mm/memblock: Do some refactoring, enhance API
[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;
30
31struct memblock memblock __initdata_memblock = {
32 .memory.regions = memblock_memory_init_regions,
33 .memory.cnt = 1, /* empty dummy entry */
34 .memory.max = INIT_MEMBLOCK_REGIONS,
35
36 .reserved.regions = memblock_reserved_init_regions,
37 .reserved.cnt = 1, /* empty dummy entry */
38 .reserved.max = INIT_MEMBLOCK_REGIONS,
39
79442ed1 40 .bottom_up = false,
fe091c20
TH
41 .current_limit = MEMBLOCK_ALLOC_ANYWHERE,
42};
95f72d1e 43
10d06439 44int memblock_debug __initdata_memblock;
55ac590c
TC
45#ifdef CONFIG_MOVABLE_NODE
46bool movable_node_enabled __initdata_memblock = false;
47#endif
1aadc056 48static int memblock_can_resize __initdata_memblock;
181eb394
GS
49static int memblock_memory_in_slab __initdata_memblock = 0;
50static int memblock_reserved_in_slab __initdata_memblock = 0;
95f72d1e 51
142b45a7 52/* inline so we don't get a warning when pr_debug is compiled out */
c2233116
RP
53static __init_memblock const char *
54memblock_type_name(struct memblock_type *type)
142b45a7
BH
55{
56 if (type == &memblock.memory)
57 return "memory";
58 else if (type == &memblock.reserved)
59 return "reserved";
60 else
61 return "unknown";
62}
63
eb18f1b5
TH
64/* adjust *@size so that (@base + *@size) doesn't overflow, return new size */
65static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size)
66{
67 return *size = min(*size, (phys_addr_t)ULLONG_MAX - base);
68}
69
6ed311b2
BH
70/*
71 * Address comparison utilities
72 */
10d06439 73static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1,
2898cc4c 74 phys_addr_t base2, phys_addr_t size2)
95f72d1e
YL
75{
76 return ((base1 < (base2 + size2)) && (base2 < (base1 + size1)));
77}
78
2d7d3eb2
HS
79static long __init_memblock memblock_overlaps_region(struct memblock_type *type,
80 phys_addr_t base, phys_addr_t size)
6ed311b2
BH
81{
82 unsigned long i;
83
84 for (i = 0; i < type->cnt; i++) {
85 phys_addr_t rgnbase = type->regions[i].base;
86 phys_addr_t rgnsize = type->regions[i].size;
87 if (memblock_addrs_overlap(base, size, rgnbase, rgnsize))
88 break;
89 }
90
91 return (i < type->cnt) ? i : -1;
92}
93
79442ed1
TC
94/*
95 * __memblock_find_range_bottom_up - find free area utility in bottom-up
96 * @start: start of candidate range
97 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
98 * @size: size of free area to find
99 * @align: alignment of free area to find
b1154233 100 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
79442ed1
TC
101 *
102 * Utility called from memblock_find_in_range_node(), find free area bottom-up.
103 *
104 * RETURNS:
105 * Found address on success, 0 on failure.
106 */
107static phys_addr_t __init_memblock
108__memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
109 phys_addr_t size, phys_addr_t align, int nid)
110{
111 phys_addr_t this_start, this_end, cand;
112 u64 i;
113
114 for_each_free_mem_range(i, nid, &this_start, &this_end, NULL) {
115 this_start = clamp(this_start, start, end);
116 this_end = clamp(this_end, start, end);
117
118 cand = round_up(this_start, align);
119 if (cand < this_end && this_end - cand >= size)
120 return cand;
121 }
122
123 return 0;
124}
125
7bd0b0f0 126/**
1402899e 127 * __memblock_find_range_top_down - find free area utility, in top-down
7bd0b0f0
TH
128 * @start: start of candidate range
129 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
130 * @size: size of free area to find
131 * @align: alignment of free area to find
b1154233 132 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
7bd0b0f0 133 *
1402899e 134 * Utility called from memblock_find_in_range_node(), find free area top-down.
7bd0b0f0
TH
135 *
136 * RETURNS:
79442ed1 137 * Found address on success, 0 on failure.
6ed311b2 138 */
1402899e
TC
139static phys_addr_t __init_memblock
140__memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
141 phys_addr_t size, phys_addr_t align, int nid)
f7210e6c
TC
142{
143 phys_addr_t this_start, this_end, cand;
144 u64 i;
145
f7210e6c
TC
146 for_each_free_mem_range_reverse(i, nid, &this_start, &this_end, NULL) {
147 this_start = clamp(this_start, start, end);
148 this_end = clamp(this_end, start, end);
149
150 if (this_end < size)
151 continue;
152
153 cand = round_down(this_end - size, align);
154 if (cand >= this_start)
155 return cand;
156 }
1402899e 157
f7210e6c
TC
158 return 0;
159}
6ed311b2 160
1402899e
TC
161/**
162 * memblock_find_in_range_node - find free area in given range and node
1402899e
TC
163 * @size: size of free area to find
164 * @align: alignment of free area to find
87029ee9
GS
165 * @start: start of candidate range
166 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
b1154233 167 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1402899e
TC
168 *
169 * Find @size free area aligned to @align in the specified range and node.
170 *
79442ed1
TC
171 * When allocation direction is bottom-up, the @start should be greater
172 * than the end of the kernel image. Otherwise, it will be trimmed. The
173 * reason is that we want the bottom-up allocation just near the kernel
174 * image so it is highly likely that the allocated memory and the kernel
175 * will reside in the same node.
176 *
177 * If bottom-up allocation failed, will try to allocate memory top-down.
178 *
1402899e 179 * RETURNS:
79442ed1 180 * Found address on success, 0 on failure.
1402899e 181 */
87029ee9
GS
182phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
183 phys_addr_t align, phys_addr_t start,
184 phys_addr_t end, int nid)
1402899e 185{
79442ed1
TC
186 int ret;
187 phys_addr_t kernel_end;
188
1402899e
TC
189 /* pump up @end */
190 if (end == MEMBLOCK_ALLOC_ACCESSIBLE)
191 end = memblock.current_limit;
192
193 /* avoid allocating the first page */
194 start = max_t(phys_addr_t, start, PAGE_SIZE);
195 end = max(start, end);
79442ed1
TC
196 kernel_end = __pa_symbol(_end);
197
198 /*
199 * try bottom-up allocation only when bottom-up mode
200 * is set and @end is above the kernel image.
201 */
202 if (memblock_bottom_up() && end > kernel_end) {
203 phys_addr_t bottom_up_start;
204
205 /* make sure we will allocate above the kernel */
206 bottom_up_start = max(start, kernel_end);
207
208 /* ok, try bottom-up allocation first */
209 ret = __memblock_find_range_bottom_up(bottom_up_start, end,
210 size, align, nid);
211 if (ret)
212 return ret;
213
214 /*
215 * we always limit bottom-up allocation above the kernel,
216 * but top-down allocation doesn't have the limit, so
217 * retrying top-down allocation may succeed when bottom-up
218 * allocation failed.
219 *
220 * bottom-up allocation is expected to be fail very rarely,
221 * so we use WARN_ONCE() here to see the stack trace if
222 * fail happens.
