xref: /openbmc/linux/crypto/vmac.c (revision cf21eb6a)
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/cipher.h>
40 #include <crypto/internal/hash.h>
41 
42 /*
43  * User definable settings.
44  */
45 #define VMAC_TAG_LEN	64
46 #define VMAC_KEY_SIZE	128/* Must be 128, 192 or 256			*/
47 #define VMAC_KEY_LEN	(VMAC_KEY_SIZE/8)
48 #define VMAC_NHBYTES	128/* Must 2^i for any 3 < i < 13 Standard = 128*/
49 #define VMAC_NONCEBYTES	16
50 
51 /* per-transform (per-key) context */
52 struct vmac_tfm_ctx {
53 	struct crypto_cipher *cipher;
54 	u64 nhkey[(VMAC_NHBYTES/8)+2*(VMAC_TAG_LEN/64-1)];
55 	u64 polykey[2*VMAC_TAG_LEN/64];
56 	u64 l3key[2*VMAC_TAG_LEN/64];
57 };
58 
59 /* per-request context */
60 struct vmac_desc_ctx {
61 	union {
62 		u8 partial[VMAC_NHBYTES];	/* partial block */
63 		__le64 partial_words[VMAC_NHBYTES / 8];
64 	};
65 	unsigned int partial_size;	/* size of the partial block */
66 	bool first_block_processed;
67 	u64 polytmp[2*VMAC_TAG_LEN/64];	/* running total of L2-hash */
68 	union {
69 		u8 bytes[VMAC_NONCEBYTES];
70 		__be64 pads[VMAC_NONCEBYTES / 8];
71 	} nonce;
72 	unsigned int nonce_size; /* nonce bytes filled so far */
73 };
74 
75 /*
76  * Constants and masks
77  */
78 #define UINT64_C(x) x##ULL
79 static const u64 p64   = UINT64_C(0xfffffffffffffeff);	/* 2^64 - 257 prime  */
80 static const u64 m62   = UINT64_C(0x3fffffffffffffff);	/* 62-bit mask       */
81 static const u64 m63   = UINT64_C(0x7fffffffffffffff);	/* 63-bit mask       */
82 static const u64 m64   = UINT64_C(0xffffffffffffffff);	/* 64-bit mask       */
83 static const u64 mpoly = UINT64_C(0x1fffffff1fffffff);	/* Poly key mask     */
84 
85 #define pe64_to_cpup le64_to_cpup		/* Prefer little endian */
86 
87 #ifdef __LITTLE_ENDIAN
88 #define INDEX_HIGH 1
89 #define INDEX_LOW 0
90 #else
91 #define INDEX_HIGH 0
92 #define INDEX_LOW 1
93 #endif
94 
95 /*
96  * The following routines are used in this implementation. They are
97  * written via macros to simulate zero-overhead call-by-reference.
98  *
99  * MUL64: 64x64->128-bit multiplication
100  * PMUL64: assumes top bits cleared on inputs
101  * ADD128: 128x128->128-bit addition
102  */
103 
104 #define ADD128(rh, rl, ih, il)						\
105 	do {								\
106 		u64 _il = (il);						\
107 		(rl) += (_il);						\
108 		if ((rl) < (_il))					\
109 			(rh)++;						\
110 		(rh) += (ih);						\
111 	} while (0)
112 
113 #define MUL32(i1, i2)	((u64)(u32)(i1)*(u32)(i2))
114 
115 #define PMUL64(rh, rl, i1, i2)	/* Assumes m doesn't overflow */	\
116 	do {								\
117 		u64 _i1 = (i1), _i2 = (i2);				\
118 		u64 m = MUL32(_i1, _i2>>32) + MUL32(_i1>>32, _i2);	\
119 		rh = MUL32(_i1>>32, _i2>>32);				\
120 		rl = MUL32(_i1, _i2);					\
121 		ADD128(rh, rl, (m >> 32), (m << 32));			\
122 	} while (0)
123 
124 #define MUL64(rh, rl, i1, i2)						\
125 	do {								\
126 		u64 _i1 = (i1), _i2 = (i2);				\
127 		u64 m1 = MUL32(_i1, _i2>>32);				\
128 		u64 m2 = MUL32(_i1>>32, _i2);				\
129 		rh = MUL32(_i1>>32, _i2>>32);				\
130 		rl = MUL32(_i1, _i2);					\
131 		ADD128(rh, rl, (m1 >> 32), (m1 << 32));			\
132 		ADD128(rh, rl, (m2 >> 32), (m2 << 32));			\
133 	} while (0)
134 
135 /*
136  * For highest performance the L1 NH and L2 polynomial hashes should be
137  * carefully implemented to take advantage of one's target architecture.
