Commit | Line | Data |
---|---|---|
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> |
e786e86a FBH |
19 | |
20 | #include <asm/bug.h> | |
1da177e4 | 21 | #include <asm/io.h> |
dfd54cbc | 22 | #include <asm/processor.h> |
e786e86a | 23 | |
1da177e4 LT |
24 | #include "internal.h" |
25 | ||
e782ab42 YL |
26 | #ifndef CONFIG_NEED_MULTIPLE_NODES |
27 | struct pglist_data __refdata contig_page_data = { | |
28 | .bdata = &bootmem_node_data[0] | |
29 | }; | |
30 | EXPORT_SYMBOL(contig_page_data); | |
31 | #endif | |
32 | ||
1da177e4 LT |
33 | unsigned long max_low_pfn; |
34 | unsigned long min_low_pfn; | |
35 | unsigned long max_pfn; | |
36 | ||
b61bfa3c JW |
37 | bootmem_data_t bootmem_node_data[MAX_NUMNODES] __initdata; |
38 | ||
636cc40c JW |
39 | static struct list_head bdata_list __initdata = LIST_HEAD_INIT(bdata_list); |
40 | ||
2e5237da JW |
41 | static int bootmem_debug; |
42 | ||
43 | static int __init bootmem_debug_setup(char *buf) | |
44 | { | |
45 | bootmem_debug = 1; | |
46 | return 0; | |
47 | } | |
48 | early_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 | 57 | static 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 | 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 | { |
636cc40c | 80 | struct list_head *iter; |
f71bf0ca | 81 | |
636cc40c JW |
82 | list_for_each(iter, &bdata_list) { |
83 | bootmem_data_t *ent; | |
84 | ||
85 | ent = list_entry(iter, bootmem_data_t, list); | |
3560e249 | 86 | if (bdata->node_min_pfn < ent->node_min_pfn) |
636cc40c | 87 | break; |
679bc9fb | 88 | } |
636cc40c | 89 | list_add_tail(&bdata->list, iter); |
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 | * | |
126 | * Returns the number of bytes needed to hold the bitmap for this node. | |
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 | * | |
139 | * Returns the number of bytes needed to hold the bitmap. | |
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 | |
9f993ac3 FT |
148 | /* |
149 | * free_bootmem_late - free bootmem pages directly to page allocator | |
150 | * @addr: starting address of the range | |
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 | */ | |
157 | void __init free_bootmem_late(unsigned long addr, unsigned long size) | |
158 | { | |
159 | unsigned long cursor, end; | |
160 | ||
161 | kmemleak_free_part(__va(addr), size); | |
162 | ||
163 | cursor = PFN_UP(addr); | |
164 | end = PFN_DOWN(addr + size); | |
165 | ||
166 | for (; cursor < end; cursor++) { | |
167 | __free_pages_bootmem(pfn_to_page(cursor), 0); | |
168 | totalram_pages++; | |
169 | } | |
170 | } | |
171 | ||
223e8dc9 JW |
172 | static unsigned long __init free_all_bootmem_core(bootmem_data_t *bdata) |
173 | { | |
174 | struct page *page; | |
41546c17 JW |
175 | unsigned long start, end, pages, count = 0; |
176 | ||
177 | if (!bdata->node_bootmem_map) | |
178 | return 0; | |
179 | ||
3560e249 | 180 | start = bdata->node_min_pfn; |
41546c17 JW |
181 | end = bdata->node_low_pfn; |
182 | ||
799f933a JW |
183 | bdebug("nid=%td start=%lx end=%lx\n", |
184 | bdata - bootmem_node_data, start, end); | |
223e8dc9 | 185 | |
41546c17 JW |
186 | while (start < end) { |
187 | unsigned long *map, idx, vec; | |
223e8dc9 | 188 | |
41546c17 | 189 | map = bdata->node_bootmem_map; |
3560e249 | 190 | idx = start - bdata->node_min_pfn; |
41546c17 | 191 | vec = ~map[idx / BITS_PER_LONG]; |
