Page MenuHome GnuPG
Files F34123826

No OneTemporary
Actions

View Options

This file is larger than 256 KB, so syntax highlighting was skipped.
diff --git a/cipher/sha1.c b/cipher/sha1.c
index 19e75b23..d15c2a29 100644
--- a/cipher/sha1.c
+++ b/cipher/sha1.c
@@ -1,562 +1,600 @@
/* sha1.c - SHA1 hash function
* Copyright (C) 1998, 2001, 2002, 2003, 2008 Free Software Foundation, Inc.
*
* This file is part of Libgcrypt.
*
* Libgcrypt is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of
* the License, or (at your option) any later version.
*
* Libgcrypt is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
/* Test vectors:
*
* "abc"
* A999 3E36 4706 816A BA3E 2571 7850 C26C 9CD0 D89D
*
* "abcdbcdecdefdefgefghfghighijhijkijkljklmklmnlmnomnopnopq"
* 8498 3E44 1C3B D26E BAAE 4AA1 F951 29E5 E546 70F1
*/
#include <config.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#ifdef HAVE_STDINT_H
# include <stdint.h>
#endif
#include "g10lib.h"
#include "bithelp.h"
#include "bufhelp.h"
#include "cipher.h"
#include "sha1.h"
/* USE_SSSE3 indicates whether to compile with Intel SSSE3 code. */
#undef USE_SSSE3
#if defined(__x86_64__) && defined(HAVE_GCC_INLINE_ASM_SSSE3) && \
(defined(HAVE_COMPATIBLE_GCC_AMD64_PLATFORM_AS) || \
defined(HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS))
# define USE_SSSE3 1
#endif
/* USE_AVX indicates whether to compile with Intel AVX code. */
#undef USE_AVX
#if defined(__x86_64__) && defined(HAVE_GCC_INLINE_ASM_AVX) && \
(defined(HAVE_COMPATIBLE_GCC_AMD64_PLATFORM_AS) || \
defined(HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS))
# define USE_AVX 1
#endif
/* USE_BMI2 indicates whether to compile with Intel AVX/BMI2 code. */
#undef USE_BMI2
#if defined(__x86_64__) && defined(HAVE_GCC_INLINE_ASM_AVX) && \
defined(HAVE_GCC_INLINE_ASM_BMI2) && \
(defined(HAVE_COMPATIBLE_GCC_AMD64_PLATFORM_AS) || \
defined(HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS))
# define USE_BMI2 1
#endif
/* USE_NEON indicates whether to enable ARM NEON assembly code. */
#undef USE_NEON
#ifdef ENABLE_NEON_SUPPORT
# if defined(HAVE_ARM_ARCH_V6) && defined(__ARMEL__) \
&& defined(HAVE_COMPATIBLE_GCC_ARM_PLATFORM_AS) \
&& defined(HAVE_GCC_INLINE_ASM_NEON)
# define USE_NEON 1
# endif
#endif /*ENABLE_NEON_SUPPORT*/
/* A macro to test whether P is properly aligned for an u32 type.
Note that config.h provides a suitable replacement for uintptr_t if
it does not exist in stdint.h. */
/* #if __GNUC__ >= 2 */
/* # define U32_ALIGNED_P(p) (!(((uintptr_t)p) % __alignof__ (u32))) */
/* #else */
/* # define U32_ALIGNED_P(p) (!(((uintptr_t)p) % sizeof (u32))) */
/* #endif */
static unsigned int
transform (void *c, const unsigned char *data, size_t nblks);
static void
sha1_init (void *context, unsigned int flags)
{
SHA1_CONTEXT *hd = context;
unsigned int features = _gcry_get_hw_features ();
(void)flags;
hd->h0 = 0x67452301;
hd->h1 = 0xefcdab89;
hd->h2 = 0x98badcfe;
hd->h3 = 0x10325476;
hd->h4 = 0xc3d2e1f0;
hd->bctx.nblocks = 0;
hd->bctx.nblocks_high = 0;
hd->bctx.count = 0;
hd->bctx.blocksize = 64;
hd->bctx.bwrite = transform;
#ifdef USE_SSSE3
hd->use_ssse3 = (features & HWF_INTEL_SSSE3) != 0;
#endif
#ifdef USE_AVX
/* AVX implementation uses SHLD which is known to be slow on non-Intel CPUs.
* Therefore use this implementation on Intel CPUs only. */
hd->use_avx = (features & HWF_INTEL_AVX) && (features & HWF_INTEL_FAST_SHLD);
#endif
#ifdef USE_BMI2
hd->use_bmi2 = (features & HWF_INTEL_AVX) && (features & HWF_INTEL_BMI2);
#endif
#ifdef USE_NEON
hd->use_neon = (features & HWF_ARM_NEON) != 0;
#endif
(void)features;
}
+/*
+ * Initialize the context HD. This is used to prepare the use of
+ * _gcry_sha1_mixblock. WARNING: This is a special purpose function
+ * for exclusive use by random-csprng.c.
+ */
+void
+_gcry_sha1_mixblock_init (SHA1_CONTEXT *hd)
+{
+ sha1_init (hd, 0);
+}
+
/* Round function macros. */
#define K1 0x5A827999L
#define K2 0x6ED9EBA1L
#define K3 0x8F1BBCDCL
#define K4 0xCA62C1D6L
#define F1(x,y,z) ( z ^ ( x & ( y ^ z ) ) )
#define F2(x,y,z) ( x ^ y ^ z )
#define F3(x,y,z) ( ( x & y ) | ( z & ( x | y ) ) )
#define F4(x,y,z) ( x ^ y ^ z )
#define M(i) ( tm = x[ i &0x0f] \
^ x[(i-14)&0x0f] \
^ x[(i-8) &0x0f] \
^ x[(i-3) &0x0f], \
(x[i&0x0f] = rol(tm, 1)))
#define R(a,b,c,d,e,f,k,m) do { e += rol( a, 5 ) \
+ f( b, c, d ) \
+ k \
+ m; \
b = rol( b, 30 ); \
} while(0)
#ifdef USE_NEON
unsigned int
_gcry_sha1_transform_armv7_neon (void *state, const unsigned char *data,
size_t nblks);
#endif
/*
* Transform NBLOCKS of each 64 bytes (16 32-bit words) at DATA.
*/
static unsigned int
transform_blk (void *ctx, const unsigned char *data)
{
SHA1_CONTEXT *hd = ctx;
const u32 *idata = (const void *)data;
register u32 a, b, c, d, e; /* Local copies of the chaining variables. */
register u32 tm; /* Helper. */
u32 x[16]; /* The array we work on. */
#define I(i) (x[i] = buf_get_be32(idata + i))
/* Get the values of the chaining variables. */
a = hd->h0;
b = hd->h1;
c = hd->h2;
d = hd->h3;
e = hd->h4;
/* Transform. */
R( a, b, c, d, e, F1, K1, I( 0) );
R( e, a, b, c, d, F1, K1, I( 1) );
R( d, e, a, b, c, F1, K1, I( 2) );
R( c, d, e, a, b, F1, K1, I( 3) );
R( b, c, d, e, a, F1, K1, I( 4) );
R( a, b, c, d, e, F1, K1, I( 5) );
R( e, a, b, c, d, F1, K1, I( 6) );
R( d, e, a, b, c, F1, K1, I( 7) );
R( c, d, e, a, b, F1, K1, I( 8) );
R( b, c, d, e, a, F1, K1, I( 9) );
R( a, b, c, d, e, F1, K1, I(10) );
R( e, a, b, c, d, F1, K1, I(11) );
R( d, e, a, b, c, F1, K1, I(12) );
R( c, d, e, a, b, F1, K1, I(13) );
R( b, c, d, e, a, F1, K1, I(14) );
R( a, b, c, d, e, F1, K1, I(15) );
R( e, a, b, c, d, F1, K1, M(16) );
R( d, e, a, b, c, F1, K1, M(17) );
R( c, d, e, a, b, F1, K1, M(18) );
R( b, c, d, e, a, F1, K1, M(19) );
R( a, b, c, d, e, F2, K2, M(20) );
R( e, a, b, c, d, F2, K2, M(21) );
R( d, e, a, b, c, F2, K2, M(22) );
R( c, d, e, a, b, F2, K2, M(23) );
R( b, c, d, e, a, F2, K2, M(24) );
R( a, b, c, d, e, F2, K2, M(25) );
R( e, a, b, c, d, F2, K2, M(26) );
R( d, e, a, b, c, F2, K2, M(27) );
R( c, d, e, a, b, F2, K2, M(28) );
R( b, c, d, e, a, F2, K2, M(29) );
R( a, b, c, d, e, F2, K2, M(30) );
R( e, a, b, c, d, F2, K2, M(31) );
R( d, e, a, b, c, F2, K2, M(32) );
R( c, d, e, a, b, F2, K2, M(33) );
R( b, c, d, e, a, F2, K2, M(34) );
R( a, b, c, d, e, F2, K2, M(35) );
R( e, a, b, c, d, F2, K2, M(36) );
R( d, e, a, b, c, F2, K2, M(37) );
R( c, d, e, a, b, F2, K2, M(38) );
R( b, c, d, e, a, F2, K2, M(39) );
R( a, b, c, d, e, F3, K3, M(40) );
R( e, a, b, c, d, F3, K3, M(41) );
R( d, e, a, b, c, F3, K3, M(42) );
R( c, d, e, a, b, F3, K3, M(43) );
R( b, c, d, e, a, F3, K3, M(44) );
R( a, b, c, d, e, F3, K3, M(45) );
R( e, a, b, c, d, F3, K3, M(46) );
R( d, e, a, b, c, F3, K3, M(47) );
R( c, d, e, a, b, F3, K3, M(48) );
R( b, c, d, e, a, F3, K3, M(49) );
R( a, b, c, d, e, F3, K3, M(50) );
R( e, a, b, c, d, F3, K3, M(51) );
R( d, e, a, b, c, F3, K3, M(52) );
R( c, d, e, a, b, F3, K3, M(53) );
R( b, c, d, e, a, F3, K3, M(54) );
R( a, b, c, d, e, F3, K3, M(55) );
R( e, a, b, c, d, F3, K3, M(56) );
R( d, e, a, b, c, F3, K3, M(57) );
R( c, d, e, a, b, F3, K3, M(58) );
R( b, c, d, e, a, F3, K3, M(59) );
R( a, b, c, d, e, F4, K4, M(60) );
R( e, a, b, c, d, F4, K4, M(61) );
R( d, e, a, b, c, F4, K4, M(62) );
R( c, d, e, a, b, F4, K4, M(63) );
R( b, c, d, e, a, F4, K4, M(64) );
R( a, b, c, d, e, F4, K4, M(65) );
R( e, a, b, c, d, F4, K4, M(66) );
R( d, e, a, b, c, F4, K4, M(67) );
R( c, d, e, a, b, F4, K4, M(68) );
R( b, c, d, e, a, F4, K4, M(69) );
R( a, b, c, d, e, F4, K4, M(70) );
R( e, a, b, c, d, F4, K4, M(71) );
R( d, e, a, b, c, F4, K4, M(72) );
R( c, d, e, a, b, F4, K4, M(73) );
R( b, c, d, e, a, F4, K4, M(74) );
R( a, b, c, d, e, F4, K4, M(75) );
R( e, a, b, c, d, F4, K4, M(76) );
R( d, e, a, b, c, F4, K4, M(77) );
R( c, d, e, a, b, F4, K4, M(78) );
R( b, c, d, e, a, F4, K4, M(79) );
/* Update the chaining variables. */
hd->h0 += a;
hd->h1 += b;
hd->h2 += c;
hd->h3 += d;
hd->h4 += e;
return /* burn_stack */ 88+4*sizeof(void*);
}
/* Assembly implementations use SystemV ABI, ABI conversion and additional
* stack to store XMM6-XMM15 needed on Win64. */
#undef ASM_FUNC_ABI
#undef ASM_EXTRA_STACK
#if defined(USE_SSSE3) || defined(USE_AVX) || defined(USE_BMI2)
# ifdef HAVE_COMPATIBLE_GCC_WIN64_PLATFORM_AS
# define ASM_FUNC_ABI __attribute__((sysv_abi))
# define ASM_EXTRA_STACK (10 * 16)
# else
# define ASM_FUNC_ABI
# define ASM_EXTRA_STACK 0
# endif
#endif
#ifdef USE_SSSE3
unsigned int
_gcry_sha1_transform_amd64_ssse3 (void *state, const unsigned char *data,
size_t nblks) ASM_FUNC_ABI;
#endif
#ifdef USE_AVX
unsigned int
_gcry_sha1_transform_amd64_avx (void *state, const unsigned char *data,
size_t nblks) ASM_FUNC_ABI;
#endif
#ifdef USE_BMI2
unsigned int
_gcry_sha1_transform_amd64_avx_bmi2 (void *state, const unsigned char *data,
size_t nblks) ASM_FUNC_ABI;
#endif
static unsigned int
transform (void *ctx, const unsigned char *data, size_t nblks)
{
SHA1_CONTEXT *hd = ctx;
unsigned int burn;
#ifdef USE_BMI2
if (hd->use_bmi2)
return _gcry_sha1_transform_amd64_avx_bmi2 (&hd->h0, data, nblks)
+ 4 * sizeof(void*) + ASM_EXTRA_STACK;
#endif
#ifdef USE_AVX
if (hd->use_avx)
return _gcry_sha1_transform_amd64_avx (&hd->h0, data, nblks)
+ 4 * sizeof(void*) + ASM_EXTRA_STACK;
#endif
#ifdef USE_SSSE3
if (hd->use_ssse3)
return _gcry_sha1_transform_amd64_ssse3 (&hd->h0, data, nblks)
+ 4 * sizeof(void*) + ASM_EXTRA_STACK;
#endif
#ifdef USE_NEON
if (hd->use_neon)
return _gcry_sha1_transform_armv7_neon (&hd->h0, data, nblks)
+ 4 * sizeof(void*);
#endif
do
{
burn = transform_blk (hd, data);
data += 64;
}
while (--nblks);
#ifdef ASM_EXTRA_STACK
/* 'transform_blk' is typically inlined and XMM6-XMM15 are stored at
* the prologue of this function. Therefore need to add ASM_EXTRA_STACK to
* here too.
*/
burn += ASM_EXTRA_STACK;
#endif
return burn;
}
+/*
+ * Apply the SHA-1 transform function on the buffer BLOCKOF64BYTE
+ * which must have a length 64 bytes. BLOCKOF64BYTE must be 32-bit
+ * aligned. Updates the 20 bytes in BLOCKOF64BYTE with its mixed
+ * content. Returns the number of bytes which should be burned on the
+ * stack. You need to use _gcry_sha1_mixblock_init to initialize the
+ * context.
+ * WARNING: This is a special purpose function for exclusive use by
+ * random-csprng.c.
+ */
+unsigned int
+_gcry_sha1_mixblock (SHA1_CONTEXT *hd, void *blockof64byte)
+{
+ u32 *p = blockof64byte;
+ unsigned int nburn;
+
+ nburn = transform (hd, blockof64byte, 1);
+ p[0] = hd->h0;
+ p[1] = hd->h1;
+ p[2] = hd->h2;
+ p[3] = hd->h3;
+ p[4] = hd->h4;
+
+ return nburn;
+}
+
+
/* The routine final terminates the computation and
* returns the digest.
* The handle is prepared for a new cycle, but adding bytes to the
* handle will the destroy the returned buffer.
* Returns: 20 bytes representing the digest.
*/
static void
sha1_final(void *context)
{
SHA1_CONTEXT *hd = context;
u32 t, th, msb, lsb;
unsigned char *p;
unsigned int burn;
_gcry_md_block_write (hd, NULL, 0); /* flush */;
t = hd->bctx.nblocks;
if (sizeof t == sizeof hd->bctx.nblocks)
th = hd->bctx.nblocks_high;
else
th = hd->bctx.nblocks >> 32;
/* multiply by 64 to make a byte count */
lsb = t << 6;
msb = (th << 6) | (t >> 26);
/* add the count */
t = lsb;
if( (lsb += hd->bctx.count) < t )
msb++;
/* multiply by 8 to make a bit count */
t = lsb;
lsb <<= 3;
msb <<= 3;
msb |= t >> 29;
if( hd->bctx.count < 56 ) /* enough room */
{
hd->bctx.buf[hd->bctx.count++] = 0x80; /* pad */
while( hd->bctx.count < 56 )
hd->bctx.buf[hd->bctx.count++] = 0; /* pad */
}
else /* need one extra block */
{
hd->bctx.buf[hd->bctx.count++] = 0x80; /* pad character */
while( hd->bctx.count < 64 )
hd->bctx.buf[hd->bctx.count++] = 0;
_gcry_md_block_write(hd, NULL, 0); /* flush */;
memset(hd->bctx.buf, 0, 56 ); /* fill next block with zeroes */
}
/* append the 64 bit count */
buf_put_be32(hd->bctx.buf + 56, msb);
buf_put_be32(hd->bctx.buf + 60, lsb);
burn = transform( hd, hd->bctx.buf, 1 );
_gcry_burn_stack (burn);
p = hd->bctx.buf;
#define X(a) do { buf_put_be32(p, hd->h##a); p += 4; } while(0)
X(0);
X(1);
X(2);
X(3);
X(4);
#undef X
}
static unsigned char *
sha1_read( void *context )
{
SHA1_CONTEXT *hd = context;
return hd->bctx.buf;
}
/****************
* Shortcut functions which puts the hash value of the supplied buffer
* into outbuf which must have a size of 20 bytes.
*/
void
_gcry_sha1_hash_buffer (void *outbuf, const void *buffer, size_t length)
{
SHA1_CONTEXT hd;
sha1_init (&hd, 0);
_gcry_md_block_write (&hd, buffer, length);
sha1_final (&hd);
memcpy (outbuf, hd.bctx.buf, 20);
}
/* Variant of the above shortcut function using a multiple buffers. */
void
_gcry_sha1_hash_buffers (void *outbuf, const gcry_buffer_t *iov, int iovcnt)
{
SHA1_CONTEXT hd;
sha1_init (&hd, 0);
for (;iovcnt > 0; iov++, iovcnt--)
_gcry_md_block_write (&hd,
(const char*)iov[0].data + iov[0].off, iov[0].len);
sha1_final (&hd);
memcpy (outbuf, hd.bctx.buf, 20);
}
/*
Self-test section.
*/
static gpg_err_code_t
selftests_sha1 (int extended, selftest_report_func_t report)
{
const char *what;
const char *errtxt;
what = "short string";
errtxt = _gcry_hash_selftest_check_one
(GCRY_MD_SHA1, 0,
"abc", 3,
"\xA9\x99\x3E\x36\x47\x06\x81\x6A\xBA\x3E"
"\x25\x71\x78\x50\xC2\x6C\x9C\xD0\xD8\x9D", 20);
if (errtxt)
goto failed;
if (extended)
{
what = "long string";
errtxt = _gcry_hash_selftest_check_one
(GCRY_MD_SHA1, 0,
"abcdbcdecdefdefgefghfghighijhijkijkljklmklmnlmnomnopnopq", 56,
"\x84\x98\x3E\x44\x1C\x3B\xD2\x6E\xBA\xAE"
"\x4A\xA1\xF9\x51\x29\xE5\xE5\x46\x70\xF1", 20);
if (errtxt)
goto failed;
what = "one million \"a\"";
errtxt = _gcry_hash_selftest_check_one
(GCRY_MD_SHA1, 1,
NULL, 0,
"\x34\xAA\x97\x3C\xD4\xC4\xDA\xA4\xF6\x1E"
"\xEB\x2B\xDB\xAD\x27\x31\x65\x34\x01\x6F", 20);
if (errtxt)
goto failed;
}
return 0; /* Succeeded. */
failed:
if (report)
report ("digest", GCRY_MD_SHA1, what, errtxt);
return GPG_ERR_SELFTEST_FAILED;
}
/* Run a full self-test for ALGO and return 0 on success. */
static gpg_err_code_t
run_selftests (int algo, int extended, selftest_report_func_t report)
{
gpg_err_code_t ec;
switch (algo)
{
case GCRY_MD_SHA1:
ec = selftests_sha1 (extended, report);
break;
default:
ec = GPG_ERR_DIGEST_ALGO;
break;
}
return ec;
}
static unsigned char asn[15] = /* Object ID is 1.3.14.3.2.26 */
{ 0x30, 0x21, 0x30, 0x09, 0x06, 0x05, 0x2b, 0x0e, 0x03,
0x02, 0x1a, 0x05, 0x00, 0x04, 0x14 };
static gcry_md_oid_spec_t oid_spec_sha1[] =
{
/* iso.member-body.us.rsadsi.pkcs.pkcs-1.5 (sha1WithRSAEncryption) */
{ "1.2.840.113549.1.1.5" },
/* iso.member-body.us.x9-57.x9cm.3 (dsaWithSha1)*/
{ "1.2.840.10040.4.3" },
/* from NIST's OIW (sha1) */
{ "1.3.14.3.2.26" },
/* from NIST OIW (sha-1WithRSAEncryption) */
{ "1.3.14.3.2.29" },
/* iso.member-body.us.ansi-x9-62.signatures.ecdsa-with-sha1 */
{ "1.2.840.10045.4.1" },
{ NULL },
};
gcry_md_spec_t _gcry_digest_spec_sha1 =
{
GCRY_MD_SHA1, {0, 1},
"SHA1", asn, DIM (asn), oid_spec_sha1, 20,
sha1_init, _gcry_md_block_write, sha1_final, sha1_read, NULL,
sizeof (SHA1_CONTEXT),
run_selftests
};
diff --git a/cipher/sha1.h b/cipher/sha1.h
index c023e15c..6b87631c 100644
--- a/cipher/sha1.h
+++ b/cipher/sha1.h
@@ -1,36 +1,39 @@
/* sha1.h - SHA-1 context definition
* Copyright (C) 1998, 2001, 2002, 2003, 2008 Free Software Foundation, Inc.
*
* This file is part of Libgcrypt.
*
* Libgcrypt is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of
* the License, or (at your option) any later version.
*
* Libgcrypt is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
#ifndef GCRY_SHA1_H
#define GCRY_SHA1_H
#include "hash-common.h"
/* We need this here for direct use by random-csprng.c. */
typedef struct
{
gcry_md_block_ctx_t bctx;
u32 h0,h1,h2,h3,h4;
unsigned int use_ssse3:1;
unsigned int use_avx:1;
unsigned int use_bmi2:1;
unsigned int use_neon:1;
} SHA1_CONTEXT;
+void _gcry_sha1_mixblock_init (SHA1_CONTEXT *hd);
+unsigned int _gcry_sha1_mixblock (SHA1_CONTEXT *hd, void *blockof64byte);
+
#endif /*GCRY_SHA1_H*/
diff --git a/doc/gcrypt.texi b/doc/gcrypt.texi
index 5433ae89..437dddb7 100644
--- a/doc/gcrypt.texi
+++ b/doc/gcrypt.texi
@@ -1,6460 +1,6460 @@
\input texinfo @c -*- Texinfo -*-
@c %**start of header
@setfilename gcrypt.info
@include version.texi
@settitle The Libgcrypt Reference Manual
@c Unify some of the indices.
@syncodeindex tp fn
@syncodeindex pg fn
@c %**end of header
@copying
This manual is for Libgcrypt
(version @value{VERSION}, @value{UPDATED}),
which is GNU's library of cryptographic building blocks.
@noindent
Copyright @copyright{} 2000, 2002, 2003, 2004, 2006, 2007, 2008, 2009, 2011, 2012 Free Software Foundation, Inc. @*
Copyright @copyright{} 2012, 2013 g10 Code GmbH
@quotation
Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2 of the License, or (at your
option) any later version. The text of the license can be found in the
section entitled ``GNU General Public License''.
@end quotation
@end copying
@dircategory GNU Libraries
@direntry
* libgcrypt: (gcrypt). Cryptographic function library.
@end direntry
@c A couple of macros with no effect on texinfo
@c but used by the yat2m processor.
@macro manpage {a}
@end macro
@macro mansect {a}
@end macro
@macro manpause
@end macro
@macro mancont
@end macro
@c
@c Printing stuff taken from gcc.
@c
@macro gnupgtabopt{body}
@code{\body\}
@end macro
@c
@c Titlepage
@c
@setchapternewpage odd
@titlepage
@title The Libgcrypt Reference Manual
@subtitle Version @value{VERSION}
@subtitle @value{UPDATED}
@author Werner Koch (@email{wk@@gnupg.org})
@author Moritz Schulte (@email{mo@@g10code.com})
@page
@vskip 0pt plus 1filll
@insertcopying
@end titlepage
@ifnothtml
@summarycontents
@contents
@page
@end ifnothtml
@ifnottex
@node Top
@top The Libgcrypt Library
@insertcopying
@end ifnottex
@menu
* Introduction:: What is Libgcrypt.
* Preparation:: What you should do before using the library.
* Generalities:: General library functions and data types.
* Handler Functions:: Working with handler functions.
* Symmetric cryptography:: How to use symmetric cryptography.
* Public Key cryptography:: How to use public key cryptography.
* Hashing:: How to use hash algorithms.
* Message Authentication Codes:: How to use MAC algorithms.
* Key Derivation:: How to derive keys from strings
* Random Numbers:: How to work with random numbers.
* S-expressions:: How to manage S-expressions.
* MPI library:: How to work with multi-precision-integers.
* Prime numbers:: How to use the Prime number related functions.
* Utilities:: Utility functions.
* Tools:: Utility tools
* Architecture:: How Libgcrypt works internally.
Appendices
* Self-Tests:: Description of the self-tests.
* FIPS Mode:: Description of the FIPS mode.
* Library Copying:: The GNU Lesser General Public License
says how you can copy and share Libgcrypt.
* Copying:: The GNU General Public License says how you
can copy and share some parts of Libgcrypt.
Indices
* Figures and Tables:: Index of figures and tables.
* Concept Index:: Index of concepts and programs.
* Function and Data Index:: Index of functions, variables and data types.
@end menu
@ifhtml
@page
@summarycontents
@contents
@end ifhtml
@c **********************************************************
@c ******************* Introduction ***********************
@c **********************************************************
@node Introduction
@chapter Introduction
Libgcrypt is a library providing cryptographic building blocks.
@menu
* Getting Started:: How to use this manual.
* Features:: A glance at Libgcrypt's features.
* Overview:: Overview about the library.
@end menu
@node Getting Started
@section Getting Started
This manual documents the Libgcrypt library application programming
interface (API). All functions and data types provided by the library
are explained.
@noindent
The reader is assumed to possess basic knowledge about applied
cryptography.
This manual can be used in several ways. If read from the beginning
to the end, it gives a good introduction into the library and how it
can be used in an application. Forward references are included where
necessary. Later on, the manual can be used as a reference manual to
get just the information needed about any particular interface of the
library. Experienced programmers might want to start looking at the
examples at the end of the manual, and then only read up those parts
of the interface which are unclear.
@node Features
@section Features
Libgcrypt might have a couple of advantages over other libraries doing
a similar job.
@table @asis
@item It's Free Software
Anybody can use, modify, and redistribute it under the terms of the GNU
Lesser General Public License (@pxref{Library Copying}). Note, that
some parts (which are in general not needed by applications) are subject
to the terms of the GNU General Public License (@pxref{Copying}); please
see the README file of the distribution for of list of these parts.
@item It encapsulates the low level cryptography
Libgcrypt provides a high level interface to cryptographic
building blocks using an extensible and flexible API.
@end table
@node Overview
@section Overview
@noindent
The Libgcrypt library is fully thread-safe, where it makes
sense to be thread-safe. Not thread-safe are some cryptographic
functions that modify a certain context stored in handles. If the
user really intents to use such functions from different threads on
the same handle, he has to take care of the serialization of such
functions himself. If not described otherwise, every function is
thread-safe.
Libgcrypt depends on the library `libgpg-error', which contains some
common code used by other GnuPG components.
@c **********************************************************
@c ******************* Preparation ************************
@c **********************************************************
@node Preparation
@chapter Preparation
To use Libgcrypt, you have to perform some changes to your
sources and the build system. The necessary changes are small and
explained in the following sections. At the end of this chapter, it
is described how the library is initialized, and how the requirements
of the library are verified.
@menu
* Header:: What header file you need to include.
* Building sources:: How to build sources using the library.
* Building sources using Automake:: How to build sources with the help of Automake.
* Initializing the library:: How to initialize the library.
* Multi-Threading:: How Libgcrypt can be used in a MT environment.
* Enabling FIPS mode:: How to enable the FIPS mode.
* Hardware features:: How to disable hardware features.
@end menu
@node Header
@section Header
All interfaces (data types and functions) of the library are defined
in the header file @file{gcrypt.h}. You must include this in all source
files using the library, either directly or through some other header
file, like this:
@example
#include <gcrypt.h>
@end example
The name space of Libgcrypt is @code{gcry_*} for function
and type names and @code{GCRY*} for other symbols. In addition the
same name prefixes with one prepended underscore are reserved for
internal use and should never be used by an application. Note that
Libgcrypt uses libgpg-error, which uses @code{gpg_*} as
name space for function and type names and @code{GPG_*} for other
symbols, including all the error codes.
@noindent
Certain parts of gcrypt.h may be excluded by defining these macros:
@table @code
@item GCRYPT_NO_MPI_MACROS
Do not define the shorthand macros @code{mpi_*} for @code{gcry_mpi_*}.
@item GCRYPT_NO_DEPRECATED
Do not include definitions for deprecated features. This is useful to
make sure that no deprecated features are used.
@end table
@node Building sources
@section Building sources
If you want to compile a source file including the `gcrypt.h' header
file, you must make sure that the compiler can find it in the
directory hierarchy. This is accomplished by adding the path to the
directory in which the header file is located to the compilers include
file search path (via the @option{-I} option).
However, the path to the include file is determined at the time the
source is configured. To solve this problem, Libgcrypt ships with a small
helper program @command{libgcrypt-config} that knows the path to the
include file and other configuration options. The options that need
to be added to the compiler invocation at compile time are output by
the @option{--cflags} option to @command{libgcrypt-config}. The following
example shows how it can be used at the command line:
@example
gcc -c foo.c `libgcrypt-config --cflags`
@end example
Adding the output of @samp{libgcrypt-config --cflags} to the
compiler’s command line will ensure that the compiler can find the
Libgcrypt header file.
A similar problem occurs when linking the program with the library.
Again, the compiler has to find the library files. For this to work,
the path to the library files has to be added to the library search path
(via the @option{-L} option). For this, the option @option{--libs} to
@command{libgcrypt-config} can be used. For convenience, this option
also outputs all other options that are required to link the program
with the Libgcrypt libraries (in particular, the @samp{-lgcrypt}
option). The example shows how to link @file{foo.o} with the Libgcrypt
library to a program @command{foo}.
@example
gcc -o foo foo.o `libgcrypt-config --libs`
@end example
Of course you can also combine both examples to a single command by
specifying both options to @command{libgcrypt-config}:
@example
gcc -o foo foo.c `libgcrypt-config --cflags --libs`
@end example
@node Building sources using Automake
@section Building sources using Automake
It is much easier if you use GNU Automake instead of writing your own
Makefiles. If you do that, you do not have to worry about finding and
invoking the @command{libgcrypt-config} script at all.
Libgcrypt provides an extension to Automake that does all
the work for you.
@c A simple macro for optional variables.
@macro ovar{varname}
@r{[}@var{\varname\}@r{]}
@end macro
@defmac AM_PATH_LIBGCRYPT (@ovar{minimum-version}, @ovar{action-if-found}, @ovar{action-if-not-found})
Check whether Libgcrypt (at least version
@var{minimum-version}, if given) exists on the host system. If it is
found, execute @var{action-if-found}, otherwise do
@var{action-if-not-found}, if given.
Additionally, the function defines @code{LIBGCRYPT_CFLAGS} to the
flags needed for compilation of the program to find the
@file{gcrypt.h} header file, and @code{LIBGCRYPT_LIBS} to the linker
flags needed to link the program to the Libgcrypt library. If the
used helper script does not match the target type you are building for
a warning is printed and the string @code{libgcrypt} is appended to the
variable @code{gpg_config_script_warn}.
This macro searches for @command{libgcrypt-config} along the PATH. If
you are cross-compiling, it is useful to set the environment variable
@code{SYSROOT} to the top directory of your target. The macro will
then first look for the helper program in the @file{bin} directory
below that top directory. An absolute directory name must be used for
@code{SYSROOT}. Finally, if the configure command line option
@code{--with-libgcrypt-prefix} is used, only its value is used for the top
directory below which the helper script is expected.
@end defmac
You can use the defined Autoconf variables like this in your
@file{Makefile.am}:
@example
AM_CPPFLAGS = $(LIBGCRYPT_CFLAGS)
LDADD = $(LIBGCRYPT_LIBS)
@end example
@node Initializing the library
@section Initializing the library
Before the library can be used, it must initialize itself. This is
achieved by invoking the function @code{gcry_check_version} described
below.
Also, it is often desirable to check that the version of
Libgcrypt used is indeed one which fits all requirements.
Even with binary compatibility, new features may have been introduced,
but due to problem with the dynamic linker an old version may actually
be used. So you may want to check that the version is okay right
after program startup.
@deftypefun {const char *} gcry_check_version (const char *@var{req_version})
The function @code{gcry_check_version} initializes some subsystems used
by Libgcrypt and must be invoked before any other function in the
library.
@xref{Multi-Threading}.
Furthermore, this function returns the version number of the library.
It can also verify that the version number is higher than a certain
required version number @var{req_version}, if this value is not a null
pointer.
@end deftypefun
Libgcrypt uses a concept known as secure memory, which is a region of
memory set aside for storing sensitive data. Because such memory is a
scarce resource, it needs to be setup in advanced to a fixed size.
Further, most operating systems have special requirements on how that
secure memory can be used. For example, it might be required to install
an application as ``setuid(root)'' to allow allocating such memory.
Libgcrypt requires a sequence of initialization steps to make sure that
this works correctly. The following examples show the necessary steps.
