]>
Commit | Line | Data |
---|---|---|
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 | 19 | #include <linux/debugfs.h> |
514c6032 | 20 | #include <linux/kmemleak.h> |
6d03b885 | 21 | #include <linux/seq_file.h> |
95f72d1e YL |
22 | #include <linux/memblock.h> |
23 | ||
c4c5ad6b | 24 | #include <asm/sections.h> |
26f09e9b SS |
25 | #include <linux/io.h> |
26 | ||
27 | #include "internal.h" | |
79442ed1 | 28 | |
8a5b403d AB |
29 | #define INIT_MEMBLOCK_REGIONS 128 |
30 | #define INIT_PHYSMEM_REGIONS 4 | |
31 | ||
32 | #ifndef INIT_MEMBLOCK_RESERVED_REGIONS | |
33 | # define INIT_MEMBLOCK_RESERVED_REGIONS INIT_MEMBLOCK_REGIONS | |
34 | #endif | |
35 | ||
3e039c5c MR |
36 | /** |
37 | * DOC: memblock overview | |
38 | * | |
39 | * Memblock is a method of managing memory regions during the early | |
40 | * boot period when the usual kernel memory allocators are not up and | |
41 | * running. | |
42 | * | |
43 | * Memblock views the system memory as collections of contiguous | |
44 | * regions. There are several types of these collections: | |
45 | * | |
46 | * * ``memory`` - describes the physical memory available to the | |
47 | * kernel; this may differ from the actual physical memory installed | |
48 | * in the system, for instance when the memory is restricted with | |
49 | * ``mem=`` command line parameter | |
50 | * * ``reserved`` - describes the regions that were allocated | |
51 | * * ``physmap`` - describes the actual physical memory regardless of | |
52 | * the possible restrictions; the ``physmap`` type is only available | |
53 | * on some architectures. | |
54 | * | |
55 | * Each region is represented by :c:type:`struct memblock_region` that | |
56 | * defines the region extents, its attributes and NUMA node id on NUMA | |
57 | * systems. Every memory type is described by the :c:type:`struct | |
58 | * memblock_type` which contains an array of memory regions along with | |
59 | * the allocator metadata. The memory types are nicely wrapped with | |
60 | * :c:type:`struct memblock`. This structure is statically initialzed | |
61 | * at build time. The region arrays for the "memory" and "reserved" | |
62 | * types are initially sized to %INIT_MEMBLOCK_REGIONS and for the | |
63 | * "physmap" type to %INIT_PHYSMEM_REGIONS. | |
64 | * The :c:func:`memblock_allow_resize` enables automatic resizing of | |
65 | * the region arrays during addition of new regions. This feature | |
66 | * should be used with care so that memory allocated for the region | |
67 | * array will not overlap with areas that should be reserved, for | |
68 | * example initrd. | |
69 | * | |
70 | * The early architecture setup should tell memblock what the physical | |
71 | * memory layout is by using :c:func:`memblock_add` or | |
72 | * :c:func:`memblock_add_node` functions. The first function does not | |
73 | * assign the region to a NUMA node and it is appropriate for UMA | |
74 | * systems. Yet, it is possible to use it on NUMA systems as well and | |
75 | * assign the region to a NUMA node later in the setup process using | |
76 | * :c:func:`memblock_set_node`. The :c:func:`memblock_add_node` | |
77 | * performs such an assignment directly. | |
78 | * | |
a2974133 MR |
79 | * Once memblock is setup the memory can be allocated using one of the |
80 | * API variants: | |
81 | * | |
82 | * * :c:func:`memblock_phys_alloc*` - these functions return the | |
83 | * **physical** address of the allocated memory | |
84 | * * :c:func:`memblock_alloc*` - these functions return the **virtual** | |
85 | * address of the allocated memory. | |
86 | * | |
87 | * Note, that both API variants use implict assumptions about allowed | |
88 | * memory ranges and the fallback methods. Consult the documentation | |
89 | * of :c:func:`memblock_alloc_internal` and | |
90 | * :c:func:`memblock_alloc_range_nid` functions for more elaboarte | |
91 | * description. | |
3e039c5c MR |
92 | * |
93 | * As the system boot progresses, the architecture specific | |
94 | * :c:func:`mem_init` function frees all the memory to the buddy page | |
95 | * allocator. | |
96 | * | |
350e88ba | 97 | * Unless an architecure enables %CONFIG_ARCH_KEEP_MEMBLOCK, the |
3e039c5c MR |
98 | * memblock data structures will be discarded after the system |
99 | * initialization compltes. | |
100 | */ | |
101 | ||
bda49a81 MR |
102 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
103 | struct pglist_data __refdata contig_page_data; | |
104 | EXPORT_SYMBOL(contig_page_data); | |
105 | #endif | |
106 | ||
107 | unsigned long max_low_pfn; | |
108 | unsigned long min_low_pfn; | |
109 | unsigned long max_pfn; | |
110 | unsigned long long max_possible_pfn; | |
111 | ||
fe091c20 | 112 | static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock; |
8a5b403d | 113 | static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_RESERVED_REGIONS] __initdata_memblock; |
70210ed9 PH |
114 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
115 | static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock; | |
116 | #endif | |
fe091c20 TH |
117 | |
118 | struct memblock memblock __initdata_memblock = { | |
119 | .memory.regions = memblock_memory_init_regions, | |
120 | .memory.cnt = 1, /* empty dummy entry */ | |
121 | .memory.max = INIT_MEMBLOCK_REGIONS, | |
0262d9c8 | 122 | .memory.name = "memory", |
fe091c20 TH |
123 | |
124 | .reserved.regions = memblock_reserved_init_regions, | |
125 | .reserved.cnt = 1, /* empty dummy entry */ | |
8a5b403d | 126 | .reserved.max = INIT_MEMBLOCK_RESERVED_REGIONS, |
0262d9c8 | 127 | .reserved.name = "reserved", |
fe091c20 | 128 | |
70210ed9 PH |
129 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
130 | .physmem.regions = memblock_physmem_init_regions, | |
131 | .physmem.cnt = 1, /* empty dummy entry */ | |
132 | .physmem.max = INIT_PHYSMEM_REGIONS, | |
0262d9c8 | 133 | .physmem.name = "physmem", |
70210ed9 PH |
134 | #endif |
135 | ||
79442ed1 | 136 | .bottom_up = false, |
fe091c20 TH |
137 | .current_limit = MEMBLOCK_ALLOC_ANYWHERE, |
138 | }; | |
95f72d1e | 139 | |
10d06439 | 140 | int memblock_debug __initdata_memblock; |
a3f5bafc | 141 | static bool system_has_some_mirror __initdata_memblock = false; |
1aadc056 | 142 | static int memblock_can_resize __initdata_memblock; |
181eb394 GS |
143 | static int memblock_memory_in_slab __initdata_memblock = 0; |
144 | static int memblock_reserved_in_slab __initdata_memblock = 0; | |
95f72d1e | 145 | |
c366ea89 | 146 | static enum memblock_flags __init_memblock choose_memblock_flags(void) |
a3f5bafc TL |
147 | { |
148 | return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE; | |
149 | } | |
150 | ||
eb18f1b5 TH |
151 | /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */ |
152 | static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size) | |
153 | { | |
1c4bc43d | 154 | return *size = min(*size, PHYS_ADDR_MAX - base); |
eb18f1b5 TH |
155 | } |
156 | ||
6ed311b2 BH |
157 | /* |
158 | * Address comparison utilities | |
159 | */ | |
10d06439 | 160 | static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1, |
2898cc4c | 161 | phys_addr_t base2, phys_addr_t size2) |
95f72d1e YL |
162 | { |
163 | return ((base1 < (base2 + size2)) && (base2 < (base1 + size1))); | |
164 | } | |
165 | ||
95cf82ec | 166 | bool __init_memblock memblock_overlaps_region(struct memblock_type *type, |
2d7d3eb2 | 167 | phys_addr_t base, phys_addr_t size) |
6ed311b2 BH |
168 | { |
169 | unsigned long i; | |
170 | ||
f14516fb AK |
171 | for (i = 0; i < type->cnt; i++) |
172 | if (memblock_addrs_overlap(base, size, type->regions[i].base, | |
173 | type->regions[i].size)) | |
6ed311b2 | 174 | break; |
c5c5c9d1 | 175 | return i < type->cnt; |
6ed311b2 BH |
176 | } |
177 | ||
47cec443 | 178 | /** |
79442ed1 TC |
179 | * __memblock_find_range_bottom_up - find free area utility in bottom-up |
180 | * @start: start of candidate range | |
47cec443 MR |
181 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
182 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
79442ed1 TC |
183 | * @size: size of free area to find |
184 | * @align: alignment of free area to find | |
b1154233 | 185 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 186 | * @flags: pick from blocks based on memory attributes |
79442ed1 TC |
187 | * |
188 | * Utility called from memblock_find_in_range_node(), find free area bottom-up. | |
189 | * | |
47cec443 | 190 | * Return: |
79442ed1 TC |
191 | * Found address on success, 0 on failure. |
192 | */ | |
193 | static phys_addr_t __init_memblock | |
194 | __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 195 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 196 | enum memblock_flags flags) |
79442ed1 TC |
197 | { |
198 | phys_addr_t this_start, this_end, cand; | |
199 | u64 i; | |
200 | ||
fc6daaf9 | 201 | for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { |
79442ed1 TC |
202 | this_start = clamp(this_start, start, end); |
203 | this_end = clamp(this_end, start, end); | |
204 | ||
205 | cand = round_up(this_start, align); | |
206 | if (cand < this_end && this_end - cand >= size) | |
207 | return cand; | |
208 | } | |
209 | ||
210 | return 0; | |
211 | } | |
212 | ||
7bd0b0f0 | 213 | /** |
1402899e | 214 | * __memblock_find_range_top_down - find free area utility, in top-down |
7bd0b0f0 | 215 | * @start: start of candidate range |
47cec443 MR |
216 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
217 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
218 | * @size: size of free area to find |
219 | * @align: alignment of free area to find | |
b1154233 | 220 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 221 | * @flags: pick from blocks based on memory attributes |
7bd0b0f0 | 222 | * |
1402899e | 223 | * Utility called from memblock_find_in_range_node(), find free area top-down. |
7bd0b0f0 | 224 | * |
47cec443 | 225 | * Return: |
79442ed1 | 226 | * Found address on success, 0 on failure. |
6ed311b2 | 227 | */ |
1402899e TC |
228 | static phys_addr_t __init_memblock |
229 | __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 230 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 231 | enum memblock_flags flags) |
f7210e6c TC |
232 | { |
233 | phys_addr_t this_start, this_end, cand; | |
