crypto: vmac - New hash algorithm for intel_txt support
Shane Wang [Wed, 2 Sep 2009 10:05:22 +0000 (20:05 +1000)]
This patch adds VMAC (a fast MAC) support into crypto framework.

Signed-off-by: Shane Wang <shane.wang@intel.com>
Signed-off-by: Joseph Cihula <joseph.cihula@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

crypto/Kconfig
crypto/Makefile
crypto/tcrypt.c
crypto/testmgr.c
crypto/testmgr.h
crypto/vmac.c [new file with mode: 0644]
include/crypto/vmac.h [new file with mode: 0644]

index 7623442..26b5dd0 100644 (file)
@@ -269,6 +269,18 @@ config CRYPTO_XCBC
                http://csrc.nist.gov/encryption/modes/proposedmodes/
                 xcbc-mac/xcbc-mac-spec.pdf
 
+config CRYPTO_VMAC
+       tristate "VMAC support"
+       depends on EXPERIMENTAL
+       select CRYPTO_HASH
+       select CRYPTO_MANAGER
+       help
+         VMAC is a message authentication algorithm designed for
+         very high speed on 64-bit architectures.
+
+         See also:
+         <http://fastcrypto.org/vmac>
+
 comment "Digest"
 
 config CRYPTO_CRC32C
index c2ca721..9e8f619 100644 (file)
@@ -32,6 +32,7 @@ cryptomgr-objs := algboss.o testmgr.o
 
 obj-$(CONFIG_CRYPTO_MANAGER2) += cryptomgr.o
 obj-$(CONFIG_CRYPTO_HMAC) += hmac.o
+obj-$(CONFIG_CRYPTO_VMAC) += vmac.o
 obj-$(CONFIG_CRYPTO_XCBC) += xcbc.o
 obj-$(CONFIG_CRYPTO_NULL) += crypto_null.o
 obj-$(CONFIG_CRYPTO_MD4) += md4.o
index 5a375e8..aa3f84c 100644 (file)
@@ -719,6 +719,10 @@ static int do_test(int m)
                ret += tcrypt_test("hmac(rmd160)");
                break;
 
+       case 109:
+               ret += tcrypt_test("vmac(aes)");
+               break;
+
        case 150:
                ret += tcrypt_test("ansi_cprng");
                break;
index 29b228d..6d5b746 100644 (file)
@@ -2248,6 +2248,15 @@ static const struct alg_test_desc alg_test_descs[] = {
                        }
                }
        }, {
+               .alg = "vmac(aes)",
+               .test = alg_test_hash,
+               .suite = {
+                       .hash = {
+                               .vecs = aes_vmac128_tv_template,
+                               .count = VMAC_AES_TEST_VECTORS
+                       }
+               }
+       }, {
                .alg = "wp256",
                .test = alg_test_hash,
                .suite = {
index 6931622..9963b18 100644 (file)
@@ -1654,6 +1654,22 @@ static struct hash_testvec aes_xcbc128_tv_template[] = {
        }
 };
 
+#define VMAC_AES_TEST_VECTORS  1
+static char vmac_string[128] = {'\x01', '\x01', '\x01', '\x01',
+                               '\x02', '\x03', '\x02', '\x02',
+                               '\x02', '\x04', '\x01', '\x07',
+                               '\x04', '\x01', '\x04', '\x03',};
+static struct hash_testvec aes_vmac128_tv_template[] = {
+       {
+               .key    = "\x00\x01\x02\x03\x04\x05\x06\x07"
+                         "\x08\x09\x0a\x0b\x0c\x0d\x0e\x0f",
+               .plaintext = vmac_string,
+               .digest = "\xcb\xd7\x8a\xfd\xb7\x33\x79\xe7",
+               .psize  = 128,
+               .ksize  = 16,
+       },
+};
+
 /*
  * SHA384 HMAC test vectors from RFC4231
  */
diff --git a/crypto/vmac.c b/crypto/vmac.c
new file mode 100644 (file)
index 0000000..0a9468e
--- /dev/null
@@ -0,0 +1,678 @@
+/*
+ * Modified to interface to the Linux kernel
+ * Copyright (c) 2009, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ */
+
+/* --------------------------------------------------------------------------
+ * VMAC and VHASH Implementation by Ted Krovetz (tdk@acm.org) and Wei Dai.
+ * This implementation is herby placed in the public domain.
