]>
Commit | Line | Data |
---|---|---|
2874c5fd | 1 | // SPDX-License-Identifier: GPL-2.0-or-later |
95f72d1e YL |
2 | /* |
3 | * Procedures for maintaining information about logical memory blocks. | |
4 | * | |
5 | * Peter Bergner, IBM Corp. June 2001. | |
6 | * Copyright (C) 2001 Peter Bergner. | |
95f72d1e YL |
7 | */ |
8 | ||
9 | #include <linux/kernel.h> | |
142b45a7 | 10 | #include <linux/slab.h> |
95f72d1e YL |
11 | #include <linux/init.h> |
12 | #include <linux/bitops.h> | |
449e8df3 | 13 | #include <linux/poison.h> |
c196f76f | 14 | #include <linux/pfn.h> |
6d03b885 | 15 | #include <linux/debugfs.h> |
514c6032 | 16 | #include <linux/kmemleak.h> |
6d03b885 | 17 | #include <linux/seq_file.h> |
95f72d1e YL |
18 | #include <linux/memblock.h> |
19 | ||
c4c5ad6b | 20 | #include <asm/sections.h> |
26f09e9b SS |
21 | #include <linux/io.h> |
22 | ||
23 | #include "internal.h" | |
79442ed1 | 24 | |
8a5b403d AB |
25 | #define INIT_MEMBLOCK_REGIONS 128 |
26 | #define INIT_PHYSMEM_REGIONS 4 | |
27 | ||
28 | #ifndef INIT_MEMBLOCK_RESERVED_REGIONS | |
29 | # define INIT_MEMBLOCK_RESERVED_REGIONS INIT_MEMBLOCK_REGIONS | |
30 | #endif | |
31 | ||
3e039c5c MR |
32 | /** |
33 | * DOC: memblock overview | |
34 | * | |
35 | * Memblock is a method of managing memory regions during the early | |
36 | * boot period when the usual kernel memory allocators are not up and | |
37 | * running. | |
38 | * | |
39 | * Memblock views the system memory as collections of contiguous | |
40 | * regions. There are several types of these collections: | |
41 | * | |
42 | * * ``memory`` - describes the physical memory available to the | |
43 | * kernel; this may differ from the actual physical memory installed | |
44 | * in the system, for instance when the memory is restricted with | |
45 | * ``mem=`` command line parameter | |
46 | * * ``reserved`` - describes the regions that were allocated | |
47 | * * ``physmap`` - describes the actual physical memory regardless of | |
48 | * the possible restrictions; the ``physmap`` type is only available | |
49 | * on some architectures. | |
50 | * | |
51 | * Each region is represented by :c:type:`struct memblock_region` that | |
52 | * defines the region extents, its attributes and NUMA node id on NUMA | |
53 | * systems. Every memory type is described by the :c:type:`struct | |
54 | * memblock_type` which contains an array of memory regions along with | |
55 | * the allocator metadata. The memory types are nicely wrapped with | |
56 | * :c:type:`struct memblock`. This structure is statically initialzed | |
57 | * at build time. The region arrays for the "memory" and "reserved" | |
58 | * types are initially sized to %INIT_MEMBLOCK_REGIONS and for the | |
59 | * "physmap" type to %INIT_PHYSMEM_REGIONS. | |
6e5af9a8 C |
60 | * The memblock_allow_resize() enables automatic resizing of the region |
61 | * arrays during addition of new regions. This feature should be used | |
62 | * with care so that memory allocated for the region array will not | |
63 | * overlap with areas that should be reserved, for example initrd. | |
3e039c5c MR |
64 | * |
65 | * The early architecture setup should tell memblock what the physical | |
6e5af9a8 C |
66 | * memory layout is by using memblock_add() or memblock_add_node() |
67 | * functions. The first function does not assign the region to a NUMA | |
68 | * node and it is appropriate for UMA systems. Yet, it is possible to | |
69 | * use it on NUMA systems as well and assign the region to a NUMA node | |
70 | * later in the setup process using memblock_set_node(). The | |
71 | * memblock_add_node() performs such an assignment directly. | |
3e039c5c | 72 | * |
a2974133 MR |
73 | * Once memblock is setup the memory can be allocated using one of the |
74 | * API variants: | |
75 | * | |
6e5af9a8 C |
76 | * * memblock_phys_alloc*() - these functions return the **physical** |
77 | * address of the allocated memory | |
78 | * * memblock_alloc*() - these functions return the **virtual** address | |
79 | * of the allocated memory. | |
a2974133 MR |
80 | * |
81 | * Note, that both API variants use implict assumptions about allowed | |
82 | * memory ranges and the fallback methods. Consult the documentation | |
6e5af9a8 C |
83 | * of memblock_alloc_internal() and memblock_alloc_range_nid() |
84 | * functions for more elaborate description. | |
3e039c5c | 85 | * |
6e5af9a8 C |
86 | * As the system boot progresses, the architecture specific mem_init() |
87 | * function frees all the memory to the buddy page allocator. | |
3e039c5c | 88 | * |
6e5af9a8 | 89 | * Unless an architecture enables %CONFIG_ARCH_KEEP_MEMBLOCK, the |
3e039c5c | 90 | * memblock data structures will be discarded after the system |
6e5af9a8 | 91 | * initialization completes. |
3e039c5c MR |
92 | */ |
93 | ||
bda49a81 MR |
94 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
95 | struct pglist_data __refdata contig_page_data; | |
96 | EXPORT_SYMBOL(contig_page_data); | |
97 | #endif | |
98 | ||
99 | unsigned long max_low_pfn; | |
100 | unsigned long min_low_pfn; | |
101 | unsigned long max_pfn; | |
102 | unsigned long long max_possible_pfn; | |
103 | ||
fe091c20 | 104 | static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock; |
8a5b403d | 105 | static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_RESERVED_REGIONS] __initdata_memblock; |
70210ed9 PH |
106 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
107 | static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock; | |
108 | #endif | |
fe091c20 TH |
109 | |
110 | struct memblock memblock __initdata_memblock = { | |
111 | .memory.regions = memblock_memory_init_regions, | |
112 | .memory.cnt = 1, /* empty dummy entry */ | |
113 | .memory.max = INIT_MEMBLOCK_REGIONS, | |
0262d9c8 | 114 | .memory.name = "memory", |
fe091c20 TH |
115 | |
116 | .reserved.regions = memblock_reserved_init_regions, | |
117 | .reserved.cnt = 1, /* empty dummy entry */ | |
8a5b403d | 118 | .reserved.max = INIT_MEMBLOCK_RESERVED_REGIONS, |
0262d9c8 | 119 | .reserved.name = "reserved", |
fe091c20 | 120 | |
70210ed9 PH |
121 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
122 | .physmem.regions = memblock_physmem_init_regions, | |
123 | .physmem.cnt = 1, /* empty dummy entry */ | |
124 | .physmem.max = INIT_PHYSMEM_REGIONS, | |
0262d9c8 | 125 | .physmem.name = "physmem", |
70210ed9 PH |
126 | #endif |
127 | ||
79442ed1 | 128 | .bottom_up = false, |
fe091c20 TH |
129 | .current_limit = MEMBLOCK_ALLOC_ANYWHERE, |
130 | }; | |
95f72d1e | 131 | |
10d06439 | 132 | int memblock_debug __initdata_memblock; |
a3f5bafc | 133 | static bool system_has_some_mirror __initdata_memblock = false; |
1aadc056 | 134 | static int memblock_can_resize __initdata_memblock; |
181eb394 GS |
135 | static int memblock_memory_in_slab __initdata_memblock = 0; |
136 | static int memblock_reserved_in_slab __initdata_memblock = 0; | |
95f72d1e | 137 | |
c366ea89 | 138 | static enum memblock_flags __init_memblock choose_memblock_flags(void) |
a3f5bafc TL |
139 | { |
140 | return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE; | |
141 | } | |
142 | ||
eb18f1b5 TH |
143 | /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */ |
144 | static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size) | |
145 | { | |
1c4bc43d | 146 | return *size = min(*size, PHYS_ADDR_MAX - base); |
eb18f1b5 TH |
147 | } |
148 | ||
6ed311b2 BH |
149 | /* |
150 | * Address comparison utilities | |
151 | */ | |
10d06439 | 152 | static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1, |
2898cc4c | 153 | phys_addr_t base2, phys_addr_t size2) |
95f72d1e YL |
154 | { |
155 | return ((base1 < (base2 + size2)) && (base2 < (base1 + size1))); | |
156 | } | |
157 | ||
95cf82ec | 158 | bool __init_memblock memblock_overlaps_region(struct memblock_type *type, |
2d7d3eb2 | 159 | phys_addr_t base, phys_addr_t size) |
6ed311b2 BH |
160 | { |
161 | unsigned long i; | |
162 | ||
f14516fb AK |
163 | for (i = 0; i < type->cnt; i++) |
164 | if (memblock_addrs_overlap(base, size, type->regions[i].base, | |
165 | type->regions[i].size)) | |
6ed311b2 | 166 | break; |
c5c5c9d1 | 167 | return i < type->cnt; |
6ed311b2 BH |
168 | } |
169 | ||
47cec443 | 170 | /** |
79442ed1 TC |
171 | * __memblock_find_range_bottom_up - find free area utility in bottom-up |
172 | * @start: start of candidate range | |
47cec443 MR |
173 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
174 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
79442ed1 TC |
175 | * @size: size of free area to find |
176 | * @align: alignment of free area to find | |
b1154233 | 177 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 178 | * @flags: pick from blocks based on memory attributes |
79442ed1 TC |
179 | * |
180 | * Utility called from memblock_find_in_range_node(), find free area bottom-up. | |
181 | * | |
47cec443 | 182 | * Return: |
79442ed1 TC |
183 | * Found address on success, 0 on failure. |
184 | */ | |
185 | static phys_addr_t __init_memblock | |
186 | __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 187 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 188 | enum memblock_flags flags) |
79442ed1 TC |
189 | { |
190 | phys_addr_t this_start, this_end, cand; | |
191 | u64 i; | |
192 | ||
fc6daaf9 | 193 | for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { |
79442ed1 TC |
194 | this_start = clamp(this_start, start, end); |
195 | this_end = clamp(this_end, start, end); | |
196 | ||
197 | cand = round_up(this_start, align); | |
198 | if (cand < this_end && this_end - cand >= size) | |
199 | return cand; | |
200 | } | |
201 | ||
202 | return 0; | |
203 | } | |
204 | ||
7bd0b0f0 | 205 | /** |
1402899e | 206 | * __memblock_find_range_top_down - find free area utility, in top-down |
7bd0b0f0 | 207 | * @start: start of candidate range |
47cec443 MR |
208 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
209 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
210 | * @size: size of free area to find |
211 | * @align: alignment of free area to find | |
b1154233 | 212 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 213 | * @flags: pick from blocks based on memory attributes |
7bd0b0f0 | 214 | * |
1402899e | 215 | * Utility called from memblock_find_in_range_node(), find free area top-down. |
7bd0b0f0 | 216 | * |
47cec443 | 217 | * Return: |
79442ed1 | 218 | * Found address on success, 0 on failure. |
6ed311b2 | 219 | */ |
1402899e TC |
220 | static phys_addr_t __init_memblock |
