]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - mm/memblock.c
btrfs: cleaner_kthread() doesn't need explicit freeze
[mirror_ubuntu-artful-kernel.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
a3f5bafc 57static bool system_has_some_mirror __initdata_memblock = false;
1aadc056 58static int memblock_can_resize __initdata_memblock;
181eb394
GS
59static int memblock_memory_in_slab __initdata_memblock = 0;
60static int memblock_reserved_in_slab __initdata_memblock = 0;
95f72d1e 61
a3f5bafc
TL
62ulong __init_memblock choose_memblock_flags(void)
63{
64 return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE;
65}
66
142b45a7 67/* inline so we don't get a warning when pr_debug is compiled out */
c2233116
RP
68static __init_memblock const char *
69memblock_type_name(struct memblock_type *type)
142b45a7
BH
70{
71 if (type == &memblock.memory)
72 return "memory";
73 else if (type == &memblock.reserved)
74 return "reserved";
75 else
76 return "unknown";
77}
78
eb18f1b5
TH
79/* adjust *@size so that (@base + *@size) doesn't overflow, return new size */
80static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size)
81{
82 return *size = min(*size, (phys_addr_t)ULLONG_MAX - base);
83}
84
6ed311b2
BH
85/*
86 * Address comparison utilities
87 */
10d06439 88static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1,
2898cc4c 89 phys_addr_t base2, phys_addr_t size2)
95f72d1e
YL
90{
91 return ((base1 < (base2 + size2)) && (base2 < (base1 + size1)));
92}
93
95cf82ec 94bool __init_memblock memblock_overlaps_region(struct memblock_type *type,
2d7d3eb2 95 phys_addr_t base, phys_addr_t size)
6ed311b2
BH
96{
97 unsigned long i;
98
f14516fb
AK
99 for (i = 0; i < type->cnt; i++)
100 if (memblock_addrs_overlap(base, size, type->regions[i].base,
101 type->regions[i].size))
6ed311b2 102 break;
c5c5c9d1 103 return i < type->cnt;
6ed311b2
BH
104}
105
79442ed1
TC
106/*
107 * __memblock_find_range_bottom_up - find free area utility in bottom-up
108 * @start: start of candidate range
109 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
110 * @size: size of free area to find
111 * @align: alignment of free area to find
b1154233 112 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
fc6daaf9 113 * @flags: pick from blocks based on memory attributes
79442ed1
TC
114 *
115 * Utility called from memblock_find_in_range_node(), find free area bottom-up.
116 *
117 * RETURNS:
118 * Found address on success, 0 on failure.
119 */
120static phys_addr_t __init_memblock
121__memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
fc6daaf9
TL
122 phys_addr_t size, phys_addr_t align, int nid,
123 ulong flags)
79442ed1
TC
124{
125 phys_addr_t this_start, this_end, cand;
126 u64 i;
127
fc6daaf9 128 for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) {
79442ed1
TC
129 this_start = clamp(this_start, start, end);
130 this_end = clamp(this_end, start, end);
131
132 cand = round_up(this_start, align);
133 if (cand < this_end && this_end - cand >= size)
134 return cand;
135 }
136
137 return 0;
138}
139
7bd0b0f0 140/**
1402899e 141 * __memblock_find_range_top_down - find free area utility, in top-down
7bd0b0f0
TH
142 * @start: start of candidate range
143 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
144 * @size: size of free area to find
145 * @align: alignment of free area to find
b1154233 146 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
fc6daaf9 147 * @flags: pick from blocks based on memory attributes
7bd0b0f0 148 *
1402899e 149 * Utility called from memblock_find_in_range_node(), find free area top-down.
7bd0b0f0
TH
150 *
151 * RETURNS:
79442ed1 152 * Found address on success, 0 on failure.
6ed311b2 153 */
1402899e
TC
154static phys_addr_t __init_memblock
155__memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
fc6daaf9
TL
156 phys_addr_t size, phys_addr_t align, int nid,
157 ulong flags)
f7210e6c
TC
158{
159 phys_addr_t this_start, this_end, cand;
160 u64 i;
161
fc6daaf9
TL
162 for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end,
163 NULL) {
f7210e6c
TC
164 this_start = clamp(this_start, start, end);
165 this_end = clamp(this_end, start, end);
166
167 if (this_end < size)
168 continue;
169
170 cand = round_down(this_end - size, align);
171 if (cand >= this_start)
172 return cand;
173 }
1402899e 174
f7210e6c
TC
175 return 0;
176}
6ed311b2 177
1402899e
TC
178/**
179 * memblock_find_in_range_node - find free area in given range and node
1402899e
TC
180 * @size: size of free area to find
181 * @align: alignment of free area to find
87029ee9
GS
182 * @start: start of candidate range
183 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
b1154233 184 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
fc6daaf9 185 * @flags: pick from blocks based on memory attributes
1402899e
TC
186 *
187 * Find @size free area aligned to @align in the specified range and node.
188 *
79442ed1
TC
189 * When allocation direction is bottom-up, the @start should be greater
190 * than the end of the kernel image. Otherwise, it will be trimmed. The
191 * reason is that we want the bottom-up allocation just near the kernel
192 * image so it is highly likely that the allocated memory and the kernel
193 * will reside in the same node.
194 *
195 * If bottom-up allocation failed, will try to allocate memory top-down.
196 *
1402899e 197 * RETURNS:
79442ed1 198 * Found address on success, 0 on failure.
1402899e 199 */
87029ee9
GS
200phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
201 phys_addr_t align, phys_addr_t start,
fc6daaf9 202 phys_addr_t end, int nid, ulong flags)
1402899e 203{
0cfb8f0c 204 phys_addr_t kernel_end, ret;
79442ed1 205
1402899e
TC
206 /* pump up @end */
207 if (end == MEMBLOCK_ALLOC_ACCESSIBLE)
208 end = memblock.current_limit;
209
210 /* avoid allocating the first page */
211 start = max_t(phys_addr_t, start, PAGE_SIZE);
212 end = max(start, end);
79442ed1
TC
213 kernel_end = __pa_symbol(_end);
214
215 /*
216 * try bottom-up allocation only when bottom-up mode
217 * is set and @end is above the kernel image.
218 */
219 if (memblock_bottom_up() && end > kernel_end) {
220 phys_addr_t bottom_up_start;
221
222 /* make sure we will allocate above the kernel */
223 bottom_up_start = max(start, kernel_end);
224
225 /* ok, try bottom-up allocation first */
226 ret = __memblock_find_range_bottom_up(bottom_up_start, end,
fc6daaf9 227 size, align, nid, flags);
79442ed1
TC
228 if (ret)
229 return ret;
230
231 /*
232 * we always limit bottom-up allocation above the kernel,
233 * but top-down allocation doesn't have the limit, so
234 * retrying top-down allocation may succeed when bottom-up
235 * allocation failed.
236 *
237 * bottom-up allocation is expected to be fail very rarely,
238 * so we use WARN_ONCE() here to see the stack trace if
239 * fail happens.
240 */
241 WARN_ONCE(1, "memblock: bottom-up allocation failed, "
242 "memory hotunplug may be affected\n");
243 }
1402899e 244
fc6daaf9
TL
245 return __memblock_find_range_top_down(start, end, size, align, nid,
246 flags);
1402899e
TC
247}
248
7bd0b0f0
TH
249/**
250 * memblock_find_in_range - find free area in given range
251 * @start: start of candidate range
252 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
253 * @size: size of free area to find
254 * @align: alignment of free area to find
255 *
256 * Find @size free area aligned to @align in the specified range.
