Commit | Line | Data |
---|---|---|
95f72d1e YL |
1 | /* |
2 | * Procedures for maintaining information about logical memory blocks. | |
3 | * | |
4 | * Peter Bergner, IBM Corp. June 2001. | |
5 | * Copyright (C) 2001 Peter Bergner. | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or | |
8 | * modify it under the terms of the GNU General Public License | |
9 | * as published by the Free Software Foundation; either version | |
10 | * 2 of the License, or (at your option) any later version. | |
11 | */ | |
12 | ||
13 | #include <linux/kernel.h> | |
142b45a7 | 14 | #include <linux/slab.h> |
95f72d1e YL |
15 | #include <linux/init.h> |
16 | #include <linux/bitops.h> | |
449e8df3 | 17 | #include <linux/poison.h> |
c196f76f | 18 | #include <linux/pfn.h> |
6d03b885 | 19 | #include <linux/debugfs.h> |
514c6032 | 20 | #include <linux/kmemleak.h> |
6d03b885 | 21 | #include <linux/seq_file.h> |
95f72d1e YL |
22 | #include <linux/memblock.h> |
23 | ||
c4c5ad6b | 24 | #include <asm/sections.h> |
26f09e9b SS |
25 | #include <linux/io.h> |
26 | ||
27 | #include "internal.h" | |
79442ed1 | 28 | |
fe091c20 TH |
29 | static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock; |
30 | static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock; | |
70210ed9 PH |
31 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
32 | static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock; | |
33 | #endif | |
fe091c20 TH |
34 | |
35 | struct memblock memblock __initdata_memblock = { | |
36 | .memory.regions = memblock_memory_init_regions, | |
37 | .memory.cnt = 1, /* empty dummy entry */ | |
38 | .memory.max = INIT_MEMBLOCK_REGIONS, | |
0262d9c8 | 39 | .memory.name = "memory", |
fe091c20 TH |
40 | |
41 | .reserved.regions = memblock_reserved_init_regions, | |
42 | .reserved.cnt = 1, /* empty dummy entry */ | |
43 | .reserved.max = INIT_MEMBLOCK_REGIONS, | |
0262d9c8 | 44 | .reserved.name = "reserved", |
fe091c20 | 45 | |
70210ed9 PH |
46 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
47 | .physmem.regions = memblock_physmem_init_regions, | |
48 | .physmem.cnt = 1, /* empty dummy entry */ | |
49 | .physmem.max = INIT_PHYSMEM_REGIONS, | |
0262d9c8 | 50 | .physmem.name = "physmem", |
70210ed9 PH |
51 | #endif |
52 | ||
79442ed1 | 53 | .bottom_up = false, |
fe091c20 TH |
54 | .current_limit = MEMBLOCK_ALLOC_ANYWHERE, |
55 | }; | |
95f72d1e | 56 | |
10d06439 | 57 | int memblock_debug __initdata_memblock; |
a3f5bafc | 58 | static bool system_has_some_mirror __initdata_memblock = false; |
1aadc056 | 59 | static int memblock_can_resize __initdata_memblock; |
181eb394 GS |
60 | static int memblock_memory_in_slab __initdata_memblock = 0; |
61 | static int memblock_reserved_in_slab __initdata_memblock = 0; | |
95f72d1e | 62 | |
e1720fee | 63 | enum memblock_flags __init_memblock choose_memblock_flags(void) |
a3f5bafc TL |
64 | { |
65 | return system_has_some_mirror ? MEMBLOCK_MIRROR : MEMBLOCK_NONE; | |
66 | } | |
67 | ||
eb18f1b5 TH |
68 | /* adjust *@size so that (@base + *@size) doesn't overflow, return new size */ |
69 | static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size) | |
70 | { | |
1c4bc43d | 71 | return *size = min(*size, PHYS_ADDR_MAX - base); |
eb18f1b5 TH |
72 | } |
73 | ||
6ed311b2 BH |
74 | /* |
75 | * Address comparison utilities | |
76 | */ | |
10d06439 | 77 | static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1, |
2898cc4c | 78 | phys_addr_t base2, phys_addr_t size2) |
95f72d1e YL |
79 | { |
80 | return ((base1 < (base2 + size2)) && (base2 < (base1 + size1))); | |
81 | } | |
82 | ||
95cf82ec | 83 | bool __init_memblock memblock_overlaps_region(struct memblock_type *type, |
2d7d3eb2 | 84 | phys_addr_t base, phys_addr_t size) |
6ed311b2 BH |
85 | { |
86 | unsigned long i; | |
87 | ||
f14516fb AK |
88 | for (i = 0; i < type->cnt; i++) |
89 | if (memblock_addrs_overlap(base, size, type->regions[i].base, | |
90 | type->regions[i].size)) | |
6ed311b2 | 91 | break; |
c5c5c9d1 | 92 | return i < type->cnt; |
6ed311b2 BH |
93 | } |
94 | ||
47cec443 | 95 | /** |
79442ed1 TC |
96 | * __memblock_find_range_bottom_up - find free area utility in bottom-up |
97 | * @start: start of candidate range | |
47cec443 MR |
98 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
99 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
79442ed1 TC |
100 | * @size: size of free area to find |
101 | * @align: alignment of free area to find | |
b1154233 | 102 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 103 | * @flags: pick from blocks based on memory attributes |
79442ed1 TC |
104 | * |
105 | * Utility called from memblock_find_in_range_node(), find free area bottom-up. | |
106 | * | |
47cec443 | 107 | * Return: |
79442ed1 TC |
108 | * Found address on success, 0 on failure. |
109 | */ | |
110 | static phys_addr_t __init_memblock | |
111 | __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 112 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 113 | enum memblock_flags flags) |
79442ed1 TC |
114 | { |
115 | phys_addr_t this_start, this_end, cand; | |
116 | u64 i; | |
117 | ||
fc6daaf9 | 118 | for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { |
79442ed1 TC |
119 | this_start = clamp(this_start, start, end); |
120 | this_end = clamp(this_end, start, end); | |
121 | ||
122 | cand = round_up(this_start, align); | |
123 | if (cand < this_end && this_end - cand >= size) | |
124 | return cand; | |
125 | } | |
126 | ||
127 | return 0; | |
128 | } | |
129 | ||
7bd0b0f0 | 130 | /** |
1402899e | 131 | * __memblock_find_range_top_down - find free area utility, in top-down |
7bd0b0f0 | 132 | * @start: start of candidate range |
47cec443 MR |
133 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
134 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
135 | * @size: size of free area to find |
136 | * @align: alignment of free area to find | |
b1154233 | 137 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 138 | * @flags: pick from blocks based on memory attributes |
7bd0b0f0 | 139 | * |
1402899e | 140 | * Utility called from memblock_find_in_range_node(), find free area top-down. |
7bd0b0f0 | 141 | * |
47cec443 | 142 | * Return: |
79442ed1 | 143 | * Found address on success, 0 on failure. |
6ed311b2 | 144 | */ |
1402899e TC |
145 | static phys_addr_t __init_memblock |
146 | __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 | 147 | phys_addr_t size, phys_addr_t align, int nid, |
e1720fee | 148 | enum memblock_flags flags) |
f7210e6c TC |
149 | { |
150 | phys_addr_t this_start, this_end, cand; | |
151 | u64 i; | |
152 | ||
fc6daaf9 TL |
153 | for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, |
154 | NULL) { | |
f7210e6c TC |
155 | this_start = clamp(this_start, start, end); |
156 | this_end = clamp(this_end, start, end); | |
157 | ||
158 | if (this_end < size) | |
159 | continue; | |
160 | ||
161 | cand = round_down(this_end - size, align); | |
162 | if (cand >= this_start) | |
163 | return cand; | |
164 | } | |
1402899e | 165 | |
f7210e6c TC |
166 | return 0; |
167 | } | |
6ed311b2 | 168 | |
1402899e TC |
169 | /** |
170 | * memblock_find_in_range_node - find free area in given range and node | |
1402899e TC |
171 | * @size: size of free area to find |
172 | * @align: alignment of free area to find | |
87029ee9 | 173 | * @start: start of candidate range |
47cec443 MR |
174 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
175 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
b1154233 | 176 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 177 | * @flags: pick from blocks based on memory attributes |
1402899e TC |
178 | * |
179 | * Find @size free area aligned to @align in the specified range and node. | |
180 | * | |
79442ed1 TC |
181 | * When allocation direction is bottom-up, the @start should be greater |
182 | * than the end of the kernel image. Otherwise, it will be trimmed. The | |
183 | * reason is that we want the bottom-up allocation just near the kernel | |
184 | * image so it is highly likely that the allocated memory and the kernel | |
185 | * will reside in the same node. | |
186 | * | |
187 | * If bottom-up allocation failed, will try to allocate memory top-down. | |
188 | * | |
47cec443 | 189 | * Return: |
79442ed1 | 190 | * Found address on success, 0 on failure. |
1402899e | 191 | */ |
87029ee9 GS |
192 | phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, |
193 | phys_addr_t align, phys_addr_t start, | |
e1720fee MR |
194 | phys_addr_t end, int nid, |
195 | enum memblock_flags flags) | |
1402899e | 196 | { |
0cfb8f0c | 197 | phys_addr_t kernel_end, ret; |
79442ed1 | 198 | |
1402899e TC |
199 | /* pump up @end */ |
200 | if (end == MEMBLOCK_ALLOC_ACCESSIBLE) | |
201 | end = memblock.current_limit; | |
202 | ||
203 | /* avoid allocating the first page */ | |
204 | start = max_t(phys_addr_t, start, PAGE_SIZE); | |
205 | end = max(start, end); | |
79442ed1 TC |
206 | kernel_end = __pa_symbol(_end); |
207 | ||
208 | /* | |
209 | * try bottom-up allocation only when bottom-up mode | |
210 | * is set and @end is above the kernel image. | |
211 | */ | |
212 | if (memblock_bottom_up() && end > kernel_end) { | |
213 | phys_addr_t bottom_up_start; | |
214 | ||
215 | /* make sure we will allocate above the kernel */ | |
216 | bottom_up_start = max(start, kernel_end); | |
217 | ||
218 | /* ok, try bottom-up allocation first */ | |
219 | ret = __memblock_find_range_bottom_up(bottom_up_start, end, | |
