blob: e1e8c63f7607d4ff7363b62fe3e69b465dd8ff82 [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>
Paul Gortmakerb95f1b312011-10-16 02:01:52 -040014#include <linux/export.h>
Yinghai Lu09325872011-02-24 14:43:05 +010015#include <linux/kmemleak.h>
16#include <linux/range.h>
17#include <linux/memblock.h>
zijun_hu23827052016-10-07 16:59:24 -070018#include <linux/bootmem.h>
Yinghai Lu09325872011-02-24 14:43:05 +010019
20#include <asm/bug.h>
21#include <asm/io.h>
Yinghai Lu09325872011-02-24 14:43:05 +010022
23#include "internal.h"
24
zijun_hu23827052016-10-07 16:59:24 -070025#ifndef CONFIG_HAVE_MEMBLOCK
26#error CONFIG_HAVE_MEMBLOCK not defined
27#endif
28
Yinghai Lue782ab42011-02-24 14:43:06 +010029#ifndef CONFIG_NEED_MULTIPLE_NODES
30struct pglist_data __refdata contig_page_data;
31EXPORT_SYMBOL(contig_page_data);
32#endif
33
Yinghai Lu09325872011-02-24 14:43:05 +010034unsigned long max_low_pfn;
35unsigned long min_low_pfn;
36unsigned long max_pfn;
Igor Mammedov8dd33032015-12-04 14:07:05 +010037unsigned long long max_possible_pfn;
Yinghai Lu09325872011-02-24 14:43:05 +010038
Yinghai Lu8bc1f912011-02-24 14:43:06 +010039static void * __init __alloc_memory_core_early(int nid, u64 size, u64 align,
40 u64 goal, u64 limit)
41{
42 void *ptr;
43 u64 addr;
Tony Lucka3f5baf2015-06-24 16:58:12 -070044 ulong flags = choose_memblock_flags();
Yinghai Lu8bc1f912011-02-24 14:43:06 +010045
46 if (limit > memblock.current_limit)
47 limit = memblock.current_limit;
48
Tony Lucka3f5baf2015-06-24 16:58:12 -070049again:
Tony Luckfc6daaf2015-06-24 16:58:09 -070050 addr = memblock_find_in_range_node(size, align, goal, limit, nid,
Tony Lucka3f5baf2015-06-24 16:58:12 -070051 flags);
52 if (!addr && (flags & MEMBLOCK_MIRROR)) {
53 flags &= ~MEMBLOCK_MIRROR;
54 pr_warn("Could not allocate %pap bytes of mirrored memory\n",
55 &size);
56 goto again;
57 }
Tejun Heo1f5026a2011-07-12 09:58:09 +020058 if (!addr)
Yinghai Lu8bc1f912011-02-24 14:43:06 +010059 return NULL;
60
Philipp Hachtmann87379ec2014-01-23 15:53:10 -080061 if (memblock_reserve(addr, size))
62 return NULL;
63
Yinghai Lu8bc1f912011-02-24 14:43:06 +010064 ptr = phys_to_virt(addr);
65 memset(ptr, 0, size);
Yinghai Lu8bc1f912011-02-24 14:43:06 +010066 /*
67 * The min_count is set to 0 so that bootmem allocated blocks
68 * are never reported as leaks.
69 */
70 kmemleak_alloc(ptr, size, 0, 0);
71 return ptr;
72}
73
Yinghai Lu09325872011-02-24 14:43:05 +010074/*
75 * free_bootmem_late - free bootmem pages directly to page allocator
76 * @addr: starting address of the range
77 * @size: size of the range in bytes
78 *
79 * This is only useful when the bootmem allocator has already been torn
80 * down, but we are still initializing the system. Pages are given directly
81 * to the page allocator, no bootmem metadata is updated because it is gone.
