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 | ||
450d0e74 ZG |
32 | #ifndef INIT_MEMBLOCK_MEMORY_REGIONS |
33 | #define INIT_MEMBLOCK_MEMORY_REGIONS INIT_MEMBLOCK_REGIONS | |
34 | #endif | |
35 | ||
3e039c5c MR |
36 | /** |
37 | * DOC: memblock overview | |
38 | * | |
39 | * Memblock is a method of managing memory regions during the early | |
40 | * boot period when the usual kernel memory allocators are not up and | |
41 | * running. | |
42 | * | |
43 | * Memblock views the system memory as collections of contiguous | |
44 | * regions. There are several types of these collections: | |
45 | * | |
46 | * * ``memory`` - describes the physical memory available to the | |
47 | * kernel; this may differ from the actual physical memory installed | |
48 | * in the system, for instance when the memory is restricted with | |
49 | * ``mem=`` command line parameter | |
50 | * * ``reserved`` - describes the regions that were allocated | |
77649905 DH |
51 | * * ``physmem`` - describes the actual physical memory available during |
52 | * boot regardless of the possible restrictions and memory hot(un)plug; | |
53 | * the ``physmem`` type is only available on some architectures. | |
3e039c5c | 54 | * |
9303c9d5 | 55 | * Each region is represented by struct memblock_region that |
3e039c5c | 56 | * defines the region extents, its attributes and NUMA node id on NUMA |
1bf162e4 MCC |
57 | * systems. Every memory type is described by the struct memblock_type |
58 | * which contains an array of memory regions along with | |
77649905 | 59 | * the allocator metadata. The "memory" and "reserved" types are nicely |
9303c9d5 | 60 | * wrapped with struct memblock. This structure is statically |
77649905 | 61 | * initialized at build time. The region arrays are initially sized to |
450d0e74 ZG |
62 | * %INIT_MEMBLOCK_MEMORY_REGIONS for "memory" and |
63 | * %INIT_MEMBLOCK_RESERVED_REGIONS for "reserved". The region array | |
64 | * for "physmem" is initially sized to %INIT_PHYSMEM_REGIONS. | |
6e5af9a8 C |
65 | * The memblock_allow_resize() enables automatic resizing of the region |
66 | * arrays during addition of new regions. This feature should be used | |
67 | * with care so that memory allocated for the region array will not | |
68 | * overlap with areas that should be reserved, for example initrd. | |
3e039c5c MR |
69 | * |
70 | * The early architecture setup should tell memblock what the physical | |
6e5af9a8 C |
71 | * memory layout is by using memblock_add() or memblock_add_node() |
72 | * functions. The first function does not assign the region to a NUMA | |
73 | * node and it is appropriate for UMA systems. Yet, it is possible to | |
74 | * use it on NUMA systems as well and assign the region to a NUMA node | |
75 | * later in the setup process using memblock_set_node(). The | |
76 | * memblock_add_node() performs such an assignment directly. | |
3e039c5c | 77 | * |
a2974133 MR |
78 | * Once memblock is setup the memory can be allocated using one of the |
79 | * API variants: | |
80 | * | |
6e5af9a8 C |
81 | * * memblock_phys_alloc*() - these functions return the **physical** |
82 | * address of the allocated memory | |
83 | * * memblock_alloc*() - these functions return the **virtual** address | |
84 | * of the allocated memory. | |
a2974133 | 85 | * |
df1758d9 | 86 | * Note, that both API variants use implicit assumptions about allowed |
a2974133 | 87 | * memory ranges and the fallback methods. Consult the documentation |
6e5af9a8 C |
88 | * of memblock_alloc_internal() and memblock_alloc_range_nid() |
89 | * functions for more elaborate description. | |
3e039c5c | 90 | * |
6e5af9a8 C |
91 | * As the system boot progresses, the architecture specific mem_init() |
92 | * function frees all the memory to the buddy page allocator. | |
3e039c5c | 93 | * |
6e5af9a8 | 94 | * Unless an architecture enables %CONFIG_ARCH_KEEP_MEMBLOCK, the |
77649905 DH |
95 | * memblock data structures (except "physmem") will be discarded after the |
96 | * system initialization completes. | |
3e039c5c MR |
97 | */ |
98 | ||
a9ee6cf5 | 99 | #ifndef CONFIG_NUMA |
bda49a81 MR |
100 | struct pglist_data __refdata contig_page_data; |
101 | EXPORT_SYMBOL(contig_page_data); | |
102 | #endif | |
103 | ||
104 | unsigned long max_low_pfn; | |
105 | unsigned long min_low_pfn; | |
106 | unsigned long max_pfn; | |
107 | unsigned long long max_possible_pfn; | |
108 | ||
450d0e74 | 109 | static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_MEMORY_REGIONS] __initdata_memblock; |
8a5b403d | 110 | static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_RESERVED_REGIONS] __initdata_memblock; |
70210ed9 | 111 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
77649905 | 112 | static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS]; |
70210ed9 | 113 | #endif |
fe091c20 TH |
114 | |
115 | struct memblock memblock __initdata_memblock = { | |
116 | .memory.regions = memblock_memory_init_regions, | |
117 | .memory.cnt = 1, /* empty dummy entry */ | |
450d0e74 | 118 | .memory.max = INIT_MEMBLOCK_MEMORY_REGIONS, |
0262d9c8 | 119 | .memory.name = "memory", |
fe091c20 TH |
120 | |
121 | .reserved.regions = memblock_reserved_init_regions, | |
122 | .reserved.cnt = 1, /* empty dummy entry */ | |
8a5b403d | 123 | .reserved.max = INIT_MEMBLOCK_RESERVED_REGIONS, |
0262d9c8 | 124 | .reserved.name = "reserved", |
fe091c20 | 125 | |
79442ed1 | 126 | .bottom_up = false, |
fe091c20 TH |
127 | .current_limit = MEMBLOCK_ALLOC_ANYWHERE, |
128 | }; | |
95f72d1e | 129 | |
77649905 DH |
130 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
131 | struct memblock_type physmem = { | |
132 | .regions = memblock_physmem_init_regions, | |
133 | .cnt = 1, /* empty dummy entry */ | |
134 | .max = INIT_PHYSMEM_REGIONS, | |
135 | .name = "physmem", | |
136 | }; | |
137 | #endif | |
138 | ||
9f3d5eaa MR |
139 | /* |
140 | * keep a pointer to &memblock.memory in the text section to use it in | |
141 | * __next_mem_range() and its helpers. | |
142 | * For architectures that do not keep memblock data after init, this | |
143 | * pointer will be reset to NULL at memblock_discard() | |
144 | */ | |
145 | static __refdata struct memblock_type *memblock_memory = &memblock.memory; | |
146 | ||
cd991db8 MR |
147 | #define for_each_memblock_type(i, memblock_type, rgn) \ |
148 | for (i = 0, rgn = &memblock_type->regions[0]; \ | |
149 | i < memblock_type->cnt; \ | |
150 | i++, rgn = &memblock_type->regions[i]) | |
151 | ||
87c55870 MR |
152 | #define memblock_dbg(fmt, ...) \ |
153 | do { \ | |
154 | if (memblock_debug) \ | |
155 | pr_info(fmt, ##__VA_ARGS__); \ | |
156 | } while (0) | |
157 | ||
158 | static int memblock_debug __initdata_memblock; | |
a3f5bafc | 159 | static bool system_has_some_mirror __initdata_memblock = false; |
1aadc056 | 160 | static int memblock_can_resize __initdata_memblock; |
181eb394 GS |
161 | static int memblock_memory_in_slab __initdata_memblock = 0; |
162 | static int memblock_reserved_in_slab __initdata_memblock = 0; | |
95f72d1e | 163 | |
c366ea89 | 164 | static enum memblock_flags __init_memblock choose_memblock_flags(void) |
a3f5bafc TL |
165 | { |
166 | return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE; | |
167 | } | |
168 | ||
eb18f1b5 TH |
169 | /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */ |
170 | static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size) | |
171 | { | |
1c4bc43d | 172 | return *size = min(*size, PHYS_ADDR_MAX - base); |
eb18f1b5 TH |
173 | } |
174 | ||
6ed311b2 BH |
175 | /* |
176 | * Address comparison utilities | |
177 | */ | |
10d06439 | 178 | static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1, |
2898cc4c | 179 | phys_addr_t base2, phys_addr_t size2) |
95f72d1e YL |
180 | { |
181 | return ((base1 < (base2 + size2)) && (base2 < (base1 + size1))); | |
182 | } | |
183 | ||
95cf82ec | 184 | bool __init_memblock memblock_overlaps_region(struct memblock_type *type, |
2d7d3eb2 | 185 | phys_addr_t base, phys_addr_t size) |
6ed311b2 BH |
186 | { |
187 | unsigned long i; | |
188 | ||
023accf5 MR |
189 | memblock_cap_size(base, &size); |
190 | ||
f14516fb AK |
191 | for (i = 0; i < type->cnt; i++) |
192 | if (memblock_addrs_overlap(base, size, type->regions[i].base, | |
193 | type->regions[i].size)) | |
6ed311b2 | 194 | break; |
c5c5c9d1 | 195 | return i < type->cnt; |
6ed311b2 BH |
196 | } |
197 | ||
47cec443 | 198 | /** |
79442ed1 TC |
199 | * __memblock_find_range_bottom_up - find free area utility in bottom-up |
200 | * @start: start of candidate range | |
47cec443 MR |
201 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
202 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
79442ed1 TC |
203 | * @size: size of free area to find |
204 | * @align: alignment of free area to find | |
b1154233 | 205 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 206 | * @flags: pick from blocks based on memory attributes |
79442ed1 TC |
207 | * |
208 | * Utility called from memblock_find_in_range_node(), find free area bottom-up. | |
209 | * | |
47cec443 | 210 | * Return: |
79442ed1 TC |
211 | * Found address on success, 0 on failure. |
212 | */ | |
213 | static phys_addr_t __init_memblock | |
214 | __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 215 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 216 | enum memblock_flags flags) |
79442ed1 TC |
217 | { |
218 | phys_addr_t this_start, this_end, cand; | |
219 | u64 i; | |
220 | ||
fc6daaf9 | 221 | for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { |
79442ed1 TC |
222 | this_start = clamp(this_start, start, end); |
223 | this_end = clamp(this_end, start, end); | |
224 | ||
225 | cand = round_up(this_start, align); | |
226 | if (cand < this_end && this_end - cand >= size) | |
227 | return cand; | |
228 | } | |
229 | ||
230 | return 0; | |
231 | } | |
232 | ||
7bd0b0f0 | 233 | /** |
1402899e | 234 | * __memblock_find_range_top_down - find free area utility, in top-down |
7bd0b0f0 | 235 | * @start: start of candidate range |
47cec443 MR |
236 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
237 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
238 | * @size: size of free area to find |
239 | * @align: alignment of free area to find | |
b1154233 | 240 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 241 | * @flags: pick from blocks based on memory attributes |
7bd0b0f0 | 242 | * |
1402899e | 243 | * Utility called from memblock_find_in_range_node(), find free area top-down. |
7bd0b0f0 | 244 | * |
47cec443 | 245 | * Return: |
79442ed1 | 246 | * Found address on success, 0 on failure. |
6ed311b2 | 247 | */ |
1402899e TC |
248 | static phys_addr_t __init_memblock |
249 | __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 250 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 251 | enum memblock_flags flags) |
f7210e6c TC |
252 | { |
253 | phys_addr_t this_start, this_end, cand; | |
254 | u64 i; | |
255 | ||
fc6daaf9 TL |
256 | for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, |
257 | NULL) { | |
f7210e6c TC |
258 | this_start = clamp(this_start, start, end); |
259 | this_end = clamp(this_end, start, end); | |
260 | ||
261 | if (this_end < size) | |
262 | continue; | |
263 | ||