fc6daaf9 | 220 | size, align, nid, flags); |
79442ed1 TC |
221 | if (ret) |
222 | return ret; | |
223 | ||
224 | /* | |
225 | * we always limit bottom-up allocation above the kernel, | |
226 | * but top-down allocation doesn't have the limit, so | |
227 | * retrying top-down allocation may succeed when bottom-up | |
228 | * allocation failed. | |
229 | * | |
230 | * bottom-up allocation is expected to be fail very rarely, | |
231 | * so we use WARN_ONCE() here to see the stack trace if | |
232 | * fail happens. | |
233 | */ | |
756a025f | 234 | WARN_ONCE(1, "memblock: bottom-up allocation failed, memory hotunplug may be affected\n"); |
79442ed1 | 235 | } |
1402899e | 236 | |
fc6daaf9 TL |
237 | return __memblock_find_range_top_down(start, end, size, align, nid, |
238 | flags); | |
1402899e TC |
239 | } |
240 | ||
7bd0b0f0 TH |
241 | /** |
242 | * memblock_find_in_range - find free area in given range | |
243 | * @start: start of candidate range | |
47cec443 MR |
244 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_ANYWHERE or |
245 | * %MEMBLOCK_ALLOC_ACCESSIBLE | |
7bd0b0f0 TH |
246 | * @size: size of free area to find |
247 | * @align: alignment of free area to find | |
248 | * | |
249 | * Find @size free area aligned to @align in the specified range. | |
250 | * | |
47cec443 | 251 | * Return: |
79442ed1 | 252 | * Found address on success, 0 on failure. |
fc769a8e | 253 | */ |
7bd0b0f0 TH |
254 | phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, |
255 | phys_addr_t end, phys_addr_t size, | |
256 | phys_addr_t align) | |
6ed311b2 | 257 | { |
a3f5bafc | 258 | phys_addr_t ret; |
e1720fee | 259 | enum memblock_flags flags = choose_memblock_flags(); |
a3f5bafc TL |
260 | |
261 | again: | |
262 | ret = memblock_find_in_range_node(size, align, start, end, | |
263 | NUMA_NO_NODE, flags); | |
264 | ||
265 | if (!ret && (flags & MEMBLOCK_MIRROR)) { | |
266 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
267 | &size); | |
268 | flags &= ~MEMBLOCK_MIRROR; | |
269 | goto again; | |
270 | } | |
271 | ||
272 | return ret; | |
6ed311b2 BH |
273 | } |
274 | ||
10d06439 | 275 | static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) |
95f72d1e | 276 | { |
1440c4e2 | 277 | type->total_size -= type->regions[r].size; |
7c0caeb8 TH |
278 | memmove(&type->regions[r], &type->regions[r + 1], |
279 | (type->cnt - (r + 1)) * sizeof(type->regions[r])); | |
e3239ff9 | 280 | type->cnt--; |
95f72d1e | 281 | |
8f7a6605 BH |
282 | /* Special case for empty arrays */ |
283 | if (type->cnt == 0) { | |
1440c4e2 | 284 | WARN_ON(type->total_size != 0); |
8f7a6605 BH |
285 | type->cnt = 1; |
286 | type->regions[0].base = 0; | |
287 | type->regions[0].size = 0; | |
66a20757 | 288 | type->regions[0].flags = 0; |
7c0caeb8 | 289 | memblock_set_region_node(&type->regions[0], MAX_NUMNODES); |
8f7a6605 | 290 | } |
95f72d1e YL |
291 | } |
292 | ||
354f17e1 | 293 | #ifdef CONFIG_ARCH_DISCARD_MEMBLOCK |
3010f876 | 294 | /** |
47cec443 | 295 | * memblock_discard - discard memory and reserved arrays if they were allocated |
3010f876 PT |
296 | */ |
297 | void __init memblock_discard(void) | |
5e270e25 | 298 | { |
3010f876 | 299 | phys_addr_t addr, size; |
5e270e25 | 300 | |
3010f876 PT |
301 | if (memblock.reserved.regions != memblock_reserved_init_regions) { |
302 | addr = __pa(memblock.reserved.regions); | |
303 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
304 | memblock.reserved.max); | |
305 | __memblock_free_late(addr, size); | |
306 | } | |
5e270e25 | 307 | |
91b540f9 | 308 | if (memblock.memory.regions != memblock_memory_init_regions) { |
3010f876 PT |
309 | addr = __pa(memblock.memory.regions); |
310 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
311 | memblock.memory.max); | |
312 | __memblock_free_late(addr, size); | |
313 | } | |
5e270e25 | 314 | } |
5e270e25 PH |
315 | #endif |
316 | ||
48c3b583 GP |
317 | /** |
318 | * memblock_double_array - double the size of the memblock regions array | |
319 | * @type: memblock type of the regions array being doubled | |
320 | * @new_area_start: starting address of memory range to avoid overlap with | |
321 | * @new_area_size: size of memory range to avoid overlap with | |
322 | * | |
323 | * Double the size of the @type regions array. If memblock is being used to | |
324 | * allocate memory for a new reserved regions array and there is a previously | |
47cec443 | 325 | * allocated memory range [@new_area_start, @new_area_start + @new_area_size] |
48c3b583 GP |
326 | * waiting to be reserved, ensure the memory used by the new array does |
327 | * not overlap. | |
328 | * | |
47cec443 | 329 | * Return: |
48c3b583 GP |
330 | * 0 on success, -1 on failure. |
331 | */ | |
332 | static int __init_memblock memblock_double_array(struct memblock_type *type, | |
333 | phys_addr_t new_area_start, | |
334 | phys_addr_t new_area_size) | |
142b45a7 BH |
335 | { |
336 | struct memblock_region *new_array, *old_array; | |
29f67386 | 337 | phys_addr_t old_alloc_size, new_alloc_size; |
142b45a7 BH |
338 | phys_addr_t old_size, new_size, addr; |
339 | int use_slab = slab_is_available(); | |
181eb394 | 340 | int *in_slab; |
142b45a7 BH |
341 | |
342 | /* We don't allow resizing until we know about the reserved regions | |
343 | * of memory that aren't suitable for allocation | |
344 | */ | |
345 | if (!memblock_can_resize) | |
346 | return -1; | |
347 | ||
142b45a7 BH |
348 | /* Calculate new doubled size */ |
349 | old_size = type->max * sizeof(struct memblock_region); | |
350 | new_size = old_size << 1; | |
29f67386 YL |
351 | /* |
352 | * We need to allocated new one align to PAGE_SIZE, | |
353 | * so we can free them completely later. | |
354 | */ | |
355 | old_alloc_size = PAGE_ALIGN(old_size); | |
356 | new_alloc_size = PAGE_ALIGN(new_size); | |
142b45a7 | 357 | |
181eb394 GS |
358 | /* Retrieve the slab flag */ |
359 | if (type == &memblock.memory) | |
360 | in_slab = &memblock_memory_in_slab; | |
361 | else | |
362 | in_slab = &memblock_reserved_in_slab; | |
363 | ||
142b45a7 BH |
364 | /* Try to find some space for it. |
365 | * | |
366 | * WARNING: We assume that either slab_is_available() and we use it or | |
fd07383b AM |
367 | * we use MEMBLOCK for allocations. That means that this is unsafe to |
368 | * use when bootmem is currently active (unless bootmem itself is | |
369 | * implemented on top of MEMBLOCK which isn't the case yet) | |
142b45a7 BH |
370 | * |
371 | * This should however not be an issue for now, as we currently only | |
fd07383b AM |
372 | * call into MEMBLOCK while it's still active, or much later when slab |
373 | * is active for memory hotplug operations | |
142b45a7 BH |
374 | */ |
375 | if (use_slab) { | |
376 | new_array = kmalloc(new_size, GFP_KERNEL); | |
1f5026a7 | 377 | addr = new_array ? __pa(new_array) : 0; |
4e2f0775 | 378 | } else { |
48c3b583 GP |
379 | /* only exclude range when trying to double reserved.regions */ |
380 | if (type != &memblock.reserved) | |
381 | new_area_start = new_area_size = 0; | |
382 | ||
383 | addr = memblock_find_in_range(new_area_start + new_area_size, | |
384 | memblock.current_limit, | |
29f67386 | 385 | new_alloc_size, PAGE_SIZE); |
48c3b583 GP |
386 | if (!addr && new_area_size) |
387 | addr = memblock_find_in_range(0, | |
fd07383b AM |
388 | min(new_area_start, memblock.current_limit), |
389 | new_alloc_size, PAGE_SIZE); | |
48c3b583 | 390 | |
15674868 | 391 | new_array = addr ? __va(addr) : NULL; |
4e2f0775 | 392 | } |
1f5026a7 | 393 | if (!addr) { |
142b45a7 | 394 | pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", |
0262d9c8 | 395 | type->name, type->max, type->max * 2); |
142b45a7 BH |
396 | return -1; |
397 | } | |
142b45a7 | 398 | |
fd07383b | 399 | memblock_dbg("memblock: %s is doubled to %ld at [%#010llx-%#010llx]", |
0262d9c8 | 400 | type->name, type->max * 2, (u64)addr, |
fd07383b | 401 | (u64)addr + new_size - 1); |
ea9e4376 | 402 | |
fd07383b AM |
403 | /* |
404 | * Found space, we now need to move the array over before we add the | |
405 | * reserved region since it may be our reserved array itself that is | |
406 | * full. | |
142b45a7 BH |
407 | */ |
408 | memcpy(new_array, type->regions, old_size); | |
409 | memset(new_array + type->max, 0, old_size); | |
410 | old_array = type->regions; | |
411 | type->regions = new_array; | |
412 | type->max <<= 1; | |
413 | ||
fd07383b | 414 | /* Free old array. We needn't free it if the array is the static one */ |
181eb394 GS |
415 | if (*in_slab) |
416 | kfree(old_array); | |
417 | else if (old_array != memblock_memory_init_regions && | |
418 | old_array != memblock_reserved_init_regions) | |
29f67386 | 419 | memblock_free(__pa(old_array), old_alloc_size); |
142b45a7 | 420 | |
fd07383b AM |
421 | /* |
422 | * Reserve the new array if that comes from the memblock. Otherwise, we | |
423 | * needn't do it | |
181eb394 GS |
424 | */ |
425 | if (!use_slab) | |
29f67386 | 426 | BUG_ON(memblock_reserve(addr, new_alloc_size)); |
181eb394 GS |
427 | |
428 | /* Update slab flag */ | |
429 | *in_slab = use_slab; | |
430 | ||
142b45a7 BH |
431 | return 0; |
432 | } | |
433 | ||
784656f9 TH |
434 | /** |
435 | * memblock_merge_regions - merge neighboring compatible regions | |
436 | * @type: memblock type to scan | |
437 | * | |
438 | * Scan @type and merge neighboring compatible regions. | |
439 | */ | |
440 | static void __init_memblock memblock_merge_regions(struct memblock_type *type) | |
