[thirdparty/openssl.git] / crypto / x86_64cpuid.pl

#! /usr/bin/env perl
# Copyright 2005-2018 The OpenSSL Project Authors. All Rights Reserved.
#
# Licensed under the Apache License 2.0 (the "License").  You may not use
# this file except in compliance with the License.  You can obtain a copy
# in the file LICENSE in the source distribution or at
# https://www.openssl.org/source/license.html


# $output is the last argument if it looks like a file (it has an extension)
# $flavour is the first argument if it doesn't look like a file
$output = $#ARGV >= 0 && $ARGV[$#ARGV] =~ m|\.\w+$| ? pop : undef;
$flavour = $#ARGV >= 0 && $ARGV[0] !~ m|\.| ? shift : undef;

$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 OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""
     or die "can't call $xlate: $!";
*STDOUT=*OUT;

($arg1,$arg2,$arg3,$arg4)=$win64?("%rcx","%rdx","%r8", "%r9") :	# Win64 order
				 ("%rdi","%rsi","%rdx","%rcx");	# Unix order

print<<___;
.extern		OPENSSL_cpuid_setup
.hidden		OPENSSL_cpuid_setup
.section	.init
	call	OPENSSL_cpuid_setup

.hidden	OPENSSL_ia32cap_P
.comm	OPENSSL_ia32cap_P,16,4

.text

.globl	OPENSSL_atomic_add
.type	OPENSSL_atomic_add,\@abi-omnipotent
.align	16
OPENSSL_atomic_add:
	movl	($arg1),%eax
.Lspin:	leaq	($arg2,%rax),%r8
	.byte	0xf0		# lock
	cmpxchgl	%r8d,($arg1)
	jne	.Lspin
	movl	%r8d,%eax
	.byte	0x48,0x98	# cltq/cdqe
	ret
.size	OPENSSL_atomic_add,.-OPENSSL_atomic_add

.globl	OPENSSL_rdtsc
.type	OPENSSL_rdtsc,\@abi-omnipotent
.align	16
OPENSSL_rdtsc:
	rdtsc
	shl	\$32,%rdx
	or	%rdx,%rax
	ret
.size	OPENSSL_rdtsc,.-OPENSSL_rdtsc

.globl	OPENSSL_ia32_cpuid
.type	OPENSSL_ia32_cpuid,\@function,1
.align	16
OPENSSL_ia32_cpuid:
.cfi_startproc
	mov	%rbx,%r8		# save %rbx
.cfi_register	%rbx,%r8

	xor	%eax,%eax
	mov	%rax,8(%rdi)		# clear extended feature flags
	cpuid
	mov	%eax,%r11d		# max value for standard query level

	xor	%eax,%eax
	cmp	\$0x756e6547,%ebx	# "Genu"
	setne	%al
	mov	%eax,%r9d
	cmp	\$0x49656e69,%edx	# "ineI"
	setne	%al
	or	%eax,%r9d
	cmp	\$0x6c65746e,%ecx	# "ntel"
	setne	%al
	or	%eax,%r9d		# 0 indicates Intel CPU
	jz	.Lintel

	cmp	\$0x68747541,%ebx	# "Auth"
	setne	%al
	mov	%eax,%r10d
	cmp	\$0x69746E65,%edx	# "enti"
	setne	%al
	or	%eax,%r10d
	cmp	\$0x444D4163,%ecx	# "cAMD"
	setne	%al
	or	%eax,%r10d		# 0 indicates AMD CPU
	jnz	.Lintel

	# AMD specific
	mov	\$0x80000000,%eax
	cpuid
	cmp	\$0x80000001,%eax
	jb	.Lintel
	mov	%eax,%r10d
	mov	\$0x80000001,%eax
	cpuid
	or	%ecx,%r9d
	and	\$0x00000801,%r9d	# isolate AMD XOP bit, 1<<11

	cmp	\$0x80000008,%r10d
	jb	.Lintel

	mov	\$0x80000008,%eax
	cpuid
	movzb	%cl,%r10		# number of cores - 1
	inc	%r10			# number of cores

	mov	\$1,%eax
	cpuid
	bt	\$28,%edx		# test hyper-threading bit
	jnc	.Lgeneric
	shr	\$16,%ebx		# number of logical processors
	cmp	%r10b,%bl
	ja	.Lgeneric
	and	\$0xefffffff,%edx	# ~(1<<28)
	jmp	.Lgeneric

.Lintel:
	cmp	\$4,%r11d
	mov	\$-1,%r10d
	jb	.Lnocacheinfo

	mov	\$4,%eax
	mov	\$0,%ecx		# query L1D
	cpuid
	mov	%eax,%r10d
	shr	\$14,%r10d
	and	\$0xfff,%r10d		# number of cores -1 per L1D

