]>
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 | |
77649905 DH |
47 | * * ``physmem`` - describes the actual physical memory available during |
48 | * boot regardless of the possible restrictions and memory hot(un)plug; | |
49 | * the ``physmem`` type is only available on some architectures. | |
3e039c5c | 50 | * |
9303c9d5 | 51 | * Each region is represented by struct memblock_region that |
3e039c5c | 52 | * defines the region extents, its attributes and NUMA node id on NUMA |
1bf162e4 MCC |
53 | * systems. Every memory type is described by the struct memblock_type |
54 | * which contains an array of memory regions along with | |
77649905 | 55 | * the allocator metadata. The "memory" and "reserved" types are nicely |
9303c9d5 | 56 | * wrapped with struct memblock. This structure is statically |
77649905 DH |
57 | * initialized at build time. The region arrays are initially sized to |
58 | * %INIT_MEMBLOCK_REGIONS for "memory" and %INIT_MEMBLOCK_RESERVED_REGIONS | |
59 | * for "reserved". The region array for "physmem" is initially sized to | |
60 | * %INIT_PHYSMEM_REGIONS. | |
6e5af9a8 C |
61 | * The memblock_allow_resize() enables automatic resizing of the region |
62 | * arrays during addition of new regions. This feature should be used | |
63 | * with care so that memory allocated for the region array will not | |
64 | * overlap with areas that should be reserved, for example initrd. | |
3e039c5c MR |
65 | * |
66 | * The early architecture setup should tell memblock what the physical | |
6e5af9a8 C |
67 | * memory layout is by using memblock_add() or memblock_add_node() |
68 | * functions. The first function does not assign the region to a NUMA | |
69 | * node and it is appropriate for UMA systems. Yet, it is possible to | |
70 | * use it on NUMA systems as well and assign the region to a NUMA node | |
71 | * later in the setup process using memblock_set_node(). The | |
72 | * memblock_add_node() performs such an assignment directly. | |
3e039c5c | 73 | * |
a2974133 MR |
74 | * Once memblock is setup the memory can be allocated using one of the |
75 | * API variants: | |
76 | * | |
6e5af9a8 C |
77 | * * memblock_phys_alloc*() - these functions return the **physical** |
78 | * address of the allocated memory | |
79 | * * memblock_alloc*() - these functions return the **virtual** address | |
80 | * of the allocated memory. | |
a2974133 | 81 | * |
df1758d9 | 82 | * Note, that both API variants use implicit assumptions about allowed |
a2974133 | 83 | * memory ranges and the fallback methods. Consult the documentation |
6e5af9a8 C |
84 | * of memblock_alloc_internal() and memblock_alloc_range_nid() |
85 | * functions for more elaborate description. | |
3e039c5c | 86 | * |
6e5af9a8 C |
87 | * As the system boot progresses, the architecture specific mem_init() |
88 | * function frees all the memory to the buddy page allocator. | |
3e039c5c | 89 | * |
6e5af9a8 | 90 | * Unless an architecture enables %CONFIG_ARCH_KEEP_MEMBLOCK, the |
77649905 DH |
91 | * memblock data structures (except "physmem") will be discarded after the |
92 | * system initialization completes. | |
3e039c5c MR |
93 | */ |
94 | ||
a9ee6cf5 | 95 | #ifndef CONFIG_NUMA |
bda49a81 MR |
96 | struct pglist_data __refdata contig_page_data; |
97 | EXPORT_SYMBOL(contig_page_data); | |
98 | #endif | |
99 | ||
100 | unsigned long max_low_pfn; | |
101 | unsigned long min_low_pfn; | |
102 | unsigned long max_pfn; | |
103 | unsigned long long max_possible_pfn; | |
104 | ||
fe091c20 | 105 | static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock; |
8a5b403d | 106 | static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_RESERVED_REGIONS] __initdata_memblock; |
70210ed9 | 107 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
77649905 | 108 | static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS]; |
70210ed9 | 109 | #endif |
fe091c20 TH |
110 | |
111 | struct memblock memblock __initdata_memblock = { | |
112 | .memory.regions = memblock_memory_init_regions, | |
113 | .memory.cnt = 1, /* empty dummy entry */ | |
114 | .memory.max = INIT_MEMBLOCK_REGIONS, | |
0262d9c8 | 115 | .memory.name = "memory", |
fe091c20 TH |
116 | |
117 | .reserved.regions = memblock_reserved_init_regions, | |
118 | .reserved.cnt = 1, /* empty dummy entry */ | |
8a5b403d | 119 | .reserved.max = INIT_MEMBLOCK_RESERVED_REGIONS, |
0262d9c8 | 120 | .reserved.name = "reserved", |
fe091c20 | 121 | |
79442ed1 | 122 | .bottom_up = false, |
fe091c20 TH |
123 | .current_limit = MEMBLOCK_ALLOC_ANYWHERE, |
124 | }; | |
95f72d1e | 125 | |
77649905 DH |
126 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
127 | struct memblock_type physmem = { | |
128 | .regions = memblock_physmem_init_regions, | |
129 | .cnt = 1, /* empty dummy entry */ | |
130 | .max = INIT_PHYSMEM_REGIONS, | |
131 | .name = "physmem", | |
132 | }; | |
133 | #endif | |
134 | ||
9f3d5eaa MR |
135 | /* |
136 | * keep a pointer to &memblock.memory in the text section to use it in | |
137 | * __next_mem_range() and its helpers. | |
138 | * For architectures that do not keep memblock data after init, this | |
139 | * pointer will be reset to NULL at memblock_discard() | |
140 | */ | |
141 | static __refdata struct memblock_type *memblock_memory = &memblock.memory; | |
142 | ||
cd991db8 MR |
143 | #define for_each_memblock_type(i, memblock_type, rgn) \ |
144 | for (i = 0, rgn = &memblock_type->regions[0]; \ | |
145 | i < memblock_type->cnt; \ | |
146 | i++, rgn = &memblock_type->regions[i]) | |
147 | ||
87c55870 MR |
148 | #define memblock_dbg(fmt, ...) \ |
149 | do { \ | |
150 | if (memblock_debug) \ | |
151 | pr_info(fmt, ##__VA_ARGS__); \ | |
152 | } while (0) | |
153 | ||
154 | static int memblock_debug __initdata_memblock; | |
a3f5bafc | 155 | static bool system_has_some_mirror __initdata_memblock = false; |
1aadc056 | 156 | static int memblock_can_resize __initdata_memblock; |
181eb394 GS |
157 | static int memblock_memory_in_slab __initdata_memblock = 0; |
158 | static int memblock_reserved_in_slab __initdata_memblock = 0; | |
95f72d1e | 159 | |
c366ea89 | 160 | static enum memblock_flags __init_memblock choose_memblock_flags(void) |
a3f5bafc TL |
161 | { |
162 | return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE; | |
163 | } | |
164 | ||
eb18f1b5 TH |
165 | /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */ |
166 | static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size) | |
167 | { | |
1c4bc43d | 168 | return *size = min(*size, PHYS_ADDR_MAX - base); |
eb18f1b5 TH |
169 | } |
170 | ||
6ed311b2 BH |
171 | /* |
172 | * Address comparison utilities | |
173 | */ | |
10d06439 | 174 | static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1, |
2898cc4c | 175 | phys_addr_t base2, phys_addr_t size2) |
95f72d1e YL |
176 | { |
177 | return ((base1 < (base2 + size2)) && (base2 < (base1 + size1))); | |
178 | } | |
179 | ||
95cf82ec | 180 | bool __init_memblock memblock_overlaps_region(struct memblock_type *type, |
2d7d3eb2 | 181 | phys_addr_t base, phys_addr_t size) |
6ed311b2 BH |
182 | { |
183 | unsigned long i; | |
184 | ||
023accf5 MR |
185 | memblock_cap_size(base, &size); |
186 | ||
f14516fb AK |
187 | for (i = 0; i < type->cnt; i++) |
188 | if (memblock_addrs_overlap(base, size, type->regions[i].base, | |
189 | type->regions[i].size)) | |
6ed311b2 | 190 | break; |
c5c5c9d1 | 191 | return i < type->cnt; |
6ed311b2 BH |
192 | } |
193 | ||
47cec443 | 194 | /** |
79442ed1 TC |
195 | * __memblock_find_range_bottom_up - find free area utility in bottom-up |
196 | * @start: start of candidate range | |
47cec443 MR |
197 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
198 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
79442ed1 TC |
199 | * @size: size of free area to find |
200 | * @align: alignment of free area to find | |
b1154233 | 201 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 202 | * @flags: pick from blocks based on memory attributes |
79442ed1 TC |
203 | * |
204 | * Utility called from memblock_find_in_range_node(), find free area bottom-up. | |
205 | * | |
47cec443 | 206 | * Return: |
79442ed1 TC |
207 | * Found address on success, 0 on failure. |
208 | */ | |
209 | static phys_addr_t __init_memblock | |
210 | __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 211 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 212 | enum memblock_flags flags) |
79442ed1 TC |
213 | { |
214 | phys_addr_t this_start, this_end, cand; | |
215 | u64 i; | |
216 | ||
fc6daaf9 | 217 | for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { |
79442ed1 TC |
218 | this_start = clamp(this_start, start, end); |
219 | this_end = clamp(this_end, start, end); | |
220 | ||
221 | cand = round_up(this_start, align); | |
222 | if (cand < this_end && this_end - cand >= size) | |
223 | return cand; | |
224 | } | |
225 | ||
226 | return 0; | |
227 | } | |
228 | ||
7bd0b0f0 | 229 | /** |
1402899e | 230 | * __memblock_find_range_top_down - find free area utility, in top-down |
7bd0b0f0 | 231 | * @start: start of candidate range |
47cec443 MR |
232 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
233 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
234 | * @size: size of free area to find |
235 | * @align: alignment of free area to find | |
b1154233 | 236 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 237 | * @flags: pick from blocks based on memory attributes |
7bd0b0f0 | 238 | * |
1402899e | 239 | * Utility called from memblock_find_in_range_node(), find free area top-down. |
7bd0b0f0 | 240 | * |
47cec443 | 241 | * Return: |
79442ed1 | 242 | * Found address on success, 0 on failure. |
6ed311b2 | 243 | */ |
1402899e TC |
244 | static phys_addr_t __init_memblock |
245 | __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 246 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 247 | enum memblock_flags flags) |
f7210e6c TC |
248 | { |
249 | phys_addr_t this_start, this_end, cand; | |
250 | u64 i; | |
251 | ||
fc6daaf9 TL |
252 | for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, |
253 | NULL) { | |
f7210e6c TC |
254 | this_start = clamp(this_start, start, end); |
255 | this_end = clamp(this_end, start, end); | |
256 | ||
257 | if (this_end < size) | |
258 | continue; | |
259 | ||
260 | cand = round_down(this_end - size, align); | |
261 | if (cand >= this_start) | |
262 | return cand; | |
263 | } | |
1402899e | 264 | |
f7210e6c TC |
265 | return 0; |
266 | } | |
6ed311b2 | 267 | |
