Commit | Line | Data |
---|---|---|
3947be19 DH |
1 | /* |
2 | * linux/mm/memory_hotplug.c | |
3 | * | |
4 | * Copyright (C) | |
5 | */ | |
6 | ||
3947be19 DH |
7 | #include <linux/stddef.h> |
8 | #include <linux/mm.h> | |
9 | #include <linux/swap.h> | |
10 | #include <linux/interrupt.h> | |
11 | #include <linux/pagemap.h> | |
12 | #include <linux/bootmem.h> | |
13 | #include <linux/compiler.h> | |
b95f1b31 | 14 | #include <linux/export.h> |
3947be19 | 15 | #include <linux/pagevec.h> |
2d1d43f6 | 16 | #include <linux/writeback.h> |
3947be19 DH |
17 | #include <linux/slab.h> |
18 | #include <linux/sysctl.h> | |
19 | #include <linux/cpu.h> | |
20 | #include <linux/memory.h> | |
21 | #include <linux/memory_hotplug.h> | |
22 | #include <linux/highmem.h> | |
23 | #include <linux/vmalloc.h> | |
0a547039 | 24 | #include <linux/ioport.h> |
0c0e6195 KH |
25 | #include <linux/delay.h> |
26 | #include <linux/migrate.h> | |
27 | #include <linux/page-isolation.h> | |
71088785 | 28 | #include <linux/pfn.h> |
6ad696d2 | 29 | #include <linux/suspend.h> |
6d9c285a | 30 | #include <linux/mm_inline.h> |
d96ae530 | 31 | #include <linux/firmware-map.h> |
3947be19 DH |
32 | |
33 | #include <asm/tlbflush.h> | |
34 | ||
1e5ad9a3 AB |
35 | #include "internal.h" |
36 | ||
9d0ad8ca DK |
37 | /* |
38 | * online_page_callback contains pointer to current page onlining function. | |
39 | * Initially it is generic_online_page(). If it is required it could be | |
40 | * changed by calling set_online_page_callback() for callback registration | |
41 | * and restore_online_page_callback() for generic callback restore. | |
42 | */ | |
43 | ||
44 | static void generic_online_page(struct page *page); | |
45 | ||
46 | static online_page_callback_t online_page_callback = generic_online_page; | |
47 | ||
20d6c96b KM |
48 | DEFINE_MUTEX(mem_hotplug_mutex); |
49 | ||
50 | void lock_memory_hotplug(void) | |
51 | { | |
52 | mutex_lock(&mem_hotplug_mutex); | |
53 | ||
54 | /* for exclusive hibernation if CONFIG_HIBERNATION=y */ | |
55 | lock_system_sleep(); | |
56 | } | |
57 | ||
58 | void unlock_memory_hotplug(void) | |
59 | { | |
60 | unlock_system_sleep(); | |
61 | mutex_unlock(&mem_hotplug_mutex); | |
62 | } | |
63 | ||
64 | ||
45e0b78b KM |
65 | /* add this memory to iomem resource */ |
66 | static struct resource *register_memory_resource(u64 start, u64 size) | |
67 | { | |
68 | struct resource *res; | |
69 | res = kzalloc(sizeof(struct resource), GFP_KERNEL); | |
70 | BUG_ON(!res); | |
71 | ||
72 | res->name = "System RAM"; | |
73 | res->start = start; | |
74 | res->end = start + size - 1; | |
887c3cb1 | 75 | res->flags = IORESOURCE_MEM | IORESOURCE_BUSY; |
45e0b78b | 76 | if (request_resource(&iomem_resource, res) < 0) { |
a62e2f4f | 77 | printk("System RAM resource %pR cannot be added\n", res); |
45e0b78b KM |
78 | kfree(res); |
79 | res = NULL; | |
80 | } | |
81 | return res; | |
82 | } | |
83 | ||
84 | static void release_memory_resource(struct resource *res) | |
85 | { | |
86 | if (!res) | |
87 | return; | |
88 | release_resource(res); | |
89 | kfree(res); | |
90 | return; | |
91 | } | |
92 | ||
53947027 | 93 | #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE |
04753278 | 94 | #ifndef CONFIG_SPARSEMEM_VMEMMAP |
5f24ce5f AA |
95 | static void get_page_bootmem(unsigned long info, struct page *page, |
96 | unsigned long type) | |
04753278 | 97 | { |
5f24ce5f | 98 | page->lru.next = (struct list_head *) type; |
04753278 YG |
99 | SetPagePrivate(page); |
100 | set_page_private(page, info); | |
101 | atomic_inc(&page->_count); | |
102 | } | |
103 | ||
23ce932a RM |
104 | /* reference to __meminit __free_pages_bootmem is valid |
105 | * so use __ref to tell modpost not to generate a warning */ | |
106 | void __ref put_page_bootmem(struct page *page) | |
04753278 | 107 | { |
5f24ce5f | 108 | unsigned long type; |
04753278 | 109 | |
5f24ce5f AA |
110 | type = (unsigned long) page->lru.next; |
111 | BUG_ON(type < MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE || | |
112 | type > MEMORY_HOTPLUG_MAX_BOOTMEM_TYPE); | |
04753278 YG |
113 | |
114 | if (atomic_dec_return(&page->_count) == 1) { | |
115 | ClearPagePrivate(page); | |
116 | set_page_private(page, 0); | |
5f24ce5f | 117 | INIT_LIST_HEAD(&page->lru); |
04753278 YG |
118 | __free_pages_bootmem(page, 0); |
119 | } | |
120 | ||
121 | } | |
122 | ||
d92bc318 | 123 | static void register_page_bootmem_info_section(unsigned long start_pfn) |
04753278 YG |
124 | { |
125 | unsigned long *usemap, mapsize, section_nr, i; | |
126 | struct mem_section *ms; | |
127 | struct page *page, *memmap; | |
128 | ||
04753278 YG |
129 | section_nr = pfn_to_section_nr(start_pfn); |
130 | ms = __nr_to_section(section_nr); | |
131 | ||
132 | /* Get section's memmap address */ | |
133 | memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr); | |
134 | ||
135 | /* | |
136 | * Get page for the memmap's phys address | |
137 | * XXX: need more consideration for sparse_vmemmap... | |
138 | */ | |
139 | page = virt_to_page(memmap); | |
140 | mapsize = sizeof(struct page) * PAGES_PER_SECTION; | |
141 | mapsize = PAGE_ALIGN(mapsize) >> PAGE_SHIFT; | |
142 | ||
143 | /* remember memmap's page */ | |
144 | for (i = 0; i < mapsize; i++, page++) | |
145 | get_page_bootmem(section_nr, page, SECTION_INFO); | |
146 | ||
147 | usemap = __nr_to_section(section_nr)->pageblock_flags; | |
148 | page = virt_to_page(usemap); | |
149 | ||
150 | mapsize = PAGE_ALIGN(usemap_size()) >> PAGE_SHIFT; | |
151 | ||
152 | for (i = 0; i < mapsize; i++, page++) | |
af370fb8 | 153 | get_page_bootmem(section_nr, page, MIX_SECTION_INFO); |
04753278 YG |
154 | |
155 | } | |
156 | ||
157 | void register_page_bootmem_info_node(struct pglist_data *pgdat) | |
158 | { | |
159 | unsigned long i, pfn, end_pfn, nr_pages; | |
160 | int node = pgdat->node_id; | |
