Hakase/openssl-custom
1
0 Fork 0
Code Issues Pull Requests Releases Wiki Activity

Latest update

Browse Source
This commit is contained in:
Hakase
2019-02-22 12:53:31 +09:00
parent 6523a76e2e
commit 25fbda8e8b
186 changed files with 9758 additions and 1640 deletions
Download Patch File Download Diff File Expand all files Collapse all files
crypto/sha/asm/keccak1600-armv4.pl
+31 -3
View File
@@ -113,8 +113,6 @@ my @T = map([ 8*$_, 8*($_+1), 8*($_+2), 8*($_+3), 8*($_+4) ], (30,35,40,45,50));
$code.=<<___;
#include "arm_arch.h"
.text
#if defined(__thumb2__)
.syntax unified
.thumb
@@ -122,6 +120,8 @@ $code.=<<___;
.code 32
#endif
.text
.type iotas32, %object
.align 5
iotas32:
@@ -691,7 +691,14 @@ ___
$code.=<<___;
blo .Lround2x
#if __ARM_ARCH__>=5
ldr pc,[sp,#440]
#else
ldr lr,[sp,#440]
tst lr,#1
moveq pc,lr @ be binary compatible with V4, yet
bx lr @ interoperable with Thumb ISA:-)
#endif
.size KeccakF1600_int,.-KeccakF1600_int
.type KeccakF1600, %function
@@ -730,7 +737,14 @@ KeccakF1600:
stmia @E[1], {@C[0]-@C[9]}
add sp,sp,#440+20
#if __ARM_ARCH__>=5
ldmia sp!,{r4-r11,pc}
#else
ldmia sp!,{r4-r11,lr}
tst lr,#1
moveq pc,lr @ be binary compatible with V4, yet
bx lr @ interoperable with Thumb ISA:-)
#endif
.size KeccakF1600,.-KeccakF1600
___
{ my ($A_flat,$inp,$len,$bsz) = map("r$_",(10..12,14));
@@ -905,7 +919,14 @@ SHA3_absorb:
.Labsorb_abort:
add sp,sp,#456+32
mov r0,$len @ return value
#if __ARM_ARCH__>=5
ldmia sp!,{r4-r12,pc}
#else
ldmia sp!,{r4-r12,lr}
tst lr,#1
moveq pc,lr @ be binary compatible with V4, yet
bx lr @ interoperable with Thumb ISA:-)
#endif
.size SHA3_absorb,.-SHA3_absorb
___
}
@@ -1055,7 +1076,14 @@ SHA3_squeeze:
.align 4
.Lsqueeze_done:
add sp,sp,#24
#if __ARM_ARCH__>=5
ldmia sp!,{r4-r10,pc}
#else
ldmia sp!,{r4-r10,lr}
tst lr,#1
moveq pc,lr @ be binary compatible with V4, yet
bx lr @ interoperable with Thumb ISA:-)
#endif
.size SHA3_squeeze,.-SHA3_squeeze
___
}
@@ -1265,7 +1293,7 @@ KeccakF1600_neon:
subs r3, r3, #1
bne .Loop_neon
bx lr
ret
.size KeccakF1600_neon,.-KeccakF1600_neon
.global SHA3_absorb_neon
crypto/sha/asm/keccak1600-armv8.pl
+14
View File
@@ -121,6 +121,7 @@ $code.=<<___;
.align 5
KeccakF1600_int:
adr $C[2],iotas
.inst 0xd503233f // paciasp
stp $C[2],x30,[sp,#16] // 32 bytes on top are mine
b .Loop
.align 4
@@ -292,12 +293,14 @@ $code.=<<___;
bne .Loop
ldr x30,[sp,#24]
.inst 0xd50323bf // autiasp
ret
.size KeccakF1600_int,.-KeccakF1600_int
.type KeccakF1600,%function
.align 5
KeccakF1600:
.inst 0xd503233f // paciasp
stp x29,x30,[sp,#-128]!
add x29,sp,#0
stp x19,x20,[sp,#16]
@@ -347,6 +350,7 @@ KeccakF1600:
ldp x25,x26,[x29,#64]
ldp x27,x28,[x29,#80]
ldp x29,x30,[sp],#128
.inst 0xd50323bf // autiasp
ret
.size KeccakF1600,.-KeccakF1600
@@ -354,6 +358,7 @@ KeccakF1600:
.type SHA3_absorb,%function
.align 5
SHA3_absorb:
.inst 0xd503233f // paciasp
stp x29,x30,[sp,#-128]!