799f933a JW |
192 | /* |
193 | * If we have a properly aligned and fully unreserved | |
194 | * BITS_PER_LONG block of pages in front of us, free | |
195 | * it in one go. | |
196 | */ | |
197 | if (IS_ALIGNED(start, BITS_PER_LONG) && vec == ~0UL) { | |
41546c17 JW |
198 | int order = ilog2(BITS_PER_LONG); |
199 | ||
200 | __free_pages_bootmem(pfn_to_page(start), order); | |
223e8dc9 | 201 | count += BITS_PER_LONG; |
799f933a | 202 | start += BITS_PER_LONG; |
41546c17 JW |
203 | } else { |
204 | unsigned long off = 0; | |
205 | ||
6dccdcbe | 206 | vec >>= start & (BITS_PER_LONG - 1); |
549381e1 | 207 | while (vec) { |
41546c17 JW |
208 | if (vec & 1) { |
209 | page = pfn_to_page(start + off); | |
223e8dc9 | 210 | __free_pages_bootmem(page, 0); |
41546c17 | 211 | count++; |
223e8dc9 | 212 | } |
41546c17 JW |
213 | vec >>= 1; |
214 | off++; | |
223e8dc9 | 215 | } |
799f933a | 216 | start = ALIGN(start + 1, BITS_PER_LONG); |
223e8dc9 | 217 | } |
223e8dc9 JW |
218 | } |
219 | ||
223e8dc9 | 220 | page = virt_to_page(bdata->node_bootmem_map); |
3560e249 | 221 | pages = bdata->node_low_pfn - bdata->node_min_pfn; |
41546c17 JW |
222 | pages = bootmem_bootmap_pages(pages); |
223 | count += pages; | |
224 | while (pages--) | |
225 | __free_pages_bootmem(page++, 0); | |
223e8dc9 | 226 | |
2e5237da JW |
227 | bdebug("nid=%td released=%lx\n", bdata - bootmem_node_data, count); |
228 | ||
223e8dc9 JW |
229 | return count; |
230 | } | |
231 | ||
a66fd7da JW |
232 | /** |
233 | * free_all_bootmem_node - release a node's free pages to the buddy allocator | |
234 | * @pgdat: node to be released | |
235 | * | |
236 | * Returns the number of pages actually released. | |
237 | */ | |
223e8dc9 JW |
238 | unsigned long __init free_all_bootmem_node(pg_data_t *pgdat) |
239 | { | |
240 | register_page_bootmem_info_node(pgdat); | |
241 | return free_all_bootmem_core(pgdat->bdata); | |
242 | } | |
243 | ||
a66fd7da JW |
244 | /** |
245 | * free_all_bootmem - release free pages to the buddy allocator | |
246 | * | |
247 | * Returns the number of pages actually released. | |
248 | */ | |
223e8dc9 JW |
249 | unsigned long __init free_all_bootmem(void) |
250 | { | |
aa235fc7 YL |
251 | unsigned long total_pages = 0; |
252 | bootmem_data_t *bdata; | |
253 | ||
254 | list_for_each_entry(bdata, &bdata_list, list) | |
255 | total_pages += free_all_bootmem_core(bdata); | |
256 | ||
257 | return total_pages; | |
223e8dc9 JW |
258 | } |
259 | ||
d747fa4b JW |
260 | static void __init __free(bootmem_data_t *bdata, |
261 | unsigned long sidx, unsigned long eidx) | |
262 | { | |
263 | unsigned long idx; | |
264 | ||
265 | bdebug("nid=%td start=%lx end=%lx\n", bdata - bootmem_node_data, | |
3560e249 JW |
266 | sidx + bdata->node_min_pfn, |
267 | eidx + bdata->node_min_pfn); | |
d747fa4b | 268 | |
e2bf3cae JW |
269 | if (bdata->hint_idx > sidx) |
270 | bdata->hint_idx = sidx; | |
271 | ||
d747fa4b JW |
272 | for (idx = sidx; idx < eidx; idx++) |
273 | if (!test_and_clear_bit(idx, bdata->node_bootmem_map)) | |
274 | BUG(); | |
275 | } | |
276 | ||
277 | static int __init __reserve(bootmem_data_t *bdata, unsigned long sidx, | |
278 | unsigned long eidx, int flags) | |
279 | { | |
280 | unsigned long idx; | |
281 | int exclusive = flags & BOOTMEM_EXCLUSIVE; | |
282 | ||
283 | bdebug("nid=%td start=%lx end=%lx flags=%x\n", | |
