import openssl-0.9.7-beta1

5 files changed, 291 insertions, 399 deletions

diff --git a/src/lib/libcrypto/md5/asm/md5-586.pl b/src/lib/libcrypto/md5/asm/md5-586.pl
index 2c7fb7dd98..5fc6a205ce 100644
--- a/src/lib/libcrypto/md5/asm/md5-586.pl
+++ b/src/lib/libcrypto/md5/asm/md5-586.pl
@@ -1,4 +1,4 @@
-#!/usr/bin/perl
+#!/usr/local/bin/perl
 
 # Normal is the
 # md5_block_x86(MD5_CTX *c, ULONG *X);
@@ -29,7 +29,7 @@ $X="esi";
  0, 7, 14, 5, 12, 3, 10, 1, 8, 15, 6, 13, 4, 11, 2, 9,  # R3
  );
 
-&md5_block("md5_block_x86");
+&md5_block("md5_block_asm_host_order");
 &asm_finish();
 
 sub Np
@@ -44,7 +44,7 @@ sub R0
         local($pos,$a,$b,$c,$d,$K,$ki,$s,$t)=@_;
 
         &mov($tmp1,$C)  if $pos < 0;
-         &mov($tmp2,&DWP($xo[$ki]*4,$K,"",0)) if $pos < 0; # very first one 
+        &mov($tmp2,&DWP($xo[$ki]*4,$K,"",0)) if $pos < 0; # very first one 
 
         # body proper
 
@@ -54,7 +54,6 @@ sub R0
         &and($tmp1,$b); # F function - part 3
         &lea($a,&DWP($t,$a,$tmp2,1));
 
-        &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if ($pos != 2);
         &xor($tmp1,$d); # F function - part 4
 
         &add($a,$tmp1);
@@ -62,8 +61,10 @@ sub R0
         &mov($tmp1,&Np($c)) if $pos == 1;       # next tmp1 for R1
 
         &rotl($a,$s);
-        &add($a,$b);
 
+        &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if ($pos != 2);
+
+        &add($a,$b);
         }
 
 sub R1
@@ -100,16 +101,16 @@ if (($n & 1) == 0)
         # make sure to do 'D' first, not 'B', else we clash with
         # the last add from the previous round.
 
-         &xor($tmp1,$d); # H function - part 2
+        &xor($tmp1,$d); # H function - part 2
 
         &xor($tmp1,$b); # H function - part 3
-         &lea($a,&DWP($t,$a,$tmp2,1));
+        &lea($a,&DWP($t,$a,$tmp2,1));
 
         &add($a,$tmp1);
-         &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0));
 
         &rotl($a,$s);
 
+        &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0));
         &mov($tmp1,&Np($c));
         }
 else
@@ -118,17 +119,17 @@ else
         # make sure to do 'D' first, not 'B', else we clash with
         # the last add from the previous round.
 
-         &lea($a,&DWP($t,$a,$tmp2,1));
+        &lea($a,&DWP($t,$a,$tmp2,1));
 
         &add($b,$c);                    # MOVED FORWARD
-         &xor($tmp1,$d); # H function - part 2
+        &xor($tmp1,$d); # H function - part 2
 
         &xor($tmp1,$b); # H function - part 3
-         &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if ($pos != 2);
+        &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if ($pos != 2);
 
         &add($a,$tmp1);
-         &mov($tmp1,&Np($c)) if $pos < 1;       # H function - part 1
-         &mov($tmp1,-1) if $pos == 1;           # I function - part 1
+        &mov($tmp1,&Np($c)) if $pos < 1;        # H function - part 1
+        &mov($tmp1,-1) if $pos == 1;            # I function - part 1
 
         &rotl($a,$s);
 
@@ -146,21 +147,21 @@ sub R3
         &xor($tmp1,$d) if $pos < 0;     # I function - part 2
 
         &or($tmp1,$b);                          # I function - part 3
-         &lea($a,&DWP($t,$a,$tmp2,1));
+        &lea($a,&DWP($t,$a,$tmp2,1));
 
         &xor($tmp1,$c);                         # I function - part 4
-         &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0)) if $pos != 2; # load X/k value
-         &mov($tmp2,&wparam(0)) if $pos == 2;
+        &mov($tmp2,&DWP($xo[$ki+1]*4,$K,"",0))  if $pos != 2; # load X/k value
+        &mov($tmp2,&wparam(0)) if $pos == 2;
 
         &add($a,$tmp1);
-         &mov($tmp1,-1) if $pos < 1;    # H function - part 1
-         &add($K,64) if $pos >=1 && !$normal;
+        &mov($tmp1,-1) if $pos < 1;     # H function - part 1
+        &add($K,64) if $pos >=1 && !$normal;
 