161 | struct page *page; | |
162 | struct zone *zone; | |
163 | ||
164 | nr_pages = PAGE_ALIGN(sizeof(struct pglist_data)) >> PAGE_SHIFT; | |
165 | page = virt_to_page(pgdat); | |
166 | ||
167 | for (i = 0; i < nr_pages; i++, page++) | |
168 | get_page_bootmem(node, page, NODE_INFO); | |
169 | ||
170 | zone = &pgdat->node_zones[0]; | |
171 | for (; zone < pgdat->node_zones + MAX_NR_ZONES - 1; zone++) { | |
172 | if (zone->wait_table) { | |
173 | nr_pages = zone->wait_table_hash_nr_entries | |
174 | * sizeof(wait_queue_head_t); | |
175 | nr_pages = PAGE_ALIGN(nr_pages) >> PAGE_SHIFT; | |
176 | page = virt_to_page(zone->wait_table); | |
177 | ||
178 | for (i = 0; i < nr_pages; i++, page++) | |
179 | get_page_bootmem(node, page, NODE_INFO); | |
180 | } | |
181 | } | |
182 | ||
183 | pfn = pgdat->node_start_pfn; | |
184 | end_pfn = pfn + pgdat->node_spanned_pages; | |
185 | ||
186 | /* register_section info */ | |
f14851af | 187 | for (; pfn < end_pfn; pfn += PAGES_PER_SECTION) { |
188 | /* | |
189 | * Some platforms can assign the same pfn to multiple nodes - on | |
190 | * node0 as well as nodeN. To avoid registering a pfn against | |
191 | * multiple nodes we check that this pfn does not already | |
192 | * reside in some other node. | |
193 | */ | |
194 | if (pfn_valid(pfn) && (pfn_to_nid(pfn) == node)) | |
195 | register_page_bootmem_info_section(pfn); | |
196 | } | |
04753278 YG |
197 | } |
198 | #endif /* !CONFIG_SPARSEMEM_VMEMMAP */ | |
199 | ||
76cdd58e HC |
200 | static void grow_zone_span(struct zone *zone, unsigned long start_pfn, |
201 | unsigned long end_pfn) | |
202 | { | |
203 | unsigned long old_zone_end_pfn; | |
204 | ||
205 | zone_span_writelock(zone); | |
206 | ||
207 | old_zone_end_pfn = zone->zone_start_pfn + zone->spanned_pages; | |
712cd386 | 208 | if (!zone->spanned_pages || start_pfn < zone->zone_start_pfn) |
76cdd58e HC |
209 | zone->zone_start_pfn = start_pfn; |
210 | ||
211 | zone->spanned_pages = max(old_zone_end_pfn, end_pfn) - | |
212 | zone->zone_start_pfn; | |
213 | ||
214 | zone_span_writeunlock(zone); | |
215 | } | |
216 | ||
511c2aba LJ |
217 | static void resize_zone(struct zone *zone, unsigned long start_pfn, |
218 | unsigned long end_pfn) | |
219 | { | |
220 | zone_span_writelock(zone); | |
221 | ||
222 | zone->zone_start_pfn = start_pfn; | |
223 | zone->spanned_pages = end_pfn - start_pfn; | |
224 | ||
225 | zone_span_writeunlock(zone); | |
226 | } | |
227 | ||
228 | static void fix_zone_id(struct zone *zone, unsigned long start_pfn, | |
229 | unsigned long end_pfn) | |
230 | { | |
231 | enum zone_type zid = zone_idx(zone); | |
232 | int nid = zone->zone_pgdat->node_id; | |
233 | unsigned long pfn; | |
234 | ||
235 | for (pfn = start_pfn; pfn < end_pfn; pfn++) | |
236 | set_page_links(pfn_to_page(pfn), zid, nid, pfn); | |
237 | } | |
238 | ||
239 | static int move_pfn_range_left(struct zone *z1, struct zone *z2, | |
240 | unsigned long start_pfn, unsigned long end_pfn) | |
241 | { | |
242 | unsigned long flags; | |
243 | ||
244 | pgdat_resize_lock(z1->zone_pgdat, &flags); | |
245 | ||
246 | /* can't move pfns which are higher than @z2 */ | |
247 | if (end_pfn > z2->zone_start_pfn + z2->spanned_pages) | |
248 | goto out_fail; | |
249 | /* the move out part mast at the left most of @z2 */ | |
250 | if (start_pfn > z2->zone_start_pfn) | |
251 | goto out_fail; | |
252 | /* must included/overlap */ | |
253 | if (end_pfn <= z2->zone_start_pfn) | |
254 | goto out_fail; | |
255 | ||
256 | resize_zone(z1, z1->zone_start_pfn, end_pfn); | |
257 | resize_zone(z2, end_pfn, z2->zone_start_pfn + z2->spanned_pages); | |
258 | ||
259 | pgdat_resize_unlock(z1->zone_pgdat, &flags); | |
260 | ||
261 | fix_zone_id(z1, start_pfn, end_pfn); | |
262 | ||
263 | return 0; | |
264 | out_fail: | |
265 | pgdat_resize_unlock(z1->zone_pgdat, &flags); | |
266 | return -1; | |
267 | } | |
268 | ||
269 | static int move_pfn_range_right(struct zone *z1, struct zone *z2, | |
270 | unsigned long start_pfn, unsigned long end_pfn) | |
271 | { | |
272 | unsigned long flags; | |
273 | ||
274 | pgdat_resize_lock(z1->zone_pgdat, &flags); | |
275 | ||
276 | /* can't move pfns which are lower than @z1 */ | |
277 | if (z1->zone_start_pfn > start_pfn) | |
278 | goto out_fail; | |
279 | /* the move out part mast at the right most of @z1 */ | |
280 | if (z1->zone_start_pfn + z1->spanned_pages > end_pfn) | |
281 | goto out_fail; | |
282 | /* must included/overlap */ | |
283 | if (start_pfn >= z1->zone_start_pfn + z1->spanned_pages) | |
284 | goto out_fail; | |
285 | ||
286 | resize_zone(z1, z1->zone_start_pfn, start_pfn); | |
287 | resize_zone(z2, start_pfn, z2->zone_start_pfn + z2->spanned_pages); | |
288 | ||
289 | pgdat_resize_unlock(z1->zone_pgdat, &flags); | |
290 | ||
291 | fix_zone_id(z2, start_pfn, end_pfn); | |
292 | ||
293 | return 0; | |
294 | out_fail: | |
295 | pgdat_resize_unlock(z1->zone_pgdat, &flags); | |
296 | return -1; | |
297 | } | |
298 | ||
76cdd58e HC |
299 | static void grow_pgdat_span(struct pglist_data *pgdat, unsigned long start_pfn, |
300 | unsigned long end_pfn) | |
301 | { | |
302 | unsigned long old_pgdat_end_pfn = | |
303 | pgdat->node_start_pfn + pgdat->node_spanned_pages; | |
304 | ||
712cd386 | 305 | if (!pgdat->node_spanned_pages || start_pfn < pgdat->node_start_pfn) |
76cdd58e HC |
306 | pgdat->node_start_pfn = start_pfn; |
307 | ||
308 | pgdat->node_spanned_pages = max(old_pgdat_end_pfn, end_pfn) - | |
309 | pgdat->node_start_pfn; | |
310 | } | |
311 | ||
31168481 | 312 | static int __meminit __add_zone(struct zone *zone, unsigned long phys_start_pfn) |
3947be19 DH |
313 | { |
314 | struct pglist_data *pgdat = zone->zone_pgdat; | |
315 | int nr_pages = PAGES_PER_SECTION; | |
316 | int nid = pgdat->node_id; | |
317 | int zone_type; | |
76cdd58e | 318 | unsigned long flags; |