221 | __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 222 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 223 | enum memblock_flags flags) |
f7210e6c TC |
224 | { |
225 | phys_addr_t this_start, this_end, cand; | |
226 | u64 i; | |
227 | ||
fc6daaf9 TL |
228 | for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, |
229 | NULL) { | |
f7210e6c TC |
230 | this_start = clamp(this_start, start, end); |
231 | this_end = clamp(this_end, start, end); | |
232 | ||
233 | if (this_end < size) | |
234 | continue; | |
235 | ||
236 | cand = round_down(this_end - size, align); | |
237 | if (cand >= this_start) | |
238 | return cand; | |
239 | } | |
1402899e | 240 | |
f7210e6c TC |
241 | return 0; |
242 | } | |
6ed311b2 | 243 | |
1402899e TC |
244 | /** |
245 | * memblock_find_in_range_node - find free area in given range and node | |
1402899e TC |
246 | * @size: size of free area to find |
247 | * @align: alignment of free area to find | |
87029ee9 | 248 | * @start: start of candidate range |
47cec443 MR |
249 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
250 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
b1154233 | 251 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 252 | * @flags: pick from blocks based on memory attributes |
1402899e TC |
253 | * |
254 | * Find @size free area aligned to @align in the specified range and node. | |
255 | * | |
79442ed1 TC |
256 | * When allocation direction is bottom-up, the @start should be greater |
257 | * than the end of the kernel image. Otherwise, it will be trimmed. The | |
258 | * reason is that we want the bottom-up allocation just near the kernel | |
259 | * image so it is highly likely that the allocated memory and the kernel | |
260 | * will reside in the same node. | |
261 | * | |
262 | * If bottom-up allocation failed, will try to allocate memory top-down. | |
263 | * | |
47cec443 | 264 | * Return: |
79442ed1 | 265 | * Found address on success, 0 on failure. |
1402899e | 266 | */ |
c366ea89 | 267 | static phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, |
87029ee9 | 268 | phys_addr_t align, phys_addr_t start, |
e1720fee MR |
269 | phys_addr_t end, int nid, |
270 | enum memblock_flags flags) | |
1402899e | 271 | { |
0cfb8f0c | 272 | phys_addr_t kernel_end, ret; |
79442ed1 | 273 | |
1402899e | 274 | /* pump up @end */ |
fed84c78 QC |
275 | if (end == MEMBLOCK_ALLOC_ACCESSIBLE || |
276 | end == MEMBLOCK_ALLOC_KASAN) | |
1402899e TC |
277 | end = memblock.current_limit; |
278 | ||
279 | /* avoid allocating the first page */ | |
280 | start = max_t(phys_addr_t, start, PAGE_SIZE); | |
281 | end = max(start, end); | |
79442ed1 TC |
282 | kernel_end = __pa_symbol(_end); |
283 | ||
284 | /* | |
285 | * try bottom-up allocation only when bottom-up mode | |
286 | * is set and @end is above the kernel image. | |
287 | */ | |
288 | if (memblock_bottom_up() && end > kernel_end) { | |
289 | phys_addr_t bottom_up_start; | |
290 | ||
291 | /* make sure we will allocate above the kernel */ | |
292 | bottom_up_start = max(start, kernel_end); | |
293 | ||
294 | /* ok, try bottom-up allocation first */ | |
295 | ret = __memblock_find_range_bottom_up(bottom_up_start, end, | |
fc6daaf9 | 296 | size, align, nid, flags); |
79442ed1 TC |
297 | if (ret) |
298 | return ret; | |
299 | ||
300 | /* | |
301 | * we always limit bottom-up allocation above the kernel, | |
302 | * but top-down allocation doesn't have the limit, so | |
303 | * retrying top-down allocation may succeed when bottom-up | |
304 | * allocation failed. | |
305 | * | |
306 | * bottom-up allocation is expected to be fail very rarely, | |
307 | * so we use WARN_ONCE() here to see the stack trace if | |
308 | * fail happens. | |
309 | */ | |
e3d301ca MH |
310 | WARN_ONCE(IS_ENABLED(CONFIG_MEMORY_HOTREMOVE), |
311 | "memblock: bottom-up allocation failed, memory hotremove may be affected\n"); | |
79442ed1 | 312 | } |
1402899e | 313 | |
fc6daaf9 TL |
314 | return __memblock_find_range_top_down(start, end, size, align, nid, |
315 | flags); | |
1402899e TC |
316 | } |
317 | ||
7bd0b0f0 TH |
318 | /** |
319 | * memblock_find_in_range - find free area in given range | |
320 | * @start: start of candidate range | |
47cec443 MR |
321 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
322 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
323 | * @size: size of free area to find |
324 | * @align: alignment of free area to find | |
325 | * | |
326 | * Find @size free area aligned to @align in the specified range. | |
327 | * | |
47cec443 | 328 | * Return: |
79442ed1 | 329 | * Found address on success, 0 on failure. |
fc769a8e | 330 | */ |
7bd0b0f0 TH |
331 | phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, |
332 | phys_addr_t end, phys_addr_t size, | |
333 | phys_addr_t align) | |
6ed311b2 | 334 | { |
a3f5bafc | 335 | phys_addr_t ret; |
e1720fee | 336 | enum memblock_flags flags = choose_memblock_flags(); |
a3f5bafc TL |
337 | |
338 | again: | |
339 | ret = memblock_find_in_range_node(size, align, start, end, | |
340 | NUMA_NO_NODE, flags); | |
341 | ||
342 | if (!ret && (flags & MEMBLOCK_MIRROR)) { | |
343 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
344 | &size); | |
345 | flags &= ~MEMBLOCK_MIRROR; | |
346 | goto again; | |
347 | } | |
348 | ||
349 | return ret; | |
6ed311b2 BH |
350 | } |
351 | ||
10d06439 | 352 | static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) |
95f72d1e | 353 | { |
1440c4e2 | 354 | type->total_size -= type->regions[r].size; |
7c0caeb8 TH |
355 | memmove(&type->regions[r], &type->regions[r + 1], |
356 | (type->cnt - (r + 1)) * sizeof(type->regions[r])); | |
e3239ff9 | 357 | type->cnt--; |
95f72d1e | 358 | |
8f7a6605 BH |
359 | /* Special case for empty arrays */ |
360 | if (type->cnt == 0) { | |
1440c4e2 | 361 | WARN_ON(type->total_size != 0); |
8f7a6605 BH |
362 | type->cnt = 1; |
363 | type->regions[0].base = 0; | |
364 | type->regions[0].size = 0; | |
66a20757 | 365 | type->regions[0].flags = 0; |
7c0caeb8 | 366 | memblock_set_region_node(&type->regions[0], MAX_NUMNODES); |
8f7a6605 | 367 | } |
95f72d1e YL |
368 | } |
369 | ||
350e88ba | 370 | #ifndef CONFIG_ARCH_KEEP_MEMBLOCK |
3010f876 | 371 | /** |
47cec443 | 372 | * memblock_discard - discard memory and reserved arrays if they were allocated |
3010f876 PT |
373 | */ |
374 | void __init memblock_discard(void) | |
5e270e25 | 375 | { |
3010f876 | 376 | phys_addr_t addr, size; |
5e270e25 | 377 | |
3010f876 PT |
378 | if (memblock.reserved.regions != memblock_reserved_init_regions) { |
379 | addr = __pa(memblock.reserved.regions); | |
380 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
381 | memblock.reserved.max); | |
382 | __memblock_free_late(addr, size); | |
383 | } | |
5e270e25 | 384 | |
91b540f9 | 385 | if (memblock.memory.regions != memblock_memory_init_regions) { |
3010f876 PT |
386 | addr = __pa(memblock.memory.regions); |
387 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
388 | memblock.memory.max); | |
389 | __memblock_free_late(addr, size); | |
390 | } | |
5e270e25 | 391 | } |
5e270e25 PH |
392 | #endif |
393 | ||
48c3b583 GP |
394 | /** |
395 | * memblock_double_array - double the size of the memblock regions array | |
396 | * @type: memblock type of the regions array being doubled | |
397 | * @new_area_start: starting address of memory range to avoid overlap with | |
398 | * @new_area_size: size of memory range to avoid overlap with | |
399 | * | |
400 | * Double the size of the @type regions array. If memblock is being used to | |
401 | * allocate memory for a new reserved regions array and there is a previously | |
47cec443 | 402 | * allocated memory range [@new_area_start, @new_area_start + @new_area_size] |
48c3b583 GP |
403 | * waiting to be reserved, ensure the memory used by the new array does |
404 | * not overlap. | |
405 | * | |
47cec443 | 406 | * Return: |
48c3b583 GP |
407 | * 0 on success, -1 on failure. |
408 | */ | |
409 | static int __init_memblock memblock_double_array(struct memblock_type *type, | |
410 | phys_addr_t new_area_start, | |
411 | phys_addr_t new_area_size) | |
142b45a7 BH |
412 | { |
413 | struct memblock_region *new_array, *old_array; | |
29f67386 | 414 | phys_addr_t old_alloc_size, new_alloc_size; |
a36aab89 | 415 | phys_addr_t old_size, new_size, addr, new_end; |
142b45a7 | 416 | int use_slab = slab_is_available(); |
181eb394 | 417 | int *in_slab; |
142b45a7 BH |
418 | |
419 | /* We don't allow resizing until we know about the reserved regions | |
420 | * of memory that aren't suitable for allocation | |
421 | */ | |
422 | if (!memblock_can_resize) | |
423 | return -1; | |
424 | ||
142b45a7 BH |
425 | /* Calculate new doubled size */ |
426 | old_size = type->max * sizeof(struct memblock_region); | |
427 | new_size = old_size << 1; | |
29f67386 YL |
428 | /* |
429 | * We need to allocated new one align to PAGE_SIZE, | |
430 | * so we can free them completely later. | |
431 | */ | |
432 | old_alloc_size = PAGE_ALIGN(old_size); | |
433 | new_alloc_size = PAGE_ALIGN(new_size); | |
142b45a7 | 434 | |
181eb394 GS |
435 | /* Retrieve the slab flag */ |
436 | if (type == &memblock.memory) | |
437 | in_slab = &memblock_memory_in_slab; | |
438 | else | |
439 | in_slab = &memblock_reserved_in_slab; | |
440 | ||
a2974133 | 441 | /* Try to find some space for it */ |
142b45a7 BH |
442 | if (use_slab) { |
443 | new_array = kmalloc(new_size, GFP_KERNEL); | |
1f5026a7 | 444 | addr = new_array ? __pa(new_array) : 0; |
4e2f0775 | 445 | } else { |
48c3b583 GP |
446 | /* only exclude range when trying to double reserved.regions */ |
447 | if (type != &memblock.reserved) | |
448 | new_area_start = new_area_size = 0; | |
449 | ||
450 | addr = memblock_find_in_range(new_area_start + new_area_size, | |
451 | memblock.current_limit, | |
29f67386 | 452 | new_alloc_size, PAGE_SIZE); |
48c3b583 GP |
453 | if (!addr && new_area_size) |
454 | addr = memblock_find_in_range(0, | |
fd07383b AM |
455 | min(new_area_start, memblock.current_limit), |
456 | new_alloc_size, PAGE_SIZE); | |
48c3b583 | 457 | |
15674868 | 458 | new_array = addr ? __va(addr) : NULL; |
4e2f0775 | 459 | } |
1f5026a7 | 460 | if (!addr) { |
142b45a7 | 461 | pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", |
0262d9c8 | 462 | type->name, type->max, type->max * 2); |
142b45a7 BH |
463 | return -1; |
464 | } | |
142b45a7 | 465 | |
a36aab89 MR |
466 | new_end = addr + new_size - 1; |
467 | memblock_dbg("memblock: %s is doubled to %ld at [%pa-%pa]", | |
468 | type->name, type->max * 2, &addr, &new_end); | |
ea9e4376 | 469 | |
fd07383b AM |
470 | /* |
