From 5a3c0a05c7f2c5d3c584b7c8d6aec836dd724c80 Mon Sep 17 00:00:00 2001 From: djm <> Date: Sat, 6 Sep 2008 12:15:56 +0000 Subject: import of OpenSSL 0.9.8h --- src/lib/libcrypto/sha/asm/sha1-586.pl | 472 ++++++-------------- src/lib/libcrypto/sha/asm/sha1-ia64.pl | 346 +++------------ src/lib/libcrypto/sha/asm/sha1-x86_64.pl | 242 +++++++++++ src/lib/libcrypto/sha/asm/sha512-ia64.pl | 672 +++++++++++++++++++++++++++++ src/lib/libcrypto/sha/asm/sha512-x86_64.pl | 344 +++++++++++++++ src/lib/libcrypto/sha/sha.h | 90 +++- src/lib/libcrypto/sha/sha1_one.c | 4 +- src/lib/libcrypto/sha/sha1dgst.c | 11 +- src/lib/libcrypto/sha/sha256.c | 282 ++++++++++++ src/lib/libcrypto/sha/sha512.c | 537 +++++++++++++++++++++++ src/lib/libcrypto/sha/sha_locl.h | 306 ++++++------- 11 files changed, 2478 insertions(+), 828 deletions(-) create mode 100755 src/lib/libcrypto/sha/asm/sha1-x86_64.pl create mode 100755 src/lib/libcrypto/sha/asm/sha512-ia64.pl create mode 100755 src/lib/libcrypto/sha/asm/sha512-x86_64.pl create mode 100644 src/lib/libcrypto/sha/sha256.c create mode 100644 src/lib/libcrypto/sha/sha512.c (limited to 'src/lib/libcrypto/sha') diff --git a/src/lib/libcrypto/sha/asm/sha1-586.pl b/src/lib/libcrypto/sha/asm/sha1-586.pl index 041acc0348..0b4dab2bd5 100644 --- a/src/lib/libcrypto/sha/asm/sha1-586.pl +++ b/src/lib/libcrypto/sha/asm/sha1-586.pl @@ -1,4 +1,16 @@ -#!/usr/local/bin/perl +#!/usr/bin/env perl + +# ==================================================================== +# [Re]written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== + +# "[Re]written" was achieved in two major overhauls. In 2004 BODY_* +# functions were re-implemented to address P4 performance issue [see +# commentary below], and in 2006 the rest was rewritten in order to +# gain freedom to liberate licensing terms. # It was noted that Intel IA-32 C compiler generates code which # performs ~30% *faster* on P4 CPU than original *hand-coded* @@ -9,7 +21,7 @@ # # compared with original compared with Intel cc # assembler impl. generated code -# Pentium -25% +37% +# Pentium -16% +48% # PIII/AMD +8% +16% # P4 +85%(!) +45% # @@ -17,174 +29,115 @@ # improvement on P4 outweights the loss and incorporate this # re-tuned code to 0.9.7 and later. # ---------------------------------------------------------------- -# Those who for any particular reason absolutely must score on -# Pentium can replace this module with one from 0.9.6 distribution. -# This "offer" shall be revoked the moment programming interface to -# this module is changed, in which case this paragraph should be -# removed. -# ---------------------------------------------------------------- # -$normal=0; - -push(@INC,"perlasm","../../perlasm"); +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +push(@INC,"${dir}","${dir}../../perlasm"); require "x86asm.pl"; &asm_init($ARGV[0],"sha1-586.pl",$ARGV[$#ARGV] eq "386"); $A="eax"; -$B="ecx"; -$C="ebx"; +$B="ebx"; +$C="ecx"; $D="edx"; $E="edi"; $T="esi"; $tmp1="ebp"; -$off=9*4; - -@K=(0x5a827999,0x6ed9eba1,0x8f1bbcdc,0xca62c1d6); - -&sha1_block_data("sha1_block_asm_data_order"); - -&asm_finish(); - -sub Nn - { - local($p)=@_; - local(%n)=($A,$T,$B,$A,$C,$B,$D,$C,$E,$D,$T,$E); - return($n{$p}); - } - -sub Np - { - local($p)=@_; - local(%n)=($A,$T,$B,$A,$C,$B,$D,$C,$E,$D,$T,$E); - local(%n)=($A,$B,$B,$C,$C,$D,$D,$E,$E,$T,$T,$A); - return($n{$p}); - } - -sub Na - { - local($n)=@_; - return( (($n )&0x0f), - (($n+ 2)&0x0f), - (($n+ 8)&0x0f), - (($n+13)&0x0f), - (($n+ 1)&0x0f)); - } - -sub X_expand - { - local($in)=@_; - - &comment("First, load the words onto the stack in network byte order"); - for ($i=0; $i<16; $i+=2) - { - &mov($A,&DWP(($i+0)*4,$in,"",0));# unless $i == 0; - &mov($B,&DWP(($i+1)*4,$in,"",0)); - &bswap($A); - &bswap($B); - &mov(&swtmp($i+0),$A); - &mov(&swtmp($i+1),$B); - } - - &comment("We now have the X array on the stack"); - &comment("starting at sp-4"); - } - -# Rules of engagement -# F is always trashable at the start, the running total. -# E becomes the next F so it can be trashed after it has been 'accumulated' -# F becomes A in the next round. We don't need to access it much. -# During the X update part, the result ends up in $X[$n0]. +@V=($A,$B,$C,$D,$E,$T); sub BODY_00_15 { - local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_; + local($n,$a,$b,$c,$d,$e,$f)=@_; &comment("00_15 $n"); - &mov($tmp1,$a); - &mov($f,$c); # f to hold F_00_19(b,c,d) + &mov($f,$c); # f to hold F_00_19(b,c,d) + if ($n==0) { &mov($tmp1,$a); } + else { &mov($a,$tmp1); } &rotl($tmp1,5); # tmp1=ROTATE(a,5) &xor($f,$d); - &and($f,$b); - &rotr($b,2); # b=ROTATE(b,30) &add($tmp1,$e); # tmp1+=e; - &mov($e,&swtmp($n)); # e becomes volatile and - # is loaded with xi - &xor($f,$d); # f holds F_00_19(b,c,d) - &lea($tmp1,&DWP($K,$tmp1,$e,1));# tmp1+=K_00_19+xi - - &add($f,$tmp1); # f+=tmp1 + &and($f,$b); + &mov($e,&swtmp($n%16)); # e becomes volatile and is loaded + # with xi, also note that e becomes + # f in next round... + &xor($f,$d); # f holds F_00_19(b,c,d) + &rotr($b,2); # b=ROTATE(b,30) + &lea($tmp1,&DWP(0x5a827999,$tmp1,$e)); # tmp1+=K_00_19+xi + + if ($n==15) { &add($f,$tmp1); } # f+=tmp1 + else { &add($tmp1,$f); } # f becomes a in next round } sub BODY_16_19 { - local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_; - local($n0,$n1,$n2,$n3,$np)=&Na($n); + local($n,$a,$b,$c,$d,$e,$f)=@_; &comment("16_19 $n"); - &mov($f,&swtmp($n1)); # f to hold Xupdate(xi,xa,xb,xc,xd) + &mov($f,&swtmp($n%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) &mov($tmp1,$c); # tmp1 to hold F_00_19(b,c,d) - &xor($f,&swtmp($n0)); + &xor($f,&swtmp(($n+2)%16)); &xor($tmp1,$d); - &xor($f,&swtmp($n2)); + &xor($f,&swtmp(($n+8)%16)); &and($tmp1,$b); # tmp1 holds F_00_19(b,c,d) - &xor($f,&swtmp($n3)); # f holds xa^xb^xc^xd - &rotr($b,2); # b=ROTATE(b,30) - &xor($tmp1,$d); # tmp1=F_00_19(b,c,d) - &rotl($f,1); # f=ROATE(f,1) - &mov(&swtmp($n0),$f); # xi=f - &lea($f,&DWP($K,$f,$e,1)); # f+=K_00_19+e + &rotr($b,2); # b=ROTATE(b,30) + &xor($f,&swtmp(($n+13)%16)); # f holds xa^xb^xc^xd + &rotl($f,1); # f=ROTATE(f,1) + &xor($tmp1,$d); # tmp1=F_00_19(b,c,d) + &mov(&swtmp($n%16),$f); # xi=f + &lea($f,&DWP(0x5a827999,$f,$e));# f+=K_00_19+e &mov($e,$a); # e becomes volatile - &add($f,$tmp1); # f+=F_00_19(b,c,d) - &rotl($e,5); # e=ROTATE(a,5) + &rotl($e,5); # e=ROTATE(a,5) + &add($f,$tmp1); # f+=F_00_19(b,c,d) &add($f,$e); # f+=ROTATE(a,5) } sub BODY_20_39 { - local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_; + local($n,$a,$b,$c,$d,$e,$f)=@_; + local $K=($n<40)?0x6ed9eba1:0xca62c1d6; &comment("20_39 $n"); - local($n0,$n1,$n2,$n3,$np)=&Na($n); - &mov($f,&swtmp($n0)); # f to hold Xupdate(xi,xa,xb,xc,xd) - &mov($tmp1,$b); # tmp1 to hold F_20_39(b,c,d) - &xor($f,&swtmp($n1)); - &rotr($b,2); # b=ROTATE(b,30) - &xor($f,&swtmp($n2)); - &xor($tmp1,$c); - &xor($f,&swtmp($n3)); # f holds xa^xb^xc^xd - &xor($tmp1,$d); # tmp1 holds F_20_39(b,c,d) + &mov($tmp1,$b); # tmp1 to hold F_20_39(b,c,d) + &mov($f,&swtmp($n%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) + &rotr($b,2); # b=ROTATE(b,30) + &xor($f,&swtmp(($n+2)%16)); + &xor($tmp1,$c); + &xor($f,&swtmp(($n+8)%16)); + &xor($tmp1,$d); # tmp1 holds F_20_39(b,c,d) + &xor($f,&swtmp(($n+13)%16)); # f holds xa^xb^xc^xd &rotl($f,1); # f=ROTATE(f,1) - &mov(&swtmp($n0),$f); # xi=f - &lea($f,&DWP($K,$f,$e,1)); # f+=K_20_39+e + &add($tmp1,$e); + &mov(&swtmp($n%16),$f); # xi=f &mov($e,$a); # e becomes volatile &rotl($e,5); # e=ROTATE(a,5) - &add($f,$tmp1); # f+=F_20_39(b,c,d) + &lea($f,&DWP($K,$f,$tmp1)); # f+=K_20_39+e &add($f,$e); # f+=ROTATE(a,5) } sub BODY_40_59 { - local($pos,$K,$X,$n,$a,$b,$c,$d,$e,$f)=@_; + local($n,$a,$b,$c,$d,$e,$f)=@_; &comment("40_59 $n"); - local($n0,$n1,$n2,$n3,$np)=&Na($n); - &mov($f,&swtmp($n0)); # f to hold Xupdate(xi,xa,xb,xc,xd) + &mov($f,&swtmp($n%16)); # f to hold Xupdate(xi,xa,xb,xc,xd) + &mov($tmp1,&swtmp(($n+2)%16)); + &xor($f,$tmp1); + &mov($tmp1,&swtmp(($n+8)%16)); + &xor($f,$tmp1); + &mov($tmp1,&swtmp(($n+13)%16)); + &xor($f,$tmp1); # f holds xa^xb^xc^xd &mov($tmp1,$b); # tmp1 to hold F_40_59(b,c,d) - &xor($f,&swtmp($n1)); + &rotl($f,1); # f=ROTATE(f,1) &or($tmp1,$c); - &xor($f,&swtmp($n2)); + &mov(&swtmp($n%16),$f); # xi=f &and($tmp1,$d); - &xor($f,&swtmp($n3)); # f holds xa^xb^xc^xd - &rotl($f,1); # f=ROTATE(f,1) - &mov(&swtmp($n0),$f); # xi=f - &lea($f,&DWP($K,$f,$e,1)); # f+=K_40_59+e + &lea($f,&DWP(0x8f1bbcdc,$f,$e));# f+=K_40_59+e &mov($e,$b); # e becomes volatile and is used # to calculate F_40_59(b,c,d) &rotr($b,2); # b=ROTATE(b,30) @@ -192,234 +145,75 @@ sub BODY_40_59 &or($tmp1,$e); # tmp1 holds F_40_59(b,c,d) &mov($e,$a); &rotl($e,5); # e=ROTATE(a,5) - &add($tmp1,$e); # tmp1+=ROTATE(a,5) - &add($f,$tmp1); # f+=tmp1; - } - -sub BODY_60_79 - { - &BODY_20_39(@_); + &add($f,$tmp1); # f+=tmp1; + &add($f,$e); # f+=ROTATE(a,5) } -sub sha1_block_host - { - local($name, $sclabel)=@_; +&function_begin("sha1_block_data_order",16); + &mov($tmp1,&wparam(0)); # SHA_CTX *c + &mov($T,&wparam(1)); # const void *input + &mov($A,&wparam(2)); # size_t num + &stack_push(16); # allocate X[16] + &shl($A,6); + &add($A,$T); + &mov(&wparam(2),$A); # pointer beyond the end of input + &mov($E,&DWP(16,$tmp1));# pre-load E - &function_begin_B($name,""); + &set_label("loop",16); - # parameter 1 is the MD5_CTX structure. - # A 0 - # B 4 - # C 8 - # D 12 - # E 16 - - &mov("ecx", &wparam(2)); - &push("esi"); - &shl("ecx",6); - &mov("esi", &wparam(1)); - &push("ebp"); - &add("ecx","esi"); # offset to leave on - &push("ebx"); - &mov("ebp", &wparam(0)); - &push("edi"); - &mov($D, &DWP(12,"ebp","",0)); - &stack_push(18+9); - &mov($E, &DWP(16,"ebp","",0)); - &mov($C, &DWP( 8,"ebp","",0)); - &mov(&swtmp(17),"ecx"); - - &comment("First we need to setup the X array"); - - for ($i=0; $i<16; $i+=2) + # copy input chunk to X, but reversing byte order! + for ($i=0; $i<16; $i+=4) { - &mov($A,&DWP(($i+0)*4,"esi","",0));# unless $i == 0; - &mov($B,&DWP(($i+1)*4,"esi","",0)); + &mov($A,&DWP(4*($i+0),$T)); + &mov($B,&DWP(4*($i+1),$T)); + &mov($C,&DWP(4*($i+2),$T)); + &mov($D,&DWP(4*($i+3),$T)); + &bswap($A); + &bswap($B); + &bswap($C); + &bswap($D); &mov(&swtmp($i+0),$A); - &mov(&swtmp($i+1),$B); + &mov(&swtmp($i+1),$B); + &mov(&swtmp($i+2),$C); + &mov(&swtmp($i+3),$D); } - &jmp($sclabel); - &function_end_B($name); - } - - -sub sha1_block_data - { - local($name)=@_; - - &function_begin_B($name,""); - - # parameter 1 is the MD5_CTX structure. - # A 0 - # B 4 - # C 8 - # D 12 - # E 16 - - &mov("ecx", &wparam(2)); - &push("esi"); - &shl("ecx",6); - &mov("esi", &wparam(1)); - &push("ebp"); - &add("ecx","esi"); # offset to leave on - &push("ebx"); - &mov("ebp", &wparam(0)); - &push("edi"); - &mov($D, &DWP(12,"ebp","",0)); - &stack_push(18+9); - &mov($E, &DWP(16,"ebp","",0)); - &mov($C, &DWP( 8,"ebp","",0)); - &mov(&swtmp(17),"ecx"); - - &comment("First we need to setup the X array"); - - &set_label("start") unless $normal; - - &X_expand("esi"); - &mov(&wparam(1),"esi"); - - &set_label("shortcut", 0, 1); - &comment(""); - &comment("Start processing"); - - # odd start - &mov($A, &DWP( 0,"ebp","",0)); - &mov($B, &DWP( 4,"ebp","",0)); - $X="esp"; - &BODY_00_15(-2,$K[0],$X, 0,$A,$B,$C,$D,$E,$T); - &BODY_00_15( 0,$K[0],$X, 1,$T,$A,$B,$C,$D,$E); - &BODY_00_15( 0,$K[0],$X, 2,$E,$T,$A,$B,$C,$D); - &BODY_00_15( 0,$K[0],$X, 3,$D,$E,$T,$A,$B,$C); - &BODY_00_15( 0,$K[0],$X, 4,$C,$D,$E,$T,$A,$B); - &BODY_00_15( 0,$K[0],$X, 5,$B,$C,$D,$E,$T,$A); - &BODY_00_15( 0,$K[0],$X, 6,$A,$B,$C,$D,$E,$T); - &BODY_00_15( 0,$K[0],$X, 7,$T,$A,$B,$C,$D,$E); - &BODY_00_15( 0,$K[0],$X, 8,$E,$T,$A,$B,$C,$D); - &BODY_00_15( 0,$K[0],$X, 9,$D,$E,$T,$A,$B,$C); - &BODY_00_15( 0,$K[0],$X,10,$C,$D,$E,$T,$A,$B); - &BODY_00_15( 0,$K[0],$X,11,$B,$C,$D,$E,$T,$A); - &BODY_00_15( 0,$K[0],$X,12,$A,$B,$C,$D,$E,$T); - &BODY_00_15( 0,$K[0],$X,13,$T,$A,$B,$C,$D,$E); - &BODY_00_15( 0,$K[0],$X,14,$E,$T,$A,$B,$C,$D); - &BODY_00_15( 1,$K[0],$X,15,$D,$E,$T,$A,$B,$C); - &BODY_16_19(-1,$K[0],$X,16,$C,$D,$E,$T,$A,$B); - &BODY_16_19( 0,$K[0],$X,17,$B,$C,$D,$E,$T,$A); - &BODY_16_19( 0,$K[0],$X,18,$A,$B,$C,$D,$E,$T); - &BODY_16_19( 1,$K[0],$X,19,$T,$A,$B,$C,$D,$E); - - &BODY_20_39(-1,$K[1],$X,20,$E,$T,$A,$B,$C,$D); - &BODY_20_39( 0,$K[1],$X,21,$D,$E,$T,$A,$B,$C); - &BODY_20_39( 0,$K[1],$X,22,$C,$D,$E,$T,$A,$B); - &BODY_20_39( 0,$K[1],$X,23,$B,$C,$D,$E,$T,$A); - &BODY_20_39( 0,$K[1],$X,24,$A,$B,$C,$D,$E,$T); - &BODY_20_39( 0,$K[1],$X,25,$T,$A,$B,$C,$D,$E); - &BODY_20_39( 0,$K[1],$X,26,$E,$T,$A,$B,$C,$D); - &BODY_20_39( 0,$K[1],$X,27,$D,$E,$T,$A,$B,$C); - &BODY_20_39( 0,$K[1],$X,28,$C,$D,$E,$T,$A,$B); - &BODY_20_39( 0,$K[1],$X,29,$B,$C,$D,$E,$T,$A); - &BODY_20_39( 0,$K[1],$X,30,$A,$B,$C,$D,$E,$T); - &BODY_20_39( 0,$K[1],$X,31,$T,$A,$B,$C,$D,$E); - &BODY_20_39( 0,$K[1],$X,32,$E,$T,$A,$B,$C,$D); - &BODY_20_39( 0,$K[1],$X,33,$D,$E,$T,$A,$B,$C); - &BODY_20_39( 0,$K[1],$X,34,$C,$D,$E,$T,$A,$B); - &BODY_20_39( 0,$K[1],$X,35,$B,$C,$D,$E,$T,$A); - &BODY_20_39( 0,$K[1],$X,36,$A,$B,$C,$D,$E,$T); - &BODY_20_39( 0,$K[1],$X,37,$T,$A,$B,$C,$D,$E); - &BODY_20_39( 0,$K[1],$X,38,$E,$T,$A,$B,$C,$D); - &BODY_20_39( 1,$K[1],$X,39,$D,$E,$T,$A,$B,$C); - - &BODY_40_59(-1,$K[2],$X,40,$C,$D,$E,$T,$A,$B); - &BODY_40_59( 0,$K[2],$X,41,$B,$C,$D,$E,$T,$A); - &BODY_40_59( 0,$K[2],$X,42,$A,$B,$C,$D,$E,$T); - &BODY_40_59( 0,$K[2],$X,43,$T,$A,$B,$C,$D,$E); - &BODY_40_59( 0,$K[2],$X,44,$E,$T,$A,$B,$C,$D); - &BODY_40_59( 0,$K[2],$X,45,$D,$E,$T,$A,$B,$C); - &BODY_40_59( 0,$K[2],$X,46,$C,$D,$E,$T,$A,$B); - &BODY_40_59( 0,$K[2],$X,47,$B,$C,$D,$E,$T,$A); - &BODY_40_59( 0,$K[2],$X,48,$A,$B,$C,$D,$E,$T); - &BODY_40_59( 0,$K[2],$X,49,$T,$A,$B,$C,$D,$E); - &BODY_40_59( 0,$K[2],$X,50,$E,$T,$A,$B,$C,$D); - &BODY_40_59( 0,$K[2],$X,51,$D,$E,$T,$A,$B,$C); - &BODY_40_59( 0,$K[2],$X,52,$C,$D,$E,$T,$A,$B); - &BODY_40_59( 0,$K[2],$X,53,$B,$C,$D,$E,$T,$A); - &BODY_40_59( 0,$K[2],$X,54,$A,$B,$C,$D,$E,$T); - &BODY_40_59( 0,$K[2],$X,55,$T,$A,$B,$C,$D,$E); - &BODY_40_59( 0,$K[2],$X,56,$E,$T,$A,$B,$C,$D); - &BODY_40_59( 0,$K[2],$X,57,$D,$E,$T,$A,$B,$C); - &BODY_40_59( 0,$K[2],$X,58,$C,$D,$E,$T,$A,$B); - &BODY_40_59( 1,$K[2],$X,59,$B,$C,$D,$E,$T,$A); - - &BODY_60_79(-1,$K[3],$X,60,$A,$B,$C,$D,$E,$T); - &BODY_60_79( 0,$K[3],$X,61,$T,$A,$B,$C,$D,$E); - &BODY_60_79( 0,$K[3],$X,62,$E,$T,$A,$B,$C,$D); - &BODY_60_79( 0,$K[3],$X,63,$D,$E,$T,$A,$B,$C); - &BODY_60_79( 0,$K[3],$X,64,$C,$D,$E,$T,$A,$B); - &BODY_60_79( 0,$K[3],$X,65,$B,$C,$D,$E,$T,$A); - &BODY_60_79( 