blob: 6ecb0d937fb5f9bde4015d20b620484f5c965f84 [file] [log] [blame]
Yinghai Lu95f72d12010-07-12 14:36:09 +10001/*
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>
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -070014#include <linux/slab.h>
Yinghai Lu95f72d12010-07-12 14:36:09 +100015#include <linux/init.h>
16#include <linux/bitops.h>
Benjamin Herrenschmidt449e8df2010-07-06 15:39:07 -070017#include <linux/poison.h>
Benjamin Herrenschmidtc196f762010-07-06 15:39:16 -070018#include <linux/pfn.h>
Benjamin Herrenschmidt6d03b882010-07-06 15:39:19 -070019#include <linux/debugfs.h>
20#include <linux/seq_file.h>
Yinghai Lu95f72d12010-07-12 14:36:09 +100021#include <linux/memblock.h>
22
Tang Chen79442ed2013-11-12 15:07:59 -080023#include <asm-generic/sections.h>
Santosh Shilimkar26f09e92014-01-21 15:50:19 -080024#include <linux/io.h>
25
26#include "internal.h"
Tang Chen79442ed2013-11-12 15:07:59 -080027
Tejun Heofe091c22011-12-08 10:22:07 -080028static struct memblock_region memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock;
29static struct memblock_region memblock_reserved_init_regions[INIT_MEMBLOCK_REGIONS] __initdata_memblock;
Philipp Hachtmann70210ed2014-01-29 18:16:01 +010030#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
31static struct memblock_region memblock_physmem_init_regions[INIT_PHYSMEM_REGIONS] __initdata_memblock;
32#endif
Tejun Heofe091c22011-12-08 10:22:07 -080033
34struct memblock memblock __initdata_memblock = {
35 .memory.regions = memblock_memory_init_regions,
36 .memory.cnt = 1, /* empty dummy entry */
37 .memory.max = INIT_MEMBLOCK_REGIONS,
38
39 .reserved.regions = memblock_reserved_init_regions,
40 .reserved.cnt = 1, /* empty dummy entry */
41 .reserved.max = INIT_MEMBLOCK_REGIONS,
42
Philipp Hachtmann70210ed2014-01-29 18:16:01 +010043#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
44 .physmem.regions = memblock_physmem_init_regions,
45 .physmem.cnt = 1, /* empty dummy entry */
46 .physmem.max = INIT_PHYSMEM_REGIONS,
47#endif
48
Tang Chen79442ed2013-11-12 15:07:59 -080049 .bottom_up = false,
Tejun Heofe091c22011-12-08 10:22:07 -080050 .current_limit = MEMBLOCK_ALLOC_ANYWHERE,
51};
Yinghai Lu95f72d12010-07-12 14:36:09 +100052
Yinghai Lu10d06432010-07-28 15:43:02 +100053int memblock_debug __initdata_memblock;
Tang Chen55ac5902014-01-21 15:49:35 -080054#ifdef CONFIG_MOVABLE_NODE
55bool movable_node_enabled __initdata_memblock = false;
56#endif
Tejun Heo1aadc052011-12-08 10:22:08 -080057static int memblock_can_resize __initdata_memblock;
Gavin Shan181eb392012-05-29 15:06:50 -070058static int memblock_memory_in_slab __initdata_memblock = 0;
59static int memblock_reserved_in_slab __initdata_memblock = 0;
Yinghai Lu95f72d12010-07-12 14:36:09 +100060
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -070061/* inline so we don't get a warning when pr_debug is compiled out */
Raghavendra D Prabhuc2233112012-10-08 16:33:55 -070062static __init_memblock const char *
63memblock_type_name(struct memblock_type *type)
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -070064{
65 if (type == &memblock.memory)
66 return "memory";
67 else if (type == &memblock.reserved)
68 return "reserved";
69 else
70 return "unknown";
71}
72
Tejun Heoeb18f1b2011-12-08 10:22:07 -080073/* adjust *@size so that (@base + *@size) doesn't overflow, return new size */
74static inline phys_addr_t memblock_cap_size(phys_addr_t base, phys_addr_t *size)
75{
76 return *size = min(*size, (phys_addr_t)ULLONG_MAX - base);
77}
78
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +100079/*
80 * Address comparison utilities
81 */
Yinghai Lu10d06432010-07-28 15:43:02 +100082static unsigned long __init_memblock memblock_addrs_overlap(phys_addr_t base1, phys_addr_t size1,
Benjamin Herrenschmidt2898cc42010-08-04 13:34:42 +100083 phys_addr_t base2, phys_addr_t size2)
Yinghai Lu95f72d12010-07-12 14:36:09 +100084{
85 return ((base1 < (base2 + size2)) && (base2 < (base1 + size1)));
86}
87
H Hartley Sweeten2d7d3eb2011-10-31 17:09:15 -070088static long __init_memblock memblock_overlaps_region(struct memblock_type *type,
89 phys_addr_t base, phys_addr_t size)
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +100090{
91 unsigned long i;
92
93 for (i = 0; i < type->cnt; i++) {
94 phys_addr_t rgnbase = type->regions[i].base;
95 phys_addr_t rgnsize = type->regions[i].size;
96 if (memblock_addrs_overlap(base, size, rgnbase, rgnsize))
97 break;
98 }
99
100 return (i < type->cnt) ? i : -1;
101}
102
Tang Chen79442ed2013-11-12 15:07:59 -0800103/*
104 * __memblock_find_range_bottom_up - find free area utility in bottom-up
105 * @start: start of candidate range
106 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
107 * @size: size of free area to find
108 * @align: alignment of free area to find
Grygorii Strashkob1154232014-01-21 15:50:16 -0800109 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
Tang Chen79442ed2013-11-12 15:07:59 -0800110 *
111 * Utility called from memblock_find_in_range_node(), find free area bottom-up.
112 *
113 * RETURNS:
114 * Found address on success, 0 on failure.
115 */
116static phys_addr_t __init_memblock
117__memblock_find_range_bottom_up(phys_addr_t start, phys_addr_t end,
118 phys_addr_t size, phys_addr_t align, int nid)
119{
120 phys_addr_t this_start, this_end, cand;
121 u64 i;
122
123 for_each_free_mem_range(i, nid, &this_start, &this_end, NULL) {
124 this_start = clamp(this_start, start, end);
125 this_end = clamp(this_end, start, end);
126
127 cand = round_up(this_start, align);
128 if (cand < this_end && this_end - cand >= size)
129 return cand;
130 }
131
132 return 0;
133}
134
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800135/**
Tang Chen14028992013-11-12 15:07:57 -0800136 * __memblock_find_range_top_down - find free area utility, in top-down
137 * @start: start of candidate range
138 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
139 * @size: size of free area to find
140 * @align: alignment of free area to find
Grygorii Strashkob1154232014-01-21 15:50:16 -0800141 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
Tang Chen14028992013-11-12 15:07:57 -0800142 *
143 * Utility called from memblock_find_in_range_node(), find free area top-down.
144 *
145 * RETURNS:
Tang Chen79442ed2013-11-12 15:07:59 -0800146 * Found address on success, 0 on failure.
Tang Chen14028992013-11-12 15:07:57 -0800147 */
148static phys_addr_t __init_memblock
149__memblock_find_range_top_down(phys_addr_t start, phys_addr_t end,
150 phys_addr_t size, phys_addr_t align, int nid)
151{
152 phys_addr_t this_start, this_end, cand;
153 u64 i;
154
155 for_each_free_mem_range_reverse(i, nid, &this_start, &this_end, NULL) {
156 this_start = clamp(this_start, start, end);
157 this_end = clamp(this_end, start, end);
158
159 if (this_end < size)
160 continue;
161
162 cand = round_down(this_end - size, align);
163 if (cand >= this_start)
164 return cand;
165 }
166
167 return 0;
168}
169
170/**
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800171 * memblock_find_in_range_node - find free area in given range and node
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800172 * @size: size of free area to find
173 * @align: alignment of free area to find
Grygorii Strashko87029ee2014-01-21 15:50:14 -0800174 * @start: start of candidate range
175 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
Grygorii Strashkob1154232014-01-21 15:50:16 -0800176 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800177 *
178 * Find @size free area aligned to @align in the specified range and node.
179 *
Tang Chen79442ed2013-11-12 15:07:59 -0800180 * When allocation direction is bottom-up, the @start should be greater
181 * than the end of the kernel image. Otherwise, it will be trimmed. The
182 * reason is that we want the bottom-up allocation just near the kernel
183 * image so it is highly likely that the allocated memory and the kernel
184 * will reside in the same node.
185 *
186 * If bottom-up allocation failed, will try to allocate memory top-down.
187 *
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800188 * RETURNS:
Tang Chen79442ed2013-11-12 15:07:59 -0800189 * Found address on success, 0 on failure.
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +1000190 */
Grygorii Strashko87029ee2014-01-21 15:50:14 -0800191phys_addr_t __init_memblock memblock_find_in_range_node(phys_addr_t size,
192 phys_addr_t align, phys_addr_t start,
193 phys_addr_t end, int nid)
Tang Chenf7210e62013-02-22 16:33:51 -0800194{
Tang Chen0cfb8f02014-08-29 15:18:31 -0700195 phys_addr_t kernel_end, ret;
Tang Chen79442ed2013-11-12 15:07:59 -0800196
Tang Chenf7210e62013-02-22 16:33:51 -0800197 /* pump up @end */
198 if (end == MEMBLOCK_ALLOC_ACCESSIBLE)
199 end = memblock.current_limit;
200
201 /* avoid allocating the first page */
202 start = max_t(phys_addr_t, start, PAGE_SIZE);
203 end = max(start, end);
Tang Chen79442ed2013-11-12 15:07:59 -0800204 kernel_end = __pa_symbol(_end);
205
206 /*
207 * try bottom-up allocation only when bottom-up mode
208 * is set and @end is above the kernel image.
