1 files changed, 196 insertions, 140 deletions

diff --git a/src/lib/libcrypto/bn/bn_lib.c b/src/lib/libcrypto/bn/bn_lib.c
index 7767d65170..a016cb7f53 100644
--- a/src/lib/libcrypto/bn/bn_lib.c
+++ b/src/lib/libcrypto/bn/bn_lib.c
@@ -128,7 +128,7 @@ int BN_get_params(int which)
         else return(0);
         }
 
-BIGNUM *BN_value_one(void)
+const BIGNUM *BN_value_one(void)
         {
         static BN_ULONG data_one=1L;
         static BIGNUM const_one={&data_one,1,1,0};
@@ -305,172 +305,168 @@ BIGNUM *BN_new(void)
         return(ret);
         }
 
-/* This is an internal function that should not be used in applications.
- * It ensures that 'b' has enough room for a 'words' word number number.
- * It is mostly used by the various BIGNUM routines. If there is an error,
- * NULL is returned. If not, 'b' is returned. */
-
-BIGNUM *bn_expand2(BIGNUM *b, int words)
+/* This is used both by bn_expand2() and bn_dup_expand() */
+/* The caller MUST check that words > b->dmax before calling this */
+static BN_ULONG *bn_expand_internal(const BIGNUM *b, int words)
         {
-        BN_ULONG *A,*a;
+        BN_ULONG *A,*a = NULL;
         const BN_ULONG *B;
         int i;
 
-        bn_check_top(b);
+        if (words > (INT_MAX/(4*BN_BITS2)))
+                {
+                BNerr(BN_F_BN_EXPAND_INTERNAL,BN_R_BIGNUM_TOO_LONG);
+                return NULL;
+                }
 