0,$K[3],$X,66,$A,$B,$C,$D,$E,$T); - &BODY_60_79( 0,$K[3],$X,67,$T,$A,$B,$C,$D,$E); - &BODY_60_79( 0,$K[3],$X,68,$E,$T,$A,$B,$C,$D); - &BODY_60_79( 0,$K[3],$X,69,$D,$E,$T,$A,$B,$C); - &BODY_60_79( 0,$K[3],$X,70,$C,$D,$E,$T,$A,$B); - &BODY_60_79( 0,$K[3],$X,71,$B,$C,$D,$E,$T,$A); - &BODY_60_79( 0,$K[3],$X,72,$A,$B,$C,$D,$E,$T); - &BODY_60_79( 0,$K[3],$X,73,$T,$A,$B,$C,$D,$E); - &BODY_60_79( 0,$K[3],$X,74,$E,$T,$A,$B,$C,$D); - &BODY_60_79( 0,$K[3],$X,75,$D,$E,$T,$A,$B,$C); - &BODY_60_79( 0,$K[3],$X,76,$C,$D,$E,$T,$A,$B); - &BODY_60_79( 0,$K[3],$X,77,$B,$C,$D,$E,$T,$A); - &BODY_60_79( 0,$K[3],$X,78,$A,$B,$C,$D,$E,$T); - &BODY_60_79( 2,$K[3],$X,79,$T,$A,$B,$C,$D,$E); - - &comment("End processing"); - &comment(""); - # D is the tmp value - - # E -> A - # T -> B - # A -> C - # B -> D - # C -> E - # D -> T - - &mov($tmp1,&wparam(0)); - - &mov($D, &DWP(12,$tmp1,"",0)); - &add($D,$B); - &mov($B, &DWP( 4,$tmp1,"",0)); - &add($B,$T); - &mov($T, $A); - &mov($A, &DWP( 0,$tmp1,"",0)); - &mov(&DWP(12,$tmp1,"",0),$D); - - &add($A,$E); - &mov($E, &DWP(16,$tmp1,"",0)); - &add($E,$C); - &mov($C, &DWP( 8,$tmp1,"",0)); - &add($C,$T); - - &mov(&DWP( 0,$tmp1,"",0),$A); - &mov("esi",&wparam(1)); - &mov(&DWP( 8,$tmp1,"",0),$C); - &add("esi",64); - &mov("eax",&swtmp(17)); - &mov(&DWP(16,$tmp1,"",0),$E); - &cmp("esi","eax"); - &mov(&DWP( 4,$tmp1,"",0),$B); - &jb(&label("start")); - - &stack_pop(18+9); - &pop("edi"); - &pop("ebx"); - &pop("ebp"); - &pop("esi"); - &ret(); - - # keep a note of shortcut label so it can be used outside - # block. - my $sclabel = &label("shortcut"); - - &function_end_B($name); - # Putting this here avoids problems with MASM in debugging mode - &sha1_block_host("sha1_block_asm_host_order", $sclabel); - } + &mov(&wparam(1),$T); # redundant in 1st spin + + &mov($A,&DWP(0,$tmp1)); # load SHA_CTX + &mov($B,&DWP(4,$tmp1)); + &mov($C,&DWP(8,$tmp1)); + &mov($D,&DWP(12,$tmp1)); + # E is pre-loaded + + for($i=0;$i<16;$i++) { &BODY_00_15($i,@V); unshift(@V,pop(@V)); } + for(;$i<20;$i++) { &BODY_16_19($i,@V); unshift(@V,pop(@V)); } + for(;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); } + for(;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); } + for(;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); } + + (($V[5] eq $D) and ($V[0] eq $E)) or die; # double-check + + &mov($tmp1,&wparam(0)); # re-load SHA_CTX* + &mov($D,&wparam(1)); # D is last "T" and is discarded + + &add($E,&DWP(0,$tmp1)); # E is last "A"... + &add($T,&DWP(4,$tmp1)); + &add($A,&DWP(8,$tmp1)); + &add($B,&DWP(12,$tmp1)); + &add($C,&DWP(16,$tmp1)); + + &mov(&DWP(0,$tmp1),$E); # update SHA_CTX + &add($D,64); # advance input pointer + &mov(&DWP(4,$tmp1),$T); + &cmp($D,&wparam(2)); # have we reached the end yet? + &mov(&DWP(8,$tmp1),$A); + &mov($E,$C); # C is last "E" which needs to be "pre-loaded" + &mov(&DWP(12,$tmp1),$B); + &mov($T,$D); # input pointer + &mov(&DWP(16,$tmp1),$C); + &jb(&label("loop")); + + &stack_pop(16); +&function_end("sha1_block_data_order"); +&asm_finish(); diff --git a/src/lib/libcrypto/sha/asm/sha1-ia64.pl b/src/lib/libcrypto/sha/asm/sha1-ia64.pl index cb9dfad124..aa18c1089b 100644 --- a/src/lib/libcrypto/sha/asm/sha1-ia64.pl +++ b/src/lib/libcrypto/sha/asm/sha1-ia64.pl @@ -2,8 +2,9 @@ # # ==================================================================== # Written by Andy Polyakov for the OpenSSL -# project. Rights for redistribution and usage in source and binary -# forms are granted according to the OpenSSL license. +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. # ==================================================================== # # Eternal question is what's wrong with compiler generated code? The @@ -11,15 +12,10 @@ # to perform rotations by maintaining copy of 32-bit value in upper # bits of 64-bit register. Just follow mux2 and shrp instructions... # Performance under big-endian OS such as HP-UX is 179MBps*1GHz, which -# is >50% better than HP C and >2x better than gcc. As of this moment -# performance under little-endian OS such as Linux and Windows will be -# a bit lower, because data has to be picked in reverse byte-order. -# It's possible to resolve this issue by implementing third function, -# sha1_block_asm_data_order_aligned, which would temporarily flip -# BE field in User Mask register... +# is >50% better than HP C and >2x better than gcc. $code=<<___; -.ident \"sha1-ia64.s, version 1.0\" +.ident \"sha1-ia64.s, version 1.2\" .ident \"IA-64 ISA artwork by Andy Polyakov \" .explicit @@ -55,63 +51,55 @@ else { sub BODY_00_15 { local *code=shift; -local ($i,$a,$b,$c,$d,$e,$f,$unaligned)=@_; +local ($i,$a,$b,$c,$d,$e,$f)=@_; -if ($unaligned) { - $code.=<<___; -{ .mmi; ld1 tmp0=[inp],2 // MSB - ld1 tmp1=[tmp3],2 };; -{ .mmi; ld1 tmp2=[inp],2 - ld1 $X[$i&0xf]=[tmp3],2 // LSB - dep tmp1=tmp0,tmp1,8,8 };; -{ .mii; cmp.ne p16,p0=r0,r0 // no misaligned prefetch - dep $X[$i&0xf]=tmp2,$X[$i&0xf],8,8;; - dep $X[$i&0xf]=tmp1,$X[$i&0xf],16,16 };; -{ .mmi; nop.m 0 -___ - } -elsif ($i<15) { - $code.=<<___; -{ .mmi; ld4 $X[($i+1)&0xf]=[inp],4 // prefetch -___ - } -else { - $code.=<<___; -{ .mmi; nop.m 0 +$code.=<<___ if ($i==0); +{ .mmi; ld1 $X[$i&0xf]=[inp],2 // MSB + ld1 tmp2=[tmp3],2 };; +{ .mmi; ld1 tmp0=[inp],2 + ld1 tmp4=[tmp3],2 // LSB + dep $X[$i&0xf]=$X[$i&0xf],tmp2,8,8 };; ___ - } if ($i<15) { $code.=<<___; - and tmp0=$c,$b - dep.z tmp5=$a,5,27 } // a<<5 +{ .mmi; ld1 $X[($i+1)&0xf]=[inp],2 // +1 + dep tmp1=tmp0,tmp4,8,8 };; +{ .mmi; ld1 tmp2=[tmp3],2 // +1 + and tmp4=$c,$b + dep $X[$i&0xf]=$X[$i&0xf],tmp1,16,16 } //;; { .mmi; andcm tmp1=$d,$b - add tmp4=$e,$K_00_19 };; -{ .mmi; or tmp0=tmp0,tmp1 // F_00_19(b,c,d)=(b&c)|(~b&d) - add $f=tmp4,$X[$i&0xf] // f=xi+e+K_00_19 + add tmp0=$e,$K_00_19 + dep.z tmp5=$a,5,27 };; // a<<5 +{ .mmi; or tmp4=tmp4,tmp1 // F_00_19(b,c,d)=(b&c)|(~b&d) + add $f=tmp0,$X[$i&0xf] // f=xi+e+K_00_19 extr.u tmp1=$a,27,5 };; // a>>27 -{ .mib; add $f=$f,tmp0 // f+=F_00_19(b,c,d) +{ .mmi; ld1 tmp0=[inp],2 // +1 + add $f=$f,tmp4 // f+=F_00_19(b,c,d) shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30) -{ .mib; or tmp1=tmp1,tmp5 // ROTATE(a,5) +{ .mmi; ld1 tmp4=[tmp3],2 // +1 + or tmp5=tmp1,tmp5 // ROTATE(a,5) mux2 tmp6=$a,0x44 };; // see b in next iteration -{ .mii; add $f=$f,tmp1 // f+=ROTATE(a,5) - mux2 $X[$i&0xf]=$X[$i&0xf],0x44 - nop.i 0 };; +{ .mii; add $f=$f,tmp5 // f+=ROTATE(a,5) + dep $X[($i+1)&0xf]=$X[($i+1)&0xf],tmp2,8,8 // +1 + mux2 $X[$i&0xf]=$X[$i&0xf],0x44 } //;; ___ } else { $code.=<<___; - and tmp0=$c,$b - dep.z tmp5=$a,5,27 } // a<<5 ;;? +{ .mii; and tmp3=$c,$b + dep tmp1=tmp0,tmp4,8,8;; + dep $X[$i&0xf]=$X[$i&0xf],tmp1,16,16 } //;; { .mmi; andcm tmp1=$d,$b - add tmp4=$e,$K_00_19 };; -{ .mmi; or tmp0=tmp0,tmp1 // F_00_19(b,c,d)=(b&c)|(~b&d) - add $f=tmp4,$X[$i&0xf] // f=xi+e+K_00_19 + add tmp0=$e,$K_00_19 + dep.z tmp5=$a,5,27 };; // a<<5 +{ .mmi; or tmp4=tmp3,tmp1 // F_00_19(b,c,d)=(b&c)|(~b&d) + add $f=tmp0,$X[$i&0xf] // f=xi+e+K_00_19 extr.u tmp1=$a,27,5 } // a>>27 { .mmi; xor tmp2=$X[($i+0+1)&0xf],$X[($i+2+1)&0xf] // +1 xor tmp3=$X[($i+8+1)&0xf],$X[($i+13+1)&0xf] // +1 nop.i 0 };; -{ .mmi; add $f=$f,tmp0 // f+=F_00_19(b,c,d) +{ .mmi; add $f=$f,tmp4 // f+=F_00_19(b,c,d) xor tmp2=tmp2,tmp3 // +1 shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30) { .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5) @@ -190,9 +178,7 @@ $code.=<<___; extr.u tmp1=$a,27,5 } // a>>27 { .mib; add $f=$f,tmp4 // f+=e+K_20_39 add $h1=$h1,$a };; // wrap up -{ .mmi; -(p16) ld4.s $X[0]=[inp],4 // non-faulting prefetch - add $f=$f,tmp0 // f+=F_20_39(b,c,d) +{ .mmi; add $f=$f,tmp0 // f+=F_20_39(b,c,d) shrp $b=tmp6,tmp6,2 } // b=ROTATE(b,30) ;;? { .mmi; or tmp1=tmp1,tmp5 // ROTATE(a,5) add $h3=$h3,$c };; // wrap up @@ -245,172 +231,15 @@ tmp3=r11; ctx=r32; // in0 inp=r33; // in1 -// void sha1_block_asm_host_order(SHA_CTX *c,const void *p,size_t num); -.global sha1_block_asm_host_order# -.proc sha1_block_asm_host_order# +// void sha1_block_data_order(SHA_CTX *c,const void *p,size_t num); +.global sha1_block_data_order# +.proc sha1_block_data_order# .align 32 -sha1_block_asm_host_order: +sha1_block_data_order: .prologue - .fframe 0 - .save ar.pfs,r0 - .save ar.lc,r3 { .mmi; alloc tmp1=ar.pfs,3,15,0,0 $ADDP tmp0=4,ctx - mov r3=ar.lc } -{ .mmi; $ADDP ctx=0,ctx - $ADDP inp=0,inp - mov r2=pr };; -tmp4=in2; -tmp5=loc13; -tmp6=loc14; - .body -{ .mlx; ld4 $h0=[ctx],8 - movl $K_00_19=0x5a827999 } -{ .mlx; ld4 $h1=[tmp0],8 - movl $K_20_39=0x6ed9eba1 };; -{ .mlx; ld4 $h2=[ctx],8 - movl $K_40_59=0x8f1bbcdc } -{ .mlx; ld4 $h3=[tmp0] - movl $K_60_79=0xca62c1d6 };; -{ .mmi; ld4 $h4=[ctx],-16 - add in2=-1,in2 // adjust num for ar.lc - mov ar.ec=1 };; -{ .mmi; ld4 $X[0]=[inp],4 // prefetch - cmp.ne p16,p0=r0,in2 // prefecth at loop end - mov ar.lc=in2 };; // brp.loop.imp: too far - -.Lhtop: -{ .mmi; mov $A=$h0 - mov $B=$h1 - mux2 tmp6=$h1,0x44 } -{ .mmi; mov $C=$h2 - mov $D=$h3 - mov $E=$h4 };; - -___ - - &BODY_00_15(\$code, 0,$A,$B,$C,$D,$E,$T); - &BODY_00_15(\$code, 1,$T,$A,$B,$C,$D,$E); - &BODY_00_15(\$code, 2,$E,$T,$A,$B,$C,$D); - &BODY_00_15(\$code, 3,$D,$E,$T,$A,$B,$C); - &BODY_00_15(\$code, 4,$C,$D,$E,$T,$A,$B); - &BODY_00_15(\$code, 5,$B,$C,$D,$E,$T,$A); - &BODY_00_15(\$code, 6,$A,$B,$C,$D,$E,$T); - &BODY_00_15(\$code, 7,$T,$A,$B,$C,$D,$E); - &BODY_00_15(\$code, 8,$E,$T,$A,$B,$C,$D); - &BODY_00_15(\$code, 9,$D,$E,$T,$A,$B,$C); - &BODY_00_15(\$code,10,$C,$D,$E,$T,$A,$B); - &BODY_00_15(\$code,11,$B,$C,$D,$E,$T,$A); - &BODY_00_15(\$code,12,$A,$B,$C,$D,$E,$T); - &BODY_00_15(\$code,13,$T,$A,$B,$C,$D,$E); - &BODY_00_15(\$code,14,$E,$T,$A,$B,$C,$D); - &BODY_00_15(\$code,15,$D,$E,$T,$A,$B,$C); - - &BODY_16_19(\$code,16,$C,$D,$E,$T,$A,$B); - &BODY_16_19(\$code,17,$B,$C,$D,$E,$T,$A); - &BODY_16_19(\$code,18,$A,$B,$C,$D,$E,$T); - &BODY_16_19(\$code,19,$T,$A,$B,$C,$D,$E); - - &BODY_20_39(\$code,20,$E,$T,$A,$B,$C,$D); - &BODY_20_39(\$code,21,$D,$E,$T,$A,$B,$C); - &BODY_20_39(\$code,22,$C,$D,$E,$T,$A,$B); - &BODY_20_39(\$code,23,$B,$C,$D,$E,$T,$A); - &BODY_20_39(\$code,24,$A,$B,$C,$D,$E,$T); - &BODY_20_39(\$code,25,$T,$A,$B,$C,$D,$E); - &BODY_20_39(\$code,26,$E,$T,$A,$B,$C,$D); - &BODY_20_39(\$code,27,$D,$E,$T,$A,$B,$C); - &BODY_20_39(\$code,28,$C,$D,$E,$T,$A,$B); - &BODY_20_39(\$code,29,$B,$C,$D,$E,$T,$A); - &BODY_20_39(\$code,30,$A,$B,$C,$D,$E,$T); - &BODY_20_39(\$code,31,$T,$A,$B,$C,$D,$E); - &BODY_20_39(\$code,32,$E,$T,$A,$B,$C,$D); - &BODY_20_39(\$code,33,$D,$E,$T,$A,$B,$C); - &BODY_20_39(\$code,34,$C,$D,$E,$T,$A,$B); - &BODY_20_39(\$code,35,$B,$C,$D,$E,$T,$A); - &BODY_20_39(\$code,36,$A,$B,$C,$D,$E,$T); - &BODY_20_39(\$code,37,$T,$A,$B,$C,$D,$E); - &BODY_20_39(\$code,38,$E,$T,$A,$B,$C,$D); - &BODY_20_39(\$code,39,$D,$E,$T,$A,$B,$C); - - &BODY_40_59(\$code,40,$C,$D,$E,$T,$A,$B); - &BODY_40_59(\$code,41,$B,$C,$D,$E,$T,$A); - &BODY_40_59(\$code,42,$A,$B,$C,$D,$E,$T); - &BODY_40_59(\$code,43,$T,$A,$B,$C,$D,$E); - &BODY_40_59(\$code,44,$E,$T,$A,$B,$C,$D); - &BODY_40_59(\$code,45,$D,$E,$T,$A,$B,$C); - &BODY_40_59(\$code,46,$C,$D,$E,$T,$A,$B); - &BODY_40_59(\$code,47,$B,$C,$D,$E,$T,$A); - &BODY_40_59(\$code,48,$A,$B,$C,$D,$E,$T); - &BODY_40_59(\$code,49,$T,$A,$B,$C,$D,$E); - &BODY_40_59(\$code,50,$E,$T,$A,$B,$C,$D); - &BODY_40_59(\$code,51,$D,$E,$T,$A,$B,$C); - &BODY_40_59(\$code,52,$C,$D,$E,$T,$A,$B); - &BODY_40_59(\$code,53,$B,$C,$D,$E,$T,$A); - &BODY_40_59(\$code,54,$A,$B,$C,$D,$E,$T); - &BODY_40_59(\$code,55,$T,$A,$B,$C,$D,$E); - &BODY_40_59(\$code,56,$E,$T,$A,$B,$C,$D); - &BODY_40_59(\$code,57,$D,$E,$T,$A,$B,$C); - &BODY_40_59(\$code,58,$C,$D,$E,$T,$A,$B); - &BODY_40_59(\$code,59,$B,$C,$D,$E,$T,$A); - - &BODY_60_79(\$code,60,$A,$B,$C,$D,$E,$T); - &BODY_60_79(\$code,61,$T,$A,$B,$C,$D,$E); - &BODY_60_79(\$code,62,$E,$T,$A,$B,$C,$D); - &BODY_60_79(\$code,63,$D,$E,$T,$A,$B,$C); - &BODY_60_79(\$code,64,$C,$D,$E,$T,$A,$B); - &BODY_60_79(\$code,65,$B,$C,$D,$E,$T,$A); - &BODY_60_79(\$code,66,$A,$B,$C,$D,$E,$T); - &BODY_60_79(\$code,67,$T,$A,$B,$C,$D,$E); - &BODY_60_79(\$code,68,$E,$T,$A,$B,$C,$D); - &BODY_60_79(\$code,69,$D,$E,$T,$A,$B,$C); - &BODY_60_79(\$code,70,$C,$D,$E,$T,$A,$B); - &BODY_60_79(\$code,71,$B,$C,$D,$E,$T,$A); - &BODY_60_79(\$code,72,$A,$B,$C,$D,$E,$T); - &BODY_60_79(\$code,73,$T,$A,$B,$C,$D,$E); - &BODY_60_79(\$code,74,$E,$T,$A,$B,$C,$D); - &BODY_60_79(\$code,75,$D,$E,$T,$A,$B,$C); - &BODY_60_79(\$code,76,$C,$D,$E,$T,$A,$B); - &BODY_60_79(\$code,77,$B,$C,$D,$E,$T,$A); - &BODY_60_79(\$code,78,$A,$B,$C,$D,$E,$T); - &BODY_60_79(\$code,79,$T,$A,$B,$C,$D,$E); - -$code.=<<___; -{ .mmb; add $h0=$h0,$E - nop.m 0 - br.ctop.dptk.many .Lhtop };; -.Lhend: -{ .mmi; add tmp0=4,ctx - mov ar.lc=r3 };; -{ .mmi; st4 [ctx]=$h0,8 - st4 [tmp0]=$h1,8 };; -{ .mmi; st4 [ctx]=$h2,8 - st4 [tmp0]=$h3 };; -{ .mib; st4 [ctx]=$h4,-16 - mov pr=r2,0x1ffff - br.ret.sptk.many b0 };; -.endp sha1_block_asm_host_order# -___ - - -$code.=<<___; -// void sha1_block_asm_data_order(SHA_CTX *c,const void *p,size_t num); -.global sha1_block_asm_data_order# -.proc sha1_block_asm_data_order# -.align 32 -sha1_block_asm_data_order: -___ -$code.=<<___ if ($big_endian); -{ .mmi; and r2=3,inp };; -{ .mib; cmp.eq p6,p0=r0,r2 -(p6) br.dptk.many sha1_block_asm_host_order };; -___ -$code.