209 */
210 if (memblock_bottom_up() && end > kernel_end) {
211 phys_addr_t bottom_up_start;
212
213 /* make sure we will allocate above the kernel */
214 bottom_up_start = max(start, kernel_end);
215
216 /* ok, try bottom-up allocation first */
217 ret = __memblock_find_range_bottom_up(bottom_up_start, end,
218 size, align, nid);
219 if (ret)
220 return ret;
221
222 /*
223 * we always limit bottom-up allocation above the kernel,
224 * but top-down allocation doesn't have the limit, so
225 * retrying top-down allocation may succeed when bottom-up
226 * allocation failed.
227 *
228 * bottom-up allocation is expected to be fail very rarely,
229 * so we use WARN_ONCE() here to see the stack trace if
230 * fail happens.
231 */
232 WARN_ONCE(1, "memblock: bottom-up allocation failed, "
233 "memory hotunplug may be affected\n");
234 }
Tang Chenf7210e62013-02-22 16:33:51 -0800235
Tang Chen14028992013-11-12 15:07:57 -0800236 return __memblock_find_range_top_down(start, end, size, align, nid);
Tang Chenf7210e62013-02-22 16:33:51 -0800237}
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +1000238
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800239/**
240 * memblock_find_in_range - find free area in given range
241 * @start: start of candidate range
242 * @end: end of candidate range, can be %MEMBLOCK_ALLOC_{ANYWHERE|ACCESSIBLE}
243 * @size: size of free area to find
244 * @align: alignment of free area to find
245 *
246 * Find @size free area aligned to @align in the specified range.
247 *
248 * RETURNS:
Tang Chen79442ed2013-11-12 15:07:59 -0800249 * Found address on success, 0 on failure.
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800250 */
251phys_addr_t __init_memblock memblock_find_in_range(phys_addr_t start,
252 phys_addr_t end, phys_addr_t size,
253 phys_addr_t align)
254{
Grygorii Strashko87029ee2014-01-21 15:50:14 -0800255 return memblock_find_in_range_node(size, align, start, end,
Grygorii Strashkob1154232014-01-21 15:50:16 -0800256 NUMA_NO_NODE);
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800257}
258
Yinghai Lu10d06432010-07-28 15:43:02 +1000259static void __init_memblock memblock_remove_region(struct memblock_type *type, unsigned long r)
Yinghai Lu95f72d12010-07-12 14:36:09 +1000260{
Tejun Heo1440c4e2011-12-08 10:22:08 -0800261 type->total_size -= type->regions[r].size;
Tejun Heo7c0caeb2011-07-14 11:43:42 +0200262 memmove(&type->regions[r], &type->regions[r + 1],
263 (type->cnt - (r + 1)) * sizeof(type->regions[r]));
Benjamin Herrenschmidte3239ff2010-08-04 14:06:41 +1000264 type->cnt--;
Yinghai Lu95f72d12010-07-12 14:36:09 +1000265
Benjamin Herrenschmidt8f7a6602011-03-22 16:33:43 -0700266 /* Special case for empty arrays */
267 if (type->cnt == 0) {
Tejun Heo1440c4e2011-12-08 10:22:08 -0800268 WARN_ON(type->total_size != 0);
Benjamin Herrenschmidt8f7a6602011-03-22 16:33:43 -0700269 type->cnt = 1;
270 type->regions[0].base = 0;
271 type->regions[0].size = 0;
Tang Chen66a20752014-01-21 15:49:20 -0800272 type->regions[0].flags = 0;
Tejun Heo7c0caeb2011-07-14 11:43:42 +0200273 memblock_set_region_node(&type->regions[0], MAX_NUMNODES);
Benjamin Herrenschmidt8f7a6602011-03-22 16:33:43 -0700274 }
Yinghai Lu95f72d12010-07-12 14:36:09 +1000275}
276
Philipp Hachtmann354f17e2014-01-23 15:53:24 -0800277#ifdef CONFIG_ARCH_DISCARD_MEMBLOCK
278
Yinghai Lu29f67382012-07-11 14:02:56 -0700279phys_addr_t __init_memblock get_allocated_memblock_reserved_regions_info(
280 phys_addr_t *addr)
281{
282 if (memblock.reserved.regions == memblock_reserved_init_regions)
283 return 0;
284
285 *addr = __pa(memblock.reserved.regions);
286
287 return PAGE_ALIGN(sizeof(struct memblock_region) *
288 memblock.reserved.max);
289}
290
Philipp Hachtmann5e270e22014-01-23 15:53:11 -0800291phys_addr_t __init_memblock get_allocated_memblock_memory_regions_info(
292 phys_addr_t *addr)
293{
294 if (memblock.memory.regions == memblock_memory_init_regions)
295 return 0;
296
297 *addr = __pa(memblock.memory.regions);
298
299 return PAGE_ALIGN(sizeof(struct memblock_region) *
300 memblock.memory.max);
301}
302
303#endif
304
Greg Pearson48c3b582012-06-20 12:53:05 -0700305/**
306 * memblock_double_array - double the size of the memblock regions array
307 * @type: memblock type of the regions array being doubled
308 * @new_area_start: starting address of memory range to avoid overlap with
309 * @new_area_size: size of memory range to avoid overlap with
310 *
311 * Double the size of the @type regions array. If memblock is being used to
312 * allocate memory for a new reserved regions array and there is a previously
313 * allocated memory range [@new_area_start,@new_area_start+@new_area_size]
314 * waiting to be reserved, ensure the memory used by the new array does
315 * not overlap.
316 *
317 * RETURNS:
318 * 0 on success, -1 on failure.
319 */
320static int __init_memblock memblock_double_array(struct memblock_type *type,
321 phys_addr_t new_area_start,
322 phys_addr_t new_area_size)
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700323{
324 struct memblock_region *new_array, *old_array;
Yinghai Lu29f67382012-07-11 14:02:56 -0700325 phys_addr_t old_alloc_size, new_alloc_size;
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700326 phys_addr_t old_size, new_size, addr;
327 int use_slab = slab_is_available();
Gavin Shan181eb392012-05-29 15:06:50 -0700328 int *in_slab;
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700329
330 /* We don't allow resizing until we know about the reserved regions
331 * of memory that aren't suitable for allocation
332 */
333 if (!memblock_can_resize)
334 return -1;
335
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700336 /* Calculate new doubled size */
337 old_size = type->max * sizeof(struct memblock_region);
338 new_size = old_size << 1;
Yinghai Lu29f67382012-07-11 14:02:56 -0700339 /*
340 * We need to allocated new one align to PAGE_SIZE,
341 * so we can free them completely later.
342 */
343 old_alloc_size = PAGE_ALIGN(old_size);
344 new_alloc_size = PAGE_ALIGN(new_size);
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700345
Gavin Shan181eb392012-05-29 15:06:50 -0700346 /* Retrieve the slab flag */
347 if (type == &memblock.memory)
348 in_slab = &memblock_memory_in_slab;
349 else
350 in_slab = &memblock_reserved_in_slab;
351
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700352 /* Try to find some space for it.
353 *
354 * WARNING: We assume that either slab_is_available() and we use it or
Andrew Mortonfd073832012-07-31 16:42:40 -0700355 * we use MEMBLOCK for allocations. That means that this is unsafe to
356 * use when bootmem is currently active (unless bootmem itself is
357 * implemented on top of MEMBLOCK which isn't the case yet)
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700358 *
359 * This should however not be an issue for now, as we currently only
Andrew Mortonfd073832012-07-31 16:42:40 -0700360 * call into MEMBLOCK while it's still active, or much later when slab
361 * is active for memory hotplug operations
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700362 */
363 if (use_slab) {
364 new_array = kmalloc(new_size, GFP_KERNEL);
Tejun Heo1f5026a2011-07-12 09:58:09 +0200365 addr = new_array ? __pa(new_array) : 0;
Gavin Shan4e2f0772012-05-29 15:06:50 -0700366 } else {
Greg Pearson48c3b582012-06-20 12:53:05 -0700367 /* only exclude range when trying to double reserved.regions */
368 if (type != &memblock.reserved)
369 new_area_start = new_area_size = 0;
370
371 addr = memblock_find_in_range(new_area_start + new_area_size,
372 memblock.current_limit,
Yinghai Lu29f67382012-07-11 14:02:56 -0700373 new_alloc_size, PAGE_SIZE);
Greg Pearson48c3b582012-06-20 12:53:05 -0700374 if (!addr && new_area_size)
375 addr = memblock_find_in_range(0,
Andrew Mortonfd073832012-07-31 16:42:40 -0700376 min(new_area_start, memblock.current_limit),
377 new_alloc_size, PAGE_SIZE);
Greg Pearson48c3b582012-06-20 12:53:05 -0700378
Sachin Kamat15674862012-09-04 13:55:05 +0530379 new_array = addr ? __va(addr) : NULL;
Gavin Shan4e2f0772012-05-29 15:06:50 -0700380 }
Tejun Heo1f5026a2011-07-12 09:58:09 +0200381 if (!addr) {
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700382 pr_err("memblock: Failed to double %s array from %ld to %ld entries !\n",
383 memblock_type_name(type), type->max, type->max * 2);
384 return -1;
385 }
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700386
Andrew Mortonfd073832012-07-31 16:42:40 -0700387 memblock_dbg("memblock: %s is doubled to %ld at [%#010llx-%#010llx]",
388 memblock_type_name(type), type->max * 2, (u64)addr,
389 (u64)addr + new_size - 1);
Yinghai Luea9e4372010-07-28 15:13:22 +1000390
Andrew Mortonfd073832012-07-31 16:42:40 -0700391 /*
392 * Found space, we now need to move the array over before we add the
393 * reserved region since it may be our reserved array itself that is
394 * full.