-        if (words > b->dmax)
+        bn_check_top(b);        
+        if (BN_get_flags(b,BN_FLG_STATIC_DATA))
                 {
-                if (words > (INT_MAX/(4*BN_BITS2)))
-                        {
-                        BNerr(BN_F_BN_EXPAND2,BN_R_BIGNUM_TOO_LONG);
-                        return NULL;
-                        }
-                        
-                bn_check_top(b);        
-                if (BN_get_flags(b,BN_FLG_STATIC_DATA))
+                BNerr(BN_F_BN_EXPAND_INTERNAL,BN_R_EXPAND_ON_STATIC_BIGNUM_DATA);
+                return(NULL);
+                }
+        a=A=(BN_ULONG *)OPENSSL_malloc(sizeof(BN_ULONG)*(words+1));
+        if (A == NULL)
+                {
+                BNerr(BN_F_BN_EXPAND_INTERNAL,ERR_R_MALLOC_FAILURE);
+                return(NULL);
+                }
+#if 1
+        B=b->d;
+        /* Check if the previous number needs to be copied */
+        if (B != NULL)
+                {
+                for (i=b->top>>2; i>0; i--,A+=4,B+=4)
                         {
-                        BNerr(BN_F_BN_EXPAND2,BN_R_EXPAND_ON_STATIC_BIGNUM_DATA);
-                        return(NULL);
+                        /*
+                         * The fact that the loop is unrolled
+                         * 4-wise is a tribute to Intel. It's
+                         * the one that doesn't have enough
+                         * registers to accomodate more data.
+                         * I'd unroll it 8-wise otherwise:-)
+                         *
+                         *              <appro@fy.chalmers.se>
+                         */
+                        BN_ULONG a0,a1,a2,a3;
+                        a0=B[0]; a1=B[1]; a2=B[2]; a3=B[3];
+                        A[0]=a0; A[1]=a1; A[2]=a2; A[3]=a3;
                         }
-                a=A=(BN_ULONG *)OPENSSL_malloc(sizeof(BN_ULONG)*(words+1));
-                if (A == NULL)
+                switch (b->top&3)
                         {
-                        BNerr(BN_F_BN_EXPAND2,ERR_R_MALLOC_FAILURE);
-                        return(NULL);
+                case 3: A[2]=B[2];
+                case 2: A[1]=B[1];
+                case 1: A[0]=B[0];
+                case 0: /* workaround for ultrix cc: without 'case 0', the optimizer does
+                         * the switch table by doing a=top&3; a--; goto jump_table[a];
+                         * which fails for top== 0 */
+                        ;
                         }
-#if 1
-                B=b->d;
-                /* Check if the previous number needs to be copied */
-                if (B != NULL)
-                        {
-#if 0
-                        /* This lot is an unrolled loop to copy b->top 
-                         * BN_ULONGs from B to A
-                         */
-/*
- * I have nothing against unrolling but it's usually done for
- * several reasons, namely:
- * - minimize percentage of decision making code, i.e. branches;
- * - avoid cache trashing;
- * - make it possible to schedule loads earlier;
- * Now let's examine the code below. The cornerstone of C is
- * "programmer is always right" and that's what we love it for:-)
- * For this very reason C compilers have to be paranoid when it
- * comes to data aliasing and assume the worst. Yeah, but what
- * does it mean in real life? This means that loop body below will
- * be compiled to sequence of loads immediately followed by stores
- * as compiler assumes the worst, something in A==B+1 style. As a
- * result CPU pipeline is going to starve for incoming data. Secondly
- * if A and B happen to share same cache line such code is going to
- * cause severe cache trashing. Both factors have severe impact on
- * performance of modern CPUs and this is the reason why this
- * particular piece of code is #ifdefed away and replaced by more
- * "friendly" version found in #else section below. This comment
- * also applies to BN_copy function.
- *
- *                                      <appro@fy.chalmers.se>
- */
-                        for (i=b->top&(~7); i>0; i-=8)
-                                {
-                                A[0]=B[0]; A[1]=B[1]; A[2]=B[2]; A[3]=B[3];
-                                A[4]=B[4]; A[5]=B[5]; A[6]=B[6]; A[7]=B[7];
-                                A+=8;
-                                B+=8;
-                                }
-                        switch (b->top&7)
-                                {
-                        case 7:
-                                A[6]=B[6];
-                        case 6:
-                                A[5]=B[5];
-                        case 5:
-                                A[4]=B[4];
-                        case 4:
-                                A[3]=B[3];
-                        case 3:
-                                A[2]=B[2];
-                        case 2:
-                                A[1]=B[1];
-                        case 1:
-                                A[0]=B[0];
-                        case 0:
-                                /* I need the 'case 0' entry for utrix cc.
-                                 * If the optimizer is turned on, it does the
-                                 * switch table by doing
-                                 * a=top&7
-                                 * a--;
-                                 * goto jump_table[a];
-                                 * If top is 0, this makes us jump to 0xffffffc 
-                                 * which is rather bad :-(.
-                                 * eric 23-Apr-1998
-                                 */
-                                ;
-                                }
+                }
+
+        /* Now need to zero any data between b->top and b->max */
+        /* XXX Why? */
+
+        A= &(a[b->top]);
+        for (i=(words - b->top)>>3; i>0; i--,A+=8)
+                {
+                A[0]=0; A[1]=0; A[2]=0; A[3]=0;
+                A[4]=0; A[5]=0; A[6]=0; A[7]=0;
+                }
+        for (i=(words - b->top)&7; i>0; i--,A++)
+                A[0]=0;
 #else
-                        for (i=b->top>>2; i>0; i--,A+=4,B+=4)
+        memset(A,0,sizeof(BN_ULONG)*(words+1));
+        memcpy(A,b->d,sizeof(b->d[0])*b->top);
+#endif
+                
+        return(a);
+        }
+
+/* This is an internal function that can be used instead of bn_expand2()
+ * when there is a need to copy BIGNUMs instead of only expanding the
+ * data part, while still expanding them.
+ * Especially useful when needing to expand BIGNUMs that are declared
+ * 'const' and should therefore not be changed.
+ * The reason to use this instead of a BN_dup() followed by a bn_expand2()
+ * is memory allocation overhead.  A BN_dup() followed by a bn_expand2()
+ * will allocate new memory for the BIGNUM data twice, and free it once,
+ * while bn_dup_expand() makes sure allocation is made only once.
+ */
+
+BIGNUM *bn_dup_expand(const BIGNUM *b, int words)
+        {
+        BIGNUM *r = NULL;
+
+        if (words > b->dmax)
+                {
+                BN_ULONG *a = bn_expand_internal(b, words);
+
+                if (a)
+                        {
+                        r = BN_new();
+                        if (r)
                                 {
-                                /*
-                                 * The fact that the loop is unrolled
-                                 * 4-wise is a tribute to Intel. It's
-                                 * the one that doesn't have enough
-                                 * registers to accomodate more data.
-                                 * I'd unroll it 8-wise otherwise:-)
-                                 *
-                                 *              <appro@fy.chalmers.se>
-                                 */
-                                BN_ULONG a0,a1,a2,a3;
-                                a0=B[0]; a1=B[1]; a2=B[2]; a3=B[3];
-                                A[0]=a0; A[1]=a1; A[2]=a2; A[3]=a3;
+                                r->top = b->top;
+                                r->dmax = words;
+                                r->neg = b->neg;
+                                r->d = a;
                                 }
-                        switch (b->top&3)
+                        else
                                 {
-                                case 3: A[2]=B[2];
-                                case 2: A[1]=B[1];
-                                case 1: A[0]=B[0];
-                                case 0: ; /* ultrix cc workaround, see above */
+                                /* r == NULL, BN_new failure */
+                                OPENSSL_free(a);
                                 }
-#endif
-                        OPENSSL_free(b->d);
                         }
+                /* If a == NULL, there was an error in allocation in
+                   bn_expand_internal(), and NULL should be returned */
+                }
+        else
+                {
+                r = BN_dup(b);
+                }
 
