#!/usr/bin/perl
#
# This file is part of Cygwin.
#
# This software is a copyrighted work licensed under the terms of the
# Cygwin license.  Please consult the file "CYGWIN_LICENSE" for
# details.
#
use strict;
use integer;
use Getopt::Long;

sub cleanup(@);

my $cpu;
my $output_def;
GetOptions('cpu=s'=>\$cpu, 'output-def=s'=>\$output_def);

$main::first = 0;
if (!defined($cpu) || !defined($output_def)) {
    die "$0: missing required option\n";
}

my $is_x86_64 = $cpu eq 'x86_64';
# FIXME? Do other (non-32 bit) arches on Windows still use symbol prefixes?
my $sym_prefix = '';

my @top = ();
while (<>) {
    push(@top, cleanup $_);
    last if /^\s*exports$/oi;
}
my @in = cleanup <>;

my %sigfe = ();
my @data = ();
my @nosigfuncs = ();
my @text = ();
for (@in) {
    chomp;
    s/\s+DATA$//o and do {
	push @data, $_;
	next;
    };
    if (/=/o) {
	if (s/\s+NOSIGFE\s*$//) {
	    # nothing
	} elsif (s/\s+SIGFE(_MAYBE)?$//) {
	    my $func = (split(' '))[2];
	    my $maybe = (defined($1) ? lc $1 : '') . '_';
	    $sigfe{$func} = '_sigfe' . $maybe . $func;
	}
    } else {
	my ($func, $sigfe) = m%^\s*(\S+)(?:\s+((?:NO)?SIGFE(?:_MAYBE)?))?$%o;
	if (defined($sigfe) && $sigfe =~ /^NO/o) {
	    $_ = $func;
	} else {
	    $sigfe ||= 'sigfe';
	    $_ = '_' . lc($sigfe) . '_' . $func;
	    $sigfe{$func} = $_;
	    $_ = $func . ' = ' . $_;
	}
    }
    s/(\S)\s+(\S)/$1 $2/go;
    s/(\S)\s+$/$1/o;
    s/^\s+(\S)/$1/o;
    push @text, $_;
}

for (@text) {
    my ($alias, $func) = /^(\S+)\s+=\s+(\S+)\s*$/o;
    $_ = $alias . ' = ' . $sigfe{$func}
      if defined($func) && $sigfe{$func};
}

open OUT, '>', $output_def or die "$0: couldn't open \"$output_def\" - $!\n";
push @top, (map {$_ . " DATA\n"} @data), (map {$_ . "\n"} @text);
print OUT @top;
close OUT;

open SIGFE, '>', 'sigfe.s' or die "$0: couldn't open 'sigfe.s' file for writing - $!\n";

for my $k (sort keys %sigfe) {
    print SIGFE fefunc($k, $sigfe{$k});
}
close SIGFE;

sub fefunc {
    my $func = $sym_prefix . shift;
    my $fe = $sym_prefix . shift;
    my $sigfe_func;
    if ($is_x86_64) {
	$sigfe_func = ($fe =~ /^(.*)_${func}$/)[0];
    }
    my $extra;
    my $res;
    if ($is_x86_64) {
	$res = <<EOF;
	.extern	$func
	.global	$fe
	.seh_proc $fe
$fe:
	leaq	$func(%rip),%r10
	pushq	%r10
	.seh_pushreg %r10
	.seh_endprologue
	jmp	$sigfe_func
	.seh_endproc

EOF
    }
    if (!$main::first++) {
	if ($is_x86_64) {
	  $res = <<EOF . longjmp () . $res;
	.include "tlsoffsets"
	.text

	.seh_proc _sigfe_maybe
_sigfe_maybe:					# stack is aligned on entry!
	.seh_endprologue
	movq	%gs:8,%r10			# location of bottom of stack
	leaq	_cygtls.initialized(%r10),%r11	# where we will be looking
	cmpq	%r11,%rsp			# stack loc > than tls
	jge	0f				# yep.  we don't have a tls.
	movl	_cygtls.initialized(%r10),%r11d
	cmpl	\$0xc763173f,%r11d		# initialized?
	je	1f
0:	ret
	.seh_endproc