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700395 */
396 memcpy(new_array, type->regions, old_size);
397 memset(new_array + type->max, 0, old_size);
398 old_array = type->regions;
399 type->regions = new_array;
400 type->max <<= 1;
401
Andrew Mortonfd073832012-07-31 16:42:40 -0700402 /* Free old array. We needn't free it if the array is the static one */
Gavin Shan181eb392012-05-29 15:06:50 -0700403 if (*in_slab)
404 kfree(old_array);
405 else if (old_array != memblock_memory_init_regions &&
406 old_array != memblock_reserved_init_regions)
Yinghai Lu29f67382012-07-11 14:02:56 -0700407 memblock_free(__pa(old_array), old_alloc_size);
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700408
Andrew Mortonfd073832012-07-31 16:42:40 -0700409 /*
410 * Reserve the new array if that comes from the memblock. Otherwise, we
411 * needn't do it
Gavin Shan181eb392012-05-29 15:06:50 -0700412 */
413 if (!use_slab)
Yinghai Lu29f67382012-07-11 14:02:56 -0700414 BUG_ON(memblock_reserve(addr, new_alloc_size));
Gavin Shan181eb392012-05-29 15:06:50 -0700415
416 /* Update slab flag */
417 *in_slab = use_slab;
418
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700419 return 0;
420}
421
Tejun Heo784656f92011-07-12 11:15:55 +0200422/**
423 * memblock_merge_regions - merge neighboring compatible regions
424 * @type: memblock type to scan
425 *
426 * Scan @type and merge neighboring compatible regions.
427 */
428static void __init_memblock memblock_merge_regions(struct memblock_type *type)
429{
430 int i = 0;
431
432 /* cnt never goes below 1 */
433 while (i < type->cnt - 1) {
434 struct memblock_region *this = &type->regions[i];
435 struct memblock_region *next = &type->regions[i + 1];
436
Tejun Heo7c0caeb2011-07-14 11:43:42 +0200437 if (this->base + this->size != next->base ||
438 memblock_get_region_node(this) !=
Tang Chen66a20752014-01-21 15:49:20 -0800439 memblock_get_region_node(next) ||
440 this->flags != next->flags) {
Tejun Heo784656f92011-07-12 11:15:55 +0200441 BUG_ON(this->base + this->size > next->base);
442 i++;
443 continue;
444 }
445
446 this->size += next->size;
Lin Fengc0232ae2013-01-11 14:31:44 -0800447 /* move forward from next + 1, index of which is i + 2 */
448 memmove(next, next + 1, (type->cnt - (i + 2)) * sizeof(*next));
Tejun Heo784656f92011-07-12 11:15:55 +0200449 type->cnt--;
450 }
451}
452
453/**
454 * memblock_insert_region - insert new memblock region
Tang Chen209ff862013-04-29 15:08:41 -0700455 * @type: memblock type to insert into
456 * @idx: index for the insertion point
457 * @base: base address of the new region
458 * @size: size of the new region
459 * @nid: node id of the new region
Tang Chen66a20752014-01-21 15:49:20 -0800460 * @flags: flags of the new region
Tejun Heo784656f92011-07-12 11:15:55 +0200461 *
462 * Insert new memblock region [@base,@base+@size) into @type at @idx.
463 * @type must already have extra room to accomodate the new region.
464 */
465static void __init_memblock memblock_insert_region(struct memblock_type *type,
466 int idx, phys_addr_t base,
Tang Chen66a20752014-01-21 15:49:20 -0800467 phys_addr_t size,
468 int nid, unsigned long flags)
Tejun Heo784656f92011-07-12 11:15:55 +0200469{
470 struct memblock_region *rgn = &type->regions[idx];
471
472 BUG_ON(type->cnt >= type->max);
473 memmove(rgn + 1, rgn, (type->cnt - idx) * sizeof(*rgn));
474 rgn->base = base;
475 rgn->size = size;
Tang Chen66a20752014-01-21 15:49:20 -0800476 rgn->flags = flags;
Tejun Heo7c0caeb2011-07-14 11:43:42 +0200477 memblock_set_region_node(rgn, nid);
Tejun Heo784656f92011-07-12 11:15:55 +0200478 type->cnt++;
Tejun Heo1440c4e2011-12-08 10:22:08 -0800479 type->total_size += size;
Tejun Heo784656f92011-07-12 11:15:55 +0200480}
481
482/**
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100483 * memblock_add_range - add new memblock region
Tejun Heo784656f92011-07-12 11:15:55 +0200484 * @type: memblock type to add new region into
485 * @base: base address of the new region
486 * @size: size of the new region
Tejun Heo7fb0bc32011-12-08 10:22:08 -0800487 * @nid: nid of the new region
Tang Chen66a20752014-01-21 15:49:20 -0800488 * @flags: flags of the new region
Tejun Heo784656f92011-07-12 11:15:55 +0200489 *
490 * Add new memblock region [@base,@base+@size) into @type. The new region
491 * is allowed to overlap with existing ones - overlaps don't affect already
492 * existing regions. @type is guaranteed to be minimal (all neighbouring
493 * compatible regions are merged) after the addition.
494 *
495 * RETURNS:
496 * 0 on success, -errno on failure.
497 */
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100498int __init_memblock memblock_add_range(struct memblock_type *type,
Tang Chen66a20752014-01-21 15:49:20 -0800499 phys_addr_t base, phys_addr_t size,
500 int nid, unsigned long flags)
Yinghai Lu95f72d12010-07-12 14:36:09 +1000501{
Tejun Heo784656f92011-07-12 11:15:55 +0200502 bool insert = false;
Tejun Heoeb18f1b2011-12-08 10:22:07 -0800503 phys_addr_t obase = base;
504 phys_addr_t end = base + memblock_cap_size(base, &size);
Tejun Heo784656f92011-07-12 11:15:55 +0200505 int i, nr_new;
Yinghai Lu95f72d12010-07-12 14:36:09 +1000506
Tejun Heob3dc6272012-04-20 08:31:34 -0700507 if (!size)
508 return 0;
509
Tejun Heo784656f92011-07-12 11:15:55 +0200510 /* special case for empty array */
511 if (type->regions[0].size == 0) {
Tejun Heo1440c4e2011-12-08 10:22:08 -0800512 WARN_ON(type->cnt != 1 || type->total_size);
Benjamin Herrenschmidte3239ff2010-08-04 14:06:41 +1000513 type->regions[0].base = base;
514 type->regions[0].size = size;
Tang Chen66a20752014-01-21 15:49:20 -0800515 type->regions[0].flags = flags;
Tejun Heo7fb0bc32011-12-08 10:22:08 -0800516 memblock_set_region_node(&type->regions[0], nid);
Tejun Heo1440c4e2011-12-08 10:22:08 -0800517 type->total_size = size;
Yinghai Lu95f72d12010-07-12 14:36:09 +1000518 return 0;
519 }
Tejun Heo784656f92011-07-12 11:15:55 +0200520repeat:
521 /*
522 * The following is executed twice. Once with %false @insert and
523 * then with %true. The first counts the number of regions needed
524 * to accomodate the new area. The second actually inserts them.
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700525 */
Tejun Heo784656f92011-07-12 11:15:55 +0200526 base = obase;
527 nr_new = 0;
Yinghai Lu95f72d12010-07-12 14:36:09 +1000528
Tejun Heo784656f92011-07-12 11:15:55 +0200529 for (i = 0; i < type->cnt; i++) {
530 struct memblock_region *rgn = &type->regions[i];
531 phys_addr_t rbase = rgn->base;
532 phys_addr_t rend = rbase + rgn->size;
533
534 if (rbase >= end)
Yinghai Lu95f72d12010-07-12 14:36:09 +1000535 break;
Tejun Heo784656f92011-07-12 11:15:55 +0200536 if (rend <= base)
537 continue;
538 /*
539 * @rgn overlaps. If it separates the lower part of new
540 * area, insert that portion.
541 */
542 if (rbase > base) {
543 nr_new++;
544 if (insert)
545 memblock_insert_region(type, i++, base,
Tang Chen66a20752014-01-21 15:49:20 -0800546 rbase - base, nid,
547 flags);
Yinghai Lu95f72d12010-07-12 14:36:09 +1000548 }
Tejun Heo784656f92011-07-12 11:15:55 +0200549 /* area below @rend is dealt with, forget about it */
550 base = min(rend, end);
Yinghai Lu95f72d12010-07-12 14:36:09 +1000551 }
Yinghai Lu95f72d12010-07-12 14:36:09 +1000552
Tejun Heo784656f92011-07-12 11:15:55 +0200553 /* insert the remaining portion */
554 if (base < end) {
555 nr_new++;
556 if (insert)
Tang Chen66a20752014-01-21 15:49:20 -0800557 memblock_insert_region(type, i, base, end - base,
558 nid, flags);
Tejun Heo784656f92011-07-12 11:15:55 +0200559 }
560
561 /*
562 * If this was the first round, resize array and repeat for actual
563 * insertions; otherwise, merge and return.
