1
2
3
4
5
6
7
8#include <linux/stddef.h>
9#include <linux/mm.h>
10#include <linux/sched/signal.h>
11#include <linux/swap.h>
12#include <linux/interrupt.h>
13#include <linux/pagemap.h>
14#include <linux/compiler.h>
15#include <linux/export.h>
16#include <linux/pagevec.h>
17#include <linux/writeback.h>
18#include <linux/slab.h>
19#include <linux/sysctl.h>
20#include <linux/cpu.h>
21#include <linux/memory.h>
22#include <linux/memremap.h>
23#include <linux/memory_hotplug.h>
24#include <linux/highmem.h>
25#include <linux/vmalloc.h>
26#include <linux/ioport.h>
27#include <linux/delay.h>
28#include <linux/migrate.h>
29#include <linux/page-isolation.h>
30#include <linux/pfn.h>
31#include <linux/suspend.h>
32#include <linux/mm_inline.h>
33#include <linux/firmware-map.h>
34#include <linux/stop_machine.h>
35#include <linux/hugetlb.h>
36#include <linux/memblock.h>
37#include <linux/compaction.h>
38#include <linux/rmap.h>
39
40#include <asm/tlbflush.h>
41
42#include "internal.h"
43#include "shuffle.h"
44
45
46
47
48
49
50
51
52static void generic_online_page(struct page *page, unsigned int order);
53
54static online_page_callback_t online_page_callback = generic_online_page;
55static DEFINE_MUTEX(online_page_callback_lock);
56
57DEFINE_STATIC_PERCPU_RWSEM(mem_hotplug_lock);
58
59void get_online_mems(void)
60{
61 percpu_down_read(&mem_hotplug_lock);
62}
63
64void put_online_mems(void)
65{
66 percpu_up_read(&mem_hotplug_lock);
67}
68
69bool movable_node_enabled = false;
70
71#ifndef CONFIG_MEMORY_HOTPLUG_DEFAULT_ONLINE
72bool memhp_auto_online;
73#else
74bool memhp_auto_online = true;
75#endif
76EXPORT_SYMBOL_GPL(memhp_auto_online);
77
78static int __init setup_memhp_default_state(char *str)
79{
80 if (!strcmp(str, "online"))
81 memhp_auto_online = true;
82 else if (!strcmp(str, "offline"))
83 memhp_auto_online = false;
84
85 return 1;
86}
87__setup("memhp_default_state=", setup_memhp_default_state);
88
89void mem_hotplug_begin(void)
90{
91 cpus_read_lock();
92 percpu_down_write(&mem_hotplug_lock);
93}
94
95void mem_hotplug_done(void)
96{
97 percpu_up_write(&mem_hotplug_lock);
98 cpus_read_unlock();
99}
100
101u64 max_mem_size = U64_MAX;
102
103
104static struct resource *register_memory_resource(u64 start, u64 size)
105{
106 struct resource *res;
107 unsigned long flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
108 char *resource_name = "System RAM";
109
110 if (start + size > max_mem_size)
111 return ERR_PTR(-E2BIG);
112
113
114
115
116
117
118 res = __request_region(&iomem_resource, start, size,
119 resource_name, flags);
120
121 if (!res) {
122 pr_debug("Unable to reserve System RAM region: %016llx->%016llx\n",
123 start, start + size);
124 return ERR_PTR(-EEXIST);
125 }
126 return res;
127}
128
129static void release_memory_resource(struct resource *res)
130{
131 if (!res)
132 return;
133 release_resource(res);
134 kfree(res);
135}
136
137#ifdef CONFIG_MEMORY_HOTPLUG_SPARSE
138void get_page_bootmem(unsigned long info, struct page *page,
139 unsigned long type)
140{
141 page->freelist = (void *)type;
142 SetPagePrivate(page);
143 set_page_private(page, info);
144 page_ref_inc(page);
145}
146
147void put_page_bootmem(struct page *page)
148{
149 unsigned long type;
150
151 type = (unsigned long) page->freelist;
152 BUG_ON(type < MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE ||
153 type > MEMORY_HOTPLUG_MAX_BOOTMEM_TYPE);
154
155 if (page_ref_dec_return(page) == 1) {
156 page->freelist = NULL;
157 ClearPagePrivate(page);
158 set_page_private(page, 0);
159 INIT_LIST_HEAD(&page->lru);
160 free_reserved_page(page);
161 }
162}
163
164#ifdef CONFIG_HAVE_BOOTMEM_INFO_NODE
165#ifndef CONFIG_SPARSEMEM_VMEMMAP
166static void register_page_bootmem_info_section(unsigned long start_pfn)
167{
168 unsigned long mapsize, section_nr, i;
169 struct mem_section *ms;
170 struct page *page, *memmap;
171 struct mem_section_usage *usage;
172
173 section_nr = pfn_to_section_nr(start_pfn);
174 ms = __nr_to_section(section_nr);
175
176
177 memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr);
178
179
180
181
182
183 page = virt_to_page(memmap);
184 mapsize = sizeof(struct page) * PAGES_PER_SECTION;
185 mapsize = PAGE_ALIGN(mapsize) >> PAGE_SHIFT;
186
187
188 for (i = 0; i < mapsize; i++, page++)
189 get_page_bootmem(section_nr, page, SECTION_INFO);
190
191 usage = ms->usage;
192 page = virt_to_page(usage);
193
194 mapsize = PAGE_ALIGN(mem_section_usage_size()) >> PAGE_SHIFT;
195
196 for (i = 0; i < mapsize; i++, page++)
197 get_page_bootmem(section_nr, page, MIX_SECTION_INFO);
198
199}
200#else
201static void register_page_bootmem_info_section(unsigned long start_pfn)
202{
203 unsigned long mapsize, section_nr, i;
204 struct mem_section *ms;
205 struct page *page, *memmap;
206 struct mem_section_usage *usage;
207
208 section_nr = pfn_to_section_nr(start_pfn);
209 ms = __nr_to_section(section_nr);
210
211 memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr);
212
213 register_page_bootmem_memmap(section_nr, memmap, PAGES_PER_SECTION);
214
215 usage = ms->usage;
216 page = virt_to_page(usage);
217