1402899e TC |
268 | /** |
269 | * memblock_find_in_range_node - find free area in given range and node | |
1402899e TC |
270 | * @size: size of free area to find |
271 | * @align: alignment of free area to find | |
87029ee9 | 272 | * @start: start of candidate range |
47cec443 MR |
273 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
274 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
b1154233 | 275 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 276 | * @flags: pick from blocks based on memory attributes |
1402899e TC |
277 | * |
278 | * Find @size free area aligned to @align in the specified range and node. | |
279 | * | |
47cec443 | 280 | * Return: |
79442ed1 | 281 | * Found address on success, 0 on failure. |
1402899e | 282 | */ |
c366ea89 | 283 | static phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, |
87029ee9 | 284 | phys_addr_t align, phys_addr_t start, |
e1720fee MR |
285 | phys_addr_t end, int nid, |
286 | enum memblock_flags flags) | |
1402899e TC |
287 | { |
288 | /* pump up @end */ | |
fed84c78 QC |
289 | if (end == MEMBLOCK_ALLOC_ACCESSIBLE || |
290 | end == MEMBLOCK_ALLOC_KASAN) | |
1402899e TC |
291 | end = memblock.current_limit; |
292 | ||
293 | /* avoid allocating the first page */ | |
294 | start = max_t(phys_addr_t, start, PAGE_SIZE); | |
295 | end = max(start, end); | |
296 | ||
2dcb3964 RG |
297 | if (memblock_bottom_up()) |
298 | return __memblock_find_range_bottom_up(start, end, size, align, | |
299 | nid, flags); | |
300 | else | |
301 | return __memblock_find_range_top_down(start, end, size, align, | |
302 | nid, flags); | |
1402899e TC |
303 | } |
304 | ||
7bd0b0f0 TH |
305 | /** |
306 | * memblock_find_in_range - find free area in given range | |
307 | * @start: start of candidate range | |
47cec443 MR |
308 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
309 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
310 | * @size: size of free area to find |
311 | * @align: alignment of free area to find | |
312 | * | |
313 | * Find @size free area aligned to @align in the specified range. | |
314 | * | |
47cec443 | 315 | * Return: |
79442ed1 | 316 | * Found address on success, 0 on failure. |
fc769a8e | 317 | */ |
a7259df7 | 318 | static phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, |
7bd0b0f0 TH |
319 | phys_addr_t end, phys_addr_t size, |
320 | phys_addr_t align) | |
6ed311b2 | 321 | { |
a3f5bafc | 322 | phys_addr_t ret; |
e1720fee | 323 | enum memblock_flags flags = choose_memblock_flags(); |
a3f5bafc TL |
324 | |
325 | again: | |
326 | ret = memblock_find_in_range_node(size, align, start, end, | |
327 | NUMA_NO_NODE, flags); | |
328 | ||
329 | if (!ret && (flags & MEMBLOCK_MIRROR)) { | |
330 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
331 | &size); | |
332 | flags &= ~MEMBLOCK_MIRROR; | |
333 | goto again; | |
334 | } | |
335 | ||
336 | return ret; | |
6ed311b2 BH |
337 | } |
338 | ||
10d06439 | 339 | static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) |
95f72d1e | 340 | { |
1440c4e2 | 341 | type->total_size -= type->regions[r].size; |
7c0caeb8 TH |
342 | memmove(&type->regions[r], &type->regions[r + 1], |
343 | (type->cnt - (r + 1)) * sizeof(type->regions[r])); | |
e3239ff9 | 344 | type->cnt--; |
95f72d1e | 345 | |
8f7a6605 BH |
346 | /* Special case for empty arrays */ |
347 | if (type->cnt == 0) { | |
1440c4e2 | 348 | WARN_ON(type->total_size != 0); |
8f7a6605 BH |
349 | type->cnt = 1; |
350 | type->regions[0].base = 0; | |
351 | type->regions[0].size = 0; | |
66a20757 | 352 | type->regions[0].flags = 0; |
7c0caeb8 | 353 | memblock_set_region_node(&type->regions[0], MAX_NUMNODES); |
8f7a6605 | 354 | } |
95f72d1e YL |
355 | } |
356 | ||
350e88ba | 357 | #ifndef CONFIG_ARCH_KEEP_MEMBLOCK |
3010f876 | 358 | /** |
47cec443 | 359 | * memblock_discard - discard memory and reserved arrays if they were allocated |
3010f876 PT |
360 | */ |
361 | void __init memblock_discard(void) | |
5e270e25 | 362 | { |
3010f876 | 363 | phys_addr_t addr, size; |
5e270e25 | 364 | |
3010f876 PT |
365 | if (memblock.reserved.regions != memblock_reserved_init_regions) { |
366 | addr = __pa(memblock.reserved.regions); | |
367 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
368 | memblock.reserved.max); | |
369 | __memblock_free_late(addr, size); | |
370 | } | |
5e270e25 | 371 | |
91b540f9 | 372 | if (memblock.memory.regions != memblock_memory_init_regions) { |
3010f876 PT |
373 | addr = __pa(memblock.memory.regions); |
374 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
375 | memblock.memory.max); | |
376 | __memblock_free_late(addr, size); | |
377 | } | |
9f3d5eaa MR |
378 | |
379 | memblock_memory = NULL; | |
5e270e25 | 380 | } |
5e270e25 PH |
381 | #endif |
382 | ||
48c3b583 GP |
383 | /** |
384 | * memblock_double_array - double the size of the memblock regions array | |
385 | * @type: memblock type of the regions array being doubled | |
386 | * @new_area_start: starting address of memory range to avoid overlap with | |
387 | * @new_area_size: size of memory range to avoid overlap with | |
388 | * | |
389 | * Double the size of the @type regions array. If memblock is being used to | |
390 | * allocate memory for a new reserved regions array and there is a previously | |
47cec443 | 391 | * allocated memory range [@new_area_start, @new_area_start + @new_area_size] |
48c3b583 GP |
392 | * waiting to be reserved, ensure the memory used by the new array does |
393 | * not overlap. | |
394 | * | |
47cec443 | 395 | * Return: |
48c3b583 GP |
396 | * 0 on success, -1 on failure. |
397 | */ | |
398 | static int __init_memblock memblock_double_array(struct memblock_type *type, | |
399 | phys_addr_t new_area_start, | |
400 | phys_addr_t new_area_size) | |
142b45a7 BH |
401 | { |
402 | struct memblock_region *new_array, *old_array; | |
29f67386 | 403 | phys_addr_t old_alloc_size, new_alloc_size; |
a36aab89 | 404 | phys_addr_t old_size, new_size, addr, new_end; |
142b45a7 | 405 | int use_slab = slab_is_available(); |
181eb394 | 406 | int *in_slab; |
142b45a7 BH |
407 | |
408 | /* We don't allow resizing until we know about the reserved regions | |
409 | * of memory that aren't suitable for allocation | |
410 | */ | |
411 | if (!memblock_can_resize) | |
412 | return -1; | |
413 | ||
142b45a7 BH |
414 | /* Calculate new doubled size */ |
415 | old_size = type->max * sizeof(struct memblock_region); | |
416 | new_size = old_size << 1; | |
29f67386 YL |
417 | /* |
418 | * We need to allocated new one align to PAGE_SIZE, | |
419 | * so we can free them completely later. | |
420 | */ | |
421 | old_alloc_size = PAGE_ALIGN(old_size); | |
422 | new_alloc_size = PAGE_ALIGN(new_size); | |
142b45a7 | 423 | |
181eb394 GS |
424 | /* Retrieve the slab flag */ |
425 | if (type == &memblock.memory) | |
426 | in_slab = &memblock_memory_in_slab; | |
427 | else | |
428 | in_slab = &memblock_reserved_in_slab; | |
429 | ||
a2974133 | 430 | /* Try to find some space for it */ |
142b45a7 BH |
431 | if (use_slab) { |
432 | new_array = kmalloc(new_size, GFP_KERNEL); | |
1f5026a7 | 433 | addr = new_array ? __pa(new_array) : 0; |
4e2f0775 | 434 | } else { |
48c3b583 GP |
435 | /* only exclude range when trying to double reserved.regions */ |
436 | if (type != &memblock.reserved) | |
437 | new_area_start = new_area_size = 0; | |
438 | ||
439 | addr = memblock_find_in_range(new_area_start + new_area_size, | |
440 | memblock.current_limit, | |
29f67386 | 441 | new_alloc_size, PAGE_SIZE); |
48c3b583 GP |
442 | if (!addr && new_area_size) |
443 | addr = memblock_find_in_range(0, | |
fd07383b AM |
444 | min(new_area_start, memblock.current_limit), |
445 | new_alloc_size, PAGE_SIZE); | |
48c3b583 | 446 | |
15674868 | 447 | new_array = addr ? __va(addr) : NULL; |
4e2f0775 | 448 | } |
1f5026a7 | 449 | if (!addr) { |
142b45a7 | 450 | pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", |
0262d9c8 | 451 | type->name, type->max, type->max * 2); |
142b45a7 BH |
452 | return -1; |
453 | } | |
142b45a7 | 454 | |
a36aab89 MR |
455 | new_end = addr + new_size - 1; |
456 | memblock_dbg("memblock: %s is doubled to %ld at [%pa-%pa]", | |
457 | type->name, type->max * 2, &addr, &new_end); | |
ea9e4376 | 458 | |
fd07383b AM |
459 | /* |
460 | * Found space, we now need to move the array over before we add the | |
461 | * reserved region since it may be our reserved array itself that is | |
462 | * full. | |
142b45a7 BH |
463 | */ |
464 | memcpy(new_array, type->regions, old_size); | |
465 | memset(new_array + type->max, 0, old_size); | |
466 | old_array = type->regions; | |
467 | type->regions = new_array; | |
468 | type->max <<= 1; | |
469 | ||
fd07383b | 470 | /* Free old array. We needn't free it if the array is the static one */ |
181eb394 GS |
471 | if (*in_slab) |
472 | kfree(old_array); | |
473 | else if (old_array != memblock_memory_init_regions && | |
474 | old_array != memblock_reserved_init_regions) | |
77e02cf5 | 475 | memblock_free_ptr(old_array, old_alloc_size); |
142b45a7 | 476 | |
fd07383b AM |
477 | /* |
478 | * Reserve the new array if that comes from the memblock. Otherwise, we | |
479 | * needn't do it | |
181eb394 GS |
480 | */ |
481 | if (!use_slab) | |
29f67386 | 482 | BUG_ON(memblock_reserve(addr, new_alloc_size)); |
181eb394 GS |
483 | |
484 | /* Update slab flag */ | |
485 | *in_slab = use_slab; | |
486 | ||
142b45a7 BH |
487 | return 0; |
488 | } | |
489 | ||
784656f9 TH |
490 | /** |
491 | * memblock_merge_regions - merge neighboring compatible regions | |
492 | * @type: memblock type to scan | |
493 | * | |
494 | * Scan @type and merge neighboring compatible regions. | |
495 | */ | |
496 | static void __init_memblock memblock_merge_regions(struct memblock_type *type) | |
95f72d1e | 497 | { |
784656f9 | 498 | int i = 0; |
95f72d1e | 499 | |
784656f9 TH |
500 | /* cnt never goes below 1 */ |
501 | while (i < type->cnt - 1) { | |
502 | struct memblock_region *this = &type->regions[i]; | |
503 | struct memblock_region *next = &type->regions[i + 1]; | |
95f72d1e | 504 | |
7c0caeb8 TH |
505 | if (this->base + this->size != next->base || |
506 | memblock_get_region_node(this) != | |
66a20757 TC |
507 | memblock_get_region_node(next) || |
508 | this->flags != next->flags) { | |
784656f9 TH |
509 | BUG_ON(this->base + this->size > next->base); |
510 | i++; | |
511 | continue; | |
8f7a6605 BH |
512 | } |
513 | ||
784656f9 | 514 | this->size += next->size; |
c0232ae8 LF |