3947be19 DH |
319 | |
320 | zone_type = zone - pgdat->node_zones; | |
76cdd58e HC |
321 | if (!zone->wait_table) { |
322 | int ret; | |
323 | ||
324 | ret = init_currently_empty_zone(zone, phys_start_pfn, | |
325 | nr_pages, MEMMAP_HOTPLUG); | |
326 | if (ret) | |
327 | return ret; | |
328 | } | |
329 | pgdat_resize_lock(zone->zone_pgdat, &flags); | |
330 | grow_zone_span(zone, phys_start_pfn, phys_start_pfn + nr_pages); | |
331 | grow_pgdat_span(zone->zone_pgdat, phys_start_pfn, | |
332 | phys_start_pfn + nr_pages); | |
333 | pgdat_resize_unlock(zone->zone_pgdat, &flags); | |
a2f3aa02 DH |
334 | memmap_init_zone(nr_pages, nid, zone_type, |
335 | phys_start_pfn, MEMMAP_HOTPLUG); | |
718127cc | 336 | return 0; |
3947be19 DH |
337 | } |
338 | ||
c04fc586 GH |
339 | static int __meminit __add_section(int nid, struct zone *zone, |
340 | unsigned long phys_start_pfn) | |
3947be19 | 341 | { |
3947be19 | 342 | int nr_pages = PAGES_PER_SECTION; |
3947be19 DH |
343 | int ret; |
344 | ||
ebd15302 KH |
345 | if (pfn_valid(phys_start_pfn)) |
346 | return -EEXIST; | |
347 | ||
0b0acbec | 348 | ret = sparse_add_one_section(zone, phys_start_pfn, nr_pages); |
3947be19 DH |
349 | |
350 | if (ret < 0) | |
351 | return ret; | |
352 | ||
718127cc YG |
353 | ret = __add_zone(zone, phys_start_pfn); |
354 | ||
355 | if (ret < 0) | |
356 | return ret; | |
357 | ||
c04fc586 | 358 | return register_new_memory(nid, __pfn_to_section(phys_start_pfn)); |
3947be19 DH |
359 | } |
360 | ||
0c0a4a51 YG |
361 | #ifdef CONFIG_SPARSEMEM_VMEMMAP |
362 | static int __remove_section(struct zone *zone, struct mem_section *ms) | |
363 | { | |
364 | /* | |
365 | * XXX: Freeing memmap with vmemmap is not implement yet. | |
366 | * This should be removed later. | |
367 | */ | |
368 | return -EBUSY; | |
369 | } | |
370 | #else | |
ea01ea93 BP |
371 | static int __remove_section(struct zone *zone, struct mem_section *ms) |
372 | { | |
373 | unsigned long flags; | |
374 | struct pglist_data *pgdat = zone->zone_pgdat; | |
375 | int ret = -EINVAL; | |
376 | ||
377 | if (!valid_section(ms)) | |
378 | return ret; | |
379 | ||
380 | ret = unregister_memory_section(ms); | |
381 | if (ret) | |
382 | return ret; | |
383 | ||
384 | pgdat_resize_lock(pgdat, &flags); | |
385 | sparse_remove_one_section(zone, ms); | |
386 | pgdat_resize_unlock(pgdat, &flags); | |
387 | return 0; | |
388 | } | |
0c0a4a51 | 389 | #endif |
ea01ea93 | 390 | |
3947be19 DH |
391 | /* |
392 | * Reasonably generic function for adding memory. It is | |
393 | * expected that archs that support memory hotplug will | |
394 | * call this function after deciding the zone to which to | |
395 | * add the new pages. | |
396 | */ | |
c04fc586 GH |
397 | int __ref __add_pages(int nid, struct zone *zone, unsigned long phys_start_pfn, |
398 | unsigned long nr_pages) | |
3947be19 DH |
399 | { |
400 | unsigned long i; | |
401 | int err = 0; | |
6f712711 KH |
402 | int start_sec, end_sec; |
403 | /* during initialize mem_map, align hot-added range to section */ | |
404 | start_sec = pfn_to_section_nr(phys_start_pfn); | |
405 | end_sec = pfn_to_section_nr(phys_start_pfn + nr_pages - 1); | |
3947be19 | 406 | |
6f712711 | 407 | for (i = start_sec; i <= end_sec; i++) { |
c04fc586 | 408 | err = __add_section(nid, zone, i << PFN_SECTION_SHIFT); |
3947be19 | 409 | |
6f712711 | 410 | /* |
183ff22b | 411 | * EEXIST is finally dealt with by ioresource collision |
6f712711 KH |
412 | * check. see add_memory() => register_memory_resource() |
413 | * Warning will be printed if there is collision. | |
bed120c6 JS |
414 | */ |
415 | if (err && (err != -EEXIST)) | |
3947be19 | 416 | break; |
6f712711 | 417 | err = 0; |
3947be19 DH |
418 | } |
419 | ||
420 | return err; | |
421 | } | |
bed120c6 | 422 | EXPORT_SYMBOL_GPL(__add_pages); |
3947be19 | 423 | |
ea01ea93 BP |
424 | /** |
425 | * __remove_pages() - remove sections of pages from a zone | |
426 | * @zone: zone from which pages need to be removed | |
427 | * @phys_start_pfn: starting pageframe (must be aligned to start of a section) | |
428 | * @nr_pages: number of pages to remove (must be multiple of section size) | |
429 | * | |
430 | * Generic helper function to remove section mappings and sysfs entries | |
431 | * for the section of the memory we are removing. Caller needs to make | |
432 | * sure that pages are marked reserved and zones are adjust properly by | |
433 | * calling offline_pages(). | |
434 | */ | |
435 | int __remove_pages(struct zone *zone, unsigned long phys_start_pfn, | |
436 | unsigned long nr_pages) | |
437 | { | |
438 | unsigned long i, ret = 0; | |
439 | int sections_to_remove; | |
440 | ||
441 | /* | |
442 | * We can only remove entire sections | |
443 | */ | |
444 | BUG_ON(phys_start_pfn & ~PAGE_SECTION_MASK); | |
445 | BUG_ON(nr_pages % PAGES_PER_SECTION); | |
446 | ||
d760afd4 YI |
447 | release_mem_region(phys_start_pfn << PAGE_SHIFT, nr_pages * PAGE_SIZE); |
448 | ||
ea01ea93 BP |
449 | sections_to_remove = nr_pages / PAGES_PER_SECTION; |
450 | for (i = 0; i < sections_to_remove; i++) { | |
451 | unsigned long pfn = phys_start_pfn + i*PAGES_PER_SECTION; | |
452 | ret = __remove_section(zone, __pfn_to_section(pfn)); | |
453 | if (ret) | |
454 | break; | |
455 | } | |
456 | return ret; | |
457 | } | |
458 | EXPORT_SYMBOL_GPL(__remove_pages); | |
459 | ||
9d0ad8ca DK |
460 | int set_online_page_callback(online_page_callback_t callback) |
461 | { | |
462 | int rc = -EINVAL; | |
463 | ||
464 | lock_memory_hotplug(); | |
465 | ||
466 | if (online_page_callback == generic_online_page) { | |
467 | online_page_callback = callback; | |
468 | rc = 0; | |
469 | } | |
470 | ||
471 | unlock_memory_hotplug(); | |
472 | ||
473 | return rc; | |
474 | } | |
475 | EXPORT_SYMBOL_GPL(set_online_page_callback); | |
476 | ||
