Merge tag 'microblaze-3.17-rc5' of git://git.monstr.eu/linux-2.6-microblaze
[deliverable/linux.git] / mm / memblock.c
CommitLineData
95f72d1e
YL
1/*
2 * Procedures for maintaining information about logical memory blocks.
3 *
4 * Peter Bergner, IBM Corp. June 2001.
5 * Copyright (C) 2001 Peter Bergner.
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 */
12
13#include <linux/kernel.h>
142b45a7 14#include <linux/slab.h>
95f72d1e
YL
15#include <linux/init.h>
16#include <linux/bitops.h>
449e8df3 17#include <linux/poison.h>
c196f76f 18#include <linux/pfn.h>
6d03b885
BH
19#include <linux/debugfs.h>
20#include <linux/seq_file.h>
95f72d1e
YL
21#include <linux/memblock.h>
22
79442ed1 23#include <asm-generic/sections.h>
26f09e9b
SS
24#include <linux/io.h>
25
26#include "internal.h"
79442ed1 27
fe091c20
TH
28static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock;
29static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock;
70210ed9
PH
30#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
31static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock;
32#endif
fe091c20
TH
33
34struct memblock memblock __initdata_memblock = {
35 .memory.regions = memblock_memory_init_regions,
36 .memory.cnt = 1, /* empty dummy entry */
37 .memory.max = INIT_MEMBLOCK_REGIONS,
38
39 .reserved.regions = memblock_reserved_init_regions,
40 .reserved.cnt = 1, /* empty dummy entry */
41 .reserved.max = INIT_MEMBLOCK_REGIONS,
42
70210ed9
PH
43#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
44 .physmem.regions = memblock_physmem_init_regions,
45 .physmem.cnt = 1, /* empty dummy entry */
46 .physmem.max = INIT_PHYSMEM_REGIONS,
47#endif
48
79442ed1 49 .bottom_up = false,
fe091c20
TH
50 .current_limit = MEMBLOCK_ALLOC_ANYWHERE,
51};
95f72d1e 52
10d06439 53int memblock_debug __initdata_memblock;
55ac590c
TC
54#ifdef CONFIG_MOVABLE_NODE
55bool movable_node_enabled __initdata_memblock = false;
56#endif
1aadc056 57static int memblock_can_resize __initdata_memblock;
181eb394
GS
58static int memblock_memory_in_slab __initdata_memblock = 0;
59static int memblock_reserved_in_slab __initdata_memblock = 0;
95f72d1e 60
142b45a7 61/* inline so we don't get a warning when pr_debug is compiled out */
c2233116
RP
62static __init_memblock const char *
63memblock_type_name(struct memblock_type *type)
142b45a7
BH
64{
65 if (type == &memblock.memory)
66 return "memory";
67 else if (type == &memblock.reserved)
68 return "reserved";
69 else
70 return "unknown";
71}
72
eb18f1b5
TH
73/* adjust *@size so that (@base + *@size) doesn't overflow, return new size */
74static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size)
75{
76 return *size = min(*size, (phys_addr_t)ULLONG_MAX - base);
77}
78
6ed311b2
BH
79/*
80 * Address comparison utilities
81 */
10d06439 82static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1,
2898cc4c 83 phys_addr_t base2, phys_addr_t size2)
95f72d1e
YL
84{
85 return ((base1 < (base2 + size2)) && (base2 < (base1 + size1)));
86}
87
2d7d3eb2
HS
88static long __init_memblock memblock_overlaps_region(struct memblock_type *type,
89 phys_addr_t base, phys_addr_t size)
6ed311b2
BH
90{
91 unsigned long i;
92
93 for (i = 0; i < type->cnt; i++) {
94 phys_addr_t rgnbase = type->regions[i].base;
95 phys_addr_t rgnsize = type->regions[i].size;
96 if (memblock_addrs_overlap(base, size, rgnbase, rgnsize))
97 break;
98 }
99
100 return (i < type->cnt) ? i : -1;
101}
102
79442ed1
TC
103/*
104 * __memblock_find_range_bottom_up - find free area utility in bottom-up
105 * @start: start of candidate range
106 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
107 * @size: size of free area to find
108 * @align: alignment of free area to find
b1154233 109 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
79442ed1
TC
110 *
111 * Utility called from memblock_find_in_range_node(), find free area bottom-up.
112 *
113 * RETURNS:
114 * Found address on success, 0 on failure.
115 */
116static phys_addr_t __init_memblock
117__memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
118 phys_addr_t size, phys_addr_t align, int nid)
119{
120 phys_addr_t this_start, this_end, cand;
121 u64 i;
122
123 for_each_free_mem_range(i, nid, &this_start, &this_end, NULL) {
124 this_start = clamp(this_start, start, end);
125 this_end = clamp(this_end, start, end);
126
127 cand = round_up(this_start, align);
128 if (cand < this_end && this_end - cand >= size)
129 return cand;
130 }
131
132 return 0;
133}
134
7bd0b0f0 135/**
1402899e 136 * __memblock_find_range_top_down - find free area utility, in top-down
7bd0b0f0
TH
137 * @start: start of candidate range
138 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
139 * @size: size of free area to find
140 * @align: alignment of free area to find
b1154233 141 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
7bd0b0f0 142 *
1402899e 143 * Utility called from memblock_find_in_range_node(), find free area top-down.
7bd0b0f0
TH
144 *
145 * RETURNS:
79442ed1 146 * Found address on success, 0 on failure.
6ed311b2 147 */
1402899e
TC
148static phys_addr_t __init_memblock
149__memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
150 phys_addr_t size, phys_addr_t align, int nid)
f7210e6c
TC
151{
152 phys_addr_t this_start, this_end, cand;
153 u64 i;
154
f7210e6c
TC
155 for_each_free_mem_range_reverse(i, nid, &this_start, &this_end, NULL) {
156 this_start = clamp(this_start, start, end);
157 this_end = clamp(this_end, start, end);
158
159 if (this_end < size)
160 continue;
161
162 cand = round_down(this_end - size, align);
163 if (cand >= this_start)
164 return cand;
165 }
1402899e 166
f7210e6c
TC
167 return 0;
168}
6ed311b2 169
1402899e
TC
170/**
171 * memblock_find_in_range_node - find free area in given range and node
1402899e
TC
172 * @size: size of free area to find
173 * @align: alignment of free area to find
87029ee9
GS
174 * @start: start of candidate range
175 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
b1154233 176 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1402899e
TC
177 *
178 * Find @size free area aligned to @align in the specified range and node.
179 *
79442ed1
TC
180 * When allocation direction is bottom-up, the @start should be greater
181 * than the end of the kernel image. Otherwise, it will be trimmed. The
182 * reason is that we want the bottom-up allocation just near the kernel
183 * image so it is highly likely that the allocated memory and the kernel
184 * will reside in the same node.
185 *
186 * If bottom-up allocation failed, will try to allocate memory top-down.
187 *
1402899e 188 * RETURNS:
79442ed1 189 * Found address on success, 0 on failure.
1402899e 190 */
87029ee9
GS
191phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
192 phys_addr_t align, phys_addr_t start,
193 phys_addr_t end, int nid)
1402899e 194{
0cfb8f0c 195 phys_addr_t kernel_end, ret;
79442ed1 196
1402899e
TC
197 /* pump up @end */
198 if (end == MEMBLOCK_ALLOC_ACCESSIBLE)
199 end = memblock.current_limit;
200
201 /* avoid allocating the first page */
202 start = max_t(phys_addr_t, start, PAGE_SIZE);
203 end = max(start, end);
79442ed1
TC
204 kernel_end = __pa_symbol(_end);
205
206 /*
207 * try bottom-up allocation only when bottom-up mode
208 * is set and @end is above the kernel image.
