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> |
1da177e4 | 13 | #include <linux/bootmem.h> |
1da177e4 | 14 | #include <linux/module.h> |
ec3a354b | 15 | #include <linux/kmemleak.h> |
e786e86a FBH |
16 | |
17 | #include <asm/bug.h> | |
1da177e4 | 18 | #include <asm/io.h> |
dfd54cbc | 19 | #include <asm/processor.h> |
e786e86a | 20 | |
1da177e4 LT |
21 | #include "internal.h" |
22 | ||
1da177e4 LT |
23 | unsigned long max_low_pfn; |
24 | unsigned long min_low_pfn; | |
25 | unsigned long max_pfn; | |
26 | ||
92aa63a5 VG |
27 | #ifdef CONFIG_CRASH_DUMP |
28 | /* | |
29 | * If we have booted due to a crash, max_pfn will be a very low value. We need | |
30 | * to know the amount of memory that the previous kernel used. | |
31 | */ | |
32 | unsigned long saved_max_pfn; | |
33 | #endif | |
34 | ||
b61bfa3c JW |
35 | bootmem_data_t bootmem_node_data[MAX_NUMNODES] __initdata; |
36 | ||
636cc40c JW |
37 | static struct list_head bdata_list __initdata = LIST_HEAD_INIT(bdata_list); |
38 | ||
2e5237da JW |
39 | static int bootmem_debug; |
40 | ||
41 | static int __init bootmem_debug_setup(char *buf) | |
42 | { | |
43 | bootmem_debug = 1; | |
44 | return 0; | |
45 | } | |
46 | early_param("bootmem_debug", bootmem_debug_setup); | |
47 | ||
48 | #define bdebug(fmt, args...) ({ \ | |
49 | if (unlikely(bootmem_debug)) \ | |
50 | printk(KERN_INFO \ | |
51 | "bootmem::%s " fmt, \ | |
80a914dc | 52 | __func__, ## args); \ |
2e5237da JW |
53 | }) |
54 | ||
df049a5f | 55 | static unsigned long __init bootmap_bytes(unsigned long pages) |
223e8dc9 | 56 | { |
df049a5f | 57 | unsigned long bytes = (pages + 7) / 8; |
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 | 66 | unsigned 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 | 76 | static void __init link_bootmem(bootmem_data_t *bdata) |
679bc9fb | 77 | { |
636cc40c | 78 | struct list_head *iter; |
f71bf0ca | 79 | |
636cc40c JW |
80 | list_for_each(iter, &bdata_list) { |
81 | bootmem_data_t *ent; | |
82 | ||
83 | ent = list_entry(iter, bootmem_data_t, list); | |
3560e249 | 84 | if (bdata->node_min_pfn < ent->node_min_pfn) |
636cc40c | 85 | break; |
679bc9fb | 86 | } |
636cc40c | 87 | list_add_tail(&bdata->list, iter); |
679bc9fb KH |
88 | } |
89 | ||
1da177e4 LT |
90 | /* |
91 | * Called once to set up the allocator itself. | |
92 | */ | |
8ae04463 | 93 | static 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 |
126 | unsigned 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 |
139 | unsigned 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 | } | |
145 | ||
146 | static unsigned long __init free_all_bootmem_core(bootmem_data_t *bdata) | |
147 | { | |
41546c17 | 148 | int aligned; |
223e8dc9 | 149 | struct page *page; |
41546c17 JW |
150 | unsigned long start, end, pages, count = 0; |
151 | ||
152 | if (!bdata->node_bootmem_map) | |
153 | return 0; | |
154 | ||
3560e249 | 155 | start = bdata->node_min_pfn; |
41546c17 JW |
156 | end = bdata->node_low_pfn; |
157 | ||
223e8dc9 | 158 | /* |
41546c17 JW |
159 | * If the start is aligned to the machines wordsize, we might |
160 | * be able to free pages in bulks of that order. | |
223e8dc9 | 161 | */ |
41546c17 | 162 | aligned = !(start & (BITS_PER_LONG - 1)); |
223e8dc9 | 163 | |
