, t2 >> 32);			\
226			ADD128(rh, rl, MUL32(t1 >> 32, t2 >> 32),	\
227				MUL32(t1, t2));				\
228			rh += (u64)(u32)(m1 >> 32)			\
229				+ (u32)(m2 >> 32);			\
230			t += (u64)(u32)m1 + (u32)m2;			\
231		}							\
232		ADD128(rh, rl, (t >> 32), (t << 32));			\
233	} while (0)
234#endif
235
236static void poly_step_func(u64 *ahi, u64 *alo,
237			const u64 *kh, const u64 *kl,
238			const u64 *mh, const u64 *ml)
239{
240#define a0 (*(((u32 *)alo)+INDEX_LOW))
241#define a1 (*(((u32 *)alo)+INDEX_HIGH))
242#define a2 (*(((u32 *)ahi)+INDEX_LOW))
243#define a3 (*(((u32 *)ahi)+INDEX_HIGH))
244#define k0 (*(((u32 *)kl)+INDEX_LOW))
245#define k1 (*(((u32 *)kl)+INDEX_HIGH))
246#define k2 (*(((u32 *)kh)+INDEX_LOW))
247#define k3 (*(((u32 *)kh)+INDEX_HIGH))
248
249	u64 p, q, t;
250	u32 t2;
251
252	p = MUL32(a3, k3);
253	p += p;
254	p += *(u64 *)mh;
255	p += MUL32(a0, k2);
256	p += MUL32(a1, k1);
257	p += MUL32(a2, k0);
258	t = (u32)(p);
259	p >>= 32;
260	p += MUL32(a0, k3);
261	p += MUL32(a1, k2);
262	p += MUL32(a2, k1);
263	p += MUL32(a3, k0);
264	t |= ((u64)((u32)p & 0x7fffffff)) << 32;
265	p >>= 31;
266	p += (u64)(((u32 *)ml)[INDEX_LOW]);
267	p += MUL32(a0, k0);
268	q =  MUL32(a1, k3);
269	q += MUL32(a2, k2);
270	q += MUL32(a3, k1);
271	q += q;
272	p += q;
273	t2 = (u32)(p);
274	p >>= 32;
275	p += (u64)(((u32 *)ml)[INDEX_HIGH]);
276	p += MUL32(a0, k1);
277	p += MUL32(a1, k0);
278	q =  MUL32(a2, k3);
279	q += MUL32(a3, k2);
280	q += q;
281	p += q;
282	*(u64 *)(alo) = (p << 32) | t2;
283	p >>= 32;
284	*(u64 *)(ahi) = p + t;
285
286#undef a0
287#undef a1
288#undef a2
289#undef a3
290#undef k0
291#undef k1
292#undef k2
293#undef k3
294}
295
296#define poly_step(ah, al, kh, kl, mh, ml)				\
297	poly_step_func(&(ah), &(al), &(kh), &(kl), &(mh), &(ml))
298
299#endif  /* end of specialized NH and poly definitions */
300
301/* At least nh_16 is defined. Defined others as needed here */
302#ifndef nh_16_2
303#define nh_16_2(mp, kp, nw, rh, rl, rh2, rl2)				\
304	do { 								\
305		nh_16(mp, kp, nw, rh, rl);				\
306		nh_16(mp, ((kp)+2), nw, rh2, rl2);			\
307	} while (0)
308#endif
309#ifndef nh_vmac_nhbytes
310#define nh_vmac_nhbytes(mp, kp, nw, rh, rl)				\
311	nh_16(mp, kp, nw, rh, rl)
312#endif
313#ifndef nh_vmac_nhbytes_2
314#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh2, rl2)			\
315	do {								\
316		nh_vmac_nhbytes(mp, kp, nw, rh, rl);			\
317		nh_vmac_nhbytes(mp, ((kp)+2), nw, rh2, rl2);		\
318	} while (0)
319#endif
320
321static void vhash_abort(struct vmac_ctx *ctx)
322{
323	ctx->polytmp[0] = ctx->polykey[0] ;
324	ctx->polytmp[1] = ctx->polykey[1] ;
325	ctx->first_block_processed = 0;
326}
327
328static u64 l3hash(u64 p1, u64 p2, u64 k1, u64 k2, u64 len)
329{
330	u64 rh, rl, t, z = 0;
331
332	/* fully reduce (p1,p2)+(len,0) mod p127 */
333	t = p1 >> 63;
334	p1 &= m63;
335	ADD128(p1, p2, len, t);
336	/* At this point, (p1,p2) is at most 2^127+(len<<64) */