=<<___; - .prologue - .fframe 0 - .save ar.pfs,r0 .save ar.lc,r3 -{ .mmi; alloc tmp1=ar.pfs,3,15,0,0 - $ADDP tmp0=4,ctx mov r3=ar.lc } { .mmi; $ADDP ctx=0,ctx $ADDP inp=0,inp @@ -444,90 +273,16 @@ tmp6=loc14; ___ - &BODY_00_15(\$code, 0,$A,$B,$C,$D,$E,$T,1); - &BODY_00_15(\$code, 1,$T,$A,$B,$C,$D,$E,1); - &BODY_00_15(\$code, 2,$E,$T,$A,$B,$C,$D,1); - &BODY_00_15(\$code, 3,$D,$E,$T,$A,$B,$C,1); - &BODY_00_15(\$code, 4,$C,$D,$E,$T,$A,$B,1); - &BODY_00_15(\$code, 5,$B,$C,$D,$E,$T,$A,1); - &BODY_00_15(\$code, 6,$A,$B,$C,$D,$E,$T,1); - &BODY_00_15(\$code, 7,$T,$A,$B,$C,$D,$E,1); - &BODY_00_15(\$code, 8,$E,$T,$A,$B,$C,$D,1); - &BODY_00_15(\$code, 9,$D,$E,$T,$A,$B,$C,1); - &BODY_00_15(\$code,10,$C,$D,$E,$T,$A,$B,1); - &BODY_00_15(\$code,11,$B,$C,$D,$E,$T,$A,1); - &BODY_00_15(\$code,12,$A,$B,$C,$D,$E,$T,1); - &BODY_00_15(\$code,13,$T,$A,$B,$C,$D,$E,1); - &BODY_00_15(\$code,14,$E,$T,$A,$B,$C,$D,1); - &BODY_00_15(\$code,15,$D,$E,$T,$A,$B,$C,1); - - &BODY_16_19(\$code,16,$C,$D,$E,$T,$A,$B); - &BODY_16_19(\$code,17,$B,$C,$D,$E,$T,$A); - &BODY_16_19(\$code,18,$A,$B,$C,$D,$E,$T); - &BODY_16_19(\$code,19,$T,$A,$B,$C,$D,$E); +{ my $i,@V=($A,$B,$C,$D,$E,$T); - &BODY_20_39(\$code,20,$E,$T,$A,$B,$C,$D); - &BODY_20_39(\$code,21,$D,$E,$T,$A,$B,$C); - &BODY_20_39(\$code,22,$C,$D,$E,$T,$A,$B); - &BODY_20_39(\$code,23,$B,$C,$D,$E,$T,$A); - &BODY_20_39(\$code,24,$A,$B,$C,$D,$E,$T); - &BODY_20_39(\$code,25,$T,$A,$B,$C,$D,$E); - &BODY_20_39(\$code,26,$E,$T,$A,$B,$C,$D); - &BODY_20_39(\$code,27,$D,$E,$T,$A,$B,$C); - &BODY_20_39(\$code,28,$C,$D,$E,$T,$A,$B); - &BODY_20_39(\$code,29,$B,$C,$D,$E,$T,$A); - &BODY_20_39(\$code,30,$A,$B,$C,$D,$E,$T); - &BODY_20_39(\$code,31,$T,$A,$B,$C,$D,$E); - &BODY_20_39(\$code,32,$E,$T,$A,$B,$C,$D); - &BODY_20_39(\$code,33,$D,$E,$T,$A,$B,$C); - &BODY_20_39(\$code,34,$C,$D,$E,$T,$A,$B); - &BODY_20_39(\$code,35,$B,$C,$D,$E,$T,$A); - &BODY_20_39(\$code,36,$A,$B,$C,$D,$E,$T); - &BODY_20_39(\$code,37,$T,$A,$B,$C,$D,$E); - &BODY_20_39(\$code,38,$E,$T,$A,$B,$C,$D); - &BODY_20_39(\$code,39,$D,$E,$T,$A,$B,$C); + for($i=0;$i<16;$i++) { &BODY_00_15(\$code,$i,@V); unshift(@V,pop(@V)); } + for(;$i<20;$i++) { &BODY_16_19(\$code,$i,@V); unshift(@V,pop(@V)); } + for(;$i<40;$i++) { &BODY_20_39(\$code,$i,@V); unshift(@V,pop(@V)); } + for(;$i<60;$i++) { &BODY_40_59(\$code,$i,@V); unshift(@V,pop(@V)); } + for(;$i<80;$i++) { &BODY_60_79(\$code,$i,@V); unshift(@V,pop(@V)); } - &BODY_40_59(\$code,40,$C,$D,$E,$T,$A,$B); - &BODY_40_59(\$code,41,$B,$C,$D,$E,$T,$A); - &BODY_40_59(\$code,42,$A,$B,$C,$D,$E,$T); - &BODY_40_59(\$code,43,$T,$A,$B,$C,$D,$E); - &BODY_40_59(\$code,44,$E,$T,$A,$B,$C,$D); - &BODY_40_59(\$code,45,$D,$E,$T,$A,$B,$C); - &BODY_40_59(\$code,46,$C,$D,$E,$T,$A,$B); - &BODY_40_59(\$code,47,$B,$C,$D,$E,$T,$A); - &BODY_40_59(\$code,48,$A,$B,$C,$D,$E,$T); - &BODY_40_59(\$code,49,$T,$A,$B,$C,$D,$E); - &BODY_40_59(\$code,50,$E,$T,$A,$B,$C,$D); - &BODY_40_59(\$code,51,$D,$E,$T,$A,$B,$C); - &BODY_40_59(\$code,52,$C,$D,$E,$T,$A,$B); - &BODY_40_59(\$code,53,$B,$C,$D,$E,$T,$A); - &BODY_40_59(\$code,54,$A,$B,$C,$D,$E,$T); - &BODY_40_59(\$code,55,$T,$A,$B,$C,$D,$E); - &BODY_40_59(\$code,56,$E,$T,$A,$B,$C,$D); - &BODY_40_59(\$code,57,$D,$E,$T,$A,$B,$C); - &BODY_40_59(\$code,58,$C,$D,$E,$T,$A,$B); - &BODY_40_59(\$code,59,$B,$C,$D,$E,$T,$A); - - &BODY_60_79(\$code,60,$A,$B,$C,$D,$E,$T); - &BODY_60_79(\$code,61,$T,$A,$B,$C,$D,$E); - &BODY_60_79(\$code,62,$E,$T,$A,$B,$C,$D); - &BODY_60_79(\$code,63,$D,$E,$T,$A,$B,$C); - &BODY_60_79(\$code,64,$C,$D,$E,$T,$A,$B); - &BODY_60_79(\$code,65,$B,$C,$D,$E,$T,$A); - &BODY_60_79(\$code,66,$A,$B,$C,$D,$E,$T); - &BODY_60_79(\$code,67,$T,$A,$B,$C,$D,$E); - &BODY_60_79(\$code,68,$E,$T,$A,$B,$C,$D); - &BODY_60_79(\$code,69,$D,$E,$T,$A,$B,$C); - &BODY_60_79(\$code,70,$C,$D,$E,$T,$A,$B); - &BODY_60_79(\$code,71,$B,$C,$D,$E,$T,$A); - &BODY_60_79(\$code,72,$A,$B,$C,$D,$E,$T); - &BODY_60_79(\$code,73,$T,$A,$B,$C,$D,$E); - &BODY_60_79(\$code,74,$E,$T,$A,$B,$C,$D); - &BODY_60_79(\$code,75,$D,$E,$T,$A,$B,$C); - &BODY_60_79(\$code,76,$C,$D,$E,$T,$A,$B); - &BODY_60_79(\$code,77,$B,$C,$D,$E,$T,$A); - &BODY_60_79(\$code,78,$A,$B,$C,$D,$E,$T); - &BODY_60_79(\$code,79,$T,$A,$B,$C,$D,$E); + (($V[5] eq $D) and ($V[0] eq $E)) or die; # double-check +} $code.=<<___; { .mmb; add $h0=$h0,$E @@ -543,7 +298,8 @@ $code.=<<___; { .mib; st4 [ctx]=$h4,-16 mov pr=r2,0x1ffff br.ret.sptk.many b0 };; -.endp sha1_block_asm_data_order# +.endp sha1_block_data_order# +stringz "SHA1 block transform for IA64, CRYPTOGAMS by " ___ print $code; diff --git a/src/lib/libcrypto/sha/asm/sha1-x86_64.pl b/src/lib/libcrypto/sha/asm/sha1-x86_64.pl new file mode 100755 index 0000000000..f7ed67a726 --- /dev/null +++ b/src/lib/libcrypto/sha/asm/sha1-x86_64.pl @@ -0,0 +1,242 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# sha1_block procedure for x86_64. +# +# It was brought to my attention that on EM64T compiler-generated code +# was far behind 32-bit assembler implementation. This is unlike on +# Opteron where compiler-generated code was only 15% behind 32-bit +# assembler, which originally made it hard to motivate the effort. +# There was suggestion to mechanically translate 32-bit code, but I +# dismissed it, reasoning that x86_64 offers enough register bank +# capacity to fully utilize SHA-1 parallelism. Therefore this fresh +# implementation:-) However! While 64-bit code does performs better +# on Opteron, I failed to beat 32-bit assembler on EM64T core. Well, +# x86_64 does offer larger *addressable* bank, but out-of-order core +# reaches for even more registers through dynamic aliasing, and EM64T +# core must have managed to run-time optimize even 32-bit code just as +# good as 64-bit one. Performance improvement is summarized in the +# following table: +# +# gcc 3.4 32-bit asm cycles/byte +# Opteron +45% +20% 6.8 +# Xeon P4 +65% +0% 9.9 +# Core2 +60% +10% 7.0 + +$output=shift; + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open STDOUT,"| $^X $xlate $output"; + +$ctx="%rdi"; # 1st arg +$inp="%rsi"; # 2nd arg +$num="%rdx"; # 3rd arg + +# reassign arguments in order to produce more compact code +$ctx="%r8"; +$inp="%r9"; +$num="%r10"; + +$xi="%eax"; +$t0="%ebx"; +$t1="%ecx"; +$A="%edx"; +$B="%esi"; +$C="%edi"; +$D="%ebp"; +$E="%r11d"; +$T="%r12d"; + +@V=($A,$B,$C,$D,$E,$T); + +sub PROLOGUE { +my $func=shift; +$code.=<<___; +.globl $func +.type $func,\@function,3 +.align 16 +$func: + push %rbx + push %rbp + push %r12 + mov %rsp,%rax + mov %rdi,$ctx # reassigned argument + sub \$`8+16*4`,%rsp + mov %rsi,$inp # reassigned argument + and \$-64,%rsp + mov %rdx,$num # reassigned argument + mov %rax,`16*4`(%rsp) + + mov 0($ctx),$A + mov 4($ctx),$B + mov 8($ctx),$C + mov 12($ctx),$D + mov 16($ctx),$E +___ +} + +sub EPILOGUE { +my $func=shift; +$code.=<<___; + mov `16*4`(%rsp),%rsp + pop %r12 + pop %rbp + pop %rbx + ret +.size $func,.-$func +___ +} + +sub BODY_00_19 { +my ($i,$a,$b,$c,$d,$e,$f,$host)=@_; +my $j=$i+1; +$code.=<<___ if ($i==0); + mov `4*$i`($inp),$xi + `"bswap $xi" if(!defined($host))` + mov $xi,`4*$i`(%rsp) +___ +$code.=<<___ if ($i<15); + lea 0x5a827999($xi,$e),$f + mov $c,$t0 + mov `4*$j`($inp),$xi + mov $a,$e + xor $d,$t0 + `"bswap $xi" if(!defined($host))` + rol \$5,$e + and $b,$t0 + mov $xi,`4*$j`(%rsp) + add $e,$f + xor $d,$t0 + rol \$30,$b + add $t0,$f +___ +$code.=<<___ if ($i>=15); + lea 0x5a827999($xi,$e),$f + mov `4*($j%16)`(%rsp),$xi + mov $c,$t0 + mov $a,$e + xor `4*(($j+2)%16)`(%rsp),$xi + xor $d,$t0 + rol \$5,$e + xor `4*(($j+8)%16)`(%rsp),$xi + and $b,$t0 + add $e,$f + xor `4*(($j+13)%16)`(%rsp),$xi + xor $d,$t0 + rol \$30,$b + add $t0,$f + rol \$1,$xi + mov $xi,`4*($j%16)`(%rsp) +___ +} + +sub BODY_20_39 { +my ($i,$a,$b,$c,$d,$e,$f)=@_; +my $j=$i+1; +my $K=($i<40)?0x6ed9eba1:0xca62c1d6; +$code.=<<___ if ($i<79); + lea $K($xi,$e),$f + mov `4*($j%16)`(%rsp),$xi + mov $c,$t0 + mov $a,$e + xor `4*(($j+2)%16)`(%rsp),$xi + xor $b,$t0 + rol \$5,$e + xor `4*(($j+8)%16)`(%rsp),$xi + xor $d,$t0 + add $e,$f + xor `4*(($j+13)%16)`(%rsp),$xi + rol \$30,$b + add $t0,$f + rol \$1,$xi +___ +$code.=<<___ if ($i<76); + mov $xi,`4*($j%16)`(%rsp) +___ +$code.=<<___ if ($i==79); + lea $K($xi,$e),$f + mov $c,$t0 + mov $a,$e + xor $b,$t0 + rol \$5,$e + xor $d,$t0 + add $e,$f + rol \$30,$b + add $t0,$f +___ +} + +sub BODY_40_59 { +my ($i,$a,$b,$c,$d,$e,$f)=@_; +my $j=$i+1; +$code.=<<___; + lea 0x8f1bbcdc($xi,$e),$f + mov `4*($j%16)`(%rsp),$xi + mov $b,$t0 + mov $b,$t1 + xor `4*(($j+2)%16)`(%rsp),$xi + mov $a,$e + and $c,$t0 + xor `4*(($j+8)%16)`(%rsp),$xi + or $c,$t1 + rol \$5,$e + xor `4*(($j+13)%16)`(%rsp),$xi + and $d,$t1 + add $e,$f + rol \$1,$xi + or $t1,$t0 + rol \$30,$b + mov $xi,`4*($j%16)`(%rsp) + add $t0,$f +___ +} + +$code=".text\n"; + +&PROLOGUE("sha1_block_data_order"); +$code.=".align 4\n.Lloop:\n"; +for($i=0;$i<20;$i++) { &BODY_00_19($i,@V); unshift(@V,pop(@V)); } +for(;$i<40;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); } +for(;$i<60;$i++) { &BODY_40_59($i,@V); unshift(@V,pop(@V)); } +for(;$i<80;$i++) { &BODY_20_39($i,@V); unshift(@V,pop(@V)); } +$code.=<<___; + add 0($ctx),$E + add 4($ctx),$T + add 8($ctx),$A + add 12($ctx),$B + add 16($ctx),$C + mov $E,0($ctx) + mov $T,4($ctx) + mov $A,8($ctx) + mov $B,12($ctx) + mov $C,16($ctx) + + xchg $E,$A # mov $E,$A + xchg $T,$B # mov $T,$B + xchg $E,$C # mov $A,$C + xchg $T,$D # mov $B,$D + # mov $C,$E + lea `16*4`($inp),$inp + sub \$1,$num + jnz .Lloop +___ +&EPILOGUE("sha1_block_data_order"); +$code.=<<___; +.asciz "SHA1 block transform for x86_64, CRYPTOGAMS by " +___ + +#################################################################### + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +print $code; +close STDOUT; diff --git a/src/lib/libcrypto/sha/asm/sha512-ia64.pl b/src/lib/libcrypto/sha/asm/sha512-ia64.pl new file mode 100755 index 0000000000..1c6ce56522 --- /dev/null +++ b/src/lib/libcrypto/sha/asm/sha512-ia64.pl @@ -0,0 +1,672 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. The module is, however, dual licensed under OpenSSL and +# CRYPTOGAMS licenses depending on where you obtain it. For further +# details see http://www.openssl.org/~appro/cryptogams/. +# ==================================================================== +# +# SHA256/512_Transform for Itanium. +# +# sha512_block runs in 1003 cycles on Itanium 2, which is almost 50% +# faster than gcc and >60%(!) faster than code generated by HP-UX +# compiler (yes, HP-UX is generating slower code, because unlike gcc, +# it failed to deploy "shift right pair," 'shrp' instruction, which +# substitutes for 64-bit rotate). +# +# 924 cycles long sha256_block outperforms gcc by over factor of 2(!) +# and HP-UX compiler - by >40% (yes, gcc won sha512_block, but lost +# this one big time). Note that "formally" 924 is about 100 cycles +# too much. I mean it's 64 32-bit rounds vs. 80 virtually identical +# 64-bit ones and 1003*64/80 gives 802. Extra cycles, 2 per round, +# are spent on extra work to provide for 32-bit rotations. 32-bit +# rotations are still handled by 'shrp' instruction and for this +# reason lower 32 bits are deposited to upper half of 64-bit register +# prior 'shrp' issue. And in order to minimize the amount of such +# operations, X[16] values are *maintained* with copies of lower +# halves in upper halves, which is why you'll spot such instructions +# as custom 'mux2', "parallel 32-bit add," 'padd4' and "parallel +# 32-bit unsigned right shift," 'pshr4.u' instructions here. +# +# Rules of engagement. +# +# There is only one integer shifter meaning that if I have two rotate, +# deposit or extract instructions in adjacent bundles, they shall +# split [at run-time if they have to]. But note that variable and +# parallel shifts are performed by multi-media ALU and *are* pairable +# with rotates [and alike]. On the backside MMALU is rather slow: it +# takes 2 extra cycles before the result of integer operation is +# available *to* MMALU and 2(*) extra cycles before the result of MM +# operation is available "back" *to* integer ALU, not to mention that +# MMALU itself has 2 cycles latency. However! I explicitly scheduled +# these MM instructions to avoid MM stalls, so that all these extra +# latencies get "hidden" in instruction-level parallelism. +# +# (*) 2 cycles on Itanium 1 and 1 cycle on Itanium 2. But I schedule +# for 2 in order to provide for best *overall* performance, +# because on Itanium 1 stall on MM result is accompanied by +# pipeline flush, which takes 6 cycles:-( +# +# Resulting performance numbers for 900MHz Itanium 2 system: +# +# The 'numbers' are in 1000s of bytes per second processed. +# type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes +# sha1(*) 6210.14k 20376.30k 52447.83k 85870.05k 105478.12k +# sha256 7476.45k 20572.05k 41538.34k 56062.29k 62093.18k +# sha512 4996.56k 20026.28k 47597.20k 85278.79k 111501.31k +# +# (*) SHA1 numbers are for HP-UX compiler and are presented purely +# for reference purposes. I bet it can improved too... +# +# To generate code, pass the file name with either 256 or 512 in its +# name and compiler flags. + +$output=shift; + +if ($output =~ /512.*\.[s|asm]/) { + $SZ=8; + $BITS=8*$SZ; + $LDW="ld8"; + $STW="st8"; + $ADD="add"; + $SHRU="shr.u"; + $TABLE="K512"; + $func="sha512_block_data_order"; + @Sigma0=(28,34,39); + @Sigma1=(14,18,41); + @sigma0=(1, 8, 7); + @sigma1=(19,61, 6); + $rounds=80; +} elsif ($output =~ /256.*\.[s|asm]/) { + $SZ=4; + $BITS=8*$SZ; + $LDW="ld4"; + $STW="st4"; + $ADD="padd4"; + $SHRU="pshr4.u"; + $TABLE="K256"; + $func="sha256_block_data_order"; + @Sigma0=( 2,13,22); + @Sigma1=( 6,11,25); + @sigma0=( 7,18, 3); + @sigma1=(17,19,10); + $rounds=64; +} else { die "nonsense $output"; } + +open STDOUT,">$output" || die "can't open $output: $!"; + +if ($^O eq "hpux") { + $ADDP="addp4"; + for (@ARGV) { $ADDP="add" if (/[\+DD|\-mlp]64/); } +} else { $ADDP="add"; } +for (@ARGV) { $big_endian=1 if (/\-DB_ENDIAN/); + $big_endian=0 if (/\-DL_ENDIAN/); } +if (!defined($big_endian)) + { $big_endian=(unpack('L',pack('N',1))==1); } + +$code=<<___; +.ident \"$output, version 1.1\" +.ident \"IA-64 ISA artwork by Andy Polyakov \" +.explicit +.text + +pfssave=r2; +lcsave=r3; +prsave=r14; +K=r15; +A=r16; B=r17; C=r18; D=r19; +E=r20; F=r21; G=r22; H=r23; +T1=r24; T2=r25; +s0=r26; s1=r27; t0=r28; t1=r29; +Ktbl=r30; +ctx=r31; // 1st arg +input=r48; // 2nd arg +num=r49; // 3rd arg +sgm0=r50; sgm1=r51; // small constants +A_=r54; B_=r55; C_=r56; D_=r57; +E_=r58; F_=r59; G_=r60; H_=r61; + +// void $func (SHA_CTX *ctx, const void *in,size_t num[,int host]) +.global $func# +.proc $func# +.align 32 +$func: + .prologue + .save ar.pfs,pfssave +{ .mmi; alloc pfssave=ar.pfs,3,27,0,16 + $ADDP ctx=0,r32 // 1st arg + .save ar.lc,lcsave + mov lcsave=ar.lc } +{ .mmi; $ADDP input=0,r33 // 2nd arg + mov num=r34 // 3rd arg + .save pr,prsave + mov prsave=pr };; + + .body +{ .mib; add r8=0*$SZ,ctx + add r9=1*$SZ,ctx + brp.loop.imp .L_first16,.L_first16_end-16 } +{ .mib; add r10=2*$SZ,ctx + add r11=3*$SZ,ctx + brp.loop.imp .L_rest,.L_rest_end-16 };; + +// load A-H +.Lpic_point: +{ .mmi; $LDW A_=[r8],4*$SZ + $LDW B_=[r9],4*$SZ + mov Ktbl=ip } +{ .mmi; $LDW C_=[r10],4*$SZ + $LDW D_=[r11],4*$SZ + mov sgm0=$sigma0[2] };; +{ .mmi; $LDW E_=[r8] + $LDW F_=[r9] + add Ktbl=($TABLE#-.Lpic_point),Ktbl } +{ .mmi; $LDW G_=[r10] + $LDW H_=[r11] + cmp.ne p0,p16=0,r0 };; // used in sha256_block +___ +$code.=<<___ if ($BITS==64); +{ .mii; and r8=7,input + and input=~7,input;; + cmp.eq p9,p0=1,r8 } +{ .mmi; cmp.eq p10,p0=2,r8 + cmp.eq p11,p0=3,r8 + cmp.eq p12,p0=4,r8 } +{ .mmi; cmp.eq p13,p0=5,r8 + cmp.eq p14,p0=6,r8 + cmp.eq p15,p0=7,r8 };; +___ +$code.