138  * Here these two hash functions are defined multiple time; once for
139  * 64-bit architectures, once for 32-bit SSE2 architectures, and once
140  * for the rest (32-bit) architectures.
141  * For each, nh_16 *must* be defined (works on multiples of 16 bytes).
142  * Optionally, nh_vmac_nhbytes can be defined (for multiples of
143  * VMAC_NHBYTES), and nh_16_2 and nh_vmac_nhbytes_2 (versions that do two
144  * NH computations at once).
145  */
146 
147 #ifdef CONFIG_64BIT
148 
149 #define nh_16(mp, kp, nw, rh, rl)					\
150 	do {								\
151 		int i; u64 th, tl;					\
152 		rh = rl = 0;						\
153 		for (i = 0; i < nw; i += 2) {				\
154 			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
155 				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
156 			ADD128(rh, rl, th, tl);				\
157 		}							\
158 	} while (0)
159 
160 #define nh_16_2(mp, kp, nw, rh, rl, rh1, rl1)				\
161 	do {								\
162 		int i; u64 th, tl;					\
163 		rh1 = rl1 = rh = rl = 0;				\
164 		for (i = 0; i < nw; i += 2) {				\
165 			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
166 				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
167 			ADD128(rh, rl, th, tl);				\
168 			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2],	\
169 				pe64_to_cpup((mp)+i+1)+(kp)[i+3]);	\
170 			ADD128(rh1, rl1, th, tl);			\
171 		}							\
172 	} while (0)
173 
174 #if (VMAC_NHBYTES >= 64) /* These versions do 64-bytes of message at a time */
175 #define nh_vmac_nhbytes(mp, kp, nw, rh, rl)				\
176 	do {								\
177 		int i; u64 th, tl;					\
178 		rh = rl = 0;						\
179 		for (i = 0; i < nw; i += 8) {				\
180 			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
181 				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
182 			ADD128(rh, rl, th, tl);				\
183 			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2],	\
184 				pe64_to_cpup((mp)+i+3)+(kp)[i+3]);	\
185 			ADD128(rh, rl, th, tl);				\
186 			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4],	\
187 				pe64_to_cpup((mp)+i+5)+(kp)[i+5]);	\
188 			ADD128(rh, rl, th, tl);				\
189 			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6],	\
190 				pe64_to_cpup((mp)+i+7)+(kp)[i+7]);	\
191 			ADD128(rh, rl, th, tl);				\
192 		}							\
193 	} while (0)
194 
195 #define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh1, rl1)			\
196 	do {								\
197 		int i; u64 th, tl;					\
198 		rh1 = rl1 = rh = rl = 0;				\
199 		for (i = 0; i < nw; i += 8) {				\
200 			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i],	\
201 				pe64_to_cpup((mp)+i+1)+(kp)[i+1]);	\
202 			ADD128(rh, rl, th, tl);				\
203 			MUL64(th, tl, pe64_to_cpup((mp)+i)+(kp)[i+2],	\
204 				pe64_to_cpup((mp)+i+1)+(kp)[i+3]);	\
205 			ADD128(rh1, rl1, th, tl);			\
