xref: /openbmc/linux/arch/arm/mm/cache-v7.S (revision da30e0ac)
1/*
2 *  linux/arch/arm/mm/cache-v7.S
3 *
4 *  Copyright (C) 2001 Deep Blue Solutions Ltd.
5 *  Copyright (C) 2005 ARM Ltd.
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 *
11 *  This is the "shell" of the ARMv7 processor support.
12 */
13#include <linux/linkage.h>
14#include <linux/init.h>
15#include <asm/assembler.h>
16#include <asm/unwind.h>
17
18#include "proc-macros.S"
19
20/*
21 *	v7_flush_icache_all()
22 *
23 *	Flush the whole I-cache.
24 *
25 *	Registers:
26 *	r0 - set to 0
27 */
28ENTRY(v7_flush_icache_all)
29	mov	r0, #0
30	ALT_SMP(mcr	p15, 0, r0, c7, c1, 0)		@ invalidate I-cache inner shareable
31	ALT_UP(mcr	p15, 0, r0, c7, c5, 0)		@ I+BTB cache invalidate
32	mov	pc, lr
33ENDPROC(v7_flush_icache_all)
34
35/*
36 *	v7_flush_dcache_all()
37 *
38 *	Flush the whole D-cache.
39 *
40 *	Corrupted registers: r0-r7, r9-r11 (r6 only in Thumb mode)
41 *
42 *	- mm    - mm_struct describing address space
43 */
44ENTRY(v7_flush_dcache_all)
45	dmb					@ ensure ordering with previous memory accesses
46	mrc	p15, 1, r0, c0, c0, 1		@ read clidr
47	ands	r3, r0, #0x7000000		@ extract loc from clidr
48	mov	r3, r3, lsr #23			@ left align loc bit field
49	beq	finished			@ if loc is 0, then no need to clean
50	mov	r10, #0				@ start clean at cache level 0
51loop1:
52	add	r2, r10, r10, lsr #1		@ work out 3x current cache level
53	mov	r1, r0, lsr r2			@ extract cache type bits from clidr
54	and	r1, r1, #7			@ mask of the bits for current cache only
55	cmp	r1, #2				@ see what cache we have at this level
56	blt	skip				@ skip if no cache, or just i-cache
57	mcr	p15, 2, r10, c0, c0, 0		@ select current cache level in cssr
58	isb					@ isb to sych the new cssr&csidr
59	mrc	p15, 1, r1, c0, c0, 0		@ read the new csidr
60	and	r2, r1, #7			@ extract the length of the cache lines
61	add	r2, r2, #4			@ add 4 (line length offset)
62	ldr	r4, =0x3ff
63	ands	r4, r4, r1, lsr #3		@ find maximum number on the way size
64	clz	r5, r4				@ find bit position of way size increment
65	ldr	r7, =0x7fff
66	ands	r7, r7, r1, lsr #13		@ extract max number of the index size
67loop2:
68	mov	r9, r4				@ create working copy of max way size
69loop3:
70 ARM(	orr	r11, r10, r9, lsl r5	)	@ factor way and cache number into r11
71 THUMB(	lsl	r6, r9, r5		)
72 THUMB(	orr	r11, r10, r6		)	@ factor way and cache number into r11
73 ARM(	orr	r11, r11, r7, lsl r2	)	@ factor index number into r11
74 THUMB(	lsl	r6, r7, r2		)
75 THUMB(	orr	r11, r11, r6		)	@ factor index number into r11
76	mcr	p15, 0, r11, c7, c14, 2		@ clean & invalidate by set/way
77	subs	r9, r9, #1			@ decrement the way
78	bge	loop3
79	subs	r7, r7, #1			@ decrement the index
80	bge	loop2
81skip:
82	add	r10, r10, #2			@ increment cache number
83	cmp	r3, r10
84	bgt	loop1
85finished:
86	mov	r10, #0				@ swith back to cache level 0
87	mcr	p15, 2, r10, c0, c0, 0		@ select current cache level in cssr
88	dsb
89	isb
90	mov	pc, lr
91ENDPROC(v7_flush_dcache_all)
92
93/*
94 *	v7_flush_cache_all()
95 *
96 *	Flush the entire cache system.
97 *  The data cache flush is now achieved using atomic clean / invalidates
98 *  working outwards from L1 cache. This is done using Set/Way based cache
99 *  maintainance instructions.
