Commit | Line | Data |
---|---|---|
3947be19 DH |
1 | /* |
2 | * linux/mm/memory_hotplug.c | |
3 | * | |
4 | * Copyright (C) | |
5 | */ | |
6 | ||
3947be19 DH |
7 | #include <linux/stddef.h> |
8 | #include <linux/mm.h> | |
9 | #include <linux/swap.h> | |
10 | #include <linux/interrupt.h> | |
11 | #include <linux/pagemap.h> | |
12 | #include <linux/bootmem.h> | |
13 | #include <linux/compiler.h> | |
b95f1b31 | 14 | #include <linux/export.h> |
3947be19 | 15 | #include <linux/pagevec.h> |
2d1d43f6 | 16 | #include <linux/writeback.h> |
3947be19 DH |
17 | #include <linux/slab.h> |
18 | #include <linux/sysctl.h> | |
19 | #include <linux/cpu.h> | |
20 | #include <linux/memory.h> | |
21 | #include <linux/memory_hotplug.h> | |
22 | #include <linux/highmem.h> | |
23 | #include <linux/vmalloc.h> | |
0a547039 | 24 | #include <linux/ioport.h> |
0c0e6195 KH |
25 | #include <linux/delay.h> |
26 | #include <linux/migrate.h> | |
27 | #include <linux/page-isolation.h> | |
71088785 | 28 | #include <linux/pfn.h> |
6ad696d2 | 29 | #include <linux/suspend.h> |
6d9c285a | 30 | #include <linux/mm_inline.h> |
d96ae530 | 31 | #include <linux/firmware-map.h> |
3947be19 DH |
32 | |
33 | #include <asm/tlbflush.h> | |
34 | ||
1e5ad9a3 AB |
35 | #include "internal.h" |
36 | ||
9d0ad8ca DK |
37 | /* |
38 | * online_page_callback contains pointer to current page onlining function. | |
39 | * Initially it is generic_online_page(). If it is required it could be | |
40 | * changed by calling set_online_page_callback() for callback registration | |
41 | * and restore_online_page_callback() for generic callback restore. | |
42 | */ | |
43 | ||
44 | static void generic_online_page(struct page *page); | |
45 | ||
46 | static online_page_callback_t online_page_callback = generic_online_page; | |
47 | ||
20d6c96b KM |
48 | DEFINE_MUTEX(mem_hotplug_mutex); |
49 | ||
50 | void lock_memory_hotplug(void) | |
51 | { | |
52 | mutex_lock(&mem_hotplug_mutex); | |
53 | ||
54 | /* for exclusive hibernation if CONFIG_HIBERNATION=y */ | |
55 | lock_system_sleep(); | |
56 | } | |
57 | ||
58 | void unlock_memory_hotplug(void) | |
59 | { | |
60 | unlock_system_sleep(); | |
61 | mutex_unlock(&mem_hotplug_mutex); | |
62 | } | |
63 | ||
64 | ||
45e0b78b KM |
65 | /* add this memory to iomem resource */ |
66 | static struct resource *register_memory_resource(u64 start, u64 size) | |
67 | { | |
68 | struct resource *res; | |
69 | res = kzalloc(sizeof(struct resource), GFP_KERNEL); | |
70 | BUG_ON(!res); | |
71 | ||
72 | res->name = "System RAM"; | |
73 | res->start = start; | |
74 | res->end = start + size - 1; | |
887c3cb1 | 75 | res->flags = IORESOURCE_MEM | IORESOURCE_BUSY; |
45e0b78b | 76 | if (request_resource(&iomem_resource, res) < 0) { |
a62e2f4f | 77 | printk("System RAM resource %pR cannot be added\n", res); |
45e0b78b KM |
78 | kfree(res); |
79 | res = NULL; | |
80 | } | |
81 | return res; | |
82 | } | |
83 | ||
84 | static void release_memory_resource(struct resource *res) | |
85 | { | |
86 | if (!res) | |
87 | return; | |
88 | release_resource(res); | |
89 | kfree(res); | |
90 | return; | |
91 | } | |
92 | ||
53947027 | 93 | #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE |
46723bfa YI |
94 | void get_page_bootmem(unsigned long info, struct page *page, |
95 | unsigned long type) | |
04753278 | 96 | { |
5f24ce5f | 97 | page->lru.next = (struct list_head *) type; |
04753278 YG |
98 | SetPagePrivate(page); |
99 | set_page_private(page, info); | |
100 | atomic_inc(&page->_count); | |
101 | } | |
102 | ||
23ce932a RM |
103 | /* reference to __meminit __free_pages_bootmem is valid |
104 | * so use __ref to tell modpost not to generate a warning */ | |
105 | void __ref put_page_bootmem(struct page *page) | |
04753278 | 106 | { |
5f24ce5f | 107 | unsigned long type; |
9feedc9d | 108 | static DEFINE_MUTEX(ppb_lock); |
04753278 | 109 | |
5f24ce5f AA |
110 | type = (unsigned long) page->lru.next; |
111 | BUG_ON(type < MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE || | |
112 | type > MEMORY_HOTPLUG_MAX_BOOTMEM_TYPE); | |
04753278 YG |
113 | |
114 | if (atomic_dec_return(&page->_count) == 1) { | |
115 | ClearPagePrivate(page); | |
116 | set_page_private(page, 0); | |
5f24ce5f | 117 | INIT_LIST_HEAD(&page->lru); |
9feedc9d JL |
118 | |
119 | /* | |
120 | * Please refer to comment for __free_pages_bootmem() | |
121 | * for why we serialize here. | |
122 | */ | |
123 | mutex_lock(&ppb_lock); | |
04753278 | 124 | __free_pages_bootmem(page, 0); |
9feedc9d | 125 | mutex_unlock(&ppb_lock); |
04753278 YG |
126 | } |
127 | ||
128 | } | |
129 | ||
46723bfa YI |
130 | #ifdef CONFIG_HAVE_BOOTMEM_INFO_NODE |
131 | #ifndef CONFIG_SPARSEMEM_VMEMMAP | |
d92bc318 | 132 | static void register_page_bootmem_info_section(unsigned long start_pfn) |
04753278 YG |
133 | { |
134 | unsigned long *usemap, mapsize, section_nr, i; | |
135 | struct mem_section *ms; | |
136 | struct page *page, *memmap; | |
137 | ||
04753278 YG |
138 | section_nr = pfn_to_section_nr(start_pfn); |
139 | ms = __nr_to_section(section_nr); | |
140 | ||
141 | /* Get section's memmap address */ | |
142 | memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr); | |
143 | ||
144 | /* | |
145 | * Get page for the memmap's phys address | |
146 | * XXX: need more consideration for sparse_vmemmap... | |
147 | */ | |
148 | page = virt_to_page(memmap); | |
149 | mapsize = sizeof(struct page) * PAGES_PER_SECTION; | |
150 | mapsize = PAGE_ALIGN(mapsize) >> PAGE_SHIFT; | |
151 | ||
152 | /* remember memmap's page */ | |
153 | for (i = 0; i < mapsize; i++, page++) | |
154 | get_page_bootmem(section_nr, page, SECTION_INFO); | |
155 | ||
156 | usemap = __nr_to_section(section_nr)->pageblock_flags; | |
157 | page = virt_to_page(usemap); | |
158 | ||
159 | mapsize = PAGE_ALIGN(usemap_size()) >> PAGE_SHIFT; | |
160 | ||
161 | for (i = 0; i < mapsize; i++, page++) | |
af370fb8 | 162 | get_page_bootmem(section_nr, page, MIX_SECTION_INFO); |
04753278 YG |
163 | |
164 | } | |
46723bfa YI |
165 | #else /* CONFIG_SPARSEMEM_VMEMMAP */ |
166 | static void register_page_bootmem_info_section(unsigned long start_pfn) | |
167 | { | |
168 | unsigned long *usemap, mapsize, section_nr, i; | |
169 | struct mem_section *ms; | |
170 | struct page *page, *memmap; | |
171 | ||
172 | if (!pfn_valid(start_pfn)) | |
173 | return; | |
174 | ||
175 | section_nr = pfn_to_section_nr(start_pfn); | |
176 | ms = __nr_to_section(section_nr); | |
177 | ||
178 | memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr); | |
179 | ||
180 | register_page_bootmem_memmap(section_nr, memmap, PAGES_PER_SECTION); | |
181 | ||
182 | usemap = __nr_to_section(section_nr)->pageblock_flags; | |
183 | page = virt_to_page(usemap); | |
184 | ||
185 | mapsize = PAGE_ALIGN(usemap_size()) >> PAGE_SHIFT; | |
186 | ||
187 | for (i = 0; i < mapsize; i++, page++) | |
188 | get_page_bootmem(section_nr, page, MIX_SECTION_INFO); | |
189 | } | |
190 | #endif /* !CONFIG_SPARSEMEM_VMEMMAP */ | |
04753278 YG |
191 | |
192 | void register_page_bootmem_info_node(struct pglist_data *pgdat) | |
193 | { | |
194 | unsigned long i, pfn, end_pfn, nr_pages; | |
195 | int node = pgdat->node_id; | |
196 | struct page *page; | |
197 | struct zone *zone; | |
198 | ||
199 | nr_pages = PAGE_ALIGN(sizeof(struct pglist_data)) >> PAGE_SHIFT; | |
