]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - mm/bootmem.c
media: dvb_ca_en50221: prevent using slot_info for Spectre attacs
[mirror_ubuntu-bionic-kernel.git] / mm / bootmem.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
1da177e4 2/*
57cfc29e 3 * bootmem - A boot-time physical memory allocator and configurator
1da177e4
LT
4 *
5 * Copyright (C) 1999 Ingo Molnar
57cfc29e
JW
6 * 1999 Kanoj Sarcar, SGI
7 * 2008 Johannes Weiner
1da177e4 8 *
57cfc29e
JW
9 * Access to this subsystem has to be serialized externally (which is true
10 * for the boot process anyway).
1da177e4 11 */
1da177e4 12#include <linux/init.h>
bbc7b92e 13#include <linux/pfn.h>
5a0e3ad6 14#include <linux/slab.h>
b95f1b31 15#include <linux/export.h>
ec3a354b 16#include <linux/kmemleak.h>
08677214 17#include <linux/range.h>
d85fbee8
PM
18#include <linux/bug.h>
19#include <linux/io.h>
1d8bf926 20#include <linux/bootmem.h>
e786e86a 21
1da177e4
LT
22#include "internal.h"
23
e782ab42
YL
24#ifndef CONFIG_NEED_MULTIPLE_NODES
25struct pglist_data __refdata contig_page_data = {
26 .bdata = &bootmem_node_data[0]
27};
28EXPORT_SYMBOL(contig_page_data);
29#endif
30
1da177e4
LT
31unsigned long max_low_pfn;
32unsigned long min_low_pfn;
33unsigned long max_pfn;
8dd33030 34unsigned long long max_possible_pfn;
1da177e4 35
b61bfa3c
JW
36bootmem_data_t bootmem_node_data[MAX_NUMNODES] __initdata;
37
636cc40c
JW
38static struct list_head bdata_list __initdata = LIST_HEAD_INIT(bdata_list);
39
2e5237da
JW
40static int bootmem_debug;
41
42static int __init bootmem_debug_setup(char *buf)
43{
44 bootmem_debug = 1;
45 return 0;
46}
47early_param("bootmem_debug", bootmem_debug_setup);
48
49#define bdebug(fmt, args...) ({ \
50 if (unlikely(bootmem_debug)) \
1170532b 51 pr_info("bootmem::%s " fmt, \
80a914dc 52 __func__, ## args); \
2e5237da
JW
53})
54
df049a5f 55static unsigned long __init bootmap_bytes(unsigned long pages)
223e8dc9 56{
d3a9d7a3 57 unsigned long bytes = DIV_ROUND_UP(pages, BITS_PER_BYTE);
223e8dc9 58
df049a5f 59 return ALIGN(bytes, sizeof(long));
223e8dc9
JW
60}
61
a66fd7da
JW
62/**
63 * bootmem_bootmap_pages - calculate bitmap size in pages
64 * @pages: number of pages the bitmap has to represent
65 */
f71bf0ca 66unsigned long __init bootmem_bootmap_pages(unsigned long pages)
1da177e4 67{
df049a5f 68 unsigned long bytes = bootmap_bytes(pages);
1da177e4 69
df049a5f 70 return PAGE_ALIGN(bytes) >> PAGE_SHIFT;
1da177e4 71}
f71bf0ca 72
679bc9fb
KH
73/*
74 * link bdata in order
75 */
69d49e68 76static void __init link_bootmem(bootmem_data_t *bdata)
679bc9fb 77{
5c2b8a16 78 bootmem_data_t *ent;
f71bf0ca 79
5c2b8a16
GS
80 list_for_each_entry(ent, &bdata_list, list) {
81 if (bdata->node_min_pfn < ent->node_min_pfn) {
82 list_add_tail(&bdata->list, &ent->list);
83 return;
84 }
679bc9fb 85 }
5c2b8a16
GS
86
87 list_add_tail(&bdata->list, &bdata_list);
679bc9fb
KH
88}
89
1da177e4
LT
90/*
91 * Called once to set up the allocator itself.