234 | u64 i; | |
235 | ||
fc6daaf9 TL |
236 | for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, |
237 | NULL) { | |
f7210e6c TC |
238 | this_start = clamp(this_start, start, end); |
239 | this_end = clamp(this_end, start, end); | |
240 | ||
241 | if (this_end < size) | |
242 | continue; | |
243 | ||
244 | cand = round_down(this_end - size, align); | |
245 | if (cand >= this_start) | |
246 | return cand; | |
247 | } | |
1402899e | 248 | |
f7210e6c TC |
249 | return 0; |
250 | } | |
6ed311b2 | 251 | |
1402899e TC |
252 | /** |
253 | * memblock_find_in_range_node - find free area in given range and node | |
1402899e TC |
254 | * @size: size of free area to find |
255 | * @align: alignment of free area to find | |
87029ee9 | 256 | * @start: start of candidate range |
47cec443 MR |
257 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
258 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
b1154233 | 259 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 260 | * @flags: pick from blocks based on memory attributes |
1402899e TC |
261 | * |
262 | * Find @size free area aligned to @align in the specified range and node. | |
263 | * | |
79442ed1 TC |
264 | * When allocation direction is bottom-up, the @start should be greater |
265 | * than the end of the kernel image. Otherwise, it will be trimmed. The | |
266 | * reason is that we want the bottom-up allocation just near the kernel | |
267 | * image so it is highly likely that the allocated memory and the kernel | |
268 | * will reside in the same node. | |
269 | * | |
270 | * If bottom-up allocation failed, will try to allocate memory top-down. | |
271 | * | |
47cec443 | 272 | * Return: |
79442ed1 | 273 | * Found address on success, 0 on failure. |
1402899e | 274 | */ |
c366ea89 | 275 | static phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, |
87029ee9 | 276 | phys_addr_t align, phys_addr_t start, |
e1720fee MR |
277 | phys_addr_t end, int nid, |
278 | enum memblock_flags flags) | |
1402899e | 279 | { |
0cfb8f0c | 280 | phys_addr_t kernel_end, ret; |
79442ed1 | 281 | |
1402899e | 282 | /* pump up @end */ |
fed84c78 QC |
283 | if (end == MEMBLOCK_ALLOC_ACCESSIBLE || |
284 | end == MEMBLOCK_ALLOC_KASAN) | |
1402899e TC |
285 | end = memblock.current_limit; |
286 | ||
287 | /* avoid allocating the first page */ | |
288 | start = max_t(phys_addr_t, start, PAGE_SIZE); | |
289 | end = max(start, end); | |
79442ed1 TC |
290 | kernel_end = __pa_symbol(_end); |
291 | ||
292 | /* | |
293 | * try bottom-up allocation only when bottom-up mode | |
294 | * is set and @end is above the kernel image. | |
295 | */ | |
296 | if (memblock_bottom_up() && end > kernel_end) { | |
297 | phys_addr_t bottom_up_start; | |
298 | ||
299 | /* make sure we will allocate above the kernel */ | |
300 | bottom_up_start = max(start, kernel_end); | |
301 | ||
302 | /* ok, try bottom-up allocation first */ | |
303 | ret = __memblock_find_range_bottom_up(bottom_up_start, end, | |
fc6daaf9 | 304 | size, align, nid, flags); |
79442ed1 TC |
305 | if (ret) |
306 | return ret; | |
307 | ||
308 | /* | |
309 | * we always limit bottom-up allocation above the kernel, | |
310 | * but top-down allocation doesn't have the limit, so | |
311 | * retrying top-down allocation may succeed when bottom-up | |
312 | * allocation failed. | |
313 | * | |
314 | * bottom-up allocation is expected to be fail very rarely, | |
315 | * so we use WARN_ONCE() here to see the stack trace if | |
316 | * fail happens. | |
317 | */ | |
e3d301ca MH |
318 | WARN_ONCE(IS_ENABLED(CONFIG_MEMORY_HOTREMOVE), |
319 | "memblock: bottom-up allocation failed, memory hotremove may be affected\n"); | |
79442ed1 | 320 | } |
1402899e | 321 | |
fc6daaf9 TL |
322 | return __memblock_find_range_top_down(start, end, size, align, nid, |
323 | flags); | |
1402899e TC |
324 | } |
325 | ||
7bd0b0f0 TH |
326 | /** |
327 | * memblock_find_in_range - find free area in given range | |
328 | * @start: start of candidate range | |
47cec443 MR |
329 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
330 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
331 | * @size: size of free area to find |
332 | * @align: alignment of free area to find | |
333 | * | |
334 | * Find @size free area aligned to @align in the specified range. | |
335 | * | |
47cec443 | 336 | * Return: |
79442ed1 | 337 | * Found address on success, 0 on failure. |
fc769a8e | 338 | */ |
7bd0b0f0 TH |
339 | phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, |
340 | phys_addr_t end, phys_addr_t size, | |
341 | phys_addr_t align) | |
6ed311b2 | 342 | { |
a3f5bafc | 343 | phys_addr_t ret; |
e1720fee | 344 | enum memblock_flags flags = choose_memblock_flags(); |
a3f5bafc TL |
345 | |
346 | again: | |
347 | ret = memblock_find_in_range_node(size, align, start, end, | |
348 | NUMA_NO_NODE, flags); | |
349 | ||
350 | if (!ret && (flags & MEMBLOCK_MIRROR)) { | |
351 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
352 | &size); | |
353 | flags &= ~MEMBLOCK_MIRROR; | |
354 | goto again; | |
355 | } | |
356 | ||
357 | return ret; | |
6ed311b2 BH |
358 | } |
359 | ||
10d06439 | 360 | static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) |
95f72d1e | 361 | { |
1440c4e2 | 362 | type->total_size -= type->regions[r].size; |
7c0caeb8 TH |
363 | memmove(&type->regions[r], &type->regions[r + 1], |
364 | (type->cnt - (r + 1)) * sizeof(type->regions[r])); | |
e3239ff9 | 365 | type->cnt--; |
95f72d1e | 366 | |
8f7a6605 BH |
367 | /* Special case for empty arrays */ |
368 | if (type->cnt == 0) { | |
1440c4e2 | 369 | WARN_ON(type->total_size != 0); |
8f7a6605 BH |
370 | type->cnt = 1; |
371 | type->regions[0].base = 0; | |
372 | type->regions[0].size = 0; | |
66a20757 | 373 | type->regions[0].flags = 0; |
7c0caeb8 | 374 | memblock_set_region_node(&type->regions[0], MAX_NUMNODES); |
8f7a6605 | 375 | } |
95f72d1e YL |
376 | } |
377 | ||
350e88ba | 378 | #ifndef CONFIG_ARCH_KEEP_MEMBLOCK |
3010f876 | 379 | /** |
47cec443 | 380 | * memblock_discard - discard memory and reserved arrays if they were allocated |
3010f876 PT |
381 | */ |
382 | void __init memblock_discard(void) | |
5e270e25 | 383 | { |
3010f876 | 384 | phys_addr_t addr, size; |
5e270e25 | 385 | |
3010f876 PT |
386 | if (memblock.reserved.regions != memblock_reserved_init_regions) { |
387 | addr = __pa(memblock.reserved.regions); | |
388 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
389 | memblock.reserved.max); | |
390 | __memblock_free_late(addr, size); | |
391 | } | |
5e270e25 | 392 | |
91b540f9 | 393 | if (memblock.memory.regions != memblock_memory_init_regions) { |
3010f876 PT |
394 | addr = __pa(memblock.memory.regions); |
395 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
396 | memblock.memory.max); | |
397 | __memblock_free_late(addr, size); | |
398 | } | |
5e270e25 | 399 | } |
5e270e25 PH |
400 | #endif |
401 | ||
48c3b583 GP |
402 | /** |
403 | * memblock_double_array - double the size of the memblock regions array | |
404 | * @type: memblock type of the regions array being doubled | |
405 | * @new_area_start: starting address of memory range to avoid overlap with | |
406 | * @new_area_size: size of memory range to avoid overlap with | |
407 | * | |
408 | * Double the size of the @type regions array. If memblock is being used to | |
409 | * allocate memory for a new reserved regions array and there is a previously | |
47cec443 | 410 | * allocated memory range [@new_area_start, @new_area_start + @new_area_size] |
48c3b583 GP |
411 | * waiting to be reserved, ensure the memory used by the new array does |
412 | * not overlap. | |
413 | * | |
47cec443 | 414 | * Return: |
48c3b583 GP |
415 | * 0 on success, -1 on failure. |
416 | */ | |
417 | static int __init_memblock memblock_double_array(struct memblock_type *type, | |
418 | phys_addr_t new_area_start, | |
419 | phys_addr_t new_area_size) | |
142b45a7 BH |
420 | { |
421 | struct memblock_region *new_array, *old_array; | |
29f67386 | 422 | phys_addr_t old_alloc_size, new_alloc_size; |
a36aab89 | 423 | phys_addr_t old_size, new_size, addr, new_end; |
142b45a7 | 424 | int use_slab = slab_is_available(); |
181eb394 | 425 | int *in_slab; |
142b45a7 BH |
426 | |
427 | /* We don't allow resizing until we know about the reserved regions | |
428 | * of memory that aren't suitable for allocation | |
429 | */ | |
430 | if (!memblock_can_resize) | |
431 | return -1; | |
432 | ||
142b45a7 BH |
433 | /* Calculate new doubled size */ |
434 | old_size = type->max * sizeof(struct memblock_region); | |
435 | new_size = old_size << 1; | |
29f67386 YL |
436 | /* |
437 | * We need to allocated new one align to PAGE_SIZE, | |
438 | * so we can free them completely later. | |
439 | */ | |
440 | old_alloc_size = PAGE_ALIGN(old_size); | |
441 | new_alloc_size = PAGE_ALIGN(new_size); | |
142b45a7 | 442 | |
181eb394 GS |
443 | /* Retrieve the slab flag */ |
444 | if (type == &memblock.memory) | |
445 | in_slab = &memblock_memory_in_slab; | |
446 | else | |
447 | in_slab = &memblock_reserved_in_slab; | |
448 | ||
a2974133 | 449 | /* Try to find some space for it */ |
142b45a7 BH |
450 | if (use_slab) { |
451 | new_array = kmalloc(new_size, GFP_KERNEL); | |
1f5026a7 | 452 | addr = new_array ? __pa(new_array) : 0; |
4e2f0775 | 453 | } else { |
48c3b583 GP |
454 | /* only exclude range when trying to double reserved.regions */ |
455 | if (type != &memblock.reserved) | |
456 | new_area_start = new_area_size = 0; | |
457 | ||
458 | addr = memblock_find_in_range(new_area_start + new_area_size, | |
459 | memblock.current_limit, | |
29f67386 | 460 | new_alloc_size, PAGE_SIZE); |
48c3b583 GP |
461 | if (!addr && new_area_size) |
462 | addr = memblock_find_in_range(0, | |
fd07383b AM |
463 | min(new_area_start, memblock.current_limit), |
464 | new_alloc_size, PAGE_SIZE); | |
48c3b583 | 465 | |
15674868 | 466 | new_array = addr ? __va(addr) : NULL; |
4e2f0775 | 467 | } |
1f5026a7 | 468 | if (!addr) { |
142b45a7 | 469 | pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", |
0262d9c8 | 470 | type->name, type->max, type->max * 2); |
142b45a7 BH |
471 | return -1; |
472 | } | |
142b45a7 | 473 | |
a36aab89 MR |
474 | new_end = addr + new_size - 1; |
