blob: 2168489c0bc9f309ff3347d6dac05a09039ad7d2 [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>
Dave Hansen3947be12005-10-29 18:16:54 -070032
33#include <asm/tlbflush.h>
34
Adrian Bunk1e5ad9a2008-04-28 20:40:08 +030035#include "internal.h"
36
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -070037/*
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
44static void generic_online_page(struct page *page);
45
46static online_page_callback_t online_page_callback = generic_online_page;
47
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -080048DEFINE_MUTEX(mem_hotplug_mutex);
49
50void 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
58void unlock_memory_hotplug(void)
59{
60 unlock_system_sleep();
61 mutex_unlock(&mem_hotplug_mutex);
62}
63
64
Keith Mannthey45e0b782006-09-30 23:27:09 -070065/* add this memory to iomem resource */
66static 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;
Yasunori Goto887c3cb2007-11-14 16:59:20 -080075 res->flags = IORESOURCE_MEM | IORESOURCE_BUSY;
Keith Mannthey45e0b782006-09-30 23:27:09 -070076 if (request_resource(&iomem_resource, res) < 0) {
77 printk("System RAM resource %llx - %llx cannot be added\n",
78 (unsigned long long)res->start, (unsigned long long)res->end);
79 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
Yasunori Goto04753272008-04-28 02:13:31 -070095#ifndef CONFIG_SPARSEMEM_VMEMMAP
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -080096static void get_page_bootmem(unsigned long info, struct page *page,
97 unsigned long type)
Yasunori Goto04753272008-04-28 02:13:31 -070098{
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -080099 page->lru.next = (struct list_head *) type;
Yasunori Goto04753272008-04-28 02:13:31 -0700100 SetPagePrivate(page);
101 set_page_private(page, info);
102 atomic_inc(&page->_count);
103}
104
Rakib Mullick23ce9322009-12-14 17:59:44 -0800105/* reference to __meminit __free_pages_bootmem is valid
106 * so use __ref to tell modpost not to generate a warning */
107void __ref put_page_bootmem(struct page *page)
Yasunori Goto04753272008-04-28 02:13:31 -0700108{
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800109 unsigned long type;
Yasunori Goto04753272008-04-28 02:13:31 -0700110
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800111 type = (unsigned long) page->lru.next;
112 BUG_ON(type < MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE ||
113 type > MEMORY_HOTPLUG_MAX_BOOTMEM_TYPE);
Yasunori Goto04753272008-04-28 02:13:31 -0700114
115 if (atomic_dec_return(&page->_count) == 1) {
116 ClearPagePrivate(page);
117 set_page_private(page, 0);
Andrea Arcangeli5f24ce52011-01-13 15:47:00 -0800118 INIT_LIST_HEAD(&page->lru);
Yasunori Goto04753272008-04-28 02:13:31 -0700119 __free_pages_bootmem(page, 0);
120 }
121
122}
123
Adrian Bunkd92bc312008-07-23 21:28:12 -0700124static void register_page_bootmem_info_section(unsigned long start_pfn)
Yasunori Goto04753272008-04-28 02:13:31 -0700125{
126 unsigned long *usemap, mapsize, section_nr, i;
127 struct mem_section *ms;
128 struct page *page, *memmap;
129
130 if (!pfn_valid(start_pfn))
131 return;
132
133 section_nr = pfn_to_section_nr(start_pfn);
134 ms = __nr_to_section(section_nr);
135
136 /* Get section's memmap address */
137 memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr);
138
139 /*
140 * Get page for the memmap's phys address
141 * XXX: need more consideration for sparse_vmemmap...
142 */
143 page = virt_to_page(memmap);
144 mapsize = sizeof(struct page) * PAGES_PER_SECTION;
145 mapsize = PAGE_ALIGN(mapsize) >> PAGE_SHIFT;
146
147 /* remember memmap's page */
148 for (i = 0; i < mapsize; i++, page++)
149 get_page_bootmem(section_nr, page, SECTION_INFO);
150
151 usemap = __nr_to_section(section_nr)->pageblock_flags;
152 page = virt_to_page(usemap);
153
154 mapsize = PAGE_ALIGN(usemap_size()) >> PAGE_SHIFT;
155
156 for (i = 0; i < mapsize; i++, page++)
Yasunori Gotoaf370fb2008-07-23 21:28:17 -0700157 get_page_bootmem(section_nr, page, MIX_SECTION_INFO);
Yasunori Goto04753272008-04-28 02:13:31 -0700158
159}
160
161void register_page_bootmem_info_node(struct pglist_data *pgdat)
162{
163 unsigned long i, pfn, end_pfn, nr_pages;
164 int node = pgdat->node_id;
165 struct page *page;
166 struct zone *zone;
167
168 nr_pages = PAGE_ALIGN(sizeof(struct pglist_data)) >> PAGE_SHIFT;
169 page = virt_to_page(pgdat);
170
171 for (i = 0; i < nr_pages; i++, page++)
172 get_page_bootmem(node, page, NODE_INFO);
173
174 zone = &pgdat->node_zones[0];
175 for (; zone < pgdat->node_zones + MAX_NR_ZONES - 1; zone++) {
176 if (zone->wait_table) {
177 nr_pages = zone->wait_table_hash_nr_entries
178 * sizeof(wait_queue_head_t);