337	t = (p1 > m63) + ((p1 == m63) && (p2 == m64));
338	ADD128(p1, p2, z, t);
339	p1 &= m63;
340
341	/* compute (p1,p2)/(2^64-2^32) and (p1,p2)%(2^64-2^32) */
342	t = p1 + (p2 >> 32);
343	t += (t >> 32);
344	t += (u32)t > 0xfffffffeu;
345	p1 += (t >> 32);
346	p2 += (p1 << 32);
347
348	/* compute (p1+k1)%p64 and (p2+k2)%p64 */
349	p1 += k1;
350	p1 += (0 - (p1 < k1)) & 257;
351	p2 += k2;
352	p2 += (0 - (p2 < k2)) & 257;
353
354	/* compute (p1+k1)*(p2+k2)%p64 */
355	MUL64(rh, rl, p1, p2);
356	t = rh >> 56;
357	ADD128(t, rl, z, rh);
358	rh <<= 8;
359	ADD128(t, rl, z, rh);
360	t += t << 8;
361	rl += t;
362	rl += (0 - (rl < t)) & 257;
363	rl += (0 - (rl > p64-1)) & 257;
364	return rl;
365}
366
367static void vhash_update(const unsigned char *m,
368			unsigned int mbytes, /* Pos multiple of VMAC_NHBYTES */
369			struct vmac_ctx *ctx)
370{
371	u64 rh, rl, *mptr;
372	const u64 *kptr = (u64 *)ctx->nhkey;
373	int i;
374	u64 ch, cl;
375	u64 pkh = ctx->polykey[0];
376	u64 pkl = ctx->polykey[1];
377
378	if (!mbytes)
379		return;
380
381	BUG_ON(mbytes % VMAC_NHBYTES);
382
383	mptr = (u64 *)m;
384	i = mbytes / VMAC_NHBYTES;  /* Must be non-zero */
385
386	ch = ctx->polytmp[0];
387	cl = ctx->polytmp[1];
388
389	if (!ctx->first_block_processed) {
390		ctx->first_block_processed = 1;
391		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
392		rh &= m62;
393		ADD128(ch, cl, rh, rl);
394		mptr += (VMAC_NHBYTES/sizeof(u64));
395		i--;
396	}
397
398	while (i--) {
399		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
400		rh &= m62;
401		poly_step(ch, cl, pkh, pkl, rh, rl);
402		mptr += (VMAC_NHBYTES/sizeof(u64));
403	}
404
405	ctx->polytmp[0] = ch;
406	ctx->polytmp[1] = cl;
407}
408
409static u64 vhash(unsigned char m[], unsigned int mbytes,
410			u64 *tagl, struct vmac_ctx *ctx)
411{
412	u64 rh, rl, *mptr;
413	const u64 *kptr = (u64 *)ctx->nhkey;
414	int i, remaining;
415	u64 ch, cl;
416	u64 pkh = ctx->polykey[0];
417	u64 pkl = ctx->polykey[1];
418
419	mptr = (u64 *)m;
420	i = mbytes / VMAC_NHBYTES;
421	remaining = mbytes % VMAC_NHBYTES;
422
423	if (ctx->first_block_processed) {
424		ch = ctx->polytmp[0];
425		cl = ctx->polytmp[1];
426	} else if (i) {
427		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, ch, cl);
428		ch &= m62;
429		ADD128(ch, cl, pkh, pkl);
430		mptr += (VMAC_NHBYTES/sizeof(u64));
431		i--;
432	} else if (remaining) {
433		nh_16(mptr, kptr, 2*((remaining+15)/16), ch, cl);
434		ch &= m62;
435		ADD128(ch, cl, pkh, pkl);
436		mptr += (VMAC_NHBYTES/sizeof(u64));
437		goto do_l3;
438	} else {/* Empty String */
439		ch = pkh; cl = pkl;
440		goto do_l3;
441	}
442
443	while (i--) {
444		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
445		rh &= m62;
446		poly_step(ch, cl, pkh, pkl, rh, rl);
447		mptr += (VMAC_NHBYTES/sizeof(u64));
448	}
449	if (remaining) {
450		nh_16(mptr, kptr, 2*((remaining+15)/16), rh, rl);
451		rh &= m62;
452		poly_step(ch, cl, pkh, pkl, rh, rl);
453	}
454
