1/*	$OpenBSD: _memcpy.S,v 1.6 2016/08/06 19:16:09 guenther Exp $	*/
2/*	$NetBSD: _memcpy.S,v 1.4 2003/04/05 23:08:52 bjh21 Exp $	*/
3
4/*-
5 * Copyright (c) 1997 The NetBSD Foundation, Inc.
6 * All rights reserved.
7 *
8 * This code is derived from software contributed to The NetBSD Foundation
9 * by Neil A. Carson and Mark Brinicombe
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 *    notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 *    notice, this list of conditions and the following disclaimer in the
18 *    documentation and/or other materials provided with the distribution.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30 * POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <private/bionic_asm.h>
34
35  .syntax unified
36
37/*
38 * This is one fun bit of code ...
39 * Some easy listening music is suggested while trying to understand this
40 * code e.g. Iron Maiden
41 *
42 * For anyone attempting to understand it :
43 *
44 * The core code is implemented here with simple stubs for memcpy()
45 * memmove() and bcopy().
46 *
47 * All local labels are prefixed with Lmemcpy_
48 * Following the prefix a label starting f is used in the forward copy code
49 * while a label using b is used in the backwards copy code
50 * The source and destination addresses determine whether a forward or
51 * backward copy is performed.
52 * Separate bits of code are used to deal with the following situations
53 * for both the forward and backwards copy.
54 * unaligned source address
55 * unaligned destination address
56 * Separate copy routines are used to produce an optimised result for each
57 * of these cases.
58 * The copy code will use LDM/STM instructions to copy up to 32 bytes at
59 * a time where possible.
60 *
61 * Note: r12 (aka ip) can be trashed during the function along with
62 * r0-r3 although r0-r2 have defined uses i.e. src, dest, len through out.
63 * Additional registers are preserved prior to use i.e. r4, r5 & lr
64 *
65 * Apologies for the state of the comments ;-)
66 */
67
68ENTRY_PRIVATE(bsd_safe_memcpy)
69	/* Determine copy direction */
70	cmp	r1, r0
71	bcc	.Lmemcpy_backwards
72
73	moveq	r0, #0			/* Quick abort for len=0 */
74	moveq	pc, lr
75
76	stmdb	sp!, {r0, lr}		/* memcpy() returns dest addr */
77	subs	r2, r2, #4
78	blt	.Lmemcpy_fl4		/* less than 4 bytes */
79	ands	r12, r0, #3
80	bne	.Lmemcpy_fdestul	/* oh unaligned destination addr */
81	ands	r12, r1, #3
82	bne	.Lmemcpy_fsrcul		/* oh unaligned source addr */
83
84.Lmemcpy_ft8:
85	/* We have aligned source and destination */
86	subs	r2, r2, #8
87	blt	.Lmemcpy_fl12		/* less than 12 bytes (4 from above) */
88	subs	r2, r2, #0x14
89	blt	.Lmemcpy_fl32		/* less than 32 bytes (12 from above) */
90	stmdb	sp!, {r4}		/* borrow r4 */
91
92	/* blat 32 bytes at a time */
93	/* XXX for really big copies perhaps we should use more registers */
94.Lmemcpy_floop32:
95	ldmia	r1!, {r3, r4, r12, lr}
96	stmia	r0!, {r3, r4, r12, lr}
97	ldmia	r1!, {r3, r4, r12, lr}
98	stmia	r0!, {r3, r4, r12, lr}
