xref: /openbmc/linux/arch/arm/mm/proc-feroceon.S (revision aeb64ff3)
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 *  linux/arch/arm/mm/proc-feroceon.S: MMU functions for Feroceon
4 *
5 *  Heavily based on proc-arm926.S
6 *  Maintainer: Assaf Hoffman <hoffman@marvell.com>
7 */
8
9#include <linux/linkage.h>
10#include <linux/init.h>
11#include <asm/assembler.h>
12#include <asm/hwcap.h>
13#include <asm/pgtable-hwdef.h>
14#include <asm/pgtable.h>
15#include <asm/page.h>
16#include <asm/ptrace.h>
17#include "proc-macros.S"
18
19/*
20 * This is the maximum size of an area which will be invalidated
21 * using the single invalidate entry instructions.  Anything larger
22 * than this, and we go for the whole cache.
23 *
24 * This value should be chosen such that we choose the cheapest
25 * alternative.
26 */
27#define CACHE_DLIMIT	16384
28
29/*
30 * the cache line size of the I and D cache
31 */
32#define CACHE_DLINESIZE	32
33
34	.bss
35	.align 3
36__cache_params_loc:
37	.space	8
38
39	.text
40__cache_params:
41	.word	__cache_params_loc
42
43/*
44 * cpu_feroceon_proc_init()
45 */
46ENTRY(cpu_feroceon_proc_init)
47	mrc	p15, 0, r0, c0, c0, 1		@ read cache type register
48	ldr	r1, __cache_params
49	mov	r2, #(16 << 5)
50	tst	r0, #(1 << 16)			@ get way
51	mov	r0, r0, lsr #18			@ get cache size order
52	movne	r3, #((4 - 1) << 30)		@ 4-way
53	and	r0, r0, #0xf
54	moveq	r3, #0				@ 1-way
55	mov	r2, r2, lsl r0			@ actual cache size
56	movne	r2, r2, lsr #2			@ turned into # of sets
57	sub	r2, r2, #(1 << 5)
58	stmia	r1, {r2, r3}
59	ret	lr
60
61/*
62 * cpu_feroceon_proc_fin()
63 */
64ENTRY(cpu_feroceon_proc_fin)
65#if defined(CONFIG_CACHE_FEROCEON_L2) && \
66	!defined(CONFIG_CACHE_FEROCEON_L2_WRITETHROUGH)
67	mov	r0, #0
68	mcr	p15, 1, r0, c15, c9, 0		@ clean L2
69	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
70#endif
71
72	mrc	p15, 0, r0, c1, c0, 0		@ ctrl register
73	bic	r0, r0, #0x1000			@ ...i............
74	bic	r0, r0, #0x000e			@ ............wca.
75	mcr	p15, 0, r0, c1, c0, 0		@ disable caches
76	ret	lr
77
78/*
79 * cpu_feroceon_reset(loc)
80 *
81 * Perform a soft reset of the system.  Put the CPU into the
82 * same state as it would be if it had been reset, and branch
83 * to what would be the reset vector.
84 *
85 * loc: location to jump to for soft reset
86 */
87	.align	5
88	.pushsection	.idmap.text, "ax"
89ENTRY(cpu_feroceon_reset)
90	mov	ip, #0
91	mcr	p15, 0, ip, c7, c7, 0		@ invalidate I,D caches
92	mcr	p15, 0, ip, c7, c10, 4		@ drain WB
93#ifdef CONFIG_MMU
94	mcr	p15, 0, ip, c8, c7, 0		@ invalidate I & D TLBs
95#endif
96	mrc	p15, 0, ip, c1, c0, 0		@ ctrl register
97	bic	ip, ip, #0x000f			@ ............wcam
98	bic	ip, ip, #0x1100			@ ...i...s........
