blob: d595606728f9c0d90b9b91e8c49971fbd26c1dc7 [file] [log] [blame]
Dave Hansen3947be12005-10-29 18:16:54 -07001/*
2 * linux/mm/memory_hotplug.c
3 *
4 * Copyright (C)
5 */
6
Dave Hansen3947be12005-10-29 18:16:54 -07007#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>
Paul Gortmakerb95f1b312011-10-16 02:01:52 -040014#include <linux/export.h>
Dave Hansen3947be12005-10-29 18:16:54 -070015#include <linux/pagevec.h>
Chandra Seetharaman2d1d43f2006-09-29 02:01:25 -070016#include <linux/writeback.h>
Dave Hansen3947be12005-10-29 18:16:54 -070017#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>
KAMEZAWA Hiroyuki0a547032006-06-27 02:53:35 -070024#include <linux/ioport.h>
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -070025#include <linux/delay.h>
26#include <linux/migrate.h>
27#include <linux/page-isolation.h>
Badari Pulavarty71088782008-10-18 20:25:58 -070028#include <linux/pfn.h>
Andi Kleen6ad696d2009-11-17 14:06:22 -080029#include <linux/suspend.h>
KOSAKI Motohiro6d9c2852009-12-14 17:58:11 -080030#include <linux/mm_inline.h>
akpm@linux-foundation.orgd96ae532010-03-05 13:41:58 -080031#include <linux/firmware-map.h>
Tang Chen60a5a192013-02-22 16:33:14 -080032#include <linux/stop_machine.h>
Dave Hansen3947be12005-10-29 18:16:54 -070033
34#include <asm/tlbflush.h>
35
Adrian Bunk1e5ad9a2008-04-28 20:40:08 +030036#include "internal.h"
37
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -070038/*
39 * online_page_callback contains pointer to current page onlining function.
40 * Initially it is generic_online_page(). If it is required it could be
41 * changed by calling set_online_page_callback() for callback registration
42 * and restore_online_page_callback() for generic callback restore.
43 */
44
45static void generic_online_page(struct page *page);
46
47static online_page_callback_t online_page_callback = generic_online_page;
48
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -080049DEFINE_MUTEX(mem_hotplug_mutex);
50
51void lock_memory_hotplug(void)
52{
53 mutex_lock(&mem_hotplug_mutex);
54
55 /* for exclusive hibernation if CONFIG_HIBERNATION=y */
56 lock_system_sleep();
57}
58
59void unlock_memory_hotplug(void)
60{
61 unlock_system_sleep();
62 mutex_unlock(&mem_hotplug_mutex);
63}
64
65
Keith Mannthey45e0b782006-09-30 23:27:09 -070066/* add this memory to iomem resource */
67static struct resource *register_memory_resource(u64 start, u64 size)
68{
69 struct resource *res;
70 res = kzalloc(sizeof(struct resource), GFP_KERNEL);
71 BUG_ON(!res);
72
73 res->name = "System RAM";
74 res->start = start;
75 res->end = start + size - 1;
Yasunori Goto887c3cb2007-11-14 16:59:20 -080076 res->flags = IORESOURCE_MEM | IORESOURCE_BUSY;
Keith Mannthey45e0b782006-09-30 23:27:09 -070077 if (request_resource(&iomem_resource, res) < 0) {
Toshi Kani4996eed2013-07-03 15:02:39 -070078 pr_debug("System RAM resource %pR cannot be added\n", res);
Keith Mannthey45e0b782006-09-30 23:27:09 -070079 kfree(res);
80 res = NULL;
81 }
82 return res;
83}
84
85static void release_memory_resource(struct resource *res)
86{
87 if (!res)
88 return;
89 release_resource(res);
90 kfree(res);
91 return;
92}
93
Keith Mannthey53947022006-09-30 23:27:08 -070094#ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
Yasuaki Ishimatsu46723bf2013-02-22 16:33:00 -080095void get_page_bootmem(unsigned long info, struct page *page,
96 unsigned long type)
Yasunori Goto04753272008-04-28 02:13:31 -070097{
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -080098 page->lru.next = (struct list_head *) type;
Yasunori Goto04753272008-04-28 02:13:31 -070099 SetPagePrivate(page);
100 set_page_private(page, info);
101 atomic_inc(&page->_count);
102}
103
Jiang Liu170a5a72013-07-03 15:03:17 -0700104void put_page_bootmem(struct page *page)
Yasunori Goto04753272008-04-28 02:13:31 -0700105{
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800106 unsigned long type;
Yasunori Goto04753272008-04-28 02:13:31 -0700107
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800108 type = (unsigned long) page->lru.next;
109 BUG_ON(type < MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE ||
110 type > MEMORY_HOTPLUG_MAX_BOOTMEM_TYPE);
Yasunori Goto04753272008-04-28 02:13:31 -0700111
112 if (atomic_dec_return(&page->_count) == 1) {
113 ClearPagePrivate(page);
114 set_page_private(page, 0);
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800115 INIT_LIST_HEAD(&page->lru);
Jiang Liu170a5a72013-07-03 15:03:17 -0700116 free_reserved_page(page);
Yasunori Goto04753272008-04-28 02:13:31 -0700117 }
Yasunori Goto04753272008-04-28 02:13:31 -0700118}
119
Yasuaki Ishimatsu46723bf2013-02-22 16:33:00 -0800120#ifdef CONFIG_HAVE_BOOTMEM_INFO_NODE
121#ifndef CONFIG_SPARSEMEM_VMEMMAP
Adrian Bunkd92bc312008-07-23 21:28:12 -0700122static void register_page_bootmem_info_section(unsigned long start_pfn)
Yasunori Goto04753272008-04-28 02:13:31 -0700123{
124 unsigned long *usemap, mapsize, section_nr, i;
125 struct mem_section *ms;
126 struct page *page, *memmap;
127
Yasunori Goto04753272008-04-28 02:13:31 -0700128 section_nr = pfn_to_section_nr(start_pfn);
129 ms = __nr_to_section(section_nr);
130
131 /* Get section's memmap address */
132 memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr);
133
134 /*
135 * Get page for the memmap's phys address
136 * XXX: need more consideration for sparse_vmemmap...
137 */
138 page = virt_to_page(memmap);
139 mapsize = sizeof(struct page) * PAGES_PER_SECTION;
140 mapsize = PAGE_ALIGN(mapsize) >> PAGE_SHIFT;
141
142 /* remember memmap's page */
143 for (i = 0; i < mapsize; i++, page++)
144 get_page_bootmem(section_nr, page, SECTION_INFO);
145
146 usemap = __nr_to_section(section_nr)->pageblock_flags;
147 page = virt_to_page(usemap);
148
149 mapsize = PAGE_ALIGN(usemap_size()) >> PAGE_SHIFT;
150
151 for (i = 0; i < mapsize; i++, page++)
Yasunori Gotoaf370fb2008-07-23 21:28:17 -0700152 get_page_bootmem(section_nr, page, MIX_SECTION_INFO);
Yasunori Goto04753272008-04-28 02:13:31 -0700153
154}
Yasuaki Ishimatsu46723bf2013-02-22 16:33:00 -0800155#else /* CONFIG_SPARSEMEM_VMEMMAP */
156static void register_page_bootmem_info_section(unsigned long start_pfn)
157{
158 unsigned long *usemap, mapsize, section_nr, i;
159 struct mem_section *ms;
160 struct page *page, *memmap;
161
162 if (!pfn_valid(start_pfn))
163 return;
164
165 section_nr = pfn_to_section_nr(start_pfn);
166 ms = __nr_to_section(section_nr);
167
168 memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr);
169
170 register_page_bootmem_memmap(section_nr, memmap, PAGES_PER_SECTION);
171
172 usemap = __nr_to_section(section_nr)->pageblock_flags;
173 page = virt_to_page(usemap);
174
175 mapsize = PAGE_ALIGN(usemap_size()) >> PAGE_SHIFT;
176
177 for (i = 0; i < mapsize; i++, page++)
178 get_page_bootmem(section_nr, page, MIX_SECTION_INFO);
179}
180#endif /* !CONFIG_SPARSEMEM_VMEMMAP */
Yasunori Goto04753272008-04-28 02:13:31 -0700181
182void register_page_bootmem_info_node(struct pglist_data *pgdat)
183{
184 unsigned long i, pfn, end_pfn, nr_pages;
185 int node = pgdat->node_id;
186 struct page *page;
187 struct zone *zone;
188
189 nr_pages = PAGE_ALIGN(sizeof(struct pglist_data)) >> PAGE_SHIFT;
190 page = virt_to_page(pgdat);
191
192 for (i = 0; i < nr_pages; i++, page++)
193 get_page_bootmem(node, page, NODE_INFO);
194
195 zone = &pgdat->node_zones[0];
196 for (; zone < pgdat->node_zones + MAX_NR_ZONES - 1; zone++) {
Xishi Qiu139c2d72013-09-11 14:21:46 -0700197 if (zone_is_initialized(zone)) {
Yasunori Goto04753272008-04-28 02:13:31 -0700198 nr_pages = zone->wait_table_hash_nr_entries
199 * sizeof(wait_queue_head_t);
200 nr_pages = PAGE_ALIGN(nr_pages) >> PAGE_SHIFT;
201 page = virt_to_page(zone->wait_table);
202
203 for (i = 0; i < nr_pages; i++, page++)
204 get_page_bootmem(node, page, NODE_INFO);
205 }
206 }
207
208 pfn = pgdat->node_start_pfn;
Cody P Schaferc1f19492013-02-22 16:35:32 -0800209 end_pfn = pgdat_end_pfn(pgdat);
Yasunori Goto04753272008-04-28 02:13:31 -0700210
Tang Chen7e9f5eb2013-07-08 16:00:23 -0700211 /* register section info */
qiuxishif14851a2012-09-17 14:09:24 -0700212 for (; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
213 /*
214 * Some platforms can assign the same pfn to multiple nodes - on
215 * node0 as well as nodeN. To avoid registering a pfn against
216 * multiple nodes we check that this pfn does not already
Tang Chen7e9f5eb2013-07-08 16:00:23 -0700217 * reside in some other nodes.