264 | cand = round_down(this_end - size, align); | |
265 | if (cand >= this_start) | |
266 | return cand; | |
267 | } | |
1402899e | 268 | |
f7210e6c TC |
269 | return 0; |
270 | } | |
6ed311b2 | 271 | |
1402899e TC |
272 | /** |
273 | * memblock_find_in_range_node - find free area in given range and node | |
1402899e TC |
274 | * @size: size of free area to find |
275 | * @align: alignment of free area to find | |
87029ee9 | 276 | * @start: start of candidate range |
47cec443 MR |
277 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
278 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
b1154233 | 279 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 280 | * @flags: pick from blocks based on memory attributes |
1402899e TC |
281 | * |
282 | * Find @size free area aligned to @align in the specified range and node. | |
283 | * | |
47cec443 | 284 | * Return: |
79442ed1 | 285 | * Found address on success, 0 on failure. |
1402899e | 286 | */ |
c366ea89 | 287 | static phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, |
87029ee9 | 288 | phys_addr_t align, phys_addr_t start, |
e1720fee MR |
289 | phys_addr_t end, int nid, |
290 | enum memblock_flags flags) | |
1402899e TC |
291 | { |
292 | /* pump up @end */ | |
fed84c78 | 293 | if (end == MEMBLOCK_ALLOC_ACCESSIBLE || |
c6975d7c | 294 | end == MEMBLOCK_ALLOC_NOLEAKTRACE) |
1402899e TC |
295 | end = memblock.current_limit; |
296 | ||
297 | /* avoid allocating the first page */ | |
298 | start = max_t(phys_addr_t, start, PAGE_SIZE); | |
299 | end = max(start, end); | |
300 | ||
2dcb3964 RG |
301 | if (memblock_bottom_up()) |
302 | return __memblock_find_range_bottom_up(start, end, size, align, | |
303 | nid, flags); | |
304 | else | |
305 | return __memblock_find_range_top_down(start, end, size, align, | |
306 | nid, flags); | |
1402899e TC |
307 | } |
308 | ||
7bd0b0f0 TH |
309 | /** |
310 | * memblock_find_in_range - find free area in given range | |
311 | * @start: start of candidate range | |
47cec443 MR |
312 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
313 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
314 | * @size: size of free area to find |
315 | * @align: alignment of free area to find | |
316 | * | |
317 | * Find @size free area aligned to @align in the specified range. | |
318 | * | |
47cec443 | 319 | * Return: |
79442ed1 | 320 | * Found address on success, 0 on failure. |
fc769a8e | 321 | */ |
a7259df7 | 322 | static phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, |
7bd0b0f0 TH |
323 | phys_addr_t end, phys_addr_t size, |
324 | phys_addr_t align) | |
6ed311b2 | 325 | { |
a3f5bafc | 326 | phys_addr_t ret; |
e1720fee | 327 | enum memblock_flags flags = choose_memblock_flags(); |
a3f5bafc TL |
328 | |
329 | again: | |
330 | ret = memblock_find_in_range_node(size, align, start, end, | |
331 | NUMA_NO_NODE, flags); | |
332 | ||
333 | if (!ret && (flags & MEMBLOCK_MIRROR)) { | |
14d9a675 | 334 | pr_warn_ratelimited("Could not allocate %pap bytes of mirrored memory\n", |
a3f5bafc TL |
335 | &size); |
336 | flags &= ~MEMBLOCK_MIRROR; | |
337 | goto again; | |
338 | } | |
339 | ||
340 | return ret; | |
6ed311b2 BH |
341 | } |
342 | ||
10d06439 | 343 | static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) |
95f72d1e | 344 | { |
1440c4e2 | 345 | type->total_size -= type->regions[r].size; |
7c0caeb8 TH |
346 | memmove(&type->regions[r], &type->regions[r + 1], |
347 | (type->cnt - (r + 1)) * sizeof(type->regions[r])); | |
e3239ff9 | 348 | type->cnt--; |
95f72d1e | 349 | |
8f7a6605 BH |
350 | /* Special case for empty arrays */ |
351 | if (type->cnt == 0) { | |
1440c4e2 | 352 | WARN_ON(type->total_size != 0); |
8f7a6605 BH |
353 | type->cnt = 1; |
354 | type->regions[0].base = 0; | |
355 | type->regions[0].size = 0; | |
66a20757 | 356 | type->regions[0].flags = 0; |
7c0caeb8 | 357 | memblock_set_region_node(&type->regions[0], MAX_NUMNODES); |
8f7a6605 | 358 | } |
95f72d1e YL |
359 | } |
360 | ||
350e88ba | 361 | #ifndef CONFIG_ARCH_KEEP_MEMBLOCK |
3010f876 | 362 | /** |
47cec443 | 363 | * memblock_discard - discard memory and reserved arrays if they were allocated |
3010f876 PT |
364 | */ |
365 | void __init memblock_discard(void) | |
5e270e25 | 366 | { |
3010f876 | 367 | phys_addr_t addr, size; |
5e270e25 | 368 | |
3010f876 PT |
369 | if (memblock.reserved.regions != memblock_reserved_init_regions) { |
370 | addr = __pa(memblock.reserved.regions); | |
371 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
372 | memblock.reserved.max); | |
c94afc46 ML |
373 | if (memblock_reserved_in_slab) |
374 | kfree(memblock.reserved.regions); | |
375 | else | |
376 | memblock_free_late(addr, size); | |
3010f876 | 377 | } |
5e270e25 | 378 | |
91b540f9 | 379 | if (memblock.memory.regions != memblock_memory_init_regions) { |
3010f876 PT |
380 | addr = __pa(memblock.memory.regions); |
381 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
382 | memblock.memory.max); | |
c94afc46 ML |
383 | if (memblock_memory_in_slab) |
384 | kfree(memblock.memory.regions); | |
385 | else | |
386 | memblock_free_late(addr, size); | |
3010f876 | 387 | } |
9f3d5eaa MR |
388 | |
389 | memblock_memory = NULL; | |
5e270e25 | 390 | } |
5e270e25 PH |
391 | #endif |
392 | ||
48c3b583 GP |
393 | /** |
394 | * memblock_double_array - double the size of the memblock regions array | |
395 | * @type: memblock type of the regions array being doubled | |
396 | * @new_area_start: starting address of memory range to avoid overlap with | |
397 | * @new_area_size: size of memory range to avoid overlap with | |
398 | * | |
399 | * Double the size of the @type regions array. If memblock is being used to | |
400 | * allocate memory for a new reserved regions array and there is a previously | |
47cec443 | 401 | * allocated memory range [@new_area_start, @new_area_start + @new_area_size] |
48c3b583 GP |
402 | * waiting to be reserved, ensure the memory used by the new array does |
403 | * not overlap. | |
404 | * | |
47cec443 | 405 | * Return: |
48c3b583 GP |
406 | * 0 on success, -1 on failure. |
407 | */ | |
408 | static int __init_memblock memblock_double_array(struct memblock_type *type, | |
409 | phys_addr_t new_area_start, | |
410 | phys_addr_t new_area_size) | |
142b45a7 BH |
411 | { |
412 | struct memblock_region *new_array, *old_array; | |
29f67386 | 413 | phys_addr_t old_alloc_size, new_alloc_size; |
a36aab89 | 414 | phys_addr_t old_size, new_size, addr, new_end; |
142b45a7 | 415 | int use_slab = slab_is_available(); |
181eb394 | 416 | int *in_slab; |
142b45a7 BH |
417 | |
418 | /* We don't allow resizing until we know about the reserved regions | |
419 | * of memory that aren't suitable for allocation | |
420 | */ | |
421 | if (!memblock_can_resize) | |
422 | return -1; | |
423 | ||
142b45a7 BH |
424 | /* Calculate new doubled size */ |
425 | old_size = type->max * sizeof(struct memblock_region); | |
426 | new_size = old_size << 1; | |
29f67386 YL |
427 | /* |
428 | * We need to allocated new one align to PAGE_SIZE, | |
429 | * so we can free them completely later. | |
430 | */ | |
431 | old_alloc_size = PAGE_ALIGN(old_size); | |
432 | new_alloc_size = PAGE_ALIGN(new_size); | |
142b45a7 | 433 | |
181eb394 GS |
434 | /* Retrieve the slab flag */ |
435 | if (type == &memblock.memory) | |
436 | in_slab = &memblock_memory_in_slab; | |
437 | else | |
438 | in_slab = &memblock_reserved_in_slab; | |
439 | ||
a2974133 | 440 | /* Try to find some space for it */ |
142b45a7 BH |
441 | if (use_slab) { |
442 | new_array = kmalloc(new_size, GFP_KERNEL); | |
1f5026a7 | 443 | addr = new_array ? __pa(new_array) : 0; |
4e2f0775 | 444 | } else { |
48c3b583 GP |
445 | /* only exclude range when trying to double reserved.regions */ |
446 | if (type != &memblock.reserved) | |
447 | new_area_start = new_area_size = 0; | |
448 | ||
449 | addr = memblock_find_in_range(new_area_start + new_area_size, | |
450 | memblock.current_limit, | |
29f67386 | 451 | new_alloc_size, PAGE_SIZE); |
48c3b583 GP |
452 | if (!addr && new_area_size) |
453 | addr = memblock_find_in_range(0, | |
fd07383b AM |
454 | min(new_area_start, memblock.current_limit), |
455 | new_alloc_size, PAGE_SIZE); | |
48c3b583 | 456 | |
15674868 | 457 | new_array = addr ? __va(addr) : NULL; |
4e2f0775 | 458 | } |
1f5026a7 | 459 | if (!addr) { |
142b45a7 | 460 | pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", |
0262d9c8 | 461 | type->name, type->max, type->max * 2); |
142b45a7 BH |
462 | return -1; |
463 | } | |
142b45a7 | 464 | |
a36aab89 MR |
465 | new_end = addr + new_size - 1; |
466 | memblock_dbg("memblock: %s is doubled to %ld at [%pa-%pa]", | |
467 | type->name, type->max * 2, &addr, &new_end); | |
ea9e4376 | 468 | |
fd07383b AM |
469 | /* |
470 | * Found space, we now need to move the array over before we add the | |
471 | * reserved region since it may be our reserved array itself that is | |
472 | * full. | |
142b45a7 BH |
473 | */ |
474 | memcpy(new_array, type->regions, old_size); | |
475 | memset(new_array + type->max, 0, old_size); | |
476 | old_array = type->regions; | |
477 | type->regions = new_array; | |
478 | type->max <<= 1; | |
479 | ||
fd07383b | 480 | /* Free old array. We needn't free it if the array is the static one */ |
181eb394 GS |
481 | if (*in_slab) |
482 | kfree(old_array); | |
483 | else if (old_array != memblock_memory_init_regions && | |
484 | old_array != memblock_reserved_init_regions) | |
4421cca0 | 485 | memblock_free(old_array, old_alloc_size); |
142b45a7 | 486 | |
fd07383b AM |
487 | /* |
488 | * Reserve the new array if that comes from the memblock. Otherwise, we | |
489 | * needn't do it | |
181eb394 GS |
490 | */ |
491 | if (!use_slab) | |
29f67386 | 492 | BUG_ON(memblock_reserve(addr, new_alloc_size)); |
181eb394 GS |
493 | |
494 | /* Update slab flag */ | |
495 | *in_slab = use_slab; | |
496 | ||
142b45a7 BH |
497 | return 0; |
498 | } | |
499 | ||
784656f9 TH |
500 | /** |
501 | * memblock_merge_regions - merge neighboring compatible regions | |
502 | * @type: memblock type to scan | |
2fe03412 PZ |
503 | * @start_rgn: start scanning from (@start_rgn - 1) |
504 | * @end_rgn: end scanning at (@end_rgn - 1) | |
505 | * Scan @type and merge neighboring compatible regions in [@start_rgn - 1, @end_rgn) | |
784656f9 | 506 | */ |
2fe03412 PZ |
507 | static void __init_memblock memblock_merge_regions(struct memblock_type *type, |
508 | unsigned long start_rgn, | |
509 | unsigned long end_rgn) | |
95f72d1e | 510 | { |
784656f9 | 511 | int i = 0; |
2fe03412 PZ |
512 | if (start_rgn) |
513 | i = start_rgn - 1; | |
514 | end_rgn = min(end_rgn, type->cnt - 1); | |
515 | while (i < end_rgn) { | |
784656f9 TH |
516 | struct memblock_region *this = &type->regions[i]; |
517 | struct memblock_region *next = &type->regions[i + 1]; | |
95f72d1e | 518 | |
7c0caeb8 TH |
519 | if (this->base + this->size != next->base || |
520 | memblock_get_region_node(this) != | |
66a20757 TC |
521 | memblock_get_region_node(next) || |
522 | this->flags != next->flags) { | |
784656f9 TH |
523 | BUG_ON(this->base + this->size > next->base); |
524 | i++; | |
525 | continue; | |
8f7a6605 BH |
526 | } |
527 | ||
784656f9 | 528 | this->size += next->size; |
c0232ae8 LF |