200 | page = virt_to_page(pgdat); | |
201 | ||
202 | for (i = 0; i < nr_pages; i++, page++) | |
203 | get_page_bootmem(node, page, NODE_INFO); | |
204 | ||
205 | zone = &pgdat->node_zones[0]; | |
206 | for (; zone < pgdat->node_zones + MAX_NR_ZONES - 1; zone++) { | |
207 | if (zone->wait_table) { | |
208 | nr_pages = zone->wait_table_hash_nr_entries | |
209 | * sizeof(wait_queue_head_t); | |
210 | nr_pages = PAGE_ALIGN(nr_pages) >> PAGE_SHIFT; | |
211 | page = virt_to_page(zone->wait_table); | |
212 | ||
213 | for (i = 0; i < nr_pages; i++, page++) | |
214 | get_page_bootmem(node, page, NODE_INFO); | |
215 | } | |
216 | } | |
217 | ||
218 | pfn = pgdat->node_start_pfn; | |
219 | end_pfn = pfn + pgdat->node_spanned_pages; | |
220 | ||
221 | /* register_section info */ | |
f14851af | 222 | for (; pfn < end_pfn; pfn += PAGES_PER_SECTION) { |
223 | /* | |
224 | * Some platforms can assign the same pfn to multiple nodes - on | |
225 | * node0 as well as nodeN. To avoid registering a pfn against | |
226 | * multiple nodes we check that this pfn does not already | |
227 | * reside in some other node. | |
228 | */ | |
229 | if (pfn_valid(pfn) && (pfn_to_nid(pfn) == node)) | |
230 | register_page_bootmem_info_section(pfn); | |
231 | } | |
04753278 | 232 | } |
46723bfa | 233 | #endif /* CONFIG_HAVE_BOOTMEM_INFO_NODE */ |
04753278 | 234 | |
76cdd58e HC |
235 | static void grow_zone_span(struct zone *zone, unsigned long start_pfn, |
236 | unsigned long end_pfn) | |
237 | { | |
238 | unsigned long old_zone_end_pfn; | |
239 | ||
240 | zone_span_writelock(zone); | |
241 | ||
242 | old_zone_end_pfn = zone->zone_start_pfn + zone->spanned_pages; | |
712cd386 | 243 | if (!zone->spanned_pages || start_pfn < zone->zone_start_pfn) |
76cdd58e HC |
244 | zone->zone_start_pfn = start_pfn; |
245 | ||
246 | zone->spanned_pages = max(old_zone_end_pfn, end_pfn) - | |
247 | zone->zone_start_pfn; | |
248 | ||
249 | zone_span_writeunlock(zone); | |
250 | } | |
251 | ||
511c2aba LJ |
252 | static void resize_zone(struct zone *zone, unsigned long start_pfn, |
253 | unsigned long end_pfn) | |
254 | { | |
255 | zone_span_writelock(zone); | |
256 | ||
e455a9b9 LJ |
257 | if (end_pfn - start_pfn) { |
258 | zone->zone_start_pfn = start_pfn; | |
259 | zone->spanned_pages = end_pfn - start_pfn; | |
260 | } else { | |
261 | /* | |
262 | * make it consist as free_area_init_core(), | |
263 | * if spanned_pages = 0, then keep start_pfn = 0 | |
264 | */ | |
265 | zone->zone_start_pfn = 0; | |
266 | zone->spanned_pages = 0; | |
267 | } | |
511c2aba LJ |
268 | |
269 | zone_span_writeunlock(zone); | |
270 | } | |
271 | ||
272 | static void fix_zone_id(struct zone *zone, unsigned long start_pfn, | |
273 | unsigned long end_pfn) | |
274 | { | |
275 | enum zone_type zid = zone_idx(zone); | |
276 | int nid = zone->zone_pgdat->node_id; | |
277 | unsigned long pfn; | |
278 | ||
279 | for (pfn = start_pfn; pfn < end_pfn; pfn++) | |
280 | set_page_links(pfn_to_page(pfn), zid, nid, pfn); | |
281 | } | |
282 | ||
e455a9b9 | 283 | static int __meminit move_pfn_range_left(struct zone *z1, struct zone *z2, |
511c2aba LJ |
284 | unsigned long start_pfn, unsigned long end_pfn) |
285 | { | |
e455a9b9 | 286 | int ret; |
511c2aba | 287 | unsigned long flags; |
e455a9b9 LJ |
288 | unsigned long z1_start_pfn; |
289 | ||
290 | if (!z1->wait_table) { | |
291 | ret = init_currently_empty_zone(z1, start_pfn, | |
292 | end_pfn - start_pfn, MEMMAP_HOTPLUG); | |
293 | if (ret) | |
294 | return ret; | |
295 | } | |
511c2aba LJ |
296 | |
297 | pgdat_resize_lock(z1->zone_pgdat, &flags); | |
298 | ||
299 | /* can't move pfns which are higher than @z2 */ | |
300 | if (end_pfn > z2->zone_start_pfn + z2->spanned_pages) | |
301 | goto out_fail; | |
302 | /* the move out part mast at the left most of @z2 */ | |
303 | if (start_pfn > z2->zone_start_pfn) | |
304 | goto out_fail; | |
305 | /* must included/overlap */ | |
306 | if (end_pfn <= z2->zone_start_pfn) | |
307 | goto out_fail; | |
308 | ||
e455a9b9 LJ |
309 | /* use start_pfn for z1's start_pfn if z1 is empty */ |
310 | if (z1->spanned_pages) | |
311 | z1_start_pfn = z1->zone_start_pfn; | |
312 | else | |
313 | z1_start_pfn = start_pfn; | |
314 | ||
315 | resize_zone(z1, z1_start_pfn, end_pfn); | |
511c2aba LJ |
316 | resize_zone(z2, end_pfn, z2->zone_start_pfn + z2->spanned_pages); |
317 | ||
318 | pgdat_resize_unlock(z1->zone_pgdat, &flags); | |
319 | ||
320 | fix_zone_id(z1, start_pfn, end_pfn); | |
321 | ||
322 | return 0; | |
323 | out_fail: | |
324 | pgdat_resize_unlock(z1->zone_pgdat, &flags); | |
325 | return -1; | |
326 | } | |
327 | ||
e455a9b9 | 328 | static int __meminit move_pfn_range_right(struct zone *z1, struct zone *z2, |
511c2aba LJ |
329 | unsigned long start_pfn, unsigned long end_pfn) |
330 | { | |
e455a9b9 | 331 | int ret; |
511c2aba | 332 | unsigned long flags; |
e455a9b9 LJ |
333 | unsigned long z2_end_pfn; |
334 | ||
335 | if (!z2->wait_table) { | |
336 | ret = init_currently_empty_zone(z2, start_pfn, | |
337 | end_pfn - start_pfn, MEMMAP_HOTPLUG); | |
338 | if (ret) | |
339 | return ret; | |
340 | } | |
511c2aba LJ |
341 | |
342 | pgdat_resize_lock(z1->zone_pgdat, &flags); | |
343 | ||
344 | /* can't move pfns which are lower than @z1 */ | |
345 | if (z1->zone_start_pfn > start_pfn) | |
346 | goto out_fail; | |
347 | /* the move out part mast at the right most of @z1 */ | |
348 | if (z1->zone_start_pfn + z1->spanned_pages > end_pfn) | |
349 | goto out_fail; | |
350 | /* must included/overlap */ | |
351 | if (start_pfn >= z1->zone_start_pfn + z1->spanned_pages) | |
352 | goto out_fail; | |
353 | ||
e455a9b9 LJ |
354 | /* use end_pfn for z2's end_pfn if z2 is empty */ |
355 | if (z2->spanned_pages) | |
356 | z2_end_pfn = z2->zone_start_pfn + z2->spanned_pages; | |
357 | else | |
358 | z2_end_pfn = end_pfn; | |
359 | ||
511c2aba | 360 | resize_zone(z1, z1->zone_start_pfn, start_pfn); |
e455a9b9 | 361 | resize_zone(z2, start_pfn, z2_end_pfn); |
511c2aba LJ |
362 | |
363 | pgdat_resize_unlock(z1->zone_pgdat, &flags); | |
364 | ||
365 | fix_zone_id(z2, start_pfn, end_pfn); | |
366 | ||
367 | return 0; | |
368 | out_fail: | |
369 | pgdat_resize_unlock(z1->zone_pgdat, &flags); | |
370 | return -1; | |
371 | } | |
372 | ||
76cdd58e HC |
373 | static void grow_pgdat_span(struct pglist_data *pgdat, unsigned long start_pfn, |
374 | unsigned long end_pfn) | |
375 | { | |
376 | unsigned long old_pgdat_end_pfn = | |
377 | pgdat->node_start_pfn + pgdat->node_spanned_pages; | |
378 | ||
712cd386 | 379 | if (!pgdat->node_spanned_pages || start_pfn < pgdat->node_start_pfn) |
76cdd58e HC |
380 | pgdat->node_start_pfn = start_pfn; |
381 | ||
382 | pgdat->node_spanned_pages = max(old_pgdat_end_pfn, end_pfn) - | |
383 | pgdat->node_start_pfn; | |
384 | } | |
385 | ||
31168481 | 386 | static int __meminit __add_zone(struct zone *zone, unsigned long phys_start_pfn) |
3947be19 DH |
387 | { |
388 | struct pglist_data *pgdat = zone->zone_pgdat; | |
389 | int nr_pages = PAGES_PER_SECTION; | |
390 | int nid = pgdat->node_id; | |
391 | int zone_type; | |
76cdd58e | 392 | unsigned long flags; |
3947be19 DH |
393 | |
394 | zone_type = zone - pgdat->node_zones; | |
76cdd58e HC |
395 | if (!zone->wait_table) { |
396 | int ret; | |