471 | * Found space, we now need to move the array over before we add the | |
472 | * reserved region since it may be our reserved array itself that is | |
473 | * full. | |
142b45a7 BH |
474 | */ |
475 | memcpy(new_array, type->regions, old_size); | |
476 | memset(new_array + type->max, 0, old_size); | |
477 | old_array = type->regions; | |
478 | type->regions = new_array; | |
479 | type->max <<= 1; | |
480 | ||
fd07383b | 481 | /* Free old array. We needn't free it if the array is the static one */ |
181eb394 GS |
482 | if (*in_slab) |
483 | kfree(old_array); | |
484 | else if (old_array != memblock_memory_init_regions && | |
485 | old_array != memblock_reserved_init_regions) | |
29f67386 | 486 | memblock_free(__pa(old_array), old_alloc_size); |
142b45a7 | 487 | |
fd07383b AM |
488 | /* |
489 | * Reserve the new array if that comes from the memblock. Otherwise, we | |
490 | * needn't do it | |
181eb394 GS |
491 | */ |
492 | if (!use_slab) | |
29f67386 | 493 | BUG_ON(memblock_reserve(addr, new_alloc_size)); |
181eb394 GS |
494 | |
495 | /* Update slab flag */ | |
496 | *in_slab = use_slab; | |
497 | ||
142b45a7 BH |
498 | return 0; |
499 | } | |
500 | ||
784656f9 TH |
501 | /** |
502 | * memblock_merge_regions - merge neighboring compatible regions | |
503 | * @type: memblock type to scan | |
504 | * | |
505 | * Scan @type and merge neighboring compatible regions. | |
506 | */ | |
507 | static void __init_memblock memblock_merge_regions(struct memblock_type *type) | |
95f72d1e | 508 | { |
784656f9 | 509 | int i = 0; |
95f72d1e | 510 | |
784656f9 TH |
511 | /* cnt never goes below 1 */ |
512 | while (i < type->cnt - 1) { | |
513 | struct memblock_region *this = &type->regions[i]; | |
514 | struct memblock_region *next = &type->regions[i + 1]; | |
95f72d1e | 515 | |
7c0caeb8 TH |
516 | if (this->base + this->size != next->base || |
517 | memblock_get_region_node(this) != | |
66a20757 TC |
518 | memblock_get_region_node(next) || |
519 | this->flags != next->flags) { | |
784656f9 TH |
520 | BUG_ON(this->base + this->size > next->base); |
521 | i++; | |
522 | continue; | |
8f7a6605 BH |
523 | } |
524 | ||
784656f9 | 525 | this->size += next->size; |
c0232ae8 LF |
526 | /* move forward from next + 1, index of which is i + 2 */ |
527 | memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); | |
784656f9 | 528 | type->cnt--; |
95f72d1e | 529 | } |
784656f9 | 530 | } |
95f72d1e | 531 | |
784656f9 TH |
532 | /** |
533 | * memblock_insert_region - insert new memblock region | |
209ff86d TC |
534 | * @type: memblock type to insert into |
535 | * @idx: index for the insertion point | |
536 | * @base: base address of the new region | |
537 | * @size: size of the new region | |
538 | * @nid: node id of the new region | |
66a20757 | 539 | * @flags: flags of the new region |
784656f9 | 540 | * |
47cec443 | 541 | * Insert new memblock region [@base, @base + @size) into @type at @idx. |
412d0008 | 542 | * @type must already have extra room to accommodate the new region. |
784656f9 TH |
543 | */ |
544 | static void __init_memblock memblock_insert_region(struct memblock_type *type, | |
545 | int idx, phys_addr_t base, | |
66a20757 | 546 | phys_addr_t size, |
e1720fee MR |
547 | int nid, |
548 | enum memblock_flags flags) | |
784656f9 TH |
549 | { |
550 | struct memblock_region *rgn = &type->regions[idx]; | |
551 | ||
552 | BUG_ON(type->cnt >= type->max); | |
553 | memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); | |
554 | rgn->base = base; | |
555 | rgn->size = size; | |
66a20757 | 556 | rgn->flags = flags; |
7c0caeb8 | 557 | memblock_set_region_node(rgn, nid); |
784656f9 | 558 | type->cnt++; |
1440c4e2 | 559 | type->total_size += size; |
784656f9 TH |
560 | } |
561 | ||
562 | /** | |
f1af9d3a | 563 | * memblock_add_range - add new memblock region |
784656f9 TH |
564 | * @type: memblock type to add new region into |
565 | * @base: base address of the new region | |
566 | * @size: size of the new region | |
7fb0bc3f | 567 | * @nid: nid of the new region |
66a20757 | 568 | * @flags: flags of the new region |
784656f9 | 569 | * |
47cec443 | 570 | * Add new memblock region [@base, @base + @size) into @type. The new region |
784656f9 TH |
571 | * is allowed to overlap with existing ones - overlaps don't affect already |
572 | * existing regions. @type is guaranteed to be minimal (all neighbouring | |
573 | * compatible regions are merged) after the addition. | |
574 | * | |
47cec443 | 575 | * Return: |
784656f9 TH |
576 | * 0 on success, -errno on failure. |
577 | */ | |
f1af9d3a | 578 | int __init_memblock memblock_add_range(struct memblock_type *type, |
66a20757 | 579 | phys_addr_t base, phys_addr_t size, |
e1720fee | 580 | int nid, enum memblock_flags flags) |
784656f9 TH |
581 | { |
582 | bool insert = false; | |
eb18f1b5 TH |
583 | phys_addr_t obase = base; |
584 | phys_addr_t end = base + memblock_cap_size(base, &size); | |
8c9c1701 AK |
585 | int idx, nr_new; |
586 | struct memblock_region *rgn; | |
784656f9 | 587 | |
b3dc627c TH |
588 | if (!size) |
589 | return 0; | |
590 | ||
784656f9 TH |
591 | /* special case for empty array */ |
592 | if (type->regions[0].size == 0) { | |
1440c4e2 | 593 | WARN_ON(type->cnt != 1 || type->total_size); |
8f7a6605 BH |
594 | type->regions[0].base = base; |
595 | type->regions[0].size = size; | |
66a20757 | 596 | type->regions[0].flags = flags; |
7fb0bc3f | 597 | memblock_set_region_node(&type->regions[0], nid); |
1440c4e2 | 598 | type->total_size = size; |
8f7a6605 | 599 | return 0; |
95f72d1e | 600 | } |
784656f9 TH |
601 | repeat: |
602 | /* | |
603 | * The following is executed twice. Once with %false @insert and | |
604 | * then with %true. The first counts the number of regions needed | |
412d0008 | 605 | * to accommodate the new area. The second actually inserts them. |
142b45a7 | 606 | */ |
784656f9 TH |
607 | base = obase; |
608 | nr_new = 0; | |
95f72d1e | 609 | |
66e8b438 | 610 | for_each_memblock_type(idx, type, rgn) { |
784656f9 TH |
611 | phys_addr_t rbase = rgn->base; |
612 | phys_addr_t rend = rbase + rgn->size; | |
613 | ||
614 | if (rbase >= end) | |
95f72d1e | 615 | break; |
784656f9 TH |
616 | if (rend <= base) |
617 | continue; | |
618 | /* | |
619 | * @rgn overlaps. If it separates the lower part of new | |
620 | * area, insert that portion. | |
621 | */ | |
622 | if (rbase > base) { | |
c0a29498 WY |
623 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
624 | WARN_ON(nid != memblock_get_region_node(rgn)); | |
625 | #endif | |
4fcab5f4 | 626 | WARN_ON(flags != rgn->flags); |
784656f9 TH |
627 | nr_new++; |
628 | if (insert) | |
8c9c1701 | 629 | memblock_insert_region(type, idx++, base, |
66a20757 TC |
630 | rbase - base, nid, |
631 | flags); | |
95f72d1e | 632 | } |
784656f9 TH |
633 | /* area below @rend is dealt with, forget about it */ |
634 | base = min(rend, end); | |
95f72d1e | 635 | } |
784656f9 TH |
636 | |
637 | /* insert the remaining portion */ | |
638 | if (base < end) { | |
639 | nr_new++; | |
640 | if (insert) | |
8c9c1701 | 641 | memblock_insert_region(type, idx, base, end - base, |
66a20757 | 642 | nid, flags); |
95f72d1e | 643 | } |
95f72d1e | 644 | |
ef3cc4db | 645 | if (!nr_new) |
646 | return 0; | |
647 | ||
784656f9 TH |
648 | /* |
649 | * If this was the first round, resize array and repeat for actual | |
650 | * insertions; otherwise, merge and return. | |
142b45a7 | 651 | */ |
784656f9 TH |
652 | if (!insert) { |
653 | while (type->cnt + nr_new > type->max) | |
48c3b583 | 654 | if (memblock_double_array(type, obase, size) < 0) |
784656f9 TH |
655 | return -ENOMEM; |
656 | insert = true; | |
657 | goto repeat; | |
658 | } else { | |
659 | memblock_merge_regions(type); | |
660 | return 0; | |
142b45a7 | 661 | } |
95f72d1e YL |
662 | } |
663 | ||
48a833cc MR |
664 | /** |
665 | * memblock_add_node - add new memblock region within a NUMA node | |
666 | * @base: base address of the new region | |
667 | * @size: size of the new region | |
668 | * @nid: nid of the new region | |
669 | * | |
670 | * Add new memblock region [@base, @base + @size) to the "memory" | |
671 | * type. See memblock_add_range() description for mode details | |
672 | * | |
673 | * Return: | |
674 | * 0 on success, -errno on failure. | |
675 | */ | |
7fb0bc3f TH |
676 | int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, |
677 | int nid) | |
678 | { | |
f1af9d3a | 679 | return memblock_add_range(&memblock.memory, base, size, nid, 0); |
7fb0bc3f TH |
680 | } |
681 | ||
48a833cc MR |
682 | /** |
683 | * memblock_add - add new memblock region | |
684 | * @base: base address of the new region | |
685 | * @size: size of the new region | |
686 | * | |
687 | * Add new memblock region [@base, @base + @size) to the "memory" | |
688 | * type. See memblock_add_range() description for mode details | |
689 | * | |
690 | * Return: | |
691 | * 0 on success, -errno on failure. | |
692 | */ | |
f705ac4b | 693 | int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) |
6a4055bc | 694 | { |
5d63f81c MC |
695 | phys_addr_t end = base + size - 1; |
696 | ||
d75f773c | 697 | memblock_dbg("memblock_add: [%pa-%pa] %pS\n", |
5d63f81c | 698 | &base, &end, (void *)_RET_IP_); |
6a4055bc | 699 | |
f705ac4b | 700 | return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
701 | } |
702 | ||
6a9ceb31 TH |
703 | /** |
704 | * memblock_isolate_range - isolate given range into disjoint memblocks | |
705 | * @type: memblock type to isolate range for | |
706 | * @base: base of range to isolate | |
707 | * @size: size of range to isolate | |
708 | * @start_rgn: out parameter for the start of isolated region | |
709 | * @end_rgn: out parameter for the end of isolated region | |
710 | * | |
711 | * Walk @type and ensure that regions don't cross the boundaries defined by | |
47cec443 | 712 | * [@base, @base + @size). Crossing regions are split at the boundaries, |
6a9ceb31 TH |
713 | * which may create at most two more regions. The index of the first |
714 | * region inside the range is returned in *@start_rgn and end in *@end_rgn. | |
715 | * | |
47cec443 | 716 | * Return: |
6a9ceb31 TH |
717 | * 0 on success, -errno on failure. |
718 | */ | |
719 | static int __init_memblock memblock_isolate_range(struct memblock_type *type, | |
720 | phys_addr_t base, phys_addr_t size, | |
721 | int *start_rgn, int *end_rgn) | |
722 | { | |
eb18f1b5 | 723 | phys_addr_t end = base + memblock_cap_size(base, &size); |
8c9c1701 AK |
724 | int idx; |
725 | struct memblock_region *rgn; | |
6a9ceb31 TH |
726 | |
727 | *start_rgn = *end_rgn = 0; | |
728 | ||