-                b->d=a;
-                b->dmax=words;
+        return r;
+        }
+
+/* This is an internal function that should not be used in applications.
+ * It ensures that 'b' has enough room for a 'words' word number number.
+ * It is mostly used by the various BIGNUM routines. If there is an error,
+ * NULL is returned. If not, 'b' is returned. */
 
-                /* Now need to zero any data between b->top and b->max */
+BIGNUM *bn_expand2(BIGNUM *b, int words)
+        {
+        if (words > b->dmax)
+                {
+                BN_ULONG *a = bn_expand_internal(b, words);
 
-                A= &(b->d[b->top]);
-                for (i=(b->dmax - b->top)>>3; i>0; i--,A+=8)
+                if (a)
                         {
-                        A[0]=0; A[1]=0; A[2]=0; A[3]=0;
-                        A[4]=0; A[5]=0; A[6]=0; A[7]=0;
-                        }
-                for (i=(b->dmax - b->top)&7; i>0; i--,A++)
-                        A[0]=0;
-#else
-                        memset(A,0,sizeof(BN_ULONG)*(words+1));
-                        memcpy(A,b->d,sizeof(b->d[0])*b->top);
+                        if (b->d)
+                                OPENSSL_free(b->d);
                         b->d=a;
-                        b->max=words;
-#endif
-                
-/*              memset(&(p[b->max]),0,((words+1)-b->max)*sizeof(BN_ULONG)); */
-/*      { int i; for (i=b->max; i<words+1; i++) p[i]=i;} */
-
+                        b->dmax=words;
+                        }
+                else
+                        b = NULL;
                 }
-        return(b);
+        return b;
         }
 
 BIGNUM *BN_dup(const BIGNUM *a)
         {
-        BIGNUM *r;
+        BIGNUM *r, *t;
 
         if (a == NULL) return NULL;
 
         bn_check_top(a);
 
-        r=BN_new();
-        if (r == NULL) return(NULL);
-        return((BIGNUM *)BN_copy(r,a));
+        t = BN_new();
+        if (t == NULL) return(NULL);
+        r = BN_copy(t, a);
+        /* now  r == t || r == NULL */
+        if (r == NULL)
+                BN_free(t);
+        return r;
         }
 
