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> |
e786e86a FBH |
15 | |
16 | #include <asm/bug.h> | |
1da177e4 | 17 | #include <asm/io.h> |
dfd54cbc | 18 | #include <asm/processor.h> |
e786e86a | 19 | |
1da177e4 LT |
20 | #include "internal.h" |
21 | ||
1da177e4 LT |
22 | unsigned long max_low_pfn; |
23 | unsigned long min_low_pfn; | |
24 | unsigned long max_pfn; | |
25 | ||
92aa63a5 VG |
26 | #ifdef CONFIG_CRASH_DUMP |
27 | /* | |
28 | * If we have booted due to a crash, max_pfn will be a very low value. We need | |
29 | * to know the amount of memory that the previous kernel used. | |
30 | */ | |
31 | unsigned long saved_max_pfn; | |
32 | #endif | |
33 | ||
b61bfa3c JW |
34 | bootmem_data_t bootmem_node_data[MAX_NUMNODES] __initdata; |
35 | ||
636cc40c JW |
36 | static struct list_head bdata_list __initdata = LIST_HEAD_INIT(bdata_list); |
37 | ||
2e5237da JW |
38 | static int bootmem_debug; |
39 | ||
40 | static int __init bootmem_debug_setup(char *buf) | |
41 | { | |
42 | bootmem_debug = 1; | |
43 | return 0; | |
44 | } | |
45 | early_param("bootmem_debug", bootmem_debug_setup); | |
46 | ||
47 | #define bdebug(fmt, args...) ({ \ | |
48 | if (unlikely(bootmem_debug)) \ | |
49 | printk(KERN_INFO \ | |
50 | "bootmem::%s " fmt, \ | |
51 | __FUNCTION__, ## args); \ | |
52 | }) | |
53 | ||
df049a5f | 54 | static unsigned long __init bootmap_bytes(unsigned long pages) |
223e8dc9 | 55 | { |
df049a5f | 56 | unsigned long bytes = (pages + 7) / 8; |
223e8dc9 | 57 | |
df049a5f | 58 | return ALIGN(bytes, sizeof(long)); |
223e8dc9 JW |
59 | } |
60 | ||
a66fd7da JW |
61 | /** |
62 | * bootmem_bootmap_pages - calculate bitmap size in pages | |
63 | * @pages: number of pages the bitmap has to represent | |
64 | */ | |
f71bf0ca | 65 | unsigned long __init bootmem_bootmap_pages(unsigned long pages) |
1da177e4 | 66 | { |
df049a5f | 67 | unsigned long bytes = bootmap_bytes(pages); |
1da177e4 | 68 | |
df049a5f | 69 | return PAGE_ALIGN(bytes) >> PAGE_SHIFT; |
1da177e4 | 70 | } |
f71bf0ca | 71 | |
679bc9fb KH |
72 | /* |
73 | * link bdata in order | |
74 | */ | |
69d49e68 | 75 | static void __init link_bootmem(bootmem_data_t *bdata) |
679bc9fb | 76 | { |
636cc40c | 77 | struct list_head *iter; |
f71bf0ca | 78 | |
636cc40c JW |
79 | list_for_each(iter, &bdata_list) { |
80 | bootmem_data_t *ent; | |
81 | ||
82 | ent = list_entry(iter, bootmem_data_t, list); | |
83 | if (bdata->node_boot_start < ent->node_boot_start) | |
84 | break; | |
679bc9fb | 85 | } |
636cc40c | 86 | list_add_tail(&bdata->list, iter); |
679bc9fb KH |
87 | } |
88 | ||
1da177e4 LT |
89 | /* |
90 | * Called once to set up the allocator itself. | |
91 | */ | |
8ae04463 | 92 | static unsigned long __init init_bootmem_core(bootmem_data_t *bdata, |
1da177e4 LT |
93 | unsigned long mapstart, unsigned long start, unsigned long end) |
94 | { | |
bbc7b92e | 95 | unsigned long mapsize; |
1da177e4 | 96 | |
2dbb51c4 | 97 | mminit_validate_memmodel_limits(&start, &end); |
bbc7b92e FBH |
98 | bdata->node_bootmem_map = phys_to_virt(PFN_PHYS(mapstart)); |
99 | bdata->node_boot_start = PFN_PHYS(start); | |
1da177e4 | 100 | bdata->node_low_pfn = end; |
679bc9fb | 101 | link_bootmem(bdata); |
1da177e4 LT |
102 | |
103 | /* | |
104 | * Initially all pages are reserved - setup_arch() has to | |
105 | * register free RAM areas explicitly. | |
106 | */ | |
df049a5f | 107 | mapsize = bootmap_bytes(end - start); |
