/* Set U into W and release U. If W is NULL only U will be released. */
void
_gcry_mpi_snatch (gcry_mpi_t w, gcry_mpi_t u)
{
if (w)
{
if (mpi_is_immutable (w))
{
mpi_immutable_failed ();
return;
}
_gcry_mpi_assign_limb_space (w, u->d, u->alloced);
w->nlimbs = u->nlimbs;
w->sign = u->sign;
w->flags = u->flags;
u->alloced = 0;
u->nlimbs = 0;
u->d = NULL;
}
_gcry_mpi_free (u);
}
/****************
* RES = BASE ^ EXPO mod MOD
*/
void
gcry_mpi_powm( gcry_mpi_t res, gcry_mpi_t base, gcry_mpi_t expo, gcry_mpi_t mod)
{
mpi_ptr_t rp, ep, mp, bp;
mpi_size_t esize, msize, bsize, rsize;
int esign, msign, bsign, rsign;
int esec, msec, bsec, rsec;
mpi_size_t size;
int mod_shift_cnt;
int negative_result;
mpi_ptr_t mp_marker=NULL, bp_marker=NULL, ep_marker=NULL;
mpi_ptr_t xp_marker=NULL;
unsigned int mp_nlimbs = 0, bp_nlimbs = 0, ep_nlimbs = 0;
unsigned int xp_nlimbs = 0;
int assign_rp = 0;
mpi_ptr_t tspace = NULL;
mpi_size_t tsize=0; /* to avoid compiler warning */
/* fixme: we should check that the warning is void*/
esize = expo->nlimbs;
msize = mod->nlimbs;
size = 2 * msize;
esign = expo->sign;
msign = mod->sign;
esec = mpi_is_secure(expo);
msec = mpi_is_secure(mod);
bsec = mpi_is_secure(base);
rsec = mpi_is_secure(res);
rp = res->d;
ep = expo->d;
if( !msize )
msize = 1 / msize; /* provoke a signal */
if( !esize ) {
/* Exponent is zero, result is 1 mod MOD, i.e., 1 or 0
* depending on if MOD equals 1. */
rp[0] = 1;
res->nlimbs = (msize == 1 && mod->d[0] == 1) ? 0 : 1;
res->sign = 0;
goto leave;
}
/* Normalize MOD (i.e. make its most significant bit set) as required by
* mpn_divrem. This will make the intermediate values in the calculation
* slightly larger, but the correct result is obtained after a final
* reduction using the original MOD value. */
mp_nlimbs = msec? msize:0;
mp = mp_marker = mpi_alloc_limb_space(msize, msec);
count_leading_zeros( mod_shift_cnt, mod->d[msize-1] );
if( mod_shift_cnt )
_gcry_mpih_lshift( mp, mod->d, msize, mod_shift_cnt );
else
MPN_COPY( mp, mod->d, msize );
bsize = base->nlimbs;
bsign = base->sign;
if( bsize > msize ) { /* The base is larger than the module. Reduce it. */
/* Allocate (BSIZE + 1) with space for remainder and quotient.
* (The quotient is (bsize - msize + 1) limbs.) */
bp_nlimbs = bsec ? (bsize + 1):0;
bp = bp_marker = mpi_alloc_limb_space( bsize + 1, bsec );
MPN_COPY( bp, base->d, bsize );
/* We don't care about the quotient, store it above the remainder,
* at BP + MSIZE. */
_gcry_mpih_divrem( bp + msize, 0, bp, bsize, mp, msize );
bsize = msize;
/* Canonicalize the base, since we are going to multiply with it
* quite a few times. */
MPN_NORMALIZE( bp, bsize );
}
else
bp = base->d;
if( !bsize ) {
res->nlimbs = 0;
res->sign = 0;
goto leave;
}
if( res->alloced < size ) {
/* We have to allocate more space for RES. If any of the input
* parameters are identical to RES, defer deallocation of the old
* space. */
if( rp == ep || rp == mp || rp == bp ) {