218 mapsize = PAGE_ALIGN(mem_section_usage_size()) >> PAGE_SHIFT;
219
220 for (i = 0; i < mapsize; i++, page++)
221 get_page_bootmem(section_nr, page, MIX_SECTION_INFO);
222}
223#endif
224
225void __init register_page_bootmem_info_node(struct pglist_data *pgdat)
226{
227 unsigned long i, pfn, end_pfn, nr_pages;
228 int node = pgdat->node_id;
229 struct page *page;
230
231 nr_pages = PAGE_ALIGN(sizeof(struct pglist_data)) >> PAGE_SHIFT;
232 page = virt_to_page(pgdat);
233
234 for (i = 0; i < nr_pages; i++, page++)
235 get_page_bootmem(node, page, NODE_INFO);
236
237 pfn = pgdat->node_start_pfn;
238 end_pfn = pgdat_end_pfn(pgdat);
239
240
241 for (; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
242
243
244
245
246
247
248 if (pfn_valid(pfn) && (early_pfn_to_nid(pfn) == node))
249 register_page_bootmem_info_section(pfn);
250 }
251}
252#endif
253
254static int check_pfn_span(unsigned long pfn, unsigned long nr_pages,
255 const char *reason)
256{
257
258
259
260
261
262
263
264
265
266 unsigned long min_align;
267
268 if (IS_ENABLED(CONFIG_SPARSEMEM_VMEMMAP))
269 min_align = PAGES_PER_SUBSECTION;
270 else
271 min_align = PAGES_PER_SECTION;
272 if (!IS_ALIGNED(pfn, min_align)
273 || !IS_ALIGNED(nr_pages, min_align)) {
274 WARN(1, "Misaligned __%s_pages start: %#lx end: #%lx\n",
275 reason, pfn, pfn + nr_pages - 1);
276 return -EINVAL;
277 }
278 return 0;
279}
280
281
282
283
284
285
286
287int __ref __add_pages(int nid, unsigned long pfn, unsigned long nr_pages,
288 struct mhp_restrictions *restrictions)
289{
290 int err;
291 unsigned long nr, start_sec, end_sec;
292 struct vmem_altmap *altmap = restrictions->altmap;
293
294 if (altmap) {
295
296
297
298 if (altmap->base_pfn != pfn
299 || vmem_altmap_offset(altmap) > nr_pages) {
300 pr_warn_once("memory add fail, invalid altmap\n");
301 return -EINVAL;
302 }
303 altmap->alloc = 0;
304 }
305
306 err = check_pfn_span(pfn, nr_pages, "add");
307 if (err)
308 return err;
309
310 start_sec = pfn_to_section_nr(pfn);
311 end_sec = pfn_to_section_nr(pfn + nr_pages - 1);
312 for (nr = start_sec; nr <= end_sec; nr++) {
313 unsigned long pfns;
314
315 pfns = min(nr_pages, PAGES_PER_SECTION
316 - (pfn & ~PAGE_SECTION_MASK));
317 err = sparse_add_section(nid, pfn, pfns, altmap);
318 if (err)
319 break;
320 pfn += pfns;
321 nr_pages -= pfns;
322 cond_resched();
323 }
324 vmemmap_populate_print_last();
325 return err;
326}
327
328
329static unsigned long find_smallest_section_pfn(int nid, struct zone *zone,
330 unsigned long start_pfn,
331 unsigned long end_pfn)
332{
333 for (; start_pfn < end_pfn; start_pfn += PAGES_PER_SUBSECTION) {
334 if (unlikely(!pfn_valid(start_pfn)))
335 continue;
336
337 if (unlikely(pfn_to_nid(start_pfn) != nid))
338 continue;
339
340 if (zone && zone != page_zone(pfn_to_page(start_pfn)))
341 continue;
342
343 return start_pfn;
344 }
345
346 return 0;
347}
348
349
350static unsigned long find_biggest_section_pfn(int nid, struct zone *zone,
351 unsigned long start_pfn,
352 unsigned long end_pfn)
353{
354 unsigned long pfn;
355
356
357 pfn = end_pfn - 1;
358 for (; pfn >= start_pfn; pfn -= PAGES_PER_SUBSECTION) {
359 if (unlikely(!pfn_valid(pfn)))
360 continue;
361
362 if (unlikely(pfn_to_nid(pfn) != nid))
363 continue;
364
365 if (zone && zone != page_zone(pfn_to_page(pfn)))
366 continue;
367
368 return pfn;
369 }
370
371 return 0;
372}
373
374static void shrink_zone_span(struct zone *zone, unsigned long start_pfn,
375 unsigned long end_pfn)
376{
377 unsigned long zone_start_pfn = zone->zone_start_pfn;
378 unsigned long z = zone_end_pfn(zone);
379 unsigned long zone_end_pfn = z;
380 unsigned long pfn;
381 int nid = zone_to_nid(zone);
382
383 zone_span_writelock(zone);
384 if (zone_start_pfn == start_pfn) {
385
386
387
388
389
390
391 pfn = find_smallest_section_pfn(nid, zone, end_pfn,
392 zone_end_pfn);
393 if (pfn) {
394 zone->zone_start_pfn = pfn;
395 zone->spanned_pages = zone_end_pfn - pfn;
396 }
397 } else if (zone_end_pfn == end_pfn) {
398
399
400
401
402
403
404 pfn = find_biggest_section_pfn(nid, zone, zone_start_pfn,
405 start_pfn);
406 if (pfn)
407 zone->spanned_pages = pfn - zone_start_pfn + 1;
408 }
409
410
411
412
413
414
415
416 pfn = zone_start_pfn;
417 for (; pfn < zone_end_pfn; pfn += PAGES_PER_SUBSECTION) {
418 if (unlikely(!pfn_valid(pfn)))
419 continue;
420
421 if (page_zone(pfn_to_page(pfn)) != zone)
422 continue;
423
424
425 if (pfn >= start_pfn && pfn < end_pfn)
426 continue;
427
428
429 zone_span_writeunlock(zone);
430 return;
431 }
432
433
434 zone->zone_start_pfn = 0;
435 zone->spanned_pages = 0;
436 zone_span_writeunlock(zone);
437}
438
439static void shrink_pgdat_span(struct pglist_data *pgdat,
440 unsigned long start_pfn, unsigned long end_pfn)
441{
442 unsigned long pgdat_start_pfn = pgdat->node_start_pfn;
443 unsigned long p = pgdat_end_pfn(pgdat);
444 unsigned long pgdat_end_pfn = p;
445 unsigned long pfn;
446 int nid = pgdat->node_id;
447
448 if (pgdat_start_pfn == start_pfn) {
449
450
451
452
453
454
455 pfn = find_smallest_section_pfn(nid, NULL, end_pfn,
456 pgdat_end_pfn);
457 if (pfn) {
