x86_64 assembly pack: Win64 SEH face-lift.

- harmonize handlers with guidelines and themselves;
- fix some bugs in handlers;
- add missing handlers in chacha and ecp_nistz256 modules;

Reviewed-by: Rich Salz <rsalz@openssl.org>

2 files changed, 105 insertions, 113 deletions

diff --git a/crypto/sha/asm/sha1-x86_64.pl b/crypto/sha/asm/sha1-x86_64.pl
index f06fa515a2..c0cc7b3898 100755
--- a/crypto/sha/asm/sha1-x86_64.pl
+++ b/crypto/sha/asm/sha1-x86_64.pl
@@ -462,7 +462,8 @@ my @V=($A,$B,$C,$D,$E)=("%eax","%ebx","%ecx","%edx","%ebp");	# size optimization
 my @T=("%esi","%edi");
 my $j=0;
 my $rx=0;
-my $K_XX_XX="%r11";
+my $K_XX_XX="%r14";
+my $fp="%r11";
 
 my $_rol=sub { &rol(@_) };
 my $_ror=sub { &ror(@_) };
@@ -483,7 +484,7 @@ $code.=<<___;
 .align	16
 sha1_block_data_order_ssse3:
 _ssse3_shortcut:
-	mov	%rsp,%rax
+	mov	%rsp,$fp	# frame pointer
 	push	%rbx
 	push	%rbp
 	push	%r12
@@ -492,16 +493,15 @@ _ssse3_shortcut:
 	lea	`-64-($win64?6*16:0)`(%rsp),%rsp
 ___
 $code.=<<___ if ($win64);
-	movaps	%xmm6,-40-6*16(%rax)
-	movaps	%xmm7,-40-5*16(%rax)
-	movaps	%xmm8,-40-4*16(%rax)
-	movaps	%xmm9,-40-3*16(%rax)
-	movaps	%xmm10,-40-2*16(%rax)
-	movaps	%xmm11,-40-1*16(%rax)
+	movaps	%xmm6,-40-6*16($fp)
+	movaps	%xmm7,-40-5*16($fp)
+	movaps	%xmm8,-40-4*16($fp)
+	movaps	%xmm9,-40-3*16($fp)
+	movaps	%xmm10,-40-2*16($fp)
+	movaps	%xmm11,-40-1*16($fp)
 .Lprologue_ssse3:
 ___
 $code.=<<___;
-	mov	%rax,%r14	# original %rsp
 	and	\$-64,%rsp
 	mov	%rdi,$ctx	# reassigned argument
 	mov	%rsi,$inp	# reassigned argument
@@ -908,21 +908,20 @@ $code.=<<___;
 	mov	$E,16($ctx)
 ___
 $code.=<<___ if ($win64);
-	movaps	-40-6*16(%r14),%xmm6
-	movaps	-40-5*16(%r14),%xmm7
-	movaps	-40-4*16(%r14),%xmm8
-	movaps	-40-3*16(%r14),%xmm9
-	movaps	-40-2*16(%r14),%xmm10
-	movaps	-40-1*16(%r14),%xmm11
+	movaps	-40-6*16($fp),%xmm6
+	movaps	-40-5*16($fp),%xmm7
+	movaps	-40-4*16($fp),%xmm8
+	movaps	-40-3*16($fp),%xmm9
+	movaps	-40-2*16($fp),%xmm10
+	movaps	-40-1*16($fp),%xmm11
 ___
 $code.=<<___;
-	lea	(%r14),%rsi
-	mov	-40(%rsi),%r14
-	mov	-32(%rsi),%r13
-	mov	-24(%rsi),%r12
-	mov	-16(%rsi),%rbp
-	mov	-8(%rsi),%rbx
-	lea	(%rsi),%rsp
+	mov	-40($fp),%r14
+	mov	-32($fp),%r13
+	mov	-24($fp),%r12
+	mov	-16($fp),%rbp
+	mov	-8($fp),%rbx
+	lea	($fp),%rsp
 .Lepilogue_ssse3:
 	ret
 .size	sha1_block_data_order_ssse3,.-sha1_block_data_order_ssse3
@@ -945,7 +944,7 @@ $code.=<<___;
 .align	16
 sha1_block_data_order_avx:
 _avx_shortcut:
-	mov	%rsp,%rax
+	mov	%rsp,$fp
 	push	%rbx
 	push	%rbp
 	push	%r12
@@ -955,16 +954,15 @@ _avx_shortcut:
 	vzeroupper
 ___
 $code.=<<___ if ($win64);
-	vmovaps	%xmm6,-40-6*16(%rax)
-	vmovaps	%xmm7,-40-5*16(%rax)
-	vmovaps	%xmm8,-40-4*16(%rax)
-	vmovaps	%xmm9,-40-3*16(%rax)
-	vmovaps	%xmm10,-40-2*16(%rax)
-	vmovaps	%xmm11,-40-1*16(%rax)
+	vmovaps	%xmm6,-40-6*16($fp)
+	vmovaps	%xmm7,-40-5*16($fp)