477 | int restore_online_page_callback(online_page_callback_t callback) | |
478 | { | |
479 | int rc = -EINVAL; | |
480 | ||
481 | lock_memory_hotplug(); | |
482 | ||
483 | if (online_page_callback == callback) { | |
484 | online_page_callback = generic_online_page; | |
485 | rc = 0; | |
486 | } | |
487 | ||
488 | unlock_memory_hotplug(); | |
489 | ||
490 | return rc; | |
491 | } | |
492 | EXPORT_SYMBOL_GPL(restore_online_page_callback); | |
493 | ||
494 | void __online_page_set_limits(struct page *page) | |
180c06ef | 495 | { |
4738e1b9 JB |
496 | unsigned long pfn = page_to_pfn(page); |
497 | ||
4738e1b9 JB |
498 | if (pfn >= num_physpages) |
499 | num_physpages = pfn + 1; | |
9d0ad8ca DK |
500 | } |
501 | EXPORT_SYMBOL_GPL(__online_page_set_limits); | |
502 | ||
503 | void __online_page_increment_counters(struct page *page) | |
504 | { | |
505 | totalram_pages++; | |
180c06ef JF |
506 | |
507 | #ifdef CONFIG_HIGHMEM | |
508 | if (PageHighMem(page)) | |
509 | totalhigh_pages++; | |
510 | #endif | |
9d0ad8ca DK |
511 | } |
512 | EXPORT_SYMBOL_GPL(__online_page_increment_counters); | |
180c06ef | 513 | |
9d0ad8ca DK |
514 | void __online_page_free(struct page *page) |
515 | { | |
180c06ef JF |
516 | ClearPageReserved(page); |
517 | init_page_count(page); | |
518 | __free_page(page); | |
519 | } | |
9d0ad8ca DK |
520 | EXPORT_SYMBOL_GPL(__online_page_free); |
521 | ||
522 | static void generic_online_page(struct page *page) | |
523 | { | |
524 | __online_page_set_limits(page); | |
525 | __online_page_increment_counters(page); | |
526 | __online_page_free(page); | |
527 | } | |
180c06ef | 528 | |
75884fb1 KH |
529 | static int online_pages_range(unsigned long start_pfn, unsigned long nr_pages, |
530 | void *arg) | |
3947be19 DH |
531 | { |
532 | unsigned long i; | |
75884fb1 KH |
533 | unsigned long onlined_pages = *(unsigned long *)arg; |
534 | struct page *page; | |
535 | if (PageReserved(pfn_to_page(start_pfn))) | |
536 | for (i = 0; i < nr_pages; i++) { | |
537 | page = pfn_to_page(start_pfn + i); | |
9d0ad8ca | 538 | (*online_page_callback)(page); |
75884fb1 KH |
539 | onlined_pages++; |
540 | } | |
541 | *(unsigned long *)arg = onlined_pages; | |
542 | return 0; | |
543 | } | |
544 | ||
d9713679 LJ |
545 | /* check which state of node_states will be changed when online memory */ |
546 | static void node_states_check_changes_online(unsigned long nr_pages, | |
547 | struct zone *zone, struct memory_notify *arg) | |
548 | { | |
549 | int nid = zone_to_nid(zone); | |
550 | enum zone_type zone_last = ZONE_NORMAL; | |
551 | ||
552 | /* | |
553 | * If we have HIGHMEM, node_states[N_NORMAL_MEMORY] contains nodes | |
554 | * which have 0...ZONE_NORMAL, set zone_last to ZONE_NORMAL. | |
555 | * | |
556 | * If we don't have HIGHMEM, node_states[N_NORMAL_MEMORY] contains nodes | |
557 | * which have 0...ZONE_MOVABLE, set zone_last to ZONE_MOVABLE. | |
558 | */ | |
559 | if (N_HIGH_MEMORY == N_NORMAL_MEMORY) | |
560 | zone_last = ZONE_MOVABLE; | |
561 | ||
562 | /* | |
563 | * if the memory to be online is in a zone of 0...zone_last, and | |
564 | * the zones of 0...zone_last don't have memory before online, we will | |
565 | * need to set the node to node_states[N_NORMAL_MEMORY] after | |
566 | * the memory is online. | |
567 | */ | |
568 | if (zone_idx(zone) <= zone_last && !node_state(nid, N_NORMAL_MEMORY)) | |
569 | arg->status_change_nid_normal = nid; | |
570 | else | |
571 | arg->status_change_nid_normal = -1; | |
572 | ||
573 | /* | |
574 | * if the node don't have memory befor online, we will need to | |
575 | * set the node to node_states[N_HIGH_MEMORY] after the memory | |
576 | * is online. | |
577 | */ | |
578 | if (!node_state(nid, N_HIGH_MEMORY)) | |
579 | arg->status_change_nid = nid; | |
580 | else | |
581 | arg->status_change_nid = -1; | |
582 | } | |
583 | ||
584 | static void node_states_set_node(int node, struct memory_notify *arg) | |
585 | { | |
586 | if (arg->status_change_nid_normal >= 0) | |
587 | node_set_state(node, N_NORMAL_MEMORY); | |
588 | ||
589 | node_set_state(node, N_HIGH_MEMORY); | |
590 | } | |
591 | ||
75884fb1 | 592 | |
511c2aba | 593 | int __ref online_pages(unsigned long pfn, unsigned long nr_pages, int online_type) |
75884fb1 | 594 | { |
3947be19 DH |
595 | unsigned long onlined_pages = 0; |
596 | struct zone *zone; | |
6811378e | 597 | int need_zonelists_rebuild = 0; |
7b78d335 YG |
598 | int nid; |
599 | int ret; | |
600 | struct memory_notify arg; | |
601 | ||
925268a0 | 602 | lock_memory_hotplug(); |
d9713679 LJ |
603 | /* |
604 | * This doesn't need a lock to do pfn_to_page(). | |
605 | * The section can't be removed here because of the | |
606 | * memory_block->state_mutex. | |
607 | */ | |
608 | zone = page_zone(pfn_to_page(pfn)); | |
609 | ||
511c2aba LJ |
610 | if (online_type == ONLINE_KERNEL && zone_idx(zone) == ZONE_MOVABLE) { |
611 | if (move_pfn_range_left(zone - 1, zone, pfn, pfn + nr_pages)) { | |
612 | unlock_memory_hotplug(); | |
613 | return -1; | |
614 | } | |
615 | } | |
616 | if (online_type == ONLINE_MOVABLE && zone_idx(zone) == ZONE_MOVABLE - 1) { | |
617 | if (move_pfn_range_right(zone, zone + 1, pfn, pfn + nr_pages)) { | |
618 | unlock_memory_hotplug(); | |
619 | return -1; | |
620 | } | |
621 | } | |
622 | ||
623 | /* Previous code may changed the zone of the pfn range */ | |
624 | zone = page_zone(pfn_to_page(pfn)); | |
625 | ||
7b78d335 YG |
626 | arg.start_pfn = pfn; |
627 | arg.nr_pages = nr_pages; | |
d9713679 | 628 | node_states_check_changes_online(nr_pages, zone, &arg); |
7b78d335 YG |
629 | |
630 | nid = page_to_nid(pfn_to_page(pfn)); | |
3947be19 | 631 | |
7b78d335 YG |
632 | ret = memory_notify(MEM_GOING_ONLINE, &arg); |
633 | ret = notifier_to_errno(ret); | |
634 | if (ret) { | |
635 | memory_notify(MEM_CANCEL_ONLINE, &arg); | |
925268a0 | 636 | unlock_memory_hotplug(); |
