x86_64 assembler pack: add support for Win64 SEH.

8 files changed, 1174 insertions, 153 deletions

diff --git a/crypto/aes/asm/aes-x86_64.pl b/crypto/aes/asm/aes-x86_64.pl
index c2b040a84f..c75e0ea2f8 100755
--- a/crypto/aes/asm/aes-x86_64.pl
+++ b/crypto/aes/asm/aes-x86_64.pl
@@ -25,17 +25,22 @@
 #
 # (*) with hyper-threading off
 
-$verticalspin=1;	# unlike 32-bit version $verticalspin performs
-			# ~15% better on both AMD and Intel cores
-$speed_limit=512;	# see aes-586.pl for details
-$output=shift;
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
 
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $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";
+open STDOUT,"| $^X $xlate $flavour $output";
+
+$verticalspin=1;	# unlike 32-bit version $verticalspin performs
+			# ~15% better on both AMD and Intel cores
+$speed_limit=512;	# see aes-586.pl for details
 
 $code=".text\n";
 
@@ -592,18 +597,20 @@ AES_encrypt:
 	push	%r15
 
 	# allocate frame "above" key schedule
-	mov	%rsp,%rax
-	mov	%rdx,$key
-	lea	-63(%rdx),%rcx
+	mov	%rsp,%r10
+	lea	-63(%rdx),%rcx	# %rdx is key argument
 	and	\$-64,%rsp
 	sub	%rsp,%rcx
 	neg	%rcx
 	and	\$0x3c0,%rcx
 	sub	%rcx,%rsp
+	sub	\$32,%rsp
 
-	push	%rax		# save real stack pointer
-	push	%rsi		# save out
+	mov	%rsi,16(%rsp)	# save out
+	mov	%r10,24(%rsp)	# save real stack pointer
+.Lenc_prologue:
 
+	mov	%rdx,$key
 	mov	240($key),$rnds	# load rounds
 
 	mov	0(%rdi),$s0	# load input vector
@@ -613,8 +620,8 @@ AES_encrypt:
 
 	shl	\$4,$rnds
 	lea	($key,$rnds),%rbp
-	push	%rbp
-	push	$key
+	mov	$key,(%rsp)	# key schedule
+	mov	%rbp,8(%rsp)	# end of key schedule
 
 	# pick Te4 copy which can't "overlap" with stack frame or key schedule
 	lea	.LAES_Te+2048(%rip),$sbox
@@ -626,18 +633,20 @@ AES_encrypt:
 	call	_x86_64_AES_encrypt_compact
 
 	mov	16(%rsp),$out	# restore out
-	mov	24(%rsp),%rsp
+	mov	24(%rsp),%rsi	# restore saved stack pointer
 	mov	$s0,0($out)	# write output vector
 	mov	$s1,4($out)
 	mov	$s2,8($out)
 	mov	$s3,12($out)
 
-	pop	%r15
-	pop	%r14
-	pop	%r13
-	pop	%r12
-	pop	%rbp
-	pop	%rbx
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lenc_epilogue:
 	ret
 .size	AES_encrypt,.-AES_encrypt
 ___
@@ -1184,18 +1193,20 @@ AES_decrypt:
 	push	%r15
 
 	# allocate frame "above" key schedule
-	mov	%rsp,%rax
-	mov	%rdx,$key
-	lea	-63(%rdx),%rcx
+	mov	%rsp,%r10
+	lea	-63(%rdx),%rcx	# %rdx is key argument
 	and	\$-64,%rsp
 	sub	%rsp,%rcx
 	neg	%rcx
 	and	\$0x3c0,%rcx
 	sub	%rcx,%rsp
+	sub	\$32,%rsp
 
-	push	%rax		# save real stack pointer
-	push	%rsi		# save out
+	mov	%rsi,16(%rsp)	# save out
+	mov	%r10,24(%rsp)	# save real stack pointer
+.Ldec_prologue:
 
+	mov	%rdx,$key
 	mov	240($key),$rnds	# load rounds
 
 	mov	0(%rdi),$s0	# load input vector
@@ -1205,8 +1216,8 @@ AES_decrypt:
 
 	shl	\$4,$rnds
 	lea	($key,$rnds),%rbp
-	push	%rbp
-	push	$key
+	mov	$key,(%rsp)	# key schedule
+	mov	%rbp,8(%rsp)	# end of key schedule
 
 	# pick Td4 copy which can't "overlap" with stack frame or key schedule
 	lea	.LAES_Td+2048(%rip),$sbox
@@ -1220,18 +1231,20 @@ AES_decrypt:
 	call	_x86_64_AES_decrypt_compact
 
 	mov	16(%rsp),$out	# restore out
-	mov	24(%rsp),%rsp
+	mov	24(%rsp),%rsi	# restore saved stack pointer
 	mov	$s0,0($out)	# write output vector
 	mov	$s1,4($out)
 	mov	$s2,8($out)
 	mov	$s3,12($out)
 
-	pop	%r15
-	pop	%r14
-	pop	%r13
-	pop	%r12
-	pop	%rbp
-	pop	%rbx
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Ldec_epilogue:
 	ret
 .size	AES_decrypt,.-AES_decrypt
 ___
@@ -1271,16 +1284,31 @@ $code.=<<___;
 .type	AES_set_encrypt_key,\@function,3
 .align	16
 AES_set_encrypt_key:
+	push	%rbx
+	push	%rbp
+	push	%r12			# redundant, but allows to share 
+	push	%r13			# exception handler...
+	push	%r14
+	push	%r15
+	sub	\$8,%rsp
+.Lenc_key_prologue:
+
 	call	_x86_64_AES_set_encrypt_key
+
+	mov	8(%rsp),%r15
+	mov	16(%rsp),%r14
+	mov	24(%rsp),%r13
+	mov	32(%rsp),%r12
+	mov	40(%rsp),%rbp
+	mov	48(%rsp),%rbx
+	add	\$56,%rsp
+.Lenc_key_epilogue:
 	ret
 .size	AES_set_encrypt_key,.-AES_set_encrypt_key
 