+	vmovaps	%xmm8,-40-4*16($fp)
+	vmovaps	%xmm9,-40-3*16($fp)
+	vmovaps	%xmm10,-40-2*16($fp)
+	vmovaps	%xmm11,-40-1*16($fp)
 .Lprologue_avx:
 ___
 $code.=<<___;
-	mov	%rax,%r14	# original %rsp
 	and	\$-64,%rsp
 	mov	%rdi,$ctx	# reassigned argument
 	mov	%rsi,$inp	# reassigned argument
@@ -1272,21 +1270,20 @@ $code.=<<___;
 	mov	$E,16($ctx)
 ___
 $code.=<<___ if ($win64);
-	movaps	-40-6*16(%r14),%xmm6
-	movaps	-40-5*16(%r14),%xmm7
-	movaps	-40-4*16(%r14),%xmm8
-	movaps	-40-3*16(%r14),%xmm9
-	movaps	-40-2*16(%r14),%xmm10
-	movaps	-40-1*16(%r14),%xmm11
+	movaps	-40-6*16($fp),%xmm6
+	movaps	-40-5*16($fp),%xmm7
+	movaps	-40-4*16($fp),%xmm8
+	movaps	-40-3*16($fp),%xmm9
+	movaps	-40-2*16($fp),%xmm10
+	movaps	-40-1*16($fp),%xmm11
 ___
 $code.=<<___;
-	lea	(%r14),%rsi
-	mov	-40(%rsi),%r14
-	mov	-32(%rsi),%r13
-	mov	-24(%rsi),%r12
-	mov	-16(%rsi),%rbp
-	mov	-8(%rsi),%rbx
-	lea	(%rsi),%rsp
+	mov	-40($fp),%r14
+	mov	-32($fp),%r13
+	mov	-24($fp),%r12
+	mov	-16($fp),%rbp
+	mov	-8($fp),%rbx
+	lea	($fp),%rsp
 .Lepilogue_avx:
 	ret
 .size	sha1_block_data_order_avx,.-sha1_block_data_order_avx
@@ -1312,7 +1309,7 @@ $code.=<<___;
 .align	16
 sha1_block_data_order_avx2:
 _avx2_shortcut:
-	mov	%rsp,%rax
+	mov	%rsp,$fp
 	push	%rbx
 	push	%rbp
 	push	%r12
@@ -1322,16 +1319,15 @@ _avx2_shortcut:
 ___
 $code.=<<___ if ($win64);
 	lea	-6*16(%rsp),%rsp
-	vmovaps	%xmm6,-40-6*16(%rax)
-	vmovaps	%xmm7,-40-5*16(%rax)
-	vmovaps	%xmm8,-40-4*16(%rax)
-	vmovaps	%xmm9,-40-3*16(%rax)
-	vmovaps	%xmm10,-40-2*16(%rax)
-	vmovaps	%xmm11,-40-1*16(%rax)
+	vmovaps	%xmm6,-40-6*16($fp)
+	vmovaps	%xmm7,-40-5*16($fp)
+	vmovaps	%xmm8,-40-4*16($fp)
+	vmovaps	%xmm9,-40-3*16($fp)
+	vmovaps	%xmm10,-40-2*16($fp)
+	vmovaps	%xmm11,-40-1*16($fp)
 .Lprologue_avx2:
 ___
 $code.=<<___;
-	mov	%rax,%r14		# original %rsp
 	mov	%rdi,$ctx		# reassigned argument
 	mov	%rsi,$inp		# reassigned argument
 	mov	%rdx,$num		# reassigned argument
@@ -1751,21 +1747,20 @@ $code.=<<___;
 	vzeroupper
 ___
 $code.=<<___ if ($win64);
-	movaps	-40-6*16(%r14),%xmm6
-	movaps	-40-5*16(%r14),%xmm7
-	movaps	-40-4*16(%r14),%xmm8
-	movaps	-40-3*16(%r14),%xmm9
-	movaps	-40-2*16(%r14),%xmm10
-	movaps	-40-1*16(%r14),%xmm11
+	movaps	-40-6*16($fp),%xmm6
+	movaps	-40-5*16($fp),%xmm7
+	movaps	-40-4*16($fp),%xmm8
+	movaps	-40-3*16($fp),%xmm9
+	movaps	-40-2*16($fp),%xmm10
+	movaps	-40-1*16($fp),%xmm11
 ___
 $code.=<<___;
-	lea	(%r14),%rsi
-	mov	-40(%rsi),%r14
-	mov	-32(%rsi),%r13
-	mov	-24(%rsi),%r12
-	mov	-16(%rsi),%rbp
-	mov	-8(%rsi),%rbx
-	lea	(%rsi),%rsp
+	mov	-40($fp),%r14
+	mov	-32($fp),%r13
+	mov	-24($fp),%r12
+	mov	-16($fp),%rbp
+	mov	-8($fp),%rbx
+	lea	($fp),%rsp
 .Lepilogue_avx2:
 	ret
 .size	sha1_block_data_order_avx2,.-sha1_block_data_order_avx2
@@ -1908,15 +1903,13 @@ ssse3_handler:
 	cmp	%r10,%rbx		# context->Rip<prologue label
 	jb	.Lcommon_seh_tail
 