95f72d1e | 441 | { |
784656f9 | 442 | int i = 0; |
95f72d1e | 443 | |
784656f9 TH |
444 | /* cnt never goes below 1 */ |
445 | while (i < type->cnt - 1) { | |
446 | struct memblock_region *this = &type->regions[i]; | |
447 | struct memblock_region *next = &type->regions[i + 1]; | |
95f72d1e | 448 | |
7c0caeb8 TH |
449 | if (this->base + this->size != next->base || |
450 | memblock_get_region_node(this) != | |
66a20757 TC |
451 | memblock_get_region_node(next) || |
452 | this->flags != next->flags) { | |
784656f9 TH |
453 | BUG_ON(this->base + this->size > next->base); |
454 | i++; | |
455 | continue; | |
8f7a6605 BH |
456 | } |
457 | ||
784656f9 | 458 | this->size += next->size; |
c0232ae8 LF |
459 | /* move forward from next + 1, index of which is i + 2 */ |
460 | memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); | |
784656f9 | 461 | type->cnt--; |
95f72d1e | 462 | } |
784656f9 | 463 | } |
95f72d1e | 464 | |
784656f9 TH |
465 | /** |
466 | * memblock_insert_region - insert new memblock region | |
209ff86d TC |
467 | * @type: memblock type to insert into |
468 | * @idx: index for the insertion point | |
469 | * @base: base address of the new region | |
470 | * @size: size of the new region | |
471 | * @nid: node id of the new region | |
66a20757 | 472 | * @flags: flags of the new region |
784656f9 | 473 | * |
47cec443 | 474 | * Insert new memblock region [@base, @base + @size) into @type at @idx. |
412d0008 | 475 | * @type must already have extra room to accommodate the new region. |
784656f9 TH |
476 | */ |
477 | static void __init_memblock memblock_insert_region(struct memblock_type *type, | |
478 | int idx, phys_addr_t base, | |
66a20757 | 479 | phys_addr_t size, |
e1720fee MR |
480 | int nid, |
481 | enum memblock_flags flags) | |
784656f9 TH |
482 | { |
483 | struct memblock_region *rgn = &type->regions[idx]; | |
484 | ||
485 | BUG_ON(type->cnt >= type->max); | |
486 | memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); | |
487 | rgn->base = base; | |
488 | rgn->size = size; | |
66a20757 | 489 | rgn->flags = flags; |
7c0caeb8 | 490 | memblock_set_region_node(rgn, nid); |
784656f9 | 491 | type->cnt++; |
1440c4e2 | 492 | type->total_size += size; |
784656f9 TH |
493 | } |
494 | ||
495 | /** | |
f1af9d3a | 496 | * memblock_add_range - add new memblock region |
784656f9 TH |
497 | * @type: memblock type to add new region into |
498 | * @base: base address of the new region | |
499 | * @size: size of the new region | |
7fb0bc3f | 500 | * @nid: nid of the new region |
66a20757 | 501 | * @flags: flags of the new region |
784656f9 | 502 | * |
47cec443 | 503 | * Add new memblock region [@base, @base + @size) into @type. The new region |
784656f9 TH |
504 | * is allowed to overlap with existing ones - overlaps don't affect already |
505 | * existing regions. @type is guaranteed to be minimal (all neighbouring | |
506 | * compatible regions are merged) after the addition. | |
507 | * | |
47cec443 | 508 | * Return: |
784656f9 TH |
509 | * 0 on success, -errno on failure. |
510 | */ | |
f1af9d3a | 511 | int __init_memblock memblock_add_range(struct memblock_type *type, |
66a20757 | 512 | phys_addr_t base, phys_addr_t size, |
e1720fee | 513 | int nid, enum memblock_flags flags) |
784656f9 TH |
514 | { |
515 | bool insert = false; | |
eb18f1b5 TH |
516 | phys_addr_t obase = base; |
517 | phys_addr_t end = base + memblock_cap_size(base, &size); | |
8c9c1701 AK |
518 | int idx, nr_new; |
519 | struct memblock_region *rgn; | |
784656f9 | 520 | |
b3dc627c TH |
521 | if (!size) |
522 | return 0; | |
523 | ||
784656f9 TH |
524 | /* special case for empty array */ |
525 | if (type->regions[0].size == 0) { | |
1440c4e2 | 526 | WARN_ON(type->cnt != 1 || type->total_size); |
8f7a6605 BH |
527 | type->regions[0].base = base; |
528 | type->regions[0].size = size; | |
66a20757 | 529 | type->regions[0].flags = flags; |
7fb0bc3f | 530 | memblock_set_region_node(&type->regions[0], nid); |
1440c4e2 | 531 | type->total_size = size; |
8f7a6605 | 532 | return 0; |
95f72d1e | 533 | } |
784656f9 TH |
534 | repeat: |
535 | /* | |
536 | * The following is executed twice. Once with %false @insert and | |
537 | * then with %true. The first counts the number of regions needed | |
412d0008 | 538 | * to accommodate the new area. The second actually inserts them. |
142b45a7 | 539 | */ |
784656f9 TH |
540 | base = obase; |
541 | nr_new = 0; | |
95f72d1e | 542 | |
66e8b438 | 543 | for_each_memblock_type(idx, type, rgn) { |
784656f9 TH |
544 | phys_addr_t rbase = rgn->base; |
545 | phys_addr_t rend = rbase + rgn->size; | |
546 | ||
547 | if (rbase >= end) | |
95f72d1e | 548 | break; |
784656f9 TH |
549 | if (rend <= base) |
550 | continue; | |
551 | /* | |
552 | * @rgn overlaps. If it separates the lower part of new | |
553 | * area, insert that portion. | |
554 | */ | |
555 | if (rbase > base) { | |
c0a29498 WY |
556 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
557 | WARN_ON(nid != memblock_get_region_node(rgn)); | |
558 | #endif | |
4fcab5f4 | 559 | WARN_ON(flags != rgn->flags); |
784656f9 TH |
560 | nr_new++; |
561 | if (insert) | |
8c9c1701 | 562 | memblock_insert_region(type, idx++, base, |
66a20757 TC |
563 | rbase - base, nid, |
564 | flags); | |
95f72d1e | 565 | } |
784656f9 TH |
566 | /* area below @rend is dealt with, forget about it */ |
567 | base = min(rend, end); | |
95f72d1e | 568 | } |
784656f9 TH |
569 | |
570 | /* insert the remaining portion */ | |
571 | if (base < end) { | |
572 | nr_new++; | |
573 | if (insert) | |
8c9c1701 | 574 | memblock_insert_region(type, idx, base, end - base, |
66a20757 | 575 | nid, flags); |
95f72d1e | 576 | } |
95f72d1e | 577 | |
ef3cc4db | 578 | if (!nr_new) |
579 | return 0; | |
580 | ||
784656f9 TH |
581 | /* |
582 | * If this was the first round, resize array and repeat for actual | |
583 | * insertions; otherwise, merge and return. | |
142b45a7 | 584 | */ |
784656f9 TH |
585 | if (!insert) { |
586 | while (type->cnt + nr_new > type->max) | |
48c3b583 | 587 | if (memblock_double_array(type, obase, size) < 0) |
784656f9 TH |
588 | return -ENOMEM; |
589 | insert = true; | |
590 | goto repeat; | |
591 | } else { | |
592 | memblock_merge_regions(type); | |
593 | return 0; | |
142b45a7 | 594 | } |
95f72d1e YL |
595 | } |
596 | ||
48a833cc MR |
597 | /** |
598 | * memblock_add_node - add new memblock region within a NUMA node | |
599 | * @base: base address of the new region | |
600 | * @size: size of the new region | |
601 | * @nid: nid of the new region | |
602 | * | |
603 | * Add new memblock region [@base, @base + @size) to the "memory" | |
604 | * type. See memblock_add_range() description for mode details | |
605 | * | |
606 | * Return: | |
607 | * 0 on success, -errno on failure. | |
608 | */ | |
7fb0bc3f TH |
609 | int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, |
610 | int nid) | |
611 | { | |
f1af9d3a | 612 | return memblock_add_range(&memblock.memory, base, size, nid, 0); |
7fb0bc3f TH |
613 | } |
614 | ||
48a833cc MR |
615 | /** |
616 | * memblock_add - add new memblock region | |
617 | * @base: base address of the new region | |
618 | * @size: size of the new region | |
619 | * | |
620 | * Add new memblock region [@base, @base + @size) to the "memory" | |
621 | * type. See memblock_add_range() description for mode details | |
622 | * | |
623 | * Return: | |
624 | * 0 on success, -errno on failure. | |
625 | */ | |
f705ac4b | 626 | int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) |
6a4055bc | 627 | { |
5d63f81c MC |
628 | phys_addr_t end = base + size - 1; |
629 | ||
630 | memblock_dbg("memblock_add: [%pa-%pa] %pF\n", | |
631 | &base, &end, (void *)_RET_IP_); | |
6a4055bc | 632 | |
f705ac4b | 633 | return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
634 | } |
635 | ||
6a9ceb31 TH |
636 | /** |
637 | * memblock_isolate_range - isolate given range into disjoint memblocks | |
638 | * @type: memblock type to isolate range for | |
639 | * @base: base of range to isolate | |
640 | * @size: size of range to isolate | |
641 | * @start_rgn: out parameter for the start of isolated region | |
642 | * @end_rgn: out parameter for the end of isolated region | |
643 | * | |
644 | * Walk @type and ensure that regions don't cross the boundaries defined by | |
47cec443 | 645 | * [@base, @base + @size). Crossing regions are split at the boundaries, |
6a9ceb31 TH |
646 | * which may create at most two more regions. The index of the first |
647 | * region inside the range is returned in *@start_rgn and end in *@end_rgn. | |
648 | * | |
47cec443 | 649 | * Return: |
6a9ceb31 TH |
650 | * 0 on success, -errno on failure. |
651 | */ | |
652 | static int __init_memblock memblock_isolate_range(struct memblock_type *type, | |
653 | phys_addr_t base, phys_addr_t size, | |
654 | int *start_rgn, int *end_rgn) | |
655 | { | |
eb18f1b5 | 656 | phys_addr_t end = base + memblock_cap_size(base, &size); |
8c9c1701 AK |
657 | int idx; |
658 | struct memblock_region *rgn; | |
6a9ceb31 TH |
659 | |
660 | *start_rgn = *end_rgn = 0; | |
661 | ||
b3dc627c TH |
662 | if (!size) |
663 | return 0; | |
664 | ||
6a9ceb31 TH |
665 | /* we'll create at most two more regions */ |
666 | while (type->cnt + 2 > type->max) | |
48c3b583 | 667 | if (memblock_double_array(type, base, size) < 0) |
6a9ceb31 TH |
668 | return -ENOMEM; |
669 | ||
66e8b438 | 670 | for_each_memblock_type(idx, type, rgn) { |