179 nr_pages = PAGE_ALIGN(nr_pages) >> PAGE_SHIFT;
180 page = virt_to_page(zone->wait_table);
181
182 for (i = 0; i < nr_pages; i++, page++)
183 get_page_bootmem(node, page, NODE_INFO);
184 }
185 }
186
187 pfn = pgdat->node_start_pfn;
188 end_pfn = pfn + pgdat->node_spanned_pages;
189
190 /* register_section info */
191 for (; pfn < end_pfn; pfn += PAGES_PER_SECTION)
192 register_page_bootmem_info_section(pfn);
193
194}
195#endif /* !CONFIG_SPARSEMEM_VMEMMAP */
196
Heiko Carstens76cdd582008-05-14 16:05:52 -0700197static void grow_zone_span(struct zone *zone, unsigned long start_pfn,
198 unsigned long end_pfn)
199{
200 unsigned long old_zone_end_pfn;
201
202 zone_span_writelock(zone);
203
204 old_zone_end_pfn = zone->zone_start_pfn + zone->spanned_pages;
205 if (start_pfn < zone->zone_start_pfn)
206 zone->zone_start_pfn = start_pfn;
207
208 zone->spanned_pages = max(old_zone_end_pfn, end_pfn) -
209 zone->zone_start_pfn;
210
211 zone_span_writeunlock(zone);
212}
213
214static void grow_pgdat_span(struct pglist_data *pgdat, unsigned long start_pfn,
215 unsigned long end_pfn)
216{
217 unsigned long old_pgdat_end_pfn =
218 pgdat->node_start_pfn + pgdat->node_spanned_pages;
219
220 if (start_pfn < pgdat->node_start_pfn)
221 pgdat->node_start_pfn = start_pfn;
222
223 pgdat->node_spanned_pages = max(old_pgdat_end_pfn, end_pfn) -
224 pgdat->node_start_pfn;
225}
226
Al Viro31168482008-11-22 17:33:24 +0000227static int __meminit __add_zone(struct zone *zone, unsigned long phys_start_pfn)
Dave Hansen3947be12005-10-29 18:16:54 -0700228{
229 struct pglist_data *pgdat = zone->zone_pgdat;
230 int nr_pages = PAGES_PER_SECTION;
231 int nid = pgdat->node_id;
232 int zone_type;
Heiko Carstens76cdd582008-05-14 16:05:52 -0700233 unsigned long flags;
Dave Hansen3947be12005-10-29 18:16:54 -0700234
235 zone_type = zone - pgdat->node_zones;
Heiko Carstens76cdd582008-05-14 16:05:52 -0700236 if (!zone->wait_table) {
237 int ret;
238
239 ret = init_currently_empty_zone(zone, phys_start_pfn,
240 nr_pages, MEMMAP_HOTPLUG);
241 if (ret)
242 return ret;
243 }
244 pgdat_resize_lock(zone->zone_pgdat, &flags);
245 grow_zone_span(zone, phys_start_pfn, phys_start_pfn + nr_pages);
246 grow_pgdat_span(zone->zone_pgdat, phys_start_pfn,
247 phys_start_pfn + nr_pages);
248 pgdat_resize_unlock(zone->zone_pgdat, &flags);
Dave Hansena2f3aa022007-01-10 23:15:30 -0800249 memmap_init_zone(nr_pages, nid, zone_type,
250 phys_start_pfn, MEMMAP_HOTPLUG);
Yasunori Goto718127c2006-06-23 02:03:10 -0700251 return 0;
Dave Hansen3947be12005-10-29 18:16:54 -0700252}
253
Gary Hadec04fc582009-01-06 14:39:14 -0800254static int __meminit __add_section(int nid, struct zone *zone,
255 unsigned long phys_start_pfn)
Dave Hansen3947be12005-10-29 18:16:54 -0700256{
Dave Hansen3947be12005-10-29 18:16:54 -0700257 int nr_pages = PAGES_PER_SECTION;
Dave Hansen3947be12005-10-29 18:16:54 -0700258 int ret;
259
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -0700260 if (pfn_valid(phys_start_pfn))
261 return -EEXIST;
262
Dave Hansen0b0acbe2005-10-29 18:16:55 -0700263 ret = sparse_add_one_section(zone, phys_start_pfn, nr_pages);
Dave Hansen3947be12005-10-29 18:16:54 -0700264
265 if (ret < 0)
266 return ret;
267
Yasunori Goto718127c2006-06-23 02:03:10 -0700268 ret = __add_zone(zone, phys_start_pfn);
269
270 if (ret < 0)
271 return ret;
272
Gary Hadec04fc582009-01-06 14:39:14 -0800273 return register_new_memory(nid, __pfn_to_section(phys_start_pfn));
Dave Hansen3947be12005-10-29 18:16:54 -0700274}
275
Yasunori Goto0c0a4a52008-04-28 02:13:34 -0700276#ifdef CONFIG_SPARSEMEM_VMEMMAP
277static int __remove_section(struct zone *zone, struct mem_section *ms)
278{
279 /*
280 * XXX: Freeing memmap with vmemmap is not implement yet.
281 * This should be removed later.
282 */
283 return -EBUSY;
284}
285#else
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700286static int __remove_section(struct zone *zone, struct mem_section *ms)
287{
288 unsigned long flags;
289 struct pglist_data *pgdat = zone->zone_pgdat;
290 int ret = -EINVAL;
291
292 if (!valid_section(ms))
293 return ret;
294
295 ret = unregister_memory_section(ms);
296 if (ret)
297 return ret;
298
299 pgdat_resize_lock(pgdat, &flags);
300 sparse_remove_one_section(zone, ms);
301 pgdat_resize_unlock(pgdat, &flags);
302 return 0;
303}
Yasunori Goto0c0a4a52008-04-28 02:13:34 -0700304#endif
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700305
Dave Hansen3947be12005-10-29 18:16:54 -0700306/*
307 * Reasonably generic function for adding memory. It is
308 * expected that archs that support memory hotplug will
309 * call this function after deciding the zone to which to
310 * add the new pages.