206 			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+2],	\
207 				pe64_to_cpup((mp)+i+3)+(kp)[i+3]);	\
208 			ADD128(rh, rl, th, tl);				\
209 			MUL64(th, tl, pe64_to_cpup((mp)+i+2)+(kp)[i+4],	\
210 				pe64_to_cpup((mp)+i+3)+(kp)[i+5]);	\
211 			ADD128(rh1, rl1, th, tl);			\
212 			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+4],	\
213 				pe64_to_cpup((mp)+i+5)+(kp)[i+5]);	\
214 			ADD128(rh, rl, th, tl);				\
215 			MUL64(th, tl, pe64_to_cpup((mp)+i+4)+(kp)[i+6],	\
216 				pe64_to_cpup((mp)+i+5)+(kp)[i+7]);	\
217 			ADD128(rh1, rl1, th, tl);			\
218 			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+6],	\
219 				pe64_to_cpup((mp)+i+7)+(kp)[i+7]);	\
220 			ADD128(rh, rl, th, tl);				\
221 			MUL64(th, tl, pe64_to_cpup((mp)+i+6)+(kp)[i+8],	\
222 				pe64_to_cpup((mp)+i+7)+(kp)[i+9]);	\
223 			ADD128(rh1, rl1, th, tl);			\
224 		}							\
225 	} while (0)
226 #endif
227 
228 #define poly_step(ah, al, kh, kl, mh, ml)				\
229 	do {								\
230 		u64 t1h, t1l, t2h, t2l, t3h, t3l, z = 0;		\
231 		/* compute ab*cd, put bd into result registers */	\
232 		PMUL64(t3h, t3l, al, kh);				\
233 		PMUL64(t2h, t2l, ah, kl);				\
234 		PMUL64(t1h, t1l, ah, 2*kh);				\
235 		PMUL64(ah, al, al, kl);					\
236 		/* add 2 * ac to result */				\
237 		ADD128(ah, al, t1h, t1l);				\
238 		/* add together ad + bc */				\
239 		ADD128(t2h, t2l, t3h, t3l);				\
240 		/* now (ah,al), (t2l,2*t2h) need summing */		\
241 		/* first add the high registers, carrying into t2h */	\
242 		ADD128(t2h, ah, z, t2l);				\
243 		/* double t2h and add top bit of ah */			\
244 		t2h = 2 * t2h + (ah >> 63);				\
245 		ah &= m63;						\
246 		/* now add the low registers */				\
247 		ADD128(ah, al, mh, ml);					\
248 		ADD128(ah, al, z, t2h);					\
249 	} while (0)
250 
251 #else /* ! CONFIG_64BIT */
252 
253 #ifndef nh_16
254 #define nh_16(mp, kp, nw, rh, rl)					\
255 	do {								\
256 		u64 t1, t2, m1, m2, t;					\
257 		int i;							\
258 		rh = rl = t = 0;					\
259 		for (i = 0; i < nw; i += 2)  {				\
260 			t1 = pe64_to_cpup(mp+i) + kp[i];		\
261 			t2 = pe64_to_cpup(mp+i+1) + kp[i+1];		\
262 			m2 = MUL32(t1 >> 32, t2);			\
263 			m1 = MUL32(t1, t2 >> 32);			\
264 			ADD128(rh, rl, MUL32(t1 >> 32, t2 >> 32),	\
265 				MUL32(t1, t2));				\
266 			rh += (u64)(u32)(m1 >> 32)			\
267 				+ (u32)(m2 >> 32);			\
268 			t += (u64)(u32)m1 + (u32)m2;			\
269 		}							\
270 		ADD128(rh, rl, (t >> 32), (t << 32));			\
271 	} while (0)
272 #endif
273 
274 static void poly_step_func(u64 *ahi, u64 *alo,
275 			const u64 *kh, const u64 *kl,
276 			const u64 *mh, const u64 *ml)
277 {
278 #define a0 (*(((u32 *)alo)+INDEX_LOW))
279 #define a1 (*(((u32 *)alo)+INDEX_HIGH))
280 #define a2 (*(((u32 *)ahi)+INDEX_LOW))
281 #define a3 (*(((u32 *)ahi)+INDEX_HIGH))
282 #define k0 (*(((u32 *)kl)+INDEX_LOW))