92 */
8ae04463 93static unsigned long __init init_bootmem_core(bootmem_data_t *bdata,
1da177e4
LT
94 unsigned long mapstart, unsigned long start, unsigned long end)
95{
bbc7b92e 96 unsigned long mapsize;
1da177e4 97
2dbb51c4 98 mminit_validate_memmodel_limits(&start, &end);
bbc7b92e 99 bdata->node_bootmem_map = phys_to_virt(PFN_PHYS(mapstart));
3560e249 100 bdata->node_min_pfn = start;
1da177e4 101 bdata->node_low_pfn = end;
679bc9fb 102 link_bootmem(bdata);
1da177e4
LT
103
104 /*
105 * Initially all pages are reserved - setup_arch() has to
106 * register free RAM areas explicitly.
107 */
df049a5f 108 mapsize = bootmap_bytes(end - start);
1da177e4
LT
109 memset(bdata->node_bootmem_map, 0xff, mapsize);
110
2e5237da
JW
111 bdebug("nid=%td start=%lx map=%lx end=%lx mapsize=%lx\n",
112 bdata - bootmem_node_data, start, mapstart, end, mapsize);
113
1da177e4
LT
114 return mapsize;
115}
116
a66fd7da
JW
117/**
118 * init_bootmem_node - register a node as boot memory
119 * @pgdat: node to register
120 * @freepfn: pfn where the bitmap for this node is to be placed
121 * @startpfn: first pfn on the node
122 * @endpfn: first pfn after the node
123 *
124 * Returns the number of bytes needed to hold the bitmap for this node.
125 */
223e8dc9
JW
126unsigned long __init init_bootmem_node(pg_data_t *pgdat, unsigned long freepfn,
127 unsigned long startpfn, unsigned long endpfn)
128{
129 return init_bootmem_core(pgdat->bdata, freepfn, startpfn, endpfn);
130}
131
a66fd7da
JW
132/**
133 * init_bootmem - register boot memory
134 * @start: pfn where the bitmap is to be placed
135 * @pages: number of available physical pages
136 *
137 * Returns the number of bytes needed to hold the bitmap.
138 */
223e8dc9
JW
139unsigned long __init init_bootmem(unsigned long start, unsigned long pages)
140{
141 max_low_pfn = pages;
142 min_low_pfn = start;
143 return init_bootmem_core(NODE_DATA(0)->bdata, start, 0, pages);
144}
09325873 145
9f993ac3
FT
146/*
147 * free_bootmem_late - free bootmem pages directly to page allocator
81df9bff 148 * @addr: starting physical address of the range
9f993ac3
FT
149 * @size: size of the range in bytes
150 *
151 * This is only useful when the bootmem allocator has already been torn
152 * down, but we are still initializing the system. Pages are given directly
153 * to the page allocator, no bootmem metadata is updated because it is gone.
154 */
81df9bff 155void __init free_bootmem_late(unsigned long physaddr, unsigned long size)
9f993ac3
FT
156{
157 unsigned long cursor, end;
158
9099daed 159 kmemleak_free_part_phys(physaddr, size);
9f993ac3 160
81df9bff
JK
161 cursor = PFN_UP(physaddr);
162 end = PFN_DOWN(physaddr + size);
9f993ac3
FT
163
164 for (; cursor < end; cursor++) {
d70ddd7a 165 __free_pages_bootmem(pfn_to_page(cursor), cursor, 0);
9f993ac3
FT
166 totalram_pages++;
167 }
168}
169
223e8dc9
JW
170static unsigned long __init free_all_bootmem_core(bootmem_data_t *bdata)
171{
172 struct page *page;
d70ddd7a 173 unsigned long *map, start, end, pages, cur, count = 0;
41546c17
JW
174
175 if (!bdata->node_bootmem_map)
176 return 0;
177
4a099fb4 178 map = bdata->node_bootmem_map;
3560e249 179 start = bdata->node_min_pfn;
41546c17
JW
180 end = bdata->node_low_pfn;
181
799f933a
JW
182 bdebug("nid=%td start=%lx end=%lx\n",
183 bdata - bootmem_node_data, start, end);
223e8dc9 184
41546c17 185 while (start < end) {
4a099fb4 186 unsigned long idx, vec;
10d73e65 187 unsigned shift;
223e8dc9 188
3560e249 189 idx = start - bdata->node_min_pfn;
10d73e65
MF
190 shift = idx & (BITS_PER_LONG - 1);
191 /*
192 * vec holds at most BITS_PER_LONG map bits,
193 * bit 0 corresponds to start.