458 pgdat->node_start_pfn = pfn;
459 pgdat->node_spanned_pages = pgdat_end_pfn - pfn;
460 }
461 } else if (pgdat_end_pfn == end_pfn) {
462
463
464
465
466
467
468 pfn = find_biggest_section_pfn(nid, NULL, pgdat_start_pfn,
469 start_pfn);
470 if (pfn)
471 pgdat->node_spanned_pages = pfn - pgdat_start_pfn + 1;
472 }
473
474
475
476
477
478
479
480
481 pfn = pgdat_start_pfn;
482 for (; pfn < pgdat_end_pfn; pfn += PAGES_PER_SUBSECTION) {
483 if (unlikely(!pfn_valid(pfn)))
484 continue;
485
486 if (pfn_to_nid(pfn) != nid)
487 continue;
488
489
490 if (pfn >= start_pfn && pfn < end_pfn)
491 continue;
492
493
494 return;
495 }
496
497
498 pgdat->node_start_pfn = 0;
499 pgdat->node_spanned_pages = 0;
500}
501
502static void __remove_zone(struct zone *zone, unsigned long start_pfn,
503 unsigned long nr_pages)
504{
505 struct pglist_data *pgdat = zone->zone_pgdat;
506 unsigned long flags;
507
508 pgdat_resize_lock(zone->zone_pgdat, &flags);
509 shrink_zone_span(zone, start_pfn, start_pfn + nr_pages);
510 shrink_pgdat_span(pgdat, start_pfn, start_pfn + nr_pages);
511 pgdat_resize_unlock(zone->zone_pgdat, &flags);
512}
513
514static void __remove_section(struct zone *zone, unsigned long pfn,
515 unsigned long nr_pages, unsigned long map_offset,
516 struct vmem_altmap *altmap)
517{
518 struct mem_section *ms = __nr_to_section(pfn_to_section_nr(pfn));
519
520 if (WARN_ON_ONCE(!valid_section(ms)))
521 return;
522
523 __remove_zone(zone, pfn, nr_pages);
524 sparse_remove_section(ms, pfn, nr_pages, map_offset, altmap);
525}
526
527
528
529
530
531
532
533
534
535
536
537
538
539void __remove_pages(struct zone *zone, unsigned long pfn,
540 unsigned long nr_pages, struct vmem_altmap *altmap)
541{
542 unsigned long map_offset = 0;
543 unsigned long nr, start_sec, end_sec;
544
545 map_offset = vmem_altmap_offset(altmap);
546
547 clear_zone_contiguous(zone);
548
549 if (check_pfn_span(pfn, nr_pages, "remove"))
550 return;
551
552 start_sec = pfn_to_section_nr(pfn);
553 end_sec = pfn_to_section_nr(pfn + nr_pages - 1);
554 for (nr = start_sec; nr <= end_sec; nr++) {
555 unsigned long pfns;
556
557 cond_resched();
558 pfns = min(nr_pages, PAGES_PER_SECTION
559 - (pfn & ~PAGE_SECTION_MASK));
560 __remove_section(zone, pfn, pfns, map_offset, altmap);
561 pfn += pfns;
562 nr_pages -= pfns;
563 map_offset = 0;
564 }
565
566 set_zone_contiguous(zone);
567}
568
569int set_online_page_callback(online_page_callback_t callback)
570{
571 int rc = -EINVAL;
572
573 get_online_mems();
574 mutex_lock(&online_page_callback_lock);
575
576 if (online_page_callback == generic_online_page) {
577 online_page_callback = callback;
578 rc = 0;
579 }
580
581 mutex_unlock(&online_page_callback_lock);
582 put_online_mems();
583
584 return rc;
585}
586EXPORT_SYMBOL_GPL(set_online_page_callback);
587
588int restore_online_page_callback(online_page_callback_t callback)
589{
590 int rc = -EINVAL;
591
592 get_online_mems();
593 mutex_lock(&online_page_callback_lock);
594
595 if (online_page_callback == callback) {
596 online_page_callback = generic_online_page;
597 rc = 0;
598 }
599
600 mutex_unlock(&online_page_callback_lock);
601 put_online_mems();
602
603 return rc;
604}
605EXPORT_SYMBOL_GPL(restore_online_page_callback);
606
607void __online_page_set_limits(struct page *page)
608{
609}
610EXPORT_SYMBOL_GPL(__online_page_set_limits);
611
612void __online_page_increment_counters(struct page *page)
613{
614 adjust_managed_page_count(page, 1);
615}
616EXPORT_SYMBOL_GPL(__online_page_increment_counters);
617
618void __online_page_free(struct page *page)
619{
620 __free_reserved_page(page);
621}
622EXPORT_SYMBOL_GPL(__online_page_free);
623
624static void generic_online_page(struct page *page, unsigned int order)
625{
626 kernel_map_pages(page, 1 << order, 1);
627 __free_pages_core(page, order);
628 totalram_pages_add(1UL << order);
629#ifdef CONFIG_HIGHMEM
630 if (PageHighMem(page))
631 totalhigh_pages_add(1UL << order);
632#endif
633}
634
635static int online_pages_blocks(unsigned long start, unsigned long nr_pages)
636{
637 unsigned long end = start + nr_pages;
638 int order, onlined_pages = 0;
639
640 while (start < end) {
641 order = min(MAX_ORDER - 1,
642 get_order(PFN_PHYS(end) - PFN_PHYS(start)));
643 (*online_page_callback)(pfn_to_page(start), order);
644
645 onlined_pages += (1UL << order);
646 start += (1UL << order);
647 }
648 return onlined_pages;
649}
650
651static int online_pages_range(unsigned long start_pfn, unsigned long nr_pages,
652 void *arg)
653{
654 unsigned long onlined_pages = *(unsigned long *)arg;
655
656 if (PageReserved(pfn_to_page(start_pfn)))
657 onlined_pages += online_pages_blocks(start_pfn, nr_pages);
658
659 online_mem_sections(start_pfn, start_pfn + nr_pages);
660
661 *(unsigned long *)arg = onlined_pages;
662 return 0;
663}
664
665
666static void node_states_check_changes_online(unsigned long nr_pages,
667 struct zone *zone, struct memory_notify *arg)
668{
669 int nid = zone_to_nid(zone);
670
671 arg->status_change_nid = NUMA_NO_NODE;
672 arg->status_change_nid_normal = NUMA_NO_NODE;
