mm/memblock: add extra "flags" to memblock to allow selection of memory based on...
[deliverable/linux.git] / mm / memblock.c
CommitLineData
95f72d1e
YL
1/*
2 * Procedures for maintaining information about logical memory blocks.
3 *
4 * Peter Bergner, IBM Corp. June 2001.
5 * Copyright (C) 2001 Peter Bergner.
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 */
12
13#include <linux/kernel.h>
142b45a7 14#include <linux/slab.h>
95f72d1e
YL
15#include <linux/init.h>
16#include <linux/bitops.h>
449e8df3 17#include <linux/poison.h>
c196f76f 18#include <linux/pfn.h>
6d03b885
BH
19#include <linux/debugfs.h>
20#include <linux/seq_file.h>
95f72d1e
YL
21#include <linux/memblock.h>
22
79442ed1 23#include <asm-generic/sections.h>
26f09e9b
SS
24#include <linux/io.h>
25
26#include "internal.h"
79442ed1 27
fe091c20
TH
28static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock;
29static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock;
70210ed9
PH
30#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
31static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock;
32#endif
fe091c20
TH
33
34struct memblock memblock __initdata_memblock = {
35 .memory.regions = memblock_memory_init_regions,
36 .memory.cnt = 1, /* empty dummy entry */
37 .memory.max = INIT_MEMBLOCK_REGIONS,
38
39 .reserved.regions = memblock_reserved_init_regions,
40 .reserved.cnt = 1, /* empty dummy entry */
41 .reserved.max = INIT_MEMBLOCK_REGIONS,
42
70210ed9
PH
43#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
44 .physmem.regions = memblock_physmem_init_regions,
45 .physmem.cnt = 1, /* empty dummy entry */
46 .physmem.max = INIT_PHYSMEM_REGIONS,
47#endif
48
79442ed1 49 .bottom_up = false,
fe091c20
TH
50 .current_limit = MEMBLOCK_ALLOC_ANYWHERE,
51};
95f72d1e 52
10d06439 53int memblock_debug __initdata_memblock;
55ac590c
TC
54#ifdef CONFIG_MOVABLE_NODE
55bool movable_node_enabled __initdata_memblock = false;
56#endif
1aadc056 57static int memblock_can_resize __initdata_memblock;
181eb394
GS
58static int memblock_memory_in_slab __initdata_memblock = 0;
59static int memblock_reserved_in_slab __initdata_memblock = 0;
95f72d1e 60
142b45a7 61/* inline so we don't get a warning when pr_debug is compiled out */
c2233116
RP
62static __init_memblock const char *
63memblock_type_name(struct memblock_type *type)
142b45a7
BH
64{
65 if (type == &memblock.memory)
66 return "memory";
67 else if (type == &memblock.reserved)
68 return "reserved";
69 else
70 return "unknown";
71}
72
eb18f1b5
TH
73/* adjust *@size so that (@base + *@size) doesn't overflow, return new size */
74static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size)
75{
76 return *size = min(*size, (phys_addr_t)ULLONG_MAX - base);
77}
78
6ed311b2
BH
79/*
80 * Address comparison utilities
81 */
10d06439 82static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1,
2898cc4c 83 phys_addr_t base2, phys_addr_t size2)
95f72d1e
YL
84{
85 return ((base1 < (base2 + size2)) && (base2 < (base1 + size1)));
86}
87
2d7d3eb2
HS
88static long __init_memblock memblock_overlaps_region(struct memblock_type *type,
89 phys_addr_t base, phys_addr_t size)
6ed311b2
BH
90{
91 unsigned long i;
92
93 for (i = 0; i < type->cnt; i++) {
94 phys_addr_t rgnbase = type->regions[i].base;
95 phys_addr_t rgnsize = type->regions[i].size;
96 if (memblock_addrs_overlap(base, size, rgnbase, rgnsize))
97 break;
98 }
99
100 return (i < type->cnt) ? i : -1;
101}
102
79442ed1
TC
103/*
104 * __memblock_find_range_bottom_up - find free area utility in bottom-up
105 * @start: start of candidate range
106 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
107 * @size: size of free area to find
108 * @align: alignment of free area to find
b1154233 109 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
fc6daaf9 110 * @flags: pick from blocks based on memory attributes
79442ed1
TC
111 *
112 * Utility called from memblock_find_in_range_node(), find free area bottom-up.
113 *
114 * RETURNS:
115 * Found address on success, 0 on failure.
116 */
117static phys_addr_t __init_memblock
118__memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
fc6daaf9
TL
119 phys_addr_t size, phys_addr_t align, int nid,
120 ulong flags)
79442ed1
TC
121{
122 phys_addr_t this_start, this_end, cand;
123 u64 i;
124
fc6daaf9 125 for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) {
79442ed1
TC
126 this_start = clamp(this_start, start, end);
127 this_end = clamp(this_end, start, end);
128
129 cand = round_up(this_start, align);
130 if (cand < this_end && this_end - cand >= size)
131 return cand;
132 }
133
134 return 0;
135}
136
7bd0b0f0 137/**
1402899e 138 * __memblock_find_range_top_down - find free area utility, in top-down
7bd0b0f0
TH
139 * @start: start of candidate range
140 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
141 * @size: size of free area to find
142 * @align: alignment of free area to find
b1154233 143 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
fc6daaf9 144 * @flags: pick from blocks based on memory attributes
7bd0b0f0 145 *
1402899e 146 * Utility called from memblock_find_in_range_node(), find free area top-down.
7bd0b0f0
TH
147 *
148 * RETURNS:
79442ed1 149 * Found address on success, 0 on failure.
6ed311b2 150 */
1402899e
TC
151static phys_addr_t __init_memblock
152__memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
fc6daaf9
TL
153 phys_addr_t size, phys_addr_t align, int nid,
154 ulong flags)
f7210e6c
TC
155{
156 phys_addr_t this_start, this_end, cand;
157 u64 i;
158
fc6daaf9
TL
159 for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end,
160 NULL) {
f7210e6c
TC
161 this_start = clamp(this_start, start, end);
162 this_end = clamp(this_end, start, end);
163
164 if (this_end < size)
165 continue;
166
167 cand = round_down(this_end - size, align);
168 if (cand >= this_start)
169 return cand;
170 }
1402899e 171
f7210e6c
TC
172 return 0;
173}
6ed311b2 174
1402899e
TC
175/**
176 * memblock_find_in_range_node - find free area in given range and node
1402899e
TC
177 * @size: size of free area to find
178 * @align: alignment of free area to find
87029ee9
GS
179 * @start: start of candidate range
180 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
b1154233 181 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
fc6daaf9 182 * @flags: pick from blocks based on memory attributes
1402899e
TC
183 *
184 * Find @size free area aligned to @align in the specified range and node.
185 *
79442ed1
TC
186 * When allocation direction is bottom-up, the @start should be greater
187 * than the end of the kernel image. Otherwise, it will be trimmed. The
188 * reason is that we want the bottom-up allocation just near the kernel
189 * image so it is highly likely that the allocated memory and the kernel
190 * will reside in the same node.
191 *
192 * If bottom-up allocation failed, will try to allocate memory top-down.
193 *
1402899e 194 * RETURNS:
79442ed1 195 * Found address on success, 0 on failure.
1402899e 196 */
87029ee9
GS
197phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
198 phys_addr_t align, phys_addr_t start,
fc6daaf9 199 phys_addr_t end, int nid, ulong flags)
1402899e 200{
0cfb8f0c 201 phys_addr_t kernel_end, ret;
79442ed1 202
1402899e
TC
203 /* pump up @end */
204 if (end == MEMBLOCK_ALLOC_ACCESSIBLE)
205 end = memblock.current_limit;
206
207 /* avoid allocating the first page */
208 start = max_t(phys_addr_t, start, PAGE_SIZE);
209 end = max(start, end);
79442ed1
TC
210 kernel_end = __pa_symbol(_end);
211
212 /*
213 * try bottom-up allocation only when bottom-up mode
214 * is set and @end is above the kernel image.