223 */
224 WARN_ONCE(1, "memblock: bottom-up allocation failed, "
225 "memory hotunplug may be affected\n");
226 }
1402899e
TC
227
228 return __memblock_find_range_top_down(start, end, size, align, nid);
229}
230
7bd0b0f0
TH
231/**
232 * memblock_find_in_range - find free area in given range
233 * @start: start of candidate range
234 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
235 * @size: size of free area to find
236 * @align: alignment of free area to find
237 *
238 * Find @size free area aligned to @align in the specified range.
239 *
240 * RETURNS:
79442ed1 241 * Found address on success, 0 on failure.
fc769a8e 242 */
7bd0b0f0
TH
243phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start,
244 phys_addr_t end, phys_addr_t size,
245 phys_addr_t align)
6ed311b2 246{
87029ee9 247 return memblock_find_in_range_node(size, align, start, end,
b1154233 248 NUMA_NO_NODE);
6ed311b2
BH
249}
250
10d06439 251static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r)
95f72d1e 252{
1440c4e2 253 type->total_size -= type->regions[r].size;
7c0caeb8
TH
254 memmove(&type->regions[r], &type->regions[r + 1],
255 (type->cnt - (r + 1)) * sizeof(type->regions[r]));
e3239ff9 256 type->cnt--;
95f72d1e 257
8f7a6605
BH
258 /* Special case for empty arrays */
259 if (type->cnt == 0) {
1440c4e2 260 WARN_ON(type->total_size != 0);
8f7a6605
BH
261 type->cnt = 1;
262 type->regions[0].base = 0;
263 type->regions[0].size = 0;
66a20757 264 type->regions[0].flags = 0;
7c0caeb8 265 memblock_set_region_node(&type->regions[0], MAX_NUMNODES);
8f7a6605 266 }
95f72d1e
YL
267}
268
354f17e1
PH
269#ifdef CONFIG_ARCH_DISCARD_MEMBLOCK
270
29f67386
YL
271phys_addr_t __init_memblock get_allocated_memblock_reserved_regions_info(
272 phys_addr_t *addr)
273{
274 if (memblock.reserved.regions == memblock_reserved_init_regions)
275 return 0;
276
277 *addr = __pa(memblock.reserved.regions);
278
279 return PAGE_ALIGN(sizeof(struct memblock_region) *
280 memblock.reserved.max);
281}
282
5e270e25
PH
283phys_addr_t __init_memblock get_allocated_memblock_memory_regions_info(
284 phys_addr_t *addr)
285{
286 if (memblock.memory.regions == memblock_memory_init_regions)
287 return 0;
288
289 *addr = __pa(memblock.memory.regions);
290
291 return PAGE_ALIGN(sizeof(struct memblock_region) *
292 memblock.memory.max);
293}
294
295#endif
296
48c3b583
GP
297/**
298 * memblock_double_array - double the size of the memblock regions array
299 * @type: memblock type of the regions array being doubled
300 * @new_area_start: starting address of memory range to avoid overlap with
301 * @new_area_size: size of memory range to avoid overlap with
302 *
303 * Double the size of the @type regions array. If memblock is being used to
304 * allocate memory for a new reserved regions array and there is a previously
305 * allocated memory range [@new_area_start,@new_area_start+@new_area_size]
306 * waiting to be reserved, ensure the memory used by the new array does
307 * not overlap.
308 *
309 * RETURNS:
310 * 0 on success, -1 on failure.
311 */
312static int __init_memblock memblock_double_array(struct memblock_type *type,
313 phys_addr_t new_area_start,
314 phys_addr_t new_area_size)
142b45a7
BH
315{
316 struct memblock_region *new_array, *old_array;
29f67386 317 phys_addr_t old_alloc_size, new_alloc_size;
142b45a7
BH
318 phys_addr_t old_size, new_size, addr;
319 int use_slab = slab_is_available();
181eb394 320 int *in_slab;
142b45a7
BH
321
322 /* We don't allow resizing until we know about the reserved regions
323 * of memory that aren't suitable for allocation
324 */
325 if (!memblock_can_resize)
326 return -1;
327
142b45a7
BH
328 /* Calculate new doubled size */
329 old_size = type->max * sizeof(struct memblock_region);
330 new_size = old_size << 1;
29f67386
YL
331 /*
332 * We need to allocated new one align to PAGE_SIZE,
333 * so we can free them completely later.
334 */
335 old_alloc_size = PAGE_ALIGN(old_size);
336 new_alloc_size = PAGE_ALIGN(new_size);
142b45a7 337
181eb394
GS
338 /* Retrieve the slab flag */
339 if (type == &memblock.memory)
340 in_slab = &memblock_memory_in_slab;
341 else
342 in_slab = &memblock_reserved_in_slab;
343
142b45a7
BH
344 /* Try to find some space for it.
345 *
346 * WARNING: We assume that either slab_is_available() and we use it or
fd07383b
AM
347 * we use MEMBLOCK for allocations. That means that this is unsafe to
348 * use when bootmem is currently active (unless bootmem itself is
349 * implemented on top of MEMBLOCK which isn't the case yet)
142b45a7
BH
350 *
351 * This should however not be an issue for now, as we currently only
fd07383b
AM
352 * call into MEMBLOCK while it's still active, or much later when slab
353 * is active for memory hotplug operations
142b45a7
BH
354 */
355 if (use_slab) {
356 new_array = kmalloc(new_size, GFP_KERNEL);
1f5026a7 357 addr = new_array ? __pa(new_array) : 0;
4e2f0775 358 } else {
48c3b583
GP
359 /* only exclude range when trying to double reserved.regions */
360 if (type != &memblock.reserved)
361 new_area_start = new_area_size = 0;
362
363 addr = memblock_find_in_range(new_area_start + new_area_size,
364 memblock.current_limit,
29f67386 365 new_alloc_size, PAGE_SIZE);
48c3b583
GP
366 if (!addr && new_area_size)
367 addr = memblock_find_in_range(0,
fd07383b
AM
368 min(new_area_start, memblock.current_limit),
369 new_alloc_size, PAGE_SIZE);
48c3b583 370
15674868 371 new_array = addr ? __va(addr) : NULL;
4e2f0775 372 }
1f5026a7 373 if (!addr) {
142b45a7
BH
374 pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n",
375 memblock_type_name(type), type->max, type->max * 2);
376 return -1;
377 }
142b45a7 378
fd07383b
AM
379 memblock_dbg("memblock: %s is doubled to %ld at [%#010llx-%#010llx]",
380 memblock_type_name(type), type->max * 2, (u64)addr,
381 (u64)addr + new_size - 1);
ea9e4376 382
fd07383b
AM
383 /*
384 * Found space, we now need to move the array over before we add the
385 * reserved region since it may be our reserved array itself that is
386 * full.
142b45a7
BH
387 */
388 memcpy(new_array, type->regions, old_size);
389 memset(new_array + type->max, 0, old_size);
390 old_array = type->regions;
391 type->regions = new_array;
392 type->max <<= 1;
393
fd07383b 394 /* Free old array. We needn't free it if the array is the static one */
181eb394
GS
395 if (*in_slab)
396 kfree(old_array);
397 else if (old_array != memblock_memory_init_regions &&
398 old_array != memblock_reserved_init_regions)
29f67386 399 memblock_free(__pa(old_array), old_alloc_size);
142b45a7 400
fd07383b
AM
401 /*
402 * Reserve the new array if that comes from the memblock. Otherwise, we
403 * needn't do it
181eb394
GS
404 */
405 if (!use_slab)
29f67386 406 BUG_ON(memblock_reserve(addr, new_alloc_size));
181eb394
GS
407
408 /* Update slab flag */
409 *in_slab = use_slab;
410
142b45a7
BH
411 return 0;
412}
413
784656f9
TH
414/**
415 * memblock_merge_regions - merge neighboring compatible regions
416 * @type: memblock type to scan
417 *
418 * Scan @type and merge neighboring compatible regions.