41546c17 JW |
164 | bdebug("nid=%td start=%lx end=%lx aligned=%d\n", |
165 | bdata - bootmem_node_data, start, end, aligned); | |
223e8dc9 | 166 | |
41546c17 JW |
167 | while (start < end) { |
168 | unsigned long *map, idx, vec; | |
223e8dc9 | 169 | |
41546c17 | 170 | map = bdata->node_bootmem_map; |
3560e249 | 171 | idx = start - bdata->node_min_pfn; |
41546c17 JW |
172 | vec = ~map[idx / BITS_PER_LONG]; |
173 | ||
174 | if (aligned && vec == ~0UL && start + BITS_PER_LONG < end) { | |
175 | int order = ilog2(BITS_PER_LONG); | |
176 | ||
177 | __free_pages_bootmem(pfn_to_page(start), order); | |
223e8dc9 | 178 | count += BITS_PER_LONG; |
41546c17 JW |
179 | } else { |
180 | unsigned long off = 0; | |
181 | ||
182 | while (vec && off < BITS_PER_LONG) { | |
183 | if (vec & 1) { | |
184 | page = pfn_to_page(start + off); | |
223e8dc9 | 185 | __free_pages_bootmem(page, 0); |
41546c17 | 186 | count++; |
223e8dc9 | 187 | } |
41546c17 JW |
188 | vec >>= 1; |
189 | off++; | |
223e8dc9 | 190 | } |
223e8dc9 | 191 | } |
41546c17 | 192 | start += BITS_PER_LONG; |
223e8dc9 JW |
193 | } |
194 | ||
223e8dc9 | 195 | page = virt_to_page(bdata->node_bootmem_map); |
3560e249 | 196 | pages = bdata->node_low_pfn - bdata->node_min_pfn; |
41546c17 JW |
197 | pages = bootmem_bootmap_pages(pages); |
198 | count += pages; | |
199 | while (pages--) | |
200 | __free_pages_bootmem(page++, 0); | |
223e8dc9 | 201 | |
2e5237da JW |
202 | bdebug("nid=%td released=%lx\n", bdata - bootmem_node_data, count); |
203 | ||
223e8dc9 JW |
204 | return count; |
205 | } | |
206 | ||
a66fd7da JW |
207 | /** |
208 | * free_all_bootmem_node - release a node's free pages to the buddy allocator | |
209 | * @pgdat: node to be released | |
210 | * | |
211 | * Returns the number of pages actually released. | |
212 | */ | |
223e8dc9 JW |
213 | unsigned long __init free_all_bootmem_node(pg_data_t *pgdat) |
214 | { | |
215 | register_page_bootmem_info_node(pgdat); | |
216 | return free_all_bootmem_core(pgdat->bdata); | |
217 | } | |
218 | ||
a66fd7da JW |
219 | /** |
220 | * free_all_bootmem - release free pages to the buddy allocator | |
221 | * | |
222 | * Returns the number of pages actually released. | |
223 | */ | |
223e8dc9 JW |
224 | unsigned long __init free_all_bootmem(void) |
225 | { | |
226 | return free_all_bootmem_core(NODE_DATA(0)->bdata); | |
227 | } | |
228 | ||
d747fa4b JW |
229 | static void __init __free(bootmem_data_t *bdata, |
230 | unsigned long sidx, unsigned long eidx) | |
231 | { | |
232 | unsigned long idx; | |
233 | ||
234 | bdebug("nid=%td start=%lx end=%lx\n", bdata - bootmem_node_data, | |
3560e249 JW |
235 | sidx + bdata->node_min_pfn, |
236 | eidx + bdata->node_min_pfn); | |
d747fa4b | 237 | |
e2bf3cae JW |
238 | if (bdata->hint_idx > sidx) |
239 | bdata->hint_idx = sidx; | |
240 | ||
d747fa4b JW |
241 | for (idx = sidx; idx < eidx; idx++) |
242 | if (!test_and_clear_bit(idx, bdata->node_bootmem_map)) | |
243 | BUG(); | |
244 | } | |
245 | ||
246 | static int __init __reserve(bootmem_data_t *bdata, unsigned long sidx, | |
247 | unsigned long eidx, int flags) | |
248 | { | |
249 | unsigned long idx; | |
250 | int exclusive = flags & BOOTMEM_EXCLUSIVE; | |
251 | ||
252 | bdebug("nid=%td start=%lx end=%lx flags=%x\n", | |