284 | bdata - bootmem_node_data, | |
3560e249 JW |
285 | sidx + bdata->node_min_pfn, |
286 | eidx + bdata->node_min_pfn, | |
d747fa4b JW |
287 | flags); |
288 | ||
289 | for (idx = sidx; idx < eidx; idx++) | |
290 | if (test_and_set_bit(idx, bdata->node_bootmem_map)) { | |
291 | if (exclusive) { | |
292 | __free(bdata, sidx, idx); | |
293 | return -EBUSY; | |
294 | } | |
295 | bdebug("silent double reserve of PFN %lx\n", | |
3560e249 | 296 | idx + bdata->node_min_pfn); |
d747fa4b JW |
297 | } |
298 | return 0; | |
299 | } | |
300 | ||
e2bf3cae JW |
301 | static int __init mark_bootmem_node(bootmem_data_t *bdata, |
302 | unsigned long start, unsigned long end, | |
303 | int reserve, int flags) | |
223e8dc9 JW |
304 | { |
305 | unsigned long sidx, eidx; | |
223e8dc9 | 306 | |
e2bf3cae JW |
307 | bdebug("nid=%td start=%lx end=%lx reserve=%d flags=%x\n", |
308 | bdata - bootmem_node_data, start, end, reserve, flags); | |
223e8dc9 | 309 | |
3560e249 | 310 | BUG_ON(start < bdata->node_min_pfn); |
e2bf3cae | 311 | BUG_ON(end > bdata->node_low_pfn); |
223e8dc9 | 312 | |
3560e249 JW |
313 | sidx = start - bdata->node_min_pfn; |
314 | eidx = end - bdata->node_min_pfn; | |
223e8dc9 | 315 | |
e2bf3cae JW |
316 | if (reserve) |
317 | return __reserve(bdata, sidx, eidx, flags); | |
223e8dc9 | 318 | else |
e2bf3cae JW |
319 | __free(bdata, sidx, eidx); |
320 | return 0; | |
321 | } | |
322 | ||
323 | static int __init mark_bootmem(unsigned long start, unsigned long end, | |
324 | int reserve, int flags) | |
325 | { | |
326 | unsigned long pos; | |
327 | bootmem_data_t *bdata; | |
328 | ||
329 | pos = start; | |
330 | list_for_each_entry(bdata, &bdata_list, list) { | |
331 | int err; | |
332 | unsigned long max; | |
333 | ||
3560e249 JW |
334 | if (pos < bdata->node_min_pfn || |
335 | pos >= bdata->node_low_pfn) { | |
e2bf3cae JW |
336 | BUG_ON(pos != start); |
337 | continue; | |
338 | } | |
339 | ||
340 | max = min(bdata->node_low_pfn, end); | |
223e8dc9 | 341 | |
e2bf3cae JW |
342 | err = mark_bootmem_node(bdata, pos, max, reserve, flags); |
343 | if (reserve && err) { | |
344 | mark_bootmem(start, pos, 0, 0); | |
345 | return err; | |
346 | } | |
223e8dc9 | 347 | |
e2bf3cae JW |
348 | if (max == end) |
349 | return 0; | |
350 | pos = bdata->node_low_pfn; | |
351 | } | |
352 | BUG(); | |
223e8dc9 JW |
353 | } |
354 | ||
a66fd7da JW |
355 | /** |
356 | * free_bootmem_node - mark a page range as usable | |
357 | * @pgdat: node the range resides on | |
358 | * @physaddr: starting address of the range | |
359 | * @size: size of the range in bytes | |
360 | * | |
361 | * Partial pages will be considered reserved and left as they are. | |
362 | * | |
e2bf3cae | 363 | * The range must reside completely on the specified node. |
a66fd7da | 364 | */ |
223e8dc9 JW |
365 | void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
366 | unsigned long size) | |
367 | { | |
e2bf3cae JW |
368 | unsigned long start, end; |
369 | ||
ec3a354b CM |
370 | kmemleak_free_part(__va(physaddr), size); |
371 | ||
e2bf3cae JW |
372 | start = PFN_UP(physaddr); |
373 | end = PFN_DOWN(physaddr + size); | |
374 | ||
375 | mark_bootmem_node(pgdat->bdata, start, end, 0, 0); | |
223e8dc9 JW |
376 | } |
377 | ||
a66fd7da JW |
378 | /** |
379 | * free_bootmem - mark a page range as usable | |
380 | * @addr: starting address of the range | |