215 */
216 if (memblock_bottom_up() && end > kernel_end) {
217 phys_addr_t bottom_up_start;
218
219 /* make sure we will allocate above the kernel */
220 bottom_up_start = max(start, kernel_end);
221
222 /* ok, try bottom-up allocation first */
223 ret = __memblock_find_range_bottom_up(bottom_up_start, end,
fc6daaf9 224 size, align, nid, flags);
79442ed1
TC
225 if (ret)
226 return ret;
227
228 /*
229 * we always limit bottom-up allocation above the kernel,
230 * but top-down allocation doesn't have the limit, so
231 * retrying top-down allocation may succeed when bottom-up
232 * allocation failed.
233 *
234 * bottom-up allocation is expected to be fail very rarely,
235 * so we use WARN_ONCE() here to see the stack trace if
236 * fail happens.
237 */
238 WARN_ONCE(1, "memblock: bottom-up allocation failed, "
239 "memory hotunplug may be affected\n");
240 }
1402899e 241
fc6daaf9
TL
242 return __memblock_find_range_top_down(start, end, size, align, nid,
243 flags);
1402899e
TC
244}
245
7bd0b0f0
TH
246/**
247 * memblock_find_in_range - find free area in given range
248 * @start: start of candidate range
249 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
250 * @size: size of free area to find
251 * @align: alignment of free area to find
252 *
253 * Find @size free area aligned to @align in the specified range.
254 *
255 * RETURNS:
79442ed1 256 * Found address on success, 0 on failure.
fc769a8e 257 */
7bd0b0f0
TH
258phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start,
259 phys_addr_t end, phys_addr_t size,
260 phys_addr_t align)
6ed311b2 261{
87029ee9 262 return memblock_find_in_range_node(size, align, start, end,
fc6daaf9 263 NUMA_NO_NODE, MEMBLOCK_NONE);
6ed311b2
BH
264}
265
10d06439 266static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r)
95f72d1e 267{
1440c4e2 268 type->total_size -= type->regions[r].size;
7c0caeb8
TH
269 memmove(&type->regions[r], &type->regions[r + 1],
270 (type->cnt - (r + 1)) * sizeof(type->regions[r]));
e3239ff9 271 type->cnt--;
95f72d1e 272
8f7a6605
BH
273 /* Special case for empty arrays */
274 if (type->cnt == 0) {
1440c4e2 275 WARN_ON(type->total_size != 0);
8f7a6605
BH
276 type->cnt = 1;
277 type->regions[0].base = 0;
278 type->regions[0].size = 0;
66a20757 279 type->regions[0].flags = 0;
7c0caeb8 280 memblock_set_region_node(&type->regions[0], MAX_NUMNODES);
8f7a6605 281 }
95f72d1e
YL
282}
283
354f17e1
PH
284#ifdef CONFIG_ARCH_DISCARD_MEMBLOCK
285
29f67386
YL
286phys_addr_t __init_memblock get_allocated_memblock_reserved_regions_info(
287 phys_addr_t *addr)
288{
289 if (memblock.reserved.regions == memblock_reserved_init_regions)
290 return 0;
291
292 *addr = __pa(memblock.reserved.regions);
293
294 return PAGE_ALIGN(sizeof(struct memblock_region) *
295 memblock.reserved.max);
296}
297
5e270e25
PH
298phys_addr_t __init_memblock get_allocated_memblock_memory_regions_info(
299 phys_addr_t *addr)
300{
301 if (memblock.memory.regions == memblock_memory_init_regions)
302 return 0;
303
304 *addr = __pa(memblock.memory.regions);
305
306 return PAGE_ALIGN(sizeof(struct memblock_region) *
307 memblock.memory.max);
308}
309
310#endif
311
48c3b583
GP
312/**
313 * memblock_double_array - double the size of the memblock regions array
314 * @type: memblock type of the regions array being doubled
315 * @new_area_start: starting address of memory range to avoid overlap with
316 * @new_area_size: size of memory range to avoid overlap with
317 *
318 * Double the size of the @type regions array. If memblock is being used to
319 * allocate memory for a new reserved regions array and there is a previously
320 * allocated memory range [@new_area_start,@new_area_start+@new_area_size]
321 * waiting to be reserved, ensure the memory used by the new array does
322 * not overlap.
323 *
324 * RETURNS:
325 * 0 on success, -1 on failure.
326 */
327static int __init_memblock memblock_double_array(struct memblock_type *type,
328 phys_addr_t new_area_start,
329 phys_addr_t new_area_size)
142b45a7
BH
330{
331 struct memblock_region *new_array, *old_array;
29f67386 332 phys_addr_t old_alloc_size, new_alloc_size;
142b45a7
BH
333 phys_addr_t old_size, new_size, addr;
334 int use_slab = slab_is_available();
181eb394 335 int *in_slab;
142b45a7
BH
336
337 /* We don't allow resizing until we know about the reserved regions
338 * of memory that aren't suitable for allocation
339 */
340 if (!memblock_can_resize)
341 return -1;
342
142b45a7
BH
343 /* Calculate new doubled size */
344 old_size = type->max * sizeof(struct memblock_region);
345 new_size = old_size << 1;
29f67386
YL
346 /*
347 * We need to allocated new one align to PAGE_SIZE,
348 * so we can free them completely later.
349 */
350 old_alloc_size = PAGE_ALIGN(old_size);
351 new_alloc_size = PAGE_ALIGN(new_size);
142b45a7 352
181eb394
GS
353 /* Retrieve the slab flag */
354 if (type == &memblock.memory)
355 in_slab = &memblock_memory_in_slab;
356 else
357 in_slab = &memblock_reserved_in_slab;
358
142b45a7
BH
359 /* Try to find some space for it.
360 *
361 * WARNING: We assume that either slab_is_available() and we use it or
fd07383b
AM
362 * we use MEMBLOCK for allocations. That means that this is unsafe to
363 * use when bootmem is currently active (unless bootmem itself is
364 * implemented on top of MEMBLOCK which isn't the case yet)
142b45a7
BH
365 *
366 * This should however not be an issue for now, as we currently only
fd07383b
AM
367 * call into MEMBLOCK while it's still active, or much later when slab
368 * is active for memory hotplug operations
142b45a7
BH
369 */
370 if (use_slab) {
371 new_array = kmalloc(new_size, GFP_KERNEL);
1f5026a7 372 addr = new_array ? __pa(new_array) : 0;
4e2f0775 373 } else {
48c3b583
GP
374 /* only exclude range when trying to double reserved.regions */
375 if (type != &memblock.reserved)
376 new_area_start = new_area_size = 0;
377
378 addr = memblock_find_in_range(new_area_start + new_area_size,
379 memblock.current_limit,
29f67386 380 new_alloc_size, PAGE_SIZE);
48c3b583
GP
381 if (!addr && new_area_size)
382 addr = memblock_find_in_range(0,
fd07383b
AM
383 min(new_area_start, memblock.current_limit),
384 new_alloc_size, PAGE_SIZE);
48c3b583 385
15674868 386 new_array = addr ? __va(addr) : NULL;
4e2f0775 387 }
1f5026a7 388 if (!addr) {
142b45a7
BH
389 pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n",
390 memblock_type_name(type), type->max, type->max * 2);
391 return -1;
392 }
142b45a7 393
fd07383b
AM
394 memblock_dbg("memblock: %s is doubled to %ld at [%#010llx-%#010llx]",
395 memblock_type_name(type), type->max * 2, (u64)addr,
396 (u64)addr + new_size - 1);
ea9e4376 397
fd07383b
AM
398 /*
399 * Found space, we now need to move the array over before we add the
400 * reserved region since it may be our reserved array itself that is
401 * full.
