Commit | Line | Data |
---|---|---|
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 |
25 | struct pglist_data __refdata contig_page_data = { | |
26 | .bdata = &bootmem_node_data[0] | |
27 | }; | |
28 | EXPORT_SYMBOL(contig_page_data); | |
29 | #endif | |
30 | ||
1da177e4 LT |
31 | unsigned long max_low_pfn; |
32 | unsigned long min_low_pfn; | |
33 | unsigned long max_pfn; | |
8dd33030 | 34 | unsigned long long max_possible_pfn; |
1da177e4 | 35 | |
b61bfa3c JW |
36 | bootmem_data_t bootmem_node_data[MAX_NUMNODES] __initdata; |
37 | ||
636cc40c JW |
38 | static struct list_head bdata_list __initdata = LIST_HEAD_INIT(bdata_list); |
39 | ||
2e5237da JW |
40 | static int bootmem_debug; |
41 | ||
42 | static int __init bootmem_debug_setup(char *buf) | |
43 | { | |
44 | bootmem_debug = 1; | |
45 | return 0; | |
46 | } | |
47 | early_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 | 55 | static 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 | |
7c757207 MR |
65 | * |
66 | * Return: the number of pages needed to hold the bitmap. | |
a66fd7da | 67 | */ |
f71bf0ca | 68 | unsigned 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 | 78 | static 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 | 95 | static 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 | * | |
7c757207 | 126 | * Return: the number of bytes needed to hold the bitmap for this node. |
a66fd7da | 127 | */ |
223e8dc9 JW |
128 | unsigned 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 | * | |
7c757207 | 139 | * Return: the number of bytes needed to hold the bitmap. |
a66fd7da | 140 | */ |
223e8dc9 JW |
141 | unsigned 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 | |
81df9bff | 148 | void __init free_bootmem_late(unsigned long physaddr, unsigned long size) |
9f993ac3 FT |
149 | { |
150 | unsigned long cursor, end; | |
151 | ||
9099daed | 152 | kmemleak_free_part_phys(physaddr, size); |
9f993ac3 | 153 | |
81df9bff JK |
154 | cursor = PFN_UP(physaddr); |
155 | end = PFN_DOWN(physaddr + size); | |
9f993ac3 FT |
156 | |
157 | for (; cursor < end; cursor++) { | |
d70ddd7a | 158 | __free_pages_bootmem(pfn_to_page(cursor), cursor, 0); |
9f993ac3 FT |
159 | totalram_pages++; |
160 | } | |
161 | } | |
162 | ||
223e8dc9 JW |
163 | static unsigned long __init free_all_bootmem_core(bootmem_data_t *bdata) |
164 | { | |
165 | struct page *page; | |
d70ddd7a | 166 | unsigned long *map, start, end, pages, cur, count = 0; |
41546c17 JW |
167 | |
168 | if (!bdata->node_bootmem_map) | |
169 | return 0; | |
170 | ||
4a099fb4 | 171 | map = bdata->node_bootmem_map; |
3560e249 | 172 | start = bdata->node_min_pfn; |
41546c17 JW |
173 | end = bdata->node_low_pfn; |
174 | ||
799f933a JW |
175 | bdebug("nid=%td start=%lx end=%lx\n", |
176 | bdata - bootmem_node_data, start, end); | |
223e8dc9 | 177 | |
41546c17 | 178 | while (start < end) { |
4a099fb4 | 179 | unsigned long idx, vec; |
10d73e65 | 180 | unsigned shift; |
223e8dc9 | 181 | |