311 */
Gary Hadec04fc582009-01-06 14:39:14 -0800312int __ref __add_pages(int nid, struct zone *zone, unsigned long phys_start_pfn,
313 unsigned long nr_pages)
Dave Hansen3947be12005-10-29 18:16:54 -0700314{
315 unsigned long i;
316 int err = 0;
KAMEZAWA Hiroyuki6f712712006-08-05 12:14:58 -0700317 int start_sec, end_sec;
318 /* during initialize mem_map, align hot-added range to section */
319 start_sec = pfn_to_section_nr(phys_start_pfn);
320 end_sec = pfn_to_section_nr(phys_start_pfn + nr_pages - 1);
Dave Hansen3947be12005-10-29 18:16:54 -0700321
KAMEZAWA Hiroyuki6f712712006-08-05 12:14:58 -0700322 for (i = start_sec; i <= end_sec; i++) {
Gary Hadec04fc582009-01-06 14:39:14 -0800323 err = __add_section(nid, zone, i << PFN_SECTION_SHIFT);
Dave Hansen3947be12005-10-29 18:16:54 -0700324
KAMEZAWA Hiroyuki6f712712006-08-05 12:14:58 -0700325 /*
Simon Arlott183ff222007-10-20 01:27:18 +0200326 * EEXIST is finally dealt with by ioresource collision
KAMEZAWA Hiroyuki6f712712006-08-05 12:14:58 -0700327 * check. see add_memory() => register_memory_resource()
328 * Warning will be printed if there is collision.
Joel H Schoppbed120c2006-05-01 12:16:11 -0700329 */
330 if (err && (err != -EEXIST))
Dave Hansen3947be12005-10-29 18:16:54 -0700331 break;
KAMEZAWA Hiroyuki6f712712006-08-05 12:14:58 -0700332 err = 0;
Dave Hansen3947be12005-10-29 18:16:54 -0700333 }
334
335 return err;
336}
Joel H Schoppbed120c2006-05-01 12:16:11 -0700337EXPORT_SYMBOL_GPL(__add_pages);
Dave Hansen3947be12005-10-29 18:16:54 -0700338
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700339/**
340 * __remove_pages() - remove sections of pages from a zone
341 * @zone: zone from which pages need to be removed
342 * @phys_start_pfn: starting pageframe (must be aligned to start of a section)
343 * @nr_pages: number of pages to remove (must be multiple of section size)
344 *
345 * Generic helper function to remove section mappings and sysfs entries
346 * for the section of the memory we are removing. Caller needs to make
347 * sure that pages are marked reserved and zones are adjust properly by
348 * calling offline_pages().
349 */
350int __remove_pages(struct zone *zone, unsigned long phys_start_pfn,
351 unsigned long nr_pages)
352{
353 unsigned long i, ret = 0;
354 int sections_to_remove;
355
356 /*
357 * We can only remove entire sections
358 */
359 BUG_ON(phys_start_pfn & ~PAGE_SECTION_MASK);
360 BUG_ON(nr_pages % PAGES_PER_SECTION);
361
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700362 sections_to_remove = nr_pages / PAGES_PER_SECTION;
363 for (i = 0; i < sections_to_remove; i++) {
364 unsigned long pfn = phys_start_pfn + i*PAGES_PER_SECTION;
Nathan Fontenotde7f0cb2008-10-18 20:27:14 -0700365 release_mem_region(pfn << PAGE_SHIFT,
366 PAGES_PER_SECTION << PAGE_SHIFT);
Badari Pulavartyea01ea92008-04-28 02:12:01 -0700367 ret = __remove_section(zone, __pfn_to_section(pfn));
368 if (ret)
369 break;
370 }
371 return ret;
372}
373EXPORT_SYMBOL_GPL(__remove_pages);
374
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700375int set_online_page_callback(online_page_callback_t callback)
376{
377 int rc = -EINVAL;
378
379 lock_memory_hotplug();
380
381 if (online_page_callback == generic_online_page) {
382 online_page_callback = callback;
383 rc = 0;
384 }
385
386 unlock_memory_hotplug();
387
388 return rc;
389}
390EXPORT_SYMBOL_GPL(set_online_page_callback);
391
392int restore_online_page_callback(online_page_callback_t callback)
393{
394 int rc = -EINVAL;
395
396 lock_memory_hotplug();
397
398 if (online_page_callback == callback) {
399 online_page_callback = generic_online_page;
400 rc = 0;
401 }
402
403 unlock_memory_hotplug();
404
405 return rc;
406}
407EXPORT_SYMBOL_GPL(restore_online_page_callback);
408
409void __online_page_set_limits(struct page *page)
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700410{
Jan Beulich4738e1b2009-09-21 17:03:03 -0700411 unsigned long pfn = page_to_pfn(page);
412
Jan Beulich4738e1b2009-09-21 17:03:03 -0700413 if (pfn >= num_physpages)
414 num_physpages = pfn + 1;
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700415}
416EXPORT_SYMBOL_GPL(__online_page_set_limits);
417
418void __online_page_increment_counters(struct page *page)
419{
420 totalram_pages++;
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700421
422#ifdef CONFIG_HIGHMEM
423 if (PageHighMem(page))
424 totalhigh_pages++;
425#endif
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700426}
427EXPORT_SYMBOL_GPL(__online_page_increment_counters);
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700428
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700429void __online_page_free(struct page *page)
430{
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700431 ClearPageReserved(page);
