xref: /linux/arch/arm/lib/memmove.S (revision 26fbb4c8c7c3ee9a4c3b4de555a8587b5a19154e)
1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 *  linux/arch/arm/lib/memmove.S
4 *
5 *  Author:	Nicolas Pitre
6 *  Created:	Sep 28, 2005
7 *  Copyright:	(C) MontaVista Software Inc.
8 */
9
10#include <linux/linkage.h>
11#include <asm/assembler.h>
12#include <asm/unwind.h>
13
14		.text
15
16/*
17 * Prototype: void *memmove(void *dest, const void *src, size_t n);
18 *
19 * Note:
20 *
21 * If the memory regions don't overlap, we simply branch to memcpy which is
22 * normally a bit faster. Otherwise the copy is done going downwards.  This
23 * is a transposition of the code from copy_template.S but with the copy
24 * occurring in the opposite direction.
25 */
26
27ENTRY(__memmove)
28WEAK(memmove)
29	UNWIND(	.fnstart			)
30
31		subs	ip, r0, r1
32		cmphi	r2, ip
33		bls	__memcpy
34
35		stmfd	sp!, {r0, r4, lr}
36	UNWIND(	.fnend				)
37
38	UNWIND(	.fnstart			)
39	UNWIND(	.save	{r0, r4, lr}		) @ in first stmfd block
40		add	r1, r1, r2
41		add	r0, r0, r2
42		subs	r2, r2, #4
43		blt	8f
44		ands	ip, r0, #3
45	PLD(	pld	[r1, #-4]		)
46		bne	9f
47		ands	ip, r1, #3
48		bne	10f
49
501:		subs	r2, r2, #(28)
51		stmfd	sp!, {r5 - r8}
52	UNWIND(	.fnend				)
53
54	UNWIND(	.fnstart			)
55	UNWIND(	.save	{r0, r4, lr}		)
56	UNWIND(	.save	{r5 - r8}		) @ in second stmfd block
57		blt	5f
58
59	CALGN(	ands	ip, r0, #31		)
60	CALGN(	sbcsne	r4, ip, r2		)  @ C is always set here
61	CALGN(	bcs	2f			)
62	CALGN(	adr	r4, 6f			)
63	CALGN(	subs	r2, r2, ip		)  @ C is set here
64	CALGN(	rsb	ip, ip, #32		)
65	CALGN(	add	pc, r4, ip		)
66
67	PLD(	pld	[r1, #-4]		)
682:	PLD(	subs	r2, r2, #96		)
69	PLD(	pld	[r1, #-32]		)
70	PLD(	blt	4f			)
71	PLD(	pld	[r1, #-64]		)
72	PLD(	pld	[r1, #-96]		)
73
743:	PLD(	pld	[r1, #-128]		)
754:		ldmdb	r1!, {r3, r4, r5, r6, r7, r8, ip, lr}
76		subs	r2, r2, #32
77		stmdb	r0!, {r3, r4, r5, r6, r7, r8, ip, lr}
78		bge	3b
79	PLD(	cmn	r2, #96			)
80	PLD(	bge	4b			)
81
825:		ands	ip, r2, #28
83		rsb	ip, ip, #32
84		addne	pc, pc, ip		@ C is always clear here
85		b	7f
866:		W(nop)
87		W(ldr)	r3, [r1, #-4]!
88		W(ldr)	r4, [r1, #-4]!
89		W(ldr)	r5, [r1, #-4]!
90		W(ldr)	r6, [r1, #-4]!
91		W(ldr)	r7, [r1, #-4]!
92		W(ldr)	r8, [r1, #-4]!
93		W(ldr)	lr, [r1, #-4]!
94
95		add	pc, pc, ip
96		nop
97		W(nop)
98		W(str)	r3, [r0, #-4]!
99		W(str)	r4, [r0, #-4]!
100		W(str)	r5, [r0, #-4]!
101		W(str)	r6, [r0, #-4]!
102		W(str)	r7, [r0, #-4]!
103		W(str)	r8, [r0, #-4]!
104		W(str)	lr, [r0, #-4]!
105
106	CALGN(	bcs	2b			)
107
1087:		ldmfd	sp!, {r5 - r8}
109	UNWIND(	.fnend				) @ end of second stmfd block
110
111	UNWIND(	.fnstart			)
112	UNWIND(	.save	{r0, r4, lr}		) @ still in first stmfd block
113
1148:		movs	r2, r2, lsl #31
115		ldrbne	r3, [r1, #-1]!