+ * The authors offers no warranty. Use at your own risk.
+ * Please send bug reports to the authors.
+ * Last modified: 17 APR 08, 1700 PDT
+ * ----------------------------------------------------------------------- */
+
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/crypto.h>
+#include <linux/scatterlist.h>
+#include <asm/byteorder.h>
+#include <crypto/scatterwalk.h>
+#include <crypto/vmac.h>
+#include <crypto/internal/hash.h>
+
+/*
+ * Constants and masks
+ */
+#define UINT64_C(x) x##ULL
+const u64 p64   = UINT64_C(0xfffffffffffffeff);  /* 2^64 - 257 prime  */
+const u64 m62   = UINT64_C(0x3fffffffffffffff);  /* 62-bit mask       */
+const u64 m63   = UINT64_C(0x7fffffffffffffff);  /* 63-bit mask       */
+const u64 m64   = UINT64_C(0xffffffffffffffff);  /* 64-bit mask       */
+const u64 mpoly = UINT64_C(0x1fffffff1fffffff);  /* Poly key mask     */
+
+#ifdef __LITTLE_ENDIAN
+#define INDEX_HIGH 1
+#define INDEX_LOW 0
+#else
+#define INDEX_HIGH 0
+#define INDEX_LOW 1
+#endif
+
+/*
+ * The following routines are used in this implementation. They are
+ * written via macros to simulate zero-overhead call-by-reference.
+ *
+ * MUL64: 64x64->128-bit multiplication
+ * PMUL64: assumes top bits cleared on inputs
+ * ADD128: 128x128->128-bit addition
+ */
+
+#define ADD128(rh, rl, ih, il)                                         \
+       do {                                                            \
+               u64 _il = (il);                                         \
+               (rl) += (_il);                                          \
+               if ((rl) < (_il))                                       \
+                       (rh)++;                                         \
+               (rh) += (ih);                                           \
+       } while (0)
+
+#define MUL32(i1, i2)  ((u64)(u32)(i1)*(u32)(i2))
+
+#define PMUL64(rh, rl, i1, i2) /* Assumes m doesn't overflow */        \
+       do {                                                            \
+               u64 _i1 = (i1), _i2 = (i2);                             \
+               u64 m = MUL32(_i1, _i2>>32) + MUL32(_i1>>32, _i2);      \
+               rh = MUL32(_i1>>32, _i2>>32);                           \
+               rl = MUL32(_i1, _i2);                                   \
+               ADD128(rh, rl, (m >> 32), (m << 32));                   \
+       } while (0)
+
+#define MUL64(rh, rl, i1, i2)                                          \
+       do {                                                            \
+               u64 _i1 = (i1), _i2 = (i2);                             \
+               u64 m1 = MUL32(_i1, _i2>>32);                           \
+               u64 m2 = MUL32(_i1>>32, _i2);                           \
+               rh = MUL32(_i1>>32, _i2>>32);                           \
+               rl = MUL32(_i1, _i2);                                   \
+               ADD128(rh, rl, (m1 >> 32), (m1 << 32));                 \
+               ADD128(rh, rl, (m2 >> 32), (m2 << 32));                 \
+       } while (0)
+
+/*
+ * For highest performance the L1 NH and L2 polynomial hashes should be
+ * carefully implemented to take advantage of one's target architechture.
+ * Here these two hash functions are defined multiple time; once for
+ * 64-bit architectures, once for 32-bit SSE2 architectures, and once
+ * for the rest (32-bit) architectures.
+ * For each, nh_16 *must* be defined (works on multiples of 16 bytes).
+ * Optionally, nh_vmac_nhbytes can be defined (for multiples of
+ * VMAC_NHBYTES), and nh_16_2 and nh_vmac_nhbytes_2 (versions that do two
+ * NH computations at once).