253 | bdata - bootmem_node_data, | |
3560e249 JW |
254 | sidx + bdata->node_min_pfn, |
255 | eidx + bdata->node_min_pfn, | |
d747fa4b JW |
256 | flags); |
257 | ||
258 | for (idx = sidx; idx < eidx; idx++) | |
259 | if (test_and_set_bit(idx, bdata->node_bootmem_map)) { | |
260 | if (exclusive) { | |
261 | __free(bdata, sidx, idx); | |
262 | return -EBUSY; | |
263 | } | |
264 | bdebug("silent double reserve of PFN %lx\n", | |
3560e249 | 265 | idx + bdata->node_min_pfn); |
d747fa4b JW |
266 | } |
267 | return 0; | |
268 | } | |
269 | ||
e2bf3cae JW |
270 | static int __init mark_bootmem_node(bootmem_data_t *bdata, |
271 | unsigned long start, unsigned long end, | |
272 | int reserve, int flags) | |
223e8dc9 JW |
273 | { |
274 | unsigned long sidx, eidx; | |
223e8dc9 | 275 | |
e2bf3cae JW |
276 | bdebug("nid=%td start=%lx end=%lx reserve=%d flags=%x\n", |
277 | bdata - bootmem_node_data, start, end, reserve, flags); | |
223e8dc9 | 278 | |
3560e249 | 279 | BUG_ON(start < bdata->node_min_pfn); |
e2bf3cae | 280 | BUG_ON(end > bdata->node_low_pfn); |
223e8dc9 | 281 | |
3560e249 JW |
282 | sidx = start - bdata->node_min_pfn; |
283 | eidx = end - bdata->node_min_pfn; | |
223e8dc9 | 284 | |
e2bf3cae JW |
285 | if (reserve) |
286 | return __reserve(bdata, sidx, eidx, flags); | |
223e8dc9 | 287 | else |
e2bf3cae JW |
288 | __free(bdata, sidx, eidx); |
289 | return 0; | |
290 | } | |
291 | ||
292 | static int __init mark_bootmem(unsigned long start, unsigned long end, | |
293 | int reserve, int flags) | |
294 | { | |
295 | unsigned long pos; | |
296 | bootmem_data_t *bdata; | |
297 | ||
298 | pos = start; | |
299 | list_for_each_entry(bdata, &bdata_list, list) { | |
300 | int err; | |
301 | unsigned long max; | |
302 | ||
3560e249 JW |
303 | if (pos < bdata->node_min_pfn || |
304 | pos >= bdata->node_low_pfn) { | |
e2bf3cae JW |
305 | BUG_ON(pos != start); |
306 | continue; | |
307 | } | |
308 | ||
309 | max = min(bdata->node_low_pfn, end); | |
223e8dc9 | 310 | |
e2bf3cae JW |
311 | err = mark_bootmem_node(bdata, pos, max, reserve, flags); |
312 | if (reserve && err) { | |
313 | mark_bootmem(start, pos, 0, 0); | |
314 | return err; | |
315 | } | |
223e8dc9 | 316 | |
e2bf3cae JW |
317 | if (max == end) |
318 | return 0; | |
319 | pos = bdata->node_low_pfn; | |
320 | } | |
321 | BUG(); | |
223e8dc9 JW |
322 | } |
323 | ||
a66fd7da JW |
324 | /** |
325 | * free_bootmem_node - mark a page range as usable | |
326 | * @pgdat: node the range resides on | |
327 | * @physaddr: starting address of the range | |
328 | * @size: size of the range in bytes | |
329 | * | |
330 | * Partial pages will be considered reserved and left as they are. | |
331 | * | |
e2bf3cae | 332 | * The range must reside completely on the specified node. |
a66fd7da | 333 | */ |
223e8dc9 JW |
334 | void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
335 | unsigned long size) | |
336 | { | |
e2bf3cae JW |
337 | unsigned long start, end; |
338 | ||
ec3a354b CM |
339 | kmemleak_free_part(__va(physaddr), size); |
340 | ||
e2bf3cae JW |
341 | start = PFN_UP(physaddr); |
342 | end = PFN_DOWN(physaddr + size); | |
343 | ||
344 | mark_bootmem_node(pgdat->bdata, start, end, 0, 0); | |
223e8dc9 JW |