381 | * @size: size of the range in bytes | |
382 | * | |
383 | * Partial pages will be considered reserved and left as they are. | |
384 | * | |
e2bf3cae | 385 | * The range must be contiguous but may span node boundaries. |
a66fd7da | 386 | */ |
223e8dc9 JW |
387 | void __init free_bootmem(unsigned long addr, unsigned long size) |
388 | { | |
e2bf3cae | 389 | unsigned long start, end; |
a5645a61 | 390 | |
ec3a354b CM |
391 | kmemleak_free_part(__va(addr), size); |
392 | ||
e2bf3cae JW |
393 | start = PFN_UP(addr); |
394 | end = PFN_DOWN(addr + size); | |
1da177e4 | 395 | |
e2bf3cae | 396 | mark_bootmem(start, end, 0, 0); |
1da177e4 LT |
397 | } |
398 | ||
a66fd7da JW |
399 | /** |
400 | * reserve_bootmem_node - mark a page range as reserved | |
401 | * @pgdat: node the range resides on | |
402 | * @physaddr: starting address of the range | |
403 | * @size: size of the range in bytes | |
404 | * @flags: reservation flags (see linux/bootmem.h) | |
405 | * | |
406 | * Partial pages will be reserved. | |
407 | * | |
e2bf3cae | 408 | * The range must reside completely on the specified node. |
a66fd7da | 409 | */ |
223e8dc9 JW |
410 | int __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
411 | unsigned long size, int flags) | |
1da177e4 | 412 | { |
e2bf3cae | 413 | unsigned long start, end; |
1da177e4 | 414 | |
e2bf3cae JW |
415 | start = PFN_DOWN(physaddr); |
416 | end = PFN_UP(physaddr + size); | |
417 | ||
418 | return mark_bootmem_node(pgdat->bdata, start, end, 1, flags); | |
223e8dc9 | 419 | } |
5a982cbc | 420 | |
a66fd7da JW |
421 | /** |
422 | * reserve_bootmem - mark a page range as usable | |
423 | * @addr: starting address of the range | |
424 | * @size: size of the range in bytes | |
425 | * @flags: reservation flags (see linux/bootmem.h) | |
426 | * | |
427 | * Partial pages will be reserved. | |
428 | * | |
e2bf3cae | 429 | * The range must be contiguous but may span node boundaries. |
a66fd7da | 430 | */ |
223e8dc9 JW |
431 | int __init reserve_bootmem(unsigned long addr, unsigned long size, |
432 | int flags) | |
433 | { | |
e2bf3cae | 434 | unsigned long start, end; |
1da177e4 | 435 | |
e2bf3cae JW |
436 | start = PFN_DOWN(addr); |
437 | end = PFN_UP(addr + size); | |
223e8dc9 | 438 | |
e2bf3cae | 439 | return mark_bootmem(start, end, 1, flags); |
1da177e4 LT |
440 | } |
441 | ||
f88eff74 YL |
442 | int __weak __init reserve_bootmem_generic(unsigned long phys, unsigned long len, |
443 | int flags) | |
444 | { | |
445 | return reserve_bootmem(phys, len, flags); | |
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 | ||
d0c4f570 TH |
578 | static void * __init alloc_arch_preferred_bootmem(bootmem_data_t *bdata, |
579 | unsigned long size, unsigned long align, | |
580 | unsigned long goal, unsigned long limit) | |
581 | { | |
441c7e0a PE |
582 | if (WARN_ON_ONCE(slab_is_available())) |
583 | return kzalloc(size, GFP_NOWAIT); | |
584 | ||
d0c4f570 | 585 | #ifdef CONFIG_HAVE_ARCH_BOOTMEM |
433f13a7 JP |
586 | { |
587 | bootmem_data_t *p_bdata; | |
588 | ||
589 | p_bdata = bootmem_arch_preferred_node(bdata, size, align, | |
590 | goal, limit); | |
591 | if (p_bdata) | |
c6785b6b | 592 | return alloc_bootmem_bdata(p_bdata, size, align, |
433f13a7 JP |
593 | goal, limit); |
594 | } | |
d0c4f570 TH |
595 | #endif |
596 | return NULL; | |
597 | } | |
598 | ||