209 */
210 if (memblock_bottom_up() && end > kernel_end) {
211 phys_addr_t bottom_up_start;
212
213 /* make sure we will allocate above the kernel */
214 bottom_up_start = max(start, kernel_end);
215
216 /* ok, try bottom-up allocation first */
217 ret = __memblock_find_range_bottom_up(bottom_up_start, end,
218 size, align, nid);
219 if (ret)
220 return ret;
221
222 /*
223 * we always limit bottom-up allocation above the kernel,
224 * but top-down allocation doesn't have the limit, so
225 * retrying top-down allocation may succeed when bottom-up
226 * allocation failed.
227 *
228 * bottom-up allocation is expected to be fail very rarely,
229 * so we use WARN_ONCE() here to see the stack trace if
230 * fail happens.
231 */
232 WARN_ONCE(1, "memblock: bottom-up allocation failed, "
233 "memory hotunplug may be affected\n");
234 }
1402899e
TC
235
236 return __memblock_find_range_top_down(start, end, size, align, nid);
237}
238
7bd0b0f0
TH
239/**
240 * memblock_find_in_range - find free area in given range
241 * @start: start of candidate range
242 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
243 * @size: size of free area to find
244 * @align: alignment of free area to find
245 *
246 * Find @size free area aligned to @align in the specified range.
247 *
248 * RETURNS:
79442ed1 249 * Found address on success, 0 on failure.
fc769a8e 250 */
7bd0b0f0
TH
251phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start,
252 phys_addr_t end, phys_addr_t size,
253 phys_addr_t align)
6ed311b2 254{
87029ee9 255 return memblock_find_in_range_node(size, align, start, end,
b1154233 256 NUMA_NO_NODE);
6ed311b2
BH
257}
258
10d06439 259static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r)
95f72d1e 260{
1440c4e2 261 type->total_size -= type->regions[r].size;
7c0caeb8
TH
262 memmove(&type->regions[r], &type->regions[r + 1],
263 (type->cnt - (r + 1)) * sizeof(type->regions[r]));
e3239ff9 264 type->cnt--;
95f72d1e 265
8f7a6605
BH
266 /* Special case for empty arrays */
267 if (type->cnt == 0) {
1440c4e2 268 WARN_ON(type->total_size != 0);
8f7a6605
BH
269 type->cnt = 1;
270 type->regions[0].base = 0;
271 type->regions[0].size = 0;
66a20757 272 type->regions[0].flags = 0;
7c0caeb8 273 memblock_set_region_node(&type->regions[0], MAX_NUMNODES);
8f7a6605 274 }
95f72d1e
YL
275}
276
354f17e1
PH
277#ifdef CONFIG_ARCH_DISCARD_MEMBLOCK
278
29f67386
YL
279phys_addr_t __init_memblock get_allocated_memblock_reserved_regions_info(
280 phys_addr_t *addr)
281{
282 if (memblock.reserved.regions == memblock_reserved_init_regions)
283 return 0;
284
285 *addr = __pa(memblock.reserved.regions);
286
287 return PAGE_ALIGN(sizeof(struct memblock_region) *
288 memblock.reserved.max);
289}
290
5e270e25
PH
291phys_addr_t __init_memblock get_allocated_memblock_memory_regions_info(
292 phys_addr_t *addr)
293{
294 if (memblock.memory.regions == memblock_memory_init_regions)
295 return 0;
296
297 *addr = __pa(memblock.memory.regions);
298
299 return PAGE_ALIGN(sizeof(struct memblock_region) *
300 memblock.memory.max);
301}
302
303#endif
304
48c3b583
GP
305/**
306 * memblock_double_array - double the size of the memblock regions array
307 * @type: memblock type of the regions array being doubled
308 * @new_area_start: starting address of memory range to avoid overlap with
309 * @new_area_size: size of memory range to avoid overlap with
310 *
311 * Double the size of the @type regions array. If memblock is being used to
312 * allocate memory for a new reserved regions array and there is a previously
313 * allocated memory range [@new_area_start,@new_area_start+@new_area_size]
314 * waiting to be reserved, ensure the memory used by the new array does
315 * not overlap.
316 *
317 * RETURNS:
318 * 0 on success, -1 on failure.
319 */
320static int __init_memblock memblock_double_array(struct memblock_type *type,
321 phys_addr_t new_area_start,
322 phys_addr_t new_area_size)
142b45a7
BH
323{
324 struct memblock_region *new_array, *old_array;
29f67386 325 phys_addr_t old_alloc_size, new_alloc_size;
142b45a7
BH
326 phys_addr_t old_size, new_size, addr;
327 int use_slab = slab_is_available();
181eb394 328 int *in_slab;
142b45a7
BH
329
330 /* We don't allow resizing until we know about the reserved regions
331 * of memory that aren't suitable for allocation
332 */
333 if (!memblock_can_resize)
334 return -1;
335
142b45a7
BH
336 /* Calculate new doubled size */
337 old_size = type->max * sizeof(struct memblock_region);
338 new_size = old_size << 1;
29f67386
YL
339 /*
340 * We need to allocated new one align to PAGE_SIZE,
341 * so we can free them completely later.
342 */
343 old_alloc_size = PAGE_ALIGN(old_size);
344 new_alloc_size = PAGE_ALIGN(new_size);
142b45a7 345
181eb394
GS
346 /* Retrieve the slab flag */
347 if (type == &memblock.memory)
348 in_slab = &memblock_memory_in_slab;
349 else
350 in_slab = &memblock_reserved_in_slab;
351
142b45a7
BH
352 /* Try to find some space for it.
353 *
354 * WARNING: We assume that either slab_is_available() and we use it or
fd07383b
AM
355 * we use MEMBLOCK for allocations. That means that this is unsafe to
356 * use when bootmem is currently active (unless bootmem itself is
357 * implemented on top of MEMBLOCK which isn't the case yet)
142b45a7
BH
358 *
359 * This should however not be an issue for now, as we currently only
fd07383b
AM
360 * call into MEMBLOCK while it's still active, or much later when slab
361 * is active for memory hotplug operations
142b45a7
BH
362 */
363 if (use_slab) {
364 new_array = kmalloc(new_size, GFP_KERNEL);
1f5026a7 365 addr = new_array ? __pa(new_array) : 0;
4e2f0775 366 } else {
48c3b583
GP
367 /* only exclude range when trying to double reserved.regions */
368 if (type != &memblock.reserved)
369 new_area_start = new_area_size = 0;
370
371 addr = memblock_find_in_range(new_area_start + new_area_size,
372 memblock.current_limit,
29f67386 373 new_alloc_size, PAGE_SIZE);
48c3b583
GP
374 if (!addr && new_area_size)
375 addr = memblock_find_in_range(0,
fd07383b
AM
376 min(new_area_start, memblock.current_limit),
377 new_alloc_size, PAGE_SIZE);
48c3b583 378
15674868 379 new_array = addr ? __va(addr) : NULL;
4e2f0775 380 }
1f5026a7 381 if (!addr) {
142b45a7
BH
382 pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n",
383 memblock_type_name(type), type->max, type->max * 2);
384 return -1;
385 }
142b45a7 386
fd07383b
AM
387 memblock_dbg("memblock: %s is doubled to %ld at [%#010llx-%#010llx]",
388 memblock_type_name(type), type->max * 2, (u64)addr,
389 (u64)addr + new_size - 1);
ea9e4376 390
fd07383b
AM
391 /*
392 * Found space, we now need to move the array over before we add the
393 * reserved region since it may be our reserved array itself that is
394 * full.