 .type	_x86_64_AES_set_encrypt_key,\@abi-omnipotent
 .align	16
 _x86_64_AES_set_encrypt_key:
-	push	%rbx
-	push	%rbp
-
 	mov	%esi,%ecx			# %ecx=bits
 	mov	%rdi,%rsi			# %rsi=userKey
 	mov	%rdx,%rdi			# %rdi=key
@@ -1461,8 +1489,6 @@ $code.=<<___;
 .Lbadpointer:
 	mov	\$-1,%rax
 .Lexit:
-	pop	%rbp
-	pop	%rbx
 	.byte	0xf3,0xc3			# rep ret
 .size	_x86_64_AES_set_encrypt_key,.-_x86_64_AES_set_encrypt_key
 ___
@@ -1528,18 +1554,19 @@ $code.=<<___;
 .type	AES_set_decrypt_key,\@function,3
 .align	16
 AES_set_decrypt_key:
-	push	%rdx			# save key schedule
-	call	_x86_64_AES_set_encrypt_key
-	cmp	\$0,%eax
-	pop	%r8			# restore key schedule
-	jne	.Labort
-
 	push	%rbx
 	push	%rbp
 	push	%r12
 	push	%r13
 	push	%r14
 	push	%r15
+	push	%rdx			# save key schedule
+.Ldec_key_prologue:
+
+	call	_x86_64_AES_set_encrypt_key
+	mov	(%rsp),%r8		# restore key schedule
+	cmp	\$0,%eax
+	jne	.Labort
 
 	mov	240(%r8),%r14d		# pull number of rounds
 	xor	%rdi,%rdi
@@ -1585,13 +1612,15 @@ $code.=<<___;
 	jnz	.Lpermute
 
 	xor	%rax,%rax
-	pop	%r15
-	pop	%r14
-	pop	%r13
-	pop	%r12
-	pop	%rbp
-	pop	%rbx
 .Labort:
+	mov	8(%rsp),%r15
+	mov	16(%rsp),%r14
+	mov	24(%rsp),%r13
+	mov	32(%rsp),%r12
+	mov	40(%rsp),%rbp
+	mov	48(%rsp),%rbx
+	add	\$56,%rsp
+.Ldec_key_epilogue:
 	ret
 .size	AES_set_decrypt_key,.-AES_set_decrypt_key
 ___
@@ -1621,14 +1650,16 @@ $code.=<<___;
 .extern	OPENSSL_ia32cap_P
 AES_cbc_encrypt:
 	cmp	\$0,%rdx	# check length
-	je	.Lcbc_just_ret
+	je	.Lcbc_epilogue
+	pushfq
 	push	%rbx
 	push	%rbp
 	push	%r12
 	push	%r13
 	push	%r14
 	push	%r15
-	pushfq
+.Lcbc_prologue:
+
 	cld
 	mov	%r9d,%r9d	# clear upper half of enc
 
@@ -1638,13 +1669,13 @@ AES_cbc_encrypt:
 	lea	.LAES_Td(%rip),$sbox
 .Lcbc_picked_te:
 
-	mov	OPENSSL_ia32cap_P(%rip),%eax
+	mov	OPENSSL_ia32cap_P(%rip),%r10d
 	cmp	\$$speed_limit,%rdx
-	jb	.Lcbc_slow_way
+	jb	.Lcbc_slow_prologue
 	test	\$15,%rdx
-	jnz	.Lcbc_slow_way
-	bt	\$28,%eax
-	jc	.Lcbc_slow_way
+	jnz	.Lcbc_slow_prologue
+	bt	\$28,%r10d
+	jc	.Lcbc_slow_prologue
 
 	# allocate aligned stack frame...
 	lea	-88-248(%rsp),$key
@@ -1672,8 +1703,9 @@ AES_cbc_encrypt:
 .Lcbc_te_ok:
 
 	xchg	%rsp,$key
-	add	\$8,%rsp	# reserve for return address!
+	#add	\$8,%rsp	# reserve for return address!
 	mov	$key,$_rsp	# save %rsp
+.Lcbc_fast_body:
 	mov	%rdi,$_inp	# save copy of inp
 	mov	%rsi,$_out	# save copy of out
 	mov	%rdx,$_len	# save copy of len
@@ -1757,25 +1789,7 @@ AES_cbc_encrypt:
 	mov	$s2,8(%rbp)
 	mov	$s3,12(%rbp)
 
-.align	4
-.Lcbc_cleanup:
-	cmpl	\$0,$mark	# was the key schedule copied?
-	lea	$aes_key,%rdi
-	je	.Lcbc_exit
-		mov	\$240/8,%ecx
-		xor	%rax,%rax
-		.long	0x90AB48F3	# rep stosq
-.Lcbc_exit:
-	mov	$_rsp,%rsp
-	popfq
-	pop	%r15
-	pop	%r14
-	pop	%r13
-	pop	%r12
-	pop	%rbp
-	pop	%rbx
-.Lcbc_just_ret:
-	ret
+	jmp	.Lcbc_fast_cleanup
 