419 */
420static void __init_memblock memblock_merge_regions(struct memblock_type *type)
95f72d1e 421{
784656f9 422 int i = 0;
95f72d1e 423
784656f9
TH
424 /* cnt never goes below 1 */
425 while (i < type->cnt - 1) {
426 struct memblock_region *this = &type->regions[i];
427 struct memblock_region *next = &type->regions[i + 1];
95f72d1e 428
7c0caeb8
TH
429 if (this->base + this->size != next->base ||
430 memblock_get_region_node(this) !=
66a20757
TC
431 memblock_get_region_node(next) ||
432 this->flags != next->flags) {
784656f9
TH
433 BUG_ON(this->base + this->size > next->base);
434 i++;
435 continue;
8f7a6605
BH
436 }
437
784656f9 438 this->size += next->size;
c0232ae8
LF
439 /* move forward from next + 1, index of which is i + 2 */
440 memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next));
784656f9 441 type->cnt--;
95f72d1e 442 }
784656f9 443}
95f72d1e 444
784656f9
TH
445/**
446 * memblock_insert_region - insert new memblock region
209ff86d
TC
447 * @type: memblock type to insert into
448 * @idx: index for the insertion point
449 * @base: base address of the new region
450 * @size: size of the new region
451 * @nid: node id of the new region
66a20757 452 * @flags: flags of the new region
784656f9
TH
453 *
454 * Insert new memblock region [@base,@base+@size) into @type at @idx.
455 * @type must already have extra room to accomodate the new region.
456 */
457static void __init_memblock memblock_insert_region(struct memblock_type *type,
458 int idx, phys_addr_t base,
66a20757
TC
459 phys_addr_t size,
460 int nid, unsigned long flags)
784656f9
TH
461{
462 struct memblock_region *rgn = &type->regions[idx];
463
464 BUG_ON(type->cnt >= type->max);
465 memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn));
466 rgn->base = base;
467 rgn->size = size;
66a20757 468 rgn->flags = flags;
7c0caeb8 469 memblock_set_region_node(rgn, nid);
784656f9 470 type->cnt++;
1440c4e2 471 type->total_size += size;
784656f9
TH
472}
473
474/**
f1af9d3a 475 * memblock_add_range - add new memblock region
784656f9
TH
476 * @type: memblock type to add new region into
477 * @base: base address of the new region
478 * @size: size of the new region
7fb0bc3f 479 * @nid: nid of the new region
66a20757 480 * @flags: flags of the new region
784656f9
TH
481 *
482 * Add new memblock region [@base,@base+@size) into @type. The new region
483 * is allowed to overlap with existing ones - overlaps don't affect already
484 * existing regions. @type is guaranteed to be minimal (all neighbouring
485 * compatible regions are merged) after the addition.
486 *
487 * RETURNS:
488 * 0 on success, -errno on failure.
489 */
f1af9d3a 490int __init_memblock memblock_add_range(struct memblock_type *type,
66a20757
TC
491 phys_addr_t base, phys_addr_t size,
492 int nid, unsigned long flags)
784656f9
TH
493{
494 bool insert = false;
eb18f1b5
TH
495 phys_addr_t obase = base;
496 phys_addr_t end = base + memblock_cap_size(base, &size);
784656f9
TH
497 int i, nr_new;
498
b3dc627c
TH
499 if (!size)
500 return 0;
501
784656f9
TH
502 /* special case for empty array */
503 if (type->regions[0].size == 0) {
1440c4e2 504 WARN_ON(type->cnt != 1 || type->total_size);
8f7a6605
BH
505 type->regions[0].base = base;
506 type->regions[0].size = size;
66a20757 507 type->regions[0].flags = flags;
7fb0bc3f 508 memblock_set_region_node(&type->regions[0], nid);
1440c4e2 509 type->total_size = size;
8f7a6605 510 return 0;
95f72d1e 511 }
784656f9
TH
512repeat:
513 /*
514 * The following is executed twice. Once with %false @insert and
515 * then with %true. The first counts the number of regions needed
516 * to accomodate the new area. The second actually inserts them.
142b45a7 517 */
784656f9
TH
518 base = obase;
519 nr_new = 0;
95f72d1e 520
784656f9
TH
521 for (i = 0; i < type->cnt; i++) {
522 struct memblock_region *rgn = &type->regions[i];
523 phys_addr_t rbase = rgn->base;
524 phys_addr_t rend = rbase + rgn->size;
525
526 if (rbase >= end)
95f72d1e 527 break;
784656f9
TH
528 if (rend <= base)
529 continue;
530 /*
531 * @rgn overlaps. If it separates the lower part of new
532 * area, insert that portion.
533 */
534 if (rbase > base) {
535 nr_new++;
536 if (insert)
537 memblock_insert_region(type, i++, base,
66a20757
TC
538 rbase - base, nid,
539 flags);
95f72d1e 540 }
784656f9
TH
541 /* area below @rend is dealt with, forget about it */
542 base = min(rend, end);
95f72d1e 543 }
784656f9
TH
544
545 /* insert the remaining portion */
546 if (base < end) {
547 nr_new++;
548 if (insert)
66a20757
TC
549 memblock_insert_region(type, i, base, end - base,
550 nid, flags);
95f72d1e 551 }
95f72d1e 552
784656f9
TH
553 /*
554 * If this was the first round, resize array and repeat for actual
555 * insertions; otherwise, merge and return.
142b45a7 556 */
784656f9
TH
557 if (!insert) {
558 while (type->cnt + nr_new > type->max)
48c3b583 559 if (memblock_double_array(type, obase, size) < 0)
784656f9
TH
560 return -ENOMEM;
561 insert = true;
562 goto repeat;
563 } else {
564 memblock_merge_regions(type);
565 return 0;
142b45a7 566 }
95f72d1e
YL
567}
568
7fb0bc3f
TH
569int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size,
570 int nid)
571{
f1af9d3a 572 return memblock_add_range(&memblock.memory, base, size, nid, 0);
7fb0bc3f
TH
573}
574
581adcbe 575int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
95f72d1e 576{
f1af9d3a 577 return memblock_add_range(&memblock.memory, base, size,
66a20757 578 MAX_NUMNODES, 0);
95f72d1e
YL
579}
580
6a9ceb31
TH
581/**
582 * memblock_isolate_range - isolate given range into disjoint memblocks
583 * @type: memblock type to isolate range for
584 * @base: base of range to isolate
585 * @size: size of range to isolate
586 * @start_rgn: out parameter for the start of isolated region
587 * @end_rgn: out parameter for the end of isolated region
588 *
589 * Walk @type and ensure that regions don't cross the boundaries defined by
590 * [@base,@base+@size). Crossing regions are split at the boundaries,
591 * which may create at most two more regions. The index of the first
592 * region inside the range is returned in *@start_rgn and end in *@end_rgn.
593 *
594 * RETURNS:
595 * 0 on success, -errno on failure.
