forked from zephyrproject-rtos/zephyr
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmmu.c
1565 lines (1341 loc) · 44.2 KB
/
mmu.c
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
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*
* Copyright (c) 2020 Intel Corporation
*
* SPDX-License-Identifier: Apache-2.0
*
* Routines for managing virtual address spaces
*/
#include <stdint.h>
#include <kernel_arch_interface.h>
#include <zephyr/spinlock.h>
#include <mmu.h>
#include <zephyr/init.h>
#include <kernel_internal.h>
#include <zephyr/syscall_handler.h>
#include <zephyr/toolchain.h>
#include <zephyr/linker/linker-defs.h>
#include <zephyr/sys/bitarray.h>
#include <zephyr/timing/timing.h>
#include <zephyr/logging/log.h>
LOG_MODULE_DECLARE(os, CONFIG_KERNEL_LOG_LEVEL);
/*
* General terminology:
* - A page frame is a page-sized physical memory region in RAM. It is a
* container where a data page may be placed. It is always referred to by
* physical address. We have a convention of using uintptr_t for physical
* addresses. We instantiate a struct z_page_frame to store metadata for
* every page frame.
*
* - A data page is a page-sized region of data. It may exist in a page frame,
* or be paged out to some backing store. Its location can always be looked
* up in the CPU's page tables (or equivalent) by virtual address.
* The data type will always be void * or in some cases uint8_t * when we
* want to do pointer arithmetic.
*/
/* Spinlock to protect any globals in this file and serialize page table
* updates in arch code
*/
struct k_spinlock z_mm_lock;
/*
* General page frame management
*/
/* Database of all RAM page frames */
struct z_page_frame z_page_frames[Z_NUM_PAGE_FRAMES];
#if __ASSERT_ON
/* Indicator that z_page_frames has been initialized, many of these APIs do
* not work before POST_KERNEL
*/
static bool page_frames_initialized;
#endif
/* Add colors to page table dumps to indicate mapping type */
#define COLOR_PAGE_FRAMES 1
#if COLOR_PAGE_FRAMES
#define ANSI_DEFAULT "\x1B" "[0m"
#define ANSI_RED "\x1B" "[1;31m"
#define ANSI_GREEN "\x1B" "[1;32m"
#define ANSI_YELLOW "\x1B" "[1;33m"
#define ANSI_BLUE "\x1B" "[1;34m"
#define ANSI_MAGENTA "\x1B" "[1;35m"
#define ANSI_CYAN "\x1B" "[1;36m"
#define ANSI_GREY "\x1B" "[1;90m"
#define COLOR(x) printk(_CONCAT(ANSI_, x))
#else
#define COLOR(x) do { } while (false)
#endif
/* LCOV_EXCL_START */
static void page_frame_dump(struct z_page_frame *pf)
{
if (z_page_frame_is_reserved(pf)) {
COLOR(CYAN);
printk("R");
} else if (z_page_frame_is_busy(pf)) {
COLOR(MAGENTA);
printk("B");
} else if (z_page_frame_is_pinned(pf)) {
COLOR(YELLOW);
printk("P");
} else if (z_page_frame_is_available(pf)) {
COLOR(GREY);
printk(".");
} else if (z_page_frame_is_mapped(pf)) {
COLOR(DEFAULT);
printk("M");
} else {
COLOR(RED);
printk("?");
}
}
void z_page_frames_dump(void)
{
int column = 0;
__ASSERT(page_frames_initialized, "%s called too early", __func__);
printk("Physical memory from 0x%lx to 0x%lx\n",
Z_PHYS_RAM_START, Z_PHYS_RAM_END);
for (int i = 0; i < Z_NUM_PAGE_FRAMES; i++) {
struct z_page_frame *pf = &z_page_frames[i];
page_frame_dump(pf);
column++;
if (column == 64) {
column = 0;
printk("\n");
}
}
COLOR(DEFAULT);
if (column != 0) {
printk("\n");
}
}
/* LCOV_EXCL_STOP */
#define VIRT_FOREACH(_base, _size, _pos) \
for (_pos = _base; \
_pos < ((uint8_t *)_base + _size); _pos += CONFIG_MMU_PAGE_SIZE)
#define PHYS_FOREACH(_base, _size, _pos) \
for (_pos = _base; \
_pos < ((uintptr_t)_base + _size); _pos += CONFIG_MMU_PAGE_SIZE)
/*
* Virtual address space management
*
* Call all of these functions with z_mm_lock held.