99	mcr	p15, 0, ip, c1, c0, 0		@ ctrl register
100	ret	r0
101ENDPROC(cpu_feroceon_reset)
102	.popsection
103
104/*
105 * cpu_feroceon_do_idle()
106 *
107 * Called with IRQs disabled
108 */
109	.align	5
110ENTRY(cpu_feroceon_do_idle)
111	mov	r0, #0
112	mcr	p15, 0, r0, c7, c10, 4		@ Drain write buffer
113	mcr	p15, 0, r0, c7, c0, 4		@ Wait for interrupt
114	ret	lr
115
116/*
117 *	flush_icache_all()
118 *
119 *	Unconditionally clean and invalidate the entire icache.
120 */
121ENTRY(feroceon_flush_icache_all)
122	mov	r0, #0
123	mcr	p15, 0, r0, c7, c5, 0		@ invalidate I cache
124	ret	lr
125ENDPROC(feroceon_flush_icache_all)
126
127/*
128 *	flush_user_cache_all()
129 *
130 *	Clean and invalidate all cache entries in a particular
131 *	address space.
132 */
133	.align	5
134ENTRY(feroceon_flush_user_cache_all)
135	/* FALLTHROUGH */
136
137/*
138 *	flush_kern_cache_all()
139 *
140 *	Clean and invalidate the entire cache.
141 */
142ENTRY(feroceon_flush_kern_cache_all)
143	mov	r2, #VM_EXEC
144
145__flush_whole_cache:
146	ldr	r1, __cache_params
147	ldmia	r1, {r1, r3}
1481:	orr	ip, r1, r3
1492:	mcr	p15, 0, ip, c7, c14, 2		@ clean + invalidate D set/way
150	subs	ip, ip, #(1 << 30)		@ next way
151	bcs	2b
152	subs	r1, r1, #(1 << 5)		@ next set
153	bcs	1b
154
155	tst	r2, #VM_EXEC
156	mov	ip, #0
157	mcrne	p15, 0, ip, c7, c5, 0		@ invalidate I cache
158	mcrne	p15, 0, ip, c7, c10, 4		@ drain WB
159	ret	lr
160
161/*
162 *	flush_user_cache_range(start, end, flags)
163 *
164 *	Clean and invalidate a range of cache entries in the
165 *	specified address range.
166 *
167 *	- start	- start address (inclusive)
168 *	- end	- end address (exclusive)
169 *	- flags	- vm_flags describing address space
170 */
171	.align	5
172ENTRY(feroceon_flush_user_cache_range)
173	sub	r3, r1, r0			@ calculate total size
174	cmp	r3, #CACHE_DLIMIT
175	bgt	__flush_whole_cache
1761:	tst	r2, #VM_EXEC
177	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D entry
178	mcrne	p15, 0, r0, c7, c5, 1		@ invalidate I entry
179	add	r0, r0, #CACHE_DLINESIZE
180	mcr	p15, 0, r0, c7, c14, 1		@ clean and invalidate D entry
181	mcrne	p15, 0, r0, c7, c5, 1		@ invalidate I entry
182	add	r0, r0, #CACHE_DLINESIZE
183	cmp	r0, r1
184	blo	1b
185	tst	r2, #VM_EXEC
186	mov	ip, #0
187	mcrne	p15, 0, ip, c7, c10, 4		@ drain WB
188	ret	lr
189
190/*
191 *	coherent_kern_range(start, end)
192 *
193 *	Ensure coherency between the Icache and the Dcache in the
194 *	region described by start, end.  If you have non-snooping
195 *	Harvard caches, you need to implement this function.
196 *
197 *	- start	- virtual start address
198 *	- end	- virtual end address
199 */
200	.align	5
201ENTRY(feroceon_coherent_kern_range)
202	/* FALLTHROUGH */
203
204/*
205 *	coherent_user_range(start, end)
206 *
207 *	Ensure coherency between the Icache and the Dcache in the
208 *	region described by start, end.  If you have non-snooping
209 *	Harvard caches, you need to implement this function.