qiuxishif14851a2012-09-17 14:09:24 -0700218 */
219 if (pfn_valid(pfn) && (pfn_to_nid(pfn) == node))
220 register_page_bootmem_info_section(pfn);
221 }
Yasunori Goto04753272008-04-28 02:13:31 -0700222}
Yasuaki Ishimatsu46723bf2013-02-22 16:33:00 -0800223#endif /* CONFIG_HAVE_BOOTMEM_INFO_NODE */
Yasunori Goto04753272008-04-28 02:13:31 -0700224
Heiko Carstens76cdd582008-05-14 16:05:52 -0700225static void grow_zone_span(struct zone *zone, unsigned long start_pfn,
226 unsigned long end_pfn)
227{
228 unsigned long old_zone_end_pfn;
229
230 zone_span_writelock(zone);
231
Xishi Qiuc33bc312013-09-11 14:21:44 -0700232 old_zone_end_pfn = zone_end_pfn(zone);
Xishi Qiu8080fc02013-09-11 14:21:45 -0700233 if (zone_is_empty(zone) || start_pfn < zone->zone_start_pfn)
Heiko Carstens76cdd582008-05-14 16:05:52 -0700234 zone->zone_start_pfn = start_pfn;
235
236 zone->spanned_pages = max(old_zone_end_pfn, end_pfn) -
237 zone->zone_start_pfn;
238
239 zone_span_writeunlock(zone);
240}
241
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800242static void resize_zone(struct zone *zone, unsigned long start_pfn,
243 unsigned long end_pfn)
244{
245 zone_span_writelock(zone);
246
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800247 if (end_pfn - start_pfn) {
248 zone->zone_start_pfn = start_pfn;
249 zone->spanned_pages = end_pfn - start_pfn;
250 } else {
251 /*
252 * make it consist as free_area_init_core(),
253 * if spanned_pages = 0, then keep start_pfn = 0
254 */
255 zone->zone_start_pfn = 0;
256 zone->spanned_pages = 0;
257 }
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800258
259 zone_span_writeunlock(zone);
260}
261
262static void fix_zone_id(struct zone *zone, unsigned long start_pfn,
263 unsigned long end_pfn)
264{
265 enum zone_type zid = zone_idx(zone);
266 int nid = zone->zone_pgdat->node_id;
267 unsigned long pfn;
268
269 for (pfn = start_pfn; pfn < end_pfn; pfn++)
270 set_page_links(pfn_to_page(pfn), zid, nid, pfn);
271}
272
Cody P Schaferf6bbb782013-02-22 16:35:30 -0800273/* Can fail with -ENOMEM from allocating a wait table with vmalloc() or
274 * alloc_bootmem_node_nopanic() */
275static int __ref ensure_zone_is_initialized(struct zone *zone,
276 unsigned long start_pfn, unsigned long num_pages)
277{
278 if (!zone_is_initialized(zone))
279 return init_currently_empty_zone(zone, start_pfn, num_pages,
280 MEMMAP_HOTPLUG);
281 return 0;
282}
283
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800284static int __meminit move_pfn_range_left(struct zone *z1, struct zone *z2,
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800285 unsigned long start_pfn, unsigned long end_pfn)
286{
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800287 int ret;
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800288 unsigned long flags;
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800289 unsigned long z1_start_pfn;
290
Cody P Schafer64dd1b22013-02-22 16:35:31 -0800291 ret = ensure_zone_is_initialized(z1, start_pfn, end_pfn - start_pfn);
292 if (ret)
293 return ret;
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800294
295 pgdat_resize_lock(z1->zone_pgdat, &flags);
296
297 /* can't move pfns which are higher than @z2 */
Cody P Schafer108bcc92013-02-22 16:35:23 -0800298 if (end_pfn > zone_end_pfn(z2))
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800299 goto out_fail;
Jiang Liu834405c2013-07-03 15:03:04 -0700300 /* the move out part must be at the left most of @z2 */
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800301 if (start_pfn > z2->zone_start_pfn)
302 goto out_fail;
303 /* must included/overlap */
304 if (end_pfn <= z2->zone_start_pfn)
305 goto out_fail;
306
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800307 /* use start_pfn for z1's start_pfn if z1 is empty */
Xishi Qiu8080fc02013-09-11 14:21:45 -0700308 if (!zone_is_empty(z1))
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800309 z1_start_pfn = z1->zone_start_pfn;
310 else
311 z1_start_pfn = start_pfn;
312
313 resize_zone(z1, z1_start_pfn, end_pfn);
Cody P Schafer108bcc92013-02-22 16:35:23 -0800314 resize_zone(z2, end_pfn, zone_end_pfn(z2));
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800315
316 pgdat_resize_unlock(z1->zone_pgdat, &flags);
317
318 fix_zone_id(z1, start_pfn, end_pfn);
319
320 return 0;
321out_fail:
322 pgdat_resize_unlock(z1->zone_pgdat, &flags);
323 return -1;
324}
325
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800326static int __meminit move_pfn_range_right(struct zone *z1, struct zone *z2,
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800327 unsigned long start_pfn, unsigned long end_pfn)
328{
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800329 int ret;
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800330 unsigned long flags;
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800331 unsigned long z2_end_pfn;
332
Cody P Schafer64dd1b22013-02-22 16:35:31 -0800333 ret = ensure_zone_is_initialized(z2, start_pfn, end_pfn - start_pfn);
334 if (ret)
335 return ret;
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800336
337 pgdat_resize_lock(z1->zone_pgdat, &flags);
338
339 /* can't move pfns which are lower than @z1 */
340 if (z1->zone_start_pfn > start_pfn)
341 goto out_fail;
342 /* the move out part mast at the right most of @z1 */
Cody P Schafer108bcc92013-02-22 16:35:23 -0800343 if (zone_end_pfn(z1) > end_pfn)
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800344 goto out_fail;
345 /* must included/overlap */
Cody P Schafer108bcc92013-02-22 16:35:23 -0800346 if (start_pfn >= zone_end_pfn(z1))
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800347 goto out_fail;
348
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800349 /* use end_pfn for z2's end_pfn if z2 is empty */
Xishi Qiu8080fc02013-09-11 14:21:45 -0700350 if (!zone_is_empty(z2))
Cody P Schafer108bcc92013-02-22 16:35:23 -0800351 z2_end_pfn = zone_end_pfn(z2);
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800352 else
353 z2_end_pfn = end_pfn;
354
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800355 resize_zone(z1, z1->zone_start_pfn, start_pfn);
Lai Jiangshane455a9b2012-12-11 16:03:20 -0800356 resize_zone(z2, start_pfn, z2_end_pfn);
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800357
358 pgdat_resize_unlock(z1->zone_pgdat, &flags);
359
360 fix_zone_id(z2, start_pfn, end_pfn);
361
362 return 0;
363out_fail:
364 pgdat_resize_unlock(z1->zone_pgdat, &flags);
365 return -1;
366}
367
Heiko Carstens76cdd582008-05-14 16:05:52 -0700368static void grow_pgdat_span(struct pglist_data *pgdat, unsigned long start_pfn,
369 unsigned long end_pfn)
370{
371 unsigned long old_pgdat_end_pfn =
372 pgdat->node_start_pfn + pgdat->node_spanned_pages;
373
Tang Chen712cd382012-12-11 16:01:07 -0800374 if (!pgdat->node_spanned_pages || start_pfn < pgdat->node_start_pfn)
Heiko Carstens76cdd582008-05-14 16:05:52 -0700375 pgdat->node_start_pfn = start_pfn;
376
377 pgdat->node_spanned_pages = max(old_pgdat_end_pfn, end_pfn) -
378 pgdat->node_start_pfn;
379}
380
Al Viro31168482008-11-22 17:33:24 +0000381static int __meminit __add_zone(struct zone *zone, unsigned long phys_start_pfn)
Dave Hansen3947be12005-10-29 18:16:54 -0700382{
383 struct pglist_data *pgdat = zone->zone_pgdat;
384 int nr_pages = PAGES_PER_SECTION;
385 int nid = pgdat->node_id;
386 int zone_type;
Heiko Carstens76cdd582008-05-14 16:05:52 -0700387 unsigned long flags;
Cody P Schafer64dd1b22013-02-22 16:35:31 -0800388 int ret;
Dave Hansen3947be12005-10-29 18:16:54 -0700389
390 zone_type = zone - pgdat->node_zones;
Cody P Schafer64dd1b22013-02-22 16:35:31 -0800391 ret = ensure_zone_is_initialized(zone, phys_start_pfn, nr_pages);
392 if (ret)
393 return ret;
Heiko Carstens76cdd582008-05-14 16:05:52 -0700394
Heiko Carstens76cdd582008-05-14 16:05:52 -0700395 pgdat_resize_lock(zone->zone_pgdat, &flags);
396 grow_zone_span(zone, phys_start_pfn, phys_start_pfn + nr_pages);
397 grow_pgdat_span(zone->zone_pgdat, phys_start_pfn,
398 phys_start_pfn + nr_pages);
399 pgdat_resize_unlock(zone->zone_pgdat, &flags);
Dave Hansena2f3aa022007-01-10 23:15:30 -0800400 memmap_init_zone(nr_pages, nid, zone_type,
401 phys_start_pfn, MEMMAP_HOTPLUG);
Yasunori Goto718127c2006-06-23 02:03:10 -0700402 return 0;
Dave Hansen3947be12005-10-29 18:16:54 -0700403}
404
Gary Hadec04fc582009-01-06 14:39:14 -0800405static int __meminit __add_section(int nid, struct zone *zone,
406 unsigned long phys_start_pfn)
Dave Hansen3947be12005-10-29 18:16:54 -0700407{
Dave Hansen3947be12005-10-29 18:16:54 -0700408 int nr_pages = PAGES_PER_SECTION;
Dave Hansen3947be12005-10-29 18:16:54 -0700409 int ret;
410
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -0700411 if (pfn_valid(phys_start_pfn))
412 return -EEXIST;
413
Dave Hansen0b0acbe2005-10-29 18:16:55 -0700414 ret = sparse_add_one_section(zone, phys_start_pfn, nr_pages);
Dave Hansen3947be12005-10-29 18:16:54 -0700415
416 if (ret < 0)
417 return ret;
418
Yasunori Goto718127c2006-06-23 02:03:10 -0700419 ret = __add_zone(zone, phys_start_pfn);
420
421 if (ret < 0)
422 return ret;
423
Gary Hadec04fc582009-01-06 14:39:14 -0800424 return register_new_memory(nid, __pfn_to_section(phys_start_pfn));
Dave Hansen3947be12005-10-29 18:16:54 -0700425}
426
David Rientjes4edd7ce2013-04-29 15:08:22 -0700427/*
428 * Reasonably generic function for adding memory. It is
429 * expected that archs that support memory hotplug will
430 * call this function after deciding the zone to which to
431 * add the new pages.
432 */
433int __ref __add_pages(int nid, struct zone *zone, unsigned long phys_start_pfn,
434 unsigned long nr_pages)
435{
436 unsigned long i;
437 int err = 0;
438 int start_sec, end_sec;
439 /* during initialize mem_map, align hot-added range to section */
440 start_sec = pfn_to_section_nr(phys_start_pfn);
441 end_sec = pfn_to_section_nr(phys_start_pfn + nr_pages - 1);
442
443 for (i = start_sec; i <= end_sec; i++) {
444 err = __add_section(nid, zone, i << PFN_SECTION_SHIFT);
445
446 /*
447 * EEXIST is finally dealt with by ioresource collision
448 * check. see add_memory() => register_memory_resource()
449 * Warning will be printed if there is collision.