194 */
41546c17 195 vec = ~map[idx / BITS_PER_LONG];
10d73e65
MF
196
197 if (shift) {
198 vec >>= shift;
199 if (end - start >= BITS_PER_LONG)
200 vec |= ~map[idx / BITS_PER_LONG + 1] <<
201 (BITS_PER_LONG - shift);
202 }
799f933a
JW
203 /*
204 * If we have a properly aligned and fully unreserved
205 * BITS_PER_LONG block of pages in front of us, free
206 * it in one go.
207 */
208 if (IS_ALIGNED(start, BITS_PER_LONG) && vec == ~0UL) {
41546c17
JW
209 int order = ilog2(BITS_PER_LONG);
210
d70ddd7a 211 __free_pages_bootmem(pfn_to_page(start), start, order);
223e8dc9 212 count += BITS_PER_LONG;
799f933a 213 start += BITS_PER_LONG;
41546c17 214 } else {
d70ddd7a 215 cur = start;
41546c17 216
10d73e65
MF
217 start = ALIGN(start + 1, BITS_PER_LONG);
218 while (vec && cur != start) {
41546c17 219 if (vec & 1) {
10d73e65 220 page = pfn_to_page(cur);
d70ddd7a 221 __free_pages_bootmem(page, cur, 0);
41546c17 222 count++;
223e8dc9 223 }
41546c17 224 vec >>= 1;
10d73e65 225 ++cur;
223e8dc9 226 }
223e8dc9 227 }
223e8dc9
JW
228 }
229
d70ddd7a 230 cur = bdata->node_min_pfn;
223e8dc9 231 page = virt_to_page(bdata->node_bootmem_map);
3560e249 232 pages = bdata->node_low_pfn - bdata->node_min_pfn;
41546c17
JW
233 pages = bootmem_bootmap_pages(pages);
234 count += pages;
5576646f 235 while (pages--)
d70ddd7a 236 __free_pages_bootmem(page++, cur++, 0);
1b4ace41 237 bdata->node_bootmem_map = NULL;
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
1b4ace41
CM
296 if (WARN_ON(bdata->node_bootmem_map == NULL))
297 return;
298
e2bf3cae
JW
299 if (bdata->hint_idx > sidx)
300 bdata->hint_idx = sidx;
301
d747fa4b
JW
302 for (idx = sidx; idx < eidx; idx++)
303 if (!test_and_clear_bit(idx, bdata->node_bootmem_map))
304 BUG();
305}
306
307static int __init __reserve(bootmem_data_t *bdata, unsigned long sidx,
308 unsigned long eidx, int flags)
309{
310 unsigned long idx;
311 int exclusive = flags & BOOTMEM_EXCLUSIVE;
312
313 bdebug("nid=%td start=%lx end=%lx flags=%x\n",
314 bdata - bootmem_node_data,
3560e249
JW
315 sidx + bdata->node_min_pfn,
316 eidx + bdata->node_min_pfn,
d747fa4b
JW
317 flags);
318
1b4ace41
CM
319 if (WARN_ON(bdata->node_bootmem_map == NULL))
320 return 0;
321
d747fa4b
JW
322 for (idx = sidx; idx < eidx; idx++)
323 if (test_and_set_bit(idx, bdata->node_bootmem_map)) {
324 if (exclusive) {
325 __free(bdata, sidx, idx);
326 return -EBUSY;
327 }
328 bdebug("silent double reserve of PFN %lx\n",
3560e249 329 idx + bdata->node_min_pfn);
d747fa4b
JW
330 }
331 return 0;
332}
333
e2bf3cae
JW
334static int __init mark_bootmem_node(bootmem_data_t *bdata,
335 unsigned long start, unsigned long end,
336 int reserve, int flags)
223e8dc9
JW
337{
338 unsigned long sidx, eidx;
223e8dc9 339
e2bf3cae
JW
340 bdebug("nid=%td start=%lx end=%lx reserve=%d flags=%x\n",
341 bdata - bootmem_node_data, start, end, reserve, flags);
223e8dc9 342