6a9ceb31 TH |
671 | phys_addr_t rbase = rgn->base; |
672 | phys_addr_t rend = rbase + rgn->size; | |
673 | ||
674 | if (rbase >= end) | |
675 | break; | |
676 | if (rend <= base) | |
677 | continue; | |
678 | ||
679 | if (rbase < base) { | |
680 | /* | |
681 | * @rgn intersects from below. Split and continue | |
682 | * to process the next region - the new top half. | |
683 | */ | |
684 | rgn->base = base; | |
1440c4e2 TH |
685 | rgn->size -= base - rbase; |
686 | type->total_size -= base - rbase; | |
8c9c1701 | 687 | memblock_insert_region(type, idx, rbase, base - rbase, |
66a20757 TC |
688 | memblock_get_region_node(rgn), |
689 | rgn->flags); | |
6a9ceb31 TH |
690 | } else if (rend > end) { |
691 | /* | |
692 | * @rgn intersects from above. Split and redo the | |
693 | * current region - the new bottom half. | |
694 | */ | |
695 | rgn->base = end; | |
1440c4e2 TH |
696 | rgn->size -= end - rbase; |
697 | type->total_size -= end - rbase; | |
8c9c1701 | 698 | memblock_insert_region(type, idx--, rbase, end - rbase, |
66a20757 TC |
699 | memblock_get_region_node(rgn), |
700 | rgn->flags); | |
6a9ceb31 TH |
701 | } else { |
702 | /* @rgn is fully contained, record it */ | |
703 | if (!*end_rgn) | |
8c9c1701 AK |
704 | *start_rgn = idx; |
705 | *end_rgn = idx + 1; | |
6a9ceb31 TH |
706 | } |
707 | } | |
708 | ||
709 | return 0; | |
710 | } | |
6a9ceb31 | 711 | |
35bd16a2 | 712 | static int __init_memblock memblock_remove_range(struct memblock_type *type, |
f1af9d3a | 713 | phys_addr_t base, phys_addr_t size) |
95f72d1e | 714 | { |
71936180 TH |
715 | int start_rgn, end_rgn; |
716 | int i, ret; | |
95f72d1e | 717 | |
71936180 TH |
718 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
719 | if (ret) | |
720 | return ret; | |
95f72d1e | 721 | |
71936180 TH |
722 | for (i = end_rgn - 1; i >= start_rgn; i--) |
723 | memblock_remove_region(type, i); | |
8f7a6605 | 724 | return 0; |
95f72d1e YL |
725 | } |
726 | ||
581adcbe | 727 | int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) |
95f72d1e | 728 | { |
25cf23d7 MK |
729 | phys_addr_t end = base + size - 1; |
730 | ||
731 | memblock_dbg("memblock_remove: [%pa-%pa] %pS\n", | |
732 | &base, &end, (void *)_RET_IP_); | |
733 | ||
f1af9d3a | 734 | return memblock_remove_range(&memblock.memory, base, size); |
95f72d1e YL |
735 | } |
736 | ||
f1af9d3a | 737 | |
581adcbe | 738 | int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size) |
95f72d1e | 739 | { |
5d63f81c MC |
740 | phys_addr_t end = base + size - 1; |
741 | ||
742 | memblock_dbg(" memblock_free: [%pa-%pa] %pF\n", | |
743 | &base, &end, (void *)_RET_IP_); | |
24aa0788 | 744 | |
9099daed | 745 | kmemleak_free_part_phys(base, size); |
f1af9d3a | 746 | return memblock_remove_range(&memblock.reserved, base, size); |
95f72d1e YL |
747 | } |
748 | ||
f705ac4b | 749 | int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size) |
95f72d1e | 750 | { |
5d63f81c MC |
751 | phys_addr_t end = base + size - 1; |
752 | ||
753 | memblock_dbg("memblock_reserve: [%pa-%pa] %pF\n", | |
754 | &base, &end, (void *)_RET_IP_); | |
95f72d1e | 755 | |
f705ac4b | 756 | return memblock_add_range(&memblock.reserved, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
757 | } |
758 | ||
66b16edf | 759 | /** |
47cec443 MR |
760 | * memblock_setclr_flag - set or clear flag for a memory region |
761 | * @base: base address of the region | |
762 | * @size: size of the region | |
763 | * @set: set or clear the flag | |
764 | * @flag: the flag to udpate | |
66b16edf | 765 | * |
4308ce17 | 766 | * This function isolates region [@base, @base + @size), and sets/clears flag |
66b16edf | 767 | * |
47cec443 | 768 | * Return: 0 on success, -errno on failure. |
66b16edf | 769 | */ |
4308ce17 TL |
770 | static int __init_memblock memblock_setclr_flag(phys_addr_t base, |
771 | phys_addr_t size, int set, int flag) | |
66b16edf TC |
772 | { |
773 | struct memblock_type *type = &memblock.memory; | |
774 | int i, ret, start_rgn, end_rgn; | |
775 | ||
776 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); | |
777 | if (ret) | |
778 | return ret; | |
779 | ||
780 | for (i = start_rgn; i < end_rgn; i++) | |
4308ce17 TL |
781 | if (set) |
782 | memblock_set_region_flags(&type->regions[i], flag); | |
783 | else | |
784 | memblock_clear_region_flags(&type->regions[i], flag); | |
66b16edf TC |
785 | |
786 | memblock_merge_regions(type); | |
787 | return 0; | |
788 | } | |
789 | ||
790 | /** | |
4308ce17 | 791 | * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. |
66b16edf TC |
792 | * @base: the base phys addr of the region |
793 | * @size: the size of the region | |
794 | * | |
47cec443 | 795 | * Return: 0 on success, -errno on failure. |
4308ce17 TL |
796 | */ |
797 | int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) | |
798 | { | |
799 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); | |
800 | } | |
801 | ||
802 | /** | |
803 | * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. | |
804 | * @base: the base phys addr of the region | |
805 | * @size: the size of the region | |
66b16edf | 806 | * |
47cec443 | 807 | * Return: 0 on success, -errno on failure. |
66b16edf TC |
808 | */ |
809 | int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) | |
810 | { | |
4308ce17 | 811 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); |
66b16edf TC |
812 | } |
813 | ||
a3f5bafc TL |
814 | /** |
815 | * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. | |
816 | * @base: the base phys addr of the region | |
817 | * @size: the size of the region | |
818 | * | |
47cec443 | 819 | * Return: 0 on success, -errno on failure. |
a3f5bafc TL |
820 | */ |
821 | int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) | |
822 | { | |
823 | system_has_some_mirror = true; | |
824 | ||
825 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); | |
826 | } | |
827 | ||
bf3d3cc5 AB |
828 | /** |
829 | * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. | |
830 | * @base: the base phys addr of the region | |
831 | * @size: the size of the region | |
832 | * | |
47cec443 | 833 | * Return: 0 on success, -errno on failure. |
bf3d3cc5 AB |
834 | */ |
835 | int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) | |
836 | { | |
837 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); | |
838 | } | |
a3f5bafc | 839 | |
4c546b8a AT |
840 | /** |
841 | * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. | |
842 | * @base: the base phys addr of the region | |
843 | * @size: the size of the region | |
844 | * | |
47cec443 | 845 | * Return: 0 on success, -errno on failure. |
4c546b8a AT |
846 | */ |
847 | int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) | |
848 | { | |
849 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); | |
850 | } | |
851 | ||
8e7a7f86 RH |
852 | /** |
853 | * __next_reserved_mem_region - next function for for_each_reserved_region() | |
854 | * @idx: pointer to u64 loop variable | |
855 | * @out_start: ptr to phys_addr_t for start address of the region, can be %NULL | |
856 | * @out_end: ptr to phys_addr_t for end address of the region, can be %NULL | |
857 | * | |
858 | * Iterate over all reserved memory regions. | |
859 | */ | |
860 | void __init_memblock __next_reserved_mem_region(u64 *idx, | |
861 | phys_addr_t *out_start, | |
862 | phys_addr_t *out_end) | |
863 | { | |
567d117b | 864 | struct memblock_type *type = &memblock.reserved; |
8e7a7f86 | 865 | |
cd33a76b | 866 | if (*idx < type->cnt) { |
567d117b | 867 | struct memblock_region *r = &type->regions[*idx]; |
8e7a7f86 RH |
868 | phys_addr_t base = r->base; |
869 | phys_addr_t size = r->size; | |
870 | ||
871 | if (out_start) | |
872 | *out_start = base; | |
873 | if (out_end) | |
874 | *out_end = base + size - 1; | |
875 | ||
876 | *idx += 1; | |
877 | return; | |
878 | } | |
879 | ||
880 | /* signal end of iteration */ | |
881 | *idx = ULLONG_MAX; | |
882 | } | |
883 | ||
35fd0808 | 884 | /** |
f1af9d3a | 885 | * __next__mem_range - next function for for_each_free_mem_range() etc. |
35fd0808 | 886 | * @idx: pointer to u64 loop variable |
b1154233 | 887 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 888 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
889 | * @type_a: pointer to memblock_type from where the range is taken |
890 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
891 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
892 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
893 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
35fd0808 | 894 | * |
f1af9d3a | 895 | * Find the first area from *@idx which matches @nid, fill the out |
35fd0808 | 896 | * parameters, and update *@idx for the next iteration. The lower 32bit of |
f1af9d3a PH |
897 | * *@idx contains index into type_a and the upper 32bit indexes the |
898 | * areas before each region in type_b. For example, if type_b regions | |
35fd0808 TH |
899 | * look like the following, |
900 | * | |
901 | * 0:[0-16), 1:[32-48), 2:[128-130) | |
902 | * | |
903 | * The upper 32bit indexes the following regions. | |
904 | * | |
905 | * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) | |
906 | * | |
907 | * As both region arrays are sorted, the function advances the two indices | |
908 | * in lockstep and returns each intersection. | |
909 | */ | |
e1720fee MR |
910 | void __init_memblock __next_mem_range(u64 *idx, int nid, |
911 | enum memblock_flags flags, | |
f1af9d3a PH |