 #----------------------------- DECRYPT -----------------------------#
 .align	16
@@ -1821,7 +1835,7 @@ AES_cbc_encrypt:
 	mov	8(%rbp),%r11
 	mov	%r10,0(%r12)		# copy back to user
 	mov	%r11,8(%r12)
-	jmp	.Lcbc_cleanup
+	jmp	.Lcbc_fast_cleanup
 
 .align	16
 .Lcbc_fast_dec_in_place:
@@ -1874,24 +1888,34 @@ AES_cbc_encrypt:
 	mov	$s2,8($out)
 	mov	$s3,12($out)
 
-	jmp	.Lcbc_cleanup
+.align	4
+.Lcbc_fast_cleanup:
+	cmpl	\$0,$mark	# was the key schedule copied?
+	lea	$aes_key,%rdi
+	je	.Lcbc_exit
+		mov	\$240/8,%ecx
+		xor	%rax,%rax
+		.long	0x90AB48F3	# rep stosq
+
+	jmp	.Lcbc_exit
 
 #--------------------------- SLOW ROUTINE ---------------------------#
 .align	16
-.Lcbc_slow_way:
+.Lcbc_slow_prologue:
 	# allocate aligned stack frame...
 	lea	-88(%rsp),%rbp
 	and	\$-64,%rbp
 	# ... just "above" key schedule
-	lea	-88-63(%rcx),%rax
-	sub	%rbp,%rax
-	neg	%rax
-	and	\$0x3c0,%rax
-	sub	%rax,%rbp
+	lea	-88-63(%rcx),%r10
+	sub	%rbp,%r10
+	neg	%r10
+	and	\$0x3c0,%r10
+	sub	%r10,%rbp
 
 	xchg	%rsp,%rbp
-	add	\$8,%rsp	# reserve for return address!
+	#add	\$8,%rsp	# reserve for return address!
 	mov	%rbp,$_rsp	# save %rsp
+.Lcbc_slow_body:
 	#mov	%rdi,$_inp	# save copy of inp
 	#mov	%rsi,$_out	# save copy of out
 	#mov	%rdx,$_len	# save copy of len
@@ -1963,6 +1987,7 @@ AES_cbc_encrypt:
 	mov	$s3,12(%rbp)
 
 	jmp	.Lcbc_exit
+
 .align	4
 .Lcbc_slow_enc_tail:
 	mov	%r10,%rcx
@@ -2053,6 +2078,21 @@ AES_cbc_encrypt:
 	lea	16(%r10),%rcx
 	.long	0x9066A4F3	# rep movsb
 	jmp	.Lcbc_exit
+
+.align	16
+.Lcbc_exit:
+	mov	$_rsp,%rsi
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lcbc_popfq:
+	popfq
+.Lcbc_epilogue:
+	ret
 .size	AES_cbc_encrypt,.-AES_cbc_encrypt
 ___
 }
@@ -2482,6 +2522,282 @@ $code.=<<___;
 .align	64
 ___
 