432 init_page_count(page);
433 __free_page(page);
434}
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700435EXPORT_SYMBOL_GPL(__online_page_free);
436
437static void generic_online_page(struct page *page)
438{
439 __online_page_set_limits(page);
440 __online_page_increment_counters(page);
441 __online_page_free(page);
442}
Jeremy Fitzhardinge180c06e2008-04-28 02:12:03 -0700443
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700444static int online_pages_range(unsigned long start_pfn, unsigned long nr_pages,
445 void *arg)
Dave Hansen3947be12005-10-29 18:16:54 -0700446{
447 unsigned long i;
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700448 unsigned long onlined_pages = *(unsigned long *)arg;
449 struct page *page;
450 if (PageReserved(pfn_to_page(start_pfn)))
451 for (i = 0; i < nr_pages; i++) {
452 page = pfn_to_page(start_pfn + i);
Daniel Kiper9d0ad8c2011-07-25 17:12:05 -0700453 (*online_page_callback)(page);
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700454 onlined_pages++;
455 }
456 *(unsigned long *)arg = onlined_pages;
457 return 0;
458}
459
460
KOSAKI Motohiro839a4fc2011-05-24 17:11:31 -0700461int __ref online_pages(unsigned long pfn, unsigned long nr_pages)
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700462{
Dave Hansen3947be12005-10-29 18:16:54 -0700463 unsigned long onlined_pages = 0;
464 struct zone *zone;
Yasunori Goto68113782006-06-23 02:03:11 -0700465 int need_zonelists_rebuild = 0;
Yasunori Goto7b78d332007-10-21 16:41:36 -0700466 int nid;
467 int ret;
468 struct memory_notify arg;
Dave Hansen3947be12005-10-29 18:16:54 -0700469
KAMEZAWA Hiroyuki925268a2011-01-11 16:44:01 +0900470 lock_memory_hotplug();
Yasunori Goto7b78d332007-10-21 16:41:36 -0700471 arg.start_pfn = pfn;
472 arg.nr_pages = nr_pages;
473 arg.status_change_nid = -1;
474
475 nid = page_to_nid(pfn_to_page(pfn));
476 if (node_present_pages(nid) == 0)
477 arg.status_change_nid = nid;
478
479 ret = memory_notify(MEM_GOING_ONLINE, &arg);
480 ret = notifier_to_errno(ret);
481 if (ret) {
482 memory_notify(MEM_CANCEL_ONLINE, &arg);
KAMEZAWA Hiroyuki925268a2011-01-11 16:44:01 +0900483 unlock_memory_hotplug();
Yasunori Goto7b78d332007-10-21 16:41:36 -0700484 return ret;
485 }
Dave Hansen3947be12005-10-29 18:16:54 -0700486 /*
487 * This doesn't need a lock to do pfn_to_page().
488 * The section can't be removed here because of the
Daniel Walkerda19cbc2008-02-04 23:35:47 -0800489 * memory_block->state_mutex.
Dave Hansen3947be12005-10-29 18:16:54 -0700490 */
491 zone = page_zone(pfn_to_page(pfn));
Yasunori Goto68113782006-06-23 02:03:11 -0700492 /*
493 * If this zone is not populated, then it is not in zonelist.
494 * This means the page allocator ignores this zone.
495 * So, zonelist must be updated after online.
496 */
Haicheng Li4eaf3f62010-05-24 14:32:52 -0700497 mutex_lock(&zonelists_mutex);
Yasunori Goto68113782006-06-23 02:03:11 -0700498 if (!populated_zone(zone))
499 need_zonelists_rebuild = 1;
500
KAMEZAWA Hiroyuki908eedc2009-09-22 16:45:46 -0700501 ret = walk_system_ram_range(pfn, nr_pages, &onlined_pages,
KAMEZAWA Hiroyuki75884fb2007-10-16 01:26:10 -0700502 online_pages_range);
Geoff Levandfd8a4222008-05-14 16:05:50 -0700503 if (ret) {
Haicheng Li4eaf3f62010-05-24 14:32:52 -0700504 mutex_unlock(&zonelists_mutex);
Geoff Levandfd8a4222008-05-14 16:05:50 -0700505 printk(KERN_DEBUG "online_pages %lx at %lx failed\n",
506 nr_pages, pfn);
507 memory_notify(MEM_CANCEL_ONLINE, &arg);
KAMEZAWA Hiroyuki925268a2011-01-11 16:44:01 +0900508 unlock_memory_hotplug();
Geoff Levandfd8a4222008-05-14 16:05:50 -0700509 return ret;
510 }
511
Dave Hansen3947be12005-10-29 18:16:54 -0700512 zone->present_pages += onlined_pages;
Yasunori Gotof2937be2006-03-09 17:33:51 -0800513 zone->zone_pgdat->node_present_pages += onlined_pages;
Haicheng Li1f522502010-05-24 14:32:51 -0700514 if (need_zonelists_rebuild)
515 build_all_zonelists(zone);
516 else
517 zone_pcp_update(zone);
Dave Hansen3947be12005-10-29 18:16:54 -0700518
Haicheng Li4eaf3f62010-05-24 14:32:52 -0700519 mutex_unlock(&zonelists_mutex);
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -0700520
521 init_per_zone_wmark_min();
522
Christoph Lameter7ea15302007-10-16 01:25:29 -0700523 if (onlined_pages) {
524 kswapd_run(zone_to_nid(zone));
525 node_set_state(zone_to_nid(zone), N_HIGH_MEMORY);
526 }
Dave Hansen61b13992005-10-29 18:16:56 -0700527
Haicheng Li1f522502010-05-24 14:32:51 -0700528 vm_total_pages = nr_free_pagecache_pages();
Kent Liu2f7f24e2008-07-23 21:28:18 -0700529
Chandra Seetharaman2d1d43f2006-09-29 02:01:25 -0700530 writeback_set_ratelimit();
Yasunori Goto7b78d332007-10-21 16:41:36 -0700531
532 if (onlined_pages)
533 memory_notify(MEM_ONLINE, &arg);
