Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/mm/memory_hotplug.c |
| 3 | * |
| 4 | * Copyright (C) |
| 5 | */ |
| 6 | |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 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> |
| 14 | #include <linux/module.h> |
| 15 | #include <linux/pagevec.h> |
Chandra Seetharaman | 2d1d43f | 2006-09-29 02:01:25 -0700 | [diff] [blame] | 16 | #include <linux/writeback.h> |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 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> |
KAMEZAWA Hiroyuki | 0a54703 | 2006-06-27 02:53:35 -0700 | [diff] [blame] | 24 | #include <linux/ioport.h> |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 25 | #include <linux/delay.h> |
| 26 | #include <linux/migrate.h> |
| 27 | #include <linux/page-isolation.h> |
Badari Pulavarty | 7108878 | 2008-10-18 20:25:58 -0700 | [diff] [blame] | 28 | #include <linux/pfn.h> |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 29 | #include <linux/suspend.h> |
KOSAKI Motohiro | 6d9c285 | 2009-12-14 17:58:11 -0800 | [diff] [blame] | 30 | #include <linux/mm_inline.h> |
akpm@linux-foundation.org | d96ae53 | 2010-03-05 13:41:58 -0800 | [diff] [blame] | 31 | #include <linux/firmware-map.h> |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 32 | |
| 33 | #include <asm/tlbflush.h> |
| 34 | |
Adrian Bunk | 1e5ad9a | 2008-04-28 20:40:08 +0300 | [diff] [blame] | 35 | #include "internal.h" |
| 36 | |
KOSAKI Motohiro | 20d6c96 | 2010-12-02 14:31:19 -0800 | [diff] [blame] | 37 | DEFINE_MUTEX(mem_hotplug_mutex); |
| 38 | |
| 39 | void lock_memory_hotplug(void) |
| 40 | { |
| 41 | mutex_lock(&mem_hotplug_mutex); |
| 42 | |
| 43 | /* for exclusive hibernation if CONFIG_HIBERNATION=y */ |
| 44 | lock_system_sleep(); |
| 45 | } |
| 46 | |
| 47 | void unlock_memory_hotplug(void) |
| 48 | { |
| 49 | unlock_system_sleep(); |
| 50 | mutex_unlock(&mem_hotplug_mutex); |
| 51 | } |
| 52 | |
| 53 | |
Keith Mannthey | 45e0b78 | 2006-09-30 23:27:09 -0700 | [diff] [blame] | 54 | /* add this memory to iomem resource */ |
| 55 | static struct resource *register_memory_resource(u64 start, u64 size) |
| 56 | { |
| 57 | struct resource *res; |
| 58 | res = kzalloc(sizeof(struct resource), GFP_KERNEL); |
| 59 | BUG_ON(!res); |
| 60 | |
| 61 | res->name = "System RAM"; |
| 62 | res->start = start; |
| 63 | res->end = start + size - 1; |
Yasunori Goto | 887c3cb | 2007-11-14 16:59:20 -0800 | [diff] [blame] | 64 | res->flags = IORESOURCE_MEM | IORESOURCE_BUSY; |
Keith Mannthey | 45e0b78 | 2006-09-30 23:27:09 -0700 | [diff] [blame] | 65 | if (request_resource(&iomem_resource, res) < 0) { |
| 66 | printk("System RAM resource %llx - %llx cannot be added\n", |
| 67 | (unsigned long long)res->start, (unsigned long long)res->end); |
| 68 | kfree(res); |
| 69 | res = NULL; |
| 70 | } |
| 71 | return res; |
| 72 | } |
| 73 | |
| 74 | static void release_memory_resource(struct resource *res) |
| 75 | { |
| 76 | if (!res) |
| 77 | return; |
| 78 | release_resource(res); |
| 79 | kfree(res); |
| 80 | return; |
| 81 | } |
| 82 | |
Keith Mannthey | 5394702 | 2006-09-30 23:27:08 -0700 | [diff] [blame] | 83 | #ifdef CONFIG_MEMORY_HOTPLUG_SPARSE |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 84 | #ifndef CONFIG_SPARSEMEM_VMEMMAP |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 85 | static void get_page_bootmem(unsigned long info, struct page *page, |
| 86 | unsigned long type) |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 87 | { |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 88 | page->lru.next = (struct list_head *) type; |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 89 | SetPagePrivate(page); |
| 90 | set_page_private(page, info); |
| 91 | atomic_inc(&page->_count); |
| 92 | } |
| 93 | |
Rakib Mullick | 23ce932 | 2009-12-14 17:59:44 -0800 | [diff] [blame] | 94 | /* reference to __meminit __free_pages_bootmem is valid |
| 95 | * so use __ref to tell modpost not to generate a warning */ |
| 96 | void __ref put_page_bootmem(struct page *page) |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 97 | { |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 98 | unsigned long type; |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 99 | |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 100 | type = (unsigned long) page->lru.next; |
| 101 | BUG_ON(type < MEMORY_HOTPLUG_MIN_BOOTMEM_TYPE || |
| 102 | type > MEMORY_HOTPLUG_MAX_BOOTMEM_TYPE); |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 103 | |
| 104 | if (atomic_dec_return(&page->_count) == 1) { |
| 105 | ClearPagePrivate(page); |
| 106 | set_page_private(page, 0); |
Andrea Arcangeli | 5f24ce5 | 2011-01-13 15:47:00 -0800 | [diff] [blame] | 107 | INIT_LIST_HEAD(&page->lru); |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 108 | __free_pages_bootmem(page, 0); |
| 109 | } |
| 110 | |
| 111 | } |
| 112 | |
Adrian Bunk | d92bc31 | 2008-07-23 21:28:12 -0700 | [diff] [blame] | 113 | static void register_page_bootmem_info_section(unsigned long start_pfn) |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 114 | { |
| 115 | unsigned long *usemap, mapsize, section_nr, i; |
| 116 | struct mem_section *ms; |
| 117 | struct page *page, *memmap; |
| 118 | |
| 119 | if (!pfn_valid(start_pfn)) |
| 120 | return; |
| 121 | |
| 122 | section_nr = pfn_to_section_nr(start_pfn); |
| 123 | ms = __nr_to_section(section_nr); |
| 124 | |
| 125 | /* Get section's memmap address */ |
| 126 | memmap = sparse_decode_mem_map(ms->section_mem_map, section_nr); |
| 127 | |
| 128 | /* |
| 129 | * Get page for the memmap's phys address |
| 130 | * XXX: need more consideration for sparse_vmemmap... |
| 131 | */ |
| 132 | page = virt_to_page(memmap); |
| 133 | mapsize = sizeof(struct page) * PAGES_PER_SECTION; |
| 134 | mapsize = PAGE_ALIGN(mapsize) >> PAGE_SHIFT; |
| 135 | |
| 136 | /* remember memmap's page */ |