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern	__imp_RtlVirtualUnwind
+.type	block_se_handler,\@abi-omnipotent
+.align	16
+block_se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	mov	8($disp),%rsi		# disp->ImageBase
+	mov	56($disp),%r11		# disp->HandlerData
+
+	mov	0(%r11),%r10d		# HandlerData[0]
+	lea	(%rsi,%r10),%r10	# prologue label
+	cmp	%r10,%rbx		# context->Rip<prologue label
+	jb	.Lin_block_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	mov	4(%r11),%r10d		# HandlerData[1]
+	lea	(%rsi,%r10),%r10	# epilogue label
+	cmp	%r10,%rbx		# context->Rip>=epilogue label
+	jae	.Lin_block_prologue
+
+	mov	24(%rax),%rax		# pull saved real stack pointer
+	lea	48(%rax),%rax		# adjust...
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
+	mov	-40(%rax),%r14
+	mov	-48(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_block_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	jmp	.Lcommon_seh_exit
+.size	block_se_handler,.-block_se_handler
+
+.type	key_se_handler,\@abi-omnipotent
+.align	16
+key_se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	mov	8($disp),%rsi		# disp->ImageBase
+	mov	56($disp),%r11		# disp->HandlerData
+
+	mov	0(%r11),%r10d		# HandlerData[0]
+	lea	(%rsi,%r10),%r10	# prologue label
+	cmp	%r10,%rbx		# context->Rip<prologue label
+	jb	.Lin_key_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	mov	4(%r11),%r10d		# HandlerData[1]
+	lea	(%rsi,%r10),%r10	# epilogue label
+	cmp	%r10,%rbx		# context->Rip>=epilogue label
+	jae	.Lin_key_prologue
+
+	lea	56(%rax),%rax
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
+	mov	-40(%rax),%r14
+	mov	-48(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_key_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	jmp	.Lcommon_seh_exit
+.size	key_se_handler,.-key_se_handler
+
+.type	cbc_se_handler,\@abi-omnipotent
+.align	16
+cbc_se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	lea	.Lcbc_prologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lcbc_prologue
+	jb	.Lin_cbc_prologue
+
+	lea	.Lcbc_fast_body(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lcbc_fast_body
+	jb	.Lin_cbc_frame_setup
+
+	lea	.Lcbc_slow_prologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lcbc_slow_prologue
+	jb	.Lin_cbc_body
+
+	lea	.Lcbc_slow_body(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lcbc_slow_body
+	jb	.Lin_cbc_frame_setup
+
+.Lin_cbc_body:
+	mov	152($context),%rax	# pull context->Rsp
+
+	lea	.Lcbc_epilogue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lcbc_epilogue
+	jae	.Lin_cbc_prologue
+
+	lea	8(%rax),%rax
+
+	lea	.Lcbc_popfq(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lcbc_popfq
+	jae	.Lin_cbc_prologue
+
+	mov	`16-8`(%rax),%rax	# biased $_rsp
+	lea	56(%rax),%rax
+
+.Lin_cbc_frame_setup:
+	mov	-16(%rax),%rbx
+	mov	-24(%rax),%rbp
+	mov	-32(%rax),%r12
+	mov	-40(%rax),%r13
+	mov	-48(%rax),%r14
+	mov	-56(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_cbc_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+.Lcommon_seh_exit:
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$`1232/8`,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
+	pop	%r15
+	pop	%r14
+	pop	%r13
+	pop	%r12
+	pop	%rbp
+	pop	%rbx
+	pop	%rdi
+	pop	%rsi
+	ret
+.size	cbc_se_handler,.-cbc_se_handler
+
+.section	.pdata
+.align	4
+	.rva	.LSEH_begin_AES_encrypt
+	.rva	.LSEH_end_AES_encrypt
+	.rva	.LSEH_info_AES_encrypt
+
+	.rva	.LSEH_begin_AES_decrypt
+	.rva	.LSEH_end_AES_decrypt
+	.rva	.LSEH_info_AES_decrypt
+
+	.rva	.LSEH_begin_AES_set_encrypt_key
+	.rva	.LSEH_end_AES_set_encrypt_key
+	.rva	.LSEH_info_AES_set_encrypt_key
+
+	.rva	.LSEH_begin_AES_set_decrypt_key
+	.rva	.LSEH_end_AES_set_decrypt_key
+	.rva	.LSEH_info_AES_set_decrypt_key
+
+	.rva	.LSEH_begin_AES_cbc_encrypt
+	.rva	.LSEH_end_AES_cbc_encrypt
+	.rva	.LSEH_info_AES_cbc_encrypt
+
+.section	.xdata
+.align	8
+.LSEH_info_AES_encrypt:
+	.byte	9,0,0,0
+	.rva	block_se_handler
+	.rva	.Lenc_prologue,.Lenc_epilogue	# HandlerData[]
+.LSEH_info_AES_decrypt:
+	.byte	9,0,0,0
+	.rva	block_se_handler
+	.rva	.Ldec_prologue,.Ldec_epilogue	# HandlerData[]
+.LSEH_info_AES_set_encrypt_key:
+	.byte	9,0,0,0
+	.rva	key_se_handler
+	.rva	.Lenc_key_prologue,.Lenc_key_epilogue	# HandlerData[]
+.LSEH_info_AES_set_decrypt_key:
+	.byte	9,0,0,0
+	.rva	key_se_handler
+	.rva	.Ldec_key_prologue,.Ldec_key_epilogue	# HandlerData[]
+.LSEH_info_AES_cbc_encrypt:
+	.byte	9,0,0,0
+	.rva	cbc_se_handler
+___
+}
+
 $code =~ s/\`([^\`]*)\`/eval($1)/gem;
 
 print $code;
diff --git a/crypto/bn/asm/x86_64-mont.pl b/crypto/bn/asm/x86_64-mont.pl
index c43b69592a..3b7a6f243f 100755
--- a/crypto/bn/asm/x86_64-mont.pl
+++ b/crypto/bn/asm/x86_64-mont.pl
@@ -15,14 +15,18 @@
 # respectful 50%. It remains to be seen if loop unrolling and
 # dedicated squaring routine can provide further improvement...
 
-$output=shift;
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
 
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $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";
+open STDOUT,"| $^X $xlate $flavour $output";
 
 # int bn_mul_mont(
 $rp="%rdi";	# BN_ULONG *rp,
@@ -55,13 +59,14 @@ bn_mul_mont:
 	push	%r15
 
 	mov	${num}d,${num}d
-	lea	2($num),%rax
-	mov	%rsp,%rbp
-	neg	%rax
-	lea	(%rsp,%rax,8),%rsp	# tp=alloca(8*(num+2))
+	lea	2($num),%r10
+	mov	%rsp,%r11
+	neg	%r10
+	lea	(%rsp,%r10,8),%rsp	# tp=alloca(8*(num+2))
 	and	\$-1024,%rsp		# minimize TLB usage
 
-	mov	%rbp,8(%rsp,$num,8)	# tp[num+1]=%rsp
+	mov	%r11,8(%rsp,$num,8)	# tp[num+1]=%rsp
+.Lprologue:
 	mov	%rdx,$bp		# $bp reassigned, remember?
 
 	mov	($n0),$n0		# pull n0[0] value
@@ -197,18 +202,129 @@ bn_mul_mont:
 	dec	$j
 	jge	.Lcopy
 