3560e249 343 BUG_ON(start < bdata->node_min_pfn);
e2bf3cae 344 BUG_ON(end > bdata->node_low_pfn);
223e8dc9 345
3560e249
JW
346 sidx = start - bdata->node_min_pfn;
347 eidx = end - bdata->node_min_pfn;
223e8dc9 348
e2bf3cae
JW
349 if (reserve)
350 return __reserve(bdata, sidx, eidx, flags);
223e8dc9 351 else
e2bf3cae
JW
352 __free(bdata, sidx, eidx);
353 return 0;
354}
355
356static int __init mark_bootmem(unsigned long start, unsigned long end,
357 int reserve, int flags)
358{
359 unsigned long pos;
360 bootmem_data_t *bdata;
361
362 pos = start;
363 list_for_each_entry(bdata, &bdata_list, list) {
364 int err;
365 unsigned long max;
366
3560e249
JW
367 if (pos < bdata->node_min_pfn ||
368 pos >= bdata->node_low_pfn) {
e2bf3cae
JW
369 BUG_ON(pos != start);
370 continue;
371 }
372
373 max = min(bdata->node_low_pfn, end);
223e8dc9 374
e2bf3cae
JW
375 err = mark_bootmem_node(bdata, pos, max, reserve, flags);
376 if (reserve && err) {
377 mark_bootmem(start, pos, 0, 0);
378 return err;
379 }
223e8dc9 380
e2bf3cae
JW
381 if (max == end)
382 return 0;
383 pos = bdata->node_low_pfn;
384 }
385 BUG();
223e8dc9
JW
386}
387
a66fd7da
JW
388/**
389 * free_bootmem_node - mark a page range as usable
390 * @pgdat: node the range resides on
391 * @physaddr: starting address of the range
392 * @size: size of the range in bytes
393 *
394 * Partial pages will be considered reserved and left as they are.
395 *
e2bf3cae 396 * The range must reside completely on the specified node.
a66fd7da 397 */
223e8dc9
JW
398void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
399 unsigned long size)
400{
e2bf3cae
JW
401 unsigned long start, end;
402
9099daed 403 kmemleak_free_part_phys(physaddr, size);
ec3a354b 404
e2bf3cae
JW
405 start = PFN_UP(physaddr);
406 end = PFN_DOWN(physaddr + size);
407
408 mark_bootmem_node(pgdat->bdata, start, end, 0, 0);
223e8dc9
JW
409}
410
a66fd7da
JW
411/**
412 * free_bootmem - mark a page range as usable
81df9bff 413 * @addr: starting physical address of the range
a66fd7da
JW
414 * @size: size of the range in bytes
415 *
416 * Partial pages will be considered reserved and left as they are.
417 *
e2bf3cae 418 * The range must be contiguous but may span node boundaries.
a66fd7da 419 */
81df9bff 420void __init free_bootmem(unsigned long physaddr, unsigned long size)
223e8dc9 421{
e2bf3cae 422 unsigned long start, end;
a5645a61 423
9099daed 424 kmemleak_free_part_phys(physaddr, size);
ec3a354b 425
81df9bff
JK
426 start = PFN_UP(physaddr);
427 end = PFN_DOWN(physaddr + size);
1da177e4 428
e2bf3cae 429 mark_bootmem(start, end, 0, 0);
1da177e4
LT
430}
431
a66fd7da
JW
432/**
433 * reserve_bootmem_node - mark a page range as reserved
434 * @pgdat: node the range resides on
435 * @physaddr: starting address of the range
436 * @size: size of the range in bytes
437 * @flags: reservation flags (see linux/bootmem.h)
438 *
439 * Partial pages will be reserved.