.Lnocacheinfo:
	mov	\$1,%eax
	cpuid
	movd	%eax,%xmm0		# put aside processor id
	and	\$0xbfefffff,%edx	# force reserved bits to 0
	cmp	\$0,%r9d
	jne	.Lnotintel
	or	\$0x40000000,%edx	# set reserved bit#30 on Intel CPUs
	and	\$15,%ah
	cmp	\$15,%ah		# examine Family ID
	jne	.LnotP4
	or	\$0x00100000,%edx	# set reserved bit#20 to engage RC4_CHAR
.LnotP4:
	cmp	\$6,%ah
	jne	.Lnotintel
	and	\$0x0fff0ff0,%eax
	cmp	\$0x00050670,%eax	# Knights Landing
	je	.Lknights
	cmp	\$0x00080650,%eax	# Knights Mill (according to sde)
	jne	.Lnotintel
.Lknights:
	and	\$0xfbffffff,%ecx	# clear XSAVE flag to mimic Silvermont

.Lnotintel:
	bt	\$28,%edx		# test hyper-threading bit
	jnc	.Lgeneric
	and	\$0xefffffff,%edx	# ~(1<<28)
	cmp	\$0,%r10d
	je	.Lgeneric

	or	\$0x10000000,%edx	# 1<<28
	shr	\$16,%ebx
	cmp	\$1,%bl			# see if cache is shared
	ja	.Lgeneric
	and	\$0xefffffff,%edx	# ~(1<<28)
.Lgeneric:
	and	\$0x00000800,%r9d	# isolate AMD XOP flag
	and	\$0xfffff7ff,%ecx
	or	%ecx,%r9d		# merge AMD XOP flag

	mov	%edx,%r10d		# %r9d:%r10d is copy of %ecx:%edx

	cmp	\$7,%r11d
	jb	.Lno_extended_info
	mov	\$7,%eax
	xor	%ecx,%ecx
	cpuid
	bt	\$26,%r9d		# check XSAVE bit, cleared on Knights
	jc	.Lnotknights
	and	\$0xfff7ffff,%ebx	# clear ADCX/ADOX flag
.Lnotknights:
	movd	%xmm0,%eax		# restore processor id
	and	\$0x0fff0ff0,%eax
	cmp	\$0x00050650,%eax	# Skylake-X
	jne	.Lnotskylakex
	and	\$0xfffeffff,%ebx	# ~(1<<16)
					# suppress AVX512F flag on Skylake-X
.Lnotskylakex:
	mov	%ebx,8(%rdi)		# save extended feature flags
	mov	%ecx,12(%rdi)
.Lno_extended_info:

	bt	\$27,%r9d		# check OSXSAVE bit
	jnc	.Lclear_avx
	xor	%ecx,%ecx		# XCR0
	.byte	0x0f,0x01,0xd0		# xgetbv
	and	\$0xe6,%eax		# isolate XMM, YMM and ZMM state support
	cmp	\$0xe6,%eax
	je	.Ldone
	andl	\$0x3fdeffff,8(%rdi)	# ~(1<<31|1<<30|1<<21|1<<16)
					# clear AVX512F+BW+VL+FIMA, all of
					# them are EVEX-encoded, which requires
					# ZMM state support even if one uses
					# only XMM and YMM :-(
	and	\$6,%eax		# isolate XMM and YMM state support
	cmp	\$6,%eax
	je	.Ldone
.Lclear_avx:
	mov	\$0xefffe7ff,%eax	# ~(1<<28|1<<12|1<<11)
	and	%eax,%r9d		# clear AVX, FMA and AMD XOP bits
	mov	\$0x3fdeffdf,%eax	# ~(1<<31|1<<30|1<<21|1<<16|1<<5)
	and	%eax,8(%rdi)		# clear AVX2 and AVX512* bits
.Ldone:
	shl	\$32,%r9
	mov	%r10d,%eax
	mov	%r8,%rbx		# restore %rbx
.cfi_restore	%rbx
	or	%r9,%rax
	ret
.cfi_endproc
.size	OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid

.globl  OPENSSL_cleanse
.type   OPENSSL_cleanse,\@abi-omnipotent
.align  16
OPENSSL_cleanse:
	xor	%rax,%rax
	cmp	\$15,$arg2
	jae	.Lot
	cmp	\$0,$arg2
	je	.Lret
.Little:
	mov	%al,($arg1)
	sub	\$1,$arg2
	lea	1($arg1),$arg1
	jnz	.Little
.Lret:
	ret
.align	16
.Lot:
	test	\$7,$arg1
	jz	.Laligned
	mov	%al,($arg1)
	lea	-1($arg2),$arg2
	lea	1($arg1),$arg1
	jmp	.Lot
.Laligned:
	mov	%rax,($arg1)
	lea	-8($arg2),$arg2
	test	\$-8,$arg2
	lea	8($arg1),$arg1
	jnz	.Laligned
	cmp	\$0,$arg2
	jne	.Little
	ret
.size	OPENSSL_cleanse,.-OPENSSL_cleanse

