1/*
2 *  linux/arch/arm/lib/csumpartial.S
3 *
4 *  Copyright (C) 1995-1998 Russell King
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 */
10#include <linux/linkage.h>
11#include <asm/assembler.h>
12
13		.text
14
15/*
16 * Function: __u32 csum_partial(const char *src, int len, __u32 sum)
17 * Params  : r0 = buffer, r1 = len, r2 = checksum
18 * Returns : r0 = new checksum
19 */
20
21buf	.req	r0
22len	.req	r1
23sum	.req	r2
24td0	.req	r3
25td1	.req	r4	@ save before use
26td2	.req	r5	@ save before use
27td3	.req	lr
28
29.zero:		mov	r0, sum
30		add	sp, sp, #4
31		ldr	pc, [sp], #4
32
33		/*
34		 * Handle 0 to 7 bytes, with any alignment of source and
35		 * destination pointers.  Note that when we get here, C = 0
36		 */
37.less8:		teq	len, #0			@ check for zero count
38		beq	.zero
39
40		/* we must have at least one byte. */
41		tst	buf, #1			@ odd address?
42		ldrneb	td0, [buf], #1
43		subne	len, len, #1
44		adcnes	sum, sum, td0, lsl #8
45
46.less4:		tst	len, #6
47		beq	.less8_byte
48
49		/* we are now half-word aligned */
50
51.less8_wordlp:
52#ifdef __ARM_ARCH_4__
53		ldrh	td0, [buf], #2
54		sub	len, len, #2
55#else
56		ldrb	td0, [buf], #1
57		ldrb	td3, [buf], #1
58		sub	len, len, #2
59		orr	td0, td0, td3, lsl #8
60#endif
61		adcs	sum, sum, td0
62		tst	len, #6
63		bne	.less8_wordlp
64
65.less8_byte:	tst	len, #1			@ odd number of bytes
66		ldrneb	td0, [buf], #1		@ include last byte
67		adcnes	sum, sum, td0		@ update checksum
68
69.done:		adc	r0, sum, #0		@ collect up the last carry
70		ldr	td0, [sp], #4
71		tst	td0, #1			@ check buffer alignment
72		movne	td0, r0, lsl #8		@ rotate checksum by 8 bits
73		orrne	r0, td0, r0, lsr #24
74		ldr	pc, [sp], #4		@ return
75
76.not_aligned:	tst	buf, #1			@ odd address
77		ldrneb	td0, [buf], #1		@ make even
78		subne	len, len, #1
79		adcnes	sum, sum, td0, lsl #8	@ update checksum
80
81		tst	buf, #2			@ 32-bit aligned?
82#ifdef __ARM_ARCH_4__
83		ldrneh	td0, [buf], #2		@ make 32-bit aligned
84		subne	len, len, #2
85#else
86		ldrneb	td0, [buf], #1
87		ldrneb	ip, [buf], #1
88		subne	len, len, #2
89		orrne	td0, td0, ip, lsl #8
90#endif
91		adcnes	sum, sum, td0		@ update checksum
92		mov	pc, lr
93
94ENTRY(csum_partial)
95		stmfd	sp!, {buf, lr}
96		cmp	len, #8			@ Ensure that we have at least
97		blo	.less8			@ 8 bytes to copy.
98
99		adds	sum, sum, #0		@ C = 0
100		tst	buf, #3			@ Test destination alignment
101		blne	.not_aligned		@ aligh destination, return here
102
1031:		bics	ip, len, #31
104		beq	3f
105
106		stmfd	sp!, {r4 - r5}
1072:		ldmia	buf!, {td0, td1, td2, td3}
108		adcs	sum, sum, td0
109		adcs	sum, sum, td1
110		adcs	sum, sum, td2
111		adcs	sum, sum, td3
112		ldmia	buf!, {td0, td1, td2, td3}
113		adcs	sum, sum, td0
114		adcs	sum, sum, td1
115		adcs	sum, sum, td2
116		adcs	sum, sum, td3
117		sub	ip, ip, #32
118		teq	ip, #0
119		bne	2b
120		ldmfd	sp!, {r4 - r5}
121
1223:		tst	len, #0x1c		@ should not change C
123		beq	.less4
124
1254:		ldr	td0, [buf], #4
126		sub	len, len, #4
127		adcs	sum, sum, td0
128		tst	len, #0x1c
129		bne	4b
130		b	.less4
131