mirror of
https://github.com/rn10950/RetroZilla.git
synced 2024-11-14 03:30:17 +01:00
108 lines
3.8 KiB
NASM
108 lines
3.8 KiB
NASM
; This Source Code Form is subject to the terms of the Mozilla Public
|
|
; License, v. 2.0. If a copy of the MPL was not distributed with this
|
|
; file, You can obtain one at http://mozilla.org/MPL/2.0/.
|
|
|
|
; ** ARCFOUR implementation optimized for AMD64.
|
|
; **
|
|
; ** The throughput achieved by this code is about 320 MBytes/sec, on
|
|
; ** a 1.8 GHz AMD Opteron (rev C0) processor.
|
|
|
|
.CODE
|
|
|
|
; extern void ARCFOUR(RC4Context *cx, unsigned long long inputLen,
|
|
; const unsigned char *input, unsigned char *output);
|
|
|
|
|
|
ARCFOUR PROC
|
|
|
|
push rbp
|
|
push rbx
|
|
push rsi
|
|
push rdi
|
|
|
|
mov rbp, rcx ; key = ARG(key)
|
|
mov rbx, rdx ; rbx = ARG(len)
|
|
mov rsi, r8 ; in = ARG(in)
|
|
mov rdi, r9 ; out = ARG(out)
|
|
mov rcx, [rbp] ; x = key->x
|
|
mov rdx, [rbp+8] ; y = key->y
|
|
add rbp, 16 ; d = key->data
|
|
inc rcx ; x++
|
|
and rcx, 0ffh ; x &= 0xff
|
|
lea rbx, [rbx+rsi-8] ; rbx = in+len-8
|
|
mov r9, rbx ; tmp = in+len-8
|
|
mov rax, [rbp+rcx*8] ; tx = d[x]
|
|
cmp rbx, rsi ; cmp in with in+len-8
|
|
jl Lend ; jump if (in+len-8 < in)
|
|
|
|
Lstart:
|
|
add rsi, 8 ; increment in
|
|
add rdi, 8 ; increment out
|
|
|
|
;
|
|
; generate the next 8 bytes of the rc4 stream into r8
|
|
;
|
|
|
|
mov r11, 8 ; byte counter
|
|
|
|
@@:
|
|
add dl, al ; y += tx
|
|
mov ebx, [rbp+rdx*8] ; ty = d[y]
|
|
mov [rbp+rcx*8], ebx ; d[x] = ty
|
|
add bl, al ; val = ty + tx
|
|
mov [rbp+rdx*8], eax ; d[y] = tx
|
|
inc cl ; x++ (NEXT ROUND)
|
|
mov eax, [rbp+rcx*8] ; tx = d[x] (NEXT ROUND)
|
|
mov r8b, [rbp+rbx*8] ; val = d[val]
|
|
dec r11b
|
|
ror r8, 8 ; (ror does not change ZF)
|
|
jnz @b
|
|
|
|
;
|
|
; xor 8 bytes
|
|
;
|
|
|
|
xor r8, [rsi-8]
|
|
cmp rsi, r9 ; cmp in+len-8 with in
|
|
mov [rdi-8], r8
|
|
jle Lstart
|
|
|
|
Lend:
|
|
add r9, 8 ; tmp = in+len
|
|
|
|
;
|
|
; handle the last bytes, one by one
|
|
;
|
|
|
|
@@:
|
|
cmp r9, rsi ; cmp in with in+len
|
|
jle Lfinished ; jump if (in+len <= in)
|
|
add dl, al ; y += tx
|
|
mov ebx, [rbp+rdx*8] ; ty = d[y]
|
|
mov [rbp+rcx*8], ebx ; d[x] = ty
|
|
add bl, al ; val = ty + tx
|
|
mov [rbp+rdx*8], eax ; d[y] = tx
|
|
inc cl ; x++ (NEXT ROUND)
|
|
mov eax, [rbp+rcx*8] ; tx = d[x] (NEXT ROUND)
|
|
mov r8b, [rbp+rbx*8] ; val = d[val]
|
|
xor r8b, [rsi] ; xor 1 byte
|
|
mov [rdi], r8b
|
|
inc rsi ; in++
|
|
inc rdi
|
|
jmp @b
|
|
|
|
Lfinished:
|
|
dec rcx ; x--
|
|
mov [rbp-8], dl ; key->y = y
|
|
mov [rbp-16], cl ; key->x = x
|
|
|
|
pop rdi
|
|
pop rsi
|
|
pop rbx
|
|
pop rbp
|
|
ret
|
|
|
|
ARCFOUR ENDP
|
|
|
|
END
|