blob: 9b02fda0886b8a87152fbdf54eab859899d75195 [file] [log] [blame]
Greg Kroah-Hartmanb2441312017-11-01 15:07:57 +01001// SPDX-License-Identifier: GPL-2.0
Yinghai Lu09325872011-02-24 14:43:05 +01002/*
3 * bootmem - A boot-time physical memory allocator and configurator
4 *
5 * Copyright (C) 1999 Ingo Molnar
6 * 1999 Kanoj Sarcar, SGI
7 * 2008 Johannes Weiner
8 *
9 * Access to this subsystem has to be serialized externally (which is true
10 * for the boot process anyway).
11 */
12#include <linux/init.h>
13#include <linux/pfn.h>
14#include <linux/slab.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>
zijun_hu23827052016-10-07 16:59:24 -070019#include <linux/bootmem.h>
Yinghai Lu09325872011-02-24 14:43:05 +010020
21#include <asm/bug.h>
22#include <asm/io.h>
Yinghai Lu09325872011-02-24 14:43:05 +010023
24#include "internal.h"
25
zijun_hu23827052016-10-07 16:59:24 -070026#ifndef CONFIG_HAVE_MEMBLOCK
27#error CONFIG_HAVE_MEMBLOCK not defined
28#endif
29
Yinghai Lue782ab42011-02-24 14:43:06 +010030#ifndef CONFIG_NEED_MULTIPLE_NODES
31struct pglist_data __refdata contig_page_data;
32EXPORT_SYMBOL(contig_page_data);
33#endif
34
Yinghai Lu09325872011-02-24 14:43:05 +010035unsigned long max_low_pfn;
36unsigned long min_low_pfn;
37unsigned long max_pfn;
Igor Mammedov8dd33032015-12-04 14:07:05 +010038unsigned long long max_possible_pfn;
Yinghai Lu09325872011-02-24 14:43:05 +010039
Yinghai Lu8bc1f912011-02-24 14:43:06 +010040static void * __init __alloc_memory_core_early(int nid, u64 size, u64 align,
41 u64 goal, u64 limit)
42{
43 void *ptr;
44 u64 addr;
Tony Lucka3f5baf2015-06-24 16:58:12 -070045 ulong flags = choose_memblock_flags();
Yinghai Lu8bc1f912011-02-24 14:43:06 +010046
47 if (limit > memblock.current_limit)
48 limit = memblock.current_limit;
49
Tony Lucka3f5baf2015-06-24 16:58:12 -070050again:
Tony Luckfc6daaf2015-06-24 16:58:09 -070051 addr = memblock_find_in_range_node(size, align, goal, limit, nid,
Tony Lucka3f5baf2015-06-24 16:58:12 -070052 flags);
53 if (!addr && (flags & MEMBLOCK_MIRROR)) {
54 flags &= ~MEMBLOCK_MIRROR;
55 pr_warn("Could not allocate %pap bytes of mirrored memory\n",
56 &size);
57 goto again;
58 }
Tejun Heo1f5026a2011-07-12 09:58:09 +020059 if (!addr)
Yinghai Lu8bc1f912011-02-24 14:43:06 +010060 return NULL;
61
Philipp Hachtmann87379ec2014-01-23 15:53:10 -080062 if (memblock_reserve(addr, size))
63 return NULL;
64
Yinghai Lu8bc1f912011-02-24 14:43:06 +010065 ptr = phys_to_virt(addr);
66 memset(ptr, 0, size);
Yinghai Lu8bc1f912011-02-24 14:43:06 +010067 /*
68 * The min_count is set to 0 so that bootmem allocated blocks
69 * are never reported as leaks.
70 */
71 kmemleak_alloc(ptr, size, 0, 0);
72 return ptr;
73}
74
Yinghai Lu09325872011-02-24 14:43:05 +010075/*
76 * free_bootmem_late - free bootmem pages directly to page allocator
77 * @addr: starting address of the range
78 * @size: size of the range in bytes
79 *
80 * This is only useful when the bootmem allocator has already been torn
81 * down, but we are still initializing the system. Pages are given directly
82 * to the page allocator, no bootmem metadata is updated because it is gone.