KAMEZAWA Hiroyuki925268a2011-01-11 16:44:01 +0900534 unlock_memory_hotplug();
Yasunori Goto7b78d332007-10-21 16:41:36 -0700535
Dave Hansen3947be12005-10-29 18:16:54 -0700536 return 0;
537}
Keith Mannthey53947022006-09-30 23:27:08 -0700538#endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */
Yasunori Gotobc02af92006-06-27 02:53:30 -0700539
Hidetoshi Setoe1319332009-11-17 14:06:18 -0800540/* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */
541static pg_data_t __ref *hotadd_new_pgdat(int nid, u64 start)
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700542{
543 struct pglist_data *pgdat;
544 unsigned long zones_size[MAX_NR_ZONES] = {0};
545 unsigned long zholes_size[MAX_NR_ZONES] = {0};
546 unsigned long start_pfn = start >> PAGE_SHIFT;
547
548 pgdat = arch_alloc_nodedata(nid);
549 if (!pgdat)
550 return NULL;
551
552 arch_refresh_nodedata(nid, pgdat);
553
554 /* we can use NODE_DATA(nid) from here */
555
556 /* init node's zones as empty zones, we don't have any present pages.*/
Johannes Weiner9109fb72008-07-23 21:27:20 -0700557 free_area_init_node(nid, zones_size, start_pfn, zholes_size);
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700558
KAMEZAWA Hiroyuki959ecc42011-06-15 15:08:38 -0700559 /*
560 * The node we allocated has no zone fallback lists. For avoiding
561 * to access not-initialized zonelist, build here.
562 */
David Rientjesf957db42011-06-22 18:13:04 -0700563 mutex_lock(&zonelists_mutex);
KAMEZAWA Hiroyuki959ecc42011-06-15 15:08:38 -0700564 build_all_zonelists(NULL);
David Rientjesf957db42011-06-22 18:13:04 -0700565 mutex_unlock(&zonelists_mutex);
KAMEZAWA Hiroyuki959ecc42011-06-15 15:08:38 -0700566
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700567 return pgdat;
568}
569
570static void rollback_node_hotadd(int nid, pg_data_t *pgdat)
571{
572 arch_refresh_nodedata(nid, NULL);
573 arch_free_nodedata(pgdat);
574 return;
575}
576
KAMEZAWA Hiroyuki0a547032006-06-27 02:53:35 -0700577
minskey guocf234222010-05-24 14:32:41 -0700578/*
579 * called by cpu_up() to online a node without onlined memory.
580 */
581int mem_online_node(int nid)
582{
583 pg_data_t *pgdat;
584 int ret;
585
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -0800586 lock_memory_hotplug();
minskey guocf234222010-05-24 14:32:41 -0700587 pgdat = hotadd_new_pgdat(nid, 0);
David Rientjes7553e8f2011-06-22 18:13:01 -0700588 if (!pgdat) {
minskey guocf234222010-05-24 14:32:41 -0700589 ret = -ENOMEM;
590 goto out;
591 }
592 node_set_online(nid);
593 ret = register_one_node(nid);
594 BUG_ON(ret);
595
596out:
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -0800597 unlock_memory_hotplug();
minskey guocf234222010-05-24 14:32:41 -0700598 return ret;
599}
600
Al Viro31168482008-11-22 17:33:24 +0000601/* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */
602int __ref add_memory(int nid, u64 start, u64 size)
Yasunori Gotobc02af92006-06-27 02:53:30 -0700603{
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700604 pg_data_t *pgdat = NULL;
605 int new_pgdat = 0;
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -0700606 struct resource *res;
Yasunori Gotobc02af92006-06-27 02:53:30 -0700607 int ret;
608
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -0800609 lock_memory_hotplug();
Andi Kleen6ad696d2009-11-17 14:06:22 -0800610
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -0700611 res = register_memory_resource(start, size);
Andi Kleen6ad696d2009-11-17 14:06:22 -0800612 ret = -EEXIST;
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -0700613 if (!res)
Andi Kleen6ad696d2009-11-17 14:06:22 -0800614 goto out;
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -0700615
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700616 if (!node_online(nid)) {
617 pgdat = hotadd_new_pgdat(nid, start);
Andi Kleen6ad696d2009-11-17 14:06:22 -0800618 ret = -ENOMEM;
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700619 if (!pgdat)
Andi Kleen6ad696d2009-11-17 14:06:22 -0800620 goto out;
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700621 new_pgdat = 1;
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700622 }
623
Yasunori Gotobc02af92006-06-27 02:53:30 -0700624 /* call arch's memory hotadd */
625 ret = arch_add_memory(nid, start, size);
626
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700627 if (ret < 0)
628 goto error;
629
Yasunori Goto0fc44152006-06-27 02:53:38 -0700630 /* we online node here. we can't roll back from here. */
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700631 node_set_online(nid);
632
Yasunori Goto0fc44152006-06-27 02:53:38 -0700633 if (new_pgdat) {
634 ret = register_one_node(nid);
635 /*
636 * If sysfs file of new node can't create, cpu on the node
637 * can't be hot-added. There is no rollback way now.
