blob: bd05a70f44b96c81d4d7c119fd4a807f1af15bd6 [file] [log] [blame]
Yinghai Lu09325872011-02-24 14:43:05 +01001/*
2 * bootmem - A boot-time physical memory allocator and configurator
3 *
4 * Copyright (C) 1999 Ingo Molnar
5 * 1999 Kanoj Sarcar, SGI
6 * 2008 Johannes Weiner
7 *
8 * Access to this subsystem has to be serialized externally (which is true
9 * for the boot process anyway).
10 */
11#include <linux/init.h>
12#include <linux/pfn.h>
13#include <linux/slab.h>
14#include <linux/bootmem.h>
Paul Gortmakerb95f1b312011-10-16 02:01:52 -040015#include <linux/export.h>
Yinghai Lu09325872011-02-24 14:43:05 +010016#include <linux/kmemleak.h>
17#include <linux/range.h>
18#include <linux/memblock.h>
19
20#include <asm/bug.h>
21#include <asm/io.h>
22#include <asm/processor.h>
23
24#include "internal.h"
25
Yinghai Lue782ab42011-02-24 14:43:06 +010026#ifndef CONFIG_NEED_MULTIPLE_NODES
27struct pglist_data __refdata contig_page_data;
28EXPORT_SYMBOL(contig_page_data);
29#endif
30
Yinghai Lu09325872011-02-24 14:43:05 +010031unsigned long max_low_pfn;
32unsigned long min_low_pfn;
33unsigned long max_pfn;
Igor Mammedov8dd33032015-12-04 14:07:05 +010034unsigned long long max_possible_pfn;
Yinghai Lu09325872011-02-24 14:43:05 +010035
Yinghai Lu8bc1f912011-02-24 14:43:06 +010036static void * __init __alloc_memory_core_early(int nid, u64 size, u64 align,
37 u64 goal, u64 limit)
38{
39 void *ptr;
40 u64 addr;
Tony Lucka3f5baf2015-06-24 16:58:12 -070041 ulong flags = choose_memblock_flags();
Yinghai Lu8bc1f912011-02-24 14:43:06 +010042
43 if (limit > memblock.current_limit)
44 limit = memblock.current_limit;
45
Tony Lucka3f5baf2015-06-24 16:58:12 -070046again:
Tony Luckfc6daaf2015-06-24 16:58:09 -070047 addr = memblock_find_in_range_node(size, align, goal, limit, nid,
Tony Lucka3f5baf2015-06-24 16:58:12 -070048 flags);
49 if (!addr && (flags & MEMBLOCK_MIRROR)) {
50 flags &= ~MEMBLOCK_MIRROR;
51 pr_warn("Could not allocate %pap bytes of mirrored memory\n",
52 &size);
53 goto again;
54 }
Tejun Heo1f5026a2011-07-12 09:58:09 +020055 if (!addr)
Yinghai Lu8bc1f912011-02-24 14:43:06 +010056 return NULL;
57
Philipp Hachtmann87379ec2014-01-23 15:53:10 -080058 if (memblock_reserve(addr, size))
59 return NULL;
60
Yinghai Lu8bc1f912011-02-24 14:43:06 +010061 ptr = phys_to_virt(addr);
62 memset(ptr, 0, size);
Yinghai Lu8bc1f912011-02-24 14:43:06 +010063 /*
64 * The min_count is set to 0 so that bootmem allocated blocks
65 * are never reported as leaks.
66 */
67 kmemleak_alloc(ptr, size, 0, 0);
68 return ptr;
69}
70
Yinghai Lu09325872011-02-24 14:43:05 +010071/*
72 * free_bootmem_late - free bootmem pages directly to page allocator
73 * @addr: starting address of the range
74 * @size: size of the range in bytes
75 *
76 * This is only useful when the bootmem allocator has already been torn
77 * down, but we are still initializing the system. Pages are given directly
78 * to the page allocator, no bootmem metadata is updated because it is gone.