475 | memblock_dbg("memblock: %s is doubled to %ld at [%pa-%pa]", | |
476 | type->name, type->max * 2, &addr, &new_end); | |
ea9e4376 | 477 | |
fd07383b AM |
478 | /* |
479 | * Found space, we now need to move the array over before we add the | |
480 | * reserved region since it may be our reserved array itself that is | |
481 | * full. | |
142b45a7 BH |
482 | */ |
483 | memcpy(new_array, type->regions, old_size); | |
484 | memset(new_array + type->max, 0, old_size); | |
485 | old_array = type->regions; | |
486 | type->regions = new_array; | |
487 | type->max <<= 1; | |
488 | ||
fd07383b | 489 | /* Free old array. We needn't free it if the array is the static one */ |
181eb394 GS |
490 | if (*in_slab) |
491 | kfree(old_array); | |
492 | else if (old_array != memblock_memory_init_regions && | |
493 | old_array != memblock_reserved_init_regions) | |
29f67386 | 494 | memblock_free(__pa(old_array), old_alloc_size); |
142b45a7 | 495 | |
fd07383b AM |
496 | /* |
497 | * Reserve the new array if that comes from the memblock. Otherwise, we | |
498 | * needn't do it | |
181eb394 GS |
499 | */ |
500 | if (!use_slab) | |
29f67386 | 501 | BUG_ON(memblock_reserve(addr, new_alloc_size)); |
181eb394 GS |
502 | |
503 | /* Update slab flag */ | |
504 | *in_slab = use_slab; | |
505 | ||
142b45a7 BH |
506 | return 0; |
507 | } | |
508 | ||
784656f9 TH |
509 | /** |
510 | * memblock_merge_regions - merge neighboring compatible regions | |
511 | * @type: memblock type to scan | |
512 | * | |
513 | * Scan @type and merge neighboring compatible regions. | |
514 | */ | |
515 | static void __init_memblock memblock_merge_regions(struct memblock_type *type) | |
95f72d1e | 516 | { |
784656f9 | 517 | int i = 0; |
95f72d1e | 518 | |
784656f9 TH |
519 | /* cnt never goes below 1 */ |
520 | while (i < type->cnt - 1) { | |
521 | struct memblock_region *this = &type->regions[i]; | |
522 | struct memblock_region *next = &type->regions[i + 1]; | |
95f72d1e | 523 | |
7c0caeb8 TH |
524 | if (this->base + this->size != next->base || |
525 | memblock_get_region_node(this) != | |
66a20757 TC |
526 | memblock_get_region_node(next) || |
527 | this->flags != next->flags) { | |
784656f9 TH |
528 | BUG_ON(this->base + this->size > next->base); |
529 | i++; | |
530 | continue; | |
8f7a6605 BH |
531 | } |
532 | ||
784656f9 | 533 | this->size += next->size; |
c0232ae8 LF |
534 | /* move forward from next + 1, index of which is i + 2 */ |
535 | memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); | |
784656f9 | 536 | type->cnt--; |
95f72d1e | 537 | } |
784656f9 | 538 | } |
95f72d1e | 539 | |
784656f9 TH |
540 | /** |
541 | * memblock_insert_region - insert new memblock region | |
209ff86d TC |
542 | * @type: memblock type to insert into |
543 | * @idx: index for the insertion point | |
544 | * @base: base address of the new region | |
545 | * @size: size of the new region | |
546 | * @nid: node id of the new region | |
66a20757 | 547 | * @flags: flags of the new region |
784656f9 | 548 | * |
47cec443 | 549 | * Insert new memblock region [@base, @base + @size) into @type at @idx. |
412d0008 | 550 | * @type must already have extra room to accommodate the new region. |
784656f9 TH |
551 | */ |
552 | static void __init_memblock memblock_insert_region(struct memblock_type *type, | |
553 | int idx, phys_addr_t base, | |
66a20757 | 554 | phys_addr_t size, |
e1720fee MR |
555 | int nid, |
556 | enum memblock_flags flags) | |
784656f9 TH |
557 | { |
558 | struct memblock_region *rgn = &type->regions[idx]; | |
559 | ||
560 | BUG_ON(type->cnt >= type->max); | |
561 | memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); | |
562 | rgn->base = base; | |
563 | rgn->size = size; | |
66a20757 | 564 | rgn->flags = flags; |
7c0caeb8 | 565 | memblock_set_region_node(rgn, nid); |
784656f9 | 566 | type->cnt++; |
1440c4e2 | 567 | type->total_size += size; |
784656f9 TH |
568 | } |
569 | ||
570 | /** | |
f1af9d3a | 571 | * memblock_add_range - add new memblock region |
784656f9 TH |
572 | * @type: memblock type to add new region into |
573 | * @base: base address of the new region | |
574 | * @size: size of the new region | |
7fb0bc3f | 575 | * @nid: nid of the new region |
66a20757 | 576 | * @flags: flags of the new region |
784656f9 | 577 | * |
47cec443 | 578 | * Add new memblock region [@base, @base + @size) into @type. The new region |
784656f9 TH |
579 | * is allowed to overlap with existing ones - overlaps don't affect already |
580 | * existing regions. @type is guaranteed to be minimal (all neighbouring | |
581 | * compatible regions are merged) after the addition. | |
582 | * | |
47cec443 | 583 | * Return: |
784656f9 TH |
584 | * 0 on success, -errno on failure. |
585 | */ | |
f1af9d3a | 586 | int __init_memblock memblock_add_range(struct memblock_type *type, |
66a20757 | 587 | phys_addr_t base, phys_addr_t size, |
e1720fee | 588 | int nid, enum memblock_flags flags) |
784656f9 TH |
589 | { |
590 | bool insert = false; | |
eb18f1b5 TH |
591 | phys_addr_t obase = base; |
592 | phys_addr_t end = base + memblock_cap_size(base, &size); | |
8c9c1701 AK |
593 | int idx, nr_new; |
594 | struct memblock_region *rgn; | |
784656f9 | 595 | |
b3dc627c TH |
596 | if (!size) |
597 | return 0; | |
598 | ||
784656f9 TH |
599 | /* special case for empty array */ |
600 | if (type->regions[0].size == 0) { | |
1440c4e2 | 601 | WARN_ON(type->cnt != 1 || type->total_size); |
8f7a6605 BH |
602 | type->regions[0].base = base; |
603 | type->regions[0].size = size; | |
66a20757 | 604 | type->regions[0].flags = flags; |
7fb0bc3f | 605 | memblock_set_region_node(&type->regions[0], nid); |
1440c4e2 | 606 | type->total_size = size; |
8f7a6605 | 607 | return 0; |
95f72d1e | 608 | } |
784656f9 TH |
609 | repeat: |
610 | /* | |
611 | * The following is executed twice. Once with %false @insert and | |
612 | * then with %true. The first counts the number of regions needed | |
412d0008 | 613 | * to accommodate the new area. The second actually inserts them. |
142b45a7 | 614 | */ |
784656f9 TH |
615 | base = obase; |
616 | nr_new = 0; | |
95f72d1e | 617 | |
66e8b438 | 618 | for_each_memblock_type(idx, type, rgn) { |
784656f9 TH |
619 | phys_addr_t rbase = rgn->base; |
620 | phys_addr_t rend = rbase + rgn->size; | |
621 | ||
622 | if (rbase >= end) | |
95f72d1e | 623 | break; |
784656f9 TH |
624 | if (rend <= base) |
625 | continue; | |
626 | /* | |
627 | * @rgn overlaps. If it separates the lower part of new | |
628 | * area, insert that portion. | |
629 | */ | |
630 | if (rbase > base) { | |
c0a29498 WY |
631 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
632 | WARN_ON(nid != memblock_get_region_node(rgn)); | |
633 | #endif | |
4fcab5f4 | 634 | WARN_ON(flags != rgn->flags); |
784656f9 TH |
635 | nr_new++; |
636 | if (insert) | |
8c9c1701 | 637 | memblock_insert_region(type, idx++, base, |
66a20757 TC |
638 | rbase - base, nid, |
639 | flags); | |
95f72d1e | 640 | } |
784656f9 TH |
641 | /* area below @rend is dealt with, forget about it */ |
642 | base = min(rend, end); | |
95f72d1e | 643 | } |
784656f9 TH |
644 | |
645 | /* insert the remaining portion */ | |
646 | if (base < end) { | |
647 | nr_new++; | |
648 | if (insert) | |
8c9c1701 | 649 | memblock_insert_region(type, idx, base, end - base, |
66a20757 | 650 | nid, flags); |
95f72d1e | 651 | } |
95f72d1e | 652 | |
ef3cc4db | 653 | if (!nr_new) |
654 | return 0; | |
655 | ||
784656f9 TH |
656 | /* |
657 | * If this was the first round, resize array and repeat for actual | |
658 | * insertions; otherwise, merge and return. | |
142b45a7 | 659 | */ |
784656f9 TH |
660 | if (!insert) { |
661 | while (type->cnt + nr_new > type->max) | |
48c3b583 | 662 | if (memblock_double_array(type, obase, size) < 0) |
784656f9 TH |
663 | return -ENOMEM; |
664 | insert = true; | |
665 | goto repeat; | |
666 | } else { | |
667 | memblock_merge_regions(type); | |
668 | return 0; | |
142b45a7 | 669 | } |
95f72d1e YL |
670 | } |
671 | ||
48a833cc MR |
672 | /** |
673 | * memblock_add_node - add new memblock region within a NUMA node | |
674 | * @base: base address of the new region | |
675 | * @size: size of the new region | |
676 | * @nid: nid of the new region | |
677 | * | |
678 | * Add new memblock region [@base, @base + @size) to the "memory" | |
679 | * type. See memblock_add_range() description for mode details | |
680 | * | |
681 | * Return: | |
682 | * 0 on success, -errno on failure. | |
683 | */ | |
7fb0bc3f TH |
684 | int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, |
685 | int nid) | |
686 | { | |
f1af9d3a | 687 | return memblock_add_range(&memblock.memory, base, size, nid, 0); |
7fb0bc3f TH |
688 | } |
689 | ||
48a833cc MR |
690 | /** |
691 | * memblock_add - add new memblock region | |
692 | * @base: base address of the new region | |
693 | * @size: size of the new region | |
694 | * | |
695 | * Add new memblock region [@base, @base + @size) to the "memory" | |
696 | * type. See memblock_add_range() description for mode details | |
697 | * | |
698 | * Return: | |
699 | * 0 on success, -errno on failure. | |
700 | */ | |
f705ac4b | 701 | int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) |
6a4055bc | 702 | { |
5d63f81c MC |
703 | phys_addr_t end = base + size - 1; |
704 | ||
d75f773c | 705 | memblock_dbg("memblock_add: [%pa-%pa] %pS\n", |
5d63f81c | 706 | &base, &end, (void *)_RET_IP_); |
6a4055bc | 707 | |
f705ac4b | 708 | return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
709 | } |
710 | ||
6a9ceb31 TH |
711 | /** |
712 | * memblock_isolate_range - isolate given range into disjoint memblocks | |
713 | * @type: memblock type to isolate range for | |
714 | * @base: base of range to isolate | |
715 | * @size: size of range to isolate | |
716 | * @start_rgn: out parameter for the start of isolated region | |
717 | * @end_rgn: out parameter for the end of isolated region | |
718 | * | |
719 | * Walk @type and ensure that regions don't cross the boundaries defined by | |
47cec443 | 720 | * [@base, @base + @size). Crossing regions are split at the boundaries, |
6a9ceb31 TH |
721 | * which may create at most two more regions. The index of the first |
722 | * region inside the range is returned in *@start_rgn and end in *@end_rgn. | |
723 | * | |
47cec443 | 724 | * Return: |