450 */
451 if (err && (err != -EEXIST))
452 break;
453 err = 0;
454 }
455
456 return err;
457}
458EXPORT_SYMBOL_GPL(__add_pages);
459
460#ifdef CONFIG_MEMORY_HOTREMOVE
Yasuaki Ishimatsu815121d2013-02-22 16:33:12 -0800461/* find the smallest valid pfn in the range [start_pfn, end_pfn) */
462static int find_smallest_section_pfn(int nid, struct zone *zone,
463 unsigned long start_pfn,
464 unsigned long end_pfn)
465{
466 struct mem_section *ms;
467
468 for (; start_pfn < end_pfn; start_pfn += PAGES_PER_SECTION) {
469 ms = __pfn_to_section(start_pfn);
470
471 if (unlikely(!valid_section(ms)))
472 continue;
473
474 if (unlikely(pfn_to_nid(start_pfn) != nid))
475 continue;
476
477 if (zone && zone != page_zone(pfn_to_page(start_pfn)))
478 continue;
479
480 return start_pfn;
481 }
482
483 return 0;
484}
485
486/* find the biggest valid pfn in the range [start_pfn, end_pfn). */
487static int find_biggest_section_pfn(int nid, struct zone *zone,
488 unsigned long start_pfn,
489 unsigned long end_pfn)
490{
491 struct mem_section *ms;
492 unsigned long pfn;
493
494 /* pfn is the end pfn of a memory section. */
495 pfn = end_pfn - 1;
496 for (; pfn >= start_pfn; pfn -= PAGES_PER_SECTION) {
497 ms = __pfn_to_section(pfn);
498
499 if (unlikely(!valid_section(ms)))
500 continue;
501
502 if (unlikely(pfn_to_nid(pfn) != nid))
503 continue;
504
505 if (zone && zone != page_zone(pfn_to_page(pfn)))
506 continue;
507
508 return pfn;
509 }
510
511 return 0;
512}
513
514static void shrink_zone_span(struct zone *zone, unsigned long start_pfn,
515 unsigned long end_pfn)
516{
Xishi Qiuc33bc312013-09-11 14:21:44 -0700517 unsigned long zone_start_pfn = zone->zone_start_pfn;
518 unsigned long z = zone_end_pfn(zone); /* zone_end_pfn namespace clash */
519 unsigned long zone_end_pfn = z;
Yasuaki Ishimatsu815121d2013-02-22 16:33:12 -0800520 unsigned long pfn;
521 struct mem_section *ms;
522 int nid = zone_to_nid(zone);
523
524 zone_span_writelock(zone);
525 if (zone_start_pfn == start_pfn) {
526 /*
527 * If the section is smallest section in the zone, it need
528 * shrink zone->zone_start_pfn and zone->zone_spanned_pages.
529 * In this case, we find second smallest valid mem_section
530 * for shrinking zone.
531 */
532 pfn = find_smallest_section_pfn(nid, zone, end_pfn,
533 zone_end_pfn);
534 if (pfn) {
535 zone->zone_start_pfn = pfn;
536 zone->spanned_pages = zone_end_pfn - pfn;
537 }
538 } else if (zone_end_pfn == end_pfn) {
539 /*
540 * If the section is biggest section in the zone, it need
541 * shrink zone->spanned_pages.
542 * In this case, we find second biggest valid mem_section for
543 * shrinking zone.
544 */
545 pfn = find_biggest_section_pfn(nid, zone, zone_start_pfn,
546 start_pfn);
547 if (pfn)
548 zone->spanned_pages = pfn - zone_start_pfn + 1;
549 }
550
551 /*
552 * The section is not biggest or smallest mem_section in the zone, it
553 * only creates a hole in the zone. So in this case, we need not
554 * change the zone. But perhaps, the zone has only hole data. Thus
555 * it check the zone has only hole or not.
556 */
557 pfn = zone_start_pfn;
558 for (; pfn < zone_end_pfn; pfn += PAGES_PER_SECTION) {
559 ms = __pfn_to_section(pfn);
560
561 if (unlikely(!valid_section(ms)))
562 continue;
563
564 if (page_zone(pfn_to_page(pfn)) != zone)
565 continue;
566
567 /* If the section is current section, it continues the loop */
568 if (start_pfn == pfn)
569 continue;
570
571 /* If we find valid section, we have nothing to do */
572 zone_span_writeunlock(zone);
573 return;
574 }
575
576 /* The zone has no valid section */
577 zone->zone_start_pfn = 0;
578 zone->spanned_pages = 0;
579 zone_span_writeunlock(zone);
580}
581
582static void shrink_pgdat_span(struct pglist_data *pgdat,
583 unsigned long start_pfn, unsigned long end_pfn)
584{
585 unsigned long pgdat_start_pfn = pgdat->node_start_pfn;
586 unsigned long pgdat_end_pfn =
587 pgdat->node_start_pfn + pgdat->node_spanned_pages;
588 unsigned long pfn;
589 struct mem_section *ms;
590 int nid = pgdat->node_id;
591
592 if (pgdat_start_pfn == start_pfn) {
593 /*
594 * If the section is smallest section in the pgdat, it need
595 * shrink pgdat->node_start_pfn and pgdat->node_spanned_pages.
596 * In this case, we find second smallest valid mem_section
597 * for shrinking zone.
598 */
599 pfn = find_smallest_section_pfn(nid, NULL, end_pfn,
600 pgdat_end_pfn);
601 if (pfn) {
602 pgdat->node_start_pfn = pfn;
603 pgdat->node_spanned_pages = pgdat_end_pfn - pfn;
604 }
605 } else if (pgdat_end_pfn == end_pfn) {
606 /*
607 * If the section is biggest section in the pgdat, it need
608 * shrink pgdat->node_spanned_pages.
609 * In this case, we find second biggest valid mem_section for
610 * shrinking zone.
611 */
612 pfn = find_biggest_section_pfn(nid, NULL, pgdat_start_pfn,
613 start_pfn);
614 if (pfn)
615 pgdat->node_spanned_pages = pfn - pgdat_start_pfn + 1;
616 }
617
618 /*
619 * If the section is not biggest or smallest mem_section in the pgdat,
620 * it only creates a hole in the pgdat. So in this case, we need not
621 * change the pgdat.
622 * But perhaps, the pgdat has only hole data. Thus it check the pgdat
623 * has only hole or not.
624 */
625 pfn = pgdat_start_pfn;
626 for (; pfn < pgdat_end_pfn; pfn += PAGES_PER_SECTION) {
627 ms = __pfn_to_section(pfn);
628
629 if (unlikely(!valid_section(ms)))
630 continue;
631
632 if (pfn_to_nid(pfn) != nid)
633 continue;
634
635 /* If the section is current section, it continues the loop */
636 if (start_pfn == pfn)
637 continue;
638
639 /* If we find valid section, we have nothing to do */
640 return;
641 }
642
643 /* The pgdat has no valid section */
644 pgdat->node_start_pfn = 0;
645 pgdat->node_spanned_pages = 0;
646}
647
648static void __remove_zone(struct zone *zone, unsigned long start_pfn)
649{
650 struct pglist_data *pgdat = zone->zone_pgdat;
651 int nr_pages = PAGES_PER_SECTION;
652 int zone_type;
653 unsigned long flags;
654
655 zone_type = zone - pgdat->node_zones;
656
657 pgdat_resize_lock(zone->zone_pgdat, &flags);
658 shrink_zone_span(zone, start_pfn, start_pfn + nr_pages);
659 shrink_pgdat_span(pgdat, start_pfn, start_pfn + nr_pages);
660 pgdat_resize_unlock(zone->zone_pgdat, &flags);
661}
662
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700663static int __remove_section(struct zone *zone, struct mem_section *ms)
664{
Yasuaki Ishimatsu815121d2013-02-22 16:33:12 -0800665 unsigned long start_pfn;
666 int scn_nr;
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700667 int ret = -EINVAL;
668
669 if (!valid_section(ms))
670 return ret;
671
672 ret = unregister_memory_section(ms);
673 if (ret)
674 return ret;
675
Yasuaki Ishimatsu815121d2013-02-22 16:33:12 -0800676 scn_nr = __section_nr(ms);
677 start_pfn = section_nr_to_pfn(scn_nr);
678 __remove_zone(zone, start_pfn);
679
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700680 sparse_remove_one_section(zone, ms);
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700681 return 0;
682}
683
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700684/**
685 * __remove_pages() - remove sections of pages from a zone
686 * @zone: zone from which pages need to be removed
687 * @phys_start_pfn: starting pageframe (must be aligned to start of a section)
688 * @nr_pages: number of pages to remove (must be multiple of section size)
689 *
690 * Generic helper function to remove section mappings and sysfs entries
691 * for the section of the memory we are removing. Caller needs to make
692 * sure that pages are marked reserved and zones are adjust properly by
693 * calling offline_pages().
694 */
695int __remove_pages(struct zone *zone, unsigned long phys_start_pfn,
696 unsigned long nr_pages)
697{
Toshi Kanife74ebb2013-04-29 15:08:20 -0700698 unsigned long i;
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700699 int sections_to_remove;
Toshi Kanife74ebb2013-04-29 15:08:20 -0700700 resource_size_t start, size;
701 int ret = 0;
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700702
703 /*
704 * We can only remove entire sections
705 */
706 BUG_ON(phys_start_pfn & ~PAGE_SECTION_MASK);
707 BUG_ON(nr_pages % PAGES_PER_SECTION);
708
Toshi Kanife74ebb2013-04-29 15:08:20 -0700709 start = phys_start_pfn << PAGE_SHIFT;
710 size = nr_pages * PAGE_SIZE;
711 ret = release_mem_region_adjustable(&iomem_resource, start, size);
Randy Dunlap348f9f02013-05-24 15:55:30 -0700712 if (ret) {
713 resource_size_t endres = start + size - 1;
714
715 pr_warn("Unable to release resource <%pa-%pa> (%d)\n",
716 &start, &endres, ret);
717 }
Yasuaki Ishimatsud760afd2012-10-08 16:34:14 -0700718
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700719 sections_to_remove = nr_pages / PAGES_PER_SECTION;
720 for (i = 0; i < sections_to_remove; i++) {
721 unsigned long pfn = phys_start_pfn + i*PAGES_PER_SECTION;
722 ret = __remove_section(zone, __pfn_to_section(pfn));
723 if (ret)
724 break;
725 }
726 return ret;
727}
728EXPORT_SYMBOL_GPL(__remove_pages);
David Rientjes4edd7ce2013-04-29 15:08:22 -0700729#endif /* CONFIG_MEMORY_HOTREMOVE */
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700730
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700731int set_online_page_callback(online_page_callback_t callback)
732{
733 int rc = -EINVAL;
734
735 lock_memory_hotplug();
736
737 if (online_page_callback == generic_online_page) {
738 online_page_callback = callback;
739 rc = 0;
740 }
741
742 unlock_memory_hotplug();
743
744 return rc;
745}
746EXPORT_SYMBOL_GPL(set_online_page_callback);
747
748int restore_online_page_callback(online_page_callback_t callback)
749{
750 int rc = -EINVAL;
751
752 lock_memory_hotplug();
753
754 if (online_page_callback == callback) {
755 online_page_callback = generic_online_page;
756 rc = 0;
757 }
758
759 unlock_memory_hotplug();
760
761 return rc;
762}
763EXPORT_SYMBOL_GPL(restore_online_page_callback);
764
765void __online_page_set_limits(struct page *page)
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700766{
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700767}
768EXPORT_SYMBOL_GPL(__online_page_set_limits);
769
770void __online_page_increment_counters(struct page *page)
771{
Jiang Liu3dcc0572013-07-03 15:03:21 -0700772 adjust_managed_page_count(page, 1);
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700773}
774EXPORT_SYMBOL_GPL(__online_page_increment_counters);
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700775
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700776void __online_page_free(struct page *page)
777{
Jiang Liu3dcc0572013-07-03 15:03:21 -0700778 __free_reserved_page(page);
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700779}
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700780EXPORT_SYMBOL_GPL(__online_page_free);
781
782static void generic_online_page(struct page *page)
783{
784 __online_page_set_limits(page);
785 __online_page_increment_counters(page);
786 __online_page_free(page);
787}
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700788
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700789static int online_pages_range(unsigned long start_pfn, unsigned long nr_pages,
790 void *arg)
Dave Hansen3947be12005-10-29 18:16:54 -0700791{
792 unsigned long i;
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700793 unsigned long onlined_pages = *(unsigned long *)arg;
794 struct page *page;
795 if (PageReserved(pfn_to_page(start_pfn)))
796 for (i = 0; i < nr_pages; i++) {
797 page = pfn_to_page(start_pfn + i);
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700798 (*online_page_callback)(page);
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700799 onlined_pages++;
800 }
801 *(unsigned long *)arg = onlined_pages;
802 return 0;
803}
804
Lai Jiangshan09285af2012-12-12 13:52:04 -0800805#ifdef CONFIG_MOVABLE_NODE
Tang Chen79a4dce2012-12-18 14:23:24 -0800806/*
807 * When CONFIG_MOVABLE_NODE, we permit onlining of a node which doesn't have
808 * normal memory.