         &rotl($a,$s);
 
         &xor($tmp1,&Np($d)) if $pos <= 0;       # I function - part = first time
         &mov($tmp1,&DWP( 0,$tmp2,"",0)) if $pos > 0;
-         &add($a,$b);
+        &add($a,$b);
         }
 
 
@@ -182,6 +183,7 @@ sub md5_block
          &mov($X,       &wparam(1)); # esi
         &mov($C,        &wparam(2));
          &push("ebp");
+        &shl($C,        6);
         &push("ebx");
          &add($C,       $X); # offset we end at
         &sub($C,        64);
diff --git a/src/lib/libcrypto/md5/md5.h b/src/lib/libcrypto/md5/md5.h
index 357c6c625d..52cb753e6a 100644
--- a/src/lib/libcrypto/md5/md5.h
+++ b/src/lib/libcrypto/md5/md5.h
@@ -63,35 +63,50 @@
 extern "C" {
 #endif
 
+#ifdef OPENSSL_NO_MD5
+#error MD5 is disabled.
+#endif
+
+/*
+ * !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
+ * ! MD5_LONG has to be at least 32 bits wide. If it's wider, then !
+ * ! MD5_LONG_LOG2 has to be defined along.                        !
+ * !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
+ */
+
+#if defined(OPENSSL_SYS_WIN16) || defined(__LP32__)
+#define MD5_LONG unsigned long
+#elif defined(OENSSL_SYS_CRAY) || defined(__ILP64__)
+#define MD5_LONG unsigned long
+#define MD5_LONG_LOG2 3
+/*
+ * _CRAY note. I could declare short, but I have no idea what impact
+ * does it have on performance on none-T3E machines. I could declare
+ * int, but at least on C90 sizeof(int) can be chosen at compile time.
+ * So I've chosen long...
+ *                                      <appro@fy.chalmers.se>
+ */
+#else
+#define MD5_LONG unsigned int
+#endif
+
 #define MD5_CBLOCK      64
-#define MD5_LBLOCK      16
-#define MD5_BLOCK       16
-#define MD5_LAST_BLOCK  56
-#define MD5_LENGTH_BLOCK 8
+#define MD5_LBLOCK      (MD5_CBLOCK/4)
 #define MD5_DIGEST_LENGTH 16
 
 typedef struct MD5state_st
         {
-        unsigned long A,B,C,D;
-        unsigned long Nl,Nh;
-        unsigned long data[MD5_LBLOCK];
+        MD5_LONG A,B,C,D;
+        MD5_LONG Nl,Nh;
+        MD5_LONG data[MD5_LBLOCK];
         int num;
         } MD5_CTX;
 
-#ifndef NOPROTO
-void MD5_Init(MD5_CTX *c);
-void MD5_Update(MD5_CTX *c, unsigned char *data, unsigned long len);
-void MD5_Final(unsigned char *md, MD5_CTX *c);
-unsigned char *MD5(unsigned char *d, unsigned long n, unsigned char *md);
-void MD5_Transform(MD5_CTX *c, unsigned char *b);
-#else
-void MD5_Init();
-void MD5_Update();
-void MD5_Final();
-unsigned char *MD5();
-void MD5_Transform();
-#endif
-
+int MD5_Init(MD5_CTX *c);
+int MD5_Update(MD5_CTX *c, const void *data, unsigned long len);
+int MD5_Final(unsigned char *md, MD5_CTX *c);
+unsigned char *MD5(const unsigned char *d, unsigned long n, unsigned char *md);
+void MD5_Transform(MD5_CTX *c, const unsigned char *b);
 #ifdef  __cplusplus
 }
 #endif
diff --git a/src/lib/libcrypto/md5/md5_dgst.c b/src/lib/libcrypto/md5/md5_dgst.c
index 43b3498d92..c38a3f021e 100644
--- a/src/lib/libcrypto/md5/md5_dgst.c
+++ b/src/lib/libcrypto/md5/md5_dgst.c
@@ -58,8 +58,9 @@
 
 #include <stdio.h>
 #include "md5_locl.h"
+#include <openssl/opensslv.h>
 
-char *MD5_version="MD5 part of SSLeay 0.9.0b 29-Jun-1998";
+const char *MD5_version="MD5" OPENSSL_VERSION_PTEXT;
 
 /* Implemented from RFC1321 The MD5 Message-Digest Algorithm
  */
@@ -69,24 +70,7 @@ char *MD5_version="MD5 part of SSLeay 0.9.0b 29-Jun-1998";
 #define INIT_DATA_C (unsigned long)0x98badcfeL
 #define INIT_DATA_D (unsigned long)0x10325476L
 