142b45a7
BH
395 */
396 memcpy(new_array, type->regions, old_size);
397 memset(new_array + type->max, 0, old_size);
398 old_array = type->regions;
399 type->regions = new_array;
400 type->max <<= 1;
401
fd07383b 402 /* Free old array. We needn't free it if the array is the static one */
181eb394
GS
403 if (*in_slab)
404 kfree(old_array);
405 else if (old_array != memblock_memory_init_regions &&
406 old_array != memblock_reserved_init_regions)
29f67386 407 memblock_free(__pa(old_array), old_alloc_size);
142b45a7 408
fd07383b
AM
409 /*
410 * Reserve the new array if that comes from the memblock. Otherwise, we
411 * needn't do it
181eb394
GS
412 */
413 if (!use_slab)
29f67386 414 BUG_ON(memblock_reserve(addr, new_alloc_size));
181eb394
GS
415
416 /* Update slab flag */
417 *in_slab = use_slab;
418
142b45a7
BH
419 return 0;
420}
421
784656f9
TH
422/**
423 * memblock_merge_regions - merge neighboring compatible regions
424 * @type: memblock type to scan
425 *
426 * Scan @type and merge neighboring compatible regions.
427 */
428static void __init_memblock memblock_merge_regions(struct memblock_type *type)
95f72d1e 429{
784656f9 430 int i = 0;
95f72d1e 431
784656f9
TH
432 /* cnt never goes below 1 */
433 while (i < type->cnt - 1) {
434 struct memblock_region *this = &type->regions[i];
435 struct memblock_region *next = &type->regions[i + 1];
95f72d1e 436
7c0caeb8
TH
437 if (this->base + this->size != next->base ||
438 memblock_get_region_node(this) !=
66a20757
TC
439 memblock_get_region_node(next) ||
440 this->flags != next->flags) {
784656f9
TH
441 BUG_ON(this->base + this->size > next->base);
442 i++;
443 continue;
8f7a6605
BH
444 }
445
784656f9 446 this->size += next->size;
c0232ae8
LF
447 /* move forward from next + 1, index of which is i + 2 */
448 memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next));
784656f9 449 type->cnt--;
95f72d1e 450 }
784656f9 451}
95f72d1e 452
784656f9
TH
453/**
454 * memblock_insert_region - insert new memblock region
209ff86d
TC
455 * @type: memblock type to insert into
456 * @idx: index for the insertion point
457 * @base: base address of the new region
458 * @size: size of the new region
459 * @nid: node id of the new region
66a20757 460 * @flags: flags of the new region
784656f9
TH
461 *
462 * Insert new memblock region [@base,@base+@size) into @type at @idx.
463 * @type must already have extra room to accomodate the new region.
464 */
465static void __init_memblock memblock_insert_region(struct memblock_type *type,
466 int idx, phys_addr_t base,
66a20757
TC
467 phys_addr_t size,
468 int nid, unsigned long flags)
784656f9
TH
469{
470 struct memblock_region *rgn = &type->regions[idx];
471
472 BUG_ON(type->cnt >= type->max);
473 memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn));
474 rgn->base = base;
475 rgn->size = size;
66a20757 476 rgn->flags = flags;
7c0caeb8 477 memblock_set_region_node(rgn, nid);
784656f9 478 type->cnt++;
1440c4e2 479 type->total_size += size;
784656f9
TH
480}
481
482/**
f1af9d3a 483 * memblock_add_range - add new memblock region
784656f9
TH
484 * @type: memblock type to add new region into
485 * @base: base address of the new region
486 * @size: size of the new region
7fb0bc3f 487 * @nid: nid of the new region
66a20757 488 * @flags: flags of the new region
784656f9
TH
489 *
490 * Add new memblock region [@base,@base+@size) into @type. The new region
491 * is allowed to overlap with existing ones - overlaps don't affect already
492 * existing regions. @type is guaranteed to be minimal (all neighbouring
493 * compatible regions are merged) after the addition.
494 *
495 * RETURNS:
496 * 0 on success, -errno on failure.
497 */
f1af9d3a 498int __init_memblock memblock_add_range(struct memblock_type *type,
66a20757
TC
499 phys_addr_t base, phys_addr_t size,
500 int nid, unsigned long flags)
784656f9
TH
501{
502 bool insert = false;
eb18f1b5
TH
503 phys_addr_t obase = base;
504 phys_addr_t end = base + memblock_cap_size(base, &size);
784656f9
TH
505 int i, nr_new;
506
b3dc627c
TH
507 if (!size)
508 return 0;
509
784656f9
TH
510 /* special case for empty array */
511 if (type->regions[0].size == 0) {
1440c4e2 512 WARN_ON(type->cnt != 1 || type->total_size);
8f7a6605
BH
513 type->regions[0].base = base;
514 type->regions[0].size = size;
66a20757 515 type->regions[0].flags = flags;
7fb0bc3f 516 memblock_set_region_node(&type->regions[0], nid);
1440c4e2 517 type->total_size = size;
8f7a6605 518 return 0;
95f72d1e 519 }
784656f9
TH
520repeat:
521 /*
522 * The following is executed twice. Once with %false @insert and
523 * then with %true. The first counts the number of regions needed
524 * to accomodate the new area. The second actually inserts them.
142b45a7 525 */
784656f9
TH
526 base = obase;
527 nr_new = 0;
95f72d1e 528
784656f9
TH
529 for (i = 0; i < type->cnt; i++) {
530 struct memblock_region *rgn = &type->regions[i];
531 phys_addr_t rbase = rgn->base;
532 phys_addr_t rend = rbase + rgn->size;
533
534 if (rbase >= end)
95f72d1e 535 break;
784656f9
TH
536 if (rend <= base)
537 continue;
538 /*
539 * @rgn overlaps. If it separates the lower part of new
540 * area, insert that portion.
541 */
542 if (rbase > base) {
543 nr_new++;
544 if (insert)
545 memblock_insert_region(type, i++, base,
66a20757
TC
546 rbase - base, nid,
547 flags);
95f72d1e 548 }
784656f9
TH
549 /* area below @rend is dealt with, forget about it */
550 base = min(rend, end);
95f72d1e 551 }
784656f9
TH
552
553 /* insert the remaining portion */
554 if (base < end) {
555 nr_new++;
556 if (insert)
66a20757
TC
557 memblock_insert_region(type, i, base, end - base,
558 nid, flags);
95f72d1e 559 }
95f72d1e 560
784656f9
TH
561 /*
562 * If this was the first round, resize array and repeat for actual
563 * insertions; otherwise, merge and return.
142b45a7 564 */
784656f9
TH
565 if (!insert) {
566 while (type->cnt + nr_new > type->max)
48c3b583 567 if (memblock_double_array(type, obase, size) < 0)
784656f9
TH
568 return -ENOMEM;
569 insert = true;
570 goto repeat;
571 } else {
572 memblock_merge_regions(type);
573 return 0;
142b45a7 574 }
95f72d1e
YL
575}
576
7fb0bc3f
TH
577int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size,
578 int nid)
579{
f1af9d3a 580 return memblock_add_range(&memblock.memory, base, size, nid, 0);
7fb0bc3f
TH
581}
582
581adcbe 583int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
95f72d1e 584{
f1af9d3a 585 return memblock_add_range(&memblock.memory, base, size,
66a20757 586 MAX_NUMNODES, 0);
95f72d1e
YL
587}
588
6a9ceb31
TH
589/**
590 * memblock_isolate_range - isolate given range into disjoint memblocks
591 * @type: memblock type to isolate range for
592 * @base: base of range to isolate
593 * @size: size of range to isolate
594 * @start_rgn: out parameter for the start of isolated region
595 * @end_rgn: out parameter for the end of isolated region
596 *
597 * Walk @type and ensure that regions don't cross the boundaries defined by
598 * [@base,@base+@size). Crossing regions are split at the boundaries,
599 * which may create at most two more regions. The index of the first
600 * region inside the range is returned in *@start_rgn and end in *@end_rgn.