596 */
597static int __init_memblock memblock_isolate_range(struct memblock_type *type,
598 phys_addr_t base, phys_addr_t size,
599 int *start_rgn, int *end_rgn)
600{
eb18f1b5 601 phys_addr_t end = base + memblock_cap_size(base, &size);
6a9ceb31
TH
602 int i;
603
604 *start_rgn = *end_rgn = 0;
605
b3dc627c
TH
606 if (!size)
607 return 0;
608
6a9ceb31
TH
609 /* we'll create at most two more regions */
610 while (type->cnt + 2 > type->max)
48c3b583 611 if (memblock_double_array(type, base, size) < 0)
6a9ceb31
TH
612 return -ENOMEM;
613
614 for (i = 0; i < type->cnt; i++) {
615 struct memblock_region *rgn = &type->regions[i];
616 phys_addr_t rbase = rgn->base;
617 phys_addr_t rend = rbase + rgn->size;
618
619 if (rbase >= end)
620 break;
621 if (rend <= base)
622 continue;
623
624 if (rbase < base) {
625 /*
626 * @rgn intersects from below. Split and continue
627 * to process the next region - the new top half.
628 */
629 rgn->base = base;
1440c4e2
TH
630 rgn->size -= base - rbase;
631 type->total_size -= base - rbase;
6a9ceb31 632 memblock_insert_region(type, i, rbase, base - rbase,
66a20757
TC
633 memblock_get_region_node(rgn),
634 rgn->flags);
6a9ceb31
TH
635 } else if (rend > end) {
636 /*
637 * @rgn intersects from above. Split and redo the
638 * current region - the new bottom half.
639 */
640 rgn->base = end;
1440c4e2
TH
641 rgn->size -= end - rbase;
642 type->total_size -= end - rbase;
6a9ceb31 643 memblock_insert_region(type, i--, rbase, end - rbase,
66a20757
TC
644 memblock_get_region_node(rgn),
645 rgn->flags);
6a9ceb31
TH
646 } else {
647 /* @rgn is fully contained, record it */
648 if (!*end_rgn)
649 *start_rgn = i;
650 *end_rgn = i + 1;
651 }
652 }
653
654 return 0;
655}
6a9ceb31 656
f1af9d3a
PH
657int __init_memblock memblock_remove_range(struct memblock_type *type,
658 phys_addr_t base, phys_addr_t size)
95f72d1e 659{
71936180
TH
660 int start_rgn, end_rgn;
661 int i, ret;
95f72d1e 662
71936180
TH
663 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
664 if (ret)
665 return ret;
95f72d1e 666
71936180
TH
667 for (i = end_rgn - 1; i >= start_rgn; i--)
668 memblock_remove_region(type, i);
8f7a6605 669 return 0;
95f72d1e
YL
670}
671
581adcbe 672int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size)
95f72d1e 673{
f1af9d3a 674 return memblock_remove_range(&memblock.memory, base, size);
95f72d1e
YL
675}
676
f1af9d3a 677
581adcbe 678int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size)
95f72d1e 679{
24aa0788 680 memblock_dbg(" memblock_free: [%#016llx-%#016llx] %pF\n",
a150439c 681 (unsigned long long)base,
931d13f5 682 (unsigned long long)base + size - 1,
a150439c 683 (void *)_RET_IP_);
24aa0788 684
f1af9d3a 685 return memblock_remove_range(&memblock.reserved, base, size);
95f72d1e
YL
686}
687
66a20757
TC
688static int __init_memblock memblock_reserve_region(phys_addr_t base,
689 phys_addr_t size,
690 int nid,
691 unsigned long flags)
95f72d1e 692{
e3239ff9 693 struct memblock_type *_rgn = &memblock.reserved;
95f72d1e 694
66a20757 695 memblock_dbg("memblock_reserve: [%#016llx-%#016llx] flags %#02lx %pF\n",
a150439c 696 (unsigned long long)base,
931d13f5 697 (unsigned long long)base + size - 1,
66a20757
TC
698 flags, (void *)_RET_IP_);
699
f1af9d3a 700 return memblock_add_range(_rgn, base, size, nid, flags);
66a20757 701}
95f72d1e 702
66a20757
TC
703int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size)
704{
705 return memblock_reserve_region(base, size, MAX_NUMNODES, 0);
95f72d1e
YL
706}
707
66b16edf
TC
708/**
709 * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG.
710 * @base: the base phys addr of the region
711 * @size: the size of the region
712 *
713 * This function isolates region [@base, @base + @size), and mark it with flag
714 * MEMBLOCK_HOTPLUG.
715 *
716 * Return 0 on succees, -errno on failure.
717 */
718int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size)
719{
720 struct memblock_type *type = &memblock.memory;
721 int i, ret, start_rgn, end_rgn;
722
723 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
724 if (ret)
725 return ret;
726
727 for (i = start_rgn; i < end_rgn; i++)
728 memblock_set_region_flags(&type->regions[i], MEMBLOCK_HOTPLUG);
729
730 memblock_merge_regions(type);
731 return 0;
732}
733
734/**
735 * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region.
736 * @base: the base phys addr of the region
737 * @size: the size of the region
738 *
739 * This function isolates region [@base, @base + @size), and clear flag
740 * MEMBLOCK_HOTPLUG for the isolated regions.
741 *
742 * Return 0 on succees, -errno on failure.
743 */
744int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size)
745{
746 struct memblock_type *type = &memblock.memory;
747 int i, ret, start_rgn, end_rgn;
748
749 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
750 if (ret)
751 return ret;
752
753 for (i = start_rgn; i < end_rgn; i++)
754 memblock_clear_region_flags(&type->regions[i],
755 MEMBLOCK_HOTPLUG);
756
757 memblock_merge_regions(type);
758 return 0;
759}
760
35fd0808 761/**
f1af9d3a 762 * __next__mem_range - next function for for_each_free_mem_range() etc.
35fd0808 763 * @idx: pointer to u64 loop variable
b1154233 764 * @nid: node selector, %NUMA_NO_NODE for all nodes
f1af9d3a
PH
765 * @type_a: pointer to memblock_type from where the range is taken
766 * @type_b: pointer to memblock_type which excludes memory from being taken
dad7557e
WL
767 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
768 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
769 * @out_nid: ptr to int for nid of the range, can be %NULL
35fd0808 770 *
f1af9d3a 771 * Find the first area from *@idx which matches @nid, fill the out
35fd0808 772 * parameters, and update *@idx for the next iteration. The lower 32bit of
f1af9d3a
PH
773 * *@idx contains index into type_a and the upper 32bit indexes the
774 * areas before each region in type_b. For example, if type_b regions
35fd0808
TH
775 * look like the following,
776 *
777 * 0:[0-16), 1:[32-48), 2:[128-130)
778 *
779 * The upper 32bit indexes the following regions.
780 *
781 * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX)
782 *
783 * As both region arrays are sorted, the function advances the two indices
784 * in lockstep and returns each intersection.