b3dc627c TH |
729 | if (!size) |
730 | return 0; | |
731 | ||
6a9ceb31 TH |
732 | /* we'll create at most two more regions */ |
733 | while (type->cnt + 2 > type->max) | |
48c3b583 | 734 | if (memblock_double_array(type, base, size) < 0) |
6a9ceb31 TH |
735 | return -ENOMEM; |
736 | ||
66e8b438 | 737 | for_each_memblock_type(idx, type, rgn) { |
6a9ceb31 TH |
738 | phys_addr_t rbase = rgn->base; |
739 | phys_addr_t rend = rbase + rgn->size; | |
740 | ||
741 | if (rbase >= end) | |
742 | break; | |
743 | if (rend <= base) | |
744 | continue; | |
745 | ||
746 | if (rbase < base) { | |
747 | /* | |
748 | * @rgn intersects from below. Split and continue | |
749 | * to process the next region - the new top half. | |
750 | */ | |
751 | rgn->base = base; | |
1440c4e2 TH |
752 | rgn->size -= base - rbase; |
753 | type->total_size -= base - rbase; | |
8c9c1701 | 754 | memblock_insert_region(type, idx, rbase, base - rbase, |
66a20757 TC |
755 | memblock_get_region_node(rgn), |
756 | rgn->flags); | |
6a9ceb31 TH |
757 | } else if (rend > end) { |
758 | /* | |
759 | * @rgn intersects from above. Split and redo the | |
760 | * current region - the new bottom half. | |
761 | */ | |
762 | rgn->base = end; | |
1440c4e2 TH |
763 | rgn->size -= end - rbase; |
764 | type->total_size -= end - rbase; | |
8c9c1701 | 765 | memblock_insert_region(type, idx--, rbase, end - rbase, |
66a20757 TC |
766 | memblock_get_region_node(rgn), |
767 | rgn->flags); | |
6a9ceb31 TH |
768 | } else { |
769 | /* @rgn is fully contained, record it */ | |
770 | if (!*end_rgn) | |
8c9c1701 AK |
771 | *start_rgn = idx; |
772 | *end_rgn = idx + 1; | |
6a9ceb31 TH |
773 | } |
774 | } | |
775 | ||
776 | return 0; | |
777 | } | |
6a9ceb31 | 778 | |
35bd16a2 | 779 | static int __init_memblock memblock_remove_range(struct memblock_type *type, |
f1af9d3a | 780 | phys_addr_t base, phys_addr_t size) |
95f72d1e | 781 | { |
71936180 TH |
782 | int start_rgn, end_rgn; |
783 | int i, ret; | |
95f72d1e | 784 | |
71936180 TH |
785 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
786 | if (ret) | |
787 | return ret; | |
95f72d1e | 788 | |
71936180 TH |
789 | for (i = end_rgn - 1; i >= start_rgn; i--) |
790 | memblock_remove_region(type, i); | |
8f7a6605 | 791 | return 0; |
95f72d1e YL |
792 | } |
793 | ||
581adcbe | 794 | int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) |
95f72d1e | 795 | { |
25cf23d7 MK |
796 | phys_addr_t end = base + size - 1; |
797 | ||
798 | memblock_dbg("memblock_remove: [%pa-%pa] %pS\n", | |
799 | &base, &end, (void *)_RET_IP_); | |
800 | ||
f1af9d3a | 801 | return memblock_remove_range(&memblock.memory, base, size); |
95f72d1e YL |
802 | } |
803 | ||
4d72868c MR |
804 | /** |
805 | * memblock_free - free boot memory block | |
806 | * @base: phys starting address of the boot memory block | |
807 | * @size: size of the boot memory block in bytes | |
808 | * | |
809 | * Free boot memory block previously allocated by memblock_alloc_xx() API. | |
810 | * The freeing memory will not be released to the buddy allocator. | |
811 | */ | |
581adcbe | 812 | int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size) |
95f72d1e | 813 | { |
5d63f81c MC |
814 | phys_addr_t end = base + size - 1; |
815 | ||
d75f773c | 816 | memblock_dbg(" memblock_free: [%pa-%pa] %pS\n", |
5d63f81c | 817 | &base, &end, (void *)_RET_IP_); |
24aa0788 | 818 | |
9099daed | 819 | kmemleak_free_part_phys(base, size); |
f1af9d3a | 820 | return memblock_remove_range(&memblock.reserved, base, size); |
95f72d1e YL |
821 | } |
822 | ||
f705ac4b | 823 | int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size) |
95f72d1e | 824 | { |
5d63f81c MC |
825 | phys_addr_t end = base + size - 1; |
826 | ||
d75f773c | 827 | memblock_dbg("memblock_reserve: [%pa-%pa] %pS\n", |
5d63f81c | 828 | &base, &end, (void *)_RET_IP_); |
95f72d1e | 829 | |
f705ac4b | 830 | return memblock_add_range(&memblock.reserved, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
831 | } |
832 | ||
66b16edf | 833 | /** |
47cec443 MR |
834 | * memblock_setclr_flag - set or clear flag for a memory region |
835 | * @base: base address of the region | |
836 | * @size: size of the region | |
837 | * @set: set or clear the flag | |
838 | * @flag: the flag to udpate | |
66b16edf | 839 | * |
4308ce17 | 840 | * This function isolates region [@base, @base + @size), and sets/clears flag |
66b16edf | 841 | * |
47cec443 | 842 | * Return: 0 on success, -errno on failure. |
66b16edf | 843 | */ |
4308ce17 TL |
844 | static int __init_memblock memblock_setclr_flag(phys_addr_t base, |
845 | phys_addr_t size, int set, int flag) | |
66b16edf TC |
846 | { |
847 | struct memblock_type *type = &memblock.memory; | |
848 | int i, ret, start_rgn, end_rgn; | |
849 | ||
850 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); | |
851 | if (ret) | |
852 | return ret; | |
853 | ||
fe145124 MR |
854 | for (i = start_rgn; i < end_rgn; i++) { |
855 | struct memblock_region *r = &type->regions[i]; | |
856 | ||
4308ce17 | 857 | if (set) |
fe145124 | 858 | r->flags |= flag; |
4308ce17 | 859 | else |
fe145124 MR |
860 | r->flags &= ~flag; |
861 | } | |
66b16edf TC |
862 | |
863 | memblock_merge_regions(type); | |
864 | return 0; | |
865 | } | |
866 | ||
867 | /** | |
4308ce17 | 868 | * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. |
66b16edf TC |
869 | * @base: the base phys addr of the region |
870 | * @size: the size of the region | |
871 | * | |
47cec443 | 872 | * Return: 0 on success, -errno on failure. |
4308ce17 TL |
873 | */ |
874 | int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) | |
875 | { | |
876 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); | |
877 | } | |
878 | ||
879 | /** | |
880 | * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. | |
881 | * @base: the base phys addr of the region | |
882 | * @size: the size of the region | |
66b16edf | 883 | * |
47cec443 | 884 | * Return: 0 on success, -errno on failure. |
66b16edf TC |
885 | */ |
886 | int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) | |
887 | { | |
4308ce17 | 888 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); |
66b16edf TC |
889 | } |
890 | ||
a3f5bafc TL |
891 | /** |
892 | * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. | |
893 | * @base: the base phys addr of the region | |
894 | * @size: the size of the region | |
895 | * | |
47cec443 | 896 | * Return: 0 on success, -errno on failure. |
a3f5bafc TL |
897 | */ |
898 | int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) | |
899 | { | |
900 | system_has_some_mirror = true; | |
901 | ||
902 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); | |
903 | } | |
904 | ||
bf3d3cc5 AB |
905 | /** |
906 | * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. | |
907 | * @base: the base phys addr of the region | |
908 | * @size: the size of the region | |
909 | * | |
47cec443 | 910 | * Return: 0 on success, -errno on failure. |
bf3d3cc5 AB |
911 | */ |
912 | int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) | |
913 | { | |
914 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); | |
915 | } | |
a3f5bafc | 916 | |
4c546b8a AT |
917 | /** |
918 | * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. | |
919 | * @base: the base phys addr of the region | |
920 | * @size: the size of the region | |
921 | * | |
47cec443 | 922 | * Return: 0 on success, -errno on failure. |
4c546b8a AT |
923 | */ |
924 | int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) | |
925 | { | |
926 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); | |
927 | } | |
928 | ||
8e7a7f86 RH |
929 | /** |
930 | * __next_reserved_mem_region - next function for for_each_reserved_region() | |
931 | * @idx: pointer to u64 loop variable | |
932 | * @out_start: ptr to phys_addr_t for start address of the region, can be %NULL | |
933 | * @out_end: ptr to phys_addr_t for end address of the region, can be %NULL | |
934 | * | |
935 | * Iterate over all reserved memory regions. | |
936 | */ | |
937 | void __init_memblock __next_reserved_mem_region(u64 *idx, | |
938 | phys_addr_t *out_start, | |
939 | phys_addr_t *out_end) | |
940 | { | |
567d117b | 941 | struct memblock_type *type = &memblock.reserved; |
8e7a7f86 | 942 | |
cd33a76b | 943 | if (*idx < type->cnt) { |
567d117b | 944 | struct memblock_region *r = &type->regions[*idx]; |
8e7a7f86 RH |
945 | phys_addr_t base = r->base; |
946 | phys_addr_t size = r->size; | |
947 | ||
948 | if (out_start) | |
949 | *out_start = base; | |
950 | if (out_end) | |
951 | *out_end = base + size - 1; | |
952 | ||
953 | *idx += 1; | |
954 | return; | |
955 | } | |
956 | ||
957 | /* signal end of iteration */ | |
958 | *idx = ULLONG_MAX; | |
959 | } | |
960 | ||
c9a688a3 MR |
961 | static bool should_skip_region(struct memblock_region *m, int nid, int flags) |
962 | { | |
963 | int m_nid = memblock_get_region_node(m); | |
964 | ||
965 | /* only memory regions are associated with nodes, check it */ | |
966 | if (nid != NUMA_NO_NODE && nid != m_nid) | |
967 | return true; | |
968 | ||
969 | /* skip hotpluggable memory regions if needed */ | |
970 | if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) | |
971 | return true; | |
972 | ||
973 | /* if we want mirror memory skip non-mirror memory regions */ | |
974 | if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) | |
975 | return true; | |
976 | ||
977 | /* skip nomap memory unless we were asked for it explicitly */ | |
978 | if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) | |
979 | return true; | |
980 | ||
981 | return false; | |
982 | } | |
983 | ||
35fd0808 | 984 | /** |
a2974133 | 985 | * __next_mem_range - next function for for_each_free_mem_range() etc. |
35fd0808 | 986 | * @idx: pointer to u64 loop variable |
b1154233 | 987 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 988 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
989 | * @type_a: pointer to memblock_type from where the range is taken |
990 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
991 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
992 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
993 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
35fd0808 | 994 | * |
f1af9d3a | 995 | * Find the first area from *@idx which matches @nid, fill the out |
35fd0808 | 996 | * parameters, and update *@idx for the next iteration. The lower 32bit of |
f1af9d3a PH |
997 | * *@idx contains index into type_a and the upper 32bit indexes the |
998 | * areas before each region in type_b. For example, if type_b regions | |