257 *
258 * RETURNS:
79442ed1 259 * Found address on success, 0 on failure.
fc769a8e 260 */
7bd0b0f0
TH
261phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start,
262 phys_addr_t end, phys_addr_t size,
263 phys_addr_t align)
6ed311b2 264{
a3f5bafc
TL
265 phys_addr_t ret;
266 ulong flags = choose_memblock_flags();
267
268again:
269 ret = memblock_find_in_range_node(size, align, start, end,
270 NUMA_NO_NODE, flags);
271
272 if (!ret && (flags & MEMBLOCK_MIRROR)) {
273 pr_warn("Could not allocate %pap bytes of mirrored memory\n",
274 &size);
275 flags &= ~MEMBLOCK_MIRROR;
276 goto again;
277 }
278
279 return ret;
6ed311b2
BH
280}
281
10d06439 282static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r)
95f72d1e 283{
1440c4e2 284 type->total_size -= type->regions[r].size;
7c0caeb8
TH
285 memmove(&type->regions[r], &type->regions[r + 1],
286 (type->cnt - (r + 1)) * sizeof(type->regions[r]));
e3239ff9 287 type->cnt--;
95f72d1e 288
8f7a6605
BH
289 /* Special case for empty arrays */
290 if (type->cnt == 0) {
1440c4e2 291 WARN_ON(type->total_size != 0);
8f7a6605
BH
292 type->cnt = 1;
293 type->regions[0].base = 0;
294 type->regions[0].size = 0;
66a20757 295 type->regions[0].flags = 0;
7c0caeb8 296 memblock_set_region_node(&type->regions[0], MAX_NUMNODES);
8f7a6605 297 }
95f72d1e
YL
298}
299
354f17e1
PH
300#ifdef CONFIG_ARCH_DISCARD_MEMBLOCK
301
29f67386
YL
302phys_addr_t __init_memblock get_allocated_memblock_reserved_regions_info(
303 phys_addr_t *addr)
304{
305 if (memblock.reserved.regions == memblock_reserved_init_regions)
306 return 0;
307
308 *addr = __pa(memblock.reserved.regions);
309
310 return PAGE_ALIGN(sizeof(struct memblock_region) *
311 memblock.reserved.max);
312}
313
5e270e25
PH
314phys_addr_t __init_memblock get_allocated_memblock_memory_regions_info(
315 phys_addr_t *addr)
316{
317 if (memblock.memory.regions == memblock_memory_init_regions)
318 return 0;
319
320 *addr = __pa(memblock.memory.regions);
321
322 return PAGE_ALIGN(sizeof(struct memblock_region) *
323 memblock.memory.max);
324}
325
326#endif
327
48c3b583
GP
328/**
329 * memblock_double_array - double the size of the memblock regions array
330 * @type: memblock type of the regions array being doubled
331 * @new_area_start: starting address of memory range to avoid overlap with
332 * @new_area_size: size of memory range to avoid overlap with
333 *
334 * Double the size of the @type regions array. If memblock is being used to
335 * allocate memory for a new reserved regions array and there is a previously
336 * allocated memory range [@new_area_start,@new_area_start+@new_area_size]
337 * waiting to be reserved, ensure the memory used by the new array does
338 * not overlap.
339 *
340 * RETURNS:
341 * 0 on success, -1 on failure.
342 */
343static int __init_memblock memblock_double_array(struct memblock_type *type,
344 phys_addr_t new_area_start,
345 phys_addr_t new_area_size)
142b45a7
BH
346{
347 struct memblock_region *new_array, *old_array;
29f67386 348 phys_addr_t old_alloc_size, new_alloc_size;
142b45a7
BH
349 phys_addr_t old_size, new_size, addr;
350 int use_slab = slab_is_available();
181eb394 351 int *in_slab;
142b45a7
BH
352
353 /* We don't allow resizing until we know about the reserved regions
354 * of memory that aren't suitable for allocation
355 */
356 if (!memblock_can_resize)
357 return -1;
358
142b45a7
BH
359 /* Calculate new doubled size */
360 old_size = type->max * sizeof(struct memblock_region);
361 new_size = old_size << 1;
29f67386
YL
362 /*
363 * We need to allocated new one align to PAGE_SIZE,
364 * so we can free them completely later.
365 */
366 old_alloc_size = PAGE_ALIGN(old_size);
367 new_alloc_size = PAGE_ALIGN(new_size);
142b45a7 368
181eb394
GS
369 /* Retrieve the slab flag */
370 if (type == &memblock.memory)
371 in_slab = &memblock_memory_in_slab;
372 else
373 in_slab = &memblock_reserved_in_slab;
374
142b45a7
BH
375 /* Try to find some space for it.
376 *
377 * WARNING: We assume that either slab_is_available() and we use it or
fd07383b
AM
378 * we use MEMBLOCK for allocations. That means that this is unsafe to
379 * use when bootmem is currently active (unless bootmem itself is
380 * implemented on top of MEMBLOCK which isn't the case yet)
142b45a7
BH
381 *
382 * This should however not be an issue for now, as we currently only
fd07383b
AM
383 * call into MEMBLOCK while it's still active, or much later when slab
384 * is active for memory hotplug operations
142b45a7
BH
385 */
386 if (use_slab) {
387 new_array = kmalloc(new_size, GFP_KERNEL);
1f5026a7 388 addr = new_array ? __pa(new_array) : 0;
4e2f0775 389 } else {
48c3b583
GP
390 /* only exclude range when trying to double reserved.regions */
391 if (type != &memblock.reserved)
392 new_area_start = new_area_size = 0;
393
394 addr = memblock_find_in_range(new_area_start + new_area_size,
395 memblock.current_limit,
29f67386 396 new_alloc_size, PAGE_SIZE);
48c3b583
GP
397 if (!addr && new_area_size)
398 addr = memblock_find_in_range(0,
fd07383b
AM
399 min(new_area_start, memblock.current_limit),
400 new_alloc_size, PAGE_SIZE);
48c3b583 401
15674868 402 new_array = addr ? __va(addr) : NULL;
4e2f0775 403 }
1f5026a7 404 if (!addr) {
142b45a7
BH
405 pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n",
406 memblock_type_name(type), type->max, type->max * 2);
407 return -1;
408 }
142b45a7 409
fd07383b
AM
410 memblock_dbg("memblock: %s is doubled to %ld at [%#010llx-%#010llx]",
411 memblock_type_name(type), type->max * 2, (u64)addr,
412 (u64)addr + new_size - 1);
ea9e4376 413
fd07383b
AM
414 /*
415 * Found space, we now need to move the array over before we add the
416 * reserved region since it may be our reserved array itself that is
417 * full.
142b45a7
BH
418 */
419 memcpy(new_array, type->regions, old_size);
420 memset(new_array + type->max, 0, old_size);
421 old_array = type->regions;
422 type->regions = new_array;
423 type->max <<= 1;
424
fd07383b 425 /* Free old array. We needn't free it if the array is the static one */
181eb394
GS
426 if (*in_slab)
427 kfree(old_array);
428 else if (old_array != memblock_memory_init_regions &&
429 old_array != memblock_reserved_init_regions)
29f67386 430 memblock_free(__pa(old_array), old_alloc_size);
142b45a7 431
fd07383b
AM
432 /*
433 * Reserve the new array if that comes from the memblock. Otherwise, we
434 * needn't do it
181eb394
GS
435 */
436 if (!use_slab)
29f67386 437 BUG_ON(memblock_reserve(addr, new_alloc_size));
181eb394
GS
438
439 /* Update slab flag */
440 *in_slab = use_slab;
441
142b45a7
BH
442 return 0;
443}
444
784656f9
TH
445/**
446 * memblock_merge_regions - merge neighboring compatible regions
447 * @type: memblock type to scan
448 *
449 * Scan @type and merge neighboring compatible regions.
450 */
451static void __init_memblock memblock_merge_regions(struct memblock_type *type)
95f72d1e 452{
784656f9 453 int i = 0;
95f72d1e 454
784656f9
TH
455 /* cnt never goes below 1 */
456 while (i < type->cnt - 1) {
457 struct memblock_region *this = &type->regions[i];
458 struct memblock_region *next = &type->regions[i + 1];
95f72d1e 459
7c0caeb8
TH
460 if (this->base + this->size != next->base ||
461 memblock_get_region_node(this) !=
66a20757
TC
462 memblock_get_region_node(next) ||
463 this->flags != next->flags) {
784656f9
TH
464 BUG_ON(this->base + this->size > next->base);
465 i++;
466 continue;
8f7a6605
BH
467 }
468
784656f9 469 this->size += next->size;
c0232ae8
LF
470 /* move forward from next + 1, index of which is i + 2 */
471 memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next));
784656f9 472 type->cnt--;
95f72d1e 473 }
784656f9 474}
95f72d1e 475
784656f9
TH
476/**
477 * memblock_insert_region - insert new memblock region
209ff86d
TC
478 * @type: memblock type to insert into
479 * @idx: index for the insertion point
480 * @base: base address of the new region
481 * @size: size of the new region
482 * @nid: node id of the new region
66a20757 483 * @flags: flags of the new region
784656f9
TH
484 *
485 * Insert new memblock region [@base,@base+@size) into @type at @idx.