397 | ||
398 | ret = init_currently_empty_zone(zone, phys_start_pfn, | |
399 | nr_pages, MEMMAP_HOTPLUG); | |
400 | if (ret) | |
401 | return ret; | |
402 | } | |
403 | pgdat_resize_lock(zone->zone_pgdat, &flags); | |
404 | grow_zone_span(zone, phys_start_pfn, phys_start_pfn + nr_pages); | |
405 | grow_pgdat_span(zone->zone_pgdat, phys_start_pfn, | |
406 | phys_start_pfn + nr_pages); | |
407 | pgdat_resize_unlock(zone->zone_pgdat, &flags); | |
a2f3aa02 DH |
408 | memmap_init_zone(nr_pages, nid, zone_type, |
409 | phys_start_pfn, MEMMAP_HOTPLUG); | |
718127cc | 410 | return 0; |
3947be19 DH |
411 | } |
412 | ||
c04fc586 GH |
413 | static int __meminit __add_section(int nid, struct zone *zone, |
414 | unsigned long phys_start_pfn) | |
3947be19 | 415 | { |
3947be19 | 416 | int nr_pages = PAGES_PER_SECTION; |
3947be19 DH |
417 | int ret; |
418 | ||
ebd15302 KH |
419 | if (pfn_valid(phys_start_pfn)) |
420 | return -EEXIST; | |
421 | ||
0b0acbec | 422 | ret = sparse_add_one_section(zone, phys_start_pfn, nr_pages); |
3947be19 DH |
423 | |
424 | if (ret < 0) | |
425 | return ret; | |
426 | ||
718127cc YG |
427 | ret = __add_zone(zone, phys_start_pfn); |
428 | ||
429 | if (ret < 0) | |
430 | return ret; | |
431 | ||
c04fc586 | 432 | return register_new_memory(nid, __pfn_to_section(phys_start_pfn)); |
3947be19 DH |
433 | } |
434 | ||
0c0a4a51 YG |
435 | #ifdef CONFIG_SPARSEMEM_VMEMMAP |
436 | static int __remove_section(struct zone *zone, struct mem_section *ms) | |
437 | { | |
438 | /* | |
439 | * XXX: Freeing memmap with vmemmap is not implement yet. | |
440 | * This should be removed later. | |
441 | */ | |
442 | return -EBUSY; | |
443 | } | |
444 | #else | |
ea01ea93 BP |
445 | static int __remove_section(struct zone *zone, struct mem_section *ms) |
446 | { | |
447 | unsigned long flags; | |
448 | struct pglist_data *pgdat = zone->zone_pgdat; | |
449 | int ret = -EINVAL; | |
450 | ||
451 | if (!valid_section(ms)) | |
452 | return ret; | |
453 | ||
454 | ret = unregister_memory_section(ms); | |
455 | if (ret) | |
456 | return ret; | |
457 | ||
458 | pgdat_resize_lock(pgdat, &flags); | |
459 | sparse_remove_one_section(zone, ms); | |
460 | pgdat_resize_unlock(pgdat, &flags); | |
461 | return 0; | |
462 | } | |
0c0a4a51 | 463 | #endif |
ea01ea93 | 464 | |
3947be19 DH |
465 | /* |
466 | * Reasonably generic function for adding memory. It is | |
467 | * expected that archs that support memory hotplug will | |
468 | * call this function after deciding the zone to which to | |
469 | * add the new pages. | |
470 | */ | |
c04fc586 GH |
471 | int __ref __add_pages(int nid, struct zone *zone, unsigned long phys_start_pfn, |
472 | unsigned long nr_pages) | |
3947be19 DH |
473 | { |
474 | unsigned long i; | |
475 | int err = 0; | |
6f712711 KH |
476 | int start_sec, end_sec; |
477 | /* during initialize mem_map, align hot-added range to section */ | |
478 | start_sec = pfn_to_section_nr(phys_start_pfn); | |
479 | end_sec = pfn_to_section_nr(phys_start_pfn + nr_pages - 1); | |
3947be19 | 480 | |
6f712711 | 481 | for (i = start_sec; i <= end_sec; i++) { |
c04fc586 | 482 | err = __add_section(nid, zone, i << PFN_SECTION_SHIFT); |
3947be19 | 483 | |
6f712711 | 484 | /* |
183ff22b | 485 | * EEXIST is finally dealt with by ioresource collision |
6f712711 KH |
486 | * check. see add_memory() => register_memory_resource() |
487 | * Warning will be printed if there is collision. | |
bed120c6 JS |
488 | */ |
489 | if (err && (err != -EEXIST)) | |
3947be19 | 490 | break; |
6f712711 | 491 | err = 0; |
3947be19 DH |
492 | } |
493 | ||
494 | return err; | |
495 | } | |
bed120c6 | 496 | EXPORT_SYMBOL_GPL(__add_pages); |
3947be19 | 497 | |
ea01ea93 BP |
498 | /** |
499 | * __remove_pages() - remove sections of pages from a zone | |
500 | * @zone: zone from which pages need to be removed | |
501 | * @phys_start_pfn: starting pageframe (must be aligned to start of a section) | |
502 | * @nr_pages: number of pages to remove (must be multiple of section size) | |
503 | * | |
504 | * Generic helper function to remove section mappings and sysfs entries | |
505 | * for the section of the memory we are removing. Caller needs to make | |
506 | * sure that pages are marked reserved and zones are adjust properly by | |
507 | * calling offline_pages(). | |
508 | */ | |
509 | int __remove_pages(struct zone *zone, unsigned long phys_start_pfn, | |
510 | unsigned long nr_pages) | |
511 | { | |
512 | unsigned long i, ret = 0; | |
513 | int sections_to_remove; | |
514 | ||
515 | /* | |
516 | * We can only remove entire sections | |
517 | */ | |
518 | BUG_ON(phys_start_pfn & ~PAGE_SECTION_MASK); | |
519 | BUG_ON(nr_pages % PAGES_PER_SECTION); | |
520 | ||
d760afd4 YI |
521 | release_mem_region(phys_start_pfn << PAGE_SHIFT, nr_pages * PAGE_SIZE); |
522 | ||
ea01ea93 BP |
523 | sections_to_remove = nr_pages / PAGES_PER_SECTION; |
524 | for (i = 0; i < sections_to_remove; i++) { | |
525 | unsigned long pfn = phys_start_pfn + i*PAGES_PER_SECTION; | |
526 | ret = __remove_section(zone, __pfn_to_section(pfn)); | |
527 | if (ret) | |
528 | break; | |
529 | } | |
530 | return ret; | |
531 | } | |
532 | EXPORT_SYMBOL_GPL(__remove_pages); | |
533 | ||
9d0ad8ca DK |
534 | int set_online_page_callback(online_page_callback_t callback) |
535 | { | |
536 | int rc = -EINVAL; | |
537 | ||
538 | lock_memory_hotplug(); | |
539 | ||
540 | if (online_page_callback == generic_online_page) { | |
541 | online_page_callback = callback; | |
542 | rc = 0; | |
543 | } | |
544 | ||
545 | unlock_memory_hotplug(); | |
546 | ||
547 | return rc; | |
548 | } | |
549 | EXPORT_SYMBOL_GPL(set_online_page_callback); | |
550 | ||
551 | int restore_online_page_callback(online_page_callback_t callback) | |
552 | { | |
553 | int rc = -EINVAL; | |
554 | ||
555 | lock_memory_hotplug(); | |
556 | ||
557 | if (online_page_callback == callback) { | |
558 | online_page_callback = generic_online_page; | |
559 | rc = 0; | |
560 | } | |
561 | ||
562 | unlock_memory_hotplug(); | |
563 | ||
564 | return rc; | |
565 | } | |
566 | EXPORT_SYMBOL_GPL(restore_online_page_callback); | |
567 | ||
568 | void __online_page_set_limits(struct page *page) | |
180c06ef | 569 | { |
4738e1b9 JB |
570 | unsigned long pfn = page_to_pfn(page); |
571 | ||
4738e1b9 JB |
572 | if (pfn >= num_physpages) |
573 | num_physpages = pfn + 1; | |
9d0ad8ca DK |
574 | } |
575 | EXPORT_SYMBOL_GPL(__online_page_set_limits); | |
576 | ||
577 | void __online_page_increment_counters(struct page *page) | |
578 | { | |
579 | totalram_pages++; | |
180c06ef JF |
580 | |
581 | #ifdef CONFIG_HIGHMEM | |
582 | if (PageHighMem(page)) | |
583 | totalhigh_pages++; | |
584 | #endif | |
9d0ad8ca DK |
585 | } |
586 | EXPORT_SYMBOL_GPL(__online_page_increment_counters); | |
180c06ef | 587 | |
9d0ad8ca DK |
588 | void __online_page_free(struct page *page) |
589 | { | |
180c06ef JF |
590 | ClearPageReserved(page); |
591 | init_page_count(page); | |
592 | __free_page(page); | |
593 | } | |
9d0ad8ca DK |
594 | EXPORT_SYMBOL_GPL(__online_page_free); |
595 | ||
596 | static void generic_online_page(struct page *page) | |
597 | { | |
598 | __online_page_set_limits(page); | |
599 | __online_page_increment_counters(page); | |
600 | __online_page_free(page); | |
601 | } | |