210 *
211 *	- start	- virtual start address
212 *	- end	- virtual end address
213 */
214ENTRY(feroceon_coherent_user_range)
215	bic	r0, r0, #CACHE_DLINESIZE - 1
2161:	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
217	mcr	p15, 0, r0, c7, c5, 1		@ invalidate I entry
218	add	r0, r0, #CACHE_DLINESIZE
219	cmp	r0, r1
220	blo	1b
221	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
222	mov	r0, #0
223	ret	lr
224
225/*
226 *	flush_kern_dcache_area(void *addr, size_t size)
227 *
228 *	Ensure no D cache aliasing occurs, either with itself or
229 *	the I cache
230 *
231 *	- addr	- kernel address
232 *	- size	- region size
233 */
234	.align	5
235ENTRY(feroceon_flush_kern_dcache_area)
236	add	r1, r0, r1
2371:	mcr	p15, 0, r0, c7, c14, 1		@ clean+invalidate D entry
238	add	r0, r0, #CACHE_DLINESIZE
239	cmp	r0, r1
240	blo	1b
241	mov	r0, #0
242	mcr	p15, 0, r0, c7, c5, 0		@ invalidate I cache
243	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
244	ret	lr
245
246	.align	5
247ENTRY(feroceon_range_flush_kern_dcache_area)
248	mrs	r2, cpsr
249	add	r1, r0, #PAGE_SZ - CACHE_DLINESIZE	@ top addr is inclusive
250	orr	r3, r2, #PSR_I_BIT
251	msr	cpsr_c, r3			@ disable interrupts
252	mcr	p15, 5, r0, c15, c15, 0		@ D clean/inv range start
253	mcr	p15, 5, r1, c15, c15, 1		@ D clean/inv range top
254	msr	cpsr_c, r2			@ restore interrupts
255	mov	r0, #0
256	mcr	p15, 0, r0, c7, c5, 0		@ invalidate I cache
257	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
258	ret	lr
259
260/*
261 *	dma_inv_range(start, end)
262 *
263 *	Invalidate (discard) the specified virtual address range.
264 *	May not write back any entries.  If 'start' or 'end'
265 *	are not cache line aligned, those lines must be written
266 *	back.
267 *
268 *	- start	- virtual start address
269 *	- end	- virtual end address
270 *
271 * (same as v4wb)
272 */
273	.align	5
274feroceon_dma_inv_range:
275	tst	r0, #CACHE_DLINESIZE - 1
276	bic	r0, r0, #CACHE_DLINESIZE - 1
277	mcrne	p15, 0, r0, c7, c10, 1		@ clean D entry
278	tst	r1, #CACHE_DLINESIZE - 1
279	mcrne	p15, 0, r1, c7, c10, 1		@ clean D entry
2801:	mcr	p15, 0, r0, c7, c6, 1		@ invalidate D entry
281	add	r0, r0, #CACHE_DLINESIZE
282	cmp	r0, r1
283	blo	1b
284	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
285	ret	lr
286
287	.align	5
288feroceon_range_dma_inv_range:
289	mrs	r2, cpsr
290	tst	r0, #CACHE_DLINESIZE - 1
291	mcrne	p15, 0, r0, c7, c10, 1		@ clean D entry
292	tst	r1, #CACHE_DLINESIZE - 1
293	mcrne	p15, 0, r1, c7, c10, 1		@ clean D entry
294	cmp	r1, r0
295	subne	r1, r1, #1			@ top address is inclusive
296	orr	r3, r2, #PSR_I_BIT
297	msr	cpsr_c, r3			@ disable interrupts
298	mcr	p15, 5, r0, c15, c14, 0		@ D inv range start
299	mcr	p15, 5, r1, c15, c14, 1		@ D inv range top
300	msr	cpsr_c, r2			@ restore interrupts
301	ret	lr
302
303/*
304 *	dma_clean_range(start, end)
305 *
306 *	Clean the specified virtual address range.
307 *
308 *	- start	- virtual start address
309 *	- end	- virtual end address
310 *
311 * (same as v4wb)
312 */
313	.align	5
314feroceon_dma_clean_range:
315	bic	r0, r0, #CACHE_DLINESIZE - 1
3161:	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
317	add	r0, r0, #CACHE_DLINESIZE
318	cmp	r0, r1
319	blo	1b
320	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
321	ret	lr
322
323	.align	5
324feroceon_range_dma_clean_range:
325	mrs	r2, cpsr
326	cmp	r1, r0
327	subne	r1, r1, #1			@ top address is inclusive
328	orr	r3, r2, #PSR_I_BIT
329	msr	cpsr_c, r3			@ disable interrupts
330	mcr	p15, 5, r0, c15, c13, 0		@ D clean range start
331	mcr	p15, 5, r1, c15, c13, 1		@ D clean range top
332	msr	cpsr_c, r2			@ restore interrupts
333	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
334	ret	lr
335
336/*
337 *	dma_flush_range(start, end)
338 *
339 *	Clean and invalidate the specified virtual address range.