515 | /* move forward from next + 1, index of which is i + 2 */ |
516 | memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); | |
784656f9 | 517 | type->cnt--; |
95f72d1e | 518 | } |
784656f9 | 519 | } |
95f72d1e | 520 | |
784656f9 TH |
521 | /** |
522 | * memblock_insert_region - insert new memblock region | |
209ff86d TC |
523 | * @type: memblock type to insert into |
524 | * @idx: index for the insertion point | |
525 | * @base: base address of the new region | |
526 | * @size: size of the new region | |
527 | * @nid: node id of the new region | |
66a20757 | 528 | * @flags: flags of the new region |
784656f9 | 529 | * |
47cec443 | 530 | * Insert new memblock region [@base, @base + @size) into @type at @idx. |
412d0008 | 531 | * @type must already have extra room to accommodate the new region. |
784656f9 TH |
532 | */ |
533 | static void __init_memblock memblock_insert_region(struct memblock_type *type, | |
534 | int idx, phys_addr_t base, | |
66a20757 | 535 | phys_addr_t size, |
e1720fee MR |
536 | int nid, |
537 | enum memblock_flags flags) | |
784656f9 TH |
538 | { |
539 | struct memblock_region *rgn = &type->regions[idx]; | |
540 | ||
541 | BUG_ON(type->cnt >= type->max); | |
542 | memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); | |
543 | rgn->base = base; | |
544 | rgn->size = size; | |
66a20757 | 545 | rgn->flags = flags; |
7c0caeb8 | 546 | memblock_set_region_node(rgn, nid); |
784656f9 | 547 | type->cnt++; |
1440c4e2 | 548 | type->total_size += size; |
784656f9 TH |
549 | } |
550 | ||
551 | /** | |
f1af9d3a | 552 | * memblock_add_range - add new memblock region |
784656f9 TH |
553 | * @type: memblock type to add new region into |
554 | * @base: base address of the new region | |
555 | * @size: size of the new region | |
7fb0bc3f | 556 | * @nid: nid of the new region |
66a20757 | 557 | * @flags: flags of the new region |
784656f9 | 558 | * |
47cec443 | 559 | * Add new memblock region [@base, @base + @size) into @type. The new region |
784656f9 TH |
560 | * is allowed to overlap with existing ones - overlaps don't affect already |
561 | * existing regions. @type is guaranteed to be minimal (all neighbouring | |
562 | * compatible regions are merged) after the addition. | |
563 | * | |
47cec443 | 564 | * Return: |
784656f9 TH |
565 | * 0 on success, -errno on failure. |
566 | */ | |
02634a44 | 567 | static int __init_memblock memblock_add_range(struct memblock_type *type, |
66a20757 | 568 | phys_addr_t base, phys_addr_t size, |
e1720fee | 569 | int nid, enum memblock_flags flags) |
784656f9 TH |
570 | { |
571 | bool insert = false; | |
eb18f1b5 TH |
572 | phys_addr_t obase = base; |
573 | phys_addr_t end = base + memblock_cap_size(base, &size); | |
8c9c1701 AK |
574 | int idx, nr_new; |
575 | struct memblock_region *rgn; | |
784656f9 | 576 | |
b3dc627c TH |
577 | if (!size) |
578 | return 0; | |
579 | ||
784656f9 TH |
580 | /* special case for empty array */ |
581 | if (type->regions[0].size == 0) { | |
1440c4e2 | 582 | WARN_ON(type->cnt != 1 || type->total_size); |
8f7a6605 BH |
583 | type->regions[0].base = base; |
584 | type->regions[0].size = size; | |
66a20757 | 585 | type->regions[0].flags = flags; |
7fb0bc3f | 586 | memblock_set_region_node(&type->regions[0], nid); |
1440c4e2 | 587 | type->total_size = size; |
8f7a6605 | 588 | return 0; |
95f72d1e | 589 | } |
784656f9 TH |
590 | repeat: |
591 | /* | |
592 | * The following is executed twice. Once with %false @insert and | |
593 | * then with %true. The first counts the number of regions needed | |
412d0008 | 594 | * to accommodate the new area. The second actually inserts them. |
142b45a7 | 595 | */ |
784656f9 TH |
596 | base = obase; |
597 | nr_new = 0; | |
95f72d1e | 598 | |
66e8b438 | 599 | for_each_memblock_type(idx, type, rgn) { |
784656f9 TH |
600 | phys_addr_t rbase = rgn->base; |
601 | phys_addr_t rend = rbase + rgn->size; | |
602 | ||
603 | if (rbase >= end) | |
95f72d1e | 604 | break; |
784656f9 TH |
605 | if (rend <= base) |
606 | continue; | |
607 | /* | |
608 | * @rgn overlaps. If it separates the lower part of new | |
609 | * area, insert that portion. | |
610 | */ | |
611 | if (rbase > base) { | |
a9ee6cf5 | 612 | #ifdef CONFIG_NUMA |
c0a29498 WY |
613 | WARN_ON(nid != memblock_get_region_node(rgn)); |
614 | #endif | |
4fcab5f4 | 615 | WARN_ON(flags != rgn->flags); |
784656f9 TH |
616 | nr_new++; |
617 | if (insert) | |
8c9c1701 | 618 | memblock_insert_region(type, idx++, base, |
66a20757 TC |
619 | rbase - base, nid, |
620 | flags); | |
95f72d1e | 621 | } |
784656f9 TH |
622 | /* area below @rend is dealt with, forget about it */ |
623 | base = min(rend, end); | |
95f72d1e | 624 | } |
784656f9 TH |
625 | |
626 | /* insert the remaining portion */ | |
627 | if (base < end) { | |
628 | nr_new++; | |
629 | if (insert) | |
8c9c1701 | 630 | memblock_insert_region(type, idx, base, end - base, |
66a20757 | 631 | nid, flags); |
95f72d1e | 632 | } |
95f72d1e | 633 | |
ef3cc4db | 634 | if (!nr_new) |
635 | return 0; | |
636 | ||
784656f9 TH |
637 | /* |
638 | * If this was the first round, resize array and repeat for actual | |
639 | * insertions; otherwise, merge and return. | |
142b45a7 | 640 | */ |
784656f9 TH |
641 | if (!insert) { |
642 | while (type->cnt + nr_new > type->max) | |
48c3b583 | 643 | if (memblock_double_array(type, obase, size) < 0) |
784656f9 TH |
644 | return -ENOMEM; |
645 | insert = true; | |
646 | goto repeat; | |
647 | } else { | |
648 | memblock_merge_regions(type); | |
649 | return 0; | |
142b45a7 | 650 | } |
95f72d1e YL |
651 | } |
652 | ||
48a833cc MR |
653 | /** |
654 | * memblock_add_node - add new memblock region within a NUMA node | |
655 | * @base: base address of the new region | |
656 | * @size: size of the new region | |
657 | * @nid: nid of the new region | |
658 | * | |
659 | * Add new memblock region [@base, @base + @size) to the "memory" | |
660 | * type. See memblock_add_range() description for mode details | |
661 | * | |
662 | * Return: | |
663 | * 0 on success, -errno on failure. | |
664 | */ | |
7fb0bc3f TH |
665 | int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, |
666 | int nid) | |
667 | { | |
00974b9a GU |
668 | phys_addr_t end = base + size - 1; |
669 | ||
670 | memblock_dbg("%s: [%pa-%pa] nid=%d %pS\n", __func__, | |
671 | &base, &end, nid, (void *)_RET_IP_); | |
672 | ||
f1af9d3a | 673 | return memblock_add_range(&memblock.memory, base, size, nid, 0); |
7fb0bc3f TH |
674 | } |
675 | ||
48a833cc MR |
676 | /** |
677 | * memblock_add - add new memblock region | |
678 | * @base: base address of the new region | |
679 | * @size: size of the new region | |
680 | * | |
681 | * Add new memblock region [@base, @base + @size) to the "memory" | |
682 | * type. See memblock_add_range() description for mode details | |
683 | * | |
684 | * Return: | |
685 | * 0 on success, -errno on failure. | |
686 | */ | |
f705ac4b | 687 | int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) |
6a4055bc | 688 | { |
5d63f81c MC |
689 | phys_addr_t end = base + size - 1; |
690 | ||
a090d711 | 691 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, |
5d63f81c | 692 | &base, &end, (void *)_RET_IP_); |
6a4055bc | 693 | |
f705ac4b | 694 | return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
695 | } |
696 | ||
6a9ceb31 TH |
697 | /** |
698 | * memblock_isolate_range - isolate given range into disjoint memblocks | |
699 | * @type: memblock type to isolate range for | |
700 | * @base: base of range to isolate | |
701 | * @size: size of range to isolate | |
702 | * @start_rgn: out parameter for the start of isolated region | |
703 | * @end_rgn: out parameter for the end of isolated region | |
704 | * | |
705 | * Walk @type and ensure that regions don't cross the boundaries defined by | |
47cec443 | 706 | * [@base, @base + @size). Crossing regions are split at the boundaries, |
6a9ceb31 TH |
707 | * which may create at most two more regions. The index of the first |
708 | * region inside the range is returned in *@start_rgn and end in *@end_rgn. | |
709 | * | |
47cec443 | 710 | * Return: |
6a9ceb31 TH |
711 | * 0 on success, -errno on failure. |
712 | */ | |
713 | static int __init_memblock memblock_isolate_range(struct memblock_type *type, | |
714 | phys_addr_t base, phys_addr_t size, | |
715 | int *start_rgn, int *end_rgn) | |
716 | { | |
eb18f1b5 | 717 | phys_addr_t end = base + memblock_cap_size(base, &size); |
8c9c1701 AK |
718 | int idx; |
719 | struct memblock_region *rgn; | |
6a9ceb31 TH |
720 | |
721 | *start_rgn = *end_rgn = 0; | |
722 | ||
b3dc627c TH |
723 | if (!size) |
724 | return 0; | |
725 | ||
6a9ceb31 TH |
726 | /* we'll create at most two more regions */ |
727 | while (type->cnt + 2 > type->max) | |
48c3b583 | 728 | if (memblock_double_array(type, base, size) < 0) |
6a9ceb31 TH |
729 | return -ENOMEM; |
730 | ||
66e8b438 | 731 | for_each_memblock_type(idx, type, rgn) { |
6a9ceb31 TH |
732 | phys_addr_t rbase = rgn->base; |
733 | phys_addr_t rend = rbase + rgn->size; | |
734 | ||
735 | if (rbase >= end) | |
736 | break; | |
737 | if (rend <= base) | |
738 | continue; | |
739 | ||
740 | if (rbase < base) { | |
741 | /* | |
742 | * @rgn intersects from below. Split and continue | |
743 | * to process the next region - the new top half. | |
744 | */ | |
745 | rgn->base = base; | |
1440c4e2 TH |
746 | rgn->size -= base - rbase; |
747 | type->total_size -= base - rbase; | |
8c9c1701 | 748 | memblock_insert_region(type, idx, rbase, base - rbase, |
66a20757 TC |
749 | memblock_get_region_node(rgn), |
750 | rgn->flags); | |
6a9ceb31 TH |
751 | } else if (rend > end) { |
752 | /* | |
753 | * @rgn intersects from above. Split and redo the | |
754 | * current region - the new bottom half. | |
755 | */ | |
756 | rgn->base = end; | |
1440c4e2 TH |
757 | rgn->size -= end - rbase; |
758 | type->total_size -= end - rbase; | |
8c9c1701 | 759 | memblock_insert_region(type, idx--, rbase, end - rbase, |
66a20757 TC |
760 | memblock_get_region_node(rgn), |
761 | rgn->flags); | |
6a9ceb31 TH |
762 | } else { |
763 | /* @rgn is fully contained, record it */ | |
764 | if (!*end_rgn) | |
8c9c1701 AK |
765 | *start_rgn = idx; |
766 | *end_rgn = idx + 1; | |
6a9ceb31 TH |
767 | } |
768 | } | |
769 | ||
770 | return 0; | |
771 | } | |
6a9ceb31 | 772 | |
35bd16a2 | 773 | static int __init_memblock memblock_remove_range(struct memblock_type *type, |
f1af9d3a | 774 | phys_addr_t base, phys_addr_t size) |
95f72d1e | 775 | { |
71936180 TH |