	.seh_proc _sigfe
_sigfe:						# stack is aligned on entry!
	.seh_endprologue
	movq	%gs:8,%r10			# location of bottom of stack
1:	movl	\$1,%r11d			# potential lock value
	xchgl	%r11d,_cygtls.stacklock(%r10)	# see if we can grab it
	movl	%r11d,_cygtls.spinning(%r10)	# flag if we are waiting for lock
	testl	%r11d,%r11d			# it will be zero
	jz	2f				#  if so
	pause
	jmp	1b				# loop
2:	movq	\$8,%rax			# have the lock, now increment the
	xaddq	%rax,_cygtls.stackptr(%r10)	#  stack pointer and get pointer
	leaq	_sigbe(%rip),%r11		# new place to return to
	xchgq	%r11,8(%rsp)			# exchange with real return value
	movq	%r11,(%rax)			# store real return value on alt stack
	incl	_cygtls.incyg(%r10)
	decl	_cygtls.stacklock(%r10)		# remove lock
	popq	%rax				# pop real function address from stack
	jmp	*%rax				# and jmp to it
	.seh_endproc

	.global _sigbe
	.seh_proc _sigbe
_sigbe:						# return here after cygwin syscall
						# stack is aligned on entry!
	.seh_endprologue
	movq	%gs:8,%r10			# address of bottom of tls
1:	movl	\$1,%r11d			# potential lock value
	xchgl	%r11d,_cygtls.stacklock(%r10)	# see if we can grab it
	movl	%r11d,_cygtls.spinning(%r10)	# flag if we are waiting for lock
	testl	%r11d,%r11d			# it will be zero
	jz	2f				#  if so
	pause
	jmp	1b				#  and loop
2:	movq	\$-8,%r11			# now decrement aux stack
	xaddq	%r11,_cygtls.stackptr(%r10)	#  and get pointer
	movq	-8(%r11),%r11			# get return address from signal stack
	decl	_cygtls.incyg(%r10)
	decl	_cygtls.stacklock(%r10)		# release lock
	jmp	*%r11				# "return" to caller
	.seh_endproc

	.global	sigdelayed
	.seh_proc sigdelayed
sigdelayed:
	pushq	%r10				# used for return address injection
	.seh_pushreg %r10
	pushq	%rbp
	.seh_pushreg %rbp
	movq	%rsp,%rbp
	pushf
	.seh_pushreg %rax			# fake, there's no .seh_pushreg for the flags
	# stack is aligned or unaligned on entry!
	# make sure it is aligned from here on
	# We could be called from an interrupted thread which doesn't know
	# about his fate, so save and restore everything and the kitchen sink.
	andq	\$0xfffffffffffffff0,%rsp
	.seh_setframe %rbp,0
	pushq	%r15
	.seh_pushreg %r15
	pushq	%r14
	.seh_pushreg %r14
	pushq	%r13
	.seh_pushreg %r13
	pushq	%r12
	.seh_pushreg %r12
	pushq	%r11
	.seh_pushreg %r11
	pushq	%r9
	.seh_pushreg %r9
	pushq	%r8
	.seh_pushreg %r8
	pushq	%rsi
	.seh_pushreg %rsi
	pushq	%rdi
	.seh_pushreg %rdi
	pushq	%rdx
	.seh_pushreg %rdx
	pushq	%rcx
	.seh_pushreg %rcx
	pushq	%rbx
	.seh_pushreg %rbx
	pushq	%rax
	.seh_pushreg %rax
	subq	\$0x128,%rsp
	.seh_stackalloc 0x128
	stmxcsr	0x124(%rsp)
	fnstcw	0x120(%rsp)
	movdqa	%xmm15,0x110(%rsp)
	movdqa	%xmm14,0x100(%rsp)
	movdqa	%xmm13,0xf0(%rsp)
	movdqa	%xmm12,0xe0(%rsp)
	movdqa	%xmm11,0xd0(%rsp)
	movdqa	%xmm10,0xc0(%rsp)
	movdqa	%xmm9,0xb0(%rsp)
	movdqa	%xmm8,0xa0(%rsp)
	movdqa	%xmm7,0x90(%rsp)
	movdqa	%xmm6,0x80(%rsp)
	movdqa	%xmm5,0x70(%rsp)
	movdqa	%xmm4,0x60(%rsp)
	movdqa	%xmm3,0x50(%rsp)
	movdqa	%xmm2,0x40(%rsp)
	movdqa	%xmm1,0x30(%rsp)
	movdqa	%xmm0,0x20(%rsp)
	.seh_endprologue