1da177e4 LT |
108 | memset(bdata->node_bootmem_map, 0xff, mapsize); |
109 | ||
2e5237da JW |
110 | bdebug("nid=%td start=%lx map=%lx end=%lx mapsize=%lx\n", |
111 | bdata - bootmem_node_data, start, mapstart, end, mapsize); | |
112 | ||
1da177e4 LT |
113 | return mapsize; |
114 | } | |
115 | ||
a66fd7da JW |
116 | /** |
117 | * init_bootmem_node - register a node as boot memory | |
118 | * @pgdat: node to register | |
119 | * @freepfn: pfn where the bitmap for this node is to be placed | |
120 | * @startpfn: first pfn on the node | |
121 | * @endpfn: first pfn after the node | |
122 | * | |
123 | * Returns the number of bytes needed to hold the bitmap for this node. | |
124 | */ | |
223e8dc9 JW |
125 | unsigned long __init init_bootmem_node(pg_data_t *pgdat, unsigned long freepfn, |
126 | unsigned long startpfn, unsigned long endpfn) | |
127 | { | |
128 | return init_bootmem_core(pgdat->bdata, freepfn, startpfn, endpfn); | |
129 | } | |
130 | ||
a66fd7da JW |
131 | /** |
132 | * init_bootmem - register boot memory | |
133 | * @start: pfn where the bitmap is to be placed | |
134 | * @pages: number of available physical pages | |
135 | * | |
136 | * Returns the number of bytes needed to hold the bitmap. | |
137 | */ | |
223e8dc9 JW |
138 | unsigned long __init init_bootmem(unsigned long start, unsigned long pages) |
139 | { | |
140 | max_low_pfn = pages; | |
141 | min_low_pfn = start; | |
142 | return init_bootmem_core(NODE_DATA(0)->bdata, start, 0, pages); | |
143 | } | |
144 | ||
145 | static unsigned long __init free_all_bootmem_core(bootmem_data_t *bdata) | |
146 | { | |
41546c17 | 147 | int aligned; |
223e8dc9 | 148 | struct page *page; |
41546c17 JW |
149 | unsigned long start, end, pages, count = 0; |
150 | ||
151 | if (!bdata->node_bootmem_map) | |
152 | return 0; | |
153 | ||
154 | start = PFN_DOWN(bdata->node_boot_start); | |
155 | end = bdata->node_low_pfn; | |
156 | ||
223e8dc9 | 157 | /* |
41546c17 JW |
158 | * If the start is aligned to the machines wordsize, we might |
159 | * be able to free pages in bulks of that order. | |
223e8dc9 | 160 | */ |
41546c17 | 161 | aligned = !(start & (BITS_PER_LONG - 1)); |
223e8dc9 | 162 | |
41546c17 JW |
163 | bdebug("nid=%td start=%lx end=%lx aligned=%d\n", |
164 | bdata - bootmem_node_data, start, end, aligned); | |
223e8dc9 | 165 | |
41546c17 JW |
166 | while (start < end) { |
167 | unsigned long *map, idx, vec; | |
223e8dc9 | 168 | |
41546c17 JW |
169 | map = bdata->node_bootmem_map; |
170 | idx = start - PFN_DOWN(bdata->node_boot_start); | |
171 | vec = ~map[idx / BITS_PER_LONG]; | |
172 | ||
173 | if (aligned && vec == ~0UL && start + BITS_PER_LONG < end) { | |
174 | int order = ilog2(BITS_PER_LONG); | |
175 | ||
176 | __free_pages_bootmem(pfn_to_page(start), order); | |
223e8dc9 | 177 | count += BITS_PER_LONG; |
41546c17 JW |
178 | } else { |
179 | unsigned long off = 0; | |
180 | ||
181 | while (vec && off < BITS_PER_LONG) { | |
182 | if (vec & 1) { | |
183 | page = pfn_to_page(start + off); | |
223e8dc9 | 184 | __free_pages_bootmem(page, 0); |
41546c17 | 185 | count++; |
223e8dc9 | 186 | } |
41546c17 JW |
187 | vec >>= 1; |
188 | off++; | |
223e8dc9 | 189 | } |
223e8dc9 | 190 | } |
41546c17 | 191 | start += BITS_PER_LONG; |
223e8dc9 JW |
192 | } |
193 | ||
223e8dc9 | 194 | page = virt_to_page(bdata->node_bootmem_map); |