776 | int start_rgn, end_rgn; |
777 | int i, ret; | |
95f72d1e | 778 | |
71936180 TH |
779 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
780 | if (ret) | |
781 | return ret; | |
95f72d1e | 782 | |
71936180 TH |
783 | for (i = end_rgn - 1; i >= start_rgn; i--) |
784 | memblock_remove_region(type, i); | |
8f7a6605 | 785 | return 0; |
95f72d1e YL |
786 | } |
787 | ||
581adcbe | 788 | int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) |
95f72d1e | 789 | { |
25cf23d7 MK |
790 | phys_addr_t end = base + size - 1; |
791 | ||
a090d711 | 792 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, |
25cf23d7 MK |
793 | &base, &end, (void *)_RET_IP_); |
794 | ||
f1af9d3a | 795 | return memblock_remove_range(&memblock.memory, base, size); |
95f72d1e YL |
796 | } |
797 | ||
77e02cf5 LT |
798 | /** |
799 | * memblock_free_ptr - free boot memory allocation | |
800 | * @ptr: starting address of the boot memory allocation | |
801 | * @size: size of the boot memory block in bytes | |
802 | * | |
803 | * Free boot memory block previously allocated by memblock_alloc_xx() API. | |
804 | * The freeing memory will not be released to the buddy allocator. | |
805 | */ | |
806 | void __init_memblock memblock_free_ptr(void *ptr, size_t size) | |
807 | { | |
808 | if (ptr) | |
809 | memblock_free(__pa(ptr), size); | |
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 | ||
a090d711 | 824 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, |
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 | ||
a090d711 | 835 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, |
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 | ||
02634a44 AK |
841 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
842 | int __init_memblock memblock_physmem_add(phys_addr_t base, phys_addr_t size) | |
843 | { | |
844 | phys_addr_t end = base + size - 1; | |
845 | ||
846 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, | |
847 | &base, &end, (void *)_RET_IP_); | |
848 | ||
77649905 | 849 | return memblock_add_range(&physmem, base, size, MAX_NUMNODES, 0); |
02634a44 AK |
850 | } |
851 | #endif | |
852 | ||
66b16edf | 853 | /** |
47cec443 MR |
854 | * memblock_setclr_flag - set or clear flag for a memory region |
855 | * @base: base address of the region | |
856 | * @size: size of the region | |
857 | * @set: set or clear the flag | |
8958b249 | 858 | * @flag: the flag to update |
66b16edf | 859 | * |
4308ce17 | 860 | * This function isolates region [@base, @base + @size), and sets/clears flag |
66b16edf | 861 | * |
47cec443 | 862 | * Return: 0 on success, -errno on failure. |
66b16edf | 863 | */ |
4308ce17 TL |
864 | static int __init_memblock memblock_setclr_flag(phys_addr_t base, |
865 | phys_addr_t size, int set, int flag) | |
66b16edf TC |
866 | { |
867 | struct memblock_type *type = &memblock.memory; | |
868 | int i, ret, start_rgn, end_rgn; | |
869 | ||
870 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); | |
871 | if (ret) | |
872 | return ret; | |
873 | ||
fe145124 MR |
874 | for (i = start_rgn; i < end_rgn; i++) { |
875 | struct memblock_region *r = &type->regions[i]; | |
876 | ||
4308ce17 | 877 | if (set) |
fe145124 | 878 | r->flags |= flag; |
4308ce17 | 879 | else |
fe145124 MR |
880 | r->flags &= ~flag; |
881 | } | |
66b16edf TC |
882 | |
883 | memblock_merge_regions(type); | |
884 | return 0; | |
885 | } | |
886 | ||
887 | /** | |
4308ce17 | 888 | * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. |
66b16edf TC |
889 | * @base: the base phys addr of the region |
890 | * @size: the size of the region | |
891 | * | |
47cec443 | 892 | * Return: 0 on success, -errno on failure. |
4308ce17 TL |
893 | */ |
894 | int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) | |
895 | { | |
896 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); | |
897 | } | |
898 | ||
899 | /** | |
900 | * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. | |
901 | * @base: the base phys addr of the region | |
902 | * @size: the size of the region | |
66b16edf | 903 | * |
47cec443 | 904 | * Return: 0 on success, -errno on failure. |
66b16edf TC |
905 | */ |
906 | int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) | |
907 | { | |
4308ce17 | 908 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); |
66b16edf TC |
909 | } |
910 | ||
a3f5bafc TL |
911 | /** |
912 | * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. | |
913 | * @base: the base phys addr of the region | |
914 | * @size: the size of the region | |
915 | * | |
47cec443 | 916 | * Return: 0 on success, -errno on failure. |
a3f5bafc TL |
917 | */ |
918 | int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) | |
919 | { | |
920 | system_has_some_mirror = true; | |
921 | ||
922 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); | |
923 | } | |
924 | ||
bf3d3cc5 AB |
925 | /** |
926 | * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. | |
927 | * @base: the base phys addr of the region | |
928 | * @size: the size of the region | |
929 | * | |
9092d4f7 MR |
930 | * The memory regions marked with %MEMBLOCK_NOMAP will not be added to the |
931 | * direct mapping of the physical memory. These regions will still be | |
932 | * covered by the memory map. The struct page representing NOMAP memory | |
933 | * frames in the memory map will be PageReserved() | |
934 | * | |
658aafc8 MR |
935 | * Note: if the memory being marked %MEMBLOCK_NOMAP was allocated from |
936 | * memblock, the caller must inform kmemleak to ignore that memory | |
937 | * | |
47cec443 | 938 | * Return: 0 on success, -errno on failure. |
bf3d3cc5 AB |
939 | */ |
940 | int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) | |
941 | { | |
6c9a5455 | 942 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); |
bf3d3cc5 | 943 | } |
a3f5bafc | 944 | |
4c546b8a AT |
945 | /** |
946 | * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. | |
947 | * @base: the base phys addr of the region | |
948 | * @size: the size of the region | |
949 | * | |
47cec443 | 950 | * Return: 0 on success, -errno on failure. |
4c546b8a AT |
951 | */ |
952 | int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) | |
953 | { | |
954 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); | |
955 | } | |
956 | ||
9f3d5eaa MR |
957 | static bool should_skip_region(struct memblock_type *type, |
958 | struct memblock_region *m, | |
959 | int nid, int flags) | |
c9a688a3 MR |
960 | { |
961 | int m_nid = memblock_get_region_node(m); | |
962 | ||
9f3d5eaa MR |
963 | /* we never skip regions when iterating memblock.reserved or physmem */ |
964 | if (type != memblock_memory) | |
965 | return false; | |
966 | ||
c9a688a3 MR |
967 | /* only memory regions are associated with nodes, check it */ |
968 | if (nid != NUMA_NO_NODE && nid != m_nid) | |
969 | return true; | |
970 | ||
971 | /* skip hotpluggable memory regions if needed */ | |
79e482e9 MR |
972 | if (movable_node_is_enabled() && memblock_is_hotpluggable(m) && |
973 | !(flags & MEMBLOCK_HOTPLUG)) | |
c9a688a3 MR |
974 | return true; |
975 | ||
976 | /* if we want mirror memory skip non-mirror memory regions */ | |
977 | if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) | |
978 | return true; | |
979 | ||
980 | /* skip nomap memory unless we were asked for it explicitly */ | |
981 | if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) | |
982 | return true; | |
983 | ||
984 | return false; | |
985 | } | |
986 | ||
35fd0808 | 987 | /** |
a2974133 | 988 | * __next_mem_range - next function for for_each_free_mem_range() etc. |
35fd0808 | 989 | * @idx: pointer to u64 loop variable |
b1154233 | 990 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 991 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
992 | * @type_a: pointer to memblock_type from where the range is taken |
993 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
994 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
995 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
996 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
35fd0808 | 997 | * |
f1af9d3a | 998 | * Find the first area from *@idx which matches @nid, fill the out |
35fd0808 | 999 | * parameters, and update *@idx for the next iteration. The lower 32bit of |
f1af9d3a PH |
1000 | * *@idx contains index into type_a and the upper 32bit indexes the |
1001 | * areas before each region in type_b. For example, if type_b regions | |
35fd0808 TH |
1002 | * look like the following, |
1003 | * | |
1004 | * 0:[0-16), 1:[32-48), 2:[128-130) | |
1005 | * | |
1006 | * The upper 32bit indexes the following regions. | |
1007 | * | |
1008 | * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) | |
1009 | * | |
1010 | * As both region arrays are sorted, the function advances the two indices | |
1011 | * in lockstep and returns each intersection. | |
1012 | */ | |
77649905 DH |
1013 | void __next_mem_range(u64 *idx, int nid, enum memblock_flags flags, |
1014 | struct memblock_type *type_a, | |
1015 | struct memblock_type *type_b, phys_addr_t *out_start, | |
1016 | phys_addr_t *out_end, int *out_nid) | |
35fd0808 | 1017 | { |
f1af9d3a PH |
1018 | int idx_a = *idx & 0xffffffff; |
1019 | int idx_b = *idx >> 32; | |
b1154233 | 1020 | |
f1af9d3a PH |
1021 | if (WARN_ONCE(nid == MAX_NUMNODES, |
1022 | "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) | |
560dca27 | 1023 | nid = NUMA_NO_NODE; |
35fd0808 | 1024 | |
f1af9d3a PH |
1025 | for (; idx_a < type_a->cnt; idx_a++) { |
1026 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1027 | ||
35fd0808 TH |
1028 | phys_addr_t m_start = m->base; |
1029 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1030 | int m_nid = memblock_get_region_node(m); |
35fd0808 | 1031 | |
9f3d5eaa | 1032 | if (should_skip_region(type_a, m, nid, flags)) |
bf3d3cc5 AB |
1033 | continue; |
1034 | ||
f1af9d3a PH |
1035 | if (!type_b) { |
1036 | if (out_start) | |
1037 | *out_start = m_start; | |
1038 | if (out_end) | |
1039 | *out_end = m_end; | |
1040 | if (out_nid) | |
1041 | *out_nid = m_nid; | |
1042 | idx_a++; | |
1043 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
1044 | return; | |
1045 | } | |
1046 | ||
1047 | /* scan areas before each reservation */ | |
1048 | for (; idx_b < type_b->cnt + 1; idx_b++) { | |
1049 | struct memblock_region *r; | |
1050 | phys_addr_t r_start; | |
1051 | phys_addr_t r_end; | |
1052 | ||
1053 | r = &type_b->regions[idx_b]; | |