6a9ceb31 TH |
725 | * 0 on success, -errno on failure. |
726 | */ | |
727 | static int __init_memblock memblock_isolate_range(struct memblock_type *type, | |
728 | phys_addr_t base, phys_addr_t size, | |
729 | int *start_rgn, int *end_rgn) | |
730 | { | |
eb18f1b5 | 731 | phys_addr_t end = base + memblock_cap_size(base, &size); |
8c9c1701 AK |
732 | int idx; |
733 | struct memblock_region *rgn; | |
6a9ceb31 TH |
734 | |
735 | *start_rgn = *end_rgn = 0; | |
736 | ||
b3dc627c TH |
737 | if (!size) |
738 | return 0; | |
739 | ||
6a9ceb31 TH |
740 | /* we'll create at most two more regions */ |
741 | while (type->cnt + 2 > type->max) | |
48c3b583 | 742 | if (memblock_double_array(type, base, size) < 0) |
6a9ceb31 TH |
743 | return -ENOMEM; |
744 | ||
66e8b438 | 745 | for_each_memblock_type(idx, type, rgn) { |
6a9ceb31 TH |
746 | phys_addr_t rbase = rgn->base; |
747 | phys_addr_t rend = rbase + rgn->size; | |
748 | ||
749 | if (rbase >= end) | |
750 | break; | |
751 | if (rend <= base) | |
752 | continue; | |
753 | ||
754 | if (rbase < base) { | |
755 | /* | |
756 | * @rgn intersects from below. Split and continue | |
757 | * to process the next region - the new top half. | |
758 | */ | |
759 | rgn->base = base; | |
1440c4e2 TH |
760 | rgn->size -= base - rbase; |
761 | type->total_size -= base - rbase; | |
8c9c1701 | 762 | memblock_insert_region(type, idx, rbase, base - rbase, |
66a20757 TC |
763 | memblock_get_region_node(rgn), |
764 | rgn->flags); | |
6a9ceb31 TH |
765 | } else if (rend > end) { |
766 | /* | |
767 | * @rgn intersects from above. Split and redo the | |
768 | * current region - the new bottom half. | |
769 | */ | |
770 | rgn->base = end; | |
1440c4e2 TH |
771 | rgn->size -= end - rbase; |
772 | type->total_size -= end - rbase; | |
8c9c1701 | 773 | memblock_insert_region(type, idx--, rbase, end - rbase, |
66a20757 TC |
774 | memblock_get_region_node(rgn), |
775 | rgn->flags); | |
6a9ceb31 TH |
776 | } else { |
777 | /* @rgn is fully contained, record it */ | |
778 | if (!*end_rgn) | |
8c9c1701 AK |
779 | *start_rgn = idx; |
780 | *end_rgn = idx + 1; | |
6a9ceb31 TH |
781 | } |
782 | } | |
783 | ||
784 | return 0; | |
785 | } | |
6a9ceb31 | 786 | |
35bd16a2 | 787 | static int __init_memblock memblock_remove_range(struct memblock_type *type, |
f1af9d3a | 788 | phys_addr_t base, phys_addr_t size) |
95f72d1e | 789 | { |
71936180 TH |
790 | int start_rgn, end_rgn; |
791 | int i, ret; | |
95f72d1e | 792 | |
71936180 TH |
793 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
794 | if (ret) | |
795 | return ret; | |
95f72d1e | 796 | |
71936180 TH |
797 | for (i = end_rgn - 1; i >= start_rgn; i--) |
798 | memblock_remove_region(type, i); | |
8f7a6605 | 799 | return 0; |
95f72d1e YL |
800 | } |
801 | ||
581adcbe | 802 | int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) |
95f72d1e | 803 | { |
25cf23d7 MK |
804 | phys_addr_t end = base + size - 1; |
805 | ||
806 | memblock_dbg("memblock_remove: [%pa-%pa] %pS\n", | |
807 | &base, &end, (void *)_RET_IP_); | |
808 | ||
f1af9d3a | 809 | return memblock_remove_range(&memblock.memory, base, size); |
95f72d1e YL |
810 | } |
811 | ||
4d72868c MR |
812 | /** |
813 | * memblock_free - free boot memory block | |
814 | * @base: phys starting address of the boot memory block | |
815 | * @size: size of the boot memory block in bytes | |
816 | * | |
817 | * Free boot memory block previously allocated by memblock_alloc_xx() API. | |
818 | * The freeing memory will not be released to the buddy allocator. | |
819 | */ | |
581adcbe | 820 | int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size) |
95f72d1e | 821 | { |
5d63f81c MC |
822 | phys_addr_t end = base + size - 1; |
823 | ||
d75f773c | 824 | memblock_dbg(" memblock_free: [%pa-%pa] %pS\n", |
5d63f81c | 825 | &base, &end, (void *)_RET_IP_); |
24aa0788 | 826 | |
9099daed | 827 | kmemleak_free_part_phys(base, size); |
f1af9d3a | 828 | return memblock_remove_range(&memblock.reserved, base, size); |
95f72d1e YL |
829 | } |
830 | ||
f705ac4b | 831 | int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size) |
95f72d1e | 832 | { |
5d63f81c MC |
833 | phys_addr_t end = base + size - 1; |
834 | ||
d75f773c | 835 | memblock_dbg("memblock_reserve: [%pa-%pa] %pS\n", |
5d63f81c | 836 | &base, &end, (void *)_RET_IP_); |
95f72d1e | 837 | |
f705ac4b | 838 | return memblock_add_range(&memblock.reserved, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
839 | } |
840 | ||
66b16edf | 841 | /** |
47cec443 MR |
842 | * memblock_setclr_flag - set or clear flag for a memory region |
843 | * @base: base address of the region | |
844 | * @size: size of the region | |
845 | * @set: set or clear the flag | |
846 | * @flag: the flag to udpate | |
66b16edf | 847 | * |
4308ce17 | 848 | * This function isolates region [@base, @base + @size), and sets/clears flag |
66b16edf | 849 | * |
47cec443 | 850 | * Return: 0 on success, -errno on failure. |
66b16edf | 851 | */ |
4308ce17 TL |
852 | static int __init_memblock memblock_setclr_flag(phys_addr_t base, |
853 | phys_addr_t size, int set, int flag) | |
66b16edf TC |
854 | { |
855 | struct memblock_type *type = &memblock.memory; | |
856 | int i, ret, start_rgn, end_rgn; | |
857 | ||
858 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); | |
859 | if (ret) | |
860 | return ret; | |
861 | ||
fe145124 MR |
862 | for (i = start_rgn; i < end_rgn; i++) { |
863 | struct memblock_region *r = &type->regions[i]; | |
864 | ||
4308ce17 | 865 | if (set) |
fe145124 | 866 | r->flags |= flag; |
4308ce17 | 867 | else |
fe145124 MR |
868 | r->flags &= ~flag; |
869 | } | |
66b16edf TC |
870 | |
871 | memblock_merge_regions(type); | |
872 | return 0; | |
873 | } | |
874 | ||
875 | /** | |
4308ce17 | 876 | * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. |
66b16edf TC |
877 | * @base: the base phys addr of the region |
878 | * @size: the size of the region | |
879 | * | |
47cec443 | 880 | * Return: 0 on success, -errno on failure. |
4308ce17 TL |
881 | */ |
882 | int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) | |
883 | { | |
884 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); | |
885 | } | |
886 | ||
887 | /** | |
888 | * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. | |
889 | * @base: the base phys addr of the region | |
890 | * @size: the size of the region | |
66b16edf | 891 | * |
47cec443 | 892 | * Return: 0 on success, -errno on failure. |
66b16edf TC |
893 | */ |
894 | int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) | |
895 | { | |
4308ce17 | 896 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); |
66b16edf TC |
897 | } |
898 | ||
a3f5bafc TL |
899 | /** |
900 | * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. | |
901 | * @base: the base phys addr of the region | |
902 | * @size: the size of the region | |
903 | * | |
47cec443 | 904 | * Return: 0 on success, -errno on failure. |
a3f5bafc TL |
905 | */ |
906 | int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) | |
907 | { | |
908 | system_has_some_mirror = true; | |
909 | ||
910 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); | |
911 | } | |
912 | ||
bf3d3cc5 AB |
913 | /** |
914 | * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. | |
915 | * @base: the base phys addr of the region | |
916 | * @size: the size of the region | |
917 | * | |
47cec443 | 918 | * Return: 0 on success, -errno on failure. |
bf3d3cc5 AB |
919 | */ |
920 | int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) | |
921 | { | |
922 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); | |
923 | } | |
a3f5bafc | 924 | |
4c546b8a AT |
925 | /** |
926 | * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. | |
927 | * @base: the base phys addr of the region | |
928 | * @size: the size of the region | |
929 | * | |
47cec443 | 930 | * Return: 0 on success, -errno on failure. |
4c546b8a AT |
931 | */ |
932 | int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) | |
933 | { | |
934 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); | |
935 | } | |
936 | ||
8e7a7f86 RH |
937 | /** |
938 | * __next_reserved_mem_region - next function for for_each_reserved_region() | |
939 | * @idx: pointer to u64 loop variable | |
940 | * @out_start: ptr to phys_addr_t for start address of the region, can be %NULL | |
941 | * @out_end: ptr to phys_addr_t for end address of the region, can be %NULL | |
942 | * | |
943 | * Iterate over all reserved memory regions. | |
944 | */ | |
945 | void __init_memblock __next_reserved_mem_region(u64 *idx, | |
946 | phys_addr_t *out_start, | |
947 | phys_addr_t *out_end) | |
948 | { | |
567d117b | 949 | struct memblock_type *type = &memblock.reserved; |
8e7a7f86 | 950 | |
cd33a76b | 951 | if (*idx < type->cnt) { |
567d117b | 952 | struct memblock_region *r = &type->regions[*idx]; |
8e7a7f86 RH |
953 | phys_addr_t base = r->base; |
954 | phys_addr_t size = r->size; | |
955 | ||
956 | if (out_start) | |
957 | *out_start = base; | |
958 | if (out_end) | |
959 | *out_end = base + size - 1; | |
960 | ||
961 | *idx += 1; | |
962 | return; | |
963 | } | |
964 | ||
965 | /* signal end of iteration */ | |
966 | *idx = ULLONG_MAX; | |
967 | } | |
968 | ||
c9a688a3 MR |
969 | static bool should_skip_region(struct memblock_region *m, int nid, int flags) |
970 | { | |
971 | int m_nid = memblock_get_region_node(m); | |
972 | ||
973 | /* only memory regions are associated with nodes, check it */ | |
974 | if (nid != NUMA_NO_NODE && nid != m_nid) | |
975 | return true; | |
976 | ||
977 | /* skip hotpluggable memory regions if needed */ | |
978 | if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) | |
979 | return true; | |
980 | ||
981 | /* if we want mirror memory skip non-mirror memory regions */ | |
982 | if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) | |
983 | return true; | |
984 | ||
985 | /* skip nomap memory unless we were asked for it explicitly */ | |
986 | if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) | |
987 | return true; | |
988 | ||
989 | return false; | |
990 | } | |
991 | ||
35fd0808 | 992 | /** |
a2974133 | 993 | * __next_mem_range - next function for for_each_free_mem_range() etc. |
35fd0808 | 994 | * @idx: pointer to u64 loop variable |
