ibphoenix/tomsfastmath
6
0
Fork 1
Code Issues Pull Requests Packages Releases Wiki Activity

Remove usage of malloc() from fp_prime_random_ex.

Browse Source 
This is the only instance of dynamic memory in TFM. It can be avoided
by writing directly into the dp[] array, and it does not violate C
aliasing rules which specifically allow access on objects through char*.
It does not matter the platform-specific representation of digits since
we are filling them with random data anyway.
This commit is contained in:
Giovanni Bajo 2011-09-21 17:45:18 +02:00
parent 08b3654ac4
commit 279da69658
2 changed files with 48 additions and 53 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
src/headers/tfm.h
View File

@ -1,10 +1,10 @@
/* TomsFastMath, a fast ISO C bignum library.
*
*
* This project is meant to fill in where LibTomMath
* falls short. That is speed ;-)
*
* This project is public domain and free for all purposes.
*
*
* Tom St Denis, tomstdenis@gmail.com
*/
#ifndef TFM_H_
@ -27,13 +27,13 @@
/* externally define this symbol to ignore the default settings, useful for changing the build from the make process */
#ifndef TFM_ALREADY_SET
/* do we want the large set of small multiplications ?
/* do we want the large set of small multiplications ?
Enable these if you are going to be doing a lot of small (<= 16 digit) multiplications say in ECC
Or if you're on a 64-bit machine doing RSA as a 1024-bit integer == 16 digits ;-)
*/
#define TFM_SMALL_SET
/* do we want huge code
/* do we want huge code
Enable these if you are doing 20, 24, 28, 32, 48, 64 digit multiplications (useful for RSA)
Less important on 64-bit machines as 32 digits == 2048 bits
*/
@ -81,7 +81,7 @@
/* #define TFM_PRESCOTT */
/* Do we want timing resistant fp_exptmod() ?
* This makes it slower but also timing invariant with respect to the exponent
* This makes it slower but also timing invariant with respect to the exponent
*/
/* #define TFM_TIMING_RESISTANT */
@ -106,7 +106,7 @@
/* autodetect x86-64 and make sure we are using 64-bit digits with x86-64 asm */
#if defined(__x86_64__)
#if defined(TFM_X86) || defined(TFM_SSE2) || defined(TFM_ARM)
#if defined(TFM_X86) || defined(TFM_SSE2) || defined(TFM_ARM)
#error x86-64 detected, x86-32/SSE2/ARM optimizations are not valid!
#endif
#if !defined(TFM_X86_64) && !defined(TFM_NO_ASM)
@ -121,7 +121,7 @@
/* try to detect x86-32 */
#if defined(__i386__) && !defined(TFM_SSE2)
#if defined(TFM_X86_64) || defined(TFM_ARM)
#if defined(TFM_X86_64) || defined(TFM_ARM)
#error x86-32 detected, x86-64/ARM optimizations are not valid!
#endif
#if !defined(TFM_X86) && !defined(TFM_NO_ASM)
@ -185,7 +185,7 @@
#undef TFM_PPC32
#undef TFM_PPC64
#undef TFM_AVR32
#undef TFM_ASM
#undef TFM_ASM
#endif
/* ECC helpers */
@ -249,6 +249,7 @@
#endif
typedef ulong64 fp_digit;
typedef unsigned long fp_word __attribute__ ((mode(TI)));
#define DIGIT_SHIFT 6
#else
/* this is to make porting into LibTomCrypt easier :-) */
#ifndef CRYPT
@ -262,6 +263,7 @@
#endif
typedef unsigned long fp_digit;
typedef ulong64 fp_word;
#define DIGIT_SHIFT 5
#endif
/* # of digits this is */
@ -290,7 +292,7 @@
/* a FP type */
typedef struct {
fp_digit dp[FP_SIZE];
int used,
int used,
sign;
} fp_int;
@ -434,9 +436,9 @@ int fp_isprime(fp_int *a);
/* callback for fp_prime_random, should fill dst with random bytes and return how many read [upto len] */
typedef int tfm_prime_callback(unsigned char *dst, int len, void *dat);
#define fp_prime_random(a, t, size, bbs, cb, dat) fp_prime_random_ex(a, t, ((size) * 8) + 1, (bbs==1)?TFM_PRIME_BBS:0, cb, dat)
#define fp_prime_random(a, size, bbs, cb, dat) fp_prime_random_ex(a, ((size) * 8) + 1, (bbs==1)?TFM_PRIME_BBS:0, cb, dat)
int fp_prime_random_ex(fp_int *a, int t, int size, int flags, tfm_prime_callback cb, void *dat);
int fp_prime_random_ex(fp_int *a, int size, int flags, tfm_prime_callback cb, void *dat);
/* radix conersions */
int fp_count_bits(fp_int *a);