.globl  CRYPTO_memcmp
.type   CRYPTO_memcmp,\@abi-omnipotent
.align  16
CRYPTO_memcmp:
	xor	%rax,%rax
	xor	%r10,%r10
	cmp	\$0,$arg3
	je	.Lno_data
	cmp	\$16,$arg3
	jne	.Loop_cmp
	mov	($arg1),%r10
	mov	8($arg1),%r11
	mov	\$1,$arg3
	xor	($arg2),%r10
	xor	8($arg2),%r11
	or	%r11,%r10
	cmovnz	$arg3,%rax
	ret

.align	16
.Loop_cmp:
	mov	($arg1),%r10b
	lea	1($arg1),$arg1
	xor	($arg2),%r10b
	lea	1($arg2),$arg2
	or	%r10b,%al
	dec	$arg3
	jnz	.Loop_cmp
	neg	%rax
	shr	\$63,%rax
.Lno_data:
	ret
.size	CRYPTO_memcmp,.-CRYPTO_memcmp
___

print<<___ if (!$win64);
.globl	OPENSSL_wipe_cpu
.type	OPENSSL_wipe_cpu,\@abi-omnipotent
.align	16
OPENSSL_wipe_cpu:
	pxor	%xmm0,%xmm0
	pxor	%xmm1,%xmm1
	pxor	%xmm2,%xmm2
	pxor	%xmm3,%xmm3
	pxor	%xmm4,%xmm4
	pxor	%xmm5,%xmm5
	pxor	%xmm6,%xmm6
	pxor	%xmm7,%xmm7
	pxor	%xmm8,%xmm8
	pxor	%xmm9,%xmm9
	pxor	%xmm10,%xmm10
	pxor	%xmm11,%xmm11
	pxor	%xmm12,%xmm12
	pxor	%xmm13,%xmm13
	pxor	%xmm14,%xmm14
	pxor	%xmm15,%xmm15
	xorq	%rcx,%rcx
	xorq	%rdx,%rdx
	xorq	%rsi,%rsi
	xorq	%rdi,%rdi
	xorq	%r8,%r8
	xorq	%r9,%r9
	xorq	%r10,%r10
	xorq	%r11,%r11
	leaq	8(%rsp),%rax
	ret
.size	OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
___
print<<___ if ($win64);
.globl	OPENSSL_wipe_cpu
.type	OPENSSL_wipe_cpu,\@abi-omnipotent
.align	16
OPENSSL_wipe_cpu:
	pxor	%xmm0,%xmm0
	pxor	%xmm1,%xmm1
	pxor	%xmm2,%xmm2
	pxor	%xmm3,%xmm3
	pxor	%xmm4,%xmm4
	pxor	%xmm5,%xmm5
	xorq	%rcx,%rcx
	xorq	%rdx,%rdx
	xorq	%r8,%r8
	xorq	%r9,%r9
	xorq	%r10,%r10
	xorq	%r11,%r11
	leaq	8(%rsp),%rax
	ret
.size	OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
___
{
my $out="%r10";
my $cnt="%rcx";
my $max="%r11";
my $lasttick="%r8d";
my $lastdiff="%r9d";
my $redzone=win64?8:-8;

print<<___;
.globl	OPENSSL_instrument_bus
.type	OPENSSL_instrument_bus,\@abi-omnipotent
.align	16
OPENSSL_instrument_bus:
	mov	$arg1,$out	# tribute to Win64
	mov	$arg2,$cnt
	mov	$arg2,$max

	rdtsc			# collect 1st tick
	mov	%eax,$lasttick	# lasttick = tick
	mov	\$0,$lastdiff	# lastdiff = 0
	clflush	($out)
	.byte	0xf0		# lock
	add	$lastdiff,($out)
	jmp	.Loop
.align	16
.Loop:	rdtsc
	mov	%eax,%edx
	sub	$lasttick,%eax
	mov	%edx,$lasttick
	mov	%eax,$lastdiff
	clflush	($out)
	.byte	0xf0		# lock
	add	%eax,($out)
	lea	4($out),$out
	sub	\$1,$cnt
	jnz	.Loop

	mov	$max,%rax
	ret
.size	OPENSSL_instrument_bus,.-OPENSSL_instrument_bus

.globl	OPENSSL_instrument_bus2
.type	OPENSSL_instrument_bus2,\@abi-omnipotent
.align	16
OPENSSL_instrument_bus2:
	mov	$arg1,$out	# tribute to Win64
	mov	$arg2,$cnt
	mov	$arg3,$max
	mov	$cnt,$redzone(%rsp)

	rdtsc			# collect 1st tick
	mov	%eax,$lasttick	# lasttick = tick
	mov	\$0,$lastdiff	# lastdiff = 0

	clflush	($out)
	.byte	0xf0		# lock
	add	$lastdiff,($out)

	rdtsc			# collect 1st diff
	mov	%eax,%edx
	sub	$lasttick,%eax	# diff
	mov	%edx,$lasttick	# lasttick = tick
	mov	%eax,$lastdiff	# lastdiff = diff
.Loop2:
	clflush	($out)
	.byte	0xf0		# lock
	add	%eax,($out)	# accumulate diff

	sub	\$1,$max
	jz	.Ldone2

	rdtsc
	mov	%eax,%edx
	sub	$lasttick,%eax	# diff
	mov	%edx,$lasttick	# lasttick = tick
	cmp	$lastdiff,%eax
	mov	%eax,$lastdiff	# lastdiff = diff
	mov	\$0,%edx
	setne	%dl
	sub	%rdx,$cnt	# conditional --$cnt
	lea	($out,%rdx,4),$out	# conditional ++$out
	jnz	.Loop2