455do_l3:
456	vhash_abort(ctx);
457	remaining *= 8;
458	return l3hash(ch, cl, ctx->l3key[0], ctx->l3key[1], remaining);
459}
460
461static u64 vmac(unsigned char m[], unsigned int mbytes,
462			const unsigned char n[16], u64 *tagl,
463			struct vmac_ctx_t *ctx)
464{
465	u64 *in_n, *out_p;
466	u64 p, h;
467	int i;
468
469	in_n = ctx->__vmac_ctx.cached_nonce;
470	out_p = ctx->__vmac_ctx.cached_aes;
471
472	i = n[15] & 1;
473	if ((*(u64 *)(n+8) != in_n[1]) || (*(u64 *)(n) != in_n[0])) {
474		in_n[0] = *(u64 *)(n);
475		in_n[1] = *(u64 *)(n+8);
476		((unsigned char *)in_n)[15] &= 0xFE;
477		crypto_cipher_encrypt_one(ctx->child,
478			(unsigned char *)out_p, (unsigned char *)in_n);
479
480		((unsigned char *)in_n)[15] |= (unsigned char)(1-i);
481	}
482	p = be64_to_cpup(out_p + i);
483	h = vhash(m, mbytes, (u64 *)0, &ctx->__vmac_ctx);
484	return le64_to_cpu(p + h);
485}
486
487static int vmac_set_key(unsigned char user_key[], struct vmac_ctx_t *ctx)
488{
489	u64 in[2] = {0}, out[2];
490	unsigned i;
491	int err = 0;
492
493	err = crypto_cipher_setkey(ctx->child, user_key, VMAC_KEY_LEN);
494	if (err)
495		return err;
496
497	/* Fill nh key */
498	((unsigned char *)in)[0] = 0x80;
499	for (i = 0; i < sizeof(ctx->__vmac_ctx.nhkey)/8; i += 2) {
500		crypto_cipher_encrypt_one(ctx->child,
501			(unsigned char *)out, (unsigned char *)in);
502		ctx->__vmac_ctx.nhkey[i] = be64_to_cpup(out);
503		ctx->__vmac_ctx.nhkey[i+1] = be64_to_cpup(out+1);
504		((unsigned char *)in)[15] += 1;
505	}
506
507	/* Fill poly key */
508	((unsigned char *)in)[0] = 0xC0;
509	in[1] = 0;
510	for (i = 0; i < sizeof(ctx->__vmac_ctx.polykey)/8; i += 2) {
511		crypto_cipher_encrypt_one(ctx->child,
512			(unsigned char *)out, (unsigned char *)in);
513		ctx->__vmac_ctx.polytmp[i] =
514			ctx->__vmac_ctx.polykey[i] =
515				be64_to_cpup(out) & mpoly;
516		ctx->__vmac_ctx.polytmp[i+1] =
517			ctx->__vmac_ctx.polykey[i+1] =
518				be64_to_cpup(out+1) & mpoly;
519		((unsigned char *)in)[15] += 1;
520	}
521
522	/* Fill ip key */
523	((unsigned char *)in)[0] = 0xE0;
524	in[1] = 0;
525	for (i = 0; i < sizeof(ctx->__vmac_ctx.l3key)/8; i += 2) {
526		do {
527			crypto_cipher_encrypt_one(ctx->child,
528				(unsigned char *)out, (unsigned char *)in);
529			ctx->__vmac_ctx.l3key[i] = be64_to_cpup(out);
530			ctx->__vmac_ctx.l3key[i+1] = be64_to_cpup(out+1);
531			((unsigned char *)in)[15] += 1;
532		} while (ctx->__vmac_ctx.l3key[i] >= p64
533			|| ctx->__vmac_ctx.l3key[i+1] >= p64);
534	}
535
536	/* Invalidate nonce/aes cache and reset other elements */
537	ctx->__vmac_ctx.cached_nonce[0] = (u64)-1; /* Ensure illegal nonce */
538	ctx->__vmac_ctx.cached_nonce[1] = (u64)0;  /* Ensure illegal nonce */
539	ctx->__vmac_ctx.first_block_processed = 0;
540
541	return err;
542}
543
544static int vmac_setkey(struct crypto_shash *parent,
545		const u8 *key, unsigned int keylen)
546{
547	struct vmac_ctx_t *ctx = crypto_shash_ctx(parent);
 