283 #define k1 (*(((u32 *)kl)+INDEX_HIGH))
284 #define k2 (*(((u32 *)kh)+INDEX_LOW))
285 #define k3 (*(((u32 *)kh)+INDEX_HIGH))
286 
287 	u64 p, q, t;
288 	u32 t2;
289 
290 	p = MUL32(a3, k3);
291 	p += p;
292 	p += *(u64 *)mh;
293 	p += MUL32(a0, k2);
294 	p += MUL32(a1, k1);
295 	p += MUL32(a2, k0);
296 	t = (u32)(p);
297 	p >>= 32;
298 	p += MUL32(a0, k3);
299 	p += MUL32(a1, k2);
300 	p += MUL32(a2, k1);
301 	p += MUL32(a3, k0);
302 	t |= ((u64)((u32)p & 0x7fffffff)) << 32;
303 	p >>= 31;
304 	p += (u64)(((u32 *)ml)[INDEX_LOW]);
305 	p += MUL32(a0, k0);
306 	q =  MUL32(a1, k3);
307 	q += MUL32(a2, k2);
308 	q += MUL32(a3, k1);
309 	q += q;
310 	p += q;
311 	t2 = (u32)(p);
312 	p >>= 32;
313 	p += (u64)(((u32 *)ml)[INDEX_HIGH]);
314 	p += MUL32(a0, k1);
315 	p += MUL32(a1, k0);
316 	q =  MUL32(a2, k3);
317 	q += MUL32(a3, k2);
318 	q += q;
319 	p += q;
320 	*(u64 *)(alo) = (p << 32) | t2;
321 	p >>= 32;
322 	*(u64 *)(ahi) = p + t;
323 
324 #undef a0
325 #undef a1
326 #undef a2
327 #undef a3
328 #undef k0
329 #undef k1
330 #undef k2
331 #undef k3
332 }
333 
334 #define poly_step(ah, al, kh, kl, mh, ml)				\
335 	poly_step_func(&(ah), &(al), &(kh), &(kl), &(mh), &(ml))
336 
337 #endif  /* end of specialized NH and poly definitions */
338 
339 /* At least nh_16 is defined. Defined others as needed here */
340 #ifndef nh_16_2
341 #define nh_16_2(mp, kp, nw, rh, rl, rh2, rl2)				\
342 	do { 								\
343 		nh_16(mp, kp, nw, rh, rl);				\
344 		nh_16(mp, ((kp)+2), nw, rh2, rl2);			\
345 	} while (0)
346 #endif
347 #ifndef nh_vmac_nhbytes
348 #define nh_vmac_nhbytes(mp, kp, nw, rh, rl)				\
349 	nh_16(mp, kp, nw, rh, rl)
350 #endif
351 #ifndef nh_vmac_nhbytes_2
352 #define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh2, rl2)			\
353 	do {								\
354 		nh_vmac_nhbytes(mp, kp, nw, rh, rl);			\
355 		nh_vmac_nhbytes(mp, ((kp)+2), nw, rh2, rl2);		\
356 	} while (0)
357 #endif
358 
359 static u64 l3hash(u64 p1, u64 p2, u64 k1, u64 k2, u64 len)
360 {
361 	u64 rh, rl, t, z = 0;
362 
363 	/* fully reduce (p1,p2)+(len,0) mod p127 */
364 	t = p1 >> 63;
365 	p1 &= m63;
366 	ADD128(p1, p2, len, t);
367 	/* At this point, (p1,p2) is at most 2^127+(len<<64) */
368 	t = (p1 > m63) + ((p1 == m63) && (p2 == m64));
369 	ADD128(p1, p2, z, t);
370 	p1 &= m63;
371 
372 	/* compute (p1,p2)/(2^64-2^32) and (p1,p2)%(2^64-2^32) */
373 	t = p1 + (p2 >> 32);
374 	t += (t >> 32);
375 	t += (u32)t > 0xfffffffeu;
376 	p1 += (t >> 32);
377 	p2 += (p1 << 32);
378 
379 	/* compute (p1+k1)%p64 and (p2+k2)%p64 */
380 	p1 += k1;
381 	p1 += (0 - (p1 < k1)) & 257;
382 	p2 += k2;
383 	p2 += (0 - (p2 < k2)) & 257;
384 
385 	/* compute (p1+k1)*(p2+k2)%p64 */
386 	MUL64(rh, rl, p1, p2);
387 	t = rh >> 56;
388 	ADD128(t, rl, z, rh);
389 	rh <<= 8;