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700564 */
Tejun Heo784656f92011-07-12 11:15:55 +0200565 if (!insert) {
566 while (type->cnt + nr_new > type->max)
Greg Pearson48c3b582012-06-20 12:53:05 -0700567 if (memblock_double_array(type, obase, size) < 0)
Tejun Heo784656f92011-07-12 11:15:55 +0200568 return -ENOMEM;
569 insert = true;
570 goto repeat;
571 } else {
572 memblock_merge_regions(type);
573 return 0;
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -0700574 }
Yinghai Lu95f72d12010-07-12 14:36:09 +1000575}
576
Tejun Heo7fb0bc32011-12-08 10:22:08 -0800577int __init_memblock memblock_add_node(phys_addr_t base, phys_addr_t size,
578 int nid)
579{
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100580 return memblock_add_range(&memblock.memory, base, size, nid, 0);
Tejun Heo7fb0bc32011-12-08 10:22:08 -0800581}
582
Tejun Heo581adcb2011-12-08 10:22:06 -0800583int __init_memblock memblock_add(phys_addr_t base, phys_addr_t size)
Yinghai Lu95f72d12010-07-12 14:36:09 +1000584{
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100585 return memblock_add_range(&memblock.memory, base, size,
Tang Chen66a20752014-01-21 15:49:20 -0800586 MAX_NUMNODES, 0);
Yinghai Lu95f72d12010-07-12 14:36:09 +1000587}
588
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800589/**
590 * memblock_isolate_range - isolate given range into disjoint memblocks
591 * @type: memblock type to isolate range for
592 * @base: base of range to isolate
593 * @size: size of range to isolate
594 * @start_rgn: out parameter for the start of isolated region
595 * @end_rgn: out parameter for the end of isolated region
596 *
597 * Walk @type and ensure that regions don't cross the boundaries defined by
598 * [@base,@base+@size). Crossing regions are split at the boundaries,
599 * which may create at most two more regions. The index of the first
600 * region inside the range is returned in *@start_rgn and end in *@end_rgn.
601 *
602 * RETURNS:
603 * 0 on success, -errno on failure.
604 */
605static int __init_memblock memblock_isolate_range(struct memblock_type *type,
606 phys_addr_t base, phys_addr_t size,
607 int *start_rgn, int *end_rgn)
608{
Tejun Heoeb18f1b2011-12-08 10:22:07 -0800609 phys_addr_t end = base + memblock_cap_size(base, &size);
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800610 int i;
611
612 *start_rgn = *end_rgn = 0;
613
Tejun Heob3dc6272012-04-20 08:31:34 -0700614 if (!size)
615 return 0;
616
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800617 /* we'll create at most two more regions */
618 while (type->cnt + 2 > type->max)
Greg Pearson48c3b582012-06-20 12:53:05 -0700619 if (memblock_double_array(type, base, size) < 0)
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800620 return -ENOMEM;
621
622 for (i = 0; i < type->cnt; i++) {
623 struct memblock_region *rgn = &type->regions[i];
624 phys_addr_t rbase = rgn->base;
625 phys_addr_t rend = rbase + rgn->size;
626
627 if (rbase >= end)
628 break;
629 if (rend <= base)
630 continue;
631
632 if (rbase < base) {
633 /*
634 * @rgn intersects from below. Split and continue
635 * to process the next region - the new top half.
636 */
637 rgn->base = base;
Tejun Heo1440c4e2011-12-08 10:22:08 -0800638 rgn->size -= base - rbase;
639 type->total_size -= base - rbase;
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800640 memblock_insert_region(type, i, rbase, base - rbase,
Tang Chen66a20752014-01-21 15:49:20 -0800641 memblock_get_region_node(rgn),
642 rgn->flags);
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800643 } else if (rend > end) {
644 /*
645 * @rgn intersects from above. Split and redo the
646 * current region - the new bottom half.
647 */
648 rgn->base = end;
Tejun Heo1440c4e2011-12-08 10:22:08 -0800649 rgn->size -= end - rbase;
650 type->total_size -= end - rbase;
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800651 memblock_insert_region(type, i--, rbase, end - rbase,
Tang Chen66a20752014-01-21 15:49:20 -0800652 memblock_get_region_node(rgn),
653 rgn->flags);
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800654 } else {
655 /* @rgn is fully contained, record it */
656 if (!*end_rgn)
657 *start_rgn = i;
658 *end_rgn = i + 1;
659 }
660 }
661
662 return 0;
663}
Tejun Heo6a9ceb32011-12-08 10:22:07 -0800664
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100665int __init_memblock memblock_remove_range(struct memblock_type *type,
666 phys_addr_t base, phys_addr_t size)
Yinghai Lu95f72d12010-07-12 14:36:09 +1000667{
Tejun Heo71936182011-12-08 10:22:07 -0800668 int start_rgn, end_rgn;
669 int i, ret;
Yinghai Lu95f72d12010-07-12 14:36:09 +1000670
Tejun Heo71936182011-12-08 10:22:07 -0800671 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
672 if (ret)
673 return ret;
Yinghai Lu95f72d12010-07-12 14:36:09 +1000674
Tejun Heo71936182011-12-08 10:22:07 -0800675 for (i = end_rgn - 1; i >= start_rgn; i--)
676 memblock_remove_region(type, i);
Benjamin Herrenschmidt8f7a6602011-03-22 16:33:43 -0700677 return 0;
Yinghai Lu95f72d12010-07-12 14:36:09 +1000678}
679
Tejun Heo581adcb2011-12-08 10:22:06 -0800680int __init_memblock memblock_remove(phys_addr_t base, phys_addr_t size)
Yinghai Lu95f72d12010-07-12 14:36:09 +1000681{
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100682 return memblock_remove_range(&memblock.memory, base, size);
Yinghai Lu95f72d12010-07-12 14:36:09 +1000683}
684
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100685
Tejun Heo581adcb2011-12-08 10:22:06 -0800686int __init_memblock memblock_free(phys_addr_t base, phys_addr_t size)
Yinghai Lu95f72d12010-07-12 14:36:09 +1000687{
Tejun Heo24aa0782011-07-12 11:16:06 +0200688 memblock_dbg(" memblock_free: [%#016llx-%#016llx] %pF\n",
H. Peter Anvina1504392011-07-14 11:57:10 -0700689 (unsigned long long)base,
Grygorii Strashko931d13f2014-01-21 15:49:17 -0800690 (unsigned long long)base + size - 1,
H. Peter Anvina1504392011-07-14 11:57:10 -0700691 (void *)_RET_IP_);
Tejun Heo24aa0782011-07-12 11:16:06 +0200692
Catalin Marinasaedf95e2014-06-06 14:38:20 -0700693 kmemleak_free_part(__va(base), size);
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100694 return memblock_remove_range(&memblock.reserved, base, size);
Yinghai Lu95f72d12010-07-12 14:36:09 +1000695}
696
Tang Chen66a20752014-01-21 15:49:20 -0800697static int __init_memblock memblock_reserve_region(phys_addr_t base,
698 phys_addr_t size,
699 int nid,
700 unsigned long flags)
Yinghai Lu95f72d12010-07-12 14:36:09 +1000701{
Benjamin Herrenschmidte3239ff2010-08-04 14:06:41 +1000702 struct memblock_type *_rgn = &memblock.reserved;
Yinghai Lu95f72d12010-07-12 14:36:09 +1000703
Tang Chen66a20752014-01-21 15:49:20 -0800704 memblock_dbg("memblock_reserve: [%#016llx-%#016llx] flags %#02lx %pF\n",
H. Peter Anvina1504392011-07-14 11:57:10 -0700705 (unsigned long long)base,
Grygorii Strashko931d13f2014-01-21 15:49:17 -0800706 (unsigned long long)base + size - 1,
Tang Chen66a20752014-01-21 15:49:20 -0800707 flags, (void *)_RET_IP_);
Yinghai Lu95f72d12010-07-12 14:36:09 +1000708
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100709 return memblock_add_range(_rgn, base, size, nid, flags);
Tang Chen66a20752014-01-21 15:49:20 -0800710}
711
712int __init_memblock memblock_reserve(phys_addr_t base, phys_addr_t size)
713{
714 return memblock_reserve_region(base, size, MAX_NUMNODES, 0);
Yinghai Lu95f72d12010-07-12 14:36:09 +1000715}
716
Tejun Heo35fd0802011-07-12 11:15:59 +0200717/**
Tang Chen66b16ed2014-01-21 15:49:23 -0800718 * memblock_mark_hotplug - Mark hotpluggable memory with flag MEMBLOCK_HOTPLUG.
719 * @base: the base phys addr of the region
720 * @size: the size of the region
721 *
722 * This function isolates region [@base, @base + @size), and mark it with flag
723 * MEMBLOCK_HOTPLUG.
724 *
725 * Return 0 on succees, -errno on failure.
726 */
727int __init_memblock memblock_mark_hotplug(phys_addr_t base, phys_addr_t size)
728{
729 struct memblock_type *type = &memblock.memory;
730 int i, ret, start_rgn, end_rgn;
731
732 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
733 if (ret)
734 return ret;
735
736 for (i = start_rgn; i < end_rgn; i++)
737 memblock_set_region_flags(&type->regions[i], MEMBLOCK_HOTPLUG);
738
739 memblock_merge_regions(type);
740 return 0;
741}
742
743/**
744 * memblock_clear_hotplug - Clear flag MEMBLOCK_HOTPLUG for a specified region.
745 * @base: the base phys addr of the region
746 * @size: the size of the region
747 *
748 * This function isolates region [@base, @base + @size), and clear flag
749 * MEMBLOCK_HOTPLUG for the isolated regions.
750 *
751 * Return 0 on succees, -errno on failure.
752 */
753int __init_memblock memblock_clear_hotplug(phys_addr_t base, phys_addr_t size)
754{
755 struct memblock_type *type = &memblock.memory;
756 int i, ret, start_rgn, end_rgn;
757
758 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
759 if (ret)
760 return ret;
761
762 for (i = start_rgn; i < end_rgn; i++)
763 memblock_clear_region_flags(&type->regions[i],
764 MEMBLOCK_HOTPLUG);
765
766 memblock_merge_regions(type);
767 return 0;
768}
769
770/**
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100771 * __next__mem_range - next function for for_each_free_mem_range() etc.