180c06ef | 602 | |
75884fb1 KH |
603 | static int online_pages_range(unsigned long start_pfn, unsigned long nr_pages, |
604 | void *arg) | |
3947be19 DH |
605 | { |
606 | unsigned long i; | |
75884fb1 KH |
607 | unsigned long onlined_pages = *(unsigned long *)arg; |
608 | struct page *page; | |
609 | if (PageReserved(pfn_to_page(start_pfn))) | |
610 | for (i = 0; i < nr_pages; i++) { | |
611 | page = pfn_to_page(start_pfn + i); | |
9d0ad8ca | 612 | (*online_page_callback)(page); |
75884fb1 KH |
613 | onlined_pages++; |
614 | } | |
615 | *(unsigned long *)arg = onlined_pages; | |
616 | return 0; | |
617 | } | |
618 | ||
09285af7 | 619 | #ifdef CONFIG_MOVABLE_NODE |
79a4dcef TC |
620 | /* |
621 | * When CONFIG_MOVABLE_NODE, we permit onlining of a node which doesn't have | |
622 | * normal memory. | |
623 | */ | |
09285af7 LJ |
624 | static bool can_online_high_movable(struct zone *zone) |
625 | { | |
626 | return true; | |
627 | } | |
79a4dcef | 628 | #else /* CONFIG_MOVABLE_NODE */ |
74d42d8f LJ |
629 | /* ensure every online node has NORMAL memory */ |
630 | static bool can_online_high_movable(struct zone *zone) | |
631 | { | |
632 | return node_state(zone_to_nid(zone), N_NORMAL_MEMORY); | |
633 | } | |
79a4dcef | 634 | #endif /* CONFIG_MOVABLE_NODE */ |
74d42d8f | 635 | |
d9713679 LJ |
636 | /* check which state of node_states will be changed when online memory */ |
637 | static void node_states_check_changes_online(unsigned long nr_pages, | |
638 | struct zone *zone, struct memory_notify *arg) | |
639 | { | |
640 | int nid = zone_to_nid(zone); | |
641 | enum zone_type zone_last = ZONE_NORMAL; | |
642 | ||
643 | /* | |
6715ddf9 LJ |
644 | * If we have HIGHMEM or movable node, node_states[N_NORMAL_MEMORY] |
645 | * contains nodes which have zones of 0...ZONE_NORMAL, | |
646 | * set zone_last to ZONE_NORMAL. | |
d9713679 | 647 | * |
6715ddf9 LJ |
648 | * If we don't have HIGHMEM nor movable node, |
649 | * node_states[N_NORMAL_MEMORY] contains nodes which have zones of | |
650 | * 0...ZONE_MOVABLE, set zone_last to ZONE_MOVABLE. | |
d9713679 | 651 | */ |
6715ddf9 | 652 | if (N_MEMORY == N_NORMAL_MEMORY) |
d9713679 LJ |
653 | zone_last = ZONE_MOVABLE; |
654 | ||
655 | /* | |
656 | * if the memory to be online is in a zone of 0...zone_last, and | |
657 | * the zones of 0...zone_last don't have memory before online, we will | |
658 | * need to set the node to node_states[N_NORMAL_MEMORY] after | |
659 | * the memory is online. | |
660 | */ | |
661 | if (zone_idx(zone) <= zone_last && !node_state(nid, N_NORMAL_MEMORY)) | |
662 | arg->status_change_nid_normal = nid; | |
663 | else | |
664 | arg->status_change_nid_normal = -1; | |
665 | ||
6715ddf9 LJ |
666 | #ifdef CONFIG_HIGHMEM |
667 | /* | |
668 | * If we have movable node, node_states[N_HIGH_MEMORY] | |
669 | * contains nodes which have zones of 0...ZONE_HIGHMEM, | |
670 | * set zone_last to ZONE_HIGHMEM. | |
671 | * | |
672 | * If we don't have movable node, node_states[N_NORMAL_MEMORY] | |
673 | * contains nodes which have zones of 0...ZONE_MOVABLE, | |
674 | * set zone_last to ZONE_MOVABLE. | |
675 | */ | |
676 | zone_last = ZONE_HIGHMEM; | |
677 | if (N_MEMORY == N_HIGH_MEMORY) | |
678 | zone_last = ZONE_MOVABLE; | |
679 | ||
680 | if (zone_idx(zone) <= zone_last && !node_state(nid, N_HIGH_MEMORY)) | |
681 | arg->status_change_nid_high = nid; | |
682 | else | |
683 | arg->status_change_nid_high = -1; | |
684 | #else | |
685 | arg->status_change_nid_high = arg->status_change_nid_normal; | |
686 | #endif | |
687 | ||
d9713679 LJ |
688 | /* |
689 | * if the node don't have memory befor online, we will need to | |
6715ddf9 | 690 | * set the node to node_states[N_MEMORY] after the memory |
d9713679 LJ |
691 | * is online. |
692 | */ | |
6715ddf9 | 693 | if (!node_state(nid, N_MEMORY)) |
d9713679 LJ |
694 | arg->status_change_nid = nid; |
695 | else | |
696 | arg->status_change_nid = -1; | |
697 | } | |
698 | ||
699 | static void node_states_set_node(int node, struct memory_notify *arg) | |
700 | { | |
701 | if (arg->status_change_nid_normal >= 0) | |
702 | node_set_state(node, N_NORMAL_MEMORY); | |
703 | ||
6715ddf9 LJ |
704 | if (arg->status_change_nid_high >= 0) |
705 | node_set_state(node, N_HIGH_MEMORY); | |
706 | ||
707 | node_set_state(node, N_MEMORY); | |
d9713679 LJ |
708 | } |
709 | ||
75884fb1 | 710 | |
511c2aba | 711 | int __ref online_pages(unsigned long pfn, unsigned long nr_pages, int online_type) |
75884fb1 | 712 | { |
3947be19 DH |
713 | unsigned long onlined_pages = 0; |
714 | struct zone *zone; | |
6811378e | 715 | int need_zonelists_rebuild = 0; |
7b78d335 YG |
716 | int nid; |
717 | int ret; | |
718 | struct memory_notify arg; | |
719 | ||
925268a0 | 720 | lock_memory_hotplug(); |
d9713679 LJ |
721 | /* |
722 | * This doesn't need a lock to do pfn_to_page(). | |
723 | * The section can't be removed here because of the | |
724 | * memory_block->state_mutex. | |
725 | */ | |
726 | zone = page_zone(pfn_to_page(pfn)); | |
727 | ||
74d42d8f LJ |
728 | if ((zone_idx(zone) > ZONE_NORMAL || online_type == ONLINE_MOVABLE) && |
729 | !can_online_high_movable(zone)) { | |
730 | unlock_memory_hotplug(); | |
731 | return -1; | |
732 | } | |
733 | ||
511c2aba LJ |
734 | if (online_type == ONLINE_KERNEL && zone_idx(zone) == ZONE_MOVABLE) { |
735 | if (move_pfn_range_left(zone - 1, zone, pfn, pfn + nr_pages)) { | |
736 | unlock_memory_hotplug(); | |
737 | return -1; | |
738 | } | |
739 | } | |
740 | if (online_type == ONLINE_MOVABLE && zone_idx(zone) == ZONE_MOVABLE - 1) { | |
741 | if (move_pfn_range_right(zone, zone + 1, pfn, pfn + nr_pages)) { | |
742 | unlock_memory_hotplug(); | |
743 | return -1; | |
744 | } | |
745 | } | |
746 | ||
747 | /* Previous code may changed the zone of the pfn range */ | |
748 | zone = page_zone(pfn_to_page(pfn)); | |
749 | ||
7b78d335 YG |
750 | arg.start_pfn = pfn; |
751 | arg.nr_pages = nr_pages; | |
d9713679 | 752 | node_states_check_changes_online(nr_pages, zone, &arg); |
7b78d335 YG |
753 | |
754 | nid = page_to_nid(pfn_to_page(pfn)); | |
3947be19 | 755 | |
7b78d335 YG |
756 | ret = memory_notify(MEM_GOING_ONLINE, &arg); |
757 | ret = notifier_to_errno(ret); | |
758 | if (ret) { | |
759 | memory_notify(MEM_CANCEL_ONLINE, &arg); | |
925268a0 | 760 | unlock_memory_hotplug(); |
7b78d335 YG |
761 | return ret; |
762 | } | |
6811378e YG |
763 | /* |
764 | * If this zone is not populated, then it is not in zonelist. | |
765 | * This means the page allocator ignores this zone. | |
766 | * So, zonelist must be updated after online. | |
767 | */ | |
4eaf3f64 | 768 | mutex_lock(&zonelists_mutex); |
6dcd73d7 | 769 | if (!populated_zone(zone)) { |
6811378e | 770 | need_zonelists_rebuild = 1; |
6dcd73d7 WC |
771 | build_all_zonelists(NULL, zone); |
772 | } | |
6811378e | 773 | |
908eedc6 | 774 | ret = walk_system_ram_range(pfn, nr_pages, &onlined_pages, |
75884fb1 | 775 | online_pages_range); |
fd8a4221 | 776 | if (ret) { |
6dcd73d7 WC |
777 | if (need_zonelists_rebuild) |
778 | zone_pcp_reset(zone); | |
4eaf3f64 | 779 | mutex_unlock(&zonelists_mutex); |
a62e2f4f BH |
780 | printk(KERN_DEBUG "online_pages [mem %#010llx-%#010llx] failed\n", |
781 | (unsigned long long) pfn << PAGE_SHIFT, | |
782 | (((unsigned long long) pfn + nr_pages) | |