| 137 | for (i = 0; i < mapsize; i++, page++) |
| 138 | get_page_bootmem(section_nr, page, SECTION_INFO); |
| 139 | |
| 140 | usemap = __nr_to_section(section_nr)->pageblock_flags; |
| 141 | page = virt_to_page(usemap); |
| 142 | |
| 143 | mapsize = PAGE_ALIGN(usemap_size()) >> PAGE_SHIFT; |
| 144 | |
| 145 | for (i = 0; i < mapsize; i++, page++) |
Yasunori Goto | af370fb | 2008-07-23 21:28:17 -0700 | [diff] [blame] | 146 | get_page_bootmem(section_nr, page, MIX_SECTION_INFO); |
Yasunori Goto | 0475327 | 2008-04-28 02:13:31 -0700 | [diff] [blame] | 147 | |
| 148 | } |
| 149 | |
| 150 | void register_page_bootmem_info_node(struct pglist_data *pgdat) |
| 151 | { |
| 152 | unsigned long i, pfn, end_pfn, nr_pages; |
| 153 | int node = pgdat->node_id; |
| 154 | struct page *page; |
| 155 | struct zone *zone; |
| 156 | |
| 157 | nr_pages = PAGE_ALIGN(sizeof(struct pglist_data)) >> PAGE_SHIFT; |
| 158 | page = virt_to_page(pgdat); |
| 159 | |
| 160 | for (i = 0; i < nr_pages; i++, page++) |
| 161 | get_page_bootmem(node, page, NODE_INFO); |
| 162 | |
| 163 | zone = &pgdat->node_zones[0]; |
| 164 | for (; zone < pgdat->node_zones + MAX_NR_ZONES - 1; zone++) { |
| 165 | if (zone->wait_table) { |
| 166 | nr_pages = zone->wait_table_hash_nr_entries |
| 167 | * sizeof(wait_queue_head_t); |
| 168 | nr_pages = PAGE_ALIGN(nr_pages) >> PAGE_SHIFT; |
| 169 | page = virt_to_page(zone->wait_table); |
| 170 | |
| 171 | for (i = 0; i < nr_pages; i++, page++) |
| 172 | get_page_bootmem(node, page, NODE_INFO); |
| 173 | } |
| 174 | } |
| 175 | |
| 176 | pfn = pgdat->node_start_pfn; |
| 177 | end_pfn = pfn + pgdat->node_spanned_pages; |
| 178 | |
| 179 | /* register_section info */ |
| 180 | for (; pfn < end_pfn; pfn += PAGES_PER_SECTION) |
| 181 | register_page_bootmem_info_section(pfn); |
| 182 | |
| 183 | } |
| 184 | #endif /* !CONFIG_SPARSEMEM_VMEMMAP */ |
| 185 | |
Heiko Carstens | 76cdd58 | 2008-05-14 16:05:52 -0700 | [diff] [blame] | 186 | static void grow_zone_span(struct zone *zone, unsigned long start_pfn, |
| 187 | unsigned long end_pfn) |
| 188 | { |
| 189 | unsigned long old_zone_end_pfn; |
| 190 | |
| 191 | zone_span_writelock(zone); |
| 192 | |
| 193 | old_zone_end_pfn = zone->zone_start_pfn + zone->spanned_pages; |
| 194 | if (start_pfn < zone->zone_start_pfn) |
| 195 | zone->zone_start_pfn = start_pfn; |
| 196 | |
| 197 | zone->spanned_pages = max(old_zone_end_pfn, end_pfn) - |
| 198 | zone->zone_start_pfn; |
| 199 | |
| 200 | zone_span_writeunlock(zone); |
| 201 | } |
| 202 | |
| 203 | static void grow_pgdat_span(struct pglist_data *pgdat, unsigned long start_pfn, |
| 204 | unsigned long end_pfn) |
| 205 | { |
| 206 | unsigned long old_pgdat_end_pfn = |
| 207 | pgdat->node_start_pfn + pgdat->node_spanned_pages; |
| 208 | |
| 209 | if (start_pfn < pgdat->node_start_pfn) |
| 210 | pgdat->node_start_pfn = start_pfn; |
| 211 | |
| 212 | pgdat->node_spanned_pages = max(old_pgdat_end_pfn, end_pfn) - |
| 213 | pgdat->node_start_pfn; |
| 214 | } |
| 215 | |
Al Viro | 3116848 | 2008-11-22 17:33:24 +0000 | [diff] [blame] | 216 | static int __meminit __add_zone(struct zone *zone, unsigned long phys_start_pfn) |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 217 | { |
| 218 | struct pglist_data *pgdat = zone->zone_pgdat; |
| 219 | int nr_pages = PAGES_PER_SECTION; |
| 220 | int nid = pgdat->node_id; |
| 221 | int zone_type; |
Heiko Carstens | 76cdd58 | 2008-05-14 16:05:52 -0700 | [diff] [blame] | 222 | unsigned long flags; |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 223 | |
| 224 | zone_type = zone - pgdat->node_zones; |
Heiko Carstens | 76cdd58 | 2008-05-14 16:05:52 -0700 | [diff] [blame] | 225 | if (!zone->wait_table) { |
| 226 | int ret; |
| 227 | |
| 228 | ret = init_currently_empty_zone(zone, phys_start_pfn, |
| 229 | nr_pages, MEMMAP_HOTPLUG); |
| 230 | if (ret) |
| 231 | return ret; |
| 232 | } |
| 233 | pgdat_resize_lock(zone->zone_pgdat, &flags); |
| 234 | grow_zone_span(zone, phys_start_pfn, phys_start_pfn + nr_pages); |
| 235 | grow_pgdat_span(zone->zone_pgdat, phys_start_pfn, |
| 236 | phys_start_pfn + nr_pages); |
| 237 | pgdat_resize_unlock(zone->zone_pgdat, &flags); |
Dave Hansen | a2f3aa02 | 2007-01-10 23:15:30 -0800 | [diff] [blame] | 238 | memmap_init_zone(nr_pages, nid, zone_type, |
| 239 | phys_start_pfn, MEMMAP_HOTPLUG); |
Yasunori Goto | 718127c | 2006-06-23 02:03:10 -0700 | [diff] [blame] | 240 | return 0; |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 241 | } |
| 242 | |
Gary Hade | c04fc58 | 2009-01-06 14:39:14 -0800 | [diff] [blame] | 243 | static int __meminit __add_section(int nid, struct zone *zone, |
| 244 | unsigned long phys_start_pfn) |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 245 | { |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 246 | int nr_pages = PAGES_PER_SECTION; |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 247 | int ret; |
| 248 | |
KAMEZAWA Hiroyuki | ebd1530 | 2006-08-05 12:15:06 -0700 | [diff] [blame] | 249 | if (pfn_valid(phys_start_pfn)) |
| 250 | return -EEXIST; |
| 251 | |
Dave Hansen | 0b0acbe | 2005-10-29 18:16:55 -0700 | [diff] [blame] | 252 | ret = sparse_add_one_section(zone, phys_start_pfn, nr_pages); |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 253 | |
| 254 | if (ret < 0) |
| 255 | return ret; |
| 256 | |
Yasunori Goto | 718127c | 2006-06-23 02:03:10 -0700 | [diff] [blame] | 257 | ret = __add_zone(zone, phys_start_pfn); |
| 258 | |
| 259 | if (ret < 0) |
| 260 | return ret; |
| 261 | |
Gary Hade | c04fc58 | 2009-01-06 14:39:14 -0800 | [diff] [blame] | 262 | return register_new_memory(nid, __pfn_to_section(phys_start_pfn)); |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 263 | } |
| 264 | |
Yasunori Goto | 0c0a4a5 | 2008-04-28 02:13:34 -0700 | [diff] [blame] | 265 | #ifdef CONFIG_SPARSEMEM_VMEMMAP |
| 266 | static int __remove_section(struct zone *zone, struct mem_section *ms) |