3560e249 | 182 | idx = start - bdata->node_min_pfn; |
10d73e65 MF |
183 | shift = idx & (BITS_PER_LONG - 1); |
184 | /* | |
185 | * vec holds at most BITS_PER_LONG map bits, | |
186 | * bit 0 corresponds to start. | |
187 | */ | |
41546c17 | 188 | vec = ~map[idx / BITS_PER_LONG]; |
10d73e65 MF |
189 | |
190 | if (shift) { | |
191 | vec >>= shift; | |
192 | if (end - start >= BITS_PER_LONG) | |
193 | vec |= ~map[idx / BITS_PER_LONG + 1] << | |
194 | (BITS_PER_LONG - shift); | |
195 | } | |
799f933a JW |
196 | /* |
197 | * If we have a properly aligned and fully unreserved | |
198 | * BITS_PER_LONG block of pages in front of us, free | |
199 | * it in one go. | |
200 | */ | |
201 | if (IS_ALIGNED(start, BITS_PER_LONG) && vec == ~0UL) { | |
41546c17 JW |
202 | int order = ilog2(BITS_PER_LONG); |
203 | ||
d70ddd7a | 204 | __free_pages_bootmem(pfn_to_page(start), start, order); |
223e8dc9 | 205 | count += BITS_PER_LONG; |
799f933a | 206 | start += BITS_PER_LONG; |
41546c17 | 207 | } else { |
d70ddd7a | 208 | cur = start; |
41546c17 | 209 | |
10d73e65 MF |
210 | start = ALIGN(start + 1, BITS_PER_LONG); |
211 | while (vec && cur != start) { | |
41546c17 | 212 | if (vec & 1) { |
10d73e65 | 213 | page = pfn_to_page(cur); |
d70ddd7a | 214 | __free_pages_bootmem(page, cur, 0); |
41546c17 | 215 | count++; |
223e8dc9 | 216 | } |
41546c17 | 217 | vec >>= 1; |
10d73e65 | 218 | ++cur; |
223e8dc9 | 219 | } |
223e8dc9 | 220 | } |
223e8dc9 JW |
221 | } |
222 | ||
d70ddd7a | 223 | cur = bdata->node_min_pfn; |
223e8dc9 | 224 | page = virt_to_page(bdata->node_bootmem_map); |
3560e249 | 225 | pages = bdata->node_low_pfn - bdata->node_min_pfn; |
41546c17 JW |
226 | pages = bootmem_bootmap_pages(pages); |
227 | count += pages; | |
5576646f | 228 | while (pages--) |
d70ddd7a | 229 | __free_pages_bootmem(page++, cur++, 0); |
1b4ace41 | 230 | bdata->node_bootmem_map = NULL; |
223e8dc9 | 231 | |
2e5237da JW |
232 | bdebug("nid=%td released=%lx\n", bdata - bootmem_node_data, count); |
233 | ||
223e8dc9 JW |
234 | return count; |
235 | } | |
236 | ||
7b4b2a0d JL |
237 | static int reset_managed_pages_done __initdata; |
238 | ||
f784a3f1 | 239 | void reset_node_managed_pages(pg_data_t *pgdat) |
9feedc9d JL |
240 | { |
241 | struct zone *z; | |
242 | ||
9feedc9d | 243 | for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++) |
7b4b2a0d JL |
244 | z->managed_pages = 0; |
245 | } | |
246 | ||
247 | void __init reset_all_zones_managed_pages(void) | |
248 | { | |
249 | struct pglist_data *pgdat; | |
250 | ||
f784a3f1 TC |
251 | if (reset_managed_pages_done) |
252 | return; | |
253 | ||
7b4b2a0d JL |
254 | for_each_online_pgdat(pgdat) |
255 | reset_node_managed_pages(pgdat); | |
f784a3f1 | 256 | |
7b4b2a0d | 257 | reset_managed_pages_done = 1; |
9feedc9d JL |
258 | } |
259 | ||
223e8dc9 JW |
260 | unsigned long __init free_all_bootmem(void) |
261 | { | |
aa235fc7 YL |
262 | unsigned long total_pages = 0; |
263 | bootmem_data_t *bdata; | |
9feedc9d | 264 | |