486 * @type must already have extra room to accomodate the new region.
487 */
488static void __init_memblock memblock_insert_region(struct memblock_type *type,
489 int idx, phys_addr_t base,
66a20757
TC
490 phys_addr_t size,
491 int nid, unsigned long flags)
784656f9
TH
492{
493 struct memblock_region *rgn = &type->regions[idx];
494
495 BUG_ON(type->cnt >= type->max);
496 memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn));
497 rgn->base = base;
498 rgn->size = size;
66a20757 499 rgn->flags = flags;
7c0caeb8 500 memblock_set_region_node(rgn, nid);
784656f9 501 type->cnt++;
1440c4e2 502 type->total_size += size;
784656f9
TH
503}
504
505/**
f1af9d3a 506 * memblock_add_range - add new memblock region
784656f9
TH
507 * @type: memblock type to add new region into
508 * @base: base address of the new region
509 * @size: size of the new region
7fb0bc3f 510 * @nid: nid of the new region
66a20757 511 * @flags: flags of the new region
784656f9
TH
512 *
513 * Add new memblock region [@base,@base+@size) into @type. The new region
514 * is allowed to overlap with existing ones - overlaps don't affect already
515 * existing regions. @type is guaranteed to be minimal (all neighbouring
516 * compatible regions are merged) after the addition.
517 *
518 * RETURNS:
519 * 0 on success, -errno on failure.
520 */
f1af9d3a 521int __init_memblock memblock_add_range(struct memblock_type *type,
66a20757
TC
522 phys_addr_t base, phys_addr_t size,
523 int nid, unsigned long flags)
784656f9
TH
524{
525 bool insert = false;
eb18f1b5
TH
526 phys_addr_t obase = base;
527 phys_addr_t end = base + memblock_cap_size(base, &size);
8c9c1701
AK
528 int idx, nr_new;
529 struct memblock_region *rgn;
784656f9 530
b3dc627c
TH
531 if (!size)
532 return 0;
533
784656f9
TH
534 /* special case for empty array */
535 if (type->regions[0].size == 0) {
1440c4e2 536 WARN_ON(type->cnt != 1 || type->total_size);
8f7a6605
BH
537 type->regions[0].base = base;
538 type->regions[0].size = size;
66a20757 539 type->regions[0].flags = flags;
7fb0bc3f 540 memblock_set_region_node(&type->regions[0], nid);
1440c4e2 541 type->total_size = size;
8f7a6605 542 return 0;
95f72d1e 543 }
784656f9
TH
544repeat:
545 /*
546 * The following is executed twice. Once with %false @insert and
547 * then with %true. The first counts the number of regions needed
548 * to accomodate the new area. The second actually inserts them.
142b45a7 549 */
784656f9
TH
550 base = obase;
551 nr_new = 0;
95f72d1e 552
8c9c1701 553 for_each_memblock_type(type, rgn) {
784656f9
TH
554 phys_addr_t rbase = rgn->base;
555 phys_addr_t rend = rbase + rgn->size;
556
557 if (rbase >= end)
95f72d1e 558 break;
784656f9
TH
559 if (rend <= base)
560 continue;
561 /*
562 * @rgn overlaps. If it separates the lower part of new
563 * area, insert that portion.
564 */
565 if (rbase > base) {
c0a29498
WY
566#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
567 WARN_ON(nid != memblock_get_region_node(rgn));
568#endif
4fcab5f4 569 WARN_ON(flags != rgn->flags);
784656f9
TH
570 nr_new++;
571 if (insert)
8c9c1701 572 memblock_insert_region(type, idx++, base,
66a20757
TC
573 rbase - base, nid,
574 flags);
95f72d1e 575 }
784656f9
TH
576 /* area below @rend is dealt with, forget about it */
577 base = min(rend, end);
95f72d1e 578 }
784656f9
TH
579
580 /* insert the remaining portion */
581 if (base < end) {
582 nr_new++;
583 if (insert)
8c9c1701 584 memblock_insert_region(type, idx, base, end - base,
66a20757 585 nid, flags);
95f72d1e 586 }
95f72d1e 587
784656f9
TH
588 /*
589 * If this was the first round, resize array and repeat for actual
590 * insertions; otherwise, merge and return.
142b45a7 591 */
784656f9
TH
592 if (!insert) {
593 while (type->cnt + nr_new > type->max)
48c3b583 594 if (memblock_double_array(type, obase, size) < 0)
784656f9
TH
595 return -ENOMEM;
596 insert = true;
597 goto repeat;
598 } else {
599 memblock_merge_regions(type);
600 return 0;
142b45a7 601 }
95f72d1e
YL
602}
603
7fb0bc3f
TH
604int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size,
605 int nid)
606{
f1af9d3a 607 return memblock_add_range(&memblock.memory, base, size, nid, 0);
7fb0bc3f
TH
608}
609
6a4055bc
AK
610static int __init_memblock memblock_add_region(phys_addr_t base,
611 phys_addr_t size,
612 int nid,
613 unsigned long flags)
614{
567d117b 615 struct memblock_type *type = &memblock.memory;
6a4055bc
AK
616
617 memblock_dbg("memblock_add: [%#016llx-%#016llx] flags %#02lx %pF\n",
618 (unsigned long long)base,
619 (unsigned long long)base + size - 1,
620 flags, (void *)_RET_IP_);
621
567d117b 622 return memblock_add_range(type, base, size, nid, flags);
6a4055bc
AK
623}
624
581adcbe 625int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
95f72d1e 626{
6a4055bc 627 return memblock_add_region(base, size, MAX_NUMNODES, 0);
95f72d1e
YL
628}
629
6a9ceb31
TH
630/**
631 * memblock_isolate_range - isolate given range into disjoint memblocks
632 * @type: memblock type to isolate range for
633 * @base: base of range to isolate
634 * @size: size of range to isolate
635 * @start_rgn: out parameter for the start of isolated region
636 * @end_rgn: out parameter for the end of isolated region
637 *
638 * Walk @type and ensure that regions don't cross the boundaries defined by
639 * [@base,@base+@size). Crossing regions are split at the boundaries,
640 * which may create at most two more regions. The index of the first
641 * region inside the range is returned in *@start_rgn and end in *@end_rgn.
642 *
643 * RETURNS:
644 * 0 on success, -errno on failure.
645 */
646static int __init_memblock memblock_isolate_range(struct memblock_type *type,
647 phys_addr_t base, phys_addr_t size,
648 int *start_rgn, int *end_rgn)
649{
eb18f1b5 650 phys_addr_t end = base + memblock_cap_size(base, &size);
8c9c1701
AK
651 int idx;
652 struct memblock_region *rgn;
6a9ceb31
TH
653
654 *start_rgn = *end_rgn = 0;
655
b3dc627c
TH
656 if (!size)
657 return 0;
658
6a9ceb31
TH
659 /* we'll create at most two more regions */
660 while (type->cnt + 2 > type->max)
48c3b583 661 if (memblock_double_array(type, base, size) < 0)
6a9ceb31
TH
662 return -ENOMEM;
663
8c9c1701 664 for_each_memblock_type(type, rgn) {
6a9ceb31
TH
665 phys_addr_t rbase = rgn->base;
666 phys_addr_t rend = rbase + rgn->size;
667
668 if (rbase >= end)
669 break;
670 if (rend <= base)
671 continue;
672
673 if (rbase < base) {
674 /*
675 * @rgn intersects from below. Split and continue
676 * to process the next region - the new top half.