If you don't have a need for secure memory, for example if your
application does not use secret keys or other confidential data or it
runs in a controlled environment where key material floating around in
memory is not a problem, you should initialize Libgcrypt this way:
@example
/* Version check should be the very first call because it
makes sure that important subsystems are initialized. */
if (!gcry_check_version (GCRYPT_VERSION))
@{
fputs ("libgcrypt version mismatch\n", stderr);
exit (2);
@}
/* Disable secure memory. */
gcry_control (GCRYCTL_DISABLE_SECMEM, 0);
/* ... If required, other initialization goes here. */
/* Tell Libgcrypt that initialization has completed. */
gcry_control (GCRYCTL_INITIALIZATION_FINISHED, 0);
@end example
If you have to protect your keys or other information in memory against
being swapped out to disk and to enable an automatic overwrite of used
and freed memory, you need to initialize Libgcrypt this way:
@example
/* Version check should be the very first call because it
makes sure that important subsystems are initialized. */
if (!gcry_check_version (GCRYPT_VERSION))
@{
fputs ("libgcrypt version mismatch\n", stderr);
exit (2);
@}
@anchor{sample-use-suspend-secmem}
/* We don't want to see any warnings, e.g. because we have not yet
parsed program options which might be used to suppress such
warnings. */
gcry_control (GCRYCTL_SUSPEND_SECMEM_WARN);
/* ... If required, other initialization goes here. Note that the
process might still be running with increased privileges and that
the secure memory has not been initialized. */
/* Allocate a pool of 16k secure memory. This make the secure memory
available and also drops privileges where needed. */
gcry_control (GCRYCTL_INIT_SECMEM, 16384, 0);
@anchor{sample-use-resume-secmem}
/* It is now okay to let Libgcrypt complain when there was/is
a problem with the secure memory. */
gcry_control (GCRYCTL_RESUME_SECMEM_WARN);
/* ... If required, other initialization goes here. */
/* Tell Libgcrypt that initialization has completed. */
gcry_control (GCRYCTL_INITIALIZATION_FINISHED, 0);
@end example
It is important that these initialization steps are not done by a
library but by the actual application. A library using Libgcrypt might
want to check for finished initialization using:
@example
if (!gcry_control (GCRYCTL_INITIALIZATION_FINISHED_P))
@{
fputs ("libgcrypt has not been initialized\n", stderr);
abort ();
@}
@end example
Instead of terminating the process, the library may instead print a
warning and try to initialize Libgcrypt itself. See also the section on
multi-threading below for more pitfalls.
@node Multi-Threading
@section Multi-Threading
As mentioned earlier, the Libgcrypt library is
thread-safe if you adhere to the following requirements:
@itemize @bullet
@item
If you use pthread and your applications forks and does not directly
call exec (even calling stdio functions), all kind of problems may
occur. Future versions of Libgcrypt will try to cleanup using
pthread_atfork but even that may lead to problems. This is a common
problem with almost all applications using pthread and fork.
@item
The function @code{gcry_check_version} must be called before any other
function in the library. To
achieve this in multi-threaded programs, you must synchronize the
memory with respect to other threads that also want to use
Libgcrypt. For this, it is sufficient to call
@code{gcry_check_version} before creating the other threads using
Libgcrypt@footnote{At least this is true for POSIX threads,
as @code{pthread_create} is a function that synchronizes memory with
respects to other threads. There are many functions which have this
property, a complete list can be found in POSIX, IEEE Std 1003.1-2003,
Base Definitions, Issue 6, in the definition of the term ``Memory
Synchronization''. For other thread packages, more relaxed or more
strict rules may apply.}.
@item
Just like the function @code{gpg_strerror}, the function
@code{gcry_strerror} is not thread safe. You have to use
@code{gpg_strerror_r} instead.
@end itemize
@node Enabling FIPS mode
@section How to enable the FIPS mode
@cindex FIPS mode
@cindex FIPS 140
Libgcrypt may be used in a FIPS 140-2 mode. Note, that this does not
necessary mean that Libcgrypt is an appoved FIPS 140-2 module. Check the
NIST database at @url{http://csrc.nist.gov/groups/STM/cmvp/} to see what
versions of Libgcrypt are approved.
Because FIPS 140 has certain restrictions on the use of cryptography
which are not always wanted, Libgcrypt needs to be put into FIPS mode
explicitly. Three alternative mechanisms are provided to switch
Libgcrypt into this mode:
@itemize
@item
If the file @file{/proc/sys/crypto/fips_enabled} exists and contains a
numeric value other than @code{0}, Libgcrypt is put into FIPS mode at
initialization time. Obviously this works only on systems with a
@code{proc} file system (i.e. GNU/Linux).
@item
If the file @file{/etc/gcrypt/fips_enabled} exists, Libgcrypt is put
into FIPS mode at initialization time. Note that this filename is
hardwired and does not depend on any configuration options.
@item
If the application requests FIPS mode using the control command
@code{GCRYCTL_FORCE_FIPS_MODE}. This must be done prior to any
initialization (i.e. before @code{gcry_check_version}).
@end itemize
@cindex Enforced FIPS mode
In addition to the standard FIPS mode, Libgcrypt may also be put into
an Enforced FIPS mode by writing a non-zero value into the file
@file{/etc/gcrypt/fips_enabled} or by using the control command
@code{GCRYCTL_SET_ENFORCED_FIPS_FLAG} before any other calls to
libgcrypt. The Enforced FIPS mode helps to detect applications
which don't fulfill all requirements for using
Libgcrypt in FIPS mode (@pxref{FIPS Mode}).
Once Libgcrypt has been put into FIPS mode, it is not possible to
switch back to standard mode without terminating the process first.
If the logging verbosity level of Libgcrypt has been set to at least
2, the state transitions and the self-tests are logged.
@node Hardware features
@section How to disable hardware features
@cindex hardware features
Libgcrypt makes use of certain hardware features. If the use of a
feature is not desired it may be either be disabled by a program or
globally using a configuration file. The currently supported features
are
@table @code
@item padlock-rng
@item padlock-aes
@item padlock-sha
@item padlock-mmul
@item intel-cpu
@item intel-fast-shld
@item intel-bmi2
@item intel-ssse3
@item intel-pclmul
@item intel-aesni
@item intel-rdrand
@item intel-avx
@item intel-avx2
@item arm-neon
@end table
To disable a feature for all processes using Libgcrypt 1.6 or newer,
create the file @file{/etc/gcrypt/hwf.deny} and put each feature not
to be used on a single line. Empty lines, white space, and lines
prefixed with a hash mark are ignored. The file should be world
readable.
To disable a feature specifically for a program that program must tell
it Libgcrypt before before calling @code{gcry_check_version}.
Example:@footnote{NB. Libgcrypt uses the RDRAND feature only as one
source of entropy. A CPU with a broken RDRAND will thus not
compromise of the random number generator}
@example
gcry_control (GCRYCTL_DISABLE_HWF, "intel-rdrand", NULL);
@end example
@noindent
To print the list of active features you may use this command:
@example
mpicalc --print-config | grep ^hwflist: | tr : '\n' | tail -n +2
@end example
@c **********************************************************
@c ******************* General ****************************
@c **********************************************************
@node Generalities
@chapter Generalities
@menu
* Controlling the library:: Controlling Libgcrypt's behavior.
* Error Handling:: Error codes and such.
@end menu
@node Controlling the library
@section Controlling the library
@deftypefun gcry_error_t gcry_control (enum gcry_ctl_cmds @var{cmd}, ...)
This function can be used to influence the general behavior of
Libgcrypt in several ways. Depending on @var{cmd}, more
arguments can or have to be provided.
@table @code
@item GCRYCTL_ENABLE_M_GUARD; Arguments: none
This command enables the built-in memory guard. It must not be used
to activate the memory guard after the memory management has already
been used; therefore it can ONLY be used before
@code{gcry_check_version}. Note that the memory guard is NOT used
when the user of the library has set his own memory management
callbacks.
@item GCRYCTL_ENABLE_QUICK_RANDOM; Arguments: none
This command inhibits the use the very secure random quality level
(@code{GCRY_VERY_STRONG_RANDOM}) and degrades all request down to
@code{GCRY_STRONG_RANDOM}. In general this is not recommended. However,
for some applications the extra quality random Libgcrypt tries to create
is not justified and this option may help to get better performance.
Please check with a crypto expert whether this option can be used for
your application.
This option can only be used at initialization time.
@item GCRYCTL_DUMP_RANDOM_STATS; Arguments: none
This command dumps random number generator related statistics to the
library's logging stream.
@item GCRYCTL_DUMP_MEMORY_STATS; Arguments: none
This command dumps memory management related statistics to the library's
logging stream.
@item GCRYCTL_DUMP_SECMEM_STATS; Arguments: none
This command dumps secure memory management related statistics to the
library's logging stream.
@item GCRYCTL_DROP_PRIVS; Arguments: none
This command disables the use of secure memory and drops the privileges
of the current process. This command has not much use; the suggested way
to disable secure memory is to use @code{GCRYCTL_DISABLE_SECMEM} right
after initialization.
@item GCRYCTL_DISABLE_SECMEM; Arguments: none
This command disables the use of secure memory. If this command is
used in FIPS mode, FIPS mode will be disabled and the function
@code{gcry_fips_mode_active} returns false. However, in Enforced FIPS
mode this command has no effect at all.
Many applications do not require secure memory, so they should disable
it right away. This command should be executed right after
@code{gcry_check_version}.
@item GCRYCTL_DISABLE_LOCKED_SECMEM; Arguments: none
This command disables the use of the mlock call for secure memory.
Disabling the use of mlock may for example be done if an encrypted
swap space is in use. This command should be executed right after
@code{gcry_check_version}.
@item GCRYCTL_DISABLE_PRIV_DROP; Arguments: none
This command sets a global flag to tell the secure memory subsystem
that it shall not drop privileges after secure memory has been
allocated. This command is commonly used right after
@code{gcry_check_version} but may also be used right away at program
startup. It won't have an effect after the secure memory pool has
been initialized. WARNING: A process running setuid(root) is a severe
security risk. Processes making use of Libgcrypt or other complex
code should drop these extra privileges as soon as possible. If this
command has been used the caller is responsible for dropping the
privileges.
@item GCRYCTL_INIT_SECMEM; Arguments: int nbytes
This command is used to allocate a pool of secure memory and thus
enabling the use of secure memory. It also drops all extra privileges
the process has (i.e. if it is run as setuid (root)). If the argument
@var{nbytes} is 0, secure memory will be disabled. The minimum amount
of secure memory allocated is currently 16384 bytes; you may thus use a
value of 1 to request that default size.
@item GCRYCTL_TERM_SECMEM; Arguments: none
This command zeroises the secure memory and destroys the handler. The
secure memory pool may not be used anymore after running this command.
If the secure memory pool as already been destroyed, this command has
no effect. Applications might want to run this command from their
exit handler to make sure that the secure memory gets properly
destroyed. This command is not necessarily thread-safe but that
should not be needed in cleanup code. It may be called from a signal
handler.
@item GCRYCTL_DISABLE_SECMEM_WARN; Arguments: none
Disable warning messages about problems with the secure memory
subsystem. This command should be run right after
@code{gcry_check_version}.
@item GCRYCTL_SUSPEND_SECMEM_WARN; Arguments: none
Postpone warning messages from the secure memory subsystem.
@xref{sample-use-suspend-secmem,,the initialization example}, on how to
use it.
@item GCRYCTL_RESUME_SECMEM_WARN; Arguments: none
Resume warning messages from the secure memory subsystem.
@xref{sample-use-resume-secmem,,the initialization example}, on how to
use it.
@item GCRYCTL_USE_SECURE_RNDPOOL; Arguments: none
This command tells the PRNG to store random numbers in secure memory.
This command should be run right after @code{gcry_check_version} and not
later than the command GCRYCTL_INIT_SECMEM. Note that in FIPS mode the
secure memory is always used.
@item GCRYCTL_SET_RANDOM_SEED_FILE; Arguments: const char *filename
This command specifies the file, which is to be used as seed file for
the PRNG. If the seed file is registered prior to initialization of the
PRNG, the seed file's content (if it exists and seems to be valid) is
fed into the PRNG pool. After the seed file has been registered, the
PRNG can be signalled to write out the PRNG pool's content into the seed
file with the following command.
@item GCRYCTL_UPDATE_RANDOM_SEED_FILE; Arguments: none
Write out the PRNG pool's content into the registered seed file.
Multiple instances of the applications sharing the same random seed file
can be started in parallel, in which case they will read out the same
pool and then race for updating it (the last update overwrites earlier
updates). They will differentiate only by the weak entropy that is
added in read_seed_file based on the PID and clock, and up to 16 bytes
of weak random non-blockingly. The consequence is that the output of
these different instances is correlated to some extent. In a perfect
attack scenario, the attacker can control (or at least guess) the PID
and clock of the application, and drain the system's entropy pool to
reduce the "up to 16 bytes" above to 0. Then the dependencies of the
initial states of the pools are completely known. Note that this is not
an issue if random of @code{GCRY_VERY_STRONG_RANDOM} quality is
requested as in this case enough extra entropy gets mixed. It is also
not an issue when using Linux (rndlinux driver), because this one
guarantees to read full 16 bytes from /dev/urandom and thus there is no
way for an attacker without kernel access to control these 16 bytes.
@item GCRYCTL_CLOSE_RANDOM_DEVICE; Arguments: none
Try to close the random device. If on Unix system you call fork(),
the child process does no call exec(), and you do not intend to use
Libgcrypt in the child, it might be useful to use this control code to
close the inherited file descriptors of the random device. If
Libgcrypt is later used again by the child, the device will be
re-opened. On non-Unix systems this control code is ignored.
@item GCRYCTL_SET_VERBOSITY; Arguments: int level
This command sets the verbosity of the logging. A level of 0 disables
all extra logging whereas positive numbers enable more verbose logging.
The level may be changed at any time but be aware that no memory
synchronization is done so the effect of this command might not
immediately show up in other threads. This command may even be used
prior to @code{gcry_check_version}.
@item GCRYCTL_SET_DEBUG_FLAGS; Arguments: unsigned int flags
Set the debug flag bits as given by the argument. Be aware that that no
memory synchronization is done so the effect of this command might not
immediately show up in other threads. The debug flags are not
considered part of the API and thus may change without notice. As of
now bit 0 enables debugging of cipher functions and bit 1 debugging of
multi-precision-integers. This command may even be used prior to
@code{gcry_check_version}.
@item GCRYCTL_CLEAR_DEBUG_FLAGS; Arguments: unsigned int flags
Set the debug flag bits as given by the argument. Be aware that that no
memory synchronization is done so the effect of this command might not
immediately show up in other threads. This command may even be used
prior to @code{gcry_check_version}.
@item GCRYCTL_DISABLE_INTERNAL_LOCKING; Arguments: none
This command does nothing. It exists only for backward compatibility.
@item GCRYCTL_ANY_INITIALIZATION_P; Arguments: none
This command returns true if the library has been basically initialized.
Such a basic initialization happens implicitly with many commands to get
certain internal subsystems running. The common and suggested way to
do this basic initialization is by calling gcry_check_version.
@item GCRYCTL_INITIALIZATION_FINISHED; Arguments: none
This command tells the library that the application has finished the
initialization.
@item GCRYCTL_INITIALIZATION_FINISHED_P; Arguments: none
This command returns true if the command@*
GCRYCTL_INITIALIZATION_FINISHED has already been run.
@item GCRYCTL_SET_THREAD_CBS; Arguments: struct ath_ops *ath_ops
This command is obsolete since version 1.6.
@item GCRYCTL_FAST_POLL; Arguments: none
Run a fast random poll.
@item GCRYCTL_SET_RNDEGD_SOCKET; Arguments: const char *filename
This command may be used to override the default name of the EGD socket
to connect to. It may be used only during initialization as it is not
thread safe. Changing the socket name again is not supported. The
function may return an error if the given filename is too long for a
local socket name.
EGD is an alternative random gatherer, used only on systems lacking a
proper random device.
@item GCRYCTL_PRINT_CONFIG; Arguments: FILE *stream
This command dumps information pertaining to the configuration of the
library to the given stream. If NULL is given for @var{stream}, the log
system is used. This command may be used before the initialization has
been finished but not before a @code{gcry_check_version}.
@item GCRYCTL_OPERATIONAL_P; Arguments: none
This command returns true if the library is in an operational state.
This information makes only sense in FIPS mode. In contrast to other
functions, this is a pure test function and won't put the library into
FIPS mode or change the internal state. This command may be used before
the initialization has been finished but not before a @code{gcry_check_version}.
@item GCRYCTL_FIPS_MODE_P; Arguments: none
This command returns true if the library is in FIPS mode. Note, that
this is no indication about the current state of the library. This
command may be used before the initialization has been finished but not
before a @code{gcry_check_version}. An application may use this command or
the convenience macro below to check whether FIPS mode is actually
active.
@deftypefun int gcry_fips_mode_active (void)
Returns true if the FIPS mode is active. Note that this is
implemented as a macro.
@end deftypefun
@item GCRYCTL_FORCE_FIPS_MODE; Arguments: none
Running this command puts the library into FIPS mode. If the library is
already in FIPS mode, a self-test is triggered and thus the library will
be put into operational state. This command may be used before a call
to @code{gcry_check_version} and that is actually the recommended way to let an
application switch the library into FIPS mode. Note that Libgcrypt will
reject an attempt to switch to fips mode during or after the initialization.
@item GCRYCTL_SET_ENFORCED_FIPS_FLAG; Arguments: none
Running this command sets the internal flag that puts the library into
the enforced FIPS mode during the FIPS mode initialization. This command
does not affect the library if the library is not put into the FIPS mode and
it must be used before any other libgcrypt library calls that initialize
the library such as @code{gcry_check_version}. Note that Libgcrypt will
reject an attempt to switch to the enforced fips mode during or after
the initialization.
@item GCRYCTL_SET_PREFERRED_RNG_TYPE; Arguments: int
These are advisory commands to select a certain random number
generator. They are only advisory because libraries may not know what
an application actually wants or vice versa. Thus Libgcrypt employs a
priority check to select the actually used RNG. If an applications
selects a lower priority RNG but a library requests a higher priority
RNG Libgcrypt will switch to the higher priority RNG. Applications
and libraries should use these control codes before
@code{gcry_check_version}. The available generators are:
@table @code
@item GCRY_RNG_TYPE_STANDARD
A conservative standard generator based on the ``Continuously Seeded
Pseudo Random Number Generator'' designed by Peter Gutmann.
@item GCRY_RNG_TYPE_FIPS
A deterministic random number generator conforming to he document
``NIST-Recommended Random Number Generator Based on ANSI X9.31
Appendix A.2.4 Using the 3-Key Triple DES and AES Algorithms''
(2005-01-31). This implementation uses the AES variant.
@item GCRY_RNG_TYPE_SYSTEM
A wrapper around the system's native RNG. On Unix system these are
usually the /dev/random and /dev/urandom devices.
@end table
The default is @code{GCRY_RNG_TYPE_STANDARD} unless FIPS mode as been
enabled; in which case @code{GCRY_RNG_TYPE_FIPS} is used and locked
against further changes.
@item GCRYCTL_GET_CURRENT_RNG_TYPE; Arguments: int *
This command stores the type of the currently used RNG as an integer
value at the provided address.
@item GCRYCTL_SELFTEST; Arguments: none
This may be used at anytime to have the library run all implemented
self-tests. It works in standard and in FIPS mode. Returns 0 on
success or an error code on failure.
@item GCRYCTL_DISABLE_HWF; Arguments: const char *name
Libgcrypt detects certain features of the CPU at startup time. For
performance tests it is sometimes required not to use such a feature.
This option may be used to disable a certain feature; i.e. Libgcrypt
behaves as if this feature has not been detected. Note that the
detection code might be run if the feature has been disabled. This
command must be used at initialization time; i.e. before calling
@code{gcry_check_version}.
@end table
@end deftypefun
@c **********************************************************
@c ******************* Errors ****************************
@c **********************************************************
@node Error Handling
@section Error Handling
Many functions in Libgcrypt can return an error if they
fail. For this reason, the application should always catch the error
condition and take appropriate measures, for example by releasing the
resources and passing the error up to the caller, or by displaying a
descriptive message to the user and cancelling the operation.
Some error values do not indicate a system error or an error in the
operation, but the result of an operation that failed properly. For
example, if you try to decrypt a tempered message, the decryption will
fail. Another error value actually means that the end of a data
buffer or list has been reached. The following descriptions explain
for many error codes what they mean usually. Some error values have
specific meanings if returned by a certain functions. Such cases are
described in the documentation of those functions.
Libgcrypt uses the @code{libgpg-error} library. This allows to share
the error codes with other components of the GnuPG system, and to pass
error values transparently from the crypto engine, or some helper
application of the crypto engine, to the user. This way no
information is lost. As a consequence, Libgcrypt does not use its own
identifiers for error codes, but uses those provided by
@code{libgpg-error}. They usually start with @code{GPG_ERR_}.
However, Libgcrypt does provide aliases for the functions
defined in libgpg-error, which might be preferred for name space
consistency.
Most functions in Libgcrypt return an error code in the case
of failure. For this reason, the application should always catch the
error condition and take appropriate measures, for example by
releasing the resources and passing the error up to the caller, or by
displaying a descriptive message to the user and canceling the
operation.
Some error values do not indicate a system error or an error in the
operation, but the result of an operation that failed properly.
GnuPG components, including Libgcrypt, use an extra library named
libgpg-error to provide a common error handling scheme. For more
information on libgpg-error, see the according manual.
@menu
* Error Values:: The error value and what it means.
* Error Sources:: A list of important error sources.
* Error Codes:: A list of important error codes.
* Error Strings:: How to get a descriptive string from a value.
@end menu
@node Error Values
@subsection Error Values
@cindex error values
@cindex error codes
@cindex error sources
@deftp {Data type} {gcry_err_code_t}
The @code{gcry_err_code_t} type is an alias for the
@code{libgpg-error} type @code{gpg_err_code_t}. The error code
indicates the type of an error, or the reason why an operation failed.
A list of important error codes can be found in the next section.
@end deftp
@deftp {Data type} {gcry_err_source_t}
The @code{gcry_err_source_t} type is an alias for the
@code{libgpg-error} type @code{gpg_err_source_t}. The error source
has not a precisely defined meaning. Sometimes it is the place where
the error happened, sometimes it is the place where an error was
encoded into an error value. Usually the error source will give an
indication to where to look for the problem. This is not always true,
but it is attempted to achieve this goal.
A list of important error sources can be found in the next section.
@end deftp
@deftp {Data type} {gcry_error_t}
The @code{gcry_error_t} type is an alias for the @code{libgpg-error}
type @code{gpg_error_t}. An error value like this has always two
components, an error code and an error source. Both together form the
error value.
Thus, the error value can not be directly compared against an error
code, but the accessor functions described below must be used.
However, it is guaranteed that only 0 is used to indicate success
(@code{GPG_ERR_NO_ERROR}), and that in this case all other parts of
the error value are set to 0, too.
Note that in Libgcrypt, the error source is used purely for
diagnostic purposes. Only the error code should be checked to test
for a certain outcome of a function. The manual only documents the
error code part of an error value. The error source is left
unspecified and might be anything.
@end deftp
@deftypefun {gcry_err_code_t} gcry_err_code (@w{gcry_error_t @var{err}})
The static inline function @code{gcry_err_code} returns the
@code{gcry_err_code_t} component of the error value @var{err}. This
function must be used to extract the error code from an error value in
order to compare it with the @code{GPG_ERR_*} error code macros.
@end deftypefun
@deftypefun {gcry_err_source_t} gcry_err_source (@w{gcry_error_t @var{err}})
The static inline function @code{gcry_err_source} returns the
@code{gcry_err_source_t} component of the error value @var{err}. This
function must be used to extract the error source from an error value in
order to compare it with the @code{GPG_ERR_SOURCE_*} error source macros.
@end deftypefun
@deftypefun {gcry_error_t} gcry_err_make (@w{gcry_err_source_t @var{source}}, @w{gcry_err_code_t @var{code}})
The static inline function @code{gcry_err_make} returns the error
value consisting of the error source @var{source} and the error code
@var{code}.
This function can be used in callback functions to construct an error
value to return it to the library.
@end deftypefun
@deftypefun {gcry_error_t} gcry_error (@w{gcry_err_code_t @var{code}})
The static inline function @code{gcry_error} returns the error value
consisting of the default error source and the error code @var{code}.
For @acronym{GCRY} applications, the default error source is
@code{GPG_ERR_SOURCE_USER_1}. You can define
@code{GCRY_ERR_SOURCE_DEFAULT} before including @file{gcrypt.h} to
change this default.
This function can be used in callback functions to construct an error
value to return it to the library.
@end deftypefun
The @code{libgpg-error} library provides error codes for all system
error numbers it knows about. If @var{err} is an unknown error
number, the error code @code{GPG_ERR_UNKNOWN_ERRNO} is used. The
following functions can be used to construct error values from system
errno numbers.
@deftypefun {gcry_error_t} gcry_err_make_from_errno (@w{gcry_err_source_t @var{source}}, @w{int @var{err}})
The function @code{gcry_err_make_from_errno} is like
@code{gcry_err_make}, but it takes a system error like @code{errno}
instead of a @code{gcry_err_code_t} error code.
@end deftypefun
@deftypefun {gcry_error_t} gcry_error_from_errno (@w{int @var{err}})
The function @code{gcry_error_from_errno} is like @code{gcry_error},
but it takes a system error like @code{errno} instead of a
@code{gcry_err_code_t} error code.
@end deftypefun
Sometimes you might want to map system error numbers to error codes
directly, or map an error code representing a system error back to the
system error number. The following functions can be used to do that.
@deftypefun {gcry_err_code_t} gcry_err_code_from_errno (@w{int @var{err}})
The function @code{gcry_err_code_from_errno} returns the error code
for the system error @var{err}. If @var{err} is not a known system
error, the function returns @code{GPG_ERR_UNKNOWN_ERRNO}.
@end deftypefun
@deftypefun {int} gcry_err_code_to_errno (@w{gcry_err_code_t @var{err}})
The function @code{gcry_err_code_to_errno} returns the system error
for the error code @var{err}. If @var{err} is not an error code
representing a system error, or if this system error is not defined on
this system, the function returns @code{0}.
@end deftypefun
@node Error Sources
@subsection Error Sources
@cindex error codes, list of
The library @code{libgpg-error} defines an error source for every
component of the GnuPG system. The error source part of an error
value is not well defined. As such it is mainly useful to improve the
diagnostic error message for the user.
If the error code part of an error value is @code{0}, the whole error
value will be @code{0}. In this case the error source part is of
course @code{GPG_ERR_SOURCE_UNKNOWN}.
The list of error sources that might occur in applications using
@acronym{Libgcrypt} is:
@table @code
@item GPG_ERR_SOURCE_UNKNOWN
The error source is not known. The value of this error source is
@code{0}.
@item GPG_ERR_SOURCE_GPGME
The error source is @acronym{GPGME} itself.
@item GPG_ERR_SOURCE_GPG
The error source is GnuPG, which is the crypto engine used for the
OpenPGP protocol.
@item GPG_ERR_SOURCE_GPGSM
The error source is GPGSM, which is the crypto engine used for the
OpenPGP protocol.
@item GPG_ERR_SOURCE_GCRYPT
The error source is @code{libgcrypt}, which is used by crypto engines
to perform cryptographic operations.
@item GPG_ERR_SOURCE_GPGAGENT
The error source is @command{gpg-agent}, which is used by crypto
engines to perform operations with the secret key.
@item GPG_ERR_SOURCE_PINENTRY
The error source is @command{pinentry}, which is used by
@command{gpg-agent} to query the passphrase to unlock a secret key.
@item GPG_ERR_SOURCE_SCD
The error source is the SmartCard Daemon, which is used by
@command{gpg-agent} to delegate operations with the secret key to a
SmartCard.
@item GPG_ERR_SOURCE_KEYBOX
The error source is @code{libkbx}, a library used by the crypto
engines to manage local keyrings.
@item GPG_ERR_SOURCE_USER_1
@item GPG_ERR_SOURCE_USER_2
@item GPG_ERR_SOURCE_USER_3
@item GPG_ERR_SOURCE_USER_4
These error sources are not used by any GnuPG component and can be
used by other software. For example, applications using
Libgcrypt can use them to mark error values coming from callback
handlers. Thus @code{GPG_ERR_SOURCE_USER_1} is the default for errors
created with @code{gcry_error} and @code{gcry_error_from_errno},
unless you define @code{GCRY_ERR_SOURCE_DEFAULT} before including
@file{gcrypt.h}.
@end table
@node Error Codes
@subsection Error Codes
@cindex error codes, list of
The library @code{libgpg-error} defines many error values. The
following list includes the most important error codes.
@table @code
@item GPG_ERR_EOF
This value indicates the end of a list, buffer or file.
@item GPG_ERR_NO_ERROR
This value indicates success. The value of this error code is
@code{0}. Also, it is guaranteed that an error value made from the
error code @code{0} will be @code{0} itself (as a whole). This means
that the error source information is lost for this error code,
however, as this error code indicates that no error occurred, this is
generally not a problem.
@item GPG_ERR_GENERAL
This value means that something went wrong, but either there is not
enough information about the problem to return a more useful error
value, or there is no separate error value for this type of problem.
@item GPG_ERR_ENOMEM
This value means that an out-of-memory condition occurred.
@item GPG_ERR_E...
System errors are mapped to GPG_ERR_EFOO where FOO is the symbol for
the system error.
@item GPG_ERR_INV_VALUE
This value means that some user provided data was out of range.
@item GPG_ERR_UNUSABLE_PUBKEY
This value means that some recipients for a message were invalid.
@item GPG_ERR_UNUSABLE_SECKEY
This value means that some signers were invalid.
@item GPG_ERR_NO_DATA
This value means that data was expected where no data was found.
@item GPG_ERR_CONFLICT
This value means that a conflict of some sort occurred.
@item GPG_ERR_NOT_IMPLEMENTED
This value indicates that the specific function (or operation) is not
implemented. This error should never happen. It can only occur if
you use certain values or configuration options which do not work,
but for which we think that they should work at some later time.
@item GPG_ERR_DECRYPT_FAILED
This value indicates that a decryption operation was unsuccessful.
@item GPG_ERR_WRONG_KEY_USAGE
This value indicates that a key is not used appropriately.
@item GPG_ERR_NO_SECKEY
This value indicates that no secret key for the user ID is available.
@item GPG_ERR_UNSUPPORTED_ALGORITHM
This value means a verification failed because the cryptographic
algorithm is not supported by the crypto backend.
@item GPG_ERR_BAD_SIGNATURE
This value means a verification failed because the signature is bad.
@item GPG_ERR_NO_PUBKEY
This value means a verification failed because the public key is not
available.
@item GPG_ERR_NOT_OPERATIONAL
This value means that the library is not yet in state which allows to
use this function. This error code is in particular returned if
Libgcrypt is operated in FIPS mode and the internal state of the
library does not yet or not anymore allow the use of a service.
This error code is only available with newer libgpg-error versions, thus
you might see ``invalid error code'' when passing this to
@code{gpg_strerror}. The numeric value of this error code is 176.
@item GPG_ERR_USER_1
@item GPG_ERR_USER_2
@item ...
@item GPG_ERR_USER_16
These error codes are not used by any GnuPG component and can be
freely used by other software. Applications using Libgcrypt
might use them to mark specific errors returned by callback handlers
if no suitable error codes (including the system errors) for these
errors exist already.
@end table
@node Error Strings
@subsection Error Strings
@cindex error values, printing of
@cindex error codes, printing of
@cindex error sources, printing of
@cindex error strings
@deftypefun {const char *} gcry_strerror (@w{gcry_error_t @var{err}})
The function @code{gcry_strerror} returns a pointer to a statically
allocated string containing a description of the error code contained
in the error value @var{err}. This string can be used to output a
diagnostic message to the user.
@end deftypefun
@deftypefun {const char *} gcry_strsource (@w{gcry_error_t @var{err}})
The function @code{gcry_strsource} returns a pointer to a statically
allocated string containing a description of the error source
contained in the error value @var{err}. This string can be used to
output a diagnostic message to the user.
@end deftypefun
The following example illustrates the use of the functions described
above:
@example
@{
gcry_cipher_hd_t handle;
gcry_error_t err = 0;
err = gcry_cipher_open (&handle, GCRY_CIPHER_AES,
GCRY_CIPHER_MODE_CBC, 0);
if (err)
@{
fprintf (stderr, "Failure: %s/%s\n",
gcry_strsource (err),
gcry_strerror (err));
@}
@}
@end example
@c **********************************************************
@c ******************* General ****************************
@c **********************************************************
@node Handler Functions
@chapter Handler Functions
Libgcrypt makes it possible to install so called `handler functions',
which get called by Libgcrypt in case of certain events.
@menu
* Progress handler:: Using a progress handler function.
* Allocation handler:: Using special memory allocation functions.
* Error handler:: Using error handler functions.
* Logging handler:: Using a special logging function.
@end menu
@node Progress handler
@section Progress handler
It is often useful to retrieve some feedback while long running
operations are performed.
@deftp {Data type} gcry_handler_progress_t
Progress handler functions have to be of the type
@code{gcry_handler_progress_t}, which is defined as:
@code{void (*gcry_handler_progress_t) (void *, const char *, int, int, int)}
@end deftp
The following function may be used to register a handler function for
this purpose.
@deftypefun void gcry_set_progress_handler (gcry_handler_progress_t @var{cb}, void *@var{cb_data})
This function installs @var{cb} as the `Progress handler' function.
It may be used only during initialization. @var{cb} must be defined
as follows:
@example
void
my_progress_handler (void *@var{cb_data}, const char *@var{what},
int @var{printchar}, int @var{current}, int @var{total})
@{
/* Do something. */
@}
@end example
A description of the arguments of the progress handler function follows.
@table @var
@item cb_data
The argument provided in the call to @code{gcry_set_progress_handler}.
@item what
A string identifying the type of the progress output. The following
values for @var{what} are defined:
@table @code
@item need_entropy
Not enough entropy is available. @var{total} holds the number of
required bytes.
@item wait_dev_random
Waiting to re-open a random device. @var{total} gives the number of
seconds until the next try.
@item primegen
Values for @var{printchar}:
@table @code
@item \n
Prime generated.
@item !
Need to refresh the pool of prime numbers.
@item <, >
Number of bits adjusted.
@item ^
Searching for a generator.
@item .
Fermat test on 10 candidates failed.
@item :
Restart with a new random value.
@item +
Rabin Miller test passed.
@end table
@end table
@end table
@end deftypefun
@node Allocation handler
@section Allocation handler
It is possible to make Libgcrypt use special memory
allocation functions instead of the built-in ones.
Memory allocation functions are of the following types:
@deftp {Data type} gcry_handler_alloc_t
This type is defined as: @code{void *(*gcry_handler_alloc_t) (size_t n)}.
@end deftp
@deftp {Data type} gcry_handler_secure_check_t
This type is defined as: @code{int *(*gcry_handler_secure_check_t) (const void *)}.
@end deftp
@deftp {Data type} gcry_handler_realloc_t
This type is defined as: @code{void *(*gcry_handler_realloc_t) (void *p, size_t n)}.
@end deftp
@deftp {Data type} gcry_handler_free_t
This type is defined as: @code{void *(*gcry_handler_free_t) (void *)}.
@end deftp
Special memory allocation functions can be installed with the
following function:
@deftypefun void gcry_set_allocation_handler (gcry_handler_alloc_t @var{func_alloc}, gcry_handler_alloc_t @var{func_alloc_secure}, gcry_handler_secure_check_t @var{func_secure_check}, gcry_handler_realloc_t @var{func_realloc}, gcry_handler_free_t @var{func_free})
Install the provided functions and use them instead of the built-in
functions for doing memory allocation. Using this function is in
general not recommended because the standard Libgcrypt allocation
functions are guaranteed to zeroize memory if needed.
This function may be used only during initialization and may not be
used in fips mode.
@end deftypefun
@node Error handler
@section Error handler
The following functions may be used to register handler functions that
are called by Libgcrypt in case certain error conditions occur. They
may and should be registered prior to calling @code{gcry_check_version}.
