• file
• revisions
• annotate
• diff
• comparison
• raw

security/nss/lib/freebl/arcfour-amd64-masm.asm@b8a032363ba2 (annotated)

security/nss/lib/freebl/arcfour-amd64-masm.asm

Thu, 22 Jan 2015 13:21:57 +0100

author

Michael Schloh von Bennewitz <michael@schloh.com>

date

Thu, 22 Jan 2015 13:21:57 +0100

branch

TOR_BUG_9701

changeset 15

b8a032363ba2

permissions

-rw-r--r--

Incorporate requested changes from Mozilla in review:
https://bugzilla.mozilla.org/show_bug.cgi?id=1123480#c6

michael@0	1	; This Source Code Form is subject to the terms of the Mozilla Public
michael@0	2	; License, v. 2.0. If a copy of the MPL was not distributed with this
michael@0	3	; file, You can obtain one at http://mozilla.org/MPL/2.0/.
michael@0	4
michael@0	5	; ** ARCFOUR implementation optimized for AMD64.
michael@0	6	; **
michael@0	7	; ** The throughput achieved by this code is about 320 MBytes/sec, on
michael@0	8	; ** a 1.8 GHz AMD Opteron (rev C0) processor.
michael@0	9
michael@0	10	.CODE
michael@0	11
michael@0	12	; extern void ARCFOUR(RC4Context *cx, unsigned long long inputLen,
michael@0	13	; const unsigned char *input, unsigned char *output);
michael@0	14
michael@0	15
michael@0	16	ARCFOUR PROC
michael@0	17
michael@0	18	push rbp
michael@0	19	push rbx
michael@0	20	push rsi
michael@0	21	push rdi
michael@0	22
michael@0	23	mov rbp, rcx ; key = ARG(key)
michael@0	24	mov rbx, rdx ; rbx = ARG(len)
michael@0	25	mov rsi, r8 ; in = ARG(in)
michael@0	26	mov rdi, r9 ; out = ARG(out)
michael@0	27	mov rcx, [rbp] ; x = key->x
michael@0	28	mov rdx, [rbp+8] ; y = key->y
michael@0	29	add rbp, 16 ; d = key->data
michael@0	30	inc rcx ; x++
michael@0	31	and rcx, 0ffh ; x &= 0xff
michael@0	32	lea rbx, [rbx+rsi-8] ; rbx = in+len-8
michael@0	33	mov r9, rbx ; tmp = in+len-8
michael@0	34	mov rax, [rbp+rcx*8] ; tx = d[x]
michael@0	35	cmp rbx, rsi ; cmp in with in+len-8
michael@0	36	jl Lend ; jump if (in+len-8 < in)
michael@0	37
michael@0	38	Lstart:
michael@0	39	add rsi, 8 ; increment in
michael@0	40	add rdi, 8 ; increment out
michael@0	41
michael@0	42	;
michael@0	43	; generate the next 8 bytes of the rc4 stream into r8
michael@0	44	;
michael@0	45
michael@0	46	mov r11, 8 ; byte counter
michael@0	47
michael@0	48	@@:
michael@0	49	add dl, al ; y += tx
michael@0	50	mov ebx, [rbp+rdx*8] ; ty = d[y]
michael@0	51	mov [rbp+rcx*8], ebx ; d[x] = ty
michael@0	52	add bl, al ; val = ty + tx
michael@0	53	mov [rbp+rdx*8], eax ; d[y] = tx
michael@0	54	inc cl ; x++ (NEXT ROUND)
michael@0	55	mov eax, [rbp+rcx*8] ; tx = d[x] (NEXT ROUND)
michael@0	56	mov r8b, [rbp+rbx*8] ; val = d[val]
michael@0	57	dec r11b
michael@0	58	ror r8, 8 ; (ror does not change ZF)
michael@0	59	jnz @b
michael@0	60
michael@0	61	;
michael@0	62	; xor 8 bytes
michael@0	63	;
michael@0	64
michael@0	65	xor r8, [rsi-8]
michael@0	66	cmp rsi, r9 ; cmp in+len-8 with in
michael@0	67	mov [rdi-8], r8
michael@0	68	jle Lstart
michael@0	69
michael@0	70	Lend:
michael@0	71	add r9, 8 ; tmp = in+len
michael@0	72
michael@0	73	;
michael@0	74	; handle the last bytes, one by one
michael@0	75	;
michael@0	76
michael@0	77	@@:
michael@0	78	cmp r9, rsi ; cmp in with in+len
michael@0	79	jle Lfinished ; jump if (in+len <= in)
michael@0	80	add dl, al ; y += tx
michael@0	81	mov ebx, [rbp+rdx*8] ; ty = d[y]
michael@0	82	mov [rbp+rcx*8], ebx ; d[x] = ty
michael@0	83	add bl, al ; val = ty + tx
michael@0	84	mov [rbp+rdx*8], eax ; d[y] = tx
michael@0	85	inc cl ; x++ (NEXT ROUND)
michael@0	86	mov eax, [rbp+rcx*8] ; tx = d[x] (NEXT ROUND)
michael@0	87	mov r8b, [rbp+rbx*8] ; val = d[val]
michael@0	88	xor r8b, [rsi] ; xor 1 byte
michael@0	89	mov [rdi], r8b
michael@0	90	inc rsi ; in++
michael@0	91	inc rdi
michael@0	92	jmp @b
michael@0	93
michael@0	94	Lfinished:
michael@0	95	dec rcx ; x--
michael@0	96	mov [rbp-8], dl ; key->y = y
michael@0	97	mov [rbp-16], cl ; key->x = x
michael@0	98
michael@0	99	pop rdi
michael@0	100	pop rsi
michael@0	101	pop rbx
michael@0	102	pop rbp
michael@0	103	ret
michael@0	104
michael@0	105	ARCFOUR ENDP
michael@0	106
michael@0	107	END