df049a5f | 195 | pages = bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start); |
41546c17 JW |
196 | pages = bootmem_bootmap_pages(pages); |
197 | count += pages; | |
198 | while (pages--) | |
199 | __free_pages_bootmem(page++, 0); | |
223e8dc9 | 200 | |
2e5237da JW |
201 | bdebug("nid=%td released=%lx\n", bdata - bootmem_node_data, count); |
202 | ||
223e8dc9 JW |
203 | return count; |
204 | } | |
205 | ||
a66fd7da JW |
206 | /** |
207 | * free_all_bootmem_node - release a node's free pages to the buddy allocator | |
208 | * @pgdat: node to be released | |
209 | * | |
210 | * Returns the number of pages actually released. | |
211 | */ | |
223e8dc9 JW |
212 | unsigned long __init free_all_bootmem_node(pg_data_t *pgdat) |
213 | { | |
214 | register_page_bootmem_info_node(pgdat); | |
215 | return free_all_bootmem_core(pgdat->bdata); | |
216 | } | |
217 | ||
a66fd7da JW |
218 | /** |
219 | * free_all_bootmem - release free pages to the buddy allocator | |
220 | * | |
221 | * Returns the number of pages actually released. | |
222 | */ | |
223e8dc9 JW |
223 | unsigned long __init free_all_bootmem(void) |
224 | { | |
225 | return free_all_bootmem_core(NODE_DATA(0)->bdata); | |
226 | } | |
227 | ||
d747fa4b JW |
228 | static void __init __free(bootmem_data_t *bdata, |
229 | unsigned long sidx, unsigned long eidx) | |
230 | { | |
231 | unsigned long idx; | |
232 | ||
233 | bdebug("nid=%td start=%lx end=%lx\n", bdata - bootmem_node_data, | |
234 | sidx + PFN_DOWN(bdata->node_boot_start), | |
235 | eidx + PFN_DOWN(bdata->node_boot_start)); | |
236 | ||
237 | for (idx = sidx; idx < eidx; idx++) | |
238 | if (!test_and_clear_bit(idx, bdata->node_bootmem_map)) | |
239 | BUG(); | |
240 | } | |
241 | ||
242 | static int __init __reserve(bootmem_data_t *bdata, unsigned long sidx, | |
243 | unsigned long eidx, int flags) | |
244 | { | |
245 | unsigned long idx; | |
246 | int exclusive = flags & BOOTMEM_EXCLUSIVE; | |
247 | ||
248 | bdebug("nid=%td start=%lx end=%lx flags=%x\n", | |
249 | bdata - bootmem_node_data, | |
250 | sidx + PFN_DOWN(bdata->node_boot_start), | |
251 | eidx + PFN_DOWN(bdata->node_boot_start), | |
252 | flags); | |
253 | ||
254 | for (idx = sidx; idx < eidx; idx++) | |
255 | if (test_and_set_bit(idx, bdata->node_bootmem_map)) { | |
256 | if (exclusive) { | |
257 | __free(bdata, sidx, idx); | |
258 | return -EBUSY; | |
259 | } | |
260 | bdebug("silent double reserve of PFN %lx\n", | |
261 | idx + PFN_DOWN(bdata->node_boot_start)); | |
262 | } | |
263 | return 0; | |
264 | } | |
265 | ||
223e8dc9 JW |
266 | static void __init free_bootmem_core(bootmem_data_t *bdata, unsigned long addr, |
267 | unsigned long size) | |
268 | { | |
269 | unsigned long sidx, eidx; | |
270 | unsigned long i; | |
271 | ||
272 | BUG_ON(!size); | |
273 | ||
274 | /* out range */ | |
275 | if (addr + size < bdata->node_boot_start || | |
276 | PFN_DOWN(addr) > bdata->node_low_pfn) | |
277 | return; | |
278 | /* | |
279 | * round down end of usable mem, partially free pages are | |
280 | * considered reserved. | |
281 | */ | |
282 | ||
5f2809e6 JW |
283 | if (addr >= bdata->node_boot_start && |
284 | PFN_DOWN(addr - bdata->node_boot_start) < bdata->hint_idx) | |
285 | bdata->hint_idx = PFN_DOWN(addr - bdata->node_boot_start); | |
223e8dc9 JW |
286 | |
287 | /* | |
288 | * Round up to index to the range. | |
289 | */ | |
290 | if (PFN_UP(addr) > PFN_DOWN(bdata->node_boot_start)) | |