677 */
678 rgn->base = base;
1440c4e2
TH
679 rgn->size -= base - rbase;
680 type->total_size -= base - rbase;
8c9c1701 681 memblock_insert_region(type, idx, rbase, base - rbase,
66a20757
TC
682 memblock_get_region_node(rgn),
683 rgn->flags);
6a9ceb31
TH
684 } else if (rend > end) {
685 /*
686 * @rgn intersects from above. Split and redo the
687 * current region - the new bottom half.
688 */
689 rgn->base = end;
1440c4e2
TH
690 rgn->size -= end - rbase;
691 type->total_size -= end - rbase;
8c9c1701 692 memblock_insert_region(type, idx--, rbase, end - rbase,
66a20757
TC
693 memblock_get_region_node(rgn),
694 rgn->flags);
6a9ceb31
TH
695 } else {
696 /* @rgn is fully contained, record it */
697 if (!*end_rgn)
8c9c1701
AK
698 *start_rgn = idx;
699 *end_rgn = idx + 1;
6a9ceb31
TH
700 }
701 }
702
703 return 0;
704}
6a9ceb31 705
35bd16a2 706static int __init_memblock memblock_remove_range(struct memblock_type *type,
f1af9d3a 707 phys_addr_t base, phys_addr_t size)
95f72d1e 708{
71936180
TH
709 int start_rgn, end_rgn;
710 int i, ret;
95f72d1e 711
71936180
TH
712 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
713 if (ret)
714 return ret;
95f72d1e 715
71936180
TH
716 for (i = end_rgn - 1; i >= start_rgn; i--)
717 memblock_remove_region(type, i);
8f7a6605 718 return 0;
95f72d1e
YL
719}
720
581adcbe 721int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size)
95f72d1e 722{
f1af9d3a 723 return memblock_remove_range(&memblock.memory, base, size);
95f72d1e
YL
724}
725
f1af9d3a 726
581adcbe 727int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size)
95f72d1e 728{
24aa0788 729 memblock_dbg(" memblock_free: [%#016llx-%#016llx] %pF\n",
a150439c 730 (unsigned long long)base,
931d13f5 731 (unsigned long long)base + size - 1,
a150439c 732 (void *)_RET_IP_);
24aa0788 733
aedf95ea 734 kmemleak_free_part(__va(base), size);
f1af9d3a 735 return memblock_remove_range(&memblock.reserved, base, size);
95f72d1e
YL
736}
737
66a20757
TC
738static int __init_memblock memblock_reserve_region(phys_addr_t base,
739 phys_addr_t size,
740 int nid,
741 unsigned long flags)
95f72d1e 742{
7fc825b4 743 struct memblock_type *type = &memblock.reserved;
95f72d1e 744
66a20757 745 memblock_dbg("memblock_reserve: [%#016llx-%#016llx] flags %#02lx %pF\n",
a150439c 746 (unsigned long long)base,
931d13f5 747 (unsigned long long)base + size - 1,
66a20757
TC
748 flags, (void *)_RET_IP_);
749
7fc825b4 750 return memblock_add_range(type, base, size, nid, flags);
66a20757 751}
95f72d1e 752
66a20757
TC
753int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size)
754{
755 return memblock_reserve_region(base, size, MAX_NUMNODES, 0);
95f72d1e
YL
756}
757
66b16edf 758/**
66b16edf 759 *
4308ce17 760 * This function isolates region [@base, @base + @size), and sets/clears flag
66b16edf 761 *
c1153931 762 * Return 0 on success, -errno on failure.
66b16edf 763 */
4308ce17
TL
764static int __init_memblock memblock_setclr_flag(phys_addr_t base,
765 phys_addr_t size, int set, int flag)
66b16edf
TC
766{
767 struct memblock_type *type = &memblock.memory;
768 int i, ret, start_rgn, end_rgn;
769
770 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
771 if (ret)
772 return ret;
773
774 for (i = start_rgn; i < end_rgn; i++)
4308ce17
TL
775 if (set)
776 memblock_set_region_flags(&type->regions[i], flag);
777 else
778 memblock_clear_region_flags(&type->regions[i], flag);
66b16edf
TC
779
780 memblock_merge_regions(type);
781 return 0;
782}
783
784/**
4308ce17 785 * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG.
66b16edf
TC
786 * @base: the base phys addr of the region
787 * @size: the size of the region
788 *
c1153931 789 * Return 0 on success, -errno on failure.
4308ce17
TL
790 */
791int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size)
792{
793 return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG);
794}
795
796/**
797 * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region.
798 * @base: the base phys addr of the region
799 * @size: the size of the region
66b16edf 800 *
c1153931 801 * Return 0 on success, -errno on failure.
66b16edf
TC
802 */
803int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size)
804{
4308ce17 805 return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG);
66b16edf
TC
806}
807
a3f5bafc
TL
808/**
809 * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR.
810 * @base: the base phys addr of the region
811 * @size: the size of the region
812 *
c1153931 813 * Return 0 on success, -errno on failure.
a3f5bafc
TL
814 */
815int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size)
816{
817 system_has_some_mirror = true;
818
819 return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR);
820}
821
bf3d3cc5
AB
822/**
823 * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP.
824 * @base: the base phys addr of the region
825 * @size: the size of the region
826 *
827 * Return 0 on success, -errno on failure.
828 */
829int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size)
830{
831 return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP);
832}
a3f5bafc 833
8e7a7f86
RH
834/**
835 * __next_reserved_mem_region - next function for for_each_reserved_region()
836 * @idx: pointer to u64 loop variable
837 * @out_start: ptr to phys_addr_t for start address of the region, can be %NULL
838 * @out_end: ptr to phys_addr_t for end address of the region, can be %NULL
839 *
840 * Iterate over all reserved memory regions.
841 */
842void __init_memblock __next_reserved_mem_region(u64 *idx,
843 phys_addr_t *out_start,
844 phys_addr_t *out_end)
845{
567d117b 846 struct memblock_type *type = &memblock.reserved;
8e7a7f86 847
567d117b
AK
848 if (*idx >= 0 && *idx < type->cnt) {
849 struct memblock_region *r = &type->regions[*idx];
8e7a7f86
RH
850 phys_addr_t base = r->base;
851 phys_addr_t size = r->size;
852
853 if (out_start)
854 *out_start = base;
855 if (out_end)
856 *out_end = base + size - 1;
857
858 *idx += 1;
859 return;
860 }
861
862 /* signal end of iteration */
863 *idx = ULLONG_MAX;
864}
865
35fd0808 866/**
f1af9d3a 867 * __next__mem_range - next function for for_each_free_mem_range() etc.
35fd0808 868 * @idx: pointer to u64 loop variable
b1154233 869 * @nid: node selector, %NUMA_NO_NODE for all nodes
fc6daaf9 870 * @flags: pick from blocks based on memory attributes
f1af9d3a
PH
871 * @type_a: pointer to memblock_type from where the range is taken
872 * @type_b: pointer to memblock_type which excludes memory from being taken
dad7557e
WL
873 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
874 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
875 * @out_nid: ptr to int for nid of the range, can be %NULL
35fd0808 876 *
f1af9d3a 877 * Find the first area from *@idx which matches @nid, fill the out
35fd0808 878 * parameters, and update *@idx for the next iteration. The lower 32bit of
f1af9d3a
PH
879 * *@idx contains index into type_a and the upper 32bit indexes the
880 * areas before each region in type_b. For example, if type_b regions
35fd0808
TH
881 * look like the following,
882 *
883 * 0:[0-16), 1:[32-48), 2:[128-130)
884 *
885 * The upper 32bit indexes the following regions.
886 *
887 * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX)
888 *
889 * As both region arrays are sorted, the function advances the two indices
890 * in lockstep and returns each intersection.
