mm: rcu-protected get_mm_exe_file()
[linux-2.6-block.git] / mm / bootmem.c
CommitLineData
1da177e4 1/*
57cfc29e 2 * bootmem - A boot-time physical memory allocator and configurator
1da177e4
LT
3 *
4 * Copyright (C) 1999 Ingo Molnar
57cfc29e
JW
5 * 1999 Kanoj Sarcar, SGI
6 * 2008 Johannes Weiner
1da177e4 7 *
57cfc29e
JW
8 * Access to this subsystem has to be serialized externally (which is true
9 * for the boot process anyway).
1da177e4 10 */
1da177e4 11#include <linux/init.h>
bbc7b92e 12#include <linux/pfn.h>
5a0e3ad6 13#include <linux/slab.h>
1da177e4 14#include <linux/bootmem.h>
b95f1b31 15#include <linux/export.h>
ec3a354b 16#include <linux/kmemleak.h>
08677214 17#include <linux/range.h>
72d7c3b3 18#include <linux/memblock.h>
d85fbee8
PM
19#include <linux/bug.h>
20#include <linux/io.h>
e786e86a 21
dfd54cbc 22#include <asm/processor.h>
e786e86a 23
1da177e4
LT
24#include "internal.h"
25
e782ab42
YL
26#ifndef CONFIG_NEED_MULTIPLE_NODES
27struct pglist_data __refdata contig_page_data = {
28 .bdata = &bootmem_node_data[0]
29};
30EXPORT_SYMBOL(contig_page_data);
31#endif
32
1da177e4
LT
33unsigned long max_low_pfn;
34unsigned long min_low_pfn;
35unsigned long max_pfn;
36
b61bfa3c
JW
37bootmem_data_t bootmem_node_data[MAX_NUMNODES] __initdata;
38
636cc40c
JW
39static struct list_head bdata_list __initdata = LIST_HEAD_INIT(bdata_list);
40
2e5237da
JW
41static int bootmem_debug;
42
43static int __init bootmem_debug_setup(char *buf)
44{
45 bootmem_debug = 1;
46 return 0;
47}
48early_param("bootmem_debug", bootmem_debug_setup);
49
50#define bdebug(fmt, args...) ({ \
51 if (unlikely(bootmem_debug)) \
52 printk(KERN_INFO \
53 "bootmem::%s " fmt, \
80a914dc 54 __func__, ## args); \
2e5237da
JW
55})
56
df049a5f 57static unsigned long __init bootmap_bytes(unsigned long pages)
223e8dc9 58{
9571a982 59 unsigned long bytes = DIV_ROUND_UP(pages, 8);
223e8dc9 60
df049a5f 61 return ALIGN(bytes, sizeof(long));
223e8dc9
JW
62}
63
a66fd7da
JW
64/**
65 * bootmem_bootmap_pages - calculate bitmap size in pages
66 * @pages: number of pages the bitmap has to represent
67 */
f71bf0ca 68unsigned long __init bootmem_bootmap_pages(unsigned long pages)
1da177e4 69{
df049a5f 70 unsigned long bytes = bootmap_bytes(pages);
1da177e4 71
df049a5f 72 return PAGE_ALIGN(bytes) >> PAGE_SHIFT;
1da177e4 73}
f71bf0ca 74
679bc9fb
KH
75/*
76 * link bdata in order
77 */
69d49e68 78static void __init link_bootmem(bootmem_data_t *bdata)
679bc9fb 79{
5c2b8a16 80 bootmem_data_t *ent;
f71bf0ca 81
5c2b8a16
GS
82 list_for_each_entry(ent, &bdata_list, list) {
83 if (bdata->node_min_pfn < ent->node_min_pfn) {
84 list_add_tail(&bdata->list, &ent->list);
85 return;
86 }
679bc9fb 87 }
5c2b8a16
GS
88
89 list_add_tail(&bdata->list, &bdata_list);
679bc9fb
KH
90}
91
1da177e4
LT
92/*
93 * Called once to set up the allocator itself.