*
* Overall virtual memory map: When the kernel starts, it resides in
* virtual memory in the region Z_KERNEL_VIRT_START to
* Z_KERNEL_VIRT_END. Unused virtual memory past this, up to the limit
* noted by CONFIG_KERNEL_VM_SIZE may be used for runtime memory mappings.
*
* If CONFIG_ARCH_MAPS_ALL_RAM is set, we do not just map the kernel image,
* but have a mapping for all RAM in place. This is for special architectural
* purposes and does not otherwise affect page frame accounting or flags;
* the only guarantee is that such RAM mapping outside of the Zephyr image
* won't be disturbed by subsequent memory mapping calls.
*
* +--------------+ <- Z_VIRT_RAM_START
* | Undefined VM | <- May contain ancillary regions like x86_64's locore
* +--------------+ <- Z_KERNEL_VIRT_START (often == Z_VIRT_RAM_START)
* | Mapping for |
* | main kernel |
* | image |
* | |
* | |
* +--------------+ <- Z_FREE_VM_START
* | |
* | Unused, |
* | Available VM |
* | |
* |..............| <- mapping_pos (grows downward as more mappings are made)
* | Mapping |
* +--------------+
* | Mapping |
* +--------------+
* | ... |
* +--------------+
* | Mapping |
* +--------------+ <- mappings start here
* | Reserved | <- special purpose virtual page(s) of size Z_VM_RESERVED
* +--------------+ <- Z_VIRT_RAM_END
*/
/* Bitmap of virtual addresses where one bit corresponds to one page.
* This is being used for virt_region_alloc() to figure out which
* region of virtual addresses can be used for memory mapping.
*
* Note that bit #0 is the highest address so that allocation is
* done in reverse from highest address.
*/
SYS_BITARRAY_DEFINE_STATIC(virt_region_bitmap,
CONFIG_KERNEL_VM_SIZE / CONFIG_MMU_PAGE_SIZE);
static bool virt_region_inited;
#define Z_VIRT_REGION_START_ADDR Z_FREE_VM_START
#define Z_VIRT_REGION_END_ADDR (Z_VIRT_RAM_END - Z_VM_RESERVED)
static inline uintptr_t virt_from_bitmap_offset(size_t offset, size_t size)
{
return POINTER_TO_UINT(Z_VIRT_RAM_END)
- (offset * CONFIG_MMU_PAGE_SIZE) - size;
}
static inline size_t virt_to_bitmap_offset(void *vaddr, size_t size)
{
return (POINTER_TO_UINT(Z_VIRT_RAM_END)
- POINTER_TO_UINT(vaddr) - size) / CONFIG_MMU_PAGE_SIZE;
}
static void virt_region_init(void)
{
size_t offset, num_bits;
/* There are regions where we should never map via
* k_mem_map() and z_phys_map(). Mark them as
* already allocated so they will never be used.
*/
if (Z_VM_RESERVED > 0) {
/* Mark reserved region at end of virtual address space */
num_bits = Z_VM_RESERVED / CONFIG_MMU_PAGE_SIZE;
(void)sys_bitarray_set_region(&virt_region_bitmap,
num_bits, 0);
}
/* Mark all bits up to Z_FREE_VM_START as allocated */
num_bits = POINTER_TO_UINT(Z_FREE_VM_START)
- POINTER_TO_UINT(Z_VIRT_RAM_START);
offset = virt_to_bitmap_offset(Z_VIRT_RAM_START, num_bits);
num_bits /= CONFIG_MMU_PAGE_SIZE;
(void)sys_bitarray_set_region(&virt_region_bitmap,
num_bits, offset);
virt_region_inited = true;
}
static void virt_region_free(void *vaddr, size_t size)
{
size_t offset, num_bits;
uint8_t *vaddr_u8 = (uint8_t *)vaddr;
if (unlikely(!virt_region_inited)) {
virt_region_init();
}
#ifndef CONFIG_KERNEL_DIRECT_MAP
/* Without the need to support K_DIRECT_MAP, the region must be
* able to be represented in the bitmap. So this case is
* simple.