891 */
fc6daaf9 892void __init_memblock __next_mem_range(u64 *idx, int nid, ulong flags,
f1af9d3a
PH
893 struct memblock_type *type_a,
894 struct memblock_type *type_b,
895 phys_addr_t *out_start,
896 phys_addr_t *out_end, int *out_nid)
35fd0808 897{
f1af9d3a
PH
898 int idx_a = *idx & 0xffffffff;
899 int idx_b = *idx >> 32;
b1154233 900
f1af9d3a
PH
901 if (WARN_ONCE(nid == MAX_NUMNODES,
902 "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
560dca27 903 nid = NUMA_NO_NODE;
35fd0808 904
f1af9d3a
PH
905 for (; idx_a < type_a->cnt; idx_a++) {
906 struct memblock_region *m = &type_a->regions[idx_a];
907
35fd0808
TH
908 phys_addr_t m_start = m->base;
909 phys_addr_t m_end = m->base + m->size;
f1af9d3a 910 int m_nid = memblock_get_region_node(m);
35fd0808
TH
911
912 /* only memory regions are associated with nodes, check it */
f1af9d3a 913 if (nid != NUMA_NO_NODE && nid != m_nid)
35fd0808
TH
914 continue;
915
0a313a99
XQ
916 /* skip hotpluggable memory regions if needed */
917 if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
918 continue;
919
a3f5bafc
TL
920 /* if we want mirror memory skip non-mirror memory regions */
921 if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m))
922 continue;
923
bf3d3cc5
AB
924 /* skip nomap memory unless we were asked for it explicitly */
925 if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m))
926 continue;
927
f1af9d3a
PH
928 if (!type_b) {
929 if (out_start)
930 *out_start = m_start;
931 if (out_end)
932 *out_end = m_end;
933 if (out_nid)
934 *out_nid = m_nid;
935 idx_a++;
936 *idx = (u32)idx_a | (u64)idx_b << 32;
937 return;
938 }
939
940 /* scan areas before each reservation */
941 for (; idx_b < type_b->cnt + 1; idx_b++) {
942 struct memblock_region *r;
943 phys_addr_t r_start;
944 phys_addr_t r_end;
945
946 r = &type_b->regions[idx_b];
947 r_start = idx_b ? r[-1].base + r[-1].size : 0;
948 r_end = idx_b < type_b->cnt ?
949 r->base : ULLONG_MAX;
35fd0808 950
f1af9d3a
PH
951 /*
952 * if idx_b advanced past idx_a,
953 * break out to advance idx_a
954 */
35fd0808
TH
955 if (r_start >= m_end)
956 break;
957 /* if the two regions intersect, we're done */
958 if (m_start < r_end) {
959 if (out_start)
f1af9d3a
PH
960 *out_start =
961 max(m_start, r_start);
35fd0808
TH
962 if (out_end)
963 *out_end = min(m_end, r_end);
964 if (out_nid)
f1af9d3a 965 *out_nid = m_nid;
35fd0808 966 /*
f1af9d3a
PH
967 * The region which ends first is
968 * advanced for the next iteration.
35fd0808
TH
969 */
970 if (m_end <= r_end)
f1af9d3a 971 idx_a++;
35fd0808 972 else
f1af9d3a
PH
973 idx_b++;
974 *idx = (u32)idx_a | (u64)idx_b << 32;
35fd0808
TH
975 return;
976 }
977 }
978 }
979
980 /* signal end of iteration */
981 *idx = ULLONG_MAX;
982}
983
7bd0b0f0 984/**
f1af9d3a
PH
985 * __next_mem_range_rev - generic next function for for_each_*_range_rev()
986 *
987 * Finds the next range from type_a which is not marked as unsuitable
988 * in type_b.
989 *
7bd0b0f0 990 * @idx: pointer to u64 loop variable
ad5ea8cd 991 * @nid: node selector, %NUMA_NO_NODE for all nodes
fc6daaf9 992 * @flags: pick from blocks based on memory attributes
f1af9d3a
PH
993 * @type_a: pointer to memblock_type from where the range is taken
994 * @type_b: pointer to memblock_type which excludes memory from being taken
dad7557e
WL
995 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
996 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
997 * @out_nid: ptr to int for nid of the range, can be %NULL
7bd0b0f0 998 *
f1af9d3a 999 * Reverse of __next_mem_range().
7bd0b0f0 1000 */
fc6daaf9 1001void __init_memblock __next_mem_range_rev(u64 *idx, int nid, ulong flags,
f1af9d3a
PH
1002 struct memblock_type *type_a,
1003 struct memblock_type *type_b,
1004 phys_addr_t *out_start,
1005 phys_addr_t *out_end, int *out_nid)
7bd0b0f0 1006{
f1af9d3a
PH
1007 int idx_a = *idx & 0xffffffff;
1008 int idx_b = *idx >> 32;
b1154233 1009
560dca27
GS
1010 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
1011 nid = NUMA_NO_NODE;
7bd0b0f0
TH
1012
1013 if (*idx == (u64)ULLONG_MAX) {
f1af9d3a
PH
1014 idx_a = type_a->cnt - 1;
1015 idx_b = type_b->cnt;
7bd0b0f0
TH
1016 }
1017
f1af9d3a
PH
1018 for (; idx_a >= 0; idx_a--) {
1019 struct memblock_region *m = &type_a->regions[idx_a];
1020
7bd0b0f0
TH
1021 phys_addr_t m_start = m->base;
1022 phys_addr_t m_end = m->base + m->size;
f1af9d3a 1023 int m_nid = memblock_get_region_node(m);
7bd0b0f0
TH
1024
1025 /* only memory regions are associated with nodes, check it */
f1af9d3a 1026 if (nid != NUMA_NO_NODE && nid != m_nid)
7bd0b0f0
TH
1027 continue;
1028
55ac590c
TC
1029 /* skip hotpluggable memory regions if needed */
1030 if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
1031 continue;
1032
a3f5bafc
TL
1033 /* if we want mirror memory skip non-mirror memory regions */
1034 if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m))
1035 continue;
1036
bf3d3cc5
AB
1037 /* skip nomap memory unless we were asked for it explicitly */
1038 if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m))
1039 continue;
1040
f1af9d3a
PH
1041 if (!type_b) {
1042 if (out_start)
1043 *out_start = m_start;
1044 if (out_end)
1045 *out_end = m_end;
1046 if (out_nid)
1047 *out_nid = m_nid;
1048 idx_a++;
1049 *idx = (u32)idx_a | (u64)idx_b << 32;
1050 return;
1051 }
1052
1053 /* scan areas before each reservation */
1054 for (; idx_b >= 0; idx_b--) {
1055 struct memblock_region *r;
1056 phys_addr_t r_start;
1057 phys_addr_t r_end;
1058
1059 r = &type_b->regions[idx_b];
1060 r_start = idx_b ? r[-1].base + r[-1].size : 0;
1061 r_end = idx_b < type_b->cnt ?
1062 r->base : ULLONG_MAX;
1063 /*
1064 * if idx_b advanced past idx_a,
1065 * break out to advance idx_a
1066 */
7bd0b0f0 1067
7bd0b0f0
TH
1068 if (r_end <= m_start)
1069 break;
1070 /* if the two regions intersect, we're done */
1071 if (m_end > r_start) {
1072 if (out_start)
1073 *out_start = max(m_start, r_start);
1074 if (out_end)
1075 *out_end = min(m_end, r_end);
1076 if (out_nid)
f1af9d3a 1077 *out_nid = m_nid;
7bd0b0f0 1078 if (m_start >= r_start)
f1af9d3a 1079 idx_a--;
7bd0b0f0 1080 else
f1af9d3a
PH
1081 idx_b--;
1082 *idx = (u32)idx_a | (u64)idx_b << 32;
7bd0b0f0
TH
1083 return;
1084 }
1085 }
1086 }
f1af9d3a 1087 /* signal end of iteration */
7bd0b0f0
TH
1088 *idx = ULLONG_MAX;
1089}
1090
7c0caeb8
TH
1091#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1092/*
1093 * Common iterator interface used to define for_each_mem_range().
1094 */
1095void __init_memblock __next_mem_pfn_range(int *idx, int nid,
1096 unsigned long *out_start_pfn,
1097 unsigned long *out_end_pfn, int *out_nid)
1098{
1099 struct memblock_type *type = &memblock.memory;
1100 struct memblock_region *r;
1101
1102 while (++*idx < type->cnt) {
1103 r = &type->regions[*idx];
1104
1105 if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size))
1106 continue;
1107 if (nid == MAX_NUMNODES || nid == r->nid)
1108 break;
1109 }
1110 if (*idx >= type->cnt) {
1111 *idx = -1;
1112 return;
1113 }
1114
1115 if (out_start_pfn)
1116 *out_start_pfn = PFN_UP(r->base);
1117 if (out_end_pfn)
1118 *out_end_pfn = PFN_DOWN(r->base + r->size);
1119 if (out_nid)
1120 *out_nid = r->nid;
1121}
1122
1123/**
1124 * memblock_set_node - set node ID on memblock regions
1125 * @base: base of area to set node ID for
1126 * @size: size of area to set node ID for
e7e8de59 1127 * @type: memblock type to set node ID for
7c0caeb8
TH
1128 * @nid: node ID to set
1129 *
e7e8de59 1130 * Set the nid of memblock @type regions in [@base,@base+@size) to @nid.