	movq	%gs:8,%r12			# get tls
	movl	_cygtls.saved_errno(%r12),%r15d	# temporarily save saved_errno
	movq	\$_cygtls.start_offset,%rcx	# point to beginning of tls block
	addq	%r12,%rcx			#  and store as first arg to method
	call	_ZN7_cygtls19call_signal_handlerEv	# call handler

1:	movl	\$1,%r11d			# potential lock value
	xchgl	%r11d,_cygtls.stacklock(%r12)	# see if we can grab it
	movl	%r11d,_cygtls.spinning(%r12)	# flag if we are waiting for lock
	testl	%r11d,%r11d			# it will be zero
	jz	2f				#  if so
	pause
	jmp	1b				#  and loop
2:	testl	%r15d,%r15d			# was saved_errno < 0
	jl	3f				# yup.  ignore it
	movq	_cygtls.errno_addr(%r12),%r11
	movl	%r15d,(%r11)
3:	movq	\$-8,%r11			# now decrement aux stack
	xaddq	%r11,_cygtls.stackptr(%r12)	#  and get pointer
	xorq	%r10,%r10
	xchgq	%r10,-8(%r11)			# get return address from signal stack
	xorl	%r11d,%r11d
	movl	%r11d,_cygtls.incyg(%r12)
	movl	%r11d,_cygtls.stacklock(%r12)	# unlock
	movdqa	0x20(%rsp),%xmm0
	movdqa	0x30(%rsp),%xmm1
	movdqa	0x40(%rsp),%xmm2
	movdqa	0x50(%rsp),%xmm3
	movdqa	0x60(%rsp),%xmm4
	movdqa	0x70(%rsp),%xmm5
	movdqa	0x80(%rsp),%xmm6
	movdqa	0x90(%rsp),%xmm7
	movdqa	0xa0(%rsp),%xmm8
	movdqa	0xb0(%rsp),%xmm9
	movdqa	0xc0(%rsp),%xmm10
	movdqa	0xd0(%rsp),%xmm11
	movdqa	0xe0(%rsp),%xmm12
	movdqa	0xf0(%rsp),%xmm13
	movdqa	0x100(%rsp),%xmm14
	movdqa	0x110(%rsp),%xmm15
	fninit
	fldcw	0x120(%rsp)
	ldmxcsr	0x124(%rsp)
	addq	\$0x128,%rsp
	popq	%rax
	popq	%rbx
	popq	%rcx
	popq	%rdx
	popq	%rdi
	popq	%rsi
	popq	%r8
	popq	%r9
	popq	%r11
	popq	%r12
	popq	%r13
	popq	%r14
	popq	%r15
	movq	%rbp,%rsp
	subq	\$8, %rsp
	popf
	popq	%rbp
	xchgq	%r10,(%rsp)
	ret
	.seh_endproc
_sigdelayed_end:
	.global _sigdelayed_end

# _cygtls::pop
	.global _ZN7_cygtls3popEv
	.seh_proc _ZN7_cygtls3popEv
_ZN7_cygtls3popEv:
	.seh_endprologue
	movq	\$-8,%r11
	xaddq	%r11,_cygtls.stackptr_p(%rcx)
	movq	-8(%r11),%rax
	ret
	.seh_endproc