809 */
Lai Jiangshan09285af2012-12-12 13:52:04 -0800810static bool can_online_high_movable(struct zone *zone)
811{
812 return true;
813}
Tang Chen79a4dce2012-12-18 14:23:24 -0800814#else /* CONFIG_MOVABLE_NODE */
Lai Jiangshan74d42d82012-12-11 16:03:23 -0800815/* ensure every online node has NORMAL memory */
816static bool can_online_high_movable(struct zone *zone)
817{
818 return node_state(zone_to_nid(zone), N_NORMAL_MEMORY);
819}
Tang Chen79a4dce2012-12-18 14:23:24 -0800820#endif /* CONFIG_MOVABLE_NODE */
Lai Jiangshan74d42d82012-12-11 16:03:23 -0800821
Lai Jiangshand9713672012-12-11 16:01:03 -0800822/* check which state of node_states will be changed when online memory */
823static void node_states_check_changes_online(unsigned long nr_pages,
824 struct zone *zone, struct memory_notify *arg)
825{
826 int nid = zone_to_nid(zone);
827 enum zone_type zone_last = ZONE_NORMAL;
828
829 /*
Lai Jiangshan6715ddf2012-12-12 13:51:49 -0800830 * If we have HIGHMEM or movable node, node_states[N_NORMAL_MEMORY]
831 * contains nodes which have zones of 0...ZONE_NORMAL,
832 * set zone_last to ZONE_NORMAL.
Lai Jiangshand9713672012-12-11 16:01:03 -0800833 *
Lai Jiangshan6715ddf2012-12-12 13:51:49 -0800834 * If we don't have HIGHMEM nor movable node,
835 * node_states[N_NORMAL_MEMORY] contains nodes which have zones of
836 * 0...ZONE_MOVABLE, set zone_last to ZONE_MOVABLE.
Lai Jiangshand9713672012-12-11 16:01:03 -0800837 */
Lai Jiangshan6715ddf2012-12-12 13:51:49 -0800838 if (N_MEMORY == N_NORMAL_MEMORY)
Lai Jiangshand9713672012-12-11 16:01:03 -0800839 zone_last = ZONE_MOVABLE;
840
841 /*
842 * if the memory to be online is in a zone of 0...zone_last, and
843 * the zones of 0...zone_last don't have memory before online, we will
844 * need to set the node to node_states[N_NORMAL_MEMORY] after
845 * the memory is online.
846 */
847 if (zone_idx(zone) <= zone_last && !node_state(nid, N_NORMAL_MEMORY))
848 arg->status_change_nid_normal = nid;
849 else
850 arg->status_change_nid_normal = -1;
851
Lai Jiangshan6715ddf2012-12-12 13:51:49 -0800852#ifdef CONFIG_HIGHMEM
853 /*
854 * If we have movable node, node_states[N_HIGH_MEMORY]
855 * contains nodes which have zones of 0...ZONE_HIGHMEM,
856 * set zone_last to ZONE_HIGHMEM.
857 *
858 * If we don't have movable node, node_states[N_NORMAL_MEMORY]
859 * contains nodes which have zones of 0...ZONE_MOVABLE,
860 * set zone_last to ZONE_MOVABLE.
861 */
862 zone_last = ZONE_HIGHMEM;
863 if (N_MEMORY == N_HIGH_MEMORY)
864 zone_last = ZONE_MOVABLE;
865
866 if (zone_idx(zone) <= zone_last && !node_state(nid, N_HIGH_MEMORY))
867 arg->status_change_nid_high = nid;
868 else
869 arg->status_change_nid_high = -1;
870#else
871 arg->status_change_nid_high = arg->status_change_nid_normal;
872#endif
873
Lai Jiangshand9713672012-12-11 16:01:03 -0800874 /*
875 * if the node don't have memory befor online, we will need to
Lai Jiangshan6715ddf2012-12-12 13:51:49 -0800876 * set the node to node_states[N_MEMORY] after the memory
Lai Jiangshand9713672012-12-11 16:01:03 -0800877 * is online.
878 */
Lai Jiangshan6715ddf2012-12-12 13:51:49 -0800879 if (!node_state(nid, N_MEMORY))
Lai Jiangshand9713672012-12-11 16:01:03 -0800880 arg->status_change_nid = nid;
881 else
882 arg->status_change_nid = -1;
883}
884
885static void node_states_set_node(int node, struct memory_notify *arg)
886{
887 if (arg->status_change_nid_normal >= 0)
888 node_set_state(node, N_NORMAL_MEMORY);
889
Lai Jiangshan6715ddf2012-12-12 13:51:49 -0800890 if (arg->status_change_nid_high >= 0)
891 node_set_state(node, N_HIGH_MEMORY);
892
893 node_set_state(node, N_MEMORY);
Lai Jiangshand9713672012-12-11 16:01:03 -0800894}
895
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700896
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800897int __ref online_pages(unsigned long pfn, unsigned long nr_pages, int online_type)
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700898{
Cody P Schaferaa472282013-07-03 15:02:10 -0700899 unsigned long flags;
Dave Hansen3947be12005-10-29 18:16:54 -0700900 unsigned long onlined_pages = 0;
901 struct zone *zone;
Yasunori Goto68113782006-06-23 02:03:11 -0700902 int need_zonelists_rebuild = 0;
Yasunori Goto7b78d332007-10-21 16:41:36 -0700903 int nid;
904 int ret;
905 struct memory_notify arg;
Dave Hansen3947be12005-10-29 18:16:54 -0700906
KAMEZAWA Hiroyuki925268a2011-01-11 16:44:01 +0900907 lock_memory_hotplug();
Lai Jiangshand9713672012-12-11 16:01:03 -0800908 /*
909 * This doesn't need a lock to do pfn_to_page().
910 * The section can't be removed here because of the
911 * memory_block->state_mutex.
912 */
913 zone = page_zone(pfn_to_page(pfn));
914
Lai Jiangshan74d42d82012-12-11 16:03:23 -0800915 if ((zone_idx(zone) > ZONE_NORMAL || online_type == ONLINE_MOVABLE) &&
916 !can_online_high_movable(zone)) {
917 unlock_memory_hotplug();
Toshi Kani0a1be152013-07-08 16:00:41 -0700918 return -EINVAL;
Lai Jiangshan74d42d82012-12-11 16:03:23 -0800919 }
920
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800921 if (online_type == ONLINE_KERNEL && zone_idx(zone) == ZONE_MOVABLE) {
922 if (move_pfn_range_left(zone - 1, zone, pfn, pfn + nr_pages)) {
923 unlock_memory_hotplug();
Toshi Kani0a1be152013-07-08 16:00:41 -0700924 return -EINVAL;
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800925 }
926 }
927 if (online_type == ONLINE_MOVABLE && zone_idx(zone) == ZONE_MOVABLE - 1) {
928 if (move_pfn_range_right(zone, zone + 1, pfn, pfn + nr_pages)) {
929 unlock_memory_hotplug();
Toshi Kani0a1be152013-07-08 16:00:41 -0700930 return -EINVAL;
Lai Jiangshan511c2ab2012-12-11 16:03:16 -0800931 }
932 }
933
934 /* Previous code may changed the zone of the pfn range */
935 zone = page_zone(pfn_to_page(pfn));
936
Yasunori Goto7b78d332007-10-21 16:41:36 -0700937 arg.start_pfn = pfn;
938 arg.nr_pages = nr_pages;
Lai Jiangshand9713672012-12-11 16:01:03 -0800939 node_states_check_changes_online(nr_pages, zone, &arg);
Yasunori Goto7b78d332007-10-21 16:41:36 -0700940
941 nid = page_to_nid(pfn_to_page(pfn));
Yasunori Goto7b78d332007-10-21 16:41:36 -0700942
943 ret = memory_notify(MEM_GOING_ONLINE, &arg);
944 ret = notifier_to_errno(ret);
945 if (ret) {
946 memory_notify(MEM_CANCEL_ONLINE, &arg);
KAMEZAWA Hiroyuki925268a2011-01-11 16:44:01 +0900947 unlock_memory_hotplug();
Yasunori Goto7b78d332007-10-21 16:41:36 -0700948 return ret;
949 }
Dave Hansen3947be12005-10-29 18:16:54 -0700950 /*
Yasunori Goto68113782006-06-23 02:03:11 -0700951 * If this zone is not populated, then it is not in zonelist.
952 * This means the page allocator ignores this zone.
953 * So, zonelist must be updated after online.