94 */
8ae04463 95static unsigned long __init init_bootmem_core(bootmem_data_t *bdata,
1da177e4
LT
96 unsigned long mapstart, unsigned long start, unsigned long end)
97{
bbc7b92e 98 unsigned long mapsize;
1da177e4 99
2dbb51c4 100 mminit_validate_memmodel_limits(&start, &end);
bbc7b92e 101 bdata->node_bootmem_map = phys_to_virt(PFN_PHYS(mapstart));
3560e249 102 bdata->node_min_pfn = start;
1da177e4 103 bdata->node_low_pfn = end;
679bc9fb 104 link_bootmem(bdata);
1da177e4
LT
105
106 /*
107 * Initially all pages are reserved - setup_arch() has to
108 * register free RAM areas explicitly.
109 */
df049a5f 110 mapsize = bootmap_bytes(end - start);
1da177e4
LT
111 memset(bdata->node_bootmem_map, 0xff, mapsize);
112
2e5237da
JW
113 bdebug("nid=%td start=%lx map=%lx end=%lx mapsize=%lx\n",
114 bdata - bootmem_node_data, start, mapstart, end, mapsize);
115
1da177e4
LT
116 return mapsize;
117}
118
a66fd7da
JW
119/**
120 * init_bootmem_node - register a node as boot memory
121 * @pgdat: node to register
122 * @freepfn: pfn where the bitmap for this node is to be placed
123 * @startpfn: first pfn on the node
124 * @endpfn: first pfn after the node
125 *
126 * Returns the number of bytes needed to hold the bitmap for this node.
127 */
223e8dc9
JW
128unsigned long __init init_bootmem_node(pg_data_t *pgdat, unsigned long freepfn,
129 unsigned long startpfn, unsigned long endpfn)
130{
131 return init_bootmem_core(pgdat->bdata, freepfn, startpfn, endpfn);
132}
133
a66fd7da
JW
134/**
135 * init_bootmem - register boot memory
136 * @start: pfn where the bitmap is to be placed
137 * @pages: number of available physical pages
138 *
139 * Returns the number of bytes needed to hold the bitmap.
140 */
223e8dc9
JW
141unsigned long __init init_bootmem(unsigned long start, unsigned long pages)
142{
143 max_low_pfn = pages;
144 min_low_pfn = start;
145 return init_bootmem_core(NODE_DATA(0)->bdata, start, 0, pages);
146}
09325873 147
9f993ac3
FT
148/*
149 * free_bootmem_late - free bootmem pages directly to page allocator
81df9bff 150 * @addr: starting physical address of the range
9f993ac3
FT
151 * @size: size of the range in bytes
152 *
153 * This is only useful when the bootmem allocator has already been torn
154 * down, but we are still initializing the system. Pages are given directly
155 * to the page allocator, no bootmem metadata is updated because it is gone.
156 */
81df9bff 157void __init free_bootmem_late(unsigned long physaddr, unsigned long size)
9f993ac3
FT
158{
159 unsigned long cursor, end;
160
81df9bff 161 kmemleak_free_part(__va(physaddr), size);
9f993ac3 162
81df9bff
JK
163 cursor = PFN_UP(physaddr);
164 end = PFN_DOWN(physaddr + size);
9f993ac3
FT
165
166 for (; cursor < end; cursor++) {
167 __free_pages_bootmem(pfn_to_page(cursor), 0);
168 totalram_pages++;
169 }
170}
171
223e8dc9
JW
172static unsigned long __init free_all_bootmem_core(bootmem_data_t *bdata)
173{
174 struct page *page;
4a099fb4 175 unsigned long *map, start, end, pages, count = 0;
41546c17
JW
176
177 if (!bdata->node_bootmem_map)
178 return 0;
179
4a099fb4 180 map = bdata->node_bootmem_map;
3560e249 181 start = bdata->node_min_pfn;
41546c17
JW
182 end = bdata->node_low_pfn;
183
799f933a
JW
184 bdebug("nid=%td start=%lx end=%lx\n",
185 bdata - bootmem_node_data, start, end);
223e8dc9 186
41546c17 187 while (start < end) {
4a099fb4 188 unsigned long idx, vec;
10d73e65 189 unsigned shift;
223e8dc9 190
3560e249 191 idx = start - bdata->node_min_pfn;
10d73e65
MF
192 shift = idx & (BITS_PER_LONG - 1);
193 /*
194 * vec holds at most BITS_PER_LONG map bits,
195 * bit 0 corresponds to start.