79 */
80void __init free_bootmem_late(unsigned long addr, unsigned long size)
81{
82 unsigned long cursor, end;
83
84 kmemleak_free_part(__va(addr), size);
85
86 cursor = PFN_UP(addr);
87 end = PFN_DOWN(addr + size);
88
89 for (; cursor < end; cursor++) {
Mel Gormand70ddd72015-06-30 14:56:52 -070090 __free_pages_bootmem(pfn_to_page(cursor), cursor, 0);
Yinghai Lu09325872011-02-24 14:43:05 +010091 totalram_pages++;
92 }
93}
94
95static void __init __free_pages_memory(unsigned long start, unsigned long end)
96{
Robin Holt309d0b32013-11-12 15:07:23 -080097 int order;
Yinghai Lu09325872011-02-24 14:43:05 +010098
Robin Holt309d0b32013-11-12 15:07:23 -080099 while (start < end) {
100 order = min(MAX_ORDER - 1UL, __ffs(start));
Yinghai Lu09325872011-02-24 14:43:05 +0100101
Robin Holt309d0b32013-11-12 15:07:23 -0800102 while (start + (1UL << order) > end)
103 order--;
Yinghai Lu09325872011-02-24 14:43:05 +0100104
Mel Gormand70ddd72015-06-30 14:56:52 -0700105 __free_pages_bootmem(pfn_to_page(start), start, order);
Robin Holt309d0b32013-11-12 15:07:23 -0800106
107 start += (1UL << order);
Yinghai Lu09325872011-02-24 14:43:05 +0100108 }
Yinghai Lu09325872011-02-24 14:43:05 +0100109}
110
Yinghai Lu29f67382012-07-11 14:02:56 -0700111static unsigned long __init __free_memory_core(phys_addr_t start,
112 phys_addr_t end)
113{
114 unsigned long start_pfn = PFN_UP(start);
115 unsigned long end_pfn = min_t(unsigned long,
116 PFN_DOWN(end), max_low_pfn);
117
118 if (start_pfn > end_pfn)
119 return 0;
120
121 __free_pages_memory(start_pfn, end_pfn);
122
123 return end_pfn - start_pfn;
124}
125
Joonsoo Kimb4def352013-04-29 15:08:52 -0700126static unsigned long __init free_low_memory_core_early(void)
Yinghai Lu09325872011-02-24 14:43:05 +0100127{
Yinghai Lu09325872011-02-24 14:43:05 +0100128 unsigned long count = 0;
Philipp Hachtmann354f17e2014-01-23 15:53:24 -0800129 phys_addr_t start, end;
Tejun Heo8a9ca342011-07-12 11:16:02 +0200130 u64 i;
Yinghai Lu09325872011-02-24 14:43:05 +0100131
Xishi Qiu0a313a92014-09-09 14:50:46 -0700132 memblock_clear_hotplug(0, -1);
133
Nathan Zimmer92923ca2015-06-30 14:56:48 -0700134 for_each_reserved_mem_region(i, &start, &end)
135 reserve_bootmem_region(start, end);
136
Tony Luckfc6daaf2015-06-24 16:58:09 -0700137 for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end,
138 NULL)
Yinghai Lu29f67382012-07-11 14:02:56 -0700139 count += __free_memory_core(start, end);
Yinghai Lu09325872011-02-24 14:43:05 +0100140
Philipp Hachtmann5e270e22014-01-23 15:53:11 -0800141#ifdef CONFIG_ARCH_DISCARD_MEMBLOCK
Philipp Hachtmann354f17e2014-01-23 15:53:24 -0800142 {
143 phys_addr_t size;
Philipp Hachtmann5e270e22014-01-23 15:53:11 -0800144
Philipp Hachtmann354f17e2014-01-23 15:53:24 -0800145 /* Free memblock.reserved array if it was allocated */
146 size = get_allocated_memblock_reserved_regions_info(&start);
147 if (size)
148 count += __free_memory_core(start, start + size);
149
150 /* Free memblock.memory array if it was allocated */
151 size = get_allocated_memblock_memory_regions_info(&start);