548
549	if (keylen != VMAC_KEY_LEN) {
550		crypto_shash_set_flags(parent, CRYPTO_TFM_RES_BAD_KEY_LEN);
551		return -EINVAL;
552	}
553
554	return vmac_set_key((u8 *)key, ctx);
555}
556
557static int vmac_init(struct shash_desc *pdesc)
558{
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
559	return 0;
560}
561
562static int vmac_update(struct shash_desc *pdesc, const u8 *p,
563		unsigned int len)
564{
565	struct crypto_shash *parent = pdesc->tfm;
566	struct vmac_ctx_t *ctx = crypto_shash_ctx(parent);
567	int expand;
568	int min;
 
 
 
 
 
569
570	expand = VMAC_NHBYTES - ctx->partial_size > 0 ?
571			VMAC_NHBYTES - ctx->partial_size : 0;
 
 
 
 
 
 
 
572
573	min = len < expand ? len : expand;
 
 
574
575	memcpy(ctx->partial + ctx->partial_size, p, min);
576	ctx->partial_size += min;
 
 
 
 
577
578	if (len < expand)
579		return 0;
580
581	vhash_update(ctx->partial, VMAC_NHBYTES, &ctx->__vmac_ctx);
582	ctx->partial_size = 0;
 
 
 
 
 
 
583
584	len -= expand;
585	p += expand;
586
587	if (len % VMAC_NHBYTES) {
588		memcpy(ctx->partial, p + len - (len % VMAC_NHBYTES),
589			len % VMAC_NHBYTES);
590		ctx->partial_size = len % VMAC_NHBYTES;
591	}
592
593	vhash_update(p, len - len % VMAC_NHBYTES, &ctx->__vmac_ctx);
 
 
 
 
 
 
594
595	return 0;
596}
597
598static int vmac_final(struct shash_desc *pdesc, u8 *out)
599{
600	struct crypto_shash *parent = pdesc->tfm;
601	struct vmac_ctx_t *ctx = crypto_shash_ctx(parent);
602	vmac_t mac;
603	u8 nonce[16] = {};
604
605	/* vmac() ends up accessing outside the array bounds that
606	 * we specify.  In appears to access up to the next 2-word
607	 * boundary.  We'll just be uber cautious and zero the
608	 * unwritten bytes in the buffer.
609	 */
610	if (ctx->partial_size) {
611		memset(ctx->partial + ctx->partial_size, 0,
612			VMAC_NHBYTES - ctx->partial_size);
613	}
614	mac = vmac(ctx->partial, ctx->partial_size, nonce, NULL, ctx);
615	memcpy(out, &mac, sizeof(vmac_t));
616	memzero_explicit(&mac, sizeof(vmac_t));
617	memset(&ctx->__vmac_ctx, 0, sizeof(struct vmac_ctx));
618	ctx->partial_size = 0;
619	return 0;
620}
621
622static int vmac_init_tfm(struct crypto_tfm *tfm)
623{
 