196 */
41546c17 197 vec = ~map[idx / BITS_PER_LONG];
10d73e65
MF
198
199 if (shift) {
200 vec >>= shift;
201 if (end - start >= BITS_PER_LONG)
202 vec |= ~map[idx / BITS_PER_LONG + 1] <<
203 (BITS_PER_LONG - shift);
204 }
799f933a
JW
205 /*
206 * If we have a properly aligned and fully unreserved
207 * BITS_PER_LONG block of pages in front of us, free
208 * it in one go.
209 */
210 if (IS_ALIGNED(start, BITS_PER_LONG) && vec == ~0UL) {
41546c17
JW
211 int order = ilog2(BITS_PER_LONG);
212
213 __free_pages_bootmem(pfn_to_page(start), order);
223e8dc9 214 count += BITS_PER_LONG;
799f933a 215 start += BITS_PER_LONG;
41546c17 216 } else {
10d73e65 217 unsigned long cur = start;
41546c17 218
10d73e65
MF
219 start = ALIGN(start + 1, BITS_PER_LONG);
220 while (vec && cur != start) {
41546c17 221 if (vec & 1) {
10d73e65 222 page = pfn_to_page(cur);
223e8dc9 223 __free_pages_bootmem(page, 0);
41546c17 224 count++;
223e8dc9 225 }
41546c17 226 vec >>= 1;
10d73e65 227 ++cur;
223e8dc9 228 }
223e8dc9 229 }
223e8dc9
JW
230 }
231
223e8dc9 232 page = virt_to_page(bdata->node_bootmem_map);
3560e249 233 pages = bdata->node_low_pfn - bdata->node_min_pfn;
41546c17
JW
234 pages = bootmem_bootmap_pages(pages);
235 count += pages;
5576646f 236 while (pages--)
41546c17 237 __free_pages_bootmem(page++, 0);
223e8dc9 238
2e5237da
JW
239 bdebug("nid=%td released=%lx\n", bdata - bootmem_node_data, count);
240
223e8dc9
JW
241 return count;
242}
243
7b4b2a0d
JL
244static int reset_managed_pages_done __initdata;
245
f784a3f1 246void reset_node_managed_pages(pg_data_t *pgdat)
9feedc9d
JL
247{
248 struct zone *z;
249
9feedc9d 250 for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++)
7b4b2a0d
JL
251 z->managed_pages = 0;
252}
253
254void __init reset_all_zones_managed_pages(void)
255{
256 struct pglist_data *pgdat;
257
f784a3f1
TC
258 if (reset_managed_pages_done)
259 return;
260
7b4b2a0d
JL
261 for_each_online_pgdat(pgdat)
262 reset_node_managed_pages(pgdat);
f784a3f1 263
7b4b2a0d 264 reset_managed_pages_done = 1;
9feedc9d
JL
265}
266
a66fd7da
JW
267/**
268 * free_all_bootmem - release free pages to the buddy allocator
269 *
270 * Returns the number of pages actually released.