345 | } |
346 | ||
a66fd7da JW |
347 | /** |
348 | * free_bootmem - mark a page range as usable | |
349 | * @addr: starting address of the range | |
350 | * @size: size of the range in bytes | |
351 | * | |
352 | * Partial pages will be considered reserved and left as they are. | |
353 | * | |
e2bf3cae | 354 | * The range must be contiguous but may span node boundaries. |
a66fd7da | 355 | */ |
223e8dc9 JW |
356 | void __init free_bootmem(unsigned long addr, unsigned long size) |
357 | { | |
e2bf3cae | 358 | unsigned long start, end; |
a5645a61 | 359 | |
ec3a354b CM |
360 | kmemleak_free_part(__va(addr), size); |
361 | ||
e2bf3cae JW |
362 | start = PFN_UP(addr); |
363 | end = PFN_DOWN(addr + size); | |
1da177e4 | 364 | |
e2bf3cae | 365 | mark_bootmem(start, end, 0, 0); |
1da177e4 LT |
366 | } |
367 | ||
a66fd7da JW |
368 | /** |
369 | * reserve_bootmem_node - mark a page range as reserved | |
370 | * @pgdat: node the range resides on | |
371 | * @physaddr: starting address of the range | |
372 | * @size: size of the range in bytes | |
373 | * @flags: reservation flags (see linux/bootmem.h) | |
374 | * | |
375 | * Partial pages will be reserved. | |
376 | * | |
e2bf3cae | 377 | * The range must reside completely on the specified node. |
a66fd7da | 378 | */ |
223e8dc9 JW |
379 | int __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
380 | unsigned long size, int flags) | |
1da177e4 | 381 | { |
e2bf3cae | 382 | unsigned long start, end; |
1da177e4 | 383 | |
e2bf3cae JW |
384 | start = PFN_DOWN(physaddr); |
385 | end = PFN_UP(physaddr + size); | |
386 | ||
387 | return mark_bootmem_node(pgdat->bdata, start, end, 1, flags); | |
223e8dc9 | 388 | } |
5a982cbc | 389 | |
a66fd7da JW |
390 | /** |
391 | * reserve_bootmem - mark a page range as usable | |
392 | * @addr: starting address of the range | |
393 | * @size: size of the range in bytes | |
394 | * @flags: reservation flags (see linux/bootmem.h) | |
395 | * | |
396 | * Partial pages will be reserved. | |
397 | * | |
e2bf3cae | 398 | * The range must be contiguous but may span node boundaries. |
a66fd7da | 399 | */ |
223e8dc9 JW |
400 | int __init reserve_bootmem(unsigned long addr, unsigned long size, |
401 | int flags) | |
402 | { | |
e2bf3cae | 403 | unsigned long start, end; |
1da177e4 | 404 | |
e2bf3cae JW |
405 | start = PFN_DOWN(addr); |
406 | end = PFN_UP(addr + size); | |
223e8dc9 | 407 | |
e2bf3cae | 408 | return mark_bootmem(start, end, 1, flags); |
1da177e4 LT |
409 | } |
410 | ||
481ebd0d JW |
411 | static unsigned long align_idx(struct bootmem_data *bdata, unsigned long idx, |
412 | unsigned long step) | |
413 | { | |
414 | unsigned long base = bdata->node_min_pfn; | |
415 | ||
416 | /* | |
417 | * Align the index with respect to the node start so that the | |
418 | * combination of both satisfies the requested alignment. | |
419 | */ | |
420 | ||
421 | return ALIGN(base + idx, step) - base; | |
422 | } | |
423 | ||
424 | static unsigned long align_off(struct bootmem_data *bdata, unsigned long off, | |
425 | unsigned long align) | |
426 | { | |
427 | unsigned long base = PFN_PHYS(bdata->node_min_pfn); | |
428 | ||
429 | /* Same as align_idx for byte offsets */ | |
430 | ||
431 | return ALIGN(base + off, align) - base; | |