7b78d335 YG |
637 | return ret; |
638 | } | |
6811378e YG |
639 | /* |
640 | * If this zone is not populated, then it is not in zonelist. | |
641 | * This means the page allocator ignores this zone. | |
642 | * So, zonelist must be updated after online. | |
643 | */ | |
4eaf3f64 | 644 | mutex_lock(&zonelists_mutex); |
6dcd73d7 | 645 | if (!populated_zone(zone)) { |
6811378e | 646 | need_zonelists_rebuild = 1; |
6dcd73d7 WC |
647 | build_all_zonelists(NULL, zone); |
648 | } | |
6811378e | 649 | |
908eedc6 | 650 | ret = walk_system_ram_range(pfn, nr_pages, &onlined_pages, |
75884fb1 | 651 | online_pages_range); |
fd8a4221 | 652 | if (ret) { |
6dcd73d7 WC |
653 | if (need_zonelists_rebuild) |
654 | zone_pcp_reset(zone); | |
4eaf3f64 | 655 | mutex_unlock(&zonelists_mutex); |
a62e2f4f BH |
656 | printk(KERN_DEBUG "online_pages [mem %#010llx-%#010llx] failed\n", |
657 | (unsigned long long) pfn << PAGE_SHIFT, | |
658 | (((unsigned long long) pfn + nr_pages) | |
659 | << PAGE_SHIFT) - 1); | |
fd8a4221 | 660 | memory_notify(MEM_CANCEL_ONLINE, &arg); |
925268a0 | 661 | unlock_memory_hotplug(); |
fd8a4221 GL |
662 | return ret; |
663 | } | |
664 | ||
3947be19 | 665 | zone->present_pages += onlined_pages; |
f2937be5 | 666 | zone->zone_pgdat->node_present_pages += onlined_pages; |
08dff7b7 | 667 | if (onlined_pages) { |
d9713679 | 668 | node_states_set_node(zone_to_nid(zone), &arg); |
08dff7b7 | 669 | if (need_zonelists_rebuild) |
6dcd73d7 | 670 | build_all_zonelists(NULL, NULL); |
08dff7b7 JL |
671 | else |
672 | zone_pcp_update(zone); | |
673 | } | |
3947be19 | 674 | |
4eaf3f64 | 675 | mutex_unlock(&zonelists_mutex); |
1b79acc9 KM |
676 | |
677 | init_per_zone_wmark_min(); | |
678 | ||
08dff7b7 | 679 | if (onlined_pages) |
7ea1530a | 680 | kswapd_run(zone_to_nid(zone)); |
61b13993 | 681 | |
1f522509 | 682 | vm_total_pages = nr_free_pagecache_pages(); |
2f7f24ec | 683 | |
2d1d43f6 | 684 | writeback_set_ratelimit(); |
7b78d335 YG |
685 | |
686 | if (onlined_pages) | |
687 | memory_notify(MEM_ONLINE, &arg); | |
925268a0 | 688 | unlock_memory_hotplug(); |
7b78d335 | 689 | |
3947be19 DH |
690 | return 0; |
691 | } | |
53947027 | 692 | #endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */ |
bc02af93 | 693 | |
e1319331 HS |
694 | /* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */ |
695 | static pg_data_t __ref *hotadd_new_pgdat(int nid, u64 start) | |
9af3c2de YG |
696 | { |
697 | struct pglist_data *pgdat; | |
698 | unsigned long zones_size[MAX_NR_ZONES] = {0}; | |
699 | unsigned long zholes_size[MAX_NR_ZONES] = {0}; | |
700 | unsigned long start_pfn = start >> PAGE_SHIFT; | |
701 | ||
702 | pgdat = arch_alloc_nodedata(nid); | |
703 | if (!pgdat) | |
704 | return NULL; | |
705 | ||
706 | arch_refresh_nodedata(nid, pgdat); | |
707 | ||
708 | /* we can use NODE_DATA(nid) from here */ | |
709 | ||
710 | /* init node's zones as empty zones, we don't have any present pages.*/ | |
9109fb7b | 711 | free_area_init_node(nid, zones_size, start_pfn, zholes_size); |
9af3c2de | 712 | |
959ecc48 KH |
713 | /* |
714 | * The node we allocated has no zone fallback lists. For avoiding | |
715 | * to access not-initialized zonelist, build here. | |
716 | */ | |
f957db4f | 717 | mutex_lock(&zonelists_mutex); |
9adb62a5 | 718 | build_all_zonelists(pgdat, NULL); |
f957db4f | 719 | mutex_unlock(&zonelists_mutex); |
959ecc48 | 720 | |
9af3c2de YG |
721 | return pgdat; |
722 | } | |
723 | ||
724 | static void rollback_node_hotadd(int nid, pg_data_t *pgdat) | |
725 | { | |
726 | arch_refresh_nodedata(nid, NULL); | |
727 | arch_free_nodedata(pgdat); | |
728 | return; | |
729 | } | |
730 | ||
0a547039 | 731 | |
cf23422b | 732 | /* |
733 | * called by cpu_up() to online a node without onlined memory. | |
734 | */ | |
735 | int mem_online_node(int nid) | |
736 | { | |
737 | pg_data_t *pgdat; | |
738 | int ret; | |
739 | ||
20d6c96b | 740 | lock_memory_hotplug(); |
cf23422b | 741 | pgdat = hotadd_new_pgdat(nid, 0); |
7553e8f2 | 742 | if (!pgdat) { |
cf23422b | 743 | ret = -ENOMEM; |
744 | goto out; | |
745 | } | |
746 | node_set_online(nid); | |
747 | ret = register_one_node(nid); | |
748 | BUG_ON(ret); | |
749 | ||
750 | out: | |
20d6c96b | 751 | unlock_memory_hotplug(); |
cf23422b | 752 | return ret; |
753 | } | |
754 | ||
31168481 AV |
755 | /* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */ |
756 | int __ref add_memory(int nid, u64 start, u64 size) | |
bc02af93 | 757 | { |
9af3c2de YG |
758 | pg_data_t *pgdat = NULL; |
759 | int new_pgdat = 0; | |
ebd15302 | 760 | struct resource *res; |
bc02af93 YG |
761 | int ret; |
762 | ||
20d6c96b | 763 | lock_memory_hotplug(); |
6ad696d2 | 764 | |
ebd15302 | 765 | res = register_memory_resource(start, size); |
6ad696d2 | 766 | ret = -EEXIST; |
ebd15302 | 767 | if (!res) |
6ad696d2 | 768 | goto out; |
ebd15302 | 769 | |
9af3c2de YG |
770 | if (!node_online(nid)) { |
771 | pgdat = hotadd_new_pgdat(nid, start); | |
6ad696d2 | 772 | ret = -ENOMEM; |
9af3c2de | 773 | if (!pgdat) |
41b9e2d7 | 774 | goto error; |
9af3c2de | 775 | new_pgdat = 1; |
9af3c2de YG |
776 | } |
777 | ||
bc02af93 YG |
778 | /* call arch's memory hotadd */ |
779 | ret = arch_add_memory(nid, start, size); | |
780 | ||
9af3c2de YG |
781 | if (ret < 0) |
782 | goto error; | |
783 | ||
0fc44159 | 784 | /* we online node here. we can't roll back from here. */ |
9af3c2de YG |
785 | node_set_online(nid); |
786 | ||
0fc44159 YG |
787 | if (new_pgdat) { |
788 | ret = register_one_node(nid); | |
789 | /* | |
790 | * If sysfs file of new node can't create, cpu on the node | |
791 | * can't be hot-added. There is no rollback way now. | |
792 | * So, check by BUG_ON() to catch it reluctantly.. | |
793 | */ | |
794 | BUG_ON(ret); | |
795 | } | |
796 | ||
d96ae530 | 797 | /* create new memmap entry */ |