440 *
e2bf3cae 441 * The range must reside completely on the specified node.
a66fd7da 442 */
223e8dc9
JW
443int __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr,
444 unsigned long size, int flags)
1da177e4 445{
e2bf3cae 446 unsigned long start, end;
1da177e4 447
e2bf3cae
JW
448 start = PFN_DOWN(physaddr);
449 end = PFN_UP(physaddr + size);
450
451 return mark_bootmem_node(pgdat->bdata, start, end, 1, flags);
223e8dc9 452}
5a982cbc 453
a66fd7da 454/**
0d4ba4d7 455 * reserve_bootmem - mark a page range as reserved
a66fd7da
JW
456 * @addr: starting address of the range
457 * @size: size of the range in bytes
458 * @flags: reservation flags (see linux/bootmem.h)
459 *
460 * Partial pages will be reserved.
461 *
e2bf3cae 462 * The range must be contiguous but may span node boundaries.
a66fd7da 463 */
223e8dc9
JW
464int __init reserve_bootmem(unsigned long addr, unsigned long size,
465 int flags)
466{
e2bf3cae 467 unsigned long start, end;
1da177e4 468
e2bf3cae
JW
469 start = PFN_DOWN(addr);
470 end = PFN_UP(addr + size);
223e8dc9 471
e2bf3cae 472 return mark_bootmem(start, end, 1, flags);
1da177e4
LT
473}
474
8aa043d7
JB
475static unsigned long __init align_idx(struct bootmem_data *bdata,
476 unsigned long idx, unsigned long step)
481ebd0d
JW
477{
478 unsigned long base = bdata->node_min_pfn;
479
480 /*
481 * Align the index with respect to the node start so that the
482 * combination of both satisfies the requested alignment.
483 */
484
485 return ALIGN(base + idx, step) - base;
486}
487
8aa043d7
JB
488static unsigned long __init align_off(struct bootmem_data *bdata,
489 unsigned long off, unsigned long align)
481ebd0d
JW
490{
491 unsigned long base = PFN_PHYS(bdata->node_min_pfn);
492
493 /* Same as align_idx for byte offsets */
494
495 return ALIGN(base + off, align) - base;
496}
497
c6785b6b 498static void * __init alloc_bootmem_bdata(struct bootmem_data *bdata,
d0c4f570
TH
499 unsigned long size, unsigned long align,
500 unsigned long goal, unsigned long limit)
1da177e4 501{
0f3caba2 502 unsigned long fallback = 0;
5f2809e6
JW
503 unsigned long min, max, start, sidx, midx, step;
504
594fe1a0
JW
505 bdebug("nid=%td size=%lx [%lu pages] align=%lx goal=%lx limit=%lx\n",
506 bdata - bootmem_node_data, size, PAGE_ALIGN(size) >> PAGE_SHIFT,
507 align, goal, limit);
508
5f2809e6
JW
509 BUG_ON(!size);
510 BUG_ON(align & (align - 1));
511 BUG_ON(limit && goal + size > limit);
1da177e4 512
7c309a64
CK
513 if (!bdata->node_bootmem_map)
514 return NULL;
515
3560e249 516 min = bdata->node_min_pfn;
5f2809e6 517 max = bdata->node_low_pfn;
9a2dc04c 518
5f2809e6
JW
519 goal >>= PAGE_SHIFT;
520 limit >>= PAGE_SHIFT;
521
522 if (limit && max > limit)
523 max = limit;
524 if (max <= min)
9a2dc04c
YL
525 return NULL;
526
5f2809e6 527 step = max(align >> PAGE_SHIFT, 1UL);
281dd25c 528
5f2809e6
JW
529 if (goal && min < goal && goal < max)
530 start = ALIGN(goal, step);
531 else
532 start = ALIGN(min, step);
1da177e4 533
481ebd0d 534 sidx = start - bdata->node_min_pfn;
3560e249 535 midx = max - bdata->node_min_pfn;
1da177e4 536
5f2809e6 537 if (bdata->hint_idx > sidx) {
0f3caba2
JW
538 /*
539 * Handle the valid case of sidx being zero and still
540 * catch the fallback below.