152 if (size)
153 count += __free_memory_core(start, start + size);
154 }
Philipp Hachtmann5e270e22014-01-23 15:53:11 -0800155#endif
156
Yinghai Lu09325872011-02-24 14:43:05 +0100157 return count;
158}
159
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700160static int reset_managed_pages_done __initdata;
161
Tang Chenf784a3f2014-11-13 15:19:39 -0800162void reset_node_managed_pages(pg_data_t *pgdat)
Jiang Liu9feedc92012-12-12 13:52:12 -0800163{
164 struct zone *z;
165
Jiang Liu9feedc92012-12-12 13:52:12 -0800166 for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++)
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700167 z->managed_pages = 0;
168}
169
170void __init reset_all_zones_managed_pages(void)
171{
172 struct pglist_data *pgdat;
173
Tang Chenf784a3f2014-11-13 15:19:39 -0800174 if (reset_managed_pages_done)
175 return;
176
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700177 for_each_online_pgdat(pgdat)
178 reset_node_managed_pages(pgdat);
Tang Chenf784a3f2014-11-13 15:19:39 -0800179
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700180 reset_managed_pages_done = 1;
Jiang Liu9feedc92012-12-12 13:52:12 -0800181}
182
Yinghai Lu09325872011-02-24 14:43:05 +0100183/**
Yinghai Lu09325872011-02-24 14:43:05 +0100184 * free_all_bootmem - release free pages to the buddy allocator
185 *
186 * Returns the number of pages actually released.
187 */
188unsigned long __init free_all_bootmem(void)
189{
Jiang Liu0c988532013-07-03 15:03:24 -0700190 unsigned long pages;
191
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700192 reset_all_zones_managed_pages();
Jiang Liu9feedc92012-12-12 13:52:12 -0800193
Yinghai Lu09325872011-02-24 14:43:05 +0100194 /*
Grygorii Strashkob1154232014-01-21 15:50:16 -0800195 * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300196 * because in some case like Node0 doesn't have RAM installed
Yinghai Lu09325872011-02-24 14:43:05 +0100197 * low ram will be on Node1
Yinghai Lu09325872011-02-24 14:43:05 +0100198 */
Jiang Liu0c988532013-07-03 15:03:24 -0700199 pages = free_low_memory_core_early();
200 totalram_pages += pages;
201
202 return pages;
Yinghai Lu09325872011-02-24 14:43:05 +0100203}
204
205/**
206 * free_bootmem_node - mark a page range as usable
207 * @pgdat: node the range resides on
208 * @physaddr: starting address of the range
209 * @size: size of the range in bytes
210 *
211 * Partial pages will be considered reserved and left as they are.
212 *
213 * The range must reside completely on the specified node.
214 */
215void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
216 unsigned long size)
217{
Tejun Heo24aa0782011-07-12 11:16:06 +0200218 memblock_free(physaddr, size);
Yinghai Lu09325872011-02-24 14:43:05 +0100219}
220
221/**
222 * free_bootmem - mark a page range as usable
223 * @addr: starting address of the range
224 * @size: size of the range in bytes
225 *
226 * Partial pages will be considered reserved and left as they are.
227 *
228 * The range must be contiguous but may span node boundaries.
