x86_64-xlate.pl

OpenSSL 0.9.8k 最新版OpenSSL

#!/usr/bin/env perl

# Ascetic x86_64 AT&T to MASM assembler translator by <appro>.
#
# Why AT&T to MASM and not vice versa? Several reasons. Because AT&T
# format is way easier to parse. Because it's simpler to "gear" from
# Unix ABI to Windows one [see cross-reference "card" at the end of
# file]. Because Linux targets were available first...
#
# In addition the script also "distills" code suitable for GNU
# assembler, so that it can be compiled with more rigid assemblers,
# such as Solaris /usr/ccs/bin/as.
#
# This translator is not designed to convert *arbitrary* assembler
# code from AT&T format to MASM one. It's designed to convert just
# enough to provide for dual-ABI OpenSSL modules development...
# There *are* limitations and you might have to modify your assembler
# code or this script to achieve the desired result...
#
# Currently recognized limitations:
#
# - can't use multiple ops per line;
# - indirect calls and jumps are not supported;
#
# Dual-ABI styling rules.
#
# 1. Adhere to Unix register and stack layout [see the end for
#    explanation].
# 2. Forget about "red zone," stick to more traditional blended
#    stack frame allocation. If volatile storage is actually required
#    that is. If not, just leave the stack as is.
# 3. Functions tagged with ".type name,@function" get crafted with
#    unified Win64 prologue and epilogue automatically. If you want
#    to take care of ABI differences yourself, tag functions as
#    ".type name,@abi-omnipotent" instead.
# 4. To optimize the Win64 prologue you can specify number of input
#    arguments as ".type name,@function,N." Keep in mind that if N is
#    larger than 6, then you *have to* write "abi-omnipotent" code,
#    because >6 cases can't be addressed with unified prologue.
# 5. Name local labels as .L*, do *not* use dynamic labels such as 1:
#    (sorry about latter).
# 6. Don't use [or hand-code with .byte] "rep ret." "ret" mnemonic is
#    required to identify the spots, where to inject Win64 epilogue!
#    But on the pros, it's then prefixed with rep automatically:-)
# 7. Due to MASM limitations [and certain general counter-intuitivity
#    of ip-relative addressing] generation of position-independent
#    code is assisted by synthetic directive, .picmeup, which puts
#    address of the *next* instruction into target register.
#
#    Example 1:
#		.picmeup	%rax
#		lea		.Label-.(%rax),%rax
#    Example 2:
#		.picmeup	%rcx
#	.Lpic_point:
#		...
#		lea		.Label-.Lpic_point(%rcx),%rbp

my $output = shift;

{ my ($stddev,$stdino,@junk)=stat(STDOUT);
  my ($outdev,$outino,@junk)=stat($output);

    open STDOUT,">$output" || die "can't open $output: $!"
	if ($stddev!=$outdev || $stdino!=$outino);
}

my $masmref=8 + 50727*2**-32;	# 8.00.50727 shipped with VS2005
my $masm=$masmref if ($output =~ /\.asm/);
if ($masm && `ml64 2>&1` =~ m/Version ([0-9]+)\.([0-9]+)(\.([0-9]+))?/)
{   $masm=$1 + $2*2**-16 + $4*2**-32;   }

my $current_segment;
my $current_function;

{ package opcode;	# pick up opcodes
    sub re {
	my	$self = shift;	# single instance in enough...
	local	*line = shift;
	undef	$ret;

	if ($line =~ /^([a-z][a-z0-9]*)/i) {
	    $self->{op} = $1;
	    $ret = $self;
	    $line = substr($line,@+[0]); $line =~ s/^\s+//;

	    undef $self->{sz};
	    if ($self->{op} =~ /^(movz)b.*/) {	# movz is pain...
		$self->{op} = $1;
		$self->{sz} = "b";
	    } elsif ($self->{op} =~ /call/) {
		$self->{sz} = ""
	    } elsif ($self->{op} =~ /([a-z]{3,})([qlwb])$/) {
		$self->{op} = $1;
		$self->{sz} = $2;
	    }
	}
	$ret;
    }
    sub size {
	my $self = shift;
	my $sz   = shift;
	$self->{sz} = $sz if (defined($sz) && !defined($self->{sz}));
	$self->{sz};
    }
    sub out {
	my $self = shift;
	if (!$masm) {
	    if ($self->{op} eq "movz") {	# movz is pain...
		sprintf "%s%s%s",$self->{op},$self->{sz},shift;
	    } elsif ($self->{op} =~ /^set/) { 
		"$self->{op}";
	    } elsif ($self->{op} eq "ret") {
	    	".byte	0xf3,0xc3";
	    } else {
		"$self->{op}$self->{sz}";
	    }
	} else {
	    $self->{op} =~ s/^movz/movzx/;
	    if ($self->{op} eq "ret") {
		$self->{op} = "";
		if ($current_function->{abi} eq "svr4") {
		    $self->{op} = "mov	rdi,QWORD PTR 8[rsp]\t;WIN64 epilogue\n\t".
				  "mov	rsi,QWORD PTR 16[rsp]\n\t";
	    	}
		$self->{op} .= "DB\t0F3h,0C3h\t\t;repret";
	    }
	    $self->{op};
	}
    }
}
{ package const;	# pick up constants, which start with $
    sub re {
	my	$self = shift;	# single instance in enough...
	local	*line = shift;
	undef	$ret;

