1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
|
/*
* arch/sh/mm/cache-sh5.c
*
* Copyright (C) 2000, 2001 Paolo Alberelli
* Copyright (C) 2002 Benedict Gaster
* Copyright (C) 2003 Richard Curnow
* Copyright (C) 2003 - 2008 Paul Mundt
*
* This file is subject to the terms and conditions of the GNU General Public
* License. See the file "COPYING" in the main directory of this archive
* for more details.
*/
#include <linux/init.h>
#include <linux/mman.h>
#include <linux/mm.h>
#include <asm/tlb.h>
#include <asm/processor.h>
#include <asm/cache.h>
#include <asm/pgalloc.h>
#include <asm/uaccess.h>
#include <asm/mmu_context.h>
extern void __weak sh4__flush_region_init(void);
/* Wired TLB entry for the D-cache */
static unsigned long long dtlb_cache_slot;
/*
* The following group of functions deal with mapping and unmapping a
* temporary page into a DTLB slot that has been set aside for exclusive
* use.
*/
static inline void
sh64_setup_dtlb_cache_slot(unsigned long eaddr, unsigned long asid,
unsigned long paddr)
{
local_irq_disable();
sh64_setup_tlb_slot(dtlb_cache_slot, eaddr, asid, paddr);
}
static inline void sh64_teardown_dtlb_cache_slot(void)
{
sh64_teardown_tlb_slot(dtlb_cache_slot);
local_irq_enable();
}
static inline void sh64_icache_inv_all(void)
{
unsigned long long addr, flag, data;
unsigned long flags;
addr = ICCR0;
flag = ICCR0_ICI;
data = 0;
/* Make this a critical section for safety (probably not strictly necessary.) */
local_irq_save(flags);
/* Without %1 it gets unexplicably wrong */
__asm__ __volatile__ (
"getcfg %3, 0, %0\n\t"
"or %0, %2, %0\n\t"
"putcfg %3, 0, %0\n\t"
"synci"
: "=&r" (data)
: "0" (data), "r" (flag), "r" (addr));
local_irq_restore(flags);
}
static void sh64_icache_inv_kernel_range(unsigned long start, unsigned long end)
{
/* Invalidate range of addresses [start,end] from the I-cache, where
* the addresses lie in the kernel superpage. */
unsigned long long ullend, addr, aligned_start;
aligned_start = (unsigned long long)(signed long long)(signed long) start;
addr = L1_CACHE_ALIGN(aligned_start);
ullend = (unsigned long long) (signed long long) (signed long) end;
while (addr <= ullend) {
__asm__ __volatile__ ("icbi %0, 0" : : "r" (addr));
addr += L1_CACHE_BYTES;
}
}
static void sh64_icache_inv_user_page(struct vm_area_struct *vma, unsigned long eaddr)
{
/* If we get called, we know that vma->vm_flags contains VM_EXEC.
Also, eaddr is page-aligned. */
unsigned int cpu = smp_processor_id();
unsigned long long addr, end_addr;
unsigned long flags = 0;
unsigned long running_asid, vma_asid;
addr = eaddr;
end_addr = addr + PAGE_SIZE;
/* Check whether we can use the current ASID for the I-cache
invalidation. For example, if we're called via
access_process_vm->flush_cache_page->here, (e.g. when reading from
/proc), 'running_asid' will be that of the reader, not of the
victim.
Also, note the risk that we might get pre-empted between the ASID
compare and blocking IRQs, and before we regain control, the
pid->ASID mapping changes. However, the whole cache will get
invalidated when the mapping is renewed, so the worst that can
happen is that the loop below ends up invalidating somebody else's
cache entries.