7b4b2a0d | 265 | reset_all_zones_managed_pages(); |
aa235fc7 YL |
266 | |
267 | list_for_each_entry(bdata, &bdata_list, list) | |
268 | total_pages += free_all_bootmem_core(bdata); | |
269 | ||
0c988534 JL |
270 | totalram_pages += total_pages; |
271 | ||
aa235fc7 | 272 | return total_pages; |
223e8dc9 JW |
273 | } |
274 | ||
d747fa4b JW |
275 | static void __init __free(bootmem_data_t *bdata, |
276 | unsigned long sidx, unsigned long eidx) | |
277 | { | |
278 | unsigned long idx; | |
279 | ||
280 | bdebug("nid=%td start=%lx end=%lx\n", bdata - bootmem_node_data, | |
3560e249 JW |
281 | sidx + bdata->node_min_pfn, |
282 | eidx + bdata->node_min_pfn); | |
d747fa4b | 283 | |
1b4ace41 CM |
284 | if (WARN_ON(bdata->node_bootmem_map == NULL)) |
285 | return; | |
286 | ||
e2bf3cae JW |
287 | if (bdata->hint_idx > sidx) |
288 | bdata->hint_idx = sidx; | |
289 | ||
d747fa4b JW |
290 | for (idx = sidx; idx < eidx; idx++) |
291 | if (!test_and_clear_bit(idx, bdata->node_bootmem_map)) | |
292 | BUG(); | |
293 | } | |
294 | ||
295 | static int __init __reserve(bootmem_data_t *bdata, unsigned long sidx, | |
296 | unsigned long eidx, int flags) | |
297 | { | |
298 | unsigned long idx; | |
299 | int exclusive = flags & BOOTMEM_EXCLUSIVE; | |
300 | ||
301 | bdebug("nid=%td start=%lx end=%lx flags=%x\n", | |
302 | bdata - bootmem_node_data, | |
3560e249 JW |
303 | sidx + bdata->node_min_pfn, |
304 | eidx + bdata->node_min_pfn, | |
d747fa4b JW |
305 | flags); |
306 | ||
1b4ace41 CM |
307 | if (WARN_ON(bdata->node_bootmem_map == NULL)) |
308 | return 0; | |
309 | ||
d747fa4b JW |
310 | for (idx = sidx; idx < eidx; idx++) |
311 | if (test_and_set_bit(idx, bdata->node_bootmem_map)) { | |
312 | if (exclusive) { | |
313 | __free(bdata, sidx, idx); | |
314 | return -EBUSY; | |
315 | } | |
316 | bdebug("silent double reserve of PFN %lx\n", | |
3560e249 | 317 | idx + bdata->node_min_pfn); |
d747fa4b JW |
318 | } |
319 | return 0; | |
320 | } | |
321 | ||
e2bf3cae JW |
322 | static int __init mark_bootmem_node(bootmem_data_t *bdata, |
323 | unsigned long start, unsigned long end, | |
324 | int reserve, int flags) | |
223e8dc9 JW |
325 | { |
326 | unsigned long sidx, eidx; | |
223e8dc9 | 327 | |
e2bf3cae JW |
328 | bdebug("nid=%td start=%lx end=%lx reserve=%d flags=%x\n", |
329 | bdata - bootmem_node_data, start, end, reserve, flags); | |
223e8dc9 | 330 | |
3560e249 | 331 | BUG_ON(start < bdata->node_min_pfn); |
e2bf3cae | 332 | BUG_ON(end > bdata->node_low_pfn); |
223e8dc9 | 333 | |
3560e249 JW |
334 | sidx = start - bdata->node_min_pfn; |
335 | eidx = end - bdata->node_min_pfn; | |
223e8dc9 | 336 | |
e2bf3cae JW |
337 | if (reserve) |
338 | return __reserve(bdata, sidx, eidx, flags); | |
223e8dc9 | 339 | else |
e2bf3cae JW |
340 | __free(bdata, sidx, eidx); |
341 | return 0; | |
342 | } | |
343 | ||
344 | static int __init mark_bootmem(unsigned long start, unsigned long end, | |
345 | int reserve, int flags) | |
346 | { | |
347 | unsigned long pos; | |