c12ab504 | 599 | static 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 | |
d0c4f570 TH |
607 | region = alloc_arch_preferred_bootmem(NULL, size, align, goal, limit); |
608 | if (region) | |
609 | return region; | |
0f3caba2 | 610 | |
d0c4f570 | 611 | list_for_each_entry(bdata, &bdata_list, list) { |
0f3caba2 JW |
612 | if (goal && bdata->node_low_pfn <= PFN_DOWN(goal)) |
613 | continue; | |
3560e249 | 614 | if (limit && bdata->node_min_pfn >= PFN_DOWN(limit)) |
0f3caba2 JW |
615 | break; |
616 | ||
c6785b6b | 617 | region = alloc_bootmem_bdata(bdata, size, align, goal, limit); |
0f3caba2 JW |
618 | if (region) |
619 | return region; | |
620 | } | |
621 | ||
c12ab504 JW |
622 | return NULL; |
623 | } | |
624 | ||
625 | static void * __init ___alloc_bootmem_nopanic(unsigned long size, | |
626 | unsigned long align, | |
627 | unsigned long goal, | |
628 | unsigned long limit) | |
629 | { | |
630 | void *ptr; | |
631 | ||
632 | restart: | |
633 | ptr = alloc_bootmem_core(size, align, goal, limit); | |
634 | if (ptr) | |
635 | return ptr; | |
5f2809e6 JW |
636 | if (goal) { |
637 | goal = 0; | |
0f3caba2 | 638 | goto restart; |
5f2809e6 | 639 | } |
2e5237da | 640 | |
5f2809e6 | 641 | return NULL; |
1da177e4 LT |
642 | } |
643 | ||
a66fd7da JW |
644 | /** |
645 | * __alloc_bootmem_nopanic - allocate boot memory without panicking | |
646 | * @size: size of the request in bytes | |
647 | * @align: alignment of the region | |
648 | * @goal: preferred starting address of the region | |
649 | * | |
650 | * The goal is dropped if it can not be satisfied and the allocation will | |
651 | * fall back to memory below @goal. | |
652 | * | |
653 | * Allocation may happen on any node in the system. | |
654 | * | |
655 | * Returns NULL on failure. | |
656 | */ | |
bb0923a6 | 657 | void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align, |
0f3caba2 | 658 | unsigned long goal) |
1da177e4 | 659 | { |
08677214 YL |
660 | unsigned long limit = 0; |
661 | ||
08677214 | 662 | return ___alloc_bootmem_nopanic(size, align, goal, limit); |
0f3caba2 | 663 | } |
1da177e4 | 664 | |
0f3caba2 JW |
665 | static void * __init ___alloc_bootmem(unsigned long size, unsigned long align, |
666 | unsigned long goal, unsigned long limit) | |
667 | { | |
668 | void *mem = ___alloc_bootmem_nopanic(size, align, goal, limit); | |
669 | ||
670 | if (mem) | |
671 | return mem; | |
672 | /* | |
673 | * Whoops, we cannot satisfy the allocation request. | |
674 | */ | |
675 | printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size); | |
676 | panic("Out of memory"); | |
a8062231 AK |
677 | return NULL; |
678 | } | |
1da177e4 | 679 | |
a66fd7da JW |
680 | /** |
681 | * __alloc_bootmem - allocate boot memory | |
682 | * @size: size of the request in bytes | |
683 | * @align: alignment of the region | |
684 | * @goal: preferred starting address of the region | |
685 | * | |
686 | * The goal is dropped if it can not be satisfied and the allocation will | |
687 | * fall back to memory below @goal. | |
688 | * | |
689 | * Allocation may happen on any node in the system. | |
690 | * | |
691 | * The function panics if the request can not be satisfied. | |
692 | */ | |
bb0923a6 FBH |
693 | void * __init __alloc_bootmem(unsigned long size, unsigned long align, |
694 | unsigned long goal) | |
a8062231 | 695 | { |
08677214 YL |
696 | unsigned long limit = 0; |
697 | ||