@deftp {Data type} gcry_handler_no_mem_t
This type is defined as: @code{int (*gcry_handler_no_mem_t) (void *, size_t, unsigned int)}
@end deftp
@deftypefun void gcry_set_outofcore_handler (gcry_handler_no_mem_t @var{func_no_mem}, void *@var{cb_data})
This function registers @var{func_no_mem} as `out-of-core handler',
which means that it will be called in the case of not having enough
memory available. The handler is called with 3 arguments: The first
one is the pointer @var{cb_data} as set with this function, the second
is the requested memory size and the last being a flag. If bit 0 of
the flag is set, secure memory has been requested. The handler should
either return true to indicate that Libgcrypt should try again
allocating memory or return false to let Libgcrypt use its default
fatal error handler.
@end deftypefun
@deftp {Data type} gcry_handler_error_t
This type is defined as: @code{void (*gcry_handler_error_t) (void *, int, const char *)}
@end deftp
@deftypefun void gcry_set_fatalerror_handler (gcry_handler_error_t @var{func_error}, void *@var{cb_data})
This function registers @var{func_error} as `error handler',
which means that it will be called in error conditions.
@end deftypefun
@node Logging handler
@section Logging handler
@deftp {Data type} gcry_handler_log_t
This type is defined as: @code{void (*gcry_handler_log_t) (void *, int, const char *, va_list)}
@end deftp
@deftypefun void gcry_set_log_handler (gcry_handler_log_t @var{func_log}, void *@var{cb_data})
This function registers @var{func_log} as `logging handler', which means
that it will be called in case Libgcrypt wants to log a message. This
function may and should be used prior to calling
@code{gcry_check_version}.
@end deftypefun
@c **********************************************************
@c ******************* Ciphers ****************************
@c **********************************************************
@c @include cipher-ref.texi
@node Symmetric cryptography
@chapter Symmetric cryptography
The cipher functions are used for symmetrical cryptography,
i.e. cryptography using a shared key. The programming model follows
an open/process/close paradigm and is in that similar to other
building blocks provided by Libgcrypt.
@menu
* Available ciphers:: List of ciphers supported by the library.
* Available cipher modes:: List of cipher modes supported by the library.
* Working with cipher handles:: How to perform operations related to cipher handles.
* General cipher functions:: General cipher functions independent of cipher handles.
@end menu
@node Available ciphers
@section Available ciphers
@table @code
@item GCRY_CIPHER_NONE
This is not a real algorithm but used by some functions as error return.
The value always evaluates to false.
@item GCRY_CIPHER_IDEA
@cindex IDEA
This is the IDEA algorithm.
@item GCRY_CIPHER_3DES
@cindex 3DES
@cindex Triple-DES
@cindex DES-EDE
@cindex Digital Encryption Standard
Triple-DES with 3 Keys as EDE. The key size of this algorithm is 168 but
you have to pass 192 bits because the most significant bits of each byte
are ignored.
@item GCRY_CIPHER_CAST5
@cindex CAST5
CAST128-5 block cipher algorithm. The key size is 128 bits.
@item GCRY_CIPHER_BLOWFISH
@cindex Blowfish
The blowfish algorithm. The current implementation allows only for a key
size of 128 bits.
@item GCRY_CIPHER_SAFER_SK128
Reserved and not currently implemented.
@item GCRY_CIPHER_DES_SK
Reserved and not currently implemented.
@item GCRY_CIPHER_AES
@itemx GCRY_CIPHER_AES128
@itemx GCRY_CIPHER_RIJNDAEL
@itemx GCRY_CIPHER_RIJNDAEL128
@cindex Rijndael
@cindex AES
@cindex Advanced Encryption Standard
AES (Rijndael) with a 128 bit key.
@item GCRY_CIPHER_AES192
@itemx GCRY_CIPHER_RIJNDAEL192
AES (Rijndael) with a 192 bit key.
@item GCRY_CIPHER_AES256
@itemx GCRY_CIPHER_RIJNDAEL256
AES (Rijndael) with a 256 bit key.
@item GCRY_CIPHER_TWOFISH
@cindex Twofish
The Twofish algorithm with a 256 bit key.
@item GCRY_CIPHER_TWOFISH128
The Twofish algorithm with a 128 bit key.
@item GCRY_CIPHER_ARCFOUR
@cindex Arcfour
@cindex RC4
An algorithm which is 100% compatible with RSA Inc.'s RC4 algorithm.
Note that this is a stream cipher and must be used very carefully to
avoid a couple of weaknesses.
@item GCRY_CIPHER_DES
@cindex DES
Standard DES with a 56 bit key. You need to pass 64 bit but the high
bits of each byte are ignored. Note, that this is a weak algorithm
which can be broken in reasonable time using a brute force approach.
@item GCRY_CIPHER_SERPENT128
@itemx GCRY_CIPHER_SERPENT192
@itemx GCRY_CIPHER_SERPENT256
@cindex Serpent
The Serpent cipher from the AES contest.
@item GCRY_CIPHER_RFC2268_40
@itemx GCRY_CIPHER_RFC2268_128
@cindex rfc-2268
@cindex RC2
Ron's Cipher 2 in the 40 and 128 bit variants.
@item GCRY_CIPHER_SEED
@cindex Seed (cipher)
A 128 bit cipher as described by RFC4269.
@item GCRY_CIPHER_CAMELLIA128
@itemx GCRY_CIPHER_CAMELLIA192
@itemx GCRY_CIPHER_CAMELLIA256
@cindex Camellia
The Camellia cipher by NTT. See
@uref{http://info.isl.ntt.co.jp/@/crypt/@/eng/@/camellia/@/specifications.html}.
@item GCRY_CIPHER_SALSA20
@cindex Salsa20
This is the Salsa20 stream cipher.
@item GCRY_CIPHER_SALSA20R12
@cindex Salsa20/12
This is the Salsa20/12 - reduced round version of Salsa20 stream cipher.
@item GCRY_CIPHER_GOST28147
@cindex GOST 28147-89
The GOST 28147-89 cipher, defined in the respective GOST standard.
Translation of this GOST into English is provided in the RFC-5830.
@item GCRY_CIPHER_CHACHA20
@cindex ChaCha20
This is the ChaCha20 stream cipher.
@end table
@node Available cipher modes
@section Available cipher modes
@table @code
@item GCRY_CIPHER_MODE_NONE
No mode specified. This should not be used. The only exception is that
if Libgcrypt is not used in FIPS mode and if any debug flag has been
set, this mode may be used to bypass the actual encryption.
@item GCRY_CIPHER_MODE_ECB
@cindex ECB, Electronic Codebook mode
Electronic Codebook mode.
@item GCRY_CIPHER_MODE_CFB
@cindex CFB, Cipher Feedback mode
Cipher Feedback mode. The shift size equals the block size of the
cipher (e.g. for AES it is CFB-128).
@item GCRY_CIPHER_MODE_CBC
@cindex CBC, Cipher Block Chaining mode
Cipher Block Chaining mode.
@item GCRY_CIPHER_MODE_STREAM
Stream mode, only to be used with stream cipher algorithms.
@item GCRY_CIPHER_MODE_OFB
@cindex OFB, Output Feedback mode
Output Feedback mode.
@item GCRY_CIPHER_MODE_CTR
@cindex CTR, Counter mode
Counter mode.
@item GCRY_CIPHER_MODE_AESWRAP
@cindex AES-Wrap mode
This mode is used to implement the AES-Wrap algorithm according to
RFC-3394. It may be used with any 128 bit block length algorithm,
however the specs require one of the 3 AES algorithms. These special
conditions apply: If @code{gcry_cipher_setiv} has not been used the
standard IV is used; if it has been used the lower 64 bit of the IV
are used as the Alternative Initial Value. On encryption the provided
output buffer must be 64 bit (8 byte) larger than the input buffer;
in-place encryption is still allowed. On decryption the output buffer
may be specified 64 bit (8 byte) shorter than then input buffer. As
per specs the input length must be at least 128 bits and the length
must be a multiple of 64 bits.
@item GCRY_CIPHER_MODE_CCM
@cindex CCM, Counter with CBC-MAC mode
Counter with CBC-MAC mode is an Authenticated Encryption with
Associated Data (AEAD) block cipher mode, which is specified in
'NIST Special Publication 800-38C' and RFC 3610.
@item GCRY_CIPHER_MODE_GCM
@cindex GCM, Galois/Counter Mode
Galois/Counter Mode (GCM) is an Authenticated Encryption with
Associated Data (AEAD) block cipher mode, which is specified in
'NIST Special Publication 800-38D'.
@item GCRY_CIPHER_MODE_POLY1305
@cindex Poly1305 based AEAD mode with ChaCha20
This mode implements the Poly1305 Authenticated Encryption with Associated
Data (AEAD) mode according to RFC-7539. This mode can be used with ChaCha20
stream cipher.
@item GCRY_CIPHER_MODE_OCB
@cindex OCB, OCB3
OCB is an Authenticated Encryption with Associated Data (AEAD) block
cipher mode, which is specified in RFC-7253. Supported tag lengths
are 128, 96, and 64 bit with the default being 128 bit. To switch to
a different tag length @code{gcry_cipher_ctl} using the command
@code{GCRYCTL_SET_TAGLEN} and the address of an @code{int} variable
set to 12 (for 96 bit) or 8 (for 64 bit) provided for the
@code{buffer} argument and @code{sizeof(int)} for @code{buflen}.
Note that the use of @code{gcry_cipher_final} is required.
@end table
@node Working with cipher handles
@section Working with cipher handles
To use a cipher algorithm, you must first allocate an according
handle. This is to be done using the open function:
@deftypefun gcry_error_t gcry_cipher_open (gcry_cipher_hd_t *@var{hd}, int @var{algo}, int @var{mode}, unsigned int @var{flags})
This function creates the context handle required for most of the
other cipher functions and returns a handle to it in `hd'. In case of
an error, an according error code is returned.
The ID of algorithm to use must be specified via @var{algo}. See
@xref{Available ciphers}, for a list of supported ciphers and the
according constants.
Besides using the constants directly, the function
@code{gcry_cipher_map_name} may be used to convert the textual name of
an algorithm into the according numeric ID.
The cipher mode to use must be specified via @var{mode}. See
@xref{Available cipher modes}, for a list of supported cipher modes
and the according constants. Note that some modes are incompatible
with some algorithms - in particular, stream mode
(@code{GCRY_CIPHER_MODE_STREAM}) only works with stream ciphers.
Poly1305 AEAD mode (@code{GCRY_CIPHER_MODE_POLY1305}) only works with
ChaCha20 stream cipher. The block cipher modes
(@code{GCRY_CIPHER_MODE_ECB}, @code{GCRY_CIPHER_MODE_CBC},
@code{GCRY_CIPHER_MODE_CFB}, @code{GCRY_CIPHER_MODE_OFB} and
@code{GCRY_CIPHER_MODE_CTR}) will work with any block cipher
algorithm. GCM mode (@code{GCRY_CIPHER_MODE_CCM}), CCM mode
(@code{GCRY_CIPHER_MODE_GCM}), and OCB mode
(@code{GCRY_CIPHER_MODE_OCB}) will only work with block cipher
algorithms which have the block size of 16 bytes.
The third argument @var{flags} can either be passed as @code{0} or as
the bit-wise OR of the following constants.
@table @code
@item GCRY_CIPHER_SECURE
Make sure that all operations are allocated in secure memory. This is
useful when the key material is highly confidential.
@item GCRY_CIPHER_ENABLE_SYNC
@cindex sync mode (OpenPGP)
This flag enables the CFB sync mode, which is a special feature of
Libgcrypt's CFB mode implementation to allow for OpenPGP's CFB variant.
See @code{gcry_cipher_sync}.
@item GCRY_CIPHER_CBC_CTS
@cindex cipher text stealing
Enable cipher text stealing (CTS) for the CBC mode. Cannot be used
simultaneous as GCRY_CIPHER_CBC_MAC. CTS mode makes it possible to
transform data of almost arbitrary size (only limitation is that it
must be greater than the algorithm's block size).
@item GCRY_CIPHER_CBC_MAC
@cindex CBC-MAC
Compute CBC-MAC keyed checksums. This is the same as CBC mode, but
only output the last block. Cannot be used simultaneous as
GCRY_CIPHER_CBC_CTS.
@end table
@end deftypefun
Use the following function to release an existing handle:
@deftypefun void gcry_cipher_close (gcry_cipher_hd_t @var{h})
This function releases the context created by @code{gcry_cipher_open}.
It also zeroises all sensitive information associated with this cipher
handle.
@end deftypefun
In order to use a handle for performing cryptographic operations, a
`key' has to be set first:
@deftypefun gcry_error_t gcry_cipher_setkey (gcry_cipher_hd_t @var{h}, const void *@var{k}, size_t @var{l})
Set the key @var{k} used for encryption or decryption in the context
denoted by the handle @var{h}. The length @var{l} (in bytes) of the
key @var{k} must match the required length of the algorithm set for
this context or be in the allowed range for algorithms with variable
key size. The function checks this and returns an error if there is a
problem. A caller should always check for an error.
@end deftypefun
Most crypto modes requires an initialization vector (IV), which
usually is a non-secret random string acting as a kind of salt value.
The CTR mode requires a counter, which is also similar to a salt
value. To set the IV or CTR, use these functions:
@deftypefun gcry_error_t gcry_cipher_setiv (gcry_cipher_hd_t @var{h}, const void *@var{k}, size_t @var{l})
Set the initialization vector used for encryption or decryption. The
vector is passed as the buffer @var{K} of length @var{l} bytes and
copied to internal data structures. The function checks that the IV
matches the requirement of the selected algorithm and mode.
This function is also used by AEAD modes and with Salsa20 and ChaCha20
stream ciphers to set or update the required nonce. In these cases it
needs to be called after setting the key.
@end deftypefun
@deftypefun gcry_error_t gcry_cipher_setctr (gcry_cipher_hd_t @var{h}, const void *@var{c}, size_t @var{l})
Set the counter vector used for encryption or decryption. The counter
is passed as the buffer @var{c} of length @var{l} bytes and copied to
internal data structures. The function checks that the counter
matches the requirement of the selected algorithm (i.e., it must be
the same size as the block size).
@end deftypefun
@deftypefun gcry_error_t gcry_cipher_reset (gcry_cipher_hd_t @var{h})
Set the given handle's context back to the state it had after the last
call to gcry_cipher_setkey and clear the initialization vector.
Note that gcry_cipher_reset is implemented as a macro.
@end deftypefun
Authenticated Encryption with Associated Data (AEAD) block cipher
modes require the handling of the authentication tag and the additional
authenticated data, which can be done by using the following
functions:
@deftypefun gcry_error_t gcry_cipher_authenticate (gcry_cipher_hd_t @var{h}, const void *@var{abuf}, size_t @var{abuflen})
Process the buffer @var{abuf} of length @var{abuflen} as the additional
authenticated data (AAD) for AEAD cipher modes.
@end deftypefun
@deftypefun {gcry_error_t} gcry_cipher_gettag @
(@w{gcry_cipher_hd_t @var{h}}, @
@w{void *@var{tag}}, @w{size_t @var{taglen}})
This function is used to read the authentication tag after encryption.
The function finalizes and outputs the authentication tag to the buffer
@var{tag} of length @var{taglen} bytes.
Depending on the used mode certain restrictions for @var{taglen} are
enforced: For GCM @var{taglen} must be at least 16 or one of the
allowed truncated lengths (4, 8, 12, 13, 14, or 15).
@end deftypefun
@deftypefun {gcry_error_t} gcry_cipher_checktag @
(@w{gcry_cipher_hd_t @var{h}}, @
@w{const void *@var{tag}}, @w{size_t @var{taglen}})
Check the authentication tag after decryption. The authentication
tag is passed as the buffer @var{tag} of length @var{taglen} bytes
and compared to internal authentication tag computed during
decryption. Error code @code{GPG_ERR_CHECKSUM} is returned if
the authentication tag in the buffer @var{tag} does not match
the authentication tag calculated during decryption.
Depending on the used mode certain restrictions for @var{taglen} are
enforced: For GCM @var{taglen} must either be 16 or one of the allowed
truncated lengths (4, 8, 12, 13, 14, or 15).
@end deftypefun
The actual encryption and decryption is done by using one of the
following functions. They may be used as often as required to process
all the data.
@deftypefun gcry_error_t gcry_cipher_encrypt (gcry_cipher_hd_t @var{h}, unsigned char *{out}, size_t @var{outsize}, const unsigned char *@var{in}, size_t @var{inlen})
@code{gcry_cipher_encrypt} is used to encrypt the data. This function
can either work in place or with two buffers. It uses the cipher
context already setup and described by the handle @var{h}. There are 2
ways to use the function: If @var{in} is passed as @code{NULL} and
@var{inlen} is @code{0}, in-place encryption of the data in @var{out} of
length @var{outsize} takes place. With @var{in} being not @code{NULL},
@var{inlen} bytes are encrypted to the buffer @var{out} which must have
at least a size of @var{inlen}. @var{outsize} must be set to the
allocated size of @var{out}, so that the function can check that there
is sufficient space. Note that overlapping buffers are not allowed.
Depending on the selected algorithms and encryption mode, the length of
the buffers must be a multiple of the block size.
Some encryption modes require that @code{gcry_cipher_final} is used
before the final data chunk is passed to this function.
The function returns @code{0} on success or an error code.
@end deftypefun
@deftypefun gcry_error_t gcry_cipher_decrypt (gcry_cipher_hd_t @var{h}, unsigned char *{out}, size_t @var{outsize}, const unsigned char *@var{in}, size_t @var{inlen})
@code{gcry_cipher_decrypt} is used to decrypt the data. This function
can either work in place or with two buffers. It uses the cipher
context already setup and described by the handle @var{h}. There are 2
ways to use the function: If @var{in} is passed as @code{NULL} and
@var{inlen} is @code{0}, in-place decryption of the data in @var{out} or
length @var{outsize} takes place. With @var{in} being not @code{NULL},
@var{inlen} bytes are decrypted to the buffer @var{out} which must have
at least a size of @var{inlen}. @var{outsize} must be set to the
allocated size of @var{out}, so that the function can check that there
is sufficient space. Note that overlapping buffers are not allowed.
Depending on the selected algorithms and encryption mode, the length of
the buffers must be a multiple of the block size.
Some encryption modes require that @code{gcry_cipher_final} is used
before the final data chunk is passed to this function.
The function returns @code{0} on success or an error code.
@end deftypefun
The OCB mode features integrated padding and must thus be told about
the end of the input data. This is done with:
@deftypefun gcry_error_t gcry_cipher_final (gcry_cipher_hd_t @var{h})
Set a flag in the context to tell the encrypt and decrypt functions
that their next call will provide the last chunk of data. Only the
first call to this function has an effect and only for modes which
support it. Checking the error in in general not necessary. This is
implemented as a macro.
@end deftypefun
OpenPGP (as defined in RFC-4880) requires a special sync operation in
some places. The following function is used for this:
@deftypefun gcry_error_t gcry_cipher_sync (gcry_cipher_hd_t @var{h})
Perform the OpenPGP sync operation on context @var{h}. Note that this
is a no-op unless the context was created with the flag
@code{GCRY_CIPHER_ENABLE_SYNC}
@end deftypefun
Some of the described functions are implemented as macros utilizing a
catch-all control function. This control function is rarely used
directly but there is nothing which would inhibit it:
@deftypefun gcry_error_t gcry_cipher_ctl (gcry_cipher_hd_t @var{h}, int @var{cmd}, void *@var{buffer}, size_t @var{buflen})
@code{gcry_cipher_ctl} controls various aspects of the cipher module and
specific cipher contexts. Usually some more specialized functions or
macros are used for this purpose. The semantics of the function and its
parameters depends on the the command @var{cmd} and the passed context
handle @var{h}. Please see the comments in the source code
(@code{src/global.c}) for details.
@end deftypefun
@deftypefun gcry_error_t gcry_cipher_info (gcry_cipher_hd_t @var{h}, @
int @var{what}, void *@var{buffer}, size_t *@var{nbytes})
@code{gcry_cipher_info} is used to retrieve various
information about a cipher context or the cipher module in general.
@c begin constants for gcry_cipher_info
@table @code
@item GCRYCTL_GET_TAGLEN:
Return the length of the tag for an AE algorithm mode. An error is
returned for modes which do not support a tag. @var{buffer} must be
given as NULL. On success the result is stored @var{nbytes}. The
taglen is returned in bytes.
@end table
@c end constants for gcry_cipher_info
@end deftypefun
@node General cipher functions
@section General cipher functions
To work with the algorithms, several functions are available to map
algorithm names to the internal identifiers, as well as ways to
retrieve information about an algorithm or the current cipher context.
@deftypefun gcry_error_t gcry_cipher_algo_info (int @var{algo}, int @var{what}, void *@var{buffer}, size_t *@var{nbytes})
This function is used to retrieve information on a specific algorithm.
You pass the cipher algorithm ID as @var{algo} and the type of
information requested as @var{what}. The result is either returned as
the return code of the function or copied to the provided @var{buffer}
whose allocated length must be available in an integer variable with the
address passed in @var{nbytes}. This variable will also receive the
actual used length of the buffer.
Here is a list of supported codes for @var{what}:
@c begin constants for gcry_cipher_algo_info
@table @code
@item GCRYCTL_GET_KEYLEN:
Return the length of the key. If the algorithm supports multiple key
lengths, the maximum supported value is returned. The length is
returned as number of octets (bytes) and not as number of bits in
@var{nbytes}; @var{buffer} must be zero. Note that it is usually
better to use the convenience function
@code{gcry_cipher_get_algo_keylen}.
@item GCRYCTL_GET_BLKLEN:
Return the block length of the algorithm. The length is returned as a
number of octets in @var{nbytes}; @var{buffer} must be zero. Note
that it is usually better to use the convenience function
@code{gcry_cipher_get_algo_blklen}.
@item GCRYCTL_TEST_ALGO:
Returns @code{0} when the specified algorithm is available for use.
@var{buffer} and @var{nbytes} must be zero.
@end table
@c end constants for gcry_cipher_algo_info
@end deftypefun
@c end gcry_cipher_algo_info
@deftypefun size_t gcry_cipher_get_algo_keylen (@var{algo})
This function returns length of the key for algorithm @var{algo}. If
the algorithm supports multiple key lengths, the maximum supported key
length is returned. On error @code{0} is returned. The key length is
returned as number of octets.
This is a convenience functions which should be preferred over
@code{gcry_cipher_algo_info} because it allows for proper type
checking.
@end deftypefun
@c end gcry_cipher_get_algo_keylen
@deftypefun size_t gcry_cipher_get_algo_blklen (int @var{algo})
This functions returns the block-length of the algorithm @var{algo}
counted in octets. On error @code{0} is returned.
This is a convenience functions which should be preferred over
@code{gcry_cipher_algo_info} because it allows for proper type
checking.
@end deftypefun
@c end gcry_cipher_get_algo_blklen
@deftypefun {const char *} gcry_cipher_algo_name (int @var{algo})
@code{gcry_cipher_algo_name} returns a string with the name of the
cipher algorithm @var{algo}. If the algorithm is not known or another
error occurred, the string @code{"?"} is returned. This function should
not be used to test for the availability of an algorithm.
@end deftypefun
@deftypefun int gcry_cipher_map_name (const char *@var{name})
@code{gcry_cipher_map_name} returns the algorithm identifier for the
cipher algorithm described by the string @var{name}. If this algorithm
is not available @code{0} is returned.
@end deftypefun
@deftypefun int gcry_cipher_mode_from_oid (const char *@var{string})
Return the cipher mode associated with an @acronym{ASN.1} object
identifier. The object identifier is expected to be in the
@acronym{IETF}-style dotted decimal notation. The function returns
@code{0} for an unknown object identifier or when no mode is associated
with it.
@end deftypefun
@c **********************************************************
@c ******************* Public Key *************************
@c **********************************************************
@node Public Key cryptography
@chapter Public Key cryptography
Public key cryptography, also known as asymmetric cryptography, is an
easy way for key management and to provide digital signatures.
Libgcrypt provides two completely different interfaces to
public key cryptography, this chapter explains the one based on
S-expressions.
@menu
* Available algorithms:: Algorithms supported by the library.
* Used S-expressions:: Introduction into the used S-expression.
* Cryptographic Functions:: Functions for performing the cryptographic actions.
* General public-key related Functions:: General functions, not implementing any cryptography.
@end menu
@node Available algorithms
@section Available algorithms
Libgcrypt supports the RSA (Rivest-Shamir-Adleman) algorithms as well
as DSA (Digital Signature Algorithm) and Elgamal. The versatile
interface allows to add more algorithms in the future.
@node Used S-expressions
@section Used S-expressions
Libgcrypt's API for asymmetric cryptography is based on data structures
called S-expressions (see
@uref{http://people.csail.mit.edu/@/rivest/@/sexp.html}) and does not work
with contexts as most of the other building blocks of Libgcrypt do.
@noindent
The following information are stored in S-expressions:
@itemize
@item keys
@item plain text data
@item encrypted data
@item signatures
@end itemize
@noindent
To describe how Libgcrypt expect keys, we use examples. Note that
words in
@ifnottex
uppercase
@end ifnottex
@iftex
italics
@end iftex
indicate parameters whereas lowercase words are literals.
Note that all MPI (multi-precision-integers) values are expected to be in
@code{GCRYMPI_FMT_USG} format. An easy way to create S-expressions is
by using @code{gcry_sexp_build} which allows to pass a string with
printf-like escapes to insert MPI values.
@menu
* RSA key parameters:: Parameters used with an RSA key.
* DSA key parameters:: Parameters used with a DSA key.
* ECC key parameters:: Parameters used with ECC keys.
@end menu
@node RSA key parameters
@subsection RSA key parameters
@noindent
An RSA private key is described by this S-expression:
@example
(private-key
(rsa
(n @var{n-mpi})
(e @var{e-mpi})
(d @var{d-mpi})
(p @var{p-mpi})
(q @var{q-mpi})
(u @var{u-mpi})))
@end example
@noindent
An RSA public key is described by this S-expression:
@example
(public-key
(rsa
(n @var{n-mpi})
(e @var{e-mpi})))
@end example
@table @var
@item n-mpi
RSA public modulus @math{n}.
@item e-mpi
RSA public exponent @math{e}.
@item d-mpi
RSA secret exponent @math{d = e^{-1} \bmod (p-1)(q-1)}.
@item p-mpi
RSA secret prime @math{p}.
@item q-mpi
RSA secret prime @math{q} with @math{p < q}.
@item u-mpi
Multiplicative inverse @math{u = p^{-1} \bmod q}.
@end table
For signing and decryption the parameters @math{(p, q, u)} are optional
but greatly improve the performance. Either all of these optional
parameters must be given or none of them. They are mandatory for
gcry_pk_testkey.
Note that OpenSSL uses slighly different parameters: @math{q < p} and
@math{u = q^{-1} \bmod p}. To use these parameters you will need to
swap the values and recompute @math{u}. Here is example code to do this:
@example
if (gcry_mpi_cmp (p, q) > 0)
@{
gcry_mpi_swap (p, q);
gcry_mpi_invm (u, p, q);
@}
@end example
@node DSA key parameters
@subsection DSA key parameters
@noindent
A DSA private key is described by this S-expression:
@example
(private-key
(dsa
(p @var{p-mpi})
(q @var{q-mpi})
(g @var{g-mpi})
(y @var{y-mpi})
(x @var{x-mpi})))
@end example
@table @var
@item p-mpi
DSA prime @math{p}.
@item q-mpi
DSA group order @math{q} (which is a prime divisor of @math{p-1}).
@item g-mpi
DSA group generator @math{g}.
@item y-mpi
DSA public key value @math{y = g^x \bmod p}.
@item x-mpi
DSA secret exponent x.
@end table
The public key is similar with "private-key" replaced by "public-key"
and no @var{x-mpi}.
@node ECC key parameters
@subsection ECC key parameters
@anchor{ecc_keyparam}
@noindent
An ECC private key is described by this S-expression:
@example
(private-key
(ecc
(p @var{p-mpi})
(a @var{a-mpi})
(b @var{b-mpi})
(g @var{g-point})
(n @var{n-mpi})
(q @var{q-point})
(d @var{d-mpi})))
@end example
@table @var
@item p-mpi
Prime specifying the field @math{GF(p)}.
@item a-mpi
@itemx b-mpi
The two coefficients of the Weierstrass equation @math{y^2 = x^3 + ax + b}
@item g-point
Base point @math{g}.
@item n-mpi
Order of @math{g}
@item q-point
The point representing the public key @math{Q = dG}.
@item d-mpi
The private key @math{d}
@end table
All point values are encoded in standard format; Libgcrypt does in
general only support uncompressed points, thus the first byte needs to
be @code{0x04}. However ``EdDSA'' describes its own compression
scheme which is used by default; the non-standard first byte
@code{0x40} may optionally be used to explicit flag the use of the
algorithm’s native compression method.
The public key is similar with "private-key" replaced by "public-key"
and no @var{d-mpi}.
If the domain parameters are well-known, the name of this curve may be
used. For example
@example
(private-key
(ecc
(curve "NIST P-192")
(q @var{q-point})
(d @var{d-mpi})))
@end example
Note that @var{q-point} is optional for a private key. The
@code{curve} parameter may be given in any case and is used to replace
missing parameters.
@noindent
Currently implemented curves are:
@table @code
@item NIST P-192
@itemx 1.2.840.10045.3.1.1
@itemx prime192v1
@itemx secp192r1
The NIST 192 bit curve, its OID, X9.62 and SECP aliases.
@item NIST P-224
@itemx secp224r1
The NIST 224 bit curve and its SECP alias.
@item NIST P-256
@itemx 1.2.840.10045.3.1.7
@itemx prime256v1
@itemx secp256r1
The NIST 256 bit curve, its OID, X9.62 and SECP aliases.
@item NIST P-384
@itemx secp384r1
The NIST 384 bit curve and its SECP alias.
@item NIST P-521
@itemx secp521r1
The NIST 521 bit curve and its SECP alias.
@end table
As usual the OIDs may optionally be prefixed with the string @code{OID.}
or @code{oid.}.
@node Cryptographic Functions
@section Cryptographic Functions
@noindent
Some functions operating on S-expressions support `flags' to influence
the operation. These flags have to be listed in a sub-S-expression
named `flags'. Flag names are case-sensitive. The following flags
are known:
@table @code
@item comp
@itemx nocomp
@cindex comp
@cindex nocomp
If supported by the algorithm and curve the @code{comp} flag requests
that points are returned in compact (compressed) representation. The
@code{nocomp} flag requests that points are returned with full
coordinates. The default depends on the the algorithm and curve. The
compact representation requires a small overhead before a point can be
used but halves the size of a to be conveyed public key. If
@code{comp} is used with the ``EdDSA'' algorithm the key generation
prefix the public key with a @code{0x40} byte.
@item pkcs1
@cindex PKCS1
Use PKCS#1 block type 2 padding for encryption, block type 1 padding
for signing.
@item oaep
@cindex OAEP
Use RSA-OAEP padding for encryption.
@item pss
@cindex PSS
Use RSA-PSS padding for signing.
@item eddsa
@cindex EdDSA
Use the EdDSA scheme signing instead of the default ECDSA algorithm.
Note that the EdDSA uses a special form of the public key.
@item rfc6979
@cindex RFC6979
For DSA and ECDSA use a deterministic scheme for the k parameter.
@item no-blinding
@cindex no-blinding
Do not use a technique called `blinding', which is used by default in
order to prevent leaking of secret information. Blinding is only
implemented by RSA, but it might be implemented by other algorithms in
the future as well, when necessary.
@item param
@cindex param
For ECC key generation also return the domain parameters. For ECC
signing and verification override default parameters by provided
domain parameters of the public or private key.
@item transient-key
@cindex transient-key
This flag is only meaningful for RSA, DSA, and ECC key generation. If
given the key is created using a faster and a somewhat less secure
random number generator. This flag may be used for keys which are
only used for a short time or per-message and do not require full
cryptographic strength.
@item no-keytest
@cindex no-keytest
This flag skips internal failsafe tests to assert that a generated key
is properly working. It currently has an effect only for standard ECC
key generation. It is mostly useful along with transient-key to
achieve fastest ECC key generation.
@item use-x931
@cindex X9.31
Force the use of the ANSI X9.31 key generation algorithm instead of
the default algorithm. This flag is only meaningful for RSA key
generation and usually not required. Note that this algorithm is
implicitly used if either @code{derive-parms} is given or Libgcrypt is
in FIPS mode.
@item use-fips186
@cindex FIPS 186
Force the use of the FIPS 186 key generation algorithm instead of the
default algorithm. This flag is only meaningful for DSA and usually
not required. Note that this algorithm is implicitly used if either
@code{derive-parms} is given or Libgcrypt is in FIPS mode. As of now
FIPS 186-2 is implemented; after the approval of FIPS 186-3 the code
will be changed to implement 186-3.
@item use-fips186-2
@cindex FIPS 186-2
Force the use of the FIPS 186-2 key generation algorithm instead of
the default algorithm. This algorithm is slightly different from
FIPS 186-3 and allows only 1024 bit keys. This flag is only meaningful
for DSA and only required for FIPS testing backward compatibility.
@end table
@noindent
Now that we know the key basics, we can carry on and explain how to
encrypt and decrypt data. In almost all cases the data is a random
session key which is in turn used for the actual encryption of the real
data. There are 2 functions to do this:
@deftypefun gcry_error_t gcry_pk_encrypt (@w{gcry_sexp_t *@var{r_ciph},} @w{gcry_sexp_t @var{data},} @w{gcry_sexp_t @var{pkey}})
Obviously a public key must be provided for encryption. It is
expected as an appropriate S-expression (see above) in @var{pkey}.
The data to be encrypted can either be in the simple old format, which
is a very simple S-expression consisting only of one MPI, or it may be
a more complex S-expression which also allows to specify flags for
operation, like e.g. padding rules.
@noindent
If you don't want to let Libgcrypt handle the padding, you must pass an
appropriate MPI using this expression for @var{data}:
@example
(data
(flags raw)
(value @var{mpi}))
@end example
@noindent
This has the same semantics as the old style MPI only way. @var{MPI}
is the actual data, already padded appropriate for your protocol.
Most RSA based systems however use PKCS#1 padding and so you can use
this S-expression for @var{data}:
@example
(data
(flags pkcs1)
(value @var{block}))
@end example
@noindent
Here, the "flags" list has the "pkcs1" flag which let the function know
that it should provide PKCS#1 block type 2 padding. The actual data to
be encrypted is passed as a string of octets in @var{block}. The
function checks that this data actually can be used with the given key,
does the padding and encrypts it.
If the function could successfully perform the encryption, the return
value will be 0 and a new S-expression with the encrypted result is
allocated and assigned to the variable at the address of @var{r_ciph}.
The caller is responsible to release this value using
@code{gcry_sexp_release}. In case of an error, an error code is
returned and @var{r_ciph} will be set to @code{NULL}.