142b45a7
BH
402 */
403 memcpy(new_array, type->regions, old_size);
404 memset(new_array + type->max, 0, old_size);
405 old_array = type->regions;
406 type->regions = new_array;
407 type->max <<= 1;
408
fd07383b 409 /* Free old array. We needn't free it if the array is the static one */
181eb394
GS
410 if (*in_slab)
411 kfree(old_array);
412 else if (old_array != memblock_memory_init_regions &&
413 old_array != memblock_reserved_init_regions)
29f67386 414 memblock_free(__pa(old_array), old_alloc_size);
142b45a7 415
fd07383b
AM
416 /*
417 * Reserve the new array if that comes from the memblock. Otherwise, we
418 * needn't do it
181eb394
GS
419 */
420 if (!use_slab)
29f67386 421 BUG_ON(memblock_reserve(addr, new_alloc_size));
181eb394
GS
422
423 /* Update slab flag */
424 *in_slab = use_slab;
425
142b45a7
BH
426 return 0;
427}
428
784656f9
TH
429/**
430 * memblock_merge_regions - merge neighboring compatible regions
431 * @type: memblock type to scan
432 *
433 * Scan @type and merge neighboring compatible regions.
434 */
435static void __init_memblock memblock_merge_regions(struct memblock_type *type)
95f72d1e 436{
784656f9 437 int i = 0;
95f72d1e 438
784656f9
TH
439 /* cnt never goes below 1 */
440 while (i < type->cnt - 1) {
441 struct memblock_region *this = &type->regions[i];
442 struct memblock_region *next = &type->regions[i + 1];
95f72d1e 443
7c0caeb8
TH
444 if (this->base + this->size != next->base ||
445 memblock_get_region_node(this) !=
66a20757
TC
446 memblock_get_region_node(next) ||
447 this->flags != next->flags) {
784656f9
TH
448 BUG_ON(this->base + this->size > next->base);
449 i++;
450 continue;
8f7a6605
BH
451 }
452
784656f9 453 this->size += next->size;
c0232ae8
LF
454 /* move forward from next + 1, index of which is i + 2 */
455 memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next));
784656f9 456 type->cnt--;
95f72d1e 457 }
784656f9 458}
95f72d1e 459
784656f9
TH
460/**
461 * memblock_insert_region - insert new memblock region
209ff86d
TC
462 * @type: memblock type to insert into
463 * @idx: index for the insertion point
464 * @base: base address of the new region
465 * @size: size of the new region
466 * @nid: node id of the new region
66a20757 467 * @flags: flags of the new region
784656f9
TH
468 *
469 * Insert new memblock region [@base,@base+@size) into @type at @idx.
470 * @type must already have extra room to accomodate the new region.
471 */
472static void __init_memblock memblock_insert_region(struct memblock_type *type,
473 int idx, phys_addr_t base,
66a20757
TC
474 phys_addr_t size,
475 int nid, unsigned long flags)
784656f9
TH
476{
477 struct memblock_region *rgn = &type->regions[idx];
478
479 BUG_ON(type->cnt >= type->max);
480 memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn));
481 rgn->base = base;
482 rgn->size = size;
66a20757 483 rgn->flags = flags;
7c0caeb8 484 memblock_set_region_node(rgn, nid);
784656f9 485 type->cnt++;
1440c4e2 486 type->total_size += size;
784656f9
TH
487}
488
489/**
f1af9d3a 490 * memblock_add_range - add new memblock region
784656f9
TH
491 * @type: memblock type to add new region into
492 * @base: base address of the new region
493 * @size: size of the new region
7fb0bc3f 494 * @nid: nid of the new region
66a20757 495 * @flags: flags of the new region
784656f9
TH
496 *
497 * Add new memblock region [@base,@base+@size) into @type. The new region
498 * is allowed to overlap with existing ones - overlaps don't affect already
499 * existing regions. @type is guaranteed to be minimal (all neighbouring
500 * compatible regions are merged) after the addition.
501 *
502 * RETURNS:
503 * 0 on success, -errno on failure.
504 */
f1af9d3a 505int __init_memblock memblock_add_range(struct memblock_type *type,
66a20757
TC
506 phys_addr_t base, phys_addr_t size,
507 int nid, unsigned long flags)
784656f9
TH
508{
509 bool insert = false;
eb18f1b5
TH
510 phys_addr_t obase = base;
511 phys_addr_t end = base + memblock_cap_size(base, &size);
784656f9
TH
512 int i, nr_new;
513
b3dc627c
TH
514 if (!size)
515 return 0;
516
784656f9
TH
517 /* special case for empty array */
518 if (type->regions[0].size == 0) {
1440c4e2 519 WARN_ON(type->cnt != 1 || type->total_size);
8f7a6605
BH
520 type->regions[0].base = base;
521 type->regions[0].size = size;
66a20757 522 type->regions[0].flags = flags;
7fb0bc3f 523 memblock_set_region_node(&type->regions[0], nid);
1440c4e2 524 type->total_size = size;
8f7a6605 525 return 0;
95f72d1e 526 }
784656f9
TH
527repeat:
528 /*
529 * The following is executed twice. Once with %false @insert and
530 * then with %true. The first counts the number of regions needed
531 * to accomodate the new area. The second actually inserts them.
142b45a7 532 */
784656f9
TH
533 base = obase;
534 nr_new = 0;
95f72d1e 535
784656f9
TH
536 for (i = 0; i < type->cnt; i++) {
537 struct memblock_region *rgn = &type->regions[i];
538 phys_addr_t rbase = rgn->base;
539 phys_addr_t rend = rbase + rgn->size;
540
541 if (rbase >= end)
95f72d1e 542 break;
784656f9
TH
543 if (rend <= base)
544 continue;
545 /*
546 * @rgn overlaps. If it separates the lower part of new
547 * area, insert that portion.
548 */
549 if (rbase > base) {
550 nr_new++;
551 if (insert)
552 memblock_insert_region(type, i++, base,
66a20757
TC
553 rbase - base, nid,
554 flags);
95f72d1e 555 }
784656f9
TH
556 /* area below @rend is dealt with, forget about it */
557 base = min(rend, end);
95f72d1e 558 }
784656f9
TH
559
560 /* insert the remaining portion */
561 if (base < end) {
562 nr_new++;
563 if (insert)
66a20757
TC
564 memblock_insert_region(type, i, base, end - base,
565 nid, flags);
95f72d1e 566 }
95f72d1e 567
784656f9
TH
568 /*
569 * If this was the first round, resize array and repeat for actual
570 * insertions; otherwise, merge and return.
142b45a7 571 */
784656f9
TH
572 if (!insert) {
573 while (type->cnt + nr_new > type->max)
48c3b583 574 if (memblock_double_array(type, obase, size) < 0)
784656f9
TH
575 return -ENOMEM;
576 insert = true;
577 goto repeat;
578 } else {
579 memblock_merge_regions(type);
580 return 0;
142b45a7 581 }
95f72d1e
YL
582}
583
7fb0bc3f
TH
584int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size,
585 int nid)
586{
f1af9d3a 587 return memblock_add_range(&memblock.memory, base, size, nid, 0);
7fb0bc3f
TH
588}
589
6a4055bc
AK
590static int __init_memblock memblock_add_region(phys_addr_t base,
591 phys_addr_t size,
592 int nid,
593 unsigned long flags)
594{
595 struct memblock_type *_rgn = &memblock.memory;
596
597 memblock_dbg("memblock_add: [%#016llx-%#016llx] flags %#02lx %pF\n",
598 (unsigned long long)base,
599 (unsigned long long)base + size - 1,
600 flags, (void *)_RET_IP_);
601
602 return memblock_add_range(_rgn, base, size, nid, flags);
603}
604
581adcbe 605int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
95f72d1e 606{
6a4055bc 607 return memblock_add_region(base, size, MAX_NUMNODES, 0);
95f72d1e
YL
608}
609
6a9ceb31
TH
610/**
611 * memblock_isolate_range - isolate given range into disjoint memblocks
612 * @type: memblock type to isolate range for
613 * @base: base of range to isolate
614 * @size: size of range to isolate
615 * @start_rgn: out parameter for the start of isolated region
616 * @end_rgn: out parameter for the end of isolated region
617 *
618 * Walk @type and ensure that regions don't cross the boundaries defined by
619 * [@base,@base+@size). Crossing regions are split at the boundaries,
620 * which may create at most two more regions. The index of the first
621 * region inside the range is returned in *@start_rgn and end in *@end_rgn.