340 *
341 *	- start	- virtual start address
342 *	- end	- virtual end address
343 */
344	.align	5
345ENTRY(feroceon_dma_flush_range)
346	bic	r0, r0, #CACHE_DLINESIZE - 1
3471:	mcr	p15, 0, r0, c7, c14, 1		@ clean+invalidate D entry
348	add	r0, r0, #CACHE_DLINESIZE
349	cmp	r0, r1
350	blo	1b
351	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
352	ret	lr
353
354	.align	5
355ENTRY(feroceon_range_dma_flush_range)
356	mrs	r2, cpsr
357	cmp	r1, r0
358	subne	r1, r1, #1			@ top address is inclusive
359	orr	r3, r2, #PSR_I_BIT
360	msr	cpsr_c, r3			@ disable interrupts
361	mcr	p15, 5, r0, c15, c15, 0		@ D clean/inv range start
362	mcr	p15, 5, r1, c15, c15, 1		@ D clean/inv range top
363	msr	cpsr_c, r2			@ restore interrupts
364	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
365	ret	lr
366
367/*
368 *	dma_map_area(start, size, dir)
369 *	- start	- kernel virtual start address
370 *	- size	- size of region
371 *	- dir	- DMA direction
372 */
373ENTRY(feroceon_dma_map_area)
374	add	r1, r1, r0
375	cmp	r2, #DMA_TO_DEVICE
376	beq	feroceon_dma_clean_range
377	bcs	feroceon_dma_inv_range
378	b	feroceon_dma_flush_range
379ENDPROC(feroceon_dma_map_area)
380
381/*
382 *	dma_map_area(start, size, dir)
383 *	- start	- kernel virtual start address
384 *	- size	- size of region
385 *	- dir	- DMA direction
386 */
387ENTRY(feroceon_range_dma_map_area)
388	add	r1, r1, r0
389	cmp	r2, #DMA_TO_DEVICE
390	beq	feroceon_range_dma_clean_range
391	bcs	feroceon_range_dma_inv_range
392	b	feroceon_range_dma_flush_range
393ENDPROC(feroceon_range_dma_map_area)
394
395/*
396 *	dma_unmap_area(start, size, dir)
397 *	- start	- kernel virtual start address
398 *	- size	- size of region
399 *	- dir	- DMA direction
400 */
401ENTRY(feroceon_dma_unmap_area)
402	ret	lr
403ENDPROC(feroceon_dma_unmap_area)
404
405	.globl	feroceon_flush_kern_cache_louis
406	.equ	feroceon_flush_kern_cache_louis, feroceon_flush_kern_cache_all
407
408	@ define struct cpu_cache_fns (see <asm/cacheflush.h> and proc-macros.S)
409	define_cache_functions feroceon
410
411.macro range_alias basename
412	.globl feroceon_range_\basename
413	.type feroceon_range_\basename , %function
414	.equ feroceon_range_\basename , feroceon_\basename
415.endm
416
417/*
418 * Most of the cache functions are unchanged for this case.