.Ldone2:
	mov	$redzone(%rsp),%rax
	sub	$cnt,%rax
	ret
.size	OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
___
}

sub gen_random {
my $rdop = shift;
print<<___;
.globl	OPENSSL_ia32_${rdop}_bytes
.type	OPENSSL_ia32_${rdop}_bytes,\@abi-omnipotent
.align	16
OPENSSL_ia32_${rdop}_bytes:
	xor	%rax, %rax	# return value
	cmp	\$0,$arg2
	je	.Ldone_${rdop}_bytes

	mov	\$8,%r11
.Loop_${rdop}_bytes:
	${rdop}	%r10
	jc	.Lbreak_${rdop}_bytes
	dec	%r11
	jnz	.Loop_${rdop}_bytes
	jmp	.Ldone_${rdop}_bytes

.align	16
.Lbreak_${rdop}_bytes:
	cmp	\$8,$arg2
	jb	.Ltail_${rdop}_bytes
	mov	%r10,($arg1)
	lea	8($arg1),$arg1
	add	\$8,%rax
	sub	\$8,$arg2
	jz	.Ldone_${rdop}_bytes
	mov	\$8,%r11
	jmp	.Loop_${rdop}_bytes

.align	16
.Ltail_${rdop}_bytes:
	mov	%r10b,($arg1)
	lea	1($arg1),$arg1
	inc	%rax
	shr	\$8,%r10
	dec	$arg2
	jnz	.Ltail_${rdop}_bytes

.Ldone_${rdop}_bytes:
	xor	%r10,%r10	# Clear sensitive data from register
	ret
.size	OPENSSL_ia32_${rdop}_bytes,.-OPENSSL_ia32_${rdop}_bytes
___
}
gen_random("rdrand");
gen_random("rdseed");