82 */
Laura Abbott7d0717562013-06-28 12:52:17 -070083void free_bootmem_late(unsigned long addr, unsigned long size)
Yinghai Lu09325872011-02-24 14:43:05 +010084{
85 unsigned long cursor, end;
86
Catalin Marinas9099dae2016-10-11 13:55:11 -070087 kmemleak_free_part_phys(addr, size);
Yinghai Lu09325872011-02-24 14:43:05 +010088
89 cursor = PFN_UP(addr);
90 end = PFN_DOWN(addr + size);
91
92 for (; cursor < end; cursor++) {
Mel Gormand70ddd72015-06-30 14:56:52 -070093 __free_pages_bootmem(pfn_to_page(cursor), cursor, 0);
Yinghai Lu09325872011-02-24 14:43:05 +010094 totalram_pages++;
95 }
96}
97
98static void __init __free_pages_memory(unsigned long start, unsigned long end)
99{
Robin Holt309d0b32013-11-12 15:07:23 -0800100 int order;
Yinghai Lu09325872011-02-24 14:43:05 +0100101
Robin Holt309d0b32013-11-12 15:07:23 -0800102 while (start < end) {
103 order = min(MAX_ORDER - 1UL, __ffs(start));
Yinghai Lu09325872011-02-24 14:43:05 +0100104
Robin Holt309d0b32013-11-12 15:07:23 -0800105 while (start + (1UL << order) > end)
106 order--;
Yinghai Lu09325872011-02-24 14:43:05 +0100107
Mel Gormand70ddd72015-06-30 14:56:52 -0700108 __free_pages_bootmem(pfn_to_page(start), start, order);
Robin Holt309d0b32013-11-12 15:07:23 -0800109
110 start += (1UL << order);
Yinghai Lu09325872011-02-24 14:43:05 +0100111 }
Yinghai Lu09325872011-02-24 14:43:05 +0100112}
113
Yinghai Lu29f67382012-07-11 14:02:56 -0700114static unsigned long __init __free_memory_core(phys_addr_t start,
115 phys_addr_t end)
116{
117 unsigned long start_pfn = PFN_UP(start);
118 unsigned long end_pfn = min_t(unsigned long,
119 PFN_DOWN(end), max_low_pfn);
120
121 if (start_pfn > end_pfn)
122 return 0;
123
124 __free_pages_memory(start_pfn, end_pfn);
125
126 return end_pfn - start_pfn;
127}
128
Joonsoo Kimb4def352013-04-29 15:08:52 -0700129static unsigned long __init free_low_memory_core_early(void)
Yinghai Lu09325872011-02-24 14:43:05 +0100130{
Yinghai Lu09325872011-02-24 14:43:05 +0100131 unsigned long count = 0;
Philipp Hachtmann354f17e2014-01-23 15:53:24 -0800132 phys_addr_t start, end;
Tejun Heo8a9ca342011-07-12 11:16:02 +0200133 u64 i;
Yinghai Lu09325872011-02-24 14:43:05 +0100134
Xishi Qiu0a313a92014-09-09 14:50:46 -0700135 memblock_clear_hotplug(0, -1);
136
Nathan Zimmer92923ca2015-06-30 14:56:48 -0700137 for_each_reserved_mem_region(i, &start, &end)
138 reserve_bootmem_region(start, end);
139
Wanlong Gao914a0512016-10-07 17:01:04 -0700140 /*
141 * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id
142 * because in some case like Node0 doesn't have RAM installed
143 * low ram will be on Node1
144 */
Tony Luckfc6daaf2015-06-24 16:58:09 -0700145 for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end,
146 NULL)
Yinghai Lu29f67382012-07-11 14:02:56 -0700147 count += __free_memory_core(start, end);
Yinghai Lu09325872011-02-24 14:43:05 +0100148
Philipp Hachtmann5e270e22014-01-23 15:53:11 -0800149#ifdef CONFIG_ARCH_DISCARD_MEMBLOCK
Philipp Hachtmann354f17e2014-01-23 15:53:24 -0800150 {
151 phys_addr_t size;
Philipp Hachtmann5e270e22014-01-23 15:53:11 -0800152
Philipp Hachtmann354f17e2014-01-23 15:53:24 -0800153 /* Free memblock.reserved array if it was allocated */
154 size = get_allocated_memblock_reserved_regions_info(&start);
155 if (size)
156 count += __free_memory_core(start, start + size);
157
158 /* Free memblock.memory array if it was allocated */
159 size = get_allocated_memblock_memory_regions_info(&start);
160 if (size)
161 count += __free_memory_core(start, start + size);
162 }
Philipp Hachtmann5e270e22014-01-23 15:53:11 -0800163#endif
164
Yinghai Lu09325872011-02-24 14:43:05 +0100165 return count;
166}
167
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700168static int reset_managed_pages_done __initdata;
169
Tang Chenf784a3f2014-11-13 15:19:39 -0800170void reset_node_managed_pages(pg_data_t *pgdat)
Jiang Liu9feedc92012-12-12 13:52:12 -0800171{
172 struct zone *z;
173
Jiang Liu9feedc92012-12-12 13:52:12 -0800174 for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++)
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700175 z->managed_pages = 0;
176}
177
178void __init reset_all_zones_managed_pages(void)
179{
180 struct pglist_data *pgdat;
181
Tang Chenf784a3f2014-11-13 15:19:39 -0800182 if (reset_managed_pages_done)
183 return;
184
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700185 for_each_online_pgdat(pgdat)
186 reset_node_managed_pages(pgdat);
Tang Chenf784a3f2014-11-13 15:19:39 -0800187
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700188 reset_managed_pages_done = 1;
Jiang Liu9feedc92012-12-12 13:52:12 -0800189}
190
Yinghai Lu09325872011-02-24 14:43:05 +0100191/**
Yinghai Lu09325872011-02-24 14:43:05 +0100192 * free_all_bootmem - release free pages to the buddy allocator
193 *
194 * Returns the number of pages actually released.