783 | << PAGE_SHIFT) - 1); | |
fd8a4221 | 784 | memory_notify(MEM_CANCEL_ONLINE, &arg); |
925268a0 | 785 | unlock_memory_hotplug(); |
fd8a4221 GL |
786 | return ret; |
787 | } | |
788 | ||
9feedc9d | 789 | zone->managed_pages += onlined_pages; |
3947be19 | 790 | zone->present_pages += onlined_pages; |
f2937be5 | 791 | zone->zone_pgdat->node_present_pages += onlined_pages; |
08dff7b7 | 792 | if (onlined_pages) { |
d9713679 | 793 | node_states_set_node(zone_to_nid(zone), &arg); |
08dff7b7 | 794 | if (need_zonelists_rebuild) |
6dcd73d7 | 795 | build_all_zonelists(NULL, NULL); |
08dff7b7 JL |
796 | else |
797 | zone_pcp_update(zone); | |
798 | } | |
3947be19 | 799 | |
4eaf3f64 | 800 | mutex_unlock(&zonelists_mutex); |
1b79acc9 KM |
801 | |
802 | init_per_zone_wmark_min(); | |
803 | ||
08dff7b7 | 804 | if (onlined_pages) |
7ea1530a | 805 | kswapd_run(zone_to_nid(zone)); |
61b13993 | 806 | |
1f522509 | 807 | vm_total_pages = nr_free_pagecache_pages(); |
2f7f24ec | 808 | |
2d1d43f6 | 809 | writeback_set_ratelimit(); |
7b78d335 YG |
810 | |
811 | if (onlined_pages) | |
812 | memory_notify(MEM_ONLINE, &arg); | |
925268a0 | 813 | unlock_memory_hotplug(); |
7b78d335 | 814 | |
3947be19 DH |
815 | return 0; |
816 | } | |
53947027 | 817 | #endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */ |
bc02af93 | 818 | |
e1319331 HS |
819 | /* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */ |
820 | static pg_data_t __ref *hotadd_new_pgdat(int nid, u64 start) | |
9af3c2de YG |
821 | { |
822 | struct pglist_data *pgdat; | |
823 | unsigned long zones_size[MAX_NR_ZONES] = {0}; | |
824 | unsigned long zholes_size[MAX_NR_ZONES] = {0}; | |
825 | unsigned long start_pfn = start >> PAGE_SHIFT; | |
826 | ||
827 | pgdat = arch_alloc_nodedata(nid); | |
828 | if (!pgdat) | |
829 | return NULL; | |
830 | ||
831 | arch_refresh_nodedata(nid, pgdat); | |
832 | ||
833 | /* we can use NODE_DATA(nid) from here */ | |
834 | ||
835 | /* init node's zones as empty zones, we don't have any present pages.*/ | |
9109fb7b | 836 | free_area_init_node(nid, zones_size, start_pfn, zholes_size); |
9af3c2de | 837 | |
959ecc48 KH |
838 | /* |
839 | * The node we allocated has no zone fallback lists. For avoiding | |
840 | * to access not-initialized zonelist, build here. | |
841 | */ | |
f957db4f | 842 | mutex_lock(&zonelists_mutex); |
9adb62a5 | 843 | build_all_zonelists(pgdat, NULL); |
f957db4f | 844 | mutex_unlock(&zonelists_mutex); |
959ecc48 | 845 | |
9af3c2de YG |
846 | return pgdat; |
847 | } | |
848 | ||
849 | static void rollback_node_hotadd(int nid, pg_data_t *pgdat) | |
850 | { | |
851 | arch_refresh_nodedata(nid, NULL); | |
852 | arch_free_nodedata(pgdat); | |
853 | return; | |
854 | } | |
855 | ||
0a547039 | 856 | |
cf23422b | 857 | /* |
858 | * called by cpu_up() to online a node without onlined memory. | |
859 | */ | |
860 | int mem_online_node(int nid) | |
861 | { | |
862 | pg_data_t *pgdat; | |
863 | int ret; | |
864 | ||
20d6c96b | 865 | lock_memory_hotplug(); |
cf23422b | 866 | pgdat = hotadd_new_pgdat(nid, 0); |
7553e8f2 | 867 | if (!pgdat) { |
cf23422b | 868 | ret = -ENOMEM; |
869 | goto out; | |
870 | } | |
871 | node_set_online(nid); | |
872 | ret = register_one_node(nid); | |
873 | BUG_ON(ret); | |
874 | ||
875 | out: | |
20d6c96b | 876 | unlock_memory_hotplug(); |
cf23422b | 877 | return ret; |
878 | } | |
879 | ||
31168481 AV |
880 | /* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */ |
881 | int __ref add_memory(int nid, u64 start, u64 size) | |
bc02af93 | 882 | { |
9af3c2de YG |
883 | pg_data_t *pgdat = NULL; |
884 | int new_pgdat = 0; | |
ebd15302 | 885 | struct resource *res; |
bc02af93 YG |
886 | int ret; |
887 | ||
20d6c96b | 888 | lock_memory_hotplug(); |
6ad696d2 | 889 | |
ebd15302 | 890 | res = register_memory_resource(start, size); |
6ad696d2 | 891 | ret = -EEXIST; |
ebd15302 | 892 | if (!res) |
6ad696d2 | 893 | goto out; |
ebd15302 | 894 | |
9af3c2de YG |
895 | if (!node_online(nid)) { |
896 | pgdat = hotadd_new_pgdat(nid, start); | |
6ad696d2 | 897 | ret = -ENOMEM; |
9af3c2de | 898 | if (!pgdat) |
41b9e2d7 | 899 | goto error; |
9af3c2de | 900 | new_pgdat = 1; |
9af3c2de YG |
901 | } |
902 | ||
bc02af93 YG |
903 | /* call arch's memory hotadd */ |
904 | ret = arch_add_memory(nid, start, size); | |
905 | ||
9af3c2de YG |
906 | if (ret < 0) |
907 | goto error; | |
908 | ||
0fc44159 | 909 | /* we online node here. we can't roll back from here. */ |
9af3c2de YG |
910 | node_set_online(nid); |
911 | ||
0fc44159 YG |
912 | if (new_pgdat) { |
913 | ret = register_one_node(nid); | |
914 | /* | |
915 | * If sysfs file of new node can't create, cpu on the node | |
916 | * can't be hot-added. There is no rollback way now. | |
917 | * So, check by BUG_ON() to catch it reluctantly.. | |
918 | */ | |
919 | BUG_ON(ret); | |
920 | } | |
921 | ||
d96ae530 | 922 | /* create new memmap entry */ |
923 | firmware_map_add_hotplug(start, start + size, "System RAM"); | |
924 | ||
6ad696d2 AK |
925 | goto out; |
926 | ||
9af3c2de YG |
927 | error: |
928 | /* rollback pgdat allocation and others */ | |
929 | if (new_pgdat) | |
930 | rollback_node_hotadd(nid, pgdat); | |
a864b9d0 | 931 | release_memory_resource(res); |
9af3c2de | 932 | |
6ad696d2 | 933 | out: |
20d6c96b | 934 | unlock_memory_hotplug(); |
bc02af93 YG |
935 | return ret; |
936 | } | |
937 | EXPORT_SYMBOL_GPL(add_memory); | |
0c0e6195 KH |
938 | |
939 | #ifdef CONFIG_MEMORY_HOTREMOVE | |
5c755e9f BP |
940 | /* |
941 | * A free page on the buddy free lists (not the per-cpu lists) has PageBuddy | |
942 | * set and the size of the free page is given by page_order(). Using this, | |
943 | * the function determines if the pageblock contains only free pages. | |
944 | * Due to buddy contraints, a free page at least the size of a pageblock will | |
945 | * be located at the start of the pageblock | |
946 | */ | |
947 | static inline int pageblock_free(struct page *page) | |
948 | { | |
949 | return PageBuddy(page) && page_order(page) >= pageblock_order; | |
950 | } | |
951 | ||
952 | /* Return the start of the next active pageblock after a given page */ | |
953 | static struct page *next_active_pageblock(struct page *page) | |
954 | { | |
5c755e9f BP |
955 | /* Ensure the starting page is pageblock-aligned */ |
956 | BUG_ON(page_to_pfn(page) & (pageblock_nr_pages - 1)); | |
957 | ||
5c755e9f | 958 | /* If the entire pageblock is free, move to the end of free page */ |
0dcc48c1 KH |
959 | if (pageblock_free(page)) { |
960 | int order; | |
961 | /* be careful. we don't have locks, page_order can be changed.*/ | |
962 | order = page_order(page); | |
963 | if ((order < MAX_ORDER) && (order >= pageblock_order)) | |
964 | return page + (1 << order); | |
965 | } | |
5c755e9f | 966 | |
0dcc48c1 | 967 | return page + pageblock_nr_pages; |
5c755e9f BP |
968 | } |
969 | ||
970 | /* Checks if this range of memory is likely to be hot-removable. */ | |
971 | int is_mem_section_removable(unsigned long start_pfn, unsigned long nr_pages) | |
972 | { | |
5c755e9f BP |
973 | struct page *page = pfn_to_page(start_pfn); |
974 | struct page *end_page = page + nr_pages; | |
975 | ||
976 | /* Check the starting page of each pageblock within the range */ | |
977 | for (; page < end_page; page = next_active_pageblock(page)) { | |