 
 
624	struct crypto_cipher *cipher;
625	struct crypto_instance *inst = (void *)tfm->__crt_alg;
626	struct crypto_spawn *spawn = crypto_instance_ctx(inst);
627	struct vmac_ctx_t *ctx = crypto_tfm_ctx(tfm);
628
629	cipher = crypto_spawn_cipher(spawn);
630	if (IS_ERR(cipher))
631		return PTR_ERR(cipher);
632
633	ctx->child = cipher;
634	return 0;
635}
636
637static void vmac_exit_tfm(struct crypto_tfm *tfm)
638{
639	struct vmac_ctx_t *ctx = crypto_tfm_ctx(tfm);
640	crypto_free_cipher(ctx->child);
 
641}
642
643static int vmac_create(struct crypto_template *tmpl, struct rtattr **tb)
644{
645	struct shash_instance *inst;
 
646	struct crypto_alg *alg;
 
647	int err;
648
649	err = crypto_check_attr_type(tb, CRYPTO_ALG_TYPE_SHASH);
650	if (err)
651		return err;
652
653	alg = crypto_get_attr_alg(tb, CRYPTO_ALG_TYPE_CIPHER,
654			CRYPTO_ALG_TYPE_MASK);
655	if (IS_ERR(alg))
656		return PTR_ERR(alg);
657
658	inst = shash_alloc_instance("vmac", alg);
659	err = PTR_ERR(inst);
660	if (IS_ERR(inst))
661		goto out_put_alg;
662
663	err = crypto_init_spawn(shash_instance_ctx(inst), alg,
664			shash_crypto_instance(inst),
665			CRYPTO_ALG_TYPE_MASK);
 
 
 
666	if (err)
667		goto out_free_inst;
668
669	inst->alg.base.cra_priority = alg->cra_priority;
670	inst->alg.base.cra_blocksize = alg->cra_blocksize;
671	inst->alg.base.cra_alignmask = alg->cra_alignmask;
672
673	inst->alg.digestsize = sizeof(vmac_t);
674	inst->alg.base.cra_ctxsize = sizeof(struct vmac_ctx_t);
675	inst->alg.base.cra_init = vmac_init_tfm;
676	inst->alg.base.cra_exit = vmac_exit_tfm;
677
 
 
678	inst->alg.init = vmac_init;
679	inst->alg.update = vmac_update;
680	inst->alg.final = vmac_final;
681	inst->alg.setkey = vmac_setkey;
682
 
 
683	err = shash_register_instance(tmpl, inst);
684	if (err) {
685out_free_inst:
686		shash_free_instance(shash_crypto_instance(inst));
687	}
688
689out_put_alg:
690	crypto_mod_put(alg);
691	return err;
692}
693
694static struct crypto_template vmac_tmpl = {
695	.name = "vmac",
696	.create = vmac_create,
697	.free = shash_free_instance,
698	.module = THIS_MODULE,
699};
700
701static int __init vmac_module_init(void)
702{
703	return crypto_register_template(&vmac_tmpl);
704}
705
706static void __exit vmac_module_exit(void)
707{
708	crypto_unregister_template(&vmac_tmpl);
709}
710
711module_init(vmac_module_init);
712module_exit(vmac_module_exit);
713
714MODULE_LICENSE("GPL");
715MODULE_DESCRIPTION("VMAC hash algorithm");
716MODULE_ALIAS_CRYPTO("vmac");