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 | |
a3f5bafc TL |
63 | ulong __init_memblock choose_memblock_flags(void) |
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 | ||
79442ed1 TC |
95 | /* |
96 | * __memblock_find_range_bottom_up - find free area utility in bottom-up | |
97 | * @start: start of candidate range | |
98 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE} | |
99 | * @size: size of free area to find | |
100 | * @align: alignment of free area to find | |
b1154233 | 101 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 102 | * @flags: pick from blocks based on memory attributes |
79442ed1 TC |
103 | * |
104 | * Utility called from memblock_find_in_range_node(), find free area bottom-up. | |
105 | * | |
106 | * RETURNS: | |
107 | * Found address on success, 0 on failure. | |
108 | */ | |
109 | static phys_addr_t __init_memblock | |
110 | __memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 TL |
111 | phys_addr_t size, phys_addr_t align, int nid, |
112 | ulong flags) | |
79442ed1 TC |
113 | { |
114 | phys_addr_t this_start, this_end, cand; | |
115 | u64 i; | |
116 | ||
fc6daaf9 | 117 | for_each_free_mem_range(i, nid, flags, &this_start, &this_end, NULL) { |
79442ed1 TC |
118 | this_start = clamp(this_start, start, end); |
119 | this_end = clamp(this_end, start, end); | |
120 | ||
121 | cand = round_up(this_start, align); | |
122 | if (cand < this_end && this_end - cand >= size) | |
123 | return cand; | |
124 | } | |
125 | ||
126 | return 0; | |
127 | } | |
128 | ||
7bd0b0f0 | 129 | /** |
1402899e | 130 | * __memblock_find_range_top_down - find free area utility, in top-down |
7bd0b0f0 TH |
131 | * @start: start of candidate range |
132 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE} | |
133 | * @size: size of free area to find | |
134 | * @align: alignment of free area to find | |
b1154233 | 135 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 136 | * @flags: pick from blocks based on memory attributes |
7bd0b0f0 | 137 | * |
1402899e | 138 | * Utility called from memblock_find_in_range_node(), find free area top-down. |
7bd0b0f0 TH |
139 | * |
140 | * RETURNS: | |
79442ed1 | 141 | * Found address on success, 0 on failure. |
6ed311b2 | 142 | */ |
1402899e TC |
143 | static phys_addr_t __init_memblock |
144 | __memblock_find_range_top_down(phys_addr_t start, phys_addr_t end, | |
fc6daaf9 TL |
145 | phys_addr_t size, phys_addr_t align, int nid, |
146 | ulong flags) | |
f7210e6c TC |
147 | { |
148 | phys_addr_t this_start, this_end, cand; | |
149 | u64 i; | |
150 | ||
fc6daaf9 TL |
151 | for_each_free_mem_range_reverse(i, nid, flags, &this_start, &this_end, |
152 | NULL) { | |
f7210e6c TC |
153 | this_start = clamp(this_start, start, end); |
154 | this_end = clamp(this_end, start, end); | |
155 | ||
156 | if (this_end < size) | |
157 | continue; | |
158 | ||
159 | cand = round_down(this_end - size, align); | |
160 | if (cand >= this_start) | |
161 | return cand; | |
162 | } | |
1402899e | 163 | |
f7210e6c TC |
164 | return 0; |
165 | } | |
6ed311b2 | 166 | |
1402899e TC |
167 | /** |
168 | * memblock_find_in_range_node - find free area in given range and node | |
1402899e TC |
169 | * @size: size of free area to find |
170 | * @align: alignment of free area to find | |
87029ee9 GS |
171 | * @start: start of candidate range |
172 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE} | |
b1154233 | 173 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node |
fc6daaf9 | 174 | * @flags: pick from blocks based on memory attributes |
1402899e TC |
175 | * |
176 | * Find @size free area aligned to @align in the specified range and node. | |
177 | * | |
79442ed1 TC |
178 | * When allocation direction is bottom-up, the @start should be greater |
179 | * than the end of the kernel image. Otherwise, it will be trimmed. The | |
180 | * reason is that we want the bottom-up allocation just near the kernel | |
181 | * image so it is highly likely that the allocated memory and the kernel | |
182 | * will reside in the same node. | |
183 | * | |
184 | * If bottom-up allocation failed, will try to allocate memory top-down. | |
185 | * | |
1402899e | 186 | * RETURNS: |
79442ed1 | 187 | * Found address on success, 0 on failure. |
1402899e | 188 | */ |
87029ee9 GS |
189 | phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size, |
190 | phys_addr_t align, phys_addr_t start, | |
fc6daaf9 | 191 | phys_addr_t end, int nid, ulong flags) |
1402899e | 192 | { |
0cfb8f0c | 193 | phys_addr_t kernel_end, ret; |
79442ed1 | 194 | |
1402899e TC |
195 | /* pump up @end */ |
196 | if (end == MEMBLOCK_ALLOC_ACCESSIBLE) | |
197 | end = memblock.current_limit; | |
198 | ||
199 | /* avoid allocating the first page */ | |
200 | start = max_t(phys_addr_t, start, PAGE_SIZE); | |
201 | end = max(start, end); | |
79442ed1 TC |
202 | kernel_end = __pa_symbol(_end); |
203 | ||
204 | /* | |
205 | * try bottom-up allocation only when bottom-up mode | |
206 | * is set and @end is above the kernel image. | |
207 | */ | |
208 | if (memblock_bottom_up() && end > kernel_end) { | |
209 | phys_addr_t bottom_up_start; | |
210 | ||
211 | /* make sure we will allocate above the kernel */ | |
212 | bottom_up_start = max(start, kernel_end); | |
213 | ||
214 | /* ok, try bottom-up allocation first */ | |
215 | ret = __memblock_find_range_bottom_up(bottom_up_start, end, | |
fc6daaf9 | 216 | size, align, nid, flags); |
79442ed1 TC |
217 | if (ret) |
218 | return ret; | |
219 | ||
220 | /* | |
221 | * we always limit bottom-up allocation above the kernel, | |
222 | * but top-down allocation doesn't have the limit, so | |
223 | * retrying top-down allocation may succeed when bottom-up | |
224 | * allocation failed. | |
225 | * | |
226 | * bottom-up allocation is expected to be fail very rarely, | |
227 | * so we use WARN_ONCE() here to see the stack trace if | |
228 | * fail happens. | |
229 | */ | |
756a025f | 230 | WARN_ONCE(1, "memblock: bottom-up allocation failed, memory hotunplug may be affected\n"); |
79442ed1 | 231 | } |
1402899e | 232 | |
fc6daaf9 TL |
233 | return __memblock_find_range_top_down(start, end, size, align, nid, |
234 | flags); | |
1402899e TC |
235 | } |
236 | ||
7bd0b0f0 TH |
237 | /** |
238 | * memblock_find_in_range - find free area in given range | |
239 | * @start: start of candidate range | |
240 | * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE} | |
241 | * @size: size of free area to find | |
242 | * @align: alignment of free area to find | |
243 | * | |
244 | * Find @size free area aligned to @align in the specified range. | |
245 | * | |
246 | * RETURNS: | |
79442ed1 | 247 | * Found address on success, 0 on failure. |
fc769a8e | 248 | */ |
7bd0b0f0 TH |
249 | phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start, |
250 | phys_addr_t end, phys_addr_t size, | |
251 | phys_addr_t align) | |
6ed311b2 | 252 | { |
a3f5bafc TL |
253 | phys_addr_t ret; |
254 | ulong flags = choose_memblock_flags(); | |
255 | ||
256 | again: | |
257 | ret = memblock_find_in_range_node(size, align, start, end, | |
258 | NUMA_NO_NODE, flags); | |
259 | ||
260 | if (!ret && (flags & MEMBLOCK_MIRROR)) { | |
261 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
262 | &size); | |
263 | flags &= ~MEMBLOCK_MIRROR; | |
264 | goto again; | |
265 | } | |
266 | ||
267 | return ret; | |
6ed311b2 BH |
268 | } |
269 | ||
10d06439 | 270 | static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r) |
95f72d1e | 271 | { |
1440c4e2 | 272 | type->total_size -= type->regions[r].size; |
7c0caeb8 TH |
273 | memmove(&type->regions[r], &type->regions[r + 1], |
274 | (type->cnt - (r + 1)) * sizeof(type->regions[r])); | |
e3239ff9 | 275 | type->cnt--; |
95f72d1e | 276 | |
8f7a6605 BH |
277 | /* Special case for empty arrays */ |
278 | if (type->cnt == 0) { | |
1440c4e2 | 279 | WARN_ON(type->total_size != 0); |
8f7a6605 BH |
280 | type->cnt = 1; |
281 | type->regions[0].base = 0; | |
282 | type->regions[0].size = 0; | |
66a20757 | 283 | type->regions[0].flags = 0; |
7c0caeb8 | 284 | memblock_set_region_node(&type->regions[0], MAX_NUMNODES); |
8f7a6605 | 285 | } |
95f72d1e YL |
286 | } |
287 | ||
354f17e1 | 288 | #ifdef CONFIG_ARCH_DISCARD_MEMBLOCK |
3010f876 PT |
289 | /** |
290 | * Discard memory and reserved arrays if they were allocated | |
291 | */ | |
292 | void __init memblock_discard(void) | |
5e270e25 | 293 | { |
3010f876 | 294 | phys_addr_t addr, size; |
5e270e25 | 295 | |
3010f876 PT |
296 | if (memblock.reserved.regions != memblock_reserved_init_regions) { |
297 | addr = __pa(memblock.reserved.regions); | |
298 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
299 | memblock.reserved.max); | |
300 | __memblock_free_late(addr, size); | |