541 */
542 fallback = sidx + 1;
481ebd0d 543 sidx = align_idx(bdata, bdata->hint_idx, step);
5f2809e6 544 }
1da177e4 545
5f2809e6
JW
546 while (1) {
547 int merge;
548 void *region;
549 unsigned long eidx, i, start_off, end_off;
550find_block:
551 sidx = find_next_zero_bit(bdata->node_bootmem_map, midx, sidx);
481ebd0d 552 sidx = align_idx(bdata, sidx, step);
5f2809e6 553 eidx = sidx + PFN_UP(size);
ad09315c 554
5f2809e6 555 if (sidx >= midx || eidx > midx)
66d43e98 556 break;
1da177e4 557
5f2809e6
JW
558 for (i = sidx; i < eidx; i++)
559 if (test_bit(i, bdata->node_bootmem_map)) {
481ebd0d 560 sidx = align_idx(bdata, i, step);
5f2809e6
JW
561 if (sidx == i)
562 sidx += step;
563 goto find_block;
564 }
1da177e4 565
627240aa 566 if (bdata->last_end_off & (PAGE_SIZE - 1) &&
5f2809e6 567 PFN_DOWN(bdata->last_end_off) + 1 == sidx)
481ebd0d 568 start_off = align_off(bdata, bdata->last_end_off, align);
5f2809e6
JW
569 else
570 start_off = PFN_PHYS(sidx);
571
572 merge = PFN_DOWN(start_off) < sidx;
573 end_off = start_off + size;
574
575 bdata->last_end_off = end_off;
576 bdata->hint_idx = PFN_UP(end_off);
577
578 /*
579 * Reserve the area now:
580 */
d747fa4b
JW
581 if (__reserve(bdata, PFN_DOWN(start_off) + merge,
582 PFN_UP(end_off), BOOTMEM_EXCLUSIVE))
583 BUG();
5f2809e6 584
3560e249
JW
585 region = phys_to_virt(PFN_PHYS(bdata->node_min_pfn) +
586 start_off);
5f2809e6 587 memset(region, 0, size);
008139d9
CM
588 /*
589 * The min_count is set to 0 so that bootmem allocated blocks
590 * are never reported as leaks.
591 */
592 kmemleak_alloc(region, size, 0, 0);
5f2809e6 593 return region;
1da177e4
LT
594 }
595
0f3caba2 596 if (fallback) {
481ebd0d 597 sidx = align_idx(bdata, fallback - 1, step);
0f3caba2
JW
598 fallback = 0;
599 goto find_block;
600 }
601
602 return NULL;
603}
604
c12ab504 605static void * __init alloc_bootmem_core(unsigned long size,
0f3caba2
JW
606 unsigned long align,
607 unsigned long goal,
608 unsigned long limit)
609{
610 bootmem_data_t *bdata;
d0c4f570 611 void *region;
0f3caba2 612
3f7dfe24
JK
613 if (WARN_ON_ONCE(slab_is_available()))
614 return kzalloc(size, GFP_NOWAIT);
0f3caba2 615
d0c4f570 616 list_for_each_entry(bdata, &bdata_list, list) {
0f3caba2
JW
617 if (goal && bdata->node_low_pfn <= PFN_DOWN(goal))
618 continue;
3560e249 619 if (limit && bdata->node_min_pfn >= PFN_DOWN(limit))
0f3caba2
JW
620 break;
621
c6785b6b 622 region = alloc_bootmem_bdata(bdata, size, align, goal, limit);
0f3caba2
JW
623 if (region)
624 return region;
625 }
626
c12ab504
JW
627 return NULL;
628}
629
630static void * __init ___alloc_bootmem_nopanic(unsigned long size,
631 unsigned long align,
632 unsigned long goal,
633 unsigned long limit)
634{
635 void *ptr;
636
637restart:
638 ptr = alloc_bootmem_core(size, align, goal, limit);
639 if (ptr)
640 return ptr;
5f2809e6
JW
641 if (goal) {
642 goal = 0;
0f3caba2 643 goto restart;
5f2809e6 644 }
2e5237da 645
5f2809e6 646 return NULL;
1da177e4
LT
647}
648
a66fd7da
JW
649/**
650 * __alloc_bootmem_nopanic - allocate boot memory without panicking
651 * @size: size of the request in bytes
652 * @align: alignment of the region
653 * @goal: preferred starting address of the region
654 *
655 * The goal is dropped if it can not be satisfied and the allocation will
656 * fall back to memory below @goal.