798 | firmware_map_add_hotplug(start, start + size, "System RAM"); | |
799 | ||
6ad696d2 AK |
800 | goto out; |
801 | ||
9af3c2de YG |
802 | error: |
803 | /* rollback pgdat allocation and others */ | |
804 | if (new_pgdat) | |
805 | rollback_node_hotadd(nid, pgdat); | |
ebd15302 KH |
806 | if (res) |
807 | release_memory_resource(res); | |
9af3c2de | 808 | |
6ad696d2 | 809 | out: |
20d6c96b | 810 | unlock_memory_hotplug(); |
bc02af93 YG |
811 | return ret; |
812 | } | |
813 | EXPORT_SYMBOL_GPL(add_memory); | |
0c0e6195 KH |
814 | |
815 | #ifdef CONFIG_MEMORY_HOTREMOVE | |
5c755e9f BP |
816 | /* |
817 | * A free page on the buddy free lists (not the per-cpu lists) has PageBuddy | |
818 | * set and the size of the free page is given by page_order(). Using this, | |
819 | * the function determines if the pageblock contains only free pages. | |
820 | * Due to buddy contraints, a free page at least the size of a pageblock will | |
821 | * be located at the start of the pageblock | |
822 | */ | |
823 | static inline int pageblock_free(struct page *page) | |
824 | { | |
825 | return PageBuddy(page) && page_order(page) >= pageblock_order; | |
826 | } | |
827 | ||
828 | /* Return the start of the next active pageblock after a given page */ | |
829 | static struct page *next_active_pageblock(struct page *page) | |
830 | { | |
5c755e9f BP |
831 | /* Ensure the starting page is pageblock-aligned */ |
832 | BUG_ON(page_to_pfn(page) & (pageblock_nr_pages - 1)); | |
833 | ||
5c755e9f | 834 | /* If the entire pageblock is free, move to the end of free page */ |
0dcc48c1 KH |
835 | if (pageblock_free(page)) { |
836 | int order; | |
837 | /* be careful. we don't have locks, page_order can be changed.*/ | |
838 | order = page_order(page); | |
839 | if ((order < MAX_ORDER) && (order >= pageblock_order)) | |
840 | return page + (1 << order); | |
841 | } | |
5c755e9f | 842 | |
0dcc48c1 | 843 | return page + pageblock_nr_pages; |
5c755e9f BP |
844 | } |
845 | ||
846 | /* Checks if this range of memory is likely to be hot-removable. */ | |
847 | int is_mem_section_removable(unsigned long start_pfn, unsigned long nr_pages) | |
848 | { | |
5c755e9f BP |
849 | struct page *page = pfn_to_page(start_pfn); |
850 | struct page *end_page = page + nr_pages; | |
851 | ||
852 | /* Check the starting page of each pageblock within the range */ | |
853 | for (; page < end_page; page = next_active_pageblock(page)) { | |
49ac8255 | 854 | if (!is_pageblock_removable_nolock(page)) |
5c755e9f | 855 | return 0; |
49ac8255 | 856 | cond_resched(); |
5c755e9f BP |
857 | } |
858 | ||
859 | /* All pageblocks in the memory block are likely to be hot-removable */ | |
860 | return 1; | |
861 | } | |
862 | ||
0c0e6195 KH |
863 | /* |
864 | * Confirm all pages in a range [start, end) is belongs to the same zone. | |
865 | */ | |
866 | static int test_pages_in_a_zone(unsigned long start_pfn, unsigned long end_pfn) | |
867 | { | |
868 | unsigned long pfn; | |
869 | struct zone *zone = NULL; | |
870 | struct page *page; | |
871 | int i; | |
872 | for (pfn = start_pfn; | |
873 | pfn < end_pfn; | |
874 | pfn += MAX_ORDER_NR_PAGES) { | |
875 | i = 0; | |
876 | /* This is just a CONFIG_HOLES_IN_ZONE check.*/ | |
877 | while ((i < MAX_ORDER_NR_PAGES) && !pfn_valid_within(pfn + i)) | |
878 | i++; | |
879 | if (i == MAX_ORDER_NR_PAGES) | |
880 | continue; | |
881 | page = pfn_to_page(pfn + i); | |
882 | if (zone && page_zone(page) != zone) | |
883 | return 0; | |
884 | zone = page_zone(page); | |
885 | } | |
886 | return 1; | |
887 | } | |
888 | ||
889 | /* | |
890 | * Scanning pfn is much easier than scanning lru list. | |
891 | * Scan pfn from start to end and Find LRU page. | |
892 | */ | |
7bbc0905 | 893 | static unsigned long scan_lru_pages(unsigned long start, unsigned long end) |
0c0e6195 KH |
894 | { |
895 | unsigned long pfn; | |
896 | struct page *page; | |
897 | for (pfn = start; pfn < end; pfn++) { | |
898 | if (pfn_valid(pfn)) { | |
899 | page = pfn_to_page(pfn); | |
900 | if (PageLRU(page)) | |
901 | return pfn; | |
902 | } | |
903 | } | |
904 | return 0; | |
905 | } | |
906 | ||
0c0e6195 KH |
907 | #define NR_OFFLINE_AT_ONCE_PAGES (256) |
908 | static int | |
909 | do_migrate_range(unsigned long start_pfn, unsigned long end_pfn) | |
910 | { | |
911 | unsigned long pfn; | |
912 | struct page *page; | |
913 | int move_pages = NR_OFFLINE_AT_ONCE_PAGES; | |
914 | int not_managed = 0; | |
915 | int ret = 0; | |
916 | LIST_HEAD(source); | |
917 | ||
918 | for (pfn = start_pfn; pfn < end_pfn && move_pages > 0; pfn++) { | |
919 | if (!pfn_valid(pfn)) | |
920 | continue; | |
921 | page = pfn_to_page(pfn); | |
700c2a46 | 922 | if (!get_page_unless_zero(page)) |
0c0e6195 KH |
923 | continue; |
924 | /* | |
925 | * We can skip free pages. And we can only deal with pages on | |
926 | * LRU. | |
927 | */ | |
62695a84 | 928 | ret = isolate_lru_page(page); |
0c0e6195 | 929 | if (!ret) { /* Success */ |
700c2a46 | 930 | put_page(page); |
62695a84 | 931 | list_add_tail(&page->lru, &source); |
0c0e6195 | 932 | move_pages--; |
6d9c285a KM |
933 | inc_zone_page_state(page, NR_ISOLATED_ANON + |
934 | page_is_file_cache(page)); | |
935 | ||
0c0e6195 | 936 | } else { |
0c0e6195 | 937 | #ifdef CONFIG_DEBUG_VM |
718a3821 WF |
938 | printk(KERN_ALERT "removing pfn %lx from LRU failed\n", |
939 | pfn); | |
940 | dump_page(page); | |
0c0e6195 | 941 | #endif |
700c2a46 | 942 | put_page(page); |
25985edc | 943 | /* Because we don't have big zone->lock. we should |
809c4449 BL |
944 | check this again here. */ |
945 | if (page_count(page)) { | |
946 | not_managed++; | |
f3ab2636 | 947 | ret = -EBUSY; |
809c4449 BL |
948 | break; |
949 | } | |
0c0e6195 KH |
950 | } |
951 | } | |
f3ab2636 BL |
952 | if (!list_empty(&source)) { |
953 | if (not_managed) { | |
954 | putback_lru_pages(&source); | |