301 | } | |
5e270e25 | 302 | |
91b540f9 | 303 | if (memblock.memory.regions != memblock_memory_init_regions) { |
3010f876 PT |
304 | addr = __pa(memblock.memory.regions); |
305 | size = PAGE_ALIGN(sizeof(struct memblock_region) * | |
306 | memblock.memory.max); | |
307 | __memblock_free_late(addr, size); | |
308 | } | |
5e270e25 | 309 | } |
5e270e25 PH |
310 | #endif |
311 | ||
48c3b583 GP |
312 | /** |
313 | * memblock_double_array - double the size of the memblock regions array | |
314 | * @type: memblock type of the regions array being doubled | |
315 | * @new_area_start: starting address of memory range to avoid overlap with | |
316 | * @new_area_size: size of memory range to avoid overlap with | |
317 | * | |
318 | * Double the size of the @type regions array. If memblock is being used to | |
319 | * allocate memory for a new reserved regions array and there is a previously | |
320 | * allocated memory range [@new_area_start,@new_area_start+@new_area_size] | |
321 | * waiting to be reserved, ensure the memory used by the new array does | |
322 | * not overlap. | |
323 | * | |
324 | * RETURNS: | |
325 | * 0 on success, -1 on failure. | |
326 | */ | |
327 | static int __init_memblock memblock_double_array(struct memblock_type *type, | |
328 | phys_addr_t new_area_start, | |
329 | phys_addr_t new_area_size) | |
142b45a7 BH |
330 | { |
331 | struct memblock_region *new_array, *old_array; | |
29f67386 | 332 | phys_addr_t old_alloc_size, new_alloc_size; |
142b45a7 BH |
333 | phys_addr_t old_size, new_size, addr; |
334 | int use_slab = slab_is_available(); | |
181eb394 | 335 | int *in_slab; |
142b45a7 BH |
336 | |
337 | /* We don't allow resizing until we know about the reserved regions | |
338 | * of memory that aren't suitable for allocation | |
339 | */ | |
340 | if (!memblock_can_resize) | |
341 | return -1; | |
342 | ||
142b45a7 BH |
343 | /* Calculate new doubled size */ |
344 | old_size = type->max * sizeof(struct memblock_region); | |
345 | new_size = old_size << 1; | |
29f67386 YL |
346 | /* |
347 | * We need to allocated new one align to PAGE_SIZE, | |
348 | * so we can free them completely later. | |
349 | */ | |
350 | old_alloc_size = PAGE_ALIGN(old_size); | |
351 | new_alloc_size = PAGE_ALIGN(new_size); | |
142b45a7 | 352 | |
181eb394 GS |
353 | /* Retrieve the slab flag */ |
354 | if (type == &memblock.memory) | |
355 | in_slab = &memblock_memory_in_slab; | |
356 | else | |
357 | in_slab = &memblock_reserved_in_slab; | |
358 | ||
142b45a7 BH |
359 | /* Try to find some space for it. |
360 | * | |
361 | * WARNING: We assume that either slab_is_available() and we use it or | |
fd07383b AM |
362 | * we use MEMBLOCK for allocations. That means that this is unsafe to |
363 | * use when bootmem is currently active (unless bootmem itself is | |
364 | * implemented on top of MEMBLOCK which isn't the case yet) | |
142b45a7 BH |
365 | * |
366 | * This should however not be an issue for now, as we currently only | |
fd07383b AM |
367 | * call into MEMBLOCK while it's still active, or much later when slab |
368 | * is active for memory hotplug operations | |
142b45a7 BH |
369 | */ |
370 | if (use_slab) { | |
371 | new_array = kmalloc(new_size, GFP_KERNEL); | |
1f5026a7 | 372 | addr = new_array ? __pa(new_array) : 0; |
4e2f0775 | 373 | } else { |
48c3b583 GP |
374 | /* only exclude range when trying to double reserved.regions */ |
375 | if (type != &memblock.reserved) | |
376 | new_area_start = new_area_size = 0; | |
377 | ||
378 | addr = memblock_find_in_range(new_area_start + new_area_size, | |
379 | memblock.current_limit, | |
29f67386 | 380 | new_alloc_size, PAGE_SIZE); |
48c3b583 GP |
381 | if (!addr && new_area_size) |
382 | addr = memblock_find_in_range(0, | |
fd07383b AM |
383 | min(new_area_start, memblock.current_limit), |
384 | new_alloc_size, PAGE_SIZE); | |
48c3b583 | 385 | |
15674868 | 386 | new_array = addr ? __va(addr) : NULL; |
4e2f0775 | 387 | } |
1f5026a7 | 388 | if (!addr) { |
142b45a7 | 389 | pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n", |
0262d9c8 | 390 | type->name, type->max, type->max * 2); |
142b45a7 BH |
391 | return -1; |
392 | } | |
142b45a7 | 393 | |
fd07383b | 394 | memblock_dbg("memblock: %s is doubled to %ld at [%#010llx-%#010llx]", |
0262d9c8 | 395 | type->name, type->max * 2, (u64)addr, |
fd07383b | 396 | (u64)addr + new_size - 1); |
ea9e4376 | 397 | |
fd07383b AM |
398 | /* |
399 | * Found space, we now need to move the array over before we add the | |
400 | * reserved region since it may be our reserved array itself that is | |
401 | * full. | |
142b45a7 BH |
402 | */ |
403 | memcpy(new_array, type->regions, old_size); | |
404 | memset(new_array + type->max, 0, old_size); | |
405 | old_array = type->regions; | |
406 | type->regions = new_array; | |
407 | type->max <<= 1; | |
408 | ||
fd07383b | 409 | /* Free old array. We needn't free it if the array is the static one */ |
181eb394 GS |
410 | if (*in_slab) |
411 | kfree(old_array); | |
412 | else if (old_array != memblock_memory_init_regions && | |
413 | old_array != memblock_reserved_init_regions) | |
29f67386 | 414 | memblock_free(__pa(old_array), old_alloc_size); |
142b45a7 | 415 | |
fd07383b AM |
416 | /* |
417 | * Reserve the new array if that comes from the memblock. Otherwise, we | |
418 | * needn't do it | |
181eb394 GS |
419 | */ |
420 | if (!use_slab) | |
29f67386 | 421 | BUG_ON(memblock_reserve(addr, new_alloc_size)); |
181eb394 GS |
422 | |
423 | /* Update slab flag */ | |
424 | *in_slab = use_slab; | |
425 | ||
142b45a7 BH |
426 | return 0; |
427 | } | |
428 | ||
784656f9 TH |
429 | /** |
430 | * memblock_merge_regions - merge neighboring compatible regions | |
431 | * @type: memblock type to scan | |
432 | * | |
433 | * Scan @type and merge neighboring compatible regions. | |
434 | */ | |
435 | static void __init_memblock memblock_merge_regions(struct memblock_type *type) | |
95f72d1e | 436 | { |
784656f9 | 437 | int i = 0; |
95f72d1e | 438 | |
784656f9 TH |
439 | /* cnt never goes below 1 */ |
440 | while (i < type->cnt - 1) { | |
441 | struct memblock_region *this = &type->regions[i]; | |
442 | struct memblock_region *next = &type->regions[i + 1]; | |
95f72d1e | 443 | |
7c0caeb8 TH |
444 | if (this->base + this->size != next->base || |
445 | memblock_get_region_node(this) != | |
66a20757 TC |
446 | memblock_get_region_node(next) || |
447 | this->flags != next->flags) { | |
784656f9 TH |
448 | BUG_ON(this->base + this->size > next->base); |
449 | i++; | |
450 | continue; | |
8f7a6605 BH |
451 | } |
452 | ||
784656f9 | 453 | this->size += next->size; |
c0232ae8 LF |
454 | /* move forward from next + 1, index of which is i + 2 */ |
455 | memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next)); | |
784656f9 | 456 | type->cnt--; |
95f72d1e | 457 | } |
784656f9 | 458 | } |
95f72d1e | 459 | |
784656f9 TH |
460 | /** |
461 | * memblock_insert_region - insert new memblock region | |
209ff86d TC |
462 | * @type: memblock type to insert into |
463 | * @idx: index for the insertion point | |
464 | * @base: base address of the new region | |
465 | * @size: size of the new region | |
466 | * @nid: node id of the new region | |
66a20757 | 467 | * @flags: flags of the new region |
784656f9 TH |
468 | * |
469 | * Insert new memblock region [@base,@base+@size) into @type at @idx. | |
412d0008 | 470 | * @type must already have extra room to accommodate the new region. |
784656f9 TH |
471 | */ |
472 | static void __init_memblock memblock_insert_region(struct memblock_type *type, | |
473 | int idx, phys_addr_t base, | |
66a20757 TC |
474 | phys_addr_t size, |
475 | int nid, unsigned long flags) | |
784656f9 TH |
476 | { |
477 | struct memblock_region *rgn = &type->regions[idx]; | |
478 | ||
479 | BUG_ON(type->cnt >= type->max); | |
480 | memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn)); | |
481 | rgn->base = base; | |
482 | rgn->size = size; | |
66a20757 | 483 | rgn->flags = flags; |
7c0caeb8 | 484 | memblock_set_region_node(rgn, nid); |
784656f9 | 485 | type->cnt++; |
1440c4e2 | 486 | type->total_size += size; |
784656f9 TH |
487 | } |
488 | ||
489 | /** | |
f1af9d3a | 490 | * memblock_add_range - add new memblock region |
784656f9 TH |
491 | * @type: memblock type to add new region into |
492 | * @base: base address of the new region | |
493 | * @size: size of the new region | |
7fb0bc3f | 494 | * @nid: nid of the new region |
66a20757 | 495 | * @flags: flags of the new region |
784656f9 TH |
496 | * |
497 | * Add new memblock region [@base,@base+@size) into @type. The new region | |
498 | * is allowed to overlap with existing ones - overlaps don't affect already | |
499 | * existing regions. @type is guaranteed to be minimal (all neighbouring | |