1054 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1055 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1056 | r->base : PHYS_ADDR_MAX; |
35fd0808 | 1057 | |
f1af9d3a PH |
1058 | /* |
1059 | * if idx_b advanced past idx_a, | |
1060 | * break out to advance idx_a | |
1061 | */ | |
35fd0808 TH |
1062 | if (r_start >= m_end) |
1063 | break; | |
1064 | /* if the two regions intersect, we're done */ | |
1065 | if (m_start < r_end) { | |
1066 | if (out_start) | |
f1af9d3a PH |
1067 | *out_start = |
1068 | max(m_start, r_start); | |
35fd0808 TH |
1069 | if (out_end) |
1070 | *out_end = min(m_end, r_end); | |
1071 | if (out_nid) | |
f1af9d3a | 1072 | *out_nid = m_nid; |
35fd0808 | 1073 | /* |
f1af9d3a PH |
1074 | * The region which ends first is |
1075 | * advanced for the next iteration. | |
35fd0808 TH |
1076 | */ |
1077 | if (m_end <= r_end) | |
f1af9d3a | 1078 | idx_a++; |
35fd0808 | 1079 | else |
f1af9d3a PH |
1080 | idx_b++; |
1081 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
35fd0808 TH |
1082 | return; |
1083 | } | |
1084 | } | |
1085 | } | |
1086 | ||
1087 | /* signal end of iteration */ | |
1088 | *idx = ULLONG_MAX; | |
1089 | } | |
1090 | ||
7bd0b0f0 | 1091 | /** |
f1af9d3a PH |
1092 | * __next_mem_range_rev - generic next function for for_each_*_range_rev() |
1093 | * | |
7bd0b0f0 | 1094 | * @idx: pointer to u64 loop variable |
ad5ea8cd | 1095 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 1096 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
1097 | * @type_a: pointer to memblock_type from where the range is taken |
1098 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
1099 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
1100 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
1101 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
7bd0b0f0 | 1102 | * |
47cec443 MR |
1103 | * Finds the next range from type_a which is not marked as unsuitable |
1104 | * in type_b. | |
1105 | * | |
f1af9d3a | 1106 | * Reverse of __next_mem_range(). |
7bd0b0f0 | 1107 | */ |
e1720fee MR |
1108 | void __init_memblock __next_mem_range_rev(u64 *idx, int nid, |
1109 | enum memblock_flags flags, | |
f1af9d3a PH |
1110 | struct memblock_type *type_a, |
1111 | struct memblock_type *type_b, | |
1112 | phys_addr_t *out_start, | |
1113 | phys_addr_t *out_end, int *out_nid) | |
7bd0b0f0 | 1114 | { |
f1af9d3a PH |
1115 | int idx_a = *idx & 0xffffffff; |
1116 | int idx_b = *idx >> 32; | |
b1154233 | 1117 | |
560dca27 GS |
1118 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1119 | nid = NUMA_NO_NODE; | |
7bd0b0f0 TH |
1120 | |
1121 | if (*idx == (u64)ULLONG_MAX) { | |
f1af9d3a | 1122 | idx_a = type_a->cnt - 1; |
e47608ab | 1123 | if (type_b != NULL) |
1124 | idx_b = type_b->cnt; | |
1125 | else | |
1126 | idx_b = 0; | |
7bd0b0f0 TH |
1127 | } |
1128 | ||
f1af9d3a PH |
1129 | for (; idx_a >= 0; idx_a--) { |
1130 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1131 | ||
7bd0b0f0 TH |
1132 | phys_addr_t m_start = m->base; |
1133 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1134 | int m_nid = memblock_get_region_node(m); |
7bd0b0f0 | 1135 | |
9f3d5eaa | 1136 | if (should_skip_region(type_a, m, nid, flags)) |
bf3d3cc5 AB |
1137 | continue; |
1138 | ||
f1af9d3a PH |
1139 | if (!type_b) { |
1140 | if (out_start) | |
1141 | *out_start = m_start; | |
1142 | if (out_end) | |
1143 | *out_end = m_end; | |
1144 | if (out_nid) | |
1145 | *out_nid = m_nid; | |
fb399b48 | 1146 | idx_a--; |
f1af9d3a PH |
1147 | *idx = (u32)idx_a | (u64)idx_b << 32; |
1148 | return; | |
1149 | } | |
1150 | ||
1151 | /* scan areas before each reservation */ | |
1152 | for (; idx_b >= 0; idx_b--) { | |
1153 | struct memblock_region *r; | |
1154 | phys_addr_t r_start; | |
1155 | phys_addr_t r_end; | |
1156 | ||
1157 | r = &type_b->regions[idx_b]; | |
1158 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1159 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1160 | r->base : PHYS_ADDR_MAX; |
f1af9d3a PH |
1161 | /* |
1162 | * if idx_b advanced past idx_a, | |
1163 | * break out to advance idx_a | |
1164 | */ | |
7bd0b0f0 | 1165 | |
7bd0b0f0 TH |
1166 | if (r_end <= m_start) |
1167 | break; | |
1168 | /* if the two regions intersect, we're done */ | |
1169 | if (m_end > r_start) { | |
1170 | if (out_start) | |
1171 | *out_start = max(m_start, r_start); | |
1172 | if (out_end) | |
1173 | *out_end = min(m_end, r_end); | |
1174 | if (out_nid) | |
f1af9d3a | 1175 | *out_nid = m_nid; |
7bd0b0f0 | 1176 | if (m_start >= r_start) |
f1af9d3a | 1177 | idx_a--; |
7bd0b0f0 | 1178 | else |
f1af9d3a PH |
1179 | idx_b--; |
1180 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
7bd0b0f0 TH |
1181 | return; |
1182 | } | |
1183 | } | |
1184 | } | |
f1af9d3a | 1185 | /* signal end of iteration */ |
7bd0b0f0 TH |
1186 | *idx = ULLONG_MAX; |
1187 | } | |
1188 | ||
7c0caeb8 | 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; | |
d622abf7 | 1198 | int r_nid; |
7c0caeb8 TH |
1199 | |
1200 | while (++*idx < type->cnt) { | |
1201 | r = &type->regions[*idx]; | |
d622abf7 | 1202 | r_nid = memblock_get_region_node(r); |
7c0caeb8 TH |
1203 | |
1204 | if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) | |
1205 | continue; | |
d622abf7 | 1206 | if (nid == MAX_NUMNODES || nid == r_nid) |
7c0caeb8 TH |
1207 | break; |
1208 | } | |
1209 | if (*idx >= type->cnt) { | |
1210 | *idx = -1; | |
1211 | return; | |
1212 | } | |
1213 | ||
1214 | if (out_start_pfn) | |
1215 | *out_start_pfn = PFN_UP(r->base); | |
1216 | if (out_end_pfn) | |
1217 | *out_end_pfn = PFN_DOWN(r->base + r->size); | |
1218 | if (out_nid) | |
d622abf7 | 1219 | *out_nid = r_nid; |
7c0caeb8 TH |
1220 | } |
1221 | ||
1222 | /** | |
1223 | * memblock_set_node - set node ID on memblock regions | |
1224 | * @base: base of area to set node ID for | |
1225 | * @size: size of area to set node ID for | |
e7e8de59 | 1226 | * @type: memblock type to set node ID for |
7c0caeb8 TH |
1227 | * @nid: node ID to set |
1228 | * | |
47cec443 | 1229 | * Set the nid of memblock @type regions in [@base, @base + @size) to @nid. |
7c0caeb8 TH |
1230 | * Regions which cross the area boundaries are split as necessary. |
1231 | * | |
47cec443 | 1232 | * Return: |
7c0caeb8 TH |
1233 | * 0 on success, -errno on failure. |
1234 | */ | |
1235 | int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, | |
e7e8de59 | 1236 | struct memblock_type *type, int nid) |
7c0caeb8 | 1237 | { |
a9ee6cf5 | 1238 | #ifdef CONFIG_NUMA |
6a9ceb31 TH |
1239 | int start_rgn, end_rgn; |
1240 | int i, ret; | |
7c0caeb8 | 1241 | |
6a9ceb31 TH |
1242 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
1243 | if (ret) | |
1244 | return ret; | |
7c0caeb8 | 1245 | |
6a9ceb31 | 1246 | for (i = start_rgn; i < end_rgn; i++) |
e9d24ad3 | 1247 | memblock_set_region_node(&type->regions[i], nid); |
7c0caeb8 TH |
1248 | |
1249 | memblock_merge_regions(type); | |
3f08a302 | 1250 | #endif |
7c0caeb8 TH |
1251 | return 0; |
1252 | } | |
3f08a302 | 1253 | |
837566e7 AD |
1254 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT |
1255 | /** | |
1256 | * __next_mem_pfn_range_in_zone - iterator for for_each_*_range_in_zone() | |
1257 | * | |
1258 | * @idx: pointer to u64 loop variable | |
1259 | * @zone: zone in which all of the memory blocks reside | |
1260 | * @out_spfn: ptr to ulong for start pfn of the range, can be %NULL | |
1261 | * @out_epfn: ptr to ulong for end pfn of the range, can be %NULL | |
1262 | * | |
1263 | * This function is meant to be a zone/pfn specific wrapper for the | |
1264 | * for_each_mem_range type iterators. Specifically they are used in the | |
1265 | * deferred memory init routines and as such we were duplicating much of | |
1266 | * this logic throughout the code. So instead of having it in multiple | |
1267 | * locations it seemed like it would make more sense to centralize this to | |
1268 | * one new iterator that does everything they need. | |
1269 | */ | |
1270 | void __init_memblock | |
1271 | __next_mem_pfn_range_in_zone(u64 *idx, struct zone *zone, | |
1272 | unsigned long *out_spfn, unsigned long *out_epfn) | |
1273 | { | |
1274 | int zone_nid = zone_to_nid(zone); | |
1275 | phys_addr_t spa, epa; | |
1276 | int nid; | |
1277 | ||
1278 | __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, | |
1279 | &memblock.memory, &memblock.reserved, | |
1280 | &spa, &epa, &nid); | |
1281 | ||
1282 | while (*idx != U64_MAX) { | |
1283 | unsigned long epfn = PFN_DOWN(epa); | |
1284 | unsigned long spfn = PFN_UP(spa); | |
1285 | ||
1286 | /* | |
1287 | * Verify the end is at least past the start of the zone and | |
1288 | * that we have at least one PFN to initialize. | |
1289 | */ | |
1290 | if (zone->zone_start_pfn < epfn && spfn < epfn) { | |
1291 | /* if we went too far just stop searching */ | |
1292 | if (zone_end_pfn(zone) <= spfn) { | |
1293 | *idx = U64_MAX; | |
1294 | break; | |
1295 | } | |
1296 | ||
1297 | if (out_spfn) | |
1298 | *out_spfn = max(zone->zone_start_pfn, spfn); | |
1299 | if (out_epfn) | |
1300 | *out_epfn = min(zone_end_pfn(zone), epfn); | |
1301 | ||
1302 | return; | |
1303 | } | |
1304 | ||
1305 | __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, | |
1306 | &memblock.memory, &memblock.reserved, | |
1307 | &spa, &epa, &nid); | |
1308 | } | |
1309 | ||
1310 | /* signal end of iteration */ | |
1311 | if (out_spfn) | |
1312 | *out_spfn = ULONG_MAX; | |
1313 | if (out_epfn) | |
1314 | *out_epfn = 0; | |
1315 | } | |
1316 | ||
1317 | #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ | |
7c0caeb8 | 1318 | |
92d12f95 MR |
1319 | /** |
1320 | * memblock_alloc_range_nid - allocate boot memory block | |
1321 | * @size: size of memory block to be allocated in bytes | |
1322 | * @align: alignment of the region and block's size | |
1323 | * @start: the lower bound of the memory region to allocate (phys address) | |
1324 | * @end: the upper bound of the memory region to allocate (phys address) | |
1325 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
0ac398b1 | 1326 | * @exact_nid: control the allocation fall back to other nodes |
92d12f95 MR |
1327 | * |
1328 | * The allocation is performed from memory region limited by | |
95830666 | 1329 | * memblock.current_limit if @end == %MEMBLOCK_ALLOC_ACCESSIBLE. |
92d12f95 | 1330 | * |
0ac398b1 YY |
1331 | * If the specified node can not hold the requested memory and @exact_nid |
1332 | * is false, the allocation falls back to any node in the system. | |
92d12f95 MR |
1333 | * |