195 */
196unsigned long __init free_all_bootmem(void)
197{
Jiang Liu0c988532013-07-03 15:03:24 -0700198 unsigned long pages;
199
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700200 reset_all_zones_managed_pages();
Jiang Liu9feedc92012-12-12 13:52:12 -0800201
Jiang Liu0c988532013-07-03 15:03:24 -0700202 pages = free_low_memory_core_early();
203 totalram_pages += pages;
204
205 return pages;
Yinghai Lu09325872011-02-24 14:43:05 +0100206}
207
208/**
209 * free_bootmem_node - mark a page range as usable
210 * @pgdat: node the range resides on
211 * @physaddr: starting address of the range
212 * @size: size of the range in bytes
213 *
214 * Partial pages will be considered reserved and left as they are.
215 *
216 * The range must reside completely on the specified node.
217 */
218void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
219 unsigned long size)
220{
Tejun Heo24aa0782011-07-12 11:16:06 +0200221 memblock_free(physaddr, size);
Yinghai Lu09325872011-02-24 14:43:05 +0100222}
223
224/**
225 * free_bootmem - mark a page range as usable
226 * @addr: starting address of the range
227 * @size: size of the range in bytes
228 *
229 * Partial pages will be considered reserved and left as they are.
230 *
231 * The range must be contiguous but may span node boundaries.
232 */
233void __init free_bootmem(unsigned long addr, unsigned long size)
234{
Tejun Heo24aa0782011-07-12 11:16:06 +0200235 memblock_free(addr, size);
Yinghai Lu09325872011-02-24 14:43:05 +0100236}
237
238static void * __init ___alloc_bootmem_nopanic(unsigned long size,
239 unsigned long align,
240 unsigned long goal,
241 unsigned long limit)
242{
243 void *ptr;
244
245 if (WARN_ON_ONCE(slab_is_available()))
246 return kzalloc(size, GFP_NOWAIT);
247
248restart:
249
Grygorii Strashkob1154232014-01-21 15:50:16 -0800250 ptr = __alloc_memory_core_early(NUMA_NO_NODE, size, align, goal, limit);
Yinghai Lu09325872011-02-24 14:43:05 +0100251
252 if (ptr)
253 return ptr;
254
255 if (goal != 0) {
256 goal = 0;
257 goto restart;
258 }
259
260 return NULL;
261}
262
263/**
264 * __alloc_bootmem_nopanic - allocate boot memory without panicking
265 * @size: size of the request in bytes
266 * @align: alignment of the region
267 * @goal: preferred starting address of the region
268 *
269 * The goal is dropped if it can not be satisfied and the allocation will
270 * fall back to memory below @goal.
271 *
272 * Allocation may happen on any node in the system.
273 *
274 * Returns NULL on failure.
275 */
276void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align,
277 unsigned long goal)
278{
279 unsigned long limit = -1UL;
280
281 return ___alloc_bootmem_nopanic(size, align, goal, limit);
282}
283
284static void * __init ___alloc_bootmem(unsigned long size, unsigned long align,
285 unsigned long goal, unsigned long limit)
286{
287 void *mem = ___alloc_bootmem_nopanic(size, align, goal, limit);
288
289 if (mem)
290 return mem;
291 /*
292 * Whoops, we cannot satisfy the allocation request.