7c0caeb8
TH
1131 * Regions which cross the area boundaries are split as necessary.
1132 *
1133 * RETURNS:
1134 * 0 on success, -errno on failure.
1135 */
1136int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size,
e7e8de59 1137 struct memblock_type *type, int nid)
7c0caeb8 1138{
6a9ceb31
TH
1139 int start_rgn, end_rgn;
1140 int i, ret;
7c0caeb8 1141
6a9ceb31
TH
1142 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
1143 if (ret)
1144 return ret;
7c0caeb8 1145
6a9ceb31 1146 for (i = start_rgn; i < end_rgn; i++)
e9d24ad3 1147 memblock_set_region_node(&type->regions[i], nid);
7c0caeb8
TH
1148
1149 memblock_merge_regions(type);
1150 return 0;
1151}
1152#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1153
2bfc2862
AM
1154static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
1155 phys_addr_t align, phys_addr_t start,
fc6daaf9 1156 phys_addr_t end, int nid, ulong flags)
95f72d1e 1157{
6ed311b2 1158 phys_addr_t found;
95f72d1e 1159
79f40fab
GS
1160 if (!align)
1161 align = SMP_CACHE_BYTES;
94f3d3af 1162
fc6daaf9
TL
1163 found = memblock_find_in_range_node(size, align, start, end, nid,
1164 flags);
aedf95ea
CM
1165 if (found && !memblock_reserve(found, size)) {
1166 /*
1167 * The min_count is set to 0 so that memblock allocations are
1168 * never reported as leaks.
1169 */
1170 kmemleak_alloc(__va(found), size, 0, 0);
6ed311b2 1171 return found;
aedf95ea 1172 }
6ed311b2 1173 return 0;
95f72d1e
YL
1174}
1175
2bfc2862 1176phys_addr_t __init memblock_alloc_range(phys_addr_t size, phys_addr_t align,
fc6daaf9
TL
1177 phys_addr_t start, phys_addr_t end,
1178 ulong flags)
2bfc2862 1179{
fc6daaf9
TL
1180 return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE,
1181 flags);
2bfc2862
AM
1182}
1183
1184static phys_addr_t __init memblock_alloc_base_nid(phys_addr_t size,
1185 phys_addr_t align, phys_addr_t max_addr,
fc6daaf9 1186 int nid, ulong flags)
2bfc2862 1187{
fc6daaf9 1188 return memblock_alloc_range_nid(size, align, 0, max_addr, nid, flags);
2bfc2862
AM
1189}
1190
7bd0b0f0
TH
1191phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid)
1192{
a3f5bafc
TL
1193 ulong flags = choose_memblock_flags();
1194 phys_addr_t ret;
1195
1196again:
1197 ret = memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE,
1198 nid, flags);
1199
1200 if (!ret && (flags & MEMBLOCK_MIRROR)) {
1201 flags &= ~MEMBLOCK_MIRROR;
1202 goto again;
1203 }
1204 return ret;
7bd0b0f0
TH
1205}
1206
1207phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
1208{
fc6daaf9
TL
1209 return memblock_alloc_base_nid(size, align, max_addr, NUMA_NO_NODE,
1210 MEMBLOCK_NONE);
7bd0b0f0
TH
1211}
1212
6ed311b2 1213phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
95f72d1e 1214{
6ed311b2
BH
1215 phys_addr_t alloc;
1216
1217 alloc = __memblock_alloc_base(size, align, max_addr);
1218
1219 if (alloc == 0)
1220 panic("ERROR: Failed to allocate 0x%llx bytes below 0x%llx.\n",
1221 (unsigned long long) size, (unsigned long long) max_addr);
1222
1223 return alloc;
95f72d1e
YL
1224}
1225
6ed311b2 1226phys_addr_t __init memblock_alloc(phys_addr_t size, phys_addr_t align)
95f72d1e 1227{
6ed311b2
BH
1228 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
1229}
95f72d1e 1230
9d1e2492
BH
1231phys_addr_t __init memblock_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid)
1232{
1233 phys_addr_t res = memblock_alloc_nid(size, align, nid);
1234
1235 if (res)
1236 return res;
15fb0972 1237 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
95f72d1e
YL
1238}
1239
26f09e9b
SS
1240/**
1241 * memblock_virt_alloc_internal - allocate boot memory block
1242 * @size: size of memory block to be allocated in bytes
1243 * @align: alignment of the region and block's size
1244 * @min_addr: the lower bound of the memory region to allocate (phys address)
1245 * @max_addr: the upper bound of the memory region to allocate (phys address)
1246 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1247 *
1248 * The @min_addr limit is dropped if it can not be satisfied and the allocation
1249 * will fall back to memory below @min_addr. Also, allocation may fall back
1250 * to any node in the system if the specified node can not
1251 * hold the requested memory.
1252 *
1253 * The allocation is performed from memory region limited by
1254 * memblock.current_limit if @max_addr == %BOOTMEM_ALLOC_ACCESSIBLE.
1255 *
1256 * The memory block is aligned on SMP_CACHE_BYTES if @align == 0.
1257 *
1258 * The phys address of allocated boot memory block is converted to virtual and
1259 * allocated memory is reset to 0.
1260 *
1261 * In addition, function sets the min_count to 0 using kmemleak_alloc for
1262 * allocated boot memory block, so that it is never reported as leaks.
1263 *
1264 * RETURNS:
1265 * Virtual address of allocated memory block on success, NULL on failure.
1266 */
1267static void * __init memblock_virt_alloc_internal(
1268 phys_addr_t size, phys_addr_t align,
1269 phys_addr_t min_addr, phys_addr_t max_addr,
1270 int nid)
1271{
1272 phys_addr_t alloc;
1273 void *ptr;
a3f5bafc 1274 ulong flags = choose_memblock_flags();
26f09e9b 1275
560dca27
GS
1276 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
1277 nid = NUMA_NO_NODE;
26f09e9b
SS
1278
1279 /*
1280 * Detect any accidental use of these APIs after slab is ready, as at
1281 * this moment memblock may be deinitialized already and its
1282 * internal data may be destroyed (after execution of free_all_bootmem)
1283 */
1284 if (WARN_ON_ONCE(slab_is_available()))
1285 return kzalloc_node(size, GFP_NOWAIT, nid);
1286
1287 if (!align)
1288 align = SMP_CACHE_BYTES;
1289
f544e14f
YL
1290 if (max_addr > memblock.current_limit)
1291 max_addr = memblock.current_limit;
1292
26f09e9b
SS
1293again:
1294 alloc = memblock_find_in_range_node(size, align, min_addr, max_addr,
a3f5bafc 1295 nid, flags);
26f09e9b
SS
1296 if (alloc)
1297 goto done;
1298
1299 if (nid != NUMA_NO_NODE) {
1300 alloc = memblock_find_in_range_node(size, align, min_addr,
fc6daaf9 1301 max_addr, NUMA_NO_NODE,
a3f5bafc 1302 flags);
26f09e9b
SS
1303 if (alloc)
1304 goto done;
1305 }
1306
1307 if (min_addr) {
1308 min_addr = 0;
1309 goto again;
26f09e9b
SS
1310 }
1311
a3f5bafc
TL
1312 if (flags & MEMBLOCK_MIRROR) {
1313 flags &= ~MEMBLOCK_MIRROR;
1314 pr_warn("Could not allocate %pap bytes of mirrored memory\n",
1315 &size);
1316 goto again;
1317 }
1318
1319 return NULL;
26f09e9b
SS
1320done:
1321 memblock_reserve(alloc, size);
1322 ptr = phys_to_virt(alloc);
1323 memset(ptr, 0, size);
1324
1325 /*
1326 * The min_count is set to 0 so that bootmem allocated blocks
1327 * are never reported as leaks. This is because many of these blocks
1328 * are only referred via the physical address which is not
1329 * looked up by kmemleak.