785 */
f1af9d3a
PH
786void __init_memblock __next_mem_range(u64 *idx, int nid,
787 struct memblock_type *type_a,
788 struct memblock_type *type_b,
789 phys_addr_t *out_start,
790 phys_addr_t *out_end, int *out_nid)
35fd0808 791{
f1af9d3a
PH
792 int idx_a = *idx & 0xffffffff;
793 int idx_b = *idx >> 32;
b1154233 794
f1af9d3a
PH
795 if (WARN_ONCE(nid == MAX_NUMNODES,
796 "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
560dca27 797 nid = NUMA_NO_NODE;
35fd0808 798
f1af9d3a
PH
799 for (; idx_a < type_a->cnt; idx_a++) {
800 struct memblock_region *m = &type_a->regions[idx_a];
801
35fd0808
TH
802 phys_addr_t m_start = m->base;
803 phys_addr_t m_end = m->base + m->size;
f1af9d3a 804 int m_nid = memblock_get_region_node(m);
35fd0808
TH
805
806 /* only memory regions are associated with nodes, check it */
f1af9d3a 807 if (nid != NUMA_NO_NODE && nid != m_nid)
35fd0808
TH
808 continue;
809
f1af9d3a
PH
810 if (!type_b) {
811 if (out_start)
812 *out_start = m_start;
813 if (out_end)
814 *out_end = m_end;
815 if (out_nid)
816 *out_nid = m_nid;
817 idx_a++;
818 *idx = (u32)idx_a | (u64)idx_b << 32;
819 return;
820 }
821
822 /* scan areas before each reservation */
823 for (; idx_b < type_b->cnt + 1; idx_b++) {
824 struct memblock_region *r;
825 phys_addr_t r_start;
826 phys_addr_t r_end;
827
828 r = &type_b->regions[idx_b];
829 r_start = idx_b ? r[-1].base + r[-1].size : 0;
830 r_end = idx_b < type_b->cnt ?
831 r->base : ULLONG_MAX;
35fd0808 832
f1af9d3a
PH
833 /*
834 * if idx_b advanced past idx_a,
835 * break out to advance idx_a
836 */
35fd0808
TH
837 if (r_start >= m_end)
838 break;
839 /* if the two regions intersect, we're done */
840 if (m_start < r_end) {
841 if (out_start)
f1af9d3a
PH
842 *out_start =
843 max(m_start, r_start);
35fd0808
TH
844 if (out_end)
845 *out_end = min(m_end, r_end);
846 if (out_nid)
f1af9d3a 847 *out_nid = m_nid;
35fd0808 848 /*
f1af9d3a
PH
849 * The region which ends first is
850 * advanced for the next iteration.
35fd0808
TH
851 */
852 if (m_end <= r_end)
f1af9d3a 853 idx_a++;
35fd0808 854 else
f1af9d3a
PH
855 idx_b++;
856 *idx = (u32)idx_a | (u64)idx_b << 32;
35fd0808
TH
857 return;
858 }
859 }
860 }
861
862 /* signal end of iteration */
863 *idx = ULLONG_MAX;
864}
865
7bd0b0f0 866/**
f1af9d3a
PH
867 * __next_mem_range_rev - generic next function for for_each_*_range_rev()
868 *
869 * Finds the next range from type_a which is not marked as unsuitable
870 * in type_b.
871 *
7bd0b0f0 872 * @idx: pointer to u64 loop variable
b1154233 873 * @nid: nid: node selector, %NUMA_NO_NODE for all nodes
f1af9d3a
PH
874 * @type_a: pointer to memblock_type from where the range is taken
875 * @type_b: pointer to memblock_type which excludes memory from being taken
dad7557e
WL
876 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
877 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
878 * @out_nid: ptr to int for nid of the range, can be %NULL
7bd0b0f0 879 *
f1af9d3a 880 * Reverse of __next_mem_range().
7bd0b0f0 881 */
f1af9d3a
PH
882void __init_memblock __next_mem_range_rev(u64 *idx, int nid,
883 struct memblock_type *type_a,
884 struct memblock_type *type_b,
885 phys_addr_t *out_start,
886 phys_addr_t *out_end, int *out_nid)
7bd0b0f0 887{
f1af9d3a
PH
888 int idx_a = *idx & 0xffffffff;
889 int idx_b = *idx >> 32;
b1154233 890
560dca27
GS
891 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
892 nid = NUMA_NO_NODE;
7bd0b0f0
TH
893
894 if (*idx == (u64)ULLONG_MAX) {
f1af9d3a
PH
895 idx_a = type_a->cnt - 1;
896 idx_b = type_b->cnt;
7bd0b0f0
TH
897 }
898
f1af9d3a
PH
899 for (; idx_a >= 0; idx_a--) {
900 struct memblock_region *m = &type_a->regions[idx_a];
901
7bd0b0f0
TH
902 phys_addr_t m_start = m->base;
903 phys_addr_t m_end = m->base + m->size;
f1af9d3a 904 int m_nid = memblock_get_region_node(m);
7bd0b0f0
TH
905
906 /* only memory regions are associated with nodes, check it */
f1af9d3a 907 if (nid != NUMA_NO_NODE && nid != m_nid)
7bd0b0f0
TH
908 continue;
909
55ac590c
TC
910 /* skip hotpluggable memory regions if needed */
911 if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
912 continue;
913
f1af9d3a
PH
914 if (!type_b) {
915 if (out_start)
916 *out_start = m_start;
917 if (out_end)
918 *out_end = m_end;
919 if (out_nid)
920 *out_nid = m_nid;
921 idx_a++;
922 *idx = (u32)idx_a | (u64)idx_b << 32;
923 return;
924 }
925
926 /* scan areas before each reservation */
927 for (; idx_b >= 0; idx_b--) {
928 struct memblock_region *r;
929 phys_addr_t r_start;
930 phys_addr_t r_end;
931
932 r = &type_b->regions[idx_b];
933 r_start = idx_b ? r[-1].base + r[-1].size : 0;
934 r_end = idx_b < type_b->cnt ?
935 r->base : ULLONG_MAX;
936 /*
937 * if idx_b advanced past idx_a,
938 * break out to advance idx_a
939 */
7bd0b0f0 940
7bd0b0f0
TH
941 if (r_end <= m_start)
942 break;
943 /* if the two regions intersect, we're done */
944 if (m_end > r_start) {
945 if (out_start)
946 *out_start = max(m_start, r_start);
947 if (out_end)
948 *out_end = min(m_end, r_end);
949 if (out_nid)
f1af9d3a 950 *out_nid = m_nid;
7bd0b0f0 951 if (m_start >= r_start)
f1af9d3a 952 idx_a--;
7bd0b0f0 953 else
f1af9d3a
PH
954 idx_b--;
955 *idx = (u32)idx_a | (u64)idx_b << 32;
7bd0b0f0
TH
956 return;
957 }
958 }
959 }
f1af9d3a 960 /* signal end of iteration */
7bd0b0f0
TH
961 *idx = ULLONG_MAX;
962}
963
7c0caeb8
TH
964#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
965/*
966 * Common iterator interface used to define for_each_mem_range().
967 */
968void __init_memblock __next_mem_pfn_range(int *idx, int nid,
969 unsigned long *out_start_pfn,
970 unsigned long *out_end_pfn, int *out_nid)
971{
972 struct memblock_type *type = &memblock.memory;
973 struct memblock_region *r;
974
975 while (++*idx < type->cnt) {
976 r = &type->regions[*idx];
977
978 if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size))
979 continue;
980 if (nid == MAX_NUMNODES || nid == r->nid)
981 break;
982 }
983 if (*idx >= type->cnt) {
984 *idx = -1;
985 return;
986 }
987
988 if (out_start_pfn)
989 *out_start_pfn = PFN_UP(r->base);
990 if (out_end_pfn)
991 *out_end_pfn = PFN_DOWN(r->base + r->size);
992 if (out_nid)
993 *out_nid = r->nid;
994}
995
996/**
997 * memblock_set_node - set node ID on memblock regions
998 * @base: base of area to set node ID for
999 * @size: size of area to set node ID for
e7e8de59 1000 * @type: memblock type to set node ID for
7c0caeb8
TH
1001 * @nid: node ID to set
1002 *
e7e8de59 1003 * Set the nid of memblock @type regions in [@base,@base+@size) to @nid.