657 *
658 * Allocation may happen on any node in the system.
659 *
660 * Returns NULL on failure.
661 */
bb0923a6 662void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align,
0f3caba2 663 unsigned long goal)
1da177e4 664{
08677214
YL
665 unsigned long limit = 0;
666
08677214 667 return ___alloc_bootmem_nopanic(size, align, goal, limit);
0f3caba2 668}
1da177e4 669
0f3caba2
JW
670static void * __init ___alloc_bootmem(unsigned long size, unsigned long align,
671 unsigned long goal, unsigned long limit)
672{
673 void *mem = ___alloc_bootmem_nopanic(size, align, goal, limit);
674
675 if (mem)
676 return mem;
677 /*
678 * Whoops, we cannot satisfy the allocation request.
679 */
1170532b 680 pr_alert("bootmem alloc of %lu bytes failed!\n", size);
0f3caba2 681 panic("Out of memory");
a8062231
AK
682 return NULL;
683}
1da177e4 684
a66fd7da
JW
685/**
686 * __alloc_bootmem - allocate boot memory
687 * @size: size of the request in bytes
688 * @align: alignment of the region
689 * @goal: preferred starting address of the region
690 *
691 * The goal is dropped if it can not be satisfied and the allocation will
692 * fall back to memory below @goal.
693 *
694 * Allocation may happen on any node in the system.
695 *
696 * The function panics if the request can not be satisfied.
697 */
bb0923a6
FBH
698void * __init __alloc_bootmem(unsigned long size, unsigned long align,
699 unsigned long goal)
a8062231 700{
08677214
YL
701 unsigned long limit = 0;
702
08677214 703 return ___alloc_bootmem(size, align, goal, limit);
1da177e4
LT
704}
705
99ab7b19 706void * __init ___alloc_bootmem_node_nopanic(pg_data_t *pgdat,
4cc278b7
JW
707 unsigned long size, unsigned long align,
708 unsigned long goal, unsigned long limit)
709{
710 void *ptr;
711
3f7dfe24 712 if (WARN_ON_ONCE(slab_is_available()))
1d8bf926 713 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
ab381843 714again:
d0c4f570 715
c8f4a2d0
YL
716 /* do not panic in alloc_bootmem_bdata() */
717 if (limit && goal + size > limit)
718 limit = 0;
719
e9079911 720 ptr = alloc_bootmem_bdata(pgdat->bdata, size, align, goal, limit);
4cc278b7
JW
721 if (ptr)
722 return ptr;
723
ab381843
JW
724 ptr = alloc_bootmem_core(size, align, goal, limit);
725 if (ptr)
726 return ptr;
727
728 if (goal) {
729 goal = 0;
730 goto again;
731 }
732
421456ed
JW
733 return NULL;
734}
735
736void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size,
737 unsigned long align, unsigned long goal)
738{
e9079911 739 return ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0);
421456ed
JW
740}
741
e9079911 742void * __init ___alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
421456ed
JW
743 unsigned long align, unsigned long goal,
744 unsigned long limit)
745{
746 void *ptr;
747
e9079911 748 ptr = ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0);
421456ed
JW
749 if (ptr)
750 return ptr;
751
1170532b 752 pr_alert("bootmem alloc of %lu bytes failed!\n", size);
ab381843
JW
753 panic("Out of memory");
754 return NULL;
4cc278b7
JW
755}
756
a66fd7da
JW
757/**
758 * __alloc_bootmem_node - allocate boot memory from a specific node
759 * @pgdat: node to allocate from
760 * @size: size of the request in bytes
761 * @align: alignment of the region
762 * @goal: preferred starting address of the region
763 *
764 * The goal is dropped if it can not be satisfied and the allocation will
765 * fall back to memory below @goal.