	if ($line =~ /^\$([^,]+)/) {
	    $self->{value} = $1;
	    $ret = $self;
	    $line = substr($line,@+[0]); $line =~ s/^\s+//;
	}
	$ret;
    }
    sub out {
    	my $self = shift;

	if (!$masm) {
	    # Solaris /usr/ccs/bin/as can't handle multiplications
	    # in $self->{value}
	    $self->{value} =~ s/(?<![0-9a-f])(0[x0-9a-f]+)/oct($1)/egi;
	    $self->{value} =~ s/([0-9]+\s*[\*\/\%]\s*[0-9]+)/eval($1)/eg;
	    sprintf "\$%s",$self->{value};
	} else {
	    $self->{value} =~ s/0x([0-9a-f]+)/0$1h/ig;
	    sprintf "%s",$self->{value};
	}
    }
}
{ package ea;		# pick up effective addresses: expr(%reg,%reg,scale)
    sub re {
	my	$self = shift;	# single instance in enough...
	local	*line = shift;
	undef	$ret;

	if ($line =~ /^([^\(,]*)\(([%\w,]+)\)/) {
	    $self->{label} = $1;
	    ($self->{base},$self->{index},$self->{scale})=split(/,/,$2);
	    $self->{scale} = 1 if (!defined($self->{scale}));
	    $ret = $self;
	    $line = substr($line,@+[0]); $line =~ s/^\s+//;

	    $self->{base}  =~ s/^%//;
	    $self->{index} =~ s/^%// if (defined($self->{index}));
	}
	$ret;
    }
    sub size {}
    sub out {
    	my $self = shift;
	my $sz = shift;

	# Silently convert all EAs to 64-bit. This is required for
	# elder GNU assembler and results in more compact code,
	# *but* most importantly AES module depends on this feature!
	$self->{index} =~ s/^[er](.?[0-9xpi])[d]?$/r\1/;
	$self->{base}  =~ s/^[er](.?[0-9xpi])[d]?$/r\1/;

	if (!$masm) {
	    # Solaris /usr/ccs/bin/as can't handle multiplications
	    # in $self->{label}
	    $self->{label} =~ s/(?<![0-9a-f])(0[x0-9a-f]+)/oct($1)/egi;
	    $self->{label} =~ s/([0-9]+\s*[\*\/\%]\s*[0-9]+)/eval($1)/eg;

	    if (defined($self->{index})) {
		sprintf "%s(%%%s,%%%s,%d)",
					$self->{label},$self->{base},
					$self->{index},$self->{scale};
	    } else {
		sprintf "%s(%%%s)",	$self->{label},$self->{base};
	    }
	} else {
	    %szmap = ( b=>"BYTE", w=>"WORD", l=>"DWORD", q=>"QWORD" );

	    $self->{label} =~ s/\./\$/g;
	    $self->{label} =~ s/0x([0-9a-f]+)/0$1h/ig;
	    $self->{label} = "($self->{label})" if ($self->{label} =~ /[\*\+\-\/]/);

	    if (defined($self->{index})) {
		sprintf "%s PTR %s[%s*%d+%s]",$szmap{$sz},
					$self->{label},
					$self->{index},$self->{scale},
					$self->{base};
	    } elsif ($self->{base} eq "rip") {
		sprintf "%s PTR %s",$szmap{$sz},$self->{label};
	    } else {
		sprintf "%s PTR %s[%s]",$szmap{$sz},
					$self->{label},$self->{base};
	    }
	}
    }
}
{ package register;	# pick up registers, which start with %.
    sub re {
	my	$class = shift;	# muliple instances...
	my	$self = {};
	local	*line = shift;
	undef	$ret;

	if ($line =~ /^%(\w+)/) {
	    bless $self,$class;
	    $self->{value} = $1;
	    $ret = $self;
	    $line = substr($line,@+[0]); $line =~ s/^\s+//;
	}
	$ret;
    }
    sub size {
	my	$self = shift;
	undef	$ret;

	if    ($self->{value} =~ /^r[\d]+b$/i)	{ $ret="b"; }
	elsif ($self->{value} =~ /^r[\d]+w$/i)	{ $ret="w"; }
	elsif ($self->{value} =~ /^r[\d]+d$/i)	{ $ret="l"; }
	elsif ($self->{value} =~ /^r[\w]+$/i)	{ $ret="q"; }
	elsif ($self->{value} =~ /^[a-d][hl]$/i){ $ret="b"; }
	elsif ($self->{value} =~ /^[\w]{2}l$/i)	{ $ret="b"; }
	elsif ($self->{value} =~ /^[\w]{2}$/i)	{ $ret="w"; }
	elsif ($self->{value} =~ /^e[a-z]{2}$/i){ $ret="l"; }

	$ret;
    }
    sub out {
    	my $self = shift;
	sprintf $masm?"%s":"%%%s",$self->{value};
    }
}
{ package label;	# pick up labels, which end with :
    sub re {
	my	$self = shift;	# single instance is enough...
	local	*line = shift;
	undef	$ret;

	if ($line =~ /(^[\.\w]+\:)/) {
	    $self->{value} = $1;
	    $ret = $self;
	    $line = substr($line,@+[0]); $line =~ s/^\s+//;

	    $self->{value} =~ s/\.L/\$L/ if ($masm);
	}
	$ret;
    }
    sub out {
	my $self = shift;

	if (!$masm) {
	    $self->{value};
	} elsif ($self->{value} ne "$current_function->{name}:") {