]>
Commit | Line | Data |
---|---|---|
09325873 YL |
1 | /* |
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> | |
b95f1b31 | 15 | #include <linux/export.h> |
09325873 YL |
16 | #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 | ||
e782ab42 YL |
26 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
27 | struct pglist_data __refdata contig_page_data; | |
28 | EXPORT_SYMBOL(contig_page_data); | |
29 | #endif | |
30 | ||
09325873 YL |
31 | unsigned long max_low_pfn; |
32 | unsigned long min_low_pfn; | |
33 | unsigned long max_pfn; | |
8dd33030 | 34 | unsigned long long max_possible_pfn; |
09325873 | 35 | |
8bc1f91e YL |
36 | static void * __init __alloc_memory_core_early(int nid, u64 size, u64 align, |
37 | u64 goal, u64 limit) | |
38 | { | |
39 | void *ptr; | |
40 | u64 addr; | |
a3f5bafc | 41 | ulong flags = choose_memblock_flags(); |
8bc1f91e YL |
42 | |
43 | if (limit > memblock.current_limit) | |
44 | limit = memblock.current_limit; | |
45 | ||
a3f5bafc | 46 | again: |
fc6daaf9 | 47 | addr = memblock_find_in_range_node(size, align, goal, limit, nid, |
a3f5bafc TL |
48 | 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 | } | |
1f5026a7 | 55 | if (!addr) |
8bc1f91e YL |
56 | return NULL; |
57 | ||
87379ec8 PH |
58 | if (memblock_reserve(addr, size)) |
59 | return NULL; | |
60 | ||
8bc1f91e YL |
61 | ptr = phys_to_virt(addr); |
62 | memset(ptr, 0, size); | |
8bc1f91e YL |
63 | /* |
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 | ||
09325873 YL |
71 | /* |
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 | */ | |
80 | void __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++) { | |
d70ddd7a | 90 | __free_pages_bootmem(pfn_to_page(cursor), cursor, 0); |
09325873 YL |
91 | totalram_pages++; |
92 | } | |
93 | } | |
94 | ||
95 | static void __init __free_pages_memory(unsigned long start, unsigned long end) | |
96 | { | |
309d0b39 | 97 | int order; |
09325873 | 98 | |
309d0b39 RH |
99 | while (start < end) { |
100 | order = min(MAX_ORDER - 1UL, __ffs(start)); | |
09325873 | 101 | |
309d0b39 RH |
102 | while (start + (1UL << order) > end) |
103 | order--; | |
09325873 | 104 | |
d70ddd7a | 105 | __free_pages_bootmem(pfn_to_page(start), start, order); |
09325873 | 106 | |
309d0b39 RH |
107 | start += (1UL << order); |
108 | } | |
09325873 YL |
109 | } |
110 | ||
29f67386 YL |
111 | static 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 | ||
b4def350 | 126 | static unsigned long __init free_low_memory_core_early(void) |
09325873 | 127 | { |
09325873 | 128 | unsigned long count = 0; |
354f17e1 | 129 | phys_addr_t start, end; |
8a9ca34c TH |
130 | u64 i; |
131 | ||
0a313a99 XQ |
132 | memblock_clear_hotplug(0, -1); |
133 | ||
92923ca3 NZ |
134 | for_each_reserved_mem_region(i, &start, &end) |
135 | reserve_bootmem_region(start, end); | |
136 | ||
fc6daaf9 TL |
137 | for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, |
138 | NULL) | |
29f67386 YL |
139 | count += __free_memory_core(start, end); |
140 | ||
5e270e25 | 141 | #ifdef CONFIG_ARCH_DISCARD_MEMBLOCK |
354f17e1 PH |
142 | { |
143 | phys_addr_t size; | |
144 | ||
145 | /* 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 | } | |
5e270e25 PH |
155 | #endif |
156 | ||
09325873 YL |
157 | return count; |
158 | } | |
159 | ||
7b4b2a0d JL |
160 | static int reset_managed_pages_done __initdata; |
161 | ||
f784a3f1 | 162 | void reset_node_managed_pages(pg_data_t *pgdat) |
9feedc9d JL |
163 | { |
164 | struct zone *z; | |
165 | ||
9feedc9d | 166 | for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) |
7b4b2a0d JL |
167 | z->managed_pages = 0; |
168 | } | |
169 | ||
170 | void __init reset_all_zones_managed_pages(void) | |
171 | { | |
172 | struct pglist_data *pgdat; | |
173 | ||
f784a3f1 TC |
174 | if (reset_managed_pages_done) |
175 | return; | |
176 | ||
7b4b2a0d JL |
177 | for_each_online_pgdat(pgdat) |
178 | reset_node_managed_pages(pgdat); | |
f784a3f1 | 179 | |
7b4b2a0d | 180 | reset_managed_pages_done = 1; |
9feedc9d JL |
181 | } |
182 | ||
09325873 YL |
183 | /** |
184 | * free_all_bootmem - release free pages to the buddy allocator | |
185 | * | |
186 | * Returns the number of pages actually released. | |