601 *
602 * RETURNS:
603 * 0 on success, -errno on failure.
604 */
605static int __init_memblock memblock_isolate_range(struct memblock_type *type,
606 phys_addr_t base, phys_addr_t size,
607 int *start_rgn, int *end_rgn)
608{
eb18f1b5 609 phys_addr_t end = base + memblock_cap_size(base, &size);
6a9ceb31
TH
610 int i;
611
612 *start_rgn = *end_rgn = 0;
613
b3dc627c
TH
614 if (!size)
615 return 0;
616
6a9ceb31
TH
617 /* we'll create at most two more regions */
618 while (type->cnt + 2 > type->max)
48c3b583 619 if (memblock_double_array(type, base, size) < 0)
6a9ceb31
TH
620 return -ENOMEM;
621
622 for (i = 0; i < type->cnt; i++) {
623 struct memblock_region *rgn = &type->regions[i];
624 phys_addr_t rbase = rgn->base;
625 phys_addr_t rend = rbase + rgn->size;
626
627 if (rbase >= end)
628 break;
629 if (rend <= base)
630 continue;
631
632 if (rbase < base) {
633 /*
634 * @rgn intersects from below. Split and continue
635 * to process the next region - the new top half.
636 */
637 rgn->base = base;
1440c4e2
TH
638 rgn->size -= base - rbase;
639 type->total_size -= base - rbase;
6a9ceb31 640 memblock_insert_region(type, i, rbase, base - rbase,
66a20757
TC
641 memblock_get_region_node(rgn),
642 rgn->flags);
6a9ceb31
TH
643 } else if (rend > end) {
644 /*
645 * @rgn intersects from above. Split and redo the
646 * current region - the new bottom half.
647 */
648 rgn->base = end;
1440c4e2
TH
649 rgn->size -= end - rbase;
650 type->total_size -= end - rbase;
6a9ceb31 651 memblock_insert_region(type, i--, rbase, end - rbase,
66a20757
TC
652 memblock_get_region_node(rgn),
653 rgn->flags);
6a9ceb31
TH
654 } else {
655 /* @rgn is fully contained, record it */
656 if (!*end_rgn)
657 *start_rgn = i;
658 *end_rgn = i + 1;
659 }
660 }
661
662 return 0;
663}
6a9ceb31 664
f1af9d3a
PH
665int __init_memblock memblock_remove_range(struct memblock_type *type,
666 phys_addr_t base, phys_addr_t size)
95f72d1e 667{
71936180
TH
668 int start_rgn, end_rgn;
669 int i, ret;
95f72d1e 670
71936180
TH
671 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
672 if (ret)
673 return ret;
95f72d1e 674
71936180
TH
675 for (i = end_rgn - 1; i >= start_rgn; i--)
676 memblock_remove_region(type, i);
8f7a6605 677 return 0;
95f72d1e
YL
678}
679
581adcbe 680int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size)
95f72d1e 681{
f1af9d3a 682 return memblock_remove_range(&memblock.memory, base, size);
95f72d1e
YL
683}
684
f1af9d3a 685
581adcbe 686int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size)
95f72d1e 687{
24aa0788 688 memblock_dbg(" memblock_free: [%#016llx-%#016llx] %pF\n",
a150439c 689 (unsigned long long)base,
931d13f5 690 (unsigned long long)base + size - 1,
a150439c 691 (void *)_RET_IP_);
24aa0788 692
aedf95ea 693 kmemleak_free_part(__va(base), size);
f1af9d3a 694 return memblock_remove_range(&memblock.reserved, base, size);
95f72d1e
YL
695}
696
66a20757
TC
697static int __init_memblock memblock_reserve_region(phys_addr_t base,
698 phys_addr_t size,
699 int nid,
700 unsigned long flags)
95f72d1e 701{
e3239ff9 702 struct memblock_type *_rgn = &memblock.reserved;
95f72d1e 703
66a20757 704 memblock_dbg("memblock_reserve: [%#016llx-%#016llx] flags %#02lx %pF\n",
a150439c 705 (unsigned long long)base,
931d13f5 706 (unsigned long long)base + size - 1,
66a20757
TC
707 flags, (void *)_RET_IP_);
708
f1af9d3a 709 return memblock_add_range(_rgn, base, size, nid, flags);
66a20757 710}
95f72d1e 711
66a20757
TC
712int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size)
713{
714 return memblock_reserve_region(base, size, MAX_NUMNODES, 0);
95f72d1e
YL
715}
716
66b16edf
TC
717/**
718 * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG.
719 * @base: the base phys addr of the region
720 * @size: the size of the region
721 *
722 * This function isolates region [@base, @base + @size), and mark it with flag
723 * MEMBLOCK_HOTPLUG.
724 *
725 * Return 0 on succees, -errno on failure.
726 */
727int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size)
728{
729 struct memblock_type *type = &memblock.memory;
730 int i, ret, start_rgn, end_rgn;
731
732 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
733 if (ret)
734 return ret;
735
736 for (i = start_rgn; i < end_rgn; i++)
737 memblock_set_region_flags(&type->regions[i], MEMBLOCK_HOTPLUG);
738
739 memblock_merge_regions(type);
740 return 0;
741}
742
743/**
744 * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region.
745 * @base: the base phys addr of the region
746 * @size: the size of the region
747 *
748 * This function isolates region [@base, @base + @size), and clear flag
749 * MEMBLOCK_HOTPLUG for the isolated regions.
750 *
751 * Return 0 on succees, -errno on failure.
752 */
753int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size)
754{
755 struct memblock_type *type = &memblock.memory;
756 int i, ret, start_rgn, end_rgn;
757
758 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
759 if (ret)
760 return ret;
761
762 for (i = start_rgn; i < end_rgn; i++)
763 memblock_clear_region_flags(&type->regions[i],
764 MEMBLOCK_HOTPLUG);
765
766 memblock_merge_regions(type);
767 return 0;
768}
769
35fd0808 770/**
f1af9d3a 771 * __next__mem_range - next function for for_each_free_mem_range() etc.
35fd0808 772 * @idx: pointer to u64 loop variable
b1154233 773 * @nid: node selector, %NUMA_NO_NODE for all nodes
f1af9d3a
PH
774 * @type_a: pointer to memblock_type from where the range is taken
775 * @type_b: pointer to memblock_type which excludes memory from being taken
dad7557e
WL
776 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
777 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
778 * @out_nid: ptr to int for nid of the range, can be %NULL
35fd0808 779 *
f1af9d3a 780 * Find the first area from *@idx which matches @nid, fill the out
35fd0808 781 * parameters, and update *@idx for the next iteration. The lower 32bit of
f1af9d3a
PH
782 * *@idx contains index into type_a and the upper 32bit indexes the
783 * areas before each region in type_b. For example, if type_b regions
35fd0808
TH
784 * look like the following,
785 *
786 * 0:[0-16), 1:[32-48), 2:[128-130)
787 *
788 * The upper 32bit indexes the following regions.
789 *
790 * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX)
791 *
792 * As both region arrays are sorted, the function advances the two indices
793 * in lockstep and returns each intersection.