271 */
223e8dc9
JW
272unsigned long __init free_all_bootmem(void)
273{
aa235fc7
YL
274 unsigned long total_pages = 0;
275 bootmem_data_t *bdata;
9feedc9d 276
7b4b2a0d 277 reset_all_zones_managed_pages();
aa235fc7
YL
278
279 list_for_each_entry(bdata, &bdata_list, list)
280 total_pages += free_all_bootmem_core(bdata);
281
0c988534
JL
282 totalram_pages += total_pages;
283
aa235fc7 284 return total_pages;
223e8dc9
JW
285}
286
d747fa4b
JW
287static void __init __free(bootmem_data_t *bdata,
288 unsigned long sidx, unsigned long eidx)
289{
290 unsigned long idx;
291
292 bdebug("nid=%td start=%lx end=%lx\n", bdata - bootmem_node_data,
3560e249
JW
293 sidx + bdata->node_min_pfn,
294 eidx + bdata->node_min_pfn);
d747fa4b 295
e2bf3cae
JW
296 if (bdata->hint_idx > sidx)
297 bdata->hint_idx = sidx;
298
d747fa4b
JW
299 for (idx = sidx; idx < eidx; idx++)
300 if (!test_and_clear_bit(idx, bdata->node_bootmem_map))
301 BUG();
302}
303
304static int __init __reserve(bootmem_data_t *bdata, unsigned long sidx,
305 unsigned long eidx, int flags)
306{
307 unsigned long idx;
308 int exclusive = flags & BOOTMEM_EXCLUSIVE;
309
310 bdebug("nid=%td start=%lx end=%lx flags=%x\n",
311 bdata - bootmem_node_data,
3560e249
JW
312 sidx + bdata->node_min_pfn,
313 eidx + bdata->node_min_pfn,
d747fa4b
JW
314 flags);
315
316 for (idx = sidx; idx < eidx; idx++)
317 if (test_and_set_bit(idx, bdata->node_bootmem_map)) {
318 if (exclusive) {
319 __free(bdata, sidx, idx);
320 return -EBUSY;
321 }
322 bdebug("silent double reserve of PFN %lx\n",
3560e249 323 idx + bdata->node_min_pfn);
d747fa4b
JW
324 }
325 return 0;
326}
327
e2bf3cae
JW
328static int __init mark_bootmem_node(bootmem_data_t *bdata,
329 unsigned long start, unsigned long end,
330 int reserve, int flags)
223e8dc9
JW
331{
332 unsigned long sidx, eidx;
223e8dc9 333
e2bf3cae
JW
334 bdebug("nid=%td start=%lx end=%lx reserve=%d flags=%x\n",
335 bdata - bootmem_node_data, start, end, reserve, flags);
223e8dc9 336
3560e249 337 BUG_ON(start < bdata->node_min_pfn);
e2bf3cae 338 BUG_ON(end > bdata->node_low_pfn);
223e8dc9 339
3560e249
JW
340 sidx = start - bdata->node_min_pfn;
341 eidx = end - bdata->node_min_pfn;
223e8dc9 342
e2bf3cae
JW
343 if (reserve)
344 return __reserve(bdata, sidx, eidx, flags);
223e8dc9 345 else
e2bf3cae
JW
346 __free(bdata, sidx, eidx);
347 return 0;
348}
349
350static int __init mark_bootmem(unsigned long start, unsigned long end,
351 int reserve, int flags)
352{
353 unsigned long pos;
354 bootmem_data_t *bdata;
355
356 pos = start;
357 list_for_each_entry(bdata, &bdata_list, list) {
358 int err;
359 unsigned long max;
360
3560e249
JW
361 if (pos < bdata->node_min_pfn ||
362 pos >= bdata->node_low_pfn) {
e2bf3cae
JW
363 BUG_ON(pos != start);
364 continue;
365 }
366
367 max = min(bdata->node_low_pfn, end);
223e8dc9 368
e2bf3cae
JW
369 err = mark_bootmem_node(bdata, pos, max, reserve, flags);
370 if (reserve && err) {
371 mark_bootmem(start, pos, 0, 0);
372 return err;
373 }
223e8dc9 374
e2bf3cae
JW
375 if (max == end)
376 return 0;
377 pos = bdata->node_low_pfn;
378 }
379 BUG();
223e8dc9
JW
380}
381
a66fd7da
JW
382/**
383 * free_bootmem_node - mark a page range as usable
384 * @pgdat: node the range resides on
385 * @physaddr: starting address of the range
386 * @size: size of the range in bytes
387 *
388 * Partial pages will be considered reserved and left as they are.
389 *
e2bf3cae 390 * The range must reside completely on the specified node.
a66fd7da 391 */
223e8dc9
JW
392void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
393 unsigned long size)
394{
e2bf3cae
JW
395 unsigned long start, end;
396
ec3a354b
CM
397 kmemleak_free_part(__va(physaddr), size);
398
e2bf3cae
JW
399 start = PFN_UP(physaddr);
400 end = PFN_DOWN(physaddr + size);
401
402 mark_bootmem_node(pgdat->bdata, start, end, 0, 0);
223e8dc9
JW
403}
404
a66fd7da
JW
405/**
406 * free_bootmem - mark a page range as usable
81df9bff 407 * @addr: starting physical address of the range
a66fd7da
JW
408 * @size: size of the range in bytes
409 *
410 * Partial pages will be considered reserved and left as they are.