+ */
+
+#ifdef CONFIG_64BIT
+
+#define nh_16(mp, kp, nw, rh, rl)                                      \
+       do {                                                            \
+               int i; u64 th, tl;                                      \
+               rh = rl = 0;                                            \
+               for (i = 0; i < nw; i += 2) {                           \
+                       MUL64(th, tl, le64_to_cpup((mp)+i)+(kp)[i],     \
+                               le64_to_cpup((mp)+i+1)+(kp)[i+1]);      \
+                       ADD128(rh, rl, th, tl);                         \
+               }                                                       \
+       } while (0)
+
+#define nh_16_2(mp, kp, nw, rh, rl, rh1, rl1)                          \
+       do {                                                            \
+               int i; u64 th, tl;                                      \
+               rh1 = rl1 = rh = rl = 0;                                \
+               for (i = 0; i < nw; i += 2) {                           \
+                       MUL64(th, tl, le64_to_cpup((mp)+i)+(kp)[i],     \
+                               le64_to_cpup((mp)+i+1)+(kp)[i+1]);      \
+                       ADD128(rh, rl, th, tl);                         \
+                       MUL64(th, tl, le64_to_cpup((mp)+i)+(kp)[i+2],   \
+                               le64_to_cpup((mp)+i+1)+(kp)[i+3]);      \
+                       ADD128(rh1, rl1, th, tl);                       \
+               }                                                       \
+       } while (0)
+
+#if (VMAC_NHBYTES >= 64) /* These versions do 64-bytes of message at a time */
+#define nh_vmac_nhbytes(mp, kp, nw, rh, rl)                            \
+       do {                                                            \
+               int i; u64 th, tl;                                      \
+               rh = rl = 0;                                            \
+               for (i = 0; i < nw; i += 8) {                           \
+                       MUL64(th, tl, le64_to_cpup((mp)+i)+(kp)[i],     \
+                               le64_to_cpup((mp)+i+1)+(kp)[i+1]);      \
+                       ADD128(rh, rl, th, tl);                         \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+2)+(kp)[i+2], \
+                               le64_to_cpup((mp)+i+3)+(kp)[i+3]);      \
+                       ADD128(rh, rl, th, tl);                         \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+4)+(kp)[i+4], \
+                               le64_to_cpup((mp)+i+5)+(kp)[i+5]);      \
+                       ADD128(rh, rl, th, tl);                         \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+6)+(kp)[i+6], \
+                               le64_to_cpup((mp)+i+7)+(kp)[i+7]);      \
+                       ADD128(rh, rl, th, tl);                         \
+               }                                                       \
+       } while (0)
+
+#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh1, rl1)                        \
+       do {                                                            \
+               int i; u64 th, tl;                                      \
+               rh1 = rl1 = rh = rl = 0;                                \
+               for (i = 0; i < nw; i += 8) {                           \
+                       MUL64(th, tl, le64_to_cpup((mp)+i)+(kp)[i],     \
+                               le64_to_cpup((mp)+i+1)+(kp)[i+1]);      \
+                       ADD128(rh, rl, th, tl);                         \
+                       MUL64(th, tl, le64_to_cpup((mp)+i)+(kp)[i+2],   \
+                               le64_to_cpup((mp)+i+1)+(kp)[i+3]);      \
+                       ADD128(rh1, rl1, th, tl);                       \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+2)+(kp)[i+2], \