| 267 | { |
| 268 | /* |
| 269 | * XXX: Freeing memmap with vmemmap is not implement yet. |
| 270 | * This should be removed later. |
| 271 | */ |
| 272 | return -EBUSY; |
| 273 | } |
| 274 | #else |
Badari Pulavarty | ea01ea9 | 2008-04-28 02:12:01 -0700 | [diff] [blame] | 275 | static int __remove_section(struct zone *zone, struct mem_section *ms) |
| 276 | { |
| 277 | unsigned long flags; |
| 278 | struct pglist_data *pgdat = zone->zone_pgdat; |
| 279 | int ret = -EINVAL; |
| 280 | |
| 281 | if (!valid_section(ms)) |
| 282 | return ret; |
| 283 | |
| 284 | ret = unregister_memory_section(ms); |
| 285 | if (ret) |
| 286 | return ret; |
| 287 | |
| 288 | pgdat_resize_lock(pgdat, &flags); |
| 289 | sparse_remove_one_section(zone, ms); |
| 290 | pgdat_resize_unlock(pgdat, &flags); |
| 291 | return 0; |
| 292 | } |
Yasunori Goto | 0c0a4a5 | 2008-04-28 02:13:34 -0700 | [diff] [blame] | 293 | #endif |
Badari Pulavarty | ea01ea9 | 2008-04-28 02:12:01 -0700 | [diff] [blame] | 294 | |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 295 | /* |
| 296 | * Reasonably generic function for adding memory. It is |
| 297 | * expected that archs that support memory hotplug will |
| 298 | * call this function after deciding the zone to which to |
| 299 | * add the new pages. |
| 300 | */ |
Gary Hade | c04fc58 | 2009-01-06 14:39:14 -0800 | [diff] [blame] | 301 | int __ref __add_pages(int nid, struct zone *zone, unsigned long phys_start_pfn, |
| 302 | unsigned long nr_pages) |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 303 | { |
| 304 | unsigned long i; |
| 305 | int err = 0; |
KAMEZAWA Hiroyuki | 6f71271 | 2006-08-05 12:14:58 -0700 | [diff] [blame] | 306 | int start_sec, end_sec; |
| 307 | /* during initialize mem_map, align hot-added range to section */ |
| 308 | start_sec = pfn_to_section_nr(phys_start_pfn); |
| 309 | end_sec = pfn_to_section_nr(phys_start_pfn + nr_pages - 1); |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 310 | |
KAMEZAWA Hiroyuki | 6f71271 | 2006-08-05 12:14:58 -0700 | [diff] [blame] | 311 | for (i = start_sec; i <= end_sec; i++) { |
Gary Hade | c04fc58 | 2009-01-06 14:39:14 -0800 | [diff] [blame] | 312 | err = __add_section(nid, zone, i << PFN_SECTION_SHIFT); |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 313 | |
KAMEZAWA Hiroyuki | 6f71271 | 2006-08-05 12:14:58 -0700 | [diff] [blame] | 314 | /* |
Simon Arlott | 183ff22 | 2007-10-20 01:27:18 +0200 | [diff] [blame] | 315 | * EEXIST is finally dealt with by ioresource collision |
KAMEZAWA Hiroyuki | 6f71271 | 2006-08-05 12:14:58 -0700 | [diff] [blame] | 316 | * check. see add_memory() => register_memory_resource() |
| 317 | * Warning will be printed if there is collision. |
Joel H Schopp | bed120c | 2006-05-01 12:16:11 -0700 | [diff] [blame] | 318 | */ |
| 319 | if (err && (err != -EEXIST)) |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 320 | break; |
KAMEZAWA Hiroyuki | 6f71271 | 2006-08-05 12:14:58 -0700 | [diff] [blame] | 321 | err = 0; |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 322 | } |
| 323 | |
| 324 | return err; |
| 325 | } |
Joel H Schopp | bed120c | 2006-05-01 12:16:11 -0700 | [diff] [blame] | 326 | EXPORT_SYMBOL_GPL(__add_pages); |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 327 | |
Badari Pulavarty | ea01ea9 | 2008-04-28 02:12:01 -0700 | [diff] [blame] | 328 | /** |
| 329 | * __remove_pages() - remove sections of pages from a zone |
| 330 | * @zone: zone from which pages need to be removed |
| 331 | * @phys_start_pfn: starting pageframe (must be aligned to start of a section) |
| 332 | * @nr_pages: number of pages to remove (must be multiple of section size) |
| 333 | * |
| 334 | * Generic helper function to remove section mappings and sysfs entries |
| 335 | * for the section of the memory we are removing. Caller needs to make |
| 336 | * sure that pages are marked reserved and zones are adjust properly by |
| 337 | * calling offline_pages(). |
| 338 | */ |
| 339 | int __remove_pages(struct zone *zone, unsigned long phys_start_pfn, |
| 340 | unsigned long nr_pages) |
| 341 | { |
| 342 | unsigned long i, ret = 0; |
| 343 | int sections_to_remove; |
| 344 | |
| 345 | /* |
| 346 | * We can only remove entire sections |
| 347 | */ |
| 348 | BUG_ON(phys_start_pfn & ~PAGE_SECTION_MASK); |
| 349 | BUG_ON(nr_pages % PAGES_PER_SECTION); |
| 350 | |
Badari Pulavarty | ea01ea9 | 2008-04-28 02:12:01 -0700 | [diff] [blame] | 351 | sections_to_remove = nr_pages / PAGES_PER_SECTION; |
| 352 | for (i = 0; i < sections_to_remove; i++) { |
| 353 | unsigned long pfn = phys_start_pfn + i*PAGES_PER_SECTION; |
Nathan Fontenot | de7f0cb | 2008-10-18 20:27:14 -0700 | [diff] [blame] | 354 | release_mem_region(pfn << PAGE_SHIFT, |
| 355 | PAGES_PER_SECTION << PAGE_SHIFT); |
Badari Pulavarty | ea01ea9 | 2008-04-28 02:12:01 -0700 | [diff] [blame] | 356 | ret = __remove_section(zone, __pfn_to_section(pfn)); |
| 357 | if (ret) |
| 358 | break; |
| 359 | } |
| 360 | return ret; |
| 361 | } |
| 362 | EXPORT_SYMBOL_GPL(__remove_pages); |
| 363 | |
Jeremy Fitzhardinge | 180c06e | 2008-04-28 02:12:03 -0700 | [diff] [blame] | 364 | void online_page(struct page *page) |
| 365 | { |
Jan Beulich | 4738e1b | 2009-09-21 17:03:03 -0700 | [diff] [blame] | 366 | unsigned long pfn = page_to_pfn(page); |
| 367 | |
Jeremy Fitzhardinge | 180c06e | 2008-04-28 02:12:03 -0700 | [diff] [blame] | 368 | totalram_pages++; |
Jan Beulich | 4738e1b | 2009-09-21 17:03:03 -0700 | [diff] [blame] | 369 | if (pfn >= num_physpages) |
| 370 | num_physpages = pfn + 1; |
Jeremy Fitzhardinge | 180c06e | 2008-04-28 02:12:03 -0700 | [diff] [blame] | 371 | |
| 372 | #ifdef CONFIG_HIGHMEM |
| 373 | if (PageHighMem(page)) |
| 374 | totalhigh_pages++; |
| 375 | #endif |
| 376 | |
Jeremy Fitzhardinge | 180c06e | 2008-04-28 02:12:03 -0700 | [diff] [blame] | 377 | ClearPageReserved(page); |