83 */
84void __init free_bootmem_late(unsigned long addr, unsigned long size)
85{
86 unsigned long cursor, end;
87
Catalin Marinas9099dae2016-10-11 13:55:11 -070088 kmemleak_free_part_phys(addr, size);
Yinghai Lu09325872011-02-24 14:43:05 +010089
90 cursor = PFN_UP(addr);
91 end = PFN_DOWN(addr + size);
92
93 for (; cursor < end; cursor++) {
Mel Gormand70ddd72015-06-30 14:56:52 -070094 __free_pages_bootmem(pfn_to_page(cursor), cursor, 0);
Yinghai Lu09325872011-02-24 14:43:05 +010095 totalram_pages++;
96 }
97}
98
99static void __init __free_pages_memory(unsigned long start, unsigned long end)
100{
Robin Holt309d0b32013-11-12 15:07:23 -0800101 int order;
Yinghai Lu09325872011-02-24 14:43:05 +0100102
Robin Holt309d0b32013-11-12 15:07:23 -0800103 while (start < end) {
104 order = min(MAX_ORDER - 1UL, __ffs(start));
Yinghai Lu09325872011-02-24 14:43:05 +0100105
Robin Holt309d0b32013-11-12 15:07:23 -0800106 while (start + (1UL << order) > end)
107 order--;
Yinghai Lu09325872011-02-24 14:43:05 +0100108
Mel Gormand70ddd72015-06-30 14:56:52 -0700109 __free_pages_bootmem(pfn_to_page(start), start, order);
Robin Holt309d0b32013-11-12 15:07:23 -0800110
111 start += (1UL << order);
Yinghai Lu09325872011-02-24 14:43:05 +0100112 }
Yinghai Lu09325872011-02-24 14:43:05 +0100113}
114
Yinghai Lu29f67382012-07-11 14:02:56 -0700115static unsigned long __init __free_memory_core(phys_addr_t start,
116 phys_addr_t end)
117{
118 unsigned long start_pfn = PFN_UP(start);
119 unsigned long end_pfn = min_t(unsigned long,
120 PFN_DOWN(end), max_low_pfn);
121
Wei Yang172ffeb2017-07-06 15:36:53 -0700122 if (start_pfn >= end_pfn)
Yinghai Lu29f67382012-07-11 14:02:56 -0700123 return 0;
124
125 __free_pages_memory(start_pfn, end_pfn);
126
127 return end_pfn - start_pfn;
128}
129
Joonsoo Kimb4def352013-04-29 15:08:52 -0700130static unsigned long __init free_low_memory_core_early(void)
Yinghai Lu09325872011-02-24 14:43:05 +0100131{
Yinghai Lu09325872011-02-24 14:43:05 +0100132 unsigned long count = 0;
Philipp Hachtmann354f17e2014-01-23 15:53:24 -0800133 phys_addr_t start, end;
Tejun Heo8a9ca342011-07-12 11:16:02 +0200134 u64 i;
Yinghai Lu09325872011-02-24 14:43:05 +0100135
Xishi Qiu0a313a92014-09-09 14:50:46 -0700136 memblock_clear_hotplug(0, -1);
137
Nathan Zimmer92923ca2015-06-30 14:56:48 -0700138 for_each_reserved_mem_region(i, &start, &end)
139 reserve_bootmem_region(start, end);
140
Wanlong Gao914a0512016-10-07 17:01:04 -0700141 /*
142 * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id
143 * because in some case like Node0 doesn't have RAM installed
144 * low ram will be on Node1
145 */
Tony Luckfc6daaf2015-06-24 16:58:09 -0700146 for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end,
147 NULL)
Yinghai Lu29f67382012-07-11 14:02:56 -0700148 count += __free_memory_core(start, end);
Yinghai Lu09325872011-02-24 14:43:05 +0100149
Yinghai Lu09325872011-02-24 14:43:05 +0100150 return count;
151}
152
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700153static int reset_managed_pages_done __initdata;
154
Tang Chenf784a3f2014-11-13 15:19:39 -0800155void reset_node_managed_pages(pg_data_t *pgdat)
Jiang Liu9feedc92012-12-12 13:52:12 -0800156{
157 struct zone *z;
158
Jiang Liu9feedc92012-12-12 13:52:12 -0800159 for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++)
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700160 z->managed_pages = 0;
161}
162
163void __init reset_all_zones_managed_pages(void)
164{
165 struct pglist_data *pgdat;
166
Tang Chenf784a3f2014-11-13 15:19:39 -0800167 if (reset_managed_pages_done)
168 return;
169
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700170 for_each_online_pgdat(pgdat)
171 reset_node_managed_pages(pgdat);
Tang Chenf784a3f2014-11-13 15:19:39 -0800172
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700173 reset_managed_pages_done = 1;
Jiang Liu9feedc92012-12-12 13:52:12 -0800174}
175
Yinghai Lu09325872011-02-24 14:43:05 +0100176/**
Yinghai Lu09325872011-02-24 14:43:05 +0100177 * free_all_bootmem - release free pages to the buddy allocator
178 *
179 * Returns the number of pages actually released.