+                               le64_to_cpup((mp)+i+3)+(kp)[i+3]);      \
+                       ADD128(rh, rl, th, tl);                         \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+2)+(kp)[i+4], \
+                               le64_to_cpup((mp)+i+3)+(kp)[i+5]);      \
+                       ADD128(rh1, rl1, th, tl);                       \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+4)+(kp)[i+4], \
+                               le64_to_cpup((mp)+i+5)+(kp)[i+5]);      \
+                       ADD128(rh, rl, th, tl);                         \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+4)+(kp)[i+6], \
+                               le64_to_cpup((mp)+i+5)+(kp)[i+7]);      \
+                       ADD128(rh1, rl1, th, tl);                       \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+6)+(kp)[i+6], \
+                               le64_to_cpup((mp)+i+7)+(kp)[i+7]);      \
+                       ADD128(rh, rl, th, tl);                         \
+                       MUL64(th, tl, le64_to_cpup((mp)+i+6)+(kp)[i+8], \
+                               le64_to_cpup((mp)+i+7)+(kp)[i+9]);      \
+                       ADD128(rh1, rl1, th, tl);                       \
+               }                                                       \
+       } while (0)
+#endif
+
+#define poly_step(ah, al, kh, kl, mh, ml)                              \
+       do {                                                            \
+               u64 t1h, t1l, t2h, t2l, t3h, t3l, z = 0;                \
+               /* compute ab*cd, put bd into result registers */       \
+               PMUL64(t3h, t3l, al, kh);                               \
+               PMUL64(t2h, t2l, ah, kl);                               \
+               PMUL64(t1h, t1l, ah, 2*kh);                             \
+               PMUL64(ah, al, al, kl);                                 \
+               /* add 2 * ac to result */                              \
+               ADD128(ah, al, t1h, t1l);                               \
+               /* add together ad + bc */                              \
+               ADD128(t2h, t2l, t3h, t3l);                             \
+               /* now (ah,al), (t2l,2*t2h) need summing */             \
+               /* first add the high registers, carrying into t2h */   \
+               ADD128(t2h, ah, z, t2l);                                \
+               /* double t2h and add top bit of ah */                  \
+               t2h = 2 * t2h + (ah >> 63);                             \
+               ah &= m63;                                              \
+               /* now add the low registers */                         \
+               ADD128(ah, al, mh, ml);                                 \
+               ADD128(ah, al, z, t2h);                                 \
+       } while (0)
+
+#else /* ! CONFIG_64BIT */
+
+#ifndef nh_16
+#define nh_16(mp, kp, nw, rh, rl)                                      \
+       do {                                                            \
+               u64 t1, t2, m1, m2, t;                                  \
+               int i;                                                  \
+               rh = rl = t = 0;                                        \
+               for (i = 0; i < nw; i += 2)  {                          \
+                       t1 = le64_to_cpup(mp+i) + kp[i];                \
+                       t2 = le64_to_cpup(mp+i+1) + kp[i+1];            \
+                       m2 = MUL32(t1 >> 32, t2);                       \
+                       m1 = MUL32(t1, t2 >> 32);                       \
+                       ADD128(rh, rl, MUL32(t1 >> 32, t2 >> 32),       \
+                               MUL32(t1, t2));                         \
+                       rh += (u64)(u32)(m1 >> 32)                      \
+                               + (u32)(m2 >> 32);                      \