-#ifndef NOPROTO
-#  ifdef MD5_ASM
-     void md5_block_x86(MD5_CTX *c, unsigned long *p,int num);
-#    define md5_block md5_block_x86
-#  else
-     static void md5_block(MD5_CTX *c, unsigned long *p,int num);
-#  endif
-#else
-#  ifdef MD5_ASM
-     void md5_block_x86();
-#    define md5_block md5_block_x86
-#  else
-     static void md5_block();
-#  endif
-#endif
-
-void MD5_Init(c)
-MD5_CTX *c;
+int MD5_Init(MD5_CTX *c)
         {
         c->A=INIT_DATA_A;
         c->B=INIT_DATA_B;
@@ -95,192 +79,35 @@ MD5_CTX *c;
         c->Nl=0;
         c->Nh=0;
         c->num=0;
+        return 1;
         }
 
-void MD5_Update(c, data, len)
-MD5_CTX *c;
-register unsigned char *data;
-unsigned long len;
-        {
-        register ULONG *p;
-        int sw,sc;
-        ULONG l;
-
-        if (len == 0) return;
-
-        l=(c->Nl+(len<<3))&0xffffffffL;
-        /* 95-05-24 eay Fixed a bug with the overflow handling, thanks to
-         * Wei Dai <weidai@eskimo.com> for pointing it out. */
-        if (l < c->Nl) /* overflow */
-                c->Nh++;
-        c->Nh+=(len>>29);
-        c->Nl=l;
-
-        if (c->num != 0)
-                {
-                p=c->data;
-                sw=c->num>>2;
-                sc=c->num&0x03;
-
-                if ((c->num+len) >= MD5_CBLOCK)
-                        {
-                        l= p[sw];
-                        p_c2l(data,l,sc);
-                        p[sw++]=l;
-                        for (; sw<MD5_LBLOCK; sw++)
-                                {
-                                c2l(data,l);
-                                p[sw]=l;
-                                }
-                        len-=(MD5_CBLOCK-c->num);
-
-                        md5_block(c,p,64);
-                        c->num=0;
-                        /* drop through and do the rest */
-                        }
-                else
-                        {
-                        int ew,ec;
-
-                        c->num+=(int)len;
-                        if ((sc+len) < 4) /* ugly, add char's to a word */
-                                {
-                                l= p[sw];
-                                p_c2l_p(data,l,sc,len);
-                                p[sw]=l;
-                                }
-                        else
-                                {
-                                ew=(c->num>>2);
-                                ec=(c->num&0x03);
-                                l= p[sw];
-                                p_c2l(data,l,sc);
-                                p[sw++]=l;
-                                for (; sw < ew; sw++)
-                                        { c2l(data,l); p[sw]=l; }
-                                if (ec)
-                                        {
-                                        c2l_p(data,l,ec);
-                                        p[sw]=l;
-                                        }
-                                }
-                        return;
-                        }
-                }
-        /* we now can process the input data in blocks of MD5_CBLOCK
-         * chars and save the leftovers to c->data. */
-#ifdef L_ENDIAN
-        if ((((unsigned long)data)%sizeof(ULONG)) == 0)
-                {
-                sw=(int)len/MD5_CBLOCK;
-                if (sw > 0)
-                        {
-                        sw*=MD5_CBLOCK;
-                        md5_block(c,(ULONG *)data,sw);
-                        data+=sw;
-                        len-=sw;
-                        }
-                }
-#endif
-        p=c->data;
-        while (len >= MD5_CBLOCK)
-                {
-#if defined(L_ENDIAN) || defined(B_ENDIAN)
-                if (p != (unsigned long *)data)
-                        memcpy(p,data,MD5_CBLOCK);
-                data+=MD5_CBLOCK;
-#ifdef B_ENDIAN
-                for (sw=(MD5_LBLOCK/4); sw; sw--)
-                        {
-                        Endian_Reverse32(p[0]);
-                        Endian_Reverse32(p[1]);
-                        Endian_Reverse32(p[2]);
-                        Endian_Reverse32(p[3]);
-                        p+=4;
-                        }
-#endif
-#else
-                for (sw=(MD5_LBLOCK/4); sw; sw--)
-                        {
-                        c2l(data,l); *(p++)=l;
-                        c2l(data,l); *(p++)=l;
-                        c2l(data,l); *(p++)=l;
-                        c2l(data,l); *(p++)=l; 
-                        } 
-#endif
-                p=c->data;
-                md5_block(c,p,64);
-                len-=MD5_CBLOCK;
-                }
-        sc=(int)len;
-        c->num=sc;
-        if (sc)
-                {
-                sw=sc>>2;       /* words to copy */
-#ifdef L_ENDIAN
-                p[sw]=0;
-                memcpy(p,data,sc);
-#else
-                sc&=0x03;
-                for ( ; sw; sw--)
-                        { c2l(data,l); *(p++)=l; }
-                c2l_p(data,l,sc);
-                *p=l;
-#endif
-                }
-        }
-
-void MD5_Transform(c,b)
-MD5_CTX *c;
-unsigned char *b;
-        {
-        ULONG p[16];
-#if !defined(L_ENDIAN)
-        ULONG *q;
-        int i;
-#endif
-
-#if defined(B_ENDIAN) || defined(L_ENDIAN)
-        memcpy(p,b,64);
-#ifdef B_ENDIAN
-        q=p;
-        for (i=(MD5_LBLOCK/4); i; i--)
-                {
-                Endian_Reverse32(q[0]);
-                Endian_Reverse32(q[1]);
-                Endian_Reverse32(q[2]);
-                Endian_Reverse32(q[3]);
-                q+=4;
-                }
-#endif
-#else
-        q=p;
-        for (i=(MD5_LBLOCK/4); i; i--)
-                {
-                ULONG l;
-                c2l(b,l); *(q++)=l;
-                c2l(b,l); *(q++)=l;
-                c2l(b,l); *(q++)=l;
-                c2l(b,l); *(q++)=l; 
-                } 
-#endif
-        md5_block(c,p,64);
-        }
-
-#ifndef MD5_ASM
-
-static void md5_block(c, X, num)
-MD5_CTX *c;
-register ULONG *X;
-int num;
+#ifndef md5_block_host_order
+void md5_block_host_order (MD5_CTX *c, const void *data, int num)
         {
-        register ULONG A,B,C,D;
+        const MD5_LONG *X=data;
+        register unsigned long A,B,C,D;
+        /*
+         * In case you wonder why A-D are declared as long and not
+         * as MD5_LONG. Doing so results in slight performance
+         * boost on LP64 architectures. The catch is we don't
+         * really care if 32 MSBs of a 64-bit register get polluted
+         * with eventual overflows as we *save* only 32 LSBs in
+         * *either* case. Now declaring 'em long excuses the compiler
+         * from keeping 32 MSBs zeroed resulting in 13% performance
+         * improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
+         * Well, to be honest it should say that this *prevents* 
+         * performance degradation.
+         *
+         *                              <appro@fy.chalmers.se>
+         */
 