| 378 | init_page_count(page); |
| 379 | __free_page(page); |
| 380 | } |
| 381 | |
KAMEZAWA Hiroyuki | 75884fb | 2007-10-16 01:26:10 -0700 | [diff] [blame] | 382 | static int online_pages_range(unsigned long start_pfn, unsigned long nr_pages, |
| 383 | void *arg) |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 384 | { |
| 385 | unsigned long i; |
KAMEZAWA Hiroyuki | 75884fb | 2007-10-16 01:26:10 -0700 | [diff] [blame] | 386 | unsigned long onlined_pages = *(unsigned long *)arg; |
| 387 | struct page *page; |
| 388 | if (PageReserved(pfn_to_page(start_pfn))) |
| 389 | for (i = 0; i < nr_pages; i++) { |
| 390 | page = pfn_to_page(start_pfn + i); |
| 391 | online_page(page); |
| 392 | onlined_pages++; |
| 393 | } |
| 394 | *(unsigned long *)arg = onlined_pages; |
| 395 | return 0; |
| 396 | } |
| 397 | |
| 398 | |
KOSAKI Motohiro | 839a4fc | 2011-05-24 17:11:31 -0700 | [diff] [blame] | 399 | int __ref online_pages(unsigned long pfn, unsigned long nr_pages) |
KAMEZAWA Hiroyuki | 75884fb | 2007-10-16 01:26:10 -0700 | [diff] [blame] | 400 | { |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 401 | unsigned long onlined_pages = 0; |
| 402 | struct zone *zone; |
Yasunori Goto | 6811378 | 2006-06-23 02:03:11 -0700 | [diff] [blame] | 403 | int need_zonelists_rebuild = 0; |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 404 | int nid; |
| 405 | int ret; |
| 406 | struct memory_notify arg; |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 407 | |
KAMEZAWA Hiroyuki | 925268a | 2011-01-11 16:44:01 +0900 | [diff] [blame] | 408 | lock_memory_hotplug(); |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 409 | arg.start_pfn = pfn; |
| 410 | arg.nr_pages = nr_pages; |
| 411 | arg.status_change_nid = -1; |
| 412 | |
| 413 | nid = page_to_nid(pfn_to_page(pfn)); |
| 414 | if (node_present_pages(nid) == 0) |
| 415 | arg.status_change_nid = nid; |
| 416 | |
| 417 | ret = memory_notify(MEM_GOING_ONLINE, &arg); |
| 418 | ret = notifier_to_errno(ret); |
| 419 | if (ret) { |
| 420 | memory_notify(MEM_CANCEL_ONLINE, &arg); |
KAMEZAWA Hiroyuki | 925268a | 2011-01-11 16:44:01 +0900 | [diff] [blame] | 421 | unlock_memory_hotplug(); |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 422 | return ret; |
| 423 | } |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 424 | /* |
| 425 | * This doesn't need a lock to do pfn_to_page(). |
| 426 | * The section can't be removed here because of the |
Daniel Walker | da19cbc | 2008-02-04 23:35:47 -0800 | [diff] [blame] | 427 | * memory_block->state_mutex. |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 428 | */ |
| 429 | zone = page_zone(pfn_to_page(pfn)); |
Yasunori Goto | 6811378 | 2006-06-23 02:03:11 -0700 | [diff] [blame] | 430 | /* |
| 431 | * If this zone is not populated, then it is not in zonelist. |
| 432 | * This means the page allocator ignores this zone. |
| 433 | * So, zonelist must be updated after online. |
| 434 | */ |
Haicheng Li | 4eaf3f6 | 2010-05-24 14:32:52 -0700 | [diff] [blame] | 435 | mutex_lock(&zonelists_mutex); |
Yasunori Goto | 6811378 | 2006-06-23 02:03:11 -0700 | [diff] [blame] | 436 | if (!populated_zone(zone)) |
| 437 | need_zonelists_rebuild = 1; |
| 438 | |
KAMEZAWA Hiroyuki | 908eedc | 2009-09-22 16:45:46 -0700 | [diff] [blame] | 439 | ret = walk_system_ram_range(pfn, nr_pages, &onlined_pages, |
KAMEZAWA Hiroyuki | 75884fb | 2007-10-16 01:26:10 -0700 | [diff] [blame] | 440 | online_pages_range); |
Geoff Levand | fd8a422 | 2008-05-14 16:05:50 -0700 | [diff] [blame] | 441 | if (ret) { |
Haicheng Li | 4eaf3f6 | 2010-05-24 14:32:52 -0700 | [diff] [blame] | 442 | mutex_unlock(&zonelists_mutex); |
Geoff Levand | fd8a422 | 2008-05-14 16:05:50 -0700 | [diff] [blame] | 443 | printk(KERN_DEBUG "online_pages %lx at %lx failed\n", |
| 444 | nr_pages, pfn); |
| 445 | memory_notify(MEM_CANCEL_ONLINE, &arg); |
KAMEZAWA Hiroyuki | 925268a | 2011-01-11 16:44:01 +0900 | [diff] [blame] | 446 | unlock_memory_hotplug(); |
Geoff Levand | fd8a422 | 2008-05-14 16:05:50 -0700 | [diff] [blame] | 447 | return ret; |
| 448 | } |
| 449 | |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 450 | zone->present_pages += onlined_pages; |
Yasunori Goto | f2937be | 2006-03-09 17:33:51 -0800 | [diff] [blame] | 451 | zone->zone_pgdat->node_present_pages += onlined_pages; |
Haicheng Li | 1f52250 | 2010-05-24 14:32:51 -0700 | [diff] [blame] | 452 | if (need_zonelists_rebuild) |
| 453 | build_all_zonelists(zone); |
| 454 | else |
| 455 | zone_pcp_update(zone); |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 456 | |
Haicheng Li | 4eaf3f6 | 2010-05-24 14:32:52 -0700 | [diff] [blame] | 457 | mutex_unlock(&zonelists_mutex); |
KOSAKI Motohiro | 1b79acc | 2011-05-24 17:11:32 -0700 | [diff] [blame] | 458 | |
| 459 | init_per_zone_wmark_min(); |
| 460 | |
Christoph Lameter | 7ea1530 | 2007-10-16 01:25:29 -0700 | [diff] [blame] | 461 | if (onlined_pages) { |
| 462 | kswapd_run(zone_to_nid(zone)); |
| 463 | node_set_state(zone_to_nid(zone), N_HIGH_MEMORY); |
| 464 | } |
Dave Hansen | 61b1399 | 2005-10-29 18:16:56 -0700 | [diff] [blame] | 465 | |
Haicheng Li | 1f52250 | 2010-05-24 14:32:51 -0700 | [diff] [blame] | 466 | vm_total_pages = nr_free_pagecache_pages(); |
Kent Liu | 2f7f24e | 2008-07-23 21:28:18 -0700 | [diff] [blame] | 467 | |
Chandra Seetharaman | 2d1d43f | 2006-09-29 02:01:25 -0700 | [diff] [blame] | 468 | writeback_set_ratelimit(); |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 469 | |
| 470 | if (onlined_pages) |
| 471 | memory_notify(MEM_ONLINE, &arg); |
KAMEZAWA Hiroyuki | 925268a | 2011-01-11 16:44:01 +0900 | [diff] [blame] | 472 | unlock_memory_hotplug(); |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 473 | |
Dave Hansen | 3947be1 | 2005-10-29 18:16:54 -0700 | [diff] [blame] | 474 | return 0; |
| 475 | } |
Keith Mannthey | 5394702 | 2006-09-30 23:27:08 -0700 | [diff] [blame] | 476 | #endif /* CONFIG_MEMORY_HOTPLUG_SPARSE */ |