291 | sidx = PFN_UP(addr) - PFN_DOWN(bdata->node_boot_start); | |
292 | else | |
293 | sidx = 0; | |
294 | ||
295 | eidx = PFN_DOWN(addr + size - bdata->node_boot_start); | |
296 | if (eidx > bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start)) | |
297 | eidx = bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start); | |
298 | ||
d747fa4b | 299 | __free(bdata, sidx, eidx); |
223e8dc9 JW |
300 | } |
301 | ||
a66fd7da JW |
302 | /** |
303 | * free_bootmem_node - mark a page range as usable | |
304 | * @pgdat: node the range resides on | |
305 | * @physaddr: starting address of the range | |
306 | * @size: size of the range in bytes | |
307 | * | |
308 | * Partial pages will be considered reserved and left as they are. | |
309 | * | |
310 | * Only physical pages that actually reside on @pgdat are marked. | |
311 | */ | |
223e8dc9 JW |
312 | void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
313 | unsigned long size) | |
314 | { | |
315 | free_bootmem_core(pgdat->bdata, physaddr, size); | |
316 | } | |
317 | ||
a66fd7da JW |
318 | /** |
319 | * free_bootmem - mark a page range as usable | |
320 | * @addr: starting address of the range | |
321 | * @size: size of the range in bytes | |
322 | * | |
323 | * Partial pages will be considered reserved and left as they are. | |
324 | * | |
325 | * All physical pages within the range are marked, no matter what | |
326 | * node they reside on. | |
327 | */ | |
223e8dc9 JW |
328 | void __init free_bootmem(unsigned long addr, unsigned long size) |
329 | { | |
330 | bootmem_data_t *bdata; | |
331 | list_for_each_entry(bdata, &bdata_list, list) | |
332 | free_bootmem_core(bdata, addr, size); | |
333 | } | |
334 | ||
1da177e4 LT |
335 | /* |
336 | * Marks a particular physical memory range as unallocatable. Usable RAM | |
337 | * might be used for boot-time allocations - or it might get added | |
338 | * to the free page pool later on. | |
339 | */ | |
a5645a61 | 340 | static int __init can_reserve_bootmem_core(bootmem_data_t *bdata, |
72a7fe39 | 341 | unsigned long addr, unsigned long size, int flags) |
1da177e4 | 342 | { |
bbc7b92e | 343 | unsigned long sidx, eidx; |
1da177e4 | 344 | unsigned long i; |
a5645a61 YL |
345 | |
346 | BUG_ON(!size); | |
347 | ||
348 | /* out of range, don't hold other */ | |
349 | if (addr + size < bdata->node_boot_start || | |
350 | PFN_DOWN(addr) > bdata->node_low_pfn) | |
351 | return 0; | |
bbc7b92e | 352 | |
1da177e4 | 353 | /* |
a5645a61 | 354 | * Round up to index to the range. |
1da177e4 | 355 | */ |
a5645a61 YL |
356 | if (addr > bdata->node_boot_start) |
357 | sidx= PFN_DOWN(addr - bdata->node_boot_start); | |
358 | else | |
359 | sidx = 0; | |
360 | ||
361 | eidx = PFN_UP(addr + size - bdata->node_boot_start); | |
362 | if (eidx > bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start)) | |
363 | eidx = bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start); | |
364 | ||
365 | for (i = sidx; i < eidx; i++) { | |
366 | if (test_bit(i, bdata->node_bootmem_map)) { | |
367 | if (flags & BOOTMEM_EXCLUSIVE) | |
368 | return -EBUSY; | |
369 | } | |
370 | } | |
371 | ||
372 | return 0; | |
373 | ||
374 | } | |
375 | ||
376 | static void __init reserve_bootmem_core(bootmem_data_t *bdata, | |
377 | unsigned long addr, unsigned long size, int flags) | |
378 | { | |
379 | unsigned long sidx, eidx; | |
380 | unsigned long i; | |
381 | ||