180 */
181unsigned long __init free_all_bootmem(void)
182{
Jiang Liu0c988532013-07-03 15:03:24 -0700183 unsigned long pages;
184
Jiang Liu7b4b2a02013-07-03 15:03:11 -0700185 reset_all_zones_managed_pages();
Jiang Liu9feedc92012-12-12 13:52:12 -0800186
Jiang Liu0c988532013-07-03 15:03:24 -0700187 pages = free_low_memory_core_early();
188 totalram_pages += pages;
189
190 return pages;
Yinghai Lu09325872011-02-24 14:43:05 +0100191}
192
193/**
194 * free_bootmem_node - mark a page range as usable
195 * @pgdat: node the range resides on
196 * @physaddr: starting address of the range
197 * @size: size of the range in bytes
198 *
199 * Partial pages will be considered reserved and left as they are.
200 *
201 * The range must reside completely on the specified node.
202 */
203void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
204 unsigned long size)
205{
Tejun Heo24aa0782011-07-12 11:16:06 +0200206 memblock_free(physaddr, size);
Yinghai Lu09325872011-02-24 14:43:05 +0100207}
208
209/**
210 * free_bootmem - mark a page range as usable
211 * @addr: 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 be contiguous but may span node boundaries.
217 */
218void __init free_bootmem(unsigned long addr, unsigned long size)
219{
Tejun Heo24aa0782011-07-12 11:16:06 +0200220 memblock_free(addr, size);
Yinghai Lu09325872011-02-24 14:43:05 +0100221}
222
223static void * __init ___alloc_bootmem_nopanic(unsigned long size,
224 unsigned long align,
225 unsigned long goal,
226 unsigned long limit)
227{
228 void *ptr;
229
230 if (WARN_ON_ONCE(slab_is_available()))
231 return kzalloc(size, GFP_NOWAIT);
232
233restart:
234
Grygorii Strashkob1154232014-01-21 15:50:16 -0800235 ptr = __alloc_memory_core_early(NUMA_NO_NODE, size, align, goal, limit);
Yinghai Lu09325872011-02-24 14:43:05 +0100236
237 if (ptr)
238 return ptr;
239
240 if (goal != 0) {
241 goal = 0;
242 goto restart;
243 }
244
245 return NULL;
246}
247
248/**
249 * __alloc_bootmem_nopanic - allocate boot memory without panicking
250 * @size: size of the request in bytes
251 * @align: alignment of the region
252 * @goal: preferred starting address of the region
253 *
254 * The goal is dropped if it can not be satisfied and the allocation will
255 * fall back to memory below @goal.
256 *
257 * Allocation may happen on any node in the system.
258 *
259 * Returns NULL on failure.
260 */
261void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align,
262 unsigned long goal)
263{
264 unsigned long limit = -1UL;
265
266 return ___alloc_bootmem_nopanic(size, align, goal, limit);
267}
268
269static void * __init ___alloc_bootmem(unsigned long size, unsigned long align,
270 unsigned long goal, unsigned long limit)
271{
272 void *mem = ___alloc_bootmem_nopanic(size, align, goal, limit);
273
274 if (mem)
275 return mem;
276 /*
277 * Whoops, we cannot satisfy the allocation request.
278 */
Joe Perches11705322016-03-17 14:19:50 -0700279 pr_alert("bootmem alloc of %lu bytes failed!\n", size);
Yinghai Lu09325872011-02-24 14:43:05 +0100280 panic("Out of memory");
281 return NULL;
282}
283
284/**
285 * __alloc_bootmem - allocate boot memory
286 * @size: size of the request in bytes
287 * @align: alignment of the region
288 * @goal: preferred starting address of the region
289 *
290 * The goal is dropped if it can not be satisfied and the allocation will
291 * fall back to memory below @goal.
292 *
293 * Allocation may happen on any node in the system.
294 *
295 * The function panics if the request can not be satisfied.