49ac8255 | 978 | if (!is_pageblock_removable_nolock(page)) |
5c755e9f | 979 | return 0; |
49ac8255 | 980 | cond_resched(); |
5c755e9f BP |
981 | } |
982 | ||
983 | /* All pageblocks in the memory block are likely to be hot-removable */ | |
984 | return 1; | |
985 | } | |
986 | ||
0c0e6195 KH |
987 | /* |
988 | * Confirm all pages in a range [start, end) is belongs to the same zone. | |
989 | */ | |
990 | static int test_pages_in_a_zone(unsigned long start_pfn, unsigned long end_pfn) | |
991 | { | |
992 | unsigned long pfn; | |
993 | struct zone *zone = NULL; | |
994 | struct page *page; | |
995 | int i; | |
996 | for (pfn = start_pfn; | |
997 | pfn < end_pfn; | |
998 | pfn += MAX_ORDER_NR_PAGES) { | |
999 | i = 0; | |
1000 | /* This is just a CONFIG_HOLES_IN_ZONE check.*/ | |
1001 | while ((i < MAX_ORDER_NR_PAGES) && !pfn_valid_within(pfn + i)) | |
1002 | i++; | |
1003 | if (i == MAX_ORDER_NR_PAGES) | |
1004 | continue; | |
1005 | page = pfn_to_page(pfn + i); | |
1006 | if (zone && page_zone(page) != zone) | |
1007 | return 0; | |
1008 | zone = page_zone(page); | |
1009 | } | |
1010 | return 1; | |
1011 | } | |
1012 | ||
1013 | /* | |
1014 | * Scanning pfn is much easier than scanning lru list. | |
1015 | * Scan pfn from start to end and Find LRU page. | |
1016 | */ | |
7bbc0905 | 1017 | static unsigned long scan_lru_pages(unsigned long start, unsigned long end) |
0c0e6195 KH |
1018 | { |
1019 | unsigned long pfn; | |
1020 | struct page *page; | |
1021 | for (pfn = start; pfn < end; pfn++) { | |
1022 | if (pfn_valid(pfn)) { | |
1023 | page = pfn_to_page(pfn); | |
1024 | if (PageLRU(page)) | |
1025 | return pfn; | |
1026 | } | |
1027 | } | |
1028 | return 0; | |
1029 | } | |
1030 | ||
0c0e6195 KH |
1031 | #define NR_OFFLINE_AT_ONCE_PAGES (256) |
1032 | static int | |
1033 | do_migrate_range(unsigned long start_pfn, unsigned long end_pfn) | |
1034 | { | |
1035 | unsigned long pfn; | |
1036 | struct page *page; | |
1037 | int move_pages = NR_OFFLINE_AT_ONCE_PAGES; | |
1038 | int not_managed = 0; | |
1039 | int ret = 0; | |
1040 | LIST_HEAD(source); | |
1041 | ||
1042 | for (pfn = start_pfn; pfn < end_pfn && move_pages > 0; pfn++) { | |
1043 | if (!pfn_valid(pfn)) | |
1044 | continue; | |
1045 | page = pfn_to_page(pfn); | |
700c2a46 | 1046 | if (!get_page_unless_zero(page)) |
0c0e6195 KH |
1047 | continue; |
1048 | /* | |
1049 | * We can skip free pages. And we can only deal with pages on | |
1050 | * LRU. | |
1051 | */ | |
62695a84 | 1052 | ret = isolate_lru_page(page); |
0c0e6195 | 1053 | if (!ret) { /* Success */ |
700c2a46 | 1054 | put_page(page); |
62695a84 | 1055 | list_add_tail(&page->lru, &source); |
0c0e6195 | 1056 | move_pages--; |
6d9c285a KM |
1057 | inc_zone_page_state(page, NR_ISOLATED_ANON + |
1058 | page_is_file_cache(page)); | |
1059 | ||
0c0e6195 | 1060 | } else { |
0c0e6195 | 1061 | #ifdef CONFIG_DEBUG_VM |
718a3821 WF |
1062 | printk(KERN_ALERT "removing pfn %lx from LRU failed\n", |
1063 | pfn); | |
1064 | dump_page(page); | |
0c0e6195 | 1065 | #endif |
700c2a46 | 1066 | put_page(page); |
25985edc | 1067 | /* Because we don't have big zone->lock. we should |
809c4449 BL |
1068 | check this again here. */ |
1069 | if (page_count(page)) { | |
1070 | not_managed++; | |
f3ab2636 | 1071 | ret = -EBUSY; |
809c4449 BL |
1072 | break; |
1073 | } | |
0c0e6195 KH |
1074 | } |
1075 | } | |
f3ab2636 BL |
1076 | if (!list_empty(&source)) { |
1077 | if (not_managed) { | |
1078 | putback_lru_pages(&source); | |
1079 | goto out; | |
1080 | } | |
74c08f98 MK |
1081 | |
1082 | /* | |
1083 | * alloc_migrate_target should be improooooved!! | |
1084 | * migrate_pages returns # of failed pages. | |
1085 | */ | |
1086 | ret = migrate_pages(&source, alloc_migrate_target, 0, | |
7b2a2d4a MG |
1087 | true, MIGRATE_SYNC, |
1088 | MR_MEMORY_HOTPLUG); | |
f3ab2636 | 1089 | if (ret) |
0c0e6195 | 1090 | putback_lru_pages(&source); |
0c0e6195 | 1091 | } |
0c0e6195 KH |
1092 | out: |
1093 | return ret; | |
1094 | } | |
1095 | ||
1096 | /* | |
1097 | * remove from free_area[] and mark all as Reserved. | |
1098 | */ | |
1099 | static int | |
1100 | offline_isolated_pages_cb(unsigned long start, unsigned long nr_pages, | |
1101 | void *data) | |
1102 | { | |
1103 | __offline_isolated_pages(start, start + nr_pages); | |
1104 | return 0; | |
1105 | } | |
1106 | ||
1107 | static void | |
1108 | offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn) | |
1109 | { | |
908eedc6 | 1110 | walk_system_ram_range(start_pfn, end_pfn - start_pfn, NULL, |
0c0e6195 KH |
1111 | offline_isolated_pages_cb); |
1112 | } | |
1113 | ||
1114 | /* | |
1115 | * Check all pages in range, recoreded as memory resource, are isolated. | |
1116 | */ | |
1117 | static int | |
1118 | check_pages_isolated_cb(unsigned long start_pfn, unsigned long nr_pages, | |
1119 | void *data) | |
1120 | { | |
1121 | int ret; | |
1122 | long offlined = *(long *)data; | |
b023f468 | 1123 | ret = test_pages_isolated(start_pfn, start_pfn + nr_pages, true); |
0c0e6195 KH |
1124 | offlined = nr_pages; |
1125 | if (!ret) | |
1126 | *(long *)data += offlined; | |
1127 | return ret; | |
1128 | } | |
1129 | ||
1130 | static long | |
1131 | check_pages_isolated(unsigned long start_pfn, unsigned long end_pfn) | |
1132 | { | |
1133 | long offlined = 0; | |
1134 | int ret; | |
1135 | ||
908eedc6 | 1136 | ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn, &offlined, |
0c0e6195 KH |
1137 | check_pages_isolated_cb); |
1138 | if (ret < 0) | |
1139 | offlined = (long)ret; | |
1140 | return offlined; | |
1141 | } | |
1142 | ||
09285af7 | 1143 | #ifdef CONFIG_MOVABLE_NODE |
79a4dcef TC |
1144 | /* |
1145 | * When CONFIG_MOVABLE_NODE, we permit offlining of a node which doesn't have | |
1146 | * normal memory. | |
1147 | */ | |
09285af7 LJ |
1148 | static bool can_offline_normal(struct zone *zone, unsigned long nr_pages) |
1149 | { | |
1150 | return true; | |
1151 | } | |
79a4dcef | 1152 | #else /* CONFIG_MOVABLE_NODE */ |
74d42d8f LJ |
1153 | /* ensure the node has NORMAL memory if it is still online */ |
1154 | static bool can_offline_normal(struct zone *zone, unsigned long nr_pages) | |
1155 | { | |
1156 | struct pglist_data *pgdat = zone->zone_pgdat; | |
1157 | unsigned long present_pages = 0; | |
1158 | enum zone_type zt; | |
1159 | ||
1160 | for (zt = 0; zt <= ZONE_NORMAL; zt++) | |
1161 | present_pages += pgdat->node_zones[zt].present_pages; | |
1162 | ||
1163 | if (present_pages > nr_pages) | |
1164 | return true; | |
1165 | ||
1166 | present_pages = 0; | |
1167 | for (; zt <= ZONE_MOVABLE; zt++) | |
1168 | present_pages += pgdat->node_zones[zt].present_pages; | |
1169 | ||
1170 | /* | |
1171 | * we can't offline the last normal memory until all | |
1172 | * higher memory is offlined. | |
1173 | */ | |
1174 | return present_pages == 0; | |
1175 | } | |
79a4dcef | 1176 | #endif /* CONFIG_MOVABLE_NODE */ |
74d42d8f | 1177 | |
d9713679 LJ |
1178 | /* check which state of node_states will be changed when offline memory */ |
1179 | static void node_states_check_changes_offline(unsigned long nr_pages, | |
1180 | struct zone *zone, struct memory_notify *arg) | |
1181 | { | |
1182 | struct pglist_data *pgdat = zone->zone_pgdat; | |
1183 | unsigned long present_pages = 0; | |