912 | struct memblock_type *type_a, |
913 | struct memblock_type *type_b, | |
914 | phys_addr_t *out_start, | |
915 | phys_addr_t *out_end, int *out_nid) | |
35fd0808 | 916 | { |
f1af9d3a PH |
917 | int idx_a = *idx & 0xffffffff; |
918 | int idx_b = *idx >> 32; | |
b1154233 | 919 | |
f1af9d3a PH |
920 | if (WARN_ONCE(nid == MAX_NUMNODES, |
921 | "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) | |
560dca27 | 922 | nid = NUMA_NO_NODE; |
35fd0808 | 923 | |
f1af9d3a PH |
924 | for (; idx_a < type_a->cnt; idx_a++) { |
925 | struct memblock_region *m = &type_a->regions[idx_a]; | |
926 | ||
35fd0808 TH |
927 | phys_addr_t m_start = m->base; |
928 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 929 | int m_nid = memblock_get_region_node(m); |
35fd0808 TH |
930 | |
931 | /* only memory regions are associated with nodes, check it */ | |
f1af9d3a | 932 | if (nid != NUMA_NO_NODE && nid != m_nid) |
35fd0808 TH |
933 | continue; |
934 | ||
0a313a99 XQ |
935 | /* skip hotpluggable memory regions if needed */ |
936 | if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) | |
937 | continue; | |
938 | ||
a3f5bafc TL |
939 | /* if we want mirror memory skip non-mirror memory regions */ |
940 | if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) | |
941 | continue; | |
942 | ||
bf3d3cc5 AB |
943 | /* skip nomap memory unless we were asked for it explicitly */ |
944 | if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) | |
945 | continue; | |
946 | ||
f1af9d3a PH |
947 | if (!type_b) { |
948 | if (out_start) | |
949 | *out_start = m_start; | |
950 | if (out_end) | |
951 | *out_end = m_end; | |
952 | if (out_nid) | |
953 | *out_nid = m_nid; | |
954 | idx_a++; | |
955 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
956 | return; | |
957 | } | |
958 | ||
959 | /* scan areas before each reservation */ | |
960 | for (; idx_b < type_b->cnt + 1; idx_b++) { | |
961 | struct memblock_region *r; | |
962 | phys_addr_t r_start; | |
963 | phys_addr_t r_end; | |
964 | ||
965 | r = &type_b->regions[idx_b]; | |
966 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
967 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 968 | r->base : PHYS_ADDR_MAX; |
35fd0808 | 969 | |
f1af9d3a PH |
970 | /* |
971 | * if idx_b advanced past idx_a, | |
972 | * break out to advance idx_a | |
973 | */ | |
35fd0808 TH |
974 | if (r_start >= m_end) |
975 | break; | |
976 | /* if the two regions intersect, we're done */ | |
977 | if (m_start < r_end) { | |
978 | if (out_start) | |
f1af9d3a PH |
979 | *out_start = |
980 | max(m_start, r_start); | |
35fd0808 TH |
981 | if (out_end) |
982 | *out_end = min(m_end, r_end); | |
983 | if (out_nid) | |
f1af9d3a | 984 | *out_nid = m_nid; |
35fd0808 | 985 | /* |
f1af9d3a PH |
986 | * The region which ends first is |
987 | * advanced for the next iteration. | |
35fd0808 TH |
988 | */ |
989 | if (m_end <= r_end) | |
f1af9d3a | 990 | idx_a++; |
35fd0808 | 991 | else |
f1af9d3a PH |
992 | idx_b++; |
993 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
35fd0808 TH |
994 | return; |
995 | } | |
996 | } | |
997 | } | |
998 | ||
999 | /* signal end of iteration */ | |
1000 | *idx = ULLONG_MAX; | |
1001 | } | |
1002 | ||
7bd0b0f0 | 1003 | /** |
f1af9d3a PH |
1004 | * __next_mem_range_rev - generic next function for for_each_*_range_rev() |
1005 | * | |
7bd0b0f0 | 1006 | * @idx: pointer to u64 loop variable |
ad5ea8cd | 1007 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 1008 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
1009 | * @type_a: pointer to memblock_type from where the range is taken |
1010 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
1011 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
1012 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
1013 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
7bd0b0f0 | 1014 | * |
47cec443 MR |
1015 | * Finds the next range from type_a which is not marked as unsuitable |
1016 | * in type_b. | |
1017 | * | |
f1af9d3a | 1018 | * Reverse of __next_mem_range(). |
7bd0b0f0 | 1019 | */ |
e1720fee MR |
1020 | void __init_memblock __next_mem_range_rev(u64 *idx, int nid, |
1021 | enum memblock_flags flags, | |
f1af9d3a PH |
1022 | struct memblock_type *type_a, |
1023 | struct memblock_type *type_b, | |
1024 | phys_addr_t *out_start, | |
1025 | phys_addr_t *out_end, int *out_nid) | |
7bd0b0f0 | 1026 | { |
f1af9d3a PH |
1027 | int idx_a = *idx & 0xffffffff; |
1028 | int idx_b = *idx >> 32; | |
b1154233 | 1029 | |
560dca27 GS |
1030 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1031 | nid = NUMA_NO_NODE; | |
7bd0b0f0 TH |
1032 | |
1033 | if (*idx == (u64)ULLONG_MAX) { | |
f1af9d3a | 1034 | idx_a = type_a->cnt - 1; |
e47608ab | 1035 | if (type_b != NULL) |
1036 | idx_b = type_b->cnt; | |
1037 | else | |
1038 | idx_b = 0; | |
7bd0b0f0 TH |
1039 | } |
1040 | ||
f1af9d3a PH |
1041 | for (; idx_a >= 0; idx_a--) { |
1042 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1043 | ||
7bd0b0f0 TH |
1044 | phys_addr_t m_start = m->base; |
1045 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1046 | int m_nid = memblock_get_region_node(m); |
7bd0b0f0 TH |
1047 | |
1048 | /* only memory regions are associated with nodes, check it */ | |
f1af9d3a | 1049 | if (nid != NUMA_NO_NODE && nid != m_nid) |
7bd0b0f0 TH |
1050 | continue; |
1051 | ||
55ac590c TC |
1052 | /* skip hotpluggable memory regions if needed */ |
1053 | if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) | |
1054 | continue; | |
1055 | ||
a3f5bafc TL |
1056 | /* if we want mirror memory skip non-mirror memory regions */ |
1057 | if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) | |
1058 | continue; | |
1059 | ||
bf3d3cc5 AB |
1060 | /* skip nomap memory unless we were asked for it explicitly */ |
1061 | if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) | |
1062 | continue; | |
1063 | ||
f1af9d3a PH |
1064 | if (!type_b) { |
1065 | if (out_start) | |
1066 | *out_start = m_start; | |
1067 | if (out_end) | |
1068 | *out_end = m_end; | |
1069 | if (out_nid) | |
1070 | *out_nid = m_nid; | |
fb399b48 | 1071 | idx_a--; |
f1af9d3a PH |
1072 | *idx = (u32)idx_a | (u64)idx_b << 32; |
1073 | return; | |
1074 | } | |
1075 | ||
1076 | /* scan areas before each reservation */ | |
1077 | for (; idx_b >= 0; idx_b--) { | |
1078 | struct memblock_region *r; | |
1079 | phys_addr_t r_start; | |
1080 | phys_addr_t r_end; | |
1081 | ||
1082 | r = &type_b->regions[idx_b]; | |
1083 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1084 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1085 | r->base : PHYS_ADDR_MAX; |
f1af9d3a PH |
1086 | /* |
1087 | * if idx_b advanced past idx_a, | |
1088 | * break out to advance idx_a | |
1089 | */ | |
7bd0b0f0 | 1090 | |
7bd0b0f0 TH |
1091 | if (r_end <= m_start) |
1092 | break; | |
1093 | /* if the two regions intersect, we're done */ | |
1094 | if (m_end > r_start) { | |
1095 | if (out_start) | |
1096 | *out_start = max(m_start, r_start); | |
1097 | if (out_end) | |
1098 | *out_end = min(m_end, r_end); | |
1099 | if (out_nid) | |
f1af9d3a | 1100 | *out_nid = m_nid; |
7bd0b0f0 | 1101 | if (m_start >= r_start) |
f1af9d3a | 1102 | idx_a--; |
7bd0b0f0 | 1103 | else |
f1af9d3a PH |
1104 | idx_b--; |
1105 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
7bd0b0f0 TH |
1106 | return; |
1107 | } | |
1108 | } | |
1109 | } | |
f1af9d3a | 1110 | /* signal end of iteration */ |
7bd0b0f0 TH |
1111 | *idx = ULLONG_MAX; |
1112 | } | |
1113 | ||
7c0caeb8 TH |
1114 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1115 | /* | |
1116 | * Common iterator interface used to define for_each_mem_range(). | |
1117 | */ | |
1118 | void __init_memblock __next_mem_pfn_range(int *idx, int nid, | |
1119 | unsigned long *out_start_pfn, | |
1120 | unsigned long *out_end_pfn, int *out_nid) | |
1121 | { | |
1122 | struct memblock_type *type = &memblock.memory; | |
1123 | struct memblock_region *r; | |
1124 | ||
1125 | while (++*idx < type->cnt) { | |
1126 | r = &type->regions[*idx]; | |
1127 | ||
1128 | if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) | |
1129 | continue; | |
1130 | if (nid == MAX_NUMNODES || nid == r->nid) | |
1131 | break; | |
1132 | } | |
1133 | if (*idx >= type->cnt) { | |
1134 | *idx = -1; | |
1135 | return; | |
1136 | } | |
1137 | ||
1138 | if (out_start_pfn) | |
1139 | *out_start_pfn = PFN_UP(r->base); | |
1140 | if (out_end_pfn) | |
1141 | *out_end_pfn = PFN_DOWN(r->base + r->size); | |
1142 | if (out_nid) | |
1143 | *out_nid = r->nid; | |
1144 | } | |
1145 | ||
1146 | /** | |
1147 | * memblock_set_node - set node ID on memblock regions | |
1148 | * @base: base of area to set node ID for | |
1149 | * @size: size of area to set node ID for | |
e7e8de59 | 1150 | * @type: memblock type to set node ID for |
7c0caeb8 TH |
1151 | * @nid: node ID to set |
1152 | * | |
47cec443 | 1153 | * Set the nid of memblock @type regions in [@base, @base + @size) to @nid. |
7c0caeb8 TH |
1154 | * Regions which cross the area boundaries are split as necessary. |
1155 | * | |
47cec443 | 1156 | * Return: |
7c0caeb8 TH |
1157 | * 0 on success, -errno on failure. |
1158 | */ | |
1159 | int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, | |
e7e8de59 | 1160 | struct memblock_type *type, int nid) |
7c0caeb8 | 1161 | { |
6a9ceb31 TH |
1162 | int start_rgn, end_rgn; |
1163 | int i, ret; | |
7c0caeb8 | 1164 | |
6a9ceb31 TH |
1165 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