955 | goto out; | |
956 | } | |
74c08f98 MK |
957 | |
958 | /* | |
959 | * alloc_migrate_target should be improooooved!! | |
960 | * migrate_pages returns # of failed pages. | |
961 | */ | |
962 | ret = migrate_pages(&source, alloc_migrate_target, 0, | |
a6bc32b8 | 963 | true, MIGRATE_SYNC); |
f3ab2636 | 964 | if (ret) |
0c0e6195 | 965 | putback_lru_pages(&source); |
0c0e6195 | 966 | } |
0c0e6195 KH |
967 | out: |
968 | return ret; | |
969 | } | |
970 | ||
971 | /* | |
972 | * remove from free_area[] and mark all as Reserved. | |
973 | */ | |
974 | static int | |
975 | offline_isolated_pages_cb(unsigned long start, unsigned long nr_pages, | |
976 | void *data) | |
977 | { | |
978 | __offline_isolated_pages(start, start + nr_pages); | |
979 | return 0; | |
980 | } | |
981 | ||
982 | static void | |
983 | offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn) | |
984 | { | |
908eedc6 | 985 | walk_system_ram_range(start_pfn, end_pfn - start_pfn, NULL, |
0c0e6195 KH |
986 | offline_isolated_pages_cb); |
987 | } | |
988 | ||
989 | /* | |
990 | * Check all pages in range, recoreded as memory resource, are isolated. | |
991 | */ | |
992 | static int | |
993 | check_pages_isolated_cb(unsigned long start_pfn, unsigned long nr_pages, | |
994 | void *data) | |
995 | { | |
996 | int ret; | |
997 | long offlined = *(long *)data; | |
b023f468 | 998 | ret = test_pages_isolated(start_pfn, start_pfn + nr_pages, true); |
0c0e6195 KH |
999 | offlined = nr_pages; |
1000 | if (!ret) | |
1001 | *(long *)data += offlined; | |
1002 | return ret; | |
1003 | } | |
1004 | ||
1005 | static long | |
1006 | check_pages_isolated(unsigned long start_pfn, unsigned long end_pfn) | |
1007 | { | |
1008 | long offlined = 0; | |
1009 | int ret; | |
1010 | ||
908eedc6 | 1011 | ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn, &offlined, |
0c0e6195 KH |
1012 | check_pages_isolated_cb); |
1013 | if (ret < 0) | |
1014 | offlined = (long)ret; | |
1015 | return offlined; | |
1016 | } | |
1017 | ||
d9713679 LJ |
1018 | /* check which state of node_states will be changed when offline memory */ |
1019 | static void node_states_check_changes_offline(unsigned long nr_pages, | |
1020 | struct zone *zone, struct memory_notify *arg) | |
1021 | { | |
1022 | struct pglist_data *pgdat = zone->zone_pgdat; | |
1023 | unsigned long present_pages = 0; | |
1024 | enum zone_type zt, zone_last = ZONE_NORMAL; | |
1025 | ||
1026 | /* | |
1027 | * If we have HIGHMEM, node_states[N_NORMAL_MEMORY] contains nodes | |
1028 | * which have 0...ZONE_NORMAL, set zone_last to ZONE_NORMAL. | |
1029 | * | |
1030 | * If we don't have HIGHMEM, node_states[N_NORMAL_MEMORY] contains nodes | |
1031 | * which have 0...ZONE_MOVABLE, set zone_last to ZONE_MOVABLE. | |
1032 | */ | |
1033 | if (N_HIGH_MEMORY == N_NORMAL_MEMORY) | |
1034 | zone_last = ZONE_MOVABLE; | |
1035 | ||
1036 | /* | |
1037 | * check whether node_states[N_NORMAL_MEMORY] will be changed. | |
1038 | * If the memory to be offline is in a zone of 0...zone_last, | |
1039 | * and it is the last present memory, 0...zone_last will | |
1040 | * become empty after offline , thus we can determind we will | |
1041 | * need to clear the node from node_states[N_NORMAL_MEMORY]. | |
1042 | */ | |
1043 | for (zt = 0; zt <= zone_last; zt++) | |
1044 | present_pages += pgdat->node_zones[zt].present_pages; | |
1045 | if (zone_idx(zone) <= zone_last && nr_pages >= present_pages) | |
1046 | arg->status_change_nid_normal = zone_to_nid(zone); | |
1047 | else | |
1048 | arg->status_change_nid_normal = -1; | |
1049 | ||
1050 | /* | |
1051 | * node_states[N_HIGH_MEMORY] contains nodes which have 0...ZONE_MOVABLE | |
1052 | */ | |
1053 | zone_last = ZONE_MOVABLE; | |
1054 | ||
1055 | /* | |
1056 | * check whether node_states[N_HIGH_MEMORY] will be changed | |
1057 | * If we try to offline the last present @nr_pages from the node, | |
1058 | * we can determind we will need to clear the node from | |
1059 | * node_states[N_HIGH_MEMORY]. | |
1060 | */ | |
1061 | for (; zt <= zone_last; zt++) | |
1062 | present_pages += pgdat->node_zones[zt].present_pages; | |
1063 | if (nr_pages >= present_pages) | |
1064 | arg->status_change_nid = zone_to_nid(zone); | |
1065 | else | |
1066 | arg->status_change_nid = -1; | |
1067 | } | |
1068 | ||
1069 | static void node_states_clear_node(int node, struct memory_notify *arg) | |
1070 | { | |
1071 | if (arg->status_change_nid_normal >= 0) | |
1072 | node_clear_state(node, N_NORMAL_MEMORY); | |
1073 | ||
1074 | if ((N_HIGH_MEMORY != N_NORMAL_MEMORY) && | |
1075 | (arg->status_change_nid >= 0)) | |
1076 | node_clear_state(node, N_HIGH_MEMORY); | |
1077 | } | |
1078 | ||
a16cee10 | 1079 | static int __ref __offline_pages(unsigned long start_pfn, |
0c0e6195 KH |
1080 | unsigned long end_pfn, unsigned long timeout) |
1081 | { | |
1082 | unsigned long pfn, nr_pages, expire; | |
1083 | long offlined_pages; | |
7b78d335 | 1084 | int ret, drain, retry_max, node; |
0c0e6195 | 1085 | struct zone *zone; |
7b78d335 | 1086 | struct memory_notify arg; |
0c0e6195 KH |
1087 | |
1088 | BUG_ON(start_pfn >= end_pfn); | |
1089 | /* at least, alignment against pageblock is necessary */ | |
1090 | if (!IS_ALIGNED(start_pfn, pageblock_nr_pages)) | |
1091 | return -EINVAL; | |
1092 | if (!IS_ALIGNED(end_pfn, pageblock_nr_pages)) | |
1093 | return -EINVAL; | |
1094 | /* This makes hotplug much easier...and readable. | |
1095 | we assume this for now. .*/ | |
1096 | if (!test_pages_in_a_zone(start_pfn, end_pfn)) | |
1097 | return -EINVAL; | |
7b78d335 | 1098 | |
20d6c96b | 1099 | lock_memory_hotplug(); |
6ad696d2 | 1100 | |
7b78d335 YG |
1101 | zone = page_zone(pfn_to_page(start_pfn)); |
1102 | node = zone_to_nid(zone); | |
1103 | nr_pages = end_pfn - start_pfn; | |
1104 | ||
0c0e6195 | 1105 | /* set above range as isolated */ |