348 | bootmem_data_t *bdata; | |
349 | ||
350 | pos = start; | |
351 | list_for_each_entry(bdata, &bdata_list, list) { | |
352 | int err; | |
353 | unsigned long max; | |
354 | ||
3560e249 JW |
355 | if (pos < bdata->node_min_pfn || |
356 | pos >= bdata->node_low_pfn) { | |
e2bf3cae JW |
357 | BUG_ON(pos != start); |
358 | continue; | |
359 | } | |
360 | ||
361 | max = min(bdata->node_low_pfn, end); | |
223e8dc9 | 362 | |
e2bf3cae JW |
363 | err = mark_bootmem_node(bdata, pos, max, reserve, flags); |
364 | if (reserve && err) { | |
365 | mark_bootmem(start, pos, 0, 0); | |
366 | return err; | |
367 | } | |
223e8dc9 | 368 | |
e2bf3cae JW |
369 | if (max == end) |
370 | return 0; | |
371 | pos = bdata->node_low_pfn; | |
372 | } | |
373 | BUG(); | |
223e8dc9 JW |
374 | } |
375 | ||
376 | void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, | |
377 | unsigned long size) | |
378 | { | |
e2bf3cae JW |
379 | unsigned long start, end; |
380 | ||
9099daed | 381 | kmemleak_free_part_phys(physaddr, size); |
ec3a354b | 382 | |
e2bf3cae JW |
383 | start = PFN_UP(physaddr); |
384 | end = PFN_DOWN(physaddr + size); | |
385 | ||
386 | mark_bootmem_node(pgdat->bdata, start, end, 0, 0); | |
223e8dc9 JW |
387 | } |
388 | ||
81df9bff | 389 | void __init free_bootmem(unsigned long physaddr, unsigned long size) |
223e8dc9 | 390 | { |
e2bf3cae | 391 | unsigned long start, end; |
a5645a61 | 392 | |
9099daed | 393 | kmemleak_free_part_phys(physaddr, size); |
ec3a354b | 394 | |
81df9bff JK |
395 | start = PFN_UP(physaddr); |
396 | end = PFN_DOWN(physaddr + size); | |
1da177e4 | 397 | |
e2bf3cae | 398 | mark_bootmem(start, end, 0, 0); |
1da177e4 LT |
399 | } |
400 | ||
a66fd7da JW |
401 | /** |
402 | * reserve_bootmem_node - mark a page range as reserved | |
403 | * @pgdat: node the range resides on | |
404 | * @physaddr: starting address of the range | |
405 | * @size: size of the range in bytes | |
406 | * @flags: reservation flags (see linux/bootmem.h) | |
407 | * | |
408 | * Partial pages will be reserved. | |
409 | * | |
e2bf3cae | 410 | * The range must reside completely on the specified node. |
7c757207 MR |
411 | * |
412 | * Return: 0 on success, -errno on failure. | |
a66fd7da | 413 | */ |
223e8dc9 JW |
414 | int __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
415 | unsigned long size, int flags) | |
1da177e4 | 416 | { |
e2bf3cae | 417 | unsigned long start, end; |
1da177e4 | 418 | |
e2bf3cae JW |
419 | start = PFN_DOWN(physaddr); |
420 | end = PFN_UP(physaddr + size); | |
421 | ||
422 | return mark_bootmem_node(pgdat->bdata, start, end, 1, flags); | |
223e8dc9 | 423 | } |
5a982cbc | 424 | |
a66fd7da | 425 | /** |
0d4ba4d7 | 426 | * reserve_bootmem - mark a page range as reserved |
a66fd7da JW |
427 | * @addr: starting address of the range |
428 | * @size: size of the range in bytes | |
429 | * @flags: reservation flags (see linux/bootmem.h) | |
430 | * | |
431 | * Partial pages will be reserved. | |
432 | * | |