1330 */
1331 kmemleak_alloc(ptr, size, 0, 0);
1332
1333 return ptr;
26f09e9b
SS
1334}
1335
1336/**
1337 * memblock_virt_alloc_try_nid_nopanic - allocate boot memory block
1338 * @size: size of memory block to be allocated in bytes
1339 * @align: alignment of the region and block's size
1340 * @min_addr: the lower bound of the memory region from where the allocation
1341 * is preferred (phys address)
1342 * @max_addr: the upper bound of the memory region from where the allocation
1343 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1344 * allocate only from memory limited by memblock.current_limit value
1345 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1346 *
1347 * Public version of _memblock_virt_alloc_try_nid_nopanic() which provides
1348 * additional debug information (including caller info), if enabled.
1349 *
1350 * RETURNS:
1351 * Virtual address of allocated memory block on success, NULL on failure.
1352 */
1353void * __init memblock_virt_alloc_try_nid_nopanic(
1354 phys_addr_t size, phys_addr_t align,
1355 phys_addr_t min_addr, phys_addr_t max_addr,
1356 int nid)
1357{
1358 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1359 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1360 (u64)max_addr, (void *)_RET_IP_);
1361 return memblock_virt_alloc_internal(size, align, min_addr,
1362 max_addr, nid);
1363}
1364
1365/**
1366 * memblock_virt_alloc_try_nid - allocate boot memory block with panicking
1367 * @size: size of memory block to be allocated in bytes
1368 * @align: alignment of the region and block's size
1369 * @min_addr: the lower bound of the memory region from where the allocation
1370 * is preferred (phys address)
1371 * @max_addr: the upper bound of the memory region from where the allocation
1372 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1373 * allocate only from memory limited by memblock.current_limit value
1374 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1375 *
1376 * Public panicking version of _memblock_virt_alloc_try_nid_nopanic()
1377 * which provides debug information (including caller info), if enabled,
1378 * and panics if the request can not be satisfied.
1379 *
1380 * RETURNS:
1381 * Virtual address of allocated memory block on success, NULL on failure.
1382 */
1383void * __init memblock_virt_alloc_try_nid(
1384 phys_addr_t size, phys_addr_t align,
1385 phys_addr_t min_addr, phys_addr_t max_addr,
1386 int nid)
1387{
1388 void *ptr;
1389
1390 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1391 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1392 (u64)max_addr, (void *)_RET_IP_);
1393 ptr = memblock_virt_alloc_internal(size, align,
1394 min_addr, max_addr, nid);
1395 if (ptr)
1396 return ptr;
1397
1398 panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx\n",
1399 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1400 (u64)max_addr);
1401 return NULL;
1402}
1403
1404/**
1405 * __memblock_free_early - free boot memory block
1406 * @base: phys starting address of the boot memory block
1407 * @size: size of the boot memory block in bytes
1408 *
1409 * Free boot memory block previously allocated by memblock_virt_alloc_xx() API.
1410 * The freeing memory will not be released to the buddy allocator.
1411 */
1412void __init __memblock_free_early(phys_addr_t base, phys_addr_t size)
1413{
1414 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1415 __func__, (u64)base, (u64)base + size - 1,
1416 (void *)_RET_IP_);
1417 kmemleak_free_part(__va(base), size);
f1af9d3a 1418 memblock_remove_range(&memblock.reserved, base, size);
26f09e9b
SS
1419}
1420
1421/*
1422 * __memblock_free_late - free bootmem block pages directly to buddy allocator
1423 * @addr: phys starting address of the boot memory block
1424 * @size: size of the boot memory block in bytes
1425 *
1426 * This is only useful when the bootmem allocator has already been torn
1427 * down, but we are still initializing the system. Pages are released directly
1428 * to the buddy allocator, no bootmem metadata is updated because it is gone.
1429 */
1430void __init __memblock_free_late(phys_addr_t base, phys_addr_t size)
1431{
1432 u64 cursor, end;
1433
1434 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1435 __func__, (u64)base, (u64)base + size - 1,
1436 (void *)_RET_IP_);
1437 kmemleak_free_part(__va(base), size);
1438 cursor = PFN_UP(base);
1439 end = PFN_DOWN(base + size);
1440
1441 for (; cursor < end; cursor++) {
d70ddd7a 1442 __free_pages_bootmem(pfn_to_page(cursor), cursor, 0);
26f09e9b
SS
1443 totalram_pages++;
1444 }
1445}
9d1e2492
BH
1446
1447/*
1448 * Remaining API functions
1449 */
1450
1f1ffb8a 1451phys_addr_t __init_memblock memblock_phys_mem_size(void)
95f72d1e 1452{
1440c4e2 1453 return memblock.memory.total_size;
95f72d1e
YL
1454}
1455
595ad9af
YL
1456phys_addr_t __init memblock_mem_size(unsigned long limit_pfn)
1457{
1458 unsigned long pages = 0;
1459 struct memblock_region *r;
1460 unsigned long start_pfn, end_pfn;
1461
1462 for_each_memblock(memory, r) {
1463 start_pfn = memblock_region_memory_base_pfn(r);
1464 end_pfn = memblock_region_memory_end_pfn(r);
1465 start_pfn = min_t(unsigned long, start_pfn, limit_pfn);
1466 end_pfn = min_t(unsigned long, end_pfn, limit_pfn);
1467 pages += end_pfn - start_pfn;
1468 }
1469
16763230 1470 return PFN_PHYS(pages);
595ad9af
YL
1471}
1472
0a93ebef
SR
1473/* lowest address */
1474phys_addr_t __init_memblock memblock_start_of_DRAM(void)
1475{
1476 return memblock.memory.regions[0].base;
1477}
1478
10d06439 1479phys_addr_t __init_memblock memblock_end_of_DRAM(void)
95f72d1e
YL
1480{
1481 int idx = memblock.memory.cnt - 1;
1482
e3239ff9 1483 return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size);
95f72d1e
YL
1484}
1485
c0ce8fef 1486void __init memblock_enforce_memory_limit(phys_addr_t limit)
95f72d1e 1487{
c0ce8fef 1488 phys_addr_t max_addr = (phys_addr_t)ULLONG_MAX;
136199f0 1489 struct memblock_region *r;
95f72d1e 1490
c0ce8fef 1491 if (!limit)
95f72d1e
YL
1492 return;
1493
c0ce8fef 1494 /* find out max address */
136199f0 1495 for_each_memblock(memory, r) {
c0ce8fef
TH
1496 if (limit <= r->size) {
1497 max_addr = r->base + limit;
1498 break;
95f72d1e 1499 }
c0ce8fef 1500 limit -= r->size;
95f72d1e 1501 }
c0ce8fef
TH
1502
1503 /* truncate both memory and reserved regions */
f1af9d3a
PH
1504 memblock_remove_range(&memblock.memory, max_addr,
1505 (phys_addr_t)ULLONG_MAX);
1506 memblock_remove_range(&memblock.reserved, max_addr,
1507 (phys_addr_t)ULLONG_MAX);
95f72d1e
YL
1508}
1509
cd79481d 1510static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr)
72d4b0b4
BH
1511{
1512 unsigned int left = 0, right = type->cnt;
1513
1514 do {
1515 unsigned int mid = (right + left) / 2;
1516
1517 if (addr < type->regions[mid].base)
1518 right = mid;
1519 else if (addr >= (type->regions[mid].base +
1520 type->regions[mid].size))
1521 left = mid + 1;
1522 else
1523 return mid;
1524 } while (left < right);
1525 return -1;
1526}
1527
b4ad0c7e 1528bool __init memblock_is_reserved(phys_addr_t addr)
95f72d1e 1529{
72d4b0b4
BH
1530 return memblock_search(&memblock.reserved, addr) != -1;
1531}
95f72d1e 1532
b4ad0c7e 1533bool __init_memblock memblock_is_memory(phys_addr_t addr)
72d4b0b4
BH
1534{
1535 return memblock_search(&memblock.memory, addr) != -1;
1536}
1537
bf3d3cc5
AB
1538int __init_memblock memblock_is_map_memory(phys_addr_t addr)
1539{
1540 int i = memblock_search(&memblock.memory, addr);
1541
1542 if (i == -1)
1543 return false;
1544 return !memblock_is_nomap(&memblock.memory.regions[i]);
1545}
1546
e76b63f8
YL
1547#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1548int __init_memblock memblock_search_pfn_nid(unsigned long pfn,
1549 unsigned long *start_pfn, unsigned long *end_pfn)
1550{
1551 struct memblock_type *type = &memblock.memory;
16763230 1552 int mid = memblock_search(type, PFN_PHYS(pfn));
e76b63f8
YL
1553
1554 if (mid == -1)
1555 return -1;
1556
f7e2f7e8
FF
1557 *start_pfn = PFN_DOWN(type->regions[mid].base);
1558 *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size);
e76b63f8
YL
1559
1560 return type->regions[mid].nid;
1561}
1562#endif
1563
eab30949
SB
1564/**
1565 * memblock_is_region_memory - check if a region is a subset of memory
1566 * @base: base of region to check
1567 * @size: size of region to check
1568 *
1569 * Check if the region [@base, @base+@size) is a subset of a memory block.