673 arg->status_change_nid_high = NUMA_NO_NODE;
674
675 if (!node_state(nid, N_MEMORY))
676 arg->status_change_nid = nid;
677 if (zone_idx(zone) <= ZONE_NORMAL && !node_state(nid, N_NORMAL_MEMORY))
678 arg->status_change_nid_normal = nid;
679#ifdef CONFIG_HIGHMEM
680 if (zone_idx(zone) <= ZONE_HIGHMEM && !node_state(nid, N_HIGH_MEMORY))
681 arg->status_change_nid_high = nid;
682#endif
683}
684
685static void node_states_set_node(int node, struct memory_notify *arg)
686{
687 if (arg->status_change_nid_normal >= 0)
688 node_set_state(node, N_NORMAL_MEMORY);
689
690 if (arg->status_change_nid_high >= 0)
691 node_set_state(node, N_HIGH_MEMORY);
692
693 if (arg->status_change_nid >= 0)
694 node_set_state(node, N_MEMORY);
695}
696
697static void __meminit resize_zone_range(struct zone *zone, unsigned long start_pfn,
698 unsigned long nr_pages)
699{
700 unsigned long old_end_pfn = zone_end_pfn(zone);
701
702 if (zone_is_empty(zone) || start_pfn < zone->zone_start_pfn)
703 zone->zone_start_pfn = start_pfn;
704
705 zone->spanned_pages = max(start_pfn + nr_pages, old_end_pfn) - zone->zone_start_pfn;
706}
707
708static void __meminit resize_pgdat_range(struct pglist_data *pgdat, unsigned long start_pfn,
709 unsigned long nr_pages)
710{
711 unsigned long old_end_pfn = pgdat_end_pfn(pgdat);
712
713 if (!pgdat->node_spanned_pages || start_pfn < pgdat->node_start_pfn)
714 pgdat->node_start_pfn = start_pfn;
715
716 pgdat->node_spanned_pages = max(start_pfn + nr_pages, old_end_pfn) - pgdat->node_start_pfn;
717}
718
719void __ref move_pfn_range_to_zone(struct zone *zone, unsigned long start_pfn,
720 unsigned long nr_pages, struct vmem_altmap *altmap)
721{
722 struct pglist_data *pgdat = zone->zone_pgdat;
723 int nid = pgdat->node_id;
724 unsigned long flags;
725
726 clear_zone_contiguous(zone);
727
728
729 pgdat_resize_lock(pgdat, &flags);
730 zone_span_writelock(zone);
731 if (zone_is_empty(zone))
732 init_currently_empty_zone(zone, start_pfn, nr_pages);
733 resize_zone_range(zone, start_pfn, nr_pages);
734 zone_span_writeunlock(zone);
735 resize_pgdat_range(pgdat, start_pfn, nr_pages);
736 pgdat_resize_unlock(pgdat, &flags);
737
738
739
740
741
742
743
744 memmap_init_zone(nr_pages, nid, zone_idx(zone), start_pfn,
745 MEMMAP_HOTPLUG, altmap);
746
747 set_zone_contiguous(zone);
748}
749
750
751
752
753
754
755static struct zone *default_kernel_zone_for_pfn(int nid, unsigned long start_pfn,
756 unsigned long nr_pages)
757{
758 struct pglist_data *pgdat = NODE_DATA(nid);
759 int zid;
760
761 for (zid = 0; zid <= ZONE_NORMAL; zid++) {
762 struct zone *zone = &pgdat->node_zones[zid];
763
764 if (zone_intersects(zone, start_pfn, nr_pages))
765 return zone;
766 }
767
768 return &pgdat->node_zones[ZONE_NORMAL];
769}
770
771static inline struct zone *default_zone_for_pfn(int nid, unsigned long start_pfn,
772 unsigned long nr_pages)
773{
774 struct zone *kernel_zone = default_kernel_zone_for_pfn(nid, start_pfn,
775 nr_pages);
776 struct zone *movable_zone = &NODE_DATA(nid)->node_zones[ZONE_MOVABLE];
777 bool in_kernel = zone_intersects(kernel_zone, start_pfn, nr_pages);
778 bool in_movable = zone_intersects(movable_zone, start_pfn, nr_pages);
779
780
781
782
783
784 if (in_kernel ^ in_movable)
785 return (in_kernel) ? kernel_zone : movable_zone;
786
787
788
789
790
791
792 return movable_node_enabled ? movable_zone : kernel_zone;
793}
794
795struct zone * zone_for_pfn_range(int online_type, int nid, unsigned start_pfn,
796 unsigned long nr_pages)
797{
798 if (online_type == MMOP_ONLINE_KERNEL)
799 return default_kernel_zone_for_pfn(nid, start_pfn, nr_pages);
800
801 if (online_type == MMOP_ONLINE_MOVABLE)
802 return &NODE_DATA(nid)->node_zones[ZONE_MOVABLE];
803
804 return default_zone_for_pfn(nid, start_pfn, nr_pages);
805}
806
807
808
809
810
811static struct zone * __meminit move_pfn_range(int online_type, int nid,
812 unsigned long start_pfn, unsigned long nr_pages)
813{
814 struct zone *zone;
815
816 zone = zone_for_pfn_range(online_type, nid, start_pfn, nr_pages);
817 move_pfn_range_to_zone(zone, start_pfn, nr_pages, NULL);
818 return zone;
819}
820
821int __ref online_pages(unsigned long pfn, unsigned long nr_pages, int online_type)
822{
823 unsigned long flags;
824 unsigned long onlined_pages = 0;
825 struct zone *zone;
826 int need_zonelists_rebuild = 0;
827 int nid;
828 int ret;
829 struct memory_notify arg;
830 struct memory_block *mem;
831
832 mem_hotplug_begin();
833
834
835
836
837
838 mem = find_memory_block(__pfn_to_section(pfn));
839 nid = mem->nid;
840 put_device(&mem->dev);
841
842
843 zone = move_pfn_range(online_type, nid, pfn, nr_pages);
844
845 arg.start_pfn = pfn;
846 arg.nr_pages = nr_pages;
847 node_states_check_changes_online(nr_pages, zone, &arg);
848
849 ret = memory_notify(MEM_GOING_ONLINE, &arg);
850 ret = notifier_to_errno(ret);
851 if (ret)
852 goto failed_addition;
853
854
855
856
857
858
859 if (!populated_zone(zone)) {
860 need_zonelists_rebuild = 1;
861 setup_zone_pageset(zone);
862 }
863
864 ret = walk_system_ram_range(pfn, nr_pages, &onlined_pages,
865 online_pages_range);
866 if (ret) {