-	mov	152($context),%rax	# pull context->Rsp
+	mov	208($context),%rax	# pull context->R11
 
 	mov	4(%r11),%r10d		# HandlerData[1]
 	lea	(%rsi,%r10),%r10	# epilogue label
 	cmp	%r10,%rbx		# context->Rip>=epilogue label
 	jae	.Lcommon_seh_tail
 
-	mov	232($context),%rax	# pull context->R14
-
 	lea	-40-6*16(%rax),%rsi
 	lea	512($context),%rdi	# &context.Xmm6
 	mov	\$12,%ecx
diff --git a/crypto/sha/asm/sha512-x86_64.pl b/crypto/sha/asm/sha512-x86_64.pl
index 5a1cbcf0ca..7a8ed7c7f2 100755
--- a/crypto/sha/asm/sha512-x86_64.pl
+++ b/crypto/sha/asm/sha512-x86_64.pl
@@ -301,13 +301,13 @@ $code.=<<___ if ($SZ==4);
 	jnz	.Lssse3_shortcut
 ___
 $code.=<<___;
+	mov	%rsp,%rax		# copy %rsp
 	push	%rbx
 	push	%rbp
 	push	%r12
 	push	%r13
 	push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
 	shl	\$4,%rdx		# num*16
 	sub	\$$framesz,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -315,7 +315,7 @@ $code.=<<___;
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
 .Lprologue:
 