500 | * compatible regions are merged) after the addition. | |
501 | * | |
502 | * RETURNS: | |
503 | * 0 on success, -errno on failure. | |
504 | */ | |
f1af9d3a | 505 | int __init_memblock memblock_add_range(struct memblock_type *type, |
66a20757 TC |
506 | phys_addr_t base, phys_addr_t size, |
507 | int nid, unsigned long flags) | |
784656f9 TH |
508 | { |
509 | bool insert = false; | |
eb18f1b5 TH |
510 | phys_addr_t obase = base; |
511 | phys_addr_t end = base + memblock_cap_size(base, &size); | |
8c9c1701 AK |
512 | int idx, nr_new; |
513 | struct memblock_region *rgn; | |
784656f9 | 514 | |
b3dc627c TH |
515 | if (!size) |
516 | return 0; | |
517 | ||
784656f9 TH |
518 | /* special case for empty array */ |
519 | if (type->regions[0].size == 0) { | |
1440c4e2 | 520 | WARN_ON(type->cnt != 1 || type->total_size); |
8f7a6605 BH |
521 | type->regions[0].base = base; |
522 | type->regions[0].size = size; | |
66a20757 | 523 | type->regions[0].flags = flags; |
7fb0bc3f | 524 | memblock_set_region_node(&type->regions[0], nid); |
1440c4e2 | 525 | type->total_size = size; |
8f7a6605 | 526 | return 0; |
95f72d1e | 527 | } |
784656f9 TH |
528 | repeat: |
529 | /* | |
530 | * The following is executed twice. Once with %false @insert and | |
531 | * then with %true. The first counts the number of regions needed | |
412d0008 | 532 | * to accommodate the new area. The second actually inserts them. |
142b45a7 | 533 | */ |
784656f9 TH |
534 | base = obase; |
535 | nr_new = 0; | |
95f72d1e | 536 | |
66e8b438 | 537 | for_each_memblock_type(idx, type, rgn) { |
784656f9 TH |
538 | phys_addr_t rbase = rgn->base; |
539 | phys_addr_t rend = rbase + rgn->size; | |
540 | ||
541 | if (rbase >= end) | |
95f72d1e | 542 | break; |
784656f9 TH |
543 | if (rend <= base) |
544 | continue; | |
545 | /* | |
546 | * @rgn overlaps. If it separates the lower part of new | |
547 | * area, insert that portion. | |
548 | */ | |
549 | if (rbase > base) { | |
c0a29498 WY |
550 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
551 | WARN_ON(nid != memblock_get_region_node(rgn)); | |
552 | #endif | |
4fcab5f4 | 553 | WARN_ON(flags != rgn->flags); |
784656f9 TH |
554 | nr_new++; |
555 | if (insert) | |
8c9c1701 | 556 | memblock_insert_region(type, idx++, base, |
66a20757 TC |
557 | rbase - base, nid, |
558 | flags); | |
95f72d1e | 559 | } |
784656f9 TH |
560 | /* area below @rend is dealt with, forget about it */ |
561 | base = min(rend, end); | |
95f72d1e | 562 | } |
784656f9 TH |
563 | |
564 | /* insert the remaining portion */ | |
565 | if (base < end) { | |
566 | nr_new++; | |
567 | if (insert) | |
8c9c1701 | 568 | memblock_insert_region(type, idx, base, end - base, |
66a20757 | 569 | nid, flags); |
95f72d1e | 570 | } |
95f72d1e | 571 | |
ef3cc4db | 572 | if (!nr_new) |
573 | return 0; | |
574 | ||
784656f9 TH |
575 | /* |
576 | * If this was the first round, resize array and repeat for actual | |
577 | * insertions; otherwise, merge and return. | |
142b45a7 | 578 | */ |
784656f9 TH |
579 | if (!insert) { |
580 | while (type->cnt + nr_new > type->max) | |
48c3b583 | 581 | if (memblock_double_array(type, obase, size) < 0) |
784656f9 TH |
582 | return -ENOMEM; |
583 | insert = true; | |
584 | goto repeat; | |
585 | } else { | |
586 | memblock_merge_regions(type); | |
587 | return 0; | |
142b45a7 | 588 | } |
95f72d1e YL |
589 | } |
590 | ||
7fb0bc3f TH |
591 | int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size, |
592 | int nid) | |
593 | { | |
f1af9d3a | 594 | return memblock_add_range(&memblock.memory, base, size, nid, 0); |
7fb0bc3f TH |
595 | } |
596 | ||
f705ac4b | 597 | int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size) |
6a4055bc | 598 | { |
5d63f81c MC |
599 | phys_addr_t end = base + size - 1; |
600 | ||
601 | memblock_dbg("memblock_add: [%pa-%pa] %pF\n", | |
602 | &base, &end, (void *)_RET_IP_); | |
6a4055bc | 603 | |
f705ac4b | 604 | return memblock_add_range(&memblock.memory, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
605 | } |
606 | ||
6a9ceb31 TH |
607 | /** |
608 | * memblock_isolate_range - isolate given range into disjoint memblocks | |
609 | * @type: memblock type to isolate range for | |
610 | * @base: base of range to isolate | |
611 | * @size: size of range to isolate | |
612 | * @start_rgn: out parameter for the start of isolated region | |
613 | * @end_rgn: out parameter for the end of isolated region | |
614 | * | |
615 | * Walk @type and ensure that regions don't cross the boundaries defined by | |
616 | * [@base,@base+@size). Crossing regions are split at the boundaries, | |
617 | * which may create at most two more regions. The index of the first | |
618 | * region inside the range is returned in *@start_rgn and end in *@end_rgn. | |
619 | * | |
620 | * RETURNS: | |
621 | * 0 on success, -errno on failure. | |
622 | */ | |
623 | static int __init_memblock memblock_isolate_range(struct memblock_type *type, | |
624 | phys_addr_t base, phys_addr_t size, | |
625 | int *start_rgn, int *end_rgn) | |
626 | { | |
eb18f1b5 | 627 | phys_addr_t end = base + memblock_cap_size(base, &size); |
8c9c1701 AK |
628 | int idx; |
629 | struct memblock_region *rgn; | |
6a9ceb31 TH |
630 | |
631 | *start_rgn = *end_rgn = 0; | |
632 | ||
b3dc627c TH |
633 | if (!size) |
634 | return 0; | |
635 | ||
6a9ceb31 TH |
636 | /* we'll create at most two more regions */ |
637 | while (type->cnt + 2 > type->max) | |
48c3b583 | 638 | if (memblock_double_array(type, base, size) < 0) |
6a9ceb31 TH |
639 | return -ENOMEM; |
640 | ||
66e8b438 | 641 | for_each_memblock_type(idx, type, rgn) { |
6a9ceb31 TH |
642 | phys_addr_t rbase = rgn->base; |
643 | phys_addr_t rend = rbase + rgn->size; | |
644 | ||
645 | if (rbase >= end) | |
646 | break; | |
647 | if (rend <= base) | |
648 | continue; | |
649 | ||
650 | if (rbase < base) { | |
651 | /* | |
652 | * @rgn intersects from below. Split and continue | |
653 | * to process the next region - the new top half. | |
654 | */ | |
655 | rgn->base = base; | |
1440c4e2 TH |
656 | rgn->size -= base - rbase; |
657 | type->total_size -= base - rbase; | |
8c9c1701 | 658 | memblock_insert_region(type, idx, rbase, base - rbase, |
66a20757 TC |
659 | memblock_get_region_node(rgn), |
660 | rgn->flags); | |
6a9ceb31 TH |
661 | } else if (rend > end) { |
662 | /* | |
663 | * @rgn intersects from above. Split and redo the | |
664 | * current region - the new bottom half. | |
665 | */ | |
666 | rgn->base = end; | |
1440c4e2 TH |
667 | rgn->size -= end - rbase; |
668 | type->total_size -= end - rbase; | |
8c9c1701 | 669 | memblock_insert_region(type, idx--, rbase, end - rbase, |
66a20757 TC |
670 | memblock_get_region_node(rgn), |
671 | rgn->flags); | |
6a9ceb31 TH |
672 | } else { |
673 | /* @rgn is fully contained, record it */ | |
674 | if (!*end_rgn) | |
8c9c1701 AK |
675 | *start_rgn = idx; |
676 | *end_rgn = idx + 1; | |
6a9ceb31 TH |
677 | } |
678 | } | |
679 | ||
680 | return 0; | |
681 | } | |
6a9ceb31 | 682 | |
35bd16a2 | 683 | static int __init_memblock memblock_remove_range(struct memblock_type *type, |
f1af9d3a | 684 | phys_addr_t base, phys_addr_t size) |
95f72d1e | 685 | { |
71936180 TH |
686 | int start_rgn, end_rgn; |
687 | int i, ret; | |
95f72d1e | 688 | |
71936180 TH |
689 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
690 | if (ret) | |
691 | return ret; | |
95f72d1e | 692 | |
71936180 TH |
693 | for (i = end_rgn - 1; i >= start_rgn; i--) |
694 | memblock_remove_region(type, i); | |
8f7a6605 | 695 | return 0; |
95f72d1e YL |
696 | } |
697 | ||
581adcbe | 698 | int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size) |
95f72d1e | 699 | { |
25cf23d7 MK |
700 | phys_addr_t end = base + size - 1; |
701 | ||
702 | memblock_dbg("memblock_remove: [%pa-%pa] %pS\n", | |
703 | &base, &end, (void *)_RET_IP_); | |
704 | ||
f1af9d3a | 705 | return memblock_remove_range(&memblock.memory, base, size); |
95f72d1e YL |
706 | } |
707 | ||
f1af9d3a | 708 | |
581adcbe | 709 | int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size) |
95f72d1e | 710 | { |
5d63f81c MC |
711 | phys_addr_t end = base + size - 1; |
712 | ||
713 | memblock_dbg(" memblock_free: [%pa-%pa] %pF\n", | |
714 | &base, &end, (void *)_RET_IP_); | |
24aa0788 | 715 | |
9099daed | 716 | kmemleak_free_part_phys(base, size); |
f1af9d3a | 717 | return memblock_remove_range(&memblock.reserved, base, size); |
95f72d1e YL |
718 | } |
719 | ||
f705ac4b | 720 | int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size) |
95f72d1e | 721 | { |
5d63f81c MC |
722 | phys_addr_t end = base + size - 1; |
723 | ||
724 | memblock_dbg("memblock_reserve: [%pa-%pa] %pF\n", | |
725 | &base, &end, (void *)_RET_IP_); | |