Yasunori Goto | bc02af9 | 2006-06-27 02:53:30 -0700 | [diff] [blame] | 477 | |
Hidetoshi Seto | e131933 | 2009-11-17 14:06:18 -0800 | [diff] [blame] | 478 | /* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */ |
| 479 | static pg_data_t __ref *hotadd_new_pgdat(int nid, u64 start) |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 480 | { |
| 481 | struct pglist_data *pgdat; |
| 482 | unsigned long zones_size[MAX_NR_ZONES] = {0}; |
| 483 | unsigned long zholes_size[MAX_NR_ZONES] = {0}; |
| 484 | unsigned long start_pfn = start >> PAGE_SHIFT; |
| 485 | |
| 486 | pgdat = arch_alloc_nodedata(nid); |
| 487 | if (!pgdat) |
| 488 | return NULL; |
| 489 | |
| 490 | arch_refresh_nodedata(nid, pgdat); |
| 491 | |
| 492 | /* we can use NODE_DATA(nid) from here */ |
| 493 | |
| 494 | /* init node's zones as empty zones, we don't have any present pages.*/ |
Johannes Weiner | 9109fb7 | 2008-07-23 21:27:20 -0700 | [diff] [blame] | 495 | free_area_init_node(nid, zones_size, start_pfn, zholes_size); |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 496 | |
KAMEZAWA Hiroyuki | 959ecc4 | 2011-06-15 15:08:38 -0700 | [diff] [blame] | 497 | /* |
| 498 | * The node we allocated has no zone fallback lists. For avoiding |
| 499 | * to access not-initialized zonelist, build here. |
| 500 | */ |
| 501 | build_all_zonelists(NULL); |
| 502 | |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 503 | return pgdat; |
| 504 | } |
| 505 | |
| 506 | static void rollback_node_hotadd(int nid, pg_data_t *pgdat) |
| 507 | { |
| 508 | arch_refresh_nodedata(nid, NULL); |
| 509 | arch_free_nodedata(pgdat); |
| 510 | return; |
| 511 | } |
| 512 | |
KAMEZAWA Hiroyuki | 0a54703 | 2006-06-27 02:53:35 -0700 | [diff] [blame] | 513 | |
minskey guo | cf23422 | 2010-05-24 14:32:41 -0700 | [diff] [blame] | 514 | /* |
| 515 | * called by cpu_up() to online a node without onlined memory. |
| 516 | */ |
| 517 | int mem_online_node(int nid) |
| 518 | { |
| 519 | pg_data_t *pgdat; |
| 520 | int ret; |
| 521 | |
KOSAKI Motohiro | 20d6c96 | 2010-12-02 14:31:19 -0800 | [diff] [blame] | 522 | lock_memory_hotplug(); |
minskey guo | cf23422 | 2010-05-24 14:32:41 -0700 | [diff] [blame] | 523 | pgdat = hotadd_new_pgdat(nid, 0); |
| 524 | if (pgdat) { |
| 525 | ret = -ENOMEM; |
| 526 | goto out; |
| 527 | } |
| 528 | node_set_online(nid); |
| 529 | ret = register_one_node(nid); |
| 530 | BUG_ON(ret); |
| 531 | |
| 532 | out: |
KOSAKI Motohiro | 20d6c96 | 2010-12-02 14:31:19 -0800 | [diff] [blame] | 533 | unlock_memory_hotplug(); |
minskey guo | cf23422 | 2010-05-24 14:32:41 -0700 | [diff] [blame] | 534 | return ret; |
| 535 | } |
| 536 | |
Al Viro | 3116848 | 2008-11-22 17:33:24 +0000 | [diff] [blame] | 537 | /* we are OK calling __meminit stuff here - we have CONFIG_MEMORY_HOTPLUG */ |
| 538 | int __ref add_memory(int nid, u64 start, u64 size) |
Yasunori Goto | bc02af9 | 2006-06-27 02:53:30 -0700 | [diff] [blame] | 539 | { |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 540 | pg_data_t *pgdat = NULL; |
| 541 | int new_pgdat = 0; |
KAMEZAWA Hiroyuki | ebd1530 | 2006-08-05 12:15:06 -0700 | [diff] [blame] | 542 | struct resource *res; |
Yasunori Goto | bc02af9 | 2006-06-27 02:53:30 -0700 | [diff] [blame] | 543 | int ret; |
| 544 | |
KOSAKI Motohiro | 20d6c96 | 2010-12-02 14:31:19 -0800 | [diff] [blame] | 545 | lock_memory_hotplug(); |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 546 | |
KAMEZAWA Hiroyuki | ebd1530 | 2006-08-05 12:15:06 -0700 | [diff] [blame] | 547 | res = register_memory_resource(start, size); |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 548 | ret = -EEXIST; |
KAMEZAWA Hiroyuki | ebd1530 | 2006-08-05 12:15:06 -0700 | [diff] [blame] | 549 | if (!res) |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 550 | goto out; |
KAMEZAWA Hiroyuki | ebd1530 | 2006-08-05 12:15:06 -0700 | [diff] [blame] | 551 | |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 552 | if (!node_online(nid)) { |
| 553 | pgdat = hotadd_new_pgdat(nid, start); |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 554 | ret = -ENOMEM; |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 555 | if (!pgdat) |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 556 | goto out; |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 557 | new_pgdat = 1; |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 558 | } |
| 559 | |
Yasunori Goto | bc02af9 | 2006-06-27 02:53:30 -0700 | [diff] [blame] | 560 | /* call arch's memory hotadd */ |
| 561 | ret = arch_add_memory(nid, start, size); |
| 562 | |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 563 | if (ret < 0) |
| 564 | goto error; |
| 565 | |
Yasunori Goto | 0fc4415 | 2006-06-27 02:53:38 -0700 | [diff] [blame] | 566 | /* we online node here. we can't roll back from here. */ |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 567 | node_set_online(nid); |
| 568 | |
Yasunori Goto | 0fc4415 | 2006-06-27 02:53:38 -0700 | [diff] [blame] | 569 | if (new_pgdat) { |
| 570 | ret = register_one_node(nid); |
| 571 | /* |
| 572 | * If sysfs file of new node can't create, cpu on the node |
| 573 | * can't be hot-added. There is no rollback way now. |
| 574 | * So, check by BUG_ON() to catch it reluctantly.. |
| 575 | */ |
| 576 | BUG_ON(ret); |
| 577 | } |
| 578 | |
akpm@linux-foundation.org | d96ae53 | 2010-03-05 13:41:58 -0800 | [diff] [blame] | 579 | /* create new memmap entry */ |
| 580 | firmware_map_add_hotplug(start, start + size, "System RAM"); |
| 581 | |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 582 | goto out; |
| 583 | |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 584 | error: |
| 585 | /* rollback pgdat allocation and others */ |
| 586 | if (new_pgdat) |
| 587 | rollback_node_hotadd(nid, pgdat); |
KAMEZAWA Hiroyuki | ebd1530 | 2006-08-05 12:15:06 -0700 | [diff] [blame] | 588 | if (res) |