867 if (need_zonelists_rebuild)
868 zone_pcp_reset(zone);
869 goto failed_addition;
870 }
871
872 zone->present_pages += onlined_pages;
873
874 pgdat_resize_lock(zone->zone_pgdat, &flags);
875 zone->zone_pgdat->node_present_pages += onlined_pages;
876 pgdat_resize_unlock(zone->zone_pgdat, &flags);
877
878 shuffle_zone(zone);
879
880 if (onlined_pages) {
881 node_states_set_node(nid, &arg);
882 if (need_zonelists_rebuild)
883 build_all_zonelists(NULL);
884 else
885 zone_pcp_update(zone);
886 }
887
888 init_per_zone_wmark_min();
889
890 if (onlined_pages) {
891 kswapd_run(nid);
892 kcompactd_run(nid);
893 }
894
895 vm_total_pages = nr_free_pagecache_pages();
896
897 writeback_set_ratelimit();
898
899 if (onlined_pages)
900 memory_notify(MEM_ONLINE, &arg);
901 mem_hotplug_done();
902 return 0;
903
904failed_addition:
905 pr_debug("online_pages [mem %#010llx-%#010llx] failed\n",
906 (unsigned long long) pfn << PAGE_SHIFT,
907 (((unsigned long long) pfn + nr_pages) << PAGE_SHIFT) - 1);
908 memory_notify(MEM_CANCEL_ONLINE, &arg);
909 mem_hotplug_done();
910 return ret;
911}
912#endif
913
914static void reset_node_present_pages(pg_data_t *pgdat)
915{
916 struct zone *z;
917
918 for (z = pgdat->node_zones; z < pgdat->node_zones + MAX_NR_ZONES; z++)
919 z->present_pages = 0;
920
921 pgdat->node_present_pages = 0;
922}
923
924
925static pg_data_t __ref *hotadd_new_pgdat(int nid, u64 start)
926{
927 struct pglist_data *pgdat;
928 unsigned long start_pfn = PFN_DOWN(start);
929
930 pgdat = NODE_DATA(nid);
931 if (!pgdat) {
932 pgdat = arch_alloc_nodedata(nid);
933 if (!pgdat)
934 return NULL;
935
936 arch_refresh_nodedata(nid, pgdat);
937 } else {
938
939
940
941
942
943 pgdat->nr_zones = 0;
944 pgdat->kswapd_order = 0;
945 pgdat->kswapd_classzone_idx = 0;
946 }
947
948
949
950 pgdat->node_id = nid;
951 pgdat->node_start_pfn = start_pfn;
952
953
954 free_area_init_core_hotplug(nid);
955 pgdat->per_cpu_nodestats = alloc_percpu(struct per_cpu_nodestat);
956
957
958
959
960
961 build_all_zonelists(pgdat);
962
963
964
965
966
967
968 reset_node_managed_pages(pgdat);
969 reset_node_present_pages(pgdat);
970
971 return pgdat;
972}
973
974static void rollback_node_hotadd(int nid)
975{
976 pg_data_t *pgdat = NODE_DATA(nid);
977
978 arch_refresh_nodedata(nid, NULL);
979 free_percpu(pgdat->per_cpu_nodestats);
980 arch_free_nodedata(pgdat);
981}
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996static int __try_online_node(int nid, u64 start, bool set_node_online)
997{
998 pg_data_t *pgdat;
999 int ret = 1;
1000
1001 if (node_online(nid))
1002 return 0;
1003
1004 pgdat = hotadd_new_pgdat(nid, start);
1005 if (!pgdat) {
1006 pr_err("Cannot online node %d due to NULL pgdat\n", nid);
1007 ret = -ENOMEM;
1008 goto out;
1009 }
1010
1011 if (set_node_online) {
1012 node_set_online(nid);
1013 ret = register_one_node(nid);
1014 BUG_ON(ret);
1015 }
1016out:
1017 return ret;
1018}
1019
1020
1021
1022
1023int try_online_node(int nid)
1024{
1025 int ret;
1026
1027 mem_hotplug_begin();
1028 ret = __try_online_node(nid, 0, true);
1029 mem_hotplug_done();
1030 return ret;
1031}
1032
1033static int check_hotplug_memory_range(u64 start, u64 size)
1034{
1035
1036 if (!size || !IS_ALIGNED(start, memory_block_size_bytes()) ||
1037 !IS_ALIGNED(size, memory_block_size_bytes())) {
1038 pr_err("Block size [%#lx] unaligned hotplug range: start %#llx, size %#llx",
1039 memory_block_size_bytes(), start, size);
1040 return -EINVAL;
1041 }
1042
1043 return 0;
1044}
1045
1046static int online_memory_block(struct memory_block *mem, void *arg)
1047{
1048 return device_online(&mem->dev);
1049}
1050
1051
1052
1053
1054
1055
1056
1057int __ref add_memory_resource(int nid, struct resource *res)
1058{
1059 struct mhp_restrictions restrictions = {};
1060 u64 start, size;
1061 bool new_node = false;
1062 int ret;
1063
1064 start = res->start;
1065 size = resource_size(res);
1066
1067 ret = check_hotplug_memory_range(start, size);
1068 if (ret)
1069 return ret;
1070
1071 mem_hotplug_begin();
1072
1073
1074
1075
1076
1077
1078
1079 memblock_add_node(start, size, nid);
1080
1081 ret = __try_online_node(nid, start, false);
1082 if (ret < 0)
1083 goto error;
1084 new_node = ret;
1085
1086
1087 ret = arch_add_memory(nid, start, size, &restrictions);
1088 if (ret < 0)
1089 goto error;
1090
1091
1092 ret = create_memory_block_devices(start, size);
1093 if (ret) {
1094 arch_remove_memory(nid, start, size, NULL);
1095 goto error;
1096 }
1097
1098 if (new_node) {
1099
1100
1101
1102
1103
1104 node_set_online(nid);
1105 ret = __register_one_node(nid);
1106 BUG_ON(ret);
1107 }
1108
1109
1110 ret = link_mem_sections(nid, PFN_DOWN(start), PFN_UP(start + size - 1));
1111 BUG_ON(ret);
1112
1113
1114 firmware_map_add_hotplug(start, start + size, "System RAM");
1115
1116
1117 mem_hotplug_done();
1118
1119
1120 if (memhp_auto_online)
1121 walk_memory_blocks(start, size, NULL, online_memory_block);
1122
1123 return ret;
1124error:
1125
1126 if (new_node)
1127 rollback_node_hotadd(nid);
1128 memblock_remove(start, size);
1129 mem_hotplug_done();
1130 return ret;
1131}