99	subs	r2, r2, #0x20
100	bge	.Lmemcpy_floop32
101
102	cmn	r2, #0x10
103	ldmiage	r1!, {r3, r4, r12, lr}	/* blat a remaining 16 bytes */
104	stmiage	r0!, {r3, r4, r12, lr}
105	subge	r2, r2, #0x10
106	ldmia	sp!, {r4}		/* return r4 */
107
108.Lmemcpy_fl32:
109	adds	r2, r2, #0x14
110
111	/* blat 12 bytes at a time */
112.Lmemcpy_floop12:
113	ldmiage	r1!, {r3, r12, lr}
114	stmiage	r0!, {r3, r12, lr}
115	subsge	r2, r2, #0x0c
116	bge	.Lmemcpy_floop12
117
118.Lmemcpy_fl12:
119	adds	r2, r2, #8
120	blt	.Lmemcpy_fl4
121
122	subs	r2, r2, #4
123	ldrlt	r3, [r1], #4
124	strlt	r3, [r0], #4
125	ldmiage	r1!, {r3, r12}
126	stmiage	r0!, {r3, r12}
127	subge	r2, r2, #4
128
129.Lmemcpy_fl4:
130	/* less than 4 bytes to go */
131	adds	r2, r2, #4
132	ldmiaeq	sp!, {r0, pc}		/* done */
133
134	/* copy the crud byte at a time */
135	cmp	r2, #2
136	ldrb	r3, [r1], #1
137	strb	r3, [r0], #1
138	ldrbge	r3, [r1], #1
139	strbge	r3, [r0], #1
140	ldrbgt	r3, [r1], #1
141	strbgt	r3, [r0], #1
142	ldmia	sp!, {r0, pc}
143
144	/* erg - unaligned destination */
145.Lmemcpy_fdestul:
146	rsb	r12, r12, #4
147	cmp	r12, #2
148
149	/* align destination with byte copies */
150	ldrb	r3, [r1], #1
151	strb	r3, [r0], #1
152	ldrbge	r3, [r1], #1
153	strbge	r3, [r0], #1
154	ldrbgt	r3, [r1], #1
155	strbgt	r3, [r0], #1
156	subs	r2, r2, r12
157	blt	.Lmemcpy_fl4		/* less the 4 bytes */
158
159	ands	r12, r1, #3
160	beq	.Lmemcpy_ft8		/* we have an aligned source */
161
162	/* erg - unaligned source */
163	/* This is where it gets nasty ... */
164.Lmemcpy_fsrcul:
165	bic	r1, r1, #3
166	ldr	lr, [r1], #4
167	cmp	r12, #2
168	bgt	.Lmemcpy_fsrcul3
169	beq	.Lmemcpy_fsrcul2
170	cmp	r2, #0x0c
171	blt	.Lmemcpy_fsrcul1loop4
172	sub	r2, r2, #0x0c
173	stmdb	sp!, {r4, r5}
174
175.Lmemcpy_fsrcul1loop16:
176	mov	r3, lr, lsr #8
177	ldmia	r1!, {r4, r5, r12, lr}
178	orr	r3, r3, r4, lsl #24
179	mov	r4, r4, lsr #8
180	orr	r4, r4, r5, lsl #24
181	mov	r5, r5, lsr #8
182	orr	r5, r5, r12, lsl #24
183	mov	r12, r12, lsr #8
184	orr	r12, r12, lr, lsl #24
185	stmia	r0!, {r3-r5, r12}
186	subs	r2, r2, #0x10
187	bge	.Lmemcpy_fsrcul1loop16
188	ldmia	sp!, {r4, r5}
189	adds	r2, r2, #0x0c
190	blt	.Lmemcpy_fsrcul1l4
191
192.Lmemcpy_fsrcul1loop4:
193	mov	r12, lr, lsr #8
194	ldr	lr, [r1], #4
195	orr	r12, r12, lr, lsl #24
196	str	r12, [r0], #4
197	subs	r2, r2, #4
198	bge	.Lmemcpy_fsrcul1loop4
199
200.Lmemcpy_fsrcul1l4:
201	sub	r1, r1, #3
202	b	.Lmemcpy_fl4
203
204.Lmemcpy_fsrcul2:
205	cmp	r2, #0x0c
206	blt	.Lmemcpy_fsrcul2loop4
207	sub	r2, r2, #0x0c
208	stmdb	sp!, {r4, r5}
209
210.Lmemcpy_fsrcul2loop16:
211	mov	r3, lr, lsr #16
212	ldmia	r1!, {r4, r5, r12, lr}
213	orr	r3, r3, r4, lsl #16
214	mov	r4, r4, lsr #16
215	orr	r4, r4, r5, lsl #16
216	mov	r5, r5, lsr #16
217	orr	r5, r5, r12, lsl #16
218	mov	r12, r12, lsr #16
219	orr	r12, r12, lr, lsl #16
220	stmia	r0!, {r3-r5, r12}
221	subs	r2, r2, #0x10
222	bge	.Lmemcpy_fsrcul2loop16
223	ldmia	sp!, {r4, r5}
224	adds	r2, r2, #0x0c
225	blt	.Lmemcpy_fsrcul2l4
226
227.Lmemcpy_fsrcul2loop4:
228	mov	r12, lr, lsr #16