390 	ADD128(t, rl, z, rh);
391 	t += t << 8;
392 	rl += t;
393 	rl += (0 - (rl < t)) & 257;
394 	rl += (0 - (rl > p64-1)) & 257;
395 	return rl;
396 }
397 
398 /* L1 and L2-hash one or more VMAC_NHBYTES-byte blocks */
399 static void vhash_blocks(const struct vmac_tfm_ctx *tctx,
400 			 struct vmac_desc_ctx *dctx,
401 			 const __le64 *mptr, unsigned int blocks)
402 {
403 	const u64 *kptr = tctx->nhkey;
404 	const u64 pkh = tctx->polykey[0];
405 	const u64 pkl = tctx->polykey[1];
406 	u64 ch = dctx->polytmp[0];
407 	u64 cl = dctx->polytmp[1];
408 	u64 rh, rl;
409 
410 	if (!dctx->first_block_processed) {
411 		dctx->first_block_processed = true;
412 		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
413 		rh &= m62;
414 		ADD128(ch, cl, rh, rl);
415 		mptr += (VMAC_NHBYTES/sizeof(u64));
416 		blocks--;
417 	}
418 
419 	while (blocks--) {
420 		nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
421 		rh &= m62;
422 		poly_step(ch, cl, pkh, pkl, rh, rl);
423 		mptr += (VMAC_NHBYTES/sizeof(u64));
424 	}
425 
426 	dctx->polytmp[0] = ch;
427 	dctx->polytmp[1] = cl;
428 }
429 
430 static int vmac_setkey(struct crypto_shash *tfm,
431 		       const u8 *key, unsigned int keylen)
432 {
433 	struct vmac_tfm_ctx *tctx = crypto_shash_ctx(tfm);
434 	__be64 out[2];
435 	u8 in[16] = { 0 };
436 	unsigned int i;
437 	int err;
438 
439 	if (keylen != VMAC_KEY_LEN)
440 		return -EINVAL;
441 
442 	err = crypto_cipher_setkey(tctx->cipher, key, keylen);
443 	if (err)
444 		return err;
445 
446 	/* Fill nh key */
447 	in[0] = 0x80;
448 	for (i = 0; i < ARRAY_SIZE(tctx->nhkey); i += 2) {
449 		crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
450 		tctx->nhkey[i] = be64_to_cpu(out[0]);
451 		tctx->nhkey[i+1] = be64_to_cpu(out[1]);
452 		in[15]++;
453 	}
454 
455 	/* Fill poly key */
456 	in[0] = 0xC0;
457 	in[15] = 0;
458 	for (i = 0; i < ARRAY_SIZE(tctx->polykey); i += 2) {
459 		crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
460 		tctx->polykey[i] = be64_to_cpu(out[0]) & mpoly;
461 		tctx->polykey[i+1] = be64_to_cpu(out[1]) & mpoly;
462 		in[15]++;
463 	}
464 
465 	/* Fill ip key */
466 	in[0] = 0xE0;
467 	in[15] = 0;
468 	for (i = 0; i < ARRAY_SIZE(tctx->l3key); i += 2) {
469 		do {
470 			crypto_cipher_encrypt_one(tctx->cipher, (u8 *)out, in);
471 			tctx->l3key[i] = be64_to_cpu(out[0]);
472 			tctx->l3key[i+1] = be64_to_cpu(out[1]);
473 			in[15]++;
474 		} while (tctx->l3key[i] >= p64 || tctx->l3key[i+1] >= p64);
475 	}
476 
477 	return 0;
478 }
479 
480 static int vmac_init(struct shash_desc *desc)
481 {
482 	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
483 	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
484 
485 	dctx->partial_size = 0;
486 	dctx->first_block_processed = false;
487 	memcpy(dctx->polytmp, tctx->polykey, sizeof(dctx->polytmp));