1334 | * For systems with memory mirroring, the allocation is attempted first | |
1335 | * from the regions with mirroring enabled and then retried from any | |
1336 | * memory region. | |
1337 | * | |
1338 | * In addition, function sets the min_count to 0 using kmemleak_alloc_phys for | |
1339 | * allocated boot memory block, so that it is never reported as leaks. | |
1340 | * | |
1341 | * Return: | |
1342 | * Physical address of allocated memory block on success, %0 on failure. | |
1343 | */ | |
8676af1f | 1344 | phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, |
2bfc2862 | 1345 | phys_addr_t align, phys_addr_t start, |
0ac398b1 YY |
1346 | phys_addr_t end, int nid, |
1347 | bool exact_nid) | |
95f72d1e | 1348 | { |
92d12f95 | 1349 | enum memblock_flags flags = choose_memblock_flags(); |
6ed311b2 | 1350 | phys_addr_t found; |
95f72d1e | 1351 | |
92d12f95 MR |
1352 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1353 | nid = NUMA_NO_NODE; | |
1354 | ||
2f770806 MR |
1355 | if (!align) { |
1356 | /* Can't use WARNs this early in boot on powerpc */ | |
1357 | dump_stack(); | |
1358 | align = SMP_CACHE_BYTES; | |
1359 | } | |
1360 | ||
92d12f95 | 1361 | again: |
fc6daaf9 TL |
1362 | found = memblock_find_in_range_node(size, align, start, end, nid, |
1363 | flags); | |
92d12f95 MR |
1364 | if (found && !memblock_reserve(found, size)) |
1365 | goto done; | |
1366 | ||
0ac398b1 | 1367 | if (nid != NUMA_NO_NODE && !exact_nid) { |
92d12f95 MR |
1368 | found = memblock_find_in_range_node(size, align, start, |
1369 | end, NUMA_NO_NODE, | |
1370 | flags); | |
1371 | if (found && !memblock_reserve(found, size)) | |
1372 | goto done; | |
1373 | } | |
1374 | ||
1375 | if (flags & MEMBLOCK_MIRROR) { | |
1376 | flags &= ~MEMBLOCK_MIRROR; | |
1377 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
1378 | &size); | |
1379 | goto again; | |
1380 | } | |
1381 | ||
1382 | return 0; | |
1383 | ||
1384 | done: | |
1385 | /* Skip kmemleak for kasan_init() due to high volume. */ | |
1386 | if (end != MEMBLOCK_ALLOC_KASAN) | |
aedf95ea | 1387 | /* |
92d12f95 MR |
1388 | * The min_count is set to 0 so that memblock allocated |
1389 | * blocks are never reported as leaks. This is because many | |
1390 | * of these blocks are only referred via the physical | |
1391 | * address which is not looked up by kmemleak. | |
aedf95ea | 1392 | */ |
9099daed | 1393 | kmemleak_alloc_phys(found, size, 0, 0); |
92d12f95 MR |
1394 | |
1395 | return found; | |
95f72d1e YL |
1396 | } |
1397 | ||
a2974133 MR |
1398 | /** |
1399 | * memblock_phys_alloc_range - allocate a memory block inside specified range | |
1400 | * @size: size of memory block to be allocated in bytes | |
1401 | * @align: alignment of the region and block's size | |
1402 | * @start: the lower bound of the memory region to allocate (physical address) | |
1403 | * @end: the upper bound of the memory region to allocate (physical address) | |
1404 | * | |
1405 | * Allocate @size bytes in the between @start and @end. | |
1406 | * | |
1407 | * Return: physical address of the allocated memory block on success, | |
1408 | * %0 on failure. | |
1409 | */ | |
8a770c2a MR |
1410 | phys_addr_t __init memblock_phys_alloc_range(phys_addr_t size, |
1411 | phys_addr_t align, | |
1412 | phys_addr_t start, | |
1413 | phys_addr_t end) | |
2bfc2862 | 1414 | { |
b5cf2d6c FM |
1415 | memblock_dbg("%s: %llu bytes align=0x%llx from=%pa max_addr=%pa %pS\n", |
1416 | __func__, (u64)size, (u64)align, &start, &end, | |
1417 | (void *)_RET_IP_); | |
0ac398b1 YY |
1418 | return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE, |
1419 | false); | |
7bd0b0f0 TH |
1420 | } |
1421 | ||
a2974133 | 1422 | /** |
17cbe038 | 1423 | * memblock_phys_alloc_try_nid - allocate a memory block from specified NUMA node |
a2974133 MR |
1424 | * @size: size of memory block to be allocated in bytes |
1425 | * @align: alignment of the region and block's size | |
1426 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1427 | * | |
1428 | * Allocates memory block from the specified NUMA node. If the node | |
1429 | * has no available memory, attempts to allocated from any node in the | |
1430 | * system. | |
1431 | * | |
1432 | * Return: physical address of the allocated memory block on success, | |
1433 | * %0 on failure. | |
1434 | */ | |
9a8dd708 | 1435 | phys_addr_t __init memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) |
9d1e2492 | 1436 | { |
33755574 | 1437 | return memblock_alloc_range_nid(size, align, 0, |
0ac398b1 | 1438 | MEMBLOCK_ALLOC_ACCESSIBLE, nid, false); |
95f72d1e YL |
1439 | } |
1440 | ||
26f09e9b | 1441 | /** |
eb31d559 | 1442 | * memblock_alloc_internal - allocate boot memory block |
26f09e9b SS |
1443 | * @size: size of memory block to be allocated in bytes |
1444 | * @align: alignment of the region and block's size | |
1445 | * @min_addr: the lower bound of the memory region to allocate (phys address) | |
1446 | * @max_addr: the upper bound of the memory region to allocate (phys address) | |
1447 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
0ac398b1 | 1448 | * @exact_nid: control the allocation fall back to other nodes |
26f09e9b | 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, | |
0ac398b1 | 1464 | int nid, bool exact_nid) |
26f09e9b SS |
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 | ||
f3057ad7 MR |
1476 | if (max_addr > memblock.current_limit) |
1477 | max_addr = memblock.current_limit; | |
1478 | ||
0ac398b1 YY |
1479 | alloc = memblock_alloc_range_nid(size, align, min_addr, max_addr, nid, |
1480 | exact_nid); | |
26f09e9b | 1481 | |
92d12f95 MR |
1482 | /* retry allocation without lower limit */ |
1483 | if (!alloc && min_addr) | |
0ac398b1 YY |
1484 | alloc = memblock_alloc_range_nid(size, align, 0, max_addr, nid, |
1485 | exact_nid); | |
26f09e9b | 1486 | |
92d12f95 MR |
1487 | if (!alloc) |
1488 | return NULL; | |
26f09e9b | 1489 | |
92d12f95 | 1490 | return phys_to_virt(alloc); |
26f09e9b SS |
1491 | } |
1492 | ||
0ac398b1 YY |
1493 | /** |
1494 | * memblock_alloc_exact_nid_raw - allocate boot memory block on the exact node | |
1495 | * without zeroing memory | |
1496 | * @size: size of memory block to be allocated in bytes | |
1497 | * @align: alignment of the region and block's size | |
1498 | * @min_addr: the lower bound of the memory region from where the allocation | |
1499 | * is preferred (phys address) | |
1500 | * @max_addr: the upper bound of the memory region from where the allocation | |
1501 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to | |
1502 | * allocate only from memory limited by memblock.current_limit value | |
1503 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1504 | * | |
1505 | * Public function, provides additional debug information (including caller | |
1506 | * info), if enabled. Does not zero allocated memory. | |
1507 | * | |
1508 | * Return: | |
1509 | * Virtual address of allocated memory block on success, NULL on failure. | |
1510 | */ | |
1511 | void * __init memblock_alloc_exact_nid_raw( | |
1512 | phys_addr_t size, phys_addr_t align, | |
1513 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1514 | int nid) | |
1515 | { | |
0ac398b1 YY |
1516 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
1517 | __func__, (u64)size, (u64)align, nid, &min_addr, | |
1518 | &max_addr, (void *)_RET_IP_); | |
1519 | ||
08678804 MR |
1520 | return memblock_alloc_internal(size, align, min_addr, max_addr, nid, |
1521 | true); | |
0ac398b1 YY |
1522 | } |
1523 | ||
ea1f5f37 | 1524 | /** |
eb31d559 | 1525 | * memblock_alloc_try_nid_raw - allocate boot memory block without zeroing |
ea1f5f37 PT |
1526 | * memory and without panicking |
1527 | * @size: size of memory block to be allocated in bytes | |
1528 | * @align: alignment of the region and block's size | |
1529 | * @min_addr: the lower bound of the memory region from where the allocation | |
1530 | * is preferred (phys address) | |
1531 | * @max_addr: the upper bound of the memory region from where the allocation | |
97ad1087 | 1532 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to |
ea1f5f37 PT |
1533 | * allocate only from memory limited by memblock.current_limit value |
1534 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1535 | * | |
1536 | * Public function, provides additional debug information (including caller | |
1537 | * info), if enabled. Does not zero allocated memory, does not panic if request | |
1538 | * cannot be satisfied. | |
1539 | * | |
47cec443 | 1540 | * Return: |
ea1f5f37 PT |
1541 | * Virtual address of allocated memory block on success, NULL on failure. |
1542 | */ | |
eb31d559 | 1543 | void * __init memblock_alloc_try_nid_raw( |
ea1f5f37 PT |
1544 | phys_addr_t size, phys_addr_t align, |
1545 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1546 | int nid) | |
1547 | { | |
d75f773c | 1548 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
a36aab89 MR |
1549 | __func__, (u64)size, (u64)align, nid, &min_addr, |
1550 | &max_addr, (void *)_RET_IP_); | |
ea1f5f37 | 1551 | |
08678804 MR |
1552 | return memblock_alloc_internal(size, align, min_addr, max_addr, nid, |
1553 | false); | |
ea1f5f37 PT |
1554 | } |
1555 | ||
26f09e9b | 1556 | /** |
c0dbe825 | 1557 | * memblock_alloc_try_nid - allocate boot memory block |
26f09e9b SS |
1558 | * @size: size of memory block to be allocated in bytes |
1559 | * @align: alignment of the region and block's size | |
1560 | * @min_addr: the lower bound of the memory region from where the allocation | |
1561 | * is preferred (phys address) | |
1562 | * @max_addr: the upper bound of the memory region from where the allocation | |
97ad1087 | 1563 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to |
26f09e9b SS |
1564 | * allocate only from memory limited by memblock.current_limit value |
1565 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1566 | * | |
c0dbe825 MR |
1567 | * Public function, provides additional debug information (including caller |
1568 | * info), if enabled. This function zeroes the allocated memory. | |
26f09e9b | 1569 | * |
47cec443 | 1570 | * Return: |
26f09e9b SS |
1571 | * Virtual address of allocated memory block on success, NULL on failure. |
1572 | */ | |
eb31d559 | 1573 | void * __init memblock_alloc_try_nid( |
26f09e9b SS |
1574 | phys_addr_t size, phys_addr_t align, |
1575 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1576 | int nid) | |
1577 | { | |
1578 | void *ptr; | |
1579 | ||