rp = mpi_alloc_limb_space( size, rsec );
assign_rp = 1;
}
else {
mpi_resize( res, size );
rp = res->d;
}
}
else { /* Make BASE, EXPO and MOD not overlap with RES. */
if( rp == bp ) {
/* RES and BASE are identical. Allocate temp. space for BASE. */
assert( !bp_marker );
bp_nlimbs = bsec? bsize:0;
bp = bp_marker = mpi_alloc_limb_space( bsize, bsec );
MPN_COPY(bp, rp, bsize);
}
if( rp == ep ) {
/* RES and EXPO are identical. Allocate temp. space for EXPO. */
ep_nlimbs = esec? esize:0;
ep = ep_marker = mpi_alloc_limb_space( esize, esec );
MPN_COPY(ep, rp, esize);
}
if( rp == mp ) {
/* RES and MOD are identical. Allocate temporary space for MOD.*/
assert( !mp_marker );
mp_nlimbs = msec?msize:0;
mp = mp_marker = mpi_alloc_limb_space( msize, msec );
MPN_COPY(mp, rp, msize);
}
}
MPN_COPY( rp, bp, bsize );
rsize = bsize;
rsign = bsign;
{
mpi_size_t i;
mpi_ptr_t xp;
int c;
mpi_limb_t e;
mpi_limb_t carry_limb;
struct karatsuba_ctx karactx;
xp_nlimbs = msec? (2 * (msize + 1)):0;
xp = xp_marker = mpi_alloc_limb_space( 2 * (msize + 1), msec );
memset( &karactx, 0, sizeof karactx );
negative_result = (ep[0] & 1) && base->sign;
i = esize - 1;
e = ep[i];
count_leading_zeros (c, e);
e = (e << c) << 1; /* shift the expo bits to the left, lose msb */
c = BITS_PER_MPI_LIMB - 1 - c;
/* Main loop.
*
* Make the result be pointed to alternately by XP and RP. This
* helps us avoid block copying, which would otherwise be necessary
* with the overlap restrictions of _gcry_mpih_divmod. With 50% probability
* the result after this loop will be in the area originally pointed
* by RP (==RES->d), and with 50% probability in the area originally
* pointed to by XP.
*/
for(;;) {
while( c ) {
mpi_ptr_t tp;
mpi_size_t xsize;
/*mpih_mul_n(xp, rp, rp, rsize);*/
if( rsize < KARATSUBA_THRESHOLD )
_gcry_mpih_sqr_n_basecase( xp, rp, rsize );
else {
if( !tspace ) {
tsize = 2 * rsize;
tspace = mpi_alloc_limb_space( tsize, 0 );
}
else if( tsize < (2*rsize) ) {
_gcry_mpi_free_limb_space (tspace, 0);
tsize = 2 * rsize;
tspace = mpi_alloc_limb_space( tsize, 0 );
}
_gcry_mpih_sqr_n( xp, rp, rsize, tspace );
}
xsize = 2 * rsize;
if( xsize > msize ) {
_gcry_mpih_divrem(xp + msize, 0, xp, xsize, mp, msize);
xsize = msize;
}
tp = rp; rp = xp; xp = tp;
rsize = xsize;
if( (mpi_limb_signed_t)e < 0 ) {
/*mpih_mul( xp, rp, rsize, bp, bsize );*/
if( bsize < KARATSUBA_THRESHOLD ) {
_gcry_mpih_mul( xp, rp, rsize, bp, bsize );
}
else {
_gcry_mpih_mul_karatsuba_case(
xp, rp, rsize, bp, bsize, &karactx );
}
xsize = rsize + bsize;
if( xsize > msize ) {
_gcry_mpih_divrem(xp + msize, 0, xp, xsize, mp, msize);
xsize = msize;
}
tp = rp; rp = xp; xp = tp;
rsize = xsize;
}
e <<= 1;
c--;
}
i--;
if( i < 0 )
break;
e = ep[i];
c = BITS_PER_MPI_LIMB;
}
/* We shifted MOD, the modulo reduction argument, left MOD_SHIFT_CNT
* steps. Adjust the result by reducing it with the original MOD.
*
* Also make sure the result is put in RES->d (where it already
* might be, see above).