100 *  The instruction cache can still be invalidated back to the point of
101 *  unification in a single instruction.
102 *
103 */
104ENTRY(v7_flush_kern_cache_all)
105 ARM(	stmfd	sp!, {r4-r5, r7, r9-r11, lr}	)
106 THUMB(	stmfd	sp!, {r4-r7, r9-r11, lr}	)
107	bl	v7_flush_dcache_all
108	mov	r0, #0
109	ALT_SMP(mcr	p15, 0, r0, c7, c1, 0)	@ invalidate I-cache inner shareable
110	ALT_UP(mcr	p15, 0, r0, c7, c5, 0)	@ I+BTB cache invalidate
111 ARM(	ldmfd	sp!, {r4-r5, r7, r9-r11, lr}	)
112 THUMB(	ldmfd	sp!, {r4-r7, r9-r11, lr}	)
113	mov	pc, lr
114ENDPROC(v7_flush_kern_cache_all)
115
116/*
117 *	v7_flush_cache_all()
118 *
119 *	Flush all TLB entries in a particular address space
120 *
121 *	- mm    - mm_struct describing address space
122 */
123ENTRY(v7_flush_user_cache_all)
124	/*FALLTHROUGH*/
125
126/*
127 *	v7_flush_cache_range(start, end, flags)
128 *
129 *	Flush a range of TLB entries in the specified address space.
130 *
131 *	- start - start address (may not be aligned)
132 *	- end   - end address (exclusive, may not be aligned)
133 *	- flags	- vm_area_struct flags describing address space
134 *
135 *	It is assumed that:
136 *	- we have a VIPT cache.
137 */
138ENTRY(v7_flush_user_cache_range)
139	mov	pc, lr
140ENDPROC(v7_flush_user_cache_all)
141ENDPROC(v7_flush_user_cache_range)
142
143/*
144 *	v7_coherent_kern_range(start,end)
145 *
146 *	Ensure that the I and D caches are coherent within specified
147 *	region.  This is typically used when code has been written to
148 *	a memory region, and will be executed.
149 *
150 *	- start   - virtual start address of region
151 *	- end     - virtual end address of region
152 *
153 *	It is assumed that:
154 *	- the Icache does not read data from the write buffer
155 */
156ENTRY(v7_coherent_kern_range)
157	/* FALLTHROUGH */
158
159/*
160 *	v7_coherent_user_range(start,end)
161 *
162 *	Ensure that the I and D caches are coherent within specified
163 *	region.  This is typically used when code has been written to
164 *	a memory region, and will be executed.
165 *
166 *	- start   - virtual start address of region
167 *	- end     - virtual end address of region
168 *
169 *	It is assumed that:
170 *	- the Icache does not read data from the write buffer
171 */
172ENTRY(v7_coherent_user_range)
173 UNWIND(.fnstart		)
174	dcache_line_size r2, r3
175	sub	r3, r2, #1
176	bic	r12, r0, r3
1771:
178 USER(	mcr	p15, 0, r12, c7, c11, 1	)	@ clean D line to the point of unification
179	add	r12, r12, r2
180	cmp	r12, r1
181	blo	1b
182	dsb
183	icache_line_size r2, r3
184	sub	r3, r2, #1
185	bic	r12, r0, r3
1862:
187 USER(	mcr	p15, 0, r12, c7, c5, 1	)	@ invalidate I line
188	add	r12, r12, r2
189	cmp	r12, r1
190	blo	2b
1913:
192	mov	r0, #0
193	ALT_SMP(mcr	p15, 0, r0, c7, c1, 6)	@ invalidate BTB Inner Shareable
194	ALT_UP(mcr	p15, 0, r0, c7, c5, 6)	@ invalidate BTB
195	dsb
196	isb
197	mov	pc, lr
198
199/*
200 * Fault handling for the cache operation above. If the virtual address in r0
201 * isn't mapped, just try the next page.
202 */
2039001:
204	mov	r12, r12, lsr #12
205	mov	r12, r12, lsl #12
206	add	r12, r12, #4096
207	b	3b
208 UNWIND(.fnend		)
209ENDPROC(v7_coherent_kern_range)
210ENDPROC(v7_coherent_user_range)
211
212/*
213 *	v7_flush_kern_dcache_area(void *addr, size_t size)
214 *
215 *	Ensure that the data held in the page kaddr is written back
216 *	to the page in question.