1166 | if (ret) | |
1167 | return ret; | |
7c0caeb8 | 1168 | |
6a9ceb31 | 1169 | for (i = start_rgn; i < end_rgn; i++) |
e9d24ad3 | 1170 | memblock_set_region_node(&type->regions[i], nid); |
7c0caeb8 TH |
1171 | |
1172 | memblock_merge_regions(type); | |
1173 | return 0; | |
1174 | } | |
1175 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ | |
1176 | ||
2bfc2862 AM |
1177 | static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, |
1178 | phys_addr_t align, phys_addr_t start, | |
e1720fee MR |
1179 | phys_addr_t end, int nid, |
1180 | enum memblock_flags flags) | |
95f72d1e | 1181 | { |
6ed311b2 | 1182 | phys_addr_t found; |
95f72d1e | 1183 | |
79f40fab GS |
1184 | if (!align) |
1185 | align = SMP_CACHE_BYTES; | |
94f3d3af | 1186 | |
fc6daaf9 TL |
1187 | found = memblock_find_in_range_node(size, align, start, end, nid, |
1188 | flags); | |
aedf95ea CM |
1189 | if (found && !memblock_reserve(found, size)) { |
1190 | /* | |
1191 | * The min_count is set to 0 so that memblock allocations are | |
1192 | * never reported as leaks. | |
1193 | */ | |
9099daed | 1194 | kmemleak_alloc_phys(found, size, 0, 0); |
6ed311b2 | 1195 | return found; |
aedf95ea | 1196 | } |
6ed311b2 | 1197 | return 0; |
95f72d1e YL |
1198 | } |
1199 | ||
2bfc2862 | 1200 | phys_addr_t __init memblock_alloc_range(phys_addr_t size, phys_addr_t align, |
fc6daaf9 | 1201 | phys_addr_t start, phys_addr_t end, |
e1720fee | 1202 | enum memblock_flags flags) |
2bfc2862 | 1203 | { |
fc6daaf9 TL |
1204 | return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE, |
1205 | flags); | |
2bfc2862 AM |
1206 | } |
1207 | ||
b575454f | 1208 | phys_addr_t __init memblock_alloc_base_nid(phys_addr_t size, |
2bfc2862 | 1209 | phys_addr_t align, phys_addr_t max_addr, |
e1720fee | 1210 | int nid, enum memblock_flags flags) |
2bfc2862 | 1211 | { |
fc6daaf9 | 1212 | return memblock_alloc_range_nid(size, align, 0, max_addr, nid, flags); |
2bfc2862 AM |
1213 | } |
1214 | ||
7bd0b0f0 TH |
1215 | phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid) |
1216 | { | |
e1720fee | 1217 | enum memblock_flags flags = choose_memblock_flags(); |
a3f5bafc TL |
1218 | phys_addr_t ret; |
1219 | ||
1220 | again: | |
1221 | ret = memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE, | |
1222 | nid, flags); | |
1223 | ||
1224 | if (!ret && (flags & MEMBLOCK_MIRROR)) { | |
1225 | flags &= ~MEMBLOCK_MIRROR; | |
1226 | goto again; | |
1227 | } | |
1228 | return ret; | |
7bd0b0f0 TH |
1229 | } |
1230 | ||
1231 | phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr) | |
1232 | { | |
fc6daaf9 TL |
1233 | return memblock_alloc_base_nid(size, align, max_addr, NUMA_NO_NODE, |
1234 | MEMBLOCK_NONE); | |
7bd0b0f0 TH |
1235 | } |
1236 | ||
6ed311b2 | 1237 | phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr) |
95f72d1e | 1238 | { |
6ed311b2 BH |
1239 | phys_addr_t alloc; |
1240 | ||
1241 | alloc = __memblock_alloc_base(size, align, max_addr); | |
1242 | ||
1243 | if (alloc == 0) | |
5d63f81c MC |
1244 | panic("ERROR: Failed to allocate %pa bytes below %pa.\n", |
1245 | &size, &max_addr); | |
6ed311b2 BH |
1246 | |
1247 | return alloc; | |
95f72d1e YL |
1248 | } |
1249 | ||
6ed311b2 | 1250 | phys_addr_t __init memblock_alloc(phys_addr_t size, phys_addr_t align) |
95f72d1e | 1251 | { |
6ed311b2 BH |
1252 | return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE); |
1253 | } | |
95f72d1e | 1254 | |
9d1e2492 BH |
1255 | phys_addr_t __init memblock_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) |
1256 | { | |
1257 | phys_addr_t res = memblock_alloc_nid(size, align, nid); | |
1258 | ||
1259 | if (res) | |
1260 | return res; | |
15fb0972 | 1261 | return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE); |
95f72d1e YL |
1262 | } |
1263 | ||
26f09e9b SS |
1264 | /** |
1265 | * memblock_virt_alloc_internal - allocate boot memory block | |
1266 | * @size: size of memory block to be allocated in bytes | |
1267 | * @align: alignment of the region and block's size | |
1268 | * @min_addr: the lower bound of the memory region to allocate (phys address) | |
1269 | * @max_addr: the upper bound of the memory region to allocate (phys address) | |
1270 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1271 | * | |
1272 | * The @min_addr limit is dropped if it can not be satisfied and the allocation | |
1273 | * will fall back to memory below @min_addr. Also, allocation may fall back | |
1274 | * to any node in the system if the specified node can not | |
1275 | * hold the requested memory. | |
1276 | * | |
1277 | * The allocation is performed from memory region limited by | |
1278 | * memblock.current_limit if @max_addr == %BOOTMEM_ALLOC_ACCESSIBLE. | |
1279 | * | |
47cec443 | 1280 | * The memory block is aligned on %SMP_CACHE_BYTES if @align == 0. |
26f09e9b SS |
1281 | * |
1282 | * The phys address of allocated boot memory block is converted to virtual and | |
1283 | * allocated memory is reset to 0. | |
1284 | * | |
1285 | * In addition, function sets the min_count to 0 using kmemleak_alloc for | |
1286 | * allocated boot memory block, so that it is never reported as leaks. | |
1287 | * | |
47cec443 | 1288 | * Return: |
26f09e9b SS |
1289 | * Virtual address of allocated memory block on success, NULL on failure. |
1290 | */ | |
1291 | static void * __init memblock_virt_alloc_internal( | |
1292 | phys_addr_t size, phys_addr_t align, | |
1293 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1294 | int nid) | |
1295 | { | |
1296 | phys_addr_t alloc; | |
1297 | void *ptr; | |
e1720fee | 1298 | enum memblock_flags flags = choose_memblock_flags(); |
26f09e9b | 1299 | |
560dca27 GS |
1300 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1301 | nid = NUMA_NO_NODE; | |
26f09e9b SS |
1302 | |
1303 | /* | |
1304 | * Detect any accidental use of these APIs after slab is ready, as at | |
1305 | * this moment memblock may be deinitialized already and its | |
1306 | * internal data may be destroyed (after execution of free_all_bootmem) | |
1307 | */ | |
1308 | if (WARN_ON_ONCE(slab_is_available())) | |
1309 | return kzalloc_node(size, GFP_NOWAIT, nid); | |
1310 | ||
1311 | if (!align) | |
1312 | align = SMP_CACHE_BYTES; | |
1313 | ||
f544e14f YL |
1314 | if (max_addr > memblock.current_limit) |
1315 | max_addr = memblock.current_limit; | |
26f09e9b SS |
1316 | again: |
1317 | alloc = memblock_find_in_range_node(size, align, min_addr, max_addr, | |
a3f5bafc | 1318 | nid, flags); |
7d41c03e | 1319 | if (alloc && !memblock_reserve(alloc, size)) |
26f09e9b SS |
1320 | goto done; |
1321 | ||
1322 | if (nid != NUMA_NO_NODE) { | |
1323 | alloc = memblock_find_in_range_node(size, align, min_addr, | |
fc6daaf9 | 1324 | max_addr, NUMA_NO_NODE, |
a3f5bafc | 1325 | flags); |
7d41c03e | 1326 | if (alloc && !memblock_reserve(alloc, size)) |
26f09e9b SS |
1327 | goto done; |
1328 | } | |
1329 | ||
1330 | if (min_addr) { | |
1331 | min_addr = 0; | |
1332 | goto again; | |
26f09e9b SS |
1333 | } |
1334 | ||
a3f5bafc TL |
1335 | if (flags & MEMBLOCK_MIRROR) { |
1336 | flags &= ~MEMBLOCK_MIRROR; | |
1337 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
1338 | &size); | |
1339 | goto again; | |
1340 | } | |
1341 | ||
1342 | return NULL; | |
26f09e9b | 1343 | done: |
26f09e9b | 1344 | ptr = phys_to_virt(alloc); |
26f09e9b SS |
1345 | |
1346 | /* | |
1347 | * The min_count is set to 0 so that bootmem allocated blocks | |
1348 | * are never reported as leaks. This is because many of these blocks | |
1349 | * are only referred via the physical address which is not | |
1350 | * looked up by kmemleak. | |
1351 | */ | |
1352 | kmemleak_alloc(ptr, size, 0, 0); | |
1353 | ||
1354 | return ptr; | |
26f09e9b SS |
1355 | } |
1356 | ||
ea1f5f37 PT |
1357 | /** |
1358 | * memblock_virt_alloc_try_nid_raw - allocate boot memory block without zeroing | |
1359 | * memory and without panicking | |
1360 | * @size: size of memory block to be allocated in bytes | |
1361 | * @align: alignment of the region and block's size | |
1362 | * @min_addr: the lower bound of the memory region from where the allocation | |
1363 | * is preferred (phys address) | |
1364 | * @max_addr: the upper bound of the memory region from where the allocation | |
1365 | * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to | |
1366 | * allocate only from memory limited by memblock.current_limit value | |
1367 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1368 | * | |
1369 | * Public function, provides additional debug information (including caller | |
1370 | * info), if enabled. Does not zero allocated memory, does not panic if request | |
1371 | * cannot be satisfied. | |
1372 | * | |
47cec443 | 1373 | * Return: |
ea1f5f37 PT |
1374 | * Virtual address of allocated memory block on success, NULL on failure. |
1375 | */ | |
1376 | void * __init memblock_virt_alloc_try_nid_raw( | |
1377 | phys_addr_t size, phys_addr_t align, | |
1378 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1379 | int nid) | |
1380 | { | |
1381 | void *ptr; | |
1382 | ||
1383 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n", | |
1384 | __func__, (u64)size, (u64)align, nid, (u64)min_addr, | |
1385 | (u64)max_addr, (void *)_RET_IP_); | |
1386 | ||
1387 | ptr = memblock_virt_alloc_internal(size, align, | |
1388 | min_addr, max_addr, nid); | |
1389 | #ifdef CONFIG_DEBUG_VM | |
1390 | if (ptr && size > 0) | |
f165b378 | 1391 | memset(ptr, PAGE_POISON_PATTERN, size); |