-	mov	8(%rsp,$num,8),%rsp	# restore %rsp
+	mov	8(%rsp,$num,8),%rsi	# restore %rsp
 	mov	\$1,%rax
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lepilogue:
+	ret
+.size	bn_mul_mont,.-bn_mul_mont
+.asciz	"Montgomery Multiplication for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align	16
+___
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern	__imp_RtlVirtualUnwind
+.type	se_handler,\@abi-omnipotent
+.align	16
+se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	lea	.Lprologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lprologue
+	jb	.Lin_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	lea	.Lepilogue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lepilogue
+	jae	.Lin_prologue
+
+	mov	192($context),%r10	# pull $num
+	mov	8(%rax,%r10,8),%rax	# pull saved stack pointer
+	lea	48(%rax),%rax
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
+	mov	-40(%rax),%r14
+	mov	-48(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$154,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
 	pop	%r15
 	pop	%r14
 	pop	%r13
 	pop	%r12
 	pop	%rbp
 	pop	%rbx
+	pop	%rdi
+	pop	%rsi
 	ret
-.size	bn_mul_mont,.-bn_mul_mont
-.asciz	"Montgomery Multiplication for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.size	se_handler,.-se_handler
+
+.section	.pdata
+.align	4
+	.rva	.LSEH_begin_bn_mul_mont
+	.rva	.LSEH_end_bn_mul_mont
+	.rva	.LSEH_info_bn_mul_mont
+
+.section	.xdata
+.align	8
+.LSEH_info_bn_mul_mont:
+	.byte	9,0,0,0
+	.rva	se_handler
 ___
+}
 
 print $code;
 close STDOUT;
diff --git a/crypto/md5/asm/md5-x86_64.pl b/crypto/md5/asm/md5-x86_64.pl
index 493231401e..867885435e 100755
--- a/crypto/md5/asm/md5-x86_64.pl
+++ b/crypto/md5/asm/md5-x86_64.pl
@@ -108,14 +108,19 @@ sub round4_step
 EOF
 }
 
-my $output = shift;
+my $flavour = shift;
+my $output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+my $win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
 
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; my $dir=$1; my $xlate;
 ( $xlate="${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";
+no warnings qw(uninitialized);
+open STDOUT,"| $^X $xlate $flavour $output";
 
 $code .= <<EOF;
 .text
@@ -129,6 +134,7 @@ md5_block_asm_data_order:
 	push	%r12
 	push	%r14
 	push	%r15
+.Lprologue:
 
 	# rdi = arg #1 (ctx, MD5_CTX pointer)
 	# rsi = arg #2 (ptr, data pointer)
@@ -243,14 +249,120 @@ $code .= <<EOF;
 	mov	%ecx,		2*4(%rbp)	# ctx->C = C
 	mov	%edx,		3*4(%rbp)	# ctx->D = D
 
+	mov	(%rsp),%r15
+	mov	8(%rsp),%r14
+	mov	16(%rsp),%r12
+	mov	24(%rsp),%rbx
+	mov	32(%rsp),%rbp
+	add	\$40,%rsp
+.Lepilogue:
+	ret
+.size md5_block_asm_data_order,.-md5_block_asm_data_order
+EOF
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+my $rec="%rcx";
+my $frame="%rdx";
+my $context="%r8";
+my $disp="%r9";
+
+$code.=<<___;
+.extern	__imp_RtlVirtualUnwind
+.type	se_handler,\@abi-omnipotent
+.align	16
+se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	lea	.Lprologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lprologue
+	jb	.Lin_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	lea	.Lepilogue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lepilogue
+	jae	.Lin_prologue
+
+	lea	40(%rax),%rax
+
+	mov	-8(%rax),%rbp
+	mov	-16(%rax),%rbx
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r14
+	mov	-40(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$154,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
 	pop	%r15
 	pop	%r14
+	pop	%r13
 	pop	%r12
-	pop	%rbx
 	pop	%rbp
+	pop	%rbx
+	pop	%rdi
+	pop	%rsi
 	ret
-.size md5_block_asm_data_order,.-md5_block_asm_data_order
-EOF
+.size	se_handler,.-se_handler
+
+.section	.pdata
+.align	4
+	.rva	.LSEH_begin_md5_block_asm_data_order
+	.rva	.LSEH_end_md5_block_asm_data_order
+	.rva	.LSEH_info_md5_block_asm_data_order
+
+.section	.xdata
+.align	8
+.LSEH_info_md5_block_asm_data_order:
+	.byte	9,0,0,0
+	.rva	se_handler
+___
+}
 
 print $code;
 
diff --git a/crypto/rc4/asm/rc4-x86_64.pl b/crypto/rc4/asm/rc4-x86_64.pl
index 959a67a868..af2093643f 100755
--- a/crypto/rc4/asm/rc4-x86_64.pl
+++ b/crypto/rc4/asm/rc4-x86_64.pl
@@ -58,14 +58,18 @@
 # fit for Core2 and therefore the code was modified to skip cloop8 on
 # this CPU.
 
-$output=shift;
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
 
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $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";
+open STDOUT,"| $^X $xlate $flavour $output";
 
 $dat="%rdi";	    # arg1
 $len="%rsi";	    # arg2
@@ -89,6 +93,8 @@ RC4:	or	$len,$len
 .Lentry:
 	push	%r12
 	push	%r13
+	sub	\$8,%rsp
+.Lprologue:
 
 	add	\$8,$dat
 	movl	-8($dat),$XX[0]#d
@@ -133,16 +139,8 @@ $code.=<<___;
 	jnz	.Lloop8
 	cmp	\$0,$len
 	jne	.Lloop1
-___
-$code.=<<___;
-.Lexit:
-	sub	\$1,$XX[0]#b
-	movl	$XX[0]#d,-8($dat)
-	movl	$YY#d,-4($dat)
+	jmp	.Lexit
 
-	pop	%r13
-	pop	%r12
-	ret
 .align	16
 .Lloop1:
 	add	$TX[0]#b,$YY#b
@@ -249,6 +247,18 @@ $code.=<<___;
 	sub	\$1,$len
 	jnz	.Lcloop1
 	jmp	.Lexit
+
+.align	16
+.Lexit:
+	sub	\$1,$XX[0]#b
+	movl	$XX[0]#d,-8($dat)
+	movl	$YY#d,-4($dat)
+
+	mov	8(%rsp),%r13
+	mov	16(%rsp),%r12
+	add	\$24,%rsp
+.Lepilogue:
+	ret
 .size	RC4,.-RC4
 ___
 