*/
__ASSERT((vaddr_u8 >= Z_VIRT_REGION_START_ADDR)
&& ((vaddr_u8 + size - 1) < Z_VIRT_REGION_END_ADDR),
"invalid virtual address region %p (%zu)", vaddr_u8, size);
if (!((vaddr_u8 >= Z_VIRT_REGION_START_ADDR)
&& ((vaddr_u8 + size - 1) < Z_VIRT_REGION_END_ADDR))) {
return;
}
offset = virt_to_bitmap_offset(vaddr, size);
num_bits = size / CONFIG_MMU_PAGE_SIZE;
(void)sys_bitarray_free(&virt_region_bitmap, num_bits, offset);
#else /* !CONFIG_KERNEL_DIRECT_MAP */
/* With K_DIRECT_MAP, the region can be outside of the virtual
* memory space, wholly within it, or overlap partially.
* So additional processing is needed to make sure we only
* mark the pages within the bitmap.
*/
if (((vaddr_u8 >= Z_VIRT_REGION_START_ADDR) &&
(vaddr_u8 < Z_VIRT_REGION_END_ADDR)) ||
(((vaddr_u8 + size - 1) >= Z_VIRT_REGION_START_ADDR) &&
((vaddr_u8 + size - 1) < Z_VIRT_REGION_END_ADDR))) {
uint8_t *adjusted_start = MAX(vaddr_u8, Z_VIRT_REGION_START_ADDR);
uint8_t *adjusted_end = MIN(vaddr_u8 + size,
Z_VIRT_REGION_END_ADDR);
size_t adjusted_sz = adjusted_end - adjusted_start;
offset = virt_to_bitmap_offset(adjusted_start, adjusted_sz);
num_bits = adjusted_sz / CONFIG_MMU_PAGE_SIZE;
(void)sys_bitarray_free(&virt_region_bitmap, num_bits, offset);
}
#endif /* !CONFIG_KERNEL_DIRECT_MAP */
}
static void *virt_region_alloc(size_t size, size_t align)
{
uintptr_t dest_addr;
size_t alloc_size;
size_t offset;
size_t num_bits;
int ret;
if (unlikely(!virt_region_inited)) {
virt_region_init();
}
/* Possibly request more pages to ensure we can get an aligned virtual address */
num_bits = (size + align - CONFIG_MMU_PAGE_SIZE) / CONFIG_MMU_PAGE_SIZE;
alloc_size = num_bits * CONFIG_MMU_PAGE_SIZE;
ret = sys_bitarray_alloc(&virt_region_bitmap, num_bits, &offset);
if (ret != 0) {
LOG_ERR("insufficient virtual address space (requested %zu)",
size);
return NULL;
}
/* Remember that bit #0 in bitmap corresponds to the highest
* virtual address. So here we need to go downwards (backwards?)
* to get the starting address of the allocated region.
*/
dest_addr = virt_from_bitmap_offset(offset, alloc_size);
if (alloc_size > size) {
uintptr_t aligned_dest_addr = ROUND_UP(dest_addr, align);
/* Here is the memory organization when trying to get an aligned
* virtual address:
*
* +--------------+ <- Z_VIRT_RAM_START
* | Undefined VM |
* +--------------+ <- Z_KERNEL_VIRT_START (often == Z_VIRT_RAM_START)
* | Mapping for |
* | main kernel |
* | image |
* | |
* | |
* +--------------+ <- Z_FREE_VM_START
* | ... |
* +==============+ <- dest_addr
* | Unused |
* |..............| <- aligned_dest_addr
* | |
* | Aligned |
* | Mapping |
* | |
* |..............| <- aligned_dest_addr + size
* | Unused |
* +==============+ <- offset from Z_VIRT_RAM_END == dest_addr + alloc_size
* | ... |
* +--------------+
* | Mapping |
* +--------------+
* | Reserved |
* +--------------+ <- Z_VIRT_RAM_END
*/
/* Free the two unused regions */
virt_region_free(UINT_TO_POINTER(dest_addr),
aligned_dest_addr - dest_addr);
if (((dest_addr + alloc_size) - (aligned_dest_addr + size)) > 0) {
virt_region_free(UINT_TO_POINTER(aligned_dest_addr + size),
(dest_addr + alloc_size) - (aligned_dest_addr + size));
}
dest_addr = aligned_dest_addr;
}
/* Need to make sure this does not step into kernel memory */
if (dest_addr < POINTER_TO_UINT(Z_VIRT_REGION_START_ADDR)) {
(void)sys_bitarray_free(&virt_region_bitmap, size, offset);
return NULL;
}
return UINT_TO_POINTER(dest_addr);
}
/*
* Free page frames management
*
* Call all of these functions with z_mm_lock held.
*/
/* Linked list of unused and available page frames.
*
* TODO: This is very simple and treats all free page frames as being equal.