95f72d1e | 726 | |
f705ac4b | 727 | return memblock_add_range(&memblock.reserved, base, size, MAX_NUMNODES, 0); |
95f72d1e YL |
728 | } |
729 | ||
66b16edf | 730 | /** |
66b16edf | 731 | * |
4308ce17 | 732 | * This function isolates region [@base, @base + @size), and sets/clears flag |
66b16edf | 733 | * |
c1153931 | 734 | * Return 0 on success, -errno on failure. |
66b16edf | 735 | */ |
4308ce17 TL |
736 | static int __init_memblock memblock_setclr_flag(phys_addr_t base, |
737 | phys_addr_t size, int set, int flag) | |
66b16edf TC |
738 | { |
739 | struct memblock_type *type = &memblock.memory; | |
740 | int i, ret, start_rgn, end_rgn; | |
741 | ||
742 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); | |
743 | if (ret) | |
744 | return ret; | |
745 | ||
746 | for (i = start_rgn; i < end_rgn; i++) | |
4308ce17 TL |
747 | if (set) |
748 | memblock_set_region_flags(&type->regions[i], flag); | |
749 | else | |
750 | memblock_clear_region_flags(&type->regions[i], flag); | |
66b16edf TC |
751 | |
752 | memblock_merge_regions(type); | |
753 | return 0; | |
754 | } | |
755 | ||
756 | /** | |
4308ce17 | 757 | * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG. |
66b16edf TC |
758 | * @base: the base phys addr of the region |
759 | * @size: the size of the region | |
760 | * | |
c1153931 | 761 | * Return 0 on success, -errno on failure. |
4308ce17 TL |
762 | */ |
763 | int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size) | |
764 | { | |
765 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_HOTPLUG); | |
766 | } | |
767 | ||
768 | /** | |
769 | * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region. | |
770 | * @base: the base phys addr of the region | |
771 | * @size: the size of the region | |
66b16edf | 772 | * |
c1153931 | 773 | * Return 0 on success, -errno on failure. |
66b16edf TC |
774 | */ |
775 | int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size) | |
776 | { | |
4308ce17 | 777 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_HOTPLUG); |
66b16edf TC |
778 | } |
779 | ||
a3f5bafc TL |
780 | /** |
781 | * memblock_mark_mirror - Mark mirrored memory with flag MEMBLOCK_MIRROR. | |
782 | * @base: the base phys addr of the region | |
783 | * @size: the size of the region | |
784 | * | |
c1153931 | 785 | * Return 0 on success, -errno on failure. |
a3f5bafc TL |
786 | */ |
787 | int __init_memblock memblock_mark_mirror(phys_addr_t base, phys_addr_t size) | |
788 | { | |
789 | system_has_some_mirror = true; | |
790 | ||
791 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_MIRROR); | |
792 | } | |
793 | ||
bf3d3cc5 AB |
794 | /** |
795 | * memblock_mark_nomap - Mark a memory region with flag MEMBLOCK_NOMAP. | |
796 | * @base: the base phys addr of the region | |
797 | * @size: the size of the region | |
798 | * | |
799 | * Return 0 on success, -errno on failure. | |
800 | */ | |
801 | int __init_memblock memblock_mark_nomap(phys_addr_t base, phys_addr_t size) | |
802 | { | |
803 | return memblock_setclr_flag(base, size, 1, MEMBLOCK_NOMAP); | |
804 | } | |
a3f5bafc | 805 | |
4c546b8a AT |
806 | /** |
807 | * memblock_clear_nomap - Clear flag MEMBLOCK_NOMAP for a specified region. | |
808 | * @base: the base phys addr of the region | |
809 | * @size: the size of the region | |
810 | * | |
811 | * Return 0 on success, -errno on failure. | |
812 | */ | |
813 | int __init_memblock memblock_clear_nomap(phys_addr_t base, phys_addr_t size) | |
814 | { | |
815 | return memblock_setclr_flag(base, size, 0, MEMBLOCK_NOMAP); | |
816 | } | |
817 | ||
8e7a7f86 RH |
818 | /** |
819 | * __next_reserved_mem_region - next function for for_each_reserved_region() | |
820 | * @idx: pointer to u64 loop variable | |
821 | * @out_start: ptr to phys_addr_t for start address of the region, can be %NULL | |
822 | * @out_end: ptr to phys_addr_t for end address of the region, can be %NULL | |
823 | * | |
824 | * Iterate over all reserved memory regions. | |
825 | */ | |
826 | void __init_memblock __next_reserved_mem_region(u64 *idx, | |
827 | phys_addr_t *out_start, | |
828 | phys_addr_t *out_end) | |
829 | { | |
567d117b | 830 | struct memblock_type *type = &memblock.reserved; |
8e7a7f86 | 831 | |
cd33a76b | 832 | if (*idx < type->cnt) { |
567d117b | 833 | struct memblock_region *r = &type->regions[*idx]; |
8e7a7f86 RH |
834 | phys_addr_t base = r->base; |
835 | phys_addr_t size = r->size; | |
836 | ||
837 | if (out_start) | |
838 | *out_start = base; | |
839 | if (out_end) | |
840 | *out_end = base + size - 1; | |
841 | ||
842 | *idx += 1; | |
843 | return; | |
844 | } | |
845 | ||
846 | /* signal end of iteration */ | |
847 | *idx = ULLONG_MAX; | |
848 | } | |
849 | ||
35fd0808 | 850 | /** |
f1af9d3a | 851 | * __next__mem_range - next function for for_each_free_mem_range() etc. |
35fd0808 | 852 | * @idx: pointer to u64 loop variable |
b1154233 | 853 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 854 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
855 | * @type_a: pointer to memblock_type from where the range is taken |
856 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
857 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
858 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
859 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
35fd0808 | 860 | * |
f1af9d3a | 861 | * Find the first area from *@idx which matches @nid, fill the out |
35fd0808 | 862 | * parameters, and update *@idx for the next iteration. The lower 32bit of |
f1af9d3a PH |
863 | * *@idx contains index into type_a and the upper 32bit indexes the |
864 | * areas before each region in type_b. For example, if type_b regions | |
35fd0808 TH |
865 | * look like the following, |
866 | * | |
867 | * 0:[0-16), 1:[32-48), 2:[128-130) | |
868 | * | |
869 | * The upper 32bit indexes the following regions. | |
870 | * | |
871 | * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX) | |
872 | * | |
873 | * As both region arrays are sorted, the function advances the two indices | |
874 | * in lockstep and returns each intersection. | |
875 | */ | |
fc6daaf9 | 876 | void __init_memblock __next_mem_range(u64 *idx, int nid, ulong flags, |
f1af9d3a PH |
877 | struct memblock_type *type_a, |
878 | struct memblock_type *type_b, | |
879 | phys_addr_t *out_start, | |
880 | phys_addr_t *out_end, int *out_nid) | |
35fd0808 | 881 | { |
f1af9d3a PH |
882 | int idx_a = *idx & 0xffffffff; |
883 | int idx_b = *idx >> 32; | |
b1154233 | 884 | |
f1af9d3a PH |
885 | if (WARN_ONCE(nid == MAX_NUMNODES, |
886 | "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) | |
560dca27 | 887 | nid = NUMA_NO_NODE; |
35fd0808 | 888 | |
f1af9d3a PH |
889 | for (; idx_a < type_a->cnt; idx_a++) { |
890 | struct memblock_region *m = &type_a->regions[idx_a]; | |
891 | ||
35fd0808 TH |
892 | phys_addr_t m_start = m->base; |
893 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 894 | int m_nid = memblock_get_region_node(m); |
35fd0808 TH |
895 | |
896 | /* only memory regions are associated with nodes, check it */ | |
f1af9d3a | 897 | if (nid != NUMA_NO_NODE && nid != m_nid) |
35fd0808 TH |
898 | continue; |
899 | ||
0a313a99 XQ |
900 | /* skip hotpluggable memory regions if needed */ |
901 | if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) | |
902 | continue; | |
903 | ||
a3f5bafc TL |
904 | /* if we want mirror memory skip non-mirror memory regions */ |
905 | if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) | |
906 | continue; | |
907 | ||
bf3d3cc5 AB |
908 | /* skip nomap memory unless we were asked for it explicitly */ |
909 | if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) | |
910 | continue; | |
911 | ||
f1af9d3a PH |
912 | if (!type_b) { |
913 | if (out_start) | |
914 | *out_start = m_start; | |
915 | if (out_end) | |
916 | *out_end = m_end; | |
917 | if (out_nid) | |
918 | *out_nid = m_nid; | |
919 | idx_a++; | |
920 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
921 | return; | |
922 | } | |
923 | ||
924 | /* scan areas before each reservation */ | |
925 | for (; idx_b < type_b->cnt + 1; idx_b++) { | |
926 | struct memblock_region *r; | |
927 | phys_addr_t r_start; | |
928 | phys_addr_t r_end; | |
929 | ||
930 | r = &type_b->regions[idx_b]; | |
931 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
932 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 933 | r->base : PHYS_ADDR_MAX; |
35fd0808 | 934 | |
f1af9d3a PH |
935 | /* |
936 | * if idx_b advanced past idx_a, | |
937 | * break out to advance idx_a | |
938 | */ | |
35fd0808 TH |
939 | if (r_start >= m_end) |
940 | break; | |
941 | /* if the two regions intersect, we're done */ | |
942 | if (m_start < r_end) { | |
943 | if (out_start) | |
f1af9d3a PH |
944 | *out_start = |