116		ldrbcs	r4, [r1, #-1]!
117		ldrbcs	ip, [r1, #-1]
118		strbne	r3, [r0, #-1]!
119		strbcs	r4, [r0, #-1]!
120		strbcs	ip, [r0, #-1]
121		ldmfd	sp!, {r0, r4, pc}
122
1239:		cmp	ip, #2
124		ldrbgt	r3, [r1, #-1]!
125		ldrbge	r4, [r1, #-1]!
126		ldrb	lr, [r1, #-1]!
127		strbgt	r3, [r0, #-1]!
128		strbge	r4, [r0, #-1]!
129		subs	r2, r2, ip
130		strb	lr, [r0, #-1]!
131		blt	8b
132		ands	ip, r1, #3
133		beq	1b
134
13510:		bic	r1, r1, #3
136		cmp	ip, #2
137		ldr	r3, [r1, #0]
138		beq	17f
139		blt	18f
140	UNWIND(	.fnend				)
141
142
143		.macro	backward_copy_shift push pull
144
145	UNWIND(	.fnstart			)
146	UNWIND(	.save	{r0, r4, lr}		) @ still in first stmfd block
147		subs	r2, r2, #28
148		blt	14f
149
150	CALGN(	ands	ip, r0, #31		)
151	CALGN(	sbcsne	r4, ip, r2		)  @ C is always set here
152	CALGN(	subcc	r2, r2, ip		)
153	CALGN(	bcc	15f			)
154
15511:		stmfd	sp!, {r5 - r9}
156	UNWIND(	.fnend				)
157
158	UNWIND(	.fnstart			)
159	UNWIND(	.save	{r0, r4, lr}		)
160	UNWIND(	.save	{r5 - r9}		) @ in new second stmfd block
161
162	PLD(	pld	[r1, #-4]		)
163	PLD(	subs	r2, r2, #96		)
164	PLD(	pld	[r1, #-32]		)
165	PLD(	blt	13f			)
166	PLD(	pld	[r1, #-64]		)
167	PLD(	pld	[r1, #-96]		)
168
16912:	PLD(	pld	[r1, #-128]		)
17013:		ldmdb   r1!, {r7, r8, r9, ip}
171		mov     lr, r3, lspush #\push
172		subs    r2, r2, #32
173		ldmdb   r1!, {r3, r4, r5, r6}
174		orr     lr, lr, ip, lspull #\pull
175		mov     ip, ip, lspush #\push
176		orr     ip, ip, r9, lspull #\pull
177		mov     r9, r9, lspush #\push
178		orr     r9, r9, r8, lspull #\pull
179		mov     r8, r8, lspush #\push
180		orr     r8, r8, r7, lspull #\pull
181		mov     r7, r7, lspush #\push
182		orr     r7, r7, r6, lspull #\pull
183		mov     r6, r6, lspush #\push
184		orr     r6, r6, r5, lspull #\pull
185		mov     r5, r5, lspush #\push
186		orr     r5, r5, r4, lspull #\pull
187		mov     r4, r4, lspush #\push
188		orr     r4, r4, r3, lspull #\pull
189		stmdb   r0!, {r4 - r9, ip, lr}
190		bge	12b
191	PLD(	cmn	r2, #96			)
192	PLD(	bge	13b			)
193
194		ldmfd	sp!, {r5 - r9}
195	UNWIND(	.fnend				) @ end of the second stmfd block
196
197	UNWIND(	.fnstart			)
198	UNWIND(	.save {r0, r4, lr}		) @ still in first stmfd block
199
20014:		ands	ip, r2, #28
201		beq	16f
202
20315:		mov     lr, r3, lspush #\push
204		ldr	r3, [r1, #-4]!
205		subs	ip, ip, #4
206		orr	lr, lr, r3, lspull #\pull
207		str	lr, [r0, #-4]!
208		bgt	15b
209	CALGN(	cmp	r2, #0			)
210	CALGN(	bge	11b			)
211
21216:		add	r1, r1, #(\pull / 8)
213		b	8b
214	UNWIND(	.fnend				)
215
216		.endm
217
218
219		backward_copy_shift	push=8	pull=24
220
22117:		backward_copy_shift	push=16	pull=16
222
22318:		backward_copy_shift	push=24	pull=8
224
225ENDPROC(memmove)
226ENDPROC(__memmove)
227