766 *
767 * Allocation may fall back to any node in the system if the specified node
768 * can not hold the requested memory.
769 *
770 * The function panics if the request can not be satisfied.
771 */
bb0923a6
FBH
772void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size,
773 unsigned long align, unsigned long goal)
1da177e4 774{
c91c4773
PE
775 if (WARN_ON_ONCE(slab_is_available()))
776 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
777
e9079911 778 return ___alloc_bootmem_node(pgdat, size, align, goal, 0);
08677214
YL
779}
780
781void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size,
782 unsigned long align, unsigned long goal)
783{
784#ifdef MAX_DMA32_PFN
785 unsigned long end_pfn;
786
787 if (WARN_ON_ONCE(slab_is_available()))
788 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
789
790 /* update goal according ...MAX_DMA32_PFN */
83285c72 791 end_pfn = pgdat_end_pfn(pgdat);
08677214
YL
792
793 if (end_pfn > MAX_DMA32_PFN + (128 >> (20 - PAGE_SHIFT)) &&
794 (goal >> PAGE_SHIFT) < MAX_DMA32_PFN) {
795 void *ptr;
796 unsigned long new_goal;
797
798 new_goal = MAX_DMA32_PFN << PAGE_SHIFT;
c6785b6b 799 ptr = alloc_bootmem_bdata(pgdat->bdata, size, align,
08677214 800 new_goal, 0);
08677214
YL
801 if (ptr)
802 return ptr;
803 }
804#endif
805
806 return __alloc_bootmem_node(pgdat, size, align, goal);
807
1da177e4
LT
808}
809
a66fd7da
JW
810/**
811 * __alloc_bootmem_low - allocate low boot memory
812 * @size: size of the request in bytes
813 * @align: alignment of the region
814 * @goal: preferred starting address of the region
815 *
816 * The goal is dropped if it can not be satisfied and the allocation will
817 * fall back to memory below @goal.
818 *
819 * Allocation may happen on any node in the system.
820 *
821 * The function panics if the request can not be satisfied.
822 */
bb0923a6
FBH
823void * __init __alloc_bootmem_low(unsigned long size, unsigned long align,
824 unsigned long goal)
008857c1 825{
0f3caba2 826 return ___alloc_bootmem(size, align, goal, ARCH_LOW_ADDRESS_LIMIT);
008857c1
RT
827}
828
38fa4175
YL
829void * __init __alloc_bootmem_low_nopanic(unsigned long size,
830 unsigned long align,
831 unsigned long goal)
832{
833 return ___alloc_bootmem_nopanic(size, align, goal,
834 ARCH_LOW_ADDRESS_LIMIT);
835}
836
a66fd7da
JW
837/**
838 * __alloc_bootmem_low_node - allocate low boot memory from a specific node
839 * @pgdat: node to allocate from
840 * @size: size of the request in bytes
841 * @align: alignment of the region
842 * @goal: preferred starting address of the region
843 *
844 * The goal is dropped if it can not be satisfied and the allocation will
845 * fall back to memory below @goal.
846 *
847 * Allocation may fall back to any node in the system if the specified node
848 * can not hold the requested memory.
849 *
850 * The function panics if the request can not be satisfied.
851 */
008857c1
RT
852void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size,
853 unsigned long align, unsigned long goal)
854{
c91c4773
PE
855 if (WARN_ON_ONCE(slab_is_available()))
856 return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id);
857
e9079911
JW
858 return ___alloc_bootmem_node(pgdat, size, align,
859 goal, ARCH_LOW_ADDRESS_LIMIT);
008857c1 860}