954 */
Haicheng Li4eaf3f62010-05-24 14:32:52 -0700955 mutex_lock(&zonelists_mutex);
Wen Congyang6dcd73d2012-12-11 16:01:01 -0800956 if (!populated_zone(zone)) {
Yasunori Goto68113782006-06-23 02:03:11 -0700957 need_zonelists_rebuild = 1;
Wen Congyang6dcd73d2012-12-11 16:01:01 -0800958 build_all_zonelists(NULL, zone);
959 }
Yasunori Goto68113782006-06-23 02:03:11 -0700960
KAMEZAWA Hiroyuki908eedc2009-09-22 16:45:46 -0700961 ret = walk_system_ram_range(pfn, nr_pages, &onlined_pages,
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700962 online_pages_range);
Geoff Levandfd8a4222008-05-14 16:05:50 -0700963 if (ret) {
Wen Congyang6dcd73d2012-12-11 16:01:01 -0800964 if (need_zonelists_rebuild)
965 zone_pcp_reset(zone);
Haicheng Li4eaf3f62010-05-24 14:32:52 -0700966 mutex_unlock(&zonelists_mutex);
Bjorn Helgaasa62e2f42012-05-29 15:06:30 -0700967 printk(KERN_DEBUG "online_pages [mem %#010llx-%#010llx] failed\n",
968 (unsigned long long) pfn << PAGE_SHIFT,
969 (((unsigned long long) pfn + nr_pages)
970 << PAGE_SHIFT) - 1);
Geoff Levandfd8a4222008-05-14 16:05:50 -0700971 memory_notify(MEM_CANCEL_ONLINE, &arg);
KAMEZAWA Hiroyuki925268a2011-01-11 16:44:01 +0900972 unlock_memory_hotplug();
Geoff Levandfd8a4222008-05-14 16:05:50 -0700973 return ret;
974 }
975
Dave Hansen3947be12005-10-29 18:16:54 -0700976 zone->present_pages += onlined_pages;
Cody P Schaferaa472282013-07-03 15:02:10 -0700977
978 pgdat_resize_lock(zone->zone_pgdat, &flags);
Yasunori Gotof2937be2006-03-09 17:33:51 -0800979 zone->zone_pgdat->node_present_pages += onlined_pages;
Cody P Schaferaa472282013-07-03 15:02:10 -0700980 pgdat_resize_unlock(zone->zone_pgdat, &flags);
981
Jiang Liu08dff7b2012-07-31 16:43:30 -0700982 if (onlined_pages) {
Lai Jiangshand9713672012-12-11 16:01:03 -0800983 node_states_set_node(zone_to_nid(zone), &arg);
Jiang Liu08dff7b2012-07-31 16:43:30 -0700984 if (need_zonelists_rebuild)
Wen Congyang6dcd73d2012-12-11 16:01:01 -0800985 build_all_zonelists(NULL, NULL);
Jiang Liu08dff7b2012-07-31 16:43:30 -0700986 else
987 zone_pcp_update(zone);
988 }
Dave Hansen3947be12005-10-29 18:16:54 -0700989
Haicheng Li4eaf3f62010-05-24 14:32:52 -0700990 mutex_unlock(&zonelists_mutex);
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -0700991
992 init_per_zone_wmark_min();
993
Jiang Liu08dff7b2012-07-31 16:43:30 -0700994 if (onlined_pages)
Christoph Lameter7ea15302007-10-16 01:25:29 -0700995 kswapd_run(zone_to_nid(zone));
Dave Hansen61b13992005-10-29 18:16:56 -0700996
Haicheng Li1f522502010-05-24 14:32:51 -0700997 vm_total_pages = nr_free_pagecache_pages();
Kent Liu2f7f24e2008-07-23 21:28:18 -0700998
Chandra Seetharaman2d1d43f2006-09-29 02:01:25 -0700999 writeback_set_ratelimit();
Yasunori Goto7b78d332007-10-21 16:41:36 -07001000
1001 if (onlined_pages)
1002 memory_notify(MEM_ONLINE, &arg);
KAMEZAWA Hiroyuki925268a2011-01-11 16:44:01 +09001003 unlock_memory_hotplug();
Yasunori Goto7b78d332007-10-21 16:41:36 -07001004
Dave Hansen3947be12005-10-29 18:16:54 -07001005 return 0;
1006}
Keith Mannthey53947022006-09-30 23:27:08 -07001007#endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */
Yasunori Gotobc02af92006-06-27 02:53:30 -07001008
Hidetoshi Setoe1319332009-11-17 14:06:18 -08001009/* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */
1010static pg_data_t __ref *hotadd_new_pgdat(int nid, u64 start)
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001011{
1012 struct pglist_data *pgdat;
1013 unsigned long zones_size[MAX_NR_ZONES] = {0};
1014 unsigned long zholes_size[MAX_NR_ZONES] = {0};
1015 unsigned long start_pfn = start >> PAGE_SHIFT;
1016
Tang Chena1e565a2013-02-22 16:33:18 -08001017 pgdat = NODE_DATA(nid);
1018 if (!pgdat) {
1019 pgdat = arch_alloc_nodedata(nid);
1020 if (!pgdat)
1021 return NULL;
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001022
Tang Chena1e565a2013-02-22 16:33:18 -08001023 arch_refresh_nodedata(nid, pgdat);
1024 }
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001025
1026 /* we can use NODE_DATA(nid) from here */
1027
1028 /* init node's zones as empty zones, we don't have any present pages.*/
Johannes Weiner9109fb72008-07-23 21:27:20 -07001029 free_area_init_node(nid, zones_size, start_pfn, zholes_size);
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001030
KAMEZAWA Hiroyuki959ecc42011-06-15 15:08:38 -07001031 /*
1032 * The node we allocated has no zone fallback lists. For avoiding
1033 * to access not-initialized zonelist, build here.
1034 */
David Rientjesf957db42011-06-22 18:13:04 -07001035 mutex_lock(&zonelists_mutex);
Jiang Liu9adb62a2012-07-31 16:43:28 -07001036 build_all_zonelists(pgdat, NULL);
David Rientjesf957db42011-06-22 18:13:04 -07001037 mutex_unlock(&zonelists_mutex);
KAMEZAWA Hiroyuki959ecc42011-06-15 15:08:38 -07001038
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001039 return pgdat;
1040}
1041
1042static void rollback_node_hotadd(int nid, pg_data_t *pgdat)
1043{
1044 arch_refresh_nodedata(nid, NULL);
1045 arch_free_nodedata(pgdat);
1046 return;
1047}
1048
KAMEZAWA Hiroyuki0a547032006-06-27 02:53:35 -07001049
minskey guocf234222010-05-24 14:32:41 -07001050/*
1051 * called by cpu_up() to online a node without onlined memory.
1052 */
1053int mem_online_node(int nid)
1054{
1055 pg_data_t *pgdat;
1056 int ret;
1057
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -08001058 lock_memory_hotplug();
minskey guocf234222010-05-24 14:32:41 -07001059 pgdat = hotadd_new_pgdat(nid, 0);
David Rientjes7553e8f2011-06-22 18:13:01 -07001060 if (!pgdat) {
minskey guocf234222010-05-24 14:32:41 -07001061 ret = -ENOMEM;
1062 goto out;
1063 }
1064 node_set_online(nid);
1065 ret = register_one_node(nid);
1066 BUG_ON(ret);
1067
1068out:
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -08001069 unlock_memory_hotplug();
minskey guocf234222010-05-24 14:32:41 -07001070 return ret;
1071}
1072
Toshi Kani27356f52013-09-11 14:21:49 -07001073static int check_hotplug_memory_range(u64 start, u64 size)
1074{
1075 u64 start_pfn = start >> PAGE_SHIFT;
1076 u64 nr_pages = size >> PAGE_SHIFT;
1077
1078 /* Memory range must be aligned with section */
1079 if ((start_pfn & ~PAGE_SECTION_MASK) ||
1080 (nr_pages % PAGES_PER_SECTION) || (!nr_pages)) {
1081 pr_err("Section-unaligned hotplug range: start 0x%llx, size 0x%llx\n",
1082 (unsigned long long)start,
1083 (unsigned long long)size);
1084 return -EINVAL;
1085 }
1086
1087 return 0;
1088}
1089
Al Viro31168482008-11-22 17:33:24 +00001090/* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */
1091int __ref add_memory(int nid, u64 start, u64 size)
Yasunori Gotobc02af92006-06-27 02:53:30 -07001092{
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001093 pg_data_t *pgdat = NULL;
Tang Chena1e565a2013-02-22 16:33:18 -08001094 bool new_pgdat;
1095 bool new_node;
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -07001096 struct resource *res;
Yasunori Gotobc02af92006-06-27 02:53:30 -07001097 int ret;
1098
Toshi Kani27356f52013-09-11 14:21:49 -07001099 ret = check_hotplug_memory_range(start, size);
1100 if (ret)
1101 return ret;
1102
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -08001103 lock_memory_hotplug();
Andi Kleen6ad696d2009-11-17 14:06:22 -08001104
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -07001105 res = register_memory_resource(start, size);
Andi Kleen6ad696d2009-11-17 14:06:22 -08001106 ret = -EEXIST;
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -07001107 if (!res)
Andi Kleen6ad696d2009-11-17 14:06:22 -08001108 goto out;
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -07001109
Tang Chena1e565a2013-02-22 16:33:18 -08001110 { /* Stupid hack to suppress address-never-null warning */
1111 void *p = NODE_DATA(nid);
1112 new_pgdat = !p;
1113 }
1114 new_node = !node_online(nid);
1115 if (new_node) {
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001116 pgdat = hotadd_new_pgdat(nid, start);
Andi Kleen6ad696d2009-11-17 14:06:22 -08001117 ret = -ENOMEM;
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001118 if (!pgdat)
Wen Congyang41b9e2d2012-07-11 14:02:31 -07001119 goto error;
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001120 }
1121
Yasunori Gotobc02af92006-06-27 02:53:30 -07001122 /* call arch's memory hotadd */
1123 ret = arch_add_memory(nid, start, size);
1124
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001125 if (ret < 0)
1126 goto error;
1127
Yasunori Goto0fc44152006-06-27 02:53:38 -07001128 /* we online node here. we can't roll back from here. */
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001129 node_set_online(nid);
1130
Tang Chena1e565a2013-02-22 16:33:18 -08001131 if (new_node) {
Yasunori Goto0fc44152006-06-27 02:53:38 -07001132 ret = register_one_node(nid);
1133 /*
1134 * If sysfs file of new node can't create, cpu on the node
1135 * can't be hot-added. There is no rollback way now.
1136 * So, check by BUG_ON() to catch it reluctantly..
1137 */
1138 BUG_ON(ret);
1139 }
1140
akpm@linux-foundation.orgd96ae532010-03-05 13:41:58 -08001141 /* create new memmap entry */
1142 firmware_map_add_hotplug(start, start + size, "System RAM");
1143
Andi Kleen6ad696d2009-11-17 14:06:22 -08001144 goto out;
1145
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001146error:
1147 /* rollback pgdat allocation and others */
1148 if (new_pgdat)
1149 rollback_node_hotadd(nid, pgdat);
Sasha Levina864b9d2013-02-22 16:32:48 -08001150 release_memory_resource(res);
Yasunori Goto9af3c2d2006-06-27 02:53:34 -07001151
Andi Kleen6ad696d2009-11-17 14:06:22 -08001152out:
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -08001153 unlock_memory_hotplug();
Yasunori Gotobc02af92006-06-27 02:53:30 -07001154 return ret;
1155}
1156EXPORT_SYMBOL_GPL(add_memory);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001157
1158#ifdef CONFIG_MEMORY_HOTREMOVE
1159/*
Badari Pulavarty5c755e92008-07-23 21:28:19 -07001160 * A free page on the buddy free lists (not the per-cpu lists) has PageBuddy
1161 * set and the size of the free page is given by page_order(). Using this,
1162 * the function determines if the pageblock contains only free pages.