close STDOUT;	# flush
Commit	Line	Data
e0a65194	1	#! /usr/bin/env perl
082193ef	2	# Copyright 2005-2018 The OpenSSL Project Authors. All Rights Reserved.
e0a65194	3	#
0e9725bc	4	# Licensed under the Apache License 2.0 (the "License"). You may not use
e0a65194 RS	5	# this file except in compliance with the License. You can obtain a copy
	6	# in the file LICENSE in the source distribution or at
	7	# https://www.openssl.org/source/license.html
	8
14e21f86	9
1aa89a7a RL	10	# $output is the last argument if it looks like a file (it has an extension)
	11	# $flavour is the first argument if it doesn't look like a file
	12	$output = $#ARGV >= 0 && $ARGV[$#ARGV] =~ m\|\.\w+$\| ? pop : undef;
	13	$flavour = $#ARGV >= 0 && $ARGV[0] !~ m\|\.\| ? shift : undef;
14e21f86	14
aa8f38e4	15	$win64=0; $win64=1 if ($flavour =~ /[nm]asm\|mingw64/ \|\| $output =~ /\.asm$/);
376729e1	16
aa8f38e4	17	$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
301799b8 AP	18	( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
	19	( $xlate="${dir}perlasm/x86_64-xlate.pl" and -f $xlate) or
	20	die "can't locate x86_64-xlate.pl";
	21
1aa89a7a RL	22	open OUT,"\| \"$^X\" \"$xlate\" $flavour \"$output\""
1aa89a7a RL	23	or die "can't call $xlate: $!";
46bf83f0	24	STDOUT=OUT;
376729e1	25
5fabb88a AP	26	($arg1,$arg2,$arg3,$arg4)=$win64?("%rcx","%rdx","%r8", "%r9") : # Win64 order
	27	("%rdi","%rsi","%rdx","%rcx"); # Unix order
	28
aa8f38e4 AP	29	print<<___;
aa8f38e4 AP	30	.extern OPENSSL_cpuid_setup
ddc20d4d	31	.hidden OPENSSL_cpuid_setup
aa8f38e4 AP	32	.section .init
aa8f38e4 AP	33	call OPENSSL_cpuid_setup
932cc129	34
ddc20d4d	35	.hidden OPENSSL_ia32cap_P
c5cd28bd	36	.comm OPENSSL_ia32cap_P,16,4
ddc20d4d	37
14e21f86	38	.text
376729e1 AP	39
376729e1 AP	40	.globl OPENSSL_atomic_add
aa8f38e4	41	.type OPENSSL_atomic_add,\@abi-omnipotent
376729e1 AP	42	.align 16
376729e1 AP	43	OPENSSL_atomic_add:
aa8f38e4 AP	44	movl ($arg1),%eax
	45	.Lspin: leaq ($arg2,%rax),%r8
	46	.byte 0xf0 # lock
	47	cmpxchgl %r8d,($arg1)
376729e1	48	jne .Lspin
e442c362	49	movl %r8d,%eax
aa8f38e4	50	.byte 0x48,0x98 # cltq/cdqe
376729e1 AP	51	ret
	52	.size OPENSSL_atomic_add,.-OPENSSL_atomic_add
	53
932cc129 AP	54	.globl OPENSSL_rdtsc
	55	.type OPENSSL_rdtsc,\@abi-omnipotent
	56	.align 16
	57	OPENSSL_rdtsc:
	58	rdtsc
	59	shl \$32,%rdx
	60	or %rdx,%rax
	61	ret
	62	.size OPENSSL_rdtsc,.-OPENSSL_rdtsc
	63
376729e1	64	.globl OPENSSL_ia32_cpuid
c5cd28bd	65	.type OPENSSL_ia32_cpuid,\@function,1
376729e1 AP	66	.align 16
376729e1 AP	67	OPENSSL_ia32_cpuid:
5e32cfb2	68	.cfi_startproc
b9064221	69	mov %rbx,%r8 # save %rbx
5e32cfb2	70	.cfi_register %rbx,%r8
9babf392 AP	71
9babf392 AP	72	xor %eax,%eax
d6ee8f3d	73	mov %rax,8(%rdi) # clear extended feature flags
9babf392	74	cpuid
761393bb AP	75	mov %eax,%r11d # max value for standard query level
761393bb AP	76
9babf392 AP	77	xor %eax,%eax
	78	cmp \$0x756e6547,%ebx # "Genu"
	79	setne %al
	80	mov %eax,%r9d
	81	cmp \$0x49656e69,%edx # "ineI"
	82	setne %al
	83	or %eax,%r9d
	84	cmp \$0x6c65746e,%ecx # "ntel"
	85	setne %al
5cd91b50	86	or %eax,%r9d # 0 indicates Intel CPU
5cd91b50 AP	87	jz .Lintel
	88
	89	cmp \$0x68747541,%ebx # "Auth"
	90	setne %al
	91	mov %eax,%r10d
	92	cmp \$0x69746E65,%edx # "enti"
	93	setne %al
	94	or %eax,%r10d
	95	cmp \$0x444D4163,%ecx # "cAMD"
	96	setne %al
	97	or %eax,%r10d # 0 indicates AMD CPU
	98	jnz .Lintel
	99
761393bb	100	# AMD specific
5cd91b50 AP	101	mov \$0x80000000,%eax
5cd91b50 AP	102	cpuid
b9064221 AP	103	cmp \$0x80000001,%eax
	104	jb .Lintel
	105	mov %eax,%r10d