+                       t += (u64)(u32)m1 + (u32)m2;                    \
+               }                                                       \
+               ADD128(rh, rl, (t >> 32), (t << 32));                   \
+       } while (0)
+#endif
+
+static void poly_step_func(u64 *ahi, u64 *alo,
+                       const u64 *kh, const u64 *kl,
+                       const u64 *mh, const u64 *ml)
+{
+#define a0 (*(((u32 *)alo)+INDEX_LOW))
+#define a1 (*(((u32 *)alo)+INDEX_HIGH))
+#define a2 (*(((u32 *)ahi)+INDEX_LOW))
+#define a3 (*(((u32 *)ahi)+INDEX_HIGH))
+#define k0 (*(((u32 *)kl)+INDEX_LOW))
+#define k1 (*(((u32 *)kl)+INDEX_HIGH))
+#define k2 (*(((u32 *)kh)+INDEX_LOW))
+#define k3 (*(((u32 *)kh)+INDEX_HIGH))
+
+       u64 p, q, t;
+       u32 t2;
+
+       p = MUL32(a3, k3);
+       p += p;
+       p += *(u64 *)mh;
+       p += MUL32(a0, k2);
+       p += MUL32(a1, k1);
+       p += MUL32(a2, k0);
+       t = (u32)(p);
+       p >>= 32;
+       p += MUL32(a0, k3);
+       p += MUL32(a1, k2);
+       p += MUL32(a2, k1);
+       p += MUL32(a3, k0);
+       t |= ((u64)((u32)p & 0x7fffffff)) << 32;
+       p >>= 31;
+       p += (u64)(((u32 *)ml)[INDEX_LOW]);
+       p += MUL32(a0, k0);
+       q =  MUL32(a1, k3);
+       q += MUL32(a2, k2);
+       q += MUL32(a3, k1);
+       q += q;
+       p += q;
+       t2 = (u32)(p);
+       p >>= 32;
+       p += (u64)(((u32 *)ml)[INDEX_HIGH]);
+       p += MUL32(a0, k1);
+       p += MUL32(a1, k0);
+       q =  MUL32(a2, k3);
+       q += MUL32(a3, k2);
+       q += q;
+       p += q;
+       *(u64 *)(alo) = (p << 32) | t2;
+       p >>= 32;
+       *(u64 *)(ahi) = p + t;
+
+#undef a0
+#undef a1
+#undef a2
+#undef a3
+#undef k0
+#undef k1
+#undef k2
+#undef k3
+}
+
+#define poly_step(ah, al, kh, kl, mh, ml)                              \
+       poly_step_func(&(ah), &(al), &(kh), &(kl), &(mh), &(ml))
+
+#endif  /* end of specialized NH and poly definitions */
+
+/* At least nh_16 is defined. Defined others as needed here */
+#ifndef nh_16_2
+#define nh_16_2(mp, kp, nw, rh, rl, rh2, rl2)                          \
+       do {                                                            \
+               nh_16(mp, kp, nw, rh, rl);                              \
+               nh_16(mp, ((kp)+2), nw, rh2, rl2);                      \
+       } while (0)
+#endif
+#ifndef nh_vmac_nhbytes
+#define nh_vmac_nhbytes(mp, kp, nw, rh, rl)                            \
+       nh_16(mp, kp, nw, rh, rl)
+#endif
+#ifndef nh_vmac_nhbytes_2
+#define nh_vmac_nhbytes_2(mp, kp, nw, rh, rl, rh2, rl2)                        \
+       do {                                                            \
+               nh_vmac_nhbytes(mp, kp, nw, rh, rl);                    \
+               nh_vmac_nhbytes(mp, ((kp)+2), nw, rh2, rl2);            \
+       } while (0)
+#endif
+
+static void vhash_abort(struct vmac_ctx *ctx)
+{
+       ctx->polytmp[0] = ctx->polykey[0] ;
+       ctx->polytmp[1] = ctx->polykey[1] ;
+       ctx->first_block_processed = 0;
+}
+
+static u64 l3hash(u64 p1, u64 p2,
+                       u64 k1, u64 k2, u64 len)
+{
+       u64 rh, rl, t, z = 0;
+
+       /* fully reduce (p1,p2)+(len,0) mod p127 */
+       t = p1 >> 63;
+       p1 &= m63;
+       ADD128(p1, p2, len, t);
+       /* At this point, (p1,p2) is at most 2^127+(len<<64) */
+       t = (p1 > m63) + ((p1 == m63) && (p2 == m64));
+       ADD128(p1, p2, z, t);
+       p1 &= m63;
+
+       /* compute (p1,p2)/(2^64-2^32) and (p1,p2)%(2^64-2^32) */
+       t = p1 + (p2 >> 32);
+       t += (t >> 32);
+       t += (u32)t > 0xfffffffeu;
+       p1 += (t >> 32);
+       p2 += (p1 << 32);
+
+       /* compute (p1+k1)%p64 and (p2+k2)%p64 */
+       p1 += k1;
+       p1 += (0 - (p1 < k1)) & 257;
+       p2 += k2;
+       p2 += (0 - (p2 < k2)) & 257;
+
+       /* compute (p1+k1)*(p2+k2)%p64 */
+       MUL64(rh, rl, p1, p2);
+       t = rh >> 56;
+       ADD128(t, rl, z, rh);
+       rh <<= 8;
+       ADD128(t, rl, z, rh);