d75f773c | 1580 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
a36aab89 MR |
1581 | __func__, (u64)size, (u64)align, nid, &min_addr, |
1582 | &max_addr, (void *)_RET_IP_); | |
eb31d559 | 1583 | ptr = memblock_alloc_internal(size, align, |
0ac398b1 | 1584 | min_addr, max_addr, nid, false); |
c0dbe825 | 1585 | if (ptr) |
ea1f5f37 | 1586 | memset(ptr, 0, size); |
26f09e9b | 1587 | |
c0dbe825 | 1588 | return ptr; |
26f09e9b SS |
1589 | } |
1590 | ||
48a833cc | 1591 | /** |
a2974133 | 1592 | * __memblock_free_late - free pages directly to buddy allocator |
48a833cc | 1593 | * @base: phys starting address of the boot memory block |
26f09e9b SS |
1594 | * @size: size of the boot memory block in bytes |
1595 | * | |
a2974133 | 1596 | * This is only useful when the memblock allocator has already been torn |
26f09e9b | 1597 | * down, but we are still initializing the system. Pages are released directly |
a2974133 | 1598 | * to the buddy allocator. |
26f09e9b SS |
1599 | */ |
1600 | void __init __memblock_free_late(phys_addr_t base, phys_addr_t size) | |
1601 | { | |
a36aab89 | 1602 | phys_addr_t cursor, end; |
26f09e9b | 1603 | |
a36aab89 | 1604 | end = base + size - 1; |
d75f773c | 1605 | memblock_dbg("%s: [%pa-%pa] %pS\n", |
a36aab89 | 1606 | __func__, &base, &end, (void *)_RET_IP_); |
9099daed | 1607 | kmemleak_free_part_phys(base, size); |
26f09e9b SS |
1608 | cursor = PFN_UP(base); |
1609 | end = PFN_DOWN(base + size); | |
1610 | ||
1611 | for (; cursor < end; cursor++) { | |
7c2ee349 | 1612 | memblock_free_pages(pfn_to_page(cursor), cursor, 0); |
ca79b0c2 | 1613 | totalram_pages_inc(); |
26f09e9b SS |
1614 | } |
1615 | } | |
9d1e2492 BH |
1616 | |
1617 | /* | |
1618 | * Remaining API functions | |
1619 | */ | |
1620 | ||
1f1ffb8a | 1621 | phys_addr_t __init_memblock memblock_phys_mem_size(void) |
95f72d1e | 1622 | { |
1440c4e2 | 1623 | return memblock.memory.total_size; |
95f72d1e YL |
1624 | } |
1625 | ||
8907de5d SD |
1626 | phys_addr_t __init_memblock memblock_reserved_size(void) |
1627 | { | |
1628 | return memblock.reserved.total_size; | |
1629 | } | |
1630 | ||
0a93ebef SR |
1631 | /* lowest address */ |
1632 | phys_addr_t __init_memblock memblock_start_of_DRAM(void) | |
1633 | { | |
1634 | return memblock.memory.regions[0].base; | |
1635 | } | |
1636 | ||
10d06439 | 1637 | phys_addr_t __init_memblock memblock_end_of_DRAM(void) |
95f72d1e YL |
1638 | { |
1639 | int idx = memblock.memory.cnt - 1; | |
1640 | ||
e3239ff9 | 1641 | return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); |
95f72d1e YL |
1642 | } |
1643 | ||
a571d4eb | 1644 | static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) |
95f72d1e | 1645 | { |
1c4bc43d | 1646 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
136199f0 | 1647 | struct memblock_region *r; |
95f72d1e | 1648 | |
a571d4eb DC |
1649 | /* |
1650 | * translate the memory @limit size into the max address within one of | |
1651 | * the memory memblock regions, if the @limit exceeds the total size | |
1c4bc43d | 1652 | * of those regions, max_addr will keep original value PHYS_ADDR_MAX |
a571d4eb | 1653 | */ |
cc6de168 | 1654 | for_each_mem_region(r) { |
c0ce8fef TH |
1655 | if (limit <= r->size) { |
1656 | max_addr = r->base + limit; | |
1657 | break; | |
95f72d1e | 1658 | } |
c0ce8fef | 1659 | limit -= r->size; |
95f72d1e | 1660 | } |
c0ce8fef | 1661 | |
a571d4eb DC |
1662 | return max_addr; |
1663 | } | |
1664 | ||
1665 | void __init memblock_enforce_memory_limit(phys_addr_t limit) | |
1666 | { | |
49aef717 | 1667 | phys_addr_t max_addr; |
a571d4eb DC |
1668 | |
1669 | if (!limit) | |
1670 | return; | |
1671 | ||
1672 | max_addr = __find_max_addr(limit); | |
1673 | ||
1674 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1675 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1676 | return; |
1677 | ||
c0ce8fef | 1678 | /* truncate both memory and reserved regions */ |
f1af9d3a | 1679 | memblock_remove_range(&memblock.memory, max_addr, |
1c4bc43d | 1680 | PHYS_ADDR_MAX); |
f1af9d3a | 1681 | memblock_remove_range(&memblock.reserved, max_addr, |
1c4bc43d | 1682 | PHYS_ADDR_MAX); |
95f72d1e YL |
1683 | } |
1684 | ||
c9ca9b4e AT |
1685 | void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) |
1686 | { | |
1687 | int start_rgn, end_rgn; | |
1688 | int i, ret; | |
1689 | ||
1690 | if (!size) | |
1691 | return; | |
1692 | ||
5173ed72 | 1693 | if (!memblock_memory->total_size) { |
e888fa7b GU |
1694 | pr_warn("%s: No memory registered yet\n", __func__); |
1695 | return; | |
1696 | } | |
1697 | ||
c9ca9b4e AT |
1698 | ret = memblock_isolate_range(&memblock.memory, base, size, |
1699 | &start_rgn, &end_rgn); | |
1700 | if (ret) | |
1701 | return; | |
1702 | ||
1703 | /* remove all the MAP regions */ | |
1704 | for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) | |
1705 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1706 | memblock_remove_region(&memblock.memory, i); | |
1707 | ||
1708 | for (i = start_rgn - 1; i >= 0; i--) | |
1709 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1710 | memblock_remove_region(&memblock.memory, i); | |
1711 | ||
1712 | /* truncate the reserved regions */ | |
1713 | memblock_remove_range(&memblock.reserved, 0, base); | |
1714 | memblock_remove_range(&memblock.reserved, | |
1c4bc43d | 1715 | base + size, PHYS_ADDR_MAX); |
c9ca9b4e AT |
1716 | } |
1717 | ||
a571d4eb DC |
1718 | void __init memblock_mem_limit_remove_map(phys_addr_t limit) |
1719 | { | |
a571d4eb | 1720 | phys_addr_t max_addr; |
a571d4eb DC |
1721 | |
1722 | if (!limit) | |
1723 | return; | |
1724 | ||
1725 | max_addr = __find_max_addr(limit); | |
1726 | ||
1727 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1728 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1729 | return; |
1730 | ||
c9ca9b4e | 1731 | memblock_cap_memory_range(0, max_addr); |
a571d4eb DC |
1732 | } |
1733 | ||
cd79481d | 1734 | static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) |
72d4b0b4 BH |
1735 | { |
1736 | unsigned int left = 0, right = type->cnt; | |
1737 | ||
1738 | do { | |
1739 | unsigned int mid = (right + left) / 2; | |
1740 | ||
1741 | if (addr < type->regions[mid].base) | |
1742 | right = mid; | |
1743 | else if (addr >= (type->regions[mid].base + | |
1744 | type->regions[mid].size)) | |
1745 | left = mid + 1; | |
1746 | else | |
1747 | return mid; | |
1748 | } while (left < right); | |
1749 | return -1; | |
1750 | } | |
1751 | ||
f5a222dc | 1752 | bool __init_memblock memblock_is_reserved(phys_addr_t addr) |
95f72d1e | 1753 | { |
72d4b0b4 BH |
1754 | return memblock_search(&memblock.reserved, addr) != -1; |
1755 | } | |
95f72d1e | 1756 | |
b4ad0c7e | 1757 | bool __init_memblock memblock_is_memory(phys_addr_t addr) |
72d4b0b4 BH |
1758 | { |
1759 | return memblock_search(&memblock.memory, addr) != -1; | |
1760 | } | |
1761 | ||
937f0c26 | 1762 | bool __init_memblock memblock_is_map_memory(phys_addr_t addr) |
bf3d3cc5 AB |
1763 | { |
1764 | int i = memblock_search(&memblock.memory, addr); | |
1765 | ||
1766 | if (i == -1) | |
1767 | return false; | |
1768 | return !memblock_is_nomap(&memblock.memory.regions[i]); | |
1769 | } | |
1770 | ||
e76b63f8 YL |
1771 | int __init_memblock memblock_search_pfn_nid(unsigned long pfn, |
1772 | unsigned long *start_pfn, unsigned long *end_pfn) | |
1773 | { | |
1774 | struct memblock_type *type = &memblock.memory; | |
16763230 | 1775 | int mid = memblock_search(type, PFN_PHYS(pfn)); |
e76b63f8 YL |
1776 | |
1777 | if (mid == -1) | |
1778 | return -1; | |
1779 | ||
f7e2f7e8 FF |
1780 | *start_pfn = PFN_DOWN(type->regions[mid].base); |
1781 | *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); | |
e76b63f8 | 1782 | |
d622abf7 | 1783 | return memblock_get_region_node(&type->regions[mid]); |
e76b63f8 | 1784 | } |
e76b63f8 | 1785 | |
eab30949 SB |
1786 | /** |
1787 | * memblock_is_region_memory - check if a region is a subset of memory | |
1788 | * @base: base of region to check | |
1789 | * @size: size of region to check | |
1790 | * | |
47cec443 | 1791 | * Check if the region [@base, @base + @size) is a subset of a memory block. |
eab30949 | 1792 | * |
47cec443 | 1793 | * Return: |
eab30949 SB |
1794 | * 0 if false, non-zero if true |
1795 | */ | |
937f0c26 | 1796 | bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) |
72d4b0b4 | 1797 | { |
abb65272 | 1798 | int idx = memblock_search(&memblock.memory, base); |
eb18f1b5 | 1799 | phys_addr_t end = base + memblock_cap_size(base, &size); |
72d4b0b4 BH |
1800 | |
1801 | if (idx == -1) | |
937f0c26 | 1802 | return false; |
ef415ef4 | 1803 | return (memblock.memory.regions[idx].base + |
eb18f1b5 | 1804 | memblock.memory.regions[idx].size) >= end; |
95f72d1e YL |
1805 | } |
1806 | ||
eab30949 SB |
1807 | /** |
1808 | * memblock_is_region_reserved - check if a region intersects reserved memory | |
1809 | * @base: base of region to check | |
1810 | * @size: size of region to check | |
1811 | * | |
47cec443 MR |
1812 | * Check if the region [@base, @base + @size) intersects a reserved |
1813 | * memory block. | |
eab30949 | 1814 | * |
47cec443 | 1815 | * Return: |
c5c5c9d1 | 1816 | * True if they intersect, false if not. |
eab30949 | 1817 | */ |
c5c5c9d1 | 1818 | bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) |
95f72d1e | 1819 | { |
c5c5c9d1 | 1820 | return memblock_overlaps_region(&memblock.reserved, base, size); |
95f72d1e YL |
1821 | } |
1822 | ||
6ede1fd3 YL |
1823 | void __init_memblock memblock_trim_memory(phys_addr_t align) |
1824 | { | |
6ede1fd3 | 1825 | phys_addr_t start, end, orig_start, orig_end; |
136199f0 | 1826 | struct memblock_region *r; |
6ede1fd3 | 1827 | |
cc6de168 | 1828 | for_each_mem_region(r) { |
136199f0 EM |
1829 | orig_start = r->base; |
1830 | orig_end = r->base + r->size; | |
6ede1fd3 YL |
1831 | start = round_up(orig_start, align); |
1832 | end = round_down(orig_end, align); | |
1833 | ||
1834 | if (start == orig_start && end == orig_end) | |
1835 | continue; | |
1836 | ||
1837 | if (start < end) { | |
136199f0 EM |
1838 | r->base = start; |
1839 | r->size = end - start; | |
6ede1fd3 | 1840 | } else { |
136199f0 EM |
1841 | memblock_remove_region(&memblock.memory, |
1842 | r - memblock.memory.regions); | |
1843 | r--; | |
6ede1fd3 YL |
1844 | } |
1845 | } | |
1846 | } | |
e63075a3 | 1847 | |
3661ca66 | 1848 | void __init_memblock memblock_set_current_limit(phys_addr_t limit) |
e63075a3 BH |
1849 | { |
1850 | memblock.current_limit = limit; | |
1851 | } | |
1852 | ||
fec51014 LA |