1da177e4 | 382 | BUG_ON(!size); |
bbc7b92e | 383 | |
a5645a61 YL |
384 | /* out of range */ |
385 | if (addr + size < bdata->node_boot_start || | |
386 | PFN_DOWN(addr) > bdata->node_low_pfn) | |
387 | return; | |
388 | ||
389 | /* | |
390 | * Round up to index to the range. | |
391 | */ | |
392 | if (addr > bdata->node_boot_start) | |
393 | sidx= PFN_DOWN(addr - bdata->node_boot_start); | |
394 | else | |
395 | sidx = 0; | |
396 | ||
bbc7b92e | 397 | eidx = PFN_UP(addr + size - bdata->node_boot_start); |
a5645a61 YL |
398 | if (eidx > bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start)) |
399 | eidx = bdata->node_low_pfn - PFN_DOWN(bdata->node_boot_start); | |
1da177e4 | 400 | |
d747fa4b | 401 | return __reserve(bdata, sidx, eidx, flags); |
1da177e4 LT |
402 | } |
403 | ||
a66fd7da JW |
404 | /** |
405 | * reserve_bootmem_node - mark a page range as reserved | |
406 | * @pgdat: node the range resides on | |
407 | * @physaddr: starting address of the range | |
408 | * @size: size of the range in bytes | |
409 | * @flags: reservation flags (see linux/bootmem.h) | |
410 | * | |
411 | * Partial pages will be reserved. | |
412 | * | |
413 | * Only physical pages that actually reside on @pgdat are marked. | |
414 | */ | |
223e8dc9 JW |
415 | int __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
416 | unsigned long size, int flags) | |
1da177e4 | 417 | { |
223e8dc9 | 418 | int ret; |
1da177e4 | 419 | |
223e8dc9 JW |
420 | ret = can_reserve_bootmem_core(pgdat->bdata, physaddr, size, flags); |
421 | if (ret < 0) | |
422 | return -ENOMEM; | |
423 | reserve_bootmem_core(pgdat->bdata, physaddr, size, flags); | |
424 | return 0; | |
425 | } | |
5a982cbc | 426 | |
223e8dc9 | 427 | #ifndef CONFIG_HAVE_ARCH_BOOTMEM_NODE |
a66fd7da JW |
428 | /** |
429 | * reserve_bootmem - mark a page range as usable | |
430 | * @addr: starting address of the range | |
431 | * @size: size of the range in bytes | |
432 | * @flags: reservation flags (see linux/bootmem.h) | |
433 | * | |
434 | * Partial pages will be reserved. | |
435 | * | |
436 | * All physical pages within the range are marked, no matter what | |
437 | * node they reside on. | |
438 | */ | |
223e8dc9 JW |
439 | int __init reserve_bootmem(unsigned long addr, unsigned long size, |
440 | int flags) | |
441 | { | |
442 | bootmem_data_t *bdata; | |
443 | int ret; | |
1da177e4 | 444 | |
223e8dc9 JW |
445 | list_for_each_entry(bdata, &bdata_list, list) { |
446 | ret = can_reserve_bootmem_core(bdata, addr, size, flags); | |
447 | if (ret < 0) | |
448 | return ret; | |
1da177e4 | 449 | } |
223e8dc9 JW |
450 | list_for_each_entry(bdata, &bdata_list, list) |
451 | reserve_bootmem_core(bdata, addr, size, flags); | |
452 | ||
453 | return 0; | |
1da177e4 | 454 | } |
223e8dc9 | 455 | #endif /* !CONFIG_HAVE_ARCH_BOOTMEM_NODE */ |
1da177e4 | 456 | |
5f2809e6 JW |
457 | static void * __init alloc_bootmem_core(struct bootmem_data *bdata, |
458 | unsigned long size, unsigned long align, | |
459 | unsigned long goal, unsigned long limit) | |
1da177e4 | 460 | { |
5f2809e6 JW |
461 | unsigned long min, max, start, sidx, midx, step; |
462 | ||
463 | BUG_ON(!size); | |
464 | BUG_ON(align & (align - 1)); | |
465 | BUG_ON(limit && goal + size > limit); | |
1da177e4 | 466 | |
7c309a64 CK |
467 | if (!bdata->node_bootmem_map) |
468 | return NULL; | |
469 | ||
2e5237da JW |
470 | bdebug("nid=%td size=%lx [%lu pages] align=%lx goal=%lx limit=%lx\n", |
