tools: Add sparse context/locking annotations in compiler-types.h
[linux-2.6-microblaze.git] / crypto / vmac.c
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);