419 * Export suitable alias symbols for the unchanged functions:
420 */
421	range_alias flush_icache_all
422	range_alias flush_user_cache_all
423	range_alias flush_kern_cache_all
424	range_alias flush_kern_cache_louis
425	range_alias flush_user_cache_range
426	range_alias coherent_kern_range
427	range_alias coherent_user_range
428	range_alias dma_unmap_area
429
430	define_cache_functions feroceon_range
431
432	.align	5
433ENTRY(cpu_feroceon_dcache_clean_area)
434#if defined(CONFIG_CACHE_FEROCEON_L2) && \
435	!defined(CONFIG_CACHE_FEROCEON_L2_WRITETHROUGH)
436	mov	r2, r0
437	mov	r3, r1
438#endif
4391:	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
440	add	r0, r0, #CACHE_DLINESIZE
441	subs	r1, r1, #CACHE_DLINESIZE
442	bhi	1b
443#if defined(CONFIG_CACHE_FEROCEON_L2) && \
444	!defined(CONFIG_CACHE_FEROCEON_L2_WRITETHROUGH)
4451:	mcr	p15, 1, r2, c15, c9, 1		@ clean L2 entry
446	add	r2, r2, #CACHE_DLINESIZE
447	subs	r3, r3, #CACHE_DLINESIZE
448	bhi	1b
449#endif
450	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
451	ret	lr
452
453/* =============================== PageTable ============================== */
454
455/*
456 * cpu_feroceon_switch_mm(pgd)
457 *
458 * Set the translation base pointer to be as described by pgd.
459 *
460 * pgd: new page tables
461 */
462	.align	5
463ENTRY(cpu_feroceon_switch_mm)
464#ifdef CONFIG_MMU
465	/*
466	 * Note: we wish to call __flush_whole_cache but we need to preserve
467	 * lr to do so.  The only way without touching main memory is to
468	 * use r2 which is normally used to test the VM_EXEC flag, and
469	 * compensate locally for the skipped ops if it is not set.
470	 */
471	mov	r2, lr				@ abuse r2 to preserve lr
472	bl	__flush_whole_cache
473	@ if r2 contains the VM_EXEC bit then the next 2 ops are done already
474	tst	r2, #VM_EXEC
475	mcreq	p15, 0, ip, c7, c5, 0		@ invalidate I cache
476	mcreq	p15, 0, ip, c7, c10, 4		@ drain WB
477
478	mcr	p15, 0, r0, c2, c0, 0		@ load page table pointer
479	mcr	p15, 0, ip, c8, c7, 0		@ invalidate I & D TLBs
480	ret	r2
481#else
482	ret	lr
483#endif
484
485/*
486 * cpu_feroceon_set_pte_ext(ptep, pte, ext)
487 *
488 * Set a PTE and flush it out
489 */
490	.align	5
491ENTRY(cpu_feroceon_set_pte_ext)
492#ifdef CONFIG_MMU
493	armv3_set_pte_ext wc_disable=0
494	mov	r0, r0
495	mcr	p15, 0, r0, c7, c10, 1		@ clean D entry
496#if defined(CONFIG_CACHE_FEROCEON_L2) && \
497	!defined(CONFIG_CACHE_FEROCEON_L2_WRITETHROUGH)
498	mcr	p15, 1, r0, c15, c9, 1		@ clean L2 entry
499#endif
500	mcr	p15, 0, r0, c7, c10, 4		@ drain WB
501#endif
502	ret	lr
503
504/* Suspend/resume support: taken from arch/arm/mm/proc-arm926.S */
505.globl	cpu_feroceon_suspend_size
506.equ	cpu_feroceon_suspend_size, 4 * 3
507#ifdef CONFIG_ARM_CPU_SUSPEND
508ENTRY(cpu_feroceon_do_suspend)
509	stmfd	sp!, {r4 - r6, lr}
510	mrc	p15, 0, r4, c13, c0, 0	@ PID
511	mrc	p15, 0, r5, c3, c0, 0	@ Domain ID
512	mrc	p15, 0, r6, c1, c0, 0	@ Control register
513	stmia	r0, {r4 - r6}
514	ldmfd	sp!, {r4 - r6, pc}