471 | bdata - bootmem_node_data, size, PAGE_ALIGN(size) >> PAGE_SHIFT, | |
472 | align, goal, limit); | |
473 | ||
5f2809e6 JW |
474 | min = PFN_DOWN(bdata->node_boot_start); |
475 | max = bdata->node_low_pfn; | |
9a2dc04c | 476 | |
5f2809e6 JW |
477 | goal >>= PAGE_SHIFT; |
478 | limit >>= PAGE_SHIFT; | |
479 | ||
480 | if (limit && max > limit) | |
481 | max = limit; | |
482 | if (max <= min) | |
9a2dc04c YL |
483 | return NULL; |
484 | ||
5f2809e6 | 485 | step = max(align >> PAGE_SHIFT, 1UL); |
281dd25c | 486 | |
5f2809e6 JW |
487 | if (goal && min < goal && goal < max) |
488 | start = ALIGN(goal, step); | |
489 | else | |
490 | start = ALIGN(min, step); | |
1da177e4 | 491 | |
5f2809e6 JW |
492 | sidx = start - PFN_DOWN(bdata->node_boot_start); |
493 | midx = max - PFN_DOWN(bdata->node_boot_start); | |
1da177e4 | 494 | |
5f2809e6 JW |
495 | if (bdata->hint_idx > sidx) { |
496 | /* Make sure we retry on failure */ | |
497 | goal = 1; | |
498 | sidx = ALIGN(bdata->hint_idx, step); | |
499 | } | |
1da177e4 | 500 | |
5f2809e6 JW |
501 | while (1) { |
502 | int merge; | |
503 | void *region; | |
504 | unsigned long eidx, i, start_off, end_off; | |
505 | find_block: | |
506 | sidx = find_next_zero_bit(bdata->node_bootmem_map, midx, sidx); | |
507 | sidx = ALIGN(sidx, step); | |
508 | eidx = sidx + PFN_UP(size); | |
ad09315c | 509 | |
5f2809e6 | 510 | if (sidx >= midx || eidx > midx) |
66d43e98 | 511 | break; |
1da177e4 | 512 | |
5f2809e6 JW |
513 | for (i = sidx; i < eidx; i++) |
514 | if (test_bit(i, bdata->node_bootmem_map)) { | |
515 | sidx = ALIGN(i, step); | |
516 | if (sidx == i) | |
517 | sidx += step; | |
518 | goto find_block; | |
519 | } | |
1da177e4 | 520 | |
5f2809e6 JW |
521 | if (bdata->last_end_off && |
522 | PFN_DOWN(bdata->last_end_off) + 1 == sidx) | |
523 | start_off = ALIGN(bdata->last_end_off, align); | |
524 | else | |
525 | start_off = PFN_PHYS(sidx); | |
526 | ||
527 | merge = PFN_DOWN(start_off) < sidx; | |
528 | end_off = start_off + size; | |
529 | ||
530 | bdata->last_end_off = end_off; | |
531 | bdata->hint_idx = PFN_UP(end_off); | |
532 | ||
533 | /* | |
534 | * Reserve the area now: | |
535 | */ | |
d747fa4b JW |
536 | if (__reserve(bdata, PFN_DOWN(start_off) + merge, |
537 | PFN_UP(end_off), BOOTMEM_EXCLUSIVE)) | |
538 | BUG(); | |
5f2809e6 JW |
539 | |
540 | region = phys_to_virt(bdata->node_boot_start + start_off); | |
541 | memset(region, 0, size); | |
542 | return region; | |
1da177e4 LT |
543 | } |
544 | ||
5f2809e6 JW |
545 | if (goal) { |
546 | goal = 0; | |
547 | sidx = 0; | |
548 | goto find_block; | |
549 | } | |
2e5237da | 550 | |
5f2809e6 | 551 | return NULL; |
1da177e4 LT |
552 | } |
553 | ||
a66fd7da JW |
554 | /** |
555 | * __alloc_bootmem_nopanic - allocate boot memory without panicking | |
556 | * @size: size of the request in bytes | |
557 | * @align: alignment of the region | |
558 | * @goal: preferred starting address of the region | |
559 | * | |
560 | * The goal is dropped if it can not be satisfied and the allocation will | |
561 | * fall back to memory below @goal. | |
562 | * | |
563 | * Allocation may happen on any node in the system. | |
564 | * | |
565 | * Returns NULL on failure. | |
566 | */ | |
bb0923a6 FBH |
567 | void * __init __alloc_bootmem_nopanic(unsigned long size, unsigned long align, |
568 | unsigned long goal) | |
1da177e4 | 569 | { |
