+ if (!BN_from_montgomery(rr,r,mont,ctx)) goto err;
+ ret=1;
+err:
+ if ((in_mont == NULL) && (mont != NULL)) BN_MONT_CTX_free(mont);
+ BN_CTX_end(ctx);
+ bn_check_top(rr);
+ return(ret);
+ }
+
+
+/* BN_mod_exp_mont_consttime() stores the precomputed powers in a specific layout
+ * so that accessing any of these table values shows the same access pattern as far
+ * as cache lines are concerned. The following functions are used to transfer a BIGNUM
+ * from/to that table. */
+
+static int MOD_EXP_CTIME_COPY_TO_PREBUF(const BIGNUM *b, int top, unsigned char *buf, int idx, int width)
+ {
+ size_t i, j;
+
+ if (top > b->top)
+ top = b->top; /* this works because 'buf' is explicitly zeroed */
+ for (i = 0, j=idx; i < top * sizeof b->d[0]; i++, j+=width)
+ {
+ buf[j] = ((unsigned char*)b->d)[i];
+ }
+
+ return 1;
+ }
+
+static int MOD_EXP_CTIME_COPY_FROM_PREBUF(BIGNUM *b, int top, unsigned char *buf, int idx, int width)
+ {
+ size_t i, j;
+
+ if (bn_wexpand(b, top) == NULL)
+ return 0;
+
+ for (i=0, j=idx; i < top * sizeof b->d[0]; i++, j+=width)
+ {
+ ((unsigned char*)b->d)[i] = buf[j];
+ }
+
+ b->top = top;
+ bn_correct_top(b);
+ return 1;
+ }
+
+/* Given a pointer value, compute the next address that is a cache line multiple. */
+#define MOD_EXP_CTIME_ALIGN(x_) \
+ ((unsigned char*)(x_) + (MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH - (((size_t)(x_)) & (MOD_EXP_CTIME_MIN_CACHE_LINE_MASK))))
+
+/* This variant of BN_mod_exp_mont() uses fixed windows and the special
+ * precomputation memory layout to limit data-dependency to a minimum
+ * to protect secret exponents (cf. the hyper-threading timing attacks
+ * pointed out by Colin Percival,
+ * http://www.daemonology.net/hyperthreading-considered-harmful/)
+ */
+int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
+ const BIGNUM *m, BN_CTX *ctx, BN_MONT_CTX *in_mont)
+ {
+ int i,bits,ret=0,window,wvalue;
+ int top;
+ BN_MONT_CTX *mont=NULL;
+
+ int numPowers;
+ unsigned char *powerbufFree=NULL;
+ int powerbufLen = 0;
+ unsigned char *powerbuf=NULL;
+ BIGNUM tmp, am;
+
+ bn_check_top(a);
+ bn_check_top(p);
+ bn_check_top(m);
+
+ top = m->top;
+
+ if (!(m->d[0] & 1))
+ {
+ BNerr(BN_F_BN_MOD_EXP_MONT_CONSTTIME,BN_R_CALLED_WITH_EVEN_MODULUS);
+ return(0);
+ }
+ bits=BN_num_bits(p);
+ if (bits == 0)
+ {
+ ret = BN_one(rr);
+ return ret;
+ }
+
+ BN_CTX_start(ctx);
+
+ /* Allocate a montgomery context if it was not supplied by the caller.
+ * If this is not done, things will break in the montgomery part.
+ */
+ if (in_mont != NULL)
+ mont=in_mont;
+ else
+ {
+ if ((mont=BN_MONT_CTX_new()) == NULL) goto err;
+ if (!BN_MONT_CTX_set(mont,m,ctx)) goto err;
+ }
+
+ /* Get the window size to use with size of p. */
+ window = BN_window_bits_for_ctime_exponent_size(bits);
+#if defined(OPENSSL_BN_ASM_MONT5)
+ if (window==6 && bits<=1024) window=5; /* ~5% improvement of 2048-bit RSA sign */
+#endif
+
+ /* Allocate a buffer large enough to hold all of the pre-computed
+ * powers of am, am itself and tmp.