* However, there are use-cases to consolidate free pages such that entire
* SRAM banks can be switched off to save power, and so obtaining free pages
* may require a more complex ontology which prefers page frames in RAM banks
* which are still active.
*
* This implies in the future there may be multiple slists managing physical
* pages. Each page frame will still just have one snode link.
*/
static sys_slist_t free_page_frame_list;
/* Number of unused and available free page frames */
size_t z_free_page_count;
#define PF_ASSERT(pf, expr, fmt, ...) \
__ASSERT(expr, "page frame 0x%lx: " fmt, z_page_frame_to_phys(pf), \
##__VA_ARGS__)
/* Get an unused page frame. don't care which one, or NULL if there are none */
static struct z_page_frame *free_page_frame_list_get(void)
{
sys_snode_t *node;
struct z_page_frame *pf = NULL;
node = sys_slist_get(&free_page_frame_list);
if (node != NULL) {
z_free_page_count--;
pf = CONTAINER_OF(node, struct z_page_frame, node);
PF_ASSERT(pf, z_page_frame_is_available(pf),
"unavailable but somehow on free list");
}
return pf;
}
/* Release a page frame back into the list of free pages */
static void free_page_frame_list_put(struct z_page_frame *pf)
{
PF_ASSERT(pf, z_page_frame_is_available(pf),
"unavailable page put on free list");
/* The structure is packed, which ensures that this is true */
void *node = pf;
sys_slist_append(&free_page_frame_list, node);
z_free_page_count++;
}
static void free_page_frame_list_init(void)
{
sys_slist_init(&free_page_frame_list);
}
static void page_frame_free_locked(struct z_page_frame *pf)
{
pf->flags = 0;
free_page_frame_list_put(pf);
}
/*
* Memory Mapping
*/
/* Called after the frame is mapped in the arch layer, to update our
* local ontology (and do some assertions while we're at it)
*/
static void frame_mapped_set(struct z_page_frame *pf, void *addr)
{
PF_ASSERT(pf, !z_page_frame_is_reserved(pf),
"attempted to map a reserved page frame");
/* We do allow multiple mappings for pinned page frames
* since we will never need to reverse map them.
* This is uncommon, use-cases are for things like the
* Zephyr equivalent of VSDOs
*/
PF_ASSERT(pf, !z_page_frame_is_mapped(pf) || z_page_frame_is_pinned(pf),
"non-pinned and already mapped to %p", pf->addr);
pf->flags |= Z_PAGE_FRAME_MAPPED;
pf->addr = addr;
}
/* LCOV_EXCL_START */
/* Go through page frames to find the physical address mapped
* by a virtual address.
*
* @param[in] virt Virtual Address
* @param[out] phys Physical address mapped to the input virtual address
* if such mapping exists.
*
* @retval 0 if mapping is found and valid
* @retval -EFAULT if virtual address is not mapped
*/
static int virt_to_page_frame(void *virt, uintptr_t *phys)
{
uintptr_t paddr;
struct z_page_frame *pf;
int ret = -EFAULT;
Z_PAGE_FRAME_FOREACH(paddr, pf) {
if (z_page_frame_is_mapped(pf)) {
if (virt == pf->addr) {
ret = 0;
*phys = z_page_frame_to_phys(pf);
break;
}
}
}
return ret;
}
/* LCOV_EXCL_STOP */
__weak FUNC_ALIAS(virt_to_page_frame, arch_page_phys_get, int);
#ifdef CONFIG_DEMAND_PAGING
static int page_frame_prepare_locked(struct z_page_frame *pf, bool *dirty_ptr,
bool page_in, uintptr_t *location_ptr);
static inline void do_backing_store_page_in(uintptr_t location);
static inline void do_backing_store_page_out(uintptr_t location);
#endif /* CONFIG_DEMAND_PAGING */
/* Allocate a free page frame, and map it to a specified virtual address
*
* TODO: Add optional support for copy-on-write mappings to a zero page instead
* of allocating, in which case page frames will be allocated lazily as
* the mappings to the zero page get touched. This will avoid expensive
* page-ins as memory is mapped and physical RAM or backing store storage will
* not be used if the mapped memory is unused. The cost is an empty physical
* page of zeroes.