529 | /* move forward from next + 1, index of which is i + 2 */ |
530 | memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); | |
784656f9 | 531 | type->cnt--; |
2fe03412 | 532 | end_rgn--; |
95f72d1e | 533 | } |
784656f9 | 534 | } |
95f72d1e | 535 | |
784656f9 TH |
536 | /** |
537 | * memblock_insert_region - insert new memblock region | |
209ff86d TC |
538 | * @type: memblock type to insert into |
539 | * @idx: index for the insertion point | |
540 | * @base: base address of the new region | |
541 | * @size: size of the new region | |
542 | * @nid: node id of the new region | |
66a20757 | 543 | * @flags: flags of the new region |
784656f9 | 544 | * |
47cec443 | 545 | * Insert new memblock region [@base, @base + @size) into @type at @idx. |
412d0008 | 546 | * @type must already have extra room to accommodate the new region. |
784656f9 TH |
547 | */ |
548 | static void __init_memblock memblock_insert_region(struct memblock_type *type, | |
549 | int idx, phys_addr_t base, | |
66a20757 | 550 | phys_addr_t size, |
e1720fee MR |
551 | int nid, |
552 | enum memblock_flags flags) | |
784656f9 TH |
553 | { |
554 | struct memblock_region *rgn = &type->regions[idx]; | |
555 | ||
556 | BUG_ON(type->cnt >= type->max); | |
557 | memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); | |
558 | rgn->base = base; | |
559 | rgn->size = size; | |
66a20757 | 560 | rgn->flags = flags; |
7c0caeb8 | 561 | memblock_set_region_node(rgn, nid); |
784656f9 | 562 | type->cnt++; |
1440c4e2 | 563 | type->total_size += size; |
784656f9 TH |
564 | } |
565 | ||
566 | /** | |
f1af9d3a | 567 | * memblock_add_range - add new memblock region |
784656f9 TH |
568 | * @type: memblock type to add new region into |
569 | * @base: base address of the new region | |
570 | * @size: size of the new region | |
7fb0bc3f | 571 | * @nid: nid of the new region |
66a20757 | 572 | * @flags: flags of the new region |
784656f9 | 573 | * |
47cec443 | 574 | * Add new memblock region [@base, @base + @size) into @type. The new region |
784656f9 TH |
575 | * is allowed to overlap with existing ones - overlaps don't affect already |
576 | * existing regions. @type is guaranteed to be minimal (all neighbouring | |
577 | * compatible regions are merged) after the addition. | |
578 | * | |
47cec443 | 579 | * Return: |
784656f9 TH |
580 | * 0 on success, -errno on failure. |
581 | */ | |
02634a44 | 582 | static int __init_memblock memblock_add_range(struct memblock_type *type, |
66a20757 | 583 | phys_addr_t base, phys_addr_t size, |
e1720fee | 584 | int nid, enum memblock_flags flags) |
784656f9 TH |
585 | { |
586 | bool insert = false; | |
eb18f1b5 TH |
587 | phys_addr_t obase = base; |
588 | phys_addr_t end = base + memblock_cap_size(base, &size); | |
2fe03412 | 589 | int idx, nr_new, start_rgn = -1, end_rgn; |
8c9c1701 | 590 | struct memblock_region *rgn; |
784656f9 | 591 | |
b3dc627c TH |
592 | if (!size) |
593 | return 0; | |
594 | ||
784656f9 TH |
595 | /* special case for empty array */ |
596 | if (type->regions[0].size == 0) { | |
1440c4e2 | 597 | WARN_ON(type->cnt != 1 || type->total_size); |
8f7a6605 BH |
598 | type->regions[0].base = base; |
599 | type->regions[0].size = size; | |
66a20757 | 600 | type->regions[0].flags = flags; |
7fb0bc3f | 601 | memblock_set_region_node(&type->regions[0], nid); |
1440c4e2 | 602 | type->total_size = size; |
8f7a6605 | 603 | return 0; |
95f72d1e | 604 | } |
28e1a8f4 JT |
605 | |
606 | /* | |
607 | * The worst case is when new range overlaps all existing regions, | |
608 | * then we'll need type->cnt + 1 empty regions in @type. So if | |
ad500fb2 | 609 | * type->cnt * 2 + 1 is less than or equal to type->max, we know |
28e1a8f4 JT |
610 | * that there is enough empty regions in @type, and we can insert |
611 | * regions directly. | |
612 | */ | |
ad500fb2 | 613 | if (type->cnt * 2 + 1 <= type->max) |
28e1a8f4 JT |
614 | insert = true; |
615 | ||
784656f9 TH |
616 | repeat: |
617 | /* | |
618 | * The following is executed twice. Once with %false @insert and | |
619 | * then with %true. The first counts the number of regions needed | |
412d0008 | 620 | * to accommodate the new area. The second actually inserts them. |
142b45a7 | 621 | */ |
784656f9 TH |
622 | base = obase; |
623 | nr_new = 0; | |
95f72d1e | 624 | |
66e8b438 | 625 | for_each_memblock_type(idx, type, rgn) { |
784656f9 TH |
626 | phys_addr_t rbase = rgn->base; |
627 | phys_addr_t rend = rbase + rgn->size; | |
628 | ||
629 | if (rbase >= end) | |
95f72d1e | 630 | break; |
784656f9 TH |
631 | if (rend <= base) |
632 | continue; | |
633 | /* | |
634 | * @rgn overlaps. If it separates the lower part of new | |
635 | * area, insert that portion. | |
636 | */ | |
637 | if (rbase > base) { | |
a9ee6cf5 | 638 | #ifdef CONFIG_NUMA |
c0a29498 WY |
639 | WARN_ON(nid != memblock_get_region_node(rgn)); |
640 | #endif | |
4fcab5f4 | 641 | WARN_ON(flags != rgn->flags); |
784656f9 | 642 | nr_new++; |
2fe03412 PZ |
643 | if (insert) { |
644 | if (start_rgn == -1) | |
645 | start_rgn = idx; | |
646 | end_rgn = idx + 1; | |
8c9c1701 | 647 | memblock_insert_region(type, idx++, base, |
66a20757 TC |
648 | rbase - base, nid, |
649 | flags); | |
2fe03412 | 650 | } |
95f72d1e | 651 | } |
784656f9 TH |
652 | /* area below @rend is dealt with, forget about it */ |
653 | base = min(rend, end); | |
95f72d1e | 654 | } |
784656f9 TH |
655 | |
656 | /* insert the remaining portion */ | |
657 | if (base < end) { | |
658 | nr_new++; | |
2fe03412 PZ |
659 | if (insert) { |
660 | if (start_rgn == -1) | |
661 | start_rgn = idx; | |
662 | end_rgn = idx + 1; | |
8c9c1701 | 663 | memblock_insert_region(type, idx, base, end - base, |
66a20757 | 664 | nid, flags); |
2fe03412 | 665 | } |
95f72d1e | 666 | } |
95f72d1e | 667 | |
ef3cc4db | 668 | if (!nr_new) |
669 | return 0; | |
670 | ||
784656f9 TH |
671 | /* |
672 | * If this was the first round, resize array and repeat for actual | |
673 | * insertions; otherwise, merge and return. | |
142b45a7 | 674 | */ |
784656f9 TH |
675 | if (!insert) { |
676 | while (type->cnt + nr_new > type->max) | |
48c3b583 | 677 | if (memblock_double_array(type, obase, size) < 0) |
784656f9 TH |
678 | return -ENOMEM; |
679 | insert = true; | |
680 | goto repeat; | |
681 | } else { | |
2fe03412 | 682 | memblock_merge_regions(type, start_rgn, end_rgn); |
784656f9 | 683 | return 0; |
142b45a7 | 684 | } |
95f72d1e YL |
685 | } |
686 | ||
48a833cc MR |
687 | /** |
688 | * memblock_add_node - add new memblock region within a NUMA node | |
689 | * @base: base address of the new region | |
690 | * @size: size of the new region | |
691 | * @nid: nid of the new region | |
952eea9b | 692 | * @flags: flags of the new region |
48a833cc MR |
693 | * |
694 | * Add new memblock region [@base, @base + @size) to the "memory" | |
695 | * type. See memblock_add_range() description for mode details | |
696 | * | |
697 | * Return: | |
698 | * 0 on success, -errno on failure. | |
699 | */ | |
7fb0bc3f | 700 | int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, |
952eea9b | 701 | int nid, enum memblock_flags flags) |
7fb0bc3f | 702 | { |
00974b9a GU |
703 | phys_addr_t end = base + size - 1; |
704 | ||
952eea9b DH |
705 | memblock_dbg("%s: [%pa-%pa] nid=%d flags=%x %pS\n", __func__, |
706 | &base, &end, nid, flags, (void *)_RET_IP_); | |
00974b9a | 707 | |
952eea9b | 708 | return memblock_add_range(&memblock.memory, base, size, nid, flags); |
7fb0bc3f TH |
709 | } |
710 | ||
48a833cc MR |
711 | /** |
712 | * memblock_add - add new memblock region | |
713 | * @base: base address of the new region | |
714 | * @size: size of the new region | |
715 | * | |
716 | * Add new memblock region [@base, @base + @size) to the "memory" | |
717 | * type. See memblock_add_range() description for mode details | |
718 | * | |
719 | * Return: | |
720 | * 0 on success, -errno on failure. | |
721 | */ | |
f705ac4b | 722 | int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) |
6a4055bc | 723 | { |
5d63f81c MC |
724 | phys_addr_t end = base + size - 1; |
725 | ||
a090d711 | 726 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, |
5d63f81c | 727 | &base, &end, (void *)_RET_IP_); |
6a4055bc | 728 | |
f705ac4b | 729 | return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
730 | } |
731 | ||
6a9ceb31 TH |
732 | /** |
733 | * memblock_isolate_range - isolate given range into disjoint memblocks | |
734 | * @type: memblock type to isolate range for | |
735 | * @base: base of range to isolate | |
736 | * @size: size of range to isolate | |
737 | * @start_rgn: out parameter for the start of isolated region | |
738 | * @end_rgn: out parameter for the end of isolated region | |
739 | * | |
740 | * Walk @type and ensure that regions don't cross the boundaries defined by | |
47cec443 | 741 | * [@base, @base + @size). Crossing regions are split at the boundaries, |
6a9ceb31 TH |
742 | * which may create at most two more regions. The index of the first |
743 | * region inside the range is returned in *@start_rgn and end in *@end_rgn. | |
744 | * | |
47cec443 | 745 | * Return: |
6a9ceb31 TH |
746 | * 0 on success, -errno on failure. |
747 | */ | |
748 | static int __init_memblock memblock_isolate_range(struct memblock_type *type, | |
749 | phys_addr_t base, phys_addr_t size, | |
750 | int *start_rgn, int *end_rgn) | |
751 | { | |
eb18f1b5 | 752 | phys_addr_t end = base + memblock_cap_size(base, &size); |
8c9c1701 AK |
753 | int idx; |
754 | struct memblock_region *rgn; | |
6a9ceb31 TH |
755 | |
756 | *start_rgn = *end_rgn = 0; | |
757 | ||
b3dc627c TH |
758 | if (!size) |
759 | return 0; | |
760 | ||
6a9ceb31 TH |
761 | /* we'll create at most two more regions */ |
762 | while (type->cnt + 2 > type->max) | |
48c3b583 | 763 | if (memblock_double_array(type, base, size) < 0) |
6a9ceb31 TH |
764 | return -ENOMEM; |
765 | ||
66e8b438 | 766 | for_each_memblock_type(idx, type, rgn) { |
6a9ceb31 TH |
767 | phys_addr_t rbase = rgn->base; |
768 | phys_addr_t rend = rbase + rgn->size; | |
769 | ||
770 | if (rbase >= end) | |
771 | break; | |
772 | if (rend <= base) | |
773 | continue; | |
774 | ||
775 | if (rbase < base) { | |
776 | /* | |
777 | * @rgn intersects from below. Split and continue | |
778 | * to process the next region - the new top half. | |
779 | */ | |
780 | rgn->base = base; | |
1440c4e2 TH |
781 | rgn->size -= base - rbase; |
782 | type->total_size -= base - rbase; | |
8c9c1701 | 783 | memblock_insert_region(type, idx, rbase, base - rbase, |
66a20757 TC |
784 | memblock_get_region_node(rgn), |
785 | rgn->flags); | |
6a9ceb31 TH |
786 | } else if (rend > end) { |
787 | /* | |
788 | * @rgn intersects from above. Split and redo the | |
789 | * current region - the new bottom half. | |
790 | */ | |
791 | rgn->base = end; | |
1440c4e2 TH |
792 | rgn->size -= end - rbase; |
793 | type->total_size -= end - rbase; | |
8c9c1701 | 794 | memblock_insert_region(type, idx--, rbase, end - rbase, |
66a20757 TC |
795 | memblock_get_region_node(rgn), |
796 | rgn->flags); | |
6a9ceb31 TH |
797 | } else { |
798 | /* @rgn is fully contained, record it */ | |
799 | if (!*end_rgn) | |
8c9c1701 AK |
800 | *start_rgn = idx; |
801 | *end_rgn = idx + 1; | |
6a9ceb31 TH |
802 | } |
803 | } | |
804 | ||
805 | return 0; | |