Tejun Heo35fd0802011-07-12 11:15:59 +0200772 * @idx: pointer to u64 loop variable
Grygorii Strashkob1154232014-01-21 15:50:16 -0800773 * @nid: node selector, %NUMA_NO_NODE for all nodes
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100774 * @type_a: pointer to memblock_type from where the range is taken
775 * @type_b: pointer to memblock_type which excludes memory from being taken
Wanpeng Lidad75572012-06-20 12:53:01 -0700776 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
777 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
778 * @out_nid: ptr to int for nid of the range, can be %NULL
Tejun Heo35fd0802011-07-12 11:15:59 +0200779 *
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100780 * Find the first area from *@idx which matches @nid, fill the out
Tejun Heo35fd0802011-07-12 11:15:59 +0200781 * parameters, and update *@idx for the next iteration. The lower 32bit of
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100782 * *@idx contains index into type_a and the upper 32bit indexes the
783 * areas before each region in type_b. For example, if type_b regions
Tejun Heo35fd0802011-07-12 11:15:59 +0200784 * look like the following,
785 *
786 * 0:[0-16), 1:[32-48), 2:[128-130)
787 *
788 * The upper 32bit indexes the following regions.
789 *
790 * 0:[0-0), 1:[16-32), 2:[48-128), 3:[130-MAX)
791 *
792 * As both region arrays are sorted, the function advances the two indices
793 * in lockstep and returns each intersection.
794 */
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100795void __init_memblock __next_mem_range(u64 *idx, int nid,
796 struct memblock_type *type_a,
797 struct memblock_type *type_b,
798 phys_addr_t *out_start,
799 phys_addr_t *out_end, int *out_nid)
Tejun Heo35fd0802011-07-12 11:15:59 +0200800{
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100801 int idx_a = *idx & 0xffffffff;
802 int idx_b = *idx >> 32;
Grygorii Strashkob1154232014-01-21 15:50:16 -0800803
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100804 if (WARN_ONCE(nid == MAX_NUMNODES,
805 "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
Grygorii Strashko560dca272014-01-21 15:50:55 -0800806 nid = NUMA_NO_NODE;
Tejun Heo35fd0802011-07-12 11:15:59 +0200807
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100808 for (; idx_a < type_a->cnt; idx_a++) {
809 struct memblock_region *m = &type_a->regions[idx_a];
810
Tejun Heo35fd0802011-07-12 11:15:59 +0200811 phys_addr_t m_start = m->base;
812 phys_addr_t m_end = m->base + m->size;
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100813 int m_nid = memblock_get_region_node(m);
Tejun Heo35fd0802011-07-12 11:15:59 +0200814
815 /* only memory regions are associated with nodes, check it */
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100816 if (nid != NUMA_NO_NODE && nid != m_nid)
Tejun Heo35fd0802011-07-12 11:15:59 +0200817 continue;
818
Xishi Qiu0a313a92014-09-09 14:50:46 -0700819 /* skip hotpluggable memory regions if needed */
820 if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
821 continue;
822
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100823 if (!type_b) {
824 if (out_start)
825 *out_start = m_start;
826 if (out_end)
827 *out_end = m_end;
828 if (out_nid)
829 *out_nid = m_nid;
830 idx_a++;
831 *idx = (u32)idx_a | (u64)idx_b << 32;
832 return;
833 }
Tejun Heo35fd0802011-07-12 11:15:59 +0200834
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100835 /* scan areas before each reservation */
836 for (; idx_b < type_b->cnt + 1; idx_b++) {
837 struct memblock_region *r;
838 phys_addr_t r_start;
839 phys_addr_t r_end;
840
841 r = &type_b->regions[idx_b];
842 r_start = idx_b ? r[-1].base + r[-1].size : 0;
843 r_end = idx_b < type_b->cnt ?
844 r->base : ULLONG_MAX;
845
846 /*
847 * if idx_b advanced past idx_a,
848 * break out to advance idx_a
849 */
Tejun Heo35fd0802011-07-12 11:15:59 +0200850 if (r_start >= m_end)
851 break;
852 /* if the two regions intersect, we're done */
853 if (m_start < r_end) {
854 if (out_start)
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100855 *out_start =
856 max(m_start, r_start);
Tejun Heo35fd0802011-07-12 11:15:59 +0200857 if (out_end)
858 *out_end = min(m_end, r_end);
859 if (out_nid)
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100860 *out_nid = m_nid;
Tejun Heo35fd0802011-07-12 11:15:59 +0200861 /*
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100862 * The region which ends first is
863 * advanced for the next iteration.
Tejun Heo35fd0802011-07-12 11:15:59 +0200864 */
865 if (m_end <= r_end)
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100866 idx_a++;
Tejun Heo35fd0802011-07-12 11:15:59 +0200867 else
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100868 idx_b++;
869 *idx = (u32)idx_a | (u64)idx_b << 32;
Tejun Heo35fd0802011-07-12 11:15:59 +0200870 return;
871 }
872 }
873 }
874
875 /* signal end of iteration */
876 *idx = ULLONG_MAX;
877}
878
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800879/**
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100880 * __next_mem_range_rev - generic next function for for_each_*_range_rev()
881 *
882 * Finds the next range from type_a which is not marked as unsuitable
883 * in type_b.
884 *
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800885 * @idx: pointer to u64 loop variable
Grygorii Strashkob1154232014-01-21 15:50:16 -0800886 * @nid: nid: node selector, %NUMA_NO_NODE for all nodes
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100887 * @type_a: pointer to memblock_type from where the range is taken
888 * @type_b: pointer to memblock_type which excludes memory from being taken
Wanpeng Lidad75572012-06-20 12:53:01 -0700889 * @out_start: ptr to phys_addr_t for start address of the range, can be %NULL
890 * @out_end: ptr to phys_addr_t for end address of the range, can be %NULL
891 * @out_nid: ptr to int for nid of the range, can be %NULL
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800892 *
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100893 * Reverse of __next_mem_range().
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800894 */
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100895void __init_memblock __next_mem_range_rev(u64 *idx, int nid,
896 struct memblock_type *type_a,
897 struct memblock_type *type_b,
898 phys_addr_t *out_start,
899 phys_addr_t *out_end, int *out_nid)
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800900{
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100901 int idx_a = *idx & 0xffffffff;
902 int idx_b = *idx >> 32;
Grygorii Strashkob1154232014-01-21 15:50:16 -0800903
Grygorii Strashko560dca272014-01-21 15:50:55 -0800904 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
905 nid = NUMA_NO_NODE;
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800906
907 if (*idx == (u64)ULLONG_MAX) {
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100908 idx_a = type_a->cnt - 1;
909 idx_b = type_b->cnt;
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800910 }
911
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100912 for (; idx_a >= 0; idx_a--) {
913 struct memblock_region *m = &type_a->regions[idx_a];
914
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800915 phys_addr_t m_start = m->base;
916 phys_addr_t m_end = m->base + m->size;
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100917 int m_nid = memblock_get_region_node(m);
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800918
919 /* only memory regions are associated with nodes, check it */
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100920 if (nid != NUMA_NO_NODE && nid != m_nid)
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800921 continue;
922
Tang Chen55ac5902014-01-21 15:49:35 -0800923 /* skip hotpluggable memory regions if needed */
924 if (movable_node_is_enabled() && memblock_is_hotpluggable(m))
925 continue;
926
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100927 if (!type_b) {
928 if (out_start)
929 *out_start = m_start;
930 if (out_end)
931 *out_end = m_end;
932 if (out_nid)
933 *out_nid = m_nid;
934 idx_a++;
935 *idx = (u32)idx_a | (u64)idx_b << 32;
936 return;
937 }
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800938
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100939 /* scan areas before each reservation */
940 for (; idx_b >= 0; idx_b--) {
941 struct memblock_region *r;
942 phys_addr_t r_start;
943 phys_addr_t r_end;
944
945 r = &type_b->regions[idx_b];
946 r_start = idx_b ? r[-1].base + r[-1].size : 0;
947 r_end = idx_b < type_b->cnt ?
948 r->base : ULLONG_MAX;
949 /*
950 * if idx_b advanced past idx_a,
951 * break out to advance idx_a
952 */
953
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800954 if (r_end <= m_start)
955 break;
956 /* if the two regions intersect, we're done */
957 if (m_end > r_start) {
958 if (out_start)
959 *out_start = max(m_start, r_start);
960 if (out_end)
961 *out_end = min(m_end, r_end);
962 if (out_nid)
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100963 *out_nid = m_nid;
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800964 if (m_start >= r_start)
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100965 idx_a--;
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800966 else
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100967 idx_b--;
968 *idx = (u32)idx_a | (u64)idx_b << 32;
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800969 return;
970 }
971 }
972 }
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +0100973 /* signal end of iteration */
Tejun Heo7bd0b0f2011-12-08 10:22:09 -0800974 *idx = ULLONG_MAX;
975}
976
Tejun Heo7c0caeb2011-07-14 11:43:42 +0200977#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
978/*
979 * Common iterator interface used to define for_each_mem_range().