b1154233 | 995 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 996 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
997 | * @type_a: pointer to memblock_type from where the range is taken |
998 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
999 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
1000 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
1001 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
35fd0808 | 1002 | * |
f1af9d3a | 1003 | * Find the first area from *@idx which matches @nid, fill the out |
35fd0808 | 1004 | * parameters, and update *@idx for the next iteration. The lower 32bit of |
f1af9d3a PH |
1005 | * *@idx contains index into type_a and the upper 32bit indexes the |
1006 | * areas before each region in type_b. For example, if type_b regions | |
35fd0808 TH |
1007 | * look like the following, |
1008 | * | |
1009 | * 0:[0-16), 1:[32-48), 2:[128-130) | |
1010 | * | |
1011 | * The upper 32bit indexes the following regions. | |
1012 | * | |
1013 | * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) | |
1014 | * | |
1015 | * As both region arrays are sorted, the function advances the two indices | |
1016 | * in lockstep and returns each intersection. | |
1017 | */ | |
e1720fee MR |
1018 | void __init_memblock __next_mem_range(u64 *idx, int nid, |
1019 | enum memblock_flags flags, | |
f1af9d3a PH |
1020 | struct memblock_type *type_a, |
1021 | struct memblock_type *type_b, | |
1022 | phys_addr_t *out_start, | |
1023 | phys_addr_t *out_end, int *out_nid) | |
35fd0808 | 1024 | { |
f1af9d3a PH |
1025 | int idx_a = *idx & 0xffffffff; |
1026 | int idx_b = *idx >> 32; | |
b1154233 | 1027 | |
f1af9d3a PH |
1028 | if (WARN_ONCE(nid == MAX_NUMNODES, |
1029 | "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) | |
560dca27 | 1030 | nid = NUMA_NO_NODE; |
35fd0808 | 1031 | |
f1af9d3a PH |
1032 | for (; idx_a < type_a->cnt; idx_a++) { |
1033 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1034 | ||
35fd0808 TH |
1035 | phys_addr_t m_start = m->base; |
1036 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1037 | int m_nid = memblock_get_region_node(m); |
35fd0808 | 1038 | |
c9a688a3 | 1039 | if (should_skip_region(m, nid, flags)) |
bf3d3cc5 AB |
1040 | continue; |
1041 | ||
f1af9d3a PH |
1042 | if (!type_b) { |
1043 | if (out_start) | |
1044 | *out_start = m_start; | |
1045 | if (out_end) | |
1046 | *out_end = m_end; | |
1047 | if (out_nid) | |
1048 | *out_nid = m_nid; | |
1049 | idx_a++; | |
1050 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
1051 | return; | |
1052 | } | |
1053 | ||
1054 | /* scan areas before each reservation */ | |
1055 | for (; idx_b < type_b->cnt + 1; idx_b++) { | |
1056 | struct memblock_region *r; | |
1057 | phys_addr_t r_start; | |
1058 | phys_addr_t r_end; | |
1059 | ||
1060 | r = &type_b->regions[idx_b]; | |
1061 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1062 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1063 | r->base : PHYS_ADDR_MAX; |
35fd0808 | 1064 | |
f1af9d3a PH |
1065 | /* |
1066 | * if idx_b advanced past idx_a, | |
1067 | * break out to advance idx_a | |
1068 | */ | |
35fd0808 TH |
1069 | if (r_start >= m_end) |
1070 | break; | |
1071 | /* if the two regions intersect, we're done */ | |
1072 | if (m_start < r_end) { | |
1073 | if (out_start) | |
f1af9d3a PH |
1074 | *out_start = |
1075 | max(m_start, r_start); | |
35fd0808 TH |
1076 | if (out_end) |
1077 | *out_end = min(m_end, r_end); | |
1078 | if (out_nid) | |
f1af9d3a | 1079 | *out_nid = m_nid; |
35fd0808 | 1080 | /* |
f1af9d3a PH |
1081 | * The region which ends first is |
1082 | * advanced for the next iteration. | |
35fd0808 TH |
1083 | */ |
1084 | if (m_end <= r_end) | |
f1af9d3a | 1085 | idx_a++; |
35fd0808 | 1086 | else |
f1af9d3a PH |
1087 | idx_b++; |
1088 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
35fd0808 TH |
1089 | return; |
1090 | } | |
1091 | } | |
1092 | } | |
1093 | ||
1094 | /* signal end of iteration */ | |
1095 | *idx = ULLONG_MAX; | |
1096 | } | |
1097 | ||
7bd0b0f0 | 1098 | /** |
f1af9d3a PH |
1099 | * __next_mem_range_rev - generic next function for for_each_*_range_rev() |
1100 | * | |
7bd0b0f0 | 1101 | * @idx: pointer to u64 loop variable |
ad5ea8cd | 1102 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 1103 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
1104 | * @type_a: pointer to memblock_type from where the range is taken |
1105 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
1106 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
1107 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
1108 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
7bd0b0f0 | 1109 | * |
47cec443 MR |
1110 | * Finds the next range from type_a which is not marked as unsuitable |
1111 | * in type_b. | |
1112 | * | |
f1af9d3a | 1113 | * Reverse of __next_mem_range(). |
7bd0b0f0 | 1114 | */ |
e1720fee MR |
1115 | void __init_memblock __next_mem_range_rev(u64 *idx, int nid, |
1116 | enum memblock_flags flags, | |
f1af9d3a PH |
1117 | struct memblock_type *type_a, |
1118 | struct memblock_type *type_b, | |
1119 | phys_addr_t *out_start, | |
1120 | phys_addr_t *out_end, int *out_nid) | |
7bd0b0f0 | 1121 | { |
f1af9d3a PH |
1122 | int idx_a = *idx & 0xffffffff; |
1123 | int idx_b = *idx >> 32; | |
b1154233 | 1124 | |
560dca27 GS |
1125 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1126 | nid = NUMA_NO_NODE; | |
7bd0b0f0 TH |
1127 | |
1128 | if (*idx == (u64)ULLONG_MAX) { | |
f1af9d3a | 1129 | idx_a = type_a->cnt - 1; |
e47608ab | 1130 | if (type_b != NULL) |
1131 | idx_b = type_b->cnt; | |
1132 | else | |
1133 | idx_b = 0; | |
7bd0b0f0 TH |
1134 | } |
1135 | ||
f1af9d3a PH |
1136 | for (; idx_a >= 0; idx_a--) { |
1137 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1138 | ||
7bd0b0f0 TH |
1139 | phys_addr_t m_start = m->base; |
1140 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1141 | int m_nid = memblock_get_region_node(m); |
7bd0b0f0 | 1142 | |
c9a688a3 | 1143 | if (should_skip_region(m, nid, flags)) |
bf3d3cc5 AB |
1144 | continue; |
1145 | ||
f1af9d3a PH |
1146 | if (!type_b) { |
1147 | if (out_start) | |
1148 | *out_start = m_start; | |
1149 | if (out_end) | |
1150 | *out_end = m_end; | |
1151 | if (out_nid) | |
1152 | *out_nid = m_nid; | |
fb399b48 | 1153 | idx_a--; |
f1af9d3a PH |
1154 | *idx = (u32)idx_a | (u64)idx_b << 32; |
1155 | return; | |
1156 | } | |
1157 | ||
1158 | /* scan areas before each reservation */ | |
1159 | for (; idx_b >= 0; idx_b--) { | |
1160 | struct memblock_region *r; | |
1161 | phys_addr_t r_start; | |
1162 | phys_addr_t r_end; | |
1163 | ||
1164 | r = &type_b->regions[idx_b]; | |
1165 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1166 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1167 | r->base : PHYS_ADDR_MAX; |
f1af9d3a PH |
1168 | /* |
1169 | * if idx_b advanced past idx_a, | |
1170 | * break out to advance idx_a | |
1171 | */ | |
7bd0b0f0 | 1172 | |
7bd0b0f0 TH |
1173 | if (r_end <= m_start) |
1174 | break; | |
1175 | /* if the two regions intersect, we're done */ | |
1176 | if (m_end > r_start) { | |
1177 | if (out_start) | |
1178 | *out_start = max(m_start, r_start); | |
1179 | if (out_end) | |
1180 | *out_end = min(m_end, r_end); | |
1181 | if (out_nid) | |
f1af9d3a | 1182 | *out_nid = m_nid; |
7bd0b0f0 | 1183 | if (m_start >= r_start) |
f1af9d3a | 1184 | idx_a--; |
7bd0b0f0 | 1185 | else |
f1af9d3a PH |
1186 | idx_b--; |
1187 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
7bd0b0f0 TH |
1188 | return; |
1189 | } | |
1190 | } | |
1191 | } | |
f1af9d3a | 1192 | /* signal end of iteration */ |
7bd0b0f0 TH |
1193 | *idx = ULLONG_MAX; |
1194 | } | |
1195 | ||
7c0caeb8 TH |
1196 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1197 | /* | |
45e79815 | 1198 | * Common iterator interface used to define for_each_mem_pfn_range(). |
7c0caeb8 TH |
1199 | */ |
1200 | void __init_memblock __next_mem_pfn_range(int *idx, int nid, | |
1201 | unsigned long *out_start_pfn, | |
1202 | unsigned long *out_end_pfn, int *out_nid) | |
1203 | { | |
1204 | struct memblock_type *type = &memblock.memory; | |
1205 | struct memblock_region *r; | |
1206 | ||
1207 | while (++*idx < type->cnt) { | |
1208 | r = &type->regions[*idx]; | |
1209 | ||
1210 | if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) | |
1211 | continue; | |
1212 | if (nid == MAX_NUMNODES || nid == r->nid) | |
1213 | break; | |
1214 | } | |
1215 | if (*idx >= type->cnt) { | |
1216 | *idx = -1; | |
1217 | return; | |
1218 | } | |
1219 | ||
1220 | if (out_start_pfn) | |
1221 | *out_start_pfn = PFN_UP(r->base); | |
1222 | if (out_end_pfn) | |
1223 | *out_end_pfn = PFN_DOWN(r->base + r->size); | |
1224 | if (out_nid) | |
1225 | *out_nid = r->nid; | |
1226 | } | |
1227 | ||
1228 | /** | |
1229 | * memblock_set_node - set node ID on memblock regions | |
1230 | * @base: base of area to set node ID for | |
1231 | * @size: size of area to set node ID for | |
e7e8de59 | 1232 | * @type: memblock type to set node ID for |
7c0caeb8 TH |
1233 | * @nid: node ID to set |
1234 | * | |
47cec443 | 1235 | * Set the nid of memblock @type regions in [@base, @base + @size) to @nid. |
7c0caeb8 TH |
1236 | * Regions which cross the area boundaries are split as necessary. |
1237 | * | |
47cec443 | 1238 | * Return: |
7c0caeb8 TH |
1239 | * 0 on success, -errno on failure. |
1240 | */ | |
1241 | int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, | |
e7e8de59 | 1242 | struct memblock_type *type, int nid) |
7c0caeb8 | 1243 | { |
6a9ceb31 TH |
1244 | int start_rgn, end_rgn; |
1245 | int i, ret; | |
7c0caeb8 | 1246 | |
6a9ceb31 TH |
1247 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
1248 | if (ret) | |
1249 | return ret; | |
7c0caeb8 | 1250 | |
6a9ceb31 | 1251 | for (i = start_rgn; i < end_rgn; i++) |
e9d24ad3 | 1252 | memblock_set_region_node(&type->regions[i], nid); |
7c0caeb8 TH |
1253 | |
1254 | memblock_merge_regions(type); | |
1255 | return 0; | |
1256 | } | |
1257 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ | |
837566e7 AD |
1258 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT |
1259 | /** | |
1260 | * __next_mem_pfn_range_in_zone - iterator for for_each_*_range_in_zone() | |
1261 | * | |