638 * So, check by BUG_ON() to catch it reluctantly..
639 */
640 BUG_ON(ret);
641 }
642
akpm@linux-foundation.orgd96ae532010-03-05 13:41:58 -0800643 /* create new memmap entry */
644 firmware_map_add_hotplug(start, start + size, "System RAM");
645
Andi Kleen6ad696d2009-11-17 14:06:22 -0800646 goto out;
647
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700648error:
649 /* rollback pgdat allocation and others */
650 if (new_pgdat)
651 rollback_node_hotadd(nid, pgdat);
KAMEZAWA Hiroyukiebd15302006-08-05 12:15:06 -0700652 if (res)
653 release_memory_resource(res);
Yasunori Goto9af3c2d2006-06-27 02:53:34 -0700654
Andi Kleen6ad696d2009-11-17 14:06:22 -0800655out:
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -0800656 unlock_memory_hotplug();
Yasunori Gotobc02af92006-06-27 02:53:30 -0700657 return ret;
658}
659EXPORT_SYMBOL_GPL(add_memory);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700660
661#ifdef CONFIG_MEMORY_HOTREMOVE
662/*
Badari Pulavarty5c755e92008-07-23 21:28:19 -0700663 * A free page on the buddy free lists (not the per-cpu lists) has PageBuddy
664 * set and the size of the free page is given by page_order(). Using this,
665 * the function determines if the pageblock contains only free pages.
666 * Due to buddy contraints, a free page at least the size of a pageblock will
667 * be located at the start of the pageblock
668 */
669static inline int pageblock_free(struct page *page)
670{
671 return PageBuddy(page) && page_order(page) >= pageblock_order;
672}
673
674/* Return the start of the next active pageblock after a given page */
675static struct page *next_active_pageblock(struct page *page)
676{
Badari Pulavarty5c755e92008-07-23 21:28:19 -0700677 /* Ensure the starting page is pageblock-aligned */
678 BUG_ON(page_to_pfn(page) & (pageblock_nr_pages - 1));
679
Badari Pulavarty5c755e92008-07-23 21:28:19 -0700680 /* If the entire pageblock is free, move to the end of free page */
KAMEZAWA Hiroyuki0dcc48c2010-09-09 16:38:01 -0700681 if (pageblock_free(page)) {
682 int order;
683 /* be careful. we don't have locks, page_order can be changed.*/
684 order = page_order(page);
685 if ((order < MAX_ORDER) && (order >= pageblock_order))
686 return page + (1 << order);
687 }
Badari Pulavarty5c755e92008-07-23 21:28:19 -0700688
KAMEZAWA Hiroyuki0dcc48c2010-09-09 16:38:01 -0700689 return page + pageblock_nr_pages;
Badari Pulavarty5c755e92008-07-23 21:28:19 -0700690}
691
692/* Checks if this range of memory is likely to be hot-removable. */
693int is_mem_section_removable(unsigned long start_pfn, unsigned long nr_pages)
694{
Badari Pulavarty5c755e92008-07-23 21:28:19 -0700695 struct page *page = pfn_to_page(start_pfn);
696 struct page *end_page = page + nr_pages;
697
698 /* Check the starting page of each pageblock within the range */
699 for (; page < end_page; page = next_active_pageblock(page)) {
KAMEZAWA Hiroyuki49ac8252010-10-26 14:21:30 -0700700 if (!is_pageblock_removable_nolock(page))
Badari Pulavarty5c755e92008-07-23 21:28:19 -0700701 return 0;
KAMEZAWA Hiroyuki49ac8252010-10-26 14:21:30 -0700702 cond_resched();
Badari Pulavarty5c755e92008-07-23 21:28:19 -0700703 }
704
705 /* All pageblocks in the memory block are likely to be hot-removable */
706 return 1;
707}
708
709/*
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700710 * Confirm all pages in a range [start, end) is belongs to the same zone.
711 */
712static int test_pages_in_a_zone(unsigned long start_pfn, unsigned long end_pfn)
713{
714 unsigned long pfn;
715 struct zone *zone = NULL;
716 struct page *page;
717 int i;
718 for (pfn = start_pfn;
719 pfn < end_pfn;
720 pfn += MAX_ORDER_NR_PAGES) {
721 i = 0;
722 /* This is just a CONFIG_HOLES_IN_ZONE check.*/
723 while ((i < MAX_ORDER_NR_PAGES) && !pfn_valid_within(pfn + i))
724 i++;
725 if (i == MAX_ORDER_NR_PAGES)
726 continue;
727 page = pfn_to_page(pfn + i);
728 if (zone && page_zone(page) != zone)
729 return 0;
730 zone = page_zone(page);
731 }
732 return 1;
733}
734
735/*
736 * Scanning pfn is much easier than scanning lru list.
737 * Scan pfn from start to end and Find LRU page.