980 */
981void __init_memblock __next_mem_pfn_range(int *idx, int nid,
982 unsigned long *out_start_pfn,
983 unsigned long *out_end_pfn, int *out_nid)
984{
985 struct memblock_type *type = &memblock.memory;
986 struct memblock_region *r;
987
988 while (++*idx < type->cnt) {
989 r = &type->regions[*idx];
990
991 if (PFN_UP(r->base) >= PFN_DOWN(r->base + r->size))
992 continue;
993 if (nid == MAX_NUMNODES || nid == r->nid)
994 break;
995 }
996 if (*idx >= type->cnt) {
997 *idx = -1;
998 return;
999 }
1000
1001 if (out_start_pfn)
1002 *out_start_pfn = PFN_UP(r->base);
1003 if (out_end_pfn)
1004 *out_end_pfn = PFN_DOWN(r->base + r->size);
1005 if (out_nid)
1006 *out_nid = r->nid;
1007}
1008
1009/**
1010 * memblock_set_node - set node ID on memblock regions
1011 * @base: base of area to set node ID for
1012 * @size: size of area to set node ID for
Tang Chene7e8de52014-01-21 15:49:26 -08001013 * @type: memblock type to set node ID for
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001014 * @nid: node ID to set
1015 *
Tang Chene7e8de52014-01-21 15:49:26 -08001016 * Set the nid of memblock @type regions in [@base,@base+@size) to @nid.
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001017 * Regions which cross the area boundaries are split as necessary.
1018 *
1019 * RETURNS:
1020 * 0 on success, -errno on failure.
1021 */
1022int __init_memblock memblock_set_node(phys_addr_t base, phys_addr_t size,
Tang Chene7e8de52014-01-21 15:49:26 -08001023 struct memblock_type *type, int nid)
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001024{
Tejun Heo6a9ceb32011-12-08 10:22:07 -08001025 int start_rgn, end_rgn;
1026 int i, ret;
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001027
Tejun Heo6a9ceb32011-12-08 10:22:07 -08001028 ret = memblock_isolate_range(type, base, size, &start_rgn, &end_rgn);
1029 if (ret)
1030 return ret;
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001031
Tejun Heo6a9ceb32011-12-08 10:22:07 -08001032 for (i = start_rgn; i < end_rgn; i++)
Wanpeng Lie9d24ad2012-10-08 16:32:21 -07001033 memblock_set_region_node(&type->regions[i], nid);
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001034
1035 memblock_merge_regions(type);
1036 return 0;
1037}
1038#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1039
Akinobu Mita2bfc2862014-06-04 16:06:53 -07001040static phys_addr_t __init memblock_alloc_range_nid(phys_addr_t size,
1041 phys_addr_t align, phys_addr_t start,
1042 phys_addr_t end, int nid)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001043{
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001044 phys_addr_t found;
Yinghai Lu95f72d12010-07-12 14:36:09 +10001045
Grygorii Strashko79f40fa2014-01-21 15:50:12 -08001046 if (!align)
1047 align = SMP_CACHE_BYTES;
Vineet Gupta94f3d3a2013-04-29 15:06:15 -07001048
Akinobu Mita2bfc2862014-06-04 16:06:53 -07001049 found = memblock_find_in_range_node(size, align, start, end, nid);
Catalin Marinasaedf95e2014-06-06 14:38:20 -07001050 if (found && !memblock_reserve(found, size)) {
1051 /*
1052 * The min_count is set to 0 so that memblock allocations are
1053 * never reported as leaks.
1054 */
1055 kmemleak_alloc(__va(found), size, 0, 0);
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001056 return found;
Catalin Marinasaedf95e2014-06-06 14:38:20 -07001057 }
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001058 return 0;
Yinghai Lu95f72d12010-07-12 14:36:09 +10001059}
1060
Akinobu Mita2bfc2862014-06-04 16:06:53 -07001061phys_addr_t __init memblock_alloc_range(phys_addr_t size, phys_addr_t align,
1062 phys_addr_t start, phys_addr_t end)
1063{
1064 return memblock_alloc_range_nid(size, align, start, end, NUMA_NO_NODE);
1065}
1066
1067static phys_addr_t __init memblock_alloc_base_nid(phys_addr_t size,
1068 phys_addr_t align, phys_addr_t max_addr,
1069 int nid)
1070{
1071 return memblock_alloc_range_nid(size, align, 0, max_addr, nid);
1072}
1073
Tejun Heo7bd0b0f2011-12-08 10:22:09 -08001074phys_addr_t __init memblock_alloc_nid(phys_addr_t size, phys_addr_t align, int nid)
1075{
1076 return memblock_alloc_base_nid(size, align, MEMBLOCK_ALLOC_ACCESSIBLE, nid);
1077}
1078
1079phys_addr_t __init __memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
1080{
Grygorii Strashkob1154232014-01-21 15:50:16 -08001081 return memblock_alloc_base_nid(size, align, max_addr, NUMA_NO_NODE);
Tejun Heo7bd0b0f2011-12-08 10:22:09 -08001082}
1083
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001084phys_addr_t __init memblock_alloc_base(phys_addr_t size, phys_addr_t align, phys_addr_t max_addr)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001085{
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001086 phys_addr_t alloc;
1087
1088 alloc = __memblock_alloc_base(size, align, max_addr);
1089
1090 if (alloc == 0)
1091 panic("ERROR: Failed to allocate 0x%llx bytes below 0x%llx.\n",
1092 (unsigned long long) size, (unsigned long long) max_addr);
1093
1094 return alloc;
Yinghai Lu95f72d12010-07-12 14:36:09 +10001095}
1096
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001097phys_addr_t __init memblock_alloc(phys_addr_t size, phys_addr_t align)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001098{
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001099 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
Yinghai Lu95f72d12010-07-12 14:36:09 +10001100}
1101
Benjamin Herrenschmidt9d1e2492010-07-06 15:39:17 -07001102phys_addr_t __init memblock_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid)
1103{
1104 phys_addr_t res = memblock_alloc_nid(size, align, nid);
1105
1106 if (res)
1107 return res;
Tejun Heo15fb0972011-07-12 09:58:07 +02001108 return memblock_alloc_base(size, align, MEMBLOCK_ALLOC_ACCESSIBLE);
Yinghai Lu95f72d12010-07-12 14:36:09 +10001109}
1110
Santosh Shilimkar26f09e92014-01-21 15:50:19 -08001111/**
1112 * memblock_virt_alloc_internal - allocate boot memory block
1113 * @size: size of memory block to be allocated in bytes
1114 * @align: alignment of the region and block's size
1115 * @min_addr: the lower bound of the memory region to allocate (phys address)
1116 * @max_addr: the upper bound of the memory region to allocate (phys address)
1117 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1118 *
1119 * The @min_addr limit is dropped if it can not be satisfied and the allocation
1120 * will fall back to memory below @min_addr. Also, allocation may fall back
1121 * to any node in the system if the specified node can not
1122 * hold the requested memory.
1123 *
1124 * The allocation is performed from memory region limited by
1125 * memblock.current_limit if @max_addr == %BOOTMEM_ALLOC_ACCESSIBLE.
1126 *
1127 * The memory block is aligned on SMP_CACHE_BYTES if @align == 0.
1128 *
1129 * The phys address of allocated boot memory block is converted to virtual and
1130 * allocated memory is reset to 0.
1131 *
1132 * In addition, function sets the min_count to 0 using kmemleak_alloc for
1133 * allocated boot memory block, so that it is never reported as leaks.
1134 *
1135 * RETURNS:
1136 * Virtual address of allocated memory block on success, NULL on failure.
1137 */
1138static void * __init memblock_virt_alloc_internal(
1139 phys_addr_t size, phys_addr_t align,
1140 phys_addr_t min_addr, phys_addr_t max_addr,
1141 int nid)
1142{
1143 phys_addr_t alloc;
1144 void *ptr;
1145
Grygorii Strashko560dca272014-01-21 15:50:55 -08001146 if (WARN_ONCE(nid == MAX_NUMNODES, "Usage of MAX_NUMNODES is deprecated. Use NUMA_NO_NODE instead\n"))
1147 nid = NUMA_NO_NODE;
Santosh Shilimkar26f09e92014-01-21 15:50:19 -08001148
1149 /*
1150 * Detect any accidental use of these APIs after slab is ready, as at
1151 * this moment memblock may be deinitialized already and its
1152 * internal data may be destroyed (after execution of free_all_bootmem)
1153 */
1154 if (WARN_ON_ONCE(slab_is_available()))
1155 return kzalloc_node(size, GFP_NOWAIT, nid);
1156
1157 if (!align)
1158 align = SMP_CACHE_BYTES;
1159
Yinghai Luf544e142014-01-29 14:05:52 -08001160 if (max_addr > memblock.current_limit)
1161 max_addr = memblock.current_limit;
1162
Santosh Shilimkar26f09e92014-01-21 15:50:19 -08001163again:
1164 alloc = memblock_find_in_range_node(size, align, min_addr, max_addr,
1165 nid);
1166 if (alloc)
1167 goto done;
1168
1169 if (nid != NUMA_NO_NODE) {
1170 alloc = memblock_find_in_range_node(size, align, min_addr,
1171 max_addr, NUMA_NO_NODE);
1172 if (alloc)
1173 goto done;
1174 }
1175
1176 if (min_addr) {
1177 min_addr = 0;
1178 goto again;
1179 } else {
1180 goto error;
1181 }
1182
1183done:
1184 memblock_reserve(alloc, size);
1185 ptr = phys_to_virt(alloc);
1186 memset(ptr, 0, size);
1187
1188 /*
1189 * The min_count is set to 0 so that bootmem allocated blocks
1190 * are never reported as leaks. This is because many of these blocks
1191 * are only referred via the physical address which is not
1192 * looked up by kmemleak.
1193 */
1194 kmemleak_alloc(ptr, size, 0, 0);
1195
1196 return ptr;
1197
1198error:
1199 return NULL;
1200}
1201
1202/**
1203 * memblock_virt_alloc_try_nid_nopanic - allocate boot memory block
1204 * @size: size of memory block to be allocated in bytes
1205 * @align: alignment of the region and block's size
1206 * @min_addr: the lower bound of the memory region from where the allocation
1207 * is preferred (phys address)
1208 * @max_addr: the upper bound of the memory region from where the allocation
1209 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1210 * allocate only from memory limited by memblock.current_limit value
1211 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1212 *
1213 * Public version of _memblock_virt_alloc_try_nid_nopanic() which provides
1214 * additional debug information (including caller info), if enabled.