35fd0808 TH |
999 | * look like the following, |
1000 | * | |
1001 | * 0:[0-16), 1:[32-48), 2:[128-130) | |
1002 | * | |
1003 | * The upper 32bit indexes the following regions. | |
1004 | * | |
1005 | * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) | |
1006 | * | |
1007 | * As both region arrays are sorted, the function advances the two indices | |
1008 | * in lockstep and returns each intersection. | |
1009 | */ | |
e1720fee MR |
1010 | void __init_memblock __next_mem_range(u64 *idx, int nid, |
1011 | enum memblock_flags flags, | |
f1af9d3a PH |
1012 | struct memblock_type *type_a, |
1013 | struct memblock_type *type_b, | |
1014 | phys_addr_t *out_start, | |
1015 | phys_addr_t *out_end, int *out_nid) | |
35fd0808 | 1016 | { |
f1af9d3a PH |
1017 | int idx_a = *idx & 0xffffffff; |
1018 | int idx_b = *idx >> 32; | |
b1154233 | 1019 | |
f1af9d3a PH |
1020 | if (WARN_ONCE(nid == MAX_NUMNODES, |
1021 | "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) | |
560dca27 | 1022 | nid = NUMA_NO_NODE; |
35fd0808 | 1023 | |
f1af9d3a PH |
1024 | for (; idx_a < type_a->cnt; idx_a++) { |
1025 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1026 | ||
35fd0808 TH |
1027 | phys_addr_t m_start = m->base; |
1028 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1029 | int m_nid = memblock_get_region_node(m); |
35fd0808 | 1030 | |
c9a688a3 | 1031 | if (should_skip_region(m, nid, flags)) |
bf3d3cc5 AB |
1032 | continue; |
1033 | ||
f1af9d3a PH |
1034 | if (!type_b) { |
1035 | if (out_start) | |
1036 | *out_start = m_start; | |
1037 | if (out_end) | |
1038 | *out_end = m_end; | |
1039 | if (out_nid) | |
1040 | *out_nid = m_nid; | |
1041 | idx_a++; | |
1042 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
1043 | return; | |
1044 | } | |
1045 | ||
1046 | /* scan areas before each reservation */ | |
1047 | for (; idx_b < type_b->cnt + 1; idx_b++) { | |
1048 | struct memblock_region *r; | |
1049 | phys_addr_t r_start; | |
1050 | phys_addr_t r_end; | |
1051 | ||
1052 | r = &type_b->regions[idx_b]; | |
1053 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1054 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1055 | r->base : PHYS_ADDR_MAX; |
35fd0808 | 1056 | |
f1af9d3a PH |
1057 | /* |
1058 | * if idx_b advanced past idx_a, | |
1059 | * break out to advance idx_a | |
1060 | */ | |
35fd0808 TH |
1061 | if (r_start >= m_end) |
1062 | break; | |
1063 | /* if the two regions intersect, we're done */ | |
1064 | if (m_start < r_end) { | |
1065 | if (out_start) | |
f1af9d3a PH |
1066 | *out_start = |
1067 | max(m_start, r_start); | |
35fd0808 TH |
1068 | if (out_end) |
1069 | *out_end = min(m_end, r_end); | |
1070 | if (out_nid) | |
f1af9d3a | 1071 | *out_nid = m_nid; |
35fd0808 | 1072 | /* |
f1af9d3a PH |
1073 | * The region which ends first is |
1074 | * advanced for the next iteration. | |
35fd0808 TH |
1075 | */ |
1076 | if (m_end <= r_end) | |
f1af9d3a | 1077 | idx_a++; |
35fd0808 | 1078 | else |
f1af9d3a PH |
1079 | idx_b++; |
1080 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
35fd0808 TH |
1081 | return; |
1082 | } | |
1083 | } | |
1084 | } | |
1085 | ||
1086 | /* signal end of iteration */ | |
1087 | *idx = ULLONG_MAX; | |
1088 | } | |
1089 | ||
7bd0b0f0 | 1090 | /** |
f1af9d3a PH |
1091 | * __next_mem_range_rev - generic next function for for_each_*_range_rev() |
1092 | * | |
7bd0b0f0 | 1093 | * @idx: pointer to u64 loop variable |
ad5ea8cd | 1094 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 1095 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
1096 | * @type_a: pointer to memblock_type from where the range is taken |
1097 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
1098 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
1099 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
1100 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
7bd0b0f0 | 1101 | * |
47cec443 MR |
1102 | * Finds the next range from type_a which is not marked as unsuitable |
1103 | * in type_b. | |
1104 | * | |
f1af9d3a | 1105 | * Reverse of __next_mem_range(). |
7bd0b0f0 | 1106 | */ |
e1720fee MR |
1107 | void __init_memblock __next_mem_range_rev(u64 *idx, int nid, |
1108 | enum memblock_flags flags, | |
f1af9d3a PH |
1109 | struct memblock_type *type_a, |
1110 | struct memblock_type *type_b, | |
1111 | phys_addr_t *out_start, | |
1112 | phys_addr_t *out_end, int *out_nid) | |
7bd0b0f0 | 1113 | { |
f1af9d3a PH |
1114 | int idx_a = *idx & 0xffffffff; |
1115 | int idx_b = *idx >> 32; | |
b1154233 | 1116 | |
560dca27 GS |
1117 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1118 | nid = NUMA_NO_NODE; | |
7bd0b0f0 TH |
1119 | |
1120 | if (*idx == (u64)ULLONG_MAX) { | |
f1af9d3a | 1121 | idx_a = type_a->cnt - 1; |
e47608ab | 1122 | if (type_b != NULL) |
1123 | idx_b = type_b->cnt; | |
1124 | else | |
1125 | idx_b = 0; | |
7bd0b0f0 TH |
1126 | } |
1127 | ||
f1af9d3a PH |
1128 | for (; idx_a >= 0; idx_a--) { |
1129 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1130 | ||
7bd0b0f0 TH |
1131 | phys_addr_t m_start = m->base; |
1132 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1133 | int m_nid = memblock_get_region_node(m); |
7bd0b0f0 | 1134 | |
c9a688a3 | 1135 | if (should_skip_region(m, nid, flags)) |
bf3d3cc5 AB |
1136 | continue; |
1137 | ||
f1af9d3a PH |
1138 | if (!type_b) { |
1139 | if (out_start) | |
1140 | *out_start = m_start; | |
1141 | if (out_end) | |
1142 | *out_end = m_end; | |
1143 | if (out_nid) | |
1144 | *out_nid = m_nid; | |
fb399b48 | 1145 | idx_a--; |
f1af9d3a PH |
1146 | *idx = (u32)idx_a | (u64)idx_b << 32; |
1147 | return; | |
1148 | } | |
1149 | ||
1150 | /* scan areas before each reservation */ | |
1151 | for (; idx_b >= 0; idx_b--) { | |
1152 | struct memblock_region *r; | |
1153 | phys_addr_t r_start; | |
1154 | phys_addr_t r_end; | |
1155 | ||
1156 | r = &type_b->regions[idx_b]; | |
1157 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1158 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1159 | r->base : PHYS_ADDR_MAX; |
f1af9d3a PH |
1160 | /* |
1161 | * if idx_b advanced past idx_a, | |
1162 | * break out to advance idx_a | |
1163 | */ | |
7bd0b0f0 | 1164 | |
7bd0b0f0 TH |
1165 | if (r_end <= m_start) |
1166 | break; | |
1167 | /* if the two regions intersect, we're done */ | |
1168 | if (m_end > r_start) { | |
1169 | if (out_start) | |
1170 | *out_start = max(m_start, r_start); | |
1171 | if (out_end) | |
1172 | *out_end = min(m_end, r_end); | |
1173 | if (out_nid) | |
f1af9d3a | 1174 | *out_nid = m_nid; |
7bd0b0f0 | 1175 | if (m_start >= r_start) |
f1af9d3a | 1176 | idx_a--; |
7bd0b0f0 | 1177 | else |
f1af9d3a PH |
1178 | idx_b--; |
1179 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
7bd0b0f0 TH |
1180 | return; |
1181 | } | |
1182 | } | |
1183 | } | |
f1af9d3a | 1184 | /* signal end of iteration */ |
7bd0b0f0 TH |
1185 | *idx = ULLONG_MAX; |
1186 | } | |
1187 | ||
7c0caeb8 TH |
1188 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1189 | /* | |
45e79815 | 1190 | * Common iterator interface used to define for_each_mem_pfn_range(). |
7c0caeb8 TH |
1191 | */ |
1192 | void __init_memblock __next_mem_pfn_range(int *idx, int nid, | |
1193 | unsigned long *out_start_pfn, | |
1194 | unsigned long *out_end_pfn, int *out_nid) | |
1195 | { | |
1196 | struct memblock_type *type = &memblock.memory; | |
1197 | struct memblock_region *r; | |
1198 | ||
1199 | while (++*idx < type->cnt) { | |
1200 | r = &type->regions[*idx]; | |
1201 | ||
1202 | if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) | |
1203 | continue; | |
1204 | if (nid == MAX_NUMNODES || nid == r->nid) | |
1205 | break; | |
1206 | } | |
1207 | if (*idx >= type->cnt) { | |
1208 | *idx = -1; | |
1209 | return; | |
1210 | } | |
1211 | ||
1212 | if (out_start_pfn) | |
1213 | *out_start_pfn = PFN_UP(r->base); | |
1214 | if (out_end_pfn) | |
1215 | *out_end_pfn = PFN_DOWN(r->base + r->size); | |
1216 | if (out_nid) | |
1217 | *out_nid = r->nid; | |
1218 | } | |
1219 | ||
1220 | /** | |
1221 | * memblock_set_node - set node ID on memblock regions | |
1222 | * @base: base of area to set node ID for | |
1223 | * @size: size of area to set node ID for | |
e7e8de59 | 1224 | * @type: memblock type to set node ID for |
7c0caeb8 TH |
1225 | * @nid: node ID to set |
1226 | * | |
47cec443 | 1227 | * Set the nid of memblock @type regions in [@base, @base + @size) to @nid. |
7c0caeb8 TH |
1228 | * Regions which cross the area boundaries are split as necessary. |
1229 | * | |
47cec443 | 1230 | * Return: |
7c0caeb8 TH |
1231 | * 0 on success, -errno on failure. |
1232 | */ | |
1233 | int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, | |
e7e8de59 | 1234 | struct memblock_type *type, int nid) |
7c0caeb8 | 1235 | { |
6a9ceb31 TH |
1236 | int start_rgn, end_rgn; |
1237 | int i, ret; | |
7c0caeb8 | 1238 | |
6a9ceb31 TH |
1239 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
1240 | if (ret) | |
1241 | return ret; | |
7c0caeb8 | 1242 | |
6a9ceb31 | 1243 | for (i = start_rgn; i < end_rgn; i++) |
e9d24ad3 | 1244 | memblock_set_region_node(&type->regions[i], nid); |
7c0caeb8 TH |
1245 | |
1246 | memblock_merge_regions(type); | |
1247 | return 0; | |
1248 | } | |
1249 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ | |
837566e7 AD |
1250 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT |
1251 | /** | |
1252 | * __next_mem_pfn_range_in_zone - iterator for for_each_*_range_in_zone() | |
1253 | * | |
1254 | * @idx: pointer to u64 loop variable | |
1255 | * @zone: zone in which all of the memory blocks reside | |
1256 | * @out_spfn: ptr to ulong for start pfn of the range, can be %NULL | |
1257 | * @out_epfn: ptr to ulong for end pfn of the range, can be %NULL | |
1258 | * | |
1259 | * This function is meant to be a zone/pfn specific wrapper for the | |
1260 | * for_each_mem_range type iterators. Specifically they are used in the | |
1261 | * deferred memory init routines and as such we were duplicating much of | |
1262 | * this logic throughout the code. So instead of having it in multiple | |
1263 | * locations it seemed like it would make more sense to centralize this to | |
1264 | * one new iterator that does everything they need. | |
1265 | */ | |
1266 | void __init_memblock | |
1267 | __next_mem_pfn_range_in_zone(u64 *idx, struct zone *zone, | |
1268 | unsigned long *out_spfn, unsigned long *out_epfn) | |
1269 | { | |
1270 | int zone_nid = zone_to_nid(zone); | |
1271 | phys_addr_t spa, epa; | |
1272 | int nid; | |
1273 | ||
1274 | __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, | |