229 */
230void __init free_bootmem(unsigned long addr, unsigned long size)
231{
Tejun Heo24aa0782011-07-12 11:16:06 +0200232 memblock_free(addr, size);
Yinghai Lu09325872011-02-24 14:43:05 +0100233}
234
235static void * __init ___alloc_bootmem_nopanic(unsigned long size,
236 unsigned long align,
237 unsigned long goal,
238 unsigned long limit)
239{
240 void *ptr;
241
242 if (WARN_ON_ONCE(slab_is_available()))
243 return kzalloc(size, GFP_NOWAIT);
244
245restart:
246
Grygorii Strashkob1154232014-01-21 15:50:16 -0800247 ptr = __alloc_memory_core_early(NUMA_NO_NODE, size, align, goal, limit);
Yinghai Lu09325872011-02-24 14:43:05 +0100248
249 if (ptr)
250 return ptr;
251
252 if (goal != 0) {
253 goal = 0;
254 goto restart;
255 }
256
257 return NULL;
258}
259
260/**
261 * __alloc_bootmem_nopanic - allocate boot memory without panicking
262 * @size: size of the request in bytes
263 * @align: alignment of the region
264 * @goal: preferred starting address of the region
265 *
266 * The goal is dropped if it can not be satisfied and the allocation will
267 * fall back to memory below @goal.
268 *
269 * Allocation may happen on any node in the system.
270 *
271 * Returns NULL on failure.
272 */
273void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align,
274 unsigned long goal)
275{
276 unsigned long limit = -1UL;
277
278 return ___alloc_bootmem_nopanic(size, align, goal, limit);
279}
280
281static void * __init ___alloc_bootmem(unsigned long size, unsigned long align,
282 unsigned long goal, unsigned long limit)
283{
284 void *mem = ___alloc_bootmem_nopanic(size, align, goal, limit);
285
286 if (mem)
287 return mem;
288 /*
289 * Whoops, we cannot satisfy the allocation request.
290 */
Joe Perches11705322016-03-17 14:19:50 -0700291 pr_alert("bootmem alloc of %lu bytes failed!\n", size);
Yinghai Lu09325872011-02-24 14:43:05 +0100292 panic("Out of memory");
293 return NULL;
294}
295
296/**
297 * __alloc_bootmem - allocate boot memory
298 * @size: size of the request in bytes
299 * @align: alignment of the region
300 * @goal: preferred starting address of the region
301 *
302 * The goal is dropped if it can not be satisfied and the allocation will
303 * fall back to memory below @goal.
304 *
305 * Allocation may happen on any node in the system.
306 *
307 * The function panics if the request can not be satisfied.
308 */
309void * __init __alloc_bootmem(unsigned long size, unsigned long align,
310 unsigned long goal)
311{
312 unsigned long limit = -1UL;
313
314 return ___alloc_bootmem(size, align, goal, limit);
315}
316
Yinghai Lu99ab7b12012-07-11 14:02:53 -0700317void * __init ___alloc_bootmem_node_nopanic(pg_data_t *pgdat,
Johannes Weinerba539862012-05-29 15:06:35 -0700318 unsigned long size,
319 unsigned long align,
320 unsigned long goal,
321 unsigned long limit)
322{
323 void *ptr;
324
325again:
326 ptr = __alloc_memory_core_early(pgdat->node_id, size, align,
327 goal, limit);
328 if (ptr)
329 return ptr;
330
Grygorii Strashkob1154232014-01-21 15:50:16 -0800331 ptr = __alloc_memory_core_early(NUMA_NO_NODE, size, align,
Johannes Weinerba539862012-05-29 15:06:35 -0700332 goal, limit);
333 if (ptr)
334 return ptr;
335
336 if (goal) {
337 goal = 0;
338 goto again;
339 }
340
341 return NULL;
342}
343
344void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size,
345 unsigned long align, unsigned long goal)
346{
347 if (WARN_ON_ONCE(slab_is_available()))
348 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
349
350 return ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0);
351}
352
Rashika Kheriade498502014-04-03 14:48:06 -0700353static void * __init ___alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
Johannes Weinerba539862012-05-29 15:06:35 -0700354 unsigned long align, unsigned long goal,
355 unsigned long limit)
356{
357 void *ptr;
358
359 ptr = ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, limit);
360 if (ptr)
361 return ptr;
362
Joe Perches11705322016-03-17 14:19:50 -0700363 pr_alert("bootmem alloc of %lu bytes failed!\n", size);
Johannes Weinerba539862012-05-29 15:06:35 -0700364 panic("Out of memory");
365 return NULL;
366}
367
Yinghai Lu09325872011-02-24 14:43:05 +0100368/**
369 * __alloc_bootmem_node - allocate boot memory from a specific node
370 * @pgdat: node to allocate from
371 * @size: size of the request in bytes
372 * @align: alignment of the region
373 * @goal: preferred starting address of the region
374 *
375 * The goal is dropped if it can not be satisfied and the allocation will
376 * fall back to memory below @goal.