794 */
f1af9d3a
PH
795void __init_memblock __next_mem_range(u64 *idx, int nid,
796 struct memblock_type *type_a,
797 struct memblock_type *type_b,
798 phys_addr_t *out_start,
799 phys_addr_t *out_end, int *out_nid)
35fd0808 800{
f1af9d3a
PH
801 int idx_a = *idx & 0xffffffff;
802 int idx_b = *idx >> 32;
b1154233 803
f1af9d3a
PH
804 if (WARN_ONCE(nid == MAX_NUMNODES,
805 "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
560dca27 806 nid = NUMA_NO_NODE;
35fd0808 807
f1af9d3a
PH
808 for (; idx_a < type_a->cnt; idx_a++) {
809 struct memblock_region *m = &type_a->regions[idx_a];
810
35fd0808
TH
811 phys_addr_t m_start = m->base;
812 phys_addr_t m_end = m->base + m->size;
f1af9d3a 813 int m_nid = memblock_get_region_node(m);
35fd0808
TH
814
815 /* only memory regions are associated with nodes, check it */
f1af9d3a 816 if (nid != NUMA_NO_NODE && nid != m_nid)
35fd0808
TH
817 continue;
818
f1af9d3a
PH
819 if (!type_b) {
820 if (out_start)
821 *out_start = m_start;
822 if (out_end)
823 *out_end = m_end;
824 if (out_nid)
825 *out_nid = m_nid;
826 idx_a++;
827 *idx = (u32)idx_a | (u64)idx_b << 32;
828 return;
829 }
830
831 /* scan areas before each reservation */
832 for (; idx_b < type_b->cnt + 1; idx_b++) {
833 struct memblock_region *r;
834 phys_addr_t r_start;
835 phys_addr_t r_end;
836
837 r = &type_b->regions[idx_b];
838 r_start = idx_b ? r[-1].base + r[-1].size : 0;
839 r_end = idx_b < type_b->cnt ?
840 r->base : ULLONG_MAX;
35fd0808 841
f1af9d3a
PH
842 /*
843 * if idx_b advanced past idx_a,
844 * break out to advance idx_a
845 */
35fd0808
TH
846 if (r_start >= m_end)
847 break;
848 /* if the two regions intersect, we're done */
849 if (m_start < r_end) {
850 if (out_start)
f1af9d3a
PH
851 *out_start =
852 max(m_start, r_start);
35fd0808
TH
853 if (out_end)
854 *out_end = min(m_end, r_end);
855 if (out_nid)
f1af9d3a 856 *out_nid = m_nid;
35fd0808 857 /*
f1af9d3a
PH
858 * The region which ends first is
859 * advanced for the next iteration.
35fd0808
TH
860 */
861 if (m_end <= r_end)
f1af9d3a 862 idx_a++;
35fd0808 863 else
f1af9d3a
PH
864 idx_b++;
865 *idx = (u32)idx_a | (u64)idx_b << 32;
35fd0808
TH
866 return;
867 }
868 }
869 }
870
871 /* signal end of iteration */
872 *idx = ULLONG_MAX;
873}
874
7bd0b0f0 875/**
f1af9d3a
PH
876 * __next_mem_range_rev - generic next function for for_each_*_range_rev()
877 *
878 * Finds the next range from type_a which is not marked as unsuitable
879 * in type_b.
880 *
7bd0b0f0 881 * @idx: pointer to u64 loop variable
b1154233 882 * @nid: nid: node selector, %NUMA_NO_NODE for all nodes
f1af9d3a
PH
883 * @type_a: pointer to memblock_type from where the range is taken
884 * @type_b: pointer to memblock_type which excludes memory from being taken
dad7557e
WL
885 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
886 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
887 * @out_nid: ptr to int for nid of the range, can be %NULL
7bd0b0f0 888 *
f1af9d3a 889 * Reverse of __next_mem_range().
7bd0b0f0 890 */
f1af9d3a
PH
891void __init_memblock __next_mem_range_rev(u64 *idx, int nid,
892 struct memblock_type *type_a,
893 struct memblock_type *type_b,
894 phys_addr_t *out_start,
895 phys_addr_t *out_end, int *out_nid)
7bd0b0f0 896{
f1af9d3a
PH
897 int idx_a = *idx & 0xffffffff;
898 int idx_b = *idx >> 32;
b1154233 899
560dca27
GS
900 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
901 nid = NUMA_NO_NODE;
7bd0b0f0
TH
902
903 if (*idx == (u64)ULLONG_MAX) {
f1af9d3a
PH
904 idx_a = type_a->cnt - 1;
905 idx_b = type_b->cnt;
7bd0b0f0
TH
906 }
907
f1af9d3a
PH
908 for (; idx_a >= 0; idx_a--) {
909 struct memblock_region *m = &type_a->regions[idx_a];
910
7bd0b0f0
TH
911 phys_addr_t m_start = m->base;
912 phys_addr_t m_end = m->base + m->size;
f1af9d3a 913 int m_nid = memblock_get_region_node(m);
7bd0b0f0
TH
914
915 /* only memory regions are associated with nodes, check it */
f1af9d3a 916 if (nid != NUMA_NO_NODE && nid != m_nid)
7bd0b0f0
TH
917 continue;
918
55ac590c
TC
919 /* skip hotpluggable memory regions if needed */
920 if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
921 continue;
922
f1af9d3a
PH
923 if (!type_b) {
924 if (out_start)
925 *out_start = m_start;
926 if (out_end)
927 *out_end = m_end;
928 if (out_nid)
929 *out_nid = m_nid;
930 idx_a++;
931 *idx = (u32)idx_a | (u64)idx_b << 32;
932 return;
933 }
934
935 /* scan areas before each reservation */
936 for (; idx_b >= 0; idx_b--) {
937 struct memblock_region *r;
938 phys_addr_t r_start;
939 phys_addr_t r_end;
940
941 r = &type_b->regions[idx_b];
942 r_start = idx_b ? r[-1].base + r[-1].size : 0;
943 r_end = idx_b < type_b->cnt ?
944 r->base : ULLONG_MAX;
945 /*
946 * if idx_b advanced past idx_a,
947 * break out to advance idx_a
948 */
7bd0b0f0 949
7bd0b0f0
TH
950 if (r_end <= m_start)
951 break;
952 /* if the two regions intersect, we're done */
953 if (m_end > r_start) {
954 if (out_start)
955 *out_start = max(m_start, r_start);
956 if (out_end)
957 *out_end = min(m_end, r_end);
958 if (out_nid)
f1af9d3a 959 *out_nid = m_nid;
7bd0b0f0 960 if (m_start >= r_start)
f1af9d3a 961 idx_a--;
7bd0b0f0 962 else
f1af9d3a
PH
963 idx_b--;
964 *idx = (u32)idx_a | (u64)idx_b << 32;
7bd0b0f0
TH
965 return;
966 }
967 }
968 }
f1af9d3a 969 /* signal end of iteration */
7bd0b0f0
TH
970 *idx = ULLONG_MAX;
971}
972
7c0caeb8
TH
973#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
974/*
975 * Common iterator interface used to define for_each_mem_range().