e2bf3cae | 433 | * The range must be contiguous but may span node boundaries. |
7c757207 MR |
434 | * |
435 | * Return: 0 on success, -errno on failure. | |
a66fd7da | 436 | */ |
223e8dc9 JW |
437 | int __init reserve_bootmem(unsigned long addr, unsigned long size, |
438 | int flags) | |
439 | { | |
e2bf3cae | 440 | unsigned long start, end; |
1da177e4 | 441 | |
e2bf3cae JW |
442 | start = PFN_DOWN(addr); |
443 | end = PFN_UP(addr + size); | |
223e8dc9 | 444 | |
e2bf3cae | 445 | return mark_bootmem(start, end, 1, flags); |
1da177e4 LT |
446 | } |
447 | ||
8aa043d7 JB |
448 | static unsigned long __init align_idx(struct bootmem_data *bdata, |
449 | unsigned long idx, unsigned long step) | |
481ebd0d JW |
450 | { |
451 | unsigned long base = bdata->node_min_pfn; | |
452 | ||
453 | /* | |
454 | * Align the index with respect to the node start so that the | |
455 | * combination of both satisfies the requested alignment. | |
456 | */ | |
457 | ||
458 | return ALIGN(base + idx, step) - base; | |
459 | } | |
460 | ||
8aa043d7 JB |
461 | static unsigned long __init align_off(struct bootmem_data *bdata, |
462 | unsigned long off, unsigned long align) | |
481ebd0d JW |
463 | { |
464 | unsigned long base = PFN_PHYS(bdata->node_min_pfn); | |
465 | ||
466 | /* Same as align_idx for byte offsets */ | |
467 | ||
468 | return ALIGN(base + off, align) - base; | |
469 | } | |
470 | ||
c6785b6b | 471 | static void * __init alloc_bootmem_bdata(struct bootmem_data *bdata, |
d0c4f570 TH |
472 | unsigned long size, unsigned long align, |
473 | unsigned long goal, unsigned long limit) | |
1da177e4 | 474 | { |
0f3caba2 | 475 | unsigned long fallback = 0; |
5f2809e6 JW |
476 | unsigned long min, max, start, sidx, midx, step; |
477 | ||
594fe1a0 JW |
478 | bdebug("nid=%td size=%lx [%lu pages] align=%lx goal=%lx limit=%lx\n", |
479 | bdata - bootmem_node_data, size, PAGE_ALIGN(size) >> PAGE_SHIFT, | |
480 | align, goal, limit); | |
481 | ||
5f2809e6 JW |
482 | BUG_ON(!size); |
483 | BUG_ON(align & (align - 1)); | |
484 | BUG_ON(limit && goal + size > limit); | |
1da177e4 | 485 | |
7c309a64 CK |
486 | if (!bdata->node_bootmem_map) |
487 | return NULL; | |
488 | ||
3560e249 | 489 | min = bdata->node_min_pfn; |
5f2809e6 | 490 | max = bdata->node_low_pfn; |
9a2dc04c | 491 | |
5f2809e6 JW |
492 | goal >>= PAGE_SHIFT; |
493 | limit >>= PAGE_SHIFT; | |
494 | ||
495 | if (limit && max > limit) | |
496 | max = limit; | |
497 | if (max <= min) | |
9a2dc04c YL |
498 | return NULL; |
499 | ||
5f2809e6 | 500 | step = max(align >> PAGE_SHIFT, 1UL); |
281dd25c | 501 | |
5f2809e6 JW |
502 | if (goal && min < goal && goal < max) |
503 | start = ALIGN(goal, step); | |
504 | else | |
505 | start = ALIGN(min, step); | |
1da177e4 | 506 | |
481ebd0d | 507 | sidx = start - bdata->node_min_pfn; |
3560e249 | 508 | midx = max - bdata->node_min_pfn; |
1da177e4 | 509 | |
5f2809e6 | 510 | if (bdata->hint_idx > sidx) { |
0f3caba2 JW |
511 | /* |
512 | * Handle the valid case of sidx being zero and still | |