411 *
e2bf3cae 412 * The range must be contiguous but may span node boundaries.
a66fd7da 413 */
81df9bff 414void __init free_bootmem(unsigned long physaddr, unsigned long size)
223e8dc9 415{
e2bf3cae 416 unsigned long start, end;
a5645a61 417
81df9bff 418 kmemleak_free_part(__va(physaddr), size);
ec3a354b 419
81df9bff
JK
420 start = PFN_UP(physaddr);
421 end = PFN_DOWN(physaddr + size);
1da177e4 422
e2bf3cae 423 mark_bootmem(start, end, 0, 0);
1da177e4
LT
424}
425
a66fd7da
JW
426/**
427 * reserve_bootmem_node - mark a page range as reserved
428 * @pgdat: node the range resides on
429 * @physaddr: starting address of the range
430 * @size: size of the range in bytes
431 * @flags: reservation flags (see linux/bootmem.h)
432 *
433 * Partial pages will be reserved.
434 *
e2bf3cae 435 * The range must reside completely on the specified node.
a66fd7da 436 */
223e8dc9
JW
437int __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
438 unsigned long size, int flags)
1da177e4 439{
e2bf3cae 440 unsigned long start, end;
1da177e4 441
e2bf3cae
JW
442 start = PFN_DOWN(physaddr);
443 end = PFN_UP(physaddr + size);
444
445 return mark_bootmem_node(pgdat->bdata, start, end, 1, flags);
223e8dc9 446}
5a982cbc 447
a66fd7da 448/**
0d4ba4d7 449 * reserve_bootmem - mark a page range as reserved
a66fd7da
JW
450 * @addr: starting address of the range
451 * @size: size of the range in bytes
452 * @flags: reservation flags (see linux/bootmem.h)
453 *
454 * Partial pages will be reserved.
455 *
e2bf3cae 456 * The range must be contiguous but may span node boundaries.
a66fd7da 457 */
223e8dc9
JW
458int __init reserve_bootmem(unsigned long addr, unsigned long size,
459 int flags)
460{
e2bf3cae 461 unsigned long start, end;
1da177e4 462
e2bf3cae
JW
463 start = PFN_DOWN(addr);
464 end = PFN_UP(addr + size);
223e8dc9 465
e2bf3cae 466 return mark_bootmem(start, end, 1, flags);
1da177e4
LT
467}
468
8aa043d7
JB
469static unsigned long __init align_idx(struct bootmem_data *bdata,
470 unsigned long idx, unsigned long step)
481ebd0d
JW
471{
472 unsigned long base = bdata->node_min_pfn;
473
474 /*
475 * Align the index with respect to the node start so that the
476 * combination of both satisfies the requested alignment.
477 */
478
479 return ALIGN(base + idx, step) - base;
480}
481
8aa043d7
JB
482static unsigned long __init align_off(struct bootmem_data *bdata,
483 unsigned long off, unsigned long align)
481ebd0d
JW
484{
485 unsigned long base = PFN_PHYS(bdata->node_min_pfn);
486
487 /* Same as align_idx for byte offsets */
488
489 return ALIGN(base + off, align) - base;
490}
491
c6785b6b 492static void * __init alloc_bootmem_bdata(struct bootmem_data *bdata,
d0c4f570
TH
493 unsigned long size, unsigned long align,
494 unsigned long goal, unsigned long limit)
1da177e4 495{
0f3caba2 496 unsigned long fallback = 0;
5f2809e6
JW
497 unsigned long min, max, start, sidx, midx, step;
498
594fe1a0
JW
499 bdebug("nid=%td size=%lx [%lu pages] align=%lx goal=%lx limit=%lx\n",
500 bdata - bootmem_node_data, size, PAGE_ALIGN(size) >> PAGE_SHIFT,
501 align, goal, limit);
502
5f2809e6
JW
503 BUG_ON(!size);
504 BUG_ON(align & (align - 1));
505 BUG_ON(limit && goal + size > limit);
1da177e4 506
7c309a64
CK
507 if (!bdata->node_bootmem_map)
508 return NULL;
509
3560e249 510 min = bdata->node_min_pfn;
5f2809e6 511 max = bdata->node_low_pfn;
9a2dc04c 512
5f2809e6
JW
513 goal >>= PAGE_SHIFT;
514 limit >>= PAGE_SHIFT;
515
516 if (limit && max > limit)
517 max = limit;
518 if (max <= min)
9a2dc04c
YL
519 return NULL;
520
5f2809e6 521 step = max(align >> PAGE_SHIFT, 1UL);
281dd25c 522
5f2809e6
JW
523 if (goal && min < goal && goal < max)
524 start = ALIGN(goal, step);
525 else
526 start = ALIGN(min, step);
1da177e4 527
481ebd0d 528 sidx = start - bdata->node_min_pfn;
3560e249 529 midx = max - bdata->node_min_pfn;
1da177e4 530
5f2809e6 531 if (bdata->hint_idx > sidx) {
0f3caba2
JW
532 /*
533 * Handle the valid case of sidx being zero and still
534 * catch the fallback below.