945 | max(m_start, r_start); | |
35fd0808 TH |
946 | if (out_end) |
947 | *out_end = min(m_end, r_end); | |
948 | if (out_nid) | |
f1af9d3a | 949 | *out_nid = m_nid; |
35fd0808 | 950 | /* |
f1af9d3a PH |
951 | * The region which ends first is |
952 | * advanced for the next iteration. | |
35fd0808 TH |
953 | */ |
954 | if (m_end <= r_end) | |
f1af9d3a | 955 | idx_a++; |
35fd0808 | 956 | else |
f1af9d3a PH |
957 | idx_b++; |
958 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
35fd0808 TH |
959 | return; |
960 | } | |
961 | } | |
962 | } | |
963 | ||
964 | /* signal end of iteration */ | |
965 | *idx = ULLONG_MAX; | |
966 | } | |
967 | ||
7bd0b0f0 | 968 | /** |
f1af9d3a PH |
969 | * __next_mem_range_rev - generic next function for for_each_*_range_rev() |
970 | * | |
971 | * Finds the next range from type_a which is not marked as unsuitable | |
972 | * in type_b. | |
973 | * | |
7bd0b0f0 | 974 | * @idx: pointer to u64 loop variable |
ad5ea8cd | 975 | * @nid: node selector, %NUMA_NO_NODE for all nodes |
fc6daaf9 | 976 | * @flags: pick from blocks based on memory attributes |
f1af9d3a PH |
977 | * @type_a: pointer to memblock_type from where the range is taken |
978 | * @type_b: pointer to memblock_type which excludes memory from being taken | |
dad7557e WL |
979 | * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL |
980 | * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL | |
981 | * @out_nid: ptr to int for nid of the range, can be %NULL | |
7bd0b0f0 | 982 | * |
f1af9d3a | 983 | * Reverse of __next_mem_range(). |
7bd0b0f0 | 984 | */ |
fc6daaf9 | 985 | void __init_memblock __next_mem_range_rev(u64 *idx, int nid, ulong flags, |
f1af9d3a PH |
986 | struct memblock_type *type_a, |
987 | struct memblock_type *type_b, | |
988 | phys_addr_t *out_start, | |
989 | phys_addr_t *out_end, int *out_nid) | |
7bd0b0f0 | 990 | { |
f1af9d3a PH |
991 | int idx_a = *idx & 0xffffffff; |
992 | int idx_b = *idx >> 32; | |
b1154233 | 993 | |
560dca27 GS |
994 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
995 | nid = NUMA_NO_NODE; | |
7bd0b0f0 TH |
996 | |
997 | if (*idx == (u64)ULLONG_MAX) { | |
f1af9d3a | 998 | idx_a = type_a->cnt - 1; |
e47608ab | 999 | if (type_b != NULL) |
1000 | idx_b = type_b->cnt; | |
1001 | else | |
1002 | idx_b = 0; | |
7bd0b0f0 TH |
1003 | } |
1004 | ||
f1af9d3a PH |
1005 | for (; idx_a >= 0; idx_a--) { |
1006 | struct memblock_region *m = &type_a->regions[idx_a]; | |
1007 | ||
7bd0b0f0 TH |
1008 | phys_addr_t m_start = m->base; |
1009 | phys_addr_t m_end = m->base + m->size; | |
f1af9d3a | 1010 | int m_nid = memblock_get_region_node(m); |
7bd0b0f0 TH |
1011 | |
1012 | /* only memory regions are associated with nodes, check it */ | |
f1af9d3a | 1013 | if (nid != NUMA_NO_NODE && nid != m_nid) |
7bd0b0f0 TH |
1014 | continue; |
1015 | ||
55ac590c TC |
1016 | /* skip hotpluggable memory regions if needed */ |
1017 | if (movable_node_is_enabled() && memblock_is_hotpluggable(m)) | |
1018 | continue; | |
1019 | ||
a3f5bafc TL |
1020 | /* if we want mirror memory skip non-mirror memory regions */ |
1021 | if ((flags & MEMBLOCK_MIRROR) && !memblock_is_mirror(m)) | |
1022 | continue; | |
1023 | ||
bf3d3cc5 AB |
1024 | /* skip nomap memory unless we were asked for it explicitly */ |
1025 | if (!(flags & MEMBLOCK_NOMAP) && memblock_is_nomap(m)) | |
1026 | continue; | |
1027 | ||
f1af9d3a PH |
1028 | if (!type_b) { |
1029 | if (out_start) | |
1030 | *out_start = m_start; | |
1031 | if (out_end) | |
1032 | *out_end = m_end; | |
1033 | if (out_nid) | |
1034 | *out_nid = m_nid; | |
fb399b48 | 1035 | idx_a--; |
f1af9d3a PH |
1036 | *idx = (u32)idx_a | (u64)idx_b << 32; |
1037 | return; | |
1038 | } | |
1039 | ||
1040 | /* scan areas before each reservation */ | |
1041 | for (; idx_b >= 0; idx_b--) { | |
1042 | struct memblock_region *r; | |
1043 | phys_addr_t r_start; | |
1044 | phys_addr_t r_end; | |
1045 | ||
1046 | r = &type_b->regions[idx_b]; | |
1047 | r_start = idx_b ? r[-1].base + r[-1].size : 0; | |
1048 | r_end = idx_b < type_b->cnt ? | |
1c4bc43d | 1049 | r->base : PHYS_ADDR_MAX; |
f1af9d3a PH |
1050 | /* |
1051 | * if idx_b advanced past idx_a, | |
1052 | * break out to advance idx_a | |
1053 | */ | |
7bd0b0f0 | 1054 | |
7bd0b0f0 TH |
1055 | if (r_end <= m_start) |
1056 | break; | |
1057 | /* if the two regions intersect, we're done */ | |
1058 | if (m_end > r_start) { | |
1059 | if (out_start) | |
1060 | *out_start = max(m_start, r_start); | |
1061 | if (out_end) | |
1062 | *out_end = min(m_end, r_end); | |
1063 | if (out_nid) | |
f1af9d3a | 1064 | *out_nid = m_nid; |
7bd0b0f0 | 1065 | if (m_start >= r_start) |
f1af9d3a | 1066 | idx_a--; |
7bd0b0f0 | 1067 | else |
f1af9d3a PH |
1068 | idx_b--; |
1069 | *idx = (u32)idx_a | (u64)idx_b << 32; | |
7bd0b0f0 TH |
1070 | return; |
1071 | } | |
1072 | } | |
1073 | } | |
f1af9d3a | 1074 | /* signal end of iteration */ |
7bd0b0f0 TH |
1075 | *idx = ULLONG_MAX; |
1076 | } | |
1077 | ||
7c0caeb8 TH |
1078 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1079 | /* | |
1080 | * Common iterator interface used to define for_each_mem_range(). | |
1081 | */ | |
1082 | void __init_memblock __next_mem_pfn_range(int *idx, int nid, | |
1083 | unsigned long *out_start_pfn, | |
1084 | unsigned long *out_end_pfn, int *out_nid) | |
1085 | { | |
1086 | struct memblock_type *type = &memblock.memory; | |
1087 | struct memblock_region *r; | |
1088 | ||
1089 | while (++*idx < type->cnt) { | |
1090 | r = &type->regions[*idx]; | |
1091 | ||
1092 | if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size)) | |
1093 | continue; | |
1094 | if (nid == MAX_NUMNODES || nid == r->nid) | |
1095 | break; | |
1096 | } | |
1097 | if (*idx >= type->cnt) { | |
1098 | *idx = -1; | |
1099 | return; | |
1100 | } | |
1101 | ||
1102 | if (out_start_pfn) | |
1103 | *out_start_pfn = PFN_UP(r->base); | |
1104 | if (out_end_pfn) | |
1105 | *out_end_pfn = PFN_DOWN(r->base + r->size); | |
1106 | if (out_nid) | |
1107 | *out_nid = r->nid; | |
1108 | } | |
1109 | ||
1110 | /** | |
1111 | * memblock_set_node - set node ID on memblock regions | |
1112 | * @base: base of area to set node ID for | |
1113 | * @size: size of area to set node ID for | |
e7e8de59 | 1114 | * @type: memblock type to set node ID for |
7c0caeb8 TH |
1115 | * @nid: node ID to set |
1116 | * | |
e7e8de59 | 1117 | * Set the nid of memblock @type regions in [@base,@base+@size) to @nid. |
7c0caeb8 TH |
1118 | * Regions which cross the area boundaries are split as necessary. |
1119 | * | |
1120 | * RETURNS: | |
1121 | * 0 on success, -errno on failure. | |
1122 | */ | |
1123 | int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size, | |
e7e8de59 | 1124 | struct memblock_type *type, int nid) |
7c0caeb8 | 1125 | { |
6a9ceb31 TH |
1126 | int start_rgn, end_rgn; |
1127 | int i, ret; | |
7c0caeb8 | 1128 | |
6a9ceb31 TH |
1129 | ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn); |
1130 | if (ret) | |
1131 | return ret; | |
7c0caeb8 | 1132 | |
6a9ceb31 | 1133 | for (i = start_rgn; i < end_rgn; i++) |
e9d24ad3 | 1134 | memblock_set_region_node(&type->regions[i], nid); |
7c0caeb8 TH |
1135 | |
1136 | memblock_merge_regions(type); | |
1137 | return 0; | |
1138 | } | |
1139 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ | |
1140 | ||
2bfc2862 AM |
1141 | static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size, |
1142 | phys_addr_t align, phys_addr_t start, | |
fc6daaf9 | 1143 | phys_addr_t end, int nid, ulong flags) |
95f72d1e | 1144 | { |
6ed311b2 | 1145 | phys_addr_t found; |
95f72d1e | 1146 | |
79f40fab GS |
1147 | if (!align) |
1148 | align = SMP_CACHE_BYTES; | |
94f3d3af | 1149 | |
fc6daaf9 TL |
1150 | found = memblock_find_in_range_node(size, align, start, end, nid, |
1151 | flags); | |
aedf95ea CM |
1152 | if (found && !memblock_reserve(found, size)) { |
1153 | /* | |
1154 | * The min_count is set to 0 so that memblock allocations are | |
1155 | * never reported as leaks. | |
1156 | */ | |
9099daed | 1157 | kmemleak_alloc_phys(found, size, 0, 0); |
6ed311b2 | 1158 | return found; |
aedf95ea | 1159 | } |
6ed311b2 | 1160 | return 0; |
95f72d1e YL |
1161 | } |
1162 | ||
2bfc2862 | 1163 | phys_addr_t __init memblock_alloc_range(phys_addr_t size, phys_addr_t align, |
fc6daaf9 TL |
1164 | phys_addr_t start, phys_addr_t end, |
1165 | ulong flags) | |
2bfc2862 | 1166 | { |
fc6daaf9 TL |
1167 | return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE, |
1168 | flags); | |
2bfc2862 AM |
1169 | } |
1170 | ||
b575454f | 1171 | phys_addr_t __init memblock_alloc_base_nid(phys_addr_t size, |
2bfc2862 | 1172 | phys_addr_t align, phys_addr_t max_addr, |
fc6daaf9 | 1173 | int nid, ulong flags) |
2bfc2862 | 1174 | { |