622 *
623 * RETURNS:
624 * 0 on success, -errno on failure.
625 */
626static int __init_memblock memblock_isolate_range(struct memblock_type *type,
627 phys_addr_t base, phys_addr_t size,
628 int *start_rgn, int *end_rgn)
629{
eb18f1b5 630 phys_addr_t end = base + memblock_cap_size(base, &size);
6a9ceb31
TH
631 int i;
632
633 *start_rgn = *end_rgn = 0;
634
b3dc627c
TH
635 if (!size)
636 return 0;
637
6a9ceb31
TH
638 /* we'll create at most two more regions */
639 while (type->cnt + 2 > type->max)
48c3b583 640 if (memblock_double_array(type, base, size) < 0)
6a9ceb31
TH
641 return -ENOMEM;
642
643 for (i = 0; i < type->cnt; i++) {
644 struct memblock_region *rgn = &type->regions[i];
645 phys_addr_t rbase = rgn->base;
646 phys_addr_t rend = rbase + rgn->size;
647
648 if (rbase >= end)
649 break;
650 if (rend <= base)
651 continue;
652
653 if (rbase < base) {
654 /*
655 * @rgn intersects from below. Split and continue
656 * to process the next region - the new top half.
657 */
658 rgn->base = base;
1440c4e2
TH
659 rgn->size -= base - rbase;
660 type->total_size -= base - rbase;
6a9ceb31 661 memblock_insert_region(type, i, rbase, base - rbase,
66a20757
TC
662 memblock_get_region_node(rgn),
663 rgn->flags);
6a9ceb31
TH
664 } else if (rend > end) {
665 /*
666 * @rgn intersects from above. Split and redo the
667 * current region - the new bottom half.
668 */
669 rgn->base = end;
1440c4e2
TH
670 rgn->size -= end - rbase;
671 type->total_size -= end - rbase;
6a9ceb31 672 memblock_insert_region(type, i--, rbase, end - rbase,
66a20757
TC
673 memblock_get_region_node(rgn),
674 rgn->flags);
6a9ceb31
TH
675 } else {
676 /* @rgn is fully contained, record it */
677 if (!*end_rgn)
678 *start_rgn = i;
679 *end_rgn = i + 1;
680 }
681 }
682
683 return 0;
684}
6a9ceb31 685
f1af9d3a
PH
686int __init_memblock memblock_remove_range(struct memblock_type *type,
687 phys_addr_t base, phys_addr_t size)
95f72d1e 688{
71936180
TH
689 int start_rgn, end_rgn;
690 int i, ret;
95f72d1e 691
71936180
TH
692 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
693 if (ret)
694 return ret;
95f72d1e 695
71936180
TH
696 for (i = end_rgn - 1; i >= start_rgn; i--)
697 memblock_remove_region(type, i);
8f7a6605 698 return 0;
95f72d1e
YL
699}
700
581adcbe 701int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size)
95f72d1e 702{
f1af9d3a 703 return memblock_remove_range(&memblock.memory, base, size);
95f72d1e
YL
704}
705
f1af9d3a 706
581adcbe 707int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size)
95f72d1e 708{
24aa0788 709 memblock_dbg(" memblock_free: [%#016llx-%#016llx] %pF\n",
a150439c 710 (unsigned long long)base,
931d13f5 711 (unsigned long long)base + size - 1,
a150439c 712 (void *)_RET_IP_);
24aa0788 713
aedf95ea 714 kmemleak_free_part(__va(base), size);
f1af9d3a 715 return memblock_remove_range(&memblock.reserved, base, size);
95f72d1e
YL
716}
717
66a20757
TC
718static int __init_memblock memblock_reserve_region(phys_addr_t base,
719 phys_addr_t size,
720 int nid,
721 unsigned long flags)
95f72d1e 722{
7fc825b4 723 struct memblock_type *type = &memblock.reserved;
95f72d1e 724
66a20757 725 memblock_dbg("memblock_reserve: [%#016llx-%#016llx] flags %#02lx %pF\n",
a150439c 726 (unsigned long long)base,
931d13f5 727 (unsigned long long)base + size - 1,
66a20757
TC
728 flags, (void *)_RET_IP_);
729
7fc825b4 730 return memblock_add_range(type, base, size, nid, flags);
66a20757 731}
95f72d1e 732
66a20757
TC
733int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size)
734{
735 return memblock_reserve_region(base, size, MAX_NUMNODES, 0);
95f72d1e
YL
736}
737
66b16edf 738/**
66b16edf 739 *
4308ce17 740 * This function isolates region [@base, @base + @size), and sets/clears flag
66b16edf
TC
741 *
742 * Return 0 on succees, -errno on failure.
743 */
4308ce17
TL
744static int __init_memblock memblock_setclr_flag(phys_addr_t base,
745 phys_addr_t size, int set, int flag)
66b16edf
TC
746{
747 struct memblock_type *type = &memblock.memory;
748 int i, ret, start_rgn, end_rgn;
749
750 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
751 if (ret)
752 return ret;
753
754 for (i = start_rgn; i < end_rgn; i++)
4308ce17
TL
755 if (set)
756 memblock_set_region_flags(&type->regions[i], flag);
757 else
758 memblock_clear_region_flags(&type->regions[i], flag);
66b16edf
TC
759
760 memblock_merge_regions(type);
761 return 0;
762}
763
764/**
4308ce17 765 * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG.
66b16edf
TC
766 * @base: the base phys addr of the region
767 * @size: the size of the region
768 *
4308ce17
TL
769 * Return 0 on succees, -errno on failure.
770 */
771int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size)
772{
773 return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG);
774}
775
776/**
777 * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region.
778 * @base: the base phys addr of the region
779 * @size: the size of the region
66b16edf
TC
780 *
781 * Return 0 on succees, -errno on failure.
782 */
783int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size)
784{
4308ce17 785 return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG);
66b16edf
TC
786}
787
35fd0808 788/**
f1af9d3a 789 * __next__mem_range - next function for for_each_free_mem_range() etc.
35fd0808 790 * @idx: pointer to u64 loop variable
b1154233 791 * @nid: node selector, %NUMA_NO_NODE for all nodes
fc6daaf9 792 * @flags: pick from blocks based on memory attributes
f1af9d3a
PH
793 * @type_a: pointer to memblock_type from where the range is taken
794 * @type_b: pointer to memblock_type which excludes memory from being taken
dad7557e
WL
795 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
796 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
797 * @out_nid: ptr to int for nid of the range, can be %NULL
35fd0808 798 *
f1af9d3a 799 * Find the first area from *@idx which matches @nid, fill the out
35fd0808 800 * parameters, and update *@idx for the next iteration. The lower 32bit of
f1af9d3a
PH
801 * *@idx contains index into type_a and the upper 32bit indexes the
802 * areas before each region in type_b. For example, if type_b regions
35fd0808
TH
803 * look like the following,
804 *
805 * 0:[0-16), 1:[32-48), 2:[128-130)
806 *
807 * The upper 32bit indexes the following regions.
808 *
809 * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX)
810 *
811 * As both region arrays are sorted, the function advances the two indices
812 * in lockstep and returns each intersection.