1132
1133
1134int __ref __add_memory(int nid, u64 start, u64 size)
1135{
1136 struct resource *res;
1137 int ret;
1138
1139 res = register_memory_resource(start, size);
1140 if (IS_ERR(res))
1141 return PTR_ERR(res);
1142
1143 ret = add_memory_resource(nid, res);
1144 if (ret < 0)
1145 release_memory_resource(res);
1146 return ret;
1147}
1148
1149int add_memory(int nid, u64 start, u64 size)
1150{
1151 int rc;
1152
1153 lock_device_hotplug();
1154 rc = __add_memory(nid, start, size);
1155 unlock_device_hotplug();
1156
1157 return rc;
1158}
1159EXPORT_SYMBOL_GPL(add_memory);
1160
1161#ifdef CONFIG_MEMORY_HOTREMOVE
1162
1163
1164
1165
1166
1167
1168
1169static inline int pageblock_free(struct page *page)
1170{
1171 return PageBuddy(page) && page_order(page) >= pageblock_order;
1172}
1173
1174
1175static unsigned long next_active_pageblock(unsigned long pfn)
1176{
1177 struct page *page = pfn_to_page(pfn);
1178
1179
1180 BUG_ON(pfn & (pageblock_nr_pages - 1));
1181
1182
1183 if (pageblock_free(page)) {
1184 int order;
1185
1186 order = page_order(page);
1187 if ((order < MAX_ORDER) && (order >= pageblock_order))
1188 return pfn + (1 << order);
1189 }
1190
1191 return pfn + pageblock_nr_pages;
1192}
1193
1194static bool is_pageblock_removable_nolock(unsigned long pfn)
1195{
1196 struct page *page = pfn_to_page(pfn);
1197 struct zone *zone;
1198
1199
1200
1201
1202
1203
1204
1205
1206 if (!node_online(page_to_nid(page)))
1207 return false;
1208
1209 zone = page_zone(page);
1210 pfn = page_to_pfn(page);
1211 if (!zone_spans_pfn(zone, pfn))
1212 return false;
1213
1214 return !has_unmovable_pages(zone, page, 0, MIGRATE_MOVABLE, SKIP_HWPOISON);
1215}
1216
1217
1218bool is_mem_section_removable(unsigned long start_pfn, unsigned long nr_pages)
1219{
1220 unsigned long end_pfn, pfn;
1221
1222 end_pfn = min(start_pfn + nr_pages,
1223 zone_end_pfn(page_zone(pfn_to_page(start_pfn))));
1224
1225
1226 for (pfn = start_pfn; pfn < end_pfn; pfn = next_active_pageblock(pfn)) {
1227 if (!is_pageblock_removable_nolock(pfn))
1228 return false;
1229 cond_resched();
1230 }
1231
1232
1233 return true;
1234}
1235
1236
1237
1238
1239
1240int test_pages_in_a_zone(unsigned long start_pfn, unsigned long end_pfn,
1241 unsigned long *valid_start, unsigned long *valid_end)
1242{
1243 unsigned long pfn, sec_end_pfn;
1244 unsigned long start, end;
1245 struct zone *zone = NULL;
1246 struct page *page;
1247 int i;
1248 for (pfn = start_pfn, sec_end_pfn = SECTION_ALIGN_UP(start_pfn + 1);
1249 pfn < end_pfn;
1250 pfn = sec_end_pfn, sec_end_pfn += PAGES_PER_SECTION) {
1251
1252 if (!present_section_nr(pfn_to_section_nr(pfn)))
1253 continue;
1254 for (; pfn < sec_end_pfn && pfn < end_pfn;
1255 pfn += MAX_ORDER_NR_PAGES) {
1256 i = 0;
1257
1258 while ((i < MAX_ORDER_NR_PAGES) &&
1259 !pfn_valid_within(pfn + i))
1260 i++;
1261 if (i == MAX_ORDER_NR_PAGES || pfn + i >= end_pfn)
1262 continue;
1263
1264 if (zone && !zone_spans_pfn(zone, pfn + i))
1265 return 0;
1266 page = pfn_to_page(pfn + i);
1267 if (zone && page_zone(page) != zone)
1268 return 0;
1269 if (!zone)
1270 start = pfn + i;
1271 zone = page_zone(page);
1272 end = pfn + MAX_ORDER_NR_PAGES;
1273 }
1274 }
1275
1276 if (zone) {
1277 *valid_start = start;
1278 *valid_end = min(end, end_pfn);
1279 return 1;
1280 } else {
1281 return 0;
1282 }
1283}
1284
1285
1286
1287
1288
1289
1290
1291static unsigned long scan_movable_pages(unsigned long start, unsigned long end)
1292{
1293 unsigned long pfn;
1294
1295 for (pfn = start; pfn < end; pfn++) {
1296 struct page *page, *head;
1297 unsigned long skip;
1298
1299 if (!pfn_valid(pfn))
1300 continue;
1301 page = pfn_to_page(pfn);
1302 if (PageLRU(page))
1303 return pfn;
1304 if (__PageMovable(page))
1305 return pfn;
1306
1307 if (!PageHuge(page))
1308 continue;
1309 head = compound_head(page);
1310 if (page_huge_active(head))
1311 return pfn;
1312 skip = (1 << compound_order(head)) - (page - head);
1313 pfn += skip - 1;
1314 }
1315 return 0;
1316}
1317
1318static struct page *new_node_page(struct page *page, unsigned long private)
1319{
1320 int nid = page_to_nid(page);
1321 nodemask_t nmask = node_states[N_MEMORY];
1322
1323
1324
1325
1326
1327
1328 node_clear(nid, nmask);
1329 if (nodes_empty(nmask))
1330 node_set(nid, nmask);
1331
1332 return new_page_nodemask(page, nid, &nmask);
1333}
1334
1335static int
1336do_migrate_range(unsigned long start_pfn, unsigned long end_pfn)
1337{
1338 unsigned long pfn;
1339 struct page *page;
1340 int ret = 0;
1341 LIST_HEAD(source);
1342
1343 for (pfn = start_pfn; pfn < end_pfn; pfn++) {
1344 if (!pfn_valid(pfn))
1345 continue;
1346 page = pfn_to_page(pfn);
1347
1348 if (PageHuge(page)) {
1349 struct page *head = compound_head(page);
1350 pfn = page_to_pfn(head) + (1<<compound_order(head)) - 1;
1351 isolate_huge_page(head, &source);
1352 continue;
1353 } else if (PageTransHuge(page))
1354 pfn = page_to_pfn(compound_head(page))
1355 + hpage_nr_pages(page) - 1;
1356
1357
1358
1359
1360
1361
1362
1363
1364 if (PageHWPoison(page)) {
1365 if (WARN_ON(PageLRU(page)))
1366 isolate_lru_page(page);