432 | } | |
433 | ||
d0c4f570 TH |
434 | static void * __init alloc_bootmem_core(struct bootmem_data *bdata, |
435 | unsigned long size, unsigned long align, | |
436 | unsigned long goal, unsigned long limit) | |
1da177e4 | 437 | { |
0f3caba2 | 438 | unsigned long fallback = 0; |
5f2809e6 JW |
439 | unsigned long min, max, start, sidx, midx, step; |
440 | ||
594fe1a0 JW |
441 | bdebug("nid=%td size=%lx [%lu pages] align=%lx goal=%lx limit=%lx\n", |
442 | bdata - bootmem_node_data, size, PAGE_ALIGN(size) >> PAGE_SHIFT, | |
443 | align, goal, limit); | |
444 | ||
5f2809e6 JW |
445 | BUG_ON(!size); |
446 | BUG_ON(align & (align - 1)); | |
447 | BUG_ON(limit && goal + size > limit); | |
1da177e4 | 448 | |
7c309a64 CK |
449 | if (!bdata->node_bootmem_map) |
450 | return NULL; | |
451 | ||
3560e249 | 452 | min = bdata->node_min_pfn; |
5f2809e6 | 453 | max = bdata->node_low_pfn; |
9a2dc04c | 454 | |
5f2809e6 JW |
455 | goal >>= PAGE_SHIFT; |
456 | limit >>= PAGE_SHIFT; | |
457 | ||
458 | if (limit && max > limit) | |
459 | max = limit; | |
460 | if (max <= min) | |
9a2dc04c YL |
461 | return NULL; |
462 | ||
5f2809e6 | 463 | step = max(align >> PAGE_SHIFT, 1UL); |
281dd25c | 464 | |
5f2809e6 JW |
465 | if (goal && min < goal && goal < max) |
466 | start = ALIGN(goal, step); | |
467 | else | |
468 | start = ALIGN(min, step); | |
1da177e4 | 469 | |
481ebd0d | 470 | sidx = start - bdata->node_min_pfn; |
3560e249 | 471 | midx = max - bdata->node_min_pfn; |
1da177e4 | 472 | |
5f2809e6 | 473 | if (bdata->hint_idx > sidx) { |
0f3caba2 JW |
474 | /* |
475 | * Handle the valid case of sidx being zero and still | |
476 | * catch the fallback below. | |
477 | */ | |
478 | fallback = sidx + 1; | |
481ebd0d | 479 | sidx = align_idx(bdata, bdata->hint_idx, step); |
5f2809e6 | 480 | } |
1da177e4 | 481 | |
5f2809e6 JW |
482 | while (1) { |
483 | int merge; | |
484 | void *region; | |
485 | unsigned long eidx, i, start_off, end_off; | |
486 | find_block: | |
487 | sidx = find_next_zero_bit(bdata->node_bootmem_map, midx, sidx); | |
481ebd0d | 488 | sidx = align_idx(bdata, sidx, step); |
5f2809e6 | 489 | eidx = sidx + PFN_UP(size); |
ad09315c | 490 | |
5f2809e6 | 491 | if (sidx >= midx || eidx > midx) |
66d43e98 | 492 | break; |
1da177e4 | 493 | |
5f2809e6 JW |
494 | for (i = sidx; i < eidx; i++) |
495 | if (test_bit(i, bdata->node_bootmem_map)) { | |
481ebd0d | 496 | sidx = align_idx(bdata, i, step); |
5f2809e6 JW |
497 | if (sidx == i) |
498 | sidx += step; | |
499 | goto find_block; | |
500 | } | |
1da177e4 | 501 | |
627240aa | 502 | if (bdata->last_end_off & (PAGE_SIZE - 1) && |
5f2809e6 | 503 | PFN_DOWN(bdata->last_end_off) + 1 == sidx) |
481ebd0d | 504 | start_off = align_off(bdata, bdata->last_end_off, align); |
5f2809e6 JW |
505 | else |
506 | start_off = PFN_PHYS(sidx); | |
507 | ||
508 | merge = PFN_DOWN(start_off) < sidx; | |
509 | end_off = start_off + size; | |
510 | ||
511 | bdata->last_end_off = end_off; | |
512 | bdata->hint_idx = PFN_UP(end_off); | |
513 | ||
514 | /* | |
515 | * Reserve the area now: | |
516 | */ | |
d747fa4b JW |
517 | if (__reserve(bdata, PFN_DOWN(start_off) + merge, |
518 | PFN_UP(end_off), BOOTMEM_EXCLUSIVE)) | |
519 | BUG(); | |
5f2809e6 | 520 | |
3560e249 JW |