@noindent
The returned S-expression has this format when used with RSA:
@example
(enc-val
(rsa
(a @var{a-mpi})))
@end example
@noindent
Where @var{a-mpi} is an MPI with the result of the RSA operation. When
using the Elgamal algorithm, the return value will have this format:
@example
(enc-val
(elg
(a @var{a-mpi})
(b @var{b-mpi})))
@end example
@noindent
Where @var{a-mpi} and @var{b-mpi} are MPIs with the result of the
Elgamal encryption operation.
@end deftypefun
@c end gcry_pk_encrypt
@deftypefun gcry_error_t gcry_pk_decrypt (@w{gcry_sexp_t *@var{r_plain},} @w{gcry_sexp_t @var{data},} @w{gcry_sexp_t @var{skey}})
Obviously a private key must be provided for decryption. It is expected
as an appropriate S-expression (see above) in @var{skey}. The data to
be decrypted must match the format of the result as returned by
@code{gcry_pk_encrypt}, but should be enlarged with a @code{flags}
element:
@example
(enc-val
(flags)
(elg
(a @var{a-mpi})
(b @var{b-mpi})))
@end example
@noindent
This function does not remove padding from the data by default. To
let Libgcrypt remove padding, give a hint in `flags' telling which
padding method was used when encrypting:
@example
(flags @var{padding-method})
@end example
@noindent
Currently @var{padding-method} is either @code{pkcs1} for PKCS#1 block
type 2 padding, or @code{oaep} for RSA-OAEP padding.
@noindent
The function returns 0 on success or an error code. The variable at the
address of @var{r_plain} will be set to NULL on error or receive the
decrypted value on success. The format of @var{r_plain} is a
simple S-expression part (i.e. not a valid one) with just one MPI if
there was no @code{flags} element in @var{data}; if at least an empty
@code{flags} is passed in @var{data}, the format is:
@example
(value @var{plaintext})
@end example
@end deftypefun
@c end gcry_pk_decrypt
Another operation commonly performed using public key cryptography is
signing data. In some sense this is even more important than
encryption because digital signatures are an important instrument for
key management. Libgcrypt supports digital signatures using
2 functions, similar to the encryption functions:
@deftypefun gcry_error_t gcry_pk_sign (@w{gcry_sexp_t *@var{r_sig},} @w{gcry_sexp_t @var{data},} @w{gcry_sexp_t @var{skey}})
This function creates a digital signature for @var{data} using the
private key @var{skey} and place it into the variable at the address of
@var{r_sig}. @var{data} may either be the simple old style S-expression
with just one MPI or a modern and more versatile S-expression which
allows to let Libgcrypt handle padding:
@example
(data
(flags pkcs1)
(hash @var{hash-algo} @var{block}))
@end example
@noindent
This example requests to sign the data in @var{block} after applying
PKCS#1 block type 1 style padding. @var{hash-algo} is a string with the
hash algorithm to be encoded into the signature, this may be any hash
algorithm name as supported by Libgcrypt. Most likely, this will be
"sha256" or "sha1". It is obvious that the length of @var{block} must
match the size of that message digests; the function checks that this
and other constraints are valid.
@noindent
If PKCS#1 padding is not required (because the caller does already
provide a padded value), either the old format or better the following
format should be used:
@example
(data
(flags raw)
(value @var{mpi}))
@end example
@noindent
Here, the data to be signed is directly given as an @var{MPI}.
@noindent
For DSA the input data is expected in this format:
@example
(data
(flags raw)
(value @var{mpi}))
@end example
@noindent
Here, the data to be signed is directly given as an @var{MPI}. It is
expect that this MPI is the the hash value. For the standard DSA
using a MPI is not a problem in regard to leading zeroes because the
hash value is directly used as an MPI. For better standard
conformance it would be better to explicit use a memory string (like
with pkcs1) but that is currently not supported. However, for
deterministic DSA as specified in RFC6979 this can't be used. Instead
the following input is expected.
@example
(data
(flags rfc6979)
(hash @var{hash-algo} @var{block}))
@end example
Note that the provided hash-algo is used for the internal HMAC; it
should match the hash-algo used to create @var{block}.
@noindent
The signature is returned as a newly allocated S-expression in
@var{r_sig} using this format for RSA:
@example
(sig-val
(rsa
(s @var{s-mpi})))
@end example
Where @var{s-mpi} is the result of the RSA sign operation. For DSA the
S-expression returned is:
@example
(sig-val
(dsa
(r @var{r-mpi})
(s @var{s-mpi})))
@end example
Where @var{r-mpi} and @var{s-mpi} are the result of the DSA sign
operation.
For Elgamal signing (which is slow, yields large numbers and probably
is not as secure as the other algorithms), the same format is used
with "elg" replacing "dsa"; for ECDSA signing, the same format is used
with "ecdsa" replacing "dsa".
For the EdDSA algorithm (cf. Ed25515) the required input parameters are:
@example
(data
(flags eddsa)
(hash-algo sha512)
(value @var{message}))
@end example
Note that the @var{message} may be of any length; hashing is part of
the algorithm. Using a large data block for @var{message} is not
suggested; in that case the used protocol should better require that a
hash of the message is used as input to the EdDSA algorithm.
@end deftypefun
@c end gcry_pk_sign
@noindent
The operation most commonly used is definitely the verification of a
signature. Libgcrypt provides this function:
@deftypefun gcry_error_t gcry_pk_verify (@w{gcry_sexp_t @var{sig}}, @w{gcry_sexp_t @var{data}}, @w{gcry_sexp_t @var{pkey}})
This is used to check whether the signature @var{sig} matches the
@var{data}. The public key @var{pkey} must be provided to perform this
verification. This function is similar in its parameters to
@code{gcry_pk_sign} with the exceptions that the public key is used
instead of the private key and that no signature is created but a
signature, in a format as created by @code{gcry_pk_sign}, is passed to
the function in @var{sig}.
@noindent
The result is 0 for success (i.e. the data matches the signature), or an
error code where the most relevant code is @code{GCRY_ERR_BAD_SIGNATURE}
to indicate that the signature does not match the provided data.
@end deftypefun
@c end gcry_pk_verify
@node General public-key related Functions
@section General public-key related Functions
@noindent
A couple of utility functions are available to retrieve the length of
the key, map algorithm identifiers and perform sanity checks:
@deftypefun {const char *} gcry_pk_algo_name (int @var{algo})
Map the public key algorithm id @var{algo} to a string representation of
the algorithm name. For unknown algorithms this functions returns the
string @code{"?"}. This function should not be used to test for the
availability of an algorithm.
@end deftypefun
@deftypefun int gcry_pk_map_name (const char *@var{name})
Map the algorithm @var{name} to a public key algorithm Id. Returns 0 if
the algorithm name is not known.
@end deftypefun
@deftypefun int gcry_pk_test_algo (int @var{algo})
Return 0 if the public key algorithm @var{algo} is available for use.
Note that this is implemented as a macro.
@end deftypefun
@deftypefun {unsigned int} gcry_pk_get_nbits (gcry_sexp_t @var{key})
Return what is commonly referred as the key length for the given
public or private in @var{key}.
@end deftypefun
@deftypefun {unsigned char *} gcry_pk_get_keygrip (@w{gcry_sexp_t @var{key}}, @w{unsigned char *@var{array}})
Return the so called "keygrip" which is the SHA-1 hash of the public key
parameters expressed in a way depended on the algorithm. @var{array}
must either provide space for 20 bytes or be @code{NULL}. In the latter
case a newly allocated array of that size is returned. On success a
pointer to the newly allocated space or to @var{array} is returned.
@code{NULL} is returned to indicate an error which is most likely an
unknown algorithm or one where a "keygrip" has not yet been defined.
The function accepts public or secret keys in @var{key}.
@end deftypefun
@deftypefun gcry_error_t gcry_pk_testkey (gcry_sexp_t @var{key})
Return zero if the private key @var{key} is `sane', an error code otherwise.
Note that it is not possible to check the `saneness' of a public key.
@end deftypefun
@deftypefun gcry_error_t gcry_pk_algo_info (@w{int @var{algo}}, @w{int @var{what}}, @w{void *@var{buffer}}, @w{size_t *@var{nbytes}})
Depending on the value of @var{what} return various information about
the public key algorithm with the id @var{algo}. Note that the
function returns @code{-1} on error and the actual error code must be
retrieved using the function @code{gcry_errno}. The currently defined
values for @var{what} are:
@table @code
@item GCRYCTL_TEST_ALGO:
Return 0 if the specified algorithm is available for use.
@var{buffer} must be @code{NULL}, @var{nbytes} may be passed as
@code{NULL} or point to a variable with the required usage of the
algorithm. This may be 0 for "don't care" or the bit-wise OR of these
flags:
@table @code
@item GCRY_PK_USAGE_SIGN
Algorithm is usable for signing.
@item GCRY_PK_USAGE_ENCR
Algorithm is usable for encryption.
@end table
Unless you need to test for the allowed usage, it is in general better
to use the macro gcry_pk_test_algo instead.
@item GCRYCTL_GET_ALGO_USAGE:
Return the usage flags for the given algorithm. An invalid algorithm
return 0. Disabled algorithms are ignored here because we
want to know whether the algorithm is at all capable of a certain usage.
@item GCRYCTL_GET_ALGO_NPKEY
Return the number of elements the public key for algorithm @var{algo}
consist of. Return 0 for an unknown algorithm.
@item GCRYCTL_GET_ALGO_NSKEY
Return the number of elements the private key for algorithm @var{algo}
consist of. Note that this value is always larger than that of the
public key. Return 0 for an unknown algorithm.
@item GCRYCTL_GET_ALGO_NSIGN
Return the number of elements a signature created with the algorithm
@var{algo} consists of. Return 0 for an unknown algorithm or for an
algorithm not capable of creating signatures.
@item GCRYCTL_GET_ALGO_NENC
Return the number of elements a encrypted message created with the algorithm
@var{algo} consists of. Return 0 for an unknown algorithm or for an
algorithm not capable of encryption.
@end table
@noindent
Please note that parameters not required should be passed as @code{NULL}.
@end deftypefun
@c end gcry_pk_algo_info
@deftypefun gcry_error_t gcry_pk_ctl (@w{int @var{cmd}}, @w{void *@var{buffer}}, @w{size_t @var{buflen}})
This is a general purpose function to perform certain control
operations. @var{cmd} controls what is to be done. The return value is
0 for success or an error code. Currently supported values for
@var{cmd} are:
@table @code
@item GCRYCTL_DISABLE_ALGO
Disable the algorithm given as an algorithm id in @var{buffer}.
@var{buffer} must point to an @code{int} variable with the algorithm
id and @var{buflen} must have the value @code{sizeof (int)}. This
function is not thread safe and should thus be used before any other
threads are started.
@end table
@end deftypefun
@c end gcry_pk_ctl
@noindent
Libgcrypt also provides a function to generate public key
pairs:
@deftypefun gcry_error_t gcry_pk_genkey (@w{gcry_sexp_t *@var{r_key}}, @w{gcry_sexp_t @var{parms}})
This function create a new public key pair using information given in
the S-expression @var{parms} and stores the private and the public key
in one new S-expression at the address given by @var{r_key}. In case of
an error, @var{r_key} is set to @code{NULL}. The return code is 0 for
success or an error code otherwise.
@noindent
Here is an example for @var{parms} to create an 2048 bit RSA key:
@example
(genkey
(rsa
(nbits 4:2048)))
@end example
@noindent
To create an Elgamal key, substitute "elg" for "rsa" and to create a DSA
key use "dsa". Valid ranges for the key length depend on the
algorithms; all commonly used key lengths are supported. Currently
supported parameters are:
@table @code
@item nbits
This is always required to specify the length of the key. The
argument is a string with a number in C-notation. The value should be
a multiple of 8. Note that the S-expression syntax requires that a
number is prefixed with its string length; thus the @code{4:} in the
above example.
@item curve @var{name}
For ECC a named curve may be used instead of giving the number of
requested bits. This allows to request a specific curve to override a
default selection Libgcrypt would have taken if @code{nbits} has been
given. The available names are listed with the description of the ECC
public key parameters.
@item rsa-use-e @var{value}
This is only used with RSA to give a hint for the public exponent. The
@var{value} will be used as a base to test for a usable exponent. Some
values are special:
@table @samp
@item 0
Use a secure and fast value. This is currently the number 41.
@item 1
Use a value as required by some crypto policies. This is currently
the number 65537.
@item 2
Reserved
@item > 2
Use the given value.
@end table
@noindent
If this parameter is not used, Libgcrypt uses for historic reasons
65537.
@item qbits @var{n}
This is only meanigful for DSA keys. If it is given the DSA key is
generated with a Q parameyer of size @var{n} bits. If it is not given
or zero Q is deduced from NBITS in this way:
@table @samp
@item 512 <= N <= 1024
Q = 160
@item N = 2048
Q = 224
@item N = 3072
Q = 256
@item N = 7680
Q = 384
@item N = 15360
Q = 512
@end table
Note that in this case only the values for N, as given in the table,
are allowed. When specifying Q all values of N in the range 512 to
15680 are valid as long as they are multiples of 8.
@item domain @var{list}
This is only meaningful for DLP algorithms. If specified keys are
generated with domain parameters taken from this list. The exact
format of this parameter depends on the actual algorithm. It is
currently only implemented for DSA using this format:
@example
(genkey
(dsa
(domain
(p @var{p-mpi})
(q @var{q-mpi})
(g @var{q-mpi}))))
@end example
@code{nbits} and @code{qbits} may not be specified because they are
derived from the domain parameters.
@item derive-parms @var{list}
This is currently only implemented for RSA and DSA keys. It is not
allowed to use this together with a @code{domain} specification. If
given, it is used to derive the keys using the given parameters.
If given for an RSA key the X9.31 key generation algorithm is used
even if libgcrypt is not in FIPS mode. If given for a DSA key, the
FIPS 186 algorithm is used even if libgcrypt is not in FIPS mode.
@example
(genkey
(rsa
(nbits 4:1024)
(rsa-use-e 1:3)
(derive-parms
(Xp1 #1A1916DDB29B4EB7EB6732E128#)
(Xp2 #192E8AAC41C576C822D93EA433#)
(Xp #D8CD81F035EC57EFE822955149D3BFF70C53520D
769D6D76646C7A792E16EBD89FE6FC5B605A6493
39DFC925A86A4C6D150B71B9EEA02D68885F5009
B98BD984#)
(Xq1 #1A5CF72EE770DE50CB09ACCEA9#)
(Xq2 #134E4CAA16D2350A21D775C404#)
(Xq #CC1092495D867E64065DEE3E7955F2EBC7D47A2D
7C9953388F97DDDC3E1CA19C35CA659EDC2FC325
6D29C2627479C086A699A49C4C9CEE7EF7BD1B34
321DE34A#))))
@end example
@example
(genkey
(dsa
(nbits 4:1024)
(derive-parms
(seed @var{seed-mpi}))))
@end example
@item flags @var{flaglist}
This is preferred way to define flags. @var{flaglist} may contain any
number of flags. See above for a specification of these flags.
Here is an example on how to create a key using curve Ed25519 with the
ECDSA signature algorithm. Note that the use of ECDSA with that curve
is in general not recommended.
@example
(genkey
(ecc
(flags transient-key)))
@end example
@item transient-key
@itemx use-x931
@itemx use-fips186
@itemx use-fips186-2
These are deprecated ways to set a flag with that name; see above for
a description of each flag.
@end table
@c end table of parameters
@noindent
The key pair is returned in a format depending on the algorithm. Both
private and public keys are returned in one container and may be
accompanied by some miscellaneous information.
@noindent
Here are two examples; the first for Elgamal and the second for
elliptic curve key generation:
@example
(key-data
(public-key
(elg
(p @var{p-mpi})
(g @var{g-mpi})
(y @var{y-mpi})))
(private-key
(elg
(p @var{p-mpi})
(g @var{g-mpi})
(y @var{y-mpi})
(x @var{x-mpi})))
(misc-key-info
(pm1-factors @var{n1 n2 ... nn}))
@end example
@example
(key-data
(public-key
(ecc
(curve Ed25519)
(flags eddsa)
(q @var{q-value})))
(private-key
(ecc
(curve Ed25519)
(flags eddsa)
(q @var{q-value})
(d @var{d-value}))))
@end example
@noindent
As you can see, some of the information is duplicated, but this
provides an easy way to extract either the public or the private key.
Note that the order of the elements is not defined, e.g. the private
key may be stored before the public key. @var{n1 n2 ... nn} is a list
of prime numbers used to composite @var{p-mpi}; this is in general not
a very useful information and only available if the key generation
algorithm provides them.
@end deftypefun
@c end gcry_pk_genkey
@noindent
Future versions of Libgcrypt will have extended versions of the public
key interfaced which will take an additional context to allow for
pre-computations, special operations, and other optimization. As a
first step a new function is introduced to help using the ECC
algorithms in new ways:
@deftypefun gcry_error_t gcry_pubkey_get_sexp (@w{gcry_sexp_t *@var{r_sexp}}, @
@w{int @var{mode}}, @w{gcry_ctx_t @var{ctx}})
Return an S-expression representing the context @var{ctx}. Depending
on the state of that context, the S-expression may either be a public
key, a private key or any other object used with public key
operations. On success 0 is returned and a new S-expression is stored
at @var{r_sexp}; on error an error code is returned and NULL is stored
at @var{r_sexp}. @var{mode} must be one of:
@table @code
@item 0
Decide what to return depending on the context. For example if the
private key parameter is available a private key is returned, if not a
public key is returned.
@item GCRY_PK_GET_PUBKEY
Return the public key even if the context has the private key
parameter.
@item GCRY_PK_GET_SECKEY
Return the private key or the error @code{GPG_ERR_NO_SECKEY} if it is
not possible.
@end table
As of now this function supports only certain ECC operations because a
context object is right now only defined for ECC. Over time this
function will be extended to cover more algorithms.
@end deftypefun
@c end gcry_pubkey_get_sexp
@c **********************************************************
@c ******************* Hash Functions *********************
@c **********************************************************
@node Hashing
@chapter Hashing
Libgcrypt provides an easy and consistent to use interface for hashing.
Hashing is buffered and several hash algorithms can be updated at once.
It is possible to compute a HMAC using the same routines. The
programming model follows an open/process/close paradigm and is in that
similar to other building blocks provided by Libgcrypt.
For convenience reasons, a few cyclic redundancy check value operations
are also supported.
@menu
* Available hash algorithms:: List of hash algorithms supported by the library.
* Working with hash algorithms:: List of functions related to hashing.
@end menu
@node Available hash algorithms
@section Available hash algorithms
@c begin table of hash algorithms
@cindex SHA-1
@cindex SHA-224, SHA-256, SHA-384, SHA-512
@cindex SHA3-224, SHA3-256, SHA3-384, SHA3-512, SHAKE128, SHAKE256
@cindex RIPE-MD-160
@cindex MD2, MD4, MD5
@cindex TIGER, TIGER1, TIGER2
@cindex HAVAL
@cindex Whirlpool
@cindex CRC32
@table @code
@item GCRY_MD_NONE
This is not a real algorithm but used by some functions as an error
return value. This constant is guaranteed to have the value @code{0}.
@item GCRY_MD_SHA1
This is the SHA-1 algorithm which yields a message digest of 20 bytes.
Note that SHA-1 begins to show some weaknesses and it is suggested to
fade out its use if strong cryptographic properties are required.
@item GCRY_MD_RMD160
This is the 160 bit version of the RIPE message digest (RIPE-MD-160).
Like SHA-1 it also yields a digest of 20 bytes. This algorithm share a
lot of design properties with SHA-1 and thus it is advisable not to use
it for new protocols.
@item GCRY_MD_MD5
This is the well known MD5 algorithm, which yields a message digest of
16 bytes. Note that the MD5 algorithm has severe weaknesses, for
example it is easy to compute two messages yielding the same hash
(collision attack). The use of this algorithm is only justified for
non-cryptographic application.
@item GCRY_MD_MD4
This is the MD4 algorithm, which yields a message digest of 16 bytes.
This algorithm has severe weaknesses and should not be used.
@item GCRY_MD_MD2
This is an reserved identifier for MD-2; there is no implementation yet.
This algorithm has severe weaknesses and should not be used.
@item GCRY_MD_TIGER
This is the TIGER/192 algorithm which yields a message digest of 24
bytes. Actually this is a variant of TIGER with a different output
print order as used by GnuPG up to version 1.3.2.
@item GCRY_MD_TIGER1
This is the TIGER variant as used by the NESSIE project. It uses the
most commonly used output print order.
@item GCRY_MD_TIGER2
This is another variant of TIGER with a different padding scheme.
@item GCRY_MD_HAVAL
This is an reserved value for the HAVAL algorithm with 5 passes and 160
bit. It yields a message digest of 20 bytes. Note that there is no
implementation yet available.
@item GCRY_MD_SHA224
This is the SHA-224 algorithm which yields a message digest of 28 bytes.
See Change Notice 1 for FIPS 180-2 for the specification.
@item GCRY_MD_SHA256
This is the SHA-256 algorithm which yields a message digest of 32 bytes.
See FIPS 180-2 for the specification.
@item GCRY_MD_SHA384
This is the SHA-384 algorithm which yields a message digest of 48 bytes.
See FIPS 180-2 for the specification.
@item GCRY_MD_SHA512
This is the SHA-384 algorithm which yields a message digest of 64 bytes.
See FIPS 180-2 for the specification.
@item GCRY_MD_SHA3_224
This is the SHA3-224 algorithm which yields a message digest of 28 bytes.
See FIPS 202 for the specification.
@item GCRY_MD_SHA3_256
This is the SHA3-256 algorithm which yields a message digest of 32 bytes.
See FIPS 202 for the specification.
@item GCRY_MD_SHA3_384
This is the SHA3-384 algorithm which yields a message digest of 48 bytes.
See FIPS 202 for the specification.
@item GCRY_MD_SHA3_512
This is the SHA3-384 algorithm which yields a message digest of 64 bytes.
See FIPS 202 for the specification.
@item GCRY_MD_SHAKE128
This is the SHAKE128 extendable-output function (XOF) algorithm with 128 bit
security strength.
See FIPS 202 for the specification.
@item GCRY_MD_SHAKE256
This is the SHAKE256 extendable-output function (XOF) algorithm with 256 bit
security strength.
See FIPS 202 for the specification.
@item GCRY_MD_CRC32
This is the ISO 3309 and ITU-T V.42 cyclic redundancy check. It yields
an output of 4 bytes. Note that this is not a hash algorithm in the
cryptographic sense.
@item GCRY_MD_CRC32_RFC1510
This is the above cyclic redundancy check function, as modified by RFC
1510. It yields an output of 4 bytes. Note that this is not a hash
algorithm in the cryptographic sense.
@item GCRY_MD_CRC24_RFC2440
This is the OpenPGP cyclic redundancy check function. It yields an
output of 3 bytes. Note that this is not a hash algorithm in the
cryptographic sense.
@item GCRY_MD_WHIRLPOOL
This is the Whirlpool algorithm which yields a message digest of 64
bytes.
@item GCRY_MD_GOSTR3411_94
This is the hash algorithm described in GOST R 34.11-94 which yields a
message digest of 32 bytes.
@item GCRY_MD_STRIBOG256
This is the 256-bit version of hash algorithm described in GOST R 34.11-2012
which yields a message digest of 32 bytes.
@item GCRY_MD_STRIBOG512
This is the 512-bit version of hash algorithm described in GOST R 34.11-2012
which yields a message digest of 64 bytes.
@end table
@c end table of hash algorithms
@node Working with hash algorithms
@section Working with hash algorithms
To use most of these function it is necessary to create a context;
this is done using:
@deftypefun gcry_error_t gcry_md_open (gcry_md_hd_t *@var{hd}, int @var{algo}, unsigned int @var{flags})
Create a message digest object for algorithm @var{algo}. @var{flags}
may be given as an bitwise OR of constants described below. @var{algo}
may be given as @code{0} if the algorithms to use are later set using
@code{gcry_md_enable}. @var{hd} is guaranteed to either receive a valid
handle or NULL.
For a list of supported algorithms, see @xref{Available hash
algorithms}.
The flags allowed for @var{mode} are:
@c begin table of hash flags
@table @code
@item GCRY_MD_FLAG_SECURE
Allocate all buffers and the resulting digest in "secure memory". Use
this is the hashed data is highly confidential.
@item GCRY_MD_FLAG_HMAC
@cindex HMAC
Turn the algorithm into a HMAC message authentication algorithm. This
only works if just one algorithm is enabled for the handle and that
algorithm is not an extendable-output function. Note that the function
@code{gcry_md_setkey} must be used to set the MAC key. The size of the
MAC is equal to the message digest of the underlying hash algorithm.
If you want CBC message authentication codes based on a cipher,
see @xref{Working with cipher handles}.
@item GCRY_MD_FLAG_BUGEMU1
@cindex bug emulation
Versions of Libgcrypt before 1.6.0 had a bug in the Whirlpool code
which led to a wrong result for certain input sizes and write
patterns. Using this flag emulates that bug. This may for example be
useful for applications which use Whirlpool as part of their key
generation. It is strongly suggested to use this flag only if really
needed and if possible to the data should be re-processed using the
regular Whirlpool algorithm.
Note that this flag works for the entire hash context. If needed
arises it may be used to enable bug emulation for other hash
algorithms. Thus you should not use this flag for a multi-algorithm
hash context.
@end table
@c begin table of hash flags
You may use the function @code{gcry_md_is_enabled} to later check
whether an algorithm has been enabled.
@end deftypefun
@c end function gcry_md_open
If you want to calculate several hash algorithms at the same time, you
have to use the following function right after the @code{gcry_md_open}:
@deftypefun gcry_error_t gcry_md_enable (gcry_md_hd_t @var{h}, int @var{algo})
Add the message digest algorithm @var{algo} to the digest object
described by handle @var{h}. Duplicated enabling of algorithms is
detected and ignored.
@end deftypefun
If the flag @code{GCRY_MD_FLAG_HMAC} was used, the key for the MAC must
be set using the function:
@deftypefun gcry_error_t gcry_md_setkey (gcry_md_hd_t @var{h}, const void *@var{key}, size_t @var{keylen})
For use with the HMAC feature, set the MAC key to the value of
@var{key} of length @var{keylen} bytes. There is no restriction on
the length of the key.
@end deftypefun
After you are done with the hash calculation, you should release the
resources by using:
@deftypefun void gcry_md_close (gcry_md_hd_t @var{h})
Release all resources of hash context @var{h}. @var{h} should not be
used after a call to this function. A @code{NULL} passed as @var{h} is
ignored. The function also zeroises all sensitive information
associated with this handle.
@end deftypefun
Often you have to do several hash operations using the same algorithm.
To avoid the overhead of creating and releasing context, a reset function
is provided:
@deftypefun void gcry_md_reset (gcry_md_hd_t @var{h})
Reset the current context to its initial state. This is effectively
identical to a close followed by an open and enabling all currently
active algorithms.
@end deftypefun
Often it is necessary to start hashing some data and then continue to
hash different data. To avoid hashing the same data several times (which
might not even be possible if the data is received from a pipe), a
snapshot of the current hash context can be taken and turned into a new
context:
@deftypefun gcry_error_t gcry_md_copy (gcry_md_hd_t *@var{handle_dst}, gcry_md_hd_t @var{handle_src})
Create a new digest object as an exact copy of the object described by
handle @var{handle_src} and store it in @var{handle_dst}. The context
is not reset and you can continue to hash data using this context and
independently using the original context.
@end deftypefun
Now that we have prepared everything to calculate hashes, it is time to
see how it is actually done. There are two ways for this, one to
update the hash with a block of memory and one macro to update the hash
by just one character. Both methods can be used on the same hash context.
@deftypefun void gcry_md_write (gcry_md_hd_t @var{h}, const void *@var{buffer}, size_t @var{length})
Pass @var{length} bytes of the data in @var{buffer} to the digest object
with handle @var{h} to update the digest values. This
function should be used for large blocks of data. If this function is
used after the context has been finalized, it will keep on pushing
the data through the algorithm specific transform function and change
the context; however the results are not meaningful and this feature
is only available to mitigate timing attacks.
@end deftypefun
@deftypefun void gcry_md_putc (gcry_md_hd_t @var{h}, int @var{c})
Pass the byte in @var{c} to the digest object with handle @var{h} to
update the digest value. This is an efficient function, implemented as
a macro to buffer the data before an actual update.
@end deftypefun
The semantics of the hash functions do not provide for reading out intermediate
message digests because the calculation must be finalized first. This
finalization may for example include the number of bytes hashed in the
message digest or some padding.
@deftypefun void gcry_md_final (gcry_md_hd_t @var{h})
Finalize the message digest calculation. This is not really needed
because @code{gcry_md_read} and @code{gcry_md_extract} do this implicitly.
After this has been done no further updates (by means of @code{gcry_md_write}
or @code{gcry_md_putc} should be done; However, to mitigate timing
attacks it is sometimes useful to keep on updating the context after
having stored away the actual digest. Only the first call to this function
has an effect. It is implemented as a macro.
@end deftypefun
The way to read out the calculated message digest is by using the
function:
@deftypefun {unsigned char *} gcry_md_read (gcry_md_hd_t @var{h}, int @var{algo})
@code{gcry_md_read} returns the message digest after finalizing the
calculation. This function may be used as often as required but it will
always return the same value for one handle. The returned message digest
is allocated within the message context and therefore valid until the
handle is released or reset-ed (using @code{gcry_md_close} or
@code{gcry_md_reset} or it has been updated as a mitigation measure
against timing attacks. @var{algo} may be given as 0 to return the only
enabled message digest or it may specify one of the enabled algorithms.
The function does return @code{NULL} if the requested algorithm has not
been enabled.
@end deftypefun
The way to read output of extendable-output function is by using the
function:
@deftypefun gpg_err_code_t gcry_md_extract (gcry_md_hd_t @var{h}, @
int @var{algo}, void *@var{buffer}, size_t @var{length})
@code{gcry_mac_read} returns output from extendable-output function.
This function may be used as often as required to generate more output
byte stream from the algorithm. Function extracts the new output bytes
to @var{buffer} of the length @var{length}. Buffer will be fully
populated with new output. @var{algo} may be given as 0 to return the only
enabled message digest or it may specify one of the enabled algorithms.
The function does return non-zero value if the requested algorithm has not
been enabled.
@end deftypefun
Because it is often necessary to get the message digest of blocks of
memory, two fast convenience function are available for this task:
@deftypefun gpg_err_code_t gcry_md_hash_buffers ( @
@w{int @var{algo}}, @w{unsigned int @var{flags}}, @
@w{void *@var{digest}}, @
@w{const gcry_buffer_t *@var{iov}}, @w{int @var{iovcnt}} )
@code{gcry_md_hash_buffers} is a shortcut function to calculate a
message digest from several buffers. This function does not require a
context and immediately returns the message digest of the data
described by @var{iov} and @var{iovcnt}. @var{digest} must be
allocated by the caller, large enough to hold the message digest
yielded by the the specified algorithm @var{algo}. This required size
may be obtained by using the function @code{gcry_md_get_algo_dlen}.
@var{iov} is an array of buffer descriptions with @var{iovcnt} items.
The caller should zero out the structures in this array and for each
array item set the fields @code{.data} to the address of the data to
be hashed, @code{.len} to number of bytes to be hashed. If @var{.off}
is also set, the data is taken starting at @var{.off} bytes from the
begin of the buffer. The field @code{.size} is not used.
The only supported flag value for @var{flags} is
@var{GCRY_MD_FLAG_HMAC} which turns this function into a HMAC
function; the first item in @var{iov} is then used as the key.
On success the function returns 0 and stores the resulting hash or MAC
at @var{digest}.
@end deftypefun
@deftypefun void gcry_md_hash_buffer (int @var{algo}, void *@var{digest}, const void *@var{buffer}, size_t @var{length});
@code{gcry_md_hash_buffer} is a shortcut function to calculate a message
digest of a buffer. This function does not require a context and
immediately returns the message digest of the @var{length} bytes at
@var{buffer}. @var{digest} must be allocated by the caller, large
enough to hold the message digest yielded by the the specified algorithm
@var{algo}. This required size may be obtained by using the function
@code{gcry_md_get_algo_dlen}.
Note that in contrast to @code{gcry_md_hash_buffers} this function
will abort the process if an unavailable algorithm is used.
@end deftypefun
@c ***********************************
@c ***** MD info functions ***********
@c ***********************************
Hash algorithms are identified by internal algorithm numbers (see
@code{gcry_md_open} for a list). However, in most applications they are
used by names, so two functions are available to map between string
representations and hash algorithm identifiers.
@deftypefun {const char *} gcry_md_algo_name (int @var{algo})
Map the digest algorithm id @var{algo} to a string representation of the
algorithm name. For unknown algorithms this function returns the
string @code{"?"}. This function should not be used to test for the
availability of an algorithm.
@end deftypefun
@deftypefun int gcry_md_map_name (const char *@var{name})
Map the algorithm with @var{name} to a digest algorithm identifier.
Returns 0 if the algorithm name is not known. Names representing
@acronym{ASN.1} object identifiers are recognized if the @acronym{IETF}
dotted format is used and the OID is prefixed with either "@code{oid.}"
or "@code{OID.}". For a list of supported OIDs, see the source code at
@file{cipher/md.c}. This function should not be used to test for the
availability of an algorithm.
@end deftypefun
@deftypefun gcry_error_t gcry_md_get_asnoid (int @var{algo}, void *@var{buffer}, size_t *@var{length})
Return an DER encoded ASN.1 OID for the algorithm @var{algo} in the
user allocated @var{buffer}. @var{length} must point to variable with
the available size of @var{buffer} and receives after return the
actual size of the returned OID. The returned error code may be
@code{GPG_ERR_TOO_SHORT} if the provided buffer is to short to receive
the OID; it is possible to call the function with @code{NULL} for
@var{buffer} to have it only return the required size. The function
returns 0 on success.
@end deftypefun
To test whether an algorithm is actually available for use, the
following macro should be used:
@deftypefun gcry_error_t gcry_md_test_algo (int @var{algo})
The macro returns 0 if the algorithm @var{algo} is available for use.
@end deftypefun
If the length of a message digest is not known, it can be retrieved
using the following function:
@deftypefun {unsigned int} gcry_md_get_algo_dlen (int @var{algo})
Retrieve the length in bytes of the digest yielded by algorithm
@var{algo}. This is often used prior to @code{gcry_md_read} to allocate
sufficient memory for the digest.
@end deftypefun
In some situations it might be hard to remember the algorithm used for
the ongoing hashing. The following function might be used to get that
information:
@deftypefun int gcry_md_get_algo (gcry_md_hd_t @var{h})
Retrieve the algorithm used with the handle @var{h}. Note that this
does not work reliable if more than one algorithm is enabled in @var{h}.
@end deftypefun
The following macro might also be useful:
@deftypefun int gcry_md_is_secure (gcry_md_hd_t @var{h})
This function returns true when the digest object @var{h} is allocated
in "secure memory"; i.e. @var{h} was created with the
@code{GCRY_MD_FLAG_SECURE}.