# _cygtls::lock
	.global _ZN7_cygtls4lockEv
	.seh_proc _ZN7_cygtls4lockEv
_ZN7_cygtls4lockEv:
	pushq	%r12
	.seh_pushreg %r12
	.seh_endprologue
	movq	%rcx,%r12
1:	movl	\$1,%r11d
	xchgl	%r11d,_cygtls.stacklock_p(%r12)
	testl   %r11d,%r11d
	jz	2f
	pause
	jmp	1b
2:	popq	%r12
	ret
	.seh_endproc

# _cygtls::unlock
	.global _ZN7_cygtls6unlockEv
	.seh_proc _ZN7_cygtls6unlockEv
_ZN7_cygtls6unlockEv:
	.seh_endprologue
	decl	_cygtls.stacklock_p(%rcx)
	ret
	.seh_endproc

# _cygtls::locked
	.global _ZN7_cygtls6lockedEv
	.seh_proc _ZN7_cygtls6lockedEv
_ZN7_cygtls6lockedEv:
	.seh_endprologue
	movl	_cygtls.stacklock_p(%rcx),%eax
	ret
	.seh_endproc

	.seh_proc stabilize_sig_stack
stabilize_sig_stack:
	pushq	%r12
	.seh_pushreg %r12
	subq	\$0x20,%rsp
	.seh_stackalloc 32
	.seh_endprologue
	movq	%gs:8,%r12
1:	movl	\$1,%r10d
	xchgl	%r10d,_cygtls.stacklock(%r12)
	movl	%r10d,_cygtls.spinning(%r12)	# flag if we are waiting for lock
	testl	%r10d,%r10d
	jz	2f
	pause
	jmp	1b
2:	incl	_cygtls.incyg(%r12)
	cmpl	\$0,_cygtls.sig(%r12)
	jz	3f
	decl	_cygtls.stacklock(%r12)		# unlock
	movq	\$_cygtls.start_offset,%rcx	# point to beginning
	addq	%r12,%rcx			#  of tls block
	call	_ZN7_cygtls19call_signal_handlerEv
	jmp	1b
3:	decl	_cygtls.incyg(%r12)
	addq	\$0x20,%rsp
	movq	%r12,%r11			# return tls addr in r11
	popq	%r12
	ret
	.seh_endproc
EOF
	}
    }
    return $res;
}

sub longjmp {
    if ($is_x86_64) {
	return <<EOF;

	.globl	sigsetjmp
	.seh_proc sigsetjmp
sigsetjmp:
	.seh_endprologue
	movl	%edx,0x100(%rcx)		# store savemask
	testl	%edx,%edx			# savemask != 0?
	je	setjmp				# no, skip fetching sigmask
	pushq	%rcx
	subq	\$0x20,%rsp
	leaq	0x108(%rcx),%r8			# &sigjmp_buf.sigmask
	xorq	%rdx,%rdx			# NULL
	xorl	%ecx,%ecx			# SIG_SETMASK
	call	pthread_sigmask
	addq	\$0x20,%rsp
	popq	%rcx
	jmp	setjmp
	.seh_endproc