| 589 | release_memory_resource(res); |
Yasunori Goto | 9af3c2d | 2006-06-27 02:53:34 -0700 | [diff] [blame] | 590 | |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 591 | out: |
KOSAKI Motohiro | 20d6c96 | 2010-12-02 14:31:19 -0800 | [diff] [blame] | 592 | unlock_memory_hotplug(); |
Yasunori Goto | bc02af9 | 2006-06-27 02:53:30 -0700 | [diff] [blame] | 593 | return ret; |
| 594 | } |
| 595 | EXPORT_SYMBOL_GPL(add_memory); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 596 | |
| 597 | #ifdef CONFIG_MEMORY_HOTREMOVE |
| 598 | /* |
Badari Pulavarty | 5c755e9 | 2008-07-23 21:28:19 -0700 | [diff] [blame] | 599 | * A free page on the buddy free lists (not the per-cpu lists) has PageBuddy |
| 600 | * set and the size of the free page is given by page_order(). Using this, |
| 601 | * the function determines if the pageblock contains only free pages. |
| 602 | * Due to buddy contraints, a free page at least the size of a pageblock will |
| 603 | * be located at the start of the pageblock |
| 604 | */ |
| 605 | static inline int pageblock_free(struct page *page) |
| 606 | { |
| 607 | return PageBuddy(page) && page_order(page) >= pageblock_order; |
| 608 | } |
| 609 | |
| 610 | /* Return the start of the next active pageblock after a given page */ |
| 611 | static struct page *next_active_pageblock(struct page *page) |
| 612 | { |
Badari Pulavarty | 5c755e9 | 2008-07-23 21:28:19 -0700 | [diff] [blame] | 613 | /* Ensure the starting page is pageblock-aligned */ |
| 614 | BUG_ON(page_to_pfn(page) & (pageblock_nr_pages - 1)); |
| 615 | |
Badari Pulavarty | 5c755e9 | 2008-07-23 21:28:19 -0700 | [diff] [blame] | 616 | /* If the entire pageblock is free, move to the end of free page */ |
KAMEZAWA Hiroyuki | 0dcc48c | 2010-09-09 16:38:01 -0700 | [diff] [blame] | 617 | if (pageblock_free(page)) { |
| 618 | int order; |
| 619 | /* be careful. we don't have locks, page_order can be changed.*/ |
| 620 | order = page_order(page); |
| 621 | if ((order < MAX_ORDER) && (order >= pageblock_order)) |
| 622 | return page + (1 << order); |
| 623 | } |
Badari Pulavarty | 5c755e9 | 2008-07-23 21:28:19 -0700 | [diff] [blame] | 624 | |
KAMEZAWA Hiroyuki | 0dcc48c | 2010-09-09 16:38:01 -0700 | [diff] [blame] | 625 | return page + pageblock_nr_pages; |
Badari Pulavarty | 5c755e9 | 2008-07-23 21:28:19 -0700 | [diff] [blame] | 626 | } |
| 627 | |
| 628 | /* Checks if this range of memory is likely to be hot-removable. */ |
| 629 | int is_mem_section_removable(unsigned long start_pfn, unsigned long nr_pages) |
| 630 | { |
Badari Pulavarty | 5c755e9 | 2008-07-23 21:28:19 -0700 | [diff] [blame] | 631 | struct page *page = pfn_to_page(start_pfn); |
| 632 | struct page *end_page = page + nr_pages; |
| 633 | |
| 634 | /* Check the starting page of each pageblock within the range */ |
| 635 | for (; page < end_page; page = next_active_pageblock(page)) { |
KAMEZAWA Hiroyuki | 49ac825 | 2010-10-26 14:21:30 -0700 | [diff] [blame] | 636 | if (!is_pageblock_removable_nolock(page)) |
Badari Pulavarty | 5c755e9 | 2008-07-23 21:28:19 -0700 | [diff] [blame] | 637 | return 0; |
KAMEZAWA Hiroyuki | 49ac825 | 2010-10-26 14:21:30 -0700 | [diff] [blame] | 638 | cond_resched(); |
Badari Pulavarty | 5c755e9 | 2008-07-23 21:28:19 -0700 | [diff] [blame] | 639 | } |
| 640 | |
| 641 | /* All pageblocks in the memory block are likely to be hot-removable */ |
| 642 | return 1; |
| 643 | } |
| 644 | |
| 645 | /* |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 646 | * Confirm all pages in a range [start, end) is belongs to the same zone. |
| 647 | */ |
| 648 | static int test_pages_in_a_zone(unsigned long start_pfn, unsigned long end_pfn) |
| 649 | { |
| 650 | unsigned long pfn; |
| 651 | struct zone *zone = NULL; |
| 652 | struct page *page; |
| 653 | int i; |
| 654 | for (pfn = start_pfn; |
| 655 | pfn < end_pfn; |
| 656 | pfn += MAX_ORDER_NR_PAGES) { |
| 657 | i = 0; |
| 658 | /* This is just a CONFIG_HOLES_IN_ZONE check.*/ |
| 659 | while ((i < MAX_ORDER_NR_PAGES) && !pfn_valid_within(pfn + i)) |
| 660 | i++; |
| 661 | if (i == MAX_ORDER_NR_PAGES) |
| 662 | continue; |
| 663 | page = pfn_to_page(pfn + i); |
| 664 | if (zone && page_zone(page) != zone) |
| 665 | return 0; |
| 666 | zone = page_zone(page); |
| 667 | } |
| 668 | return 1; |
| 669 | } |
| 670 | |
| 671 | /* |
| 672 | * Scanning pfn is much easier than scanning lru list. |
| 673 | * Scan pfn from start to end and Find LRU page. |
| 674 | */ |
Andrew Morton | 7bbc090 | 2010-10-26 14:22:05 -0700 | [diff] [blame] | 675 | static unsigned long scan_lru_pages(unsigned long start, unsigned long end) |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 676 | { |
| 677 | unsigned long pfn; |
| 678 | struct page *page; |
| 679 | for (pfn = start; pfn < end; pfn++) { |
| 680 | if (pfn_valid(pfn)) { |
| 681 | page = pfn_to_page(pfn); |
| 682 | if (PageLRU(page)) |
| 683 | return pfn; |
| 684 | } |
| 685 | } |
| 686 | return 0; |
| 687 | } |
| 688 | |
| 689 | static struct page * |
Hugh Dickins | 3c1d437 | 2009-01-06 14:39:23 -0800 | [diff] [blame] | 690 | hotremove_migrate_alloc(struct page *page, unsigned long private, int **x) |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 691 | { |
Hugh Dickins | 3c1d437 | 2009-01-06 14:39:23 -0800 | [diff] [blame] | 692 | /* This should be improooooved!! */ |
| 693 | return alloc_page(GFP_HIGHUSER_MOVABLE); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 694 | } |
| 695 | |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 696 | #define NR_OFFLINE_AT_ONCE_PAGES (256) |
| 697 | static int |
| 698 | do_migrate_range(unsigned long start_pfn, unsigned long end_pfn) |
| 699 | { |
| 700 | unsigned long pfn; |
| 701 | struct page *page; |
| 702 | int move_pages = NR_OFFLINE_AT_ONCE_PAGES; |
| 703 | int not_managed = 0; |
| 704 | int ret = 0; |
| 705 | LIST_HEAD(source); |
| 706 | |