1163 * Due to buddy contraints, a free page at least the size of a pageblock will
1164 * be located at the start of the pageblock
1165 */
1166static inline int pageblock_free(struct page *page)
1167{
1168 return PageBuddy(page) && page_order(page) >= pageblock_order;
1169}
1170
1171/* Return the start of the next active pageblock after a given page */
1172static struct page *next_active_pageblock(struct page *page)
1173{
Badari Pulavarty5c755e92008-07-23 21:28:19 -07001174 /* Ensure the starting page is pageblock-aligned */
1175 BUG_ON(page_to_pfn(page) & (pageblock_nr_pages - 1));
1176
Badari Pulavarty5c755e92008-07-23 21:28:19 -07001177 /* If the entire pageblock is free, move to the end of free page */
KAMEZAWA Hiroyuki0dcc48c2010-09-09 16:38:01 -07001178 if (pageblock_free(page)) {
1179 int order;
1180 /* be careful. we don't have locks, page_order can be changed.*/
1181 order = page_order(page);
1182 if ((order < MAX_ORDER) && (order >= pageblock_order))
1183 return page + (1 << order);
1184 }
Badari Pulavarty5c755e92008-07-23 21:28:19 -07001185
KAMEZAWA Hiroyuki0dcc48c2010-09-09 16:38:01 -07001186 return page + pageblock_nr_pages;
Badari Pulavarty5c755e92008-07-23 21:28:19 -07001187}
1188
1189/* Checks if this range of memory is likely to be hot-removable. */
1190int is_mem_section_removable(unsigned long start_pfn, unsigned long nr_pages)
1191{
Badari Pulavarty5c755e92008-07-23 21:28:19 -07001192 struct page *page = pfn_to_page(start_pfn);
1193 struct page *end_page = page + nr_pages;
1194
1195 /* Check the starting page of each pageblock within the range */
1196 for (; page < end_page; page = next_active_pageblock(page)) {
KAMEZAWA Hiroyuki49ac8252010-10-26 14:21:30 -07001197 if (!is_pageblock_removable_nolock(page))
Badari Pulavarty5c755e92008-07-23 21:28:19 -07001198 return 0;
KAMEZAWA Hiroyuki49ac8252010-10-26 14:21:30 -07001199 cond_resched();
Badari Pulavarty5c755e92008-07-23 21:28:19 -07001200 }
1201
1202 /* All pageblocks in the memory block are likely to be hot-removable */
1203 return 1;
1204}
1205
1206/*
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001207 * Confirm all pages in a range [start, end) is belongs to the same zone.
1208 */
1209static int test_pages_in_a_zone(unsigned long start_pfn, unsigned long end_pfn)
1210{
1211 unsigned long pfn;
1212 struct zone *zone = NULL;
1213 struct page *page;
1214 int i;
1215 for (pfn = start_pfn;
1216 pfn < end_pfn;
1217 pfn += MAX_ORDER_NR_PAGES) {
1218 i = 0;
1219 /* This is just a CONFIG_HOLES_IN_ZONE check.*/
1220 while ((i < MAX_ORDER_NR_PAGES) && !pfn_valid_within(pfn + i))
1221 i++;
1222 if (i == MAX_ORDER_NR_PAGES)
1223 continue;
1224 page = pfn_to_page(pfn + i);
1225 if (zone && page_zone(page) != zone)
1226 return 0;
1227 zone = page_zone(page);
1228 }
1229 return 1;
1230}
1231
1232/*
1233 * Scanning pfn is much easier than scanning lru list.
1234 * Scan pfn from start to end and Find LRU page.
1235 */
Andrew Morton7bbc0902010-10-26 14:22:05 -07001236static unsigned long scan_lru_pages(unsigned long start, unsigned long end)
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001237{
1238 unsigned long pfn;
1239 struct page *page;
1240 for (pfn = start; pfn < end; pfn++) {
1241 if (pfn_valid(pfn)) {
1242 page = pfn_to_page(pfn);
1243 if (PageLRU(page))
1244 return pfn;
1245 }
1246 }
1247 return 0;
1248}
1249
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001250#define NR_OFFLINE_AT_ONCE_PAGES (256)
1251static int
1252do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
1253{
1254 unsigned long pfn;
1255 struct page *page;
1256 int move_pages = NR_OFFLINE_AT_ONCE_PAGES;
1257 int not_managed = 0;
1258 int ret = 0;
1259 LIST_HEAD(source);
1260
1261 for (pfn = start_pfn; pfn < end_pfn && move_pages > 0; pfn++) {
1262 if (!pfn_valid(pfn))
1263 continue;
1264 page = pfn_to_page(pfn);
Konstantin Khlebnikov700c2a42011-05-24 17:12:19 -07001265 if (!get_page_unless_zero(page))
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001266 continue;
1267 /*
1268 * We can skip free pages. And we can only deal with pages on
1269 * LRU.
1270 */
Nick Piggin62695a82008-10-18 20:26:09 -07001271 ret = isolate_lru_page(page);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001272 if (!ret) { /* Success */
Konstantin Khlebnikov700c2a42011-05-24 17:12:19 -07001273 put_page(page);
Nick Piggin62695a82008-10-18 20:26:09 -07001274 list_add_tail(&page->lru, &source);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001275 move_pages--;
KOSAKI Motohiro6d9c2852009-12-14 17:58:11 -08001276 inc_zone_page_state(page, NR_ISOLATED_ANON +
1277 page_is_file_cache(page));
1278
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001279 } else {
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001280#ifdef CONFIG_DEBUG_VM
Wu Fengguang718a3822010-03-10 15:20:43 -08001281 printk(KERN_ALERT "removing pfn %lx from LRU failed\n",
1282 pfn);
1283 dump_page(page);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001284#endif
Konstantin Khlebnikov700c2a42011-05-24 17:12:19 -07001285 put_page(page);
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001286 /* Because we don't have big zone->lock. we should
Bob Liu809c4442010-10-26 14:22:10 -07001287 check this again here. */
1288 if (page_count(page)) {
1289 not_managed++;
Bob Liuf3ab2632010-10-26 14:22:10 -07001290 ret = -EBUSY;
Bob Liu809c4442010-10-26 14:22:10 -07001291 break;
1292 }
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001293 }
1294 }
Bob Liuf3ab2632010-10-26 14:22:10 -07001295 if (!list_empty(&source)) {
1296 if (not_managed) {
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001297 putback_lru_pages(&source);
Bob Liuf3ab2632010-10-26 14:22:10 -07001298 goto out;
1299 }
Minchan Kim74c08f92012-10-08 16:32:54 -07001300
1301 /*
1302 * alloc_migrate_target should be improooooved!!
1303 * migrate_pages returns # of failed pages.
1304 */
1305 ret = migrate_pages(&source, alloc_migrate_target, 0,
Hugh Dickins9c620e22013-02-22 16:35:14 -08001306 MIGRATE_SYNC, MR_MEMORY_HOTPLUG);
Bob Liuf3ab2632010-10-26 14:22:10 -07001307 if (ret)
1308 putback_lru_pages(&source);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001309 }
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001310out:
1311 return ret;
1312}
1313
1314/*
1315 * remove from free_area[] and mark all as Reserved.
1316 */
1317static int
1318offline_isolated_pages_cb(unsigned long start, unsigned long nr_pages,
1319 void *data)
1320{
1321 __offline_isolated_pages(start, start + nr_pages);
1322 return 0;
1323}
1324
1325static void
1326offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn)
1327{
KAMEZAWA Hiroyuki908eedc2009-09-22 16:45:46 -07001328 walk_system_ram_range(start_pfn, end_pfn - start_pfn, NULL,
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001329 offline_isolated_pages_cb);
1330}
1331
1332/*
1333 * Check all pages in range, recoreded as memory resource, are isolated.
1334 */
1335static int
1336check_pages_isolated_cb(unsigned long start_pfn, unsigned long nr_pages,
1337 void *data)
1338{
1339 int ret;
1340 long offlined = *(long *)data;
Wen Congyangb023f462012-12-11 16:00:45 -08001341 ret = test_pages_isolated(start_pfn, start_pfn + nr_pages, true);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001342 offlined = nr_pages;
1343 if (!ret)
1344 *(long *)data += offlined;
1345 return ret;
1346}
1347
1348static long
1349check_pages_isolated(unsigned long start_pfn, unsigned long end_pfn)
1350{
1351 long offlined = 0;
1352 int ret;
1353
KAMEZAWA Hiroyuki908eedc2009-09-22 16:45:46 -07001354 ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn, &offlined,
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001355 check_pages_isolated_cb);
1356 if (ret < 0)
1357 offlined = (long)ret;
1358 return offlined;
1359}
1360
Lai Jiangshan09285af2012-12-12 13:52:04 -08001361#ifdef CONFIG_MOVABLE_NODE
Tang Chen79a4dce2012-12-18 14:23:24 -08001362/*
1363 * When CONFIG_MOVABLE_NODE, we permit offlining of a node which doesn't have
1364 * normal memory.
1365 */
Lai Jiangshan09285af2012-12-12 13:52:04 -08001366static bool can_offline_normal(struct zone *zone, unsigned long nr_pages)
1367{
1368 return true;
1369}
Tang Chen79a4dce2012-12-18 14:23:24 -08001370#else /* CONFIG_MOVABLE_NODE */
Lai Jiangshan74d42d82012-12-11 16:03:23 -08001371/* ensure the node has NORMAL memory if it is still online */
1372static bool can_offline_normal(struct zone *zone, unsigned long nr_pages)
1373{
1374 struct pglist_data *pgdat = zone->zone_pgdat;
1375 unsigned long present_pages = 0;
1376 enum zone_type zt;
1377
1378 for (zt = 0; zt <= ZONE_NORMAL; zt++)
1379 present_pages += pgdat->node_zones[zt].present_pages;
1380
1381 if (present_pages > nr_pages)
1382 return true;
1383
1384 present_pages = 0;
1385 for (; zt <= ZONE_MOVABLE; zt++)
1386 present_pages += pgdat->node_zones[zt].present_pages;
1387
1388 /*
1389 * we can't offline the last normal memory until all
1390 * higher memory is offlined.
1391 */
1392 return present_pages == 0;
1393}
Tang Chen79a4dce2012-12-18 14:23:24 -08001394#endif /* CONFIG_MOVABLE_NODE */
Lai Jiangshan74d42d82012-12-11 16:03:23 -08001395
Lai Jiangshand9713672012-12-11 16:01:03 -08001396/* check which state of node_states will be changed when offline memory */
1397static void node_states_check_changes_offline(unsigned long nr_pages,
1398 struct zone *zone, struct memory_notify *arg)
1399{
1400 struct pglist_data *pgdat = zone->zone_pgdat;
1401 unsigned long present_pages = 0;
1402 enum zone_type zt, zone_last = ZONE_NORMAL;
1403
1404 /*
Lai Jiangshan6715ddf2012-12-12 13:51:49 -08001405 * If we have HIGHMEM or movable node, node_states[N_NORMAL_MEMORY]
1406 * contains nodes which have zones of 0...ZONE_NORMAL,
1407 * set zone_last to ZONE_NORMAL.
Lai Jiangshand9713672012-12-11 16:01:03 -08001408 *
Lai Jiangshan6715ddf2012-12-12 13:51:49 -08001409 * If we don't have HIGHMEM nor movable node,
1410 * node_states[N_NORMAL_MEMORY] contains nodes which have zones of
1411 * 0...ZONE_MOVABLE, set zone_last to ZONE_MOVABLE.
Lai Jiangshand9713672012-12-11 16:01:03 -08001412 */
Lai Jiangshan6715ddf2012-12-12 13:51:49 -08001413 if (N_MEMORY == N_NORMAL_MEMORY)
Lai Jiangshand9713672012-12-11 16:01:03 -08001414 zone_last = ZONE_MOVABLE;
1415
1416 /*
1417 * check whether node_states[N_NORMAL_MEMORY] will be changed.
1418 * If the memory to be offline is in a zone of 0...zone_last,
1419 * and it is the last present memory, 0...zone_last will
1420 * become empty after offline , thus we can determind we will
1421 * need to clear the node from node_states[N_NORMAL_MEMORY].