7c0caeb8
TH
1004 * Regions which cross the area boundaries are split as necessary.
1005 *
1006 * RETURNS:
1007 * 0 on success, -errno on failure.
1008 */
1009int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size,
e7e8de59 1010 struct memblock_type *type, int nid)
7c0caeb8 1011{
6a9ceb31
TH
1012 int start_rgn, end_rgn;
1013 int i, ret;
7c0caeb8 1014
6a9ceb31
TH
1015 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
1016 if (ret)
1017 return ret;
7c0caeb8 1018
6a9ceb31 1019 for (i = start_rgn; i < end_rgn; i++)
e9d24ad3 1020 memblock_set_region_node(&type->regions[i], nid);
7c0caeb8
TH
1021
1022 memblock_merge_regions(type);
1023 return 0;
1024}
1025#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1026
7bd0b0f0
TH
1027static phys_addr_t __init memblock_alloc_base_nid(phys_addr_t size,
1028 phys_addr_t align, phys_addr_t max_addr,
1029 int nid)
95f72d1e 1030{
6ed311b2 1031 phys_addr_t found;
95f72d1e 1032
79f40fab
GS
1033 if (!align)
1034 align = SMP_CACHE_BYTES;
94f3d3af 1035
87029ee9 1036 found = memblock_find_in_range_node(size, align, 0, max_addr, nid);
9c8c27e2 1037 if (found && !memblock_reserve(found, size))
6ed311b2 1038 return found;
95f72d1e 1039
6ed311b2 1040 return 0;
95f72d1e
YL
1041}
1042
7bd0b0f0
TH
1043phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid)
1044{
1045 return memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE, nid);
1046}
1047
1048phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
1049{
b1154233 1050 return memblock_alloc_base_nid(size, align, max_addr, NUMA_NO_NODE);
7bd0b0f0
TH
1051}
1052
6ed311b2 1053phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
95f72d1e 1054{
6ed311b2
BH
1055 phys_addr_t alloc;
1056
1057 alloc = __memblock_alloc_base(size, align, max_addr);
1058
1059 if (alloc == 0)
1060 panic("ERROR: Failed to allocate 0x%llx bytes below 0x%llx.\n",
1061 (unsigned long long) size, (unsigned long long) max_addr);
1062
1063 return alloc;
95f72d1e
YL
1064}
1065
6ed311b2 1066phys_addr_t __init memblock_alloc(phys_addr_t size, phys_addr_t align)
95f72d1e 1067{
6ed311b2
BH
1068 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
1069}
95f72d1e 1070
9d1e2492
BH
1071phys_addr_t __init memblock_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid)
1072{
1073 phys_addr_t res = memblock_alloc_nid(size, align, nid);
1074
1075 if (res)
1076 return res;
15fb0972 1077 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
95f72d1e
YL
1078}
1079
26f09e9b
SS
1080/**
1081 * memblock_virt_alloc_internal - allocate boot memory block
1082 * @size: size of memory block to be allocated in bytes
1083 * @align: alignment of the region and block's size
1084 * @min_addr: the lower bound of the memory region to allocate (phys address)
1085 * @max_addr: the upper bound of the memory region to allocate (phys address)
1086 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1087 *
1088 * The @min_addr limit is dropped if it can not be satisfied and the allocation
1089 * will fall back to memory below @min_addr. Also, allocation may fall back
1090 * to any node in the system if the specified node can not
1091 * hold the requested memory.
1092 *
1093 * The allocation is performed from memory region limited by
1094 * memblock.current_limit if @max_addr == %BOOTMEM_ALLOC_ACCESSIBLE.
1095 *
1096 * The memory block is aligned on SMP_CACHE_BYTES if @align == 0.
1097 *
1098 * The phys address of allocated boot memory block is converted to virtual and
1099 * allocated memory is reset to 0.
1100 *
1101 * In addition, function sets the min_count to 0 using kmemleak_alloc for
1102 * allocated boot memory block, so that it is never reported as leaks.
1103 *
1104 * RETURNS:
1105 * Virtual address of allocated memory block on success, NULL on failure.
1106 */
1107static void * __init memblock_virt_alloc_internal(
1108 phys_addr_t size, phys_addr_t align,
1109 phys_addr_t min_addr, phys_addr_t max_addr,
1110 int nid)
1111{
1112 phys_addr_t alloc;
1113 void *ptr;
1114
560dca27
GS
1115 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
1116 nid = NUMA_NO_NODE;
26f09e9b
SS
1117
1118 /*
1119 * Detect any accidental use of these APIs after slab is ready, as at
1120 * this moment memblock may be deinitialized already and its
1121 * internal data may be destroyed (after execution of free_all_bootmem)
1122 */
1123 if (WARN_ON_ONCE(slab_is_available()))
1124 return kzalloc_node(size, GFP_NOWAIT, nid);
1125
1126 if (!align)
1127 align = SMP_CACHE_BYTES;
1128
f544e14f
YL
1129 if (max_addr > memblock.current_limit)
1130 max_addr = memblock.current_limit;
1131
26f09e9b
SS
1132again:
1133 alloc = memblock_find_in_range_node(size, align, min_addr, max_addr,
1134 nid);
1135 if (alloc)
1136 goto done;
1137
1138 if (nid != NUMA_NO_NODE) {
1139 alloc = memblock_find_in_range_node(size, align, min_addr,
1140 max_addr, NUMA_NO_NODE);
1141 if (alloc)
1142 goto done;
1143 }
1144
1145 if (min_addr) {
1146 min_addr = 0;
1147 goto again;
1148 } else {
1149 goto error;
1150 }
1151
1152done:
1153 memblock_reserve(alloc, size);
1154 ptr = phys_to_virt(alloc);
1155 memset(ptr, 0, size);
1156
1157 /*
1158 * The min_count is set to 0 so that bootmem allocated blocks
1159 * are never reported as leaks. This is because many of these blocks
1160 * are only referred via the physical address which is not
1161 * looked up by kmemleak.
1162 */
1163 kmemleak_alloc(ptr, size, 0, 0);
1164
1165 return ptr;
1166
1167error:
1168 return NULL;
1169}
1170
1171/**
1172 * memblock_virt_alloc_try_nid_nopanic - allocate boot memory block
1173 * @size: size of memory block to be allocated in bytes
1174 * @align: alignment of the region and block's size
1175 * @min_addr: the lower bound of the memory region from where the allocation
1176 * is preferred (phys address)
1177 * @max_addr: the upper bound of the memory region from where the allocation
1178 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1179 * allocate only from memory limited by memblock.current_limit value
1180 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1181 *
1182 * Public version of _memblock_virt_alloc_try_nid_nopanic() which provides
1183 * additional debug information (including caller info), if enabled.
1184 *
1185 * RETURNS:
1186 * Virtual address of allocated memory block on success, NULL on failure.