229	ldr	lr, [r1], #4
230	orr	r12, r12, lr, lsl #16
231	str	r12, [r0], #4
232	subs	r2, r2, #4
233	bge	.Lmemcpy_fsrcul2loop4
234
235.Lmemcpy_fsrcul2l4:
236	sub	r1, r1, #2
237	b	.Lmemcpy_fl4
238
239.Lmemcpy_fsrcul3:
240	cmp	r2, #0x0c
241	blt	.Lmemcpy_fsrcul3loop4
242	sub	r2, r2, #0x0c
243	stmdb	sp!, {r4, r5}
244
245.Lmemcpy_fsrcul3loop16:
246	mov	r3, lr, lsr #24
247	ldmia	r1!, {r4, r5, r12, lr}
248	orr	r3, r3, r4, lsl #8
249	mov	r4, r4, lsr #24
250	orr	r4, r4, r5, lsl #8
251	mov	r5, r5, lsr #24
252	orr	r5, r5, r12, lsl #8
253	mov	r12, r12, lsr #24
254	orr	r12, r12, lr, lsl #8
255	stmia	r0!, {r3-r5, r12}
256	subs	r2, r2, #0x10
257	bge	.Lmemcpy_fsrcul3loop16
258	ldmia	sp!, {r4, r5}
259	adds	r2, r2, #0x0c
260	blt	.Lmemcpy_fsrcul3l4
261
262.Lmemcpy_fsrcul3loop4:
263	mov	r12, lr, lsr #24
264	ldr	lr, [r1], #4
265	orr	r12, r12, lr, lsl #8
266	str	r12, [r0], #4
267	subs	r2, r2, #4
268	bge	.Lmemcpy_fsrcul3loop4
269
270.Lmemcpy_fsrcul3l4:
271	sub	r1, r1, #1
272	b	.Lmemcpy_fl4
273
274.Lmemcpy_backwards:
275	add	r1, r1, r2
276	add	r0, r0, r2
277	subs	r2, r2, #4
278	blt	.Lmemcpy_bl4		/* less than 4 bytes */
279	ands	r12, r0, #3
280	bne	.Lmemcpy_bdestul	/* oh unaligned destination addr */
281	ands	r12, r1, #3
282	bne	.Lmemcpy_bsrcul		/* oh unaligned source addr */
283
284.Lmemcpy_bt8:
285	/* We have aligned source and destination */
286	subs	r2, r2, #8
287	blt	.Lmemcpy_bl12		/* less than 12 bytes (4 from above) */
288	stmdb	sp!, {r4, lr}
289	subs	r2, r2, #0x14		/* less than 32 bytes (12 from above) */
290	blt	.Lmemcpy_bl32
291
292	/* blat 32 bytes at a time */
293	/* XXX for really big copies perhaps we should use more registers */
294.Lmemcpy_bloop32:
295	ldmdb	r1!, {r3, r4, r12, lr}
296	stmdb	r0!, {r3, r4, r12, lr}
297	ldmdb	r1!, {r3, r4, r12, lr}
298	stmdb	r0!, {r3, r4, r12, lr}
299	subs	r2, r2, #0x20
300	bge	.Lmemcpy_bloop32
301
302.Lmemcpy_bl32:
303	cmn	r2, #0x10
304	ldmdbge	r1!, {r3, r4, r12, lr}	/* blat a remaining 16 bytes */
305	stmdbge	r0!, {r3, r4, r12, lr}
306	subge	r2, r2, #0x10
307	adds	r2, r2, #0x14
308	ldmdbge	r1!, {r3, r12, lr}	/* blat a remaining 12 bytes */
309	stmdbge	r0!, {r3, r12, lr}
310	subge	r2, r2, #0x0c
311	ldmia	sp!, {r4, lr}
312
313.Lmemcpy_bl12:
314	adds	r2, r2, #8
315	blt	.Lmemcpy_bl4
316	subs	r2, r2, #4
317	ldrlt	r3, [r1, #-4]!
318	strlt	r3, [r0, #-4]!
319	ldmdbge	r1!, {r3, r12}
320	stmdbge	r0!, {r3, r12}
321	subge	r2, r2, #4
322
323.Lmemcpy_bl4:
324	/* less than 4 bytes to go */
325	adds	r2, r2, #4
326	moveq	pc, lr			/* done */
327
328	/* copy the crud byte at a time */
329	cmp	r2, #2
330	ldrb	r3, [r1, #-1]!
331	strb	r3, [r0, #-1]!
332	ldrbge	r3, [r1, #-1]!
333	strbge	r3, [r0, #-1]!
334	ldrbgt	r3, [r1, #-1]!
335	strbgt	r3, [r0, #-1]!
336	mov	pc, lr
337
338	/* erg - unaligned destination */
339.Lmemcpy_bdestul:
340	cmp	r12, #2
341
342	/* align destination with byte copies */
343	ldrb	r3, [r1, #-1]!
344	strb	r3, [r0, #-1]!
345	ldrbge	r3, [r1, #-1]!
346	strbge	r3, [r0, #-1]!
347	ldrbgt	r3, [r1, #-1]!
348	strbgt	r3, [r0, #-1]!