738 */
Andrew Morton7bbc0902010-10-26 14:22:05 -0700739static unsigned long scan_lru_pages(unsigned long start, unsigned long end)
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700740{
741 unsigned long pfn;
742 struct page *page;
743 for (pfn = start; pfn < end; pfn++) {
744 if (pfn_valid(pfn)) {
745 page = pfn_to_page(pfn);
746 if (PageLRU(page))
747 return pfn;
748 }
749 }
750 return 0;
751}
752
753static struct page *
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800754hotremove_migrate_alloc(struct page *page, unsigned long private, int **x)
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700755{
Hugh Dickins3c1d4372009-01-06 14:39:23 -0800756 /* This should be improooooved!! */
757 return alloc_page(GFP_HIGHUSER_MOVABLE);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700758}
759
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700760#define NR_OFFLINE_AT_ONCE_PAGES (256)
761static int
762do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
763{
764 unsigned long pfn;
765 struct page *page;
766 int move_pages = NR_OFFLINE_AT_ONCE_PAGES;
767 int not_managed = 0;
768 int ret = 0;
769 LIST_HEAD(source);
770
771 for (pfn = start_pfn; pfn < end_pfn && move_pages > 0; pfn++) {
772 if (!pfn_valid(pfn))
773 continue;
774 page = pfn_to_page(pfn);
Konstantin Khlebnikov700c2a42011-05-24 17:12:19 -0700775 if (!get_page_unless_zero(page))
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700776 continue;
777 /*
778 * We can skip free pages. And we can only deal with pages on
779 * LRU.
780 */
Nick Piggin62695a82008-10-18 20:26:09 -0700781 ret = isolate_lru_page(page);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700782 if (!ret) { /* Success */
Konstantin Khlebnikov700c2a42011-05-24 17:12:19 -0700783 put_page(page);
Nick Piggin62695a82008-10-18 20:26:09 -0700784 list_add_tail(&page->lru, &source);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700785 move_pages--;
KOSAKI Motohiro6d9c2852009-12-14 17:58:11 -0800786 inc_zone_page_state(page, NR_ISOLATED_ANON +
787 page_is_file_cache(page));
788
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700789 } else {
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700790#ifdef CONFIG_DEBUG_VM
Wu Fengguang718a3822010-03-10 15:20:43 -0800791 printk(KERN_ALERT "removing pfn %lx from LRU failed\n",
792 pfn);
793 dump_page(page);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700794#endif
Konstantin Khlebnikov700c2a42011-05-24 17:12:19 -0700795 put_page(page);
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300796 /* Because we don't have big zone->lock. we should
Bob Liu809c4442010-10-26 14:22:10 -0700797 check this again here. */
798 if (page_count(page)) {
799 not_managed++;
Bob Liuf3ab2632010-10-26 14:22:10 -0700800 ret = -EBUSY;
Bob Liu809c4442010-10-26 14:22:10 -0700801 break;
802 }
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700803 }
804 }
Bob Liuf3ab2632010-10-26 14:22:10 -0700805 if (!list_empty(&source)) {
806 if (not_managed) {
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700807 putback_lru_pages(&source);
Bob Liuf3ab2632010-10-26 14:22:10 -0700808 goto out;
809 }
810 /* this function returns # of failed pages */
Mel Gorman77f1fe62011-01-13 15:45:57 -0800811 ret = migrate_pages(&source, hotremove_migrate_alloc, 0,
Mel Gorman7f0f2492011-01-13 15:45:58 -0800812 true, true);
Bob Liuf3ab2632010-10-26 14:22:10 -0700813 if (ret)
814 putback_lru_pages(&source);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700815 }
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700816out:
817 return ret;
818}
819
820/*
821 * remove from free_area[] and mark all as Reserved.
822 */
823static int
824offline_isolated_pages_cb(unsigned long start, unsigned long nr_pages,
825 void *data)
826{
827 __offline_isolated_pages(start, start + nr_pages);
828 return 0;
829}
830
831static void
832offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn)
833{
KAMEZAWA Hiroyuki908eedc2009-09-22 16:45:46 -0700834 walk_system_ram_range(start_pfn, end_pfn - start_pfn, NULL,
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700835 offline_isolated_pages_cb);
836}
837
838/*
839 * Check all pages in range, recoreded as memory resource, are isolated.
840 */
841static int
842check_pages_isolated_cb(unsigned long start_pfn, unsigned long nr_pages,
843 void *data)
844{
845 int ret;
846 long offlined = *(long *)data;
847 ret = test_pages_isolated(start_pfn, start_pfn + nr_pages);
848 offlined = nr_pages;
849 if (!ret)
850 *(long *)data += offlined;
851 return ret;
852}
853
854static long
855check_pages_isolated(unsigned long start_pfn, unsigned long end_pfn)
856{
857 long offlined = 0;
858 int ret;
859
KAMEZAWA Hiroyuki908eedc2009-09-22 16:45:46 -0700860 ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn, &offlined,
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700861 check_pages_isolated_cb);
862 if (ret < 0)
863 offlined = (long)ret;
864 return offlined;
865}
866
KOSAKI Motohiro839a4fc2011-05-24 17:11:31 -0700867static int __ref offline_pages(unsigned long start_pfn,
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700868 unsigned long end_pfn, unsigned long timeout)
869{
870 unsigned long pfn, nr_pages, expire;
871 long offlined_pages;
Yasunori Goto7b78d332007-10-21 16:41:36 -0700872 int ret, drain, retry_max, node;
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700873 struct zone *zone;
Yasunori Goto7b78d332007-10-21 16:41:36 -0700874 struct memory_notify arg;
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700875
876 BUG_ON(start_pfn >= end_pfn);
877 /* at least, alignment against pageblock is necessary */
878 if (!IS_ALIGNED(start_pfn, pageblock_nr_pages))
879 return -EINVAL;
880 if (!IS_ALIGNED(end_pfn, pageblock_nr_pages))
881 return -EINVAL;
882 /* This makes hotplug much easier...and readable.