535 */
536 fallback = sidx + 1;
481ebd0d 537 sidx = align_idx(bdata, bdata->hint_idx, step);
5f2809e6 538 }
1da177e4 539
5f2809e6
JW
540 while (1) {
541 int merge;
542 void *region;
543 unsigned long eidx, i, start_off, end_off;
544find_block:
545 sidx = find_next_zero_bit(bdata->node_bootmem_map, midx, sidx);
481ebd0d 546 sidx = align_idx(bdata, sidx, step);
5f2809e6 547 eidx = sidx + PFN_UP(size);
ad09315c 548
5f2809e6 549 if (sidx >= midx || eidx > midx)
66d43e98 550 break;
1da177e4 551
5f2809e6
JW
552 for (i = sidx; i < eidx; i++)
553 if (test_bit(i, bdata->node_bootmem_map)) {
481ebd0d 554 sidx = align_idx(bdata, i, step);
5f2809e6
JW
555 if (sidx == i)
556 sidx += step;
557 goto find_block;
558 }
1da177e4 559
627240aa 560 if (bdata->last_end_off & (PAGE_SIZE - 1) &&
5f2809e6 561 PFN_DOWN(bdata->last_end_off) + 1 == sidx)
481ebd0d 562 start_off = align_off(bdata, bdata->last_end_off, align);
5f2809e6
JW
563 else
564 start_off = PFN_PHYS(sidx);
565
566 merge = PFN_DOWN(start_off) < sidx;
567 end_off = start_off + size;
568
569 bdata->last_end_off = end_off;
570 bdata->hint_idx = PFN_UP(end_off);
571
572 /*
573 * Reserve the area now:
574 */
d747fa4b
JW
575 if (__reserve(bdata, PFN_DOWN(start_off) + merge,
576 PFN_UP(end_off), BOOTMEM_EXCLUSIVE))
577 BUG();
5f2809e6 578
3560e249
JW
579 region = phys_to_virt(PFN_PHYS(bdata->node_min_pfn) +
580 start_off);
5f2809e6 581 memset(region, 0, size);
008139d9
CM
582 /*
583 * The min_count is set to 0 so that bootmem allocated blocks
584 * are never reported as leaks.
585 */
586 kmemleak_alloc(region, size, 0, 0);
5f2809e6 587 return region;
1da177e4
LT
588 }
589
0f3caba2 590 if (fallback) {
481ebd0d 591 sidx = align_idx(bdata, fallback - 1, step);
0f3caba2
JW
592 fallback = 0;
593 goto find_block;
594 }
595
596 return NULL;
597}
598
c12ab504 599static void * __init alloc_bootmem_core(unsigned long size,
0f3caba2
JW
600 unsigned long align,
601 unsigned long goal,
602 unsigned long limit)
603{
604 bootmem_data_t *bdata;
d0c4f570 605 void *region;
0f3caba2 606
3f7dfe24
JK
607 if (WARN_ON_ONCE(slab_is_available()))
608 return kzalloc(size, GFP_NOWAIT);
0f3caba2 609
d0c4f570 610 list_for_each_entry(bdata, &bdata_list, list) {
0f3caba2
JW
611 if (goal && bdata->node_low_pfn <= PFN_DOWN(goal))
612 continue;
3560e249 613 if (limit && bdata->node_min_pfn >= PFN_DOWN(limit))
0f3caba2
JW
614 break;
615
c6785b6b 616 region = alloc_bootmem_bdata(bdata, size, align, goal, limit);
0f3caba2
JW
617 if (region)
618 return region;
619 }
620
c12ab504
JW
621 return NULL;
622}
623
624static void * __init ___alloc_bootmem_nopanic(unsigned long size,
625 unsigned long align,
626 unsigned long goal,
627 unsigned long limit)
628{
629 void *ptr;
630
631restart:
632 ptr = alloc_bootmem_core(size, align, goal, limit);
633 if (ptr)
634 return ptr;
5f2809e6
JW
635 if (goal) {
636 goal = 0;
0f3caba2 637 goto restart;
5f2809e6 638 }
2e5237da 639
5f2809e6 640 return NULL;
1da177e4
LT
641}
642
a66fd7da
JW
643/**
644 * __alloc_bootmem_nopanic - allocate boot memory without panicking
645 * @size: size of the request in bytes
646 * @align: alignment of the region
647 * @goal: preferred starting address of the region
648 *
649 * The goal is dropped if it can not be satisfied and the allocation will
650 * fall back to memory below @goal.