976 */
977void __init_memblock __next_mem_pfn_range(int *idx, int nid,
978 unsigned long *out_start_pfn,
979 unsigned long *out_end_pfn, int *out_nid)
980{
981 struct memblock_type *type = &memblock.memory;
982 struct memblock_region *r;
983
984 while (++*idx < type->cnt) {
985 r = &type->regions[*idx];
986
987 if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size))
988 continue;
989 if (nid == MAX_NUMNODES || nid == r->nid)
990 break;
991 }
992 if (*idx >= type->cnt) {
993 *idx = -1;
994 return;
995 }
996
997 if (out_start_pfn)
998 *out_start_pfn = PFN_UP(r->base);
999 if (out_end_pfn)
1000 *out_end_pfn = PFN_DOWN(r->base + r->size);
1001 if (out_nid)
1002 *out_nid = r->nid;
1003}
1004
1005/**
1006 * memblock_set_node - set node ID on memblock regions
1007 * @base: base of area to set node ID for
1008 * @size: size of area to set node ID for
e7e8de59 1009 * @type: memblock type to set node ID for
7c0caeb8
TH
1010 * @nid: node ID to set
1011 *
e7e8de59 1012 * Set the nid of memblock @type regions in [@base,@base+@size) to @nid.
7c0caeb8
TH
1013 * Regions which cross the area boundaries are split as necessary.
1014 *
1015 * RETURNS:
1016 * 0 on success, -errno on failure.
1017 */
1018int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size,
e7e8de59 1019 struct memblock_type *type, int nid)
7c0caeb8 1020{
6a9ceb31
TH
1021 int start_rgn, end_rgn;
1022 int i, ret;
7c0caeb8 1023
6a9ceb31
TH
1024 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
1025 if (ret)
1026 return ret;
7c0caeb8 1027
6a9ceb31 1028 for (i = start_rgn; i < end_rgn; i++)
e9d24ad3 1029 memblock_set_region_node(&type->regions[i], nid);
7c0caeb8
TH
1030
1031 memblock_merge_regions(type);
1032 return 0;
1033}
1034#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1035
2bfc2862
AM
1036static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
1037 phys_addr_t align, phys_addr_t start,
1038 phys_addr_t end, int nid)
95f72d1e 1039{
6ed311b2 1040 phys_addr_t found;
95f72d1e 1041
79f40fab
GS
1042 if (!align)
1043 align = SMP_CACHE_BYTES;
94f3d3af 1044
2bfc2862 1045 found = memblock_find_in_range_node(size, align, start, end, nid);
aedf95ea
CM
1046 if (found && !memblock_reserve(found, size)) {
1047 /*
1048 * The min_count is set to 0 so that memblock allocations are
1049 * never reported as leaks.
1050 */
1051 kmemleak_alloc(__va(found), size, 0, 0);
6ed311b2 1052 return found;
aedf95ea 1053 }
6ed311b2 1054 return 0;
95f72d1e
YL
1055}
1056
2bfc2862
AM
1057phys_addr_t __init memblock_alloc_range(phys_addr_t size, phys_addr_t align,
1058 phys_addr_t start, phys_addr_t end)
1059{
1060 return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE);
1061}
1062
1063static phys_addr_t __init memblock_alloc_base_nid(phys_addr_t size,
1064 phys_addr_t align, phys_addr_t max_addr,
1065 int nid)
1066{
1067 return memblock_alloc_range_nid(size, align, 0, max_addr, nid);
1068}
1069
7bd0b0f0
TH
1070phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid)
1071{
1072 return memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE, nid);
1073}
1074
1075phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
1076{
b1154233 1077 return memblock_alloc_base_nid(size, align, max_addr, NUMA_NO_NODE);
7bd0b0f0
TH
1078}
1079
6ed311b2 1080phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
95f72d1e 1081{
6ed311b2
BH
1082 phys_addr_t alloc;
1083
1084 alloc = __memblock_alloc_base(size, align, max_addr);
1085
1086 if (alloc == 0)
1087 panic("ERROR: Failed to allocate 0x%llx bytes below 0x%llx.\n",
1088 (unsigned long long) size, (unsigned long long) max_addr);
1089
1090 return alloc;
95f72d1e
YL
1091}
1092
6ed311b2 1093phys_addr_t __init memblock_alloc(phys_addr_t size, phys_addr_t align)
95f72d1e 1094{
6ed311b2
BH
1095 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
1096}
95f72d1e 1097
9d1e2492
BH
1098phys_addr_t __init memblock_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid)
1099{
1100 phys_addr_t res = memblock_alloc_nid(size, align, nid);
1101
1102 if (res)
1103 return res;
15fb0972 1104 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
95f72d1e
YL
1105}
1106
26f09e9b
SS
1107/**
1108 * memblock_virt_alloc_internal - allocate boot memory block
1109 * @size: size of memory block to be allocated in bytes
1110 * @align: alignment of the region and block's size
1111 * @min_addr: the lower bound of the memory region to allocate (phys address)
1112 * @max_addr: the upper bound of the memory region to allocate (phys address)
1113 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1114 *
1115 * The @min_addr limit is dropped if it can not be satisfied and the allocation
1116 * will fall back to memory below @min_addr. Also, allocation may fall back
1117 * to any node in the system if the specified node can not
1118 * hold the requested memory.
1119 *
1120 * The allocation is performed from memory region limited by
1121 * memblock.current_limit if @max_addr == %BOOTMEM_ALLOC_ACCESSIBLE.
1122 *
1123 * The memory block is aligned on SMP_CACHE_BYTES if @align == 0.
1124 *
1125 * The phys address of allocated boot memory block is converted to virtual and
1126 * allocated memory is reset to 0.
1127 *
1128 * In addition, function sets the min_count to 0 using kmemleak_alloc for
1129 * allocated boot memory block, so that it is never reported as leaks.
1130 *
1131 * RETURNS:
1132 * Virtual address of allocated memory block on success, NULL on failure.
1133 */
1134static void * __init memblock_virt_alloc_internal(
1135 phys_addr_t size, phys_addr_t align,
1136 phys_addr_t min_addr, phys_addr_t max_addr,
1137 int nid)
1138{
1139 phys_addr_t alloc;
1140 void *ptr;
1141
560dca27
GS
1142 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
1143 nid = NUMA_NO_NODE;
26f09e9b
SS
1144
1145 /*
1146 * Detect any accidental use of these APIs after slab is ready, as at
1147 * this moment memblock may be deinitialized already and its
1148 * internal data may be destroyed (after execution of free_all_bootmem)
1149 */
1150 if (WARN_ON_ONCE(slab_is_available()))
1151 return kzalloc_node(size, GFP_NOWAIT, nid);
1152
1153 if (!align)
1154 align = SMP_CACHE_BYTES;
1155
f544e14f
YL
1156 if (max_addr > memblock.current_limit)
1157 max_addr = memblock.current_limit;
1158
26f09e9b
SS
1159again:
1160 alloc = memblock_find_in_range_node(size, align, min_addr, max_addr,
1161 nid);
1162 if (alloc)
1163 goto done;
1164
1165 if (nid != NUMA_NO_NODE) {
1166 alloc = memblock_find_in_range_node(size, align, min_addr,
1167 max_addr, NUMA_NO_NODE);
1168 if (alloc)
1169 goto done;
1170 }
1171
1172 if (min_addr) {
1173 min_addr = 0;
1174 goto again;
1175 } else {
1176 goto error;
1177 }
1178
1179done:
1180 memblock_reserve(alloc, size);
1181 ptr = phys_to_virt(alloc);
1182 memset(ptr, 0, size);
1183
1184 /*
1185 * The min_count is set to 0 so that bootmem allocated blocks
1186 * are never reported as leaks. This is because many of these blocks
1187 * are only referred via the physical address which is not
1188 * looked up by kmemleak.
1189 */
1190 kmemleak_alloc(ptr, size, 0, 0);
1191
1192 return ptr;
1193
1194error:
1195 return NULL;
1196}
1197
1198/**
1199 * memblock_virt_alloc_try_nid_nopanic - allocate boot memory block
1200 * @size: size of memory block to be allocated in bytes
1201 * @align: alignment of the region and block's size
1202 * @min_addr: the lower bound of the memory region from where the allocation
1203 * is preferred (phys address)
1204 * @max_addr: the upper bound of the memory region from where the allocation
1205 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1206 * allocate only from memory limited by memblock.current_limit value
1207 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1208 *
1209 * Public version of _memblock_virt_alloc_try_nid_nopanic() which provides
1210 * additional debug information (including caller info), if enabled.