806 | } | |
6a9ceb31 | 807 | |
35bd16a2 | 808 | static int __init_memblock memblock_remove_range(struct memblock_type *type, |
f1af9d3a | 809 | phys_addr_t base, phys_addr_t size) |
95f72d1e | 810 | { |
71936180 TH |
811 | int start_rgn, end_rgn; |
812 | int i, ret; | |
95f72d1e | 813 | |
71936180 TH |
814 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
815 | if (ret) | |
816 | return ret; | |
95f72d1e | 817 | |
71936180 TH |
818 | for (i = end_rgn - 1; i >= start_rgn; i--) |
819 | memblock_remove_region(type, i); | |
8f7a6605 | 820 | return 0; |
95f72d1e YL |
821 | } |
822 | ||
581adcbe | 823 | int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) |
95f72d1e | 824 | { |
25cf23d7 MK |
825 | phys_addr_t end = base + size - 1; |
826 | ||
a090d711 | 827 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, |
25cf23d7 MK |
828 | &base, &end, (void *)_RET_IP_); |
829 | ||
f1af9d3a | 830 | return memblock_remove_range(&memblock.memory, base, size); |
95f72d1e YL |
831 | } |
832 | ||
77e02cf5 | 833 | /** |
4421cca0 | 834 | * memblock_free - free boot memory allocation |
77e02cf5 LT |
835 | * @ptr: starting address of the boot memory allocation |
836 | * @size: size of the boot memory block in bytes | |
837 | * | |
838 | * Free boot memory block previously allocated by memblock_alloc_xx() API. | |
839 | * The freeing memory will not be released to the buddy allocator. | |
840 | */ | |
4421cca0 | 841 | void __init_memblock memblock_free(void *ptr, size_t size) |
77e02cf5 LT |
842 | { |
843 | if (ptr) | |
3ecc6834 | 844 | memblock_phys_free(__pa(ptr), size); |
77e02cf5 LT |
845 | } |
846 | ||
4d72868c | 847 | /** |
3ecc6834 | 848 | * memblock_phys_free - free boot memory block |
4d72868c MR |
849 | * @base: phys starting address of the boot memory block |
850 | * @size: size of the boot memory block in bytes | |
851 | * | |
fa81ab49 | 852 | * Free boot memory block previously allocated by memblock_phys_alloc_xx() API. |
4d72868c MR |
853 | * The freeing memory will not be released to the buddy allocator. |
854 | */ | |
3ecc6834 | 855 | int __init_memblock memblock_phys_free(phys_addr_t base, phys_addr_t size) |
95f72d1e | 856 | { |
5d63f81c MC |
857 | phys_addr_t end = base + size - 1; |
858 | ||
a090d711 | 859 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, |
5d63f81c | 860 | &base, &end, (void *)_RET_IP_); |
24aa0788 | 861 | |
9099daed | 862 | kmemleak_free_part_phys(base, size); |
f1af9d3a | 863 | return memblock_remove_range(&memblock.reserved, base, size); |
95f72d1e YL |
864 | } |
865 | ||
f705ac4b | 866 | int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size) |
95f72d1e | 867 | { |
5d63f81c MC |
868 | phys_addr_t end = base + size - 1; |
869 | ||
a090d711 | 870 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, |
5d63f81c | 871 | &base, &end, (void *)_RET_IP_); |
95f72d1e | 872 | |
f705ac4b | 873 | return memblock_add_range(&memblock.reserved, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
874 | } |
875 | ||
02634a44 AK |
876 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
877 | int __init_memblock memblock_physmem_add(phys_addr_t base, phys_addr_t size) | |
878 | { | |
879 | phys_addr_t end = base + size - 1; | |
880 | ||
881 | memblock_dbg("%s: [%pa-%pa] %pS\n", __func__, | |
882 | &base, &end, (void *)_RET_IP_); | |
883 | ||
77649905 | 884 | return memblock_add_range(&physmem, base, size, MAX_NUMNODES, 0); |
02634a44 AK |
885 | } |
886 | #endif | |
887 | ||
66b16edf | 888 | /** |
47cec443 MR |
889 | * memblock_setclr_flag - set or clear flag for a memory region |
890 | * @base: base address of the region | |
891 | * @size: size of the region | |
892 | * @set: set or clear the flag | |
8958b249 | 893 | * @flag: the flag to update |
66b16edf | 894 | * |
4308ce17 | 895 | * This function isolates region [@base, @base + @size), and sets/clears flag |
66b16edf | 896 | * |
47cec443 | 897 | * Return: 0 on success, -errno on failure. |
66b16edf | 898 | */ |
4308ce17 TL |
899 | static int __init_memblock memblock_setclr_flag(phys_addr_t base, |
900 | phys_addr_t size, int set, int flag) | |
66b16edf TC |
901 | { |
902 | struct memblock_type *type = &memblock.memory; | |
903 | int i, ret, start_rgn, end_rgn; | |
904 | ||
905 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); | |
906 | if (ret) | |
907 | return ret; | |
908 | ||
fe145124 MR |
909 | for (i = start_rgn; i < end_rgn; i++) { |
910 | struct memblock_region *r = &type->regions[i]; | |
911 | ||
4308ce17 | 912 | if (set) |
fe145124 | 913 | r->flags |= flag; |
4308ce17 | 914 | else |
fe145124 MR |
915 | r->flags &= ~flag; |
916 | } | |
66b16edf | 917 | |
2fe03412 | 918 | memblock_merge_regions(type, start_rgn, end_rgn); |
66b16edf TC |
919 | return 0; |
920 | } | |
921 | ||
922 | /** | |
4308ce17 | 923 | * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. |
66b16edf TC |
924 | * @base: the base phys addr of the region |
925 | * @size: the size of the region | |
926 | * | |
47cec443 | 927 | * Return: 0 on success, -errno on failure. |
4308ce17 TL |
928 | */ |
929 | int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) | |
930 | { | |
931 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); | |
932 | } | |
933 | ||
934 | /** | |
935 | * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. | |
936 | * @base: the base phys addr of the region | |
937 | * @size: the size of the region | |
66b16edf | 938 | * |
47cec443 | 939 | * Return: 0 on success, -errno on failure. |
66b16edf TC |
940 | */ |
941 | int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) | |
942 | { | |
4308ce17 | 943 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); |
66b16edf TC |
944 | } |
945 | ||
a3f5bafc TL |
946 | /** |
947 | * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. | |
948 | * @base: the base phys addr of the region | |
949 | * @size: the size of the region | |
950 | * | |
47cec443 | 951 | * Return: 0 on success, -errno on failure. |
a3f5bafc TL |
952 | */ |
953 | int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) | |
954 | { | |
902c2d91 MW |
955 | if (!mirrored_kernelcore) |
956 | return 0; | |
957 | ||
a3f5bafc TL |
958 | system_has_some_mirror = true; |
959 | ||
960 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); | |
961 | } | |
962 | ||
bf3d3cc5 AB |
963 | /** |
964 | * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. | |
965 | * @base: the base phys addr of the region | |
966 | * @size: the size of the region | |
967 | * | |
9092d4f7 MR |
968 | * The memory regions marked with %MEMBLOCK_NOMAP will not be added to the |
969 | * direct mapping of the physical memory. These regions will still be | |
970 | * covered by the memory map. The struct page representing NOMAP memory | |
971 | * frames in the memory map will be PageReserved() | |
972 | * | |
658aafc8 MR |
973 | * Note: if the memory being marked %MEMBLOCK_NOMAP was allocated from |
974 | * memblock, the caller must inform kmemleak to ignore that memory | |
975 | * | |
47cec443 | 976 | * Return: 0 on success, -errno on failure. |
bf3d3cc5 AB |
977 | */ |
978 | int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) | |
979 | { | |
6c9a5455 | 980 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); |
bf3d3cc5 | 981 | } |
a3f5bafc | 982 | |
4c546b8a AT |
983 | /** |
984 | * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. | |
985 | * @base: the base phys addr of the region | |
986 | * @size: the size of the region | |
987 | * | |
47cec443 | 988 | * Return: 0 on success, -errno on failure. |
4c546b8a AT |
989 | */ |
990 | int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) | |
991 | { | |
992 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); | |
993 | } | |
994 | ||
9f3d5eaa MR |
995 | static bool should_skip_region(struct memblock_type *type, |
996 | struct memblock_region *m, | |
997 | int nid, int flags) | |
c9a688a3 MR |
998 | { |
999 | int m_nid = memblock_get_region_node(m); | |
1000 | ||
9f3d5eaa MR |
1001 | /* we never skip regions when iterating memblock.reserved or physmem */ |
1002 | if (type != memblock_memory) | |
1003 | return false; | |
1004 | ||
c9a688a3 MR |
1005 | /* only memory regions are associated with nodes, check it */ |
1006 | if (nid != NUMA_NO_NODE && nid != m_nid) | |
1007 | return true; | |
1008 | ||
1009 | /* skip hotpluggable memory regions if needed */ | |
79e482e9 MR |
1010 | if (movable_node_is_enabled() && memblock_is_hotpluggable(m) && |
1011 | !(flags & MEMBLOCK_HOTPLUG)) | |
c9a688a3 MR |
1012 | return true; |
1013 | ||
1014 | /* if we want mirror memory skip non-mirror memory regions */ | |
1015 | if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) | |
1016 | return true; | |
1017 | ||
1018 | /* skip nomap memory unless we were asked for it explicitly */ | |
1019 | if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) | |
1020 | return true; | |
1021 | ||
f7892d8e DH |
1022 | /* skip driver-managed memory unless we were asked for it explicitly */ |
1023 | if (!(flags & MEMBLOCK_DRIVER_MANAGED) && memblock_is_driver_managed(m)) | |
1024 | return true; | |
1025 | ||
c9a688a3 MR |
1026 | return false; |
1027 | } | |
1028 | ||
35fd0808 | 1029 | /** |
a2974133 | 1030 | * __next_mem_range - next function for for_each_free_mem_range() etc. |
35fd0808 | 1031 | * @idx: pointer to u64 loop variable |
b1154233 | 1032 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 1033 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
1034 | * @type_a: pointer to memblock_type from where the range is taken |
1035 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
1036 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
1037 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
1038 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
35fd0808 | 1039 | * |
f1af9d3a | 1040 | * Find the first area from *@idx which matches @nid, fill the out |
35fd0808 | 1041 | * parameters, and update *@idx for the next iteration. The lower 32bit of |
f1af9d3a PH |
1042 | * *@idx contains index into type_a and the upper 32bit indexes the |
1043 | * areas before each region in type_b. For example, if type_b regions | |
35fd0808 TH |
1044 | * look like the following, |
1045 | * | |
1046 | * 0:[0-16), 1:[32-48), 2:[128-130) | |
1047 | * | |
1048 | * The upper 32bit indexes the following regions. | |
1049 | * | |
1050 | * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) | |
1051 | * | |
1052 | * As both region arrays are sorted, the function advances the two indices | |
1053 | * in lockstep and returns each intersection. | |
1054 | */ | |
77649905 DH |
1055 | void __next_mem_range(u64 *idx, int nid, enum memblock_flags flags, |
1056 | struct memblock_type *type_a, | |
1057 | struct memblock_type *type_b, phys_addr_t *out_start, | |
1058 | phys_addr_t *out_end, int *out_nid) | |
35fd0808 | 1059 | { |
f1af9d3a PH |
1060 | int idx_a = *idx & 0xffffffff; |
1061 | int idx_b = *idx >> 32; | |
b1154233 | 1062 | |
f1af9d3a PH |
1063 | if (WARN_ONCE(nid == MAX_NUMNODES, |
1064 | "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) | |
560dca27 | 1065 | nid = NUMA_NO_NODE; |
35fd0808 | 1066 | |
f1af9d3a PH |
1067 | for (; idx_a < type_a->cnt; idx_a++) { |