1422 */
1423 for (zt = 0; zt <= zone_last; zt++)
1424 present_pages += pgdat->node_zones[zt].present_pages;
1425 if (zone_idx(zone) <= zone_last && nr_pages >= present_pages)
1426 arg->status_change_nid_normal = zone_to_nid(zone);
1427 else
1428 arg->status_change_nid_normal = -1;
1429
Lai Jiangshan6715ddf2012-12-12 13:51:49 -08001430#ifdef CONFIG_HIGHMEM
1431 /*
1432 * If we have movable node, node_states[N_HIGH_MEMORY]
1433 * contains nodes which have zones of 0...ZONE_HIGHMEM,
1434 * set zone_last to ZONE_HIGHMEM.
1435 *
1436 * If we don't have movable node, node_states[N_NORMAL_MEMORY]
1437 * contains nodes which have zones of 0...ZONE_MOVABLE,
1438 * set zone_last to ZONE_MOVABLE.
1439 */
1440 zone_last = ZONE_HIGHMEM;
1441 if (N_MEMORY == N_HIGH_MEMORY)
1442 zone_last = ZONE_MOVABLE;
1443
1444 for (; zt <= zone_last; zt++)
1445 present_pages += pgdat->node_zones[zt].present_pages;
1446 if (zone_idx(zone) <= zone_last && nr_pages >= present_pages)
1447 arg->status_change_nid_high = zone_to_nid(zone);
1448 else
1449 arg->status_change_nid_high = -1;
1450#else
1451 arg->status_change_nid_high = arg->status_change_nid_normal;
1452#endif
1453
Lai Jiangshand9713672012-12-11 16:01:03 -08001454 /*
1455 * node_states[N_HIGH_MEMORY] contains nodes which have 0...ZONE_MOVABLE
1456 */
1457 zone_last = ZONE_MOVABLE;
1458
1459 /*
1460 * check whether node_states[N_HIGH_MEMORY] will be changed
1461 * If we try to offline the last present @nr_pages from the node,
1462 * we can determind we will need to clear the node from
1463 * node_states[N_HIGH_MEMORY].
1464 */
1465 for (; zt <= zone_last; zt++)
1466 present_pages += pgdat->node_zones[zt].present_pages;
1467 if (nr_pages >= present_pages)
1468 arg->status_change_nid = zone_to_nid(zone);
1469 else
1470 arg->status_change_nid = -1;
1471}
1472
1473static void node_states_clear_node(int node, struct memory_notify *arg)
1474{
1475 if (arg->status_change_nid_normal >= 0)
1476 node_clear_state(node, N_NORMAL_MEMORY);
1477
Lai Jiangshan6715ddf2012-12-12 13:51:49 -08001478 if ((N_MEMORY != N_NORMAL_MEMORY) &&
1479 (arg->status_change_nid_high >= 0))
Lai Jiangshand9713672012-12-11 16:01:03 -08001480 node_clear_state(node, N_HIGH_MEMORY);
Lai Jiangshan6715ddf2012-12-12 13:51:49 -08001481
1482 if ((N_MEMORY != N_HIGH_MEMORY) &&
1483 (arg->status_change_nid >= 0))
1484 node_clear_state(node, N_MEMORY);
Lai Jiangshand9713672012-12-11 16:01:03 -08001485}
1486
Wen Congyanga16cee12012-10-08 16:33:58 -07001487static int __ref __offline_pages(unsigned long start_pfn,
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001488 unsigned long end_pfn, unsigned long timeout)
1489{
1490 unsigned long pfn, nr_pages, expire;
1491 long offlined_pages;
Yasunori Goto7b78d332007-10-21 16:41:36 -07001492 int ret, drain, retry_max, node;
Cody P Schaferd7029092013-07-03 15:02:11 -07001493 unsigned long flags;
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001494 struct zone *zone;
Yasunori Goto7b78d332007-10-21 16:41:36 -07001495 struct memory_notify arg;
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001496
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001497 /* at least, alignment against pageblock is necessary */
1498 if (!IS_ALIGNED(start_pfn, pageblock_nr_pages))
1499 return -EINVAL;
1500 if (!IS_ALIGNED(end_pfn, pageblock_nr_pages))
1501 return -EINVAL;
1502 /* This makes hotplug much easier...and readable.
1503 we assume this for now. .*/
1504 if (!test_pages_in_a_zone(start_pfn, end_pfn))
1505 return -EINVAL;
Yasunori Goto7b78d332007-10-21 16:41:36 -07001506
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -08001507 lock_memory_hotplug();
Andi Kleen6ad696d2009-11-17 14:06:22 -08001508
Yasunori Goto7b78d332007-10-21 16:41:36 -07001509 zone = page_zone(pfn_to_page(start_pfn));
1510 node = zone_to_nid(zone);
1511 nr_pages = end_pfn - start_pfn;
1512
Lai Jiangshan74d42d82012-12-11 16:03:23 -08001513 ret = -EINVAL;
1514 if (zone_idx(zone) <= ZONE_NORMAL && !can_offline_normal(zone, nr_pages))
1515 goto out;
1516
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001517 /* set above range as isolated */
Wen Congyangb023f462012-12-11 16:00:45 -08001518 ret = start_isolate_page_range(start_pfn, end_pfn,
1519 MIGRATE_MOVABLE, true);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001520 if (ret)
Andi Kleen6ad696d2009-11-17 14:06:22 -08001521 goto out;
Yasunori Goto7b78d332007-10-21 16:41:36 -07001522
1523 arg.start_pfn = start_pfn;
1524 arg.nr_pages = nr_pages;
Lai Jiangshand9713672012-12-11 16:01:03 -08001525 node_states_check_changes_offline(nr_pages, zone, &arg);
Yasunori Goto7b78d332007-10-21 16:41:36 -07001526
1527 ret = memory_notify(MEM_GOING_OFFLINE, &arg);
1528 ret = notifier_to_errno(ret);
1529 if (ret)
1530 goto failed_removal;
1531
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001532 pfn = start_pfn;
1533 expire = jiffies + timeout;
1534 drain = 0;
1535 retry_max = 5;
1536repeat:
1537 /* start memory hot removal */
1538 ret = -EAGAIN;
1539 if (time_after(jiffies, expire))
1540 goto failed_removal;
1541 ret = -EINTR;
1542 if (signal_pending(current))
1543 goto failed_removal;
1544 ret = 0;
1545 if (drain) {
1546 lru_add_drain_all();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001547 cond_resched();
Christoph Lameter9f8f2172008-02-04 22:29:11 -08001548 drain_all_pages();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001549 }
1550
1551 pfn = scan_lru_pages(start_pfn, end_pfn);
1552 if (pfn) { /* We have page on LRU */
1553 ret = do_migrate_range(pfn, end_pfn);
1554 if (!ret) {
1555 drain = 1;
1556 goto repeat;
1557 } else {
1558 if (ret < 0)
1559 if (--retry_max == 0)
1560 goto failed_removal;
1561 yield();
1562 drain = 1;
1563 goto repeat;
1564 }
1565 }
Adam Buchbinderb3834be2012-09-19 21:48:02 -04001566 /* drain all zone's lru pagevec, this is asynchronous... */
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001567 lru_add_drain_all();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001568 yield();
Adam Buchbinderb3834be2012-09-19 21:48:02 -04001569 /* drain pcp pages, this is synchronous. */
Christoph Lameter9f8f2172008-02-04 22:29:11 -08001570 drain_all_pages();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001571 /* check again */
1572 offlined_pages = check_pages_isolated(start_pfn, end_pfn);
1573 if (offlined_pages < 0) {
1574 ret = -EBUSY;
1575 goto failed_removal;
1576 }
1577 printk(KERN_INFO "Offlined Pages %ld\n", offlined_pages);
Adam Buchbinderb3834be2012-09-19 21:48:02 -04001578 /* Ok, all of our target is isolated.
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001579 We cannot do rollback at this point. */
1580 offline_isolated_pages(start_pfn, end_pfn);
KAMEZAWA Hiroyukidbc0e4c2007-11-14 16:59:12 -08001581 /* reset pagetype flags and makes migrate type to be MOVABLE */
Michal Nazarewicz0815f3d2012-04-03 15:06:15 +02001582 undo_isolate_page_range(start_pfn, end_pfn, MIGRATE_MOVABLE);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001583 /* removal success */
Jiang Liu3dcc0572013-07-03 15:03:21 -07001584 adjust_managed_page_count(pfn_to_page(start_pfn), -offlined_pages);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001585 zone->present_pages -= offlined_pages;
Cody P Schaferd7029092013-07-03 15:02:11 -07001586
1587 pgdat_resize_lock(zone->zone_pgdat, &flags);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001588 zone->zone_pgdat->node_present_pages -= offlined_pages;
Cody P Schaferd7029092013-07-03 15:02:11 -07001589 pgdat_resize_unlock(zone->zone_pgdat, &flags);
Yasunori Goto7b78d332007-10-21 16:41:36 -07001590
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -07001591 init_per_zone_wmark_min();
1592
Xishi Qiu1e8537b2012-10-08 16:31:51 -07001593 if (!populated_zone(zone)) {
Jiang Liu340175b2012-07-31 16:43:32 -07001594 zone_pcp_reset(zone);
Xishi Qiu1e8537b2012-10-08 16:31:51 -07001595 mutex_lock(&zonelists_mutex);
1596 build_all_zonelists(NULL, NULL);
1597 mutex_unlock(&zonelists_mutex);
1598 } else
1599 zone_pcp_update(zone);
Jiang Liu340175b2012-07-31 16:43:32 -07001600
Lai Jiangshand9713672012-12-11 16:01:03 -08001601 node_states_clear_node(node, &arg);
1602 if (arg.status_change_nid >= 0)
David Rientjes8fe23e02009-12-14 17:58:33 -08001603 kswapd_stop(node);
Minchan Kimbce73942009-06-16 15:32:50 -07001604
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001605 vm_total_pages = nr_free_pagecache_pages();
1606 writeback_set_ratelimit();
Yasunori Goto7b78d332007-10-21 16:41:36 -07001607
1608 memory_notify(MEM_OFFLINE, &arg);
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -08001609 unlock_memory_hotplug();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001610 return 0;
1611
1612failed_removal:
Bjorn Helgaasa62e2f42012-05-29 15:06:30 -07001613 printk(KERN_INFO "memory offlining [mem %#010llx-%#010llx] failed\n",
1614 (unsigned long long) start_pfn << PAGE_SHIFT,
1615 ((unsigned long long) end_pfn << PAGE_SHIFT) - 1);
Yasunori Goto7b78d332007-10-21 16:41:36 -07001616 memory_notify(MEM_CANCEL_OFFLINE, &arg);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001617 /* pushback to free area */
Michal Nazarewicz0815f3d2012-04-03 15:06:15 +02001618 undo_isolate_page_range(start_pfn, end_pfn, MIGRATE_MOVABLE);
Yasunori Goto7b78d332007-10-21 16:41:36 -07001619
Andi Kleen6ad696d2009-11-17 14:06:22 -08001620out:
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -08001621 unlock_memory_hotplug();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001622 return ret;
1623}
Badari Pulavarty71088782008-10-18 20:25:58 -07001624
Wen Congyanga16cee12012-10-08 16:33:58 -07001625int offline_pages(unsigned long start_pfn, unsigned long nr_pages)
1626{
1627 return __offline_pages(start_pfn, start_pfn + nr_pages, 120 * HZ);
1628}
Rafael J. Wysockie2ff3942013-05-08 00:29:49 +02001629#endif /* CONFIG_MEMORY_HOTREMOVE */
Wen Congyanga16cee12012-10-08 16:33:58 -07001630
Wen Congyangbbc76be2013-02-22 16:32:54 -08001631/**
1632 * walk_memory_range - walks through all mem sections in [start_pfn, end_pfn)
1633 * @start_pfn: start pfn of the memory range
Toshi Kanie05c4bb2013-04-29 15:06:16 -07001634 * @end_pfn: end pfn of the memory range
Wen Congyangbbc76be2013-02-22 16:32:54 -08001635 * @arg: argument passed to func
1636 * @func: callback for each memory section walked
1637 *
1638 * This function walks through all present mem sections in range
1639 * [start_pfn, end_pfn) and call func on each mem section.