+ */
+ numPowers = 1 << window;
+ powerbufLen = sizeof(m->d[0])*(top*numPowers +
+ ((2*top)>numPowers?(2*top):numPowers));
+#ifdef alloca
+ if (powerbufLen < 3072)
+ powerbufFree = alloca(powerbufLen+MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH);
+ else
+#endif
+ if ((powerbufFree=(unsigned char*)OPENSSL_malloc(powerbufLen+MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH)) == NULL)
+ goto err;
+
+ powerbuf = MOD_EXP_CTIME_ALIGN(powerbufFree);
+ memset(powerbuf, 0, powerbufLen);
+
+#ifdef alloca
+ if (powerbufLen < 3072)
+ powerbufFree = NULL;
+#endif
+
+ /* lay down tmp and am right after powers table */
+ tmp.d = (BN_ULONG *)(powerbuf + sizeof(m->d[0])*top*numPowers);
+ am.d = tmp.d + top;
+ tmp.top = am.top = 0;
+ tmp.dmax = am.dmax = top;
+ tmp.neg = am.neg = 0;
+ tmp.flags = am.flags = BN_FLG_STATIC_DATA;
+
+ /* prepare a^0 in Montgomery domain */
+#if 1
+ if (!BN_to_montgomery(&tmp,BN_value_one(),mont,ctx)) goto err;
+#else
+ tmp.d[0] = (0-m->d[0])&BN_MASK2; /* 2^(top*BN_BITS2) - m */
+ for (i=1;i<top;i++)
+ tmp.d[i] = (~m->d[i])&BN_MASK2;
+ tmp.top = top;
+#endif
+
+ /* prepare a^1 in Montgomery domain */
+ if (a->neg || BN_ucmp(a,m) >= 0)
+ {
+ if (!BN_mod(&am,a,m,ctx)) goto err;
+ if (!BN_to_montgomery(&am,&am,mont,ctx)) goto err;
+ }
+ else if (!BN_to_montgomery(&am,a,mont,ctx)) goto err;
+
+#if defined(OPENSSL_BN_ASM_MONT5)
+ /* This optimization uses ideas from http://eprint.iacr.org/2011/239,
+ * specifically optimization of cache-timing attack countermeasures
+ * and pre-computation optimization. */
+
+ /* Dedicated window==4 case improves 512-bit RSA sign by ~15%, but as
+ * 512-bit RSA is hardly relevant, we omit it to spare size... */
+ if (window==5)
+ {
+ void bn_mul_mont_gather5(BN_ULONG *rp,const BN_ULONG *ap,
+ const void *table,const BN_ULONG *np,
+ const BN_ULONG *n0,int num,int power);
+ void bn_scatter5(const BN_ULONG *inp,size_t num,
+ void *table,size_t power);
+ void bn_gather5(BN_ULONG *out,size_t num,
+ void *table,size_t power);
+
+ BN_ULONG *np=mont->N.d, *n0=mont->n0;
+
+ /* BN_to_montgomery can contaminate words above .top
+ * [in BN_DEBUG[_DEBUG] build]... */
+ for (i=am.top; i<top; i++) am.d[i]=0;
+ for (i=tmp.top; i<top; i++) tmp.d[i]=0;
+
+ bn_scatter5(tmp.d,top,powerbuf,0);
+ bn_scatter5(am.d,am.top,powerbuf,1);
+ bn_mul_mont(tmp.d,am.d,am.d,np,n0,top);
+ bn_scatter5(tmp.d,top,powerbuf,2);
+
+#if 0
+ for (i=3; i<32; i++)
+ {
+ /* Calculate a^i = a^(i-1) * a */
+ bn_mul_mont_gather5(tmp.d,am.d,powerbuf,np,n0,top,i-1);
+ bn_scatter5(tmp.d,top,powerbuf,i);
+ }
+#else
+ /* same as above, but uses squaring for 1/2 of operations */
+ for (i=4; i<32; i*=2)
+ {
+ bn_mul_mont(tmp.d,tmp.d,tmp.d,np,n0,top);
+ bn_scatter5(tmp.d,top,powerbuf,i);
+ }
+ for (i=3; i<8; i+=2)
+ {
+ int j;
+ bn_mul_mont_gather5(tmp.d,am.d,powerbuf,np,n0,top,i-1);
+ bn_scatter5(tmp.d,top,powerbuf,i);
+ for (j=2*i; j<32; j*=2)
+ {
+ bn_mul_mont(tmp.d,tmp.d,tmp.d,np,n0,top);
+ bn_scatter5(tmp.d,top,powerbuf,j);
+ }
+ }
+ for (; i<16; i+=2)
+ {
+ bn_mul_mont_gather5(tmp.d,am.d,powerbuf,np,n0,top,i-1);
+ bn_scatter5(tmp.d,top,powerbuf,i);
+ bn_mul_mont(tmp.d,tmp.d,tmp.d,np,n0,top);
+ bn_scatter5(tmp.d,top,powerbuf,2*i);
+ }
+ for (; i<32; i+=2)
+ {
+ bn_mul_mont_gather5(tmp.d,am.d,powerbuf,np,n0,top,i-1);
+ bn_scatter5(tmp.d,top,powerbuf,i);
+ }
+#endif
+ bits--;
+ for (wvalue=0, i=bits%5; i>=0; i--,bits--)
+ wvalue = (wvalue<<1)+BN_is_bit_set(p,bits);
+ bn_gather5(tmp.d,top,powerbuf,wvalue);
+
+ /* Scan the exponent one window at a time starting from the most
+ * significant bits.