1068 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1069 | ||
35fd0808 TH |
1070 | phys_addr_t m_start = m->base; |
1071 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1072 | int m_nid = memblock_get_region_node(m); |
35fd0808 | 1073 | |
9f3d5eaa | 1074 | if (should_skip_region(type_a, m, nid, flags)) |
bf3d3cc5 AB |
1075 | continue; |
1076 | ||
f1af9d3a PH |
1077 | if (!type_b) { |
1078 | if (out_start) | |
1079 | *out_start = m_start; | |
1080 | if (out_end) | |
1081 | *out_end = m_end; | |
1082 | if (out_nid) | |
1083 | *out_nid = m_nid; | |
1084 | idx_a++; | |
1085 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
1086 | return; | |
1087 | } | |
1088 | ||
1089 | /* scan areas before each reservation */ | |
1090 | for (; idx_b < type_b->cnt + 1; idx_b++) { | |
1091 | struct memblock_region *r; | |
1092 | phys_addr_t r_start; | |
1093 | phys_addr_t r_end; | |
1094 | ||
1095 | r = &type_b->regions[idx_b]; | |
1096 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1097 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1098 | r->base : PHYS_ADDR_MAX; |
35fd0808 | 1099 | |
f1af9d3a PH |
1100 | /* |
1101 | * if idx_b advanced past idx_a, | |
1102 | * break out to advance idx_a | |
1103 | */ | |
35fd0808 TH |
1104 | if (r_start >= m_end) |
1105 | break; | |
1106 | /* if the two regions intersect, we're done */ | |
1107 | if (m_start < r_end) { | |
1108 | if (out_start) | |
f1af9d3a PH |
1109 | *out_start = |
1110 | max(m_start, r_start); | |
35fd0808 TH |
1111 | if (out_end) |
1112 | *out_end = min(m_end, r_end); | |
1113 | if (out_nid) | |
f1af9d3a | 1114 | *out_nid = m_nid; |
35fd0808 | 1115 | /* |
f1af9d3a PH |
1116 | * The region which ends first is |
1117 | * advanced for the next iteration. | |
35fd0808 TH |
1118 | */ |
1119 | if (m_end <= r_end) | |
f1af9d3a | 1120 | idx_a++; |
35fd0808 | 1121 | else |
f1af9d3a PH |
1122 | idx_b++; |
1123 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
35fd0808 TH |
1124 | return; |
1125 | } | |
1126 | } | |
1127 | } | |
1128 | ||
1129 | /* signal end of iteration */ | |
1130 | *idx = ULLONG_MAX; | |
1131 | } | |
1132 | ||
7bd0b0f0 | 1133 | /** |
f1af9d3a PH |
1134 | * __next_mem_range_rev - generic next function for for_each_*_range_rev() |
1135 | * | |
7bd0b0f0 | 1136 | * @idx: pointer to u64 loop variable |
ad5ea8cd | 1137 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 1138 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
1139 | * @type_a: pointer to memblock_type from where the range is taken |
1140 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
1141 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
1142 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
1143 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
7bd0b0f0 | 1144 | * |
47cec443 MR |
1145 | * Finds the next range from type_a which is not marked as unsuitable |
1146 | * in type_b. | |
1147 | * | |
f1af9d3a | 1148 | * Reverse of __next_mem_range(). |
7bd0b0f0 | 1149 | */ |
e1720fee MR |
1150 | void __init_memblock __next_mem_range_rev(u64 *idx, int nid, |
1151 | enum memblock_flags flags, | |
f1af9d3a PH |
1152 | struct memblock_type *type_a, |
1153 | struct memblock_type *type_b, | |
1154 | phys_addr_t *out_start, | |
1155 | phys_addr_t *out_end, int *out_nid) | |
7bd0b0f0 | 1156 | { |
f1af9d3a PH |
1157 | int idx_a = *idx & 0xffffffff; |
1158 | int idx_b = *idx >> 32; | |
b1154233 | 1159 | |
560dca27 GS |
1160 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1161 | nid = NUMA_NO_NODE; | |
7bd0b0f0 TH |
1162 | |
1163 | if (*idx == (u64)ULLONG_MAX) { | |
f1af9d3a | 1164 | idx_a = type_a->cnt - 1; |
e47608ab | 1165 | if (type_b != NULL) |
1166 | idx_b = type_b->cnt; | |
1167 | else | |
1168 | idx_b = 0; | |
7bd0b0f0 TH |
1169 | } |
1170 | ||
f1af9d3a PH |
1171 | for (; idx_a >= 0; idx_a--) { |
1172 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1173 | ||
7bd0b0f0 TH |
1174 | phys_addr_t m_start = m->base; |
1175 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1176 | int m_nid = memblock_get_region_node(m); |
7bd0b0f0 | 1177 | |
9f3d5eaa | 1178 | if (should_skip_region(type_a, m, nid, flags)) |
bf3d3cc5 AB |
1179 | continue; |
1180 | ||
f1af9d3a PH |
1181 | if (!type_b) { |
1182 | if (out_start) | |
1183 | *out_start = m_start; | |
1184 | if (out_end) | |
1185 | *out_end = m_end; | |
1186 | if (out_nid) | |
1187 | *out_nid = m_nid; | |
fb399b48 | 1188 | idx_a--; |
f1af9d3a PH |
1189 | *idx = (u32)idx_a | (u64)idx_b << 32; |
1190 | return; | |
1191 | } | |
1192 | ||
1193 | /* scan areas before each reservation */ | |
1194 | for (; idx_b >= 0; idx_b--) { | |
1195 | struct memblock_region *r; | |
1196 | phys_addr_t r_start; | |
1197 | phys_addr_t r_end; | |
1198 | ||
1199 | r = &type_b->regions[idx_b]; | |
1200 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1201 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1202 | r->base : PHYS_ADDR_MAX; |
f1af9d3a PH |
1203 | /* |
1204 | * if idx_b advanced past idx_a, | |
1205 | * break out to advance idx_a | |
1206 | */ | |
7bd0b0f0 | 1207 | |
7bd0b0f0 TH |
1208 | if (r_end <= m_start) |
1209 | break; | |
1210 | /* if the two regions intersect, we're done */ | |
1211 | if (m_end > r_start) { | |
1212 | if (out_start) | |
1213 | *out_start = max(m_start, r_start); | |
1214 | if (out_end) | |
1215 | *out_end = min(m_end, r_end); | |
1216 | if (out_nid) | |
f1af9d3a | 1217 | *out_nid = m_nid; |
7bd0b0f0 | 1218 | if (m_start >= r_start) |
f1af9d3a | 1219 | idx_a--; |
7bd0b0f0 | 1220 | else |
f1af9d3a PH |
1221 | idx_b--; |
1222 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
7bd0b0f0 TH |
1223 | return; |
1224 | } | |
1225 | } | |
1226 | } | |
f1af9d3a | 1227 | /* signal end of iteration */ |
7bd0b0f0 TH |
1228 | *idx = ULLONG_MAX; |
1229 | } | |
1230 | ||
7c0caeb8 | 1231 | /* |
45e79815 | 1232 | * Common iterator interface used to define for_each_mem_pfn_range(). |
7c0caeb8 TH |
1233 | */ |
1234 | void __init_memblock __next_mem_pfn_range(int *idx, int nid, | |
1235 | unsigned long *out_start_pfn, | |
1236 | unsigned long *out_end_pfn, int *out_nid) | |
1237 | { | |
1238 | struct memblock_type *type = &memblock.memory; | |
1239 | struct memblock_region *r; | |
d622abf7 | 1240 | int r_nid; |
7c0caeb8 TH |
1241 | |
1242 | while (++*idx < type->cnt) { | |
1243 | r = &type->regions[*idx]; | |
d622abf7 | 1244 | r_nid = memblock_get_region_node(r); |
7c0caeb8 TH |
1245 | |
1246 | if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) | |
1247 | continue; | |
d622abf7 | 1248 | if (nid == MAX_NUMNODES || nid == r_nid) |
7c0caeb8 TH |
1249 | break; |
1250 | } | |
1251 | if (*idx >= type->cnt) { | |
1252 | *idx = -1; | |
1253 | return; | |
1254 | } | |
1255 | ||
1256 | if (out_start_pfn) | |
1257 | *out_start_pfn = PFN_UP(r->base); | |
1258 | if (out_end_pfn) | |
1259 | *out_end_pfn = PFN_DOWN(r->base + r->size); | |
1260 | if (out_nid) | |
d622abf7 | 1261 | *out_nid = r_nid; |
7c0caeb8 TH |
1262 | } |
1263 | ||
1264 | /** | |
1265 | * memblock_set_node - set node ID on memblock regions | |
1266 | * @base: base of area to set node ID for | |
1267 | * @size: size of area to set node ID for | |
e7e8de59 | 1268 | * @type: memblock type to set node ID for |
7c0caeb8 TH |
1269 | * @nid: node ID to set |
1270 | * | |
47cec443 | 1271 | * Set the nid of memblock @type regions in [@base, @base + @size) to @nid. |
7c0caeb8 TH |
1272 | * Regions which cross the area boundaries are split as necessary. |
1273 | * | |
47cec443 | 1274 | * Return: |
7c0caeb8 TH |
1275 | * 0 on success, -errno on failure. |
1276 | */ | |
1277 | int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, | |
e7e8de59 | 1278 | struct memblock_type *type, int nid) |
7c0caeb8 | 1279 | { |
a9ee6cf5 | 1280 | #ifdef CONFIG_NUMA |
6a9ceb31 TH |
1281 | int start_rgn, end_rgn; |
1282 | int i, ret; | |
7c0caeb8 | 1283 | |
6a9ceb31 TH |
1284 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
1285 | if (ret) | |
1286 | return ret; | |
7c0caeb8 | 1287 | |
6a9ceb31 | 1288 | for (i = start_rgn; i < end_rgn; i++) |
e9d24ad3 | 1289 | memblock_set_region_node(&type->regions[i], nid); |
7c0caeb8 | 1290 | |
2fe03412 | 1291 | memblock_merge_regions(type, start_rgn, end_rgn); |
3f08a302 | 1292 | #endif |
7c0caeb8 TH |
1293 | return 0; |
1294 | } | |
3f08a302 | 1295 | |
837566e7 AD |
1296 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT |
1297 | /** | |
1298 | * __next_mem_pfn_range_in_zone - iterator for for_each_*_range_in_zone() | |
1299 | * | |
1300 | * @idx: pointer to u64 loop variable | |
1301 | * @zone: zone in which all of the memory blocks reside | |
1302 | * @out_spfn: ptr to ulong for start pfn of the range, can be %NULL | |
1303 | * @out_epfn: ptr to ulong for end pfn of the range, can be %NULL | |
1304 | * | |
1305 | * This function is meant to be a zone/pfn specific wrapper for the | |
1306 | * for_each_mem_range type iterators. Specifically they are used in the | |
1307 | * deferred memory init routines and as such we were duplicating much of | |
1308 | * this logic throughout the code. So instead of having it in multiple | |
1309 | * locations it seemed like it would make more sense to centralize this to | |
1310 | * one new iterator that does everything they need. | |
1311 | */ | |
1312 | void __init_memblock | |
1313 | __next_mem_pfn_range_in_zone(u64 *idx, struct zone *zone, | |
1314 | unsigned long *out_spfn, unsigned long *out_epfn) | |
1315 | { | |
1316 | int zone_nid = zone_to_nid(zone); | |
1317 | phys_addr_t spa, epa; | |
837566e7 AD |
1318 | |
1319 | __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, | |
1320 | &memblock.memory, &memblock.reserved, | |
f30b002c | 1321 | &spa, &epa, NULL); |
837566e7 AD |
1322 | |
1323 | while (*idx != U64_MAX) { | |
1324 | unsigned long epfn = PFN_DOWN(epa); | |
1325 | unsigned long spfn = PFN_UP(spa); | |
1326 | ||
1327 | /* | |
1328 | * Verify the end is at least past the start of the zone and | |
1329 | * that we have at least one PFN to initialize. | |
1330 | */ | |
1331 | if (zone->zone_start_pfn < epfn && spfn < epfn) { | |
1332 | /* if we went too far just stop searching */ | |
1333 | if (zone_end_pfn(zone) <= spfn) { | |
1334 | *idx = U64_MAX; | |
1335 | break; | |
1336 | } | |
1337 | ||
1338 | if (out_spfn) | |
1339 | *out_spfn = max(zone->zone_start_pfn, spfn); | |
1340 | if (out_epfn) | |
1341 | *out_epfn = min(zone_end_pfn(zone), epfn); | |
1342 | ||
1343 | return; | |
1344 | } | |
1345 | ||
1346 | __next_mem_range(idx, zone_nid, MEMBLOCK_NONE, | |
1347 | &memblock.memory, &memblock.reserved, | |
f30b002c | 1348 | &spa, &epa, NULL); |
837566e7 AD |
1349 | } |
1350 | ||
1351 | /* signal end of iteration */ | |
1352 | if (out_spfn) | |
1353 | *out_spfn = ULONG_MAX; | |
1354 | if (out_epfn) | |
1355 | *out_epfn = 0; | |
1356 | } | |
1357 | ||
1358 | #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ | |
7c0caeb8 | 1359 | |
92d12f95 MR |
1360 | /** |
1361 | * memblock_alloc_range_nid - allocate boot memory block | |
1362 | * @size: size of memory block to be allocated in bytes | |
1363 | * @align: alignment of the region and block's size | |
1364 | * @start: the lower bound of the memory region to allocate (phys address) | |