	.globl  setjmp
	.seh_proc setjmp
setjmp:
	.seh_endprologue
	# We use the Windows jmp_buf layout with two small twists.
	# - we store the tls stackptr in Frame, MSVCRT stores a second copy
	#   of %rbp in Frame (twice? why?)
	# - we just store %rsp as is, MSVCRT stores %rsp of the caller in Rsp
	movq	%rbx,0x8(%rcx)
	movq	%rsp,0x10(%rcx)
	movq	%rbp,0x18(%rcx)
	movq	%rsi,0x20(%rcx)
	movq	%rdi,0x28(%rcx)
	movq	%r12,0x30(%rcx)
	movq	%r13,0x38(%rcx)
	movq	%r14,0x40(%rcx)
	movq	%r15,0x48(%rcx)
	movq	(%rsp),%r10
	movq	%r10,0x50(%rcx)
	stmxcsr	0x58(%rcx)
	fnstcw	0x5c(%rcx)
	# jmp_buf is potentially unaligned!
	movdqu	%xmm6,0x60(%rcx)
	movdqu	%xmm7,0x70(%rcx)
	movdqu	%xmm8,0x80(%rcx)
	movdqu	%xmm9,0x90(%rcx)
	movdqu	%xmm10,0xa0(%rcx)
	movdqu	%xmm11,0xb0(%rcx)
	movdqu	%xmm12,0xc0(%rcx)
	movdqu	%xmm13,0xd0(%rcx)
	movdqu	%xmm14,0xe0(%rcx)
	movdqu	%xmm15,0xf0(%rcx)
	pushq	%rcx
	.seh_pushreg %rcx
	call	stabilize_sig_stack		# returns tls in r11
	popq	%rcx
	movq	_cygtls.stackptr(%r11),%r10
	movq	%r10,(%rcx)
	decl	_cygtls.stacklock(%r11)
	xorl	%eax,%eax
	ret
	.seh_endproc

	.globl	siglongjmp
	.seh_proc siglongjmp
siglongjmp:
	pushq	%rcx
	.seh_pushreg %rcx
	.seh_endprologue
	movl	%edx, %r12d
	movl	0x100(%rcx),%r8d		# savemask
	testl	%r8d,%r8d			# savemask != 0?
	je	1f				# no, jmp to longjmp
	xorq	%r8,%r8				# NULL
	leaq    0x108(%rcx),%rdx		# &sigjmp_buf.sigmask
	xorl	%ecx,%ecx			# SIG_SETMASK
	subq	\$0x20,%rsp
	call	pthread_sigmask
	addq	\$0x20,%rsp
	jmp	1f
	.seh_endproc

	.globl  longjmp
	.seh_proc longjmp
longjmp:
	pushq	%rcx
	.seh_pushreg %rcx
	.seh_endprologue
	movl	%edx,%r12d			# save return value
1:
	call	stabilize_sig_stack		# returns tls in r11
	popq	%rcx
	movl	%r12d,%eax			# restore return value
	movq	(%rcx),%r10			# get old signal stack
	movq	%r10,_cygtls.stackptr(%r11)	# restore
	decl	_cygtls.stacklock(%r11)		# relinquish lock
	xorl	%r10d,%r10d
	movl	%r10d,_cygtls.incyg(%r11)		# we're not in cygwin anymore
	movq	0x8(%rcx),%rbx
	movq	0x10(%rcx),%rsp
	movq	0x18(%rcx),%rbp
	movq	0x20(%rcx),%rsi
	movq	0x28(%rcx),%rdi
	movq	0x30(%rcx),%r12
	movq	0x38(%rcx),%r13
	movq	0x40(%rcx),%r14
	movq	0x48(%rcx),%r15
	movq	0x50(%rcx),%r10
	movq	%r10,(%rsp)
	ldmxcsr	0x58(%rcx)
	fnclex
	fldcw	0x5c(%rcx)
	# jmp_buf is potentially unaligned!
	movdqu	0x60(%rcx),%xmm6
	movdqu	0x70(%rcx),%xmm7
	movdqu	0x80(%rcx),%xmm8
	movdqu	0x90(%rcx),%xmm9
	movdqu	0xa0(%rcx),%xmm10
	movdqu	0xb0(%rcx),%xmm11
	movdqu	0xc0(%rcx),%xmm12
	movdqu	0xd0(%rcx),%xmm13
	movdqu	0xe0(%rcx),%xmm14
	movdqu	0xf0(%rcx),%xmm15
	testl	%eax,%eax
	jne	0f
	incl	%eax
0:	ret
	.seh_endproc
EOF
    }
}

sub cleanup(@) {
    grep {s/\r//og; s/#.*//og; s/\s+\n//sog; !/^$/o && $_} @_;
}