1640 *
1641 * Returns the return value of func.
1642 */
Rafael J. Wysockie2ff3942013-05-08 00:29:49 +02001643int walk_memory_range(unsigned long start_pfn, unsigned long end_pfn,
Wen Congyangbbc76be2013-02-22 16:32:54 -08001644 void *arg, int (*func)(struct memory_block *, void *))
Badari Pulavarty71088782008-10-18 20:25:58 -07001645{
Wen Congyange90bdb72012-10-08 16:34:01 -07001646 struct memory_block *mem = NULL;
1647 struct mem_section *section;
Wen Congyange90bdb72012-10-08 16:34:01 -07001648 unsigned long pfn, section_nr;
1649 int ret;
Badari Pulavarty71088782008-10-18 20:25:58 -07001650
Wen Congyange90bdb72012-10-08 16:34:01 -07001651 for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
1652 section_nr = pfn_to_section_nr(pfn);
1653 if (!present_section_nr(section_nr))
1654 continue;
1655
1656 section = __nr_to_section(section_nr);
1657 /* same memblock? */
1658 if (mem)
1659 if ((section_nr >= mem->start_section_nr) &&
1660 (section_nr <= mem->end_section_nr))
1661 continue;
1662
1663 mem = find_memory_block_hinted(section, mem);
1664 if (!mem)
1665 continue;
1666
Wen Congyangbbc76be2013-02-22 16:32:54 -08001667 ret = func(mem, arg);
Wen Congyange90bdb72012-10-08 16:34:01 -07001668 if (ret) {
Wen Congyangbbc76be2013-02-22 16:32:54 -08001669 kobject_put(&mem->dev.kobj);
1670 return ret;
Wen Congyange90bdb72012-10-08 16:34:01 -07001671 }
1672 }
1673
1674 if (mem)
1675 kobject_put(&mem->dev.kobj);
1676
Wen Congyangbbc76be2013-02-22 16:32:54 -08001677 return 0;
1678}
1679
Rafael J. Wysockie2ff3942013-05-08 00:29:49 +02001680#ifdef CONFIG_MEMORY_HOTREMOVE
Wen Congyangbbc76be2013-02-22 16:32:54 -08001681static int is_memblock_offlined_cb(struct memory_block *mem, void *arg)
1682{
1683 int ret = !is_memblock_offlined(mem);
1684
Randy Dunlap349daa02013-04-29 15:08:49 -07001685 if (unlikely(ret)) {
1686 phys_addr_t beginpa, endpa;
1687
1688 beginpa = PFN_PHYS(section_nr_to_pfn(mem->start_section_nr));
1689 endpa = PFN_PHYS(section_nr_to_pfn(mem->end_section_nr + 1))-1;
Wen Congyangbbc76be2013-02-22 16:32:54 -08001690 pr_warn("removing memory fails, because memory "
Randy Dunlap349daa02013-04-29 15:08:49 -07001691 "[%pa-%pa] is onlined\n",
1692 &beginpa, &endpa);
1693 }
Wen Congyangbbc76be2013-02-22 16:32:54 -08001694
1695 return ret;
1696}
1697
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001698static int check_cpu_on_node(pg_data_t *pgdat)
Tang Chen60a5a192013-02-22 16:33:14 -08001699{
Tang Chen60a5a192013-02-22 16:33:14 -08001700 int cpu;
1701
1702 for_each_present_cpu(cpu) {
1703 if (cpu_to_node(cpu) == pgdat->node_id)
1704 /*
1705 * the cpu on this node isn't removed, and we can't
1706 * offline this node.
1707 */
1708 return -EBUSY;
1709 }
1710
1711 return 0;
1712}
1713
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001714static void unmap_cpu_on_node(pg_data_t *pgdat)
Wen Congyange13fe862013-02-22 16:33:31 -08001715{
1716#ifdef CONFIG_ACPI_NUMA
Wen Congyange13fe862013-02-22 16:33:31 -08001717 int cpu;
1718
1719 for_each_possible_cpu(cpu)
1720 if (cpu_to_node(cpu) == pgdat->node_id)
1721 numa_clear_node(cpu);
1722#endif
1723}
1724
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001725static int check_and_unmap_cpu_on_node(pg_data_t *pgdat)
Wen Congyange13fe862013-02-22 16:33:31 -08001726{
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001727 int ret;
Wen Congyange13fe862013-02-22 16:33:31 -08001728
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001729 ret = check_cpu_on_node(pgdat);
Wen Congyange13fe862013-02-22 16:33:31 -08001730 if (ret)
1731 return ret;
1732
1733 /*
1734 * the node will be offlined when we come here, so we can clear
1735 * the cpu_to_node() now.
1736 */
1737
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001738 unmap_cpu_on_node(pgdat);
Wen Congyange13fe862013-02-22 16:33:31 -08001739 return 0;
1740}
1741
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001742/**
1743 * try_offline_node
1744 *
1745 * Offline a node if all memory sections and cpus of the node are removed.
1746 *
1747 * NOTE: The caller must call lock_device_hotplug() to serialize hotplug
1748 * and online/offline operations before this call.
1749 */
Wen Congyang90b30cd2013-02-22 16:33:27 -08001750void try_offline_node(int nid)
Tang Chen60a5a192013-02-22 16:33:14 -08001751{
Wen Congyangd822b862013-02-22 16:33:16 -08001752 pg_data_t *pgdat = NODE_DATA(nid);
1753 unsigned long start_pfn = pgdat->node_start_pfn;
1754 unsigned long end_pfn = start_pfn + pgdat->node_spanned_pages;
Tang Chen60a5a192013-02-22 16:33:14 -08001755 unsigned long pfn;
Wen Congyangd822b862013-02-22 16:33:16 -08001756 struct page *pgdat_page = virt_to_page(pgdat);
1757 int i;
Tang Chen60a5a192013-02-22 16:33:14 -08001758
1759 for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
1760 unsigned long section_nr = pfn_to_section_nr(pfn);
1761
1762 if (!present_section_nr(section_nr))
1763 continue;
1764
1765 if (pfn_to_nid(pfn) != nid)
1766 continue;
1767
1768 /*
1769 * some memory sections of this node are not removed, and we
1770 * can't offline node now.
1771 */
1772 return;
1773 }
1774
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001775 if (check_and_unmap_cpu_on_node(pgdat))
Tang Chen60a5a192013-02-22 16:33:14 -08001776 return;
1777
1778 /*
1779 * all memory/cpu of this node are removed, we can offline this
1780 * node now.
1781 */
1782 node_set_offline(nid);
1783 unregister_one_node(nid);
Wen Congyangd822b862013-02-22 16:33:16 -08001784
1785 if (!PageSlab(pgdat_page) && !PageCompound(pgdat_page))
1786 /* node data is allocated from boot memory */
1787 return;
1788
1789 /* free waittable in each zone */
1790 for (i = 0; i < MAX_NR_ZONES; i++) {
1791 struct zone *zone = pgdat->node_zones + i;
1792
Jianguo Wuca4b3f32013-03-22 15:04:50 -07001793 /*
1794 * wait_table may be allocated from boot memory,
1795 * here only free if it's allocated by vmalloc.
1796 */
1797 if (is_vmalloc_addr(zone->wait_table))
Wen Congyangd822b862013-02-22 16:33:16 -08001798 vfree(zone->wait_table);
1799 }
1800
1801 /*
1802 * Since there is no way to guarentee the address of pgdat/zone is not
1803 * on stack of any kernel threads or used by other kernel objects
1804 * without reference counting or other symchronizing method, do not
1805 * reset node_data and free pgdat here. Just reset it to 0 and reuse
1806 * the memory when the node is online again.
1807 */
1808 memset(pgdat, 0, sizeof(*pgdat));
Tang Chen60a5a192013-02-22 16:33:14 -08001809}
Wen Congyang90b30cd2013-02-22 16:33:27 -08001810EXPORT_SYMBOL(try_offline_node);
Tang Chen60a5a192013-02-22 16:33:14 -08001811
Toshi Kani0f1cfe92013-09-11 14:21:50 -07001812/**
1813 * remove_memory
1814 *
1815 * NOTE: The caller must call lock_device_hotplug() to serialize hotplug
1816 * and online/offline operations before this call, as required by
1817 * try_offline_node().
1818 */
Rafael J. Wysocki242831e2013-05-27 12:58:46 +02001819void __ref remove_memory(int nid, u64 start, u64 size)
Wen Congyangbbc76be2013-02-22 16:32:54 -08001820{
Rafael J. Wysocki242831e2013-05-27 12:58:46 +02001821 int ret;
Wen Congyang993c1aa2013-02-22 16:32:50 -08001822
Toshi Kani27356f52013-09-11 14:21:49 -07001823 BUG_ON(check_hotplug_memory_range(start, size));
1824
Yasuaki Ishimatsu6677e3e2013-02-22 16:32:52 -08001825 lock_memory_hotplug();
1826
1827 /*
Rafael J. Wysocki242831e2013-05-27 12:58:46 +02001828 * All memory blocks must be offlined before removing memory. Check
1829 * whether all memory blocks in question are offline and trigger a BUG()
1830 * if this is not the case.
Yasuaki Ishimatsu6677e3e2013-02-22 16:32:52 -08001831 */
Rafael J. Wysocki242831e2013-05-27 12:58:46 +02001832 ret = walk_memory_range(PFN_DOWN(start), PFN_UP(start + size - 1), NULL,
Wen Congyangbbc76be2013-02-22 16:32:54 -08001833 is_memblock_offlined_cb);
1834 if (ret) {
1835 unlock_memory_hotplug();
Rafael J. Wysocki242831e2013-05-27 12:58:46 +02001836 BUG();
Yasuaki Ishimatsu6677e3e2013-02-22 16:32:52 -08001837 }
1838
Yasuaki Ishimatsu46c66c42013-02-22 16:32:56 -08001839 /* remove memmap entry */
1840 firmware_map_remove(start, start + size, "System RAM");
1841
Wen Congyang24d335c2013-02-22 16:32:58 -08001842 arch_remove_memory(start, size);
1843
Tang Chen60a5a192013-02-22 16:33:14 -08001844 try_offline_node(nid);
1845
Yasuaki Ishimatsu6677e3e2013-02-22 16:32:52 -08001846 unlock_memory_hotplug();
Badari Pulavarty71088782008-10-18 20:25:58 -07001847}
Badari Pulavarty71088782008-10-18 20:25:58 -07001848EXPORT_SYMBOL_GPL(remove_memory);
Rafael J. Wysockiaba6efc2013-06-01 22:24:07 +02001849#endif /* CONFIG_MEMORY_HOTREMOVE */