fc6daaf9 | 1175 | return memblock_alloc_range_nid(size, align, 0, max_addr, nid, flags); |
2bfc2862 AM |
1176 | } |
1177 | ||
7bd0b0f0 TH |
1178 | phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid) |
1179 | { | |
a3f5bafc TL |
1180 | ulong flags = choose_memblock_flags(); |
1181 | phys_addr_t ret; | |
1182 | ||
1183 | again: | |
1184 | ret = memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE, | |
1185 | nid, flags); | |
1186 | ||
1187 | if (!ret && (flags & MEMBLOCK_MIRROR)) { | |
1188 | flags &= ~MEMBLOCK_MIRROR; | |
1189 | goto again; | |
1190 | } | |
1191 | return ret; | |
7bd0b0f0 TH |
1192 | } |
1193 | ||
1194 | phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr) | |
1195 | { | |
fc6daaf9 TL |
1196 | return memblock_alloc_base_nid(size, align, max_addr, NUMA_NO_NODE, |
1197 | MEMBLOCK_NONE); | |
7bd0b0f0 TH |
1198 | } |
1199 | ||
6ed311b2 | 1200 | phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr) |
95f72d1e | 1201 | { |
6ed311b2 BH |
1202 | phys_addr_t alloc; |
1203 | ||
1204 | alloc = __memblock_alloc_base(size, align, max_addr); | |
1205 | ||
1206 | if (alloc == 0) | |
5d63f81c MC |
1207 | panic("ERROR: Failed to allocate %pa bytes below %pa.\n", |
1208 | &size, &max_addr); | |
6ed311b2 BH |
1209 | |
1210 | return alloc; | |
95f72d1e YL |
1211 | } |
1212 | ||
6ed311b2 | 1213 | phys_addr_t __init memblock_alloc(phys_addr_t size, phys_addr_t align) |
95f72d1e | 1214 | { |
6ed311b2 BH |
1215 | return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE); |
1216 | } | |
95f72d1e | 1217 | |
9d1e2492 BH |
1218 | phys_addr_t __init memblock_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid) |
1219 | { | |
1220 | phys_addr_t res = memblock_alloc_nid(size, align, nid); | |
1221 | ||
1222 | if (res) | |
1223 | return res; | |
15fb0972 | 1224 | return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE); |
95f72d1e YL |
1225 | } |
1226 | ||
26f09e9b SS |
1227 | /** |
1228 | * memblock_virt_alloc_internal - allocate boot memory block | |
1229 | * @size: size of memory block to be allocated in bytes | |
1230 | * @align: alignment of the region and block's size | |
1231 | * @min_addr: the lower bound of the memory region to allocate (phys address) | |
1232 | * @max_addr: the upper bound of the memory region to allocate (phys address) | |
1233 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1234 | * | |
1235 | * The @min_addr limit is dropped if it can not be satisfied and the allocation | |
1236 | * will fall back to memory below @min_addr. Also, allocation may fall back | |
1237 | * to any node in the system if the specified node can not | |
1238 | * hold the requested memory. | |
1239 | * | |
1240 | * The allocation is performed from memory region limited by | |
1241 | * memblock.current_limit if @max_addr == %BOOTMEM_ALLOC_ACCESSIBLE. | |
1242 | * | |
1243 | * The memory block is aligned on SMP_CACHE_BYTES if @align == 0. | |
1244 | * | |
1245 | * The phys address of allocated boot memory block is converted to virtual and | |
1246 | * allocated memory is reset to 0. | |
1247 | * | |
1248 | * In addition, function sets the min_count to 0 using kmemleak_alloc for | |
1249 | * allocated boot memory block, so that it is never reported as leaks. | |
1250 | * | |
1251 | * RETURNS: | |
1252 | * Virtual address of allocated memory block on success, NULL on failure. | |
1253 | */ | |
1254 | static void * __init memblock_virt_alloc_internal( | |
1255 | phys_addr_t size, phys_addr_t align, | |
1256 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1257 | int nid) | |
1258 | { | |
1259 | phys_addr_t alloc; | |
1260 | void *ptr; | |
a3f5bafc | 1261 | ulong flags = choose_memblock_flags(); |
26f09e9b | 1262 | |
560dca27 GS |
1263 | if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n")) |
1264 | nid = NUMA_NO_NODE; | |
26f09e9b SS |
1265 | |
1266 | /* | |
1267 | * Detect any accidental use of these APIs after slab is ready, as at | |
1268 | * this moment memblock may be deinitialized already and its | |
1269 | * internal data may be destroyed (after execution of free_all_bootmem) | |
1270 | */ | |
1271 | if (WARN_ON_ONCE(slab_is_available())) | |
1272 | return kzalloc_node(size, GFP_NOWAIT, nid); | |
1273 | ||
1274 | if (!align) | |
1275 | align = SMP_CACHE_BYTES; | |
1276 | ||
f544e14f YL |
1277 | if (max_addr > memblock.current_limit) |
1278 | max_addr = memblock.current_limit; | |
26f09e9b SS |
1279 | again: |
1280 | alloc = memblock_find_in_range_node(size, align, min_addr, max_addr, | |
a3f5bafc | 1281 | nid, flags); |
7d41c03e | 1282 | if (alloc && !memblock_reserve(alloc, size)) |
26f09e9b SS |
1283 | goto done; |
1284 | ||
1285 | if (nid != NUMA_NO_NODE) { | |
1286 | alloc = memblock_find_in_range_node(size, align, min_addr, | |
fc6daaf9 | 1287 | max_addr, NUMA_NO_NODE, |
a3f5bafc | 1288 | flags); |
7d41c03e | 1289 | if (alloc && !memblock_reserve(alloc, size)) |
26f09e9b SS |
1290 | goto done; |
1291 | } | |
1292 | ||
1293 | if (min_addr) { | |
1294 | min_addr = 0; | |
1295 | goto again; | |
26f09e9b SS |
1296 | } |
1297 | ||
a3f5bafc TL |
1298 | if (flags & MEMBLOCK_MIRROR) { |
1299 | flags &= ~MEMBLOCK_MIRROR; | |
1300 | pr_warn("Could not allocate %pap bytes of mirrored memory\n", | |
1301 | &size); | |
1302 | goto again; | |
1303 | } | |
1304 | ||
1305 | return NULL; | |
26f09e9b | 1306 | done: |
26f09e9b | 1307 | ptr = phys_to_virt(alloc); |
26f09e9b SS |
1308 | |
1309 | /* | |
1310 | * The min_count is set to 0 so that bootmem allocated blocks | |
1311 | * are never reported as leaks. This is because many of these blocks | |
1312 | * are only referred via the physical address which is not | |
1313 | * looked up by kmemleak. | |
1314 | */ | |
1315 | kmemleak_alloc(ptr, size, 0, 0); | |
1316 | ||
1317 | return ptr; | |
26f09e9b SS |
1318 | } |
1319 | ||
ea1f5f37 PT |
1320 | /** |
1321 | * memblock_virt_alloc_try_nid_raw - allocate boot memory block without zeroing | |
1322 | * memory and without panicking | |
1323 | * @size: size of memory block to be allocated in bytes | |
1324 | * @align: alignment of the region and block's size | |
1325 | * @min_addr: the lower bound of the memory region from where the allocation | |
1326 | * is preferred (phys address) | |
1327 | * @max_addr: the upper bound of the memory region from where the allocation | |
1328 | * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to | |
1329 | * allocate only from memory limited by memblock.current_limit value | |
1330 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1331 | * | |
1332 | * Public function, provides additional debug information (including caller | |
1333 | * info), if enabled. Does not zero allocated memory, does not panic if request | |
1334 | * cannot be satisfied. | |
1335 | * | |
1336 | * RETURNS: | |
1337 | * Virtual address of allocated memory block on success, NULL on failure. | |
1338 | */ | |
1339 | void * __init memblock_virt_alloc_try_nid_raw( | |
1340 | phys_addr_t size, phys_addr_t align, | |
1341 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1342 | int nid) | |
1343 | { | |
1344 | void *ptr; | |
1345 | ||
1346 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n", | |
1347 | __func__, (u64)size, (u64)align, nid, (u64)min_addr, | |
1348 | (u64)max_addr, (void *)_RET_IP_); | |
1349 | ||
1350 | ptr = memblock_virt_alloc_internal(size, align, | |
1351 | min_addr, max_addr, nid); | |
1352 | #ifdef CONFIG_DEBUG_VM | |
1353 | if (ptr && size > 0) | |
f165b378 | 1354 | memset(ptr, PAGE_POISON_PATTERN, size); |
ea1f5f37 PT |
1355 | #endif |
1356 | return ptr; | |
1357 | } | |
1358 | ||
26f09e9b SS |
1359 | /** |
1360 | * memblock_virt_alloc_try_nid_nopanic - allocate boot memory block | |
1361 | * @size: size of memory block to be allocated in bytes | |
1362 | * @align: alignment of the region and block's size | |
1363 | * @min_addr: the lower bound of the memory region from where the allocation | |
1364 | * is preferred (phys address) | |
1365 | * @max_addr: the upper bound of the memory region from where the allocation | |
1366 | * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to | |
1367 | * allocate only from memory limited by memblock.current_limit value | |
1368 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1369 | * | |
ea1f5f37 PT |
1370 | * Public function, provides additional debug information (including caller |
1371 | * info), if enabled. This function zeroes the allocated memory. | |
26f09e9b SS |
1372 | * |
1373 | * RETURNS: | |
1374 | * Virtual address of allocated memory block on success, NULL on failure. | |
1375 | */ | |
1376 | void * __init memblock_virt_alloc_try_nid_nopanic( | |
1377 | phys_addr_t size, phys_addr_t align, | |
1378 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1379 | int nid) | |