1365 | * @end: the upper bound of the memory region to allocate (phys address) | |
1366 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
0ac398b1 | 1367 | * @exact_nid: control the allocation fall back to other nodes |
92d12f95 MR |
1368 | * |
1369 | * The allocation is performed from memory region limited by | |
95830666 | 1370 | * memblock.current_limit if @end == %MEMBLOCK_ALLOC_ACCESSIBLE. |
92d12f95 | 1371 | * |
0ac398b1 YY |
1372 | * If the specified node can not hold the requested memory and @exact_nid |
1373 | * is false, the allocation falls back to any node in the system. | |
92d12f95 MR |
1374 | * |
1375 | * For systems with memory mirroring, the allocation is attempted first | |
1376 | * from the regions with mirroring enabled and then retried from any | |
1377 | * memory region. | |
1378 | * | |
c200d900 PW |
1379 | * In addition, function using kmemleak_alloc_phys for allocated boot |
1380 | * memory block, it is never reported as leaks. | |
92d12f95 MR |
1381 | * |
1382 | * Return: | |
1383 | * Physical address of allocated memory block on success, %0 on failure. | |
1384 | */ | |
8676af1f | 1385 | phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, |
2bfc2862 | 1386 | phys_addr_t align, phys_addr_t start, |
0ac398b1 YY |
1387 | phys_addr_t end, int nid, |
1388 | bool exact_nid) | |
95f72d1e | 1389 | { |
92d12f95 | 1390 | enum memblock_flags flags = choose_memblock_flags(); |
6ed311b2 | 1391 | phys_addr_t found; |
95f72d1e | 1392 | |
92d12f95 MR |
1393 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1394 | nid = NUMA_NO_NODE; | |
1395 | ||
2f770806 MR |
1396 | if (!align) { |
1397 | /* Can't use WARNs this early in boot on powerpc */ | |
1398 | dump_stack(); | |
1399 | align = SMP_CACHE_BYTES; | |
1400 | } | |
1401 | ||
92d12f95 | 1402 | again: |
fc6daaf9 TL |
1403 | found = memblock_find_in_range_node(size, align, start, end, nid, |
1404 | flags); | |
92d12f95 MR |
1405 | if (found && !memblock_reserve(found, size)) |
1406 | goto done; | |
1407 | ||
0ac398b1 | 1408 | if (nid != NUMA_NO_NODE && !exact_nid) { |
92d12f95 MR |
1409 | found = memblock_find_in_range_node(size, align, start, |
1410 | end, NUMA_NO_NODE, | |
1411 | flags); | |
1412 | if (found && !memblock_reserve(found, size)) | |
1413 | goto done; | |
1414 | } | |
1415 | ||
1416 | if (flags & MEMBLOCK_MIRROR) { | |
1417 | flags &= ~MEMBLOCK_MIRROR; | |
14d9a675 | 1418 | pr_warn_ratelimited("Could not allocate %pap bytes of mirrored memory\n", |
92d12f95 MR |
1419 | &size); |
1420 | goto again; | |
1421 | } | |
1422 | ||
1423 | return 0; | |
1424 | ||
1425 | done: | |
c6975d7c QC |
1426 | /* |
1427 | * Skip kmemleak for those places like kasan_init() and | |
1428 | * early_pgtable_alloc() due to high volume. | |
1429 | */ | |
1430 | if (end != MEMBLOCK_ALLOC_NOLEAKTRACE) | |
aedf95ea | 1431 | /* |
c200d900 PW |
1432 | * Memblock allocated blocks are never reported as |
1433 | * leaks. This is because many of these blocks are | |
1434 | * only referred via the physical address which is | |
1435 | * not looked up by kmemleak. | |
aedf95ea | 1436 | */ |
c200d900 | 1437 | kmemleak_alloc_phys(found, size, 0); |
92d12f95 MR |
1438 | |
1439 | return found; | |
95f72d1e YL |
1440 | } |
1441 | ||
a2974133 MR |
1442 | /** |
1443 | * memblock_phys_alloc_range - allocate a memory block inside specified range | |
1444 | * @size: size of memory block to be allocated in bytes | |
1445 | * @align: alignment of the region and block's size | |
1446 | * @start: the lower bound of the memory region to allocate (physical address) | |
1447 | * @end: the upper bound of the memory region to allocate (physical address) | |
1448 | * | |
1449 | * Allocate @size bytes in the between @start and @end. | |
1450 | * | |
1451 | * Return: physical address of the allocated memory block on success, | |
1452 | * %0 on failure. | |
1453 | */ | |
8a770c2a MR |
1454 | phys_addr_t __init memblock_phys_alloc_range(phys_addr_t size, |
1455 | phys_addr_t align, | |
1456 | phys_addr_t start, | |
1457 | phys_addr_t end) | |
2bfc2862 | 1458 | { |
b5cf2d6c FM |
1459 | memblock_dbg("%s: %llu bytes align=0x%llx from=%pa max_addr=%pa %pS\n", |
1460 | __func__, (u64)size, (u64)align, &start, &end, | |
1461 | (void *)_RET_IP_); | |
0ac398b1 YY |
1462 | return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE, |
1463 | false); | |
7bd0b0f0 TH |
1464 | } |
1465 | ||
a2974133 | 1466 | /** |
17cbe038 | 1467 | * memblock_phys_alloc_try_nid - allocate a memory block from specified NUMA node |
a2974133 MR |
1468 | * @size: size of memory block to be allocated in bytes |
1469 | * @align: alignment of the region and block's size | |
1470 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1471 | * | |
1472 | * Allocates memory block from the specified NUMA node. If the node | |
1473 | * has no available memory, attempts to allocated from any node in the | |
1474 | * system. | |
1475 | * | |
1476 | * Return: physical address of the allocated memory block on success, | |
1477 | * %0 on failure. | |
1478 | */ | |
9a8dd708 | 1479 | phys_addr_t __init memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) |
9d1e2492 | 1480 | { |
33755574 | 1481 | return memblock_alloc_range_nid(size, align, 0, |
0ac398b1 | 1482 | MEMBLOCK_ALLOC_ACCESSIBLE, nid, false); |
95f72d1e YL |
1483 | } |
1484 | ||
26f09e9b | 1485 | /** |
eb31d559 | 1486 | * memblock_alloc_internal - allocate boot memory block |
26f09e9b SS |
1487 | * @size: size of memory block to be allocated in bytes |
1488 | * @align: alignment of the region and block's size | |
1489 | * @min_addr: the lower bound of the memory region to allocate (phys address) | |
1490 | * @max_addr: the upper bound of the memory region to allocate (phys address) | |
1491 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
0ac398b1 | 1492 | * @exact_nid: control the allocation fall back to other nodes |
26f09e9b | 1493 | * |
92d12f95 MR |
1494 | * Allocates memory block using memblock_alloc_range_nid() and |
1495 | * converts the returned physical address to virtual. | |
26f09e9b | 1496 | * |
92d12f95 MR |
1497 | * The @min_addr limit is dropped if it can not be satisfied and the allocation |
1498 | * will fall back to memory below @min_addr. Other constraints, such | |
1499 | * as node and mirrored memory will be handled again in | |
1500 | * memblock_alloc_range_nid(). | |
26f09e9b | 1501 | * |
47cec443 | 1502 | * Return: |
26f09e9b SS |
1503 | * Virtual address of allocated memory block on success, NULL on failure. |
1504 | */ | |
eb31d559 | 1505 | static void * __init memblock_alloc_internal( |
26f09e9b SS |
1506 | phys_addr_t size, phys_addr_t align, |
1507 | phys_addr_t min_addr, phys_addr_t max_addr, | |
0ac398b1 | 1508 | int nid, bool exact_nid) |
26f09e9b SS |
1509 | { |
1510 | phys_addr_t alloc; | |
26f09e9b SS |
1511 | |
1512 | /* | |
1513 | * Detect any accidental use of these APIs after slab is ready, as at | |
1514 | * this moment memblock may be deinitialized already and its | |
c6ffc5ca | 1515 | * internal data may be destroyed (after execution of memblock_free_all) |
26f09e9b SS |
1516 | */ |
1517 | if (WARN_ON_ONCE(slab_is_available())) | |
1518 | return kzalloc_node(size, GFP_NOWAIT, nid); | |
1519 | ||
f3057ad7 MR |
1520 | if (max_addr > memblock.current_limit) |
1521 | max_addr = memblock.current_limit; | |
1522 | ||
0ac398b1 YY |
1523 | alloc = memblock_alloc_range_nid(size, align, min_addr, max_addr, nid, |
1524 | exact_nid); | |
26f09e9b | 1525 | |
92d12f95 MR |
1526 | /* retry allocation without lower limit */ |
1527 | if (!alloc && min_addr) | |
0ac398b1 YY |
1528 | alloc = memblock_alloc_range_nid(size, align, 0, max_addr, nid, |
1529 | exact_nid); | |
26f09e9b | 1530 | |
92d12f95 MR |
1531 | if (!alloc) |
1532 | return NULL; | |
26f09e9b | 1533 | |
92d12f95 | 1534 | return phys_to_virt(alloc); |
26f09e9b SS |
1535 | } |
1536 | ||
0ac398b1 YY |
1537 | /** |
1538 | * memblock_alloc_exact_nid_raw - allocate boot memory block on the exact node | |
1539 | * without zeroing memory | |
1540 | * @size: size of memory block to be allocated in bytes | |
1541 | * @align: alignment of the region and block's size | |
1542 | * @min_addr: the lower bound of the memory region from where the allocation | |
1543 | * is preferred (phys address) | |
1544 | * @max_addr: the upper bound of the memory region from where the allocation | |
1545 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to | |
1546 | * allocate only from memory limited by memblock.current_limit value | |
1547 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1548 | * | |
1549 | * Public function, provides additional debug information (including caller | |
1550 | * info), if enabled. Does not zero allocated memory. | |
1551 | * | |
1552 | * Return: | |
1553 | * Virtual address of allocated memory block on success, NULL on failure. | |
1554 | */ | |
1555 | void * __init memblock_alloc_exact_nid_raw( | |
1556 | phys_addr_t size, phys_addr_t align, | |
1557 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1558 | int nid) | |
1559 | { | |
0ac398b1 YY |
1560 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
1561 | __func__, (u64)size, (u64)align, nid, &min_addr, | |
1562 | &max_addr, (void *)_RET_IP_); | |
1563 | ||
08678804 MR |
1564 | return memblock_alloc_internal(size, align, min_addr, max_addr, nid, |
1565 | true); | |
0ac398b1 YY |
1566 | } |
1567 | ||
ea1f5f37 | 1568 | /** |
eb31d559 | 1569 | * memblock_alloc_try_nid_raw - allocate boot memory block without zeroing |
ea1f5f37 PT |
1570 | * memory and without panicking |
1571 | * @size: size of memory block to be allocated in bytes | |
1572 | * @align: alignment of the region and block's size | |
1573 | * @min_addr: the lower bound of the memory region from where the allocation | |
1574 | * is preferred (phys address) | |
1575 | * @max_addr: the upper bound of the memory region from where the allocation | |
97ad1087 | 1576 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to |
ea1f5f37 PT |
1577 | * allocate only from memory limited by memblock.current_limit value |
1578 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1579 | * | |
1580 | * Public function, provides additional debug information (including caller | |
1581 | * info), if enabled. Does not zero allocated memory, does not panic if request | |
1582 | * cannot be satisfied. | |
1583 | * | |
47cec443 | 1584 | * Return: |
ea1f5f37 PT |
1585 | * Virtual address of allocated memory block on success, NULL on failure. |
1586 | */ | |
eb31d559 | 1587 | void * __init memblock_alloc_try_nid_raw( |
ea1f5f37 PT |
1588 | phys_addr_t size, phys_addr_t align, |
1589 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1590 | int nid) | |
1591 | { | |
d75f773c | 1592 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
a36aab89 MR |
1593 | __func__, (u64)size, (u64)align, nid, &min_addr, |
1594 | &max_addr, (void *)_RET_IP_); | |
ea1f5f37 | 1595 | |
08678804 MR |
1596 | return memblock_alloc_internal(size, align, min_addr, max_addr, nid, |
1597 | false); | |
ea1f5f37 PT |
1598 | } |
1599 | ||
26f09e9b | 1600 | /** |
c0dbe825 | 1601 | * memblock_alloc_try_nid - allocate boot memory block |
26f09e9b SS |
1602 | * @size: size of memory block to be allocated in bytes |