| 707 | for (pfn = start_pfn; pfn < end_pfn && move_pages > 0; pfn++) { |
| 708 | if (!pfn_valid(pfn)) |
| 709 | continue; |
| 710 | page = pfn_to_page(pfn); |
Konstantin Khlebnikov | 700c2a4 | 2011-05-24 17:12:19 -0700 | [diff] [blame] | 711 | if (!get_page_unless_zero(page)) |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 712 | continue; |
| 713 | /* |
| 714 | * We can skip free pages. And we can only deal with pages on |
| 715 | * LRU. |
| 716 | */ |
Nick Piggin | 62695a8 | 2008-10-18 20:26:09 -0700 | [diff] [blame] | 717 | ret = isolate_lru_page(page); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 718 | if (!ret) { /* Success */ |
Konstantin Khlebnikov | 700c2a4 | 2011-05-24 17:12:19 -0700 | [diff] [blame] | 719 | put_page(page); |
Nick Piggin | 62695a8 | 2008-10-18 20:26:09 -0700 | [diff] [blame] | 720 | list_add_tail(&page->lru, &source); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 721 | move_pages--; |
KOSAKI Motohiro | 6d9c285 | 2009-12-14 17:58:11 -0800 | [diff] [blame] | 722 | inc_zone_page_state(page, NR_ISOLATED_ANON + |
| 723 | page_is_file_cache(page)); |
| 724 | |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 725 | } else { |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 726 | #ifdef CONFIG_DEBUG_VM |
Wu Fengguang | 718a382 | 2010-03-10 15:20:43 -0800 | [diff] [blame] | 727 | printk(KERN_ALERT "removing pfn %lx from LRU failed\n", |
| 728 | pfn); |
| 729 | dump_page(page); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 730 | #endif |
Konstantin Khlebnikov | 700c2a4 | 2011-05-24 17:12:19 -0700 | [diff] [blame] | 731 | put_page(page); |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 732 | /* Because we don't have big zone->lock. we should |
Bob Liu | 809c444 | 2010-10-26 14:22:10 -0700 | [diff] [blame] | 733 | check this again here. */ |
| 734 | if (page_count(page)) { |
| 735 | not_managed++; |
Bob Liu | f3ab263 | 2010-10-26 14:22:10 -0700 | [diff] [blame] | 736 | ret = -EBUSY; |
Bob Liu | 809c444 | 2010-10-26 14:22:10 -0700 | [diff] [blame] | 737 | break; |
| 738 | } |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 739 | } |
| 740 | } |
Bob Liu | f3ab263 | 2010-10-26 14:22:10 -0700 | [diff] [blame] | 741 | if (!list_empty(&source)) { |
| 742 | if (not_managed) { |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 743 | putback_lru_pages(&source); |
Bob Liu | f3ab263 | 2010-10-26 14:22:10 -0700 | [diff] [blame] | 744 | goto out; |
| 745 | } |
| 746 | /* this function returns # of failed pages */ |
Mel Gorman | 77f1fe6 | 2011-01-13 15:45:57 -0800 | [diff] [blame] | 747 | ret = migrate_pages(&source, hotremove_migrate_alloc, 0, |
Mel Gorman | 7f0f249 | 2011-01-13 15:45:58 -0800 | [diff] [blame] | 748 | true, true); |
Bob Liu | f3ab263 | 2010-10-26 14:22:10 -0700 | [diff] [blame] | 749 | if (ret) |
| 750 | putback_lru_pages(&source); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 751 | } |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 752 | out: |
| 753 | return ret; |
| 754 | } |
| 755 | |
| 756 | /* |
| 757 | * remove from free_area[] and mark all as Reserved. |
| 758 | */ |
| 759 | static int |
| 760 | offline_isolated_pages_cb(unsigned long start, unsigned long nr_pages, |
| 761 | void *data) |
| 762 | { |
| 763 | __offline_isolated_pages(start, start + nr_pages); |
| 764 | return 0; |
| 765 | } |
| 766 | |
| 767 | static void |
| 768 | offline_isolated_pages(unsigned long start_pfn, unsigned long end_pfn) |
| 769 | { |
KAMEZAWA Hiroyuki | 908eedc | 2009-09-22 16:45:46 -0700 | [diff] [blame] | 770 | walk_system_ram_range(start_pfn, end_pfn - start_pfn, NULL, |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 771 | offline_isolated_pages_cb); |
| 772 | } |
| 773 | |
| 774 | /* |
| 775 | * Check all pages in range, recoreded as memory resource, are isolated. |
| 776 | */ |
| 777 | static int |
| 778 | check_pages_isolated_cb(unsigned long start_pfn, unsigned long nr_pages, |
| 779 | void *data) |
| 780 | { |
| 781 | int ret; |
| 782 | long offlined = *(long *)data; |
| 783 | ret = test_pages_isolated(start_pfn, start_pfn + nr_pages); |
| 784 | offlined = nr_pages; |
| 785 | if (!ret) |
| 786 | *(long *)data += offlined; |
| 787 | return ret; |
| 788 | } |
| 789 | |
| 790 | static long |
| 791 | check_pages_isolated(unsigned long start_pfn, unsigned long end_pfn) |
| 792 | { |
| 793 | long offlined = 0; |
| 794 | int ret; |
| 795 | |
KAMEZAWA Hiroyuki | 908eedc | 2009-09-22 16:45:46 -0700 | [diff] [blame] | 796 | ret = walk_system_ram_range(start_pfn, end_pfn - start_pfn, &offlined, |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 797 | check_pages_isolated_cb); |
| 798 | if (ret < 0) |
| 799 | offlined = (long)ret; |
| 800 | return offlined; |
| 801 | } |
| 802 | |
KOSAKI Motohiro | 839a4fc | 2011-05-24 17:11:31 -0700 | [diff] [blame] | 803 | static int __ref offline_pages(unsigned long start_pfn, |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 804 | unsigned long end_pfn, unsigned long timeout) |
| 805 | { |
| 806 | unsigned long pfn, nr_pages, expire; |
| 807 | long offlined_pages; |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 808 | int ret, drain, retry_max, node; |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 809 | struct zone *zone; |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 810 | struct memory_notify arg; |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 811 | |
| 812 | BUG_ON(start_pfn >= end_pfn); |
| 813 | /* at least, alignment against pageblock is necessary */ |
| 814 | if (!IS_ALIGNED(start_pfn, pageblock_nr_pages)) |
| 815 | return -EINVAL; |
| 816 | if (!IS_ALIGNED(end_pfn, pageblock_nr_pages)) |
| 817 | return -EINVAL; |
| 818 | /* This makes hotplug much easier...and readable. |
| 819 | we assume this for now. .*/ |
| 820 | if (!test_pages_in_a_zone(start_pfn, end_pfn)) |
| 821 | return -EINVAL; |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 822 | |