         A=c->A;
         B=c->B;
         C=c->C;
         D=c->D;
-        for (;;)
+
+        for (;num--;X+=HASH_LBLOCK)
                 {
         /* Round 0 */
         R0(A,B,C,D,X[ 0], 7,0xd76aa478L);
@@ -351,80 +178,133 @@ int num;
         R3(C,D,A,B,X[ 2],15,0x2ad7d2bbL);
         R3(B,C,D,A,X[ 9],21,0xeb86d391L);
 
-        A+=c->A&0xffffffffL;
-        B+=c->B&0xffffffffL;
-        c->A=A;
-        c->B=B;
-        C+=c->C&0xffffffffL;
-        D+=c->D&0xffffffffL;
-        c->C=C;
-        c->D=D;
-        X+=16;
-        num-=64;
-        if (num <= 0) break;
+        A = c->A += A;
+        B = c->B += B;
+        C = c->C += C;
+        D = c->D += D;
                 }
         }
 #endif
 
-void MD5_Final(md, c)
-unsigned char *md;
-MD5_CTX *c;
+#ifndef md5_block_data_order
+#ifdef X
+#undef X
+#endif
+void md5_block_data_order (MD5_CTX *c, const void *data_, int num)
         {
-        register int i,j;
-        register ULONG l;
-        register ULONG *p;
-        static unsigned char end[4]={0x80,0x00,0x00,0x00};
-        unsigned char *cp=end;
+        const unsigned char *data=data_;
+        register unsigned long A,B,C,D,l;
+        /*
+         * In case you wonder why A-D are declared as long and not
+         * as MD5_LONG. Doing so results in slight performance
+         * boost on LP64 architectures. The catch is we don't
+         * really care if 32 MSBs of a 64-bit register get polluted
+         * with eventual overflows as we *save* only 32 LSBs in
+         * *either* case. Now declaring 'em long excuses the compiler
+         * from keeping 32 MSBs zeroed resulting in 13% performance
+         * improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
+         * Well, to be honest it should say that this *prevents* 
+         * performance degradation.
+         *
+         *                              <appro@fy.chalmers.se>
+         */
+#ifndef MD32_XARRAY
+        /* See comment in crypto/sha/sha_locl.h for details. */
+        unsigned long   XX0, XX1, XX2, XX3, XX4, XX5, XX6, XX7,
+                        XX8, XX9,XX10,XX11,XX12,XX13,XX14,XX15;
+# define X(i)   XX##i
+#else
+        MD5_LONG XX[MD5_LBLOCK];
+# define X(i)   XX[i]
+#endif
 