513 | * catch the fallback below. | |
514 | */ | |
515 | fallback = sidx + 1; | |
481ebd0d | 516 | sidx = align_idx(bdata, bdata->hint_idx, step); |
5f2809e6 | 517 | } |
1da177e4 | 518 | |
5f2809e6 JW |
519 | while (1) { |
520 | int merge; | |
521 | void *region; | |
522 | unsigned long eidx, i, start_off, end_off; | |
523 | find_block: | |
524 | sidx = find_next_zero_bit(bdata->node_bootmem_map, midx, sidx); | |
481ebd0d | 525 | sidx = align_idx(bdata, sidx, step); |
5f2809e6 | 526 | eidx = sidx + PFN_UP(size); |
ad09315c | 527 | |
5f2809e6 | 528 | if (sidx >= midx || eidx > midx) |
66d43e98 | 529 | break; |
1da177e4 | 530 | |
5f2809e6 JW |
531 | for (i = sidx; i < eidx; i++) |
532 | if (test_bit(i, bdata->node_bootmem_map)) { | |
481ebd0d | 533 | sidx = align_idx(bdata, i, step); |
5f2809e6 JW |
534 | if (sidx == i) |
535 | sidx += step; | |
536 | goto find_block; | |
537 | } | |
1da177e4 | 538 | |
627240aa | 539 | if (bdata->last_end_off & (PAGE_SIZE - 1) && |
5f2809e6 | 540 | PFN_DOWN(bdata->last_end_off) + 1 == sidx) |
481ebd0d | 541 | start_off = align_off(bdata, bdata->last_end_off, align); |
5f2809e6 JW |
542 | else |
543 | start_off = PFN_PHYS(sidx); | |
544 | ||
545 | merge = PFN_DOWN(start_off) < sidx; | |
546 | end_off = start_off + size; | |
547 | ||
548 | bdata->last_end_off = end_off; | |
549 | bdata->hint_idx = PFN_UP(end_off); | |
550 | ||
551 | /* | |
552 | * Reserve the area now: | |
553 | */ | |
d747fa4b JW |
554 | if (__reserve(bdata, PFN_DOWN(start_off) + merge, |
555 | PFN_UP(end_off), BOOTMEM_EXCLUSIVE)) | |
556 | BUG(); | |
5f2809e6 | 557 | |
3560e249 JW |
558 | region = phys_to_virt(PFN_PHYS(bdata->node_min_pfn) + |
559 | start_off); | |
5f2809e6 | 560 | memset(region, 0, size); |
008139d9 CM |
561 | /* |
562 | * The min_count is set to 0 so that bootmem allocated blocks | |
563 | * are never reported as leaks. | |
564 | */ | |
565 | kmemleak_alloc(region, size, 0, 0); | |
5f2809e6 | 566 | return region; |
1da177e4 LT |
567 | } |
568 | ||
0f3caba2 | 569 | if (fallback) { |
481ebd0d | 570 | sidx = align_idx(bdata, fallback - 1, step); |
0f3caba2 JW |
571 | fallback = 0; |
572 | goto find_block; | |
573 | } | |
574 | ||
575 | return NULL; | |
576 | } | |
577 | ||
c12ab504 | 578 | static void * __init alloc_bootmem_core(unsigned long size, |
0f3caba2 JW |
579 | unsigned long align, |
580 | unsigned long goal, | |
581 | unsigned long limit) | |
582 | { | |
583 | bootmem_data_t *bdata; | |
d0c4f570 | 584 | void *region; |
0f3caba2 | 585 | |
3f7dfe24 JK |
586 | if (WARN_ON_ONCE(slab_is_available())) |
587 | return kzalloc(size, GFP_NOWAIT); | |
0f3caba2 | 588 | |
d0c4f570 | 589 | list_for_each_entry(bdata, &bdata_list, list) { |
0f3caba2 JW |
590 | if (goal && bdata->node_low_pfn <= PFN_DOWN(goal)) |
591 | continue; | |
3560e249 | 592 | if (limit && bdata->node_min_pfn >= PFN_DOWN(limit)) |
0f3caba2 JW |
593 | break; |
594 | ||
c6785b6b | 595 | region = alloc_bootmem_bdata(bdata, size, align, goal, limit); |