1184 | enum zone_type zt, zone_last = ZONE_NORMAL; | |
1185 | ||
1186 | /* | |
6715ddf9 LJ |
1187 | * If we have HIGHMEM or movable node, node_states[N_NORMAL_MEMORY] |
1188 | * contains nodes which have zones of 0...ZONE_NORMAL, | |
1189 | * set zone_last to ZONE_NORMAL. | |
d9713679 | 1190 | * |
6715ddf9 LJ |
1191 | * If we don't have HIGHMEM nor movable node, |
1192 | * node_states[N_NORMAL_MEMORY] contains nodes which have zones of | |
1193 | * 0...ZONE_MOVABLE, set zone_last to ZONE_MOVABLE. | |
d9713679 | 1194 | */ |
6715ddf9 | 1195 | if (N_MEMORY == N_NORMAL_MEMORY) |
d9713679 LJ |
1196 | zone_last = ZONE_MOVABLE; |
1197 | ||
1198 | /* | |
1199 | * check whether node_states[N_NORMAL_MEMORY] will be changed. | |
1200 | * If the memory to be offline is in a zone of 0...zone_last, | |
1201 | * and it is the last present memory, 0...zone_last will | |
1202 | * become empty after offline , thus we can determind we will | |
1203 | * need to clear the node from node_states[N_NORMAL_MEMORY]. | |
1204 | */ | |
1205 | for (zt = 0; zt <= zone_last; zt++) | |
1206 | present_pages += pgdat->node_zones[zt].present_pages; | |
1207 | if (zone_idx(zone) <= zone_last && nr_pages >= present_pages) | |
1208 | arg->status_change_nid_normal = zone_to_nid(zone); | |
1209 | else | |
1210 | arg->status_change_nid_normal = -1; | |
1211 | ||
6715ddf9 LJ |
1212 | #ifdef CONFIG_HIGHMEM |
1213 | /* | |
1214 | * If we have movable node, node_states[N_HIGH_MEMORY] | |
1215 | * contains nodes which have zones of 0...ZONE_HIGHMEM, | |
1216 | * set zone_last to ZONE_HIGHMEM. | |
1217 | * | |
1218 | * If we don't have movable node, node_states[N_NORMAL_MEMORY] | |
1219 | * contains nodes which have zones of 0...ZONE_MOVABLE, | |
1220 | * set zone_last to ZONE_MOVABLE. | |
1221 | */ | |
1222 | zone_last = ZONE_HIGHMEM; | |
1223 | if (N_MEMORY == N_HIGH_MEMORY) | |
1224 | zone_last = ZONE_MOVABLE; | |
1225 | ||
1226 | for (; zt <= zone_last; zt++) | |
1227 | present_pages += pgdat->node_zones[zt].present_pages; | |
1228 | if (zone_idx(zone) <= zone_last && nr_pages >= present_pages) | |
1229 | arg->status_change_nid_high = zone_to_nid(zone); | |
1230 | else | |
1231 | arg->status_change_nid_high = -1; | |
1232 | #else | |
1233 | arg->status_change_nid_high = arg->status_change_nid_normal; | |
1234 | #endif | |
1235 | ||
d9713679 LJ |
1236 | /* |
1237 | * node_states[N_HIGH_MEMORY] contains nodes which have 0...ZONE_MOVABLE | |
1238 | */ | |
1239 | zone_last = ZONE_MOVABLE; | |
1240 | ||
1241 | /* | |
1242 | * check whether node_states[N_HIGH_MEMORY] will be changed | |
1243 | * If we try to offline the last present @nr_pages from the node, | |
1244 | * we can determind we will need to clear the node from | |
1245 | * node_states[N_HIGH_MEMORY]. | |
1246 | */ | |
1247 | for (; zt <= zone_last; zt++) | |
1248 | present_pages += pgdat->node_zones[zt].present_pages; | |
1249 | if (nr_pages >= present_pages) | |
1250 | arg->status_change_nid = zone_to_nid(zone); | |
1251 | else | |
1252 | arg->status_change_nid = -1; | |
1253 | } | |
1254 | ||
1255 | static void node_states_clear_node(int node, struct memory_notify *arg) | |
1256 | { | |
1257 | if (arg->status_change_nid_normal >= 0) | |
1258 | node_clear_state(node, N_NORMAL_MEMORY); | |
1259 | ||
6715ddf9 LJ |
1260 | if ((N_MEMORY != N_NORMAL_MEMORY) && |
1261 | (arg->status_change_nid_high >= 0)) | |
d9713679 | 1262 | node_clear_state(node, N_HIGH_MEMORY); |
6715ddf9 LJ |
1263 | |
1264 | if ((N_MEMORY != N_HIGH_MEMORY) && | |
1265 | (arg->status_change_nid >= 0)) | |
1266 | node_clear_state(node, N_MEMORY); | |
d9713679 LJ |
1267 | } |
1268 | ||
a16cee10 | 1269 | static int __ref __offline_pages(unsigned long start_pfn, |
0c0e6195 KH |
1270 | unsigned long end_pfn, unsigned long timeout) |
1271 | { | |
1272 | unsigned long pfn, nr_pages, expire; | |
1273 | long offlined_pages; | |
7b78d335 | 1274 | int ret, drain, retry_max, node; |
0c0e6195 | 1275 | struct zone *zone; |
7b78d335 | 1276 | struct memory_notify arg; |
0c0e6195 KH |
1277 | |
1278 | BUG_ON(start_pfn >= end_pfn); | |
1279 | /* at least, alignment against pageblock is necessary */ | |
1280 | if (!IS_ALIGNED(start_pfn, pageblock_nr_pages)) | |
1281 | return -EINVAL; | |
1282 | if (!IS_ALIGNED(end_pfn, pageblock_nr_pages)) | |
1283 | return -EINVAL; | |
1284 | /* This makes hotplug much easier...and readable. | |
1285 | we assume this for now. .*/ | |
1286 | if (!test_pages_in_a_zone(start_pfn, end_pfn)) | |
1287 | return -EINVAL; | |
7b78d335 | 1288 | |
20d6c96b | 1289 | lock_memory_hotplug(); |
6ad696d2 | 1290 | |
7b78d335 YG |
1291 | zone = page_zone(pfn_to_page(start_pfn)); |
1292 | node = zone_to_nid(zone); | |
1293 | nr_pages = end_pfn - start_pfn; | |
1294 | ||
74d42d8f LJ |
1295 | ret = -EINVAL; |
1296 | if (zone_idx(zone) <= ZONE_NORMAL && !can_offline_normal(zone, nr_pages)) | |
1297 | goto out; | |
1298 | ||
0c0e6195 | 1299 | /* set above range as isolated */ |
b023f468 WC |
1300 | ret = start_isolate_page_range(start_pfn, end_pfn, |
1301 | MIGRATE_MOVABLE, true); | |
0c0e6195 | 1302 | if (ret) |
6ad696d2 | 1303 | goto out; |
7b78d335 YG |
1304 | |
1305 | arg.start_pfn = start_pfn; | |
1306 | arg.nr_pages = nr_pages; | |
d9713679 | 1307 | node_states_check_changes_offline(nr_pages, zone, &arg); |
7b78d335 YG |
1308 | |
1309 | ret = memory_notify(MEM_GOING_OFFLINE, &arg); | |
1310 | ret = notifier_to_errno(ret); | |
1311 | if (ret) | |
1312 | goto failed_removal; | |
1313 | ||
0c0e6195 KH |
1314 | pfn = start_pfn; |
1315 | expire = jiffies + timeout; | |
1316 | drain = 0; | |
1317 | retry_max = 5; | |
1318 | repeat: | |
1319 | /* start memory hot removal */ | |
1320 | ret = -EAGAIN; | |
1321 | if (time_after(jiffies, expire)) | |
1322 | goto failed_removal; | |
1323 | ret = -EINTR; | |
1324 | if (signal_pending(current)) | |
1325 | goto failed_removal; | |
1326 | ret = 0; | |
1327 | if (drain) { | |
1328 | lru_add_drain_all(); | |
0c0e6195 | 1329 | cond_resched(); |
9f8f2172 | 1330 | drain_all_pages(); |
0c0e6195 KH |
1331 | } |
1332 | ||
1333 | pfn = scan_lru_pages(start_pfn, end_pfn); | |
1334 | if (pfn) { /* We have page on LRU */ | |
1335 | ret = do_migrate_range(pfn, end_pfn); | |
1336 | if (!ret) { | |
1337 | drain = 1; | |
1338 | goto repeat; | |
1339 | } else { | |
1340 | if (ret < 0) | |
1341 | if (--retry_max == 0) | |
1342 | goto failed_removal; | |
1343 | yield(); | |
1344 | drain = 1; | |
1345 | goto repeat; | |
1346 | } | |
1347 | } | |
b3834be5 | 1348 | /* drain all zone's lru pagevec, this is asynchronous... */ |
0c0e6195 | 1349 | lru_add_drain_all(); |
0c0e6195 | 1350 | yield(); |
b3834be5 | 1351 | /* drain pcp pages, this is synchronous. */ |
9f8f2172 | 1352 | drain_all_pages(); |
0c0e6195 KH |
1353 | /* check again */ |
1354 | offlined_pages = check_pages_isolated(start_pfn, end_pfn); | |
1355 | if (offlined_pages < 0) { | |
1356 | ret = -EBUSY; | |
1357 | goto failed_removal; | |
1358 | } | |
1359 | printk(KERN_INFO "Offlined Pages %ld\n", offlined_pages); | |
b3834be5 | 1360 | /* Ok, all of our target is isolated. |
0c0e6195 KH |
1361 | We cannot do rollback at this point. */ |
1362 | offline_isolated_pages(start_pfn, end_pfn); | |
dbc0e4ce | 1363 | /* reset pagetype flags and makes migrate type to be MOVABLE */ |