377 *
378 * Allocation may fall back to any node in the system if the specified node
379 * can not hold the requested memory.
380 *
381 * The function panics if the request can not be satisfied.
382 */
383void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
384 unsigned long align, unsigned long goal)
385{
Yinghai Lu09325872011-02-24 14:43:05 +0100386 if (WARN_ON_ONCE(slab_is_available()))
387 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
388
Johannes Weinerba539862012-05-29 15:06:35 -0700389 return ___alloc_bootmem_node(pgdat, size, align, goal, 0);
Yinghai Lu09325872011-02-24 14:43:05 +0100390}
391
392void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size,
393 unsigned long align, unsigned long goal)
394{
Yinghai Lu09325872011-02-24 14:43:05 +0100395 return __alloc_bootmem_node(pgdat, size, align, goal);
Yinghai Lu09325872011-02-24 14:43:05 +0100396}
397
Yinghai Lu09325872011-02-24 14:43:05 +0100398#ifndef ARCH_LOW_ADDRESS_LIMIT
399#define ARCH_LOW_ADDRESS_LIMIT 0xffffffffUL
400#endif
401
402/**
403 * __alloc_bootmem_low - allocate low boot memory
404 * @size: size of the request in bytes
405 * @align: alignment of the region
406 * @goal: preferred starting address of the region
407 *
408 * The goal is dropped if it can not be satisfied and the allocation will
409 * fall back to memory below @goal.
410 *
411 * Allocation may happen on any node in the system.
412 *
413 * The function panics if the request can not be satisfied.
414 */
415void * __init __alloc_bootmem_low(unsigned long size, unsigned long align,
416 unsigned long goal)
417{
418 return ___alloc_bootmem(size, align, goal, ARCH_LOW_ADDRESS_LIMIT);
419}
420
Yinghai Lu38fa4172013-01-24 12:20:15 -0800421void * __init __alloc_bootmem_low_nopanic(unsigned long size,
422 unsigned long align,
423 unsigned long goal)
424{
425 return ___alloc_bootmem_nopanic(size, align, goal,
426 ARCH_LOW_ADDRESS_LIMIT);
427}
428
Yinghai Lu09325872011-02-24 14:43:05 +0100429/**
430 * __alloc_bootmem_low_node - allocate low boot memory from a specific node
431 * @pgdat: node to allocate from
432 * @size: size of the request in bytes
433 * @align: alignment of the region
434 * @goal: preferred starting address of the region
435 *
436 * The goal is dropped if it can not be satisfied and the allocation will
437 * fall back to memory below @goal.
438 *
439 * Allocation may fall back to any node in the system if the specified node
440 * can not hold the requested memory.
441 *
442 * The function panics if the request can not be satisfied.
443 */
444void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size,
445 unsigned long align, unsigned long goal)
446{
Yinghai Lu09325872011-02-24 14:43:05 +0100447 if (WARN_ON_ONCE(slab_is_available()))
448 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
449
Johannes Weinerba539862012-05-29 15:06:35 -0700450 return ___alloc_bootmem_node(pgdat, size, align, goal,
451 ARCH_LOW_ADDRESS_LIMIT);
Yinghai Lu09325872011-02-24 14:43:05 +0100452}