521 | region = phys_to_virt(PFN_PHYS(bdata->node_min_pfn) + |
522 | start_off); | |
5f2809e6 | 523 | memset(region, 0, size); |
ec3a354b | 524 | kmemleak_alloc(region, size, 1, 0); |
5f2809e6 | 525 | return region; |
1da177e4 LT |
526 | } |
527 | ||
0f3caba2 | 528 | if (fallback) { |
481ebd0d | 529 | sidx = align_idx(bdata, fallback - 1, step); |
0f3caba2 JW |
530 | fallback = 0; |
531 | goto find_block; | |
532 | } | |
533 | ||
534 | return NULL; | |
535 | } | |
536 | ||
d0c4f570 TH |
537 | static void * __init alloc_arch_preferred_bootmem(bootmem_data_t *bdata, |
538 | unsigned long size, unsigned long align, | |
539 | unsigned long goal, unsigned long limit) | |
540 | { | |
441c7e0a PE |
541 | if (WARN_ON_ONCE(slab_is_available())) |
542 | return kzalloc(size, GFP_NOWAIT); | |
543 | ||
d0c4f570 | 544 | #ifdef CONFIG_HAVE_ARCH_BOOTMEM |
433f13a7 JP |
545 | { |
546 | bootmem_data_t *p_bdata; | |
547 | ||
548 | p_bdata = bootmem_arch_preferred_node(bdata, size, align, | |
549 | goal, limit); | |
550 | if (p_bdata) | |
551 | return alloc_bootmem_core(p_bdata, size, align, | |
552 | goal, limit); | |
553 | } | |
d0c4f570 TH |
554 | #endif |
555 | return NULL; | |
556 | } | |
557 | ||
0f3caba2 JW |
558 | static void * __init ___alloc_bootmem_nopanic(unsigned long size, |
559 | unsigned long align, | |
560 | unsigned long goal, | |
561 | unsigned long limit) | |
562 | { | |
563 | bootmem_data_t *bdata; | |
d0c4f570 | 564 | void *region; |
0f3caba2 JW |
565 | |
566 | restart: | |
d0c4f570 TH |
567 | region = alloc_arch_preferred_bootmem(NULL, size, align, goal, limit); |
568 | if (region) | |
569 | return region; | |
0f3caba2 | 570 | |
d0c4f570 | 571 | list_for_each_entry(bdata, &bdata_list, list) { |
0f3caba2 JW |
572 | if (goal && bdata->node_low_pfn <= PFN_DOWN(goal)) |
573 | continue; | |
3560e249 | 574 | if (limit && bdata->node_min_pfn >= PFN_DOWN(limit)) |
0f3caba2 JW |
575 | break; |
576 | ||
577 | region = alloc_bootmem_core(bdata, size, align, goal, limit); | |
578 | if (region) | |
579 | return region; | |
580 | } | |
581 | ||
5f2809e6 JW |
582 | if (goal) { |
583 | goal = 0; | |
0f3caba2 | 584 | goto restart; |
5f2809e6 | 585 | } |
2e5237da | 586 | |
5f2809e6 | 587 | return NULL; |
1da177e4 LT |
588 | } |
589 | ||
a66fd7da JW |
590 | /** |
591 | * __alloc_bootmem_nopanic - allocate boot memory without panicking | |
592 | * @size: size of the request in bytes | |
593 | * @align: alignment of the region | |
594 | * @goal: preferred starting address of the region | |
595 | * | |
596 | * The goal is dropped if it can not be satisfied and the allocation will | |
597 | * fall back to memory below @goal. | |
598 | * | |
599 | * Allocation may happen on any node in the system. | |
600 | * | |
601 | * Returns NULL on failure. | |
602 | */ | |
bb0923a6 | 603 | void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align, |
0f3caba2 | 604 | unsigned long goal) |
1da177e4 | 605 | { |
0f3caba2 JW |
606 | return ___alloc_bootmem_nopanic(size, align, goal, 0); |
607 | } | |
1da177e4 | 608 | |
0f3caba2 JW |
609 | static void * __init ___alloc_bootmem(unsigned long size, unsigned long align, |
610 | unsigned long goal, unsigned long limit) | |
611 | { | |