 	mov	$SZ*0($ctx),$A
@@ -382,13 +382,13 @@ $code.=<<___;
 	jb	.Lloop
 
 	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
+	mov	-48(%rsi),%r15
+	mov	-40(%rsi),%r14
+	mov	-32(%rsi),%r13
+	mov	-24(%rsi),%r12
+	mov	-16(%rsi),%rbp
+	mov	-8(%rsi),%rbx
+	lea	(%rsi),%rsp
 .Lepilogue:
 	ret
 .size	$func,.-$func
@@ -761,13 +761,13 @@ $code.=<<___;
 .align	64
 ${func}_ssse3:
 .Lssse3_shortcut:
+	mov	%rsp,%rax		# copy %rsp
 	push	%rbx
 	push	%rbp
 	push	%r12
 	push	%r13
 	push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
 	shl	\$4,%rdx		# num*16
 	sub	\$`$framesz+$win64*16*4`,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -775,7 +775,7 @@ ${func}_ssse3:
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
 ___
 $code.=<<___ if ($win64);
 	movaps	%xmm6,16*$SZ+32(%rsp)
@@ -1082,13 +1082,13 @@ $code.=<<___ if ($win64);
 	movaps	16*$SZ+80(%rsp),%xmm9
 ___
 $code.=<<___;
-	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
+	mov	-48(%rsi),%r15
+	mov	-40(%rsi),%r14
+	mov	-32(%rsi),%r13
+	mov	-24(%rsi),%r12
+	mov	-16(%rsi),%rbp
+	mov	-8(%rsi),%rbx
+	lea	(%rsi),%rsp
 .Lepilogue_ssse3:
 	ret
 .size	${func}_ssse3,.-${func}_ssse3
@@ -1105,13 +1105,13 @@ $code.=<<___;
 .align	64
 ${func}_xop:
 .Lxop_shortcut:
+	mov	%rsp,%rax		# copy %rsp
 	push	%rbx
 	push	%rbp
 	push	%r12
 	push	%r13
 	push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
 	shl	\$4,%rdx		# num*16
 	sub	\$`$framesz+$win64*16*($SZ==4?4:6)`,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -1119,7 +1119,7 @@ ${func}_xop:
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
 ___
 $code.=<<___ if ($win64);
 	movaps	%xmm6,16*$SZ+32(%rsp)
@@ -1459,13 +1459,13 @@ $code.=<<___ if ($win64 && $SZ>4);
 	movaps	16*$SZ+112(%rsp),%xmm11
 ___
 $code.=<<___;
-	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
+	mov	-48(%rsi),%r15
+	mov	-40(%rsi),%r14
+	mov	-32(%rsi),%r13
+	mov	-24(%rsi),%r12
+	mov	-16(%rsi),%rbp
+	mov	-8(%rsi),%rbx
+	lea	(%rsi),%rsp
 .Lepilogue_xop:
 	ret
 .size	${func}_xop,.-${func}_xop
@@ -1481,13 +1481,13 @@ $code.=<<___;
 .align	64
 ${func}_avx:
 .Lavx_shortcut:
+	mov	%rsp,%rax		# copy %rsp
 	push	%rbx
 	push	%rbp
 	push	%r12
 	push	%r13
 	push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
 	shl	\$4,%rdx		# num*16
 	sub	\$`$framesz+$win64*16*($SZ==4?4:6)`,%rsp
 	lea	($inp,%rdx,$SZ),%rdx	# inp+num*16*$SZ
@@ -1495,7 +1495,7 @@ ${func}_avx:
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
 ___
 $code.=<<___ if ($win64);
 	movaps	%xmm6,16*$SZ+32(%rsp)
@@ -1767,13 +1767,13 @@ $code.=<<___ if ($win64 && $SZ>4);
 	movaps	16*$SZ+112(%rsp),%xmm11
 ___
 $code.=<<___;
-	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
+	mov	-48(%rsi),%r15
+	mov	-40(%rsi),%r14
+	mov	-32(%rsi),%r13
+	mov	-24(%rsi),%r12
+	mov	-16(%rsi),%rbp
+	mov	-8(%rsi),%rbx
+	lea	(%rsi),%rsp
 .Lepilogue_avx:
 	ret
 .size	${func}_avx,.-${func}_avx
@@ -1832,13 +1832,13 @@ $code.=<<___;
 .align	64
 ${func}_avx2:
 .Lavx2_shortcut:
+	mov	%rsp,%rax		# copy %rsp
 	push	%rbx
 	push	%rbp
 	push	%r12
 	push	%r13
 	push	%r14
 	push	%r15
-	mov	%rsp,%r11		# copy %rsp
 	sub	\$`2*$SZ*$rounds+4*8+$win64*16*($SZ==4?4:6)`,%rsp
 	shl	\$4,%rdx		# num*16
 	and	\$-256*$SZ,%rsp		# align stack frame
@@ -1847,7 +1847,7 @@ ${func}_avx2:
 	mov	$ctx,$_ctx		# save ctx, 1st arg
 	mov	$inp,$_inp		# save inp, 2nd arh
 	mov	%rdx,$_end		# save end pointer, "3rd" arg
-	mov	%r11,$_rsp		# save copy of %rsp
+	mov	%rax,$_rsp		# save copy of %rsp
 ___
 $code.=<<___ if ($win64);
 	movaps	%xmm6,16*$SZ+32(%rsp)
@@ -2141,13 +2141,13 @@ $code.=<<___ if ($win64 && $SZ>4);
 	movaps	16*$SZ+112(%rsp),%xmm11
 ___
 $code.=<<___;
-	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
+	mov	-48(%rsi),%r15
+	mov	-40(%rsi),%r14
+	mov	-32(%rsi),%r13
+	mov	-24(%rsi),%r12
+	mov	-16(%rsi),%rbp
+	mov	-8(%rsi),%rbx
+	lea	(%rsi),%rsp
 .Lepilogue_avx2:
 	ret
 .size	${func}_avx2,.-${func}_avx2
@@ -2209,7 +2209,6 @@ ___
 $code.=<<___;
 	mov	%rax,%rsi		# put aside Rsp
 	mov	16*$SZ+3*8(%rax),%rax	# pull $_rsp
-	lea	48(%rax),%rax
 
 	mov	-8(%rax),%rbx
 	mov	-16(%rax),%rbp