add x29,sp,#0
stp x19,x20,[sp,#16]
@@ -451,6 +456,7 @@ $code.=<<___;
ldp x25,x26,[x29,#64]
ldp x27,x28,[x29,#80]
ldp x29,x30,[sp],#128
.inst 0xd50323bf // autiasp
ret
.size SHA3_absorb,.-SHA3_absorb
___
@@ -461,6 +467,7 @@ $code.=<<___;
.type SHA3_squeeze,%function
.align 5
SHA3_squeeze:
.inst 0xd503233f // paciasp
stp x29,x30,[sp,#-48]!
add x29,sp,#0
stp x19,x20,[sp,#16]
@@ -523,6 +530,7 @@ SHA3_squeeze:
ldp x19,x20,[sp,#16]
ldp x21,x22,[sp,#32]
ldp x29,x30,[sp],#48
.inst 0xd50323bf // autiasp
ret
.size SHA3_squeeze,.-SHA3_squeeze
___
@@ -641,6 +649,7 @@ KeccakF1600_ce:
.type KeccakF1600_cext,%function
.align 5
KeccakF1600_cext:
.inst 0xd503233f // paciasp
stp x29,x30,[sp,#-80]!
add x29,sp,#0
stp d8,d9,[sp,#16] // per ABI requirement
@@ -673,6 +682,7 @@ $code.=<<___;
ldp d12,d13,[sp,#48]
ldp d14,d15,[sp,#64]
ldr x29,[sp],#80
.inst 0xd50323bf // autiasp
ret
.size KeccakF1600_cext,.-KeccakF1600_cext
___
@@ -685,6 +695,7 @@ $code.=<<___;
.type SHA3_absorb_cext,%function
.align 5
SHA3_absorb_cext:
.inst 0xd503233f // paciasp
stp x29,x30,[sp,#-80]!
add x29,sp,#0
stp d8,d9,[sp,#16] // per ABI requirement
@@ -756,6 +767,7 @@ $code.=<<___;
ldp d12,d13,[sp,#48]
ldp d14,d15,[sp,#64]
ldp x29,x30,[sp],#80
.inst 0xd50323bf // autiasp
ret
.size SHA3_absorb_cext,.-SHA3_absorb_cext
___
@@ -767,6 +779,7 @@ $code.=<<___;
.type SHA3_squeeze_cext,%function
.align 5
SHA3_squeeze_cext:
.inst 0xd503233f // paciasp
stp x29,x30,[sp,#-16]!
add x29,sp,#0
mov x9,$ctx
@@ -822,6 +835,7 @@ SHA3_squeeze_cext:
.Lsqueeze_done_ce:
ldr x29,[sp],#16
.inst 0xd50323bf // autiasp
ret
.size SHA3_squeeze_cext,.-SHA3_squeeze_cext
___
crypto/sha/asm/sha1-armv4-large.pl
+14 -6
View File
@@ -187,7 +187,6 @@ ___
$code=<<___;
#include "arm_arch.h"
.text
#if defined(__thumb2__)
.syntax unified
.thumb
@@ -195,6 +194,8 @@ $code=<<___;
.code 32
#endif
.text
.global sha1_block_data_order
.type sha1_block_data_order,%function
@@ -202,12 +203,14 @@ $code=<<___;
sha1_block_data_order:
#if __ARM_MAX_ARCH__>=7
.Lsha1_block:
adr r3,.Lsha1_block
ldr r12,.LOPENSSL_armcap
# if !defined(_WIN32)
adr r3,.Lsha1_block
ldr r12,[r3,r12] @ OPENSSL_armcap_P
#ifdef __APPLE__
# endif
# if defined(__APPLE__) || defined(_WIN32)
ldr r12,[r12]
#endif
# endif
tst r12,#ARMV8_SHA1
bne .LARMv8
tst r12,#ARMV7_NEON
@@ -311,7 +314,11 @@ $code.=<<___;
.LK_60_79: .word 0xca62c1d6
#if __ARM_MAX_ARCH__>=7
.LOPENSSL_armcap:
# ifdef _WIN32
.word OPENSSL_armcap_P
# else
.word OPENSSL_armcap_P-.Lsha1_block
# endif
#endif
.asciz "SHA1 block transform for ARMv4/NEON/ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
.align 5
@@ -613,14 +620,15 @@ my ($ABCD,$E,$E0,$E1)=map("q$_",(0..3));
my @MSG=map("q$_",(4..7));
my @Kxx=map("q$_",(8..11));
my ($W0,$W1,$ABCD_SAVE)=map("q$_",(12..14));
my $_byte = ($flavour =~ /win/ ? "DCB" : ".byte");
$code.=<<___;
#if __ARM_MAX_ARCH__>=7