*/
static int map_anon_page(void *addr, uint32_t flags)
{
struct z_page_frame *pf;
uintptr_t phys;
bool lock = (flags & K_MEM_MAP_LOCK) != 0U;
bool uninit = (flags & K_MEM_MAP_UNINIT) != 0U;
pf = free_page_frame_list_get();
if (pf == NULL) {
#ifdef CONFIG_DEMAND_PAGING
uintptr_t location;
bool dirty;
int ret;
pf = k_mem_paging_eviction_select(&dirty);
__ASSERT(pf != NULL, "failed to get a page frame");
LOG_DBG("evicting %p at 0x%lx", pf->addr,
z_page_frame_to_phys(pf));
ret = page_frame_prepare_locked(pf, &dirty, false, &location);
if (ret != 0) {
return -ENOMEM;
}
if (dirty) {
do_backing_store_page_out(location);
}
pf->flags = 0;
#else
return -ENOMEM;
#endif /* CONFIG_DEMAND_PAGING */
}
phys = z_page_frame_to_phys(pf);
arch_mem_map(addr, phys, CONFIG_MMU_PAGE_SIZE, flags | K_MEM_CACHE_WB);
if (lock) {
pf->flags |= Z_PAGE_FRAME_PINNED;
}
frame_mapped_set(pf, addr);
LOG_DBG("memory mapping anon page %p -> 0x%lx", addr, phys);
if (!uninit) {
/* If we later implement mappings to a copy-on-write
* zero page, won't need this step
*/
memset(addr, 0, CONFIG_MMU_PAGE_SIZE);
}
return 0;
}
void *k_mem_map(size_t size, uint32_t flags)
{
uint8_t *dst;
size_t total_size;
int ret;
k_spinlock_key_t key;
uint8_t *pos;
__ASSERT(!(((flags & K_MEM_PERM_USER) != 0U) &&
((flags & K_MEM_MAP_UNINIT) != 0U)),
"user access to anonymous uninitialized pages is forbidden");
__ASSERT(size % CONFIG_MMU_PAGE_SIZE == 0U,
"unaligned size %zu passed to %s", size, __func__);
__ASSERT(size != 0, "zero sized memory mapping");
__ASSERT(page_frames_initialized, "%s called too early", __func__);
__ASSERT((flags & K_MEM_CACHE_MASK) == 0U,
"%s does not support explicit cache settings", __func__);
key = k_spin_lock(&z_mm_lock);
/* Need extra for the guard pages (before and after) which we
* won't map.
*/
total_size = size + CONFIG_MMU_PAGE_SIZE * 2;
dst = virt_region_alloc(total_size, CONFIG_MMU_PAGE_SIZE);
if (dst == NULL) {
/* Address space has no free region */
goto out;
}
/* Unmap both guard pages to make sure accessing them
* will generate fault.
*/
arch_mem_unmap(dst, CONFIG_MMU_PAGE_SIZE);
arch_mem_unmap(dst + CONFIG_MMU_PAGE_SIZE + size,
CONFIG_MMU_PAGE_SIZE);
/* Skip over the "before" guard page in returned address. */
dst += CONFIG_MMU_PAGE_SIZE;
VIRT_FOREACH(dst, size, pos) {
ret = map_anon_page(pos, flags);
if (ret != 0) {
/* TODO: call k_mem_unmap(dst, pos - dst) when
* implemented in #28990 and release any guard virtual
* page as well.
*/
dst = NULL;
goto out;
}
}
out:
k_spin_unlock(&z_mm_lock, key);
return dst;
}
void k_mem_unmap(void *addr, size_t size)
{
uintptr_t phys;
uint8_t *pos;
struct z_page_frame *pf;
k_spinlock_key_t key;
size_t total_size;
int ret;
/* Need space for the "before" guard page */
__ASSERT_NO_MSG(POINTER_TO_UINT(addr) >= CONFIG_MMU_PAGE_SIZE);
/* Make sure address range is still valid after accounting
* for two guard pages.
*/
pos = (uint8_t *)addr - CONFIG_MMU_PAGE_SIZE;
z_mem_assert_virtual_region(pos, size + (CONFIG_MMU_PAGE_SIZE * 2));
key = k_spin_lock(&z_mm_lock);
/* Check if both guard pages are unmapped.
* Bail if not, as this is probably a region not mapped
* using k_mem_map().