KOSAKI Motohiro | 20d6c96 | 2010-12-02 14:31:19 -0800 | [diff] [blame] | 823 | lock_memory_hotplug(); |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 824 | |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 825 | zone = page_zone(pfn_to_page(start_pfn)); |
| 826 | node = zone_to_nid(zone); |
| 827 | nr_pages = end_pfn - start_pfn; |
| 828 | |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 829 | /* set above range as isolated */ |
| 830 | ret = start_isolate_page_range(start_pfn, end_pfn); |
| 831 | if (ret) |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 832 | goto out; |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 833 | |
| 834 | arg.start_pfn = start_pfn; |
| 835 | arg.nr_pages = nr_pages; |
| 836 | arg.status_change_nid = -1; |
| 837 | if (nr_pages >= node_present_pages(node)) |
| 838 | arg.status_change_nid = node; |
| 839 | |
| 840 | ret = memory_notify(MEM_GOING_OFFLINE, &arg); |
| 841 | ret = notifier_to_errno(ret); |
| 842 | if (ret) |
| 843 | goto failed_removal; |
| 844 | |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 845 | pfn = start_pfn; |
| 846 | expire = jiffies + timeout; |
| 847 | drain = 0; |
| 848 | retry_max = 5; |
| 849 | repeat: |
| 850 | /* start memory hot removal */ |
| 851 | ret = -EAGAIN; |
| 852 | if (time_after(jiffies, expire)) |
| 853 | goto failed_removal; |
| 854 | ret = -EINTR; |
| 855 | if (signal_pending(current)) |
| 856 | goto failed_removal; |
| 857 | ret = 0; |
| 858 | if (drain) { |
| 859 | lru_add_drain_all(); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 860 | cond_resched(); |
Christoph Lameter | 9f8f217 | 2008-02-04 22:29:11 -0800 | [diff] [blame] | 861 | drain_all_pages(); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 862 | } |
| 863 | |
| 864 | pfn = scan_lru_pages(start_pfn, end_pfn); |
| 865 | if (pfn) { /* We have page on LRU */ |
| 866 | ret = do_migrate_range(pfn, end_pfn); |
| 867 | if (!ret) { |
| 868 | drain = 1; |
| 869 | goto repeat; |
| 870 | } else { |
| 871 | if (ret < 0) |
| 872 | if (--retry_max == 0) |
| 873 | goto failed_removal; |
| 874 | yield(); |
| 875 | drain = 1; |
| 876 | goto repeat; |
| 877 | } |
| 878 | } |
| 879 | /* drain all zone's lru pagevec, this is asyncronous... */ |
| 880 | lru_add_drain_all(); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 881 | yield(); |
| 882 | /* drain pcp pages , this is synchrouns. */ |
Christoph Lameter | 9f8f217 | 2008-02-04 22:29:11 -0800 | [diff] [blame] | 883 | drain_all_pages(); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 884 | /* check again */ |
| 885 | offlined_pages = check_pages_isolated(start_pfn, end_pfn); |
| 886 | if (offlined_pages < 0) { |
| 887 | ret = -EBUSY; |
| 888 | goto failed_removal; |
| 889 | } |
| 890 | printk(KERN_INFO "Offlined Pages %ld\n", offlined_pages); |
| 891 | /* Ok, all of our target is islaoted. |
| 892 | We cannot do rollback at this point. */ |
| 893 | offline_isolated_pages(start_pfn, end_pfn); |
KAMEZAWA Hiroyuki | dbc0e4c | 2007-11-14 16:59:12 -0800 | [diff] [blame] | 894 | /* reset pagetype flags and makes migrate type to be MOVABLE */ |
| 895 | undo_isolate_page_range(start_pfn, end_pfn); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 896 | /* removal success */ |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 897 | zone->present_pages -= offlined_pages; |
| 898 | zone->zone_pgdat->node_present_pages -= offlined_pages; |
| 899 | totalram_pages -= offlined_pages; |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 900 | |
KOSAKI Motohiro | 1b79acc | 2011-05-24 17:11:32 -0700 | [diff] [blame] | 901 | init_per_zone_wmark_min(); |
| 902 | |
David Rientjes | 8fe23e0 | 2009-12-14 17:58:33 -0800 | [diff] [blame] | 903 | if (!node_present_pages(node)) { |
| 904 | node_clear_state(node, N_HIGH_MEMORY); |
| 905 | kswapd_stop(node); |
| 906 | } |
Minchan Kim | bce7394 | 2009-06-16 15:32:50 -0700 | [diff] [blame] | 907 | |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 908 | vm_total_pages = nr_free_pagecache_pages(); |
| 909 | writeback_set_ratelimit(); |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 910 | |
| 911 | memory_notify(MEM_OFFLINE, &arg); |
KOSAKI Motohiro | 20d6c96 | 2010-12-02 14:31:19 -0800 | [diff] [blame] | 912 | unlock_memory_hotplug(); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 913 | return 0; |
| 914 | |
| 915 | failed_removal: |
| 916 | printk(KERN_INFO "memory offlining %lx to %lx failed\n", |
| 917 | start_pfn, end_pfn); |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 918 | memory_notify(MEM_CANCEL_OFFLINE, &arg); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 919 | /* pushback to free area */ |
| 920 | undo_isolate_page_range(start_pfn, end_pfn); |
Yasunori Goto | 7b78d33 | 2007-10-21 16:41:36 -0700 | [diff] [blame] | 921 | |
Andi Kleen | 6ad696d | 2009-11-17 14:06:22 -0800 | [diff] [blame] | 922 | out: |
KOSAKI Motohiro | 20d6c96 | 2010-12-02 14:31:19 -0800 | [diff] [blame] | 923 | unlock_memory_hotplug(); |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 924 | return ret; |
| 925 | } |
Badari Pulavarty | 7108878 | 2008-10-18 20:25:58 -0700 | [diff] [blame] | 926 | |
| 927 | int remove_memory(u64 start, u64 size) |
| 928 | { |
| 929 | unsigned long start_pfn, end_pfn; |
| 930 | |
| 931 | start_pfn = PFN_DOWN(start); |
| 932 | end_pfn = start_pfn + PFN_DOWN(size); |
| 933 | return offline_pages(start_pfn, end_pfn, 120 * HZ); |
| 934 | } |
KAMEZAWA Hiroyuki | 48e9419 | 2007-10-16 01:26:14 -0700 | [diff] [blame] | 935 | #else |
| 936 | int remove_memory(u64 start, u64 size) |
| 937 | { |
| 938 | return -EINVAL; |
| 939 | } |
KAMEZAWA Hiroyuki | 0c0e619 | 2007-10-16 01:26:12 -0700 | [diff] [blame] | 940 | #endif /* CONFIG_MEMORY_HOTREMOVE */ |
Badari Pulavarty | 7108878 | 2008-10-18 20:25:58 -0700 | [diff] [blame] | 941 | EXPORT_SYMBOL_GPL(remove_memory); |