	106	mov \$0x80000001,%eax
	107	cpuid
	108	or %ecx,%r9d
	109	and \$0x00000801,%r9d # isolate AMD XOP bit, 1<<11
	110
	111	cmp \$0x80000008,%r10d
5cd91b50 AP	112	jb .Lintel
	113
	114	mov \$0x80000008,%eax
	115	cpuid
	116	movzb %cl,%r10 # number of cores - 1
	117	inc %r10 # number of cores
9babf392	118
761393bb AP	119	mov \$1,%eax
	120	cpuid
	121	bt \$28,%edx # test hyper-threading bit
b9064221	122	jnc .Lgeneric
761393bb AP	123	shr \$16,%ebx # number of logical processors
761393bb AP	124	cmp %r10b,%bl
b9064221	125	ja .Lgeneric
761393bb	126	and \$0xefffffff,%edx # ~(1<<28)
b9064221	127	jmp .Lgeneric
761393bb	128
5cd91b50	129	.Lintel:
761393bb AP	130	cmp \$4,%r11d
	131	mov \$-1,%r10d
	132	jb .Lnocacheinfo
	133
	134	mov \$4,%eax
	135	mov \$0,%ecx # query L1D
	136	cpuid
	137	mov %eax,%r10d
	138	shr \$14,%r10d
	139	and \$0xfff,%r10d # number of cores -1 per L1D
	140
	141	.Lnocacheinfo:
932cc129	142	mov \$1,%eax
376729e1	143	cpuid
79337628	144	movd %eax,%xmm0 # put aside processor id
4bb90087	145	and \$0xbfefffff,%edx # force reserved bits to 0
932cc129	146	cmp \$0,%r9d
9babf392	147	jne .Lnotintel
4bb90087	148	or \$0x40000000,%edx # set reserved bit#30 on Intel CPUs
9babf392 AP	149	and \$15,%ah
9babf392 AP	150	cmp \$15,%ah # examine Family ID
64d92d74	151	jne .LnotP4
4bb90087	152	or \$0x00100000,%edx # set reserved bit#20 to engage RC4_CHAR
64d92d74 AP	153	.LnotP4:
	154	cmp \$6,%ah
	155	jne .Lnotintel
d84df594	156	and \$0x0fff0ff0,%eax
64d92d74 AP	157	cmp \$0x00050670,%eax # Knights Landing
	158	je .Lknights
	159	cmp \$0x00080650,%eax # Knights Mill (according to sde)
	160	jne .Lnotintel
	161	.Lknights:
	162	and \$0xfbffffff,%ecx # clear XSAVE flag to mimic Silvermont
	163
9babf392	164	.Lnotintel:
3df2eff4	165	bt \$28,%edx # test hyper-threading bit
b9064221	166	jnc .Lgeneric
761393bb AP	167	and \$0xefffffff,%edx # ~(1<<28)
761393bb AP	168	cmp \$0,%r10d
b9064221	169	je .Lgeneric
761393bb AP	170
761393bb AP	171	or \$0x10000000,%edx # 1<<28
9babf392	172	shr \$16,%ebx
761393bb	173	cmp \$1,%bl # see if cache is shared
b9064221	174	ja .Lgeneric
932cc129	175	and \$0xefffffff,%edx # ~(1<<28)
b9064221 AP	176	.Lgeneric:
	177	and \$0x00000800,%r9d # isolate AMD XOP flag
	178	and \$0xfffff7ff,%ecx
2bc3ad28	179	or %ecx,%r9d # merge AMD XOP flag
b9064221	180
2bc3ad28	181	mov %edx,%r10d # %r9d:%r10d is copy of %ecx:%edx
1aed5e1a AP	182
	183	cmp \$7,%r11d
	184	jb .Lno_extended_info
	185	mov \$7,%eax
	186	xor %ecx,%ecx
	187	cpuid
64d92d74 AP	188	bt \$26,%r9d # check XSAVE bit, cleared on Knights
	189	jc .Lnotknights
	190	and \$0xfff7ffff,%ebx # clear ADCX/ADOX flag
	191	.Lnotknights:
79337628 AP	192	movd %xmm0,%eax # restore processor id
	193	and \$0x0fff0ff0,%eax
	194	cmp \$0x00050650,%eax # Skylake-X
	195	jne .Lnotskylakex
	196	and \$0xfffeffff,%ebx # ~(1<<16)
	197	# suppress AVX512F flag on Skylake-X
	198	.Lnotskylakex:
1aed5e1a	199	mov %ebx,8(%rdi) # save extended feature flags
d6ee8f3d	200	mov %ecx,12(%rdi)
1aed5e1a AP	201	.Lno_extended_info:
1aed5e1a AP	202
2bc3ad28	203	bt \$27,%r9d # check OSXSAVE bit
b9064221 AP	204	jnc .Lclear_avx
	205	xor %ecx,%ecx # XCR0
	206	.byte 0x0f,0x01,0xd0 # xgetbv
66bee01c AP	207	and \$0xe6,%eax # isolate XMM, YMM and ZMM state support
	208	cmp \$0xe6,%eax
	209	je .Ldone
88ac224c AP	210	andl \$0x3fdeffff,8(%rdi) # ~(1<<31\|1<<30\|1<<21\|1<<16)
	211	# clear AVX512F+BW+VL+FIMA, all of
	212	# them are EVEX-encoded, which requires
	213	# ZMM state support even if one uses
	214	# only XMM and YMM :-(
b9064221 AP	215	and \$6,%eax # isolate XMM and YMM state support