-        /* c->num should definitly have room for at least one more byte. */
-        p=c->data;
-        j=c->num;
-        i=j>>2;
+        A=c->A;
+        B=c->B;
+        C=c->C;
+        D=c->D;
 
-        /* purify often complains about the following line as an
-         * Uninitialized Memory Read.  While this can be true, the
-         * following p_c2l macro will reset l when that case is true.
-         * This is because j&0x03 contains the number of 'valid' bytes
-         * already in p[i].  If and only if j&0x03 == 0, the UMR will
-         * occur but this is also the only time p_c2l will do
-         * l= *(cp++) instead of l|= *(cp++)
-         * Many thanks to Alex Tang <altitude@cic.net> for pickup this
-         * 'potential bug' */
-#ifdef PURIFY
-        if ((j&0x03) == 0) p[i]=0;
-#endif
-        l=p[i];
-        p_c2l(cp,l,j&0x03);
-        p[i]=l;
-        i++;
-        /* i is the next 'undefined word' */
-        if (c->num >= MD5_LAST_BLOCK)
+        for (;num--;)
                 {
-                for (; i<MD5_LBLOCK; i++)
-                        p[i]=0;
-                md5_block(c,p,64);
-                i=0;
-                }
-        for (; i<(MD5_LBLOCK-2); i++)
-                p[i]=0;
-        p[MD5_LBLOCK-2]=c->Nl;
-        p[MD5_LBLOCK-1]=c->Nh;
-        md5_block(c,p,64);
-        cp=md;
-        l=c->A; l2c(l,cp);
-        l=c->B; l2c(l,cp);
-        l=c->C; l2c(l,cp);
-        l=c->D; l2c(l,cp);
+        HOST_c2l(data,l); X( 0)=l;              HOST_c2l(data,l); X( 1)=l;
+        /* Round 0 */
+        R0(A,B,C,D,X( 0), 7,0xd76aa478L);       HOST_c2l(data,l); X( 2)=l;
+        R0(D,A,B,C,X( 1),12,0xe8c7b756L);       HOST_c2l(data,l); X( 3)=l;
+        R0(C,D,A,B,X( 2),17,0x242070dbL);       HOST_c2l(data,l); X( 4)=l;
+        R0(B,C,D,A,X( 3),22,0xc1bdceeeL);       HOST_c2l(data,l); X( 5)=l;
+        R0(A,B,C,D,X( 4), 7,0xf57c0fafL);       HOST_c2l(data,l); X( 6)=l;
+        R0(D,A,B,C,X( 5),12,0x4787c62aL);       HOST_c2l(data,l); X( 7)=l;
+        R0(C,D,A,B,X( 6),17,0xa8304613L);       HOST_c2l(data,l); X( 8)=l;
+        R0(B,C,D,A,X( 7),22,0xfd469501L);       HOST_c2l(data,l); X( 9)=l;
+        R0(A,B,C,D,X( 8), 7,0x698098d8L);       HOST_c2l(data,l); X(10)=l;
+        R0(D,A,B,C,X( 9),12,0x8b44f7afL);       HOST_c2l(data,l); X(11)=l;
+        R0(C,D,A,B,X(10),17,0xffff5bb1L);       HOST_c2l(data,l); X(12)=l;
+        R0(B,C,D,A,X(11),22,0x895cd7beL);       HOST_c2l(data,l); X(13)=l;
+        R0(A,B,C,D,X(12), 7,0x6b901122L);       HOST_c2l(data,l); X(14)=l;
+        R0(D,A,B,C,X(13),12,0xfd987193L);       HOST_c2l(data,l); X(15)=l;
+        R0(C,D,A,B,X(14),17,0xa679438eL);
+        R0(B,C,D,A,X(15),22,0x49b40821L);
+        /* Round 1 */
+        R1(A,B,C,D,X( 1), 5,0xf61e2562L);
+        R1(D,A,B,C,X( 6), 9,0xc040b340L);
+        R1(C,D,A,B,X(11),14,0x265e5a51L);
+        R1(B,C,D,A,X( 0),20,0xe9b6c7aaL);
+        R1(A,B,C,D,X( 5), 5,0xd62f105dL);
+        R1(D,A,B,C,X(10), 9,0x02441453L);
+        R1(C,D,A,B,X(15),14,0xd8a1e681L);
+        R1(B,C,D,A,X( 4),20,0xe7d3fbc8L);
+        R1(A,B,C,D,X( 9), 5,0x21e1cde6L);
+        R1(D,A,B,C,X(14), 9,0xc33707d6L);
+        R1(C,D,A,B,X( 3),14,0xf4d50d87L);
+        R1(B,C,D,A,X( 8),20,0x455a14edL);
+        R1(A,B,C,D,X(13), 5,0xa9e3e905L);
+        R1(D,A,B,C,X( 2), 9,0xfcefa3f8L);
+        R1(C,D,A,B,X( 7),14,0x676f02d9L);
+        R1(B,C,D,A,X(12),20,0x8d2a4c8aL);
+        /* Round 2 */
+        R2(A,B,C,D,X( 5), 4,0xfffa3942L);
+        R2(D,A,B,C,X( 8),11,0x8771f681L);
+        R2(C,D,A,B,X(11),16,0x6d9d6122L);
+        R2(B,C,D,A,X(14),23,0xfde5380cL);
+        R2(A,B,C,D,X( 1), 4,0xa4beea44L);
+        R2(D,A,B,C,X( 4),11,0x4bdecfa9L);
+        R2(C,D,A,B,X( 7),16,0xf6bb4b60L);
+        R2(B,C,D,A,X(10),23,0xbebfbc70L);
+        R2(A,B,C,D,X(13), 4,0x289b7ec6L);
+        R2(D,A,B,C,X( 0),11,0xeaa127faL);
+        R2(C,D,A,B,X( 3),16,0xd4ef3085L);
+        R2(B,C,D,A,X( 6),23,0x04881d05L);
+        R2(A,B,C,D,X( 9), 4,0xd9d4d039L);
+        R2(D,A,B,C,X(12),11,0xe6db99e5L);
+        R2(C,D,A,B,X(15),16,0x1fa27cf8L);
+        R2(B,C,D,A,X( 2),23,0xc4ac5665L);
+        /* Round 3 */
+        R3(A,B,C,D,X( 0), 6,0xf4292244L);
+        R3(D,A,B,C,X( 7),10,0x432aff97L);
+        R3(C,D,A,B,X(14),15,0xab9423a7L);
+        R3(B,C,D,A,X( 5),21,0xfc93a039L);
+        R3(A,B,C,D,X(12), 6,0x655b59c3L);
+        R3(D,A,B,C,X( 3),10,0x8f0ccc92L);
+        R3(C,D,A,B,X(10),15,0xffeff47dL);
+        R3(B,C,D,A,X( 1),21,0x85845dd1L);
+        R3(A,B,C,D,X( 8), 6,0x6fa87e4fL);
+        R3(D,A,B,C,X(15),10,0xfe2ce6e0L);
+        R3(C,D,A,B,X( 6),15,0xa3014314L);
+        R3(B,C,D,A,X(13),21,0x4e0811a1L);
+        R3(A,B,C,D,X( 4), 6,0xf7537e82L);
+        R3(D,A,B,C,X(11),10,0xbd3af235L);
+        R3(C,D,A,B,X( 2),15,0x2ad7d2bbL);
+        R3(B,C,D,A,X( 9),21,0xeb86d391L);
 