1262 | * @idx: pointer to u64 loop variable | |
1263 | * @zone: zone in which all of the memory blocks reside | |
1264 | * @out_spfn: ptr to ulong for start pfn of the range, can be %NULL | |
1265 | * @out_epfn: ptr to ulong for end pfn of the range, can be %NULL | |
1266 | * | |
1267 | * This function is meant to be a zone/pfn specific wrapper for the | |
1268 | * for_each_mem_range type iterators. Specifically they are used in the | |
1269 | * deferred memory init routines and as such we were duplicating much of | |
1270 | * this logic throughout the code. So instead of having it in multiple | |
1271 | * locations it seemed like it would make more sense to centralize this to | |
1272 | * one new iterator that does everything they need. | |
1273 | */ | |
1274 | void __init_memblock | |
1275 | __next_mem_pfn_range_in_zone(u64 *idx, struct zone *zone, | |
1276 | unsigned long *out_spfn, unsigned long *out_epfn) | |
1277 | { | |
1278 | int zone_nid = zone_to_nid(zone); | |
1279 | phys_addr_t spa, epa; | |
1280 | int nid; | |
1281 | ||
1282 | __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, | |
1283 | &memblock.memory, &memblock.reserved, | |
1284 | &spa, &epa, &nid); | |
1285 | ||
1286 | while (*idx != U64_MAX) { | |
1287 | unsigned long epfn = PFN_DOWN(epa); | |
1288 | unsigned long spfn = PFN_UP(spa); | |
1289 | ||
1290 | /* | |
1291 | * Verify the end is at least past the start of the zone and | |
1292 | * that we have at least one PFN to initialize. | |
1293 | */ | |
1294 | if (zone->zone_start_pfn < epfn && spfn < epfn) { | |
1295 | /* if we went too far just stop searching */ | |
1296 | if (zone_end_pfn(zone) <= spfn) { | |
1297 | *idx = U64_MAX; | |
1298 | break; | |
1299 | } | |
1300 | ||
1301 | if (out_spfn) | |
1302 | *out_spfn = max(zone->zone_start_pfn, spfn); | |
1303 | if (out_epfn) | |
1304 | *out_epfn = min(zone_end_pfn(zone), epfn); | |
1305 | ||
1306 | return; | |
1307 | } | |
1308 | ||
1309 | __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, | |
1310 | &memblock.memory, &memblock.reserved, | |
1311 | &spa, &epa, &nid); | |
1312 | } | |
1313 | ||
1314 | /* signal end of iteration */ | |
1315 | if (out_spfn) | |
1316 | *out_spfn = ULONG_MAX; | |
1317 | if (out_epfn) | |
1318 | *out_epfn = 0; | |
1319 | } | |
1320 | ||
1321 | #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ | |
7c0caeb8 | 1322 | |
92d12f95 MR |
1323 | /** |
1324 | * memblock_alloc_range_nid - allocate boot memory block | |
1325 | * @size: size of memory block to be allocated in bytes | |
1326 | * @align: alignment of the region and block's size | |
1327 | * @start: the lower bound of the memory region to allocate (phys address) | |
1328 | * @end: the upper bound of the memory region to allocate (phys address) | |
1329 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1330 | * | |
1331 | * The allocation is performed from memory region limited by | |
1332 | * memblock.current_limit if @max_addr == %MEMBLOCK_ALLOC_ACCESSIBLE. | |
1333 | * | |
1334 | * If the specified node can not hold the requested memory the | |
1335 | * allocation falls back to any node in the system | |
1336 | * | |
1337 | * For systems with memory mirroring, the allocation is attempted first | |
1338 | * from the regions with mirroring enabled and then retried from any | |
1339 | * memory region. | |
1340 | * | |
1341 | * In addition, function sets the min_count to 0 using kmemleak_alloc_phys for | |
1342 | * allocated boot memory block, so that it is never reported as leaks. | |
1343 | * | |
1344 | * Return: | |
1345 | * Physical address of allocated memory block on success, %0 on failure. | |
1346 | */ | |
2bfc2862 AM |
1347 | static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, |
1348 | phys_addr_t align, phys_addr_t start, | |
92d12f95 | 1349 | phys_addr_t end, int nid) |
95f72d1e | 1350 | { |
92d12f95 | 1351 | enum memblock_flags flags = choose_memblock_flags(); |
6ed311b2 | 1352 | phys_addr_t found; |
95f72d1e | 1353 | |
92d12f95 MR |
1354 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1355 | nid = NUMA_NO_NODE; | |
1356 | ||
2f770806 MR |
1357 | if (!align) { |
1358 | /* Can't use WARNs this early in boot on powerpc */ | |
1359 | dump_stack(); | |
1360 | align = SMP_CACHE_BYTES; | |
1361 | } | |
1362 | ||
92d12f95 MR |
1363 | if (end > memblock.current_limit) |
1364 | end = memblock.current_limit; | |
1365 | ||
1366 | again: | |
fc6daaf9 TL |
1367 | found = memblock_find_in_range_node(size, align, start, end, nid, |
1368 | flags); | |
92d12f95 MR |
1369 | if (found && !memblock_reserve(found, size)) |
1370 | goto done; | |
1371 | ||
1372 | if (nid != NUMA_NO_NODE) { | |
1373 | found = memblock_find_in_range_node(size, align, start, | |
1374 | end, NUMA_NO_NODE, | |
1375 | flags); | |
1376 | if (found && !memblock_reserve(found, size)) | |
1377 | goto done; | |
1378 | } | |
1379 | ||
1380 | if (flags & MEMBLOCK_MIRROR) { | |
1381 | flags &= ~MEMBLOCK_MIRROR; | |
1382 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
1383 | &size); | |
1384 | goto again; | |
1385 | } | |
1386 | ||
1387 | return 0; | |
1388 | ||
1389 | done: | |
1390 | /* Skip kmemleak for kasan_init() due to high volume. */ | |
1391 | if (end != MEMBLOCK_ALLOC_KASAN) | |
aedf95ea | 1392 | /* |
92d12f95 MR |
1393 | * The min_count is set to 0 so that memblock allocated |
1394 | * blocks are never reported as leaks. This is because many | |
1395 | * of these blocks are only referred via the physical | |
1396 | * address which is not looked up by kmemleak. | |
aedf95ea | 1397 | */ |
9099daed | 1398 | kmemleak_alloc_phys(found, size, 0, 0); |
92d12f95 MR |
1399 | |
1400 | return found; | |
95f72d1e YL |
1401 | } |
1402 | ||
a2974133 MR |
1403 | /** |
1404 | * memblock_phys_alloc_range - allocate a memory block inside specified range | |
1405 | * @size: size of memory block to be allocated in bytes | |
1406 | * @align: alignment of the region and block's size | |
1407 | * @start: the lower bound of the memory region to allocate (physical address) | |
1408 | * @end: the upper bound of the memory region to allocate (physical address) | |
1409 | * | |
1410 | * Allocate @size bytes in the between @start and @end. | |
1411 | * | |
1412 | * Return: physical address of the allocated memory block on success, | |
1413 | * %0 on failure. | |
1414 | */ | |
8a770c2a MR |
1415 | phys_addr_t __init memblock_phys_alloc_range(phys_addr_t size, |
1416 | phys_addr_t align, | |
1417 | phys_addr_t start, | |
1418 | phys_addr_t end) | |
2bfc2862 | 1419 | { |
92d12f95 | 1420 | return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE); |
7bd0b0f0 TH |
1421 | } |
1422 | ||
a2974133 MR |
1423 | /** |
1424 | * memblock_phys_alloc_try_nid - allocate a memory block from specified MUMA node | |
1425 | * @size: size of memory block to be allocated in bytes | |
1426 | * @align: alignment of the region and block's size | |
1427 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1428 | * | |
1429 | * Allocates memory block from the specified NUMA node. If the node | |
1430 | * has no available memory, attempts to allocated from any node in the | |
1431 | * system. | |
1432 | * | |
1433 | * Return: physical address of the allocated memory block on success, | |
1434 | * %0 on failure. | |
1435 | */ | |
9a8dd708 | 1436 | phys_addr_t __init memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) |
9d1e2492 | 1437 | { |
33755574 | 1438 | return memblock_alloc_range_nid(size, align, 0, |
92d12f95 | 1439 | MEMBLOCK_ALLOC_ACCESSIBLE, nid); |
95f72d1e YL |
1440 | } |
1441 | ||
26f09e9b | 1442 | /** |
eb31d559 | 1443 | * memblock_alloc_internal - allocate boot memory block |
26f09e9b SS |
1444 | * @size: size of memory block to be allocated in bytes |
1445 | * @align: alignment of the region and block's size | |
1446 | * @min_addr: the lower bound of the memory region to allocate (phys address) | |
1447 | * @max_addr: the upper bound of the memory region to allocate (phys address) | |
1448 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1449 | * | |
92d12f95 MR |
1450 | * Allocates memory block using memblock_alloc_range_nid() and |
1451 | * converts the returned physical address to virtual. | |
26f09e9b | 1452 | * |
92d12f95 MR |
1453 | * The @min_addr limit is dropped if it can not be satisfied and the allocation |
1454 | * will fall back to memory below @min_addr. Other constraints, such | |
1455 | * as node and mirrored memory will be handled again in | |
1456 | * memblock_alloc_range_nid(). | |
26f09e9b | 1457 | * |
47cec443 | 1458 | * Return: |
26f09e9b SS |
1459 | * Virtual address of allocated memory block on success, NULL on failure. |
1460 | */ | |
eb31d559 | 1461 | static void * __init memblock_alloc_internal( |
26f09e9b SS |
1462 | phys_addr_t size, phys_addr_t align, |
1463 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1464 | int nid) | |
1465 | { | |
1466 | phys_addr_t alloc; | |
26f09e9b SS |
1467 | |
1468 | /* | |
1469 | * Detect any accidental use of these APIs after slab is ready, as at | |
1470 | * this moment memblock may be deinitialized already and its | |
c6ffc5ca | 1471 | * internal data may be destroyed (after execution of memblock_free_all) |
26f09e9b SS |
1472 | */ |
1473 | if (WARN_ON_ONCE(slab_is_available())) | |
1474 | return kzalloc_node(size, GFP_NOWAIT, nid); | |
1475 | ||
92d12f95 | 1476 | alloc = memblock_alloc_range_nid(size, align, min_addr, max_addr, nid); |
26f09e9b | 1477 | |
92d12f95 MR |
1478 | /* retry allocation without lower limit */ |
1479 | if (!alloc && min_addr) | |
1480 | alloc = memblock_alloc_range_nid(size, align, 0, max_addr, nid); | |
26f09e9b | 1481 | |
92d12f95 MR |
1482 | if (!alloc) |
1483 | return NULL; | |
26f09e9b | 1484 | |
92d12f95 | 1485 | return phys_to_virt(alloc); |
26f09e9b SS |
1486 | } |
1487 | ||
ea1f5f37 | 1488 | /** |
eb31d559 | 1489 | * memblock_alloc_try_nid_raw - allocate boot memory block without zeroing |
ea1f5f37 PT |
1490 | * memory and without panicking |
1491 | * @size: size of memory block to be allocated in bytes | |
1492 | * @align: alignment of the region and block's size | |
1493 | * @min_addr: the lower bound of the memory region from where the allocation | |
1494 | * is preferred (phys address) | |