1380 | { | |
ea1f5f37 PT |
1381 | void *ptr; |
1382 | ||
26f09e9b SS |
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_); | |
ea1f5f37 PT |
1386 | |
1387 | ptr = memblock_virt_alloc_internal(size, align, | |
1388 | min_addr, max_addr, nid); | |
1389 | if (ptr) | |
1390 | memset(ptr, 0, size); | |
1391 | return ptr; | |
26f09e9b SS |
1392 | } |
1393 | ||
1394 | /** | |
1395 | * memblock_virt_alloc_try_nid - allocate boot memory block with panicking | |
1396 | * @size: size of memory block to be allocated in bytes | |
1397 | * @align: alignment of the region and block's size | |
1398 | * @min_addr: the lower bound of the memory region from where the allocation | |
1399 | * is preferred (phys address) | |
1400 | * @max_addr: the upper bound of the memory region from where the allocation | |
1401 | * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to | |
1402 | * allocate only from memory limited by memblock.current_limit value | |
1403 | * @nid: nid of the free area to find, %NUMA_NO_NODE for any node | |
1404 | * | |
ea1f5f37 | 1405 | * Public panicking version of memblock_virt_alloc_try_nid_nopanic() |
26f09e9b SS |
1406 | * which provides debug information (including caller info), if enabled, |
1407 | * and panics if the request can not be satisfied. | |
1408 | * | |
1409 | * RETURNS: | |
1410 | * Virtual address of allocated memory block on success, NULL on failure. | |
1411 | */ | |
1412 | void * __init memblock_virt_alloc_try_nid( | |
1413 | phys_addr_t size, phys_addr_t align, | |
1414 | phys_addr_t min_addr, phys_addr_t max_addr, | |
1415 | int nid) | |
1416 | { | |
1417 | void *ptr; | |
1418 | ||
1419 | memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n", | |
1420 | __func__, (u64)size, (u64)align, nid, (u64)min_addr, | |
1421 | (u64)max_addr, (void *)_RET_IP_); | |
1422 | ptr = memblock_virt_alloc_internal(size, align, | |
1423 | min_addr, max_addr, nid); | |
ea1f5f37 PT |
1424 | if (ptr) { |
1425 | memset(ptr, 0, size); | |
26f09e9b | 1426 | return ptr; |
ea1f5f37 | 1427 | } |
26f09e9b SS |
1428 | |
1429 | panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx\n", | |
1430 | __func__, (u64)size, (u64)align, nid, (u64)min_addr, | |
1431 | (u64)max_addr); | |
1432 | return NULL; | |
1433 | } | |
1434 | ||
1435 | /** | |
1436 | * __memblock_free_early - free boot memory block | |
1437 | * @base: phys starting address of the boot memory block | |
1438 | * @size: size of the boot memory block in bytes | |
1439 | * | |
1440 | * Free boot memory block previously allocated by memblock_virt_alloc_xx() API. | |
1441 | * The freeing memory will not be released to the buddy allocator. | |
1442 | */ | |
1443 | void __init __memblock_free_early(phys_addr_t base, phys_addr_t size) | |
1444 | { | |
1445 | memblock_dbg("%s: [%#016llx-%#016llx] %pF\n", | |
1446 | __func__, (u64)base, (u64)base + size - 1, | |
1447 | (void *)_RET_IP_); | |
9099daed | 1448 | kmemleak_free_part_phys(base, size); |
f1af9d3a | 1449 | memblock_remove_range(&memblock.reserved, base, size); |
26f09e9b SS |
1450 | } |
1451 | ||
1452 | /* | |
1453 | * __memblock_free_late - free bootmem block pages directly to buddy allocator | |
1454 | * @addr: phys starting address of the boot memory block | |
1455 | * @size: size of the boot memory block in bytes | |
1456 | * | |
1457 | * This is only useful when the bootmem allocator has already been torn | |
1458 | * down, but we are still initializing the system. Pages are released directly | |
1459 | * to the buddy allocator, no bootmem metadata is updated because it is gone. | |
1460 | */ | |
1461 | void __init __memblock_free_late(phys_addr_t base, phys_addr_t size) | |
1462 | { | |
1463 | u64 cursor, end; | |
1464 | ||
1465 | memblock_dbg("%s: [%#016llx-%#016llx] %pF\n", | |
1466 | __func__, (u64)base, (u64)base + size - 1, | |
1467 | (void *)_RET_IP_); | |
9099daed | 1468 | kmemleak_free_part_phys(base, size); |
26f09e9b SS |
1469 | cursor = PFN_UP(base); |
1470 | end = PFN_DOWN(base + size); | |
1471 | ||
1472 | for (; cursor < end; cursor++) { | |
d70ddd7a | 1473 | __free_pages_bootmem(pfn_to_page(cursor), cursor, 0); |
26f09e9b SS |
1474 | totalram_pages++; |
1475 | } | |
1476 | } | |
9d1e2492 BH |
1477 | |
1478 | /* | |
1479 | * Remaining API functions | |
1480 | */ | |
1481 | ||
1f1ffb8a | 1482 | phys_addr_t __init_memblock memblock_phys_mem_size(void) |
95f72d1e | 1483 | { |
1440c4e2 | 1484 | return memblock.memory.total_size; |
95f72d1e YL |
1485 | } |
1486 | ||
8907de5d SD |
1487 | phys_addr_t __init_memblock memblock_reserved_size(void) |
1488 | { | |
1489 | return memblock.reserved.total_size; | |
1490 | } | |
1491 | ||
595ad9af YL |
1492 | phys_addr_t __init memblock_mem_size(unsigned long limit_pfn) |
1493 | { | |
1494 | unsigned long pages = 0; | |
1495 | struct memblock_region *r; | |
1496 | unsigned long start_pfn, end_pfn; | |
1497 | ||
1498 | for_each_memblock(memory, r) { | |
1499 | start_pfn = memblock_region_memory_base_pfn(r); | |
1500 | end_pfn = memblock_region_memory_end_pfn(r); | |
1501 | start_pfn = min_t(unsigned long, start_pfn, limit_pfn); | |
1502 | end_pfn = min_t(unsigned long, end_pfn, limit_pfn); | |
1503 | pages += end_pfn - start_pfn; | |
1504 | } | |
1505 | ||
16763230 | 1506 | return PFN_PHYS(pages); |
595ad9af YL |
1507 | } |
1508 | ||
0a93ebef SR |
1509 | /* lowest address */ |
1510 | phys_addr_t __init_memblock memblock_start_of_DRAM(void) | |
1511 | { | |
1512 | return memblock.memory.regions[0].base; | |
1513 | } | |
1514 | ||
10d06439 | 1515 | phys_addr_t __init_memblock memblock_end_of_DRAM(void) |
95f72d1e YL |
1516 | { |
1517 | int idx = memblock.memory.cnt - 1; | |
1518 | ||
e3239ff9 | 1519 | return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size); |
95f72d1e YL |
1520 | } |
1521 | ||
a571d4eb | 1522 | static phys_addr_t __init_memblock __find_max_addr(phys_addr_t limit) |
95f72d1e | 1523 | { |
1c4bc43d | 1524 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
136199f0 | 1525 | struct memblock_region *r; |
95f72d1e | 1526 | |
a571d4eb DC |
1527 | /* |
1528 | * translate the memory @limit size into the max address within one of | |
1529 | * the memory memblock regions, if the @limit exceeds the total size | |
1c4bc43d | 1530 | * of those regions, max_addr will keep original value PHYS_ADDR_MAX |
a571d4eb | 1531 | */ |
136199f0 | 1532 | for_each_memblock(memory, r) { |
c0ce8fef TH |
1533 | if (limit <= r->size) { |
1534 | max_addr = r->base + limit; | |
1535 | break; | |
95f72d1e | 1536 | } |
c0ce8fef | 1537 | limit -= r->size; |
95f72d1e | 1538 | } |
c0ce8fef | 1539 | |
a571d4eb DC |
1540 | return max_addr; |
1541 | } | |
1542 | ||
1543 | void __init memblock_enforce_memory_limit(phys_addr_t limit) | |
1544 | { | |
1c4bc43d | 1545 | phys_addr_t max_addr = PHYS_ADDR_MAX; |
a571d4eb DC |
1546 | |
1547 | if (!limit) | |
1548 | return; | |
1549 | ||
1550 | max_addr = __find_max_addr(limit); | |
1551 | ||
1552 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1553 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1554 | return; |
1555 | ||
c0ce8fef | 1556 | /* truncate both memory and reserved regions */ |
f1af9d3a | 1557 | memblock_remove_range(&memblock.memory, max_addr, |
1c4bc43d | 1558 | PHYS_ADDR_MAX); |
f1af9d3a | 1559 | memblock_remove_range(&memblock.reserved, max_addr, |
1c4bc43d | 1560 | PHYS_ADDR_MAX); |
95f72d1e YL |
1561 | } |
1562 | ||
c9ca9b4e AT |
1563 | void __init memblock_cap_memory_range(phys_addr_t base, phys_addr_t size) |
1564 | { | |
1565 | int start_rgn, end_rgn; | |
1566 | int i, ret; | |
1567 | ||
1568 | if (!size) | |
1569 | return; | |
1570 | ||
1571 | ret = memblock_isolate_range(&memblock.memory, base, size, | |
1572 | &start_rgn, &end_rgn); | |
1573 | if (ret) | |
1574 | return; | |
1575 | ||
1576 | /* remove all the MAP regions */ | |
1577 | for (i = memblock.memory.cnt - 1; i >= end_rgn; i--) | |
1578 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1579 | memblock_remove_region(&memblock.memory, i); | |
1580 | ||
1581 | for (i = start_rgn - 1; i >= 0; i--) | |
1582 | if (!memblock_is_nomap(&memblock.memory.regions[i])) | |
1583 | memblock_remove_region(&memblock.memory, i); | |
1584 | ||
1585 | /* truncate the reserved regions */ | |
1586 | memblock_remove_range(&memblock.reserved, 0, base); | |
1587 | memblock_remove_range(&memblock.reserved, | |
1c4bc43d | 1588 | base + size, PHYS_ADDR_MAX); |
c9ca9b4e AT |
1589 | } |
1590 | ||
a571d4eb DC |
1591 | void __init memblock_mem_limit_remove_map(phys_addr_t limit) |
1592 | { | |
a571d4eb | 1593 | phys_addr_t max_addr; |
a571d4eb DC |
1594 | |
1595 | if (!limit) | |
1596 | return; | |
1597 | ||
1598 | max_addr = __find_max_addr(limit); | |