+       t += t << 8;
+       rl += t;
+       rl += (0 - (rl < t)) & 257;
+       rl += (0 - (rl > p64-1)) & 257;
+       return rl;
+}
+
+static void vhash_update(const unsigned char *m,
+                       unsigned int mbytes, /* Pos multiple of VMAC_NHBYTES */
+                       struct vmac_ctx *ctx)
+{
+       u64 rh, rl, *mptr;
+       const u64 *kptr = (u64 *)ctx->nhkey;
+       int i;
+       u64 ch, cl;
+       u64 pkh = ctx->polykey[0];
+       u64 pkl = ctx->polykey[1];
+
+       mptr = (u64 *)m;
+       i = mbytes / VMAC_NHBYTES;  /* Must be non-zero */
+
+       ch = ctx->polytmp[0];
+       cl = ctx->polytmp[1];
+
+       if (!ctx->first_block_processed) {
+               ctx->first_block_processed = 1;
+               nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
+               rh &= m62;
+               ADD128(ch, cl, rh, rl);
+               mptr += (VMAC_NHBYTES/sizeof(u64));
+               i--;
+       }
+
+       while (i--) {
+               nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
+               rh &= m62;
+               poly_step(ch, cl, pkh, pkl, rh, rl);
+               mptr += (VMAC_NHBYTES/sizeof(u64));
+       }
+
+       ctx->polytmp[0] = ch;
+       ctx->polytmp[1] = cl;
+}
+
+static u64 vhash(unsigned char m[], unsigned int mbytes,
+                       u64 *tagl, struct vmac_ctx *ctx)
+{
+       u64 rh, rl, *mptr;
+       const u64 *kptr = (u64 *)ctx->nhkey;
+       int i, remaining;
+       u64 ch, cl;
+       u64 pkh = ctx->polykey[0];
+       u64 pkl = ctx->polykey[1];
+
+       mptr = (u64 *)m;
+       i = mbytes / VMAC_NHBYTES;
+       remaining = mbytes % VMAC_NHBYTES;
+
+       if (ctx->first_block_processed) {
+               ch = ctx->polytmp[0];
+               cl = ctx->polytmp[1];
+       } else if (i) {
+               nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, ch, cl);
+               ch &= m62;
+               ADD128(ch, cl, pkh, pkl);
+               mptr += (VMAC_NHBYTES/sizeof(u64));
+               i--;
+       } else if (remaining) {
+               nh_16(mptr, kptr, 2*((remaining+15)/16), ch, cl);
+               ch &= m62;
+               ADD128(ch, cl, pkh, pkl);
+               mptr += (VMAC_NHBYTES/sizeof(u64));
+               goto do_l3;
+       } else {/* Empty String */
+               ch = pkh; cl = pkl;
+               goto do_l3;
+       }
+
+       while (i--) {
+               nh_vmac_nhbytes(mptr, kptr, VMAC_NHBYTES/8, rh, rl);
+               rh &= m62;
+               poly_step(ch, cl, pkh, pkl, rh, rl);
+               mptr += (VMAC_NHBYTES/sizeof(u64));
+       }
+       if (remaining) {
+               nh_16(mptr, kptr, 2*((remaining+15)/16), rh, rl);
+               rh &= m62;
+               poly_step(ch, cl, pkh, pkl, rh, rl);
+       }
+
+do_l3:
+       vhash_abort(ctx);
+       remaining *= 8;
+       return l3hash(ch, cl, ctx->l3key[0], ctx->l3key[1], remaining);
+}
+
+static u64 vmac(unsigned char m[], unsigned int mbytes,
+                       unsigned char n[16], u64 *tagl,
+                       struct vmac_ctx_t *ctx)
+{
+       u64 *in_n, *out_p;
+       u64 p, h;
+       int i;
+
+       in_n = ctx->__vmac_ctx.cached_nonce;
+       out_p = ctx->__vmac_ctx.cached_aes;
+
+       i = n[15] & 1;
+       if ((*(u64 *)(n+8) != in_n[1]) || (*(u64 *)(n) != in_n[0])) {
+               in_n[0] = *(u64 *)(n);
+               in_n[1] = *(u64 *)(n+8);
+               ((unsigned char *)in_n)[15] &= 0xFE;
+               crypto_cipher_encrypt_one(ctx->child,
+                       (unsigned char *)out_p, (unsigned char *)in_n);
+
+               ((unsigned char *)in_n)[15] |= (unsigned char)(1-i);
+       }
+       p = be64_to_cpup(out_p + i);
+       h = vhash(m, mbytes, (u64 *)0, &ctx->__vmac_ctx);
+       return p + h;
+}
+
+static int vmac_set_key(unsigned char user_key[], struct vmac_ctx_t *ctx)
+{
+       u64 in[2] = {0}, out[2];
+       unsigned i;
+       int err = 0;
+