@end deftypefun
@deftypefun int gcry_md_is_enabled (gcry_md_hd_t @var{h}, int @var{algo})
This function returns true when the algorithm @var{algo} has been
enabled for the digest object @var{h}.
@end deftypefun
Tracking bugs related to hashing is often a cumbersome task which
requires to add a lot of printf statements into the code.
Libgcrypt provides an easy way to avoid this. The actual data
hashed can be written to files on request.
@deftypefun void gcry_md_debug (gcry_md_hd_t @var{h}, const char *@var{suffix})
Enable debugging for the digest object with handle @var{h}. This
creates files named @file{dbgmd-<n>.<string>} while doing the
actual hashing. @var{suffix} is the string part in the filename. The
number is a counter incremented for each new hashing. The data in the
file is the raw data as passed to @code{gcry_md_write} or
@code{gcry_md_putc}. If @code{NULL} is used for @var{suffix}, the
debugging is stopped and the file closed. This is only rarely required
because @code{gcry_md_close} implicitly stops debugging.
@end deftypefun
@c **********************************************************
@c ******************* MAC Functions **********************
@c **********************************************************
@node Message Authentication Codes
@chapter Message Authentication Codes
Libgcrypt provides an easy and consistent to use interface for generating
Message Authentication Codes (MAC). MAC generation is buffered and interface
similar to the one used with hash algorithms. The programming model follows
an open/process/close paradigm and is in that similar to other building blocks
provided by Libgcrypt.
@menu
* Available MAC algorithms:: List of MAC algorithms supported by the library.
* Working with MAC algorithms:: List of functions related to MAC algorithms.
@end menu
@node Available MAC algorithms
@section Available MAC algorithms
@c begin table of MAC algorithms
@cindex HMAC-SHA-1
@cindex HMAC-SHA-224, HMAC-SHA-256, HMAC-SHA-384, HMAC-SHA-512
@cindex HMAC-SHA3-224, HMAC-SHA3-256, HMAC-SHA3-384, HMAC-SHA3-512
@cindex HMAC-RIPE-MD-160
@cindex HMAC-MD2, HMAC-MD4, HMAC-MD5
@cindex HMAC-TIGER1
@cindex HMAC-Whirlpool
@cindex HMAC-Stribog-256, HMAC-Stribog-512
@cindex HMAC-GOSTR-3411-94
@table @code
@item GCRY_MAC_NONE
This is not a real algorithm but used by some functions as an error
return value. This constant is guaranteed to have the value @code{0}.
@item GCRY_MAC_HMAC_SHA256
This is keyed-hash message authentication code (HMAC) message authentication
algorithm based on the SHA-256 hash algorithm.
@item GCRY_MAC_HMAC_SHA224
This is HMAC message authentication algorithm based on the SHA-224 hash
algorithm.
@item GCRY_MAC_HMAC_SHA512
This is HMAC message authentication algorithm based on the SHA-512 hash
algorithm.
@item GCRY_MAC_HMAC_SHA384
This is HMAC message authentication algorithm based on the SHA-384 hash
algorithm.
@item GCRY_MAC_HMAC_SHA3_256
This is HMAC message authentication algorithm based on the SHA3-384 hash
algorithm.
@item GCRY_MAC_HMAC_SHA3_224
This is HMAC message authentication algorithm based on the SHA3-224 hash
algorithm.
@item GCRY_MAC_HMAC_SHA3_512
This is HMAC message authentication algorithm based on the SHA3-512 hash
algorithm.
@item GCRY_MAC_HMAC_SHA3_384
This is HMAC message authentication algorithm based on the SHA3-384 hash
algorithm.
@item GCRY_MAC_HMAC_SHA1
This is HMAC message authentication algorithm based on the SHA-1 hash
algorithm.
@item GCRY_MAC_HMAC_MD5
This is HMAC message authentication algorithm based on the MD5 hash
algorithm.
@item GCRY_MAC_HMAC_MD4
This is HMAC message authentication algorithm based on the MD4 hash
algorithm.
@item GCRY_MAC_HMAC_RMD160
This is HMAC message authentication algorithm based on the RIPE-MD-160 hash
algorithm.
@item GCRY_MAC_HMAC_WHIRLPOOL
This is HMAC message authentication algorithm based on the WHIRLPOOL hash
algorithm.
@item GCRY_MAC_HMAC_GOSTR3411_94
This is HMAC message authentication algorithm based on the GOST R 34.11-94 hash
algorithm.
@item GCRY_MAC_HMAC_STRIBOG256
This is HMAC message authentication algorithm based on the 256-bit hash
algorithm described in GOST R 34.11-2012.
@item GCRY_MAC_HMAC_STRIBOG512
This is HMAC message authentication algorithm based on the 512-bit hash
algorithm described in GOST R 34.11-2012.
@item GCRY_MAC_CMAC_AES
This is CMAC (Cipher-based MAC) message authentication algorithm based on
the AES block cipher algorithm.
@item GCRY_MAC_CMAC_3DES
This is CMAC message authentication algorithm based on the three-key EDE
Triple-DES block cipher algorithm.
@item GCRY_MAC_CMAC_CAMELLIA
This is CMAC message authentication algorithm based on the Camellia block cipher
algorithm.
@item GCRY_MAC_CMAC_CAST5
This is CMAC message authentication algorithm based on the CAST128-5
block cipher algorithm.
@item GCRY_MAC_CMAC_BLOWFISH
This is CMAC message authentication algorithm based on the Blowfish
block cipher algorithm.
@item GCRY_MAC_CMAC_TWOFISH
This is CMAC message authentication algorithm based on the Twofish
block cipher algorithm.
@item GCRY_MAC_CMAC_SERPENT
This is CMAC message authentication algorithm based on the Serpent
block cipher algorithm.
@item GCRY_MAC_CMAC_SEED
This is CMAC message authentication algorithm based on the SEED
block cipher algorithm.
@item GCRY_MAC_CMAC_RFC2268
This is CMAC message authentication algorithm based on the Ron's Cipher 2
block cipher algorithm.
@item GCRY_MAC_CMAC_IDEA
This is CMAC message authentication algorithm based on the IDEA
block cipher algorithm.
@item GCRY_MAC_CMAC_GOST28147
This is CMAC message authentication algorithm based on the GOST 28147-89
block cipher algorithm.
@item GCRY_MAC_GMAC_AES
This is GMAC (GCM mode based MAC) message authentication algorithm based on
the AES block cipher algorithm.
@item GCRY_MAC_GMAC_CAMELLIA
This is GMAC message authentication algorithm based on the Camellia
block cipher algorithm.
@item GCRY_MAC_GMAC_TWOFISH
This is GMAC message authentication algorithm based on the Twofish
block cipher algorithm.
@item GCRY_MAC_GMAC_SERPENT
This is GMAC message authentication algorithm based on the Serpent
block cipher algorithm.
@item GCRY_MAC_GMAC_SEED
This is GMAC message authentication algorithm based on the SEED
block cipher algorithm.
@item GCRY_MAC_POLY1305
This is plain Poly1305 message authentication algorithm, used with
one-time key.
@item GCRY_MAC_POLY1305_AES
This is Poly1305-AES message authentication algorithm, used with
key and one-time nonce.
@item GCRY_MAC_POLY1305_CAMELLIA
This is Poly1305-Camellia message authentication algorithm, used with
key and one-time nonce.
@item GCRY_MAC_POLY1305_TWOFISH
This is Poly1305-Twofish message authentication algorithm, used with
key and one-time nonce.
@item GCRY_MAC_POLY1305_SERPENT
This is Poly1305-Serpent message authentication algorithm, used with
key and one-time nonce.
@item GCRY_MAC_POLY1305_SEED
This is Poly1305-SEED message authentication algorithm, used with
key and one-time nonce.
@end table
@c end table of MAC algorithms
@node Working with MAC algorithms
@section Working with MAC algorithms
To use most of these function it is necessary to create a context;
this is done using:
@deftypefun gcry_error_t gcry_mac_open (gcry_mac_hd_t *@var{hd}, int @var{algo}, unsigned int @var{flags}, gcry_ctx_t @var{ctx})
Create a MAC object for algorithm @var{algo}. @var{flags} may be given as an
bitwise OR of constants described below. @var{hd} is guaranteed to either
receive a valid handle or NULL. @var{ctx} is context object to associate MAC
object with. @var{ctx} maybe set to NULL.
For a list of supported algorithms, see @xref{Available MAC algorithms}.
The flags allowed for @var{mode} are:
@c begin table of MAC flags
@table @code
@item GCRY_MAC_FLAG_SECURE
Allocate all buffers and the resulting MAC in "secure memory". Use this if the
MAC data is highly confidential.
@end table
@c begin table of MAC flags
@end deftypefun
@c end function gcry_mac_open
In order to use a handle for performing MAC algorithm operations, a
`key' has to be set first:
@deftypefun gcry_error_t gcry_mac_setkey (gcry_mac_hd_t @var{h}, const void *@var{key}, size_t @var{keylen})
Set the MAC key to the value of @var{key} of length @var{keylen} bytes. With
HMAC algorithms, there is no restriction on the length of the key. With CMAC
algorithms, the length of the key is restricted to those supported by the
underlying block cipher.
@end deftypefun
GMAC algorithms and Poly1305-with-cipher algorithms need initialization vector to be set,
which can be performed with function:
@deftypefun gcry_error_t gcry_mac_setiv (gcry_mac_hd_t @var{h}, const void *@var{iv}, size_t @var{ivlen})
Set the IV to the value of @var{iv} of length @var{ivlen} bytes.
@end deftypefun
After you are done with the MAC calculation, you should release the resources
by using:
@deftypefun void gcry_mac_close (gcry_mac_hd_t @var{h})
Release all resources of MAC context @var{h}. @var{h} should not be
used after a call to this function. A @code{NULL} passed as @var{h} is
ignored. The function also clears all sensitive information associated
with this handle.
@end deftypefun
Often you have to do several MAC operations using the same algorithm.
To avoid the overhead of creating and releasing context, a reset function
is provided:
@deftypefun gcry_error_t gcry_mac_reset (gcry_mac_hd_t @var{h})
Reset the current context to its initial state. This is effectively identical
to a close followed by an open and setting same key.
Note that gcry_mac_reset is implemented as a macro.
@end deftypefun
Now that we have prepared everything to calculate MAC, it is time to
see how it is actually done.
@deftypefun gcry_error_t gcry_mac_write (gcry_mac_hd_t @var{h}, const void *@var{buffer}, size_t @var{length})
Pass @var{length} bytes of the data in @var{buffer} to the MAC object
with handle @var{h} to update the MAC values. If this function is
used after the context has been finalized, it will keep on pushing the
data through the algorithm specific transform function and thereby
change the context; however the results are not meaningful and this
feature is only available to mitigate timing attacks.
@end deftypefun
The way to read out the calculated MAC is by using the function:
@deftypefun gcry_error_t gcry_mac_read (gcry_mac_hd_t @var{h}, void *@var{buffer}, size_t *@var{length})
@code{gcry_mac_read} returns the MAC after finalizing the calculation.
Function copies the resulting MAC value to @var{buffer} of the length
@var{length}. If @var{length} is larger than length of resulting MAC value,
then length of MAC is returned through @var{length}.
@end deftypefun
To compare existing MAC value with recalculated MAC, one is to use the function:
@deftypefun gcry_error_t gcry_mac_verify (gcry_mac_hd_t @var{h}, void *@var{buffer}, size_t @var{length})
@code{gcry_mac_verify} finalizes MAC calculation and compares result with
@var{length} bytes of data in @var{buffer}. Error code @code{GPG_ERR_CHECKSUM}
is returned if the MAC value in the buffer @var{buffer} does not match
the MAC calculated in object @var{h}.
@end deftypefun
In some situations it might be hard to remember the algorithm used for
the MAC calculation. The following function might be used to get that
information:
@deftypefun {int} gcry_mac_get_algo (gcry_mac_hd_t @var{h})
Retrieve the algorithm used with the handle @var{h}.
@end deftypefun
@c ***********************************
@c ***** MAC info functions **********
@c ***********************************
MAC algorithms are identified by internal algorithm numbers (see
@code{gcry_mac_open} for a list). However, in most applications they are
used by names, so two functions are available to map between string
representations and MAC algorithm identifiers.
@deftypefun {const char *} gcry_mac_algo_name (int @var{algo})
Map the MAC algorithm id @var{algo} to a string representation of the
algorithm name. For unknown algorithms this function returns the
string @code{"?"}. This function should not be used to test for the
availability of an algorithm.
@end deftypefun
@deftypefun int gcry_mac_map_name (const char *@var{name})
Map the algorithm with @var{name} to a MAC algorithm identifier.
Returns 0 if the algorithm name is not known. This function should not
be used to test for the availability of an algorithm.
@end deftypefun
To test whether an algorithm is actually available for use, the
following macro should be used:
@deftypefun gcry_error_t gcry_mac_test_algo (int @var{algo})
The macro returns 0 if the MAC algorithm @var{algo} is available for use.
@end deftypefun
If the length of a message digest is not known, it can be retrieved
using the following function:
@deftypefun {unsigned int} gcry_mac_get_algo_maclen (int @var{algo})
Retrieve the length in bytes of the MAC yielded by algorithm @var{algo}.
This is often used prior to @code{gcry_mac_read} to allocate sufficient memory
for the MAC value. On error @code{0} is returned.
@end deftypefun
@deftypefun {unsigned int} gcry_mac_get_algo_keylen (@var{algo})
This function returns length of the key for MAC algorithm @var{algo}. If
the algorithm supports multiple key lengths, the default supported key
length is returned. On error @code{0} is returned. The key length is
returned as number of octets.
@end deftypefun
@c *******************************************************
@c ******************* KDF *****************************
@c *******************************************************
@node Key Derivation
@chapter Key Derivation
@acronym{Libgcypt} provides a general purpose function to derive keys
from strings.
@deftypefun gpg_error_t gcry_kdf_derive ( @
@w{const void *@var{passphrase}}, @w{size_t @var{passphraselen}}, @
@w{int @var{algo}}, @w{int @var{subalgo}}, @
@w{const void *@var{salt}}, @w{size_t @var{saltlen}}, @
@w{unsigned long @var{iterations}}, @
@w{size_t @var{keysize}}, @w{void *@var{keybuffer}} )
Derive a key from a passphrase. @var{keysize} gives the requested
size of the keys in octets. @var{keybuffer} is a caller provided
buffer filled on success with the derived key. The input passphrase
is taken from @var{passphrase} which is an arbitrary memory buffer of
@var{passphraselen} octets. @var{algo} specifies the KDF algorithm to
use; see below. @var{subalgo} specifies an algorithm used internally
by the KDF algorithms; this is usually a hash algorithm but certain
KDF algorithms may use it differently. @var{salt} is a salt of length
@var{saltlen} octets, as needed by most KDF algorithms.
@var{iterations} is a positive integer parameter to most KDFs.
@noindent
On success 0 is returned; on failure an error code.
@noindent
Currently supported KDFs (parameter @var{algo}):
@table @code
@item GCRY_KDF_SIMPLE_S2K
The OpenPGP simple S2K algorithm (cf. RFC4880). Its use is strongly
deprecated. @var{salt} and @var{iterations} are not needed and may be
passed as @code{NULL}/@code{0}.
@item GCRY_KDF_SALTED_S2K
The OpenPGP salted S2K algorithm (cf. RFC4880). Usually not used.
@var{iterations} is not needed and may be passed as @code{0}. @var{saltlen}
must be given as 8.
@item GCRY_KDF_ITERSALTED_S2K
The OpenPGP iterated+salted S2K algorithm (cf. RFC4880). This is the
default for most OpenPGP applications. @var{saltlen} must be given as
8. Note that OpenPGP defines a special encoding of the
@var{iterations}; however this function takes the plain decoded
iteration count.
@item GCRY_KDF_PBKDF2
The PKCS#5 Passphrase Based Key Derivation Function number 2.
@item GCRY_KDF_SCRYPT
The SCRYPT Key Derivation Function. The subalgorithm is used to specify
the CPU/memory cost parameter N, and the number of iterations
is used for the parallelization parameter p. The block size is fixed
at 8 in the current implementation.
@end table
@end deftypefun
@c **********************************************************
@c ******************* Random *****************************
@c **********************************************************
@node Random Numbers
@chapter Random Numbers
@menu
* Quality of random numbers:: Libgcrypt uses different quality levels.
* Retrieving random numbers:: How to retrieve random numbers.
@end menu
@node Quality of random numbers
@section Quality of random numbers
@acronym{Libgcypt} offers random numbers of different quality levels:
@deftp {Data type} gcry_random_level_t
The constants for the random quality levels are of this enum type.
@end deftp
@table @code
@item GCRY_WEAK_RANDOM
For all functions, except for @code{gcry_mpi_randomize}, this level maps
to GCRY_STRONG_RANDOM. If you do not want this, consider using
@code{gcry_create_nonce}.
@item GCRY_STRONG_RANDOM
Use this level for session keys and similar purposes.
@item GCRY_VERY_STRONG_RANDOM
Use this level for long term key material.
@end table
@node Retrieving random numbers
@section Retrieving random numbers
@deftypefun void gcry_randomize (unsigned char *@var{buffer}, size_t @var{length}, enum gcry_random_level @var{level})
Fill @var{buffer} with @var{length} random bytes using a random quality
as defined by @var{level}.
@end deftypefun
@deftypefun {void *} gcry_random_bytes (size_t @var{nbytes}, enum gcry_random_level @var{level})
Convenience function to allocate a memory block consisting of
@var{nbytes} fresh random bytes using a random quality as defined by
@var{level}.
@end deftypefun
@deftypefun {void *} gcry_random_bytes_secure (size_t @var{nbytes}, enum gcry_random_level @var{level})
Convenience function to allocate a memory block consisting of
@var{nbytes} fresh random bytes using a random quality as defined by
@var{level}. This function differs from @code{gcry_random_bytes} in
that the returned buffer is allocated in a ``secure'' area of the
memory.
@end deftypefun
@deftypefun void gcry_create_nonce (unsigned char *@var{buffer}, size_t @var{length})
Fill @var{buffer} with @var{length} unpredictable bytes. This is
commonly called a nonce and may also be used for initialization
vectors and padding. This is an extra function nearly independent of
the other random function for 3 reasons: It better protects the
regular random generator's internal state, provides better performance
and does not drain the precious entropy pool.
@end deftypefun
@c **********************************************************
@c ******************* S-Expressions ***********************
@c **********************************************************
@node S-expressions
@chapter S-expressions
S-expressions are used by the public key functions to pass complex data
structures around. These LISP like objects are used by some
cryptographic protocols (cf. RFC-2692) and Libgcrypt provides functions
to parse and construct them. For detailed information, see
@cite{Ron Rivest, code and description of S-expressions,
@uref{http://theory.lcs.mit.edu/~rivest/sexp.html}}.
@menu
* Data types for S-expressions:: Data types related with S-expressions.
* Working with S-expressions:: How to work with S-expressions.
@end menu
@node Data types for S-expressions
@section Data types for S-expressions
@deftp {Data type} gcry_sexp_t
The @code{gcry_sexp_t} type describes an object with the Libgcrypt internal
representation of an S-expression.
@end deftp
@node Working with S-expressions
@section Working with S-expressions
@noindent
There are several functions to create an Libgcrypt S-expression object
from its external representation or from a string template. There is
also a function to convert the internal representation back into one of
the external formats:
@deftypefun gcry_error_t gcry_sexp_new (@w{gcry_sexp_t *@var{r_sexp}}, @w{const void *@var{buffer}}, @w{size_t @var{length}}, @w{int @var{autodetect}})
This is the generic function to create an new S-expression object from
its external representation in @var{buffer} of @var{length} bytes. On
success the result is stored at the address given by @var{r_sexp}.
With @var{autodetect} set to 0, the data in @var{buffer} is expected to
be in canonized format, with @var{autodetect} set to 1 the parses any of
the defined external formats. If @var{buffer} does not hold a valid
S-expression an error code is returned and @var{r_sexp} set to
@code{NULL}.
Note that the caller is responsible for releasing the newly allocated
S-expression using @code{gcry_sexp_release}.
@end deftypefun
@deftypefun gcry_error_t gcry_sexp_create (@w{gcry_sexp_t *@var{r_sexp}}, @w{void *@var{buffer}}, @w{size_t @var{length}}, @w{int @var{autodetect}}, @w{void (*@var{freefnc})(void*)})
This function is identical to @code{gcry_sexp_new} but has an extra
argument @var{freefnc}, which, when not set to @code{NULL}, is expected
to be a function to release the @var{buffer}; most likely the standard
@code{free} function is used for this argument. This has the effect of
transferring the ownership of @var{buffer} to the created object in
@var{r_sexp}. The advantage of using this function is that Libgcrypt
might decide to directly use the provided buffer and thus avoid extra
copying.
@end deftypefun
@deftypefun gcry_error_t gcry_sexp_sscan (@w{gcry_sexp_t *@var{r_sexp}}, @w{size_t *@var{erroff}}, @w{const char *@var{buffer}}, @w{size_t @var{length}})
This is another variant of the above functions. It behaves nearly
identical but provides an @var{erroff} argument which will receive the
offset into the buffer where the parsing stopped on error.
@end deftypefun
@deftypefun gcry_error_t gcry_sexp_build (@w{gcry_sexp_t *@var{r_sexp}}, @w{size_t *@var{erroff}}, @w{const char *@var{format}, ...})
This function creates an internal S-expression from the string template
@var{format} and stores it at the address of @var{r_sexp}. If there is a
parsing error, the function returns an appropriate error code and stores
the offset into @var{format} where the parsing stopped in @var{erroff}.
The function supports a couple of printf-like formatting characters and
expects arguments for some of these escape sequences right after
@var{format}. The following format characters are defined:
@table @samp
@item %m
The next argument is expected to be of type @code{gcry_mpi_t} and a copy of
its value is inserted into the resulting S-expression. The MPI is
stored as a signed integer.
@item %M
The next argument is expected to be of type @code{gcry_mpi_t} and a copy of
its value is inserted into the resulting S-expression. The MPI is
stored as an unsigned integer.
@item %s
The next argument is expected to be of type @code{char *} and that
string is inserted into the resulting S-expression.
@item %d
The next argument is expected to be of type @code{int} and its value is
inserted into the resulting S-expression.
@item %u
The next argument is expected to be of type @code{unsigned int} and
its value is inserted into the resulting S-expression.
@item %b
The next argument is expected to be of type @code{int} directly
followed by an argument of type @code{char *}. This represents a
buffer of given length to be inserted into the resulting S-expression.
@item %S
The next argument is expected to be of type @code{gcry_sexp_t} and a
copy of that S-expression is embedded in the resulting S-expression.
The argument needs to be a regular S-expression, starting with a
parenthesis.
@end table
@noindent
No other format characters are defined and would return an error. Note
that the format character @samp{%%} does not exists, because a percent
sign is not a valid character in an S-expression.
@end deftypefun
@deftypefun void gcry_sexp_release (@w{gcry_sexp_t @var{sexp}})
Release the S-expression object @var{sexp}. If the S-expression is
stored in secure memory it explicitly zeroises that memory; note that
this is done in addition to the zeroisation always done when freeing
secure memory.
@end deftypefun
@noindent
The next 2 functions are used to convert the internal representation
back into a regular external S-expression format and to show the
structure for debugging.
@deftypefun size_t gcry_sexp_sprint (@w{gcry_sexp_t @var{sexp}}, @w{int @var{mode}}, @w{char *@var{buffer}}, @w{size_t @var{maxlength}})
Copies the S-expression object @var{sexp} into @var{buffer} using the
format specified in @var{mode}. @var{maxlength} must be set to the
allocated length of @var{buffer}. The function returns the actual
length of valid bytes put into @var{buffer} or 0 if the provided buffer
is too short. Passing @code{NULL} for @var{buffer} returns the required
length for @var{buffer}. For convenience reasons an extra byte with
value 0 is appended to the buffer.
@noindent
The following formats are supported:
@table @code
@item GCRYSEXP_FMT_DEFAULT
Returns a convenient external S-expression representation.
@item GCRYSEXP_FMT_CANON
Return the S-expression in canonical format.
@item GCRYSEXP_FMT_BASE64
Not currently supported.
@item GCRYSEXP_FMT_ADVANCED
Returns the S-expression in advanced format.
@end table
@end deftypefun
@deftypefun void gcry_sexp_dump (@w{gcry_sexp_t @var{sexp}})
Dumps @var{sexp} in a format suitable for debugging to Libgcrypt's
logging stream.
@end deftypefun
@noindent
Often canonical encoding is used in the external representation. The
following function can be used to check for valid encoding and to learn
the length of the S-expression.
@deftypefun size_t gcry_sexp_canon_len (@w{const unsigned char *@var{buffer}}, @w{size_t @var{length}}, @w{size_t *@var{erroff}}, @w{int *@var{errcode}})
Scan the canonical encoded @var{buffer} with implicit length values and
return the actual length this S-expression uses. For a valid S-expression
it should never return 0. If @var{length} is not 0, the maximum
length to scan is given; this can be used for syntax checks of
data passed from outside. @var{errcode} and @var{erroff} may both be
passed as @code{NULL}.
@end deftypefun
@noindent
There are functions to parse S-expressions and retrieve elements:
@deftypefun gcry_sexp_t gcry_sexp_find_token (@w{const gcry_sexp_t @var{list}}, @w{const char *@var{token}}, @w{size_t @var{toklen}})
Scan the S-expression for a sublist with a type (the car of the list)
matching the string @var{token}. If @var{toklen} is not 0, the token is
assumed to be raw memory of this length. The function returns a newly
allocated S-expression consisting of the found sublist or @code{NULL}
when not found.
@end deftypefun
@deftypefun int gcry_sexp_length (@w{const gcry_sexp_t @var{list}})
Return the length of the @var{list}. For a valid S-expression this
should be at least 1.
@end deftypefun
@deftypefun gcry_sexp_t gcry_sexp_nth (@w{const gcry_sexp_t @var{list}}, @w{int @var{number}})
Create and return a new S-expression from the element with index @var{number} in
@var{list}. Note that the first element has the index 0. If there is
no such element, @code{NULL} is returned.
@end deftypefun
@deftypefun gcry_sexp_t gcry_sexp_car (@w{const gcry_sexp_t @var{list}})
Create and return a new S-expression from the first element in
@var{list}; this is called the "type" and should always exist per
S-expression specification and in general be a string. @code{NULL} is
returned in case of a problem.
@end deftypefun
@deftypefun gcry_sexp_t gcry_sexp_cdr (@w{const gcry_sexp_t @var{list}})
Create and return a new list form all elements except for the first one.
Note that this function may return an invalid S-expression because it
is not guaranteed, that the type exists and is a string. However, for
parsing a complex S-expression it might be useful for intermediate
lists. Returns @code{NULL} on error.
@end deftypefun
@deftypefun {const char *} gcry_sexp_nth_data (@w{const gcry_sexp_t @var{list}}, @w{int @var{number}}, @w{size_t *@var{datalen}})
This function is used to get data from a @var{list}. A pointer to the
actual data with index @var{number} is returned and the length of this
data will be stored to @var{datalen}. If there is no data at the given
index or the index represents another list, @code{NULL} is returned.
@strong{Caution:} The returned pointer is valid as long as @var{list} is
not modified or released.
@noindent
Here is an example on how to extract and print the surname (Meier) from
the S-expression @samp{(Name Otto Meier (address Burgplatz 3))}:
@example
size_t len;
const char *name;
name = gcry_sexp_nth_data (list, 2, &len);
printf ("my name is %.*s\n", (int)len, name);
@end example
@end deftypefun
@deftypefun {void *} gcry_sexp_nth_buffer (@w{const gcry_sexp_t @var{list}}, @w{int @var{number}}, @w{size_t *@var{rlength}})
This function is used to get data from a @var{list}. A malloced
buffer with the actual data at list index @var{number} is returned and
the length of this buffer will be stored to @var{rlength}. If there
is no data at the given index or the index represents another list,
@code{NULL} is returned. The caller must release the result using
@code{gcry_free}.
@noindent
Here is an example on how to extract and print the CRC value from the
S-expression @samp{(hash crc32 #23ed00d7)}:
@example
size_t len;
char *value;
value = gcry_sexp_nth_buffer (list, 2, &len);
if (value)
fwrite (value, len, 1, stdout);
gcry_free (value);
@end example
@end deftypefun
@deftypefun {char *} gcry_sexp_nth_string (@w{gcry_sexp_t @var{list}}, @w{int @var{number}})
This function is used to get and convert data from a @var{list}. The
data is assumed to be a Nul terminated string. The caller must
release this returned value using @code{gcry_free}. If there is
no data at the given index, the index represents a list or the value
can't be converted to a string, @code{NULL} is returned.
@end deftypefun
@deftypefun gcry_mpi_t gcry_sexp_nth_mpi (@w{gcry_sexp_t @var{list}}, @w{int @var{number}}, @w{int @var{mpifmt}})
This function is used to get and convert data from a @var{list}. This
data is assumed to be an MPI stored in the format described by
@var{mpifmt} and returned as a standard Libgcrypt MPI. The caller must
release this returned value using @code{gcry_mpi_release}. If there is
no data at the given index, the index represents a list or the value
can't be converted to an MPI, @code{NULL} is returned. If you use
this function to parse results of a public key function, you most
likely want to use @code{GCRYMPI_FMT_USG}.
@end deftypefun
@deftypefun gpg_error_t gcry_sexp_extract_param ( @
@w{gcry_sexp_t @var{sexp}}, @
@w{const char *@var{path}}, @
@w{const char *@var{list}}, ...)
Extract parameters from an S-expression using a list of parameter
names. The names of these parameters are specified in LIST. White
space between the parameter names are ignored. Some special characters
may be given to control the conversion:
@table @samp
@item +
Switch to unsigned integer format (GCRYMPI_FMT_USG). This is the
default mode.
@item -
Switch to standard signed format (GCRYMPI_FMT_STD).
@item /
Switch to opaque MPI format. The resulting MPIs may not be used for
computations; see @code{gcry_mpi_get_opaque} for details.
@item &
Switch to buffer descriptor mode. See below for details.
@item ?
If immediately following a parameter letter (no white space allowed),
that parameter is considered optional.
@end table
In general parameter names are single letters. To use a string for a
parameter name, enclose the name in single quotes.
Unless in buffer descriptor mode for each parameter name a pointer to
an @code{gcry_mpi_t} variable is expected that must be set to
@code{NULL} prior to invoking this function, and finally a @code{NULL}
is expected. For example
@example
_gcry_sexp_extract_param (key, NULL, "n/x+e d-'foo'",
&mpi_n, &mpi_x, &mpi_e, &mpi_foo, NULL)
@end example
stores the parameter 'n' from @var{key} as an unsigned MPI into
@var{mpi_n}, the parameter 'x' as an opaque MPI into @var{mpi_x}, the
parameter 'e' again as an unsigned MPI into @var{mpi_e}, and the
parameter 'foo' as a signed MPI.
@var{path} is an optional string used to locate a token. The
exclamation mark separated tokens are used via
@code{gcry_sexp_find_token} to find a start point inside the
S-expression.
In buffer descriptor mode a pointer to a @code{gcry_buffer_t}
descriptor is expected instead of a pointer to an MPI. The caller may
use two different operation modes here: If the @var{data} field of the
provided descriptor is @code{NULL}, the function allocates a new
buffer and stores it at @var{data}; the other fields are set
accordingly with @var{off} set to 0. If @var{data} is not
@code{NULL}, the function assumes that the @var{data}, @var{size}, and
@var{off} fields specify a buffer where to but the value of the
respective parameter; on return the @var{len} field receives the
number of bytes copied to that buffer; in case the buffer is too
small, the function immediately returns with an error code (and
@var{len} is set to 0).
The function returns 0 on success. On error an error code is
returned, all passed MPIs that might have been allocated up to this
point are deallocated and set to @code{NULL}, and all passed buffers
are either truncated if the caller supplied the buffer, or deallocated
if the function allocated the buffer.
@end deftypefun
@c **********************************************************
@c ******************* MPIs ******** ***********************
@c **********************************************************
@node MPI library
@chapter MPI library
@menu
* Data types:: MPI related data types.
* Basic functions:: First steps with MPI numbers.
* MPI formats:: External representation of MPIs.
* Calculations:: Performing MPI calculations.
* Comparisons:: How to compare MPI values.
* Bit manipulations:: How to access single bits of MPI values.
* EC functions:: Elliptic curve related functions.
* Miscellaneous:: Miscellaneous MPI functions.
@end menu
Public key cryptography is based on mathematics with large numbers. To
implement the public key functions, a library for handling these large
numbers is required. Because of the general usefulness of such a
library, its interface is exposed by Libgcrypt.
In the context of Libgcrypt and in most other applications, these large
numbers are called MPIs (multi-precision-integers).
@node Data types
@section Data types
@deftp {Data type} {gcry_mpi_t}
This type represents an object to hold an MPI.
@end deftp
@deftp {Data type} {gcry_mpi_point_t}
This type represents an object to hold a point for elliptic curve math.
@end deftp
@node Basic functions
@section Basic functions
@noindent
To work with MPIs, storage must be allocated and released for the
numbers. This can be done with one of these functions:
@deftypefun gcry_mpi_t gcry_mpi_new (@w{unsigned int @var{nbits}})
Allocate a new MPI object, initialize it to 0 and initially allocate
enough memory for a number of at least @var{nbits}. This pre-allocation is
only a small performance issue and not actually necessary because
Libgcrypt automatically re-allocates the required memory.
@end deftypefun
@deftypefun gcry_mpi_t gcry_mpi_snew (@w{unsigned int @var{nbits}})
This is identical to @code{gcry_mpi_new} but allocates the MPI in the so
called "secure memory" which in turn will take care that all derived
values will also be stored in this "secure memory". Use this for highly
confidential data like private key parameters.
@end deftypefun
@deftypefun gcry_mpi_t gcry_mpi_copy (@w{const gcry_mpi_t @var{a}})
Create a new MPI as the exact copy of @var{a} but with the constant
and immutable flags cleared.
@end deftypefun
@deftypefun void gcry_mpi_release (@w{gcry_mpi_t @var{a}})
Release the MPI @var{a} and free all associated resources. Passing
@code{NULL} is allowed and ignored. When a MPI stored in the "secure
memory" is released, that memory gets wiped out immediately.
@end deftypefun
@noindent
The simplest operations are used to assign a new value to an MPI:
@deftypefun gcry_mpi_t gcry_mpi_set (@w{gcry_mpi_t @var{w}}, @w{const gcry_mpi_t @var{u}})
Assign the value of @var{u} to @var{w} and return @var{w}. If
@code{NULL} is passed for @var{w}, a new MPI is allocated, set to the
value of @var{u} and returned.
@end deftypefun
@deftypefun gcry_mpi_t gcry_mpi_set_ui (@w{gcry_mpi_t @var{w}}, @w{unsigned long @var{u}})
Assign the value of @var{u} to @var{w} and return @var{w}. If
@code{NULL} is passed for @var{w}, a new MPI is allocated, set to the
value of @var{u} and returned. This function takes an @code{unsigned
int} as type for @var{u} and thus it is only possible to set @var{w} to
small values (usually up to the word size of the CPU).