ea1f5f37 PT |
1392 | #endif |
1393 | return ptr; | |
1394 | } | |
1395 | ||
26f09e9b SS |
1396 | /** |
1397 | * memblock_virt_alloc_try_nid_nopanic - allocate boot memory block | |
1398 | * @size: size of memory block to be allocated in bytes | |
1399 | * @align: alignment of the region and block's size | |
1400 | * @min_addr: the lower bound of the memory region from where the allocation | |
1401 | * is preferred (phys address) | |
1402 | * @max_addr: the upper bound of the memory region from where the allocation | |
1403 | * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to | |
1404 | * allocate only from memory limited by memblock.current_limit value | |
1405 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1406 | * | |
ea1f5f37 PT |
1407 | * Public function, provides additional debug information (including caller |
1408 | * info), if enabled. This function zeroes the allocated memory. | |
26f09e9b | 1409 | * |
47cec443 | 1410 | * Return: |
26f09e9b SS |
1411 | * Virtual address of allocated memory block on success, NULL on failure. |
1412 | */ | |
1413 | void * __init memblock_virt_alloc_try_nid_nopanic( | |
1414 | phys_addr_t size, phys_addr_t align, | |
1415 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1416 | int nid) | |
1417 | { | |
ea1f5f37 PT |
1418 | void *ptr; |
1419 | ||
26f09e9b SS |
1420 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n", |
1421 | __func__, (u64)size, (u64)align, nid, (u64)min_addr, | |
1422 | (u64)max_addr, (void *)_RET_IP_); | |
ea1f5f37 PT |
1423 | |
1424 | ptr = memblock_virt_alloc_internal(size, align, | |
1425 | min_addr, max_addr, nid); | |
1426 | if (ptr) | |
1427 | memset(ptr, 0, size); | |
1428 | return ptr; | |
26f09e9b SS |
1429 | } |
1430 | ||
1431 | /** | |
1432 | * memblock_virt_alloc_try_nid - allocate boot memory block with panicking | |
1433 | * @size: size of memory block to be allocated in bytes | |
1434 | * @align: alignment of the region and block's size | |
1435 | * @min_addr: the lower bound of the memory region from where the allocation | |
1436 | * is preferred (phys address) | |
1437 | * @max_addr: the upper bound of the memory region from where the allocation | |
1438 | * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to | |
1439 | * allocate only from memory limited by memblock.current_limit value | |
1440 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1441 | * | |
ea1f5f37 | 1442 | * Public panicking version of memblock_virt_alloc_try_nid_nopanic() |
26f09e9b SS |
1443 | * which provides debug information (including caller info), if enabled, |
1444 | * and panics if the request can not be satisfied. | |
1445 | * | |
47cec443 | 1446 | * Return: |
26f09e9b SS |
1447 | * Virtual address of allocated memory block on success, NULL on failure. |
1448 | */ | |
1449 | void * __init memblock_virt_alloc_try_nid( | |
1450 | phys_addr_t size, phys_addr_t align, | |
1451 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1452 | int nid) | |
1453 | { | |
1454 | void *ptr; | |
1455 | ||
1456 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n", | |
1457 | __func__, (u64)size, (u64)align, nid, (u64)min_addr, | |
1458 | (u64)max_addr, (void *)_RET_IP_); | |
1459 | ptr = memblock_virt_alloc_internal(size, align, | |
1460 | min_addr, max_addr, nid); | |
ea1f5f37 PT |
1461 | if (ptr) { |
1462 | memset(ptr, 0, size); | |
26f09e9b | 1463 | return ptr; |
ea1f5f37 | 1464 | } |
26f09e9b SS |
1465 | |
1466 | panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx\n", | |
1467 | __func__, (u64)size, (u64)align, nid, (u64)min_addr, | |
1468 | (u64)max_addr); | |
1469 | return NULL; | |
1470 | } | |
1471 | ||
1472 | /** | |
1473 | * __memblock_free_early - free boot memory block | |
1474 | * @base: phys starting address of the boot memory block | |
1475 | * @size: size of the boot memory block in bytes | |
1476 | * | |
1477 | * Free boot memory block previously allocated by memblock_virt_alloc_xx() API. | |
1478 | * The freeing memory will not be released to the buddy allocator. | |
1479 | */ | |
1480 | void __init __memblock_free_early(phys_addr_t base, phys_addr_t size) | |
1481 | { | |
1482 | memblock_dbg("%s: [%#016llx-%#016llx] %pF\n", | |
1483 | __func__, (u64)base, (u64)base + size - 1, | |
1484 | (void *)_RET_IP_); | |
9099daed | 1485 | kmemleak_free_part_phys(base, size); |
f1af9d3a | 1486 | memblock_remove_range(&memblock.reserved, base, size); |
26f09e9b SS |
1487 | } |
1488 | ||
48a833cc | 1489 | /** |
26f09e9b | 1490 | * __memblock_free_late - free bootmem block pages directly to buddy allocator |
48a833cc | 1491 | * @base: phys starting address of the boot memory block |
26f09e9b SS |
1492 | * @size: size of the boot memory block in bytes |
1493 | * | |
1494 | * This is only useful when the bootmem allocator has already been torn | |
1495 | * down, but we are still initializing the system. Pages are released directly | |
1496 | * to the buddy allocator, no bootmem metadata is updated because it is gone. | |
1497 | */ | |
1498 | void __init __memblock_free_late(phys_addr_t base, phys_addr_t size) | |
1499 | { | |
1500 | u64 cursor, end; | |
1501 | ||
1502 | memblock_dbg("%s: [%#016llx-%#016llx] %pF\n", | |
1503 | __func__, (u64)base, (u64)base + size - 1, | |
1504 | (void *)_RET_IP_); | |
9099daed | 1505 | kmemleak_free_part_phys(base, size); |
26f09e9b SS |
1506 | cursor = PFN_UP(base); |
1507 | end = PFN_DOWN(base + size); | |
1508 | ||
1509 | for (; cursor < end; cursor++) { | |
d70ddd7a | 1510 | __free_pages_bootmem(pfn_to_page(cursor), cursor, 0); |
26f09e9b SS |
1511 | totalram_pages++; |
1512 | } | |
1513 | } | |
9d1e2492 BH |
1514 | |
1515 | /* | |
1516 | * Remaining API functions | |
1517 | */ | |
1518 | ||
1f1ffb8a | 1519 | phys_addr_t __init_memblock memblock_phys_mem_size(void) |
95f72d1e | 1520 | { |
1440c4e2 | 1521 | return memblock.memory.total_size; |
95f72d1e YL |
1522 | } |
1523 | ||
8907de5d SD |
1524 | phys_addr_t __init_memblock memblock_reserved_size(void) |
1525 | { | |
1526 | return memblock.reserved.total_size; | |
1527 | } | |
1528 | ||
595ad9af YL |
1529 | phys_addr_t __init memblock_mem_size(unsigned long limit_pfn) |
1530 | { | |
1531 | unsigned long pages = 0; | |
1532 | struct memblock_region *r; | |
1533 | unsigned long start_pfn, end_pfn; | |
1534 | ||
1535 | for_each_memblock(memory, r) { | |
1536 | start_pfn = memblock_region_memory_base_pfn(r); | |
1537 | end_pfn = memblock_region_memory_end_pfn(r); | |
1538 | start_pfn = min_t(unsigned long, start_pfn, limit_pfn); | |
1539 | end_pfn = min_t(unsigned long, end_pfn, limit_pfn); | |
1540 | pages += end_pfn - start_pfn; | |
1541 | } | |
1542 | ||
16763230 | 1543 | return PFN_PHYS(pages); |
595ad9af YL |
1544 | } |
1545 | ||
0a93ebef SR |
1546 | /* lowest address */ |
1547 | phys_addr_t __init_memblock memblock_start_of_DRAM(void) | |
1548 | { | |
1549 | return memblock.memory.regions[0].base; | |
1550 | } | |
1551 | ||
10d06439 | 1552 | phys_addr_t __init_memblock memblock_end_of_DRAM(void) |
95f72d1e YL |
1553 | { |
1554 | int idx = memblock.memory.cnt - 1; | |
1555 | ||
e3239ff9 | 1556 | return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); |
95f72d1e YL |
1557 | } |
1558 | ||
a571d4eb | 1559 | static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) |
95f72d1e | 1560 | { |
1c4bc43d | 1561 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
136199f0 | 1562 | struct memblock_region *r; |
95f72d1e | 1563 | |
a571d4eb DC |
1564 | /* |
1565 | * translate the memory @limit size into the max address within one of | |
1566 | * the memory memblock regions, if the @limit exceeds the total size | |
1c4bc43d | 1567 | * of those regions, max_addr will keep original value PHYS_ADDR_MAX |
a571d4eb | 1568 | */ |
136199f0 | 1569 | for_each_memblock(memory, r) { |
c0ce8fef TH |
1570 | if (limit <= r->size) { |
1571 | max_addr = r->base + limit; | |
1572 | break; | |
95f72d1e | 1573 | } |
c0ce8fef | 1574 | limit -= r->size; |
95f72d1e | 1575 | } |
c0ce8fef | 1576 | |
a571d4eb DC |
1577 | return max_addr; |
1578 | } | |
1579 | ||
1580 | void __init memblock_enforce_memory_limit(phys_addr_t limit) | |
1581 | { | |
1c4bc43d | 1582 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
a571d4eb DC |
1583 | |
1584 | if (!limit) | |
1585 | return; | |
1586 | ||
1587 | max_addr = __find_max_addr(limit); | |
1588 | ||
1589 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1590 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1591 | return; |
1592 | ||
c0ce8fef | 1593 | /* truncate both memory and reserved regions */ |
f1af9d3a | 1594 | memblock_remove_range(&memblock.memory, max_addr, |
1c4bc43d | 1595 | PHYS_ADDR_MAX); |
f1af9d3a | 1596 | memblock_remove_range(&memblock.reserved, max_addr, |
1c4bc43d | 1597 | PHYS_ADDR_MAX); |
95f72d1e YL |
1598 | } |
1599 | ||
c9ca9b4e AT |
1600 | void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) |
1601 | { | |
1602 | int start_rgn, end_rgn; | |
1603 | int i, ret; | |
1604 | ||
1605 | if (!size) | |
1606 | return; | |
1607 | ||
1608 | ret = memblock_isolate_range(&memblock.memory, base, size, | |
1609 | &start_rgn, &end_rgn); | |
1610 | if (ret) | |
1611 | return; | |
1612 | ||
1613 | /* remove all the MAP regions */ | |
1614 | for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) | |
1615 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1616 | memblock_remove_region(&memblock.memory, i); | |
1617 | ||