+       err = crypto_cipher_setkey(ctx->child, user_key, VMAC_KEY_LEN);
+       if (err)
+               return err;
+
+       /* Fill nh key */
+       ((unsigned char *)in)[0] = 0x80;
+       for (i = 0; i < sizeof(ctx->__vmac_ctx.nhkey)/8; i += 2) {
+               crypto_cipher_encrypt_one(ctx->child,
+                       (unsigned char *)out, (unsigned char *)in);
+               ctx->__vmac_ctx.nhkey[i] = be64_to_cpup(out);
+               ctx->__vmac_ctx.nhkey[i+1] = be64_to_cpup(out+1);
+               ((unsigned char *)in)[15] += 1;
+       }
+
+       /* Fill poly key */
+       ((unsigned char *)in)[0] = 0xC0;
+       in[1] = 0;
+       for (i = 0; i < sizeof(ctx->__vmac_ctx.polykey)/8; i += 2) {
+               crypto_cipher_encrypt_one(ctx->child,
+                       (unsigned char *)out, (unsigned char *)in);
+               ctx->__vmac_ctx.polytmp[i] =
+                       ctx->__vmac_ctx.polykey[i] =
+                               be64_to_cpup(out) & mpoly;
+               ctx->__vmac_ctx.polytmp[i+1] =
+                       ctx->__vmac_ctx.polykey[i+1] =
+                               be64_to_cpup(out+1) & mpoly;
+               ((unsigned char *)in)[15] += 1;
+       }
+
+       /* Fill ip key */
+       ((unsigned char *)in)[0] = 0xE0;
+       in[1] = 0;
+       for (i = 0; i < sizeof(ctx->__vmac_ctx.l3key)/8; i += 2) {
+               do {
+                       crypto_cipher_encrypt_one(ctx->child,
+                               (unsigned char *)out, (unsigned char *)in);
+                       ctx->__vmac_ctx.l3key[i] = be64_to_cpup(out);
+                       ctx->__vmac_ctx.l3key[i+1] = be64_to_cpup(out+1);
+                       ((unsigned char *)in)[15] += 1;
+               } while (ctx->__vmac_ctx.l3key[i] >= p64
+                       || ctx->__vmac_ctx.l3key[i+1] >= p64);
+       }
+
+       /* Invalidate nonce/aes cache and reset other elements */
+       ctx->__vmac_ctx.cached_nonce[0] = (u64)-1; /* Ensure illegal nonce */
+       ctx->__vmac_ctx.cached_nonce[1] = (u64)0;  /* Ensure illegal nonce */
+       ctx->__vmac_ctx.first_block_processed = 0;
+
+       return err;
+}
+
+static int vmac_setkey(struct crypto_shash *parent,
+               const u8 *key, unsigned int keylen)
+{
+       struct vmac_ctx_t *ctx = crypto_shash_ctx(parent);
+
+       if (keylen != VMAC_KEY_LEN) {
+               crypto_shash_set_flags(parent, CRYPTO_TFM_RES_BAD_KEY_LEN);
+               return -EINVAL;
+       }
+
+       return vmac_set_key((u8 *)key, ctx);
+}
+
+static int vmac_init(struct shash_desc *pdesc)
+{
+       struct crypto_shash *parent = pdesc->tfm;
+       struct vmac_ctx_t *ctx = crypto_shash_ctx(parent);
+
+       memset(&ctx->__vmac_ctx, 0, sizeof(struct vmac_ctx));
+       return 0;
+}
+
+static int vmac_update(struct shash_desc *pdesc, const u8 *p,
+               unsigned int len)
+{
+       struct crypto_shash *parent = pdesc->tfm;
+       struct vmac_ctx_t *ctx = crypto_shash_ctx(parent);
+
+       vhash_update(p, len, &ctx->__vmac_ctx);
+
+       return 0;
+}
+
+static int vmac_final(struct shash_desc *pdesc, u8 *out)
+{
+       struct crypto_shash *parent = pdesc->tfm;
+       struct vmac_ctx_t *ctx = crypto_shash_ctx(parent);
+       vmac_t mac;
+       u8 nonce[16] = {};
+
+       mac = vmac(NULL, 0, nonce, NULL, ctx);
+       memcpy(out, &mac, sizeof(vmac_t));
+       memset(&mac, 0, sizeof(vmac_t));
+       memset(&ctx->__vmac_ctx, 0, sizeof(struct vmac_ctx));
+       return 0;
+}
+
+static int vmac_init_tfm(struct crypto_tfm *tfm)
+{
+       struct crypto_cipher *cipher;
+       struct crypto_instance *inst = (void *)tfm->__crt_alg;
+       struct crypto_spawn *spawn = crypto_instance_ctx(inst);
+       struct vmac_ctx_t *ctx = crypto_tfm_ctx(tfm);
+
+       cipher = crypto_spawn_cipher(spawn);
+       if (IS_ERR(cipher))
+               return PTR_ERR(cipher);
+
+       ctx->child = cipher;
+       return 0;
+}
+
+static void vmac_exit_tfm(struct crypto_tfm *tfm)