679bc9fb | 570 | bootmem_data_t *bdata; |
1da177e4 LT |
571 | void *ptr; |
572 | ||
f71bf0ca | 573 | list_for_each_entry(bdata, &bdata_list, list) { |
ffc6421f | 574 | ptr = alloc_bootmem_core(bdata, size, align, goal, 0); |
f71bf0ca FBH |
575 | if (ptr) |
576 | return ptr; | |
577 | } | |
a8062231 AK |
578 | return NULL; |
579 | } | |
1da177e4 | 580 | |
a66fd7da JW |
581 | /** |
582 | * __alloc_bootmem - allocate boot memory | |
583 | * @size: size of the request in bytes | |
584 | * @align: alignment of the region | |
585 | * @goal: preferred starting address of the region | |
586 | * | |
587 | * The goal is dropped if it can not be satisfied and the allocation will | |
588 | * fall back to memory below @goal. | |
589 | * | |
590 | * Allocation may happen on any node in the system. | |
591 | * | |
592 | * The function panics if the request can not be satisfied. | |
593 | */ | |
bb0923a6 FBH |
594 | void * __init __alloc_bootmem(unsigned long size, unsigned long align, |
595 | unsigned long goal) | |
a8062231 AK |
596 | { |
597 | void *mem = __alloc_bootmem_nopanic(size,align,goal); | |
f71bf0ca | 598 | |
a8062231 AK |
599 | if (mem) |
600 | return mem; | |
1da177e4 LT |
601 | /* |
602 | * Whoops, we cannot satisfy the allocation request. | |
603 | */ | |
604 | printk(KERN_ALERT "bootmem alloc of %lu bytes failed!\n", size); | |
605 | panic("Out of memory"); | |
606 | return NULL; | |
607 | } | |
608 | ||
a66fd7da JW |
609 | /** |
610 | * __alloc_bootmem_node - allocate boot memory from a specific node | |
611 | * @pgdat: node to allocate from | |
612 | * @size: size of the request in bytes | |
613 | * @align: alignment of the region | |
614 | * @goal: preferred starting address of the region | |
615 | * | |
616 | * The goal is dropped if it can not be satisfied and the allocation will | |
617 | * fall back to memory below @goal. | |
618 | * | |
619 | * Allocation may fall back to any node in the system if the specified node | |
620 | * can not hold the requested memory. | |
621 | * | |
622 | * The function panics if the request can not be satisfied. | |
623 | */ | |
bb0923a6 FBH |
624 | void * __init __alloc_bootmem_node(pg_data_t *pgdat, unsigned long size, |
625 | unsigned long align, unsigned long goal) | |
1da177e4 LT |
626 | { |
627 | void *ptr; | |
628 | ||
ffc6421f | 629 | ptr = alloc_bootmem_core(pgdat->bdata, size, align, goal, 0); |
1da177e4 | 630 | if (ptr) |
f71bf0ca | 631 | return ptr; |
1da177e4 | 632 | |
008857c1 | 633 | return __alloc_bootmem(size, align, goal); |
1da177e4 LT |
634 | } |
635 | ||
e70260aa | 636 | #ifdef CONFIG_SPARSEMEM |
a66fd7da JW |
637 | /** |
638 | * alloc_bootmem_section - allocate boot memory from a specific section | |
639 | * @size: size of the request in bytes | |
640 | * @section_nr: sparse map section to allocate from | |
641 | * | |
642 | * Return NULL on failure. | |
643 | */ | |
e70260aa YG |
644 | void * __init alloc_bootmem_section(unsigned long size, |
645 | unsigned long section_nr) | |
646 | { | |
647 | void *ptr; | |
648 | unsigned long limit, goal, start_nr, end_nr, pfn; | |
649 | struct pglist_data *pgdat; | |
650 | ||
651 | pfn = section_nr_to_pfn(section_nr); | |
652 | goal = PFN_PHYS(pfn); | |
653 | limit = PFN_PHYS(section_nr_to_pfn(section_nr + 1)) - 1; | |
654 | pgdat = NODE_DATA(early_pfn_to_nid(pfn)); | |
ffc6421f JW |