813 */
fc6daaf9 814void __init_memblock __next_mem_range(u64 *idx, int nid, ulong flags,
f1af9d3a
PH
815 struct memblock_type *type_a,
816 struct memblock_type *type_b,
817 phys_addr_t *out_start,
818 phys_addr_t *out_end, int *out_nid)
35fd0808 819{
f1af9d3a
PH
820 int idx_a = *idx & 0xffffffff;
821 int idx_b = *idx >> 32;
b1154233 822
f1af9d3a
PH
823 if (WARN_ONCE(nid == MAX_NUMNODES,
824 "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
560dca27 825 nid = NUMA_NO_NODE;
35fd0808 826
f1af9d3a
PH
827 for (; idx_a < type_a->cnt; idx_a++) {
828 struct memblock_region *m = &type_a->regions[idx_a];
829
35fd0808
TH
830 phys_addr_t m_start = m->base;
831 phys_addr_t m_end = m->base + m->size;
f1af9d3a 832 int m_nid = memblock_get_region_node(m);
35fd0808
TH
833
834 /* only memory regions are associated with nodes, check it */
f1af9d3a 835 if (nid != NUMA_NO_NODE && nid != m_nid)
35fd0808
TH
836 continue;
837
0a313a99
XQ
838 /* skip hotpluggable memory regions if needed */
839 if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
840 continue;
841
f1af9d3a
PH
842 if (!type_b) {
843 if (out_start)
844 *out_start = m_start;
845 if (out_end)
846 *out_end = m_end;
847 if (out_nid)
848 *out_nid = m_nid;
849 idx_a++;
850 *idx = (u32)idx_a | (u64)idx_b << 32;
851 return;
852 }
853
854 /* scan areas before each reservation */
855 for (; idx_b < type_b->cnt + 1; idx_b++) {
856 struct memblock_region *r;
857 phys_addr_t r_start;
858 phys_addr_t r_end;
859
860 r = &type_b->regions[idx_b];
861 r_start = idx_b ? r[-1].base + r[-1].size : 0;
862 r_end = idx_b < type_b->cnt ?
863 r->base : ULLONG_MAX;
35fd0808 864
f1af9d3a
PH
865 /*
866 * if idx_b advanced past idx_a,
867 * break out to advance idx_a
868 */
35fd0808
TH
869 if (r_start >= m_end)
870 break;
871 /* if the two regions intersect, we're done */
872 if (m_start < r_end) {
873 if (out_start)
f1af9d3a
PH
874 *out_start =
875 max(m_start, r_start);
35fd0808
TH
876 if (out_end)
877 *out_end = min(m_end, r_end);
878 if (out_nid)
f1af9d3a 879 *out_nid = m_nid;
35fd0808 880 /*
f1af9d3a
PH
881 * The region which ends first is
882 * advanced for the next iteration.
35fd0808
TH
883 */
884 if (m_end <= r_end)
f1af9d3a 885 idx_a++;
35fd0808 886 else
f1af9d3a
PH
887 idx_b++;
888 *idx = (u32)idx_a | (u64)idx_b << 32;
35fd0808
TH
889 return;
890 }
891 }
892 }
893
894 /* signal end of iteration */
895 *idx = ULLONG_MAX;
896}
897
7bd0b0f0 898/**
f1af9d3a
PH
899 * __next_mem_range_rev - generic next function for for_each_*_range_rev()
900 *
901 * Finds the next range from type_a which is not marked as unsuitable
902 * in type_b.
903 *
7bd0b0f0 904 * @idx: pointer to u64 loop variable
b1154233 905 * @nid: nid: node selector, %NUMA_NO_NODE for all nodes
fc6daaf9 906 * @flags: pick from blocks based on memory attributes
f1af9d3a
PH
907 * @type_a: pointer to memblock_type from where the range is taken
908 * @type_b: pointer to memblock_type which excludes memory from being taken
dad7557e
WL
909 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
910 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
911 * @out_nid: ptr to int for nid of the range, can be %NULL
7bd0b0f0 912 *
f1af9d3a 913 * Reverse of __next_mem_range().
7bd0b0f0 914 */
fc6daaf9 915void __init_memblock __next_mem_range_rev(u64 *idx, int nid, ulong flags,
f1af9d3a
PH
916 struct memblock_type *type_a,
917 struct memblock_type *type_b,
918 phys_addr_t *out_start,
919 phys_addr_t *out_end, int *out_nid)
7bd0b0f0 920{
f1af9d3a
PH
921 int idx_a = *idx & 0xffffffff;
922 int idx_b = *idx >> 32;
b1154233 923
560dca27
GS
924 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
925 nid = NUMA_NO_NODE;
7bd0b0f0
TH
926
927 if (*idx == (u64)ULLONG_MAX) {
f1af9d3a
PH
928 idx_a = type_a->cnt - 1;
929 idx_b = type_b->cnt;
7bd0b0f0
TH
930 }
931
f1af9d3a
PH
932 for (; idx_a >= 0; idx_a--) {
933 struct memblock_region *m = &type_a->regions[idx_a];
934
7bd0b0f0
TH
935 phys_addr_t m_start = m->base;
936 phys_addr_t m_end = m->base + m->size;
f1af9d3a 937 int m_nid = memblock_get_region_node(m);
7bd0b0f0
TH
938
939 /* only memory regions are associated with nodes, check it */
f1af9d3a 940 if (nid != NUMA_NO_NODE && nid != m_nid)
7bd0b0f0
TH
941 continue;
942
55ac590c
TC
943 /* skip hotpluggable memory regions if needed */
944 if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
945 continue;
946
f1af9d3a
PH
947 if (!type_b) {
948 if (out_start)
949 *out_start = m_start;
950 if (out_end)
951 *out_end = m_end;
952 if (out_nid)
953 *out_nid = m_nid;
954 idx_a++;
955 *idx = (u32)idx_a | (u64)idx_b << 32;
956 return;
957 }
958
959 /* scan areas before each reservation */
960 for (; idx_b >= 0; idx_b--) {
961 struct memblock_region *r;
962 phys_addr_t r_start;
963 phys_addr_t r_end;
964
965 r = &type_b->regions[idx_b];
966 r_start = idx_b ? r[-1].base + r[-1].size : 0;
967 r_end = idx_b < type_b->cnt ?
968 r->base : ULLONG_MAX;
969 /*
970 * if idx_b advanced past idx_a,
971 * break out to advance idx_a
972 */
7bd0b0f0 973
7bd0b0f0
TH
974 if (r_end <= m_start)
975 break;
976 /* if the two regions intersect, we're done */
977 if (m_end > r_start) {
978 if (out_start)
979 *out_start = max(m_start, r_start);
980 if (out_end)
981 *out_end = min(m_end, r_end);
982 if (out_nid)
f1af9d3a 983 *out_nid = m_nid;
7bd0b0f0 984 if (m_start >= r_start)
f1af9d3a 985 idx_a--;
7bd0b0f0 986 else
f1af9d3a
PH
987 idx_b--;
988 *idx = (u32)idx_a | (u64)idx_b << 32;
7bd0b0f0
TH
989 return;
990 }
991 }
992 }
f1af9d3a 993 /* signal end of iteration */
7bd0b0f0
TH
994 *idx = ULLONG_MAX;
995}
996
7c0caeb8
TH
997#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
998/*
999 * Common iterator interface used to define for_each_mem_range().