+{
+       struct vmac_ctx_t *ctx = crypto_tfm_ctx(tfm);
+       crypto_free_cipher(ctx->child);
+}
+
+static int vmac_create(struct crypto_template *tmpl, struct rtattr **tb)
+{
+       struct shash_instance *inst;
+       struct crypto_alg *alg;
+       int err;
+
+       err = crypto_check_attr_type(tb, CRYPTO_ALG_TYPE_SHASH);
+       if (err)
+               return err;
+
+       alg = crypto_get_attr_alg(tb, CRYPTO_ALG_TYPE_CIPHER,
+                       CRYPTO_ALG_TYPE_MASK);
+       if (IS_ERR(alg))
+               return PTR_ERR(alg);
+
+       inst = shash_alloc_instance("vmac", alg);
+       err = PTR_ERR(inst);
+       if (IS_ERR(inst))
+               goto out_put_alg;
+
+       err = crypto_init_spawn(shash_instance_ctx(inst), alg,
+                       shash_crypto_instance(inst),
+                       CRYPTO_ALG_TYPE_MASK);
+       if (err)
+               goto out_free_inst;
+
+       inst->alg.base.cra_priority = alg->cra_priority;
+       inst->alg.base.cra_blocksize = alg->cra_blocksize;
+       inst->alg.base.cra_alignmask = alg->cra_alignmask;
+
+       inst->alg.digestsize = sizeof(vmac_t);
+       inst->alg.base.cra_ctxsize = sizeof(struct vmac_ctx_t);
+       inst->alg.base.cra_init = vmac_init_tfm;
+       inst->alg.base.cra_exit = vmac_exit_tfm;
+
+       inst->alg.init = vmac_init;
+       inst->alg.update = vmac_update;
+       inst->alg.final = vmac_final;
+       inst->alg.setkey = vmac_setkey;
+
+       err = shash_register_instance(tmpl, inst);
+       if (err) {
+out_free_inst:
+               shash_free_instance(shash_crypto_instance(inst));
+       }
+
+out_put_alg:
+       crypto_mod_put(alg);
+       return err;
+}
+
+static struct crypto_template vmac_tmpl = {
+       .name = "vmac",
+       .create = vmac_create,
+       .free = shash_free_instance,
+       .module = THIS_MODULE,
+};
+
+static int __init vmac_module_init(void)
+{
+       return crypto_register_template(&vmac_tmpl);
+}
+
+static void __exit vmac_module_exit(void)
+{
+       crypto_unregister_template(&vmac_tmpl);
+}
+
+module_init(vmac_module_init);
+module_exit(vmac_module_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("VMAC hash algorithm");
+
diff --git a/include/crypto/vmac.h b/include/crypto/vmac.h
new file mode 100644 (file)
index 0000000..c4467c5
--- /dev/null
@@ -0,0 +1,61 @@
+/*
+ * Modified to interface to the Linux kernel
+ * Copyright (c) 2009, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
+ * Place - Suite 330, Boston, MA 02111-1307 USA.
+ */
+
+#ifndef __CRYPTO_VMAC_H
+#define __CRYPTO_VMAC_H
+
+/* --------------------------------------------------------------------------
+ * VMAC and VHASH Implementation by Ted Krovetz (tdk@acm.org) and Wei Dai.
+ * This implementation is herby placed in the public domain.
+ * The authors offers no warranty. Use at your own risk.
+ * Please send bug reports to the authors.
+ * Last modified: 17 APR 08, 1700 PDT
+ * ----------------------------------------------------------------------- */
+
+/*
+ * User definable settings.
+ */
+#define VMAC_TAG_LEN   64
+#define VMAC_KEY_SIZE  128/* Must be 128, 192 or 256                   */
+#define VMAC_KEY_LEN   (VMAC_KEY_SIZE/8)
+#define VMAC_NHBYTES   128/* Must 2^i for any 3 < i < 13 Standard = 128*/
+
+/*
+ * This implementation uses u32 and u64 as names for unsigned 32-
+ * and 64-bit integer types. These are defined in C99 stdint.h. The
+ * following may need adaptation if you are not running a C99 or
+ * Microsoft C environment.
+ */
+struct vmac_ctx {
+       u64 nhkey[(VMAC_NHBYTES/8)+2*(VMAC_TAG_LEN/64-1)];
+       u64 polykey[2*VMAC_TAG_LEN/64];
+       u64 l3key[2*VMAC_TAG_LEN/64];
+       u64 polytmp[2*VMAC_TAG_LEN/64];
+       u64 cached_nonce[2];
+       u64 cached_aes[2];
+       int first_block_processed;
+};
+
+typedef u64 vmac_t;
+
+struct vmac_ctx_t {
+       struct crypto_cipher *child;
+       struct vmac_ctx __vmac_ctx;
+};
+
+#endif /* __CRYPTO_VMAC_H */