0815f3d8 | 1364 | undo_isolate_page_range(start_pfn, end_pfn, MIGRATE_MOVABLE); |
0c0e6195 | 1365 | /* removal success */ |
9feedc9d | 1366 | zone->managed_pages -= offlined_pages; |
0c0e6195 KH |
1367 | zone->present_pages -= offlined_pages; |
1368 | zone->zone_pgdat->node_present_pages -= offlined_pages; | |
1369 | totalram_pages -= offlined_pages; | |
7b78d335 | 1370 | |
1b79acc9 KM |
1371 | init_per_zone_wmark_min(); |
1372 | ||
1e8537ba | 1373 | if (!populated_zone(zone)) { |
340175b7 | 1374 | zone_pcp_reset(zone); |
1e8537ba XQ |
1375 | mutex_lock(&zonelists_mutex); |
1376 | build_all_zonelists(NULL, NULL); | |
1377 | mutex_unlock(&zonelists_mutex); | |
1378 | } else | |
1379 | zone_pcp_update(zone); | |
340175b7 | 1380 | |
d9713679 LJ |
1381 | node_states_clear_node(node, &arg); |
1382 | if (arg.status_change_nid >= 0) | |
8fe23e05 | 1383 | kswapd_stop(node); |
bce7394a | 1384 | |
0c0e6195 KH |
1385 | vm_total_pages = nr_free_pagecache_pages(); |
1386 | writeback_set_ratelimit(); | |
7b78d335 YG |
1387 | |
1388 | memory_notify(MEM_OFFLINE, &arg); | |
20d6c96b | 1389 | unlock_memory_hotplug(); |
0c0e6195 KH |
1390 | return 0; |
1391 | ||
1392 | failed_removal: | |
a62e2f4f BH |
1393 | printk(KERN_INFO "memory offlining [mem %#010llx-%#010llx] failed\n", |
1394 | (unsigned long long) start_pfn << PAGE_SHIFT, | |
1395 | ((unsigned long long) end_pfn << PAGE_SHIFT) - 1); | |
7b78d335 | 1396 | memory_notify(MEM_CANCEL_OFFLINE, &arg); |
0c0e6195 | 1397 | /* pushback to free area */ |
0815f3d8 | 1398 | undo_isolate_page_range(start_pfn, end_pfn, MIGRATE_MOVABLE); |
7b78d335 | 1399 | |
6ad696d2 | 1400 | out: |
20d6c96b | 1401 | unlock_memory_hotplug(); |
0c0e6195 KH |
1402 | return ret; |
1403 | } | |
71088785 | 1404 | |
a16cee10 WC |
1405 | int offline_pages(unsigned long start_pfn, unsigned long nr_pages) |
1406 | { | |
1407 | return __offline_pages(start_pfn, start_pfn + nr_pages, 120 * HZ); | |
1408 | } | |
1409 | ||
bbc76be6 WC |
1410 | /** |
1411 | * walk_memory_range - walks through all mem sections in [start_pfn, end_pfn) | |
1412 | * @start_pfn: start pfn of the memory range | |
1413 | * @end_pfn: end pft of the memory range | |
1414 | * @arg: argument passed to func | |
1415 | * @func: callback for each memory section walked | |
1416 | * | |
1417 | * This function walks through all present mem sections in range | |
1418 | * [start_pfn, end_pfn) and call func on each mem section. | |
1419 | * | |
1420 | * Returns the return value of func. | |
1421 | */ | |
1422 | static int walk_memory_range(unsigned long start_pfn, unsigned long end_pfn, | |
1423 | void *arg, int (*func)(struct memory_block *, void *)) | |
71088785 | 1424 | { |
e90bdb7f WC |
1425 | struct memory_block *mem = NULL; |
1426 | struct mem_section *section; | |
e90bdb7f WC |
1427 | unsigned long pfn, section_nr; |
1428 | int ret; | |
e90bdb7f WC |
1429 | |
1430 | for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) { | |
1431 | section_nr = pfn_to_section_nr(pfn); | |
1432 | if (!present_section_nr(section_nr)) | |
1433 | continue; | |
1434 | ||
1435 | section = __nr_to_section(section_nr); | |
1436 | /* same memblock? */ | |
1437 | if (mem) | |
1438 | if ((section_nr >= mem->start_section_nr) && | |
1439 | (section_nr <= mem->end_section_nr)) | |
1440 | continue; | |
1441 | ||
1442 | mem = find_memory_block_hinted(section, mem); | |
1443 | if (!mem) | |
1444 | continue; | |
1445 | ||
bbc76be6 | 1446 | ret = func(mem, arg); |
e90bdb7f | 1447 | if (ret) { |
bbc76be6 WC |
1448 | kobject_put(&mem->dev.kobj); |
1449 | return ret; | |
e90bdb7f WC |
1450 | } |
1451 | } | |
1452 | ||
1453 | if (mem) | |
1454 | kobject_put(&mem->dev.kobj); | |
1455 | ||
bbc76be6 WC |
1456 | return 0; |
1457 | } | |
1458 | ||
1459 | /** | |
1460 | * offline_memory_block_cb - callback function for offlining memory block | |
1461 | * @mem: the memory block to be offlined | |
1462 | * @arg: buffer to hold error msg | |
1463 | * | |
1464 | * Always return 0, and put the error msg in arg if any. | |
1465 | */ | |
1466 | static int offline_memory_block_cb(struct memory_block *mem, void *arg) | |
1467 | { | |
1468 | int *ret = arg; | |
1469 | int error = offline_memory_block(mem); | |
1470 | ||
1471 | if (error != 0 && *ret == 0) | |
1472 | *ret = error; | |
1473 | ||
1474 | return 0; | |
1475 | } | |
1476 | ||
1477 | static int is_memblock_offlined_cb(struct memory_block *mem, void *arg) | |
1478 | { | |
1479 | int ret = !is_memblock_offlined(mem); | |
1480 | ||
1481 | if (unlikely(ret)) | |
1482 | pr_warn("removing memory fails, because memory " | |
1483 | "[%#010llx-%#010llx] is onlined\n", | |
1484 | PFN_PHYS(section_nr_to_pfn(mem->start_section_nr)), | |
1485 | PFN_PHYS(section_nr_to_pfn(mem->end_section_nr + 1))-1); | |
1486 | ||
1487 | return ret; | |
1488 | } | |
1489 | ||
46c66c4b | 1490 | int __ref remove_memory(u64 start, u64 size) |
bbc76be6 WC |
1491 | { |
1492 | unsigned long start_pfn, end_pfn; | |
1493 | int ret = 0; | |
1494 | int retry = 1; | |
1495 | ||
1496 | start_pfn = PFN_DOWN(start); | |
1497 | end_pfn = start_pfn + PFN_DOWN(size); | |
1498 | ||
1499 | /* | |
1500 | * When CONFIG_MEMCG is on, one memory block may be used by other | |
1501 | * blocks to store page cgroup when onlining pages. But we don't know | |
1502 | * in what order pages are onlined. So we iterate twice to offline | |
1503 | * memory: | |
1504 | * 1st iterate: offline every non primary memory block. | |
1505 | * 2nd iterate: offline primary (i.e. first added) memory block. | |
1506 | */ | |
1507 | repeat: | |
1508 | walk_memory_range(start_pfn, end_pfn, &ret, | |
1509 | offline_memory_block_cb); | |
1510 | if (ret) { | |
1511 | if (!retry) | |
1512 | return ret; | |
1513 | ||
1514 | retry = 0; | |
1515 | ret = 0; | |
993c1aad WC |
1516 | goto repeat; |
1517 | } | |
1518 | ||
6677e3ea YI |
1519 | lock_memory_hotplug(); |
1520 | ||
1521 | /* | |
1522 | * we have offlined all memory blocks like this: | |
1523 | * 1. lock memory hotplug | |
1524 | * 2. offline a memory block | |
1525 | * 3. unlock memory hotplug | |
1526 | * | |
1527 | * repeat step1-3 to offline the memory block. All memory blocks | |
1528 | * must be offlined before removing memory. But we don't hold the | |
1529 | * lock in the whole operation. So we should check whether all | |
1530 | * memory blocks are offlined. | |
1531 | */ | |
1532 | ||
bbc76be6 WC |
1533 | ret = walk_memory_range(start_pfn, end_pfn, NULL, |
1534 | is_memblock_offlined_cb); | |
1535 | if (ret) { | |
1536 | unlock_memory_hotplug(); | |
1537 | return ret; | |
6677e3ea YI |
1538 | } |
1539 | ||
46c66c4b YI |
1540 | /* remove memmap entry */ |
1541 | firmware_map_remove(start, start + size, "System RAM"); | |
1542 | ||
24d335ca WC |
1543 | arch_remove_memory(start, size); |
1544 | ||
6677e3ea YI |
1545 | unlock_memory_hotplug(); |
1546 | ||
e90bdb7f | 1547 | return 0; |
71088785 | 1548 | } |
48e94196 | 1549 | #else |
a16cee10 WC |
1550 | int offline_pages(unsigned long start_pfn, unsigned long nr_pages) |
1551 | { | |
1552 | return -EINVAL; | |
1553 | } | |
48e94196 KH |
1554 | int remove_memory(u64 start, u64 size) |
1555 | { | |
1556 | return -EINVAL; | |
1557 | } | |
0c0e6195 | 1558 | #endif /* CONFIG_MEMORY_HOTREMOVE */ |
71088785 | 1559 | EXPORT_SYMBOL_GPL(remove_memory); |