655 | ptr = alloc_bootmem_core(pgdat->bdata, size, SMP_CACHE_BYTES, goal, |
656 | limit); | |
e70260aa YG |
657 | |
658 | if (!ptr) | |
659 | return NULL; | |
660 | ||
661 | start_nr = pfn_to_section_nr(PFN_DOWN(__pa(ptr))); | |
662 | end_nr = pfn_to_section_nr(PFN_DOWN(__pa(ptr) + size)); | |
663 | if (start_nr != section_nr || end_nr != section_nr) { | |
664 | printk(KERN_WARNING "alloc_bootmem failed on section %ld.\n", | |
665 | section_nr); | |
666 | free_bootmem_core(pgdat->bdata, __pa(ptr), size); | |
667 | ptr = NULL; | |
668 | } | |
669 | ||
670 | return ptr; | |
671 | } | |
672 | #endif | |
673 | ||
b54bbf7b AK |
674 | void * __init __alloc_bootmem_node_nopanic(pg_data_t *pgdat, unsigned long size, |
675 | unsigned long align, unsigned long goal) | |
676 | { | |
677 | void *ptr; | |
678 | ||
679 | ptr = alloc_bootmem_core(pgdat->bdata, size, align, goal, 0); | |
680 | if (ptr) | |
681 | return ptr; | |
682 | ||
683 | return __alloc_bootmem_nopanic(size, align, goal); | |
684 | } | |
685 | ||
dfd54cbc HC |
686 | #ifndef ARCH_LOW_ADDRESS_LIMIT |
687 | #define ARCH_LOW_ADDRESS_LIMIT 0xffffffffUL | |
688 | #endif | |
008857c1 | 689 | |
a66fd7da JW |
690 | /** |
691 | * __alloc_bootmem_low - allocate low boot memory | |
692 | * @size: size of the request in bytes | |
693 | * @align: alignment of the region | |
694 | * @goal: preferred starting address of the region | |
695 | * | |
696 | * The goal is dropped if it can not be satisfied and the allocation will | |
697 | * fall back to memory below @goal. | |
698 | * | |
699 | * Allocation may happen on any node in the system. | |
700 | * | |
701 | * The function panics if the request can not be satisfied. | |
702 | */ | |
bb0923a6 FBH |
703 | void * __init __alloc_bootmem_low(unsigned long size, unsigned long align, |
704 | unsigned long goal) | |
008857c1 | 705 | { |
679bc9fb | 706 | bootmem_data_t *bdata; |
008857c1 RT |
707 | void *ptr; |
708 | ||
f71bf0ca | 709 | list_for_each_entry(bdata, &bdata_list, list) { |
ffc6421f JW |
710 | ptr = alloc_bootmem_core(bdata, size, align, goal, |
711 | ARCH_LOW_ADDRESS_LIMIT); | |
f71bf0ca FBH |
712 | if (ptr) |
713 | return ptr; | |
714 | } | |
008857c1 RT |
715 | |
716 | /* | |
717 | * Whoops, we cannot satisfy the allocation request. | |
718 | */ | |
719 | printk(KERN_ALERT "low bootmem alloc of %lu bytes failed!\n", size); | |
720 | panic("Out of low memory"); | |
721 | return NULL; | |
722 | } | |
723 | ||
a66fd7da JW |
724 | /** |
725 | * __alloc_bootmem_low_node - allocate low boot memory from a specific node | |
726 | * @pgdat: node to allocate from | |
727 | * @size: size of the request in bytes | |
728 | * @align: alignment of the region | |
729 | * @goal: preferred starting address of the region | |
730 | * | |
731 | * The goal is dropped if it can not be satisfied and the allocation will | |
732 | * fall back to memory below @goal. | |
733 | * | |
734 | * Allocation may fall back to any node in the system if the specified node | |
735 | * can not hold the requested memory. | |
736 | * | |
737 | * The function panics if the request can not be satisfied. | |
738 | */ | |
008857c1 RT |
739 | void * __init __alloc_bootmem_low_node(pg_data_t *pgdat, unsigned long size, |
740 | unsigned long align, unsigned long goal) | |
741 | { | |
ffc6421f JW |
742 | return alloc_bootmem_core(pgdat->bdata, size, align, goal, |
743 | ARCH_LOW_ADDRESS_LIMIT); | |
008857c1 | 744 | } |