1000 */
1001void __init_memblock __next_mem_pfn_range(int *idx, int nid,
1002 unsigned long *out_start_pfn,
1003 unsigned long *out_end_pfn, int *out_nid)
1004{
1005 struct memblock_type *type = &memblock.memory;
1006 struct memblock_region *r;
1007
1008 while (++*idx < type->cnt) {
1009 r = &type->regions[*idx];
1010
1011 if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size))
1012 continue;
1013 if (nid == MAX_NUMNODES || nid == r->nid)
1014 break;
1015 }
1016 if (*idx >= type->cnt) {
1017 *idx = -1;
1018 return;
1019 }
1020
1021 if (out_start_pfn)
1022 *out_start_pfn = PFN_UP(r->base);
1023 if (out_end_pfn)
1024 *out_end_pfn = PFN_DOWN(r->base + r->size);
1025 if (out_nid)
1026 *out_nid = r->nid;
1027}
1028
1029/**
1030 * memblock_set_node - set node ID on memblock regions
1031 * @base: base of area to set node ID for
1032 * @size: size of area to set node ID for
e7e8de59 1033 * @type: memblock type to set node ID for
7c0caeb8
TH
1034 * @nid: node ID to set
1035 *
e7e8de59 1036 * Set the nid of memblock @type regions in [@base,@base+@size) to @nid.
7c0caeb8
TH
1037 * Regions which cross the area boundaries are split as necessary.
1038 *
1039 * RETURNS:
1040 * 0 on success, -errno on failure.
1041 */
1042int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size,
e7e8de59 1043 struct memblock_type *type, int nid)
7c0caeb8 1044{
6a9ceb31
TH
1045 int start_rgn, end_rgn;
1046 int i, ret;
7c0caeb8 1047
6a9ceb31
TH
1048 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
1049 if (ret)
1050 return ret;
7c0caeb8 1051
6a9ceb31 1052 for (i = start_rgn; i < end_rgn; i++)
e9d24ad3 1053 memblock_set_region_node(&type->regions[i], nid);
7c0caeb8
TH
1054
1055 memblock_merge_regions(type);
1056 return 0;
1057}
1058#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1059
2bfc2862
AM
1060static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
1061 phys_addr_t align, phys_addr_t start,
fc6daaf9 1062 phys_addr_t end, int nid, ulong flags)
95f72d1e 1063{
6ed311b2 1064 phys_addr_t found;
95f72d1e 1065
79f40fab
GS
1066 if (!align)
1067 align = SMP_CACHE_BYTES;
94f3d3af 1068
fc6daaf9
TL
1069 found = memblock_find_in_range_node(size, align, start, end, nid,
1070 flags);
aedf95ea
CM
1071 if (found && !memblock_reserve(found, size)) {
1072 /*
1073 * The min_count is set to 0 so that memblock allocations are
1074 * never reported as leaks.
1075 */
1076 kmemleak_alloc(__va(found), size, 0, 0);
6ed311b2 1077 return found;
aedf95ea 1078 }
6ed311b2 1079 return 0;
95f72d1e
YL
1080}
1081
2bfc2862 1082phys_addr_t __init memblock_alloc_range(phys_addr_t size, phys_addr_t align,
fc6daaf9
TL
1083 phys_addr_t start, phys_addr_t end,
1084 ulong flags)
2bfc2862 1085{
fc6daaf9
TL
1086 return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE,
1087 flags);
2bfc2862
AM
1088}
1089
1090static phys_addr_t __init memblock_alloc_base_nid(phys_addr_t size,
1091 phys_addr_t align, phys_addr_t max_addr,
fc6daaf9 1092 int nid, ulong flags)
2bfc2862 1093{
fc6daaf9 1094 return memblock_alloc_range_nid(size, align, 0, max_addr, nid, flags);
2bfc2862
AM
1095}
1096
7bd0b0f0
TH
1097phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid)
1098{
fc6daaf9
TL
1099 return memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE,
1100 nid, MEMBLOCK_NONE);
7bd0b0f0
TH
1101}
1102
1103phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
1104{
fc6daaf9
TL
1105 return memblock_alloc_base_nid(size, align, max_addr, NUMA_NO_NODE,
1106 MEMBLOCK_NONE);
7bd0b0f0
TH
1107}
1108
6ed311b2 1109phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
95f72d1e 1110{
6ed311b2
BH
1111 phys_addr_t alloc;
1112
1113 alloc = __memblock_alloc_base(size, align, max_addr);
1114
1115 if (alloc == 0)
1116 panic("ERROR: Failed to allocate 0x%llx bytes below 0x%llx.\n",
1117 (unsigned long long) size, (unsigned long long) max_addr);
1118
1119 return alloc;
95f72d1e
YL
1120}
1121
6ed311b2 1122phys_addr_t __init memblock_alloc(phys_addr_t size, phys_addr_t align)
95f72d1e 1123{
6ed311b2
BH
1124 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
1125}
95f72d1e 1126
9d1e2492
BH
1127phys_addr_t __init memblock_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid)
1128{
1129 phys_addr_t res = memblock_alloc_nid(size, align, nid);
1130
1131 if (res)
1132 return res;
15fb0972 1133 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
95f72d1e
YL
1134}
1135
26f09e9b
SS
1136/**
1137 * memblock_virt_alloc_internal - allocate boot memory block
1138 * @size: size of memory block to be allocated in bytes
1139 * @align: alignment of the region and block's size
1140 * @min_addr: the lower bound of the memory region to allocate (phys address)
1141 * @max_addr: the upper bound of the memory region to allocate (phys address)
1142 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1143 *
1144 * The @min_addr limit is dropped if it can not be satisfied and the allocation
1145 * will fall back to memory below @min_addr. Also, allocation may fall back
1146 * to any node in the system if the specified node can not
1147 * hold the requested memory.
1148 *
1149 * The allocation is performed from memory region limited by
1150 * memblock.current_limit if @max_addr == %BOOTMEM_ALLOC_ACCESSIBLE.
1151 *
1152 * The memory block is aligned on SMP_CACHE_BYTES if @align == 0.
1153 *
1154 * The phys address of allocated boot memory block is converted to virtual and
1155 * allocated memory is reset to 0.
1156 *
1157 * In addition, function sets the min_count to 0 using kmemleak_alloc for
1158 * allocated boot memory block, so that it is never reported as leaks.
1159 *
1160 * RETURNS:
1161 * Virtual address of allocated memory block on success, NULL on failure.
1162 */
1163static void * __init memblock_virt_alloc_internal(
1164 phys_addr_t size, phys_addr_t align,
1165 phys_addr_t min_addr, phys_addr_t max_addr,
1166 int nid)
1167{
1168 phys_addr_t alloc;
1169 void *ptr;
1170
560dca27
GS
1171 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
1172 nid = NUMA_NO_NODE;
26f09e9b
SS
1173
1174 /*
1175 * Detect any accidental use of these APIs after slab is ready, as at
1176 * this moment memblock may be deinitialized already and its
1177 * internal data may be destroyed (after execution of free_all_bootmem)
1178 */
1179 if (WARN_ON_ONCE(slab_is_available()))
1180 return kzalloc_node(size, GFP_NOWAIT, nid);
1181
1182 if (!align)
1183 align = SMP_CACHE_BYTES;
1184
f544e14f
YL
1185 if (max_addr > memblock.current_limit)
1186 max_addr = memblock.current_limit;
1187
26f09e9b
SS
1188again:
1189 alloc = memblock_find_in_range_node(size, align, min_addr, max_addr,
fc6daaf9 1190 nid, MEMBLOCK_NONE);
26f09e9b
SS
1191 if (alloc)
1192 goto done;
1193
1194 if (nid != NUMA_NO_NODE) {
1195 alloc = memblock_find_in_range_node(size, align, min_addr,
fc6daaf9
TL
1196 max_addr, NUMA_NO_NODE,
1197 MEMBLOCK_NONE);
26f09e9b
SS
1198 if (alloc)
1199 goto done;
1200 }
1201
1202 if (min_addr) {
1203 min_addr = 0;
1204 goto again;
1205 } else {
1206 goto error;
1207 }
1208
1209done:
1210 memblock_reserve(alloc, size);
1211 ptr = phys_to_virt(alloc);
1212 memset(ptr, 0, size);
1213
1214 /*
1215 * The min_count is set to 0 so that bootmem allocated blocks
1216 * are never reported as leaks. This is because many of these blocks
1217 * are only referred via the physical address which is not
1218 * looked up by kmemleak.