=<<___; +.L_outer: +.rotr X[16] +{ .mmi; mov A=A_ + mov B=B_ + mov ar.lc=14 } +{ .mmi; mov C=C_ + mov D=D_ + mov E=E_ } +{ .mmi; mov F=F_ + mov G=G_ + mov ar.ec=2 } +{ .mmi; ld1 X[15]=[input],$SZ // eliminated in 64-bit + mov H=H_ + mov sgm1=$sigma1[2] };; + +___ +$t0="t0", $t1="t1", $code.=<<___ if ($BITS==32); +.align 32 +.L_first16: +{ .mmi; add r9=1-$SZ,input + add r10=2-$SZ,input + add r11=3-$SZ,input };; +{ .mmi; ld1 r9=[r9] + ld1 r10=[r10] + dep.z $t1=E,32,32 } +{ .mmi; $LDW K=[Ktbl],$SZ + ld1 r11=[r11] + zxt4 E=E };; +{ .mii; or $t1=$t1,E + dep X[15]=X[15],r9,8,8 + dep r11=r10,r11,8,8 };; +{ .mmi; and T1=F,E + and T2=A,B + dep X[15]=X[15],r11,16,16 } +{ .mmi; andcm r8=G,E + and r9=A,C + mux2 $t0=A,0x44 };; // copy lower half to upper +{ .mmi; (p16) ld1 X[15-1]=[input],$SZ // prefetch + xor T1=T1,r8 // T1=((e & f) ^ (~e & g)) + _rotr r11=$t1,$Sigma1[0] } // ROTR(e,14) +{ .mib; and r10=B,C + xor T2=T2,r9 };; +___ +$t0="A", $t1="E", $code.=<<___ if ($BITS==64); +// in 64-bit mode I load whole X[16] at once and take care of alignment... +{ .mmi; add r8=1*$SZ,input + add r9=2*$SZ,input + add r10=3*$SZ,input };; +{ .mmb; $LDW X[15]=[input],4*$SZ + $LDW X[14]=[r8],4*$SZ +(p9) br.cond.dpnt.many .L1byte };; +{ .mmb; $LDW X[13]=[r9],4*$SZ + $LDW X[12]=[r10],4*$SZ +(p10) br.cond.dpnt.many .L2byte };; +{ .mmb; $LDW X[11]=[input],4*$SZ + $LDW X[10]=[r8],4*$SZ +(p11) br.cond.dpnt.many .L3byte };; +{ .mmb; $LDW X[ 9]=[r9],4*$SZ + $LDW X[ 8]=[r10],4*$SZ +(p12) br.cond.dpnt.many .L4byte };; +{ .mmb; $LDW X[ 7]=[input],4*$SZ + $LDW X[ 6]=[r8],4*$SZ +(p13) br.cond.dpnt.many .L5byte };; +{ .mmb; $LDW X[ 5]=[r9],4*$SZ + $LDW X[ 4]=[r10],4*$SZ +(p14) br.cond.dpnt.many .L6byte };; +{ .mmb; $LDW X[ 3]=[input],4*$SZ + $LDW X[ 2]=[r8],4*$SZ +(p15) br.cond.dpnt.many .L7byte };; +{ .mmb; $LDW X[ 1]=[r9],4*$SZ + $LDW X[ 0]=[r10],4*$SZ + br.many .L_first16 };; +.L1byte: +{ .mmi; $LDW X[13]=[r9],4*$SZ + $LDW X[12]=[r10],4*$SZ + shrp X[15]=X[15],X[14],56 };; +{ .mmi; $LDW X[11]=[input],4*$SZ + $LDW X[10]=[r8],4*$SZ + shrp X[14]=X[14],X[13],56 } +{ .mmi; $LDW X[ 9]=[r9],4*$SZ + $LDW X[ 8]=[r10],4*$SZ + shrp X[13]=X[13],X[12],56 };; +{ .mmi; $LDW X[ 7]=[input],4*$SZ + $LDW X[ 6]=[r8],4*$SZ + shrp X[12]=X[12],X[11],56 } +{ .mmi; $LDW X[ 5]=[r9],4*$SZ + $LDW X[ 4]=[r10],4*$SZ + shrp X[11]=X[11],X[10],56 };; +{ .mmi; $LDW X[ 3]=[input],4*$SZ + $LDW X[ 2]=[r8],4*$SZ + shrp X[10]=X[10],X[ 9],56 } +{ .mmi; $LDW X[ 1]=[r9],4*$SZ + $LDW X[ 0]=[r10],4*$SZ + shrp X[ 9]=X[ 9],X[ 8],56 };; +{ .mii; $LDW T1=[input] + shrp X[ 8]=X[ 8],X[ 7],56 + shrp X[ 7]=X[ 7],X[ 6],56 } +{ .mii; shrp X[ 6]=X[ 6],X[ 5],56 + shrp X[ 5]=X[ 5],X[ 4],56 };; +{ .mii; shrp X[ 4]=X[ 4],X[ 3],56 + shrp X[ 3]=X[ 3],X[ 2],56 } +{ .mii; shrp X[ 2]=X[ 2],X[ 1],56 + shrp X[ 1]=X[ 1],X[ 0],56 } +{ .mib; shrp X[ 0]=X[ 0],T1,56 + br.many .L_first16 };; +.L2byte: +{ .mmi; $LDW X[11]=[input],4*$SZ + $LDW X[10]=[r8],4*$SZ + shrp X[15]=X[15],X[14],48 } +{ .mmi; $LDW X[ 9]=[r9],4*$SZ + $LDW X[ 8]=[r10],4*$SZ + shrp X[14]=X[14],X[13],48 };; +{ .mmi; $LDW X[ 7]=[input],4*$SZ + $LDW X[ 6]=[r8],4*$SZ + shrp X[13]=X[13],X[12],48 } +{ .mmi; $LDW X[ 5]=[r9],4*$SZ + $LDW X[ 4]=[r10],4*$SZ + shrp X[12]=X[12],X[11],48 };; +{ .mmi; $LDW X[ 3]=[input],4*$SZ + $LDW X[ 2]=[r8],4*$SZ + shrp X[11]=X[11],X[10],48 } +{ .mmi; $LDW X[ 1]=[r9],4*$SZ + $LDW X[ 0]=[r10],4*$SZ + shrp X[10]=X[10],X[ 9],48 };; +{ .mii; $LDW T1=[input] + shrp X[ 9]=X[ 9],X[ 8],48 + shrp X[ 8]=X[ 8],X[ 7],48 } +{ .mii; shrp X[ 7]=X[ 7],X[ 6],48 + shrp X[ 6]=X[ 6],X[ 5],48 };; +{ .mii; shrp X[ 5]=X[ 5],X[ 4],48 + shrp X[ 4]=X[ 4],X[ 3],48 } +{ .mii; shrp X[ 3]=X[ 3],X[ 2],48 + shrp X[ 2]=X[ 2],X[ 1],48 } +{ .mii; shrp X[ 1]=X[ 1],X[ 0],48 + shrp X[ 0]=X[ 0],T1,48 } +{ .mfb; br.many .L_first16 };; +.L3byte: +{ .mmi; $LDW X[ 9]=[r9],4*$SZ + $LDW X[ 8]=[r10],4*$SZ + shrp X[15]=X[15],X[14],40 };; +{ .mmi; $LDW X[ 7]=[input],4*$SZ + $LDW X[ 6]=[r8],4*$SZ + shrp X[14]=X[14],X[13],40 } +{ .mmi; $LDW X[ 5]=[r9],4*$SZ + $LDW X[ 4]=[r10],4*$SZ + shrp X[13]=X[13],X[12],40 };; +{ .mmi; $LDW X[ 3]=[input],4*$SZ + $LDW X[ 2]=[r8],4*$SZ + shrp X[12]=X[12],X[11],40 } +{ .mmi; $LDW X[ 1]=[r9],4*$SZ + $LDW X[ 0]=[r10],4*$SZ + shrp X[11]=X[11],X[10],40 };; +{ .mii; $LDW T1=[input] + shrp X[10]=X[10],X[ 9],40 + shrp X[ 9]=X[ 9],X[ 8],40 } +{ .mii; shrp X[ 8]=X[ 8],X[ 7],40 + shrp X[ 7]=X[ 7],X[ 6],40 };; +{ .mii; shrp X[ 6]=X[ 6],X[ 5],40 + shrp X[ 5]=X[ 5],X[ 4],40 } +{ .mii; shrp X[ 4]=X[ 4],X[ 3],40 + shrp X[ 3]=X[ 3],X[ 2],40 } +{ .mii; shrp X[ 2]=X[ 2],X[ 1],40 + shrp X[ 1]=X[ 1],X[ 0],40 } +{ .mib; shrp X[ 0]=X[ 0],T1,40 + br.many .L_first16 };; +.L4byte: +{ .mmi; $LDW X[ 7]=[input],4*$SZ + $LDW X[ 6]=[r8],4*$SZ + shrp X[15]=X[15],X[14],32 } +{ .mmi; $LDW X[ 5]=[r9],4*$SZ + $LDW X[ 4]=[r10],4*$SZ + shrp X[14]=X[14],X[13],32 };; +{ .mmi; $LDW X[ 3]=[input],4*$SZ + $LDW X[ 2]=[r8],4*$SZ + shrp X[13]=X[13],X[12],32 } +{ .mmi; $LDW X[ 1]=[r9],4*$SZ + $LDW X[ 0]=[r10],4*$SZ + shrp X[12]=X[12],X[11],32 };; +{ .mii; $LDW T1=[input] + shrp X[11]=X[11],X[10],32 + shrp X[10]=X[10],X[ 9],32 } +{ .mii; shrp X[ 9]=X[ 9],X[ 8],32 + shrp X[ 8]=X[ 8],X[ 7],32 };; +{ .mii; shrp X[ 7]=X[ 7],X[ 6],32 + shrp X[ 6]=X[ 6],X[ 5],32 } +{ .mii; shrp X[ 5]=X[ 5],X[ 4],32 + shrp X[ 4]=X[ 4],X[ 3],32 } +{ .mii; shrp X[ 3]=X[ 3],X[ 2],32 + shrp X[ 2]=X[ 2],X[ 1],32 } +{ .mii; shrp X[ 1]=X[ 1],X[ 0],32 + shrp X[ 0]=X[ 0],T1,32 } +{ .mfb; br.many .L_first16 };; +.L5byte: +{ .mmi; $LDW X[ 5]=[r9],4*$SZ + $LDW X[ 4]=[r10],4*$SZ + shrp X[15]=X[15],X[14],24 };; +{ .mmi; $LDW X[ 3]=[input],4*$SZ + $LDW X[ 2]=[r8],4*$SZ + shrp X[14]=X[14],X[13],24 } +{ .mmi; $LDW X[ 1]=[r9],4*$SZ + $LDW X[ 0]=[r10],4*$SZ + shrp X[13]=X[13],X[12],24 };; +{ .mii; $LDW T1=[input] + shrp X[12]=X[12],X[11],24 + shrp X[11]=X[11],X[10],24 } +{ .mii; shrp X[10]=X[10],X[ 9],24 + shrp X[ 9]=X[ 9],X[ 8],24 };; +{ .mii; shrp X[ 8]=X[ 8],X[ 7],24 + shrp X[ 7]=X[ 7],X[ 6],24 } +{ .mii; shrp X[ 6]=X[ 6],X[ 5],24 + shrp X[ 5]=X[ 5],X[ 4],24 } +{ .mii; shrp X[ 4]=X[ 4],X[ 3],24 + shrp X[ 3]=X[ 3],X[ 2],24 } +{ .mii; shrp X[ 2]=X[ 2],X[ 1],24 + shrp X[ 1]=X[ 1],X[ 0],24 } +{ .mib; shrp X[ 0]=X[ 0],T1,24 + br.many .L_first16 };; +.L6byte: +{ .mmi; $LDW X[ 3]=[input],4*$SZ + $LDW X[ 2]=[r8],4*$SZ + shrp X[15]=X[15],X[14],16 } +{ .mmi; $LDW X[ 1]=[r9],4*$SZ + $LDW X[ 0]=[r10],4*$SZ + shrp X[14]=X[14],X[13],16 };; +{ .mii; $LDW T1=[input] + shrp X[13]=X[13],X[12],16 + shrp X[12]=X[12],X[11],16 } +{ .mii; shrp X[11]=X[11],X[10],16 + shrp X[10]=X[10],X[ 9],16 };; +{ .mii; shrp X[ 9]=X[ 9],X[ 8],16 + shrp X[ 8]=X[ 8],X[ 7],16 } +{ .mii; shrp X[ 7]=X[ 7],X[ 6],16 + shrp X[ 6]=X[ 6],X[ 5],16 } +{ .mii; shrp X[ 5]=X[ 5],X[ 4],16 + shrp X[ 4]=X[ 4],X[ 3],16 } +{ .mii; shrp X[ 3]=X[ 3],X[ 2],16 + shrp X[ 2]=X[ 2],X[ 1],16 } +{ .mii; shrp X[ 1]=X[ 1],X[ 0],16 + shrp X[ 0]=X[ 0],T1,16 } +{ .mfb; br.many .L_first16 };; +.L7byte: +{ .mmi; $LDW X[ 1]=[r9],4*$SZ + $LDW X[ 0]=[r10],4*$SZ + shrp X[15]=X[15],X[14],8 };; +{ .mii; $LDW T1=[input] + shrp X[14]=X[14],X[13],8 + shrp X[13]=X[13],X[12],8 } +{ .mii; shrp X[12]=X[12],X[11],8 + shrp X[11]=X[11],X[10],8 };; +{ .mii; shrp X[10]=X[10],X[ 9],8 + shrp X[ 9]=X[ 9],X[ 8],8 } +{ .mii; shrp X[ 8]=X[ 8],X[ 7],8 + shrp X[ 7]=X[ 7],X[ 6],8 } +{ .mii; shrp X[ 6]=X[ 6],X[ 5],8 + shrp X[ 5]=X[ 5],X[ 4],8 } +{ .mii; shrp X[ 4]=X[ 4],X[ 3],8 + shrp X[ 3]=X[ 3],X[ 2],8 } +{ .mii; shrp X[ 2]=X[ 2],X[ 1],8 + shrp X[ 1]=X[ 1],X[ 0],8 } +{ .mib; shrp X[ 0]=X[ 0],T1,8 + br.many .L_first16 };; + +.align 32 +.L_first16: +{ .mmi; $LDW K=[Ktbl],$SZ + and T1=F,E + and T2=A,B } +{ .mmi; //$LDW X[15]=[input],$SZ // X[i]=*input++ + andcm r8=G,E + and r9=A,C };; +{ .mmi; xor T1=T1,r8 //T1=((e & f) ^ (~e & g)) + and r10=B,C + _rotr r11=$t1,$Sigma1[0] } // ROTR(e,14) +{ .mmi; xor T2=T2,r9 + mux1 X[15]=X[15],\@rev };; // eliminated in big-endian +___ +$code.=<<___; +{ .mib; add T1=T1,H // T1=Ch(e,f,g)+h + _rotr r8=$t1,$Sigma1[1] } // ROTR(e,18) +{ .mib; xor T2=T2,r10 // T2=((a & b) ^ (a & c) ^ (b & c)) + mov H=G };; +{ .mib; xor r11=r8,r11 + _rotr r9=$t1,$Sigma1[2] } // ROTR(e,41) +{ .mib; mov G=F + mov F=E };; +{ .mib; xor r9=r9,r11 // r9=Sigma1(e) + _rotr r10=$t0,$Sigma0[0] } // ROTR(a,28) +{ .mib; add T1=T1,K // T1=Ch(e,f,g)+h+K512[i] + mov E=D };; +{ .mib; add T1=T1,r9 // T1+=Sigma1(e) + _rotr r11=$t0,$Sigma0[1] } // ROTR(a,34) +{ .mib; mov D=C + mov C=B };; +{ .mib; add T1=T1,X[15] // T1+=X[i] + _rotr r8=$t0,$Sigma0[2] } // ROTR(a,39) +{ .mib; xor r10=r10,r11 + mux2 X[15]=X[15],0x44 };; // eliminated in 64-bit +{ .mmi; xor r10=r8,r10 // r10=Sigma0(a) + mov B=A + add A=T1,T2 };; +{ .mib; add E=E,T1 + add A=A,r10 // T2=Maj(a,b,c)+Sigma0(a) + br.ctop.sptk .L_first16 };; +.L_first16_end: + +{ .mii; mov ar.lc=$rounds-17 + mov ar.ec=1 };; + +.align 32 +.L_rest: +.rotr X[16] +{ .mib; $LDW K=[Ktbl],$SZ + _rotr r8=X[15-1],$sigma0[0] } // ROTR(s0,1) +{ .mib; $ADD X[15]=X[15],X[15-9] // X[i&0xF]+=X[(i+9)&0xF] + $SHRU s0=X[15-1],sgm0 };; // s0=X[(i+1)&0xF]>>7 +{ .mib; and T1=F,E + _rotr r9=X[15-1],$sigma0[1] } // ROTR(s0,8) +{ .mib; andcm r10=G,E + $SHRU s1=X[15-14],sgm1 };; // s1=X[(i+14)&0xF]>>6 +{ .mmi; xor T1=T1,r10 // T1=((e & f) ^ (~e & g)) + xor r9=r8,r9 + _rotr r10=X[15-14],$sigma1[0] };;// ROTR(s1,19) +{ .mib; and T2=A,B + _rotr r11=X[15-14],$sigma1[1] }// ROTR(s1,61) +{ .mib; and r8=A,C };; +___ +$t0="t0", $t1="t1", $code.=<<___ if ($BITS==32); +// I adhere to mmi; in order to hold Itanium 1 back and avoid 6 cycle +// pipeline flush in last bundle. Note that even on Itanium2 the +// latter stalls for one clock cycle... +{ .mmi; xor s0=s0,r9 // s0=sigma0(X[(i+1)&0xF]) + dep.z $t1=E,32,32 } +{ .mmi; xor r10=r11,r10 + zxt4 E=E };; +{ .mmi; or $t1=$t1,E + xor s1=s1,r10 // s1=sigma1(X[(i+14)&0xF]) + mux2 $t0=A,0x44 };; // copy lower half to upper +{ .mmi; xor T2=T2,r8 + _rotr r9=$t1,$Sigma1[0] } // ROTR(e,14) +{ .mmi; and r10=B,C + add T1=T1,H // T1=Ch(e,f,g)+h + $ADD X[15]=X[15],s0 };; // X[i&0xF]+=sigma0(X[(i+1)&0xF]) +___ +$t0="A", $t1="E", $code.=<<___ if ($BITS==64); +{ .mib; xor s0=s0,r9 // s0=sigma0(X[(i+1)&0xF]) + _rotr r9=$t1,$Sigma1[0] } // ROTR(e,14) +{ .mib; xor r10=r11,r10 + xor T2=T2,r8 };; +{ .mib; xor s1=s1,r10 // s1=sigma1(X[(i+14)&0xF]) + add T1=T1,H } +{ .mib; and r10=B,C + $ADD X[15]=X[15],s0 };; // X[i&0xF]+=sigma0(X[(i+1)&0xF]) +___ +$code.=<<___; +{ .mmi; xor T2=T2,r10 // T2=((a & b) ^ (a & c) ^ (b & c)) + mov H=G + _rotr r8=$t1,$Sigma1[1] };; // ROTR(e,18) +{ .mmi; xor r11=r8,r9 + $ADD X[15]=X[15],s1 // X[i&0xF]+=sigma1(X[(i+14)&0xF]) + _rotr r9=$t1,$Sigma1[2] } // ROTR(e,41) +{ .mmi; mov G=F + mov F=E };; +{ .mib; xor r9=r9,r11 // r9=Sigma1(e) + _rotr r10=$t0,$Sigma0[0] } // ROTR(a,28) +{ .mib; add T1=T1,K // T1=Ch(e,f,g)+h+K512[i] + mov E=D };; +{ .mib; add T1=T1,r9 // T1+=Sigma1(e) + _rotr r11=$t0,$Sigma0[1] } // ROTR(a,34) +{ .mib; mov D=C + mov C=B };; +{ .mmi; add T1=T1,X[15] // T1+=X[i] + xor r10=r10,r11 + _rotr r8=$t0,$Sigma0[2] };; // ROTR(a,39) +{ .mmi; xor r10=r8,r10 // r10=Sigma0(a) + mov B=A + add A=T1,T2 };; +{ .mib; add E=E,T1 + add A=A,r10 // T2=Maj(a,b,c)+Sigma0(a) + br.ctop.sptk .L_rest };; +.L_rest_end: + +{ .mmi; add A_=A_,A + add B_=B_,B + add C_=C_,C } +{ .mmi; add D_=D_,D + add E_=E_,E + cmp.ltu p16,p0=1,num };; +{ .mmi; add F_=F_,F + add G_=G_,G + add H_=H_,H } +{ .mmb; add Ktbl=-$SZ*$rounds,Ktbl +(p16) add num=-1,num +(p16) br.dptk.many .L_outer };; + +{ .mib; add r8=0*$SZ,ctx + add r9=1*$SZ,ctx } +{ .mib; add r10=2*$SZ,ctx + add r11=3*$SZ,ctx };; +{ .mmi; $STW [r8]=A_,4*$SZ + $STW [r9]=B_,4*$SZ + mov ar.lc=lcsave } +{ .mmi; $STW [r10]=C_,4*$SZ + $STW [r11]=D_,4*$SZ + mov pr=prsave,0x1ffff };; +{ .mmb; $STW [r8]=E_ + $STW [r9]=F_ } +{ .mmb; $STW [r10]=G_ + $STW [r11]=H_ + br.ret.sptk.many b0 };; +.endp $func# +___ + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +$code =~ s/_rotr(\s+)([^=]+)=([^,]+),([0-9]+)/shrp$1$2=$3,$3,$4/gm; +if ($BITS==64) { + $code =~ s/mux2(\s+)\S+/nop.i$1 0x0/gm; + $code =~ s/mux1(\s+)\S+/nop.i$1 0x0/gm if ($big_endian); + $code =~ s/(shrp\s+X\[[^=]+)=([^,]+),([^,]+),([1-9]+)/$1=$3,$2,64-$4/gm + if (!$big_endian); + $code =~ s/ld1(\s+)X\[\S+/nop.m$1 0x0/gm; +} + +print $code; + +print<<___ if ($BITS==32); +.align 64 +.type K256#,\@object +K256: data4 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5 + data4 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5 + data4 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3 + data4 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174 + data4 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc + data4 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da + data4 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7 + data4 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967 + data4 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13 + data4 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85 + data4 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3 + data4 0xd192e819,0xd6990624,0xf40e3585,0x106aa070 + data4 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5 + data4 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3 + data4 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208 + data4 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2 +.size K256#,$SZ*$rounds +stringz "SHA256 block transform for IA64, CRYPTOGAMS by " +___ +print<<___ if ($BITS==64); +.align 64 +.type K512#,\@object +K512: data8 0x428a2f98d728ae22,0x7137449123ef65cd + data8 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc + data8 0x3956c25bf348b538,0x59f111f1b605d019 + data8 0x923f82a4af194f9b,0xab1c5ed5da6d8118 + data8 0xd807aa98a3030242,0x12835b0145706fbe + data8 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2 + data8 0x72be5d74f27b896f,0x80deb1fe3b1696b1 + data8 0x9bdc06a725c71235,0xc19bf174cf692694 + data8 0xe49b69c19ef14ad2,0xefbe4786384f25e3 + data8 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65 + data8 0x2de92c6f592b0275,0x4a7484aa6ea6e483 + data8 0x5cb0a9dcbd41fbd4,0x76f988da831153b5 + data8 0x983e5152ee66dfab,0xa831c66d2db43210 + data8 0xb00327c898fb213f,0xbf597fc7beef0ee4 + data8 0xc6e00bf33da88fc2,0xd5a79147930aa725 + data8 0x06ca6351e003826f,0x142929670a0e6e70 + data8 0x27b70a8546d22ffc,0x2e1b21385c26c926 + data8 0x4d2c6dfc5ac42aed,0x53380d139d95b3df + data8 0x650a73548baf63de,0x766a0abb3c77b2a8 + data8 0x81c2c92e47edaee6,0x92722c851482353b + data8 0xa2bfe8a14cf10364,0xa81a664bbc423001 + data8 0xc24b8b70d0f89791,0xc76c51a30654be30 + data8 0xd192e819d6ef5218,0xd69906245565a910 + data8 0xf40e35855771202a,0x106aa07032bbd1b8 + data8 0x19a4c116b8d2d0c8,0x1e376c085141ab53 + data8 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8 + data8 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb + data8 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3 + data8 0x748f82ee5defb2fc,0x78a5636f43172f60 + data8 0x84c87814a1f0ab72,0x8cc702081a6439ec + data8 0x90befffa23631e28,0xa4506cebde82bde9 + data8 0xbef9a3f7b2c67915,0xc67178f2e372532b + data8 0xca273eceea26619c,0xd186b8c721c0c207 + data8 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178 + data8 0x06f067aa72176fba,0x0a637dc5a2c898a6 + data8 0x113f9804bef90dae,0x1b710b35131c471b + data8 0x28db77f523047d84,0x32caab7b40c72493 + data8 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c + data8 0x4cc5d4becb3e42b6,0x597f299cfc657e2a + data8 0x5fcb6fab3ad6faec,0x6c44198c4a475817 +.size K512#,$SZ*$rounds +stringz "SHA512 block transform for IA64, CRYPTOGAMS by " +___ diff --git a/src/lib/libcrypto/sha/asm/sha512-x86_64.pl b/src/lib/libcrypto/sha/asm/sha512-x86_64.pl new file mode 100755 index 0000000000..b6252d31ec --- /dev/null +++ b/src/lib/libcrypto/sha/asm/sha512-x86_64.pl @@ -0,0 +1,344 @@ +#!