@end deftypefun
@deftypefun void gcry_mpi_swap (@w{gcry_mpi_t @var{a}}, @w{gcry_mpi_t @var{b}})
Swap the values of @var{a} and @var{b}.
@end deftypefun
@deftypefun void gcry_mpi_snatch (@w{gcry_mpi_t @var{w}}, @
@w{const gcry_mpi_t @var{u}})
Set @var{u} into @var{w} and release @var{u}. If @var{w} is
@code{NULL} only @var{u} will be released.
@end deftypefun
@deftypefun void gcry_mpi_neg (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}})
Set the sign of @var{w} to the negative of @var{u}.
@end deftypefun
@deftypefun void gcry_mpi_abs (@w{gcry_mpi_t @var{w}})
Clear the sign of @var{w}.
@end deftypefun
@node MPI formats
@section MPI formats
@noindent
The following functions are used to convert between an external
representation of an MPI and the internal one of Libgcrypt.
@deftypefun gcry_error_t gcry_mpi_scan (@w{gcry_mpi_t *@var{r_mpi}}, @w{enum gcry_mpi_format @var{format}}, @w{const unsigned char *@var{buffer}}, @w{size_t @var{buflen}}, @w{size_t *@var{nscanned}})
Convert the external representation of an integer stored in @var{buffer}
with a length of @var{buflen} into a newly created MPI returned which
will be stored at the address of @var{r_mpi}. For certain formats the
length argument is not required and should be passed as @code{0}. After a
successful operation the variable @var{nscanned} receives the number of
bytes actually scanned unless @var{nscanned} was given as
@code{NULL}. @var{format} describes the format of the MPI as stored in
@var{buffer}:
@table @code
@item GCRYMPI_FMT_STD
2-complement stored without a length header. Note that
@code{gcry_mpi_print} stores a @code{0} as a string of zero length.
@item GCRYMPI_FMT_PGP
As used by OpenPGP (only defined as unsigned). This is basically
@code{GCRYMPI_FMT_STD} with a 2 byte big endian length header.
@item GCRYMPI_FMT_SSH
As used in the Secure Shell protocol. This is @code{GCRYMPI_FMT_STD}
with a 4 byte big endian header.
@item GCRYMPI_FMT_HEX
Stored as a string with each byte of the MPI encoded as 2 hex digits.
Negative numbers are prefix with a minus sign and in addition the
high bit is always zero to make clear that an explicit sign ist used.
When using this format, @var{buflen} must be zero.
@item GCRYMPI_FMT_USG
Simple unsigned integer.
@end table
@noindent
Note that all of the above formats store the integer in big-endian
format (MSB first).
@end deftypefun
@deftypefun gcry_error_t gcry_mpi_print (@w{enum gcry_mpi_format @var{format}}, @w{unsigned char *@var{buffer}}, @w{size_t @var{buflen}}, @w{size_t *@var{nwritten}}, @w{const gcry_mpi_t @var{a}})
Convert the MPI @var{a} into an external representation described by
@var{format} (see above) and store it in the provided @var{buffer}
which has a usable length of at least the @var{buflen} bytes. If
@var{nwritten} is not NULL, it will receive the number of bytes
actually stored in @var{buffer} after a successful operation.
@end deftypefun
@deftypefun gcry_error_t gcry_mpi_aprint (@w{enum gcry_mpi_format @var{format}}, @w{unsigned char **@var{buffer}}, @w{size_t *@var{nbytes}}, @w{const gcry_mpi_t @var{a}})
Convert the MPI @var{a} into an external representation described by
@var{format} (see above) and store it in a newly allocated buffer which
address will be stored in the variable @var{buffer} points to. The
number of bytes stored in this buffer will be stored in the variable
@var{nbytes} points to, unless @var{nbytes} is @code{NULL}.
Even if @var{nbytes} is zero, the function allocates at least one byte
and store a zero there. Thus with formats @code{GCRYMPI_FMT_STD} and
@code{GCRYMPI_FMT_USG} the caller may safely set a returned length of
0 to 1 to represent a zero as a 1 byte string.
@end deftypefun
@deftypefun void gcry_mpi_dump (@w{const gcry_mpi_t @var{a}})
Dump the value of @var{a} in a format suitable for debugging to
Libgcrypt's logging stream. Note that one leading space but no trailing
space or linefeed will be printed. It is okay to pass @code{NULL} for
@var{a}.
@end deftypefun
@node Calculations
@section Calculations
@noindent
Basic arithmetic operations:
@deftypefun void gcry_mpi_add (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}})
@math{@var{w} = @var{u} + @var{v}}.
@end deftypefun
@deftypefun void gcry_mpi_add_ui (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{unsigned long @var{v}})
@math{@var{w} = @var{u} + @var{v}}. Note that @var{v} is an unsigned integer.
@end deftypefun
@deftypefun void gcry_mpi_addm (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}}, @w{gcry_mpi_t @var{m}})
@math{@var{w} = @var{u} + @var{v} \bmod @var{m}}.
@end deftypefun
@deftypefun void gcry_mpi_sub (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}})
@math{@var{w} = @var{u} - @var{v}}.
@end deftypefun
@deftypefun void gcry_mpi_sub_ui (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{unsigned long @var{v}})
@math{@var{w} = @var{u} - @var{v}}. @var{v} is an unsigned integer.
@end deftypefun
@deftypefun void gcry_mpi_subm (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}}, @w{gcry_mpi_t @var{m}})
@math{@var{w} = @var{u} - @var{v} \bmod @var{m}}.
@end deftypefun
@deftypefun void gcry_mpi_mul (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}})
@math{@var{w} = @var{u} * @var{v}}.
@end deftypefun
@deftypefun void gcry_mpi_mul_ui (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{unsigned long @var{v}})
@math{@var{w} = @var{u} * @var{v}}. @var{v} is an unsigned integer.
@end deftypefun
@deftypefun void gcry_mpi_mulm (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{gcry_mpi_t @var{v}}, @w{gcry_mpi_t @var{m}})
@math{@var{w} = @var{u} * @var{v} \bmod @var{m}}.
@end deftypefun
@deftypefun void gcry_mpi_mul_2exp (@w{gcry_mpi_t @var{w}}, @w{gcry_mpi_t @var{u}}, @w{unsigned long @var{e}})
@c FIXME: I am in need for a real TeX{info} guru:
@c I don't know why TeX can grok @var{e} here.
@math{@var{w} = @var{u} * 2^e}.
@end deftypefun
@deftypefun void gcry_mpi_div (@w{gcry_mpi_t @var{q}}, @w{gcry_mpi_t @var{r}}, @w{gcry_mpi_t @var{dividend}}, @w{gcry_mpi_t @var{divisor}}, @w{int @var{round}})
@math{@var{q} = @var{dividend} / @var{divisor}}, @math{@var{r} =
@var{dividend} \bmod @var{divisor}}. @var{q} and @var{r} may be passed
as @code{NULL}. @var{round} should be negative or 0.
@end deftypefun
@deftypefun void gcry_mpi_mod (@w{gcry_mpi_t @var{r}}, @w{gcry_mpi_t @var{dividend}}, @w{gcry_mpi_t @var{divisor}})
@math{@var{r} = @var{dividend} \bmod @var{divisor}}.
@end deftypefun
@deftypefun void gcry_mpi_powm (@w{gcry_mpi_t @var{w}}, @w{const gcry_mpi_t @var{b}}, @w{const gcry_mpi_t @var{e}}, @w{const gcry_mpi_t @var{m}})
@c I don't know why TeX can grok @var{e} here.
@math{@var{w} = @var{b}^e \bmod @var{m}}.
@end deftypefun
@deftypefun int gcry_mpi_gcd (@w{gcry_mpi_t @var{g}}, @w{gcry_mpi_t @var{a}}, @w{gcry_mpi_t @var{b}})
Set @var{g} to the greatest common divisor of @var{a} and @var{b}.
Return true if the @var{g} is 1.
@end deftypefun
@deftypefun int gcry_mpi_invm (@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{a}}, @w{gcry_mpi_t @var{m}})
Set @var{x} to the multiplicative inverse of @math{@var{a} \bmod @var{m}}.
Return true if the inverse exists.
@end deftypefun
@node Comparisons
@section Comparisons
@noindent
The next 2 functions are used to compare MPIs:
@deftypefun int gcry_mpi_cmp (@w{const gcry_mpi_t @var{u}}, @w{const gcry_mpi_t @var{v}})
Compare the multi-precision-integers number @var{u} and @var{v}
returning 0 for equality, a positive value for @var{u} > @var{v} and a
negative for @var{u} < @var{v}. If both numbers are opaque values
(cf, gcry_mpi_set_opaque) the comparison is done by checking the bit
sizes using memcmp. If only one number is an opaque value, the opaque
value is less than the other number.
@end deftypefun
@deftypefun int gcry_mpi_cmp_ui (@w{const gcry_mpi_t @var{u}}, @w{unsigned long @var{v}})
Compare the multi-precision-integers number @var{u} with the unsigned
integer @var{v} returning 0 for equality, a positive value for @var{u} >
@var{v} and a negative for @var{u} < @var{v}.
@end deftypefun
@deftypefun int gcry_mpi_is_neg (@w{const gcry_mpi_t @var{a}})
Return 1 if @var{a} is less than zero; return 0 if zero or positive.
@end deftypefun
@node Bit manipulations
@section Bit manipulations
@noindent
There are a couple of functions to get information on arbitrary bits
in an MPI and to set or clear them:
@deftypefun {unsigned int} gcry_mpi_get_nbits (@w{gcry_mpi_t @var{a}})
Return the number of bits required to represent @var{a}.
@end deftypefun
@deftypefun int gcry_mpi_test_bit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})
Return true if bit number @var{n} (counting from 0) is set in @var{a}.
@end deftypefun
@deftypefun void gcry_mpi_set_bit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})
Set bit number @var{n} in @var{a}.
@end deftypefun
@deftypefun void gcry_mpi_clear_bit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})
Clear bit number @var{n} in @var{a}.
@end deftypefun
@deftypefun void gcry_mpi_set_highbit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})
Set bit number @var{n} in @var{a} and clear all bits greater than @var{n}.
@end deftypefun
@deftypefun void gcry_mpi_clear_highbit (@w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})
Clear bit number @var{n} in @var{a} and all bits greater than @var{n}.
@end deftypefun
@deftypefun void gcry_mpi_rshift (@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})
Shift the value of @var{a} by @var{n} bits to the right and store the
result in @var{x}.
@end deftypefun
@deftypefun void gcry_mpi_lshift (@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{a}}, @w{unsigned int @var{n}})
Shift the value of @var{a} by @var{n} bits to the left and store the
result in @var{x}.
@end deftypefun
@node EC functions
@section EC functions
@noindent
Libgcrypt provides an API to access low level functions used by its
elliptic curve implementation. These functions allow to implement
elliptic curve methods for which no explicit support is available.
@deftypefun gcry_mpi_point_t gcry_mpi_point_new (@w{unsigned int @var{nbits}})
Allocate a new point object, initialize it to 0, and allocate enough
memory for a points of at least @var{nbits}. This pre-allocation
yields only a small performance win and is not really necessary
because Libgcrypt automatically re-allocates the required memory.
Using 0 for @var{nbits} is usually the right thing to do.
@end deftypefun
@deftypefun void gcry_mpi_point_release (@w{gcry_mpi_point_t @var{point}})
Release @var{point} and free all associated resources. Passing
@code{NULL} is allowed and ignored.
@end deftypefun
@deftypefun void gcry_mpi_point_get (@w{gcry_mpi_t @var{x}}, @
@w{gcry_mpi_t @var{y}}, @w{gcry_mpi_t @var{z}}, @
@w{gcry_mpi_point_t @var{point}})
Store the projective coordinates from @var{point} into the MPIs
@var{x}, @var{y}, and @var{z}. If a coordinate is not required,
@code{NULL} may be used for @var{x}, @var{y}, or @var{z}.
@end deftypefun
@deftypefun void gcry_mpi_point_snatch_get (@w{gcry_mpi_t @var{x}}, @
@w{gcry_mpi_t @var{y}}, @w{gcry_mpi_t @var{z}}, @
@w{gcry_mpi_point_t @var{point}})
Store the projective coordinates from @var{point} into the MPIs
@var{x}, @var{y}, and @var{z}. If a coordinate is not required,
@code{NULL} may be used for @var{x}, @var{y}, or @var{z}. The object
@var{point} is then released. Using this function instead of
@code{gcry_mpi_point_get} and @code{gcry_mpi_point_release} has the
advantage of avoiding some extra memory allocations and copies.
@end deftypefun
@deftypefun gcry_mpi_point_t gcry_mpi_point_set ( @
@w{gcry_mpi_point_t @var{point}}, @
@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{y}}, @w{gcry_mpi_t @var{z}})
Store the projective coordinates from @var{x}, @var{y}, and @var{z}
into @var{point}. If a coordinate is given as @code{NULL}, the value
0 is used. If @code{NULL} is used for @var{point} a new point object
is allocated and returned. Returns @var{point} or the newly allocated
point object.
@end deftypefun
@deftypefun gcry_mpi_point_t gcry_mpi_point_snatch_set ( @
@w{gcry_mpi_point_t @var{point}}, @
@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{y}}, @w{gcry_mpi_t @var{z}})
Store the projective coordinates from @var{x}, @var{y}, and @var{z}
into @var{point}. If a coordinate is given as @code{NULL}, the value
0 is used. If @code{NULL} is used for @var{point} a new point object
is allocated and returned. The MPIs @var{x}, @var{y}, and @var{z} are
released. Using this function instead of @code{gcry_mpi_point_set}
and 3 calls to @code{gcry_mpi_release} has the advantage of avoiding
some extra memory allocations and copies. Returns @var{point} or the
newly allocated point object.
@end deftypefun
@anchor{gcry_mpi_ec_new}
@deftypefun gpg_error_t gcry_mpi_ec_new (@w{gcry_ctx_t *@var{r_ctx}}, @
@w{gcry_sexp_t @var{keyparam}}, @w{const char *@var{curvename}})
Allocate a new context for elliptic curve operations. If
@var{keyparam} is given it specifies the parameters of the curve
(@pxref{ecc_keyparam}). If @var{curvename} is given in addition to
@var{keyparam} and the key parameters do not include a named curve
reference, the string @var{curvename} is used to fill in missing
parameters. If only @var{curvename} is given, the context is
initialized for this named curve.
If a parameter specifying a point (e.g. @code{g} or @code{q}) is not
found, the parser looks for a non-encoded point by appending
@code{.x}, @code{.y}, and @code{.z} to the parameter name and looking
them all up to create a point. A parameter with the suffix @code{.z}
is optional and defaults to 1.
On success the function returns 0 and stores the new context object at
@var{r_ctx}; this object eventually needs to be released
(@pxref{gcry_ctx_release}). On error the function stores @code{NULL} at
@var{r_ctx} and returns an error code.
@end deftypefun
@deftypefun gcry_mpi_t gcry_mpi_ec_get_mpi ( @
@w{const char *@var{name}}, @w{gcry_ctx_t @var{ctx}}, @w{int @var{copy}})
Return the MPI with @var{name} from the context @var{ctx}. If not
found @code{NULL} is returned. If the returned MPI may later be
modified, it is suggested to pass @code{1} to @var{copy}, so that the
function guarantees that a modifiable copy of the MPI is returned. If
@code{0} is used for @var{copy}, this function may return a constant
flagged MPI. In any case @code{gcry_mpi_release} needs to be called
to release the result. For valid names @ref{ecc_keyparam}. If the
public key @code{q} is requested but only the private key @code{d} is
available, @code{q} will be recomputed on the fly. If a point
parameter is requested it is returned as an uncompressed
encoded point unless these special names are used:
@table @var
@item q@@eddsa
Return an EdDSA style compressed point. This is only supported for
Twisted Edwards curves.
@end table
@end deftypefun
@deftypefun gcry_mpi_point_t gcry_mpi_ec_get_point ( @
@w{const char *@var{name}}, @w{gcry_ctx_t @var{ctx}}, @w{int @var{copy}})
Return the point with @var{name} from the context @var{ctx}. If not
found @code{NULL} is returned. If the returned MPI may later be
modified, it is suggested to pass @code{1} to @var{copy}, so that the
function guarantees that a modifiable copy of the MPI is returned. If
@code{0} is used for @var{copy}, this function may return a constant
flagged point. In any case @code{gcry_mpi_point_release} needs to be
called to release the result. If the public key @code{q} is requested
but only the private key @code{d} is available, @code{q} will be
recomputed on the fly.
@end deftypefun
@deftypefun gpg_error_t gcry_mpi_ec_set_mpi ( @
@w{const char *@var{name}}, @w{gcry_mpi_t @var{newvalue}}, @
@w{gcry_ctx_t @var{ctx}})
Store the MPI @var{newvalue} at @var{name} into the context @var{ctx}.
On success @code{0} is returned; on error an error code. Valid names
are the MPI parameters of an elliptic curve (@pxref{ecc_keyparam}).
@end deftypefun
@deftypefun gpg_error_t gcry_mpi_ec_set_point ( @
@w{const char *@var{name}}, @w{gcry_mpi_point_t @var{newvalue}}, @
@w{gcry_ctx_t @var{ctx}})
Store the point @var{newvalue} at @var{name} into the context
@var{ctx}. On success @code{0} is returned; on error an error code.
Valid names are the point parameters of an elliptic curve
(@pxref{ecc_keyparam}).
@end deftypefun
@deftypefun gpg_err_code_t gcry_mpi_ec_decode_point ( @
@w{mpi_point_t @var{result}}, @w{gcry_mpi_t @var{value}}, @
@w{gcry_ctx_t @var{ctx}})
Decode the point given as an MPI in @var{value} and store at
@var{result}. To decide which encoding is used the function takes a
context @var{ctx} which can be created with @code{gcry_mpi_ec_new}.
If @code{NULL} is given for the context the function assumes a 0x04
prefixed uncompressed encoding. On error an error code is returned
and @var{result} might be changed.
@end deftypefun
@deftypefun int gcry_mpi_ec_get_affine ( @
@w{gcry_mpi_t @var{x}}, @w{gcry_mpi_t @var{y}}, @
@w{gcry_mpi_point_t @var{point}}, @w{gcry_ctx_t @var{ctx}})
Compute the affine coordinates from the projective coordinates in
@var{point} and store them into @var{x} and @var{y}. If one
coordinate is not required, @code{NULL} may be passed to @var{x} or
@var{y}. @var{ctx} is the context object which has been created using
@code{gcry_mpi_ec_new}. Returns 0 on success or not 0 if @var{point}
is at infinity.
Note that you can use @code{gcry_mpi_ec_set_point} with the value
@code{GCRYMPI_CONST_ONE} for @var{z} to convert affine coordinates
back into projective coordinates.
@end deftypefun
@deftypefun void gcry_mpi_ec_dup ( @
@w{gcry_mpi_point_t @var{w}}, @w{gcry_mpi_point_t @var{u}}, @
@w{gcry_ctx_t @var{ctx}})
Double the point @var{u} of the elliptic curve described by @var{ctx}
and store the result into @var{w}.
@end deftypefun
@deftypefun void gcry_mpi_ec_add ( @
@w{gcry_mpi_point_t @var{w}}, @w{gcry_mpi_point_t @var{u}}, @
@w{gcry_mpi_point_t @var{v}}, @w{gcry_ctx_t @var{ctx}})
Add the points @var{u} and @var{v} of the elliptic curve described by
@var{ctx} and store the result into @var{w}.
@end deftypefun
@deftypefun void gcry_mpi_ec_sub ( @
@w{gcry_mpi_point_t @var{w}}, @w{gcry_mpi_point_t @var{u}}, @
@w{gcry_mpi_point_t @var{v}}, @w{gcry_ctx_t @var{ctx}})
Subtracts the point @var{v} from the point @var{u} of the elliptic
curve described by @var{ctx} and store the result into @var{w}. Only
Twisted Edwards curves are supported for now.
@end deftypefun
@deftypefun void gcry_mpi_ec_mul ( @
@w{gcry_mpi_point_t @var{w}}, @w{gcry_mpi_t @var{n}}, @
@w{gcry_mpi_point_t @var{u}}, @w{gcry_ctx_t @var{ctx}})
Multiply the point @var{u} of the elliptic curve described by
@var{ctx} by @var{n} and store the result into @var{w}.
@end deftypefun
@deftypefun int gcry_mpi_ec_curve_point ( @
@w{gcry_mpi_point_t @var{point}}, @w{gcry_ctx_t @var{ctx}})
Return true if @var{point} is on the elliptic curve described by
@var{ctx}.
@end deftypefun
@node Miscellaneous
@section Miscellaneous
An MPI data type is allowed to be ``misused'' to store an arbitrary
value. Two functions implement this kludge:
@deftypefun gcry_mpi_t gcry_mpi_set_opaque (@w{gcry_mpi_t @var{a}}, @w{void *@var{p}}, @w{unsigned int @var{nbits}})
Store @var{nbits} of the value @var{p} points to in @var{a} and mark
@var{a} as an opaque value (i.e. an value that can't be used for any
math calculation and is only used to store an arbitrary bit pattern in
@var{a}). Ownership of @var{p} is taken by this function and thus the
user may not use dereference the passed value anymore. It is required
that them memory referenced by @var{p} has been allocated in a way
that @code{gcry_free} is able to release it.
WARNING: Never use an opaque MPI for actual math operations. The only
valid functions are gcry_mpi_get_opaque and gcry_mpi_release. Use
gcry_mpi_scan to convert a string of arbitrary bytes into an MPI.
@end deftypefun
@deftypefun gcry_mpi_t gcry_mpi_set_opaque_copy (@w{gcry_mpi_t @var{a}}, @w{const void *@var{p}}, @w{unsigned int @var{nbits}})
Same as @code{gcry_mpi_set_opaque} but ownership of @var{p} is not
taken instead a copy of @var{p} is used.
@end deftypefun
@deftypefun {void *} gcry_mpi_get_opaque (@w{gcry_mpi_t @var{a}}, @w{unsigned int *@var{nbits}})
Return a pointer to an opaque value stored in @var{a} and return its
size in @var{nbits}. Note that the returned pointer is still owned by
@var{a} and that the function should never be used for an non-opaque
MPI.
@end deftypefun
Each MPI has an associated set of flags for special purposes. The
currently defined flags are:
@table @code
@item GCRYMPI_FLAG_SECURE
Setting this flag converts @var{a} into an MPI stored in "secure
memory". Clearing this flag is not allowed.
@item GCRYMPI_FLAG_OPAQUE
This is an interanl flag, indicating the an opaque valuue and not an
integer is stored. This is an read-only flag; it may not be set or
cleared.
@item GCRYMPI_FLAG_IMMUTABLE
If this flag is set, the MPI is marked as immutable. Setting or
changing the value of that MPI is ignored and an error message is
logged. The flag is sometimes useful for debugging.
@item GCRYMPI_FLAG_CONST
If this flag is set, the MPI is marked as a constant and as immutable
Setting or changing the value of that MPI is ignored and an error
message is logged. Such an MPI will never be deallocated and may thus
be used without copying. Note that using gcry_mpi_copy will return a
copy of that constant with this and the immutable flag cleared. A few
commonly used constants are pre-defined and accessible using the
macros @code{GCRYMPI_CONST_ONE}, @code{GCRYMPI_CONST_TWO},
@code{GCRYMPI_CONST_THREE}, @code{GCRYMPI_CONST_FOUR}, and
@code{GCRYMPI_CONST_EIGHT}.
@item GCRYMPI_FLAG_USER1
@itemx GCRYMPI_FLAG_USER2
@itemx GCRYMPI_FLAG_USER3
@itemx GCRYMPI_FLAG_USER4
These flags are reserved for use by the application.
@end table
@deftypefun void gcry_mpi_set_flag (@w{gcry_mpi_t @var{a}}, @
@w{enum gcry_mpi_flag @var{flag}})
Set the @var{flag} for the MPI @var{a}. The only allowed flags are
@code{GCRYMPI_FLAG_SECURE}, @code{GCRYMPI_FLAG_IMMUTABLE}, and
@code{GCRYMPI_FLAG_CONST}.
@end deftypefun
@deftypefun void gcry_mpi_clear_flag (@w{gcry_mpi_t @var{a}}, @
@w{enum gcry_mpi_flag @var{flag}})
Clear @var{flag} for the multi-precision-integers @var{a}. The only
allowed flag is @code{GCRYMPI_FLAG_IMMUTABLE} but only if
@code{GCRYMPI_FLAG_CONST} is not set. If @code{GCRYMPI_FLAG_CONST} is
set, clearing @code{GCRYMPI_FLAG_IMMUTABLE} will simply be ignored.
@end deftypefun
o
@deftypefun int gcry_mpi_get_flag (@w{gcry_mpi_t @var{a}}, @
@w{enum gcry_mpi_flag @var{flag}})
Return true if @var{flag} is set for @var{a}.
@end deftypefun
To put a random value into an MPI, the following convenience function
may be used:
@deftypefun void gcry_mpi_randomize (@w{gcry_mpi_t @var{w}}, @w{unsigned int @var{nbits}}, @w{enum gcry_random_level @var{level}})
Set the multi-precision-integers @var{w} to a random non-negative number of
@var{nbits}, using random data quality of level @var{level}. In case
@var{nbits} is not a multiple of a byte, @var{nbits} is rounded up to
the next byte boundary. When using a @var{level} of
@code{GCRY_WEAK_RANDOM} this function makes use of
@code{gcry_create_nonce}.
@end deftypefun
@c **********************************************************
@c ******************** Prime numbers ***********************
@c **********************************************************
@node Prime numbers
@chapter Prime numbers
@menu
* Generation:: Generation of new prime numbers.
* Checking:: Checking if a given number is prime.
@end menu
@node Generation
@section Generation
@deftypefun gcry_error_t gcry_prime_generate (gcry_mpi_t *@var{prime},unsigned int @var{prime_bits}, unsigned int @var{factor_bits}, gcry_mpi_t **@var{factors}, gcry_prime_check_func_t @var{cb_func}, void *@var{cb_arg}, gcry_random_level_t @var{random_level}, unsigned int @var{flags})
Generate a new prime number of @var{prime_bits} bits and store it in
@var{prime}. If @var{factor_bits} is non-zero, one of the prime factors
of (@var{prime} - 1) / 2 must be @var{factor_bits} bits long. If
@var{factors} is non-zero, allocate a new, @code{NULL}-terminated array
holding the prime factors and store it in @var{factors}. @var{flags}
might be used to influence the prime number generation process.
@end deftypefun
@deftypefun gcry_error_t gcry_prime_group_generator (gcry_mpi_t *@var{r_g}, gcry_mpi_t @var{prime}, gcry_mpi_t *@var{factors}, gcry_mpi_t @var{start_g})
Find a generator for @var{prime} where the factorization of
(@var{prime}-1) is in the @code{NULL} terminated array @var{factors}.
Return the generator as a newly allocated MPI in @var{r_g}. If
@var{start_g} is not NULL, use this as the start for the search.
@end deftypefun
@deftypefun void gcry_prime_release_factors (gcry_mpi_t *@var{factors})
Convenience function to release the @var{factors} array.
@end deftypefun
@node Checking
@section Checking
@deftypefun gcry_error_t gcry_prime_check (gcry_mpi_t @var{p}, unsigned int @var{flags})
Check whether the number @var{p} is prime. Returns zero in case @var{p}
is indeed a prime, returns @code{GPG_ERR_NO_PRIME} in case @var{p} is
not a prime and a different error code in case something went horribly
wrong.
@end deftypefun
@c **********************************************************
@c ******************** Utilities ***************************
@c **********************************************************
@node Utilities
@chapter Utilities
@menu
* Memory allocation:: Functions related with memory allocation.
* Context management:: Functions related with context management.
* Buffer description:: A data type to describe buffers.
@end menu
@node Memory allocation
@section Memory allocation
@deftypefun {void *} gcry_malloc (size_t @var{n})
This function tries to allocate @var{n} bytes of memory. On success
it returns a pointer to the memory area, in an out-of-core condition,
it returns NULL.
@end deftypefun
@deftypefun {void *} gcry_malloc_secure (size_t @var{n})
Like @code{gcry_malloc}, but uses secure memory.
@end deftypefun
@deftypefun {void *} gcry_calloc (size_t @var{n}, size_t @var{m})
This function allocates a cleared block of memory (i.e. initialized with
zero bytes) long enough to contain a vector of @var{n} elements, each of
size @var{m} bytes. On success it returns a pointer to the memory
block; in an out-of-core condition, it returns NULL.
@end deftypefun
@deftypefun {void *} gcry_calloc_secure (size_t @var{n}, size_t @var{m})
Like @code{gcry_calloc}, but uses secure memory.
@end deftypefun
@deftypefun {void *} gcry_realloc (void *@var{p}, size_t @var{n})
This function tries to resize the memory area pointed to by @var{p} to
@var{n} bytes. On success it returns a pointer to the new memory
area, in an out-of-core condition, it returns NULL. Depending on
whether the memory pointed to by @var{p} is secure memory or not,
gcry_realloc tries to use secure memory as well.
@end deftypefun
@deftypefun void gcry_free (void *@var{p})
Release the memory area pointed to by @var{p}.
@end deftypefun
@node Context management
@section Context management
Some function make use of a context object. As of now there are only
a few math functions. However, future versions of Libgcrypt may make
more use of this context object.
@deftp {Data type} {gcry_ctx_t}
This type is used to refer to the general purpose context object.
@end deftp
@anchor{gcry_ctx_release}
@deftypefun void gcry_ctx_release (gcry_ctx_t @var{ctx})
Release the context object @var{ctx} and all associated resources. A
@code{NULL} passed as @var{ctx} is ignored.
@end deftypefun
@node Buffer description
@section Buffer description
To help hashing non-contiguous areas of memory a general purpose data
type is defined:
@deftp {Data type} {gcry_buffer_t}
This type is a structure to describe a buffer. The user should make
sure that this structure is initialized to zero. The available fields
of this structure are:
@table @code
@item .size
This is either 0 for no information available or indicates the
allocated length of the buffer.
@item .off
This is the offset into the buffer.
@item .len
This is the valid length of the buffer starting at @code{.off}.
@item .data
This is the address of the buffer.
@end table
@end deftp
@c **********************************************************
@c ********************* Tools ****************************
@c **********************************************************
@node Tools
@chapter Tools
@menu
* hmac256:: A standalone HMAC-SHA-256 implementation
@end menu
@manpage hmac256.1
@node hmac256
@section A HMAC-SHA-256 tool
@ifset manverb
.B hmac256
\- Compute an HMAC-SHA-256 MAC
@end ifset
@mansect synopsis
@ifset manverb
.B hmac256
.RB [ \-\-binary ]
.I key
.I [FILENAME]
@end ifset
@mansect description
This is a standalone HMAC-SHA-256 implementation used to compute an
HMAC-SHA-256 message authentication code. The tool has originally
been developed as a second implementation for Libgcrypt to allow
comparing against the primary implementation and to be used for
internal consistency checks. It should not be used for sensitive data
because no mechanisms to clear the stack etc are used.
The code has been written in a highly portable manner and requires
only a few standard definitions to be provided in a config.h file.
@noindent
@command{hmac256} is commonly invoked as
@example
hmac256 "This is my key" foo.txt
@end example
@noindent
This compute the MAC on the file @file{foo.txt} using the key given on
the command line.
@mansect options
@noindent
@command{hmac256} understands these options:
@table @gnupgtabopt
@item --binary
Print the MAC as a binary string. The default is to print the MAC
encoded has lower case hex digits.
@item --version
Print version of the program and exit.
@end table
@mansect see also
@ifset isman
@command{sha256sum}(1)
@end ifset
@manpause
@c **********************************************************
@c ***************** Architecure Overview *****************
@c **********************************************************
@node Architecture
@chapter Architecture
This chapter describes the internal architecture of Libgcrypt.
Libgcrypt is a function library written in ISO C-90. Any compliant
compiler should be able to build Libgcrypt as long as the target is
either a POSIX platform or compatible to the API used by Windows NT.
Provisions have been take so that the library can be directly used from
C++ applications; however building with a C++ compiler is not supported.
Building Libgcrypt is done by using the common @code{./configure && make}
approach. The configure command is included in the source distribution
and as a portable shell script it works on any Unix-alike system. The
result of running the configure script are a C header file
(@file{config.h}), customized Makefiles, the setup of symbolic links and
a few other things. After that the make tool builds and optionally
installs the library and the documentation. See the files
@file{INSTALL} and @file{README} in the source distribution on how to do
this.
Libgcrypt is developed using a Subversion@footnote{A version control
system available for many platforms} repository. Although all released
versions are tagged in this repository, they should not be used to build
production versions of Libgcrypt. Instead released tarballs should be
used. These tarballs are available from several places with the master
copy at @indicateurl{ftp://ftp.gnupg.org/gcrypt/libgcrypt/}.
Announcements of new releases are posted to the
@indicateurl{gnupg-announce@@gnupg.org} mailing list@footnote{See
@url{http://www.gnupg.org/documentation/mailing-lists.en.html} for
details.}.
@float Figure,fig:subsystems
@caption{Libgcrypt subsystems}
@center @image{libgcrypt-modules, 150mm,,Libgcrypt subsystems}
@end float
Libgcrypt consists of several subsystems (@pxref{fig:subsystems}) and
all these subsystems provide a public API; this includes the helper
subsystems like the one for S-expressions. The API style depends on the
subsystem; in general an open-use-close approach is implemented. The
open returns a handle to a context used for all further operations on
this handle, several functions may then be used on this handle and a
final close function releases all resources associated with the handle.
@menu
* Public-Key Subsystem Architecture:: About public keys.
* Symmetric Encryption Subsystem Architecture:: About standard ciphers.
* Hashing and MACing Subsystem Architecture:: About hashing.
* Multi-Precision-Integer Subsystem Architecture:: About big integers.
* Prime-Number-Generator Subsystem Architecture:: About prime numbers.
* Random-Number Subsystem Architecture:: About random stuff.
@c * Helper Subsystems Architecture:: About other stuff.
@end menu
@node Public-Key Subsystem Architecture
@section Public-Key Architecture
Because public key cryptography is almost always used to process small
amounts of data (hash values or session keys), the interface is not
implemented using the open-use-close paradigm, but with single
self-contained functions. Due to the wide variety of parameters
required by different algorithms S-expressions, as flexible way to
convey these parameters, are used. There is a set of helper functions
to work with these S-expressions.
@c see @xref{S-expression Subsystem Architecture}.
Aside of functions to register new algorithms, map algorithms names to
algorithms identifiers and to lookup properties of a key, the
following main functions are available:
@table @code
@item gcry_pk_encrypt
Encrypt data using a public key.
@item gcry_pk_decrypt
Decrypt data using a private key.
@item gcry_pk_sign
Sign data using a private key.
@item gcry_pk_verify
Verify that a signature matches the data.
@item gcry_pk_testkey
Perform a consistency over a public or private key.
@item gcry_pk_genkey
Create a new public/private key pair.