0f3caba2 JW |
596 | if (region) |
597 | return region; | |
598 | } | |
599 | ||
c12ab504 JW |
600 | return NULL; |
601 | } | |
602 | ||
603 | static void * __init ___alloc_bootmem_nopanic(unsigned long size, | |
604 | unsigned long align, | |
605 | unsigned long goal, | |
606 | unsigned long limit) | |
607 | { | |
608 | void *ptr; | |
609 | ||
610 | restart: | |
611 | ptr = alloc_bootmem_core(size, align, goal, limit); | |
612 | if (ptr) | |
613 | return ptr; | |
5f2809e6 JW |
614 | if (goal) { |
615 | goal = 0; | |
0f3caba2 | 616 | goto restart; |
5f2809e6 | 617 | } |
2e5237da | 618 | |
5f2809e6 | 619 | return NULL; |
1da177e4 LT |
620 | } |
621 | ||
bb0923a6 | 622 | void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align, |
0f3caba2 | 623 | unsigned long goal) |
1da177e4 | 624 | { |
08677214 YL |
625 | unsigned long limit = 0; |
626 | ||
08677214 | 627 | return ___alloc_bootmem_nopanic(size, align, goal, limit); |
0f3caba2 | 628 | } |
1da177e4 | 629 | |
0f3caba2 JW |
630 | static void * __init ___alloc_bootmem(unsigned long size, unsigned long align, |
631 | unsigned long goal, unsigned long limit) | |
632 | { | |
633 | void *mem = ___alloc_bootmem_nopanic(size, align, goal, limit); | |
634 | ||
635 | if (mem) | |
636 | return mem; | |
637 | /* | |
638 | * Whoops, we cannot satisfy the allocation request. | |
639 | */ | |
1170532b | 640 | pr_alert("bootmem alloc of %lu bytes failed!\n", size); |
0f3caba2 | 641 | panic("Out of memory"); |
a8062231 AK |
642 | return NULL; |
643 | } | |
1da177e4 | 644 | |
bb0923a6 FBH |
645 | void * __init __alloc_bootmem(unsigned long size, unsigned long align, |
646 | unsigned long goal) | |
a8062231 | 647 | { |
08677214 YL |
648 | unsigned long limit = 0; |
649 | ||
08677214 | 650 | return ___alloc_bootmem(size, align, goal, limit); |
1da177e4 LT |
651 | } |
652 | ||
99ab7b19 | 653 | void * __init ___alloc_bootmem_node_nopanic(pg_data_t *pgdat, |
4cc278b7 JW |
654 | unsigned long size, unsigned long align, |
655 | unsigned long goal, unsigned long limit) | |
656 | { | |
657 | void *ptr; | |
658 | ||
3f7dfe24 | 659 | if (WARN_ON_ONCE(slab_is_available())) |
1d8bf926 | 660 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); |
ab381843 | 661 | again: |
d0c4f570 | 662 | |
c8f4a2d0 YL |
663 | /* do not panic in alloc_bootmem_bdata() */ |
664 | if (limit && goal + size > limit) | |
665 | limit = 0; | |
666 | ||
e9079911 | 667 | ptr = alloc_bootmem_bdata(pgdat->bdata, size, align, goal, limit); |
4cc278b7 JW |
668 | if (ptr) |
669 | return ptr; | |
670 | ||
ab381843 JW |
671 | ptr = alloc_bootmem_core(size, align, goal, limit); |
672 | if (ptr) | |
673 | return ptr; | |
674 | ||
675 | if (goal) { | |
676 | goal = 0; | |
677 | goto again; | |
678 | } | |
679 | ||
421456ed JW |
680 | return NULL; |
681 | } | |
682 | ||
683 | void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size, | |
684 | unsigned long align, unsigned long goal) | |
685 | { | |