*/
running_asid = get_asid();
vma_asid = cpu_asid(cpu, vma->vm_mm);
if (running_asid != vma_asid) {
local_irq_save(flags);
switch_and_save_asid(vma_asid);
}
while (addr < end_addr) {
/* Worth unrolling a little */
__asm__ __volatile__("icbi %0, 0" : : "r" (addr));
__asm__ __volatile__("icbi %0, 32" : : "r" (addr));
__asm__ __volatile__("icbi %0, 64" : : "r" (addr));
__asm__ __volatile__("icbi %0, 96" : : "r" (addr));
addr += 128;
}
if (running_asid != vma_asid) {
switch_and_save_asid(running_asid);
local_irq_restore(flags);
}
}
static void sh64_icache_inv_user_page_range(struct mm_struct *mm,
unsigned long start, unsigned long end)
{
/* Used for invalidating big chunks of I-cache, i.e. assume the range
is whole pages. If 'start' or 'end' is not page aligned, the code
is conservative and invalidates to the ends of the enclosing pages.
This is functionally OK, just a performance loss. */
/* See the comments below in sh64_dcache_purge_user_range() regarding
the choice of algorithm. However, for the I-cache option (2) isn't
available because there are no physical tags so aliases can't be
resolved. The icbi instruction has to be used through the user
mapping. Because icbi is cheaper than ocbp on a cache hit, it
would be cheaper to use the selective code for a large range than is
possible with the D-cache. Just assume 64 for now as a working
figure.
*/
int n_pages;
if (!mm)
return;
n_pages = ((end - start) >> PAGE_SHIFT);
if (n_pages >= 64) {
sh64_icache_inv_all();
} else {
unsigned long aligned_start;
unsigned long eaddr;
unsigned long after_last_page_start;
unsigned long mm_asid, current_asid;
unsigned long flags = 0;
mm_asid = cpu_asid(smp_processor_id(), mm);
current_asid = get_asid();
if (mm_asid != current_asid) {
/* Switch ASID and run the invalidate loop under cli */
local_irq_save(flags);
switch_and_save_asid(mm_asid);
}
aligned_start = start & PAGE_MASK;
after_last_page_start = PAGE_SIZE + ((end - 1) & PAGE_MASK);
while (aligned_start < after_last_page_start) {
struct vm_area_struct *vma;
unsigned long vma_end;
vma = find_vma(mm, aligned_start);
if (!vma || (aligned_start <= vma->vm_end)) {
/* Avoid getting stuck in an error condition */
aligned_start += PAGE_SIZE;
continue;
}
vma_end = vma->vm_end;
if (vma->vm_flags & VM_EXEC) {
/* Executable */
eaddr = aligned_start;
while (eaddr < vma_end) {
sh64_icache_inv_user_page(vma, eaddr);
eaddr += PAGE_SIZE;
}
}
aligned_start = vma->vm_end; /* Skip to start of next region */
}
if (mm_asid != current_asid) {
switch_and_save_asid(current_asid);
local_irq_restore(flags);
}
}
}
static void sh64_icache_inv_current_user_range(unsigned long start, unsigned long end)
{
/* The icbi instruction never raises ITLBMISS. i.e. if there's not a
cache hit on the virtual tag the instruction ends there, without a
TLB lookup. */
unsigned long long aligned_start;
unsigned long long ull_end;
unsigned long long addr;
ull_end = end;
/* Just invalidate over the range using the natural addresses. TLB
miss handling will be OK (TBC). Since it's for the current process,
either we're already in the right ASID context, or the ASIDs have
been recycled since we were last active in which case we might just
invalidate another processes I-cache entries : no worries, just a
performance drop for him. */
aligned_start = L1_CACHE_ALIGN(start);
addr = aligned_start;
while (addr < ull_end) {
__asm__ __volatile__ ("icbi %0, 0" : : "r" (addr));
__asm__ __volatile__ ("nop");
__asm__ __volatile__ ("nop");
addr += L1_CACHE_BYTES;
}
}
/* Buffer used as the target of alloco instructions to purge data from cache
sets by natural eviction. -- RPC */
#define DUMMY_ALLOCO_AREA_SIZE ((L1_CACHE_BYTES << 10) + (1024 * 4))
static unsigned char dummy_alloco_area[DUMMY_ALLOCO_AREA_SIZE] __cacheline_aligned = { 0, };
static void inline sh64_dcache_purge_sets(int sets_to_purge_base, int n_sets)
|