# if defined(__thumb2__)
# define INST(a,b,c,d) .byte c,d|0xf,a,b
# define INST(a,b,c,d) $_byte c,d|0xf,a,b
# else
# define INST(a,b,c,d) .byte a,b,c,d|0x10
# define INST(a,b,c,d) $_byte a,b,c,d|0x10
# endif
.type sha1_block_data_order_armv8,%function
crypto/sha/asm/sha1-armv8.pl
+11 -19
View File
@@ -58,10 +58,10 @@ $code.=<<___ if ($i<15 && !($i&1));
lsr @Xx[$i+1],@Xx[$i],#32
___
$code.=<<___ if ($i<14 && !($i&1));
ldr @Xx[$i+2],[$inp,#`($i+2)*4-64`]
ldur @Xx[$i+2],[$inp,#`($i+2)*4-64`]
___
$code.=<<___ if ($i<14 && ($i&1));
#ifdef __ARMEB__
#ifdef __AARCH64EB__
ror @Xx[$i+1],@Xx[$i+1],#32
#else
rev32 @Xx[$i+1],@Xx[$i+1]
@@ -171,23 +171,19 @@ ___
}
$code.=<<___;
#include "arm_arch.h"
#ifndef __KERNEL__
# include "arm_arch.h"
.extern OPENSSL_armcap_P
#endif
.text
.extern OPENSSL_armcap_P
.globl sha1_block_data_order
.type sha1_block_data_order,%function
.align 6
sha1_block_data_order:
#ifdef __ILP32__
ldrsw x16,.LOPENSSL_armcap_P
#else
ldr x16,.LOPENSSL_armcap_P
#endif
adr x17,.LOPENSSL_armcap_P
add x16,x16,x17
ldr w16,[x16]
adrp x16,OPENSSL_armcap_P
ldr w16,[x16,#:lo12:OPENSSL_armcap_P]
tst w16,#ARMV8_SHA1
b.ne .Lv8_entry
@@ -208,7 +204,7 @@ sha1_block_data_order:
movz $K,#0x7999
sub $num,$num,#1
movk $K,#0x5a82,lsl#16
#ifdef __ARMEB__
#ifdef __AARCH64EB__
ror $Xx[0],@Xx[0],#32
#else
rev32 @Xx[0],@Xx[0]
@@ -321,15 +317,11 @@ $code.=<<___;
.long 0x6ed9eba1,0x6ed9eba1,0x6ed9eba1,0x6ed9eba1 //K_20_39
.long 0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc,0x8f1bbcdc //K_40_59
.long 0xca62c1d6,0xca62c1d6,0xca62c1d6,0xca62c1d6 //K_60_79
.LOPENSSL_armcap_P:
#ifdef __ILP32__
.long OPENSSL_armcap_P-.
#else
.quad OPENSSL_armcap_P-.
#endif
.asciz "SHA1 block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
.align 2
#if !defined(__KERNELL__) && !defined(_WIN64)
.comm OPENSSL_armcap_P,4,4
#endif
___
}}}
crypto/sha/asm/sha256-armv4.pl
+13 -5
View File
@@ -181,7 +181,6 @@ $code=<<___;
# define __ARM_MAX_ARCH__ 7
#endif
.text
#if defined(__thumb2__)
.syntax unified
.thumb
@@ -189,6 +188,8 @@ $code=<<___;
.code 32
#endif
.text
.type K256,%object
.align 5
K256:
@@ -212,7 +213,11 @@ K256:
.word 0 @ terminator
#if __ARM_MAX_ARCH__>=7 && !defined(__KERNEL__)
.LOPENSSL_armcap:
# ifdef _WIN32
.word OPENSSL_armcap_P
# else
.word OPENSSL_armcap_P-.Lsha256_block_data_order
# endif
#endif
.align 5
@@ -227,10 +232,12 @@ sha256_block_data_order:
#endif
#if __ARM_MAX_ARCH__>=7 && !defined(__KERNEL__)
ldr r12,.LOPENSSL_armcap
# if !defined(_WIN32)
ldr r12,[r3,r12] @ OPENSSL_armcap_P
#ifdef __APPLE__
# endif
# if defined(__APPLE__) || defined(_WIN32)
ldr r12,[r12]
#endif
# endif
tst r12,#ARMV8_SHA256
bne .LARMv8
tst r12,#ARMV7_NEON
@@ -598,14 +605,15 @@ my ($ABCD,$EFGH,$abcd)=map("q$_",(0..2));
my @MSG=map("q$_",(8..11));
my ($W0,$W1,$ABCD_SAVE,$EFGH_SAVE)=map("q$_",(12..15));
my $Ktbl="r3";
my $_byte = ($flavour =~ /win/ ? "DCB" : ".byte");
$code.=<<___;