77
src/numtheory/fp_prime_random_ex.c
View File

@ -1,22 +1,28 @@
/* TomsFastMath, a fast ISO C bignum library.
*
*
* This project is meant to fill in where LibTomMath
* falls short. That is speed ;-)
*
* This project is public domain and free for all purposes.
*
*
* Tom St Denis, tomstdenis@gmail.com
*/
#include <tfm.h>
#define fp_on_bitnum(a, bitnum) \
a->dp[(bitnum) >> DIGIT_SHIFT] |= 1 << ((bitnum) & (DIGIT_BIT-1));
#define fp_off_bitnum(a, bitnum) \
a->dp[(bitnum) >> DIGIT_SHIFT] &= ~(1 << ((bitnum) & (DIGIT_BIT-1)));
/* This is possibly the mother of all prime generation functions, muahahahahaha! */
int fp_prime_random_ex(fp_int *a, int t, int size, int flags, tfm_prime_callback cb, void *dat)
int fp_prime_random_ex(fp_int *a, int size, int flags, tfm_prime_callback cb, void *dat)
{
unsigned char *tmp, maskAND, maskOR_msb, maskOR_lsb;
int res, err, bsize, maskOR_msb_offset;
fp_digit maskAND_msb, maskOR_lsb;
int res, err, bsize, dsize;
/* sanity check the input */
if (size <= 1 || t <= 0) {
if (size <= 1) {
return FP_VAL;
}
@ -25,26 +31,11 @@ int fp_prime_random_ex(fp_int *a, int t, int size, int flags, tfm_prime_callback
flags |= TFM_PRIME_BBS;
}
/* calc the byte size */
bsize = (size>>3)+(size&7?1:0);
/* calc the digit size */
dsize = (size + DIGIT_BIT - 1) >> DIGIT_SHIFT;
/* we need a buffer of bsize bytes */
tmp = malloc(bsize);
if (tmp == NULL) {
return FP_MEM;
}
/* calc the maskAND value for the MSbyte*/
maskAND = 0xFF >> ((8 - (size & 7)) & 7);
/* calc the maskOR_msb */
maskOR_msb = 0;
maskOR_msb_offset = (size - 2) >> 3;
if (flags & TFM_PRIME_2MSB_ON) {
maskOR_msb |= 1 << ((size - 2) & 7);
} else if (flags & TFM_PRIME_2MSB_OFF) {
maskAND &= ~(1 << ((size - 2) & 7));
}
/* calc the maskAND value for the MSbyte */
maskAND_msb = FP_MASK >> ((DIGIT_BIT - (size & (DIGIT_BIT-1))) & (DIGIT_BIT-1));
/* get the maskOR_lsb */
maskOR_lsb = 1;
@ -54,21 +45,26 @@ int fp_prime_random_ex(fp_int *a, int t, int size, int flags, tfm_prime_callback
do {
/* read the bytes */
if (cb(tmp, bsize, dat) != bsize) {
err = FP_VAL;
goto error;
if (cb((unsigned char*)&a->dp[0], dsize*DIGIT_BIT, dat) != dsize*DIGIT_BIT) {
return FP_VAL;
}
/* work over the MSbyte */
tmp[0] &= maskAND;
tmp[0] |= 1 << ((size - 1) & 7);
a->used = dsize;
/* mix in the maskORs */
tmp[maskOR_msb_offset] |= maskOR_msb;
tmp[bsize-1] |= maskOR_lsb;
/* make sure the MSbyte has the required number of bits */
a->dp[dsize-1] &= maskAND_msb;
/* read it in */
fp_read_unsigned_bin(a, tmp, bsize);
/* modify the LSbyte as requested */
a->dp[0] |= maskOR_lsb;
/* turn on the MSbit to force the requested magnitude */
fp_on_bitnum(a, size-1);
/* modify the 2nd MSBit */
if (flags & TFM_PRIME_2MSB_ON) {
fp_on_bitnum(a, size-2);
} else if (flags & TFM_PRIME_2MSB_OFF) {
fp_off_bitnum(a, size-2);
}
/* is it prime? */
res = fp_isprime(a);
@ -78,7 +74,7 @@ int fp_prime_random_ex(fp_int *a, int t, int size, int flags, tfm_prime_callback
/* see if (a-1)/2 is prime */
fp_sub_d(a, 1, a);
fp_div_2(a, a);
/* is it prime? */
res = fp_isprime(a);
}
@ -90,10 +86,7 @@ int fp_prime_random_ex(fp_int *a, int t, int size, int flags, tfm_prime_callback
fp_add_d(a, 1, a);
}
err = FP_OKAY;
error:
free(tmp);
return err;
return FP_OKAY;
}
/* $Source$ */