488 	dctx->nonce_size = 0;
489 	return 0;
490 }
491 
492 static int vmac_update(struct shash_desc *desc, const u8 *p, unsigned int len)
493 {
494 	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
495 	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
496 	unsigned int n;
497 
498 	/* Nonce is passed as first VMAC_NONCEBYTES bytes of data */
499 	if (dctx->nonce_size < VMAC_NONCEBYTES) {
500 		n = min(len, VMAC_NONCEBYTES - dctx->nonce_size);
501 		memcpy(&dctx->nonce.bytes[dctx->nonce_size], p, n);
502 		dctx->nonce_size += n;
503 		p += n;
504 		len -= n;
505 	}
506 
507 	if (dctx->partial_size) {
508 		n = min(len, VMAC_NHBYTES - dctx->partial_size);
509 		memcpy(&dctx->partial[dctx->partial_size], p, n);
510 		dctx->partial_size += n;
511 		p += n;
512 		len -= n;
513 		if (dctx->partial_size == VMAC_NHBYTES) {
514 			vhash_blocks(tctx, dctx, dctx->partial_words, 1);
515 			dctx->partial_size = 0;
516 		}
517 	}
518 
519 	if (len >= VMAC_NHBYTES) {
520 		n = round_down(len, VMAC_NHBYTES);
521 		/* TODO: 'p' may be misaligned here */
522 		vhash_blocks(tctx, dctx, (const __le64 *)p, n / VMAC_NHBYTES);
523 		p += n;
524 		len -= n;
525 	}
526 
527 	if (len) {
528 		memcpy(dctx->partial, p, len);
529 		dctx->partial_size = len;
530 	}
531 
532 	return 0;
533 }
534 
535 static u64 vhash_final(const struct vmac_tfm_ctx *tctx,
536 		       struct vmac_desc_ctx *dctx)
537 {
538 	unsigned int partial = dctx->partial_size;
539 	u64 ch = dctx->polytmp[0];
540 	u64 cl = dctx->polytmp[1];
541 
542 	/* L1 and L2-hash the final block if needed */
543 	if (partial) {
544 		/* Zero-pad to next 128-bit boundary */
545 		unsigned int n = round_up(partial, 16);
546 		u64 rh, rl;
547 
548 		memset(&dctx->partial[partial], 0, n - partial);
549 		nh_16(dctx->partial_words, tctx->nhkey, n / 8, rh, rl);
550 		rh &= m62;
551 		if (dctx->first_block_processed)
552 			poly_step(ch, cl, tctx->polykey[0], tctx->polykey[1],
553 				  rh, rl);
554 		else
555 			ADD128(ch, cl, rh, rl);
556 	}
557 
558 	/* L3-hash the 128-bit output of L2-hash */
559 	return l3hash(ch, cl, tctx->l3key[0], tctx->l3key[1], partial * 8);
560 }
561 
562 static int vmac_final(struct shash_desc *desc, u8 *out)
563 {
564 	const struct vmac_tfm_ctx *tctx = crypto_shash_ctx(desc->tfm);
565 	struct vmac_desc_ctx *dctx = shash_desc_ctx(desc);
566 	int index;
567 	u64 hash, pad;
568 
569 	if (dctx->nonce_size != VMAC_NONCEBYTES)
570 		return -EINVAL;
571 
572 	/*
573 	 * The VMAC specification requires a nonce at least 1 bit shorter than
574 	 * the block cipher's block length, so we actually only accept a 127-bit
575 	 * nonce.  We define the unused bit to be the first one and require that
576 	 * it be 0, so the needed prepending of a 0 bit is implicit.