1211 *
1212 * RETURNS:
1213 * Virtual address of allocated memory block on success, NULL on failure.
1214 */
1215void * __init memblock_virt_alloc_try_nid_nopanic(
1216 phys_addr_t size, phys_addr_t align,
1217 phys_addr_t min_addr, phys_addr_t max_addr,
1218 int nid)
1219{
1220 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1221 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1222 (u64)max_addr, (void *)_RET_IP_);
1223 return memblock_virt_alloc_internal(size, align, min_addr,
1224 max_addr, nid);
1225}
1226
1227/**
1228 * memblock_virt_alloc_try_nid - allocate boot memory block with panicking
1229 * @size: size of memory block to be allocated in bytes
1230 * @align: alignment of the region and block's size
1231 * @min_addr: the lower bound of the memory region from where the allocation
1232 * is preferred (phys address)
1233 * @max_addr: the upper bound of the memory region from where the allocation
1234 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1235 * allocate only from memory limited by memblock.current_limit value
1236 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1237 *
1238 * Public panicking version of _memblock_virt_alloc_try_nid_nopanic()
1239 * which provides debug information (including caller info), if enabled,
1240 * and panics if the request can not be satisfied.
1241 *
1242 * RETURNS:
1243 * Virtual address of allocated memory block on success, NULL on failure.
1244 */
1245void * __init memblock_virt_alloc_try_nid(
1246 phys_addr_t size, phys_addr_t align,
1247 phys_addr_t min_addr, phys_addr_t max_addr,
1248 int nid)
1249{
1250 void *ptr;
1251
1252 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1253 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1254 (u64)max_addr, (void *)_RET_IP_);
1255 ptr = memblock_virt_alloc_internal(size, align,
1256 min_addr, max_addr, nid);
1257 if (ptr)
1258 return ptr;
1259
1260 panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx\n",
1261 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1262 (u64)max_addr);
1263 return NULL;
1264}
1265
1266/**
1267 * __memblock_free_early - free boot memory block
1268 * @base: phys starting address of the boot memory block
1269 * @size: size of the boot memory block in bytes
1270 *
1271 * Free boot memory block previously allocated by memblock_virt_alloc_xx() API.
1272 * The freeing memory will not be released to the buddy allocator.
1273 */
1274void __init __memblock_free_early(phys_addr_t base, phys_addr_t size)
1275{
1276 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1277 __func__, (u64)base, (u64)base + size - 1,
1278 (void *)_RET_IP_);
1279 kmemleak_free_part(__va(base), size);
f1af9d3a 1280 memblock_remove_range(&memblock.reserved, base, size);
26f09e9b
SS
1281}
1282
1283/*
1284 * __memblock_free_late - free bootmem block pages directly to buddy allocator
1285 * @addr: phys starting address of the boot memory block
1286 * @size: size of the boot memory block in bytes
1287 *
1288 * This is only useful when the bootmem allocator has already been torn
1289 * down, but we are still initializing the system. Pages are released directly
1290 * to the buddy allocator, no bootmem metadata is updated because it is gone.
1291 */
1292void __init __memblock_free_late(phys_addr_t base, phys_addr_t size)
1293{
1294 u64 cursor, end;
1295
1296 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1297 __func__, (u64)base, (u64)base + size - 1,
1298 (void *)_RET_IP_);
1299 kmemleak_free_part(__va(base), size);
1300 cursor = PFN_UP(base);
1301 end = PFN_DOWN(base + size);
1302
1303 for (; cursor < end; cursor++) {
1304 __free_pages_bootmem(pfn_to_page(cursor), 0);
1305 totalram_pages++;
1306 }
1307}
9d1e2492
BH
1308
1309/*
1310 * Remaining API functions
1311 */
1312
2898cc4c 1313phys_addr_t __init memblock_phys_mem_size(void)
95f72d1e 1314{
1440c4e2 1315 return memblock.memory.total_size;
95f72d1e
YL
1316}
1317
595ad9af
YL
1318phys_addr_t __init memblock_mem_size(unsigned long limit_pfn)
1319{
1320 unsigned long pages = 0;
1321 struct memblock_region *r;
1322 unsigned long start_pfn, end_pfn;
1323
1324 for_each_memblock(memory, r) {
1325 start_pfn = memblock_region_memory_base_pfn(r);
1326 end_pfn = memblock_region_memory_end_pfn(r);
1327 start_pfn = min_t(unsigned long, start_pfn, limit_pfn);
1328 end_pfn = min_t(unsigned long, end_pfn, limit_pfn);
1329 pages += end_pfn - start_pfn;
1330 }
1331
16763230 1332 return PFN_PHYS(pages);
595ad9af
YL
1333}
1334
0a93ebef
SR
1335/* lowest address */
1336phys_addr_t __init_memblock memblock_start_of_DRAM(void)
1337{
1338 return memblock.memory.regions[0].base;
1339}
1340
10d06439 1341phys_addr_t __init_memblock memblock_end_of_DRAM(void)
95f72d1e
YL
1342{
1343 int idx = memblock.memory.cnt - 1;
1344
e3239ff9 1345 return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size);
95f72d1e
YL
1346}
1347
c0ce8fef 1348void __init memblock_enforce_memory_limit(phys_addr_t limit)
95f72d1e 1349{
c0ce8fef 1350 phys_addr_t max_addr = (phys_addr_t)ULLONG_MAX;
136199f0 1351 struct memblock_region *r;
95f72d1e 1352
c0ce8fef 1353 if (!limit)
95f72d1e
YL
1354 return;
1355
c0ce8fef 1356 /* find out max address */
136199f0 1357 for_each_memblock(memory, r) {
c0ce8fef
TH
1358 if (limit <= r->size) {
1359 max_addr = r->base + limit;
1360 break;
95f72d1e 1361 }
c0ce8fef 1362 limit -= r->size;
95f72d1e 1363 }
c0ce8fef
TH
1364
1365 /* truncate both memory and reserved regions */
f1af9d3a
PH
1366 memblock_remove_range(&memblock.memory, max_addr,
1367 (phys_addr_t)ULLONG_MAX);
1368 memblock_remove_range(&memblock.reserved, max_addr,
1369 (phys_addr_t)ULLONG_MAX);
95f72d1e
YL
1370}
1371
cd79481d 1372static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr)
72d4b0b4
BH
1373{
1374 unsigned int left = 0, right = type->cnt;
1375
1376 do {
1377 unsigned int mid = (right + left) / 2;
1378
1379 if (addr < type->regions[mid].base)
1380 right = mid;
1381 else if (addr >= (type->regions[mid].base +
1382 type->regions[mid].size))
1383 left = mid + 1;
1384 else
1385 return mid;
1386 } while (left < right);
1387 return -1;
1388}
1389
2898cc4c 1390int __init memblock_is_reserved(phys_addr_t addr)
95f72d1e 1391{
72d4b0b4
BH
1392 return memblock_search(&memblock.reserved, addr) != -1;
1393}
95f72d1e 1394
3661ca66 1395int __init_memblock memblock_is_memory(phys_addr_t addr)
72d4b0b4
BH
1396{
1397 return memblock_search(&memblock.memory, addr) != -1;
1398}
1399
e76b63f8
YL
1400#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1401int __init_memblock memblock_search_pfn_nid(unsigned long pfn,
1402 unsigned long *start_pfn, unsigned long *end_pfn)
1403{
1404 struct memblock_type *type = &memblock.memory;
16763230 1405 int mid = memblock_search(type, PFN_PHYS(pfn));
e76b63f8
YL
1406
1407 if (mid == -1)
1408 return -1;
1409
f7e2f7e8
FF
1410 *start_pfn = PFN_DOWN(type->regions[mid].base);
1411 *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size);
e76b63f8
YL
1412
1413 return type->regions[mid].nid;
1414}
1415#endif
1416
eab30949
SB
1417/**
1418 * memblock_is_region_memory - check if a region is a subset of memory
1419 * @base: base of region to check
1420 * @size: size of region to check
1421 *
1422 * Check if the region [@base, @base+@size) is a subset of a memory block.