#if __ARM_MAX_ARCH__>=7 && !defined(__KERNEL__)
# if defined(__thumb2__)
# define INST(a,b,c,d) .byte c,d|0xc,a,b
# define INST(a,b,c,d) $_byte c,d|0xc,a,b
# else
# define INST(a,b,c,d) .byte a,b,c,d
# define INST(a,b,c,d) $_byte a,b,c,d
# endif
.type sha256_block_data_order_armv8,%function
crypto/sha/asm/sha512-armv4.pl
+15 -5
View File
@@ -196,6 +196,9 @@ $code.=<<___;
add $Ktbl,$Ktbl,#8
___
}
my $_word = ($flavour =~ /win/ ? "DCDU" : ".word");
$code=<<___;
#ifndef __KERNEL__
# include "arm_arch.h"
@@ -211,14 +214,13 @@ $code=<<___;
#ifdef __ARMEL__
# define LO 0
# define HI 4
# define WORD64(hi0,lo0,hi1,lo1) .word lo0,hi0, lo1,hi1
# define WORD64(hi0,lo0,hi1,lo1) $_word lo0,hi0, lo1,hi1
#else
# define HI 0
# define LO 4
# define WORD64(hi0,lo0,hi1,lo1) .word hi0,lo0, hi1,lo1
# define WORD64(hi0,lo0,hi1,lo1) $_word hi0,lo0, hi1,lo1
#endif
.text
#if defined(__thumb2__)
.syntax unified
.thumb
@@ -227,6 +229,8 @@ $code=<<___;
.code 32
#endif
.text
.type K512,%object
.align 5
K512:
@@ -273,7 +277,11 @@ WORD64(0x5fcb6fab,0x3ad6faec, 0x6c44198c,0x4a475817)
.size K512,.-K512
#if __ARM_MAX_ARCH__>=7 && !defined(__KERNEL__)
.LOPENSSL_armcap:
# ifdef _WIN32
.word OPENSSL_armcap_P
# else
.word OPENSSL_armcap_P-.Lsha512_block_data_order
# endif
.skip 32-4
#else
.skip 32
@@ -290,10 +298,12 @@ sha512_block_data_order:
#endif
#if __ARM_MAX_ARCH__>=7 && !defined(__KERNEL__)
ldr r12,.LOPENSSL_armcap
# if !defined(_WIN32)
ldr r12,[r3,r12] @ OPENSSL_armcap_P
#ifdef __APPLE__
# endif
# if defined(__APPLE__) || defined(_WIN32)
ldr r12,[r12]
#endif
# endif
tst r12,#ARMV7_NEON
bne .LNEON
#endif
crypto/sha/asm/sha512-armv8.pl
+6 -19
View File
@@ -188,24 +188,18 @@ ___
$code.=<<___;
#ifndef __KERNEL__
# include "arm_arch.h"
.extern OPENSSL_armcap_P
#endif
.text
.extern OPENSSL_armcap_P
.globl $func
.type $func,%function
.align 6
$func:
#ifndef __KERNEL__
# ifdef __ILP32__
ldrsw x16,.LOPENSSL_armcap_P
# else
ldr x16,.LOPENSSL_armcap_P
# endif
adr x17,.LOPENSSL_armcap_P
add x16,x16,x17
ldr w16,[x16]
adrp x16,OPENSSL_armcap_P
ldr w16,[x16,#:lo12:OPENSSL_armcap_P]
___
$code.=<<___ if ($SZ==4);
tst w16,#ARMV8_SHA256
@@ -219,6 +213,7 @@ $code.=<<___ if ($SZ==8);
___
$code.=<<___;
#endif
.inst 0xd503233f // paciasp
stp x29,x30,[sp,#-128]!
add x29,sp,#0
@@ -280,6 +275,7 @@ $code.=<<___;
ldp x25,x26,[x29,#64]
ldp x27,x28,[x29,#80]
ldp x29,x30,[sp],#128
.inst 0xd50323bf // autiasp
ret
.size $func,.-$func
@@ -351,15 +347,6 @@ $code.=<<___ if ($SZ==4);
___
$code.=<<___;
.size .LK$BITS,.-.LK$BITS
#ifndef __KERNEL__
.align 3
.LOPENSSL_armcap_P:
# ifdef __ILP32__
.long OPENSSL_armcap_P-.
# else
.quad OPENSSL_armcap_P-.
# endif
#endif
.asciz "SHA$BITS block transform for ARMv8, CRYPTOGAMS by <appro\@openssl.org>"
.align 2
___
@@ -839,7 +826,7 @@ ___
}
$code.=<<___;
#ifndef __KERNEL__
#if !defined(__KERNEL__) && !defined(_WIN64)
.comm OPENSSL_armcap_P,4,4
#endif
___