08677214 | 698 | return ___alloc_bootmem(size, align, goal, limit); |
1da177e4 LT |
699 | } |
700 | ||
4cc278b7 JW |
701 | static void * __init ___alloc_bootmem_node(bootmem_data_t *bdata, |
702 | unsigned long size, unsigned long align, | |
703 | unsigned long goal, unsigned long limit) | |
704 | { | |
705 | void *ptr; | |
706 | ||
ab381843 | 707 | again: |
d0c4f570 TH |
708 | ptr = alloc_arch_preferred_bootmem(bdata, size, align, goal, limit); |
709 | if (ptr) | |
710 | return ptr; | |
711 | ||
c6785b6b | 712 | ptr = alloc_bootmem_bdata(bdata, size, align, goal, limit); |
4cc278b7 JW |
713 | if (ptr) |
714 | return ptr; | |
715 | ||
ab381843 JW |
716 | ptr = alloc_bootmem_core(size, align, goal, limit); |
717 | if (ptr) | |
718 | return ptr; | |
719 | ||
720 | if (goal) { | |
721 | goal = 0; | |
722 | goto again; | |
723 | } | |
724 | ||
725 | printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size); | |
726 | panic("Out of memory"); | |
727 | return NULL; | |
4cc278b7 JW |
728 | } |
729 | ||
a66fd7da JW |
730 | /** |
731 | * __alloc_bootmem_node - allocate boot memory from a specific node | |
732 | * @pgdat: node to allocate from | |
733 | * @size: size of the request in bytes | |
734 | * @align: alignment of the region | |
735 | * @goal: preferred starting address of the region | |
736 | * | |
737 | * The goal is dropped if it can not be satisfied and the allocation will | |
738 | * fall back to memory below @goal. | |
739 | * | |
740 | * Allocation may fall back to any node in the system if the specified node | |
741 | * can not hold the requested memory. | |
742 | * | |
743 | * The function panics if the request can not be satisfied. | |
744 | */ | |
bb0923a6 FBH |
745 | void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size, |
746 | unsigned long align, unsigned long goal) | |
1da177e4 | 747 | { |
c91c4773 PE |
748 | if (WARN_ON_ONCE(slab_is_available())) |
749 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
750 | ||
09325873 | 751 | return ___alloc_bootmem_node(pgdat->bdata, size, align, goal, 0); |
08677214 YL |
752 | } |
753 | ||
754 | void * __init __alloc_bootmem_node_high(pg_data_t *pgdat, unsigned long size, | |
755 | unsigned long align, unsigned long goal) | |
756 | { | |
757 | #ifdef MAX_DMA32_PFN | |
758 | unsigned long end_pfn; | |
759 | ||
760 | if (WARN_ON_ONCE(slab_is_available())) | |
761 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
762 | ||
763 | /* update goal according ...MAX_DMA32_PFN */ | |
764 | end_pfn = pgdat->node_start_pfn + pgdat->node_spanned_pages; | |
765 | ||
766 | if (end_pfn > MAX_DMA32_PFN + (128 >> (20 - PAGE_SHIFT)) && | |
767 | (goal >> PAGE_SHIFT) < MAX_DMA32_PFN) { | |
768 | void *ptr; | |
769 | unsigned long new_goal; | |
770 | ||
771 | new_goal = MAX_DMA32_PFN << PAGE_SHIFT; | |
c6785b6b | 772 | ptr = alloc_bootmem_bdata(pgdat->bdata, size, align, |
08677214 | 773 | new_goal, 0); |
08677214 YL |
774 | if (ptr) |
775 | return ptr; | |
776 | } | |
777 | #endif | |
778 | ||
779 | return __alloc_bootmem_node(pgdat, size, align, goal); | |
780 | ||
1da177e4 LT |
781 | } |
782 | ||
e70260aa | 783 | #ifdef CONFIG_SPARSEMEM |
a66fd7da JW |
784 | /** |
785 | * alloc_bootmem_section - allocate boot memory from a specific section | |
786 | * @size: size of the request in bytes | |
787 | * @section_nr: sparse map section to allocate from | |