1603 | * @align: alignment of the region and block's size | |
1604 | * @min_addr: the lower bound of the memory region from where the allocation | |
1605 | * is preferred (phys address) | |
1606 | * @max_addr: the upper bound of the memory region from where the allocation | |
97ad1087 | 1607 | * is preferred (phys address), or %MEMBLOCK_ALLOC_ACCESSIBLE to |
26f09e9b SS |
1608 | * allocate only from memory limited by memblock.current_limit value |
1609 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1610 | * | |
c0dbe825 MR |
1611 | * Public function, provides additional debug information (including caller |
1612 | * info), if enabled. This function zeroes the allocated memory. | |
26f09e9b | 1613 | * |
47cec443 | 1614 | * Return: |
26f09e9b SS |
1615 | * Virtual address of allocated memory block on success, NULL on failure. |
1616 | */ | |
eb31d559 | 1617 | void * __init memblock_alloc_try_nid( |
26f09e9b SS |
1618 | phys_addr_t size, phys_addr_t align, |
1619 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1620 | int nid) | |
1621 | { | |
1622 | void *ptr; | |
1623 | ||
d75f773c | 1624 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=%pa max_addr=%pa %pS\n", |
a36aab89 MR |
1625 | __func__, (u64)size, (u64)align, nid, &min_addr, |
1626 | &max_addr, (void *)_RET_IP_); | |
eb31d559 | 1627 | ptr = memblock_alloc_internal(size, align, |
0ac398b1 | 1628 | min_addr, max_addr, nid, false); |
c0dbe825 | 1629 | if (ptr) |
ea1f5f37 | 1630 | memset(ptr, 0, size); |
26f09e9b | 1631 | |
c0dbe825 | 1632 | return ptr; |
26f09e9b SS |
1633 | } |
1634 | ||
48a833cc | 1635 | /** |
621d9739 | 1636 | * memblock_free_late - free pages directly to buddy allocator |
48a833cc | 1637 | * @base: phys starting address of the boot memory block |
26f09e9b SS |
1638 | * @size: size of the boot memory block in bytes |
1639 | * | |
a2974133 | 1640 | * This is only useful when the memblock allocator has already been torn |
26f09e9b | 1641 | * down, but we are still initializing the system. Pages are released directly |
a2974133 | 1642 | * to the buddy allocator. |
26f09e9b | 1643 | */ |
621d9739 | 1644 | void __init memblock_free_late(phys_addr_t base, phys_addr_t size) |
26f09e9b | 1645 | { |
a36aab89 | 1646 | phys_addr_t cursor, end; |
26f09e9b | 1647 | |
a36aab89 | 1648 | end = base + size - 1; |
d75f773c | 1649 | memblock_dbg("%s: [%pa-%pa] %pS\n", |
a36aab89 | 1650 | __func__, &base, &end, (void *)_RET_IP_); |
9099daed | 1651 | kmemleak_free_part_phys(base, size); |
26f09e9b SS |
1652 | cursor = PFN_UP(base); |
1653 | end = PFN_DOWN(base + size); | |
1654 | ||
1655 | for (; cursor < end; cursor++) { | |
647037ad | 1656 | memblock_free_pages(pfn_to_page(cursor), cursor, 0); |
ca79b0c2 | 1657 | totalram_pages_inc(); |
26f09e9b SS |
1658 | } |
1659 | } | |
9d1e2492 BH |
1660 | |
1661 | /* | |
1662 | * Remaining API functions | |
1663 | */ | |
1664 | ||
1f1ffb8a | 1665 | phys_addr_t __init_memblock memblock_phys_mem_size(void) |
95f72d1e | 1666 | { |
1440c4e2 | 1667 | return memblock.memory.total_size; |
95f72d1e YL |
1668 | } |
1669 | ||
8907de5d SD |
1670 | phys_addr_t __init_memblock memblock_reserved_size(void) |
1671 | { | |
1672 | return memblock.reserved.total_size; | |
1673 | } | |
1674 | ||
0a93ebef SR |
1675 | /* lowest address */ |
1676 | phys_addr_t __init_memblock memblock_start_of_DRAM(void) | |
1677 | { | |
1678 | return memblock.memory.regions[0].base; | |
1679 | } | |
1680 | ||
10d06439 | 1681 | phys_addr_t __init_memblock memblock_end_of_DRAM(void) |
95f72d1e YL |
1682 | { |
1683 | int idx = memblock.memory.cnt - 1; | |
1684 | ||
e3239ff9 | 1685 | return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); |
95f72d1e YL |
1686 | } |
1687 | ||
a571d4eb | 1688 | static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) |
95f72d1e | 1689 | { |
1c4bc43d | 1690 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
136199f0 | 1691 | struct memblock_region *r; |
95f72d1e | 1692 | |
a571d4eb DC |
1693 | /* |
1694 | * translate the memory @limit size into the max address within one of | |
1695 | * the memory memblock regions, if the @limit exceeds the total size | |
1c4bc43d | 1696 | * of those regions, max_addr will keep original value PHYS_ADDR_MAX |
a571d4eb | 1697 | */ |
cc6de168 | 1698 | for_each_mem_region(r) { |
c0ce8fef TH |
1699 | if (limit <= r->size) { |
1700 | max_addr = r->base + limit; | |
1701 | break; | |
95f72d1e | 1702 | } |
c0ce8fef | 1703 | limit -= r->size; |
95f72d1e | 1704 | } |
c0ce8fef | 1705 | |
a571d4eb DC |
1706 | return max_addr; |
1707 | } | |
1708 | ||
1709 | void __init memblock_enforce_memory_limit(phys_addr_t limit) | |
1710 | { | |
49aef717 | 1711 | phys_addr_t max_addr; |
a571d4eb DC |
1712 | |
1713 | if (!limit) | |
1714 | return; | |
1715 | ||
1716 | max_addr = __find_max_addr(limit); | |
1717 | ||
1718 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1719 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1720 | return; |
1721 | ||
c0ce8fef | 1722 | /* truncate both memory and reserved regions */ |
f1af9d3a | 1723 | memblock_remove_range(&memblock.memory, max_addr, |
1c4bc43d | 1724 | PHYS_ADDR_MAX); |
f1af9d3a | 1725 | memblock_remove_range(&memblock.reserved, max_addr, |
1c4bc43d | 1726 | PHYS_ADDR_MAX); |
95f72d1e YL |
1727 | } |
1728 | ||
c9ca9b4e AT |
1729 | void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) |
1730 | { | |
1731 | int start_rgn, end_rgn; | |
1732 | int i, ret; | |
1733 | ||
1734 | if (!size) | |
1735 | return; | |
1736 | ||
5173ed72 | 1737 | if (!memblock_memory->total_size) { |
e888fa7b GU |
1738 | pr_warn("%s: No memory registered yet\n", __func__); |
1739 | return; | |
1740 | } | |
1741 | ||
c9ca9b4e AT |
1742 | ret = memblock_isolate_range(&memblock.memory, base, size, |
1743 | &start_rgn, &end_rgn); | |
1744 | if (ret) | |
1745 | return; | |
1746 | ||
1747 | /* remove all the MAP regions */ | |
1748 | for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) | |
1749 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1750 | memblock_remove_region(&memblock.memory, i); | |
1751 | ||
1752 | for (i = start_rgn - 1; i >= 0; i--) | |
1753 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1754 | memblock_remove_region(&memblock.memory, i); | |
1755 | ||
1756 | /* truncate the reserved regions */ | |
1757 | memblock_remove_range(&memblock.reserved, 0, base); | |
1758 | memblock_remove_range(&memblock.reserved, | |
1c4bc43d | 1759 | base + size, PHYS_ADDR_MAX); |
c9ca9b4e AT |
1760 | } |
1761 | ||
a571d4eb DC |
1762 | void __init memblock_mem_limit_remove_map(phys_addr_t limit) |
1763 | { | |
a571d4eb | 1764 | phys_addr_t max_addr; |
a571d4eb DC |
1765 | |
1766 | if (!limit) | |
1767 | return; | |
1768 | ||
1769 | max_addr = __find_max_addr(limit); | |
1770 | ||
1771 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1772 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1773 | return; |
1774 | ||
c9ca9b4e | 1775 | memblock_cap_memory_range(0, max_addr); |
a571d4eb DC |
1776 | } |
1777 | ||
cd79481d | 1778 | static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) |
72d4b0b4 BH |
1779 | { |
1780 | unsigned int left = 0, right = type->cnt; | |
1781 | ||
1782 | do { | |
1783 | unsigned int mid = (right + left) / 2; | |
1784 | ||
1785 | if (addr < type->regions[mid].base) | |
1786 | right = mid; | |
1787 | else if (addr >= (type->regions[mid].base + | |
1788 | type->regions[mid].size)) | |
1789 | left = mid + 1; | |
1790 | else | |
1791 | return mid; | |
1792 | } while (left < right); | |
1793 | return -1; | |
1794 | } | |
1795 | ||
f5a222dc | 1796 | bool __init_memblock memblock_is_reserved(phys_addr_t addr) |
95f72d1e | 1797 | { |
72d4b0b4 BH |
1798 | return memblock_search(&memblock.reserved, addr) != -1; |
1799 | } | |
95f72d1e | 1800 | |
b4ad0c7e | 1801 | bool __init_memblock memblock_is_memory(phys_addr_t addr) |
72d4b0b4 BH |
1802 | { |
1803 | return memblock_search(&memblock.memory, addr) != -1; | |
1804 | } | |
1805 | ||
937f0c26 | 1806 | bool __init_memblock memblock_is_map_memory(phys_addr_t addr) |
bf3d3cc5 AB |
1807 | { |
1808 | int i = memblock_search(&memblock.memory, addr); | |
1809 | ||
1810 | if (i == -1) | |
1811 | return false; | |
1812 | return !memblock_is_nomap(&memblock.memory.regions[i]); | |
1813 | } | |
1814 | ||
e76b63f8 YL |
1815 | int __init_memblock memblock_search_pfn_nid(unsigned long pfn, |
1816 | unsigned long *start_pfn, unsigned long *end_pfn) | |
1817 | { | |
1818 | struct memblock_type *type = &memblock.memory; | |
16763230 | 1819 | int mid = memblock_search(type, PFN_PHYS(pfn)); |
e76b63f8 YL |
1820 | |
1821 | if (mid == -1) | |
1822 | return -1; | |
1823 | ||
f7e2f7e8 FF |
1824 | *start_pfn = PFN_DOWN(type->regions[mid].base); |
1825 | *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); | |
e76b63f8 | 1826 | |
d622abf7 | 1827 | return memblock_get_region_node(&type->regions[mid]); |
e76b63f8 | 1828 | } |
e76b63f8 | 1829 | |
eab30949 SB |
1830 | /** |
1831 | * memblock_is_region_memory - check if a region is a subset of memory | |
1832 | * @base: base of region to check | |
1833 | * @size: size of region to check | |
1834 | * | |
47cec443 | 1835 | * Check if the region [@base, @base + @size) is a subset of a memory block. |
eab30949 | 1836 | * |
47cec443 | 1837 | * Return: |
eab30949 SB |
1838 | * 0 if false, non-zero if true |
1839 | */ | |
937f0c26 | 1840 | bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) |
72d4b0b4 | 1841 | { |
abb65272 | 1842 | int idx = memblock_search(&memblock.memory, base); |
eb18f1b5 | 1843 | phys_addr_t end = base + memblock_cap_size(base, &size); |
72d4b0b4 BH |
1844 | |
1845 | if (idx == -1) | |
937f0c26 | 1846 | return false; |
ef415ef4 | 1847 | return (memblock.memory.regions[idx].base + |
eb18f1b5 | 1848 | memblock.memory.regions[idx].size) >= end; |
95f72d1e YL |
1849 | } |
1850 | ||
eab30949 SB |
1851 | /** |
1852 | * memblock_is_region_reserved - check if a region intersects reserved memory | |
1853 | * @base: base of region to check | |
1854 | * @size: size of region to check | |
1855 | * | |
47cec443 MR |
1856 | * Check if the region [@base, @base + @size) intersects a reserved |
1857 | * memory block. | |
eab30949 | 1858 | * |
47cec443 | 1859 | * Return: |
c5c5c9d1 | 1860 | * True if they intersect, false if not. |
eab30949 | 1861 | */ |
c5c5c9d1 | 1862 | bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) |
95f72d1e | 1863 | { |
c5c5c9d1 | 1864 | return memblock_overlaps_region(&memblock.reserved, base, size); |
95f72d1e YL |
1865 | } |
1866 | ||
6ede1fd3 YL |
1867 | void __init_memblock memblock_trim_memory(phys_addr_t align) |
1868 | { | |
6ede1fd3 | 1869 | phys_addr_t start, end, orig_start, orig_end; |
136199f0 | 1870 | struct memblock_region *r; |
6ede1fd3 | 1871 | |
cc6de168 | 1872 | for_each_mem_region(r) { |
136199f0 EM |
1873 | orig_start = r->base; |
1874 | orig_end = r->base + r->size; | |
6ede1fd3 YL |
1875 | start = round_up(orig_start, align); |
1876 | end = round_down(orig_end, align); | |
1877 | ||
1878 | if (start == orig_start && end == orig_end) | |
1879 | continue; | |
1880 | ||
1881 | if (start < end) { | |
136199f0 EM |
1882 | r->base = start; |
1883 | r->size = end - start; | |