b023f468 WC |
1106 | ret = start_isolate_page_range(start_pfn, end_pfn, |
1107 | MIGRATE_MOVABLE, true); | |
0c0e6195 | 1108 | if (ret) |
6ad696d2 | 1109 | goto out; |
7b78d335 YG |
1110 | |
1111 | arg.start_pfn = start_pfn; | |
1112 | arg.nr_pages = nr_pages; | |
d9713679 | 1113 | node_states_check_changes_offline(nr_pages, zone, &arg); |
7b78d335 YG |
1114 | |
1115 | ret = memory_notify(MEM_GOING_OFFLINE, &arg); | |
1116 | ret = notifier_to_errno(ret); | |
1117 | if (ret) | |
1118 | goto failed_removal; | |
1119 | ||
0c0e6195 KH |
1120 | pfn = start_pfn; |
1121 | expire = jiffies + timeout; | |
1122 | drain = 0; | |
1123 | retry_max = 5; | |
1124 | repeat: | |
1125 | /* start memory hot removal */ | |
1126 | ret = -EAGAIN; | |
1127 | if (time_after(jiffies, expire)) | |
1128 | goto failed_removal; | |
1129 | ret = -EINTR; | |
1130 | if (signal_pending(current)) | |
1131 | goto failed_removal; | |
1132 | ret = 0; | |
1133 | if (drain) { | |
1134 | lru_add_drain_all(); | |
0c0e6195 | 1135 | cond_resched(); |
9f8f2172 | 1136 | drain_all_pages(); |
0c0e6195 KH |
1137 | } |
1138 | ||
1139 | pfn = scan_lru_pages(start_pfn, end_pfn); | |
1140 | if (pfn) { /* We have page on LRU */ | |
1141 | ret = do_migrate_range(pfn, end_pfn); | |
1142 | if (!ret) { | |
1143 | drain = 1; | |
1144 | goto repeat; | |
1145 | } else { | |
1146 | if (ret < 0) | |
1147 | if (--retry_max == 0) | |
1148 | goto failed_removal; | |
1149 | yield(); | |
1150 | drain = 1; | |
1151 | goto repeat; | |
1152 | } | |
1153 | } | |
1154 | /* drain all zone's lru pagevec, this is asyncronous... */ | |
1155 | lru_add_drain_all(); | |
0c0e6195 KH |
1156 | yield(); |
1157 | /* drain pcp pages , this is synchrouns. */ | |
9f8f2172 | 1158 | drain_all_pages(); |
0c0e6195 KH |
1159 | /* check again */ |
1160 | offlined_pages = check_pages_isolated(start_pfn, end_pfn); | |
1161 | if (offlined_pages < 0) { | |
1162 | ret = -EBUSY; | |
1163 | goto failed_removal; | |
1164 | } | |
1165 | printk(KERN_INFO "Offlined Pages %ld\n", offlined_pages); | |
1166 | /* Ok, all of our target is islaoted. | |
1167 | We cannot do rollback at this point. */ | |
1168 | offline_isolated_pages(start_pfn, end_pfn); | |
dbc0e4ce | 1169 | /* reset pagetype flags and makes migrate type to be MOVABLE */ |
0815f3d8 | 1170 | undo_isolate_page_range(start_pfn, end_pfn, MIGRATE_MOVABLE); |
0c0e6195 | 1171 | /* removal success */ |
0c0e6195 KH |
1172 | zone->present_pages -= offlined_pages; |
1173 | zone->zone_pgdat->node_present_pages -= offlined_pages; | |
1174 | totalram_pages -= offlined_pages; | |
7b78d335 | 1175 | |
1b79acc9 KM |
1176 | init_per_zone_wmark_min(); |
1177 | ||
1e8537ba | 1178 | if (!populated_zone(zone)) { |
340175b7 | 1179 | zone_pcp_reset(zone); |
1e8537ba XQ |
1180 | mutex_lock(&zonelists_mutex); |
1181 | build_all_zonelists(NULL, NULL); | |
1182 | mutex_unlock(&zonelists_mutex); | |
1183 | } else | |
1184 | zone_pcp_update(zone); | |
340175b7 | 1185 | |
d9713679 LJ |
1186 | node_states_clear_node(node, &arg); |
1187 | if (arg.status_change_nid >= 0) | |
8fe23e05 | 1188 | kswapd_stop(node); |
bce7394a | 1189 | |
0c0e6195 KH |
1190 | vm_total_pages = nr_free_pagecache_pages(); |
1191 | writeback_set_ratelimit(); | |
7b78d335 YG |
1192 | |
1193 | memory_notify(MEM_OFFLINE, &arg); | |
20d6c96b | 1194 | unlock_memory_hotplug(); |
0c0e6195 KH |
1195 | return 0; |
1196 | ||
1197 | failed_removal: | |
a62e2f4f BH |
1198 | printk(KERN_INFO "memory offlining [mem %#010llx-%#010llx] failed\n", |
1199 | (unsigned long long) start_pfn << PAGE_SHIFT, | |
1200 | ((unsigned long long) end_pfn << PAGE_SHIFT) - 1); | |
7b78d335 | 1201 | memory_notify(MEM_CANCEL_OFFLINE, &arg); |
0c0e6195 | 1202 | /* pushback to free area */ |
0815f3d8 | 1203 | undo_isolate_page_range(start_pfn, end_pfn, MIGRATE_MOVABLE); |
7b78d335 | 1204 | |
6ad696d2 | 1205 | out: |
20d6c96b | 1206 | unlock_memory_hotplug(); |
0c0e6195 KH |
1207 | return ret; |
1208 | } | |
71088785 | 1209 | |
a16cee10 WC |
1210 | int offline_pages(unsigned long start_pfn, unsigned long nr_pages) |
1211 | { | |
1212 | return __offline_pages(start_pfn, start_pfn + nr_pages, 120 * HZ); | |
1213 | } | |
1214 | ||
71088785 BP |
1215 | int remove_memory(u64 start, u64 size) |
1216 | { | |
e90bdb7f WC |
1217 | struct memory_block *mem = NULL; |
1218 | struct mem_section *section; | |
71088785 | 1219 | unsigned long start_pfn, end_pfn; |
e90bdb7f WC |
1220 | unsigned long pfn, section_nr; |
1221 | int ret; | |
71088785 BP |
1222 | |
1223 | start_pfn = PFN_DOWN(start); | |
1224 | end_pfn = start_pfn + PFN_DOWN(size); | |
e90bdb7f WC |
1225 | |
1226 | for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) { | |
1227 | section_nr = pfn_to_section_nr(pfn); | |
1228 | if (!present_section_nr(section_nr)) | |
1229 | continue; | |
1230 | ||
1231 | section = __nr_to_section(section_nr); | |
1232 | /* same memblock? */ | |
1233 | if (mem) | |
1234 | if ((section_nr >= mem->start_section_nr) && | |
1235 | (section_nr <= mem->end_section_nr)) | |
1236 | continue; | |
1237 | ||
1238 | mem = find_memory_block_hinted(section, mem); | |
1239 | if (!mem) | |
1240 | continue; | |
1241 | ||
1242 | ret = offline_memory_block(mem); | |
1243 | if (ret) { | |
1244 | kobject_put(&mem->dev.kobj); | |
1245 | return ret; | |
1246 | } | |
1247 | } | |
1248 | ||
1249 | if (mem) | |
1250 | kobject_put(&mem->dev.kobj); | |
1251 | ||
1252 | return 0; | |
71088785 | 1253 | } |
48e94196 | 1254 | #else |
a16cee10 WC |
1255 | int offline_pages(unsigned long start_pfn, unsigned long nr_pages) |
1256 | { | |
1257 | return -EINVAL; | |
1258 | } | |
48e94196 KH |
1259 | int remove_memory(u64 start, u64 size) |
1260 | { | |
1261 | return -EINVAL; | |
1262 | } | |
0c0e6195 | 1263 | #endif /* CONFIG_MEMORY_HOTREMOVE */ |
71088785 | 1264 | EXPORT_SYMBOL_GPL(remove_memory); |