1599 | ||
1600 | /* @limit exceeds the total size of the memory, do nothing */ | |
1c4bc43d | 1601 | if (max_addr == PHYS_ADDR_MAX) |
a571d4eb DC |
1602 | return; |
1603 | ||
c9ca9b4e | 1604 | memblock_cap_memory_range(0, max_addr); |
a571d4eb DC |
1605 | } |
1606 | ||
cd79481d | 1607 | static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr) |
72d4b0b4 BH |
1608 | { |
1609 | unsigned int left = 0, right = type->cnt; | |
1610 | ||
1611 | do { | |
1612 | unsigned int mid = (right + left) / 2; | |
1613 | ||
1614 | if (addr < type->regions[mid].base) | |
1615 | right = mid; | |
1616 | else if (addr >= (type->regions[mid].base + | |
1617 | type->regions[mid].size)) | |
1618 | left = mid + 1; | |
1619 | else | |
1620 | return mid; | |
1621 | } while (left < right); | |
1622 | return -1; | |
1623 | } | |
1624 | ||
b4ad0c7e | 1625 | bool __init memblock_is_reserved(phys_addr_t addr) |
95f72d1e | 1626 | { |
72d4b0b4 BH |
1627 | return memblock_search(&memblock.reserved, addr) != -1; |
1628 | } | |
95f72d1e | 1629 | |
b4ad0c7e | 1630 | bool __init_memblock memblock_is_memory(phys_addr_t addr) |
72d4b0b4 BH |
1631 | { |
1632 | return memblock_search(&memblock.memory, addr) != -1; | |
1633 | } | |
1634 | ||
937f0c26 | 1635 | bool __init_memblock memblock_is_map_memory(phys_addr_t addr) |
bf3d3cc5 AB |
1636 | { |
1637 | int i = memblock_search(&memblock.memory, addr); | |
1638 | ||
1639 | if (i == -1) | |
1640 | return false; | |
1641 | return !memblock_is_nomap(&memblock.memory.regions[i]); | |
1642 | } | |
1643 | ||
e76b63f8 YL |
1644 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1645 | int __init_memblock memblock_search_pfn_nid(unsigned long pfn, | |
1646 | unsigned long *start_pfn, unsigned long *end_pfn) | |
1647 | { | |
1648 | struct memblock_type *type = &memblock.memory; | |
16763230 | 1649 | int mid = memblock_search(type, PFN_PHYS(pfn)); |
e76b63f8 YL |
1650 | |
1651 | if (mid == -1) | |
1652 | return -1; | |
1653 | ||
f7e2f7e8 FF |
1654 | *start_pfn = PFN_DOWN(type->regions[mid].base); |
1655 | *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size); | |
e76b63f8 YL |
1656 | |
1657 | return type->regions[mid].nid; | |
1658 | } | |
1659 | #endif | |
1660 | ||
eab30949 SB |
1661 | /** |
1662 | * memblock_is_region_memory - check if a region is a subset of memory | |
1663 | * @base: base of region to check | |
1664 | * @size: size of region to check | |
1665 | * | |
1666 | * Check if the region [@base, @base+@size) is a subset of a memory block. | |
1667 | * | |
1668 | * RETURNS: | |
1669 | * 0 if false, non-zero if true | |
1670 | */ | |
937f0c26 | 1671 | bool __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size) |
72d4b0b4 | 1672 | { |
abb65272 | 1673 | int idx = memblock_search(&memblock.memory, base); |
eb18f1b5 | 1674 | phys_addr_t end = base + memblock_cap_size(base, &size); |
72d4b0b4 BH |
1675 | |
1676 | if (idx == -1) | |
937f0c26 | 1677 | return false; |
ef415ef4 | 1678 | return (memblock.memory.regions[idx].base + |
eb18f1b5 | 1679 | memblock.memory.regions[idx].size) >= end; |
95f72d1e YL |
1680 | } |
1681 | ||
eab30949 SB |
1682 | /** |
1683 | * memblock_is_region_reserved - check if a region intersects reserved memory | |
1684 | * @base: base of region to check | |
1685 | * @size: size of region to check | |
1686 | * | |
1687 | * Check if the region [@base, @base+@size) intersects a reserved memory block. | |
1688 | * | |
1689 | * RETURNS: | |
c5c5c9d1 | 1690 | * True if they intersect, false if not. |
eab30949 | 1691 | */ |
c5c5c9d1 | 1692 | bool __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size) |
95f72d1e | 1693 | { |
eb18f1b5 | 1694 | memblock_cap_size(base, &size); |
c5c5c9d1 | 1695 | return memblock_overlaps_region(&memblock.reserved, base, size); |
95f72d1e YL |
1696 | } |
1697 | ||
6ede1fd3 YL |
1698 | void __init_memblock memblock_trim_memory(phys_addr_t align) |
1699 | { | |
6ede1fd3 | 1700 | phys_addr_t start, end, orig_start, orig_end; |
136199f0 | 1701 | struct memblock_region *r; |
6ede1fd3 | 1702 | |
136199f0 EM |
1703 | for_each_memblock(memory, r) { |
1704 | orig_start = r->base; | |
1705 | orig_end = r->base + r->size; | |
6ede1fd3 YL |
1706 | start = round_up(orig_start, align); |
1707 | end = round_down(orig_end, align); | |
1708 | ||
1709 | if (start == orig_start && end == orig_end) | |
1710 | continue; | |
1711 | ||
1712 | if (start < end) { | |
136199f0 EM |
1713 | r->base = start; |
1714 | r->size = end - start; | |
6ede1fd3 | 1715 | } else { |
136199f0 EM |
1716 | memblock_remove_region(&memblock.memory, |
1717 | r - memblock.memory.regions); | |
1718 | r--; | |
6ede1fd3 YL |
1719 | } |
1720 | } | |
1721 | } | |
e63075a3 | 1722 | |
3661ca66 | 1723 | void __init_memblock memblock_set_current_limit(phys_addr_t limit) |
e63075a3 BH |
1724 | { |
1725 | memblock.current_limit = limit; | |
1726 | } | |
1727 | ||
fec51014 LA |
1728 | phys_addr_t __init_memblock memblock_get_current_limit(void) |
1729 | { | |
1730 | return memblock.current_limit; | |
1731 | } | |
1732 | ||
0262d9c8 | 1733 | static void __init_memblock memblock_dump(struct memblock_type *type) |
6ed311b2 | 1734 | { |
5d63f81c | 1735 | phys_addr_t base, end, size; |
66a20757 | 1736 | unsigned long flags; |
8c9c1701 AK |
1737 | int idx; |
1738 | struct memblock_region *rgn; | |
6ed311b2 | 1739 | |
0262d9c8 | 1740 | pr_info(" %s.cnt = 0x%lx\n", type->name, type->cnt); |
6ed311b2 | 1741 | |
66e8b438 | 1742 | for_each_memblock_type(idx, type, rgn) { |
7c0caeb8 TH |
1743 | char nid_buf[32] = ""; |
1744 | ||
1745 | base = rgn->base; | |
1746 | size = rgn->size; | |
5d63f81c | 1747 | end = base + size - 1; |
66a20757 | 1748 | flags = rgn->flags; |
7c0caeb8 TH |
1749 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP |
1750 | if (memblock_get_region_node(rgn) != MAX_NUMNODES) | |
1751 | snprintf(nid_buf, sizeof(nid_buf), " on node %d", | |
1752 | memblock_get_region_node(rgn)); | |
1753 | #endif | |
5d63f81c | 1754 | pr_info(" %s[%#x]\t[%pa-%pa], %pa bytes%s flags: %#lx\n", |
0262d9c8 | 1755 | type->name, idx, &base, &end, &size, nid_buf, flags); |
6ed311b2 BH |
1756 | } |
1757 | } | |
1758 | ||
4ff7b82f | 1759 | void __init_memblock __memblock_dump_all(void) |
6ed311b2 | 1760 | { |
6ed311b2 | 1761 | pr_info("MEMBLOCK configuration:\n"); |
5d63f81c MC |
1762 | pr_info(" memory size = %pa reserved size = %pa\n", |
1763 | &memblock.memory.total_size, | |
1764 | &memblock.reserved.total_size); | |
6ed311b2 | 1765 | |
0262d9c8 HC |
1766 | memblock_dump(&memblock.memory); |
1767 | memblock_dump(&memblock.reserved); | |
409efd4c | 1768 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
0262d9c8 | 1769 | memblock_dump(&memblock.physmem); |
409efd4c | 1770 | #endif |
6ed311b2 BH |
1771 | } |
1772 | ||
1aadc056 | 1773 | void __init memblock_allow_resize(void) |
6ed311b2 | 1774 | { |
142b45a7 | 1775 | memblock_can_resize = 1; |
6ed311b2 BH |
1776 | } |
1777 | ||
6ed311b2 BH |
1778 | static int __init early_memblock(char *p) |
1779 | { | |
1780 | if (p && strstr(p, "debug")) | |
1781 | memblock_debug = 1; | |
1782 | return 0; | |
1783 | } | |
1784 | early_param("memblock", early_memblock); | |
1785 | ||
c378ddd5 | 1786 | #if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK) |
6d03b885 BH |
1787 | |
1788 | static int memblock_debug_show(struct seq_file *m, void *private) | |
1789 | { | |
1790 | struct memblock_type *type = m->private; | |
1791 | struct memblock_region *reg; | |
1792 | int i; | |
5d63f81c | 1793 | phys_addr_t end; |
6d03b885 BH |
1794 | |
1795 | for (i = 0; i < type->cnt; i++) { | |
1796 | reg = &type->regions[i]; | |
5d63f81c | 1797 | end = reg->base + reg->size - 1; |
6d03b885 | 1798 | |
5d63f81c MC |
1799 | seq_printf(m, "%4d: ", i); |
1800 | seq_printf(m, "%pa..%pa\n", ®->base, &end); | |
6d03b885 BH |
1801 | } |
1802 | return 0; | |
1803 | } | |
5ad35093 | 1804 | DEFINE_SHOW_ATTRIBUTE(memblock_debug); |
6d03b885 BH |
1805 | |
1806 | static int __init memblock_init_debugfs(void) | |
1807 | { | |
1808 | struct dentry *root = debugfs_create_dir("memblock", NULL); | |
1809 | if (!root) | |
1810 | return -ENXIO; | |
0825a6f9 JP |
1811 | debugfs_create_file("memory", 0444, root, |
1812 | &memblock.memory, &memblock_debug_fops); | |
1813 | debugfs_create_file("reserved", 0444, root, | |
1814 | &memblock.reserved, &memblock_debug_fops); | |
70210ed9 | 1815 | #ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP |
0825a6f9 JP |
1816 | debugfs_create_file("physmem", 0444, root, |
1817 | &memblock.physmem, &memblock_debug_fops); | |
70210ed9 | 1818 | #endif |
6d03b885 BH |
1819 | |
1820 | return 0; | |
1821 | } | |
1822 | __initcall(memblock_init_debugfs); | |
1823 | ||
1824 | #endif /* CONFIG_DEBUG_FS */ |