1275 | &memblock.memory, &memblock.reserved, | |
1276 | &spa, &epa, &nid); | |
1277 | ||
1278 | while (*idx != U64_MAX) { | |
1279 | unsigned long epfn = PFN_DOWN(epa); | |
1280 | unsigned long spfn = PFN_UP(spa); | |
1281 | ||
1282 | /* | |
1283 | * Verify the end is at least past the start of the zone and | |
1284 | * that we have at least one PFN to initialize. | |
1285 | */ | |
1286 | if (zone->zone_start_pfn < epfn && spfn < epfn) { | |
1287 | /* if we went too far just stop searching */ | |
1288 | if (zone_end_pfn(zone) <= spfn) { | |
1289 | *idx = U64_MAX; | |
1290 | break; | |
1291 | } | |
1292 | ||
1293 | if (out_spfn) | |
1294 | *out_spfn = max(zone->zone_start_pfn, spfn); | |
1295 | if (out_epfn) | |
1296 | *out_epfn = min(zone_end_pfn(zone), epfn); | |
1297 | ||
1298 | return; | |
1299 | } | |
1300 | ||
1301 | __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, | |
1302 | &memblock.memory, &memblock.reserved, | |
1303 | &spa, &epa, &nid); | |
1304 | } | |
1305 | ||
1306 | /* signal end of iteration */ | |
1307 | if (out_spfn) | |
1308 | *out_spfn = ULONG_MAX; | |
1309 | if (out_epfn) | |
1310 | *out_epfn = 0; | |
1311 | } | |
1312 | ||
1313 | #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ | |
7c0caeb8 | 1314 | |
92d12f95 MR |
1315 | /** |
1316 | * memblock_alloc_range_nid - allocate boot memory block | |
1317 | * @size: size of memory block to be allocated in bytes | |
1318 | * @align: alignment of the region and block's size | |
1319 | * @start: the lower bound of the memory region to allocate (phys address) | |
1320 | * @end: the upper bound of the memory region to allocate (phys address) | |
1321 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
0ac398b1 | 1322 | * @exact_nid: control the allocation fall back to other nodes |
92d12f95 MR |
1323 | * |
1324 | * The allocation is performed from memory region limited by | |
95830666 | 1325 | * memblock.current_limit if @end == %MEMBLOCK_ALLOC_ACCESSIBLE. |
92d12f95 | 1326 | * |
0ac398b1 YY |
1327 | * If the specified node can not hold the requested memory and @exact_nid |
1328 | * is false, the allocation falls back to any node in the system. | |
92d12f95 MR |
1329 | * |
1330 | * For systems with memory mirroring, the allocation is attempted first | |
1331 | * from the regions with mirroring enabled and then retried from any | |
1332 | * memory region. | |
1333 | * | |
1334 | * In addition, function sets the min_count to 0 using kmemleak_alloc_phys for | |
1335 | * allocated boot memory block, so that it is never reported as leaks. | |
1336 | * | |
1337 | * Return: | |
1338 | * Physical address of allocated memory block on success, %0 on failure. | |
1339 | */ | |
2bfc2862 AM |
1340 | static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, |
1341 | phys_addr_t align, phys_addr_t start, | |
0ac398b1 YY |
1342 | phys_addr_t end, int nid, |
1343 | bool exact_nid) | |
95f72d1e | 1344 | { |
92d12f95 | 1345 | enum memblock_flags flags = choose_memblock_flags(); |
6ed311b2 | 1346 | phys_addr_t found; |
95f72d1e | 1347 | |
92d12f95 MR |
1348 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1349 | nid = NUMA_NO_NODE; | |
1350 | ||
2f770806 MR |
1351 | if (!align) { |
1352 | /* Can't use WARNs this early in boot on powerpc */ | |
1353 | dump_stack(); | |
1354 | align = SMP_CACHE_BYTES; | |
1355 | } | |
1356 | ||
92d12f95 | 1357 | again: |
fc6daaf9 TL |
1358 | found = memblock_find_in_range_node(size, align, start, end, nid, |
1359 | flags); | |
92d12f95 MR |
1360 | if (found && !memblock_reserve(found, size)) |
1361 | goto done; | |
1362 | ||
0ac398b1 | 1363 | if (nid != NUMA_NO_NODE && !exact_nid) { |
92d12f95 MR |
1364 | found = memblock_find_in_range_node(size, align, start, |
1365 | end, NUMA_NO_NODE, | |
1366 | flags); | |
1367 | if (found && !memblock_reserve(found, size)) | |
1368 | goto done; | |
1369 | } | |
1370 | ||
1371 | if (flags & MEMBLOCK_MIRROR) { | |
1372 | flags &= ~MEMBLOCK_MIRROR; | |
1373 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
1374 | &size); | |
1375 | goto again; | |
1376 | } | |
1377 | ||
1378 | return 0; | |
1379 | ||
1380 | done: | |
1381 | /* Skip kmemleak for kasan_init() due to high volume. */ | |
1382 | if (end != MEMBLOCK_ALLOC_KASAN) | |
aedf95ea | 1383 | /* |
92d12f95 MR |
1384 | * The min_count is set to 0 so that memblock allocated |
1385 | * blocks are never reported as leaks. This is because many | |
1386 | * of these blocks are only referred via the physical | |
1387 | * address which is not looked up by kmemleak. | |
aedf95ea | 1388 | */ |
9099daed | 1389 | kmemleak_alloc_phys(found, size, 0, 0); |
92d12f95 MR |
1390 | |
1391 | return found; | |
95f72d1e YL |
1392 | } |
1393 | ||
a2974133 MR |
1394 | /** |
1395 | * memblock_phys_alloc_range - allocate a memory block inside specified range | |
1396 | * @size: size of memory block to be allocated in bytes | |
1397 | * @align: alignment of the region and block's size | |
1398 | * @start: the lower bound of the memory region to allocate (physical address) | |
1399 | * @end: the upper bound of the memory region to allocate (physical address) | |
1400 | * | |
1401 | * Allocate @size bytes in the between @start and @end. | |
1402 | * | |
1403 | * Return: physical address of the allocated memory block on success, | |
1404 | * %0 on failure. | |
1405 | */ | |
8a770c2a MR |
1406 | phys_addr_t __init memblock_phys_alloc_range(phys_addr_t size, |
1407 | phys_addr_t align, | |
1408 | phys_addr_t start, | |
1409 | phys_addr_t end) | |
2bfc2862 | 1410 | { |
0ac398b1 YY |
1411 | return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE, |
1412 | false); | |
7bd0b0f0 TH |
1413 | } |
1414 | ||
a2974133 MR |
1415 | /** |
1416 | * memblock_phys_alloc_try_nid - allocate a memory block from specified MUMA node | |
1417 | * @size: size of memory block to be allocated in bytes | |
1418 | * @align: alignment of the region and block's size | |
1419 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1420 | * | |
1421 | * Allocates memory block from the specified NUMA node. If the node | |
1422 | * has no available memory, attempts to allocated from any node in the | |
1423 | * system. | |
1424 | * | |
1425 | * Return: physical address of the allocated memory block on success, | |
1426 | * %0 on failure. | |
1427 | */ | |
9a8dd708 | 1428 | phys_addr_t __init memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) |
9d1e2492 | 1429 | { |
33755574 | 1430 | return memblock_alloc_range_nid(size, align, 0, |
0ac398b1 | 1431 | MEMBLOCK_ALLOC_ACCESSIBLE, nid, false); |
95f72d1e YL |
1432 | } |
1433 | ||
26f09e9b | 1434 | /** |
eb31d559 | 1435 | * memblock_alloc_internal - allocate boot memory block |
26f09e9b SS |
1436 | * @size: size of memory block to be allocated in bytes |
1437 | * @align: alignment of the region and block's size | |
1438 | * @min_addr: the lower bound of the memory region to allocate (phys address) | |
1439 | * @max_addr: the upper bound of the memory region to allocate (phys address) | |
1440 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
0ac398b1 | 1441 | * @exact_nid: control the allocation fall back to other nodes |
26f09e9b | 1442 | * |
92d12f95 MR |
1443 | * Allocates memory block using memblock_alloc_range_nid() and |
1444 | * converts the returned physical address to virtual. | |
26f09e9b | 1445 | * |
92d12f95 MR |
1446 | * The @min_addr limit is dropped if it can not be satisfied and the allocation |
1447 | * will fall back to memory below @min_addr. Other constraints, such | |
1448 | * as node and mirrored memory will be handled again in | |
1449 | * memblock_alloc_range_nid(). | |
26f09e9b | 1450 | * |
47cec443 | 1451 | * Return: |
26f09e9b SS |
1452 | * Virtual address of allocated memory block on success, NULL on failure. |
1453 | */ | |
eb31d559 | 1454 | static void * __init memblock_alloc_internal( |
26f09e9b SS |
1455 | phys_addr_t size, phys_addr_t align, |
1456 | phys_addr_t min_addr, phys_addr_t max_addr, | |
0ac398b1 | 1457 | int nid, bool exact_nid) |
26f09e9b SS |
1458 | { |
1459 | phys_addr_t alloc; | |
26f09e9b SS |
1460 | |
1461 | /* | |
1462 | * Detect any accidental use of these APIs after slab is ready, as at | |
1463 | * this moment memblock may be deinitialized already and its | |
c6ffc5ca | 1464 | * internal data may be destroyed (after execution of memblock_free_all) |
26f09e9b SS |
1465 | */ |
1466 | if (WARN_ON_ONCE(slab_is_available())) | |
1467 | return kzalloc_node(size, GFP_NOWAIT, nid); | |
1468 | ||
f3057ad7 MR |
1469 | if (max_addr > memblock.current_limit) |
1470 | max_addr = memblock.current_limit; | |
1471 | ||
0ac398b1 YY |
1472 | alloc = memblock_alloc_range_nid(size, align, min_addr, max_addr, nid, |
1473 | exact_nid); | |
26f09e9b | 1474 | |
92d12f95 MR |
1475 | /* retry allocation without lower limit */ |
1476 | if (!alloc && min_addr) | |
0ac398b1 YY |
1477 | alloc = memblock_alloc_range_nid(size, align, 0, max_addr, nid, |
1478 | exact_nid); | |
26f09e9b | 1479 | |
92d12f95 MR |
1480 | if (!alloc) |
1481 | return NULL; | |
26f09e9b | 1482 | |
92d12f95 | 1483 | return phys_to_virt(alloc); |
26f09e9b SS |
1484 | } |
1485 | ||
0ac398b1 YY |
1486 | /** |
1487 | * memblock_alloc_exact_nid_raw - allocate boot memory block on the exact node | |
1488 | * without zeroing memory | |
1489 | * @size: size of memory block to be allocated in bytes | |
1490 | * @align: alignment of the region and block's size | |
1491 | * @min_addr: the lower bound of the memory region from where the allocation | |
1492 | * is preferred (phys address) | |
1493 | * @max_addr: the upper bound of the memory region from where the allocation | |
1494 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to | |
1495 | * allocate only from memory limited by memblock.current_limit value | |
1496 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1497 | * | |
1498 | * Public function, provides additional debug information (including caller | |
1499 | * info), if enabled. Does not zero allocated memory. | |
1500 | * | |
1501 | * Return: | |
1502 | * Virtual address of allocated memory block on success, NULL on failure. | |
1503 | */ | |
1504 | void * __init memblock_alloc_exact_nid_raw( | |
1505 | phys_addr_t size, phys_addr_t align, | |
1506 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1507 | int nid) | |
1508 | { | |
1509 | void *ptr; | |
1510 | ||
1511 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", | |
1512 | __func__, (u64)size, (u64)align, nid, &min_addr, | |
1513 | &max_addr, (void *)_RET_IP_); | |
1514 | ||