e9079911 | 686 | return ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0); |
421456ed JW |
687 | } |
688 | ||
e9079911 | 689 | void * __init ___alloc_bootmem_node(pg_data_t *pgdat, unsigned long size, |
421456ed JW |
690 | unsigned long align, unsigned long goal, |
691 | unsigned long limit) | |
692 | { | |
693 | void *ptr; | |
694 | ||
e9079911 | 695 | ptr = ___alloc_bootmem_node_nopanic(pgdat, size, align, goal, 0); |
421456ed JW |
696 | if (ptr) |
697 | return ptr; | |
698 | ||
1170532b | 699 | pr_alert("bootmem alloc of %lu bytes failed!\n", size); |
ab381843 JW |
700 | panic("Out of memory"); |
701 | return NULL; | |
4cc278b7 JW |
702 | } |
703 | ||
bb0923a6 FBH |
704 | void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size, |
705 | unsigned long align, unsigned long goal) | |
1da177e4 | 706 | { |
c91c4773 PE |
707 | if (WARN_ON_ONCE(slab_is_available())) |
708 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
709 | ||
e9079911 | 710 | return ___alloc_bootmem_node(pgdat, size, align, goal, 0); |
08677214 YL |
711 | } |
712 | ||
713 | void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size, | |
714 | unsigned long align, unsigned long goal) | |
715 | { | |
716 | #ifdef MAX_DMA32_PFN | |
717 | unsigned long end_pfn; | |
718 | ||
719 | if (WARN_ON_ONCE(slab_is_available())) | |
720 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
721 | ||
722 | /* update goal according ...MAX_DMA32_PFN */ | |
83285c72 | 723 | end_pfn = pgdat_end_pfn(pgdat); |
08677214 YL |
724 | |
725 | if (end_pfn > MAX_DMA32_PFN + (128 >> (20 - PAGE_SHIFT)) && | |
726 | (goal >> PAGE_SHIFT) < MAX_DMA32_PFN) { | |
727 | void *ptr; | |
728 | unsigned long new_goal; | |
729 | ||
730 | new_goal = MAX_DMA32_PFN << PAGE_SHIFT; | |
c6785b6b | 731 | ptr = alloc_bootmem_bdata(pgdat->bdata, size, align, |
08677214 | 732 | new_goal, 0); |
08677214 YL |
733 | if (ptr) |
734 | return ptr; | |
735 | } | |
736 | #endif | |
737 | ||
738 | return __alloc_bootmem_node(pgdat, size, align, goal); | |
739 | ||
1da177e4 LT |
740 | } |
741 | ||
bb0923a6 FBH |
742 | void * __init __alloc_bootmem_low(unsigned long size, unsigned long align, |
743 | unsigned long goal) | |
008857c1 | 744 | { |
0f3caba2 | 745 | return ___alloc_bootmem(size, align, goal, ARCH_LOW_ADDRESS_LIMIT); |
008857c1 RT |
746 | } |
747 | ||
38fa4175 YL |
748 | void * __init __alloc_bootmem_low_nopanic(unsigned long size, |
749 | unsigned long align, | |
750 | unsigned long goal) | |
751 | { | |
752 | return ___alloc_bootmem_nopanic(size, align, goal, | |
753 | ARCH_LOW_ADDRESS_LIMIT); | |
754 | } | |
755 | ||
008857c1 RT |
756 | void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size, |
757 | unsigned long align, unsigned long goal) | |
758 | { | |
c91c4773 PE |
759 | if (WARN_ON_ONCE(slab_is_available())) |
760 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
761 | ||
e9079911 JW |
762 | return ___alloc_bootmem_node(pgdat, size, align, |
763 | goal, ARCH_LOW_ADDRESS_LIMIT); | |
008857c1 | 764 | } |