*/
pos = addr;
ret = arch_page_phys_get(pos - CONFIG_MMU_PAGE_SIZE, NULL);
if (ret == 0) {
__ASSERT(ret == 0,
"%s: cannot find preceding guard page for (%p, %zu)",
__func__, addr, size);
goto out;
}
ret = arch_page_phys_get(pos + size, NULL);
if (ret == 0) {
__ASSERT(ret == 0,
"%s: cannot find succeeding guard page for (%p, %zu)",
__func__, addr, size);
goto out;
}
VIRT_FOREACH(addr, size, pos) {
ret = arch_page_phys_get(pos, &phys);
__ASSERT(ret == 0,
"%s: cannot unmap an unmapped address %p",
__func__, pos);
if (ret != 0) {
/* Found an address not mapped. Do not continue. */
goto out;
}
__ASSERT(z_is_page_frame(phys),
"%s: 0x%lx is not a page frame", __func__, phys);
if (!z_is_page_frame(phys)) {
/* Physical address has no corresponding page frame
* description in the page frame array.
* This should not happen. Do not continue.
*/
goto out;
}
/* Grab the corresponding page frame from physical address */
pf = z_phys_to_page_frame(phys);
__ASSERT(z_page_frame_is_mapped(pf),
"%s: 0x%lx is not a mapped page frame", __func__, phys);
if (!z_page_frame_is_mapped(pf)) {
/* Page frame is not marked mapped.
* This should not happen. Do not continue.
*/
goto out;
}
arch_mem_unmap(pos, CONFIG_MMU_PAGE_SIZE);
/* Put the page frame back into free list */
page_frame_free_locked(pf);
}
/* There are guard pages just before and after the mapped
* region. So we also need to free them from the bitmap.
*/
pos = (uint8_t *)addr - CONFIG_MMU_PAGE_SIZE;
total_size = size + CONFIG_MMU_PAGE_SIZE * 2;
virt_region_free(pos, total_size);
out:
k_spin_unlock(&z_mm_lock, key);
}
size_t k_mem_free_get(void)
{
size_t ret;
k_spinlock_key_t key;
__ASSERT(page_frames_initialized, "%s called too early", __func__);
key = k_spin_lock(&z_mm_lock);
#ifdef CONFIG_DEMAND_PAGING
if (z_free_page_count > CONFIG_DEMAND_PAGING_PAGE_FRAMES_RESERVE) {
ret = z_free_page_count - CONFIG_DEMAND_PAGING_PAGE_FRAMES_RESERVE;
} else {
ret = 0;
}
#else
ret = z_free_page_count;
#endif
k_spin_unlock(&z_mm_lock, key);
return ret * (size_t)CONFIG_MMU_PAGE_SIZE;
}
/* Get the default virtual region alignment, here the default MMU page size
*
* @param[in] phys Physical address of region to be mapped, aligned to MMU_PAGE_SIZE
* @param[in] size Size of region to be mapped, aligned to MMU_PAGE_SIZE
*
* @retval alignment to apply on the virtual address of this region
*/
static size_t virt_region_align(uintptr_t phys, size_t size)
{
ARG_UNUSED(phys);
ARG_UNUSED(size);
return CONFIG_MMU_PAGE_SIZE;
}
__weak FUNC_ALIAS(virt_region_align, arch_virt_region_align, size_t);
/* This may be called from arch early boot code before z_cstart() is invoked.
* Data will be copied and BSS zeroed, but this must not rely on any
* initialization functions being called prior to work correctly.
*/
void z_phys_map(uint8_t **virt_ptr, uintptr_t phys, size_t size, uint32_t flags)
{
uintptr_t aligned_phys, addr_offset;
size_t aligned_size, align_boundary;
k_spinlock_key_t key;
uint8_t *dest_addr;
size_t num_bits;
size_t offset;
#ifndef CONFIG_KERNEL_DIRECT_MAP
__ASSERT(!(flags & K_MEM_DIRECT_MAP), "The direct-map is not enabled");
#endif
addr_offset = k_mem_region_align(&aligned_phys, &aligned_size,
phys, size,
CONFIG_MMU_PAGE_SIZE);
__ASSERT(aligned_size != 0U, "0-length mapping at 0x%lx", aligned_phys);
__ASSERT(aligned_phys < (aligned_phys + (aligned_size - 1)),
"wraparound for physical address 0x%lx (size %zu)",
aligned_phys, aligned_size);
align_boundary = arch_virt_region_align(aligned_phys, aligned_size);
key = k_spin_lock(&z_mm_lock);
if (IS_ENABLED(CONFIG_KERNEL_DIRECT_MAP) &&
(flags & K_MEM_DIRECT_MAP)) {
dest_addr = (uint8_t *)aligned_phys;
/* Mark the region of virtual memory bitmap as used
* if the region overlaps the virtual memory space.