1515 | ptr = memblock_alloc_internal(size, align, | |
1516 | min_addr, max_addr, nid, true); | |
1517 | if (ptr && size > 0) | |
1518 | page_init_poison(ptr, size); | |
1519 | ||
1520 | return ptr; | |
1521 | } | |
1522 | ||
ea1f5f37 | 1523 | /** |
eb31d559 | 1524 | * memblock_alloc_try_nid_raw - allocate boot memory block without zeroing |
ea1f5f37 PT |
1525 | * memory and without panicking |
1526 | * @size: size of memory block to be allocated in bytes | |
1527 | * @align: alignment of the region and block's size | |
1528 | * @min_addr: the lower bound of the memory region from where the allocation | |
1529 | * is preferred (phys address) | |
1530 | * @max_addr: the upper bound of the memory region from where the allocation | |
97ad1087 | 1531 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to |
ea1f5f37 PT |
1532 | * allocate only from memory limited by memblock.current_limit value |
1533 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1534 | * | |
1535 | * Public function, provides additional debug information (including caller | |
1536 | * info), if enabled. Does not zero allocated memory, does not panic if request | |
1537 | * cannot be satisfied. | |
1538 | * | |
47cec443 | 1539 | * Return: |
ea1f5f37 PT |
1540 | * Virtual address of allocated memory block on success, NULL on failure. |
1541 | */ | |
eb31d559 | 1542 | void * __init memblock_alloc_try_nid_raw( |
ea1f5f37 PT |
1543 | phys_addr_t size, phys_addr_t align, |
1544 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1545 | int nid) | |
1546 | { | |
1547 | void *ptr; | |
1548 | ||
d75f773c | 1549 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
a36aab89 MR |
1550 | __func__, (u64)size, (u64)align, nid, &min_addr, |
1551 | &max_addr, (void *)_RET_IP_); | |
ea1f5f37 | 1552 | |
eb31d559 | 1553 | ptr = memblock_alloc_internal(size, align, |
0ac398b1 | 1554 | min_addr, max_addr, nid, false); |
ea1f5f37 | 1555 | if (ptr && size > 0) |
f682a97a AD |
1556 | page_init_poison(ptr, size); |
1557 | ||
ea1f5f37 PT |
1558 | return ptr; |
1559 | } | |
1560 | ||
26f09e9b | 1561 | /** |
c0dbe825 | 1562 | * memblock_alloc_try_nid - allocate boot memory block |
26f09e9b SS |
1563 | * @size: size of memory block to be allocated in bytes |
1564 | * @align: alignment of the region and block's size | |
1565 | * @min_addr: the lower bound of the memory region from where the allocation | |
1566 | * is preferred (phys address) | |
1567 | * @max_addr: the upper bound of the memory region from where the allocation | |
97ad1087 | 1568 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to |
26f09e9b SS |
1569 | * allocate only from memory limited by memblock.current_limit value |
1570 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1571 | * | |
c0dbe825 MR |
1572 | * Public function, provides additional debug information (including caller |
1573 | * info), if enabled. This function zeroes the allocated memory. | |
26f09e9b | 1574 | * |
47cec443 | 1575 | * Return: |
26f09e9b SS |
1576 | * Virtual address of allocated memory block on success, NULL on failure. |
1577 | */ | |
eb31d559 | 1578 | void * __init memblock_alloc_try_nid( |
26f09e9b SS |
1579 | phys_addr_t size, phys_addr_t align, |
1580 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1581 | int nid) | |
1582 | { | |
1583 | void *ptr; | |
1584 | ||
d75f773c | 1585 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
a36aab89 MR |
1586 | __func__, (u64)size, (u64)align, nid, &min_addr, |
1587 | &max_addr, (void *)_RET_IP_); | |
eb31d559 | 1588 | ptr = memblock_alloc_internal(size, align, |
0ac398b1 | 1589 | min_addr, max_addr, nid, false); |
c0dbe825 | 1590 | if (ptr) |
ea1f5f37 | 1591 | memset(ptr, 0, size); |
26f09e9b | 1592 | |
c0dbe825 | 1593 | return ptr; |
26f09e9b SS |
1594 | } |
1595 | ||
48a833cc | 1596 | /** |
a2974133 | 1597 | * __memblock_free_late - free pages directly to buddy allocator |
48a833cc | 1598 | * @base: phys starting address of the boot memory block |
26f09e9b SS |
1599 | * @size: size of the boot memory block in bytes |
1600 | * | |
a2974133 | 1601 | * This is only useful when the memblock allocator has already been torn |
26f09e9b | 1602 | * down, but we are still initializing the system. Pages are released directly |
a2974133 | 1603 | * to the buddy allocator. |
26f09e9b SS |
1604 | */ |
1605 | void __init __memblock_free_late(phys_addr_t base, phys_addr_t size) | |
1606 | { | |
a36aab89 | 1607 | phys_addr_t cursor, end; |
26f09e9b | 1608 | |
a36aab89 | 1609 | end = base + size - 1; |
d75f773c | 1610 | memblock_dbg("%s: [%pa-%pa] %pS\n", |
a36aab89 | 1611 | __func__, &base, &end, (void *)_RET_IP_); |
9099daed | 1612 | kmemleak_free_part_phys(base, size); |
26f09e9b SS |
1613 | cursor = PFN_UP(base); |
1614 | end = PFN_DOWN(base + size); | |
1615 | ||
1616 | for (; cursor < end; cursor++) { | |
7c2ee349 | 1617 | memblock_free_pages(pfn_to_page(cursor), cursor, 0); |
ca79b0c2 | 1618 | totalram_pages_inc(); |
26f09e9b SS |
1619 | } |
1620 | } | |
9d1e2492 BH |
1621 | |
1622 | /* | |
1623 | * Remaining API functions | |
1624 | */ | |
1625 | ||
1f1ffb8a | 1626 | phys_addr_t __init_memblock memblock_phys_mem_size(void) |
95f72d1e | 1627 | { |
1440c4e2 | 1628 | return memblock.memory.total_size; |
95f72d1e YL |
1629 | } |
1630 | ||
8907de5d SD |
1631 | phys_addr_t __init_memblock memblock_reserved_size(void) |
1632 | { | |
1633 | return memblock.reserved.total_size; | |
1634 | } | |
1635 | ||
595ad9af YL |
1636 | phys_addr_t __init memblock_mem_size(unsigned long limit_pfn) |
1637 | { | |
1638 | unsigned long pages = 0; | |
1639 | struct memblock_region *r; | |
1640 | unsigned long start_pfn, end_pfn; | |
1641 | ||
1642 | for_each_memblock(memory, r) { | |
1643 | start_pfn = memblock_region_memory_base_pfn(r); | |
1644 | end_pfn = memblock_region_memory_end_pfn(r); | |
1645 | start_pfn = min_t(unsigned long, start_pfn, limit_pfn); | |
1646 | end_pfn = min_t(unsigned long, end_pfn, limit_pfn); | |
1647 | pages += end_pfn - start_pfn; | |
1648 | } | |
1649 | ||
16763230 | 1650 | return PFN_PHYS(pages); |
595ad9af YL |
1651 | } |
1652 | ||
0a93ebef SR |
1653 | /* lowest address */ |
1654 | phys_addr_t __init_memblock memblock_start_of_DRAM(void) | |
1655 | { | |
1656 | return memblock.memory.regions[0].base; | |
1657 | } | |
1658 | ||
10d06439 | 1659 | phys_addr_t __init_memblock memblock_end_of_DRAM(void) |
95f72d1e YL |
1660 | { |
1661 | int idx = memblock.memory.cnt - 1; | |
1662 | ||
e3239ff9 | 1663 | return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); |
95f72d1e YL |
1664 | } |
1665 | ||
a571d4eb | 1666 | static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) |
95f72d1e | 1667 | { |
1c4bc43d | 1668 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
136199f0 | 1669 | struct memblock_region *r; |
95f72d1e | 1670 | |
a571d4eb DC |
1671 | /* |
1672 | * translate the memory @limit size into the max address within one of | |
1673 | * the memory memblock regions, if the @limit exceeds the total size | |
1c4bc43d | 1674 | * of those regions, max_addr will keep original value PHYS_ADDR_MAX |
a571d4eb | 1675 | */ |
136199f0 | 1676 | for_each_memblock(memory, r) { |
c0ce8fef TH |
1677 | if (limit <= r->size) { |
1678 | max_addr = r->base + limit; | |
1679 | break; | |
95f72d1e | 1680 | } |
c0ce8fef | 1681 | limit -= r->size; |
95f72d1e | 1682 | } |
c0ce8fef | 1683 | |
a571d4eb DC |
1684 | return max_addr; |
1685 | } | |
1686 | ||
1687 | void __init memblock_enforce_memory_limit(phys_addr_t limit) | |
1688 | { | |
1c4bc43d | 1689 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
a571d4eb DC |
1690 | |
1691 | if (!limit) | |
1692 | return; | |
1693 | ||
1694 | max_addr = __find_max_addr(limit); | |
1695 | ||
1696 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1697 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1698 | return; |
1699 | ||
c0ce8fef | 1700 | /* truncate both memory and reserved regions */ |
f1af9d3a | 1701 | memblock_remove_range(&memblock.memory, max_addr, |
1c4bc43d | 1702 | PHYS_ADDR_MAX); |
f1af9d3a | 1703 | memblock_remove_range(&memblock.reserved, max_addr, |
1c4bc43d | 1704 | PHYS_ADDR_MAX); |
95f72d1e YL |
1705 | } |
1706 | ||
c9ca9b4e AT |
1707 | void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) |
1708 | { | |
1709 | int start_rgn, end_rgn; | |
1710 | int i, ret; | |
1711 | ||
1712 | if (!size) | |
1713 | return; | |
1714 | ||
1715 | ret = memblock_isolate_range(&memblock.memory, base, size, | |
1716 | &start_rgn, &end_rgn); | |
1717 | if (ret) | |
1718 | return; | |
1719 | ||
1720 | /* remove all the MAP regions */ | |
1721 | for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) | |
1722 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1723 | memblock_remove_region(&memblock.memory, i); | |
1724 | ||
1725 | for (i = start_rgn - 1; i >= 0; i--) | |
1726 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1727 | memblock_remove_region(&memblock.memory, i); | |
1728 | ||
1729 | /* truncate the reserved regions */ | |
1730 | memblock_remove_range(&memblock.reserved, 0, base); | |
1731 | memblock_remove_range(&memblock.reserved, | |
1c4bc43d | 1732 | base + size, PHYS_ADDR_MAX); |
c9ca9b4e AT |
1733 | } |
1734 | ||
a571d4eb DC |
1735 | void __init memblock_mem_limit_remove_map(phys_addr_t limit) |
1736 | { | |
a571d4eb | 1737 | phys_addr_t max_addr; |
a571d4eb DC |
1738 | |
1739 | if (!limit) | |
1740 | return; | |
1741 | ||
1742 | max_addr = __find_max_addr(limit); | |
1743 | ||
1744 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1745 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1746 | return; |
1747 | ||
c9ca9b4e | 1748 | memblock_cap_memory_range(0, max_addr); |
a571d4eb DC |
1749 | } |
1750 | ||
cd79481d | 1751 | static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) |
72d4b0b4 BH |
1752 | { |
1753 | unsigned int left = 0, right = type->cnt; | |
1754 | ||
1755 | do { | |
1756 | unsigned int mid = (right + left) / 2; | |
1757 | ||
1758 | if (addr < type->regions[mid].base) | |
1759 | right = mid; | |
1760 | else if (addr >= (type->regions[mid].base + | |
1761 | type->regions[mid].size)) | |
1762 | left = mid + 1; | |
1763 | else | |
1764 | return mid; | |
1765 | } while (left < right); | |
1766 | return -1; | |
1767 | } | |
1768 | ||
f5a222dc | 1769 | bool __init_memblock memblock_is_reserved(phys_addr_t addr) |
95f72d1e | 1770 | { |
72d4b0b4 BH |
1771 | return memblock_search(&memblock.reserved, addr) != -1; |
1772 | } | |
95f72d1e | 1773 | |
b4ad0c7e | 1774 | bool __init_memblock memblock_is_memory(phys_addr_t addr) |