1367 if (page_mapped(page))
1368 try_to_unmap(page, TTU_IGNORE_MLOCK | TTU_IGNORE_ACCESS);
1369 continue;
1370 }
1371
1372 if (!get_page_unless_zero(page))
1373 continue;
1374
1375
1376
1377
1378 if (PageLRU(page))
1379 ret = isolate_lru_page(page);
1380 else
1381 ret = isolate_movable_page(page, ISOLATE_UNEVICTABLE);
1382 if (!ret) {
1383 list_add_tail(&page->lru, &source);
1384 if (!__PageMovable(page))
1385 inc_node_page_state(page, NR_ISOLATED_ANON +
1386 page_is_file_cache(page));
1387
1388 } else {
1389 pr_warn("failed to isolate pfn %lx\n", pfn);
1390 dump_page(page, "isolation failed");
1391 }
1392 put_page(page);
1393 }
1394 if (!list_empty(&source)) {
1395
1396 ret = migrate_pages(&source, new_node_page, NULL, 0,
1397 MIGRATE_SYNC, MR_MEMORY_HOTPLUG);
1398 if (ret) {
1399 list_for_each_entry(page, &source, lru) {
1400 pr_warn("migrating pfn %lx failed ret:%d ",
1401 page_to_pfn(page), ret);
1402 dump_page(page, "migration failure");
1403 }
1404 putback_movable_pages(&source);
1405 }
1406 }
1407
1408 return ret;
1409}
1410
1411
1412
1413
1414static int
1415offline_isolated_pages_cb(unsigned long start, unsigned long nr_pages,
1416 void *data)
1417{
1418 unsigned long *offlined_pages = (unsigned long *)data;
1419
1420 *offlined_pages += __offline_isolated_pages(start, start + nr_pages);
1421 return 0;
1422}
1423
1424
1425
1426
1427static int
1428check_pages_isolated_cb(unsigned long start_pfn, unsigned long nr_pages,
1429 void *data)
1430{
1431 return test_pages_isolated(start_pfn, start_pfn + nr_pages, true);
1432}
1433
1434static int __init cmdline_parse_movable_node(char *p)
1435{
1436#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1437 movable_node_enabled = true;
1438#else
1439 pr_warn("movable_node parameter depends on CONFIG_HAVE_MEMBLOCK_NODE_MAP to work properly\n");
1440#endif
1441 return 0;
1442}
1443early_param("movable_node", cmdline_parse_movable_node);
1444
1445
1446static void node_states_check_changes_offline(unsigned long nr_pages,
1447 struct zone *zone, struct memory_notify *arg)
1448{
1449 struct pglist_data *pgdat = zone->zone_pgdat;
1450 unsigned long present_pages = 0;
1451 enum zone_type zt;
1452
1453 arg->status_change_nid = NUMA_NO_NODE;
1454 arg->status_change_nid_normal = NUMA_NO_NODE;
1455 arg->status_change_nid_high = NUMA_NO_NODE;
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465 for (zt = 0; zt <= ZONE_NORMAL; zt++)
1466 present_pages += pgdat->node_zones[zt].present_pages;
1467 if (zone_idx(zone) <= ZONE_NORMAL && nr_pages >= present_pages)
1468 arg->status_change_nid_normal = zone_to_nid(zone);
1469
1470#ifdef CONFIG_HIGHMEM
1471
1472
1473
1474
1475
1476
1477
1478
1479 present_pages += pgdat->node_zones[ZONE_HIGHMEM].present_pages;
1480 if (zone_idx(zone) <= ZONE_HIGHMEM && nr_pages >= present_pages)
1481 arg->status_change_nid_high = zone_to_nid(zone);
1482#endif
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494 present_pages += pgdat->node_zones[ZONE_MOVABLE].present_pages;
1495
1496 if (nr_pages >= present_pages)
1497 arg->status_change_nid = zone_to_nid(zone);
1498}
1499
1500static void node_states_clear_node(int node, struct memory_notify *arg)
1501{
1502 if (arg->status_change_nid_normal >= 0)
1503 node_clear_state(node, N_NORMAL_MEMORY);
1504
1505 if (arg->status_change_nid_high >= 0)
1506 node_clear_state(node, N_HIGH_MEMORY);
1507
1508 if (arg->status_change_nid >= 0)
1509 node_clear_state(node, N_MEMORY);
1510}
1511
1512static int __ref __offline_pages(unsigned long start_pfn,
1513 unsigned long end_pfn)
1514{
1515 unsigned long pfn, nr_pages;
1516 unsigned long offlined_pages = 0;
1517 int ret, node, nr_isolate_pageblock;
1518 unsigned long flags;
1519 unsigned long valid_start, valid_end;
1520 struct zone *zone;
1521 struct memory_notify arg;
1522 char *reason;
1523
1524 mem_hotplug_begin();
1525
1526
1527
1528 if (!test_pages_in_a_zone(start_pfn, end_pfn, &valid_start,
1529 &valid_end)) {
1530 ret = -EINVAL;
1531 reason = "multizone range";
1532 goto failed_removal;
1533 }
1534
1535 zone = page_zone(pfn_to_page(valid_start));
1536 node = zone_to_nid(zone);
1537 nr_pages = end_pfn - start_pfn;
1538
1539
1540 ret = start_isolate_page_range(start_pfn, end_pfn,
1541 MIGRATE_MOVABLE,
1542 SKIP_HWPOISON | REPORT_FAILURE);
1543 if (ret < 0) {
1544 reason = "failure to isolate range";
1545 goto failed_removal;
1546 }
1547 nr_isolate_pageblock = ret;
1548
1549 arg.start_pfn = start_pfn;
1550 arg.nr_pages = nr_pages;
1551 node_states_check_changes_offline(nr_pages, zone, &arg);
1552
1553 ret = memory_notify(MEM_GOING_OFFLINE, &arg);
1554 ret = notifier_to_errno(ret);
1555 if (ret) {
1556 reason = "notifier failure";
1557 goto failed_removal_isolated;
1558 }
1559
1560 do {
1561 for (pfn = start_pfn; pfn;) {
1562 if (signal_pending(current)) {
1563 ret = -EINTR;
1564 reason = "signal backoff";
1565 goto failed_removal_isolated;
1566 }
1567
1568 cond_resched();
1569 lru_add_drain_all();
1570
1571 pfn = scan_movable_pages(pfn, end_pfn);
1572 if (pfn) {
1573
1574
1575
1576
1577 do_migrate_range(pfn, end_pfn);
1578 }
1579 }
1580
1581