@end table
All these functions
lookup the module implementing the algorithm and pass the actual work
to that module. The parsing of the S-expression input and the
construction of S-expression for the return values is done by the high
level code (@file{cipher/pubkey.c}). Thus the internal interface
between the algorithm modules and the high level functions passes data
in a custom format.
By default Libgcrypt uses a blinding technique for RSA decryption to
mitigate real world timing attacks over a network: Instead of using
the RSA decryption directly, a blinded value @math{y = x r^{e} \bmod n}
is decrypted and the unblinded value @math{x' = y' r^{-1} \bmod n}
returned. The blinding value @math{r} is a random value with the size
of the modulus @math{n} and generated with @code{GCRY_WEAK_RANDOM}
random level.
@cindex X9.31
@cindex FIPS 186
The algorithm used for RSA and DSA key generation depends on whether
Libgcrypt is operated in standard or in FIPS mode. In standard mode
an algorithm based on the Lim-Lee prime number generator is used. In
FIPS mode RSA keys are generated as specified in ANSI X9.31 (1998) and
DSA keys as specified in FIPS 186-2.
@node Symmetric Encryption Subsystem Architecture
@section Symmetric Encryption Subsystem Architecture
The interface to work with symmetric encryption algorithms is made up
of functions from the @code{gcry_cipher_} name space. The
implementation follows the open-use-close paradigm and uses registered
algorithm modules for the actual work. Unless a module implements
optimized cipher mode implementations, the high level code
(@file{cipher/cipher.c}) implements the modes and calls the core
algorithm functions to process each block.
The most important functions are:
@table @code
@item gcry_cipher_open
Create a new instance to encrypt or decrypt using a specified
algorithm and mode.
@item gcry_cipher_close
Release an instance.
@item gcry_cipher_setkey
Set a key to be used for encryption or decryption.
@item gcry_cipher_setiv
Set an initialization vector to be used for encryption or decryption.
@item gcry_cipher_encrypt
@itemx gcry_cipher_decrypt
Encrypt or decrypt data. These functions may be called with arbitrary
amounts of data and as often as needed to encrypt or decrypt all data.
@end table
There are also functions to query properties of algorithms or context,
like block length, key length, map names or to enable features like
padding methods.
@node Hashing and MACing Subsystem Architecture
@section Hashing and MACing Subsystem Architecture
The interface to work with message digests and CRC algorithms is made
up of functions from the @code{gcry_md_} name space. The
implementation follows the open-use-close paradigm and uses registered
algorithm modules for the actual work. Although CRC algorithms are
not considered cryptographic hash algorithms, they share enough
properties so that it makes sense to handle them in the same way.
It is possible to use several algorithms at once with one context and
thus compute them all on the same data.
The most important functions are:
@table @code
@item gcry_md_open
Create a new message digest instance and optionally enable one
algorithm. A flag may be used to turn the message digest algorithm
into a HMAC algorithm.
@item gcry_md_enable
Enable an additional algorithm for the instance.
@item gcry_md_setkey
Set the key for the MAC.
@item gcry_md_write
Pass more data for computing the message digest to an instance.
@item gcry_md_putc
Buffered version of @code{gcry_md_write} implemented as a macro.
@item gcry_md_read
Finalize the computation of the message digest or HMAC and return the
result.
@item gcry_md_close
Release an instance
@item gcry_md_hash_buffer
Convenience function to directly compute a message digest over a
memory buffer without the need to create an instance first.
@end table
There are also functions to query properties of algorithms or the
instance, like enabled algorithms, digest length, map algorithm names.
it is also possible to reset an instance or to copy the current state
of an instance at any time. Debug functions to write the hashed data
to files are available as well.
@node Multi-Precision-Integer Subsystem Architecture
@section Multi-Precision-Integer Subsystem Architecture
The implementation of Libgcrypt's big integer computation code is
based on an old release of GNU Multi-Precision Library (GMP). The
decision not to use the GMP library directly was due to stalled
development at that time and due to security requirements which could
not be provided by the code in GMP. As GMP does, Libgcrypt provides
high performance assembler implementations of low level code for
several CPUS to gain much better performance than with a generic C
implementation.
@noindent
Major features of Libgcrypt's multi-precision-integer code compared to
GMP are:
@itemize
@item
Avoidance of stack based allocations to allow protection against
swapping out of sensitive data and for easy zeroing of sensitive
intermediate results.
@item
Optional use of secure memory and tracking of its use so that results
are also put into secure memory.
@item
MPIs are identified by a handle (implemented as a pointer) to give
better control over allocations and to augment them with extra
properties like opaque data.
@item
Removal of unnecessary code to reduce complexity.
@item
Functions specialized for public key cryptography.
@end itemize
@node Prime-Number-Generator Subsystem Architecture
@section Prime-Number-Generator Subsystem Architecture
Libgcrypt provides an interface to its prime number generator. These
functions make use of the internal prime number generator which is
required for the generation for public key key pairs. The plain prime
checking function is exported as well.
The generation of random prime numbers is based on the Lim and Lee
algorithm to create practically save primes.@footnote{Chae Hoon Lim
and Pil Joong Lee. A key recovery attack on discrete log-based schemes
using a prime order subgroup. In Burton S. Kaliski Jr., editor,
Advances in Cryptology: Crypto '97, pages 249­-263, Berlin /
Heidelberg / New York, 1997. Springer-Verlag. Described on page 260.}
This algorithm creates a pool of smaller primes, select a few of them
to create candidate primes of the form @math{2 * p_0 * p_1 * ... * p_n
+ 1}, tests the candidate for primality and permutates the pool until
a prime has been found. It is possible to clamp one of the small
primes to a certain size to help DSA style algorithms. Because most
of the small primes in the pool are not used for the resulting prime
number, they are saved for later use (see @code{save_pool_prime} and
@code{get_pool_prime} in @file{cipher/primegen.c}). The prime
generator optionally supports the finding of an appropriate generator.
@noindent
The primality test works in three steps:
@enumerate
@item
The standard sieve algorithm using the primes up to 4999 is used as a
quick first check.
@item
A Fermat test filters out almost all non-primes.
@item
A 5 round Rabin-Miller test is finally used. The first round uses a
witness of 2, whereas the next rounds use a random witness.
@end enumerate
To support the generation of RSA and DSA keys in FIPS mode according
to X9.31 and FIPS 186-2, Libgcrypt implements two additional prime
generation functions: @code{_gcry_derive_x931_prime} and
@code{_gcry_generate_fips186_2_prime}. These functions are internal
and not available through the public API.
@node Random-Number Subsystem Architecture
@section Random-Number Subsystem Architecture
Libgcrypt provides 3 levels or random quality: The level
@code{GCRY_VERY_STRONG_RANDOM} usually used for key generation, the
level @code{GCRY_STRONG_RANDOM} for all other strong random
requirements and the function @code{gcry_create_nonce} which is used
for weaker usages like nonces. There is also a level
@code{GCRY_WEAK_RANDOM} which in general maps to
@code{GCRY_STRONG_RANDOM} except when used with the function
@code{gcry_mpi_randomize}, where it randomizes an
multi-precision-integer using the @code{gcry_create_nonce} function.
@noindent
There are two distinct random generators available:
@itemize
@item
The Continuously Seeded Pseudo Random Number Generator (CSPRNG), which
is based on the classic GnuPG derived big pool implementation.
Implemented in @code{random/random-csprng.c} and used by default.
@item
A FIPS approved ANSI X9.31 PRNG using AES with a 128 bit key. Implemented in
@code{random/random-fips.c} and used if Libgcrypt is in FIPS mode.
@end itemize
@noindent
Both generators make use of so-called entropy gathering modules:
@table @asis
@item rndlinux
Uses the operating system provided
@file{/dev/random} and @file{/dev/urandom} devices.
@item rndunix
Runs several operating system commands to collect entropy from sources
like virtual machine and process statistics. It is a kind of
poor-man's @code{/dev/random} implementation. It is not available in
FIPS mode.
@item rndegd
Uses the operating system provided Entropy Gathering Daemon (EGD).
The EGD basically uses the same algorithms as rndunix does. However
as a system daemon it keeps on running and thus can serve several
processes requiring entropy input and does not waste collected entropy
if the application does not need all the collected entropy. It is not
available in FIPS mode.
@item rndw32
Targeted for the Microsoft Windows OS. It uses certain properties of
that system and is the only gathering module available for that OS.
@item rndhw
Extra module to collect additional entropy by utilizing a hardware
random number generator. As of now the supported hardware RNG is
the Padlock engine of VIA (Centaur) CPUs and x86 CPUs with the RDRAND
instruction. It is not available in FIPS mode.
@end table
@menu
* CSPRNG Description:: Description of the CSPRNG.
* FIPS PRNG Description:: Description of the FIPS X9.31 PRNG.
@end menu
@node CSPRNG Description
@subsection Description of the CSPRNG
This random number generator is loosely modelled after the one
described in Peter Gutmann's paper: "Software Generation of
Practically Strong Random Numbers".@footnote{Also described in chapter
6 of his book "Cryptographic Security Architecture", New York, 2004,
ISBN 0-387-95387-6.}
-A pool of 600 bytes is used and mixed using the core RIPE-MD160 hash
+A pool of 600 bytes is used and mixed using the core SHA-1 hash
transform function. Several extra features are used to make the
robust against a wide variety of attacks and to protect against
failures of subsystems. The state of the generator may be saved to a
file and initially seed form a file.
Depending on how Libgcrypt was build the generator is able to select
the best working entropy gathering module. It makes use of the slow
and fast collection methods and requires the pool to initially seeded
form the slow gatherer or a seed file. An entropy estimation is used
to mix in enough data from the gather modules before returning the
actual random output. Process fork detection and protection is
implemented.
@c FIXME: The design and implementaion needs a more verbose description.
The implementation of the nonce generator (for
@code{gcry_create_nonce}) is a straightforward repeated hash design: A
28 byte buffer is initially seeded with the PID and the time in
seconds in the first 20 bytes and with 8 bytes of random taken from
the @code{GCRY_STRONG_RANDOM} generator. Random numbers are then
created by hashing all the 28 bytes with SHA-1 and saving that again
in the first 20 bytes. The hash is also returned as result.
@node FIPS PRNG Description
@subsection Description of the FIPS X9.31 PRNG
The core of this deterministic random number generator is implemented
according to the document ``NIST-Recommended Random Number Generator
Based on ANSI X9.31 Appendix A.2.4 Using the 3-Key Triple DES and AES
Algorithms'', dated 2005-01-31. This implementation uses the AES
variant.
The generator is based on contexts to utilize the same core functions
for all random levels as required by the high-level interface. All
random generators return their data in 128 bit blocks. If the caller
requests less bits, the extra bits are not used. The key for each
generator is only set once at the first time a generator context is
used. The seed value is set along with the key and again after 1000
output blocks.
On Unix like systems the @code{GCRY_VERY_STRONG_RANDOM} and
@code{GCRY_STRONG_RANDOM} generators are keyed and seeded using the
rndlinux module with the @file{/dev/random} device. Thus these
generators may block until the OS kernel has collected enough entropy.
When used with Microsoft Windows the rndw32 module is used instead.
The generator used for @code{gcry_create_nonce} is keyed and seeded
from the @code{GCRY_STRONG_RANDOM} generator. Thus is may also block
if the @code{GCRY_STRONG_RANDOM} generator has not yet been used
before and thus gets initialized on the first use by
@code{gcry_create_nonce}. This special treatment is justified by the
weaker requirements for a nonce generator and to save precious kernel
entropy for use by the ``real'' random generators.
A self-test facility uses a separate context to check the
functionality of the core X9.31 functions using a known answers test.
During runtime each output block is compared to the previous one to
detect a stuck generator.
The DT value for the generator is made up of the current time down to
microseconds (if available) and a free running 64 bit counter. When
used with the test context the DT value is taken from the context and
incremented on each use.
@c @node Helper Subsystems Architecture
@c @section Helper Subsystems Architecture
@c
@c There are a few smaller subsystems which are mainly used internally by
@c Libgcrypt but also available to applications.
@c
@c @menu
@c * S-expression Subsystem Architecture:: Details about the S-expression architecture.
@c * Memory Subsystem Architecture:: Details about the memory allocation architecture.
@c * Miscellaneous Subsystems Architecture:: Details about other subsystems.
@c @end menu
@c
@c @node S-expression Subsystem Architecture
@c @subsection S-expression Subsystem Architecture
@c
@c Libgcrypt provides an interface to S-expression to create and parse
@c them. To use an S-expression with Libgcrypt it needs first be
@c converted into the internal representation used by Libgcrypt (the type
@c @code{gcry_sexp_t}). The conversion functions support a large subset
@c of the S-expression specification and further feature a printf like
@c function to convert a list of big integers or other binary data into
@c an S-expression.
@c
@c Libgcrypt currently implements S-expressions using a tagged linked
@c list. However this is not exposed to an application and may be
@c changed in future releases to reduce overhead when already working
@c with canonically encoded S-expressions. Secure memory is supported by
@c this S-expressions implementation.
@c
@c @node Memory Subsystem Architecture
@c @subsection Memory Subsystem Architecture
@c
@c TBD.
@c
@c
@c @node Miscellaneous Subsystems Architecture
@c @subsection Miscellaneous Subsystems Architecture
@c
@c TBD.
@c
@c
@c **********************************************************
@c ******************* Appendices *************************
@c **********************************************************
@c ********************************************
@node Self-Tests
@appendix Description of the Self-Tests
In addition to the build time regression test suite, Libgcrypt
implements self-tests to be performed at runtime. Which self-tests
are actually used depends on the mode Libgcrypt is used in. In
standard mode a limited set of self-tests is run at the time an
algorithm is first used. Note that not all algorithms feature a
self-test in standard mode. The @code{GCRYCTL_SELFTEST} control
command may be used to run all implemented self-tests at any time;
this will even run more tests than those run in FIPS mode.
If any of the self-tests fails, the library immediately returns an
error code to the caller. If Libgcrypt is in FIPS mode the self-tests
will be performed within the ``Self-Test'' state and any failure puts
the library into the ``Error'' state.
@c --------------------------------
@section Power-Up Tests
Power-up tests are only performed if Libgcrypt is in FIPS mode.
@subsection Symmetric Cipher Algorithm Power-Up Tests
The following symmetric encryption algorithm tests are run during
power-up:
@table @asis
@item 3DES
To test the 3DES 3-key EDE encryption in ECB mode these tests are
run:
@enumerate
@item
A known answer test is run on a 64 bit test vector processed by 64
rounds of Single-DES block encryption and decryption using a key
changed with each round.
@item
A known answer test is run on a 64 bit test vector processed by 16
rounds of 2-key and 3-key Triple-DES block encryption and decryptions
using a key changed with each round.
@item
10 known answer tests using 3-key Triple-DES EDE encryption, comparing
the ciphertext to the known value, then running a decryption and
comparing it to the initial plaintext.
@end enumerate
(@code{cipher/des.c:selftest})
@item AES-128
A known answer tests is run using one test vector and one test
key with AES in ECB mode. (@code{cipher/rijndael.c:selftest_basic_128})
@item AES-192
A known answer tests is run using one test vector and one test
key with AES in ECB mode. (@code{cipher/rijndael.c:selftest_basic_192})
@item AES-256
A known answer tests is run using one test vector and one test key
with AES in ECB mode. (@code{cipher/rijndael.c:selftest_basic_256})
@end table
@subsection Hash Algorithm Power-Up Tests
The following hash algorithm tests are run during power-up:
@table @asis
@item SHA-1
A known answer test using the string @code{"abc"} is run.
(@code{cipher/@/sha1.c:@/selftests_sha1})
@item SHA-224
A known answer test using the string @code{"abc"} is run.
(@code{cipher/@/sha256.c:@/selftests_sha224})
@item SHA-256
A known answer test using the string @code{"abc"} is run.
(@code{cipher/@/sha256.c:@/selftests_sha256})
@item SHA-384
A known answer test using the string @code{"abc"} is run.
(@code{cipher/@/sha512.c:@/selftests_sha384})
@item SHA-512
A known answer test using the string @code{"abc"} is run.
(@code{cipher/@/sha512.c:@/selftests_sha512})
@end table
@subsection MAC Algorithm Power-Up Tests
The following MAC algorithm tests are run during power-up:
@table @asis
@item HMAC SHA-1
A known answer test using 9 byte of data and a 64 byte key is run.
(@code{cipher/hmac-tests.c:selftests_sha1})
@item HMAC SHA-224
A known answer test using 28 byte of data and a 4 byte key is run.
(@code{cipher/hmac-tests.c:selftests_sha224})
@item HMAC SHA-256
A known answer test using 28 byte of data and a 4 byte key is run.
(@code{cipher/hmac-tests.c:selftests_sha256})
@item HMAC SHA-384
A known answer test using 28 byte of data and a 4 byte key is run.
(@code{cipher/hmac-tests.c:selftests_sha384})
@item HMAC SHA-512
A known answer test using 28 byte of data and a 4 byte key is run.
(@code{cipher/hmac-tests.c:selftests_sha512})
@end table
@subsection Random Number Power-Up Test
The DRNG is tested during power-up this way:
@enumerate
@item
Requesting one block of random using the public interface to check
general working and the duplicated block detection.
@item
3 know answer tests using pre-defined keys, seed and initial DT
values. For each test 3 blocks of 16 bytes are requested and compared
to the expected result. The DT value is incremented for each block.
@end enumerate
@subsection Public Key Algorithm Power-Up Tests
The public key algorithms are tested during power-up:
@table @asis
@item RSA
A pre-defined 1024 bit RSA key is used and these tests are run
in turn:
@enumerate
@item
Conversion of S-expression to internal format.
(@code{cipher/@/rsa.c:@/selftests_rsa})
@item
Private key consistency check.
(@code{cipher/@/rsa.c:@/selftests_rsa})
@item
A pre-defined 20 byte value is signed with PKCS#1 padding for SHA-1.
The result is verified using the public key against the original data
and against modified data. (@code{cipher/@/rsa.c:@/selftest_sign_1024})
@item
A 1000 bit random value is encrypted and checked that it does not
match the original random value. The encrypted result is then
decrypted and checked that it matches the original random value.
(@code{cipher/@/rsa.c:@/selftest_encr_1024})
@end enumerate
@item DSA
A pre-defined 1024 bit DSA key is used and these tests are run in turn:
@enumerate
@item
Conversion of S-expression to internal format.
(@code{cipher/@/dsa.c:@/selftests_dsa})
@item
Private key consistency check.
(@code{cipher/@/dsa.c:@/selftests_dsa})
@item
A pre-defined 20 byte value is signed with PKCS#1 padding for
SHA-1. The result is verified using the public key against the
original data and against modified data.
(@code{cipher/@/dsa.c:@/selftest_sign_1024})
@end enumerate
@end table
@subsection Integrity Power-Up Tests
The integrity of the Libgcrypt is tested during power-up but only if
checking has been enabled at build time. The check works by computing
a HMAC SHA-256 checksum over the file used to load Libgcrypt into
memory. That checksum is compared against a checksum stored in a file
of the same name but with a single dot as a prefix and a suffix of
@file{.hmac}.
@subsection Critical Functions Power-Up Tests
The 3DES weak key detection is tested during power-up by calling the
detection function with keys taken from a table listening all weak
keys. The table itself is protected using a SHA-1 hash.
(@code{cipher/@/des.c:@/selftest})
@c --------------------------------
@section Conditional Tests
The conditional tests are performed if a certain condition is met.
This may occur at any time; the library does not necessary enter the
``Self-Test'' state to run these tests but will transit to the
``Error'' state if a test failed.
@subsection Key-Pair Generation Tests
After an asymmetric key-pair has been generated, Libgcrypt runs a
pair-wise consistency tests on the generated key. On failure the
generated key is not used, an error code is returned and, if in FIPS
mode, the library is put into the ``Error'' state.
@table @asis
@item RSA
The test uses a random number 64 bits less the size of the modulus as
plaintext and runs an encryption and decryption operation in turn. The
encrypted value is checked to not match the plaintext and the result
of the decryption is checked to match the plaintext.
A new random number of the same size is generated, signed and verified
to test the correctness of the signing operation. As a second signing
test, the signature is modified by incrementing its value and then
verified with the expected result that the verification fails.
(@code{cipher/@/rsa.c:@/test_keys})
@item DSA
The test uses a random number of the size of the Q parameter to create
a signature and then checks that the signature verifies. As a second
signing test, the data is modified by incrementing its value and then
verified against the signature with the expected result that the
verification fails. (@code{cipher/@/dsa.c:@/test_keys})
@end table
@subsection Software Load Tests
No code is loaded at runtime.
@subsection Manual Key Entry Tests
A manual key entry feature is not implemented in Libgcrypt.
@subsection Continuous RNG Tests
The continuous random number test is only used in FIPS mode. The RNG
generates blocks of 128 bit size; the first block generated per
context is saved in the context and another block is generated to be
returned to the caller. Each block is compared against the saved
block and then stored in the context. If a duplicated block is
detected an error is signaled and the library is put into the
``Fatal-Error'' state.
(@code{random/@/random-fips.c:@/x931_aes_driver})
@c --------------------------------
@section Application Requested Tests
The application may requests tests at any time by means of the
@code{GCRYCTL_SELFTEST} control command. Note that using these tests
is not FIPS conform: Although Libgcrypt rejects all application
requests for services while running self-tests, it does not ensure
that no other operations of Libgcrypt are still being executed. Thus,
in FIPS mode an application requesting self-tests needs to power-cycle
Libgcrypt instead.
When self-tests are requested, Libgcrypt runs all the tests it does
during power-up as well as a few extra checks as described below.
@subsection Symmetric Cipher Algorithm Tests
The following symmetric encryption algorithm tests are run in addition
to the power-up tests:
@table @asis
@item AES-128
A known answer tests with test vectors taken from NIST SP800-38a and
using the high level functions is run for block modes CFB and OFB.
@end table
@subsection Hash Algorithm Tests
The following hash algorithm tests are run in addition to the
power-up tests:
@table @asis
@item SHA-1
@itemx SHA-224
@itemx SHA-256
@enumerate
@item
A known answer test using a 56 byte string is run.
@item
A known answer test using a string of one million letters "a" is run.
@end enumerate
(@code{cipher/@/sha1.c:@/selftests_sha1},
@code{cipher/@/sha256.c:@/selftests_sha224},
@code{cipher/@/sha256.c:@/selftests_sha256})
@item SHA-384
@item SHA-512
@enumerate
@item
A known answer test using a 112 byte string is run.
@item
A known answer test using a string of one million letters "a" is run.
@end enumerate
(@code{cipher/@/sha512.c:@/selftests_sha384},
@code{cipher/@/sha512.c:@/selftests_sha512})
@end table
@subsection MAC Algorithm Tests
The following MAC algorithm tests are run in addition to the power-up
tests:
@table @asis
@item HMAC SHA-1
@enumerate
@item
A known answer test using 9 byte of data and a 20 byte key is run.
@item
A known answer test using 9 byte of data and a 100 byte key is run.
@item
A known answer test using 9 byte of data and a 49 byte key is run.
@end enumerate
(@code{cipher/hmac-tests.c:selftests_sha1})
@item HMAC SHA-224
@itemx HMAC SHA-256
@itemx HMAC SHA-384
@itemx HMAC SHA-512
@enumerate
@item
A known answer test using 9 byte of data and a 20 byte key is run.
@item
A known answer test using 50 byte of data and a 20 byte key is run.
@item
A known answer test using 50 byte of data and a 26 byte key is run.
@item
A known answer test using 54 byte of data and a 131 byte key is run.
@item
A known answer test using 152 byte of data and a 131 byte key is run.
@end enumerate
(@code{cipher/@/hmac-tests.c:@/selftests_sha224},
@code{cipher/@/hmac-tests.c:@/selftests_sha256},
@code{cipher/@/hmac-tests.c:@/selftests_sha384},
@code{cipher/@/hmac-tests.c:@/selftests_sha512})
@end table
@c ********************************************
@node FIPS Mode
@appendix Description of the FIPS Mode
This appendix gives detailed information pertaining to the FIPS mode.
In particular, the changes to the standard mode and the finite state
machine are described. The self-tests required in this mode are
described in the appendix on self-tests.
@c -------------------------------
@section Restrictions in FIPS Mode
@noindent
If Libgcrypt is used in FIPS mode these restrictions are effective:
@itemize
@item
The cryptographic algorithms are restricted to this list:
@table @asis
@item GCRY_CIPHER_3DES
3 key EDE Triple-DES symmetric encryption.
@item GCRY_CIPHER_AES128
AES 128 bit symmetric encryption.
@item GCRY_CIPHER_AES192
AES 192 bit symmetric encryption.
@item GCRY_CIPHER_AES256
AES 256 bit symmetric encryption.
@item GCRY_MD_SHA1
SHA-1 message digest.
@item GCRY_MD_SHA224
SHA-224 message digest.
@item GCRY_MD_SHA256
SHA-256 message digest.
@item GCRY_MD_SHA384
SHA-384 message digest.
@item GCRY_MD_SHA512
SHA-512 message digest.
@item GCRY_MD_SHA1,GCRY_MD_FLAG_HMAC
HMAC using a SHA-1 message digest.
@item GCRY_MD_SHA224,GCRY_MD_FLAG_HMAC
HMAC using a SHA-224 message digest.
@item GCRY_MD_SHA256,GCRY_MD_FLAG_HMAC
HMAC using a SHA-256 message digest.
@item GCRY_MD_SHA384,GCRY_MD_FLAG_HMAC
HMAC using a SHA-384 message digest.
@item GCRY_MD_SHA512,GCRY_MD_FLAG_HMAC
HMAC using a SHA-512 message digest.
@item GCRY_PK_RSA
RSA encryption and signing.
@item GCRY_PK_DSA
DSA signing.
@end table
Note that the CRC algorithms are not considered cryptographic algorithms
and thus are in addition available.
@item
RSA key generation refuses to create a key with a keysize of
less than 1024 bits.
@item
DSA key generation refuses to create a key with a keysize other
than 1024 bits.
@item
The @code{transient-key} flag for RSA and DSA key generation is ignored.
@item
Support for the VIA Padlock engine is disabled.
@item
FIPS mode may only be used on systems with a /dev/random device.
Switching into FIPS mode on other systems will fail at runtime.
@item
Saving and loading a random seed file is ignored.
@item
An X9.31 style random number generator is used in place of the
large-pool-CSPRNG generator.
@item
The command @code{GCRYCTL_ENABLE_QUICK_RANDOM} is ignored.
@item
Message digest debugging is disabled.
@item
All debug output related to cryptographic data is suppressed.
@item
On-the-fly self-tests are not performed, instead self-tests are run
before entering operational state.
@item
The function @code{gcry_set_allocation_handler} may not be used. If
it is used Libgcrypt disables FIPS mode unless Enforced FIPS mode is
enabled, in which case Libgcrypt will enter the error state.
@item
The digest algorithm MD5 may not be used. If it is used Libgcrypt
disables FIPS mode unless Enforced FIPS mode is enabled, in which case
Libgcrypt will enter the error state.
@item
In Enforced FIPS mode the command @code{GCRYCTL_DISABLE_SECMEM} is
ignored. In standard FIPS mode it disables FIPS mode.
@item
A handler set by @code{gcry_set_outofcore_handler} is ignored.
@item
A handler set by @code{gcry_set_fatalerror_handler} is ignored.
@end itemize
Note that when we speak about disabling FIPS mode, it merely means
that the function @code{gcry_fips_mode_active} returns false; it does
not mean that any non FIPS algorithms are allowed.
@c ********************************************
@section FIPS Finite State Machine
The FIPS mode of libgcrypt implements a finite state machine (FSM) using
8 states (@pxref{tbl:fips-states}) and checks at runtime that only valid
transitions (@pxref{tbl:fips-state-transitions}) may happen.
@float Figure,fig:fips-fsm
@caption{FIPS mode state diagram}
@center @image{fips-fsm,150mm,,FIPS FSM Diagram}
@end float
@float Table,tbl:fips-states
@caption{FIPS mode states}
@noindent
States used by the FIPS FSM:
@table @asis
@item Power-Off
Libgcrypt is not runtime linked to another application. This usually
means that the library is not loaded into main memory. This state is
documentation only.
@item Power-On
Libgcrypt is loaded into memory and API calls may be made. Compiler
introduced constructor functions may be run. Note that Libgcrypt does
not implement any arbitrary constructor functions to be called by the
operating system
@item Init
The Libgcrypt initialization functions are performed and the library has
not yet run any self-test.
@item Self-Test
Libgcrypt is performing self-tests.
@item Operational
Libgcrypt is in the operational state and all interfaces may be used.
@item Error
Libgrypt is in the error state. When calling any FIPS relevant
interfaces they either return an error (@code{GPG_ERR_NOT_OPERATIONAL})
or put Libgcrypt into the Fatal-Error state and won't return.
@item Fatal-Error
Libgcrypt is in a non-recoverable error state and
will automatically transit into the Shutdown state.
@item Shutdown
Libgcrypt is about to be terminated and removed from the memory. The
application may at this point still running cleanup handlers.
@end table
@end float
@float Table,tbl:fips-state-transitions
@caption{FIPS mode state transitions}
@noindent
The valid state transitions (@pxref{fig:fips-fsm}) are:
@table @code
@item 1
Power-Off to Power-On is implicitly done by the OS loading Libgcrypt as
a shared library and having it linked to an application.
@item 2
Power-On to Init is triggered by the application calling the
Libgcrypt initialization function @code{gcry_check_version}.
@item 3
Init to Self-Test is either triggered by a dedicated API call or implicit
by invoking a libgrypt service controlled by the FSM.
@item 4
Self-Test to Operational is triggered after all self-tests passed
successfully.
@item 5
Operational to Shutdown is an artificial state without any direct action
in Libgcrypt. When reaching the Shutdown state the library is
deinitialized and can't return to any other state again.
@item 6
Shutdown to Power-off is the process of removing Libgcrypt from the
computer's memory. For obvious reasons the Power-Off state can't be
represented within Libgcrypt and thus this transition is for
documentation only.
@item 7
Operational to Error is triggered if Libgcrypt detected an application
error which can't be returned to the caller but still allows Libgcrypt
to properly run. In the Error state all FIPS relevant interfaces return
an error code.
@item 8
Error to Shutdown is similar to the Operational to Shutdown transition
(5).
@item 9
Error to Fatal-Error is triggered if Libgrypt detects an fatal error
while already being in Error state.
@item 10
Fatal-Error to Shutdown is automatically entered by Libgcrypt
after having reported the error.
@item 11
Power-On to Shutdown is an artificial state to document that Libgcrypt
has not ye been initialized but the process is about to terminate.
@item 12
Power-On to Fatal-Error will be triggered if certain Libgcrypt functions
are used without having reached the Init state.
@item 13
Self-Test to Fatal-Error is triggered by severe errors in Libgcrypt while
running self-tests.
@item 14
Self-Test to Error is triggered by a failed self-test.
@item 15
Operational to Fatal-Error is triggered if Libcrypt encountered a
non-recoverable error.
@item 16
Operational to Self-Test is triggered if the application requested to run
the self-tests again.
@item 17
Error to Self-Test is triggered if the application has requested to run
self-tests to get to get back into operational state after an error.
@item 18
Init to Error is triggered by errors in the initialization code.
@item 19
Init to Fatal-Error is triggered by non-recoverable errors in the
initialization code.
@item 20
Error to Error is triggered by errors while already in the Error
state.
@end table
@end float
@c ********************************************
@section FIPS Miscellaneous Information
Libgcrypt does not do any key management on itself; the application
needs to care about it. Keys which are passed to Libgcrypt should be
allocated in secure memory as available with the functions
@code{gcry_malloc_secure} and @code{gcry_calloc_secure}. By calling
@code{gcry_free} on this memory, the memory and thus the keys are
overwritten with zero bytes before releasing the memory.
For use with the random number generator, Libgcrypt generates 3
internal keys which are stored in the encryption contexts used by the
RNG. These keys are stored in secure memory for the lifetime of the
process. Application are required to use @code{GCRYCTL_TERM_SECMEM}
before process termination. This will zero out the entire secure
memory and thus also the encryption contexts with these keys.
@c **********************************************************
@c ************* Appendices (license etc.) ****************
@c **********************************************************
@include lgpl.texi
@include gpl.texi
@node Figures and Tables
@unnumbered List of Figures and Tables
@listoffloats Figure
@listoffloats Table
@node Concept Index
@unnumbered Concept Index
@printindex cp
@node Function and Data Index
@unnumbered Function and Data Index
@printindex fn
@bye
GCRYCTL_SET_RANDOM_DAEMON_SOCKET
GCRYCTL_USE_RANDOM_DAEMON
The random daemon is still a bit experimental, thus we do not document
them. Note that they should be used during initialization and that
these functions are not really thread safe.
@c LocalWords: int HD
diff --git a/random/random-csprng.c b/random/random-csprng.c
index e7b751a2..a0bfc783 100644
--- a/random/random-csprng.c
+++ b/random/random-csprng.c
@@ -1,1272 +1,1273 @@
/* random-csprng.c - CSPRNG style random number generator (libgcrypt classic)
* Copyright (C) 1998, 2000, 2001, 2002, 2003, 2004, 2005, 2006,
* 2007, 2008, 2010, 2012 Free Software Foundation, Inc.
*
* This file is part of Libgcrypt.
*
* Libgcrypt is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of
* the License, or (at your option) any later version.
*
* Libgcrypt is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this program; if not, see <http://www.gnu.org/licenses/>.
*/
/*
This random number generator is modelled after the one described in
Peter Gutmann's 1998 Usenix Security Symposium paper: "Software
Generation of Practically Strong Random Numbers". See also chapter
6 in his book "Cryptographic Security Architecture", New York,
2004, ISBN 0-387-95387-6.
Note that the acronym CSPRNG stands for "Continuously Seeded
PseudoRandom Number Generator" as used in Peter's implementation of
the paper and not only for "Cryptographically Secure PseudoRandom
Number Generator".