-        /* clear stuff, md5_block may be leaving some stuff on the stack
-         * but I'm not worried :-) */
-        c->num=0;
-/*      memset((char *)&c,0,sizeof(c));*/
+        A = c->A += A;
+        B = c->B += B;
+        C = c->C += C;
+        D = c->D += D;
+                }
         }
+#endif
 
 #ifdef undef
-int printit(l)
-unsigned long *l;
+int printit(unsigned long *l)
         {
         int i,ii;
 
diff --git a/src/lib/libcrypto/md5/md5_locl.h b/src/lib/libcrypto/md5/md5_locl.h
index dbbe1b71ca..34c5257306 100644
--- a/src/lib/libcrypto/md5/md5_locl.h
+++ b/src/lib/libcrypto/md5/md5_locl.h
@@ -56,109 +56,94 @@
  * [including the GNU Public Licence.]
  */
 
-/* On sparc, this actually slows things down :-( */
-#if defined(sun)
-#undef B_ENDIAN
-#endif
-
 #include <stdlib.h>
 #include <string.h>
-#include "md5.h"
-
-#define ULONG   unsigned long
-#define UCHAR   unsigned char
-#define UINT    unsigned int
+#include <openssl/opensslconf.h>
+#include <openssl/md5.h>
 
-#if defined(NOCONST)
-#define const
+#ifndef MD5_LONG_LOG2
+#define MD5_LONG_LOG2 2 /* default to 32 bits */
 #endif
 