217 *
218 *	- addr	- kernel address
219 *	- size	- region size
220 */
221ENTRY(v7_flush_kern_dcache_area)
222	dcache_line_size r2, r3
223	add	r1, r0, r1
2241:
225	mcr	p15, 0, r0, c7, c14, 1		@ clean & invalidate D line / unified line
226	add	r0, r0, r2
227	cmp	r0, r1
228	blo	1b
229	dsb
230	mov	pc, lr
231ENDPROC(v7_flush_kern_dcache_area)
232
233/*
234 *	v7_dma_inv_range(start,end)
235 *
236 *	Invalidate the data cache within the specified region; we will
237 *	be performing a DMA operation in this region and we want to
238 *	purge old data in the cache.
239 *
240 *	- start   - virtual start address of region
241 *	- end     - virtual end address of region
242 */
243v7_dma_inv_range:
244	dcache_line_size r2, r3
245	sub	r3, r2, #1
246	tst	r0, r3
247	bic	r0, r0, r3
248	mcrne	p15, 0, r0, c7, c14, 1		@ clean & invalidate D / U line
249
250	tst	r1, r3
251	bic	r1, r1, r3
252	mcrne	p15, 0, r1, c7, c14, 1		@ clean & invalidate D / U line
2531:
254	mcr	p15, 0, r0, c7, c6, 1		@ invalidate D / U line
255	add	r0, r0, r2
256	cmp	r0, r1
257	blo	1b
258	dsb
259	mov	pc, lr
260ENDPROC(v7_dma_inv_range)
261
262/*
263 *	v7_dma_clean_range(start,end)
264 *	- start   - virtual start address of region
265 *	- end     - virtual end address of region
266 */
267v7_dma_clean_range:
268	dcache_line_size r2, r3
269	sub	r3, r2, #1
270	bic	r0, r0, r3
2711:
272	mcr	p15, 0, r0, c7, c10, 1		@ clean D / U line
273	add	r0, r0, r2
274	cmp	r0, r1
275	blo	1b
276	dsb
277	mov	pc, lr
278ENDPROC(v7_dma_clean_range)
279
280/*
281 *	v7_dma_flush_range(start,end)
282 *	- start   - virtual start address of region
283 *	- end     - virtual end address of region
284 */
285ENTRY(v7_dma_flush_range)
286	dcache_line_size r2, r3
287	sub	r3, r2, #1
288	bic	r0, r0, r3
2891:
290	mcr	p15, 0, r0, c7, c14, 1		@ clean & invalidate D / U line
291	add	r0, r0, r2
292	cmp	r0, r1
293	blo	1b
294	dsb
295	mov	pc, lr
296ENDPROC(v7_dma_flush_range)
297
298/*
299 *	dma_map_area(start, size, dir)
300 *	- start	- kernel virtual start address
301 *	- size	- size of region
302 *	- dir	- DMA direction
303 */
304ENTRY(v7_dma_map_area)
305	add	r1, r1, r0
306	teq	r2, #DMA_FROM_DEVICE
307	beq	v7_dma_inv_range
308	b	v7_dma_clean_range
309ENDPROC(v7_dma_map_area)
310
311/*
312 *	dma_unmap_area(start, size, dir)
313 *	- start	- kernel virtual start address
314 *	- size	- size of region
315 *	- dir	- DMA direction
316 */
317ENTRY(v7_dma_unmap_area)
318	add	r1, r1, r0
319	teq	r2, #DMA_TO_DEVICE
320	bne	v7_dma_inv_range
321	mov	pc, lr
322ENDPROC(v7_dma_unmap_area)
323
324	__INITDATA
325
326	.type	v7_cache_fns, #object
327ENTRY(v7_cache_fns)
328	.long	v7_flush_icache_all
329	.long	v7_flush_kern_cache_all
330	.long	v7_flush_user_cache_all
331	.long	v7_flush_user_cache_range
332	.long	v7_coherent_kern_range
333	.long	v7_coherent_user_range
334	.long	v7_flush_kern_dcache_area
335	.long	v7_dma_map_area
336	.long	v7_dma_unmap_area
337	.long	v7_dma_flush_range
338	.size	v7_cache_fns, . - v7_cache_fns
339