+ */
+ while (bits >= 0)
+ {
+ for (wvalue=0, i=0; i<5; i++,bits--)
+ wvalue = (wvalue<<1)+BN_is_bit_set(p,bits);
+
+ bn_mul_mont(tmp.d,tmp.d,tmp.d,np,n0,top);
+ bn_mul_mont(tmp.d,tmp.d,tmp.d,np,n0,top);
+ bn_mul_mont(tmp.d,tmp.d,tmp.d,np,n0,top);
+ bn_mul_mont(tmp.d,tmp.d,tmp.d,np,n0,top);
+ bn_mul_mont(tmp.d,tmp.d,tmp.d,np,n0,top);
+ bn_mul_mont_gather5(tmp.d,tmp.d,powerbuf,np,n0,top,wvalue);
+ }
+
+ tmp.top=top;
+ bn_correct_top(&tmp);
+ }
+ else
+#endif
+ {
+ if (!MOD_EXP_CTIME_COPY_TO_PREBUF(&tmp, top, powerbuf, 0, numPowers)) goto err;
+ if (!MOD_EXP_CTIME_COPY_TO_PREBUF(&am, top, powerbuf, 1, numPowers)) goto err;
+
+ /* If the window size is greater than 1, then calculate
+ * val[i=2..2^winsize-1]. Powers are computed as a*a^(i-1)
+ * (even powers could instead be computed as (a^(i/2))^2
+ * to use the slight performance advantage of sqr over mul).
+ */
+ if (window > 1)
+ {
+ if (!BN_mod_mul_montgomery(&tmp,&am,&am,mont,ctx)) goto err;
+ if (!MOD_EXP_CTIME_COPY_TO_PREBUF(&tmp, top, powerbuf, 2, numPowers)) goto err;
+ for (i=3; i<numPowers; i++)
+ {
+ /* Calculate a^i = a^(i-1) * a */
+ if (!BN_mod_mul_montgomery(&tmp,&am,&tmp,mont,ctx))
+ goto err;
+ if (!MOD_EXP_CTIME_COPY_TO_PREBUF(&tmp, top, powerbuf, i, numPowers)) goto err;
+ }
+ }
+
+ bits--;
+ for (wvalue=0, i=bits%window; i>=0; i--,bits--)
+ wvalue = (wvalue<<1)+BN_is_bit_set(p,bits);
+ if (!MOD_EXP_CTIME_COPY_FROM_PREBUF(&tmp,top,powerbuf,wvalue,numPowers)) goto err;
+
+ /* Scan the exponent one window at a time starting from the most
+ * significant bits.
+ */
+ while (bits >= 0)
+ {
+ wvalue=0; /* The 'value' of the window */
+
+ /* Scan the window, squaring the result as we go */
+ for (i=0; i<window; i++,bits--)
+ {
+ if (!BN_mod_mul_montgomery(&tmp,&tmp,&tmp,mont,ctx)) goto err;
+ wvalue = (wvalue<<1)+BN_is_bit_set(p,bits);
+ }
+
+ /* Fetch the appropriate pre-computed value from the pre-buf */
+ if (!MOD_EXP_CTIME_COPY_FROM_PREBUF(&am, top, powerbuf, wvalue, numPowers)) goto err;
+
+ /* Multiply the result into the intermediate result */
+ if (!BN_mod_mul_montgomery(&tmp,&tmp,&am,mont,ctx)) goto err;
+ }
+ }
+
+ /* Convert the final result from montgomery to standard format */
+ if (!BN_from_montgomery(rr,&tmp,mont,ctx)) goto err;