1219 */
1220 kmemleak_alloc(ptr, size, 0, 0);
1221
1222 return ptr;
1223
1224error:
1225 return NULL;
1226}
1227
1228/**
1229 * memblock_virt_alloc_try_nid_nopanic - allocate boot memory block
1230 * @size: size of memory block to be allocated in bytes
1231 * @align: alignment of the region and block's size
1232 * @min_addr: the lower bound of the memory region from where the allocation
1233 * is preferred (phys address)
1234 * @max_addr: the upper bound of the memory region from where the allocation
1235 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1236 * allocate only from memory limited by memblock.current_limit value
1237 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1238 *
1239 * Public version of _memblock_virt_alloc_try_nid_nopanic() which provides
1240 * additional debug information (including caller info), if enabled.
1241 *
1242 * RETURNS:
1243 * Virtual address of allocated memory block on success, NULL on failure.
1244 */
1245void * __init memblock_virt_alloc_try_nid_nopanic(
1246 phys_addr_t size, phys_addr_t align,
1247 phys_addr_t min_addr, phys_addr_t max_addr,
1248 int nid)
1249{
1250 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1251 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1252 (u64)max_addr, (void *)_RET_IP_);
1253 return memblock_virt_alloc_internal(size, align, min_addr,
1254 max_addr, nid);
1255}
1256
1257/**
1258 * memblock_virt_alloc_try_nid - allocate boot memory block with panicking
1259 * @size: size of memory block to be allocated in bytes
1260 * @align: alignment of the region and block's size
1261 * @min_addr: the lower bound of the memory region from where the allocation
1262 * is preferred (phys address)
1263 * @max_addr: the upper bound of the memory region from where the allocation
1264 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1265 * allocate only from memory limited by memblock.current_limit value
1266 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1267 *
1268 * Public panicking version of _memblock_virt_alloc_try_nid_nopanic()
1269 * which provides debug information (including caller info), if enabled,
1270 * and panics if the request can not be satisfied.
1271 *
1272 * RETURNS:
1273 * Virtual address of allocated memory block on success, NULL on failure.
1274 */
1275void * __init memblock_virt_alloc_try_nid(
1276 phys_addr_t size, phys_addr_t align,
1277 phys_addr_t min_addr, phys_addr_t max_addr,
1278 int nid)
1279{
1280 void *ptr;
1281
1282 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1283 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1284 (u64)max_addr, (void *)_RET_IP_);
1285 ptr = memblock_virt_alloc_internal(size, align,
1286 min_addr, max_addr, nid);
1287 if (ptr)
1288 return ptr;
1289
1290 panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx\n",
1291 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1292 (u64)max_addr);
1293 return NULL;
1294}
1295
1296/**
1297 * __memblock_free_early - free boot memory block
1298 * @base: phys starting address of the boot memory block
1299 * @size: size of the boot memory block in bytes
1300 *
1301 * Free boot memory block previously allocated by memblock_virt_alloc_xx() API.
1302 * The freeing memory will not be released to the buddy allocator.
1303 */
1304void __init __memblock_free_early(phys_addr_t base, phys_addr_t size)
1305{
1306 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1307 __func__, (u64)base, (u64)base + size - 1,
1308 (void *)_RET_IP_);
1309 kmemleak_free_part(__va(base), size);
f1af9d3a 1310 memblock_remove_range(&memblock.reserved, base, size);
26f09e9b
SS
1311}
1312
1313/*
1314 * __memblock_free_late - free bootmem block pages directly to buddy allocator
1315 * @addr: phys starting address of the boot memory block
1316 * @size: size of the boot memory block in bytes
1317 *
1318 * This is only useful when the bootmem allocator has already been torn
1319 * down, but we are still initializing the system. Pages are released directly
1320 * to the buddy allocator, no bootmem metadata is updated because it is gone.
1321 */
1322void __init __memblock_free_late(phys_addr_t base, phys_addr_t size)
1323{
1324 u64 cursor, end;
1325
1326 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1327 __func__, (u64)base, (u64)base + size - 1,
1328 (void *)_RET_IP_);
1329 kmemleak_free_part(__va(base), size);
1330 cursor = PFN_UP(base);
1331 end = PFN_DOWN(base + size);
1332
1333 for (; cursor < end; cursor++) {
1334 __free_pages_bootmem(pfn_to_page(cursor), 0);
1335 totalram_pages++;
1336 }
1337}
9d1e2492
BH
1338
1339/*
1340 * Remaining API functions
1341 */
1342
2898cc4c 1343phys_addr_t __init memblock_phys_mem_size(void)
95f72d1e 1344{
1440c4e2 1345 return memblock.memory.total_size;
95f72d1e
YL
1346}
1347
595ad9af
YL
1348phys_addr_t __init memblock_mem_size(unsigned long limit_pfn)
1349{
1350 unsigned long pages = 0;
1351 struct memblock_region *r;
1352 unsigned long start_pfn, end_pfn;
1353
1354 for_each_memblock(memory, r) {
1355 start_pfn = memblock_region_memory_base_pfn(r);
1356 end_pfn = memblock_region_memory_end_pfn(r);
1357 start_pfn = min_t(unsigned long, start_pfn, limit_pfn);
1358 end_pfn = min_t(unsigned long, end_pfn, limit_pfn);
1359 pages += end_pfn - start_pfn;
1360 }
1361
16763230 1362 return PFN_PHYS(pages);
595ad9af
YL
1363}
1364
0a93ebef
SR
1365/* lowest address */
1366phys_addr_t __init_memblock memblock_start_of_DRAM(void)
1367{
1368 return memblock.memory.regions[0].base;
1369}
1370
10d06439 1371phys_addr_t __init_memblock memblock_end_of_DRAM(void)
95f72d1e
YL
1372{
1373 int idx = memblock.memory.cnt - 1;
1374
e3239ff9 1375 return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size);
95f72d1e
YL
1376}
1377
c0ce8fef 1378void __init memblock_enforce_memory_limit(phys_addr_t limit)
95f72d1e 1379{
c0ce8fef 1380 phys_addr_t max_addr = (phys_addr_t)ULLONG_MAX;
136199f0 1381 struct memblock_region *r;
95f72d1e 1382
c0ce8fef 1383 if (!limit)
95f72d1e
YL
1384 return;
1385
c0ce8fef 1386 /* find out max address */
136199f0 1387 for_each_memblock(memory, r) {
c0ce8fef
TH
1388 if (limit <= r->size) {
1389 max_addr = r->base + limit;
1390 break;
95f72d1e 1391 }
c0ce8fef 1392 limit -= r->size;
95f72d1e 1393 }
c0ce8fef
TH
1394
1395 /* truncate both memory and reserved regions */
f1af9d3a
PH
1396 memblock_remove_range(&memblock.memory, max_addr,
1397 (phys_addr_t)ULLONG_MAX);
1398 memblock_remove_range(&memblock.reserved, max_addr,
1399 (phys_addr_t)ULLONG_MAX);
95f72d1e
YL
1400}
1401
cd79481d 1402static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr)
72d4b0b4
BH
1403{
1404 unsigned int left = 0, right = type->cnt;
1405
1406 do {
1407 unsigned int mid = (right + left) / 2;
1408
1409 if (addr < type->regions[mid].base)
1410 right = mid;
1411 else if (addr >= (type->regions[mid].base +
1412 type->regions[mid].size))
1413 left = mid + 1;
1414 else
1415 return mid;
1416 } while (left < right);
1417 return -1;
1418}
1419
2898cc4c 1420int __init memblock_is_reserved(phys_addr_t addr)
95f72d1e 1421{
72d4b0b4
BH
1422 return memblock_search(&memblock.reserved, addr) != -1;
1423}
95f72d1e 1424
3661ca66 1425int __init_memblock memblock_is_memory(phys_addr_t addr)
72d4b0b4
BH
1426{
1427 return memblock_search(&memblock.memory, addr) != -1;
1428}
1429
e76b63f8
YL
1430#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1431int __init_memblock memblock_search_pfn_nid(unsigned long pfn,
1432 unsigned long *start_pfn, unsigned long *end_pfn)
1433{
1434 struct memblock_type *type = &memblock.memory;
16763230 1435 int mid = memblock_search(type, PFN_PHYS(pfn));
e76b63f8
YL
1436
1437 if (mid == -1)
1438 return -1;
1439
f7e2f7e8
FF
1440 *start_pfn = PFN_DOWN(type->regions[mid].base);
1441 *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size);
e76b63f8
YL
1442
1443 return type->regions[mid].nid;
1444}
1445#endif
1446
eab30949
SB
1447/**
1448 * memblock_is_region_memory - check if a region is a subset of memory
1449 * @base: base of region to check
1450 * @size: size of region to check
1451 *
1452 * Check if the region [@base, @base+@size) is a subset of a memory block.