1853 | phys_addr_t __init_memblock memblock_get_current_limit(void) |
1854 | { | |
1855 | return memblock.current_limit; | |
1856 | } | |
1857 | ||
0262d9c8 | 1858 | static void __init_memblock memblock_dump(struct memblock_type *type) |
6ed311b2 | 1859 | { |
5d63f81c | 1860 | phys_addr_t base, end, size; |
e1720fee | 1861 | enum memblock_flags flags; |
8c9c1701 AK |
1862 | int idx; |
1863 | struct memblock_region *rgn; | |
6ed311b2 | 1864 | |
0262d9c8 | 1865 | pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); |
6ed311b2 | 1866 | |
66e8b438 | 1867 | for_each_memblock_type(idx, type, rgn) { |
7c0caeb8 TH |
1868 | char nid_buf[32] = ""; |
1869 | ||
1870 | base = rgn->base; | |
1871 | size = rgn->size; | |
5d63f81c | 1872 | end = base + size - 1; |
66a20757 | 1873 | flags = rgn->flags; |
a9ee6cf5 | 1874 | #ifdef CONFIG_NUMA |
7c0caeb8 TH |
1875 | if (memblock_get_region_node(rgn) != MAX_NUMNODES) |
1876 | snprintf(nid_buf, sizeof(nid_buf), " on node %d", | |
1877 | memblock_get_region_node(rgn)); | |
1878 | #endif | |
e1720fee | 1879 | pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#x\n", |
0262d9c8 | 1880 | type->name, idx, &base, &end, &size, nid_buf, flags); |
6ed311b2 BH |
1881 | } |
1882 | } | |
1883 | ||
87c55870 | 1884 | static void __init_memblock __memblock_dump_all(void) |
6ed311b2 | 1885 | { |
6ed311b2 | 1886 | pr_info("MEMBLOCK configuration:\n"); |
5d63f81c MC |
1887 | pr_info(" memory size = %pa reserved size = %pa\n", |
1888 | &memblock.memory.total_size, | |
1889 | &memblock.reserved.total_size); | |
6ed311b2 | 1890 | |
0262d9c8 HC |
1891 | memblock_dump(&memblock.memory); |
1892 | memblock_dump(&memblock.reserved); | |
409efd4c | 1893 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
77649905 | 1894 | memblock_dump(&physmem); |
409efd4c | 1895 | #endif |
6ed311b2 BH |
1896 | } |
1897 | ||
87c55870 MR |
1898 | void __init_memblock memblock_dump_all(void) |
1899 | { | |
1900 | if (memblock_debug) | |
1901 | __memblock_dump_all(); | |
1902 | } | |
1903 | ||
1aadc056 | 1904 | void __init memblock_allow_resize(void) |
6ed311b2 | 1905 | { |
142b45a7 | 1906 | memblock_can_resize = 1; |
6ed311b2 BH |
1907 | } |
1908 | ||
6ed311b2 BH |
1909 | static int __init early_memblock(char *p) |
1910 | { | |
1911 | if (p && strstr(p, "debug")) | |
1912 | memblock_debug = 1; | |
1913 | return 0; | |
1914 | } | |
1915 | early_param("memblock", early_memblock); | |
1916 | ||
4f5b0c17 MR |
1917 | static void __init free_memmap(unsigned long start_pfn, unsigned long end_pfn) |
1918 | { | |
1919 | struct page *start_pg, *end_pg; | |
1920 | phys_addr_t pg, pgend; | |
1921 | ||
1922 | /* | |
1923 | * Convert start_pfn/end_pfn to a struct page pointer. | |
1924 | */ | |
1925 | start_pg = pfn_to_page(start_pfn - 1) + 1; | |
1926 | end_pg = pfn_to_page(end_pfn - 1) + 1; | |
1927 | ||
1928 | /* | |
1929 | * Convert to physical addresses, and round start upwards and end | |
1930 | * downwards. | |
1931 | */ | |
1932 | pg = PAGE_ALIGN(__pa(start_pg)); | |
1933 | pgend = __pa(end_pg) & PAGE_MASK; | |
1934 | ||
1935 | /* | |
1936 | * If there are free pages between these, free the section of the | |
1937 | * memmap array. | |
1938 | */ | |
1939 | if (pg < pgend) | |
1940 | memblock_free(pg, pgend - pg); | |
1941 | } | |
1942 | ||
1943 | /* | |
1944 | * The mem_map array can get very big. Free the unused area of the memory map. | |
1945 | */ | |
1946 | static void __init free_unused_memmap(void) | |
1947 | { | |
1948 | unsigned long start, end, prev_end = 0; | |
1949 | int i; | |
1950 | ||
1951 | if (!IS_ENABLED(CONFIG_HAVE_ARCH_PFN_VALID) || | |
1952 | IS_ENABLED(CONFIG_SPARSEMEM_VMEMMAP)) | |
1953 | return; | |
1954 | ||
1955 | /* | |
1956 | * This relies on each bank being in address order. | |
1957 | * The banks are sorted previously in bootmem_init(). | |
1958 | */ | |
1959 | for_each_mem_pfn_range(i, MAX_NUMNODES, &start, &end, NULL) { | |
1960 | #ifdef CONFIG_SPARSEMEM | |
1961 | /* | |
1962 | * Take care not to free memmap entries that don't exist | |
1963 | * due to SPARSEMEM sections which aren't present. | |
1964 | */ | |
1965 | start = min(start, ALIGN(prev_end, PAGES_PER_SECTION)); | |
f921f53e | 1966 | #endif |
4f5b0c17 | 1967 | /* |
e2a86800 MR |
1968 | * Align down here since many operations in VM subsystem |
1969 | * presume that there are no holes in the memory map inside | |
1970 | * a pageblock | |
4f5b0c17 | 1971 | */ |
e2a86800 | 1972 | start = round_down(start, pageblock_nr_pages); |
4f5b0c17 MR |
1973 | |
1974 | /* | |
1975 | * If we had a previous bank, and there is a space | |
1976 | * between the current bank and the previous, free it. | |
1977 | */ | |
1978 | if (prev_end && prev_end < start) | |
1979 | free_memmap(prev_end, start); | |
1980 | ||
1981 | /* | |
e2a86800 MR |
1982 | * Align up here since many operations in VM subsystem |
1983 | * presume that there are no holes in the memory map inside | |
1984 | * a pageblock | |
4f5b0c17 | 1985 | */ |
e2a86800 | 1986 | prev_end = ALIGN(end, pageblock_nr_pages); |
4f5b0c17 MR |
1987 | } |
1988 | ||
1989 | #ifdef CONFIG_SPARSEMEM | |
f921f53e MR |
1990 | if (!IS_ALIGNED(prev_end, PAGES_PER_SECTION)) { |
1991 | prev_end = ALIGN(end, pageblock_nr_pages); | |
4f5b0c17 | 1992 | free_memmap(prev_end, ALIGN(prev_end, PAGES_PER_SECTION)); |
f921f53e | 1993 | } |
4f5b0c17 MR |
1994 | #endif |
1995 | } | |
1996 | ||
bda49a81 MR |
1997 | static void __init __free_pages_memory(unsigned long start, unsigned long end) |
1998 | { | |
1999 | int order; | |
2000 | ||
2001 | while (start < end) { | |
2002 | order = min(MAX_ORDER - 1UL, __ffs(start)); | |
2003 | ||
2004 | while (start + (1UL << order) > end) | |
2005 | order--; | |
2006 | ||
2007 | memblock_free_pages(pfn_to_page(start), start, order); | |
2008 | ||
2009 | start += (1UL << order); | |
2010 | } | |
2011 | } | |
2012 | ||
2013 | static unsigned long __init __free_memory_core(phys_addr_t start, | |
2014 | phys_addr_t end) | |
2015 | { | |
2016 | unsigned long start_pfn = PFN_UP(start); | |
2017 | unsigned long end_pfn = min_t(unsigned long, | |
2018 | PFN_DOWN(end), max_low_pfn); | |
2019 | ||
2020 | if (start_pfn >= end_pfn) | |
2021 | return 0; | |
2022 | ||
2023 | __free_pages_memory(start_pfn, end_pfn); | |
2024 | ||
2025 | return end_pfn - start_pfn; | |
2026 | } | |
2027 | ||
9092d4f7 MR |
2028 | static void __init memmap_init_reserved_pages(void) |
2029 | { | |
2030 | struct memblock_region *region; | |
2031 | phys_addr_t start, end; | |
2032 | u64 i; | |
2033 | ||
2034 | /* initialize struct pages for the reserved regions */ | |
2035 | for_each_reserved_mem_range(i, &start, &end) | |
2036 | reserve_bootmem_region(start, end); | |
2037 | ||
2038 | /* and also treat struct pages for the NOMAP regions as PageReserved */ | |
2039 | for_each_mem_region(region) { | |
2040 | if (memblock_is_nomap(region)) { | |
2041 | start = region->base; | |
2042 | end = start + region->size; | |
2043 | reserve_bootmem_region(start, end); | |
2044 | } | |
2045 | } | |
2046 | } | |
2047 | ||
bda49a81 MR |
2048 | static unsigned long __init free_low_memory_core_early(void) |
2049 | { | |
2050 | unsigned long count = 0; | |
2051 | phys_addr_t start, end; | |
2052 | u64 i; | |
2053 | ||
2054 | memblock_clear_hotplug(0, -1); | |
2055 | ||
9092d4f7 | 2056 | memmap_init_reserved_pages(); |
bda49a81 MR |
2057 | |
2058 | /* | |
2059 | * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id | |
2060 | * because in some case like Node0 doesn't have RAM installed | |
2061 | * low ram will be on Node1 | |
2062 | */ | |
2063 | for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, | |
2064 | NULL) | |
2065 | count += __free_memory_core(start, end); | |
2066 | ||
2067 | return count; | |
2068 | } | |
2069 | ||
2070 | static int reset_managed_pages_done __initdata; | |
2071 | ||
2072 | void reset_node_managed_pages(pg_data_t *pgdat) | |
2073 | { | |
2074 | struct zone *z; | |
2075 | ||
2076 | for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) | |
9705bea5 | 2077 | atomic_long_set(&z->managed_pages, 0); |
bda49a81 MR |
2078 | } |
2079 | ||
2080 | void __init reset_all_zones_managed_pages(void) | |
2081 | { | |
2082 | struct pglist_data *pgdat; | |
2083 | ||
2084 | if (reset_managed_pages_done) | |
2085 | return; | |
2086 | ||
2087 | for_each_online_pgdat(pgdat) | |
2088 | reset_node_managed_pages(pgdat); | |
2089 | ||
2090 | reset_managed_pages_done = 1; | |
2091 | } | |
2092 | ||
2093 | /** | |
2094 | * memblock_free_all - release free pages to the buddy allocator | |
bda49a81 | 2095 | */ |
097d43d8 | 2096 | void __init memblock_free_all(void) |
bda49a81 MR |
2097 | { |
2098 | unsigned long pages; | |
2099 | ||
4f5b0c17 | 2100 | free_unused_memmap(); |
bda49a81 MR |
2101 | reset_all_zones_managed_pages(); |
2102 | ||
2103 | pages = free_low_memory_core_early(); | |
ca79b0c2 | 2104 | totalram_pages_add(pages); |
bda49a81 MR |
2105 | } |
2106 | ||
350e88ba | 2107 | #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_ARCH_KEEP_MEMBLOCK) |
6d03b885 BH |
2108 | |
2109 | static int memblock_debug_show(struct seq_file *m, void *private) | |
2110 | { | |
2111 | struct memblock_type *type = m->private; | |
2112 | struct memblock_region *reg; | |
2113 | int i; | |
5d63f81c | 2114 | phys_addr_t end; |
6d03b885 BH |
2115 | |
2116 | for (i = 0; i < type->cnt; i++) { | |
2117 | reg = &type->regions[i]; | |
5d63f81c | 2118 | end = reg->base + reg->size - 1; |
6d03b885 | 2119 | |
5d63f81c MC |
2120 | seq_printf(m, "%4d: ", i); |
2121 | seq_printf(m, "%pa..%pa\n", ®->base, &end); | |
6d03b885 BH |
2122 | } |
2123 | return 0; | |
2124 | } | |
5ad35093 | 2125 | DEFINE_SHOW_ATTRIBUTE(memblock_debug); |
6d03b885 BH |
2126 | |
2127 | static int __init memblock_init_debugfs(void) | |
2128 | { | |
2129 | struct dentry *root = debugfs_create_dir("memblock", NULL); | |
d9f7979c | 2130 | |
0825a6f9 JP |
2131 | debugfs_create_file("memory", 0444, root, |
2132 | &memblock.memory, &memblock_debug_fops); | |
2133 | debugfs_create_file("reserved", 0444, root, | |
2134 | &memblock.reserved, &memblock_debug_fops); | |
70210ed9 | 2135 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
77649905 DH |
2136 | debugfs_create_file("physmem", 0444, root, &physmem, |
2137 | &memblock_debug_fops); | |
70210ed9 | 2138 | #endif |
6d03b885 BH |
2139 | |
2140 | return 0; | |
2141 | } | |
2142 | __initcall(memblock_init_debugfs); | |
2143 | ||
2144 | #endif /* CONFIG_DEBUG_FS */ |