*/
#include <config.h>
#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <string.h>
#include <sys/time.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <fcntl.h>
#include <time.h>
#ifdef HAVE_GETHRTIME
#include <sys/times.h>
#endif
#ifdef HAVE_GETTIMEOFDAY
#include <sys/time.h>
#endif
#ifdef HAVE_GETRUSAGE
#include <sys/resource.h>
#endif
#ifdef __MINGW32__
#include <process.h>
#endif
#include "g10lib.h"
-#include "../cipher/rmd.h"
#include "random.h"
#include "rand-internal.h"
-#include "cipher.h" /* Required for the rmd160_hash_buffer() prototype. */
+#include "cipher.h" /* _gcry_sha1_hash_buffer */
+#include "../cipher/sha1.h" /* _gcry_sha1_mixblock */
#ifndef RAND_MAX /* For SunOS. */
#define RAND_MAX 32767
#endif
/* Check whether we can lock the seed file read write. */
#if defined(HAVE_FCNTL) && defined(HAVE_FTRUNCATE) && !defined(HAVE_W32_SYSTEM)
#define LOCK_SEED_FILE 1
#else
#define LOCK_SEED_FILE 0
#endif
/* Define the constant we use for transforming the pool at read-out. */
#if SIZEOF_UNSIGNED_LONG == 8
#define ADD_VALUE 0xa5a5a5a5a5a5a5a5
#elif SIZEOF_UNSIGNED_LONG == 4
#define ADD_VALUE 0xa5a5a5a5
#else
#error weird size for an unsigned long
#endif
/* Contstants pertaining to the hash pool. */
#define BLOCKLEN 64 /* Hash this amount of bytes... */
-#define DIGESTLEN 20 /* ... into a digest of this length (rmd160). */
+#define DIGESTLEN 20 /* ... into a digest of this length (sha-1). */
/* POOLBLOCKS is the number of digests which make up the pool. */
#define POOLBLOCKS 30
/* POOLSIZE must be a multiple of the digest length to make the AND
operations faster, the size should also be a multiple of unsigned
long. */
#define POOLSIZE (POOLBLOCKS*DIGESTLEN)
#if (POOLSIZE % SIZEOF_UNSIGNED_LONG)
#error Please make sure that poolsize is a multiple of unsigned long
#endif
#define POOLWORDS (POOLSIZE / SIZEOF_UNSIGNED_LONG)
/* RNDPOOL is the pool we use to collect the entropy and to stir it
up. Its allocated size is POOLSIZE+BLOCKLEN. Note that this is
also an indication on whether the module has been fully
initialized. */
static unsigned char *rndpool;
/* KEYPOOL is used as a scratch copy to read out random from RNDPOOL.
Its allocated size is also POOLSIZE+BLOCKLEN. */
static unsigned char *keypool;
/* This is the offset into RNDPOOL where the next random bytes are to
be mixed in. */
static size_t pool_writepos;
/* When reading data out of KEYPOOL, we start the read at different
positions. This variable keeps track on where to read next. */
static size_t pool_readpos;
/* This flag is set to true as soon as the pool has been completely
filled the first time. This may happen either by rereading a seed
file or by adding enough entropy. */
static int pool_filled;
/* This counter is used to track whether the initial seeding has been
done with enough bytes from a reliable entropy source. */
static size_t pool_filled_counter;
/* If random of level GCRY_VERY_STRONG_RANDOM has been requested we
have stricter requirements on what kind of entropy is in the pool.
In particular POOL_FILLED is not sufficient. Thus we add some
extra seeding and set this flag to true if the extra seeding has
been done. */
static int did_initial_extra_seeding;
/* This variable is used to estimated the amount of fresh entropy
available in RNDPOOL. */
static int pool_balance;
/* After a mixing operation this variable will be set to true and
cleared if new entropy has been added or a remix is required for
other reasons. */
static int just_mixed;
/* The name of the seed file or NULL if no seed file has been defined.
The seed file needs to be regsitered at initialiation time. We
keep a malloced copy here. */
static char *seed_file_name;
/* If a seed file has been registered and maybe updated on exit this
flag set. */
static int allow_seed_file_update;
/* Option flag set at initialiation time to force allocation of the
pool in secure memory. */
static int secure_alloc;
/* This function pointer is set to the actual entropy gathering
function during initialization. After initialization it is
guaranteed to point to function. (On systems without a random
gatherer module a dummy function is used).*/
static int (*slow_gather_fnc)(void (*)(const void*, size_t,
enum random_origins),
enum random_origins, size_t, int);
/* This function is set to the actual fast entropy gathering function
during initialization. If it is NULL, no such function is
available. */
static void (*fast_gather_fnc)(void (*)(const void*, size_t,
enum random_origins),
enum random_origins);
/* Option flag useful for debugging and the test suite. If set
requests for very strong random are degraded to strong random. Not
used by regular applications. */
static int quick_test;
/* This is the lock we use to protect all pool operations. */
GPGRT_LOCK_DEFINE (pool_lock);
/* This is a helper for assert calls. These calls are used to assert
that functions are called in a locked state. It is not meant to be
thread-safe but as a method to get aware of missing locks in the
test suite. */
static int pool_is_locked;
/* We keep some counters in this structure for the sake of the
_gcry_random_dump_stats () function. */
static struct
{
unsigned long mixrnd;
unsigned long mixkey;
unsigned long slowpolls;
unsigned long fastpolls;
unsigned long getbytes1;
unsigned long ngetbytes1;
unsigned long getbytes2;
unsigned long ngetbytes2;
unsigned long addbytes;
unsigned long naddbytes;
} rndstats;
/* --- Stuff pertaining to the random daemon support. --- */
#ifdef USE_RANDOM_DAEMON
/* If ALLOW_DAEMON is true, the module will try to use the random
daemon first. If the daemon has failed, this variable is set to
back to false and the code continues as normal. Note, we don't
test this flag in a locked state because a wrong value does not
harm and the trhead will find out itself that the daemon does not
work and set it (again) to false. */
static int allow_daemon;
/* During initialization, the user may set a non-default socket name
for accessing the random daemon. If this value is NULL, the
default name will be used. */
static char *daemon_socket_name;
#endif /*USE_RANDOM_DAEMON*/
/* --- Prototypes --- */
static void read_pool (byte *buffer, size_t length, int level );
static void add_randomness (const void *buffer, size_t length,
enum random_origins origin);
static void random_poll (void);
static void do_fast_random_poll (void);
static int (*getfnc_gather_random (void))(void (*)(const void*, size_t,
enum random_origins),
enum random_origins, size_t, int);
static void (*getfnc_fast_random_poll (void))(void (*)(const void*, size_t,
enum random_origins),
enum random_origins);
static void read_random_source (enum random_origins origin,
size_t length, int level);
/* --- Functions --- */
/* Basic initialization which is required to initialize mutexes and
such. It does not run a full initialization so that the filling of
the random pool can be delayed until it is actually needed. We
assume that this function is used before any concurrent access
happens. */
static void
initialize_basics(void)
{
static int initialized;
if (!initialized)
{
initialized = 1;
#ifdef USE_RANDOM_DAEMON
_gcry_daemon_initialize_basics ();
#endif /*USE_RANDOM_DAEMON*/
/* Make sure that we are still using the values we have
traditionally used for the random levels. */
gcry_assert (GCRY_WEAK_RANDOM == 0
&& GCRY_STRONG_RANDOM == 1
&& GCRY_VERY_STRONG_RANDOM == 2);
}
}
/* Take the pool lock. */
static void
lock_pool (void)
{
int err;
err = gpgrt_lock_lock (&pool_lock);
if (err)
log_fatal ("failed to acquire the pool lock: %s\n", gpg_strerror (err));
pool_is_locked = 1;
}
/* Release the pool lock. */
static void
unlock_pool (void)
{
int err;
pool_is_locked = 0;
err = gpgrt_lock_unlock (&pool_lock);
if (err)
log_fatal ("failed to release the pool lock: %s\n", gpg_strerror (err));
}
/* Full initialization of this module. */
static void
initialize(void)
{
/* Although the basic initialization should have happened already,
we call it here to make sure that all prerequisites are met. */
initialize_basics ();
/* Now we can look the pool and complete the initialization if
necessary. */
lock_pool ();
if (!rndpool)
{
/* The data buffer is allocated somewhat larger, so that we can
use this extra space (which is allocated in secure memory) as
a temporary hash buffer */
rndpool = (secure_alloc
? xcalloc_secure (1, POOLSIZE + BLOCKLEN)
: xcalloc (1, POOLSIZE + BLOCKLEN));
keypool = (secure_alloc
? xcalloc_secure (1, POOLSIZE + BLOCKLEN)
: xcalloc (1, POOLSIZE + BLOCKLEN));
/* Setup the slow entropy gathering function. The code requires
that this function exists. */
slow_gather_fnc = getfnc_gather_random ();
/* Setup the fast entropy gathering function. */
fast_gather_fnc = getfnc_fast_random_poll ();
}
unlock_pool ();
}
/* Initialize this random subsystem. If FULL is false, this function
merely calls the initialize and does not do anything more. Doing
this is not really required but when running in a threaded
environment we might get a race condition otherwise. */
void
_gcry_rngcsprng_initialize (int full)
{
if (!full)
initialize_basics ();
else
initialize ();
}
/* Try to close the FDs of the random gather module. This is
currently only implemented for rndlinux. */
void
_gcry_rngcsprng_close_fds (void)
{
lock_pool ();
#if USE_RNDLINUX
_gcry_rndlinux_gather_random (NULL, 0, 0, 0);
pool_filled = 0; /* Force re-open on next use. */
#endif
unlock_pool ();
}
void
_gcry_rngcsprng_dump_stats (void)
{
/* In theory we would need to lock the stats here. However this
function is usually called during cleanup and then we _might_ run
into problems. */
log_info ("random usage: poolsize=%d mixed=%lu polls=%lu/%lu added=%lu/%lu\n"
" outmix=%lu getlvl1=%lu/%lu getlvl2=%lu/%lu%s\n",
POOLSIZE, rndstats.mixrnd, rndstats.slowpolls, rndstats.fastpolls,
rndstats.naddbytes, rndstats.addbytes,
rndstats.mixkey, rndstats.ngetbytes1, rndstats.getbytes1,
rndstats.ngetbytes2, rndstats.getbytes2,
_gcry_rndhw_failed_p()? " (hwrng failed)":"");
}
/* This function should be called during initialization and before
initialization of this module to place the random pools into secure
memory. */
void
_gcry_rngcsprng_secure_alloc (void)
{
secure_alloc = 1;
}
/* This may be called before full initialization to degrade the
quality of the RNG for the sake of a faster running test suite. */
void
_gcry_rngcsprng_enable_quick_gen (void)
{
quick_test = 1;
}
void
_gcry_rngcsprng_set_daemon_socket (const char *socketname)
{
#ifdef USE_RANDOM_DAEMON
if (daemon_socket_name)
BUG ();
daemon_socket_name = gcry_xstrdup (socketname);
#else /*!USE_RANDOM_DAEMON*/
(void)socketname;
#endif /*!USE_RANDOM_DAEMON*/
}
/* With ONOFF set to 1, enable the use of the daemon. With ONOFF set
to 0, disable the use of the daemon. With ONOF set to -1, return
whether the daemon has been enabled. */
int
_gcry_rngcsprng_use_daemon (int onoff)
{
#ifdef USE_RANDOM_DAEMON
int last;
/* This is not really thread safe. However it is expected that this
function is being called during initialization and at that point
we are for other reasons not really thread safe. We do not want
to lock it because we might eventually decide that this function
may even be called prior to gcry_check_version. */
last = allow_daemon;
if (onoff != -1)
allow_daemon = onoff;
return last;
#else /*!USE_RANDOM_DAEMON*/
(void)onoff;
return 0;
#endif /*!USE_RANDOM_DAEMON*/
}
/* This function returns true if no real RNG is available or the
quality of the RNG has been degraded for test purposes. */
int
_gcry_rngcsprng_is_faked (void)
{
/* We need to initialize due to the runtime determination of
available entropy gather modules. */
initialize();
return quick_test;
}
/* Add BUFLEN bytes from BUF to the internal random pool. QUALITY
should be in the range of 0..100 to indicate the goodness of the
entropy added, or -1 for goodness not known. */
gcry_error_t
_gcry_rngcsprng_add_bytes (const void *buf, size_t buflen, int quality)
{
size_t nbytes;
const char *bufptr;
if (quality == -1)
quality = 35;
else if (quality > 100)
quality = 100;
else if (quality < 0)
quality = 0;
if (!buf)
return gpg_error (GPG_ERR_INV_ARG);
if (!buflen || quality < 10)
return 0; /* Take a shortcut. */
/* Because we don't increment the entropy estimation with FASTPOLL,
we don't need to take lock that estimation while adding from an
external source. This limited entropy estimation also means that
we can't take QUALITY into account. */
initialize_basics ();
bufptr = buf;
while (buflen)
{
nbytes = buflen > POOLSIZE? POOLSIZE : buflen;
lock_pool ();
if (rndpool)
add_randomness (bufptr, nbytes, RANDOM_ORIGIN_EXTERNAL);
unlock_pool ();
bufptr += nbytes;
buflen -= nbytes;
}
return 0;
}
/* Public function to fill the buffer with LENGTH bytes of
cryptographically strong random bytes. Level GCRY_WEAK_RANDOM is
not very strong, GCRY_STRONG_RANDOM is strong enough for most
usage, GCRY_VERY_STRONG_RANDOM is good for key generation stuff but
may be very slow. */
void
_gcry_rngcsprng_randomize (void *buffer, size_t length,
enum gcry_random_level level)
{
unsigned char *p;
/* Make sure we are initialized. */
initialize ();
/* Handle our hack used for regression tests of Libgcrypt. */
if ( quick_test && level > GCRY_STRONG_RANDOM )
level = GCRY_STRONG_RANDOM;
/* Make sure the level is okay. */
level &= 3;
#ifdef USE_RANDOM_DAEMON
if (allow_daemon
&& !_gcry_daemon_randomize (daemon_socket_name, buffer, length, level))
return; /* The daemon succeeded. */
allow_daemon = 0; /* Daemon failed - switch off. */
#endif /*USE_RANDOM_DAEMON*/
/* Acquire the pool lock. */
lock_pool ();
/* Update the statistics. */
if (level >= GCRY_VERY_STRONG_RANDOM)
{
rndstats.getbytes2 += length;
rndstats.ngetbytes2++;
}
else
{
rndstats.getbytes1 += length;
rndstats.ngetbytes1++;
}
/* Read the random into the provided buffer. */
for (p = buffer; length > 0;)
{
size_t n;
n = length > POOLSIZE? POOLSIZE : length;
read_pool (p, n, level);
length -= n;
p += n;
}
/* Release the pool lock. */
unlock_pool ();
}
/*
Mix the pool:
|........blocks*20byte........|20byte|..44byte..|
<..44byte..> <20byte>
| |
| +------+
+---------------------------|----------+
v v
|........blocks*20byte........|20byte|..44byte..|
<.....64bytes.....>
|
+----------------------------------+
Hash
v
|.............................|20byte|..44byte..|
<20byte><20byte><..44byte..>
| |
| +---------------------+
+-----------------------------+ |
v v
|.............................|20byte|..44byte..|
<.....64byte......>
|
+-------------------------+
Hash
v
|.............................|20byte|..44byte..|
<20byte><20byte><..44byte..>
and so on until we did this for all blocks.
To better protect against implementation errors in this code, we
xor a digest of the entire pool into the pool before mixing.
Note: this function must only be called with a locked pool.
*/
static void
mix_pool(unsigned char *pool)
{
static unsigned char failsafe_digest[DIGESTLEN];
static int failsafe_digest_valid;
unsigned char *hashbuf = pool + POOLSIZE;
unsigned char *p, *pend;
int i, n;
- RMD160_CONTEXT md;
+ SHA1_CONTEXT md;
+ unsigned int nburn;
#if DIGESTLEN != 20
-#error must have a digest length of 20 for ripe-md-160
+#error must have a digest length of 20 for SHA-1
#endif
gcry_assert (pool_is_locked);
- _gcry_rmd160_init( &md );
+ _gcry_sha1_mixblock_init (&md);
/* Loop over the pool. */
pend = pool + POOLSIZE;
memcpy(hashbuf, pend - DIGESTLEN, DIGESTLEN );
memcpy(hashbuf+DIGESTLEN, pool, BLOCKLEN-DIGESTLEN);
- _gcry_rmd160_mixblock( &md, hashbuf);
+ nburn = _gcry_sha1_mixblock (&md, hashbuf);
memcpy(pool, hashbuf, 20 );
if (failsafe_digest_valid && pool == rndpool)
{
for (i=0; i < 20; i++)
pool[i] ^= failsafe_digest[i];
}
p = pool;
for (n=1; n < POOLBLOCKS; n++)
{
memcpy (hashbuf, p, DIGESTLEN);
p += DIGESTLEN;
if (p+DIGESTLEN+BLOCKLEN < pend)
memcpy (hashbuf+DIGESTLEN, p+DIGESTLEN, BLOCKLEN-DIGESTLEN);
else
{
unsigned char *pp = p + DIGESTLEN;
for (i=DIGESTLEN; i < BLOCKLEN; i++ )
{
if ( pp >= pend )
pp = pool;
hashbuf[i] = *pp++;
}
}
- _gcry_rmd160_mixblock ( &md, hashbuf);
+ _gcry_sha1_mixblock (&md, hashbuf);
memcpy(p, hashbuf, 20 );
}
- /* Our hash implementation does only leave small parts (64 bytes)
- of the pool on the stack, so it is okay not to require secure
- memory here. Before we use this pool, it will be copied to the
- help buffer anyway. */
- if ( pool == rndpool)
- {
- _gcry_rmd160_hash_buffer (failsafe_digest, pool, POOLSIZE);
- failsafe_digest_valid = 1;
- }
-
- _gcry_burn_stack (384); /* for the rmd160_mixblock(), rmd160_hash_buffer */
+ /* Our hash implementation does only leave small parts (64 bytes)
+ of the pool on the stack, so it is okay not to require secure
+ memory here. Before we use this pool, it will be copied to the
+ help buffer anyway. */
+ if ( pool == rndpool)
+ {
+ _gcry_sha1_hash_buffer (failsafe_digest, pool, POOLSIZE);
+ failsafe_digest_valid = 1;
+ }
+
+ _gcry_burn_stack (nburn);
}
void
_gcry_rngcsprng_set_seed_file (const char *name)
{
if (seed_file_name)
BUG ();
seed_file_name = xstrdup (name);
}
/* Lock an open file identified by file descriptor FD and wait a
reasonable time to succeed. With FOR_WRITE set to true a write
lock will be taken. FNAME is used only for diagnostics. Returns 0
on success or -1 on error. */
static int
lock_seed_file (int fd, const char *fname, int for_write)
{
#ifdef __GCC__
#warning Check whether we can lock on Windows.
#endif
#if LOCK_SEED_FILE
struct flock lck;
struct timeval tv;
int backoff=0;
/* We take a lock on the entire file. */
memset (&lck, 0, sizeof lck);
lck.l_type = for_write? F_WRLCK : F_RDLCK;
lck.l_whence = SEEK_SET;
while (fcntl (fd, F_SETLK, &lck) == -1)
{
if (errno != EAGAIN && errno != EACCES)
{
log_info (_("can't lock `%s': %s\n"), fname, strerror (errno));
return -1;
}
if (backoff > 2) /* Show the first message after ~2.25 seconds. */
log_info( _("waiting for lock on `%s'...\n"), fname);
tv.tv_sec = backoff;
tv.tv_usec = 250000;
select (0, NULL, NULL, NULL, &tv);
if (backoff < 10)
backoff++ ;
}
#endif /*!LOCK_SEED_FILE*/
return 0;
}
/* Read in a seed from the random_seed file and return true if this
was successful.
Note: Multiple instances of applications sharing the same random
seed file can be started in parallel, in which case they will read
out the same pool and then race for updating it (the last update
overwrites earlier updates). They will differentiate only by the
weak entropy that is added in read_seed_file based on the PID and
clock, and up to 16 bytes of weak random non-blockingly. The
consequence is that the output of these different instances is
correlated to some extent. In the perfect scenario, the attacker
can control (or at least guess) the PID and clock of the
application, and drain the system's entropy pool to reduce the "up
to 16 bytes" above to 0. Then the dependencies of the initial
states of the pools are completely known. */
static int
read_seed_file (void)
{
int fd;
struct stat sb;
unsigned char buffer[POOLSIZE];
int n;
gcry_assert (pool_is_locked);
if (!seed_file_name)
return 0;
#ifdef HAVE_DOSISH_SYSTEM
fd = open( seed_file_name, O_RDONLY | O_BINARY );
#else
fd = open( seed_file_name, O_RDONLY );
#endif
if( fd == -1 && errno == ENOENT)
{
allow_seed_file_update = 1;
return 0;
}
if (fd == -1 )
{
log_info(_("can't open `%s': %s\n"), seed_file_name, strerror(errno) );
return 0;
}
if (lock_seed_file (fd, seed_file_name, 0))
{
close (fd);
return 0;
}
if (fstat( fd, &sb ) )
{
log_info(_("can't stat `%s': %s\n"), seed_file_name, strerror(errno) );
close(fd);
return 0;
}
if (!S_ISREG(sb.st_mode) )
{
log_info(_("`%s' is not a regular file - ignored\n"), seed_file_name );
close(fd);
return 0;
}
if (!sb.st_size )
{
log_info(_("note: random_seed file is empty\n") );
close(fd);
allow_seed_file_update = 1;
return 0;
}
if (sb.st_size != POOLSIZE )
{
log_info(_("warning: invalid size of random_seed file - not used\n") );
close(fd);
return 0;
}
do
{
n = read( fd, buffer, POOLSIZE );
}
while (n == -1 && errno == EINTR );
if (n != POOLSIZE)
{
log_fatal(_("can't read `%s': %s\n"), seed_file_name,strerror(errno) );
close(fd);/*NOTREACHED*/
return 0;
}
close(fd);
add_randomness( buffer, POOLSIZE, RANDOM_ORIGIN_INIT );
/* add some minor entropy to the pool now (this will also force a mixing) */
{
pid_t x = getpid();
add_randomness( &x, sizeof(x), RANDOM_ORIGIN_INIT );
}
{
time_t x = time(NULL);
add_randomness( &x, sizeof(x), RANDOM_ORIGIN_INIT );
}
{
clock_t x = clock();
add_randomness( &x, sizeof(x), RANDOM_ORIGIN_INIT );
}
/* And read a few bytes from our entropy source. By using a level
* of 0 this will not block and might not return anything with some
* entropy drivers, however the rndlinux driver will use
* /dev/urandom and return some stuff - Do not read too much as we
* want to be friendly to the scare system entropy resource. */
read_random_source ( RANDOM_ORIGIN_INIT, 16, GCRY_WEAK_RANDOM );
allow_seed_file_update = 1;
return 1;
}
void
_gcry_rngcsprng_update_seed_file (void)
{
unsigned long *sp, *dp;
int fd, i;
/* We do only a basic initialization so that we can lock the pool.
This is required to cope with the case that this function is
called by some cleanup code at a point where the RNG has never
been initialized. */
initialize_basics ();
lock_pool ();
if ( !seed_file_name || !rndpool || !pool_filled )
{
unlock_pool ();
return;
}
if ( !allow_seed_file_update )
{
unlock_pool ();
log_info(_("note: random_seed file not updated\n"));
return;
}
/* At this point we know that there is something in the pool and
thus we can conclude that the pool has been fully initialized. */
/* Copy the entropy pool to a scratch pool and mix both of them. */
for (i=0,dp=(unsigned long*)(void*)keypool, sp=(unsigned long*)(void*)rndpool;
i < POOLWORDS; i++, dp++, sp++ )
{
*dp = *sp + ADD_VALUE;
}
mix_pool(rndpool); rndstats.mixrnd++;
mix_pool(keypool); rndstats.mixkey++;
#if defined(HAVE_DOSISH_SYSTEM) || defined(__CYGWIN__)
fd = open (seed_file_name, O_WRONLY|O_CREAT|O_TRUNC|O_BINARY,
S_IRUSR|S_IWUSR );
#else
# if LOCK_SEED_FILE
fd = open (seed_file_name, O_WRONLY|O_CREAT, S_IRUSR|S_IWUSR );
# else
fd = open (seed_file_name, O_WRONLY|O_CREAT|O_TRUNC, S_IRUSR|S_IWUSR );
# endif
#endif
if (fd == -1 )
log_info (_("can't create `%s': %s\n"), seed_file_name, strerror(errno) );
else if (lock_seed_file (fd, seed_file_name, 1))
{
close (fd);
}
#if LOCK_SEED_FILE
else if (ftruncate (fd, 0))
{
log_info(_("can't write `%s': %s\n"), seed_file_name, strerror(errno));
close (fd);
}
#endif /*LOCK_SEED_FILE*/
else
{
do
{
i = write (fd, keypool, POOLSIZE );
}
while (i == -1 && errno == EINTR);
if (i != POOLSIZE)
log_info (_("can't write `%s': %s\n"),seed_file_name, strerror(errno));
if (close(fd))
log_info (_("can't close `%s': %s\n"),seed_file_name, strerror(errno));
}
unlock_pool ();
}
/* Read random out of the pool. This function is the core of the
public random functions. Note that Level GCRY_WEAK_RANDOM is not
anymore handled special and in fact is an alias in the API for
level GCRY_STRONG_RANDOM. Must be called with the pool already
locked. */
static void
read_pool (byte *buffer, size_t length, int level)
{
int i;
unsigned long *sp, *dp;
/* The volatile is there to make sure the compiler does not optimize
the code away in case the getpid function is badly attributed.
Note that we keep a pid in a static variable as well as in a
stack based one; the latter is to detect ill behaving thread
libraries, ignoring the pool mutexes. */
static volatile pid_t my_pid = (pid_t)(-1);
volatile pid_t my_pid2;
gcry_assert (pool_is_locked);
retry:
/* Get our own pid, so that we can detect a fork. */
my_pid2 = getpid ();
if (my_pid == (pid_t)(-1))
my_pid = my_pid2;
if ( my_pid != my_pid2 )
{
/* We detected a plain fork; i.e. we are now the child. Update
the static pid and add some randomness. */
pid_t x;
my_pid = my_pid2;
x = my_pid;
add_randomness (&x, sizeof(x), RANDOM_ORIGIN_INIT);
just_mixed = 0; /* Make sure it will get mixed. */
}
gcry_assert (pool_is_locked);
/* Our code does not allow to extract more than POOLSIZE. Better
check it here. */
if (length > POOLSIZE)
{
log_bug("too many random bits requested\n");
}
if (!pool_filled)
{
if (read_seed_file() )
pool_filled = 1;
}
/* For level 2 quality (key generation) we always make sure that the
pool has been seeded enough initially. */
if (level == GCRY_VERY_STRONG_RANDOM && !did_initial_extra_seeding)
{
size_t needed;
pool_balance = 0;
needed = length - pool_balance;
if (needed < 16) /* At least 128 bits. */
needed = 16;
else if( needed > POOLSIZE )
BUG ();
read_random_source (RANDOM_ORIGIN_EXTRAPOLL, needed,
GCRY_VERY_STRONG_RANDOM);
pool_balance += needed;
did_initial_extra_seeding = 1;
}
/* For level 2 make sure that there is enough random in the pool. */
if (level == GCRY_VERY_STRONG_RANDOM && pool_balance < length)
{
size_t needed;
if (pool_balance < 0)
pool_balance = 0;
needed = length - pool_balance;
if (needed > POOLSIZE)
BUG ();
read_random_source (RANDOM_ORIGIN_EXTRAPOLL, needed,
GCRY_VERY_STRONG_RANDOM);
pool_balance += needed;
}
/* Make sure the pool is filled. */
while (!pool_filled)
random_poll();
/* Always do a fast random poll (we have to use the unlocked version). */
do_fast_random_poll();
/* Mix the pid in so that we for sure won't deliver the same random
after a fork. */
{
pid_t apid = my_pid;
add_randomness (&apid, sizeof (apid), RANDOM_ORIGIN_INIT);
}
/* Mix the pool (if add_randomness() didn't it). */
if (!just_mixed)
{
mix_pool(rndpool);
rndstats.mixrnd++;
}
/* Create a new pool. */
for(i=0,dp=(unsigned long*)(void*)keypool, sp=(unsigned long*)(void*)rndpool;
i < POOLWORDS; i++, dp++, sp++ )
*dp = *sp + ADD_VALUE;
/* Mix both pools. */
mix_pool(rndpool); rndstats.mixrnd++;
mix_pool(keypool); rndstats.mixkey++;
/* Read the requested data. We use a read pointer to read from a
different position each time. */
while (length--)
{
*buffer++ = keypool[pool_readpos++];
if (pool_readpos >= POOLSIZE)
pool_readpos = 0;
pool_balance--;
}
if (pool_balance < 0)
pool_balance = 0;
/* Clear the keypool. */
memset (keypool, 0, POOLSIZE);
/* We need to detect whether a fork has happened. A fork might have
an identical pool and thus the child and the parent could emit
the very same random number. This test here is to detect forks
in a multi-threaded process. It does not work with all thread
implementations in particular not with pthreads. However it is
good enough for GNU Pth. */
if ( getpid () != my_pid2 )
{
pid_t x = getpid();
add_randomness (&x, sizeof(x), RANDOM_ORIGIN_INIT);
just_mixed = 0; /* Make sure it will get mixed. */
my_pid = x; /* Also update the static pid. */
goto retry;
}
}
/* Add LENGTH bytes of randomness from buffer to the pool. ORIGIN is
used to specify the randomness origin. This is one of the
RANDOM_ORIGIN_* values. */
static void
add_randomness (const void *buffer, size_t length, enum random_origins origin)
{
const unsigned char *p = buffer;
size_t count = 0;
gcry_assert (pool_is_locked);
rndstats.addbytes += length;
rndstats.naddbytes++;
while (length-- )
{
rndpool[pool_writepos++] ^= *p++;
count++;
if (pool_writepos >= POOLSIZE )
{
/* It is possible that we are invoked before the pool is
filled using an unreliable origin of entropy, for example
the fast random poll. To avoid flagging the pool as
filled in this case, we track the initial filling state
separately. See also the remarks about the seed file. */
if (origin >= RANDOM_ORIGIN_SLOWPOLL && !pool_filled)
{
pool_filled_counter += count;
count = 0;
if (pool_filled_counter >= POOLSIZE)
pool_filled = 1;
}
pool_writepos = 0;
mix_pool(rndpool); rndstats.mixrnd++;
just_mixed = !length;
}
}
}
static void
random_poll()
{
rndstats.slowpolls++;
read_random_source (RANDOM_ORIGIN_SLOWPOLL, POOLSIZE/5, GCRY_STRONG_RANDOM);
}
/* Runtime determination of the slow entropy gathering module. */
static int (*
getfnc_gather_random (void))(void (*)(const void*, size_t,
enum random_origins),
enum random_origins, size_t, int)
{
int (*fnc)(void (*)(const void*, size_t, enum random_origins),
enum random_origins, size_t, int);
#if USE_RNDLINUX
if ( !access (NAME_OF_DEV_RANDOM, R_OK)
&& !access (NAME_OF_DEV_URANDOM, R_OK))
{
fnc = _gcry_rndlinux_gather_random;
return fnc;
}
#endif
#if USE_RNDEGD
if ( _gcry_rndegd_connect_socket (1) != -1 )
{
fnc = _gcry_rndegd_gather_random;
return fnc;
}
#endif
#if USE_RNDUNIX
fnc = _gcry_rndunix_gather_random;
return fnc;
#endif
#if USE_RNDW32
fnc = _gcry_rndw32_gather_random;
return fnc;
#endif
#if USE_RNDW32CE
fnc = _gcry_rndw32ce_gather_random;
return fnc;
#endif
log_fatal (_("no entropy gathering module detected\n"));
return NULL; /*NOTREACHED*/
}
/* Runtime determination of the fast entropy gathering function.
(Currently a compile time method is used.) */
static void (*
getfnc_fast_random_poll (void))( void (*)(const void*, size_t,
enum random_origins),
enum random_origins)
{
#if USE_RNDW32
return _gcry_rndw32_gather_random_fast;
#endif
#if USE_RNDW32CE
return _gcry_rndw32ce_gather_random_fast;
#endif
return NULL;
}
static void
do_fast_random_poll (void)
{
gcry_assert (pool_is_locked);
rndstats.fastpolls++;
if (fast_gather_fnc)
fast_gather_fnc (add_randomness, RANDOM_ORIGIN_FASTPOLL);
/* Continue with the generic functions. */
#if HAVE_GETHRTIME
{
hrtime_t tv;
tv = gethrtime();
add_randomness( &tv, sizeof(tv), RANDOM_ORIGIN_FASTPOLL );
}
#elif HAVE_GETTIMEOFDAY
{
struct timeval tv;
if( gettimeofday( &tv, NULL ) )
BUG();
add_randomness( &tv.tv_sec, sizeof(tv.tv_sec), RANDOM_ORIGIN_FASTPOLL );
add_randomness( &tv.tv_usec, sizeof(tv.tv_usec), RANDOM_ORIGIN_FASTPOLL );
}
#elif HAVE_CLOCK_GETTIME
{ struct timespec tv;
if( clock_gettime( CLOCK_REALTIME, &tv ) == -1 )
BUG();
add_randomness( &tv.tv_sec, sizeof(tv.tv_sec), RANDOM_ORIGIN_FASTPOLL );
add_randomness( &tv.tv_nsec, sizeof(tv.tv_nsec), RANDOM_ORIGIN_FASTPOLL );
}
#else /* use times */
# ifndef HAVE_DOSISH_SYSTEM
{ struct tms buf;
times( &buf );
add_randomness( &buf, sizeof buf, RANDOM_ORIGIN_FASTPOLL );
}
# endif
#endif
#ifdef HAVE_GETRUSAGE
# ifdef RUSAGE_SELF
{
struct rusage buf;
/* QNX/Neutrino does return ENOSYS - so we just ignore it and add
whatever is in buf. In a chroot environment it might not work
at all (i.e. because /proc/ is not accessible), so we better
ignore all error codes and hope for the best. */
getrusage (RUSAGE_SELF, &buf );
add_randomness( &buf, sizeof buf, RANDOM_ORIGIN_FASTPOLL );
memset( &buf, 0, sizeof buf );
}
# else /*!RUSAGE_SELF*/
# ifdef __GCC__
# warning There is no RUSAGE_SELF on this system
# endif
# endif /*!RUSAGE_SELF*/
#endif /*HAVE_GETRUSAGE*/
/* Time and clock are available on all systems - so we better do it
just in case one of the above functions didn't work. */
{
time_t x = time(NULL);
add_randomness( &x, sizeof(x), RANDOM_ORIGIN_FASTPOLL );
}
{
clock_t x = clock();
add_randomness( &x, sizeof(x), RANDOM_ORIGIN_FASTPOLL );
}
/* If the system features a fast hardware RNG, read some bytes from
there. */
_gcry_rndhw_poll_fast (add_randomness, RANDOM_ORIGIN_FASTPOLL);
}
/* The fast random pool function as called at some places in
libgcrypt. This is merely a wrapper to make sure that this module
is initialized and to lock the pool. Note, that this function is a
NOP unless a random function has been used or _gcry_initialize (1)
has been used. We use this hack so that the internal use of this
function in cipher_open and md_open won't start filling up the
random pool, even if no random will be required by the process. */
void
_gcry_rngcsprng_fast_poll (void)
{
initialize_basics ();
lock_pool ();
if (rndpool)
{
/* Yes, we are fully initialized. */
do_fast_random_poll ();
}
unlock_pool ();
}
static void
read_random_source (enum random_origins origin, size_t length, int level)
{
if ( !slow_gather_fnc )
log_fatal ("Slow entropy gathering module not yet initialized\n");
if (slow_gather_fnc (add_randomness, origin, length, level) < 0)
log_fatal ("No way to gather entropy for the RNG\n");
}

File Metadata

Mime Type
text/x-diff
Expires
Sat, Dec 6, 11:57 PM (1 d, 2 h)
Storage Engine
local-disk
Storage Format
Raw Data
Storage Handle
a5/24/6637d8a451b7504579e77fdab40e

Event Timeline