-#undef c2l
-#define c2l(c,l)        (l = ((unsigned long)(*((c)++)))     , \
-                         l|=(((unsigned long)(*((c)++)))<< 8), \
-                         l|=(((unsigned long)(*((c)++)))<<16), \
-                         l|=(((unsigned long)(*((c)++)))<<24))
-
-#undef p_c2l
-#define p_c2l(c,l,n)    { \
-                        switch (n) { \
-                        case 0: l =((unsigned long)(*((c)++))); \
-                        case 1: l|=((unsigned long)(*((c)++)))<< 8; \
-                        case 2: l|=((unsigned long)(*((c)++)))<<16; \
-                        case 3: l|=((unsigned long)(*((c)++)))<<24; \
-                                } \
-                        }
+#ifdef MD5_ASM
+# if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__)
+#  define md5_block_host_order md5_block_asm_host_order
+# elif defined(__sparc) && defined(OPENSSL_SYS_ULTRASPARC)
+   void md5_block_asm_data_order_aligned (MD5_CTX *c, const MD5_LONG *p,int num);
+#  define HASH_BLOCK_DATA_ORDER_ALIGNED md5_block_asm_data_order_aligned
+# endif
+#endif
 
-/* NOTE the pointer is not incremented at the end of this */
-#undef c2l_p
-#define c2l_p(c,l,n)    { \
-                        l=0; \
-                        (c)+=n; \
-                        switch (n) { \
-                        case 3: l =((unsigned long)(*(--(c))))<<16; \
-                        case 2: l|=((unsigned long)(*(--(c))))<< 8; \
-                        case 1: l|=((unsigned long)(*(--(c))))    ; \
-                                } \
-                        }
+void md5_block_host_order (MD5_CTX *c, const void *p,int num);
+void md5_block_data_order (MD5_CTX *c, const void *p,int num);
 
-#undef p_c2l_p
-#define p_c2l_p(c,l,sc,len) { \
-                        switch (sc) \
-                                { \
-                        case 0: l =((unsigned long)(*((c)++))); \
-                                if (--len == 0) break; \
-                        case 1: l|=((unsigned long)(*((c)++)))<< 8; \
-                                if (--len == 0) break; \
-                        case 2: l|=((unsigned long)(*((c)++)))<<16; \
-                                } \
-                        }
+#if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__)
+/*
+ * *_block_host_order is expected to handle aligned data while
+ * *_block_data_order - unaligned. As algorithm and host (x86)
+ * are in this case of the same "endianness" these two are
+ * otherwise indistinguishable. But normally you don't want to
+ * call the same function because unaligned access in places
+ * where alignment is expected is usually a "Bad Thing". Indeed,
+ * on RISCs you get punished with BUS ERROR signal or *severe*
+ * performance degradation. Intel CPUs are in turn perfectly
+ * capable of loading unaligned data without such drastic side
+ * effect. Yes, they say it's slower than aligned load, but no
+ * exception is generated and therefore performance degradation
+ * is *incomparable* with RISCs. What we should weight here is
+ * costs of unaligned access against costs of aligning data.
+ * According to my measurements allowing unaligned access results
+ * in ~9% performance improvement on Pentium II operating at
+ * 266MHz. I won't be surprised if the difference will be higher
+ * on faster systems:-)
+ *
+ *                              <appro@fy.chalmers.se>
+ */
+#define md5_block_data_order md5_block_host_order
+#endif
 
-#undef l2c
-#define l2c(l,c)        (*((c)++)=(unsigned char)(((l)    )&0xff), \
-                         *((c)++)=(unsigned char)(((l)>> 8)&0xff), \
-                         *((c)++)=(unsigned char)(((l)>>16)&0xff), \
-                         *((c)++)=(unsigned char)(((l)>>24)&0xff))
+#define DATA_ORDER_IS_LITTLE_ENDIAN
+
+#define HASH_LONG               MD5_LONG
+#define HASH_LONG_LOG2          MD5_LONG_LOG2
+#define HASH_CTX                MD5_CTX
+#define HASH_CBLOCK             MD5_CBLOCK
+#define HASH_LBLOCK             MD5_LBLOCK
+#define HASH_UPDATE             MD5_Update
+#define HASH_TRANSFORM          MD5_Transform
+#define HASH_FINAL              MD5_Final
+#define HASH_MAKE_STRING(c,s)   do {    \
+        unsigned long ll;               \
+        ll=(c)->A; HOST_l2c(ll,(s));    \
+        ll=(c)->B; HOST_l2c(ll,(s));    \
+        ll=(c)->C; HOST_l2c(ll,(s));    \
+        ll=(c)->D; HOST_l2c(ll,(s));    \
+        } while (0)
+#define HASH_BLOCK_HOST_ORDER   md5_block_host_order
+#if !defined(L_ENDIAN) || defined(md5_block_data_order)
+#define HASH_BLOCK_DATA_ORDER   md5_block_data_order
+/*
+ * Little-endians (Intel and Alpha) feel better without this.
+ * It looks like memcpy does better job than generic
+ * md5_block_data_order on copying-n-aligning input data.
+ * But frankly speaking I didn't expect such result on Alpha.
+ * On the other hand I've got this with egcs-1.0.2 and if
+ * program is compiled with another (better?) compiler it
+ * might turn out other way around.
+ *
+ *                              <appro@fy.chalmers.se>
+ */
+#endif
 