577 	 */
578 	if (dctx->nonce.bytes[0] & 0x80)
579 		return -EINVAL;
580 
581 	/* Finish calculating the VHASH of the message */
582 	hash = vhash_final(tctx, dctx);
583 
584 	/* Generate pseudorandom pad by encrypting the nonce */
585 	BUILD_BUG_ON(VMAC_NONCEBYTES != 2 * (VMAC_TAG_LEN / 8));
586 	index = dctx->nonce.bytes[VMAC_NONCEBYTES - 1] & 1;
587 	dctx->nonce.bytes[VMAC_NONCEBYTES - 1] &= ~1;
588 	crypto_cipher_encrypt_one(tctx->cipher, dctx->nonce.bytes,
589 				  dctx->nonce.bytes);
590 	pad = be64_to_cpu(dctx->nonce.pads[index]);
591 
592 	/* The VMAC is the sum of VHASH and the pseudorandom pad */
593 	put_unaligned_be64(hash + pad, out);
594 	return 0;
595 }
596 
597 static int vmac_init_tfm(struct crypto_tfm *tfm)
598 {
599 	struct crypto_instance *inst = crypto_tfm_alg_instance(tfm);
600 	struct crypto_cipher_spawn *spawn = crypto_instance_ctx(inst);
601 	struct vmac_tfm_ctx *tctx = crypto_tfm_ctx(tfm);
602 	struct crypto_cipher *cipher;
603 
604 	cipher = crypto_spawn_cipher(spawn);
605 	if (IS_ERR(cipher))
606 		return PTR_ERR(cipher);
607 
608 	tctx->cipher = cipher;
609 	return 0;
610 }
611 
612 static void vmac_exit_tfm(struct crypto_tfm *tfm)
613 {
614 	struct vmac_tfm_ctx *tctx = crypto_tfm_ctx(tfm);
615 
616 	crypto_free_cipher(tctx->cipher);
617 }
618 
619 static int vmac_create(struct crypto_template *tmpl, struct rtattr **tb)
620 {
621 	struct shash_instance *inst;
622 	struct crypto_cipher_spawn *spawn;
623 	struct crypto_alg *alg;
624 	u32 mask;
625 	int err;
626 
627 	err = crypto_check_attr_type(tb, CRYPTO_ALG_TYPE_SHASH, &mask);
628 	if (err)
629 		return err;
630 
631 	inst = kzalloc(sizeof(*inst) + sizeof(*spawn), GFP_KERNEL);
632 	if (!inst)
633 		return -ENOMEM;
634 	spawn = shash_instance_ctx(inst);
635 
636 	err = crypto_grab_cipher(spawn, shash_crypto_instance(inst),
637 				 crypto_attr_alg_name(tb[1]), 0, mask);
638 	if (err)
639 		goto err_free_inst;
640 	alg = crypto_spawn_cipher_alg(spawn);
641 
642 	err = -EINVAL;
643 	if (alg->cra_blocksize != VMAC_NONCEBYTES)
644 		goto err_free_inst;
645 
646 	err = crypto_inst_setname(shash_crypto_instance(inst), tmpl->name, alg);
647 	if (err)
648 		goto err_free_inst;
649 
650 	inst->alg.base.cra_priority = alg->cra_priority;
651 	inst->alg.base.cra_blocksize = alg->cra_blocksize;
652 	inst->alg.base.cra_alignmask = alg->cra_alignmask;
653 
654 	inst->alg.base.cra_ctxsize = sizeof(struct vmac_tfm_ctx);
655 	inst->alg.base.cra_init = vmac_init_tfm;
656 	inst->alg.base.cra_exit = vmac_exit_tfm;
657 
658 	inst->alg.descsize = sizeof(struct vmac_desc_ctx);
659 	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 	inst->alg.setkey = vmac_setkey;
664 
665 	inst->free = shash_free_singlespawn_instance;
666 
667 	err = shash_register_instance(tmpl, inst);
668 	if (err) {
669 err_free_inst:
670 		shash_free_singlespawn_instance(inst);
671 	}
672 	return err;
673 }
674 
675 static struct crypto_template vmac64_tmpl = {
676 	.name = "vmac64",
677 	.create = vmac_create,
678 	.module = THIS_MODULE,
679 };
680 
681 static int __init vmac_module_init(void)
682 {
683 	return crypto_register_template(&vmac64_tmpl);
684 }
685 
686 static void __exit vmac_module_exit(void)
687 {
688 	crypto_unregister_template(&vmac64_tmpl);
689 }
690 
691 subsys_initcall(vmac_module_init);
692 module_exit(vmac_module_exit);
693 
694 MODULE_LICENSE("GPL");
695 MODULE_DESCRIPTION("VMAC hash algorithm");
696 MODULE_ALIAS_CRYPTO("vmac64");
697 MODULE_IMPORT_NS(CRYPTO_INTERNAL);
698