	216	cmp \$6,%eax
	217	je .Ldone
	218	.Lclear_avx:
	219	mov \$0xefffe7ff,%eax # ~(1<<28\|1<<12\|1<<11)
2bc3ad28	220	and %eax,%r9d # clear AVX, FMA and AMD XOP bits
66bee01c	221	mov \$0x3fdeffdf,%eax # ~(1<<31\|1<<30\|1<<21\|1<<16\|1<<5)
d67e7554	222	and %eax,8(%rdi) # clear AVX2 and AVX512* bits
b9064221	223	.Ldone:
2bc3ad28 AP	224	shl \$32,%r9
2bc3ad28 AP	225	mov %r10d,%eax
b9064221	226	mov %r8,%rbx # restore %rbx
5e32cfb2	227	.cfi_restore %rbx
2bc3ad28	228	or %r9,%rax
376729e1	229	ret
5e32cfb2	230	.cfi_endproc
376729e1	231	.size OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
b2dba9bf AP	232
b2dba9bf AP	233	.globl OPENSSL_cleanse
aa8f38e4	234	.type OPENSSL_cleanse,\@abi-omnipotent
b2dba9bf AP	235	.align 16
	236	OPENSSL_cleanse:
	237	xor %rax,%rax
aa8f38e4	238	cmp \$15,$arg2
b2dba9bf	239	jae .Lot
7676eebf AP	240	cmp \$0,$arg2
7676eebf AP	241	je .Lret
b2dba9bf	242	.Little:
aa8f38e4 AP	243	mov %al,($arg1)
	244	sub \$1,$arg2
	245	lea 1($arg1),$arg1
b2dba9bf	246	jnz .Little
1fd79f66 AP	247	.Lret:
1fd79f66 AP	248	ret
b2dba9bf AP	249	.align 16
b2dba9bf AP	250	.Lot:
aa8f38e4	251	test \$7,$arg1
b2dba9bf	252	jz .Laligned
aa8f38e4 AP	253	mov %al,($arg1)
	254	lea -1($arg2),$arg2
	255	lea 1($arg1),$arg1
b2dba9bf AP	256	jmp .Lot
b2dba9bf AP	257	.Laligned:
aa8f38e4 AP	258	mov %rax,($arg1)
	259	lea -8($arg2),$arg2
	260	test \$-8,$arg2
	261	lea 8($arg1),$arg1
b2dba9bf	262	jnz .Laligned
aa8f38e4	263	cmp \$0,$arg2
b2dba9bf AP	264	jne .Little
	265	ret
	266	.size OPENSSL_cleanse,.-OPENSSL_cleanse
e33826f0 AP	267
	268	.globl CRYPTO_memcmp
	269	.type CRYPTO_memcmp,\@abi-omnipotent
	270	.align 16
	271	CRYPTO_memcmp:
	272	xor %rax,%rax
	273	xor %r10,%r10
	274	cmp \$0,$arg3
	275	je .Lno_data
9a708bf9 AP	276	cmp \$16,$arg3
	277	jne .Loop_cmp
	278	mov ($arg1),%r10
	279	mov 8($arg1),%r11
	280	mov \$1,$arg3
	281	xor ($arg2),%r10
	282	xor 8($arg2),%r11
	283	or %r11,%r10
	284	cmovnz $arg3,%rax
	285	ret
	286
	287	.align 16
e33826f0 AP	288	.Loop_cmp:
	289	mov ($arg1),%r10b
	290	lea 1($arg1),$arg1
	291	xor ($arg2),%r10b
	292	lea 1($arg2),$arg2
	293	or %r10b,%al
	294	dec $arg3
	295	jnz .Loop_cmp
	296	neg %rax
	297	shr \$63,%rax
	298	.Lno_data:
	299	ret
	300	.size CRYPTO_memcmp,.-CRYPTO_memcmp
14e21f86	301	___
aa8f38e4 AP	302
	303	print<<___ if (!$win64);
	304	.globl OPENSSL_wipe_cpu
	305	.type OPENSSL_wipe_cpu,\@abi-omnipotent
	306	.align 16
	307	OPENSSL_wipe_cpu:
	308	pxor %xmm0,%xmm0
	309	pxor %xmm1,%xmm1
	310	pxor %xmm2,%xmm2
	311	pxor %xmm3,%xmm3
	312	pxor %xmm4,%xmm4
	313	pxor %xmm5,%xmm5
	314	pxor %xmm6,%xmm6
	315	pxor %xmm7,%xmm7
	316	pxor %xmm8,%xmm8
	317	pxor %xmm9,%xmm9
	318	pxor %xmm10,%xmm10
	319	pxor %xmm11,%xmm11
	320	pxor %xmm12,%xmm12
	321	pxor %xmm13,%xmm13
	322	pxor %xmm14,%xmm14
	323	pxor %xmm15,%xmm15
	324	xorq %rcx,%rcx
	325	xorq %rdx,%rdx
	326	xorq %rsi,%rsi
	327	xorq %rdi,%rdi
	328	xorq %r8,%r8
	329	xorq %r9,%r9
	330	xorq %r10,%r10
	331	xorq %r11,%r11
	332	leaq 8(%rsp),%rax
	333	ret
	334	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	335	___
	336	print<<___ if ($win64);
	337	.globl OPENSSL_wipe_cpu
	338	.type OPENSSL_wipe_cpu,\@abi-omnipotent
	339	.align 16
	340	OPENSSL_wipe_cpu:
	341	pxor %xmm0,%xmm0
	342	pxor %xmm1,%xmm1
	343	pxor %xmm2,%xmm2
	344	pxor %xmm3,%xmm3
	345	pxor %xmm4,%xmm4
	346	pxor %xmm5,%xmm5
	347	xorq %rcx,%rcx
	348	xorq %rdx,%rdx
	349	xorq %r8,%r8
	350	xorq %r9,%r9
	351	xorq %r10,%r10
	352	xorq %r11,%r11
	353	leaq 8(%rsp),%rax
	354	ret
	355	.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
	356	___