651 *
652 * Allocation may happen on any node in the system.
653 *
654 * Returns NULL on failure.
655 */
bb0923a6 656void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align,
0f3caba2 657 unsigned long goal)
1da177e4 658{
08677214
YL
659 unsigned long limit = 0;
660
08677214 661 return ___alloc_bootmem_nopanic(size, align, goal, limit);
0f3caba2 662}
1da177e4 663
0f3caba2
JW
664static void * __init ___alloc_bootmem(unsigned long size, unsigned long align,
665 unsigned long goal, unsigned long limit)
666{
667 void *mem = ___alloc_bootmem_nopanic(size, align, goal, limit);
668
669 if (mem)
670 return mem;
671 /*
672 * Whoops, we cannot satisfy the allocation request.
673 */
674 printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size);
675 panic("Out of memory");
a8062231
AK
676 return NULL;
677}
1da177e4 678
a66fd7da
JW
679/**
680 * __alloc_bootmem - allocate boot memory
681 * @size: size of the request in bytes
682 * @align: alignment of the region
683 * @goal: preferred starting address of the region
684 *
685 * The goal is dropped if it can not be satisfied and the allocation will
686 * fall back to memory below @goal.
687 *
688 * Allocation may happen on any node in the system.
689 *
690 * The function panics if the request can not be satisfied.
691 */
bb0923a6
FBH
692void * __init __alloc_bootmem(unsigned long size, unsigned long align,
693 unsigned long goal)
a8062231 694{
08677214
YL
695 unsigned long limit = 0;
696
08677214 697 return ___alloc_bootmem(size, align, goal, limit);
1da177e4
LT
698}
699
99ab7b19 700void * __init ___alloc_bootmem_node_nopanic(pg_data_t *pgdat,
4cc278b7
JW
701 unsigned long size, unsigned long align,
702 unsigned long goal, unsigned long limit)
703{
704 void *ptr;
705
3f7dfe24
JK
706 if (WARN_ON_ONCE(slab_is_available()))
707 return kzalloc(size, GFP_NOWAIT);
ab381843 708again:
d0c4f570 709
c8f4a2d0
YL
710 /* do not panic in alloc_bootmem_bdata() */
711 if (limit && goal + size > limit)
712 limit = 0;
713
e9079911 714 ptr = alloc_bootmem_bdata(pgdat->bdata, size, align, goal, limit);
4cc278b7
JW
715 if (ptr)
716 return ptr;
717
ab381843
JW
718 ptr = alloc_bootmem_core(size, align, goal, limit);
719 if (ptr)
720 return ptr;
721
722 if (goal) {
723 goal = 0;
724 goto again;
725 }
726
421456ed
JW
727 return NULL;
728}
729
730void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size,
731 unsigned long align, unsigned long goal)
732{
733 if (WARN_ON_ONCE(slab_is_available()))
734 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
735
e9079911 736 return ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0);
421456ed
JW
737}
738
e9079911 739void * __init ___alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
421456ed
JW
740 unsigned long align, unsigned long goal,
741 unsigned long limit)
742{
743 void *ptr;
744
e9079911 745 ptr = ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0);
421456ed
JW
746 if (ptr)
747 return ptr;
748
ab381843
JW
749 printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size);
750 panic("Out of memory");
751 return NULL;
4cc278b7
JW
752}
753
a66fd7da
JW
754/**
755 * __alloc_bootmem_node - allocate boot memory from a specific node
756 * @pgdat: node to allocate from
757 * @size: size of the request in bytes
758 * @align: alignment of the region
759 * @goal: preferred starting address of the region
760 *
761 * The goal is dropped if it can not be satisfied and the allocation will
762 * fall back to memory below @goal.