@@ -333,7 +343,7 @@ RC4_set_key:
 .size	RC4_set_key,.-RC4_set_key
 
 .globl	RC4_options
-.type	RC4_options,\@function,0
+.type	RC4_options,\@abi-omnipotent
 .align	16
 RC4_options:
 	lea	.Lopts(%rip),%rax
@@ -356,6 +366,136 @@ RC4_options:
 .size	RC4_options,.-RC4_options
 ___
 
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern	__imp_RtlVirtualUnwind
+.type	stream_se_handler,\@abi-omnipotent
+.align	16
+stream_se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	lea	.Lprologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<prologue label
+	jb	.Lin_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	lea	.Lepilogue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<prologue label
+	jae	.Lin_prologue
+
+	lea	24(%rax),%rax
+
+	mov	-8(%rax),%r12
+	mov	-16(%rax),%r13
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+
+.Lin_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	jmp	.Lcommon_seh_exit
+.size	stream_se_handler,.-stream_se_handler
+
+.type	key_se_handler,\@abi-omnipotent
+.align	16
+key_se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	152($context),%rax	# pull context->Rsp
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+.Lcommon_seh_exit:
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$154,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
+	pop	%r15
+	pop	%r14
+	pop	%r13
+	pop	%r12
+	pop	%rbp
+	pop	%rbx
+	pop	%rdi
+	pop	%rsi
+	ret
+.size	key_se_handler,.-key_se_handler
+
+.section	.pdata
+.align	4
+	.rva	.LSEH_begin_RC4
+	.rva	.LSEH_end_RC4
+	.rva	.LSEH_info_RC4
+
+	.rva	.LSEH_begin_RC4_set_key
+	.rva	.LSEH_end_RC4_set_key
+	.rva	.LSEH_info_RC4_set_key
+
+.section	.xdata
+.align	8
+.LSEH_info_RC4:
+	.byte	9,0,0,0
+	.rva	stream_se_handler
+.LSEH_info_RC4_set_key:
+	.byte	9,0,0,0
+	.rva	key_se_handler
+___
+}
+
 $code =~ s/#([bwd])/$1/gm;
 
 print $code;
diff --git a/crypto/sha/asm/sha1-x86_64.pl b/crypto/sha/asm/sha1-x86_64.pl
index f7ed67a726..4edc5ea9ad 100755
--- a/crypto/sha/asm/sha1-x86_64.pl
+++ b/crypto/sha/asm/sha1-x86_64.pl
@@ -29,14 +29,18 @@
 # Xeon P4	+65%		+0%		9.9
 # Core2		+60%		+10%		7.0
 
-$output=shift;
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
 
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $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";
+open STDOUT,"| $^X $xlate $flavour $output";
 
 $ctx="%rdi";	# 1st arg
 $inp="%rsi";	# 2nd arg
@@ -69,13 +73,14 @@ $func:
 	push	%rbx
 	push	%rbp
 	push	%r12
-	mov	%rsp,%rax
+	mov	%rsp,%r11
 	mov	%rdi,$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	%r11,`16*4`(%rsp)
+.Lprologue:
 
 	mov	0($ctx),$A
 	mov	4($ctx),$B
@@ -88,10 +93,12 @@ ___
 sub EPILOGUE {
 my $func=shift;
 $code.=<<___;
-	mov	`16*4`(%rsp),%rsp
-	pop	%r12
-	pop	%rbp
-	pop	%rbx
+	mov	`16*4`(%rsp),%rsi
+	mov	(%rsi),%r12
+	mov	8(%rsi),%rbp
+	mov	16(%rsi),%rbx
+	lea	24(%rsi),%rsp
+.Lepilogue:
 	ret
 .size	$func,.-$func
 ___
@@ -233,7 +240,109 @@ ___
 &EPILOGUE("sha1_block_data_order");
 $code.=<<___;
 .asciz	"SHA1 block transform for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
+.align	16
+___
+
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern	__imp_RtlVirtualUnwind
+.type	se_handler,\@abi-omnipotent
+.align	16
+se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	lea	.Lprologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lprologue
+	jb	.Lin_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	lea	.Lepilogue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lepilogue
+	jae	.Lin_prologue
+
+	mov	`16*4`(%rax),%rax	# pull saved stack pointer
+	lea	24(%rax),%rax
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+
+.Lin_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$154,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
+	pop	%r15
+	pop	%r14
+	pop	%r13
+	pop	%r12
+	pop	%rbp
+	pop	%rbx
+	pop	%rdi
+	pop	%rsi
+	ret
+.size	se_handler,.-se_handler
+
+.section	.pdata
+.align	4
+	.rva	.LSEH_begin_sha1_block_data_order
+	.rva	.LSEH_end_sha1_block_data_order
+	.rva	.LSEH_info_sha1_block_data_order
+
+.section	.xdata
+.align	8
+.LSEH_info_sha1_block_data_order:
+	.byte	9,0,0,0
+	.rva	se_handler
 ___
+}
 
 ####################################################################
 
diff --git a/crypto/sha/asm/sha512-x86_64.pl b/crypto/sha/asm/sha512-x86_64.pl
index 10fd2abb65..e6643f8cf6 100755
--- a/crypto/sha/asm/sha512-x86_64.pl
+++ b/crypto/sha/asm/sha512-x86_64.pl
@@ -44,6 +44,8 @@ $flavour = shift;
 $output  = shift;
 if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
 