1618 | for (i = start_rgn - 1; i >= 0; i--) | |
1619 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1620 | memblock_remove_region(&memblock.memory, i); | |
1621 | ||
1622 | /* truncate the reserved regions */ | |
1623 | memblock_remove_range(&memblock.reserved, 0, base); | |
1624 | memblock_remove_range(&memblock.reserved, | |
1c4bc43d | 1625 | base + size, PHYS_ADDR_MAX); |
c9ca9b4e AT |
1626 | } |
1627 | ||
a571d4eb DC |
1628 | void __init memblock_mem_limit_remove_map(phys_addr_t limit) |
1629 | { | |
a571d4eb | 1630 | phys_addr_t max_addr; |
a571d4eb DC |
1631 | |
1632 | if (!limit) | |
1633 | return; | |
1634 | ||
1635 | max_addr = __find_max_addr(limit); | |
1636 | ||
1637 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1638 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1639 | return; |
1640 | ||
c9ca9b4e | 1641 | memblock_cap_memory_range(0, max_addr); |
a571d4eb DC |
1642 | } |
1643 | ||
cd79481d | 1644 | static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) |
72d4b0b4 BH |
1645 | { |
1646 | unsigned int left = 0, right = type->cnt; | |
1647 | ||
1648 | do { | |
1649 | unsigned int mid = (right + left) / 2; | |
1650 | ||
1651 | if (addr < type->regions[mid].base) | |
1652 | right = mid; | |
1653 | else if (addr >= (type->regions[mid].base + | |
1654 | type->regions[mid].size)) | |
1655 | left = mid + 1; | |
1656 | else | |
1657 | return mid; | |
1658 | } while (left < right); | |
1659 | return -1; | |
1660 | } | |
1661 | ||
b4ad0c7e | 1662 | bool __init memblock_is_reserved(phys_addr_t addr) |
95f72d1e | 1663 | { |
72d4b0b4 BH |
1664 | return memblock_search(&memblock.reserved, addr) != -1; |
1665 | } | |
95f72d1e | 1666 | |
b4ad0c7e | 1667 | bool __init_memblock memblock_is_memory(phys_addr_t addr) |
72d4b0b4 BH |
1668 | { |
1669 | return memblock_search(&memblock.memory, addr) != -1; | |
1670 | } | |
1671 | ||
937f0c26 | 1672 | bool __init_memblock memblock_is_map_memory(phys_addr_t addr) |
bf3d3cc5 AB |
1673 | { |
1674 | int i = memblock_search(&memblock.memory, addr); | |
1675 | ||
1676 | if (i == -1) | |
1677 | return false; | |
1678 | return !memblock_is_nomap(&memblock.memory.regions[i]); | |
1679 | } | |
1680 | ||
e76b63f8 YL |
1681 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1682 | int __init_memblock memblock_search_pfn_nid(unsigned long pfn, | |
1683 | unsigned long *start_pfn, unsigned long *end_pfn) | |
1684 | { | |
1685 | struct memblock_type *type = &memblock.memory; | |
16763230 | 1686 | int mid = memblock_search(type, PFN_PHYS(pfn)); |
e76b63f8 YL |
1687 | |
1688 | if (mid == -1) | |
1689 | return -1; | |
1690 | ||
f7e2f7e8 FF |
1691 | *start_pfn = PFN_DOWN(type->regions[mid].base); |
1692 | *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); | |
e76b63f8 YL |
1693 | |
1694 | return type->regions[mid].nid; | |
1695 | } | |
1696 | #endif | |
1697 | ||
eab30949 SB |
1698 | /** |
1699 | * memblock_is_region_memory - check if a region is a subset of memory | |
1700 | * @base: base of region to check | |
1701 | * @size: size of region to check | |
1702 | * | |
47cec443 | 1703 | * Check if the region [@base, @base + @size) is a subset of a memory block. |
eab30949 | 1704 | * |
47cec443 | 1705 | * Return: |
eab30949 SB |
1706 | * 0 if false, non-zero if true |
1707 | */ | |
937f0c26 | 1708 | bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) |
72d4b0b4 | 1709 | { |
abb65272 | 1710 | int idx = memblock_search(&memblock.memory, base); |
eb18f1b5 | 1711 | phys_addr_t end = base + memblock_cap_size(base, &size); |
72d4b0b4 BH |
1712 | |
1713 | if (idx == -1) | |
937f0c26 | 1714 | return false; |
ef415ef4 | 1715 | return (memblock.memory.regions[idx].base + |
eb18f1b5 | 1716 | memblock.memory.regions[idx].size) >= end; |
95f72d1e YL |
1717 | } |
1718 | ||
eab30949 SB |
1719 | /** |
1720 | * memblock_is_region_reserved - check if a region intersects reserved memory | |
1721 | * @base: base of region to check | |
1722 | * @size: size of region to check | |
1723 | * | |
47cec443 MR |
1724 | * Check if the region [@base, @base + @size) intersects a reserved |
1725 | * memory block. | |
eab30949 | 1726 | * |
47cec443 | 1727 | * Return: |
c5c5c9d1 | 1728 | * True if they intersect, false if not. |
eab30949 | 1729 | */ |
c5c5c9d1 | 1730 | bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) |
95f72d1e | 1731 | { |
eb18f1b5 | 1732 | memblock_cap_size(base, &size); |
c5c5c9d1 | 1733 | return memblock_overlaps_region(&memblock.reserved, base, size); |
95f72d1e YL |
1734 | } |
1735 | ||
6ede1fd3 YL |
1736 | void __init_memblock memblock_trim_memory(phys_addr_t align) |
1737 | { | |
6ede1fd3 | 1738 | phys_addr_t start, end, orig_start, orig_end; |
136199f0 | 1739 | struct memblock_region *r; |
6ede1fd3 | 1740 | |
136199f0 EM |
1741 | for_each_memblock(memory, r) { |
1742 | orig_start = r->base; | |
1743 | orig_end = r->base + r->size; | |
6ede1fd3 YL |
1744 | start = round_up(orig_start, align); |
1745 | end = round_down(orig_end, align); | |
1746 | ||
1747 | if (start == orig_start && end == orig_end) | |
1748 | continue; | |
1749 | ||
1750 | if (start < end) { | |
136199f0 EM |
1751 | r->base = start; |
1752 | r->size = end - start; | |
6ede1fd3 | 1753 | } else { |
136199f0 EM |
1754 | memblock_remove_region(&memblock.memory, |
1755 | r - memblock.memory.regions); | |
1756 | r--; | |
6ede1fd3 YL |
1757 | } |
1758 | } | |
1759 | } | |
e63075a3 | 1760 | |
3661ca66 | 1761 | void __init_memblock memblock_set_current_limit(phys_addr_t limit) |
e63075a3 BH |
1762 | { |
1763 | memblock.current_limit = limit; | |
1764 | } | |
1765 | ||
fec51014 LA |
1766 | phys_addr_t __init_memblock memblock_get_current_limit(void) |
1767 | { | |
1768 | return memblock.current_limit; | |
1769 | } | |
1770 | ||
0262d9c8 | 1771 | static void __init_memblock memblock_dump(struct memblock_type *type) |
6ed311b2 | 1772 | { |
5d63f81c | 1773 | phys_addr_t base, end, size; |
e1720fee | 1774 | enum memblock_flags flags; |
8c9c1701 AK |
1775 | int idx; |
1776 | struct memblock_region *rgn; | |
6ed311b2 | 1777 | |
0262d9c8 | 1778 | pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); |
6ed311b2 | 1779 | |
66e8b438 | 1780 | for_each_memblock_type(idx, type, rgn) { |
7c0caeb8 TH |
1781 | char nid_buf[32] = ""; |
1782 | ||
1783 | base = rgn->base; | |
1784 | size = rgn->size; | |
5d63f81c | 1785 | end = base + size - 1; |
66a20757 | 1786 | flags = rgn->flags; |
7c0caeb8 TH |
1787 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1788 | if (memblock_get_region_node(rgn) != MAX_NUMNODES) | |
1789 | snprintf(nid_buf, sizeof(nid_buf), " on node %d", | |
1790 | memblock_get_region_node(rgn)); | |
1791 | #endif | |
e1720fee | 1792 | pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#x\n", |
0262d9c8 | 1793 | type->name, idx, &base, &end, &size, nid_buf, flags); |
6ed311b2 BH |
1794 | } |
1795 | } | |
1796 | ||
4ff7b82f | 1797 | void __init_memblock __memblock_dump_all(void) |
6ed311b2 | 1798 | { |
6ed311b2 | 1799 | pr_info("MEMBLOCK configuration:\n"); |
5d63f81c MC |
1800 | pr_info(" memory size = %pa reserved size = %pa\n", |
1801 | &memblock.memory.total_size, | |
1802 | &memblock.reserved.total_size); | |
6ed311b2 | 1803 | |
0262d9c8 HC |
1804 | memblock_dump(&memblock.memory); |
1805 | memblock_dump(&memblock.reserved); | |
409efd4c | 1806 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
0262d9c8 | 1807 | memblock_dump(&memblock.physmem); |
409efd4c | 1808 | #endif |
6ed311b2 BH |
1809 | } |
1810 | ||
1aadc056 | 1811 | void __init memblock_allow_resize(void) |
6ed311b2 | 1812 | { |
142b45a7 | 1813 | memblock_can_resize = 1; |
6ed311b2 BH |
1814 | } |
1815 | ||
6ed311b2 BH |
1816 | static int __init early_memblock(char *p) |
1817 | { | |
1818 | if (p && strstr(p, "debug")) | |
1819 | memblock_debug = 1; | |
1820 | return 0; | |
1821 | } | |
1822 | early_param("memblock", early_memblock); | |
1823 | ||
c378ddd5 | 1824 | #if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK) |
6d03b885 BH |
1825 | |
1826 | static int memblock_debug_show(struct seq_file *m, void *private) | |
1827 | { | |
1828 | struct memblock_type *type = m->private; | |
1829 | struct memblock_region *reg; | |
1830 | int i; | |
5d63f81c | 1831 | phys_addr_t end; |
6d03b885 BH |
1832 | |
1833 | for (i = 0; i < type->cnt; i++) { | |
1834 | reg = &type->regions[i]; | |
5d63f81c | 1835 | end = reg->base + reg->size - 1; |
6d03b885 | 1836 | |
5d63f81c MC |
1837 | seq_printf(m, "%4d: ", i); |
1838 | seq_printf(m, "%pa..%pa\n", ®->base, &end); | |
6d03b885 BH |
1839 | } |
1840 | return 0; | |
1841 | } | |
5ad35093 | 1842 | DEFINE_SHOW_ATTRIBUTE(memblock_debug); |
6d03b885 BH |
1843 | |
1844 | static int __init memblock_init_debugfs(void) | |
1845 | { | |
1846 | struct dentry *root = debugfs_create_dir("memblock", NULL); | |
1847 | if (!root) | |
1848 | return -ENXIO; | |
0825a6f9 JP |
1849 | debugfs_create_file("memory", 0444, root, |
1850 | &memblock.memory, &memblock_debug_fops); | |
1851 | debugfs_create_file("reserved", 0444, root, | |
1852 | &memblock.reserved, &memblock_debug_fops); | |
70210ed9 | 1853 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
0825a6f9 JP |
1854 | debugfs_create_file("physmem", 0444, root, |
1855 | &memblock.physmem, &memblock_debug_fops); | |
70210ed9 | 1856 | #endif |
6d03b885 BH |
1857 | |
1858 | return 0; | |
1859 | } | |
1860 | __initcall(memblock_init_debugfs); | |
1861 | ||
1862 | #endif /* CONFIG_DEBUG_FS */ |