1453 *
1454 * RETURNS:
1455 * 0 if false, non-zero if true
1456 */
3661ca66 1457int __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size)
72d4b0b4 1458{
abb65272 1459 int idx = memblock_search(&memblock.memory, base);
eb18f1b5 1460 phys_addr_t end = base + memblock_cap_size(base, &size);
72d4b0b4
BH
1461
1462 if (idx == -1)
1463 return 0;
abb65272
TV
1464 return memblock.memory.regions[idx].base <= base &&
1465 (memblock.memory.regions[idx].base +
eb18f1b5 1466 memblock.memory.regions[idx].size) >= end;
95f72d1e
YL
1467}
1468
eab30949
SB
1469/**
1470 * memblock_is_region_reserved - check if a region intersects reserved memory
1471 * @base: base of region to check
1472 * @size: size of region to check
1473 *
1474 * Check if the region [@base, @base+@size) intersects a reserved memory block.
1475 *
1476 * RETURNS:
1477 * 0 if false, non-zero if true
1478 */
10d06439 1479int __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size)
95f72d1e 1480{
eb18f1b5 1481 memblock_cap_size(base, &size);
f1c2c19c 1482 return memblock_overlaps_region(&memblock.reserved, base, size) >= 0;
95f72d1e
YL
1483}
1484
6ede1fd3
YL
1485void __init_memblock memblock_trim_memory(phys_addr_t align)
1486{
6ede1fd3 1487 phys_addr_t start, end, orig_start, orig_end;
136199f0 1488 struct memblock_region *r;
6ede1fd3 1489
136199f0
EM
1490 for_each_memblock(memory, r) {
1491 orig_start = r->base;
1492 orig_end = r->base + r->size;
6ede1fd3
YL
1493 start = round_up(orig_start, align);
1494 end = round_down(orig_end, align);
1495
1496 if (start == orig_start && end == orig_end)
1497 continue;
1498
1499 if (start < end) {
136199f0
EM
1500 r->base = start;
1501 r->size = end - start;
6ede1fd3 1502 } else {
136199f0
EM
1503 memblock_remove_region(&memblock.memory,
1504 r - memblock.memory.regions);
1505 r--;
6ede1fd3
YL
1506 }
1507 }
1508}
e63075a3 1509
3661ca66 1510void __init_memblock memblock_set_current_limit(phys_addr_t limit)
e63075a3
BH
1511{
1512 memblock.current_limit = limit;
1513}
1514
fec51014
LA
1515phys_addr_t __init_memblock memblock_get_current_limit(void)
1516{
1517 return memblock.current_limit;
1518}
1519
7c0caeb8 1520static void __init_memblock memblock_dump(struct memblock_type *type, char *name)
6ed311b2
BH
1521{
1522 unsigned long long base, size;
66a20757 1523 unsigned long flags;
6ed311b2
BH
1524 int i;
1525
7c0caeb8 1526 pr_info(" %s.cnt = 0x%lx\n", name, type->cnt);
6ed311b2 1527
7c0caeb8
TH
1528 for (i = 0; i < type->cnt; i++) {
1529 struct memblock_region *rgn = &type->regions[i];
1530 char nid_buf[32] = "";
1531
1532 base = rgn->base;
1533 size = rgn->size;
66a20757 1534 flags = rgn->flags;
7c0caeb8
TH
1535#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1536 if (memblock_get_region_node(rgn) != MAX_NUMNODES)
1537 snprintf(nid_buf, sizeof(nid_buf), " on node %d",
1538 memblock_get_region_node(rgn));
1539#endif
66a20757
TC
1540 pr_info(" %s[%#x]\t[%#016llx-%#016llx], %#llx bytes%s flags: %#lx\n",
1541 name, i, base, base + size - 1, size, nid_buf, flags);
6ed311b2
BH
1542 }
1543}
1544
4ff7b82f 1545void __init_memblock __memblock_dump_all(void)
6ed311b2 1546{
6ed311b2 1547 pr_info("MEMBLOCK configuration:\n");
1440c4e2
TH
1548 pr_info(" memory size = %#llx reserved size = %#llx\n",
1549 (unsigned long long)memblock.memory.total_size,
1550 (unsigned long long)memblock.reserved.total_size);
6ed311b2
BH
1551
1552 memblock_dump(&memblock.memory, "memory");
1553 memblock_dump(&memblock.reserved, "reserved");
1554}
1555
1aadc056 1556void __init memblock_allow_resize(void)
6ed311b2 1557{
142b45a7 1558 memblock_can_resize = 1;
6ed311b2
BH
1559}
1560
6ed311b2
BH
1561static int __init early_memblock(char *p)
1562{
1563 if (p && strstr(p, "debug"))
1564 memblock_debug = 1;
1565 return 0;
1566}
1567early_param("memblock", early_memblock);
1568
c378ddd5 1569#if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK)
6d03b885
BH
1570
1571static int memblock_debug_show(struct seq_file *m, void *private)
1572{
1573 struct memblock_type *type = m->private;
1574 struct memblock_region *reg;
1575 int i;
1576
1577 for (i = 0; i < type->cnt; i++) {
1578 reg = &type->regions[i];
1579 seq_printf(m, "%4d: ", i);
1580 if (sizeof(phys_addr_t) == 4)
1581 seq_printf(m, "0x%08lx..0x%08lx\n",
1582 (unsigned long)reg->base,
1583 (unsigned long)(reg->base + reg->size - 1));
1584 else
1585 seq_printf(m, "0x%016llx..0x%016llx\n",
1586 (unsigned long long)reg->base,
1587 (unsigned long long)(reg->base + reg->size - 1));
1588
1589 }
1590 return 0;
1591}
1592
1593static int memblock_debug_open(struct inode *inode, struct file *file)
1594{
1595 return single_open(file, memblock_debug_show, inode->i_private);
1596}
1597
1598static const struct file_operations memblock_debug_fops = {
1599 .open = memblock_debug_open,
1600 .read = seq_read,
1601 .llseek = seq_lseek,
1602 .release = single_release,
1603};
1604
1605static int __init memblock_init_debugfs(void)
1606{
1607 struct dentry *root = debugfs_create_dir("memblock", NULL);
1608 if (!root)
1609 return -ENXIO;
1610 debugfs_create_file("memory", S_IRUGO, root, &memblock.memory, &memblock_debug_fops);
1611 debugfs_create_file("reserved", S_IRUGO, root, &memblock.reserved, &memblock_debug_fops);
70210ed9
PH
1612#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
1613 debugfs_create_file("physmem", S_IRUGO, root, &memblock.physmem, &memblock_debug_fops);
1614#endif
6d03b885
BH
1615
1616 return 0;
1617}
1618__initcall(memblock_init_debugfs);
1619
1620#endif /* CONFIG_DEBUG_FS */
This page took 0.430622 seconds and 5 git commands to generate.