1570 *
1571 * RETURNS:
1572 * 0 if false, non-zero if true
1573 */
3661ca66 1574int __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size)
72d4b0b4 1575{
abb65272 1576 int idx = memblock_search(&memblock.memory, base);
eb18f1b5 1577 phys_addr_t end = base + memblock_cap_size(base, &size);
72d4b0b4
BH
1578
1579 if (idx == -1)
1580 return 0;
abb65272
TV
1581 return memblock.memory.regions[idx].base <= base &&
1582 (memblock.memory.regions[idx].base +
eb18f1b5 1583 memblock.memory.regions[idx].size) >= end;
95f72d1e
YL
1584}
1585
eab30949
SB
1586/**
1587 * memblock_is_region_reserved - check if a region intersects reserved memory
1588 * @base: base of region to check
1589 * @size: size of region to check
1590 *
1591 * Check if the region [@base, @base+@size) intersects a reserved memory block.
1592 *
1593 * RETURNS:
c5c5c9d1 1594 * True if they intersect, false if not.
eab30949 1595 */
c5c5c9d1 1596bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size)
95f72d1e 1597{
eb18f1b5 1598 memblock_cap_size(base, &size);
c5c5c9d1 1599 return memblock_overlaps_region(&memblock.reserved, base, size);
95f72d1e
YL
1600}
1601
6ede1fd3
YL
1602void __init_memblock memblock_trim_memory(phys_addr_t align)
1603{
6ede1fd3 1604 phys_addr_t start, end, orig_start, orig_end;
136199f0 1605 struct memblock_region *r;
6ede1fd3 1606
136199f0
EM
1607 for_each_memblock(memory, r) {
1608 orig_start = r->base;
1609 orig_end = r->base + r->size;
6ede1fd3
YL
1610 start = round_up(orig_start, align);
1611 end = round_down(orig_end, align);
1612
1613 if (start == orig_start && end == orig_end)
1614 continue;
1615
1616 if (start < end) {
136199f0
EM
1617 r->base = start;
1618 r->size = end - start;
6ede1fd3 1619 } else {
136199f0
EM
1620 memblock_remove_region(&memblock.memory,
1621 r - memblock.memory.regions);
1622 r--;
6ede1fd3
YL
1623 }
1624 }
1625}
e63075a3 1626
3661ca66 1627void __init_memblock memblock_set_current_limit(phys_addr_t limit)
e63075a3
BH
1628{
1629 memblock.current_limit = limit;
1630}
1631
fec51014
LA
1632phys_addr_t __init_memblock memblock_get_current_limit(void)
1633{
1634 return memblock.current_limit;
1635}
1636
7c0caeb8 1637static void __init_memblock memblock_dump(struct memblock_type *type, char *name)
6ed311b2
BH
1638{
1639 unsigned long long base, size;
66a20757 1640 unsigned long flags;
8c9c1701
AK
1641 int idx;
1642 struct memblock_region *rgn;
6ed311b2 1643
7c0caeb8 1644 pr_info(" %s.cnt = 0x%lx\n", name, type->cnt);
6ed311b2 1645
8c9c1701 1646 for_each_memblock_type(type, rgn) {
7c0caeb8
TH
1647 char nid_buf[32] = "";
1648
1649 base = rgn->base;
1650 size = rgn->size;
66a20757 1651 flags = rgn->flags;
7c0caeb8
TH
1652#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1653 if (memblock_get_region_node(rgn) != MAX_NUMNODES)
1654 snprintf(nid_buf, sizeof(nid_buf), " on node %d",
1655 memblock_get_region_node(rgn));
1656#endif
66a20757 1657 pr_info(" %s[%#x]\t[%#016llx-%#016llx], %#llx bytes%s flags: %#lx\n",
8c9c1701 1658 name, idx, base, base + size - 1, size, nid_buf, flags);
6ed311b2
BH
1659 }
1660}
1661
4ff7b82f 1662void __init_memblock __memblock_dump_all(void)
6ed311b2 1663{
6ed311b2 1664 pr_info("MEMBLOCK configuration:\n");
1440c4e2
TH
1665 pr_info(" memory size = %#llx reserved size = %#llx\n",
1666 (unsigned long long)memblock.memory.total_size,
1667 (unsigned long long)memblock.reserved.total_size);
6ed311b2
BH
1668
1669 memblock_dump(&memblock.memory, "memory");
1670 memblock_dump(&memblock.reserved, "reserved");
1671}
1672
1aadc056 1673void __init memblock_allow_resize(void)
6ed311b2 1674{
142b45a7 1675 memblock_can_resize = 1;
6ed311b2
BH
1676}
1677
6ed311b2
BH
1678static int __init early_memblock(char *p)
1679{
1680 if (p && strstr(p, "debug"))
1681 memblock_debug = 1;
1682 return 0;
1683}
1684early_param("memblock", early_memblock);
1685
c378ddd5 1686#if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK)
6d03b885
BH
1687
1688static int memblock_debug_show(struct seq_file *m, void *private)
1689{
1690 struct memblock_type *type = m->private;
1691 struct memblock_region *reg;
1692 int i;
1693
1694 for (i = 0; i < type->cnt; i++) {
1695 reg = &type->regions[i];
1696 seq_printf(m, "%4d: ", i);
1697 if (sizeof(phys_addr_t) == 4)
1698 seq_printf(m, "0x%08lx..0x%08lx\n",
1699 (unsigned long)reg->base,
1700 (unsigned long)(reg->base + reg->size - 1));
1701 else
1702 seq_printf(m, "0x%016llx..0x%016llx\n",
1703 (unsigned long long)reg->base,
1704 (unsigned long long)(reg->base + reg->size - 1));
1705
1706 }
1707 return 0;
1708}
1709
1710static int memblock_debug_open(struct inode *inode, struct file *file)
1711{
1712 return single_open(file, memblock_debug_show, inode->i_private);
1713}
1714
1715static const struct file_operations memblock_debug_fops = {
1716 .open = memblock_debug_open,
1717 .read = seq_read,
1718 .llseek = seq_lseek,
1719 .release = single_release,
1720};
1721
1722static int __init memblock_init_debugfs(void)
1723{
1724 struct dentry *root = debugfs_create_dir("memblock", NULL);
1725 if (!root)
1726 return -ENXIO;
1727 debugfs_create_file("memory", S_IRUGO, root, &memblock.memory, &memblock_debug_fops);
1728 debugfs_create_file("reserved", S_IRUGO, root, &memblock.reserved, &memblock_debug_fops);
70210ed9
PH
1729#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
1730 debugfs_create_file("physmem", S_IRUGO, root, &memblock.physmem, &memblock_debug_fops);
1731#endif
6d03b885
BH
1732
1733 return 0;
1734}
1735__initcall(memblock_init_debugfs);
1736
1737#endif /* CONFIG_DEBUG_FS */