1495 | * @max_addr: the upper bound of the memory region from where the allocation | |
97ad1087 | 1496 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to |
ea1f5f37 PT |
1497 | * allocate only from memory limited by memblock.current_limit value |
1498 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1499 | * | |
1500 | * Public function, provides additional debug information (including caller | |
1501 | * info), if enabled. Does not zero allocated memory, does not panic if request | |
1502 | * cannot be satisfied. | |
1503 | * | |
47cec443 | 1504 | * Return: |
ea1f5f37 PT |
1505 | * Virtual address of allocated memory block on success, NULL on failure. |
1506 | */ | |
eb31d559 | 1507 | void * __init memblock_alloc_try_nid_raw( |
ea1f5f37 PT |
1508 | phys_addr_t size, phys_addr_t align, |
1509 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1510 | int nid) | |
1511 | { | |
1512 | void *ptr; | |
1513 | ||
d75f773c | 1514 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
a36aab89 MR |
1515 | __func__, (u64)size, (u64)align, nid, &min_addr, |
1516 | &max_addr, (void *)_RET_IP_); | |
ea1f5f37 | 1517 | |
eb31d559 | 1518 | ptr = memblock_alloc_internal(size, align, |
ea1f5f37 | 1519 | min_addr, max_addr, nid); |
ea1f5f37 | 1520 | if (ptr && size > 0) |
f682a97a AD |
1521 | page_init_poison(ptr, size); |
1522 | ||
ea1f5f37 PT |
1523 | return ptr; |
1524 | } | |
1525 | ||
26f09e9b | 1526 | /** |
c0dbe825 | 1527 | * memblock_alloc_try_nid - allocate boot memory block |
26f09e9b SS |
1528 | * @size: size of memory block to be allocated in bytes |
1529 | * @align: alignment of the region and block's size | |
1530 | * @min_addr: the lower bound of the memory region from where the allocation | |
1531 | * is preferred (phys address) | |
1532 | * @max_addr: the upper bound of the memory region from where the allocation | |
97ad1087 | 1533 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to |
26f09e9b SS |
1534 | * allocate only from memory limited by memblock.current_limit value |
1535 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1536 | * | |
c0dbe825 MR |
1537 | * Public function, provides additional debug information (including caller |
1538 | * info), if enabled. This function zeroes the allocated memory. | |
26f09e9b | 1539 | * |
47cec443 | 1540 | * Return: |
26f09e9b SS |
1541 | * Virtual address of allocated memory block on success, NULL on failure. |
1542 | */ | |
eb31d559 | 1543 | void * __init memblock_alloc_try_nid( |
26f09e9b SS |
1544 | phys_addr_t size, phys_addr_t align, |
1545 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1546 | int nid) | |
1547 | { | |
1548 | void *ptr; | |
1549 | ||
d75f773c | 1550 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
a36aab89 MR |
1551 | __func__, (u64)size, (u64)align, nid, &min_addr, |
1552 | &max_addr, (void *)_RET_IP_); | |
eb31d559 | 1553 | ptr = memblock_alloc_internal(size, align, |
26f09e9b | 1554 | min_addr, max_addr, nid); |
c0dbe825 | 1555 | if (ptr) |
ea1f5f37 | 1556 | memset(ptr, 0, size); |
26f09e9b | 1557 | |
c0dbe825 | 1558 | return ptr; |
26f09e9b SS |
1559 | } |
1560 | ||
48a833cc | 1561 | /** |
a2974133 | 1562 | * __memblock_free_late - free pages directly to buddy allocator |
48a833cc | 1563 | * @base: phys starting address of the boot memory block |
26f09e9b SS |
1564 | * @size: size of the boot memory block in bytes |
1565 | * | |
a2974133 | 1566 | * This is only useful when the memblock allocator has already been torn |
26f09e9b | 1567 | * down, but we are still initializing the system. Pages are released directly |
a2974133 | 1568 | * to the buddy allocator. |
26f09e9b SS |
1569 | */ |
1570 | void __init __memblock_free_late(phys_addr_t base, phys_addr_t size) | |
1571 | { | |
a36aab89 | 1572 | phys_addr_t cursor, end; |
26f09e9b | 1573 | |
a36aab89 | 1574 | end = base + size - 1; |
d75f773c | 1575 | memblock_dbg("%s: [%pa-%pa] %pS\n", |
a36aab89 | 1576 | __func__, &base, &end, (void *)_RET_IP_); |
9099daed | 1577 | kmemleak_free_part_phys(base, size); |
26f09e9b SS |
1578 | cursor = PFN_UP(base); |
1579 | end = PFN_DOWN(base + size); | |
1580 | ||
1581 | for (; cursor < end; cursor++) { | |
7c2ee349 | 1582 | memblock_free_pages(pfn_to_page(cursor), cursor, 0); |
ca79b0c2 | 1583 | totalram_pages_inc(); |
26f09e9b SS |
1584 | } |
1585 | } | |
9d1e2492 BH |
1586 | |
1587 | /* | |
1588 | * Remaining API functions | |
1589 | */ | |
1590 | ||
1f1ffb8a | 1591 | phys_addr_t __init_memblock memblock_phys_mem_size(void) |
95f72d1e | 1592 | { |
1440c4e2 | 1593 | return memblock.memory.total_size; |
95f72d1e YL |
1594 | } |
1595 | ||
8907de5d SD |
1596 | phys_addr_t __init_memblock memblock_reserved_size(void) |
1597 | { | |
1598 | return memblock.reserved.total_size; | |
1599 | } | |
1600 | ||
595ad9af YL |
1601 | phys_addr_t __init memblock_mem_size(unsigned long limit_pfn) |
1602 | { | |
1603 | unsigned long pages = 0; | |
1604 | struct memblock_region *r; | |
1605 | unsigned long start_pfn, end_pfn; | |
1606 | ||
1607 | for_each_memblock(memory, r) { | |
1608 | start_pfn = memblock_region_memory_base_pfn(r); | |
1609 | end_pfn = memblock_region_memory_end_pfn(r); | |
1610 | start_pfn = min_t(unsigned long, start_pfn, limit_pfn); | |
1611 | end_pfn = min_t(unsigned long, end_pfn, limit_pfn); | |
1612 | pages += end_pfn - start_pfn; | |
1613 | } | |
1614 | ||
16763230 | 1615 | return PFN_PHYS(pages); |
595ad9af YL |
1616 | } |
1617 | ||
0a93ebef SR |
1618 | /* lowest address */ |
1619 | phys_addr_t __init_memblock memblock_start_of_DRAM(void) | |
1620 | { | |
1621 | return memblock.memory.regions[0].base; | |
1622 | } | |
1623 | ||
10d06439 | 1624 | phys_addr_t __init_memblock memblock_end_of_DRAM(void) |
95f72d1e YL |
1625 | { |
1626 | int idx = memblock.memory.cnt - 1; | |
1627 | ||
e3239ff9 | 1628 | return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); |
95f72d1e YL |
1629 | } |
1630 | ||
a571d4eb | 1631 | static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) |
95f72d1e | 1632 | { |
1c4bc43d | 1633 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
136199f0 | 1634 | struct memblock_region *r; |
95f72d1e | 1635 | |
a571d4eb DC |
1636 | /* |
1637 | * translate the memory @limit size into the max address within one of | |
1638 | * the memory memblock regions, if the @limit exceeds the total size | |
1c4bc43d | 1639 | * of those regions, max_addr will keep original value PHYS_ADDR_MAX |
a571d4eb | 1640 | */ |
136199f0 | 1641 | for_each_memblock(memory, r) { |
c0ce8fef TH |
1642 | if (limit <= r->size) { |
1643 | max_addr = r->base + limit; | |
1644 | break; | |
95f72d1e | 1645 | } |
c0ce8fef | 1646 | limit -= r->size; |
95f72d1e | 1647 | } |
c0ce8fef | 1648 | |
a571d4eb DC |
1649 | return max_addr; |
1650 | } | |
1651 | ||
1652 | void __init memblock_enforce_memory_limit(phys_addr_t limit) | |
1653 | { | |
1c4bc43d | 1654 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
a571d4eb DC |
1655 | |
1656 | if (!limit) | |
1657 | return; | |
1658 | ||
1659 | max_addr = __find_max_addr(limit); | |
1660 | ||
1661 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1662 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1663 | return; |
1664 | ||
c0ce8fef | 1665 | /* truncate both memory and reserved regions */ |
f1af9d3a | 1666 | memblock_remove_range(&memblock.memory, max_addr, |
1c4bc43d | 1667 | PHYS_ADDR_MAX); |
f1af9d3a | 1668 | memblock_remove_range(&memblock.reserved, max_addr, |
1c4bc43d | 1669 | PHYS_ADDR_MAX); |
95f72d1e YL |
1670 | } |
1671 | ||
c9ca9b4e AT |
1672 | void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) |
1673 | { | |
1674 | int start_rgn, end_rgn; | |
1675 | int i, ret; | |
1676 | ||
1677 | if (!size) | |
1678 | return; | |
1679 | ||
1680 | ret = memblock_isolate_range(&memblock.memory, base, size, | |
1681 | &start_rgn, &end_rgn); | |
1682 | if (ret) | |
1683 | return; | |
1684 | ||
1685 | /* remove all the MAP regions */ | |
1686 | for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) | |
1687 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1688 | memblock_remove_region(&memblock.memory, i); | |
1689 | ||
1690 | for (i = start_rgn - 1; i >= 0; i--) | |
1691 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1692 | memblock_remove_region(&memblock.memory, i); | |
1693 | ||
1694 | /* truncate the reserved regions */ | |
1695 | memblock_remove_range(&memblock.reserved, 0, base); | |
1696 | memblock_remove_range(&memblock.reserved, | |
1c4bc43d | 1697 | base + size, PHYS_ADDR_MAX); |
c9ca9b4e AT |
1698 | } |
1699 | ||
a571d4eb DC |
1700 | void __init memblock_mem_limit_remove_map(phys_addr_t limit) |
1701 | { | |
a571d4eb | 1702 | phys_addr_t max_addr; |
a571d4eb DC |
1703 | |
1704 | if (!limit) | |
1705 | return; | |
1706 | ||
1707 | max_addr = __find_max_addr(limit); | |
1708 | ||
1709 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1710 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1711 | return; |
1712 | ||
c9ca9b4e | 1713 | memblock_cap_memory_range(0, max_addr); |
a571d4eb DC |
1714 | } |
1715 | ||
cd79481d | 1716 | static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) |
72d4b0b4 BH |
1717 | { |
1718 | unsigned int left = 0, right = type->cnt; | |
1719 | ||
1720 | do { | |
1721 | unsigned int mid = (right + left) / 2; | |
1722 | ||
1723 | if (addr < type->regions[mid].base) | |
1724 | right = mid; | |
1725 | else if (addr >= (type->regions[mid].base + | |
1726 | type->regions[mid].size)) | |
1727 | left = mid + 1; | |
1728 | else | |
1729 | return mid; | |
1730 | } while (left < right); | |
1731 | return -1; | |
1732 | } | |
1733 | ||
f5a222dc | 1734 | bool __init_memblock memblock_is_reserved(phys_addr_t addr) |
95f72d1e | 1735 | { |
72d4b0b4 BH |
1736 | return memblock_search(&memblock.reserved, addr) != -1; |
1737 | } | |
95f72d1e | 1738 | |
b4ad0c7e | 1739 | bool __init_memblock memblock_is_memory(phys_addr_t addr) |
72d4b0b4 BH |
1740 | { |
1741 | return memblock_search(&memblock.memory, addr) != -1; | |
1742 | } | |
1743 | ||
937f0c26 | 1744 | bool __init_memblock memblock_is_map_memory(phys_addr_t addr) |
bf3d3cc5 AB |
1745 | { |
1746 | int i = memblock_search(&memblock.memory, addr); | |
1747 | ||
1748 | if (i == -1) | |