515ENDPROC(cpu_feroceon_do_suspend)
516
517ENTRY(cpu_feroceon_do_resume)
518	mov	ip, #0
519	mcr	p15, 0, ip, c8, c7, 0	@ invalidate I+D TLBs
520	mcr	p15, 0, ip, c7, c7, 0	@ invalidate I+D caches
521	ldmia	r0, {r4 - r6}
522	mcr	p15, 0, r4, c13, c0, 0	@ PID
523	mcr	p15, 0, r5, c3, c0, 0	@ Domain ID
524	mcr	p15, 0, r1, c2, c0, 0	@ TTB address
525	mov	r0, r6			@ control register
526	b	cpu_resume_mmu
527ENDPROC(cpu_feroceon_do_resume)
528#endif
529
530	.type	__feroceon_setup, #function
531__feroceon_setup:
532	mov	r0, #0
533	mcr	p15, 0, r0, c7, c7		@ invalidate I,D caches on v4
534	mcr	p15, 0, r0, c7, c10, 4		@ drain write buffer on v4
535#ifdef CONFIG_MMU
536	mcr	p15, 0, r0, c8, c7		@ invalidate I,D TLBs on v4
537#endif
538
539	adr	r5, feroceon_crval
540	ldmia	r5, {r5, r6}
541	mrc	p15, 0, r0, c1, c0		@ get control register v4
542	bic	r0, r0, r5
543	orr	r0, r0, r6
544	ret	lr
545	.size	__feroceon_setup, . - __feroceon_setup
546
547	/*
548	 *      B
549	 *  R   P
550	 * .RVI UFRS BLDP WCAM
551	 * .011 .001 ..11 0101
552	 *
553	 */
554	.type	feroceon_crval, #object
555feroceon_crval:
556	crval	clear=0x0000773f, mmuset=0x00003135, ucset=0x00001134
557
558	__INITDATA
559
560	@ define struct processor (see <asm/proc-fns.h> and proc-macros.S)
561	define_processor_functions feroceon, dabort=v5t_early_abort, pabort=legacy_pabort
562
563	.section ".rodata"
564
565	string	cpu_arch_name, "armv5te"
566	string	cpu_elf_name, "v5"
567	string	cpu_feroceon_name, "Feroceon"
568	string	cpu_88fr531_name, "Feroceon 88FR531-vd"
569	string	cpu_88fr571_name, "Feroceon 88FR571-vd"
570	string	cpu_88fr131_name, "Feroceon 88FR131"
571
572	.align
573
574	.section ".proc.info.init", "a"
575
576.macro feroceon_proc_info name:req, cpu_val:req, cpu_mask:req, cpu_name:req, cache:req
577	.type	__\name\()_proc_info,#object
578__\name\()_proc_info:
579	.long	\cpu_val
580	.long	\cpu_mask
581	.long	PMD_TYPE_SECT | \
582		PMD_SECT_BUFFERABLE | \
583		PMD_SECT_CACHEABLE | \
584		PMD_BIT4 | \
585		PMD_SECT_AP_WRITE | \
586		PMD_SECT_AP_READ
587	.long	PMD_TYPE_SECT | \
588		PMD_BIT4 | \
589		PMD_SECT_AP_WRITE | \
590		PMD_SECT_AP_READ
591	initfn	__feroceon_setup, __\name\()_proc_info
592	.long	cpu_arch_name
593	.long	cpu_elf_name
594	.long	HWCAP_SWP|HWCAP_HALF|HWCAP_THUMB|HWCAP_FAST_MULT|HWCAP_EDSP
595	.long	\cpu_name
596	.long	feroceon_processor_functions
597	.long	v4wbi_tlb_fns
598	.long	feroceon_user_fns
599	.long	\cache
600	 .size	__\name\()_proc_info, . - __\name\()_proc_info
601.endm
602
603#ifdef CONFIG_CPU_FEROCEON_OLD_ID
604	feroceon_proc_info feroceon_old_id, 0x41009260, 0xff00fff0, \
605		cpu_name=cpu_feroceon_name, cache=feroceon_cache_fns
606#endif
607
608	feroceon_proc_info 88fr531, 0x56055310, 0xfffffff0, cpu_88fr531_name, \
609		cache=feroceon_cache_fns
610	feroceon_proc_info 88fr571, 0x56155710, 0xfffffff0, cpu_88fr571_name, \
611		cache=feroceon_range_cache_fns
612	feroceon_proc_info 88fr131, 0x56251310, 0xfffffff0, cpu_88fr131_name, \
613		cache=feroceon_range_cache_fns
614