1215 *
1216 * RETURNS:
1217 * Virtual address of allocated memory block on success, NULL on failure.
1218 */
1219void * __init memblock_virt_alloc_try_nid_nopanic(
1220 phys_addr_t size, phys_addr_t align,
1221 phys_addr_t min_addr, phys_addr_t max_addr,
1222 int nid)
1223{
1224 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1225 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1226 (u64)max_addr, (void *)_RET_IP_);
1227 return memblock_virt_alloc_internal(size, align, min_addr,
1228 max_addr, nid);
1229}
1230
1231/**
1232 * memblock_virt_alloc_try_nid - allocate boot memory block with panicking
1233 * @size: size of memory block to be allocated in bytes
1234 * @align: alignment of the region and block's size
1235 * @min_addr: the lower bound of the memory region from where the allocation
1236 * is preferred (phys address)
1237 * @max_addr: the upper bound of the memory region from where the allocation
1238 * is preferred (phys address), or %BOOTMEM_ALLOC_ACCESSIBLE to
1239 * allocate only from memory limited by memblock.current_limit value
1240 * @nid: nid of the free area to find, %NUMA_NO_NODE for any node
1241 *
1242 * Public panicking version of _memblock_virt_alloc_try_nid_nopanic()
1243 * which provides debug information (including caller info), if enabled,
1244 * and panics if the request can not be satisfied.
1245 *
1246 * RETURNS:
1247 * Virtual address of allocated memory block on success, NULL on failure.
1248 */
1249void * __init memblock_virt_alloc_try_nid(
1250 phys_addr_t size, phys_addr_t align,
1251 phys_addr_t min_addr, phys_addr_t max_addr,
1252 int nid)
1253{
1254 void *ptr;
1255
1256 memblock_dbg("%s: %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx %pF\n",
1257 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1258 (u64)max_addr, (void *)_RET_IP_);
1259 ptr = memblock_virt_alloc_internal(size, align,
1260 min_addr, max_addr, nid);
1261 if (ptr)
1262 return ptr;
1263
1264 panic("%s: Failed to allocate %llu bytes align=0x%llx nid=%d from=0x%llx max_addr=0x%llx\n",
1265 __func__, (u64)size, (u64)align, nid, (u64)min_addr,
1266 (u64)max_addr);
1267 return NULL;
1268}
1269
1270/**
1271 * __memblock_free_early - free boot memory block
1272 * @base: phys starting address of the boot memory block
1273 * @size: size of the boot memory block in bytes
1274 *
1275 * Free boot memory block previously allocated by memblock_virt_alloc_xx() API.
1276 * The freeing memory will not be released to the buddy allocator.
1277 */
1278void __init __memblock_free_early(phys_addr_t base, phys_addr_t size)
1279{
1280 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1281 __func__, (u64)base, (u64)base + size - 1,
1282 (void *)_RET_IP_);
1283 kmemleak_free_part(__va(base), size);
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +01001284 memblock_remove_range(&memblock.reserved, base, size);
Santosh Shilimkar26f09e92014-01-21 15:50:19 -08001285}
1286
1287/*
1288 * __memblock_free_late - free bootmem block pages directly to buddy allocator
1289 * @addr: phys starting address of the boot memory block
1290 * @size: size of the boot memory block in bytes
1291 *
1292 * This is only useful when the bootmem allocator has already been torn
1293 * down, but we are still initializing the system. Pages are released directly
1294 * to the buddy allocator, no bootmem metadata is updated because it is gone.
1295 */
1296void __init __memblock_free_late(phys_addr_t base, phys_addr_t size)
1297{
1298 u64 cursor, end;
1299
1300 memblock_dbg("%s: [%#016llx-%#016llx] %pF\n",
1301 __func__, (u64)base, (u64)base + size - 1,
1302 (void *)_RET_IP_);
1303 kmemleak_free_part(__va(base), size);
1304 cursor = PFN_UP(base);
1305 end = PFN_DOWN(base + size);
1306
1307 for (; cursor < end; cursor++) {
1308 __free_pages_bootmem(pfn_to_page(cursor), 0);
1309 totalram_pages++;
1310 }
1311}
Benjamin Herrenschmidt9d1e2492010-07-06 15:39:17 -07001312
1313/*
1314 * Remaining API functions
1315 */
1316
Benjamin Herrenschmidt2898cc42010-08-04 13:34:42 +10001317phys_addr_t __init memblock_phys_mem_size(void)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001318{
Tejun Heo1440c4e2011-12-08 10:22:08 -08001319 return memblock.memory.total_size;
Yinghai Lu95f72d12010-07-12 14:36:09 +10001320}
1321
Yinghai Lu595ad9a2013-01-24 12:20:09 -08001322phys_addr_t __init memblock_mem_size(unsigned long limit_pfn)
1323{
1324 unsigned long pages = 0;
1325 struct memblock_region *r;
1326 unsigned long start_pfn, end_pfn;
1327
1328 for_each_memblock(memory, r) {
1329 start_pfn = memblock_region_memory_base_pfn(r);
1330 end_pfn = memblock_region_memory_end_pfn(r);
1331 start_pfn = min_t(unsigned long, start_pfn, limit_pfn);
1332 end_pfn = min_t(unsigned long, end_pfn, limit_pfn);
1333 pages += end_pfn - start_pfn;
1334 }
1335
Fabian Frederick16763232014-04-07 15:37:53 -07001336 return PFN_PHYS(pages);
Yinghai Lu595ad9a2013-01-24 12:20:09 -08001337}
1338
Sam Ravnborg0a93ebe2011-10-31 17:08:16 -07001339/* lowest address */
1340phys_addr_t __init_memblock memblock_start_of_DRAM(void)
1341{
1342 return memblock.memory.regions[0].base;
1343}
1344
Yinghai Lu10d06432010-07-28 15:43:02 +10001345phys_addr_t __init_memblock memblock_end_of_DRAM(void)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001346{
1347 int idx = memblock.memory.cnt - 1;
1348
Benjamin Herrenschmidte3239ff2010-08-04 14:06:41 +10001349 return (memblock.memory.regions[idx].base + memblock.memory.regions[idx].size);
Yinghai Lu95f72d12010-07-12 14:36:09 +10001350}
1351
Tejun Heoc0ce8fe2011-12-08 10:22:07 -08001352void __init memblock_enforce_memory_limit(phys_addr_t limit)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001353{
Tejun Heoc0ce8fe2011-12-08 10:22:07 -08001354 phys_addr_t max_addr = (phys_addr_t)ULLONG_MAX;
Emil Medve136199f2014-04-07 15:37:52 -07001355 struct memblock_region *r;
Yinghai Lu95f72d12010-07-12 14:36:09 +10001356
Tejun Heoc0ce8fe2011-12-08 10:22:07 -08001357 if (!limit)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001358 return;
1359
Tejun Heoc0ce8fe2011-12-08 10:22:07 -08001360 /* find out max address */
Emil Medve136199f2014-04-07 15:37:52 -07001361 for_each_memblock(memory, r) {
Tejun Heoc0ce8fe2011-12-08 10:22:07 -08001362 if (limit <= r->size) {
1363 max_addr = r->base + limit;
1364 break;
1365 }
1366 limit -= r->size;
Yinghai Lu95f72d12010-07-12 14:36:09 +10001367 }
1368
Tejun Heoc0ce8fe2011-12-08 10:22:07 -08001369 /* truncate both memory and reserved regions */
Philipp Hachtmannf1af9d32014-01-29 18:16:01 +01001370 memblock_remove_range(&memblock.memory, max_addr,
1371 (phys_addr_t)ULLONG_MAX);
1372 memblock_remove_range(&memblock.reserved, max_addr,
1373 (phys_addr_t)ULLONG_MAX);
Yinghai Lu95f72d12010-07-12 14:36:09 +10001374}
1375
Yinghai Lucd794812010-10-11 12:34:09 -07001376static int __init_memblock memblock_search(struct memblock_type *type, phys_addr_t addr)
Benjamin Herrenschmidt72d4b0b2010-08-04 14:38:47 +10001377{
1378 unsigned int left = 0, right = type->cnt;
1379
1380 do {
1381 unsigned int mid = (right + left) / 2;
1382
1383 if (addr < type->regions[mid].base)
1384 right = mid;
1385 else if (addr >= (type->regions[mid].base +
1386 type->regions[mid].size))
1387 left = mid + 1;
1388 else
1389 return mid;
1390 } while (left < right);
1391 return -1;
1392}
1393
Benjamin Herrenschmidt2898cc42010-08-04 13:34:42 +10001394int __init memblock_is_reserved(phys_addr_t addr)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001395{
Benjamin Herrenschmidt72d4b0b2010-08-04 14:38:47 +10001396 return memblock_search(&memblock.reserved, addr) != -1;
1397}
Yinghai Lu95f72d12010-07-12 14:36:09 +10001398
Yinghai Lu3661ca62010-09-15 13:05:29 -07001399int __init_memblock memblock_is_memory(phys_addr_t addr)
Benjamin Herrenschmidt72d4b0b2010-08-04 14:38:47 +10001400{
1401 return memblock_search(&memblock.memory, addr) != -1;
1402}
1403
Yinghai Lue76b63f2013-09-11 14:22:17 -07001404#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1405int __init_memblock memblock_search_pfn_nid(unsigned long pfn,
1406 unsigned long *start_pfn, unsigned long *end_pfn)
1407{
1408 struct memblock_type *type = &memblock.memory;
Fabian Frederick16763232014-04-07 15:37:53 -07001409 int mid = memblock_search(type, PFN_PHYS(pfn));
Yinghai Lue76b63f2013-09-11 14:22:17 -07001410
1411 if (mid == -1)
1412 return -1;
1413
Fabian Frederickf7e2f7e2014-06-04 16:07:51 -07001414 *start_pfn = PFN_DOWN(type->regions[mid].base);
1415 *end_pfn = PFN_DOWN(type->regions[mid].base + type->regions[mid].size);
Yinghai Lue76b63f2013-09-11 14:22:17 -07001416
1417 return type->regions[mid].nid;
1418}
1419#endif
1420
Stephen Boydeab30942012-05-24 00:45:21 -07001421/**
1422 * memblock_is_region_memory - check if a region is a subset of memory
1423 * @base: base of region to check
1424 * @size: size of region to check
1425 *
1426 * Check if the region [@base, @base+@size) is a subset of a memory block.