 BIGNUM *BN_copy(BIGNUM *a, const BIGNUM *b)
@@ -498,7 +494,7 @@ BIGNUM *BN_copy(BIGNUM *a, const BIGNUM *b)
                 case 3: A[2]=B[2];
                 case 2: A[1]=B[1];
                 case 1: A[0]=B[0];
-                case 0: ; /* ultrix cc workaround, see comments in bn_expand2 */
+                case 0: ; /* ultrix cc workaround, see comments in bn_expand_internal */
                 }
 #else
         memcpy(a->d,b->d,sizeof(b->d[0])*b->top);
@@ -512,6 +508,35 @@ BIGNUM *BN_copy(BIGNUM *a, const BIGNUM *b)
         return(a);
         }
 
+void BN_swap(BIGNUM *a, BIGNUM *b)
+        {
+        int flags_old_a, flags_old_b;
+        BN_ULONG *tmp_d;
+        int tmp_top, tmp_dmax, tmp_neg;
+        
+        flags_old_a = a->flags;
+        flags_old_b = b->flags;
+
+        tmp_d = a->d;
+        tmp_top = a->top;
+        tmp_dmax = a->dmax;
+        tmp_neg = a->neg;
+        
+        a->d = b->d;
+        a->top = b->top;
+        a->dmax = b->dmax;
+        a->neg = b->neg;
+        
+        b->d = tmp_d;
+        b->top = tmp_top;
+        b->dmax = tmp_dmax;
+        b->neg = tmp_neg;
+        
+        a->flags = (flags_old_a & BN_FLG_MALLOCED) | (flags_old_b & BN_FLG_STATIC_DATA);
+        b->flags = (flags_old_b & BN_FLG_MALLOCED) | (flags_old_a & BN_FLG_STATIC_DATA);
+        }
+
+
 void BN_clear(BIGNUM *a)
         {
         if (a->d != NULL)
@@ -520,7 +545,7 @@ void BN_clear(BIGNUM *a)
         a->neg=0;
         }
 
-BN_ULONG BN_get_word(BIGNUM *a)
+BN_ULONG BN_get_word(const BIGNUM *a)
         {
         int i,n;
         BN_ULONG ret=0;
@@ -568,7 +593,6 @@ int BN_set_word(BIGNUM *a, BN_ULONG w)
         return(1);
         }
 
-/* ignore negative */
 BIGNUM *BN_bin2bn(const unsigned char *s, int len, BIGNUM *ret)
         {
         unsigned int i,m;
@@ -589,6 +613,7 @@ BIGNUM *BN_bin2bn(const unsigned char *s, int len, BIGNUM *ret)
         i=((n-1)/BN_BYTES)+1;
         m=((n-1)%(BN_BYTES));
         ret->top=i;
+        ret->neg=0;
         while (n-- > 0)
                 {
                 l=(l<<8L)| *(s++);
@@ -743,7 +768,7 @@ int BN_mask_bits(BIGNUM *a, int n)
         return(1);
         }
 
-int bn_cmp_words(BN_ULONG *a, BN_ULONG *b, int n)
+int bn_cmp_words(const BN_ULONG *a, const BN_ULONG *b, int n)
         {
         int i;
         BN_ULONG aa,bb;
@@ -760,3 +785,34 @@ int bn_cmp_words(BN_ULONG *a, BN_ULONG *b, int n)
         return(0);
         }
 
+/* Here follows a specialised variants of bn_cmp_words().  It has the
+   property of performing the operation on arrays of different sizes.
+   The sizes of those arrays is expressed through cl, which is the
+   common length ( basicall, min(len(a),len(b)) ), and dl, which is the
+   delta between the two lengths, calculated as len(a)-len(b).
+   All lengths are the number of BN_ULONGs...  */
+
+int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b,
+        int cl, int dl)
+        {
+        int n,i;
+        n = cl-1;
+
+        if (dl < 0)
+                {
+                for (i=dl; i<0; i++)
+                        {
+                        if (b[n-i] != 0)
+                                return -1; /* a < b */
+                        }
+                }
+        if (dl > 0)
+                {
+                for (i=dl; i>0; i--)
+                        {
+                        if (a[n+i] != 0)
+                                return 1; /* a > b */
+                        }
+                }
+        return bn_cmp_words(a,b,cl);
+        }