296 */
297void * __init __alloc_bootmem(unsigned long size, unsigned long align,
298 unsigned long goal)
299{
300 unsigned long limit = -1UL;
301
302 return ___alloc_bootmem(size, align, goal, limit);
303}
304
Yinghai Lu99ab7b12012-07-11 14:02:53 -0700305void * __init ___alloc_bootmem_node_nopanic(pg_data_t *pgdat,
Johannes Weinerba539862012-05-29 15:06:35 -0700306 unsigned long size,
307 unsigned long align,
308 unsigned long goal,
309 unsigned long limit)
310{
311 void *ptr;
312
313again:
314 ptr = __alloc_memory_core_early(pgdat->node_id, size, align,
315 goal, limit);
316 if (ptr)
317 return ptr;
318
Grygorii Strashkob1154232014-01-21 15:50:16 -0800319 ptr = __alloc_memory_core_early(NUMA_NO_NODE, size, align,
Johannes Weinerba539862012-05-29 15:06:35 -0700320 goal, limit);
321 if (ptr)
322 return ptr;
323
324 if (goal) {
325 goal = 0;
326 goto again;
327 }
328
329 return NULL;
330}
331
332void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size,
333 unsigned long align, unsigned long goal)
334{
335 if (WARN_ON_ONCE(slab_is_available()))
336 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
337
338 return ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0);
339}
340
Rashika Kheriade498502014-04-03 14:48:06 -0700341static void * __init ___alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
Johannes Weinerba539862012-05-29 15:06:35 -0700342 unsigned long align, unsigned long goal,
343 unsigned long limit)
344{
345 void *ptr;
346
347 ptr = ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, limit);
348 if (ptr)
349 return ptr;
350
Joe Perches11705322016-03-17 14:19:50 -0700351 pr_alert("bootmem alloc of %lu bytes failed!\n", size);
Johannes Weinerba539862012-05-29 15:06:35 -0700352 panic("Out of memory");
353 return NULL;
354}
355
Yinghai Lu09325872011-02-24 14:43:05 +0100356/**
357 * __alloc_bootmem_node - allocate boot memory from a specific node
358 * @pgdat: node to allocate from
359 * @size: size of the request in bytes
360 * @align: alignment of the region
361 * @goal: preferred starting address of the region
362 *
363 * The goal is dropped if it can not be satisfied and the allocation will
364 * fall back to memory below @goal.
365 *
366 * Allocation may fall back to any node in the system if the specified node
367 * can not hold the requested memory.
368 *
369 * The function panics if the request can not be satisfied.
370 */
371void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
372 unsigned long align, unsigned long goal)
373{
Yinghai Lu09325872011-02-24 14:43:05 +0100374 if (WARN_ON_ONCE(slab_is_available()))
375 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
376
Johannes Weinerba539862012-05-29 15:06:35 -0700377 return ___alloc_bootmem_node(pgdat, size, align, goal, 0);
Yinghai Lu09325872011-02-24 14:43:05 +0100378}
379
380void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size,
381 unsigned long align, unsigned long goal)
382{
Yinghai Lu09325872011-02-24 14:43:05 +0100383 return __alloc_bootmem_node(pgdat, size, align, goal);
Yinghai Lu09325872011-02-24 14:43:05 +0100384}
385
Yinghai Lu09325872011-02-24 14:43:05 +0100386
387/**
388 * __alloc_bootmem_low - allocate low boot memory
389 * @size: size of the request in bytes
390 * @align: alignment of the region
391 * @goal: preferred starting address of the region
392 *
393 * The goal is dropped if it can not be satisfied and the allocation will
394 * fall back to memory below @goal.
395 *
396 * Allocation may happen on any node in the system.
397 *
398 * The function panics if the request can not be satisfied.
399 */
400void * __init __alloc_bootmem_low(unsigned long size, unsigned long align,
401 unsigned long goal)
402{
403 return ___alloc_bootmem(size, align, goal, ARCH_LOW_ADDRESS_LIMIT);
404}
405
Yinghai Lu38fa4172013-01-24 12:20:15 -0800406void * __init __alloc_bootmem_low_nopanic(unsigned long size,
407 unsigned long align,
408 unsigned long goal)
409{
410 return ___alloc_bootmem_nopanic(size, align, goal,
411 ARCH_LOW_ADDRESS_LIMIT);
412}
413
Yinghai Lu09325872011-02-24 14:43:05 +0100414/**
415 * __alloc_bootmem_low_node - allocate low boot memory from a specific node
416 * @pgdat: node to allocate from
417 * @size: size of the request in bytes
418 * @align: alignment of the region
419 * @goal: preferred starting address of the region
420 *
421 * The goal is dropped if it can not be satisfied and the allocation will
422 * fall back to memory below @goal.
423 *
424 * Allocation may fall back to any node in the system if the specified node
425 * can not hold the requested memory.
426 *
427 * The function panics if the request can not be satisfied.
428 */
429void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size,
430 unsigned long align, unsigned long goal)
431{
Yinghai Lu09325872011-02-24 14:43:05 +0100432 if (WARN_ON_ONCE(slab_is_available()))
433 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
434
Johannes Weinerba539862012-05-29 15:06:35 -0700435 return ___alloc_bootmem_node(pgdat, size, align, goal,
436 ARCH_LOW_ADDRESS_LIMIT);
Yinghai Lu09325872011-02-24 14:43:05 +0100437}