1749 | return false; | |
1750 | return !memblock_is_nomap(&memblock.memory.regions[i]); | |
1751 | } | |
1752 | ||
e76b63f8 YL |
1753 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1754 | int __init_memblock memblock_search_pfn_nid(unsigned long pfn, | |
1755 | unsigned long *start_pfn, unsigned long *end_pfn) | |
1756 | { | |
1757 | struct memblock_type *type = &memblock.memory; | |
16763230 | 1758 | int mid = memblock_search(type, PFN_PHYS(pfn)); |
e76b63f8 YL |
1759 | |
1760 | if (mid == -1) | |
1761 | return -1; | |
1762 | ||
f7e2f7e8 FF |
1763 | *start_pfn = PFN_DOWN(type->regions[mid].base); |
1764 | *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); | |
e76b63f8 YL |
1765 | |
1766 | return type->regions[mid].nid; | |
1767 | } | |
1768 | #endif | |
1769 | ||
eab30949 SB |
1770 | /** |
1771 | * memblock_is_region_memory - check if a region is a subset of memory | |
1772 | * @base: base of region to check | |
1773 | * @size: size of region to check | |
1774 | * | |
47cec443 | 1775 | * Check if the region [@base, @base + @size) is a subset of a memory block. |
eab30949 | 1776 | * |
47cec443 | 1777 | * Return: |
eab30949 SB |
1778 | * 0 if false, non-zero if true |
1779 | */ | |
937f0c26 | 1780 | bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) |
72d4b0b4 | 1781 | { |
abb65272 | 1782 | int idx = memblock_search(&memblock.memory, base); |
eb18f1b5 | 1783 | phys_addr_t end = base + memblock_cap_size(base, &size); |
72d4b0b4 BH |
1784 | |
1785 | if (idx == -1) | |
937f0c26 | 1786 | return false; |
ef415ef4 | 1787 | return (memblock.memory.regions[idx].base + |
eb18f1b5 | 1788 | memblock.memory.regions[idx].size) >= end; |
95f72d1e YL |
1789 | } |
1790 | ||
eab30949 SB |
1791 | /** |
1792 | * memblock_is_region_reserved - check if a region intersects reserved memory | |
1793 | * @base: base of region to check | |
1794 | * @size: size of region to check | |
1795 | * | |
47cec443 MR |
1796 | * Check if the region [@base, @base + @size) intersects a reserved |
1797 | * memory block. | |
eab30949 | 1798 | * |
47cec443 | 1799 | * Return: |
c5c5c9d1 | 1800 | * True if they intersect, false if not. |
eab30949 | 1801 | */ |
c5c5c9d1 | 1802 | bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) |
95f72d1e | 1803 | { |
eb18f1b5 | 1804 | memblock_cap_size(base, &size); |
c5c5c9d1 | 1805 | return memblock_overlaps_region(&memblock.reserved, base, size); |
95f72d1e YL |
1806 | } |
1807 | ||
6ede1fd3 YL |
1808 | void __init_memblock memblock_trim_memory(phys_addr_t align) |
1809 | { | |
6ede1fd3 | 1810 | phys_addr_t start, end, orig_start, orig_end; |
136199f0 | 1811 | struct memblock_region *r; |
6ede1fd3 | 1812 | |
136199f0 EM |
1813 | for_each_memblock(memory, r) { |
1814 | orig_start = r->base; | |
1815 | orig_end = r->base + r->size; | |
6ede1fd3 YL |
1816 | start = round_up(orig_start, align); |
1817 | end = round_down(orig_end, align); | |
1818 | ||
1819 | if (start == orig_start && end == orig_end) | |
1820 | continue; | |
1821 | ||
1822 | if (start < end) { | |
136199f0 EM |
1823 | r->base = start; |
1824 | r->size = end - start; | |
6ede1fd3 | 1825 | } else { |
136199f0 EM |
1826 | memblock_remove_region(&memblock.memory, |
1827 | r - memblock.memory.regions); | |
1828 | r--; | |
6ede1fd3 YL |
1829 | } |
1830 | } | |
1831 | } | |
e63075a3 | 1832 | |
3661ca66 | 1833 | void __init_memblock memblock_set_current_limit(phys_addr_t limit) |
e63075a3 BH |
1834 | { |
1835 | memblock.current_limit = limit; | |
1836 | } | |
1837 | ||
fec51014 LA |
1838 | phys_addr_t __init_memblock memblock_get_current_limit(void) |
1839 | { | |
1840 | return memblock.current_limit; | |
1841 | } | |
1842 | ||
0262d9c8 | 1843 | static void __init_memblock memblock_dump(struct memblock_type *type) |
6ed311b2 | 1844 | { |
5d63f81c | 1845 | phys_addr_t base, end, size; |
e1720fee | 1846 | enum memblock_flags flags; |
8c9c1701 AK |
1847 | int idx; |
1848 | struct memblock_region *rgn; | |
6ed311b2 | 1849 | |
0262d9c8 | 1850 | pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); |
6ed311b2 | 1851 | |
66e8b438 | 1852 | for_each_memblock_type(idx, type, rgn) { |
7c0caeb8 TH |
1853 | char nid_buf[32] = ""; |
1854 | ||
1855 | base = rgn->base; | |
1856 | size = rgn->size; | |
5d63f81c | 1857 | end = base + size - 1; |
66a20757 | 1858 | flags = rgn->flags; |
7c0caeb8 TH |
1859 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1860 | if (memblock_get_region_node(rgn) != MAX_NUMNODES) | |
1861 | snprintf(nid_buf, sizeof(nid_buf), " on node %d", | |
1862 | memblock_get_region_node(rgn)); | |
1863 | #endif | |
e1720fee | 1864 | pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#x\n", |
0262d9c8 | 1865 | type->name, idx, &base, &end, &size, nid_buf, flags); |
6ed311b2 BH |
1866 | } |
1867 | } | |
1868 | ||
4ff7b82f | 1869 | void __init_memblock __memblock_dump_all(void) |
6ed311b2 | 1870 | { |
6ed311b2 | 1871 | pr_info("MEMBLOCK configuration:\n"); |
5d63f81c MC |
1872 | pr_info(" memory size = %pa reserved size = %pa\n", |
1873 | &memblock.memory.total_size, | |
1874 | &memblock.reserved.total_size); | |
6ed311b2 | 1875 | |
0262d9c8 HC |
1876 | memblock_dump(&memblock.memory); |
1877 | memblock_dump(&memblock.reserved); | |
409efd4c | 1878 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
0262d9c8 | 1879 | memblock_dump(&memblock.physmem); |
409efd4c | 1880 | #endif |
6ed311b2 BH |
1881 | } |
1882 | ||
1aadc056 | 1883 | void __init memblock_allow_resize(void) |
6ed311b2 | 1884 | { |
142b45a7 | 1885 | memblock_can_resize = 1; |
6ed311b2 BH |
1886 | } |
1887 | ||
6ed311b2 BH |
1888 | static int __init early_memblock(char *p) |
1889 | { | |
1890 | if (p && strstr(p, "debug")) | |
1891 | memblock_debug = 1; | |
1892 | return 0; | |
1893 | } | |
1894 | early_param("memblock", early_memblock); | |
1895 | ||
bda49a81 MR |
1896 | static void __init __free_pages_memory(unsigned long start, unsigned long end) |
1897 | { | |
1898 | int order; | |
1899 | ||
1900 | while (start < end) { | |
1901 | order = min(MAX_ORDER - 1UL, __ffs(start)); | |
1902 | ||
1903 | while (start + (1UL << order) > end) | |
1904 | order--; | |
1905 | ||
1906 | memblock_free_pages(pfn_to_page(start), start, order); | |
1907 | ||
1908 | start += (1UL << order); | |
1909 | } | |
1910 | } | |
1911 | ||
1912 | static unsigned long __init __free_memory_core(phys_addr_t start, | |
1913 | phys_addr_t end) | |
1914 | { | |
1915 | unsigned long start_pfn = PFN_UP(start); | |
1916 | unsigned long end_pfn = min_t(unsigned long, | |
1917 | PFN_DOWN(end), max_low_pfn); | |
1918 | ||
1919 | if (start_pfn >= end_pfn) | |
1920 | return 0; | |
1921 | ||
1922 | __free_pages_memory(start_pfn, end_pfn); | |
1923 | ||
1924 | return end_pfn - start_pfn; | |
1925 | } | |
1926 | ||
1927 | static unsigned long __init free_low_memory_core_early(void) | |
1928 | { | |
1929 | unsigned long count = 0; | |
1930 | phys_addr_t start, end; | |
1931 | u64 i; | |
1932 | ||
1933 | memblock_clear_hotplug(0, -1); | |
1934 | ||
1935 | for_each_reserved_mem_region(i, &start, &end) | |
1936 | reserve_bootmem_region(start, end); | |
1937 | ||
1938 | /* | |
1939 | * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id | |
1940 | * because in some case like Node0 doesn't have RAM installed | |
1941 | * low ram will be on Node1 | |
1942 | */ | |
1943 | for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, | |
1944 | NULL) | |
1945 | count += __free_memory_core(start, end); | |
1946 | ||
1947 | return count; | |
1948 | } | |
1949 | ||
1950 | static int reset_managed_pages_done __initdata; | |
1951 | ||
1952 | void reset_node_managed_pages(pg_data_t *pgdat) | |
1953 | { | |
1954 | struct zone *z; | |
1955 | ||
1956 | for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) | |
9705bea5 | 1957 | atomic_long_set(&z->managed_pages, 0); |
bda49a81 MR |
1958 | } |
1959 | ||
1960 | void __init reset_all_zones_managed_pages(void) | |
1961 | { | |
1962 | struct pglist_data *pgdat; | |
1963 | ||
1964 | if (reset_managed_pages_done) | |
1965 | return; | |
1966 | ||
1967 | for_each_online_pgdat(pgdat) | |
1968 | reset_node_managed_pages(pgdat); | |
1969 | ||
1970 | reset_managed_pages_done = 1; | |
1971 | } | |
1972 | ||
1973 | /** | |
1974 | * memblock_free_all - release free pages to the buddy allocator | |
1975 | * | |
1976 | * Return: the number of pages actually released. | |
1977 | */ | |
1978 | unsigned long __init memblock_free_all(void) | |
1979 | { | |
1980 | unsigned long pages; | |
1981 | ||
1982 | reset_all_zones_managed_pages(); | |
1983 | ||
1984 | pages = free_low_memory_core_early(); | |
ca79b0c2 | 1985 | totalram_pages_add(pages); |
bda49a81 MR |
1986 | |
1987 | return pages; | |
1988 | } | |
1989 | ||
350e88ba | 1990 | #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_ARCH_KEEP_MEMBLOCK) |
6d03b885 BH |
1991 | |
1992 | static int memblock_debug_show(struct seq_file *m, void *private) | |
1993 | { | |
1994 | struct memblock_type *type = m->private; | |
1995 | struct memblock_region *reg; | |
1996 | int i; | |
5d63f81c | 1997 | phys_addr_t end; |
6d03b885 BH |
1998 | |
1999 | for (i = 0; i < type->cnt; i++) { | |
2000 | reg = &type->regions[i]; | |
5d63f81c | 2001 | end = reg->base + reg->size - 1; |
6d03b885 | 2002 | |
5d63f81c MC |
2003 | seq_printf(m, "%4d: ", i); |
2004 | seq_printf(m, "%pa..%pa\n", ®->base, &end); | |
6d03b885 BH |
2005 | } |
2006 | return 0; | |
2007 | } | |
5ad35093 | 2008 | DEFINE_SHOW_ATTRIBUTE(memblock_debug); |
6d03b885 BH |
2009 | |
2010 | static int __init memblock_init_debugfs(void) | |
2011 | { | |
2012 | struct dentry *root = debugfs_create_dir("memblock", NULL); | |
d9f7979c | 2013 | |
0825a6f9 JP |
2014 | debugfs_create_file("memory", 0444, root, |
2015 | &memblock.memory, &memblock_debug_fops); | |
2016 | debugfs_create_file("reserved", 0444, root, | |
2017 | &memblock.reserved, &memblock_debug_fops); | |
70210ed9 | 2018 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
0825a6f9 JP |
2019 | debugfs_create_file("physmem", 0444, root, |
2020 | &memblock.physmem, &memblock_debug_fops); | |
70210ed9 | 2021 | #endif |
6d03b885 BH |
2022 | |
2023 | return 0; | |
2024 | } | |
2025 | __initcall(memblock_init_debugfs); | |
2026 | ||
2027 | #endif /* CONFIG_DEBUG_FS */ |