*/
if( mod_shift_cnt ) {
carry_limb = _gcry_mpih_lshift( res->d, rp, rsize, mod_shift_cnt);
rp = res->d;
if( carry_limb ) {
rp[rsize] = carry_limb;
rsize++;
}
}
else {
MPN_COPY( res->d, rp, rsize);
rp = res->d;
}
if( rsize >= msize ) {
_gcry_mpih_divrem(rp + msize, 0, rp, rsize, mp, msize);
rsize = msize;
}
/* Remove any leading zero words from the result. */
if( mod_shift_cnt )
_gcry_mpih_rshift( rp, rp, rsize, mod_shift_cnt);
MPN_NORMALIZE (rp, rsize);
_gcry_mpih_release_karatsuba_ctx( &karactx );
}
if( negative_result && rsize ) {
if( mod_shift_cnt )
_gcry_mpih_rshift( mp, mp, msize, mod_shift_cnt);
_gcry_mpih_sub( rp, mp, msize, rp, rsize);
rsize = msize;
rsign = msign;
MPN_NORMALIZE(rp, rsize);
}
res->nlimbs = rsize;
res->sign = rsign;
leave:
if( assign_rp ) _gcry_mpi_assign_limb_space( res, rp, size );
if( mp_marker ) _gcry_mpi_free_limb_space( mp_marker, mp_nlimbs );
if( bp_marker ) _gcry_mpi_free_limb_space( bp_marker, bp_nlimbs );
if( ep_marker ) _gcry_mpi_free_limb_space( ep_marker, ep_nlimbs );
if( xp_marker ) _gcry_mpi_free_limb_space( xp_marker, xp_nlimbs );
if( tspace ) _gcry_mpi_free_limb_space( tspace, 0 );
}
void
gcry_mpi_mul( gcry_mpi_t w, gcry_mpi_t u, gcry_mpi_t v)
{
mpi_size_t usize, vsize, wsize;
mpi_ptr_t up, vp, wp;
mpi_limb_t cy;
int usign, vsign, usecure, vsecure, sign_product;
int assign_wp=0;
mpi_ptr_t tmp_limb=NULL;
unsigned int tmp_limb_nlimbs = 0;
if( u->nlimbs < v->nlimbs ) { /* Swap U and V. */
usize = v->nlimbs;
usign = v->sign;
usecure = mpi_is_secure(v);
up = v->d;
vsize = u->nlimbs;
vsign = u->sign;
vsecure = mpi_is_secure(u);
vp = u->d;
}
else {
usize = u->nlimbs;
usign = u->sign;
usecure = mpi_is_secure(u);
up = u->d;
vsize = v->nlimbs;
vsign = v->sign;
vsecure = mpi_is_secure(v);
vp = v->d;
}
sign_product = usign ^ vsign;
wp = w->d;
/* Ensure W has space enough to store the result. */
wsize = usize + vsize;
if ( !mpi_is_secure (w) && (mpi_is_secure (u) || mpi_is_secure (v)) ) {
/* w is not allocated in secure space but u or v is. To make sure
* that no temporray results are stored in w, we temporary use
* a newly allocated limb space for w */
wp = mpi_alloc_limb_space( wsize, 1 );
assign_wp = 2; /* mark it as 2 so that we can later copy it back to
* mormal memory */
}
else if( w->alloced < wsize ) {
if( wp == up || wp == vp ) {
wp = mpi_alloc_limb_space( wsize, mpi_is_secure(w) );
assign_wp = 1;
}
else {
mpi_resize(w, wsize );
wp = w->d;
}
}
else { /* Make U and V not overlap with W. */
if( wp == up ) {
/* W and U are identical. Allocate temporary space for U. */
tmp_limb_nlimbs = usize;
up = tmp_limb = mpi_alloc_limb_space( usize, usecure );
/* Is V identical too? Keep it identical with U. */
if( wp == vp )
vp = up;
/* Copy to the temporary space. */
MPN_COPY( up, wp, usize );
}
else if( wp == vp ) {
/* W and V are identical. Allocate temporary space for V. */
tmp_limb_nlimbs = vsize;
vp = tmp_limb = mpi_alloc_limb_space( vsize, vsecure );
/* Copy to the temporary space. */
MPN_COPY( vp, wp, vsize );
}
}
if( !vsize )
wsize = 0;
else {
cy = _gcry_mpih_mul( wp, up, usize, vp, vsize );
wsize -= cy? 0:1;
}
if( assign_wp ) {
if (assign_wp == 2) {
/* copy the temp wp from secure memory back to normal memory */
mpi_ptr_t tmp_wp = mpi_alloc_limb_space (wsize, 0);
MPN_COPY (tmp_wp, wp, wsize);
_gcry_mpi_free_limb_space (wp, 0);
wp = tmp_wp;
}
_gcry_mpi_assign_limb_space( w, wp, wsize );
}
w->nlimbs = wsize;
w->sign = sign_product;
if( tmp_limb )
_gcry_mpi_free_limb_space (tmp_limb, tmp_limb_nlimbs);
}