612 | void *mem = ___alloc_bootmem_nopanic(size, align, goal, limit); | |
613 | ||
614 | if (mem) | |
615 | return mem; | |
616 | /* | |
617 | * Whoops, we cannot satisfy the allocation request. | |
618 | */ | |
619 | printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size); | |
620 | panic("Out of memory"); | |
a8062231 AK |
621 | return NULL; |
622 | } | |
1da177e4 | 623 | |
a66fd7da JW |
624 | /** |
625 | * __alloc_bootmem - allocate boot memory | |
626 | * @size: size of the request in bytes | |
627 | * @align: alignment of the region | |
628 | * @goal: preferred starting address of the region | |
629 | * | |
630 | * The goal is dropped if it can not be satisfied and the allocation will | |
631 | * fall back to memory below @goal. | |
632 | * | |
633 | * Allocation may happen on any node in the system. | |
634 | * | |
635 | * The function panics if the request can not be satisfied. | |
636 | */ | |
bb0923a6 FBH |
637 | void * __init __alloc_bootmem(unsigned long size, unsigned long align, |
638 | unsigned long goal) | |
a8062231 | 639 | { |
0f3caba2 | 640 | return ___alloc_bootmem(size, align, goal, 0); |
1da177e4 LT |
641 | } |
642 | ||
4cc278b7 JW |
643 | static void * __init ___alloc_bootmem_node(bootmem_data_t *bdata, |
644 | unsigned long size, unsigned long align, | |
645 | unsigned long goal, unsigned long limit) | |
646 | { | |
647 | void *ptr; | |
648 | ||
d0c4f570 TH |
649 | ptr = alloc_arch_preferred_bootmem(bdata, size, align, goal, limit); |
650 | if (ptr) | |
651 | return ptr; | |
652 | ||
4cc278b7 JW |
653 | ptr = alloc_bootmem_core(bdata, size, align, goal, limit); |
654 | if (ptr) | |
655 | return ptr; | |
656 | ||
657 | return ___alloc_bootmem(size, align, goal, limit); | |
658 | } | |
659 | ||
a66fd7da JW |
660 | /** |
661 | * __alloc_bootmem_node - allocate boot memory from a specific node | |
662 | * @pgdat: node to allocate from | |
663 | * @size: size of the request in bytes | |
664 | * @align: alignment of the region | |
665 | * @goal: preferred starting address of the region | |
666 | * | |
667 | * The goal is dropped if it can not be satisfied and the allocation will | |
668 | * fall back to memory below @goal. | |
669 | * | |
670 | * Allocation may fall back to any node in the system if the specified node | |
671 | * can not hold the requested memory. | |
672 | * | |
673 | * The function panics if the request can not be satisfied. | |
674 | */ | |
bb0923a6 FBH |
675 | void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size, |
676 | unsigned long align, unsigned long goal) | |
1da177e4 | 677 | { |
c91c4773 PE |
678 | if (WARN_ON_ONCE(slab_is_available())) |
679 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
680 | ||
4cc278b7 | 681 | return ___alloc_bootmem_node(pgdat->bdata, size, align, goal, 0); |
1da177e4 LT |
682 | } |
683 | ||
e70260aa | 684 | #ifdef CONFIG_SPARSEMEM |
a66fd7da JW |
685 | /** |
686 | * alloc_bootmem_section - allocate boot memory from a specific section | |
687 | * @size: size of the request in bytes | |
688 | * @section_nr: sparse map section to allocate from | |
689 | * | |
690 | * Return NULL on failure. | |
691 | */ | |
e70260aa YG |
692 | void * __init alloc_bootmem_section(unsigned long size, |
693 | unsigned long section_nr) | |