883 we assume this for now. .*/
884 if (!test_pages_in_a_zone(start_pfn, end_pfn))
885 return -EINVAL;
Yasunori Goto7b78d332007-10-21 16:41:36 -0700886
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -0800887 lock_memory_hotplug();
Andi Kleen6ad696d2009-11-17 14:06:22 -0800888
Yasunori Goto7b78d332007-10-21 16:41:36 -0700889 zone = page_zone(pfn_to_page(start_pfn));
890 node = zone_to_nid(zone);
891 nr_pages = end_pfn - start_pfn;
892
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700893 /* set above range as isolated */
894 ret = start_isolate_page_range(start_pfn, end_pfn);
895 if (ret)
Andi Kleen6ad696d2009-11-17 14:06:22 -0800896 goto out;
Yasunori Goto7b78d332007-10-21 16:41:36 -0700897
898 arg.start_pfn = start_pfn;
899 arg.nr_pages = nr_pages;
900 arg.status_change_nid = -1;
901 if (nr_pages >= node_present_pages(node))
902 arg.status_change_nid = node;
903
904 ret = memory_notify(MEM_GOING_OFFLINE, &arg);
905 ret = notifier_to_errno(ret);
906 if (ret)
907 goto failed_removal;
908
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700909 pfn = start_pfn;
910 expire = jiffies + timeout;
911 drain = 0;
912 retry_max = 5;
913repeat:
914 /* start memory hot removal */
915 ret = -EAGAIN;
916 if (time_after(jiffies, expire))
917 goto failed_removal;
918 ret = -EINTR;
919 if (signal_pending(current))
920 goto failed_removal;
921 ret = 0;
922 if (drain) {
923 lru_add_drain_all();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700924 cond_resched();
Christoph Lameter9f8f2172008-02-04 22:29:11 -0800925 drain_all_pages();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700926 }
927
928 pfn = scan_lru_pages(start_pfn, end_pfn);
929 if (pfn) { /* We have page on LRU */
930 ret = do_migrate_range(pfn, end_pfn);
931 if (!ret) {
932 drain = 1;
933 goto repeat;
934 } else {
935 if (ret < 0)
936 if (--retry_max == 0)
937 goto failed_removal;
938 yield();
939 drain = 1;
940 goto repeat;
941 }
942 }
943 /* drain all zone's lru pagevec, this is asyncronous... */
944 lru_add_drain_all();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700945 yield();
946 /* drain pcp pages , this is synchrouns. */
Christoph Lameter9f8f2172008-02-04 22:29:11 -0800947 drain_all_pages();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700948 /* check again */
949 offlined_pages = check_pages_isolated(start_pfn, end_pfn);
950 if (offlined_pages < 0) {
951 ret = -EBUSY;
952 goto failed_removal;
953 }
954 printk(KERN_INFO "Offlined Pages %ld\n", offlined_pages);
955 /* Ok, all of our target is islaoted.
956 We cannot do rollback at this point. */
957 offline_isolated_pages(start_pfn, end_pfn);
KAMEZAWA Hiroyukidbc0e4c2007-11-14 16:59:12 -0800958 /* reset pagetype flags and makes migrate type to be MOVABLE */
959 undo_isolate_page_range(start_pfn, end_pfn);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700960 /* removal success */
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700961 zone->present_pages -= offlined_pages;
962 zone->zone_pgdat->node_present_pages -= offlined_pages;
963 totalram_pages -= offlined_pages;
Yasunori Goto7b78d332007-10-21 16:41:36 -0700964
KOSAKI Motohiro1b79acc2011-05-24 17:11:32 -0700965 init_per_zone_wmark_min();
966
David Rientjes8fe23e02009-12-14 17:58:33 -0800967 if (!node_present_pages(node)) {
968 node_clear_state(node, N_HIGH_MEMORY);
969 kswapd_stop(node);
970 }
Minchan Kimbce73942009-06-16 15:32:50 -0700971
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700972 vm_total_pages = nr_free_pagecache_pages();
973 writeback_set_ratelimit();
Yasunori Goto7b78d332007-10-21 16:41:36 -0700974
975 memory_notify(MEM_OFFLINE, &arg);
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -0800976 unlock_memory_hotplug();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700977 return 0;
978
979failed_removal:
980 printk(KERN_INFO "memory offlining %lx to %lx failed\n",
981 start_pfn, end_pfn);
Yasunori Goto7b78d332007-10-21 16:41:36 -0700982 memory_notify(MEM_CANCEL_OFFLINE, &arg);
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700983 /* pushback to free area */
984 undo_isolate_page_range(start_pfn, end_pfn);
Yasunori Goto7b78d332007-10-21 16:41:36 -0700985
Andi Kleen6ad696d2009-11-17 14:06:22 -0800986out:
KOSAKI Motohiro20d6c962010-12-02 14:31:19 -0800987 unlock_memory_hotplug();
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -0700988 return ret;
989}
Badari Pulavarty71088782008-10-18 20:25:58 -0700990
991int remove_memory(u64 start, u64 size)
992{
993 unsigned long start_pfn, end_pfn;
994
995 start_pfn = PFN_DOWN(start);
996 end_pfn = start_pfn + PFN_DOWN(size);
997 return offline_pages(start_pfn, end_pfn, 120 * HZ);
998}
KAMEZAWA Hiroyuki48e94192007-10-16 01:26:14 -0700999#else
1000int remove_memory(u64 start, u64 size)
1001{
1002 return -EINVAL;
1003}
KAMEZAWA Hiroyuki0c0e6192007-10-16 01:26:12 -07001004#endif /* CONFIG_MEMORY_HOTREMOVE */
Badari Pulavarty71088782008-10-18 20:25:58 -07001005EXPORT_SYMBOL_GPL(remove_memory);