788 | * | |
789 | * Return NULL on failure. | |
790 | */ | |
e70260aa YG |
791 | void * __init alloc_bootmem_section(unsigned long size, |
792 | unsigned long section_nr) | |
793 | { | |
75a56cfe | 794 | bootmem_data_t *bdata; |
f5bf18fa | 795 | unsigned long pfn, goal; |
e70260aa YG |
796 | |
797 | pfn = section_nr_to_pfn(section_nr); | |
75a56cfe | 798 | goal = pfn << PAGE_SHIFT; |
75a56cfe | 799 | bdata = &bootmem_node_data[early_pfn_to_nid(pfn)]; |
e70260aa | 800 | |
c6785b6b | 801 | return alloc_bootmem_bdata(bdata, size, SMP_CACHE_BYTES, goal, 0); |
e70260aa YG |
802 | } |
803 | #endif | |
804 | ||
b54bbf7b AK |
805 | void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size, |
806 | unsigned long align, unsigned long goal) | |
807 | { | |
808 | void *ptr; | |
809 | ||
c91c4773 PE |
810 | if (WARN_ON_ONCE(slab_is_available())) |
811 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
812 | ||
d0c4f570 TH |
813 | ptr = alloc_arch_preferred_bootmem(pgdat->bdata, size, align, goal, 0); |
814 | if (ptr) | |
815 | return ptr; | |
816 | ||
c6785b6b | 817 | ptr = alloc_bootmem_bdata(pgdat->bdata, size, align, goal, 0); |
b54bbf7b AK |
818 | if (ptr) |
819 | return ptr; | |
820 | ||
821 | return __alloc_bootmem_nopanic(size, align, goal); | |
822 | } | |
823 | ||
dfd54cbc HC |
824 | #ifndef ARCH_LOW_ADDRESS_LIMIT |
825 | #define ARCH_LOW_ADDRESS_LIMIT 0xffffffffUL | |
826 | #endif | |
008857c1 | 827 | |
a66fd7da JW |
828 | /** |
829 | * __alloc_bootmem_low - allocate low boot memory | |
830 | * @size: size of the request in bytes | |
831 | * @align: alignment of the region | |
832 | * @goal: preferred starting address of the region | |
833 | * | |
834 | * The goal is dropped if it can not be satisfied and the allocation will | |
835 | * fall back to memory below @goal. | |
836 | * | |
837 | * Allocation may happen on any node in the system. | |
838 | * | |
839 | * The function panics if the request can not be satisfied. | |
840 | */ | |
bb0923a6 FBH |
841 | void * __init __alloc_bootmem_low(unsigned long size, unsigned long align, |
842 | unsigned long goal) | |
008857c1 | 843 | { |
0f3caba2 | 844 | return ___alloc_bootmem(size, align, goal, ARCH_LOW_ADDRESS_LIMIT); |
008857c1 RT |
845 | } |
846 | ||
a66fd7da JW |
847 | /** |
848 | * __alloc_bootmem_low_node - allocate low boot memory from a specific node | |
849 | * @pgdat: node to allocate from | |
850 | * @size: size of the request in bytes | |
851 | * @align: alignment of the region | |
852 | * @goal: preferred starting address of the region | |
853 | * | |
854 | * The goal is dropped if it can not be satisfied and the allocation will | |
855 | * fall back to memory below @goal. | |
856 | * | |
857 | * Allocation may fall back to any node in the system if the specified node | |
858 | * can not hold the requested memory. | |
859 | * | |
860 | * The function panics if the request can not be satisfied. | |
861 | */ | |
008857c1 RT |
862 | void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size, |
863 | unsigned long align, unsigned long goal) | |
864 | { | |
c91c4773 PE |
865 | if (WARN_ON_ONCE(slab_is_available())) |
866 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
867 | ||
09325873 | 868 | return ___alloc_bootmem_node(pgdat->bdata, size, align, |
b8ab9f82 | 869 | goal, ARCH_LOW_ADDRESS_LIMIT); |
008857c1 | 870 | } |