763 *
764 * Allocation may fall back to any node in the system if the specified node
765 * can not hold the requested memory.
766 *
767 * The function panics if the request can not be satisfied.
768 */
bb0923a6
FBH
769void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
770 unsigned long align, unsigned long goal)
1da177e4 771{
c91c4773
PE
772 if (WARN_ON_ONCE(slab_is_available()))
773 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
774
e9079911 775 return ___alloc_bootmem_node(pgdat, size, align, goal, 0);
08677214
YL
776}
777
778void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size,
779 unsigned long align, unsigned long goal)
780{
781#ifdef MAX_DMA32_PFN
782 unsigned long end_pfn;
783
784 if (WARN_ON_ONCE(slab_is_available()))
785 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
786
787 /* update goal according ...MAX_DMA32_PFN */
83285c72 788 end_pfn = pgdat_end_pfn(pgdat);
08677214
YL
789
790 if (end_pfn > MAX_DMA32_PFN + (128 >> (20 - PAGE_SHIFT)) &&
791 (goal >> PAGE_SHIFT) < MAX_DMA32_PFN) {
792 void *ptr;
793 unsigned long new_goal;
794
795 new_goal = MAX_DMA32_PFN << PAGE_SHIFT;
c6785b6b 796 ptr = alloc_bootmem_bdata(pgdat->bdata, size, align,
08677214 797 new_goal, 0);
08677214
YL
798 if (ptr)
799 return ptr;
800 }
801#endif
802
803 return __alloc_bootmem_node(pgdat, size, align, goal);
804
1da177e4
LT
805}
806
dfd54cbc
HC
807#ifndef ARCH_LOW_ADDRESS_LIMIT
808#define ARCH_LOW_ADDRESS_LIMIT 0xffffffffUL
809#endif
008857c1 810
a66fd7da
JW
811/**
812 * __alloc_bootmem_low - allocate low boot memory
813 * @size: size of the request in bytes
814 * @align: alignment of the region
815 * @goal: preferred starting address of the region
816 *
817 * The goal is dropped if it can not be satisfied and the allocation will
818 * fall back to memory below @goal.
819 *
820 * Allocation may happen on any node in the system.
821 *
822 * The function panics if the request can not be satisfied.
823 */
bb0923a6
FBH
824void * __init __alloc_bootmem_low(unsigned long size, unsigned long align,
825 unsigned long goal)
008857c1 826{
0f3caba2 827 return ___alloc_bootmem(size, align, goal, ARCH_LOW_ADDRESS_LIMIT);
008857c1
RT
828}
829
38fa4175
YL
830void * __init __alloc_bootmem_low_nopanic(unsigned long size,
831 unsigned long align,
832 unsigned long goal)
833{
834 return ___alloc_bootmem_nopanic(size, align, goal,
835 ARCH_LOW_ADDRESS_LIMIT);
836}
837
a66fd7da
JW
838/**
839 * __alloc_bootmem_low_node - allocate low boot memory from a specific node
840 * @pgdat: node to allocate from
841 * @size: size of the request in bytes
842 * @align: alignment of the region
843 * @goal: preferred starting address of the region
844 *
845 * The goal is dropped if it can not be satisfied and the allocation will
846 * fall back to memory below @goal.
847 *
848 * Allocation may fall back to any node in the system if the specified node
849 * can not hold the requested memory.
850 *
851 * The function panics if the request can not be satisfied.
852 */
008857c1
RT
853void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size,
854 unsigned long align, unsigned long goal)
855{
c91c4773
PE
856 if (WARN_ON_ONCE(slab_is_available()))
857 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
858
e9079911
JW
859 return ___alloc_bootmem_node(pgdat, size, align,
860 goal, ARCH_LOW_ADDRESS_LIMIT);
008857c1 861}