6ede1fd3 | 1884 | } else { |
136199f0 EM |
1885 | memblock_remove_region(&memblock.memory, |
1886 | r - memblock.memory.regions); | |
1887 | r--; | |
6ede1fd3 YL |
1888 | } |
1889 | } | |
1890 | } | |
e63075a3 | 1891 | |
3661ca66 | 1892 | void __init_memblock memblock_set_current_limit(phys_addr_t limit) |
e63075a3 BH |
1893 | { |
1894 | memblock.current_limit = limit; | |
1895 | } | |
1896 | ||
fec51014 LA |
1897 | phys_addr_t __init_memblock memblock_get_current_limit(void) |
1898 | { | |
1899 | return memblock.current_limit; | |
1900 | } | |
1901 | ||
0262d9c8 | 1902 | static void __init_memblock memblock_dump(struct memblock_type *type) |
6ed311b2 | 1903 | { |
5d63f81c | 1904 | phys_addr_t base, end, size; |
e1720fee | 1905 | enum memblock_flags flags; |
8c9c1701 AK |
1906 | int idx; |
1907 | struct memblock_region *rgn; | |
6ed311b2 | 1908 | |
0262d9c8 | 1909 | pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); |
6ed311b2 | 1910 | |
66e8b438 | 1911 | for_each_memblock_type(idx, type, rgn) { |
7c0caeb8 TH |
1912 | char nid_buf[32] = ""; |
1913 | ||
1914 | base = rgn->base; | |
1915 | size = rgn->size; | |
5d63f81c | 1916 | end = base + size - 1; |
66a20757 | 1917 | flags = rgn->flags; |
a9ee6cf5 | 1918 | #ifdef CONFIG_NUMA |
7c0caeb8 TH |
1919 | if (memblock_get_region_node(rgn) != MAX_NUMNODES) |
1920 | snprintf(nid_buf, sizeof(nid_buf), " on node %d", | |
1921 | memblock_get_region_node(rgn)); | |
1922 | #endif | |
e1720fee | 1923 | pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#x\n", |
0262d9c8 | 1924 | type->name, idx, &base, &end, &size, nid_buf, flags); |
6ed311b2 BH |
1925 | } |
1926 | } | |
1927 | ||
87c55870 | 1928 | static void __init_memblock __memblock_dump_all(void) |
6ed311b2 | 1929 | { |
6ed311b2 | 1930 | pr_info("MEMBLOCK configuration:\n"); |
5d63f81c MC |
1931 | pr_info(" memory size = %pa reserved size = %pa\n", |
1932 | &memblock.memory.total_size, | |
1933 | &memblock.reserved.total_size); | |
6ed311b2 | 1934 | |
0262d9c8 HC |
1935 | memblock_dump(&memblock.memory); |
1936 | memblock_dump(&memblock.reserved); | |
409efd4c | 1937 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
77649905 | 1938 | memblock_dump(&physmem); |
409efd4c | 1939 | #endif |
6ed311b2 BH |
1940 | } |
1941 | ||
87c55870 MR |
1942 | void __init_memblock memblock_dump_all(void) |
1943 | { | |
1944 | if (memblock_debug) | |
1945 | __memblock_dump_all(); | |
1946 | } | |
1947 | ||
1aadc056 | 1948 | void __init memblock_allow_resize(void) |
6ed311b2 | 1949 | { |
142b45a7 | 1950 | memblock_can_resize = 1; |
6ed311b2 BH |
1951 | } |
1952 | ||
6ed311b2 BH |
1953 | static int __init early_memblock(char *p) |
1954 | { | |
1955 | if (p && strstr(p, "debug")) | |
1956 | memblock_debug = 1; | |
1957 | return 0; | |
1958 | } | |
1959 | early_param("memblock", early_memblock); | |
1960 | ||
4f5b0c17 MR |
1961 | static void __init free_memmap(unsigned long start_pfn, unsigned long end_pfn) |
1962 | { | |
1963 | struct page *start_pg, *end_pg; | |
1964 | phys_addr_t pg, pgend; | |
1965 | ||
1966 | /* | |
1967 | * Convert start_pfn/end_pfn to a struct page pointer. | |
1968 | */ | |
1969 | start_pg = pfn_to_page(start_pfn - 1) + 1; | |
1970 | end_pg = pfn_to_page(end_pfn - 1) + 1; | |
1971 | ||
1972 | /* | |
1973 | * Convert to physical addresses, and round start upwards and end | |
1974 | * downwards. | |
1975 | */ | |
1976 | pg = PAGE_ALIGN(__pa(start_pg)); | |
1977 | pgend = __pa(end_pg) & PAGE_MASK; | |
1978 | ||
1979 | /* | |
1980 | * If there are free pages between these, free the section of the | |
1981 | * memmap array. | |
1982 | */ | |
1983 | if (pg < pgend) | |
3ecc6834 | 1984 | memblock_phys_free(pg, pgend - pg); |
4f5b0c17 MR |
1985 | } |
1986 | ||
1987 | /* | |
1988 | * The mem_map array can get very big. Free the unused area of the memory map. | |
1989 | */ | |
1990 | static void __init free_unused_memmap(void) | |
1991 | { | |
1992 | unsigned long start, end, prev_end = 0; | |
1993 | int i; | |
1994 | ||
1995 | if (!IS_ENABLED(CONFIG_HAVE_ARCH_PFN_VALID) || | |
1996 | IS_ENABLED(CONFIG_SPARSEMEM_VMEMMAP)) | |
1997 | return; | |
1998 | ||
1999 | /* | |
2000 | * This relies on each bank being in address order. | |
2001 | * The banks are sorted previously in bootmem_init(). | |
2002 | */ | |
2003 | for_each_mem_pfn_range(i, MAX_NUMNODES, &start, &end, NULL) { | |
2004 | #ifdef CONFIG_SPARSEMEM | |
2005 | /* | |
2006 | * Take care not to free memmap entries that don't exist | |
2007 | * due to SPARSEMEM sections which aren't present. | |
2008 | */ | |
2009 | start = min(start, ALIGN(prev_end, PAGES_PER_SECTION)); | |
f921f53e | 2010 | #endif |
4f5b0c17 | 2011 | /* |
e2a86800 MR |
2012 | * Align down here since many operations in VM subsystem |
2013 | * presume that there are no holes in the memory map inside | |
2014 | * a pageblock | |
4f5b0c17 | 2015 | */ |
4f9bc69a | 2016 | start = pageblock_start_pfn(start); |
4f5b0c17 MR |
2017 | |
2018 | /* | |
2019 | * If we had a previous bank, and there is a space | |
2020 | * between the current bank and the previous, free it. | |
2021 | */ | |
2022 | if (prev_end && prev_end < start) | |
2023 | free_memmap(prev_end, start); | |
2024 | ||
2025 | /* | |
e2a86800 MR |
2026 | * Align up here since many operations in VM subsystem |
2027 | * presume that there are no holes in the memory map inside | |
2028 | * a pageblock | |
4f5b0c17 | 2029 | */ |
5f7fa13f | 2030 | prev_end = pageblock_align(end); |
4f5b0c17 MR |
2031 | } |
2032 | ||
2033 | #ifdef CONFIG_SPARSEMEM | |
f921f53e | 2034 | if (!IS_ALIGNED(prev_end, PAGES_PER_SECTION)) { |
5f7fa13f | 2035 | prev_end = pageblock_align(end); |
4f5b0c17 | 2036 | free_memmap(prev_end, ALIGN(prev_end, PAGES_PER_SECTION)); |
f921f53e | 2037 | } |
4f5b0c17 MR |
2038 | #endif |
2039 | } | |
2040 | ||
bda49a81 MR |
2041 | static void __init __free_pages_memory(unsigned long start, unsigned long end) |
2042 | { | |
2043 | int order; | |
2044 | ||
2045 | while (start < end) { | |
59f876fb KS |
2046 | /* |
2047 | * Free the pages in the largest chunks alignment allows. | |
2048 | * | |
2049 | * __ffs() behaviour is undefined for 0. start == 0 is | |
2050 | * MAX_ORDER-aligned, set order to MAX_ORDER for the case. | |
2051 | */ | |
2052 | if (start) | |
2053 | order = min_t(int, MAX_ORDER, __ffs(start)); | |
2054 | else | |
2055 | order = MAX_ORDER; | |
bda49a81 MR |
2056 | |
2057 | while (start + (1UL << order) > end) | |
2058 | order--; | |
2059 | ||
2060 | memblock_free_pages(pfn_to_page(start), start, order); | |
2061 | ||
2062 | start += (1UL << order); | |
2063 | } | |
2064 | } | |
2065 | ||
2066 | static unsigned long __init __free_memory_core(phys_addr_t start, | |
2067 | phys_addr_t end) | |
2068 | { | |
2069 | unsigned long start_pfn = PFN_UP(start); | |
2070 | unsigned long end_pfn = min_t(unsigned long, | |
2071 | PFN_DOWN(end), max_low_pfn); | |
2072 | ||
2073 | if (start_pfn >= end_pfn) | |
2074 | return 0; | |
2075 | ||
2076 | __free_pages_memory(start_pfn, end_pfn); | |
2077 | ||
2078 | return end_pfn - start_pfn; | |
2079 | } | |
2080 | ||
9092d4f7 MR |
2081 | static void __init memmap_init_reserved_pages(void) |
2082 | { | |
2083 | struct memblock_region *region; | |
2084 | phys_addr_t start, end; | |
2085 | u64 i; | |
2086 | ||
2087 | /* initialize struct pages for the reserved regions */ | |
2088 | for_each_reserved_mem_range(i, &start, &end) | |
2089 | reserve_bootmem_region(start, end); | |
2090 | ||
2091 | /* and also treat struct pages for the NOMAP regions as PageReserved */ | |
2092 | for_each_mem_region(region) { | |
2093 | if (memblock_is_nomap(region)) { | |
2094 | start = region->base; | |
2095 | end = start + region->size; | |
2096 | reserve_bootmem_region(start, end); | |
2097 | } | |
2098 | } | |
2099 | } | |
2100 | ||
bda49a81 MR |
2101 | static unsigned long __init free_low_memory_core_early(void) |
2102 | { | |
2103 | unsigned long count = 0; | |
2104 | phys_addr_t start, end; | |
2105 | u64 i; | |
2106 | ||
2107 | memblock_clear_hotplug(0, -1); | |
2108 | ||
9092d4f7 | 2109 | memmap_init_reserved_pages(); |
bda49a81 MR |
2110 | |
2111 | /* | |
2112 | * We need to use NUMA_NO_NODE instead of NODE_DATA(0)->node_id | |
2113 | * because in some case like Node0 doesn't have RAM installed | |
2114 | * low ram will be on Node1 | |
2115 | */ | |
2116 | for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE, &start, &end, | |
2117 | NULL) | |
2118 | count += __free_memory_core(start, end); | |
2119 | ||
2120 | return count; | |
2121 | } | |
2122 | ||
2123 | static int reset_managed_pages_done __initdata; | |
2124 | ||
2125 | void reset_node_managed_pages(pg_data_t *pgdat) | |
2126 | { | |
2127 | struct zone *z; | |
2128 | ||
2129 | for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) | |
9705bea5 | 2130 | atomic_long_set(&z->managed_pages, 0); |
bda49a81 MR |
2131 | } |
2132 | ||
2133 | void __init reset_all_zones_managed_pages(void) | |
2134 | { | |
2135 | struct pglist_data *pgdat; | |
2136 | ||
2137 | if (reset_managed_pages_done) | |
2138 | return; | |
2139 | ||
2140 | for_each_online_pgdat(pgdat) | |
2141 | reset_node_managed_pages(pgdat); | |
2142 | ||
2143 | reset_managed_pages_done = 1; | |
2144 | } | |
2145 | ||
2146 | /** | |
2147 | * memblock_free_all - release free pages to the buddy allocator | |
bda49a81 | 2148 | */ |
097d43d8 | 2149 | void __init memblock_free_all(void) |
bda49a81 MR |
2150 | { |
2151 | unsigned long pages; | |
2152 | ||
4f5b0c17 | 2153 | free_unused_memmap(); |
bda49a81 MR |
2154 | reset_all_zones_managed_pages(); |
2155 | ||
2156 | pages = free_low_memory_core_early(); | |
ca79b0c2 | 2157 | totalram_pages_add(pages); |
bda49a81 MR |
2158 | } |
2159 | ||
350e88ba | 2160 | #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_ARCH_KEEP_MEMBLOCK) |
6d03b885 BH |
2161 | |
2162 | static int memblock_debug_show(struct seq_file *m, void *private) | |
2163 | { | |
2164 | struct memblock_type *type = m->private; | |
2165 | struct memblock_region *reg; | |
2166 | int i; | |
5d63f81c | 2167 | phys_addr_t end; |
6d03b885 BH |
2168 | |
2169 | for (i = 0; i < type->cnt; i++) { | |
2170 | reg = &type->regions[i]; | |
5d63f81c | 2171 | end = reg->base + reg->size - 1; |
6d03b885 | 2172 | |
5d63f81c MC |
2173 | seq_printf(m, "%4d: ", i); |
2174 | seq_printf(m, "%pa..%pa\n", ®->base, &end); | |
6d03b885 BH |
2175 | } |
2176 | return 0; | |
2177 | } | |
5ad35093 | 2178 | DEFINE_SHOW_ATTRIBUTE(memblock_debug); |
6d03b885 BH |
2179 | |
2180 | static int __init memblock_init_debugfs(void) | |
2181 | { | |
2182 | struct dentry *root = debugfs_create_dir("memblock", NULL); | |
d9f7979c | 2183 | |
0825a6f9 JP |
2184 | debugfs_create_file("memory", 0444, root, |
2185 | &memblock.memory, &memblock_debug_fops); | |
2186 | debugfs_create_file("reserved", 0444, root, | |
2187 | &memblock.reserved, &memblock_debug_fops); | |
70210ed9 | 2188 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
77649905 DH |
2189 | debugfs_create_file("physmem", 0444, root, &physmem, |
2190 | &memblock_debug_fops); | |
70210ed9 | 2191 | #endif |
6d03b885 BH |
2192 | |
2193 | return 0; | |
2194 | } | |
2195 | __initcall(memblock_init_debugfs); | |
2196 | ||
2197 | #endif /* CONFIG_DEBUG_FS */ |