/usr/bin/env perl +# +# ==================================================================== +# Written by Andy Polyakov for the OpenSSL +# project. Rights for redistribution and usage in source and binary +# forms are granted according to the OpenSSL license. +# ==================================================================== +# +# sha256/512_block procedure for x86_64. +# +# 40% improvement over compiler-generated code on Opteron. On EM64T +# sha256 was observed to run >80% faster and sha512 - >40%. No magical +# tricks, just straight implementation... I really wonder why gcc +# [being armed with inline assembler] fails to generate as fast code. +# The only thing which is cool about this module is that it's very +# same instruction sequence used for both SHA-256 and SHA-512. In +# former case the instructions operate on 32-bit operands, while in +# latter - on 64-bit ones. All I had to do is to get one flavor right, +# the other one passed the test right away:-) +# +# sha256_block runs in ~1005 cycles on Opteron, which gives you +# asymptotic performance of 64*1000/1005=63.7MBps times CPU clock +# frequency in GHz. sha512_block runs in ~1275 cycles, which results +# in 128*1000/1275=100MBps per GHz. Is there room for improvement? +# Well, if you compare it to IA-64 implementation, which maintains +# X[16] in register bank[!], tends to 4 instructions per CPU clock +# cycle and runs in 1003 cycles, 1275 is very good result for 3-way +# issue Opteron pipeline and X[16] maintained in memory. So that *if* +# there is a way to improve it, *then* the only way would be to try to +# offload X[16] updates to SSE unit, but that would require "deeper" +# loop unroll, which in turn would naturally cause size blow-up, not +# to mention increased complexity! And once again, only *if* it's +# actually possible to noticeably improve overall ILP, instruction +# level parallelism, on a given CPU implementation in this case. +# +# Special note on Intel EM64T. While Opteron CPU exhibits perfect +# perfromance ratio of 1.5 between 64- and 32-bit flavors [see above], +# [currently available] EM64T CPUs apparently are far from it. On the +# contrary, 64-bit version, sha512_block, is ~30% *slower* than 32-bit +# sha256_block:-( This is presumably because 64-bit shifts/rotates +# apparently are not atomic instructions, but implemented in microcode. + +$output=shift; + +$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open STDOUT,"| $^X $xlate $output"; + +if ($output =~ /512/) { + $func="sha512_block_data_order"; + $TABLE="K512"; + $SZ=8; + @ROT=($A,$B,$C,$D,$E,$F,$G,$H)=("%rax","%rbx","%rcx","%rdx", + "%r8", "%r9", "%r10","%r11"); + ($T1,$a0,$a1,$a2)=("%r12","%r13","%r14","%r15"); + @Sigma0=(28,34,39); + @Sigma1=(14,18,41); + @sigma0=(1, 8, 7); + @sigma1=(19,61, 6); + $rounds=80; +} else { + $func="sha256_block_data_order"; + $TABLE="K256"; + $SZ=4; + @ROT=($A,$B,$C,$D,$E,$F,$G,$H)=("%eax","%ebx","%ecx","%edx", + "%r8d","%r9d","%r10d","%r11d"); + ($T1,$a0,$a1,$a2)=("%r12d","%r13d","%r14d","%r15d"); + @Sigma0=( 2,13,22); + @Sigma1=( 6,11,25); + @sigma0=( 7,18, 3); + @sigma1=(17,19,10); + $rounds=64; +} + +$ctx="%rdi"; # 1st arg +$round="%rdi"; # zaps $ctx +$inp="%rsi"; # 2nd arg +$Tbl="%rbp"; + +$_ctx="16*$SZ+0*8(%rsp)"; +$_inp="16*$SZ+1*8(%rsp)"; +$_end="16*$SZ+2*8(%rsp)"; +$_rsp="16*$SZ+3*8(%rsp)"; +$framesz="16*$SZ+4*8"; + + +sub ROUND_00_15() +{ my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_; + +$code.=<<___; + mov $e,$a0 + mov $e,$a1 + mov $f,$a2 + + ror \$$Sigma1[0],$a0 + ror \$$Sigma1[1],$a1 + xor $g,$a2 # f^g + + xor $a1,$a0 + ror \$`$Sigma1[2]-$Sigma1[1]`,$a1 + and $e,$a2 # (f^g)&e + mov $T1,`$SZ*($i&0xf)`(%rsp) + + xor $a1,$a0 # Sigma1(e) + xor $g,$a2 # Ch(e,f,g)=((f^g)&e)^g + add $h,$T1 # T1+=h + + mov $a,$h + add $a0,$T1 # T1+=Sigma1(e) + + add $a2,$T1 # T1+=Ch(e,f,g) + mov $a,$a0 + mov $a,$a1 + + ror \$$Sigma0[0],$h + ror \$$Sigma0[1],$a0 + mov $a,$a2 + add ($Tbl,$round,$SZ),$T1 # T1+=K[round] + + xor $a0,$h + ror \$`$Sigma0[2]-$Sigma0[1]`,$a0 + or $c,$a1 # a|c + + xor $a0,$h # h=Sigma0(a) + and $c,$a2 # a&c + add $T1,$d # d+=T1 + + and $b,$a1 # (a|c)&b + add $T1,$h # h+=T1 + + or $a2,$a1 # Maj(a,b,c)=((a|c)&b)|(a&c) + lea 1($round),$round # round++ + + add $a1,$h # h+=Maj(a,b,c) +___ +} + +sub ROUND_16_XX() +{ my ($i,$a,$b,$c,$d,$e,$f,$g,$h) = @_; + +$code.=<<___; + mov `$SZ*(($i+1)&0xf)`(%rsp),$a0 + mov `$SZ*(($i+14)&0xf)`(%rsp),$T1 + + mov $a0,$a2 + + shr \$$sigma0[2],$a0 + ror \$$sigma0[0],$a2 + + xor $a2,$a0 + ror \$`$sigma0[1]-$sigma0[0]`,$a2 + + xor $a2,$a0 # sigma0(X[(i+1)&0xf]) + mov $T1,$a1 + + shr \$$sigma1[2],$T1 + ror \$$sigma1[0],$a1 + + xor $a1,$T1 + ror \$`$sigma1[1]-$sigma1[0]`,$a1 + + xor $a1,$T1 # sigma1(X[(i+14)&0xf]) + + add $a0,$T1 + + add `$SZ*(($i+9)&0xf)`(%rsp),$T1 + + add `$SZ*($i&0xf)`(%rsp),$T1 +___ + &ROUND_00_15(@_); +} + +$code=<<___; +.text + +.globl $func +.type $func,\@function,4 +.align 16 +$func: + push %rbx + push %rbp + push %r12 + push %r13 + push %r14 + push %r15 + mov %rsp,%rbp # copy %rsp + shl \$4,%rdx # num*16 + sub \$$framesz,%rsp + lea ($inp,%rdx,$SZ),%rdx # inp+num*16*$SZ + and \$-64,%rsp # align stack frame + mov $ctx,$_ctx # save ctx, 1st arg + mov $inp,$_inp # save inp, 2nd arh + mov %rdx,$_end # save end pointer, "3rd" arg + mov %rbp,$_rsp # save copy of %rsp + + .picmeup $Tbl + lea $TABLE-.($Tbl),$Tbl + + mov $SZ*0($ctx),$A + mov $SZ*1($ctx),$B + mov $SZ*2($ctx),$C + mov $SZ*3($ctx),$D + mov $SZ*4($ctx),$E + mov $SZ*5($ctx),$F + mov $SZ*6($ctx),$G + mov $SZ*7($ctx),$H + jmp .Lloop + +.align 16 +.Lloop: + xor $round,$round +___ + for($i=0;$i<16;$i++) { + $code.=" mov $SZ*$i($inp),$T1\n"; + $code.=" bswap $T1\n"; + &ROUND_00_15($i,@ROT); + unshift(@ROT,pop(@ROT)); + } +$code.=<<___; + jmp .Lrounds_16_xx +.align 16 +.Lrounds_16_xx: +___ + for(;$i<32;$i++) { + &ROUND_16_XX($i,@ROT); + unshift(@ROT,pop(@ROT)); + } + +$code.=<<___; + cmp \$$rounds,$round + jb .Lrounds_16_xx + + mov $_ctx,$ctx + lea 16*$SZ($inp),$inp + + add $SZ*0($ctx),$A + add $SZ*1($ctx),$B + add $SZ*2($ctx),$C + add $SZ*3($ctx),$D + add $SZ*4($ctx),$E + add $SZ*5($ctx),$F + add $SZ*6($ctx),$G + add $SZ*7($ctx),$H + + cmp $_end,$inp + + mov $A,$SZ*0($ctx) + mov $B,$SZ*1($ctx) + mov $C,$SZ*2($ctx) + mov $D,$SZ*3($ctx) + mov $E,$SZ*4($ctx) + mov $F,$SZ*5($ctx) + mov $G,$SZ*6($ctx) + mov $H,$SZ*7($ctx) + jb .Lloop + + mov $_rsp,%rsp + pop %r15 + pop %r14 + pop %r13 + pop %r12 + pop %rbp + pop %rbx + + ret +.size $func,.-$func +___ + +if ($SZ==4) { +$code.=<<___; +.align 64 +.type $TABLE,\@object +$TABLE: + .long 0x428a2f98,0x71374491,0xb5c0fbcf,0xe9b5dba5 + .long 0x3956c25b,0x59f111f1,0x923f82a4,0xab1c5ed5 + .long 0xd807aa98,0x12835b01,0x243185be,0x550c7dc3 + .long 0x72be5d74,0x80deb1fe,0x9bdc06a7,0xc19bf174 + .long 0xe49b69c1,0xefbe4786,0x0fc19dc6,0x240ca1cc + .long 0x2de92c6f,0x4a7484aa,0x5cb0a9dc,0x76f988da + .long 0x983e5152,0xa831c66d,0xb00327c8,0xbf597fc7 + .long 0xc6e00bf3,0xd5a79147,0x06ca6351,0x14292967 + .long 0x27b70a85,0x2e1b2138,0x4d2c6dfc,0x53380d13 + .long 0x650a7354,0x766a0abb,0x81c2c92e,0x92722c85 + .long 0xa2bfe8a1,0xa81a664b,0xc24b8b70,0xc76c51a3 + .long 0xd192e819,0xd6990624,0xf40e3585,0x106aa070 + .long 0x19a4c116,0x1e376c08,0x2748774c,0x34b0bcb5 + .long 0x391c0cb3,0x4ed8aa4a,0x5b9cca4f,0x682e6ff3 + .long 0x748f82ee,0x78a5636f,0x84c87814,0x8cc70208 + .long 0x90befffa,0xa4506ceb,0xbef9a3f7,0xc67178f2 +___ +} else { +$code.=<<___; +.align 64 +.type $TABLE,\@object +$TABLE: + .quad 0x428a2f98d728ae22,0x7137449123ef65cd + .quad 0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc + .quad 0x3956c25bf348b538,0x59f111f1b605d019 + .quad 0x923f82a4af194f9b,0xab1c5ed5da6d8118 + .quad 0xd807aa98a3030242,0x12835b0145706fbe + .quad 0x243185be4ee4b28c,0x550c7dc3d5ffb4e2 + .quad 0x72be5d74f27b896f,0x80deb1fe3b1696b1 + .quad 0x9bdc06a725c71235,0xc19bf174cf692694 + .quad 0xe49b69c19ef14ad2,0xefbe4786384f25e3 + .quad 0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65 + .quad 0x2de92c6f592b0275,0x4a7484aa6ea6e483 + .quad 0x5cb0a9dcbd41fbd4,0x76f988da831153b5 + .quad 0x983e5152ee66dfab,0xa831c66d2db43210 + .quad 0xb00327c898fb213f,0xbf597fc7beef0ee4 + .quad 0xc6e00bf33da88fc2,0xd5a79147930aa725 + .quad 0x06ca6351e003826f,0x142929670a0e6e70 + .quad 0x27b70a8546d22ffc,0x2e1b21385c26c926 + .quad 0x4d2c6dfc5ac42aed,0x53380d139d95b3df + .quad 0x650a73548baf63de,0x766a0abb3c77b2a8 + .quad 0x81c2c92e47edaee6,0x92722c851482353b + .quad 0xa2bfe8a14cf10364,0xa81a664bbc423001 + .quad 0xc24b8b70d0f89791,0xc76c51a30654be30 + .quad 0xd192e819d6ef5218,0xd69906245565a910 + .quad 0xf40e35855771202a,0x106aa07032bbd1b8 + .quad 0x19a4c116b8d2d0c8,0x1e376c085141ab53 + .quad 0x2748774cdf8eeb99,0x34b0bcb5e19b48a8 + .quad 0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb + .quad 0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3 + .quad 0x748f82ee5defb2fc,0x78a5636f43172f60 + .quad 0x84c87814a1f0ab72,0x8cc702081a6439ec + .quad 0x90befffa23631e28,0xa4506cebde82bde9 + .quad 0xbef9a3f7b2c67915,0xc67178f2e372532b + .quad 0xca273eceea26619c,0xd186b8c721c0c207 + .quad 0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178 + .quad 0x06f067aa72176fba,0x0a637dc5a2c898a6 + .quad 0x113f9804bef90dae,0x1b710b35131c471b + .quad 0x28db77f523047d84,0x32caab7b40c72493 + .quad 0x3c9ebe0a15c9bebc,0x431d67c49c100d4c + .quad 0x4cc5d4becb3e42b6,0x597f299cfc657e2a + .quad 0x5fcb6fab3ad6faec,0x6c44198c4a475817 +___ +} + +$code =~ s/\`([^\`]*)\`/eval $1/gem; +print $code; +close STDOUT; diff --git a/src/lib/libcrypto/sha/sha.h b/src/lib/libcrypto/sha/sha.h index 79c07b0fd1..eed44d7f94 100644 --- a/src/lib/libcrypto/sha/sha.h +++ b/src/lib/libcrypto/sha/sha.h @@ -60,6 +60,7 @@ #define HEADER_SHA_H #include +#include #ifdef __cplusplus extern "C" { @@ -70,7 +71,7 @@ extern "C" { #endif #if defined(OPENSSL_FIPS) -#define FIPS_SHA_SIZE_T unsigned long +#define FIPS_SHA_SIZE_T size_t #endif /* @@ -101,26 +102,97 @@ typedef struct SHAstate_st SHA_LONG h0,h1,h2,h3,h4; SHA_LONG Nl,Nh; SHA_LONG data[SHA_LBLOCK]; - int num; + unsigned int num; } SHA_CTX; #ifndef OPENSSL_NO_SHA0 -#ifdef OPENSSL_FIPS -int private_SHA_Init(SHA_CTX *c); -#endif int SHA_Init(SHA_CTX *c); -int SHA_Update(SHA_CTX *c, const void *data, unsigned long len); +int SHA_Update(SHA_CTX *c, const void *data, size_t len); int SHA_Final(unsigned char *md, SHA_CTX *c); -unsigned char *SHA(const unsigned char *d, unsigned long n,unsigned char *md); +unsigned char *SHA(const unsigned char *d, size_t n, unsigned char *md); void SHA_Transform(SHA_CTX *c, const unsigned char *data); #endif #ifndef OPENSSL_NO_SHA1 int SHA1_Init(SHA_CTX *c); -int SHA1_Update(SHA_CTX *c, const void *data, unsigned long len); +int SHA1_Update(SHA_CTX *c, const void *data, size_t len); int SHA1_Final(unsigned char *md, SHA_CTX *c); -unsigned char *SHA1(const unsigned char *d, unsigned long n,unsigned char *md); +unsigned char *SHA1(const unsigned char *d, size_t n, unsigned char *md); void SHA1_Transform(SHA_CTX *c, const unsigned char *data); #endif + +#define SHA256_CBLOCK (SHA_LBLOCK*4) /* SHA-256 treats input data as a + * contiguous array of 32 bit + * wide big-endian values. */ +#define SHA224_DIGEST_LENGTH 28 +#define SHA256_DIGEST_LENGTH 32 + +typedef struct SHA256state_st + { + SHA_LONG h[8]; + SHA_LONG Nl,Nh; + SHA_LONG data[SHA_LBLOCK]; + unsigned int num,md_len; + } SHA256_CTX; + +#ifndef OPENSSL_NO_SHA256 +int SHA224_Init(SHA256_CTX *c); +int SHA224_Update(SHA256_CTX *c, const void *data, size_t len); +int SHA224_Final(unsigned char *md, SHA256_CTX *c); +unsigned char *SHA224(const unsigned char *d, size_t n,unsigned char *md); +int SHA256_Init(SHA256_CTX *c); +int SHA256_Update(SHA256_CTX *c, const void *data, size_t len); +int SHA256_Final(unsigned char *md, SHA256_CTX *c); +unsigned char *SHA256(const unsigned char *d, size_t n,unsigned char *md); +void SHA256_Transform(SHA256_CTX *c, const unsigned char *data); +#endif + +#define SHA384_DIGEST_LENGTH 48 +#define SHA512_DIGEST_LENGTH 64 + +#ifndef OPENSSL_NO_SHA512 +/* + * Unlike 32-bit digest algorithms, SHA-512 *relies* on SHA_LONG64 + * being exactly 64-bit wide. See Implementation Notes in sha512.c + * for further details. + */ +#define SHA512_CBLOCK (SHA_LBLOCK*8) /* SHA-512 treats input data as a + * contiguous array of 64 bit + * wide big-endian values. */ +#if (defined(_WIN32) || defined(_WIN64)) && !defined(__MINGW32__) +#define SHA_LONG64 unsigned __int64 +#define U64(C) C##UI64 +#elif defined(__arch64__) +#define SHA_LONG64 unsigned long +#define U64(C) C##UL +#else +#define SHA_LONG64 unsigned long long +#define U64(C) C##ULL +#endif + +typedef struct SHA512state_st + { + SHA_LONG64 h[8]; + SHA_LONG64 Nl,Nh; + union { + SHA_LONG64 d[SHA_LBLOCK]; + unsigned char p[SHA512_CBLOCK]; + } u; + unsigned int num,md_len; + } SHA512_CTX; +#endif + +#ifndef OPENSSL_NO_SHA512 +int SHA384_Init(SHA512_CTX *c); +int SHA384_Update(SHA512_CTX *c, const void *data, size_t len); +int SHA384_Final(unsigned char *md, SHA512_CTX *c); +unsigned char *SHA384(const unsigned char *d, size_t n,unsigned char *md); +int SHA512_Init(SHA512_CTX *c); +int SHA512_Update(SHA512_CTX *c, const void *data, size_t len); +int SHA512_Final(unsigned char *md, SHA512_CTX *c); +unsigned char *SHA512(const unsigned char *d, size_t n,unsigned char *md); +void SHA512_Transform(SHA512_CTX *c, const unsigned char *data); +#endif + #ifdef __cplusplus } #endif diff --git a/src/lib/libcrypto/sha/sha1_one.c b/src/lib/libcrypto/sha/sha1_one.c index f4694b701b..7c65b60276 100644 --- a/src/lib/libcrypto/sha/sha1_one.c +++ b/src/lib/libcrypto/sha/sha1_one.c @@ -61,8 +61,8 @@ #include #include -#if !defined(OPENSSL_NO_SHA1) && !defined(OPENSSL_FIPS) -unsigned char *SHA1(const unsigned char *d, unsigned long n, unsigned char *md) +#ifndef OPENSSL_NO_SHA1 +unsigned char *SHA1(const unsigned char *d, size_t n, unsigned char *md) { SHA_CTX c; static unsigned char m[SHA_DIGEST_LENGTH]; diff --git a/src/lib/libcrypto/sha/sha1dgst.c b/src/lib/libcrypto/sha/sha1dgst.c index 1e2009b760..50d1925cde 100644 --- a/src/lib/libcrypto/sha/sha1dgst.c +++ b/src/lib/libcrypto/sha/sha1dgst.c @@ -56,26 +56,19 @@ * [including the GNU Public Licence.] */ +#include #if !defined(OPENSSL_NO_SHA1) && !defined(OPENSSL_NO_SHA) #undef SHA_0 #define SHA_1 #include -#include -#ifndef OPENSSL_FIPS -const char *SHA1_version="SHA1" OPENSSL_VERSION_PTEXT; +const char SHA1_version[]="SHA1" OPENSSL_VERSION_PTEXT; /* The implementation is in ../md32_common.h */ #include "sha_locl.h" -#else /* ndef OPENSSL_FIPS */ - -static void *dummy=&dummy; - -#endif /* ndef OPENSSL_FIPS */ - #endif diff --git a/src/lib/libcrypto/sha/sha256.c b/src/lib/libcrypto/sha/sha256.c new file mode 100644 index 0000000000..867f90cc97 --- /dev/null +++ b/src/lib/libcrypto/sha/sha256.c @@ -0,0 +1,282 @@ +/* crypto/sha/sha256.c */ +/* ==================================================================== + * Copyright (c) 2004 The OpenSSL Project. All rights reserved + * according to the OpenSSL license [found in ../../LICENSE]. + * ==================================================================== + */ +#include +#if !defined(OPENSSL_NO_SHA) && !defined(OPENSSL_NO_SHA256) + +#include +#include + +#include +#include +#include + +const char SHA256_version[]="SHA-256" OPENSSL_VERSION_PTEXT; + +int SHA224_Init (SHA256_CTX *c) + { + c->h[0]=0xc1059ed8UL; c->h[1]=0x367cd507UL; + c->h[2]=0x3070dd17UL; c->h[3]=0xf70e5939UL; + c->h[4]=0xffc00b31UL; c->h[5]=0x68581511UL; + c->h[6]=0x64f98fa7UL; c->h[7]=0xbefa4fa4UL; + c->Nl=0; c->Nh=0; + c->num=0; c->md_len=SHA224_DIGEST_LENGTH; + return 1; + } + +int SHA256_Init (SHA256_CTX *c) + { + c->h[0]=0x6a09e667UL; c->h[1]=0xbb67ae85UL; + c->h[2]=0x3c6ef372UL; c->h[3]=0xa54ff53aUL; + c->h[4]=0x510e527fUL; c->h[5]=0x9b05688cUL; + c->h[6]=0x1f83d9abUL; c->h[7]=0x5be0cd19UL; + c->Nl=0; c->Nh=0; + c->num=0; c->md_len=SHA256_DIGEST_LENGTH; + return 1; + } + +unsigned char *SHA224(const unsigned char *d, size_t n, unsigned char *md) + { + SHA256_CTX c; + static unsigned char m[SHA224_DIGEST_LENGTH]; + + if (md == NULL) md=m; + SHA224_Init(&c); + SHA256_Update(&c,d,n); + SHA256_Final(md,&c); + OPENSSL_cleanse(&c,sizeof(c)); + return(md); + } + +unsigned char *SHA256(const unsigned char *d, size_t n, unsigned char *md) + { + SHA256_CTX c; + static unsigned char m[SHA256_DIGEST_LENGTH]; + + if (md == NULL) md=m; + SHA256_Init(&c); + SHA256_Update(&c,d,n); + SHA256_Final(md,&c); + OPENSSL_cleanse(&c,sizeof(c)); + return(md); + } + +int SHA224_Update(SHA256_CTX *c, const void *data, size_t len) +{ return SHA256_Update (c,data,len); } +int SHA224_Final (unsigned char *md, SHA256_CTX *c) +{ return SHA256_Final (md,c); } + +#define DATA_ORDER_IS_BIG_ENDIAN + +#define HASH_LONG SHA_LONG +#define HASH_CTX SHA256_CTX +#define HASH_CBLOCK SHA_CBLOCK +/* + * Note that FIPS180-2 discusses "Truncation of the Hash Function Output." + * default: case below covers for it. It's not clear however if it's + * permitted to truncate to amount of bytes not divisible by 4. I bet not, + * but if it is, then default: case shall be extended. For reference. + * Idea behind separate cases for pre-defined lenghts is to let the + * compiler decide if it's appropriate to unroll small loops. + */ +#define HASH_MAKE_STRING(c,s) do { \ + unsigned long ll; \ + unsigned int xn; \ + switch ((c)->md_len) \ + { case SHA224_DIGEST_LENGTH: \ + for (xn=0;xnh[xn]; HOST_l2c(ll,(s)); } \ + break; \ + case SHA256_DIGEST_LENGTH: \ + for (xn=0;xnh[xn]; HOST_l2c(ll,(s)); } \ + break; \ + default: \ + if ((c)->md_len > SHA256_DIGEST_LENGTH) \ + return 0; \ + for (xn=0;xn<(c)->md_len/4;xn++) \ + { ll=(c)->h[xn]; HOST_l2c(ll,(s)); } \ + break; \ + } \ + } while (0) + +#define HASH_UPDATE SHA256_Update +#define HASH_TRANSFORM SHA256_Transform +#define HASH_FINAL SHA256_Final +#define HASH_BLOCK_DATA_ORDER sha256_block_data_order +#ifndef SHA256_ASM +static +#endif +void sha256_block_data_order (SHA256_CTX *ctx, const void *in, size_t num); + +#include "md32_common.h" + +#ifndef SHA256_ASM +static const SHA_LONG K256[64] = { + 0x428a2f98UL,0x71374491UL,0xb5c0fbcfUL,0xe9b5dba5UL, + 0x3956c25bUL,0x59f111f1UL,0x923f82a4UL,0xab1c5ed5UL, + 0xd807aa98UL,0x12835b01UL,0x243185beUL,0x550c7dc3UL, + 0x72be5d74UL,0x80deb1feUL,0x9bdc06a7UL,0xc19bf174UL, + 0xe49b69c1UL,0xefbe4786UL,0x0fc19dc6UL,0x240ca1ccUL, + 0x2de92c6fUL,0x4a7484aaUL,0x5cb0a9dcUL,0x76f988daUL, + 0x983e5152UL,0xa831c66dUL,0xb00327c8UL,0xbf597fc7UL, + 0xc6e00bf3UL,0xd5a79147UL,0x06ca6351UL,0x14292967UL, + 0x27b70a85UL,0x2e1b2138UL,0x4d2c6dfcUL,0x53380d13UL, + 0x650a7354UL,0x766a0abbUL,0x81c2c92eUL,0x92722c85UL, + 0xa2bfe8a1UL,0xa81a664bUL,0xc24b8b70UL,0xc76c51a3UL, + 0xd192e819UL,0xd6990624UL,0xf40e3585UL,0x106aa070UL, + 0x19a4c116UL,0x1e376c08UL,0x2748774cUL,0x34b0bcb5UL, + 0x391c0cb3UL,0x4ed8aa4aUL,0x5b9cca4fUL,0x682e6ff3UL, + 0x748f82eeUL,0x78a5636fUL,0x84c87814UL,0x8cc70208UL, + 0x90befffaUL,0xa4506cebUL,0xbef9a3f7UL,0xc67178f2UL }; + +/* + * FIPS specification refers to right rotations, while our ROTATE macro + * is left one. This is why you might notice that rotation coefficients + * differ from those observed in FIPS document by 32-N... + */ +#define Sigma0(x) (ROTATE((x),30) ^ ROTATE((x),19) ^ ROTATE((x),10)) +#define Sigma1(x) (ROTATE((x),26) ^ ROTATE((x),21) ^ ROTATE((x),7)) +#define sigma0(x) (ROTATE((x),25) ^ ROTATE((x),14) ^ ((x)>>3)) +#define sigma1(x) (ROTATE((x),15) ^ ROTATE((x),13) ^ ((x)>>10)) + +#define Ch(x,y,z) (((x) & (y)) ^ ((~(x)) & (z))) +#define Maj(x,y,z) (((x) & (y)) ^ ((x) & (z)) ^ ((y) & (z))) + +#ifdef OPENSSL_SMALL_FOOTPRINT + +static void sha256_block_data_order (SHA256_CTX *ctx, const void *in, size_t num) + { + unsigned MD32_REG_T a,b,c,d,e,f,g,h,s0,s1,T1,T2; + SHA_LONG X[16],l; + int i; + const unsigned char *data=in; + + while (num--) { + + a = ctx->h[0]; b = ctx->h[1]; c = ctx->h[2]; d = ctx->h[3]; + e = ctx->h[4]; f = ctx->h[5]; g = ctx->h[6]; h = ctx->h[7]; + + for (i=0;i<16;i++) + { + HOST_c2l(data,l); T1 = X[i] = l; + T1 += h + Sigma1(e) + Ch(e,f,g) + K256[i]; + T2 = Sigma0(a) + Maj(a,b,c); + h = g; g = f; f = e; e = d + T1; + d = c; c = b; b = a; a = T1 + T2; + } + + for (;i<64;i++) + { + s0 = X[(i+1)&0x0f]; s0 = sigma0(s0); + s1 = X[(i+14)&0x0f]; s1 = sigma1(s1); + + T1 = X[i&0xf] += s0 + s1 + X[(i+9)&0xf]; + T1 += h + Sigma1(e) + Ch(e,f,g) + K256[i]; + T2 = Sigma0(a) + Maj(a,b,c); + h = g; g = f; f = e; e = d + T1; + d = c; c = b; b = a; a = T1 + T2; + } + + ctx->h[0] += a; ctx->h[1] += b; ctx->h[2] += c; ctx->h[3] += d; + ctx->h[4] += e; ctx->h[5] += f; ctx->h[6] += g; ctx->h[7] += h; + + } +} + +#else + +#define ROUND_00_15(i,a,b,c,d,e,f,g,h) do { \ + T1 += h + Sigma1(e) + Ch(e,f,g) + K256[i]; \ + h = Sigma0(a) + Maj(a,b,c); \ + d += T1; h += T1; } while (0) + +#define ROUND_16_63(i,a,b,c,d,e,f,g,h,X) do { \ + s0 = X[(i+1)&0x0f]; s0 = sigma0(s0); \ + s1 = X[(i+14)&0x0f]; s1 = sigma1(s1); \ + T1 = X[(i)&0x0f] += s0 + s1 + X[(i+9)&0x0f]; \ + ROUND_00_15(i,a,b,c,d,e,f,g,h); } while (0) + +static void sha256_block_data_order (SHA256_CTX *ctx, const void *in, size_t num) + { + unsigned MD32_REG_T a,b,c,d,e,f,g,h,s0,s1,T1; + SHA_LONG X[16]; + int i; + const unsigned char *data=in; + const union { long one; char little; } is_endian = {1}; + + while (num--) { + + a = ctx->h[0]; b = ctx->h[1]; c = ctx->h[2]; d = ctx->h[3]; + e = ctx->h[4]; f = ctx->h[5]; g = ctx->h[6]; h = ctx->h[7]; + + if (!is_endian.little && sizeof(SHA_LONG)==4 && ((size_t)in%4)==0) + { + const SHA_LONG *W=(const SHA_LONG *)data; + + T1 = X[0] = W[0]; ROUND_00_15(0,a,b,c,d,e,f,g,h); + T1 = X[1] = W[1]; ROUND_00_15(1,h,a,b,c,d,e,f,g); + T1 = X[2] = W[2]; ROUND_00_15(2,g,h,a,b,c,d,e,f); + T1 = X[3] = W[3]; ROUND_00_15(3,f,g,h,a,b,c,d,e); + T1 = X[4] = W[4]; ROUND_00_15(4,e,f,g,h,a,b,c,d); + T1 = X[5] = W[5]; ROUND_00_15(5,d,e,f,g,h,a,b,c); + T1 = X[6] = W[6]; ROUND_00_15(6,c,d,e,f,g,h,a,b); + T1 = X[7] = W[7]; ROUND_00_15(7,b,c,d,e,f,g,h,a); + T1 = X[8] = W[8]; ROUND_00_15(8,a,b,c,d,e,f,g,h); + T1 = X[9] = W[9]; ROUND_00_15(9,h,a,b,c,d,e,f,g); + T1 = X[10] = W[10]; ROUND_00_15(10,g,h,a,b,c,d,e,f); + T1 = X[11] = W[11]; ROUND_00_15(11,f,g,h,a,b,c,d,e); + T1 = X[12] = W[12]; ROUND_00_15(12,e,f,g,h,a,b,c,d); + T1 = X[13] = W[13]; ROUND_00_15(13,d,e,f,g,h,a,b,c); + T1 = X[14] = W[14]; ROUND_00_15(14,c,d,e,f,g,h,a,b); + T1 = X[15] = W[15]; ROUND_00_15(15,b,c,d,e,f,g,h,a); + + data += SHA256_CBLOCK; + } + else + { + SHA_LONG l; + + HOST_c2l(data,l); T1 = X[0] = l; ROUND_00_15(0,a,b,c,d,e,f,g,h); + HOST_c2l(data,l); T1 = X[1] = l; ROUND_00_15(1,h,a,b,c,d,e,f,g); + HOST_c2l(data,l); T1 = X[2] = l; ROUND_00_15(2,g,h,a,b,c,d,e,f); + HOST_c2l(data,l); T1 = X[3] = l; ROUND_00_15(3,f,g,h,a,b,c,d,e); + HOST_c2l(data,l); T1 = X[4] = l; ROUND_00_15(4,e,f,g,h,a,b,c,d); + HOST_c2l(data,l); T1 = X[5] = l; ROUND_00_15(5,d,e,f,g,h,a,b,c); + HOST_c2l(data,l); T1 = X[6] = l; ROUND_00_15(6,c,d,e,f,g,h,a,b); + HOST_c2l(data,l); T1 = X[7] = l; ROUND_00_15(7,b,c,d,e,f,g,h,a); + HOST_c2l(data,l); T1 = X[8] = l; ROUND_00_15(8,a,b,c,d,e,f,g,h); + HOST_c2l(data,l); T1 = X[9] = l; ROUND_00_15(9,h,a,b,c,d,e,f,g); + HOST_c2l(data,l); T1 = X[10] = l; ROUND_00_15(10,g,h,a,b,c,d,e,f); + HOST_c2l(data,l); T1 = X[11] = l; ROUND_00_15(11,f,g,h,a,b,c,d,e); + HOST_c2l(data,l); T1 = X[12] = l; ROUND_00_15(12,e,f,g,h,a,b,c,d); + HOST_c2l(data,l); T1 = X[13] = l; ROUND_00_15(13,d,e,f,g,h,a,b,c); + HOST_c2l(data,l); T1 = X[14] = l; ROUND_00_15(14,c,d,e,f,g,h,a,b); + HOST_c2l(data,l); T1 = X[15] = l; ROUND_00_15(15,b,c,d,e,f,g,h,a); + } + + for (i=16;i<64;i+=8) + { + ROUND_16_63(i+0,a,b,c,d,e,f,g,h,X); + ROUND_16_63(i+1,h,a,b,c,d,e,f,g,X); + ROUND_16_63(i+2,g,h,a,b,c,d,e,f,X); + ROUND_16_63(i+3,f,g,h,a,b,c,d,e,X); + ROUND_16_63(i+4,e,f,g,h,a,b,c,d,X); + ROUND_16_63(i+5,d,e,f,g,h,a,b,c,X); + ROUND_16_63(i+6,c,d,e,f,g,h,a,b,X); + ROUND_16_63(i+7,b,c,d,e,f,g,h,a,X); + } + + ctx->h[0] += a; ctx->h[1] += b; ctx->h[2] += c; ctx->h[3] += d; + ctx->h[4] += e; ctx->h[5] += f; ctx->h[6] += g; ctx->h[7] += h; + + } + } + +#endif +#endif /* SHA256_ASM */ + +#endif /* OPENSSL_NO_SHA256 */ diff --git a/src/lib/libcrypto/sha/sha512.c b/src/lib/libcrypto/sha/sha512.c new file mode 100644 index 0000000000..987fc07c99 --- /dev/null +++ b/src/lib/libcrypto/sha/sha512.c @@ -0,0 +1,537 @@ +/* crypto/sha/sha512.c */ +/* ==================================================================== + * Copyright (c) 2004 The OpenSSL Project. All rights reserved + * according to the OpenSSL license [found in ../../LICENSE]. + * ==================================================================== + */ +#include +#if !defined(OPENSSL_NO_SHA) && !defined(OPENSSL_NO_SHA512) +/* + * IMPLEMENTATION NOTES. + * + * As you might have noticed 32-bit hash algorithms: + * + * - permit SHA_LONG to be wider than 32-bit (case on CRAY); + * - optimized versions implement two transform functions: one operating + * on [aligned] data in host byte order and one - on data in input + * stream byte order; + * - share common byte-order neutral collector and padding function + * implementations, ../md32_common.h; + * + * Neither of the above applies to this SHA-512 implementations. Reasons + * [in reverse order] are: + * + * - it's the only 64-bit hash algorithm for the moment of this writing, + * there is no need for common collector/padding implementation [yet]; + * - by supporting only one transform function [which operates on + * *aligned* data in input stream byte order, big-endian in this case] + * we minimize burden of maintenance in two ways: a) collector/padding + * function is simpler; b) only one transform function to stare at; + * - SHA_LONG64 is required to be exactly 64-bit in order to be able to + * apply a number of optimizations to mitigate potential performance + * penalties caused by previous design decision; + * + * Caveat lector. + * + * Implementation relies on the fact that "long long" is 64-bit on + * both 32- and 64-bit platforms. If some compiler vendor comes up + * with 128-bit long long, adjustment to sha.h would be required. + * As this implementation relies on 64-bit integer type, it's totally + * inappropriate for platforms which don't support it, most notably + * 16-bit platforms. + * + */ +#include +#include + +#include +#include +#include + +#include "cryptlib.h" + +const char SHA512_version[]="SHA-512" OPENSSL_VERSION_PTEXT; + +#if defined(__i386) || defined(__i386__) || defined(_M_IX86) || \ + defined(__x86_64) || defined(_M_AMD64) || defined(_M_X64) || \ + defined(__s390__) || defined(__s390x__) || \ + defined(SHA512_ASM) +#define SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA +#endif + +int SHA384_Init (SHA512_CTX *c) + { + c->h[0]=U64(0xcbbb9d5dc1059ed8); + c->h[1]=U64(0x629a292a367cd507); + c->h[2]=U64(0x9159015a3070dd17); + c->h[3]=U64(0x152fecd8f70e5939); + c->h[4]=U64(0x67332667ffc00b31); + c->h[5]=U64(0x8eb44a8768581511); + c->h[6]=U64(0xdb0c2e0d64f98fa7); + c->h[7]=U64(0x47b5481dbefa4fa4); + c->Nl=0; c->Nh=0; + c->num=0; c->md_len=SHA384_DIGEST_LENGTH; + return 1; + } + +int SHA512_Init (SHA512_CTX *c) + { + c->h[0]=U64(0x6a09e667f3bcc908); + c->h[1]=U64(0xbb67ae8584caa73b); + c->h[2]=U64(0x3c6ef372fe94f82b); + c->h[3]=U64(0xa54ff53a5f1d36f1); + c->h[4]=U64(0x510e527fade682d1); + c->h[5]=U64(0x9b05688c2b3e6c1f); + c->h[6]=U64(0x1f83d9abfb41bd6b); + c->h[7]=U64(0x5be0cd19137e2179); + c->Nl=0; c->Nh=0; + c->num=0; c->md_len=SHA512_DIGEST_LENGTH; + return 1; + } + +#ifndef SHA512_ASM +static +#endif +void sha512_block_data_order (SHA512_CTX *ctx, const void *in, size_t num); + +int SHA512_Final (unsigned char *md, SHA512_CTX *c) + { + unsigned char *p=(unsigned char *)c->u.p; + size_t n=c->num; + + p[n]=0x80; /* There always is a room for one */ + n++; + if (n > (sizeof(c->u)-16)) + memset (p+n,0,sizeof(c->u)-n), n=0, + sha512_block_data_order (c,p,1); + + memset (p+n,0,sizeof(c->u)-16-n); +#ifdef B_ENDIAN + c->u.d[SHA_LBLOCK-2] = c->Nh; + c->u.d[SHA_LBLOCK-1] = c->Nl; +#else + p[sizeof(c->u)-1] = (unsigned char)(c->Nl); + p[sizeof(c->u)-2] = (unsigned char)(c->Nl>>8); + p[sizeof(c->u)-3] = (unsigned char)(c->Nl>>16); + p[sizeof(c->u)-4] = (unsigned char)(c->Nl>>24); + p[sizeof(c->u)-5] = (unsigned char)(c->Nl>>32); + p[sizeof(c->u)-6] = (unsigned char)(c->Nl>>40); + p[sizeof(c->u)-7] = (unsigned char)(c->Nl>>48); + p[sizeof(c->u)-8] = (unsigned char)(c->Nl>>56); + p[sizeof(c->u)-9] = (unsigned char)(c->Nh); + p[sizeof(c->u)-10] = (unsigned char)(c->Nh>>8); + p[sizeof(c->u)-11] = (unsigned char)(c->Nh>>16); + p[sizeof(c->u)-12] = (unsigned char)(c->Nh>>24); + p[sizeof(c->u)-13] = (unsigned char)(c->Nh>>32); + p[sizeof(c->u)-14] = (unsigned char)(c->Nh>>40); + p[sizeof(c->u)-15] = (unsigned char)(c->Nh>>48); + p[sizeof(c->u)-16] = (unsigned char)(c->Nh>>56); +#endif + + sha512_block_data_order (c,p,1); + + if (md==0) return 0; + + switch (c->md_len) + { + /* Let compiler decide if it's appropriate to unroll... */ + case SHA384_DIGEST_LENGTH: + for (n=0;nh[n]; + + *(md++) = (unsigned char)(t>>56); + *(md++) = (unsigned char)(t>>48); + *(md++) = (unsigned char)(t>>40); + *(md++) = (unsigned char)(t>>32); + *(md++) = (unsigned char)(t>>24); + *(md++) = (unsigned char)(t>>16); + *(md++) = (unsigned char)(t>>8); + *(md++) = (unsigned char)(t); + } + break; + case SHA512_DIGEST_LENGTH: + for (n=0;nh[n]; + + *(md++) = (unsigned char)(t>>56); + *(md++) = (unsigned char)(t>>48); + *(md++) = (unsigned char)(t>>40); + *(md++) = (unsigned char)(t>>32); + *(md++) = (unsigned char)(t>>24); + *(md++) = (unsigned char)(t>>16); + *(md++) = (unsigned char)(t>>8); + *(md++) = (unsigned char)(t); + } + break; + /* ... as well as make sure md_len is not abused. */ + default: return 0; + } + + return 1; + } + +int SHA384_Final (unsigned char *md,SHA512_CTX *c) +{ return SHA512_Final (md,c); } + +int SHA512_Update (SHA512_CTX *c, const void *_data, size_t len) + { + SHA_LONG64 l; + unsigned char *p=c->u.p; + const unsigned char *data=(const unsigned char *)_data; + + if (len==0) return 1; + + l = (c->Nl+(((SHA_LONG64)len)<<3))&U64(0xffffffffffffffff); + if (l < c->Nl) c->Nh++; + if (sizeof(len)>=8) c->Nh+=(((SHA_LONG64)len)>>61); + c->Nl=l; + + if (c->num != 0) + { + size_t n = sizeof(c->u) - c->num; + + if (len < n) + { + memcpy (p+c->num,data,len), c->num += len; + return 1; + } + else { + memcpy (p+c->num,data,n), c->num = 0; + len-=n, data+=n; + sha512_block_data_order (c,p,1); + } + } + + if (len >= sizeof(c->u)) + { +#ifndef SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA + if ((size_t)data%sizeof(c->u.d[0]) != 0) + while (len >= sizeof(c->u)) + memcpy (p,data,sizeof(c->u)), + sha512_block_data_order (c,p,1), + len -= sizeof(c->u), + data += sizeof(c->u); + else +#endif + sha512_block_data_order (c,data,len/sizeof(c->u)), + data += len, + len %= sizeof(c->u), + data -= len; + } + + if (len != 0) memcpy (p,data,len), c->num = (int)len; + + return 1; + } + +int SHA384_Update (SHA512_CTX *c, const void *data, size_t len) +{ return SHA512_Update (c,data,len); } + +void SHA512_Transform (SHA512_CTX *c, const unsigned char *data) +{ sha512_block_data_order (c,data,1); } + +unsigned char *SHA384(const unsigned char *d, size_t n, unsigned char *md) + { + SHA512_CTX c; + static unsigned char m[SHA384_DIGEST_LENGTH]; + + if (md == NULL) md=m; + SHA384_Init(&c); + SHA512_Update(&c,d,n); + SHA512_Final(md,&c); + OPENSSL_cleanse(&c,sizeof(c)); + return(md); + } + +unsigned char *SHA512(const unsigned char *d, size_t n, unsigned char *md) + { + SHA512_CTX c; + static unsigned char m[SHA512_DIGEST_LENGTH]; + + if (md == NULL) md=m; + SHA512_Init(&c); + SHA512_Update(&c,d,n); + SHA512_Final(md,&c); + OPENSSL_cleanse(&c,sizeof(c)); + return(md); + } + +#ifndef SHA512_ASM +static const SHA_LONG64 K512[80] = { + U64(0x428a2f98d728ae22),U64(0x7137449123ef65cd), + U64(0xb5c0fbcfec4d3b2f),U64(0xe9b5dba58189dbbc), + U64(0x3956c25bf348b538),U64(0x59f111f1b605d019), + U64(0x923f82a4af194f9b),U64(0xab1c5ed5da6d8118), + U64(0xd807aa98a3030242),U64(0x12835b0145706fbe), + U64(0x243185be4ee4b28c),U64(0x550c7dc3d5ffb4e2), + U64(0x72be5d74f27b896f),U64(0x80deb1fe3b1696b1), + U64(0x9bdc06a725c71235),U64(0xc19bf174cf692694), + U64(0xe49b69c19ef14ad2),U64(0xefbe4786384f25e3), + U64(0x0fc19dc68b8cd5b5),U64(0x240ca1cc77ac9c65), + U64(0x2de92c6f592b0275),U64(0x4a7484aa6ea6e483), + U64(0x5cb0a9dcbd41fbd4),U64(0x76f988da831153b5), + U64(0x983e5152ee66dfab),U64(0xa831c66d2db43210), + U64(0xb00327c898fb213f),U64(0xbf597fc7beef0ee4), + U64(0xc6e00bf33da88fc2),U64(0xd5a79147930aa725), + U64(0x06ca6351e003826f),U64(0x142929670a0e6e70), + U64(0x27b70a8546d22ffc),U64(0x2e1b21385c26c926), + U64(0x4d2c6dfc5ac42aed),U64(0x53380d139d95b3df), + U64(0x650a73548baf63de),U64(0x766a0abb3c77b2a8), + U64(0x81c2c92e47edaee6),U64(0x92722c851482353b), + U64(0xa2bfe8a14cf10364),U64(0xa81a664bbc423001), + U64(0xc24b8b70d0f89791),U64(0xc76c51a30654be30), + U64(0xd192e819d6ef5218),U64(0xd69906245565a910), + U64(0xf40e35855771202a),U64(0x106aa07032bbd1b8), + U64(0x19a4c116b8d2d0c8),U64(0x1e376c085141ab53), + U64(0x2748774cdf8eeb99),U64(0x34b0bcb5e19b48a8), + U64(0x391c0cb3c5c95a63),U64(0x4ed8aa4ae3418acb), + U64(0x5b9cca4f7763e373),U64(0x682e6ff3d6b2b8a3), + U64(0x748f82ee5defb2fc),U64(0x78a5636f43172f60), + U64(0x84c87814a1f0ab72),U64(0x8cc702081a6439ec), + U64(0x90befffa23631e28),U64(0xa4506cebde82bde9), + U64(0xbef9a3f7b2c67915),U64(0xc67178f2e372532b), + U64(0xca273eceea26619c),U64(0xd186b8c721c0c207), + U64(0xeada7dd6cde0eb1e),U64(0xf57d4f7fee6ed178), + U64(0x06f067aa72176fba),U64(0x0a637dc5a2c898a6), + U64(0x113f9804bef90dae),U64(0x1b710b35131c471b), + U64(0x28db77f523047d84),U64(0x32caab7b40c72493), + U64(0x3c9ebe0a15c9bebc),U64(0x431d67c49c100d4c), + U64(0x4cc5d4becb3e42b6),U64(0x597f299cfc657e2a), + U64(0x5fcb6fab3ad6faec),U64(0x6c44198c4a475817) }; + +#ifndef PEDANTIC +# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) +# if defined(__x86_64) || defined(__x86_64__) +# define ROTR(a,n) ({ unsigned long ret; \ + asm ("rorq %1,%0" \ + : "=r"(ret) \ + : "J"(n),"0"(a) \ + : "cc"); ret; }) +# if !defined(B_ENDIAN) +# define PULL64(x) ({ SHA_LONG64 ret=*((const SHA_LONG64 *)(&(x))); \ + asm ("bswapq %0" \ + : "=r"(ret) \ + : "0"(ret)); ret; }) +# endif +# elif (defined(__i386) || defined(__i386__)) && !defined(B_ENDIAN) +# if defined(I386_ONLY) +# define PULL64(x) ({ const unsigned int *p=(const unsigned int *)(&(x));\ + unsigned int hi=p[0],lo=p[1]; \ + asm("xchgb %%ah,%%al;xchgb %%dh,%%dl;"\ + "roll $16,%%eax; roll $16,%%edx; "\ + "xchgb %%ah,%%al;xchgb %%dh,%%dl;" \ + : "=a"(lo),"=d"(hi) \ + : "0"(lo),"1"(hi) : "cc"); \ + ((SHA_LONG64)hi)<<32|lo; }) +# else +# define PULL64(x) ({ const unsigned int *p=(const unsigned int *)(&(x));\ + unsigned int hi=p[0],lo=p[1]; \ + asm ("bswapl %0; bswapl %1;" \ + : "=r"(lo),"=r"(hi) \ + : "0"(lo),"1"(hi)); \ + ((SHA_LONG64)hi)<<32|lo; }) +# endif +# elif (defined(_ARCH_PPC) && defined(__64BIT__)) || defined(_ARCH_PPC64) +# define ROTR(a,n) ({ unsigned long ret; \ + asm ("rotrdi %0,%1,%2" \ + : "=r"(ret) \ + : "r"(a),"K"(n)); ret; }) +# endif +# elif defined(_MSC_VER) +# if defined(_WIN64) /* applies to both IA-64 and AMD64 */ +# define ROTR(a,n) _rotr64((a),n) +# endif +# if defined(_M_IX86) && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) +# if defined(I386_ONLY) + static SHA_LONG64 __fastcall __pull64be(const void *x) + { _asm mov edx, [ecx + 0] + _asm mov eax, [ecx + 4] + _asm xchg dh,dl + _asm xchg ah,al + _asm rol edx,16 + _asm rol eax,16 + _asm xchg dh,dl + _asm xchg ah,al + } +# else + static SHA_LONG64 __fastcall __pull64be(const void *x) + { _asm mov edx, [ecx + 0] + _asm mov eax, [ecx + 4] + _asm bswap edx + _asm bswap eax + } +# endif +# define PULL64(x) __pull64be(&(x)) +# if _MSC_VER<=1200 +# pragma inline_depth(0) +# endif +# endif +# endif +#endif + +#ifndef PULL64 +#define B(x,j) (((SHA_LONG64)(*(((const unsigned char *)(&x))+j)))<<((7-j)*8)) +#define PULL64(x) (B(x,0)|B(x,1)|B(x,2)|B(x,3)|B(x,4)|B(x,5)|B(x,6)|B(x,7)) +#endif + +#ifndef ROTR +#define ROTR(x,s) (((x)>>s) | (x)<<(64-s)) +#endif + +#define Sigma0(x) (ROTR((x),28) ^ ROTR((x),34) ^ ROTR((x),39)) +#define Sigma1(x) (ROTR((x),14) ^ ROTR((x),18) ^ ROTR((x),41)) +#define sigma0(x) (ROTR((x),1) ^ ROTR((x),8) ^ ((x)>>7)) +#define sigma1(x) (ROTR((x),19) ^ ROTR((x),61) ^ ((x)>>6)) + +#define Ch(x,y,z) (((x) & (y)) ^ ((~(x)) & (z))) +#define Maj(x,y,z) (((x) & (y)) ^ ((x) & (z)) ^ ((y) & (z))) + +#if defined(OPENSSL_IA32_SSE2) && !defined(OPENSSL_NO_ASM) && !defined(I386_ONLY) +#define GO_FOR_SSE2(ctx,in,num) do { \ + void sha512_block_sse2(void *,const void *,size_t); \ + if (!(OPENSSL_ia32cap_P & (1<<26))) break; \ + sha512_block_sse2(ctx->h,in,num); return; \ + } while (0) +#endif + +#ifdef OPENSSL_SMALL_FOOTPRINT + +static void sha512_block_data_order (SHA512_CTX *ctx, const void *in, size_t num) + { + const SHA_LONG64 *W=in; + SHA_LONG64 a,b,c,d,e,f,g,h,s0,s1,T1,T2; + SHA_LONG64 X[16]; + int i; + +#ifdef GO_FOR_SSE2 + GO_FOR_SSE2(ctx,in,num); +#endif + + while (num--) { + + a = ctx->h[0]; b = ctx->h[1]; c = ctx->h[2]; d = ctx->h[3]; + e = ctx->h[4]; f = ctx->h[5]; g = ctx->h[6]; h = ctx->h[7]; + + for (i=0;i<16;i++) + { +#ifdef B_ENDIAN + T1 = X[i] = W[i]; +#else + T1 = X[i] = PULL64(W[i]); +#endif + T1 += h + Sigma1(e) + Ch(e,f,g) + K512[i]; + T2 = Sigma0(a) + Maj(a,b,c); + h = g; g = f; f = e; e = d + T1; + d = c; c = b; b = a; a = T1 + T2; + } + + for (;i<80;i++) + { + s0 = X[(i+1)&0x0f]; s0 = sigma0(s0); + s1 = X[(i+14)&0x0f]; s1 = sigma1(s1); + + T1 = X[i&0xf] += s0 + s1 + X[(i+9)&0xf]; + T1 += h + Sigma1(e) + Ch(e,f,g) + K512[i]; + T2 = Sigma0(a) + Maj(a,b,c); + h = g; g = f; f = e; e = d + T1; + d = c; c = b; b = a; a = T1 + T2; + } + + ctx->h[0] += a; ctx->h[1] += b; ctx->h[2] += c; ctx->h[3] += d; + ctx->h[4] += e; ctx->h[5] += f; ctx->h[6] += g; ctx->h[7] += h; + + W+=SHA_LBLOCK; + } + } + +#else + +#define ROUND_00_15(i,a,b,c,d,e,f,g,h) do { \ + T1 += h + Sigma1(e) + Ch(e,f,g) + K512[i]; \ + h = Sigma0(a) + Maj(a,b,c); \ + d += T1; h += T1; } while (0) + +#define ROUND_16_80(i,a,b,c,d,e,f,g,h,X) do { \ + s0 = X[(i+1)&0x0f]; s0 = sigma0(s0); \ + s1 = X[(i+14)&0x0f]; s1 = sigma1(s1); \ + T1 = X[(i)&0x0f] += s0 + s1 + X[(i+9)&0x0f]; \ + ROUND_00_15(i,a,b,c,d,e,f,g,h); } while (0) + +static void sha512_block_data_order (SHA512_CTX *ctx, const void *in, size_t num) + { + const SHA_LONG64 *W=in; + SHA_LONG64 a,b,c,d,e,f,g,h,s0,s1,T1; + SHA_LONG64 X[16]; + int i; + +#ifdef GO_FOR_SSE2 + GO_FOR_SSE2(ctx,in,num); +#endif + + while (num--) { + + a = ctx->h[0]; b = ctx->h[1]; c = ctx->h[2]; d = ctx->h[3]; + e = ctx->h[4]; f = ctx->h[5]; g = ctx->h[6]; h = ctx->h[7]; + +#ifdef B_ENDIAN + T1 = X[0] = W[0]; ROUND_00_15(0,a,b,c,d,e,f,g,h); + T1 = X[1] = W[1]; ROUND_00_15(1,h,a,b,c,d,e,f,g); + T1 = X[2] = W[2]; ROUND_00_15(2,g,h,a,b,c,d,e,f); + T1 = X[3] = W[3]; ROUND_00_15(3,f,g,h,a,b,c,d,e); + T1 = X[4] = W[4]; ROUND_00_15(4,e,f,g,h,a,b,c,d); + T1 = X[5] = W[5]; ROUND_00_15(5,d,e,f,g,h,a,b,c); + T1 = X[6] = W[6]; ROUND_00_15(6,c,d,e,f,g,h,a,b); + T1 = X[7] = W[7]; ROUND_00_15(7,b,c,d,e,f,g,h,a); + T1 = X[8] = W[8]; ROUND_00_15(8,a,b,c,d,e,f,g,h); + T1 = X[9] = W[9]; ROUND_00_15(9,h,a,b,c,d,e,f,g); + T1 = X[10] = W[10]; ROUND_00_15(10,g,h,a,b,c,d,e,f); + T1 = X[11] = W[11]; ROUND_00_15(11,f,g,h,a,b,c,d,e); + T1 = X[12] = W[12]; ROUND_00_15(12,e,f,g,h,a,b,c,d); + T1 = X[13] = W[13]; ROUND_00_15(13,d,e,f,g,h,a,b,c); + T1 = X[14] = W[14]; ROUND_00_15(14,c,d,e,f,g,h,a,b); + T1 = X[15] = W[15]; ROUND_00_15(15,b,c,d,e,f,g,h,a); +#else + T1 = X[0] = PULL64(W[0]); ROUND_00_15(0,a,b,c,d,e,f,g,h); + T1 = X[1] = PULL64(W[1]); ROUND_00_15(1,h,a,b,c,d,e,f,g); + T1 = X[2] = PULL64(W[2]); ROUND_00_15(2,g,h,a,b,c,d,e,f); + T1 = X[3] = PULL64(W[3]); ROUND_00_15(3,f,g,h,a,b,c,d,e); + T1 = X[4] = PULL64(W[4]); ROUND_00_15(4,e,f,g,h,a,b,c,d); + T1 = X[5] = PULL64(W[5]); ROUND_00_15(5,d,e,f,g,h,a,b,c); + T1 = X[6] = PULL64(W[6]); ROUND_00_15(6,c,d,e,f,g,h,a,b); + T1 = X[7] = PULL64(W[7]); ROUND_00_15(7,b,c,d,e,f,g,h,a); + T1 = X[8] = PULL64(W[8]); ROUND_00_15(8,a,b,c,d,e,f,g,h); + T1 = X[9] = PULL64(W[9]); ROUND_00_15(9,h,a,b,c,d,e,f,g); + T1 = X[10] = PULL64(W[10]); ROUND_00_15(10,g,h,a,b,c,d,e,f); + T1 = X[11] = PULL64(W[11]); ROUND_00_15(11,f,g,h,a,b,c,d,e); + T1 = X[12] = PULL64(W[12]); ROUND_00_15(12,e,f,g,h,a,b,c,d); + T1 = X[13] = PULL64(W[13]); ROUND_00_15(13,d,e,f,g,h,a,b,c); + T1 = X[14] = PULL64(W[14]); ROUND_00_15(14,c,d,e,f,g,h,a,b); + T1 = X[15] = PULL64(W[15]); ROUND_00_15(15,b,c,d,e,f,g,h,a); +#endif + + for (i=16;i<80;i+=8) + { + ROUND_16_80(i+0,a,b,c,d,e,f,g,h,X); + ROUND_16_80(i+1,h,a,b,c,d,e,f,g,X); + ROUND_16_80(i+2,g,h,a,b,c,d,e,f,X); + ROUND_16_80(i+3,f,g,h,a,b,c,d,e,X); + ROUND_16_80(i+4,e,f,g,h,a,b,c,d,X); + ROUND_16_80(i+5,d,e,f,g,h,a,b,c,X); + ROUND_16_80(i+6,c,d,e,f,g,h,a,b,X); + ROUND_16_80(i+7,b,c,d,e,f,g,h,a,X); + } + + ctx->h[0] += a; ctx->h[1] += b; ctx->h[2] += c; ctx->h[3] += d; + ctx->h[4] += e; ctx->h[5] += f; ctx->h[6] += g; ctx->h[7] += h; + + W+=SHA_LBLOCK; + } + } + +#endif + +#endif /* SHA512_ASM */ + +#endif /* OPENSSL_NO_SHA512 */ diff --git a/src/lib/libcrypto/sha/sha_locl.h b/src/lib/libcrypto/sha/sha_locl.h index a3623f72da..e37e5726e3 100644 --- a/src/lib/libcrypto/sha/sha_locl.h +++ b/src/lib/libcrypto/sha/sha_locl.h @@ -62,17 +62,11 @@ #include #include -#ifndef SHA_LONG_LOG2 -#define SHA_LONG_LOG2 2 /* default to 32 bits */ -#endif - #define DATA_ORDER_IS_BIG_ENDIAN #define HASH_LONG SHA_LONG -#define HASH_LONG_LOG2 SHA_LONG_LOG2 #define HASH_CTX SHA_CTX #define HASH_CBLOCK SHA_CBLOCK -#define HASH_LBLOCK SHA_LBLOCK #define HASH_MAKE_STRING(c,s) do { \ unsigned long ll; \ ll=(c)->h0; HOST_l2c(ll,(s)); \ @@ -88,12 +82,10 @@ # define HASH_TRANSFORM SHA_Transform # define HASH_FINAL SHA_Final # define HASH_INIT SHA_Init -# define HASH_BLOCK_HOST_ORDER sha_block_host_order # define HASH_BLOCK_DATA_ORDER sha_block_data_order # define Xupdate(a,ix,ia,ib,ic,id) (ix=(a)=(ia^ib^ic^id)) - void sha_block_host_order (SHA_CTX *c, const void *p,int num); - void sha_block_data_order (SHA_CTX *c, const void *p,int num); +static void sha_block_data_order (SHA_CTX *c, const void *p,size_t num); #elif defined(SHA_1) @@ -101,7 +93,6 @@ # define HASH_TRANSFORM SHA1_Transform # define HASH_FINAL SHA1_Final # define HASH_INIT SHA1_Init -# define HASH_BLOCK_HOST_ORDER sha1_block_host_order # define HASH_BLOCK_DATA_ORDER sha1_block_data_order # if defined(__MWERKS__) && defined(__MC68K__) /* Metrowerks for Motorola fails otherwise:-( */ @@ -114,22 +105,10 @@ ) # endif -# ifdef SHA1_ASM -# if defined(__i386) || defined(__i386__) || defined(_M_IX86) || defined(__INTEL__) -# define sha1_block_host_order sha1_block_asm_host_order -# define DONT_IMPLEMENT_BLOCK_HOST_ORDER -# define sha1_block_data_order sha1_block_asm_data_order -# define DONT_IMPLEMENT_BLOCK_DATA_ORDER -# define HASH_BLOCK_DATA_ORDER_ALIGNED sha1_block_asm_data_order -# elif defined(__ia64) || defined(__ia64__) || defined(_M_IA64) -# define sha1_block_host_order sha1_block_asm_host_order -# define DONT_IMPLEMENT_BLOCK_HOST_ORDER -# define sha1_block_data_order sha1_block_asm_data_order -# define DONT_IMPLEMENT_BLOCK_DATA_ORDER -# endif -# endif - void sha1_block_host_order (SHA_CTX *c, const void *p,int num); - void sha1_block_data_order (SHA_CTX *c, const void *p,int num); +#ifndef SHA1_ASM +static +#endif +void sha1_block_data_order (SHA_CTX *c, const void *p,size_t num); #else # error "Either SHA_0 or SHA_1 must be defined." @@ -143,11 +122,7 @@ #define INIT_DATA_h3 0x10325476UL #define INIT_DATA_h4 0xc3d2e1f0UL -#if defined(SHA_0) && defined(OPENSSL_FIPS) -FIPS_NON_FIPS_MD_Init(SHA) -#else int HASH_INIT (SHA_CTX *c) -#endif { c->h0=INIT_DATA_h0; c->h1=INIT_DATA_h1; @@ -177,6 +152,8 @@ int HASH_INIT (SHA_CTX *c) #define F_40_59(b,c,d) (((b) & (c)) | (((b)|(c)) & (d))) #define F_60_79(b,c,d) F_20_39(b,c,d) +#ifndef OPENSSL_SMALL_FOOTPRINT + #define BODY_00_15(i,a,b,c,d,e,f,xi) \ (f)=xi+(e)+K_00_19+ROTATE((a),5)+F_00_19((b),(c),(d)); \ (b)=ROTATE((b),30); @@ -229,11 +206,11 @@ int HASH_INIT (SHA_CTX *c) # define X(i) XX[i] #endif -#ifndef DONT_IMPLEMENT_BLOCK_HOST_ORDER -void HASH_BLOCK_HOST_ORDER (SHA_CTX *c, const void *d, int num) +#if !defined(SHA_1) || !defined(SHA1_ASM) +static void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num) { - const SHA_LONG *W=d; - register unsigned MD32_REG_T A,B,C,D,E,T; + const unsigned char *data=p; + register unsigned MD32_REG_T A,B,C,D,E,T,l; #ifndef MD32_XARRAY unsigned MD32_REG_T XX0, XX1, XX2, XX3, XX4, XX5, XX6, XX7, XX8, XX9,XX10,XX11,XX12,XX13,XX14,XX15; @@ -248,41 +225,71 @@ void HASH_BLOCK_HOST_ORDER (SHA_CTX *c, const void *d, int num) E=c->h4; for (;;) + { + const union { long one; char little; } is_endian = {1}; + + if (!is_endian.little && sizeof(SHA_LONG)==4 && ((size_t)p%4)==0) + { + const SHA_LONG *W=(const SHA_LONG *)data; + + X( 0) = W[0]; X( 1) = W[ 1]; + BODY_00_15( 0,A,B,C,D,E,T,X( 0)); X( 2) = W[ 2]; + BODY_00_15( 1,T,A,B,C,D,E,X( 1)); X( 3) = W[ 3]; + BODY_00_15( 2,E,T,A,B,C,D,X( 2)); X( 4) = W[ 4]; + BODY_00_15( 3,D,E,T,A,B,C,X( 3)); X( 5) = W[ 5]; + BODY_00_15( 4,C,D,E,T,A,B,X( 4)); X( 6) = W[ 6]; + BODY_00_15( 5,B,C,D,E,T,A,X( 5)); X( 7) = W[ 7]; + BODY_00_15( 6,A,B,C,D,E,T,X( 6)); X( 8) = W[ 8]; + BODY_00_15( 7,T,A,B,C,D,E,X( 7)); X( 9) = W[ 9]; + BODY_00_15( 8,E,T,A,B,C,D,X( 8)); X(10) = W[10]; + BODY_00_15( 9,D,E,T,A,B,C,X( 9)); X(11) = W[11]; + BODY_00_15(10,C,D,E,T,A,B,X(10)); X(12) = W[12]; + BODY_00_15(11,B,C,D,E,T,A,X(11)); X(13) = W[13]; + BODY_00_15(12,A,B,C,D,E,T,X(12)); X(14) = W[14]; + BODY_00_15(13,T,A,B,C,D,E,X(13)); X(15) = W[15]; + BODY_00_15(14,E,T,A,B,C,D,X(14)); + BODY_00_15(15,D,E,T,A,B,C,X(15)); + + data += SHA_CBLOCK; + } + else { - BODY_00_15( 0,A,B,C,D,E,T,W[ 0]); - BODY_00_15( 1,T,A,B,C,D,E,W[ 1]); - BODY_00_15( 2,E,T,A,B,C,D,W[ 2]); - BODY_00_15( 3,D,E,T,A,B,C,W[ 3]); - BODY_00_15( 4,C,D,E,T,A,B,W[ 4]); - BODY_00_15( 5,B,C,D,E,T,A,W[ 5]); - BODY_00_15( 6,A,B,C,D,E,T,W[ 6]); - BODY_00_15( 7,T,A,B,C,D,E,W[ 7]); - BODY_00_15( 8,E,T,A,B,C,D,W[ 8]); - BODY_00_15( 9,D,E,T,A,B,C,W[ 9]); - BODY_00_15(10,C,D,E,T,A,B,W[10]); - BODY_00_15(11,B,C,D,E,T,A,W[11]); - BODY_00_15(12,A,B,C,D,E,T,W[12]); - BODY_00_15(13,T,A,B,C,D,E,W[13]); - BODY_00_15(14,E,T,A,B,C,D,W[14]); - BODY_00_15(15,D,E,T,A,B,C,W[15]); - - BODY_16_19(16,C,D,E,T,A,B,X( 0),W[ 0],W[ 2],W[ 8],W[13]); - BODY_16_19(17,B,C,D,E,T,A,X( 1),W[ 1],W[ 3],W[ 9],W[14]); - BODY_16_19(18,A,B,C,D,E,T,X( 2),W[ 2],W[ 4],W[10],W[15]); - BODY_16_19(19,T,A,B,C,D,E,X( 3),W[ 3],W[ 5],W[11],X( 0)); - - BODY_20_31(20,E,T,A,B,C,D,X( 4),W[ 4],W[ 6],W[12],X( 1)); - BODY_20_31(21,D,E,T,A,B,C,X( 5),W[ 5],W[ 7],W[13],X( 2)); - BODY_20_31(22,C,D,E,T,A,B,X( 6),W[ 6],W[ 8],W[14],X( 3)); - BODY_20_31(23,B,C,D,E,T,A,X( 7),W[ 7],W[ 9],W[15],X( 4)); - BODY_20_31(24,A,B,C,D,E,T,X( 8),W[ 8],W[10],X( 0),X( 5)); - BODY_20_31(25,T,A,B,C,D,E,X( 9),W[ 9],W[11],X( 1),X( 6)); - BODY_20_31(26,E,T,A,B,C,D,X(10),W[10],W[12],X( 2),X( 7)); - BODY_20_31(27,D,E,T,A,B,C,X(11),W[11],W[13],X( 3),X( 8)); - BODY_20_31(28,C,D,E,T,A,B,X(12),W[12],W[14],X( 4),X( 9)); - BODY_20_31(29,B,C,D,E,T,A,X(13),W[13],W[15],X( 5),X(10)); - BODY_20_31(30,A,B,C,D,E,T,X(14),W[14],X( 0),X( 6),X(11)); - BODY_20_31(31,T,A,B,C,D,E,X(15),W[15],X( 1),X( 7),X(12)); + HOST_c2l(data,l); X( 0)=l; HOST_c2l(data,l); X( 1)=l; + BODY_00_15( 0,A,B,C,D,E,T,X( 0)); HOST_c2l(data,l); X( 2)=l; + BODY_00_15( 1,T,A,B,C,D,E,X( 1)); HOST_c2l(data,l); X( 3)=l; + BODY_00_15( 2,E,T,A,B,C,D,X( 2)); HOST_c2l(data,l); X( 4)=l; + BODY_00_15( 3,D,E,T,A,B,C,X( 3)); HOST_c2l(data,l); X( 5)=l; + BODY_00_15( 4,C,D,E,T,A,B,X( 4)); HOST_c2l(data,l); X( 6)=l; + BODY_00_15( 5,B,C,D,E,T,A,X( 5)); HOST_c2l(data,l); X( 7)=l; + BODY_00_15( 6,A,B,C,D,E,T,X( 6)); HOST_c2l(data,l); X( 8)=l; + BODY_00_15( 7,T,A,B,C,D,E,X( 7)); HOST_c2l(data,l); X( 9)=l; + BODY_00_15( 8,E,T,A,B,C,D,X( 8)); HOST_c2l(data,l); X(10)=l; + BODY_00_15( 9,D,E,T,A,B,C,X( 9)); HOST_c2l(data,l); X(11)=l; + BODY_00_15(10,C,D,E,T,A,B,X(10)); HOST_c2l(data,l); X(12)=l; + BODY_00_15(11,B,C,D,E,T,A,X(11)); HOST_c2l(data,l); X(13)=l; + BODY_00_15(12,A,B,C,D,E,T,X(12)); HOST_c2l(data,l); X(14)=l; + BODY_00_15(13,T,A,B,C,D,E,X(13)); HOST_c2l(data,l); X(15)=l; + BODY_00_15(14,E,T,A,B,C,D,X(14)); + BODY_00_15(15,D,E,T,A,B,C,X(15)); + } + + BODY_16_19(16,C,D,E,T,A,B,X( 0),X( 0),X( 2),X( 8),X(13)); + BODY_16_19(17,B,C,D,E,T,A,X( 1),X( 1),X( 3),X( 9),X(14)); + BODY_16_19(18,A,B,C,D,E,T,X( 2),X( 2),X( 4),X(10),X(15)); + BODY_16_19(19,T,A,B,C,D,E,X( 3),X( 3),X( 5),X(11),X( 0)); + + BODY_20_31(20,E,T,A,B,C,D,X( 4),X( 4),X( 6),X(12),X( 1)); + BODY_20_31(21,D,E,T,A,B,C,X( 5),X( 5),X( 7),X(13),X( 2)); + BODY_20_31(22,C,D,E,T,A,B,X( 6),X( 6),X( 8),X(14),X( 3)); + BODY_20_31(23,B,C,D,E,T,A,X( 7),X( 7),X( 9),X(15),X( 4)); + BODY_20_31(24,A,B,C,D,E,T,X( 8),X( 8),X(10),X( 0),X( 5)); + BODY_20_31(25,T,A,B,C,D,E,X( 9),X( 9),X(11),X( 1),X( 6)); + BODY_20_31(26,E,T,A,B,C,D,X(10),X(10),X(12),X( 2),X( 7)); + BODY_20_31(27,D,E,T,A,B,C,X(11),X(11),X(13),X( 3),X( 8)); + BODY_20_31(28,C,D,E,T,A,B,X(12),X(12),X(14),X( 4),X( 9)); + BODY_20_31(29,B,C,D,E,T,A,X(13),X(13),X(15),X( 5),X(10)); + BODY_20_31(30,A,B,C,D,E,T,X(14),X(14),X( 0),X( 6),X(11)); + BODY_20_31(31,T,A,B,C,D,E,X(15),X(15),X( 1),X( 7),X(12)); BODY_32_39(32,E,T,A,B,C,D,X( 0),X( 2),X( 8),X(13)); BODY_32_39(33,D,E,T,A,B,C,X( 1),X( 3),X( 9),X(14)); @@ -341,7 +348,7 @@ void HASH_BLOCK_HOST_ORDER (SHA_CTX *c, const void *d, int num) c->h3=(c->h3+B)&0xffffffffL; c->h4=(c->h4+C)&0xffffffffL; - if (--num <= 0) break; + if (--num == 0) break; A=c->h0; B=c->h1; @@ -349,22 +356,48 @@ void HASH_BLOCK_HOST_ORDER (SHA_CTX *c, const void *d, int num) D=c->h3; E=c->h4; - W+=SHA_LBLOCK; - } + } } #endif -#ifndef DONT_IMPLEMENT_BLOCK_DATA_ORDER -void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, int num) +#else /* OPENSSL_SMALL_FOOTPRINT */ + +#define BODY_00_15(xi) do { \ + T=E+K_00_19+F_00_19(B,C,D); \ + E=D, D=C, C=ROTATE(B,30), B=A; \ + A=ROTATE(A,5)+T+xi; } while(0) + +#define BODY_16_19(xa,xb,xc,xd) do { \ + Xupdate(T,xa,xa,xb,xc,xd); \ + T+=E+K_00_19+F_00_19(B,C,D); \ + E=D, D=C, C=ROTATE(B,30), B=A; \ + A=ROTATE(A,5)+T; } while(0) + +#define BODY_20_39(xa,xb,xc,xd) do { \ + Xupdate(T,xa,xa,xb,xc,xd); \ + T+=E+K_20_39+F_20_39(B,C,D); \ + E=D, D=C, C=ROTATE(B,30), B=A; \ + A=ROTATE(A,5)+T; } while(0) + +#define BODY_40_59(xa,xb,xc,xd) do { \ + Xupdate(T,xa,xa,xb,xc,xd); \ + T+=E+K_40_59+F_40_59(B,C,D); \ + E=D, D=C, C=ROTATE(B,30), B=A; \ + A=ROTATE(A,5)+T; } while(0) + +#define BODY_60_79(xa,xb,xc,xd) do { \ + Xupdate(T,xa,xa,xb,xc,xd); \ + T=E+K_60_79+F_60_79(B,C,D); \ + E=D, D=C, C=ROTATE(B,30), B=A; \ + A=ROTATE(A,5)+T+xa; } while(0) + +#if !defined(SHA_1) || !defined(SHA1_ASM) +static void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, size_t num) { const unsigned char *data=p; register unsigned MD32_REG_T A,B,C,D,E,T,l; -#ifndef MD32_XARRAY - unsigned MD32_REG_T XX0, XX1, XX2, XX3, XX4, XX5, XX6, XX7, - XX8, XX9,XX10,XX11,XX12,XX13,XX14,XX15; -#else - SHA_LONG XX[16]; -#endif + int i; + SHA_LONG X[16]; A=c->h0; B=c->h1; @@ -374,101 +407,24 @@ void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, int num) for (;;) { - - HOST_c2l(data,l); X( 0)=l; HOST_c2l(data,l); X( 1)=l; - BODY_00_15( 0,A,B,C,D,E,T,X( 0)); HOST_c2l(data,l); X( 2)=l; - BODY_00_15( 1,T,A,B,C,D,E,X( 1)); HOST_c2l(data,l); X( 3)=l; - BODY_00_15( 2,E,T,A,B,C,D,X( 2)); HOST_c2l(data,l); X( 4)=l; - BODY_00_15( 3,D,E,T,A,B,C,X( 3)); HOST_c2l(data,l); X( 5)=l; - BODY_00_15( 4,C,D,E,T,A,B,X( 4)); HOST_c2l(data,l); X( 6)=l; - BODY_00_15( 5,B,C,D,E,T,A,X( 5)); HOST_c2l(data,l); X( 7)=l; - BODY_00_15( 6,A,B,C,D,E,T,X( 6)); HOST_c2l(data,l); X( 8)=l; - BODY_00_15( 7,T,A,B,C,D,E,X( 7)); HOST_c2l(data,l); X( 9)=l; - BODY_00_15( 8,E,T,A,B,C,D,X( 8)); HOST_c2l(data,l); X(10)=l; - BODY_00_15( 9,D,E,T,A,B,C,X( 9)); HOST_c2l(data,l); X(11)=l; - BODY_00_15(10,C,D,E,T,A,B,X(10)); HOST_c2l(data,l); X(12)=l; - BODY_00_15(11,B,C,D,E,T,A,X(11)); HOST_c2l(data,l); X(13)=l; - BODY_00_15(12,A,B,C,D,E,T,X(12)); HOST_c2l(data,l); X(14)=l; - BODY_00_15(13,T,A,B,C,D,E,X(13)); HOST_c2l(data,l); X(15)=l; - BODY_00_15(14,E,T,A,B,C,D,X(14)); - BODY_00_15(15,D,E,T,A,B,C,X(15)); - - BODY_16_19(16,C,D,E,T,A,B,X( 0),X( 0),X( 2),X( 8),X(13)); - BODY_16_19(17,B,C,D,E,T,A,X( 1),X( 1),X( 3),X( 9),X(14)); - BODY_16_19(18,A,B,C,D,E,T,X( 2),X( 2),X( 4),X(10),X(15)); - BODY_16_19(19,T,A,B,C,D,E,X( 3),X( 3),X( 5),X(11),X( 0)); - - BODY_20_31(20,E,T,A,B,C,D,X( 4),X( 4),X( 6),X(12),X( 1)); - BODY_20_31(21,D,E,T,A,B,C,X( 5),X( 5),X( 7),X(13),X( 2)); - BODY_20_31(22,C,D,E,T,A,B,X( 6),X( 6),X( 8),X(14),X( 3)); - BODY_20_31(23,B,C,D,E,T,A,X( 7),X( 7),X( 9),X(15),X( 4)); - BODY_20_31(24,A,B,C,D,E,T,X( 8),X( 8),X(10),X( 0),X( 5)); - BODY_20_31(25,T,A,B,C,D,E,X( 9),X( 9),X(11),X( 1),X( 6)); - BODY_20_31(26,E,T,A,B,C,D,X(10),X(10),X(12),X( 2),X( 7)); - BODY_20_31(27,D,E,T,A,B,C,X(11),X(11),X(13),X( 3),X( 8)); - BODY_20_31(28,C,D,E,T,A,B,X(12),X(12),X(14),X( 4),X( 9)); - BODY_20_31(29,B,C,D,E,T,A,X(13),X(13),X(15),X( 5),X(10)); - BODY_20_31(30,A,B,C,D,E,T,X(14),X(14),X( 0),X( 6),X(11)); - BODY_20_31(31,T,A,B,C,D,E,X(15),X(15),X( 1),X( 7),X(12)); - - BODY_32_39(32,E,T,A,B,C,D,X( 0),X( 2),X( 8),X(13)); - BODY_32_39(33,D,E,T,A,B,C,X( 1),X( 3),X( 9),X(14)); - BODY_32_39(34,C,D,E,T,A,B,X( 2),X( 4),X(10),X(15)); - BODY_32_39(35,B,C,D,E,T,A,X( 3),X( 5),X(11),X( 0)); - BODY_32_39(36,A,B,C,D,E,T,X( 4),X( 6),X(12),X( 1)); - BODY_32_39(37,T,A,B,C,D,E,X( 5),X( 7),X(13),X( 2)); - BODY_32_39(38,E,T,A,B,C,D,X( 6),X( 8),X(14),X( 3)); - BODY_32_39(39,D,E,T,A,B,C,X( 7),X( 9),X(15),X( 4)); - - BODY_40_59(40,C,D,E,T,A,B,X( 8),X(10),X( 0),X( 5)); - BODY_40_59(41,B,C,D,E,T,A,X( 9),X(11),X( 1),X( 6)); - BODY_40_59(42,A,B,C,D,E,T,X(10),X(12),X( 2),X( 7)); - BODY_40_59(43,T,A,B,C,D,E,X(11),X(13),X( 3),X( 8)); - BODY_40_59(44,E,T,A,B,C,D,X(12),X(14),X( 4),X( 9)); - BODY_40_59(45,D,E,T,A,B,C,X(13),X(15),X( 5),X(10)); - BODY_40_59(46,C,D,E,T,A,B,X(14),X( 0),X( 6),X(11)); - BODY_40_59(47,B,C,D,E,T,A,X(15),X( 1),X( 7),X(12)); - BODY_40_59(48,A,B,C,D,E,T,X( 0),X( 2),X( 8),X(13)); - BODY_40_59(49,T,A,B,C,D,E,X( 1),X( 3),X( 9),X(14)); - BODY_40_59(50,E,T,A,B,C,D,X( 2),X( 4),X(10),X(15)); - BODY_40_59(51,D,E,T,A,B,C,X( 3),X( 5),X(11),X( 0)); - BODY_40_59(52,C,D,E,T,A,B,X( 4),X( 6),X(12),X( 1)); - BODY_40_59(53,B,C,D,E,T,A,X( 5),X( 7),X(13),X( 2)); - BODY_40_59(54,A,B,C,D,E,T,X( 6),X( 8),X(14),X( 3)); - BODY_40_59(55,T,A,B,C,D,E,X( 7),X( 9),X(15),X( 4)); - BODY_40_59(56,E,T,A,B,C,D,X( 8),X(10),X( 0),X( 5)); - BODY_40_59(57,D,E,T,A,B,C,X( 9),X(11),X( 1),X( 6)); - BODY_40_59(58,C,D,E,T,A,B,X(10),X(12),X( 2),X( 7)); - BODY_40_59(59,B,C,D,E,T,A,X(11),X(13),X( 3),X( 8)); - - BODY_60_79(60,A,B,C,D,E,T,X(12),X(14),X( 4),X( 9)); - BODY_60_79(61,T,A,B,C,D,E,X(13),X(15),X( 5),X(10)); - BODY_60_79(62,E,T,A,B,C,D,X(14),X( 0),X( 6),X(11)); - BODY_60_79(63,D,E,T,A,B,C,X(15),X( 1),X( 7),X(12)); - BODY_60_79(64,C,D,E,T,A,B,X( 0),X( 2),X( 8),X(13)); - BODY_60_79(65,B,C,D,E,T,A,X( 1),X( 3),X( 9),X(14)); - BODY_60_79(66,A,B,C,D,E,T,X( 2),X( 4),X(10),X(15)); - BODY_60_79(67,T,A,B,C,D,E,X( 3),X( 5),X(11),X( 0)); - BODY_60_79(68,E,T,A,B,C,D,X( 4),X( 6),X(12),X( 1)); - BODY_60_79(69,D,E,T,A,B,C,X( 5),X( 7),X(13),X( 2)); - BODY_60_79(70,C,D,E,T,A,B,X( 6),X( 8),X(14),X( 3)); - BODY_60_79(71,B,C,D,E,T,A,X( 7),X( 9),X(15),X( 4)); - BODY_60_79(72,A,B,C,D,E,T,X( 8),X(10),X( 0),X( 5)); - BODY_60_79(73,T,A,B,C,D,E,X( 9),X(11),X( 1),X( 6)); - BODY_60_79(74,E,T,A,B,C,D,X(10),X(12),X( 2),X( 7)); - BODY_60_79(75,D,E,T,A,B,C,X(11),X(13),X( 3),X( 8)); - BODY_60_79(76,C,D,E,T,A,B,X(12),X(14),X( 4),X( 9)); - BODY_60_79(77,B,C,D,E,T,A,X(13),X(15),X( 5),X(10)); - BODY_60_79(78,A,B,C,D,E,T,X(14),X( 0),X( 6),X(11)); - BODY_60_79(79,T,A,B,C,D,E,X(15),X( 1),X( 7),X(12)); - - c->h0=(c->h0+E)&0xffffffffL; - c->h1=(c->h1+T)&0xffffffffL; - c->h2=(c->h2+A)&0xffffffffL; - c->h3=(c->h3+B)&0xffffffffL; - c->h4=(c->h4+C)&0xffffffffL; - - if (--num <= 0) break; + for (i=0;i<16;i++) + { HOST_c2l(data,l); X[i]=l; BODY_00_15(X[i]); } + for (i=0;i<4;i++) + { BODY_16_19(X[i], X[i+2], X[i+8], X[(i+13)&15]); } + for (;i<24;i++) + { BODY_20_39(X[i&15], X[(i+2)&15], X[(i+8)&15],X[(i+13)&15]); } + for (i=0;i<20;i++) + { BODY_40_59(X[(i+8)&15],X[(i+10)&15],X[i&15], X[(i+5)&15]); } + for (i=4;i<24;i++) + { BODY_60_79(X[(i+8)&15],X[(i+10)&15],X[i&15], X[(i+5)&15]); } + + c->h0=(c->h0+A)&0xffffffffL; + c->h1=(c->h1+B)&0xffffffffL; + c->h2=(c->h2+C)&0xffffffffL; + c->h3=(c->h3+D)&0xffffffffL; + c->h4=(c->h4+E)&0xffffffffL; + + if (--num == 0) break; A=c->h0; B=c->h1; @@ -479,3 +435,5 @@ void HASH_BLOCK_DATA_ORDER (SHA_CTX *c, const void *p, int num) } } #endif + +#endif -- cgit v1.2.3-55-g6feb