1187 */
1188void * __init memblock_virt_alloc_try_nid_nopanic(
1189 phys_addr_t size, phys_addr_t align,
1190 phys_addr_t min_addr, phys_addr_t max_addr,
1191 int nid)
1192{
1193 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1194 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1195 (u64)max_addr, (void *)_RET_IP_);
1196 return memblock_virt_alloc_internal(size, align, min_addr,
1197 max_addr, nid);
1198}
1199
1200/**
1201 * memblock_virt_alloc_try_nid - allocate boot memory block with panicking
1202 * @size: size of memory block to be allocated in bytes
1203 * @align: alignment of the region and block's size
1204 * @min_addr: the lower bound of the memory region from where the allocation
1205 * is preferred (phys address)
1206 * @max_addr: the upper bound of the memory region from where the allocation
1207 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1208 * allocate only from memory limited by memblock.current_limit value
1209 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1210 *
1211 * Public panicking version of _memblock_virt_alloc_try_nid_nopanic()
1212 * which provides debug information (including caller info), if enabled,
1213 * and panics if the request can not be satisfied.
1214 *
1215 * RETURNS:
1216 * Virtual address of allocated memory block on success, NULL on failure.
1217 */
1218void * __init memblock_virt_alloc_try_nid(
1219 phys_addr_t size, phys_addr_t align,
1220 phys_addr_t min_addr, phys_addr_t max_addr,
1221 int nid)
1222{
1223 void *ptr;
1224
1225 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1226 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1227 (u64)max_addr, (void *)_RET_IP_);
1228 ptr = memblock_virt_alloc_internal(size, align,
1229 min_addr, max_addr, nid);
1230 if (ptr)
1231 return ptr;
1232
1233 panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx\n",
1234 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1235 (u64)max_addr);
1236 return NULL;
1237}
1238
1239/**
1240 * __memblock_free_early - free boot memory block
1241 * @base: phys starting address of the boot memory block
1242 * @size: size of the boot memory block in bytes
1243 *
1244 * Free boot memory block previously allocated by memblock_virt_alloc_xx() API.
1245 * The freeing memory will not be released to the buddy allocator.
1246 */
1247void __init __memblock_free_early(phys_addr_t base, phys_addr_t size)
1248{
1249 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1250 __func__, (u64)base, (u64)base + size - 1,
1251 (void *)_RET_IP_);
1252 kmemleak_free_part(__va(base), size);
f1af9d3a 1253 memblock_remove_range(&memblock.reserved, base, size);
26f09e9b
SS
1254}
1255
1256/*
1257 * __memblock_free_late - free bootmem block pages directly to buddy allocator
1258 * @addr: phys starting address of the boot memory block
1259 * @size: size of the boot memory block in bytes
1260 *
1261 * This is only useful when the bootmem allocator has already been torn
1262 * down, but we are still initializing the system. Pages are released directly
1263 * to the buddy allocator, no bootmem metadata is updated because it is gone.
1264 */
1265void __init __memblock_free_late(phys_addr_t base, phys_addr_t size)
1266{
1267 u64 cursor, end;
1268
1269 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1270 __func__, (u64)base, (u64)base + size - 1,
1271 (void *)_RET_IP_);
1272 kmemleak_free_part(__va(base), size);
1273 cursor = PFN_UP(base);
1274 end = PFN_DOWN(base + size);
1275
1276 for (; cursor < end; cursor++) {
1277 __free_pages_bootmem(pfn_to_page(cursor), 0);
1278 totalram_pages++;
1279 }
1280}
9d1e2492
BH
1281
1282/*
1283 * Remaining API functions
1284 */
1285
2898cc4c 1286phys_addr_t __init memblock_phys_mem_size(void)
95f72d1e 1287{
1440c4e2 1288 return memblock.memory.total_size;
95f72d1e
YL
1289}
1290
595ad9af
YL
1291phys_addr_t __init memblock_mem_size(unsigned long limit_pfn)
1292{
1293 unsigned long pages = 0;
1294 struct memblock_region *r;
1295 unsigned long start_pfn, end_pfn;
1296
1297 for_each_memblock(memory, r) {
1298 start_pfn = memblock_region_memory_base_pfn(r);
1299 end_pfn = memblock_region_memory_end_pfn(r);
1300 start_pfn = min_t(unsigned long, start_pfn, limit_pfn);
1301 end_pfn = min_t(unsigned long, end_pfn, limit_pfn);
1302 pages += end_pfn - start_pfn;
1303 }
1304
16763230 1305 return PFN_PHYS(pages);
595ad9af
YL
1306}
1307
0a93ebef
SR
1308/* lowest address */
1309phys_addr_t __init_memblock memblock_start_of_DRAM(void)
1310{
1311 return memblock.memory.regions[0].base;
1312}
1313
10d06439 1314phys_addr_t __init_memblock memblock_end_of_DRAM(void)
95f72d1e
YL
1315{
1316 int idx = memblock.memory.cnt - 1;
1317
e3239ff9 1318 return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size);
95f72d1e
YL
1319}
1320
c0ce8fef 1321void __init memblock_enforce_memory_limit(phys_addr_t limit)
95f72d1e 1322{
c0ce8fef 1323 phys_addr_t max_addr = (phys_addr_t)ULLONG_MAX;
136199f0 1324 struct memblock_region *r;
95f72d1e 1325
c0ce8fef 1326 if (!limit)
95f72d1e
YL
1327 return;
1328
c0ce8fef 1329 /* find out max address */
136199f0 1330 for_each_memblock(memory, r) {
c0ce8fef
TH
1331 if (limit <= r->size) {
1332 max_addr = r->base + limit;
1333 break;
95f72d1e 1334 }
c0ce8fef 1335 limit -= r->size;
95f72d1e 1336 }
c0ce8fef
TH
1337
1338 /* truncate both memory and reserved regions */
f1af9d3a
PH
1339 memblock_remove_range(&memblock.memory, max_addr,
1340 (phys_addr_t)ULLONG_MAX);
1341 memblock_remove_range(&memblock.reserved, max_addr,
1342 (phys_addr_t)ULLONG_MAX);
95f72d1e
YL
1343}
1344
cd79481d 1345static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr)
72d4b0b4
BH
1346{
1347 unsigned int left = 0, right = type->cnt;
1348
1349 do {
1350 unsigned int mid = (right + left) / 2;
1351
1352 if (addr < type->regions[mid].base)
1353 right = mid;
1354 else if (addr >= (type->regions[mid].base +
1355 type->regions[mid].size))
1356 left = mid + 1;
1357 else
1358 return mid;
1359 } while (left < right);
1360 return -1;
1361}
1362
2898cc4c 1363int __init memblock_is_reserved(phys_addr_t addr)
95f72d1e 1364{
72d4b0b4
BH
1365 return memblock_search(&memblock.reserved, addr) != -1;
1366}
95f72d1e 1367
3661ca66 1368int __init_memblock memblock_is_memory(phys_addr_t addr)
72d4b0b4
BH
1369{
1370 return memblock_search(&memblock.memory, addr) != -1;
1371}
1372
e76b63f8
YL
1373#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1374int __init_memblock memblock_search_pfn_nid(unsigned long pfn,
1375 unsigned long *start_pfn, unsigned long *end_pfn)
1376{
1377 struct memblock_type *type = &memblock.memory;
16763230 1378 int mid = memblock_search(type, PFN_PHYS(pfn));
e76b63f8
YL
1379
1380 if (mid == -1)
1381 return -1;
1382
1383 *start_pfn = type->regions[mid].base >> PAGE_SHIFT;
1384 *end_pfn = (type->regions[mid].base + type->regions[mid].size)
1385 >> PAGE_SHIFT;
1386
1387 return type->regions[mid].nid;
1388}
1389#endif
1390
eab30949
SB
1391/**
1392 * memblock_is_region_memory - check if a region is a subset of memory
1393 * @base: base of region to check
1394 * @size: size of region to check
1395 *
1396 * Check if the region [@base, @base+@size) is a subset of a memory block.