694 | { | |
75a56cfe JW |
695 | bootmem_data_t *bdata; |
696 | unsigned long pfn, goal, limit; | |
e70260aa YG |
697 | |
698 | pfn = section_nr_to_pfn(section_nr); | |
75a56cfe JW |
699 | goal = pfn << PAGE_SHIFT; |
700 | limit = section_nr_to_pfn(section_nr + 1) << PAGE_SHIFT; | |
701 | bdata = &bootmem_node_data[early_pfn_to_nid(pfn)]; | |
e70260aa | 702 | |
75a56cfe | 703 | return alloc_bootmem_core(bdata, size, SMP_CACHE_BYTES, goal, limit); |
e70260aa YG |
704 | } |
705 | #endif | |
706 | ||
b54bbf7b AK |
707 | void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size, |
708 | unsigned long align, unsigned long goal) | |
709 | { | |
710 | void *ptr; | |
711 | ||
c91c4773 PE |
712 | if (WARN_ON_ONCE(slab_is_available())) |
713 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
714 | ||
d0c4f570 TH |
715 | ptr = alloc_arch_preferred_bootmem(pgdat->bdata, size, align, goal, 0); |
716 | if (ptr) | |
717 | return ptr; | |
718 | ||
b54bbf7b AK |
719 | ptr = alloc_bootmem_core(pgdat->bdata, size, align, goal, 0); |
720 | if (ptr) | |
721 | return ptr; | |
722 | ||
723 | return __alloc_bootmem_nopanic(size, align, goal); | |
724 | } | |
725 | ||
dfd54cbc HC |
726 | #ifndef ARCH_LOW_ADDRESS_LIMIT |
727 | #define ARCH_LOW_ADDRESS_LIMIT 0xffffffffUL | |
728 | #endif | |
008857c1 | 729 | |
a66fd7da JW |
730 | /** |
731 | * __alloc_bootmem_low - allocate low boot memory | |
732 | * @size: size of the request in bytes | |
733 | * @align: alignment of the region | |
734 | * @goal: preferred starting address of the region | |
735 | * | |
736 | * The goal is dropped if it can not be satisfied and the allocation will | |
737 | * fall back to memory below @goal. | |
738 | * | |
739 | * Allocation may happen on any node in the system. | |
740 | * | |
741 | * The function panics if the request can not be satisfied. | |
742 | */ | |
bb0923a6 FBH |
743 | void * __init __alloc_bootmem_low(unsigned long size, unsigned long align, |
744 | unsigned long goal) | |
008857c1 | 745 | { |
0f3caba2 | 746 | return ___alloc_bootmem(size, align, goal, ARCH_LOW_ADDRESS_LIMIT); |
008857c1 RT |
747 | } |
748 | ||
a66fd7da JW |
749 | /** |
750 | * __alloc_bootmem_low_node - allocate low boot memory from a specific node | |
751 | * @pgdat: node to allocate from | |
752 | * @size: size of the request in bytes | |
753 | * @align: alignment of the region | |
754 | * @goal: preferred starting address of the region | |
755 | * | |
756 | * The goal is dropped if it can not be satisfied and the allocation will | |
757 | * fall back to memory below @goal. | |
758 | * | |
759 | * Allocation may fall back to any node in the system if the specified node | |
760 | * can not hold the requested memory. | |
761 | * | |
762 | * The function panics if the request can not be satisfied. | |
763 | */ | |
008857c1 RT |
764 | void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size, |
765 | unsigned long align, unsigned long goal) | |
766 | { | |
c91c4773 PE |
767 | if (WARN_ON_ONCE(slab_is_available())) |
768 | return kzalloc_node(size, GFP_NOWAIT, pgdat->node_id); | |
769 | ||
4cc278b7 JW |
770 | return ___alloc_bootmem_node(pgdat->bdata, size, align, |
771 | goal, ARCH_LOW_ADDRESS_LIMIT); | |
008857c1 | 772 | } |