187 | */ | |
188 | unsigned long __init free_all_bootmem(void) | |
189 | { | |
0c988534 JL |
190 | unsigned long pages; |
191 | ||
7b4b2a0d | 192 | reset_all_zones_managed_pages(); |
9feedc9d | 193 | |
09325873 | 194 | /* |
b1154233 | 195 | * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id |
25985edc | 196 | * because in some case like Node0 doesn't have RAM installed |
09325873 | 197 | * low ram will be on Node1 |
09325873 | 198 | */ |
0c988534 JL |
199 | pages = free_low_memory_core_early(); |
200 | totalram_pages += pages; | |
201 | ||
202 | return pages; | |
09325873 YL |
203 | } |
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 | */ | |
215 | void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, | |
216 | unsigned long size) | |
217 | { | |
24aa0788 | 218 | memblock_free(physaddr, size); |
09325873 YL |
219 | } |
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 | */ | |
230 | void __init free_bootmem(unsigned long addr, unsigned long size) | |
231 | { | |
24aa0788 | 232 | memblock_free(addr, size); |
09325873 YL |
233 | } |
234 | ||
235 | static 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 | ||
245 | restart: | |
246 | ||
b1154233 | 247 | ptr = __alloc_memory_core_early(NUMA_NO_NODE, size, align, goal, limit); |
09325873 YL |
248 | |
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 | */ | |
273 | void * __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 | ||
281 | static 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 | */ | |
291 | printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size); | |
292 | 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 | */ | |
309 | void * __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 | ||
99ab7b19 | 317 | void * __init ___alloc_bootmem_node_nopanic(pg_data_t *pgdat, |
ba539868 JW |
318 | unsigned long size, |
319 | unsigned long align, | |
320 | unsigned long goal, | |
321 | unsigned long limit) | |
322 | { | |
323 | void *ptr; | |
324 | ||
325 | again: | |
326 | ptr = __alloc_memory_core_early(pgdat->node_id, size, align, | |
327 | goal, limit); | |
328 | if (ptr) | |
329 | return ptr; | |
330 | ||
b1154233 | 331 | ptr = __alloc_memory_core_early(NUMA_NO_NODE, size, align, |
ba539868 JW |
332 | 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 | ||
344 | void * __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 | ||
de498507 | 353 | static void * __init ___alloc_bootmem_node(pg_data_t *pgdat, unsigned long size, |
ba539868 JW |
354 | 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 | ||
363 | printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size); | |
364 | panic("Out of memory"); | |
365 | return NULL; | |
366 | } | |
367 | ||
09325873 YL |
368 | /** |
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 | */ | |
383 | void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size, | |
384 | unsigned long align, unsigned long goal) | |
385 | { | |
09325873 YL |
386 | if (WARN_ON_ONCE(slab_is_available())) |
387 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
388 | ||
ba539868 | 389 | return ___alloc_bootmem_node(pgdat, size, align, goal, 0); |
09325873 YL |
390 | } |
391 | ||
392 | void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size, | |
393 | unsigned long align, unsigned long goal) | |
394 | { | |
09325873 | 395 | return __alloc_bootmem_node(pgdat, size, align, goal); |
09325873 YL |
396 | } |
397 | ||
09325873 YL |
398 | #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 | */ | |
415 | void * __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 | ||
38fa4175 YL |
421 | void * __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 | ||
09325873 YL |
429 | /** |
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 | */ | |
444 | void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size, | |
445 | unsigned long align, unsigned long goal) | |
446 | { | |
09325873 YL |
447 | if (WARN_ON_ONCE(slab_is_available())) |
448 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
449 | ||
ba539868 JW |
450 | return ___alloc_bootmem_node(pgdat, size, align, goal, |
451 | ARCH_LOW_ADDRESS_LIMIT); | |
09325873 | 452 | } |