*
* Basically if either end of region is within
* virtual memory space, we need to mark the bits.
*/
if (((dest_addr >= Z_VIRT_RAM_START) &&
(dest_addr < Z_VIRT_RAM_END)) ||
(((dest_addr + aligned_size) >= Z_VIRT_RAM_START) &&
((dest_addr + aligned_size) < Z_VIRT_RAM_END))) {
uint8_t *adjusted_start = MAX(dest_addr, Z_VIRT_RAM_START);
uint8_t *adjusted_end = MIN(dest_addr + aligned_size,
Z_VIRT_RAM_END);
size_t adjusted_sz = adjusted_end - adjusted_start;
num_bits = adjusted_sz / CONFIG_MMU_PAGE_SIZE;
offset = virt_to_bitmap_offset(adjusted_start, adjusted_sz);
if (sys_bitarray_test_and_set_region(
&virt_region_bitmap, num_bits, offset, true))
goto fail;
}
} else {
/* Obtain an appropriately sized chunk of virtual memory */
dest_addr = virt_region_alloc(aligned_size, align_boundary);
if (!dest_addr) {
goto fail;
}
}
/* If this fails there's something amiss with virt_region_get */
__ASSERT((uintptr_t)dest_addr <
((uintptr_t)dest_addr + (size - 1)),
"wraparound for virtual address %p (size %zu)",
dest_addr, size);
LOG_DBG("arch_mem_map(%p, 0x%lx, %zu, %x) offset %lu", dest_addr,
aligned_phys, aligned_size, flags, addr_offset);
arch_mem_map(dest_addr, aligned_phys, aligned_size, flags);
k_spin_unlock(&z_mm_lock, key);
*virt_ptr = dest_addr + addr_offset;
return;
fail:
/* May re-visit this in the future, but for now running out of
* virtual address space or failing the arch_mem_map() call is
* an unrecoverable situation.
*
* Other problems not related to resource exhaustion we leave as
* assertions since they are clearly programming mistakes.
*/
LOG_ERR("memory mapping 0x%lx (size %zu, flags 0x%x) failed",
phys, size, flags);
k_panic();
}
void z_phys_unmap(uint8_t *virt, size_t size)
{
uintptr_t aligned_virt, addr_offset;
size_t aligned_size;
k_spinlock_key_t key;
addr_offset = k_mem_region_align(&aligned_virt, &aligned_size,
POINTER_TO_UINT(virt), size,
CONFIG_MMU_PAGE_SIZE);
__ASSERT(aligned_size != 0U, "0-length mapping at 0x%lx", aligned_virt);
__ASSERT(aligned_virt < (aligned_virt + (aligned_size - 1)),
"wraparound for virtual address 0x%lx (size %zu)",
aligned_virt, aligned_size);
key = k_spin_lock(&z_mm_lock);
LOG_DBG("arch_mem_unmap(0x%lx, %zu) offset %lu",
aligned_virt, aligned_size, addr_offset);
arch_mem_unmap(UINT_TO_POINTER(aligned_virt), aligned_size);
virt_region_free(UINT_TO_POINTER(aligned_virt), aligned_size);
k_spin_unlock(&z_mm_lock, key);
}
/*
* Miscellaneous
*/
size_t k_mem_region_align(uintptr_t *aligned_addr, size_t *aligned_size,
uintptr_t addr, size_t size, size_t align)
{
size_t addr_offset;
/* The actual mapped region must be page-aligned. Round down the
* physical address and pad the region size appropriately
*/
*aligned_addr = ROUND_DOWN(addr, align);
addr_offset = addr - *aligned_addr;
*aligned_size = ROUND_UP(size + addr_offset, align);
return addr_offset;
}
#if defined(CONFIG_LINKER_USE_BOOT_SECTION) || defined(CONFIG_LINKER_USE_PINNED_SECTION)
static void mark_linker_section_pinned(void *start_addr, void *end_addr,
bool pin)
{
struct z_page_frame *pf;
uint8_t *addr;
uintptr_t pinned_start = ROUND_DOWN(POINTER_TO_UINT(start_addr),
CONFIG_MMU_PAGE_SIZE);
uintptr_t pinned_end = ROUND_UP(POINTER_TO_UINT(end_addr),
CONFIG_MMU_PAGE_SIZE);
size_t pinned_size = pinned_end - pinned_start;
VIRT_FOREACH(UINT_TO_POINTER(pinned_start), pinned_size, addr)
{
pf = z_phys_to_page_frame(Z_BOOT_VIRT_TO_PHYS(addr));
frame_mapped_set(pf, addr);
if (pin) {
pf->flags |= Z_PAGE_FRAME_PINNED;
} else {
pf->flags &= ~Z_PAGE_FRAME_PINNED;
}
}
}
#endif /* CONFIG_LINKER_USE_BOOT_SECTION) || CONFIG_LINKER_USE_PINNED_SECTION */
void z_mem_manage_init(void)
{
uintptr_t phys;
uint8_t *addr;
struct z_page_frame *pf;
k_spinlock_key_t key = k_spin_lock(&z_mm_lock);
free_page_frame_list_init();
ARG_UNUSED(addr);
#ifdef CONFIG_ARCH_HAS_RESERVED_PAGE_FRAMES
/* If some page frames are unavailable for use as memory, arch
* code will mark Z_PAGE_FRAME_RESERVED in their flags
*/
arch_reserved_pages_update();
#endif /* CONFIG_ARCH_HAS_RESERVED_PAGE_FRAMES */
#ifdef CONFIG_LINKER_GENERIC_SECTIONS_PRESENT_AT_BOOT
/* All pages composing the Zephyr image are mapped at boot in a
* predictable way. This can change at runtime.