1427 *
1428 * RETURNS:
1429 * 0 if false, non-zero if true
1430 */
Yinghai Lu3661ca62010-09-15 13:05:29 -07001431int __init_memblock memblock_is_region_memory(phys_addr_t base, phys_addr_t size)
Benjamin Herrenschmidt72d4b0b2010-08-04 14:38:47 +10001432{
Tomi Valkeinenabb65272011-01-20 14:44:20 -08001433 int idx = memblock_search(&memblock.memory, base);
Tejun Heoeb18f1b2011-12-08 10:22:07 -08001434 phys_addr_t end = base + memblock_cap_size(base, &size);
Benjamin Herrenschmidt72d4b0b2010-08-04 14:38:47 +10001435
1436 if (idx == -1)
1437 return 0;
Tomi Valkeinenabb65272011-01-20 14:44:20 -08001438 return memblock.memory.regions[idx].base <= base &&
1439 (memblock.memory.regions[idx].base +
Tejun Heoeb18f1b2011-12-08 10:22:07 -08001440 memblock.memory.regions[idx].size) >= end;
Yinghai Lu95f72d12010-07-12 14:36:09 +10001441}
1442
Stephen Boydeab30942012-05-24 00:45:21 -07001443/**
1444 * memblock_is_region_reserved - check if a region intersects reserved memory
1445 * @base: base of region to check
1446 * @size: size of region to check
1447 *
1448 * Check if the region [@base, @base+@size) intersects a reserved memory block.
1449 *
1450 * RETURNS:
1451 * 0 if false, non-zero if true
1452 */
Yinghai Lu10d06432010-07-28 15:43:02 +10001453int __init_memblock memblock_is_region_reserved(phys_addr_t base, phys_addr_t size)
Yinghai Lu95f72d12010-07-12 14:36:09 +10001454{
Tejun Heoeb18f1b2011-12-08 10:22:07 -08001455 memblock_cap_size(base, &size);
Benjamin Herrenschmidtf1c2c192010-08-04 14:17:17 +10001456 return memblock_overlaps_region(&memblock.reserved, base, size) >= 0;
Yinghai Lu95f72d12010-07-12 14:36:09 +10001457}
1458
Yinghai Lu6ede1fd2012-10-22 16:35:18 -07001459void __init_memblock memblock_trim_memory(phys_addr_t align)
1460{
Yinghai Lu6ede1fd2012-10-22 16:35:18 -07001461 phys_addr_t start, end, orig_start, orig_end;
Emil Medve136199f2014-04-07 15:37:52 -07001462 struct memblock_region *r;
Yinghai Lu6ede1fd2012-10-22 16:35:18 -07001463
Emil Medve136199f2014-04-07 15:37:52 -07001464 for_each_memblock(memory, r) {
1465 orig_start = r->base;
1466 orig_end = r->base + r->size;
Yinghai Lu6ede1fd2012-10-22 16:35:18 -07001467 start = round_up(orig_start, align);
1468 end = round_down(orig_end, align);
1469
1470 if (start == orig_start && end == orig_end)
1471 continue;
1472
1473 if (start < end) {
Emil Medve136199f2014-04-07 15:37:52 -07001474 r->base = start;
1475 r->size = end - start;
Yinghai Lu6ede1fd2012-10-22 16:35:18 -07001476 } else {
Emil Medve136199f2014-04-07 15:37:52 -07001477 memblock_remove_region(&memblock.memory,
1478 r - memblock.memory.regions);
1479 r--;
Yinghai Lu6ede1fd2012-10-22 16:35:18 -07001480 }
1481 }
1482}
Benjamin Herrenschmidte63075a2010-07-06 15:39:01 -07001483
Yinghai Lu3661ca62010-09-15 13:05:29 -07001484void __init_memblock memblock_set_current_limit(phys_addr_t limit)
Benjamin Herrenschmidte63075a2010-07-06 15:39:01 -07001485{
1486 memblock.current_limit = limit;
1487}
1488
Laura Abbottfec51012014-02-27 01:23:43 +01001489phys_addr_t __init_memblock memblock_get_current_limit(void)
1490{
1491 return memblock.current_limit;
1492}
1493
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001494static void __init_memblock memblock_dump(struct memblock_type *type, char *name)
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001495{
1496 unsigned long long base, size;
Tang Chen66a20752014-01-21 15:49:20 -08001497 unsigned long flags;
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001498 int i;
1499
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001500 pr_info(" %s.cnt = 0x%lx\n", name, type->cnt);
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001501
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001502 for (i = 0; i < type->cnt; i++) {
1503 struct memblock_region *rgn = &type->regions[i];
1504 char nid_buf[32] = "";
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001505
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001506 base = rgn->base;
1507 size = rgn->size;
Tang Chen66a20752014-01-21 15:49:20 -08001508 flags = rgn->flags;
Tejun Heo7c0caeb2011-07-14 11:43:42 +02001509#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1510 if (memblock_get_region_node(rgn) != MAX_NUMNODES)
1511 snprintf(nid_buf, sizeof(nid_buf), " on node %d",
1512 memblock_get_region_node(rgn));
1513#endif
Tang Chen66a20752014-01-21 15:49:20 -08001514 pr_info(" %s[%#x]\t[%#016llx-%#016llx], %#llx bytes%s flags: %#lx\n",
1515 name, i, base, base + size - 1, size, nid_buf, flags);
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001516 }
1517}
1518
Tejun Heo4ff7b822011-12-08 10:22:06 -08001519void __init_memblock __memblock_dump_all(void)
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001520{
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001521 pr_info("MEMBLOCK configuration:\n");
Tejun Heo1440c4e2011-12-08 10:22:08 -08001522 pr_info(" memory size = %#llx reserved size = %#llx\n",
1523 (unsigned long long)memblock.memory.total_size,
1524 (unsigned long long)memblock.reserved.total_size);
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001525
1526 memblock_dump(&memblock.memory, "memory");
1527 memblock_dump(&memblock.reserved, "reserved");
1528}
1529
Tejun Heo1aadc052011-12-08 10:22:08 -08001530void __init memblock_allow_resize(void)
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001531{
Benjamin Herrenschmidt142b45a2010-07-06 15:39:13 -07001532 memblock_can_resize = 1;
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001533}
1534
Benjamin Herrenschmidt6ed311b2010-07-12 14:36:48 +10001535static int __init early_memblock(char *p)
1536{
1537 if (p && strstr(p, "debug"))
1538 memblock_debug = 1;
1539 return 0;
1540}
1541early_param("memblock", early_memblock);
1542
Tejun Heoc378ddd2011-07-14 11:46:03 +02001543#if defined(CONFIG_DEBUG_FS) && !defined(CONFIG_ARCH_DISCARD_MEMBLOCK)
Benjamin Herrenschmidt6d03b882010-07-06 15:39:19 -07001544
1545static int memblock_debug_show(struct seq_file *m, void *private)
1546{
1547 struct memblock_type *type = m->private;
1548 struct memblock_region *reg;
1549 int i;
1550
1551 for (i = 0; i < type->cnt; i++) {
1552 reg = &type->regions[i];
1553 seq_printf(m, "%4d: ", i);
1554 if (sizeof(phys_addr_t) == 4)
1555 seq_printf(m, "0x%08lx..0x%08lx\n",
1556 (unsigned long)reg->base,
1557 (unsigned long)(reg->base + reg->size - 1));
1558 else
1559 seq_printf(m, "0x%016llx..0x%016llx\n",
1560 (unsigned long long)reg->base,
1561 (unsigned long long)(reg->base + reg->size - 1));
1562
1563 }
1564 return 0;
1565}
1566
1567static int memblock_debug_open(struct inode *inode, struct file *file)
1568{
1569 return single_open(file, memblock_debug_show, inode->i_private);
1570}
1571
1572static const struct file_operations memblock_debug_fops = {
1573 .open = memblock_debug_open,
1574 .read = seq_read,
1575 .llseek = seq_lseek,
1576 .release = single_release,
1577};
1578
1579static int __init memblock_init_debugfs(void)
1580{
1581 struct dentry *root = debugfs_create_dir("memblock", NULL);
1582 if (!root)
1583 return -ENXIO;
1584 debugfs_create_file("memory", S_IRUGO, root, &memblock.memory, &memblock_debug_fops);
1585 debugfs_create_file("reserved", S_IRUGO, root, &memblock.reserved, &memblock_debug_fops);
Philipp Hachtmann70210ed2014-01-29 18:16:01 +01001586#ifdef CONFIG_HAVE_MEMBLOCK_PHYS_MAP
1587 debugfs_create_file("physmem", S_IRUGO, root, &memblock.physmem, &memblock_debug_fops);
1588#endif
Benjamin Herrenschmidt6d03b882010-07-06 15:39:19 -07001589
1590 return 0;
1591}
1592__initcall(memblock_init_debugfs);
1593
1594#endif /* CONFIG_DEBUG_FS */