349	subs	r2, r2, r12
350	blt	.Lmemcpy_bl4		/* less than 4 bytes to go */
351	ands	r12, r1, #3
352	beq	.Lmemcpy_bt8		/* we have an aligned source */
353
354	/* erg - unaligned source */
355	/* This is where it gets nasty ... */
356.Lmemcpy_bsrcul:
357	bic	r1, r1, #3
358	ldr	r3, [r1, #0]
359	cmp	r12, #2
360	blt	.Lmemcpy_bsrcul1
361	beq	.Lmemcpy_bsrcul2
362	cmp	r2, #0x0c
363	blt	.Lmemcpy_bsrcul3loop4
364	sub	r2, r2, #0x0c
365	stmdb	sp!, {r4, r5, lr}
366
367.Lmemcpy_bsrcul3loop16:
368	mov	lr, r3, lsl #8
369	ldmdb	r1!, {r3-r5, r12}
370	orr	lr, lr, r12, lsr #24
371	mov	r12, r12, lsl #8
372	orr	r12, r12, r5, lsr #24
373	mov	r5, r5, lsl #8
374	orr	r5, r5, r4, lsr #24
375	mov	r4, r4, lsl #8
376	orr	r4, r4, r3, lsr #24
377	stmdb	r0!, {r4, r5, r12, lr}
378	subs	r2, r2, #0x10
379	bge	.Lmemcpy_bsrcul3loop16
380	ldmia	sp!, {r4, r5, lr}
381	adds	r2, r2, #0x0c
382	blt	.Lmemcpy_bsrcul3l4
383
384.Lmemcpy_bsrcul3loop4:
385	mov	r12, r3, lsl #8
386	ldr	r3, [r1, #-4]!
387	orr	r12, r12, r3, lsr #24
388	str	r12, [r0, #-4]!
389	subs	r2, r2, #4
390	bge	.Lmemcpy_bsrcul3loop4
391
392.Lmemcpy_bsrcul3l4:
393	add	r1, r1, #3
394	b	.Lmemcpy_bl4
395
396.Lmemcpy_bsrcul2:
397	cmp	r2, #0x0c
398	blt	.Lmemcpy_bsrcul2loop4
399	sub	r2, r2, #0x0c
400	stmdb	sp!, {r4, r5, lr}
401
402.Lmemcpy_bsrcul2loop16:
403	mov	lr, r3, lsl #16
404	ldmdb	r1!, {r3-r5, r12}
405	orr	lr, lr, r12, lsr #16
406	mov	r12, r12, lsl #16
407	orr	r12, r12, r5, lsr #16
408	mov	r5, r5, lsl #16
409	orr	r5, r5, r4, lsr #16
410	mov	r4, r4, lsl #16
411	orr	r4, r4, r3, lsr #16
412	stmdb	r0!, {r4, r5, r12, lr}
413	subs	r2, r2, #0x10
414	bge	.Lmemcpy_bsrcul2loop16
415	ldmia	sp!, {r4, r5, lr}
416	adds	r2, r2, #0x0c
417	blt	.Lmemcpy_bsrcul2l4
418
419.Lmemcpy_bsrcul2loop4:
420	mov	r12, r3, lsl #16
421	ldr	r3, [r1, #-4]!
422	orr	r12, r12, r3, lsr #16
423	str	r12, [r0, #-4]!
424	subs	r2, r2, #4
425	bge	.Lmemcpy_bsrcul2loop4
426
427.Lmemcpy_bsrcul2l4:
428	add	r1, r1, #2
429	b	.Lmemcpy_bl4
430
431.Lmemcpy_bsrcul1:
432	cmp	r2, #0x0c
433	blt	.Lmemcpy_bsrcul1loop4
434	sub	r2, r2, #0x0c
435	stmdb	sp!, {r4, r5, lr}
436
437.Lmemcpy_bsrcul1loop32:
438	mov	lr, r3, lsl #24
439	ldmdb	r1!, {r3-r5, r12}
440	orr	lr, lr, r12, lsr #8
441	mov	r12, r12, lsl #24
442	orr	r12, r12, r5, lsr #8
443	mov	r5, r5, lsl #24
444	orr	r5, r5, r4, lsr #8
445	mov	r4, r4, lsl #24
446	orr	r4, r4, r3, lsr #8
447	stmdb	r0!, {r4, r5, r12, lr}
448	subs	r2, r2, #0x10
449	bge	.Lmemcpy_bsrcul1loop32
450	ldmia	sp!, {r4, r5, lr}
451	adds	r2, r2, #0x0c
452	blt	.Lmemcpy_bsrcul1l4
453
454.Lmemcpy_bsrcul1loop4:
455	mov	r12, r3, lsl #24
456	ldr	r3, [r1, #-4]!
457	orr	r12, r12, r3, lsr #8
458	str	r12, [r0, #-4]!
459	subs	r2, r2, #4
460	bge	.Lmemcpy_bsrcul1loop4
461
462.Lmemcpy_bsrcul1l4:
463	add	r1, r1, #1
464	b	.Lmemcpy_bl4
465END(bsd_safe_memcpy)
466
467ENTRY(memmove_generic)
468        stmfd   sp!, {r0, lr}
469        bl      bsd_safe_memcpy
470        ldmfd   sp!, {r0, pc}
471END(memmove_generic)
472