*/
VIRT_FOREACH(Z_KERNEL_VIRT_START, Z_KERNEL_VIRT_SIZE, addr)
{
pf = z_phys_to_page_frame(Z_BOOT_VIRT_TO_PHYS(addr));
frame_mapped_set(pf, addr);
/* TODO: for now we pin the whole Zephyr image. Demand paging
* currently tested with anonymously-mapped pages which are not
* pinned.
*
* We will need to setup linker regions for a subset of kernel
* code/data pages which are pinned in memory and
* may not be evicted. This will contain critical CPU data
* structures, and any code used to perform page fault
* handling, page-ins, etc.
*/
pf->flags |= Z_PAGE_FRAME_PINNED;
}
#endif /* CONFIG_LINKER_GENERIC_SECTIONS_PRESENT_AT_BOOT */
#ifdef CONFIG_LINKER_USE_BOOT_SECTION
/* Pin the boot section to prevent it from being swapped out during
* boot process. Will be un-pinned once boot process completes.
*/
mark_linker_section_pinned(lnkr_boot_start, lnkr_boot_end, true);
#endif
#ifdef CONFIG_LINKER_USE_PINNED_SECTION
/* Pin the page frames correspondng to the pinned symbols */
mark_linker_section_pinned(lnkr_pinned_start, lnkr_pinned_end, true);
#endif
/* Any remaining pages that aren't mapped, reserved, or pinned get
* added to the free pages list
*/
Z_PAGE_FRAME_FOREACH(phys, pf) {
if (z_page_frame_is_available(pf)) {
free_page_frame_list_put(pf);
}
}
LOG_DBG("free page frames: %zu", z_free_page_count);
#ifdef CONFIG_DEMAND_PAGING
#ifdef CONFIG_DEMAND_PAGING_TIMING_HISTOGRAM
z_paging_histogram_init();
#endif
k_mem_paging_backing_store_init();
k_mem_paging_eviction_init();
#endif
#if __ASSERT_ON
page_frames_initialized = true;
#endif
k_spin_unlock(&z_mm_lock, key);
#ifndef CONFIG_LINKER_GENERIC_SECTIONS_PRESENT_AT_BOOT
/* If BSS section is not present in memory at boot,
* it would not have been cleared. This needs to be
* done now since paging mechanism has been initialized
* and the BSS pages can be brought into physical
* memory to be cleared.
*/
z_bss_zero();
#endif
}
void z_mem_manage_boot_finish(void)
{
#ifdef CONFIG_LINKER_USE_BOOT_SECTION
/* At the end of boot process, unpin the boot sections
* as they don't need to be in memory all the time anymore.
*/
mark_linker_section_pinned(lnkr_boot_start, lnkr_boot_end, false);
#endif
}
#ifdef CONFIG_DEMAND_PAGING
#ifdef CONFIG_DEMAND_PAGING_STATS
struct k_mem_paging_stats_t paging_stats;
extern struct k_mem_paging_histogram_t z_paging_histogram_eviction;
extern struct k_mem_paging_histogram_t z_paging_histogram_backing_store_page_in;
extern struct k_mem_paging_histogram_t z_paging_histogram_backing_store_page_out;