293 */
Joe Perches11705322016-03-17 14:19:50 -0700294 pr_alert("bootmem alloc of %lu bytes failed!\n", size);
Yinghai Lu09325872011-02-24 14:43:05 +0100295 panic("Out of memory");
296 return NULL;
297}
298
299/**
300 * __alloc_bootmem - allocate boot memory
301 * @size: size of the request in bytes
302 * @align: alignment of the region
303 * @goal: preferred starting address of the region
304 *
305 * The goal is dropped if it can not be satisfied and the allocation will
306 * fall back to memory below @goal.
307 *
308 * Allocation may happen on any node in the system.
309 *
310 * The function panics if the request can not be satisfied.
311 */
312void * __init __alloc_bootmem(unsigned long size, unsigned long align,
313 unsigned long goal)
314{
315 unsigned long limit = -1UL;
316
317 return ___alloc_bootmem(size, align, goal, limit);
318}
319
Yinghai Lu99ab7b12012-07-11 14:02:53 -0700320void * __init ___alloc_bootmem_node_nopanic(pg_data_t *pgdat,
Johannes Weinerba539862012-05-29 15:06:35 -0700321 unsigned long size,
322 unsigned long align,
323 unsigned long goal,
324 unsigned long limit)
325{
326 void *ptr;
327
328again:
329 ptr = __alloc_memory_core_early(pgdat->node_id, size, align,
330 goal, limit);
331 if (ptr)
332 return ptr;
333
Grygorii Strashkob1154232014-01-21 15:50:16 -0800334 ptr = __alloc_memory_core_early(NUMA_NO_NODE, size, align,
Johannes Weinerba539862012-05-29 15:06:35 -0700335 goal, limit);
336 if (ptr)
337 return ptr;
338
339 if (goal) {
340 goal = 0;
341 goto again;
342 }
343
344 return NULL;
345}
346
347void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size,
348 unsigned long align, unsigned long goal)
349{
350 if (WARN_ON_ONCE(slab_is_available()))
351 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
352
353 return ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0);
354}
355
Rashika Kheriade498502014-04-03 14:48:06 -0700356static void * __init ___alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
Johannes Weinerba539862012-05-29 15:06:35 -0700357 unsigned long align, unsigned long goal,
358 unsigned long limit)
359{
360 void *ptr;
361
362 ptr = ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, limit);
363 if (ptr)
364 return ptr;
365
Joe Perches11705322016-03-17 14:19:50 -0700366 pr_alert("bootmem alloc of %lu bytes failed!\n", size);
Johannes Weinerba539862012-05-29 15:06:35 -0700367 panic("Out of memory");
368 return NULL;
369}
370
Yinghai Lu09325872011-02-24 14:43:05 +0100371/**
372 * __alloc_bootmem_node - allocate boot memory from a specific node
373 * @pgdat: node to allocate from
374 * @size: size of the request in bytes
375 * @align: alignment of the region
376 * @goal: preferred starting address of the region
377 *
378 * The goal is dropped if it can not be satisfied and the allocation will
379 * fall back to memory below @goal.
380 *
381 * Allocation may fall back to any node in the system if the specified node
382 * can not hold the requested memory.
383 *
384 * The function panics if the request can not be satisfied.
385 */
386void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
387 unsigned long align, unsigned long goal)
388{
Yinghai Lu09325872011-02-24 14:43:05 +0100389 if (WARN_ON_ONCE(slab_is_available()))
390 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
391
Johannes Weinerba539862012-05-29 15:06:35 -0700392 return ___alloc_bootmem_node(pgdat, size, align, goal, 0);
Yinghai Lu09325872011-02-24 14:43:05 +0100393}
394
395void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size,
396 unsigned long align, unsigned long goal)
397{
Yinghai Lu09325872011-02-24 14:43:05 +0100398 return __alloc_bootmem_node(pgdat, size, align, goal);
Yinghai Lu09325872011-02-24 14:43:05 +0100399}
400
Yinghai Lu09325872011-02-24 14:43:05 +0100401
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}