5fabb88a AP	357	{
	358	my $out="%r10";
	359	my $cnt="%rcx";
	360	my $max="%r11";
	361	my $lasttick="%r8d";
	362	my $lastdiff="%r9d";
	363	my $redzone=win64?8:-8;
	364
	365	print<<___;
	366	.globl OPENSSL_instrument_bus
	367	.type OPENSSL_instrument_bus,\@abi-omnipotent
	368	.align 16
	369	OPENSSL_instrument_bus:
	370	mov $arg1,$out # tribute to Win64
	371	mov $arg2,$cnt
	372	mov $arg2,$max
	373
	374	rdtsc # collect 1st tick
	375	mov %eax,$lasttick # lasttick = tick
	376	mov \$0,$lastdiff # lastdiff = 0
	377	clflush ($out)
b9064221	378	.byte 0xf0 # lock
5fabb88a AP	379	add $lastdiff,($out)
	380	jmp .Loop
	381	.align 16
	382	.Loop: rdtsc
	383	mov %eax,%edx
	384	sub $lasttick,%eax
	385	mov %edx,$lasttick
	386	mov %eax,$lastdiff
	387	clflush ($out)
b9064221	388	.byte 0xf0 # lock
5fabb88a AP	389	add %eax,($out)
	390	lea 4($out),$out
	391	sub \$1,$cnt
	392	jnz .Loop
	393
	394	mov $max,%rax
	395	ret
	396	.size OPENSSL_instrument_bus,.-OPENSSL_instrument_bus
	397
	398	.globl OPENSSL_instrument_bus2
	399	.type OPENSSL_instrument_bus2,\@abi-omnipotent
	400	.align 16
	401	OPENSSL_instrument_bus2:
	402	mov $arg1,$out # tribute to Win64
	403	mov $arg2,$cnt
	404	mov $arg3,$max
	405	mov $cnt,$redzone(%rsp)
	406
	407	rdtsc # collect 1st tick
	408	mov %eax,$lasttick # lasttick = tick
	409	mov \$0,$lastdiff # lastdiff = 0
	410
	411	clflush ($out)
b9064221	412	.byte 0xf0 # lock
5fabb88a AP	413	add $lastdiff,($out)
	414
	415	rdtsc # collect 1st diff
	416	mov %eax,%edx
	417	sub $lasttick,%eax # diff
	418	mov %edx,$lasttick # lasttick = tick
	419	mov %eax,$lastdiff # lastdiff = diff
	420	.Loop2:
	421	clflush ($out)
b9064221	422	.byte 0xf0 # lock
5fabb88a AP	423	add %eax,($out) # accumulate diff
	424
	425	sub \$1,$max
	426	jz .Ldone2
	427
	428	rdtsc
	429	mov %eax,%edx
	430	sub $lasttick,%eax # diff
	431	mov %edx,$lasttick # lasttick = tick
	432	cmp $lastdiff,%eax
	433	mov %eax,$lastdiff # lastdiff = diff
	434	mov \$0,%edx
	435	setne %dl
	436	sub %rdx,$cnt # conditional --$cnt
	437	lea ($out,%rdx,4),$out # conditional ++$out
	438	jnz .Loop2
	439
	440	.Ldone2:
	441	mov $redzone(%rsp),%rax
	442	sub $cnt,%rax
	443	ret
	444	.size OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
	445	___
	446	}
aa8f38e4	447
9c940446 AP	448	sub gen_random {
9c940446 AP	449	my $rdop = shift;
301799b8	450	print<<___;
9c940446 AP	451	.globl OPENSSL_ia32_${rdop}_bytes
9c940446 AP	452	.type OPENSSL_ia32_${rdop}_bytes,\@abi-omnipotent
f4d45640	453	.align 16
9c940446 AP	454	OPENSSL_ia32_${rdop}_bytes:
	455	xor %rax, %rax # return value
	456	cmp \$0,$arg2
	457	je .Ldone_${rdop}_bytes
	458
	459	mov \$8,%r11
	460	.Loop_${rdop}_bytes:
	461	${rdop} %r10
	462	jc .Lbreak_${rdop}_bytes
	463	dec %r11
	464	jnz .Loop_${rdop}_bytes
	465	jmp .Ldone_${rdop}_bytes
	466
	467	.align 16
	468	.Lbreak_${rdop}_bytes:
	469	cmp \$8,$arg2
	470	jb .Ltail_${rdop}_bytes
	471	mov %r10,($arg1)
	472	lea 8($arg1),$arg1
	473	add \$8,%rax
	474	sub \$8,$arg2
	475	jz .Ldone_${rdop}_bytes
	476	mov \$8,%r11
	477	jmp .Loop_${rdop}_bytes
	478
	479	.align 16
	480	.Ltail_${rdop}_bytes:
	481	mov %r10b,($arg1)
	482	lea 1($arg1),$arg1
	483	inc %rax
082193ef	484	shr \$8,%r10
9c940446 AP	485	dec $arg2
	486	jnz .Ltail_${rdop}_bytes
	487
	488	.Ldone_${rdop}_bytes:
082193ef	489	xor %r10,%r10 # Clear sensitive data from register
f4d45640	490	ret
9c940446	491	.size OPENSSL_ia32_${rdop}_bytes,.-OPENSSL_ia32_${rdop}_bytes
301799b8	492	___
9c940446 AP	493	}
	494	gen_random("rdrand");
	495	gen_random("rdseed");
301799b8	496
5d863367	497	close STDOUT; # flush