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
+
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
 ( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
 ( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
@@ -188,7 +190,7 @@ $func:
 	push	%r13
 	push	%r14
 	push	%r15
-	mov	%rsp,%rbp		# copy %rsp
+	mov	%rsp,%r11		# copy %rsp
 	shl	\$4,%rdx		# num*16
 	sub	\$$framesz,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -196,7 +198,8 @@ $func:
 	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
+	mov	%r11,$_rsp		# save copy of %rsp
+.Lprologue:
 
 	lea	$TABLE(%rip),$Tbl
 
@@ -258,14 +261,15 @@ $code.=<<___;
 	mov	$H,$SZ*7($ctx)
 	jb	.Lloop
 
-	mov	$_rsp,%rsp
-	pop	%r15
-	pop	%r14
-	pop	%r13
-	pop	%r12
-	pop	%rbp
-	pop	%rbx
-
+	mov	$_rsp,%rsi
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lepilogue:
 	ret
 .size	$func,.-$func
 ___
@@ -340,6 +344,113 @@ $TABLE:
 ___
 }
 
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern	__imp_RtlVirtualUnwind
+.type	se_handler,\@abi-omnipotent
+.align	16
+se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	lea	.Lprologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lprologue
+	jb	.Lin_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	lea	.Lepilogue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lepilogue
+	jae	.Lin_prologue
+
+	mov	16*$SZ+3*8(%rax),%rax	# pull $_rsp
+	lea	48(%rax),%rax
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
+	mov	-40(%rax),%r14
+	mov	-48(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$154,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
+	pop	%r15
+	pop	%r14
+	pop	%r13
+	pop	%r12
+	pop	%rbp
+	pop	%rbx
+	pop	%rdi
+	pop	%rsi
+	ret
+.size	se_handler,.-se_handler
+
+.section	.pdata
+.align	4
+	.rva	.LSEH_begin_$func
+	.rva	.LSEH_end_$func
+	.rva	.LSEH_info_$func
+
+.section	.xdata
+.align	8
+.LSEH_info_$func:
+	.byte	9,0,0,0
+	.rva	se_handler
+___
+}
+
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
 print $code;
 close STDOUT;
diff --git a/crypto/whrlpool/asm/wp-x86_64.pl b/crypto/whrlpool/asm/wp-x86_64.pl
index aaed353419..87c0843dc1 100644
--- a/crypto/whrlpool/asm/wp-x86_64.pl
+++ b/crypto/whrlpool/asm/wp-x86_64.pl
@@ -30,14 +30,18 @@
 # 3 on Opteron] and which is *unacceptably* slow with 64-bit
 # operand.
 
-$output=shift;
+$flavour = shift;
+$output  = shift;
+if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
+
+$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
 
 $0 =~ m/(.*[\/\\])[^\/\\]+$/; my $dir=$1; my $xlate;
 ( $xlate="${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";
+open STDOUT,"| $^X $xlate $flavour $output";
 
 sub L() { $code.=".byte	".join(',',@_)."\n"; }
 sub LL(){ $code.=".byte	".join(',',@_).",".join(',',@_)."\n"; }
@@ -61,16 +65,18 @@ $func:
 	push	%r14
 	push	%r15
 
-	mov	%rsp,%rax
+	mov	%rsp,%r11
 	sub	\$128+40,%rsp
 	and	\$-64,%rsp
 
-	lea	128(%rsp),%rbx
-	mov	%rdi,0(%rbx)		# save parameter block
-	mov	%rsi,8(%rbx)
-	mov	%rdx,16(%rbx)
-	mov	%rax,32(%rbx)		# saved stack pointer
+	lea	128(%rsp),%r10
+	mov	%rdi,0(%r10)		# save parameter block
+	mov	%rsi,8(%r10)
+	mov	%rdx,16(%r10)
+	mov	%r11,32(%r10)		# saved stack pointer
+.Lprologue:
 
+	mov	%r10,%rbx
 	lea	$table(%rip),%rbp
 
 	xor	%rcx,%rcx
@@ -187,13 +193,15 @@ $code.=<<___;
 	mov	%rax,16(%rbx)
 	jmp	.Louterloop
 .Lalldone:
-	mov	32(%rbx),%rsp		# restore saved pointer
-	pop	%r15
-	pop	%r14
-	pop	%r13
-	pop	%r12
-	pop	%rbp
-	pop	%rbx
+	mov	32(%rbx),%rsi		# restore saved pointer
+	mov	(%rsi),%r15
+	mov	8(%rsi),%r14
+	mov	16(%rsi),%r13
+	mov	24(%rsi),%r12
+	mov	32(%rsi),%rbp
+	mov	40(%rsi),%rbx
+	lea	48(%rsi),%rsp
+.Lepilogue:
 	ret
 .size	$func,.-$func
 
@@ -469,6 +477,113 @@ ___
 	&L(0xfb,0xee,0x7c,0x66,0xdd,0x17,0x47,0x9e);
 	&L(0xca,0x2d,0xbf,0x07,0xad,0x5a,0x83,0x33);
 
+# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
+#		CONTEXT *context,DISPATCHER_CONTEXT *disp)
+if ($win64) {
+$rec="%rcx";
+$frame="%rdx";
+$context="%r8";
+$disp="%r9";
+
+$code.=<<___;
+.extern	__imp_RtlVirtualUnwind
+.type	se_handler,\@abi-omnipotent
+.align	16
+se_handler:
+	push	%rsi
+	push	%rdi
+	push	%rbx
+	push	%rbp
+	push	%r12
+	push	%r13
+	push	%r14
+	push	%r15
+	pushfq
+	sub	\$64,%rsp
+
+	mov	120($context),%rax	# pull context->Rax
+	mov	248($context),%rbx	# pull context->Rip
+
+	lea	.Lprologue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip<.Lprologue
+	jb	.Lin_prologue
+
+	mov	152($context),%rax	# pull context->Rsp
+
+	lea	.Lepilogue(%rip),%r10
+	cmp	%r10,%rbx		# context->Rip>=.Lepilogue
+	jae	.Lin_prologue
+
+	mov	128+32(%rax),%rax	# pull saved stack pointer
+	lea	48(%rax),%rax
+
+	mov	-8(%rax),%rbx
+	mov	-16(%rax),%rbp
+	mov	-24(%rax),%r12
+	mov	-32(%rax),%r13
+	mov	-40(%rax),%r14
+	mov	-48(%rax),%r15
+	mov	%rbx,144($context)	# restore context->Rbx
+	mov	%rbp,160($context)	# restore context->Rbp
+	mov	%r12,216($context)	# restore context->R12
+	mov	%r13,224($context)	# restore context->R13
+	mov	%r14,232($context)	# restore context->R14
+	mov	%r15,240($context)	# restore context->R15
+
+.Lin_prologue:
+	mov	8(%rax),%rdi
+	mov	16(%rax),%rsi
+	mov	%rax,152($context)	# restore context->Rsp
+	mov	%rsi,168($context)	# restore context->Rsi
+	mov	%rdi,176($context)	# restore context->Rdi
+
+	mov	40($disp),%rdi		# disp->ContextRecord
+	mov	$context,%rsi		# context
+	mov	\$154,%ecx		# sizeof(CONTEXT)
+	.long	0xa548f3fc		# cld; rep movsq
+
+	mov	$disp,%rsi
+	xor	%rcx,%rcx		# arg1, UNW_FLAG_NHANDLER
+	mov	8(%rsi),%rdx		# arg2, disp->ImageBase
+	mov	0(%rsi),%r8		# arg3, disp->ControlPc
+	mov	16(%rsi),%r9		# arg4, disp->FunctionEntry
+	mov	40(%rsi),%r10		# disp->ContextRecord
+	lea	56(%rsi),%r11		# &disp->HandlerData
+	lea	24(%rsi),%r12		# &disp->EstablisherFrame
+	mov	%r10,32(%rsp)		# arg5
+	mov	%r11,40(%rsp)		# arg6
+	mov	%r12,48(%rsp)		# arg7
+	mov	%rcx,56(%rsp)		# arg8, (NULL)
+	call	*__imp_RtlVirtualUnwind(%rip)
+
+	mov	\$1,%eax		# ExceptionContinueSearch
+	add	\$64,%rsp
+	popfq
+	pop	%r15
+	pop	%r14
+	pop	%r13
+	pop	%r12
+	pop	%rbp
+	pop	%rbx
+	pop	%rdi
+	pop	%rsi
+	ret
+.size	se_handler,.-se_handler
+
+.section	.pdata
+.align	4
+	.rva	.LSEH_begin_$func
+	.rva	.LSEH_end_$func
+	.rva	.LSEH_info_$func
+
+.section	.xdata
+.align	8
+.LSEH_info_$func:
+	.byte	9,0,0,0
+	.rva	se_handler
+___
+}
+
 $code =~ s/\`([^\`]*)\`/eval $1/gem;
 print $code;
 close STDOUT;
diff --git a/ms/uplink-x86_64.pl b/ms/uplink-x86_64.pl
index 80807ce05b..9acbf6be6f 100755
--- a/ms/uplink-x86_64.pl
+++ b/ms/uplink-x86_64.pl
@@ -12,6 +12,7 @@ $prefix="_lazy";
 print <<___;
 .text
 .extern	OPENSSL_Uplink
+.globl	OPENSSL_UplinkTable
 ___
 for ($i=1;$i<=$N;$i++) {
 print <<___;
@@ -30,8 +31,8 @@ $prefix${i}:
 	mov	56(%rsp),%rdx
 	mov	64(%rsp),%r8
 	mov	72(%rsp),%r9
-	add	\$40,%rsp
 	lea	OPENSSL_UplinkTable(%rip),%rax
+	add	\$40,%rsp
 	jmp	*8*$i(%rax)
 $prefix${i}_end:
 .size	$prefix${i},.-$prefix${i}
@@ -39,24 +40,25 @@ ___
 }
 print <<___;
 .data
-.globl  OPENSSL_UplinkTable
 OPENSSL_UplinkTable:
         .quad   $N
 ___
 for ($i=1;$i<=$N;$i++) {   print "      .quad   $prefix$i\n";   }
 print <<___;
-.section	.pdata
+.section	.pdata,"r"
+.align		4
 ___
 for ($i=1;$i<=$N;$i++) {
 print <<___;
-	.long	$prefix${i}
-	.long	$prefix${i}_end
-	.long	${prefix}_unwind_info
+	.rva	$prefix${i},$prefix${i}_end,${prefix}_unwind_info
 ___
 }
 print <<___;
-.section	.xdata
+.section	.xdata,"r"
+.align		8
 ${prefix}_unwind_info:
 	.byte	0x01,0x04,0x01,0x00
 	.byte	0x04,0x42,0x00,0x00
 ___
+
+close STDOUT;