1397 *
1398 * RETURNS:
1399 * 0 if false, non-zero if true
1400 */
3661ca66 1401int __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size)
72d4b0b4 1402{
abb65272 1403 int idx = memblock_search(&memblock.memory, base);
eb18f1b5 1404 phys_addr_t end = base + memblock_cap_size(base, &size);
72d4b0b4
BH
1405
1406 if (idx == -1)
1407 return 0;
abb65272
TV
1408 return memblock.memory.regions[idx].base <= base &&
1409 (memblock.memory.regions[idx].base +
eb18f1b5 1410 memblock.memory.regions[idx].size) >= end;
95f72d1e
YL
1411}
1412
eab30949
SB
1413/**
1414 * memblock_is_region_reserved - check if a region intersects reserved memory
1415 * @base: base of region to check
1416 * @size: size of region to check
1417 *
1418 * Check if the region [@base, @base+@size) intersects a reserved memory block.
1419 *
1420 * RETURNS:
1421 * 0 if false, non-zero if true
1422 */
10d06439 1423int __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size)
95f72d1e 1424{
eb18f1b5 1425 memblock_cap_size(base, &size);
f1c2c19c 1426 return memblock_overlaps_region(&memblock.reserved, base, size) >= 0;
95f72d1e
YL
1427}
1428
6ede1fd3
YL
1429void __init_memblock memblock_trim_memory(phys_addr_t align)
1430{
6ede1fd3 1431 phys_addr_t start, end, orig_start, orig_end;
136199f0 1432 struct memblock_region *r;
6ede1fd3 1433
136199f0
EM
1434 for_each_memblock(memory, r) {
1435 orig_start = r->base;
1436 orig_end = r->base + r->size;
6ede1fd3
YL
1437 start = round_up(orig_start, align);
1438 end = round_down(orig_end, align);
1439
1440 if (start == orig_start && end == orig_end)
1441 continue;
1442
1443 if (start < end) {
136199f0
EM
1444 r->base = start;
1445 r->size = end - start;
6ede1fd3 1446 } else {
136199f0
EM
1447 memblock_remove_region(&memblock.memory,
1448 r - memblock.memory.regions);
1449 r--;
6ede1fd3
YL
1450 }
1451 }
1452}
e63075a3 1453
3661ca66 1454void __init_memblock memblock_set_current_limit(phys_addr_t limit)
e63075a3
BH
1455{
1456 memblock.current_limit = limit;
1457}
1458
fec51014
LA
1459phys_addr_t __init_memblock memblock_get_current_limit(void)
1460{
1461 return memblock.current_limit;
1462}
1463
7c0caeb8 1464static void __init_memblock memblock_dump(struct memblock_type *type, char *name)
6ed311b2
BH
1465{
1466 unsigned long long base, size;
66a20757 1467 unsigned long flags;
6ed311b2
BH
1468 int i;
1469
7c0caeb8 1470 pr_info(" %s.cnt = 0x%lx\n", name, type->cnt);
6ed311b2 1471
7c0caeb8
TH
1472 for (i = 0; i < type->cnt; i++) {
1473 struct memblock_region *rgn = &type->regions[i];
1474 char nid_buf[32] = "";
1475
1476 base = rgn->base;
1477 size = rgn->size;
66a20757 1478 flags = rgn->flags;
7c0caeb8
TH
1479#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1480 if (memblock_get_region_node(rgn) != MAX_NUMNODES)
1481 snprintf(nid_buf, sizeof(nid_buf), " on node %d",
1482 memblock_get_region_node(rgn));
1483#endif
66a20757
TC
1484 pr_info(" %s[%#x]\t[%#016llx-%#016llx], %#llx bytes%s flags: %#lx\n",
1485 name, i, base, base + size - 1, size, nid_buf, flags);
6ed311b2
BH
1486 }
1487}
1488
4ff7b82f 1489void __init_memblock __memblock_dump_all(void)
6ed311b2 1490{
6ed311b2 1491 pr_info("MEMBLOCK configuration:\n");
1440c4e2
TH
1492 pr_info(" memory size = %#llx reserved size = %#llx\n",
1493 (unsigned long long)memblock.memory.total_size,
1494 (unsigned long long)memblock.reserved.total_size);
6ed311b2
BH
1495
1496 memblock_dump(&memblock.memory, "memory");
1497 memblock_dump(&memblock.reserved, "reserved");
1498}
1499
1aadc056 1500void __init memblock_allow_resize(void)
6ed311b2 1501{
142b45a7 1502 memblock_can_resize = 1;
6ed311b2
BH
1503}
1504
6ed311b2
BH
1505static int __init early_memblock(char *p)
1506{
1507 if (p && strstr(p, "debug"))
1508 memblock_debug = 1;
1509 return 0;
1510}
1511early_param("memblock", early_memblock);
1512
c378ddd5 1513#if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK)
6d03b885
BH
1514
1515static int memblock_debug_show(struct seq_file *m, void *private)
1516{
1517 struct memblock_type *type = m->private;
1518 struct memblock_region *reg;
1519 int i;
1520
1521 for (i = 0; i < type->cnt; i++) {
1522 reg = &type->regions[i];
1523 seq_printf(m, "%4d: ", i);
1524 if (sizeof(phys_addr_t) == 4)
1525 seq_printf(m, "0x%08lx..0x%08lx\n",
1526 (unsigned long)reg->base,
1527 (unsigned long)(reg->base + reg->size - 1));
1528 else
1529 seq_printf(m, "0x%016llx..0x%016llx\n",
1530 (unsigned long long)reg->base,
1531 (unsigned long long)(reg->base + reg->size - 1));
1532
1533 }
1534 return 0;
1535}
1536
1537static int memblock_debug_open(struct inode *inode, struct file *file)
1538{
1539 return single_open(file, memblock_debug_show, inode->i_private);
1540}
1541
1542static const struct file_operations memblock_debug_fops = {
1543 .open = memblock_debug_open,
1544 .read = seq_read,
1545 .llseek = seq_lseek,
1546 .release = single_release,
1547};
1548
1549static int __init memblock_init_debugfs(void)
1550{
1551 struct dentry *root = debugfs_create_dir("memblock", NULL);
1552 if (!root)
1553 return -ENXIO;
1554 debugfs_create_file("memory", S_IRUGO, root, &memblock.memory, &memblock_debug_fops);
1555 debugfs_create_file("reserved", S_IRUGO, root, &memblock.reserved, &memblock_debug_fops);
1556
1557 return 0;
1558}
1559__initcall(memblock_init_debugfs);
1560
1561#endif /* CONFIG_DEBUG_FS */
This page took 0.539724 seconds and 5 git commands to generate.