1423 *
1424 * RETURNS:
1425 * 0 if false, non-zero if true
1426 */
3661ca66 1427int __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size)
72d4b0b4 1428{
abb65272 1429 int idx = memblock_search(&memblock.memory, base);
eb18f1b5 1430 phys_addr_t end = base + memblock_cap_size(base, &size);
72d4b0b4
BH
1431
1432 if (idx == -1)
1433 return 0;
abb65272
TV
1434 return memblock.memory.regions[idx].base <= base &&
1435 (memblock.memory.regions[idx].base +
eb18f1b5 1436 memblock.memory.regions[idx].size) >= end;
95f72d1e
YL
1437}
1438
eab30949
SB
1439/**
1440 * memblock_is_region_reserved - check if a region intersects reserved memory
1441 * @base: base of region to check
1442 * @size: size of region to check
1443 *
1444 * Check if the region [@base, @base+@size) intersects a reserved memory block.
1445 *
1446 * RETURNS:
1447 * 0 if false, non-zero if true
1448 */
10d06439 1449int __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size)
95f72d1e 1450{
eb18f1b5 1451 memblock_cap_size(base, &size);
f1c2c19c 1452 return memblock_overlaps_region(&memblock.reserved, base, size) >= 0;
95f72d1e
YL
1453}
1454
6ede1fd3
YL
1455void __init_memblock memblock_trim_memory(phys_addr_t align)
1456{
6ede1fd3 1457 phys_addr_t start, end, orig_start, orig_end;
136199f0 1458 struct memblock_region *r;
6ede1fd3 1459
136199f0
EM
1460 for_each_memblock(memory, r) {
1461 orig_start = r->base;
1462 orig_end = r->base + r->size;
6ede1fd3
YL
1463 start = round_up(orig_start, align);
1464 end = round_down(orig_end, align);
1465
1466 if (start == orig_start && end == orig_end)
1467 continue;
1468
1469 if (start < end) {
136199f0
EM
1470 r->base = start;
1471 r->size = end - start;
6ede1fd3 1472 } else {
136199f0
EM
1473 memblock_remove_region(&memblock.memory,
1474 r - memblock.memory.regions);
1475 r--;
6ede1fd3
YL
1476 }
1477 }
1478}
e63075a3 1479
3661ca66 1480void __init_memblock memblock_set_current_limit(phys_addr_t limit)
e63075a3
BH
1481{
1482 memblock.current_limit = limit;
1483}
1484
fec51014
LA
1485phys_addr_t __init_memblock memblock_get_current_limit(void)
1486{
1487 return memblock.current_limit;
1488}
1489
7c0caeb8 1490static void __init_memblock memblock_dump(struct memblock_type *type, char *name)
6ed311b2
BH
1491{
1492 unsigned long long base, size;
66a20757 1493 unsigned long flags;
6ed311b2
BH
1494 int i;
1495
7c0caeb8 1496 pr_info(" %s.cnt = 0x%lx\n", name, type->cnt);
6ed311b2 1497
7c0caeb8
TH
1498 for (i = 0; i < type->cnt; i++) {
1499 struct memblock_region *rgn = &type->regions[i];
1500 char nid_buf[32] = "";
1501
1502 base = rgn->base;
1503 size = rgn->size;
66a20757 1504 flags = rgn->flags;
7c0caeb8
TH
1505#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1506 if (memblock_get_region_node(rgn) != MAX_NUMNODES)
1507 snprintf(nid_buf, sizeof(nid_buf), " on node %d",
1508 memblock_get_region_node(rgn));
1509#endif
66a20757
TC
1510 pr_info(" %s[%#x]\t[%#016llx-%#016llx], %#llx bytes%s flags: %#lx\n",
1511 name, i, base, base + size - 1, size, nid_buf, flags);
6ed311b2
BH
1512 }
1513}
1514
4ff7b82f 1515void __init_memblock __memblock_dump_all(void)
6ed311b2 1516{
6ed311b2 1517 pr_info("MEMBLOCK configuration:\n");
1440c4e2
TH
1518 pr_info(" memory size = %#llx reserved size = %#llx\n",
1519 (unsigned long long)memblock.memory.total_size,
1520 (unsigned long long)memblock.reserved.total_size);
6ed311b2
BH
1521
1522 memblock_dump(&memblock.memory, "memory");
1523 memblock_dump(&memblock.reserved, "reserved");
1524}
1525
1aadc056 1526void __init memblock_allow_resize(void)
6ed311b2 1527{
142b45a7 1528 memblock_can_resize = 1;
6ed311b2
BH
1529}
1530
6ed311b2
BH
1531static int __init early_memblock(char *p)
1532{
1533 if (p && strstr(p, "debug"))
1534 memblock_debug = 1;
1535 return 0;
1536}
1537early_param("memblock", early_memblock);
1538
c378ddd5 1539#if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK)
6d03b885
BH
1540
1541static int memblock_debug_show(struct seq_file *m, void *private)
1542{
1543 struct memblock_type *type = m->private;
1544 struct memblock_region *reg;
1545 int i;
1546
1547 for (i = 0; i < type->cnt; i++) {
1548 reg = &type->regions[i];
1549 seq_printf(m, "%4d: ", i);
1550 if (sizeof(phys_addr_t) == 4)
1551 seq_printf(m, "0x%08lx..0x%08lx\n",
1552 (unsigned long)reg->base,
1553 (unsigned long)(reg->base + reg->size - 1));
1554 else
1555 seq_printf(m, "0x%016llx..0x%016llx\n",
1556 (unsigned long long)reg->base,
1557 (unsigned long long)(reg->base + reg->size - 1));
1558
1559 }
1560 return 0;
1561}
1562
1563static int memblock_debug_open(struct inode *inode, struct file *file)
1564{
1565 return single_open(file, memblock_debug_show, inode->i_private);
1566}
1567
1568static const struct file_operations memblock_debug_fops = {
1569 .open = memblock_debug_open,
1570 .read = seq_read,
1571 .llseek = seq_lseek,
1572 .release = single_release,
1573};
1574
1575static int __init memblock_init_debugfs(void)
1576{
1577 struct dentry *root = debugfs_create_dir("memblock", NULL);
1578 if (!root)
1579 return -ENXIO;
1580 debugfs_create_file("memory", S_IRUGO, root, &memblock.memory, &memblock_debug_fops);
1581 debugfs_create_file("reserved", S_IRUGO, root, &memblock.reserved, &memblock_debug_fops);
70210ed9
PH
1582#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
1583 debugfs_create_file("physmem", S_IRUGO, root, &memblock.physmem, &memblock_debug_fops);
1584#endif
6d03b885
BH
1585
1586 return 0;
1587}
1588__initcall(memblock_init_debugfs);
1589
1590#endif /* CONFIG_DEBUG_FS */
This page took 0.325308 seconds and 5 git commands to generate.