1582
1583
1584
1585
1586 ret = dissolve_free_huge_pages(start_pfn, end_pfn);
1587 if (ret) {
1588 reason = "failure to dissolve huge pages";
1589 goto failed_removal_isolated;
1590 }
1591
1592 ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn,
1593 NULL, check_pages_isolated_cb);
1594 } while (ret);
1595
1596
1597
1598 walk_system_ram_range(start_pfn, end_pfn - start_pfn,
1599 &offlined_pages, offline_isolated_pages_cb);
1600 pr_info("Offlined Pages %ld\n", offlined_pages);
1601
1602
1603
1604
1605
1606 spin_lock_irqsave(&zone->lock, flags);
1607 zone->nr_isolate_pageblock -= nr_isolate_pageblock;
1608 spin_unlock_irqrestore(&zone->lock, flags);
1609
1610
1611 adjust_managed_page_count(pfn_to_page(start_pfn), -offlined_pages);
1612 zone->present_pages -= offlined_pages;
1613
1614 pgdat_resize_lock(zone->zone_pgdat, &flags);
1615 zone->zone_pgdat->node_present_pages -= offlined_pages;
1616 pgdat_resize_unlock(zone->zone_pgdat, &flags);
1617
1618 init_per_zone_wmark_min();
1619
1620 if (!populated_zone(zone)) {
1621 zone_pcp_reset(zone);
1622 build_all_zonelists(NULL);
1623 } else
1624 zone_pcp_update(zone);
1625
1626 node_states_clear_node(node, &arg);
1627 if (arg.status_change_nid >= 0) {
1628 kswapd_stop(node);
1629 kcompactd_stop(node);
1630 }
1631
1632 vm_total_pages = nr_free_pagecache_pages();
1633 writeback_set_ratelimit();
1634
1635 memory_notify(MEM_OFFLINE, &arg);
1636 mem_hotplug_done();
1637 return 0;
1638
1639failed_removal_isolated:
1640 undo_isolate_page_range(start_pfn, end_pfn, MIGRATE_MOVABLE);
1641 memory_notify(MEM_CANCEL_OFFLINE, &arg);
1642failed_removal:
1643 pr_debug("memory offlining [mem %#010llx-%#010llx] failed due to %s\n",
1644 (unsigned long long) start_pfn << PAGE_SHIFT,
1645 ((unsigned long long) end_pfn << PAGE_SHIFT) - 1,
1646 reason);
1647
1648 mem_hotplug_done();
1649 return ret;
1650}
1651
1652int offline_pages(unsigned long start_pfn, unsigned long nr_pages)
1653{
1654 return __offline_pages(start_pfn, start_pfn + nr_pages);
1655}
1656
1657static int check_memblock_offlined_cb(struct memory_block *mem, void *arg)
1658{
1659 int ret = !is_memblock_offlined(mem);
1660
1661 if (unlikely(ret)) {
1662 phys_addr_t beginpa, endpa;
1663
1664 beginpa = PFN_PHYS(section_nr_to_pfn(mem->start_section_nr));
1665 endpa = PFN_PHYS(section_nr_to_pfn(mem->end_section_nr + 1))-1;
1666 pr_warn("removing memory fails, because memory [%pa-%pa] is onlined\n",
1667 &beginpa, &endpa);
1668
1669 return -EBUSY;
1670 }
1671 return 0;
1672}
1673
1674static int check_cpu_on_node(pg_data_t *pgdat)
1675{
1676 int cpu;
1677
1678 for_each_present_cpu(cpu) {
1679 if (cpu_to_node(cpu) == pgdat->node_id)
1680
1681
1682
1683
1684 return -EBUSY;
1685 }
1686
1687 return 0;
1688}
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699void try_offline_node(int nid)
1700{
1701 pg_data_t *pgdat = NODE_DATA(nid);
1702 unsigned long start_pfn = pgdat->node_start_pfn;
1703 unsigned long end_pfn = start_pfn + pgdat->node_spanned_pages;
1704 unsigned long pfn;
1705
1706 for (pfn = start_pfn; pfn < end_pfn; pfn += PAGES_PER_SECTION) {
1707 unsigned long section_nr = pfn_to_section_nr(pfn);
1708
1709 if (!present_section_nr(section_nr))
1710 continue;
1711
1712 if (pfn_to_nid(pfn) != nid)
1713 continue;
1714
1715
1716
1717
1718
1719 return;
1720 }
1721
1722 if (check_cpu_on_node(pgdat))
1723 return;
1724
1725
1726
1727
1728
1729 node_set_offline(nid);
1730 unregister_one_node(nid);
1731}
1732EXPORT_SYMBOL(try_offline_node);
1733
1734static void __release_memory_resource(resource_size_t start,
1735 resource_size_t size)
1736{
1737 int ret;
1738
1739
1740
1741
1742
1743
1744
1745 ret = release_mem_region_adjustable(&iomem_resource, start, size);
1746 if (ret) {
1747 resource_size_t endres = start + size - 1;
1748
1749 pr_warn("Unable to release resource <%pa-%pa> (%d)\n",
1750 &start, &endres, ret);
1751 }
1752}
1753
1754static int __ref try_remove_memory(int nid, u64 start, u64 size)
1755{
1756 int rc = 0;
1757
1758 BUG_ON(check_hotplug_memory_range(start, size));
1759
1760 mem_hotplug_begin();
1761
1762
1763
1764
1765
1766
1767 rc = walk_memory_blocks(start, size, NULL, check_memblock_offlined_cb);
1768 if (rc)
1769 goto done;
1770
1771
1772 firmware_map_remove(start, start + size, "System RAM");
1773 memblock_free(start, size);
1774 memblock_remove(start, size);
1775
1776
1777 remove_memory_block_devices(start, size);
1778
1779 arch_remove_memory(nid, start, size, NULL);
1780 __release_memory_resource(start, size);
1781
1782 try_offline_node(nid);
1783
1784done:
1785 mem_hotplug_done();
1786 return rc;
1787}
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799void __remove_memory(int nid, u64 start, u64 size)
1800{
1801
1802
1803
1804
1805
1806 if (try_remove_memory(nid, start, size))
1807 BUG();
1808}
1809
1810
1811
1812
1813
1814int remove_memory(int nid, u64 start, u64 size)
1815{
1816 int rc;
1817
1818 lock_device_hotplug();
1819 rc = try_remove_memory(nid, start, size);
1820 unlock_device_hotplug();
1821
1822 return rc;
1823}
1824EXPORT_SYMBOL_GPL(remove_memory);
1825#endif
1826