-/* NOTE - c is not incremented as per l2c */
-#undef l2cn
-#define l2cn(l1,l2,c,n) { \
-                        c+=n; \
-                        switch (n) { \
-                        case 8: *(--(c))=(unsigned char)(((l2)>>24)&0xff); \
-                        case 7: *(--(c))=(unsigned char)(((l2)>>16)&0xff); \
-                        case 6: *(--(c))=(unsigned char)(((l2)>> 8)&0xff); \
-                        case 5: *(--(c))=(unsigned char)(((l2)    )&0xff); \
-                        case 4: *(--(c))=(unsigned char)(((l1)>>24)&0xff); \
-                        case 3: *(--(c))=(unsigned char)(((l1)>>16)&0xff); \
-                        case 2: *(--(c))=(unsigned char)(((l1)>> 8)&0xff); \
-                        case 1: *(--(c))=(unsigned char)(((l1)    )&0xff); \
-                                } \
-                        }
+#include "md32_common.h"
 
-/* A nice byte order reversal from Wei Dai <weidai@eskimo.com> */
-#if defined(WIN32)
-/* 5 instructions with rotate instruction, else 9 */
-#define Endian_Reverse32(a) \
-        { \
-        unsigned long l=(a); \
-        (a)=((ROTATE(l,8)&0x00FF00FF)|(ROTATE(l,24)&0xFF00FF00)); \
-        }
-#else
-/* 6 instructions with rotate instruction, else 8 */
-#define Endian_Reverse32(a) \
-        { \
-        unsigned long l=(a); \
-        l=(((l&0xFF00FF00)>>8L)|((l&0x00FF00FF)<<8L)); \
-        (a)=ROTATE(l,16L); \
-        }
-#endif
 /*
 #define F(x,y,z)        (((x) & (y))  |  ((~(x)) & (z)))
 #define G(x,y,z)        (((x) & (z))  |  ((y) & (~(z))))
 */
 
 /* As pointed out by Wei Dai <weidai@eskimo.com>, the above can be
- * simplified to the code below.  Wei attributes these optimisations
+ * simplified to the code below.  Wei attributes these optimizations
  * to Peter Gutmann's SHS code, and he attributes it to Rich Schroeppel.
  */
 #define F(b,c,d)        ((((c) ^ (d)) & (b)) ^ (d))
@@ -166,14 +151,6 @@
 #define H(b,c,d)        ((b) ^ (c) ^ (d))
 #define I(b,c,d)        (((~(d)) | (b)) ^ (c))
 
-#undef ROTATE
-#if defined(WIN32)
-#define ROTATE(a,n)     _lrotl(a,n)
-#else
-#define ROTATE(a,n)     (((a)<<(n))|(((a)&0xffffffff)>>(32-(n))))
-#endif
-
-
 #define R0(a,b,c,d,k,s,t) { \
         a+=((k)+(t)+F((b),(c),(d))); \
         a=ROTATE(a,s); \
diff --git a/src/lib/libcrypto/md5/md5_one.c b/src/lib/libcrypto/md5/md5_one.c
index ab6bb435f9..b89dec850d 100644
--- a/src/lib/libcrypto/md5/md5_one.c
+++ b/src/lib/libcrypto/md5/md5_one.c
@@ -57,19 +57,37 @@
  */
 
 #include <stdio.h>
-#include "md5_locl.h"
+#include <string.h>
+#include <openssl/md5.h>
 
-unsigned char *MD5(d, n, md)
-unsigned char *d;
-unsigned long n;
-unsigned char *md;
+#ifdef CHARSET_EBCDIC
+#include <openssl/ebcdic.h>
+#endif
+
+unsigned char *MD5(const unsigned char *d, unsigned long n, unsigned char *md)
         {
         MD5_CTX c;
         static unsigned char m[MD5_DIGEST_LENGTH];
 
         if (md == NULL) md=m;
         MD5_Init(&c);
+#ifndef CHARSET_EBCDIC
         MD5_Update(&c,d,n);
+#else
+        {
+                char temp[1024];
+                unsigned long chunk;
+
+                while (n > 0)
+                {
+                        chunk = (n > sizeof(temp)) ? sizeof(temp) : n;
+                        ebcdic2ascii(temp, d, chunk);
+                        MD5_Update(&c,temp,chunk);
+                        n -= chunk;
+                        d += chunk;
+                }
+        }
+#endif
         MD5_Final(md,&c);
         memset(&c,0,sizeof(c)); /* security consideration */
         return(md);