72d4b0b4 BH |
1775 | { |
1776 | return memblock_search(&memblock.memory, addr) != -1; | |
1777 | } | |
1778 | ||
937f0c26 | 1779 | bool __init_memblock memblock_is_map_memory(phys_addr_t addr) |
bf3d3cc5 AB |
1780 | { |
1781 | int i = memblock_search(&memblock.memory, addr); | |
1782 | ||
1783 | if (i == -1) | |
1784 | return false; | |
1785 | return !memblock_is_nomap(&memblock.memory.regions[i]); | |
1786 | } | |
1787 | ||
e76b63f8 YL |
1788 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1789 | int __init_memblock memblock_search_pfn_nid(unsigned long pfn, | |
1790 | unsigned long *start_pfn, unsigned long *end_pfn) | |
1791 | { | |
1792 | struct memblock_type *type = &memblock.memory; | |
16763230 | 1793 | int mid = memblock_search(type, PFN_PHYS(pfn)); |
e76b63f8 YL |
1794 | |
1795 | if (mid == -1) | |
1796 | return -1; | |
1797 | ||
f7e2f7e8 FF |
1798 | *start_pfn = PFN_DOWN(type->regions[mid].base); |
1799 | *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); | |
e76b63f8 YL |
1800 | |
1801 | return type->regions[mid].nid; | |
1802 | } | |
1803 | #endif | |
1804 | ||
eab30949 SB |
1805 | /** |
1806 | * memblock_is_region_memory - check if a region is a subset of memory | |
1807 | * @base: base of region to check | |
1808 | * @size: size of region to check | |
1809 | * | |
47cec443 | 1810 | * Check if the region [@base, @base + @size) is a subset of a memory block. |
eab30949 | 1811 | * |
47cec443 | 1812 | * Return: |
eab30949 SB |
1813 | * 0 if false, non-zero if true |
1814 | */ | |
937f0c26 | 1815 | bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) |
72d4b0b4 | 1816 | { |
abb65272 | 1817 | int idx = memblock_search(&memblock.memory, base); |
eb18f1b5 | 1818 | phys_addr_t end = base + memblock_cap_size(base, &size); |
72d4b0b4 BH |
1819 | |
1820 | if (idx == -1) | |
937f0c26 | 1821 | return false; |
ef415ef4 | 1822 | return (memblock.memory.regions[idx].base + |
eb18f1b5 | 1823 | memblock.memory.regions[idx].size) >= end; |
95f72d1e YL |
1824 | } |
1825 | ||
eab30949 SB |
1826 | /** |
1827 | * memblock_is_region_reserved - check if a region intersects reserved memory | |
1828 | * @base: base of region to check | |
1829 | * @size: size of region to check | |
1830 | * | |
47cec443 MR |
1831 | * Check if the region [@base, @base + @size) intersects a reserved |
1832 | * memory block. | |
eab30949 | 1833 | * |
47cec443 | 1834 | * Return: |
c5c5c9d1 | 1835 | * True if they intersect, false if not. |
eab30949 | 1836 | */ |
c5c5c9d1 | 1837 | bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) |
95f72d1e | 1838 | { |
eb18f1b5 | 1839 | memblock_cap_size(base, &size); |
c5c5c9d1 | 1840 | return memblock_overlaps_region(&memblock.reserved, base, size); |
95f72d1e YL |
1841 | } |
1842 | ||
6ede1fd3 YL |
1843 | void __init_memblock memblock_trim_memory(phys_addr_t align) |
1844 | { | |
6ede1fd3 | 1845 | phys_addr_t start, end, orig_start, orig_end; |
136199f0 | 1846 | struct memblock_region *r; |
6ede1fd3 | 1847 | |
136199f0 EM |
1848 | for_each_memblock(memory, r) { |
1849 | orig_start = r->base; | |
1850 | orig_end = r->base + r->size; | |
6ede1fd3 YL |
1851 | start = round_up(orig_start, align); |
1852 | end = round_down(orig_end, align); | |
1853 | ||
1854 | if (start == orig_start && end == orig_end) | |
1855 | continue; | |
1856 | ||
1857 | if (start < end) { | |
136199f0 EM |
1858 | r->base = start; |
1859 | r->size = end - start; | |
6ede1fd3 | 1860 | } else { |
136199f0 EM |
1861 | memblock_remove_region(&memblock.memory, |
1862 | r - memblock.memory.regions); | |
1863 | r--; | |
6ede1fd3 YL |
1864 | } |
1865 | } | |
1866 | } | |
e63075a3 | 1867 | |
3661ca66 | 1868 | void __init_memblock memblock_set_current_limit(phys_addr_t limit) |
e63075a3 BH |
1869 | { |
1870 | memblock.current_limit = limit; | |
1871 | } | |
1872 | ||
fec51014 LA |
1873 | phys_addr_t __init_memblock memblock_get_current_limit(void) |
1874 | { | |
1875 | return memblock.current_limit; | |
1876 | } | |
1877 | ||
0262d9c8 | 1878 | static void __init_memblock memblock_dump(struct memblock_type *type) |
6ed311b2 | 1879 | { |
5d63f81c | 1880 | phys_addr_t base, end, size; |
e1720fee | 1881 | enum memblock_flags flags; |
8c9c1701 AK |
1882 | int idx; |
1883 | struct memblock_region *rgn; | |
6ed311b2 | 1884 | |
0262d9c8 | 1885 | pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); |
6ed311b2 | 1886 | |
66e8b438 | 1887 | for_each_memblock_type(idx, type, rgn) { |
7c0caeb8 TH |
1888 | char nid_buf[32] = ""; |
1889 | ||
1890 | base = rgn->base; | |
1891 | size = rgn->size; | |
5d63f81c | 1892 | end = base + size - 1; |
66a20757 | 1893 | flags = rgn->flags; |
7c0caeb8 TH |
1894 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1895 | if (memblock_get_region_node(rgn) != MAX_NUMNODES) | |
1896 | snprintf(nid_buf, sizeof(nid_buf), " on node %d", | |
1897 | memblock_get_region_node(rgn)); | |
1898 | #endif | |
e1720fee | 1899 | pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#x\n", |
0262d9c8 | 1900 | type->name, idx, &base, &end, &size, nid_buf, flags); |
6ed311b2 BH |
1901 | } |
1902 | } | |
1903 | ||
4ff7b82f | 1904 | void __init_memblock __memblock_dump_all(void) |
6ed311b2 | 1905 | { |
6ed311b2 | 1906 | pr_info("MEMBLOCK configuration:\n"); |
5d63f81c MC |
1907 | pr_info(" memory size = %pa reserved size = %pa\n", |
1908 | &memblock.memory.total_size, | |
1909 | &memblock.reserved.total_size); | |
6ed311b2 | 1910 | |
0262d9c8 HC |
1911 | memblock_dump(&memblock.memory); |
1912 | memblock_dump(&memblock.reserved); | |
409efd4c | 1913 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
0262d9c8 | 1914 | memblock_dump(&memblock.physmem); |
409efd4c | 1915 | #endif |
6ed311b2 BH |
1916 | } |
1917 | ||
1aadc056 | 1918 | void __init memblock_allow_resize(void) |
6ed311b2 | 1919 | { |
142b45a7 | 1920 | memblock_can_resize = 1; |
6ed311b2 BH |
1921 | } |
1922 | ||
6ed311b2 BH |
1923 | static int __init early_memblock(char *p) |
1924 | { | |
1925 | if (p && strstr(p, "debug")) | |
1926 | memblock_debug = 1; | |
1927 | return 0; | |
1928 | } | |
1929 | early_param("memblock", early_memblock); | |
1930 | ||
bda49a81 MR |
1931 | static void __init __free_pages_memory(unsigned long start, unsigned long end) |
1932 | { | |
1933 | int order; | |
1934 | ||
1935 | while (start < end) { | |
1936 | order = min(MAX_ORDER - 1UL, __ffs(start)); | |
1937 | ||
1938 | while (start + (1UL << order) > end) | |
1939 | order--; | |
1940 | ||
1941 | memblock_free_pages(pfn_to_page(start), start, order); | |
1942 | ||
1943 | start += (1UL << order); | |
1944 | } | |
1945 | } | |
1946 | ||
1947 | static unsigned long __init __free_memory_core(phys_addr_t start, | |
1948 | phys_addr_t end) | |
1949 | { | |
1950 | unsigned long start_pfn = PFN_UP(start); | |
1951 | unsigned long end_pfn = min_t(unsigned long, | |
1952 | PFN_DOWN(end), max_low_pfn); | |
1953 | ||
1954 | if (start_pfn >= end_pfn) | |
1955 | return 0; | |
1956 | ||
1957 | __free_pages_memory(start_pfn, end_pfn); | |
1958 | ||
1959 | return end_pfn - start_pfn; | |
1960 | } | |
1961 | ||
1962 | static unsigned long __init free_low_memory_core_early(void) | |
1963 | { | |
1964 | unsigned long count = 0; | |
1965 | phys_addr_t start, end; | |
1966 | u64 i; | |
1967 | ||
1968 | memblock_clear_hotplug(0, -1); | |
1969 | ||
1970 | for_each_reserved_mem_region(i, &start, &end) | |
1971 | reserve_bootmem_region(start, end); | |
1972 | ||
1973 | /* | |
1974 | * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id | |
1975 | * because in some case like Node0 doesn't have RAM installed | |
1976 | * low ram will be on Node1 | |
1977 | */ | |
1978 | for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, | |
1979 | NULL) | |
1980 | count += __free_memory_core(start, end); | |
1981 | ||
1982 | return count; | |
1983 | } | |
1984 | ||
1985 | static int reset_managed_pages_done __initdata; | |
1986 | ||
1987 | void reset_node_managed_pages(pg_data_t *pgdat) | |
1988 | { | |
1989 | struct zone *z; | |
1990 | ||
1991 | for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) | |
9705bea5 | 1992 | atomic_long_set(&z->managed_pages, 0); |
bda49a81 MR |
1993 | } |
1994 | ||
1995 | void __init reset_all_zones_managed_pages(void) | |
1996 | { | |
1997 | struct pglist_data *pgdat; | |
1998 | ||
1999 | if (reset_managed_pages_done) | |
2000 | return; | |
2001 | ||
2002 | for_each_online_pgdat(pgdat) | |
2003 | reset_node_managed_pages(pgdat); | |
2004 | ||
2005 | reset_managed_pages_done = 1; | |
2006 | } | |
2007 | ||
2008 | /** | |
2009 | * memblock_free_all - release free pages to the buddy allocator | |
2010 | * | |
2011 | * Return: the number of pages actually released. | |
2012 | */ | |
2013 | unsigned long __init memblock_free_all(void) | |
2014 | { | |
2015 | unsigned long pages; | |
2016 | ||
2017 | reset_all_zones_managed_pages(); | |
2018 | ||
2019 | pages = free_low_memory_core_early(); | |
ca79b0c2 | 2020 | totalram_pages_add(pages); |
bda49a81 MR |
2021 | |
2022 | return pages; | |
2023 | } | |
2024 | ||
350e88ba | 2025 | #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_ARCH_KEEP_MEMBLOCK) |
6d03b885 BH |
2026 | |
2027 | static int memblock_debug_show(struct seq_file *m, void *private) | |
2028 | { | |
2029 | struct memblock_type *type = m->private; | |
2030 | struct memblock_region *reg; | |
2031 | int i; | |
5d63f81c | 2032 | phys_addr_t end; |
6d03b885 BH |
2033 | |
2034 | for (i = 0; i < type->cnt; i++) { | |
2035 | reg = &type->regions[i]; | |
5d63f81c | 2036 | end = reg->base + reg->size - 1; |
6d03b885 | 2037 | |
5d63f81c MC |
2038 | seq_printf(m, "%4d: ", i); |
2039 | seq_printf(m, "%pa..%pa\n", ®->base, &end); | |
6d03b885 BH |
2040 | } |
2041 | return 0; | |
2042 | } | |
5ad35093 | 2043 | DEFINE_SHOW_ATTRIBUTE(memblock_debug); |
6d03b885 BH |
2044 | |
2045 | static int __init memblock_init_debugfs(void) | |
2046 | { | |
2047 | struct dentry *root = debugfs_create_dir("memblock", NULL); | |
d9f7979c | 2048 | |
0825a6f9 JP |
2049 | debugfs_create_file("memory", 0444, root, |
2050 | &memblock.memory, &memblock_debug_fops); | |
2051 | debugfs_create_file("reserved", 0444, root, | |
2052 | &memblock.reserved, &memblock_debug_fops); | |
70210ed9 | 2053 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
0825a6f9 JP |
2054 | debugfs_create_file("physmem", 0444, root, |
2055 | &memblock.physmem, &memblock_debug_fops); | |
70210ed9 | 2056 | #endif |
6d03b885 BH |
2057 | |
2058 | return 0; | |
2059 | } | |
2060 | __initcall(memblock_init_debugfs); | |
2061 | ||
2062 | #endif /* CONFIG_DEBUG_FS */ |